From 060c3af12406169b3bbd4b58afe13dd10ceea9fc Mon Sep 17 00:00:00 2001 From: Carles Fernandez Date: Sat, 18 Mar 2023 11:41:43 +0100 Subject: [PATCH] cpu_features: Add Windows Arm64 support, update C++ standard for tests from C++11 to C++14, add Raptor Lake-P and Raptor Lake-HX/S detection --- .../volk_gnsssdr/cpu_features/CMakeLists.txt | 5 +- .../include/cpu_features_macros.h | 2 +- .../cpu_features/include/cpuinfo_aarch64.h | 103 +++++++++- .../include/internal/windows_utils.h | 32 +++ .../cpu_features/src/impl_aarch64_windows.c | 130 ++++++++++++ .../cpu_features/test/CMakeLists.txt | 6 +- .../cpu_features/test/cpuinfo_aarch64_test.cc | 188 ++++++++---------- .../cpu_features/test/cpuinfo_x86_test.cc | 35 ++++ 8 files changed, 388 insertions(+), 113 deletions(-) create mode 100644 src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/src/impl_aarch64_windows.c diff --git a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/CMakeLists.txt b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/CMakeLists.txt index 56c411c53..c213b6016 100644 --- a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/CMakeLists.txt +++ b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/CMakeLists.txt @@ -85,6 +85,7 @@ macro(add_cpu_features_headers_and_sources HDRS_LIST_NAME SRCS_LIST_NAME) list(APPEND ${HDRS_LIST_NAME} ${PROJECT_SOURCE_DIR}/include/cpuinfo_arm.h) elseif(PROCESSOR_IS_AARCH64) list(APPEND ${HDRS_LIST_NAME} ${PROJECT_SOURCE_DIR}/include/cpuinfo_aarch64.h) + list(APPEND ${SRCS_LIST_NAME} ${PROJECT_SOURCE_DIR}/include/internal/windows_utils.h) elseif(PROCESSOR_IS_X86) list(APPEND ${HDRS_LIST_NAME} ${PROJECT_SOURCE_DIR}/include/cpuinfo_x86.h) list(APPEND ${SRCS_LIST_NAME} ${PROJECT_SOURCE_DIR}/include/internal/cpuid_x86.h) @@ -185,9 +186,9 @@ if(BUILD_TESTING) # found. enable_language(CXX) - set(CMAKE_CXX_STANDARD 11) + set(CMAKE_CXX_STANDARD 14) set(CMAKE_CXX_STANDARD_REQUIRED ON) - set(CMAKE_CXX_EXTENSIONS OFF) # prefer use of -std11 instead of -gnustd11 + set(CMAKE_CXX_EXTENSIONS OFF) # prefer use of -std14 instead of -gnustd14 if(NOT TARGET gtest OR NOT TARGET gmock_main) # Download and unpack googletest at configure time. diff --git a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/include/cpu_features_macros.h b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/include/cpu_features_macros.h index ace763248..2a8442db9 100644 --- a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/include/cpu_features_macros.h +++ b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/include/cpu_features_macros.h @@ -28,7 +28,7 @@ #define CPU_FEATURES_ARCH_ARM #endif -#if defined(__aarch64__) +#if (defined(__aarch64__) || defined(_M_ARM64)) #define CPU_FEATURES_ARCH_AARCH64 #endif diff --git a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/include/cpuinfo_aarch64.h b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/include/cpuinfo_aarch64.h index b68869e3f..9fafccab8 100644 --- a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/include/cpuinfo_aarch64.h +++ b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/include/cpuinfo_aarch64.h @@ -1,6 +1,100 @@ // SPDX-FileCopyrightText: 2017 Google LLC // SPDX-License-Identifier: Apache-2.0 +//////////////////////////////////////////////////////////////////////////////// +// A note on Windows AArch64 implementation +//////////////////////////////////////////////////////////////////////////////// + +// Getting cpu info via EL1 system registers is not possible, so we delegate it +// to the Windows API (i.e., IsProcessorFeaturePresent and GetNativeSystemInfo). +// The `implementer`, `variant` and `part` fields of the `Aarch64Info` struct +// are not used, so they are set to 0. To get `revision` we use +// `wProcessorRevision` from `SYSTEM_INFO`. +// +// Cryptographic Extension: +// ----------------------------------------------------------------------------- +// According to documentation Arm Architecture Reference Manual for +// A-profile architecture. A2.3 The Armv8 Cryptographic Extension. The Armv8.0 +// Cryptographic Extension provides instructions for the acceleration of +// encryption and decryption, and includes the following features: FEAT_AES, +// FEAT_PMULL, FEAT_SHA1, FEAT_SHA256. +// see: https://developer.arm.com/documentation/ddi0487/latest +// +// We use `PF_ARM_V8_CRYPTO_INSTRUCTIONS_AVAILABLE` to detect all Armv8.0 crypto +// features. This value reports all features or nothing, so even if you only +// have support FEAT_AES and FEAT_PMULL, it will still return false. +// +// From Armv8.2, an implementation of the Armv8.0 Cryptographic Extension can +// include either or both of: +// +// • The AES functionality, including support for multiplication of 64-bit +// polynomials. The ID_AA64ISAR0_EL1.AES field indicates whether this +// functionality is supported. +// • The SHA1 and SHA2-256 functionality. The ID_AA64ISAR0_EL1.{SHA2, SHA1} +// fields indicate whether this functionality is supported. +// +// ID_AA64ISAR0_EL1.AES, bits [7:4]: +// Indicates support for AES instructions in AArch64 state. Defined values are: +// - 0b0000 No AES instructions implemented. +// - 0b0001 AESE, AESD, AESMC, and AESIMC instructions implemented. +// - 0b0010 As for 0b0001, plus PMULL/PMULL2 instructions operating on 64-bit +// data quantities. +// +// FEAT_AES implements the functionality identified by the value 0b0001. +// FEAT_PMULL implements the functionality identified by the value 0b0010. +// From Armv8, the permitted values are 0b0000 and 0b0010. +// +// ID_AA64ISAR0_EL1.SHA1, bits [11:8]: +// Indicates support for SHA1 instructions in AArch64 state. Defined values are: +// - 0b0000 No SHA1 instructions implemented. +// - 0b0001 SHA1C, SHA1P, SHA1M, SHA1H, SHA1SU0, and SHA1SU1 instructions +// implemented. +// +// FEAT_SHA1 implements the functionality identified by the value 0b0001. +// From Armv8, the permitted values are 0b0000 and 0b0001. +// If the value of ID_AA64ISAR0_EL1.SHA2 is 0b0000, this field must have the +// value 0b0000. +// +// ID_AA64ISAR0_EL1.SHA2, bits [15:12]: +// Indicates support for SHA2 instructions in AArch64 state. Defined values are: +// - 0b0000 No SHA2 instructions implemented. +// - 0b0001 Implements instructions: SHA256H, SHA256H2, SHA256SU0, and +// SHA256SU1. +// - 0b0010 Implements instructions: +// • SHA256H, SHA256H2, SHA256SU0, and SHA256SU1. +// • SHA512H, SHA512H2, SHA512SU0, and SHA512SU1. +// +// FEAT_SHA256 implements the