Mercurial > hg > truffle
annotate src/cpu/x86/vm/vm_version_x86.hpp @ 18177:89f97291c3a5
Hotspot: Support narrow oop relocations in constants section.
author | Roland Schatz <roland.schatz@oracle.com> |
---|---|
date | Mon, 27 Oct 2014 14:24:02 +0100 |
parents | 52b4284cb496 |
children | 7848fc12602b |
rev | line source |
---|---|
585 | 1 /* |
11080
b800986664f4
7088419: Use x86 Hardware CRC32 Instruction with java.util.zip.CRC32
drchase
parents:
7474
diff
changeset
|
2 * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved. |
585 | 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 * | |
1552
c18cbe5936b8
6941466: Oracle rebranding changes for Hotspot repositories
trims
parents:
775
diff
changeset
|
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA |
c18cbe5936b8
6941466: Oracle rebranding changes for Hotspot repositories
trims
parents:
775
diff
changeset
|
20 * or visit www.oracle.com if you need additional information or have any |
c18cbe5936b8
6941466: Oracle rebranding changes for Hotspot repositories
trims
parents:
775
diff
changeset
|
21 * questions. |
585 | 22 * |
23 */ | |
24 | |
1972 | 25 #ifndef CPU_X86_VM_VM_VERSION_X86_HPP |
26 #define CPU_X86_VM_VM_VERSION_X86_HPP | |
27 | |
28 #include "runtime/globals_extension.hpp" | |
29 #include "runtime/vm_version.hpp" | |
30 | |
585 | 31 class VM_Version : public Abstract_VM_Version { |
13517
eefb0224149d
derive CPUFeatures from VM_Version::_cpuFeatures
Tom Rodriguez <tom.rodriguez@oracle.com>
parents:
11173
diff
changeset
|
32 friend class VMStructs; |
585 | 33 public: |
34 // cpuid result register layouts. These are all unions of a uint32_t | |
35 // (in case anyone wants access to the register as a whole) and a bitfield. | |
36 | |
37 union StdCpuid1Eax { | |
38 uint32_t value; | |
39 struct { | |
40 uint32_t stepping : 4, | |
41 model : 4, | |
42 family : 4, | |
43 proc_type : 2, | |
44 : 2, | |
45 ext_model : 4, | |
46 ext_family : 8, | |
47 : 4; | |
48 } bits; | |
49 }; | |
50 | |
51 union StdCpuid1Ebx { // example, unused | |
52 uint32_t value; | |
53 struct { | |
54 uint32_t brand_id : 8, | |
55 clflush_size : 8, | |
56 threads_per_cpu : 8, | |
57 apic_id : 8; | |
58 } bits; | |
59 }; | |
60 | |
61 union StdCpuid1Ecx { | |
62 uint32_t value; | |
63 struct { | |
64 uint32_t sse3 : 1, | |
11080
b800986664f4
7088419: Use x86 Hardware CRC32 Instruction with java.util.zip.CRC32
drchase
parents:
7474
diff
changeset
|
65 clmul : 1, |
b800986664f4
7088419: Use x86 Hardware CRC32 Instruction with java.util.zip.CRC32
drchase
parents:
7474
diff
changeset
|
66 : 1, |
585 | 67 monitor : 1, |
68 : 1, | |
69 vmx : 1, | |
70 : 1, | |
71 est : 1, | |
72 : 1, | |
73 ssse3 : 1, | |
74 cid : 1, | |
75 : 2, | |
76 cmpxchg16: 1, | |
77 : 4, | |
78 dca : 1, | |
79 sse4_1 : 1, | |
80 sse4_2 : 1, | |
643
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
585
diff
changeset
|
81 : 2, |
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
585
diff
changeset
|
82 popcnt : 1, |
6894 | 83 : 1, |
84 aes : 1, | |
85 : 1, | |
4759 | 86 osxsave : 1, |
87 avx : 1, | |
88 : 3; | |
585 | 89 } bits; |
90 }; | |
91 | |
92 union StdCpuid1Edx { | |
93 uint32_t value; | |
94 struct { | |
95 uint32_t : 4, | |
96 tsc : 1, | |
97 : 3, | |
98 cmpxchg8 : 1, | |
99 : 6, | |
100 cmov : 1, | |
3787
6ae7a1561b53
6990015: Incorrect Icache line size is used for 64 bit x86
kvn
parents:
2479
diff
changeset
|
101 : 3, |
6ae7a1561b53
6990015: Incorrect Icache line size is used for 64 bit x86
kvn
parents:
2479
diff
changeset
|
102 clflush : 1, |
6ae7a1561b53
6990015: Incorrect Icache line size is used for 64 bit x86
kvn
parents:
2479
diff
changeset
|
103 : 3, |
585 | 104 mmx : 1, |
105 fxsr : 1, | |
106 sse : 1, | |
107 sse2 : 1, | |
108 : 1, | |
109 ht : 1, | |
110 : 3; | |
111 } bits; | |
112 }; | |
113 | |
114 union DcpCpuid4Eax { | |
115 uint32_t value; | |
116 struct { | |
117 uint32_t cache_type : 5, | |
118 : 21, | |
119 cores_per_cpu : 6; | |
120 } bits; | |
121 }; | |
122 | |
123 union DcpCpuid4Ebx { | |
124 uint32_t value; | |
125 struct { | |
126 uint32_t L1_line_size : 12, | |
127 partitions : 10, | |
128 associativity : 10; | |
129 } bits; | |
130 }; | |
131 | |
1622 | 132 union TplCpuidBEbx { |
133 uint32_t value; | |
134 struct { | |
135 uint32_t logical_cpus : 16, | |
136 : 16; | |
137 } bits; | |
138 }; | |
139 | |
585 | 140 union ExtCpuid1Ecx { |
141 uint32_t value; | |
142 struct { | |
143 uint32_t LahfSahf : 1, | |
144 CmpLegacy : 1, | |
14693
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
145 : 3, |
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
146 lzcnt_intel : 1, |
775
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
643
diff
changeset
|
147 lzcnt : 1, |
585 | 148 sse4a : 1, |
149 misalignsse : 1, | |
150 prefetchw : 1, | |
151 : 22; | |
152 } bits; | |
153 }; | |
154 | |
155 union ExtCpuid1Edx { | |
156 uint32_t value; | |
157 struct { | |
158 uint32_t : 22, | |
159 mmx_amd : 1, | |
160 mmx : 1, | |
161 fxsr : 1, | |
162 : 4, | |
163 long_mode : 1, | |
164 tdnow2 : 1, | |
165 tdnow : 1; | |
166 } bits; | |
167 }; | |
