src/hotspot/cpu/x86/x86.ad
author vdeshpande
Wed, 12 Dec 2018 14:48:34 -0800
changeset 52992 4bb6e0871bf7
parent 51996 84743156e780
child 53171 3ab3cb8a8d41
permissions -rw-r--r--
8214751: X86: Support for VNNI Instructions Reviewed-by: kvn Contributed-by: razvan.a.lupusoru@intel.com, vivek.r.deshpande@intel.com
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
     1
//
51078
fc6cfe40e32a 8207049: Minor improvements of compiler code.
goetz
parents: 50525
diff changeset
     2
// Copyright (c) 2011, 2018, Oracle and/or its affiliates. All rights reserved.
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
     3
// DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
     4
//
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
     5
// This code is free software; you can redistribute it and/or modify it
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
     6
// under the terms of the GNU General Public License version 2 only, as
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
     7
// published by the Free Software Foundation.
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
     8
//
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
     9
// This code is distributed in the hope that it will be useful, but WITHOUT
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    10
// ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    11
// FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    12
// version 2 for more details (a copy is included in the LICENSE file that
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    13
// accompanied this code).
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    14
//
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    15
// You should have received a copy of the GNU General Public License version
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    16
// 2 along with this work; if not, write to the Free Software Foundation,
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    17
// Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    18
//
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    19
// Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    20
// or visit www.oracle.com if you need additional information or have any
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    21
// questions.
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    22
//
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    23
//
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    24
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    25
// X86 Common Architecture Description File
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
    26
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    27
//----------REGISTER DEFINITION BLOCK------------------------------------------
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    28
// This information is used by the matcher and the register allocator to
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    29
// describe individual registers and classes of registers within the target
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    30
// archtecture.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    31
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    32
register %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    33
//----------Architecture Description Register Definitions----------------------
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    34
// General Registers
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    35
// "reg_def"  name ( register save type, C convention save type,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    36
//                   ideal register type, encoding );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    37
// Register Save Types:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    38
//
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    39
// NS  = No-Save:       The register allocator assumes that these registers
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    40
//                      can be used without saving upon entry to the method, &
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    41
//                      that they do not need to be saved at call sites.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    42
//
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    43
// SOC = Save-On-Call:  The register allocator assumes that these registers
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    44
//                      can be used without saving upon entry to the method,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    45
//                      but that they must be saved at call sites.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    46
//
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    47
// SOE = Save-On-Entry: The register allocator assumes that these registers
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    48
//                      must be saved before using them upon entry to the
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    49
//                      method, but they do not need to be saved at call
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    50
//                      sites.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    51
//
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    52
// AS  = Always-Save:   The register allocator assumes that these registers
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    53
//                      must be saved before using them upon entry to the
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    54
//                      method, & that they must be saved at call sites.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    55
//
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    56
// Ideal Register Type is used to determine how to save & restore a
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    57
// register.  Op_RegI will get spilled with LoadI/StoreI, Op_RegP will get
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    58
// spilled with LoadP/StoreP.  If the register supports both, use Op_RegI.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    59
//
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    60
// The encoding number is the actual bit-pattern placed into the opcodes.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    61
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    62
// XMM registers.  512-bit registers or 8 words each, labeled (a)-p.
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    63
// Word a in each register holds a Float, words ab hold a Double.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    64
// The whole registers are used in SSE4.2 version intrinsics,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    65
// array copy stubs and superword operations (see UseSSE42Intrinsics,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    66
// UseXMMForArrayCopy and UseSuperword flags).
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    67
// For pre EVEX enabled architectures:
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    68
//      XMM8-XMM15 must be encoded with REX (VEX for UseAVX)
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    69
// For EVEX enabled architectures:
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    70
//      XMM8-XMM31 must be encoded with REX (EVEX for UseAVX).
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    71
//
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    72
// Linux ABI:   No register preserved across function calls
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    73
//              XMM0-XMM7 might hold parameters
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    74
// Windows ABI: XMM6-XMM31 preserved across function calls
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    75
//              XMM0-XMM3 might hold parameters
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    76
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    77
reg_def XMM0 ( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
    78
reg_def XMM0b( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
    79
reg_def XMM0c( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
    80
reg_def XMM0d( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
    81
reg_def XMM0e( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
    82
reg_def XMM0f( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
    83
reg_def XMM0g( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
    84
reg_def XMM0h( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    85
reg_def XMM0i( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    86
reg_def XMM0j( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    87
reg_def XMM0k( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    88
reg_def XMM0l( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    89
reg_def XMM0m( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    90
reg_def XMM0n( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    91
reg_def XMM0o( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
    92
reg_def XMM0p( SOC, SOC, Op_RegF, 0, xmm0->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    93
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
    94
reg_def XMM1 ( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
    95
reg_def XMM1b( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
    96
reg_def XMM1c( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
    97
reg_def XMM1d( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
    98
reg_def XMM1e( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
    99
reg_def XMM1f( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   100
reg_def XMM1g( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   101
reg_def XMM1h( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   102
reg_def XMM1i( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   103
reg_def XMM1j( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   104
reg_def XMM1k( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   105
reg_def XMM1l( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   106
reg_def XMM1m( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   107
reg_def XMM1n( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   108
reg_def XMM1o( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   109
reg_def XMM1p( SOC, SOC, Op_RegF, 1, xmm1->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   110
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   111
reg_def XMM2 ( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   112
reg_def XMM2b( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   113
reg_def XMM2c( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   114
reg_def XMM2d( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   115
reg_def XMM2e( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   116
reg_def XMM2f( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   117
reg_def XMM2g( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   118
reg_def XMM2h( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   119
reg_def XMM2i( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   120
reg_def XMM2j( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   121
reg_def XMM2k( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   122
reg_def XMM2l( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   123
reg_def XMM2m( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   124
reg_def XMM2n( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   125
reg_def XMM2o( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   126
reg_def XMM2p( SOC, SOC, Op_RegF, 2, xmm2->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   127
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   128
reg_def XMM3 ( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   129
reg_def XMM3b( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   130
reg_def XMM3c( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   131
reg_def XMM3d( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   132
reg_def XMM3e( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   133
reg_def XMM3f( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   134
reg_def XMM3g( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   135
reg_def XMM3h( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   136
reg_def XMM3i( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   137
reg_def XMM3j( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   138
reg_def XMM3k( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   139
reg_def XMM3l( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   140
reg_def XMM3m( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   141
reg_def XMM3n( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   142
reg_def XMM3o( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   143
reg_def XMM3p( SOC, SOC, Op_RegF, 3, xmm3->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   144
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   145
reg_def XMM4 ( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   146
reg_def XMM4b( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   147
reg_def XMM4c( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   148
reg_def XMM4d( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   149
reg_def XMM4e( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   150
reg_def XMM4f( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   151
reg_def XMM4g( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   152
reg_def XMM4h( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   153
reg_def XMM4i( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   154
reg_def XMM4j( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   155
reg_def XMM4k( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   156
reg_def XMM4l( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   157
reg_def XMM4m( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   158
reg_def XMM4n( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   159
reg_def XMM4o( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   160
reg_def XMM4p( SOC, SOC, Op_RegF, 4, xmm4->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   161
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   162
reg_def XMM5 ( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   163
reg_def XMM5b( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   164
reg_def XMM5c( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   165
reg_def XMM5d( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   166
reg_def XMM5e( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   167
reg_def XMM5f( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   168
reg_def XMM5g( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   169
reg_def XMM5h( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   170
reg_def XMM5i( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   171
reg_def XMM5j( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   172
reg_def XMM5k( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   173
reg_def XMM5l( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   174
reg_def XMM5m( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   175
reg_def XMM5n( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   176
reg_def XMM5o( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   177
reg_def XMM5p( SOC, SOC, Op_RegF, 5, xmm5->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   178
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   179
reg_def XMM6 ( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   180
reg_def XMM6b( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   181
reg_def XMM6c( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   182
reg_def XMM6d( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   183
reg_def XMM6e( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   184
reg_def XMM6f( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   185
reg_def XMM6g( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   186
reg_def XMM6h( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   187
reg_def XMM6i( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   188
reg_def XMM6j( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   189
reg_def XMM6k( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   190
reg_def XMM6l( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   191
reg_def XMM6m( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   192
reg_def XMM6n( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   193
reg_def XMM6o( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   194
reg_def XMM6p( SOC, SOC, Op_RegF, 6, xmm6->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   195
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   196
reg_def XMM7 ( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   197
reg_def XMM7b( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   198
reg_def XMM7c( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   199
reg_def XMM7d( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   200
reg_def XMM7e( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   201
reg_def XMM7f( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   202
reg_def XMM7g( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   203
reg_def XMM7h( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   204
reg_def XMM7i( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   205
reg_def XMM7j( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   206
reg_def XMM7k( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   207
reg_def XMM7l( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   208
reg_def XMM7m( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   209
reg_def XMM7n( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   210
reg_def XMM7o( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   211
reg_def XMM7p( SOC, SOC, Op_RegF, 7, xmm7->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   212
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   213
#ifdef _LP64
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   214
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   215
reg_def XMM8 ( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   216
reg_def XMM8b( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   217
reg_def XMM8c( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   218
reg_def XMM8d( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   219
reg_def XMM8e( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   220
reg_def XMM8f( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   221
reg_def XMM8g( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   222
reg_def XMM8h( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   223
reg_def XMM8i( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   224
reg_def XMM8j( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   225
reg_def XMM8k( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   226
reg_def XMM8l( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   227
reg_def XMM8m( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   228
reg_def XMM8n( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   229
reg_def XMM8o( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   230
reg_def XMM8p( SOC, SOC, Op_RegF, 8, xmm8->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   231
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   232
reg_def XMM9 ( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   233
reg_def XMM9b( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   234
reg_def XMM9c( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   235
reg_def XMM9d( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   236
reg_def XMM9e( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   237
reg_def XMM9f( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   238
reg_def XMM9g( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   239
reg_def XMM9h( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   240
reg_def XMM9i( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   241
reg_def XMM9j( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   242
reg_def XMM9k( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   243
reg_def XMM9l( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   244
reg_def XMM9m( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   245
reg_def XMM9n( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   246
reg_def XMM9o( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   247
reg_def XMM9p( SOC, SOC, Op_RegF, 9, xmm9->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   248
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   249
reg_def XMM10 ( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   250
reg_def XMM10b( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   251
reg_def XMM10c( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   252
reg_def XMM10d( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   253
reg_def XMM10e( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   254
reg_def XMM10f( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   255
reg_def XMM10g( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   256
reg_def XMM10h( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   257
reg_def XMM10i( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   258
reg_def XMM10j( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   259
reg_def XMM10k( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   260
reg_def XMM10l( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   261
reg_def XMM10m( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   262
reg_def XMM10n( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   263
reg_def XMM10o( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   264
reg_def XMM10p( SOC, SOC, Op_RegF, 10, xmm10->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   265
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   266
reg_def XMM11 ( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   267
reg_def XMM11b( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   268
reg_def XMM11c( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   269
reg_def XMM11d( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   270
reg_def XMM11e( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   271
reg_def XMM11f( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   272
reg_def XMM11g( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   273
reg_def XMM11h( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   274
reg_def XMM11i( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   275
reg_def XMM11j( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   276
reg_def XMM11k( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   277
reg_def XMM11l( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   278
reg_def XMM11m( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   279
reg_def XMM11n( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   280
reg_def XMM11o( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   281
reg_def XMM11p( SOC, SOC, Op_RegF, 11, xmm11->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   282
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   283
reg_def XMM12 ( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   284
reg_def XMM12b( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   285
reg_def XMM12c( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   286
reg_def XMM12d( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   287
reg_def XMM12e( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   288
reg_def XMM12f( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   289
reg_def XMM12g( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   290
reg_def XMM12h( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   291
reg_def XMM12i( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   292
reg_def XMM12j( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   293
reg_def XMM12k( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   294
reg_def XMM12l( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   295
reg_def XMM12m( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   296
reg_def XMM12n( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   297
reg_def XMM12o( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   298
reg_def XMM12p( SOC, SOC, Op_RegF, 12, xmm12->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   299
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   300
reg_def XMM13 ( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   301
reg_def XMM13b( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   302
reg_def XMM13c( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   303
reg_def XMM13d( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   304
reg_def XMM13e( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   305
reg_def XMM13f( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   306
reg_def XMM13g( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   307
reg_def XMM13h( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   308
reg_def XMM13i( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   309
reg_def XMM13j( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   310
reg_def XMM13k( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   311
reg_def XMM13l( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   312
reg_def XMM13m( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   313
reg_def XMM13n( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   314
reg_def XMM13o( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   315
reg_def XMM13p( SOC, SOC, Op_RegF, 13, xmm13->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   316
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   317
reg_def XMM14 ( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   318
reg_def XMM14b( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   319
reg_def XMM14c( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   320
reg_def XMM14d( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   321
reg_def XMM14e( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   322
reg_def XMM14f( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   323
reg_def XMM14g( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   324
reg_def XMM14h( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   325
reg_def XMM14i( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   326
reg_def XMM14j( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   327
reg_def XMM14k( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   328
reg_def XMM14l( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   329
reg_def XMM14m( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   330
reg_def XMM14n( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   331
reg_def XMM14o( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   332
reg_def XMM14p( SOC, SOC, Op_RegF, 14, xmm14->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   333
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   334
reg_def XMM15 ( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg());
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   335
reg_def XMM15b( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(1));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   336
reg_def XMM15c( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(2));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   337
reg_def XMM15d( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(3));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   338
reg_def XMM15e( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(4));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   339
reg_def XMM15f( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(5));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   340
reg_def XMM15g( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(6));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
   341
reg_def XMM15h( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(7));
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   342
reg_def XMM15i( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   343
reg_def XMM15j( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   344
reg_def XMM15k( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   345
reg_def XMM15l( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   346
reg_def XMM15m( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   347
reg_def XMM15n( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   348
reg_def XMM15o( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   349
reg_def XMM15p( SOC, SOC, Op_RegF, 15, xmm15->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   350
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   351
reg_def XMM16 ( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   352
reg_def XMM16b( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   353
reg_def XMM16c( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   354
reg_def XMM16d( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   355
reg_def XMM16e( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   356
reg_def XMM16f( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   357
reg_def XMM16g( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   358
reg_def XMM16h( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   359
reg_def XMM16i( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   360
reg_def XMM16j( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   361
reg_def XMM16k( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   362
reg_def XMM16l( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   363
reg_def XMM16m( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   364
reg_def XMM16n( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   365
reg_def XMM16o( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   366
reg_def XMM16p( SOC, SOC, Op_RegF, 16, xmm16->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   367
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   368
reg_def XMM17 ( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   369
reg_def XMM17b( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   370
reg_def XMM17c( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   371
reg_def XMM17d( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   372
reg_def XMM17e( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   373
reg_def XMM17f( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   374
reg_def XMM17g( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   375
reg_def XMM17h( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   376
reg_def XMM17i( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   377
reg_def XMM17j( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   378
reg_def XMM17k( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   379
reg_def XMM17l( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   380
reg_def XMM17m( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   381
reg_def XMM17n( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   382
reg_def XMM17o( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   383
reg_def XMM17p( SOC, SOC, Op_RegF, 17, xmm17->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   384
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   385
reg_def XMM18 ( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   386
reg_def XMM18b( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   387
reg_def XMM18c( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   388
reg_def XMM18d( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   389
reg_def XMM18e( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   390
reg_def XMM18f( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   391
reg_def XMM18g( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   392
reg_def XMM18h( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   393
reg_def XMM18i( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   394
reg_def XMM18j( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   395
reg_def XMM18k( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   396
reg_def XMM18l( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   397
reg_def XMM18m( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   398
reg_def XMM18n( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   399
reg_def XMM18o( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   400
reg_def XMM18p( SOC, SOC, Op_RegF, 18, xmm18->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   401
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   402
reg_def XMM19 ( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   403
reg_def XMM19b( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   404
reg_def XMM19c( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   405
reg_def XMM19d( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   406
reg_def XMM19e( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   407
reg_def XMM19f( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   408
reg_def XMM19g( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   409
reg_def XMM19h( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   410
reg_def XMM19i( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   411
reg_def XMM19j( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   412
reg_def XMM19k( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   413
reg_def XMM19l( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   414
reg_def XMM19m( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   415
reg_def XMM19n( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   416
reg_def XMM19o( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   417
reg_def XMM19p( SOC, SOC, Op_RegF, 19, xmm19->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   418
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   419
reg_def XMM20 ( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   420
reg_def XMM20b( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   421
reg_def XMM20c( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   422
reg_def XMM20d( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   423
reg_def XMM20e( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   424
reg_def XMM20f( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   425
reg_def XMM20g( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   426
reg_def XMM20h( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   427
reg_def XMM20i( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   428
reg_def XMM20j( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   429
reg_def XMM20k( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   430
reg_def XMM20l( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   431
reg_def XMM20m( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   432
reg_def XMM20n( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   433
reg_def XMM20o( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   434
reg_def XMM20p( SOC, SOC, Op_RegF, 20, xmm20->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   435
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   436
reg_def XMM21 ( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   437
reg_def XMM21b( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   438
reg_def XMM21c( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   439
reg_def XMM21d( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   440
reg_def XMM21e( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   441
reg_def XMM21f( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   442
reg_def XMM21g( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   443
reg_def XMM21h( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   444
reg_def XMM21i( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   445
reg_def XMM21j( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   446
reg_def XMM21k( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   447
reg_def XMM21l( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   448
reg_def XMM21m( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   449
reg_def XMM21n( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   450
reg_def XMM21o( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   451
reg_def XMM21p( SOC, SOC, Op_RegF, 21, xmm21->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   452
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   453
reg_def XMM22 ( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   454
reg_def XMM22b( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   455
reg_def XMM22c( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   456
reg_def XMM22d( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   457
reg_def XMM22e( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   458
reg_def XMM22f( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   459
reg_def XMM22g( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   460
reg_def XMM22h( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   461
reg_def XMM22i( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   462
reg_def XMM22j( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   463
reg_def XMM22k( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   464
reg_def XMM22l( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   465
reg_def XMM22m( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   466
reg_def XMM22n( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   467
reg_def XMM22o( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   468
reg_def XMM22p( SOC, SOC, Op_RegF, 22, xmm22->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   469
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   470
reg_def XMM23 ( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   471
reg_def XMM23b( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   472
reg_def XMM23c( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   473
reg_def XMM23d( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   474
reg_def XMM23e( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   475
reg_def XMM23f( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   476
reg_def XMM23g( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   477
reg_def XMM23h( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   478
reg_def XMM23i( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   479
reg_def XMM23j( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   480
reg_def XMM23k( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   481
reg_def XMM23l( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   482
reg_def XMM23m( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   483
reg_def XMM23n( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   484
reg_def XMM23o( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   485
reg_def XMM23p( SOC, SOC, Op_RegF, 23, xmm23->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   486
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   487
reg_def XMM24 ( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   488
reg_def XMM24b( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   489
reg_def XMM24c( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   490
reg_def XMM24d( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   491
reg_def XMM24e( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   492
reg_def XMM24f( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   493
reg_def XMM24g( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   494
reg_def XMM24h( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   495
reg_def XMM24i( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   496
reg_def XMM24j( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   497
reg_def XMM24k( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   498
reg_def XMM24l( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   499
reg_def XMM24m( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   500
reg_def XMM24n( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   501
reg_def XMM24o( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   502
reg_def XMM24p( SOC, SOC, Op_RegF, 24, xmm24->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   503
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   504
reg_def XMM25 ( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   505
reg_def XMM25b( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   506
reg_def XMM25c( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   507
reg_def XMM25d( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   508
reg_def XMM25e( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   509
reg_def XMM25f( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   510
reg_def XMM25g( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   511
reg_def XMM25h( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   512
reg_def XMM25i( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   513
reg_def XMM25j( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   514
reg_def XMM25k( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   515
reg_def XMM25l( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   516
reg_def XMM25m( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   517
reg_def XMM25n( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   518
reg_def XMM25o( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   519
reg_def XMM25p( SOC, SOC, Op_RegF, 25, xmm25->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   520
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   521
reg_def XMM26 ( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   522
reg_def XMM26b( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   523
reg_def XMM26c( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   524
reg_def XMM26d( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   525
reg_def XMM26e( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   526
reg_def XMM26f( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   527
reg_def XMM26g( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   528
reg_def XMM26h( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   529
reg_def XMM26i( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   530
reg_def XMM26j( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   531
reg_def XMM26k( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   532
reg_def XMM26l( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   533
reg_def XMM26m( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   534
reg_def XMM26n( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   535
reg_def XMM26o( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   536
reg_def XMM26p( SOC, SOC, Op_RegF, 26, xmm26->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   537
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   538
reg_def XMM27 ( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   539
reg_def XMM27b( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   540
reg_def XMM27c( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   541
reg_def XMM27d( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   542
reg_def XMM27e( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   543
reg_def XMM27f( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   544
reg_def XMM27g( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   545
reg_def XMM27h( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   546
reg_def XMM27i( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   547
reg_def XMM27j( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   548
reg_def XMM27k( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   549
reg_def XMM27l( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   550
reg_def XMM27m( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   551
reg_def XMM27n( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   552
reg_def XMM27o( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   553
reg_def XMM27p( SOC, SOC, Op_RegF, 27, xmm27->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   554
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   555
reg_def XMM28 ( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   556
reg_def XMM28b( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   557
reg_def XMM28c( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   558
reg_def XMM28d( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   559
reg_def XMM28e( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   560
reg_def XMM28f( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   561
reg_def XMM28g( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   562
reg_def XMM28h( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   563
reg_def XMM28i( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   564
reg_def XMM28j( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   565
reg_def XMM28k( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   566
reg_def XMM28l( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   567
reg_def XMM28m( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   568
reg_def XMM28n( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   569
reg_def XMM28o( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   570
reg_def XMM28p( SOC, SOC, Op_RegF, 28, xmm28->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   571
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   572
reg_def XMM29 ( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   573
reg_def XMM29b( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   574
reg_def XMM29c( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   575
reg_def XMM29d( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   576
reg_def XMM29e( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   577
reg_def XMM29f( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   578
reg_def XMM29g( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   579
reg_def XMM29h( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   580
reg_def XMM29i( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   581
reg_def XMM29j( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   582
reg_def XMM29k( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   583
reg_def XMM29l( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   584
reg_def XMM29m( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   585
reg_def XMM29n( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   586
reg_def XMM29o( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   587
reg_def XMM29p( SOC, SOC, Op_RegF, 29, xmm29->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   588
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   589
reg_def XMM30 ( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   590
reg_def XMM30b( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   591
reg_def XMM30c( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   592
reg_def XMM30d( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   593
reg_def XMM30e( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   594
reg_def XMM30f( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   595
reg_def XMM30g( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   596
reg_def XMM30h( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   597
reg_def XMM30i( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   598
reg_def XMM30j( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   599
reg_def XMM30k( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   600
reg_def XMM30l( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   601
reg_def XMM30m( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   602
reg_def XMM30n( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   603
reg_def XMM30o( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   604
reg_def XMM30p( SOC, SOC, Op_RegF, 30, xmm30->as_VMReg()->next(15));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   605
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   606
reg_def XMM31 ( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   607
reg_def XMM31b( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(1));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   608
reg_def XMM31c( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   609
reg_def XMM31d( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(3));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   610
reg_def XMM31e( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(4));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   611
reg_def XMM31f( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(5));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   612
reg_def XMM31g( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(6));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   613
reg_def XMM31h( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(7));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   614
reg_def XMM31i( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(8));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   615
reg_def XMM31j( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(9));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   616
reg_def XMM31k( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(10));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   617
reg_def XMM31l( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(11));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   618
reg_def XMM31m( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(12));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   619
reg_def XMM31n( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(13));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   620
reg_def XMM31o( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(14));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   621
reg_def XMM31p( SOC, SOC, Op_RegF, 31, xmm31->as_VMReg()->next(15));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   622
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   623
#endif // _LP64
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   624
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   625
#ifdef _LP64
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   626
reg_def RFLAGS(SOC, SOC, 0, 16, VMRegImpl::Bad());
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   627
#else
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   628
reg_def RFLAGS(SOC, SOC, 0, 8, VMRegImpl::Bad());
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   629
#endif // _LP64
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   630
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   631
alloc_class chunk1(XMM0,  XMM0b,  XMM0c,  XMM0d,  XMM0e,  XMM0f,  XMM0g,  XMM0h,  XMM0i,  XMM0j,  XMM0k,  XMM0l,  XMM0m,  XMM0n,  XMM0o,  XMM0p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   632
                   XMM1,  XMM1b,  XMM1c,  XMM1d,  XMM1e,  XMM1f,  XMM1g,  XMM1h,  XMM1i,  XMM1j,  XMM1k,  XMM1l,  XMM1m,  XMM1n,  XMM1o,  XMM1p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   633
                   XMM2,  XMM2b,  XMM2c,  XMM2d,  XMM2e,  XMM2f,  XMM2g,  XMM2h,  XMM2i,  XMM2j,  XMM2k,  XMM2l,  XMM2m,  XMM2n,  XMM2o,  XMM2p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   634
                   XMM3,  XMM3b,  XMM3c,  XMM3d,  XMM3e,  XMM3f,  XMM3g,  XMM3h,  XMM3i,  XMM3j,  XMM3k,  XMM3l,  XMM3m,  XMM3n,  XMM3o,  XMM3p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   635
                   XMM4,  XMM4b,  XMM4c,  XMM4d,  XMM4e,  XMM4f,  XMM4g,  XMM4h,  XMM4i,  XMM4j,  XMM4k,  XMM4l,  XMM4m,  XMM4n,  XMM4o,  XMM4p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   636
                   XMM5,  XMM5b,  XMM5c,  XMM5d,  XMM5e,  XMM5f,  XMM5g,  XMM5h,  XMM5i,  XMM5j,  XMM5k,  XMM5l,  XMM5m,  XMM5n,  XMM5o,  XMM5p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   637
                   XMM6,  XMM6b,  XMM6c,  XMM6d,  XMM6e,  XMM6f,  XMM6g,  XMM6h,  XMM6i,  XMM6j,  XMM6k,  XMM6l,  XMM6m,  XMM6n,  XMM6o,  XMM6p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   638
                   XMM7,  XMM7b,  XMM7c,  XMM7d,  XMM7e,  XMM7f,  XMM7g,  XMM7h,  XMM7i,  XMM7j,  XMM7k,  XMM7l,  XMM7m,  XMM7n,  XMM7o,  XMM7p
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   639
#ifdef _LP64
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   640
                  ,XMM8,  XMM8b,  XMM8c,  XMM8d,  XMM8e,  XMM8f,  XMM8g,  XMM8h,  XMM8i,  XMM8j,  XMM8k,  XMM8l,  XMM8m,  XMM8n,  XMM8o,  XMM8p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   641
                   XMM9,  XMM9b,  XMM9c,  XMM9d,  XMM9e,  XMM9f,  XMM9g,  XMM9h,  XMM9i,  XMM9j,  XMM9k,  XMM9l,  XMM9m,  XMM9n,  XMM9o,  XMM9p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   642
                   XMM10, XMM10b, XMM10c, XMM10d, XMM10e, XMM10f, XMM10g, XMM10h, XMM10i, XMM10j, XMM10k, XMM10l, XMM10m, XMM10n, XMM10o, XMM10p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   643
                   XMM11, XMM11b, XMM11c, XMM11d, XMM11e, XMM11f, XMM11g, XMM11h, XMM11i, XMM11j, XMM11k, XMM11l, XMM11m, XMM11n, XMM11o, XMM11p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   644
                   XMM12, XMM12b, XMM12c, XMM12d, XMM12e, XMM12f, XMM12g, XMM12h, XMM12i, XMM12j, XMM12k, XMM12l, XMM12m, XMM12n, XMM12o, XMM12p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   645
                   XMM13, XMM13b, XMM13c, XMM13d, XMM13e, XMM13f, XMM13g, XMM13h, XMM13i, XMM13j, XMM13k, XMM13l, XMM13m, XMM13n, XMM13o, XMM13p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   646
                   XMM14, XMM14b, XMM14c, XMM14d, XMM14e, XMM14f, XMM14g, XMM14h, XMM14i, XMM14j, XMM14k, XMM14l, XMM14m, XMM14n, XMM14o, XMM14p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   647
                   XMM15, XMM15b, XMM15c, XMM15d, XMM15e, XMM15f, XMM15g, XMM15h, XMM15i, XMM15j, XMM15k, XMM15l, XMM15m, XMM15n, XMM15o, XMM15p
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   648
                  ,XMM16, XMM16b, XMM16c, XMM16d, XMM16e, XMM16f, XMM16g, XMM16h, XMM16i, XMM16j, XMM16k, XMM16l, XMM16m, XMM16n, XMM16o, XMM16p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   649
                   XMM17, XMM17b, XMM17c, XMM17d, XMM17e, XMM17f, XMM17g, XMM17h, XMM17i, XMM17j, XMM17k, XMM17l, XMM17m, XMM17n, XMM17o, XMM17p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   650
                   XMM18, XMM18b, XMM18c, XMM18d, XMM18e, XMM18f, XMM18g, XMM18h, XMM18i, XMM18j, XMM18k, XMM18l, XMM18m, XMM18n, XMM18o, XMM18p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   651
                   XMM19, XMM19b, XMM19c, XMM19d, XMM19e, XMM19f, XMM19g, XMM19h, XMM19i, XMM19j, XMM19k, XMM19l, XMM19m, XMM19n, XMM19o, XMM19p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   652
                   XMM20, XMM20b, XMM20c, XMM20d, XMM20e, XMM20f, XMM20g, XMM20h, XMM20i, XMM20j, XMM20k, XMM20l, XMM20m, XMM20n, XMM20o, XMM20p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   653
                   XMM21, XMM21b, XMM21c, XMM21d, XMM21e, XMM21f, XMM21g, XMM21h, XMM21i, XMM21j, XMM21k, XMM21l, XMM21m, XMM21n, XMM21o, XMM21p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   654
                   XMM22, XMM22b, XMM22c, XMM22d, XMM22e, XMM22f, XMM22g, XMM22h, XMM22i, XMM22j, XMM22k, XMM22l, XMM22m, XMM22n, XMM22o, XMM22p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   655
                   XMM23, XMM23b, XMM23c, XMM23d, XMM23e, XMM23f, XMM23g, XMM23h, XMM23i, XMM23j, XMM23k, XMM23l, XMM23m, XMM23n, XMM23o, XMM23p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   656
                   XMM24, XMM24b, XMM24c, XMM24d, XMM24e, XMM24f, XMM24g, XMM24h, XMM24i, XMM24j, XMM24k, XMM24l, XMM24m, XMM24n, XMM24o, XMM24p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   657
                   XMM25, XMM25b, XMM25c, XMM25d, XMM25e, XMM25f, XMM25g, XMM25h, XMM25i, XMM25j, XMM25k, XMM25l, XMM25m, XMM25n, XMM25o, XMM25p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   658
                   XMM26, XMM26b, XMM26c, XMM26d, XMM26e, XMM26f, XMM26g, XMM26h, XMM26i, XMM26j, XMM26k, XMM26l, XMM26m, XMM26n, XMM26o, XMM26p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   659
                   XMM27, XMM27b, XMM27c, XMM27d, XMM27e, XMM27f, XMM27g, XMM27h, XMM27i, XMM27j, XMM27k, XMM27l, XMM27m, XMM27n, XMM27o, XMM27p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   660
                   XMM28, XMM28b, XMM28c, XMM28d, XMM28e, XMM28f, XMM28g, XMM28h, XMM28i, XMM28j, XMM28k, XMM28l, XMM28m, XMM28n, XMM28o, XMM28p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   661
                   XMM29, XMM29b, XMM29c, XMM29d, XMM29e, XMM29f, XMM29g, XMM29h, XMM29i, XMM29j, XMM29k, XMM29l, XMM29m, XMM29n, XMM29o, XMM29p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   662
                   XMM30, XMM30b, XMM30c, XMM30d, XMM30e, XMM30f, XMM30g, XMM30h, XMM30i, XMM30j, XMM30k, XMM30l, XMM30m, XMM30n, XMM30o, XMM30p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   663
                   XMM31, XMM31b, XMM31c, XMM31d, XMM31e, XMM31f, XMM31g, XMM31h, XMM31i, XMM31j, XMM31k, XMM31l, XMM31m, XMM31n, XMM31o, XMM31p
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   664
#endif
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   665
                      );
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   666
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   667
// flags allocation class should be last.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   668
alloc_class chunk2(RFLAGS);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   669
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   670
// Singleton class for condition codes
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   671
reg_class int_flags(RFLAGS);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   672
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   673
// Class for pre evex float registers
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   674
reg_class float_reg_legacy(XMM0,
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   675
                    XMM1,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   676
                    XMM2,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   677
                    XMM3,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   678
                    XMM4,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   679
                    XMM5,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   680
                    XMM6,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   681
                    XMM7
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   682
#ifdef _LP64
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   683
                   ,XMM8,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   684
                    XMM9,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   685
                    XMM10,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   686
                    XMM11,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   687
                    XMM12,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   688
                    XMM13,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   689
                    XMM14,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   690
                    XMM15
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   691
#endif
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   692
                    );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   693
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   694
// Class for evex float registers
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   695
reg_class float_reg_evex(XMM0,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   696
                    XMM1,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   697
                    XMM2,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   698
                    XMM3,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   699
                    XMM4,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   700
                    XMM5,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   701
                    XMM6,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   702
                    XMM7
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   703
#ifdef _LP64
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   704
                   ,XMM8,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   705
                    XMM9,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   706
                    XMM10,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   707
                    XMM11,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   708
                    XMM12,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   709
                    XMM13,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   710
                    XMM14,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   711
                    XMM15,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   712
                    XMM16,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   713
                    XMM17,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   714
                    XMM18,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   715
                    XMM19,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   716
                    XMM20,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   717
                    XMM21,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   718
                    XMM22,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   719
                    XMM23,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   720
                    XMM24,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   721
                    XMM25,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   722
                    XMM26,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   723
                    XMM27,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   724
                    XMM28,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   725
                    XMM29,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   726
                    XMM30,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   727
                    XMM31
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   728
#endif
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   729
                    );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   730
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   731
reg_class_dynamic float_reg(float_reg_evex, float_reg_legacy, %{ VM_Version::supports_evex() %} );
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
   732
reg_class_dynamic float_reg_vl(float_reg_evex, float_reg_legacy, %{ VM_Version::supports_evex() && VM_Version::supports_avx512vl() %} );
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   733
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   734
// Class for pre evex double registers
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   735
reg_class double_reg_legacy(XMM0,  XMM0b,
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   736
                     XMM1,  XMM1b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   737
                     XMM2,  XMM2b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   738
                     XMM3,  XMM3b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   739
                     XMM4,  XMM4b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   740
                     XMM5,  XMM5b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   741
                     XMM6,  XMM6b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   742
                     XMM7,  XMM7b
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   743
#ifdef _LP64
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   744
                    ,XMM8,  XMM8b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   745
                     XMM9,  XMM9b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   746
                     XMM10, XMM10b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   747
                     XMM11, XMM11b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   748
                     XMM12, XMM12b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   749
                     XMM13, XMM13b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   750
                     XMM14, XMM14b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   751
                     XMM15, XMM15b
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   752
#endif
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   753
                     );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   754
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   755
// Class for evex double registers
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   756
reg_class double_reg_evex(XMM0,  XMM0b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   757
                     XMM1,  XMM1b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   758
                     XMM2,  XMM2b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   759
                     XMM3,  XMM3b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   760
                     XMM4,  XMM4b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   761
                     XMM5,  XMM5b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   762
                     XMM6,  XMM6b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   763
                     XMM7,  XMM7b
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   764
#ifdef _LP64
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   765
                    ,XMM8,  XMM8b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   766
                     XMM9,  XMM9b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   767
                     XMM10, XMM10b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   768
                     XMM11, XMM11b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   769
                     XMM12, XMM12b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   770
                     XMM13, XMM13b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   771
                     XMM14, XMM14b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   772
                     XMM15, XMM15b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   773
                     XMM16, XMM16b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   774
                     XMM17, XMM17b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   775
                     XMM18, XMM18b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   776
                     XMM19, XMM19b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   777
                     XMM20, XMM20b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   778
                     XMM21, XMM21b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   779
                     XMM22, XMM22b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   780
                     XMM23, XMM23b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   781
                     XMM24, XMM24b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   782
                     XMM25, XMM25b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   783
                     XMM26, XMM26b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   784
                     XMM27, XMM27b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   785
                     XMM28, XMM28b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   786
                     XMM29, XMM29b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   787
                     XMM30, XMM30b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   788
                     XMM31, XMM31b
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   789
#endif
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   790
                     );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   791
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   792
reg_class_dynamic double_reg(double_reg_evex, double_reg_legacy, %{ VM_Version::supports_evex() %} );
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
   793
reg_class_dynamic double_reg_vl(double_reg_evex, double_reg_legacy, %{ VM_Version::supports_evex() && VM_Version::supports_avx512vl() %} );
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   794
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   795
// Class for pre evex 32bit vector registers
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   796
reg_class vectors_reg_legacy(XMM0,
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   797
                      XMM1,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   798
                      XMM2,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   799
                      XMM3,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   800
                      XMM4,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   801
                      XMM5,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   802
                      XMM6,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   803
                      XMM7
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   804
#ifdef _LP64
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   805
                     ,XMM8,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   806
                      XMM9,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   807
                      XMM10,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   808
                      XMM11,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   809
                      XMM12,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   810
                      XMM13,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   811
                      XMM14,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   812
                      XMM15
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   813
#endif
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   814
                      );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   815
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   816
// Class for evex 32bit vector registers
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   817
reg_class vectors_reg_evex(XMM0,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   818
                      XMM1,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   819
                      XMM2,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   820
                      XMM3,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   821
                      XMM4,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   822
                      XMM5,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   823
                      XMM6,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   824
                      XMM7
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   825
#ifdef _LP64
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   826
                     ,XMM8,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   827
                      XMM9,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   828
                      XMM10,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   829
                      XMM11,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   830
                      XMM12,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   831
                      XMM13,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   832
                      XMM14,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   833
                      XMM15,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   834
                      XMM16,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   835
                      XMM17,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   836
                      XMM18,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   837
                      XMM19,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   838
                      XMM20,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   839
                      XMM21,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   840
                      XMM22,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   841
                      XMM23,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   842
                      XMM24,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   843
                      XMM25,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   844
                      XMM26,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   845
                      XMM27,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   846
                      XMM28,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   847
                      XMM29,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   848
                      XMM30,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   849
                      XMM31
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   850
#endif
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   851
                      );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   852
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   853
reg_class_dynamic vectors_reg(vectors_reg_evex, vectors_reg_legacy, %{ VM_Version::supports_evex() %} );
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
   854
reg_class_dynamic vectors_reg_vlbwdq(vectors_reg_evex, vectors_reg_legacy, %{ VM_Version::supports_avx512vlbwdq() %} );
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   855
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   856
// Class for all 64bit vector registers
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   857
reg_class vectord_reg_legacy(XMM0,  XMM0b,
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   858
                      XMM1,  XMM1b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   859
                      XMM2,  XMM2b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   860
                      XMM3,  XMM3b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   861
                      XMM4,  XMM4b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   862
                      XMM5,  XMM5b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   863
                      XMM6,  XMM6b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   864
                      XMM7,  XMM7b
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   865
