author | kvn |
Thu, 21 Jul 2011 11:25:07 -0700 | |
changeset 10252 | 0981ce1c3eef |
parent 10027 | 20cd71f29262 |
child 10264 | 6879f93d268d |
permissions | -rw-r--r-- |
1 | 1 |
/* |
7115
32300e243300
6987135: Performance regression on Intel platform with 32-bits edition between 6u13 and 6u14.
kvn
parents:
6272
diff
changeset
|
2 |
* Copyright (c) 1997, 2010, Oracle and/or its affiliates. All rights reserved. |
1 | 3 |
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. |
4 |
* |
|
5 |
* This code is free software; you can redistribute it and/or modify it |
|
6 |
* under the terms of the GNU General Public License version 2 only, as |
|
7 |
* published by the Free Software Foundation. |
|
8 |
* |
|
9 |
* This code is distributed in the hope that it will be useful, but WITHOUT |
|
10 |
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
|
11 |
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
|
12 |
* version 2 for more details (a copy is included in the LICENSE file that |
|
13 |
* accompanied this code). |
|
14 |
* |
|
15 |
* You should have received a copy of the GNU General Public License version |
|
16 |
* 2 along with this work; if not, write to the Free Software Foundation, |
|
17 |
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. |
|
18 |
* |
|
5547
f4b087cbb361
6941466: Oracle rebranding changes for Hotspot repositories
trims
parents:
5431
diff
changeset
|
19 |
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA |
f4b087cbb361
6941466: Oracle rebranding changes for Hotspot repositories
trims
parents:
5431
diff
changeset
|
20 |
* or visit www.oracle.com if you need additional information or have any |
f4b087cbb361
6941466: Oracle rebranding changes for Hotspot repositories
trims
parents:
5431
diff
changeset
|
21 |
* questions. |
1 | 22 |
* |
23 |
*/ |
|
24 |
||
7397 | 25 |
#include "precompiled.hpp" |
26 |
#include "assembler_sparc.inline.hpp" |
|
27 |
#include "memory/resourceArea.hpp" |
|
28 |
#include "runtime/java.hpp" |
|
29 |
#include "runtime/stubCodeGenerator.hpp" |
|
30 |
#include "vm_version_sparc.hpp" |
|
31 |
#ifdef TARGET_OS_FAMILY_linux |
|
32 |
# include "os_linux.inline.hpp" |
|
33 |
#endif |
|
34 |
#ifdef TARGET_OS_FAMILY_solaris |
|
35 |
# include "os_solaris.inline.hpp" |
|
36 |
#endif |
|
1 | 37 |
|
38 |
int VM_Version::_features = VM_Version::unknown_m; |
|
39 |
const char* VM_Version::_features_str = ""; |
|
40 |
||
41 |
void VM_Version::initialize() { |
|
42 |
_features = determine_features(); |
|
43 |
PrefetchCopyIntervalInBytes = prefetch_copy_interval_in_bytes(); |
|
44 |
PrefetchScanIntervalInBytes = prefetch_scan_interval_in_bytes(); |
|
45 |
PrefetchFieldsAhead = prefetch_fields_ahead(); |
|
46 |
||
47 |
// Allocation prefetch settings |
|
48 |
intx cache_line_size = L1_data_cache_line_size(); |
|
49 |
if( cache_line_size > AllocatePrefetchStepSize ) |
|
50 |
AllocatePrefetchStepSize = cache_line_size; |
|
51 |
if( FLAG_IS_DEFAULT(AllocatePrefetchLines) ) |
|
52 |
AllocatePrefetchLines = 3; // Optimistic value |
|
53 |
