-
Notifications
You must be signed in to change notification settings - Fork 128
/
cpuid.go
1551 lines (1427 loc) · 52.1 KB
/
cpuid.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
// Copyright (c) 2015 Klaus Post, released under MIT License. See LICENSE file.
// Package cpuid provides information about the CPU running the current program.
//
// CPU features are detected on startup, and kept for fast access through the life of the application.
// Currently x86 / x64 (AMD64) as well as arm64 is supported.
//
// You can access the CPU information by accessing the shared CPU variable of the cpuid library.
//
// Package home: https://github.com/klauspost/cpuid
package cpuid
import (
"flag"
"fmt"
"math"
"math/bits"
"os"
"runtime"
"strings"
)
// AMD refererence: https://www.amd.com/system/files/TechDocs/25481.pdf
// and Processor Programming Reference (PPR)
// Vendor is a representation of a CPU vendor.
type Vendor int
const (
VendorUnknown Vendor = iota
Intel
AMD
VIA
Transmeta
NSC
KVM // Kernel-based Virtual Machine
MSVM // Microsoft Hyper-V or Windows Virtual PC
VMware
XenHVM
Bhyve
Hygon
SiS
RDC
Ampere
ARM
Broadcom
Cavium
DEC
Fujitsu
Infineon
Motorola
NVIDIA
AMCC
Qualcomm
Marvell
QEMU
QNX
ACRN
SRE
Apple
lastVendor
)
//go:generate stringer -type=FeatureID,Vendor
// FeatureID is the ID of a specific cpu feature.
type FeatureID int
const (
// Keep index -1 as unknown
UNKNOWN = -1
// x86 features
ADX FeatureID = iota // Intel ADX (Multi-Precision Add-Carry Instruction Extensions)
AESNI // Advanced Encryption Standard New Instructions
AMD3DNOW // AMD 3DNOW
AMD3DNOWEXT // AMD 3DNowExt
AMXBF16 // Tile computational operations on BFLOAT16 numbers
AMXFP16 // Tile computational operations on FP16 numbers
AMXINT8 // Tile computational operations on 8-bit integers
AMXTILE // Tile architecture
APX_F // Intel APX
AVX // AVX functions
AVX10 // If set the Intel AVX10 Converged Vector ISA is supported
AVX10_128 // If set indicates that AVX10 128-bit vector support is present
AVX10_256 // If set indicates that AVX10 256-bit vector support is present
AVX10_512 // If set indicates that AVX10 512-bit vector support is present
AVX2 // AVX2 functions
AVX512BF16 // AVX-512 BFLOAT16 Instructions
AVX512BITALG // AVX-512 Bit Algorithms
AVX512BW // AVX-512 Byte and Word Instructions
AVX512CD // AVX-512 Conflict Detection Instructions
AVX512DQ // AVX-512 Doubleword and Quadword Instructions
AVX512ER // AVX-512 Exponential and Reciprocal Instructions
AVX512F // AVX-512 Foundation
AVX512FP16 // AVX-512 FP16 Instructions
AVX512IFMA // AVX-512 Integer Fused Multiply-Add Instructions
AVX512PF // AVX-512 Prefetch Instructions
AVX512VBMI // AVX-512 Vector Bit Manipulation Instructions
AVX512VBMI2 // AVX-512 Vector Bit Manipulation Instructions, Version 2
AVX512VL // AVX-512 Vector Length Extensions
AVX512VNNI // AVX-512 Vector Neural Network Instructions
AVX512VP2INTERSECT // AVX-512 Intersect for D/Q
AVX512VPOPCNTDQ // AVX-512 Vector Population Count Doubleword and Quadword
AVXIFMA // AVX-IFMA instructions
AVXNECONVERT // AVX-NE-CONVERT instructions
AVXSLOW // Indicates the CPU performs 2 128 bit operations instead of one
AVXVNNI // AVX (VEX encoded) VNNI neural network instructions
AVXVNNIINT8 // AVX-VNNI-INT8 instructions
AVXVNNIINT16 // AVX-VNNI-INT16 instructions
BHI_CTRL // Branch History Injection and Intra-mode Branch Target Injection / CVE-2022-0001, CVE-2022-0002 / INTEL-SA-00598
BMI1 // Bit Manipulation Instruction Set 1
BMI2 // Bit Manipulation Instruction Set 2
CETIBT // Intel CET Indirect Branch Tracking
CETSS // Intel CET Shadow Stack
CLDEMOTE // Cache Line Demote
CLMUL // Carry-less Multiplication
CLZERO // CLZERO instruction supported
CMOV // i686 CMOV
CMPCCXADD // CMPCCXADD instructions
CMPSB_SCADBS_SHORT // Fast short CMPSB and SCASB
CMPXCHG8 // CMPXCHG8 instruction
CPBOOST // Core Performance Boost
CPPC // AMD: Collaborative Processor Performance Control
CX16 // CMPXCHG16B Instruction
EFER_LMSLE_UNS // AMD: =Core::X86::Msr::EFER[LMSLE] is not supported, and MBZ
ENQCMD // Enqueue Command
ERMS // Enhanced REP MOVSB/STOSB
F16C // Half-precision floating-point conversion
FLUSH_L1D // Flush L1D cache
FMA3 // Intel FMA 3. Does not imply AVX.