functionality identified by the value 0b0001. +// FEAT_SHA512 implements the functionality identified by the value 0b0010. +// +// In Armv8, the permitted values are 0b0000 and 0b0001. +// From Armv8.2, the permitted values are 0b0000, 0b0001, and 0b0010. +// +// If the value of ID_AA64ISAR0_EL1.SHA1 is 0b0000, this field must have the +// value 0b0000. +// +// If the value of this field is 0b0010, ID_AA64ISAR0_EL1.SHA3 +// must have the value 0b0001. +// +// Other cryptographic features that we cannot detect such as sha512, sha3, sm3, +// sm4, sveaes, svepmull, svesha3, svesm4 we set to 0. +// +// FP/SIMD: +// ----------------------------------------------------------------------------- +// FP/SIMD must be implemented on all Armv8.0 implementations, but +// implementations targeting specialized markets may support the following +// combinations: +// +// • No NEON or floating-point. +// • Full floating-point and SIMD support with exception trapping. +// • Full floating-point and SIMD support without exception trapping. +// +// ref: +// https://developer.arm.com/documentation/den0024/a/AArch64-Floating-point-and-NEON +// +// So, we use `PF_ARM_VFP_32_REGISTERS_AVAILABLE`, +// `PF_ARM_NEON_INSTRUCTIONS_AVAILABLE` to detect `asimd` and `fp` + #ifndef CPU_FEATURES_INCLUDE_CPUINFO_AARCH64_H_ #define CPU_FEATURES_INCLUDE_CPUINFO_AARCH64_H_ @@ -72,10 +166,11 @@ typedef struct typedef struct { Aarch64Features features; - int implementer; - int variant; - int part; - int revision; + int implementer; // We set 0 for Windows. + int variant; // We set 0 for Windows. + int part; // We set 0 for Windows. + int revision; // We use GetNativeSystemInfo to get processor revision for + // Windows. } Aarch64Info; Aarch64Info GetAarch64Info(void); diff --git a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/include/internal/windows_utils.h b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/include/internal/windows_utils.h index 1790adc9c..cc08d3c9f 100644 --- a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/include/internal/windows_utils.h +++ b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/include/internal/windows_utils.h @@ -22,5 +22,37 @@ #define PF_SSE4_2_INSTRUCTIONS_AVAILABLE 38 #endif +#if !defined(PF_ARM_VFP_32_REGISTERS_AVAILABLE) +#define PF_ARM_VFP_32_REGISTERS_AVAILABLE 18 +#endif + +#if !defined(PF_ARM_NEON_INSTRUCTIONS_AVAILABLE) +#define PF_ARM_NEON_INSTRUCTIONS_AVAILABLE 19 +#endif + +#if !defined(PF_ARM_V8_CRYPTO_INSTRUCTIONS_AVAILABLE) +#define PF_ARM_V8_CRYPTO_INSTRUCTIONS_AVAILABLE 30 +#endif + +#if !defined(PF_ARM_V8_CRC32_INSTRUCTIONS_AVAILABLE) +#define PF_ARM_V8_CRC32_INSTRUCTIONS_AVAILABLE 31 +#endif + +#if !defined(PF_ARM_V81_ATOMIC_INSTRUCTIONS_AVAILABLE) +#define PF_ARM_V81_ATOMIC_INSTRUCTIONS_AVAILABLE 34 +#endif + +#if !defined(PF_ARM_V82_DP_INSTRUCTIONS_AVAILABLE) +#define PF_ARM_V82_DP_INSTRUCTIONS_AVAILABLE 43 +#endif + +#if !defined(PF_ARM_V83_JSCVT_INSTRUCTIONS_AVAILABLE) +#define PF_ARM_V83_JSCVT_INSTRUCTIONS_AVAILABLE 44 +#endif + +#if !defined(PF_ARM_V83_LRCPC_INSTRUCTIONS_AVAILABLE) +#define PF_ARM_V83_LRCPC_INSTRUCTIONS_AVAILABLE 45 +#endif + #endif // CPU_FEATURES_OS_WINDOWS #endif // CPU_FEATURES_INCLUDE_INTERNAL_WINDOWS_UTILS_H_ diff --git a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/src/impl_aarch64_windows.c b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/src/impl_aarch64_windows.c new file mode 100644 index 000000000..dd830863b --- /dev/null +++ b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/src/impl_aarch64_windows.c @@ -0,0 +1,130 @@ +// SPDX-FileCopyrightText: 2023 Google LLC +// SPDX-License-Identifier: Apache-2.0 + +#include "cpu_features_macros.h" + +#ifdef CPU_FEATURES_ARCH_AARCH64 +#ifdef CPU_FEATURES_OS_WINDOWS + +#include "cpuinfo_aarch64.h" + +//////////////////////////////////////////////////////////////////////////////// +// Definitions for introspection. +//////////////////////////////////////////////////////////////////////////////// +#define INTROSPECTION_TABLE \ + LINE(AARCH64_FP, fp, , , ) \ + LINE(AARCH64_ASIMD, asimd, , , ) \ + LINE(AARCH64_EVTSTRM, evtstrm, , , ) \ + LINE(AARCH64_AES, aes, , , ) \ + LINE(AARCH64_PMULL, pmull, , , ) \ + LINE(AARCH64_SHA1, sha1, , , ) \ + LINE(AARCH64_SHA2, sha2, , , ) \ + LINE(AARCH64_CRC32, crc32, , , ) \ + LINE(AARCH64_ATOMICS, atomics, , , ) \ + LINE(AARCH64_FPHP, fphp, , , ) \ + LINE(AARCH64_ASIMDHP, asimdhp, , , ) \ + LINE(AARCH64_CPUID, cpuid, , , ) \ + LINE(AARCH64_ASIMDRDM, asimdrdm, , , ) \ + LINE(AARCH64_JSCVT, jscvt, , , ) \ + LINE(AARCH64_FCMA, fcma, , , ) \ + LINE(AARCH64_LRCPC, lrcpc, , , ) \ + LINE(AARCH64_DCPOP, dcpop, , , ) \ + LINE(AARCH64_SHA3, sha3, , , ) \ + LINE(AARCH64_SM3, sm3, , , ) \ + LINE(AARCH64_SM4, sm4, , , ) \ + LINE(AARCH64_ASIMDDP, asimddp, , , ) \ + LINE(AARCH64_SHA512, sha512, , , ) \ + LINE(AARCH64_SVE, sve, , , ) \ + LINE(AARCH64_ASIMDFHM, asimdfhm, , , ) \ + LINE(AARCH64_DIT, dit, , , ) \ + LINE(AARCH64_USCAT, uscat, , , ) \ + LINE(AARCH64_ILRCPC, ilrcpc, , , ) \ + LINE(AARCH64_FLAGM, flagm, , , ) \ + LINE(AARCH64_SSBS, ssbs, , , ) \ + LINE(AARCH64_SB, sb, , , ) \ + LINE(AARCH64_PACA, paca, , , ) \ + LINE(AARCH64_PACG, pacg, , , ) \ + LINE(AARCH64_DCPODP, dcpodp, , , ) \ + LINE(AARCH64_SVE2, sve2, , , ) \ + LINE(AARCH64_SVEAES, sveaes, , , ) \ + LINE(AARCH64_SVEPMULL, svepmull, , , ) \ + LINE(AARCH64_SVEBITPERM, svebitperm, , , ) \ + LINE(AARCH64_SVESHA3, svesha3, , , ) \ + LINE(AARCH64_SVESM4, svesm4, , , ) \ + LINE(AARCH64_FLAGM2, flagm2, , , ) \ + LINE(AARCH64_FRINT, frint, , , ) \ + LINE(AARCH64_SVEI8MM, svei8mm, , , ) \ + LINE(AARCH64_SVEF32MM, svef32mm, , , ) \ + LINE(AARCH64_SVEF64MM, svef64mm, , , ) \ + LINE(AARCH64_SVEBF16, svebf16, , , ) \ + LINE(AARCH64_I8MM, i8mm, , , ) \ + LINE(AARCH64_BF16, bf16, , , ) \ + LINE(AARCH64_DGH, dgh, , , ) \ + LINE(AARCH64_RNG, rng, , , ) \ + LINE(AARCH64_BTI, bti, , , ) \ + LINE(AARCH64_MTE, mte, , , ) \ + LINE(AARCH64_ECV, ecv, , , ) \ + LINE(AARCH64_AFP, afp, , , ) \ + LINE(AARCH64_RPRES, rpres, , , ) +#define INTROSPECTION_PREFIX Aarch64 +#define INTROSPECTION_ENUM_PREFIX AARCH64 +#include "define_introspection.inl" + +//////////////////////////////////////////////////////////////////////////////// +// Implementation. +//////////////////////////////////////////////////////////////////////////////// + +#include "internal/windows_utils.h" +#include + +#ifdef CPU_FEATURES_MOCK_CPUID_AARCH64 +extern bool