glibc/sysdeps/x86/include/cpu-features.h

984 lines
29 KiB
C
Raw Normal View History

x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
/* Data structure for x86 CPU features.
Copyright (C) 2020-2023 Free Software Foundation, Inc.
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
<https://www.gnu.org/licenses/>. */
#ifndef _PRIVATE_CPU_FEATURES_H
#define _PRIVATE_CPU_FEATURES_H 1
#ifdef _CPU_FEATURES_H
# error this should be impossible
#endif
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
/* Get data structures without inline functions. */
#define _SYS_PLATFORM_X86_H
#include <bits/platform/x86.h>
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
enum
{
CPUID_INDEX_MAX = CPUID_INDEX_14_ECX_0 + 1
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
};
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
enum
{
/* The integer bit array index for the first set of preferred feature
bits. */
PREFERRED_FEATURE_INDEX_1 = 0,
/* The current maximum size of the feature integer bit array. */
PREFERRED_FEATURE_INDEX_MAX
};
/* Only used directly in cpu-features.c. */
#define CPU_FEATURE_SET(ptr, name) \
ptr->features[index_cpu_##name].active.reg_##name |= bit_cpu_##name;
#define CPU_FEATURE_UNSET(ptr, name) \
ptr->features[index_cpu_##name].active.reg_##name &= ~bit_cpu_##name;
#define CPU_FEATURE_SET_ACTIVE(ptr, name) \
ptr->features[index_cpu_##name].active.reg_##name \
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
|= ptr->features[index_cpu_##name].cpuid.reg_##name & bit_cpu_##name;
#define CPU_FEATURE_PREFERRED_P(ptr, name) \
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
((ptr->preferred[index_arch_##name] & bit_arch_##name) != 0)
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define CPU_FEATURE_CHECK_P(ptr, name, check) \
((ptr->features[index_cpu_##name].check.reg_##name \
& bit_cpu_##name) != 0)
#define CPU_FEATURE_PRESENT_P(ptr, name) \
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
CPU_FEATURE_CHECK_P (ptr, name, cpuid)
#define CPU_FEATURE_ACTIVE_P(ptr, name) \
CPU_FEATURE_CHECK_P (ptr, name, active)
#define CPU_FEATURE_CPU_P(ptr, name) \
CPU_FEATURE_PRESENT_P (ptr, name)
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define CPU_FEATURE_USABLE_P(ptr, name) \
CPU_FEATURE_ACTIVE_P (ptr, name)
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
/* HAS_CPU_FEATURE evaluates to true if CPU supports the feature. */
#define HAS_CPU_FEATURE(name) \
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
CPU_FEATURE_CPU_P (__get_cpu_features (), name)
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
/* CPU_FEATURE_USABLE evaluates to true if the feature is usable. */
#define CPU_FEATURE_USABLE(name) \
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
CPU_FEATURE_USABLE_P (__get_cpu_features (), name)
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
/* CPU_FEATURE_PREFER evaluates to true if we prefer the feature at
runtime. */
#define CPU_FEATURE_PREFERRED(name) \
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
CPU_FEATURE_PREFERRED_P(__get_cpu_features (), name)
#define CPU_FEATURES_CPU_P(ptr, name) \
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
CPU_FEATURE_CPU_P (ptr, name)
#define CPU_FEATURES_ARCH_P(ptr, name) \
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
CPU_FEATURE_PREFERRED_P (ptr, name)
#define HAS_ARCH_FEATURE(name) \
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
CPU_FEATURE_PREFERRED (name)
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
/* CPU features. */
/* CPUID_INDEX_1. */
/* ECX. */
#define bit_cpu_SSE3 (1u << 0)
#define bit_cpu_PCLMULQDQ (1u << 1)
#define bit_cpu_DTES64 (1u << 2)
#define bit_cpu_MONITOR (1u << 3)
#define bit_cpu_DS_CPL (1u << 4)
#define bit_cpu_VMX (1u << 5)
#define bit_cpu_SMX (1u << 6)
#define bit_cpu_EIST (1u << 7)
#define bit_cpu_TM2 (1u << 8)
#define bit_cpu_SSSE3 (1u << 9)
#define bit_cpu_CNXT_ID (1u << 10)
#define bit_cpu_SDBG (1u << 11)
#define bit_cpu_FMA (1u << 12)
#define bit_cpu_CMPXCHG16B (1u << 13)
#define bit_cpu_XTPRUPDCTRL (1u << 14)
#define bit_cpu_PDCM (1u << 15)
#define bit_cpu_INDEX_1_ECX_16 (1u << 16)
#define bit_cpu_PCID (1u << 17)
#define bit_cpu_DCA (1u << 18)
#define bit_cpu_SSE4_1 (1u << 19)
#define bit_cpu_SSE4_2 (1u << 20)
#define bit_cpu_X2APIC (1u << 21)
#define bit_cpu_MOVBE (1u << 22)
#define bit_cpu_POPCNT (1u << 23)
#define bit_cpu_TSC_DEADLINE (1u << 24)
#define bit_cpu_AES (1u << 25)
#define bit_cpu_XSAVE (1u << 26)
#define bit_cpu_OSXSAVE (1u << 27)
#define bit_cpu_AVX (1u << 28)
#define bit_cpu_F16C (1u << 29)
#define bit_cpu_RDRAND (1u << 30)
#define bit_cpu_INDEX_1_ECX_31 (1u << 31)