FMA4 // Bulldozer FMA4 functions
FP128 // AMD: When set, the internal FP/SIMD execution datapath is no more than 128-bits wide
FP256 // AMD: When set, the internal FP/SIMD execution datapath is no more than 256-bits wide
FSRM // Fast Short Rep Mov
FXSR // FXSAVE, FXRESTOR instructions, CR4 bit 9
FXSROPT // FXSAVE/FXRSTOR optimizations
GFNI // Galois Field New Instructions. May require other features (AVX, AVX512VL,AVX512F) based on usage.
HLE // Hardware Lock Elision
HRESET // If set CPU supports history reset and the IA32_HRESET_ENABLE MSR
HTT // Hyperthreading (enabled)
HWA // Hardware assert supported. Indicates support for MSRC001_10
HYBRID_CPU // This part has CPUs of more than one type.
HYPERVISOR // This bit has been reserved by Intel & AMD for use by hypervisors
IA32_ARCH_CAP // IA32_ARCH_CAPABILITIES MSR (Intel)
IA32_CORE_CAP // IA32_CORE_CAPABILITIES MSR
IBPB // Indirect Branch Restricted Speculation (IBRS) and Indirect Branch Predictor Barrier (IBPB)
IBPB_BRTYPE // Indicates that MSR 49h (PRED_CMD) bit 0 (IBPB) flushes all branch type predictions from the CPU branch predictor
IBRS // AMD: Indirect Branch Restricted Speculation
IBRS_PREFERRED // AMD: IBRS is preferred over software solution
IBRS_PROVIDES_SMP // AMD: IBRS provides Same Mode Protection
IBS // Instruction Based Sampling (AMD)
IBSBRNTRGT // Instruction Based Sampling Feature (AMD)
IBSFETCHSAM // Instruction Based Sampling Feature (AMD)
IBSFFV // Instruction Based Sampling Feature (AMD)
IBSOPCNT // Instruction Based Sampling Feature (AMD)
IBSOPCNTEXT // Instruction Based Sampling Feature (AMD)
IBSOPSAM // Instruction Based Sampling Feature (AMD)
IBSRDWROPCNT // Instruction Based Sampling Feature (AMD)
IBSRIPINVALIDCHK // Instruction Based Sampling Feature (AMD)
IBS_FETCH_CTLX // AMD: IBS fetch control extended MSR supported
IBS_OPDATA4 // AMD: IBS op data 4 MSR supported
IBS_OPFUSE // AMD: Indicates support for IbsOpFuse
IBS_PREVENTHOST // Disallowing IBS use by the host supported
IBS_ZEN4 // AMD: Fetch and Op IBS support IBS extensions added with Zen4
IDPRED_CTRL // IPRED_DIS
INT_WBINVD // WBINVD/WBNOINVD are interruptible.
INVLPGB // NVLPGB and TLBSYNC instruction supported
KEYLOCKER // Key locker
KEYLOCKERW // Key locker wide
LAHF // LAHF/SAHF in long mode
LAM // If set, CPU supports Linear Address Masking
LBRVIRT // LBR virtualization
LZCNT // LZCNT instruction
MCAOVERFLOW // MCA overflow recovery support.
MCDT_NO // Processor do not exhibit MXCSR Configuration Dependent Timing behavior and do not need to mitigate it.
MCOMMIT // MCOMMIT instruction supported
MD_CLEAR // VERW clears CPU buffers
MMX // standard MMX
MMXEXT // SSE integer functions or AMD MMX ext
MOVBE // MOVBE instruction (big-endian)
MOVDIR64B // Move 64 Bytes as Direct Store
MOVDIRI // Move Doubleword as Direct Store
MOVSB_ZL // Fast Zero-Length MOVSB
MOVU // AMD: MOVU SSE instructions are more efficient and should be preferred to SSE MOVL/MOVH. MOVUPS is more efficient than MOVLPS/MOVHPS. MOVUPD is more efficient than MOVLPD/MOVHPD
MPX // Intel MPX (Memory Protection Extensions)
MSRIRC // Instruction Retired Counter MSR available
MSRLIST // Read/Write List of Model Specific Registers
MSR_PAGEFLUSH // Page Flush MSR available
NRIPS // Indicates support for NRIP save on VMEXIT
NX // NX (No-Execute) bit
OSXSAVE // XSAVE enabled by OS
PCONFIG // PCONFIG for Intel Multi-Key Total Memory Encryption
POPCNT // POPCNT instruction
PPIN // AMD: Protected Processor Inventory Number support. Indicates that Protected Processor Inventory Number (PPIN) capability can be enabled
PREFETCHI // PREFETCHIT0/1 instructions
PSFD // Predictive Store Forward Disable
RDPRU // RDPRU instruction supported
RDRAND // RDRAND instruction is available
RDSEED // RDSEED instruction is available
RDTSCP // RDTSCP Instruction
RRSBA_CTRL // Restricted RSB Alternate
RTM // Restricted Transactional Memory
RTM_ALWAYS_ABORT // Indicates that the loaded microcode is forcing RTM abort.
