Mercurial > hg > truffle
comparison src/cpu/x86/vm/vm_version_x86.hpp @ 585:22e09c0f4b47
6808589: Merge vm_version_x86_{32,64}.{cpp,hpp}
Summary: There is very much duplicated code in vm_version_x86_{32,64}.{cpp,hpp}. Refactoring these would help maintainability.
Reviewed-by: kvn, never
author | twisti |
---|---|
date | Mon, 23 Feb 2009 12:02:30 -0800 |
parents | |
children | c771b7f43bbf |
comparison
equal
deleted
inserted
replaced
584:49a36a80b0c7 | 585:22e09c0f4b47 |
---|---|
1 /* | |
2 * Copyright 1997-2009 Sun Microsystems, Inc. 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 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, | |
20 * CA 95054 USA or visit www.sun.com if you need additional information or | |
21 * have any questions. | |
22 * | |
23 */ | |
24 | |
25 class VM_Version : public Abstract_VM_Version { | |
26 public: | |
27 // cpuid result register layouts. These are all unions of a uint32_t | |
28 // (in case anyone wants access to the register as a whole) and a bitfield. | |
29 | |
30 union StdCpuid1Eax { | |
31 uint32_t value; | |
32 struct { | |
33 uint32_t stepping : 4, | |
34 model : 4, | |
35 family : 4, | |
36 proc_type : 2, | |
37 : 2, | |
38 ext_model : 4, | |
39 ext_family : 8, | |
40 : 4; | |
41 } bits; | |
42 }; | |
43 | |
44 union StdCpuid1Ebx { // example, unused | |
45 uint32_t value; | |
46 struct { | |
47 uint32_t brand_id : 8, | |
48 clflush_size : 8, | |
49 threads_per_cpu : 8, | |
50 apic_id : 8; | |
51 } bits; | |
52 }; | |
53 | |
54 union StdCpuid1Ecx { | |
55 uint32_t value; | |
56 struct { | |
57 uint32_t sse3 : 1, | |
58 : 2, | |
59 monitor : 1, | |
60 : 1, | |
61 vmx : 1, | |
62 : 1, | |
63 est : 1, | |
64 : 1, | |
65 ssse3 : 1, | |
66 cid : 1, | |
67 : 2, | |
68 cmpxchg16: 1, | |
69 : 4, | |
70 dca : 1, | |
71 sse4_1 : 1, | |
72 sse4_2 : 1, | |
73 : 11; | |
74 } bits; | |
75 }; | |
76 | |
77 union StdCpuid1Edx { | |
78 uint32_t value; | |
79 struct { | |
80 uint32_t : 4, | |
81 tsc : 1, | |
82 : 3, | |
83 cmpxchg8 : 1, | |
84 : 6, | |
85 cmov : 1, | |
86 : 7, | |
87 mmx : 1, | |
88 fxsr : 1, | |
89 sse : 1, | |
90 sse2 : 1, | |
91 : 1, | |
92 ht : 1, | |
93 : 3; | |
94 } bits; | |
95 }; | |
96 | |
97 union DcpCpuid4Eax { | |
98 uint32_t value; | |
99 struct { | |
100 uint32_t cache_type : 5, | |
101 : 21, | |
102 cores_per_cpu : 6; | |
103 } bits; | |
104 }; | |
105 | |
106 union DcpCpuid4Ebx { | |
107 uint32_t value; | |
108 struct { | |
109 uint32_t L1_line_size : 12, | |
110 partitions : 10, | |
111 associativity : 10; | |
112 } bits; | |
113 }; | |
114 | |
115 union ExtCpuid1Ecx { | |
116 uint32_t value; | |
117 struct { | |
118 uint32_t LahfSahf : 1, | |
119 CmpLegacy : 1, | |
120 : 4, | |
121 abm : 1, | |
122 sse4a : 1, | |
123 misalignsse : 1, | |
124 prefetchw : 1, | |
125 : 22; | |
126 } bits; | |
127 }; | |
128 | |
129 union ExtCpuid1Edx { | |
130 uint32_t value; | |
131 struct { | |
132 uint32_t : 22, | |
133 mmx_amd : 1, | |
134 mmx : 1, | |
135 fxsr : 1, | |
136 : 4, | |
137 long_mode : 1, | |
138 tdnow2 : 1, | |
139 tdnow : 1; | |
140 } bits; | |
141 }; | |
142 | |
143 union ExtCpuid5Ex { | |
144 uint32_t value; | |
145 struct { | |
146 uint32_t L1_line_size : 8, | |