168 | |
169 union ExtCpuid5Ex { | |
170 uint32_t value; | |
171 struct { | |
172 uint32_t L1_line_size : 8, | |
173 L1_tag_lines : 8, | |
174 L1_assoc : 8, | |
175 L1_size : 8; | |
176 } bits; | |
177 }; | |
178 | |
4771 | 179 union ExtCpuid7Edx { |
180 uint32_t value; | |
181 struct { | |
182 uint32_t : 8, | |
183 tsc_invariance : 1, | |
184 : 23; | |
185 } bits; | |
186 }; | |
187 | |
585 | 188 union ExtCpuid8Ecx { |
189 uint32_t value; | |
190 struct { | |
191 uint32_t cores_per_cpu : 8, | |
192 : 24; | |
193 } bits; | |
194 }; | |
195 | |
4759 | 196 union SefCpuid7Eax { |
197 uint32_t value; | |
198 }; | |
199 | |
200 union SefCpuid7Ebx { | |
201 uint32_t value; | |
202 struct { | |
203 uint32_t fsgsbase : 1, | |
204 : 2, | |
205 bmi1 : 1, | |
206 : 1, | |
207 avx2 : 1, | |
208 : 2, | |
209 bmi2 : 1, | |
7474
00af3a3a8df4
8005522: use fast-string instructions on x86 for zeroing
kvn
parents:
6894
diff
changeset
|
210 erms : 1, |
17780 | 211 : 1, |
212 rtm : 1, | |
213 : 20; | |
4759 | 214 } bits; |
215 }; | |
216 | |
217 union XemXcr0Eax { | |
218 uint32_t value; | |
219 struct { | |
220 uint32_t x87 : 1, | |
221 sse : 1, | |
222 ymm : 1, | |
223 : 29; | |
224 } bits; | |
225 }; | |
226 | |
585 | 227 protected: |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
228 static int _cpu; |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
229 static int _model; |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
230 static int _stepping; |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
231 static int _cpuFeatures; // features returned by the "cpuid" instruction |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
232 // 0 if this instruction is not available |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
233 static const char* _features_str; |
585 | 234 |
17739
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
235 static address _cpuinfo_segv_addr; // address of instruction which causes SEGV |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
236 static address _cpuinfo_cont_addr; // address of instruction after the one which causes SEGV |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
237 |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
238 enum { |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
239 CPU_CX8 = (1 << 0), // next bits are from cpuid 1 (EDX) |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
240 CPU_CMOV = (1 << 1), |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
241 CPU_FXSR = (1 << 2), |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
242 CPU_HT = (1 << 3), |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
243 CPU_MMX = (1 << 4), |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
244 CPU_3DNOW_PREFETCH = (1 << 5), // Processor supports 3dnow prefetch and prefetchw instructions |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
245 // may not necessarily support other 3dnow instructions |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
246 CPU_SSE = (1 << 6), |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
247 CPU_SSE2 = (1 << 7), |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
248 CPU_SSE3 = (1 << 8), // SSE3 comes from cpuid 1 (ECX) |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
249 CPU_SSSE3 = (1 << 9), |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
250 CPU_SSE4A = (1 << 10), |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
251 CPU_SSE4_1 = (1 << 11), |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
252 CPU_SSE4_2 = (1 << 12), |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
253 CPU_POPCNT = (1 << 13), |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
254 CPU_LZCNT = (1 << 14), |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
255 CPU_TSC = (1 << 15), |
4771 | 256 CPU_TSCINV = (1 << 16), |
257 CPU_AVX = (1 << 17), | |
6894 | 258 CPU_AVX2 = (1 << 18), |
7474
00af3a3a8df4
8005522: use fast-string instructions on x86 for zeroing
kvn
parents:
6894
diff
changeset
|
259 CPU_AES = (1 << 19), |
11080
b800986664f4
7088419: Use x86 Hardware CRC32 Instruction with java.util.zip.CRC32
drchase
parents:
7474
diff
changeset
|
260 CPU_ERMS = (1 << 20), // enhanced 'rep movsb/stosb' instructions |
14693
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
261 CPU_CLMUL = (1 << 21), // carryless multiply for CRC |
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
262 CPU_BMI1 = (1 << 22), |
17780 | 263 CPU_BMI2 = (1 << 23), |
264 CPU_RTM = (1 << 24) // Restricted Transactional Memory instructions | |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
265 } cpuFeatureFlags; |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
266 |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
267 enum { |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
268 // AMD |
4933
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
269 CPU_FAMILY_AMD_11H = 0x11, |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
270 // Intel |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
271 CPU_FAMILY_INTEL_CORE = 6, |
4933