SBPB // Indicates support for the Selective Branch Predictor Barrier
SERIALIZE // Serialize Instruction Execution
SEV // AMD Secure Encrypted Virtualization supported
SEV_64BIT // AMD SEV guest execution only allowed from a 64-bit host
SEV_ALTERNATIVE // AMD SEV Alternate Injection supported
SEV_DEBUGSWAP // Full debug state swap supported for SEV-ES guests
SEV_ES // AMD SEV Encrypted State supported
SEV_RESTRICTED // AMD SEV Restricted Injection supported
SEV_SNP // AMD SEV Secure Nested Paging supported
SGX // Software Guard Extensions
SGXLC // Software Guard Extensions Launch Control
SHA // Intel SHA Extensions
SME // AMD Secure Memory Encryption supported
SME_COHERENT // AMD Hardware cache coherency across encryption domains enforced
SPEC_CTRL_SSBD // Speculative Store Bypass Disable
SRBDS_CTRL // SRBDS mitigation MSR available
SRSO_MSR_FIX // Indicates that software may use MSR BP_CFG[BpSpecReduce] to mitigate SRSO.
SRSO_NO // Indicates the CPU is not subject to the SRSO vulnerability
SRSO_USER_KERNEL_NO // Indicates the CPU is not subject to the SRSO vulnerability across user/kernel boundaries
SSE // SSE functions
SSE2 // P4 SSE functions
SSE3 // Prescott SSE3 functions
SSE4 // Penryn SSE4.1 functions
SSE42 // Nehalem SSE4.2 functions
SSE4A // AMD Barcelona microarchitecture SSE4a instructions
SSSE3 // Conroe SSSE3 functions
STIBP // Single Thread Indirect Branch Predictors
STIBP_ALWAYSON // AMD: Single Thread Indirect Branch Prediction Mode has Enhanced Performance and may be left Always On
STOSB_SHORT // Fast short STOSB
SUCCOR // Software uncorrectable error containment and recovery capability.
SVM // AMD Secure Virtual Machine
SVMDA // Indicates support for the SVM decode assists.
SVMFBASID // SVM, Indicates that TLB flush events, including CR3 writes and CR4.PGE toggles, flush only the current ASID's TLB entries. Also indicates support for the extended VMCBTLB_Control
SVML // AMD SVM lock. Indicates support for SVM-Lock.
SVMNP // AMD SVM nested paging
SVMPF // SVM pause intercept filter. Indicates support for the pause intercept filter
SVMPFT // SVM PAUSE filter threshold. Indicates support for the PAUSE filter cycle count threshold
SYSCALL // System-Call Extension (SCE): SYSCALL and SYSRET instructions.
SYSEE // SYSENTER and SYSEXIT instructions
TBM // AMD Trailing Bit Manipulation
TDX_GUEST // Intel Trust Domain Extensions Guest
TLB_FLUSH_NESTED // AMD: Flushing includes all the nested translations for guest translations
TME // Intel Total Memory Encryption. The following MSRs are supported: IA32_TME_CAPABILITY, IA32_TME_ACTIVATE, IA32_TME_EXCLUDE_MASK, and IA32_TME_EXCLUDE_BASE.
TOPEXT // TopologyExtensions: topology extensions support. Indicates support for CPUID Fn8000_001D_EAX_x[N:0]-CPUID Fn8000_001E_EDX.
TSCRATEMSR // MSR based TSC rate control. Indicates support for MSR TSC ratio MSRC000_0104
TSXLDTRK // Intel TSX Suspend Load Address Tracking
VAES // Vector AES. AVX(512) versions requires additional checks.
VMCBCLEAN // VMCB clean bits. Indicates support for VMCB clean bits.
VMPL // AMD VM Permission Levels supported
VMSA_REGPROT // AMD VMSA Register Protection supported
VMX // Virtual Machine Extensions
VPCLMULQDQ // Carry-Less Multiplication Quadword. Requires AVX for 3 register versions.
