Thu, 15 Aug 2013 20:04:10 -0400
8003424: Enable Class Data Sharing for CompressedOops
8016729: ObjectAlignmentInBytes=16 now forces the use of heap based compressed oops
8005933: The -Xshare:auto option is ignored for -server
Summary: Move klass metaspace above the heap and support CDS with compressed klass ptrs.
Reviewed-by: coleenp, kvn, mgerdin, tschatzl, stefank
1 /*
2 * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
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 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
25 #ifndef CPU_X86_VM_VM_VERSION_X86_HPP
26 #define CPU_X86_VM_VM_VERSION_X86_HPP
28 #include "runtime/globals_extension.hpp"
29 #include "runtime/vm_version.hpp"
31 class VM_Version : public Abstract_VM_Version {
32 public:
33 // cpuid result register layouts. These are all unions of a uint32_t
34 // (in case anyone wants access to the register as a whole) and a bitfield.
36 union StdCpuid1Eax {
37 uint32_t value;
38 struct {
39 uint32_t stepping : 4,
40 model : 4,
41 family : 4,
42 proc_type : 2,
43 : 2,
44 ext_model : 4,
45 ext_family : 8,
46 : 4;
47 } bits;
48 };
50 union StdCpuid1Ebx { // example, unused
51 uint32_t value;
52 struct {
53 uint32_t brand_id : 8,
54 clflush_size : 8,
55 threads_per_cpu : 8,
56 apic_id : 8;
57 } bits;
58 };
60 union StdCpuid1Ecx {
61 uint32_t value;
62 struct {
63 uint32_t sse3 : 1,
64 clmul : 1,
65 : 1,
66 monitor : 1,
67 : 1,
68 vmx : 1,
69 : 1,
70 est : 1,
71 : 1,
72 ssse3 : 1,
73 cid : 1,
74 : 2,
75 cmpxchg16: 1,
76 : 4,
77 dca : 1,
78 sse4_1 : 1,
79 sse4_2 : 1,
80 : 2,
81 popcnt : 1,
82 : 1,
83 aes : 1,
84 : 1,
85 osxsave : 1,
86 avx : 1,
87 : 3;
88 } bits;
89 };
91 union StdCpuid1Edx {
92 uint32_t value;
93 struct {
94 uint32_t : 4,
95 tsc : 1,
96 : 3,
97 cmpxchg8 : 1,
98 : 6,
99 cmov : 1,
100 : 3,
101 clflush : 1,
102 : 3,
103 mmx : 1,
104 fxsr : 1,
105 sse : 1,
106 sse2 : 1,
107 : 1,
108 ht : 1,
109 : 3;
110 } bits;
111 };
113 union DcpCpuid4Eax {
114 uint32_t value;
115 struct {
116 uint32_t cache_type : 5,
117 : 21,
118 cores_per_cpu : 6;
119 } bits;
120 };
122 union DcpCpuid4Ebx {
123 uint32_t value;
124 struct {
125 uint32_t L1_line_size : 12,
126 partitions : 10,
127 associativity : 10;
128 } bits;
129 };
131 union TplCpuidBEbx {
132 uint32_t value;
133 struct {
134 uint32_t logical_cpus : 16,
135 : 16;
136 } bits;
137 };
139 union ExtCpuid1Ecx {
140 uint32_t value;
141 struct {
142 uint32_t LahfSahf : 1,
143 CmpLegacy : 1,
144 : 4,
145 lzcnt : 1,
146 sse4a : 1,
147 misalignsse : 1,
148 prefetchw : 1,
149 : 22;
150 } bits;
151 };
153 union ExtCpuid1Edx {
154 uint32_t value;
155 struct {
156 uint32_t : 22,
157 mmx_amd : 1,
158 mmx : 1,
159 fxsr : 1,
160 : 4,
161 long_mode : 1,
162 tdnow2 : 1,
163 tdnow : 1;
