glibc/sysdeps/x86/cpu-features.h
H.J. Lu fb0f7a6755 X86-64: Add _dl_runtime_resolve_avx[512]_{opt|slow} [BZ #20508]
There is transition penalty when SSE instructions are mixed with 256-bit
AVX or 512-bit AVX512 load instructions.  Since _dl_runtime_resolve_avx
and _dl_runtime_profile_avx512 save/restore 256-bit YMM/512-bit ZMM
registers, there is transition penalty when SSE instructions are used
with lazy binding on AVX and AVX512 processors.

To avoid SSE transition penalty, if only the lower 128 bits of the first
8 vector registers are non-zero, we can preserve %xmm0 - %xmm7 registers
with the zero upper bits.

For AVX and AVX512 processors which support XGETBV with ECX == 1, we can
use XGETBV with ECX == 1 to check if the upper 128 bits of YMM registers
or the upper 256 bits of ZMM registers are zero.  We can restore only the
non-zero portion of vector registers with AVX/AVX512 load instructions
which will zero-extend upper bits of vector registers.

This patch adds _dl_runtime_resolve_sse_vex which saves and restores
XMM registers with 128-bit AVX store/load instructions.  It is used to
preserve YMM/ZMM registers when only the lower 128 bits are non-zero.
_dl_runtime_resolve_avx_opt and _dl_runtime_resolve_avx512_opt are added
and used on AVX/AVX512 processors supporting XGETBV with ECX == 1 so
that we store and load only the non-zero portion of vector registers.
This avoids SSE transition penalty caused by _dl_runtime_resolve_avx and
_dl_runtime_profile_avx512 when only the lower 128 bits of vector
registers are used.

_dl_runtime_resolve_avx_slow is added and used for AVX processors which
don't support XGETBV with ECX == 1.  Since there is no SSE transition
penalty on AVX512 processors which don't support XGETBV with ECX == 1,
_dl_runtime_resolve_avx512_slow isn't provided.

	[BZ #20495]
	[BZ #20508]
	* sysdeps/x86/cpu-features.c (init_cpu_features): For Intel
	processors, set Use_dl_runtime_resolve_slow and set
	Use_dl_runtime_resolve_opt if XGETBV suports ECX == 1.
	* sysdeps/x86/cpu-features.h (bit_arch_Use_dl_runtime_resolve_opt):
	New.
	(bit_arch_Use_dl_runtime_resolve_slow): Likewise.
	(index_arch_Use_dl_runtime_resolve_opt): Likewise.
	(index_arch_Use_dl_runtime_resolve_slow): Likewise.
	* sysdeps/x86_64/dl-machine.h (elf_machine_runtime_setup): Use
	_dl_runtime_resolve_avx512_opt and _dl_runtime_resolve_avx_opt
	if Use_dl_runtime_resolve_opt is set.  Use
	_dl_runtime_resolve_slow if Use_dl_runtime_resolve_slow is set.
	* sysdeps/x86_64/dl-trampoline.S: Include <cpu-features.h>.
	(_dl_runtime_resolve_opt): New.  Defined for AVX and AVX512.
	(_dl_runtime_resolve): Add one for _dl_runtime_resolve_sse_vex.
	* sysdeps/x86_64/dl-trampoline.h (_dl_runtime_resolve_avx_slow):
	New.
	(_dl_runtime_resolve_opt): Likewise.
	(_dl_runtime_profile): Define only if _dl_runtime_profile is
	defined.
2016-09-06 08:51:07 -07:00