assert( AllocatePrefetchLines > 0, "invalid value"); |
|
54 |
if( AllocatePrefetchLines < 1 ) // set valid value in product VM |
|
55 |
AllocatePrefetchLines = 1; // Conservative value |
|
56 |
||
57 |
AllocatePrefetchDistance = allocate_prefetch_distance(); |
|
58 |
AllocatePrefetchStyle = allocate_prefetch_style(); |
|
59 |
||
60 |
assert(AllocatePrefetchDistance % AllocatePrefetchStepSize == 0, "invalid value"); |
|
61 |
||
10252 | 62 |
if (AllocatePrefetchStyle == 3 && !has_blk_init()) { |
63 |
warning("BIS instructions are not available on this CPU"); |
|
64 |
FLAG_SET_DEFAULT(AllocatePrefetchStyle, 1); |
|
65 |
} |
|
66 |
||
1 | 67 |
UseSSE = 0; // Only on x86 and x64 |
68 |
||
69 |
_supports_cx8 = has_v9(); |
|
70 |
||
7704 | 71 |
if (is_niagara()) { |
1 | 72 |
// Indirect branch is the same cost as direct |
73 |
if (FLAG_IS_DEFAULT(UseInlineCaches)) { |
|
2342 | 74 |
FLAG_SET_DEFAULT(UseInlineCaches, false); |
1 | 75 |
} |
7704 | 76 |
// Align loops on a single instruction boundary. |
77 |
if (FLAG_IS_DEFAULT(OptoLoopAlignment)) { |
|
78 |
FLAG_SET_DEFAULT(OptoLoopAlignment, 4); |
|
79 |
} |
|
80 |
// When using CMS, we cannot use memset() in BOT updates because |
|
81 |
// the sun4v/CMT version in libc_psr uses BIS which exposes |
|
82 |
// "phantom zeros" to concurrent readers. See 6948537. |
|
83 |
if (FLAG_IS_DEFAULT(UseMemSetInBOT) && UseConcMarkSweepGC) { |
|
84 |
FLAG_SET_DEFAULT(UseMemSetInBOT, false); |
|
85 |
} |
|
360
21d113ecbf6a
6420645: Create a vm that uses compressed oops for up to 32gb heapsizes
coleenp
parents:
183
diff
changeset
|
86 |
#ifdef _LP64 |
2254
f13dda645a4b
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
2253
diff
changeset
|
87 |
// 32-bit oops don't make sense for the 64-bit VM on sparc |
f13dda645a4b
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
2253
diff
changeset
|
88 |
// since the 32-bit VM has the same registers and smaller objects. |
f13dda645a4b
6791178: Specialize for zero as the compressed oop vm heap base
kvn
parents:
2253
diff
changeset
|
89 |
Universe::set_narrow_oop_shift(LogMinObjAlignmentInBytes); |
360
21d113ecbf6a
6420645: Create a vm that uses compressed oops for up to 32gb heapsizes
coleenp
parents:
183
diff
changeset
|
90 |
#endif // _LP64 |
1 | 91 |
#ifdef COMPILER2 |
92 |
// Indirect branch is the same cost as direct |
|
93 |
if (FLAG_IS_DEFAULT(UseJumpTables)) { |
|
2342 | 94 |
FLAG_SET_DEFAULT(UseJumpTables, true); |
1 | 95 |
} |
96 |
// Single-issue, so entry and loop tops are |
|
97 |
// aligned on a single instruction boundary |
|
98 |
if (FLAG_IS_DEFAULT(InteriorEntryAlignment)) { |
|
2342 | 99 |
FLAG_SET_DEFAULT(InteriorEntryAlignment, 4); |
1 | 100 |
} |
7704 | 101 |
if (is_niagara_plus()) { |
7115
32300e243300
6987135: Performance regression on Intel platform with 32-bits edition between 6u13 and 6u14.
kvn
parents:
6272
diff
changeset
|
102 |
if (has_blk_init() && AllocatePrefetchStyle > 0 && |
32300e243300
6987135: Performance regression on Intel platform with 32-bits edition between 6u13 and 6u14.