#ifdef _LP64
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   866
                     ,XMM8,  XMM8b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   867
                      XMM9,  XMM9b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   868
                      XMM10, XMM10b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   869
                      XMM11, XMM11b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   870
                      XMM12, XMM12b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   871
                      XMM13, XMM13b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   872
                      XMM14, XMM14b,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   873
                      XMM15, XMM15b
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   874
#endif
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   875
                      );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   876
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   877
// Class for all 64bit vector registers
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   878
reg_class vectord_reg_evex(XMM0,  XMM0b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   879
                      XMM1,  XMM1b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   880
                      XMM2,  XMM2b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   881
                      XMM3,  XMM3b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   882
                      XMM4,  XMM4b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   883
                      XMM5,  XMM5b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   884
                      XMM6,  XMM6b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   885
                      XMM7,  XMM7b
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   886
#ifdef _LP64
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   887
                     ,XMM8,  XMM8b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   888
                      XMM9,  XMM9b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   889
                      XMM10, XMM10b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   890
                      XMM11, XMM11b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   891
                      XMM12, XMM12b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   892
                      XMM13, XMM13b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   893
                      XMM14, XMM14b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   894
                      XMM15, XMM15b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   895
                      XMM16, XMM16b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   896
                      XMM17, XMM17b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   897
                      XMM18, XMM18b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   898
                      XMM19, XMM19b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   899
                      XMM20, XMM20b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   900
                      XMM21, XMM21b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   901
                      XMM22, XMM22b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   902
                      XMM23, XMM23b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   903
                      XMM24, XMM24b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   904
                      XMM25, XMM25b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   905
                      XMM26, XMM26b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   906
                      XMM27, XMM27b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   907
                      XMM28, XMM28b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   908
                      XMM29, XMM29b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   909
                      XMM30, XMM30b,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   910
                      XMM31, XMM31b
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   911
#endif
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   912
                      );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   913
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   914
reg_class_dynamic vectord_reg(vectord_reg_evex, vectord_reg_legacy, %{ VM_Version::supports_evex() %} );
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
   915
reg_class_dynamic vectord_reg_vlbwdq(vectord_reg_evex, vectord_reg_legacy, %{ VM_Version::supports_avx512vlbwdq() %} );
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   916
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   917
// Class for all 128bit vector registers
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   918
reg_class vectorx_reg_legacy(XMM0,  XMM0b,  XMM0c,  XMM0d,
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   919
                      XMM1,  XMM1b,  XMM1c,  XMM1d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   920
                      XMM2,  XMM2b,  XMM2c,  XMM2d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   921
                      XMM3,  XMM3b,  XMM3c,  XMM3d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   922
                      XMM4,  XMM4b,  XMM4c,  XMM4d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   923
                      XMM5,  XMM5b,  XMM5c,  XMM5d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   924
                      XMM6,  XMM6b,  XMM6c,  XMM6d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   925
                      XMM7,  XMM7b,  XMM7c,  XMM7d
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   926
#ifdef _LP64
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   927
                     ,XMM8,  XMM8b,  XMM8c,  XMM8d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   928
                      XMM9,  XMM9b,  XMM9c,  XMM9d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   929
                      XMM10, XMM10b, XMM10c, XMM10d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   930
                      XMM11, XMM11b, XMM11c, XMM11d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   931
                      XMM12, XMM12b, XMM12c, XMM12d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   932
                      XMM13, XMM13b, XMM13c, XMM13d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   933
                      XMM14, XMM14b, XMM14c, XMM14d,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   934
                      XMM15, XMM15b, XMM15c, XMM15d
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   935
#endif
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   936
                      );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   937
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   938
// Class for all 128bit vector registers
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   939
reg_class vectorx_reg_evex(XMM0,  XMM0b,  XMM0c,  XMM0d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   940
                      XMM1,  XMM1b,  XMM1c,  XMM1d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   941
                      XMM2,  XMM2b,  XMM2c,  XMM2d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   942
                      XMM3,  XMM3b,  XMM3c,  XMM3d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   943
                      XMM4,  XMM4b,  XMM4c,  XMM4d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   944
                      XMM5,  XMM5b,  XMM5c,  XMM5d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   945
                      XMM6,  XMM6b,  XMM6c,  XMM6d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   946
                      XMM7,  XMM7b,  XMM7c,  XMM7d
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   947
#ifdef _LP64
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   948
                     ,XMM8,  XMM8b,  XMM8c,  XMM8d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   949
                      XMM9,  XMM9b,  XMM9c,  XMM9d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   950
                      XMM10, XMM10b, XMM10c, XMM10d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   951
                      XMM11, XMM11b, XMM11c, XMM11d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   952
                      XMM12, XMM12b, XMM12c, XMM12d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   953
                      XMM13, XMM13b, XMM13c, XMM13d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   954
                      XMM14, XMM14b, XMM14c, XMM14d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   955
                      XMM15, XMM15b, XMM15c, XMM15d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   956
                      XMM16, XMM16b, XMM16c, XMM16d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   957
                      XMM17, XMM17b, XMM17c, XMM17d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   958
                      XMM18, XMM18b, XMM18c, XMM18d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   959
                      XMM19, XMM19b, XMM19c, XMM19d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   960
                      XMM20, XMM20b, XMM20c, XMM20d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   961
                      XMM21, XMM21b, XMM21c, XMM21d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   962
                      XMM22, XMM22b, XMM22c, XMM22d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   963
                      XMM23, XMM23b, XMM23c, XMM23d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   964
                      XMM24, XMM24b, XMM24c, XMM24d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   965
                      XMM25, XMM25b, XMM25c, XMM25d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   966
                      XMM26, XMM26b, XMM26c, XMM26d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   967
                      XMM27, XMM27b, XMM27c, XMM27d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   968
                      XMM28, XMM28b, XMM28c, XMM28d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   969
                      XMM29, XMM29b, XMM29c, XMM29d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   970
                      XMM30, XMM30b, XMM30c, XMM30d,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   971
                      XMM31, XMM31b, XMM31c, XMM31d
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   972
#endif
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   973
                      );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   974
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   975
reg_class_dynamic vectorx_reg(vectorx_reg_evex, vectorx_reg_legacy, %{ VM_Version::supports_evex() %} );
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
   976
reg_class_dynamic vectorx_reg_vlbwdq(vectorx_reg_evex, vectorx_reg_legacy, %{ VM_Version::supports_avx512vlbwdq() %} );
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   977
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   978
// Class for all 256bit vector registers
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   979
reg_class vectory_reg_legacy(XMM0,  XMM0b,  XMM0c,  XMM0d,  XMM0e,  XMM0f,  XMM0g,  XMM0h,
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   980
                      XMM1,  XMM1b,  XMM1c,  XMM1d,  XMM1e,  XMM1f,  XMM1g,  XMM1h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   981
                      XMM2,  XMM2b,  XMM2c,  XMM2d,  XMM2e,  XMM2f,  XMM2g,  XMM2h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   982
                      XMM3,  XMM3b,  XMM3c,  XMM3d,  XMM3e,  XMM3f,  XMM3g,  XMM3h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   983
                      XMM4,  XMM4b,  XMM4c,  XMM4d,  XMM4e,  XMM4f,  XMM4g,  XMM4h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   984
                      XMM5,  XMM5b,  XMM5c,  XMM5d,  XMM5e,  XMM5f,  XMM5g,  XMM5h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   985
                      XMM6,  XMM6b,  XMM6c,  XMM6d,  XMM6e,  XMM6f,  XMM6g,  XMM6h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   986
                      XMM7,  XMM7b,  XMM7c,  XMM7d,  XMM7e,  XMM7f,  XMM7g,  XMM7h
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   987
#ifdef _LP64
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   988
                     ,XMM8,  XMM8b,  XMM8c,  XMM8d,  XMM8e,  XMM8f,  XMM8g,  XMM8h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   989
                      XMM9,  XMM9b,  XMM9c,  XMM9d,  XMM9e,  XMM9f,  XMM9g,  XMM9h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   990
                      XMM10, XMM10b, XMM10c, XMM10d, XMM10e, XMM10f, XMM10g, XMM10h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   991
                      XMM11, XMM11b, XMM11c, XMM11d, XMM11e, XMM11f, XMM11g, XMM11h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   992
                      XMM12, XMM12b, XMM12c, XMM12d, XMM12e, XMM12f, XMM12g, XMM12h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   993
                      XMM13, XMM13b, XMM13c, XMM13d, XMM13e, XMM13f, XMM13g, XMM13h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   994
                      XMM14, XMM14b, XMM14c, XMM14d, XMM14e, XMM14f, XMM14g, XMM14h,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   995
                      XMM15, XMM15b, XMM15c, XMM15d, XMM15e, XMM15f, XMM15g, XMM15h
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   996
#endif
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   997
                      );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
   998
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
   999
// Class for all 256bit vector registers
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1000
reg_class vectory_reg_evex(XMM0,  XMM0b,  XMM0c,  XMM0d,  XMM0e,  XMM0f,  XMM0g,  XMM0h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1001
                      XMM1,  XMM1b,  XMM1c,  XMM1d,  XMM1e,  XMM1f,  XMM1g,  XMM1h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1002
                      XMM2,  XMM2b,  XMM2c,  XMM2d,  XMM2e,  XMM2f,  XMM2g,  XMM2h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1003
                      XMM3,  XMM3b,  XMM3c,  XMM3d,  XMM3e,  XMM3f,  XMM3g,  XMM3h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1004
                      XMM4,  XMM4b,  XMM4c,  XMM4d,  XMM4e,  XMM4f,  XMM4g,  XMM4h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1005
                      XMM5,  XMM5b,  XMM5c,  XMM5d,  XMM5e,  XMM5f,  XMM5g,  XMM5h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1006
                      XMM6,  XMM6b,  XMM6c,  XMM6d,  XMM6e,  XMM6f,  XMM6g,  XMM6h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1007
                      XMM7,  XMM7b,  XMM7c,  XMM7d,  XMM7e,  XMM7f,  XMM7g,  XMM7h
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1008
#ifdef _LP64
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1009
                     ,XMM8,  XMM8b,  XMM8c,  XMM8d,  XMM8e,  XMM8f,  XMM8g,  XMM8h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1010
                      XMM9,  XMM9b,  XMM9c,  XMM9d,  XMM9e,  XMM9f,  XMM9g,  XMM9h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1011
                      XMM10, XMM10b, XMM10c, XMM10d, XMM10e, XMM10f, XMM10g, XMM10h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1012
                      XMM11, XMM11b, XMM11c, XMM11d, XMM11e, XMM11f, XMM11g, XMM11h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1013
                      XMM12, XMM12b, XMM12c, XMM12d, XMM12e, XMM12f, XMM12g, XMM12h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1014
                      XMM13, XMM13b, XMM13c, XMM13d, XMM13e, XMM13f, XMM13g, XMM13h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1015
                      XMM14, XMM14b, XMM14c, XMM14d, XMM14e, XMM14f, XMM14g, XMM14h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1016
                      XMM15, XMM15b, XMM15c, XMM15d, XMM15e, XMM15f, XMM15g, XMM15h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1017
                      XMM16, XMM16b, XMM16c, XMM16d, XMM16e, XMM16f, XMM16g, XMM16h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1018
                      XMM17, XMM17b, XMM17c, XMM17d, XMM17e, XMM17f, XMM17g, XMM17h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1019
                      XMM18, XMM18b, XMM18c, XMM18d, XMM18e, XMM18f, XMM18g, XMM18h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1020
                      XMM19, XMM19b, XMM19c, XMM19d, XMM19e, XMM19f, XMM19g, XMM19h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1021
                      XMM20, XMM20b, XMM20c, XMM20d, XMM20e, XMM20f, XMM20g, XMM20h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1022
                      XMM21, XMM21b, XMM21c, XMM21d, XMM21e, XMM21f, XMM21g, XMM21h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1023
                      XMM22, XMM22b, XMM22c, XMM22d, XMM22e, XMM22f, XMM22g, XMM22h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1024
                      XMM23, XMM23b, XMM23c, XMM23d, XMM23e, XMM23f, XMM23g, XMM23h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1025
                      XMM24, XMM24b, XMM24c, XMM24d, XMM24e, XMM24f, XMM24g, XMM24h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1026
                      XMM25, XMM25b, XMM25c, XMM25d, XMM25e, XMM25f, XMM25g, XMM25h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1027
                      XMM26, XMM26b, XMM26c, XMM26d, XMM26e, XMM26f, XMM26g, XMM26h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1028
                      XMM27, XMM27b, XMM27c, XMM27d, XMM27e, XMM27f, XMM27g, XMM27h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1029
                      XMM28, XMM28b, XMM28c, XMM28d, XMM28e, XMM28f, XMM28g, XMM28h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1030
                      XMM29, XMM29b, XMM29c, XMM29d, XMM29e, XMM29f, XMM29g, XMM29h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1031
                      XMM30, XMM30b, XMM30c, XMM30d, XMM30e, XMM30f, XMM30g, XMM30h,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1032
                      XMM31, XMM31b, XMM31c, XMM31d, XMM31e, XMM31f, XMM31g, XMM31h
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1033
#endif
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1034
                      );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1035
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1036
reg_class_dynamic vectory_reg(vectory_reg_evex, vectory_reg_legacy, %{ VM_Version::supports_evex() %} );
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1037
reg_class_dynamic vectory_reg_vlbwdq(vectory_reg_evex, vectory_reg_legacy, %{ VM_Version::supports_avx512vlbwdq() %} );
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1038
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1039
// Class for all 512bit vector registers
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1040
reg_class vectorz_reg_evex(XMM0,  XMM0b,  XMM0c,  XMM0d,  XMM0e,  XMM0f,  XMM0g,  XMM0h,  XMM0i,  XMM0j,  XMM0k,  XMM0l,  XMM0m,  XMM0n,  XMM0o,  XMM0p,
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1041
                      XMM1,  XMM1b,  XMM1c,  XMM1d,  XMM1e,  XMM1f,  XMM1g,  XMM1h,  XMM1i,  XMM1j,  XMM1k,  XMM1l,  XMM1m,  XMM1n,  XMM1o,  XMM1p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1042
                      XMM2,  XMM2b,  XMM2c,  XMM2d,  XMM2e,  XMM2f,  XMM2g,  XMM2h,  XMM2i,  XMM2j,  XMM2k,  XMM2l,  XMM2m,  XMM2n,  XMM2o,  XMM2p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1043
                      XMM3,  XMM3b,  XMM3c,  XMM3d,  XMM3e,  XMM3f,  XMM3g,  XMM3h,  XMM3i,  XMM3j,  XMM3k,  XMM3l,  XMM3m,  XMM3n,  XMM3o,  XMM3p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1044
                      XMM4,  XMM4b,  XMM4c,  XMM4d,  XMM4e,  XMM4f,  XMM4g,  XMM4h,  XMM4i,  XMM4j,  XMM4k,  XMM4l,  XMM4m,  XMM4n,  XMM4o,  XMM4p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1045
                      XMM5,  XMM5b,  XMM5c,  XMM5d,  XMM5e,  XMM5f,  XMM5g,  XMM5h,  XMM5i,  XMM5j,  XMM5k,  XMM5l,  XMM5m,  XMM5n,  XMM5o,  XMM5p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1046
                      XMM6,  XMM6b,  XMM6c,  XMM6d,  XMM6e,  XMM6f,  XMM6g,  XMM6h,  XMM6i,  XMM6j,  XMM6k,  XMM6l,  XMM6m,  XMM6n,  XMM6o,  XMM6p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1047
                      XMM7,  XMM7b,  XMM7c,  XMM7d,  XMM7e,  XMM7f,  XMM7g,  XMM7h,  XMM7i,  XMM7j,  XMM7k,  XMM7l,  XMM7m,  XMM7n,  XMM7o,  XMM7p
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1048
#ifdef _LP64
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1049
                     ,XMM8,  XMM8b,  XMM8c,  XMM8d,  XMM8e,  XMM8f,  XMM8g,  XMM8h,  XMM8i,  XMM8j,  XMM8k,  XMM8l,  XMM8m,  XMM8n,  XMM8o,  XMM8p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1050
                      XMM9,  XMM9b,  XMM9c,  XMM9d,  XMM9e,  XMM9f,  XMM9g,  XMM9h,  XMM9i,  XMM9j,  XMM9k,  XMM9l,  XMM9m,  XMM9n,  XMM9o,  XMM9p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1051
                      XMM10, XMM10b, XMM10c, XMM10d, XMM10e, XMM10f, XMM10g, XMM10h, XMM10i, XMM10j, XMM10k, XMM10l, XMM10m, XMM10n, XMM10o, XMM10p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1052
                      XMM11, XMM11b, XMM11c, XMM11d, XMM11e, XMM11f, XMM11g, XMM11h, XMM11i, XMM11j, XMM11k, XMM11l, XMM11m, XMM11n, XMM11o, XMM11p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1053
                      XMM12, XMM12b, XMM12c, XMM12d, XMM12e, XMM12f, XMM12g, XMM12h, XMM12i, XMM12j, XMM12k, XMM12l, XMM12m, XMM12n, XMM12o, XMM12p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1054
                      XMM13, XMM13b, XMM13c, XMM13d, XMM13e, XMM13f, XMM13g, XMM13h, XMM13i, XMM13j, XMM13k, XMM13l, XMM13m, XMM13n, XMM13o, XMM13p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1055
                      XMM14, XMM14b, XMM14c, XMM14d, XMM14e, XMM14f, XMM14g, XMM14h, XMM14i, XMM14j, XMM14k, XMM14l, XMM14m, XMM14n, XMM14o, XMM14p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1056
                      XMM15, XMM15b, XMM15c, XMM15d, XMM15e, XMM15f, XMM15g, XMM15h, XMM15i, XMM15j, XMM15k, XMM15l, XMM15m, XMM15n, XMM15o, XMM15p
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1057
                     ,XMM16, XMM16b, XMM16c, XMM16d, XMM16e, XMM16f, XMM16g, XMM16h, XMM16i, XMM16j, XMM16k, XMM16l, XMM16m, XMM16n, XMM16o, XMM16p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1058
                      XMM17, XMM17b, XMM17c, XMM17d, XMM17e, XMM17f, XMM17g, XMM17h, XMM17i, XMM17j, XMM17k, XMM17l, XMM17m, XMM17n, XMM17o, XMM17p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1059
                      XMM18, XMM18b, XMM18c, XMM18d, XMM18e, XMM18f, XMM18g, XMM18h, XMM18i, XMM18j, XMM18k, XMM18l, XMM18m, XMM18n, XMM18o, XMM18p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1060
                      XMM19, XMM19b, XMM19c, XMM19d, XMM19e, XMM19f, XMM19g, XMM19h, XMM19i, XMM19j, XMM19k, XMM19l, XMM19m, XMM19n, XMM19o, XMM19p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1061
                      XMM20, XMM20b, XMM20c, XMM20d, XMM20e, XMM20f, XMM20g, XMM20h, XMM20i, XMM20j, XMM20k, XMM20l, XMM20m, XMM20n, XMM20o, XMM20p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1062
                      XMM21, XMM21b, XMM21c, XMM21d, XMM21e, XMM21f, XMM21g, XMM21h, XMM21i, XMM21j, XMM21k, XMM21l, XMM21m, XMM21n, XMM21o, XMM21p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1063
                      XMM22, XMM22b, XMM22c, XMM22d, XMM22e, XMM22f, XMM22g, XMM22h, XMM22i, XMM22j, XMM22k, XMM22l, XMM22m, XMM22n, XMM22o, XMM22p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1064
                      XMM23, XMM23b, XMM23c, XMM23d, XMM23e, XMM23f, XMM23g, XMM23h, XMM23i, XMM23j, XMM23k, XMM23l, XMM23m, XMM23n, XMM23o, XMM23p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1065
                      XMM24, XMM24b, XMM24c, XMM24d, XMM24e, XMM24f, XMM24g, XMM24h, XMM24i, XMM24j, XMM24k, XMM24l, XMM24m, XMM24n, XMM24o, XMM24p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1066
                      XMM25, XMM25b, XMM25c, XMM25d, XMM25e, XMM25f, XMM25g, XMM25h, XMM25i, XMM25j, XMM25k, XMM25l, XMM25m, XMM25n, XMM25o, XMM25p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1067
                      XMM26, XMM26b, XMM26c, XMM26d, XMM26e, XMM26f, XMM26g, XMM26h, XMM26i, XMM26j, XMM26k, XMM26l, XMM26m, XMM26n, XMM26o, XMM26p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1068
                      XMM27, XMM27b, XMM27c, XMM27d, XMM27e, XMM27f, XMM27g, XMM27h, XMM27i, XMM27j, XMM27k, XMM27l, XMM27m, XMM27n, XMM27o, XMM27p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1069
                      XMM28, XMM28b, XMM28c, XMM28d, XMM28e, XMM28f, XMM28g, XMM28h, XMM28i, XMM28j, XMM28k, XMM28l, XMM28m, XMM28n, XMM28o, XMM28p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1070
                      XMM29, XMM29b, XMM29c, XMM29d, XMM29e, XMM29f, XMM29g, XMM29h, XMM29i, XMM29j, XMM29k, XMM29l, XMM29m, XMM29n, XMM29o, XMM29p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1071
                      XMM30, XMM30b, XMM30c, XMM30d, XMM30e, XMM30f, XMM30g, XMM30h, XMM30i, XMM30j, XMM30k, XMM30l, XMM30m, XMM30n, XMM30o, XMM30p,
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1072
                      XMM31, XMM31b, XMM31c, XMM31d, XMM31e, XMM31f, XMM31g, XMM31h, XMM31i, XMM31j, XMM31k, XMM31l, XMM31m, XMM31n, XMM31o, XMM31p
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1073
#endif
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1074
                      );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1075
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1076
// Class for restricted 512bit vector registers
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1077
reg_class vectorz_reg_legacy(XMM0,  XMM0b,  XMM0c,  XMM0d,  XMM0e,  XMM0f,  XMM0g,  XMM0h,  XMM0i,  XMM0j,  XMM0k,  XMM0l,  XMM0m,  XMM0n,  XMM0o,  XMM0p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1078
                      XMM1,  XMM1b,  XMM1c,  XMM1d,  XMM1e,  XMM1f,  XMM1g,  XMM1h,  XMM1i,  XMM1j,  XMM1k,  XMM1l,  XMM1m,  XMM1n,  XMM1o,  XMM1p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1079
                      XMM2,  XMM2b,  XMM2c,  XMM2d,  XMM2e,  XMM2f,  XMM2g,  XMM2h,  XMM2i,  XMM2j,  XMM2k,  XMM2l,  XMM2m,  XMM2n,  XMM2o,  XMM2p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1080
                      XMM3,  XMM3b,  XMM3c,  XMM3d,  XMM3e,  XMM3f,  XMM3g,  XMM3h,  XMM3i,  XMM3j,  XMM3k,  XMM3l,  XMM3m,  XMM3n,  XMM3o,  XMM3p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1081
                      XMM4,  XMM4b,  XMM4c,  XMM4d,  XMM4e,  XMM4f,  XMM4g,  XMM4h,  XMM4i,  XMM4j,  XMM4k,  XMM4l,  XMM4m,  XMM4n,  XMM4o,  XMM4p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1082
                      XMM5,  XMM5b,  XMM5c,  XMM5d,  XMM5e,  XMM5f,  XMM5g,  XMM5h,  XMM5i,  XMM5j,  XMM5k,  XMM5l,  XMM5m,  XMM5n,  XMM5o,  XMM5p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1083
                      XMM6,  XMM6b,  XMM6c,  XMM6d,  XMM6e,  XMM6f,  XMM6g,  XMM6h,  XMM6i,  XMM6j,  XMM6k,  XMM6l,  XMM6m,  XMM6n,  XMM6o,  XMM6p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1084
                      XMM7,  XMM7b,  XMM7c,  XMM7d,  XMM7e,  XMM7f,  XMM7g,  XMM7h,  XMM7i,  XMM7j,  XMM7k,  XMM7l,  XMM7m,  XMM7n,  XMM7o,  XMM7p
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1085
#ifdef _LP64
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1086
                     ,XMM8,  XMM8b,  XMM8c,  XMM8d,  XMM8e,  XMM8f,  XMM8g,  XMM8h,  XMM8i,  XMM8j,  XMM8k,  XMM8l,  XMM8m,  XMM8n,  XMM8o,  XMM8p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1087
                      XMM9,  XMM9b,  XMM9c,  XMM9d,  XMM9e,  XMM9f,  XMM9g,  XMM9h,  XMM9i,  XMM9j,  XMM9k,  XMM9l,  XMM9m,  XMM9n,  XMM9o,  XMM9p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1088
                      XMM10, XMM10b, XMM10c, XMM10d, XMM10e, XMM10f, XMM10g, XMM10h, XMM10i, XMM10j, XMM10k, XMM10l, XMM10m, XMM10n, XMM10o, XMM10p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1089
                      XMM11, XMM11b, XMM11c, XMM11d, XMM11e, XMM11f, XMM11g, XMM11h, XMM11i, XMM11j, XMM11k, XMM11l, XMM11m, XMM11n, XMM11o, XMM11p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1090
                      XMM12, XMM12b, XMM12c, XMM12d, XMM12e, XMM12f, XMM12g, XMM12h, XMM12i, XMM12j, XMM12k, XMM12l, XMM12m, XMM12n, XMM12o, XMM12p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1091
                      XMM13, XMM13b, XMM13c, XMM13d, XMM13e, XMM13f, XMM13g, XMM13h, XMM13i, XMM13j, XMM13k, XMM13l, XMM13m, XMM13n, XMM13o, XMM13p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1092
                      XMM14, XMM14b, XMM14c, XMM14d, XMM14e, XMM14f, XMM14g, XMM14h, XMM14i, XMM14j, XMM14k, XMM14l, XMM14m, XMM14n, XMM14o, XMM14p,
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1093
                      XMM15, XMM15b, XMM15c, XMM15d, XMM15e, XMM15f, XMM15g, XMM15h, XMM15i, XMM15j, XMM15k, XMM15l, XMM15m, XMM15n, XMM15o, XMM15p
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1094
#endif
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1095
                      );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1096
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1097
reg_class_dynamic vectorz_reg(vectorz_reg_evex, vectorz_reg_legacy, %{ VM_Version::supports_evex() %} );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1098
reg_class_dynamic vectorz_reg_vl(vectorz_reg_evex, vectorz_reg_legacy, %{ VM_Version::supports_evex() && VM_Version::supports_avx512vl() %} );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1099
50525
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1100
reg_class xmm0_reg(XMM0, XMM0b, XMM0c, XMM0d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1101
reg_class ymm0_reg(XMM0, XMM0b, XMM0c, XMM0d, XMM0e, XMM0f, XMM0g, XMM0h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1102
reg_class zmm0_reg(XMM0, XMM0b, XMM0c, XMM0d, XMM0e, XMM0f, XMM0g, XMM0h, XMM0i, XMM0j, XMM0k, XMM0l, XMM0m, XMM0n, XMM0o, XMM0p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1103
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1104
reg_class xmm1_reg(XMM1, XMM1b, XMM1c, XMM1d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1105
reg_class ymm1_reg(XMM1, XMM1b, XMM1c, XMM1d, XMM1e, XMM1f, XMM1g, XMM1h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1106
reg_class zmm1_reg(XMM1, XMM1b, XMM1c, XMM1d, XMM1e, XMM1f, XMM1g, XMM1h, XMM1i, XMM1j, XMM1k, XMM1l, XMM1m, XMM1n, XMM1o, XMM1p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1107
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1108
reg_class xmm2_reg(XMM2, XMM2b, XMM2c, XMM2d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1109
reg_class ymm2_reg(XMM2, XMM2b, XMM2c, XMM2d, XMM2e, XMM2f, XMM2g, XMM2h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1110
reg_class zmm2_reg(XMM2, XMM2b, XMM2c, XMM2d, XMM2e, XMM2f, XMM2g, XMM2h, XMM2i, XMM2j, XMM2k, XMM2l, XMM2m, XMM2n, XMM2o, XMM2p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1111
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1112
reg_class xmm3_reg(XMM3, XMM3b, XMM3c, XMM3d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1113
reg_class ymm3_reg(XMM3, XMM3b, XMM3c, XMM3d, XMM3e, XMM3f, XMM3g, XMM3h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1114
reg_class zmm3_reg(XMM3, XMM3b, XMM3c, XMM3d, XMM3e, XMM3f, XMM3g, XMM3h, XMM3i, XMM3j, XMM3k, XMM3l, XMM3m, XMM3n, XMM3o, XMM3p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1115
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1116
reg_class xmm4_reg(XMM4, XMM4b, XMM4c, XMM4d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1117
reg_class ymm4_reg(XMM4, XMM4b, XMM4c, XMM4d, XMM4e, XMM4f, XMM4g, XMM4h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1118
reg_class zmm4_reg(XMM4, XMM4b, XMM4c, XMM4d, XMM4e, XMM4f, XMM4g, XMM4h, XMM4i, XMM4j, XMM4k, XMM4l, XMM4m, XMM4n, XMM4o, XMM4p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1119
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1120
reg_class xmm5_reg(XMM5, XMM5b, XMM5c, XMM5d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1121
reg_class ymm5_reg(XMM5, XMM5b, XMM5c, XMM5d, XMM5e, XMM5f, XMM5g, XMM5h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1122
reg_class zmm5_reg(XMM5, XMM5b, XMM5c, XMM5d, XMM5e, XMM5f, XMM5g, XMM5h, XMM5i, XMM5j, XMM5k, XMM5l, XMM5m, XMM5n, XMM5o, XMM5p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1123
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1124
reg_class xmm6_reg(XMM6, XMM6b, XMM6c, XMM6d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1125
reg_class ymm6_reg(XMM6, XMM6b, XMM6c, XMM6d, XMM6e, XMM6f, XMM6g, XMM6h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1126
reg_class zmm6_reg(XMM6, XMM6b, XMM6c, XMM6d, XMM6e, XMM6f, XMM6g, XMM6h, XMM6i, XMM6j, XMM6k, XMM6l, XMM6m, XMM6n, XMM6o, XMM6p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1127
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1128
reg_class xmm7_reg(XMM7, XMM7b, XMM7c, XMM7d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1129
reg_class ymm7_reg(XMM7, XMM7b, XMM7c, XMM7d, XMM7e, XMM7f, XMM7g, XMM7h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1130
reg_class zmm7_reg(XMM7, XMM7b, XMM7c, XMM7d, XMM7e, XMM7f, XMM7g, XMM7h, XMM7i, XMM7j, XMM7k, XMM7l, XMM7m, XMM7n, XMM7o, XMM7p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1131
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1132
#ifdef _LP64
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1133
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1134
reg_class xmm8_reg(XMM8, XMM8b, XMM8c, XMM8d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1135
reg_class ymm8_reg(XMM8, XMM8b, XMM8c, XMM8d, XMM8e, XMM8f, XMM8g, XMM8h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1136
reg_class zmm8_reg(XMM8, XMM8b, XMM8c, XMM8d, XMM8e, XMM8f, XMM8g, XMM8h, XMM8i, XMM8j, XMM8k, XMM8l, XMM8m, XMM8n, XMM8o, XMM8p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1137
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1138
reg_class xmm9_reg(XMM9, XMM9b, XMM9c, XMM9d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1139
reg_class ymm9_reg(XMM9, XMM9b, XMM9c, XMM9d, XMM9e, XMM9f, XMM9g, XMM9h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1140
reg_class zmm9_reg(XMM9, XMM9b, XMM9c, XMM9d, XMM9e, XMM9f, XMM9g, XMM9h, XMM9i, XMM9j, XMM9k, XMM9l, XMM9m, XMM9n, XMM9o, XMM9p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1141
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1142
reg_class xmm10_reg(XMM10, XMM10b, XMM10c, XMM10d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1143
reg_class ymm10_reg(XMM10, XMM10b, XMM10c, XMM10d, XMM10e, XMM10f, XMM10g, XMM10h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1144
reg_class zmm10_reg(XMM10, XMM10b, XMM10c, XMM10d, XMM10e, XMM10f, XMM10g, XMM10h, XMM10i, XMM10j, XMM10k, XMM10l, XMM10m, XMM10n, XMM10o, XMM10p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1145
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1146
reg_class xmm11_reg(XMM11, XMM11b, XMM11c, XMM11d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1147
reg_class ymm11_reg(XMM11, XMM11b, XMM11c, XMM11d, XMM11e, XMM11f, XMM11g, XMM11h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1148
reg_class zmm11_reg(XMM11, XMM11b, XMM11c, XMM11d, XMM11e, XMM11f, XMM11g, XMM11h, XMM11i, XMM11j, XMM11k, XMM11l, XMM11m, XMM11n, XMM11o, XMM11p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1149
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1150
reg_class xmm12_reg(XMM12, XMM12b, XMM12c, XMM12d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1151
reg_class ymm12_reg(XMM12, XMM12b, XMM12c, XMM12d, XMM12e, XMM12f, XMM12g, XMM12h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1152
reg_class zmm12_reg(XMM12, XMM12b, XMM12c, XMM12d, XMM12e, XMM12f, XMM12g, XMM12h, XMM12i, XMM12j, XMM12k, XMM12l, XMM12m, XMM12n, XMM12o, XMM12p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1153
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1154
reg_class xmm13_reg(XMM13, XMM13b, XMM13c, XMM13d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1155
reg_class ymm13_reg(XMM13, XMM13b, XMM13c, XMM13d, XMM13e, XMM13f, XMM13g, XMM13h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1156
reg_class zmm13_reg(XMM13, XMM13b, XMM13c, XMM13d, XMM13e, XMM13f, XMM13g, XMM13h, XMM13i, XMM13j, XMM13k, XMM13l, XMM13m, XMM13n, XMM13o, XMM13p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1157
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1158
reg_class xmm14_reg(XMM14, XMM14b, XMM14c, XMM14d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1159
reg_class ymm14_reg(XMM14, XMM14b, XMM14c, XMM14d, XMM14e, XMM14f, XMM14g, XMM14h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1160
reg_class zmm14_reg(XMM14, XMM14b, XMM14c, XMM14d, XMM14e, XMM14f, XMM14g, XMM14h, XMM14i, XMM14j, XMM14k, XMM14l, XMM14m, XMM14n, XMM14o, XMM14p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1161
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1162
reg_class xmm15_reg(XMM15, XMM15b, XMM15c, XMM15d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1163
reg_class ymm15_reg(XMM15, XMM15b, XMM15c, XMM15d, XMM15e, XMM15f, XMM15g, XMM15h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1164
reg_class zmm15_reg(XMM15, XMM15b, XMM15c, XMM15d, XMM15e, XMM15f, XMM15g, XMM15h, XMM15i, XMM15j, XMM15k, XMM15l, XMM15m, XMM15n, XMM15o, XMM15p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1165
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1166
reg_class xmm16_reg(XMM16, XMM16b, XMM16c, XMM16d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1167
reg_class ymm16_reg(XMM16, XMM16b, XMM16c, XMM16d, XMM16e, XMM16f, XMM16g, XMM16h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1168
reg_class zmm16_reg(XMM16, XMM16b, XMM16c, XMM16d, XMM16e, XMM16f, XMM16g, XMM16h, XMM16i, XMM16j, XMM16k, XMM16l, XMM16m, XMM16n, XMM16o, XMM16p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1169
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1170
reg_class xmm17_reg(XMM17, XMM17b, XMM17c, XMM17d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1171
reg_class ymm17_reg(XMM17, XMM17b, XMM17c, XMM17d, XMM17e, XMM17f, XMM17g, XMM17h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1172
reg_class zmm17_reg(XMM17, XMM17b, XMM17c, XMM17d, XMM17e, XMM17f, XMM17g, XMM17h, XMM17i, XMM17j, XMM17k, XMM17l, XMM17m, XMM17n, XMM17o, XMM17p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1173
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1174
reg_class xmm18_reg(XMM18, XMM18b, XMM18c, XMM18d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1175
reg_class ymm18_reg(XMM18, XMM18b, XMM18c, XMM18d, XMM18e, XMM18f, XMM18g, XMM18h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1176
reg_class zmm18_reg(XMM18, XMM18b, XMM18c, XMM18d, XMM18e, XMM18f, XMM18g, XMM18h, XMM18i, XMM18j, XMM18k, XMM18l, XMM18m, XMM18n, XMM18o, XMM18p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1177
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1178
reg_class xmm19_reg(XMM19, XMM19b, XMM19c, XMM19d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1179
reg_class ymm19_reg(XMM19, XMM19b, XMM19c, XMM19d, XMM19e, XMM19f, XMM19g, XMM19h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1180
reg_class zmm19_reg(XMM19, XMM19b, XMM19c, XMM19d, XMM19e, XMM19f, XMM19g, XMM19h, XMM19i, XMM19j, XMM19k, XMM19l, XMM19m, XMM19n, XMM19o, XMM19p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1181
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1182
reg_class xmm20_reg(XMM20, XMM20b, XMM20c, XMM20d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1183
reg_class ymm20_reg(XMM20, XMM20b, XMM20c, XMM20d, XMM20e, XMM20f, XMM20g, XMM20h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1184
reg_class zmm20_reg(XMM20, XMM20b, XMM20c, XMM20d, XMM20e, XMM20f, XMM20g, XMM20h, XMM20i, XMM20j, XMM20k, XMM20l, XMM20m, XMM20n, XMM20o, XMM20p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1185
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1186
reg_class xmm21_reg(XMM21, XMM21b, XMM21c, XMM21d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1187
reg_class ymm21_reg(XMM21, XMM21b, XMM21c, XMM21d, XMM21e, XMM21f, XMM21g, XMM21h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1188
reg_class zmm21_reg(XMM21, XMM21b, XMM21c, XMM21d, XMM21e, XMM21f, XMM21g, XMM21h, XMM21i, XMM21j, XMM21k, XMM21l, XMM21m, XMM21n, XMM21o, XMM21p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1189
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1190
reg_class xmm22_reg(XMM22, XMM22b, XMM22c, XMM22d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1191
reg_class ymm22_reg(XMM22, XMM22b, XMM22c, XMM22d, XMM22e, XMM22f, XMM22g, XMM22h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1192
reg_class zmm22_reg(XMM22, XMM22b, XMM22c, XMM22d, XMM22e, XMM22f, XMM22g, XMM22h, XMM22i, XMM22j, XMM22k, XMM22l, XMM22m, XMM22n, XMM22o, XMM22p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1193
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1194
reg_class xmm23_reg(XMM23, XMM23b, XMM23c, XMM23d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1195
reg_class ymm23_reg(XMM23, XMM23b, XMM23c, XMM23d, XMM23e, XMM23f, XMM23g, XMM23h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1196
reg_class zmm23_reg(XMM23, XMM23b, XMM23c, XMM23d, XMM23e, XMM23f, XMM23g, XMM23h, XMM23i, XMM23j, XMM23k, XMM23l, XMM23m, XMM23n, XMM23o, XMM23p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1197
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1198
reg_class xmm24_reg(XMM24, XMM24b, XMM24c, XMM24d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1199
reg_class ymm24_reg(XMM24, XMM24b, XMM24c, XMM24d, XMM24e, XMM24f, XMM24g, XMM24h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1200
reg_class zmm24_reg(XMM24, XMM24b, XMM24c, XMM24d, XMM24e, XMM24f, XMM24g, XMM24h, XMM24i, XMM24j, XMM24k, XMM24l, XMM24m, XMM24n, XMM24o, XMM24p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1201
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1202
reg_class xmm25_reg(XMM25, XMM25b, XMM25c, XMM25d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1203
reg_class ymm25_reg(XMM25, XMM25b, XMM25c, XMM25d, XMM25e, XMM25f, XMM25g, XMM25h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1204
reg_class zmm25_reg(XMM25, XMM25b, XMM25c, XMM25d, XMM25e, XMM25f, XMM25g, XMM25h, XMM25i, XMM25j, XMM25k, XMM25l, XMM25m, XMM25n, XMM25o, XMM25p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1205
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1206
reg_class xmm26_reg(XMM26, XMM26b, XMM26c, XMM26d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1207
reg_class ymm26_reg(XMM26, XMM26b, XMM26c, XMM26d, XMM26e, XMM26f, XMM26g, XMM26h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1208
reg_class zmm26_reg(XMM26, XMM26b, XMM26c, XMM26d, XMM26e, XMM26f, XMM26g, XMM26h, XMM26i, XMM26j, XMM26k, XMM26l, XMM26m, XMM26n, XMM26o, XMM26p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1209
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1210
reg_class xmm27_reg(XMM27, XMM27b, XMM27c, XMM27d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1211
reg_class ymm27_reg(XMM27, XMM27b, XMM27c, XMM27d, XMM27e, XMM27f, XMM27g, XMM27h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1212
reg_class zmm27_reg(XMM27, XMM27b, XMM27c, XMM27d, XMM27e, XMM27f, XMM27g, XMM27h, XMM27i, XMM27j, XMM27k, XMM27l, XMM27m, XMM27n, XMM27o, XMM27p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1213
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1214
reg_class xmm28_reg(XMM28, XMM28b, XMM28c, XMM28d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1215
reg_class ymm28_reg(XMM28, XMM28b, XMM28c, XMM28d, XMM28e, XMM28f, XMM28g, XMM28h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1216
reg_class zmm28_reg(XMM28, XMM28b, XMM28c, XMM28d, XMM28e, XMM28f, XMM28g, XMM28h, XMM28i, XMM28j, XMM28k, XMM28l, XMM28m, XMM28n, XMM28o, XMM28p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1217
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1218
reg_class xmm29_reg(XMM29, XMM29b, XMM29c, XMM29d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1219
reg_class ymm29_reg(XMM29, XMM29b, XMM29c, XMM29d, XMM29e, XMM29f, XMM29g, XMM29h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1220
reg_class zmm29_reg(XMM29, XMM29b, XMM29c, XMM29d, XMM29e, XMM29f, XMM29g, XMM29h, XMM29i, XMM29j, XMM29k, XMM29l, XMM29m, XMM29n, XMM29o, XMM29p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1221
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1222
reg_class xmm30_reg(XMM30, XMM30b, XMM30c, XMM30d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1223
reg_class ymm30_reg(XMM30, XMM30b, XMM30c, XMM30d, XMM30e, XMM30f, XMM30g, XMM30h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1224
reg_class zmm30_reg(XMM30, XMM30b, XMM30c, XMM30d, XMM30e, XMM30f, XMM30g, XMM30h, XMM30i, XMM30j, XMM30k, XMM30l, XMM30m, XMM30n, XMM30o, XMM30p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1225
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1226
reg_class xmm31_reg(XMM31, XMM31b, XMM31c, XMM31d);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1227
reg_class ymm31_reg(XMM31, XMM31b, XMM31c, XMM31d, XMM31e, XMM31f, XMM31g, XMM31h);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1228
reg_class zmm31_reg(XMM31, XMM31b, XMM31c, XMM31d, XMM31e, XMM31f, XMM31g, XMM31h, XMM31i, XMM31j, XMM31k, XMM31l, XMM31m, XMM31n, XMM31o, XMM31p);
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1229
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1230
#endif
767cdb97f103 8204210: Implementation: JEP 333: ZGC: A Scalable Low-Latency Garbage Collector (Experimental)
pliden
parents: 49384
diff changeset
  1231
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1232
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1233
23498
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1234
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1235
//----------SOURCE BLOCK-------------------------------------------------------
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1236
// This is a block of C++ code which provides values, functions, and
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1237
// definitions necessary in the rest of the architecture description
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1238
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1239
source_hpp %{
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1240
// Header information of the source block.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1241
// Method declarations/definitions which are used outside
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1242
// the ad-scope can conveniently be defined here.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1243
//
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1244
// To keep related declarations/definitions/uses close together,
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1245
// we switch between source %{ }% and source_hpp %{ }% freely as needed.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1246
25715
d5a8dbdc5150 8049325: Introduce and clean up umbrella headers for the files in the cpu subdirectories.
goetz
parents: 23498
diff changeset
  1247
class NativeJump;
d5a8dbdc5150 8049325: Introduce and clean up umbrella headers for the files in the cpu subdirectories.
goetz
parents: 23498
diff changeset
  1248
23498
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1249
class CallStubImpl {
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1250
23498
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1251
  //--------------------------------------------------------------
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1252
  //---<  Used for optimization in Compile::shorten_branches  >---
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1253
  //--------------------------------------------------------------
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1254
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1255
 public:
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1256
  // Size of call trampoline stub.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1257
  static uint size_call_trampoline() {
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1258
    return 0; // no call trampolines on this platform
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1259
  }
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1260
23498
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1261
  // number of relocations needed by a call trampoline stub
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1262
  static uint reloc_call_trampoline() {
23498
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1263
    return 0; // no call trampolines on this platform
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1264
  }
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1265
};
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1266
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1267
class HandlerImpl {
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1268
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1269
 public:
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1270
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1271
  static int emit_exception_handler(CodeBuffer &cbuf);
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1272
  static int emit_deopt_handler(CodeBuffer& cbuf);
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1273
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1274
  static uint size_exception_handler() {
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1275
    // NativeCall instruction size is the same as NativeJump.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1276
    // exception handler starts out as jump and can be patched to
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1277
    // a call be deoptimization.  (4932387)
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1278
    // Note that this value is also credited (in output.cpp) to
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1279
    // the size of the code section.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1280
    return NativeJump::instruction_size;
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1281
  }
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1282
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1283
#ifdef _LP64
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1284
  static uint size_deopt_handler() {
51633
21154cb84d2a 8209594: guarantee(this->is8bit(imm8)) failed: Short forward jump exceeds 8-bit offset
kvn
parents: 51078
diff changeset
  1285
    // three 5 byte instructions plus one move for unreachable address.
21154cb84d2a 8209594: guarantee(this->is8bit(imm8)) failed: Short forward jump exceeds 8-bit offset
kvn
parents: 51078
diff changeset
  1286
    return 15+3;
23498
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1287
  }
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1288
#else
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1289
  static uint size_deopt_handler() {
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1290
    // NativeCall instruction size is the same as NativeJump.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1291
    // exception handler starts out as jump and can be patched to
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1292
    // a call be deoptimization.  (4932387)
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1293
    // Note that this value is also credited (in output.cpp) to
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1294
    // the size of the code section.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1295
    return 5 + NativeJump::instruction_size; // pushl(); jmp;
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1296
  }
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1297
#endif
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1298
};
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1299
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1300
%} // end source_hpp
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1301
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1302
source %{
23498
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1303
38286
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1304
#include "opto/addnode.hpp"
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1305
23498
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1306
// Emit exception handler code.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1307
// Stuff framesize into a register and call a VM stub routine.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1308
int HandlerImpl::emit_exception_handler(CodeBuffer& cbuf) {
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1309
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1310
  // Note that the code buffer's insts_mark is always relative to insts.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1311
  // That's why we must use the macroassembler to generate a handler.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1312
  MacroAssembler _masm(&cbuf);
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1313
  address base = __ start_a_stub(size_exception_handler());
32082
2a3323e25de1 8130309: Need to bailout cleanly if creation of stubs fails when codecache is out of space
thartmann
parents: 31410
diff changeset
  1314
  if (base == NULL) {
2a3323e25de1 8130309: Need to bailout cleanly if creation of stubs fails when codecache is out of space
thartmann
parents: 31410
diff changeset
  1315
    ciEnv::current()->record_failure("CodeCache is full");
2a3323e25de1 8130309: Need to bailout cleanly if creation of stubs fails when codecache is out of space
thartmann
parents: 31410
diff changeset
  1316
    return 0;  // CodeBuffer::expand failed
2a3323e25de1 8130309: Need to bailout cleanly if creation of stubs fails when codecache is out of space
thartmann
parents: 31410
diff changeset
  1317
  }
23498
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1318
  int offset = __ offset();
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1319
  __ jump(RuntimeAddress(OptoRuntime::exception_blob()->entry_point()));
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1320
  assert(__ offset() - offset <= (int) size_exception_handler(), "overflow");
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1321
  __ end_a_stub();
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1322
  return offset;
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1323
}
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1324
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1325
// Emit deopt handler code.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1326
int HandlerImpl::emit_deopt_handler(CodeBuffer& cbuf) {
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1327
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1328
  // Note that the code buffer's insts_mark is always relative to insts.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1329
  // That's why we must use the macroassembler to generate a handler.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1330
  MacroAssembler _masm(&cbuf);
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1331
  address base = __ start_a_stub(size_deopt_handler());
32082
2a3323e25de1 8130309: Need to bailout cleanly if creation of stubs fails when codecache is out of space
thartmann
parents: 31410
diff changeset
  1332
  if (base == NULL) {
2a3323e25de1 8130309: Need to bailout cleanly if creation of stubs fails when codecache is out of space
thartmann
parents: 31410
diff changeset
  1333
    ciEnv::current()->record_failure("CodeCache is full");
2a3323e25de1 8130309: Need to bailout cleanly if creation of stubs fails when codecache is out of space
thartmann
parents: 31410
diff changeset
  1334
    return 0;  // CodeBuffer::expand failed
2a3323e25de1 8130309: Need to bailout cleanly if creation of stubs fails when codecache is out of space
thartmann
parents: 31410
diff changeset
  1335
  }
23498
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1336
  int offset = __ offset();
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1337
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1338
#ifdef _LP64
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1339
  address the_pc = (address) __ pc();
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1340
  Label next;
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1341
  // push a "the_pc" on the stack without destroying any registers
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1342
  // as they all may be live.
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1343
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1344
  // push address of "next"
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1345
  __ call(next, relocInfo::none); // reloc none is fine since it is a disp32
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1346
  __ bind(next);
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1347
  // adjust it so it matches "the_pc"
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1348
  __ subptr(Address(rsp, 0), __ offset() - offset);
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1349
#else
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1350
  InternalAddress here(__ pc());
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1351
  __ pushptr(here.addr());
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1352
#endif
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1353
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1354
  __ jump(RuntimeAddress(SharedRuntime::deopt_blob()->unpack()));
51633
21154cb84d2a 8209594: guarantee(this->is8bit(imm8)) failed: Short forward jump exceeds 8-bit offset
kvn
parents: 51078
diff changeset
  1355
  assert(__ offset() - offset <= (int) size_deopt_handler(), "overflow %d", (__ offset() - offset));
23498
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1356
  __ end_a_stub();
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1357
  return offset;
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1358
}
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1359
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1360
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1361
//=============================================================================
a0e67b766e5c 8037821: Account for trampoline stubs when estimating code buffer sizes
goetz
parents: 22505
diff changeset
  1362
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1363
  // Float masks come from different places depending on platform.
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1364
#ifdef _LP64
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1365
  static address float_signmask()  { return StubRoutines::x86::float_sign_mask(); }
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1366
  static address float_signflip()  { return StubRoutines::x86::float_sign_flip(); }
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1367
  static address double_signmask() { return StubRoutines::x86::double_sign_mask(); }
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1368
  static address double_signflip() { return StubRoutines::x86::double_sign_flip(); }
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1369
#else
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1370
  static address float_signmask()  { return (address)float_signmask_pool; }
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1371
  static address float_signflip()  { return (address)float_signflip_pool; }
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1372
  static address double_signmask() { return (address)double_signmask_pool; }
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1373
  static address double_signflip() { return (address)double_signflip_pool; }
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1374
#endif
11794
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1375
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1376
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1377
const bool Matcher::match_rule_supported(int opcode) {
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1378
  if (!has_match_rule(opcode))
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1379
    return false;
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1380
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1381
  bool ret_value = true;
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1382
  switch (opcode) {
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1383
    case Op_PopCountI:
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1384
    case Op_PopCountL:
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1385
      if (!UsePopCountInstruction)
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1386
        ret_value = false;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1387
      break;
49384
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  1388
    case Op_PopCountVI:
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  1389
      if (!UsePopCountInstruction || !VM_Version::supports_vpopcntdq())
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  1390
        ret_value = false;
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  1391
      break;
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1392
    case Op_MulVI:
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1393
      if ((UseSSE < 4) && (UseAVX < 1)) // only with SSE4_1 or AVX
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1394
        ret_value = false;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1395
      break;
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1396
    case Op_MulVL:
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1397
    case Op_MulReductionVL:
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1398
      if (VM_Version::supports_avx512dq() == false)
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1399
        ret_value = false;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1400
      break;
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1401
    case Op_AddReductionVL:
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1402
      if (UseAVX < 3) // only EVEX : vector connectivity becomes an issue here
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1403
        ret_value = false;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1404
      break;
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1405
    case Op_AddReductionVI:
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1406
      if (UseSSE < 3) // requires at least SSE3
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1407
        ret_value = false;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1408
      break;
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1409
    case Op_MulReductionVI:
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1410
      if (UseSSE < 4) // requires at least SSE4
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1411
        ret_value = false;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1412
      break;
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1413
    case Op_AddReductionVF:
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1414
    case Op_AddReductionVD:
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1415
    case Op_MulReductionVF:
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1416
    case Op_MulReductionVD:
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  1417
      if (UseSSE < 1) // requires at least SSE
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1418
        ret_value = false;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1419
      break;
32723
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  1420
    case Op_SqrtVD:
48089
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  1421
    case Op_SqrtVF:
32723
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  1422
      if (UseAVX < 1) // enabled for AVX only
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1423
        ret_value = false;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1424
      break;
13886
8d82c4dfa722 7023898: Intrinsify AtomicLongFieldUpdater.getAndIncrement()
roland
parents: 13883
diff changeset
  1425
    case Op_CompareAndSwapL:
8d82c4dfa722 7023898: Intrinsify AtomicLongFieldUpdater.getAndIncrement()
roland
parents: 13883
diff changeset
  1426
#ifdef _LP64
8d82c4dfa722 7023898: Intrinsify AtomicLongFieldUpdater.getAndIncrement()
roland
parents: 13883
diff changeset
  1427
    case Op_CompareAndSwapP:
8d82c4dfa722 7023898: Intrinsify AtomicLongFieldUpdater.getAndIncrement()
roland
parents: 13883
diff changeset
  1428
#endif
8d82c4dfa722 7023898: Intrinsify AtomicLongFieldUpdater.getAndIncrement()
roland
parents: 13883
diff changeset
  1429
      if (!VM_Version::supports_cx8())
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1430
        ret_value = false;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1431
      break;
48309
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  1432
    case Op_CMoveVF:
33469
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  1433
    case Op_CMoveVD:
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  1434
      if (UseAVX < 1 || UseAVX > 2)
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  1435
        ret_value = false;
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  1436
      break;
35581
dd47cf4734f2 8145336: PPC64: fix string intrinsics after CompactStrings change
simonis
parents: 34162
diff changeset
  1437
    case Op_StrIndexOf:
dd47cf4734f2 8145336: PPC64: fix string intrinsics after CompactStrings change
simonis
parents: 34162
diff changeset
  1438
      if (!UseSSE42Intrinsics)
dd47cf4734f2 8145336: PPC64: fix string intrinsics after CompactStrings change
simonis
parents: 34162
diff changeset
  1439
        ret_value = false;
dd47cf4734f2 8145336: PPC64: fix string intrinsics after CompactStrings change
simonis
parents: 34162
diff changeset
  1440
      break;
dd47cf4734f2 8145336: PPC64: fix string intrinsics after CompactStrings change
simonis
parents: 34162
diff changeset
  1441
    case Op_StrIndexOfChar:
39253
bd5fe208734e 8157842: indexOfChar intrinsic is not emitted on x86
thartmann
parents: 38286
diff changeset
  1442
      if (!UseSSE42Intrinsics)
35581
dd47cf4734f2 8145336: PPC64: fix string intrinsics after CompactStrings change
simonis
parents: 34162
diff changeset
  1443
        ret_value = false;
dd47cf4734f2 8145336: PPC64: fix string intrinsics after CompactStrings change
simonis
parents: 34162
diff changeset
  1444
      break;
38017
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  1445
    case Op_OnSpinWait:
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  1446
      if (VM_Version::supports_on_spin_wait() == false)
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  1447
        ret_value = false;
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  1448
      break;
52992
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  1449
    case Op_MulAddVS2VI:
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  1450
      if (UseSSE < 2)
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  1451
        ret_value = false;
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  1452
      break;
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1453
  }
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1454
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1455
  return ret_value;  // Per default match rules are supported.
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1456
}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  1457
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1458
const bool Matcher::match_rule_supported_vector(int opcode, int vlen) {
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1459
  // identify extra cases that we might want to provide match rules for
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1460
  // e.g. Op_ vector nodes and other intrinsics while guarding with vlen
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1461
  bool ret_value = match_rule_supported(opcode);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1462
  if (ret_value) {
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1463
    switch (opcode) {
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1464
      case Op_AddVB:
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1465
      case Op_SubVB:
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1466
        if ((vlen == 64) && (VM_Version::supports_avx512bw() == false))
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1467
          ret_value = false;
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1468
        break;
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1469
      case Op_URShiftVS:
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1470
      case Op_RShiftVS:
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1471
      case Op_LShiftVS:
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1472
      case Op_MulVS:
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1473
      case Op_AddVS:
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1474
      case Op_SubVS:
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1475
        if ((vlen == 32) && (VM_Version::supports_avx512bw() == false))
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1476
          ret_value = false;
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1477
        break;
48309
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  1478
      case Op_CMoveVF:
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  1479
        if (vlen != 8)
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  1480
          ret_value  = false;
51078
fc6cfe40e32a 8207049: Minor improvements of compiler code.
goetz
parents: 50525
diff changeset
  1481
        break;
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1482
      case Op_CMoveVD:
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1483
        if (vlen != 4)
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1484
          ret_value  = false;
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1485
        break;
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1486
    }
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1487
  }
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1488
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1489
  return ret_value;  // Per default match rules are supported.
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1490
}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1491
38049
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  1492
const bool Matcher::has_predicated_vectors(void) {
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  1493
  bool ret_value = false;
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  1494
  if (UseAVX > 2) {
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  1495
    ret_value = VM_Version::supports_avx512vl();
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  1496
  }
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  1497
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  1498
  return ret_value;
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  1499
}
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  1500
33065
55892792936f 8134802: LCM register pressure scheduling
mcberg
parents: 32727
diff changeset
  1501
const int Matcher::float_pressure(int default_pressure_threshold) {
55892792936f 8134802: LCM register pressure scheduling
mcberg
parents: 32727
diff changeset
  1502
  int float_pressure_threshold = default_pressure_threshold;
55892792936f 8134802: LCM register pressure scheduling
mcberg
parents: 32727
diff changeset
  1503
#ifdef _LP64
55892792936f 8134802: LCM register pressure scheduling
mcberg
parents: 32727
diff changeset
  1504
  if (UseAVX > 2) {
55892792936f 8134802: LCM register pressure scheduling
mcberg
parents: 32727
diff changeset
  1505
    // Increase pressure threshold on machines with AVX3 which have
55892792936f 8134802: LCM register pressure scheduling
mcberg
parents: 32727
diff changeset
  1506
    // 2x more XMM registers.
55892792936f 8134802: LCM register pressure scheduling
mcberg
parents: 32727
diff changeset
  1507
    float_pressure_threshold = default_pressure_threshold * 2;
55892792936f 8134802: LCM register pressure scheduling
mcberg
parents: 32727
diff changeset
  1508
  }
55892792936f 8134802: LCM register pressure scheduling
mcberg
parents: 32727
diff changeset
  1509
#endif
55892792936f 8134802: LCM register pressure scheduling
mcberg
parents: 32727
diff changeset
  1510
  return float_pressure_threshold;
55892792936f 8134802: LCM register pressure scheduling
mcberg
parents: 32727
diff changeset
  1511
}
55892792936f 8134802: LCM register pressure scheduling
mcberg
parents: 32727
diff changeset
  1512
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1513
// Max vector size in bytes. 0 if not supported.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1514
const int Matcher::vector_width_in_bytes(BasicType bt) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1515
  assert(is_java_primitive(bt), "only primitive type vectors");
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1516
  if (UseSSE < 2) return 0;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1517
  // SSE2 supports 128bit vectors for all types.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1518
  // AVX2 supports 256bit vectors for all types.
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1519
  // AVX2/EVEX supports 512bit vectors for all types.
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1520
  int size = (UseAVX > 1) ? (1 << UseAVX) * 8 : 16;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1521
  // AVX1 supports 256bit vectors only for FLOAT and DOUBLE.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1522
  if (UseAVX > 0 && (bt == T_FLOAT || bt == T_DOUBLE))
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1523
    size = (UseAVX > 2) ? 64 : 32;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1524
  if (UseAVX > 2 && (bt == T_BYTE || bt == T_SHORT || bt == T_CHAR))
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1525
    size = (VM_Version::supports_avx512bw()) ? 64 : 32;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1526
  // Use flag to limit vector size.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1527
  size = MIN2(size,(int)MaxVectorSize);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1528
  // Minimum 2 values in vector (or 4 for bytes).
