hotspot/src/cpu/aarch64/vm/aarch64.ad
author aph
Fri, 05 May 2017 17:29:57 +0100
changeset 45053 bd6d9a8a0b61
parent 44332 5c8abd7d8b49
child 45054 c09733aaf97f
permissions -rw-r--r--
8179701: AArch64: Reinstate FP as an allocatable register Reviewed-by: roland

//
// Copyright (c) 2003, 2016, Oracle and/or its affiliates. All rights reserved.
// Copyright (c) 2014, Red Hat Inc. All rights reserved.
// DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
//
// This code is free software; you can redistribute it and/or modify it
// under the terms of the GNU General Public License version 2 only, as
// published by the Free Software Foundation.
//
// This code is distributed in the hope that it will be useful, but WITHOUT
// ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
// FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
// version 2 for more details (a copy is included in the LICENSE file that
// accompanied this code).
//
// You should have received a copy of the GNU General Public License version
// 2 along with this work; if not, write to the Free Software Foundation,
// Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
//
// Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
// or visit www.oracle.com if you need additional information or have any
// questions.
//
//

// AArch64 Architecture Description File

//----------REGISTER DEFINITION BLOCK------------------------------------------
// This information is used by the matcher and the register allocator to
// describe individual registers and classes of registers within the target
// archtecture.

register %{
//----------Architecture Description Register Definitions----------------------
// General Registers
// "reg_def"  name ( register save type, C convention save type,
//                   ideal register type, encoding );
// Register Save Types:
//
// NS  = No-Save:       The register allocator assumes that these registers
//                      can be used without saving upon entry to the method, &
//                      that they do not need to be saved at call sites.
//
// SOC = Save-On-Call:  The register allocator assumes that these registers
//                      can be used without saving upon entry to the method,
//                      but that they must be saved at call sites.
//
// SOE = Save-On-Entry: The register allocator assumes that these registers
//                      must be saved before using them upon entry to the
//                      method, but they do not need to be saved at call
//                      sites.
//
// AS  = Always-Save:   The register allocator assumes that these registers
//                      must be saved before using them upon entry to the
//                      method, & that they must be saved at call sites.
//
// Ideal Register Type is used to determine how to save & restore a
// register.  Op_RegI will get spilled with LoadI/StoreI, Op_RegP will get
// spilled with LoadP/StoreP.  If the register supports both, use Op_RegI.
//
// The encoding number is the actual bit-pattern placed into the opcodes.

// We must define the 64 bit int registers in two 32 bit halves, the
// real lower register and a virtual upper half register. upper halves
// are used by the register allocator but are not actually supplied as
// operands to memory ops.
//
// follow the C1 compiler in making registers
//
//   r0-r7,r10-r26 volatile (caller save)
//   r27-r32 system (no save, no allocate)
//   r8-r9 invisible to the allocator (so we can use them as scratch regs)
//
// as regards Java usage. we don't use any callee save registers
// because this makes it difficult to de-optimise a frame (see comment
// in x86 implementation of Deoptimization::unwind_callee_save_values)
//

// General Registers

reg_def R0      ( SOC, SOC, Op_RegI,  0, r0->as_VMReg()         );
reg_def R0_H    ( SOC, SOC, Op_RegI,  0, r0->as_VMReg()->next() );
reg_def R1      ( SOC, SOC, Op_RegI,  1, r1->as_VMReg()         );
reg_def R1_H    ( SOC, SOC, Op_RegI,  1, r1->as_VMReg()->next() );
reg_def R2      ( SOC, SOC, Op_RegI,  2, r2->as_VMReg()         );
reg_def R2_H    ( SOC, SOC, Op_RegI,  2, r2->as_VMReg()->next() );
reg_def R3      ( SOC, SOC, Op_RegI,  3, r3->as_VMReg()         );
reg_def R3_H    ( SOC, SOC, Op_RegI,  3, r3->as_VMReg()->next() );
reg_def R4      ( SOC, SOC, Op_RegI,  4, r4->as_VMReg()         );
reg_def R4_H    ( SOC, SOC, Op_RegI,  4, r4->as_VMReg()->next() );
reg_def R5      ( SOC, SOC, Op_RegI,  5, r5->as_VMReg()         );
reg_def R5_H    ( SOC, SOC, Op_RegI,  5, r5->as_VMReg()->next() );
reg_def R6      ( SOC, SOC, Op_RegI,  6, r6->as_VMReg()         );
reg_def R6_H    ( SOC, SOC, Op_RegI,  6, r6->as_VMReg()->next() );
reg_def R7      ( SOC, SOC, Op_RegI,  7, r7->as_VMReg()         );
reg_def R7_H    ( SOC, SOC, Op_RegI,  7, r7->as_VMReg()->next() );
reg_def R10     ( SOC, SOC, Op_RegI, 10, r10->as_VMReg()        );
reg_def R10_H   ( SOC, SOC, Op_RegI, 10, r10->as_VMReg()->next());
reg_def R11     ( SOC, SOC, Op_RegI, 11, r11->as_VMReg()        );
reg_def R11_H   ( SOC, SOC, Op_RegI, 11, r11->as_VMReg()->next());
reg_def R12     ( SOC, SOC, Op_RegI, 12, r12->as_VMReg()        );
reg_def R12_H   ( SOC, SOC, Op_RegI, 12, r12->as_VMReg()->next());
reg_def R13     ( SOC, SOC, Op_RegI, 13, r13->as_VMReg()        );
reg_def R13_H   ( SOC, SOC, Op_RegI, 13, r13->as_VMReg()->next());
reg_def R14     ( SOC, SOC, Op_RegI, 14, r14->as_VMReg()        );
reg_def R14_H   ( SOC, SOC, Op_RegI, 14, r14->as_VMReg()->next());
reg_def R15     ( SOC, SOC, Op_RegI, 15, r15->as_VMReg()        );
reg_def R15_H   ( SOC, SOC, Op_RegI, 15, r15->as_VMReg()->next());
reg_def R16     ( SOC, SOC, Op_RegI, 16, r16->as_VMReg()        );
reg_def R16_H   ( SOC, SOC, Op_RegI, 16, r16->as_VMReg()->next());
reg_def R17     ( SOC, SOC, Op_RegI, 17, r17->as_VMReg()        );
reg_def R17_H   ( SOC, SOC, Op_RegI, 17, r17->as_VMReg()->next());
reg_def R18     ( SOC, SOC, Op_RegI, 18, r18->as_VMReg()        );
reg_def R18_H   ( SOC, SOC, Op_RegI, 18, r18->as_VMReg()->next());
reg_def R19     ( SOC, SOE, Op_RegI, 19, r19->as_VMReg()        );
reg_def R19_H   ( SOC, SOE, Op_RegI, 19, r19->as_VMReg()->next());
reg_def R20     ( SOC, SOE, Op_RegI, 20, r20->as_VMReg()        ); // caller esp
reg_def R20_H   ( SOC, SOE, Op_RegI, 20, r20->as_VMReg()->next());
reg_def R21     ( SOC, SOE, Op_RegI, 21, r21->as_VMReg()        );
reg_def R21_H   ( SOC, SOE, Op_RegI, 21, r21->as_VMReg()->next());
reg_def R22     ( SOC, SOE, Op_RegI, 22, r22->as_VMReg()        );
reg_def R22_H   ( SOC, SOE, Op_RegI, 22, r22->as_VMReg()->next());
reg_def R23     ( SOC, SOE, Op_RegI, 23, r23->as_VMReg()        );
reg_def R23_H   ( SOC, SOE, Op_RegI, 23, r23->as_VMReg()->next());
reg_def R24     ( SOC, SOE, Op_RegI, 24, r24->as_VMReg()        );
reg_def R24_H   ( SOC, SOE, Op_RegI, 24, r24->as_VMReg()->next());
reg_def R25     ( SOC, SOE, Op_RegI, 25, r25->as_VMReg()        );
reg_def R25_H   ( SOC, SOE, Op_RegI, 25, r25->as_VMReg()->next());
reg_def R26     ( SOC, SOE, Op_RegI, 26, r26->as_VMReg()        );
reg_def R26_H   ( SOC, SOE, Op_RegI, 26, r26->as_VMReg()->next());
reg_def R27     (  NS, SOE, Op_RegI, 27, r27->as_VMReg()        ); // heapbase
reg_def R27_H   (  NS, SOE, Op_RegI, 27, r27->as_VMReg()->next());
reg_def R28     (  NS, SOE, Op_RegI, 28, r28->as_VMReg()        ); // thread
reg_def R28_H   (  NS, SOE, Op_RegI, 28, r28->as_VMReg()->next());
reg_def R29     (  NS,  NS, Op_RegI, 29, r29->as_VMReg()        ); // fp
reg_def R29_H   (  NS,  NS, Op_RegI, 29, r29->as_VMReg()->next());
reg_def R30     (  NS,  NS, Op_RegI, 30, r30->as_VMReg()        ); // lr
reg_def R30_H   (  NS,  NS, Op_RegI, 30, r30->as_VMReg()->next());
reg_def R31     (  NS,  NS, Op_RegI, 31, r31_sp->as_VMReg()     ); // sp
reg_def R31_H   (  NS,  NS, Op_RegI, 31, r31_sp->as_VMReg()->next());

// ----------------------------
// Float/Double Registers
// ----------------------------

// Double Registers

// The rules of ADL require that double registers be defined in pairs.
// Each pair must be two 32-bit values, but not necessarily a pair of
// single float registers. In each pair, ADLC-assigned register numbers
// must be adjacent, with the lower number even. Finally, when the
// CPU stores such a register pair to memory, the word associated with
// the lower ADLC-assigned number must be stored to the lower address.

// AArch64 has 32 floating-point registers. Each can store a vector of
// single or double precision floating-point values up to 8 * 32
// floats, 4 * 64 bit floats or 2 * 128 bit floats.  We currently only
// use the first float or double element of the vector.

// for Java use float registers v0-v15 are always save on call whereas
// the platform ABI treats v8-v15 as callee save). float registers
// v16-v31 are SOC as per the platform spec

  reg_def V0   ( SOC, SOC, Op_RegF,  0, v0->as_VMReg()          );
  reg_def V0_H ( SOC, SOC, Op_RegF,  0, v0->as_VMReg()->next()  );
  reg_def V0_J ( SOC, SOC, Op_RegF,  0, v0->as_VMReg()->next(2) );
  reg_def V0_K ( SOC, SOC, Op_RegF,  0, v0->as_VMReg()->next(3) );

  reg_def V1   ( SOC, SOC, Op_RegF,  1, v1->as_VMReg()          );
  reg_def V1_H ( SOC, SOC, Op_RegF,  1, v1->as_VMReg()->next()  );
  reg_def V1_J ( SOC, SOC, Op_RegF,  1, v1->as_VMReg()->next(2) );
  reg_def V1_K ( SOC, SOC, Op_RegF,  1, v1->as_VMReg()->next(3) );

  reg_def V2   ( SOC, SOC, Op_RegF,  2, v2->as_VMReg()          );
  reg_def V2_H ( SOC, SOC, Op_RegF,  2, v2->as_VMReg()->next()  );
  reg_def V2_J ( SOC, SOC, Op_RegF,  2, v2->as_VMReg()->next(2) );
  reg_def V2_K ( SOC, SOC, Op_RegF,  2, v2->as_VMReg()->next(3) );

  reg_def V3   ( SOC, SOC, Op_RegF,  3, v3->as_VMReg()          );
  reg_def V3_H ( SOC, SOC, Op_RegF,  3, v3->as_VMReg()->next()  );
  reg_def V3_J ( SOC, SOC, Op_RegF,  3, v3->as_VMReg()->next(2) );
  reg_def V3_K ( SOC, SOC, Op_RegF,  3, v3->as_VMReg()->next(3) );

  reg_def V4   ( SOC, SOC, Op_RegF,  4, v4->as_VMReg()          );
  reg_def V4_H ( SOC, SOC, Op_RegF,  4, v4->as_VMReg()->next()  );
  reg_def V4_J ( SOC, SOC, Op_RegF,  4, v4->as_VMReg()->next(2) );
  reg_def V4_K ( SOC, SOC, Op_RegF,  4, v4->as_VMReg()->next(3) );

  reg_def V5   ( SOC, SOC, Op_RegF,  5, v5->as_VMReg()          );
  reg_def V5_H ( SOC, SOC, Op_RegF,  5, v5->as_VMReg()->next()  );
  reg_def V5_J ( SOC, SOC, Op_RegF,  5, v5->as_VMReg()->next(2) );
  reg_def V5_K ( SOC, SOC, Op_RegF,  5, v5->as_VMReg()->next(3) );

  reg_def V6   ( SOC, SOC, Op_RegF,  6, v6->as_VMReg()          );
  reg_def V6_H ( SOC, SOC, Op_RegF,  6, v6->as_VMReg()->next()  );
  reg_def V6_J ( SOC, SOC, Op_RegF,  6, v6->as_VMReg()->next(2) );
  reg_def V6_K ( SOC, SOC, Op_RegF,  6, v6->as_VMReg()->next(3) );

  reg_def V7   ( SOC, SOC, Op_RegF,  7, v7->as_VMReg()          );
  reg_def V7_H ( SOC, SOC, Op_RegF,  7, v7->as_VMReg()->next()  );
  reg_def V7_J ( SOC, SOC, Op_RegF,  7, v7->as_VMReg()->next(2) );
  reg_def V7_K ( SOC, SOC, Op_RegF,  7, v7->as_VMReg()->next(3) );

  reg_def V8   ( SOC, SOC, Op_RegF,  8, v8->as_VMReg()          );
  reg_def V8_H ( SOC, SOC, Op_RegF,  8, v8->as_VMReg()->next()  );
  reg_def V8_J ( SOC, SOC, Op_RegF,  8, v8->as_VMReg()->next(2) );
  reg_def V8_K ( SOC, SOC, Op_RegF,  8, v8->as_VMReg()->next(3) );

  reg_def V9   ( SOC, SOC, Op_RegF,  9, v9->as_VMReg()          );
  reg_def V9_H ( SOC, SOC, Op_RegF,  9, v9->as_VMReg()->next()  );
  reg_def V9_J ( SOC, SOC, Op_RegF,  9, v9->as_VMReg()->next(2) );
  reg_def V9_K ( SOC, SOC, Op_RegF,  9, v9->as_VMReg()->next(3) );

  reg_def V10  ( SOC, SOC, Op_RegF, 10, v10->as_VMReg()         );
  reg_def V10_H( SOC, SOC, Op_RegF, 10, v10->as_VMReg()->next() );
  reg_def V10_J( SOC, SOC, Op_RegF, 10, v10->as_VMReg()->next(2));
  reg_def V10_K( SOC, SOC, Op_RegF, 10, v10->as_VMReg()->next(3));

  reg_def V11  ( SOC, SOC, Op_RegF, 11, v11->as_VMReg()         );
  reg_def V11_H( SOC, SOC, Op_RegF, 11, v11->as_VMReg()->next() );
  reg_def V11_J( SOC, SOC, Op_RegF, 11, v11->as_VMReg()->next(2));
  reg_def V11_K( SOC, SOC, Op_RegF, 11, v11->as_VMReg()->next(3));

  reg_def V12  ( SOC, SOC, Op_RegF, 12, v12->as_VMReg()         );
  reg_def V12_H( SOC, SOC, Op_RegF, 12, v12->as_VMReg()->next() );
  reg_def V12_J( SOC, SOC, Op_RegF, 12, v12->as_VMReg()->next(2));
  reg_def V12_K( SOC, SOC, Op_RegF, 12, v12->as_VMReg()->next(3));

  reg_def V13  ( SOC, SOC, Op_RegF, 13, v13->as_VMReg()         );
  reg_def V13_H( SOC, SOC, Op_RegF, 13, v13->as_VMReg()->next() );
  reg_def V13_J( SOC, SOC, Op_RegF, 13, v13->as_VMReg()->next(2));
  reg_def V13_K( SOC, SOC, Op_RegF, 13, v13->as_VMReg()->next(3));

  reg_def V14  ( SOC, SOC, Op_RegF, 14, v14->as_VMReg()         );
  reg_def V14_H( SOC, SOC, Op_RegF, 14, v14->as_VMReg()->next() );
  reg_def V14_J( SOC, SOC, Op_RegF, 14, v14->as_VMReg()->next(2));
  reg_def V14_K( SOC, SOC, Op_RegF, 14, v14->as_VMReg()->next(3));

  reg_def V15  ( SOC, SOC, Op_RegF, 15, v15->as_VMReg()         );
  reg_def V15_H( SOC, SOC, Op_RegF, 15, v15->as_VMReg()->next() );
  reg_def V15_J( SOC, SOC, Op_RegF, 15, v15->as_VMReg()->next(2));
  reg_def V15_K( SOC, SOC, Op_RegF, 15, v15->as_VMReg()->next(3));

  reg_def V16  ( SOC, SOC, Op_RegF, 16, v16->as_VMReg()         );
  reg_def V16_H( SOC, SOC, Op_RegF, 16, v16->as_VMReg()->next() );
  reg_def V16_J( SOC, SOC, Op_RegF, 16, v16->as_VMReg()->next(2));
  reg_def V16_K( SOC, SOC, Op_RegF, 16, v16->as_VMReg()->next(3));

  reg_def V17  ( SOC, SOC, Op_RegF, 17, v17->as_VMReg()         );
  reg_def V17_H( SOC, SOC, Op_RegF, 17, v17->as_VMReg()->next() );
  reg_def V17_J( SOC, SOC, Op_RegF, 17, v17->as_VMReg()->next(2));
  reg_def V17_K( SOC, SOC, Op_RegF, 17, v17->as_VMReg()->next(3));

  reg_def V18  ( SOC, SOC, Op_RegF, 18, v18->as_VMReg()         );
  reg_def V18_H( SOC, SOC, Op_RegF, 18, v18->as_VMReg()->next() );
  reg_def V18_J( SOC, SOC, Op_RegF, 18, v18->as_VMReg()->next(2));
  reg_def V18_K( SOC, SOC, Op_RegF, 18, v18->as_VMReg()->next(3));

  reg_def V19  ( SOC, SOC, Op_RegF, 19, v19->as_VMReg()         );
  reg_def V19_H( SOC, SOC, Op_RegF, 19, v19->as_VMReg()->next() );
  reg_def V19_J( SOC, SOC, Op_RegF, 19, v19->as_VMReg()->next(2));
  reg_def V19_K( SOC, SOC, Op_RegF, 19, v19->as_VMReg()->next(3));

  reg_def V20  ( SOC, SOC, Op_RegF, 20, v20->as_VMReg()         );
  reg_def V20_H( SOC, SOC, Op_RegF, 20, v20->as_VMReg()->next() );
  reg_def V20_J( SOC, SOC, Op_RegF, 20, v20->as_VMReg()->next(2));
  reg_def V20_K( SOC, SOC, Op_RegF, 20, v20->as_VMReg()->next(3));

  reg_def V21  ( SOC, SOC, Op_RegF, 21, v21->as_VMReg()         );
  reg_def V21_H( SOC, SOC, Op_RegF, 21, v21->as_VMReg()->next() );
  reg_def V21_J( SOC, SOC, Op_RegF, 21, v21->as_VMReg()->next(2));
  reg_def V21_K( SOC, SOC, Op_RegF, 21, v21->as_VMReg()->next(3));

  reg_def V22  ( SOC, SOC, Op_RegF, 22, v22->as_VMReg()         );
  reg_def V22_H( SOC, SOC, Op_RegF, 22, v22->as_VMReg()->next() );
  reg_def V22_J( SOC, SOC, Op_RegF, 22, v22->as_VMReg()->next(2));
  reg_def V22_K( SOC, SOC, Op_RegF, 22, v22->as_VMReg()->next(3));

  reg_def V23  ( SOC, SOC, Op_RegF, 23, v23->as_VMReg()         );
  reg_def V23_H( SOC, SOC, Op_RegF, 23, v23->as_VMReg()->next() );
  reg_def V23_J( SOC, SOC, Op_RegF, 23, v23->as_VMReg()->next(2));
  reg_def V23_K( SOC, SOC, Op_RegF, 23, v23->as_VMReg()->next(3));

  reg_def V24  ( SOC, SOC, Op_RegF, 24, v24->as_VMReg()         );
  reg_def V24_H( SOC, SOC, Op_RegF, 24, v24->as_VMReg()->next() );
  reg_def V24_J( SOC, SOC, Op_RegF, 24, v24->as_VMReg()->next(2));
  reg_def V24_K( SOC, SOC, Op_RegF, 24, v24->as_VMReg()->next(3));

  reg_def V25  ( SOC, SOC, Op_RegF, 25, v25->as_VMReg()         );
  reg_def V25_H( SOC, SOC, Op_RegF, 25, v25->as_VMReg()->next() );
  reg_def V25_J( SOC, SOC, Op_RegF, 25, v25->as_VMReg()->next(2));
  reg_def V25_K( SOC, SOC, Op_RegF, 25, v25->as_VMReg()->next(3));

  reg_def V26  ( SOC, SOC, Op_RegF, 26, v26->as_VMReg()         );
  reg_def V26_H( SOC, SOC, Op_RegF, 26, v26->as_VMReg()->next() );
  reg_def V26_J( SOC, SOC, Op_RegF, 26, v26->as_VMReg()->next(2));
  reg_def V26_K( SOC, SOC, Op_RegF, 26, v26->as_VMReg()->next(3));

  reg_def V27  ( SOC, SOC, Op_RegF, 27, v27->as_VMReg()         );
  reg_def V27_H( SOC, SOC, Op_RegF, 27, v27->as_VMReg()->next() );
  reg_def V27_J( SOC, SOC, Op_RegF, 27, v27->as_VMReg()->next(2));
  reg_def V27_K( SOC, SOC, Op_RegF, 27, v27->as_VMReg()->next(3));

  reg_def V28  ( SOC, SOC, Op_RegF, 28, v28->as_VMReg()         );
  reg_def V28_H( SOC, SOC, Op_RegF, 28, v28->as_VMReg()->next() );
  reg_def V28_J( SOC, SOC, Op_RegF, 28, v28->as_VMReg()->next(2));
  reg_def V28_K( SOC, SOC, Op_RegF, 28, v28->as_VMReg()->next(3));

  reg_def V29  ( SOC, SOC, Op_RegF, 29, v29->as_VMReg()         );
  reg_def V29_H( SOC, SOC, Op_RegF, 29, v29->as_VMReg()->next() );
  reg_def V29_J( SOC, SOC, Op_RegF, 29, v29->as_VMReg()->next(2));
  reg_def V29_K( SOC, SOC, Op_RegF, 29, v29->as_VMReg()->next(3));

  reg_def V30  ( SOC, SOC, Op_RegF, 30, v30->as_VMReg()         );
  reg_def V30_H( SOC, SOC, Op_RegF, 30, v30->as_VMReg()->next() );
  reg_def V30_J( SOC, SOC, Op_RegF, 30, v30->as_VMReg()->next(2));
  reg_def V30_K( SOC, SOC, Op_RegF, 30, v30->as_VMReg()->next(3));

  reg_def V31  ( SOC, SOC, Op_RegF, 31, v31->as_VMReg()         );
  reg_def V31_H( SOC, SOC, Op_RegF, 31, v31->as_VMReg()->next() );
  reg_def V31_J( SOC, SOC, Op_RegF, 31, v31->as_VMReg()->next(2));
  reg_def V31_K( SOC, SOC, Op_RegF, 31, v31->as_VMReg()->next(3));

// ----------------------------
// Special Registers
// ----------------------------

// the AArch64 CSPR status flag register is not directly acessible as
// instruction operand. the FPSR status flag register is a system
// register which can be written/read using MSR/MRS but again does not
// appear as an operand (a code identifying the FSPR occurs as an
// immediate value in the instruction).

reg_def RFLAGS(SOC, SOC, 0, 32, VMRegImpl::Bad());


// Specify priority of register selection within phases of register
// allocation.  Highest priority is first.  A useful heuristic is to
// give registers a low priority when they are required by machine
// instructions, like EAX and EDX on I486, and choose no-save registers
// before save-on-call, & save-on-call before save-on-entry.  Registers
// which participate in fixed calling sequences should come last.
// Registers which are used as pairs must fall on an even boundary.

alloc_class chunk0(
    // volatiles
    R10, R10_H,
    R11, R11_H,
    R12, R12_H,
    R13, R13_H,
    R14, R14_H,
    R15, R15_H,
    R16, R16_H,
    R17, R17_H,
    R18, R18_H,

    // arg registers
    R0, R0_H,
    R1, R1_H,
    R2, R2_H,
    R3, R3_H,
    R4, R4_H,
    R5, R5_H,
    R6, R6_H,
    R7, R7_H,

    // non-volatiles
    R19, R19_H,
    R20, R20_H,
    R21, R21_H,
    R22, R22_H,
    R23, R23_H,
    R24, R24_H,
    R25, R25_H,
    R26, R26_H,

    // non-allocatable registers

    R27, R27_H, // heapbase
    R28, R28_H, // thread
    R29, R29_H, // fp
    R30, R30_H, // lr
    R31, R31_H, // sp
);

alloc_class chunk1(

    // no save
    V16, V16_H, V16_J, V16_K,
    V17, V17_H, V17_J, V17_K,
    V18, V18_H, V18_J, V18_K,
    V19, V19_H, V19_J, V19_K,
    V20, V20_H, V20_J, V20_K,
    V21, V21_H, V21_J, V21_K,
    V22, V22_H, V22_J, V22_K,
    V23, V23_H, V23_J, V23_K,
    V24, V24_H, V24_J, V24_K,
    V25, V25_H, V25_J, V25_K,
    V26, V26_H, V26_J, V26_K,
    V27, V27_H, V27_J, V27_K,
    V28, V28_H, V28_J, V28_K,
    V29, V29_H, V29_J, V29_K,
    V30, V30_H, V30_J, V30_K,
    V31, V31_H, V31_J, V31_K,

    // arg registers
    V0, V0_H, V0_J, V0_K,
    V1, V1_H, V1_J, V1_K,
    V2, V2_H, V2_J, V2_K,
    V3, V3_H, V3_J, V3_K,
    V4, V4_H, V4_J, V4_K,
    V5, V5_H, V5_J, V5_K,
    V6, V6_H, V6_J, V6_K,
    V7, V7_H, V7_J, V7_K,

    // non-volatiles
    V8, V8_H, V8_J, V8_K,
    V9, V9_H, V9_J, V9_K,
    V10, V10_H, V10_J, V10_K,
    V11, V11_H, V11_J, V11_K,
    V12, V12_H, V12_J, V12_K,
    V13, V13_H, V13_J, V13_K,
    V14, V14_H, V14_J, V14_K,
    V15, V15_H, V15_J, V15_K,
);

alloc_class chunk2(RFLAGS);

//----------Architecture Description Register Classes--------------------------
// Several register classes are automatically defined based upon information in
// this architecture description.
// 1) reg_class inline_cache_reg           ( /* as def'd in frame section */ )
// 2) reg_class compiler_method_oop_reg    ( /* as def'd in frame section */ )
// 2) reg_class interpreter_method_oop_reg ( /* as def'd in frame section */ )
// 3) reg_class stack_slots( /* one chunk of stack-based "registers" */ )
//

// Class for all 32 bit integer registers -- excludes SP which will
// never be used as an integer register
reg_class any_reg32(
    R0,
    R1,
    R2,
    R3,
    R4,
    R5,
    R6,
    R7,
    R10,
    R11,
    R12,
    R13,
    R14,
    R15,
    R16,
    R17,
    R18,
    R19,
    R20,
    R21,
    R22,
    R23,
    R24,
    R25,
    R26,
    R27,
    R28,
    R29,
    R30
);

// Singleton class for R0 int register
reg_class int_r0_reg(R0);

// Singleton class for R2 int register
reg_class int_r2_reg(R2);

// Singleton class for R3 int register
reg_class int_r3_reg(R3);

// Singleton class for R4 int register
reg_class int_r4_reg(R4);

// Class for all long integer registers (including RSP)
reg_class any_reg(
    R0, R0_H,
    R1, R1_H,
    R2, R2_H,
    R3, R3_H,
    R4, R4_H,
    R5, R5_H,
    R6, R6_H,
    R7, R7_H,
    R10, R10_H,
    R11, R11_H,
    R12, R12_H,
    R13, R13_H,
    R14, R14_H,
    R15, R15_H,
    R16, R16_H,
    R17, R17_H,
    R18, R18_H,
    R19, R19_H,
    R20, R20_H,
    R21, R21_H,
    R22, R22_H,
    R23, R23_H,
    R24, R24_H,
    R25, R25_H,
    R26, R26_H,
    R27, R27_H,
    R28, R28_H,
    R29, R29_H,
    R30, R30_H,
    R31, R31_H
);

// Class for all non-special integer registers
reg_class no_special_reg32_no_fp(
    R0,
    R1,
    R2,
    R3,
    R4,
    R5,
    R6,
    R7,
    R10,
    R11,
    R12,                        // rmethod
    R13,
    R14,
    R15,
    R16,
    R17,
    R18,
    R19,
    R20,
    R21,
    R22,
    R23,
    R24,
    R25,
    R26
 /* R27, */                     // heapbase
 /* R28, */                     // thread
 /* R29, */                     // fp
 /* R30, */                     // lr
 /* R31 */                      // sp
);

reg_class no_special_reg32_with_fp(
    R0,
    R1,
    R2,
    R3,
    R4,
    R5,
    R6,
    R7,
    R10,
    R11,
    R12,                        // rmethod
    R13,
    R14,
    R15,
    R16,
    R17,
    R18,
    R19,
    R20,
    R21,
    R22,
    R23,
    R24,
    R25,
    R26
 /* R27, */                     // heapbase
 /* R28, */                     // thread
    R29,                        // fp
 /* R30, */                     // lr
 /* R31 */                      // sp
);

reg_class_dynamic no_special_reg32(no_special_reg32_no_fp, no_special_reg32_with_fp, %{ PreserveFramePointer %});

// Class for all non-special long integer registers
reg_class no_special_reg_no_fp(
    R0, R0_H,
    R1, R1_H,
    R2, R2_H,
    R3, R3_H,
    R4, R4_H,
    R5, R5_H,
    R6, R6_H,
    R7, R7_H,
    R10, R10_H,
    R11, R11_H,
    R12, R12_H,                 // rmethod
    R13, R13_H,
    R14, R14_H,
    R15, R15_H,
    R16, R16_H,
    R17, R17_H,
    R18, R18_H,
    R19, R19_H,
    R20, R20_H,
    R21, R21_H,
    R22, R22_H,
    R23, R23_H,
    R24, R24_H,
    R25, R25_H,
    R26, R26_H,
 /* R27, R27_H, */              // heapbase
 /* R28, R28_H, */              // thread
 /* R29, R29_H, */              // fp
 /* R30, R30_H, */              // lr
 /* R31, R31_H */               // sp
);

reg_class no_special_reg_with_fp(
    R0, R0_H,
    R1, R1_H,
    R2, R2_H,
    R3, R3_H,
    R4, R4_H,
    R5, R5_H,
    R6, R6_H,
    R7, R7_H,
    R10, R10_H,
    R11, R11_H,
    R12, R12_H,                 // rmethod
    R13, R13_H,
    R14, R14_H,
    R15, R15_H,
    R16, R16_H,
    R17, R17_H,
    R18, R18_H,
    R19, R19_H,
    R20, R20_H,
    R21, R21_H,
    R22, R22_H,
    R23, R23_H,
    R24, R24_H,
    R25, R25_H,
    R26, R26_H,
 /* R27, R27_H, */              // heapbase
 /* R28, R28_H, */              // thread
    R29, R29_H,                 // fp
 /* R30, R30_H, */              // lr
 /* R31, R31_H */               // sp
);

reg_class_dynamic no_special_reg(no_special_reg_no_fp, no_special_reg_with_fp, %{ PreserveFramePointer %});

// Class for 64 bit register r0
reg_class r0_reg(
    R0, R0_H
);

// Class for 64 bit register r1
reg_class r1_reg(
    R1, R1_H
);

// Class for 64 bit register r2
reg_class r2_reg(
    R2, R2_H
);

// Class for 64 bit register r3
reg_class r3_reg(
    R3, R3_H
);

// Class for 64 bit register r4
reg_class r4_reg(
    R4, R4_H
);

// Class for 64 bit register r5
reg_class r5_reg(
    R5, R5_H
);

// Class for 64 bit register r10
reg_class r10_reg(
    R10, R10_H
);

// Class for 64 bit register r11
reg_class r11_reg(
    R11, R11_H
);

// Class for method register
reg_class method_reg(
    R12, R12_H
);

// Class for heapbase register
reg_class heapbase_reg(
    R27, R27_H
);

// Class for thread register
reg_class thread_reg(
    R28, R28_H
);

// Class for frame pointer register
reg_class fp_reg(
    R29, R29_H
);

// Class for link register
reg_class lr_reg(
    R30, R30_H
);

// Class for long sp register
reg_class sp_reg(
  R31, R31_H
);

// Class for all pointer registers
reg_class ptr_reg(
    R0, R0_H,
    R1, R1_H,
    R2, R2_H,
    R3, R3_H,
    R4, R4_H,
    R5, R5_H,
    R6, R6_H,
    R7, R7_H,
    R10, R10_H,
    R11, R11_H,
    R12, R12_H,
    R13, R13_H,
    R14, R14_H,
    R15, R15_H,
    R16, R16_H,
    R17, R17_H,
    R18, R18_H,
    R19, R19_H,
    R20, R20_H,
    R21, R21_H,
    R22, R22_H,
    R23, R23_H,
    R24, R24_H,
    R25, R25_H,
    R26, R26_H,
    R27, R27_H,
    R28, R28_H,
    R29, R29_H,
    R30, R30_H,
    R31, R31_H
);

// Class for all non_special pointer registers
reg_class no_special_ptr_reg(
    R0, R0_H,
    R1, R1_H,
    R2, R2_H,
    R3, R3_H,
    R4, R4_H,
    R5, R5_H,
    R6, R6_H,
    R7, R7_H,
    R10, R10_H,
    R11, R11_H,
    R12, R12_H,
    R13, R13_H,
    R14, R14_H,
    R15, R15_H,
    R16, R16_H,
    R17, R17_H,
    R18, R18_H,
    R19, R19_H,
    R20, R20_H,
    R21, R21_H,
    R22, R22_H,
    R23, R23_H,
    R24, R24_H,
    R25, R25_H,
    R26, R26_H,
 /* R27, R27_H, */              // heapbase
 /* R28, R28_H, */              // thread
 /* R29, R29_H, */              // fp
 /* R30, R30_H, */              // lr
 /* R31, R31_H */               // sp
);

// Class for all float registers
reg_class float_reg(
    V0,
    V1,
    V2,
    V3,
    V4,
    V5,
    V6,
    V7,
    V8,
    V9,
    V10,
    V11,
    V12,
    V13,
    V14,
    V15,
    V16,
    V17,
    V18,
    V19,
    V20,
    V21,
    V22,
    V23,
    V24,
    V25,
    V26,
    V27,
    V28,
    V29,
    V30,
    V31
);

// Double precision float registers have virtual `high halves' that
// are needed by the allocator.
// Class for all double registers
reg_class double_reg(
    V0, V0_H,
    V1, V1_H,
    V2, V2_H,
    V3, V3_H,
    V4, V4_H,
    V5, V5_H,
    V6, V6_H,
    V7, V7_H,
    V8, V8_H,
    V9, V9_H,
    V10, V10_H,
    V11, V11_H,
    V12, V12_H,
    V13, V13_H,
    V14, V14_H,
    V15, V15_H,
    V16, V16_H,
    V17, V17_H,
    V18, V18_H,
    V19, V19_H,
    V20, V20_H,
    V21, V21_H,
    V22, V22_H,
    V23, V23_H,
    V24, V24_H,
    V25, V25_H,
    V26, V26_H,
    V27, V27_H,
    V28, V28_H,
    V29, V29_H,
    V30, V30_H,
    V31, V31_H
);

// Class for all 64bit vector registers
reg_class vectord_reg(
    V0, V0_H,
    V1, V1_H,
    V2, V2_H,
    V3, V3_H,
    V4, V4_H,
    V5, V5_H,
    V6, V6_H,
    V7, V7_H,
    V8, V8_H,
    V9, V9_H,
    V10, V10_H,
    V11, V11_H,
    V12, V12_H,
    V13, V13_H,
    V14, V14_H,
    V15, V15_H,
    V16, V16_H,
    V17, V17_H,
    V18, V18_H,
    V19, V19_H,
    V20, V20_H,
    V21, V21_H,
    V22, V22_H,
    V23, V23_H,
    V24, V24_H,
    V25, V25_H,
    V26, V26_H,
    V27, V27_H,
    V28, V28_H,
    V29, V29_H,
    V30, V30_H,
    V31, V31_H
);

// Class for all 128bit vector registers
reg_class vectorx_reg(
    V0, V0_H, V0_J, V0_K,
    V1, V1_H, V1_J, V1_K,
    V2, V2_H, V2_J, V2_K,
    V3, V3_H, V3_J, V3_K,
    V4, V4_H, V4_J, V4_K,
    V5, V5_H, V5_J, V5_K,
    V6, V6_H, V6_J, V6_K,
    V7, V7_H, V7_J, V7_K,
    V8, V8_H, V8_J, V8_K,
    V9, V9_H, V9_J, V9_K,
    V10, V10_H, V10_J, V10_K,
    V11, V11_H, V11_J, V11_K,
    V12, V12_H, V12_J, V12_K,
    V13, V13_H, V13_J, V13_K,
    V14, V14_H, V14_J, V14_K,
    V15, V15_H, V15_J, V15_K,
    V16, V16_H, V16_J, V16_K,
    V17, V17_H, V17_J, V17_K,
    V18, V18_H, V18_J, V18_K,
    V19, V19_H, V19_J, V19_K,
    V20, V20_H, V20_J, V20_K,
    V21, V21_H, V21_J, V21_K,
    V22, V22_H, V22_J, V22_K,
    V23, V23_H, V23_J, V23_K,
    V24, V24_H, V24_J, V24_K,
    V25, V25_H, V25_J, V25_K,
    V26, V26_H, V26_J, V26_K,
    V27, V27_H, V27_J, V27_K,
    V28, V28_H, V28_J, V28_K,
    V29, V29_H, V29_J, V29_K,
    V30, V30_H, V30_J, V30_K,
    V31, V31_H, V31_J, V31_K
);

// Class for 128 bit register v0
reg_class v0_reg(
    V0, V0_H
);

// Class for 128 bit register v1
reg_class v1_reg(
    V1, V1_H
);

// Class for 128 bit register v2
reg_class v2_reg(
    V2, V2_H
);

// Class for 128 bit register v3
reg_class v3_reg(
    V3, V3_H
);

// Singleton class for condition codes
reg_class int_flags(RFLAGS);

%}

//----------DEFINITION BLOCK---------------------------------------------------
// Define name --> value mappings to inform the ADLC of an integer valued name
// Current support includes integer values in the range [0, 0x7FFFFFFF]
// Format:
//        int_def  <name>         ( <int_value>, <expression>);
// Generated Code in ad_<arch>.hpp
//        #define  <name>   (<expression>)
//        // value == <int_value>
// Generated code in ad_<arch>.cpp adlc_verification()
//        assert( <name> == <int_value>, "Expect (<expression>) to equal <int_value>");
//

// we follow the ppc-aix port in using a simple cost model which ranks
// register operations as cheap, memory ops as more expensive and
// branches as most expensive. the first two have a low as well as a
// normal cost. huge cost appears to be a way of saying don't do
// something

definitions %{
  // The default cost (of a register move instruction).
  int_def INSN_COST            (    100,     100);
  int_def BRANCH_COST          (    200,     2 * INSN_COST);
  int_def CALL_COST            (    200,     2 * INSN_COST);
  int_def VOLATILE_REF_COST    (   1000,     10 * INSN_COST);
%}


//----------SOURCE BLOCK-------------------------------------------------------
// This is a block of C++ code which provides values, functions, and
// definitions necessary in the rest of the architecture description

source_hpp %{

#include "gc/shared/cardTableModRefBS.hpp"
#include "opto/addnode.hpp"

class CallStubImpl {

  //--------------------------------------------------------------
  //---<  Used for optimization in Compile::shorten_branches  >---
  //--------------------------------------------------------------

 public:
  // Size of call trampoline stub.
  static uint size_call_trampoline() {
    return 0; // no call trampolines on this platform
  }

  // number of relocations needed by a call trampoline stub
  static uint reloc_call_trampoline() {
    return 0; // no call trampolines on this platform
  }
};

class HandlerImpl {

 public:

  static int emit_exception_handler(CodeBuffer &cbuf);
  static int emit_deopt_handler(CodeBuffer& cbuf);

  static uint size_exception_handler() {
    return MacroAssembler::far_branch_size();
  }

  static uint size_deopt_handler() {
    // count one adr and one far branch instruction
    return 4 * NativeInstruction::instruction_size;
  }
};

  // graph traversal helpers

  MemBarNode *parent_membar(const Node *n);
  MemBarNode *child_membar(const MemBarNode *n);
  bool leading_membar(const MemBarNode *barrier);

  bool is_card_mark_membar(const MemBarNode *barrier);
  bool is_CAS(int opcode);

  MemBarNode *leading_to_normal(MemBarNode *leading);
  MemBarNode *normal_to_leading(const MemBarNode *barrier);
  MemBarNode *card_mark_to_trailing(const MemBarNode *barrier);
  MemBarNode *trailing_to_card_mark(const MemBarNode *trailing);
  MemBarNode *trailing_to_leading(const MemBarNode *trailing);

  // predicates controlling emit of ldr<x>/ldar<x> and associated dmb

  bool unnecessary_acquire(const Node *barrier);
  bool needs_acquiring_load(const Node *load);

  // predicates controlling emit of str<x>/stlr<x> and associated dmbs

  bool unnecessary_release(const Node *barrier);
  bool unnecessary_volatile(const Node *barrier);
  bool needs_releasing_store(const Node *store);

  // predicate controlling translation of CompareAndSwapX
  bool needs_acquiring_load_exclusive(const Node *load);

  // predicate controlling translation of StoreCM
  bool unnecessary_storestore(const Node *storecm);

  // predicate controlling addressing modes
  bool size_fits_all_mem_uses(AddPNode* addp, int shift);
%}

source %{

  // Optimizaton of volatile gets and puts
  // -------------------------------------
  //
  // AArch64 has ldar<x> and stlr<x> instructions which we can safely
  // use to implement volatile reads and writes. For a volatile read
  // we simply need
  //
  //   ldar<x>
  //
  // and for a volatile write we need
  //
  //   stlr<x>
  //
  // Alternatively, we can implement them by pairing a normal
  // load/store with a memory barrier. For a volatile read we need
  //
  //   ldr<x>
  //   dmb ishld
  //
  // for a volatile write
  //
  //   dmb ish
  //   str<x>
  //   dmb ish
  //
  // We can also use ldaxr and stlxr to implement compare and swap CAS
  // sequences. These are normally translated to an instruction
  // sequence like the following
  //
  //   dmb      ish
  // retry:
  //   ldxr<x>   rval raddr
  //   cmp       rval rold
  //   b.ne done
  //   stlxr<x>  rval, rnew, rold
  //   cbnz      rval retry
  // done:
  //   cset      r0, eq
  //   dmb ishld
  //
  // Note that the exclusive store is already using an stlxr
  // instruction. That is required to ensure visibility to other
  // threads of the exclusive write (assuming it succeeds) before that
  // of any subsequent writes.
  //
  // The following instruction sequence is an improvement on the above
  //
  // retry:
  //   ldaxr<x>  rval raddr
  //   cmp       rval rold
  //   b.ne done
  //   stlxr<x>  rval, rnew, rold
  //   cbnz      rval retry
  // done:
  //   cset      r0, eq
  //
  // We don't need the leading dmb ish since the stlxr guarantees
  // visibility of prior writes in the case that the swap is
  // successful. Crucially we don't have to worry about the case where
  // the swap is not successful since no valid program should be
  // relying on visibility of prior changes by the attempting thread
  // in the case where the CAS fails.
  //
  // Similarly, we don't need the trailing dmb ishld if we substitute
  // an ldaxr instruction since that will provide all the guarantees we
  // require regarding observation of changes made by other threads
  // before any change to the CAS address observed by the load.
  //
  // In order to generate the desired instruction sequence we need to
  // be able to identify specific 'signature' ideal graph node
  // sequences which i) occur as a translation of a volatile reads or
  // writes or CAS operations and ii) do not occur through any other
  // translation or graph transformation. We can then provide
  // alternative aldc matching rules which translate these node
  // sequences to the desired machine code sequences. Selection of the
  // alternative rules can be implemented by predicates which identify
  // the relevant node sequences.
  //
  // The ideal graph generator translates a volatile read to the node
  // sequence
  //
  //   LoadX[mo_acquire]
  //   MemBarAcquire
  //
  // As a special case when using the compressed oops optimization we
  // may also see this variant
  //
  //   LoadN[mo_acquire]
  //   DecodeN
  //   MemBarAcquire
  //
  // A volatile write is translated to the node sequence
  //
  //   MemBarRelease
  //   StoreX[mo_release] {CardMark}-optional
  //   MemBarVolatile
  //
  // n.b. the above node patterns are generated with a strict
  // 'signature' configuration of input and output dependencies (see
  // the predicates below for exact details). The card mark may be as
  // simple as a few extra nodes or, in a few GC configurations, may
  // include more complex control flow between the leading and
  // trailing memory barriers. However, whatever the card mark
  // configuration these signatures are unique to translated volatile
  // reads/stores -- they will not appear as a result of any other
  // bytecode translation or inlining nor as a consequence of
  // optimizing transforms.
  //
  // We also want to catch inlined unsafe volatile gets and puts and
  // be able to implement them using either ldar<x>/stlr<x> or some
  // combination of ldr<x>/stlr<x> and dmb instructions.
  //
  // Inlined unsafe volatiles puts manifest as a minor variant of the
  // normal volatile put node sequence containing an extra cpuorder
  // membar
  //
  //   MemBarRelease
  //   MemBarCPUOrder
  //   StoreX[mo_release] {CardMark}-optional
  //   MemBarVolatile
  //
  // n.b. as an aside, the cpuorder membar is not itself subject to
  // matching and translation by adlc rules.  However, the rule
  // predicates need to detect its presence in order to correctly
  // select the desired adlc rules.
  //
  // Inlined unsafe volatile gets manifest as a somewhat different
  // node sequence to a normal volatile get
  //
  //   MemBarCPUOrder
  //        ||       \\
  //   MemBarAcquire LoadX[mo_acquire]
  //        ||
  //   MemBarCPUOrder
  //
  // In this case the acquire membar does not directly depend on the
  // load. However, we can be sure that the load is generated from an
  // inlined unsafe volatile get if we see it dependent on this unique
  // sequence of membar nodes. Similarly, given an acquire membar we
  // can know that it was added because of an inlined unsafe volatile
  // get if it is fed and feeds a cpuorder membar and if its feed
  // membar also feeds an acquiring load.
  //
  // Finally an inlined (Unsafe) CAS operation is translated to the
  // following ideal graph
  //
  //   MemBarRelease
  //   MemBarCPUOrder
  //   CompareAndSwapX {CardMark}-optional
  //   MemBarCPUOrder
  //   MemBarAcquire
  //
  // So, where we can identify these volatile read and write
  // signatures we can choose to plant either of the above two code
  // sequences. For a volatile read we can simply plant a normal
  // ldr<x> and translate the MemBarAcquire to a dmb. However, we can
  // also choose to inhibit translation of the MemBarAcquire and
  // inhibit planting of the ldr<x>, instead planting an ldar<x>.
  //
  // When we recognise a volatile store signature we can choose to
  // plant at a dmb ish as a translation for the MemBarRelease, a
  // normal str<x> and then a dmb ish for the MemBarVolatile.
  // Alternatively, we can inhibit translation of the MemBarRelease
  // and MemBarVolatile and instead plant a simple stlr<x>
  // instruction.
  //
  // when we recognise a CAS signature we can choose to plant a dmb
  // ish as a translation for the MemBarRelease, the conventional
  // macro-instruction sequence for the CompareAndSwap node (which
  // uses ldxr<x>) and then a dmb ishld for the MemBarAcquire.
  // Alternatively, we can elide generation of the dmb instructions
  // and plant the alternative CompareAndSwap macro-instruction
  // sequence (which uses ldaxr<x>).
  //
  // Of course, the above only applies when we see these signature
  // configurations. We still want to plant dmb instructions in any
  // other cases where we may see a MemBarAcquire, MemBarRelease or
  // MemBarVolatile. For example, at the end of a constructor which
  // writes final/volatile fields we will see a MemBarRelease
  // instruction and this needs a 'dmb ish' lest we risk the
  // constructed object being visible without making the
  // final/volatile field writes visible.
  //
  // n.b. the translation rules below which rely on detection of the
  // volatile signatures and insert ldar<x> or stlr<x> are failsafe.
  // If we see anything other than the signature configurations we
  // always just translate the loads and stores to ldr<x> and str<x>
  // and translate acquire, release and volatile membars to the
  // relevant dmb instructions.
  //

  // graph traversal helpers used for volatile put/get and CAS
  // optimization

  // 1) general purpose helpers

  // if node n is linked to a parent MemBarNode by an intervening
  // Control and Memory ProjNode return the MemBarNode otherwise return
  // NULL.
  //
  // n may only be a Load or a MemBar.

  MemBarNode *parent_membar(const Node *n)
  {
    Node *ctl = NULL;
    Node *mem = NULL;
    Node *membar = NULL;

    if (n->is_Load()) {
      ctl = n->lookup(LoadNode::Control);
      mem = n->lookup(LoadNode::Memory);
    } else if (n->is_MemBar()) {
      ctl = n->lookup(TypeFunc::Control);
      mem = n->lookup(TypeFunc::Memory);
    } else {
	return NULL;
    }

    if (!ctl || !mem || !ctl->is_Proj() || !mem->is_Proj()) {
      return NULL;
    }

    membar = ctl->lookup(0);

    if (!membar || !membar->is_MemBar()) {
      return NULL;
    }

    if (mem->lookup(0) != membar) {
      return NULL;
    }

    return membar->as_MemBar();
  }

  // if n is linked to a child MemBarNode by intervening Control and
  // Memory ProjNodes return the MemBarNode otherwise return NULL.

  MemBarNode *child_membar(const MemBarNode *n)
  {
    ProjNode *ctl = n->proj_out(TypeFunc::Control);
    ProjNode *mem = n->proj_out(TypeFunc::Memory);

    // MemBar needs to have both a Ctl and Mem projection
    if (! ctl || ! mem)
      return NULL;

    MemBarNode *child = NULL;
    Node *x;

    for (DUIterator_Fast imax, i = ctl->fast_outs(imax); i < imax; i++) {
      x = ctl->fast_out(i);
      // if we see a membar we keep hold of it. we may also see a new
      // arena copy of the original but it will appear later
      if (x->is_MemBar()) {
	  child = x->as_MemBar();
	  break;
      }
    }

    if (child == NULL) {
      return NULL;
    }

    for (DUIterator_Fast imax, i = mem->fast_outs(imax); i < imax; i++) {
      x = mem->fast_out(i);
      // if we see a membar we keep hold of it. we may also see a new
      // arena copy of the original but it will appear later
      if (x == child) {
	return child;
      }
    }
    return NULL;
  }

  // helper predicate use to filter candidates for a leading memory
  // barrier
  //
  // returns true if barrier is a MemBarRelease or a MemBarCPUOrder
  // whose Ctl and Mem feeds come from a MemBarRelease otherwise false

  bool leading_membar(const MemBarNode *barrier)
  {
    int opcode = barrier->Opcode();
    // if this is a release membar we are ok
    if (opcode == Op_MemBarRelease) {
      return true;
    }
    // if its a cpuorder membar . . .
    if (opcode != Op_MemBarCPUOrder) {
      return false;
    }
    // then the parent has to be a release membar
    MemBarNode *parent = parent_membar(barrier);
    if (!parent) {
      return false;
    }
    opcode = parent->Opcode();
    return opcode == Op_MemBarRelease;
  }

  // 2) card mark detection helper

  // helper predicate which can be used to detect a volatile membar
  // introduced as part of a conditional card mark sequence either by
  // G1 or by CMS when UseCondCardMark is true.
  //
  // membar can be definitively determined to be part of a card mark
  // sequence if and only if all the following hold
  //
  // i) it is a MemBarVolatile
  //
  // ii) either UseG1GC or (UseConcMarkSweepGC && UseCondCardMark) is
  // true
  //
  // iii) the node's Mem projection feeds a StoreCM node.

  bool is_card_mark_membar(const MemBarNode *barrier)
  {
    if (!UseG1GC && !(UseConcMarkSweepGC && UseCondCardMark)) {
      return false;
    }

    if (barrier->Opcode() != Op_MemBarVolatile) {
      return false;
    }

    ProjNode *mem = barrier->proj_out(TypeFunc::Memory);

    for (DUIterator_Fast imax, i = mem->fast_outs(imax); i < imax ; i++) {
      Node *y = mem->fast_out(i);
      if (y->Opcode() == Op_StoreCM) {
	return true;
      }
    }

    return false;
  }


  // 3) helper predicates to traverse volatile put or CAS graphs which
  // may contain GC barrier subgraphs

  // Preamble
  // --------
  //
  // for volatile writes we can omit generating barriers and employ a
  // releasing store when we see a node sequence sequence with a
  // leading MemBarRelease and a trailing MemBarVolatile as follows
  //
  //   MemBarRelease
  //  {      ||      } -- optional
  //  {MemBarCPUOrder}
  //         ||     \\
  //         ||     StoreX[mo_release]
  //         | \     /
  //         | MergeMem
  //         | /
  //   MemBarVolatile
  //
  // where
  //  || and \\ represent Ctl and Mem feeds via Proj nodes
  //  | \ and / indicate further routing of the Ctl and Mem feeds
  //
  // this is the graph we see for non-object stores. however, for a
  // volatile Object store (StoreN/P) we may see other nodes below the
  // leading membar because of the need for a GC pre- or post-write
  // barrier.
  //
  // with most GC configurations we with see this simple variant which
  // includes a post-write barrier card mark.
  //
  //   MemBarRelease______________________________
  //         ||    \\               Ctl \        \\
  //         ||    StoreN/P[mo_release] CastP2X  StoreB/CM
  //         | \     /                       . . .  /
  //         | MergeMem
  //         | /
  //         ||      /
  //   MemBarVolatile
  //
  // i.e. the leading membar feeds Ctl to a CastP2X (which converts
  // the object address to an int used to compute the card offset) and
  // Ctl+Mem to a StoreB node (which does the actual card mark).
  //
  // n.b. a StoreCM node will only appear in this configuration when
  // using CMS. StoreCM differs from a normal card mark write (StoreB)
  // because it implies a requirement to order visibility of the card
  // mark (StoreCM) relative to the object put (StoreP/N) using a
  // StoreStore memory barrier (arguably this ought to be represented
  // explicitly in the ideal graph but that is not how it works). This
  // ordering is required for both non-volatile and volatile
  // puts. Normally that means we need to translate a StoreCM using
  // the sequence
  //
  //   dmb ishst
  //   stlrb
  //
  // However, in the case of a volatile put if we can recognise this
  // configuration and plant an stlr for the object write then we can
  // omit the dmb and just plant an strb since visibility of the stlr
  // is ordered before visibility of subsequent stores. StoreCM nodes
  // also arise when using G1 or using CMS with conditional card
  // marking. In these cases (as we shall see) we don't need to insert
  // the dmb when translating StoreCM because there is already an
  // intervening StoreLoad barrier between it and the StoreP/N.
  //
  // It is also possible to perform the card mark conditionally on it
  // currently being unmarked in which case the volatile put graph
  // will look slightly different
  //
  //   MemBarRelease____________________________________________
  //         ||    \\               Ctl \     Ctl \     \\  Mem \
  //         ||    StoreN/P[mo_release] CastP2X   If   LoadB     |
  //         | \     /                              \            |
  //         | MergeMem                            . . .      StoreB
  //         | /                                                /
  //         ||     /
  //   MemBarVolatile
  //
  // It is worth noting at this stage that both the above
  // configurations can be uniquely identified by checking that the
  // memory flow includes the following subgraph:
  //
  //   MemBarRelease
  //  {MemBarCPUOrder}
  //          |  \      . . .
  //          |  StoreX[mo_release]  . . .
  //          |   /
  //         MergeMem
  //          |
  //   MemBarVolatile
  //
  // This is referred to as a *normal* subgraph. It can easily be
  // detected starting from any candidate MemBarRelease,
  // StoreX[mo_release] or MemBarVolatile.
  //
  // A simple variation on this normal case occurs for an unsafe CAS
  // operation. The basic graph for a non-object CAS is
  //
  //   MemBarRelease
  //         ||
  //   MemBarCPUOrder
  //         ||     \\   . . .
  //         ||     CompareAndSwapX
  //         ||       |
  //         ||     SCMemProj
  //         | \     /
  //         | MergeMem
  //         | /
  //   MemBarCPUOrder
  //         ||
  //   MemBarAcquire
  //
  // The same basic variations on this arrangement (mutatis mutandis)
  // occur when a card mark is introduced. i.e. we se the same basic
  // shape but the StoreP/N is replaced with CompareAndSawpP/N and the
  // tail of the graph is a pair comprising a MemBarCPUOrder +
  // MemBarAcquire.
  //
  // So, in the case of a CAS the normal graph has the variant form
  //
  //   MemBarRelease
  //   MemBarCPUOrder
  //          |   \      . . .
  //          |  CompareAndSwapX  . . .
  //          |    |
  //          |   SCMemProj
  //          |   /  . . .
  //         MergeMem
  //          |
  //   MemBarCPUOrder
  //   MemBarAcquire
  //
  // This graph can also easily be detected starting from any
  // candidate MemBarRelease, CompareAndSwapX or MemBarAcquire.
  //
  // the code below uses two helper predicates, leading_to_normal and
  // normal_to_leading to identify these normal graphs, one validating
  // the layout starting from the top membar and searching down and
  // the other validating the layout starting from the lower membar
  // and searching up.
  //
  // There are two special case GC configurations when a normal graph
  // may not be generated: when using G1 (which always employs a
  // conditional card mark); and when using CMS with conditional card
  // marking configured. These GCs are both concurrent rather than
  // stop-the world GCs. So they introduce extra Ctl+Mem flow into the
  // graph between the leading and trailing membar nodes, in
  // particular enforcing stronger memory serialisation beween the
  // object put and the corresponding conditional card mark. CMS
  // employs a post-write GC barrier while G1 employs both a pre- and
  // post-write GC barrier. Of course the extra nodes may be absent --
  // they are only inserted for object puts. This significantly
  // complicates the task of identifying whether a MemBarRelease,
  // StoreX[mo_release] or MemBarVolatile forms part of a volatile put
  // when using these GC configurations (see below). It adds similar
  // complexity to the task of identifying whether a MemBarRelease,
  // CompareAndSwapX or MemBarAcquire forms part of a CAS.
  //
  // In both cases the post-write subtree includes an auxiliary
  // MemBarVolatile (StoreLoad barrier) separating the object put and
  // the read of the corresponding card. This poses two additional
  // problems.
  //
  // Firstly, a card mark MemBarVolatile needs to be distinguished
  // from a normal trailing MemBarVolatile. Resolving this first
  // problem is straightforward: a card mark MemBarVolatile always
  // projects a Mem feed to a StoreCM node and that is a unique marker
  //
  //      MemBarVolatile (card mark)
  //       C |    \     . . .
  //         |   StoreCM   . . .
  //       . . .
  //
  // The second problem is how the code generator is to translate the
  // card mark barrier? It always needs to be translated to a "dmb
  // ish" instruction whether or not it occurs as part of a volatile
  // put. A StoreLoad barrier is needed after the object put to ensure
  // i) visibility to GC threads of the object put and ii) visibility
  // to the mutator thread of any card clearing write by a GC
  // thread. Clearly a normal store (str) will not guarantee this
  // ordering but neither will a releasing store (stlr). The latter
  // guarantees that the object put is visible but does not guarantee
  // that writes by other threads have also been observed.
  //
  // So, returning to the task of translating the object put and the
  // leading/trailing membar nodes: what do the non-normal node graph
  // look like for these 2 special cases? and how can we determine the
  // status of a MemBarRelease, StoreX[mo_release] or MemBarVolatile
  // in both normal and non-normal cases?
  //
  // A CMS GC post-barrier wraps its card write (StoreCM) inside an If
  // which selects conditonal execution based on the value loaded
  // (LoadB) from the card. Ctl and Mem are fed to the If via an
  // intervening StoreLoad barrier (MemBarVolatile).
  //
  // So, with CMS we may see a node graph for a volatile object store
  // which looks like this
  //
  //   MemBarRelease
  //   MemBarCPUOrder_(leading)__________________
  //     C |    M \       \\                   C \
  //       |       \    StoreN/P[mo_release]  CastP2X
  //       |    Bot \    /
  //       |       MergeMem
  //       |         /
  //      MemBarVolatile (card mark)
  //     C |  ||    M |
  //       | LoadB    |
  //       |   |      |
  //       | Cmp      |\
  //       | /        | \
  //       If         |  \
  //       | \        |   \
  // IfFalse  IfTrue  |    \
  //       \     / \  |     \
  //        \   / StoreCM    |
  //         \ /      |      |
  //        Region   . . .   |
  //          | \           /
  //          |  . . .  \  / Bot
  //          |       MergeMem
  //          |          |
  //        MemBarVolatile (trailing)
  //
  // The first MergeMem merges the AliasIdxBot Mem slice from the
  // leading membar and the oopptr Mem slice from the Store into the
  // card mark membar. The trailing MergeMem merges the AliasIdxBot
  // Mem slice from the card mark membar and the AliasIdxRaw slice
  // from the StoreCM into the trailing membar (n.b. the latter
  // proceeds via a Phi associated with the If region).
  //
  // The graph for a CAS varies slightly, the obvious difference being
  // that the StoreN/P node is replaced by a CompareAndSwapP/N node
  // and the trailing MemBarVolatile by a MemBarCPUOrder +
  // MemBarAcquire pair. The other important difference is that the
  // CompareAndSwap node's SCMemProj is not merged into the card mark
  // membar - it still feeds the trailing MergeMem. This also means
  // that the card mark membar receives its Mem feed directly from the
  // leading membar rather than via a MergeMem.
  //
  //   MemBarRelease
  //   MemBarCPUOrder__(leading)_________________________
  //       ||                       \\                 C \
  //   MemBarVolatile (card mark)  CompareAndSwapN/P  CastP2X
  //     C |  ||    M |              |
  //       | LoadB    |       ______/|
  //       |   |      |      /       |
  //       | Cmp      |     /      SCMemProj
  //       | /        |    /         |
  //       If         |   /         /
  //       | \        |  /         /
  // IfFalse  IfTrue  | /         /
  //       \     / \  |/ prec    /
  //        \   / StoreCM       /
  //         \ /      |        /
  //        Region   . . .    /
  //          | \            /
  //          |  . . .  \   / Bot
  //          |       MergeMem
  //          |          |
  //        MemBarCPUOrder
  //        MemBarAcquire (trailing)
  //
  // This has a slightly different memory subgraph to the one seen
  // previously but the core of it is the same as for the CAS normal
  // sungraph
  //
  //   MemBarRelease
  //   MemBarCPUOrder____
  //      ||             \      . . .
  //   MemBarVolatile  CompareAndSwapX  . . .
  //      |  \            |
  //        . . .   SCMemProj
  //          |     /  . . .
  //         MergeMem
  //          |
  //   MemBarCPUOrder
  //   MemBarAcquire
  //
  //
  // G1 is quite a lot more complicated. The nodes inserted on behalf
  // of G1 may comprise: a pre-write graph which adds the old value to
  // the SATB queue; the releasing store itself; and, finally, a
  // post-write graph which performs a card mark.
  //
  // The pre-write graph may be omitted, but only when the put is
  // writing to a newly allocated (young gen) object and then only if
  // there is a direct memory chain to the Initialize node for the
  // object allocation. This will not happen for a volatile put since
  // any memory chain passes through the leading membar.
  //
  // The pre-write graph includes a series of 3 If tests. The outermost
  // If tests whether SATB is enabled (no else case). The next If tests
  // whether the old value is non-NULL (no else case). The third tests
  // whether the SATB queue index is > 0, if so updating the queue. The
  // else case for this third If calls out to the runtime to allocate a
  // new queue buffer.
  //
  // So with G1 the pre-write and releasing store subgraph looks like
  // this (the nested Ifs are omitted).
  //
  //  MemBarRelease (leading)____________
  //     C |  ||  M \   M \    M \  M \ . . .
  //       | LoadB   \  LoadL  LoadN   \
  //       | /        \                 \
  //       If         |\                 \
  //       | \        | \                 \
  //  IfFalse  IfTrue |  \                 \
  //       |     |    |   \                 |
  //       |     If   |   /\                |
  //       |     |          \               |
  //       |                 \              |
  //       |    . . .         \             |
  //       | /       | /       |            |
  //      Region  Phi[M]       |            |
  //       | \       |         |            |
  //       |  \_____ | ___     |            |
  //     C | C \     |   C \ M |            |
  //       | CastP2X | StoreN/P[mo_release] |
  //       |         |         |            |
  //     C |       M |       M |          M |
  //        \        |         |           /
  //                  . . .
  //          (post write subtree elided)
  //                    . . .
  //             C \         M /
  //         MemBarVolatile (trailing)
  //
  // n.b. the LoadB in this subgraph is not the card read -- it's a
  // read of the SATB queue active flag.
  //
  // Once again the CAS graph is a minor variant on the above with the
  // expected substitutions of CompareAndSawpX for StoreN/P and
  // MemBarCPUOrder + MemBarAcquire for trailing MemBarVolatile.
  //
  // The G1 post-write subtree is also optional, this time when the
  // new value being written is either null or can be identified as a
  // newly allocated (young gen) object with no intervening control
  // flow. The latter cannot happen but the former may, in which case
  // the card mark membar is omitted and the memory feeds form the
  // leading membar and the SToreN/P are merged direct into the
  // trailing membar as per the normal subgraph. So, the only special
  // case which arises is when the post-write subgraph is generated.
  //
  // The kernel of the post-write G1 subgraph is the card mark itself
  // which includes a card mark memory barrier (MemBarVolatile), a
  // card test (LoadB), and a conditional update (If feeding a
  // StoreCM). These nodes are surrounded by a series of nested Ifs
  // which try to avoid doing the card mark. The top level If skips if
  // the object reference does not cross regions (i.e. it tests if
  // (adr ^ val) >> log2(regsize) != 0) -- intra-region references
  // need not be recorded. The next If, which skips on a NULL value,
  // may be absent (it is not generated if the type of value is >=
  // OopPtr::NotNull). The 3rd If skips writes to young regions (by
  // checking if card_val != young).  n.b. although this test requires
  // a pre-read of the card it can safely be done before the StoreLoad
  // barrier. However that does not bypass the need to reread the card
  // after the barrier.
  //
  //                (pre-write subtree elided)
  //        . . .                  . . .    . . .  . . .
  //        C |                    M |     M |    M |
  //       Region                  Phi[M] StoreN    |
  //          |                     / \      |      |
  //         / \_______            /   \     |      |
  //      C / C \      . . .            \    |      |
  //       If   CastP2X . . .            |   |      |
  //       / \                           |   |      |
  //      /   \                          |   |      |
  // IfFalse IfTrue                      |   |      |
  //   |       |                         |   |     /|
  //   |       If                        |   |    / |
  //   |      / \                        |   |   /  |
  //   |     /   \                        \  |  /   |
  //   | IfFalse IfTrue                   MergeMem  |
  //   |  . . .    / \                       /      |
  //   |          /   \                     /       |
  //   |     IfFalse IfTrue                /        |
  //   |      . . .    |                  /         |
  //   |               If                /          |
  //   |               / \              /           |
  //   |              /   \            /            |
  //   |         IfFalse IfTrue       /             |
  //   |           . . .   |         /              |
  //   |                    \       /               |
  //   |                     \     /                |
  //   |             MemBarVolatile__(card mark)    |
  //   |                ||   C |  M \  M \          |
  //   |               LoadB   If    |    |         |
  //   |                      / \    |    |         |
  //   |                     . . .   |    |         |
  //   |                          \  |    |        /
  //   |                        StoreCM   |       /
  //   |                          . . .   |      /
  //   |                        _________/      /
  //   |                       /  _____________/
  //   |   . . .       . . .  |  /            /
  //   |    |                 | /   _________/
  //   |    |               Phi[M] /        /
  //   |    |                 |   /        /
  //   |    |                 |  /        /
  //   |  Region  . . .     Phi[M]  _____/
  //   |    /                 |    /
  //   |                      |   /
  //   | . . .   . . .        |  /
  //   | /                    | /
  // Region           |  |  Phi[M]
  //   |              |  |  / Bot
  //    \            MergeMem
  //     \            /
  //     MemBarVolatile
  //
  // As with CMS the initial MergeMem merges the AliasIdxBot Mem slice
  // from the leading membar and the oopptr Mem slice from the Store
  // into the card mark membar i.e. the memory flow to the card mark
  // membar still looks like a normal graph.
  //
  // The trailing MergeMem merges an AliasIdxBot Mem slice with other
  // Mem slices (from the StoreCM and other card mark queue stores).
  // However in this case the AliasIdxBot Mem slice does not come
  // direct from the card mark membar. It is merged through a series
  // of Phi nodes. These are needed to merge the AliasIdxBot Mem flow
  // from the leading membar with the Mem feed from the card mark
  // membar. Each Phi corresponds to one of the Ifs which may skip
  // around the card mark membar. So when the If implementing the NULL
  // value check has been elided the total number of Phis is 2
  // otherwise it is 3.
  //
  // The CAS graph when using G1GC also includes a pre-write subgraph
  // and an optional post-write subgraph. Teh sam evarioations are
  // introduced as for CMS with conditional card marking i.e. the
  // StoreP/N is swapped for a CompareAndSwapP/N, the tariling
  // MemBarVolatile for a MemBarCPUOrder + MemBarAcquire pair and the
  // Mem feed from the CompareAndSwapP/N includes a precedence
  // dependency feed to the StoreCM and a feed via an SCMemProj to the
  // trailing membar. So, as before the configuration includes the
  // normal CAS graph as a subgraph of the memory flow.
  //
  // So, the upshot is that in all cases the volatile put graph will
  // include a *normal* memory subgraph betwen the leading membar and
  // its child membar, either a volatile put graph (including a
  // releasing StoreX) or a CAS graph (including a CompareAndSwapX).
  // When that child is not a card mark membar then it marks the end
  // of the volatile put or CAS subgraph. If the child is a card mark
  // membar then the normal subgraph will form part of a volatile put
  // subgraph if and only if the child feeds an AliasIdxBot Mem feed
  // to a trailing barrier via a MergeMem. That feed is either direct
  // (for CMS) or via 2 or 3 Phi nodes merging the leading barrier
  // memory flow (for G1).
  //
  // The predicates controlling generation of instructions for store
  // and barrier nodes employ a few simple helper functions (described
  // below) which identify the presence or absence of all these
  // subgraph configurations and provide a means of traversing from
  // one node in the subgraph to another.