/* EDX. */
#define bit_cpu_FPU (1u << 0)
#define bit_cpu_VME (1u << 1)
#define bit_cpu_DE (1u << 2)
#define bit_cpu_PSE (1u << 3)
#define bit_cpu_TSC (1u << 4)
#define bit_cpu_MSR (1u << 5)
#define bit_cpu_PAE (1u << 6)
#define bit_cpu_MCE (1u << 7)
#define bit_cpu_CX8 (1u << 8)
#define bit_cpu_APIC (1u << 9)
#define bit_cpu_INDEX_1_EDX_10 (1u << 10)
#define bit_cpu_SEP (1u << 11)
#define bit_cpu_MTRR (1u << 12)
#define bit_cpu_PGE (1u << 13)
#define bit_cpu_MCA (1u << 14)
#define bit_cpu_CMOV (1u << 15)
#define bit_cpu_PAT (1u << 16)
#define bit_cpu_PSE_36 (1u << 17)
#define bit_cpu_PSN (1u << 18)
#define bit_cpu_CLFSH (1u << 19)
#define bit_cpu_INDEX_1_EDX_20 (1u << 20)
#define bit_cpu_DS (1u << 21)
#define bit_cpu_ACPI (1u << 22)
#define bit_cpu_MMX (1u << 23)
#define bit_cpu_FXSR (1u << 24)
#define bit_cpu_SSE (1u << 25)
#define bit_cpu_SSE2 (1u << 26)
#define bit_cpu_SS (1u << 27)
#define bit_cpu_HTT (1u << 28)
#define bit_cpu_TM (1u << 29)
#define bit_cpu_INDEX_1_EDX_30 (1u << 30)
#define bit_cpu_PBE (1u << 31)
/* CPUID_INDEX_7. */
/* EBX. */
#define bit_cpu_FSGSBASE (1u << 0)
#define bit_cpu_TSC_ADJUST (1u << 1)
#define bit_cpu_SGX (1u << 2)
#define bit_cpu_BMI1 (1u << 3)
#define bit_cpu_HLE (1u << 4)
#define bit_cpu_AVX2 (1u << 5)
#define bit_cpu_INDEX_7_EBX_6 (1u << 6)
#define bit_cpu_SMEP (1u << 7)
#define bit_cpu_BMI2 (1u << 8)
#define bit_cpu_ERMS (1u << 9)
#define bit_cpu_INVPCID (1u << 10)
#define bit_cpu_RTM (1u << 11)
#define bit_cpu_RDT_M (1u << 12)
#define bit_cpu_DEPR_FPU_CS_DS (1u << 13)
#define bit_cpu_MPX (1u << 14)
#define bit_cpu_RDT_A (1u << 15)
#define bit_cpu_AVX512F (1u << 16)
#define bit_cpu_AVX512DQ (1u << 17)
#define bit_cpu_RDSEED (1u << 18)
#define bit_cpu_ADX (1u << 19)
#define bit_cpu_SMAP (1u << 20)
#define bit_cpu_AVX512_IFMA (1u << 21)
#define bit_cpu_INDEX_7_EBX_22 (1u << 22)
#define bit_cpu_CLFLUSHOPT (1u << 23)
#define bit_cpu_CLWB (1u << 24)
#define bit_cpu_TRACE (1u << 25)
#define bit_cpu_AVX512PF (1u << 26)
#define bit_cpu_AVX512ER (1u << 27)
#define bit_cpu_AVX512CD (1u << 28)
#define bit_cpu_SHA (1u << 29)
#define bit_cpu_AVX512BW (1u << 30)
#define bit_cpu_AVX512VL (1u << 31)
/* ECX. */
#define bit_cpu_PREFETCHWT1 (1u << 0)
#define bit_cpu_AVX512_VBMI (1u << 1)
#define bit_cpu_UMIP (1u << 2)
#define bit_cpu_PKU (1u << 3)
#define bit_cpu_OSPKE (1u << 4)
#define bit_cpu_WAITPKG (1u << 5)
#define bit_cpu_AVX512_VBMI2 (1u << 6)
#define bit_cpu_SHSTK (1u << 7)
#define bit_cpu_GFNI (1u << 8)
#define bit_cpu_VAES (1u << 9)
#define bit_cpu_VPCLMULQDQ (1u << 10)
#define bit_cpu_AVX512_VNNI (1u << 11)
#define bit_cpu_AVX512_BITALG (1u << 12)
#define bit_cpu_INDEX_7_ECX_13 (1u << 13)
#define bit_cpu_AVX512_VPOPCNTDQ (1u << 14)
#define bit_cpu_INDEX_7_ECX_15 (1u << 15)
#define bit_cpu_INDEX_7_ECX_16 (1u << 16)
/* Note: Bits 17-21: The value of MAWAU used by the BNDLDX and BNDSTX
instructions in 64-bit mode. */
#define bit_cpu_RDPID (1u << 22)
#define bit_cpu_KL (1u << 23)
#define bit_cpu_INDEX_7_ECX_24 (1u << 24)
#define bit_cpu_CLDEMOTE (1u << 25)
#define bit_cpu_INDEX_7_ECX_26 (1u << 26)
#define bit_cpu_MOVDIRI (1u << 27)
#define bit_cpu_MOVDIR64B (1u << 28)
#define bit_cpu_ENQCMD (1u << 29)
#define bit_cpu_SGX_LC (1u << 30)
#define bit_cpu_PKS (1u << 31)
/* EDX. */
#define bit_cpu_INDEX_7_EDX_0 (1u << 0)
#define bit_cpu_INDEX_7_EDX_1 (1u << 1)
#define bit_cpu_AVX512_4VNNIW (1u << 2)
#define bit_cpu_AVX512_4FMAPS (1u << 3)
#define bit_cpu_FSRM (1u << 4)
#define bit_cpu_UINTR (1u << 5)
#define bit_cpu_INDEX_7_EDX_6 (1u << 6)
#define bit_cpu_INDEX_7_EDX_7 (1u << 7)
#define bit_cpu_AVX512_VP2INTERSECT (1u << 8)
#define bit_cpu_INDEX_7_EDX_9 (1u << 9)
#define bit_cpu_MD_CLEAR (1u << 10)
#define bit_cpu_RTM_ALWAYS_ABORT (1u << 11)
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define bit_cpu_INDEX_7_EDX_12 (1u << 12)
#define bit_cpu_INDEX_7_EDX_13 (1u << 13)
#define bit_cpu_SERIALIZE (1u << 14)
#define bit_cpu_HYBRID (1u << 15)
#define bit_cpu_TSXLDTRK (1u << 16)
#define bit_cpu_INDEX_7_EDX_17 (1u << 17)
#define bit_cpu_PCONFIG (1u << 18)
#define bit_cpu_INDEX_7_EDX_19 (1u << 19)
#define bit_cpu_IBT (1u << 20)
#define bit_cpu_INDEX_7_EDX_21 (1u << 21)
#define bit_cpu_AMX_BF16 (1u << 22)
#define bit_cpu_AVX512_FP16 (1u << 23)
#define bit_cpu_AMX_TILE (1u << 24)
#define bit_cpu_AMX_INT8 (1u << 25)
#define bit_cpu_IBRS_IBPB (1u << 26)
#define bit_cpu_STIBP (1u << 27)
#define bit_cpu_L1D_FLUSH (1u << 28)
#define bit_cpu_ARCH_CAPABILITIES (1u << 29)
#define bit_cpu_CORE_CAPABILITIES (1u << 30)
#define bit_cpu_SSBD (1u << 31)
/* CPUID_INDEX_80000001. */
/* ECX. */
#define bit_cpu_LAHF64_SAHF64 (1u << 0)