318 lines
12 KiB
C

/* This file is part of the GNU C Library.
Copyright (C) 2008-2016 Free Software Foundation, Inc.
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
<http://www.gnu.org/licenses/>. */
#ifndef cpu_features_h
#define cpu_features_h
#define bit_arch_Fast_Rep_String (1 << 0)
#define bit_arch_Fast_Copy_Backward (1 << 1)
#define bit_arch_Slow_BSF (1 << 2)
#define bit_arch_Fast_Unaligned_Load (1 << 4)
#define bit_arch_Prefer_PMINUB_for_stringop (1 << 5)
#define bit_arch_AVX_Usable (1 << 6)
#define bit_arch_FMA_Usable (1 << 7)
#define bit_arch_FMA4_Usable (1 << 8)
#define bit_arch_Slow_SSE4_2 (1 << 9)
#define bit_arch_AVX2_Usable (1 << 10)
#define bit_arch_AVX_Fast_Unaligned_Load (1 << 11)
#define bit_arch_AVX512F_Usable (1 << 12)
#define bit_arch_AVX512DQ_Usable (1 << 13)
#define bit_arch_I586 (1 << 14)
#define bit_arch_I686 (1 << 15)
#define bit_arch_Prefer_MAP_32BIT_EXEC (1 << 16)
#define bit_arch_Prefer_No_VZEROUPPER (1 << 17)
#define bit_arch_Fast_Unaligned_Copy (1 << 18)
#define bit_arch_Prefer_ERMS (1 << 19)
#define bit_arch_Use_dl_runtime_resolve_opt (1 << 20)
#define bit_arch_Use_dl_runtime_resolve_slow (1 << 21)
/* CPUID Feature flags. */
/* COMMON_CPUID_INDEX_1. */
#define bit_cpu_CX8 (1 << 8)
#define bit_cpu_CMOV (1 << 15)
#define bit_cpu_SSE2 (1 << 26)
#define bit_cpu_SSSE3 (1 << 9)
#define bit_cpu_SSE4_1 (1 << 19)
#define bit_cpu_SSE4_2 (1 << 20)
#define bit_cpu_OSXSAVE (1 << 27)
#define bit_cpu_AVX (1 << 28)
#define bit_cpu_POPCOUNT (1 << 23)
#define bit_cpu_FMA (1 << 12)
#define bit_cpu_FMA4 (1 << 16)
#define bit_cpu_HTT (1 << 28)
/* COMMON_CPUID_INDEX_7. */
#define bit_cpu_ERMS (1 << 9)
#define bit_cpu_RTM (1 << 11)
#define bit_cpu_AVX2 (1 << 5)
#define bit_cpu_AVX512F (1 << 16)
#define bit_cpu_AVX512DQ (1 << 17)
/* XCR0 Feature flags. */
#define bit_XMM_state (1 << 1)
#define bit_YMM_state (1 << 2)
#define bit_Opmask_state (1 << 5)
#define bit_ZMM0_15_state (1 << 6)
#define bit_ZMM16_31_state (1 << 7)
/* The integer bit array index for the first set of internal feature bits. */
#define FEATURE_INDEX_1 0
/* The current maximum size of the feature integer bit array. */
#define FEATURE_INDEX_MAX 1
#ifdef __ASSEMBLER__
# include <cpu-features-offsets.h>
# define index_cpu_CX8 COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_EDX_OFFSET
# define index_cpu_CMOV COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_EDX_OFFSET
# define index_cpu_SSE2 COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_EDX_OFFSET
# define index_cpu_SSSE3 COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET
# define index_cpu_SSE4_1 COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET
# define index_cpu_SSE4_2 COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET
# define index_cpu_AVX COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET
# define index_cpu_AVX2 COMMON_CPUID_INDEX_7*CPUID_SIZE+CPUID_EBX_OFFSET
# define index_cpu_ERMS COMMON_CPUID_INDEX_7*CPUID_SIZE+CPUID_EBX_OFFSET
# define index_arch_Fast_Rep_String FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Fast_Copy_Backward FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Slow_BSF FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Fast_Unaligned_Load FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Prefer_PMINUB_for_stringop FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_AVX_Usable FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_FMA_Usable FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_FMA4_Usable FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Slow_SSE4_2 FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_AVX2_Usable FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_AVX_Fast_Unaligned_Load FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_AVX512F_Usable FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_AVX512DQ_Usable FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_I586 FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_I686 FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Prefer_MAP_32BIT_EXEC FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Prefer_No_VZEROUPPER FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Fast_Unaligned_Copy FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Prefer_ERMS FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Use_dl_runtime_resolve_opt FEATURE_INDEX_1*FEATURE_SIZE
# define index_arch_Use_dl_runtime_resolve_slow FEATURE_INDEX_1*FEATURE_SIZE
# if defined (_LIBC) && !IS_IN (nonlib)
# ifdef __x86_64__
# ifdef SHARED
# if IS_IN (rtld)
# define LOAD_RTLD_GLOBAL_RO_RDX
# define HAS_FEATURE(offset, field, name) \
testl $(bit_##field##_##name), \
_rtld_local_ro+offset+(index_##field##_##name)(%rip)
# else
# define LOAD_RTLD_GLOBAL_RO_RDX \
mov _rtld_global_ro@GOTPCREL(%rip), %RDX_LP
# define HAS_FEATURE(offset, field, name) \
testl $(bit_##field##_##name), \
RTLD_GLOBAL_RO_DL_X86_CPU_FEATURES_OFFSET+offset+(index_##field##_##name)(%rdx)
# endif
# else /* SHARED */
# define LOAD_RTLD_GLOBAL_RO_RDX
# define HAS_FEATURE(offset, field, name) \
testl $(bit_##field##_##name), \
_dl_x86_cpu_features+offset+(index_##field##_##name)(%rip)
# endif /* !SHARED */
# else /* __x86_64__ */
# ifdef SHARED
# define LOAD_FUNC_GOT_EAX(func) \
leal func@GOTOFF(%edx), %eax
# if IS_IN (rtld)
# define LOAD_GOT_AND_RTLD_GLOBAL_RO \
LOAD_PIC_REG(dx)
# define HAS_FEATURE(offset, field, name) \
testl $(bit_##field##_##name), \
offset+(index_##field##_##name)+_rtld_local_ro@GOTOFF(%edx)
# else
# define LOAD_GOT_AND_RTLD_GLOBAL_RO \
LOAD_PIC_REG(dx); \
mov _rtld_global_ro@GOT(%edx), %ecx
# define HAS_FEATURE(offset, field, name) \
testl $(bit_##field##_##name), \
RTLD_GLOBAL_RO_DL_X86_CPU_FEATURES_OFFSET+offset+(index_##field##_##name)(%ecx)
# endif
# else /* SHARED */
# define LOAD_FUNC_GOT_EAX(func) \
leal func, %eax
# define LOAD_GOT_AND_RTLD_GLOBAL_RO
# define HAS_FEATURE(offset, field, name) \
testl $(bit_##field##_##name), \
_dl_x86_cpu_features+offset+(index_##field##_##name)
# endif /* !SHARED */
# endif /* !__x86_64__ */
# else /* _LIBC && !nonlib */
# error "Sorry, <cpu-features.h> is unimplemented for assembler"
# endif /* !_LIBC || nonlib */
/* HAS_* evaluates to true if we may use the feature at runtime. */
# define HAS_CPU_FEATURE(name) HAS_FEATURE (CPUID_OFFSET, cpu, name)
# define HAS_ARCH_FEATURE(name) HAS_FEATURE (FEATURE_OFFSET, arch, name)
#else /* __ASSEMBLER__ */
enum
{
COMMON_CPUID_INDEX_1 = 0,
COMMON_CPUID_INDEX_7,
COMMON_CPUID_INDEX_80000001, /* for AMD */
/* Keep the following line at the end. */
COMMON_CPUID_INDEX_MAX
};
struct cpu_features
{
enum cpu_features_kind
{
arch_kind_unknown = 0,
arch_kind_intel,
arch_kind_amd,
arch_kind_other
} kind;
int max_cpuid;
struct cpuid_registers
{
unsigned int eax;
unsigned int ebx;
unsigned int ecx;
unsigned int edx;
} cpuid[COMMON_CPUID_INDEX_MAX];
unsigned int family;
unsigned int model;
unsigned int feature[FEATURE_INDEX_MAX];
};
/* Used from outside of glibc to get access to the CPU features
structure. */
extern const struct cpu_features *__get_cpu_features (void)
__attribute__ ((const));
# if defined (_LIBC) && !IS_IN (nonlib)
/* Unused for x86. */
# define INIT_ARCH()
# define __get_cpu_features() (&GLRO(dl_x86_cpu_features))
# endif
/* Only used directly in cpu-features.c. */