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1529
  switch (bt) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1530
  case T_DOUBLE:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1531
  case T_LONG:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1532
    if (size < 16) return 0;
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1533
    break;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1534
  case T_FLOAT:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1535
  case T_INT:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1536
    if (size < 8) return 0;
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1537
    break;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1538
  case T_BOOLEAN:
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1539
    if (size < 4) return 0;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1540
    break;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1541
  case T_CHAR:
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1542
    if (size < 4) return 0;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1543
    break;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1544
  case T_BYTE:
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1545
    if (size < 4) return 0;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1546
    break;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1547
  case T_SHORT:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1548
    if (size < 4) return 0;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1549
    break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1550
  default:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1551
    ShouldNotReachHere();
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1552
  }
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1553
  return size;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1554
}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1555
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1556
// Limits on vector size (number of elements) loaded into vector.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1557
const int Matcher::max_vector_size(const BasicType bt) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1558
  return vector_width_in_bytes(bt)/type2aelembytes(bt);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1559
}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1560
const int Matcher::min_vector_size(const BasicType bt) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1561
  int max_size = max_vector_size(bt);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1562
  // Min size which can be loaded into vector is 4 bytes.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1563
  int size = (type2aelembytes(bt) == 1) ? 4 : 2;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1564
  return MIN2(size,max_size);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1565
}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1566
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1567
// Vector ideal reg corresponding to specified size in bytes
46378
4ccca1fdf627 8160748: Inconsistent types for ideal_reg
kbarrett
parents: 42039
diff changeset
  1568
const uint Matcher::vector_ideal_reg(int size) {
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1569
  assert(MaxVectorSize >= size, "");
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1570
  switch(size) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1571
    case  4: return Op_VecS;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1572
    case  8: return Op_VecD;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1573
    case 16: return Op_VecX;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1574
    case 32: return Op_VecY;
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1575
    case 64: return Op_VecZ;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1576
  }
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1577
  ShouldNotReachHere();
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1578
  return 0;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1579
}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1580
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  1581
// Only lowest bits of xmm reg are used for vector shift count.
46378
4ccca1fdf627 8160748: Inconsistent types for ideal_reg
kbarrett
parents: 42039
diff changeset
  1582
const uint Matcher::vector_shift_count_ideal_reg(int size) {
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  1583
  return Op_VecS;
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  1584
}
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  1585
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1586
// x86 supports misaligned vectors store/load.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1587
const bool Matcher::misaligned_vectors_ok() {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1588
  return !AlignVector; // can be changed by flag
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1589
}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1590
22505
4523090c9674 8002074: Support for AES on SPARC
kvn
parents: 14131
diff changeset
  1591
// x86 AES instructions are compatible with SunJCE expanded
4523090c9674 8002074: Support for AES on SPARC
kvn
parents: 14131
diff changeset
  1592
// keys, hence we do not need to pass the original key to stubs
4523090c9674 8002074: Support for AES on SPARC
kvn
parents: 14131
diff changeset
  1593
const bool Matcher::pass_original_key_for_aes() {
4523090c9674 8002074: Support for AES on SPARC
kvn
parents: 14131
diff changeset
  1594
  return false;
4523090c9674 8002074: Support for AES on SPARC
kvn
parents: 14131
diff changeset
  1595
}
4523090c9674 8002074: Support for AES on SPARC
kvn
parents: 14131
diff changeset
  1596
38286
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1597
38236
510f77046e00 8154943: AArch64: redundant address computation instructions with vectorization
roland
parents: 38049
diff changeset
  1598
const bool Matcher::convi2l_type_required = true;
510f77046e00 8154943: AArch64: redundant address computation instructions with vectorization
roland
parents: 38049
diff changeset
  1599
38286
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1600
// Check for shift by small constant as well
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1601
static bool clone_shift(Node* shift, Matcher* matcher, Matcher::MStack& mstack, VectorSet& address_visited) {
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1602
  if (shift->Opcode() == Op_LShiftX && shift->in(2)->is_Con() &&
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1603
      shift->in(2)->get_int() <= 3 &&
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1604
      // Are there other uses besides address expressions?
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1605
      !matcher->is_visited(shift)) {
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1606
    address_visited.set(shift->_idx); // Flag as address_visited
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1607
    mstack.push(shift->in(2), Matcher::Visit);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1608
    Node *conv = shift->in(1);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1609
#ifdef _LP64
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1610
    // Allow Matcher to match the rule which bypass
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1611
    // ConvI2L operation for an array index on LP64
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1612
    // if the index value is positive.
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1613
    if (conv->Opcode() == Op_ConvI2L &&
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1614
        conv->as_Type()->type()->is_long()->_lo >= 0 &&
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1615
        // Are there other uses besides address expressions?
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1616
        !matcher->is_visited(conv)) {
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1617
      address_visited.set(conv->_idx); // Flag as address_visited
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1618
      mstack.push(conv->in(1), Matcher::Pre_Visit);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1619
    } else
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1620
#endif
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1621
      mstack.push(conv, Matcher::Pre_Visit);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1622
    return true;
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1623
  }
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1624
  return false;
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1625
}
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1626
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1627
// Should the Matcher clone shifts on addressing modes, expecting them
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1628
// to be subsumed into complex addressing expressions or compute them
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1629
// into registers?
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1630
bool Matcher::clone_address_expressions(AddPNode* m, Matcher::MStack& mstack, VectorSet& address_visited) {
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1631
  Node *off = m->in(AddPNode::Offset);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1632
  if (off->is_Con()) {
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1633
    address_visited.test_set(m->_idx); // Flag as address_visited
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1634
    Node *adr = m->in(AddPNode::Address);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1635
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1636
    // Intel can handle 2 adds in addressing mode
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1637
    // AtomicAdd is not an addressing expression.
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1638
    // Cheap to find it by looking for screwy base.
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1639
    if (adr->is_AddP() &&
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1640
        !adr->in(AddPNode::Base)->is_top() &&
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1641
        // Are there other uses besides address expressions?
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1642
        !is_visited(adr)) {
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1643
      address_visited.set(adr->_idx); // Flag as address_visited
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1644
      Node *shift = adr->in(AddPNode::Offset);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1645
      if (!clone_shift(shift, this, mstack, address_visited)) {
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1646
        mstack.push(shift, Pre_Visit);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1647
      }
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1648
      mstack.push(adr->in(AddPNode::Address), Pre_Visit);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1649
      mstack.push(adr->in(AddPNode::Base), Pre_Visit);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1650
    } else {
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1651
      mstack.push(adr, Pre_Visit);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1652
    }
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1653
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1654
    // Clone X+offset as it also folds into most addressing expressions
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1655
    mstack.push(off, Visit);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1656
    mstack.push(m->in(AddPNode::Base), Pre_Visit);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1657
    return true;
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1658
  } else if (clone_shift(off, this, mstack, address_visited)) {
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1659
    address_visited.test_set(m->_idx); // Flag as address_visited
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1660
    mstack.push(m->in(AddPNode::Address), Pre_Visit);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1661
    mstack.push(m->in(AddPNode::Base), Pre_Visit);
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1662
    return true;
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1663
  }
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1664
  return false;
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1665
}
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1666
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1667
void Compile::reshape_address(AddPNode* addp) {
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1668
}
0ddb6f84e138 8154826: AArch64: take advantage better of base + shifted offset addressing mode
roland
parents: 38236
diff changeset
  1669
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1670
// Helper methods for MachSpillCopyNode::implementation().
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1671
static int vec_mov_helper(CodeBuffer *cbuf, bool do_size, int src_lo, int dst_lo,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1672
                          int src_hi, int dst_hi, uint ireg, outputStream* st) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1673
  // In 64-bit VM size calculation is very complex. Emitting instructions
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1674
  // into scratch buffer is used to get size in 64-bit VM.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1675
  LP64_ONLY( assert(!do_size, "this method calculates size only for 32-bit VM"); )
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1676
  assert(ireg == Op_VecS || // 32bit vector
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1677
         (src_lo & 1) == 0 && (src_lo + 1) == src_hi &&
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1678
         (dst_lo & 1) == 0 && (dst_lo + 1) == dst_hi,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1679
         "no non-adjacent vector moves" );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1680
  if (cbuf) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1681
    MacroAssembler _masm(cbuf);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1682
    int offset = __ offset();
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1683
    switch (ireg) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1684
    case Op_VecS: // copy whole register
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1685
    case Op_VecD:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1686
    case Op_VecX:
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1687
#ifndef LP64
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1688
      __ movdqu(as_XMMRegister(Matcher::_regEncode[dst_lo]), as_XMMRegister(Matcher::_regEncode[src_lo]));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1689
#else
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1690
      if ((UseAVX < 3) || VM_Version::supports_avx512vl()) {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1691
        __ movdqu(as_XMMRegister(Matcher::_regEncode[dst_lo]), as_XMMRegister(Matcher::_regEncode[src_lo]));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1692
      } else {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1693
        __ vpxor(as_XMMRegister(Matcher::_regEncode[dst_lo]), as_XMMRegister(Matcher::_regEncode[dst_lo]), 2);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1694
        __ vinserti32x4(as_XMMRegister(Matcher::_regEncode[dst_lo]), as_XMMRegister(Matcher::_regEncode[dst_lo]), as_XMMRegister(Matcher::_regEncode[src_lo]), 0x0);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1695
     }
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1696
#endif
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1697
      break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1698
    case Op_VecY:
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1699
#ifndef LP64
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1700
      __ vmovdqu(as_XMMRegister(Matcher::_regEncode[dst_lo]), as_XMMRegister(Matcher::_regEncode[src_lo]));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1701
#else
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1702
      if ((UseAVX < 3) || VM_Version::supports_avx512vl()) {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1703
        __ vmovdqu(as_XMMRegister(Matcher::_regEncode[dst_lo]), as_XMMRegister(Matcher::_regEncode[src_lo]));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1704
      } else {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1705
        __ vpxor(as_XMMRegister(Matcher::_regEncode[dst_lo]), as_XMMRegister(Matcher::_regEncode[dst_lo]), 2);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1706
        __ vinserti64x4(as_XMMRegister(Matcher::_regEncode[dst_lo]), as_XMMRegister(Matcher::_regEncode[dst_lo]), as_XMMRegister(Matcher::_regEncode[src_lo]), 0x0);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1707
     }
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1708
#endif
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1709
      break;
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1710
    case Op_VecZ:
38049
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  1711
      __ evmovdquq(as_XMMRegister(Matcher::_regEncode[dst_lo]), as_XMMRegister(Matcher::_regEncode[src_lo]), 2);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1712
      break;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1713
    default:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1714
      ShouldNotReachHere();
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1715
    }
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1716
    int size = __ offset() - offset;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1717
#ifdef ASSERT
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1718
    // VEX_2bytes prefix is used if UseAVX > 0, so it takes the same 2 bytes as SIMD prefix.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1719
    assert(!do_size || size == 4, "incorrect size calculattion");
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1720
#endif
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1721
    return size;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1722
#ifndef PRODUCT
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1723
  } else if (!do_size) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1724
    switch (ireg) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1725
    case Op_VecS:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1726
    case Op_VecD:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1727
    case Op_VecX:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1728
      st->print("movdqu  %s,%s\t# spill",Matcher::regName[dst_lo],Matcher::regName[src_lo]);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1729
      break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1730
    case Op_VecY:
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1731
    case Op_VecZ:
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1732
      st->print("vmovdqu %s,%s\t# spill",Matcher::regName[dst_lo],Matcher::regName[src_lo]);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1733
      break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1734
    default:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1735
      ShouldNotReachHere();
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1736
    }
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1737
#endif
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1738
  }
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1739
  // VEX_2bytes prefix is used if UseAVX > 0, and it takes the same 2 bytes as SIMD prefix.
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1740
  return (UseAVX > 2) ? 6 : 4;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1741
}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1742
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1743
static int vec_spill_helper(CodeBuffer *cbuf, bool do_size, bool is_load,
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1744
                            int stack_offset, int reg, uint ireg, outputStream* st) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1745
  // In 64-bit VM size calculation is very complex. Emitting instructions
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1746
  // into scratch buffer is used to get size in 64-bit VM.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1747
  LP64_ONLY( assert(!do_size, "this method calculates size only for 32-bit VM"); )
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1748
  if (cbuf) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1749
    MacroAssembler _masm(cbuf);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1750
    int offset = __ offset();
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1751
    if (is_load) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1752
      switch (ireg) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1753
      case Op_VecS:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1754
        __ movdl(as_XMMRegister(Matcher::_regEncode[reg]), Address(rsp, stack_offset));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1755
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1756
      case Op_VecD:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1757
        __ movq(as_XMMRegister(Matcher::_regEncode[reg]), Address(rsp, stack_offset));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1758
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1759
      case Op_VecX:
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1760
#ifndef LP64
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1761
        __ movdqu(as_XMMRegister(Matcher::_regEncode[reg]), Address(rsp, stack_offset));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1762
#else
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1763
        if ((UseAVX < 3) || VM_Version::supports_avx512vl()) {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1764
          __ movdqu(as_XMMRegister(Matcher::_regEncode[reg]), Address(rsp, stack_offset));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1765
        } else {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1766
          __ vpxor(as_XMMRegister(Matcher::_regEncode[reg]), as_XMMRegister(Matcher::_regEncode[reg]), 2);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1767
          __ vinserti32x4(as_XMMRegister(Matcher::_regEncode[reg]), as_XMMRegister(Matcher::_regEncode[reg]), Address(rsp, stack_offset),0x0);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1768
        }
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1769
#endif
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1770
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1771
      case Op_VecY:
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1772
#ifndef LP64
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1773
        __ vmovdqu(as_XMMRegister(Matcher::_regEncode[reg]), Address(rsp, stack_offset));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1774
#else
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1775
        if ((UseAVX < 3) || VM_Version::supports_avx512vl()) {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1776
          __ vmovdqu(as_XMMRegister(Matcher::_regEncode[reg]), Address(rsp, stack_offset));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1777
        } else {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1778
          __ vpxor(as_XMMRegister(Matcher::_regEncode[reg]), as_XMMRegister(Matcher::_regEncode[reg]), 2);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1779
          __ vinserti64x4(as_XMMRegister(Matcher::_regEncode[reg]), as_XMMRegister(Matcher::_regEncode[reg]), Address(rsp, stack_offset),0x0);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1780
        }
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1781
#endif
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1782
        break;
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1783
      case Op_VecZ:
38049
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  1784
        __ evmovdquq(as_XMMRegister(Matcher::_regEncode[reg]), Address(rsp, stack_offset), 2);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1785
        break;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1786
      default:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1787
        ShouldNotReachHere();
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1788
      }
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1789
    } else { // store
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1790
      switch (ireg) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1791
      case Op_VecS:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1792
        __ movdl(Address(rsp, stack_offset), as_XMMRegister(Matcher::_regEncode[reg]));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1793
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1794
      case Op_VecD:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1795
        __ movq(Address(rsp, stack_offset), as_XMMRegister(Matcher::_regEncode[reg]));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1796
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1797
      case Op_VecX:
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1798
#ifndef LP64
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1799
        __ movdqu(Address(rsp, stack_offset), as_XMMRegister(Matcher::_regEncode[reg]));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1800
#else
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1801
        if ((UseAVX < 3) || VM_Version::supports_avx512vl()) {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1802
          __ movdqu(Address(rsp, stack_offset), as_XMMRegister(Matcher::_regEncode[reg]));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1803
        }
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1804
        else {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1805
          __ vextracti32x4(Address(rsp, stack_offset), as_XMMRegister(Matcher::_regEncode[reg]), as_XMMRegister(Matcher::_regEncode[reg]), 0x0);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1806
        }
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1807
#endif
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1808
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1809
      case Op_VecY:
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1810
#ifndef LP64
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1811
        __ vmovdqu(Address(rsp, stack_offset), as_XMMRegister(Matcher::_regEncode[reg]));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1812
#else
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1813
        if ((UseAVX < 3) || VM_Version::supports_avx512vl()) {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1814
          __ vmovdqu(Address(rsp, stack_offset), as_XMMRegister(Matcher::_regEncode[reg]));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1815
        }
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1816
        else {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1817
          __ vextracti64x4(Address(rsp, stack_offset), as_XMMRegister(Matcher::_regEncode[reg]), as_XMMRegister(Matcher::_regEncode[reg]), 0x0);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1818
        }
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  1819
#endif
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1820
        break;
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1821
      case Op_VecZ:
38049
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  1822
        __ evmovdquq(Address(rsp, stack_offset), as_XMMRegister(Matcher::_regEncode[reg]), 2);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1823
        break;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1824
      default:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1825
        ShouldNotReachHere();
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1826
      }
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1827
    }
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1828
    int size = __ offset() - offset;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1829
#ifdef ASSERT
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1830
    int offset_size = (stack_offset == 0) ? 0 : ((stack_offset < 0x80) ? 1 : (UseAVX > 2) ? 6 : 4);
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1831
    // VEX_2bytes prefix is used if UseAVX > 0, so it takes the same 2 bytes as SIMD prefix.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1832
    assert(!do_size || size == (5+offset_size), "incorrect size calculattion");
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1833
#endif
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1834
    return size;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1835
#ifndef PRODUCT
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1836
  } else if (!do_size) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1837
    if (is_load) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1838
      switch (ireg) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1839
      case Op_VecS:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1840
        st->print("movd    %s,[rsp + %d]\t# spill", Matcher::regName[reg], stack_offset);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1841
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1842
      case Op_VecD:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1843
        st->print("movq    %s,[rsp + %d]\t# spill", Matcher::regName[reg], stack_offset);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1844
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1845
       case Op_VecX:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1846
        st->print("movdqu  %s,[rsp + %d]\t# spill", Matcher::regName[reg], stack_offset);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1847
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1848
      case Op_VecY:
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1849
      case Op_VecZ:
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1850
        st->print("vmovdqu %s,[rsp + %d]\t# spill", Matcher::regName[reg], stack_offset);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1851
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1852
      default:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1853
        ShouldNotReachHere();
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1854
      }
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1855
    } else { // store
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1856
      switch (ireg) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1857
      case Op_VecS:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1858
        st->print("movd    [rsp + %d],%s\t# spill", stack_offset, Matcher::regName[reg]);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1859
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1860
      case Op_VecD:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1861
        st->print("movq    [rsp + %d],%s\t# spill", stack_offset, Matcher::regName[reg]);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1862
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1863
       case Op_VecX:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1864
        st->print("movdqu  [rsp + %d],%s\t# spill", stack_offset, Matcher::regName[reg]);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1865
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1866
      case Op_VecY:
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  1867
      case Op_VecZ:
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1868
        st->print("vmovdqu [rsp + %d],%s\t# spill", stack_offset, Matcher::regName[reg]);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1869
        break;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1870
      default:
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1871
        ShouldNotReachHere();
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1872
      }
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1873
    }
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1874
#endif
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1875
  }
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1876
  bool is_single_byte = false;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1877
  int vec_len = 0;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1878
  if ((UseAVX > 2) && (stack_offset != 0)) {
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1879
    int tuple_type = Assembler::EVEX_FVM;
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1880
    int input_size = Assembler::EVEX_32bit;
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1881
    switch (ireg) {
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1882
    case Op_VecS:
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1883
      tuple_type = Assembler::EVEX_T1S;
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1884
      break;
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1885
    case Op_VecD:
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1886
      tuple_type = Assembler::EVEX_T1S;
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1887
      input_size = Assembler::EVEX_64bit;
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1888
      break;
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1889
    case Op_VecX:
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1890
      break;
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1891
    case Op_VecY:
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1892
      vec_len = 1;
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1893
      break;
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1894
    case Op_VecZ:
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1895
      vec_len = 2;
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1896
      break;
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1897
    }
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1898
    is_single_byte = Assembler::query_compressed_disp_byte(stack_offset, true, vec_len, tuple_type, input_size, 0);
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1899
  }
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1900
  int offset_size = 0;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1901
  int size = 5;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1902
  if (UseAVX > 2 ) {
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1903
    if (VM_Version::supports_avx512novl() && (vec_len == 2)) {
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1904
      offset_size = (stack_offset == 0) ? 0 : ((is_single_byte) ? 1 : 4);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1905
      size += 2; // Need an additional two bytes for EVEX encoding
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  1906
    } else if (VM_Version::supports_avx512novl() && (vec_len < 2)) {
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1907
      offset_size = (stack_offset == 0) ? 0 : ((stack_offset <= 127) ? 1 : 4);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1908
    } else {
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1909
      offset_size = (stack_offset == 0) ? 0 : ((is_single_byte) ? 1 : 4);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1910
      size += 2; // Need an additional two bytes for EVEX encodding
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1911
    }
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1912
  } else {
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1913
    offset_size = (stack_offset == 0) ? 0 : ((stack_offset <= 127) ? 1 : 4);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1914
  }
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1915
  // VEX_2bytes prefix is used if UseAVX > 0, so it takes the same 2 bytes as SIMD prefix.
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  1916
  return size+offset_size;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1917
}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1918
40040
7644f470d923 8160425: Vectorization with signalling NaN returns wrong result
thartmann
parents: 39253
diff changeset
  1919
static inline jint replicate4_imm(int con, int width) {
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1920
  // Load a constant of "width" (in bytes) and replicate it to fill 32bit.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1921
  assert(width == 1 || width == 2, "only byte or short types here");
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1922
  int bit_width = width * 8;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1923
  jint val = con;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1924
  val &= (1 << bit_width) - 1;  // mask off sign bits
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1925
  while(bit_width < 32) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1926
    val |= (val << bit_width);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1927
    bit_width <<= 1;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1928
  }
40040
7644f470d923 8160425: Vectorization with signalling NaN returns wrong result
thartmann
parents: 39253
diff changeset
  1929
  return val;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1930
}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1931
40040
7644f470d923 8160425: Vectorization with signalling NaN returns wrong result
thartmann
parents: 39253
diff changeset
  1932
static inline jlong replicate8_imm(int con, int width) {
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1933
  // Load a constant of "width" (in bytes) and replicate it to fill 64bit.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1934
  assert(width == 1 || width == 2 || width == 4, "only byte, short or int types here");
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1935
  int bit_width = width * 8;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1936
  jlong val = con;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1937
  val &= (((jlong) 1) << bit_width) - 1;  // mask off sign bits
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1938
  while(bit_width < 64) {
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1939
    val |= (val << bit_width);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1940
    bit_width <<= 1;
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1941
  }
40040
7644f470d923 8160425: Vectorization with signalling NaN returns wrong result
thartmann
parents: 39253
diff changeset
  1942
  return val;
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1943
}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1944
11794
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1945
#ifndef PRODUCT
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1946
  void MachNopNode::format(PhaseRegAlloc*, outputStream* st) const {
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1947
    st->print("nop \t# %d bytes pad for loops and calls", _count);
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1948
  }
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1949
#endif
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1950
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1951
  void MachNopNode::emit(CodeBuffer &cbuf, PhaseRegAlloc*) const {
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1952
    MacroAssembler _masm(&cbuf);
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1953
    __ nop(_count);
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1954
  }
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1955
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1956
  uint MachNopNode::size(PhaseRegAlloc*) const {
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1957
    return _count;
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1958
  }
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1959
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1960
#ifndef PRODUCT
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1961
  void MachBreakpointNode::format(PhaseRegAlloc*, outputStream* st) const {
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1962
    st->print("# breakpoint");
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1963
  }
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1964
#endif
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1965
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1966
  void MachBreakpointNode::emit(CodeBuffer &cbuf, PhaseRegAlloc* ra_) const {
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1967
    MacroAssembler _masm(&cbuf);
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1968
    __ int3();
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1969
  }
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1970
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1971
  uint MachBreakpointNode::size(PhaseRegAlloc* ra_) const {
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1972
    return MachNode::size(ra_);
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1973
  }
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1974
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1975
%}
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1976
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1977
encode %{
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1978
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1979
  enc_class call_epilog %{
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1980
    if (VerifyStackAtCalls) {
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1981
      // Check that stack depth is unchanged: find majik cookie on stack
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1982
      int framesize = ra_->reg2offset_unchecked(OptoReg::add(ra_->_matcher._old_SP, -3*VMRegImpl::slots_per_word));
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1983
      MacroAssembler _masm(&cbuf);
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1984
      Label L;
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1985
      __ cmpptr(Address(rsp, framesize), (int32_t)0xbadb100d);
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1986
      __ jccb(Assembler::equal, L);
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1987
      // Die if stack mismatch
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1988
      __ int3();
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1989
      __ bind(L);
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1990
    }
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1991
  %}
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  1992
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1993
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  1994
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1995
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1996
//----------OPERANDS-----------------------------------------------------------
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1997
// Operand definitions must precede instruction definitions for correct parsing
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1998
// in the ADLC because operands constitute user defined types which are used in
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  1999
// instruction definitions.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2000
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  2001
operand vecZ() %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  2002
  constraint(ALLOC_IN_RC(vectorz_reg));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  2003
  match(VecZ);
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2004
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2005
  format %{ %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2006
  interface(REG_INTER);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2007
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2008
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2009
operand legVecZ() %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2010
  constraint(ALLOC_IN_RC(vectorz_reg_vl));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2011
  match(VecZ);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2012
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2013
  format %{ %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2014
  interface(REG_INTER);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2015
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2016
33469
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2017
// Comparison Code for FP conditional move
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2018
operand cmpOp_vcmppd() %{
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2019
  match(Bool);
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2020
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2021
  predicate(n->as_Bool()->_test._test != BoolTest::overflow &&
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2022
            n->as_Bool()->_test._test != BoolTest::no_overflow);
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2023
  format %{ "" %}
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2024
  interface(COND_INTER) %{
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2025
    equal        (0x0, "eq");
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2026
    less         (0x1, "lt");
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2027
    less_equal   (0x2, "le");
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2028
    not_equal    (0xC, "ne");
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2029
    greater_equal(0xD, "ge");
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2030
    greater      (0xE, "gt");
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2031
    //TODO cannot compile (adlc breaks) without two next lines with error:
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2032
    // x86_64.ad(13987) Syntax Error: :In operand cmpOp_vcmppd: Do not support this encode constant: ' %{
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2033
    // equal' for overflow.
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2034
    overflow     (0x20, "o");  // not really supported by the instruction
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2035
    no_overflow  (0x21, "no"); // not really supported by the instruction
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2036
  %}
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2037
%}
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2038
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  2039
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2040
// INSTRUCTIONS -- Platform independent definitions (same for 32- and 64-bit)
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2041
11794
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  2042
// ============================================================================
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  2043
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  2044
instruct ShouldNotReachHere() %{
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  2045
  match(Halt);
46525
3a5c833a43de 8176506: C2: loop unswitching and unsafe accesses cause crash
roland
parents: 46378
diff changeset
  2046
  format %{ "ud2\t# ShouldNotReachHere" %}
3a5c833a43de 8176506: C2: loop unswitching and unsafe accesses cause crash
roland
parents: 46378
diff changeset
  2047
  ins_encode %{
3a5c833a43de 8176506: C2: loop unswitching and unsafe accesses cause crash
roland
parents: 46378
diff changeset
  2048
    __ ud2();
11794
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  2049
  %}
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  2050
  ins_pipe(pipe_slow);
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  2051
%}
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  2052
38049
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2053
// =================================EVEX special===============================
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2054
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2055
instruct setMask(rRegI dst, rRegI src) %{
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2056
  predicate(Matcher::has_predicated_vectors());
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2057
  match(Set dst (SetVectMaskI  src));
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2058
  effect(TEMP dst);
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2059
  format %{ "setvectmask   $dst, $src" %}
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2060
  ins_encode %{
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2061
    __ setvectmask($dst$$Register, $src$$Register);
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2062
  %}
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2063
  ins_pipe(pipe_slow);
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2064
%}
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2065
11794
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  2066
// ============================================================================
72249bf6ab83 7145346: VerifyStackAtCalls is broken
kvn
parents: 11429
diff changeset
  2067
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2068
instruct addF_reg(regF dst, regF src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2069
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2070
  match(Set dst (AddF dst src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2071
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2072
  format %{ "addss   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2073
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2074
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2075
    __ addss($dst$$XMMRegister, $src$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2076
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2077
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2078
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2079
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2080
instruct addF_mem(regF dst, memory src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2081
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2082
  match(Set dst (AddF dst (LoadF src)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2083
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2084
  format %{ "addss   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2085
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2086
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2087
    __ addss($dst$$XMMRegister, $src$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2088
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2089
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2090
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2091
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2092
instruct addF_imm(regF dst, immF con) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2093
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2094
  match(Set dst (AddF dst con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2095
  format %{ "addss   $dst, [$constantaddress]\t# load from constant table: float=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2096
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2097
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2098
    __ addss($dst$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2099
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2100
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2101
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2102
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2103
instruct addF_reg_reg(regF dst, regF src1, regF src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2104
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2105
  match(Set dst (AddF src1 src2));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2106
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2107
  format %{ "vaddss  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2108
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2109
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2110
    __ vaddss($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2111
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2112
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2113
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2114
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2115
instruct addF_reg_mem(regF dst, regF src1, memory src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2116
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2117
  match(Set dst (AddF src1 (LoadF src2)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2118
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2119
  format %{ "vaddss  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2120
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2121
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2122
    __ vaddss($dst$$XMMRegister, $src1$$XMMRegister, $src2$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2123
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2124
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2125
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2126
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2127
instruct addF_reg_imm(regF dst, regF src, immF con) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2128
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2129
  match(Set dst (AddF src con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2130
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2131
  format %{ "vaddss  $dst, $src, [$constantaddress]\t# load from constant table: float=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2132
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2133
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2134
    __ vaddss($dst$$XMMRegister, $src$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2135
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2136
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2137
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2138
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2139
instruct addD_reg(regD dst, regD src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2140
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2141
  match(Set dst (AddD dst src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2142
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2143
  format %{ "addsd   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2144
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2145
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2146
    __ addsd($dst$$XMMRegister, $src$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2147
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2148
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2149
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2150
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2151
instruct addD_mem(regD dst, memory src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2152
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2153
  match(Set dst (AddD dst (LoadD src)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2154
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2155
  format %{ "addsd   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2156
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2157
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2158
    __ addsd($dst$$XMMRegister, $src$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2159
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2160
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2161
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2162
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2163
instruct addD_imm(regD dst, immD con) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2164
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2165
  match(Set dst (AddD dst con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2166
  format %{ "addsd   $dst, [$constantaddress]\t# load from constant table: double=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2167
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2168
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2169
    __ addsd($dst$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2170
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2171
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2172
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2173
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2174
instruct addD_reg_reg(regD dst, regD src1, regD src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2175
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2176
  match(Set dst (AddD src1 src2));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2177
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2178
  format %{ "vaddsd  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2179
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2180
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2181
    __ vaddsd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2182
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2183
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2184
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2185
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2186
instruct addD_reg_mem(regD dst, regD src1, memory src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2187
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2188
  match(Set dst (AddD src1 (LoadD src2)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2189
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2190
  format %{ "vaddsd  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2191
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2192
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2193
    __ vaddsd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2194
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2195
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2196
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2197
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2198
instruct addD_reg_imm(regD dst, regD src, immD con) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2199
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2200
  match(Set dst (AddD src con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2201
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2202
  format %{ "vaddsd  $dst, $src, [$constantaddress]\t# load from constant table: double=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2203
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2204
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2205
    __ vaddsd($dst$$XMMRegister, $src$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2206
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2207
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2208
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2209
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2210
instruct subF_reg(regF dst, regF src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2211
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2212
  match(Set dst (SubF dst src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2213
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2214
  format %{ "subss   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2215
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2216
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2217
    __ subss($dst$$XMMRegister, $src$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2218
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2219
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2220
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2221
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2222
instruct subF_mem(regF dst, memory src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2223
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2224
  match(Set dst (SubF dst (LoadF src)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2225
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2226
  format %{ "subss   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2227
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2228
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2229
    __ subss($dst$$XMMRegister, $src$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2230
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2231
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2232
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2233
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2234
instruct subF_imm(regF dst, immF con) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2235
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2236
  match(Set dst (SubF dst con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2237
  format %{ "subss   $dst, [$constantaddress]\t# load from constant table: float=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2238
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2239
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2240
    __ subss($dst$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2241
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2242
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2243
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2244
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2245
instruct subF_reg_reg(regF dst, regF src1, regF src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2246
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2247
  match(Set dst (SubF src1 src2));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2248
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2249
  format %{ "vsubss  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2250
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2251
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2252
    __ vsubss($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2253
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2254
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2255
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2256
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2257
instruct subF_reg_mem(regF dst, regF src1, memory src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2258
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2259
  match(Set dst (SubF src1 (LoadF src2)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2260
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2261
  format %{ "vsubss  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2262
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2263
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2264
    __ vsubss($dst$$XMMRegister, $src1$$XMMRegister, $src2$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2265
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2266
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2267
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2268
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2269
instruct subF_reg_imm(regF dst, regF src, immF con) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2270
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2271
  match(Set dst (SubF src con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2272
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2273
  format %{ "vsubss  $dst, $src, [$constantaddress]\t# load from constant table: float=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2274
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2275
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2276
    __ vsubss($dst$$XMMRegister, $src$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2277
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2278
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2279
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2280
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2281
instruct subD_reg(regD dst, regD src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2282
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2283
  match(Set dst (SubD dst src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2284
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2285
  format %{ "subsd   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2286
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2287
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2288
    __ subsd($dst$$XMMRegister, $src$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2289
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2290
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2291
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2292
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2293
instruct subD_mem(regD dst, memory src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2294
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2295
  match(Set dst (SubD dst (LoadD src)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2296
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2297
  format %{ "subsd   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2298
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2299
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2300
    __ subsd($dst$$XMMRegister, $src$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2301
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2302
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2303
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2304
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2305
instruct subD_imm(regD dst, immD con) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2306
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2307
  match(Set dst (SubD dst con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2308
  format %{ "subsd   $dst, [$constantaddress]\t# load from constant table: double=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2309
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2310
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2311
    __ subsd($dst$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2312
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2313
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2314
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2315
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2316
instruct subD_reg_reg(regD dst, regD src1, regD src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2317
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2318
  match(Set dst (SubD src1 src2));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2319
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2320
  format %{ "vsubsd  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2321
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2322
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2323
    __ vsubsd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2324
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2325
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2326
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2327
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2328
instruct subD_reg_mem(regD dst, regD src1, memory src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2329
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2330
  match(Set dst (SubD src1 (LoadD src2)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2331
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2332
  format %{ "vsubsd  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2333
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2334
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2335
    __ vsubsd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2336
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2337
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2338
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2339
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2340
instruct subD_reg_imm(regD dst, regD src, immD con) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2341
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2342
  match(Set dst (SubD src con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2343
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2344
  format %{ "vsubsd  $dst, $src, [$constantaddress]\t# load from constant table: double=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2345
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2346
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2347
    __ vsubsd($dst$$XMMRegister, $src$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2348
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2349
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2350
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2351
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2352
instruct mulF_reg(regF dst, regF src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2353
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2354
  match(Set dst (MulF dst src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2355
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2356
  format %{ "mulss   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2357
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2358
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2359
    __ mulss($dst$$XMMRegister, $src$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2360
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2361
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2362
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2363
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2364
instruct mulF_mem(regF dst, memory src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2365
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2366
  match(Set dst (MulF dst (LoadF src)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2367
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2368
  format %{ "mulss   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2369
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2370
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2371
    __ mulss($dst$$XMMRegister, $src$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2372
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2373
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2374
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2375
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2376
instruct mulF_imm(regF dst, immF con) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2377
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2378
  match(Set dst (MulF dst con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2379
  format %{ "mulss   $dst, [$constantaddress]\t# load from constant table: float=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2380
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2381
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2382
    __ mulss($dst$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2383
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2384
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2385
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2386
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2387
instruct mulF_reg_reg(regF dst, regF src1, regF src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2388
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2389
  match(Set dst (MulF src1 src2));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2390
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2391
  format %{ "vmulss  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2392
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2393
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2394
    __ vmulss($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2395
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2396
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2397
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2398
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2399
instruct mulF_reg_mem(regF dst, regF src1, memory src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2400
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2401
  match(Set dst (MulF src1 (LoadF src2)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2402
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2403
  format %{ "vmulss  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2404
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2405
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2406
    __ vmulss($dst$$XMMRegister, $src1$$XMMRegister, $src2$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2407
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2408
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2409
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2410
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2411
instruct mulF_reg_imm(regF dst, regF src, immF con) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2412
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2413
  match(Set dst (MulF src con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2414
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2415
  format %{ "vmulss  $dst, $src, [$constantaddress]\t# load from constant table: float=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2416
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2417
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2418
    __ vmulss($dst$$XMMRegister, $src$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2419
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2420
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2421
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2422
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2423
instruct mulD_reg(regD dst, regD src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2424
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2425
  match(Set dst (MulD dst src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2426
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2427
  format %{ "mulsd   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2428
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2429
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2430
    __ mulsd($dst$$XMMRegister, $src$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2431
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2432
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2433
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2434
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2435
instruct mulD_mem(regD dst, memory src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2436
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2437
  match(Set dst (MulD dst (LoadD src)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2438
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2439
  format %{ "mulsd   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2440
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2441
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2442
    __ mulsd($dst$$XMMRegister, $src$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2443
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2444
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2445
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2446
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2447
instruct mulD_imm(regD dst, immD con) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2448
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2449
  match(Set dst (MulD dst con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2450
  format %{ "mulsd   $dst, [$constantaddress]\t# load from constant table: double=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2451
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2452
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2453
    __ mulsd($dst$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2454
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2455
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2456
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2457
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2458
instruct mulD_reg_reg(regD dst, regD src1, regD src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2459
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2460
  match(Set dst (MulD src1 src2));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2461
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2462
  format %{ "vmulsd  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2463
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2464
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2465
    __ vmulsd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2466
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2467
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2468
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2469
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2470
instruct mulD_reg_mem(regD dst, regD src1, memory src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2471
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2472
  match(Set dst (MulD src1 (LoadD src2)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2473
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2474
  format %{ "vmulsd  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2475
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2476
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2477
    __ vmulsd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2478
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2479
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2480
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2481
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2482
instruct mulD_reg_imm(regD dst, regD src, immD con) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2483
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2484
  match(Set dst (MulD src con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2485
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2486
  format %{ "vmulsd  $dst, $src, [$constantaddress]\t# load from constant table: double=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2487
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2488
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2489
    __ vmulsd($dst$$XMMRegister, $src$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2490
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2491
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2492
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2493
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2494
instruct divF_reg(regF dst, regF src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2495
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2496
  match(Set dst (DivF dst src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2497
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2498
  format %{ "divss   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2499
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2500
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2501
    __ divss($dst$$XMMRegister, $src$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2502
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2503
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2504
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2505
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2506
instruct divF_mem(regF dst, memory src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2507
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2508
  match(Set dst (DivF dst (LoadF src)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2509
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2510
  format %{ "divss   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2511
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2512
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2513
    __ divss($dst$$XMMRegister, $src$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2514
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2515
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2516
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2517
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2518
instruct divF_imm(regF dst, immF con) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2519
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2520
  match(Set dst (DivF dst con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2521
  format %{ "divss   $dst, [$constantaddress]\t# load from constant table: float=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2522
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2523
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2524
    __ divss($dst$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2525
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2526
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2527
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2528
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2529
instruct divF_reg_reg(regF dst, regF src1, regF src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2530
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2531
  match(Set dst (DivF src1 src2));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2532
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2533
  format %{ "vdivss  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2534
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2535
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2536
    __ vdivss($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2537
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2538
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2539
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2540
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2541
instruct divF_reg_mem(regF dst, regF src1, memory src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2542
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2543
  match(Set dst (DivF src1 (LoadF src2)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2544
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2545
  format %{ "vdivss  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2546
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2547
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2548
    __ vdivss($dst$$XMMRegister, $src1$$XMMRegister, $src2$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2549
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2550
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2551
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2552
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2553
instruct divF_reg_imm(regF dst, regF src, immF con) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2554
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2555
  match(Set dst (DivF src con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2556
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2557
  format %{ "vdivss  $dst, $src, [$constantaddress]\t# load from constant table: float=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2558
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2559
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2560
    __ vdivss($dst$$XMMRegister, $src$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2561
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2562
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2563
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2564
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2565
instruct divD_reg(regD dst, regD src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2566
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2567
  match(Set dst (DivD dst src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2568
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2569
  format %{ "divsd   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2570
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2571
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2572
    __ divsd($dst$$XMMRegister, $src$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2573
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2574
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2575
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2576
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2577
instruct divD_mem(regD dst, memory src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2578
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2579
  match(Set dst (DivD dst (LoadD src)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2580
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2581
  format %{ "divsd   $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2582
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2583
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2584
    __ divsd($dst$$XMMRegister, $src$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2585
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2586
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2587
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2588
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2589
instruct divD_imm(regD dst, immD con) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2590
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2591
  match(Set dst (DivD dst con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2592
  format %{ "divsd   $dst, [$constantaddress]\t# load from constant table: double=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2593
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2594
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2595
    __ divsd($dst$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2596
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2597
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2598
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2599
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2600
instruct divD_reg_reg(regD dst, regD src1, regD src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2601
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2602
  match(Set dst (DivD src1 src2));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2603
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2604
  format %{ "vdivsd  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2605
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2606
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2607
    __ vdivsd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2608
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2609
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2610
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2611
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2612
instruct divD_reg_mem(regD dst, regD src1, memory src2) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2613
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2614
  match(Set dst (DivD src1 (LoadD src2)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2615
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2616
  format %{ "vdivsd  $dst, $src1, $src2" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2617
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2618
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2619
    __ vdivsd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2620
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2621
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2622
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2623
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  2624
instruct divD_reg_imm(regD dst, regD src, immD con) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2625
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2626
  match(Set dst (DivD src con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2627
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2628
  format %{ "vdivsd  $dst, $src, [$constantaddress]\t# load from constant table: double=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2629
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2630
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2631
    __ vdivsd($dst$$XMMRegister, $src$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2632
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2633
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2634
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2635
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2636
instruct absF_reg(regF dst) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2637
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2638
  match(Set dst (AbsF dst));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2639
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2640
  format %{ "andps   $dst, [0x7fffffff]\t# abs float by sign masking" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2641
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2642
    __ andps($dst$$XMMRegister, ExternalAddress(float_signmask()));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2643
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2644
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2645
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2646
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2647
instruct absF_reg_reg(vlRegF dst, vlRegF src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2648
  predicate(UseAVX > 0);
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2649
  match(Set dst (AbsF src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2650
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2651
  format %{ "vandps  $dst, $src, [0x7fffffff]\t# abs float by sign masking" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2652
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  2653
    int vector_len = 0;
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2654
    __ vandps($dst$$XMMRegister, $src$$XMMRegister,
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  2655
              ExternalAddress(float_signmask()), vector_len);
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2656
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2657
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2658
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2659
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2660
instruct absD_reg(regD dst) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2661
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2662
  match(Set dst (AbsD dst));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2663
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2664
  format %{ "andpd   $dst, [0x7fffffffffffffff]\t"
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2665
            "# abs double by sign masking" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2666
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2667
    __ andpd($dst$$XMMRegister, ExternalAddress(double_signmask()));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2668
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2669
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2670
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2671
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2672
instruct absD_reg_reg(vlRegD dst, vlRegD src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2673
  predicate(UseAVX > 0);
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2674
  match(Set dst (AbsD src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2675
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2676
  format %{ "vandpd  $dst, $src, [0x7fffffffffffffff]\t"
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2677
            "# abs double by sign masking" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2678
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  2679
    int vector_len = 0;
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2680
    __ vandpd($dst$$XMMRegister, $src$$XMMRegister,
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  2681
              ExternalAddress(double_signmask()), vector_len);
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2682
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2683
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2684
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2685
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2686
instruct negF_reg(regF dst) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2687
  predicate((UseSSE>=1) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2688
  match(Set dst (NegF dst));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2689
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2690
  format %{ "xorps   $dst, [0x80000000]\t# neg float by sign flipping" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2691
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2692
    __ xorps($dst$$XMMRegister, ExternalAddress(float_signflip()));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2693
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2694
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2695
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2696
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2697
instruct negF_reg_reg(vlRegF dst, vlRegF src) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2698
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2699
  match(Set dst (NegF src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2700
  ins_cost(150);
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  2701
  format %{ "vnegatess  $dst, $src, [0x80000000]\t# neg float by sign flipping" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  2702
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  2703
    __ vnegatess($dst$$XMMRegister, $src$$XMMRegister,
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  2704
                 ExternalAddress(float_signflip()));
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2705
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2706
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2707
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2708
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2709
instruct negD_reg(regD dst) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2710
  predicate((UseSSE>=2) && (UseAVX == 0));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2711
  match(Set dst (NegD dst));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2712
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2713
  format %{ "xorpd   $dst, [0x8000000000000000]\t"
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2714
            "# neg double by sign flipping" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2715
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2716
    __ xorpd($dst$$XMMRegister, ExternalAddress(double_signflip()));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2717
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2718
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2719
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2720
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2721
instruct negD_reg_reg(vlRegD dst, vlRegD src) %{
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2722
  predicate(UseAVX > 0);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2723
  match(Set dst (NegD src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2724
  ins_cost(150);
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2725
  format %{ "vnegatesd  $dst, $src, [0x8000000000000000]\t"
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2726
            "# neg double by sign flipping" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2727
  ins_encode %{
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  2728
    __ vnegatesd($dst$$XMMRegister, $src$$XMMRegister,
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  2729
                 ExternalAddress(double_signflip()));
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2730
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2731
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2732
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2733
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2734
instruct sqrtF_reg(regF dst, regF src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2735
  predicate(UseSSE>=1);
48089
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  2736
  match(Set dst (SqrtF src));
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2737
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2738
  format %{ "sqrtss  $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2739
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2740
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2741
    __ sqrtss($dst$$XMMRegister, $src$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2742
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2743
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2744
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2745
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2746
instruct sqrtF_mem(regF dst, memory src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2747
  predicate(UseSSE>=1);
48089
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  2748
  match(Set dst (SqrtF (LoadF src)));
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2749
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2750
  format %{ "sqrtss  $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2751
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2752
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2753
    __ sqrtss($dst$$XMMRegister, $src$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2754
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2755
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2756
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2757
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2758
instruct sqrtF_imm(regF dst, immF con) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2759
  predicate(UseSSE>=1);
48089
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  2760
  match(Set dst (SqrtF con));
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  2761
11429
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2762
  format %{ "sqrtss  $dst, [$constantaddress]\t# load from constant table: float=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2763
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2764
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2765
    __ sqrtss($dst$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2766
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2767
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2768
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2769
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2770
instruct sqrtD_reg(regD dst, regD src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2771
  predicate(UseSSE>=2);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2772
  match(Set dst (SqrtD src));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2773
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2774
  format %{ "sqrtsd  $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2775
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2776
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2777
    __ sqrtsd($dst$$XMMRegister, $src$$XMMRegister);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2778
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2779
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2780
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2781
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2782
instruct sqrtD_mem(regD dst, memory src) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2783
  predicate(UseSSE>=2);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2784
  match(Set dst (SqrtD (LoadD src)));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2785
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2786
  format %{ "sqrtsd  $dst, $src" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2787
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2788
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2789
    __ sqrtsd($dst$$XMMRegister, $src$$Address);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2790
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2791
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2792
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2793
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2794
instruct sqrtD_imm(regD dst, immD con) %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2795
  predicate(UseSSE>=2);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2796
  match(Set dst (SqrtD con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2797
  format %{ "sqrtsd  $dst, [$constantaddress]\t# load from constant table: double=$con" %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2798
  ins_cost(150);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2799
  ins_encode %{
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2800
    __ sqrtsd($dst$$XMMRegister, $constantaddress($con));
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2801
  %}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2802
  ins_pipe(pipe_slow);
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2803
%}
e894217a5d94 7121648: Use 3-operands SIMD instructions on x86 with AVX
kvn
parents:
diff changeset
  2804
38017
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2805
instruct onspinwait() %{
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2806
  match(OnSpinWait);
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2807
  ins_cost(200);
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2808
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2809
  format %{
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2810
    $$template
51996
84743156e780 8188764: Obsolete AssumeMP and then remove all support for non-MP builds
dholmes
parents: 51857
diff changeset
  2811
    $$emit$$"pause\t! membar_onspinwait"
38017
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2812
  %}
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2813
  ins_encode %{
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2814
    __ pause();
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2815
  %}
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2816
  ins_pipe(pipe_slow);
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2817
%}
55047d16f141 8147844: new method j.l.Runtime.onSpinWait() and the corresponding x86 hotspot instrinsic
ikrylov
parents: 36561
diff changeset
  2818
41323
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2819
// a * b + c
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2820
instruct fmaD_reg(regD a, regD b, regD c) %{
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2821
  predicate(UseFMA);
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2822
  match(Set c (FmaD  c (Binary a b)));
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2823
  format %{ "fmasd $a,$b,$c\t# $c = $a * $b + $c" %}
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2824
  ins_cost(150);
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2825
  ins_encode %{
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2826
    __ fmad($c$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $c$$XMMRegister);
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2827
  %}
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2828
  ins_pipe( pipe_slow );
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2829
%}
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2830
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2831
// a * b + c
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2832
instruct fmaF_reg(regF a, regF b, regF c) %{
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2833
  predicate(UseFMA);
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2834
  match(Set c (FmaF  c (Binary a b)));
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2835
  format %{ "fmass $a,$b,$c\t# $c = $a * $b + $c" %}
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2836
  ins_cost(150);
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2837
  ins_encode %{
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2838
    __ fmaf($c$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $c$$XMMRegister);
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2839
  %}
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2840
  ins_pipe( pipe_slow );
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2841
%}
ddd5600d4762 8154122: Intrinsify fused mac operations
vdeshpande
parents: 40040
diff changeset
  2842
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2843
// ====================VECTOR INSTRUCTIONS=====================================
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2844
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2845
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2846
// Load vectors (4 bytes long)
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2847
instruct loadV4(vecS dst, memory mem) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2848
  predicate(n->as_LoadVector()->memory_size() == 4);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2849