VTE // AMD Virtual Transparent Encryption supported
WAITPKG // TPAUSE, UMONITOR, UMWAIT
WBNOINVD // Write Back and Do Not Invalidate Cache
WRMSRNS // Non-Serializing Write to Model Specific Register
X87 // FPU
XGETBV1 // Supports XGETBV with ECX = 1
XOP // Bulldozer XOP functions
XSAVE // XSAVE, XRESTOR, XSETBV, XGETBV
XSAVEC // Supports XSAVEC and the compacted form of XRSTOR.
XSAVEOPT // XSAVEOPT available
XSAVES // Supports XSAVES/XRSTORS and IA32_XSS
// ARM features:
AESARM // AES instructions
ARMCPUID // Some CPU ID registers readable at user-level
ASIMD // Advanced SIMD
ASIMDDP // SIMD Dot Product
ASIMDHP // Advanced SIMD half-precision floating point
ASIMDRDM // Rounding Double Multiply Accumulate/Subtract (SQRDMLAH/SQRDMLSH)
ATOMICS // Large System Extensions (LSE)
CRC32 // CRC32/CRC32C instructions
DCPOP // Data cache clean to Point of Persistence (DC CVAP)
EVTSTRM // Generic timer
FCMA // Floatin point complex number addition and multiplication
FP // Single-precision and double-precision floating point
FPHP // Half-precision floating point
GPA // Generic Pointer Authentication
JSCVT // Javascript-style double->int convert (FJCVTZS)
LRCPC // Weaker release consistency (LDAPR, etc)
PMULL // Polynomial Multiply instructions (PMULL/PMULL2)
SHA1 // SHA-1 instructions (SHA1C, etc)
SHA2 // SHA-2 instructions (SHA256H, etc)
SHA3 // SHA-3 instructions (EOR3, RAXI, XAR, BCAX)
SHA512 // SHA512 instructions
SM3 // SM3 instructions
SM4 // SM4 instructions
SVE // Scalable Vector Extension
// Keep it last. It automatically defines the size of []flagSet
lastID
firstID FeatureID = UNKNOWN + 1
)
// CPUInfo contains information about the detected system CPU.
type CPUInfo struct {
BrandName string // Brand name reported by the CPU
VendorID Vendor // Comparable CPU vendor ID
VendorString string // Raw vendor string.
HypervisorVendorID Vendor // Hypervisor vendor
HypervisorVendorString string // Raw hypervisor vendor string
featureSet flagSet // Features of the CPU
PhysicalCores int // Number of physical processor cores in your CPU. Will be 0 if undetectable.
ThreadsPerCore int // Number of threads per physical core. Will be 1 if undetectable.
LogicalCores int // Number of physical cores times threads that can run on each core through the use of hyperthreading. Will be 0 if undetectable.
Family int // CPU family number
Model int // CPU model number
Stepping int // CPU stepping info
CacheLine int // Cache line size in bytes. Will be 0 if undetectable.
Hz int64 // Clock speed, if known, 0 otherwise. Will attempt to contain base clock speed.
BoostFreq int64 // Max clock speed, if known, 0 otherwise
Cache struct {
L1I int // L1 Instruction Cache (per core or shared). Will be -1 if undetected
L1D int // L1 Data Cache (per core or shared). Will be -1 if undetected
L2 int // L2 Cache (per core or shared). Will be -1 if undetected
L3 int // L3 Cache (per core, per ccx or shared). Will be -1 if undetected
}
SGX SGXSupport
AMDMemEncryption AMDMemEncryptionSupport
AVX10Level uint8
maxFunc uint32
maxExFunc uint32
}
var cpuid func(op uint32) (eax, ebx, ecx, edx uint32)
var cpuidex func(op, op2 uint32) (eax, ebx, ecx, edx uint32)
var xgetbv func(index uint32) (eax, edx uint32)
var rdtscpAsm func() (eax, ebx, ecx, edx uint32)
var darwinHasAVX512 = func() bool { return false }
// CPU contains information about the CPU as detected on startup,
// or when Detect last was called.
//
// Use this as the primary entry point to you data.
var CPU CPUInfo
func init() {
initCPU()
Detect()
}
// Detect will re-detect current CPU info.
// This will replace the content of the exported CPU variable.
//
// Unless you expect the CPU to change while you are running your program
// you should not need to call this function.