GetWindowsIsProcessorFeaturePresent(DWORD); +extern WORD GetWindowsNativeSystemInfoProcessorRevision(); +#else // CPU_FEATURES_MOCK_CPUID_AARCH64 +static bool +GetWindowsIsProcessorFeaturePresent(DWORD dwProcessorFeature) +{ + return IsProcessorFeaturePresent(dwProcessorFeature); +} + +static WORD GetWindowsNativeSystemInfoProcessorRevision() +{ + SYSTEM_INFO system_info; + GetNativeSystemInfo(&system_info); + return system_info.wProcessorRevision; +} +#endif + +static const Aarch64Info kEmptyAarch64Info; + +Aarch64Info GetAarch64Info(void) +{ + Aarch64Info info = kEmptyAarch64Info; + info.revision = GetWindowsNativeSystemInfoProcessorRevision(); + info.features.fp = + GetWindowsIsProcessorFeaturePresent(PF_ARM_VFP_32_REGISTERS_AVAILABLE); + info.features.asimd = + GetWindowsIsProcessorFeaturePresent(PF_ARM_NEON_INSTRUCTIONS_AVAILABLE); + info.features.crc32 = GetWindowsIsProcessorFeaturePresent( + PF_ARM_V8_CRC32_INSTRUCTIONS_AVAILABLE); + info.features.asimddp = + GetWindowsIsProcessorFeaturePresent(PF_ARM_V82_DP_INSTRUCTIONS_AVAILABLE); + info.features.jscvt = GetWindowsIsProcessorFeaturePresent( + PF_ARM_V83_JSCVT_INSTRUCTIONS_AVAILABLE); + info.features.lrcpc = GetWindowsIsProcessorFeaturePresent( + PF_ARM_V83_LRCPC_INSTRUCTIONS_AVAILABLE); + info.features.atomics = GetWindowsIsProcessorFeaturePresent( + PF_ARM_V81_ATOMIC_INSTRUCTIONS_AVAILABLE); + + + bool is_crypto_available = GetWindowsIsProcessorFeaturePresent( + PF_ARM_V8_CRYPTO_INSTRUCTIONS_AVAILABLE); + info.features.aes = is_crypto_available; + info.features.sha1 = is_crypto_available; + info.features.sha2 = is_crypto_available; + info.features.pmull = is_crypto_available; + return info; +} + +#endif // CPU_FEATURES_OS_WINDOWS +#endif // CPU_FEATURES_ARCH_AARCH64 \ No newline at end of file diff --git a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/test/CMakeLists.txt b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/test/CMakeLists.txt index 8df134964..5d41179e5 100644 --- a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/test/CMakeLists.txt +++ b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/test/CMakeLists.txt @@ -74,7 +74,11 @@ endif() ##------------------------------------------------------------------------------ ## cpuinfo_aarch64_test if(PROCESSOR_IS_AARCH64) - add_executable(cpuinfo_aarch64_test cpuinfo_aarch64_test.cc ../src/impl_aarch64_linux_or_android.c) + add_executable(cpuinfo_aarch64_test + cpuinfo_aarch64_test.cc + ../src/impl_aarch64_linux_or_android.c + ../src/impl_aarch64_windows.c) + target_compile_definitions(cpuinfo_aarch64_test PUBLIC CPU_FEATURES_MOCK_CPUID_AARCH64) target_link_libraries(cpuinfo_aarch64_test all_libraries) add_test(NAME cpuinfo_aarch64_test COMMAND cpuinfo_aarch64_test) endif() diff --git a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/test/cpuinfo_aarch64_test.cc b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/test/cpuinfo_aarch64_test.cc index 8d88e49d3..d2450752a 100644 --- a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/test/cpuinfo_aarch64_test.cc +++ b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/test/cpuinfo_aarch64_test.cc @@ -5,150 +5,103 @@ #include "filesystem_for_testing.h" #include "gtest/gtest.h" #include "hwcaps_for_testing.h" +#include + +#if defined(CPU_FEATURES_OS_WINDOWS) +#include "internal/windows_utils.h" +#endif // CPU_FEATURES_OS_WINDOWS namespace cpu_features { -namespace -{ -#if defined(CPU_FEATURES_OS_DARWIN) - -class FakeCpu +class FakeCpuAarch64 { public: - bool GetDarwinSysCtlByName(std::string name) const +#if defined(CPU_FEATURES_OS_WINDOWS) + bool GetWindowsIsProcessorFeaturePresent(DWORD dwProcessorFeature) { - return darwin_sysctlbyname_.count(name); + return windows_isprocessorfeaturepresent_.count(dwProcessorFeature); } - int GetDarwinSysCtlByNameValue(std::string name) const + void SetWindowsIsProcessorFeaturePresent(DWORD dwProcessorFeature) { - std::map::const_iterator iter = - darwin_sysctlbynamevalue_.find(name); - if (iter != std::end(darwin_sysctlbynamevalue_)) - { - return iter->second; - } - - return 0; + windows_isprocessorfeaturepresent_.insert(dwProcessorFeature); } - void SetDarwinSysCtlByName(std::string name) + WORD GetWindowsNativeSystemInfoProcessorRevision() const { - darwin_sysctlbyname_.insert(name); + return processor_revision_; } - void SetDarwinSysCtlByNameValue(std::string name, int value) + void SetWindowsNativeSystemInfoProcessorRevision(WORD wProcessorRevision) { - darwin_sysctlbynamevalue_[name] = value; + processor_revision_ = wProcessorRevision; } private: - std::set darwin_sysctlbyname_; - std::map darwin_sysctlbynamevalue_; + std::set windows_isprocessorfeaturepresent_; + WORD processor_revision_{}; +#endif // CPU_FEATURES_OS_WINDOWS }; -FakeCpu* g_fake_cpu = nullptr; +static FakeCpuAarch64* g_fake_cpu_instance = nullptr; -extern "C" bool GetDarwinSysCtlByName(const char* name) +static FakeCpuAarch64& cpu() { - return g_fake_cpu->GetDarwinSysCtlByName(name); + assert(g_fake_cpu_instance != nullptr); + return *g_fake_cpu_instance; } -extern "C" int GetDarwinSysCtlByNameValue(const char* name) +#if defined(CPU_FEATURES_OS_WINDOWS) +extern "C" bool GetWindowsIsProcessorFeaturePresent(DWORD dwProcessorFeature) { - return g_fake_cpu->GetDarwinSysCtlByNameValue(name); + return cpu().GetWindowsIsProcessorFeaturePresent(dwProcessorFeature); } -class CpuinfoAarch64Test : public ::testing::Test +extern "C" WORD GetWindowsNativeSystemInfoProcessorRevision() +{ + return cpu().GetWindowsNativeSystemInfoProcessorRevision(); +} +#endif // CPU_FEATURES_OS_WINDOWS + +namespace +{ + +class CpuidAarch64Test : public ::testing::Test { protected: - void SetUp() override { g_fake_cpu = new FakeCpu(); } - void TearDown() override { delete g_fake_cpu; } + void SetUp() override + { + assert(g_fake_cpu_instance == nullptr); + g_fake_cpu_instance = new FakeCpuAarch64(); + } + void TearDown() override + { + delete g_fake_cpu_instance; + g_fake_cpu_instance = nullptr; + } }; -TEST_F(CpuinfoAarch64Test, FromDarwinSysctlFromName) -{ - g_fake_cpu->SetDarwinSysCtlByName("hw.optional.floatingpoint"); - g_fake_cpu->SetDarwinSysCtlByName("hw.optional.neon"); - g_fake_cpu->SetDarwinSysCtlByName("hw.optional.neon_hpfp"); - g_fake_cpu->SetDarwinSysCtlByName("hw.optional.neon_fp16"); - g_fake_cpu->SetDarwinSysCtlByName("hw.optional.armv8_1_atomics"); - g_fake_cpu->SetDarwinSysCtlByName("hw.optional.armv8_crc32"); - g_fake_cpu->SetDarwinSysCtlByName("hw.optional.armv8_2_fhm"); - g_fake_cpu->SetDarwinSysCtlByName("hw.optional.armv8_2_sha512"); - g_fake_cpu->SetDarwinSysCtlByName("hw.optional.armv8_2_sha3"); - g_fake_cpu->SetDarwinSysCtlByName("hw.optional.amx_version"); - g_fake_cpu->SetDarwinSysCtlByName("hw.optional.ucnormal_mem"); - g_fake_cpu->SetDarwinSysCtlByName("hw.optional.arm64"); - - g_fake_cpu->SetDarwinSysCtlByNameValue("hw.cputype", 