  match(Set dst (LoadVector mem));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2850
  ins_cost(125);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2851
  format %{ "movd    $dst,$mem\t! load vector (4 bytes)" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2852
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2853
    __ movdl($dst$$XMMRegister, $mem$$Address);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2854
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2855
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2856
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2857
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2858
// Load vectors (4 bytes long)
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2859
instruct MoveVecS2Leg(legVecS dst, vecS src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2860
  match(Set dst src);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2861
  format %{ "movss $dst,$src\t! load vector (4 bytes)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2862
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2863
    __ movflt($dst$$XMMRegister, $src$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2864
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2865
  ins_pipe( fpu_reg_reg );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2866
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2867
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2868
// Load vectors (4 bytes long)
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2869
instruct MoveLeg2VecS(vecS dst, legVecS src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2870
  match(Set dst src);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2871
  format %{ "movss $dst,$src\t! load vector (4 bytes)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2872
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2873
    __ movflt($dst$$XMMRegister, $src$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2874
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2875
  ins_pipe( fpu_reg_reg );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2876
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2877
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2878
// Load vectors (8 bytes long)
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2879
instruct loadV8(vecD dst, memory mem) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2880
  predicate(n->as_LoadVector()->memory_size() == 8);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2881
  match(Set dst (LoadVector mem));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2882
  ins_cost(125);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2883
  format %{ "movq    $dst,$mem\t! load vector (8 bytes)" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2884
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2885
    __ movq($dst$$XMMRegister, $mem$$Address);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2886
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2887
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2888
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2889
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2890
// Load vectors (8 bytes long)
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2891
instruct MoveVecD2Leg(legVecD dst, vecD src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2892
  match(Set dst src);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2893
  format %{ "movsd $dst,$src\t! load vector (8 bytes)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2894
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2895
    __ movdbl($dst$$XMMRegister, $src$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2896
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2897
  ins_pipe( fpu_reg_reg );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2898
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2899
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2900
// Load vectors (8 bytes long)
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2901
instruct MoveLeg2VecD(vecD dst, legVecD src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2902
  match(Set dst src);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2903
  format %{ "movsd $dst,$src\t! load vector (8 bytes)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2904
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2905
    __ movdbl($dst$$XMMRegister, $src$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2906
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2907
  ins_pipe( fpu_reg_reg );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2908
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2909
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2910
// Load vectors (16 bytes long)
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2911
instruct loadV16(vecX dst, memory mem) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2912
  predicate(n->as_LoadVector()->memory_size() == 16);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2913
  match(Set dst (LoadVector mem));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2914
  ins_cost(125);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2915
  format %{ "movdqu  $dst,$mem\t! load vector (16 bytes)" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2916
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2917
    __ movdqu($dst$$XMMRegister, $mem$$Address);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2918
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2919
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2920
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2921
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2922
// Load vectors (16 bytes long)
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2923
instruct MoveVecX2Leg(legVecX dst, vecX src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2924
  match(Set dst src);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2925
  format %{ "movdqu $dst,$src\t! load vector (16 bytes)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2926
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2927
    if (UseAVX < 2 || VM_Version::supports_avx512vl()) {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2928
      __ movdqu($dst$$XMMRegister, $src$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2929
    } else {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2930
      int vector_len = 2;
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2931
      __ evmovdquq($dst$$XMMRegister, $src$$XMMRegister, vector_len);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2932
    }
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2933
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2934
  ins_pipe( fpu_reg_reg );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2935
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2936
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2937
// Load vectors (16 bytes long)
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2938
instruct MoveLeg2VecX(vecX dst, legVecX src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2939
  match(Set dst src);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2940
  format %{ "movdqu $dst,$src\t! load vector (16 bytes)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2941
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2942
    if (UseAVX < 2 || VM_Version::supports_avx512vl()) {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2943
      __ movdqu($dst$$XMMRegister, $src$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2944
    } else {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2945
      int vector_len = 2;
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2946
      __ evmovdquq($dst$$XMMRegister, $src$$XMMRegister, vector_len);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2947
    }
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2948
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2949
  ins_pipe( fpu_reg_reg );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2950
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2951
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2952
// Load vectors (32 bytes long)
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2953
instruct loadV32(vecY dst, memory mem) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2954
  predicate(n->as_LoadVector()->memory_size() == 32);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2955
  match(Set dst (LoadVector mem));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2956
  ins_cost(125);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2957
  format %{ "vmovdqu $dst,$mem\t! load vector (32 bytes)" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2958
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2959
    __ vmovdqu($dst$$XMMRegister, $mem$$Address);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2960
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2961
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2962
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  2963
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2964
// Load vectors (32 bytes long)
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2965
instruct MoveVecY2Leg(legVecY dst, vecY src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2966
  match(Set dst src);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2967
  format %{ "vmovdqu $dst,$src\t! load vector (32 bytes)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2968
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2969
    if (UseAVX < 2 || VM_Version::supports_avx512vl()) {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2970
      __ vmovdqu($dst$$XMMRegister, $src$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2971
    } else {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2972
      int vector_len = 2;
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2973
      __ evmovdquq($dst$$XMMRegister, $src$$XMMRegister, vector_len);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2974
    }
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2975
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2976
  ins_pipe( fpu_reg_reg );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2977
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2978
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2979
// Load vectors (32 bytes long)
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2980
instruct MoveLeg2VecY(vecY dst, legVecY src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2981
  match(Set dst src);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2982
  format %{ "vmovdqu $dst,$src\t! load vector (32 bytes)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2983
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2984
    if (UseAVX < 2 || VM_Version::supports_avx512vl()) {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2985
      __ vmovdqu($dst$$XMMRegister, $src$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2986
    } else {
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2987
      int vector_len = 2;
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2988
      __ evmovdquq($dst$$XMMRegister, $src$$XMMRegister, vector_len);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2989
    }
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2990
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2991
  ins_pipe( fpu_reg_reg );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2992
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  2993
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  2994
// Load vectors (64 bytes long)
38049
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2995
instruct loadV64_dword(vecZ dst, memory mem) %{
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2996
  predicate(n->as_LoadVector()->memory_size() == 64 && n->as_LoadVector()->element_size() <= 4);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  2997
  match(Set dst (LoadVector mem));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  2998
  ins_cost(125);
38049
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  2999
  format %{ "vmovdqul $dst k0,$mem\t! load vector (64 bytes)" %}
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3000
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3001
    int vector_len = 2;
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3002
    __ evmovdqul($dst$$XMMRegister, $mem$$Address, vector_len);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3003
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3004
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3005
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3006
38049
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3007
// Load vectors (64 bytes long)
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3008
instruct loadV64_qword(vecZ dst, memory mem) %{
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3009
  predicate(n->as_LoadVector()->memory_size() == 64 && n->as_LoadVector()->element_size() > 4);
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3010
  match(Set dst (LoadVector mem));
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3011
  ins_cost(125);
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3012
  format %{ "vmovdquq $dst k0,$mem\t! load vector (64 bytes)" %}
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3013
  ins_encode %{
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3014
    int vector_len = 2;
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3015
    __ evmovdquq($dst$$XMMRegister, $mem$$Address, vector_len);
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3016
  %}
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3017
  ins_pipe( pipe_slow );
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3018
%}
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3019
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3020
instruct MoveVecZ2Leg(legVecZ dst, vecZ  src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3021
  match(Set dst src);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3022
  format %{ "vmovdquq $dst k0,$src\t! Move vector (64 bytes)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3023
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3024
    int vector_len = 2;
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3025
    __ evmovdquq($dst$$XMMRegister, $src$$XMMRegister, vector_len);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3026
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3027
  ins_pipe( fpu_reg_reg );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3028
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3029
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3030
instruct MoveLeg2VecZ(vecZ dst, legVecZ  src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3031
  match(Set dst src);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3032
  format %{ "vmovdquq $dst k0,$src\t! Move vector (64 bytes)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3033
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3034
    int vector_len = 2;
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3035
    __ evmovdquq($dst$$XMMRegister, $src$$XMMRegister, vector_len);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3036
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3037
  ins_pipe( fpu_reg_reg );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3038
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3039
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3040
// Store vectors
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3041
instruct storeV4(memory mem, vecS src) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3042
  predicate(n->as_StoreVector()->memory_size() == 4);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3043
  match(Set mem (StoreVector mem src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3044
  ins_cost(145);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3045
  format %{ "movd    $mem,$src\t! store vector (4 bytes)" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3046
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3047
    __ movdl($mem$$Address, $src$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3048
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3049
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3050
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3051
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3052
instruct storeV8(memory mem, vecD src) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3053
  predicate(n->as_StoreVector()->memory_size() == 8);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3054
  match(Set mem (StoreVector mem src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3055
  ins_cost(145);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3056
  format %{ "movq    $mem,$src\t! store vector (8 bytes)" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3057
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3058
    __ movq($mem$$Address, $src$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3059
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3060
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3061
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3062
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3063
instruct storeV16(memory mem, vecX src) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3064
  predicate(n->as_StoreVector()->memory_size() == 16);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3065
  match(Set mem (StoreVector mem src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3066
  ins_cost(145);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3067
  format %{ "movdqu  $mem,$src\t! store vector (16 bytes)" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3068
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3069
    __ movdqu($mem$$Address, $src$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3070
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3071
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3072
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3073
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3074
instruct storeV32(memory mem, vecY src) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3075
  predicate(n->as_StoreVector()->memory_size() == 32);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3076
  match(Set mem (StoreVector mem src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3077
  ins_cost(145);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3078
  format %{ "vmovdqu $mem,$src\t! store vector (32 bytes)" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3079
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3080
    __ vmovdqu($mem$$Address, $src$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3081
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3082
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3083
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3084
38049
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3085
instruct storeV64_dword(memory mem, vecZ src) %{
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3086
  predicate(n->as_StoreVector()->memory_size() == 64 && n->as_StoreVector()->element_size() <= 4);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3087
  match(Set mem (StoreVector mem src));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3088
  ins_cost(145);
38049
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3089
  format %{ "vmovdqul $mem k0,$src\t! store vector (64 bytes)" %}
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3090
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3091
    int vector_len = 2;
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3092
    __ evmovdqul($mem$$Address, $src$$XMMRegister, vector_len);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3093
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3094
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3095
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3096
38049
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3097
instruct storeV64_qword(memory mem, vecZ src) %{
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3098
  predicate(n->as_StoreVector()->memory_size() == 64 && n->as_StoreVector()->element_size() > 4);
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3099
  match(Set mem (StoreVector mem src));
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3100
  ins_cost(145);
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3101
  format %{ "vmovdquq $mem k0,$src\t! store vector (64 bytes)" %}
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3102
  ins_encode %{
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3103
    int vector_len = 2;
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3104
    __ evmovdquq($mem$$Address, $src$$XMMRegister, vector_len);
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3105
  %}
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3106
  ins_pipe( pipe_slow );
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3107
%}
e8541793960f 8153998: Masked vector post loops
mcberg
parents: 38017
diff changeset
  3108
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3109
// ====================LEGACY REPLICATE=======================================
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3110
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3111
instruct Repl4B_mem(vecS dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3112
  predicate(n->as_Vector()->length() == 4 && UseAVX > 0 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3113
  match(Set dst (ReplicateB (LoadB mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3114
  format %{ "punpcklbw $dst,$mem\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3115
            "pshuflw $dst,$dst,0x00\t! replicate4B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3116
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3117
    __ punpcklbw($dst$$XMMRegister, $mem$$Address);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3118
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3119
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3120
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3121
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3122
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3123
instruct Repl8B_mem(vecD dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3124
  predicate(n->as_Vector()->length() == 8 && UseAVX > 0 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3125
  match(Set dst (ReplicateB (LoadB mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3126
  format %{ "punpcklbw $dst,$mem\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3127
            "pshuflw $dst,$dst,0x00\t! replicate8B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3128
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3129
    __ punpcklbw($dst$$XMMRegister, $mem$$Address);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3130
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3131
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3132
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3133
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3134
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3135
instruct Repl16B(vecX dst, rRegI src) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3136
  predicate(n->as_Vector()->length() == 16 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3137
  match(Set dst (ReplicateB src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3138
  format %{ "movd    $dst,$src\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3139
            "punpcklbw $dst,$dst\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3140
            "pshuflw $dst,$dst,0x00\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3141
            "punpcklqdq $dst,$dst\t! replicate16B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3142
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3143
    __ movdl($dst$$XMMRegister, $src$$Register);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3144
    __ punpcklbw($dst$$XMMRegister, $dst$$XMMRegister);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3145
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3146
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3147
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3148
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3149
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3150
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3151
instruct Repl16B_mem(vecX dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3152
  predicate(n->as_Vector()->length() == 16 && UseAVX > 0 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3153
  match(Set dst (ReplicateB (LoadB mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3154
  format %{ "punpcklbw $dst,$mem\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3155
            "pshuflw $dst,$dst,0x00\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3156
            "punpcklqdq $dst,$dst\t! replicate16B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3157
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3158
    __ punpcklbw($dst$$XMMRegister, $mem$$Address);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3159
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3160
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3161
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3162
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3163
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3164
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3165
instruct Repl32B(vecY dst, rRegI src) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3166
  predicate(n->as_Vector()->length() == 32 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3167
  match(Set dst (ReplicateB src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3168
  format %{ "movd    $dst,$src\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3169
            "punpcklbw $dst,$dst\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3170
            "pshuflw $dst,$dst,0x00\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3171
            "punpcklqdq $dst,$dst\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3172
            "vinserti128_high $dst,$dst\t! replicate32B" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3173
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3174
    __ movdl($dst$$XMMRegister, $src$$Register);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3175
    __ punpcklbw($dst$$XMMRegister, $dst$$XMMRegister);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3176
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3177
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3178
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3179
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3180
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3181
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3182
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3183
instruct Repl32B_mem(vecY dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3184
  predicate(n->as_Vector()->length() == 32 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3185
  match(Set dst (ReplicateB (LoadB mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3186
  format %{ "punpcklbw $dst,$mem\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3187
            "pshuflw $dst,$dst,0x00\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3188
            "punpcklqdq $dst,$dst\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3189
            "vinserti128_high $dst,$dst\t! replicate32B" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3190
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3191
    __ punpcklbw($dst$$XMMRegister, $mem$$Address);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3192
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3193
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3194
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3195
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3196
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3197
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3198
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3199
instruct Repl64B(legVecZ dst, rRegI src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3200
  predicate(n->as_Vector()->length() == 64 && !VM_Version::supports_avx512vlbw());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3201
  match(Set dst (ReplicateB src));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3202
  format %{ "movd    $dst,$src\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3203
            "punpcklbw $dst,$dst\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3204
            "pshuflw $dst,$dst,0x00\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3205
            "punpcklqdq $dst,$dst\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3206
            "vinserti128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3207
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate64B" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3208
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3209
    __ movdl($dst$$XMMRegister, $src$$Register);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3210
    __ punpcklbw($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3211
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3212
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3213
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3214
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3215
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3216
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3217
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3218
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3219
instruct Repl64B_mem(legVecZ dst, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3220
  predicate(n->as_Vector()->length() == 64 && !VM_Version::supports_avx512vlbw());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3221
  match(Set dst (ReplicateB (LoadB mem)));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3222
  format %{ "punpcklbw $dst,$mem\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3223
            "pshuflw $dst,$dst,0x00\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3224
            "punpcklqdq $dst,$dst\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3225
            "vinserti128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3226
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate64B" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3227
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3228
    __ punpcklbw($dst$$XMMRegister, $mem$$Address);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3229
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3230
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3231
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3232
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3233
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3234
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3235
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3236
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3237
instruct Repl16B_imm(vecX dst, immI con) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3238
  predicate(n->as_Vector()->length() == 16 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3239
  match(Set dst (ReplicateB con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3240
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3241
            "punpcklqdq $dst,$dst\t! replicate16B($con)" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3242
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3243
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 1)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3244
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3245
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3246
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3247
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3248
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3249
instruct Repl32B_imm(vecY dst, immI con) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3250
  predicate(n->as_Vector()->length() == 32 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3251
  match(Set dst (ReplicateB con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3252
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3253
            "punpcklqdq $dst,$dst\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3254
            "vinserti128_high $dst,$dst\t! lreplicate32B($con)" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3255
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3256
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 1)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3257
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3258
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3259
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3260
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3261
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3262
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3263
instruct Repl64B_imm(legVecZ dst, immI con) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3264
  predicate(n->as_Vector()->length() == 64 && !VM_Version::supports_avx512vlbw());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3265
  match(Set dst (ReplicateB con));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3266
  format %{ "movq    $dst,[$constantaddress]\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3267
            "punpcklqdq $dst,$dst\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3268
            "vinserti128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3269
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate64B($con)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3270
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3271
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 1)));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3272
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3273
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3274
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3275
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3276
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3277
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3278
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3279
instruct Repl4S(vecD dst, rRegI src) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3280
  predicate(n->as_Vector()->length() == 4 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3281
  match(Set dst (ReplicateS src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3282
  format %{ "movd    $dst,$src\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3283
            "pshuflw $dst,$dst,0x00\t! replicate4S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3284
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3285
    __ movdl($dst$$XMMRegister, $src$$Register);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3286
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3287
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3288
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3289
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3290
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3291
instruct Repl4S_mem(vecD dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3292
  predicate(n->as_Vector()->length() == 4 && UseAVX > 0 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3293
  match(Set dst (ReplicateS (LoadS mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3294
  format %{ "pshuflw $dst,$mem,0x00\t! replicate4S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3295
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3296
    __ pshuflw($dst$$XMMRegister, $mem$$Address, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3297
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3298
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3299
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3300
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3301
instruct Repl8S(vecX dst, rRegI src) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3302
  predicate(n->as_Vector()->length() == 8 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3303
  match(Set dst (ReplicateS src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3304
  format %{ "movd    $dst,$src\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3305
            "pshuflw $dst,$dst,0x00\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3306
            "punpcklqdq $dst,$dst\t! replicate8S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3307
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3308
    __ movdl($dst$$XMMRegister, $src$$Register);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3309
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3310
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3311
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3312
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3313
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3314
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3315
instruct Repl8S_mem(vecX dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3316
  predicate(n->as_Vector()->length() == 8 && UseAVX > 0 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3317
  match(Set dst (ReplicateS (LoadS mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3318
  format %{ "pshuflw $dst,$mem,0x00\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3319
            "punpcklqdq $dst,$dst\t! replicate8S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3320
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3321
    __ pshuflw($dst$$XMMRegister, $mem$$Address, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3322
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3323
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3324
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3325
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3326
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3327
instruct Repl8S_imm(vecX dst, immI con) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3328
  predicate(n->as_Vector()->length() == 8 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3329
  match(Set dst (ReplicateS con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3330
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3331
            "punpcklqdq $dst,$dst\t! replicate8S($con)" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3332
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3333
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 2)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3334
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3335
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3336
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3337
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3338
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3339
instruct Repl16S(vecY dst, rRegI src) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3340
  predicate(n->as_Vector()->length() == 16 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3341
  match(Set dst (ReplicateS src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3342
  format %{ "movd    $dst,$src\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3343
            "pshuflw $dst,$dst,0x00\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3344
            "punpcklqdq $dst,$dst\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3345
            "vinserti128_high $dst,$dst\t! replicate16S" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3346
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3347
    __ movdl($dst$$XMMRegister, $src$$Register);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3348
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3349
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3350
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3351
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3352
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3353
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3354
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3355
instruct Repl16S_mem(vecY dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3356
  predicate(n->as_Vector()->length() == 16 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3357
  match(Set dst (ReplicateS (LoadS mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3358
  format %{ "pshuflw $dst,$mem,0x00\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3359
            "punpcklqdq $dst,$dst\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3360
            "vinserti128_high $dst,$dst\t! replicate16S" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3361
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3362
    __ pshuflw($dst$$XMMRegister, $mem$$Address, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3363
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3364
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3365
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3366
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3367
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3368
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3369
instruct Repl16S_imm(vecY dst, immI con) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3370
  predicate(n->as_Vector()->length() == 16 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3371
  match(Set dst (ReplicateS con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3372
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3373
            "punpcklqdq $dst,$dst\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3374
            "vinserti128_high $dst,$dst\t! replicate16S($con)" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3375
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3376
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 2)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3377
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3378
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3379
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3380
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3381
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3382
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3383
instruct Repl32S(legVecZ dst, rRegI src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3384
  predicate(n->as_Vector()->length() == 32 && !VM_Version::supports_avx512vlbw());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3385
  match(Set dst (ReplicateS src));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3386
  format %{ "movd    $dst,$src\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3387
            "pshuflw $dst,$dst,0x00\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3388
            "punpcklqdq $dst,$dst\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3389
            "vinserti128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3390
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate32S" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3391
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3392
    __ movdl($dst$$XMMRegister, $src$$Register);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3393
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3394
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3395
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3396
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3397
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3398
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3399
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3400
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3401
instruct Repl32S_mem(legVecZ dst, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3402
  predicate(n->as_Vector()->length() == 32 && !VM_Version::supports_avx512vlbw());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3403
  match(Set dst (ReplicateS (LoadS mem)));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3404
  format %{ "pshuflw $dst,$mem,0x00\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3405
            "punpcklqdq $dst,$dst\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3406
            "vinserti128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3407
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate32S" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3408
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3409
    __ pshuflw($dst$$XMMRegister, $mem$$Address, 0x00);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3410
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3411
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3412
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3413
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3414
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3415
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3416
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3417
instruct Repl32S_imm(legVecZ dst, immI con) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3418
  predicate(n->as_Vector()->length() == 32 && !VM_Version::supports_avx512vlbw());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3419
  match(Set dst (ReplicateS con));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3420
  format %{ "movq    $dst,[$constantaddress]\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3421
            "punpcklqdq $dst,$dst\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3422
            "vinserti128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3423
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate32S($con)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3424
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3425
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 2)));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3426
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3427
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3428
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3429
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3430
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3431
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3432
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3433
instruct Repl4I(vecX dst, rRegI src) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3434
  predicate(n->as_Vector()->length() == 4 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3435
  match(Set dst (ReplicateI src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3436
  format %{ "movd    $dst,$src\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3437
            "pshufd  $dst,$dst,0x00\t! replicate4I" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3438
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3439
    __ movdl($dst$$XMMRegister, $src$$Register);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3440
    __ pshufd($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3441
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3442
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3443
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3444
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3445
instruct Repl4I_mem(vecX dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3446
  predicate(n->as_Vector()->length() == 4 && UseAVX > 0 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3447
  match(Set dst (ReplicateI (LoadI mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3448
  format %{ "pshufd  $dst,$mem,0x00\t! replicate4I" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3449
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3450
    __ pshufd($dst$$XMMRegister, $mem$$Address, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3451
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3452
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3453
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3454
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3455
instruct Repl8I(vecY dst, rRegI src) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3456
  predicate(n->as_Vector()->length() == 8 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3457
  match(Set dst (ReplicateI src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3458
  format %{ "movd    $dst,$src\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3459
            "pshufd  $dst,$dst,0x00\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3460
            "vinserti128_high $dst,$dst\t! replicate8I" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3461
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3462
    __ movdl($dst$$XMMRegister, $src$$Register);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3463
    __ pshufd($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3464
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3465
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3466
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3467
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3468
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3469
instruct Repl8I_mem(vecY dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3470
  predicate(n->as_Vector()->length() == 8 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3471
  match(Set dst (ReplicateI (LoadI mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3472
  format %{ "pshufd  $dst,$mem,0x00\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3473
            "vinserti128_high $dst,$dst\t! replicate8I" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3474
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3475
    __ pshufd($dst$$XMMRegister, $mem$$Address, 0x00);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3476
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3477
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3478
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3479
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3480
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3481
instruct Repl16I(legVecZ dst, rRegI src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3482
  predicate(n->as_Vector()->length() == 16 && !VM_Version::supports_avx512vl());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3483
  match(Set dst (ReplicateI src));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3484
  format %{ "movd    $dst,$src\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3485
            "pshufd  $dst,$dst,0x00\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3486
            "vinserti128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3487
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate16I" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3488
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3489
    __ movdl($dst$$XMMRegister, $src$$Register);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3490
    __ pshufd($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3491
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3492
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3493
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3494
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3495
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3496
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3497
instruct Repl16I_mem(legVecZ dst, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3498
  predicate(n->as_Vector()->length() == 16 && !VM_Version::supports_avx512vl());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3499
  match(Set dst (ReplicateI (LoadI mem)));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3500
  format %{ "pshufd  $dst,$mem,0x00\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3501
            "vinserti128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3502
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate16I" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3503
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3504
    __ pshufd($dst$$XMMRegister, $mem$$Address, 0x00);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3505
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3506
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3507
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3508
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3509
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3510
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3511
instruct Repl4I_imm(vecX dst, immI con) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3512
  predicate(n->as_Vector()->length() == 4 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3513
  match(Set dst (ReplicateI con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3514
  format %{ "movq    $dst,[$constantaddress]\t! replicate4I($con)\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3515
            "punpcklqdq $dst,$dst" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3516
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3517
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 4)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3518
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3519
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3520
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3521
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3522
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3523
instruct Repl8I_imm(vecY dst, immI con) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3524
  predicate(n->as_Vector()->length() == 8 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3525
  match(Set dst (ReplicateI con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3526
  format %{ "movq    $dst,[$constantaddress]\t! replicate8I($con)\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3527
            "punpcklqdq $dst,$dst\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3528
            "vinserti128_high $dst,$dst" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3529
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3530
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 4)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3531
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3532
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3533
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3534
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3535
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3536
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3537
instruct Repl16I_imm(legVecZ dst, immI con) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3538
  predicate(n->as_Vector()->length() == 16 && !VM_Version::supports_avx512vl());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3539
  match(Set dst (ReplicateI con));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3540
  format %{ "movq    $dst,[$constantaddress]\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3541
            "punpcklqdq $dst,$dst\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3542
            "vinserti128_high $dst,$dst"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3543
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate16I($con)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3544
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3545
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 4)));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3546
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3547
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3548
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3549
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3550
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3551
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3552
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3553
// Long could be loaded into xmm register directly from memory.
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3554
instruct Repl2L_mem(vecX dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3555
  predicate(n->as_Vector()->length() == 2 && !VM_Version::supports_avx512vlbw());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3556
  match(Set dst (ReplicateL (LoadL mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3557
  format %{ "movq    $dst,$mem\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3558
            "punpcklqdq $dst,$dst\t! replicate2L" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3559
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3560
    __ movq($dst$$XMMRegister, $mem$$Address);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3561
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3562
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3563
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3564
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3565
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3566
// Replicate long (8 byte) scalar to be vector
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3567
#ifdef _LP64
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3568
instruct Repl4L(vecY dst, rRegL src) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3569
  predicate(n->as_Vector()->length() == 4 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3570
  match(Set dst (ReplicateL src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3571
  format %{ "movdq   $dst,$src\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3572
            "punpcklqdq $dst,$dst\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3573
            "vinserti128_high $dst,$dst\t! replicate4L" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3574
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3575
    __ movdq($dst$$XMMRegister, $src$$Register);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3576
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3577
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3578
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3579
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3580
%}
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3581
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3582
instruct Repl8L(legVecZ dst, rRegL src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3583
  predicate(n->as_Vector()->length() == 8 && !VM_Version::supports_avx512vl());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3584
  match(Set dst (ReplicateL src));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3585
  format %{ "movdq   $dst,$src\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3586
            "punpcklqdq $dst,$dst\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3587
            "vinserti128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3588
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate8L" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3589
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3590
    __ movdq($dst$$XMMRegister, $src$$Register);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3591
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3592
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3593
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3594
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3595
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3596
%}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3597
#else // _LP64
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3598
instruct Repl4L(vecY dst, eRegL src, vecY tmp) %{
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3599
  predicate(n->as_Vector()->length() == 4 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3600
  match(Set dst (ReplicateL src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3601
  effect(TEMP dst, USE src, TEMP tmp);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3602
  format %{ "movdl   $dst,$src.lo\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3603
            "movdl   $tmp,$src.hi\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3604
            "punpckldq $dst,$tmp\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3605
            "punpcklqdq $dst,$dst\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3606
            "vinserti128_high $dst,$dst\t! replicate4L" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3607
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3608
    __ movdl($dst$$XMMRegister, $src$$Register);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3609
    __ movdl($tmp$$XMMRegister, HIGH_FROM_LOW($src$$Register));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3610
    __ punpckldq($dst$$XMMRegister, $tmp$$XMMRegister);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3611
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3612
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3613
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3614
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3615
%}
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3616
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3617
instruct Repl8L(legVecZ dst, eRegL src, legVecZ tmp) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3618
  predicate(n->as_Vector()->length() == 8 && !VM_Version::supports_avx512vl());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3619
  match(Set dst (ReplicateL src));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3620
  effect(TEMP dst, USE src, TEMP tmp);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3621
  format %{ "movdl   $dst,$src.lo\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3622
            "movdl   $tmp,$src.hi\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3623
            "punpckldq $dst,$tmp\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3624
            "punpcklqdq $dst,$dst\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3625
            "vinserti128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3626
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate8L" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3627
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3628
    __ movdl($dst$$XMMRegister, $src$$Register);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3629
    __ movdl($tmp$$XMMRegister, HIGH_FROM_LOW($src$$Register));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3630
    __ punpckldq($dst$$XMMRegister, $tmp$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3631
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3632
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3633
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3634
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3635
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3636
%}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3637
#endif // _LP64
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3638
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3639
instruct Repl4L_imm(vecY dst, immL con) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3640
  predicate(n->as_Vector()->length() == 4 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3641
  match(Set dst (ReplicateL con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3642
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3643
            "punpcklqdq $dst,$dst\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3644
            "vinserti128_high $dst,$dst\t! replicate4L($con)" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3645
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3646
    __ movq($dst$$XMMRegister, $constantaddress($con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3647
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3648
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3649
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3650
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3651
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3652
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3653
instruct Repl8L_imm(legVecZ dst, immL con) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3654
  predicate(n->as_Vector()->length() == 8 && !VM_Version::supports_avx512vl());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3655
  match(Set dst (ReplicateL con));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3656
  format %{ "movq    $dst,[$constantaddress]\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3657
            "punpcklqdq $dst,$dst\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3658
            "vinserti128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3659
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate8L($con)" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3660
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3661
    __ movq($dst$$XMMRegister, $constantaddress($con));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3662
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3663
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3664
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3665
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3666
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3667
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3668
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3669
instruct Repl4L_mem(vecY dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3670
  predicate(n->as_Vector()->length() == 4 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3671
  match(Set dst (ReplicateL (LoadL mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3672
  format %{ "movq    $dst,$mem\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3673
            "punpcklqdq $dst,$dst\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3674
            "vinserti128_high $dst,$dst\t! replicate4L" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3675
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3676
    __ movq($dst$$XMMRegister, $mem$$Address);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3677
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3678
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3679
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3680
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3681
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3682
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3683
instruct Repl8L_mem(legVecZ dst, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3684
  predicate(n->as_Vector()->length() == 8 && !VM_Version::supports_avx512vl());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3685
  match(Set dst (ReplicateL (LoadL mem)));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3686
  format %{ "movq    $dst,$mem\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3687
            "punpcklqdq $dst,$dst\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3688
            "vinserti128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3689
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate8L" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3690
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3691
    __ movq($dst$$XMMRegister, $mem$$Address);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3692
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3693
    __ vinserti128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3694
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3695
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3696
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3697
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3698
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3699
instruct Repl2F_mem(vecD dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3700
  predicate(n->as_Vector()->length() == 2 && UseAVX > 0 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3701
  match(Set dst (ReplicateF (LoadF mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3702
  format %{ "pshufd  $dst,$mem,0x00\t! replicate2F" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3703
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3704
    __ pshufd($dst$$XMMRegister, $mem$$Address, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3705
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3706
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3707
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3708
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3709
instruct Repl4F_mem(vecX dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3710
  predicate(n->as_Vector()->length() == 4 && UseAVX > 0 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3711
  match(Set dst (ReplicateF (LoadF mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3712
  format %{ "pshufd  $dst,$mem,0x00\t! replicate4F" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3713
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3714
    __ pshufd($dst$$XMMRegister, $mem$$Address, 0x00);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3715
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3716
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3717
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3718
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3719
instruct Repl8F(vecY dst, vlRegF src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3720
  predicate(n->as_Vector()->length() == 8 && UseAVX > 0 && !VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3721
  match(Set dst (ReplicateF src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3722
  format %{ "pshufd  $dst,$src,0x00\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3723
            "vinsertf128_high $dst,$dst\t! replicate8F" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3724
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3725
    __ pshufd($dst$$XMMRegister, $src$$XMMRegister, 0x00);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3726
    __ vinsertf128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3727
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3728
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3729
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3730
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3731
instruct Repl8F_mem(vecY dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3732
  predicate(n->as_Vector()->length() == 8 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3733
  match(Set dst (ReplicateF (LoadF mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3734
  format %{ "pshufd  $dst,$mem,0x00\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3735
            "vinsertf128_high $dst,$dst\t! replicate8F" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3736
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3737
    __ pshufd($dst$$XMMRegister, $mem$$Address, 0x00);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3738
    __ vinsertf128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3739
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3740
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3741
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3742
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3743
instruct Repl16F(legVecZ dst, vlRegF src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3744
  predicate(n->as_Vector()->length() == 16 && UseAVX > 0 && !VM_Version::supports_avx512vl());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3745
  match(Set dst (ReplicateF src));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3746
  format %{ "pshufd  $dst,$src,0x00\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3747
            "vinsertf128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3748
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate16F" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3749
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3750
    __ pshufd($dst$$XMMRegister, $src$$XMMRegister, 0x00);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3751
    __ vinsertf128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3752
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3753
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3754
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3755
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3756
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3757
instruct Repl16F_mem(legVecZ dst, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3758
  predicate(n->as_Vector()->length() == 16 && !VM_Version::supports_avx512vl());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3759
  match(Set dst (ReplicateF (LoadF mem)));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3760
  format %{ "pshufd  $dst,$mem,0x00\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3761
            "vinsertf128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3762
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate16F" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3763
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3764
    __ pshufd($dst$$XMMRegister, $mem$$Address, 0x00);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3765
    __ vinsertf128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3766
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3767
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3768
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3769
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3770
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3771
instruct Repl2F_zero(vecD dst, immF0 zero) %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3772
  predicate(n->as_Vector()->length() == 2 && UseAVX < 3);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3773
  match(Set dst (ReplicateF zero));
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3774
  format %{ "xorps   $dst,$dst\t! replicate2F zero" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3775
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3776
    __ xorps($dst$$XMMRegister, $dst$$XMMRegister);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3777
  %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3778
  ins_pipe( fpu_reg_reg );
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3779
%}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3780
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3781
instruct Repl4F_zero(vecX dst, immF0 zero) %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3782
  predicate(n->as_Vector()->length() == 4 && UseAVX < 3);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3783
  match(Set dst (ReplicateF zero));
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3784
  format %{ "xorps   $dst,$dst\t! replicate4F zero" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3785
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3786
    __ xorps($dst$$XMMRegister, $dst$$XMMRegister);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3787
  %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3788
  ins_pipe( fpu_reg_reg );
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3789
%}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3790
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3791
instruct Repl8F_zero(vecY dst, immF0 zero) %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3792
  predicate(n->as_Vector()->length() == 8 && UseAVX < 3);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3793
  match(Set dst (ReplicateF zero));
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3794
  format %{ "vxorps  $dst,$dst,$dst\t! replicate8F zero" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3795
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3796
    int vector_len = 1;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3797
    __ vxorps($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, vector_len);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3798
  %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3799
  ins_pipe( fpu_reg_reg );
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3800
%}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3801
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3802
instruct Repl2D_mem(vecX dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3803
  predicate(n->as_Vector()->length() == 2 && UseAVX > 0 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3804
  match(Set dst (ReplicateD (LoadD mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3805
  format %{ "pshufd  $dst,$mem,0x44\t! replicate2D" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3806
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3807
    __ pshufd($dst$$XMMRegister, $mem$$Address, 0x44);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3808
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3809
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3810
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3811
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3812
instruct Repl4D(vecY dst, vlRegD src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3813
  predicate(n->as_Vector()->length() == 4 && UseAVX > 0 && !VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3814
  match(Set dst (ReplicateD src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3815
  format %{ "pshufd  $dst,$src,0x44\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3816
            "vinsertf128_high $dst,$dst\t! replicate4D" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3817
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3818
    __ pshufd($dst$$XMMRegister, $src$$XMMRegister, 0x44);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3819
    __ vinsertf128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3820
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3821
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3822
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3823
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3824
instruct Repl4D_mem(vecY dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3825
  predicate(n->as_Vector()->length() == 4 && !VM_Version::supports_avx512vl());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3826
  match(Set dst (ReplicateD (LoadD mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3827
  format %{ "pshufd  $dst,$mem,0x44\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3828
            "vinsertf128_high $dst,$dst\t! replicate4D" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3829
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3830
    __ pshufd($dst$$XMMRegister, $mem$$Address, 0x44);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  3831
    __ vinsertf128_high($dst$$XMMRegister, $dst$$XMMRegister);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3832
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3833
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3834
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3835
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3836
instruct Repl8D(legVecZ dst, vlRegD src) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3837
  predicate(n->as_Vector()->length() == 8 && UseAVX > 0 && !VM_Version::supports_avx512vl());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3838
  match(Set dst (ReplicateD src));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3839
  format %{ "pshufd  $dst,$src,0x44\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3840
            "vinsertf128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3841
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate8D" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3842
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3843
    __ pshufd($dst$$XMMRegister, $src$$XMMRegister, 0x44);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3844
    __ vinsertf128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3845
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3846
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3847
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3848
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3849
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3850
instruct Repl8D_mem(legVecZ dst, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3851
  predicate(n->as_Vector()->length() == 8 && !VM_Version::supports_avx512vl());
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3852
  match(Set dst (ReplicateD (LoadD mem)));
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3853
  format %{ "pshufd  $dst,$mem,0x44\n\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3854
            "vinsertf128_high $dst,$dst\t"
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3855
            "vinserti64x4 $dst,$dst,$dst,0x1\t! replicate8D" %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3856
  ins_encode %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3857
    __ pshufd($dst$$XMMRegister, $mem$$Address, 0x44);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3858
    __ vinsertf128_high($dst$$XMMRegister, $dst$$XMMRegister);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3859
    __ vinserti64x4($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, 0x1);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3860
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3861
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3862
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  3863
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3864
// Replicate double (8 byte) scalar zero to be vector
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3865
instruct Repl2D_zero(vecX dst, immD0 zero) %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3866
  predicate(n->as_Vector()->length() == 2 && UseAVX < 3);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3867
  match(Set dst (ReplicateD zero));
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3868
  format %{ "xorpd   $dst,$dst\t! replicate2D zero" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3869
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3870
    __ xorpd($dst$$XMMRegister, $dst$$XMMRegister);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3871
  %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3872
  ins_pipe( fpu_reg_reg );
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3873
%}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3874
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3875
instruct Repl4D_zero(vecY dst, immD0 zero) %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3876
  predicate(n->as_Vector()->length() == 4 && UseAVX < 3);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3877
  match(Set dst (ReplicateD zero));
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3878
  format %{ "vxorpd  $dst,$dst,$dst,vect256\t! replicate4D zero" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3879
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3880
    int vector_len = 1;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3881
    __ vxorpd($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, vector_len);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3882
  %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3883
  ins_pipe( fpu_reg_reg );
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3884
%}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  3885
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3886
// ====================GENERIC REPLICATE==========================================
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  3887
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3888
// Replicate byte scalar to be vector
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3889
instruct Repl4B(vecS dst, rRegI src) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3890
  predicate(n->as_Vector()->length() == 4);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3891
  match(Set dst (ReplicateB src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3892
  format %{ "movd    $dst,$src\n\t"
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3893
            "punpcklbw $dst,$dst\n\t"
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3894
            "pshuflw $dst,$dst,0x00\t! replicate4B" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3895
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3896
    __ movdl($dst$$XMMRegister, $src$$Register);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3897
    __ punpcklbw($dst$$XMMRegister, $dst$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3898
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3899
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3900
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3901
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3902
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3903
instruct Repl8B(vecD dst, rRegI src) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3904
  predicate(n->as_Vector()->length() == 8);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3905
  match(Set dst (ReplicateB src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3906
  format %{ "movd    $dst,$src\n\t"
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3907
            "punpcklbw $dst,$dst\n\t"
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3908
            "pshuflw $dst,$dst,0x00\t! replicate8B" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3909
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3910
    __ movdl($dst$$XMMRegister, $src$$Register);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3911
    __ punpcklbw($dst$$XMMRegister, $dst$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3912
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3913
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3914
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3915
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3916
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3917
// Replicate byte scalar immediate to be vector by loading from const table.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3918
instruct Repl4B_imm(vecS dst, immI con) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3919
  predicate(n->as_Vector()->length() == 4);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3920
  match(Set dst (ReplicateB con));
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  3921
  format %{ "movdl   $dst,[$constantaddress]\t! replicate4B($con)" %}
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3922
  ins_encode %{
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  3923
    __ movdl($dst$$XMMRegister, $constantaddress(replicate4_imm($con$$constant, 1)));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3924
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3925
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3926
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3927
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3928
instruct Repl8B_imm(vecD dst, immI con) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3929
  predicate(n->as_Vector()->length() == 8);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3930
  match(Set dst (ReplicateB con));
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  3931
  format %{ "movq    $dst,[$constantaddress]\t! replicate8B($con)" %}
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3932
  ins_encode %{
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  3933
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 1)));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3934
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3935
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3936
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3937
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3938
// Replicate byte scalar zero to be vector
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3939
instruct Repl4B_zero(vecS dst, immI0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3940
  predicate(n->as_Vector()->length() == 4);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3941
  match(Set dst (ReplicateB zero));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3942
  format %{ "pxor    $dst,$dst\t! replicate4B zero" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3943
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3944
    __ pxor($dst$$XMMRegister, $dst$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3945
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3946
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3947
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3948
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3949
instruct Repl8B_zero(vecD dst, immI0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3950
  predicate(n->as_Vector()->length() == 8);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3951
  match(Set dst (ReplicateB zero));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3952
  format %{ "pxor    $dst,$dst\t! replicate8B zero" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3953
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3954
    __ pxor($dst$$XMMRegister, $dst$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3955
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3956
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3957
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3958
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3959
instruct Repl16B_zero(vecX dst, immI0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3960
  predicate(n->as_Vector()->length() == 16);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3961
  match(Set dst (ReplicateB zero));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3962
  format %{ "pxor    $dst,$dst\t! replicate16B zero" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3963
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3964
    __ pxor($dst$$XMMRegister, $dst$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3965
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3966
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3967
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3968
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3969
instruct Repl32B_zero(vecY dst, immI0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3970
  predicate(n->as_Vector()->length() == 32);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3971
  match(Set dst (ReplicateB zero));
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  3972
  format %{ "vpxor   $dst,$dst,$dst\t! replicate32B zero" %}
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3973
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3974
    // Use vxorpd since AVX does not have vpxor for 256-bit (AVX2 will have it).
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3975
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3976
    __ vpxor($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3977
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3978
  ins_pipe( fpu_reg_reg );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3979
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  3980
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3981
// Replicate char/short (2 byte) scalar to be vector
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3982
instruct Repl2S(vecS dst, rRegI src) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3983
  predicate(n->as_Vector()->length() == 2);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3984
  match(Set dst (ReplicateS src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3985
  format %{ "movd    $dst,$src\n\t"
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3986
            "pshuflw $dst,$dst,0x00\t! replicate2S" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3987
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3988
    __ movdl($dst$$XMMRegister, $src$$Register);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3989
    __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3990
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3991
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3992
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3993
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3994
// Replicate char/short (2 byte) scalar immediate to be vector by loading from const table.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3995
instruct Repl2S_imm(vecS dst, immI con) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3996
  predicate(n->as_Vector()->length() == 2);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3997
  match(Set dst (ReplicateS con));
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  3998
  format %{ "movdl   $dst,[$constantaddress]\t! replicate2S($con)" %}
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  3999
  ins_encode %{
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4000
    __ movdl($dst$$XMMRegister, $constantaddress(replicate4_imm($con$$constant, 2)));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4001
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4002
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4003
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4004
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4005
instruct Repl4S_imm(vecD dst, immI con) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4006
  predicate(n->as_Vector()->length() == 4);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4007
  match(Set dst (ReplicateS con));
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4008
  format %{ "movq    $dst,[$constantaddress]\t! replicate4S($con)" %}
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4009
  ins_encode %{
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4010
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 2)));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4011
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4012
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4013
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4014
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4015
// Replicate char/short (2 byte) scalar zero to be vector
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4016
instruct Repl2S_zero(vecS dst, immI0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4017
  predicate(n->as_Vector()->length() == 2);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4018
  match(Set dst (ReplicateS zero));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4019
  format %{ "pxor    $dst,$dst\t! replicate2S zero" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4020
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4021
    __ pxor($dst$$XMMRegister, $dst$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4022
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4023
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4024
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4025
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4026
instruct Repl4S_zero(vecD dst, immI0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4027
  predicate(n->as_Vector()->length() == 4);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4028
  match(Set dst (ReplicateS zero));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4029
  format %{ "pxor    $dst,$dst\t! replicate4S zero" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4030
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4031
    __ pxor($dst$$XMMRegister, $dst$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4032
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4033
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4034
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4035
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4036
instruct Repl8S_zero(vecX dst, immI0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4037
  predicate(n->as_Vector()->length() == 8);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4038
  match(Set dst (ReplicateS zero));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4039
  format %{ "pxor    $dst,$dst\t! replicate8S zero" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4040
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4041
    __ pxor($dst$$XMMRegister, $dst$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4042
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4043
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4044
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4045
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4046
instruct Repl16S_zero(vecY dst, immI0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4047
  predicate(n->as_Vector()->length() == 16);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4048
  match(Set dst (ReplicateS zero));
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4049
  format %{ "vpxor   $dst,$dst,$dst\t! replicate16S zero" %}
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4050
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4051
    // Use vxorpd since AVX does not have vpxor for 256-bit (AVX2 will have it).
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4052
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4053
    __ vpxor($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4054
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4055
  ins_pipe( fpu_reg_reg );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4056
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4057
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4058
// Replicate integer (4 byte) scalar to be vector
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4059
instruct Repl2I(vecD dst, rRegI src) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4060
  predicate(n->as_Vector()->length() == 2);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4061
  match(Set dst (ReplicateI src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4062
  format %{ "movd    $dst,$src\n\t"
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4063
            "pshufd  $dst,$dst,0x00\t! replicate2I" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4064
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4065
    __ movdl($dst$$XMMRegister, $src$$Register);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4066
    __ pshufd($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4067
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4068
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4069
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4070
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4071
// Integer could be loaded into xmm register directly from memory.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4072
instruct Repl2I_mem(vecD dst, memory mem) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4073
  predicate(n->as_Vector()->length() == 2);
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4074
  match(Set dst (ReplicateI (LoadI mem)));
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4075
  format %{ "movd    $dst,$mem\n\t"
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4076
            "pshufd  $dst,$dst,0x00\t! replicate2I" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4077
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4078
    __ movdl($dst$$XMMRegister, $mem$$Address);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4079
    __ pshufd($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4080
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4081
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4082
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4083
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4084
// Replicate integer (4 byte) scalar immediate to be vector by loading from const table.
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4085
instruct Repl2I_imm(vecD dst, immI con) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4086
  predicate(n->as_Vector()->length() == 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4087
  match(Set dst (ReplicateI con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4088
  format %{ "movq    $dst,[$constantaddress]\t! replicate2I($con)" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4089
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4090
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 4)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4091
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4092
  ins_pipe( fpu_reg_reg );
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4093
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4094
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4095
// Replicate integer (4 byte) scalar zero to be vector
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4096
instruct Repl2I_zero(vecD dst, immI0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4097
  predicate(n->as_Vector()->length() == 2);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4098
  match(Set dst (ReplicateI zero));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4099
  format %{ "pxor    $dst,$dst\t! replicate2I" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4100
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4101
    __ pxor($dst$$XMMRegister, $dst$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4102
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4103
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4104
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4105
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4106
instruct Repl4I_zero(vecX dst, immI0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4107
  predicate(n->as_Vector()->length() == 4);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4108
  match(Set dst (ReplicateI zero));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4109
  format %{ "pxor    $dst,$dst\t! replicate4I zero)" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4110
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4111
    __ pxor($dst$$XMMRegister, $dst$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4112
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4113
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4114
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4115
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4116
instruct Repl8I_zero(vecY dst, immI0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4117
  predicate(n->as_Vector()->length() == 8);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4118
  match(Set dst (ReplicateI zero));
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4119
  format %{ "vpxor   $dst,$dst,$dst\t! replicate8I zero" %}
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4120
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4121
    // Use vxorpd since AVX does not have vpxor for 256-bit (AVX2 will have it).
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4122
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4123
    __ vpxor($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4124
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4125
  ins_pipe( fpu_reg_reg );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4126
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4127
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4128
// Replicate long (8 byte) scalar to be vector
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4129
#ifdef _LP64
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4130
instruct Repl2L(vecX dst, rRegL src) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4131
  predicate(n->as_Vector()->length() == 2);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4132
  match(Set dst (ReplicateL src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4133
  format %{ "movdq   $dst,$src\n\t"
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4134
            "punpcklqdq $dst,$dst\t! replicate2L" %}
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4135
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4136
    __ movdq($dst$$XMMRegister, $src$$Register);
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4137
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4138
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4139
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4140
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4141
#else // _LP64
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4142
instruct Repl2L(vecX dst, eRegL src, vecX tmp) %{
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4143
  predicate(n->as_Vector()->length() == 2);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4144
  match(Set dst (ReplicateL src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4145
  effect(TEMP dst, USE src, TEMP tmp);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4146
  format %{ "movdl   $dst,$src.lo\n\t"
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4147
            "movdl   $tmp,$src.hi\n\t"
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4148
            "punpckldq $dst,$tmp\n\t"
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4149
            "punpcklqdq $dst,$dst\t! replicate2L"%}
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4150
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4151
    __ movdl($dst$$XMMRegister, $src$$Register);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4152
    __ movdl($tmp$$XMMRegister, HIGH_FROM_LOW($src$$Register));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4153
    __ punpckldq($dst$$XMMRegister, $tmp$$XMMRegister);
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4154
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4155
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4156
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4157
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4158
#endif // _LP64
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4159
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4160
// Replicate long (8 byte) scalar immediate to be vector by loading from const table.