164 } bits;
165 };
167 union ExtCpuid5Ex {
168 uint32_t value;
169 struct {
170 uint32_t L1_line_size : 8,
171 L1_tag_lines : 8,
172 L1_assoc : 8,
173 L1_size : 8;
174 } bits;
175 };
177 union ExtCpuid7Edx {
178 uint32_t value;
179 struct {
180 uint32_t : 8,
181 tsc_invariance : 1,
182 : 23;
183 } bits;
184 };
186 union ExtCpuid8Ecx {
187 uint32_t value;
188 struct {
189 uint32_t cores_per_cpu : 8,
190 : 24;
191 } bits;
192 };
194 union SefCpuid7Eax {
195 uint32_t value;
196 };
198 union SefCpuid7Ebx {
199 uint32_t value;
200 struct {
201 uint32_t fsgsbase : 1,
202 : 2,
203 bmi1 : 1,
204 : 1,
205 avx2 : 1,
206 : 2,
207 bmi2 : 1,
208 erms : 1,
209 : 22;
210 } bits;
211 };
213 union XemXcr0Eax {
214 uint32_t value;
215 struct {
216 uint32_t x87 : 1,
217 sse : 1,
218 ymm : 1,
219 : 29;
220 } bits;
221 };
223 protected:
224 static int _cpu;
225 static int _model;
226 static int _stepping;
227 static int _cpuFeatures; // features returned by the "cpuid" instruction
228 // 0 if this instruction is not available
229 static const char* _features_str;
231 enum {
232 CPU_CX8 = (1 << 0), // next bits are from cpuid 1 (EDX)
233 CPU_CMOV = (1 << 1),
234 CPU_FXSR = (1 << 2),
235 CPU_HT = (1 << 3),
236 CPU_MMX = (1 << 4),
237 CPU_3DNOW_PREFETCH = (1 << 5), // Processor supports 3dnow prefetch and prefetchw instructions
238 // may not necessarily support other 3dnow instructions
239 CPU_SSE = (1 << 6),
240 CPU_SSE2 = (1 << 7),
241 CPU_SSE3 = (1 << 8), // SSE3 comes from cpuid 1 (ECX)
242 CPU_SSSE3 = (1 << 9),
243 CPU_SSE4A = (1 << 10),
244 CPU_SSE4_1 = (1 << 11),
245 CPU_SSE4_2 = (1 << 12),
246 CPU_POPCNT = (1 << 13),
247 CPU_LZCNT = (1 << 14),
248 CPU_TSC = (1 << 15),
249 CPU_TSCINV = (1 << 16),
250 CPU_AVX = (1 << 17),
251 CPU_AVX2 = (1 << 18),
252 CPU_AES = (1 << 19),
253 CPU_ERMS = (1 << 20), // enhanced 'rep movsb/stosb' instructions
254 CPU_CLMUL = (1 << 21) // carryless multiply for CRC
255 } cpuFeatureFlags;
257 enum {
258 // AMD
259 CPU_FAMILY_AMD_11H = 0x11,
260 // Intel
261 CPU_FAMILY_INTEL_CORE = 6,
262 CPU_MODEL_NEHALEM = 0x1e,
263 CPU_MODEL_NEHALEM_EP = 0x1a,
264 CPU_MODEL_NEHALEM_EX = 0x2e,
265 CPU_MODEL_WESTMERE = 0x25,
266 CPU_MODEL_WESTMERE_EP = 0x2c,
267 CPU_MODEL_WESTMERE_EX = 0x2f,
268 CPU_MODEL_SANDYBRIDGE = 0x2a,
269 CPU_MODEL_SANDYBRIDGE_EP = 0x2d,
270 CPU_MODEL_IVYBRIDGE_EP = 0x3a
271 } cpuExtendedFamily;
273 // cpuid information block. All info derived from executing cpuid with
274 // various function numbers is stored here. Intel and AMD info is
275 // merged in this block: accessor methods disentangle it.
276 //
277 // The info block is laid out in subblocks of 4 dwords corresponding to
278 // eax, ebx, ecx and edx, whether or not they contain anything useful.