kvn
parents:
6272
diff
changeset
|
103 |
FLAG_IS_DEFAULT(AllocatePrefetchStyle)) { |
5251
f86f7a86d761
6940726: Use BIS instruction for allocation prefetch on Sparc
kvn
parents:
5249
diff
changeset
|
104 |
// Use BIS instruction for allocation prefetch. |
f86f7a86d761
6940726: Use BIS instruction for allocation prefetch on Sparc
kvn
parents:
5249
diff
changeset
|
105 |
FLAG_SET_DEFAULT(AllocatePrefetchStyle, 3); |
f86f7a86d761
6940726: Use BIS instruction for allocation prefetch on Sparc
kvn
parents:
5249
diff
changeset
|
106 |
if (FLAG_IS_DEFAULT(AllocatePrefetchDistance)) { |
f86f7a86d761
6940726: Use BIS instruction for allocation prefetch on Sparc
kvn
parents:
5249
diff
changeset
|
107 |
// Use smaller prefetch distance on N2 with BIS |
f86f7a86d761
6940726: Use BIS instruction for allocation prefetch on Sparc
kvn
parents:
5249
diff
changeset
|
108 |
FLAG_SET_DEFAULT(AllocatePrefetchDistance, 64); |
f86f7a86d761
6940726: Use BIS instruction for allocation prefetch on Sparc
kvn
parents:
5249
diff
changeset
|
109 |
} |
f86f7a86d761
6940726: Use BIS instruction for allocation prefetch on Sparc
kvn
parents:
5249
diff
changeset
|
110 |
} |
f86f7a86d761
6940726: Use BIS instruction for allocation prefetch on Sparc
kvn
parents:
5249
diff
changeset
|
111 |
if (AllocatePrefetchStyle != 3 && FLAG_IS_DEFAULT(AllocatePrefetchDistance)) { |
f86f7a86d761
6940726: Use BIS instruction for allocation prefetch on Sparc
kvn
parents:
5249
diff
changeset
|
112 |
// Use different prefetch distance without BIS |
f86f7a86d761
6940726: Use BIS instruction for allocation prefetch on Sparc
kvn
parents:
5249
diff
changeset
|
113 |
FLAG_SET_DEFAULT(AllocatePrefetchDistance, 256); |
f86f7a86d761
6940726: Use BIS instruction for allocation prefetch on Sparc
kvn
parents:
5249
diff
changeset
|
114 |
} |
1 | 115 |
} |
116 |
#endif |
|
117 |
} |
|
118 |
||
2255
54abdf3e1055
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
2254
diff
changeset
|
119 |
// Use hardware population count instruction if available. |
54abdf3e1055
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
2254
diff
changeset
|
120 |
if (has_hardware_popc()) { |
54abdf3e1055
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
2254
diff
changeset
|
121 |
if (FLAG_IS_DEFAULT(UsePopCountInstruction)) { |
2342 | 122 |
FLAG_SET_DEFAULT(UsePopCountInstruction, true); |
2255
54abdf3e1055
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
2254
diff
changeset
|
123 |
} |
10252 | 124 |
} else if (UsePopCountInstruction) { |
125 |
warning("POPC instruction is not available on this CPU"); |
|
126 |
FLAG_SET_DEFAULT(UsePopCountInstruction, false); |
|
127 |
} |
|
128 |
||
129 |
// T4 and newer Sparc cpus have new compare and branch instruction. |
|
130 |
if (has_cbcond()) { |
|
131 |
if (FLAG_IS_DEFAULT(UseCBCond)) { |
|
132 |
FLAG_SET_DEFAULT(UseCBCond, true); |
|
133 |
} |
|
134 |
} else if (UseCBCond) { |
|
135 |
warning("CBCOND instruction is not available on this CPU"); |
|
136 |
FLAG_SET_DEFAULT(UseCBCond, false); |
|
2255
54abdf3e1055
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
2254
diff
changeset
|
137 |
} |
54abdf3e1055
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
2254
diff
changeset
|
138 |
|
6272
94a20ad0e9de
6978249: spill between cpu and fpu registers when those moves are fast
never
parents:
5702
diff
changeset
|
139 |
#ifdef COMPILER2 |
10252 | 140 |
// T4 and newer Sparc cpus have fast RDPC. |