147 L1_tag_lines : 8, | |
148 L1_assoc : 8, | |
149 L1_size : 8; | |
150 } bits; | |
151 }; | |
152 | |
153 union ExtCpuid8Ecx { | |
154 uint32_t value; | |
155 struct { | |
156 uint32_t cores_per_cpu : 8, | |
157 : 24; | |
158 } bits; | |
159 }; | |
160 | |
161 protected: | |
162 static int _cpu; | |
163 static int _model; | |
164 static int _stepping; | |
165 static int _cpuFeatures; // features returned by the "cpuid" instruction | |
166 // 0 if this instruction is not available | |
167 static const char* _features_str; | |
168 | |
169 enum { | |
170 CPU_CX8 = (1 << 0), // next bits are from cpuid 1 (EDX) | |
171 CPU_CMOV = (1 << 1), | |
172 CPU_FXSR = (1 << 2), | |
173 CPU_HT = (1 << 3), | |
174 CPU_MMX = (1 << 4), | |
175 CPU_3DNOW = (1 << 5), // 3DNow comes from cpuid 0x80000001 (EDX) | |
176 CPU_SSE = (1 << 6), | |
177 CPU_SSE2 = (1 << 7), | |
178 CPU_SSE3 = (1 << 8), // SSE3 comes from cpuid 1 (ECX) | |
179 CPU_SSSE3 = (1 << 9), | |
180 CPU_SSE4A = (1 << 10), | |
181 CPU_SSE4_1 = (1 << 11), | |
182 CPU_SSE4_2 = (1 << 12) | |
183 } cpuFeatureFlags; | |
184 | |
185 // cpuid information block. All info derived from executing cpuid with | |
186 // various function numbers is stored here. Intel and AMD info is | |
187 // merged in this block: accessor methods disentangle it. | |
188 // | |
189 // The info block is laid out in subblocks of 4 dwords corresponding to | |
190 // eax, ebx, ecx and edx, whether or not they contain anything useful. | |
191 struct CpuidInfo { | |
192 // cpuid function 0 | |
193 uint32_t std_max_function; | |
194 uint32_t std_vendor_name_0; | |
195 uint32_t std_vendor_name_1; | |
196 uint32_t std_vendor_name_2; | |
197 | |
198 // cpuid function 1 | |
199 StdCpuid1Eax std_cpuid1_eax; | |
200 StdCpuid1Ebx std_cpuid1_ebx; | |
201 StdCpuid1Ecx std_cpuid1_ecx; | |
202 StdCpuid1Edx std_cpuid1_edx; | |
203 | |
204 // cpuid function 4 (deterministic cache parameters) | |
205 DcpCpuid4Eax dcp_cpuid4_eax; | |
206 DcpCpuid4Ebx dcp_cpuid4_ebx; | |
207 uint32_t dcp_cpuid4_ecx; // unused currently | |
208 uint32_t dcp_cpuid4_edx; // unused currently | |
209 | |
210 // cpuid function 0x80000000 // example, unused | |
211 uint32_t ext_max_function; | |
212 uint32_t ext_vendor_name_0; | |
213 uint32_t ext_vendor_name_1; | |
214 uint32_t ext_vendor_name_2; | |
215 | |
216 // cpuid function 0x80000001 | |
217 uint32_t ext_cpuid1_eax; // reserved | |
218 uint32_t ext_cpuid1_ebx; // reserved | |
219 ExtCpuid1Ecx ext_cpuid1_ecx; | |
220 ExtCpuid1Edx ext_cpuid1_edx; | |
221 | |
222 // cpuid functions 0x80000002 thru 0x80000004: example, unused | |
223 uint32_t proc_name_0, proc_name_1, proc_name_2, proc_name_3; | |
224 uint32_t proc_name_4, proc_name_5, proc_name_6, proc_name_7; | |
225 uint32_t proc_name_8, proc_name_9, proc_name_10,proc_name_11; | |
226 | |
227 // cpuid function 0x80000005 //AMD L1, Intel reserved | |
228 uint32_t ext_cpuid5_eax; // unused currently | |
229 uint32_t ext_cpuid5_ebx; // reserved | |
230 ExtCpuid5Ex ext_cpuid5_ecx; // L1 data cache info (AMD) | |
231 ExtCpuid5Ex ext_cpuid5_edx; // L1 instruction cache info (AMD) | |
232 | |
233 // cpuid function 0x80000008 | |