# define CPU_FEATURES_CPU_P(ptr, name) \
((ptr->cpuid[index_cpu_##name].reg_##name & (bit_cpu_##name)) != 0)
# define CPU_FEATURES_ARCH_P(ptr, name) \
((ptr->feature[index_arch_##name] & (bit_arch_##name)) != 0)
/* HAS_* evaluates to true if we may use the feature at runtime. */
# define HAS_CPU_FEATURE(name) \
CPU_FEATURES_CPU_P (__get_cpu_features (), name)
# define HAS_ARCH_FEATURE(name) \
CPU_FEATURES_ARCH_P (__get_cpu_features (), name)
# define index_cpu_CX8 COMMON_CPUID_INDEX_1
# define index_cpu_CMOV COMMON_CPUID_INDEX_1
# define index_cpu_SSE2 COMMON_CPUID_INDEX_1
# define index_cpu_SSSE3 COMMON_CPUID_INDEX_1
# define index_cpu_SSE4_1 COMMON_CPUID_INDEX_1
# define index_cpu_SSE4_2 COMMON_CPUID_INDEX_1
# define index_cpu_AVX COMMON_CPUID_INDEX_1
# define index_cpu_AVX2 COMMON_CPUID_INDEX_7
# define index_cpu_AVX512F COMMON_CPUID_INDEX_7
# define index_cpu_AVX512DQ COMMON_CPUID_INDEX_7
# define index_cpu_ERMS COMMON_CPUID_INDEX_7
# define index_cpu_RTM COMMON_CPUID_INDEX_7
# define index_cpu_FMA COMMON_CPUID_INDEX_1
# define index_cpu_FMA4 COMMON_CPUID_INDEX_80000001
# define index_cpu_POPCOUNT COMMON_CPUID_INDEX_1
# define index_cpu_OSXSAVE COMMON_CPUID_INDEX_1
# define index_cpu_HTT COMMON_CPUID_INDEX_1
# define reg_CX8 edx
# define reg_CMOV edx
# define reg_SSE2 edx
# define reg_SSSE3 ecx
# define reg_SSE4_1 ecx
# define reg_SSE4_2 ecx
# define reg_AVX ecx
# define reg_AVX2 ebx
# define reg_AVX512F ebx
# define reg_AVX512DQ ebx
# define reg_ERMS ebx
# define reg_RTM ebx
# define reg_FMA ecx
# define reg_FMA4 ecx
# define reg_POPCOUNT ecx
# define reg_OSXSAVE ecx
# define reg_HTT edx
# define index_arch_Fast_Rep_String FEATURE_INDEX_1
# define index_arch_Fast_Copy_Backward FEATURE_INDEX_1
# define index_arch_Slow_BSF FEATURE_INDEX_1
# define index_arch_Fast_Unaligned_Load FEATURE_INDEX_1
# define index_arch_Prefer_PMINUB_for_stringop FEATURE_INDEX_1
# define index_arch_AVX_Usable FEATURE_INDEX_1
# define index_arch_FMA_Usable FEATURE_INDEX_1
# define index_arch_FMA4_Usable FEATURE_INDEX_1
# define index_arch_Slow_SSE4_2 FEATURE_INDEX_1
# define index_arch_AVX2_Usable FEATURE_INDEX_1
# define index_arch_AVX_Fast_Unaligned_Load FEATURE_INDEX_1
# define index_arch_AVX512F_Usable FEATURE_INDEX_1
# define index_arch_AVX512DQ_Usable FEATURE_INDEX_1
# define index_arch_I586 FEATURE_INDEX_1
# define index_arch_I686 FEATURE_INDEX_1
# define index_arch_Prefer_MAP_32BIT_EXEC FEATURE_INDEX_1
# define index_arch_Prefer_No_VZEROUPPER FEATURE_INDEX_1
# define index_arch_Fast_Unaligned_Copy FEATURE_INDEX_1
# define index_arch_Prefer_ERMS FEATURE_INDEX_1
# define index_arch_Use_dl_runtime_resolve_opt FEATURE_INDEX_1
# define index_arch_Use_dl_runtime_resolve_slow FEATURE_INDEX_1
#endif /* !__ASSEMBLER__ */
#ifdef __x86_64__
# define HAS_CPUID 1
#elif defined __i586__ || defined __pentium__
# define HAS_CPUID 1
# define HAS_I586 1
# define HAS_I686 HAS_ARCH_FEATURE (I686)
#elif (defined __i686__ || defined __pentiumpro__ \
|| defined __pentium4__ || defined __nocona__ \
|| defined __atom__ || defined __core2__ \
|| defined __corei7__ || defined __corei7_avx__ \
|| defined __core_avx2__ || defined __nehalem__ \
|| defined __sandybridge__ || defined __haswell__ \
|| defined __knl__ || defined __bonnell__ \
|| defined __silvermont__ \
|| defined __k6__ || defined __k8__ \
|| defined __athlon__ || defined __amdfam10__ \
|| defined __bdver1__ || defined __bdver2__ \
|| defined __bdver3__ || defined __bdver4__ \
|| defined __btver1__ || defined __btver2__)
# define HAS_CPUID 1
# define HAS_I586 1
# define HAS_I686 1
#else
# define HAS_CPUID 0
# define HAS_I586 HAS_ARCH_FEATURE (I586)
# define HAS_I686 HAS_ARCH_FEATURE (I686)
#endif
#endif /* cpu_features_h */