// If you call this, you must ensure that no other goroutine is accessing the
// exported CPU variable.
func Detect() {
// Set defaults
CPU.ThreadsPerCore = 1
CPU.Cache.L1I = -1
CPU.Cache.L1D = -1
CPU.Cache.L2 = -1
CPU.Cache.L3 = -1
safe := true
if detectArmFlag != nil {
safe = !*detectArmFlag
}
addInfo(&CPU, safe)
if displayFeats != nil && *displayFeats {
fmt.Println("cpu features:", strings.Join(CPU.FeatureSet(), ","))
// Exit with non-zero so tests will print value.
os.Exit(1)
}
if disableFlag != nil {
s := strings.Split(*disableFlag, ",")
for _, feat := range s {
feat := ParseFeature(strings.TrimSpace(feat))
if feat != UNKNOWN {
CPU.featureSet.unset(feat)
}
}
}
}
// DetectARM will detect ARM64 features.
// This is NOT done automatically since it can potentially crash
// if the OS does not handle the command.
// If in the future this can be done safely this function may not
// do anything.
func DetectARM() {
addInfo(&CPU, false)
}
var detectArmFlag *bool
var displayFeats *bool
var disableFlag *string
// Flags will enable flags.
// This must be called *before* flag.Parse AND
// Detect must be called after the flags have been parsed.
// Note that this means that any detection used in init() functions
// will not contain these flags.
func Flags() {
disableFlag = flag.String("cpu.disable", "", "disable cpu features; comma separated list")
displayFeats = flag.Bool("cpu.features", false, "lists cpu features and exits")
detectArmFlag = flag.Bool("cpu.arm", false, "allow ARM features to be detected; can potentially crash")
}
// Supports returns whether the CPU supports all of the requested features.
func (c CPUInfo) Supports(ids ...FeatureID) bool {
for _, id := range ids {
if !c.featureSet.inSet(id) {
return false
}
}
return true
}
// Has allows for checking a single feature.
// Should be inlined by the compiler.
func (c *CPUInfo) Has(id FeatureID) bool {
return c.featureSet.inSet(id)
}
// AnyOf returns whether the CPU supports one or more of the requested features.
func (c CPUInfo) AnyOf(ids ...FeatureID) bool {
for _, id := range ids {
if c.featureSet.inSet(id) {
return true
}
}
return false
}
// Features contains several features combined for a fast check using
// CpuInfo.HasAll
type Features *flagSet
// CombineFeatures allows to combine several features for a close to constant time lookup.
func CombineFeatures(ids ...FeatureID) Features {
var v flagSet
for _, id := range ids {
v.set(id)
}
return &v
}
func (c *CPUInfo) HasAll(f Features) bool {
return c.featureSet.hasSetP(f)
}
// https://en.wikipedia.org/wiki/X86-64#Microarchitecture_levels
var oneOfLevel = CombineFeatures(SYSEE, SYSCALL)
var level1Features = CombineFeatures(CMOV, CMPXCHG8, X87, FXSR, MMX, SSE, SSE2)
var level2Features = CombineFeatures(CMOV, CMPXCHG8, X87, FXSR, MMX, SSE, SSE2, CX16, LAHF, POPCNT, SSE3, SSE4, SSE42, SSSE3)
var level3Features = CombineFeatures(CMOV, CMPXCHG8, X87, FXSR, MMX, SSE, SSE2, CX16, LAHF, POPCNT, SSE3, SSE4, SSE42, SSSE3, AVX, AVX2, BMI1, BMI2, F16C, FMA3, LZCNT, MOVBE, OSXSAVE)
var level4Features = CombineFeatures(CMOV, CMPXCHG8, X87, FXSR, MMX, SSE, SSE2, CX16, LAHF, POPCNT, SSE3, SSE4, SSE42, SSSE3, AVX, AVX2, BMI1, BMI2, F16C, FMA3, LZCNT, MOVBE, OSXSAVE, AVX512F, AVX512BW, AVX512CD, AVX512DQ, AVX512VL)
// X64Level returns the microarchitecture level detected on the CPU.
// If features are lacking or non x64 mode, 0 is returned.
// See https://en.wikipedia.org/wiki/X86-64#Microarchitecture_levels
func (c CPUInfo) X64Level() int {
if !c.featureSet.hasOneOf(oneOfLevel) {
return 0
}
if c.featureSet.hasSetP(level4Features) {
return 4
}
if c.featureSet.hasSetP(level3Features) {
return 3
}
if c.featureSet.hasSetP(level2Features) {
return 2
}
if c.featureSet.hasSetP(level1Features) {
return 1
}
return 0
}
// Disable will disable one or several features.
func (c *CPUInfo) Disable(ids ...FeatureID) bool {
for _, id := range ids {
c.featureSet.unset(id)
}
return true
}
// Enable will disable one or several features even if they were undetected.