#define bit_cpu_SVM (1u << 2)
#define bit_cpu_LZCNT (1u << 5)
#define bit_cpu_SSE4A (1u << 6)
#define bit_cpu_PREFETCHW (1u << 8)
#define bit_cpu_XOP (1u << 11)
#define bit_cpu_LWP (1u << 15)
#define bit_cpu_FMA4 (1u << 16)
#define bit_cpu_TBM (1u << 21)
/* EDX. */
#define bit_cpu_SYSCALL_SYSRET (1u << 11)
#define bit_cpu_NX (1u << 20)
#define bit_cpu_PAGE1GB (1u << 26)
#define bit_cpu_RDTSCP (1u << 27)
#define bit_cpu_LM (1u << 29)
/* CPUID_INDEX_D_ECX_1. */
/* EAX. */
#define bit_cpu_XSAVEOPT (1u << 0)
#define bit_cpu_XSAVEC (1u << 1)
#define bit_cpu_XGETBV_ECX_1 (1u << 2)
#define bit_cpu_XSAVES (1u << 3)
#define bit_cpu_XFD (1u << 4)
/* CPUID_INDEX_80000007. */
/* EDX. */
#define bit_cpu_INVARIANT_TSC (1u << 8)
/* CPUID_INDEX_80000008. */
/* EBX. */
#define bit_cpu_WBNOINVD (1u << 9)
#define bit_cpu_AMD_IBPB (1u << 12)
#define bit_cpu_AMD_IBRS (1u << 14)
#define bit_cpu_AMD_STIBP (1u << 15)
#define bit_cpu_AMD_SSBD (1u << 24)
#define bit_cpu_AMD_VIRT_SSBD (1u << 25)
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
/* CPUID_INDEX_7_ECX_1. */
/* EAX. */
#define bit_cpu_RAO_INT (1u << 3)
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define bit_cpu_AVX_VNNI (1u << 4)
#define bit_cpu_AVX512_BF16 (1u << 5)
#define bit_cpu_CMPCCXADD (1u << 7)
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define bit_cpu_FZLRM (1u << 10)
#define bit_cpu_FSRS (1u << 11)
#define bit_cpu_FSRCS (1u << 12)
#define bit_cpu_AMX_FP16 (1u << 21)
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define bit_cpu_HRESET (1u << 22)
#define bit_cpu_AVX_IFMA (1u << 23)
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define bit_cpu_LAM (1u << 26)
/* EDX. */
#define bit_cpu_AVX_VNNI_INT8 (1u << 4)
#define bit_cpu_AVX_NE_CONVERT (1u << 5)
#define bit_cpu_AMX_COMPLEX (1u << 8)
#define bit_cpu_PREFETCHI (1u << 14)
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
/* CPUID_INDEX_19. */
/* EBX. */
#define bit_cpu_AESKLE (1u << 0)
#define bit_cpu_WIDE_KL (1u << 2)
/* CPUID_INDEX_14_ECX_0. */
/* EBX. */
#define bit_cpu_PTWRITE (1u << 4)
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
/* CPUID_INDEX_1. */
/* ECX. */
#define index_cpu_SSE3 CPUID_INDEX_1
#define index_cpu_PCLMULQDQ CPUID_INDEX_1
#define index_cpu_DTES64 CPUID_INDEX_1
#define index_cpu_MONITOR CPUID_INDEX_1
#define index_cpu_DS_CPL CPUID_INDEX_1
#define index_cpu_VMX CPUID_INDEX_1
#define index_cpu_SMX CPUID_INDEX_1
#define index_cpu_EIST CPUID_INDEX_1
#define index_cpu_TM2 CPUID_INDEX_1
#define index_cpu_SSSE3 CPUID_INDEX_1
#define index_cpu_CNXT_ID CPUID_INDEX_1
#define index_cpu_SDBG CPUID_INDEX_1
#define index_cpu_FMA CPUID_INDEX_1
#define index_cpu_CMPXCHG16B CPUID_INDEX_1
#define index_cpu_XTPRUPDCTRL CPUID_INDEX_1
#define index_cpu_PDCM CPUID_INDEX_1
#define index_cpu_INDEX_1_ECX_16 CPUID_INDEX_1
#define index_cpu_PCID CPUID_INDEX_1
#define index_cpu_DCA CPUID_INDEX_1
#define index_cpu_SSE4_1 CPUID_INDEX_1
#define index_cpu_SSE4_2 CPUID_INDEX_1
#define index_cpu_X2APIC CPUID_INDEX_1
#define index_cpu_MOVBE CPUID_INDEX_1
#define index_cpu_POPCNT CPUID_INDEX_1
#define index_cpu_TSC_DEADLINE CPUID_INDEX_1
#define index_cpu_AES CPUID_INDEX_1
#define index_cpu_XSAVE CPUID_INDEX_1
#define index_cpu_OSXSAVE CPUID_INDEX_1
#define index_cpu_AVX CPUID_INDEX_1
#define index_cpu_F16C CPUID_INDEX_1
#define index_cpu_RDRAND CPUID_INDEX_1
#define index_cpu_INDEX_1_ECX_31 CPUID_INDEX_1
/* ECX. */
#define index_cpu_FPU CPUID_INDEX_1
#define index_cpu_VME CPUID_INDEX_1
#define index_cpu_DE CPUID_INDEX_1
#define index_cpu_PSE CPUID_INDEX_1
#define index_cpu_TSC CPUID_INDEX_1
#define index_cpu_MSR CPUID_INDEX_1
#define index_cpu_PAE CPUID_INDEX_1
#define index_cpu_MCE CPUID_INDEX_1
#define index_cpu_CX8 CPUID_INDEX_1
#define index_cpu_APIC CPUID_INDEX_1
#define index_cpu_INDEX_1_EDX_10 CPUID_INDEX_1
#define index_cpu_SEP CPUID_INDEX_1
#define index_cpu_MTRR CPUID_INDEX_1
#define index_cpu_PGE CPUID_INDEX_1
#define index_cpu_MCA CPUID_INDEX_1
#define index_cpu_CMOV CPUID_INDEX_1
#define index_cpu_PAT CPUID_INDEX_1
#define index_cpu_PSE_36 CPUID_INDEX_1
#define index_cpu_PSN CPUID_INDEX_1
#define index_cpu_CLFSH CPUID_INDEX_1
#define index_cpu_INDEX_1_EDX_20 CPUID_INDEX_1
#define index_cpu_DS CPUID_INDEX_1
#define index_cpu_ACPI CPUID_INDEX_1
#define index_cpu_MMX CPUID_INDEX_1
#define index_cpu_FXSR CPUID_INDEX_1
#define index_cpu_SSE CPUID_INDEX_1
#define index_cpu_SSE2 CPUID_INDEX_1
#define index_cpu_SS CPUID_INDEX_1
#define index_cpu_HTT CPUID_INDEX_1
#define index_cpu_TM CPUID_INDEX_1
#define index_cpu_INDEX_1_EDX_30 CPUID_INDEX_1
#define index_cpu_PBE CPUID_INDEX_1