279 struct CpuidInfo {
280 // cpuid function 0
281 uint32_t std_max_function;
282 uint32_t std_vendor_name_0;
283 uint32_t std_vendor_name_1;
284 uint32_t std_vendor_name_2;
286 // cpuid function 1
287 StdCpuid1Eax std_cpuid1_eax;
288 StdCpuid1Ebx std_cpuid1_ebx;
289 StdCpuid1Ecx std_cpuid1_ecx;
290 StdCpuid1Edx std_cpuid1_edx;
292 // cpuid function 4 (deterministic cache parameters)
293 DcpCpuid4Eax dcp_cpuid4_eax;
294 DcpCpuid4Ebx dcp_cpuid4_ebx;
295 uint32_t dcp_cpuid4_ecx; // unused currently
296 uint32_t dcp_cpuid4_edx; // unused currently
298 // cpuid function 7 (structured extended features)
299 SefCpuid7Eax sef_cpuid7_eax;
300 SefCpuid7Ebx sef_cpuid7_ebx;
301 uint32_t sef_cpuid7_ecx; // unused currently
302 uint32_t sef_cpuid7_edx; // unused currently
304 // cpuid function 0xB (processor topology)
305 // ecx = 0
306 uint32_t tpl_cpuidB0_eax;
307 TplCpuidBEbx tpl_cpuidB0_ebx;
308 uint32_t tpl_cpuidB0_ecx; // unused currently
309 uint32_t tpl_cpuidB0_edx; // unused currently
311 // ecx = 1
312 uint32_t tpl_cpuidB1_eax;
313 TplCpuidBEbx tpl_cpuidB1_ebx;
314 uint32_t tpl_cpuidB1_ecx; // unused currently
315 uint32_t tpl_cpuidB1_edx; // unused currently
317 // ecx = 2
318 uint32_t tpl_cpuidB2_eax;
319 TplCpuidBEbx tpl_cpuidB2_ebx;
320 uint32_t tpl_cpuidB2_ecx; // unused currently
321 uint32_t tpl_cpuidB2_edx; // unused currently
323 // cpuid function 0x80000000 // example, unused
324 uint32_t ext_max_function;
325 uint32_t ext_vendor_name_0;
326 uint32_t ext_vendor_name_1;
327 uint32_t ext_vendor_name_2;
329 // cpuid function 0x80000001
330 uint32_t ext_cpuid1_eax; // reserved
331 uint32_t ext_cpuid1_ebx; // reserved
332 ExtCpuid1Ecx ext_cpuid1_ecx;
333 ExtCpuid1Edx ext_cpuid1_edx;
335 // cpuid functions 0x80000002 thru 0x80000004: example, unused
336 uint32_t proc_name_0, proc_name_1, proc_name_2, proc_name_3;
337 uint32_t proc_name_4, proc_name_5, proc_name_6, proc_name_7;
338 uint32_t proc_name_8, proc_name_9, proc_name_10,proc_name_11;
340 // cpuid function 0x80000005 // AMD L1, Intel reserved
341 uint32_t ext_cpuid5_eax; // unused currently
342 uint32_t ext_cpuid5_ebx; // reserved
343 ExtCpuid5Ex ext_cpuid5_ecx; // L1 data cache info (AMD)
344 ExtCpuid5Ex ext_cpuid5_edx; // L1 instruction cache info (AMD)
346 // cpuid function 0x80000007
347 uint32_t ext_cpuid7_eax; // reserved
348 uint32_t ext_cpuid7_ebx; // reserved
349 uint32_t ext_cpuid7_ecx; // reserved
350 ExtCpuid7Edx ext_cpuid7_edx; // tscinv
352 // cpuid function 0x80000008
353 uint32_t ext_cpuid8_eax; // unused currently
354 uint32_t ext_cpuid8_ebx; // reserved
355 ExtCpuid8Ecx ext_cpuid8_ecx;
356 uint32_t ext_cpuid8_edx; // reserved