// This is of course not recommended for obvious reasons.
func (c *CPUInfo) Enable(ids ...FeatureID) bool {
for _, id := range ids {
c.featureSet.set(id)
}
return true
}
// IsVendor returns true if vendor is recognized as Intel
func (c CPUInfo) IsVendor(v Vendor) bool {
return c.VendorID == v
}
// FeatureSet returns all available features as strings.
func (c CPUInfo) FeatureSet() []string {
s := make([]string, 0, c.featureSet.nEnabled())
s = append(s, c.featureSet.Strings()...)
return s
}
// RTCounter returns the 64-bit time-stamp counter
// Uses the RDTSCP instruction. The value 0 is returned
// if the CPU does not support the instruction.
func (c CPUInfo) RTCounter() uint64 {
if !c.Has(RDTSCP) {
return 0
}
a, _, _, d := rdtscpAsm()
return uint64(a) | (uint64(d) << 32)
}
// Ia32TscAux returns the IA32_TSC_AUX part of the RDTSCP.
// This variable is OS dependent, but on Linux contains information
// about the current cpu/core the code is running on.
// If the RDTSCP instruction isn't supported on the CPU, the value 0 is returned.
func (c CPUInfo) Ia32TscAux() uint32 {
if !c.Has(RDTSCP) {
return 0
}
_, _, ecx, _ := rdtscpAsm()
return ecx
}
// SveLengths returns arm SVE vector and predicate lengths.
// Will return 0, 0 if SVE is not enabled or otherwise unable to detect.
func (c CPUInfo) SveLengths() (vl, pl uint64) {
if !c.Has(SVE) {
return 0, 0
}
return getVectorLength()
}
// LogicalCPU will return the Logical CPU the code is currently executing on.
// This is likely to change when the OS re-schedules the running thread
// to another CPU.
// If the current core cannot be detected, -1 will be returned.
func (c CPUInfo) LogicalCPU() int {
if c.maxFunc < 1 {
return -1
}
_, ebx, _, _ := cpuid(1)
return int(ebx >> 24)
}
// frequencies tries to compute the clock speed of the CPU. If leaf 15 is
// supported, use it, otherwise parse the brand string. Yes, really.
func (c *CPUInfo) frequencies() {
c.Hz, c.BoostFreq = 0, 0
mfi := maxFunctionID()
if mfi >= 0x15 {
eax, ebx, ecx, _ := cpuid(0x15)
if eax != 0 && ebx != 0 && ecx != 0 {
c.Hz = (int64(ecx) * int64(ebx)) / int64(eax)
}
}
if mfi >= 0x16 {
a, b, _, _ := cpuid(0x16)
// Base...
if a&0xffff > 0 {
c.Hz = int64(a&0xffff) * 1_000_000
}
// Boost...
if b&0xffff > 0 {
c.BoostFreq = int64(b&0xffff) * 1_000_000
}
}
if c.Hz > 0 {
return
}
// computeHz determines the official rated speed of a CPU from its brand
// string. This insanity is *actually the official documented way to do
// this according to Intel*, prior to leaf 0x15 existing. The official
// documentation only shows this working for exactly `x.xx` or `xxxx`
// cases, e.g., `2.50GHz` or `1300MHz`; this parser will accept other
// sizes.
model := c.BrandName
hz := strings.LastIndex(model, "Hz")
if hz < 3 {
return
}
var multiplier int64
switch model[hz-1] {
case 'M':
multiplier = 1000 * 1000
case 'G':
multiplier = 1000 * 1000 * 1000
case 'T':
multiplier = 1000 * 1000 * 1000 * 1000
}
if multiplier == 0 {
return
}
freq := int64(0)
divisor := int64(0)
decimalShift := int64(1)
var i int
for i = hz - 2; i >= 0 && model[i] != ' '; i-- {
if model[i] >= '0' && model[i] <= '9' {
freq += int64(model[i]-'0') * decimalShift
decimalShift *= 10
} else if model[i] == '.' {
if divisor != 0 {
return
}
divisor = decimalShift
} else {
return
}
}
// we didn't find a space
if i < 0 {
return
}
if divisor != 0 {
c.Hz = (freq * multiplier) / divisor
return
}
c.Hz = freq * multiplier
}
// VM Will return true if the cpu id indicates we are in
// a virtual machine.
func (c CPUInfo) VM() bool {
return CPU.featureSet.inSet(HYPERVISOR)
}
// flags contains detected cpu features and characteristics
type flags uint64
// log2(bits_in_uint64)
const flagBitsLog2 = 6
const flagBits = 1 << flagBitsLog2
const flagMask = flagBits - 1
// flagSet contains detected cpu features and characteristics in an array of flags
type flagSet [(lastID + flagMask) / flagBits]flags
func (s *flagSet) inSet(feat FeatureID) bool {
return s[feat>>flagBitsLog2]&(1<<(feat&flagMask)) != 0
}
func (s *flagSet) set(feat FeatureID) {
s[feat>>flagBitsLog2] |= 1 << (feat & flagMask)
}
// setIf will set a feature if boolean is true.