/* CPUID_INDEX_7. */
/* EBX. */
#define index_cpu_FSGSBASE CPUID_INDEX_7
#define index_cpu_TSC_ADJUST CPUID_INDEX_7
#define index_cpu_SGX CPUID_INDEX_7
#define index_cpu_BMI1 CPUID_INDEX_7
#define index_cpu_HLE CPUID_INDEX_7
#define index_cpu_AVX2 CPUID_INDEX_7
#define index_cpu_INDEX_7_EBX_6 CPUID_INDEX_7
#define index_cpu_SMEP CPUID_INDEX_7
#define index_cpu_BMI2 CPUID_INDEX_7
#define index_cpu_ERMS CPUID_INDEX_7
#define index_cpu_INVPCID CPUID_INDEX_7
#define index_cpu_RTM CPUID_INDEX_7
#define index_cpu_RDT_M CPUID_INDEX_7
#define index_cpu_DEPR_FPU_CS_DS CPUID_INDEX_7
#define index_cpu_MPX CPUID_INDEX_7
#define index_cpu_RDT_A CPUID_INDEX_7
#define index_cpu_AVX512F CPUID_INDEX_7
#define index_cpu_AVX512DQ CPUID_INDEX_7
#define index_cpu_RDSEED CPUID_INDEX_7
#define index_cpu_ADX CPUID_INDEX_7
#define index_cpu_SMAP CPUID_INDEX_7
#define index_cpu_AVX512_IFMA CPUID_INDEX_7
#define index_cpu_INDEX_7_EBX_22 CPUID_INDEX_7
#define index_cpu_CLFLUSHOPT CPUID_INDEX_7
#define index_cpu_CLWB CPUID_INDEX_7
#define index_cpu_TRACE CPUID_INDEX_7
#define index_cpu_AVX512PF CPUID_INDEX_7
#define index_cpu_AVX512ER CPUID_INDEX_7
#define index_cpu_AVX512CD CPUID_INDEX_7
#define index_cpu_SHA CPUID_INDEX_7
#define index_cpu_AVX512BW CPUID_INDEX_7
#define index_cpu_AVX512VL CPUID_INDEX_7
/* ECX. */
#define index_cpu_PREFETCHWT1 CPUID_INDEX_7
#define index_cpu_AVX512_VBMI CPUID_INDEX_7
#define index_cpu_UMIP CPUID_INDEX_7
#define index_cpu_PKU CPUID_INDEX_7
#define index_cpu_OSPKE CPUID_INDEX_7
#define index_cpu_WAITPKG CPUID_INDEX_7
#define index_cpu_AVX512_VBMI2 CPUID_INDEX_7
#define index_cpu_SHSTK CPUID_INDEX_7
#define index_cpu_GFNI CPUID_INDEX_7
#define index_cpu_VAES CPUID_INDEX_7
#define index_cpu_VPCLMULQDQ CPUID_INDEX_7
#define index_cpu_AVX512_VNNI CPUID_INDEX_7
#define index_cpu_AVX512_BITALG CPUID_INDEX_7
#define index_cpu_INDEX_7_ECX_13 CPUID_INDEX_7
#define index_cpu_AVX512_VPOPCNTDQ CPUID_INDEX_7
#define index_cpu_INDEX_7_ECX_15 CPUID_INDEX_7
#define index_cpu_INDEX_7_ECX_16 CPUID_INDEX_7
#define index_cpu_RDPID CPUID_INDEX_7
#define index_cpu_KL CPUID_INDEX_7
#define index_cpu_INDEX_7_ECX_24 CPUID_INDEX_7
#define index_cpu_CLDEMOTE CPUID_INDEX_7
#define index_cpu_INDEX_7_ECX_26 CPUID_INDEX_7
#define index_cpu_MOVDIRI CPUID_INDEX_7
#define index_cpu_MOVDIR64B CPUID_INDEX_7
#define index_cpu_ENQCMD CPUID_INDEX_7
#define index_cpu_SGX_LC CPUID_INDEX_7
#define index_cpu_PKS CPUID_INDEX_7
/* EDX. */
#define index_cpu_INDEX_7_EDX_0 CPUID_INDEX_7
#define index_cpu_INDEX_7_EDX_1 CPUID_INDEX_7
#define index_cpu_AVX512_4VNNIW CPUID_INDEX_7
#define index_cpu_AVX512_4FMAPS CPUID_INDEX_7
#define index_cpu_FSRM CPUID_INDEX_7
#define index_cpu_UINTR CPUID_INDEX_7
#define index_cpu_INDEX_7_EDX_6 CPUID_INDEX_7
#define index_cpu_INDEX_7_EDX_7 CPUID_INDEX_7
#define index_cpu_AVX512_VP2INTERSECT CPUID_INDEX_7
#define index_cpu_INDEX_7_EDX_9 CPUID_INDEX_7
#define index_cpu_MD_CLEAR CPUID_INDEX_7
#define index_cpu_RTM_ALWAYS_ABORT CPUID_INDEX_7
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define index_cpu_INDEX_7_EDX_12 CPUID_INDEX_7
#define index_cpu_INDEX_7_EDX_13 CPUID_INDEX_7
#define index_cpu_SERIALIZE CPUID_INDEX_7
#define index_cpu_HYBRID CPUID_INDEX_7
#define index_cpu_TSXLDTRK CPUID_INDEX_7
#define index_cpu_INDEX_7_EDX_17 CPUID_INDEX_7
#define index_cpu_PCONFIG CPUID_INDEX_7
#define index_cpu_INDEX_7_EDX_19 CPUID_INDEX_7
#define index_cpu_IBT CPUID_INDEX_7
#define index_cpu_INDEX_7_EDX_21 CPUID_INDEX_7
#define index_cpu_AMX_BF16 CPUID_INDEX_7
#define index_cpu_AVX512_FP16 CPUID_INDEX_7
#define index_cpu_AMX_TILE CPUID_INDEX_7
#define index_cpu_AMX_INT8 CPUID_INDEX_7
#define index_cpu_IBRS_IBPB CPUID_INDEX_7
#define index_cpu_STIBP CPUID_INDEX_7
#define index_cpu_L1D_FLUSH CPUID_INDEX_7
#define index_cpu_ARCH_CAPABILITIES CPUID_INDEX_7
#define index_cpu_CORE_CAPABILITIES CPUID_INDEX_7
#define index_cpu_SSBD CPUID_INDEX_7
/* CPUID_INDEX_80000001. */
/* ECX. */
#define index_cpu_LAHF64_SAHF64 CPUID_INDEX_80000001
#define index_cpu_SVM CPUID_INDEX_80000001
#define index_cpu_LZCNT CPUID_INDEX_80000001
#define index_cpu_SSE4A CPUID_INDEX_80000001
#define index_cpu_PREFETCHW CPUID_INDEX_80000001
#define index_cpu_XOP CPUID_INDEX_80000001
#define index_cpu_LWP CPUID_INDEX_80000001
#define index_cpu_FMA4 CPUID_INDEX_80000001
#define index_cpu_TBM CPUID_INDEX_80000001
/* EDX. */
#define index_cpu_SYSCALL_SYSRET CPUID_INDEX_80000001
#define index_cpu_NX CPUID_INDEX_80000001
#define index_cpu_PAGE1GB CPUID_INDEX_80000001
#define index_cpu_RDTSCP CPUID_INDEX_80000001
#define index_cpu_LM CPUID_INDEX_80000001
/* CPUID_INDEX_D_ECX_1. */
/* EAX. */