  // is_CAS(int opcode)
  //
  // return true if opcode is one of the possible CompareAndSwapX
  // values otherwise false.

  bool is_CAS(int opcode)
  {
    switch(opcode) {
      // We handle these
    case Op_CompareAndSwapI:
    case Op_CompareAndSwapL:
    case Op_CompareAndSwapP:
    case Op_CompareAndSwapN:
 // case Op_CompareAndSwapB:
 // case Op_CompareAndSwapS:
      return true;
      // These are TBD
    case Op_WeakCompareAndSwapB:
    case Op_WeakCompareAndSwapS:
    case Op_WeakCompareAndSwapI:
    case Op_WeakCompareAndSwapL:
    case Op_WeakCompareAndSwapP:
    case Op_WeakCompareAndSwapN:
    case Op_CompareAndExchangeB:
    case Op_CompareAndExchangeS:
    case Op_CompareAndExchangeI:
    case Op_CompareAndExchangeL:
    case Op_CompareAndExchangeP:
    case Op_CompareAndExchangeN:
      return false;
    default:
      return false;
    }
  }


  // leading_to_normal
  //
  //graph traversal helper which detects the normal case Mem feed from
  // a release membar (or, optionally, its cpuorder child) to a
  // dependent volatile membar i.e. it ensures that one or other of
  // the following Mem flow subgraph is present.
  //
  //   MemBarRelease
  //   MemBarCPUOrder {leading}
  //          |  \      . . .
  //          |  StoreN/P[mo_release]  . . .
  //          |   /
  //         MergeMem
  //          |
  //   MemBarVolatile {trailing or card mark}
  //
  //   MemBarRelease
  //   MemBarCPUOrder {leading}
  //      |       \      . . .
  //      |     CompareAndSwapX  . . .
  //               |
  //     . . .    SCMemProj
  //           \   |
  //      |    MergeMem
  //      |       /
  //    MemBarCPUOrder
  //    MemBarAcquire {trailing}
  //
  // if the correct configuration is present returns the trailing
  // membar otherwise NULL.
  //
  // the input membar is expected to be either a cpuorder membar or a
  // release membar. in the latter case it should not have a cpu membar
  // child.
  //
  // the returned value may be a card mark or trailing membar
  //

  MemBarNode *leading_to_normal(MemBarNode *leading)
  {
    assert((leading->Opcode() == Op_MemBarRelease ||
	    leading->Opcode() == Op_MemBarCPUOrder),
	   "expecting a volatile or cpuroder membar!");

    // check the mem flow
    ProjNode *mem = leading->proj_out(TypeFunc::Memory);

    if (!mem) {
      return NULL;
    }

    Node *x = NULL;
    StoreNode * st = NULL;
    LoadStoreNode *cas = NULL;
    MergeMemNode *mm = NULL;

    for (DUIterator_Fast imax, i = mem->fast_outs(imax); i < imax; i++) {
      x = mem->fast_out(i);
      if (x->is_MergeMem()) {
	if (mm != NULL) {
	  return NULL;
	}
	// two merge mems is one too many
	mm = x->as_MergeMem();
      } else if (x->is_Store() && x->as_Store()->is_release() && x->Opcode() != Op_StoreCM) {
	// two releasing stores/CAS nodes is one too many
	if (st != NULL || cas != NULL) {
	  return NULL;
	}
	st = x->as_Store();
      } else if (is_CAS(x->Opcode())) {
	if (st != NULL || cas != NULL) {
	  return NULL;
	}
	cas = x->as_LoadStore();
      }
    }

    // must have a store or a cas
    if (!st && !cas) {
      return NULL;
    }

    // must have a merge if we also have st
    if (st && !mm) {
      return NULL;
    }

    Node *y = NULL;
    if (cas) {
      // look for an SCMemProj
      for (DUIterator_Fast imax, i = cas->fast_outs(imax); i < imax; i++) {
	x = cas->fast_out(i);
	if (x->is_Proj()) {
	  y = x;
	  break;
	}
      }
      if (y == NULL) {
	return NULL;
      }
      // the proj must feed a MergeMem
      for (DUIterator_Fast imax, i = y->fast_outs(imax); i < imax; i++) {
	x = y->fast_out(i);
	if (x->is_MergeMem()) {
	  mm = x->as_MergeMem();
	  break;
	}
      }
      if (mm == NULL)
	return NULL;
    } else {
      // ensure the store feeds the existing mergemem;
      for (DUIterator_Fast imax, i = st->fast_outs(imax); i < imax; i++) {
	if (st->fast_out(i) == mm) {
	  y = st;
	  break;
	}
      }
      if (y == NULL) {
	return NULL;
      }
    }

    MemBarNode *mbar = NULL;
    // ensure the merge feeds to the expected type of membar
    for (DUIterator_Fast imax, i = mm->fast_outs(imax); i < imax; i++) {
      x = mm->fast_out(i);
      if (x->is_MemBar()) {
	int opcode = x->Opcode();
	if (opcode == Op_MemBarVolatile && st) {
	  mbar = x->as_MemBar();
	} else if (cas && opcode == Op_MemBarCPUOrder) {
	  MemBarNode *y =  x->as_MemBar();
	  y = child_membar(y);
	  if (y != NULL && y->Opcode() == Op_MemBarAcquire) {
	    mbar = y;
	  }
	}
	break;
      }
    }

    return mbar;
  }

  // normal_to_leading
  //
  // graph traversal helper which detects the normal case Mem feed
  // from either a card mark or a trailing membar to a preceding
  // release membar (optionally its cpuorder child) i.e. it ensures
  // that one or other of the following Mem flow subgraphs is present.
  //
  //   MemBarRelease
  //   MemBarCPUOrder {leading}
  //          |  \      . . .
  //          |  StoreN/P[mo_release]  . . .
  //          |   /
  //         MergeMem
  //          |
  //   MemBarVolatile {card mark or trailing}
  //
  //   MemBarRelease
  //   MemBarCPUOrder {leading}
  //      |       \      . . .
  //      |     CompareAndSwapX  . . .
  //               |
  //     . . .    SCMemProj
  //           \   |
  //      |    MergeMem
  //      |        /
  //    MemBarCPUOrder
  //    MemBarAcquire {trailing}
  //
  // this predicate checks for the same flow as the previous predicate
  // but starting from the bottom rather than the top.
  //
  // if the configuration is present returns the cpuorder member for
  // preference or when absent the release membar otherwise NULL.
  //
  // n.b. the input membar is expected to be a MemBarVolatile but
  // need not be a card mark membar.

  MemBarNode *normal_to_leading(const MemBarNode *barrier)
  {
    // input must be a volatile membar
    assert((barrier->Opcode() == Op_MemBarVolatile ||
	    barrier->Opcode() == Op_MemBarAcquire),
	   "expecting a volatile or an acquire membar");
    Node *x;
    bool is_cas = barrier->Opcode() == Op_MemBarAcquire;

    // if we have an acquire membar then it must be fed via a CPUOrder
    // membar

    if (is_cas) {
      // skip to parent barrier which must be a cpuorder
      x = parent_membar(barrier);
      if (x->Opcode() != Op_MemBarCPUOrder)
	return NULL;
    } else {
      // start from the supplied barrier
      x = (Node *)barrier;
    }

    // the Mem feed to the membar should be a merge
    x = x ->in(TypeFunc::Memory);
    if (!x->is_MergeMem())
      return NULL;

    MergeMemNode *mm = x->as_MergeMem();

    if (is_cas) {
      // the merge should be fed from the CAS via an SCMemProj node
      x = NULL;
      for (uint idx = 1; idx < mm->req(); idx++) {
	if (mm->in(idx)->Opcode() == Op_SCMemProj) {
	  x = mm->in(idx);
	  break;
	}
      }
      if (x == NULL) {
	return NULL;
      }
      // check for a CAS feeding this proj
      x = x->in(0);
      int opcode = x->Opcode();
      if (!is_CAS(opcode)) {
	return NULL;
      }
      // the CAS should get its mem feed from the leading membar
      x = x->in(MemNode::Memory);
    } else {
      // the merge should get its Bottom mem feed from the leading membar
      x = mm->in(Compile::AliasIdxBot);
    }

    // ensure this is a non control projection
    if (!x->is_Proj() || x->is_CFG()) {
      return NULL;
    }
    // if it is fed by a membar that's the one we want
    x = x->in(0);

    if (!x->is_MemBar()) {
      return NULL;
    }

    MemBarNode *leading = x->as_MemBar();
    // reject invalid candidates
    if (!leading_membar(leading)) {
      return NULL;
    }

    // ok, we have a leading membar, now for the sanity clauses

    // the leading membar must feed Mem to a releasing store or CAS
    ProjNode *mem = leading->proj_out(TypeFunc::Memory);
    StoreNode *st = NULL;
    LoadStoreNode *cas = NULL;
    for (DUIterator_Fast imax, i = mem->fast_outs(imax); i < imax; i++) {
      x = mem->fast_out(i);
      if (x->is_Store() && x->as_Store()->is_release() && x->Opcode() != Op_StoreCM) {
	// two stores or CASes is one too many
	if (st != NULL || cas != NULL) {
	  return NULL;
	}
	st = x->as_Store();
      } else if (is_CAS(x->Opcode())) {
	if (st != NULL || cas != NULL) {
	  return NULL;
	}
	cas = x->as_LoadStore();
      }
    }

    // we should not have both a store and a cas
    if (st == NULL & cas == NULL) {
      return NULL;
    }

    if (st == NULL) {
      // nothing more to check
      return leading;
    } else {
      // we should not have a store if we started from an acquire
      if (is_cas) {
	return NULL;
      }

      // the store should feed the merge we used to get here
      for (DUIterator_Fast imax, i = st->fast_outs(imax); i < imax; i++) {
	if (st->fast_out(i) == mm) {
	  return leading;
	}
      }
    }

    return NULL;
  }

  // card_mark_to_trailing
  //
  // graph traversal helper which detects extra, non-normal Mem feed
  // from a card mark volatile membar to a trailing membar i.e. it
  // ensures that one of the following three GC post-write Mem flow
  // subgraphs is present.
  //
  // 1)
  //     . . .
  //       |
  //   MemBarVolatile (card mark)
  //      |          |
  //      |        StoreCM
  //      |          |
  //      |        . . .
  //  Bot |  /
  //   MergeMem
  //      |
  //      |
  //    MemBarVolatile {trailing}
  //
  // 2)
  //   MemBarRelease/CPUOrder (leading)
  //    |
  //    |
  //    |\       . . .
  //    | \        |
  //    |  \  MemBarVolatile (card mark)
  //    |   \   |     |
  //     \   \  |   StoreCM    . . .
  //      \   \ |
  //       \  Phi
  //        \ /
  //        Phi  . . .
  //     Bot |   /
  //       MergeMem
  //         |
  //    MemBarVolatile {trailing}
  //
  //
  // 3)
  //   MemBarRelease/CPUOrder (leading)
  //    |
  //    |\
  //    | \
  //    |  \      . . .
  //    |   \       |
  //    |\   \  MemBarVolatile (card mark)
  //    | \   \   |     |
  //    |  \   \  |   StoreCM    . . .
  //    |   \   \ |
  //     \   \  Phi
  //      \   \ /
  //       \  Phi
  //        \ /
  //        Phi  . . .
  //     Bot |   /
  //       MergeMem
  //         |
  //         |
  //    MemBarVolatile {trailing}
  //
  // configuration 1 is only valid if UseConcMarkSweepGC &&
  // UseCondCardMark
  //
  // configurations 2 and 3 are only valid if UseG1GC.
  //
  // if a valid configuration is present returns the trailing membar
  // otherwise NULL.
  //
  // n.b. the supplied membar is expected to be a card mark
  // MemBarVolatile i.e. the caller must ensure the input node has the
  // correct operand and feeds Mem to a StoreCM node

  MemBarNode *card_mark_to_trailing(const MemBarNode *barrier)
  {
    // input must be a card mark volatile membar
    assert(is_card_mark_membar(barrier), "expecting a card mark membar");

    Node *feed = barrier->proj_out(TypeFunc::Memory);
    Node *x;
    MergeMemNode *mm = NULL;

    const int MAX_PHIS = 3;	// max phis we will search through
    int phicount = 0; 		// current search count

    bool retry_feed = true;
    while (retry_feed) {
      // see if we have a direct MergeMem feed
      for (DUIterator_Fast imax, i = feed->fast_outs(imax); i < imax; i++) {
	x = feed->fast_out(i);
	// the correct Phi will be merging a Bot memory slice
	if (x->is_MergeMem()) {
	  mm = x->as_MergeMem();
	  break;
	}
      }
      if (mm) {
	retry_feed = false;
      } else if (UseG1GC & phicount++ < MAX_PHIS) {
	// the barrier may feed indirectly via one or two Phi nodes
	PhiNode *phi = NULL;
	for (DUIterator_Fast imax, i = feed->fast_outs(imax); i < imax; i++) {
	  x = feed->fast_out(i);
	  // the correct Phi will be merging a Bot memory slice
	  if (x->is_Phi() && x->adr_type() == TypePtr::BOTTOM) {
	    phi = x->as_Phi();
	    break;
	  }
	}
	if (!phi) {
	  return NULL;
	}
	// look for another merge below this phi
	feed = phi;
      } else {
	// couldn't find a merge
	return NULL;
      }
    }

    // sanity check this feed turns up as the expected slice
    assert(mm->as_MergeMem()->in(Compile::AliasIdxBot) == feed, "expecting membar to feed AliasIdxBot slice to Merge");

    MemBarNode *trailing = NULL;
    // be sure we have a trailing membar the merge
    for (DUIterator_Fast imax, i = mm->fast_outs(imax); i < imax; i++) {
      x = mm->fast_out(i);
      if (x->is_MemBar() && x->Opcode() == Op_MemBarVolatile) {
	trailing = x->as_MemBar();
	break;
      }
    }

    return trailing;
  }

  // trailing_to_card_mark
  //
  // graph traversal helper which detects extra, non-normal Mem feed
  // from a trailing volatile membar to a preceding card mark volatile
  // membar i.e. it identifies whether one of the three possible extra
  // GC post-write Mem flow subgraphs is present
  //
  // this predicate checks for the same flow as the previous predicate
  // but starting from the bottom rather than the top.
  //
  // if the configuration is present returns the card mark membar
  // otherwise NULL
  //
  // n.b. the supplied membar is expected to be a trailing
  // MemBarVolatile i.e. the caller must ensure the input node has the
  // correct opcode

  MemBarNode *trailing_to_card_mark(const MemBarNode *trailing)
  {
    assert(trailing->Opcode() == Op_MemBarVolatile,
	   "expecting a volatile membar");
    assert(!is_card_mark_membar(trailing),
	   "not expecting a card mark membar");

    // the Mem feed to the membar should be a merge
    Node *x = trailing->in(TypeFunc::Memory);
    if (!x->is_MergeMem()) {
      return NULL;
    }

    MergeMemNode *mm = x->as_MergeMem();

    x = mm->in(Compile::AliasIdxBot);
    // with G1 we may possibly see a Phi or two before we see a Memory
    // Proj from the card mark membar

    const int MAX_PHIS = 3;	// max phis we will search through
    int phicount = 0; 		// current search count

    bool retry_feed = !x->is_Proj();

    while (retry_feed) {
      if (UseG1GC && x->is_Phi() && phicount++ < MAX_PHIS) {
	PhiNode *phi = x->as_Phi();
	ProjNode *proj = NULL;
	PhiNode *nextphi = NULL;
	bool found_leading = false;
	for (uint i = 1; i < phi->req(); i++) {
	  x = phi->in(i);
	  if (x->is_Phi()) {
	    nextphi = x->as_Phi();
	  } else if (x->is_Proj()) {
	    int opcode = x->in(0)->Opcode();
	    if (opcode == Op_MemBarVolatile) {
	      proj = x->as_Proj();
	    } else if (opcode == Op_MemBarRelease ||
		       opcode == Op_MemBarCPUOrder) {
	      // probably a leading membar
	      found_leading = true;
	    }
	  }
	}
	// if we found a correct looking proj then retry from there
	// otherwise we must see a leading and a phi or this the
	// wrong config
	if (proj != NULL) {
	  x = proj;
	  retry_feed = false;
	} else if (found_leading && nextphi != NULL) {
	  // retry from this phi to check phi2
	  x = nextphi;
	} else {
	  // not what we were looking for
	  return NULL;
	}
      } else {
	return NULL;
      }
    }
    // the proj has to come from the card mark membar
    x = x->in(0);
    if (!x->is_MemBar()) {
      return NULL;
    }

    MemBarNode *card_mark_membar = x->as_MemBar();

    if (!is_card_mark_membar(card_mark_membar)) {
      return NULL;
    }

    return card_mark_membar;
  }

  // trailing_to_leading
  //
  // graph traversal helper which checks the Mem flow up the graph
  // from a (non-card mark) trailing membar attempting to locate and
  // return an associated leading membar. it first looks for a
  // subgraph in the normal configuration (relying on helper
  // normal_to_leading). failing that it then looks for one of the
  // possible post-write card mark subgraphs linking the trailing node
  // to a the card mark membar (relying on helper
  // trailing_to_card_mark), and then checks that the card mark membar
  // is fed by a leading membar (once again relying on auxiliary
  // predicate normal_to_leading).
  //
  // if the configuration is valid returns the cpuorder member for
  // preference or when absent the release membar otherwise NULL.
  //
  // n.b. the input membar is expected to be either a volatile or
  // acquire membar but in the former case must *not* be a card mark
  // membar.

  MemBarNode *trailing_to_leading(const MemBarNode *trailing)
  {
    assert((trailing->Opcode() == Op_MemBarAcquire ||
	    trailing->Opcode() == Op_MemBarVolatile),
	   "expecting an acquire or volatile membar");
    assert((trailing->Opcode() != Op_MemBarVolatile ||
	    !is_card_mark_membar(trailing)),
	   "not expecting a card mark membar");

    MemBarNode *leading = normal_to_leading(trailing);

    if (leading) {
      return leading;
    }

    // nothing more to do if this is an acquire
    if (trailing->Opcode() == Op_MemBarAcquire) {
      return NULL;
    }

    MemBarNode *card_mark_membar = trailing_to_card_mark(trailing);

    if (!card_mark_membar) {
      return NULL;
    }

    return normal_to_leading(card_mark_membar);
  }

  // predicates controlling emit of ldr<x>/ldar<x> and associated dmb

bool unnecessary_acquire(const Node *barrier)
{
  assert(barrier->is_MemBar(), "expecting a membar");

  if (UseBarriersForVolatile) {
    // we need to plant a dmb
    return false;
  }

  // a volatile read derived from bytecode (or also from an inlined
  // SHA field read via LibraryCallKit::load_field_from_object)
  // manifests as a LoadX[mo_acquire] followed by an acquire membar
  // with a bogus read dependency on it's preceding load. so in those
  // cases we will find the load node at the PARMS offset of the
  // acquire membar.  n.b. there may be an intervening DecodeN node.
  //
  // a volatile load derived from an inlined unsafe field access
  // manifests as a cpuorder membar with Ctl and Mem projections
  // feeding both an acquire membar and a LoadX[mo_acquire]. The
  // acquire then feeds another cpuorder membar via Ctl and Mem
  // projections. The load has no output dependency on these trailing
  // membars because subsequent nodes inserted into the graph take
  // their control feed from the final membar cpuorder meaning they
  // are all ordered after the load.

  Node *x = barrier->lookup(TypeFunc::Parms);
  if (x) {
    // we are starting from an acquire and it has a fake dependency
    //
    // need to check for
    //
    //   LoadX[mo_acquire]
    //   {  |1   }
    //   {DecodeN}
    //      |Parms
    //   MemBarAcquire*
    //
    // where * tags node we were passed
    // and |k means input k
    if (x->is_DecodeNarrowPtr()) {
      x = x->in(1);
    }

    return (x->is_Load() && x->as_Load()->is_acquire());
  }

  // now check for an unsafe volatile get

  // need to check for
  //
  //   MemBarCPUOrder
  //        ||       \\
  //   MemBarAcquire* LoadX[mo_acquire]
  //        ||
  //   MemBarCPUOrder
  //
  // where * tags node we were passed
  // and || or \\ are Ctl+Mem feeds via intermediate Proj Nodes

  // check for a parent MemBarCPUOrder
  ProjNode *ctl;
  ProjNode *mem;
  MemBarNode *parent = parent_membar(barrier);
  if (!parent || parent->Opcode() != Op_MemBarCPUOrder)
    return false;
  ctl = parent->proj_out(TypeFunc::Control);
  mem = parent->proj_out(TypeFunc::Memory);
  if (!ctl || !mem) {
    return false;
  }
  // ensure the proj nodes both feed a LoadX[mo_acquire]
  LoadNode *ld = NULL;
  for (DUIterator_Fast imax, i = ctl->fast_outs(imax); i < imax; i++) {
    x = ctl->fast_out(i);
    // if we see a load we keep hold of it and stop searching
    if (x->is_Load()) {
      ld = x->as_Load();
      break;
    }
  }
  // it must be an acquiring load
  if (ld && ld->is_acquire()) {

    for (DUIterator_Fast imax, i = mem->fast_outs(imax); i < imax; i++) {
      x = mem->fast_out(i);
      // if we see the same load we drop it and stop searching
      if (x == ld) {
	ld = NULL;
	break;
      }
    }
    // we must have dropped the load
    if (ld == NULL) {
      // check for a child cpuorder membar
      MemBarNode *child  = child_membar(barrier->as_MemBar());
      if (child && child->Opcode() == Op_MemBarCPUOrder)
	return true;
    }
  }

  // final option for unnecessary mebar is that it is a trailing node
  // belonging to a CAS

  MemBarNode *leading = trailing_to_leading(barrier->as_MemBar());

  return leading != NULL;
}

bool needs_acquiring_load(const Node *n)
{
  assert(n->is_Load(), "expecting a load");
  if (UseBarriersForVolatile) {
    // we use a normal load and a dmb
    return false;
  }

  LoadNode *ld = n->as_Load();

  if (!ld->is_acquire()) {
    return false;
  }

  // check if this load is feeding an acquire membar
  //
  //   LoadX[mo_acquire]
  //   {  |1   }
  //   {DecodeN}
  //      |Parms
  //   MemBarAcquire*
  //
  // where * tags node we were passed
  // and |k means input k

  Node *start = ld;
  Node *mbacq = NULL;

  // if we hit a DecodeNarrowPtr we reset the start node and restart
  // the search through the outputs
 restart:

  for (DUIterator_Fast imax, i = start->fast_outs(imax); i < imax; i++) {
    Node *x = start->fast_out(i);
    if (x->is_MemBar() && x->Opcode() == Op_MemBarAcquire) {
      mbacq = x;
    } else if (!mbacq &&
	       (x->is_DecodeNarrowPtr() ||
		(x->is_Mach() && x->Opcode() == Op_DecodeN))) {
      start = x;
      goto restart;
    }
  }

  if (mbacq) {
    return true;
  }

  // now check for an unsafe volatile get

  // check if Ctl and Proj feed comes from a MemBarCPUOrder
  //
  //     MemBarCPUOrder
  //        ||       \\
  //   MemBarAcquire* LoadX[mo_acquire]
  //        ||
  //   MemBarCPUOrder

  MemBarNode *membar;

  membar = parent_membar(ld);

  if (!membar || !membar->Opcode() == Op_MemBarCPUOrder) {
    return false;
  }

  // ensure that there is a CPUOrder->Acquire->CPUOrder membar chain

  membar = child_membar(membar);

  if (!membar || !membar->Opcode() == Op_MemBarAcquire) {
    return false;
  }

  membar = child_membar(membar);

  if (!membar || !membar->Opcode() == Op_MemBarCPUOrder) {
    return false;
  }

  return true;
}

bool unnecessary_release(const Node *n)
{
  assert((n->is_MemBar() &&
	  n->Opcode() == Op_MemBarRelease),
	 "expecting a release membar");

  if (UseBarriersForVolatile) {
    // we need to plant a dmb
    return false;
  }

  // if there is a dependent CPUOrder barrier then use that as the
  // leading

  MemBarNode *barrier = n->as_MemBar();
  // check for an intervening cpuorder membar
  MemBarNode *b = child_membar(barrier);
  if (b && b->Opcode() == Op_MemBarCPUOrder) {
    // ok, so start the check from the dependent cpuorder barrier
    barrier = b;
  }

  // must start with a normal feed
  MemBarNode *child_barrier = leading_to_normal(barrier);

  if (!child_barrier) {
    return false;
  }

  if (!is_card_mark_membar(child_barrier)) {
    // this is the trailing membar and we are done
    return true;
  }

  // must be sure this card mark feeds a trailing membar
  MemBarNode *trailing = card_mark_to_trailing(child_barrier);
  return (trailing != NULL);
}

bool unnecessary_volatile(const Node *n)
{
  // assert n->is_MemBar();
  if (UseBarriersForVolatile) {
    // we need to plant a dmb
    return false;
  }

  MemBarNode *mbvol = n->as_MemBar();

  // first we check if this is part of a card mark. if so then we have
  // to generate a StoreLoad barrier

  if (is_card_mark_membar(mbvol)) {
      return false;
  }

  // ok, if it's not a card mark then we still need to check if it is
  // a trailing membar of a volatile put hgraph.

  return (trailing_to_leading(mbvol) != NULL);
}

// predicates controlling emit of str<x>/stlr<x> and associated dmbs

bool needs_releasing_store(const Node *n)
{
  // assert n->is_Store();
  if (UseBarriersForVolatile) {
    // we use a normal store and dmb combination
    return false;
  }

  StoreNode *st = n->as_Store();

  // the store must be marked as releasing
  if (!st->is_release()) {
    return false;
  }

  // the store must be fed by a membar

  Node *x = st->lookup(StoreNode::Memory);

  if (! x || !x->is_Proj()) {
    return false;
  }

  ProjNode *proj = x->as_Proj();

  x = proj->lookup(0);

  if (!x || !x->is_MemBar()) {
    return false;
  }

  MemBarNode *barrier = x->as_MemBar();

  // if the barrier is a release membar or a cpuorder mmebar fed by a
  // release membar then we need to check whether that forms part of a
  // volatile put graph.

  // reject invalid candidates
  if (!leading_membar(barrier)) {
    return false;
  }

  // does this lead a normal subgraph?
  MemBarNode *mbvol = leading_to_normal(barrier);

  if (!mbvol) {
    return false;
  }

  // all done unless this is a card mark
  if (!is_card_mark_membar(mbvol)) {
    return true;
  }

  // we found a card mark -- just make sure we have a trailing barrier

  return (card_mark_to_trailing(mbvol) != NULL);
}

// predicate controlling translation of CAS
//
// returns true if CAS needs to use an acquiring load otherwise false

bool needs_acquiring_load_exclusive(const Node *n)
{
  assert(is_CAS(n->Opcode()), "expecting a compare and swap");
  if (UseBarriersForVolatile) {
    return false;
  }

  // CAS nodes only ought to turn up in inlined unsafe CAS operations
#ifdef ASSERT
  LoadStoreNode *st = n->as_LoadStore();

  // the store must be fed by a membar

  Node *x = st->lookup(StoreNode::Memory);

  assert (x && x->is_Proj(), "CAS not fed by memory proj!");

  ProjNode *proj = x->as_Proj();

  x = proj->lookup(0);

  assert (x && x->is_MemBar(), "CAS not fed by membar!");

  MemBarNode *barrier = x->as_MemBar();

  // the barrier must be a cpuorder mmebar fed by a release membar

  assert(barrier->Opcode() == Op_MemBarCPUOrder,
	 "CAS not fed by cpuorder membar!");

  MemBarNode *b = parent_membar(barrier);
  assert ((b != NULL && b->Opcode() == Op_MemBarRelease),
	  "CAS not fed by cpuorder+release membar pair!");

  // does this lead a normal subgraph?
  MemBarNode *mbar = leading_to_normal(barrier);

  assert(mbar != NULL, "CAS not embedded in normal graph!");

  assert(mbar->Opcode() == Op_MemBarAcquire, "trailing membar should be an acquire");
#endif // ASSERT
  // so we can just return true here
  return true;
}

// predicate controlling translation of StoreCM
//
// returns true if a StoreStore must precede the card write otherwise
// false

bool unnecessary_storestore(const Node *storecm)
{
  assert(storecm->Opcode()  == Op_StoreCM, "expecting a StoreCM");

  // we only ever need to generate a dmb ishst between an object put
  // and the associated card mark when we are using CMS without
  // conditional card marking

  if (!UseConcMarkSweepGC || UseCondCardMark) {
    return true;
  }

  // if we are implementing volatile puts using barriers then the
  // object put as an str so we must insert the dmb ishst

  if (UseBarriersForVolatile) {
    return false;
  }

  // we can omit the dmb ishst if this StoreCM is part of a volatile
  // put because in thta case the put will be implemented by stlr
  //
  // we need to check for a normal subgraph feeding this StoreCM.
  // that means the StoreCM must be fed Memory from a leading membar,
  // either a MemBarRelease or its dependent MemBarCPUOrder, and the
  // leading membar must be part of a normal subgraph

  Node *x = storecm->in(StoreNode::Memory);

  if (!x->is_Proj()) {
    return false;
  }

  x = x->in(0);

  if (!x->is_MemBar()) {
    return false;
  }

  MemBarNode *leading = x->as_MemBar();

  // reject invalid candidates
  if (!leading_membar(leading)) {
    return false;
  }

  // we can omit the StoreStore if it is the head of a normal subgraph
  return (leading_to_normal(leading) != NULL);
}


#define __ _masm.

// advance declarations for helper functions to convert register
// indices to register objects

// the ad file has to provide implementations of certain methods
// expected by the generic code
//
// REQUIRED FUNCTIONALITY

//=============================================================================

// !!!!! Special hack to get all types of calls to specify the byte offset
//       from the start of the call to the point where the return address
//       will point.

int MachCallStaticJavaNode::ret_addr_offset()
{
  // call should be a simple bl
  int off = 4;
  return off;
}

int MachCallDynamicJavaNode::ret_addr_offset()
{
  return 16; // movz, movk, movk, bl
}

int MachCallRuntimeNode::ret_addr_offset() {
  // for generated stubs the call will be
  //   far_call(addr)
  // for real runtime callouts it will be six instructions
  // see aarch64_enc_java_to_runtime
  //   adr(rscratch2, retaddr)
  //   lea(rscratch1, RuntimeAddress(addr)
  //   stp(zr, rscratch2, Address(__ pre(sp, -2 * wordSize)))
  //   blrt rscratch1
  CodeBlob *cb = CodeCache::find_blob(_entry_point);
  if (cb) {
    return MacroAssembler::far_branch_size();
  } else {
    return 6 * NativeInstruction::instruction_size;
  }
}

// Indicate if the safepoint node needs the polling page as an input

// the shared code plants the oop data at the start of the generated
// code for the safepoint node and that needs ot be at the load
// instruction itself. so we cannot plant a mov of the safepoint poll
// address followed by a load. setting this to true means the mov is
// scheduled as a prior instruction. that's better for scheduling
// anyway.

bool SafePointNode::needs_polling_address_input()
{
  return true;
}

//=============================================================================

#ifndef PRODUCT
void MachBreakpointNode::format(PhaseRegAlloc *ra_, outputStream *st) const {
  st->print("BREAKPOINT");
}
#endif

void MachBreakpointNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
  MacroAssembler _masm(&cbuf);
  __ brk(0);
}

uint MachBreakpointNode::size(PhaseRegAlloc *ra_) const {
  return MachNode::size(ra_);
}

//=============================================================================

#ifndef PRODUCT
  void MachNopNode::format(PhaseRegAlloc*, outputStream* st) const {
    st->print("nop \t# %d bytes pad for loops and calls", _count);
  }
#endif

  void MachNopNode::emit(CodeBuffer &cbuf, PhaseRegAlloc*) const {
    MacroAssembler _masm(&cbuf);
    for (int i = 0; i < _count; i++) {
      __ nop();
    }
  }

  uint MachNopNode::size(PhaseRegAlloc*) const {
    return _count * NativeInstruction::instruction_size;
  }

//=============================================================================
const RegMask& MachConstantBaseNode::_out_RegMask = RegMask::Empty;

int Compile::ConstantTable::calculate_table_base_offset() const {
  return 0;  // absolute addressing, no offset
}

bool MachConstantBaseNode::requires_postalloc_expand() const { return false; }
void MachConstantBaseNode::postalloc_expand(GrowableArray <Node *> *nodes, PhaseRegAlloc *ra_) {
  ShouldNotReachHere();
}

void MachConstantBaseNode::emit(CodeBuffer& cbuf, PhaseRegAlloc* ra_) const {
  // Empty encoding
}

uint MachConstantBaseNode::size(PhaseRegAlloc* ra_) const {
  return 0;
}

#ifndef PRODUCT
void MachConstantBaseNode::format(PhaseRegAlloc* ra_, outputStream* st) const {
  st->print("-- \t// MachConstantBaseNode (empty encoding)");
}
#endif

#ifndef PRODUCT
void MachPrologNode::format(PhaseRegAlloc *ra_, outputStream *st) const {
  Compile* C = ra_->C;

  int framesize = C->frame_slots() << LogBytesPerInt;

  if (C->need_stack_bang(framesize))
    st->print("# stack bang size=%d\n\t", framesize);

  if (framesize < ((1 << 9) + 2 * wordSize)) {
    st->print("sub  sp, sp, #%d\n\t", framesize);
    st->print("stp  rfp, lr, [sp, #%d]", framesize - 2 * wordSize);
    if (PreserveFramePointer) st->print("\n\tadd  rfp, sp, #%d", framesize - 2 * wordSize);
  } else {
    st->print("stp  lr, rfp, [sp, #%d]!\n\t", -(2 * wordSize));
    if (PreserveFramePointer) st->print("mov  rfp, sp\n\t");
    st->print("mov  rscratch1, #%d\n\t", framesize - 2 * wordSize);
    st->print("sub  sp, sp, rscratch1");
  }
}
#endif

void MachPrologNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
  Compile* C = ra_->C;
  MacroAssembler _masm(&cbuf);

  // n.b. frame size includes space for return pc and rfp
  const long framesize = C->frame_size_in_bytes();
  assert(framesize%(2*wordSize) == 0, "must preserve 2*wordSize alignment");

  // insert a nop at the start of the prolog so we can patch in a
  // branch if we need to invalidate the method later
  __ nop();

  int bangsize = C->bang_size_in_bytes();
  if (C->need_stack_bang(bangsize) && UseStackBanging)
    __ generate_stack_overflow_check(bangsize);

  __ build_frame(framesize);

  if (NotifySimulator) {
    __ notify(Assembler::method_entry);
  }

  if (VerifyStackAtCalls) {
    Unimplemented();
  }

  C->set_frame_complete(cbuf.insts_size());

  if (C->has_mach_constant_base_node()) {
    // NOTE: We set the table base offset here because users might be
    // emitted before MachConstantBaseNode.
    Compile::ConstantTable& constant_table = C->constant_table();
    constant_table.set_table_base_offset(constant_table.calculate_table_base_offset());
  }
}

uint MachPrologNode::size(PhaseRegAlloc* ra_) const
{
  return MachNode::size(ra_); // too many variables; just compute it
                              // the hard way
}

int MachPrologNode::reloc() const
{
  return 0;
}

//=============================================================================

#ifndef PRODUCT
void MachEpilogNode::format(PhaseRegAlloc *ra_, outputStream *st) const {
  Compile* C = ra_->C;
  int framesize = C->frame_slots() << LogBytesPerInt;

  st->print("# pop frame %d\n\t",framesize);

  if (framesize == 0) {
    st->print("ldp  lr, rfp, [sp],#%d\n\t", (2 * wordSize));
  } else if (framesize < ((1 << 9) + 2 * wordSize)) {
    st->print("ldp  lr, rfp, [sp,#%d]\n\t", framesize - 2 * wordSize);
    st->print("add  sp, sp, #%d\n\t", framesize);
  } else {
    st->print("mov  rscratch1, #%d\n\t", framesize - 2 * wordSize);
    st->print("add  sp, sp, rscratch1\n\t");
    st->print("ldp  lr, rfp, [sp],#%d\n\t", (2 * wordSize));
  }

  if (do_polling() && C->is_method_compilation()) {
    st->print("# touch polling page\n\t");
    st->print("mov  rscratch1, #0x%lx\n\t", p2i(os::get_polling_page()));
    st->print("ldr zr, [rscratch1]");
  }
}
#endif

void MachEpilogNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
  Compile* C = ra_->C;
  MacroAssembler _masm(&cbuf);
  int framesize = C->frame_slots() << LogBytesPerInt;

  __ remove_frame(framesize);

  if (NotifySimulator) {
    __ notify(Assembler::method_reentry);
  }

  if (StackReservedPages > 0 && C->has_reserved_stack_access()) {
    __ reserved_stack_check();
  }

  if (do_polling() && C->is_method_compilation()) {
    __ read_polling_page(rscratch1, os::get_polling_page(), relocInfo::poll_return_type);
  }
}

uint MachEpilogNode::size(PhaseRegAlloc *ra_) const {
  // Variable size. Determine dynamically.
  return MachNode::size(ra_);
}

int MachEpilogNode::reloc() const {
  // Return number of relocatable values contained in this instruction.
  return 1; // 1 for polling page.
}

const Pipeline * MachEpilogNode::pipeline() const {
  return MachNode::pipeline_class();
}

// This method seems to be obsolete. It is declared in machnode.hpp
// and defined in all *.ad files, but it is never called. Should we
// get rid of it?
int MachEpilogNode::safepoint_offset() const {
  assert(do_polling(), "no return for this epilog node");
  return 4;
}

//=============================================================================

// Figure out which register class each belongs in: rc_int, rc_float or
// rc_stack.
enum RC { rc_bad, rc_int, rc_float, rc_stack };

static enum RC rc_class(OptoReg::Name reg) {

  if (reg == OptoReg::Bad) {
    return rc_bad;
  }

  // we have 30 int registers * 2 halves
  // (rscratch1 and rscratch2 are omitted)

  if (reg < 60) {
    return rc_int;
  }

  // we have 32 float register * 2 halves
  if (reg < 60 + 128) {
    return rc_float;
  }

  // Between float regs & stack is the flags regs.
  assert(OptoReg::is_stack(reg), "blow up if spilling flags");

  return rc_stack;
}

uint MachSpillCopyNode::implementation(CodeBuffer *cbuf, PhaseRegAlloc *ra_, bool do_size, outputStream *st) const {
  Compile* C = ra_->C;

  // Get registers to move.
  OptoReg::Name src_hi = ra_->get_reg_second(in(1));
  OptoReg::Name src_lo = ra_->get_reg_first(in(1));
  OptoReg::Name dst_hi = ra_->get_reg_second(this);
  OptoReg::Name dst_lo = ra_->get_reg_first(this);

  enum RC src_hi_rc = rc_class(src_hi);
  enum RC src_lo_rc = rc_class(src_lo);
  enum RC dst_hi_rc = rc_class(dst_hi);
  enum RC dst_lo_rc = rc_class(dst_lo);

  assert(src_lo != OptoReg::Bad && dst_lo != OptoReg::Bad, "must move at least 1 register");

  if (src_hi != OptoReg::Bad) {
    assert((src_lo&1)==0 && src_lo+1==src_hi &&
           (dst_lo&1)==0 && dst_lo+1==dst_hi,
           "expected aligned-adjacent pairs");
  }

  if (src_lo == dst_lo && src_hi == dst_hi) {
    return 0;            // Self copy, no move.
  }

  bool is64 = (src_lo & 1) == 0 && src_lo + 1 == src_hi &&
              (dst_lo & 1) == 0 && dst_lo + 1 == dst_hi;
  int src_offset = ra_->reg2offset(src_lo);
  int dst_offset = ra_->reg2offset(dst_lo);

  if (bottom_type()->isa_vect() != NULL) {
    uint ireg = ideal_reg();
    assert(ireg == Op_VecD || ireg == Op_VecX, "must be 64 bit or 128 bit vector");
    if (cbuf) {
      MacroAssembler _masm(cbuf);
      assert((src_lo_rc != rc_int && dst_lo_rc != rc_int), "sanity");
      if (src_lo_rc == rc_stack && dst_lo_rc == rc_stack) {
        // stack->stack
        assert((src_offset & 7) == 0 && (dst_offset & 7) == 0, "unaligned stack offset");
        if (ireg == Op_VecD) {
          __ unspill(rscratch1, true, src_offset);
          __ spill(rscratch1, true, dst_offset);
        } else {
          __ spill_copy128(src_offset, dst_offset);
        }
      } else if (src_lo_rc == rc_float && dst_lo_rc == rc_float) {
        __ mov(as_FloatRegister(Matcher::_regEncode[dst_lo]),
               ireg == Op_VecD ? __ T8B : __ T16B,
               as_FloatRegister(Matcher::_regEncode[src_lo]));
      } else if (src_lo_rc == rc_float && dst_lo_rc == rc_stack) {
        __ spill(as_FloatRegister(Matcher::_regEncode[src_lo]),
                       ireg == Op_VecD ? __ D : __ Q,
                       ra_->reg2offset(dst_lo));
      } else if (src_lo_rc == rc_stack && dst_lo_rc == rc_float) {
        __ unspill(as_FloatRegister(Matcher::_regEncode[dst_lo]),
                       ireg == Op_VecD ? __ D : __ Q,
                       ra_->reg2offset(src_lo));
      } else {
        ShouldNotReachHere();
      }
    }
  } else if (cbuf) {
    MacroAssembler _masm(cbuf);
    switch (src_lo_rc) {
    case rc_int:
      if (dst_lo_rc == rc_int) {  // gpr --> gpr copy
        if (is64) {
            __ mov(as_Register(Matcher::_regEncode[dst_lo]),
                   as_Register(Matcher::_regEncode[src_lo]));
        } else {
            MacroAssembler _masm(cbuf);
            __ movw(as_Register(Matcher::_regEncode[dst_lo]),
                    as_Register(Matcher::_regEncode[src_lo]));
        }
      } else if (dst_lo_rc == rc_float) { // gpr --> fpr copy
        if (is64) {
            __ fmovd(as_FloatRegister(Matcher::_regEncode[dst_lo]),
                     as_Register(Matcher::_regEncode[src_lo]));
        } else {
            __ fmovs(as_FloatRegister(Matcher::_regEncode[dst_lo]),
                     as_Register(Matcher::_regEncode[src_lo]));
        }
      } else {                    // gpr --> stack spill
        assert(dst_lo_rc == rc_stack, "spill to bad register class");
        __ spill(as_Register(Matcher::_regEncode[src_lo]), is64, dst_offset);
      }
      break;
    case rc_float:
      if (dst_lo_rc == rc_int) {  // fpr --> gpr copy
        if (is64) {
            __ fmovd(as_Register(Matcher::_regEncode[dst_lo]),
                     as_FloatRegister(Matcher::_regEncode[src_lo]));
        } else {
            __ fmovs(as_Register(Matcher::_regEncode[dst_lo]),
                     as_FloatRegister(Matcher::_regEncode[src_lo]));
        }
      } else if (dst_lo_rc == rc_float) { // fpr --> fpr copy
          if (cbuf) {
            __ fmovd(as_FloatRegister(Matcher::_regEncode[dst_lo]),
                     as_FloatRegister(Matcher::_regEncode[src_lo]));
        } else {
            __ fmovs(as_FloatRegister(Matcher::_regEncode[dst_lo]),
                     as_FloatRegister(Matcher::_regEncode[src_lo]));
        }
      } else {                    // fpr --> stack spill
        assert(dst_lo_rc == rc_stack, "spill to bad register class");
        __ spill(as_FloatRegister(Matcher::_regEncode[src_lo]),
                 is64 ? __ D : __ S, dst_offset);
      }
      break;
    case rc_stack:
      if (dst_lo_rc == rc_int) {  // stack --> gpr load
        __ unspill(as_Register(Matcher::_regEncode[dst_lo]), is64, src_offset);
      } else if (dst_lo_rc == rc_float) { // stack --> fpr load
        __ unspill(as_FloatRegister(Matcher::_regEncode[dst_lo]),
                   is64 ? __ D : __ S, src_offset);
      } else {                    // stack --> stack copy
        assert(dst_lo_rc == rc_stack, "spill to bad register class");
        __ unspill(rscratch1, is64, src_offset);
        __ spill(rscratch1, is64, dst_offset);
      }
      break;
    default:
      assert(false, "bad rc_class for spill");
      ShouldNotReachHere();
    }
  }

  if (st) {
    st->print("spill ");
    if (src_lo_rc == rc_stack) {
      st->print("[sp, #%d] -> ", ra_->reg2offset(src_lo));
    } else {
      st->print("%s -> ", Matcher::regName[src_lo]);
    }
    if (dst_lo_rc == rc_stack) {
      st->print("[sp, #%d]", ra_->reg2offset(dst_lo));
    } else {
      st->print("%s", Matcher::regName[dst_lo]);
    }
    if (bottom_type()->isa_vect() != NULL) {
      st->print("\t# vector spill size = %d", ideal_reg()==Op_VecD ? 64:128);
    } else {
      st->print("\t# spill size = %d", is64 ? 64:32);
    }
  }

  return 0;

}

#ifndef PRODUCT
void MachSpillCopyNode::format(PhaseRegAlloc *ra_, outputStream *st) const {
  if (!ra_)
    st->print("N%d = SpillCopy(N%d)", _idx, in(1)->_idx);
  else
    implementation(NULL, ra_, false, st);
}
#endif

void MachSpillCopyNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
  implementation(&cbuf, ra_, false, NULL);
}

uint MachSpillCopyNode::size(PhaseRegAlloc *ra_) const {
  return MachNode::size(ra_);
}

//=============================================================================

#ifndef PRODUCT
void BoxLockNode::format(PhaseRegAlloc *ra_, outputStream *st) const {
  int offset = ra_->reg2offset(in_RegMask(0).find_first_elem());
  int reg = ra_->get_reg_first(this);
  st->print("add %s, rsp, #%d]\t# box lock",
            Matcher::regName[reg], offset);
}
#endif

void BoxLockNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
  MacroAssembler _masm(&cbuf);

  int offset = ra_->reg2offset(in_RegMask(0).find_first_elem());
  int reg    = ra_->get_encode(this);

  if (Assembler::operand_valid_for_add_sub_immediate(offset)) {
    __ add(as_Register(reg), sp, offset);
  } else {
    ShouldNotReachHere();
  }
}

uint BoxLockNode::size(PhaseRegAlloc *ra_) const {
  // BoxLockNode is not a MachNode, so we can't just call MachNode::size(ra_).
  return 4;
}

//=============================================================================

#ifndef PRODUCT
void MachUEPNode::format(PhaseRegAlloc* ra_, outputStream* st) const
{
  st->print_cr("# MachUEPNode");
  if (UseCompressedClassPointers) {
    st->print_cr("\tldrw rscratch1, j_rarg0 + oopDesc::klass_offset_in_bytes()]\t# compressed klass");
    if (Universe::narrow_klass_shift() != 0) {
      st->print_cr("\tdecode_klass_not_null rscratch1, rscratch1");
    }
  } else {
   st->print_cr("\tldr rscratch1, j_rarg0 + oopDesc::klass_offset_in_bytes()]\t# compressed klass");
  }
  st->print_cr("\tcmp r0, rscratch1\t # Inline cache check");
  st->print_cr("\tbne, SharedRuntime::_ic_miss_stub");
}
#endif

void MachUEPNode::emit(CodeBuffer& cbuf, PhaseRegAlloc* ra_) const
{
  // This is the unverified entry point.
  MacroAssembler _masm(&cbuf);

  __ cmp_klass(j_rarg0, rscratch2, rscratch1);
  Label skip;
  // TODO
  // can we avoid this skip and still use a reloc?
  __ br(Assembler::EQ, skip);
  __ far_jump(RuntimeAddress(SharedRuntime::get_ic_miss_stub()));
  __ bind(skip);
}

uint MachUEPNode::size(PhaseRegAlloc* ra_) const
{
  return MachNode::size(ra_);
}

// REQUIRED EMIT CODE

//=============================================================================

// Emit exception handler code.
int HandlerImpl::emit_exception_handler(CodeBuffer& cbuf)
{
  // mov rscratch1 #exception_blob_entry_point
  // br rscratch1
  // Note that the code buffer's insts_mark is always relative to insts.
  // That's why we must use the macroassembler to generate a handler.
  MacroAssembler _masm(&cbuf);
  address base = __ start_a_stub(size_exception_handler());
  if (base == NULL) {
    ciEnv::current()->record_failure("CodeCache is full");
    return 0;  // CodeBuffer::expand failed
  }
  int offset = __ offset();
  __ far_jump(RuntimeAddress(OptoRuntime::exception_blob()->entry_point()));
  assert(__ offset() - offset <= (int) size_exception_handler(), "overflow");
  __ end_a_stub();
  return offset;
}

// Emit deopt handler code.
int HandlerImpl::emit_deopt_handler(CodeBuffer& cbuf)
{
  // Note that the code buffer's insts_mark is always relative to insts.
  // That's why we must use the macroassembler to generate a handler.
  MacroAssembler _masm(&cbuf);
  address base = __ start_a_stub(size_deopt_handler());
  if (base == NULL) {
    ciEnv::current()->record_failure("CodeCache is full");
    return 0;  // CodeBuffer::expand failed
  }
  int offset = __ offset();

  __ adr(lr, __ pc());
  __ far_jump(RuntimeAddress(SharedRuntime::deopt_blob()->unpack()));

  assert(__ offset() - offset <= (int) size_deopt_handler(), "overflow");
  __ end_a_stub();
  return offset;
}

// REQUIRED MATCHER CODE

//=============================================================================

const bool Matcher::match_rule_supported(int opcode) {

  switch (opcode) {
  default:
    break;
  }

  if (!has_match_rule(opcode)) {
    return false;
  }

  return true;  // Per default match rules are supported.
}

const bool Matcher::match_rule_supported_vector(int opcode, int vlen) {

  // TODO
  // identify extra cases that we might want to provide match rules for
  // e.g. Op_ vector nodes and other intrinsics while guarding with vlen
  bool ret_value = match_rule_supported(opcode);
  // Add rules here.

  return ret_value;  // Per default match rules are supported.
}

const bool Matcher::has_predicated_vectors(void) {
  return false;
}

const int Matcher::float_pressure(int default_pressure_threshold) {
  return default_pressure_threshold;
}

int Matcher::regnum_to_fpu_offset(int regnum)
{
  Unimplemented();
  return 0;
}

// Is this branch offset short enough that a short branch can be used?
//
// NOTE: If the platform does not provide any short branch variants, then
//       this method should return false for offset 0.
bool Matcher::is_short_branch_offset(int rule, int br_size, int offset) {
  // The passed offset is relative to address of the branch.

  return (-32768 <= offset && offset < 32768);
}

const bool Matcher::isSimpleConstant64(jlong value) {
  // Will one (StoreL ConL) be cheaper than two (StoreI ConI)?.
  // Probably always true, even if a temp register is required.
  return true;
}

// true just means we have fast l2f conversion
const bool Matcher::convL2FSupported(void) {
  return true;
}

// Vector width in bytes.
const int Matcher::vector_width_in_bytes(BasicType bt) {
  int size = MIN2(16,(int)MaxVectorSize);
  // Minimum 2 values in vector
  if (size < 2*type2aelembytes(bt)) size = 0;
  // But never < 4
  if (size < 4) size = 0;
  return size;
}

// Limits on vector size (number of elements) loaded into vector.
const int Matcher::max_vector_size(const BasicType bt) {
  return vector_width_in_bytes(bt)/type2aelembytes(bt);
}
const int Matcher::min_vector_size(const BasicType bt) {
//  For the moment limit the vector size to 8 bytes
    int size = 8 / type2aelembytes(bt);
    if (size < 2) size = 2;
    return size;
}

// Vector ideal reg.
const int Matcher::vector_ideal_reg(int len) {
  switch(len) {
    case  8: return Op_VecD;
    case 16: return Op_VecX;
  }
  ShouldNotReachHere();
  return 0;
}

const int Matcher::vector_shift_count_ideal_reg(int size) {
  return Op_VecX;
}

// AES support not yet implemented
const bool Matcher::pass_original_key_for_aes() {
  return false;
}

// x86 supports misaligned vectors store/load.
const bool Matcher::misaligned_vectors_ok() {
  return !AlignVector; // can be changed by flag
}

// false => size gets scaled to BytesPerLong, ok.
const bool Matcher::init_array_count_is_in_bytes = false;

// Use conditional move (CMOVL)
const int Matcher::long_cmove_cost() {
  // long cmoves are no more expensive than int cmoves
  return 0;
}

const int Matcher::float_cmove_cost() {
  // float cmoves are no more expensive than int cmoves
  return 0;
}

// Does the CPU require late expand (see block.cpp for description of late expand)?
const bool Matcher::require_postalloc_expand = false;

// Do we need to mask the count passed to shift instructions or does
// the cpu only look at the lower 5/6 bits anyway?
const bool Matcher::need_masked_shift_count = false;

// This affects two different things:
//  - how Decode nodes are matched
//  - how ImplicitNullCheck opportunities are recognized
// If true, the matcher will try to remove all Decodes and match them
// (as operands) into nodes. NullChecks are not prepared to deal with
// Decodes by final_graph_reshaping().
// If false, final_graph_reshaping() forces the decode behind the Cmp
// for a NullCheck. The matcher matches the Decode node into a register.
// Implicit_null_check optimization moves the Decode along with the
// memory operation back up before the NullCheck.
bool Matcher::narrow_oop_use_complex_address() {
  return Universe::narrow_oop_shift() == 0;
}

bool Matcher::narrow_klass_use_complex_address() {
// TODO
// decide whether we need to set this to true
  return false;
}

bool Matcher::const_oop_prefer_decode() {
  // Prefer ConN+DecodeN over ConP in simple compressed oops mode.
  return Universe::narrow_oop_base() == NULL;
}

bool Matcher::const_klass_prefer_decode() {
  // Prefer ConNKlass+DecodeNKlass over ConP in simple compressed klass mode.
  return Universe::narrow_klass_base() == NULL;
}

// Is it better to copy float constants, or load them directly from
// memory?  Intel can load a float constant from a direct address,
// requiring no extra registers.  Most RISCs will have to materialize
// an address into a register first, so they would do better to copy
// the constant from stack.
const bool Matcher::rematerialize_float_constants = false;

// If CPU can load and store mis-aligned doubles directly then no
// fixup is needed.  Else we split the double into 2 integer pieces
// and move it piece-by-piece.  Only happens when passing doubles into
// C code as the Java calling convention forces doubles to be aligned.
const bool Matcher::misaligned_doubles_ok = true;

// No-op on amd64
void Matcher::pd_implicit_null_fixup(MachNode *node, uint idx) {
  Unimplemented();
}

// Advertise here if the CPU requires explicit rounding operations to
// implement the UseStrictFP mode.
const bool Matcher::strict_fp_requires_explicit_rounding = false;

// Are floats converted to double when stored to stack during
// deoptimization?
bool Matcher::float_in_double() { return true; }

// Do ints take an entire long register or just half?
// The relevant question is how the int is callee-saved:
// the whole long is written but de-opt'ing will have to extract
// the relevant 32 bits.
const bool Matcher::int_in_long = true;

// Return whether or not this register is ever used as an argument.
// This function is used on startup to build the trampoline stubs in
// generateOptoStub.  Registers not mentioned will be killed by the VM
// call in the trampoline, and arguments in those registers not be
// available to the callee.
bool Matcher::can_be_java_arg(int reg)
{
  return
    reg ==  R0_num || reg == R0_H_num ||
    reg ==  R1_num || reg == R1_H_num ||
    reg ==  R2_num || reg == R2_H_num ||
    reg ==  R3_num || reg == R3_H_num ||
    reg ==  R4_num || reg == R4_H_num ||
    reg ==  R5_num || reg == R5_H_num ||
    reg ==  R6_num || reg == R6_H_num ||
    reg ==  R7_num || reg == R7_H_num ||
    reg ==  V0_num || reg == V0_H_num ||
    reg ==  V1_num || reg == V1_H_num ||
    reg ==  V2_num || reg == V2_H_num ||
    reg ==  V3_num || reg == V3_H_num ||
    reg ==  V4_num || reg == V4_H_num ||
    reg ==  V5_num || reg == V5_H_num ||
    reg ==  V6_num || reg == V6_H_num ||
    reg ==  V7_num || reg == V7_H_num;
}

bool Matcher::is_spillable_arg(int reg)
{
  return can_be_java_arg(reg);
}

bool Matcher::use_asm_for_ldiv_by_con(jlong divisor) {
  return false;
}

RegMask Matcher::divI_proj_mask() {
  ShouldNotReachHere();
  return RegMask();
}

// Register for MODI projection of divmodI.
RegMask Matcher::modI_proj_mask() {
  ShouldNotReachHere();
  return RegMask();
}

// Register for DIVL projection of divmodL.
RegMask Matcher::divL_proj_mask() {
  ShouldNotReachHere();
  return RegMask();
}

// Register for MODL projection of divmodL.
RegMask Matcher::modL_proj_mask() {
  ShouldNotReachHere();
  return RegMask();
}

const RegMask Matcher::method_handle_invoke_SP_save_mask() {
  return FP_REG_mask();
}

bool size_fits_all_mem_uses(AddPNode* addp, int shift) {
  for (DUIterator_Fast imax, i = addp->fast_outs(imax); i < imax; i++) {
    Node* u = addp->fast_out(i);
    if (u->is_Mem()) {
      int opsize = u->as_Mem()->memory_size();
      assert(opsize > 0, "unexpected memory operand size");
      if (u->as_Mem()->memory_size() != (1<<shift)) {
        return false;
      }
    }
  }
  return true;
}

const bool Matcher::convi2l_type_required = false;