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
272 CPU_MODEL_NEHALEM = 0x1e, |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
273 CPU_MODEL_NEHALEM_EP = 0x1a, |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
274 CPU_MODEL_NEHALEM_EX = 0x2e, |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
275 CPU_MODEL_WESTMERE = 0x25, |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
276 CPU_MODEL_WESTMERE_EP = 0x2c, |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
277 CPU_MODEL_WESTMERE_EX = 0x2f, |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
278 CPU_MODEL_SANDYBRIDGE = 0x2a, |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
279 CPU_MODEL_SANDYBRIDGE_EP = 0x2d, |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
280 CPU_MODEL_IVYBRIDGE_EP = 0x3a |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
281 } cpuExtendedFamily; |
585 | 282 |
283 // cpuid information block. All info derived from executing cpuid with | |
284 // various function numbers is stored here. Intel and AMD info is | |
285 // merged in this block: accessor methods disentangle it. | |
286 // | |
287 // The info block is laid out in subblocks of 4 dwords corresponding to | |
288 // eax, ebx, ecx and edx, whether or not they contain anything useful. | |
289 struct CpuidInfo { | |
290 // cpuid function 0 | |
291 uint32_t std_max_function; | |
292 uint32_t std_vendor_name_0; | |
293 uint32_t std_vendor_name_1; | |
294 uint32_t std_vendor_name_2; | |
295 | |
296 // cpuid function 1 | |
297 StdCpuid1Eax std_cpuid1_eax; | |
298 StdCpuid1Ebx std_cpuid1_ebx; | |
299 StdCpuid1Ecx std_cpuid1_ecx; | |
300 StdCpuid1Edx std_cpuid1_edx; | |
301 | |
302 // cpuid function 4 (deterministic cache parameters) | |
303 DcpCpuid4Eax dcp_cpuid4_eax; | |
304 DcpCpuid4Ebx dcp_cpuid4_ebx; | |
305 uint32_t dcp_cpuid4_ecx; // unused currently | |
306 uint32_t dcp_cpuid4_edx; // unused currently | |
307 | |
4759 | 308 // cpuid function 7 (structured extended features) |
309 SefCpuid7Eax sef_cpuid7_eax; | |
310 SefCpuid7Ebx sef_cpuid7_ebx; | |
311 uint32_t sef_cpuid7_ecx; // unused currently | |
312 uint32_t sef_cpuid7_edx; // unused currently | |
313 | |
1622 | 314 // cpuid function 0xB (processor topology) |
315 // ecx = 0 | |
316 uint32_t tpl_cpuidB0_eax; | |
317 TplCpuidBEbx tpl_cpuidB0_ebx; | |
318 uint32_t tpl_cpuidB0_ecx; // unused currently | |
319 uint32_t tpl_cpuidB0_edx; // unused currently | |
320 | |
321 // ecx = 1 | |
322 uint32_t tpl_cpuidB1_eax; | |
323 TplCpuidBEbx tpl_cpuidB1_ebx; | |
324 uint32_t tpl_cpuidB1_ecx; // unused currently | |
325 uint32_t tpl_cpuidB1_edx; // unused currently | |
326 | |
327 // ecx = 2 | |
328 uint32_t tpl_cpuidB2_eax; | |
329 TplCpuidBEbx tpl_cpuidB2_ebx; | |
330 uint32_t tpl_cpuidB2_ecx; // unused currently | |
331 uint32_t tpl_cpuidB2_edx; // unused currently | |
332 | |
585 | 333 // cpuid function 0x80000000 // example, unused |
334 uint32_t ext_max_function; | |
335 uint32_t ext_vendor_name_0; | |
336 uint32_t ext_vendor_name_1; | |
337 uint32_t ext_vendor_name_2; | |
338 | |
339 // cpuid function 0x80000001 | |
340 uint32_t ext_cpuid1_eax; // reserved | |
341 uint32_t ext_cpuid1_ebx; // reserved | |
342 ExtCpuid1Ecx ext_cpuid1_ecx; | |
343 ExtCpuid1Edx ext_cpuid1_edx; | |
344 | |
345 // cpuid functions 0x80000002 thru 0x80000004: example, unused | |
346 uint32_t proc_name_0, proc_name_1, proc_name_2, proc_name_3; | |
347 uint32_t proc_name_4, proc_name_5, proc_name_6, proc_name_7; | |
348 uint32_t proc_name_8, proc_name_9, proc_name_10,proc_name_11; | |
349 | |
4933
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
350 // cpuid function 0x80000005 // AMD L1, Intel reserved |
585 | 351 uint32_t ext_cpuid5_eax; // unused currently |
352 uint32_t ext_cpuid5_ebx; // reserved | |
353 ExtCpuid5Ex ext_cpuid5_ecx; // L1 data cache info (AMD) | |
354 ExtCpuid5Ex ext_cpuid5_edx; // L1 instruction cache info (AMD) | |
355 | |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
356 // cpuid function 0x80000007 |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
357 uint32_t ext_cpuid7_eax; // reserved |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
358 uint32_t ext_cpuid7_ebx; // reserved |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
359 uint32_t ext_cpuid7_ecx; // reserved |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
360 ExtCpuid7Edx ext_cpuid7_edx; // tscinv |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
361 |
585 | 362 // cpuid function 0x80000008 |
363 uint32_t ext_cpuid8_eax; // unused currently | |
364 uint32_t ext_cpuid8_ebx; // reserved | |
365 ExtCpuid8Ecx ext_cpuid8_ecx; | |
366 uint32_t ext_cpuid8_edx; // reserved | |
4759 | 367 |
368 // extended control register XCR0 (the XFEATURE_ENABLED_MASK register) | |
369 XemXcr0Eax xem_xcr0_eax; | |
370 uint32_t xem_xcr0_edx; // reserved | |
17739
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
371 |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
372 // Space to save ymm registers after signal handle |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