358 // extended control register XCR0 (the XFEATURE_ENABLED_MASK register)
359 XemXcr0Eax xem_xcr0_eax;
360 uint32_t xem_xcr0_edx; // reserved
361 };
363 // The actual cpuid info block
364 static CpuidInfo _cpuid_info;
366 // Extractors and predicates
367 static uint32_t extended_cpu_family() {
368 uint32_t result = _cpuid_info.std_cpuid1_eax.bits.family;
369 result += _cpuid_info.std_cpuid1_eax.bits.ext_family;
370 return result;
371 }
373 static uint32_t extended_cpu_model() {
374 uint32_t result = _cpuid_info.std_cpuid1_eax.bits.model;
375 result |= _cpuid_info.std_cpuid1_eax.bits.ext_model << 4;
376 return result;
377 }
379 static uint32_t cpu_stepping() {
380 uint32_t result = _cpuid_info.std_cpuid1_eax.bits.stepping;
381 return result;
382 }
384 static uint logical_processor_count() {
385 uint result = threads_per_core();
386 return result;
387 }
389 static uint32_t feature_flags() {
390 uint32_t result = 0;
391 if (_cpuid_info.std_cpuid1_edx.bits.cmpxchg8 != 0)
392 result |= CPU_CX8;
393 if (_cpuid_info.std_cpuid1_edx.bits.cmov != 0)
394 result |= CPU_CMOV;
395 if (_cpuid_info.std_cpuid1_edx.bits.fxsr != 0 || (is_amd() &&
396 _cpuid_info.ext_cpuid1_edx.bits.fxsr != 0))
397 result |= CPU_FXSR;
398 // HT flag is set for multi-core processors also.
399 if (threads_per_core() > 1)
400 result |= CPU_HT;
401 if (_cpuid_info.std_cpuid1_edx.bits.mmx != 0 || (is_amd() &&
402 _cpuid_info.ext_cpuid1_edx.bits.mmx != 0))
403 result |= CPU_MMX;
404 if (_cpuid_info.std_cpuid1_edx.bits.sse != 0)
405 result |= CPU_SSE;
406 if (_cpuid_info.std_cpuid1_edx.bits.sse2 != 0)
407 result |= CPU_SSE2;
408 if (_cpuid_info.std_cpuid1_ecx.bits.sse3 != 0)
409 result |= CPU_SSE3;
410 if (_cpuid_info.std_cpuid1_ecx.bits.ssse3 != 0)
411 result |= CPU_SSSE3;
412 if (_cpuid_info.std_cpuid1_ecx.bits.sse4_1 != 0)
413 result |= CPU_SSE4_1;
414 if (_cpuid_info.std_cpuid1_ecx.bits.sse4_2 != 0)
415 result |= CPU_SSE4_2;
416 if (_cpuid_info.std_cpuid1_ecx.bits.popcnt != 0)
417 result |= CPU_POPCNT;
418 if (_cpuid_info.std_cpuid1_ecx.bits.avx != 0 &&
419 _cpuid_info.std_cpuid1_ecx.bits.osxsave != 0 &&
420 _cpuid_info.xem_xcr0_eax.bits.sse != 0 &&
421 _cpuid_info.xem_xcr0_eax.bits.ymm != 0) {
422 result |= CPU_AVX;
423 if (_cpuid_info.sef_cpuid7_ebx.bits.avx2 != 0)
424 result |= CPU_AVX2;
425 }
426 if (_cpuid_info.std_cpuid1_edx.bits.tsc != 0)
427 result |= CPU_TSC;
428 if (_cpuid_info.ext_cpuid7_edx.bits.tsc_invariance != 0)
429 result |= CPU_TSCINV;
430 if (_cpuid_info.std_cpuid1_ecx.bits.aes != 0)
431 result |= CPU_AES;
432 if (_cpuid_info.sef_cpuid7_ebx.bits.erms != 0)
433 result |= CPU_ERMS;
434 if (_cpuid_info.std_cpuid1_ecx.bits.clmul != 0)
435 result |= CPU_CLMUL;
437 // AMD features.