#define index_cpu_XSAVEOPT CPUID_INDEX_D_ECX_1
#define index_cpu_XSAVEC CPUID_INDEX_D_ECX_1
#define index_cpu_XGETBV_ECX_1 CPUID_INDEX_D_ECX_1
#define index_cpu_XSAVES CPUID_INDEX_D_ECX_1
#define index_cpu_XFD CPUID_INDEX_D_ECX_1
/* CPUID_INDEX_80000007. */
/* EDX. */
#define index_cpu_INVARIANT_TSC CPUID_INDEX_80000007
/* CPUID_INDEX_80000008. */
/* EBX. */
#define index_cpu_WBNOINVD CPUID_INDEX_80000008
#define index_cpu_AMD_IBPB CPUID_INDEX_80000008
#define index_cpu_AMD_IBRS CPUID_INDEX_80000008
#define index_cpu_AMD_STIBP CPUID_INDEX_80000008
#define index_cpu_AMD_SSBD CPUID_INDEX_80000008
#define index_cpu_AMD_VIRT_SSBD CPUID_INDEX_80000008
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
/* CPUID_INDEX_7_ECX_1. */
/* EAX. */
#define index_cpu_RAO_INT CPUID_INDEX_7_ECX_1
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define index_cpu_AVX_VNNI CPUID_INDEX_7_ECX_1
#define index_cpu_AVX512_BF16 CPUID_INDEX_7_ECX_1
#define index_cpu_CMPCCXADD CPUID_INDEX_7_ECX_1
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define index_cpu_FZLRM CPUID_INDEX_7_ECX_1
#define index_cpu_FSRS CPUID_INDEX_7_ECX_1
#define index_cpu_FSRCS CPUID_INDEX_7_ECX_1
#define index_cpu_AMX_FP16 CPUID_INDEX_7_ECX_1
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define index_cpu_HRESET CPUID_INDEX_7_ECX_1
#define index_cpu_AVX_IFMA CPUID_INDEX_7_ECX_1
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define index_cpu_LAM CPUID_INDEX_7_ECX_1
#define index_cpu_AVX_VNNI_INT8 CPUID_INDEX_7_ECX_1
#define index_cpu_AVX_NE_CONVERT CPUID_INDEX_7_ECX_1
#define index_cpu_AMX_COMPLEX CPUID_INDEX_7_ECX_1
#define index_cpu_PREFETCHI CPUID_INDEX_7_ECX_1
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
/* CPUID_INDEX_19. */
/* EBX. */
#define index_cpu_AESKLE CPUID_INDEX_19
#define index_cpu_WIDE_KL CPUID_INDEX_19
/* CPUID_INDEX_14_ECX_0. */
/* EBX. */
#define index_cpu_PTWRITE CPUID_INDEX_14_ECX_0
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
/* CPUID_INDEX_1. */
/* ECX. */
#define reg_SSE3 ecx
#define reg_PCLMULQDQ ecx
#define reg_DTES64 ecx
#define reg_MONITOR ecx
#define reg_DS_CPL ecx
#define reg_VMX ecx
#define reg_SMX ecx
#define reg_EIST ecx
#define reg_TM2 ecx
#define reg_SSSE3 ecx
#define reg_CNXT_ID ecx
#define reg_SDBG ecx
#define reg_FMA ecx
#define reg_CMPXCHG16B ecx
#define reg_XTPRUPDCTRL ecx
#define reg_PDCM ecx
#define reg_INDEX_1_ECX_16 ecx
#define reg_PCID ecx
#define reg_DCA ecx
#define reg_SSE4_1 ecx
#define reg_SSE4_2 ecx
#define reg_X2APIC ecx
#define reg_MOVBE ecx
#define reg_POPCNT ecx
#define reg_TSC_DEADLINE ecx
#define reg_AES ecx
#define reg_XSAVE ecx
#define reg_OSXSAVE ecx
#define reg_AVX ecx
#define reg_F16C ecx
#define reg_RDRAND ecx
#define reg_INDEX_1_ECX_31 ecx
/* EDX. */
#define reg_FPU edx
#define reg_VME edx
#define reg_DE edx
#define reg_PSE edx
#define reg_TSC edx
#define reg_MSR edx
#define reg_PAE edx
#define reg_MCE edx
#define reg_CX8 edx
#define reg_APIC edx
#define reg_INDEX_1_EDX_10 edx
#define reg_SEP edx
#define reg_MTRR edx
#define reg_PGE edx
#define reg_MCA edx
#define reg_CMOV edx
#define reg_PAT edx
#define reg_PSE_36 edx
#define reg_PSN edx
#define reg_CLFSH edx
#define reg_INDEX_1_EDX_20 edx
#define reg_DS edx
#define reg_ACPI edx
#define reg_MMX edx
#define reg_FXSR edx
#define reg_SSE edx
#define reg_SSE2 edx
#define reg_SS edx
#define reg_HTT edx
#define reg_TM edx
#define reg_INDEX_1_EDX_30 edx
#define reg_PBE edx
/* CPUID_INDEX_7. */
/* EBX. */
#define reg_FSGSBASE ebx
#define reg_TSC_ADJUST ebx
#define reg_SGX ebx
#define reg_BMI1 ebx
#define reg_HLE ebx
#define reg_BMI2 ebx
#define reg_AVX2 ebx
#define reg_INDEX_7_EBX_6 ebx
#define reg_SMEP ebx
#define reg_ERMS ebx
#define reg_INVPCID ebx
#define reg_RTM ebx
#define reg_RDT_M ebx
#define reg_DEPR_FPU_CS_DS ebx
#define reg_MPX ebx
#define reg_RDT_A ebx
#define reg_AVX512F ebx
#define reg_AVX512DQ ebx
#define reg_RDSEED ebx
#define reg_ADX ebx
#define reg_SMAP ebx
#define reg_AVX512_IFMA ebx
#define reg_INDEX_7_EBX_22 ebx
#define reg_CLFLUSHOPT ebx
#define reg_CLWB ebx
#define reg_TRACE ebx
#define reg_AVX512PF ebx
#define reg_AVX512ER ebx
#define reg_AVX512CD ebx
#define reg_SHA ebx
#define reg_AVX512BW ebx
#define reg_AVX512VL ebx
/* ECX. */
#define reg_PREFETCHWT1 ecx
#define reg_AVX512_VBMI ecx
#define reg_UMIP ecx
#define reg_PKU ecx
#define reg_OSPKE ecx
#define reg_WAITPKG ecx
#define reg_AVX512_VBMI2 ecx
#define reg_SHSTK ecx
#define reg_GFNI ecx
#define reg_VAES ecx
#define reg_VPCLMULQDQ ecx
#define reg_AVX512_VNNI ecx
#define reg_AVX512_BITALG ecx
#define reg_INDEX_7_ECX_13 ecx
#define reg_AVX512_VPOPCNTDQ ecx
#define reg_INDEX_7_ECX_15 ecx