373 int ymm_save[8*4]; // Save ymm0, ymm7, ymm8, ymm15 |
585 | 374 }; |
375 | |
376 // The actual cpuid info block | |
377 static CpuidInfo _cpuid_info; | |
378 | |
379 // Extractors and predicates | |
380 static uint32_t extended_cpu_family() { | |
381 uint32_t result = _cpuid_info.std_cpuid1_eax.bits.family; | |
382 result += _cpuid_info.std_cpuid1_eax.bits.ext_family; | |
383 return result; | |
384 } | |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
385 |
585 | 386 static uint32_t extended_cpu_model() { |
387 uint32_t result = _cpuid_info.std_cpuid1_eax.bits.model; | |
388 result |= _cpuid_info.std_cpuid1_eax.bits.ext_model << 4; | |
389 return result; | |
390 } | |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
391 |
585 | 392 static uint32_t cpu_stepping() { |
393 uint32_t result = _cpuid_info.std_cpuid1_eax.bits.stepping; | |
394 return result; | |
395 } | |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
396 |
585 | 397 static uint logical_processor_count() { |
398 uint result = threads_per_core(); | |
399 return result; | |
400 } | |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
401 |
585 | 402 static uint32_t feature_flags() { |
403 uint32_t result = 0; | |
404 if (_cpuid_info.std_cpuid1_edx.bits.cmpxchg8 != 0) | |
405 result |= CPU_CX8; | |
406 if (_cpuid_info.std_cpuid1_edx.bits.cmov != 0) | |
407 result |= CPU_CMOV; | |
1789
a83b0246bb77
6934483: GCC 4.5 errors "suggest parentheses around something..." when compiling with -Werror and -Wall
twisti
parents:
1647
diff
changeset
|
408 if (_cpuid_info.std_cpuid1_edx.bits.fxsr != 0 || (is_amd() && |
a83b0246bb77
6934483: GCC 4.5 errors "suggest parentheses around something..." when compiling with -Werror and -Wall
twisti
parents:
1647
diff
changeset
|
409 _cpuid_info.ext_cpuid1_edx.bits.fxsr != 0)) |
585 | 410 result |= CPU_FXSR; |
411 // HT flag is set for multi-core processors also. | |
412 if (threads_per_core() > 1) | |
413 result |= CPU_HT; | |
1789
a83b0246bb77
6934483: GCC 4.5 errors "suggest parentheses around something..." when compiling with -Werror and -Wall
twisti
parents:
1647
diff
changeset
|
414 if (_cpuid_info.std_cpuid1_edx.bits.mmx != 0 || (is_amd() && |
a83b0246bb77
6934483: GCC 4.5 errors "suggest parentheses around something..." when compiling with -Werror and -Wall
twisti
parents:
1647
diff
changeset
|
415 _cpuid_info.ext_cpuid1_edx.bits.mmx != 0)) |
585 | 416 result |= CPU_MMX; |
417 if (_cpuid_info.std_cpuid1_edx.bits.sse != 0) | |
418 result |= CPU_SSE; | |
419 if (_cpuid_info.std_cpuid1_edx.bits.sse2 != 0) | |
420 result |= CPU_SSE2; | |
421 if (_cpuid_info.std_cpuid1_ecx.bits.sse3 != 0) | |
422 result |= CPU_SSE3; | |
423 if (_cpuid_info.std_cpuid1_ecx.bits.ssse3 != 0) | |
424 result |= CPU_SSSE3; | |
425 if (_cpuid_info.std_cpuid1_ecx.bits.sse4_1 != 0) | |
426 result |= CPU_SSE4_1; | |
427 if (_cpuid_info.std_cpuid1_ecx.bits.sse4_2 != 0) | |
428 result |= CPU_SSE4_2; | |
643
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
585
diff
changeset
|
429 if (_cpuid_info.std_cpuid1_ecx.bits.popcnt != 0) |
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
585
diff
changeset
|
430 result |= CPU_POPCNT; |
4759 | 431 if (_cpuid_info.std_cpuid1_ecx.bits.avx != 0 && |
432 _cpuid_info.std_cpuid1_ecx.bits.osxsave != 0 && | |
433 _cpuid_info.xem_xcr0_eax.bits.sse != 0 && | |
434 _cpuid_info.xem_xcr0_eax.bits.ymm != 0) { | |
435 result |= CPU_AVX; | |
436 if (_cpuid_info.sef_cpuid7_ebx.bits.avx2 != 0) | |
437 result |= CPU_AVX2; | |
438 } | |
14693
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
439 if(_cpuid_info.sef_cpuid7_ebx.bits.bmi1 != 0) |
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
440 result |= CPU_BMI1; |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
441 if (_cpuid_info.std_cpuid1_edx.bits.tsc != 0) |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
442 result |= CPU_TSC; |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
443 if (_cpuid_info.ext_cpuid7_edx.bits.tsc_invariance != 0) |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
444 result |= CPU_TSCINV; |
6894 | 445 if (_cpuid_info.std_cpuid1_ecx.bits.aes != 0) |
446 result |= CPU_AES; | |
7474
00af3a3a8df4
8005522: use fast-string instructions on x86 for zeroing
kvn
parents:
6894
diff
changeset
|
447 if (_cpuid_info.sef_cpuid7_ebx.bits.erms != 0) |
00af3a3a8df4
8005522: use fast-string instructions on x86 for zeroing
kvn
parents:
6894
diff
changeset
|
448 result |= CPU_ERMS; |
11080
b800986664f4
7088419: Use x86 Hardware CRC32 Instruction with java.util.zip.CRC32
drchase
parents:
7474
diff
changeset
|
449 if (_cpuid_info.std_cpuid1_ecx.bits.clmul != 0) |
b800986664f4
7088419: Use x86 Hardware CRC32 Instruction with java.util.zip.CRC32
drchase
parents:
7474
diff
changeset
|
450 result |= CPU_CLMUL; |
17780 | 451 if (_cpuid_info.sef_cpuid7_ebx.bits.rtm != 0) |
452 result |= CPU_RTM; | |
775
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
643
diff
changeset
|
453 |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
643
diff
changeset
|
454 // AMD features. |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