234 uint32_t ext_cpuid8_eax; // unused currently | |
235 uint32_t ext_cpuid8_ebx; // reserved | |
236 ExtCpuid8Ecx ext_cpuid8_ecx; | |
237 uint32_t ext_cpuid8_edx; // reserved | |
238 }; | |
239 | |
240 // The actual cpuid info block | |
241 static CpuidInfo _cpuid_info; | |
242 | |
243 // Extractors and predicates | |
244 static uint32_t extended_cpu_family() { | |
245 uint32_t result = _cpuid_info.std_cpuid1_eax.bits.family; | |
246 result += _cpuid_info.std_cpuid1_eax.bits.ext_family; | |
247 return result; | |
248 } | |
249 static uint32_t extended_cpu_model() { | |
250 uint32_t result = _cpuid_info.std_cpuid1_eax.bits.model; | |
251 result |= _cpuid_info.std_cpuid1_eax.bits.ext_model << 4; | |
252 return result; | |
253 } | |
254 static uint32_t cpu_stepping() { | |
255 uint32_t result = _cpuid_info.std_cpuid1_eax.bits.stepping; | |
256 return result; | |
257 } | |
258 static uint logical_processor_count() { | |
259 uint result = threads_per_core(); | |
260 return result; | |
261 } | |
262 static uint32_t feature_flags() { | |
263 uint32_t result = 0; | |
264 if (_cpuid_info.std_cpuid1_edx.bits.cmpxchg8 != 0) | |
265 result |= CPU_CX8; | |
266 if (_cpuid_info.std_cpuid1_edx.bits.cmov != 0) | |
267 result |= CPU_CMOV; | |
268 if (_cpuid_info.std_cpuid1_edx.bits.fxsr != 0 || is_amd() && | |
269 _cpuid_info.ext_cpuid1_edx.bits.fxsr != 0) | |
270 result |= CPU_FXSR; | |
271 // HT flag is set for multi-core processors also. | |
272 if (threads_per_core() > 1) | |
273 result |= CPU_HT; | |
274 if (_cpuid_info.std_cpuid1_edx.bits.mmx != 0 || is_amd() && | |
275 _cpuid_info.ext_cpuid1_edx.bits.mmx != 0) | |
276 result |= CPU_MMX; | |
277 if (is_amd() && _cpuid_info.ext_cpuid1_edx.bits.tdnow != 0) | |
278 result |= CPU_3DNOW; | |
279 if (_cpuid_info.std_cpuid1_edx.bits.sse != 0) | |
280 result |= CPU_SSE; | |
281 if (_cpuid_info.std_cpuid1_edx.bits.sse2 != 0) | |
282 result |= CPU_SSE2; | |
283 if (_cpuid_info.std_cpuid1_ecx.bits.sse3 != 0) | |
284 result |= CPU_SSE3; | |
285 if (_cpuid_info.std_cpuid1_ecx.bits.ssse3 != 0) | |
286 result |= CPU_SSSE3; | |
287 if (is_amd() && _cpuid_info.ext_cpuid1_ecx.bits.sse4a != 0) | |
288 result |= CPU_SSE4A; | |
289 if (_cpuid_info.std_cpuid1_ecx.bits.sse4_1 != 0) | |
290 result |= CPU_SSE4_1; | |
291 if (_cpuid_info.std_cpuid1_ecx.bits.sse4_2 != 0) | |
292 result |= CPU_SSE4_2; | |
293 return result; | |
294 } | |
295 | |
296 static void get_processor_features(); | |
297 | |
298 public: | |
299 // Offsets for cpuid asm stub | |
300 static ByteSize std_cpuid0_offset() { return byte_offset_of(CpuidInfo, std_max_function); } | |
301 static ByteSize std_cpuid1_offset() { return byte_offset_of(CpuidInfo, std_cpuid1_eax); } | |
302 static ByteSize dcp_cpuid4_offset() { return byte_offset_of(CpuidInfo, dcp_cpuid4_eax); } | |
303 static ByteSize ext_cpuid1_offset() { return byte_offset_of(CpuidInfo, ext_cpuid1_eax); } | |
304 static ByteSize ext_cpuid5_offset() { return byte_offset_of(CpuidInfo, ext_cpuid5_eax); } | |
305 static ByteSize ext_cpuid8_offset() { return byte_offset_of(CpuidInfo, ext_cpuid8_eax); } | |
306 | |
307 // Initialization | |
308 static void initialize(); | |
309 | |
310 // Asserts | |