#define reg_INDEX_7_ECX_16 ecx
#define reg_RDPID ecx
#define reg_KL ecx
#define reg_INDEX_7_ECX_24 ecx
#define reg_CLDEMOTE ecx
#define reg_INDEX_7_ECX_26 ecx
#define reg_MOVDIRI ecx
#define reg_MOVDIR64B ecx
#define reg_ENQCMD ecx
#define reg_SGX_LC ecx
#define reg_PKS ecx
/* EDX. */
#define reg_INDEX_7_EDX_0 edx
#define reg_INDEX_7_EDX_1 edx
#define reg_AVX512_4VNNIW edx
#define reg_AVX512_4FMAPS edx
#define reg_FSRM edx
#define reg_UINTR edx
#define reg_INDEX_7_EDX_6 edx
#define reg_INDEX_7_EDX_7 edx
#define reg_AVX512_VP2INTERSECT edx
#define reg_INDEX_7_EDX_9 edx
#define reg_MD_CLEAR edx
#define reg_RTM_ALWAYS_ABORT edx
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define reg_INDEX_7_EDX_12 edx
#define reg_INDEX_7_EDX_13 edx
#define reg_SERIALIZE edx
#define reg_HYBRID edx
#define reg_TSXLDTRK edx
#define reg_INDEX_7_EDX_17 edx
#define reg_PCONFIG edx
#define reg_INDEX_7_EDX_19 edx
#define reg_IBT edx
#define reg_INDEX_7_EDX_21 edx
#define reg_AMX_BF16 edx
#define reg_AVX512_FP16 edx
#define reg_AMX_TILE edx
#define reg_AMX_INT8 edx
#define reg_IBRS_IBPB edx
#define reg_STIBP edx
#define reg_L1D_FLUSH edx
#define reg_ARCH_CAPABILITIES edx
#define reg_CORE_CAPABILITIES edx
#define reg_SSBD edx
/* CPUID_INDEX_80000001. */
/* ECX. */
#define reg_LAHF64_SAHF64 ecx
#define reg_SVM ecx
#define reg_LZCNT ecx
#define reg_SSE4A ecx
#define reg_PREFETCHW ecx
#define reg_XOP ecx
#define reg_LWP ecx
#define reg_FMA4 ecx
#define reg_TBM ecx
/* EDX. */
#define reg_SYSCALL_SYSRET edx
#define reg_NX edx
#define reg_PAGE1GB edx
#define reg_RDTSCP edx
#define reg_LM edx
/* CPUID_INDEX_D_ECX_1. */
/* EAX. */
#define reg_XSAVEOPT eax
#define reg_XSAVEC eax
#define reg_XGETBV_ECX_1 eax
#define reg_XSAVES eax
#define reg_XFD eax
/* CPUID_INDEX_80000007. */
/* EDX. */
#define reg_INVARIANT_TSC edx
/* CPUID_INDEX_80000008. */
/* EBX. */
#define reg_WBNOINVD ebx
#define reg_AMD_IBPB ebx
#define reg_AMD_IBRS ebx
#define reg_AMD_STIBP ebx
#define reg_AMD_SSBD ebx
#define reg_AMD_VIRT_SSBD ebx
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
/* CPUID_INDEX_7_ECX_1. */
/* EAX. */
#define reg_RAO_INT eax
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define reg_AVX_VNNI eax
#define reg_AVX512_BF16 eax
#define reg_CMPCCXADD eax
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define reg_FZLRM eax
#define reg_FSRS eax
#define reg_FSRCS eax
#define reg_AMX_FP16 eax
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define reg_HRESET eax
#define reg_AVX_IFMA eax
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
#define reg_LAM eax
/* EDX. */
#define reg_AVX_VNNI_INT8 edx
#define reg_AVX_NE_CONVERT edx
#define reg_AMX_COMPLEX edx
#define reg_PREFETCHI edx
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
/* CPUID_INDEX_19. */
/* EBX. */
#define reg_AESKLE ebx
#define reg_WIDE_KL ebx
/* CPUID_INDEX_14_ECX_0. */
/* EBX. */
#define reg_PTWRITE ebx
/* PREFERRED_FEATURE_INDEX_1. First define the bitindex values
sequentially, then define the bit_arch* and index_arch_* lookup
constants. */
enum
{
#define BIT(x) _bitindex_arch_##x ,
#include "cpu-features-preferred_feature_index_1.def"
#undef BIT
};
enum
{
#define BIT(x) \
bit_arch_##x = 1u << _bitindex_arch_##x , \
index_arch_##x = PREFERRED_FEATURE_INDEX_1,
#include "cpu-features-preferred_feature_index_1.def"
#undef BIT
};
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
/* XCR0 Feature flags. */
#define bit_XMM_state (1u << 1)
#define bit_YMM_state (1u << 2)
#define bit_Opmask_state (1u << 5)
#define bit_ZMM0_15_state (1u << 6)
#define bit_ZMM16_31_state (1u << 7)
#define bit_XTILECFG_state (1u << 17)
#define bit_XTILEDATA_state (1u << 18)
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
enum cpu_features_kind
{
arch_kind_unknown = 0,
arch_kind_intel,
arch_kind_amd,
arch_kind_zhaoxin,
arch_kind_other
};
struct cpu_features_basic
{
enum cpu_features_kind kind;
int max_cpuid;
unsigned int family;
unsigned int model;
unsigned int stepping;
};
struct cpuid_registers
{
unsigned int eax;
unsigned int ebx;
unsigned int ecx;
unsigned int edx;
};
struct cpuid_feature_internal
{
union
{
unsigned int cpuid_array[4];
struct cpuid_registers cpuid;
};
union
{
unsigned int active_array[4];
struct cpuid_registers active;
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
};
};
/* NB: When adding new fields, update sysdeps/x86/dl-diagnostics-cpu.c
to print them. */
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
struct cpu_features
{
struct cpu_features_basic basic;
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
struct cpuid_feature_internal features[CPUID_INDEX_MAX];