// Should the Matcher clone shifts on addressing modes, expecting them
// to be subsumed into complex addressing expressions or compute them
// into registers?
bool Matcher::clone_address_expressions(AddPNode* m, Matcher::MStack& mstack, VectorSet& address_visited) {
  if (clone_base_plus_offset_address(m, mstack, address_visited)) {
    return true;
  }

  Node *off = m->in(AddPNode::Offset);
  if (off->Opcode() == Op_LShiftL && off->in(2)->is_Con() &&
      size_fits_all_mem_uses(m, off->in(2)->get_int()) &&
      // Are there other uses besides address expressions?
      !is_visited(off)) {
    address_visited.set(off->_idx); // Flag as address_visited
    mstack.push(off->in(2), Visit);
    Node *conv = off->in(1);
    if (conv->Opcode() == Op_ConvI2L &&
        // Are there other uses besides address expressions?
        !is_visited(conv)) {
      address_visited.set(conv->_idx); // Flag as address_visited
      mstack.push(conv->in(1), Pre_Visit);
    } else {
      mstack.push(conv, Pre_Visit);
    }
    address_visited.test_set(m->_idx); // Flag as address_visited
    mstack.push(m->in(AddPNode::Address), Pre_Visit);
    mstack.push(m->in(AddPNode::Base), Pre_Visit);
    return true;
  } else if (off->Opcode() == Op_ConvI2L &&
             // Are there other uses besides address expressions?
             !is_visited(off)) {
    address_visited.test_set(m->_idx); // Flag as address_visited
    address_visited.set(off->_idx); // Flag as address_visited
    mstack.push(off->in(1), Pre_Visit);
    mstack.push(m->in(AddPNode::Address), Pre_Visit);
    mstack.push(m->in(AddPNode::Base), Pre_Visit);
    return true;
  }
  return false;
}

// Transform:
// (AddP base (AddP base address (LShiftL index con)) offset)
// into:
// (AddP base (AddP base offset) (LShiftL index con))
// to take full advantage of ARM's addressing modes
void Compile::reshape_address(AddPNode* addp) {
  Node *addr = addp->in(AddPNode::Address);
  if (addr->is_AddP() && addr->in(AddPNode::Base) == addp->in(AddPNode::Base)) {
    const AddPNode *addp2 = addr->as_AddP();
    if ((addp2->in(AddPNode::Offset)->Opcode() == Op_LShiftL &&
         addp2->in(AddPNode::Offset)->in(2)->is_Con() &&
         size_fits_all_mem_uses(addp, addp2->in(AddPNode::Offset)->in(2)->get_int())) ||
        addp2->in(AddPNode::Offset)->Opcode() == Op_ConvI2L) {

      // Any use that can't embed the address computation?
      for (DUIterator_Fast imax, i = addp->fast_outs(imax); i < imax; i++) {
        Node* u = addp->fast_out(i);
        if (!u->is_Mem() || u->is_LoadVector() || u->is_StoreVector() || u->Opcode() == Op_StoreCM) {
          return;
        }
      }
      
      Node* off = addp->in(AddPNode::Offset);
      Node* addr2 = addp2->in(AddPNode::Address);
      Node* base = addp->in(AddPNode::Base);
      
      Node* new_addr = NULL;
      // Check whether the graph already has the new AddP we need
      // before we create one (no GVN available here).
      for (DUIterator_Fast imax, i = addr2->fast_outs(imax); i < imax; i++) {
        Node* u = addr2->fast_out(i);
        if (u->is_AddP() &&
            u->in(AddPNode::Base) == base &&
            u->in(AddPNode::Address) == addr2 &&
            u->in(AddPNode::Offset) == off) {
          new_addr = u;
          break;
        }
      }
      
      if (new_addr == NULL) {
        new_addr = new AddPNode(base, addr2, off);
      }
      Node* new_off = addp2->in(AddPNode::Offset);
      addp->set_req(AddPNode::Address, new_addr);
      if (addr->outcnt() == 0) {
        addr->disconnect_inputs(NULL, this);
      }
      addp->set_req(AddPNode::Offset, new_off);
      if (off->outcnt() == 0) {
        off->disconnect_inputs(NULL, this);
      }
    }
  }
}

// helper for encoding java_to_runtime calls on sim
//
// this is needed to compute the extra arguments required when
// planting a call to the simulator blrt instruction. the TypeFunc
// can be queried to identify the counts for integral, and floating
// arguments and the return type

static void getCallInfo(const TypeFunc *tf, int &gpcnt, int &fpcnt, int &rtype)
{
  int gps = 0;
  int fps = 0;
  const TypeTuple *domain = tf->domain();
  int max = domain->cnt();
  for (int i = TypeFunc::Parms; i < max; i++) {
    const Type *t = domain->field_at(i);
    switch(t->basic_type()) {
    case T_FLOAT:
    case T_DOUBLE:
      fps++;
    default:
      gps++;
    }
  }
  gpcnt = gps;
  fpcnt = fps;
  BasicType rt = tf->return_type();
  switch (rt) {
  case T_VOID:
    rtype = MacroAssembler::ret_type_void;
    break;
  default:
    rtype = MacroAssembler::ret_type_integral;
    break;
  case T_FLOAT:
    rtype = MacroAssembler::ret_type_float;
    break;
  case T_DOUBLE:
    rtype = MacroAssembler::ret_type_double;
    break;
  }
}

#define MOV_VOLATILE(REG, BASE, INDEX, SCALE, DISP, SCRATCH, INSN)      \
  MacroAssembler _masm(&cbuf);                                          \
  {                                                                     \
    guarantee(INDEX == -1, "mode not permitted for volatile");          \
    guarantee(DISP == 0, "mode not permitted for volatile");            \
    guarantee(SCALE == 0, "mode not permitted for volatile");           \
    __ INSN(REG, as_Register(BASE));                                    \
  }

typedef void (MacroAssembler::* mem_insn)(Register Rt, const Address &adr);
typedef void (MacroAssembler::* mem_float_insn)(FloatRegister Rt, const Address &adr);
typedef void (MacroAssembler::* mem_vector_insn)(FloatRegister Rt,
                                  MacroAssembler::SIMD_RegVariant T, const Address &adr);

  // Used for all non-volatile memory accesses.  The use of
  // $mem->opcode() to discover whether this pattern uses sign-extended
  // offsets is something of a kludge.
  static void loadStore(MacroAssembler masm, mem_insn insn,
                         Register reg, int opcode,
                         Register base, int index, int size, int disp)
  {
    Address::extend scale;

    // Hooboy, this is fugly.  We need a way to communicate to the
    // encoder that the index needs to be sign extended, so we have to
    // enumerate all the cases.
    switch (opcode) {
    case INDINDEXSCALEDI2L:
    case INDINDEXSCALEDI2LN:
    case INDINDEXI2L:
    case INDINDEXI2LN:
      scale = Address::sxtw(size);
      break;
    default:
      scale = Address::lsl(size);
    }

    if (index == -1) {
      (masm.*insn)(reg, Address(base, disp));
    } else {
      assert(disp == 0, "unsupported address mode: disp = %d", disp);
      (masm.*insn)(reg, Address(base, as_Register(index), scale));
    }
  }

  static void loadStore(MacroAssembler masm, mem_float_insn insn,
                         FloatRegister reg, int opcode,
                         Register base, int index, int size, int disp)
  {
    Address::extend scale;

    switch (opcode) {
    case INDINDEXSCALEDI2L:
    case INDINDEXSCALEDI2LN:
      scale = Address::sxtw(size);
      break;
    default:
      scale = Address::lsl(size);
    }

     if (index == -1) {
      (masm.*insn)(reg, Address(base, disp));
    } else {
      assert(disp == 0, "unsupported address mode: disp = %d", disp);
      (masm.*insn)(reg, Address(base, as_Register(index), scale));
    }
  }

  static void loadStore(MacroAssembler masm, mem_vector_insn insn,
                         FloatRegister reg, MacroAssembler::SIMD_RegVariant T,
                         int opcode, Register base, int index, int size, int disp)
  {
    if (index == -1) {
      (masm.*insn)(reg, T, Address(base, disp));
    } else {
      assert(disp == 0, "unsupported address mode");
      (masm.*insn)(reg, T, Address(base, as_Register(index), Address::lsl(size)));
    }
  }

%}



//----------ENCODING BLOCK-----------------------------------------------------
// This block specifies the encoding classes used by the compiler to
// output byte streams.  Encoding classes are parameterized macros
// used by Machine Instruction Nodes in order to generate the bit
// encoding of the instruction.  Operands specify their base encoding
// interface with the interface keyword.  There are currently
// supported four interfaces, REG_INTER, CONST_INTER, MEMORY_INTER, &
// COND_INTER.  REG_INTER causes an operand to generate a function
// which returns its register number when queried.  CONST_INTER causes
// an operand to generate a function which returns the value of the
// constant when queried.  MEMORY_INTER causes an operand to generate
// four functions which return the Base Register, the Index Register,
// the Scale Value, and the Offset Value of the operand when queried.
// COND_INTER causes an operand to generate six functions which return
// the encoding code (ie - encoding bits for the instruction)
// associated with each basic boolean condition for a conditional
// instruction.
//
// Instructions specify two basic values for encoding.  Again, a
// function is available to check if the constant displacement is an
// oop. They use the ins_encode keyword to specify their encoding
// classes (which must be a sequence of enc_class names, and their
// parameters, specified in the encoding block), and they use the
// opcode keyword to specify, in order, their primary, secondary, and
// tertiary opcode.  Only the opcode sections which a particular
// instruction needs for encoding need to be specified.
encode %{
  // Build emit functions for each basic byte or larger field in the
  // intel encoding scheme (opcode, rm, sib, immediate), and call them
  // from C++ code in the enc_class source block.  Emit functions will
  // live in the main source block for now.  In future, we can
  // generalize this by adding a syntax that specifies the sizes of
  // fields in an order, so that the adlc can build the emit functions
  // automagically

  // catch all for unimplemented encodings
  enc_class enc_unimplemented %{
    MacroAssembler _masm(&cbuf);
    __ unimplemented("C2 catch all");
  %}

  // BEGIN Non-volatile memory access

  enc_class aarch64_enc_ldrsbw(iRegI dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrsbw, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrsb(iRegI dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrsb, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrb(iRegI dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrb, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrb(iRegL dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrb, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrshw(iRegI dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrshw, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrsh(iRegI dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrsh, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrh(iRegI dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrh, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrh(iRegL dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrh, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrw(iRegI dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrw, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrw(iRegL dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrw, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrsw(iRegL dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrsw, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldr(iRegL dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldr, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrs(vRegF dst, memory mem) %{
    FloatRegister dst_reg = as_FloatRegister($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrs, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrd(vRegD dst, memory mem) %{
    FloatRegister dst_reg = as_FloatRegister($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrd, dst_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrvS(vecD dst, memory mem) %{
    FloatRegister dst_reg = as_FloatRegister($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldr, dst_reg, MacroAssembler::S,
       $mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrvD(vecD dst, memory mem) %{
    FloatRegister dst_reg = as_FloatRegister($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldr, dst_reg, MacroAssembler::D,
       $mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_ldrvQ(vecX dst, memory mem) %{
    FloatRegister dst_reg = as_FloatRegister($dst$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldr, dst_reg, MacroAssembler::Q,
       $mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_strb(iRegI src, memory mem) %{
    Register src_reg = as_Register($src$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::strb, src_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_strb0(memory mem) %{
    MacroAssembler _masm(&cbuf);
    loadStore(_masm, &MacroAssembler::strb, zr, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_strb0_ordered(memory mem) %{
    MacroAssembler _masm(&cbuf);
    __ membar(Assembler::StoreStore);
    loadStore(_masm, &MacroAssembler::strb, zr, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_strh(iRegI src, memory mem) %{
    Register src_reg = as_Register($src$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::strh, src_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_strh0(memory mem) %{
    MacroAssembler _masm(&cbuf);
    loadStore(_masm, &MacroAssembler::strh, zr, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_strw(iRegI src, memory mem) %{
    Register src_reg = as_Register($src$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::strw, src_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_strw0(memory mem) %{
    MacroAssembler _masm(&cbuf);
    loadStore(_masm, &MacroAssembler::strw, zr, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_str(iRegL src, memory mem) %{
    Register src_reg = as_Register($src$$reg);
    // we sometimes get asked to store the stack pointer into the
    // current thread -- we cannot do that directly on AArch64
    if (src_reg == r31_sp) {
      MacroAssembler _masm(&cbuf);
      assert(as_Register($mem$$base) == rthread, "unexpected store for sp");
      __ mov(rscratch2, sp);
      src_reg = rscratch2;
    }
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::str, src_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_str0(memory mem) %{
    MacroAssembler _masm(&cbuf);
    loadStore(_masm, &MacroAssembler::str, zr, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_strs(vRegF src, memory mem) %{
    FloatRegister src_reg = as_FloatRegister($src$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::strs, src_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_strd(vRegD src, memory mem) %{
    FloatRegister src_reg = as_FloatRegister($src$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::strd, src_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_strvS(vecD src, memory mem) %{
    FloatRegister src_reg = as_FloatRegister($src$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::str, src_reg, MacroAssembler::S,
       $mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_strvD(vecD src, memory mem) %{
    FloatRegister src_reg = as_FloatRegister($src$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::str, src_reg, MacroAssembler::D,
       $mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  enc_class aarch64_enc_strvQ(vecX src, memory mem) %{
    FloatRegister src_reg = as_FloatRegister($src$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::str, src_reg, MacroAssembler::Q,
       $mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
  %}

  // END Non-volatile memory access

  // volatile loads and stores

  enc_class aarch64_enc_stlrb(iRegI src, memory mem) %{
    MOV_VOLATILE(as_Register($src$$reg), $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
                 rscratch1, stlrb);
  %}

  enc_class aarch64_enc_stlrh(iRegI src, memory mem) %{
    MOV_VOLATILE(as_Register($src$$reg), $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
                 rscratch1, stlrh);
  %}

  enc_class aarch64_enc_stlrw(iRegI src, memory mem) %{
    MOV_VOLATILE(as_Register($src$$reg), $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
                 rscratch1, stlrw);
  %}


  enc_class aarch64_enc_ldarsbw(iRegI dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    MOV_VOLATILE(dst_reg, $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldarb);
    __ sxtbw(dst_reg, dst_reg);
  %}

  enc_class aarch64_enc_ldarsb(iRegL dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    MOV_VOLATILE(dst_reg, $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldarb);
    __ sxtb(dst_reg, dst_reg);
  %}

  enc_class aarch64_enc_ldarbw(iRegI dst, memory mem) %{
    MOV_VOLATILE(as_Register($dst$$reg), $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldarb);
  %}

  enc_class aarch64_enc_ldarb(iRegL dst, memory mem) %{
    MOV_VOLATILE(as_Register($dst$$reg), $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldarb);
  %}

  enc_class aarch64_enc_ldarshw(iRegI dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    MOV_VOLATILE(dst_reg, $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldarh);
    __ sxthw(dst_reg, dst_reg);
  %}

  enc_class aarch64_enc_ldarsh(iRegL dst, memory mem) %{
    Register dst_reg = as_Register($dst$$reg);
    MOV_VOLATILE(dst_reg, $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldarh);
    __ sxth(dst_reg, dst_reg);
  %}

  enc_class aarch64_enc_ldarhw(iRegI dst, memory mem) %{
    MOV_VOLATILE(as_Register($dst$$reg), $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldarh);
  %}

  enc_class aarch64_enc_ldarh(iRegL dst, memory mem) %{
    MOV_VOLATILE(as_Register($dst$$reg), $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldarh);
  %}

  enc_class aarch64_enc_ldarw(iRegI dst, memory mem) %{
    MOV_VOLATILE(as_Register($dst$$reg), $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldarw);
  %}

  enc_class aarch64_enc_ldarw(iRegL dst, memory mem) %{
    MOV_VOLATILE(as_Register($dst$$reg), $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldarw);
  %}

  enc_class aarch64_enc_ldar(iRegL dst, memory mem) %{
    MOV_VOLATILE(as_Register($dst$$reg), $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldar);
  %}

  enc_class aarch64_enc_fldars(vRegF dst, memory mem) %{
    MOV_VOLATILE(rscratch1, $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldarw);
    __ fmovs(as_FloatRegister($dst$$reg), rscratch1);
  %}

  enc_class aarch64_enc_fldard(vRegD dst, memory mem) %{
    MOV_VOLATILE(rscratch1, $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
             rscratch1, ldar);
    __ fmovd(as_FloatRegister($dst$$reg), rscratch1);
  %}

  enc_class aarch64_enc_stlr(iRegL src, memory mem) %{
    Register src_reg = as_Register($src$$reg);
    // we sometimes get asked to store the stack pointer into the
    // current thread -- we cannot do that directly on AArch64
    if (src_reg == r31_sp) {
        MacroAssembler _masm(&cbuf);
      assert(as_Register($mem$$base) == rthread, "unexpected store for sp");
      __ mov(rscratch2, sp);
      src_reg = rscratch2;
    }
    MOV_VOLATILE(src_reg, $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
                 rscratch1, stlr);
  %}

  enc_class aarch64_enc_fstlrs(vRegF src, memory mem) %{
    {
      MacroAssembler _masm(&cbuf);
      FloatRegister src_reg = as_FloatRegister($src$$reg);
      __ fmovs(rscratch2, src_reg);
    }
    MOV_VOLATILE(rscratch2, $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
                 rscratch1, stlrw);
  %}

  enc_class aarch64_enc_fstlrd(vRegD src, memory mem) %{
    {
      MacroAssembler _masm(&cbuf);
      FloatRegister src_reg = as_FloatRegister($src$$reg);
      __ fmovd(rscratch2, src_reg);
    }
    MOV_VOLATILE(rscratch2, $mem$$base, $mem$$index, $mem$$scale, $mem$$disp,
                 rscratch1, stlr);
  %}

  // synchronized read/update encodings

  enc_class aarch64_enc_ldaxr(iRegL dst, memory mem) %{
    MacroAssembler _masm(&cbuf);
    Register dst_reg = as_Register($dst$$reg);
    Register base = as_Register($mem$$base);
    int index = $mem$$index;
    int scale = $mem$$scale;
    int disp = $mem$$disp;
    if (index == -1) {
       if (disp != 0) {
        __ lea(rscratch1, Address(base, disp));
        __ ldaxr(dst_reg, rscratch1);
      } else {
        // TODO
        // should we ever get anything other than this case?
        __ ldaxr(dst_reg, base);
      }
    } else {
      Register index_reg = as_Register(index);
      if (disp == 0) {
        __ lea(rscratch1, Address(base, index_reg, Address::lsl(scale)));
        __ ldaxr(dst_reg, rscratch1);
      } else {
        __ lea(rscratch1, Address(base, disp));
        __ lea(rscratch1, Address(rscratch1, index_reg, Address::lsl(scale)));
        __ ldaxr(dst_reg, rscratch1);
      }
    }
  %}

  enc_class aarch64_enc_stlxr(iRegLNoSp src, memory mem) %{
    MacroAssembler _masm(&cbuf);
    Register src_reg = as_Register($src$$reg);
    Register base = as_Register($mem$$base);
    int index = $mem$$index;
    int scale = $mem$$scale;
    int disp = $mem$$disp;
    if (index == -1) {
       if (disp != 0) {
        __ lea(rscratch2, Address(base, disp));
        __ stlxr(rscratch1, src_reg, rscratch2);
      } else {
        // TODO
        // should we ever get anything other than this case?
        __ stlxr(rscratch1, src_reg, base);
      }
    } else {
      Register index_reg = as_Register(index);
      if (disp == 0) {
        __ lea(rscratch2, Address(base, index_reg, Address::lsl(scale)));
        __ stlxr(rscratch1, src_reg, rscratch2);
      } else {
        __ lea(rscratch2, Address(base, disp));
        __ lea(rscratch2, Address(rscratch2, index_reg, Address::lsl(scale)));
        __ stlxr(rscratch1, src_reg, rscratch2);
      }
    }
    __ cmpw(rscratch1, zr);
  %}

  enc_class aarch64_enc_cmpxchg(memory mem, iRegLNoSp oldval, iRegLNoSp newval) %{
    MacroAssembler _masm(&cbuf);
    guarantee($mem$$index == -1 && $mem$$disp == 0, "impossible encoding");
    __ cmpxchg($mem$$base$$Register, $oldval$$Register, $newval$$Register,
               Assembler::xword, /*acquire*/ false, /*release*/ true,
               /*weak*/ false, noreg);
  %}

  enc_class aarch64_enc_cmpxchgw(memory mem, iRegINoSp oldval, iRegINoSp newval) %{
    MacroAssembler _masm(&cbuf);
    guarantee($mem$$index == -1 && $mem$$disp == 0, "impossible encoding");
    __ cmpxchg($mem$$base$$Register, $oldval$$Register, $newval$$Register,
               Assembler::word, /*acquire*/ false, /*release*/ true,
               /*weak*/ false, noreg);
  %}


  // The only difference between aarch64_enc_cmpxchg and
  // aarch64_enc_cmpxchg_acq is that we use load-acquire in the
  // CompareAndSwap sequence to serve as a barrier on acquiring a
  // lock.
  enc_class aarch64_enc_cmpxchg_acq(memory mem, iRegLNoSp oldval, iRegLNoSp newval) %{
    MacroAssembler _masm(&cbuf);
    guarantee($mem$$index == -1 && $mem$$disp == 0, "impossible encoding");
    __ cmpxchg($mem$$base$$Register, $oldval$$Register, $newval$$Register,
               Assembler::xword, /*acquire*/ true, /*release*/ true,
               /*weak*/ false, noreg);
  %}

  enc_class aarch64_enc_cmpxchgw_acq(memory mem, iRegINoSp oldval, iRegINoSp newval) %{
    MacroAssembler _masm(&cbuf);
    guarantee($mem$$index == -1 && $mem$$disp == 0, "impossible encoding");
    __ cmpxchg($mem$$base$$Register, $oldval$$Register, $newval$$Register,
               Assembler::word, /*acquire*/ true, /*release*/ true,
               /*weak*/ false, noreg);
  %}


  // auxiliary used for CompareAndSwapX to set result register
  enc_class aarch64_enc_cset_eq(iRegINoSp res) %{
    MacroAssembler _masm(&cbuf);
    Register res_reg = as_Register($res$$reg);
    __ cset(res_reg, Assembler::EQ);
  %}

  // prefetch encodings

  enc_class aarch64_enc_prefetchw(memory mem) %{
    MacroAssembler _masm(&cbuf);
    Register base = as_Register($mem$$base);
    int index = $mem$$index;
    int scale = $mem$$scale;
    int disp = $mem$$disp;
    if (index == -1) {
      __ prfm(Address(base, disp), PSTL1KEEP);
    } else {
      Register index_reg = as_Register(index);
      if (disp == 0) {
        __ prfm(Address(base, index_reg, Address::lsl(scale)), PSTL1KEEP);
      } else {
        __ lea(rscratch1, Address(base, disp));
	__ prfm(Address(rscratch1, index_reg, Address::lsl(scale)), PSTL1KEEP);
      }
    }
  %}

  /// mov envcodings

  enc_class aarch64_enc_movw_imm(iRegI dst, immI src) %{
    MacroAssembler _masm(&cbuf);
    u_int32_t con = (u_int32_t)$src$$constant;
    Register dst_reg = as_Register($dst$$reg);
    if (con == 0) {
      __ movw(dst_reg, zr);
    } else {
      __ movw(dst_reg, con);
    }
  %}

  enc_class aarch64_enc_mov_imm(iRegL dst, immL src) %{
    MacroAssembler _masm(&cbuf);
    Register dst_reg = as_Register($dst$$reg);
    u_int64_t con = (u_int64_t)$src$$constant;
    if (con == 0) {
      __ mov(dst_reg, zr);
    } else {
      __ mov(dst_reg, con);
    }
  %}

  enc_class aarch64_enc_mov_p(iRegP dst, immP src) %{
    MacroAssembler _masm(&cbuf);
    Register dst_reg = as_Register($dst$$reg);
    address con = (address)$src$$constant;
    if (con == NULL || con == (address)1) {
      ShouldNotReachHere();
    } else {
      relocInfo::relocType rtype = $src->constant_reloc();
      if (rtype == relocInfo::oop_type) {
        __ movoop(dst_reg, (jobject)con, /*immediate*/true);
      } else if (rtype == relocInfo::metadata_type) {
        __ mov_metadata(dst_reg, (Metadata*)con);
      } else {
        assert(rtype == relocInfo::none, "unexpected reloc type");
        if (con < (address)(uintptr_t)os::vm_page_size()) {
          __ mov(dst_reg, con);
        } else {
          unsigned long offset;
          __ adrp(dst_reg, con, offset);
          __ add(dst_reg, dst_reg, offset);
        }
      }
    }
  %}

  enc_class aarch64_enc_mov_p0(iRegP dst, immP0 src) %{
    MacroAssembler _masm(&cbuf);
    Register dst_reg = as_Register($dst$$reg);
    __ mov(dst_reg, zr);
  %}

  enc_class aarch64_enc_mov_p1(iRegP dst, immP_1 src) %{
    MacroAssembler _masm(&cbuf);
    Register dst_reg = as_Register($dst$$reg);
    __ mov(dst_reg, (u_int64_t)1);
  %}

  enc_class aarch64_enc_mov_poll_page(iRegP dst, immPollPage src) %{
    MacroAssembler _masm(&cbuf);
    address page = (address)$src$$constant;
    Register dst_reg = as_Register($dst$$reg);
    unsigned long off;
    __ adrp(dst_reg, Address(page, relocInfo::poll_type), off);
    assert(off == 0, "assumed offset == 0");
  %}

  enc_class aarch64_enc_mov_byte_map_base(iRegP dst, immByteMapBase src) %{
    MacroAssembler _masm(&cbuf);
    __ load_byte_map_base($dst$$Register);
  %}

  enc_class aarch64_enc_mov_n(iRegN dst, immN src) %{
    MacroAssembler _masm(&cbuf);
    Register dst_reg = as_Register($dst$$reg);
    address con = (address)$src$$constant;
    if (con == NULL) {
      ShouldNotReachHere();
    } else {
      relocInfo::relocType rtype = $src->constant_reloc();
      assert(rtype == relocInfo::oop_type, "unexpected reloc type");
      __ set_narrow_oop(dst_reg, (jobject)con);
    }
  %}

  enc_class aarch64_enc_mov_n0(iRegN dst, immN0 src) %{
    MacroAssembler _masm(&cbuf);
    Register dst_reg = as_Register($dst$$reg);
    __ mov(dst_reg, zr);
  %}

  enc_class aarch64_enc_mov_nk(iRegN dst, immNKlass src) %{
    MacroAssembler _masm(&cbuf);
    Register dst_reg = as_Register($dst$$reg);
    address con = (address)$src$$constant;
    if (con == NULL) {
      ShouldNotReachHere();
    } else {
      relocInfo::relocType rtype = $src->constant_reloc();
      assert(rtype == relocInfo::metadata_type, "unexpected reloc type");
      __ set_narrow_klass(dst_reg, (Klass *)con);
    }
  %}

  // arithmetic encodings

  enc_class aarch64_enc_addsubw_imm(iRegI dst, iRegI src1, immIAddSub src2) %{
    MacroAssembler _masm(&cbuf);
    Register dst_reg = as_Register($dst$$reg);
    Register src_reg = as_Register($src1$$reg);
    int32_t con = (int32_t)$src2$$constant;
    // add has primary == 0, subtract has primary == 1
    if ($primary) { con = -con; }
    if (con < 0) {
      __ subw(dst_reg, src_reg, -con);
    } else {
      __ addw(dst_reg, src_reg, con);
    }
  %}

  enc_class aarch64_enc_addsub_imm(iRegL dst, iRegL src1, immLAddSub src2) %{
    MacroAssembler _masm(&cbuf);
    Register dst_reg = as_Register($dst$$reg);
    Register src_reg = as_Register($src1$$reg);
    int32_t con = (int32_t)$src2$$constant;
    // add has primary == 0, subtract has primary == 1
    if ($primary) { con = -con; }
    if (con < 0) {
      __ sub(dst_reg, src_reg, -con);
    } else {
      __ add(dst_reg, src_reg, con);
    }
  %}

  enc_class aarch64_enc_divw(iRegI dst, iRegI src1, iRegI src2) %{
    MacroAssembler _masm(&cbuf);
   Register dst_reg = as_Register($dst$$reg);
   Register src1_reg = as_Register($src1$$reg);
   Register src2_reg = as_Register($src2$$reg);
    __ corrected_idivl(dst_reg, src1_reg, src2_reg, false, rscratch1);
  %}

  enc_class aarch64_enc_div(iRegI dst, iRegI src1, iRegI src2) %{
    MacroAssembler _masm(&cbuf);
   Register dst_reg = as_Register($dst$$reg);
   Register src1_reg = as_Register($src1$$reg);
   Register src2_reg = as_Register($src2$$reg);
    __ corrected_idivq(dst_reg, src1_reg, src2_reg, false, rscratch1);
  %}

  enc_class aarch64_enc_modw(iRegI dst, iRegI src1, iRegI src2) %{
    MacroAssembler _masm(&cbuf);
   Register dst_reg = as_Register($dst$$reg);
   Register src1_reg = as_Register($src1$$reg);
   Register src2_reg = as_Register($src2$$reg);
    __ corrected_idivl(dst_reg, src1_reg, src2_reg, true, rscratch1);
  %}

  enc_class aarch64_enc_mod(iRegI dst, iRegI src1, iRegI src2) %{
    MacroAssembler _masm(&cbuf);
   Register dst_reg = as_Register($dst$$reg);
   Register src1_reg = as_Register($src1$$reg);
   Register src2_reg = as_Register($src2$$reg);
    __ corrected_idivq(dst_reg, src1_reg, src2_reg, true, rscratch1);
  %}

  // compare instruction encodings

  enc_class aarch64_enc_cmpw(iRegI src1, iRegI src2) %{
    MacroAssembler _masm(&cbuf);
    Register reg1 = as_Register($src1$$reg);
    Register reg2 = as_Register($src2$$reg);
    __ cmpw(reg1, reg2);
  %}

  enc_class aarch64_enc_cmpw_imm_addsub(iRegI src1, immIAddSub src2) %{
    MacroAssembler _masm(&cbuf);
    Register reg = as_Register($src1$$reg);
    int32_t val = $src2$$constant;
    if (val >= 0) {
      __ subsw(zr, reg, val);
    } else {
      __ addsw(zr, reg, -val);
    }
  %}

  enc_class aarch64_enc_cmpw_imm(iRegI src1, immI src2) %{
    MacroAssembler _masm(&cbuf);
    Register reg1 = as_Register($src1$$reg);
    u_int32_t val = (u_int32_t)$src2$$constant;
    __ movw(rscratch1, val);
    __ cmpw(reg1, rscratch1);
  %}

  enc_class aarch64_enc_cmp(iRegL src1, iRegL src2) %{
    MacroAssembler _masm(&cbuf);
    Register reg1 = as_Register($src1$$reg);
    Register reg2 = as_Register($src2$$reg);
    __ cmp(reg1, reg2);
  %}

  enc_class aarch64_enc_cmp_imm_addsub(iRegL src1, immL12 src2) %{
    MacroAssembler _masm(&cbuf);
    Register reg = as_Register($src1$$reg);
    int64_t val = $src2$$constant;
    if (val >= 0) {
      __ subs(zr, reg, val);
    } else if (val != -val) {
      __ adds(zr, reg, -val);
    } else {
    // aargh, Long.MIN_VALUE is a special case
      __ orr(rscratch1, zr, (u_int64_t)val);
      __ subs(zr, reg, rscratch1);
    }
  %}

  enc_class aarch64_enc_cmp_imm(iRegL src1, immL src2) %{
    MacroAssembler _masm(&cbuf);
    Register reg1 = as_Register($src1$$reg);
    u_int64_t val = (u_int64_t)$src2$$constant;
    __ mov(rscratch1, val);
    __ cmp(reg1, rscratch1);
  %}

  enc_class aarch64_enc_cmpp(iRegP src1, iRegP src2) %{
    MacroAssembler _masm(&cbuf);
    Register reg1 = as_Register($src1$$reg);
    Register reg2 = as_Register($src2$$reg);
    __ cmp(reg1, reg2);
  %}

  enc_class aarch64_enc_cmpn(iRegN src1, iRegN src2) %{
    MacroAssembler _masm(&cbuf);
    Register reg1 = as_Register($src1$$reg);
    Register reg2 = as_Register($src2$$reg);
    __ cmpw(reg1, reg2);
  %}

  enc_class aarch64_enc_testp(iRegP src) %{
    MacroAssembler _masm(&cbuf);
    Register reg = as_Register($src$$reg);
    __ cmp(reg, zr);
  %}

  enc_class aarch64_enc_testn(iRegN src) %{
    MacroAssembler _masm(&cbuf);
    Register reg = as_Register($src$$reg);
    __ cmpw(reg, zr);
  %}

  enc_class aarch64_enc_b(label lbl) %{
    MacroAssembler _masm(&cbuf);
    Label *L = $lbl$$label;
    __ b(*L);
  %}

  enc_class aarch64_enc_br_con(cmpOp cmp, label lbl) %{
    MacroAssembler _masm(&cbuf);
    Label *L = $lbl$$label;
    __ br ((Assembler::Condition)$cmp$$cmpcode, *L);
  %}

  enc_class aarch64_enc_br_conU(cmpOpU cmp, label lbl) %{
    MacroAssembler _masm(&cbuf);
    Label *L = $lbl$$label;
    __ br ((Assembler::Condition)$cmp$$cmpcode, *L);
  %}

  enc_class aarch64_enc_partial_subtype_check(iRegP sub, iRegP super, iRegP temp, iRegP result)
  %{
     Register sub_reg = as_Register($sub$$reg);
     Register super_reg = as_Register($super$$reg);
     Register temp_reg = as_Register($temp$$reg);
     Register result_reg = as_Register($result$$reg);

     Label miss;
     MacroAssembler _masm(&cbuf);
     __ check_klass_subtype_slow_path(sub_reg, super_reg, temp_reg, result_reg,
                                     NULL, &miss,
                                     /*set_cond_codes:*/ true);
     if ($primary) {
       __ mov(result_reg, zr);
     }
     __ bind(miss);
  %}

  enc_class aarch64_enc_java_static_call(method meth) %{
    MacroAssembler _masm(&cbuf);

    address addr = (address)$meth$$method;
    address call;
    if (!_method) {
      // A call to a runtime wrapper, e.g. new, new_typeArray_Java, uncommon_trap.
      call = __ trampoline_call(Address(addr, relocInfo::runtime_call_type), &cbuf);
    } else {
      int method_index = resolved_method_index(cbuf);
      RelocationHolder rspec = _optimized_virtual ? opt_virtual_call_Relocation::spec(method_index)
                                                  : static_call_Relocation::spec(method_index);
      call = __ trampoline_call(Address(addr, rspec), &cbuf);

      // Emit stub for static call
      address stub = CompiledStaticCall::emit_to_interp_stub(cbuf);
      if (stub == NULL) {
        ciEnv::current()->record_failure("CodeCache is full");
        return;
      }
    }
    if (call == NULL) {
      ciEnv::current()->record_failure("CodeCache is full");
      return;
    }
  %}

  enc_class aarch64_enc_java_dynamic_call(method meth) %{
    MacroAssembler _masm(&cbuf);
    int method_index = resolved_method_index(cbuf);
    address call = __ ic_call((address)$meth$$method, method_index);
    if (call == NULL) {
      ciEnv::current()->record_failure("CodeCache is full");
      return;
    }
  %}

  enc_class aarch64_enc_call_epilog() %{
    MacroAssembler _masm(&cbuf);
    if (VerifyStackAtCalls) {
      // Check that stack depth is unchanged: find majik cookie on stack
      __ call_Unimplemented();
    }
  %}

  enc_class aarch64_enc_java_to_runtime(method meth) %{
    MacroAssembler _masm(&cbuf);

    // some calls to generated routines (arraycopy code) are scheduled
    // by C2 as runtime calls. if so we can call them using a br (they
    // will be in a reachable segment) otherwise we have to use a blrt
    // which loads the absolute address into a register.
    address entry = (address)$meth$$method;
    CodeBlob *cb = CodeCache::find_blob(entry);
    if (cb) {
      address call = __ trampoline_call(Address(entry, relocInfo::runtime_call_type));
      if (call == NULL) {
        ciEnv::current()->record_failure("CodeCache is full");
        return;
      }
    } else {
      int gpcnt;
      int fpcnt;
      int rtype;
      getCallInfo(tf(), gpcnt, fpcnt, rtype);
      Label retaddr;
      __ adr(rscratch2, retaddr);
      __ lea(rscratch1, RuntimeAddress(entry));
      // Leave a breadcrumb for JavaFrameAnchor::capture_last_Java_pc()
      __ stp(zr, rscratch2, Address(__ pre(sp, -2 * wordSize)));
      __ blrt(rscratch1, gpcnt, fpcnt, rtype);
      __ bind(retaddr);
      __ add(sp, sp, 2 * wordSize);
    }
  %}

  enc_class aarch64_enc_rethrow() %{
    MacroAssembler _masm(&cbuf);
    __ far_jump(RuntimeAddress(OptoRuntime::rethrow_stub()));
  %}

  enc_class aarch64_enc_ret() %{
    MacroAssembler _masm(&cbuf);
    __ ret(lr);
  %}

  enc_class aarch64_enc_tail_call(iRegP jump_target) %{
    MacroAssembler _masm(&cbuf);
    Register target_reg = as_Register($jump_target$$reg);
    __ br(target_reg);
  %}

  enc_class aarch64_enc_tail_jmp(iRegP jump_target) %{
    MacroAssembler _masm(&cbuf);
    Register target_reg = as_Register($jump_target$$reg);
    // exception oop should be in r0
    // ret addr has been popped into lr
    // callee expects it in r3
    __ mov(r3, lr);
    __ br(target_reg);
  %}

  enc_class aarch64_enc_fast_lock(iRegP object, iRegP box, iRegP tmp, iRegP tmp2) %{
    MacroAssembler _masm(&cbuf);
    Register oop = as_Register($object$$reg);
    Register box = as_Register($box$$reg);
    Register disp_hdr = as_Register($tmp$$reg);
    Register tmp = as_Register($tmp2$$reg);
    Label cont;
    Label object_has_monitor;
    Label cas_failed;

    assert_different_registers(oop, box, tmp, disp_hdr);

    // Load markOop from object into displaced_header.
    __ ldr(disp_hdr, Address(oop, oopDesc::mark_offset_in_bytes()));

    // Always do locking in runtime.
    if (EmitSync & 0x01) {
      __ cmp(oop, zr);
      return;
    }

    if (UseBiasedLocking && !UseOptoBiasInlining) {
      __ biased_locking_enter(box, oop, disp_hdr, tmp, true, cont);
    }

    // Handle existing monitor
    if ((EmitSync & 0x02) == 0) {
      // we can use AArch64's bit test and branch here but
      // markoopDesc does not define a bit index just the bit value
      // so assert in case the bit pos changes
#     define __monitor_value_log2 1
      assert(markOopDesc::monitor_value == (1 << __monitor_value_log2), "incorrect bit position");
      __ tbnz(disp_hdr, __monitor_value_log2, object_has_monitor);
#     undef __monitor_value_log2
    }

    // Set displaced_header to be (markOop of object | UNLOCK_VALUE).
    __ orr(disp_hdr, disp_hdr, markOopDesc::unlocked_value);

    // Load Compare Value application register.

    // Initialize the box. (Must happen before we update the object mark!)
    __ str(disp_hdr, Address(box, BasicLock::displaced_header_offset_in_bytes()));

    // Compare object markOop with mark and if equal exchange scratch1
    // with object markOop.
    if (UseLSE) {
      __ mov(tmp, disp_hdr);
      __ casal(Assembler::xword, tmp, box, oop);
      __ cmp(tmp, disp_hdr);
      __ br(Assembler::EQ, cont);
    } else {
      Label retry_load;
      if ((VM_Version::features() & VM_Version::CPU_STXR_PREFETCH))
        __ prfm(Address(oop), PSTL1STRM);
      __ bind(retry_load);
      __ ldaxr(tmp, oop);
      __ cmp(tmp, disp_hdr);
      __ br(Assembler::NE, cas_failed);
      // use stlxr to ensure update is immediately visible
      __ stlxr(tmp, box, oop);
      __ cbzw(tmp, cont);
      __ b(retry_load);
    }

    // Formerly:
    // __ cmpxchgptr(/*oldv=*/disp_hdr,
    //               /*newv=*/box,
    //               /*addr=*/oop,
    //               /*tmp=*/tmp,
    //               cont,
    //               /*fail*/NULL);

    assert(oopDesc::mark_offset_in_bytes() == 0, "offset of _mark is not 0");

    // If the compare-and-exchange succeeded, then we found an unlocked
    // object, will have now locked it will continue at label cont

    __ bind(cas_failed);
    // We did not see an unlocked object so try the fast recursive case.

    // Check if the owner is self by comparing the value in the
    // markOop of object (disp_hdr) with the stack pointer.
    __ mov(rscratch1, sp);
    __ sub(disp_hdr, disp_hdr, rscratch1);
    __ mov(tmp, (address) (~(os::vm_page_size()-1) | markOopDesc::lock_mask_in_place));
    // If condition is true we are cont and hence we can store 0 as the
    // displaced header in the box, which indicates that it is a recursive lock.
    __ ands(tmp/*==0?*/, disp_hdr, tmp);
    __ str(tmp/*==0, perhaps*/, Address(box, BasicLock::displaced_header_offset_in_bytes()));

    // Handle existing monitor.
    if ((EmitSync & 0x02) == 0) {
      __ b(cont);

      __ bind(object_has_monitor);
      // The object's monitor m is unlocked iff m->owner == NULL,
      // otherwise m->owner may contain a thread or a stack address.
      //
      // Try to CAS m->owner from NULL to current thread.
      __ add(tmp, disp_hdr, (ObjectMonitor::owner_offset_in_bytes()-markOopDesc::monitor_value));
      __ mov(disp_hdr, zr);

      if (UseLSE) {
        __ mov(rscratch1, disp_hdr);
        __ casal(Assembler::xword, rscratch1, rthread, tmp);
        __ cmp(rscratch1, disp_hdr);
      } else {
        Label retry_load, fail;
        if ((VM_Version::features() & VM_Version::CPU_STXR_PREFETCH))
          __ prfm(Address(tmp), PSTL1STRM);
        __ bind(retry_load);
        __ ldaxr(rscratch1, tmp);
        __ cmp(disp_hdr, rscratch1);
        __ br(Assembler::NE, fail);
        // use stlxr to ensure update is immediately visible
        __ stlxr(rscratch1, rthread, tmp);
        __ cbnzw(rscratch1, retry_load);
        __ bind(fail);
      }

      // Label next;
      // __ cmpxchgptr(/*oldv=*/disp_hdr,
      //               /*newv=*/rthread,
      //               /*addr=*/tmp,
      //               /*tmp=*/rscratch1,
      //               /*succeed*/next,
      //               /*fail*/NULL);
      // __ bind(next);

      // store a non-null value into the box.
      __ str(box, Address(box, BasicLock::displaced_header_offset_in_bytes()));

      // PPC port checks the following invariants
      // #ifdef ASSERT
      // bne(flag, cont);
      // We have acquired the monitor, check some invariants.
      // addw(/*monitor=*/tmp, tmp, -ObjectMonitor::owner_offset_in_bytes());
      // Invariant 1: _recursions should be 0.
      // assert(ObjectMonitor::recursions_size_in_bytes() == 8, "unexpected size");
      // assert_mem8_is_zero(ObjectMonitor::recursions_offset_in_bytes(), tmp,
      //                        "monitor->_recursions should be 0", -1);
      // Invariant 2: OwnerIsThread shouldn't be 0.
      // assert(ObjectMonitor::OwnerIsThread_size_in_bytes() == 4, "unexpected size");
      //assert_mem4_isnot_zero(ObjectMonitor::OwnerIsThread_offset_in_bytes(), tmp,
      //                           "monitor->OwnerIsThread shouldn't be 0", -1);
      // #endif
    }

    __ bind(cont);
    // flag == EQ indicates success
    // flag == NE indicates failure

  %}

  // TODO
  // reimplement this with custom cmpxchgptr code
  // which avoids some of the unnecessary branching
  enc_class aarch64_enc_fast_unlock(iRegP object, iRegP box, iRegP tmp, iRegP tmp2) %{
    MacroAssembler _masm(&cbuf);
    Register oop = as_Register($object$$reg);
    Register box = as_Register($box$$reg);
    Register disp_hdr = as_Register($tmp$$reg);
    Register tmp = as_Register($tmp2$$reg);
    Label cont;
    Label object_has_monitor;
    Label cas_failed;

    assert_different_registers(oop, box, tmp, disp_hdr);

    // Always do locking in runtime.
    if (EmitSync & 0x01) {
      __ cmp(oop, zr); // Oop can't be 0 here => always false.
      return;
    }

    if (UseBiasedLocking && !UseOptoBiasInlining) {
      __ biased_locking_exit(oop, tmp, cont);
    }

    // Find the lock address and load the displaced header from the stack.
    __ ldr(disp_hdr, Address(box, BasicLock::displaced_header_offset_in_bytes()));

    // If the displaced header is 0, we have a recursive unlock.
    __ cmp(disp_hdr, zr);
    __ br(Assembler::EQ, cont);


    // Handle existing monitor.
    if ((EmitSync & 0x02) == 0) {
      __ ldr(tmp, Address(oop, oopDesc::mark_offset_in_bytes()));
      __ tbnz(disp_hdr, exact_log2(markOopDesc::monitor_value), object_has_monitor);
    }

    // Check if it is still a light weight lock, this is is true if we
    // see the stack address of the basicLock in the markOop of the
    // object.

      if (UseLSE) {
        __ mov(tmp, box);
        __ casl(Assembler::xword, tmp, disp_hdr, oop);
        __ cmp(tmp, box);
      } else {
        Label retry_load;
        if ((VM_Version::features() & VM_Version::CPU_STXR_PREFETCH))
          __ prfm(Address(oop), PSTL1STRM);
        __ bind(retry_load);
        __ ldxr(tmp, oop);
        __ cmp(box, tmp);
        __ br(Assembler::NE, cas_failed);
        // use stlxr to ensure update is immediately visible
        __ stlxr(tmp, disp_hdr, oop);
        __ cbzw(tmp, cont);
        __ b(retry_load);
      }

    // __ cmpxchgptr(/*compare_value=*/box,
    //               /*exchange_value=*/disp_hdr,
    //               /*where=*/oop,
    //               /*result=*/tmp,
    //               cont,
    //               /*cas_failed*/NULL);
    assert(oopDesc::mark_offset_in_bytes() == 0, "offset of _mark is not 0");

    __ bind(cas_failed);

    // Handle existing monitor.
    if ((EmitSync & 0x02) == 0) {
      __ b(cont);

      __ bind(object_has_monitor);
      __ add(tmp, tmp, -markOopDesc::monitor_value); // monitor
      __ ldr(rscratch1, Address(tmp, ObjectMonitor::owner_offset_in_bytes()));
      __ ldr(disp_hdr, Address(tmp, ObjectMonitor::recursions_offset_in_bytes()));
      __ eor(rscratch1, rscratch1, rthread); // Will be 0 if we are the owner.
      __ orr(rscratch1, rscratch1, disp_hdr); // Will be 0 if there are 0 recursions
      __ cmp(rscratch1, zr);
      __ br(Assembler::NE, cont);

      __ ldr(rscratch1, Address(tmp, ObjectMonitor::EntryList_offset_in_bytes()));
      __ ldr(disp_hdr, Address(tmp, ObjectMonitor::cxq_offset_in_bytes()));
      __ orr(rscratch1, rscratch1, disp_hdr); // Will be 0 if both are 0.
      __ cmp(rscratch1, zr);
      __ cbnz(rscratch1, cont);
      // need a release store here
      __ lea(tmp, Address(tmp, ObjectMonitor::owner_offset_in_bytes()));
      __ stlr(rscratch1, tmp); // rscratch1 is zero
    }

    __ bind(cont);
    // flag == EQ indicates success
    // flag == NE indicates failure
  %}

%}

//----------FRAME--------------------------------------------------------------
// Definition of frame structure and management information.
//
//  S T A C K   L A Y O U T    Allocators stack-slot number
//                             |   (to get allocators register number
//  G  Owned by    |        |  v    add OptoReg::stack0())
//  r   CALLER     |        |
//  o     |        +--------+      pad to even-align allocators stack-slot
//  w     V        |  pad0  |        numbers; owned by CALLER
//  t   -----------+--------+----> Matcher::_in_arg_limit, unaligned
//  h     ^        |   in   |  5
//        |        |  args  |  4   Holes in incoming args owned by SELF
//  |     |        |        |  3
//  |     |        +--------+
//  V     |        | old out|      Empty on Intel, window on Sparc
//        |    old |preserve|      Must be even aligned.
//        |     SP-+--------+----> Matcher::_old_SP, even aligned
//        |        |   in   |  3   area for Intel ret address
//     Owned by    |preserve|      Empty on Sparc.
//       SELF      +--------+
//        |        |  pad2  |  2   pad to align old SP
//        |        +--------+  1
//        |        | locks  |  0
//        |        +--------+----> OptoReg::stack0(), even aligned
//        |        |  pad1  | 11   pad to align new SP
//        |        +--------+
//        |        |        | 10
//        |        | spills |  9   spills
//        V        |        |  8   (pad0 slot for callee)
//      -----------+--------+----> Matcher::_out_arg_limit, unaligned
//        ^        |  out   |  7
//        |        |  args  |  6   Holes in outgoing args owned by CALLEE
//     Owned by    +--------+
//      CALLEE     | new out|  6   Empty on Intel, window on Sparc
//        |    new |preserve|      Must be even-aligned.
//        |     SP-+--------+----> Matcher::_new_SP, even aligned
//        |        |        |
//
// Note 1: Only region 8-11 is determined by the allocator.  Region 0-5 is
//         known from SELF's arguments and the Java calling convention.
//         Region 6-7 is determined per call site.
// Note 2: If the calling convention leaves holes in the incoming argument
//         area, those holes are owned by SELF.  Holes in the outgoing area
//         are owned by the CALLEE.  Holes should not be nessecary in the
//         incoming area, as the Java calling convention is completely under
//         the control of the AD file.  Doubles can be sorted and packed to
//         avoid holes.  Holes in the outgoing arguments may be nessecary for
//         varargs C calling conventions.
// Note 3: Region 0-3 is even aligned, with pad2 as needed.  Region 3-5 is
//         even aligned with pad0 as needed.
//         Region 6 is even aligned.  Region 6-7 is NOT even aligned;
//           (the latter is true on Intel but is it false on AArch64?)
//         region 6-11 is even aligned; it may be padded out more so that
//         the region from SP to FP meets the minimum stack alignment.
// Note 4: For I2C adapters, the incoming FP may not meet the minimum stack
//         alignment.  Region 11, pad1, may be dynamically extended so that
//         SP meets the minimum alignment.

frame %{
  // What direction does stack grow in (assumed to be same for C & Java)
  stack_direction(TOWARDS_LOW);

  // These three registers define part of the calling convention
  // between compiled code and the interpreter.

  // Inline Cache Register or methodOop for I2C.
  inline_cache_reg(R12);

  // Method Oop Register when calling interpreter.
  interpreter_method_oop_reg(R12);

  // Number of stack slots consumed by locking an object
  sync_stack_slots(2);

  // Compiled code's Frame Pointer
  frame_pointer(R31);

  // Interpreter stores its frame pointer in a register which is
  // stored to the stack by I2CAdaptors.
  // I2CAdaptors convert from interpreted java to compiled java.
  interpreter_frame_pointer(R29);

  // Stack alignment requirement
  stack_alignment(StackAlignmentInBytes); // Alignment size in bytes (128-bit -> 16 bytes)

  // Number of stack slots between incoming argument block and the start of
  // a new frame.  The PROLOG must add this many slots to the stack.  The
  // EPILOG must remove this many slots. aarch64 needs two slots for
  // return address and fp.
  // TODO think this is correct but check
  in_preserve_stack_slots(4);

  // Number of outgoing stack slots killed above the out_preserve_stack_slots
  // for calls to C.  Supports the var-args backing area for register parms.
  varargs_C_out_slots_killed(frame::arg_reg_save_area_bytes/BytesPerInt);

  // The after-PROLOG location of the return address.  Location of
  // return address specifies a type (REG or STACK) and a number
  // representing the register number (i.e. - use a register name) or
  // stack slot.
  // Ret Addr is on stack in slot 0 if no locks or verification or alignment.
  // Otherwise, it is above the locks and verification slot and alignment word
  // TODO this may well be correct but need to check why that - 2 is there
  // ppc port uses 0 but we definitely need to allow for fixed_slots
  // which folds in the space used for monitors
  return_addr(STACK - 2 +
              round_to((Compile::current()->in_preserve_stack_slots() +
                        Compile::current()->fixed_slots()),
                       stack_alignment_in_slots()));

  // Body of function which returns an integer array locating
  // arguments either in registers or in stack slots.  Passed an array
  // of ideal registers called "sig" and a "length" count.  Stack-slot
  // offsets are based on outgoing arguments, i.e. a CALLER setting up
  // arguments for a CALLEE.  Incoming stack arguments are
  // automatically biased by the preserve_stack_slots field above.

  calling_convention
  %{
    // No difference between ingoing/outgoing just pass false
    SharedRuntime::java_calling_convention(sig_bt, regs, length, false);
  %}

  c_calling_convention
  %{
    // This is obviously always outgoing
    (void) SharedRuntime::c_calling_convention(sig_bt, regs, NULL, length);
  %}

  // Location of compiled Java return values.  Same as C for now.
  return_value
  %{
    // TODO do we allow ideal_reg == Op_RegN???
    assert(ideal_reg >= Op_RegI && ideal_reg <= Op_RegL,
           "only return normal values");

    static const int lo[Op_RegL + 1] = { // enum name
      0,                                 // Op_Node
      0,                                 // Op_Set
      R0_num,                            // Op_RegN
      R0_num,                            // Op_RegI
      R0_num,                            // Op_RegP
      V0_num,                            // Op_RegF
      V0_num,                            // Op_RegD
      R0_num                             // Op_RegL
    };

    static const int hi[Op_RegL + 1] = { // enum name
      0,                                 // Op_Node
      0,                                 // Op_Set
      OptoReg::Bad,                       // Op_RegN
      OptoReg::Bad,                      // Op_RegI
      R0_H_num,                          // Op_RegP
      OptoReg::Bad,                      // Op_RegF
      V0_H_num,                          // Op_RegD
      R0_H_num                           // Op_RegL
    };

    return OptoRegPair(hi[ideal_reg], lo[ideal_reg]);
  %}
%}

//----------ATTRIBUTES---------------------------------------------------------
//----------Operand Attributes-------------------------------------------------
op_attrib op_cost(1);        // Required cost attribute

//----------Instruction Attributes---------------------------------------------
ins_attrib ins_cost(INSN_COST); // Required cost attribute
ins_attrib ins_size(32);        // Required size attribute (in bits)
ins_attrib ins_short_branch(0); // Required flag: is this instruction
                                // a non-matching short branch variant
                                // of some long branch?
ins_attrib ins_alignment(4);    // Required alignment attribute (must
                                // be a power of 2) specifies the
                                // alignment that some part of the
                                // instruction (not necessarily the
                                // start) requires.  If > 1, a
                                // compute_padding() function must be
                                // provided for the instruction

//----------OPERANDS-----------------------------------------------------------
// Operand definitions must precede instruction definitions for correct parsing
// in the ADLC because operands constitute user defined types which are used in
// instruction definitions.