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4161
instruct Repl2L_imm(vecX dst, immL con) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4162
  predicate(n->as_Vector()->length() == 2);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4163
  match(Set dst (ReplicateL con));
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4164
  format %{ "movq    $dst,[$constantaddress]\n\t"
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4165
            "punpcklqdq $dst,$dst\t! replicate2L($con)" %}
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4166
  ins_encode %{
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4167
    __ movq($dst$$XMMRegister, $constantaddress($con));
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4168
    __ punpcklqdq($dst$$XMMRegister, $dst$$XMMRegister);
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4169
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4170
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4171
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4172
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4173
// Replicate long (8 byte) scalar zero to be vector
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4174
instruct Repl2L_zero(vecX dst, immL0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4175
  predicate(n->as_Vector()->length() == 2);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4176
  match(Set dst (ReplicateL zero));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4177
  format %{ "pxor    $dst,$dst\t! replicate2L zero" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4178
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4179
    __ pxor($dst$$XMMRegister, $dst$$XMMRegister);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4180
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4181
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4182
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4183
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4184
instruct Repl4L_zero(vecY dst, immL0 zero) %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4185
  predicate(n->as_Vector()->length() == 4);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4186
  match(Set dst (ReplicateL zero));
13294
80131b419f85 7181494: cleanup avx and vectors code
kvn
parents: 13108
diff changeset
  4187
  format %{ "vpxor   $dst,$dst,$dst\t! replicate4L zero" %}
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4188
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4189
    // Use vxorpd since AVX does not have vpxor for 256-bit (AVX2 will have it).
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4190
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4191
    __ vpxor($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4192
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4193
  ins_pipe( fpu_reg_reg );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4194
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4195
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4196
// Replicate float (4 byte) scalar to be vector
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4197
instruct Repl2F(vecD dst, vlRegF src) %{
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4198
  predicate(n->as_Vector()->length() == 2);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4199
  match(Set dst (ReplicateF src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4200
  format %{ "pshufd  $dst,$dst,0x00\t! replicate2F" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4201
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4202
    __ pshufd($dst$$XMMRegister, $src$$XMMRegister, 0x00);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4203
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4204
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4205
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4206
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4207
instruct Repl4F(vecX dst, vlRegF src) %{
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4208
  predicate(n->as_Vector()->length() == 4);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4209
  match(Set dst (ReplicateF src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4210
  format %{ "pshufd  $dst,$dst,0x00\t! replicate4F" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4211
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4212
    __ pshufd($dst$$XMMRegister, $src$$XMMRegister, 0x00);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4213
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4214
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4215
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4216
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4217
// Replicate double (8 bytes) scalar to be vector
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4218
instruct Repl2D(vecX dst, vlRegD src) %{
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4219
  predicate(n->as_Vector()->length() == 2);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4220
  match(Set dst (ReplicateD src));
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4221
  format %{ "pshufd  $dst,$src,0x44\t! replicate2D" %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4222
  ins_encode %{
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4223
    __ pshufd($dst$$XMMRegister, $src$$XMMRegister, 0x44);
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4224
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4225
  ins_pipe( pipe_slow );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4226
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4227
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4228
// ====================EVEX REPLICATE=============================================
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4229
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4230
instruct Repl4B_mem_evex(vecS dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4231
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4232
  match(Set dst (ReplicateB (LoadB mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4233
  format %{ "vpbroadcastb  $dst,$mem\t! replicate4B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4234
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4235
    int vector_len = 0;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4236
    __ vpbroadcastb($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4237
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4238
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4239
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4240
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4241
instruct Repl8B_mem_evex(vecD dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4242
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4243
  match(Set dst (ReplicateB (LoadB mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4244
  format %{ "vpbroadcastb  $dst,$mem\t! replicate8B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4245
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4246
    int vector_len = 0;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4247
    __ vpbroadcastb($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4248
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4249
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4250
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4251
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4252
instruct Repl16B_evex(vecX dst, rRegI src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4253
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4254
  match(Set dst (ReplicateB src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4255
  format %{ "evpbroadcastb $dst,$src\t! replicate16B" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4256
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4257
   int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4258
    __ evpbroadcastb($dst$$XMMRegister, $src$$Register, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4259
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4260
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4261
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4262
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4263
instruct Repl16B_mem_evex(vecX dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4264
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4265
  match(Set dst (ReplicateB (LoadB mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4266
  format %{ "vpbroadcastb  $dst,$mem\t! replicate16B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4267
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4268
    int vector_len = 0;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4269
    __ vpbroadcastb($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4270
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4271
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4272
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4273
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4274
instruct Repl32B_evex(vecY dst, rRegI src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4275
  predicate(n->as_Vector()->length() == 32 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4276
  match(Set dst (ReplicateB src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4277
  format %{ "evpbroadcastb $dst,$src\t! replicate32B" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4278
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4279
   int vector_len = 1;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4280
    __ evpbroadcastb($dst$$XMMRegister, $src$$Register, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4281
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4282
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4283
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4284
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4285
instruct Repl32B_mem_evex(vecY dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4286
  predicate(n->as_Vector()->length() == 32 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4287
  match(Set dst (ReplicateB (LoadB mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4288
  format %{ "vpbroadcastb  $dst,$mem\t! replicate32B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4289
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4290
    int vector_len = 1;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4291
    __ vpbroadcastb($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4292
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4293
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4294
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4295
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4296
instruct Repl64B_evex(vecZ dst, rRegI src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4297
  predicate(n->as_Vector()->length() == 64 && UseAVX > 2 && VM_Version::supports_avx512bw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4298
  match(Set dst (ReplicateB src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4299
  format %{ "evpbroadcastb $dst,$src\t! upper replicate64B" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4300
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4301
   int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4302
    __ evpbroadcastb($dst$$XMMRegister, $src$$Register, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4303
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4304
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4305
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4306
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4307
instruct Repl64B_mem_evex(vecZ dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4308
  predicate(n->as_Vector()->length() == 64 && UseAVX > 2 && VM_Version::supports_avx512bw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4309
  match(Set dst (ReplicateB (LoadB mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4310
  format %{ "vpbroadcastb  $dst,$mem\t! replicate64B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4311
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4312
    int vector_len = 2;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4313
    __ vpbroadcastb($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4314
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4315
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4316
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4317
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4318
instruct Repl16B_imm_evex(vecX dst, immI con) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4319
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4320
  match(Set dst (ReplicateB con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4321
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4322
            "vpbroadcastb $dst,$dst\t! replicate16B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4323
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4324
   int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4325
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 1)));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4326
    __ vpbroadcastb($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4327
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4328
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4329
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4330
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4331
instruct Repl32B_imm_evex(vecY dst, immI con) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4332
  predicate(n->as_Vector()->length() == 32 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4333
  match(Set dst (ReplicateB con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4334
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4335
            "vpbroadcastb $dst,$dst\t! replicate32B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4336
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4337
   int vector_len = 1;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4338
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 1)));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4339
    __ vpbroadcastb($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4340
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4341
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4342
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4343
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4344
instruct Repl64B_imm_evex(vecZ dst, immI con) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4345
  predicate(n->as_Vector()->length() == 64 && UseAVX > 2 && VM_Version::supports_avx512bw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4346
  match(Set dst (ReplicateB con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4347
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4348
            "vpbroadcastb $dst,$dst\t! upper replicate64B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4349
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4350
   int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4351
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 1)));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4352
    __ vpbroadcastb($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4353
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4354
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4355
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4356
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4357
instruct Repl64B_zero_evex(vecZ dst, immI0 zero) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4358
  predicate(n->as_Vector()->length() == 64 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4359
  match(Set dst (ReplicateB zero));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4360
  format %{ "vpxor   $dst k0,$dst,$dst\t! replicate64B zero" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4361
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4362
    // Use vxorpd since AVX does not have vpxor for 512-bit (EVEX will have it).
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4363
    int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4364
    __ vpxor($dst$$XMMRegister,$dst$$XMMRegister, $dst$$XMMRegister, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4365
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4366
  ins_pipe( fpu_reg_reg );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4367
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4368
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4369
instruct Repl4S_evex(vecD dst, rRegI src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4370
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4371
  match(Set dst (ReplicateS src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4372
  format %{ "evpbroadcastw $dst,$src\t! replicate4S" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4373
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4374
   int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4375
    __ evpbroadcastw($dst$$XMMRegister, $src$$Register, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4376
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4377
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4378
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4379
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4380
instruct Repl4S_mem_evex(vecD dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4381
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4382
  match(Set dst (ReplicateS (LoadS mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4383
  format %{ "vpbroadcastw  $dst,$mem\t! replicate4S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4384
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4385
    int vector_len = 0;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4386
    __ vpbroadcastw($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4387
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4388
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4389
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4390
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4391
instruct Repl8S_evex(vecX dst, rRegI src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4392
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4393
  match(Set dst (ReplicateS src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4394
  format %{ "evpbroadcastw $dst,$src\t! replicate8S" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4395
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4396
   int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4397
    __ evpbroadcastw($dst$$XMMRegister, $src$$Register, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4398
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4399
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4400
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4401
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4402
instruct Repl8S_mem_evex(vecX dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4403
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4404
  match(Set dst (ReplicateS (LoadS mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4405
  format %{ "vpbroadcastw  $dst,$mem\t! replicate8S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4406
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4407
    int vector_len = 0;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4408
    __ vpbroadcastw($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4409
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4410
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4411
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4412
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4413
instruct Repl16S_evex(vecY dst, rRegI src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4414
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4415
  match(Set dst (ReplicateS src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4416
  format %{ "evpbroadcastw $dst,$src\t! replicate16S" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4417
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4418
   int vector_len = 1;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4419
    __ evpbroadcastw($dst$$XMMRegister, $src$$Register, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4420
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4421
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4422
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4423
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4424
instruct Repl16S_mem_evex(vecY dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4425
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4426
  match(Set dst (ReplicateS (LoadS mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4427
  format %{ "vpbroadcastw  $dst,$mem\t! replicate16S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4428
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4429
    int vector_len = 1;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4430
    __ vpbroadcastw($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4431
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4432
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4433
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4434
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4435
instruct Repl32S_evex(vecZ dst, rRegI src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4436
  predicate(n->as_Vector()->length() == 32 && UseAVX > 2 && VM_Version::supports_avx512bw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4437
  match(Set dst (ReplicateS src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4438
  format %{ "evpbroadcastw $dst,$src\t! replicate32S" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4439
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4440
   int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4441
    __ evpbroadcastw($dst$$XMMRegister, $src$$Register, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4442
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4443
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4444
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4445
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4446
instruct Repl32S_mem_evex(vecZ dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4447
  predicate(n->as_Vector()->length() == 32 && UseAVX > 2 && VM_Version::supports_avx512bw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4448
  match(Set dst (ReplicateS (LoadS mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4449
  format %{ "vpbroadcastw  $dst,$mem\t! replicate32S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4450
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4451
    int vector_len = 2;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4452
    __ vpbroadcastw($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4453
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4454
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4455
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4456
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4457
instruct Repl8S_imm_evex(vecX dst, immI con) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4458
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4459
  match(Set dst (ReplicateS con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4460
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4461
            "vpbroadcastw $dst,$dst\t! replicate8S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4462
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4463
   int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4464
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 2)));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4465
    __ vpbroadcastw($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4466
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4467
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4468
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4469
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4470
instruct Repl16S_imm_evex(vecY dst, immI con) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4471
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2 && VM_Version::supports_avx512vlbw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4472
  match(Set dst (ReplicateS con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4473
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4474
            "vpbroadcastw $dst,$dst\t! replicate16S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4475
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4476
   int vector_len = 1;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4477
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 2)));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4478
    __ vpbroadcastw($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4479
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4480
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4481
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4482
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4483
instruct Repl32S_imm_evex(vecZ dst, immI con) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4484
  predicate(n->as_Vector()->length() == 32 && UseAVX > 2 && VM_Version::supports_avx512bw());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4485
  match(Set dst (ReplicateS con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4486
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4487
            "vpbroadcastw $dst,$dst\t! replicate32S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4488
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4489
   int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4490
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 2)));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4491
    __ vpbroadcastw($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4492
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4493
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4494
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4495
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4496
instruct Repl32S_zero_evex(vecZ dst, immI0 zero) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4497
  predicate(n->as_Vector()->length() == 32 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4498
  match(Set dst (ReplicateS zero));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4499
  format %{ "vpxor   $dst k0,$dst,$dst\t! replicate32S zero" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4500
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4501
    // Use vxorpd since AVX does not have vpxor for 512-bit (EVEX will have it).
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4502
    int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4503
    __ vpxor($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4504
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4505
  ins_pipe( fpu_reg_reg );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4506
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4507
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4508
instruct Repl4I_evex(vecX dst, rRegI src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4509
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4510
  match(Set dst (ReplicateI src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4511
  format %{ "evpbroadcastd  $dst,$src\t! replicate4I" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4512
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4513
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4514
    __ evpbroadcastd($dst$$XMMRegister, $src$$Register, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4515
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4516
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4517
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4518
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4519
instruct Repl4I_mem_evex(vecX dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4520
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4521
  match(Set dst (ReplicateI (LoadI mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4522
  format %{ "vpbroadcastd  $dst,$mem\t! replicate4I" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4523
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4524
    int vector_len = 0;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4525
    __ vpbroadcastd($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4526
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4527
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4528
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4529
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4530
instruct Repl8I_evex(vecY dst, rRegI src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4531
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4532
  match(Set dst (ReplicateI src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4533
  format %{ "evpbroadcastd  $dst,$src\t! replicate8I" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4534
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4535
    int vector_len = 1;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4536
    __ evpbroadcastd($dst$$XMMRegister, $src$$Register, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4537
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4538
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4539
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4540
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4541
instruct Repl8I_mem_evex(vecY dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4542
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4543
  match(Set dst (ReplicateI (LoadI mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4544
  format %{ "vpbroadcastd  $dst,$mem\t! replicate8I" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4545
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4546
    int vector_len = 1;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4547
    __ vpbroadcastd($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4548
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4549
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4550
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4551
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4552
instruct Repl16I_evex(vecZ dst, rRegI src) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4553
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4554
  match(Set dst (ReplicateI src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4555
  format %{ "evpbroadcastd  $dst,$src\t! replicate16I" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4556
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4557
    int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4558
    __ evpbroadcastd($dst$$XMMRegister, $src$$Register, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4559
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4560
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4561
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4562
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4563
instruct Repl16I_mem_evex(vecZ dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4564
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4565
  match(Set dst (ReplicateI (LoadI mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4566
  format %{ "vpbroadcastd  $dst,$mem\t! replicate16I" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4567
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4568
    int vector_len = 2;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4569
    __ vpbroadcastd($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4570
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4571
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4572
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4573
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4574
instruct Repl4I_imm_evex(vecX dst, immI con) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4575
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4576
  match(Set dst (ReplicateI con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4577
  format %{ "movq    $dst,[$constantaddress]\t! replicate8I($con)\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4578
            "vpbroadcastd  $dst,$dst\t! replicate4I" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4579
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4580
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4581
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 4)));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4582
    __ vpbroadcastd($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4583
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4584
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4585
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4586
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4587
instruct Repl8I_imm_evex(vecY dst, immI con) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4588
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4589
  match(Set dst (ReplicateI con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4590
  format %{ "movq    $dst,[$constantaddress]\t! replicate8I($con)\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4591
            "vpbroadcastd  $dst,$dst\t! replicate8I" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4592
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4593
    int vector_len = 1;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4594
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 4)));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4595
    __ vpbroadcastd($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4596
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4597
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4598
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4599
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4600
instruct Repl16I_imm_evex(vecZ dst, immI con) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4601
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4602
  match(Set dst (ReplicateI con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4603
  format %{ "movq    $dst,[$constantaddress]\t! replicate16I($con)\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4604
            "vpbroadcastd  $dst,$dst\t! replicate16I" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4605
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4606
    int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4607
    __ movq($dst$$XMMRegister, $constantaddress(replicate8_imm($con$$constant, 4)));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4608
    __ vpbroadcastd($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4609
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4610
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4611
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4612
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4613
instruct Repl16I_zero_evex(vecZ dst, immI0 zero) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4614
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4615
  match(Set dst (ReplicateI zero));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4616
  format %{ "vpxor   $dst k0,$dst,$dst\t! replicate16I zero" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4617
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4618
    // Use vxorpd since AVX does not have vpxor for 512-bit (AVX2 will have it).
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4619
    int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4620
    __ vpxor($dst$$XMMRegister, $dst$$XMMRegister, $dst$$XMMRegister, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4621
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4622
  ins_pipe( fpu_reg_reg );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4623
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4624
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4625
// Replicate long (8 byte) scalar to be vector
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4626
#ifdef _LP64
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4627
instruct Repl4L_evex(vecY dst, rRegL src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4628
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4629
  match(Set dst (ReplicateL src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4630
  format %{ "evpbroadcastq  $dst,$src\t! replicate4L" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4631
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4632
    int vector_len = 1;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4633
    __ evpbroadcastq($dst$$XMMRegister, $src$$Register, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4634
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4635
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4636
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4637
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4638
instruct Repl8L_evex(vecZ dst, rRegL src) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4639
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4640
  match(Set dst (ReplicateL src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4641
  format %{ "evpbroadcastq  $dst,$src\t! replicate8L" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4642
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4643
    int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4644
    __ evpbroadcastq($dst$$XMMRegister, $src$$Register, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4645
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4646
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4647
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4648
#else // _LP64
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4649
instruct Repl4L_evex(vecY dst, eRegL src, regD tmp) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4650
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4651
  match(Set dst (ReplicateL src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4652
  effect(TEMP dst, USE src, TEMP tmp);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4653
  format %{ "movdl   $dst,$src.lo\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4654
            "movdl   $tmp,$src.hi\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4655
            "punpckldq $dst,$tmp\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4656
            "vpbroadcastq  $dst,$dst\t! replicate4L" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4657
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4658
    int vector_len = 1;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4659
    __ movdl($dst$$XMMRegister, $src$$Register);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4660
    __ movdl($tmp$$XMMRegister, HIGH_FROM_LOW($src$$Register));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4661
    __ punpckldq($dst$$XMMRegister, $tmp$$XMMRegister);
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4662
    __ vpbroadcastq($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4663
  %}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4664
  ins_pipe( pipe_slow );
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4665
%}
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4666
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4667
instruct Repl8L_evex(legVecZ dst, eRegL src, legVecZ tmp) %{
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4668
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4669
  match(Set dst (ReplicateL src));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4670
  effect(TEMP dst, USE src, TEMP tmp);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4671
  format %{ "movdl   $dst,$src.lo\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4672
            "movdl   $tmp,$src.hi\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4673
            "punpckldq $dst,$tmp\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4674
            "vpbroadcastq  $dst,$dst\t! replicate8L" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4675
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4676
    int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4677
    __ movdl($dst$$XMMRegister, $src$$Register);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4678
    __ movdl($tmp$$XMMRegister, HIGH_FROM_LOW($src$$Register));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4679
    __ punpckldq($dst$$XMMRegister, $tmp$$XMMRegister);
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4680
    __ vpbroadcastq($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4681
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4682
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4683
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4684
#endif // _LP64
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4685
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4686
instruct Repl4L_imm_evex(vecY dst, immL con) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4687
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4688
  match(Set dst (ReplicateL con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4689
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4690
            "vpbroadcastq  $dst,$dst\t! replicate4L" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4691
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4692
    int vector_len = 1;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4693
    __ movq($dst$$XMMRegister, $constantaddress($con));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4694
    __ vpbroadcastq($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4695
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4696
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4697
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4698
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4699
instruct Repl8L_imm_evex(vecZ dst, immL con) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4700
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4701
  match(Set dst (ReplicateL con));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4702
  format %{ "movq    $dst,[$constantaddress]\n\t"
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4703
            "vpbroadcastq  $dst,$dst\t! replicate8L" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4704
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4705
    int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4706
    __ movq($dst$$XMMRegister, $constantaddress($con));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4707
    __ vpbroadcastq($dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4708
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4709
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4710
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4711
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4712
instruct Repl2L_mem_evex(vecX dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4713
  predicate(n->as_Vector()->length() == 2 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4714
  match(Set dst (ReplicateL (LoadL mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4715
  format %{ "vpbroadcastd  $dst,$mem\t! replicate2L" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4716
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4717
    int vector_len = 0;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4718
    __ vpbroadcastq($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4719
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4720
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4721
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4722
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4723
instruct Repl4L_mem_evex(vecY dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4724
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4725
  match(Set dst (ReplicateL (LoadL mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4726
  format %{ "vpbroadcastd  $dst,$mem\t! replicate4L" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4727
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4728
    int vector_len = 1;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4729
    __ vpbroadcastq($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4730
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4731
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4732
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4733
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4734
instruct Repl8L_mem_evex(vecZ dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4735
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4736
  match(Set dst (ReplicateL (LoadL mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4737
  format %{ "vpbroadcastd  $dst,$mem\t! replicate8L" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4738
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4739
    int vector_len = 2;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4740
    __ vpbroadcastq($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4741
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4742
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4743
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4744
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4745
instruct Repl8L_zero_evex(vecZ dst, immL0 zero) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4746
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4747
  match(Set dst (ReplicateL zero));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4748
  format %{ "vpxor   $dst k0,$dst,$dst\t! replicate8L zero" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4749
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4750
    // Use vxorpd since AVX does not have vpxor for 512-bit (EVEX will have it).
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4751
    int vector_len = 2;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4752
    __ vpxor($dst$$XMMRegister,$dst$$XMMRegister, $dst$$XMMRegister, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4753
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4754
  ins_pipe( fpu_reg_reg );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4755
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4756
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4757
instruct Repl8F_evex(vecY dst, regF src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4758
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4759
  match(Set dst (ReplicateF src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4760
  format %{ "vpbroadcastss $dst,$src\t! replicate8F" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4761
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4762
    int vector_len = 1;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4763
    __ vpbroadcastss($dst$$XMMRegister, $src$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4764
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4765
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4766
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4767
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4768
instruct Repl8F_mem_evex(vecY dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4769
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4770
  match(Set dst (ReplicateF (LoadF mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4771
  format %{ "vbroadcastss  $dst,$mem\t! replicate8F" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4772
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4773
    int vector_len = 1;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4774
    __ vpbroadcastss($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4775
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4776
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4777
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4778
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4779
instruct Repl16F_evex(vecZ dst, regF src) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4780
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4781
  match(Set dst (ReplicateF src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4782
  format %{ "vpbroadcastss $dst,$src\t! replicate16F" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4783
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4784
    int vector_len = 2;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4785
    __ vpbroadcastss($dst$$XMMRegister, $src$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4786
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4787
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4788
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4789
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4790
instruct Repl16F_mem_evex(vecZ dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4791
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4792
  match(Set dst (ReplicateF (LoadF mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4793
  format %{ "vbroadcastss  $dst,$mem\t! replicate16F" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4794
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4795
    int vector_len = 2;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4796
    __ vpbroadcastss($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4797
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4798
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4799
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4800
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4801
instruct Repl2F_zero_evex(vecD dst, immF0 zero) %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4802
  predicate(n->as_Vector()->length() == 2 && UseAVX > 2);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4803
  match(Set dst (ReplicateF zero));
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4804
  format %{ "vpxor  $dst k0,$dst,$dst\t! replicate2F zero" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4805
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4806
    // Use vpxor in place of vxorps since EVEX has a constriant on dq for vxorps: this is a 512-bit operation
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4807
    int vector_len = 2;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4808
    __ vpxor($dst$$XMMRegister,$dst$$XMMRegister, $dst$$XMMRegister, vector_len);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4809
  %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4810
  ins_pipe( fpu_reg_reg );
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4811
%}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4812
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4813
instruct Repl4F_zero_evex(vecX dst, immF0 zero) %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4814
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4815
  match(Set dst (ReplicateF zero));
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4816
  format %{ "vpxor  $dst k0,$dst,$dst\t! replicate4F zero" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4817
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4818
    // Use vpxor in place of vxorps since EVEX has a constriant on dq for vxorps: this is a 512-bit operation
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4819
    int vector_len = 2;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4820
    __ vpxor($dst$$XMMRegister,$dst$$XMMRegister, $dst$$XMMRegister, vector_len);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4821
  %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4822
  ins_pipe( fpu_reg_reg );
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4823
%}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4824
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4825
instruct Repl8F_zero_evex(vecY dst, immF0 zero) %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4826
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4827
  match(Set dst (ReplicateF zero));
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4828
  format %{ "vpxor  $dst k0,$dst,$dst\t! replicate8F zero" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4829
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4830
    // Use vpxor in place of vxorps since EVEX has a constriant on dq for vxorps: this is a 512-bit operation
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4831
    int vector_len = 2;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4832
    __ vpxor($dst$$XMMRegister,$dst$$XMMRegister, $dst$$XMMRegister, vector_len);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4833
  %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4834
  ins_pipe( fpu_reg_reg );
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4835
%}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4836
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4837
instruct Repl16F_zero_evex(vecZ dst, immF0 zero) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4838
  predicate(n->as_Vector()->length() == 16 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4839
  match(Set dst (ReplicateF zero));
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4840
  format %{ "vpxor  $dst k0,$dst,$dst\t! replicate16F zero" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4841
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4842
    // Use vpxor in place of vxorps since EVEX has a constriant on dq for vxorps: this is a 512-bit operation
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4843
    int vector_len = 2;
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4844
    __ vpxor($dst$$XMMRegister,$dst$$XMMRegister, $dst$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4845
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4846
  ins_pipe( fpu_reg_reg );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4847
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4848
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4849
instruct Repl4D_evex(vecY dst, regD src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4850
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4851
  match(Set dst (ReplicateD src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4852
  format %{ "vpbroadcastsd $dst,$src\t! replicate4D" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4853
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4854
    int vector_len = 1;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4855
    __ vpbroadcastsd($dst$$XMMRegister, $src$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4856
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4857
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4858
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4859
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4860
instruct Repl4D_mem_evex(vecY dst, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4861
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2 && VM_Version::supports_avx512vl());
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4862
  match(Set dst (ReplicateD (LoadD mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4863
  format %{ "vbroadcastsd  $dst,$mem\t! replicate4D" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4864
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4865
    int vector_len = 1;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4866
    __ vpbroadcastsd($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4867
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4868
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4869
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4870
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4871
instruct Repl8D_evex(vecZ dst, regD src) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4872
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4873
  match(Set dst (ReplicateD src));
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4874
  format %{ "vpbroadcastsd $dst,$src\t! replicate8D" %}
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4875
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4876
    int vector_len = 2;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4877
    __ vpbroadcastsd($dst$$XMMRegister, $src$$XMMRegister, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4878
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4879
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4880
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4881
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4882
instruct Repl8D_mem_evex(vecZ dst, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4883
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4884
  match(Set dst (ReplicateD (LoadD mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4885
  format %{ "vbroadcastsd  $dst,$mem\t! replicate8D" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4886
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4887
    int vector_len = 2;
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4888
    __ vpbroadcastsd($dst$$XMMRegister, $mem$$Address, vector_len);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4889
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4890
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4891
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4892
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4893
instruct Repl2D_zero_evex(vecX dst, immD0 zero) %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4894
  predicate(n->as_Vector()->length() == 2 && UseAVX > 2);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4895
  match(Set dst (ReplicateD zero));
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4896
  format %{ "vpxor  $dst k0,$dst,$dst\t! replicate2D zero" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4897
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4898
    // Use vpxor in place of vxorpd since EVEX has a constriant on dq for vxorpd: this is a 512-bit operation
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4899
    int vector_len = 2;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4900
    __ vpxor($dst$$XMMRegister,$dst$$XMMRegister, $dst$$XMMRegister, vector_len);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4901
  %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4902
  ins_pipe( fpu_reg_reg );
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4903
%}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4904
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4905
instruct Repl4D_zero_evex(vecY dst, immD0 zero) %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4906
  predicate(n->as_Vector()->length() == 4 && UseAVX > 2);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4907
  match(Set dst (ReplicateD zero));
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4908
  format %{ "vpxor  $dst k0,$dst,$dst\t! replicate4D zero" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4909
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4910
    // Use vpxor in place of vxorpd since EVEX has a constriant on dq for vxorpd: this is a 512-bit operation
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4911
    int vector_len = 2;
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4912
    __ vpxor($dst$$XMMRegister,$dst$$XMMRegister, $dst$$XMMRegister, vector_len);
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4913
  %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4914
  ins_pipe( fpu_reg_reg );
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4915
%}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4916
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4917
instruct Repl8D_zero_evex(vecZ dst, immD0 zero) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  4918
  predicate(n->as_Vector()->length() == 8 && UseAVX > 2);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4919
  match(Set dst (ReplicateD zero));
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4920
  format %{ "vpxor  $dst k0,$dst,$dst,vect512\t! replicate8D zero" %}
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4921
  ins_encode %{
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4922
    // Use vpxor in place of vxorpd since EVEX has a constriant on dq for vxorpd: this is a 512-bit operation
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4923
    int vector_len = 2;
32727
320855c2baef 8132160: support for AVX 512 call frames and stack management
mcberg
parents: 32723
diff changeset
  4924
    __ vpxor($dst$$XMMRegister,$dst$$XMMRegister, $dst$$XMMRegister, vector_len);
13104
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4925
  %}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4926
  ins_pipe( fpu_reg_reg );
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4927
%}
657b387034fb 7119644: Increase superword's vector size up to 256 bits
kvn
parents: 11794
diff changeset
  4928
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4929
// ====================REDUCTION ARITHMETIC=======================================
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4930
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4931
instruct rsadd2I_reduction_reg(rRegI dst, rRegI src1, vecD src2, vecD tmp, vecD tmp2) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4932
  predicate(UseSSE > 2 && UseAVX == 0);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4933
  match(Set dst (AddReductionVI src1 src2));
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4934
  effect(TEMP tmp2, TEMP tmp);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4935
  format %{ "movdqu  $tmp2,$src2\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4936
            "phaddd  $tmp2,$tmp2\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4937
            "movd    $tmp,$src1\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4938
            "paddd   $tmp,$tmp2\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4939
            "movd    $dst,$tmp\t! add reduction2I" %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4940
  ins_encode %{
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4941
    __ movdqu($tmp2$$XMMRegister, $src2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4942
    __ phaddd($tmp2$$XMMRegister, $tmp2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4943
    __ movdl($tmp$$XMMRegister, $src1$$Register);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4944
    __ paddd($tmp$$XMMRegister, $tmp2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4945
    __ movdl($dst$$Register, $tmp$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4946
  %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4947
  ins_pipe( pipe_slow );
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4948
%}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4949
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4950
instruct rvadd2I_reduction_reg(rRegI dst, rRegI src1, vecD src2, vecD tmp, vecD tmp2) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  4951
  predicate(VM_Version::supports_avxonly());
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4952
  match(Set dst (AddReductionVI src1 src2));
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4953
  effect(TEMP tmp, TEMP tmp2);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4954
  format %{ "vphaddd  $tmp,$src2,$src2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4955
            "movd     $tmp2,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4956
            "vpaddd   $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4957
            "movd     $dst,$tmp2\t! add reduction2I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4958
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4959
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4960
    __ vphaddd($tmp$$XMMRegister, $src2$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4961
    __ movdl($tmp2$$XMMRegister, $src1$$Register);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4962
    __ vpaddd($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4963
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4964
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4965
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4966
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4967
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4968
instruct rvadd2I_reduction_reg_evex(rRegI dst, rRegI src1, vecD src2, vecD tmp, vecD tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4969
  predicate(UseAVX > 2);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4970
  match(Set dst (AddReductionVI src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4971
  effect(TEMP tmp, TEMP tmp2);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4972
  format %{ "pshufd  $tmp2,$src2,0x1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4973
            "vpaddd  $tmp,$src2,$tmp2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4974
            "movd    $tmp2,$src1\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4975
            "vpaddd  $tmp2,$tmp,$tmp2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4976
            "movd    $dst,$tmp2\t! add reduction2I" %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4977
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4978
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4979
    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x1);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4980
    __ vpaddd($tmp$$XMMRegister, $src2$$XMMRegister, $tmp2$$XMMRegister, vector_len);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4981
    __ movdl($tmp2$$XMMRegister, $src1$$Register);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  4982
    __ vpaddd($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4983
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4984
  %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4985
  ins_pipe( pipe_slow );
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4986
%}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4987
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  4988
instruct rsadd4I_reduction_reg(rRegI dst, rRegI src1, vecX src2, vecX tmp, vecX tmp2) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4989
  predicate(UseSSE > 2 && UseAVX == 0);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  4990
  match(Set dst (AddReductionVI src1 src2));
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  4991
  effect(TEMP tmp, TEMP tmp2);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  4992
  format %{ "movdqu  $tmp,$src2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  4993
            "phaddd  $tmp,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  4994
            "phaddd  $tmp,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  4995
            "movd    $tmp2,$src1\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  4996
            "paddd   $tmp2,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  4997
            "movd    $dst,$tmp2\t! add reduction4I" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  4998
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  4999
    __ movdqu($tmp$$XMMRegister, $src2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5000
    __ phaddd($tmp$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5001
    __ phaddd($tmp$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5002
    __ movdl($tmp2$$XMMRegister, $src1$$Register);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5003
    __ paddd($tmp2$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5004
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5005
  %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5006
  ins_pipe( pipe_slow );
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5007
%}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5008
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5009
instruct rvadd4I_reduction_reg(rRegI dst, rRegI src1, vecX src2, vecX tmp, vecX tmp2) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5010
  predicate(VM_Version::supports_avxonly());
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5011
  match(Set dst (AddReductionVI src1 src2));
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5012
  effect(TEMP tmp, TEMP tmp2);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5013
  format %{ "vphaddd  $tmp,$src2,$src2\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5014
            "vphaddd  $tmp,$tmp,$tmp\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5015
            "movd     $tmp2,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5016
            "vpaddd   $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5017
            "movd     $dst,$tmp2\t! add reduction4I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5018
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5019
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5020
    __ vphaddd($tmp$$XMMRegister, $src2$$XMMRegister, $src2$$XMMRegister, vector_len);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5021
    __ vphaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp$$XMMRegister, vector_len);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5022
    __ movdl($tmp2$$XMMRegister, $src1$$Register);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5023
    __ vpaddd($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5024
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5025
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5026
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5027
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5028
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5029
instruct rvadd4I_reduction_reg_evex(rRegI dst, rRegI src1, vecX src2, vecX tmp, vecX tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5030
  predicate(UseAVX > 2);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5031
  match(Set dst (AddReductionVI src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5032
  effect(TEMP tmp, TEMP tmp2);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5033
  format %{ "pshufd  $tmp2,$src2,0xE\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5034
            "vpaddd  $tmp,$src2,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5035
            "pshufd  $tmp2,$tmp,0x1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5036
            "vpaddd  $tmp,$tmp,$tmp2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5037
            "movd    $tmp2,$src1\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5038
            "vpaddd  $tmp2,$tmp,$tmp2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5039
            "movd    $dst,$tmp2\t! add reduction4I" %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5040
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5041
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5042
    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0xE);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5043
    __ vpaddd($tmp$$XMMRegister, $src2$$XMMRegister, $tmp2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5044
    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0x1);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5045
    __ vpaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5046
    __ movdl($tmp2$$XMMRegister, $src1$$Register);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5047
    __ vpaddd($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5048
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5049
  %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5050
  ins_pipe( pipe_slow );
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5051
%}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5052
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5053
instruct rvadd8I_reduction_reg(rRegI dst, rRegI src1, vecY src2, vecY tmp, vecY tmp2) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5054
  predicate(VM_Version::supports_avxonly());
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5055
  match(Set dst (AddReductionVI src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5056
  effect(TEMP tmp, TEMP tmp2);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5057
  format %{ "vphaddd  $tmp,$src2,$src2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5058
            "vphaddd  $tmp,$tmp,$tmp2\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5059
            "vextracti128_high  $tmp2,$tmp\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5060
            "vpaddd   $tmp,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5061
            "movd     $tmp2,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5062
            "vpaddd   $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5063
            "movd     $dst,$tmp2\t! add reduction8I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5064
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5065
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5066
    __ vphaddd($tmp$$XMMRegister, $src2$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5067
    __ vphaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5068
    __ vextracti128_high($tmp2$$XMMRegister, $tmp$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5069
    __ vpaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5070
    __ movdl($tmp2$$XMMRegister, $src1$$Register);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5071
    __ vpaddd($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5072
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5073
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5074
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5075
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5076
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5077
instruct rvadd8I_reduction_reg_evex(rRegI dst, rRegI src1, vecY src2, vecY tmp, vecY tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5078
  predicate(UseAVX > 2);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5079
  match(Set dst (AddReductionVI src1 src2));
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5080
  effect(TEMP tmp, TEMP tmp2);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5081
  format %{ "vextracti128_high  $tmp,$src2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5082
            "vpaddd  $tmp,$tmp,$src2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5083
            "pshufd  $tmp2,$tmp,0xE\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5084
            "vpaddd  $tmp,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5085
            "pshufd  $tmp2,$tmp,0x1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5086
            "vpaddd  $tmp,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5087
            "movd    $tmp2,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5088
            "vpaddd  $tmp2,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5089
            "movd    $dst,$tmp2\t! add reduction8I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5090
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5091
    int vector_len = 0;
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5092
    __ vextracti128_high($tmp$$XMMRegister, $src2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5093
    __ vpaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5094
    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0xE);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5095
    __ vpaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5096
    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0x1);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5097
    __ vpaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5098
    __ movdl($tmp2$$XMMRegister, $src1$$Register);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5099
    __ vpaddd($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5100
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5101
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5102
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5103
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5104
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5105
instruct rvadd16I_reduction_reg_evex(rRegI dst, rRegI src1, legVecZ src2, legVecZ tmp, legVecZ tmp2, legVecZ tmp3) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5106
  predicate(UseAVX > 2);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5107
  match(Set dst (AddReductionVI src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5108
  effect(TEMP tmp, TEMP tmp2, TEMP tmp3);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5109
  format %{ "vextracti64x4_high  $tmp3,$src2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5110
            "vpaddd  $tmp3,$tmp3,$src2\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5111
            "vextracti128_high  $tmp,$tmp3\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5112
            "vpaddd  $tmp,$tmp,$tmp3\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5113
            "pshufd  $tmp2,$tmp,0xE\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5114
            "vpaddd  $tmp,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5115
            "pshufd  $tmp2,$tmp,0x1\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5116
            "vpaddd  $tmp,$tmp,$tmp2\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5117
            "movd    $tmp2,$src1\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5118
            "vpaddd  $tmp2,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5119
            "movd    $dst,$tmp2\t! mul reduction16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5120
  ins_encode %{
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5121
    __ vextracti64x4_high($tmp3$$XMMRegister, $src2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5122
    __ vpaddd($tmp3$$XMMRegister, $tmp3$$XMMRegister, $src2$$XMMRegister, 1);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5123
    __ vextracti128_high($tmp$$XMMRegister, $tmp3$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5124
    __ vpaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp3$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5125
    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0xE);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5126
    __ vpaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5127
    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0x1);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5128
    __ vpaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, 0);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5129
    __ movdl($tmp2$$XMMRegister, $src1$$Register);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5130
    __ vpaddd($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, 0);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5131
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5132
  %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5133
  ins_pipe( pipe_slow );
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5134
%}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5135
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5136
#ifdef _LP64
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5137
instruct rvadd2L_reduction_reg(rRegL dst, rRegL src1, vecX src2, vecX tmp, vecX tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5138
  predicate(UseAVX > 2);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5139
  match(Set dst (AddReductionVL src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5140
  effect(TEMP tmp, TEMP tmp2);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5141
  format %{ "pshufd  $tmp2,$src2,0xE\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5142
            "vpaddq  $tmp,$src2,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5143
            "movdq   $tmp2,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5144
            "vpaddq  $tmp2,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5145
            "movdq   $dst,$tmp2\t! add reduction2L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5146
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5147
    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0xE);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5148
    __ vpaddq($tmp$$XMMRegister, $src2$$XMMRegister, $tmp2$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5149
    __ movdq($tmp2$$XMMRegister, $src1$$Register);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5150
    __ vpaddq($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5151
    __ movdq($dst$$Register, $tmp2$$XMMRegister);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5152
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5153
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5154
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5155
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5156
instruct rvadd4L_reduction_reg(rRegL dst, rRegL src1, vecY src2, vecY tmp, vecY tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5157
  predicate(UseAVX > 2);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5158
  match(Set dst (AddReductionVL src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5159
  effect(TEMP tmp, TEMP tmp2);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5160
  format %{ "vextracti128_high  $tmp,$src2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5161
            "vpaddq  $tmp2,$tmp,$src2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5162
            "pshufd  $tmp,$tmp2,0xE\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5163
            "vpaddq  $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5164
            "movdq   $tmp,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5165
            "vpaddq  $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5166
            "movdq   $dst,$tmp2\t! add reduction4L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5167
  ins_encode %{
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5168
    __ vextracti128_high($tmp$$XMMRegister, $src2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5169
    __ vpaddq($tmp2$$XMMRegister, $tmp$$XMMRegister, $src2$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5170
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0xE);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5171
    __ vpaddq($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5172
    __ movdq($tmp$$XMMRegister, $src1$$Register);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5173
    __ vpaddq($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5174
    __ movdq($dst$$Register, $tmp2$$XMMRegister);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5175
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5176
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5177
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5178
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5179
instruct rvadd8L_reduction_reg(rRegL dst, rRegL src1, legVecZ src2, legVecZ tmp, legVecZ tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5180
  predicate(UseAVX > 2);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5181
  match(Set dst (AddReductionVL src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5182
  effect(TEMP tmp, TEMP tmp2);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5183
  format %{ "vextracti64x4_high  $tmp2,$src2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5184
            "vpaddq  $tmp2,$tmp2,$src2\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5185
            "vextracti128_high  $tmp,$tmp2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5186
            "vpaddq  $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5187
            "pshufd  $tmp,$tmp2,0xE\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5188
            "vpaddq  $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5189
            "movdq   $tmp,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5190
            "vpaddq  $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5191
            "movdq   $dst,$tmp2\t! add reduction8L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5192
  ins_encode %{
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5193
    __ vextracti64x4_high($tmp2$$XMMRegister, $src2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5194
    __ vpaddq($tmp2$$XMMRegister, $tmp2$$XMMRegister, $src2$$XMMRegister, 1);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5195
    __ vextracti128_high($tmp$$XMMRegister, $tmp2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5196
    __ vpaddq($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5197
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0xE);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5198
    __ vpaddq($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5199
    __ movdq($tmp$$XMMRegister, $src1$$Register);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5200
    __ vpaddq($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5201
    __ movdq($dst$$Register, $tmp2$$XMMRegister);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5202
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5203
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5204
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5205
#endif
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5206
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5207
instruct rsadd2F_reduction_reg(regF dst, vecD src2, vecD tmp) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5208
  predicate(UseSSE >= 1 && UseAVX == 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5209
  match(Set dst (AddReductionVF dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5210
  effect(TEMP dst, TEMP tmp);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5211
  format %{ "addss   $dst,$src2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5212
            "pshufd  $tmp,$src2,0x01\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5213
            "addss   $dst,$tmp\t! add reduction2F" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5214
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5215
    __ addss($dst$$XMMRegister, $src2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5216
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5217
    __ addss($dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5218
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5219
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5220
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5221
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5222
instruct rvadd2F_reduction_reg(regF dst, vecD src2, vecD tmp) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5223
  predicate(UseAVX > 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5224
  match(Set dst (AddReductionVF dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5225
  effect(TEMP dst, TEMP tmp);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5226
  format %{ "vaddss  $dst,$dst,$src2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5227
            "pshufd  $tmp,$src2,0x01\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5228
            "vaddss  $dst,$dst,$tmp\t! add reduction2F" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5229
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5230
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5231
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5232
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5233
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5234
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5235
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5236
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5237
instruct rsadd4F_reduction_reg(regF dst, vecX src2, vecX tmp) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5238
  predicate(UseSSE >= 1 && UseAVX == 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5239
  match(Set dst (AddReductionVF dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5240
  effect(TEMP dst, TEMP tmp);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5241
  format %{ "addss   $dst,$src2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5242
            "pshufd  $tmp,$src2,0x01\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5243
            "addss   $dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5244
            "pshufd  $tmp,$src2,0x02\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5245
            "addss   $dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5246
            "pshufd  $tmp,$src2,0x03\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5247
            "addss   $dst,$tmp\t! add reduction4F" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5248
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5249
    __ addss($dst$$XMMRegister, $src2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5250
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5251
    __ addss($dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5252
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x02);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5253
    __ addss($dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5254
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x03);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5255
    __ addss($dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5256
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5257
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5258
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5259
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5260
instruct rvadd4F_reduction_reg(regF dst, vecX src2, vecX tmp) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5261
  predicate(UseAVX > 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5262
  match(Set dst (AddReductionVF dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5263
  effect(TEMP tmp, TEMP dst);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5264
  format %{ "vaddss  $dst,dst,$src2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5265
            "pshufd  $tmp,$src2,0x01\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5266
            "vaddss  $dst,$dst,$tmp\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5267
            "pshufd  $tmp,$src2,0x02\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5268
            "vaddss  $dst,$dst,$tmp\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5269
            "pshufd  $tmp,$src2,0x03\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5270
            "vaddss  $dst,$dst,$tmp\t! add reduction4F" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5271
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5272
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5273
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5274
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5275
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x02);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5276
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5277
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x03);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5278
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5279
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5280
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5281
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5282
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5283
instruct radd8F_reduction_reg(regF dst, vecY src2, vecY tmp, vecY tmp2) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5284
  predicate(UseAVX > 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5285
  match(Set dst (AddReductionVF dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5286
  effect(TEMP tmp, TEMP dst, TEMP tmp2);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5287
  format %{ "vaddss  $dst,$dst,$src2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5288
            "pshufd  $tmp,$src2,0x01\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5289
            "vaddss  $dst,$dst,$tmp\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5290
            "pshufd  $tmp,$src2,0x02\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5291
            "vaddss  $dst,$dst,$tmp\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5292
            "pshufd  $tmp,$src2,0x03\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5293
            "vaddss  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5294
            "vextractf128_high  $tmp2,$src2\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5295
            "vaddss  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5296
            "pshufd  $tmp,$tmp2,0x01\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5297
            "vaddss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5298
            "pshufd  $tmp,$tmp2,0x02\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5299
            "vaddss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5300
            "pshufd  $tmp,$tmp2,0x03\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5301
            "vaddss  $dst,$dst,$tmp\t! add reduction8F" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5302
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5303
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5304
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5305
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5306
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x02);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5307
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5308
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x03);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5309
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5310
    __ vextractf128_high($tmp2$$XMMRegister, $src2$$XMMRegister);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5311
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5312
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x01);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5313
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5314
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x02);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5315
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5316
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x03);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5317
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5318
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5319
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5320
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5321
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5322
instruct radd16F_reduction_reg(regF dst, legVecZ src2, legVecZ tmp, legVecZ tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5323
  predicate(UseAVX > 2);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5324
  match(Set dst (AddReductionVF dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5325
  effect(TEMP tmp, TEMP dst, TEMP tmp2);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5326
  format %{ "vaddss  $dst,$dst,$src2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5327
            "pshufd  $tmp,$src2,0x01\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5328
            "vaddss  $dst,$dst,$tmp\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5329
            "pshufd  $tmp,$src2,0x02\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5330
            "vaddss  $dst,$dst,$tmp\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5331
            "pshufd  $tmp,$src2,0x03\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5332
            "vaddss  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5333
            "vextractf32x4  $tmp2,$src2,0x1\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5334
            "vaddss  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5335
            "pshufd  $tmp,$tmp2,0x01\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5336
            "vaddss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5337
            "pshufd  $tmp,$tmp2,0x02\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5338
            "vaddss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5339
            "pshufd  $tmp,$tmp2,0x03\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5340
            "vaddss  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5341
            "vextractf32x4  $tmp2,$src2,0x2\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5342
            "vaddss  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5343
            "pshufd  $tmp,$tmp2,0x01\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5344
            "vaddss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5345
            "pshufd  $tmp,$tmp2,0x02\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5346
            "vaddss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5347
            "pshufd  $tmp,$tmp2,0x03\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5348
            "vaddss  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5349
            "vextractf32x4  $tmp2,$src2,0x3\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5350
            "vaddss  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5351
            "pshufd  $tmp,$tmp2,0x01\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5352
            "vaddss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5353
            "pshufd  $tmp,$tmp2,0x02\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5354
            "vaddss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5355
            "pshufd  $tmp,$tmp2,0x03\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5356
            "vaddss  $dst,$dst,$tmp\t! add reduction16F" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5357
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5358
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5359
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5360
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5361
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x02);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5362
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5363
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x03);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5364
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5365
    __ vextractf32x4($tmp2$$XMMRegister, $src2$$XMMRegister, 0x1);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5366
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5367
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x01);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5368
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5369
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x02);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5370
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5371
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x03);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5372
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5373
    __ vextractf32x4($tmp2$$XMMRegister, $src2$$XMMRegister, 0x2);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5374
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5375
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x01);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5376
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5377
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x02);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5378
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5379
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x03);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5380
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5381
    __ vextractf32x4($tmp2$$XMMRegister, $src2$$XMMRegister, 0x3);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5382
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5383
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x01);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5384
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5385
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x02);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5386
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5387
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x03);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5388
    __ vaddss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5389
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5390
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5391
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5392
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5393
instruct rsadd2D_reduction_reg(regD dst, vecX src2, vecX tmp) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5394
  predicate(UseSSE >= 1 && UseAVX == 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5395
  match(Set dst (AddReductionVD dst src2));
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5396
  effect(TEMP tmp, TEMP dst);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5397
  format %{ "addsd   $dst,$src2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5398
            "pshufd  $tmp,$src2,0xE\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5399
            "addsd   $dst,$tmp\t! add reduction2D" %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5400
  ins_encode %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5401
    __ addsd($dst$$XMMRegister, $src2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5402
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0xE);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5403
    __ addsd($dst$$XMMRegister, $tmp$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5404
  %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5405
  ins_pipe( pipe_slow );
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5406
%}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5407
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5408
instruct rvadd2D_reduction_reg(regD dst, vecX src2, vecX tmp) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5409
  predicate(UseAVX > 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5410
  match(Set dst (AddReductionVD dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5411
  effect(TEMP tmp, TEMP dst);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5412
  format %{ "vaddsd  $dst,$dst,$src2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5413
            "pshufd  $tmp,$src2,0xE\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5414
            "vaddsd  $dst,$dst,$tmp\t! add reduction2D" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5415
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5416
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5417
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0xE);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5418
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5419
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5420
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5421
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5422
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5423
instruct rvadd4D_reduction_reg(regD dst, vecY src2, vecX tmp, vecX tmp2) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5424
  predicate(UseAVX > 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5425
  match(Set dst (AddReductionVD dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5426
  effect(TEMP tmp, TEMP dst, TEMP tmp2);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5427
  format %{ "vaddsd  $dst,$dst,$src2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5428
            "pshufd  $tmp,$src2,0xE\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5429
            "vaddsd  $dst,$dst,$tmp\n\t"
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5430
            "vextractf128  $tmp2,$src2,0x1\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5431
            "vaddsd  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5432
            "pshufd  $tmp,$tmp2,0xE\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5433
            "vaddsd  $dst,$dst,$tmp\t! add reduction4D" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5434
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5435
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5436
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0xE);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5437
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5438
    __ vextractf128($tmp2$$XMMRegister, $src2$$XMMRegister, 0x1);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5439
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5440
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0xE);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5441
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5442
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5443
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5444
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5445
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5446
instruct rvadd8D_reduction_reg(regD dst, legVecZ src2, legVecZ tmp, legVecZ tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5447
  predicate(UseAVX > 2);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5448
  match(Set dst (AddReductionVD dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5449
  effect(TEMP tmp, TEMP dst, TEMP tmp2);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5450
  format %{ "vaddsd  $dst,$dst,$src2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5451
            "pshufd  $tmp,$src2,0xE\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5452
            "vaddsd  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5453
            "vextractf32x4  $tmp2,$src2,0x1\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5454
            "vaddsd  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5455
            "pshufd  $tmp,$tmp2,0xE\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5456
            "vaddsd  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5457
            "vextractf32x4  $tmp2,$src2,0x2\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5458
            "vaddsd  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5459
            "pshufd  $tmp,$tmp2,0xE\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5460
            "vaddsd  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5461
            "vextractf32x4  $tmp2,$src2,0x3\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5462
            "vaddsd  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5463
            "pshufd  $tmp,$tmp2,0xE\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5464
            "vaddsd  $dst,$dst,$tmp\t! add reduction8D" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5465
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5466
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5467
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0xE);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5468
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5469
    __ vextractf32x4($tmp2$$XMMRegister, $src2$$XMMRegister, 0x1);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5470
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5471
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0xE);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5472
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5473
    __ vextractf32x4($tmp2$$XMMRegister, $src2$$XMMRegister, 0x2);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5474
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5475
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0xE);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5476
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5477
    __ vextractf32x4($tmp2$$XMMRegister, $src2$$XMMRegister, 0x3);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5478
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5479
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0xE);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5480
    __ vaddsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5481
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5482
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5483
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5484
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5485
instruct rsmul2I_reduction_reg(rRegI dst, rRegI src1, vecD src2, vecD tmp, vecD tmp2) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5486
  predicate(UseSSE > 3 && UseAVX == 0);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5487
  match(Set dst (MulReductionVI src1 src2));
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5488
  effect(TEMP tmp, TEMP tmp2);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5489
  format %{ "pshufd  $tmp2,$src2,0x1\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5490
            "pmulld  $tmp2,$src2\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5491
            "movd    $tmp,$src1\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5492
            "pmulld  $tmp2,$tmp\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5493
            "movd    $dst,$tmp2\t! mul reduction2I" %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5494
  ins_encode %{
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5495
    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x1);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5496
    __ pmulld($tmp2$$XMMRegister, $src2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5497
    __ movdl($tmp$$XMMRegister, $src1$$Register);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5498
    __ pmulld($tmp2$$XMMRegister, $tmp$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5499
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5500
  %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5501
  ins_pipe( pipe_slow );
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5502
%}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5503
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5504
instruct rvmul2I_reduction_reg(rRegI dst, rRegI src1, vecD src2, vecD tmp, vecD tmp2) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5505
  predicate(UseAVX > 0);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5506
  match(Set dst (MulReductionVI src1 src2));
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5507
  effect(TEMP tmp, TEMP tmp2);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5508
  format %{ "pshufd   $tmp2,$src2,0x1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5509
            "vpmulld  $tmp,$src2,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5510
            "movd     $tmp2,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5511
            "vpmulld  $tmp2,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5512
            "movd     $dst,$tmp2\t! mul reduction2I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5513
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5514
    int vector_len = 0;
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5515
    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x1);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5516
    __ vpmulld($tmp$$XMMRegister, $src2$$XMMRegister, $tmp2$$XMMRegister, vector_len);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5517
    __ movdl($tmp2$$XMMRegister, $src1$$Register);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5518
    __ vpmulld($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5519
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5520
  %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5521
  ins_pipe( pipe_slow );
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5522
%}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5523
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5524
instruct rsmul4I_reduction_reg(rRegI dst, rRegI src1, vecX src2, vecX tmp, vecX tmp2) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5525
  predicate(UseSSE > 3 && UseAVX == 0);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5526
  match(Set dst (MulReductionVI src1 src2));
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5527