643
diff
changeset
|
455 if (is_amd()) { |
2479 | 456 if ((_cpuid_info.ext_cpuid1_edx.bits.tdnow != 0) || |
457 (_cpuid_info.ext_cpuid1_ecx.bits.prefetchw != 0)) | |
458 result |= CPU_3DNOW_PREFETCH; | |
775
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
643
diff
changeset
|
459 if (_cpuid_info.ext_cpuid1_ecx.bits.lzcnt != 0) |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
643
diff
changeset
|
460 result |= CPU_LZCNT; |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
643
diff
changeset
|
461 if (_cpuid_info.ext_cpuid1_ecx.bits.sse4a != 0) |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
643
diff
changeset
|
462 result |= CPU_SSE4A; |
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
643
diff
changeset
|
463 } |
14693
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
464 // Intel features. |
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
465 if(is_intel()) { |
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
466 if(_cpuid_info.sef_cpuid7_ebx.bits.bmi2 != 0) |
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
467 result |= CPU_BMI2; |
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
468 if(_cpuid_info.ext_cpuid1_ecx.bits.lzcnt_intel != 0) |
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
469 result |= CPU_LZCNT; |
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
470 } |
775
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
643
diff
changeset
|
471 |
585 | 472 return result; |
473 } | |
474 | |
17739
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
475 static bool os_supports_avx_vectors() { |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
476 if (!supports_avx()) { |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
477 return false; |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
478 } |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
479 // Verify that OS save/restore all bits of AVX registers |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
480 // during signal processing. |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
481 int nreg = 2 LP64_ONLY(+2); |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
482 for (int i = 0; i < 8 * nreg; i++) { // 32 bytes per ymm register |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
483 if (_cpuid_info.ymm_save[i] != ymm_test_value()) { |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
484 return false; |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
485 } |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
486 } |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
487 return true; |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
488 } |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
489 |
585 | 490 static void get_processor_features(); |
491 | |
492 public: | |
493 // Offsets for cpuid asm stub | |
494 static ByteSize std_cpuid0_offset() { return byte_offset_of(CpuidInfo, std_max_function); } | |
495 static ByteSize std_cpuid1_offset() { return byte_offset_of(CpuidInfo, std_cpuid1_eax); } | |
496 static ByteSize dcp_cpuid4_offset() { return byte_offset_of(CpuidInfo, dcp_cpuid4_eax); } | |
4759 | 497 static ByteSize sef_cpuid7_offset() { return byte_offset_of(CpuidInfo, sef_cpuid7_eax); } |
585 | 498 static ByteSize ext_cpuid1_offset() { return byte_offset_of(CpuidInfo, ext_cpuid1_eax); } |
499 static ByteSize ext_cpuid5_offset() { return byte_offset_of(CpuidInfo, ext_cpuid5_eax); } | |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
500 static ByteSize ext_cpuid7_offset() { return byte_offset_of(CpuidInfo, ext_cpuid7_eax); } |
585 | 501 static ByteSize ext_cpuid8_offset() { return byte_offset_of(CpuidInfo, ext_cpuid8_eax); } |
1622 | 502 static ByteSize tpl_cpuidB0_offset() { return byte_offset_of(CpuidInfo, tpl_cpuidB0_eax); } |
503 static ByteSize tpl_cpuidB1_offset() { return byte_offset_of(CpuidInfo, tpl_cpuidB1_eax); } | |
504 static ByteSize tpl_cpuidB2_offset() { return byte_offset_of(CpuidInfo, tpl_cpuidB2_eax); } | |
4759 | 505 static ByteSize xem_xcr0_offset() { return byte_offset_of(CpuidInfo, xem_xcr0_eax); } |
17739
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
506 static ByteSize ymm_save_offset() { return byte_offset_of(CpuidInfo, ymm_save); } |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
507 |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
508 // The value used to check ymm register after signal handle |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
509 static int ymm_test_value() { return 0xCAFEBABE; } |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
510 |
17829
0118c8c7b80f
8038633: crash in VM_Version::get_processor_features() on startup
kvn
parents:
17780
diff
changeset
|
511 static void get_cpu_info_wrapper(); |
17739
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
512 static void set_cpuinfo_segv_addr(address pc) { _cpuinfo_segv_addr = pc; } |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
513 static bool is_cpuinfo_segv_addr(address pc) { return _cpuinfo_segv_addr == pc; } |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