//----------Simple Operands----------------------------------------------------

// Integer operands 32 bit
// 32 bit immediate
operand immI()
%{
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 32 bit zero
operand immI0()
%{
  predicate(n->get_int() == 0);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 32 bit unit increment
operand immI_1()
%{
  predicate(n->get_int() == 1);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 32 bit unit decrement
operand immI_M1()
%{
  predicate(n->get_int() == -1);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI_le_4()
%{
  predicate(n->get_int() <= 4);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI_31()
%{
  predicate(n->get_int() == 31);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI_8()
%{
  predicate(n->get_int() == 8);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI_16()
%{
  predicate(n->get_int() == 16);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI_24()
%{
  predicate(n->get_int() == 24);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI_32()
%{
  predicate(n->get_int() == 32);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI_48()
%{
  predicate(n->get_int() == 48);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI_56()
%{
  predicate(n->get_int() == 56);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI_64()
%{
  predicate(n->get_int() == 64);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI_255()
%{
  predicate(n->get_int() == 255);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI_65535()
%{
  predicate(n->get_int() == 65535);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immL_63()
%{
  predicate(n->get_int() == 63);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immL_255()
%{
  predicate(n->get_int() == 255);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immL_65535()
%{
  predicate(n->get_long() == 65535L);
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immL_4294967295()
%{
  predicate(n->get_long() == 4294967295L);
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immL_bitmask()
%{
  predicate(((n->get_long() & 0xc000000000000000l) == 0)
            && is_power_of_2(n->get_long() + 1));
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI_bitmask()
%{
  predicate(((n->get_int() & 0xc0000000) == 0)
            && is_power_of_2(n->get_int() + 1));
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Scale values for scaled offset addressing modes (up to long but not quad)
operand immIScale()
%{
  predicate(0 <= n->get_int() && (n->get_int() <= 3));
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 26 bit signed offset -- for pc-relative branches
operand immI26()
%{
  predicate(((-(1 << 25)) <= n->get_int()) && (n->get_int() < (1 << 25)));
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 19 bit signed offset -- for pc-relative loads
operand immI19()
%{
  predicate(((-(1 << 18)) <= n->get_int()) && (n->get_int() < (1 << 18)));
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 12 bit unsigned offset -- for base plus immediate loads
operand immIU12()
%{
  predicate((0 <= n->get_int()) && (n->get_int() < (1 << 12)));
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immLU12()
%{
  predicate((0 <= n->get_long()) && (n->get_long() < (1 << 12)));
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Offset for scaled or unscaled immediate loads and stores
operand immIOffset()
%{
  predicate(Address::offset_ok_for_immed(n->get_int()));
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immIOffset4()
%{
  predicate(Address::offset_ok_for_immed(n->get_int(), 2));
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immIOffset8()
%{
  predicate(Address::offset_ok_for_immed(n->get_int(), 3));
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immIOffset16()
%{
  predicate(Address::offset_ok_for_immed(n->get_int(), 4));
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immLoffset()
%{
  predicate(Address::offset_ok_for_immed(n->get_long()));
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immLoffset4()
%{
  predicate(Address::offset_ok_for_immed(n->get_long(), 2));
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immLoffset8()
%{
  predicate(Address::offset_ok_for_immed(n->get_long(), 3));
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immLoffset16()
%{
  predicate(Address::offset_ok_for_immed(n->get_long(), 4));
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 32 bit integer valid for add sub immediate
operand immIAddSub()
%{
  predicate(Assembler::operand_valid_for_add_sub_immediate((long)n->get_int()));
  match(ConI);
  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 32 bit unsigned integer valid for logical immediate
// TODO -- check this is right when e.g the mask is 0x80000000
operand immILog()
%{
  predicate(Assembler::operand_valid_for_logical_immediate(/*is32*/true, (unsigned long)n->get_int()));
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Integer operands 64 bit
// 64 bit immediate
operand immL()
%{
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 64 bit zero
operand immL0()
%{
  predicate(n->get_long() == 0);
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 64 bit unit increment
operand immL_1()
%{
  predicate(n->get_long() == 1);
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 64 bit unit decrement
operand immL_M1()
%{
  predicate(n->get_long() == -1);
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 32 bit offset of pc in thread anchor

operand immL_pc_off()
%{
  predicate(n->get_long() == in_bytes(JavaThread::frame_anchor_offset()) +
                             in_bytes(JavaFrameAnchor::last_Java_pc_offset()));
  match(ConL);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 64 bit integer valid for add sub immediate
operand immLAddSub()
%{
  predicate(Assembler::operand_valid_for_add_sub_immediate(n->get_long()));
  match(ConL);
  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// 64 bit integer valid for logical immediate
operand immLLog()
%{
  predicate(Assembler::operand_valid_for_logical_immediate(/*is32*/false, (unsigned long)n->get_long()));
  match(ConL);
  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Long Immediate: low 32-bit mask
operand immL_32bits()
%{
  predicate(n->get_long() == 0xFFFFFFFFL);
  match(ConL);
  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Pointer operands
// Pointer Immediate
operand immP()
%{
  match(ConP);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// NULL Pointer Immediate
operand immP0()
%{
  predicate(n->get_ptr() == 0);
  match(ConP);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Pointer Immediate One
// this is used in object initialization (initial object header)
operand immP_1()
%{
  predicate(n->get_ptr() == 1);
  match(ConP);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Polling Page Pointer Immediate
operand immPollPage()
%{
  predicate((address)n->get_ptr() == os::get_polling_page());
  match(ConP);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Card Table Byte Map Base
operand immByteMapBase()
%{
  // Get base of card map
  predicate((jbyte*)n->get_ptr() ==
        ((CardTableModRefBS*)(Universe::heap()->barrier_set()))->byte_map_base);
  match(ConP);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Pointer Immediate Minus One
// this is used when we want to write the current PC to the thread anchor
operand immP_M1()
%{
  predicate(n->get_ptr() == -1);
  match(ConP);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Pointer Immediate Minus Two
// this is used when we want to write the current PC to the thread anchor
operand immP_M2()
%{
  predicate(n->get_ptr() == -2);
  match(ConP);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Float and Double operands
// Double Immediate
operand immD()
%{
  match(ConD);
  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Double Immediate: +0.0d
operand immD0()
%{
  predicate(jlong_cast(n->getd()) == 0);
  match(ConD);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// constant 'double +0.0'.
operand immDPacked()
%{
  predicate(Assembler::operand_valid_for_float_immediate(n->getd()));
  match(ConD);
  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Float Immediate
operand immF()
%{
  match(ConF);
  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Float Immediate: +0.0f.
operand immF0()
%{
  predicate(jint_cast(n->getf()) == 0);
  match(ConF);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

//
operand immFPacked()
%{
  predicate(Assembler::operand_valid_for_float_immediate((double)n->getf()));
  match(ConF);
  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Narrow pointer operands
// Narrow Pointer Immediate
operand immN()
%{
  match(ConN);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Narrow NULL Pointer Immediate
operand immN0()
%{
  predicate(n->get_narrowcon() == 0);
  match(ConN);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

operand immNKlass()
%{
  match(ConNKlass);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Integer 32 bit Register Operands
// Integer 32 bitRegister (excludes SP)
operand iRegI()
%{
  constraint(ALLOC_IN_RC(any_reg32));
  match(RegI);
  match(iRegINoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Integer 32 bit Register not Special
operand iRegINoSp()
%{
  constraint(ALLOC_IN_RC(no_special_reg32));
  match(RegI);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Integer 64 bit Register Operands
// Integer 64 bit Register (includes SP)
operand iRegL()
%{
  constraint(ALLOC_IN_RC(any_reg));
  match(RegL);
  match(iRegLNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Integer 64 bit Register not Special
operand iRegLNoSp()
%{
  constraint(ALLOC_IN_RC(no_special_reg));
  match(RegL);
  match(iRegL_R0);
  format %{ %}
  interface(REG_INTER);
%}

// Pointer Register Operands
// Pointer Register
operand iRegP()
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(RegP);
  match(iRegPNoSp);
  match(iRegP_R0);
  //match(iRegP_R2);
  //match(iRegP_R4);
  //match(iRegP_R5);
  match(thread_RegP);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Pointer 64 bit Register not Special
operand iRegPNoSp()
%{
  constraint(ALLOC_IN_RC(no_special_ptr_reg));
  match(RegP);
  // match(iRegP);
  // match(iRegP_R0);
  // match(iRegP_R2);
  // match(iRegP_R4);
  // match(iRegP_R5);
  // match(thread_RegP);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Pointer 64 bit Register R0 only
operand iRegP_R0()
%{
  constraint(ALLOC_IN_RC(r0_reg));
  match(RegP);
  // match(iRegP);
  match(iRegPNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Pointer 64 bit Register R1 only
operand iRegP_R1()
%{
  constraint(ALLOC_IN_RC(r1_reg));
  match(RegP);
  // match(iRegP);
  match(iRegPNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Pointer 64 bit Register R2 only
operand iRegP_R2()
%{
  constraint(ALLOC_IN_RC(r2_reg));
  match(RegP);
  // match(iRegP);
  match(iRegPNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Pointer 64 bit Register R3 only
operand iRegP_R3()
%{
  constraint(ALLOC_IN_RC(r3_reg));
  match(RegP);
  // match(iRegP);
  match(iRegPNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Pointer 64 bit Register R4 only
operand iRegP_R4()
%{
  constraint(ALLOC_IN_RC(r4_reg));
  match(RegP);
  // match(iRegP);
  match(iRegPNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Pointer 64 bit Register R5 only
operand iRegP_R5()
%{
  constraint(ALLOC_IN_RC(r5_reg));
  match(RegP);
  // match(iRegP);
  match(iRegPNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Pointer 64 bit Register R10 only
operand iRegP_R10()
%{
  constraint(ALLOC_IN_RC(r10_reg));
  match(RegP);
  // match(iRegP);
  match(iRegPNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Long 64 bit Register R0 only
operand iRegL_R0()
%{
  constraint(ALLOC_IN_RC(r0_reg));
  match(RegL);
  match(iRegLNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Long 64 bit Register R2 only
operand iRegL_R2()
%{
  constraint(ALLOC_IN_RC(r2_reg));
  match(RegL);
  match(iRegLNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Long 64 bit Register R3 only
operand iRegL_R3()
%{
  constraint(ALLOC_IN_RC(r3_reg));
  match(RegL);
  match(iRegLNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Long 64 bit Register R11 only
operand iRegL_R11()
%{
  constraint(ALLOC_IN_RC(r11_reg));
  match(RegL);
  match(iRegLNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Pointer 64 bit Register FP only
operand iRegP_FP()
%{
  constraint(ALLOC_IN_RC(fp_reg));
  match(RegP);
  // match(iRegP);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Register R0 only
operand iRegI_R0()
%{
  constraint(ALLOC_IN_RC(int_r0_reg));
  match(RegI);
  match(iRegINoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Register R2 only
operand iRegI_R2()
%{
  constraint(ALLOC_IN_RC(int_r2_reg));
  match(RegI);
  match(iRegINoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Register R3 only
operand iRegI_R3()
%{
  constraint(ALLOC_IN_RC(int_r3_reg));
  match(RegI);
  match(iRegINoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}


// Register R4 only
operand iRegI_R4()
%{
  constraint(ALLOC_IN_RC(int_r4_reg));
  match(RegI);
  match(iRegINoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}


// Pointer Register Operands
// Narrow Pointer Register
operand iRegN()
%{
  constraint(ALLOC_IN_RC(any_reg32));
  match(RegN);
  match(iRegNNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

operand iRegN_R0()
%{
  constraint(ALLOC_IN_RC(r0_reg));
  match(iRegN);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

operand iRegN_R2()
%{
  constraint(ALLOC_IN_RC(r2_reg));
  match(iRegN);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

operand iRegN_R3()
%{
  constraint(ALLOC_IN_RC(r3_reg));
  match(iRegN);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Integer 64 bit Register not Special
operand iRegNNoSp()
%{
  constraint(ALLOC_IN_RC(no_special_reg32));
  match(RegN);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// heap base register -- used for encoding immN0

operand iRegIHeapbase()
%{
  constraint(ALLOC_IN_RC(heapbase_reg));
  match(RegI);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Float Register
// Float register operands
operand vRegF()
%{
  constraint(ALLOC_IN_RC(float_reg));
  match(RegF);

  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Double Register
// Double register operands
operand vRegD()
%{
  constraint(ALLOC_IN_RC(double_reg));
  match(RegD);

  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

operand vecD()
%{
  constraint(ALLOC_IN_RC(vectord_reg));
  match(VecD);

  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

operand vecX()
%{
  constraint(ALLOC_IN_RC(vectorx_reg));
  match(VecX);

  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

operand vRegD_V0()
%{
  constraint(ALLOC_IN_RC(v0_reg));
  match(RegD);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

operand vRegD_V1()
%{
  constraint(ALLOC_IN_RC(v1_reg));
  match(RegD);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

operand vRegD_V2()
%{
  constraint(ALLOC_IN_RC(v2_reg));
  match(RegD);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

operand vRegD_V3()
%{
  constraint(ALLOC_IN_RC(v3_reg));
  match(RegD);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Flags register, used as output of signed compare instructions

// note that on AArch64 we also use this register as the output for
// for floating point compare instructions (CmpF CmpD). this ensures
// that ordered inequality tests use GT, GE, LT or LE none of which
// pass through cases where the result is unordered i.e. one or both
// inputs to the compare is a NaN. this means that the ideal code can
// replace e.g. a GT with an LE and not end up capturing the NaN case
// (where the comparison should always fail). EQ and NE tests are
// always generated in ideal code so that unordered folds into the NE
// case, matching the behaviour of AArch64 NE.
//
// This differs from x86 where the outputs of FP compares use a
// special FP flags registers and where compares based on this
// register are distinguished into ordered inequalities (cmpOpUCF) and
// EQ/NEQ tests (cmpOpUCF2). x86 has to special case the latter tests
// to explicitly handle the unordered case in branches. x86 also has
// to include extra CMoveX rules to accept a cmpOpUCF input.

operand rFlagsReg()
%{
  constraint(ALLOC_IN_RC(int_flags));
  match(RegFlags);

  op_cost(0);
  format %{ "RFLAGS" %}
  interface(REG_INTER);
%}

// Flags register, used as output of unsigned compare instructions
operand rFlagsRegU()
%{
  constraint(ALLOC_IN_RC(int_flags));
  match(RegFlags);

  op_cost(0);
  format %{ "RFLAGSU" %}
  interface(REG_INTER);
%}

// Special Registers

// Method Register
operand inline_cache_RegP(iRegP reg)
%{
  constraint(ALLOC_IN_RC(method_reg)); // inline_cache_reg
  match(reg);
  match(iRegPNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

operand interpreter_method_oop_RegP(iRegP reg)
%{
  constraint(ALLOC_IN_RC(method_reg)); // interpreter_method_oop_reg
  match(reg);
  match(iRegPNoSp);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

// Thread Register
operand thread_RegP(iRegP reg)
%{
  constraint(ALLOC_IN_RC(thread_reg)); // link_reg
  match(reg);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

operand lr_RegP(iRegP reg)
%{
  constraint(ALLOC_IN_RC(lr_reg)); // link_reg
  match(reg);
  op_cost(0);
  format %{ %}
  interface(REG_INTER);
%}

//----------Memory Operands----------------------------------------------------

operand indirect(iRegP reg)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(reg);
  op_cost(0);
  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp(0x0);
  %}
%}

operand indIndexScaledI2L(iRegP reg, iRegI ireg, immIScale scale)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  predicate(size_fits_all_mem_uses(n->as_AddP(), n->in(AddPNode::Offset)->in(2)->get_int()));
  match(AddP reg (LShiftL (ConvI2L ireg) scale));
  op_cost(0);
  format %{ "$reg, $ireg sxtw($scale), 0, I2L" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($ireg);
    scale($scale);
    disp(0x0);
  %}
%}

operand indIndexScaled(iRegP reg, iRegL lreg, immIScale scale)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  predicate(size_fits_all_mem_uses(n->as_AddP(), n->in(AddPNode::Offset)->in(2)->get_int()));
  match(AddP reg (LShiftL lreg scale));
  op_cost(0);
  format %{ "$reg, $lreg lsl($scale)" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($lreg);
    scale($scale);
    disp(0x0);
  %}
%}

operand indIndexI2L(iRegP reg, iRegI ireg)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg (ConvI2L ireg));
  op_cost(0);
  format %{ "$reg, $ireg, 0, I2L" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($ireg);
    scale(0x0);
    disp(0x0);
  %}
%}

operand indIndex(iRegP reg, iRegL lreg)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg lreg);
  op_cost(0);
  format %{ "$reg, $lreg" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($lreg);
    scale(0x0);
    disp(0x0);
  %}
%}

operand indOffI(iRegP reg, immIOffset off)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg off);
  op_cost(0);
  format %{ "[$reg, $off]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp($off);
  %}
%}

operand indOffI4(iRegP reg, immIOffset4 off)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg off);
  op_cost(0);
  format %{ "[$reg, $off]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp($off);
  %}
%}

operand indOffI8(iRegP reg, immIOffset8 off)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg off);
  op_cost(0);
  format %{ "[$reg, $off]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp($off);
  %}
%}

operand indOffI16(iRegP reg, immIOffset16 off)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg off);
  op_cost(0);
  format %{ "[$reg, $off]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp($off);
  %}
%}

operand indOffL(iRegP reg, immLoffset off)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg off);
  op_cost(0);
  format %{ "[$reg, $off]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp($off);
  %}
%}

operand indOffL4(iRegP reg, immLoffset4 off)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg off);
  op_cost(0);
  format %{ "[$reg, $off]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp($off);
  %}
%}

operand indOffL8(iRegP reg, immLoffset8 off)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg off);
  op_cost(0);
  format %{ "[$reg, $off]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp($off);
  %}
%}

operand indOffL16(iRegP reg, immLoffset16 off)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg off);
  op_cost(0);
  format %{ "[$reg, $off]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp($off);
  %}
%}

operand indirectN(iRegN reg)
%{
  predicate(Universe::narrow_oop_shift() == 0);
  constraint(ALLOC_IN_RC(ptr_reg));
  match(DecodeN reg);
  op_cost(0);
  format %{ "[$reg]\t# narrow" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp(0x0);
  %}
%}

operand indIndexScaledI2LN(iRegN reg, iRegI ireg, immIScale scale)
%{
  predicate(Universe::narrow_oop_shift() == 0 && size_fits_all_mem_uses(n->as_AddP(), n->in(AddPNode::Offset)->in(2)->get_int()));
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (DecodeN reg) (LShiftL (ConvI2L ireg) scale));
  op_cost(0);
  format %{ "$reg, $ireg sxtw($scale), 0, I2L\t# narrow" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($ireg);
    scale($scale);
    disp(0x0);
  %}
%}

operand indIndexScaledN(iRegN reg, iRegL lreg, immIScale scale)
%{
  predicate(Universe::narrow_oop_shift() == 0 && size_fits_all_mem_uses(n->as_AddP(), n->in(AddPNode::Offset)->in(2)->get_int()));
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (DecodeN reg) (LShiftL lreg scale));
  op_cost(0);
  format %{ "$reg, $lreg lsl($scale)\t# narrow" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($lreg);
    scale($scale);
    disp(0x0);
  %}
%}

operand indIndexI2LN(iRegN reg, iRegI ireg)
%{
  predicate(Universe::narrow_oop_shift() == 0);
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (DecodeN reg) (ConvI2L ireg));
  op_cost(0);
  format %{ "$reg, $ireg, 0, I2L\t# narrow" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($ireg);
    scale(0x0);
    disp(0x0);
  %}
%}

operand indIndexN(iRegN reg, iRegL lreg)
%{
  predicate(Universe::narrow_oop_shift() == 0);
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (DecodeN reg) lreg);
  op_cost(0);
  format %{ "$reg, $lreg\t# narrow" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($lreg);
    scale(0x0);
    disp(0x0);
  %}
%}

operand indOffIN(iRegN reg, immIOffset off)
%{
  predicate(Universe::narrow_oop_shift() == 0);
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (DecodeN reg) off);
  op_cost(0);
  format %{ "[$reg, $off]\t# narrow" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp($off);
  %}
%}

operand indOffLN(iRegN reg, immLoffset off)
%{
  predicate(Universe::narrow_oop_shift() == 0);
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (DecodeN reg) off);
  op_cost(0);
  format %{ "[$reg, $off]\t# narrow" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp($off);
  %}
%}



// AArch64 opto stubs need to write to the pc slot in the thread anchor
operand thread_anchor_pc(thread_RegP reg, immL_pc_off off)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg off);
  op_cost(0);
  format %{ "[$reg, $off]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0xffffffff);
    scale(0x0);
    disp($off);
  %}
%}

//----------Special Memory Operands--------------------------------------------
// Stack Slot Operand - This operand is used for loading and storing temporary
//                      values on the stack where a match requires a value to
//                      flow through memory.
operand stackSlotP(sRegP reg)
%{
  constraint(ALLOC_IN_RC(stack_slots));
  op_cost(100);
  // No match rule because this operand is only generated in matching
  // match(RegP);
  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base(0x1e);  // RSP
    index(0x0);  // No Index
    scale(0x0);  // No Scale
    disp($reg);  // Stack Offset
  %}
%}

operand stackSlotI(sRegI reg)
%{
  constraint(ALLOC_IN_RC(stack_slots));
  // No match rule because this operand is only generated in matching
  // match(RegI);
  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base(0x1e);  // RSP
    index(0x0);  // No Index
    scale(0x0);  // No Scale
    disp($reg);  // Stack Offset
  %}
%}

operand stackSlotF(sRegF reg)
%{
  constraint(ALLOC_IN_RC(stack_slots));
  // No match rule because this operand is only generated in matching
  // match(RegF);
  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base(0x1e);  // RSP
    index(0x0);  // No Index
    scale(0x0);  // No Scale
    disp($reg);  // Stack Offset
  %}
%}

operand stackSlotD(sRegD reg)
%{
  constraint(ALLOC_IN_RC(stack_slots));
  // No match rule because this operand is only generated in matching
  // match(RegD);
  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base(0x1e);  // RSP
    index(0x0);  // No Index
    scale(0x0);  // No Scale
    disp($reg);  // Stack Offset
  %}
%}

operand stackSlotL(sRegL reg)
%{
  constraint(ALLOC_IN_RC(stack_slots));
  // No match rule because this operand is only generated in matching
  // match(RegL);
  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base(0x1e);  // RSP
    index(0x0);  // No Index
    scale(0x0);  // No Scale
    disp($reg);  // Stack Offset
  %}
%}

// Operands for expressing Control Flow
// NOTE: Label is a predefined operand which should not be redefined in
//       the AD file. It is generically handled within the ADLC.

//----------Conditional Branch Operands----------------------------------------
// Comparison Op  - This is the operation of the comparison, and is limited to
//                  the following set of codes:
//                  L (<), LE (<=), G (>), GE (>=), E (==), NE (!=)
//
// Other attributes of the comparison, such as unsignedness, are specified
// by the comparison instruction that sets a condition code flags register.
// That result is represented by a flags operand whose subtype is appropriate
// to the unsignedness (etc.) of the comparison.
//
// Later, the instruction which matches both the Comparison Op (a Bool) and
// the flags (produced by the Cmp) specifies the coding of the comparison op
// by matching a specific subtype of Bool operand below, such as cmpOpU.

// used for signed integral comparisons and fp comparisons

operand cmpOp()
%{
  match(Bool);

  format %{ "" %}
  interface(COND_INTER) %{
    equal(0x0, "eq");
    not_equal(0x1, "ne");
    less(0xb, "lt");
    greater_equal(0xa, "ge");
    less_equal(0xd, "le");
    greater(0xc, "gt");
    overflow(0x6, "vs");
    no_overflow(0x7, "vc");
  %}
%}

// used for unsigned integral comparisons

operand cmpOpU()
%{
  match(Bool);

  format %{ "" %}
  interface(COND_INTER) %{
    equal(0x0, "eq");
    not_equal(0x1, "ne");
    less(0x3, "lo");
    greater_equal(0x2, "hs");
    less_equal(0x9, "ls");
    greater(0x8, "hi");
    overflow(0x6, "vs");
    no_overflow(0x7, "vc");
  %}
%}

// used for certain integral comparisons which can be
// converted to cbxx or tbxx instructions

operand cmpOpEqNe()
%{
  match(Bool);
  match(CmpOp);
  op_cost(0);
  predicate(n->as_Bool()->_test._test == BoolTest::ne
            || n->as_Bool()->_test._test == BoolTest::eq);

  format %{ "" %}
  interface(COND_INTER) %{
    equal(0x0, "eq");
    not_equal(0x1, "ne");
    less(0xb, "lt");
    greater_equal(0xa, "ge");
    less_equal(0xd, "le");
    greater(0xc, "gt");
    overflow(0x6, "vs");
    no_overflow(0x7, "vc");
  %}
%}

// used for certain integral comparisons which can be
// converted to cbxx or tbxx instructions

operand cmpOpLtGe()
%{
  match(Bool);
  match(CmpOp);
  op_cost(0);

  predicate(n->as_Bool()->_test._test == BoolTest::lt
            || n->as_Bool()->_test._test == BoolTest::ge);

  format %{ "" %}
  interface(COND_INTER) %{
    equal(0x0, "eq");
    not_equal(0x1, "ne");
    less(0xb, "lt");
    greater_equal(0xa, "ge");
    less_equal(0xd, "le");
    greater(0xc, "gt");
    overflow(0x6, "vs");
    no_overflow(0x7, "vc");
  %}
%}

// used for certain unsigned integral comparisons which can be
// converted to cbxx or tbxx instructions

operand cmpOpUEqNeLtGe()
%{
  match(Bool);
  match(CmpOp);
  op_cost(0);

  predicate(n->as_Bool()->_test._test == BoolTest::eq
            || n->as_Bool()->_test._test == BoolTest::ne
            || n->as_Bool()->_test._test == BoolTest::lt
            || n->as_Bool()->_test._test == BoolTest::ge);

  format %{ "" %}
  interface(COND_INTER) %{
    equal(0x0, "eq");
    not_equal(0x1, "ne");
    less(0xb, "lt");
    greater_equal(0xa, "ge");
    less_equal(0xd, "le");
    greater(0xc, "gt");
    overflow(0x6, "vs");
    no_overflow(0x7, "vc");
  %}
%}

// Special operand allowing long args to int ops to be truncated for free

operand iRegL2I(iRegL reg) %{

  op_cost(0);

  match(ConvL2I reg);

  format %{ "l2i($reg)" %}

  interface(REG_INTER)
%}

opclass vmem4(indirect, indIndex, indOffI4, indOffL4);
opclass vmem8(indirect, indIndex, indOffI8, indOffL8);
opclass vmem16(indirect, indIndex, indOffI16, indOffL16);

//----------OPERAND CLASSES----------------------------------------------------
// Operand Classes are groups of operands that are used as to simplify
// instruction definitions by not requiring the AD writer to specify
// separate instructions for every form of operand when the
// instruction accepts multiple operand types with the same basic
// encoding and format. The classic case of this is memory operands.

// memory is used to define read/write location for load/store
// instruction defs. we can turn a memory op into an Address

opclass memory(indirect, indIndexScaled, indIndexScaledI2L, indIndexI2L, indIndex, indOffI, indOffL,
               indirectN, indIndexScaledN, indIndexScaledI2LN, indIndexI2LN, indIndexN, indOffIN, indOffLN);

// iRegIorL2I is used for src inputs in rules for 32 bit int (I)
// operations. it allows the src to be either an iRegI or a (ConvL2I
// iRegL). in the latter case the l2i normally planted for a ConvL2I
// can be elided because the 32-bit instruction will just employ the
// lower 32 bits anyway.
//
// n.b. this does not elide all L2I conversions. if the truncated
// value is consumed by more than one operation then the ConvL2I
// cannot be bundled into the consuming nodes so an l2i gets planted
// (actually a movw $dst $src) and the downstream instructions consume
// the result of the l2i as an iRegI input. That's a shame since the
// movw is actually redundant but its not too costly.

opclass iRegIorL2I(iRegI, iRegL2I);

//----------PIPELINE-----------------------------------------------------------
// Rules which define the behavior of the target architectures pipeline.

// For specific pipelines, eg A53, define the stages of that pipeline
//pipe_desc(ISS, EX1, EX2, WR);
#define ISS S0
#define EX1 S1
#define EX2 S2
#define WR  S3

// Integer ALU reg operation
pipeline %{

attributes %{
  // ARM instructions are of fixed length
  fixed_size_instructions;        // Fixed size instructions TODO does
  max_instructions_per_bundle = 2;   // A53 = 2, A57 = 4
  // ARM instructions come in 32-bit word units
  instruction_unit_size = 4;         // An instruction is 4 bytes long
  instruction_fetch_unit_size = 64;  // The processor fetches one line
  instruction_fetch_units = 1;       // of 64 bytes

  // List of nop instructions
  nops( MachNop );
%}

// We don't use an actual pipeline model so don't care about resources
// or description. we do use pipeline classes to introduce fixed
// latencies

//----------RESOURCES----------------------------------------------------------
// Resources are the functional units available to the machine

resources( INS0, INS1, INS01 = INS0 | INS1,
           ALU0, ALU1, ALU = ALU0 | ALU1,
           MAC,
           DIV,
           BRANCH,
           LDST,
           NEON_FP);

//----------PIPELINE DESCRIPTION-----------------------------------------------
// Pipeline Description specifies the stages in the machine's pipeline

// Define the pipeline as a generic 6 stage pipeline
pipe_desc(S0, S1, S2, S3, S4, S5);

//----------PIPELINE CLASSES---------------------------------------------------
// Pipeline Classes describe the stages in which input and output are
// referenced by the hardware pipeline.

pipe_class fp_dop_reg_reg_s(vRegF dst, vRegF src1, vRegF src2)
%{
  single_instruction;
  src1   : S1(read);
  src2   : S2(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_dop_reg_reg_d(vRegD dst, vRegD src1, vRegD src2)
%{
  single_instruction;
  src1   : S1(read);
  src2   : S2(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_uop_s(vRegF dst, vRegF src)
%{
  single_instruction;
  src    : S1(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_uop_d(vRegD dst, vRegD src)
%{
  single_instruction;
  src    : S1(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_d2f(vRegF dst, vRegD src)
%{
  single_instruction;
  src    : S1(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_f2d(vRegD dst, vRegF src)
%{
  single_instruction;
  src    : S1(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_f2i(iRegINoSp dst, vRegF src)
%{
  single_instruction;
  src    : S1(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_f2l(iRegLNoSp dst, vRegF src)
%{
  single_instruction;
  src    : S1(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_i2f(vRegF dst, iRegIorL2I src)
%{
  single_instruction;
  src    : S1(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_l2f(vRegF dst, iRegL src)
%{
  single_instruction;
  src    : S1(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_d2i(iRegINoSp dst, vRegD src)
%{
  single_instruction;
  src    : S1(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_d2l(iRegLNoSp dst, vRegD src)
%{
  single_instruction;
  src    : S1(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_i2d(vRegD dst, iRegIorL2I src)
%{
  single_instruction;
  src    : S1(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_l2d(vRegD dst, iRegIorL2I src)
%{
  single_instruction;
  src    : S1(read);
  dst    : S5(write);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class fp_div_s(vRegF dst, vRegF src1, vRegF src2)
%{
  single_instruction;
  src1   : S1(read);
  src2   : S2(read);
  dst    : S5(write);
  INS0   : ISS;
  NEON_FP : S5;
%}

pipe_class fp_div_d(vRegD dst, vRegD src1, vRegD src2)
%{
  single_instruction;
  src1   : S1(read);
  src2   : S2(read);
  dst    : S5(write);
  INS0   : ISS;
  NEON_FP : S5;
%}

pipe_class fp_cond_reg_reg_s(vRegF dst, vRegF src1, vRegF src2, rFlagsReg cr)
%{
  single_instruction;
  cr     : S1(read);
  src1   : S1(read);
  src2   : S1(read);
  dst    : S3(write);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class fp_cond_reg_reg_d(vRegD dst, vRegD src1, vRegD src2, rFlagsReg cr)
%{
  single_instruction;
  cr     : S1(read);
  src1   : S1(read);
  src2   : S1(read);
  dst    : S3(write);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class fp_imm_s(vRegF dst)
%{
  single_instruction;
  dst    : S3(write);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class fp_imm_d(vRegD dst)
%{
  single_instruction;
  dst    : S3(write);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class fp_load_constant_s(vRegF dst)
%{
  single_instruction;
  dst    : S4(write);
  INS01  : ISS;
  NEON_FP : S4;
%}

pipe_class fp_load_constant_d(vRegD dst)
%{
  single_instruction;
  dst    : S4(write);
  INS01  : ISS;
  NEON_FP : S4;
%}

pipe_class vmul64(vecD dst, vecD src1, vecD src2)
%{
  single_instruction;
  dst    : S5(write);
  src1   : S1(read);
  src2   : S1(read);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class vmul128(vecX dst, vecX src1, vecX src2)
%{
  single_instruction;
  dst    : S5(write);
  src1   : S1(read);
  src2   : S1(read);
  INS0   : ISS;
  NEON_FP : S5;
%}

pipe_class vmla64(vecD dst, vecD src1, vecD src2)
%{
  single_instruction;
  dst    : S5(write);
  src1   : S1(read);
  src2   : S1(read);
  dst    : S1(read);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class vmla128(vecX dst, vecX src1, vecX src2)
%{
  single_instruction;
  dst    : S5(write);
  src1   : S1(read);
  src2   : S1(read);
  dst    : S1(read);
  INS0   : ISS;
  NEON_FP : S5;
%}

pipe_class vdop64(vecD dst, vecD src1, vecD src2)
%{
  single_instruction;
  dst    : S4(write);
  src1   : S2(read);
  src2   : S2(read);
  INS01  : ISS;
  NEON_FP : S4;
%}

pipe_class vdop128(vecX dst, vecX src1, vecX src2)
%{
  single_instruction;
  dst    : S4(write);
  src1   : S2(read);
  src2   : S2(read);
  INS0   : ISS;
  NEON_FP : S4;
%}

pipe_class vlogical64(vecD dst, vecD src1, vecD src2)
%{
  single_instruction;
  dst    : S3(write);
  src1   : S2(read);
  src2   : S2(read);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class vlogical128(vecX dst, vecX src1, vecX src2)
%{
  single_instruction;
  dst    : S3(write);
  src1   : S2(read);
  src2   : S2(read);
  INS0   : ISS;
  NEON_FP : S3;
%}

pipe_class vshift64(vecD dst, vecD src, vecX shift)
%{
  single_instruction;
  dst    : S3(write);
  src    : S1(read);
  shift  : S1(read);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class vshift128(vecX dst, vecX src, vecX shift)
%{
  single_instruction;
  dst    : S3(write);
  src    : S1(read);
  shift  : S1(read);
  INS0   : ISS;
  NEON_FP : S3;
%}

pipe_class vshift64_imm(vecD dst, vecD src, immI shift)
%{
  single_instruction;
  dst    : S3(write);
  src    : S1(read);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class vshift128_imm(vecX dst, vecX src, immI shift)
%{
  single_instruction;
  dst    : S3(write);
  src    : S1(read);
  INS0   : ISS;
  NEON_FP : S3;
%}

pipe_class vdop_fp64(vecD dst, vecD src1, vecD src2)
%{
  single_instruction;
  dst    : S5(write);
  src1   : S1(read);
  src2   : S1(read);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class vdop_fp128(vecX dst, vecX src1, vecX src2)
%{
  single_instruction;
  dst    : S5(write);
  src1   : S1(read);
  src2   : S1(read);
  INS0   : ISS;
  NEON_FP : S5;
%}

pipe_class vmuldiv_fp64(vecD dst, vecD src1, vecD src2)
%{
  single_instruction;
  dst    : S5(write);
  src1   : S1(read);
  src2   : S1(read);
  INS0   : ISS;
  NEON_FP : S5;
%}

pipe_class vmuldiv_fp128(vecX dst, vecX src1, vecX src2)
%{
  single_instruction;
  dst    : S5(write);
  src1   : S1(read);
  src2   : S1(read);
  INS0   : ISS;
  NEON_FP : S5;
%}

pipe_class vsqrt_fp128(vecX dst, vecX src)
%{
  single_instruction;
  dst    : S5(write);
  src    : S1(read);
  INS0   : ISS;
  NEON_FP : S5;
%}

pipe_class vunop_fp64(vecD dst, vecD src)
%{
  single_instruction;
  dst    : S5(write);
  src    : S1(read);
  INS01  : ISS;
  NEON_FP : S5;
%}

pipe_class vunop_fp128(vecX dst, vecX src)
%{
  single_instruction;
  dst    : S5(write);
  src    : S1(read);
  INS0   : ISS;
  NEON_FP : S5;
%}

pipe_class vdup_reg_reg64(vecD dst, iRegI src)
%{
  single_instruction;
  dst    : S3(write);
  src    : S1(read);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class vdup_reg_reg128(vecX dst, iRegI src)
%{
  single_instruction;
  dst    : S3(write);
  src    : S1(read);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class vdup_reg_freg64(vecD dst, vRegF src)
%{
  single_instruction;
  dst    : S3(write);
  src    : S1(read);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class vdup_reg_freg128(vecX dst, vRegF src)
%{
  single_instruction;
  dst    : S3(write);
  src    : S1(read);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class vdup_reg_dreg128(vecX dst, vRegD src)
%{
  single_instruction;
  dst    : S3(write);
  src    : S1(read);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class vmovi_reg_imm64(vecD dst)
%{
  single_instruction;
  dst    : S3(write);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class vmovi_reg_imm128(vecX dst)
%{
  single_instruction;
  dst    : S3(write);
  INS0   : ISS;
  NEON_FP : S3;
%}

pipe_class vload_reg_mem64(vecD dst, vmem8 mem)
%{
  single_instruction;
  dst    : S5(write);
  mem    : ISS(read);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class vload_reg_mem128(vecX dst, vmem16 mem)
%{
  single_instruction;
  dst    : S5(write);
  mem    : ISS(read);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class vstore_reg_mem64(vecD src, vmem8 mem)
%{
  single_instruction;
  mem    : ISS(read);
  src    : S2(read);
  INS01  : ISS;
  NEON_FP : S3;
%}

pipe_class vstore_reg_mem128(vecD src, vmem16 mem)
%{
  single_instruction;
  mem    : ISS(read);
  src    : S2(read);
  INS01  : ISS;
  NEON_FP : S3;
%}

//------- Integer ALU operations --------------------------

// Integer ALU reg-reg operation
// Operands needed in EX1, result generated in EX2
// Eg.  ADD     x0, x1, x2
pipe_class ialu_reg_reg(iRegI dst, iRegI src1, iRegI src2)
%{
  single_instruction;
  dst    : EX2(write);
  src1   : EX1(read);
  src2   : EX1(read);
  INS01  : ISS; // Dual issue as instruction 0 or 1
  ALU    : EX2;
%}

// Integer ALU reg-reg operation with constant shift
// Shifted register must be available in LATE_ISS instead of EX1
// Eg.  ADD     x0, x1, x2, LSL #2
pipe_class ialu_reg_reg_shift(iRegI dst, iRegI src1, iRegI src2, immI shift)
%{
  single_instruction;
  dst    : EX2(write);
  src1   : EX1(read);
  src2   : ISS(read);
  INS01  : ISS;
  ALU    : EX2;
%}

// Integer ALU reg operation with constant shift
// Eg.  LSL     x0, x1, #shift
pipe_class ialu_reg_shift(iRegI dst, iRegI src1)
%{
  single_instruction;
  dst    : EX2(write);
  src1   : ISS(read);
  INS01  : ISS;
  ALU    : EX2;
%}

// Integer ALU reg-reg operation with variable shift
// Both operands must be available in LATE_ISS instead of EX1
// Result is available in EX1 instead of EX2
// Eg.  LSLV    x0, x1, x2
pipe_class ialu_reg_reg_vshift(iRegI dst, iRegI src1, iRegI src2)
%{
  single_instruction;
  dst    : EX1(write);
  src1   : ISS(read);
  src2   : ISS(read);
  INS01  : ISS;
  ALU    : EX1;
%}

// Integer ALU reg-reg operation with extract
// As for _vshift above, but result generated in EX2
// Eg.  EXTR    x0, x1, x2, #N
pipe_class ialu_reg_reg_extr(iRegI dst, iRegI src1, iRegI src2)
%{
  single_instruction;
  dst    : EX2(write);
  src1   : ISS(read);
  src2   : ISS(read);
  INS1   : ISS; // Can only dual issue as Instruction 1
  ALU    : EX1;
%}

// Integer ALU reg operation
// Eg.  NEG     x0, x1
pipe_class ialu_reg(iRegI dst, iRegI src)
%{
  single_instruction;
  dst    : EX2(write);
  src    : EX1(read);
  INS01  : ISS;
  ALU    : EX2;
%}

// Integer ALU reg mmediate operation
// Eg.  ADD     x0, x1, #N
pipe_class ialu_reg_imm(iRegI dst, iRegI src1)
%{
  single_instruction;
  dst    : EX2(write);
  src1   : EX1(read);
  INS01  : ISS;
  ALU    : EX2;
%}

// Integer ALU immediate operation (no source operands)
// Eg.  MOV     x0, #N
pipe_class ialu_imm(iRegI dst)
%{
  single_instruction;
  dst    : EX1(write);
  INS01  : ISS;
  ALU    : EX1;
%}

//------- Compare operation -------------------------------

// Compare reg-reg
// Eg.  CMP     x0, x1
pipe_class icmp_reg_reg(rFlagsReg cr, iRegI op1, iRegI op2)
%{
  single_instruction;
//  fixed_latency(16);
  cr     : EX2(write);
  op1    : EX1(read);
  op2    : EX1(read);
  INS01  : ISS;
  ALU    : EX2;
%}

// Compare reg-reg
// Eg.  CMP     x0, #N
pipe_class icmp_reg_imm(rFlagsReg cr, iRegI op1)
%{
  single_instruction;
//  fixed_latency(16);
  cr     : EX2(write);
  op1    : EX1(read);
  INS01  : ISS;
  ALU    : EX2;
%}

//------- Conditional instructions ------------------------

// Conditional no operands
// Eg.  CSINC   x0, zr, zr, <cond>
pipe_class icond_none(iRegI dst, rFlagsReg cr)
%{
  single_instruction;
  cr     : EX1(read);
  dst    : EX2(write);
  INS01  : ISS;
  ALU    : EX2;
%}

// Conditional 2 operand
// EG.  CSEL    X0, X1, X2, <cond>
pipe_class icond_reg_reg(iRegI dst, iRegI src1, iRegI src2, rFlagsReg cr)
%{
  single_instruction;
  cr     : EX1(read);
  src1   : EX1(read);
  src2   : EX1(read);
  dst    : EX2(write);
  INS01  : ISS;
  ALU    : EX2;
%}

// Conditional 2 operand
// EG.  CSEL    X0, X1, X2, <cond>
pipe_class icond_reg(iRegI dst, iRegI src, rFlagsReg cr)
%{
  single_instruction;
  cr     : EX1(read);
  src    : EX1(read);
  dst    : EX2(write);
  INS01  : ISS;
  ALU    : EX2;
%}

//------- Multiply pipeline operations --------------------

// Multiply reg-reg
// Eg.  MUL     w0, w1, w2
pipe_class imul_reg_reg(iRegI dst, iRegI src1, iRegI src2)
%{
  single_instruction;
  dst    : WR(write);
  src1   : ISS(read);
  src2   : ISS(read);
  INS01  : ISS;
  MAC    : WR;
%}

// Multiply accumulate
// Eg.  MADD    w0, w1, w2, w3
pipe_class imac_reg_reg(iRegI dst, iRegI src1, iRegI src2, iRegI src3)
%{
  single_instruction;
  dst    : WR(write);
  src1   : ISS(read);
  src2   : ISS(read);
  src3   : ISS(read);
  INS01  : ISS;
  MAC    : WR;
%}

// Eg.  MUL     w0, w1, w2
pipe_class lmul_reg_reg(iRegI dst, iRegI src1, iRegI src2)
%{
  single_instruction;
  fixed_latency(3); // Maximum latency for 64 bit mul
  dst    : WR(write);
  src1   : ISS(read);
  src2   : ISS(read);
  INS01  : ISS;
  MAC    : WR;
%}

// Multiply accumulate
// Eg.  MADD    w0, w1, w2, w3
pipe_class lmac_reg_reg(iRegI dst, iRegI src1, iRegI src2, iRegI src3)
%{
  single_instruction;
  fixed_latency(3); // Maximum latency for 64 bit mul
  dst    : WR(write);
  src1   : ISS(read);
  src2   : ISS(read);
  src3   : ISS(read);
  INS01  : ISS;
  MAC    : WR;
%}

//------- Divide pipeline operations --------------------

// Eg.  SDIV    w0, w1, w2
pipe_class idiv_reg_reg(iRegI dst, iRegI src1, iRegI src2)
%{
  single_instruction;
  fixed_latency(8); // Maximum latency for 32 bit divide
  dst    : WR(write);
  src1   : ISS(read);
  src2   : ISS(read);
  INS0   : ISS; // Can only dual issue as instruction 0
  DIV    : WR;
%}

// Eg.  SDIV    x0, x1, x2
pipe_class ldiv_reg_reg(iRegI dst, iRegI src1, iRegI src2)
%{
  single_instruction;
  fixed_latency(16); // Maximum latency for 64 bit divide
  dst    : WR(write);
  src1   : ISS(read);
  src2   : ISS(read);
  INS0   : ISS; // Can only dual issue as instruction 0
  DIV    : WR;
%}

//------- Load pipeline operations ------------------------

// Load - prefetch
// Eg.  PFRM    <mem>
pipe_class iload_prefetch(memory mem)
%{
  single_instruction;
  mem    : ISS(read);
  INS01  : ISS;
  LDST   : WR;
%}

// Load - reg, mem
// Eg.  LDR     x0, <mem>
pipe_class iload_reg_mem(iRegI dst, memory mem)
%{
  single_instruction;
  dst    : WR(write);
  mem    : ISS(read);
  INS01  : ISS;
  LDST   : WR;
%}

// Load - reg, reg
// Eg.  LDR     x0, [sp, x1]
pipe_class iload_reg_reg(iRegI dst, iRegI src)
%{
  single_instruction;
  dst    : WR(write);
  src    : ISS(read);
  INS01  : ISS;
  LDST   : WR;
%}

//------- Store pipeline operations -----------------------

// Store - zr, mem
// Eg.  STR     zr, <mem>
pipe_class istore_mem(memory mem)
%{
  single_instruction;
  mem    : ISS(read);
  INS01  : ISS;
  LDST   : WR;
%}

// Store - reg, mem
// Eg.  STR     x0, <mem>
pipe_class istore_reg_mem(iRegI src, memory mem)
%{
  single_instruction;
  mem    : ISS(read);
  src    : EX2(read);
  INS01  : ISS;
  LDST   : WR;
%}

// Store - reg, reg
// Eg. STR      x0, [sp, x1]
pipe_class istore_reg_reg(iRegI dst, iRegI src)
%{
  single_instruction;
  dst    : ISS(read);
  src    : EX2(read);
  INS01  : ISS;
  LDST   : WR;
%}

//------- Store pipeline operations -----------------------

// Branch
pipe_class pipe_branch()
%{
  single_instruction;
  INS01  : ISS;
  BRANCH : EX1;
%}

// Conditional branch
pipe_class pipe_branch_cond(rFlagsReg cr)
%{
  single_instruction;
  cr     : EX1(read);
  INS01  : ISS;
  BRANCH : EX1;
%}

// Compare & Branch
// EG.  CBZ/CBNZ
pipe_class pipe_cmp_branch(iRegI op1)
%{
  single_instruction;
  op1    : EX1(read);
  INS01  : ISS;
  BRANCH : EX1;
%}

//------- Synchronisation operations ----------------------

// Any operation requiring serialization.
// EG.  DMB/Atomic Ops/Load Acquire/Str Release
pipe_class pipe_serial()
%{
  single_instruction;
  force_serialization;
  fixed_latency(16);
  INS01  : ISS(2); // Cannot dual issue with any other instruction
  LDST   : WR;
%}

// Generic big/slow expanded idiom - also serialized
pipe_class pipe_slow()
%{
  instruction_count(10);
  multiple_bundles;
  force_serialization;
  fixed_latency(16);
  INS01  : ISS(2); // Cannot dual issue with any other instruction
  LDST   : WR;
%}

// Empty pipeline class
pipe_class pipe_class_empty()
%{
  single_instruction;
  fixed_latency(0);
%}

// Default pipeline class.
pipe_class pipe_class_default()
%{
  single_instruction;
  fixed_latency(2);
%}

// Pipeline class for compares.
pipe_class pipe_class_compare()
%{
  single_instruction;
  fixed_latency(16);
%}

// Pipeline class for memory operations.
pipe_class pipe_class_memory()
%{
  single_instruction;
  fixed_latency(16);
%}

// Pipeline class for call.
pipe_class pipe_class_call()
%{
  single_instruction;
  fixed_latency(100);
%}

// Define the class for the Nop node.
define %{
   MachNop = pipe_class_empty;
%}

%}
//----------INSTRUCTIONS-------------------------------------------------------
//
// match      -- States which machine-independent subtree may be replaced
//               by this instruction.
// ins_cost   -- The estimated cost of this instruction is used by instruction
//               selection to identify a minimum cost tree of machine
//               instructions that matches a tree of machine-independent
//               instructions.
// format     -- A string providing the disassembly for this instruction.
//               The value of an instruction's operand may be inserted
//               by referring to it with a '$' prefix.
// opcode     -- Three instruction opcodes may be provided.  These are referred
//               to within an encode class as $primary, $secondary, and $tertiary
//               rrspectively.  The primary opcode is commonly used to
//               indicate the type of machine instruction, while secondary
//               and tertiary are often used for prefix options or addressing
//               modes.
// ins_encode -- A list of encode classes with parameters. The encode class
//               name must have been defined in an 'enc_class' specification
//               in the encode section of the architecture description.

// ============================================================================
// Memory (Load/Store) Instructions

// Load Instructions

// Load Byte (8 bit signed)
instruct loadB(iRegINoSp dst, memory mem)
%{
  match(Set dst (LoadB mem));
  predicate(!needs_acquiring_load(n));

  ins_cost(4 * INSN_COST);
  format %{ "ldrsbw  $dst, $mem\t# byte" %}

  ins_encode(aarch64_enc_ldrsbw(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Byte (8 bit signed) into long
instruct loadB2L(iRegLNoSp dst, memory mem)
%{
  match(Set dst (ConvI2L (LoadB mem)));
  predicate(!needs_acquiring_load(n->in(1)));

  ins_cost(4 * INSN_COST);
  format %{ "ldrsb  $dst, $mem\t# byte" %}

  ins_encode(aarch64_enc_ldrsb(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Byte (8 bit unsigned)
instruct loadUB(iRegINoSp dst, memory mem)
%{
  match(Set dst (LoadUB mem));
  predicate(!needs_acquiring_load(n));

  ins_cost(4 * INSN_COST);
  format %{ "ldrbw  $dst, $mem\t# byte" %}

  ins_encode(aarch64_enc_ldrb(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Byte (8 bit unsigned) into long
instruct loadUB2L(iRegLNoSp dst, memory mem)
%{
  match(Set dst (ConvI2L (LoadUB mem)));
  predicate(!needs_acquiring_load(n->in(1)));

  ins_cost(4 * INSN_COST);
  format %{ "ldrb  $dst, $mem\t# byte" %}

  ins_encode(aarch64_enc_ldrb(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Short (16 bit signed)
instruct loadS(iRegINoSp dst, memory mem)
%{
  match(Set dst (LoadS mem));
  predicate(!needs_acquiring_load(n));

  ins_cost(4 * INSN_COST);
  format %{ "ldrshw  $dst, $mem\t# short" %}

  ins_encode(aarch64_enc_ldrshw(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Short (16 bit signed) into long
instruct loadS2L(iRegLNoSp dst, memory mem)
%{
  match(Set dst (ConvI2L (LoadS mem)));
  predicate(!needs_acquiring_load(n->in(1)));

  ins_cost(4 * INSN_COST);
  format %{ "ldrsh  $dst, $mem\t# short" %}

  ins_encode(aarch64_enc_ldrsh(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Char (16 bit unsigned)
instruct loadUS(iRegINoSp dst, memory mem)
%{
  match(Set dst (LoadUS mem));
  predicate(!needs_acquiring_load(n));

  ins_cost(4 * INSN_COST);
  format %{ "ldrh  $dst, $mem\t# short" %}

  ins_encode(aarch64_enc_ldrh(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Short/Char (16 bit unsigned) into long
instruct loadUS2L(iRegLNoSp dst, memory mem)
%{
  match(Set dst (ConvI2L (LoadUS mem)));
  predicate(!needs_acquiring_load(n->in(1)));

  ins_cost(4 * INSN_COST);
  format %{ "ldrh  $dst, $mem\t# short" %}

  ins_encode(aarch64_enc_ldrh(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Integer (32 bit signed)
instruct loadI(iRegINoSp dst, memory mem)
%{
  match(Set dst (LoadI mem));
  predicate(!needs_acquiring_load(n));

  ins_cost(4 * INSN_COST);
  format %{ "ldrw  $dst, $mem\t# int" %}

  ins_encode(aarch64_enc_ldrw(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Integer (32 bit signed) into long
instruct loadI2L(iRegLNoSp dst, memory mem)
%{
  match(Set dst (ConvI2L (LoadI mem)));
  predicate(!needs_acquiring_load(n->in(1)));

  ins_cost(4 * INSN_COST);
  format %{ "ldrsw  $dst, $mem\t# int" %}

  ins_encode(aarch64_enc_ldrsw(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Integer (32 bit unsigned) into long
instruct loadUI2L(iRegLNoSp dst, memory mem, immL_32bits mask)
%{
  match(Set dst (AndL (ConvI2L (LoadI mem)) mask));
  predicate(!needs_acquiring_load(n->in(1)->in(1)->as_Load()));

  ins_cost(4 * INSN_COST);
  format %{ "ldrw  $dst, $mem\t# int" %}

  ins_encode(aarch64_enc_ldrw(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Long (64 bit signed)
instruct loadL(iRegLNoSp dst, memory mem)
%{
  match(Set dst (LoadL mem));
  predicate(!needs_acquiring_load(n));

  ins_cost(4 * INSN_COST);
  format %{ "ldr  $dst, $mem\t# int" %}

  ins_encode(aarch64_enc_ldr(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Range
instruct loadRange(iRegINoSp dst, memory mem)
%{
  match(Set dst (LoadRange mem));

  ins_cost(4 * INSN_COST);
  format %{ "ldrw  $dst, $mem\t# range" %}

  ins_encode(aarch64_enc_ldrw(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Pointer
instruct loadP(iRegPNoSp dst, memory mem)
%{
  match(Set dst (LoadP mem));
  predicate(!needs_acquiring_load(n));

  ins_cost(4 * INSN_COST);
  format %{ "ldr  $dst, $mem\t# ptr" %}

  ins_encode(aarch64_enc_ldr(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Compressed Pointer
instruct loadN(iRegNNoSp dst, memory mem)
%{
  match(Set dst (LoadN mem));
  predicate(!needs_acquiring_load(n));

  ins_cost(4 * INSN_COST);
  format %{ "ldrw  $dst, $mem\t# compressed ptr" %}

  ins_encode(aarch64_enc_ldrw(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Klass Pointer
instruct loadKlass(iRegPNoSp dst, memory mem)
%{
  match(Set dst (LoadKlass mem));
  predicate(!needs_acquiring_load(n));

  ins_cost(4 * INSN_COST);
  format %{ "ldr  $dst, $mem\t# class" %}

  ins_encode(aarch64_enc_ldr(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Narrow Klass Pointer
instruct loadNKlass(iRegNNoSp dst, memory mem)
%{
  match(Set dst (LoadNKlass mem));
  predicate(!needs_acquiring_load(n));

  ins_cost(4 * INSN_COST);
  format %{ "ldrw  $dst, $mem\t# compressed class ptr" %}

  ins_encode(aarch64_enc_ldrw(dst, mem));

  ins_pipe(iload_reg_mem);
%}

// Load Float
instruct loadF(vRegF dst, memory mem)
%{
  match(Set dst (LoadF mem));
  predicate(!needs_acquiring_load(n));

  ins_cost(4 * INSN_COST);
  format %{ "ldrs  $dst, $mem\t# float" %}

  ins_encode( aarch64_enc_ldrs(dst, mem) );

  ins_pipe(pipe_class_memory);
%}

// Load Double
instruct loadD(vRegD dst, memory mem)
%{
  match(Set dst (LoadD mem));
  predicate(!needs_acquiring_load(n));

  ins_cost(4 * INSN_COST);
  format %{ "ldrd  $dst, $mem\t# double" %}

  ins_encode( aarch64_enc_ldrd(dst, mem) );

  ins_pipe(pipe_class_memory);
%}


// Load Int Constant
instruct loadConI(iRegINoSp dst, immI src)
%{
  match(Set dst src);

  ins_cost(INSN_COST);
  format %{ "mov $dst, $src\t# int" %}

  ins_encode( aarch64_enc_movw_imm(dst, src) );

  ins_pipe(ialu_imm);
%}

// Load Long Constant
instruct loadConL(iRegLNoSp dst, immL src)
%{
  match(Set dst src);

  ins_cost(INSN_COST);
  format %{ "mov $dst, $src\t# long" %}

  ins_encode( aarch64_enc_mov_imm(dst, src) );

  ins_pipe(ialu_imm);
%}

// Load Pointer Constant

instruct loadConP(iRegPNoSp dst, immP con)
%{
  match(Set dst con);

  ins_cost(INSN_COST * 4);
  format %{
    "mov  $dst, $con\t# ptr\n\t"
  %}

  ins_encode(aarch64_enc_mov_p(dst, con));

  ins_pipe(ialu_imm);
%}

// Load Null Pointer Constant

instruct loadConP0(iRegPNoSp dst, immP0 con)
%{
  match(Set dst con);

  ins_cost(INSN_COST);
  format %{ "mov  $dst, $con\t# NULL ptr" %}

  ins_encode(aarch64_enc_mov_p0(dst, con));

  ins_pipe(ialu_imm);
%}

// Load Pointer Constant One

instruct loadConP1(iRegPNoSp dst, immP_1 con)
%{
  match(Set dst con);

  ins_cost(INSN_COST);
  format %{ "mov  $dst, $con\t# NULL ptr" %}

  ins_encode(aarch64_enc_mov_p1(dst, con));

  ins_pipe(ialu_imm);
%}

// Load Poll Page Constant

instruct loadConPollPage(iRegPNoSp dst, immPollPage con)
%{
  match(Set dst con);

  ins_cost(INSN_COST);
  format %{ "adr  $dst, $con\t# Poll Page Ptr" %}

  ins_encode(aarch64_enc_mov_poll_page(dst, con));

  ins_pipe(ialu_imm);
%}

// Load Byte Map Base Constant

instruct loadByteMapBase(iRegPNoSp dst, immByteMapBase con)
%{
  match(Set dst con);

  ins_cost(INSN_COST);
  format %{ "adr  $dst, $con\t# Byte Map Base" %}

  ins_encode(aarch64_enc_mov_byte_map_base(dst, con));

  ins_pipe(ialu_imm);
%}

// Load Narrow Pointer Constant

instruct loadConN(iRegNNoSp dst, immN con)
%{
  match(Set dst con);

  ins_cost(INSN_COST * 4);
  format %{ "mov  $dst, $con\t# compressed ptr" %}

  ins_encode(aarch64_enc_mov_n(dst, con));

  ins_pipe(ialu_imm);
%}

// Load Narrow Null Pointer Constant

instruct loadConN0(iRegNNoSp dst, immN0 con)
%{
  match(Set dst con);

  ins_cost(INSN_COST);
  format %{ "mov  $dst, $con\t# compressed NULL ptr" %}

  ins_encode(aarch64_enc_mov_n0(dst, con));

  ins_pipe(ialu_imm);
%}

// Load Narrow Klass Constant

instruct loadConNKlass(iRegNNoSp dst, immNKlass con)
%{
  match(Set dst con);

  ins_cost(INSN_COST);
  format %{ "mov  $dst, $con\t# compressed klass ptr" %}

  ins_encode(aarch64_enc_mov_nk(dst, con));

  ins_pipe(ialu_imm);
%}

// Load Packed Float Constant

instruct loadConF_packed(vRegF dst, immFPacked con) %{
  match(Set dst con);
  ins_cost(INSN_COST * 4);
  format %{ "fmovs  $dst, $con"%}
  ins_encode %{
    __ fmovs(as_FloatRegister($dst$$reg), (double)$con$$constant);
  %}

  ins_pipe(fp_imm_s);
%}

// Load Float Constant

instruct loadConF(vRegF dst, immF con) %{
  match(Set dst con);

  ins_cost(INSN_COST * 4);

  format %{
    "ldrs $dst, [$constantaddress]\t# load from constant table: float=$con\n\t"
  %}

  ins_encode %{
    __ ldrs(as_FloatRegister($dst$$reg), $constantaddress($con));
  %}

  ins_pipe(fp_load_constant_s);
%}

// Load Packed Double Constant

instruct loadConD_packed(vRegD dst, immDPacked con) %{
  match(Set dst con);
  ins_cost(INSN_COST);
  format %{ "fmovd  $dst, $con"%}
  ins_encode %{
    __ fmovd(as_FloatRegister($dst$$reg), $con$$constant);
  %}

  ins_pipe(fp_imm_d);
%}

// Load Double Constant

instruct loadConD(vRegD dst, immD con) %{
  match(Set dst con);

  ins_cost(INSN_COST * 5);
  format %{
    "ldrd $dst, [$constantaddress]\t# load from constant table: float=$con\n\t"
  %}

  ins_encode %{
    __ ldrd(as_FloatRegister($dst$$reg), $constantaddress($con));
  %}

  ins_pipe(fp_load_constant_d);
%}

// Store Instructions

// Store CMS card-mark Immediate
instruct storeimmCM0(immI0 zero, memory mem)
%{
  match(Set mem (StoreCM mem zero));
  predicate(unnecessary_storestore(n));

  ins_cost(INSN_COST);
  format %{ "strb zr, $mem\t# byte" %}

  ins_encode(aarch64_enc_strb0(mem));

  ins_pipe(istore_mem);
%}

// Store CMS card-mark Immediate with intervening StoreStore
// needed when using CMS with no conditional card marking
instruct storeimmCM0_ordered(immI0 zero, memory mem)
%{
  match(Set mem (StoreCM mem zero));

  ins_cost(INSN_COST * 2);
  format %{ "dmb ishst"
      "\n\tstrb zr, $mem\t# byte" %}

  ins_encode(aarch64_enc_strb0_ordered(mem));

  ins_pipe(istore_mem);
%}

// Store Byte
instruct storeB(iRegIorL2I src, memory mem)
%{
  match(Set mem (StoreB mem src));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "strb  $src, $mem\t# byte" %}

  ins_encode(aarch64_enc_strb(src, mem));

  ins_pipe(istore_reg_mem);
%}


instruct storeimmB0(immI0 zero, memory mem)
%{
  match(Set mem (StoreB mem zero));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "strb rscractch2, $mem\t# byte" %}

  ins_encode(aarch64_enc_strb0(mem));

  ins_pipe(istore_mem);
%}

// Store Char/Short
instruct storeC(iRegIorL2I src, memory mem)
%{
  match(Set mem (StoreC mem src));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "strh  $src, $mem\t# short" %}

  ins_encode(aarch64_enc_strh(src, mem));

  ins_pipe(istore_reg_mem);
%}

instruct storeimmC0(immI0 zero, memory mem)
%{
  match(Set mem (StoreC mem zero));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "strh  zr, $mem\t# short" %}

  ins_encode(aarch64_enc_strh0(mem));

  ins_pipe(istore_mem);
%}

// Store Integer

instruct storeI(iRegIorL2I src, memory mem)
%{
  match(Set mem(StoreI mem src));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "strw  $src, $mem\t# int" %}

  ins_encode(aarch64_enc_strw(src, mem));

  ins_pipe(istore_reg_mem);
%}

instruct storeimmI0(immI0 zero, memory mem)
%{
  match(Set mem(StoreI mem zero));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "strw  zr, $mem\t# int" %}

  ins_encode(aarch64_enc_strw0(mem));

  ins_pipe(istore_mem);
%}

// Store Long (64 bit signed)
instruct storeL(iRegL src, memory mem)
%{
  match(Set mem (StoreL mem src));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "str  $src, $mem\t# int" %}

  ins_encode(aarch64_enc_str(src, mem));

  ins_pipe(istore_reg_mem);
%}

// Store Long (64 bit signed)
instruct storeimmL0(immL0 zero, memory mem)
%{
  match(Set mem (StoreL mem zero));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "str  zr, $mem\t# int" %}

  ins_encode(aarch64_enc_str0(mem));

  ins_pipe(istore_mem);
%}

// Store Pointer
instruct storeP(iRegP src, memory mem)
%{
  match(Set mem (StoreP mem src));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "str  $src, $mem\t# ptr" %}

  ins_encode(aarch64_enc_str(src, mem));

  ins_pipe(istore_reg_mem);
%}

// Store Pointer
instruct storeimmP0(immP0 zero, memory mem)
%{
  match(Set mem (StoreP mem zero));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "str zr, $mem\t# ptr" %}

  ins_encode(aarch64_enc_str0(mem));

  ins_pipe(istore_mem);
%}

// Store Compressed Pointer
instruct storeN(iRegN src, memory mem)
%{
  match(Set mem (StoreN mem src));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "strw  $src, $mem\t# compressed ptr" %}

  ins_encode(aarch64_enc_strw(src, mem));

  ins_pipe(istore_reg_mem);
%}

instruct storeImmN0(iRegIHeapbase heapbase, immN0 zero, memory mem)
%{
  match(Set mem (StoreN mem zero));
  predicate(Universe::narrow_oop_base() == NULL &&
            Universe::narrow_klass_base() == NULL &&
            (!needs_releasing_store(n)));

  ins_cost(INSN_COST);
  format %{ "strw  rheapbase, $mem\t# compressed ptr (rheapbase==0)" %}

  ins_encode(aarch64_enc_strw(heapbase, mem));

  ins_pipe(istore_reg_mem);
%}

// Store Float
instruct storeF(vRegF src, memory mem)
%{
  match(Set mem (StoreF mem src));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "strs  $src, $mem\t# float" %}

  ins_encode( aarch64_enc_strs(src, mem) );

  ins_pipe(pipe_class_memory);
%}

// TODO
// implement storeImmF0 and storeFImmPacked

// Store Double
instruct storeD(vRegD src, memory mem)
%{
  match(Set mem (StoreD mem src));
  predicate(!needs_releasing_store(n));

  ins_cost(INSN_COST);
  format %{ "strd  $src, $mem\t# double" %}

  ins_encode( aarch64_enc_strd(src, mem) );

  ins_pipe(pipe_class_memory);
%}

// Store Compressed Klass Pointer
instruct storeNKlass(iRegN src, memory mem)
%{
  predicate(!needs_releasing_store(n));
  match(Set mem (StoreNKlass mem src));

  ins_cost(INSN_COST);
  format %{ "strw  $src, $mem\t# compressed klass ptr" %}

  ins_encode(aarch64_enc_strw(src, mem));

  ins_pipe(istore_reg_mem);
%}

// TODO
// implement storeImmD0 and storeDImmPacked

// prefetch instructions
// Must be safe to execute with invalid address (cannot fault).