  effect(TEMP tmp, TEMP tmp2);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5528
  format %{ "pshufd  $tmp2,$src2,0xE\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5529
            "pmulld  $tmp2,$src2\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5530
            "pshufd  $tmp,$tmp2,0x1\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5531
            "pmulld  $tmp2,$tmp\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5532
            "movd    $tmp,$src1\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5533
            "pmulld  $tmp2,$tmp\n\t"
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5534
            "movd    $dst,$tmp2\t! mul reduction4I" %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5535
  ins_encode %{
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5536
    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0xE);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5537
    __ pmulld($tmp2$$XMMRegister, $src2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5538
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x1);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5539
    __ pmulld($tmp2$$XMMRegister, $tmp$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5540
    __ movdl($tmp$$XMMRegister, $src1$$Register);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5541
    __ pmulld($tmp2$$XMMRegister, $tmp$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5542
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5543
  %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5544
  ins_pipe( pipe_slow );
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5545
%}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5546
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5547
instruct rvmul4I_reduction_reg(rRegI dst, rRegI src1, vecX src2, vecX tmp, vecX tmp2) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5548
  predicate(UseAVX > 0);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5549
  match(Set dst (MulReductionVI src1 src2));
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5550
  effect(TEMP tmp, TEMP tmp2);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5551
  format %{ "pshufd   $tmp2,$src2,0xE\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5552
            "vpmulld  $tmp,$src2,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5553
            "pshufd   $tmp2,$tmp,0x1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5554
            "vpmulld  $tmp,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5555
            "movd     $tmp2,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5556
            "vpmulld  $tmp2,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5557
            "movd     $dst,$tmp2\t! mul reduction4I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5558
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5559
    int vector_len = 0;
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5560
    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0xE);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5561
    __ vpmulld($tmp$$XMMRegister, $src2$$XMMRegister, $tmp2$$XMMRegister, vector_len);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5562
    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0x1);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5563
    __ vpmulld($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5564
    __ movdl($tmp2$$XMMRegister, $src1$$Register);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5565
    __ vpmulld($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5566
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5567
  %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5568
  ins_pipe( pipe_slow );
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5569
%}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5570
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5571
instruct rvmul8I_reduction_reg(rRegI dst, rRegI src1, vecY src2, vecY tmp, vecY tmp2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5572
  predicate(UseAVX > 1);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5573
  match(Set dst (MulReductionVI src1 src2));
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5574
  effect(TEMP tmp, TEMP tmp2);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5575
  format %{ "vextracti128_high  $tmp,$src2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5576
            "vpmulld  $tmp,$tmp,$src2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5577
            "pshufd   $tmp2,$tmp,0xE\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5578
            "vpmulld  $tmp,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5579
            "pshufd   $tmp2,$tmp,0x1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5580
            "vpmulld  $tmp,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5581
            "movd     $tmp2,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5582
            "vpmulld  $tmp2,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5583
            "movd     $dst,$tmp2\t! mul reduction8I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5584
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5585
    int vector_len = 0;
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5586
    __ vextracti128_high($tmp$$XMMRegister, $src2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5587
    __ vpmulld($tmp$$XMMRegister, $tmp$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5588
    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0xE);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5589
    __ vpmulld($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5590
    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0x1);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5591
    __ vpmulld($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5592
    __ movdl($tmp2$$XMMRegister, $src1$$Register);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5593
    __ vpmulld($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5594
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5595
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5596
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5597
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5598
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5599
instruct rvmul16I_reduction_reg(rRegI dst, rRegI src1, legVecZ src2, legVecZ tmp, legVecZ tmp2, legVecZ tmp3) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5600
  predicate(UseAVX > 2);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5601
  match(Set dst (MulReductionVI src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5602
  effect(TEMP tmp, TEMP tmp2, TEMP tmp3);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5603
  format %{ "vextracti64x4_high  $tmp3,$src2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5604
            "vpmulld  $tmp3,$tmp3,$src2\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5605
            "vextracti128_high  $tmp,$tmp3\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5606
            "vpmulld  $tmp,$tmp,$src2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5607
            "pshufd   $tmp2,$tmp,0xE\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5608
            "vpmulld  $tmp,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5609
            "pshufd   $tmp2,$tmp,0x1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5610
            "vpmulld  $tmp,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5611
            "movd     $tmp2,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5612
            "vpmulld  $tmp2,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5613
            "movd     $dst,$tmp2\t! mul reduction16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5614
  ins_encode %{
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5615
    __ vextracti64x4_high($tmp3$$XMMRegister, $src2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5616
    __ vpmulld($tmp3$$XMMRegister, $tmp3$$XMMRegister, $src2$$XMMRegister, 1);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5617
    __ vextracti128_high($tmp$$XMMRegister, $tmp3$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5618
    __ vpmulld($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp3$$XMMRegister, 0);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5619
    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0xE);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5620
    __ vpmulld($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, 0);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5621
    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0x1);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5622
    __ vpmulld($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, 0);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5623
    __ movdl($tmp2$$XMMRegister, $src1$$Register);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5624
    __ vpmulld($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, 0);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5625
    __ movdl($dst$$Register, $tmp2$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5626
  %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5627
  ins_pipe( pipe_slow );
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5628
%}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5629
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5630
#ifdef _LP64
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5631
instruct rvmul2L_reduction_reg(rRegL dst, rRegL src1, vecX src2, vecX tmp, vecX tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5632
  predicate(UseAVX > 2 && VM_Version::supports_avx512dq());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5633
  match(Set dst (MulReductionVL src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5634
  effect(TEMP tmp, TEMP tmp2);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5635
  format %{ "pshufd   $tmp2,$src2,0xE\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5636
            "vpmullq  $tmp,$src2,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5637
            "movdq    $tmp2,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5638
            "vpmullq  $tmp2,$tmp,$tmp2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5639
            "movdq    $dst,$tmp2\t! mul reduction2L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5640
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5641
    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0xE);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5642
    __ vpmullq($tmp$$XMMRegister, $src2$$XMMRegister, $tmp2$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5643
    __ movdq($tmp2$$XMMRegister, $src1$$Register);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5644
    __ vpmullq($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5645
    __ movdq($dst$$Register, $tmp2$$XMMRegister);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5646
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5647
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5648
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5649
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5650
instruct rvmul4L_reduction_reg(rRegL dst, rRegL src1, vecY src2, vecY tmp, vecY tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5651
  predicate(UseAVX > 2 && VM_Version::supports_avx512dq());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5652
  match(Set dst (MulReductionVL src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5653
  effect(TEMP tmp, TEMP tmp2);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5654
  format %{ "vextracti128_high  $tmp,$src2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5655
            "vpmullq  $tmp2,$tmp,$src2\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5656
            "pshufd   $tmp,$tmp2,0xE\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5657
            "vpmullq  $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5658
            "movdq    $tmp,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5659
            "vpmullq  $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5660
            "movdq    $dst,$tmp2\t! mul reduction4L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5661
  ins_encode %{
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5662
    __ vextracti128_high($tmp$$XMMRegister, $src2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5663
    __ vpmullq($tmp2$$XMMRegister, $tmp$$XMMRegister, $src2$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5664
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0xE);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5665
    __ vpmullq($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5666
    __ movdq($tmp$$XMMRegister, $src1$$Register);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5667
    __ vpmullq($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5668
    __ movdq($dst$$Register, $tmp2$$XMMRegister);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5669
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5670
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5671
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5672
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5673
instruct rvmul8L_reduction_reg(rRegL dst, rRegL src1, legVecZ src2, legVecZ tmp, legVecZ tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5674
  predicate(UseAVX > 2 && VM_Version::supports_avx512dq());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5675
  match(Set dst (MulReductionVL src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5676
  effect(TEMP tmp, TEMP tmp2);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5677
  format %{ "vextracti64x4_high  $tmp2,$src2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5678
            "vpmullq  $tmp2,$tmp2,$src2\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5679
            "vextracti128_high  $tmp,$tmp2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5680
            "vpmullq  $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5681
            "pshufd   $tmp,$tmp2,0xE\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5682
            "vpmullq  $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5683
            "movdq    $tmp,$src1\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5684
            "vpmullq  $tmp2,$tmp2,$tmp\n\t"
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5685
            "movdq    $dst,$tmp2\t! mul reduction8L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5686
  ins_encode %{
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5687
    __ vextracti64x4_high($tmp2$$XMMRegister, $src2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5688
    __ vpmullq($tmp2$$XMMRegister, $tmp2$$XMMRegister, $src2$$XMMRegister, 1);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5689
    __ vextracti128_high($tmp$$XMMRegister, $tmp2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5690
    __ vpmullq($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5691
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0xE);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5692
    __ vpmullq($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5693
    __ movdq($tmp$$XMMRegister, $src1$$Register);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5694
    __ vpmullq($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, 0);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5695
    __ movdq($dst$$Register, $tmp2$$XMMRegister);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5696
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5697
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5698
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5699
#endif
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5700
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5701
instruct rsmul2F_reduction(regF dst, vecD src2, vecD tmp) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5702
  predicate(UseSSE >= 1 && UseAVX == 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5703
  match(Set dst (MulReductionVF dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5704
  effect(TEMP dst, TEMP tmp);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5705
  format %{ "mulss   $dst,$src2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5706
            "pshufd  $tmp,$src2,0x01\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5707
            "mulss   $dst,$tmp\t! mul reduction2F" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5708
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5709
    __ mulss($dst$$XMMRegister, $src2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5710
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5711
    __ mulss($dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5712
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5713
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5714
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5715
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5716
instruct rvmul2F_reduction_reg(regF dst, vecD src2, vecD tmp) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5717
  predicate(UseAVX > 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5718
  match(Set dst (MulReductionVF dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5719
  effect(TEMP tmp, TEMP dst);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5720
  format %{ "vmulss  $dst,$dst,$src2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5721
            "pshufd  $tmp,$src2,0x01\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5722
            "vmulss  $dst,$dst,$tmp\t! mul reduction2F" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5723
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5724
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5725
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5726
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5727
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5728
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5729
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5730
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5731
instruct rsmul4F_reduction_reg(regF dst, vecX src2, vecX tmp) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5732
  predicate(UseSSE >= 1 && UseAVX == 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5733
  match(Set dst (MulReductionVF dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5734
  effect(TEMP dst, TEMP tmp);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5735
  format %{ "mulss   $dst,$src2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5736
            "pshufd  $tmp,$src2,0x01\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5737
            "mulss   $dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5738
            "pshufd  $tmp,$src2,0x02\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5739
            "mulss   $dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5740
            "pshufd  $tmp,$src2,0x03\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5741
            "mulss   $dst,$tmp\t! mul reduction4F" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5742
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5743
    __ mulss($dst$$XMMRegister, $src2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5744
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5745
    __ mulss($dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5746
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x02);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5747
    __ mulss($dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5748
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x03);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5749
    __ mulss($dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5750
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5751
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5752
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5753
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5754
instruct rvmul4F_reduction_reg(regF dst, vecX src2, vecX tmp) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5755
  predicate(UseAVX > 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5756
  match(Set dst (MulReductionVF dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5757
  effect(TEMP tmp, TEMP dst);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5758
  format %{ "vmulss  $dst,$dst,$src2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5759
            "pshufd  $tmp,$src2,0x01\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5760
            "vmulss  $dst,$dst,$tmp\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5761
            "pshufd  $tmp,$src2,0x02\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5762
            "vmulss  $dst,$dst,$tmp\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5763
            "pshufd  $tmp,$src2,0x03\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5764
            "vmulss  $dst,$dst,$tmp\t! mul reduction4F" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5765
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5766
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5767
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5768
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5769
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x02);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5770
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5771
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x03);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5772
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5773
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5774
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5775
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5776
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5777
instruct rvmul8F_reduction_reg(regF dst, vecY src2, vecY tmp, vecY tmp2) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5778
  predicate(UseAVX > 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5779
  match(Set dst (MulReductionVF dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5780
  effect(TEMP tmp, TEMP dst, TEMP tmp2);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5781
  format %{ "vmulss  $dst,$dst,$src2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5782
            "pshufd  $tmp,$src2,0x01\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5783
            "vmulss  $dst,$dst,$tmp\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5784
            "pshufd  $tmp,$src2,0x02\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5785
            "vmulss  $dst,$dst,$tmp\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5786
            "pshufd  $tmp,$src2,0x03\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5787
            "vmulss  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5788
            "vextractf128_high  $tmp2,$src2\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5789
            "vmulss  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5790
            "pshufd  $tmp,$tmp2,0x01\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5791
            "vmulss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5792
            "pshufd  $tmp,$tmp2,0x02\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5793
            "vmulss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5794
            "pshufd  $tmp,$tmp2,0x03\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5795
            "vmulss  $dst,$dst,$tmp\t! mul reduction8F" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5796
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5797
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5798
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5799
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5800
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x02);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5801
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5802
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x03);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5803
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5804
    __ vextractf128_high($tmp2$$XMMRegister, $src2$$XMMRegister);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5805
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5806
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x01);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5807
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5808
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x02);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5809
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5810
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x03);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5811
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5812
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5813
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5814
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5815
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5816
instruct rvmul16F_reduction_reg(regF dst, legVecZ src2, legVecZ tmp, legVecZ tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5817
  predicate(UseAVX > 2);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5818
  match(Set dst (MulReductionVF dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5819
  effect(TEMP tmp, TEMP dst, TEMP tmp2);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5820
  format %{ "vmulss  $dst,$dst,$src2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5821
            "pshufd  $tmp,$src2,0x01\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5822
            "vmulss  $dst,$dst,$tmp\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5823
            "pshufd  $tmp,$src2,0x02\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5824
            "vmulss  $dst,$dst,$tmp\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5825
            "pshufd  $tmp,$src2,0x03\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5826
            "vmulss  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5827
            "vextractf32x4  $tmp2,$src2,0x1\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5828
            "vmulss  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5829
            "pshufd  $tmp,$tmp2,0x01\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5830
            "vmulss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5831
            "pshufd  $tmp,$tmp2,0x02\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5832
            "vmulss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5833
            "pshufd  $tmp,$tmp2,0x03\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5834
            "vmulss  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5835
            "vextractf32x4  $tmp2,$src2,0x2\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5836
            "vmulss  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5837
            "pshufd  $tmp,$tmp2,0x01\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5838
            "vmulss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5839
            "pshufd  $tmp,$tmp2,0x02\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5840
            "vmulss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5841
            "pshufd  $tmp,$tmp2,0x03\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5842
            "vmulss  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5843
            "vextractf32x4  $tmp2,$src2,0x3\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5844
            "vmulss  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5845
            "pshufd  $tmp,$tmp2,0x01\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5846
            "vmulss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5847
            "pshufd  $tmp,$tmp2,0x02\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5848
            "vmulss  $dst,$dst,$tmp\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5849
            "pshufd  $tmp,$tmp2,0x03\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5850
            "vmulss  $dst,$dst,$tmp\t! mul reduction16F" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5851
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5852
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5853
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5854
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5855
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x02);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5856
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5857
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x03);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5858
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5859
    __ vextractf32x4($tmp2$$XMMRegister, $src2$$XMMRegister, 0x1);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5860
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5861
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x01);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5862
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5863
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x02);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5864
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5865
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x03);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5866
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5867
    __ vextractf32x4($tmp2$$XMMRegister, $src2$$XMMRegister, 0x2);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5868
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5869
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x01);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5870
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5871
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x02);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5872
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5873
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x03);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5874
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5875
    __ vextractf32x4($tmp2$$XMMRegister, $src2$$XMMRegister, 0x3);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5876
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5877
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x01);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5878
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5879
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x02);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5880
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5881
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x03);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5882
    __ vmulss($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5883
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5884
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5885
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5886
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5887
instruct rsmul2D_reduction_reg(regD dst, vecX src2, vecX tmp) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5888
  predicate(UseSSE >= 1 && UseAVX == 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5889
  match(Set dst (MulReductionVD dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5890
  effect(TEMP dst, TEMP tmp);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5891
  format %{ "mulsd   $dst,$src2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5892
            "pshufd  $tmp,$src2,0xE\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5893
            "mulsd   $dst,$tmp\t! mul reduction2D" %}
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5894
  ins_encode %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5895
    __ mulsd($dst$$XMMRegister, $src2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5896
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0xE);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5897
    __ mulsd($dst$$XMMRegister, $tmp$$XMMRegister);
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5898
  %}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5899
  ins_pipe( pipe_slow );
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5900
%}
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5901
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5902
instruct rvmul2D_reduction_reg(regD dst, vecX src2, vecX tmp) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5903
  predicate(UseAVX > 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5904
  match(Set dst (MulReductionVD dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5905
  effect(TEMP tmp, TEMP dst);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5906
  format %{ "vmulsd  $dst,$dst,$src2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5907
            "pshufd  $tmp,$src2,0xE\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5908
            "vmulsd  $dst,$dst,$tmp\t! mul reduction2D" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5909
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5910
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5911
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0xE);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5912
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5913
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5914
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5915
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5916
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5917
instruct rvmul4D_reduction_reg(regD dst, vecY src2, vecY tmp, vecY tmp2) %{
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5918
  predicate(UseAVX > 0);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5919
  match(Set dst (MulReductionVD dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5920
  effect(TEMP tmp, TEMP dst, TEMP tmp2);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5921
  format %{ "vmulsd  $dst,$dst,$src2\n\t"
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5922
            "pshufd  $tmp,$src2,0xE\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5923
            "vmulsd  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5924
            "vextractf128_high  $tmp2,$src2\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5925
            "vmulsd  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5926
            "pshufd  $tmp,$tmp2,0xE\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5927
            "vmulsd  $dst,$dst,$tmp\t! mul reduction4D" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5928
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5929
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30211
442fbbb31f75 8074981: Integer/FP scalar reduction optimization
kvn
parents: 25715
diff changeset
  5930
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0xE);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5931
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5932
    __ vextractf128_high($tmp2$$XMMRegister, $src2$$XMMRegister);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5933
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5934
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0xE);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5935
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5936
  %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5937
  ins_pipe( pipe_slow );
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5938
%}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5939
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5940
instruct rvmul8D_reduction_reg(regD dst, legVecZ src2, legVecZ tmp, legVecZ tmp2) %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5941
  predicate(UseAVX > 2);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5942
  match(Set dst (MulReductionVD dst src2));
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5943
  effect(TEMP tmp, TEMP dst, TEMP tmp2);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5944
  format %{ "vmulsd  $dst,$dst,$src2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5945
            "pshufd  $tmp,$src2,0xE\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5946
            "vmulsd  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5947
            "vextractf32x4  $tmp2,$src2,0x1\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5948
            "vmulsd  $dst,$dst,$tmp2\n\t"
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5949
            "pshufd  $tmp,$src2,0xE\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5950
            "vmulsd  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5951
            "vextractf32x4  $tmp2,$src2,0x2\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5952
            "vmulsd  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5953
            "pshufd  $tmp,$tmp2,0xE\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5954
            "vmulsd  $dst,$dst,$tmp\n\t"
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5955
            "vextractf32x4  $tmp2,$src2,0x3\n\t"
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5956
            "vmulsd  $dst,$dst,$tmp2\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5957
            "pshufd  $tmp,$tmp2,0xE\n\t"
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5958
            "vmulsd  $dst,$dst,$tmp\t! mul reduction8D" %}
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5959
  ins_encode %{
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5960
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $src2$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5961
    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0xE);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5962
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5963
    __ vextractf32x4($tmp2$$XMMRegister, $src2$$XMMRegister, 0x1);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5964
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5965
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0xE);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5966
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5967
    __ vextractf32x4($tmp2$$XMMRegister, $src2$$XMMRegister, 0x2);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5968
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5969
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0xE);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5970
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
36561
b18243f4d955 8151002: Make Assembler methods vextract and vinsert match actual instructions
mikael
parents: 35581
diff changeset
  5971
    __ vextractf32x4($tmp2$$XMMRegister, $src2$$XMMRegister, 0x3);
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5972
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp2$$XMMRegister);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5973
    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0xE);
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5974
    __ vmulsd($dst$$XMMRegister, $dst$$XMMRegister, $tmp$$XMMRegister);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5975
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5976
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5977
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5978
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5979
// ====================VECTOR ARITHMETIC=======================================
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5980
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5981
// --------------------------------- ADD --------------------------------------
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5982
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5983
// Bytes vector add
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5984
instruct vadd4B(vecS dst, vecS src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  5985
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5986
  match(Set dst (AddVB dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5987
  format %{ "paddb   $dst,$src\t! add packed4B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5988
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5989
    __ paddb($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5990
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5991
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5992
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5993
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5994
instruct vadd4B_reg(vecS dst, vecS src1, vecS src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  5995
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5996
  match(Set dst (AddVB src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5997
  format %{ "vpaddb  $dst,$src1,$src2\t! add packed4B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  5998
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  5999
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6000
    __ vpaddb($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6001
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6002
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6003
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6004
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6005
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6006
instruct vadd4B_mem(vecS dst, vecS src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6007
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6008
  match(Set dst (AddVB src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6009
  format %{ "vpaddb  $dst,$src,$mem\t! add packed4B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6010
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6011
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6012
    __ vpaddb($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6013
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6014
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6015
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6016
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6017
instruct vadd8B(vecD dst, vecD src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  6018
  predicate(UseAVX == 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6019
  match(Set dst (AddVB dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6020
  format %{ "paddb   $dst,$src\t! add packed8B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6021
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6022
    __ paddb($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6023
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6024
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6025
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6026
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6027
instruct vadd8B_reg(vecD dst, vecD src1, vecD src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6028
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6029
  match(Set dst (AddVB src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6030
  format %{ "vpaddb  $dst,$src1,$src2\t! add packed8B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6031
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6032
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6033
    __ vpaddb($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6034
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6035
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6036
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6037
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6038
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6039
instruct vadd8B_mem(vecD dst, vecD src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6040
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6041
  match(Set dst (AddVB src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6042
  format %{ "vpaddb  $dst,$src,$mem\t! add packed8B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6043
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6044
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6045
    __ vpaddb($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6046
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6047
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6048
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6049
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6050
instruct vadd16B(vecX dst, vecX src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  6051
  predicate(UseAVX == 0 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6052
  match(Set dst (AddVB dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6053
  format %{ "paddb   $dst,$src\t! add packed16B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6054
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6055
    __ paddb($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6056
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6057
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6058
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6059
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6060
instruct vadd16B_reg(vecX dst, vecX src1, vecX src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6061
  predicate(UseAVX > 0  && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6062
  match(Set dst (AddVB src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6063
  format %{ "vpaddb  $dst,$src1,$src2\t! add packed16B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6064
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6065
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6066
    __ vpaddb($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6067
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6068
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6069
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6070
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6071
instruct vadd16B_mem(vecX dst, vecX src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6072
  predicate(UseAVX > 0 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6073
  match(Set dst (AddVB src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6074
  format %{ "vpaddb  $dst,$src,$mem\t! add packed16B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6075
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6076
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6077
    __ vpaddb($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6078
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6079
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6080
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6081
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6082
instruct vadd32B_reg(vecY dst, vecY src1, vecY src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6083
  predicate(UseAVX > 1 && n->as_Vector()->length() == 32);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6084
  match(Set dst (AddVB src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6085
  format %{ "vpaddb  $dst,$src1,$src2\t! add packed32B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6086
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6087
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6088
    __ vpaddb($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6089
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6090
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6091
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6092
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6093
instruct vadd32B_mem(vecY dst, vecY src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6094
  predicate(UseAVX > 1 && n->as_Vector()->length() == 32);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6095
  match(Set dst (AddVB src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6096
  format %{ "vpaddb  $dst,$src,$mem\t! add packed32B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6097
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6098
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6099
    __ vpaddb($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6100
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6101
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6102
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6103
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6104
instruct vadd64B_reg(vecZ dst, vecZ src1, vecZ src2) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6105
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 64);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6106
  match(Set dst (AddVB src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6107
  format %{ "vpaddb  $dst,$src1,$src2\t! add packed64B" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6108
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6109
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6110
    __ vpaddb($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6111
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6112
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6113
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6114
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6115
instruct vadd64B_mem(vecZ dst, vecZ src, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6116
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 64);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6117
  match(Set dst (AddVB src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6118
  format %{ "vpaddb  $dst,$src,$mem\t! add packed64B" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6119
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6120
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6121
    __ vpaddb($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6122
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6123
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6124
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6125
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6126
// Shorts/Chars vector add
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6127
instruct vadd2S(vecS dst, vecS src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  6128
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6129
  match(Set dst (AddVS dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6130
  format %{ "paddw   $dst,$src\t! add packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6131
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6132
    __ paddw($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6133
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6134
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6135
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6136
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6137
instruct vadd2S_reg(vecS dst, vecS src1, vecS src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6138
  predicate(UseAVX > 0  && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6139
  match(Set dst (AddVS src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6140
  format %{ "vpaddw  $dst,$src1,$src2\t! add packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6141
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6142
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6143
    __ vpaddw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6144
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6145
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6146
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6147
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6148
instruct vadd2S_mem(vecS dst, vecS src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6149
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6150
  match(Set dst (AddVS src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6151
  format %{ "vpaddw  $dst,$src,$mem\t! add packed2S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6152
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6153
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6154
    __ vpaddw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6155
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6156
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6157
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6158
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6159
instruct vadd4S(vecD dst, vecD src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  6160
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6161
  match(Set dst (AddVS dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6162
  format %{ "paddw   $dst,$src\t! add packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6163
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6164
    __ paddw($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6165
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6166
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6167
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6168
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6169
instruct vadd4S_reg(vecD dst, vecD src1, vecD src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6170
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6171
  match(Set dst (AddVS src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6172
  format %{ "vpaddw  $dst,$src1,$src2\t! add packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6173
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6174
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6175
    __ vpaddw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6176
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6177
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6178
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6179
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6180
instruct vadd4S_mem(vecD dst, vecD src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6181
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6182
  match(Set dst (AddVS src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6183
  format %{ "vpaddw  $dst,$src,$mem\t! add packed4S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6184
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6185
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6186
    __ vpaddw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6187
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6188
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6189
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6190
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6191
instruct vadd8S(vecX dst, vecX src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  6192
  predicate(UseAVX == 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6193
  match(Set dst (AddVS dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6194
  format %{ "paddw   $dst,$src\t! add packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6195
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6196
    __ paddw($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6197
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6198
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6199
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6200
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6201
instruct vadd8S_reg(vecX dst, vecX src1, vecX src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6202
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6203
  match(Set dst (AddVS src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6204
  format %{ "vpaddw  $dst,$src1,$src2\t! add packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6205
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6206
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6207
    __ vpaddw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6208
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6209
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6210
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6211
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6212
instruct vadd8S_mem(vecX dst, vecX src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6213
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6214
  match(Set dst (AddVS src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6215
  format %{ "vpaddw  $dst,$src,$mem\t! add packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6216
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6217
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6218
    __ vpaddw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6219
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6220
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6221
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6222
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6223
instruct vadd16S_reg(vecY dst, vecY src1, vecY src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6224
  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6225
  match(Set dst (AddVS src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6226
  format %{ "vpaddw  $dst,$src1,$src2\t! add packed16S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6227
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6228
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6229
    __ vpaddw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6230
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6231
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6232
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6233
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6234
instruct vadd16S_mem(vecY dst, vecY src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6235
  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6236
  match(Set dst (AddVS src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6237
  format %{ "vpaddw  $dst,$src,$mem\t! add packed16S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6238
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6239
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6240
    __ vpaddw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6241
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6242
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6243
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6244
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6245
instruct vadd32S_reg(vecZ dst, vecZ src1, vecZ src2) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6246
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 32);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6247
  match(Set dst (AddVS src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6248
  format %{ "vpaddw  $dst,$src1,$src2\t! add packed32S" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6249
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6250
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6251
    __ vpaddw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6252
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6253
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6254
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6255
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6256
instruct vadd32S_mem(vecZ dst, vecZ src, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6257
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 32);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6258
  match(Set dst (AddVS src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6259
  format %{ "vpaddw  $dst,$src,$mem\t! add packed32S" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6260
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6261
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6262
    __ vpaddw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6263
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6264
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6265
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6266
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6267
// Integers vector add
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6268
instruct vadd2I(vecD dst, vecD src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6269
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6270
  match(Set dst (AddVI dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6271
  format %{ "paddd   $dst,$src\t! add packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6272
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6273
    __ paddd($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6274
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6275
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6276
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6277
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6278
instruct vadd2I_reg(vecD dst, vecD src1, vecD src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6279
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6280
  match(Set dst (AddVI src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6281
  format %{ "vpaddd  $dst,$src1,$src2\t! add packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6282
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6283
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6284
    __ vpaddd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6285
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6286
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6287
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6288
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6289
instruct vadd2I_mem(vecD dst, vecD src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6290
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6291
  match(Set dst (AddVI src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6292
  format %{ "vpaddd  $dst,$src,$mem\t! add packed2I" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6293
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6294
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6295
    __ vpaddd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6296
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6297
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6298
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6299
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6300
instruct vadd4I(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6301
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6302
  match(Set dst (AddVI dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6303
  format %{ "paddd   $dst,$src\t! add packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6304
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6305
    __ paddd($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6306
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6307
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6308
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6309
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6310
instruct vadd4I_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6311
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6312
  match(Set dst (AddVI src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6313
  format %{ "vpaddd  $dst,$src1,$src2\t! add packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6314
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6315
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6316
    __ vpaddd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6317
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6318
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6319
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6320
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6321
instruct vadd4I_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6322
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6323
  match(Set dst (AddVI src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6324
  format %{ "vpaddd  $dst,$src,$mem\t! add packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6325
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6326
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6327
    __ vpaddd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6328
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6329
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6330
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6331
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6332
instruct vadd8I_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6333
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6334
  match(Set dst (AddVI src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6335
  format %{ "vpaddd  $dst,$src1,$src2\t! add packed8I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6336
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6337
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6338
    __ vpaddd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6339
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6340
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6341
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6342
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6343
instruct vadd8I_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6344
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6345
  match(Set dst (AddVI src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6346
  format %{ "vpaddd  $dst,$src,$mem\t! add packed8I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6347
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6348
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6349
    __ vpaddd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6350
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6351
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6352
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6353
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6354
instruct vadd16I_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6355
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6356
  match(Set dst (AddVI src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6357
  format %{ "vpaddd  $dst,$src1,$src2\t! add packed16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6358
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6359
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6360
    __ vpaddd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6361
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6362
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6363
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6364
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6365
instruct vadd16I_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6366
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6367
  match(Set dst (AddVI src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6368
  format %{ "vpaddd  $dst,$src,$mem\t! add packed16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6369
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6370
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6371
    __ vpaddd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6372
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6373
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6374
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6375
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6376
// Longs vector add
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6377
instruct vadd2L(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6378
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6379
  match(Set dst (AddVL dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6380
  format %{ "paddq   $dst,$src\t! add packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6381
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6382
    __ paddq($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6383
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6384
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6385
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6386
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6387
instruct vadd2L_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6388
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6389
  match(Set dst (AddVL src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6390
  format %{ "vpaddq  $dst,$src1,$src2\t! add packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6391
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6392
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6393
    __ vpaddq($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6394
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6395
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6396
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6397
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6398
instruct vadd2L_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6399
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6400
  match(Set dst (AddVL src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6401
  format %{ "vpaddq  $dst,$src,$mem\t! add packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6402
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6403
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6404
    __ vpaddq($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6405
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6406
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6407
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6408
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6409
instruct vadd4L_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6410
  predicate(UseAVX > 1 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6411
  match(Set dst (AddVL src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6412
  format %{ "vpaddq  $dst,$src1,$src2\t! add packed4L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6413
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6414
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6415
    __ vpaddq($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6416
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6417
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6418
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6419
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6420
instruct vadd4L_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6421
  predicate(UseAVX > 1 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6422
  match(Set dst (AddVL src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6423
  format %{ "vpaddq  $dst,$src,$mem\t! add packed4L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6424
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6425
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6426
    __ vpaddq($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6427
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6428
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6429
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6430
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6431
instruct vadd8L_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6432
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6433
  match(Set dst (AddVL src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6434
  format %{ "vpaddq  $dst,$src1,$src2\t! add packed8L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6435
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6436
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6437
    __ vpaddq($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6438
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6439
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6440
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6441
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6442
instruct vadd8L_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6443
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6444
  match(Set dst (AddVL src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6445
  format %{ "vpaddq  $dst,$src,$mem\t! add packed8L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6446
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6447
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6448
    __ vpaddq($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6449
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6450
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6451
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6452
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6453
// Floats vector add
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6454
instruct vadd2F(vecD dst, vecD src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6455
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6456
  match(Set dst (AddVF dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6457
  format %{ "addps   $dst,$src\t! add packed2F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6458
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6459
    __ addps($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6460
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6461
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6462
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6463
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6464
instruct vadd2F_reg(vecD dst, vecD src1, vecD src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6465
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6466
  match(Set dst (AddVF src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6467
  format %{ "vaddps  $dst,$src1,$src2\t! add packed2F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6468
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6469
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6470
    __ vaddps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6471
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6472
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6473
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6474
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6475
instruct vadd2F_mem(vecD dst, vecD src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6476
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6477
  match(Set dst (AddVF src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6478
  format %{ "vaddps  $dst,$src,$mem\t! add packed2F" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6479
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6480
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6481
    __ vaddps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6482
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6483
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6484
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6485
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6486
instruct vadd4F(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6487
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6488
  match(Set dst (AddVF dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6489
  format %{ "addps   $dst,$src\t! add packed4F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6490
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6491
    __ addps($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6492
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6493
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6494
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6495
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6496
instruct vadd4F_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6497
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6498
  match(Set dst (AddVF src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6499
  format %{ "vaddps  $dst,$src1,$src2\t! add packed4F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6500
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6501
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6502
    __ vaddps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6503
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6504
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6505
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6506
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6507
instruct vadd4F_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6508
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6509
  match(Set dst (AddVF src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6510
  format %{ "vaddps  $dst,$src,$mem\t! add packed4F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6511
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6512
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6513
    __ vaddps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6514
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6515
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6516
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6517
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6518
instruct vadd8F_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6519
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6520
  match(Set dst (AddVF src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6521
  format %{ "vaddps  $dst,$src1,$src2\t! add packed8F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6522
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6523
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6524
    __ vaddps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6525
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6526
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6527
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6528
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6529
instruct vadd8F_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6530
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6531
  match(Set dst (AddVF src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6532
  format %{ "vaddps  $dst,$src,$mem\t! add packed8F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6533
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6534
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6535
    __ vaddps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6536
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6537
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6538
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6539
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6540
instruct vadd16F_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6541
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6542
  match(Set dst (AddVF src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6543
  format %{ "vaddps  $dst,$src1,$src2\t! add packed16F" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6544
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6545
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6546
    __ vaddps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6547
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6548
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6549
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6550
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6551
instruct vadd16F_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6552
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6553
  match(Set dst (AddVF src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6554
  format %{ "vaddps  $dst,$src,$mem\t! add packed16F" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6555
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6556
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6557
    __ vaddps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6558
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6559
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6560
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6561
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6562
// Doubles vector add
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6563
instruct vadd2D(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6564
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6565
  match(Set dst (AddVD dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6566
  format %{ "addpd   $dst,$src\t! add packed2D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6567
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6568
    __ addpd($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6569
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6570
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6571
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6572
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6573
instruct vadd2D_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6574
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6575
  match(Set dst (AddVD src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6576
  format %{ "vaddpd  $dst,$src1,$src2\t! add packed2D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6577
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6578
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6579
    __ vaddpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6580
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6581
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6582
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6583
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6584
instruct vadd2D_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6585
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6586
  match(Set dst (AddVD src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6587
  format %{ "vaddpd  $dst,$src,$mem\t! add packed2D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6588
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6589
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6590
    __ vaddpd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6591
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6592
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6593
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6594
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6595
instruct vadd4D_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6596
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6597
  match(Set dst (AddVD src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6598
  format %{ "vaddpd  $dst,$src1,$src2\t! add packed4D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6599
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6600
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6601
    __ vaddpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6602
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6603
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6604
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6605
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6606
instruct vadd4D_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6607
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6608
  match(Set dst (AddVD src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6609
  format %{ "vaddpd  $dst,$src,$mem\t! add packed4D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6610
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6611
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6612
    __ vaddpd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6613
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6614
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6615
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6616
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6617
instruct vadd8D_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6618
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6619
  match(Set dst (AddVD src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6620
  format %{ "vaddpd  $dst,$src1,$src2\t! add packed8D" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6621
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6622
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6623
    __ vaddpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6624
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6625
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6626
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6627
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6628
instruct vadd8D_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6629
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6630
  match(Set dst (AddVD src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6631
  format %{ "vaddpd  $dst,$src,$mem\t! add packed8D" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6632
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6633
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6634
    __ vaddpd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6635
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6636
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6637
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6638
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6639
// --------------------------------- SUB --------------------------------------
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6640
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6641
// Bytes vector sub
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6642
instruct vsub4B(vecS dst, vecS src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  6643
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6644
  match(Set dst (SubVB dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6645
  format %{ "psubb   $dst,$src\t! sub packed4B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6646
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6647
    __ psubb($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6648
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6649
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6650
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6651
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6652
instruct vsub4B_reg(vecS dst, vecS src1, vecS src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6653
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6654
  match(Set dst (SubVB src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6655
  format %{ "vpsubb  $dst,$src1,$src2\t! sub packed4B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6656
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6657
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6658
    __ vpsubb($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6659
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6660
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6661
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6662
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6663
instruct vsub4B_mem(vecS dst, vecS src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6664
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6665
  match(Set dst (SubVB src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6666
  format %{ "vpsubb  $dst,$src,$mem\t! sub packed4B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6667
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6668
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6669
    __ vpsubb($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6670
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6671
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6672
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6673
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6674
instruct vsub8B(vecD dst, vecD src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  6675
  predicate(UseAVX == 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6676
  match(Set dst (SubVB dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6677
  format %{ "psubb   $dst,$src\t! sub packed8B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6678
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6679
    __ psubb($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6680
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6681
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6682
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6683
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6684
instruct vsub8B_reg(vecD dst, vecD src1, vecD src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6685
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6686
  match(Set dst (SubVB src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6687
  format %{ "vpsubb  $dst,$src1,$src2\t! sub packed8B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6688
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6689
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6690
    __ vpsubb($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6691
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6692
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6693
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6694
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6695
instruct vsub8B_mem(vecD dst, vecD src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6696
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6697
  match(Set dst (SubVB src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6698
  format %{ "vpsubb  $dst,$src,$mem\t! sub packed8B" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6699
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6700
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6701
    __ vpsubb($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6702
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6703
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6704
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6705
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6706
instruct vsub16B(vecX dst, vecX src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  6707
  predicate(UseAVX == 0 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6708
  match(Set dst (SubVB dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6709
  format %{ "psubb   $dst,$src\t! sub packed16B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6710
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6711
    __ psubb($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6712
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6713
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6714
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6715
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6716
instruct vsub16B_reg(vecX dst, vecX src1, vecX src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6717
  predicate(UseAVX > 0 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6718
  match(Set dst (SubVB src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6719
  format %{ "vpsubb  $dst,$src1,$src2\t! sub packed16B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6720
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6721
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6722
    __ vpsubb($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6723
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6724
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6725
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6726
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6727
instruct vsub16B_mem(vecX dst, vecX src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6728
  predicate(UseAVX > 0 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6729
  match(Set dst (SubVB src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6730
  format %{ "vpsubb  $dst,$src,$mem\t! sub packed16B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6731
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6732
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6733
    __ vpsubb($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6734
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6735
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6736
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6737
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6738
instruct vsub32B_reg(vecY dst, vecY src1, vecY src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6739
  predicate(UseAVX > 1 && n->as_Vector()->length() == 32);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6740
  match(Set dst (SubVB src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6741
  format %{ "vpsubb  $dst,$src1,$src2\t! sub packed32B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6742
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6743
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6744
    __ vpsubb($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6745
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6746
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6747
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6748
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6749
instruct vsub32B_mem(vecY dst, vecY src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6750
  predicate(UseAVX > 1 && n->as_Vector()->length() == 32);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6751
  match(Set dst (SubVB src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6752
  format %{ "vpsubb  $dst,$src,$mem\t! sub packed32B" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6753
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6754
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6755
    __ vpsubb($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6756
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6757
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6758
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6759
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6760
instruct vsub64B_reg(vecZ dst, vecZ src1, vecZ src2) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6761
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 64);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6762
  match(Set dst (SubVB src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6763
  format %{ "vpsubb  $dst,$src1,$src2\t! sub packed64B" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6764
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6765
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6766
    __ vpsubb($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6767
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6768
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6769
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6770
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6771
instruct vsub64B_mem(vecZ dst, vecZ src, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6772
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 64);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6773
  match(Set dst (SubVB src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6774
  format %{ "vpsubb  $dst,$src,$mem\t! sub packed64B" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6775
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6776
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6777
    __ vpsubb($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6778
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6779
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6780
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6781
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6782
// Shorts/Chars vector sub
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6783
instruct vsub2S(vecS dst, vecS src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  6784
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6785
  match(Set dst (SubVS dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6786
  format %{ "psubw   $dst,$src\t! sub packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6787
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6788
    __ psubw($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6789
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6790
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6791
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6792
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6793
instruct vsub2S_reg(vecS dst, vecS src1, vecS src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6794
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6795
  match(Set dst (SubVS src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6796
  format %{ "vpsubw  $dst,$src1,$src2\t! sub packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6797
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6798
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6799
    __ vpsubw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6800
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6801
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6802
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6803
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6804
instruct vsub2S_mem(vecS dst, vecS src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6805
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6806
  match(Set dst (SubVS src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6807
  format %{ "vpsubw  $dst,$src,$mem\t! sub packed2S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6808
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6809
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6810
    __ vpsubw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6811
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6812
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6813
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6814
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6815
instruct vsub4S(vecD dst, vecD src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  6816
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6817
  match(Set dst (SubVS dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6818
  format %{ "psubw   $dst,$src\t! sub packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6819
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6820
    __ psubw($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6821
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6822
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6823
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6824
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6825
instruct vsub4S_reg(vecD dst, vecD src1, vecD src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6826
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6827
  match(Set dst (SubVS src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6828
  format %{ "vpsubw  $dst,$src1,$src2\t! sub packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6829
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6830
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6831
    __ vpsubw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6832
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6833
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6834
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6835
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6836
instruct vsub4S_mem(vecD dst, vecD src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6837
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6838
  match(Set dst (SubVS src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6839
  format %{ "vpsubw  $dst,$src,$mem\t! sub packed4S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6840
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6841
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6842
    __ vpsubw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6843
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6844
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6845
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6846
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6847
instruct vsub8S(vecX dst, vecX src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  6848
  predicate(UseAVX == 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6849
  match(Set dst (SubVS dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6850
  format %{ "psubw   $dst,$src\t! sub packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6851
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6852
    __ psubw($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6853
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6854
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6855
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6856
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6857
instruct vsub8S_reg(vecX dst, vecX src1, vecX src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6858
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6859
  match(Set dst (SubVS src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6860
  format %{ "vpsubw  $dst,$src1,$src2\t! sub packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6861
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6862
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6863
    __ vpsubw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6864
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6865
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6866
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6867
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6868
instruct vsub8S_mem(vecX dst, vecX src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6869
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6870
  match(Set dst (SubVS src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6871
  format %{ "vpsubw  $dst,$src,$mem\t! sub packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6872
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6873
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6874
    __ vpsubw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6875
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6876
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6877
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6878
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6879
instruct vsub16S_reg(vecY dst, vecY src1, vecY src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6880
  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6881
  match(Set dst (SubVS src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6882
  format %{ "vpsubw  $dst,$src1,$src2\t! sub packed16S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6883
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6884
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6885
    __ vpsubw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6886
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6887
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6888
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6889
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6890
instruct vsub16S_mem(vecY dst, vecY src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6891
  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6892
  match(Set dst (SubVS src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6893
  format %{ "vpsubw  $dst,$src,$mem\t! sub packed16S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6894
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6895
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6896
    __ vpsubw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6897
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6898
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6899
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6900
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6901
instruct vsub32S_reg(vecZ dst, vecZ src1, vecZ src2) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6902
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 32);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6903
  match(Set dst (SubVS src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6904
  format %{ "vpsubw  $dst,$src1,$src2\t! sub packed32S" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6905
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6906
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6907
    __ vpsubw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6908
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6909
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6910
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6911
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6912
instruct vsub32S_mem(vecZ dst, vecZ src, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6913
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 32);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6914
  match(Set dst (SubVS src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6915
  format %{ "vpsubw  $dst,$src,$mem\t! sub packed32S" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6916
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6917
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6918
    __ vpsubw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6919
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6920
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6921
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6922
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6923
// Integers vector sub
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6924
instruct vsub2I(vecD dst, vecD src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6925
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6926
  match(Set dst (SubVI dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6927
  format %{ "psubd   $dst,$src\t! sub packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6928
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6929
    __ psubd($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6930
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6931
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6932
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6933
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6934
instruct vsub2I_reg(vecD dst, vecD src1, vecD src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6935
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6936
  match(Set dst (SubVI src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6937
  format %{ "vpsubd  $dst,$src1,$src2\t! sub packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6938
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6939
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6940
    __ vpsubd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6941
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6942
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6943
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6944
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6945
instruct vsub2I_mem(vecD dst, vecD src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6946
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6947
  match(Set dst (SubVI src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6948
  format %{ "vpsubd  $dst,$src,$mem\t! sub packed2I" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6949
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6950
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6951
    __ vpsubd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6952
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6953
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6954
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  6955
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6956
instruct vsub4I(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  6957
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6958
  match(Set dst (SubVI dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6959
  format %{ "psubd   $dst,$src\t! sub packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6960
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6961
    __ psubd($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6962
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6963
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6964
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6965
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6966
instruct vsub4I_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6967
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6968
  match(Set dst (SubVI src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6969
  format %{ "vpsubd  $dst,$src1,$src2\t! sub packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6970
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6971
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6972
    __ vpsubd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6973
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6974
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6975
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6976
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6977
instruct vsub4I_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6978
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6979
  match(Set dst (SubVI src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6980
  format %{ "vpsubd  $dst,$src,$mem\t! sub packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6981
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6982
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6983
    __ vpsubd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6984
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6985
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6986
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6987
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6988
instruct vsub8I_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6989
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6990
  match(Set dst (SubVI src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6991
  format %{ "vpsubd  $dst,$src1,$src2\t! sub packed8I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6992
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6993
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  6994
    __ vpsubd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6995
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6996
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6997
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6998
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  6999
instruct vsub8I_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7000
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7001
  match(Set dst (SubVI src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7002
  format %{ "vpsubd  $dst,$src,$mem\t! sub packed8I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7003
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7004
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7005
    __ vpsubd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7006
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7007
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7008
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7009
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7010
instruct vsub16I_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7011
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7012
  match(Set dst (SubVI src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7013
  format %{ "vpsubd  $dst,$src1,$src2\t! sub packed16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7014
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7015
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7016
    __ vpsubd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7017
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7018
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7019
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7020
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7021
instruct vsub16I_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7022
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7023
  match(Set dst (SubVI src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7024