141 |
if (has_fast_rdpc() && FLAG_IS_DEFAULT(UseRDPCForConstantTableBase)) { |
|
142 |
// FLAG_SET_DEFAULT(UseRDPCForConstantTableBase, true); |
|
143 |
} |
|
144 |
||
6272
94a20ad0e9de
6978249: spill between cpu and fpu registers when those moves are fast
never
parents:
5702
diff
changeset
|
145 |
// Currently not supported anywhere. |
94a20ad0e9de
6978249: spill between cpu and fpu registers when those moves are fast
never
parents:
5702
diff
changeset
|
146 |
FLAG_SET_DEFAULT(UseFPUForSpilling, false); |
94a20ad0e9de
6978249: spill between cpu and fpu registers when those moves are fast
never
parents:
5702
diff
changeset
|
147 |
#endif |
94a20ad0e9de
6978249: spill between cpu and fpu registers when those moves are fast
never
parents:
5702
diff
changeset
|
148 |
|
1 | 149 |
char buf[512]; |
10252 | 150 |
jio_snprintf(buf, sizeof(buf), "%s%s%s%s%s%s%s%s%s%s%s%s%s%s", |
151 |
(has_v9() ? ", v9" : (has_v8() ? ", v8" : "")), |
|
2255
54abdf3e1055
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
2254
diff
changeset
|
152 |
(has_hardware_popc() ? ", popc" : ""), |
10252 | 153 |
(has_vis1() ? ", vis1" : ""), |
154 |
(has_vis2() ? ", vis2" : ""), |
|
155 |
(has_vis3() ? ", vis3" : ""), |
|
156 |
(has_blk_init() ? ", blk_init" : ""), |
|
157 |
(has_cbcond() ? ", cbcond" : ""), |
|
158 |
(is_ultra3() ? ", ultra3" : ""), |
|
159 |
(is_sun4v() ? ", sun4v" : ""), |
|
160 |
(is_niagara_plus() ? ", niagara_plus" : (is_niagara() ? ", niagara" : "")), |
|
161 |
(is_sparc64() ? ", sparc64" : ""), |
|
2253
30268d00878e
6812587: Use auxv to determine SPARC hardware features on Solaris
twisti
parents:
670
diff
changeset
|
162 |
(!has_hardware_mul32() ? ", no-mul32" : ""), |
30268d00878e
6812587: Use auxv to determine SPARC hardware features on Solaris
twisti
parents:
670
diff
changeset
|
163 |
(!has_hardware_div32() ? ", no-div32" : ""), |
1 | 164 |
(!has_hardware_fsmuld() ? ", no-fsmuld" : "")); |
165 |
||
166 |
// buf is started with ", " or is empty |
|
167 |
_features_str = strdup(strlen(buf) > 2 ? buf + 2 : buf); |
|
168 |
||
10027 | 169 |
// UseVIS is set to the smallest of what hardware supports and what |
170 |
// the command line requires. I.e., you cannot set UseVIS to 3 on |
|
171 |
// older UltraSparc which do not support it. |
|
172 |
if (UseVIS > 3) UseVIS=3; |
|
173 |
if (UseVIS < 0) UseVIS=0; |
|
174 |
if (!has_vis3()) // Drop to 2 if no VIS3 support |
|
175 |
UseVIS = MIN2((intx)2,UseVIS); |
|
176 |
if (!has_vis2()) // Drop to 1 if no VIS2 support |
|
177 |
UseVIS = MIN2((intx)1,UseVIS); |
|
178 |
if (!has_vis1()) // Drop to 0 if no VIS1 support |
|
179 |
UseVIS = 0; |
|
180 |
||
1 | 181 |
#ifndef PRODUCT |
182 |
if (PrintMiscellaneous && Verbose) { |
|
183 |
tty->print("Allocation: "); |
|
184 |
if (AllocatePrefetchStyle <= 0) { |
|
185 |
tty->print_cr("no prefetching"); |
|
186 |
} else { |
|
187 |
if (AllocatePrefetchLines > 1) { |
|
188 |
tty->print_cr("PREFETCH %d, %d lines of size %d bytes", AllocatePrefetchDistance, AllocatePrefetchLines, AllocatePrefetchStepSize); |
|
189 |
} else { |
|
190 |
tty->print_cr("PREFETCH %d, one line", AllocatePrefetchDistance); |
|
191 |
} |
|
192 |
} |
|
193 |
if (PrefetchCopyIntervalInBytes > 0) { |
|
194 |
tty->print_cr("PrefetchCopyIntervalInBytes %d", PrefetchCopyIntervalInBytes); |
|
195 |
} |
|
196 |
if (PrefetchScanIntervalInBytes > 0) { |
|
197 |