instruct prefetchalloc( memory mem ) %{
  match(PrefetchAllocation mem);

  ins_cost(INSN_COST);
  format %{ "prfm $mem, PSTL1KEEP\t# Prefetch into level 1 cache write keep" %}

  ins_encode( aarch64_enc_prefetchw(mem) );

  ins_pipe(iload_prefetch);
%}

//  ---------------- volatile loads and stores ----------------

// Load Byte (8 bit signed)
instruct loadB_volatile(iRegINoSp dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (LoadB mem));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldarsb  $dst, $mem\t# byte" %}

  ins_encode(aarch64_enc_ldarsb(dst, mem));

  ins_pipe(pipe_serial);
%}

// Load Byte (8 bit signed) into long
instruct loadB2L_volatile(iRegLNoSp dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (ConvI2L (LoadB mem)));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldarsb  $dst, $mem\t# byte" %}

  ins_encode(aarch64_enc_ldarsb(dst, mem));

  ins_pipe(pipe_serial);
%}

// Load Byte (8 bit unsigned)
instruct loadUB_volatile(iRegINoSp dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (LoadUB mem));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldarb  $dst, $mem\t# byte" %}

  ins_encode(aarch64_enc_ldarb(dst, mem));

  ins_pipe(pipe_serial);
%}

// Load Byte (8 bit unsigned) into long
instruct loadUB2L_volatile(iRegLNoSp dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (ConvI2L (LoadUB mem)));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldarb  $dst, $mem\t# byte" %}

  ins_encode(aarch64_enc_ldarb(dst, mem));

  ins_pipe(pipe_serial);
%}

// Load Short (16 bit signed)
instruct loadS_volatile(iRegINoSp dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (LoadS mem));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldarshw  $dst, $mem\t# short" %}

  ins_encode(aarch64_enc_ldarshw(dst, mem));

  ins_pipe(pipe_serial);
%}

instruct loadUS_volatile(iRegINoSp dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (LoadUS mem));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldarhw  $dst, $mem\t# short" %}

  ins_encode(aarch64_enc_ldarhw(dst, mem));

  ins_pipe(pipe_serial);
%}

// Load Short/Char (16 bit unsigned) into long
instruct loadUS2L_volatile(iRegLNoSp dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (ConvI2L (LoadUS mem)));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldarh  $dst, $mem\t# short" %}

  ins_encode(aarch64_enc_ldarh(dst, mem));

  ins_pipe(pipe_serial);
%}

// Load Short/Char (16 bit signed) into long
instruct loadS2L_volatile(iRegLNoSp dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (ConvI2L (LoadS mem)));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldarh  $dst, $mem\t# short" %}

  ins_encode(aarch64_enc_ldarsh(dst, mem));

  ins_pipe(pipe_serial);
%}

// Load Integer (32 bit signed)
instruct loadI_volatile(iRegINoSp dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (LoadI mem));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldarw  $dst, $mem\t# int" %}

  ins_encode(aarch64_enc_ldarw(dst, mem));

  ins_pipe(pipe_serial);
%}

// Load Integer (32 bit unsigned) into long
instruct loadUI2L_volatile(iRegLNoSp dst, /* sync_memory*/indirect mem, immL_32bits mask)
%{
  match(Set dst (AndL (ConvI2L (LoadI mem)) mask));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldarw  $dst, $mem\t# int" %}

  ins_encode(aarch64_enc_ldarw(dst, mem));

  ins_pipe(pipe_serial);
%}

// Load Long (64 bit signed)
instruct loadL_volatile(iRegLNoSp dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (LoadL mem));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldar  $dst, $mem\t# int" %}

  ins_encode(aarch64_enc_ldar(dst, mem));

  ins_pipe(pipe_serial);
%}

// Load Pointer
instruct loadP_volatile(iRegPNoSp dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (LoadP mem));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldar  $dst, $mem\t# ptr" %}

  ins_encode(aarch64_enc_ldar(dst, mem));

  ins_pipe(pipe_serial);
%}

// Load Compressed Pointer
instruct loadN_volatile(iRegNNoSp dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (LoadN mem));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldarw  $dst, $mem\t# compressed ptr" %}

  ins_encode(aarch64_enc_ldarw(dst, mem));

  ins_pipe(pipe_serial);
%}

// Load Float
instruct loadF_volatile(vRegF dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (LoadF mem));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldars  $dst, $mem\t# float" %}

  ins_encode( aarch64_enc_fldars(dst, mem) );

  ins_pipe(pipe_serial);
%}

// Load Double
instruct loadD_volatile(vRegD dst, /* sync_memory*/indirect mem)
%{
  match(Set dst (LoadD mem));

  ins_cost(VOLATILE_REF_COST);
  format %{ "ldard  $dst, $mem\t# double" %}

  ins_encode( aarch64_enc_fldard(dst, mem) );

  ins_pipe(pipe_serial);
%}

// Store Byte
instruct storeB_volatile(iRegIorL2I src, /* sync_memory*/indirect mem)
%{
  match(Set mem (StoreB mem src));

  ins_cost(VOLATILE_REF_COST);
  format %{ "stlrb  $src, $mem\t# byte" %}

  ins_encode(aarch64_enc_stlrb(src, mem));

  ins_pipe(pipe_class_memory);
%}

// Store Char/Short
instruct storeC_volatile(iRegIorL2I src, /* sync_memory*/indirect mem)
%{
  match(Set mem (StoreC mem src));

  ins_cost(VOLATILE_REF_COST);
  format %{ "stlrh  $src, $mem\t# short" %}

  ins_encode(aarch64_enc_stlrh(src, mem));

  ins_pipe(pipe_class_memory);
%}

// Store Integer

instruct storeI_volatile(iRegIorL2I src, /* sync_memory*/indirect mem)
%{
  match(Set mem(StoreI mem src));

  ins_cost(VOLATILE_REF_COST);
  format %{ "stlrw  $src, $mem\t# int" %}

  ins_encode(aarch64_enc_stlrw(src, mem));

  ins_pipe(pipe_class_memory);
%}

// Store Long (64 bit signed)
instruct storeL_volatile(iRegL src, /* sync_memory*/indirect mem)
%{
  match(Set mem (StoreL mem src));

  ins_cost(VOLATILE_REF_COST);
  format %{ "stlr  $src, $mem\t# int" %}

  ins_encode(aarch64_enc_stlr(src, mem));

  ins_pipe(pipe_class_memory);
%}

// Store Pointer
instruct storeP_volatile(iRegP src, /* sync_memory*/indirect mem)
%{
  match(Set mem (StoreP mem src));

  ins_cost(VOLATILE_REF_COST);
  format %{ "stlr  $src, $mem\t# ptr" %}

  ins_encode(aarch64_enc_stlr(src, mem));

  ins_pipe(pipe_class_memory);
%}

// Store Compressed Pointer
instruct storeN_volatile(iRegN src, /* sync_memory*/indirect mem)
%{
  match(Set mem (StoreN mem src));

  ins_cost(VOLATILE_REF_COST);
  format %{ "stlrw  $src, $mem\t# compressed ptr" %}

  ins_encode(aarch64_enc_stlrw(src, mem));

  ins_pipe(pipe_class_memory);
%}

// Store Float
instruct storeF_volatile(vRegF src, /* sync_memory*/indirect mem)
%{
  match(Set mem (StoreF mem src));

  ins_cost(VOLATILE_REF_COST);
  format %{ "stlrs  $src, $mem\t# float" %}

  ins_encode( aarch64_enc_fstlrs(src, mem) );

  ins_pipe(pipe_class_memory);
%}

// TODO
// implement storeImmF0 and storeFImmPacked

// Store Double
instruct storeD_volatile(vRegD src, /* sync_memory*/indirect mem)
%{
  match(Set mem (StoreD mem src));

  ins_cost(VOLATILE_REF_COST);
  format %{ "stlrd  $src, $mem\t# double" %}

  ins_encode( aarch64_enc_fstlrd(src, mem) );

  ins_pipe(pipe_class_memory);
%}

//  ---------------- end of volatile loads and stores ----------------

// ============================================================================
// BSWAP Instructions

instruct bytes_reverse_int(iRegINoSp dst, iRegIorL2I src) %{
  match(Set dst (ReverseBytesI src));

  ins_cost(INSN_COST);
  format %{ "revw  $dst, $src" %}

  ins_encode %{
    __ revw(as_Register($dst$$reg), as_Register($src$$reg));
  %}

  ins_pipe(ialu_reg);
%}

instruct bytes_reverse_long(iRegLNoSp dst, iRegL src) %{
  match(Set dst (ReverseBytesL src));

  ins_cost(INSN_COST);
  format %{ "rev  $dst, $src" %}

  ins_encode %{
    __ rev(as_Register($dst$$reg), as_Register($src$$reg));
  %}

  ins_pipe(ialu_reg);
%}

instruct bytes_reverse_unsigned_short(iRegINoSp dst, iRegIorL2I src) %{
  match(Set dst (ReverseBytesUS src));

  ins_cost(INSN_COST);
  format %{ "rev16w  $dst, $src" %}

  ins_encode %{
    __ rev16w(as_Register($dst$$reg), as_Register($src$$reg));
  %}

  ins_pipe(ialu_reg);
%}

instruct bytes_reverse_short(iRegINoSp dst, iRegIorL2I src) %{
  match(Set dst (ReverseBytesS src));

  ins_cost(INSN_COST);
  format %{ "rev16w  $dst, $src\n\t"
            "sbfmw $dst, $dst, #0, #15" %}

  ins_encode %{
    __ rev16w(as_Register($dst$$reg), as_Register($src$$reg));
    __ sbfmw(as_Register($dst$$reg), as_Register($dst$$reg), 0U, 15U);
  %}

  ins_pipe(ialu_reg);
%}

// ============================================================================
// Zero Count Instructions

instruct countLeadingZerosI(iRegINoSp dst, iRegIorL2I src) %{
  match(Set dst (CountLeadingZerosI src));

  ins_cost(INSN_COST);
  format %{ "clzw  $dst, $src" %}
  ins_encode %{
    __ clzw(as_Register($dst$$reg), as_Register($src$$reg));
  %}

  ins_pipe(ialu_reg);
%}

instruct countLeadingZerosL(iRegINoSp dst, iRegL src) %{
  match(Set dst (CountLeadingZerosL src));

  ins_cost(INSN_COST);
  format %{ "clz   $dst, $src" %}
  ins_encode %{
    __ clz(as_Register($dst$$reg), as_Register($src$$reg));
  %}

  ins_pipe(ialu_reg);
%}

instruct countTrailingZerosI(iRegINoSp dst, iRegIorL2I src) %{
  match(Set dst (CountTrailingZerosI src));

  ins_cost(INSN_COST * 2);
  format %{ "rbitw  $dst, $src\n\t"
            "clzw   $dst, $dst" %}
  ins_encode %{
    __ rbitw(as_Register($dst$$reg), as_Register($src$$reg));
    __ clzw(as_Register($dst$$reg), as_Register($dst$$reg));
  %}

  ins_pipe(ialu_reg);
%}

instruct countTrailingZerosL(iRegINoSp dst, iRegL src) %{
  match(Set dst (CountTrailingZerosL src));

  ins_cost(INSN_COST * 2);
  format %{ "rbit   $dst, $src\n\t"
            "clz    $dst, $dst" %}
  ins_encode %{
    __ rbit(as_Register($dst$$reg), as_Register($src$$reg));
    __ clz(as_Register($dst$$reg), as_Register($dst$$reg));
  %}

  ins_pipe(ialu_reg);
%}

//---------- Population Count Instructions -------------------------------------
//

instruct popCountI(iRegINoSp dst, iRegIorL2I src, vRegF tmp) %{
  predicate(UsePopCountInstruction);
  match(Set dst (PopCountI src));
  effect(TEMP tmp);
  ins_cost(INSN_COST * 13);

  format %{ "movw   $src, $src\n\t"
            "mov    $tmp, $src\t# vector (1D)\n\t"
            "cnt    $tmp, $tmp\t# vector (8B)\n\t"
            "addv   $tmp, $tmp\t# vector (8B)\n\t"
            "mov    $dst, $tmp\t# vector (1D)" %}
  ins_encode %{
    __ movw($src$$Register, $src$$Register); // ensure top 32 bits 0
    __ mov($tmp$$FloatRegister, __ T1D, 0, $src$$Register);
    __ cnt($tmp$$FloatRegister, __ T8B, $tmp$$FloatRegister);
    __ addv($tmp$$FloatRegister, __ T8B, $tmp$$FloatRegister);
    __ mov($dst$$Register, $tmp$$FloatRegister, __ T1D, 0);
  %}

  ins_pipe(pipe_class_default);
%}

instruct popCountI_mem(iRegINoSp dst, memory mem, vRegF tmp) %{
  predicate(UsePopCountInstruction);
  match(Set dst (PopCountI (LoadI mem)));
  effect(TEMP tmp);
  ins_cost(INSN_COST * 13);

  format %{ "ldrs   $tmp, $mem\n\t"
            "cnt    $tmp, $tmp\t# vector (8B)\n\t"
            "addv   $tmp, $tmp\t# vector (8B)\n\t"
            "mov    $dst, $tmp\t# vector (1D)" %}
  ins_encode %{
    FloatRegister tmp_reg = as_FloatRegister($tmp$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrs, tmp_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
    __ cnt($tmp$$FloatRegister, __ T8B, $tmp$$FloatRegister);
    __ addv($tmp$$FloatRegister, __ T8B, $tmp$$FloatRegister);
    __ mov($dst$$Register, $tmp$$FloatRegister, __ T1D, 0);
  %}

  ins_pipe(pipe_class_default);
%}

// Note: Long.bitCount(long) returns an int.
instruct popCountL(iRegINoSp dst, iRegL src, vRegD tmp) %{
  predicate(UsePopCountInstruction);
  match(Set dst (PopCountL src));
  effect(TEMP tmp);
  ins_cost(INSN_COST * 13);

  format %{ "mov    $tmp, $src\t# vector (1D)\n\t"
            "cnt    $tmp, $tmp\t# vector (8B)\n\t"
            "addv   $tmp, $tmp\t# vector (8B)\n\t"
            "mov    $dst, $tmp\t# vector (1D)" %}
  ins_encode %{
    __ mov($tmp$$FloatRegister, __ T1D, 0, $src$$Register);
    __ cnt($tmp$$FloatRegister, __ T8B, $tmp$$FloatRegister);
    __ addv($tmp$$FloatRegister, __ T8B, $tmp$$FloatRegister);
    __ mov($dst$$Register, $tmp$$FloatRegister, __ T1D, 0);
  %}

  ins_pipe(pipe_class_default);
%}

instruct popCountL_mem(iRegINoSp dst, memory mem, vRegD tmp) %{
  predicate(UsePopCountInstruction);
  match(Set dst (PopCountL (LoadL mem)));
  effect(TEMP tmp);
  ins_cost(INSN_COST * 13);

  format %{ "ldrd   $tmp, $mem\n\t"
            "cnt    $tmp, $tmp\t# vector (8B)\n\t"
            "addv   $tmp, $tmp\t# vector (8B)\n\t"
            "mov    $dst, $tmp\t# vector (1D)" %}
  ins_encode %{
    FloatRegister tmp_reg = as_FloatRegister($tmp$$reg);
    loadStore(MacroAssembler(&cbuf), &MacroAssembler::ldrd, tmp_reg, $mem->opcode(),
               as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
    __ cnt($tmp$$FloatRegister, __ T8B, $tmp$$FloatRegister);
    __ addv($tmp$$FloatRegister, __ T8B, $tmp$$FloatRegister);
    __ mov($dst$$Register, $tmp$$FloatRegister, __ T1D, 0);
  %}

  ins_pipe(pipe_class_default);
%}

// ============================================================================
// MemBar Instruction

instruct load_fence() %{
  match(LoadFence);
  ins_cost(VOLATILE_REF_COST);

  format %{ "load_fence" %}

  ins_encode %{
    __ membar(Assembler::LoadLoad|Assembler::LoadStore);
  %}
  ins_pipe(pipe_serial);
%}

instruct unnecessary_membar_acquire() %{
  predicate(unnecessary_acquire(n));
  match(MemBarAcquire);
  ins_cost(0);

  format %{ "membar_acquire (elided)" %}

  ins_encode %{
    __ block_comment("membar_acquire (elided)");
  %}

  ins_pipe(pipe_class_empty);
%}

instruct membar_acquire() %{
  match(MemBarAcquire);
  ins_cost(VOLATILE_REF_COST);

  format %{ "membar_acquire" %}

  ins_encode %{
    __ block_comment("membar_acquire");
    __ membar(Assembler::LoadLoad|Assembler::LoadStore);
  %}

  ins_pipe(pipe_serial);
%}


instruct membar_acquire_lock() %{
  match(MemBarAcquireLock);
  ins_cost(VOLATILE_REF_COST);

  format %{ "membar_acquire_lock (elided)" %}

  ins_encode %{
    __ block_comment("membar_acquire_lock (elided)");
  %}

  ins_pipe(pipe_serial);
%}

instruct store_fence() %{
  match(StoreFence);
  ins_cost(VOLATILE_REF_COST);

  format %{ "store_fence" %}

  ins_encode %{
    __ membar(Assembler::LoadStore|Assembler::StoreStore);
  %}
  ins_pipe(pipe_serial);
%}

instruct unnecessary_membar_release() %{
  predicate(unnecessary_release(n));
  match(MemBarRelease);
  ins_cost(0);

  format %{ "membar_release (elided)" %}

  ins_encode %{
    __ block_comment("membar_release (elided)");
  %}
  ins_pipe(pipe_serial);
%}

instruct membar_release() %{
  match(MemBarRelease);
  ins_cost(VOLATILE_REF_COST);

  format %{ "membar_release" %}

  ins_encode %{
    __ block_comment("membar_release");
    __ membar(Assembler::LoadStore|Assembler::StoreStore);
  %}
  ins_pipe(pipe_serial);
%}

instruct membar_storestore() %{
  match(MemBarStoreStore);
  ins_cost(VOLATILE_REF_COST);

  format %{ "MEMBAR-store-store" %}

  ins_encode %{
    __ membar(Assembler::StoreStore);
  %}
  ins_pipe(pipe_serial);
%}

instruct membar_release_lock() %{
  match(MemBarReleaseLock);
  ins_cost(VOLATILE_REF_COST);

  format %{ "membar_release_lock (elided)" %}

  ins_encode %{
    __ block_comment("membar_release_lock (elided)");
  %}

  ins_pipe(pipe_serial);
%}

instruct unnecessary_membar_volatile() %{
  predicate(unnecessary_volatile(n));
  match(MemBarVolatile);
  ins_cost(0);

  format %{ "membar_volatile (elided)" %}

  ins_encode %{
    __ block_comment("membar_volatile (elided)");
  %}

  ins_pipe(pipe_serial);
%}

instruct membar_volatile() %{
  match(MemBarVolatile);
  ins_cost(VOLATILE_REF_COST*100);

  format %{ "membar_volatile" %}

  ins_encode %{
    __ block_comment("membar_volatile");
    __ membar(Assembler::StoreLoad);
  %}

  ins_pipe(pipe_serial);
%}

// ============================================================================
// Cast/Convert Instructions

instruct castX2P(iRegPNoSp dst, iRegL src) %{
  match(Set dst (CastX2P src));

  ins_cost(INSN_COST);
  format %{ "mov $dst, $src\t# long -> ptr" %}

  ins_encode %{
    if ($dst$$reg != $src$$reg) {
      __ mov(as_Register($dst$$reg), as_Register($src$$reg));
    }
  %}

  ins_pipe(ialu_reg);
%}

instruct castP2X(iRegLNoSp dst, iRegP src) %{
  match(Set dst (CastP2X src));

  ins_cost(INSN_COST);
  format %{ "mov $dst, $src\t# ptr -> long" %}

  ins_encode %{
    if ($dst$$reg != $src$$reg) {
      __ mov(as_Register($dst$$reg), as_Register($src$$reg));
    }
  %}

  ins_pipe(ialu_reg);
%}

// Convert oop into int for vectors alignment masking
instruct convP2I(iRegINoSp dst, iRegP src) %{
  match(Set dst (ConvL2I (CastP2X src)));

  ins_cost(INSN_COST);
  format %{ "movw $dst, $src\t# ptr -> int" %}
  ins_encode %{
    __ movw($dst$$Register, $src$$Register);
  %}

  ins_pipe(ialu_reg);
%}

// Convert compressed oop into int for vectors alignment masking
// in case of 32bit oops (heap < 4Gb).
instruct convN2I(iRegINoSp dst, iRegN src)
%{
  predicate(Universe::narrow_oop_shift() == 0);
  match(Set dst (ConvL2I (CastP2X (DecodeN src))));

  ins_cost(INSN_COST);
  format %{ "mov dst, $src\t# compressed ptr -> int" %}
  ins_encode %{
    __ movw($dst$$Register, $src$$Register);
  %}

  ins_pipe(ialu_reg);
%}


// Convert oop pointer into compressed form
instruct encodeHeapOop(iRegNNoSp dst, iRegP src, rFlagsReg cr) %{
  predicate(n->bottom_type()->make_ptr()->ptr() != TypePtr::NotNull);
  match(Set dst (EncodeP src));
  effect(KILL cr);
  ins_cost(INSN_COST * 3);
  format %{ "encode_heap_oop $dst, $src" %}
  ins_encode %{
    Register s = $src$$Register;
    Register d = $dst$$Register;
    __ encode_heap_oop(d, s);
  %}
  ins_pipe(ialu_reg);
%}

instruct encodeHeapOop_not_null(iRegNNoSp dst, iRegP src, rFlagsReg cr) %{
  predicate(n->bottom_type()->make_ptr()->ptr() == TypePtr::NotNull);
  match(Set dst (EncodeP src));
  ins_cost(INSN_COST * 3);
  format %{ "encode_heap_oop_not_null $dst, $src" %}
  ins_encode %{
    __ encode_heap_oop_not_null($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg);
%}

instruct decodeHeapOop(iRegPNoSp dst, iRegN src, rFlagsReg cr) %{
  predicate(n->bottom_type()->is_ptr()->ptr() != TypePtr::NotNull &&
            n->bottom_type()->is_ptr()->ptr() != TypePtr::Constant);
  match(Set dst (DecodeN src));
  ins_cost(INSN_COST * 3);
  format %{ "decode_heap_oop $dst, $src" %}
  ins_encode %{
    Register s = $src$$Register;
    Register d = $dst$$Register;
    __ decode_heap_oop(d, s);
  %}
  ins_pipe(ialu_reg);
%}

instruct decodeHeapOop_not_null(iRegPNoSp dst, iRegN src, rFlagsReg cr) %{
  predicate(n->bottom_type()->is_ptr()->ptr() == TypePtr::NotNull ||
            n->bottom_type()->is_ptr()->ptr() == TypePtr::Constant);
  match(Set dst (DecodeN src));
  ins_cost(INSN_COST * 3);
  format %{ "decode_heap_oop_not_null $dst, $src" %}
  ins_encode %{
    Register s = $src$$Register;
    Register d = $dst$$Register;
    __ decode_heap_oop_not_null(d, s);
  %}
  ins_pipe(ialu_reg);
%}

// n.b. AArch64 implementations of encode_klass_not_null and
// decode_klass_not_null do not modify the flags register so, unlike
// Intel, we don't kill CR as a side effect here

instruct encodeKlass_not_null(iRegNNoSp dst, iRegP src) %{
  match(Set dst (EncodePKlass src));

  ins_cost(INSN_COST * 3);
  format %{ "encode_klass_not_null $dst,$src" %}

  ins_encode %{
    Register src_reg = as_Register($src$$reg);
    Register dst_reg = as_Register($dst$$reg);
    __ encode_klass_not_null(dst_reg, src_reg);
  %}

   ins_pipe(ialu_reg);
%}

instruct decodeKlass_not_null(iRegPNoSp dst, iRegN src) %{
  match(Set dst (DecodeNKlass src));

  ins_cost(INSN_COST * 3);
  format %{ "decode_klass_not_null $dst,$src" %}

  ins_encode %{
    Register src_reg = as_Register($src$$reg);
    Register dst_reg = as_Register($dst$$reg);
    if (dst_reg != src_reg) {
      __ decode_klass_not_null(dst_reg, src_reg);
    } else {
      __ decode_klass_not_null(dst_reg);
    }
  %}

   ins_pipe(ialu_reg);
%}

instruct checkCastPP(iRegPNoSp dst)
%{
  match(Set dst (CheckCastPP dst));

  size(0);
  format %{ "# checkcastPP of $dst" %}
  ins_encode(/* empty encoding */);
  ins_pipe(pipe_class_empty);
%}

instruct castPP(iRegPNoSp dst)
%{
  match(Set dst (CastPP dst));

  size(0);
  format %{ "# castPP of $dst" %}
  ins_encode(/* empty encoding */);
  ins_pipe(pipe_class_empty);
%}

instruct castII(iRegI dst)
%{
  match(Set dst (CastII dst));

  size(0);
  format %{ "# castII of $dst" %}
  ins_encode(/* empty encoding */);
  ins_cost(0);
  ins_pipe(pipe_class_empty);
%}

// ============================================================================
// Atomic operation instructions
//
// Intel and SPARC both implement Ideal Node LoadPLocked and
// Store{PIL}Conditional instructions using a normal load for the
// LoadPLocked and a CAS for the Store{PIL}Conditional.
//
// The ideal code appears only to use LoadPLocked/StorePLocked as a
// pair to lock object allocations from Eden space when not using
// TLABs.
//
// There does not appear to be a Load{IL}Locked Ideal Node and the
// Ideal code appears to use Store{IL}Conditional as an alias for CAS
// and to use StoreIConditional only for 32-bit and StoreLConditional
// only for 64-bit.
//
// We implement LoadPLocked and StorePLocked instructions using,
// respectively the AArch64 hw load-exclusive and store-conditional
// instructions. Whereas we must implement each of
// Store{IL}Conditional using a CAS which employs a pair of
// instructions comprising a load-exclusive followed by a
// store-conditional.


// Locked-load (linked load) of the current heap-top
// used when updating the eden heap top
// implemented using ldaxr on AArch64

instruct loadPLocked(iRegPNoSp dst, indirect mem)
%{
  match(Set dst (LoadPLocked mem));

  ins_cost(VOLATILE_REF_COST);

  format %{ "ldaxr $dst, $mem\t# ptr linked acquire" %}

  ins_encode(aarch64_enc_ldaxr(dst, mem));

  ins_pipe(pipe_serial);
%}

// Conditional-store of the updated heap-top.
// Used during allocation of the shared heap.
// Sets flag (EQ) on success.
// implemented using stlxr on AArch64.

instruct storePConditional(memory heap_top_ptr, iRegP oldval, iRegP newval, rFlagsReg cr)
%{
  match(Set cr (StorePConditional heap_top_ptr (Binary oldval newval)));

  ins_cost(VOLATILE_REF_COST);

 // TODO
 // do we need to do a store-conditional release or can we just use a
 // plain store-conditional?

  format %{
    "stlxr rscratch1, $newval, $heap_top_ptr\t# ptr cond release"
    "cmpw rscratch1, zr\t# EQ on successful write"
  %}

  ins_encode(aarch64_enc_stlxr(newval, heap_top_ptr));

  ins_pipe(pipe_serial);
%}


// storeLConditional is used by PhaseMacroExpand::expand_lock_node
// when attempting to rebias a lock towards the current thread.  We
// must use the acquire form of cmpxchg in order to guarantee acquire
// semantics in this case.
instruct storeLConditional(indirect mem, iRegLNoSp oldval, iRegLNoSp newval, rFlagsReg cr)
%{
  match(Set cr (StoreLConditional mem (Binary oldval newval)));

  ins_cost(VOLATILE_REF_COST);

  format %{
    "cmpxchg rscratch1, $mem, $oldval, $newval, $mem\t# if $mem == $oldval then $mem <-- $newval"
    "cmpw rscratch1, zr\t# EQ on successful write"
  %}

  ins_encode(aarch64_enc_cmpxchg_acq(mem, oldval, newval));

  ins_pipe(pipe_slow);
%}

// storeIConditional also has acquire semantics, for no better reason
// than matching storeLConditional.  At the time of writing this
// comment storeIConditional was not used anywhere by AArch64.
instruct storeIConditional(indirect mem, iRegINoSp oldval, iRegINoSp newval, rFlagsReg cr)
%{
  match(Set cr (StoreIConditional mem (Binary oldval newval)));

  ins_cost(VOLATILE_REF_COST);

  format %{
    "cmpxchgw rscratch1, $mem, $oldval, $newval, $mem\t# if $mem == $oldval then $mem <-- $newval"
    "cmpw rscratch1, zr\t# EQ on successful write"
  %}

  ins_encode(aarch64_enc_cmpxchgw_acq(mem, oldval, newval));

  ins_pipe(pipe_slow);
%}

// standard CompareAndSwapX when we are using barriers
// these have higher priority than the rules selected by a predicate

// XXX No flag versions for CompareAndSwap{I,L,P,N} because matcher
// can't match them

instruct compareAndSwapI(iRegINoSp res, indirect mem, iRegINoSp oldval, iRegINoSp newval, rFlagsReg cr) %{

  match(Set res (CompareAndSwapI mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);

  effect(KILL cr);

 format %{
    "cmpxchgw $mem, $oldval, $newval\t# (int) if $mem == $oldval then $mem <-- $newval"
    "cset $res, EQ\t# $res <-- (EQ ? 1 : 0)"
 %}

 ins_encode(aarch64_enc_cmpxchgw(mem, oldval, newval),
            aarch64_enc_cset_eq(res));

  ins_pipe(pipe_slow);
%}

instruct compareAndSwapL(iRegINoSp res, indirect mem, iRegLNoSp oldval, iRegLNoSp newval, rFlagsReg cr) %{

  match(Set res (CompareAndSwapL mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);

  effect(KILL cr);

 format %{
    "cmpxchg $mem, $oldval, $newval\t# (long) if $mem == $oldval then $mem <-- $newval"
    "cset $res, EQ\t# $res <-- (EQ ? 1 : 0)"
 %}

 ins_encode(aarch64_enc_cmpxchg(mem, oldval, newval),
            aarch64_enc_cset_eq(res));

  ins_pipe(pipe_slow);
%}

instruct compareAndSwapP(iRegINoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{

  match(Set res (CompareAndSwapP mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);

  effect(KILL cr);

 format %{
    "cmpxchg $mem, $oldval, $newval\t# (ptr) if $mem == $oldval then $mem <-- $newval"
    "cset $res, EQ\t# $res <-- (EQ ? 1 : 0)"
 %}

 ins_encode(aarch64_enc_cmpxchg(mem, oldval, newval),
            aarch64_enc_cset_eq(res));

  ins_pipe(pipe_slow);
%}

instruct compareAndSwapN(iRegINoSp res, indirect mem, iRegNNoSp oldval, iRegNNoSp newval, rFlagsReg cr) %{

  match(Set res (CompareAndSwapN mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);

  effect(KILL cr);

 format %{
    "cmpxchgw $mem, $oldval, $newval\t# (narrow oop) if $mem == $oldval then $mem <-- $newval"
    "cset $res, EQ\t# $res <-- (EQ ? 1 : 0)"
 %}

 ins_encode(aarch64_enc_cmpxchgw(mem, oldval, newval),
            aarch64_enc_cset_eq(res));

  ins_pipe(pipe_slow);
%}

// alternative CompareAndSwapX when we are eliding barriers

instruct compareAndSwapIAcq(iRegINoSp res, indirect mem, iRegINoSp oldval, iRegINoSp newval, rFlagsReg cr) %{

  predicate(needs_acquiring_load_exclusive(n));
  match(Set res (CompareAndSwapI mem (Binary oldval newval)));
  ins_cost(VOLATILE_REF_COST);

  effect(KILL cr);

 format %{
    "cmpxchgw_acq $mem, $oldval, $newval\t# (int) if $mem == $oldval then $mem <-- $newval"
    "cset $res, EQ\t# $res <-- (EQ ? 1 : 0)"
 %}

 ins_encode(aarch64_enc_cmpxchgw_acq(mem, oldval, newval),
            aarch64_enc_cset_eq(res));

  ins_pipe(pipe_slow);
%}

instruct compareAndSwapLAcq(iRegINoSp res, indirect mem, iRegLNoSp oldval, iRegLNoSp newval, rFlagsReg cr) %{

  predicate(needs_acquiring_load_exclusive(n));
  match(Set res (CompareAndSwapL mem (Binary oldval newval)));
  ins_cost(VOLATILE_REF_COST);

  effect(KILL cr);

 format %{
    "cmpxchg_acq $mem, $oldval, $newval\t# (long) if $mem == $oldval then $mem <-- $newval"
    "cset $res, EQ\t# $res <-- (EQ ? 1 : 0)"
 %}

 ins_encode(aarch64_enc_cmpxchg_acq(mem, oldval, newval),
            aarch64_enc_cset_eq(res));

  ins_pipe(pipe_slow);
%}

instruct compareAndSwapPAcq(iRegINoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{

  predicate(needs_acquiring_load_exclusive(n));
  match(Set res (CompareAndSwapP mem (Binary oldval newval)));
  ins_cost(VOLATILE_REF_COST);

  effect(KILL cr);

 format %{
    "cmpxchg_acq $mem, $oldval, $newval\t# (ptr) if $mem == $oldval then $mem <-- $newval"
    "cset $res, EQ\t# $res <-- (EQ ? 1 : 0)"
 %}

 ins_encode(aarch64_enc_cmpxchg_acq(mem, oldval, newval),
            aarch64_enc_cset_eq(res));

  ins_pipe(pipe_slow);
%}

instruct compareAndSwapNAcq(iRegINoSp res, indirect mem, iRegNNoSp oldval, iRegNNoSp newval, rFlagsReg cr) %{

  predicate(needs_acquiring_load_exclusive(n));
  match(Set res (CompareAndSwapN mem (Binary oldval newval)));
  ins_cost(VOLATILE_REF_COST);

  effect(KILL cr);

 format %{
    "cmpxchgw_acq $mem, $oldval, $newval\t# (narrow oop) if $mem == $oldval then $mem <-- $newval"
    "cset $res, EQ\t# $res <-- (EQ ? 1 : 0)"
 %}

 ins_encode(aarch64_enc_cmpxchgw_acq(mem, oldval, newval),
            aarch64_enc_cset_eq(res));

  ins_pipe(pipe_slow);
%}


// ---------------------------------------------------------------------


// BEGIN This section of the file is automatically generated. Do not edit --------------

// Sundry CAS operations.  Note that release is always true,
// regardless of the memory ordering of the CAS.  This is because we
// need the volatile case to be sequentially consistent but there is
// no trailing StoreLoad barrier emitted by C2.  Unfortunately we
// can't check the type of memory ordering here, so we always emit a
// STLXR.

// This section is generated from aarch64_ad_cas.m4



instruct compareAndExchangeB(iRegINoSp res, indirect mem, iRegI oldval, iRegI newval, rFlagsReg cr) %{
  match(Set res (CompareAndExchangeB mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);
  effect(TEMP_DEF res, KILL cr);
  format %{
    "cmpxchg $res = $mem, $oldval, $newval\t# (byte, weak) if $mem == $oldval then $mem <-- $newval"
  %}
  ins_encode %{
    __ uxtbw(rscratch2, $oldval$$Register);
    __ cmpxchg($mem$$Register, rscratch2, $newval$$Register,
               Assembler::byte, /*acquire*/ false, /*release*/ true,
               /*weak*/ false, $res$$Register);
    __ sxtbw($res$$Register, $res$$Register);
  %}
  ins_pipe(pipe_slow);
%}

instruct compareAndExchangeS(iRegINoSp res, indirect mem, iRegI oldval, iRegI newval, rFlagsReg cr) %{
  match(Set res (CompareAndExchangeS mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);
  effect(TEMP_DEF res, KILL cr);
  format %{
    "cmpxchg $res = $mem, $oldval, $newval\t# (short, weak) if $mem == $oldval then $mem <-- $newval"
  %}
  ins_encode %{
    __ uxthw(rscratch2, $oldval$$Register);
    __ cmpxchg($mem$$Register, rscratch2, $newval$$Register,
               Assembler::halfword, /*acquire*/ false, /*release*/ true,
               /*weak*/ false, $res$$Register);
    __ sxthw($res$$Register, $res$$Register);
  %}
  ins_pipe(pipe_slow);
%}

instruct compareAndExchangeI(iRegINoSp res, indirect mem, iRegI oldval, iRegI newval, rFlagsReg cr) %{
  match(Set res (CompareAndExchangeI mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);
  effect(TEMP_DEF res, KILL cr);
  format %{
    "cmpxchg $res = $mem, $oldval, $newval\t# (int, weak) if $mem == $oldval then $mem <-- $newval"
  %}
  ins_encode %{
    __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register,
               Assembler::word, /*acquire*/ false, /*release*/ true,
               /*weak*/ false, $res$$Register);
  %}
  ins_pipe(pipe_slow);
%}

instruct compareAndExchangeL(iRegLNoSp res, indirect mem, iRegL oldval, iRegL newval, rFlagsReg cr) %{
  match(Set res (CompareAndExchangeL mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);
  effect(TEMP_DEF res, KILL cr);
  format %{
    "cmpxchg $res = $mem, $oldval, $newval\t# (long, weak) if $mem == $oldval then $mem <-- $newval"
  %}
  ins_encode %{
    __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register,
               Assembler::xword, /*acquire*/ false, /*release*/ true,
               /*weak*/ false, $res$$Register);
  %}
  ins_pipe(pipe_slow);
%}

instruct compareAndExchangeN(iRegNNoSp res, indirect mem, iRegN oldval, iRegN newval, rFlagsReg cr) %{
  match(Set res (CompareAndExchangeN mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);
  effect(TEMP_DEF res, KILL cr);
  format %{
    "cmpxchg $res = $mem, $oldval, $newval\t# (narrow oop, weak) if $mem == $oldval then $mem <-- $newval"
  %}
  ins_encode %{
    __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register,
               Assembler::word, /*acquire*/ false, /*release*/ true,
               /*weak*/ false, $res$$Register);
  %}
  ins_pipe(pipe_slow);
%}

instruct compareAndExchangeP(iRegPNoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{
  match(Set res (CompareAndExchangeP mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);
  effect(TEMP_DEF res, KILL cr);
  format %{
    "cmpxchg $res = $mem, $oldval, $newval\t# (ptr, weak) if $mem == $oldval then $mem <-- $newval"
  %}
  ins_encode %{
    __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register,
               Assembler::xword, /*acquire*/ false, /*release*/ true,
               /*weak*/ false, $res$$Register);
  %}
  ins_pipe(pipe_slow);
%}

instruct weakCompareAndSwapB(iRegINoSp res, indirect mem, iRegI oldval, iRegI newval, rFlagsReg cr) %{
  match(Set res (WeakCompareAndSwapB mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);
  effect(KILL cr);
  format %{
    "cmpxchg $res = $mem, $oldval, $newval\t# (byte, weak) if $mem == $oldval then $mem <-- $newval"
    "csetw $res, EQ\t# $res <-- (EQ ? 1 : 0)"
  %}
  ins_encode %{
    __ uxtbw(rscratch2, $oldval$$Register);
    __ cmpxchg($mem$$Register, rscratch2, $newval$$Register,
               Assembler::byte, /*acquire*/ false, /*release*/ true,
               /*weak*/ true, noreg);
    __ csetw($res$$Register, Assembler::EQ);
  %}
  ins_pipe(pipe_slow);
%}

instruct weakCompareAndSwapS(iRegINoSp res, indirect mem, iRegI oldval, iRegI newval, rFlagsReg cr) %{
  match(Set res (WeakCompareAndSwapS mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);
  effect(KILL cr);
  format %{
    "cmpxchg $res = $mem, $oldval, $newval\t# (short, weak) if $mem == $oldval then $mem <-- $newval"
    "csetw $res, EQ\t# $res <-- (EQ ? 1 : 0)"
  %}
  ins_encode %{
    __ uxthw(rscratch2, $oldval$$Register);
    __ cmpxchg($mem$$Register, rscratch2, $newval$$Register,
               Assembler::halfword, /*acquire*/ false, /*release*/ true,
               /*weak*/ true, noreg);
    __ csetw($res$$Register, Assembler::EQ);
  %}
  ins_pipe(pipe_slow);
%}

instruct weakCompareAndSwapI(iRegINoSp res, indirect mem, iRegI oldval, iRegI newval, rFlagsReg cr) %{
  match(Set res (WeakCompareAndSwapI mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);
  effect(KILL cr);
  format %{
    "cmpxchg $res = $mem, $oldval, $newval\t# (int, weak) if $mem == $oldval then $mem <-- $newval"
    "csetw $res, EQ\t# $res <-- (EQ ? 1 : 0)"
  %}
  ins_encode %{
    __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register,
               Assembler::word, /*acquire*/ false, /*release*/ true,
               /*weak*/ true, noreg);
    __ csetw($res$$Register, Assembler::EQ);
  %}
  ins_pipe(pipe_slow);
%}

instruct weakCompareAndSwapL(iRegINoSp res, indirect mem, iRegL oldval, iRegL newval, rFlagsReg cr) %{
  match(Set res (WeakCompareAndSwapL mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);
  effect(KILL cr);
  format %{
    "cmpxchg $res = $mem, $oldval, $newval\t# (long, weak) if $mem == $oldval then $mem <-- $newval"
    "csetw $res, EQ\t# $res <-- (EQ ? 1 : 0)"
  %}
  ins_encode %{
    __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register,
               Assembler::xword, /*acquire*/ false, /*release*/ true,
               /*weak*/ true, noreg);
    __ csetw($res$$Register, Assembler::EQ);
  %}
  ins_pipe(pipe_slow);
%}

instruct weakCompareAndSwapN(iRegINoSp res, indirect mem, iRegN oldval, iRegN newval, rFlagsReg cr) %{
  match(Set res (WeakCompareAndSwapN mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);
  effect(KILL cr);
  format %{
    "cmpxchg $res = $mem, $oldval, $newval\t# (narrow oop, weak) if $mem == $oldval then $mem <-- $newval"
    "csetw $res, EQ\t# $res <-- (EQ ? 1 : 0)"
  %}
  ins_encode %{
    __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register,
               Assembler::word, /*acquire*/ false, /*release*/ true,
               /*weak*/ true, noreg);
    __ csetw($res$$Register, Assembler::EQ);
  %}
  ins_pipe(pipe_slow);
%}

instruct weakCompareAndSwapP(iRegINoSp res, indirect mem, iRegP oldval, iRegP newval, rFlagsReg cr) %{
  match(Set res (WeakCompareAndSwapP mem (Binary oldval newval)));
  ins_cost(2 * VOLATILE_REF_COST);
  effect(KILL cr);
  format %{
    "cmpxchg $res = $mem, $oldval, $newval\t# (ptr, weak) if $mem == $oldval then $mem <-- $newval"
    "csetw $res, EQ\t# $res <-- (EQ ? 1 : 0)"
  %}
  ins_encode %{
    __ cmpxchg($mem$$Register, $oldval$$Register, $newval$$Register,
               Assembler::xword, /*acquire*/ false, /*release*/ true,
               /*weak*/ true, noreg);
    __ csetw($res$$Register, Assembler::EQ);
  %}
  ins_pipe(pipe_slow);
%}

// END This section of the file is automatically generated. Do not edit --------------
// ---------------------------------------------------------------------

instruct get_and_setI(indirect mem, iRegI newv, iRegINoSp prev) %{
  match(Set prev (GetAndSetI mem newv));
  format %{ "atomic_xchgw  $prev, $newv, [$mem]" %}
  ins_encode %{
    __ atomic_xchgw($prev$$Register, $newv$$Register, as_Register($mem$$base));
  %}
  ins_pipe(pipe_serial);
%}

instruct get_and_setL(indirect mem, iRegL newv, iRegLNoSp prev) %{
  match(Set prev (GetAndSetL mem newv));
  format %{ "atomic_xchg  $prev, $newv, [$mem]" %}
  ins_encode %{
    __ atomic_xchg($prev$$Register, $newv$$Register, as_Register($mem$$base));
  %}
  ins_pipe(pipe_serial);
%}

instruct get_and_setN(indirect mem, iRegN newv, iRegINoSp prev) %{
  match(Set prev (GetAndSetN mem newv));
  format %{ "atomic_xchgw $prev, $newv, [$mem]" %}
  ins_encode %{
    __ atomic_xchgw($prev$$Register, $newv$$Register, as_Register($mem$$base));
  %}
  ins_pipe(pipe_serial);
%}

instruct get_and_setP(indirect mem, iRegP newv, iRegPNoSp prev) %{
  match(Set prev (GetAndSetP mem newv));
  format %{ "atomic_xchg  $prev, $newv, [$mem]" %}
  ins_encode %{
    __ atomic_xchg($prev$$Register, $newv$$Register, as_Register($mem$$base));
  %}
  ins_pipe(pipe_serial);
%}


instruct get_and_addL(indirect mem, iRegLNoSp newval, iRegL incr) %{
  match(Set newval (GetAndAddL mem incr));
  ins_cost(INSN_COST * 10);
  format %{ "get_and_addL $newval, [$mem], $incr" %}
  ins_encode %{
    __ atomic_add($newval$$Register, $incr$$Register, as_Register($mem$$base));
  %}
  ins_pipe(pipe_serial);
%}

instruct get_and_addL_no_res(indirect mem, Universe dummy, iRegL incr) %{
  predicate(n->as_LoadStore()->result_not_used());
  match(Set dummy (GetAndAddL mem incr));
  ins_cost(INSN_COST * 9);
  format %{ "get_and_addL [$mem], $incr" %}
  ins_encode %{
    __ atomic_add(noreg, $incr$$Register, as_Register($mem$$base));
  %}
  ins_pipe(pipe_serial);
%}

instruct get_and_addLi(indirect mem, iRegLNoSp newval, immLAddSub incr) %{
  match(Set newval (GetAndAddL mem incr));
  ins_cost(INSN_COST * 10);
  format %{ "get_and_addL $newval, [$mem], $incr" %}
  ins_encode %{
    __ atomic_add($newval$$Register, $incr$$constant, as_Register($mem$$base));
  %}
  ins_pipe(pipe_serial);
%}

instruct get_and_addLi_no_res(indirect mem, Universe dummy, immLAddSub incr) %{
  predicate(n->as_LoadStore()->result_not_used());
  match(Set dummy (GetAndAddL mem incr));
  ins_cost(INSN_COST * 9);
  format %{ "get_and_addL [$mem], $incr" %}
  ins_encode %{
    __ atomic_add(noreg, $incr$$constant, as_Register($mem$$base));
  %}
  ins_pipe(pipe_serial);
%}

instruct get_and_addI(indirect mem, iRegINoSp newval, iRegIorL2I incr) %{
  match(Set newval (GetAndAddI mem incr));
  ins_cost(INSN_COST * 10);
  format %{ "get_and_addI $newval, [$mem], $incr" %}
  ins_encode %{
    __ atomic_addw($newval$$Register, $incr$$Register, as_Register($mem$$base));
  %}
  ins_pipe(pipe_serial);
%}

instruct get_and_addI_no_res(indirect mem, Universe dummy, iRegIorL2I incr) %{
  predicate(n->as_LoadStore()->result_not_used());
  match(Set dummy (GetAndAddI mem incr));
  ins_cost(INSN_COST * 9);
  format %{ "get_and_addI [$mem], $incr" %}
  ins_encode %{
    __ atomic_addw(noreg, $incr$$Register, as_Register($mem$$base));
  %}
  ins_pipe(pipe_serial);
%}

instruct get_and_addIi(indirect mem, iRegINoSp newval, immIAddSub incr) %{
  match(Set newval (GetAndAddI mem incr));
  ins_cost(INSN_COST * 10);
  format %{ "get_and_addI $newval, [$mem], $incr" %}
  ins_encode %{
    __ atomic_addw($newval$$Register, $incr$$constant, as_Register($mem$$base));
  %}
  ins_pipe(pipe_serial);
%}

instruct get_and_addIi_no_res(indirect mem, Universe dummy, immIAddSub incr) %{
  predicate(n->as_LoadStore()->result_not_used());
  match(Set dummy (GetAndAddI mem incr));
  ins_cost(INSN_COST * 9);
  format %{ "get_and_addI [$mem], $incr" %}
  ins_encode %{
    __ atomic_addw(noreg, $incr$$constant, as_Register($mem$$base));
  %}
  ins_pipe(pipe_serial);
%}

// Manifest a CmpL result in an integer register.
// (src1 < src2) ? -1 : ((src1 > src2) ? 1 : 0)
instruct cmpL3_reg_reg(iRegINoSp dst, iRegL src1, iRegL src2, rFlagsReg flags)
%{
  match(Set dst (CmpL3 src1 src2));
  effect(KILL flags);

  ins_cost(INSN_COST * 6);
  format %{
      "cmp $src1, $src2"
      "csetw $dst, ne"
      "cnegw $dst, lt"
  %}
  // format %{ "CmpL3 $dst, $src1, $src2" %}
  ins_encode %{
    __ cmp($src1$$Register, $src2$$Register);
    __ csetw($dst$$Register, Assembler::NE);
    __ cnegw($dst$$Register, $dst$$Register, Assembler::LT);
  %}

  ins_pipe(pipe_class_default);
%}

instruct cmpL3_reg_imm(iRegINoSp dst, iRegL src1, immLAddSub src2, rFlagsReg flags)
%{
  match(Set dst (CmpL3 src1 src2));
  effect(KILL flags);

  ins_cost(INSN_COST * 6);
  format %{
      "cmp $src1, $src2"
      "csetw $dst, ne"
      "cnegw $dst, lt"
  %}
  ins_encode %{
    int32_t con = (int32_t)$src2$$constant;
     if (con < 0) {
      __ adds(zr, $src1$$Register, -con);
    } else {
      __ subs(zr, $src1$$Register, con);
    }
    __ csetw($dst$$Register, Assembler::NE);
    __ cnegw($dst$$Register, $dst$$Register, Assembler::LT);
  %}

  ins_pipe(pipe_class_default);
%}

// ============================================================================
// Conditional Move Instructions

// n.b. we have identical rules for both a signed compare op (cmpOp)
// and an unsigned compare op (cmpOpU). it would be nice if we could
// define an op class which merged both inputs and use it to type the
// argument to a single rule. unfortunatelyt his fails because the
// opclass does not live up to the COND_INTER interface of its
// component operands. When the generic code tries to negate the
// operand it ends up running the generci Machoper::negate method
// which throws a ShouldNotHappen. So, we have to provide two flavours
// of each rule, one for a cmpOp and a second for a cmpOpU (sigh).

instruct cmovI_reg_reg(cmpOp cmp, rFlagsReg cr, iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (CMoveI (Binary cmp cr) (Binary src1 src2)));

  ins_cost(INSN_COST * 2);
  format %{ "cselw $dst, $src2, $src1 $cmp\t# signed, int"  %}

  ins_encode %{
    __ cselw(as_Register($dst$$reg),
             as_Register($src2$$reg),
             as_Register($src1$$reg),
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg_reg);
%}

instruct cmovUI_reg_reg(cmpOpU cmp, rFlagsRegU cr, iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (CMoveI (Binary cmp cr) (Binary src1 src2)));

  ins_cost(INSN_COST * 2);
  format %{ "cselw $dst, $src2, $src1 $cmp\t# unsigned, int"  %}

  ins_encode %{
    __ cselw(as_Register($dst$$reg),
             as_Register($src2$$reg),
             as_Register($src1$$reg),
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg_reg);
%}

// special cases where one arg is zero

// n.b. this is selected in preference to the rule above because it
// avoids loading constant 0 into a source register

// TODO
// we ought only to be able to cull one of these variants as the ideal
// transforms ought always to order the zero consistently (to left/right?)

instruct cmovI_zero_reg(cmpOp cmp, rFlagsReg cr, iRegINoSp dst, immI0 zero, iRegIorL2I src) %{
  match(Set dst (CMoveI (Binary cmp cr) (Binary zero src)));

  ins_cost(INSN_COST * 2);
  format %{ "cselw $dst, $src, zr $cmp\t# signed, int"  %}

  ins_encode %{
    __ cselw(as_Register($dst$$reg),
             as_Register($src$$reg),
             zr,
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovUI_zero_reg(cmpOpU cmp, rFlagsRegU cr, iRegINoSp dst, immI0 zero, iRegIorL2I src) %{
  match(Set dst (CMoveI (Binary cmp cr) (Binary zero src)));

  ins_cost(INSN_COST * 2);
  format %{ "cselw $dst, $src, zr $cmp\t# unsigned, int"  %}

  ins_encode %{
    __ cselw(as_Register($dst$$reg),
             as_Register($src$$reg),
             zr,
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovI_reg_zero(cmpOp cmp, rFlagsReg cr, iRegINoSp dst, iRegIorL2I src, immI0 zero) %{
  match(Set dst (CMoveI (Binary cmp cr) (Binary src zero)));

  ins_cost(INSN_COST * 2);
  format %{ "cselw $dst, zr, $src $cmp\t# signed, int"  %}

  ins_encode %{
    __ cselw(as_Register($dst$$reg),
             zr,
             as_Register($src$$reg),
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovUI_reg_zero(cmpOpU cmp, rFlagsRegU cr, iRegINoSp dst, iRegIorL2I src, immI0 zero) %{
  match(Set dst (CMoveI (Binary cmp cr) (Binary src zero)));

  ins_cost(INSN_COST * 2);
  format %{ "cselw $dst, zr, $src $cmp\t# unsigned, int"  %}

  ins_encode %{
    __ cselw(as_Register($dst$$reg),
             zr,
             as_Register($src$$reg),
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

// special case for creating a boolean 0 or 1

// n.b. this is selected in preference to the rule above because it
// avoids loading constants 0 and 1 into a source register

instruct cmovI_reg_zero_one(cmpOp cmp, rFlagsReg cr, iRegINoSp dst, immI0 zero, immI_1 one) %{
  match(Set dst (CMoveI (Binary cmp cr) (Binary one zero)));

  ins_cost(INSN_COST * 2);
  format %{ "csincw $dst, zr, zr $cmp\t# signed, int"  %}

  ins_encode %{
    // equivalently
    // cset(as_Register($dst$$reg),
    //      negate_condition((Assembler::Condition)$cmp$$cmpcode));
    __ csincw(as_Register($dst$$reg),
             zr,
             zr,
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_none);
%}

instruct cmovUI_reg_zero_one(cmpOpU cmp, rFlagsRegU cr, iRegINoSp dst, immI0 zero, immI_1 one) %{
  match(Set dst (CMoveI (Binary cmp cr) (Binary one zero)));

  ins_cost(INSN_COST * 2);
  format %{ "csincw $dst, zr, zr $cmp\t# unsigned, int"  %}

  ins_encode %{
    // equivalently
    // cset(as_Register($dst$$reg),
    //      negate_condition((Assembler::Condition)$cmp$$cmpcode));
    __ csincw(as_Register($dst$$reg),
             zr,
             zr,
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_none);
%}

instruct cmovL_reg_reg(cmpOp cmp, rFlagsReg cr, iRegLNoSp dst, iRegL src1, iRegL src2) %{
  match(Set dst (CMoveL (Binary cmp cr) (Binary src1 src2)));

  ins_cost(INSN_COST * 2);
  format %{ "csel $dst, $src2, $src1 $cmp\t# signed, long"  %}

  ins_encode %{
    __ csel(as_Register($dst$$reg),
            as_Register($src2$$reg),
            as_Register($src1$$reg),
            (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg_reg);
%}

instruct cmovUL_reg_reg(cmpOpU cmp, rFlagsRegU cr, iRegLNoSp dst, iRegL src1, iRegL src2) %{
  match(Set dst (CMoveL (Binary cmp cr) (Binary src1 src2)));

  ins_cost(INSN_COST * 2);
  format %{ "csel $dst, $src2, $src1 $cmp\t# unsigned, long"  %}

  ins_encode %{
    __ csel(as_Register($dst$$reg),
            as_Register($src2$$reg),
            as_Register($src1$$reg),
            (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg_reg);
%}

// special cases where one arg is zero

instruct cmovL_reg_zero(cmpOp cmp, rFlagsReg cr, iRegLNoSp dst, iRegL src, immL0 zero) %{
  match(Set dst (CMoveL (Binary cmp cr) (Binary src zero)));

  ins_cost(INSN_COST * 2);
  format %{ "csel $dst, zr, $src $cmp\t# signed, long"  %}

  ins_encode %{
    __ csel(as_Register($dst$$reg),
            zr,
            as_Register($src$$reg),
            (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovUL_reg_zero(cmpOpU cmp, rFlagsRegU cr, iRegLNoSp dst, iRegL src, immL0 zero) %{
  match(Set dst (CMoveL (Binary cmp cr) (Binary src zero)));

  ins_cost(INSN_COST * 2);
  format %{ "csel $dst, zr, $src $cmp\t# unsigned, long"  %}

  ins_encode %{
    __ csel(as_Register($dst$$reg),
            zr,
            as_Register($src$$reg),
            (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovL_zero_reg(cmpOp cmp, rFlagsReg cr, iRegLNoSp dst, immL0 zero, iRegL src) %{
  match(Set dst (CMoveL (Binary cmp cr) (Binary zero src)));

  ins_cost(INSN_COST * 2);
  format %{ "csel $dst, $src, zr $cmp\t# signed, long"  %}

  ins_encode %{
    __ csel(as_Register($dst$$reg),
            as_Register($src$$reg),
            zr,
            (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovUL_zero_reg(cmpOpU cmp, rFlagsRegU cr, iRegLNoSp dst, immL0 zero, iRegL src) %{
  match(Set dst (CMoveL (Binary cmp cr) (Binary zero src)));

  ins_cost(INSN_COST * 2);
  format %{ "csel $dst, $src, zr $cmp\t# unsigned, long"  %}

  ins_encode %{
    __ csel(as_Register($dst$$reg),
            as_Register($src$$reg),
            zr,
            (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovP_reg_reg(cmpOp cmp, rFlagsReg cr, iRegPNoSp dst, iRegP src1, iRegP src2) %{
  match(Set dst (CMoveP (Binary cmp cr) (Binary src1 src2)));

  ins_cost(INSN_COST * 2);
  format %{ "csel $dst, $src2, $src1 $cmp\t# signed, ptr"  %}

  ins_encode %{
    __ csel(as_Register($dst$$reg),
            as_Register($src2$$reg),
            as_Register($src1$$reg),
            (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg_reg);
%}

instruct cmovUP_reg_reg(cmpOpU cmp, rFlagsRegU cr, iRegPNoSp dst, iRegP src1, iRegP src2) %{
  match(Set dst (CMoveP (Binary cmp cr) (Binary src1 src2)));

  ins_cost(INSN_COST * 2);
  format %{ "csel $dst, $src2, $src1 $cmp\t# unsigned, ptr"  %}

  ins_encode %{
    __ csel(as_Register($dst$$reg),
            as_Register($src2$$reg),
            as_Register($src1$$reg),
            (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg_reg);
%}

// special cases where one arg is zero

instruct cmovP_reg_zero(cmpOp cmp, rFlagsReg cr, iRegPNoSp dst, iRegP src, immP0 zero) %{
  match(Set dst (CMoveP (Binary cmp cr) (Binary src zero)));

  ins_cost(INSN_COST * 2);
  format %{ "csel $dst, zr, $src $cmp\t# signed, ptr"  %}

  ins_encode %{
    __ csel(as_Register($dst$$reg),
            zr,
            as_Register($src$$reg),
            (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovUP_reg_zero(cmpOpU cmp, rFlagsRegU cr, iRegPNoSp dst, iRegP src, immP0 zero) %{
  match(Set dst (CMoveP (Binary cmp cr) (Binary src zero)));

  ins_cost(INSN_COST * 2);
  format %{ "csel $dst, zr, $src $cmp\t# unsigned, ptr"  %}

  ins_encode %{
    __ csel(as_Register($dst$$reg),
            zr,
            as_Register($src$$reg),
            (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovP_zero_reg(cmpOp cmp, rFlagsReg cr, iRegPNoSp dst, immP0 zero, iRegP src) %{
  match(Set dst (CMoveP (Binary cmp cr) (Binary zero src)));

  ins_cost(INSN_COST * 2);
  format %{ "csel $dst, $src, zr $cmp\t# signed, ptr"  %}

  ins_encode %{
    __ csel(as_Register($dst$$reg),
            as_Register($src$$reg),
            zr,
            (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovUP_zero_reg(cmpOpU cmp, rFlagsRegU cr, iRegPNoSp dst, immP0 zero, iRegP src) %{
  match(Set dst (CMoveP (Binary cmp cr) (Binary zero src)));

  ins_cost(INSN_COST * 2);
  format %{ "csel $dst, $src, zr $cmp\t# unsigned, ptr"  %}

  ins_encode %{
    __ csel(as_Register($dst$$reg),
            as_Register($src$$reg),
            zr,
            (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovN_reg_reg(cmpOp cmp, rFlagsReg cr, iRegNNoSp dst, iRegN src1, iRegN src2) %{
  match(Set dst (CMoveN (Binary cmp cr) (Binary src1 src2)));

  ins_cost(INSN_COST * 2);
  format %{ "cselw $dst, $src2, $src1 $cmp\t# signed, compressed ptr"  %}

  ins_encode %{
    __ cselw(as_Register($dst$$reg),
             as_Register($src2$$reg),
             as_Register($src1$$reg),
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg_reg);
%}

instruct cmovUN_reg_reg(cmpOpU cmp, rFlagsRegU cr, iRegNNoSp dst, iRegN src1, iRegN src2) %{
  match(Set dst (CMoveN (Binary cmp cr) (Binary src1 src2)));

  ins_cost(INSN_COST * 2);
  format %{ "cselw $dst, $src2, $src1 $cmp\t# signed, compressed ptr"  %}

  ins_encode %{
    __ cselw(as_Register($dst$$reg),
             as_Register($src2$$reg),
             as_Register($src1$$reg),
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg_reg);
%}

// special cases where one arg is zero

instruct cmovN_reg_zero(cmpOp cmp, rFlagsReg cr, iRegNNoSp dst, iRegN src, immN0 zero) %{
  match(Set dst (CMoveN (Binary cmp cr) (Binary src zero)));

  ins_cost(INSN_COST * 2);
  format %{ "cselw $dst, zr, $src $cmp\t# signed, compressed ptr"  %}

  ins_encode %{
    __ cselw(as_Register($dst$$reg),
             zr,
             as_Register($src$$reg),
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovUN_reg_zero(cmpOpU cmp, rFlagsRegU cr, iRegNNoSp dst, iRegN src, immN0 zero) %{
  match(Set dst (CMoveN (Binary cmp cr) (Binary src zero)));

  ins_cost(INSN_COST * 2);
  format %{ "cselw $dst, zr, $src $cmp\t# unsigned, compressed ptr"  %}

  ins_encode %{
    __ cselw(as_Register($dst$$reg),
             zr,
             as_Register($src$$reg),
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovN_zero_reg(cmpOp cmp, rFlagsReg cr, iRegNNoSp dst, immN0 zero, iRegN src) %{
  match(Set dst (CMoveN (Binary cmp cr) (Binary zero src)));

  ins_cost(INSN_COST * 2);
  format %{ "cselw $dst, $src, zr $cmp\t# signed, compressed ptr"  %}

  ins_encode %{
    __ cselw(as_Register($dst$$reg),
             as_Register($src$$reg),
             zr,
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovUN_zero_reg(cmpOpU cmp, rFlagsRegU cr, iRegNNoSp dst, immN0 zero, iRegN src) %{
  match(Set dst (CMoveN (Binary cmp cr) (Binary zero src)));

  ins_cost(INSN_COST * 2);
  format %{ "cselw $dst, $src, zr $cmp\t# unsigned, compressed ptr"  %}

  ins_encode %{
    __ cselw(as_Register($dst$$reg),
             as_Register($src$$reg),
             zr,
             (Assembler::Condition)$cmp$$cmpcode);
  %}

  ins_pipe(icond_reg);
%}

instruct cmovF_reg(cmpOp cmp, rFlagsReg cr, vRegF dst, vRegF src1,  vRegF src2)
%{
  match(Set dst (CMoveF (Binary cmp cr) (Binary src1 src2)));

  ins_cost(INSN_COST * 3);

  format %{ "fcsels $dst, $src1, $src2, $cmp\t# signed cmove float\n\t" %}
  ins_encode %{
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    __ fcsels(as_FloatRegister($dst$$reg),
              as_FloatRegister($src2$$reg),
              as_FloatRegister($src1$$reg),
              cond);
  %}

  ins_pipe(fp_cond_reg_reg_s);
%}

instruct cmovUF_reg(cmpOpU cmp, rFlagsRegU cr, vRegF dst, vRegF src1,  vRegF src2)
%{
  match(Set dst (CMoveF (Binary cmp cr) (Binary src1 src2)));

  ins_cost(INSN_COST * 3);

  format %{ "fcsels $dst, $src1, $src2, $cmp\t# unsigned cmove float\n\t" %}
  ins_encode %{
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    __ fcsels(as_FloatRegister($dst$$reg),
              as_FloatRegister($src2$$reg),
              as_FloatRegister($src1$$reg),
              cond);
  %}

  ins_pipe(fp_cond_reg_reg_s);
%}

instruct cmovD_reg(cmpOp cmp, rFlagsReg cr, vRegD dst, vRegD src1,  vRegD src2)
%{
  match(Set dst (CMoveD (Binary cmp cr) (Binary src1 src2)));

  ins_cost(INSN_COST * 3);

  format %{ "fcseld $dst, $src1, $src2, $cmp\t# signed cmove float\n\t" %}
  ins_encode %{
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    __ fcseld(as_FloatRegister($dst$$reg),
              as_FloatRegister($src2$$reg),
              as_FloatRegister($src1$$reg),
              cond);
  %}

  ins_pipe(fp_cond_reg_reg_d);
%}

instruct cmovUD_reg(cmpOpU cmp, rFlagsRegU cr, vRegD dst, vRegD src1,  vRegD src2)
%{
  match(Set dst (CMoveD (Binary cmp cr) (Binary src1 src2)));

  ins_cost(INSN_COST * 3);

  format %{ "fcseld $dst, $src1, $src2, $cmp\t# unsigned cmove float\n\t" %}
  ins_encode %{
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    __ fcseld(as_FloatRegister($dst$$reg),
              as_FloatRegister($src2$$reg),
              as_FloatRegister($src1$$reg),
              cond);
  %}

  ins_pipe(fp_cond_reg_reg_d);
%}

// ============================================================================
// Arithmetic Instructions
//

// Integer Addition

// TODO
// these currently employ operations which do not set CR and hence are
// not flagged as killing CR but we would like to isolate the cases
// where we want to set flags from those where we don't. need to work
// out how to do that.

instruct addI_reg_reg(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (AddI src1 src2));

  ins_cost(INSN_COST);
  format %{ "addw  $dst, $src1, $src2" %}

  ins_encode %{
    __ addw(as_Register($dst$$reg),
            as_Register($src1$$reg),
            as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct addI_reg_imm(iRegINoSp dst, iRegIorL2I src1, immIAddSub src2) %{
  match(Set dst (AddI src1 src2));

  ins_cost(INSN_COST);
  format %{ "addw $dst, $src1, $src2" %}

  // use opcode to indicate that this is an add not a sub
  opcode(0x0);

  ins_encode(aarch64_enc_addsubw_imm(dst, src1, src2));

  ins_pipe(ialu_reg_imm);
%}

instruct addI_reg_imm_i2l(iRegINoSp dst, iRegL src1, immIAddSub src2) %{
  match(Set dst (AddI (ConvL2I src1) src2));

  ins_cost(INSN_COST);
  format %{ "addw $dst, $src1, $src2" %}

  // use opcode to indicate that this is an add not a sub
  opcode(0x0);

  ins_encode(aarch64_enc_addsubw_imm(dst, src1, src2));

  ins_pipe(ialu_reg_imm);
%}

// Pointer Addition
instruct addP_reg_reg(iRegPNoSp dst, iRegP src1, iRegL src2) %{
  match(Set dst (AddP src1 src2));

  ins_cost(INSN_COST);
  format %{ "add $dst, $src1, $src2\t# ptr" %}

  ins_encode %{
    __ add(as_Register($dst$$reg),
           as_Register($src1$$reg),
           as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct addP_reg_reg_ext(iRegPNoSp dst, iRegP src1, iRegIorL2I src2) %{
  match(Set dst (AddP src1 (ConvI2L src2)));

  ins_cost(1.9 * INSN_COST);
  format %{ "add $dst, $src1, $src2, sxtw\t# ptr" %}

  ins_encode %{
    __ add(as_Register($dst$$reg),
           as_Register($src1$$reg),
           as_Register($src2$$reg), ext::sxtw);
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct addP_reg_reg_lsl(iRegPNoSp dst, iRegP src1, iRegL src2, immIScale scale) %{
  match(Set dst (AddP src1 (LShiftL src2 scale)));

  ins_cost(1.9 * INSN_COST);
  format %{ "add $dst, $src1, $src2, LShiftL $scale\t# ptr" %}

  ins_encode %{
    __ lea(as_Register($dst$$reg),
           Address(as_Register($src1$$reg), as_Register($src2$$reg),
                   Address::lsl($scale$$constant)));
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct addP_reg_reg_ext_shift(iRegPNoSp dst, iRegP src1, iRegIorL2I src2, immIScale scale) %{
  match(Set dst (AddP src1 (LShiftL (ConvI2L src2) scale)));

  ins_cost(1.9 * INSN_COST);
  format %{ "add $dst, $src1, $src2, I2L $scale\t# ptr" %}

  ins_encode %{
    __ lea(as_Register($dst$$reg),
           Address(as_Register($src1$$reg), as_Register($src2$$reg),
                   Address::sxtw($scale$$constant)));
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct lshift_ext(iRegLNoSp dst, iRegIorL2I src, immI scale, rFlagsReg cr) %{
  match(Set dst (LShiftL (ConvI2L src) scale));

  ins_cost(INSN_COST);
  format %{ "sbfiz $dst, $src, $scale & 63, -$scale & 63\t" %}

  ins_encode %{
    __ sbfiz(as_Register($dst$$reg),
          as_Register($src$$reg),
          $scale$$constant & 63, MIN(32, (-$scale$$constant) & 63));
  %}

  ins_pipe(ialu_reg_shift);
%}

// Pointer Immediate Addition
// n.b. this needs to be more expensive than using an indirect memory
// operand
instruct addP_reg_imm(iRegPNoSp dst, iRegP src1, immLAddSub src2) %{
  match(Set dst (AddP src1 src2));

  ins_cost(INSN_COST);
  format %{ "add $dst, $src1, $src2\t# ptr" %}

  // use opcode to indicate that this is an add not a sub
  opcode(0x0);

  ins_encode( aarch64_enc_addsub_imm(dst, src1, src2) );

  ins_pipe(ialu_reg_imm);
%}

// Long Addition
instruct addL_reg_reg(iRegLNoSp dst, iRegL src1, iRegL src2) %{

  match(Set dst (AddL src1 src2));

  ins_cost(INSN_COST);
  format %{ "add  $dst, $src1, $src2" %}

  ins_encode %{
    __ add(as_Register($dst$$reg),
           as_Register($src1$$reg),
           as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg);
%}