  format %{ "vpsubd  $dst,$src,$mem\t! sub packed16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7025
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7026
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7027
    __ vpsubd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7028
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7029
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7030
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7031
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7032
// Longs vector sub
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7033
instruct vsub2L(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7034
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7035
  match(Set dst (SubVL dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7036
  format %{ "psubq   $dst,$src\t! sub packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7037
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7038
    __ psubq($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7039
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7040
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7041
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7042
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7043
instruct vsub2L_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7044
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7045
  match(Set dst (SubVL src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7046
  format %{ "vpsubq  $dst,$src1,$src2\t! sub packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7047
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7048
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7049
    __ vpsubq($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7050
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7051
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7052
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7053
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7054
instruct vsub2L_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7055
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7056
  match(Set dst (SubVL src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7057
  format %{ "vpsubq  $dst,$src,$mem\t! sub packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7058
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7059
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7060
    __ vpsubq($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7061
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7062
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7063
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7064
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7065
instruct vsub4L_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7066
  predicate(UseAVX > 1 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7067
  match(Set dst (SubVL src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7068
  format %{ "vpsubq  $dst,$src1,$src2\t! sub packed4L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7069
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7070
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7071
    __ vpsubq($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7072
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7073
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7074
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7075
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7076
instruct vsub4L_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7077
  predicate(UseAVX > 1 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7078
  match(Set dst (SubVL src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7079
  format %{ "vpsubq  $dst,$src,$mem\t! sub packed4L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7080
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7081
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7082
    __ vpsubq($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7083
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7084
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7085
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7086
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7087
instruct vsub8L_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7088
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7089
  match(Set dst (SubVL src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7090
  format %{ "vpsubq  $dst,$src1,$src2\t! sub packed8L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7091
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7092
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7093
    __ vpsubq($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7094
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7095
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7096
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7097
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7098
instruct vsub8L_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7099
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7100
  match(Set dst (SubVL src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7101
  format %{ "vpsubq  $dst,$src,$mem\t! sub packed8L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7102
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7103
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7104
    __ vpsubq($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7105
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7106
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7107
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7108
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7109
// Floats vector sub
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7110
instruct vsub2F(vecD dst, vecD src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7111
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7112
  match(Set dst (SubVF dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7113
  format %{ "subps   $dst,$src\t! sub packed2F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7114
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7115
    __ subps($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7116
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7117
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7118
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7119
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7120
instruct vsub2F_reg(vecD dst, vecD src1, vecD src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7121
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7122
  match(Set dst (SubVF src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7123
  format %{ "vsubps  $dst,$src1,$src2\t! sub packed2F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7124
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7125
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7126
    __ vsubps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7127
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7128
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7129
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7130
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7131
instruct vsub2F_mem(vecD dst, vecD src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7132
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7133
  match(Set dst (SubVF src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7134
  format %{ "vsubps  $dst,$src,$mem\t! sub packed2F" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7135
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7136
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7137
    __ vsubps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7138
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7139
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7140
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7141
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7142
instruct vsub4F(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7143
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7144
  match(Set dst (SubVF dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7145
  format %{ "subps   $dst,$src\t! sub packed4F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7146
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7147
    __ subps($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7148
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7149
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7150
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7151
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7152
instruct vsub4F_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7153
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7154
  match(Set dst (SubVF src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7155
  format %{ "vsubps  $dst,$src1,$src2\t! sub packed4F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7156
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7157
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7158
    __ vsubps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7159
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7160
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7161
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7162
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7163
instruct vsub4F_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7164
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7165
  match(Set dst (SubVF src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7166
  format %{ "vsubps  $dst,$src,$mem\t! sub packed4F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7167
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7168
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7169
    __ vsubps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7170
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7171
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7172
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7173
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7174
instruct vsub8F_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7175
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7176
  match(Set dst (SubVF src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7177
  format %{ "vsubps  $dst,$src1,$src2\t! sub packed8F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7178
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7179
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7180
    __ vsubps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7181
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7182
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7183
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7184
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7185
instruct vsub8F_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7186
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7187
  match(Set dst (SubVF src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7188
  format %{ "vsubps  $dst,$src,$mem\t! sub packed8F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7189
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7190
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7191
    __ vsubps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7192
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7193
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7194
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7195
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7196
instruct vsub16F_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7197
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7198
  match(Set dst (SubVF src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7199
  format %{ "vsubps  $dst,$src1,$src2\t! sub packed16F" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7200
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7201
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7202
    __ vsubps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7203
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7204
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7205
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7206
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7207
instruct vsub16F_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7208
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7209
  match(Set dst (SubVF src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7210
  format %{ "vsubps  $dst,$src,$mem\t! sub packed16F" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7211
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7212
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7213
    __ vsubps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7214
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7215
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7216
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7217
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7218
// Doubles vector sub
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7219
instruct vsub2D(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7220
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7221
  match(Set dst (SubVD dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7222
  format %{ "subpd   $dst,$src\t! sub packed2D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7223
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7224
    __ subpd($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7225
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7226
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7227
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7228
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7229
instruct vsub2D_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7230
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7231
  match(Set dst (SubVD src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7232
  format %{ "vsubpd  $dst,$src1,$src2\t! sub packed2D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7233
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7234
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7235
    __ vsubpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7236
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7237
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7238
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7239
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7240
instruct vsub2D_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7241
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7242
  match(Set dst (SubVD src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7243
  format %{ "vsubpd  $dst,$src,$mem\t! sub packed2D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7244
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7245
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7246
    __ vsubpd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7247
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7248
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7249
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7250
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7251
instruct vsub4D_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7252
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7253
  match(Set dst (SubVD src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7254
  format %{ "vsubpd  $dst,$src1,$src2\t! sub packed4D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7255
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7256
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7257
    __ vsubpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7258
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7259
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7260
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7261
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7262
instruct vsub4D_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7263
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7264
  match(Set dst (SubVD src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7265
  format %{ "vsubpd  $dst,$src,$mem\t! sub packed4D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7266
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7267
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7268
    __ vsubpd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7269
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7270
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7271
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7272
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7273
instruct vsub8D_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7274
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7275
  match(Set dst (SubVD src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7276
  format %{ "vsubpd  $dst,$src1,$src2\t! sub packed8D" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7277
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7278
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7279
    __ vsubpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7280
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7281
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7282
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7283
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7284
instruct vsub8D_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7285
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7286
  match(Set dst (SubVD src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7287
  format %{ "vsubpd  $dst,$src,$mem\t! sub packed8D" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7288
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7289
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7290
    __ vsubpd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7291
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7292
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7293
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7294
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7295
// --------------------------------- MUL --------------------------------------
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7296
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7297
// Shorts/Chars vector mul
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7298
instruct vmul2S(vecS dst, vecS src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  7299
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7300
  match(Set dst (MulVS dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7301
  format %{ "pmullw $dst,$src\t! mul packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7302
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7303
    __ pmullw($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7304
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7305
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7306
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7307
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7308
instruct vmul2S_reg(vecS dst, vecS src1, vecS src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7309
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7310
  match(Set dst (MulVS src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7311
  format %{ "vpmullw $dst,$src1,$src2\t! mul packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7312
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7313
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7314
    __ vpmullw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7315
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7316
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7317
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7318
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7319
instruct vmul2S_mem(vecS dst, vecS src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7320
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7321
  match(Set dst (MulVS src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7322
  format %{ "vpmullw $dst,$src,$mem\t! mul packed2S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7323
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7324
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7325
    __ vpmullw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7326
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7327
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7328
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7329
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7330
instruct vmul4S(vecD dst, vecD src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  7331
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7332
  match(Set dst (MulVS dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7333
  format %{ "pmullw  $dst,$src\t! mul packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7334
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7335
    __ pmullw($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7336
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7337
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7338
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7339
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7340
instruct vmul4S_reg(vecD dst, vecD src1, vecD src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7341
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7342
  match(Set dst (MulVS src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7343
  format %{ "vpmullw $dst,$src1,$src2\t! mul packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7344
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7345
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7346
    __ vpmullw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7347
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7348
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7349
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7350
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7351
instruct vmul4S_mem(vecD dst, vecD src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7352
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7353
  match(Set dst (MulVS src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7354
  format %{ "vpmullw $dst,$src,$mem\t! mul packed4S" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7355
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7356
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7357
    __ vpmullw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7358
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7359
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7360
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7361
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7362
instruct vmul8S(vecX dst, vecX src) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  7363
  predicate(UseAVX == 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7364
  match(Set dst (MulVS dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7365
  format %{ "pmullw  $dst,$src\t! mul packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7366
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7367
    __ pmullw($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7368
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7369
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7370
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7371
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7372
instruct vmul8S_reg(vecX dst, vecX src1, vecX src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7373
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7374
  match(Set dst (MulVS src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7375
  format %{ "vpmullw $dst,$src1,$src2\t! mul packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7376
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7377
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7378
    __ vpmullw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7379
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7380
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7381
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7382
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7383
instruct vmul8S_mem(vecX dst, vecX src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7384
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7385
  match(Set dst (MulVS src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7386
  format %{ "vpmullw $dst,$src,$mem\t! mul packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7387
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7388
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7389
    __ vpmullw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7390
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7391
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7392
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7393
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7394
instruct vmul16S_reg(vecY dst, vecY src1, vecY src2) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7395
  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7396
  match(Set dst (MulVS src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7397
  format %{ "vpmullw $dst,$src1,$src2\t! mul packed16S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7398
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7399
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7400
    __ vpmullw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7401
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7402
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7403
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7404
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7405
instruct vmul16S_mem(vecY dst, vecY src, memory mem) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7406
  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7407
  match(Set dst (MulVS src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7408
  format %{ "vpmullw $dst,$src,$mem\t! mul packed16S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7409
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7410
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7411
    __ vpmullw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7412
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7413
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7414
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7415
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7416
instruct vmul32S_reg(vecZ dst, vecZ src1, vecZ src2) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7417
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 32);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7418
  match(Set dst (MulVS src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7419
  format %{ "vpmullw $dst,$src1,$src2\t! mul packed32S" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7420
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7421
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7422
    __ vpmullw($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7423
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7424
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7425
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7426
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7427
instruct vmul32S_mem(vecZ dst, vecZ src, memory mem) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7428
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 32);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7429
  match(Set dst (MulVS src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7430
  format %{ "vpmullw $dst,$src,$mem\t! mul packed32S" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7431
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7432
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7433
    __ vpmullw($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7434
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7435
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7436
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7437
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7438
// Integers vector mul (sse4_1)
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7439
instruct vmul2I(vecD dst, vecD src) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7440
  predicate(UseSSE > 3 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7441
  match(Set dst (MulVI dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7442
  format %{ "pmulld  $dst,$src\t! mul packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7443
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7444
    __ pmulld($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7445
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7446
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7447
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7448
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7449
instruct vmul2I_reg(vecD dst, vecD src1, vecD src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7450
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7451
  match(Set dst (MulVI src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7452
  format %{ "vpmulld $dst,$src1,$src2\t! mul packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7453
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7454
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7455
    __ vpmulld($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7456
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7457
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7458
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7459
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7460
instruct vmul2I_mem(vecD dst, vecD src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7461
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7462
  match(Set dst (MulVI src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7463
  format %{ "vpmulld $dst,$src,$mem\t! mul packed2I" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7464
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7465
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7466
    __ vpmulld($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7467
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7468
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7469
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7470
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7471
instruct vmul4I(vecX dst, vecX src) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7472
  predicate(UseSSE > 3 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7473
  match(Set dst (MulVI dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7474
  format %{ "pmulld  $dst,$src\t! mul packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7475
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7476
    __ pmulld($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7477
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7478
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7479
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7480
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7481
instruct vmul4I_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7482
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7483
  match(Set dst (MulVI src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7484
  format %{ "vpmulld $dst,$src1,$src2\t! mul packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7485
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7486
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7487
    __ vpmulld($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7488
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7489
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7490
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7491
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7492
instruct vmul4I_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7493
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7494
  match(Set dst (MulVI src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7495
  format %{ "vpmulld $dst,$src,$mem\t! mul packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7496
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7497
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7498
    __ vpmulld($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7499
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7500
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7501
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7502
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7503
instruct vmul2L_reg(vecX dst, vecX src1, vecX src2) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7504
  predicate(UseAVX > 2 && n->as_Vector()->length() == 2 && VM_Version::supports_avx512dq());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7505
  match(Set dst (MulVL src1 src2));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7506
  format %{ "vpmullq $dst,$src1,$src2\t! mul packed2L" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7507
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7508
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7509
    __ vpmullq($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7510
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7511
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7512
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7513
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7514
instruct vmul2L_mem(vecX dst, vecX src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7515
  predicate(UseAVX > 2 && n->as_Vector()->length() == 2 && VM_Version::supports_avx512dq());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7516
  match(Set dst (MulVL src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7517
  format %{ "vpmullq $dst,$src,$mem\t! mul packed2L" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7518
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7519
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7520
    __ vpmullq($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7521
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7522
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7523
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7524
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7525
instruct vmul4L_reg(vecY dst, vecY src1, vecY src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7526
  predicate(UseAVX > 2 && n->as_Vector()->length() == 4 && VM_Version::supports_avx512dq());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7527
  match(Set dst (MulVL src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7528
  format %{ "vpmullq $dst,$src1,$src2\t! mul packed4L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7529
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7530
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7531
    __ vpmullq($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7532
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7533
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7534
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7535
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7536
instruct vmul4L_mem(vecY dst, vecY src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7537
  predicate(UseAVX > 2 && n->as_Vector()->length() == 4 && VM_Version::supports_avx512dq());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7538
  match(Set dst (MulVL src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7539
  format %{ "vpmullq $dst,$src,$mem\t! mul packed4L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7540
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7541
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7542
    __ vpmullq($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7543
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7544
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7545
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7546
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7547
instruct vmul8L_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7548
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8 && VM_Version::supports_avx512dq());
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7549
  match(Set dst (MulVL src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7550
  format %{ "vpmullq $dst,$src1,$src2\t! mul packed8L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7551
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7552
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7553
    __ vpmullq($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7554
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7555
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7556
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7557
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7558
instruct vmul8L_mem(vecZ dst, vecZ src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7559
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8 && VM_Version::supports_avx512dq());
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7560
  match(Set dst (MulVL src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7561
  format %{ "vpmullq $dst,$src,$mem\t! mul packed8L" %}
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7562
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7563
    int vector_len = 2;
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7564
    __ vpmullq($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7565
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7566
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7567
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7568
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7569
instruct vmul8I_reg(vecY dst, vecY src1, vecY src2) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7570
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7571
  match(Set dst (MulVI src1 src2));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7572
  format %{ "vpmulld $dst,$src1,$src2\t! mul packed8I" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7573
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7574
    int vector_len = 1;
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7575
    __ vpmulld($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7576
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7577
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7578
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7579
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7580
instruct vmul8I_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7581
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7582
  match(Set dst (MulVI src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7583
  format %{ "vpmulld $dst,$src,$mem\t! mul packed8I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7584
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7585
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7586
    __ vpmulld($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7587
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7588
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7589
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7590
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7591
instruct vmul16I_reg(vecZ dst, vecZ src1, vecZ src2) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7592
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7593
  match(Set dst (MulVI src1 src2));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7594
  format %{ "vpmulld $dst,$src1,$src2\t! mul packed16I" %}
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7595
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7596
    int vector_len = 2;
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7597
    __ vpmulld($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7598
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7599
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7600
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7601
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7602
instruct vmul16I_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7603
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7604
  match(Set dst (MulVI src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7605
  format %{ "vpmulld $dst,$src,$mem\t! mul packed16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7606
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7607
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7608
    __ vpmulld($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7609
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7610
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7611
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7612
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7613
// Floats vector mul
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7614
instruct vmul2F(vecD dst, vecD src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7615
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7616
  match(Set dst (MulVF dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7617
  format %{ "mulps   $dst,$src\t! mul packed2F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7618
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7619
    __ mulps($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7620
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7621
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7622
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7623
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7624
instruct vmul2F_reg(vecD dst, vecD src1, vecD src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7625
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7626
  match(Set dst (MulVF src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7627
  format %{ "vmulps  $dst,$src1,$src2\t! mul packed2F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7628
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7629
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7630
    __ vmulps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7631
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7632
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7633
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7634
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7635
instruct vmul2F_mem(vecD dst, vecD src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7636
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7637
  match(Set dst (MulVF src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7638
  format %{ "vmulps  $dst,$src,$mem\t! mul packed2F" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7639
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7640
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7641
    __ vmulps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7642
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7643
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7644
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7645
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7646
instruct vmul4F(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7647
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7648
  match(Set dst (MulVF dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7649
  format %{ "mulps   $dst,$src\t! mul packed4F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7650
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7651
    __ mulps($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7652
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7653
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7654
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7655
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7656
instruct vmul4F_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7657
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7658
  match(Set dst (MulVF src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7659
  format %{ "vmulps  $dst,$src1,$src2\t! mul packed4F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7660
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7661
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7662
    __ vmulps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7663
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7664
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7665
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7666
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7667
instruct vmul4F_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7668
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7669
  match(Set dst (MulVF src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7670
  format %{ "vmulps  $dst,$src,$mem\t! mul packed4F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7671
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7672
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7673
    __ vmulps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7674
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7675
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7676
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7677
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7678
instruct vmul8F_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7679
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7680
  match(Set dst (MulVF src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7681
  format %{ "vmulps  $dst,$src1,$src2\t! mul packed8F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7682
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7683
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7684
    __ vmulps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7685
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7686
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7687
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7688
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7689
instruct vmul8F_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7690
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7691
  match(Set dst (MulVF src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7692
  format %{ "vmulps  $dst,$src,$mem\t! mul packed8F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7693
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7694
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7695
    __ vmulps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7696
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7697
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7698
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7699
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7700
instruct vmul16F_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7701
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7702
  match(Set dst (MulVF src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7703
  format %{ "vmulps  $dst,$src1,$src2\t! mul packed16F" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7704
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7705
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7706
    __ vmulps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7707
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7708
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7709
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7710
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7711
instruct vmul16F_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7712
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7713
  match(Set dst (MulVF src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7714
  format %{ "vmulps  $dst,$src,$mem\t! mul packed16F" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7715
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7716
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7717
    __ vmulps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7718
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7719
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7720
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7721
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7722
// Doubles vector mul
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7723
instruct vmul2D(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7724
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7725
  match(Set dst (MulVD dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7726
  format %{ "mulpd   $dst,$src\t! mul packed2D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7727
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7728
    __ mulpd($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7729
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7730
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7731
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7732
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7733
instruct vmul2D_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7734
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7735
  match(Set dst (MulVD src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7736
  format %{ "vmulpd  $dst,$src1,$src2\t! mul packed2D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7737
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7738
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7739
    __ vmulpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7740
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7741
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7742
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7743
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7744
instruct vmul2D_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7745
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7746
  match(Set dst (MulVD src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7747
  format %{ "vmulpd  $dst,$src,$mem\t! mul packed2D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7748
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7749
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7750
    __ vmulpd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7751
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7752
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7753
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7754
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7755
instruct vmul4D_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7756
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7757
  match(Set dst (MulVD src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7758
  format %{ "vmulpd  $dst,$src1,$src2\t! mul packed4D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7759
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7760
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7761
    __ vmulpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7762
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7763
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7764
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7765
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7766
instruct vmul4D_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7767
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7768
  match(Set dst (MulVD src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7769
  format %{ "vmulpd  $dst,$src,$mem\t! mul packed4D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7770
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7771
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7772
    __ vmulpd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7773
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7774
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7775
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7776
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7777
instruct vmul8D_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7778
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7779
  match(Set dst (MulVD src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7780
  format %{ "vmulpd  $dst k0,$src1,$src2\t! mul packed8D" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7781
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7782
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7783
    __ vmulpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7784
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7785
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7786
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7787
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7788
instruct vmul8D_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7789
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7790
  match(Set dst (MulVD src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7791
  format %{ "vmulpd  $dst k0,$src,$mem\t! mul packed8D" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7792
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7793
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7794
    __ vmulpd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7795
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7796
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7797
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7798
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7799
instruct vcmov8F_reg(legVecY dst, legVecY src1, legVecY src2, immI8 cop, cmpOp_vcmppd copnd) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7800
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
48309
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7801
  match(Set dst (CMoveVF (Binary copnd cop) (Binary src1 src2)));
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7802
  effect(TEMP dst, USE src1, USE src2);
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7803
  format %{ "cmpps.$copnd  $dst, $src1, $src2  ! vcmovevf, cond=$cop\n\t"
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7804
            "blendvps $dst,$src1,$src2,$dst ! vcmovevf\n\t"
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7805
         %}
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7806
  ins_encode %{
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7807
    int vector_len = 1;
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7808
    int cond = (Assembler::Condition)($copnd$$cmpcode);
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7809
    __ cmpps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, cond, vector_len);
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7810
    __ blendvps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, $dst$$XMMRegister, vector_len);
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7811
  %}
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7812
  ins_pipe( pipe_slow );
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7813
%}
1a0499fd252e 8192846: Support cmov vectorization for float
kvn
parents: 48089
diff changeset
  7814
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7815
instruct vcmov4D_reg(legVecY dst, legVecY src1, legVecY src2, immI8 cop, cmpOp_vcmppd copnd) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7816
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
33469
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  7817
  match(Set dst (CMoveVD (Binary copnd cop) (Binary src1 src2)));
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  7818
  effect(TEMP dst, USE src1, USE src2);
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  7819
  format %{ "cmppd.$copnd  $dst, $src1, $src2  ! vcmovevd, cond=$cop\n\t"
42039
db627462f2c9 8165381: Update for x86 SHA512 using AVX2
kvn
parents: 41333
diff changeset
  7820
            "blendvpd $dst,$src1,$src2,$dst ! vcmovevd\n\t"
33469
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  7821
         %}
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  7822
  ins_encode %{
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  7823
    int vector_len = 1;
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  7824
    int cond = (Assembler::Condition)($copnd$$cmpcode);
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  7825
    __ cmppd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, cond, vector_len);
42039
db627462f2c9 8165381: Update for x86 SHA512 using AVX2
kvn
parents: 41333
diff changeset
  7826
    __ blendvpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, $dst$$XMMRegister, vector_len);
33469
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  7827
  %}
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  7828
  ins_pipe( pipe_slow );
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  7829
%}
30f4811eded0 8139340: SuperWord enhancement to support vector conditional move (CMovVD) on Intel AVX cpu
iveresov
parents: 33065
diff changeset
  7830
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7831
// --------------------------------- DIV --------------------------------------
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7832
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7833
// Floats vector div
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7834
instruct vdiv2F(vecD dst, vecD src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7835
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7836
  match(Set dst (DivVF dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7837
  format %{ "divps   $dst,$src\t! div packed2F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7838
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7839
    __ divps($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7840
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7841
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7842
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7843
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7844
instruct vdiv2F_reg(vecD dst, vecD src1, vecD src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7845
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7846
  match(Set dst (DivVF src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7847
  format %{ "vdivps  $dst,$src1,$src2\t! div packed2F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7848
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7849
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7850
    __ vdivps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7851
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7852
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7853
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7854
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7855
instruct vdiv2F_mem(vecD dst, vecD src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7856
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7857
  match(Set dst (DivVF src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7858
  format %{ "vdivps  $dst,$src,$mem\t! div packed2F" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7859
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7860
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7861
    __ vdivps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7862
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7863
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7864
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  7865
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7866
instruct vdiv4F(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7867
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7868
  match(Set dst (DivVF dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7869
  format %{ "divps   $dst,$src\t! div packed4F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7870
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7871
    __ divps($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7872
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7873
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7874
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7875
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7876
instruct vdiv4F_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7877
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7878
  match(Set dst (DivVF src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7879
  format %{ "vdivps  $dst,$src1,$src2\t! div packed4F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7880
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7881
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7882
    __ vdivps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7883
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7884
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7885
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7886
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7887
instruct vdiv4F_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7888
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7889
  match(Set dst (DivVF src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7890
  format %{ "vdivps  $dst,$src,$mem\t! div packed4F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7891
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7892
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7893
    __ vdivps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7894
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7895
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7896
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7897
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7898
instruct vdiv8F_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7899
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7900
  match(Set dst (DivVF src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7901
  format %{ "vdivps  $dst,$src1,$src2\t! div packed8F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7902
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7903
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7904
    __ vdivps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7905
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7906
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7907
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7908
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7909
instruct vdiv8F_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7910
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7911
  match(Set dst (DivVF src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7912
  format %{ "vdivps  $dst,$src,$mem\t! div packed8F" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7913
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7914
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7915
    __ vdivps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7916
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7917
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7918
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7919
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7920
instruct vdiv16F_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7921
  predicate(UseAVX > 0 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7922
  match(Set dst (DivVF src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7923
  format %{ "vdivps  $dst,$src1,$src2\t! div packed16F" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7924
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7925
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7926
    __ vdivps($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7927
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7928
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7929
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7930
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7931
instruct vdiv16F_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7932
  predicate(UseAVX > 0 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7933
  match(Set dst (DivVF src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7934
  format %{ "vdivps  $dst,$src,$mem\t! div packed16F" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7935
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7936
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7937
    __ vdivps($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7938
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7939
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7940
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7941
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7942
// Doubles vector div
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7943
instruct vdiv2D(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  7944
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7945
  match(Set dst (DivVD dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7946
  format %{ "divpd   $dst,$src\t! div packed2D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7947
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7948
    __ divpd($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7949
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7950
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7951
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7952
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7953
instruct vdiv2D_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7954
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7955
  match(Set dst (DivVD src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7956
  format %{ "vdivpd  $dst,$src1,$src2\t! div packed2D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7957
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7958
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7959
    __ vdivpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7960
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7961
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7962
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7963
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7964
instruct vdiv2D_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7965
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7966
  match(Set dst (DivVD src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7967
  format %{ "vdivpd  $dst,$src,$mem\t! div packed2D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7968
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7969
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7970
    __ vdivpd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7971
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7972
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7973
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7974
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7975
instruct vdiv4D_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7976
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7977
  match(Set dst (DivVD src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7978
  format %{ "vdivpd  $dst,$src1,$src2\t! div packed4D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7979
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7980
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7981
    __ vdivpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7982
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7983
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7984
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7985
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7986
instruct vdiv4D_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7987
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7988
  match(Set dst (DivVD src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7989
  format %{ "vdivpd  $dst,$src,$mem\t! div packed4D" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  7990
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7991
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7992
    __ vdivpd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7993
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7994
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7995
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7996
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7997
instruct vdiv8D_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7998
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  7999
  match(Set dst (DivVD src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8000
  format %{ "vdivpd  $dst,$src1,$src2\t! div packed8D" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8001
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8002
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8003
    __ vdivpd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8004
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8005
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8006
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8007
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8008
instruct vdiv8D_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8009
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8010
  match(Set dst (DivVD src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8011
  format %{ "vdivpd  $dst,$src,$mem\t! div packed8D" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8012
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8013
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8014
    __ vdivpd($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8015
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8016
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8017
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8018
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8019
// ------------------------------ Shift ---------------------------------------
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8020
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8021
// Left and right shift count vectors are the same on x86
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8022
// (only lowest bits of xmm reg are used for count).
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8023
instruct vshiftcnt(vecS dst, rRegI cnt) %{
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8024
  match(Set dst (LShiftCntV cnt));
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8025
  match(Set dst (RShiftCntV cnt));
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8026
  format %{ "movd    $dst,$cnt\t! load shift count" %}
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8027
  ins_encode %{
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8028
    __ movdl($dst$$XMMRegister, $cnt$$Register);
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8029
  %}
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8030
  ins_pipe( pipe_slow );
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8031
%}
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8032
32723
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8033
// --------------------------------- Sqrt --------------------------------------
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8034
48089
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8035
// Floating point vector sqrt
32723
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8036
instruct vsqrt2D_reg(vecX dst, vecX src) %{
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8037
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8038
  match(Set dst (SqrtVD src));
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8039
  format %{ "vsqrtpd  $dst,$src\t! sqrt packed2D" %}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8040
  ins_encode %{
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8041
    int vector_len = 0;
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8042
    __ vsqrtpd($dst$$XMMRegister, $src$$XMMRegister, vector_len);
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8043
  %}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8044
  ins_pipe( pipe_slow );
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8045
%}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8046
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8047
instruct vsqrt2D_mem(vecX dst, memory mem) %{
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8048
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8049
  match(Set dst (SqrtVD (LoadVector mem)));
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8050
  format %{ "vsqrtpd  $dst,$mem\t! sqrt packed2D" %}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8051
  ins_encode %{
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8052
    int vector_len = 0;
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8053
    __ vsqrtpd($dst$$XMMRegister, $mem$$Address, vector_len);
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8054
  %}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8055
  ins_pipe( pipe_slow );
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8056
%}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8057
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8058
instruct vsqrt4D_reg(vecY dst, vecY src) %{
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8059
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8060
  match(Set dst (SqrtVD src));
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8061
  format %{ "vsqrtpd  $dst,$src\t! sqrt packed4D" %}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8062
  ins_encode %{
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8063
    int vector_len = 1;
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8064
    __ vsqrtpd($dst$$XMMRegister, $src$$XMMRegister, vector_len);
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8065
  %}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8066
  ins_pipe( pipe_slow );
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8067
%}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8068
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8069
instruct vsqrt4D_mem(vecY dst, memory mem) %{
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8070
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8071
  match(Set dst (SqrtVD (LoadVector mem)));
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8072
  format %{ "vsqrtpd  $dst,$mem\t! sqrt packed4D" %}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8073
  ins_encode %{
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8074
    int vector_len = 1;
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8075
    __ vsqrtpd($dst$$XMMRegister, $mem$$Address, vector_len);
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8076
  %}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8077
  ins_pipe( pipe_slow );
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8078
%}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8079
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8080
instruct vsqrt8D_reg(vecZ dst, vecZ src) %{
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8081
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8082
  match(Set dst (SqrtVD src));
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8083
  format %{ "vsqrtpd  $dst,$src\t! sqrt packed8D" %}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8084
  ins_encode %{
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8085
    int vector_len = 2;
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8086
    __ vsqrtpd($dst$$XMMRegister, $src$$XMMRegister, vector_len);
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8087
  %}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8088
  ins_pipe( pipe_slow );
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8089
%}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8090
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8091
instruct vsqrt8D_mem(vecZ dst, memory mem) %{
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8092
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8093
  match(Set dst (SqrtVD (LoadVector mem)));
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8094
  format %{ "vsqrtpd  $dst,$mem\t! sqrt packed8D" %}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8095
  ins_encode %{
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8096
    int vector_len = 2;
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8097
    __ vsqrtpd($dst$$XMMRegister, $mem$$Address, vector_len);
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8098
  %}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8099
  ins_pipe( pipe_slow );
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8100
%}
56534fb3d71a 8135028: support for vectorizing double precision sqrt
mcberg
parents: 32082
diff changeset
  8101
48089
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8102
instruct vsqrt2F_reg(vecD dst, vecD src) %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8103
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8104
  match(Set dst (SqrtVF src));
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8105
  format %{ "vsqrtps  $dst,$src\t! sqrt packed2F" %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8106
  ins_encode %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8107
    int vector_len = 0;
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8108
    __ vsqrtps($dst$$XMMRegister, $src$$XMMRegister, vector_len);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8109
  %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8110
  ins_pipe( pipe_slow );
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8111
%}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8112
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8113
instruct vsqrt2F_mem(vecD dst, memory mem) %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8114
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8115
  match(Set dst (SqrtVF (LoadVector mem)));
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8116
  format %{ "vsqrtps  $dst,$mem\t! sqrt packed2F" %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8117
  ins_encode %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8118
    int vector_len = 0;
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8119
    __ vsqrtps($dst$$XMMRegister, $mem$$Address, vector_len);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8120
  %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8121
  ins_pipe( pipe_slow );
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8122
%}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8123
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8124
instruct vsqrt4F_reg(vecX dst, vecX src) %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8125
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8126
  match(Set dst (SqrtVF src));
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8127
  format %{ "vsqrtps  $dst,$src\t! sqrt packed4F" %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8128
  ins_encode %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8129
    int vector_len = 0;
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8130
    __ vsqrtps($dst$$XMMRegister, $src$$XMMRegister, vector_len);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8131
  %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8132
  ins_pipe( pipe_slow );
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8133
%}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8134
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8135
instruct vsqrt4F_mem(vecX dst, memory mem) %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8136
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8137
  match(Set dst (SqrtVF (LoadVector mem)));
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8138
  format %{ "vsqrtps  $dst,$mem\t! sqrt packed4F" %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8139
  ins_encode %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8140
    int vector_len = 0;
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8141
    __ vsqrtps($dst$$XMMRegister, $mem$$Address, vector_len);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8142
  %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8143
  ins_pipe( pipe_slow );
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8144
%}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8145
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8146
instruct vsqrt8F_reg(vecY dst, vecY src) %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8147
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8148
  match(Set dst (SqrtVF src));
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8149
  format %{ "vsqrtps  $dst,$src\t! sqrt packed8F" %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8150
  ins_encode %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8151
    int vector_len = 1;
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8152
    __ vsqrtps($dst$$XMMRegister, $src$$XMMRegister, vector_len);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8153
  %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8154
  ins_pipe( pipe_slow );
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8155
%}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8156
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8157
instruct vsqrt8F_mem(vecY dst, memory mem) %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8158
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8159
  match(Set dst (SqrtVF (LoadVector mem)));
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8160
  format %{ "vsqrtps  $dst,$mem\t! sqrt packed8F" %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8161
  ins_encode %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8162
    int vector_len = 1;
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8163
    __ vsqrtps($dst$$XMMRegister, $mem$$Address, vector_len);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8164
  %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8165
  ins_pipe( pipe_slow );
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8166
%}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8167
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8168
instruct vsqrt16F_reg(vecZ dst, vecZ src) %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8169
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8170
  match(Set dst (SqrtVF src));
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8171
  format %{ "vsqrtps  $dst,$src\t! sqrt packed16F" %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8172
  ins_encode %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8173
    int vector_len = 2;
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8174
    __ vsqrtps($dst$$XMMRegister, $src$$XMMRegister, vector_len);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8175
  %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8176
  ins_pipe( pipe_slow );
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8177
%}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8178
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8179
instruct vsqrt16F_mem(vecZ dst, memory mem) %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8180
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8181
  match(Set dst (SqrtVF (LoadVector mem)));
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8182
  format %{ "vsqrtps  $dst,$mem\t! sqrt packed16F" %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8183
  ins_encode %{
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8184
    int vector_len = 2;
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8185
    __ vsqrtps($dst$$XMMRegister, $mem$$Address, vector_len);
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8186
  %}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8187
  ins_pipe( pipe_slow );
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8188
%}
22c9856fc2c2 8190800: Support vectorization of Math.sqrt() on floats
rlupusoru
parents: 47216
diff changeset
  8189
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8190
// ------------------------------ LeftShift -----------------------------------
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8191
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8192
// Shorts/Chars vector left shift
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8193
instruct vsll2S(vecS dst, vecS shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8194
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8195
  match(Set dst (LShiftVS dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8196
  format %{ "psllw   $dst,$shift\t! left shift packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8197
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8198
    __ psllw($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8199
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8200
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8201
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8202
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8203
instruct vsll2S_imm(vecS dst, immI8 shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8204
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8205
  match(Set dst (LShiftVS dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8206
  format %{ "psllw   $dst,$shift\t! left shift packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8207
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8208
    __ psllw($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8209
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8210
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8211
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8212
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8213
instruct vsll2S_reg(vecS dst, vecS src, vecS shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8214
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8215
  match(Set dst (LShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8216
  format %{ "vpsllw  $dst,$src,$shift\t! left shift packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8217
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8218
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8219
    __ vpsllw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8220
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8221
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8222
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8223
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8224
instruct vsll2S_reg_imm(vecS dst, vecS src, immI8 shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8225
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8226
  match(Set dst (LShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8227
  format %{ "vpsllw  $dst,$src,$shift\t! left shift packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8228
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8229
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8230
    __ vpsllw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8231
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8232
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8233
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8234
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8235
instruct vsll4S(vecD dst, vecS shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8236
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8237
  match(Set dst (LShiftVS dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8238
  format %{ "psllw   $dst,$shift\t! left shift packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8239
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8240
    __ psllw($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8241
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8242
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8243
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8244
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8245
instruct vsll4S_imm(vecD dst, immI8 shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8246
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8247
  match(Set dst (LShiftVS dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8248
  format %{ "psllw   $dst,$shift\t! left shift packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8249
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8250
    __ psllw($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8251
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8252
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8253
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8254
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8255
instruct vsll4S_reg(vecD dst, vecD src, vecS shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8256
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8257
  match(Set dst (LShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8258
  format %{ "vpsllw  $dst,$src,$shift\t! left shift packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8259
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8260
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8261
    __ vpsllw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8262
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8263
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8264
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8265
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8266
instruct vsll4S_reg_imm(vecD dst, vecD src, immI8 shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8267
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8268
  match(Set dst (LShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8269
  format %{ "vpsllw  $dst,$src,$shift\t! left shift packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8270
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8271
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8272
    __ vpsllw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8273
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8274
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8275
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8276
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8277
instruct vsll8S(vecX dst, vecS shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8278
  predicate(UseAVX == 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8279
  match(Set dst (LShiftVS dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8280
  format %{ "psllw   $dst,$shift\t! left shift packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8281
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8282
    __ psllw($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8283
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8284
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8285
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8286
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8287
instruct vsll8S_imm(vecX dst, immI8 shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8288
  predicate(UseAVX == 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8289
  match(Set dst (LShiftVS dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8290
  format %{ "psllw   $dst,$shift\t! left shift packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8291
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8292
    __ psllw($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8293
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8294
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8295
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8296
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8297
instruct vsll8S_reg(vecX dst, vecX src, vecS shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8298
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8299
  match(Set dst (LShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8300
  format %{ "vpsllw  $dst,$src,$shift\t! left shift packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8301
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8302
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8303
    __ vpsllw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8304
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8305
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8306
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8307
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8308
instruct vsll8S_reg_imm(vecX dst, vecX src, immI8 shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8309
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8310
  match(Set dst (LShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8311
  format %{ "vpsllw  $dst,$src,$shift\t! left shift packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8312
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8313
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8314
    __ vpsllw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8315
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8316
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8317
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8318
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8319
instruct vsll16S_reg(vecY dst, vecY src, vecS shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8320
  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8321
  match(Set dst (LShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8322
  format %{ "vpsllw  $dst,$src,$shift\t! left shift packed16S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8323
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8324
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8325
    __ vpsllw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8326
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8327
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8328
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8329
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8330
instruct vsll16S_reg_imm(vecY dst, vecY src, immI8 shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8331
  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8332
  match(Set dst (LShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8333
  format %{ "vpsllw  $dst,$src,$shift\t! left shift packed16S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8334
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8335
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8336
    __ vpsllw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8337
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8338
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8339
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8340
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8341
instruct vsll32S_reg(vecZ dst, vecZ src, vecS shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8342
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 32);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8343
  match(Set dst (LShiftVS src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8344
  format %{ "vpsllw  $dst,$src,$shift\t! left shift packed32S" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8345
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8346
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8347
    __ vpsllw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8348
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8349
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8350
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8351
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8352
instruct vsll32S_reg_imm(vecZ dst, vecZ src, immI8 shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8353
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 32);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8354
  match(Set dst (LShiftVS src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8355
  format %{ "vpsllw  $dst,$src,$shift\t! left shift packed32S" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8356
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8357
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8358
    __ vpsllw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8359
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8360
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8361
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8362
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8363
// Integers vector left shift
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8364
instruct vsll2I(vecD dst, vecS shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8365
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8366
  match(Set dst (LShiftVI dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8367
  format %{ "pslld   $dst,$shift\t! left shift packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8368
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8369
    __ pslld($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8370
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8371
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8372
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8373
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8374
instruct vsll2I_imm(vecD dst, immI8 shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8375
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8376
  match(Set dst (LShiftVI dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8377
  format %{ "pslld   $dst,$shift\t! left shift packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8378
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8379
    __ pslld($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8380
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8381
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8382
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8383
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8384
instruct vsll2I_reg(vecD dst, vecD src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8385
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8386
  match(Set dst (LShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8387
  format %{ "vpslld  $dst,$src,$shift\t! left shift packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8388
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8389
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8390
    __ vpslld($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8391
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8392
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8393
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8394
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8395
instruct vsll2I_reg_imm(vecD dst, vecD src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8396
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8397
  match(Set dst (LShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8398
  format %{ "vpslld  $dst,$src,$shift\t! left shift packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8399
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8400
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8401
    __ vpslld($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8402
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8403
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8404
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8405
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8406
instruct vsll4I(vecX dst, vecS shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8407
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8408
  match(Set dst (LShiftVI dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8409
  format %{ "pslld   $dst,$shift\t! left shift packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8410
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8411
    __ pslld($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8412
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8413
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8414
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8415
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8416
instruct vsll4I_imm(vecX dst, immI8 shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8417
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8418
  match(Set dst (LShiftVI dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8419
  format %{ "pslld   $dst,$shift\t! left shift packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8420
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8421
    __ pslld($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8422
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8423
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8424
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8425
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8426
instruct vsll4I_reg(vecX dst, vecX src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8427
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8428
  match(Set dst (LShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8429
  format %{ "vpslld  $dst,$src,$shift\t! left shift packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8430
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8431
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8432
    __ vpslld($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8433
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8434
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8435
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8436
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8437
instruct vsll4I_reg_imm(vecX dst, vecX src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8438
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8439
  match(Set dst (LShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8440
  format %{ "vpslld  $dst,$src,$shift\t! left shift packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8441
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8442
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8443
    __ vpslld($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8444
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8445
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8446
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8447
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8448
instruct vsll8I_reg(vecY dst, vecY src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8449
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8450
  match(Set dst (LShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8451
  format %{ "vpslld  $dst,$src,$shift\t! left shift packed8I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8452
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8453
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8454
    __ vpslld($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8455
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8456
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8457
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8458
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8459
instruct vsll8I_reg_imm(vecY dst, vecY src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8460
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8461
  match(Set dst (LShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8462
  format %{ "vpslld  $dst,$src,$shift\t! left shift packed8I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8463
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8464
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8465
    __ vpslld($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8466
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8467
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8468
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8469
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8470
instruct vsll16I_reg(vecZ dst, vecZ src, vecS shift) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8471
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8472
  match(Set dst (LShiftVI src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8473
  format %{ "vpslld  $dst,$src,$shift\t! left shift packed16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8474
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8475
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8476
    __ vpslld($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8477
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8478
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8479
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8480
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8481
instruct vsll16I_reg_imm(vecZ dst, vecZ src, immI8 shift) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8482
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8483
  match(Set dst (LShiftVI src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8484
  format %{ "vpslld  $dst,$src,$shift\t! left shift packed16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8485
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8486
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8487
    __ vpslld($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8488
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8489
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8490
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8491
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8492
// Longs vector left shift
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8493
instruct vsll2L(vecX dst, vecS shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8494
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8495
  match(Set dst (LShiftVL dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8496
  format %{ "psllq   $dst,$shift\t! left shift packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8497
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8498
    __ psllq($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8499
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8500
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8501
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8502
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8503
instruct vsll2L_imm(vecX dst, immI8 shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8504
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8505
  match(Set dst (LShiftVL dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8506
  format %{ "psllq   $dst,$shift\t! left shift packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8507
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8508
    __ psllq($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8509
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8510
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8511
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8512
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8513
instruct vsll2L_reg(vecX dst, vecX src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8514
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8515
  match(Set dst (LShiftVL src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8516
  format %{ "vpsllq  $dst,$src,$shift\t! left shift packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8517
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8518
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8519
    __ vpsllq($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8520
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8521
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8522
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8523
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8524
instruct vsll2L_reg_imm(vecX dst, vecX src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8525
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8526
  match(Set dst (LShiftVL src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8527
  format %{ "vpsllq  $dst,$src,$shift\t! left shift packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8528
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8529
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8530
    __ vpsllq($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8531
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8532
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8533
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8534
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8535
instruct vsll4L_reg(vecY dst, vecY src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8536
  predicate(UseAVX > 1 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8537
  match(Set dst (LShiftVL src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8538
  format %{ "vpsllq  $dst,$src,$shift\t! left shift packed4L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8539
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8540
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8541
    __ vpsllq($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8542
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8543
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8544
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8545
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8546
instruct vsll4L_reg_imm(vecY dst, vecY src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8547
  predicate(UseAVX > 1 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8548
  match(Set dst (LShiftVL src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8549
  format %{ "vpsllq  $dst,$src,$shift\t! left shift packed4L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8550
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8551
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8552
    __ vpsllq($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8553
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8554
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8555
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8556
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8557
instruct vsll8L_reg(vecZ dst, vecZ src, vecS shift) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8558
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8559
  match(Set dst (LShiftVL src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8560
  format %{ "vpsllq  $dst,$src,$shift\t! left shift packed8L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8561
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8562
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8563
    __ vpsllq($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8564
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8565
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8566
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8567
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8568
instruct vsll8L_reg_imm(vecZ dst, vecZ src, immI8 shift) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8569
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8570
  match(Set dst (LShiftVL src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8571
  format %{ "vpsllq  $dst,$src,$shift\t! left shift packed8L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8572
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8573
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8574
    __ vpsllq($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8575
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8576
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8577
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8578
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8579
// ----------------------- LogicalRightShift -----------------------------------
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8580
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8581
// Shorts vector logical right shift produces incorrect Java result
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8582
// for negative data because java code convert short value into int with
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8583
// sign extension before a shift. But char vectors are fine since chars are
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8584
// unsigned values.