438 if (is_amd()) {
439 if ((_cpuid_info.ext_cpuid1_edx.bits.tdnow != 0) ||
440 (_cpuid_info.ext_cpuid1_ecx.bits.prefetchw != 0))
441 result |= CPU_3DNOW_PREFETCH;
442 if (_cpuid_info.ext_cpuid1_ecx.bits.lzcnt != 0)
443 result |= CPU_LZCNT;
444 if (_cpuid_info.ext_cpuid1_ecx.bits.sse4a != 0)
445 result |= CPU_SSE4A;
446 }
448 return result;
449 }
451 static void get_processor_features();
453 public:
454 // Offsets for cpuid asm stub
455 static ByteSize std_cpuid0_offset() { return byte_offset_of(CpuidInfo, std_max_function); }
456 static ByteSize std_cpuid1_offset() { return byte_offset_of(CpuidInfo, std_cpuid1_eax); }
457 static ByteSize dcp_cpuid4_offset() { return byte_offset_of(CpuidInfo, dcp_cpuid4_eax); }
458 static ByteSize sef_cpuid7_offset() { return byte_offset_of(CpuidInfo, sef_cpuid7_eax); }
459 static ByteSize ext_cpuid1_offset() { return byte_offset_of(CpuidInfo, ext_cpuid1_eax); }
460 static ByteSize ext_cpuid5_offset() { return byte_offset_of(CpuidInfo, ext_cpuid5_eax); }
461 static ByteSize ext_cpuid7_offset() { return byte_offset_of(CpuidInfo, ext_cpuid7_eax); }
462 static ByteSize ext_cpuid8_offset() { return byte_offset_of(CpuidInfo, ext_cpuid8_eax); }
463 static ByteSize tpl_cpuidB0_offset() { return byte_offset_of(CpuidInfo, tpl_cpuidB0_eax); }
464 static ByteSize tpl_cpuidB1_offset() { return byte_offset_of(CpuidInfo, tpl_cpuidB1_eax); }
465 static ByteSize tpl_cpuidB2_offset() { return byte_offset_of(CpuidInfo, tpl_cpuidB2_eax); }
466 static ByteSize xem_xcr0_offset() { return byte_offset_of(CpuidInfo, xem_xcr0_eax); }
468 // Initialization
469 static void initialize();
471 // Asserts
472 static void assert_is_initialized() {
473 assert(_cpuid_info.std_cpuid1_eax.bits.family != 0, "VM_Version not initialized");
474 }
476 //
477 // Processor family:
478 // 3 - 386
479 // 4 - 486
480 // 5 - Pentium
481 // 6 - PentiumPro, Pentium II, Celeron, Xeon, Pentium III, Athlon,
482 // Pentium M, Core Solo, Core Duo, Core2 Duo
483 // family 6 model: 9, 13, 14, 15
484 // 0x0f - Pentium 4, Opteron
485 //
486 // Note: The cpu family should be used to select between
487 // instruction sequences which are valid on all Intel
488 // processors. Use the feature test functions below to
489 // determine whether a particular instruction is supported.
490 //
491 static int cpu_family() { return _cpu;}
492 static bool is_P6() { return cpu_family() >= 6; }
493 static bool is_amd() { assert_is_initialized(); return _cpuid_info.std_vendor_name_0 == 0x68747541; } // 'htuA'
494 static bool is_intel() { assert_is_initialized(); return _cpuid_info.std_vendor_name_0 == 0x756e6547; } // 'uneG'
496 static bool supports_processor_topology() {
497 return (_cpuid_info.std_max_function >= 0xB) &&
498 // eax[4:0] | ebx[0:15] == 0 indicates invalid topology level.
499 // Some cpus have max cpuid >= 0xB but do not support processor topology.