16777228); - g_fake_cpu->SetDarwinSysCtlByNameValue("hw.cpusubtype", 2); - g_fake_cpu->SetDarwinSysCtlByNameValue("hw.cpu64bit", 1); - g_fake_cpu->SetDarwinSysCtlByNameValue("hw.cpufamily", 458787763); - g_fake_cpu->SetDarwinSysCtlByNameValue("hw.cpusubfamily", 2); - - const auto info = GetAarch64Info(); - - EXPECT_EQ(info.implementer, 0x100000C); - EXPECT_EQ(info.variant, 2); - EXPECT_EQ(info.part, 0x1B588BB3); - EXPECT_EQ(info.revision, 2); - - EXPECT_TRUE(info.features.fp); - EXPECT_FALSE(info.features.asimd); - EXPECT_FALSE(info.features.evtstrm); - EXPECT_FALSE(info.features.aes); - EXPECT_FALSE(info.features.pmull); - EXPECT_FALSE(info.features.sha1); - EXPECT_FALSE(info.features.sha2); - EXPECT_TRUE(info.features.crc32); - EXPECT_TRUE(info.features.atomics); - EXPECT_TRUE(info.features.fphp); - EXPECT_FALSE(info.features.asimdhp); - EXPECT_FALSE(info.features.cpuid); - EXPECT_FALSE(info.features.asimdrdm); - EXPECT_FALSE(info.features.jscvt); - EXPECT_FALSE(info.features.fcma); - EXPECT_FALSE(info.features.lrcpc); - EXPECT_FALSE(info.features.dcpop); - EXPECT_TRUE(info.features.sha3); - EXPECT_FALSE(info.features.sm3); - EXPECT_FALSE(info.features.sm4); - EXPECT_FALSE(info.features.asimddp); - EXPECT_TRUE(info.features.sha512); - EXPECT_FALSE(info.features.sve); - EXPECT_TRUE(info.features.asimdfhm); - EXPECT_FALSE(info.features.dit); - EXPECT_FALSE(info.features.uscat); - EXPECT_FALSE(info.features.ilrcpc); - EXPECT_FALSE(info.features.flagm); - EXPECT_FALSE(info.features.ssbs); - EXPECT_FALSE(info.features.sb); - EXPECT_FALSE(info.features.paca); - EXPECT_FALSE(info.features.pacg); -} - -#else - -void DisableHardwareCapabilities() -{ - SetHardwareCapabilities(0, 0); -} - TEST(CpuinfoAarch64Test, Aarch64FeaturesEnum) { - const char *last_name = GetAarch64FeaturesEnumName(AARCH64_LAST_); + const char* last_name = GetAarch64FeaturesEnumName(AARCH64_LAST_); EXPECT_STREQ(last_name, "unknown_feature"); - for (int i = static_cast(AARCH64_FP); i != static_cast(AARCH64_LAST_); ++i) + for (int i = static_cast(AARCH64_FP); + i != static_cast(AARCH64_LAST_); ++i) { const auto feature = static_cast(i); - const char *name = GetAarch64FeaturesEnumName(feature); + const char* name = GetAarch64FeaturesEnumName(feature); ASSERT_FALSE(name == nullptr); EXPECT_STRNE(name, ""); EXPECT_STRNE(name, last_name); } } +#if defined(CPU_FEATURES_OS_LINUX) +void DisableHardwareCapabilities() +{ + SetHardwareCapabilities(0, 0); +} + TEST(CpuinfoAarch64Test, FromHardwareCap) { ResetHwcaps(); @@ -222,7 +175,7 @@ TEST(CpuinfoAarch64Test, FromHardwareCap2) TEST(CpuinfoAarch64Test, ARMCortexA53) { ResetHwcaps(); - auto &fs = GetEmptyFilesystem(); + auto& fs = GetEmptyFilesystem(); fs.CreateFile("/proc/cpuinfo", R"(Processor : AArch64 Processor rev 3 (aarch64) processor : 0 @@ -301,8 +254,33 @@ CPU revision : 3)"); EXPECT_FALSE(info.features.afp); EXPECT_FALSE(info.features.rpres); } +#endif // CPU_FEATURES_OS_LINUX -#endif +#if defined(CPU_FEATURES_OS_WINDOWS) +TEST_F(CpuidAarch64Test, WINDOWS_AARCH64_RPI4) +{ + cpu().SetWindowsNativeSystemInfoProcessorRevision(0x03); + cpu().SetWindowsIsProcessorFeaturePresent(PF_ARM_VFP_32_REGISTERS_AVAILABLE); + cpu().SetWindowsIsProcessorFeaturePresent(PF_ARM_NEON_INSTRUCTIONS_AVAILABLE); + cpu().SetWindowsIsProcessorFeaturePresent( + PF_ARM_V8_CRC32_INSTRUCTIONS_AVAILABLE); + + const auto info = GetAarch64Info(); + + EXPECT_EQ(info.revision, 0x03); + EXPECT_TRUE(info.features.fp); + EXPECT_TRUE(info.features.asimd); + EXPECT_TRUE(info.features.crc32); + EXPECT_FALSE(info.features.aes); + EXPECT_FALSE(info.features.sha1); + EXPECT_FALSE(info.features.sha2); + EXPECT_FALSE(info.features.pmull); + EXPECT_FALSE(info.features.atomics); + EXPECT_FALSE(info.features.asimddp); + EXPECT_FALSE(info.features.jscvt); + EXPECT_FALSE(info.features.lrcpc); +} +#endif // CPU_FEATURES_OS_WINDOWS } // namespace -} // namespace cpu_features +} // namespace cpu_features \ No newline at end of file diff --git a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/test/cpuinfo_x86_test.cc b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/test/cpuinfo_x86_test.cc index f83df2efa..3948e1399 100644 --- a/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/test/cpuinfo_x86_test.cc +++ b/src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/cpu_features/test/cpuinfo_x86_test.cc @@ -1708,6 +1708,41 @@ TEST_F(CpuidX86Test, INTEL_HASWELL_LZCNT) EXPECT_TRUE(info.features.lzcnt); } +// http://users.atw.hu/instlatx64/GenuineIntel/GenuineIntel00B06A2_RaptorLakeP_03_CPUID.txt +TEST_F(CpuidX86Test, INTEL_RAPTOR_LAKE_P) +{ + cpu().SetLeaves({ + {{0x00000000, 0}, Leaf{0x00000020, 0x756E6547, 0x6C65746E, 0x49656E69}}, + {{0x00000001, 0}, Leaf{0x000B06A3, 0x00400800, 0x7FFAFBFF, 0xBFEBFBFF}}, + {{0x80000000, 0}, Leaf{0x80000008, 0x00000000, 0x00000000, 0x00000000}}, + {{0x80000001, 0}, Leaf{0x00000000, 0x00000000, 0x00000121, 0x2C100000}}, + }); + const auto info = GetX86Info(); + + EXPECT_STREQ(info.vendor, CPU_FEATURES_VENDOR_GENUINE_INTEL); + EXPECT_EQ(info.family, 0x06); + EXPECT_EQ(info.model, 0xBA); + EXPECT_EQ(GetX86Microarchitecture(&info), X86Microarchitecture::INTEL_RPL); +} + +// http://users.atw.hu/instlatx64/GenuineIntel/GenuineIntel00B06F2_RaptorLakeS_02_CPUID.txt +TEST_F(CpuidX86Test, INTEL_RAPTOR_LAKE_S) +{ + cpu().SetLeaves({ + {{0x00000000, 0}, Leaf{0x00000020, 0x756E6547, 0x6C65746E, 0x49656E69}}, + {{0x00000001, 0}, Leaf{0x000B06F2, 0x00800800, 0x7FFAFBFF, 0xBFEBFBFF}}, + {{0x80000000, 0}, Leaf{0x80000008, 0x00000000, 0x00000000, 0x00000000}}, + {{0x80000001, 0}, Leaf{0x00000000, 0x00000000, 0x00000121, 0x2C100000}}, + }); + const auto info = GetX86Info(); + + EXPECT_STREQ(info.vendor, CPU_FEATURES_VENDOR_GENUINE_INTEL); + EXPECT_EQ(info.family, 0x06); + EXPECT_EQ(info.model, 0xBF); + EXPECT_EQ(GetX86Microarchitecture(&info), X86Microarchitecture::INTEL_RPL); +} + + // https://github.com/google/cpu_features/issues/200 // http://users.atw.hu/instlatx64/GenuineIntel/GenuineIntel00206F2_Eagleton_CPUID.txt #if defined(CPU_FEATURES_OS_WINDOWS)