311 static void assert_is_initialized() { | |
312 assert(_cpuid_info.std_cpuid1_eax.bits.family != 0, "VM_Version not initialized"); | |
313 } | |
314 | |
315 // | |
316 // Processor family: | |
317 // 3 - 386 | |
318 // 4 - 486 | |
319 // 5 - Pentium | |
320 // 6 - PentiumPro, Pentium II, Celeron, Xeon, Pentium III, Athlon, | |
321 // Pentium M, Core Solo, Core Duo, Core2 Duo | |
322 // family 6 model: 9, 13, 14, 15 | |
323 // 0x0f - Pentium 4, Opteron | |
324 // | |
325 // Note: The cpu family should be used to select between | |
326 // instruction sequences which are valid on all Intel | |
327 // processors. Use the feature test functions below to | |
328 // determine whether a particular instruction is supported. | |
329 // | |
330 static int cpu_family() { return _cpu;} | |
331 static bool is_P6() { return cpu_family() >= 6; } | |
332 | |
333 static bool is_amd() { assert_is_initialized(); return _cpuid_info.std_vendor_name_0 == 0x68747541; } // 'htuA' | |
334 static bool is_intel() { assert_is_initialized(); return _cpuid_info.std_vendor_name_0 == 0x756e6547; } // 'uneG' | |
335 | |
336 static uint cores_per_cpu() { | |
337 uint result = 1; | |
338 if (is_intel()) { | |
339 result = (_cpuid_info.dcp_cpuid4_eax.bits.cores_per_cpu + 1); | |
340 } else if (is_amd()) { | |
341 result = (_cpuid_info.ext_cpuid8_ecx.bits.cores_per_cpu + 1); | |
342 } | |
343 return result; | |
344 } | |
345 | |
346 static uint threads_per_core() { | |
347 uint result = 1; | |
348 if (_cpuid_info.std_cpuid1_edx.bits.ht != 0) { | |
349 result = _cpuid_info.std_cpuid1_ebx.bits.threads_per_cpu / | |
350 cores_per_cpu(); | |
351 } | |
352 return result; | |
353 } | |
354 | |
355 static intx L1_data_cache_line_size() { | |
356 intx result = 0; | |
357 if (is_intel()) { | |
358 result = (_cpuid_info.dcp_cpuid4_ebx.bits.L1_line_size + 1); | |
359 } else if (is_amd()) { | |
360 result = _cpuid_info.ext_cpuid5_ecx.bits.L1_line_size; | |
361 } | |
362 if (result < 32) // not defined ? | |
363 result = 32; // 32 bytes by default on x86 and other x64 | |
364 return result; | |
365 } | |
366 | |
367 // | |
368 // Feature identification | |
369 // | |
370 static bool supports_cpuid() { return _cpuFeatures != 0; } | |
371 static bool supports_cmpxchg8() { return (_cpuFeatures & CPU_CX8) != 0; } | |
372 static bool supports_cmov() { return (_cpuFeatures & CPU_CMOV) != 0; } | |
373 static bool supports_fxsr() { return (_cpuFeatures & CPU_FXSR) != 0; } | |
374 static bool supports_ht() { return (_cpuFeatures & CPU_HT) != 0; } | |
375 static bool supports_mmx() { return (_cpuFeatures & CPU_MMX) != 0; } | |
376 static bool supports_sse() { return (_cpuFeatures & CPU_SSE) != 0; } | |
377 static bool supports_sse2() { return (_cpuFeatures & CPU_SSE2) != 0; } | |
378 static bool supports_sse3() { return (_cpuFeatures & CPU_SSE3) != 0; } | |
379 static bool supports_ssse3() { return (_cpuFeatures & CPU_SSSE3)!= 0; } | |
380 static bool supports_sse4_1() { return (_cpuFeatures & CPU_SSE4_1) != 0; } | |
381 static bool supports_sse4_2() { return (_cpuFeatures & CPU_SSE4_2) != 0; } | |
382 // | |
383 // AMD features | |
384 // | |