514 static void set_cpuinfo_cont_addr(address pc) { _cpuinfo_cont_addr = pc; } |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
515 static address cpuinfo_cont_addr() { return _cpuinfo_cont_addr; } |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
516 |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
517 static void clean_cpuFeatures() { _cpuFeatures = 0; } |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
518 static void set_avx_cpuFeatures() { _cpuFeatures = (CPU_SSE | CPU_SSE2 | CPU_AVX); } |
98af1e198e73
8037226: compiler/7196199/Test7196199.java fails on 32-bit linux with MaxVectorSize > 16
kvn
parents:
14693
diff
changeset
|
519 |
585 | 520 |
521 // Initialization | |
522 static void initialize(); | |
523 | |
17780 | 524 // Override Abstract_VM_Version implementation |
525 static bool use_biased_locking(); | |
526 | |
585 | 527 // Asserts |
528 static void assert_is_initialized() { | |
529 assert(_cpuid_info.std_cpuid1_eax.bits.family != 0, "VM_Version not initialized"); | |
530 } | |
531 | |
532 // | |
533 // Processor family: | |
534 // 3 - 386 | |
535 // 4 - 486 | |
536 // 5 - Pentium | |
537 // 6 - PentiumPro, Pentium II, Celeron, Xeon, Pentium III, Athlon, | |
538 // Pentium M, Core Solo, Core Duo, Core2 Duo | |
539 // family 6 model: 9, 13, 14, 15 | |
540 // 0x0f - Pentium 4, Opteron | |
541 // | |
542 // Note: The cpu family should be used to select between | |
543 // instruction sequences which are valid on all Intel | |
544 // processors. Use the feature test functions below to | |
545 // determine whether a particular instruction is supported. | |
546 // | |
547 static int cpu_family() { return _cpu;} | |
548 static bool is_P6() { return cpu_family() >= 6; } | |
549 static bool is_amd() { assert_is_initialized(); return _cpuid_info.std_vendor_name_0 == 0x68747541; } // 'htuA' | |
550 static bool is_intel() { assert_is_initialized(); return _cpuid_info.std_vendor_name_0 == 0x756e6547; } // 'uneG' | |
551 | |
1647 | 552 static bool supports_processor_topology() { |
553 return (_cpuid_info.std_max_function >= 0xB) && | |
554 // eax[4:0] | ebx[0:15] == 0 indicates invalid topology level. | |
555 // Some cpus have max cpuid >= 0xB but do not support processor topology. | |
7474
00af3a3a8df4
8005522: use fast-string instructions on x86 for zeroing
kvn
parents:
6894
diff
changeset
|
556 (((_cpuid_info.tpl_cpuidB0_eax & 0x1f) | _cpuid_info.tpl_cpuidB0_ebx.bits.logical_cpus) != 0); |
1647 | 557 } |
558 | |
585 | 559 static uint cores_per_cpu() { |
560 uint result = 1; | |
561 if (is_intel()) { | |
1647 | 562 if (supports_processor_topology()) { |
1622 | 563 result = _cpuid_info.tpl_cpuidB1_ebx.bits.logical_cpus / |
564 _cpuid_info.tpl_cpuidB0_ebx.bits.logical_cpus; | |
565 } else { | |
566 result = (_cpuid_info.dcp_cpuid4_eax.bits.cores_per_cpu + 1); | |
567 } | |
585 | 568 } else if (is_amd()) { |
569 result = (_cpuid_info.ext_cpuid8_ecx.bits.cores_per_cpu + 1); | |
570 } | |
571 return result; | |
572 } | |
573 | |
574 static uint threads_per_core() { | |
575 uint result = 1; | |
1647 | 576 if (is_intel() && supports_processor_topology()) { |
1622 | 577 result = _cpuid_info.tpl_cpuidB0_ebx.bits.logical_cpus; |
578 } else if (_cpuid_info.std_cpuid1_edx.bits.ht != 0) { | |
585 | 579 result = _cpuid_info.std_cpuid1_ebx.bits.threads_per_cpu / |
580 cores_per_cpu(); | |
581 } | |
582 return result; | |
583 } | |
584 | |
3854 | 585 static intx prefetch_data_size() { |
585 | 586 intx result = 0; |
587 if (is_intel()) { | |
588 result = (_cpuid_info.dcp_cpuid4_ebx.bits.L1_line_size + 1); | |
589 } else if (is_amd()) { | |
590 result = _cpuid_info.ext_cpuid5_ecx.bits.L1_line_size; | |
591 } | |
592 if (result < 32) // not defined ? | |
593 result = 32; // 32 bytes by default on x86 and other x64 | |
594 return result; | |
595 } | |
596 | |
597 // | |
598 // Feature identification | |
599 // | |
600 static bool supports_cpuid() { return _cpuFeatures != 0; } | |
601 static bool supports_cmpxchg8() { return (_cpuFeatures & CPU_CX8) != 0; } | |
602 static bool supports_cmov() { return (_cpuFeatures & CPU_CMOV) != 0; } | |
603 static bool supports_fxsr() { return (_cpuFeatures & CPU_FXSR) != 0; } | |
604 static bool supports_ht() { return (_cpuFeatures & CPU_HT) != 0; } | |
605 static bool supports_mmx() { return (_cpuFeatures & CPU_MMX) != 0; } | |
606 static bool supports_sse() { return (_cpuFeatures & CPU_SSE) != 0; } | |
607 static bool supports_sse2() { return (_cpuFeatures & CPU_SSE2) != 0; } | |
608 static bool supports_sse3() { return (_cpuFeatures & CPU_SSE3) != 0; } | |
609 static bool supports_ssse3() { return (_cpuFeatures & CPU_SSSE3)!= 0; } | |
610 static bool supports_sse4_1() { return (_cpuFeatures & CPU_SSE4_1) != 0; } | |
611 static bool supports_sse4_2() { return (_cpuFeatures & CPU_SSE4_2) != 0; } | |
643
c771b7f43bbf
6378821: bitCount() should use POPC on SPARC processors and AMD+10h
twisti
parents:
585
diff
changeset
|
612 static bool supports_popcnt() { return (_cpuFeatures & CPU_POPCNT) != 0; } |