func (s *flagSet) setIf(cond bool, features ...FeatureID) {
if cond {
for _, offset := range features {
s[offset>>flagBitsLog2] |= 1 << (offset & flagMask)
}
}
}
func (s *flagSet) unset(offset FeatureID) {
bit := flags(1 << (offset & flagMask))
s[offset>>flagBitsLog2] = s[offset>>flagBitsLog2] & ^bit
}
// or with another flagset.
func (s *flagSet) or(other flagSet) {
for i, v := range other[:] {
s[i] |= v
}
}
// hasSet returns whether all features are present.
func (s *flagSet) hasSet(other flagSet) bool {
for i, v := range other[:] {
if s[i]&v != v {
return false
}
}
return true
}
// hasSet returns whether all features are present.
func (s *flagSet) hasSetP(other *flagSet) bool {
for i, v := range other[:] {
if s[i]&v != v {
return false
}
}
return true
}
// hasOneOf returns whether one or more features are present.
func (s *flagSet) hasOneOf(other *flagSet) bool {
for i, v := range other[:] {
if s[i]&v != 0 {
return true
}
}
return false
}
// nEnabled will return the number of enabled flags.
func (s *flagSet) nEnabled() (n int) {
for _, v := range s[:] {
n += bits.OnesCount64(uint64(v))
}
return n
}
func flagSetWith(feat ...FeatureID) flagSet {
var res flagSet
for _, f := range feat {
res.set(f)
}
return res
}
// ParseFeature will parse the string and return the ID of the matching feature.
// Will return UNKNOWN if not found.
func ParseFeature(s string) FeatureID {
s = strings.ToUpper(s)
for i := firstID; i < lastID; i++ {
if i.String() == s {
return i
}
}
return UNKNOWN
}
// Strings returns an array of the detected features for FlagsSet.
func (s flagSet) Strings() []string {
if len(s) == 0 {
return []string{""}
}
r := make([]string, 0)
for i := firstID; i < lastID; i++ {
if s.inSet(i) {
r = append(r, i.String())
}
}
return r
}
func maxExtendedFunction() uint32 {
eax, _, _, _ := cpuid(0x80000000)
return eax
}
func maxFunctionID() uint32 {
a, _, _, _ := cpuid(0)
return a
}
func brandName() string {
if maxExtendedFunction() >= 0x80000004 {
v := make([]uint32, 0, 48)
for i := uint32(0); i < 3; i++ {
a, b, c, d := cpuid(0x80000002 + i)
v = append(v, a, b, c, d)
}
return strings.Trim(string(valAsString(v...)), " ")
}
return "unknown"
}
func threadsPerCore() int {
mfi := maxFunctionID()
vend, _ := vendorID()
if mfi < 0x4 || (vend != Intel && vend != AMD) {
return 1
}
if mfi < 0xb {
if vend != Intel {
return 1
}
_, b, _, d := cpuid(1)
if (d & (1 << 28)) != 0 {
// v will contain logical core count
v := (b >> 16) & 255
if v > 1 {
a4, _, _, _ := cpuid(4)
// physical cores
v2 := (a4 >> 26) + 1
if v2 > 0 {
return int(v) / int(v2)
}
}
}
return 1
}
_, b, _, _ := cpuidex(0xb, 0)
if b&0xffff == 0 {
if vend == AMD {
// Workaround for AMD returning 0, assume 2 if >= Zen 2
// It will be more correct than not.
fam, _, _ := familyModel()
_, _, _, d := cpuid(1)
if (d&(1<<28)) != 0 && fam >= 23 {
return 2
}
}
return 1
}
return int(b & 0xffff)
}
func logicalCores() int {
mfi := maxFunctionID()
v, _ := vendorID()
switch v {
case Intel:
// Use this on old Intel processors
if mfi < 0xb {
if mfi < 1 {
return 0
}
// CPUID.1:EBX[23:16] represents the maximum number of addressable IDs (initial APIC ID)
// that can be assigned to logical processors in a physical package.
// The value may not be the same as the number of logical processors that are present in the hardware of a physical package.
