From 404aeae4545d2426c089a5f8d5e82dae56f5212b Mon Sep 17 00:00:00 2001 From: Rutger Broekhoff Date: Fri, 29 Dec 2023 21:31:53 +0100 Subject: Make Nix builds work --- vendor/github.com/klauspost/cpuid/v2/.gitignore | 24 + .../github.com/klauspost/cpuid/v2/.goreleaser.yml | 74 + .../github.com/klauspost/cpuid/v2/CONTRIBUTING.txt | 35 + vendor/github.com/klauspost/cpuid/v2/LICENSE | 22 + vendor/github.com/klauspost/cpuid/v2/README.md | 497 +++++++ vendor/github.com/klauspost/cpuid/v2/cpuid.go | 1473 ++++++++++++++++++++ vendor/github.com/klauspost/cpuid/v2/cpuid_386.s | 47 + vendor/github.com/klauspost/cpuid/v2/cpuid_amd64.s | 72 + vendor/github.com/klauspost/cpuid/v2/cpuid_arm64.s | 26 + .../github.com/klauspost/cpuid/v2/detect_arm64.go | 247 ++++ vendor/github.com/klauspost/cpuid/v2/detect_ref.go | 15 + vendor/github.com/klauspost/cpuid/v2/detect_x86.go | 37 + .../klauspost/cpuid/v2/featureid_string.go | 279 ++++ .../klauspost/cpuid/v2/os_darwin_arm64.go | 121 ++ .../klauspost/cpuid/v2/os_linux_arm64.go | 130 ++ .../klauspost/cpuid/v2/os_other_arm64.go | 16 + .../klauspost/cpuid/v2/os_safe_linux_arm64.go | 8 + .../klauspost/cpuid/v2/os_unsafe_linux_arm64.go | 11 + .../klauspost/cpuid/v2/test-architectures.sh | 15 + 19 files changed, 3149 insertions(+) create mode 100644 vendor/github.com/klauspost/cpuid/v2/.gitignore create mode 100644 vendor/github.com/klauspost/cpuid/v2/.goreleaser.yml create mode 100644 vendor/github.com/klauspost/cpuid/v2/CONTRIBUTING.txt create mode 100644 vendor/github.com/klauspost/cpuid/v2/LICENSE create mode 100644 vendor/github.com/klauspost/cpuid/v2/README.md create mode 100644 vendor/github.com/klauspost/cpuid/v2/cpuid.go create mode 100644 vendor/github.com/klauspost/cpuid/v2/cpuid_386.s create mode 100644 vendor/github.com/klauspost/cpuid/v2/cpuid_amd64.s create mode 100644 vendor/github.com/klauspost/cpuid/v2/cpuid_arm64.s create mode 100644 vendor/github.com/klauspost/cpuid/v2/detect_arm64.go create mode 100644 vendor/github.com/klauspost/cpuid/v2/detect_ref.go create mode 100644 vendor/github.com/klauspost/cpuid/v2/detect_x86.go create mode 100644 vendor/github.com/klauspost/cpuid/v2/featureid_string.go create mode 100644 vendor/github.com/klauspost/cpuid/v2/os_darwin_arm64.go create mode 100644 vendor/github.com/klauspost/cpuid/v2/os_linux_arm64.go create mode 100644 vendor/github.com/klauspost/cpuid/v2/os_other_arm64.go create mode 100644 vendor/github.com/klauspost/cpuid/v2/os_safe_linux_arm64.go create mode 100644 vendor/github.com/klauspost/cpuid/v2/os_unsafe_linux_arm64.go create mode 100644 vendor/github.com/klauspost/cpuid/v2/test-architectures.sh (limited to 'vendor/github.com/klauspost/cpuid/v2') diff --git a/vendor/github.com/klauspost/cpuid/v2/.gitignore b/vendor/github.com/klauspost/cpuid/v2/.gitignore new file mode 100644 index 0000000..daf913b --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/.gitignore @@ -0,0 +1,24 @@ +# Compiled Object files, Static and Dynamic libs (Shared Objects) +*.o +*.a +*.so + +# Folders +_obj +_test + +# Architecture specific extensions/prefixes +*.[568vq] +[568vq].out + +*.cgo1.go +*.cgo2.c +_cgo_defun.c +_cgo_gotypes.go +_cgo_export.* + +_testmain.go + +*.exe +*.test +*.prof diff --git a/vendor/github.com/klauspost/cpuid/v2/.goreleaser.yml b/vendor/github.com/klauspost/cpuid/v2/.goreleaser.yml new file mode 100644 index 0000000..944cc00 --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/.goreleaser.yml @@ -0,0 +1,74 @@ +# This is an example goreleaser.yaml file with some sane defaults. +# Make sure to check the documentation at http://goreleaser.com + +builds: + - + id: "cpuid" + binary: cpuid + main: ./cmd/cpuid/main.go + env: + - CGO_ENABLED=0 + flags: + - -ldflags=-s -w + goos: + - aix + - linux + - freebsd + - netbsd + - windows + - darwin + goarch: + - 386 + - amd64 + - arm64 + goarm: + - 7 + +archives: + - + id: cpuid + name_template: "cpuid-{{ .Os }}_{{ .Arch }}_{{ .Version }}" + replacements: + aix: AIX + darwin: OSX + linux: Linux + windows: Windows + 386: i386 + amd64: x86_64 + freebsd: FreeBSD + netbsd: NetBSD + format_overrides: + - goos: windows + format: zip + files: + - LICENSE +checksum: + name_template: 'checksums.txt' +snapshot: + name_template: "{{ .Tag }}-next" +changelog: + sort: asc + filters: + exclude: + - '^doc:' + - '^docs:' + - '^test:' + - '^tests:' + - '^Update\sREADME.md' + +nfpms: + - + file_name_template: "cpuid_package_{{ .Version }}_{{ .Os }}_{{ .Arch }}" + vendor: Klaus Post + homepage: https://github.com/klauspost/cpuid + maintainer: Klaus Post + description: CPUID Tool + license: BSD 3-Clause + formats: + - deb + - rpm + replacements: + darwin: Darwin + linux: Linux + freebsd: FreeBSD + amd64: x86_64 diff --git a/vendor/github.com/klauspost/cpuid/v2/CONTRIBUTING.txt b/vendor/github.com/klauspost/cpuid/v2/CONTRIBUTING.txt new file mode 100644 index 0000000..452d28e --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/CONTRIBUTING.txt @@ -0,0 +1,35 @@ +Developer Certificate of Origin +Version 1.1 + +Copyright (C) 2015- Klaus Post & Contributors. +Email: klauspost@gmail.com + +Everyone is permitted to copy and distribute verbatim copies of this +license document, but changing it is not allowed. + + +Developer's Certificate of Origin 1.1 + +By making a contribution to this project, I certify that: + +(a) The contribution was created in whole or in part by me and I + have the right to submit it under the open source license + indicated in the file; or + +(b) The contribution is based upon previous work that, to the best + of my knowledge, is covered under an appropriate open source + license and I have the right under that license to submit that + work with modifications, whether created in whole or in part + by me, under the same open source license (unless I am + permitted to submit under a different license), as indicated + in the file; or + +(c) The contribution was provided directly to me by some other + person who certified (a), (b) or (c) and I have not modified + it. + +(d) I understand and agree that this project and the contribution + are public and that a record of the contribution (including all + personal information I submit with it, including my sign-off) is + maintained indefinitely and may be redistributed consistent with + this project or the open source license(s) involved. diff --git a/vendor/github.com/klauspost/cpuid/v2/LICENSE b/vendor/github.com/klauspost/cpuid/v2/LICENSE new file mode 100644 index 0000000..5cec7ee --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/LICENSE @@ -0,0 +1,22 @@ +The MIT License (MIT) + +Copyright (c) 2015 Klaus Post + +Permission is hereby granted, free of charge, to any person obtaining a copy +of this software and associated documentation files (the "Software"), to deal +in the Software without restriction, including without limitation the rights +to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +copies of the Software, and to permit persons to whom the Software is +furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. + diff --git a/vendor/github.com/klauspost/cpuid/v2/README.md b/vendor/github.com/klauspost/cpuid/v2/README.md new file mode 100644 index 0000000..30f8d29 --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/README.md @@ -0,0 +1,497 @@ +# cpuid +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/i386) and ARM (ARM64) is supported, and no external C (cgo) code is used, which should make the library very easy to use. + +You can access the CPU information by accessing the shared CPU variable of the cpuid library. + +Package home: https://github.com/klauspost/cpuid + +[![PkgGoDev](https://pkg.go.dev/badge/github.com/klauspost/cpuid)](https://pkg.go.dev/github.com/klauspost/cpuid/v2) +[![Go](https://github.com/klauspost/cpuid/actions/workflows/go.yml/badge.svg)](https://github.com/klauspost/cpuid/actions/workflows/go.yml) + +## installing + +`go get -u github.com/klauspost/cpuid/v2` using modules. +Drop `v2` for others. + +Installing binary: + +`go install github.com/klauspost/cpuid/v2/cmd/cpuid@latest` + +Or download binaries from release page: https://github.com/klauspost/cpuid/releases + +### Homebrew + +For macOS/Linux users, you can install via [brew](https://brew.sh/) + +```sh +$ brew install cpuid +``` + +## example + +```Go +package main + +import ( + "fmt" + "strings" + + . "github.com/klauspost/cpuid/v2" +) + +func main() { + // Print basic CPU information: + fmt.Println("Name:", CPU.BrandName) + fmt.Println("PhysicalCores:", CPU.PhysicalCores) + fmt.Println("ThreadsPerCore:", CPU.ThreadsPerCore) + fmt.Println("LogicalCores:", CPU.LogicalCores) + fmt.Println("Family", CPU.Family, "Model:", CPU.Model, "Vendor ID:", CPU.VendorID) + fmt.Println("Features:", strings.Join(CPU.FeatureSet(), ",")) + fmt.Println("Cacheline bytes:", CPU.CacheLine) + fmt.Println("L1 Data Cache:", CPU.Cache.L1D, "bytes") + fmt.Println("L1 Instruction Cache:", CPU.Cache.L1I, "bytes") + fmt.Println("L2 Cache:", CPU.Cache.L2, "bytes") + fmt.Println("L3 Cache:", CPU.Cache.L3, "bytes") + fmt.Println("Frequency", CPU.Hz, "hz") + + // Test if we have these specific features: + if CPU.Supports(SSE, SSE2) { + fmt.Println("We have Streaming SIMD 2 Extensions") + } +} +``` + +Sample output: +``` +>go run main.go +Name: AMD Ryzen 9 3950X 16-Core Processor +PhysicalCores: 16 +ThreadsPerCore: 2 +LogicalCores: 32 +Family 23 Model: 113 Vendor ID: AMD +Features: ADX,AESNI,AVX,AVX2,BMI1,BMI2,CLMUL,CMOV,CX16,F16C,FMA3,HTT,HYPERVISOR,LZCNT,MMX,MMXEXT,NX,POPCNT,RDRAND,RDSEED,RDTSCP,SHA,SSE,SSE2,SSE3,SSE4,SSE42,SSE4A,SSSE3 +Cacheline bytes: 64 +L1 Data Cache: 32768 bytes +L1 Instruction Cache: 32768 bytes +L2 Cache: 524288 bytes +L3 Cache: 16777216 bytes +Frequency 0 hz +We have Streaming SIMD 2 Extensions +``` + +# usage + +The `cpuid.CPU` provides access to CPU features. Use `cpuid.CPU.Supports()` to check for CPU features. +A faster `cpuid.CPU.Has()` is provided which will usually be inlined by the gc compiler. + +To test a larger number of features, they can be combined using `f := CombineFeatures(CMOV, CMPXCHG8, X87, FXSR, MMX, SYSCALL, SSE, SSE2)`, etc. +This can be using with `cpuid.CPU.HasAll(f)` to quickly test if all features are supported. + +Note that for some cpu/os combinations some features will not be detected. +`amd64` has rather good support and should work reliably on all platforms. + +Note that hypervisors may not pass through all CPU features through to the guest OS, +so even if your host supports a feature it may not be visible on guests. + +## arm64 feature detection + +Not all operating systems provide ARM features directly +and there is no safe way to do so for the rest. + +Currently `arm64/linux` and `arm64/freebsd` should be quite reliable. +`arm64/darwin` adds features expected from the M1 processor, but a lot remains undetected. + +A `DetectARM()` can be used if you are able to control your deployment, +it will detect CPU features, but may crash if the OS doesn't intercept the calls. +A `-cpu.arm` flag for detecting unsafe ARM features can be added. See below. + +Note that currently only features are detected on ARM, +no additional information is currently available. + +## flags + +It is possible to add flags that affects cpu detection. + +For this the `Flags()` command is provided. + +This must be called *before* `flag.Parse()` AND after the flags have been parsed `Detect()` must be called. + +This means that any detection used in `init()` functions will not contain these flags. + +Example: + +```Go +package main + +import ( + "flag" + "fmt" + "strings" + + "github.com/klauspost/cpuid/v2" +) + +func main() { + cpuid.Flags() + flag.Parse() + cpuid.Detect() + + // Test if we have these specific features: + if cpuid.CPU.Supports(cpuid.SSE, cpuid.SSE2) { + fmt.Println("We have Streaming SIMD 2 Extensions") + } +} +``` + +## commandline + +Download as binary from: https://github.com/klauspost/cpuid/releases + +Install from source: + +`go install github.com/klauspost/cpuid/v2/cmd/cpuid@latest` + +### Example + +``` +λ cpuid +Name: AMD Ryzen 9 3950X 16-Core Processor +Vendor String: AuthenticAMD +Vendor ID: AMD +PhysicalCores: 16 +Threads Per Core: 2 +Logical Cores: 32 +CPU Family 23 Model: 113 +Features: ADX,AESNI,AVX,AVX2,BMI1,BMI2,CLMUL,CLZERO,CMOV,CMPXCHG8,CPBOOST,CX16,F16C,FMA3,FXSR,FXSROPT,HTT,HYPERVISOR,LAHF,LZCNT,MCAOVERFLOW,MMX,MMXEXT,MOVBE,NX,OSXSAVE,POPCNT,RDRAND,RDSEED,RDTSCP,SCE,SHA,SSE,SSE2,SSE3,SSE4,SSE42,SSE4A,SSSE3,SUCCOR,X87,XSAVE +Microarchitecture level: 3 +Cacheline bytes: 64 +L1 Instruction Cache: 32768 bytes +L1 Data Cache: 32768 bytes +L2 Cache: 524288 bytes +L3 Cache: 16777216 bytes + +``` +### JSON Output: + +``` +λ cpuid --json +{ + "BrandName": "AMD Ryzen 9 3950X 16-Core Processor", + "VendorID": 2, + "VendorString": "AuthenticAMD", + "PhysicalCores": 16, + "ThreadsPerCore": 2, + "LogicalCores": 32, + "Family": 23, + "Model": 113, + "CacheLine": 64, + "Hz": 0, + "BoostFreq": 0, + "Cache": { + "L1I": 32768, + "L1D": 32768, + "L2": 524288, + "L3": 16777216 + }, + "SGX": { + "Available": false, + "LaunchControl": false, + "SGX1Supported": false, + "SGX2Supported": false, + "MaxEnclaveSizeNot64": 0, + "MaxEnclaveSize64": 0, + "EPCSections": null + }, + "Features": [ + "ADX", + "AESNI", + "AVX", + "AVX2", + "BMI1", + "BMI2", + "CLMUL", + "CLZERO", + "CMOV", + "CMPXCHG8", + "CPBOOST", + "CX16", + "F16C", + "FMA3", + "FXSR", + "FXSROPT", + "HTT", + "HYPERVISOR", + "LAHF", + "LZCNT", + "MCAOVERFLOW", + "MMX", + "MMXEXT", + "MOVBE", + "NX", + "OSXSAVE", + "POPCNT", + "RDRAND", + "RDSEED", + "RDTSCP", + "SCE", + "SHA", + "SSE", + "SSE2", + "SSE3", + "SSE4", + "SSE42", + "SSE4A", + "SSSE3", + "SUCCOR", + "X87", + "XSAVE" + ], + "X64Level": 3 +} +``` + +### Check CPU microarch level + +``` +λ cpuid --check-level=3 +2022/03/18 17:04:40 AMD Ryzen 9 3950X 16-Core Processor +2022/03/18 17:04:40 Microarchitecture level 3 is supported. Max level is 3. +Exit Code 0 + +λ cpuid --check-level=4 +2022/03/18 17:06:18 AMD Ryzen 9 3950X 16-Core Processor +2022/03/18 17:06:18 Microarchitecture level 4 not supported. Max level is 3. +Exit Code 1 +``` + + +## Available flags + +### x86 & amd64 + +| Feature Flag | Description | +|--------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| ADX | 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 | +| AMXINT8 | Tile computational operations on 8-bit integers | +| AMXFP16 | Tile computational operations on FP16 numbers | +| 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 | +| 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 128-bits wide | +| FP256 | AMD: When set, the internal FP/SIMD execution datapath is 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) | +| 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 | 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 | +| MPX | Intel MPX (Memory Protection Extensions) | +| MOVU | 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 | +| 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. | +| 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 | +| 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: + +| Feature Flag | Description | +|--------------|------------------------------------------------------------------| +| 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 | + +# license + +This code is published under an MIT license. See LICENSE file for more information. diff --git a/vendor/github.com/klauspost/cpuid/v2/cpuid.go b/vendor/github.com/klauspost/cpuid/v2/cpuid.go new file mode 100644 index 0000000..15b7603 --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/cpuid.go @@ -0,0 +1,1473 @@ +// 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 + + 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 + + // Add 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 + 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) + 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. + 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 + 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. + 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 + 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.Supports(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.Supports(RDTSCP) { + return 0 + } + _, _, ecx, _ := rdtscpAsm() + return ecx +} + +// 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, + "KVMKVMKVMKVM": KVM, + "Microsoft Hv": MSVM, + "VMwareVMware": VMware, + "XenVMMXenVMM": XenHVM, + "bhyve bhyve ": Bhyve, + "HygonGenuine": Hygon, + "Vortex86 SoC": SiS, + "SiS SiS SiS ": SiS, + "RiseRiseRise": SiS, + "Genuine RDC": RDC, +} + +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 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 { + return + } + _, _, ecx, edx := cpuid(0x80000005) + c.Cache.L1D = int(((ecx >> 24) & 0xFF) * 1024) + c.Cache.L1I = int(((edx >> 24) & 0xFF) * 1024) + + if maxExtendedFunction() < 0x80000006 { + return + } + _, _, ecx, _ = cpuid(0x80000006) + c.Cache.L2 = int(((ecx >> 16) & 0xFFFF) * 1024) + + // CPUID Fn8000_001D_EAX_x[N:0] Cache Properties + if maxExtendedFunction() < 0x8000001D || !c.Has(TOPEXT) { + return + } + + // Xen Hypervisor is buggy and returns the same entry no matter ECX value. + // Hack: When we encounter the same entry 100 times we break. + nSame := 0 + var last uint32 + for i := uint32(0); i < math.MaxUint32; i++ { + eax, ebx, ecx, _ := cpuidex(0x8000001D, i) + + level := (eax >> 5) & 7 + cacheNumSets := ecx + 1 + cacheLineSize := 1 + (ebx & 2047) + cachePhysPartitions := 1 + ((ebx >> 12) & 511) + cacheNumWays := 1 + ((ebx >> 22) & 511) + + typ := eax & 15 + size := int(cacheNumSets * cacheLineSize * cachePhysPartitions * cacheNumWays) + if typ == 0 { + return + } + + // Check for the same value repeated. + comb := eax ^ ebx ^ ecx + if comb == last { + nSame++ + if nSame == 100 { + return + } + } + last = comb + + switch level { + case 1: + switch typ { + case 1: + // Data cache + c.Cache.L1D = size + case 2: + // Inst cache + c.Cache.L1I = size + default: + 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 + } + } + } +} + +type SGXEPCSection struct { + BaseAddress uint64 + EPCSize uint64 +} + +type SGXSupport struct { + Available bool + LaunchControl bool + SGX1Supported bool + SGX2Supported bool + MaxEnclaveSizeNot64 int64 + MaxEnclaveSize64 int64 + EPCSections []SGXEPCSection +} + +func hasSGX(available, lc bool) (rval SGXSupport) { + rval.Available = available + + if !available { + return + } + + rval.LaunchControl = lc + + a, _, _, d := cpuidex(0x12, 0) + rval.SGX1Supported = a&0x01 != 0 + rval.SGX2Supported = a&0x02 != 0 + rval.MaxEnclaveSizeNot64 = 1 << (d & 0xFF) // pow 2 + rval.MaxEnclaveSize64 = 1 << ((d >> 8) & 0xFF) // pow 2 + rval.EPCSections = make([]SGXEPCSection, 0) + + for subleaf := uint32(2); subleaf < 2+8; subleaf++ { + eax, ebx, ecx, edx := cpuidex(0x12, subleaf) + leafType := eax & 0xf + + if leafType == 0 { + // Invalid subleaf, stop iterating + break + } else if leafType == 1 { + // EPC Section subleaf + baseAddress := uint64(eax&0xfffff000) + (uint64(ebx&0x000fffff) << 32) + size := uint64(ecx&0xfffff000) + (uint64(edx&0x000fffff) << 32) + + section := SGXEPCSection{BaseAddress: baseAddress, EPCSize: size} + rval.EPCSections = append(rval.EPCSections, section) + } + } + + return +} + +func support() flagSet { + var fs flagSet + mfi := maxFunctionID() + vend, _ := vendorID() + if mfi < 0x1 { + return fs + } + family, model, _ := familyModel() + + _, _, c, d := cpuid(1) + fs.setIf((d&(1<<0)) != 0, X87) + fs.setIf((d&(1<<8)) != 0, CMPXCHG8) + fs.setIf((d&(1<<11)) != 0, SYSEE) + fs.setIf((d&(1<<15)) != 0, CMOV) + fs.setIf((d&(1<<23)) != 0, MMX) + fs.setIf((d&(1<<24)) != 0, FXSR) + fs.setIf((d&(1<<25)) != 0, FXSROPT) + fs.setIf((d&(1<<25)) != 0, SSE) + fs.setIf((d&(1<<26)) != 0, SSE2) + fs.setIf((c&1) != 0, SSE3) + fs.setIf((c&(1<<5)) != 0, VMX) + fs.setIf((c&(1<<9)) != 0, SSSE3) + fs.setIf((c&(1<<19)) != 0, SSE4) + fs.setIf((c&(1<<20)) != 0, SSE42) + fs.setIf((c&(1<<25)) != 0, AESNI) + fs.setIf((c&(1<<1)) != 0, CLMUL) + fs.setIf(c&(1<<22) != 0, MOVBE) + fs.setIf(c&(1<<23) != 0, POPCNT) + fs.setIf(c&(1<<30) != 0, RDRAND) + + // This bit has been reserved by Intel & AMD for use by hypervisors, + // and indicates the presence of a hypervisor. + fs.setIf(c&(1<<31) != 0, HYPERVISOR) + fs.setIf(c&(1<<29) != 0, F16C) + fs.setIf(c&(1<<13) != 0, CX16) + + if vend == Intel && (d&(1<<28)) != 0 && mfi >= 4 { + fs.setIf(threadsPerCore() > 1, HTT) + } + if vend == AMD && (d&(1<<28)) != 0 && mfi >= 4 { + fs.setIf(threadsPerCore() > 1, HTT) + } + fs.setIf(c&1<<26 != 0, XSAVE) + fs.setIf(c&1<<27 != 0, OSXSAVE) + // Check XGETBV/XSAVE (26), OXSAVE (27) and AVX (28) bits + const avxCheck = 1<<26 | 1<<27 | 1<<28 + if c&avxCheck == avxCheck { + // Check for OS support + eax, _ := xgetbv(0) + if (eax & 0x6) == 0x6 { + fs.set(AVX) + switch vend { + case Intel: + // Older than Haswell. + fs.setIf(family == 6 && model < 60, AVXSLOW) + case AMD: + // Older than Zen 2 + fs.setIf(family < 23 || (family == 23 && model < 49), AVXSLOW) + } + } + } + // FMA3 can be used with SSE registers, so no OS support is strictly needed. + // fma3 and OSXSAVE needed. + const fma3Check = 1<<12 | 1<<27 + fs.setIf(c&fma3Check == fma3Check, FMA3) + + // Check AVX2, AVX2 requires OS support, but BMI1/2 don't. + if mfi >= 7 { + _, ebx, ecx, edx := cpuidex(7, 0) + if fs.inSet(AVX) && (ebx&0x00000020) != 0 { + fs.set(AVX2) + } + // CPUID.(EAX=7, ECX=0).EBX + if (ebx & 0x00000008) != 0 { + fs.set(BMI1) + fs.setIf((ebx&0x00000100) != 0, BMI2) + } + fs.setIf(ebx&(1<<2) != 0, SGX) + fs.setIf(ebx&(1<<4) != 0, HLE) + fs.setIf(ebx&(1<<9) != 0, ERMS) + fs.setIf(ebx&(1<<11) != 0, RTM) + fs.setIf(ebx&(1<<14) != 0, MPX) + fs.setIf(ebx&(1<<18) != 0, RDSEED) + fs.setIf(ebx&(1<<19) != 0, ADX) + fs.setIf(ebx&(1<<29) != 0, SHA) + + // CPUID.(EAX=7, ECX=0).ECX + fs.setIf(ecx&(1<<5) != 0, WAITPKG) + fs.setIf(ecx&(1<<7) != 0, CETSS) + fs.setIf(ecx&(1<<8) != 0, GFNI) + fs.setIf(ecx&(1<<9) != 0, VAES) + fs.setIf(ecx&(1<<10) != 0, VPCLMULQDQ) + fs.setIf(ecx&(1<<13) != 0, TME) + fs.setIf(ecx&(1<<25) != 0, CLDEMOTE) + fs.setIf(ecx&(1<<23) != 0, KEYLOCKER) + fs.setIf(ecx&(1<<27) != 0, MOVDIRI) + fs.setIf(ecx&(1<<28) != 0, MOVDIR64B) + fs.setIf(ecx&(1<<29) != 0, ENQCMD) + fs.setIf(ecx&(1<<30) != 0, SGXLC) + + // CPUID.(EAX=7, ECX=0).EDX + fs.setIf(edx&(1<<4) != 0, FSRM) + fs.setIf(edx&(1<<9) != 0, SRBDS_CTRL) + fs.setIf(edx&(1<<10) != 0, MD_CLEAR) + fs.setIf(edx&(1<<11) != 0, RTM_ALWAYS_ABORT) + fs.setIf(edx&(1<<14) != 0, SERIALIZE) + fs.setIf(edx&(1<<15) != 0, HYBRID_CPU) + fs.setIf(edx&(1<<16) != 0, TSXLDTRK) + fs.setIf(edx&(1<<18) != 0, PCONFIG) + fs.setIf(edx&(1<<20) != 0, CETIBT) + fs.setIf(edx&(1<<26) != 0, IBPB) + fs.setIf(edx&(1<<27) != 0, STIBP) + fs.setIf(edx&(1<<28) != 0, FLUSH_L1D) + fs.setIf(edx&(1<<29) != 0, IA32_ARCH_CAP) + fs.setIf(edx&(1<<30) != 0, IA32_CORE_CAP) + fs.setIf(edx&(1<<31) != 0, SPEC_CTRL_SSBD) + + // CPUID.(EAX=7, ECX=1).EAX + eax1, _, _, edx1 := cpuidex(7, 1) + fs.setIf(fs.inSet(AVX) && eax1&(1<<4) != 0, AVXVNNI) + fs.setIf(eax1&(1<<7) != 0, CMPCCXADD) + fs.setIf(eax1&(1<<10) != 0, MOVSB_ZL) + fs.setIf(eax1&(1<<11) != 0, STOSB_SHORT) + fs.setIf(eax1&(1<<12) != 0, CMPSB_SCADBS_SHORT) + fs.setIf(eax1&(1<<22) != 0, HRESET) + fs.setIf(eax1&(1<<23) != 0, AVXIFMA) + fs.setIf(eax1&(1<<26) != 0, LAM) + + // CPUID.(EAX=7, ECX=1).EDX + fs.setIf(edx1&(1<<4) != 0, AVXVNNIINT8) + fs.setIf(edx1&(1<<5) != 0, AVXNECONVERT) + fs.setIf(edx1&(1<<14) != 0, PREFETCHI) + fs.setIf(edx1&(1<<19) != 0, AVX10) + fs.setIf(edx1&(1<<21) != 0, APX_F) + + // Only detect AVX-512 features if XGETBV is supported + if c&((1<<26)|(1<<27)) == (1<<26)|(1<<27) { + // Check for OS support + eax, _ := xgetbv(0) + + // Verify that XCR0[7:5] = ‘111b’ (OPMASK state, upper 256-bit of ZMM0-ZMM15 and + // ZMM16-ZMM31 state are enabled by OS) + /// and that XCR0[2:1] = ‘11b’ (XMM state and YMM state are enabled by OS). + hasAVX512 := (eax>>5)&7 == 7 && (eax>>1)&3 == 3 + if runtime.GOOS == "darwin" { + hasAVX512 = fs.inSet(AVX) && darwinHasAVX512() + } + if hasAVX512 { + fs.setIf(ebx&(1<<16) != 0, AVX512F) + fs.setIf(ebx&(1<<17) != 0, AVX512DQ) + fs.setIf(ebx&(1<<21) != 0, AVX512IFMA) + fs.setIf(ebx&(1<<26) != 0, AVX512PF) + fs.setIf(ebx&(1<<27) != 0, AVX512ER) + fs.setIf(ebx&(1<<28) != 0, AVX512CD) + fs.setIf(ebx&(1<<30) != 0, AVX512BW) + fs.setIf(ebx&(1<<31) != 0, AVX512VL) + // ecx + fs.setIf(ecx&(1<<1) != 0, AVX512VBMI) + fs.setIf(ecx&(1<<6) != 0, AVX512VBMI2) + fs.setIf(ecx&(1<<11) != 0, AVX512VNNI) + fs.setIf(ecx&(1<<12) != 0, AVX512BITALG) + fs.setIf(ecx&(1<<14) != 0, AVX512VPOPCNTDQ) + // edx + fs.setIf(edx&(1<<8) != 0, AVX512VP2INTERSECT) + fs.setIf(edx&(1<<22) != 0, AMXBF16) + fs.setIf(edx&(1<<23) != 0, AVX512FP16) + fs.setIf(edx&(1<<24) != 0, AMXTILE) + fs.setIf(edx&(1<<25) != 0, AMXINT8) + // eax1 = CPUID.(EAX=7, ECX=1).EAX + fs.setIf(eax1&(1<<5) != 0, AVX512BF16) + fs.setIf(eax1&(1<<19) != 0, WRMSRNS) + fs.setIf(eax1&(1<<21) != 0, AMXFP16) + fs.setIf(eax1&(1<<27) != 0, MSRLIST) + } + } + + // CPUID.