500 (((_cpuid_info.tpl_cpuidB0_eax & 0x1f) | _cpuid_info.tpl_cpuidB0_ebx.bits.logical_cpus) != 0);
501 }
503 static uint cores_per_cpu() {
504 uint result = 1;
505 if (is_intel()) {
506 if (supports_processor_topology()) {
507 result = _cpuid_info.tpl_cpuidB1_ebx.bits.logical_cpus /
508 _cpuid_info.tpl_cpuidB0_ebx.bits.logical_cpus;
509 } else {
510 result = (_cpuid_info.dcp_cpuid4_eax.bits.cores_per_cpu + 1);
511 }
512 } else if (is_amd()) {
513 result = (_cpuid_info.ext_cpuid8_ecx.bits.cores_per_cpu + 1);
514 }
515 return result;
516 }
518 static uint threads_per_core() {
519 uint result = 1;
520 if (is_intel() && supports_processor_topology()) {
521 result = _cpuid_info.tpl_cpuidB0_ebx.bits.logical_cpus;
522 } else if (_cpuid_info.std_cpuid1_edx.bits.ht != 0) {
523 result = _cpuid_info.std_cpuid1_ebx.bits.threads_per_cpu /
524 cores_per_cpu();
525 }
526 return result;
527 }
529 static intx prefetch_data_size() {
530 intx result = 0;
531 if (is_intel()) {
532 result = (_cpuid_info.dcp_cpuid4_ebx.bits.L1_line_size + 1);
533 } else if (is_amd()) {
534 result = _cpuid_info.ext_cpuid5_ecx.bits.L1_line_size;
535 }
536 if (result < 32) // not defined ?
537 result = 32; // 32 bytes by default on x86 and other x64
538 return result;
539 }
541 //
542 // Feature identification
543 //
544 static bool supports_cpuid() { return _cpuFeatures != 0; }
545 static bool supports_cmpxchg8() { return (_cpuFeatures & CPU_CX8) != 0; }
546 static bool supports_cmov() { return (_cpuFeatures & CPU_CMOV) != 0; }
547 static bool supports_fxsr() { return (_cpuFeatures & CPU_FXSR) != 0; }
548 static bool supports_ht() { return (_cpuFeatures & CPU_HT) != 0; }
549 static bool supports_mmx() { return (_cpuFeatures & CPU_MMX) != 0; }
550 static bool supports_sse() { return (_cpuFeatures & CPU_SSE) != 0; }
551 static bool supports_sse2() { return (_cpuFeatures & CPU_SSE2) != 0; }
552 static bool supports_sse3() { return (_cpuFeatures & CPU_SSE3) != 0; }
553 static bool supports_ssse3() { return (_cpuFeatures & CPU_SSSE3)!= 0; }
554 static bool supports_sse4_1() { return (_cpuFeatures & CPU_SSE4_1) != 0; }
555 static bool supports_sse4_2() { return (_cpuFeatures & CPU_SSE4_2) != 0; }
556 static bool supports_popcnt() { return (_cpuFeatures & CPU_POPCNT) != 0; }
557 static bool supports_avx() { return (_cpuFeatures & CPU_AVX) != 0; }
558 static bool supports_avx2() { return (_cpuFeatures & CPU_AVX2) != 0; }
559 static bool supports_tsc() { return (_cpuFeatures & CPU_TSC) != 0; }
560 static bool supports_aes() { return (_cpuFeatures & CPU_AES) != 0; }
561 static bool supports_erms() { return (_cpuFeatures & CPU_ERMS) != 0; }
562 static bool supports_clmul() { return (_cpuFeatures & CPU_CLMUL) != 0; }
564 // Intel features
565 static bool is_intel_family_core() { return is_intel() &&
566 extended_cpu_family() == CPU_FAMILY_INTEL_CORE; }
568 static bool is_intel_tsc_synched_at_init() {
569 if (is_intel_family_core()) {
570 uint32_t ext_model = extended_cpu_model();
571 if (ext_model == CPU_MODEL_NEHALEM_EP ||
572 ext_model == CPU_MODEL_WESTMERE_EP ||
573 ext_model == CPU_MODEL_SANDYBRIDGE_EP ||
574 ext_model == CPU_MODEL_IVYBRIDGE_EP) {
575 // <= 2-socket invariant tsc support. EX versions are usually used
576 // in > 2-socket systems and likely don't synchronize tscs at
577 // initialization.
578 // Code that uses tsc values must be prepared for them to arbitrarily
579 // jump forward or backward.