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
unsigned int preferred[PREFERRED_FEATURE_INDEX_MAX];
x86: Support GNU_PROPERTY_X86_ISA_1_V[234] marker [BZ #26717] GCC 11 supports -march=x86-64-v[234] to enable x86 micro-architecture ISA levels: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97250 and -mneeded to emit GNU_PROPERTY_X86_ISA_1_NEEDED property with GNU_PROPERTY_X86_ISA_1_V[234] marker: https://gitlab.com/x86-psABIs/x86-64-ABI/-/merge_requests/13 Binutils support for GNU_PROPERTY_X86_ISA_1_V[234] marker were added by commit b0ab06937385e0ae25cebf1991787d64f439bf12 Author: H.J. Lu <hjl.tools@gmail.com> Date: Fri Oct 30 06:49:57 2020 -0700 x86: Support GNU_PROPERTY_X86_ISA_1_BASELINE marker and commit 32930e4edbc06bc6f10c435dbcc63131715df678 Author: H.J. Lu <hjl.tools@gmail.com> Date: Fri Oct 9 05:05:57 2020 -0700 x86: Support GNU_PROPERTY_X86_ISA_1_V[234] marker GNU_PROPERTY_X86_ISA_1_NEEDED property in x86 ELF binaries indicate the micro-architecture ISA level required to execute the binary. The marker must be added by programmers explicitly in one of 3 ways: 1. Pass -mneeded to GCC. 2. Add the marker in the linker inputs as this patch does. 3. Pass -z x86-64-v[234] to the linker. Add GNU_PROPERTY_X86_ISA_1_BASELINE and GNU_PROPERTY_X86_ISA_1_V[234] marker support to ld.so if binutils 2.32 or newer is used to build glibc: 1. Add GNU_PROPERTY_X86_ISA_1_BASELINE and GNU_PROPERTY_X86_ISA_1_V[234] markers to elf.h. 2. Add GNU_PROPERTY_X86_ISA_1_BASELINE and GNU_PROPERTY_X86_ISA_1_V[234] marker to abi-note.o based on the ISA level used to compile abi-note.o, assuming that the same ISA level is used to compile the whole glibc. 3. Add isa_1 to cpu_features to record the supported x86 ISA level. 4. Rename _dl_process_cet_property_note to _dl_process_property_note and add GNU_PROPERTY_X86_ISA_1_V[234] marker detection. 5. Update _rtld_main_check and _dl_open_check to check loaded objects with the incompatible ISA level. 6. Add a testcase to verify that dlopen an x86-64-v4 shared object fails on lesser platforms. 7. Use <get-isa-level.h> in dl-hwcaps-subdirs.c and tst-glibc-hwcaps.c. Tested under i686, x32 and x86-64 modes on x86-64-v2, x86-64-v3 and x86-64-v4 machines. Marked elf/tst-isa-level-1 with x86-64-v4, ran it on x86-64-v3 machine and got: [hjl@gnu-cfl-2 build-x86_64-linux]$ ./elf/tst-isa-level-1 ./elf/tst-isa-level-1: CPU ISA level is lower than required [hjl@gnu-cfl-2 build-x86_64-linux]$
2020-10-09 13:06:56 +00:00
/* X86 micro-architecture ISA levels. */
unsigned int isa_1;
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
/* The state size for XSAVEC or XSAVE. The type must be unsigned long
int so that we use
sub xsave_state_size_offset(%rip) %RSP_LP
in _dl_runtime_resolve. */
unsigned long int xsave_state_size;
/* The full state size for XSAVE when XSAVEC is disabled by
GLIBC_TUNABLES=glibc.cpu.hwcaps=-XSAVEC
*/
unsigned int xsave_state_full_size;
/* Data cache size for use in memory and string routines, typically
L1 size. */
unsigned long int data_cache_size;
/* Shared cache size for use in memory and string routines, typically
L2 or L3 size. */
unsigned long int shared_cache_size;
/* Threshold to use non temporal store. */
unsigned long int non_temporal_threshold;
/* Threshold to use "rep movsb". */
unsigned long int rep_movsb_threshold;
/* Threshold to stop using "rep movsb". */
unsigned long int rep_movsb_stop_threshold;
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
/* Threshold to use "rep stosb". */
unsigned long int rep_stosb_threshold;
/* _SC_LEVEL1_ICACHE_SIZE. */
unsigned long int level1_icache_size;
/* _SC_LEVEL1_ICACHE_LINESIZE. */
unsigned long int level1_icache_linesize;
/* _SC_LEVEL1_DCACHE_SIZE. */
unsigned long int level1_dcache_size;
/* _SC_LEVEL1_DCACHE_ASSOC. */
unsigned long int level1_dcache_assoc;
/* _SC_LEVEL1_DCACHE_LINESIZE. */
unsigned long int level1_dcache_linesize;
/* _SC_LEVEL2_CACHE_ASSOC. */
unsigned long int level2_cache_size;
/* _SC_LEVEL2_DCACHE_ASSOC. */
unsigned long int level2_cache_assoc;
/* _SC_LEVEL2_CACHE_LINESIZE. */
unsigned long int level2_cache_linesize;
/* /_SC_LEVEL3_CACHE_SIZE. */
unsigned long int level3_cache_size;
/* _SC_LEVEL3_CACHE_ASSOC. */
unsigned long int level3_cache_assoc;
/* _SC_LEVEL3_CACHE_LINESIZE. */
unsigned long int level3_cache_linesize;
/* /_SC_LEVEL4_CACHE_SIZE. */
unsigned long int level4_cache_size;
/* When no user non_temporal_threshold is specified. We default to
cachesize / cachesize_non_temporal_divisor. */
unsigned long int cachesize_non_temporal_divisor;
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
};