(EAX=7, ECX=2) + _, _, _, edx = cpuidex(7, 2) + fs.setIf(edx&(1<<0) != 0, PSFD) + fs.setIf(edx&(1<<1) != 0, IDPRED_CTRL) + fs.setIf(edx&(1<<2) != 0, RRSBA_CTRL) + fs.setIf(edx&(1<<4) != 0, BHI_CTRL) + fs.setIf(edx&(1<<5) != 0, MCDT_NO) + + // Add keylocker features. + if fs.inSet(KEYLOCKER) && mfi >= 0x19 { + _, ebx, _, _ := cpuidex(0x19, 0) + fs.setIf(ebx&5 == 5, KEYLOCKERW) // Bit 0 and 2 (1+4) + } + + // Add AVX10 features. + if fs.inSet(AVX10) && mfi >= 0x24 { + _, ebx, _, _ := cpuidex(0x24, 0) + fs.setIf(ebx&(1<<16) != 0, AVX10_128) + fs.setIf(ebx&(1<<17) != 0, AVX10_256) + fs.setIf(ebx&(1<<18) != 0, AVX10_512) + } + } + + // Processor Extended State Enumeration Sub-leaf (EAX = 0DH, ECX = 1) + // EAX + // Bit 00: XSAVEOPT is available. + // Bit 01: Supports XSAVEC and the compacted form of XRSTOR if set. + // Bit 02: Supports XGETBV with ECX = 1 if set. + // Bit 03: Supports XSAVES/XRSTORS and IA32_XSS if set. + // Bits 31 - 04: Reserved. + // EBX + // Bits 31 - 00: The size in bytes of the XSAVE area containing all states enabled by XCRO | IA32_XSS. + // ECX + // Bits 31 - 00: Reports the supported bits of the lower 32 bits of the IA32_XSS MSR. IA32_XSS[n] can be set to 1 only if ECX[n] is 1. + // EDX? + // Bits 07 - 00: Used for XCR0. Bit 08: PT state. Bit 09: Used for XCR0. Bits 12 - 10: Reserved. Bit 13: HWP state. Bits 31 - 14: Reserved. + if mfi >= 0xd { + if fs.inSet(XSAVE) { + eax, _, _, _ := cpuidex(0xd, 1) + fs.setIf(eax&(1<<0) != 0, XSAVEOPT) + fs.setIf(eax&(1<<1) != 0, XSAVEC) + fs.setIf(eax&(1<<2) != 0, XGETBV1) + fs.setIf(eax&(1<<3) != 0, XSAVES) + } + } + if maxExtendedFunction() >= 0x80000001 { + _, _, c, d := cpuid(0x80000001) + if (c & (1 << 5)) != 0 { + fs.set(LZCNT) + fs.set(POPCNT) + } + // ECX + fs.setIf((c&(1<<0)) != 0, LAHF) + fs.setIf((c&(1<<2)) != 0, SVM) + fs.setIf((c&(1<<6)) != 0, SSE4A) + fs.setIf((c&(1<<10)) != 0, IBS) + fs.setIf((c&(1<<22)) != 0, TOPEXT) + + // EDX + fs.setIf(d&(1<<11) != 0, SYSCALL) + fs.setIf(d&(1<<20) != 0, NX) + fs.setIf(d&(1<<22) != 0, MMXEXT) + fs.setIf(d&(1<<23) != 0, MMX) + fs.setIf(d&(1<<24) != 0, FXSR) + fs.setIf(d&(1<<25) != 0, FXSROPT) + fs.setIf(d&(1<<27) != 0, RDTSCP) + fs.setIf(d&(1<<30) != 0, AMD3DNOWEXT) + fs.setIf(d&(1<<31) != 0, AMD3DNOW) + + /* XOP and FMA4 use the AVX instruction coding scheme, so they can't be + * used unless the OS has AVX support. */ + if fs.inSet(AVX) { + fs.setIf((c&(1<<11)) != 0, XOP) + fs.setIf((c&(1<<16)) != 0, FMA4) + } + + } + if maxExtendedFunction() >= 0x80000007 { + _, b, _, d := cpuid(0x80000007) + fs.setIf((b&(1<<0)) != 0, MCAOVERFLOW) + fs.setIf((b&(1<<1)) != 0, SUCCOR) + fs.setIf((b&(1<<2)) != 0, HWA) + fs.setIf((d&(1<<9)) != 0, CPBOOST) + } + + if maxExtendedFunction() >= 0x80000008 { + _, b, _, _ := cpuid(0x80000008) + fs.setIf(b&(1<<28) != 0, PSFD) + fs.setIf(b&(1<<27) != 0, CPPC) + fs.setIf(b&(1<<24) != 0, SPEC_CTRL_SSBD) + fs.setIf(b&(1<<23) != 0, PPIN) + fs.setIf(b&(1<<21) != 0, TLB_FLUSH_NESTED) + fs.setIf(b&(1<<20) != 0, EFER_LMSLE_UNS) + fs.setIf(b&(1<<19) != 0, IBRS_PROVIDES_SMP) + fs.setIf(b&(1<<18) != 0, IBRS_PREFERRED) + fs.setIf(b&(1<<17) != 0, STIBP_ALWAYSON) + fs.setIf(b&(1<<15) != 0, STIBP) + fs.setIf(b&(1<<14) != 0, IBRS) + fs.setIf((b&(1<<13)) != 0, INT_WBINVD) + fs.setIf(b&(1<<12) != 0, IBPB) + fs.setIf((b&(1<<9)) != 0, WBNOINVD) + fs.setIf((b&(1<<8)) != 0, MCOMMIT) + fs.setIf((b&(1<<4)) != 0, RDPRU) + fs.setIf((b&(1<<3)) != 0, INVLPGB) + fs.setIf((b&(1<<1)) != 0, MSRIRC) + fs.setIf((b&(1<<0)) != 0, CLZERO) + } + + if fs.inSet(SVM) && maxExtendedFunction() >= 0x8000000A { + _, _, _, edx := cpuid(0x8000000A) + fs.setIf((edx>>0)&1 == 1, SVMNP) + fs.setIf((edx>>1)&1 == 1, LBRVIRT) + fs.setIf((edx>>2)&1 == 1, SVML) + fs.setIf((edx>>3)&1 == 1, NRIPS) + fs.setIf((edx>>4)&1 == 1, TSCRATEMSR) + fs.setIf((edx>>5)&1 == 1, VMCBCLEAN) + fs.setIf((edx>>6)&1 == 1, SVMFBASID) + fs.setIf((edx>>7)&1 == 1, SVMDA) + fs.setIf((edx>>10)&1 == 1, SVMPF) + fs.setIf((edx>>12)&1 == 1, SVMPFT) + } + + if maxExtendedFunction() >= 0x8000001a { + eax, _, _, _ := cpuid(0x8000001a) + fs.setIf((eax>>0)&1 == 1, FP128) + fs.setIf((eax>>1)&1 == 1, MOVU) + fs.setIf((eax>>2)&1 == 1, FP256) + } + + if maxExtendedFunction() >= 0x8000001b && fs.inSet(IBS) { + eax, _, _, _ := cpuid(0x8000001b) + fs.setIf((eax>>0)&1 == 1, IBSFFV) + fs.setIf((eax>>1)&1 == 1, IBSFETCHSAM) + fs.setIf((eax>>2)&1 == 1, IBSOPSAM) + fs.setIf((eax>>3)&1 == 1, IBSRDWROPCNT) + fs.setIf((eax>>4)&1 == 1, IBSOPCNT) + fs.setIf((eax>>5)&1 == 1, IBSBRNTRGT) + fs.setIf((eax>>6)&1 == 1, IBSOPCNTEXT) + fs.setIf((eax>>7)&1 == 1, IBSRIPINVALIDCHK) + fs.setIf((eax>>8)&1 == 1, IBS_OPFUSE) + fs.setIf((eax>>9)&1 == 1, IBS_FETCH_CTLX) + fs.setIf((eax>>10)&1 == 1, IBS_OPDATA4) // Doc says "Fixed,0. IBS op data 4 MSR supported", but assuming they mean 1. + fs.setIf((eax>>11)&1 == 1, IBS_ZEN4) + } + + if maxExtendedFunction() >= 0x8000001f && vend == AMD { + a, _, _, _ := cpuid(0x8000001f) + fs.setIf((a>>0)&1 == 1, SME) + fs.setIf((a>>1)&1 == 1, SEV) + fs.setIf((a>>2)&1 == 1, MSR_PAGEFLUSH) + fs.setIf((a>>3)&1 == 1, SEV_ES) + fs.setIf((a>>4)&1 == 1, SEV_SNP) + fs.setIf((a>>5)&1 == 1, VMPL) + fs.setIf((a>>10)&1 == 1, SME_COHERENT) + fs.setIf((a>>11)&1 == 1, SEV_64BIT) + fs.setIf((a>>12)&1 == 1, SEV_RESTRICTED) + fs.setIf((a>>13)&1 == 1, SEV_ALTERNATIVE) + fs.setIf((a>>14)&1 == 1, SEV_DEBUGSWAP) + fs.setIf((a>>15)&1 == 1, IBS_PREVENTHOST) + fs.setIf((a>>16)&1 == 1, VTE) + fs.setIf((a>>24)&1 == 1, VMSA_REGPROT) + } + + if mfi >= 0x20 { + // Microsoft has decided to purposefully hide the information + // of the guest TEE when VMs are being created using Hyper-V. + // + // This leads us to check for the Hyper-V cpuid features + // (0x4000000C), and then for the `ebx` value set. + // + // For Intel TDX, `ebx` is set as `0xbe3`, being 3 the part + // we're mostly interested about,according to: + // https://github.com/torvalds/linux/blob/d2f51b3516dade79269ff45eae2a7668ae711b25/arch/x86/include/asm/hyperv-tlfs.h#L169-L174 + _, ebx, _, _ := cpuid(0x4000000C) + fs.setIf(ebx == 0xbe3, TDX_GUEST) + } + + if mfi >= 0x21 { + // Intel Trusted Domain Extensions Guests have their own cpuid leaf (0x21). + _, ebx, ecx, edx := cpuid(0x21) + identity := string(valAsString(ebx, edx, ecx)) + fs.setIf(identity == "IntelTDX ", TDX_GUEST) + } + + return fs +} + +func (c *CPUInfo) supportAVX10() uint8 { + if c.maxFunc >= 0x24 && c.featureSet.inSet(AVX10) { + _, ebx, _, _ := cpuidex(0x24, 0) + return uint8(ebx) + } + return 0 +} + +func valAsString(values ...uint32) []byte { + r := make([]byte, 4*len(values)) + for i, v := range values { + dst := r[i*4:] + dst[0] = byte(v & 0xff) + dst[1] = byte((v >> 8) & 0xff) + dst[2] = byte((v >> 16) & 0xff) + dst[3] = byte((v >> 24) & 0xff) + switch { + case dst[0] == 0: + return r[:i*4] + case dst[1] == 0: + return r[:i*4+1] + case dst[2] == 0: + return r[:i*4+2] + case dst[3] == 0: + return r[:i*4+3] + } + } + return r +} diff --git a/vendor/github.com/klauspost/cpuid/v2/cpuid_386.s b/vendor/github.com/klauspost/cpuid/v2/cpuid_386.s new file mode 100644 index 0000000..8587c3a --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/cpuid_386.s @@ -0,0 +1,47 @@ +// Copyright (c) 2015 Klaus Post, released under MIT License. See LICENSE file. + +//+build 386,!gccgo,!noasm,!appengine + +// func asmCpuid(op uint32) (eax, ebx, ecx, edx uint32) +TEXT ·asmCpuid(SB), 7, $0 + XORL CX, CX + MOVL op+0(FP), AX + CPUID + MOVL AX, eax+4(FP) + MOVL BX, ebx+8(FP) + MOVL CX, ecx+12(FP) + MOVL DX, edx+16(FP) + RET + +// func asmCpuidex(op, op2 uint32) (eax, ebx, ecx, edx uint32) +TEXT ·asmCpuidex(SB), 7, $0 + MOVL op+0(FP), AX + MOVL op2+4(FP), CX + CPUID + MOVL AX, eax+8(FP) + MOVL BX, ebx+12(FP) + MOVL CX, ecx+16(FP) + MOVL DX, edx+20(FP) + RET + +// func xgetbv(index uint32) (eax, edx uint32) +TEXT ·asmXgetbv(SB), 7, $0 + MOVL index+0(FP), CX + BYTE $0x0f; BYTE $0x01; BYTE $0xd0 // XGETBV + MOVL AX, eax+4(FP) + MOVL DX, edx+8(FP) + RET + +// func asmRdtscpAsm() (eax, ebx, ecx, edx uint32) +TEXT ·asmRdtscpAsm(SB), 7, $0 + BYTE $0x0F; BYTE $0x01; BYTE $0xF9 // RDTSCP + MOVL AX, eax+0(FP) + MOVL BX, ebx+4(FP) + MOVL CX, ecx+8(FP) + MOVL DX, edx+12(FP) + RET + +// func asmDarwinHasAVX512() bool +TEXT ·asmDarwinHasAVX512(SB), 7, $0 + MOVL $0, eax+0(FP) + RET diff --git a/vendor/github.com/klauspost/cpuid/v2/cpuid_amd64.s b/vendor/github.com/klauspost/cpuid/v2/cpuid_amd64.s new file mode 100644 index 0000000..bc11f89 --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/cpuid_amd64.s @@ -0,0 +1,72 @@ +// Copyright (c) 2015 Klaus Post, released under MIT License. See LICENSE file. + +//+build amd64,!gccgo,!noasm,!appengine + +// func asmCpuid(op uint32) (eax, ebx, ecx, edx uint32) +TEXT ·asmCpuid(SB), 7, $0 + XORQ CX, CX + MOVL op+0(FP), AX + CPUID + MOVL AX, eax+8(FP) + MOVL BX, ebx+12(FP) + MOVL CX, ecx+16(FP) + MOVL DX, edx+20(FP) + RET + +// func