4759 | 613 static bool supports_avx() { return (_cpuFeatures & CPU_AVX) != 0; } |
614 static bool supports_avx2() { return (_cpuFeatures & CPU_AVX2) != 0; } | |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
615 static bool supports_tsc() { return (_cpuFeatures & CPU_TSC) != 0; } |
6894 | 616 static bool supports_aes() { return (_cpuFeatures & CPU_AES) != 0; } |
7474
00af3a3a8df4
8005522: use fast-string instructions on x86 for zeroing
kvn
parents:
6894
diff
changeset
|
617 static bool supports_erms() { return (_cpuFeatures & CPU_ERMS) != 0; } |
11080
b800986664f4
7088419: Use x86 Hardware CRC32 Instruction with java.util.zip.CRC32
drchase
parents:
7474
diff
changeset
|
618 static bool supports_clmul() { return (_cpuFeatures & CPU_CLMUL) != 0; } |
17780 | 619 static bool supports_rtm() { return (_cpuFeatures & CPU_RTM) != 0; } |
14693
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
620 static bool supports_bmi1() { return (_cpuFeatures & CPU_BMI1) != 0; } |
9e9af3aa4278
8031321: Support Intel bit manipulation instructions
iveresov
parents:
11080
diff
changeset
|
621 static bool supports_bmi2() { return (_cpuFeatures & CPU_BMI2) != 0; } |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
622 // Intel features |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
623 static bool is_intel_family_core() { return is_intel() && |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
624 extended_cpu_family() == CPU_FAMILY_INTEL_CORE; } |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
625 |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
626 static bool is_intel_tsc_synched_at_init() { |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
627 if (is_intel_family_core()) { |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
628 uint32_t ext_model = extended_cpu_model(); |
4933
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
629 if (ext_model == CPU_MODEL_NEHALEM_EP || |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
630 ext_model == CPU_MODEL_WESTMERE_EP || |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
631 ext_model == CPU_MODEL_SANDYBRIDGE_EP || |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
632 ext_model == CPU_MODEL_IVYBRIDGE_EP) { |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
633 // <= 2-socket invariant tsc support. EX versions are usually used |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
634 // in > 2-socket systems and likely don't synchronize tscs at |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
635 // initialization. |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
636 // Code that uses tsc values must be prepared for them to arbitrarily |
4a24c4f648bd
7142113: Add Ivy Bridge to the known Intel x86 cpu families
phh
parents:
4771
diff
changeset
|
637 // jump forward or backward. |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
638 return true; |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
639 } |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
640 } |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
641 return false; |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
642 } |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
643 |
585 | 644 // AMD features |
2479 | 645 static bool supports_3dnow_prefetch() { return (_cpuFeatures & CPU_3DNOW_PREFETCH) != 0; } |
585 | 646 static bool supports_mmx_ext() { return is_amd() && _cpuid_info.ext_cpuid1_edx.bits.mmx_amd != 0; } |
775
93c14e5562c4
6823354: Add intrinsics for {Integer,Long}.{numberOfLeadingZeros,numberOfTrailingZeros}()
twisti
parents:
643
diff
changeset
|
647 static bool supports_lzcnt() { return (_cpuFeatures & CPU_LZCNT) != 0; } |
585 | 648 static bool supports_sse4a() { return (_cpuFeatures & CPU_SSE4A) != 0; } |
649 | |
4749
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
650 static bool is_amd_Barcelona() { return is_amd() && |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
651 extended_cpu_family() == CPU_FAMILY_AMD_11H; } |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
652 |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
653 // Intel and AMD newer cores support fast timestamps well |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
654 static bool supports_tscinv_bit() { |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
655 return (_cpuFeatures & CPU_TSCINV) != 0; |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
656 } |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
657 static bool supports_tscinv() { |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
658 return supports_tscinv_bit() && |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
659 ( (is_amd() && !is_amd_Barcelona()) || |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
660 is_intel_tsc_synched_at_init() ); |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
661 } |
7ab5f6318694
7125934: Add a fast unordered timestamp capability to Hotspot on x86/x64
phh
parents:
3865
diff
changeset
|
662 |
1914
ae065c367d93
6987135: Performance regression on Intel platform with 32-bits edition between 6u13 and 6u14.