// No constant pool entries requiredLong Immediate Addition.
instruct addL_reg_imm(iRegLNoSp dst, iRegL src1, immLAddSub src2) %{
  match(Set dst (AddL src1 src2));

  ins_cost(INSN_COST);
  format %{ "add $dst, $src1, $src2" %}

  // use opcode to indicate that this is an add not a sub
  opcode(0x0);

  ins_encode( aarch64_enc_addsub_imm(dst, src1, src2) );

  ins_pipe(ialu_reg_imm);
%}

// Integer Subtraction
instruct subI_reg_reg(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (SubI src1 src2));

  ins_cost(INSN_COST);
  format %{ "subw  $dst, $src1, $src2" %}

  ins_encode %{
    __ subw(as_Register($dst$$reg),
            as_Register($src1$$reg),
            as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg);
%}

// Immediate Subtraction
instruct subI_reg_imm(iRegINoSp dst, iRegIorL2I src1, immIAddSub src2) %{
  match(Set dst (SubI src1 src2));

  ins_cost(INSN_COST);
  format %{ "subw $dst, $src1, $src2" %}

  // use opcode to indicate that this is a sub not an add
  opcode(0x1);

  ins_encode(aarch64_enc_addsubw_imm(dst, src1, src2));

  ins_pipe(ialu_reg_imm);
%}

// Long Subtraction
instruct subL_reg_reg(iRegLNoSp dst, iRegL src1, iRegL src2) %{

  match(Set dst (SubL src1 src2));

  ins_cost(INSN_COST);
  format %{ "sub  $dst, $src1, $src2" %}

  ins_encode %{
    __ sub(as_Register($dst$$reg),
           as_Register($src1$$reg),
           as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg);
%}

// No constant pool entries requiredLong Immediate Subtraction.
instruct subL_reg_imm(iRegLNoSp dst, iRegL src1, immLAddSub src2) %{
  match(Set dst (SubL src1 src2));

  ins_cost(INSN_COST);
  format %{ "sub$dst, $src1, $src2" %}

  // use opcode to indicate that this is a sub not an add
  opcode(0x1);

  ins_encode( aarch64_enc_addsub_imm(dst, src1, src2) );

  ins_pipe(ialu_reg_imm);
%}

// Integer Negation (special case for sub)

instruct negI_reg(iRegINoSp dst, iRegIorL2I src, immI0 zero, rFlagsReg cr) %{
  match(Set dst (SubI zero src));

  ins_cost(INSN_COST);
  format %{ "negw $dst, $src\t# int" %}

  ins_encode %{
    __ negw(as_Register($dst$$reg),
            as_Register($src$$reg));
  %}

  ins_pipe(ialu_reg);
%}

// Long Negation

instruct negL_reg(iRegLNoSp dst, iRegIorL2I src, immL0 zero, rFlagsReg cr) %{
  match(Set dst (SubL zero src));

  ins_cost(INSN_COST);
  format %{ "neg $dst, $src\t# long" %}

  ins_encode %{
    __ neg(as_Register($dst$$reg),
           as_Register($src$$reg));
  %}

  ins_pipe(ialu_reg);
%}

// Integer Multiply

instruct mulI(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (MulI src1 src2));

  ins_cost(INSN_COST * 3);
  format %{ "mulw  $dst, $src1, $src2" %}

  ins_encode %{
    __ mulw(as_Register($dst$$reg),
            as_Register($src1$$reg),
            as_Register($src2$$reg));
  %}

  ins_pipe(imul_reg_reg);
%}

instruct smulI(iRegLNoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (MulL (ConvI2L src1) (ConvI2L src2)));

  ins_cost(INSN_COST * 3);
  format %{ "smull  $dst, $src1, $src2" %}

  ins_encode %{
    __ smull(as_Register($dst$$reg),
             as_Register($src1$$reg),
             as_Register($src2$$reg));
  %}

  ins_pipe(imul_reg_reg);
%}

// Long Multiply

instruct mulL(iRegLNoSp dst, iRegL src1, iRegL src2) %{
  match(Set dst (MulL src1 src2));

  ins_cost(INSN_COST * 5);
  format %{ "mul  $dst, $src1, $src2" %}

  ins_encode %{
    __ mul(as_Register($dst$$reg),
           as_Register($src1$$reg),
           as_Register($src2$$reg));
  %}

  ins_pipe(lmul_reg_reg);
%}

instruct mulHiL_rReg(iRegLNoSp dst, iRegL src1, iRegL src2, rFlagsReg cr)
%{
  match(Set dst (MulHiL src1 src2));

  ins_cost(INSN_COST * 7);
  format %{ "smulh   $dst, $src1, $src2, \t# mulhi" %}

  ins_encode %{
    __ smulh(as_Register($dst$$reg),
             as_Register($src1$$reg),
             as_Register($src2$$reg));
  %}

  ins_pipe(lmul_reg_reg);
%}

// Combined Integer Multiply & Add/Sub

instruct maddI(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2, iRegIorL2I src3) %{
  match(Set dst (AddI src3 (MulI src1 src2)));

  ins_cost(INSN_COST * 3);
  format %{ "madd  $dst, $src1, $src2, $src3" %}

  ins_encode %{
    __ maddw(as_Register($dst$$reg),
             as_Register($src1$$reg),
             as_Register($src2$$reg),
             as_Register($src3$$reg));
  %}

  ins_pipe(imac_reg_reg);
%}

instruct msubI(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2, iRegIorL2I src3) %{
  match(Set dst (SubI src3 (MulI src1 src2)));

  ins_cost(INSN_COST * 3);
  format %{ "msub  $dst, $src1, $src2, $src3" %}

  ins_encode %{
    __ msubw(as_Register($dst$$reg),
             as_Register($src1$$reg),
             as_Register($src2$$reg),
             as_Register($src3$$reg));
  %}

  ins_pipe(imac_reg_reg);
%}

// Combined Long Multiply & Add/Sub

instruct maddL(iRegLNoSp dst, iRegL src1, iRegL src2, iRegL src3) %{
  match(Set dst (AddL src3 (MulL src1 src2)));

  ins_cost(INSN_COST * 5);
  format %{ "madd  $dst, $src1, $src2, $src3" %}

  ins_encode %{
    __ madd(as_Register($dst$$reg),
            as_Register($src1$$reg),
            as_Register($src2$$reg),
            as_Register($src3$$reg));
  %}

  ins_pipe(lmac_reg_reg);
%}

instruct msubL(iRegLNoSp dst, iRegL src1, iRegL src2, iRegL src3) %{
  match(Set dst (SubL src3 (MulL src1 src2)));

  ins_cost(INSN_COST * 5);
  format %{ "msub  $dst, $src1, $src2, $src3" %}

  ins_encode %{
    __ msub(as_Register($dst$$reg),
            as_Register($src1$$reg),
            as_Register($src2$$reg),
            as_Register($src3$$reg));
  %}

  ins_pipe(lmac_reg_reg);
%}

// Integer Divide

instruct divI(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (DivI src1 src2));

  ins_cost(INSN_COST * 19);
  format %{ "sdivw  $dst, $src1, $src2" %}

  ins_encode(aarch64_enc_divw(dst, src1, src2));
  ins_pipe(idiv_reg_reg);
%}

instruct signExtract(iRegINoSp dst, iRegIorL2I src1, immI_31 div1, immI_31 div2) %{
  match(Set dst (URShiftI (RShiftI src1 div1) div2));
  ins_cost(INSN_COST);
  format %{ "lsrw $dst, $src1, $div1" %}
  ins_encode %{
    __ lsrw(as_Register($dst$$reg), as_Register($src1$$reg), 31);
  %}
  ins_pipe(ialu_reg_shift);
%}

instruct div2Round(iRegINoSp dst, iRegIorL2I src, immI_31 div1, immI_31 div2) %{
  match(Set dst (AddI src (URShiftI (RShiftI src div1) div2)));
  ins_cost(INSN_COST);
  format %{ "addw $dst, $src, LSR $div1" %}

  ins_encode %{
    __ addw(as_Register($dst$$reg),
              as_Register($src$$reg),
              as_Register($src$$reg),
              Assembler::LSR, 31);
  %}
  ins_pipe(ialu_reg);
%}

// Long Divide

instruct divL(iRegLNoSp dst, iRegL src1, iRegL src2) %{
  match(Set dst (DivL src1 src2));

  ins_cost(INSN_COST * 35);
  format %{ "sdiv   $dst, $src1, $src2" %}

  ins_encode(aarch64_enc_div(dst, src1, src2));
  ins_pipe(ldiv_reg_reg);
%}

instruct signExtractL(iRegLNoSp dst, iRegL src1, immL_63 div1, immL_63 div2) %{
  match(Set dst (URShiftL (RShiftL src1 div1) div2));
  ins_cost(INSN_COST);
  format %{ "lsr $dst, $src1, $div1" %}
  ins_encode %{
    __ lsr(as_Register($dst$$reg), as_Register($src1$$reg), 63);
  %}
  ins_pipe(ialu_reg_shift);
%}

instruct div2RoundL(iRegLNoSp dst, iRegL src, immL_63 div1, immL_63 div2) %{
  match(Set dst (AddL src (URShiftL (RShiftL src div1) div2)));
  ins_cost(INSN_COST);
  format %{ "add $dst, $src, $div1" %}

  ins_encode %{
    __ add(as_Register($dst$$reg),
              as_Register($src$$reg),
              as_Register($src$$reg),
              Assembler::LSR, 63);
  %}
  ins_pipe(ialu_reg);
%}

// Integer Remainder

instruct modI(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (ModI src1 src2));

  ins_cost(INSN_COST * 22);
  format %{ "sdivw  rscratch1, $src1, $src2\n\t"
            "msubw($dst, rscratch1, $src2, $src1" %}

  ins_encode(aarch64_enc_modw(dst, src1, src2));
  ins_pipe(idiv_reg_reg);
%}

// Long Remainder

instruct modL(iRegLNoSp dst, iRegL src1, iRegL src2) %{
  match(Set dst (ModL src1 src2));

  ins_cost(INSN_COST * 38);
  format %{ "sdiv   rscratch1, $src1, $src2\n"
            "msub($dst, rscratch1, $src2, $src1" %}

  ins_encode(aarch64_enc_mod(dst, src1, src2));
  ins_pipe(ldiv_reg_reg);
%}

// Integer Shifts

// Shift Left Register
instruct lShiftI_reg_reg(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (LShiftI src1 src2));

  ins_cost(INSN_COST * 2);
  format %{ "lslvw  $dst, $src1, $src2" %}

  ins_encode %{
    __ lslvw(as_Register($dst$$reg),
             as_Register($src1$$reg),
             as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg_vshift);
%}

// Shift Left Immediate
instruct lShiftI_reg_imm(iRegINoSp dst, iRegIorL2I src1, immI src2) %{
  match(Set dst (LShiftI src1 src2));

  ins_cost(INSN_COST);
  format %{ "lslw $dst, $src1, ($src2 & 0x1f)" %}

  ins_encode %{
    __ lslw(as_Register($dst$$reg),
            as_Register($src1$$reg),
            $src2$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_shift);
%}

// Shift Right Logical Register
instruct urShiftI_reg_reg(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (URShiftI src1 src2));

  ins_cost(INSN_COST * 2);
  format %{ "lsrvw  $dst, $src1, $src2" %}

  ins_encode %{
    __ lsrvw(as_Register($dst$$reg),
             as_Register($src1$$reg),
             as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg_vshift);
%}

// Shift Right Logical Immediate
instruct urShiftI_reg_imm(iRegINoSp dst, iRegIorL2I src1, immI src2) %{
  match(Set dst (URShiftI src1 src2));

  ins_cost(INSN_COST);
  format %{ "lsrw $dst, $src1, ($src2 & 0x1f)" %}

  ins_encode %{
    __ lsrw(as_Register($dst$$reg),
            as_Register($src1$$reg),
            $src2$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_shift);
%}

// Shift Right Arithmetic Register
instruct rShiftI_reg_reg(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (RShiftI src1 src2));

  ins_cost(INSN_COST * 2);
  format %{ "asrvw  $dst, $src1, $src2" %}

  ins_encode %{
    __ asrvw(as_Register($dst$$reg),
             as_Register($src1$$reg),
             as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg_vshift);
%}

// Shift Right Arithmetic Immediate
instruct rShiftI_reg_imm(iRegINoSp dst, iRegIorL2I src1, immI src2) %{
  match(Set dst (RShiftI src1 src2));

  ins_cost(INSN_COST);
  format %{ "asrw $dst, $src1, ($src2 & 0x1f)" %}

  ins_encode %{
    __ asrw(as_Register($dst$$reg),
            as_Register($src1$$reg),
            $src2$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_shift);
%}

// Combined Int Mask and Right Shift (using UBFM)
// TODO

// Long Shifts

// Shift Left Register
instruct lShiftL_reg_reg(iRegLNoSp dst, iRegL src1, iRegIorL2I src2) %{
  match(Set dst (LShiftL src1 src2));

  ins_cost(INSN_COST * 2);
  format %{ "lslv  $dst, $src1, $src2" %}

  ins_encode %{
    __ lslv(as_Register($dst$$reg),
            as_Register($src1$$reg),
            as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg_vshift);
%}

// Shift Left Immediate
instruct lShiftL_reg_imm(iRegLNoSp dst, iRegL src1, immI src2) %{
  match(Set dst (LShiftL src1 src2));

  ins_cost(INSN_COST);
  format %{ "lsl $dst, $src1, ($src2 & 0x3f)" %}

  ins_encode %{
    __ lsl(as_Register($dst$$reg),
            as_Register($src1$$reg),
            $src2$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_shift);
%}

// Shift Right Logical Register
instruct urShiftL_reg_reg(iRegLNoSp dst, iRegL src1, iRegIorL2I src2) %{
  match(Set dst (URShiftL src1 src2));

  ins_cost(INSN_COST * 2);
  format %{ "lsrv  $dst, $src1, $src2" %}

  ins_encode %{
    __ lsrv(as_Register($dst$$reg),
            as_Register($src1$$reg),
            as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg_vshift);
%}

// Shift Right Logical Immediate
instruct urShiftL_reg_imm(iRegLNoSp dst, iRegL src1, immI src2) %{
  match(Set dst (URShiftL src1 src2));

  ins_cost(INSN_COST);
  format %{ "lsr $dst, $src1, ($src2 & 0x3f)" %}

  ins_encode %{
    __ lsr(as_Register($dst$$reg),
           as_Register($src1$$reg),
           $src2$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_shift);
%}

// A special-case pattern for card table stores.
instruct urShiftP_reg_imm(iRegLNoSp dst, iRegP src1, immI src2) %{
  match(Set dst (URShiftL (CastP2X src1) src2));

  ins_cost(INSN_COST);
  format %{ "lsr $dst, p2x($src1), ($src2 & 0x3f)" %}

  ins_encode %{
    __ lsr(as_Register($dst$$reg),
           as_Register($src1$$reg),
           $src2$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_shift);
%}

// Shift Right Arithmetic Register
instruct rShiftL_reg_reg(iRegLNoSp dst, iRegL src1, iRegIorL2I src2) %{
  match(Set dst (RShiftL src1 src2));

  ins_cost(INSN_COST * 2);
  format %{ "asrv  $dst, $src1, $src2" %}

  ins_encode %{
    __ asrv(as_Register($dst$$reg),
            as_Register($src1$$reg),
            as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg_vshift);
%}

// Shift Right Arithmetic Immediate
instruct rShiftL_reg_imm(iRegLNoSp dst, iRegL src1, immI src2) %{
  match(Set dst (RShiftL src1 src2));

  ins_cost(INSN_COST);
  format %{ "asr $dst, $src1, ($src2 & 0x3f)" %}

  ins_encode %{
    __ asr(as_Register($dst$$reg),
           as_Register($src1$$reg),
           $src2$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_shift);
%}

// BEGIN This section of the file is automatically generated. Do not edit --------------

instruct regL_not_reg(iRegLNoSp dst,
                         iRegL src1, immL_M1 m1,
                         rFlagsReg cr) %{
  match(Set dst (XorL src1 m1));
  ins_cost(INSN_COST);
  format %{ "eon  $dst, $src1, zr" %}

  ins_encode %{
    __ eon(as_Register($dst$$reg),
              as_Register($src1$$reg),
              zr,
              Assembler::LSL, 0);
  %}

  ins_pipe(ialu_reg);
%}
instruct regI_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, immI_M1 m1,
                         rFlagsReg cr) %{
  match(Set dst (XorI src1 m1));
  ins_cost(INSN_COST);
  format %{ "eonw  $dst, $src1, zr" %}

  ins_encode %{
    __ eonw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              zr,
              Assembler::LSL, 0);
  %}

  ins_pipe(ialu_reg);
%}

instruct AndI_reg_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2, immI_M1 m1,
                         rFlagsReg cr) %{
  match(Set dst (AndI src1 (XorI src2 m1)));
  ins_cost(INSN_COST);
  format %{ "bicw  $dst, $src1, $src2" %}

  ins_encode %{
    __ bicw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL, 0);
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct AndL_reg_not_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2, immL_M1 m1,
                         rFlagsReg cr) %{
  match(Set dst (AndL src1 (XorL src2 m1)));
  ins_cost(INSN_COST);
  format %{ "bic  $dst, $src1, $src2" %}

  ins_encode %{
    __ bic(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL, 0);
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct OrI_reg_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2, immI_M1 m1,
                         rFlagsReg cr) %{
  match(Set dst (OrI src1 (XorI src2 m1)));
  ins_cost(INSN_COST);
  format %{ "ornw  $dst, $src1, $src2" %}

  ins_encode %{
    __ ornw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL, 0);
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct OrL_reg_not_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2, immL_M1 m1,
                         rFlagsReg cr) %{
  match(Set dst (OrL src1 (XorL src2 m1)));
  ins_cost(INSN_COST);
  format %{ "orn  $dst, $src1, $src2" %}

  ins_encode %{
    __ orn(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL, 0);
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct XorI_reg_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2, immI_M1 m1,
                         rFlagsReg cr) %{
  match(Set dst (XorI m1 (XorI src2 src1)));
  ins_cost(INSN_COST);
  format %{ "eonw  $dst, $src1, $src2" %}

  ins_encode %{
    __ eonw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL, 0);
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct XorL_reg_not_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2, immL_M1 m1,
                         rFlagsReg cr) %{
  match(Set dst (XorL m1 (XorL src2 src1)));
  ins_cost(INSN_COST);
  format %{ "eon  $dst, $src1, $src2" %}

  ins_encode %{
    __ eon(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL, 0);
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct AndI_reg_URShift_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, immI_M1 src4, rFlagsReg cr) %{
  match(Set dst (AndI src1 (XorI(URShiftI src2 src3) src4)));
  ins_cost(1.9 * INSN_COST);
  format %{ "bicw  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ bicw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AndL_reg_URShift_not_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, immL_M1 src4, rFlagsReg cr) %{
  match(Set dst (AndL src1 (XorL(URShiftL src2 src3) src4)));
  ins_cost(1.9 * INSN_COST);
  format %{ "bic  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ bic(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AndI_reg_RShift_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, immI_M1 src4, rFlagsReg cr) %{
  match(Set dst (AndI src1 (XorI(RShiftI src2 src3) src4)));
  ins_cost(1.9 * INSN_COST);
  format %{ "bicw  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ bicw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AndL_reg_RShift_not_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, immL_M1 src4, rFlagsReg cr) %{
  match(Set dst (AndL src1 (XorL(RShiftL src2 src3) src4)));
  ins_cost(1.9 * INSN_COST);
  format %{ "bic  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ bic(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AndI_reg_LShift_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, immI_M1 src4, rFlagsReg cr) %{
  match(Set dst (AndI src1 (XorI(LShiftI src2 src3) src4)));
  ins_cost(1.9 * INSN_COST);
  format %{ "bicw  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ bicw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AndL_reg_LShift_not_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, immL_M1 src4, rFlagsReg cr) %{
  match(Set dst (AndL src1 (XorL(LShiftL src2 src3) src4)));
  ins_cost(1.9 * INSN_COST);
  format %{ "bic  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ bic(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct XorI_reg_URShift_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, immI_M1 src4, rFlagsReg cr) %{
  match(Set dst (XorI src4 (XorI(URShiftI src2 src3) src1)));
  ins_cost(1.9 * INSN_COST);
  format %{ "eonw  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ eonw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct XorL_reg_URShift_not_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, immL_M1 src4, rFlagsReg cr) %{
  match(Set dst (XorL src4 (XorL(URShiftL src2 src3) src1)));
  ins_cost(1.9 * INSN_COST);
  format %{ "eon  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ eon(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct XorI_reg_RShift_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, immI_M1 src4, rFlagsReg cr) %{
  match(Set dst (XorI src4 (XorI(RShiftI src2 src3) src1)));
  ins_cost(1.9 * INSN_COST);
  format %{ "eonw  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ eonw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct XorL_reg_RShift_not_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, immL_M1 src4, rFlagsReg cr) %{
  match(Set dst (XorL src4 (XorL(RShiftL src2 src3) src1)));
  ins_cost(1.9 * INSN_COST);
  format %{ "eon  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ eon(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct XorI_reg_LShift_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, immI_M1 src4, rFlagsReg cr) %{
  match(Set dst (XorI src4 (XorI(LShiftI src2 src3) src1)));
  ins_cost(1.9 * INSN_COST);
  format %{ "eonw  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ eonw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct XorL_reg_LShift_not_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, immL_M1 src4, rFlagsReg cr) %{
  match(Set dst (XorL src4 (XorL(LShiftL src2 src3) src1)));
  ins_cost(1.9 * INSN_COST);
  format %{ "eon  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ eon(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct OrI_reg_URShift_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, immI_M1 src4, rFlagsReg cr) %{
  match(Set dst (OrI src1 (XorI(URShiftI src2 src3) src4)));
  ins_cost(1.9 * INSN_COST);
  format %{ "ornw  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ ornw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct OrL_reg_URShift_not_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, immL_M1 src4, rFlagsReg cr) %{
  match(Set dst (OrL src1 (XorL(URShiftL src2 src3) src4)));
  ins_cost(1.9 * INSN_COST);
  format %{ "orn  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ orn(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct OrI_reg_RShift_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, immI_M1 src4, rFlagsReg cr) %{
  match(Set dst (OrI src1 (XorI(RShiftI src2 src3) src4)));
  ins_cost(1.9 * INSN_COST);
  format %{ "ornw  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ ornw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct OrL_reg_RShift_not_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, immL_M1 src4, rFlagsReg cr) %{
  match(Set dst (OrL src1 (XorL(RShiftL src2 src3) src4)));
  ins_cost(1.9 * INSN_COST);
  format %{ "orn  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ orn(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct OrI_reg_LShift_not_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, immI_M1 src4, rFlagsReg cr) %{
  match(Set dst (OrI src1 (XorI(LShiftI src2 src3) src4)));
  ins_cost(1.9 * INSN_COST);
  format %{ "ornw  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ ornw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct OrL_reg_LShift_not_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, immL_M1 src4, rFlagsReg cr) %{
  match(Set dst (OrL src1 (XorL(LShiftL src2 src3) src4)));
  ins_cost(1.9 * INSN_COST);
  format %{ "orn  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ orn(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AndI_reg_URShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (AndI src1 (URShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "andw  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ andw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AndL_reg_URShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (AndL src1 (URShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "andr  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ andr(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AndI_reg_RShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (AndI src1 (RShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "andw  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ andw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AndL_reg_RShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (AndL src1 (RShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "andr  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ andr(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AndI_reg_LShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (AndI src1 (LShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "andw  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ andw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AndL_reg_LShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (AndL src1 (LShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "andr  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ andr(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct XorI_reg_URShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (XorI src1 (URShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "eorw  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ eorw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct XorL_reg_URShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (XorL src1 (URShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "eor  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ eor(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct XorI_reg_RShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (XorI src1 (RShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "eorw  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ eorw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct XorL_reg_RShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (XorL src1 (RShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "eor  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ eor(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct XorI_reg_LShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (XorI src1 (LShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "eorw  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ eorw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct XorL_reg_LShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (XorL src1 (LShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "eor  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ eor(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct OrI_reg_URShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (OrI src1 (URShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "orrw  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ orrw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct OrL_reg_URShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (OrL src1 (URShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "orr  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ orr(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct OrI_reg_RShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (OrI src1 (RShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "orrw  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ orrw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct OrL_reg_RShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (OrL src1 (RShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "orr  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ orr(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct OrI_reg_LShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (OrI src1 (LShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "orrw  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ orrw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct OrL_reg_LShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (OrL src1 (LShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "orr  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ orr(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AddI_reg_URShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (AddI src1 (URShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "addw  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ addw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AddL_reg_URShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (AddL src1 (URShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "add  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ add(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AddI_reg_RShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (AddI src1 (RShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "addw  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ addw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AddL_reg_RShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (AddL src1 (RShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "add  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ add(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AddI_reg_LShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (AddI src1 (LShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "addw  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ addw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct AddL_reg_LShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (AddL src1 (LShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "add  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ add(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct SubI_reg_URShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (SubI src1 (URShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "subw  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ subw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct SubL_reg_URShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (SubL src1 (URShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "sub  $dst, $src1, $src2, LSR $src3" %}

  ins_encode %{
    __ sub(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct SubI_reg_RShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (SubI src1 (RShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "subw  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ subw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct SubL_reg_RShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (SubL src1 (RShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "sub  $dst, $src1, $src2, ASR $src3" %}

  ins_encode %{
    __ sub(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::ASR,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct SubI_reg_LShift_reg(iRegINoSp dst,
                         iRegIorL2I src1, iRegIorL2I src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (SubI src1 (LShiftI src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "subw  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ subw(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x1f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}

instruct SubL_reg_LShift_reg(iRegLNoSp dst,
                         iRegL src1, iRegL src2,
                         immI src3, rFlagsReg cr) %{
  match(Set dst (SubL src1 (LShiftL src2 src3)));

  ins_cost(1.9 * INSN_COST);
  format %{ "sub  $dst, $src1, $src2, LSL $src3" %}

  ins_encode %{
    __ sub(as_Register($dst$$reg),
              as_Register($src1$$reg),
              as_Register($src2$$reg),
              Assembler::LSL,
              $src3$$constant & 0x3f);
  %}

  ins_pipe(ialu_reg_reg_shift);
%}



// Shift Left followed by Shift Right.
// This idiom is used by the compiler for the i2b bytecode etc.
instruct sbfmL(iRegLNoSp dst, iRegL src, immI lshift_count, immI rshift_count)
%{
  match(Set dst (RShiftL (LShiftL src lshift_count) rshift_count));
  // Make sure we are not going to exceed what sbfm can do.
  predicate((unsigned int)n->in(2)->get_int() <= 63
            && (unsigned int)n->in(1)->in(2)->get_int() <= 63);

  ins_cost(INSN_COST * 2);
  format %{ "sbfm  $dst, $src, $rshift_count - $lshift_count, #63 - $lshift_count" %}
  ins_encode %{
    int lshift = $lshift_count$$constant, rshift = $rshift_count$$constant;
    int s = 63 - lshift;
    int r = (rshift - lshift) & 63;
    __ sbfm(as_Register($dst$$reg),
            as_Register($src$$reg),
            r, s);
  %}

  ins_pipe(ialu_reg_shift);
%}

// Shift Left followed by Shift Right.
// This idiom is used by the compiler for the i2b bytecode etc.
instruct sbfmwI(iRegINoSp dst, iRegIorL2I src, immI lshift_count, immI rshift_count)
%{
  match(Set dst (RShiftI (LShiftI src lshift_count) rshift_count));
  // Make sure we are not going to exceed what sbfmw can do.
  predicate((unsigned int)n->in(2)->get_int() <= 31
            && (unsigned int)n->in(1)->in(2)->get_int() <= 31);

  ins_cost(INSN_COST * 2);
  format %{ "sbfmw  $dst, $src, $rshift_count - $lshift_count, #31 - $lshift_count" %}
  ins_encode %{
    int lshift = $lshift_count$$constant, rshift = $rshift_count$$constant;
    int s = 31 - lshift;
    int r = (rshift - lshift) & 31;
    __ sbfmw(as_Register($dst$$reg),
            as_Register($src$$reg),
            r, s);
  %}

  ins_pipe(ialu_reg_shift);
%}

// Shift Left followed by Shift Right.
// This idiom is used by the compiler for the i2b bytecode etc.
instruct ubfmL(iRegLNoSp dst, iRegL src, immI lshift_count, immI rshift_count)
%{
  match(Set dst (URShiftL (LShiftL src lshift_count) rshift_count));
  // Make sure we are not going to exceed what ubfm can do.
  predicate((unsigned int)n->in(2)->get_int() <= 63
            && (unsigned int)n->in(1)->in(2)->get_int() <= 63);

  ins_cost(INSN_COST * 2);
  format %{ "ubfm  $dst, $src, $rshift_count - $lshift_count, #63 - $lshift_count" %}
  ins_encode %{
    int lshift = $lshift_count$$constant, rshift = $rshift_count$$constant;
    int s = 63 - lshift;
    int r = (rshift - lshift) & 63;
    __ ubfm(as_Register($dst$$reg),
            as_Register($src$$reg),
            r, s);
  %}

  ins_pipe(ialu_reg_shift);
%}

// Shift Left followed by Shift Right.
// This idiom is used by the compiler for the i2b bytecode etc.
instruct ubfmwI(iRegINoSp dst, iRegIorL2I src, immI lshift_count, immI rshift_count)
%{
  match(Set dst (URShiftI (LShiftI src lshift_count) rshift_count));
  // Make sure we are not going to exceed what ubfmw can do.
  predicate((unsigned int)n->in(2)->get_int() <= 31
            && (unsigned int)n->in(1)->in(2)->get_int() <= 31);

  ins_cost(INSN_COST * 2);
  format %{ "ubfmw  $dst, $src, $rshift_count - $lshift_count, #31 - $lshift_count" %}
  ins_encode %{
    int lshift = $lshift_count$$constant, rshift = $rshift_count$$constant;
    int s = 31 - lshift;
    int r = (rshift - lshift) & 31;
    __ ubfmw(as_Register($dst$$reg),
            as_Register($src$$reg),
            r, s);
  %}

  ins_pipe(ialu_reg_shift);
%}
// Bitfield extract with shift & mask

instruct ubfxwI(iRegINoSp dst, iRegIorL2I src, immI rshift, immI_bitmask mask)
%{
  match(Set dst (AndI (URShiftI src rshift) mask));

  ins_cost(INSN_COST);
  format %{ "ubfxw $dst, $src, $mask" %}
  ins_encode %{
    int rshift = $rshift$$constant;
    long mask = $mask$$constant;
    int width = exact_log2(mask+1);
    __ ubfxw(as_Register($dst$$reg),
            as_Register($src$$reg), rshift, width);
  %}
  ins_pipe(ialu_reg_shift);
%}
instruct ubfxL(iRegLNoSp dst, iRegL src, immI rshift, immL_bitmask mask)
%{
  match(Set dst (AndL (URShiftL src rshift) mask));

  ins_cost(INSN_COST);
  format %{ "ubfx $dst, $src, $mask" %}
  ins_encode %{
    int rshift = $rshift$$constant;
    long mask = $mask$$constant;
    int width = exact_log2(mask+1);
    __ ubfx(as_Register($dst$$reg),
            as_Register($src$$reg), rshift, width);
  %}
  ins_pipe(ialu_reg_shift);
%}

// We can use ubfx when extending an And with a mask when we know mask
// is positive.  We know that because immI_bitmask guarantees it.
instruct ubfxIConvI2L(iRegLNoSp dst, iRegIorL2I src, immI rshift, immI_bitmask mask)
%{
  match(Set dst (ConvI2L (AndI (URShiftI src rshift) mask)));

  ins_cost(INSN_COST * 2);
  format %{ "ubfx $dst, $src, $mask" %}
  ins_encode %{
    int rshift = $rshift$$constant;
    long mask = $mask$$constant;
    int width = exact_log2(mask+1);
    __ ubfx(as_Register($dst$$reg),
            as_Register($src$$reg), rshift, width);
  %}
  ins_pipe(ialu_reg_shift);
%}

// Rotations

instruct extrOrL(iRegLNoSp dst, iRegL src1, iRegL src2, immI lshift, immI rshift, rFlagsReg cr)
%{
  match(Set dst (OrL (LShiftL src1 lshift) (URShiftL src2 rshift)));
  predicate(0 == ((n->in(1)->in(2)->get_int() + n->in(2)->in(2)->get_int()) & 63));

  ins_cost(INSN_COST);
  format %{ "extr $dst, $src1, $src2, #$rshift" %}

  ins_encode %{
    __ extr(as_Register($dst$$reg), as_Register($src1$$reg), as_Register($src2$$reg),
            $rshift$$constant & 63);
  %}
  ins_pipe(ialu_reg_reg_extr);
%}

instruct extrOrI(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2, immI lshift, immI rshift, rFlagsReg cr)
%{
  match(Set dst (OrI (LShiftI src1 lshift) (URShiftI src2 rshift)));
  predicate(0 == ((n->in(1)->in(2)->get_int() + n->in(2)->in(2)->get_int()) & 31));

  ins_cost(INSN_COST);
  format %{ "extr $dst, $src1, $src2, #$rshift" %}

  ins_encode %{
    __ extrw(as_Register($dst$$reg), as_Register($src1$$reg), as_Register($src2$$reg),
            $rshift$$constant & 31);
  %}
  ins_pipe(ialu_reg_reg_extr);
%}

instruct extrAddL(iRegLNoSp dst, iRegL src1, iRegL src2, immI lshift, immI rshift, rFlagsReg cr)
%{
  match(Set dst (AddL (LShiftL src1 lshift) (URShiftL src2 rshift)));
  predicate(0 == ((n->in(1)->in(2)->get_int() + n->in(2)->in(2)->get_int()) & 63));

  ins_cost(INSN_COST);
  format %{ "extr $dst, $src1, $src2, #$rshift" %}

  ins_encode %{
    __ extr(as_Register($dst$$reg), as_Register($src1$$reg), as_Register($src2$$reg),
            $rshift$$constant & 63);
  %}
  ins_pipe(ialu_reg_reg_extr);
%}

instruct extrAddI(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2, immI lshift, immI rshift, rFlagsReg cr)
%{
  match(Set dst (AddI (LShiftI src1 lshift) (URShiftI src2 rshift)));
  predicate(0 == ((n->in(1)->in(2)->get_int() + n->in(2)->in(2)->get_int()) & 31));

  ins_cost(INSN_COST);
  format %{ "extr $dst, $src1, $src2, #$rshift" %}

  ins_encode %{
    __ extrw(as_Register($dst$$reg), as_Register($src1$$reg), as_Register($src2$$reg),
            $rshift$$constant & 31);
  %}
  ins_pipe(ialu_reg_reg_extr);
%}


// rol expander

instruct rolL_rReg(iRegLNoSp dst, iRegL src, iRegI shift, rFlagsReg cr)
%{
  effect(DEF dst, USE src, USE shift);

  format %{ "rol    $dst, $src, $shift" %}
  ins_cost(INSN_COST * 3);
  ins_encode %{
    __ subw(rscratch1, zr, as_Register($shift$$reg));
    __ rorv(as_Register($dst$$reg), as_Register($src$$reg),
            rscratch1);
    %}
  ins_pipe(ialu_reg_reg_vshift);
%}

// rol expander

instruct rolI_rReg(iRegINoSp dst, iRegI src, iRegI shift, rFlagsReg cr)
%{
  effect(DEF dst, USE src, USE shift);

  format %{ "rol    $dst, $src, $shift" %}
  ins_cost(INSN_COST * 3);
  ins_encode %{
    __ subw(rscratch1, zr, as_Register($shift$$reg));
    __ rorvw(as_Register($dst$$reg), as_Register($src$$reg),
            rscratch1);
    %}
  ins_pipe(ialu_reg_reg_vshift);
%}

instruct rolL_rReg_Var_C_64(iRegLNoSp dst, iRegL src, iRegI shift, immI_64 c_64, rFlagsReg cr)
%{
  match(Set dst (OrL (LShiftL src shift) (URShiftL src (SubI c_64 shift))));

  expand %{
    rolL_rReg(dst, src, shift, cr);
  %}
%}

instruct rolL_rReg_Var_C0(iRegLNoSp dst, iRegL src, iRegI shift, immI0 c0, rFlagsReg cr)
%{
  match(Set dst (OrL (LShiftL src shift) (URShiftL src (SubI c0 shift))));

  expand %{
    rolL_rReg(dst, src, shift, cr);
  %}
%}

instruct rolI_rReg_Var_C_32(iRegINoSp dst, iRegI src, iRegI shift, immI_32 c_32, rFlagsReg cr)
%{
  match(Set dst (OrI (LShiftI src shift) (URShiftI src (SubI c_32 shift))));

  expand %{
    rolI_rReg(dst, src, shift, cr);
  %}
%}

instruct rolI_rReg_Var_C0(iRegINoSp dst, iRegI src, iRegI shift, immI0 c0, rFlagsReg cr)
%{
  match(Set dst (OrI (LShiftI src shift) (URShiftI src (SubI c0 shift))));

  expand %{
    rolI_rReg(dst, src, shift, cr);
  %}
%}

// ror expander

instruct rorL_rReg(iRegLNoSp dst, iRegL src, iRegI shift, rFlagsReg cr)
%{
  effect(DEF dst, USE src, USE shift);

  format %{ "ror    $dst, $src, $shift" %}
  ins_cost(INSN_COST);
  ins_encode %{
    __ rorv(as_Register($dst$$reg), as_Register($src$$reg),
            as_Register($shift$$reg));
    %}
  ins_pipe(ialu_reg_reg_vshift);
%}

// ror expander

instruct rorI_rReg(iRegINoSp dst, iRegI src, iRegI shift, rFlagsReg cr)
%{
  effect(DEF dst, USE src, USE shift);

  format %{ "ror    $dst, $src, $shift" %}
  ins_cost(INSN_COST);
  ins_encode %{
    __ rorvw(as_Register($dst$$reg), as_Register($src$$reg),
            as_Register($shift$$reg));
    %}
  ins_pipe(ialu_reg_reg_vshift);
%}

instruct rorL_rReg_Var_C_64(iRegLNoSp dst, iRegL src, iRegI shift, immI_64 c_64, rFlagsReg cr)
%{
  match(Set dst (OrL (URShiftL src shift) (LShiftL src (SubI c_64 shift))));

  expand %{
    rorL_rReg(dst, src, shift, cr);
  %}
%}

instruct rorL_rReg_Var_C0(iRegLNoSp dst, iRegL src, iRegI shift, immI0 c0, rFlagsReg cr)
%{
  match(Set dst (OrL (URShiftL src shift) (LShiftL src (SubI c0 shift))));

  expand %{
    rorL_rReg(dst, src, shift, cr);
  %}
%}

instruct rorI_rReg_Var_C_32(iRegINoSp dst, iRegI src, iRegI shift, immI_32 c_32, rFlagsReg cr)
%{
  match(Set dst (OrI (URShiftI src shift) (LShiftI src (SubI c_32 shift))));

  expand %{
    rorI_rReg(dst, src, shift, cr);
  %}
%}

instruct rorI_rReg_Var_C0(iRegINoSp dst, iRegI src, iRegI shift, immI0 c0, rFlagsReg cr)
%{
  match(Set dst (OrI (URShiftI src shift) (LShiftI src (SubI c0 shift))));

  expand %{
    rorI_rReg(dst, src, shift, cr);
  %}
%}

// Add/subtract (extended)

instruct AddExtI(iRegLNoSp dst, iRegL src1, iRegIorL2I src2, rFlagsReg cr)
%{
  match(Set dst (AddL src1 (ConvI2L src2)));
  ins_cost(INSN_COST);
  format %{ "add  $dst, $src1, sxtw $src2" %}

   ins_encode %{
     __ add(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::sxtw);
   %}
  ins_pipe(ialu_reg_reg);
%};

instruct SubExtI(iRegLNoSp dst, iRegL src1, iRegIorL2I src2, rFlagsReg cr)
%{
  match(Set dst (SubL src1 (ConvI2L src2)));
  ins_cost(INSN_COST);
  format %{ "sub  $dst, $src1, sxtw $src2" %}

   ins_encode %{
     __ sub(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::sxtw);
   %}
  ins_pipe(ialu_reg_reg);
%};


instruct AddExtI_sxth(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2, immI_16 lshift, immI_16 rshift, rFlagsReg cr)
%{
  match(Set dst (AddI src1 (RShiftI (LShiftI src2 lshift) rshift)));
  ins_cost(INSN_COST);
  format %{ "add  $dst, $src1, sxth $src2" %}

   ins_encode %{
     __ add(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::sxth);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct AddExtI_sxtb(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2, immI_24 lshift, immI_24 rshift, rFlagsReg cr)
%{
  match(Set dst (AddI src1 (RShiftI (LShiftI src2 lshift) rshift)));
  ins_cost(INSN_COST);
  format %{ "add  $dst, $src1, sxtb $src2" %}

   ins_encode %{
     __ add(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::sxtb);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct AddExtI_uxtb(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2, immI_24 lshift, immI_24 rshift, rFlagsReg cr)
%{
  match(Set dst (AddI src1 (URShiftI (LShiftI src2 lshift) rshift)));
  ins_cost(INSN_COST);
  format %{ "add  $dst, $src1, uxtb $src2" %}

   ins_encode %{
     __ add(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::uxtb);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct AddExtL_sxth(iRegLNoSp dst, iRegL src1, iRegL src2, immI_48 lshift, immI_48 rshift, rFlagsReg cr)
%{
  match(Set dst (AddL src1 (RShiftL (LShiftL src2 lshift) rshift)));
  ins_cost(INSN_COST);
  format %{ "add  $dst, $src1, sxth $src2" %}

   ins_encode %{
     __ add(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::sxth);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct AddExtL_sxtw(iRegLNoSp dst, iRegL src1, iRegL src2, immI_32 lshift, immI_32 rshift, rFlagsReg cr)
%{
  match(Set dst (AddL src1 (RShiftL (LShiftL src2 lshift) rshift)));
  ins_cost(INSN_COST);
  format %{ "add  $dst, $src1, sxtw $src2" %}

   ins_encode %{
     __ add(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::sxtw);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct AddExtL_sxtb(iRegLNoSp dst, iRegL src1, iRegL src2, immI_56 lshift, immI_56 rshift, rFlagsReg cr)
%{
  match(Set dst (AddL src1 (RShiftL (LShiftL src2 lshift) rshift)));
  ins_cost(INSN_COST);
  format %{ "add  $dst, $src1, sxtb $src2" %}

   ins_encode %{
     __ add(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::sxtb);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct AddExtL_uxtb(iRegLNoSp dst, iRegL src1, iRegL src2, immI_56 lshift, immI_56 rshift, rFlagsReg cr)
%{
  match(Set dst (AddL src1 (URShiftL (LShiftL src2 lshift) rshift)));
  ins_cost(INSN_COST);
  format %{ "add  $dst, $src1, uxtb $src2" %}

   ins_encode %{
     __ add(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::uxtb);
   %}
  ins_pipe(ialu_reg_reg);
%}


instruct AddExtI_uxtb_and(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2, immI_255 mask, rFlagsReg cr)
%{
  match(Set dst (AddI src1 (AndI src2 mask)));
  ins_cost(INSN_COST);
  format %{ "addw  $dst, $src1, $src2, uxtb" %}

   ins_encode %{
     __ addw(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::uxtb);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct AddExtI_uxth_and(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2, immI_65535 mask, rFlagsReg cr)
%{
  match(Set dst (AddI src1 (AndI src2 mask)));
  ins_cost(INSN_COST);
  format %{ "addw  $dst, $src1, $src2, uxth" %}

   ins_encode %{
     __ addw(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::uxth);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct AddExtL_uxtb_and(iRegLNoSp dst, iRegL src1, iRegL src2, immL_255 mask, rFlagsReg cr)
%{
  match(Set dst (AddL src1 (AndL src2 mask)));
  ins_cost(INSN_COST);
  format %{ "add  $dst, $src1, $src2, uxtb" %}

   ins_encode %{
     __ add(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::uxtb);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct AddExtL_uxth_and(iRegLNoSp dst, iRegL src1, iRegL src2, immL_65535 mask, rFlagsReg cr)
%{
  match(Set dst (AddL src1 (AndL src2 mask)));
  ins_cost(INSN_COST);
  format %{ "add  $dst, $src1, $src2, uxth" %}

   ins_encode %{
     __ add(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::uxth);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct AddExtL_uxtw_and(iRegLNoSp dst, iRegL src1, iRegL src2, immL_4294967295 mask, rFlagsReg cr)
%{
  match(Set dst (AddL src1 (AndL src2 mask)));
  ins_cost(INSN_COST);
  format %{ "add  $dst, $src1, $src2, uxtw" %}

   ins_encode %{
     __ add(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::uxtw);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct SubExtI_uxtb_and(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2, immI_255 mask, rFlagsReg cr)
%{
  match(Set dst (SubI src1 (AndI src2 mask)));
  ins_cost(INSN_COST);
  format %{ "subw  $dst, $src1, $src2, uxtb" %}

   ins_encode %{
     __ subw(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::uxtb);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct SubExtI_uxth_and(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2, immI_65535 mask, rFlagsReg cr)
%{
  match(Set dst (SubI src1 (AndI src2 mask)));
  ins_cost(INSN_COST);
  format %{ "subw  $dst, $src1, $src2, uxth" %}

   ins_encode %{
     __ subw(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::uxth);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct SubExtL_uxtb_and(iRegLNoSp dst, iRegL src1, iRegL src2, immL_255 mask, rFlagsReg cr)
%{
  match(Set dst (SubL src1 (AndL src2 mask)));
  ins_cost(INSN_COST);
  format %{ "sub  $dst, $src1, $src2, uxtb" %}

   ins_encode %{
     __ sub(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::uxtb);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct SubExtL_uxth_and(iRegLNoSp dst, iRegL src1, iRegL src2, immL_65535 mask, rFlagsReg cr)
%{
  match(Set dst (SubL src1 (AndL src2 mask)));
  ins_cost(INSN_COST);
  format %{ "sub  $dst, $src1, $src2, uxth" %}

   ins_encode %{
     __ sub(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::uxth);
   %}
  ins_pipe(ialu_reg_reg);
%}

instruct SubExtL_uxtw_and(iRegLNoSp dst, iRegL src1, iRegL src2, immL_4294967295 mask, rFlagsReg cr)
%{
  match(Set dst (SubL src1 (AndL src2 mask)));
  ins_cost(INSN_COST);
  format %{ "sub  $dst, $src1, $src2, uxtw" %}

   ins_encode %{
     __ sub(as_Register($dst$$reg), as_Register($src1$$reg),
            as_Register($src2$$reg), ext::uxtw);
   %}
  ins_pipe(ialu_reg_reg);
%}

// END This section of the file is automatically generated. Do not edit --------------

// ============================================================================
// Floating Point Arithmetic Instructions

instruct addF_reg_reg(vRegF dst, vRegF src1, vRegF src2) %{
  match(Set dst (AddF src1 src2));

  ins_cost(INSN_COST * 5);
  format %{ "fadds   $dst, $src1, $src2" %}

  ins_encode %{
    __ fadds(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg),
             as_FloatRegister($src2$$reg));
  %}

  ins_pipe(fp_dop_reg_reg_s);
%}

instruct addD_reg_reg(vRegD dst, vRegD src1, vRegD src2) %{
  match(Set dst (AddD src1 src2));

  ins_cost(INSN_COST * 5);
  format %{ "faddd   $dst, $src1, $src2" %}

  ins_encode %{
    __ faddd(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg),
             as_FloatRegister($src2$$reg));
  %}

  ins_pipe(fp_dop_reg_reg_d);
%}

instruct subF_reg_reg(vRegF dst, vRegF src1, vRegF src2) %{
  match(Set dst (SubF src1 src2));

  ins_cost(INSN_COST * 5);
  format %{ "fsubs   $dst, $src1, $src2" %}

  ins_encode %{
    __ fsubs(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg),
             as_FloatRegister($src2$$reg));
  %}

  ins_pipe(fp_dop_reg_reg_s);
%}

instruct subD_reg_reg(vRegD dst, vRegD src1, vRegD src2) %{
  match(Set dst (SubD src1 src2));

  ins_cost(INSN_COST * 5);
  format %{ "fsubd   $dst, $src1, $src2" %}

  ins_encode %{
    __ fsubd(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg),
             as_FloatRegister($src2$$reg));
  %}

  ins_pipe(fp_dop_reg_reg_d);
%}

instruct mulF_reg_reg(vRegF dst, vRegF src1, vRegF src2) %{
  match(Set dst (MulF src1 src2));

  ins_cost(INSN_COST * 6);
  format %{ "fmuls   $dst, $src1, $src2" %}

  ins_encode %{
    __ fmuls(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg),
             as_FloatRegister($src2$$reg));
  %}

  ins_pipe(fp_dop_reg_reg_s);
%}

instruct mulD_reg_reg(vRegD dst, vRegD src1, vRegD src2) %{
  match(Set dst (MulD src1 src2));

  ins_cost(INSN_COST * 6);
  format %{ "fmuld   $dst, $src1, $src2" %}

  ins_encode %{
    __ fmuld(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg),
             as_FloatRegister($src2$$reg));
  %}

  ins_pipe(fp_dop_reg_reg_d);
%}

// src1 * src2 + src3
instruct maddF_reg_reg(vRegF dst, vRegF src1, vRegF src2, vRegF src3) %{
  predicate(UseFMA);
  match(Set dst (FmaF src3 (Binary src1 src2)));

  format %{ "fmadds   $dst, $src1, $src2, $src3" %}

  ins_encode %{
    __ fmadds(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg),
             as_FloatRegister($src2$$reg),
             as_FloatRegister($src3$$reg));
  %}

  ins_pipe(pipe_class_default);
%}

// src1 * src2 + src3
instruct maddD_reg_reg(vRegD dst, vRegD src1, vRegD src2, vRegD src3) %{
  predicate(UseFMA);
  match(Set dst (FmaD src3 (Binary src1 src2)));

  format %{ "fmaddd   $dst, $src1, $src2, $src3" %}

  ins_encode %{
    __ fmaddd(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg),
             as_FloatRegister($src2$$reg),
             as_FloatRegister($src3$$reg));
  %}

  ins_pipe(pipe_class_default);
%}

// -src1 * src2 + src3
instruct msubF_reg_reg(vRegF dst, vRegF src1, vRegF src2, vRegF src3) %{
  predicate(UseFMA);
  match(Set dst (FmaF src3 (Binary (NegF src1) src2)));
  match(Set dst (FmaF src3 (Binary src1 (NegF src2))));

  format %{ "fmsubs   $dst, $src1, $src2, $src3" %}

  ins_encode %{
    __ fmsubs(as_FloatRegister($dst$$reg),
              as_FloatRegister($src1$$reg),
              as_FloatRegister($src2$$reg),
              as_FloatRegister($src3$$reg));
  %}

  ins_pipe(pipe_class_default);
%}

// -src1 * src2 + src3
instruct msubD_reg_reg(vRegD dst, vRegD src1, vRegD src2, vRegD src3) %{
  predicate(UseFMA);
  match(Set dst (FmaD src3 (Binary (NegD src1) src2)));
  match(Set dst (FmaD src3 (Binary src1 (NegD src2))));

  format %{ "fmsubd   $dst, $src1, $src2, $src3" %}

  ins_encode %{
    __ fmsubd(as_FloatRegister($dst$$reg),
              as_FloatRegister($src1$$reg),
              as_FloatRegister($src2$$reg),
              as_FloatRegister($src3$$reg));
  %}

  ins_pipe(pipe_class_default);
%}

// -src1 * src2 - src3
instruct mnaddF_reg_reg(vRegF dst, vRegF src1, vRegF src2, vRegF src3) %{
  predicate(UseFMA);
  match(Set dst (FmaF (NegF src3) (Binary (NegF src1) src2)));
  match(Set dst (FmaF (NegF src3) (Binary src1 (NegF src2))));

  format %{ "fnmadds  $dst, $src1, $src2, $src3" %}

  ins_encode %{
    __ fnmadds(as_FloatRegister($dst$$reg),
               as_FloatRegister($src1$$reg),
               as_FloatRegister($src2$$reg),
               as_FloatRegister($src3$$reg));
  %}

  ins_pipe(pipe_class_default);
%}

// -src1 * src2 - src3
instruct mnaddD_reg_reg(vRegD dst, vRegD src1, vRegD src2, vRegD src3) %{
  predicate(UseFMA);
  match(Set dst (FmaD (NegD src3) (Binary (NegD src1) src2)));
  match(Set dst (FmaD (NegD src3) (Binary src1 (NegD src2))));

  format %{ "fnmaddd   $dst, $src1, $src2, $src3" %}

  ins_encode %{
    __ fnmaddd(as_FloatRegister($dst$$reg),
               as_FloatRegister($src1$$reg),
               as_FloatRegister($src2$$reg),
               as_FloatRegister($src3$$reg));
  %}

  ins_pipe(pipe_class_default);
%}

// src1 * src2 - src3
instruct mnsubF_reg_reg(vRegF dst, vRegF src1, vRegF src2, vRegF src3, immF0 zero) %{
  predicate(UseFMA);
  match(Set dst (FmaF (NegF src3) (Binary src1 src2)));

  format %{ "fnmsubs  $dst, $src1, $src2, $src3" %}

  ins_encode %{
    __ fnmsubs(as_FloatRegister($dst$$reg),
               as_FloatRegister($src1$$reg),
               as_FloatRegister($src2$$reg),
               as_FloatRegister($src3$$reg));
  %}

  ins_pipe(pipe_class_default);
%}

// src1 * src2 - src3
instruct mnsubD_reg_reg(vRegD dst, vRegD src1, vRegD src2, vRegD src3, immD0 zero) %{
  predicate(UseFMA);
  match(Set dst (FmaD (NegD src3) (Binary src1 src2)));

  format %{ "fnmsubd   $dst, $src1, $src2, $src3" %}

  ins_encode %{
  // n.b. insn name should be fnmsubd
    __ fnmsub(as_FloatRegister($dst$$reg),
              as_FloatRegister($src1$$reg),
              as_FloatRegister($src2$$reg),
              as_FloatRegister($src3$$reg));
  %}

  ins_pipe(pipe_class_default);
%}


instruct divF_reg_reg(vRegF dst, vRegF src1, vRegF src2) %{
  match(Set dst (DivF src1  src2));

  ins_cost(INSN_COST * 18);
  format %{ "fdivs   $dst, $src1, $src2" %}

  ins_encode %{
    __ fdivs(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg),
             as_FloatRegister($src2$$reg));
  %}

  ins_pipe(fp_div_s);
%}

instruct divD_reg_reg(vRegD dst, vRegD src1, vRegD src2) %{
  match(Set dst (DivD src1  src2));

  ins_cost(INSN_COST * 32);
  format %{ "fdivd   $dst, $src1, $src2" %}

  ins_encode %{
    __ fdivd(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg),
             as_FloatRegister($src2$$reg));
  %}

  ins_pipe(fp_div_d);
%}

instruct negF_reg_reg(vRegF dst, vRegF src) %{
  match(Set dst (NegF src));

  ins_cost(INSN_COST * 3);
  format %{ "fneg   $dst, $src" %}

  ins_encode %{
    __ fnegs(as_FloatRegister($dst$$reg),
             as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_uop_s);
%}

instruct negD_reg_reg(vRegD dst, vRegD src) %{
  match(Set dst (NegD src));

  ins_cost(INSN_COST * 3);
  format %{ "fnegd   $dst, $src" %}

  ins_encode %{
    __ fnegd(as_FloatRegister($dst$$reg),
             as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_uop_d);
%}

instruct absF_reg(vRegF dst, vRegF src) %{
  match(Set dst (AbsF src));

  ins_cost(INSN_COST * 3);
  format %{ "fabss   $dst, $src" %}
  ins_encode %{
    __ fabss(as_FloatRegister($dst$$reg),
             as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_uop_s);
%}

instruct absD_reg(vRegD dst, vRegD src) %{
  match(Set dst (AbsD src));

  ins_cost(INSN_COST * 3);
  format %{ "fabsd   $dst, $src" %}
  ins_encode %{
    __ fabsd(as_FloatRegister($dst$$reg),
             as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_uop_d);
%}

instruct sqrtD_reg(vRegD dst, vRegD src) %{
  match(Set dst (SqrtD src));

  ins_cost(INSN_COST * 50);
  format %{ "fsqrtd  $dst, $src" %}
  ins_encode %{
    __ fsqrtd(as_FloatRegister($dst$$reg),
             as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_div_s);
%}

instruct sqrtF_reg(vRegF dst, vRegF src) %{
  match(Set dst (ConvD2F (SqrtD (ConvF2D src))));

  ins_cost(INSN_COST * 50);
  format %{ "fsqrts  $dst, $src" %}
  ins_encode %{
    __ fsqrts(as_FloatRegister($dst$$reg),
             as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_div_d);
%}

// ============================================================================
// Logical Instructions

// Integer Logical Instructions

// And Instructions


instruct andI_reg_reg(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2, rFlagsReg cr) %{
  match(Set dst (AndI src1 src2));

  format %{ "andw  $dst, $src1, $src2\t# int" %}

  ins_cost(INSN_COST);
  ins_encode %{
    __ andw(as_Register($dst$$reg),
            as_Register($src1$$reg),
            as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct andI_reg_imm(iRegINoSp dst, iRegIorL2I src1, immILog src2, rFlagsReg cr) %{
  match(Set dst (AndI src1 src2));

  format %{ "andsw  $dst, $src1, $src2\t# int" %}

  ins_cost(INSN_COST);
  ins_encode %{
    __ andw(as_Register($dst$$reg),
            as_Register($src1$$reg),
            (unsigned long)($src2$$constant));
  %}

  ins_pipe(ialu_reg_imm);
%}

// Or Instructions

instruct orI_reg_reg(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (OrI src1 src2));

  format %{ "orrw  $dst, $src1, $src2\t# int" %}

  ins_cost(INSN_COST);
  ins_encode %{
    __ orrw(as_Register($dst$$reg),
            as_Register($src1$$reg),
            as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct orI_reg_imm(iRegINoSp dst, iRegIorL2I src1, immILog src2) %{
  match(Set dst (OrI src1 src2));

  format %{ "orrw  $dst, $src1, $src2\t# int" %}

  ins_cost(INSN_COST);
  ins_encode %{
    __ orrw(as_Register($dst$$reg),
            as_Register($src1$$reg),
            (unsigned long)($src2$$constant));
  %}

  ins_pipe(ialu_reg_imm);
%}

// Xor Instructions

instruct xorI_reg_reg(iRegINoSp dst, iRegIorL2I src1, iRegIorL2I src2) %{
  match(Set dst (XorI src1 src2));

  format %{ "eorw  $dst, $src1, $src2\t# int" %}

  ins_cost(INSN_COST);
  ins_encode %{
    __ eorw(as_Register($dst$$reg),
            as_Register($src1$$reg),
            as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct xorI_reg_imm(iRegINoSp dst, iRegIorL2I src1, immILog src2) %{
  match(Set dst (XorI src1 src2));

  format %{ "eorw  $dst, $src1, $src2\t# int" %}

  ins_cost(INSN_COST);
  ins_encode %{
    __ eorw(as_Register($dst$$reg),
            as_Register($src1$$reg),
            (unsigned long)($src2$$constant));
  %}

  ins_pipe(ialu_reg_imm);
%}

// Long Logical Instructions
// TODO

instruct andL_reg_reg(iRegLNoSp dst, iRegL src1, iRegL src2, rFlagsReg cr) %{
  match(Set dst (AndL src1 src2));

  format %{ "and  $dst, $src1, $src2\t# int" %}

  ins_cost(INSN_COST);
  ins_encode %{
    __ andr(as_Register($dst$$reg),
            as_Register($src1$$reg),
            as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct andL_reg_imm(iRegLNoSp dst, iRegL src1, immLLog src2, rFlagsReg cr) %{
  match(Set dst (AndL src1 src2));

  format %{ "and  $dst, $src1, $src2\t# int" %}

  ins_cost(INSN_COST);
  ins_encode %{
    __ andr(as_Register($dst$$reg),
            as_Register($src1$$reg),
            (unsigned long)($src2$$constant));
  %}

  ins_pipe(ialu_reg_imm);
%}

// Or Instructions

instruct orL_reg_reg(iRegLNoSp dst, iRegL src1, iRegL src2) %{
  match(Set dst (OrL src1 src2));

  format %{ "orr  $dst, $src1, $src2\t# int" %}

  ins_cost(INSN_COST);
  ins_encode %{
    __ orr(as_Register($dst$$reg),
           as_Register($src1$$reg),
           as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct orL_reg_imm(iRegLNoSp dst, iRegL src1, immLLog src2) %{
  match(Set dst (OrL src1 src2));

  format %{ "orr  $dst, $src1, $src2\t# int" %}

  ins_cost(INSN_COST);
  ins_encode %{
    __ orr(as_Register($dst$$reg),
           as_Register($src1$$reg),
           (unsigned long)($src2$$constant));
  %}

  ins_pipe(ialu_reg_imm);
%}

// Xor Instructions

instruct xorL_reg_reg(iRegLNoSp dst, iRegL src1, iRegL src2) %{
  match(Set dst (XorL src1 src2));

  format %{ "eor  $dst, $src1, $src2\t# int" %}

  ins_cost(INSN_COST);
  ins_encode %{
    __ eor(as_Register($dst$$reg),
           as_Register($src1$$reg),
           as_Register($src2$$reg));
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct xorL_reg_imm(iRegLNoSp dst, iRegL src1, immLLog src2) %{
  match(Set dst (XorL src1 src2));

  ins_cost(INSN_COST);
  format %{ "eor  $dst, $src1, $src2\t# int" %}

  ins_encode %{
    __ eor(as_Register($dst$$reg),
           as_Register($src1$$reg),
           (unsigned long)($src2$$constant));
  %}

  ins_pipe(ialu_reg_imm);
%}

instruct convI2L_reg_reg(iRegLNoSp dst, iRegIorL2I src)
%{
  match(Set dst (ConvI2L src));

  ins_cost(INSN_COST);
  format %{ "sxtw  $dst, $src\t# i2l" %}
  ins_encode %{
    __ sbfm($dst$$Register, $src$$Register, 0, 31);
  %}
  ins_pipe(ialu_reg_shift);
%}