580 return true;
581 }
582 }
583 return false;
584 }
586 // AMD features
587 static bool supports_3dnow_prefetch() { return (_cpuFeatures & CPU_3DNOW_PREFETCH) != 0; }
588 static bool supports_mmx_ext() { return is_amd() && _cpuid_info.ext_cpuid1_edx.bits.mmx_amd != 0; }
589 static bool supports_lzcnt() { return (_cpuFeatures & CPU_LZCNT) != 0; }
590 static bool supports_sse4a() { return (_cpuFeatures & CPU_SSE4A) != 0; }
592 static bool is_amd_Barcelona() { return is_amd() &&
593 extended_cpu_family() == CPU_FAMILY_AMD_11H; }
595 // Intel and AMD newer cores support fast timestamps well
596 static bool supports_tscinv_bit() {
597 return (_cpuFeatures & CPU_TSCINV) != 0;
598 }
599 static bool supports_tscinv() {
600 return supports_tscinv_bit() &&
601 ( (is_amd() && !is_amd_Barcelona()) ||
602 is_intel_tsc_synched_at_init() );
603 }
605 // Intel Core and newer cpus have fast IDIV instruction (excluding Atom).
606 static bool has_fast_idiv() { return is_intel() && cpu_family() == 6 &&
607 supports_sse3() && _model != 0x1C; }
609 static bool supports_compare_and_exchange() { return true; }
611 static const char* cpu_features() { return _features_str; }
613 static intx allocate_prefetch_distance() {
614 // This method should be called before allocate_prefetch_style().
615 //
616 // Hardware prefetching (distance/size in bytes):
617 // Pentium 3 - 64 / 32
618 // Pentium 4 - 256 / 128
619 // Athlon - 64 / 32 ????
620 // Opteron - 128 / 64 only when 2 sequential cache lines accessed
621 // Core - 128 / 64
622 //
623 // Software prefetching (distance in bytes / instruction with best score):
624 // Pentium 3 - 128 / prefetchnta
625 // Pentium 4 - 512 / prefetchnta
626 // Athlon - 128 / prefetchnta
627 // Opteron - 256 / prefetchnta
628 // Core - 256 / prefetchnta
629 // It will be used only when AllocatePrefetchStyle > 0
631 intx count = AllocatePrefetchDistance;
632 if (count < 0) { // default ?
633 if (is_amd()) { // AMD
634 if (supports_sse2())
635 count = 256; // Opteron
636 else
637 count = 128; // Athlon
638 } else { // Intel
639 if (supports_sse2())
640 if (cpu_family() == 6) {
641 count = 256; // Pentium M, Core, Core2
642 } else {
643 count = 512; // Pentium 4
644 }
645 else
646 count = 128; // Pentium 3 (and all other old CPUs)
647 }
648 }
649 return count;
650 }
651 static intx allocate_prefetch_style() {
652 assert(AllocatePrefetchStyle >= 0, "AllocatePrefetchStyle should be positive");
653 // Return 0 if AllocatePrefetchDistance was not defined.
654 return AllocatePrefetchDistance > 0 ? AllocatePrefetchStyle : 0;
655 }
657 // Prefetch interval for gc copy/scan == 9 dcache lines. Derived from
658 // 50-warehouse specjbb runs on a 2-way 1.8ghz opteron using a 4gb heap.
659 // Tested intervals from 128 to 2048 in increments of 64 == one cache line.
660 // 256 bytes (4 dcache lines) was the nearest runner-up to 576.
662 // gc copy/scan is disabled if prefetchw isn't supported, because
663 // Prefetch::write emits an inlined prefetchw on Linux.
664 // Do not use the 3dnow prefetchw instruction. It isn't supported on em64t.
665 // The used prefetcht0 instruction works for both amd64 and em64t.
666 static intx prefetch_copy_interval_in_bytes() {
667 intx interval = PrefetchCopyIntervalInBytes;
668 return interval >= 0 ? interval : 576;
669 }
670 static intx prefetch_scan_interval_in_bytes() {
671 intx interval = PrefetchScanIntervalInBytes;
672 return interval >= 0 ? interval : 576;
673 }
674 static intx prefetch_fields_ahead() {
675 intx count = PrefetchFieldsAhead;
676 return count >= 0 ? count : 1;
677 }
678 };
680 #endif // CPU_X86_VM_VM_VERSION_X86_HPP