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8585
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8586
instruct vsrl2S(vecS dst, vecS shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8587
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8588
  match(Set dst (URShiftVS dst shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8589
  format %{ "psrlw   $dst,$shift\t! logical right shift packed2S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8590
  ins_encode %{
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8591
    __ psrlw($dst$$XMMRegister, $shift$$XMMRegister);
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8592
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8593
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8594
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8595
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8596
instruct vsrl2S_imm(vecS dst, immI8 shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8597
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8598
  match(Set dst (URShiftVS dst shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8599
  format %{ "psrlw   $dst,$shift\t! logical right shift packed2S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8600
  ins_encode %{
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8601
    __ psrlw($dst$$XMMRegister, (int)$shift$$constant);
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8602
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8603
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8604
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8605
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8606
instruct vsrl2S_reg(vecS dst, vecS src, vecS shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8607
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8608
  match(Set dst (URShiftVS src shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8609
  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed2S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8610
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8611
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8612
    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8613
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8614
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8615
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8616
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8617
instruct vsrl2S_reg_imm(vecS dst, vecS src, immI8 shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8618
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8619
  match(Set dst (URShiftVS src shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8620
  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed2S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8621
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8622
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8623
    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8624
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8625
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8626
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8627
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8628
instruct vsrl4S(vecD dst, vecS shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8629
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8630
  match(Set dst (URShiftVS dst shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8631
  format %{ "psrlw   $dst,$shift\t! logical right shift packed4S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8632
  ins_encode %{
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8633
    __ psrlw($dst$$XMMRegister, $shift$$XMMRegister);
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8634
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8635
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8636
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8637
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8638
instruct vsrl4S_imm(vecD dst, immI8 shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8639
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8640
  match(Set dst (URShiftVS dst shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8641
  format %{ "psrlw   $dst,$shift\t! logical right shift packed4S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8642
  ins_encode %{
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8643
    __ psrlw($dst$$XMMRegister, (int)$shift$$constant);
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8644
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8645
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8646
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8647
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8648
instruct vsrl4S_reg(vecD dst, vecD src, vecS shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8649
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8650
  match(Set dst (URShiftVS src shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8651
  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed4S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8652
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8653
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8654
    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8655
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8656
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8657
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8658
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8659
instruct vsrl4S_reg_imm(vecD dst, vecD src, immI8 shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8660
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8661
  match(Set dst (URShiftVS src shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8662
  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed4S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8663
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8664
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8665
    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8666
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8667
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8668
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8669
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8670
instruct vsrl8S(vecX dst, vecS shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8671
  predicate(UseAVX == 0 && n->as_Vector()->length() == 8);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8672
  match(Set dst (URShiftVS dst shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8673
  format %{ "psrlw   $dst,$shift\t! logical right shift packed8S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8674
  ins_encode %{
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8675
    __ psrlw($dst$$XMMRegister, $shift$$XMMRegister);
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8676
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8677
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8678
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8679
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8680
instruct vsrl8S_imm(vecX dst, immI8 shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8681
  predicate(UseAVX == 0 && n->as_Vector()->length() == 8);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8682
  match(Set dst (URShiftVS dst shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8683
  format %{ "psrlw   $dst,$shift\t! logical right shift packed8S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8684
  ins_encode %{
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8685
    __ psrlw($dst$$XMMRegister, (int)$shift$$constant);
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8686
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8687
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8688
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8689
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8690
instruct vsrl8S_reg(vecX dst, vecX src, vecS shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8691
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8692
  match(Set dst (URShiftVS src shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8693
  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed8S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8694
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8695
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8696
    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8697
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8698
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8699
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8700
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8701
instruct vsrl8S_reg_imm(vecX dst, vecX src, immI8 shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8702
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8703
  match(Set dst (URShiftVS src shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8704
  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed8S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8705
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8706
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8707
    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8708
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8709
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8710
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8711
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8712
instruct vsrl16S_reg(vecY dst, vecY src, vecS shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8713
  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8714
  match(Set dst (URShiftVS src shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8715
  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed16S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8716
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8717
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8718
    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8719
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8720
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8721
%}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8722
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8723
instruct vsrl16S_reg_imm(vecY dst, vecY src, immI8 shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8724
  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8725
  match(Set dst (URShiftVS src shift));
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8726
  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed16S" %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8727
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8728
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8729
    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8730
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8731
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8732
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8733
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8734
instruct vsrl32S_reg(vecZ dst, vecZ src, vecS shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8735
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 32);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8736
  match(Set dst (URShiftVS src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8737
  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed32S" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8738
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8739
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8740
    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8741
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8742
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8743
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8744
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8745
instruct vsrl32S_reg_imm(vecZ dst, vecZ src, immI8 shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8746
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 32);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8747
  match(Set dst (URShiftVS src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8748
  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed32S" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8749
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8750
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8751
    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
14131
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8752
  %}
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8753
  ins_pipe( pipe_slow );
e376e3d428c9 8001183: incorrect results of char vectors right shift operaiton
kvn
parents: 13930
diff changeset
  8754
%}
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8755
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8756
// Integers vector logical right shift
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8757
instruct vsrl2I(vecD dst, vecS shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8758
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8759
  match(Set dst (URShiftVI dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8760
  format %{ "psrld   $dst,$shift\t! logical right shift packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8761
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8762
    __ psrld($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8763
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8764
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8765
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8766
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8767
instruct vsrl2I_imm(vecD dst, immI8 shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8768
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8769
  match(Set dst (URShiftVI dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8770
  format %{ "psrld   $dst,$shift\t! logical right shift packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8771
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8772
    __ psrld($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8773
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8774
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8775
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8776
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8777
instruct vsrl2I_reg(vecD dst, vecD src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8778
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8779
  match(Set dst (URShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8780
  format %{ "vpsrld  $dst,$src,$shift\t! logical right shift packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8781
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8782
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8783
    __ vpsrld($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8784
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8785
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8786
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8787
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8788
instruct vsrl2I_reg_imm(vecD dst, vecD src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8789
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8790
  match(Set dst (URShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8791
  format %{ "vpsrld  $dst,$src,$shift\t! logical right shift packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8792
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8793
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8794
    __ vpsrld($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8795
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8796
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8797
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8798
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8799
instruct vsrl4I(vecX dst, vecS shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8800
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8801
  match(Set dst (URShiftVI dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8802
  format %{ "psrld   $dst,$shift\t! logical right shift packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8803
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8804
    __ psrld($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8805
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8806
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8807
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8808
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8809
instruct vsrl4I_imm(vecX dst, immI8 shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8810
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8811
  match(Set dst (URShiftVI dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8812
  format %{ "psrld   $dst,$shift\t! logical right shift packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8813
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8814
    __ psrld($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8815
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8816
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8817
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8818
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8819
instruct vsrl4I_reg(vecX dst, vecX src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8820
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8821
  match(Set dst (URShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8822
  format %{ "vpsrld  $dst,$src,$shift\t! logical right shift packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8823
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8824
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8825
    __ vpsrld($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8826
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8827
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8828
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8829
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8830
instruct vsrl4I_reg_imm(vecX dst, vecX src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8831
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8832
  match(Set dst (URShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8833
  format %{ "vpsrld  $dst,$src,$shift\t! logical right shift packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8834
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8835
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8836
    __ vpsrld($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8837
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8838
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8839
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8840
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8841
instruct vsrl8I_reg(vecY dst, vecY src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8842
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8843
  match(Set dst (URShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8844
  format %{ "vpsrld  $dst,$src,$shift\t! logical right shift packed8I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8845
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8846
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8847
    __ vpsrld($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8848
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8849
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8850
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8851
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8852
instruct vsrl8I_reg_imm(vecY dst, vecY src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8853
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8854
  match(Set dst (URShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8855
  format %{ "vpsrld  $dst,$src,$shift\t! logical right shift packed8I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8856
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8857
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8858
    __ vpsrld($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8859
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8860
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8861
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8862
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8863
instruct vsrl16I_reg(vecZ dst, vecZ src, vecS shift) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8864
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8865
  match(Set dst (URShiftVI src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8866
  format %{ "vpsrld  $dst,$src,$shift\t! logical right shift packed16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8867
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8868
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8869
    __ vpsrld($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8870
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8871
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8872
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8873
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8874
instruct vsrl16I_reg_imm(vecZ dst, vecZ src, immI8 shift) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8875
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8876
  match(Set dst (URShiftVI src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8877
  format %{ "vpsrld  $dst,$src,$shift\t! logical right shift packed16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8878
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8879
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8880
    __ vpsrld($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8881
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8882
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8883
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8884
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8885
// Longs vector logical right shift
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8886
instruct vsrl2L(vecX dst, vecS shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8887
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8888
  match(Set dst (URShiftVL dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8889
  format %{ "psrlq   $dst,$shift\t! logical right shift packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8890
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8891
    __ psrlq($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8892
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8893
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8894
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8895
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8896
instruct vsrl2L_imm(vecX dst, immI8 shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8897
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8898
  match(Set dst (URShiftVL dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8899
  format %{ "psrlq   $dst,$shift\t! logical right shift packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8900
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8901
    __ psrlq($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8902
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8903
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8904
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8905
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8906
instruct vsrl2L_reg(vecX dst, vecX src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8907
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8908
  match(Set dst (URShiftVL src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8909
  format %{ "vpsrlq  $dst,$src,$shift\t! logical right shift packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8910
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8911
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8912
    __ vpsrlq($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8913
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8914
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8915
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8916
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8917
instruct vsrl2L_reg_imm(vecX dst, vecX src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8918
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8919
  match(Set dst (URShiftVL src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8920
  format %{ "vpsrlq  $dst,$src,$shift\t! logical right shift packed2L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8921
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8922
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8923
    __ vpsrlq($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8924
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8925
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8926
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8927
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8928
instruct vsrl4L_reg(vecY dst, vecY src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8929
  predicate(UseAVX > 1 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8930
  match(Set dst (URShiftVL src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8931
  format %{ "vpsrlq  $dst,$src,$shift\t! logical right shift packed4L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8932
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8933
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8934
    __ vpsrlq($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8935
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8936
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8937
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8938
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8939
instruct vsrl4L_reg_imm(vecY dst, vecY src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8940
  predicate(UseAVX > 1 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8941
  match(Set dst (URShiftVL src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8942
  format %{ "vpsrlq  $dst,$src,$shift\t! logical right shift packed4L" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8943
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8944
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8945
    __ vpsrlq($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8946
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8947
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8948
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8949
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8950
instruct vsrl8L_reg(vecZ dst, vecZ src, vecS shift) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8951
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8952
  match(Set dst (URShiftVL src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8953
  format %{ "vpsrlq  $dst,$src,$shift\t! logical right shift packed8L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8954
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8955
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8956
    __ vpsrlq($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8957
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8958
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8959
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8960
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8961
instruct vsrl8L_reg_imm(vecZ dst, vecZ src, immI8 shift) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8962
  predicate(UseAVX > 2 && n->as_Vector()->length() == 8);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8963
  match(Set dst (URShiftVL src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8964
  format %{ "vpsrlq  $dst,$src,$shift\t! logical right shift packed8L" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8965
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8966
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  8967
    __ vpsrlq($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8968
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8969
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8970
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8971
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8972
// ------------------- ArithmeticRightShift -----------------------------------
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8973
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8974
// Shorts/Chars vector arithmetic right shift
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  8975
instruct vsra2S(vecS dst, vecS shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  8976
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8977
  match(Set dst (RShiftVS dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8978
  format %{ "psraw   $dst,$shift\t! arithmetic right shift packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8979
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8980
    __ psraw($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8981
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8982
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8983
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8984
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8985
instruct vsra2S_imm(vecS dst, immI8 shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8986
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8987
  match(Set dst (RShiftVS dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8988
  format %{ "psraw   $dst,$shift\t! arithmetic right shift packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8989
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8990
    __ psraw($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8991
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8992
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8993
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8994
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8995
instruct vsra2S_reg(vecS dst, vecS src, vecS shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  8996
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8997
  match(Set dst (RShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8998
  format %{ "vpsraw  $dst,$src,$shift\t! arithmetic right shift packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  8999
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9000
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9001
    __ vpsraw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9002
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9003
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9004
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9005
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9006
instruct vsra2S_reg_imm(vecS dst, vecS src, immI8 shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9007
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9008
  match(Set dst (RShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9009
  format %{ "vpsraw  $dst,$src,$shift\t! arithmetic right shift packed2S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9010
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9011
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9012
    __ vpsraw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9013
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9014
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9015
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9016
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  9017
instruct vsra4S(vecD dst, vecS shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  9018
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9019
  match(Set dst (RShiftVS dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9020
  format %{ "psraw   $dst,$shift\t! arithmetic right shift packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9021
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9022
    __ psraw($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9023
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9024
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9025
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9026
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9027
instruct vsra4S_imm(vecD dst, immI8 shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  9028
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9029
  match(Set dst (RShiftVS dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9030
  format %{ "psraw   $dst,$shift\t! arithmetic right shift packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9031
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9032
    __ psraw($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9033
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9034
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9035
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9036
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9037
instruct vsra4S_reg(vecD dst, vecD src, vecS shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9038
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9039
  match(Set dst (RShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9040
  format %{ "vpsraw  $dst,$src,$shift\t! arithmetic right shift packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9041
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9042
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9043
    __ vpsraw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9044
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9045
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9046
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9047
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9048
instruct vsra4S_reg_imm(vecD dst, vecD src, immI8 shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9049
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9050
  match(Set dst (RShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9051
  format %{ "vpsraw  $dst,$src,$shift\t! arithmetic right shift packed4S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9052
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9053
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9054
    __ vpsraw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9055
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9056
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9057
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9058
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  9059
instruct vsra8S(vecX dst, vecS shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  9060
  predicate(UseAVX == 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9061
  match(Set dst (RShiftVS dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9062
  format %{ "psraw   $dst,$shift\t! arithmetic right shift packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9063
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9064
    __ psraw($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9065
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9066
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9067
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9068
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9069
instruct vsra8S_imm(vecX dst, immI8 shift) %{
34162
16b54851eaf6 8140779: Code generation fixes for avx512
iveresov
parents: 33469
diff changeset
  9070
  predicate(UseAVX == 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9071
  match(Set dst (RShiftVS dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9072
  format %{ "psraw   $dst,$shift\t! arithmetic right shift packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9073
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9074
    __ psraw($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9075
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9076
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9077
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9078
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9079
instruct vsra8S_reg(vecX dst, vecX src, vecS shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9080
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9081
  match(Set dst (RShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9082
  format %{ "vpsraw  $dst,$src,$shift\t! arithmetic right shift packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9083
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9084
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9085
    __ vpsraw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9086
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9087
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9088
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9089
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9090
instruct vsra8S_reg_imm(vecX dst, vecX src, immI8 shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9091
  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9092
  match(Set dst (RShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9093
  format %{ "vpsraw  $dst,$src,$shift\t! arithmetic right shift packed8S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9094
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9095
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9096
    __ vpsraw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9097
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9098
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9099
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9100
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9101
instruct vsra16S_reg(vecY dst, vecY src, vecS shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9102
  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9103
  match(Set dst (RShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9104
  format %{ "vpsraw  $dst,$src,$shift\t! arithmetic right shift packed16S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9105
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9106
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9107
    __ vpsraw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9108
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9109
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9110
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9111
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9112
instruct vsra16S_reg_imm(vecY dst, vecY src, immI8 shift) %{
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9113
  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9114
  match(Set dst (RShiftVS src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9115
  format %{ "vpsraw  $dst,$src,$shift\t! arithmetic right shift packed16S" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9116
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9117
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9118
    __ vpsraw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9119
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9120
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9121
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9122
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9123
instruct vsra32S_reg(vecZ dst, vecZ src, vecS shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9124
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 32);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9125
  match(Set dst (RShiftVS src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9126
  format %{ "vpsraw  $dst,$src,$shift\t! arithmetic right shift packed32S" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9127
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9128
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9129
    __ vpsraw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9130
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9131
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9132
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9133
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9134
instruct vsra32S_reg_imm(vecZ dst, vecZ src, immI8 shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9135
  predicate(UseAVX > 2 && VM_Version::supports_avx512bw() && n->as_Vector()->length() == 32);
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9136
  match(Set dst (RShiftVS src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9137
  format %{ "vpsraw  $dst,$src,$shift\t! arithmetic right shift packed32S" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9138
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9139
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9140
    __ vpsraw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9141
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9142
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9143
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9144
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9145
// Integers vector arithmetic right shift
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  9146
instruct vsra2I(vecD dst, vecS shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9147
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9148
  match(Set dst (RShiftVI dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9149
  format %{ "psrad   $dst,$shift\t! arithmetic right shift packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9150
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9151
    __ psrad($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9152
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9153
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9154
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9155
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9156
instruct vsra2I_imm(vecD dst, immI8 shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9157
  predicate(UseAVX == 0 && n->as_Vector()->length() == 2);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9158
  match(Set dst (RShiftVI dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9159
  format %{ "psrad   $dst,$shift\t! arithmetic right shift packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9160
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9161
    __ psrad($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9162
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9163
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9164
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9165
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  9166
instruct vsra2I_reg(vecD dst, vecD src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9167
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9168
  match(Set dst (RShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9169
  format %{ "vpsrad  $dst,$src,$shift\t! arithmetic right shift packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9170
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9171
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9172
    __ vpsrad($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9173
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9174
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9175
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9176
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9177
instruct vsra2I_reg_imm(vecD dst, vecD src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9178
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9179
  match(Set dst (RShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9180
  format %{ "vpsrad  $dst,$src,$shift\t! arithmetic right shift packed2I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9181
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9182
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9183
    __ vpsrad($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9184
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9185
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9186
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9187
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  9188
instruct vsra4I(vecX dst, vecS shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9189
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9190
  match(Set dst (RShiftVI dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9191
  format %{ "psrad   $dst,$shift\t! arithmetic right shift packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9192
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9193
    __ psrad($dst$$XMMRegister, $shift$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9194
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9195
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9196
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9197
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9198
instruct vsra4I_imm(vecX dst, immI8 shift) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9199
  predicate(UseAVX == 0 && n->as_Vector()->length() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9200
  match(Set dst (RShiftVI dst shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9201
  format %{ "psrad   $dst,$shift\t! arithmetic right shift packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9202
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9203
    __ psrad($dst$$XMMRegister, (int)$shift$$constant);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9204
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9205
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9206
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9207
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  9208
instruct vsra4I_reg(vecX dst, vecX src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9209
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9210
  match(Set dst (RShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9211
  format %{ "vpsrad  $dst,$src,$shift\t! arithmetic right shift packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9212
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9213
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9214
    __ vpsrad($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9215
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9216
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9217
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9218
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9219
instruct vsra4I_reg_imm(vecX dst, vecX src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9220
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9221
  match(Set dst (RShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9222
  format %{ "vpsrad  $dst,$src,$shift\t! arithmetic right shift packed4I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9223
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9224
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9225
    __ vpsrad($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9226
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9227
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9228
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9229
13930
8df570f94294 7201026: add vector for shift count
kvn
parents: 13886
diff changeset
  9230
instruct vsra8I_reg(vecY dst, vecY src, vecS shift) %{
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9231
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9232
  match(Set dst (RShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9233
  format %{ "vpsrad  $dst,$src,$shift\t! arithmetic right shift packed8I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9234
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9235
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9236
    __ vpsrad($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9237
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9238
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9239
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9240
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9241
instruct vsra8I_reg_imm(vecY dst, vecY src, immI8 shift) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9242
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9243
  match(Set dst (RShiftVI src shift));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9244
  format %{ "vpsrad  $dst,$src,$shift\t! arithmetic right shift packed8I" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9245
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9246
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9247
    __ vpsrad($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9248
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9249
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9250
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9251
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9252
instruct vsra16I_reg(vecZ dst, vecZ src, vecS shift) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9253
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9254
  match(Set dst (RShiftVI src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9255
  format %{ "vpsrad  $dst,$src,$shift\t! arithmetic right shift packed16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9256
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9257
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9258
    __ vpsrad($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9259
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9260
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9261
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9262
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9263
instruct vsra16I_reg_imm(vecZ dst, vecZ src, immI8 shift) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9264
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9265
  match(Set dst (RShiftVI src shift));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9266
  format %{ "vpsrad  $dst,$src,$shift\t! arithmetic right shift packed16I" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9267
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9268
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9269
    __ vpsrad($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9270
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9271
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9272
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9273
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9274
// There are no longs vector arithmetic right shift instructions.
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9275
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9276
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9277
// --------------------------------- AND --------------------------------------
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9278
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9279
instruct vand4B(vecS dst, vecS src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9280
  predicate(UseAVX == 0 && n->as_Vector()->length_in_bytes() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9281
  match(Set dst (AndV dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9282
  format %{ "pand    $dst,$src\t! and vectors (4 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9283
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9284
    __ pand($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9285
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9286
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9287
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9288
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9289
instruct vand4B_reg(vecS dst, vecS src1, vecS src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9290
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9291
  match(Set dst (AndV src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9292
  format %{ "vpand   $dst,$src1,$src2\t! and vectors (4 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9293
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9294
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9295
    __ vpand($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9296
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9297
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9298
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9299
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9300
instruct vand4B_mem(vecS dst, vecS src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9301
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 4);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9302
  match(Set dst (AndV src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9303
  format %{ "vpand   $dst,$src,$mem\t! and vectors (4 bytes)" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9304
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9305
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9306
    __ vpand($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9307
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9308
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9309
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9310
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9311
instruct vand8B(vecD dst, vecD src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9312
  predicate(UseAVX == 0 && n->as_Vector()->length_in_bytes() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9313
  match(Set dst (AndV dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9314
  format %{ "pand    $dst,$src\t! and vectors (8 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9315
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9316
    __ pand($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9317
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9318
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9319
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9320
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9321
instruct vand8B_reg(vecD dst, vecD src1, vecD src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9322
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9323
  match(Set dst (AndV src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9324
  format %{ "vpand   $dst,$src1,$src2\t! and vectors (8 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9325
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9326
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9327
    __ vpand($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9328
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9329
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9330
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9331
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9332
instruct vand8B_mem(vecD dst, vecD src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9333
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 8);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9334
  match(Set dst (AndV src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9335
  format %{ "vpand   $dst,$src,$mem\t! and vectors (8 bytes)" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9336
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9337
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9338
    __ vpand($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9339
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9340
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9341
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9342
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9343
instruct vand16B(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9344
  predicate(UseAVX == 0 && n->as_Vector()->length_in_bytes() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9345
  match(Set dst (AndV dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9346
  format %{ "pand    $dst,$src\t! and vectors (16 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9347
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9348
    __ pand($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9349
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9350
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9351
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9352
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9353
instruct vand16B_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9354
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 16);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9355
  match(Set dst (AndV src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9356
  format %{ "vpand   $dst,$src1,$src2\t! and vectors (16 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9357
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9358
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9359
    __ vpand($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9360
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9361
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9362
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9363
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9364
instruct vand16B_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9365
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 16);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9366
  match(Set dst (AndV src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9367
  format %{ "vpand   $dst,$src,$mem\t! and vectors (16 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9368
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9369
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9370
    __ vpand($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9371
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9372
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9373
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9374
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9375
instruct vand32B_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9376
  predicate(UseAVX > 1 && n->as_Vector()->length_in_bytes() == 32);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9377
  match(Set dst (AndV src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9378
  format %{ "vpand   $dst,$src1,$src2\t! and vectors (32 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9379
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9380
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9381
    __ vpand($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9382
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9383
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9384
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9385
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9386
instruct vand32B_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9387
  predicate(UseAVX > 1 && n->as_Vector()->length_in_bytes() == 32);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9388
  match(Set dst (AndV src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9389
  format %{ "vpand   $dst,$src,$mem\t! and vectors (32 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9390
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9391
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9392
    __ vpand($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9393
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9394
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9395
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9396
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9397
instruct vand64B_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9398
  predicate(UseAVX > 2 && n->as_Vector()->length_in_bytes() == 64);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9399
  match(Set dst (AndV src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9400
  format %{ "vpand   $dst,$src1,$src2\t! and vectors (64 bytes)" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9401
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9402
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9403
    __ vpand($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9404
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9405
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9406
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9407
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9408
instruct vand64B_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9409
  predicate(UseAVX > 2 && n->as_Vector()->length_in_bytes() == 64);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9410
  match(Set dst (AndV src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9411
  format %{ "vpand   $dst,$src,$mem\t! and vectors (64 bytes)" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9412
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9413
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9414
    __ vpand($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9415
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9416
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9417
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9418
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9419
// --------------------------------- OR ---------------------------------------
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9420
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9421
instruct vor4B(vecS dst, vecS src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9422
  predicate(UseAVX == 0 && n->as_Vector()->length_in_bytes() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9423
  match(Set dst (OrV dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9424
  format %{ "por     $dst,$src\t! or vectors (4 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9425
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9426
    __ por($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9427
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9428
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9429
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9430
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9431
instruct vor4B_reg(vecS dst, vecS src1, vecS src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9432
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9433
  match(Set dst (OrV src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9434
  format %{ "vpor    $dst,$src1,$src2\t! or vectors (4 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9435
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9436
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9437
    __ vpor($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9438
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9439
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9440
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9441
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9442
instruct vor4B_mem(vecS dst, vecS src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9443
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 4);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9444
  match(Set dst (OrV src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9445
  format %{ "vpor    $dst,$src,$mem\t! or vectors (4 bytes)" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9446
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9447
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9448
    __ vpor($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9449
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9450
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9451
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9452
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9453
instruct vor8B(vecD dst, vecD src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9454
  predicate(UseAVX == 0 && n->as_Vector()->length_in_bytes() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9455
  match(Set dst (OrV dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9456
  format %{ "por     $dst,$src\t! or vectors (8 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9457
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9458
    __ por($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9459
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9460
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9461
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9462
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9463
instruct vor8B_reg(vecD dst, vecD src1, vecD src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9464
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9465
  match(Set dst (OrV src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9466
  format %{ "vpor    $dst,$src1,$src2\t! or vectors (8 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9467
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9468
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9469
    __ vpor($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9470
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9471
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9472
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9473
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9474
instruct vor8B_mem(vecD dst, vecD src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9475
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 4);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9476
  match(Set dst (OrV src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9477
  format %{ "vpor    $dst,$src,$mem\t! or vectors (8 bytes)" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9478
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9479
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9480
    __ vpor($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9481
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9482
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9483
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9484
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9485
instruct vor16B(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9486
  predicate(UseAVX == 0 && n->as_Vector()->length_in_bytes() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9487
  match(Set dst (OrV dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9488
  format %{ "por     $dst,$src\t! or vectors (16 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9489
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9490
    __ por($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9491
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9492
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9493
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9494
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9495
instruct vor16B_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9496
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 16);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9497
  match(Set dst (OrV src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9498
  format %{ "vpor    $dst,$src1,$src2\t! or vectors (16 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9499
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9500
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9501
    __ vpor($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9502
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9503
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9504
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9505
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9506
instruct vor16B_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9507
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 16);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9508
  match(Set dst (OrV src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9509
  format %{ "vpor    $dst,$src,$mem\t! or vectors (16 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9510
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9511
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9512
    __ vpor($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9513
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9514
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9515
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9516
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9517
instruct vor32B_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9518
  predicate(UseAVX > 1 && n->as_Vector()->length_in_bytes() == 32);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9519
  match(Set dst (OrV src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9520
  format %{ "vpor    $dst,$src1,$src2\t! or vectors (32 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9521
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9522
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9523
    __ vpor($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9524
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9525
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9526
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9527
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9528
instruct vor32B_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9529
  predicate(UseAVX > 1 && n->as_Vector()->length_in_bytes() == 32);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9530
  match(Set dst (OrV src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9531
  format %{ "vpor    $dst,$src,$mem\t! or vectors (32 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9532
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9533
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9534
    __ vpor($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9535
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9536
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9537
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9538
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9539
instruct vor64B_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9540
  predicate(UseAVX > 2 && n->as_Vector()->length_in_bytes() == 64);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9541
  match(Set dst (OrV src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9542
  format %{ "vpor    $dst,$src1,$src2\t! or vectors (64 bytes)" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9543
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9544
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9545
    __ vpor($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9546
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9547
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9548
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9549
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9550
instruct vor64B_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9551
  predicate(UseAVX > 2 && n->as_Vector()->length_in_bytes() == 64);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9552
  match(Set dst (OrV src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9553
  format %{ "vpor    $dst,$src,$mem\t! or vectors (64 bytes)" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9554
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9555
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9556
    __ vpor($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9557
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9558
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9559
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9560
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9561
// --------------------------------- XOR --------------------------------------
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9562
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9563
instruct vxor4B(vecS dst, vecS src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9564
  predicate(UseAVX == 0 && n->as_Vector()->length_in_bytes() == 4);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9565
  match(Set dst (XorV dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9566
  format %{ "pxor    $dst,$src\t! xor vectors (4 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9567
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9568
    __ pxor($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9569
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9570
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9571
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9572
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9573
instruct vxor4B_reg(vecS dst, vecS src1, vecS src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9574
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 4);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9575
  match(Set dst (XorV src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9576
  format %{ "vpxor   $dst,$src1,$src2\t! xor vectors (4 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9577
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9578
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9579
    __ vpxor($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9580
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9581
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9582
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9583
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9584
instruct vxor4B_mem(vecS dst, vecS src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9585
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 4);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9586
  match(Set dst (XorV src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9587
  format %{ "vpxor   $dst,$src,$mem\t! xor vectors (4 bytes)" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9588
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9589
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9590
    __ vpxor($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9591
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9592
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9593
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9594
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9595
instruct vxor8B(vecD dst, vecD src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9596
  predicate(UseAVX == 0 && n->as_Vector()->length_in_bytes() == 8);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9597
  match(Set dst (XorV dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9598
  format %{ "pxor    $dst,$src\t! xor vectors (8 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9599
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9600
    __ pxor($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9601
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9602
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9603
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9604
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9605
instruct vxor8B_reg(vecD dst, vecD src1, vecD src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9606
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 8);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9607
  match(Set dst (XorV src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9608
  format %{ "vpxor   $dst,$src1,$src2\t! xor vectors (8 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9609
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9610
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9611
    __ vpxor($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9612
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9613
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9614
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9615
31410
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9616
instruct vxor8B_mem(vecD dst, vecD src, memory mem) %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9617
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 8);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9618
  match(Set dst (XorV src (LoadVector mem)));
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9619
  format %{ "vpxor   $dst,$src,$mem\t! xor vectors (8 bytes)" %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9620
  ins_encode %{
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9621
    int vector_len = 0;
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9622
    __ vpxor($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9623
  %}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9624
  ins_pipe( pipe_slow );
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9625
%}
2a222ae1205f 8081247: AVX 512 extended support
mcberg
parents: 30624
diff changeset
  9626
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9627
instruct vxor16B(vecX dst, vecX src) %{
51857
9978fea8a371 8210764: Update avx512 implementation
kvn
parents: 51633
diff changeset
  9628
  predicate(UseAVX == 0 && n->as_Vector()->length_in_bytes() == 16);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9629
  match(Set dst (XorV dst src));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9630
  format %{ "pxor    $dst,$src\t! xor vectors (16 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9631
  ins_encode %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9632
    __ pxor($dst$$XMMRegister, $src$$XMMRegister);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9633
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9634
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9635
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9636
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9637
instruct vxor16B_reg(vecX dst, vecX src1, vecX src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9638
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 16);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9639
  match(Set dst (XorV src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9640
  format %{ "vpxor   $dst,$src1,$src2\t! xor vectors (16 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9641
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9642
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9643
    __ vpxor($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9644
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9645
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9646
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9647
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9648
instruct vxor16B_mem(vecX dst, vecX src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9649
  predicate(UseAVX > 0 && n->as_Vector()->length_in_bytes() == 16);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9650
  match(Set dst (XorV src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9651
  format %{ "vpxor   $dst,$src,$mem\t! xor vectors (16 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9652
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9653
    int vector_len = 0;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9654
    __ vpxor($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9655
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9656
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9657
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9658
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9659
instruct vxor32B_reg(vecY dst, vecY src1, vecY src2) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9660
  predicate(UseAVX > 1 && n->as_Vector()->length_in_bytes() == 32);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9661
  match(Set dst (XorV src1 src2));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9662
  format %{ "vpxor   $dst,$src1,$src2\t! xor vectors (32 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9663
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9664
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9665
    __ vpxor($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
13485
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9666
  %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9667
  ins_pipe( pipe_slow );
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9668
%}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9669
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9670
instruct vxor32B_mem(vecY dst, vecY src, memory mem) %{
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9671
  predicate(UseAVX > 1 && n->as_Vector()->length_in_bytes() == 32);
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9672
  match(Set dst (XorV src (LoadVector mem)));
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9673
  format %{ "vpxor   $dst,$src,$mem\t! xor vectors (32 bytes)" %}
6c7faa516fc6 6340864: Implement vectorization optimizations in hotspot-server
kvn
parents: 13294
diff changeset
  9674
  ins_encode %{
30624
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9675
    int vector_len = 1;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9676
    __ vpxor($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9677
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9678
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9679
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9680
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9681
instruct vxor64B_reg(vecZ dst, vecZ src1, vecZ src2) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9682
  predicate(UseAVX > 2 && n->as_Vector()->length_in_bytes() == 64);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9683
  match(Set dst (XorV src1 src2));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9684
  format %{ "vpxor   $dst,$src1,$src2\t! xor vectors (64 bytes)" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9685
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9686
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9687
    __ vpxor($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9688
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9689
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9690
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9691
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9692
instruct vxor64B_mem(vecZ dst, vecZ src, memory mem) %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9693
  predicate(UseAVX > 2 && n->as_Vector()->length_in_bytes() == 64);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9694
  match(Set dst (XorV src (LoadVector mem)));
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9695
  format %{ "vpxor   $dst,$src,$mem\t! xor vectors (64 bytes)" %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9696
  ins_encode %{
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9697
    int vector_len = 2;
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9698
    __ vpxor($dst$$XMMRegister, $src$$XMMRegister, $mem$$Address, vector_len);
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9699
  %}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9700
  ins_pipe( pipe_slow );
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9701
%}
2e1803c8a26d 8076276: Add support for AVX512
kvn
parents: 30305
diff changeset
  9702
46528
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9703
// --------------------------------- FMA --------------------------------------
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9704
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9705
// a * b + c
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9706
instruct vfma2D_reg(vecX a, vecX b, vecX c) %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9707
  predicate(UseFMA && n->as_Vector()->length() == 2);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9708
  match(Set c (FmaVD  c (Binary a b)));
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9709
  format %{ "fmapd $a,$b,$c\t# $c = $a * $b + $c fma packed2D" %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9710
  ins_cost(150);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9711
  ins_encode %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9712
    int vector_len = 0;
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9713
    __ vfmad($c$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $c$$XMMRegister, vector_len);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9714
  %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9715
  ins_pipe( pipe_slow );
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9716
%}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9717
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9718
// a * b + c
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9719
instruct vfma2D_mem(vecX a, memory b, vecX c) %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9720
  predicate(UseFMA && n->as_Vector()->length() == 2);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9721
  match(Set c (FmaVD  c (Binary a (LoadVector b))));
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9722
  format %{ "fmapd $a,$b,$c\t# $c = $a * $b + $c fma packed2D" %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9723
  ins_cost(150);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9724
  ins_encode %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9725
    int vector_len = 0;
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9726
    __ vfmad($c$$XMMRegister, $a$$XMMRegister, $b$$Address, $c$$XMMRegister, vector_len);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9727
  %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9728
  ins_pipe( pipe_slow );
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9729
%}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9730
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9731
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9732
// a * b + c
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9733
instruct vfma4D_reg(vecY a, vecY b, vecY c) %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9734
  predicate(UseFMA && n->as_Vector()->length() == 4);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9735
  match(Set c (FmaVD  c (Binary a b)));
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9736
  format %{ "fmapd $a,$b,$c\t# $c = $a * $b + $c fma packed4D" %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9737
  ins_cost(150);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9738
  ins_encode %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9739
    int vector_len = 1;
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9740
    __ vfmad($c$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $c$$XMMRegister, vector_len);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9741
  %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9742
  ins_pipe( pipe_slow );
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9743
%}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9744
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9745
// a * b + c
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9746
instruct vfma4D_mem(vecY a, memory b, vecY c) %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9747
  predicate(UseFMA && n->as_Vector()->length() == 4);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9748
  match(Set c (FmaVD  c (Binary a (LoadVector b))));
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9749
  format %{ "fmapd $a,$b,$c\t# $c = $a * $b + $c fma packed4D" %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9750
  ins_cost(150);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9751
  ins_encode %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9752
    int vector_len = 1;
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9753
    __ vfmad($c$$XMMRegister, $a$$XMMRegister, $b$$Address, $c$$XMMRegister, vector_len);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9754
  %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9755
  ins_pipe( pipe_slow );
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9756
%}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9757
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9758
// a * b + c
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9759
instruct vfma8D_reg(vecZ a, vecZ b, vecZ c) %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9760
  predicate(UseFMA && n->as_Vector()->length() == 8);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9761
  match(Set c (FmaVD  c (Binary a b)));
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9762
  format %{ "fmapd $a,$b,$c\t# $c = $a * $b + $c fma packed8D" %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9763
  ins_cost(150);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9764
  ins_encode %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9765
    int vector_len = 2;
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9766
    __ vfmad($c$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $c$$XMMRegister, vector_len);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9767
  %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9768
  ins_pipe( pipe_slow );
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9769
%}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9770
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9771
// a * b + c
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9772
instruct vfma8D_mem(vecZ a, memory b, vecZ c) %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9773
  predicate(UseFMA && n->as_Vector()->length() == 8);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9774
  match(Set c (FmaVD  c (Binary a (LoadVector b))));
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9775
  format %{ "fmapd $a,$b,$c\t# $c = $a * $b + $c fma packed8D" %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9776
  ins_cost(150);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9777
  ins_encode %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9778
    int vector_len = 2;
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9779
    __ vfmad($c$$XMMRegister, $a$$XMMRegister, $b$$Address, $c$$XMMRegister, vector_len);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9780
  %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9781
  ins_pipe( pipe_slow );
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9782
%}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9783
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9784
// a * b + c
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9785
instruct vfma4F_reg(vecX a, vecX b, vecX c) %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9786
  predicate(UseFMA && n->as_Vector()->length() == 4);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9787
  match(Set c (FmaVF  c (Binary a b)));
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9788
  format %{ "fmaps $a,$b,$c\t# $c = $a * $b + $c fma packed4F" %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9789
  ins_cost(150);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9790
  ins_encode %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9791
    int vector_len = 0;
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9792
    __ vfmaf($c$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $c$$XMMRegister, vector_len);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9793
  %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9794
  ins_pipe( pipe_slow );
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9795
%}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9796
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9797
// a * b + c
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9798
instruct vfma4F_mem(vecX a, memory b, vecX c) %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9799
  predicate(UseFMA && n->as_Vector()->length() == 4);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9800
  match(Set c (FmaVF  c (Binary a (LoadVector b))));
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9801
  format %{ "fmaps $a,$b,$c\t# $c = $a * $b + $c fma packed4F" %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9802
  ins_cost(150);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9803
  ins_encode %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9804
    int vector_len = 0;
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9805
    __ vfmaf($c$$XMMRegister, $a$$XMMRegister, $b$$Address, $c$$XMMRegister, vector_len);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9806
  %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9807
  ins_pipe( pipe_slow );
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9808
%}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9809
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9810
// a * b + c
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9811
instruct vfma8F_reg(vecY a, vecY b, vecY c) %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9812
  predicate(UseFMA && n->as_Vector()->length() == 8);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9813
  match(Set c (FmaVF  c (Binary a b)));
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9814
  format %{ "fmaps $a,$b,$c\t# $c = $a * $b + $c fma packed8F" %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9815
  ins_cost(150);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9816
  ins_encode %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9817
    int vector_len = 1;
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9818
    __ vfmaf($c$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $c$$XMMRegister, vector_len);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9819
  %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9820
  ins_pipe( pipe_slow );
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9821
%}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9822
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9823
// a * b + c
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9824
instruct vfma8F_mem(vecY a, memory b, vecY c) %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9825
  predicate(UseFMA && n->as_Vector()->length() == 8);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9826
  match(Set c (FmaVF  c (Binary a (LoadVector b))));
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9827
  format %{ "fmaps $a,$b,$c\t# $c = $a * $b + $c fma packed8F" %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9828
  ins_cost(150);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9829
  ins_encode %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9830
    int vector_len = 1;
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9831
    __ vfmaf($c$$XMMRegister, $a$$XMMRegister, $b$$Address, $c$$XMMRegister, vector_len);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9832
  %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9833
  ins_pipe( pipe_slow );
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9834
%}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9835
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9836
// a * b + c
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9837
instruct vfma16F_reg(vecZ a, vecZ b, vecZ c) %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9838
  predicate(UseFMA && n->as_Vector()->length() == 16);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9839
  match(Set c (FmaVF  c (Binary a b)));
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9840
  format %{ "fmaps $a,$b,$c\t# $c = $a * $b + $c fma packed16F" %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9841
  ins_cost(150);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9842
  ins_encode %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9843
    int vector_len = 2;
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9844
    __ vfmaf($c$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $c$$XMMRegister, vector_len);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9845
  %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9846
  ins_pipe( pipe_slow );
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9847
%}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9848
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9849
// a * b + c
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9850
instruct vfma16F_mem(vecZ a, memory b, vecZ c) %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9851
  predicate(UseFMA && n->as_Vector()->length() == 16);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9852
  match(Set c (FmaVF  c (Binary a (LoadVector b))));
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9853
  format %{ "fmaps $a,$b,$c\t# $c = $a * $b + $c fma packed16F" %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9854
  ins_cost(150);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9855
  ins_encode %{
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9856
    int vector_len = 2;
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9857
    __ vfmaf($c$$XMMRegister, $a$$XMMRegister, $b$$Address, $c$$XMMRegister, vector_len);
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9858
  %}
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9859
  ins_pipe( pipe_slow );
cf0da758e7b5 8181616: FMA Vectorization on x86
vdeshpande
parents: 46525
diff changeset
  9860
%}
49384
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9861
52992
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9862
// --------------------------------- Vector Multiply Add --------------------------------------
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9863
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9864
instruct smuladd4S2I_reg(vecD dst, vecD src1) %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9865
  predicate(UseSSE >= 2 && UseAVX == 0 && n->as_Vector()->length() == 2);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9866
  match(Set dst (MulAddVS2VI dst src1));
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9867
  format %{ "pmaddwd $dst,$dst,$src1\t! muladd packed4Sto2I" %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9868
  ins_encode %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9869
    __ pmaddwd($dst$$XMMRegister, $src1$$XMMRegister);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9870
  %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9871
  ins_pipe( pipe_slow );
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9872
%}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9873
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9874
instruct vmuladd4S2I_reg(vecD dst, vecD src1, vecD src2) %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9875
  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9876
  match(Set dst (MulAddVS2VI src1 src2));
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9877
  format %{ "vpmaddwd $dst,$src1,$src2\t! muladd packed4Sto2I" %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9878
  ins_encode %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9879
    int vector_len = 0;
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9880
    __ vpmaddwd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9881
  %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9882
  ins_pipe( pipe_slow );
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9883
%}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9884
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9885
instruct smuladd8S4I_reg(vecX dst, vecX src1) %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9886
  predicate(UseSSE >= 2 && UseAVX == 0 && n->as_Vector()->length() == 4);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9887
  match(Set dst (MulAddVS2VI dst src1));
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9888
  format %{ "pmaddwd $dst,$dst,$src1\t! muladd packed8Sto4I" %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9889
  ins_encode %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9890
    __ pmaddwd($dst$$XMMRegister, $src1$$XMMRegister);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9891
  %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9892
  ins_pipe( pipe_slow );
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9893
%}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9894
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9895
instruct vmuladd8S4I_reg(vecX dst, vecX src1, vecX src2) %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9896
  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9897
  match(Set dst (MulAddVS2VI src1 src2));
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9898
  format %{ "vpmaddwd $dst,$src1,$src2\t! muladd packed8Sto4I" %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9899
  ins_encode %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9900
    int vector_len = 0;
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9901
    __ vpmaddwd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9902
  %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9903
  ins_pipe( pipe_slow );
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9904
%}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9905
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9906
instruct vmuladd16S8I_reg(vecY dst, vecY src1, vecY src2) %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9907
  predicate(UseAVX > 1 && n->as_Vector()->length() == 8);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9908
  match(Set dst (MulAddVS2VI src1 src2));
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9909
  format %{ "vpmaddwd $dst,$src1,$src2\t! muladd packed16Sto8I" %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9910
  ins_encode %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9911
    int vector_len = 1;
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9912
    __ vpmaddwd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9913
  %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9914
  ins_pipe( pipe_slow );
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9915
%}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9916
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9917
instruct vmuladd32S16I_reg(vecZ dst, vecZ src1, vecZ src2) %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9918
  predicate(UseAVX > 2 && n->as_Vector()->length() == 16);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9919
  match(Set dst (MulAddVS2VI src1 src2));
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9920
  format %{ "vpmaddwd $dst,$src1,$src2\t! muladd packed32Sto16I" %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9921
  ins_encode %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9922
    int vector_len = 2;
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9923
    __ vpmaddwd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9924
  %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9925
  ins_pipe( pipe_slow );
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9926
%}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9927
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9928
// --------------------------------- Vector Multiply Add Add ----------------------------------
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9929
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9930
instruct vmuladdadd4S2I_reg(vecD dst, vecD src1, vecD src2) %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9931
  predicate(VM_Version::supports_vnni() && UseAVX > 2 && n->as_Vector()->length() == 2);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9932
  match(Set dst (AddVI (MulAddVS2VI src1 src2) dst));
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9933
  format %{ "evpdpwssd $dst,$src1,$src2\t! muladdadd packed4Sto2I" %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9934
  ins_encode %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9935
    int vector_len = 0;
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9936
    __ evpdpwssd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9937
  %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9938
  ins_pipe( pipe_slow );
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9939
%}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9940
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9941
instruct vmuladdadd8S4I_reg(vecX dst, vecX src1, vecX src2) %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9942
  predicate(VM_Version::supports_vnni() && UseAVX > 2 && n->as_Vector()->length() == 4);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9943
  match(Set dst (AddVI (MulAddVS2VI src1 src2) dst));
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9944
  format %{ "evpdpwssd $dst,$src1,$src2\t! muladdadd packed8Sto4I" %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9945
  ins_encode %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9946
    int vector_len = 0;
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9947
    __ evpdpwssd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9948
  %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9949
  ins_pipe( pipe_slow );
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9950
%}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9951
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9952
instruct vmuladdadd16S8I_reg(vecY dst, vecY src1, vecY src2) %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9953
  predicate(VM_Version::supports_vnni() && UseAVX > 2 && n->as_Vector()->length() == 8);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9954
  match(Set dst (AddVI (MulAddVS2VI src1 src2) dst));
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9955
  format %{ "evpdpwssd $dst,$src1,$src2\t! muladdadd packed16Sto8I" %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9956
  ins_encode %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9957
    int vector_len = 1;
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9958
    __ evpdpwssd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9959
  %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9960
  ins_pipe( pipe_slow );
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9961
%}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9962
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9963
instruct vmuladdadd32S16I_reg(vecZ dst, vecZ src1, vecZ src2) %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9964
  predicate(VM_Version::supports_vnni() && UseAVX > 2 && n->as_Vector()->length() == 16);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9965
  match(Set dst (AddVI (MulAddVS2VI src1 src2) dst));
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9966
  format %{ "evpdpwssd $dst,$src1,$src2\t! muladdadd packed32Sto16I" %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9967
  ins_encode %{
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9968
    int vector_len = 2;
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9969
    __ evpdpwssd($dst$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister, vector_len);
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9970
  %}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9971
  ins_pipe( pipe_slow );
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9972
%}
4bb6e0871bf7 8214751: X86: Support for VNNI Instructions
vdeshpande
parents: 51996
diff changeset
  9973
49384
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9974
// --------------------------------- PopCount --------------------------------------
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9975
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9976
instruct vpopcount2I(vecD dst, vecD src) %{
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9977
  predicate(VM_Version::supports_vpopcntdq() && UsePopCountInstruction && n->as_Vector()->length() == 2);
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9978
  match(Set dst (PopCountVI src));
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9979
  format %{ "vpopcntd  $dst,$src\t! vector popcount packed2I" %}
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9980
  ins_encode %{
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9981
    int vector_len = 0;
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9982
    __ vpopcntd($dst$$XMMRegister, $src$$XMMRegister, vector_len);
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9983
  %}
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9984
  ins_pipe( pipe_slow );
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9985
%}
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9986
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9987
instruct vpopcount4I(vecX dst, vecX src) %{
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9988
  predicate(VM_Version::supports_vpopcntdq() && UsePopCountInstruction && n->as_Vector()->length() == 4);
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9989
  match(Set dst (PopCountVI src));
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9990
  format %{ "vpopcntd  $dst,$src\t! vector popcount packed4I" %}
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9991
  ins_encode %{
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9992
    int vector_len = 0;
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9993
    __ vpopcntd($dst$$XMMRegister, $src$$XMMRegister, vector_len);
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9994
  %}
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9995
  ins_pipe( pipe_slow );
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9996
%}
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9997
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9998
instruct vpopcount8I(vecY dst, vecY src) %{
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
  9999
  predicate(VM_Version::supports_vpopcntdq() && UsePopCountInstruction && n->as_Vector()->length() == 8);
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10000
  match(Set dst (PopCountVI src));
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10001
  format %{ "vpopcntd  $dst,$src\t! vector popcount packed8I" %}
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10002
  ins_encode %{
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10003
    int vector_len = 1;
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10004
    __ vpopcntd($dst$$XMMRegister, $src$$XMMRegister, vector_len);
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10005
  %}
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10006
  ins_pipe( pipe_slow );
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10007
%}
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10008
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10009
instruct vpopcount16I(vecZ dst, vecZ src) %{
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10010
  predicate(VM_Version::supports_vpopcntdq() && UsePopCountInstruction && n->as_Vector()->length() == 16);
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10011
  match(Set dst (PopCountVI src));
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10012
  format %{ "vpopcntd  $dst,$src\t! vector popcount packed16I" %}
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10013
  ins_encode %{
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10014
    int vector_len = 2;
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10015
    __ vpopcntd($dst$$XMMRegister, $src$$XMMRegister, vector_len);
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10016
  %}
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10017
  ins_pipe( pipe_slow );
b242a1e3f9cf 8199421: Add support for vector popcount
rlupusoru
parents: 48309
diff changeset
 10018
%}