tty->print_cr("PrefetchScanIntervalInBytes %d", PrefetchScanIntervalInBytes); |
|
198 |
} |
|
199 |
if (PrefetchFieldsAhead > 0) { |
|
200 |
tty->print_cr("PrefetchFieldsAhead %d", PrefetchFieldsAhead); |
|
201 |
} |
|
202 |
} |
|
203 |
#endif // PRODUCT |
|
204 |
} |
|
205 |
||
206 |
void VM_Version::print_features() { |
|
207 |
tty->print_cr("Version:%s", cpu_features()); |
|
208 |
} |
|
209 |
||
210 |
int VM_Version::determine_features() { |
|
211 |
if (UseV8InstrsOnly) { |
|
212 |
NOT_PRODUCT(if (PrintMiscellaneous && Verbose) tty->print_cr("Version is Forced-V8");) |
|
213 |
return generic_v8_m; |
|
214 |
} |
|
215 |
||
216 |
int features = platform_features(unknown_m); // platform_features() is os_arch specific |
|
217 |
||
218 |
if (features == unknown_m) { |
|
219 |
features = generic_v9_m; |
|
220 |
warning("Cannot recognize SPARC version. Default to V9"); |
|
221 |
} |
|
222 |
||
7704 | 223 |
assert(is_T_family(features) == is_niagara(features), "Niagara should be T series"); |
224 |
if (UseNiagaraInstrs) { // Force code generation for Niagara |
|
225 |
if (is_T_family(features)) { |
|
1 | 226 |
// Happy to accomodate... |
227 |
} else { |
|
228 |
NOT_PRODUCT(if (PrintMiscellaneous && Verbose) tty->print_cr("Version is Forced-Niagara");) |
|
7704 | 229 |
features |= T_family_m; |
1 | 230 |
} |
231 |
} else { |
|
7704 | 232 |
if (is_T_family(features) && !FLAG_IS_DEFAULT(UseNiagaraInstrs)) { |
1 | 233 |
NOT_PRODUCT(if (PrintMiscellaneous && Verbose) tty->print_cr("Version is Forced-Not-Niagara");) |
7704 | 234 |
features &= ~(T_family_m | T1_model_m); |
1 | 235 |
} else { |
236 |
// Happy to accomodate... |
|
237 |
} |
|
238 |
} |
|
239 |
||
240 |
return features; |
|
241 |
} |
|
242 |
||
243 |
static int saved_features = 0; |
|
244 |
||
245 |
void VM_Version::allow_all() { |
|
246 |
saved_features = _features; |
|
247 |
_features = all_features_m; |
|
248 |
} |
|
249 |
||
250 |
void VM_Version::revert() { |
|
251 |
_features = saved_features; |
|
252 |
} |
|
183
ba55c7f3fd45
6362677: Change parallel GC collector default number of parallel GC threads.
jmasa
parents:
1
diff
changeset
|
253 |
|
ba55c7f3fd45
6362677: Change parallel GC collector default number of parallel GC threads.
jmasa
parents:
1
diff
changeset
|
254 |
unsigned int VM_Version::calc_parallel_worker_threads() { |
ba55c7f3fd45
6362677: Change parallel GC collector default number of parallel GC threads.
jmasa
parents:
1
diff
changeset
|
255 |
unsigned int result; |
7704 | 256 |
if (is_niagara_plus()) { |
183
ba55c7f3fd45
6362677: Change parallel GC collector default number of parallel GC threads.
jmasa
parents:
1
diff
changeset
|
257 |
result = nof_parallel_worker_threads(5, 16, 8); |
ba55c7f3fd45
6362677: Change parallel GC collector default number of parallel GC threads.
jmasa
parents:
1
diff
changeset
|
258 |
} else { |
ba55c7f3fd45
6362677: Change parallel GC collector default number of parallel GC threads.
jmasa
parents:
1
diff
changeset
|
259 |
result = nof_parallel_worker_threads(5, 8, 8); |
ba55c7f3fd45
6362677: Change parallel GC collector default number of parallel GC threads.
jmasa
parents:
1
diff
changeset
|
260 |
} |
ba55c7f3fd45
6362677: Change parallel GC collector default number of parallel GC threads.
jmasa
parents:
1
diff
changeset
|
261 |
return result; |
ba55c7f3fd45
6362677: Change parallel GC collector default number of parallel GC threads.
jmasa
parents:
1
diff
changeset
|
262 |
} |