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
/* Get a pointer to the CPU features structure. */
extern const struct cpu_features *_dl_x86_get_cpu_features (void)
__attribute__ ((pure));
#define __get_cpu_features() _dl_x86_get_cpu_features()
#if defined (_LIBC) && !IS_IN (nonlib)
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
/* Unused for x86. */
# define INIT_ARCH()
<sys/platform/x86.h>: Remove the C preprocessor magic In <sys/platform/x86.h>, define CPU features as enum instead of using the C preprocessor magic to make it easier to wrap this functionality in other languages. Move the C preprocessor magic to internal header for better GCC codegen when more than one features are checked in a single expression as in x86-64 dl-hwcaps-subdirs.c. 1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX. 2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h. 3. Remove struct cpu_features and __x86_get_cpu_features from <sys/platform/x86.h>. 4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it in libc. 5. Make __get_cpu_features() private to glibc. 6. Replace __x86_get_cpu_features(N) with __get_cpu_features(). 7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE. 8. Use a single enum index for each CPU feature detection. 9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf. 10. Return zero struct cpuid_feature for the older glibc binary with a smaller CPUID_INDEX_MAX [BZ #27104]. 11. Inside glibc, use the C preprocessor magic so that cpu_features data can be loaded just once leading to more compact code for glibc. 256 bits are used for each CPUID leaf. Some leaves only contain a few features. We can add exceptions to such leaves. But it will increase code sizes and it is harder to provide backward/forward compatibilities when new features are added to such leaves in the future. When new leaves are added, _rtld_global_ro offsets will change which leads to race condition during in-place updates. We may avoid in-place updates by 1. Rename the old glibc. 2. Install the new glibc. 3. Remove the old glibc. NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2020-12-25 15:30:46 +00:00
# define _dl_x86_get_cpu_features() (&GLRO(dl_x86_cpu_features))
extern void _dl_x86_init_cpu_features (void) attribute_hidden;
#endif
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
#ifdef __x86_64__
# define HAS_CPUID 1
#elif (defined __i586__ || defined __pentium__ \
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
|| defined __geode__ || defined __k6__)
# define HAS_CPUID 1
# define HAS_I586 1
# define HAS_I686 HAS_ARCH_FEATURE (I686)
#elif defined __i486__
# define HAS_CPUID 0
# define HAS_I586 HAS_ARCH_FEATURE (I586)
# define HAS_I686 HAS_ARCH_FEATURE (I686)
#else
# define HAS_CPUID 1
# define HAS_I586 1
# define HAS_I686 1
#endif
x86: Install <sys/platform/x86.h> [BZ #26124] Install <sys/platform/x86.h> so that programmers can do #if __has_include(<sys/platform/x86.h>) #include <sys/platform/x86.h> #endif ... if (CPU_FEATURE_USABLE (SSE2)) ... if (CPU_FEATURE_USABLE (AVX2)) ... <sys/platform/x86.h> exports only: enum { COMMON_CPUID_INDEX_1 = 0, COMMON_CPUID_INDEX_7, COMMON_CPUID_INDEX_80000001, COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007, COMMON_CPUID_INDEX_80000008, COMMON_CPUID_INDEX_7_ECX_1, /* Keep the following line at the end. */ COMMON_CPUID_INDEX_MAX }; struct cpuid_features { struct cpuid_registers cpuid; struct cpuid_registers usable; }; struct cpu_features { struct cpu_features_basic basic; struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; }; /* Get a pointer to the CPU features structure. */ extern const struct cpu_features *__x86_get_cpu_features (unsigned int max) __attribute__ ((const)); Since all feature checks are done through macros, programs compiled with a newer <sys/platform/x86.h> are compatible with the older glibc binaries as long as the layout of struct cpu_features is identical. The features array can be expanded with backward binary compatibility for both .o and .so files. When COMMON_CPUID_INDEX_MAX is increased to support new processor features, __x86_get_cpu_features in the older glibc binaries returns NULL and HAS_CPU_FEATURE/CPU_FEATURE_USABLE return false on the new processor feature. No new symbol version is neeeded. Both CPU_FEATURE_USABLE and HAS_CPU_FEATURE are provided. HAS_CPU_FEATURE can be used to identify processor features. Note: Although GCC has __builtin_cpu_supports, it only supports a subset of <sys/platform/x86.h> and it is equivalent to CPU_FEATURE_USABLE. It doesn't support HAS_CPU_FEATURE.
2020-06-30 01:30:54 +00:00
#endif /* include/cpu-features.h */