385 static bool supports_3dnow() { return (_cpuFeatures & CPU_3DNOW) != 0; } | |
386 static bool supports_mmx_ext() { return is_amd() && _cpuid_info.ext_cpuid1_edx.bits.mmx_amd != 0; } | |
387 static bool supports_3dnow2() { return is_amd() && _cpuid_info.ext_cpuid1_edx.bits.tdnow2 != 0; } | |
388 static bool supports_sse4a() { return (_cpuFeatures & CPU_SSE4A) != 0; } | |
389 | |
390 static bool supports_compare_and_exchange() { return true; } | |
391 | |
392 static const char* cpu_features() { return _features_str; } | |
393 | |
394 static intx allocate_prefetch_distance() { | |
395 // This method should be called before allocate_prefetch_style(). | |
396 // | |
397 // Hardware prefetching (distance/size in bytes): | |
398 // Pentium 3 - 64 / 32 | |
399 // Pentium 4 - 256 / 128 | |
400 // Athlon - 64 / 32 ???? | |
401 // Opteron - 128 / 64 only when 2 sequential cache lines accessed | |
402 // Core - 128 / 64 | |
403 // | |
404 // Software prefetching (distance in bytes / instruction with best score): | |
405 // Pentium 3 - 128 / prefetchnta | |
406 // Pentium 4 - 512 / prefetchnta | |
407 // Athlon - 128 / prefetchnta | |
408 // Opteron - 256 / prefetchnta | |
409 // Core - 256 / prefetchnta | |
410 // It will be used only when AllocatePrefetchStyle > 0 | |
411 | |
412 intx count = AllocatePrefetchDistance; | |
413 if (count < 0) { // default ? | |
414 if (is_amd()) { // AMD | |
415 if (supports_sse2()) | |
416 count = 256; // Opteron | |
417 else | |
418 count = 128; // Athlon | |
419 } else { // Intel | |
420 if (supports_sse2()) | |
421 if (cpu_family() == 6) { | |
422 count = 256; // Pentium M, Core, Core2 | |
423 } else { | |
424 count = 512; // Pentium 4 | |
425 } | |
426 else | |
427 count = 128; // Pentium 3 (and all other old CPUs) | |
428 } | |
429 } | |
430 return count; | |
431 } | |
432 static intx allocate_prefetch_style() { | |
433 assert(AllocatePrefetchStyle >= 0, "AllocatePrefetchStyle should be positive"); | |
434 // Return 0 if AllocatePrefetchDistance was not defined. | |
435 return AllocatePrefetchDistance > 0 ? AllocatePrefetchStyle : 0; | |
436 } | |
437 | |
438 // Prefetch interval for gc copy/scan == 9 dcache lines. Derived from | |
439 // 50-warehouse specjbb runs on a 2-way 1.8ghz opteron using a 4gb heap. | |
440 // Tested intervals from 128 to 2048 in increments of 64 == one cache line. | |
441 // 256 bytes (4 dcache lines) was the nearest runner-up to 576. | |
442 | |
443 // gc copy/scan is disabled if prefetchw isn't supported, because | |
444 // Prefetch::write emits an inlined prefetchw on Linux. | |
445 // Do not use the 3dnow prefetchw instruction. It isn't supported on em64t. | |
446 // The used prefetcht0 instruction works for both amd64 and em64t. | |
447 static intx prefetch_copy_interval_in_bytes() { | |
448 intx interval = PrefetchCopyIntervalInBytes; | |
449 return interval >= 0 ? interval : 576; | |
450 } | |
451 static intx prefetch_scan_interval_in_bytes() { | |
452 intx interval = PrefetchScanIntervalInBytes; | |
453 return interval >= 0 ? interval : 576; | |
454 } | |
455 static intx prefetch_fields_ahead() { | |
456 intx count = PrefetchFieldsAhead; | |
457 return count >= 0 ? count : 1; | |
458 } | |
459 }; |