// this pattern occurs in bigmath arithmetic
instruct convUI2L_reg_reg(iRegLNoSp dst, iRegIorL2I src, immL_32bits mask)
%{
  match(Set dst (AndL (ConvI2L src) mask));

  ins_cost(INSN_COST);
  format %{ "ubfm  $dst, $src, 0, 31\t# ui2l" %}
  ins_encode %{
    __ ubfm($dst$$Register, $src$$Register, 0, 31);
  %}

  ins_pipe(ialu_reg_shift);
%}

instruct convL2I_reg(iRegINoSp dst, iRegL src) %{
  match(Set dst (ConvL2I src));

  ins_cost(INSN_COST);
  format %{ "movw  $dst, $src \t// l2i" %}

  ins_encode %{
    __ movw(as_Register($dst$$reg), as_Register($src$$reg));
  %}

  ins_pipe(ialu_reg);
%}

instruct convI2B(iRegINoSp dst, iRegIorL2I src, rFlagsReg cr)
%{
  match(Set dst (Conv2B src));
  effect(KILL cr);

  format %{
    "cmpw $src, zr\n\t"
    "cset $dst, ne"
  %}

  ins_encode %{
    __ cmpw(as_Register($src$$reg), zr);
    __ cset(as_Register($dst$$reg), Assembler::NE);
  %}

  ins_pipe(ialu_reg);
%}

instruct convP2B(iRegINoSp dst, iRegP src, rFlagsReg cr)
%{
  match(Set dst (Conv2B src));
  effect(KILL cr);

  format %{
    "cmp  $src, zr\n\t"
    "cset $dst, ne"
  %}

  ins_encode %{
    __ cmp(as_Register($src$$reg), zr);
    __ cset(as_Register($dst$$reg), Assembler::NE);
  %}

  ins_pipe(ialu_reg);
%}

instruct convD2F_reg(vRegF dst, vRegD src) %{
  match(Set dst (ConvD2F src));

  ins_cost(INSN_COST * 5);
  format %{ "fcvtd  $dst, $src \t// d2f" %}

  ins_encode %{
    __ fcvtd(as_FloatRegister($dst$$reg), as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_d2f);
%}

instruct convF2D_reg(vRegD dst, vRegF src) %{
  match(Set dst (ConvF2D src));

  ins_cost(INSN_COST * 5);
  format %{ "fcvts  $dst, $src \t// f2d" %}

  ins_encode %{
    __ fcvts(as_FloatRegister($dst$$reg), as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_f2d);
%}

instruct convF2I_reg_reg(iRegINoSp dst, vRegF src) %{
  match(Set dst (ConvF2I src));

  ins_cost(INSN_COST * 5);
  format %{ "fcvtzsw  $dst, $src \t// f2i" %}

  ins_encode %{
    __ fcvtzsw(as_Register($dst$$reg), as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_f2i);
%}

instruct convF2L_reg_reg(iRegLNoSp dst, vRegF src) %{
  match(Set dst (ConvF2L src));

  ins_cost(INSN_COST * 5);
  format %{ "fcvtzs  $dst, $src \t// f2l" %}

  ins_encode %{
    __ fcvtzs(as_Register($dst$$reg), as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_f2l);
%}

instruct convI2F_reg_reg(vRegF dst, iRegIorL2I src) %{
  match(Set dst (ConvI2F src));

  ins_cost(INSN_COST * 5);
  format %{ "scvtfws  $dst, $src \t// i2f" %}

  ins_encode %{
    __ scvtfws(as_FloatRegister($dst$$reg), as_Register($src$$reg));
  %}

  ins_pipe(fp_i2f);
%}

instruct convL2F_reg_reg(vRegF dst, iRegL src) %{
  match(Set dst (ConvL2F src));

  ins_cost(INSN_COST * 5);
  format %{ "scvtfs  $dst, $src \t// l2f" %}

  ins_encode %{
    __ scvtfs(as_FloatRegister($dst$$reg), as_Register($src$$reg));
  %}

  ins_pipe(fp_l2f);
%}

instruct convD2I_reg_reg(iRegINoSp dst, vRegD src) %{
  match(Set dst (ConvD2I src));

  ins_cost(INSN_COST * 5);
  format %{ "fcvtzdw  $dst, $src \t// d2i" %}

  ins_encode %{
    __ fcvtzdw(as_Register($dst$$reg), as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_d2i);
%}

instruct convD2L_reg_reg(iRegLNoSp dst, vRegD src) %{
  match(Set dst (ConvD2L src));

  ins_cost(INSN_COST * 5);
  format %{ "fcvtzd  $dst, $src \t// d2l" %}

  ins_encode %{
    __ fcvtzd(as_Register($dst$$reg), as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_d2l);
%}

instruct convI2D_reg_reg(vRegD dst, iRegIorL2I src) %{
  match(Set dst (ConvI2D src));

  ins_cost(INSN_COST * 5);
  format %{ "scvtfwd  $dst, $src \t// i2d" %}

  ins_encode %{
    __ scvtfwd(as_FloatRegister($dst$$reg), as_Register($src$$reg));
  %}

  ins_pipe(fp_i2d);
%}

instruct convL2D_reg_reg(vRegD dst, iRegL src) %{
  match(Set dst (ConvL2D src));

  ins_cost(INSN_COST * 5);
  format %{ "scvtfd  $dst, $src \t// l2d" %}

  ins_encode %{
    __ scvtfd(as_FloatRegister($dst$$reg), as_Register($src$$reg));
  %}

  ins_pipe(fp_l2d);
%}

// stack <-> reg and reg <-> reg shuffles with no conversion

instruct MoveF2I_stack_reg(iRegINoSp dst, stackSlotF src) %{

  match(Set dst (MoveF2I src));

  effect(DEF dst, USE src);

  ins_cost(4 * INSN_COST);

  format %{ "ldrw $dst, $src\t# MoveF2I_stack_reg" %}

  ins_encode %{
    __ ldrw($dst$$Register, Address(sp, $src$$disp));
  %}

  ins_pipe(iload_reg_reg);

%}

instruct MoveI2F_stack_reg(vRegF dst, stackSlotI src) %{

  match(Set dst (MoveI2F src));

  effect(DEF dst, USE src);

  ins_cost(4 * INSN_COST);

  format %{ "ldrs $dst, $src\t# MoveI2F_stack_reg" %}

  ins_encode %{
    __ ldrs(as_FloatRegister($dst$$reg), Address(sp, $src$$disp));
  %}

  ins_pipe(pipe_class_memory);

%}

instruct MoveD2L_stack_reg(iRegLNoSp dst, stackSlotD src) %{

  match(Set dst (MoveD2L src));

  effect(DEF dst, USE src);

  ins_cost(4 * INSN_COST);

  format %{ "ldr $dst, $src\t# MoveD2L_stack_reg" %}

  ins_encode %{
    __ ldr($dst$$Register, Address(sp, $src$$disp));
  %}

  ins_pipe(iload_reg_reg);

%}

instruct MoveL2D_stack_reg(vRegD dst, stackSlotL src) %{

  match(Set dst (MoveL2D src));

  effect(DEF dst, USE src);

  ins_cost(4 * INSN_COST);

  format %{ "ldrd $dst, $src\t# MoveL2D_stack_reg" %}

  ins_encode %{
    __ ldrd(as_FloatRegister($dst$$reg), Address(sp, $src$$disp));
  %}

  ins_pipe(pipe_class_memory);

%}

instruct MoveF2I_reg_stack(stackSlotI dst, vRegF src) %{

  match(Set dst (MoveF2I src));

  effect(DEF dst, USE src);

  ins_cost(INSN_COST);

  format %{ "strs $src, $dst\t# MoveF2I_reg_stack" %}

  ins_encode %{
    __ strs(as_FloatRegister($src$$reg), Address(sp, $dst$$disp));
  %}

  ins_pipe(pipe_class_memory);

%}

instruct MoveI2F_reg_stack(stackSlotF dst, iRegI src) %{

  match(Set dst (MoveI2F src));

  effect(DEF dst, USE src);

  ins_cost(INSN_COST);

  format %{ "strw $src, $dst\t# MoveI2F_reg_stack" %}

  ins_encode %{
    __ strw($src$$Register, Address(sp, $dst$$disp));
  %}

  ins_pipe(istore_reg_reg);

%}

instruct MoveD2L_reg_stack(stackSlotL dst, vRegD src) %{

  match(Set dst (MoveD2L src));

  effect(DEF dst, USE src);

  ins_cost(INSN_COST);

  format %{ "strd $dst, $src\t# MoveD2L_reg_stack" %}

  ins_encode %{
    __ strd(as_FloatRegister($src$$reg), Address(sp, $dst$$disp));
  %}

  ins_pipe(pipe_class_memory);

%}

instruct MoveL2D_reg_stack(stackSlotD dst, iRegL src) %{

  match(Set dst (MoveL2D src));

  effect(DEF dst, USE src);

  ins_cost(INSN_COST);

  format %{ "str $src, $dst\t# MoveL2D_reg_stack" %}

  ins_encode %{
    __ str($src$$Register, Address(sp, $dst$$disp));
  %}

  ins_pipe(istore_reg_reg);

%}

instruct MoveF2I_reg_reg(iRegINoSp dst, vRegF src) %{

  match(Set dst (MoveF2I src));

  effect(DEF dst, USE src);

  ins_cost(INSN_COST);

  format %{ "fmovs $dst, $src\t# MoveF2I_reg_reg" %}

  ins_encode %{
    __ fmovs($dst$$Register, as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_f2i);

%}

instruct MoveI2F_reg_reg(vRegF dst, iRegI src) %{

  match(Set dst (MoveI2F src));

  effect(DEF dst, USE src);

  ins_cost(INSN_COST);

  format %{ "fmovs $dst, $src\t# MoveI2F_reg_reg" %}

  ins_encode %{
    __ fmovs(as_FloatRegister($dst$$reg), $src$$Register);
  %}

  ins_pipe(fp_i2f);

%}

instruct MoveD2L_reg_reg(iRegLNoSp dst, vRegD src) %{

  match(Set dst (MoveD2L src));

  effect(DEF dst, USE src);

  ins_cost(INSN_COST);

  format %{ "fmovd $dst, $src\t# MoveD2L_reg_reg" %}

  ins_encode %{
    __ fmovd($dst$$Register, as_FloatRegister($src$$reg));
  %}

  ins_pipe(fp_d2l);

%}

instruct MoveL2D_reg_reg(vRegD dst, iRegL src) %{

  match(Set dst (MoveL2D src));

  effect(DEF dst, USE src);

  ins_cost(INSN_COST);

  format %{ "fmovd $dst, $src\t# MoveL2D_reg_reg" %}

  ins_encode %{
    __ fmovd(as_FloatRegister($dst$$reg), $src$$Register);
  %}

  ins_pipe(fp_l2d);

%}

// ============================================================================
// clearing of an array

instruct clearArray_reg_reg(iRegL_R11 cnt, iRegP_R10 base, Universe dummy, rFlagsReg cr)
%{
  match(Set dummy (ClearArray cnt base));
  effect(USE_KILL cnt, USE_KILL base);

  ins_cost(4 * INSN_COST);
  format %{ "ClearArray $cnt, $base" %}

  ins_encode %{
    __ zero_words($base$$Register, $cnt$$Register);
  %}

  ins_pipe(pipe_class_memory);
%}

instruct clearArray_imm_reg(immL cnt, iRegP_R10 base, iRegL_R11 tmp, Universe dummy, rFlagsReg cr)
%{
  match(Set dummy (ClearArray cnt base));
  effect(USE_KILL base, TEMP tmp);

  ins_cost(4 * INSN_COST);
  format %{ "ClearArray $cnt, $base" %}

  ins_encode %{
    __ zero_words($base$$Register, (u_int64_t)$cnt$$constant);
  %}

  ins_pipe(pipe_class_memory);
%}

// ============================================================================
// Overflow Math Instructions

instruct overflowAddI_reg_reg(rFlagsReg cr, iRegIorL2I op1, iRegIorL2I op2)
%{
  match(Set cr (OverflowAddI op1 op2));

  format %{ "cmnw  $op1, $op2\t# overflow check int" %}
  ins_cost(INSN_COST);
  ins_encode %{
    __ cmnw($op1$$Register, $op2$$Register);
  %}

  ins_pipe(icmp_reg_reg);
%}

instruct overflowAddI_reg_imm(rFlagsReg cr, iRegIorL2I op1, immIAddSub op2)
%{
  match(Set cr (OverflowAddI op1 op2));

  format %{ "cmnw  $op1, $op2\t# overflow check int" %}
  ins_cost(INSN_COST);
  ins_encode %{
    __ cmnw($op1$$Register, $op2$$constant);
  %}

  ins_pipe(icmp_reg_imm);
%}

instruct overflowAddL_reg_reg(rFlagsReg cr, iRegL op1, iRegL op2)
%{
  match(Set cr (OverflowAddL op1 op2));

  format %{ "cmn   $op1, $op2\t# overflow check long" %}
  ins_cost(INSN_COST);
  ins_encode %{
    __ cmn($op1$$Register, $op2$$Register);
  %}

  ins_pipe(icmp_reg_reg);
%}

instruct overflowAddL_reg_imm(rFlagsReg cr, iRegL op1, immLAddSub op2)
%{
  match(Set cr (OverflowAddL op1 op2));

  format %{ "cmn   $op1, $op2\t# overflow check long" %}
  ins_cost(INSN_COST);
  ins_encode %{
    __ cmn($op1$$Register, $op2$$constant);
  %}

  ins_pipe(icmp_reg_imm);
%}

instruct overflowSubI_reg_reg(rFlagsReg cr, iRegIorL2I op1, iRegIorL2I op2)
%{
  match(Set cr (OverflowSubI op1 op2));

  format %{ "cmpw  $op1, $op2\t# overflow check int" %}
  ins_cost(INSN_COST);
  ins_encode %{
    __ cmpw($op1$$Register, $op2$$Register);
  %}

  ins_pipe(icmp_reg_reg);
%}

instruct overflowSubI_reg_imm(rFlagsReg cr, iRegIorL2I op1, immIAddSub op2)
%{
  match(Set cr (OverflowSubI op1 op2));

  format %{ "cmpw  $op1, $op2\t# overflow check int" %}
  ins_cost(INSN_COST);
  ins_encode %{
    __ cmpw($op1$$Register, $op2$$constant);
  %}

  ins_pipe(icmp_reg_imm);
%}

instruct overflowSubL_reg_reg(rFlagsReg cr, iRegL op1, iRegL op2)
%{
  match(Set cr (OverflowSubL op1 op2));

  format %{ "cmp   $op1, $op2\t# overflow check long" %}
  ins_cost(INSN_COST);
  ins_encode %{
    __ cmp($op1$$Register, $op2$$Register);
  %}

  ins_pipe(icmp_reg_reg);
%}

instruct overflowSubL_reg_imm(rFlagsReg cr, iRegL op1, immLAddSub op2)
%{
  match(Set cr (OverflowSubL op1 op2));

  format %{ "cmp   $op1, $op2\t# overflow check long" %}
  ins_cost(INSN_COST);
  ins_encode %{
    __ cmp($op1$$Register, $op2$$constant);
  %}

  ins_pipe(icmp_reg_imm);
%}

instruct overflowNegI_reg(rFlagsReg cr, immI0 zero, iRegIorL2I op1)
%{
  match(Set cr (OverflowSubI zero op1));

  format %{ "cmpw  zr, $op1\t# overflow check int" %}
  ins_cost(INSN_COST);
  ins_encode %{
    __ cmpw(zr, $op1$$Register);
  %}

  ins_pipe(icmp_reg_imm);
%}

instruct overflowNegL_reg(rFlagsReg cr, immI0 zero, iRegL op1)
%{
  match(Set cr (OverflowSubL zero op1));

  format %{ "cmp   zr, $op1\t# overflow check long" %}
  ins_cost(INSN_COST);
  ins_encode %{
    __ cmp(zr, $op1$$Register);
  %}

  ins_pipe(icmp_reg_imm);
%}

instruct overflowMulI_reg(rFlagsReg cr, iRegIorL2I op1, iRegIorL2I op2)
%{
  match(Set cr (OverflowMulI op1 op2));

  format %{ "smull rscratch1, $op1, $op2\t# overflow check int\n\t"
            "cmp   rscratch1, rscratch1, sxtw\n\t"
            "movw  rscratch1, #0x80000000\n\t"
            "cselw rscratch1, rscratch1, zr, NE\n\t"
            "cmpw  rscratch1, #1" %}
  ins_cost(5 * INSN_COST);
  ins_encode %{
    __ smull(rscratch1, $op1$$Register, $op2$$Register);
    __ subs(zr, rscratch1, rscratch1, ext::sxtw);      // NE => overflow
    __ movw(rscratch1, 0x80000000);                    // Develop 0 (EQ),
    __ cselw(rscratch1, rscratch1, zr, Assembler::NE); // or 0x80000000 (NE)
    __ cmpw(rscratch1, 1);                             // 0x80000000 - 1 => VS
  %}

  ins_pipe(pipe_slow);
%}

instruct overflowMulI_reg_branch(cmpOp cmp, iRegIorL2I op1, iRegIorL2I op2, label labl, rFlagsReg cr)
%{
  match(If cmp (OverflowMulI op1 op2));
  predicate(n->in(1)->as_Bool()->_test._test == BoolTest::overflow
            || n->in(1)->as_Bool()->_test._test == BoolTest::no_overflow);
  effect(USE labl, KILL cr);

  format %{ "smull rscratch1, $op1, $op2\t# overflow check int\n\t"
            "cmp   rscratch1, rscratch1, sxtw\n\t"
            "b$cmp   $labl" %}
  ins_cost(3 * INSN_COST); // Branch is rare so treat as INSN_COST
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    __ smull(rscratch1, $op1$$Register, $op2$$Register);
    __ subs(zr, rscratch1, rscratch1, ext::sxtw);      // NE => overflow
    __ br(cond == Assembler::VS ? Assembler::NE : Assembler::EQ, *L);
  %}

  ins_pipe(pipe_serial);
%}

instruct overflowMulL_reg(rFlagsReg cr, iRegL op1, iRegL op2)
%{
  match(Set cr (OverflowMulL op1 op2));

  format %{ "mul   rscratch1, $op1, $op2\t#overflow check long\n\t"
            "smulh rscratch2, $op1, $op2\n\t"
            "cmp   rscratch2, rscratch1, ASR #63\n\t"
            "movw  rscratch1, #0x80000000\n\t"
            "cselw rscratch1, rscratch1, zr, NE\n\t"
            "cmpw  rscratch1, #1" %}
  ins_cost(6 * INSN_COST);
  ins_encode %{
    __ mul(rscratch1, $op1$$Register, $op2$$Register);   // Result bits 0..63
    __ smulh(rscratch2, $op1$$Register, $op2$$Register); // Result bits 64..127
    __ cmp(rscratch2, rscratch1, Assembler::ASR, 63);    // Top is pure sign ext
    __ movw(rscratch1, 0x80000000);                    // Develop 0 (EQ),
    __ cselw(rscratch1, rscratch1, zr, Assembler::NE); // or 0x80000000 (NE)
    __ cmpw(rscratch1, 1);                             // 0x80000000 - 1 => VS
  %}

  ins_pipe(pipe_slow);
%}

instruct overflowMulL_reg_branch(cmpOp cmp, iRegL op1, iRegL op2, label labl, rFlagsReg cr)
%{
  match(If cmp (OverflowMulL op1 op2));
  predicate(n->in(1)->as_Bool()->_test._test == BoolTest::overflow
            || n->in(1)->as_Bool()->_test._test == BoolTest::no_overflow);
  effect(USE labl, KILL cr);

  format %{ "mul   rscratch1, $op1, $op2\t#overflow check long\n\t"
            "smulh rscratch2, $op1, $op2\n\t"
            "cmp   rscratch2, rscratch1, ASR #63\n\t"
            "b$cmp $labl" %}
  ins_cost(4 * INSN_COST); // Branch is rare so treat as INSN_COST
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    __ mul(rscratch1, $op1$$Register, $op2$$Register);   // Result bits 0..63
    __ smulh(rscratch2, $op1$$Register, $op2$$Register); // Result bits 64..127
    __ cmp(rscratch2, rscratch1, Assembler::ASR, 63);    // Top is pure sign ext
    __ br(cond == Assembler::VS ? Assembler::NE : Assembler::EQ, *L);
  %}

  ins_pipe(pipe_serial);
%}

// ============================================================================
// Compare Instructions

instruct compI_reg_reg(rFlagsReg cr, iRegI op1, iRegI op2)
%{
  match(Set cr (CmpI op1 op2));

  effect(DEF cr, USE op1, USE op2);

  ins_cost(INSN_COST);
  format %{ "cmpw  $op1, $op2" %}

  ins_encode(aarch64_enc_cmpw(op1, op2));

  ins_pipe(icmp_reg_reg);
%}

instruct compI_reg_immI0(rFlagsReg cr, iRegI op1, immI0 zero)
%{
  match(Set cr (CmpI op1 zero));

  effect(DEF cr, USE op1);

  ins_cost(INSN_COST);
  format %{ "cmpw $op1, 0" %}

  ins_encode(aarch64_enc_cmpw_imm_addsub(op1, zero));

  ins_pipe(icmp_reg_imm);
%}

instruct compI_reg_immIAddSub(rFlagsReg cr, iRegI op1, immIAddSub op2)
%{
  match(Set cr (CmpI op1 op2));

  effect(DEF cr, USE op1);

  ins_cost(INSN_COST);
  format %{ "cmpw  $op1, $op2" %}

  ins_encode(aarch64_enc_cmpw_imm_addsub(op1, op2));

  ins_pipe(icmp_reg_imm);
%}

instruct compI_reg_immI(rFlagsReg cr, iRegI op1, immI op2)
%{
  match(Set cr (CmpI op1 op2));

  effect(DEF cr, USE op1);

  ins_cost(INSN_COST * 2);
  format %{ "cmpw  $op1, $op2" %}

  ins_encode(aarch64_enc_cmpw_imm(op1, op2));

  ins_pipe(icmp_reg_imm);
%}

// Unsigned compare Instructions; really, same as signed compare
// except it should only be used to feed an If or a CMovI which takes a
// cmpOpU.

instruct compU_reg_reg(rFlagsRegU cr, iRegI op1, iRegI op2)
%{
  match(Set cr (CmpU op1 op2));

  effect(DEF cr, USE op1, USE op2);

  ins_cost(INSN_COST);
  format %{ "cmpw  $op1, $op2\t# unsigned" %}

  ins_encode(aarch64_enc_cmpw(op1, op2));

  ins_pipe(icmp_reg_reg);
%}

instruct compU_reg_immI0(rFlagsRegU cr, iRegI op1, immI0 zero)
%{
  match(Set cr (CmpU op1 zero));

  effect(DEF cr, USE op1);

  ins_cost(INSN_COST);
  format %{ "cmpw $op1, #0\t# unsigned" %}

  ins_encode(aarch64_enc_cmpw_imm_addsub(op1, zero));

  ins_pipe(icmp_reg_imm);
%}

instruct compU_reg_immIAddSub(rFlagsRegU cr, iRegI op1, immIAddSub op2)
%{
  match(Set cr (CmpU op1 op2));

  effect(DEF cr, USE op1);

  ins_cost(INSN_COST);
  format %{ "cmpw  $op1, $op2\t# unsigned" %}

  ins_encode(aarch64_enc_cmpw_imm_addsub(op1, op2));

  ins_pipe(icmp_reg_imm);
%}

instruct compU_reg_immI(rFlagsRegU cr, iRegI op1, immI op2)
%{
  match(Set cr (CmpU op1 op2));

  effect(DEF cr, USE op1);

  ins_cost(INSN_COST * 2);
  format %{ "cmpw  $op1, $op2\t# unsigned" %}

  ins_encode(aarch64_enc_cmpw_imm(op1, op2));

  ins_pipe(icmp_reg_imm);
%}

instruct compL_reg_reg(rFlagsReg cr, iRegL op1, iRegL op2)
%{
  match(Set cr (CmpL op1 op2));

  effect(DEF cr, USE op1, USE op2);

  ins_cost(INSN_COST);
  format %{ "cmp  $op1, $op2" %}

  ins_encode(aarch64_enc_cmp(op1, op2));

  ins_pipe(icmp_reg_reg);
%}

instruct compL_reg_immI0(rFlagsReg cr, iRegL op1, immI0 zero)
%{
  match(Set cr (CmpL op1 zero));

  effect(DEF cr, USE op1);

  ins_cost(INSN_COST);
  format %{ "tst  $op1" %}

  ins_encode(aarch64_enc_cmp_imm_addsub(op1, zero));

  ins_pipe(icmp_reg_imm);
%}

instruct compL_reg_immLAddSub(rFlagsReg cr, iRegL op1, immLAddSub op2)
%{
  match(Set cr (CmpL op1 op2));

  effect(DEF cr, USE op1);

  ins_cost(INSN_COST);
  format %{ "cmp  $op1, $op2" %}

  ins_encode(aarch64_enc_cmp_imm_addsub(op1, op2));

  ins_pipe(icmp_reg_imm);
%}

instruct compL_reg_immL(rFlagsReg cr, iRegL op1, immL op2)
%{
  match(Set cr (CmpL op1 op2));

  effect(DEF cr, USE op1);

  ins_cost(INSN_COST * 2);
  format %{ "cmp  $op1, $op2" %}

  ins_encode(aarch64_enc_cmp_imm(op1, op2));

  ins_pipe(icmp_reg_imm);
%}

instruct compP_reg_reg(rFlagsRegU cr, iRegP op1, iRegP op2)
%{
  match(Set cr (CmpP op1 op2));

  effect(DEF cr, USE op1, USE op2);

  ins_cost(INSN_COST);
  format %{ "cmp  $op1, $op2\t // ptr" %}

  ins_encode(aarch64_enc_cmpp(op1, op2));

  ins_pipe(icmp_reg_reg);
%}

instruct compN_reg_reg(rFlagsRegU cr, iRegN op1, iRegN op2)
%{
  match(Set cr (CmpN op1 op2));

  effect(DEF cr, USE op1, USE op2);

  ins_cost(INSN_COST);
  format %{ "cmp  $op1, $op2\t // compressed ptr" %}

  ins_encode(aarch64_enc_cmpn(op1, op2));

  ins_pipe(icmp_reg_reg);
%}

instruct testP_reg(rFlagsRegU cr, iRegP op1, immP0 zero)
%{
  match(Set cr (CmpP op1 zero));

  effect(DEF cr, USE op1, USE zero);

  ins_cost(INSN_COST);
  format %{ "cmp  $op1, 0\t // ptr" %}

  ins_encode(aarch64_enc_testp(op1));

  ins_pipe(icmp_reg_imm);
%}

instruct testN_reg(rFlagsRegU cr, iRegN op1, immN0 zero)
%{
  match(Set cr (CmpN op1 zero));

  effect(DEF cr, USE op1, USE zero);

  ins_cost(INSN_COST);
  format %{ "cmp  $op1, 0\t // compressed ptr" %}

  ins_encode(aarch64_enc_testn(op1));

  ins_pipe(icmp_reg_imm);
%}

// FP comparisons
//
// n.b. CmpF/CmpD set a normal flags reg which then gets compared
// using normal cmpOp. See declaration of rFlagsReg for details.

instruct compF_reg_reg(rFlagsReg cr, vRegF src1, vRegF src2)
%{
  match(Set cr (CmpF src1 src2));

  ins_cost(3 * INSN_COST);
  format %{ "fcmps $src1, $src2" %}

  ins_encode %{
    __ fcmps(as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
  %}

  ins_pipe(pipe_class_compare);
%}

instruct compF_reg_zero(rFlagsReg cr, vRegF src1, immF0 src2)
%{
  match(Set cr (CmpF src1 src2));

  ins_cost(3 * INSN_COST);
  format %{ "fcmps $src1, 0.0" %}

  ins_encode %{
    __ fcmps(as_FloatRegister($src1$$reg), 0.0D);
  %}

  ins_pipe(pipe_class_compare);
%}
// FROM HERE

instruct compD_reg_reg(rFlagsReg cr, vRegD src1, vRegD src2)
%{
  match(Set cr (CmpD src1 src2));

  ins_cost(3 * INSN_COST);
  format %{ "fcmpd $src1, $src2" %}

  ins_encode %{
    __ fcmpd(as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
  %}

  ins_pipe(pipe_class_compare);
%}

instruct compD_reg_zero(rFlagsReg cr, vRegD src1, immD0 src2)
%{
  match(Set cr (CmpD src1 src2));

  ins_cost(3 * INSN_COST);
  format %{ "fcmpd $src1, 0.0" %}

  ins_encode %{
    __ fcmpd(as_FloatRegister($src1$$reg), 0.0D);
  %}

  ins_pipe(pipe_class_compare);
%}

instruct compF3_reg_reg(iRegINoSp dst, vRegF src1, vRegF src2, rFlagsReg cr)
%{
  match(Set dst (CmpF3 src1 src2));
  effect(KILL cr);

  ins_cost(5 * INSN_COST);
  format %{ "fcmps $src1, $src2\n\t"
            "csinvw($dst, zr, zr, eq\n\t"
            "csnegw($dst, $dst, $dst, lt)"
  %}

  ins_encode %{
    Label done;
    FloatRegister s1 = as_FloatRegister($src1$$reg);
    FloatRegister s2 = as_FloatRegister($src2$$reg);
    Register d = as_Register($dst$$reg);
    __ fcmps(s1, s2);
    // installs 0 if EQ else -1
    __ csinvw(d, zr, zr, Assembler::EQ);
    // keeps -1 if less or unordered else installs 1
    __ csnegw(d, d, d, Assembler::LT);
    __ bind(done);
  %}

  ins_pipe(pipe_class_default);

%}

instruct compD3_reg_reg(iRegINoSp dst, vRegD src1, vRegD src2, rFlagsReg cr)
%{
  match(Set dst (CmpD3 src1 src2));
  effect(KILL cr);

  ins_cost(5 * INSN_COST);
  format %{ "fcmpd $src1, $src2\n\t"
            "csinvw($dst, zr, zr, eq\n\t"
            "csnegw($dst, $dst, $dst, lt)"
  %}

  ins_encode %{
    Label done;
    FloatRegister s1 = as_FloatRegister($src1$$reg);
    FloatRegister s2 = as_FloatRegister($src2$$reg);
    Register d = as_Register($dst$$reg);
    __ fcmpd(s1, s2);
    // installs 0 if EQ else -1
    __ csinvw(d, zr, zr, Assembler::EQ);
    // keeps -1 if less or unordered else installs 1
    __ csnegw(d, d, d, Assembler::LT);
    __ bind(done);
  %}
  ins_pipe(pipe_class_default);

%}

instruct compF3_reg_immF0(iRegINoSp dst, vRegF src1, immF0 zero, rFlagsReg cr)
%{
  match(Set dst (CmpF3 src1 zero));
  effect(KILL cr);

  ins_cost(5 * INSN_COST);
  format %{ "fcmps $src1, 0.0\n\t"
            "csinvw($dst, zr, zr, eq\n\t"
            "csnegw($dst, $dst, $dst, lt)"
  %}

  ins_encode %{
    Label done;
    FloatRegister s1 = as_FloatRegister($src1$$reg);
    Register d = as_Register($dst$$reg);
    __ fcmps(s1, 0.0D);
    // installs 0 if EQ else -1
    __ csinvw(d, zr, zr, Assembler::EQ);
    // keeps -1 if less or unordered else installs 1
    __ csnegw(d, d, d, Assembler::LT);
    __ bind(done);
  %}

  ins_pipe(pipe_class_default);

%}

instruct compD3_reg_immD0(iRegINoSp dst, vRegD src1, immD0 zero, rFlagsReg cr)
%{
  match(Set dst (CmpD3 src1 zero));
  effect(KILL cr);

  ins_cost(5 * INSN_COST);
  format %{ "fcmpd $src1, 0.0\n\t"
            "csinvw($dst, zr, zr, eq\n\t"
            "csnegw($dst, $dst, $dst, lt)"
  %}

  ins_encode %{
    Label done;
    FloatRegister s1 = as_FloatRegister($src1$$reg);
    Register d = as_Register($dst$$reg);
    __ fcmpd(s1, 0.0D);
    // installs 0 if EQ else -1
    __ csinvw(d, zr, zr, Assembler::EQ);
    // keeps -1 if less or unordered else installs 1
    __ csnegw(d, d, d, Assembler::LT);
    __ bind(done);
  %}
  ins_pipe(pipe_class_default);

%}

instruct cmpLTMask_reg_reg(iRegINoSp dst, iRegIorL2I p, iRegIorL2I q, rFlagsReg cr)
%{
  match(Set dst (CmpLTMask p q));
  effect(KILL cr);

  ins_cost(3 * INSN_COST);

  format %{ "cmpw $p, $q\t# cmpLTMask\n\t"
            "csetw $dst, lt\n\t"
            "subw $dst, zr, $dst"
  %}

  ins_encode %{
    __ cmpw(as_Register($p$$reg), as_Register($q$$reg));
    __ csetw(as_Register($dst$$reg), Assembler::LT);
    __ subw(as_Register($dst$$reg), zr, as_Register($dst$$reg));
  %}

  ins_pipe(ialu_reg_reg);
%}

instruct cmpLTMask_reg_zero(iRegINoSp dst, iRegIorL2I src, immI0 zero, rFlagsReg cr)
%{
  match(Set dst (CmpLTMask src zero));
  effect(KILL cr);

  ins_cost(INSN_COST);

  format %{ "asrw $dst, $src, #31\t# cmpLTMask0" %}

  ins_encode %{
    __ asrw(as_Register($dst$$reg), as_Register($src$$reg), 31);
  %}

  ins_pipe(ialu_reg_shift);
%}

// ============================================================================
// Max and Min

instruct minI_rReg(iRegINoSp dst, iRegI src1, iRegI src2, rFlagsReg cr)
%{
  match(Set dst (MinI src1 src2));

  effect(DEF dst, USE src1, USE src2, KILL cr);
  size(8);

  ins_cost(INSN_COST * 3);
  format %{
    "cmpw $src1 $src2\t signed int\n\t"
    "cselw $dst, $src1, $src2 lt\t"
  %}

  ins_encode %{
    __ cmpw(as_Register($src1$$reg),
            as_Register($src2$$reg));
    __ cselw(as_Register($dst$$reg),
             as_Register($src1$$reg),
             as_Register($src2$$reg),
             Assembler::LT);
  %}

  ins_pipe(ialu_reg_reg);
%}
// FROM HERE

instruct maxI_rReg(iRegINoSp dst, iRegI src1, iRegI src2, rFlagsReg cr)
%{
  match(Set dst (MaxI src1 src2));

  effect(DEF dst, USE src1, USE src2, KILL cr);
  size(8);

  ins_cost(INSN_COST * 3);
  format %{
    "cmpw $src1 $src2\t signed int\n\t"
    "cselw $dst, $src1, $src2 gt\t"
  %}

  ins_encode %{
    __ cmpw(as_Register($src1$$reg),
            as_Register($src2$$reg));
    __ cselw(as_Register($dst$$reg),
             as_Register($src1$$reg),
             as_Register($src2$$reg),
             Assembler::GT);
  %}

  ins_pipe(ialu_reg_reg);
%}

// ============================================================================
// Branch Instructions

// Direct Branch.
instruct branch(label lbl)
%{
  match(Goto);

  effect(USE lbl);

  ins_cost(BRANCH_COST);
  format %{ "b  $lbl" %}

  ins_encode(aarch64_enc_b(lbl));

  ins_pipe(pipe_branch);
%}

// Conditional Near Branch
instruct branchCon(cmpOp cmp, rFlagsReg cr, label lbl)
%{
  // Same match rule as `branchConFar'.
  match(If cmp cr);

  effect(USE lbl);

  ins_cost(BRANCH_COST);
  // If set to 1 this indicates that the current instruction is a
  // short variant of a long branch. This avoids using this
  // instruction in first-pass matching. It will then only be used in
  // the `Shorten_branches' pass.
  // ins_short_branch(1);
  format %{ "b$cmp  $lbl" %}

  ins_encode(aarch64_enc_br_con(cmp, lbl));

  ins_pipe(pipe_branch_cond);
%}

// Conditional Near Branch Unsigned
instruct branchConU(cmpOpU cmp, rFlagsRegU cr, label lbl)
%{
  // Same match rule as `branchConFar'.
  match(If cmp cr);

  effect(USE lbl);

  ins_cost(BRANCH_COST);
  // If set to 1 this indicates that the current instruction is a
  // short variant of a long branch. This avoids using this
  // instruction in first-pass matching. It will then only be used in
  // the `Shorten_branches' pass.
  // ins_short_branch(1);
  format %{ "b$cmp  $lbl\t# unsigned" %}

  ins_encode(aarch64_enc_br_conU(cmp, lbl));

  ins_pipe(pipe_branch_cond);
%}

// Make use of CBZ and CBNZ.  These instructions, as well as being
// shorter than (cmp; branch), have the additional benefit of not
// killing the flags.

instruct cmpI_imm0_branch(cmpOpEqNe cmp, iRegIorL2I op1, immI0 op2, label labl, rFlagsReg cr) %{
  match(If cmp (CmpI op1 op2));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "cbw$cmp   $op1, $labl" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    if (cond == Assembler::EQ)
      __ cbzw($op1$$Register, *L);
    else
      __ cbnzw($op1$$Register, *L);
  %}
  ins_pipe(pipe_cmp_branch);
%}

instruct cmpL_imm0_branch(cmpOpEqNe cmp, iRegL op1, immL0 op2, label labl, rFlagsReg cr) %{
  match(If cmp (CmpL op1 op2));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "cb$cmp   $op1, $labl" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    if (cond == Assembler::EQ)
      __ cbz($op1$$Register, *L);
    else
      __ cbnz($op1$$Register, *L);
  %}
  ins_pipe(pipe_cmp_branch);
%}

instruct cmpP_imm0_branch(cmpOpEqNe cmp, iRegP op1, immP0 op2, label labl, rFlagsReg cr) %{
  match(If cmp (CmpP op1 op2));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "cb$cmp   $op1, $labl" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    if (cond == Assembler::EQ)
      __ cbz($op1$$Register, *L);
    else
      __ cbnz($op1$$Register, *L);
  %}
  ins_pipe(pipe_cmp_branch);
%}

instruct cmpN_imm0_branch(cmpOpEqNe cmp, iRegN op1, immN0 op2, label labl, rFlagsReg cr) %{
  match(If cmp (CmpN op1 op2));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "cbw$cmp   $op1, $labl" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    if (cond == Assembler::EQ)
      __ cbzw($op1$$Register, *L);
    else
      __ cbnzw($op1$$Register, *L);
  %}
  ins_pipe(pipe_cmp_branch);
%}

instruct cmpP_narrowOop_imm0_branch(cmpOpEqNe cmp, iRegN oop, immP0 zero, label labl, rFlagsReg cr) %{
  match(If cmp (CmpP (DecodeN oop) zero));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "cb$cmp   $oop, $labl" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    if (cond == Assembler::EQ)
      __ cbzw($oop$$Register, *L);
    else
      __ cbnzw($oop$$Register, *L);
  %}
  ins_pipe(pipe_cmp_branch);
%}

instruct cmpUI_imm0_branch(cmpOpUEqNeLtGe cmp, iRegIorL2I op1, immI0 op2, label labl, rFlagsRegU cr) %{
  match(If cmp (CmpU op1 op2));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "cbw$cmp   $op1, $labl" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    if (cond == Assembler::EQ || cond == Assembler::LS)
      __ cbzw($op1$$Register, *L);
    else
      __ cbnzw($op1$$Register, *L);
  %}
  ins_pipe(pipe_cmp_branch);
%}

instruct cmpUL_imm0_branch(cmpOpUEqNeLtGe cmp, iRegL op1, immL0 op2, label labl, rFlagsRegU cr) %{
  match(If cmp (CmpU op1 op2));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "cb$cmp   $op1, $labl" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    if (cond == Assembler::EQ || cond == Assembler::LS)
      __ cbz($op1$$Register, *L);
    else
      __ cbnz($op1$$Register, *L);
  %}
  ins_pipe(pipe_cmp_branch);
%}

// Test bit and Branch

// Patterns for short (< 32KiB) variants
instruct cmpL_branch_sign(cmpOpLtGe cmp, iRegL op1, immL0 op2, label labl) %{
  match(If cmp (CmpL op1 op2));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "cb$cmp   $op1, $labl # long" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond =
      ((Assembler::Condition)$cmp$$cmpcode == Assembler::LT) ? Assembler::NE : Assembler::EQ;
    __ tbr(cond, $op1$$Register, 63, *L);
  %}
  ins_pipe(pipe_cmp_branch);
  ins_short_branch(1);
%}

instruct cmpI_branch_sign(cmpOpLtGe cmp, iRegIorL2I op1, immI0 op2, label labl) %{
  match(If cmp (CmpI op1 op2));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "cb$cmp   $op1, $labl # int" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond =
      ((Assembler::Condition)$cmp$$cmpcode == Assembler::LT) ? Assembler::NE : Assembler::EQ;
    __ tbr(cond, $op1$$Register, 31, *L);
  %}
  ins_pipe(pipe_cmp_branch);
  ins_short_branch(1);
%}

instruct cmpL_branch_bit(cmpOpEqNe cmp, iRegL op1, immL op2, immL0 op3, label labl) %{
  match(If cmp (CmpL (AndL op1 op2) op3));
  predicate(is_power_of_2(n->in(2)->in(1)->in(2)->get_long()));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "tb$cmp   $op1, $op2, $labl" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    int bit = exact_log2($op2$$constant);
    __ tbr(cond, $op1$$Register, bit, *L);
  %}
  ins_pipe(pipe_cmp_branch);
  ins_short_branch(1);
%}

instruct cmpI_branch_bit(cmpOpEqNe cmp, iRegIorL2I op1, immI op2, immI0 op3, label labl) %{
  match(If cmp (CmpI (AndI op1 op2) op3));
  predicate(is_power_of_2(n->in(2)->in(1)->in(2)->get_int()));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "tb$cmp   $op1, $op2, $labl" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    int bit = exact_log2($op2$$constant);
    __ tbr(cond, $op1$$Register, bit, *L);
  %}
  ins_pipe(pipe_cmp_branch);
  ins_short_branch(1);
%}

// And far variants
instruct far_cmpL_branch_sign(cmpOpLtGe cmp, iRegL op1, immL0 op2, label labl) %{
  match(If cmp (CmpL op1 op2));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "cb$cmp   $op1, $labl # long" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond =
      ((Assembler::Condition)$cmp$$cmpcode == Assembler::LT) ? Assembler::NE : Assembler::EQ;
    __ tbr(cond, $op1$$Register, 63, *L, /*far*/true);
  %}
  ins_pipe(pipe_cmp_branch);
%}

instruct far_cmpI_branch_sign(cmpOpLtGe cmp, iRegIorL2I op1, immI0 op2, label labl) %{
  match(If cmp (CmpI op1 op2));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "cb$cmp   $op1, $labl # int" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond =
      ((Assembler::Condition)$cmp$$cmpcode == Assembler::LT) ? Assembler::NE : Assembler::EQ;
    __ tbr(cond, $op1$$Register, 31, *L, /*far*/true);
  %}
  ins_pipe(pipe_cmp_branch);
%}

instruct far_cmpL_branch_bit(cmpOpEqNe cmp, iRegL op1, immL op2, immL0 op3, label labl) %{
  match(If cmp (CmpL (AndL op1 op2) op3));
  predicate(is_power_of_2(n->in(2)->in(1)->in(2)->get_long()));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "tb$cmp   $op1, $op2, $labl" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    int bit = exact_log2($op2$$constant);
    __ tbr(cond, $op1$$Register, bit, *L, /*far*/true);
  %}
  ins_pipe(pipe_cmp_branch);
%}

instruct far_cmpI_branch_bit(cmpOpEqNe cmp, iRegIorL2I op1, immI op2, immI0 op3, label labl) %{
  match(If cmp (CmpI (AndI op1 op2) op3));
  predicate(is_power_of_2(n->in(2)->in(1)->in(2)->get_int()));
  effect(USE labl);

  ins_cost(BRANCH_COST);
  format %{ "tb$cmp   $op1, $op2, $labl" %}
  ins_encode %{
    Label* L = $labl$$label;
    Assembler::Condition cond = (Assembler::Condition)$cmp$$cmpcode;
    int bit = exact_log2($op2$$constant);
    __ tbr(cond, $op1$$Register, bit, *L, /*far*/true);
  %}
  ins_pipe(pipe_cmp_branch);
%}

// Test bits

instruct cmpL_and(cmpOp cmp, iRegL op1, immL op2, immL0 op3, rFlagsReg cr) %{
  match(Set cr (CmpL (AndL op1 op2) op3));
  predicate(Assembler::operand_valid_for_logical_immediate
            (/*is_32*/false, n->in(1)->in(2)->get_long()));

  ins_cost(INSN_COST);
  format %{ "tst $op1, $op2 # long" %}
  ins_encode %{
    __ tst($op1$$Register, $op2$$constant);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct cmpI_and(cmpOp cmp, iRegIorL2I op1, immI op2, immI0 op3, rFlagsReg cr) %{
  match(Set cr (CmpI (AndI op1 op2) op3));
  predicate(Assembler::operand_valid_for_logical_immediate
            (/*is_32*/true, n->in(1)->in(2)->get_int()));

  ins_cost(INSN_COST);
  format %{ "tst $op1, $op2 # int" %}
  ins_encode %{
    __ tstw($op1$$Register, $op2$$constant);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct cmpL_and_reg(cmpOp cmp, iRegL op1, iRegL op2, immL0 op3, rFlagsReg cr) %{
  match(Set cr (CmpL (AndL op1 op2) op3));

  ins_cost(INSN_COST);
  format %{ "tst $op1, $op2 # long" %}
  ins_encode %{
    __ tst($op1$$Register, $op2$$Register);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct cmpI_and_reg(cmpOp cmp, iRegIorL2I op1, iRegIorL2I op2, immI0 op3, rFlagsReg cr) %{
  match(Set cr (CmpI (AndI op1 op2) op3));

  ins_cost(INSN_COST);
  format %{ "tstw $op1, $op2 # int" %}
  ins_encode %{
    __ tstw($op1$$Register, $op2$$Register);
  %}
  ins_pipe(ialu_reg_reg);
%}


// Conditional Far Branch
// Conditional Far Branch Unsigned
// TODO: fixme

// counted loop end branch near
instruct branchLoopEnd(cmpOp cmp, rFlagsReg cr, label lbl)
%{
  match(CountedLoopEnd cmp cr);

  effect(USE lbl);

  ins_cost(BRANCH_COST);
  // short variant.
  // ins_short_branch(1);
  format %{ "b$cmp $lbl \t// counted loop end" %}

  ins_encode(aarch64_enc_br_con(cmp, lbl));

  ins_pipe(pipe_branch);
%}

// counted loop end branch near Unsigned
instruct branchLoopEndU(cmpOpU cmp, rFlagsRegU cr, label lbl)
%{
  match(CountedLoopEnd cmp cr);

  effect(USE lbl);

  ins_cost(BRANCH_COST);
  // short variant.
  // ins_short_branch(1);
  format %{ "b$cmp $lbl \t// counted loop end unsigned" %}

  ins_encode(aarch64_enc_br_conU(cmp, lbl));

  ins_pipe(pipe_branch);
%}

// counted loop end branch far
// counted loop end branch far unsigned
// TODO: fixme

// ============================================================================
// inlined locking and unlocking

instruct cmpFastLock(rFlagsReg cr, iRegP object, iRegP box, iRegPNoSp tmp, iRegPNoSp tmp2)
%{
  match(Set cr (FastLock object box));
  effect(TEMP tmp, TEMP tmp2);

  // TODO
  // identify correct cost
  ins_cost(5 * INSN_COST);
  format %{ "fastlock $object,$box\t! kills $tmp,$tmp2" %}

  ins_encode(aarch64_enc_fast_lock(object, box, tmp, tmp2));

  ins_pipe(pipe_serial);
%}

instruct cmpFastUnlock(rFlagsReg cr, iRegP object, iRegP box, iRegPNoSp tmp, iRegPNoSp tmp2)
%{
  match(Set cr (FastUnlock object box));
  effect(TEMP tmp, TEMP tmp2);

  ins_cost(5 * INSN_COST);
  format %{ "fastunlock $object,$box\t! kills $tmp, $tmp2" %}

  ins_encode(aarch64_enc_fast_unlock(object, box, tmp, tmp2));

  ins_pipe(pipe_serial);
%}


// ============================================================================
// Safepoint Instructions

// TODO
// provide a near and far version of this code

instruct safePoint(iRegP poll)
%{
  match(SafePoint poll);

  format %{
    "ldrw zr, [$poll]\t# Safepoint: poll for GC"
  %}
  ins_encode %{
    __ read_polling_page(as_Register($poll$$reg), relocInfo::poll_type);
  %}
  ins_pipe(pipe_serial); // ins_pipe(iload_reg_mem);
%}


// ============================================================================
// Procedure Call/Return Instructions

// Call Java Static Instruction

instruct CallStaticJavaDirect(method meth)
%{
  match(CallStaticJava);

  effect(USE meth);

  ins_cost(CALL_COST);

  format %{ "call,static $meth \t// ==> " %}

  ins_encode( aarch64_enc_java_static_call(meth),
              aarch64_enc_call_epilog );

  ins_pipe(pipe_class_call);
%}

// TO HERE

// Call Java Dynamic Instruction
instruct CallDynamicJavaDirect(method meth)
%{
  match(CallDynamicJava);

  effect(USE meth);

  ins_cost(CALL_COST);

  format %{ "CALL,dynamic $meth \t// ==> " %}

  ins_encode( aarch64_enc_java_dynamic_call(meth),
               aarch64_enc_call_epilog );

  ins_pipe(pipe_class_call);
%}

// Call Runtime Instruction

instruct CallRuntimeDirect(method meth)
%{
  match(CallRuntime);

  effect(USE meth);

  ins_cost(CALL_COST);

  format %{ "CALL, runtime $meth" %}

  ins_encode( aarch64_enc_java_to_runtime(meth) );

  ins_pipe(pipe_class_call);
%}

// Call Runtime Instruction

instruct CallLeafDirect(method meth)
%{
  match(CallLeaf);

  effect(USE meth);

  ins_cost(CALL_COST);

  format %{ "CALL, runtime leaf $meth" %}

  ins_encode( aarch64_enc_java_to_runtime(meth) );

  ins_pipe(pipe_class_call);
%}

// Call Runtime Instruction

instruct CallLeafNoFPDirect(method meth)
%{
  match(CallLeafNoFP);

  effect(USE meth);

  ins_cost(CALL_COST);

  format %{ "CALL, runtime leaf nofp $meth" %}

  ins_encode( aarch64_enc_java_to_runtime(meth) );

  ins_pipe(pipe_class_call);
%}

// Tail Call; Jump from runtime stub to Java code.
// Also known as an 'interprocedural jump'.
// Target of jump will eventually return to caller.
// TailJump below removes the return address.
instruct TailCalljmpInd(iRegPNoSp jump_target, inline_cache_RegP method_oop)
%{
  match(TailCall jump_target method_oop);

  ins_cost(CALL_COST);

  format %{ "br $jump_target\t# $method_oop holds method oop" %}

  ins_encode(aarch64_enc_tail_call(jump_target));

  ins_pipe(pipe_class_call);
%}

instruct TailjmpInd(iRegPNoSp jump_target, iRegP_R0 ex_oop)
%{
  match(TailJump jump_target ex_oop);

  ins_cost(CALL_COST);

  format %{ "br $jump_target\t# $ex_oop holds exception oop" %}

  ins_encode(aarch64_enc_tail_jmp(jump_target));

  ins_pipe(pipe_class_call);
%}

// Create exception oop: created by stack-crawling runtime code.
// Created exception is now available to this handler, and is setup
// just prior to jumping to this handler. No code emitted.
// TODO check
// should ex_oop be in r0? intel uses rax, ppc cannot use r0 so uses rarg1
instruct CreateException(iRegP_R0 ex_oop)
%{
  match(Set ex_oop (CreateEx));

  format %{ " -- \t// exception oop; no code emitted" %}

  size(0);

  ins_encode( /*empty*/ );

  ins_pipe(pipe_class_empty);
%}

// Rethrow exception: The exception oop will come in the first
// argument position. Then JUMP (not call) to the rethrow stub code.
instruct RethrowException() %{
  match(Rethrow);
  ins_cost(CALL_COST);

  format %{ "b rethrow_stub" %}

  ins_encode( aarch64_enc_rethrow() );

  ins_pipe(pipe_class_call);
%}


// Return Instruction
// epilog node loads ret address into lr as part of frame pop
instruct Ret()
%{
  match(Return);

  format %{ "ret\t// return register" %}

  ins_encode( aarch64_enc_ret() );

  ins_pipe(pipe_branch);
%}

// Die now.
instruct ShouldNotReachHere() %{
  match(Halt);

  ins_cost(CALL_COST);
  format %{ "ShouldNotReachHere" %}

  ins_encode %{
    // TODO
    // implement proper trap call here
    __ brk(999);
  %}

  ins_pipe(pipe_class_default);
%}

// ============================================================================
// Partial Subtype Check
//
// superklass array for an instance of the superklass.  Set a hidden
// internal cache on a hit (cache is checked with exposed code in
// gen_subtype_check()).  Return NZ for a miss or zero for a hit.  The
// encoding ALSO sets flags.

instruct partialSubtypeCheck(iRegP_R4 sub, iRegP_R0 super, iRegP_R2 temp, iRegP_R5 result, rFlagsReg cr)
%{
  match(Set result (PartialSubtypeCheck sub super));
  effect(KILL cr, KILL temp);

  ins_cost(1100);  // slightly larger than the next version
  format %{ "partialSubtypeCheck $result, $sub, $super" %}

  ins_encode(aarch64_enc_partial_subtype_check(sub, super, temp, result));

  opcode(0x1); // Force zero of result reg on hit

  ins_pipe(pipe_class_memory);
%}

instruct partialSubtypeCheckVsZero(iRegP_R4 sub, iRegP_R0 super, iRegP_R2 temp, iRegP_R5 result, immP0 zero, rFlagsReg cr)
%{
  match(Set cr (CmpP (PartialSubtypeCheck sub super) zero));
  effect(KILL temp, KILL result);

  ins_cost(1100);  // slightly larger than the next version
  format %{ "partialSubtypeCheck $result, $sub, $super == 0" %}

  ins_encode(aarch64_enc_partial_subtype_check(sub, super, temp, result));

  opcode(0x0); // Don't zero result reg on hit

  ins_pipe(pipe_class_memory);
%}

instruct string_compareU(iRegP_R1 str1, iRegI_R2 cnt1, iRegP_R3 str2, iRegI_R4 cnt2,
                        iRegI_R0 result, iRegP_R10 tmp1, rFlagsReg cr)
%{
  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::UU);
  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(KILL tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);

  format %{ "String Compare $str1,$cnt1,$str2,$cnt2 -> $result   # KILL $tmp1" %}
  ins_encode %{
    // Count is in 8-bit bytes; non-Compact chars are 16 bits.
    __ string_compare($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register, $result$$Register,
                      $tmp1$$Register,
                      fnoreg, fnoreg, StrIntrinsicNode::UU);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_compareL(iRegP_R1 str1, iRegI_R2 cnt1, iRegP_R3 str2, iRegI_R4 cnt2,
                        iRegI_R0 result, iRegP_R10 tmp1, rFlagsReg cr)
%{
  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::LL);
  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(KILL tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);

  format %{ "String Compare $str1,$cnt1,$str2,$cnt2 -> $result   # KILL $tmp1" %}
  ins_encode %{
    __ string_compare($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register, $result$$Register,
                      $tmp1$$Register,
                      fnoreg, fnoreg, StrIntrinsicNode::LL);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_compareUL(iRegP_R1 str1, iRegI_R2 cnt1, iRegP_R3 str2, iRegI_R4 cnt2,
                        iRegI_R0 result, vRegD vtmp1, vRegD vtmp2, iRegP_R10 tmp1, rFlagsReg cr)
%{
  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::UL);
  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(KILL tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, TEMP vtmp1, TEMP vtmp2, KILL cr);

  format %{ "String Compare $str1,$cnt1,$str2,$cnt2 -> $result   # KILL $tmp1" %}
  ins_encode %{
    __ string_compare($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register, $result$$Register,
                      $tmp1$$Register,
                      $vtmp1$$FloatRegister, $vtmp2$$FloatRegister, StrIntrinsicNode::UL);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_compareLU(iRegP_R1 str1, iRegI_R2 cnt1, iRegP_R3 str2, iRegI_R4 cnt2,
                        iRegI_R0 result, vRegD vtmp1, vRegD vtmp2, iRegP_R10 tmp1, rFlagsReg cr)
%{
  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::LU);
  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(KILL tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, TEMP vtmp1, TEMP vtmp2, KILL cr);

  format %{ "String Compare $str1,$cnt1,$str2,$cnt2 -> $result   # KILL $tmp1" %}
  ins_encode %{
    __ string_compare($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register, $result$$Register,
                      $tmp1$$Register,
                      $vtmp1$$FloatRegister, $vtmp2$$FloatRegister, StrIntrinsicNode::LU);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_indexofUU(iRegP_R1 str1, iRegI_R4 cnt1, iRegP_R3 str2, iRegI_R2 cnt2,
       iRegI_R0 result, iRegINoSp tmp1, iRegINoSp tmp2, iRegINoSp tmp3, iRegINoSp tmp4, rFlagsReg cr)
%{
  predicate(((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UU);
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2,
         TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, KILL cr);
  format %{ "String IndexOf $str1,$cnt1,$str2,$cnt2 -> $result (UU)" %}

  ins_encode %{
    __ string_indexof($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register,
                      $tmp1$$Register, $tmp2$$Register,
                      $tmp3$$Register, $tmp4$$Register,
                      -1, $result$$Register, StrIntrinsicNode::UU);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_indexofLL(iRegP_R1 str1, iRegI_R4 cnt1, iRegP_R3 str2, iRegI_R2 cnt2,
       iRegI_R0 result, iRegINoSp tmp1, iRegINoSp tmp2, iRegINoSp tmp3, iRegINoSp tmp4, rFlagsReg cr)
%{
  predicate(((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::LL);
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2,
         TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, KILL cr);
  format %{ "String IndexOf $str1,$cnt1,$str2,$cnt2 -> $result (LL)" %}

  ins_encode %{
    __ string_indexof($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register,
                      $tmp1$$Register, $tmp2$$Register,
                      $tmp3$$Register, $tmp4$$Register,
                      -1, $result$$Register, StrIntrinsicNode::LL);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_indexofUL(iRegP_R1 str1, iRegI_R4 cnt1, iRegP_R3 str2, iRegI_R2 cnt2,
       iRegI_R0 result, iRegINoSp tmp1, iRegINoSp tmp2, iRegINoSp tmp3, iRegINoSp tmp4, rFlagsReg cr)
%{
  predicate(((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UL);
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2,
         TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, KILL cr);
  format %{ "String IndexOf $str1,$cnt1,$str2,$cnt2 -> $result (UL)" %}

  ins_encode %{
    __ string_indexof($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register,
                      $tmp1$$Register, $tmp2$$Register,
                      $tmp3$$Register, $tmp4$$Register,
                      -1, $result$$Register, StrIntrinsicNode::UL);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_indexofLU(iRegP_R1 str1, iRegI_R4 cnt1, iRegP_R3 str2, iRegI_R2 cnt2,
       iRegI_R0 result, iRegINoSp tmp1, iRegINoSp tmp2, iRegINoSp tmp3, iRegINoSp tmp4, rFlagsReg cr)
%{
  predicate(((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::LU);
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2,
         TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, KILL cr);
  format %{ "String IndexOf $str1,$cnt1,$str2,$cnt2 -> $result (LU)" %}

  ins_encode %{
    __ string_indexof($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register,
                      $tmp1$$Register, $tmp2$$Register,
                      $tmp3$$Register, $tmp4$$Register,
                      -1, $result$$Register, StrIntrinsicNode::LU);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_indexof_conUU(iRegP_R1 str1, iRegI_R4 cnt1, iRegP_R3 str2,
                 immI_le_4 int_cnt2, iRegI_R0 result, iRegINoSp tmp1, iRegINoSp tmp2,
                 iRegINoSp tmp3, iRegINoSp tmp4, rFlagsReg cr)
%{
  predicate(((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UU);
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1,
         TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, KILL cr);
  format %{ "String IndexOf $str1,$cnt1,$str2,$int_cnt2 -> $result (UU)" %}

  ins_encode %{
    int icnt2 = (int)$int_cnt2$$constant;
    __ string_indexof($str1$$Register, $str2$$Register,
                      $cnt1$$Register, zr,
                      $tmp1$$Register, $tmp2$$Register,
                      $tmp3$$Register, $tmp4$$Register,
                      icnt2, $result$$Register, StrIntrinsicNode::UU);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_indexof_conLL(iRegP_R1 str1, iRegI_R4 cnt1, iRegP_R3 str2,
                 immI_le_4 int_cnt2, iRegI_R0 result, iRegINoSp tmp1, iRegINoSp tmp2,
                 iRegINoSp tmp3, iRegINoSp tmp4, rFlagsReg cr)
%{
  predicate(((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::LL);
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1,
         TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, KILL cr);
  format %{ "String IndexOf $str1,$cnt1,$str2,$int_cnt2 -> $result (LL)" %}

  ins_encode %{
    int icnt2 = (int)$int_cnt2$$constant;
    __ string_indexof($str1$$Register, $str2$$Register,
                      $cnt1$$Register, zr,
                      $tmp1$$Register, $tmp2$$Register,
                      $tmp3$$Register, $tmp4$$Register,
                      icnt2, $result$$Register, StrIntrinsicNode::LL);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_indexof_conUL(iRegP_R1 str1, iRegI_R4 cnt1, iRegP_R3 str2,
                 immI_1 int_cnt2, iRegI_R0 result, iRegINoSp tmp1, iRegINoSp tmp2,
                 iRegINoSp tmp3, iRegINoSp tmp4, rFlagsReg cr)
%{
  predicate(((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UL);
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1,
         TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, KILL cr);
  format %{ "String IndexOf $str1,$cnt1,$str2,$int_cnt2 -> $result (UL)" %}

  ins_encode %{
    int icnt2 = (int)$int_cnt2$$constant;
    __ string_indexof($str1$$Register, $str2$$Register,
                      $cnt1$$Register, zr,
                      $tmp1$$Register, $tmp2$$Register,
                      $tmp3$$Register, $tmp4$$Register,
                      icnt2, $result$$Register, StrIntrinsicNode::UL);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_indexof_conLU(iRegP_R1 str1, iRegI_R4 cnt1, iRegP_R3 str2,
                 immI_1 int_cnt2, iRegI_R0 result, iRegINoSp tmp1, iRegINoSp tmp2,
                 iRegINoSp tmp3, iRegINoSp tmp4, rFlagsReg cr)
%{
  predicate(((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::LU);
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1,
         TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, KILL cr);
  format %{ "String IndexOf $str1,$cnt1,$str2,$int_cnt2 -> $result (LU)" %}

  ins_encode %{
    int icnt2 = (int)$int_cnt2$$constant;
    __ string_indexof($str1$$Register, $str2$$Register,
                      $cnt1$$Register, zr,
                      $tmp1$$Register, $tmp2$$Register,
                      $tmp3$$Register, $tmp4$$Register,
                      icnt2, $result$$Register, StrIntrinsicNode::LU);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_indexofU_char(iRegP_R1 str1, iRegI_R2 cnt1, iRegI_R3 ch,
                              iRegI_R0 result, iRegINoSp tmp1, iRegINoSp tmp2,
                              iRegINoSp tmp3, rFlagsReg cr)
%{
  match(Set result (StrIndexOfChar (Binary str1 cnt1) ch));
  effect(USE_KILL str1, USE_KILL cnt1, USE_KILL ch,
         TEMP tmp1, TEMP tmp2, TEMP tmp3, KILL cr);

  format %{ "String IndexOf char[] $str1,$cnt1,$ch -> $result" %}

  ins_encode %{
    __ string_indexof_char($str1$$Register, $cnt1$$Register, $ch$$Register,
                           $result$$Register, $tmp1$$Register, $tmp2$$Register,
                           $tmp3$$Register);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_equalsL(iRegP_R1 str1, iRegP_R3 str2, iRegI_R4 cnt,
                        iRegI_R0 result, rFlagsReg cr)
%{
  predicate(((StrEqualsNode*)n)->encoding() == StrIntrinsicNode::LL);
  match(Set result (StrEquals (Binary str1 str2) cnt));
  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt, KILL cr);

  format %{ "String Equals $str1,$str2,$cnt -> $result" %}
  ins_encode %{
    // Count is in 8-bit bytes; non-Compact chars are 16 bits.
    __ arrays_equals($str1$$Register, $str2$$Register,
                     $result$$Register, $cnt$$Register,
                     1, /*is_string*/true);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct string_equalsU(iRegP_R1 str1, iRegP_R3 str2, iRegI_R4 cnt,
                        iRegI_R0 result, rFlagsReg cr)
%{
  predicate(((StrEqualsNode*)n)->encoding() == StrIntrinsicNode::UU);
  match(Set result (StrEquals (Binary str1 str2) cnt));
  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt, KILL cr);

  format %{ "String Equals $str1,$str2,$cnt -> $result" %}
  ins_encode %{
    // Count is in 8-bit bytes; non-Compact chars are 16 bits.
    __ asrw($cnt$$Register, $cnt$$Register, 1);
    __ arrays_equals($str1$$Register, $str2$$Register,
                     $result$$Register, $cnt$$Register,
                     2, /*is_string*/true);
  %}
  ins_pipe(pipe_class_memory);
%}

instruct array_equalsB(iRegP_R1 ary1, iRegP_R2 ary2, iRegI_R0 result,
                      iRegP_R10 tmp, rFlagsReg cr)
%{
  predicate(((AryEqNode*)n)->encoding() == StrIntrinsicNode::LL);
  match(Set result (AryEq ary1 ary2));
  effect(KILL tmp, USE_KILL ary1, USE_KILL ary2, KILL cr);

  format %{ "Array Equals $ary1,ary2 -> $result    // KILL $tmp" %}
  ins_encode %{
    __ arrays_equals($ary1$$Register, $ary2$$Register,
                     $result$$Register, $tmp$$Register,
                     1, /*is_string*/false);
    %}
  ins_pipe(pipe_class_memory);
%}

instruct array_equalsC(iRegP_R1 ary1, iRegP_R2 ary2, iRegI_R0 result,
                      iRegP_R10 tmp, rFlagsReg cr)
%{
  predicate(((AryEqNode*)n)->encoding() == StrIntrinsicNode::UU);
  match(Set result (AryEq ary1 ary2));
  effect(KILL tmp, USE_KILL ary1, USE_KILL ary2, KILL cr);

  format %{ "Array Equals $ary1,ary2 -> $result    // KILL $tmp" %}
  ins_encode %{
    __ arrays_equals($ary1$$Register, $ary2$$Register,
                     $result$$Register, $tmp$$Register,
                     2, /*is_string*/false);
  %}
  ins_pipe(pipe_class_memory);
%}