_, ebx, _, _ := cpuid(1)
logical := (ebx >> 16) & 0xff
return int(logical)
}
_, b, _, _ := cpuidex(0xb, 1)
return int(b & 0xffff)
case AMD, Hygon:
_, b, _, _ := cpuid(1)
return int((b >> 16) & 0xff)
default:
return 0
}
}
func familyModel() (family, model, stepping int) {
if maxFunctionID() < 0x1 {
return 0, 0, 0
}
eax, _, _, _ := cpuid(1)
// If BaseFamily[3:0] is less than Fh then ExtendedFamily[7:0] is reserved and Family is equal to BaseFamily[3:0].
family = int((eax >> 8) & 0xf)
extFam := family == 0x6 // Intel is 0x6, needs extended model.
if family == 0xf {
// Add ExtFamily
family += int((eax >> 20) & 0xff)
extFam = true
}
// If BaseFamily[3:0] is less than 0Fh then ExtendedModel[3:0] is reserved and Model is equal to BaseModel[3:0].
model = int((eax >> 4) & 0xf)
if extFam {
// Add ExtModel
model += int((eax >> 12) & 0xf0)
}
stepping = int(eax & 0xf)
return family, model, stepping
}
func physicalCores() int {
v, _ := vendorID()
switch v {
case Intel:
return logicalCores() / threadsPerCore()
case AMD, Hygon:
lc := logicalCores()
tpc := threadsPerCore()
if lc > 0 && tpc > 0 {
return lc / tpc
}
// The following is inaccurate on AMD EPYC 7742 64-Core Processor
if maxExtendedFunction() >= 0x80000008 {
_, _, c, _ := cpuid(0x80000008)
if c&0xff > 0 {
return int(c&0xff) + 1
}
}
}
return 0
}
// Except from http://en.wikipedia.org/wiki/CPUID#EAX.3D0:_Get_vendor_ID
var vendorMapping = map[string]Vendor{
"AMDisbetter!": AMD,
"AuthenticAMD": AMD,
"CentaurHauls": VIA,
"GenuineIntel": Intel,
"TransmetaCPU": Transmeta,
"GenuineTMx86": Transmeta,
"Geode by NSC": NSC,
"VIA VIA VIA ": VIA,
"KVMKVMKVM": KVM,
"Linux KVM Hv": KVM,
"TCGTCGTCGTCG": QEMU,
"Microsoft Hv": MSVM,
"VMwareVMware": VMware,
"XenVMMXenVMM": XenHVM,
"bhyve bhyve ": Bhyve,
"HygonGenuine": Hygon,
"Vortex86 SoC": SiS,
"SiS SiS SiS ": SiS,
"RiseRiseRise": SiS,
"Genuine RDC": RDC,
"QNXQVMBSQG": QNX,
"ACRNACRNACRN": ACRN,
"SRESRESRESRE": SRE,
"Apple VZ": Apple,
}
func vendorID() (Vendor, string) {
_, b, c, d := cpuid(0)
v := string(valAsString(b, d, c))
vend, ok := vendorMapping[v]
if !ok {
return VendorUnknown, v
}
return vend, v
}
func hypervisorVendorID() (Vendor, string) {
// https://lwn.net/Articles/301888/
_, b, c, d := cpuid(0x40000000)
v := string(valAsString(b, c, d))
vend, ok := vendorMapping[v]
if !ok {
return VendorUnknown, v
}
return vend, v
}
func cacheLine() int {
if maxFunctionID() < 0x1 {
return 0
}
_, ebx, _, _ := cpuid(1)
cache := (ebx & 0xff00) >> 5 // cflush size
if cache == 0 && maxExtendedFunction() >= 0x80000006 {
_, _, ecx, _ := cpuid(0x80000006)
cache = ecx & 0xff // cacheline size
}
// TODO: Read from Cache and TLB Information
return int(cache)
}
func (c *CPUInfo) cacheSize() {
c.Cache.L1D = -1
c.Cache.L1I = -1
c.Cache.L2 = -1
c.Cache.L3 = -1
vendor, _ := vendorID()
switch vendor {
case Intel:
if maxFunctionID() < 4 {
return
}
c.Cache.L1I, c.Cache.L1D, c.Cache.L2, c.Cache.L3 = 0, 0, 0, 0
for i := uint32(0); ; i++ {
eax, ebx, ecx, _ := cpuidex(4, i)
cacheType := eax & 15
if cacheType == 0 {
break
}
cacheLevel := (eax >> 5) & 7
coherency := int(ebx&0xfff) + 1
partitions := int((ebx>>12)&0x3ff) + 1
associativity := int((ebx>>22)&0x3ff) + 1
sets := int(ecx) + 1
size := associativity * partitions * coherency * sets
switch cacheLevel {
case 1:
if cacheType == 1 {
// 1 = Data Cache
c.Cache.L1D = size
} else if cacheType == 2 {
// 2 = Instruction Cache
c.Cache.L1I = size
} else {
if c.Cache.L1D < 0 {
c.Cache.L1I = size
}
if c.Cache.L1I < 0 {
c.Cache.L1I = size
}
}
case 2:
c.Cache.L2 = size
case 3:
c.Cache.L3 = size
}
}
case AMD, Hygon:
// Untested.
if maxExtendedFunction() < 0x80000005 {