asmCpuidex(op, op2 uint32) (eax, ebx, ecx, edx uint32) +TEXT ·asmCpuidex(SB), 7, $0 + MOVL op+0(FP), AX + MOVL op2+4(FP), CX + CPUID + MOVL AX, eax+8(FP) + MOVL BX, ebx+12(FP) + MOVL CX, ecx+16(FP) + MOVL DX, edx+20(FP) + RET + +// func asmXgetbv(index uint32) (eax, edx uint32) +TEXT ·asmXgetbv(SB), 7, $0 + MOVL index+0(FP), CX + BYTE $0x0f; BYTE $0x01; BYTE $0xd0 // XGETBV + MOVL AX, eax+8(FP) + MOVL DX, edx+12(FP) + RET + +// func asmRdtscpAsm() (eax, ebx, ecx, edx uint32) +TEXT ·asmRdtscpAsm(SB), 7, $0 + BYTE $0x0F; BYTE $0x01; BYTE $0xF9 // RDTSCP + MOVL AX, eax+0(FP) + MOVL BX, ebx+4(FP) + MOVL CX, ecx+8(FP) + MOVL DX, edx+12(FP) + RET + +// From https://go-review.googlesource.com/c/sys/+/285572/ +// func asmDarwinHasAVX512() bool +TEXT ·asmDarwinHasAVX512(SB), 7, $0-1 + MOVB $0, ret+0(FP) // default to false + +#ifdef GOOS_darwin // return if not darwin +#ifdef GOARCH_amd64 // return if not amd64 +// These values from: +// https://github.com/apple/darwin-xnu/blob/xnu-4570.1.46/osfmk/i386/cpu_capabilities.h +#define commpage64_base_address 0x00007fffffe00000 +#define commpage64_cpu_capabilities64 (commpage64_base_address+0x010) +#define commpage64_version (commpage64_base_address+0x01E) +#define hasAVX512F 0x0000004000000000 + MOVQ $commpage64_version, BX + MOVW (BX), AX + CMPW AX, $13 // versions < 13 do not support AVX512 + JL no_avx512 + MOVQ $commpage64_cpu_capabilities64, BX + MOVQ (BX), AX + MOVQ $hasAVX512F, CX + ANDQ CX, AX + JZ no_avx512 + MOVB $1, ret+0(FP) + +no_avx512: +#endif +#endif + RET + diff --git a/vendor/github.com/klauspost/cpuid/v2/cpuid_arm64.s b/vendor/github.com/klauspost/cpuid/v2/cpuid_arm64.s new file mode 100644 index 0000000..b31d6ae --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/cpuid_arm64.s @@ -0,0 +1,26 @@ +// Copyright (c) 2015 Klaus Post, released under MIT License. See LICENSE file. + +//+build arm64,!gccgo,!noasm,!appengine + +// See https://www.kernel.org/doc/Documentation/arm64/cpu-feature-registers.txt + +// func getMidr +TEXT ·getMidr(SB), 7, $0 + WORD $0xd5380000 // mrs x0, midr_el1 /* Main ID Register */ + MOVD R0, midr+0(FP) + RET + +// func getProcFeatures +TEXT ·getProcFeatures(SB), 7, $0 + WORD $0xd5380400 // mrs x0, id_aa64pfr0_el1 /* Processor Feature Register 0 */ + MOVD R0, procFeatures+0(FP) + RET + +// func getInstAttributes +TEXT ·getInstAttributes(SB), 7, $0 + WORD $0xd5380600 // mrs x0, id_aa64isar0_el1 /* Instruction Set Attribute Register 0 */ + WORD $0xd5380621 // mrs x1, id_aa64isar1_el1 /* Instruction Set Attribute Register 1 */ + MOVD R0, instAttrReg0+0(FP) + MOVD R1, instAttrReg1+8(FP) + RET + diff --git a/vendor/github.com/klauspost/cpuid/v2/detect_arm64.go b/vendor/github.com/klauspost/cpuid/v2/detect_arm64.go new file mode 100644 index 0000000..9a53504 --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/detect_arm64.go @@ -0,0 +1,247 @@ +// Copyright (c) 2015 Klaus Post, released under MIT License. See LICENSE file. + +//go:build arm64 && !gccgo && !noasm && !appengine +// +build arm64,!gccgo,!noasm,!appengine + +package cpuid + +import "runtime" + +func getMidr() (midr uint64) +func getProcFeatures() (procFeatures uint64) +func getInstAttributes() (instAttrReg0, instAttrReg1 uint64) + +func initCPU() { + cpuid = func(uint32) (a, b, c, d uint32) { return 0, 0, 0, 0 } + cpuidex = func(x, y uint32) (a, b, c, d uint32) { return 0, 0, 0, 0 } + xgetbv = func(uint32) (a, b uint32) { return 0, 0 } + rdtscpAsm = func() (a, b, c, d uint32) { return 0, 0, 0, 0 } +} + +func addInfo(c *CPUInfo, safe bool) { + // Seems to be safe to assume on ARM64 + c.CacheLine = 64 + detectOS(c) + + // ARM64 disabled since it may crash if interrupt is not intercepted by OS. + if safe && !c.Supports(ARMCPUID) && runtime.GOOS != "freebsd" { + return + } + midr := getMidr() + + // MIDR_EL1 - Main ID Register + // https://developer.arm.com/docs/ddi0595/h/aarch64-system-registers/midr_el1 + // x--------------------------------------------------x + // | Name | bits | visible | + // |--------------------------------------------------| + // | Implementer | [31-24] | y | + // |--------------------------------------------------| + // | Variant | [23-20] | y | + // |--------------------------------------------------| + // | Architecture | [19-16] | y | + // |--------------------------------------------------| + // | PartNum | [15-4] | y | + // |--------------------------------------------------| + // | Revision | [3-0] | y | + // x--------------------------------------------------x + + switch (midr >> 24) & 0xff { + case 0xC0: + c.VendorString = "Ampere Computing" + c.VendorID = Ampere + case 0x41: + c.VendorString = "Arm Limited" + c.VendorID = ARM + case 0x42: + c.VendorString = "Broadcom Corporation" + c.VendorID = Broadcom + case 0x43: + c.VendorString = "Cavium Inc" + c.VendorID = Cavium + case 0x44: + c.VendorString = "Digital Equipment Corporation" + c.VendorID = DEC + case 0x46: + c.VendorString = "Fujitsu Ltd" + c.VendorID = Fujitsu + case 0x49: + c.VendorString = "Infineon Technologies AG" + c.VendorID = Infineon + case 0x4D: + c.VendorString = "Motorola or Freescale Semiconductor Inc" + c.VendorID = Motorola + case 0x4E: + c.VendorString = "NVIDIA Corporation" + c.VendorID = NVIDIA + case 0x50: + c.VendorString = "Applied Micro Circuits Corporation" + c.VendorID = AMCC + case 0x51: + c.VendorString = "Qualcomm Inc" + c.VendorID = Qualcomm + case 0x56: + c.VendorString = "Marvell International Ltd" + c.VendorID = Marvell + case 0x69: + c.VendorString = "Intel Corporation" + c.VendorID = Intel + } + + // Lower 4 bits: Architecture + // Architecture Meaning + // 0b0001 Armv4. + // 0b0010 Armv4T. + // 0b0011 Armv5 (obsolete). + // 0b0100 Armv5T. + // 0b0101 Armv5TE. + // 0b0110 Armv5TEJ. + // 0b0111 Armv6. + // 0b1111 Architectural features are individually identified in the ID_* registers, see 'ID registers'. + // Upper 4 bit: Variant + // An IMPLEMENTATION DEFINED variant number. + // Typically, this field is used to distinguish between different product variants, or major revisions of a product. + c.Family = int(midr>>16) & 0xff + + // PartNum, bits [15:4] + // An IMPLEMENTATION DEFINED primary part number for the device. + // On processors implemented by Arm, if the top four bits of the primary + // part number are 0x0 or 0x7, the variant and architecture are encoded differently. + // Revision, bits [3:0] + // An IMPLEMENTATION DEFINED revision number for the device. + c.Model = int(midr) & 0xffff + + procFeatures := getProcFeatures() + + // ID_AA64PFR0_EL1 - Processor Feature Register 0 + // x--------------------------------------------------x + // | Name | bits | visible | + // |--------------------------------------------------| + // | DIT | [51-48] | y | + // |--------------------------------------------------| + // | SVE | [35-32] | y | + // |--------------------------------------------------| + // | GIC | [27-24] | n | + // |--------------------------------------------------| + // | AdvSIMD | [23-20] | y | + // |--------------------------------------------------| + // | FP | [19-16] | y | + // |--------------------------------------------------| + // | EL3 | [15-12] | n | + // |--------------------------------------------------| + // | EL2 | [11-8] | n | + // |--------------------------------------------------| + // | EL1 | [7-4] | n | + // |--------------------------------------------------| + // | EL0 | [3-0] | n | + // x--------------------------------------------------x + + var f flagSet + // if procFeatures&(0xf<<48) != 0 { + // fmt.Println("DIT") + // } + f.setIf(procFeatures&(0xf<<32) != 0, SVE) + if procFeatures&(0xf<<20) != 15<<20 { + f.set(ASIMD) + // https://developer.arm.com/docs/ddi0595/b/aarch64-system-registers/id_aa64pfr0_el1 + // 0b0001 --> As for 0b0000, and also includes support for half-precision floating-point arithmetic. + f.setIf(procFeatures&(0xf<<20) == 1<<20, FPHP, ASIMDHP) + } + f.setIf(procFeatures&(0xf<<16) != 0, FP) + + instAttrReg0, instAttrReg1 := getInstAttributes() + + // https://developer.arm.com/docs/ddi0595/b/aarch64-system-registers/id_aa64isar0_el1 + // + // ID_AA64ISAR0_EL1 - Instruction Set Attribute Register 0 + // x--------------------------------------------------x + // | Name | bits | visible | + // |--------------------------------------------------| + // | TS | [55-52] | y | + // |--------------------------------------------------| + // | FHM | [51-48] | y | + // |--------------------------------------------------| + // | DP | [47-44] | y | + // |--------------------------------------------------| + // | SM4 | [43-40] | y | + // |--------------------------------------------------| + // | SM3 | [39-36] | y | + // |--------------------------------------------------| + // | SHA3 | [35-32] | y | + // |--------------------------------------------------| + // | RDM | [31-28] | y | + // |--------------------------------------------------| + // | ATOMICS | [23-20] | y | + // |--------------------------------------------------| + // | CRC32 | [19-16] | y | + // |--------------------------------------------------| + // | SHA2 | [15-12] | y | + // |--------------------------------------------------| + // | SHA1 | [11-8] | y | + // |--------------------------------------------------| + // | AES | [7-4] | y | + // x--------------------------------------------------x + + // if instAttrReg0&(0xf<<52) != 0 { + // fmt.Println("TS") + // } + // if instAttrReg0&(0xf<<48) != 0 { + // fmt.Println("FHM") + // } + f.setIf(instAttrReg0&(0xf<<44) != 0, ASIMDDP) + f.setIf(instAttrReg0&(0xf<<40) != 0, SM4) + f.setIf(instAttrReg0&(0xf<<36) != 0, SM3) + f.setIf(instAttrReg0&(0xf<<32) != 0, SHA3) + f.setIf(instAttrReg0&(0xf<<28) != 0, ASIMDRDM) + f.setIf(instAttrReg0&(0xf<<20) != 0, ATOMICS) + f.setIf(instAttrReg0&(0xf<<16) != 0, CRC32) + f.setIf(instAttrReg0&(0xf<<12) != 0, SHA2) + // https://developer.arm.com/docs/ddi0595/b/aarch64-system-registers/id_aa64isar0_el1 + // 0b0010 --> As 0b0001, plus SHA512H, SHA512H2, SHA512SU0, and SHA512SU1 instructions implemented. + f.setIf(instAttrReg0&(0xf<<12) == 2<<12, SHA512) + f.setIf(instAttrReg0&(0xf<<8) != 0, SHA1) + f.setIf(instAttrReg0&(0xf<<4) != 0, AESARM) + // https://developer.arm.com/docs/ddi0595/b/aarch64-system-registers/id_aa64isar0_el1 + // 0b0010 --> As for 0b0001, plus PMULL/PMULL2 instructions operating on 64-bit data quantities. + f.setIf(instAttrReg0&(0xf<<4) == 2<<4, PMULL) + + // https://developer.arm.com/docs/ddi0595/b/aarch64-system-registers/id_aa64isar1_el1 + // + // ID_AA64ISAR1_EL1 - Instruction set attribute register 1 + // x--------------------------------------------------x + // | Name | bits | visible | + // |--------------------------------------------------| + // | GPI | [31-28] | y | + // |--------------------------------------------------| + // | GPA | [27-24] | y | + // |--------------------------------------------------| + // | LRCPC | [23-20] | y | + // |--------------------------------------------------| + // | FCMA | [19-16] | y | + // |--------------------------------------------------| + // | JSCVT | [15-12] | y | + // |--------------------------------------------------| + // | API | [11-8] | y | + // |--------------------------------------------------| + // | APA | [7-4] | y | + // |--------------------------------------------------| + // | DPB | [3-0] | y | + // x--------------------------------------------------x + + // if instAttrReg1&(0xf<<28) != 0 { + // fmt.Println("GPI") + // } + f.setIf(instAttrReg1&(0xf<<28) != 24, GPA) + f.setIf(instAttrReg1&(0xf<<20) != 0, LRCPC) + f.setIf(instAttrReg1&(0xf<<16) != 0, FCMA) + f.setIf(instAttrReg1&(0xf<<12) != 0, JSCVT) + // if instAttrReg1&(0xf<<8) != 0 { + // fmt.Println("API") + // } + // if instAttrReg1&(0xf<<4) != 0 { + // fmt.Println("APA") + // } + f.setIf(instAttrReg1&(0xf<<0) != 0, DCPOP) + + // Store + c.featureSet.or(f) +} diff --git a/vendor/github.com/klauspost/cpuid/v2/detect_ref.go b/vendor/github.com/klauspost/cpuid/v2/detect_ref.go new file mode 100644 index 0000000..9636c2b --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/detect_ref.go @@ -0,0 +1,15 @@ +// Copyright (c) 2015 Klaus Post, released under MIT License. See LICENSE file. + +//go:build (!amd64 && !386 && !arm64) || gccgo || noasm || appengine +// +build !amd64,!386,!arm64 gccgo noasm appengine + +package cpuid + +func initCPU() { + cpuid = func(uint32) (a, b, c, d uint32) { return 0, 0, 0, 0 } + cpuidex = func(x, y uint32) (a, b, c, d uint32) { return 0, 0, 0, 0 } + xgetbv = func(uint32) (a, b uint32) { return 0, 0 } + rdtscpAsm = func() (a, b, c, d uint32) { return 0, 0, 0, 0 } +} + +func addInfo(info *CPUInfo, safe bool) {} diff --git a/vendor/github.com/klauspost/cpuid/v2/detect_x86.go b/vendor/github.com/klauspost/cpuid/v2/detect_x86.go new file mode 100644 index 0000000..c7dfa12 --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/detect_x86.go @@ -0,0 +1,37 @@ +// Copyright (c) 2015 Klaus Post, released under MIT License. See LICENSE file. + +//go:build (386 && !gccgo && !noasm && !appengine) || (amd64 && !gccgo && !noasm && !appengine) +// +build 386,!gccgo,!noasm,!appengine amd64,!gccgo,!noasm,!appengine + +package cpuid + +func asmCpuid(op uint32) (eax, ebx, ecx, edx uint32) +func asmCpuidex(op, op2 uint32) (eax, ebx, ecx, edx uint32) +func asmXgetbv(index uint32) (eax, edx uint32) +func asmRdtscpAsm() (eax, ebx, ecx, edx uint32) +func asmDarwinHasAVX512() bool + +func initCPU() { + cpuid = asmCpuid + cpuidex = asmCpuidex + xgetbv = asmXgetbv + rdtscpAsm = asmRdtscpAsm + darwinHasAVX512 = asmDarwinHasAVX512 +} + +func addInfo(c *CPUInfo, safe bool) { + c.maxFunc = maxFunctionID() + c.maxExFunc = maxExtendedFunction() + c.BrandName = brandName() + c.CacheLine = cacheLine() + c.Family, c.Model, c.Stepping = familyModel() + c.featureSet = support() + c.SGX = hasSGX(c.featureSet.inSet(SGX), c.featureSet.inSet(SGXLC)) + c.ThreadsPerCore = threadsPerCore() + c.LogicalCores = logicalCores() + c.PhysicalCores = physicalCores() + c.VendorID, c.VendorString = vendorID() + c.AVX10Level = c.supportAVX10() + c.cacheSize() + c.frequencies() +} diff --git a/vendor/github.com/klauspost/cpuid/v2/featureid_string.go b/vendor/github.com/klauspost/cpuid/v2/featureid_string.go new file mode 100644 index 0000000..43bd05f --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/featureid_string.go @@ -0,0 +1,279 @@ +// Code generated by "stringer -type=FeatureID,Vendor"; DO NOT EDIT. + +package cpuid + +import "strconv" + +func _() { + // An "invalid array index" compiler error signifies that the constant values have changed. + // Re-run the stringer command to generate them again. + var x [1]struct{} + _ = x[ADX-1] + _ = x[AESNI-2] + _ = x[AMD3DNOW-3] + _ = x[AMD3DNOWEXT-4] + _ = x[AMXBF16-5] + _ = x[AMXFP16-6] + _ = x[AMXINT8-7] + _ = x[AMXTILE-8] + _ = x[APX_F-9] + _ = x[AVX-10] + _ = x[AVX10-11] + _ = x[AVX10_128-12] + _ = x[AVX10_256-13] + _ = x[AVX10_512-14] + _ = x[AVX2-15] + _ = x[AVX512BF16-16] + _ = x[AVX512BITALG-17] + _ = x[AVX512BW-18] + _ = x[AVX512CD-19] + _ = x[AVX512DQ-20] + _ = x[AVX512ER-21] + _ = x[AVX512F-22] + _ = x[AVX512FP16-23] + _ = x[AVX512IFMA-24] + _ = x[AVX512PF-25] + _ = x[AVX512VBMI-26] + _ = x[AVX512VBMI2-27] + _ = x[AVX512VL-28] + _ = x[AVX512VNNI-29] + _ = x[AVX512VP2INTERSECT-30] + _ = x[AVX512VPOPCNTDQ-31] + _ = x[AVXIFMA-32] + _ = x[AVXNECONVERT-33] + _ = x[AVXSLOW-34] + _ = x[AVXVNNI-35] + _ = x[AVXVNNIINT8-36] + _ = x[BHI_CTRL-37] + _ = x[BMI1-38] + _ = x[BMI2-39] + _ = x[CETIBT-40] + _ = x[CETSS-41] + _ = x[CLDEMOTE-42] + _ = x[CLMUL-43] + _ = x[CLZERO-44] + _ = x[CMOV-45] + _ = x[CMPCCXADD-46] + _ = x[CMPSB_SCADBS_SHORT-47] + _ = x[CMPXCHG8-48] + _ = x[CPBOOST-49] + _ = x[CPPC-50] + _ = x[CX16-51] + _ = x[EFER_LMSLE_UNS-52] + _ = x[ENQCMD-53] + _ = x[ERMS-54] + _ = x[F16C-55] + _ = x[FLUSH_L1D-56] + _ = x[FMA3-57] + _ = x[FMA4-58] + _ = x[FP128-59] + _ = x[FP256-60] + _ = x[FSRM-61] + _ = x[FXSR-62] + _ = x[FXSROPT-63] + _ = x[GFNI-64] + _ = x[HLE-65] + _ = x[HRESET-66] + _ = x[HTT-67] + _ = x[HWA-68] + _ = x[HYBRID_CPU-69] + _ = x[HYPERVISOR-70] + _ = x[IA32_ARCH_CAP-71] + _ = x[IA32_CORE_CAP-72] + _ = x[IBPB-73] + _ = x[IBRS-74] + _ = x[IBRS_PREFERRED-75] + _ = x[IBRS_PROVIDES_SMP-76] + _ = x[IBS-77] + _ = x[IBSBRNTRGT-78] + _ = x[IBSFETCHSAM-79] + _ = x[IBSFFV-80] + _ = x[IBSOPCNT-81] + _ = x[IBSOPCNTEXT-82] + _ = x[IBSOPSAM-83] + _ = x[IBSRDWROPCNT-84] + _ = x[IBSRIPINVALIDCHK-85] + _ = x[IBS_FETCH_CTLX-86] + _ = x[IBS_OPDATA4-87] + _ = x[IBS_OPFUSE-88] + _ = x[IBS_PREVENTHOST-89] + _ = x[IBS_ZEN4-90] + _ = x[IDPRED_CTRL-91] + _ = x[INT_WBINVD-92] + _ = x[INVLPGB-93] + _ = x[KEYLOCKER-94] + _ = x[KEYLOCKERW-95] + _ = x[LAHF-96] + _ = x[LAM-97] + _ = x[LBRVIRT-98] + _ = x[LZCNT-99] + _ = x[MCAOVERFLOW-100] + _ = x[MCDT_NO-101] + _ = x[MCOMMIT-102] + _ = x[MD_CLEAR-103] + _ = x[MMX-104] + _ = x[MMXEXT-105] + _ = x[MOVBE-106] + _ = x[MOVDIR64B-107] + _ = x[MOVDIRI-108] + _ = x[MOVSB_ZL-109] + _ = x[MOVU-110] + _ = x[MPX-111] + _ = x[MSRIRC-112] + _ = x[MSRLIST-113] + _ = x[MSR_PAGEFLUSH-114] + _ = x[NRIPS-115] + _ = x[NX-116] + _ = x[OSXSAVE-117] + _ = x[PCONFIG-118] + _ = x[POPCNT-119] + _ = x[PPIN-120] + _ = x[PREFETCHI-121] + _ = x[PSFD-122] + _ = x[RDPRU-123] + _ = x[RDRAND-124] + _ = x[RDSEED-125] + _ = x[RDTSCP-126] + _ = x[RRSBA_CTRL-127] + _ = x[RTM-128] + _ = x[RTM_ALWAYS_ABORT-129] + _ = x[SERIALIZE-130] + _ = x[SEV-131] + _ = x[SEV_64BIT-132] + _ = x[SEV_ALTERNATIVE-133] + _ = x[SEV_DEBUGSWAP-134] + _ = x[SEV_ES-135] + _ = x[SEV_RESTRICTED-136] + _ = x[SEV_SNP-137] + _ = x[SGX-138] + _ = x[SGXLC-139] + _ = x[SHA-140] + _ = x[SME-141] + _ = x[SME_COHERENT-142] + _ = x[SPEC_CTRL_SSBD-143] + _ = x[SRBDS_CTRL-144] + _ = x[SSE-145] + _ = x[SSE2-146] + _ = x[SSE3-147] + _ = x[SSE4-148] + _ = x[SSE42-149] + _ = x[SSE4A-150] + _ = x[SSSE3-151] + _ = x[STIBP-152] + _ = x[STIBP_ALWAYSON-153] + _ = x[STOSB_SHORT-154] + _ = x[SUCCOR-155] + _ = x[SVM-156] + _ = x[SVMDA-157] + _ = x[SVMFBASID-158] + _ = x[SVML-159] + _ = x[SVMNP-160] + _ = x[SVMPF-161] + _ = x[SVMPFT-162] + _ = x[SYSCALL-163] + _ = x[SYSEE-164] + _ = x[TBM-165] + _ = x[TDX_GUEST-166] + _ = x[TLB_FLUSH_NESTED-167] + _ = x[TME-168] + _ = x[TOPEXT-169] + _ = x[TSCRATEMSR-170] + _ = x[TSXLDTRK-171] + _ = x[VAES-172] + _ = x[VMCBCLEAN-173] + _ = x[VMPL-174] + _ = x[VMSA_REGPROT-175] + _ = x[VMX-176] + _ = x[VPCLMULQDQ-177] + _ = x[VTE-178] + _ = x[WAITPKG-179] + _ = x[WBNOINVD-180] + _ = x[WRMSRNS-181] + _ = x[X87-182] + _ = x[XGETBV1-183] + _ = x[XOP-184] + _ = x[XSAVE-185] + _ = x[XSAVEC-186] + _ = x[XSAVEOPT-187] + _ = x[XSAVES-188] + _ = x[AESARM-189] + _ = x[ARMCPUID-190] + _ = x[ASIMD-191] + _ = x[ASIMDDP-192] + _ = x[ASIMDHP-193] + _ = x[ASIMDRDM-194] + _ = x[ATOMICS-195] + _ = x[CRC32-196] + _ = x[DCPOP-197] + _ = x[EVTSTRM-198] + _ = x[FCMA-199] + _ = x[FP-200] + _ = x[FPHP-201] + _ = x[GPA-202] + _ = x[JSCVT-203] + _ = x[LRCPC-204] + _ = x[PMULL-205] + _ = x[SHA1-206] + _ = x[SHA2-207] + _ = x[SHA3-208] + _ = x[SHA512-209] + _ = x[SM3-210] + _ = x[SM4-211] + _ = x[SVE-212] + _ = x[lastID-213] + _ = x[firstID-0] +} + +const _FeatureID_name = "firstIDADXAESNIAMD3DNOWAMD3DNOWEXTAMXBF16AMXFP16AMXINT8AMXTILEAPX_FAVXAVX10AVX10_128AVX10_256AVX10_512AVX2AVX512BF16AVX512BITALGAVX512BWAVX512CDAVX512DQAVX512ERAVX512FAVX512FP16AVX512IFMAAVX512PFAVX512VBMIAVX512VBMI2AVX512VLAVX512VNNIAVX512VP2INTERSECTAVX512VPOPCNTDQAVXIFMAAVXNECONVERTAVXSLOWAVXVNNIAVXVNNIINT8BHI_CTRLBMI1BMI2CETIBTCETSSCLDEMOTECLMULCLZEROCMOVCMPCCXADDCMPSB_SCADBS_SHORTCMPXCHG8CPBOOSTCPPCCX16EFER_LMSLE_UNSENQCMDERMSF16CFLUSH_L1DFMA3FMA4FP128FP256FSRMFXSRFXSROPTGFNIHLEHRESETHTTHWAHYBRID_CPUHYPERVISORIA32_ARCH_CAPIA32_CORE_CAPIBPBIBRSIBRS_PREFERREDIBRS_PROVIDES_SMPIBSIBSBRNTRGTIBSFETCHSAMIBSFFVIBSOPCNTIBSOPCNTEXTIBSOPSAMIBSRDWROPCNTIBSRIPINVALIDCHKIBS_FETCH_CTLXIBS_OPDATA4IBS_OPFUSEIBS_PREVENTHOSTIBS_ZEN4IDPRED_CTRLINT_WBINVDINVLPGBKEYLOCKERKEYLOCKERWLAHFLAMLBRVIRTLZCNTMCAOVERFLOWMCDT_NOMCOMMITMD_CLEARMMXMMXEXTMOVBEMOVDIR64BMOVDIRIMOVSB_ZLMOVUMPXMSRIRCMSRLISTMSR_PAGEFLUSHNRIPSNXOSXSAVEPCONFIGPOPCNTPPINPREFETCHIPSFDRDPRURDRANDRDSEEDRDTSCPRRSBA_CTRLRTMRTM_ALWAYS_ABORTSERIALIZESEVSEV_64BITSEV_ALTERNATIVESEV_DEBUGSWAPSEV_ESSEV_RESTRICTEDSEV_SNPSGXSGXLCSHASMESME_COHERENTSPEC_CTRL_SSBDSRBDS_CTRLSSESSE2SSE3SSE4SSE42SSE4ASSSE3STIBPSTIBP_ALWAYSONSTOSB_SHORTSUCCORSVMSVMDASVMFBASIDSVMLSVMNPSVMPFSVMPFTSYSCALLSYSEETBMTDX_GUESTTLB_FLUSH_NESTEDTMETOPEXTTSCRATEMSRTSXLDTRKVAESVMCBCLEANVMPLVMSA_REGPROTVMXVPCLMULQDQVTEWAITPKGWBNOINVDWRMSRNSX87XGETBV1XOPXSAVEXSAVECXSAVEOPTXSAVESAESARMARMCPUIDASIMDASIMDDPASIMDHPASIMDRDMATOMICSCRC32DCPOPEVTSTRMFCMAFPFPHPGPAJSCVTLRCPCPMULLSHA1SHA2SHA3SHA512SM3SM4SVElastID" + +var _FeatureID_index = [...]uint16{0, 7, 10, 15, 23, 34, 41, 48, 55, 62, 67, 70, 75, 84, 93, 102, 106, 116, 128, 136, 144, 152, 160, 167, 177, 187, 195, 205, 216, 224, 234, 252, 267, 274, 286, 293, 300, 311, 319, 323, 327, 333, 338, 346, 351, 357, 361, 370, 388, 396, 403, 407, 411, 425, 431, 435, 439, 448, 452, 456, 461, 466, 470, 474, 481, 485, 488, 494, 497, 500, 510, 520, 533, 546, 550, 554, 568, 585, 588, 598, 609, 615, 623, 634, 642, 654, 670, 684, 695, 705, 720, 728, 739, 749, 756, 765, 775, 779, 782, 789, 794, 805, 812, 819, 827, 830, 836, 841, 850, 857, 865, 869, 872, 878, 885, 898, 903, 905, 912, 919, 925, 929, 938, 942, 947, 953, 959, 965, 975, 978, 994, 1003, 1006, 1015, 1030, 1043, 1049, 1063, 1070, 1073, 1078, 1081, 1084, 1096, 1110, 1120, 1123, 1127, 1131, 1135, 1140, 1145, 1150, 1155, 1169, 1180, 1186, 1189, 1194, 1203, 1207, 1212, 1217, 1223, 1230, 1235, 1238, 1247, 1263, 1266, 1272, 1282, 1290, 1294, 1303, 1307, 1319, 1322, 1332, 1335, 1342, 1350, 1357, 1360, 1367, 1370, 1375, 1381, 1389, 1395, 1401, 1409, 1414, 1421, 1428, 1436, 1443, 1448, 1453, 1460, 1464, 1466, 1470, 1473, 1478, 1483, 1488, 1492, 1496, 1500, 1506, 1509, 1512, 1515, 1521} + +func (i FeatureID) String() string { + if i < 0 || i >= FeatureID(len(_FeatureID_index)-1) { + return "FeatureID(" + strconv.FormatInt(int64(i), 10) + ")" + } + return _FeatureID_name[_FeatureID_index[i]:_FeatureID_index[i+1]] +} +func _() { + // An "invalid array index" compiler error signifies that the constant values have changed. + // Re-run the stringer command to generate them again. + var x [1]struct{} + _ = x[VendorUnknown-0] + _ = x[Intel-1] + _ = x[AMD-2] + _ = x[VIA-3] + _ = x[Transmeta-4] + _ = x[NSC-5] + _ = x[KVM-6] + _ = x[MSVM-7] + _ = x[VMware-8] + _ = x[XenHVM-9] + _ = x[Bhyve-10] + _ = x[Hygon-11] + _ = x[SiS-12] + _ = x[RDC-13] + _ = x[Ampere-14] + _ = x[ARM-15] + _ = x[Broadcom-16] + _ = x[Cavium-17] + _ = x[DEC-18] + _ = x[Fujitsu-19] + _ = x[Infineon-20] + _ = x[Motorola-21] + _ = x[NVIDIA-22] + _ = x[AMCC-23] + _ = x[Qualcomm-24] + _ = x[Marvell-25] + _ = x[lastVendor-26] +} + +const _Vendor_name = "VendorUnknownIntelAMDVIATransmetaNSCKVMMSVMVMwareXenHVMBhyveHygonSiSRDCAmpereARMBroadcomCaviumDECFujitsuInfineonMotorolaNVIDIAAMCCQualcommMarvelllastVendor" + +var _Vendor_index = [...]uint8{0, 13, 18, 21, 24, 33, 36, 39, 43, 49, 55, 60, 65, 68, 71, 77, 80, 88, 94, 97, 104, 112, 120, 126, 130, 138, 145, 155} + +func (i Vendor) String() string { + if i < 0 || i >= Vendor(len(_Vendor_index)-1) { + return "Vendor(" + strconv.FormatInt(int64(i), 10) + ")" + } + return _Vendor_name[_Vendor_index[i]:_Vendor_index[i+1]] +} diff --git a/vendor/github.com/klauspost/cpuid/v2/os_darwin_arm64.go b/vendor/github.com/klauspost/cpuid/v2/os_darwin_arm64.go new file mode 100644 index 0000000..84b1acd --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/os_darwin_arm64.go @@ -0,0 +1,121 @@ +// Copyright (c) 2020 Klaus Post, released under MIT License. See LICENSE file. + +package cpuid + +import ( + "runtime" + "strings" + + "golang.org/x/sys/unix" +) + +func detectOS(c *CPUInfo) bool { + if runtime.GOOS != "ios" { + tryToFillCPUInfoFomSysctl(c) + } + // There are no hw.optional sysctl values for the below features on Mac OS 11.0 + // to detect their supported state dynamically. Assume the CPU features that + // Apple Silicon M1 supports to be available as a minimal set of features + // to all Go programs running on darwin/arm64. + // TODO: Add more if we know them. + c.featureSet.setIf(runtime.GOOS != "ios", AESARM, PMULL, SHA1, SHA2) + + return true +} + +func sysctlGetBool(name string) bool { + value, err := unix.SysctlUint32(name) + if err != nil { + return false + } + return value != 0 +} + +func sysctlGetString(name string) string { + value, err := unix.Sysctl(name) + if err != nil { + return "" + } + return value +} + +func sysctlGetInt(unknown int, names ...string) int { + for _, name := range names { + value, err := unix.SysctlUint32(name) + if err != nil { + continue + } + if value != 0 { + return int(value) + } + } + return unknown +} + +func sysctlGetInt64(unknown int, names ...string) int { + for _, name := range names { + value64, err := unix.SysctlUint64(name) + if err != nil { + continue + } + if int(value64) != unknown { + return int(value64) + } + } + return unknown +} + +func setFeature(c *CPUInfo, name string, feature FeatureID) { + c.featureSet.setIf(sysctlGetBool(name), feature) +} +func tryToFillCPUInfoFomSysctl(c *CPUInfo) { + c.BrandName = sysctlGetString("machdep.cpu.brand_string") + + if len(c.BrandName) != 0 { + c.VendorString = strings.Fields(c.BrandName)[0] + } + + c.PhysicalCores = sysctlGetInt(runtime.NumCPU(), "hw.physicalcpu") + c.ThreadsPerCore = sysctlGetInt(1, "machdep.cpu.thread_count", "kern.num_threads") / + sysctlGetInt(1, "hw.physicalcpu") + c.LogicalCores = sysctlGetInt(runtime.NumCPU(), "machdep.cpu.core_count") + c.Family = sysctlGetInt(0, "machdep.cpu.family", "hw.cpufamily") + c.Model = sysctlGetInt(0, "machdep.cpu.model") + c.CacheLine = sysctlGetInt64(0, "hw.cachelinesize") + c.Cache.L1I = sysctlGetInt64(-1, "hw.l1icachesize") + c.Cache.L1D = sysctlGetInt64(-1, "hw.l1dcachesize") + c.Cache.L2 = sysctlGetInt64(-1, "hw.l2cachesize") + c.Cache.L3 = sysctlGetInt64(-1, "hw.l3cachesize") + + // from https://developer.arm.com/downloads/-/exploration-tools/feature-names-for-a-profile + setFeature(c, "hw.optional.arm.FEAT_AES", AESARM) + setFeature(c, "hw.optional.AdvSIMD", ASIMD) + setFeature(c, "hw.optional.arm.FEAT_DotProd", ASIMDDP) + setFeature(c, "hw.optional.arm.FEAT_RDM", ASIMDRDM) + setFeature(c, "hw.optional.FEAT_CRC32", CRC32) + setFeature(c, "hw.optional.arm.FEAT_DPB", DCPOP) + // setFeature(c, "", EVTSTRM) + setFeature(c, "hw.optional.arm.FEAT_FCMA", FCMA) + setFeature(c, "hw.optional.arm.FEAT_FP", FP) + setFeature(c, "hw.optional.arm.FEAT_FP16", FPHP) + setFeature(c, "hw.optional.arm.FEAT_PAuth", GPA) + setFeature(c, "hw.optional.arm.FEAT_JSCVT", JSCVT) + setFeature(c, "hw.optional.arm.FEAT_LRCPC", LRCPC) + setFeature(c, "hw.optional.arm.FEAT_PMULL", PMULL) + setFeature(c, "hw.optional.arm.FEAT_SHA1", SHA1) + setFeature(c, "hw.optional.arm.FEAT_SHA256", SHA2) + setFeature(c, "hw.optional.arm.FEAT_SHA3", SHA3) + setFeature(c, "hw.optional.arm.FEAT_SHA512", SHA512) + // setFeature(c, "", SM3) + // setFeature(c, "", SM4) + setFeature(c, "hw.optional.arm.FEAT_SVE", SVE) + + // from empirical observation + setFeature(c, "hw.optional.AdvSIMD_HPFPCvt", ASIMDHP) + setFeature(c, "hw.optional.armv8_1_atomics", ATOMICS) + setFeature(c, "hw.optional.floatingpoint", FP) + setFeature(c, "hw.optional.armv8_2_sha3", SHA3) + setFeature(c, "hw.optional.armv8_2_sha512", SHA512) + setFeature(c, "hw.optional.armv8_3_compnum", FCMA) + setFeature(c, "hw.optional.armv8_crc32", CRC32) +} diff --git a/vendor/github.com/klauspost/cpuid/v2/os_linux_arm64.go b/vendor/github.com/klauspost/cpuid/v2/os_linux_arm64.go new file mode 100644 index 0000000..ee278b9 --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/os_linux_arm64.go @@ -0,0 +1,130 @@ +// Copyright (c) 2020 Klaus Post, released under MIT License. See LICENSE file. + +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file located +// here https://github.com/golang/sys/blob/master/LICENSE + +package cpuid + +import ( + "encoding/binary" + "io/ioutil" + "runtime" +) + +// HWCAP bits. +const ( + hwcap_FP = 1 << 0 + hwcap_ASIMD = 1 << 1 + hwcap_EVTSTRM = 1 << 2 + hwcap_AES = 1 << 3 + hwcap_PMULL = 1 << 4 + hwcap_SHA1 = 1 << 5 + hwcap_SHA2 = 1 << 6 + hwcap_CRC32 = 1 << 7 + hwcap_ATOMICS = 1 << 8 + hwcap_FPHP = 1 << 9 + hwcap_ASIMDHP = 1 << 10 + hwcap_CPUID = 1 << 11 + hwcap_ASIMDRDM = 1 << 12 + hwcap_JSCVT = 1 << 13 + hwcap_FCMA = 1 << 14 + hwcap_LRCPC = 1 << 15 + hwcap_DCPOP = 1 << 16 + hwcap_SHA3 = 1 << 17 + hwcap_SM3 = 1 << 18 + hwcap_SM4 = 1 << 19 + hwcap_ASIMDDP = 1 << 20 + hwcap_SHA512 = 1 << 21 + hwcap_SVE = 1 << 22 + hwcap_ASIMDFHM = 1 << 23 +) + +func detectOS(c *CPUInfo) bool { + // For now assuming no hyperthreading is reasonable. + c.LogicalCores = runtime.NumCPU() + c.PhysicalCores = c.LogicalCores + c.ThreadsPerCore = 1 + if hwcap == 0 { + // We did not get values from the runtime. + // Try reading /proc/self/auxv + + // From https://github.com/golang/sys + const ( + _AT_HWCAP = 16 + _AT_HWCAP2 = 26 + + uintSize = int(32 << (^uint(0) >> 63)) + ) + + buf, err := ioutil.ReadFile("/proc/self/auxv") + if err != nil { + // e.g. on android /proc/self/auxv is not accessible, so silently + // ignore the error and leave Initialized = false. On some + // architectures (e.g. arm64) doinit() implements a fallback + // readout and will set Initialized = true again. + return false + } + bo := binary.LittleEndian + for len(buf) >= 2*(uintSize/8) { + var tag, val uint + switch uintSize { + case 32: + tag = uint(bo.Uint32(buf[0:])) + val = uint(bo.Uint32(buf[4:])) + buf = buf[8:] + case 64: + tag = uint(bo.Uint64(buf[0:])) + val = uint(bo.Uint64(buf[8:])) + buf = buf[16:] + } + switch tag { + case _AT_HWCAP: + hwcap = val + case _AT_HWCAP2: + // Not used + } + } + if hwcap == 0 { + return false + } + } + + // HWCap was populated by the runtime from the auxiliary vector. + // Use HWCap information since reading aarch64 system registers + // is not supported in user space on older linux kernels. + c.featureSet.setIf(isSet(hwcap, hwcap_AES), AESARM) + c.featureSet.setIf(isSet(hwcap, hwcap_ASIMD), ASIMD) + c.featureSet.setIf(isSet(hwcap, hwcap_ASIMDDP), ASIMDDP) + c.featureSet.setIf(isSet(hwcap, hwcap_ASIMDHP), ASIMDHP) + c.featureSet.setIf(isSet(hwcap, hwcap_ASIMDRDM), ASIMDRDM) + c.featureSet.setIf(isSet(hwcap, hwcap_CPUID), ARMCPUID) + c.featureSet.setIf(isSet(hwcap, hwcap_CRC32), CRC32) + c.featureSet.setIf(isSet(hwcap, hwcap_DCPOP), DCPOP) + c.featureSet.setIf(isSet(hwcap, hwcap_EVTSTRM), EVTSTRM) + c.featureSet.setIf(isSet(hwcap, hwcap_FCMA), FCMA) + c.featureSet.setIf(isSet(hwcap, hwcap_FP), FP) + c.featureSet.setIf(isSet(hwcap, hwcap_FPHP), FPHP) + c.featureSet.setIf(isSet(hwcap, hwcap_JSCVT), JSCVT) + c.featureSet.setIf(isSet(hwcap, hwcap_LRCPC), LRCPC) + c.featureSet.setIf(isSet(hwcap, hwcap_PMULL), PMULL) + c.featureSet.setIf(isSet(hwcap, hwcap_SHA1), SHA1) + c.featureSet.setIf(isSet(hwcap, hwcap_SHA2), SHA2) + c.featureSet.setIf(isSet(hwcap, hwcap_SHA3), SHA3) + c.featureSet.setIf(isSet(hwcap, hwcap_SHA512), SHA512) + c.featureSet.setIf(isSet(hwcap, hwcap_SM3), SM3) + c.featureSet.setIf(isSet(hwcap, hwcap_SM4), SM4) + c.featureSet.setIf(isSet(hwcap, hwcap_SVE), SVE) + + // The Samsung S9+ kernel reports support for atomics, but not all cores + // actually support them, resulting in SIGILL. See issue #28431. + // TODO(elias.naur): Only disable the optimization on bad chipsets on android. + c.featureSet.setIf(isSet(hwcap, hwcap_ATOMICS) && runtime.GOOS != "android", ATOMICS) + + return true +} + +func isSet(hwc uint, value uint) bool { + return hwc&value != 0 +} diff --git a/vendor/github.com/klauspost/cpuid/v2/os_other_arm64.go b/vendor/github.com/klauspost/cpuid/v2/os_other_arm64.go new file mode 100644 index 0000000..8733ba3 --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/os_other_arm64.go @@ -0,0 +1,16 @@ +// Copyright (c) 2020 Klaus Post, released under MIT License. See LICENSE file. + +//go:build arm64 && !linux && !darwin +// +build arm64,!linux,!darwin + +package cpuid + +import "runtime" + +func detectOS(c *CPUInfo) bool { + c.PhysicalCores = runtime.NumCPU() + // For now assuming 1 thread per core... + c.ThreadsPerCore = 1 + c.LogicalCores = c.PhysicalCores + return false +} diff --git a/vendor/github.com/klauspost/cpuid/v2/os_safe_linux_arm64.go b/vendor/github.com/klauspost/cpuid/v2/os_safe_linux_arm64.go new file mode 100644 index 0000000..f8f201b --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/os_safe_linux_arm64.go @@ -0,0 +1,8 @@ +// Copyright (c) 2021 Klaus Post, released under MIT License. See LICENSE file. + +//go:build nounsafe +// +build nounsafe + +package cpuid + +var hwcap uint diff --git a/vendor/github.com/klauspost/cpuid/v2/os_unsafe_linux_arm64.go b/vendor/github.com/klauspost/cpuid/v2/os_unsafe_linux_arm64.go new file mode 100644 index 0000000..92af622 --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/os_unsafe_linux_arm64.go @@ -0,0 +1,11 @@ +// Copyright (c) 2021 Klaus Post, released under MIT License. See LICENSE file. + +//go:build !nounsafe +// +build !nounsafe + +package cpuid + +import _ "unsafe" // needed for go:linkname + +//go:linkname hwcap internal/cpu.HWCap +var hwcap uint diff --git a/vendor/github.com/klauspost/cpuid/v2/test-architectures.sh b/vendor/github.com/klauspost/cpuid/v2/test-architectures.sh new file mode 100644 index 0000000..471d986 --- /dev/null +++ b/vendor/github.com/klauspost/cpuid/v2/test-architectures.sh @@ -0,0 +1,15 @@ +#!/bin/sh + +set -e + +go tool dist list | while IFS=/ read os arch; do + echo "Checking $os/$arch..." + echo " normal" + GOARCH=$arch GOOS=$os go build -o /dev/null . + echo " noasm" + GOARCH=$arch GOOS=$os go build -tags noasm -o /dev/null . + echo " appengine" + GOARCH=$arch GOOS=$os go build -tags appengine -o /dev/null . + echo " noasm,appengine" + GOARCH=$arch GOOS=$os go build -tags 'appengine noasm' -o /dev/null . +done -- cgit v1.2.3