kvn
parents:
1789
diff
changeset
|
663 // Intel Core and newer cpus have fast IDIV instruction (excluding Atom). |
ae065c367d93
6987135: Performance regression on Intel platform with 32-bits edition between 6u13 and 6u14.
kvn
parents:
1789
diff
changeset
|
664 static bool has_fast_idiv() { return is_intel() && cpu_family() == 6 && |
ae065c367d93
6987135: Performance regression on Intel platform with 32-bits edition between 6u13 and 6u14.
kvn
parents:
1789
diff
changeset
|
665 supports_sse3() && _model != 0x1C; } |
ae065c367d93
6987135: Performance regression on Intel platform with 32-bits edition between 6u13 and 6u14.
kvn
parents:
1789
diff
changeset
|
666 |
585 | 667 static bool supports_compare_and_exchange() { return true; } |
668 | |
669 static const char* cpu_features() { return _features_str; } | |
670 | |
671 static intx allocate_prefetch_distance() { | |
672 // This method should be called before allocate_prefetch_style(). | |
673 // | |
674 // Hardware prefetching (distance/size in bytes): | |
675 // Pentium 3 - 64 / 32 | |
676 // Pentium 4 - 256 / 128 | |
677 // Athlon - 64 / 32 ???? | |
678 // Opteron - 128 / 64 only when 2 sequential cache lines accessed | |
679 // Core - 128 / 64 | |
680 // | |
681 // Software prefetching (distance in bytes / instruction with best score): | |
682 // Pentium 3 - 128 / prefetchnta | |
683 // Pentium 4 - 512 / prefetchnta | |
684 // Athlon - 128 / prefetchnta | |
685 // Opteron - 256 / prefetchnta | |
686 // Core - 256 / prefetchnta | |
687 // It will be used only when AllocatePrefetchStyle > 0 | |
688 | |
689 intx count = AllocatePrefetchDistance; | |
690 if (count < 0) { // default ? | |
691 if (is_amd()) { // AMD | |
692 if (supports_sse2()) | |
693 count = 256; // Opteron | |
694 else | |
695 count = 128; // Athlon | |
696 } else { // Intel | |
697 if (supports_sse2()) | |
698 if (cpu_family() == 6) { | |
699 count = 256; // Pentium M, Core, Core2 | |
700 } else { | |
701 count = 512; // Pentium 4 | |
702 } | |
703 else | |
704 count = 128; // Pentium 3 (and all other old CPUs) | |
705 } | |
706 } | |
707 return count; | |
708 } | |
709 static intx allocate_prefetch_style() { | |
710 assert(AllocatePrefetchStyle >= 0, "AllocatePrefetchStyle should be positive"); | |
711 // Return 0 if AllocatePrefetchDistance was not defined. | |
712 return AllocatePrefetchDistance > 0 ? AllocatePrefetchStyle : 0; | |
713 } | |
714 | |
715 // Prefetch interval for gc copy/scan == 9 dcache lines. Derived from | |
716 // 50-warehouse specjbb runs on a 2-way 1.8ghz opteron using a 4gb heap. | |
717 // Tested intervals from 128 to 2048 in increments of 64 == one cache line. | |
718 // 256 bytes (4 dcache lines) was the nearest runner-up to 576. | |
719 | |
720 // gc copy/scan is disabled if prefetchw isn't supported, because | |
721 // Prefetch::write emits an inlined prefetchw on Linux. | |
722 // Do not use the 3dnow prefetchw instruction. It isn't supported on em64t. | |
723 // The used prefetcht0 instruction works for both amd64 and em64t. | |
724 static intx prefetch_copy_interval_in_bytes() { | |
725 intx interval = PrefetchCopyIntervalInBytes; | |
726 return interval >= 0 ? interval : 576; | |
727 } | |
728 static intx prefetch_scan_interval_in_bytes() { | |
729 intx interval = PrefetchScanIntervalInBytes; | |
730 return interval >= 0 ? interval : 576; | |
731 } | |
732 static intx prefetch_fields_ahead() { | |
733 intx count = PrefetchFieldsAhead; | |
734 return count >= 0 ? count : 1; | |
735 } | |
736 }; | |
1972 | 737 |
738 #endif // CPU_X86_VM_VM_VERSION_X86_HPP |