// fast char[] to byte[] compression
instruct string_compress(iRegP_R2 src, iRegP_R1 dst, iRegI_R3 len,
                         vRegD_V0 tmp1, vRegD_V1 tmp2,
                         vRegD_V2 tmp3, vRegD_V3 tmp4,
                         iRegI_R0 result, rFlagsReg cr)
%{
  match(Set result (StrCompressedCopy src (Binary dst len)));
  effect(TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, USE_KILL src, USE_KILL dst, USE_KILL len, KILL cr);

  format %{ "String Compress $src,$dst -> $result    // KILL R1, R2, R3, R4" %}
  ins_encode %{
    __ char_array_compress($src$$Register, $dst$$Register, $len$$Register,
                           $tmp1$$FloatRegister, $tmp2$$FloatRegister,
                           $tmp3$$FloatRegister, $tmp4$$FloatRegister,
                           $result$$Register);
  %}
  ins_pipe( pipe_slow );
%}

// fast byte[] to char[] inflation
instruct string_inflate(Universe dummy, iRegP_R0 src, iRegP_R1 dst, iRegI_R2 len,
                        vRegD tmp1, vRegD tmp2, vRegD tmp3, iRegP_R3 tmp4, rFlagsReg cr)
%{
  match(Set dummy (StrInflatedCopy src (Binary dst len)));
  effect(TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, USE_KILL src, USE_KILL dst, USE_KILL len, KILL cr);

  format %{ "String Inflate $src,$dst    // KILL $tmp1, $tmp2" %}
  ins_encode %{
    __ byte_array_inflate($src$$Register, $dst$$Register, $len$$Register,
                          $tmp1$$FloatRegister, $tmp2$$FloatRegister, $tmp3$$FloatRegister, $tmp4$$Register);
  %}
  ins_pipe(pipe_class_memory);
%}

// encode char[] to byte[] in ISO_8859_1
instruct encode_iso_array(iRegP_R2 src, iRegP_R1 dst, iRegI_R3 len,
                          vRegD_V0 Vtmp1, vRegD_V1 Vtmp2,
                          vRegD_V2 Vtmp3, vRegD_V3 Vtmp4,
                          iRegI_R0 result, rFlagsReg cr)
%{
  match(Set result (EncodeISOArray src (Binary dst len)));
  effect(USE_KILL src, USE_KILL dst, USE_KILL len,
         KILL Vtmp1, KILL Vtmp2, KILL Vtmp3, KILL Vtmp4, KILL cr);

  format %{ "Encode array $src,$dst,$len -> $result" %}
  ins_encode %{
    __ encode_iso_array($src$$Register, $dst$$Register, $len$$Register,
         $result$$Register, $Vtmp1$$FloatRegister,  $Vtmp2$$FloatRegister,
         $Vtmp3$$FloatRegister,  $Vtmp4$$FloatRegister);
  %}
  ins_pipe( pipe_class_memory );
%}

// ============================================================================
// This name is KNOWN by the ADLC and cannot be changed.
// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
// for this guy.
instruct tlsLoadP(thread_RegP dst)
%{
  match(Set dst (ThreadLocal));

  ins_cost(0);

  format %{ " -- \t// $dst=Thread::current(), empty" %}

  size(0);

  ins_encode( /*empty*/ );

  ins_pipe(pipe_class_empty);
%}

// ====================VECTOR INSTRUCTIONS=====================================

// Load vector (32 bits)
instruct loadV4(vecD dst, vmem4 mem)
%{
  predicate(n->as_LoadVector()->memory_size() == 4);
  match(Set dst (LoadVector mem));
  ins_cost(4 * INSN_COST);
  format %{ "ldrs   $dst,$mem\t# vector (32 bits)" %}
  ins_encode( aarch64_enc_ldrvS(dst, mem) );
  ins_pipe(vload_reg_mem64);
%}

// Load vector (64 bits)
instruct loadV8(vecD dst, vmem8 mem)
%{
  predicate(n->as_LoadVector()->memory_size() == 8);
  match(Set dst (LoadVector mem));
  ins_cost(4 * INSN_COST);
  format %{ "ldrd   $dst,$mem\t# vector (64 bits)" %}
  ins_encode( aarch64_enc_ldrvD(dst, mem) );
  ins_pipe(vload_reg_mem64);
%}

// Load Vector (128 bits)
instruct loadV16(vecX dst, vmem16 mem)
%{
  predicate(n->as_LoadVector()->memory_size() == 16);
  match(Set dst (LoadVector mem));
  ins_cost(4 * INSN_COST);
  format %{ "ldrq   $dst,$mem\t# vector (128 bits)" %}
  ins_encode( aarch64_enc_ldrvQ(dst, mem) );
  ins_pipe(vload_reg_mem128);
%}

// Store Vector (32 bits)
instruct storeV4(vecD src, vmem4 mem)
%{
  predicate(n->as_StoreVector()->memory_size() == 4);
  match(Set mem (StoreVector mem src));
  ins_cost(4 * INSN_COST);
  format %{ "strs   $mem,$src\t# vector (32 bits)" %}
  ins_encode( aarch64_enc_strvS(src, mem) );
  ins_pipe(vstore_reg_mem64);
%}

// Store Vector (64 bits)
instruct storeV8(vecD src, vmem8 mem)
%{
  predicate(n->as_StoreVector()->memory_size() == 8);
  match(Set mem (StoreVector mem src));
  ins_cost(4 * INSN_COST);
  format %{ "strd   $mem,$src\t# vector (64 bits)" %}
  ins_encode( aarch64_enc_strvD(src, mem) );
  ins_pipe(vstore_reg_mem64);
%}

// Store Vector (128 bits)
instruct storeV16(vecX src, vmem16 mem)
%{
  predicate(n->as_StoreVector()->memory_size() == 16);
  match(Set mem (StoreVector mem src));
  ins_cost(4 * INSN_COST);
  format %{ "strq   $mem,$src\t# vector (128 bits)" %}
  ins_encode( aarch64_enc_strvQ(src, mem) );
  ins_pipe(vstore_reg_mem128);
%}

instruct replicate8B(vecD dst, iRegIorL2I src)
%{
  predicate(n->as_Vector()->length() == 4 ||
            n->as_Vector()->length() == 8);
  match(Set dst (ReplicateB src));
  ins_cost(INSN_COST);
  format %{ "dup  $dst, $src\t# vector (8B)" %}
  ins_encode %{
    __ dup(as_FloatRegister($dst$$reg), __ T8B, as_Register($src$$reg));
  %}
  ins_pipe(vdup_reg_reg64);
%}

instruct replicate16B(vecX dst, iRegIorL2I src)
%{
  predicate(n->as_Vector()->length() == 16);
  match(Set dst (ReplicateB src));
  ins_cost(INSN_COST);
  format %{ "dup  $dst, $src\t# vector (16B)" %}
  ins_encode %{
    __ dup(as_FloatRegister($dst$$reg), __ T16B, as_Register($src$$reg));
  %}
  ins_pipe(vdup_reg_reg128);
%}

instruct replicate8B_imm(vecD dst, immI con)
%{
  predicate(n->as_Vector()->length() == 4 ||
            n->as_Vector()->length() == 8);
  match(Set dst (ReplicateB con));
  ins_cost(INSN_COST);
  format %{ "movi  $dst, $con\t# vector(8B)" %}
  ins_encode %{
    __ mov(as_FloatRegister($dst$$reg), __ T8B, $con$$constant & 0xff);
  %}
  ins_pipe(vmovi_reg_imm64);
%}

instruct replicate16B_imm(vecX dst, immI con)
%{
  predicate(n->as_Vector()->length() == 16);
  match(Set dst (ReplicateB con));
  ins_cost(INSN_COST);
  format %{ "movi  $dst, $con\t# vector(16B)" %}
  ins_encode %{
    __ mov(as_FloatRegister($dst$$reg), __ T16B, $con$$constant & 0xff);
  %}
  ins_pipe(vmovi_reg_imm128);
%}

instruct replicate4S(vecD dst, iRegIorL2I src)
%{
  predicate(n->as_Vector()->length() == 2 ||
            n->as_Vector()->length() == 4);
  match(Set dst (ReplicateS src));
  ins_cost(INSN_COST);
  format %{ "dup  $dst, $src\t# vector (4S)" %}
  ins_encode %{
    __ dup(as_FloatRegister($dst$$reg), __ T4H, as_Register($src$$reg));
  %}
  ins_pipe(vdup_reg_reg64);
%}

instruct replicate8S(vecX dst, iRegIorL2I src)
%{
  predicate(n->as_Vector()->length() == 8);
  match(Set dst (ReplicateS src));
  ins_cost(INSN_COST);
  format %{ "dup  $dst, $src\t# vector (8S)" %}
  ins_encode %{
    __ dup(as_FloatRegister($dst$$reg), __ T8H, as_Register($src$$reg));
  %}
  ins_pipe(vdup_reg_reg128);
%}

instruct replicate4S_imm(vecD dst, immI con)
%{
  predicate(n->as_Vector()->length() == 2 ||
            n->as_Vector()->length() == 4);
  match(Set dst (ReplicateS con));
  ins_cost(INSN_COST);
  format %{ "movi  $dst, $con\t# vector(4H)" %}
  ins_encode %{
    __ mov(as_FloatRegister($dst$$reg), __ T4H, $con$$constant & 0xffff);
  %}
  ins_pipe(vmovi_reg_imm64);
%}

instruct replicate8S_imm(vecX dst, immI con)
%{
  predicate(n->as_Vector()->length() == 8);
  match(Set dst (ReplicateS con));
  ins_cost(INSN_COST);
  format %{ "movi  $dst, $con\t# vector(8H)" %}
  ins_encode %{
    __ mov(as_FloatRegister($dst$$reg), __ T8H, $con$$constant & 0xffff);
  %}
  ins_pipe(vmovi_reg_imm128);
%}

instruct replicate2I(vecD dst, iRegIorL2I src)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (ReplicateI src));
  ins_cost(INSN_COST);
  format %{ "dup  $dst, $src\t# vector (2I)" %}
  ins_encode %{
    __ dup(as_FloatRegister($dst$$reg), __ T2S, as_Register($src$$reg));
  %}
  ins_pipe(vdup_reg_reg64);
%}

instruct replicate4I(vecX dst, iRegIorL2I src)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (ReplicateI src));
  ins_cost(INSN_COST);
  format %{ "dup  $dst, $src\t# vector (4I)" %}
  ins_encode %{
    __ dup(as_FloatRegister($dst$$reg), __ T4S, as_Register($src$$reg));
  %}
  ins_pipe(vdup_reg_reg128);
%}

instruct replicate2I_imm(vecD dst, immI con)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (ReplicateI con));
  ins_cost(INSN_COST);
  format %{ "movi  $dst, $con\t# vector(2I)" %}
  ins_encode %{
    __ mov(as_FloatRegister($dst$$reg), __ T2S, $con$$constant);
  %}
  ins_pipe(vmovi_reg_imm64);
%}

instruct replicate4I_imm(vecX dst, immI con)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (ReplicateI con));
  ins_cost(INSN_COST);
  format %{ "movi  $dst, $con\t# vector(4I)" %}
  ins_encode %{
    __ mov(as_FloatRegister($dst$$reg), __ T4S, $con$$constant);
  %}
  ins_pipe(vmovi_reg_imm128);
%}

instruct replicate2L(vecX dst, iRegL src)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (ReplicateL src));
  ins_cost(INSN_COST);
  format %{ "dup  $dst, $src\t# vector (2L)" %}
  ins_encode %{
    __ dup(as_FloatRegister($dst$$reg), __ T2D, as_Register($src$$reg));
  %}
  ins_pipe(vdup_reg_reg128);
%}

instruct replicate2L_zero(vecX dst, immI0 zero)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (ReplicateI zero));
  ins_cost(INSN_COST);
  format %{ "movi  $dst, $zero\t# vector(4I)" %}
  ins_encode %{
    __ eor(as_FloatRegister($dst$$reg), __ T16B,
           as_FloatRegister($dst$$reg),
           as_FloatRegister($dst$$reg));
  %}
  ins_pipe(vmovi_reg_imm128);
%}

instruct replicate2F(vecD dst, vRegF src)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (ReplicateF src));
  ins_cost(INSN_COST);
  format %{ "dup  $dst, $src\t# vector (2F)" %}
  ins_encode %{
    __ dup(as_FloatRegister($dst$$reg), __ T2S,
           as_FloatRegister($src$$reg));
  %}
  ins_pipe(vdup_reg_freg64);
%}

instruct replicate4F(vecX dst, vRegF src)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (ReplicateF src));
  ins_cost(INSN_COST);
  format %{ "dup  $dst, $src\t# vector (4F)" %}
  ins_encode %{
    __ dup(as_FloatRegister($dst$$reg), __ T4S,
           as_FloatRegister($src$$reg));
  %}
  ins_pipe(vdup_reg_freg128);
%}

instruct replicate2D(vecX dst, vRegD src)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (ReplicateD src));
  ins_cost(INSN_COST);
  format %{ "dup  $dst, $src\t# vector (2D)" %}
  ins_encode %{
    __ dup(as_FloatRegister($dst$$reg), __ T2D,
           as_FloatRegister($src$$reg));
  %}
  ins_pipe(vdup_reg_dreg128);
%}

// ====================REDUCTION ARITHMETIC====================================

instruct reduce_add2I(iRegINoSp dst, iRegIorL2I src1, vecD src2, iRegINoSp tmp, iRegINoSp tmp2)
%{
  match(Set dst (AddReductionVI src1 src2));
  ins_cost(INSN_COST);
  effect(TEMP tmp, TEMP tmp2);
  format %{ "umov  $tmp, $src2, S, 0\n\t"
            "umov  $tmp2, $src2, S, 1\n\t"
            "addw  $dst, $src1, $tmp\n\t"
            "addw  $dst, $dst, $tmp2\t add reduction2i"
  %}
  ins_encode %{
    __ umov($tmp$$Register, as_FloatRegister($src2$$reg), __ S, 0);
    __ umov($tmp2$$Register, as_FloatRegister($src2$$reg), __ S, 1);
    __ addw($dst$$Register, $src1$$Register, $tmp$$Register);
    __ addw($dst$$Register, $dst$$Register, $tmp2$$Register);
  %}
  ins_pipe(pipe_class_default);
%}

instruct reduce_add4I(iRegINoSp dst, iRegIorL2I src1, vecX src2, vecX tmp, iRegINoSp tmp2)
%{
  match(Set dst (AddReductionVI src1 src2));
  ins_cost(INSN_COST);
  effect(TEMP tmp, TEMP tmp2);
  format %{ "addv  $tmp, T4S, $src2\n\t"
            "umov  $tmp2, $tmp, S, 0\n\t"
            "addw  $dst, $tmp2, $src1\t add reduction4i"
  %}
  ins_encode %{
    __ addv(as_FloatRegister($tmp$$reg), __ T4S,
            as_FloatRegister($src2$$reg));
    __ umov($tmp2$$Register, as_FloatRegister($tmp$$reg), __ S, 0);
    __ addw($dst$$Register, $tmp2$$Register, $src1$$Register);
  %}
  ins_pipe(pipe_class_default);
%}

instruct reduce_mul2I(iRegINoSp dst, iRegIorL2I src1, vecD src2, iRegINoSp tmp)
%{
  match(Set dst (MulReductionVI src1 src2));
  ins_cost(INSN_COST);
  effect(TEMP tmp, TEMP dst);
  format %{ "umov  $tmp, $src2, S, 0\n\t"
            "mul   $dst, $tmp, $src1\n\t"
            "umov  $tmp, $src2, S, 1\n\t"
            "mul   $dst, $tmp, $dst\t mul reduction2i\n\t"
  %}
  ins_encode %{
    __ umov($tmp$$Register, as_FloatRegister($src2$$reg), __ S, 0);
    __ mul($dst$$Register, $tmp$$Register, $src1$$Register);
    __ umov($tmp$$Register, as_FloatRegister($src2$$reg), __ S, 1);
    __ mul($dst$$Register, $tmp$$Register, $dst$$Register);
  %}
  ins_pipe(pipe_class_default);
%}

instruct reduce_mul4I(iRegINoSp dst, iRegIorL2I src1, vecX src2, vecX tmp, iRegINoSp tmp2)
%{
  match(Set dst (MulReductionVI src1 src2));
  ins_cost(INSN_COST);
  effect(TEMP tmp, TEMP tmp2, TEMP dst);
  format %{ "ins   $tmp, $src2, 0, 1\n\t"
            "mul   $tmp, $tmp, $src2\n\t"
            "umov  $tmp2, $tmp, S, 0\n\t"
            "mul   $dst, $tmp2, $src1\n\t"
            "umov  $tmp2, $tmp, S, 1\n\t"
            "mul   $dst, $tmp2, $dst\t mul reduction4i\n\t"
  %}
  ins_encode %{
    __ ins(as_FloatRegister($tmp$$reg), __ D,
           as_FloatRegister($src2$$reg), 0, 1);
    __ mulv(as_FloatRegister($tmp$$reg), __ T2S,
           as_FloatRegister($tmp$$reg), as_FloatRegister($src2$$reg));
    __ umov($tmp2$$Register, as_FloatRegister($tmp$$reg), __ S, 0);
    __ mul($dst$$Register, $tmp2$$Register, $src1$$Register);
    __ umov($tmp2$$Register, as_FloatRegister($tmp$$reg), __ S, 1);
    __ mul($dst$$Register, $tmp2$$Register, $dst$$Register);
  %}
  ins_pipe(pipe_class_default);
%}

instruct reduce_add2F(vRegF dst, vRegF src1, vecD src2, vecD tmp)
%{
  match(Set dst (AddReductionVF src1 src2));
  ins_cost(INSN_COST);
  effect(TEMP tmp, TEMP dst);
  format %{ "fadds $dst, $src1, $src2\n\t"
            "ins   $tmp, S, $src2, 0, 1\n\t"
            "fadds $dst, $dst, $tmp\t add reduction2f"
  %}
  ins_encode %{
    __ fadds(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
    __ ins(as_FloatRegister($tmp$$reg), __ S,
           as_FloatRegister($src2$$reg), 0, 1);
    __ fadds(as_FloatRegister($dst$$reg),
             as_FloatRegister($dst$$reg), as_FloatRegister($tmp$$reg));
  %}
  ins_pipe(pipe_class_default);
%}

instruct reduce_add4F(vRegF dst, vRegF src1, vecX src2, vecX tmp)
%{
  match(Set dst (AddReductionVF src1 src2));
  ins_cost(INSN_COST);
  effect(TEMP tmp, TEMP dst);
  format %{ "fadds $dst, $src1, $src2\n\t"
            "ins   $tmp, S, $src2, 0, 1\n\t"
            "fadds $dst, $dst, $tmp\n\t"
            "ins   $tmp, S, $src2, 0, 2\n\t"
            "fadds $dst, $dst, $tmp\n\t"
            "ins   $tmp, S, $src2, 0, 3\n\t"
            "fadds $dst, $dst, $tmp\t add reduction4f"
  %}
  ins_encode %{
    __ fadds(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
    __ ins(as_FloatRegister($tmp$$reg), __ S,
           as_FloatRegister($src2$$reg), 0, 1);
    __ fadds(as_FloatRegister($dst$$reg),
             as_FloatRegister($dst$$reg), as_FloatRegister($tmp$$reg));
    __ ins(as_FloatRegister($tmp$$reg), __ S,
           as_FloatRegister($src2$$reg), 0, 2);
    __ fadds(as_FloatRegister($dst$$reg),
             as_FloatRegister($dst$$reg), as_FloatRegister($tmp$$reg));
    __ ins(as_FloatRegister($tmp$$reg), __ S,
           as_FloatRegister($src2$$reg), 0, 3);
    __ fadds(as_FloatRegister($dst$$reg),
             as_FloatRegister($dst$$reg), as_FloatRegister($tmp$$reg));
  %}
  ins_pipe(pipe_class_default);
%}

instruct reduce_mul2F(vRegF dst, vRegF src1, vecD src2, vecD tmp)
%{
  match(Set dst (MulReductionVF src1 src2));
  ins_cost(INSN_COST);
  effect(TEMP tmp, TEMP dst);
  format %{ "fmuls $dst, $src1, $src2\n\t"
            "ins   $tmp, S, $src2, 0, 1\n\t"
            "fmuls $dst, $dst, $tmp\t add reduction4f"
  %}
  ins_encode %{
    __ fmuls(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
    __ ins(as_FloatRegister($tmp$$reg), __ S,
           as_FloatRegister($src2$$reg), 0, 1);
    __ fmuls(as_FloatRegister($dst$$reg),
             as_FloatRegister($dst$$reg), as_FloatRegister($tmp$$reg));
  %}
  ins_pipe(pipe_class_default);
%}

instruct reduce_mul4F(vRegF dst, vRegF src1, vecX src2, vecX tmp)
%{
  match(Set dst (MulReductionVF src1 src2));
  ins_cost(INSN_COST);
  effect(TEMP tmp, TEMP dst);
  format %{ "fmuls $dst, $src1, $src2\n\t"
            "ins   $tmp, S, $src2, 0, 1\n\t"
            "fmuls $dst, $dst, $tmp\n\t"
            "ins   $tmp, S, $src2, 0, 2\n\t"
            "fmuls $dst, $dst, $tmp\n\t"
            "ins   $tmp, S, $src2, 0, 3\n\t"
            "fmuls $dst, $dst, $tmp\t add reduction4f"
  %}
  ins_encode %{
    __ fmuls(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
    __ ins(as_FloatRegister($tmp$$reg), __ S,
           as_FloatRegister($src2$$reg), 0, 1);
    __ fmuls(as_FloatRegister($dst$$reg),
             as_FloatRegister($dst$$reg), as_FloatRegister($tmp$$reg));
    __ ins(as_FloatRegister($tmp$$reg), __ S,
           as_FloatRegister($src2$$reg), 0, 2);
    __ fmuls(as_FloatRegister($dst$$reg),
             as_FloatRegister($dst$$reg), as_FloatRegister($tmp$$reg));
    __ ins(as_FloatRegister($tmp$$reg), __ S,
           as_FloatRegister($src2$$reg), 0, 3);
    __ fmuls(as_FloatRegister($dst$$reg),
             as_FloatRegister($dst$$reg), as_FloatRegister($tmp$$reg));
  %}
  ins_pipe(pipe_class_default);
%}

instruct reduce_add2D(vRegD dst, vRegD src1, vecX src2, vecX tmp)
%{
  match(Set dst (AddReductionVD src1 src2));
  ins_cost(INSN_COST);
  effect(TEMP tmp, TEMP dst);
  format %{ "faddd $dst, $src1, $src2\n\t"
            "ins   $tmp, D, $src2, 0, 1\n\t"
            "faddd $dst, $dst, $tmp\t add reduction2d"
  %}
  ins_encode %{
    __ faddd(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
    __ ins(as_FloatRegister($tmp$$reg), __ D,
           as_FloatRegister($src2$$reg), 0, 1);
    __ faddd(as_FloatRegister($dst$$reg),
             as_FloatRegister($dst$$reg), as_FloatRegister($tmp$$reg));
  %}
  ins_pipe(pipe_class_default);
%}

instruct reduce_mul2D(vRegD dst, vRegD src1, vecX src2, vecX tmp)
%{
  match(Set dst (MulReductionVD src1 src2));
  ins_cost(INSN_COST);
  effect(TEMP tmp, TEMP dst);
  format %{ "fmuld $dst, $src1, $src2\n\t"
            "ins   $tmp, D, $src2, 0, 1\n\t"
            "fmuld $dst, $dst, $tmp\t add reduction2d"
  %}
  ins_encode %{
    __ fmuld(as_FloatRegister($dst$$reg),
             as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
    __ ins(as_FloatRegister($tmp$$reg), __ D,
           as_FloatRegister($src2$$reg), 0, 1);
    __ fmuld(as_FloatRegister($dst$$reg),
             as_FloatRegister($dst$$reg), as_FloatRegister($tmp$$reg));
  %}
  ins_pipe(pipe_class_default);
%}

// ====================VECTOR ARITHMETIC=======================================

// --------------------------------- ADD --------------------------------------

instruct vadd8B(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 4 ||
            n->as_Vector()->length() == 8);
  match(Set dst (AddVB src1 src2));
  ins_cost(INSN_COST);
  format %{ "addv  $dst,$src1,$src2\t# vector (8B)" %}
  ins_encode %{
    __ addv(as_FloatRegister($dst$$reg), __ T8B,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop64);
%}

instruct vadd16B(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 16);
  match(Set dst (AddVB src1 src2));
  ins_cost(INSN_COST);
  format %{ "addv  $dst,$src1,$src2\t# vector (16B)" %}
  ins_encode %{
    __ addv(as_FloatRegister($dst$$reg), __ T16B,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop128);
%}

instruct vadd4S(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2 ||
            n->as_Vector()->length() == 4);
  match(Set dst (AddVS src1 src2));
  ins_cost(INSN_COST);
  format %{ "addv  $dst,$src1,$src2\t# vector (4H)" %}
  ins_encode %{
    __ addv(as_FloatRegister($dst$$reg), __ T4H,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop64);
%}

instruct vadd8S(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 8);
  match(Set dst (AddVS src1 src2));
  ins_cost(INSN_COST);
  format %{ "addv  $dst,$src1,$src2\t# vector (8H)" %}
  ins_encode %{
    __ addv(as_FloatRegister($dst$$reg), __ T8H,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop128);
%}

instruct vadd2I(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (AddVI src1 src2));
  ins_cost(INSN_COST);
  format %{ "addv  $dst,$src1,$src2\t# vector (2S)" %}
  ins_encode %{
    __ addv(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop64);
%}

instruct vadd4I(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (AddVI src1 src2));
  ins_cost(INSN_COST);
  format %{ "addv  $dst,$src1,$src2\t# vector (4S)" %}
  ins_encode %{
    __ addv(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop128);
%}

instruct vadd2L(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (AddVL src1 src2));
  ins_cost(INSN_COST);
  format %{ "addv  $dst,$src1,$src2\t# vector (2L)" %}
  ins_encode %{
    __ addv(as_FloatRegister($dst$$reg), __ T2D,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop128);
%}

instruct vadd2F(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (AddVF src1 src2));
  ins_cost(INSN_COST);
  format %{ "fadd  $dst,$src1,$src2\t# vector (2S)" %}
  ins_encode %{
    __ fadd(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop_fp64);
%}

instruct vadd4F(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (AddVF src1 src2));
  ins_cost(INSN_COST);
  format %{ "fadd  $dst,$src1,$src2\t# vector (4S)" %}
  ins_encode %{
    __ fadd(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop_fp128);
%}

instruct vadd2D(vecX dst, vecX src1, vecX src2)
%{
  match(Set dst (AddVD src1 src2));
  ins_cost(INSN_COST);
  format %{ "fadd  $dst,$src1,$src2\t# vector (2D)" %}
  ins_encode %{
    __ fadd(as_FloatRegister($dst$$reg), __ T2D,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop_fp128);
%}

// --------------------------------- SUB --------------------------------------

instruct vsub8B(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 4 ||
            n->as_Vector()->length() == 8);
  match(Set dst (SubVB src1 src2));
  ins_cost(INSN_COST);
  format %{ "subv  $dst,$src1,$src2\t# vector (8B)" %}
  ins_encode %{
    __ subv(as_FloatRegister($dst$$reg), __ T8B,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop64);
%}

instruct vsub16B(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 16);
  match(Set dst (SubVB src1 src2));
  ins_cost(INSN_COST);
  format %{ "subv  $dst,$src1,$src2\t# vector (16B)" %}
  ins_encode %{
    __ subv(as_FloatRegister($dst$$reg), __ T16B,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop128);
%}

instruct vsub4S(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2 ||
            n->as_Vector()->length() == 4);
  match(Set dst (SubVS src1 src2));
  ins_cost(INSN_COST);
  format %{ "subv  $dst,$src1,$src2\t# vector (4H)" %}
  ins_encode %{
    __ subv(as_FloatRegister($dst$$reg), __ T4H,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop64);
%}

instruct vsub8S(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 8);
  match(Set dst (SubVS src1 src2));
  ins_cost(INSN_COST);
  format %{ "subv  $dst,$src1,$src2\t# vector (8H)" %}
  ins_encode %{
    __ subv(as_FloatRegister($dst$$reg), __ T8H,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop128);
%}

instruct vsub2I(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (SubVI src1 src2));
  ins_cost(INSN_COST);
  format %{ "subv  $dst,$src1,$src2\t# vector (2S)" %}
  ins_encode %{
    __ subv(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop64);
%}

instruct vsub4I(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (SubVI src1 src2));
  ins_cost(INSN_COST);
  format %{ "subv  $dst,$src1,$src2\t# vector (4S)" %}
  ins_encode %{
    __ subv(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop128);
%}

instruct vsub2L(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (SubVL src1 src2));
  ins_cost(INSN_COST);
  format %{ "subv  $dst,$src1,$src2\t# vector (2L)" %}
  ins_encode %{
    __ subv(as_FloatRegister($dst$$reg), __ T2D,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop128);
%}

instruct vsub2F(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (SubVF src1 src2));
  ins_cost(INSN_COST);
  format %{ "fsub  $dst,$src1,$src2\t# vector (2S)" %}
  ins_encode %{
    __ fsub(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop_fp64);
%}

instruct vsub4F(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (SubVF src1 src2));
  ins_cost(INSN_COST);
  format %{ "fsub  $dst,$src1,$src2\t# vector (4S)" %}
  ins_encode %{
    __ fsub(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop_fp128);
%}

instruct vsub2D(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (SubVD src1 src2));
  ins_cost(INSN_COST);
  format %{ "fsub  $dst,$src1,$src2\t# vector (2D)" %}
  ins_encode %{
    __ fsub(as_FloatRegister($dst$$reg), __ T2D,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vdop_fp128);
%}

// --------------------------------- MUL --------------------------------------

instruct vmul4S(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2 ||
            n->as_Vector()->length() == 4);
  match(Set dst (MulVS src1 src2));
  ins_cost(INSN_COST);
  format %{ "mulv  $dst,$src1,$src2\t# vector (4H)" %}
  ins_encode %{
    __ mulv(as_FloatRegister($dst$$reg), __ T4H,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmul64);
%}

instruct vmul8S(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 8);
  match(Set dst (MulVS src1 src2));
  ins_cost(INSN_COST);
  format %{ "mulv  $dst,$src1,$src2\t# vector (8H)" %}
  ins_encode %{
    __ mulv(as_FloatRegister($dst$$reg), __ T8H,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmul128);
%}

instruct vmul2I(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (MulVI src1 src2));
  ins_cost(INSN_COST);
  format %{ "mulv  $dst,$src1,$src2\t# vector (2S)" %}
  ins_encode %{
    __ mulv(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmul64);
%}

instruct vmul4I(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (MulVI src1 src2));
  ins_cost(INSN_COST);
  format %{ "mulv  $dst,$src1,$src2\t# vector (4S)" %}
  ins_encode %{
    __ mulv(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmul128);
%}

instruct vmul2F(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (MulVF src1 src2));
  ins_cost(INSN_COST);
  format %{ "fmul  $dst,$src1,$src2\t# vector (2S)" %}
  ins_encode %{
    __ fmul(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmuldiv_fp64);
%}

instruct vmul4F(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (MulVF src1 src2));
  ins_cost(INSN_COST);
  format %{ "fmul  $dst,$src1,$src2\t# vector (4S)" %}
  ins_encode %{
    __ fmul(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmuldiv_fp128);
%}

instruct vmul2D(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (MulVD src1 src2));
  ins_cost(INSN_COST);
  format %{ "fmul  $dst,$src1,$src2\t# vector (2D)" %}
  ins_encode %{
    __ fmul(as_FloatRegister($dst$$reg), __ T2D,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmuldiv_fp128);
%}

// --------------------------------- MLA --------------------------------------

instruct vmla4S(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2 ||
            n->as_Vector()->length() == 4);
  match(Set dst (AddVS dst (MulVS src1 src2)));
  ins_cost(INSN_COST);
  format %{ "mlav  $dst,$src1,$src2\t# vector (4H)" %}
  ins_encode %{
    __ mlav(as_FloatRegister($dst$$reg), __ T4H,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmla64);
%}

instruct vmla8S(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 8);
  match(Set dst (AddVS dst (MulVS src1 src2)));
  ins_cost(INSN_COST);
  format %{ "mlav  $dst,$src1,$src2\t# vector (8H)" %}
  ins_encode %{
    __ mlav(as_FloatRegister($dst$$reg), __ T8H,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmla128);
%}

instruct vmla2I(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (AddVI dst (MulVI src1 src2)));
  ins_cost(INSN_COST);
  format %{ "mlav  $dst,$src1,$src2\t# vector (2S)" %}
  ins_encode %{
    __ mlav(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmla64);
%}

instruct vmla4I(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (AddVI dst (MulVI src1 src2)));
  ins_cost(INSN_COST);
  format %{ "mlav  $dst,$src1,$src2\t# vector (4S)" %}
  ins_encode %{
    __ mlav(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmla128);
%}

// --------------------------------- MLS --------------------------------------

instruct vmls4S(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2 ||
            n->as_Vector()->length() == 4);
  match(Set dst (SubVS dst (MulVS src1 src2)));
  ins_cost(INSN_COST);
  format %{ "mlsv  $dst,$src1,$src2\t# vector (4H)" %}
  ins_encode %{
    __ mlsv(as_FloatRegister($dst$$reg), __ T4H,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmla64);
%}

instruct vmls8S(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 8);
  match(Set dst (SubVS dst (MulVS src1 src2)));
  ins_cost(INSN_COST);
  format %{ "mlsv  $dst,$src1,$src2\t# vector (8H)" %}
  ins_encode %{
    __ mlsv(as_FloatRegister($dst$$reg), __ T8H,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmla128);
%}

instruct vmls2I(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (SubVI dst (MulVI src1 src2)));
  ins_cost(INSN_COST);
  format %{ "mlsv  $dst,$src1,$src2\t# vector (2S)" %}
  ins_encode %{
    __ mlsv(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmla64);
%}

instruct vmls4I(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (SubVI dst (MulVI src1 src2)));
  ins_cost(INSN_COST);
  format %{ "mlsv  $dst,$src1,$src2\t# vector (4S)" %}
  ins_encode %{
    __ mlsv(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmla128);
%}

// --------------------------------- DIV --------------------------------------

instruct vdiv2F(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (DivVF src1 src2));
  ins_cost(INSN_COST);
  format %{ "fdiv  $dst,$src1,$src2\t# vector (2S)" %}
  ins_encode %{
    __ fdiv(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmuldiv_fp64);
%}

instruct vdiv4F(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (DivVF src1 src2));
  ins_cost(INSN_COST);
  format %{ "fdiv  $dst,$src1,$src2\t# vector (4S)" %}
  ins_encode %{
    __ fdiv(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmuldiv_fp128);
%}

instruct vdiv2D(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (DivVD src1 src2));
  ins_cost(INSN_COST);
  format %{ "fdiv  $dst,$src1,$src2\t# vector (2D)" %}
  ins_encode %{
    __ fdiv(as_FloatRegister($dst$$reg), __ T2D,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vmuldiv_fp128);
%}

// --------------------------------- SQRT -------------------------------------

instruct vsqrt2D(vecX dst, vecX src)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (SqrtVD src));
  format %{ "fsqrt  $dst, $src\t# vector (2D)" %}
  ins_encode %{
    __ fsqrt(as_FloatRegister($dst$$reg), __ T2D,
             as_FloatRegister($src$$reg));
  %}
  ins_pipe(vsqrt_fp128);
%}

// --------------------------------- ABS --------------------------------------

instruct vabs2F(vecD dst, vecD src)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (AbsVF src));
  ins_cost(INSN_COST * 3);
  format %{ "fabs  $dst,$src\t# vector (2S)" %}
  ins_encode %{
    __ fabs(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src$$reg));
  %}
  ins_pipe(vunop_fp64);
%}

instruct vabs4F(vecX dst, vecX src)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (AbsVF src));
  ins_cost(INSN_COST * 3);
  format %{ "fabs  $dst,$src\t# vector (4S)" %}
  ins_encode %{
    __ fabs(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src$$reg));
  %}
  ins_pipe(vunop_fp128);
%}

instruct vabs2D(vecX dst, vecX src)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (AbsVD src));
  ins_cost(INSN_COST * 3);
  format %{ "fabs  $dst,$src\t# vector (2D)" %}
  ins_encode %{
    __ fabs(as_FloatRegister($dst$$reg), __ T2D,
            as_FloatRegister($src$$reg));
  %}
  ins_pipe(vunop_fp128);
%}

// --------------------------------- NEG --------------------------------------

instruct vneg2F(vecD dst, vecD src)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (NegVF src));
  ins_cost(INSN_COST * 3);
  format %{ "fneg  $dst,$src\t# vector (2S)" %}
  ins_encode %{
    __ fneg(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src$$reg));
  %}
  ins_pipe(vunop_fp64);
%}

instruct vneg4F(vecX dst, vecX src)
%{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (NegVF src));
  ins_cost(INSN_COST * 3);
  format %{ "fneg  $dst,$src\t# vector (4S)" %}
  ins_encode %{
    __ fneg(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src$$reg));
  %}
  ins_pipe(vunop_fp128);
%}

instruct vneg2D(vecX dst, vecX src)
%{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (NegVD src));
  ins_cost(INSN_COST * 3);
  format %{ "fneg  $dst,$src\t# vector (2D)" %}
  ins_encode %{
    __ fneg(as_FloatRegister($dst$$reg), __ T2D,
            as_FloatRegister($src$$reg));
  %}
  ins_pipe(vunop_fp128);
%}

// --------------------------------- AND --------------------------------------

instruct vand8B(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length_in_bytes() == 4 ||
            n->as_Vector()->length_in_bytes() == 8);
  match(Set dst (AndV src1 src2));
  ins_cost(INSN_COST);
  format %{ "and  $dst,$src1,$src2\t# vector (8B)" %}
  ins_encode %{
    __ andr(as_FloatRegister($dst$$reg), __ T8B,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vlogical64);
%}

instruct vand16B(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length_in_bytes() == 16);
  match(Set dst (AndV src1 src2));
  ins_cost(INSN_COST);
  format %{ "and  $dst,$src1,$src2\t# vector (16B)" %}
  ins_encode %{
    __ andr(as_FloatRegister($dst$$reg), __ T16B,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vlogical128);
%}

// --------------------------------- OR ---------------------------------------

instruct vor8B(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length_in_bytes() == 4 ||
            n->as_Vector()->length_in_bytes() == 8);
  match(Set dst (OrV src1 src2));
  ins_cost(INSN_COST);
  format %{ "and  $dst,$src1,$src2\t# vector (8B)" %}
  ins_encode %{
    __ orr(as_FloatRegister($dst$$reg), __ T8B,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vlogical64);
%}

instruct vor16B(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length_in_bytes() == 16);
  match(Set dst (OrV src1 src2));
  ins_cost(INSN_COST);
  format %{ "orr  $dst,$src1,$src2\t# vector (16B)" %}
  ins_encode %{
    __ orr(as_FloatRegister($dst$$reg), __ T16B,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vlogical128);
%}

// --------------------------------- XOR --------------------------------------

instruct vxor8B(vecD dst, vecD src1, vecD src2)
%{
  predicate(n->as_Vector()->length_in_bytes() == 4 ||
            n->as_Vector()->length_in_bytes() == 8);
  match(Set dst (XorV src1 src2));
  ins_cost(INSN_COST);
  format %{ "xor  $dst,$src1,$src2\t# vector (8B)" %}
  ins_encode %{
    __ eor(as_FloatRegister($dst$$reg), __ T8B,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vlogical64);
%}

instruct vxor16B(vecX dst, vecX src1, vecX src2)
%{
  predicate(n->as_Vector()->length_in_bytes() == 16);
  match(Set dst (XorV src1 src2));
  ins_cost(INSN_COST);
  format %{ "xor  $dst,$src1,$src2\t# vector (16B)" %}
  ins_encode %{
    __ eor(as_FloatRegister($dst$$reg), __ T16B,
            as_FloatRegister($src1$$reg),
            as_FloatRegister($src2$$reg));
  %}
  ins_pipe(vlogical128);
%}

// ------------------------------ Shift ---------------------------------------

instruct vshiftcntL(vecX dst, iRegIorL2I cnt) %{
  match(Set dst (LShiftCntV cnt));
  format %{ "dup  $dst, $cnt\t# shift count (vecX)" %}
  ins_encode %{
    __ dup(as_FloatRegister($dst$$reg), __ T16B, as_Register($cnt$$reg));
  %}
  ins_pipe(vdup_reg_reg128);
%}

// Right shifts on aarch64 SIMD are implemented as left shift by -ve amount
instruct vshiftcntR(vecX dst, iRegIorL2I cnt) %{
  match(Set dst (RShiftCntV cnt));
  format %{ "dup  $dst, $cnt\t# shift count (vecX)\n\tneg  $dst, $dst\t T16B" %}
  ins_encode %{
    __ dup(as_FloatRegister($dst$$reg), __ T16B, as_Register($cnt$$reg));
    __ negr(as_FloatRegister($dst$$reg), __ T16B, as_FloatRegister($dst$$reg));
  %}
  ins_pipe(vdup_reg_reg128);
%}

instruct vsll8B(vecD dst, vecD src, vecX shift) %{
  predicate(n->as_Vector()->length() == 4 ||
            n->as_Vector()->length() == 8);
  match(Set dst (LShiftVB src shift));
  match(Set dst (RShiftVB src shift));
  ins_cost(INSN_COST);
  format %{ "sshl  $dst,$src,$shift\t# vector (8B)" %}
  ins_encode %{
    __ sshl(as_FloatRegister($dst$$reg), __ T8B,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift64);
%}

instruct vsll16B(vecX dst, vecX src, vecX shift) %{
  predicate(n->as_Vector()->length() == 16);
  match(Set dst (LShiftVB src shift));
  match(Set dst (RShiftVB src shift));
  ins_cost(INSN_COST);
  format %{ "sshl  $dst,$src,$shift\t# vector (16B)" %}
  ins_encode %{
    __ sshl(as_FloatRegister($dst$$reg), __ T16B,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift128);
%}

instruct vsrl8B(vecD dst, vecD src, vecX shift) %{
  predicate(n->as_Vector()->length() == 4 ||
            n->as_Vector()->length() == 8);
  match(Set dst (URShiftVB src shift));
  ins_cost(INSN_COST);
  format %{ "ushl  $dst,$src,$shift\t# vector (8B)" %}
  ins_encode %{
    __ ushl(as_FloatRegister($dst$$reg), __ T8B,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift64);
%}

instruct vsrl16B(vecX dst, vecX src, vecX shift) %{
  predicate(n->as_Vector()->length() == 16);
  match(Set dst (URShiftVB src shift));
  ins_cost(INSN_COST);
  format %{ "ushl  $dst,$src,$shift\t# vector (16B)" %}
  ins_encode %{
    __ ushl(as_FloatRegister($dst$$reg), __ T16B,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift128);
%}

instruct vsll8B_imm(vecD dst, vecD src, immI shift) %{
  predicate(n->as_Vector()->length() == 4 ||
            n->as_Vector()->length() == 8);
  match(Set dst (LShiftVB src shift));
  ins_cost(INSN_COST);
  format %{ "shl    $dst, $src, $shift\t# vector (8B)" %}
  ins_encode %{
    int sh = (int)$shift$$constant & 31;
    if (sh >= 8) {
      __ eor(as_FloatRegister($dst$$reg), __ T8B,
             as_FloatRegister($src$$reg),
             as_FloatRegister($src$$reg));
    } else {
      __ shl(as_FloatRegister($dst$$reg), __ T8B,
             as_FloatRegister($src$$reg), sh);
    }
  %}
  ins_pipe(vshift64_imm);
%}

instruct vsll16B_imm(vecX dst, vecX src, immI shift) %{
  predicate(n->as_Vector()->length() == 16);
  match(Set dst (LShiftVB src shift));
  ins_cost(INSN_COST);
  format %{ "shl    $dst, $src, $shift\t# vector (16B)" %}
  ins_encode %{
    int sh = (int)$shift$$constant & 31;
    if (sh >= 8) {
      __ eor(as_FloatRegister($dst$$reg), __ T16B,
             as_FloatRegister($src$$reg),
             as_FloatRegister($src$$reg));
    } else {
      __ shl(as_FloatRegister($dst$$reg), __ T16B,
             as_FloatRegister($src$$reg), sh);
    }
  %}
  ins_pipe(vshift128_imm);
%}

instruct vsra8B_imm(vecD dst, vecD src, immI shift) %{
  predicate(n->as_Vector()->length() == 4 ||
            n->as_Vector()->length() == 8);
  match(Set dst (RShiftVB src shift));
  ins_cost(INSN_COST);
  format %{ "sshr    $dst, $src, $shift\t# vector (8B)" %}
  ins_encode %{
    int sh = (int)$shift$$constant & 31;
    if (sh >= 8) sh = 7;
    sh = -sh & 7;
    __ sshr(as_FloatRegister($dst$$reg), __ T8B,
           as_FloatRegister($src$$reg), sh);
  %}
  ins_pipe(vshift64_imm);
%}

instruct vsra16B_imm(vecX dst, vecX src, immI shift) %{
  predicate(n->as_Vector()->length() == 16);
  match(Set dst (RShiftVB src shift));
  ins_cost(INSN_COST);
  format %{ "sshr    $dst, $src, $shift\t# vector (16B)" %}
  ins_encode %{
    int sh = (int)$shift$$constant & 31;
    if (sh >= 8) sh = 7;
    sh = -sh & 7;
    __ sshr(as_FloatRegister($dst$$reg), __ T16B,
           as_FloatRegister($src$$reg), sh);
  %}
  ins_pipe(vshift128_imm);
%}

instruct vsrl8B_imm(vecD dst, vecD src, immI shift) %{
  predicate(n->as_Vector()->length() == 4 ||
            n->as_Vector()->length() == 8);
  match(Set dst (URShiftVB src shift));
  ins_cost(INSN_COST);
  format %{ "ushr    $dst, $src, $shift\t# vector (8B)" %}
  ins_encode %{
    int sh = (int)$shift$$constant & 31;
    if (sh >= 8) {
      __ eor(as_FloatRegister($dst$$reg), __ T8B,
             as_FloatRegister($src$$reg),
             as_FloatRegister($src$$reg));
    } else {
      __ ushr(as_FloatRegister($dst$$reg), __ T8B,
             as_FloatRegister($src$$reg), -sh & 7);
    }
  %}
  ins_pipe(vshift64_imm);
%}

instruct vsrl16B_imm(vecX dst, vecX src, immI shift) %{
  predicate(n->as_Vector()->length() == 16);
  match(Set dst (URShiftVB src shift));
  ins_cost(INSN_COST);
  format %{ "ushr    $dst, $src, $shift\t# vector (16B)" %}
  ins_encode %{
    int sh = (int)$shift$$constant & 31;
    if (sh >= 8) {
      __ eor(as_FloatRegister($dst$$reg), __ T16B,
             as_FloatRegister($src$$reg),
             as_FloatRegister($src$$reg));
    } else {
      __ ushr(as_FloatRegister($dst$$reg), __ T16B,
             as_FloatRegister($src$$reg), -sh & 7);
    }
  %}
  ins_pipe(vshift128_imm);
%}

instruct vsll4S(vecD dst, vecD src, vecX shift) %{
  predicate(n->as_Vector()->length() == 2 ||
            n->as_Vector()->length() == 4);
  match(Set dst (LShiftVS src shift));
  match(Set dst (RShiftVS src shift));
  ins_cost(INSN_COST);
  format %{ "sshl  $dst,$src,$shift\t# vector (4H)" %}
  ins_encode %{
    __ sshl(as_FloatRegister($dst$$reg), __ T4H,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift64);
%}

instruct vsll8S(vecX dst, vecX src, vecX shift) %{
  predicate(n->as_Vector()->length() == 8);
  match(Set dst (LShiftVS src shift));
  match(Set dst (RShiftVS src shift));
  ins_cost(INSN_COST);
  format %{ "sshl  $dst,$src,$shift\t# vector (8H)" %}
  ins_encode %{
    __ sshl(as_FloatRegister($dst$$reg), __ T8H,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift128);
%}

instruct vsrl4S(vecD dst, vecD src, vecX shift) %{
  predicate(n->as_Vector()->length() == 2 ||
            n->as_Vector()->length() == 4);
  match(Set dst (URShiftVS src shift));
  ins_cost(INSN_COST);
  format %{ "ushl  $dst,$src,$shift\t# vector (4H)" %}
  ins_encode %{
    __ ushl(as_FloatRegister($dst$$reg), __ T4H,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift64);
%}

instruct vsrl8S(vecX dst, vecX src, vecX shift) %{
  predicate(n->as_Vector()->length() == 8);
  match(Set dst (URShiftVS src shift));
  ins_cost(INSN_COST);
  format %{ "ushl  $dst,$src,$shift\t# vector (8H)" %}
  ins_encode %{
    __ ushl(as_FloatRegister($dst$$reg), __ T8H,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift128);
%}

instruct vsll4S_imm(vecD dst, vecD src, immI shift) %{
  predicate(n->as_Vector()->length() == 2 ||
            n->as_Vector()->length() == 4);
  match(Set dst (LShiftVS src shift));
  ins_cost(INSN_COST);
  format %{ "shl    $dst, $src, $shift\t# vector (4H)" %}
  ins_encode %{
    int sh = (int)$shift$$constant & 31;
    if (sh >= 16) {
      __ eor(as_FloatRegister($dst$$reg), __ T8B,
             as_FloatRegister($src$$reg),
             as_FloatRegister($src$$reg));
    } else {
      __ shl(as_FloatRegister($dst$$reg), __ T4H,
             as_FloatRegister($src$$reg), sh);
    }
  %}
  ins_pipe(vshift64_imm);
%}

instruct vsll8S_imm(vecX dst, vecX src, immI shift) %{
  predicate(n->as_Vector()->length() == 8);
  match(Set dst (LShiftVS src shift));
  ins_cost(INSN_COST);
  format %{ "shl    $dst, $src, $shift\t# vector (8H)" %}
  ins_encode %{
    int sh = (int)$shift$$constant & 31;
    if (sh >= 16) {
      __ eor(as_FloatRegister($dst$$reg), __ T16B,
             as_FloatRegister($src$$reg),
             as_FloatRegister($src$$reg));
    } else {
      __ shl(as_FloatRegister($dst$$reg), __ T8H,
             as_FloatRegister($src$$reg), sh);
    }
  %}
  ins_pipe(vshift128_imm);
%}

instruct vsra4S_imm(vecD dst, vecD src, immI shift) %{
  predicate(n->as_Vector()->length() == 2 ||
            n->as_Vector()->length() == 4);
  match(Set dst (RShiftVS src shift));
  ins_cost(INSN_COST);
  format %{ "sshr    $dst, $src, $shift\t# vector (4H)" %}
  ins_encode %{
    int sh = (int)$shift$$constant & 31;
    if (sh >= 16) sh = 15;
    sh = -sh & 15;
    __ sshr(as_FloatRegister($dst$$reg), __ T4H,
           as_FloatRegister($src$$reg), sh);
  %}
  ins_pipe(vshift64_imm);
%}

instruct vsra8S_imm(vecX dst, vecX src, immI shift) %{
  predicate(n->as_Vector()->length() == 8);
  match(Set dst (RShiftVS src shift));
  ins_cost(INSN_COST);
  format %{ "sshr    $dst, $src, $shift\t# vector (8H)" %}
  ins_encode %{
    int sh = (int)$shift$$constant & 31;
    if (sh >= 16) sh = 15;
    sh = -sh & 15;
    __ sshr(as_FloatRegister($dst$$reg), __ T8H,
           as_FloatRegister($src$$reg), sh);
  %}
  ins_pipe(vshift128_imm);
%}

instruct vsrl4S_imm(vecD dst, vecD src, immI shift) %{
  predicate(n->as_Vector()->length() == 2 ||
            n->as_Vector()->length() == 4);
  match(Set dst (URShiftVS src shift));
  ins_cost(INSN_COST);
  format %{ "ushr    $dst, $src, $shift\t# vector (4H)" %}
  ins_encode %{
    int sh = (int)$shift$$constant & 31;
    if (sh >= 16) {
      __ eor(as_FloatRegister($dst$$reg), __ T8B,
             as_FloatRegister($src$$reg),
             as_FloatRegister($src$$reg));
    } else {
      __ ushr(as_FloatRegister($dst$$reg), __ T4H,
             as_FloatRegister($src$$reg), -sh & 15);
    }
  %}
  ins_pipe(vshift64_imm);
%}

instruct vsrl8S_imm(vecX dst, vecX src, immI shift) %{
  predicate(n->as_Vector()->length() == 8);
  match(Set dst (URShiftVS src shift));
  ins_cost(INSN_COST);
  format %{ "ushr    $dst, $src, $shift\t# vector (8H)" %}
  ins_encode %{
    int sh = (int)$shift$$constant & 31;
    if (sh >= 16) {
      __ eor(as_FloatRegister($dst$$reg), __ T16B,
             as_FloatRegister($src$$reg),
             as_FloatRegister($src$$reg));
    } else {
      __ ushr(as_FloatRegister($dst$$reg), __ T8H,
             as_FloatRegister($src$$reg), -sh & 15);
    }
  %}
  ins_pipe(vshift128_imm);
%}

instruct vsll2I(vecD dst, vecD src, vecX shift) %{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (LShiftVI src shift));
  match(Set dst (RShiftVI src shift));
  ins_cost(INSN_COST);
  format %{ "sshl  $dst,$src,$shift\t# vector (2S)" %}
  ins_encode %{
    __ sshl(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift64);
%}

instruct vsll4I(vecX dst, vecX src, vecX shift) %{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (LShiftVI src shift));
  match(Set dst (RShiftVI src shift));
  ins_cost(INSN_COST);
  format %{ "sshl  $dst,$src,$shift\t# vector (4S)" %}
  ins_encode %{
    __ sshl(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift128);
%}

instruct vsrl2I(vecD dst, vecD src, vecX shift) %{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (URShiftVI src shift));
  ins_cost(INSN_COST);
  format %{ "ushl  $dst,$src,$shift\t# vector (2S)" %}
  ins_encode %{
    __ ushl(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift64);
%}

instruct vsrl4I(vecX dst, vecX src, vecX shift) %{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (URShiftVI src shift));
  ins_cost(INSN_COST);
  format %{ "ushl  $dst,$src,$shift\t# vector (4S)" %}
  ins_encode %{
    __ ushl(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift128);
%}

instruct vsll2I_imm(vecD dst, vecD src, immI shift) %{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (LShiftVI src shift));
  ins_cost(INSN_COST);
  format %{ "shl    $dst, $src, $shift\t# vector (2S)" %}
  ins_encode %{
    __ shl(as_FloatRegister($dst$$reg), __ T2S,
           as_FloatRegister($src$$reg),
           (int)$shift$$constant & 31);
  %}
  ins_pipe(vshift64_imm);
%}

instruct vsll4I_imm(vecX dst, vecX src, immI shift) %{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (LShiftVI src shift));
  ins_cost(INSN_COST);
  format %{ "shl    $dst, $src, $shift\t# vector (4S)" %}
  ins_encode %{
    __ shl(as_FloatRegister($dst$$reg), __ T4S,
           as_FloatRegister($src$$reg),
           (int)$shift$$constant & 31);
  %}
  ins_pipe(vshift128_imm);
%}

instruct vsra2I_imm(vecD dst, vecD src, immI shift) %{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (RShiftVI src shift));
  ins_cost(INSN_COST);
  format %{ "sshr    $dst, $src, $shift\t# vector (2S)" %}
  ins_encode %{
    __ sshr(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src$$reg),
            -(int)$shift$$constant & 31);
  %}
  ins_pipe(vshift64_imm);
%}

instruct vsra4I_imm(vecX dst, vecX src, immI shift) %{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (RShiftVI src shift));
  ins_cost(INSN_COST);
  format %{ "sshr    $dst, $src, $shift\t# vector (4S)" %}
  ins_encode %{
    __ sshr(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src$$reg),
            -(int)$shift$$constant & 31);
  %}
  ins_pipe(vshift128_imm);
%}

instruct vsrl2I_imm(vecD dst, vecD src, immI shift) %{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (URShiftVI src shift));
  ins_cost(INSN_COST);
  format %{ "ushr    $dst, $src, $shift\t# vector (2S)" %}
  ins_encode %{
    __ ushr(as_FloatRegister($dst$$reg), __ T2S,
            as_FloatRegister($src$$reg),
            -(int)$shift$$constant & 31);
  %}
  ins_pipe(vshift64_imm);
%}

instruct vsrl4I_imm(vecX dst, vecX src, immI shift) %{
  predicate(n->as_Vector()->length() == 4);
  match(Set dst (URShiftVI src shift));
  ins_cost(INSN_COST);
  format %{ "ushr    $dst, $src, $shift\t# vector (4S)" %}
  ins_encode %{
    __ ushr(as_FloatRegister($dst$$reg), __ T4S,
            as_FloatRegister($src$$reg),
            -(int)$shift$$constant & 31);
  %}
  ins_pipe(vshift128_imm);
%}

instruct vsll2L(vecX dst, vecX src, vecX shift) %{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (LShiftVL src shift));
  match(Set dst (RShiftVL src shift));
  ins_cost(INSN_COST);
  format %{ "sshl  $dst,$src,$shift\t# vector (2D)" %}
  ins_encode %{
    __ sshl(as_FloatRegister($dst$$reg), __ T2D,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift128);
%}

instruct vsrl2L(vecX dst, vecX src, vecX shift) %{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (URShiftVL src shift));
  ins_cost(INSN_COST);
  format %{ "ushl  $dst,$src,$shift\t# vector (2D)" %}
  ins_encode %{
    __ ushl(as_FloatRegister($dst$$reg), __ T2D,
            as_FloatRegister($src$$reg),
            as_FloatRegister($shift$$reg));
  %}
  ins_pipe(vshift128);
%}

instruct vsll2L_imm(vecX dst, vecX src, immI shift) %{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (LShiftVL src shift));
  ins_cost(INSN_COST);
  format %{ "shl    $dst, $src, $shift\t# vector (2D)" %}
  ins_encode %{
    __ shl(as_FloatRegister($dst$$reg), __ T2D,
           as_FloatRegister($src$$reg),
           (int)$shift$$constant & 63);
  %}
  ins_pipe(vshift128_imm);
%}

instruct vsra2L_imm(vecX dst, vecX src, immI shift) %{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (RShiftVL src shift));
  ins_cost(INSN_COST);
  format %{ "sshr    $dst, $src, $shift\t# vector (2D)" %}
  ins_encode %{
    __ sshr(as_FloatRegister($dst$$reg), __ T2D,
            as_FloatRegister($src$$reg),
            -(int)$shift$$constant & 63);
  %}
  ins_pipe(vshift128_imm);
%}

instruct vsrl2L_imm(vecX dst, vecX src, immI shift) %{
  predicate(n->as_Vector()->length() == 2);
  match(Set dst (URShiftVL src shift));
  ins_cost(INSN_COST);
  format %{ "ushr    $dst, $src, $shift\t# vector (2D)" %}
  ins_encode %{
    __ ushr(as_FloatRegister($dst$$reg), __ T2D,
            as_FloatRegister($src$$reg),
            -(int)$shift$$constant & 63);
  %}
  ins_pipe(vshift128_imm);
%}

//----------PEEPHOLE RULES-----------------------------------------------------
// These must follow all instruction definitions as they use the names
// defined in the instructions definitions.
//
// peepmatch ( root_instr_name [preceding_instruction]* );
//
// peepconstraint %{
// (instruction_number.operand_name relational_op instruction_number.operand_name
//  [, ...] );
// // instruction numbers are zero-based using left to right order in peepmatch
//
// peepreplace ( instr_name  ( [instruction_number.operand_name]* ) );
// // provide an instruction_number.operand_name for each operand that appears
// // in the replacement instruction's match rule
//
// ---------VM FLAGS---------------------------------------------------------
//
// All peephole optimizations can be turned off using -XX:-OptoPeephole
//
// Each peephole rule is given an identifying number starting with zero and
// increasing by one in the order seen by the parser.  An individual peephole
// can be enabled, and all others disabled, by using -XX:OptoPeepholeAt=#
// on the command-line.
//
// ---------CURRENT LIMITATIONS----------------------------------------------
//
// Only match adjacent instructions in same basic block
// Only equality constraints
// Only constraints between operands, not (0.dest_reg == RAX_enc)
// Only one replacement instruction
//
// ---------EXAMPLE----------------------------------------------------------
//
// // pertinent parts of existing instructions in architecture description
// instruct movI(iRegINoSp dst, iRegI src)
// %{
//   match(Set dst (CopyI src));
// %}
//
// instruct incI_iReg(iRegINoSp dst, immI1 src, rFlagsReg cr)
// %{
//   match(Set dst (AddI dst src));
//   effect(KILL cr);
// %}
//
// // Change (inc mov) to lea
// peephole %{
//   // increment preceeded by register-register move
//   peepmatch ( incI_iReg movI );
//   // require that the destination register of the increment
//   // match the destination register of the move
//   peepconstraint ( 0.dst == 1.dst );
//   // construct a replacement instruction that sets
//   // the destination to ( move's source register + one )
//   peepreplace ( leaI_iReg_immI( 0.dst 1.src 0.src ) );
// %}
//

// Implementation no longer uses movX instructions since
// machine-independent system no longer uses CopyX nodes.
//
// peephole
// %{
//   peepmatch (incI_iReg movI);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaI_iReg_immI(0.dst 1.src 0.src));
// %}

// peephole
// %{
//   peepmatch (decI_iReg movI);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaI_iReg_immI(0.dst 1.src 0.src));
// %}

// peephole
// %{
//   peepmatch (addI_iReg_imm movI);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaI_iReg_immI(0.dst 1.src 0.src));
// %}

// peephole
// %{
//   peepmatch (incL_iReg movL);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaL_iReg_immL(0.dst 1.src 0.src));
// %}

// peephole
// %{
//   peepmatch (decL_iReg movL);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaL_iReg_immL(0.dst 1.src 0.src));
// %}

// peephole
// %{
//   peepmatch (addL_iReg_imm movL);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaL_iReg_immL(0.dst 1.src 0.src));
// %}

// peephole
// %{
//   peepmatch (addP_iReg_imm movP);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaP_iReg_imm(0.dst 1.src 0.src));
// %}

// // Change load of spilled value to only a spill
// instruct storeI(memory mem, iRegI src)
// %{
//   match(Set mem (StoreI mem src));
// %}
//
// instruct loadI(iRegINoSp dst, memory mem)
// %{
//   match(Set dst (LoadI mem));
// %}
//

//----------SMARTSPILL RULES---------------------------------------------------
// These must follow all instruction definitions as they use the names
// defined in the instructions definitions.

// Local Variables:
// mode: c++
// End: