diff options
| author | H.J. Lu <hjl.tools@gmail.com> | 2020-06-29 16:36:08 -0700 |
|---|---|---|
| committer | H.J. Lu <hjl.tools@gmail.com> | 2020-07-13 06:05:16 -0700 |
| commit | 107e6a3c2212ba7a3a4ec7cae8d82d73f7c95d0b (patch) | |
| tree | b23f1c5ba166bd28f519fa36f225a31407b45270 /sysdeps | |
| parent | 10b01bd4529336bffc2c398ce43a171ed94aacc7 (diff) | |
| download | glibc-107e6a3c2212ba7a3a4ec7cae8d82d73f7c95d0b.zip glibc-107e6a3c2212ba7a3a4ec7cae8d82d73f7c95d0b.tar.gz glibc-107e6a3c2212ba7a3a4ec7cae8d82d73f7c95d0b.tar.bz2 | |
x86: Support usable check for all CPU features
Support usable check for all CPU features with the following changes:
1. Change struct cpu_features to
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];
unsigned int preferred[PREFERRED_FEATURE_INDEX_MAX];
...
};
so that there is a usable bit for each cpuid bit.
2. After the cpuid bits have been initialized, copy the known bits to the
usable bits. EAX/EBX from INDEX_1 and EAX from INDEX_7 aren't used for
CPU feature detection.
3. Clear the usable bits which require OS support.
4. If the feature is supported by OS, copy its cpuid bit to its usable
bit.
5. Replace HAS_CPU_FEATURE and CPU_FEATURES_CPU_P with CPU_FEATURE_USABLE
and CPU_FEATURE_USABLE_P to check if a feature is usable.
6. Add DEPR_FPU_CS_DS for INDEX_7_EBX_13.
7. Unset MPX feature since it has been deprecated.
The results are
1. If the feature is known and doesn't requre OS support, its usable bit
is copied from the cpuid bit.
2. Otherwise, its usable bit is copied from the cpuid bit only if the
feature is known to supported by OS.
3. CPU_FEATURE_USABLE/CPU_FEATURE_USABLE_P are used to check if the
feature can be used.
4. HAS_CPU_FEATURE/CPU_FEATURE_CPU_P are used to check if CPU supports
the feature.
Diffstat (limited to 'sysdeps')
63 files changed, 854 insertions, 732 deletions
diff --git a/sysdeps/i386/fpu/fclrexcpt.c b/sysdeps/i386/fpu/fclrexcpt.c index 7bf7dd0..7dc357f 100644 --- a/sysdeps/i386/fpu/fclrexcpt.c +++ b/sysdeps/i386/fpu/fclrexcpt.c @@ -41,7 +41,7 @@ __feclearexcept (int excepts) __asm__ ("fldenv %0" : : "m" (*&temp)); /* If the CPU supports SSE, we clear the MXCSR as well. */ - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) { unsigned int xnew_exc; diff --git a/sysdeps/i386/fpu/fedisblxcpt.c b/sysdeps/i386/fpu/fedisblxcpt.c index 0e518f7..5399bc1 100644 --- a/sysdeps/i386/fpu/fedisblxcpt.c +++ b/sysdeps/i386/fpu/fedisblxcpt.c @@ -38,7 +38,7 @@ fedisableexcept (int excepts) __asm__ ("fldcw %0" : : "m" (*&new_exc)); /* If the CPU supports SSE we set the MXCSR as well. */ - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) { unsigned int xnew_exc; diff --git a/sysdeps/i386/fpu/feenablxcpt.c b/sysdeps/i386/fpu/feenablxcpt.c index b1f7081..b9d7e65 100644 --- a/sysdeps/i386/fpu/feenablxcpt.c +++ b/sysdeps/i386/fpu/feenablxcpt.c @@ -38,7 +38,7 @@ feenableexcept (int excepts) __asm__ ("fldcw %0" : : "m" (*&new_exc)); /* If the CPU supports SSE we set the MXCSR as well. */ - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) { unsigned int xnew_exc; diff --git a/sysdeps/i386/fpu/fegetenv.c b/sysdeps/i386/fpu/fegetenv.c index cb6ef35..637bc85 100644 --- a/sysdeps/i386/fpu/fegetenv.c +++ b/sysdeps/i386/fpu/fegetenv.c @@ -31,7 +31,7 @@ __fegetenv (fenv_t *envp) would block all exceptions. */ __asm__ ("fldenv %0" : : "m" (*envp)); - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) __asm__ ("stmxcsr %0" : "=m" (envp->__eip)); /* Success. */ diff --git a/sysdeps/i386/fpu/fegetmode.c b/sysdeps/i386/fpu/fegetmode.c index e147689..e5154ea 100644 --- a/sysdeps/i386/fpu/fegetmode.c +++ b/sysdeps/i386/fpu/fegetmode.c @@ -26,7 +26,7 @@ int fegetmode (femode_t *modep) { _FPU_GETCW (modep->__control_word); - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) __asm__ ("stmxcsr %0" : "=m" (modep->__mxcsr)); return 0; } diff --git a/sysdeps/i386/fpu/feholdexcpt.c b/sysdeps/i386/fpu/feholdexcpt.c index ad25339..8d2d0ee 100644 --- a/sysdeps/i386/fpu/feholdexcpt.c +++ b/sysdeps/i386/fpu/feholdexcpt.c @@ -30,7 +30,7 @@ __feholdexcept (fenv_t *envp) __asm__ volatile ("fnstenv %0; fnclex" : "=m" (*envp)); /* If the CPU supports SSE we set the MXCSR as well. */ - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) { unsigned int xwork; diff --git a/sysdeps/i386/fpu/fesetenv.c b/sysdeps/i386/fpu/fesetenv.c index 5ec7bd6..cd9afea 100644 --- a/sysdeps/i386/fpu/fesetenv.c +++ b/sysdeps/i386/fpu/fesetenv.c @@ -79,7 +79,7 @@ __fesetenv (const fenv_t *envp) __asm__ ("fldenv %0" : : "m" (temp)); - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) { unsigned int mxcsr; __asm__ ("stmxcsr %0" : "=m" (mxcsr)); diff --git a/sysdeps/i386/fpu/fesetmode.c b/sysdeps/i386/fpu/fesetmode.c index 4563da0..e3b3065 100644 --- a/sysdeps/i386/fpu/fesetmode.c +++ b/sysdeps/i386/fpu/fesetmode.c @@ -35,7 +35,7 @@ fesetmode (const femode_t *modep) else cw = modep->__control_word; _FPU_SETCW (cw); - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) { unsigned int mxcsr; __asm__ ("stmxcsr %0" : "=m" (mxcsr)); diff --git a/sysdeps/i386/fpu/fesetround.c b/sysdeps/i386/fpu/fesetround.c index 18320a6..5c3fd34 100644 --- a/sysdeps/i386/fpu/fesetround.c +++ b/sysdeps/i386/fpu/fesetround.c @@ -37,7 +37,7 @@ __fesetround (int round) __asm__ ("fldcw %0" : : "m" (*&cw)); /* If the CPU supports SSE we set the MXCSR as well. */ - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) { unsigned int xcw; diff --git a/sysdeps/i386/fpu/feupdateenv.c b/sysdeps/i386/fpu/feupdateenv.c index 7387831..ef7132e 100644 --- a/sysdeps/i386/fpu/feupdateenv.c +++ b/sysdeps/i386/fpu/feupdateenv.c @@ -32,7 +32,7 @@ __feupdateenv (const fenv_t *envp) __asm__ ("fnstsw %0" : "=m" (*&temp)); /* If the CPU supports SSE we test the MXCSR as well. */ - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) __asm__ ("stmxcsr %0" : "=m" (*&xtemp)); temp = (temp | xtemp) & FE_ALL_EXCEPT; diff --git a/sysdeps/i386/fpu/fgetexcptflg.c b/sysdeps/i386/fpu/fgetexcptflg.c index 82b2aa5..2c32c83 100644 --- a/sysdeps/i386/fpu/fgetexcptflg.c +++ b/sysdeps/i386/fpu/fgetexcptflg.c @@ -34,7 +34,7 @@ __fegetexceptflag (fexcept_t *flagp, int excepts) *flagp = temp & excepts & FE_ALL_EXCEPT; /* If the CPU supports SSE, we clear the MXCSR as well. */ - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) { unsigned int sse_exc; diff --git a/sysdeps/i386/fpu/fsetexcptflg.c b/sysdeps/i386/fpu/fsetexcptflg.c index dc257b8..02a1bd5 100644 --- a/sysdeps/i386/fpu/fsetexcptflg.c +++ b/sysdeps/i386/fpu/fsetexcptflg.c @@ -41,7 +41,7 @@ __fesetexceptflag (const fexcept_t *flagp, int excepts) __asm__ ("fldenv %0" : : "m" (*&temp)); /* If the CPU supports SSE, we set the MXCSR as well. */ - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) { unsigned int xnew_exc; diff --git a/sysdeps/i386/fpu/ftestexcept.c b/sysdeps/i386/fpu/ftestexcept.c index 9c22689..a00c44e 100644 --- a/sysdeps/i386/fpu/ftestexcept.c +++ b/sysdeps/i386/fpu/ftestexcept.c @@ -32,7 +32,7 @@ fetestexcept (int excepts) __asm__ ("fnstsw %0" : "=a" (temp)); /* If the CPU supports SSE we test the MXCSR as well. */ - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) __asm__ ("stmxcsr %0" : "=m" (*&xtemp)); return (temp | xtemp) & excepts & FE_ALL_EXCEPT; diff --git a/sysdeps/i386/i686/fpu/multiarch/s_cosf.c b/sysdeps/i386/i686/fpu/multiarch/s_cosf.c index 8da7d4b..9cd14a1 100644 --- a/sysdeps/i386/i686/fpu/multiarch/s_cosf.c +++ b/sysdeps/i386/i686/fpu/multiarch/s_cosf.c @@ -23,7 +23,7 @@ extern float __cosf_sse2 (float); extern float __cosf_ia32 (float); -libm_ifunc (__cosf, HAS_CPU_FEATURE (SSE2) ? __cosf_sse2 : __cosf_ia32); +libm_ifunc (__cosf, CPU_FEATURE_USABLE (SSE2) ? __cosf_sse2 : __cosf_ia32); libm_alias_float (__cos, cos); #define COSF __cosf_ia32 diff --git a/sysdeps/i386/i686/fpu/multiarch/s_sincosf.c b/sysdeps/i386/i686/fpu/multiarch/s_sincosf.c index 06d094d..9b47914 100644 --- a/sysdeps/i386/i686/fpu/multiarch/s_sincosf.c +++ b/sysdeps/i386/i686/fpu/multiarch/s_sincosf.c @@ -24,7 +24,7 @@ extern void __sincosf_sse2 (float, float *, float *); extern void __sincosf_ia32 (float, float *, float *); libm_ifunc (__sincosf, - HAS_CPU_FEATURE (SSE2) ? __sincosf_sse2 : __sincosf_ia32); + CPU_FEATURE_USABLE (SSE2) ? __sincosf_sse2 : __sincosf_ia32); libm_alias_float (__sincos, sincos); #define SINCOSF __sincosf_ia32 diff --git a/sysdeps/i386/i686/fpu/multiarch/s_sinf.c b/sysdeps/i386/i686/fpu/multiarch/s_sinf.c index abd355e..84977e6 100644 --- a/sysdeps/i386/i686/fpu/multiarch/s_sinf.c +++ b/sysdeps/i386/i686/fpu/multiarch/s_sinf.c @@ -23,7 +23,7 @@ extern float __sinf_sse2 (float); extern float __sinf_ia32 (float); -libm_ifunc (__sinf, HAS_CPU_FEATURE (SSE2) ? __sinf_sse2 : __sinf_ia32); +libm_ifunc (__sinf, CPU_FEATURE_USABLE (SSE2) ? __sinf_sse2 : __sinf_ia32); libm_alias_float (__sin, sin); #define SINF __sinf_ia32 #include <sysdeps/ieee754/flt-32/s_sinf.c> diff --git a/sysdeps/i386/i686/multiarch/ifunc-impl-list.c b/sysdeps/i386/i686/multiarch/ifunc-impl-list.c index 23774fb..89afdc0 100644 --- a/sysdeps/i386/i686/multiarch/ifunc-impl-list.c +++ b/sysdeps/i386/i686/multiarch/ifunc-impl-list.c @@ -38,35 +38,35 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/i386/i686/multiarch/bcopy.S. */ IFUNC_IMPL (i, name, bcopy, - IFUNC_IMPL_ADD (array, i, bcopy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, bcopy, CPU_FEATURE_USABLE (SSSE3), __bcopy_ssse3_rep) - IFUNC_IMPL_ADD (array, i, bcopy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, bcopy, CPU_FEATURE_USABLE (SSSE3), __bcopy_ssse3) - IFUNC_IMPL_ADD (array, i, bcopy, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, bcopy, CPU_FEATURE_USABLE (SSE2), __bcopy_sse2_unaligned) IFUNC_IMPL_ADD (array, i, bcopy, 1, __bcopy_ia32)) /* Support sysdeps/i386/i686/multiarch/bzero.S. */ IFUNC_IMPL (i, name, bzero, - IFUNC_IMPL_ADD (array, i, bzero, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, bzero, CPU_FEATURE_USABLE (SSE2), __bzero_sse2_rep) - IFUNC_IMPL_ADD (array, i, bzero, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, bzero, CPU_FEATURE_USABLE (SSE2), __bzero_sse2) IFUNC_IMPL_ADD (array, i, bzero, 1, __bzero_ia32)) /* Support sysdeps/i386/i686/multiarch/memchr.S. */ IFUNC_IMPL (i, name, memchr, - IFUNC_IMPL_ADD (array, i, memchr, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, memchr, CPU_FEATURE_USABLE (SSE2), __memchr_sse2_bsf) - IFUNC_IMPL_ADD (array, i, memchr, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, memchr, CPU_FEATURE_USABLE (SSE2), __memchr_sse2) IFUNC_IMPL_ADD (array, i, memchr, 1, __memchr_ia32)) /* Support sysdeps/i386/i686/multiarch/memcmp.S. */ IFUNC_IMPL (i, name, memcmp, - IFUNC_IMPL_ADD (array, i, memcmp, HAS_CPU_FEATURE (SSE4_2), + IFUNC_IMPL_ADD (array, i, memcmp, CPU_FEATURE_USABLE (SSE4_2), __memcmp_sse4_2) - IFUNC_IMPL_ADD (array, i, memcmp, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, memcmp, CPU_FEATURE_USABLE (SSSE3), __memcmp_ssse3) IFUNC_IMPL_ADD (array, i, memcmp, 1, __memcmp_ia32)) @@ -74,13 +74,13 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/i386/i686/multiarch/memmove_chk.S. */ IFUNC_IMPL (i, name, __memmove_chk, IFUNC_IMPL_ADD (array, i, __memmove_chk, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __memmove_chk_ssse3_rep) IFUNC_IMPL_ADD (array, i, __memmove_chk, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __memmove_chk_ssse3) IFUNC_IMPL_ADD (array, i, __memmove_chk, - HAS_CPU_FEATURE (SSE2), + CPU_FEATURE_USABLE (SSE2), __memmove_chk_sse2_unaligned) IFUNC_IMPL_ADD (array, i, __memmove_chk, 1, __memmove_chk_ia32)) @@ -88,19 +88,19 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/i386/i686/multiarch/memmove.S. */ IFUNC_IMPL (i, name, memmove, - IFUNC_IMPL_ADD (array, i, memmove, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, memmove, CPU_FEATURE_USABLE (SSSE3), __memmove_ssse3_rep) - IFUNC_IMPL_ADD (array, i, memmove, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, memmove, CPU_FEATURE_USABLE (SSSE3), __memmove_ssse3) - IFUNC_IMPL_ADD (array, i, memmove, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, memmove, CPU_FEATURE_USABLE (SSE2), __memmove_sse2_unaligned) IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_ia32)) /* Support sysdeps/i386/i686/multiarch/memrchr.S. */ IFUNC_IMPL (i, name, memrchr, - IFUNC_IMPL_ADD (array, i, memrchr, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, memrchr, CPU_FEATURE_USABLE (SSE2), __memrchr_sse2_bsf) - IFUNC_IMPL_ADD (array, i, memrchr, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, memrchr, CPU_FEATURE_USABLE (SSE2), __memrchr_sse2) IFUNC_IMPL_ADD (array, i, memrchr, 1, __memrchr_ia32)) @@ -108,10 +108,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/i386/i686/multiarch/memset_chk.S. */ IFUNC_IMPL (i, name, __memset_chk, IFUNC_IMPL_ADD (array, i, __memset_chk, - HAS_CPU_FEATURE (SSE2), + CPU_FEATURE_USABLE (SSE2), __memset_chk_sse2_rep) IFUNC_IMPL_ADD (array, i, __memset_chk, - HAS_CPU_FEATURE (SSE2), + CPU_FEATURE_USABLE (SSE2), __memset_chk_sse2) IFUNC_IMPL_ADD (array, i, __memset_chk, 1, __memset_chk_ia32)) @@ -119,102 +119,102 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/i386/i686/multiarch/memset.S. */ IFUNC_IMPL (i, name, memset, - IFUNC_IMPL_ADD (array, i, memset, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, memset, CPU_FEATURE_USABLE (SSE2), __memset_sse2_rep) - IFUNC_IMPL_ADD (array, i, memset, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, memset, CPU_FEATURE_USABLE (SSE2), __memset_sse2) IFUNC_IMPL_ADD (array, i, memset, 1, __memset_ia32)) /* Support sysdeps/i386/i686/multiarch/rawmemchr.S. */ IFUNC_IMPL (i, name, rawmemchr, - IFUNC_IMPL_ADD (array, i, rawmemchr, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, rawmemchr, CPU_FEATURE_USABLE (SSE2), __rawmemchr_sse2_bsf) - IFUNC_IMPL_ADD (array, i, rawmemchr, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, rawmemchr, CPU_FEATURE_USABLE (SSE2), __rawmemchr_sse2) IFUNC_IMPL_ADD (array, i, rawmemchr, 1, __rawmemchr_ia32)) /* Support sysdeps/i386/i686/multiarch/stpncpy.S. */ IFUNC_IMPL (i, name, stpncpy, - IFUNC_IMPL_ADD (array, i, stpncpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, stpncpy, CPU_FEATURE_USABLE (SSSE3), __stpncpy_ssse3) - IFUNC_IMPL_ADD (array, i, stpncpy, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, stpncpy, CPU_FEATURE_USABLE (SSE2), __stpncpy_sse2) IFUNC_IMPL_ADD (array, i, stpncpy, 1, __stpncpy_ia32)) /* Support sysdeps/i386/i686/multiarch/stpcpy.S. */ IFUNC_IMPL (i, name, stpcpy, - IFUNC_IMPL_ADD (array, i, stpcpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, stpcpy, CPU_FEATURE_USABLE (SSSE3), __stpcpy_ssse3) - IFUNC_IMPL_ADD (array, i, stpcpy, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, stpcpy, CPU_FEATURE_USABLE (SSE2), __stpcpy_sse2) IFUNC_IMPL_ADD (array, i, stpcpy, 1, __stpcpy_ia32)) /* Support sysdeps/i386/i686/multiarch/strcasecmp.S. */ IFUNC_IMPL (i, name, strcasecmp, IFUNC_IMPL_ADD (array, i, strcasecmp, - HAS_CPU_FEATURE (SSE4_2), + CPU_FEATURE_USABLE (SSE4_2), __strcasecmp_sse4_2) IFUNC_IMPL_ADD (array, i, strcasecmp, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __strcasecmp_ssse3) IFUNC_IMPL_ADD (array, i, strcasecmp, 1, __strcasecmp_ia32)) /* Support sysdeps/i386/i686/multiarch/strcasecmp_l.S. */ IFUNC_IMPL (i, name, strcasecmp_l, IFUNC_IMPL_ADD (array, i, strcasecmp_l, - HAS_CPU_FEATURE (SSE4_2), + CPU_FEATURE_USABLE (SSE4_2), __strcasecmp_l_sse4_2) IFUNC_IMPL_ADD (array, i, strcasecmp_l, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __strcasecmp_l_ssse3) IFUNC_IMPL_ADD (array, i, strcasecmp_l, 1, __strcasecmp_l_ia32)) /* Support sysdeps/i386/i686/multiarch/strcat.S. */ IFUNC_IMPL (i, name, strcat, - IFUNC_IMPL_ADD (array, i, strcat, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, strcat, CPU_FEATURE_USABLE (SSSE3), __strcat_ssse3) - IFUNC_IMPL_ADD (array, i, strcat, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, strcat, CPU_FEATURE_USABLE (SSE2), __strcat_sse2) IFUNC_IMPL_ADD (array, i, strcat, 1, __strcat_ia32)) /* Support sysdeps/i386/i686/multiarch/strchr.S. */ IFUNC_IMPL (i, name, strchr, - IFUNC_IMPL_ADD (array, i, strchr, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, strchr, CPU_FEATURE_USABLE (SSE2), __strchr_sse2_bsf) - IFUNC_IMPL_ADD (array, i, strchr, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, strchr, CPU_FEATURE_USABLE (SSE2), __strchr_sse2) IFUNC_IMPL_ADD (array, i, strchr, 1, __strchr_ia32)) /* Support sysdeps/i386/i686/multiarch/strcmp.S. */ IFUNC_IMPL (i, name, strcmp, - IFUNC_IMPL_ADD (array, i, strcmp, HAS_CPU_FEATURE (SSE4_2), + IFUNC_IMPL_ADD (array, i, strcmp, CPU_FEATURE_USABLE (SSE4_2), __strcmp_sse4_2) - IFUNC_IMPL_ADD (array, i, strcmp, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, strcmp, CPU_FEATURE_USABLE (SSSE3), __strcmp_ssse3) IFUNC_IMPL_ADD (array, i, strcmp, 1, __strcmp_ia32)) /* Support sysdeps/i386/i686/multiarch/strcpy.S. */ IFUNC_IMPL (i, name, strcpy, - IFUNC_IMPL_ADD (array, i, strcpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, strcpy, CPU_FEATURE_USABLE (SSSE3), __strcpy_ssse3) - IFUNC_IMPL_ADD (array, i, strcpy, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, strcpy, CPU_FEATURE_USABLE (SSE2), __strcpy_sse2) IFUNC_IMPL_ADD (array, i, strcpy, 1, __strcpy_ia32)) /* Support sysdeps/i386/i686/multiarch/strcspn.S. */ IFUNC_IMPL (i, name, strcspn, - IFUNC_IMPL_ADD (array, i, strcspn, HAS_CPU_FEATURE (SSE4_2), + IFUNC_IMPL_ADD (array, i, strcspn, CPU_FEATURE_USABLE (SSE4_2), __strcspn_sse42) IFUNC_IMPL_ADD (array, i, strcspn, 1, __strcspn_ia32)) /* Support sysdeps/i386/i686/multiarch/strncase.S. */ IFUNC_IMPL (i, name, strncasecmp, IFUNC_IMPL_ADD (array, i, strncasecmp, - HAS_CPU_FEATURE (SSE4_2), + CPU_FEATURE_USABLE (SSE4_2), __strncasecmp_sse4_2) IFUNC_IMPL_ADD (array, i, strncasecmp, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __strncasecmp_ssse3) IFUNC_IMPL_ADD (array, i, strncasecmp, 1, __strncasecmp_ia32)) @@ -222,91 +222,91 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/i386/i686/multiarch/strncase_l.S. */ IFUNC_IMPL (i, name, strncasecmp_l, IFUNC_IMPL_ADD (array, i, strncasecmp_l, - HAS_CPU_FEATURE (SSE4_2), + CPU_FEATURE_USABLE (SSE4_2), __strncasecmp_l_sse4_2) IFUNC_IMPL_ADD (array, i, strncasecmp_l, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __strncasecmp_l_ssse3) IFUNC_IMPL_ADD (array, i, strncasecmp_l, 1, __strncasecmp_l_ia32)) /* Support sysdeps/i386/i686/multiarch/strncat.S. */ IFUNC_IMPL (i, name, strncat, - IFUNC_IMPL_ADD (array, i, strncat, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, strncat, CPU_FEATURE_USABLE (SSSE3), __strncat_ssse3) - IFUNC_IMPL_ADD (array, i, strncat, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, strncat, CPU_FEATURE_USABLE (SSE2), __strncat_sse2) IFUNC_IMPL_ADD (array, i, strncat, 1, __strncat_ia32)) /* Support sysdeps/i386/i686/multiarch/strncpy.S. */ IFUNC_IMPL (i, name, strncpy, - IFUNC_IMPL_ADD (array, i, strncpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, strncpy, CPU_FEATURE_USABLE (SSSE3), __strncpy_ssse3) - IFUNC_IMPL_ADD (array, i, strncpy, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, strncpy, CPU_FEATURE_USABLE (SSE2), __strncpy_sse2) IFUNC_IMPL_ADD (array, i, strncpy, 1, __strncpy_ia32)) /* Support sysdeps/i386/i686/multiarch/strnlen.S. */ IFUNC_IMPL (i, name, strnlen, - IFUNC_IMPL_ADD (array, i, strnlen, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, strnlen, CPU_FEATURE_USABLE (SSE2), __strnlen_sse2) IFUNC_IMPL_ADD (array, i, strnlen, 1, __strnlen_ia32)) /* Support sysdeps/i386/i686/multiarch/strpbrk.S. */ IFUNC_IMPL (i, name, strpbrk, - IFUNC_IMPL_ADD (array, i, strpbrk, HAS_CPU_FEATURE (SSE4_2), + IFUNC_IMPL_ADD (array, i, strpbrk, CPU_FEATURE_USABLE (SSE4_2), __strpbrk_sse42) IFUNC_IMPL_ADD (array, i, strpbrk, 1, __strpbrk_ia32)) /* Support sysdeps/i386/i686/multiarch/strrchr.S. */ IFUNC_IMPL (i, name, strrchr, - IFUNC_IMPL_ADD (array, i, strrchr, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, strrchr, CPU_FEATURE_USABLE (SSE2), __strrchr_sse2_bsf) - IFUNC_IMPL_ADD (array, i, strrchr, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, strrchr, CPU_FEATURE_USABLE (SSE2), __strrchr_sse2) IFUNC_IMPL_ADD (array, i, strrchr, 1, __strrchr_ia32)) /* Support sysdeps/i386/i686/multiarch/strspn.S. */ IFUNC_IMPL (i, name, strspn, - IFUNC_IMPL_ADD (array, i, strspn, HAS_CPU_FEATURE (SSE4_2), + IFUNC_IMPL_ADD (array, i, strspn, CPU_FEATURE_USABLE (SSE4_2), __strspn_sse42) IFUNC_IMPL_ADD (array, i, strspn, 1, __strspn_ia32)) /* Support sysdeps/i386/i686/multiarch/wcschr.S. */ IFUNC_IMPL (i, name, wcschr, - IFUNC_IMPL_ADD (array, i, wcschr, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, wcschr, CPU_FEATURE_USABLE (SSE2), __wcschr_sse2) IFUNC_IMPL_ADD (array, i, wcschr, 1, __wcschr_ia32)) /* Support sysdeps/i386/i686/multiarch/wcscmp.S. */ IFUNC_IMPL (i, name, wcscmp, - IFUNC_IMPL_ADD (array, i, wcscmp, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, wcscmp, CPU_FEATURE_USABLE (SSE2), __wcscmp_sse2) IFUNC_IMPL_ADD (array, i, wcscmp, 1, __wcscmp_ia32)) /* Support sysdeps/i386/i686/multiarch/wcscpy.S. */ IFUNC_IMPL (i, name, wcscpy, - IFUNC_IMPL_ADD (array, i, wcscpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, wcscpy, CPU_FEATURE_USABLE (SSSE3), __wcscpy_ssse3) IFUNC_IMPL_ADD (array, i, wcscpy, 1, __wcscpy_ia32)) /* Support sysdeps/i386/i686/multiarch/wcslen.S. */ IFUNC_IMPL (i, name, wcslen, - IFUNC_IMPL_ADD (array, i, wcslen, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, wcslen, CPU_FEATURE_USABLE (SSE2), __wcslen_sse2) IFUNC_IMPL_ADD (array, i, wcslen, 1, __wcslen_ia32)) /* Support sysdeps/i386/i686/multiarch/wcsrchr.S. */ IFUNC_IMPL (i, name, wcsrchr, - IFUNC_IMPL_ADD (array, i, wcsrchr, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, wcsrchr, CPU_FEATURE_USABLE (SSE2), __wcsrchr_sse2) IFUNC_IMPL_ADD (array, i, wcsrchr, 1, __wcsrchr_ia32)) /* Support sysdeps/i386/i686/multiarch/wmemcmp.S. */ IFUNC_IMPL (i, name, wmemcmp, - IFUNC_IMPL_ADD (array, i, wmemcmp, HAS_CPU_FEATURE (SSE4_2), + IFUNC_IMPL_ADD (array, i, wmemcmp, CPU_FEATURE_USABLE (SSE4_2), __wmemcmp_sse4_2) - IFUNC_IMPL_ADD (array, i, wmemcmp, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, wmemcmp, CPU_FEATURE_USABLE (SSSE3), __wmemcmp_ssse3) IFUNC_IMPL_ADD (array, i, wmemcmp, 1, __wmemcmp_ia32)) @@ -314,64 +314,64 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/i386/i686/multiarch/memcpy_chk.S. */ IFUNC_IMPL (i, name, __memcpy_chk, IFUNC_IMPL_ADD (array, i, __memcpy_chk, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __memcpy_chk_ssse3_rep) IFUNC_IMPL_ADD (array, i, __memcpy_chk, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __memcpy_chk_ssse3) IFUNC_IMPL_ADD (array, i, __memcpy_chk, - HAS_CPU_FEATURE (SSE2), + CPU_FEATURE_USABLE (SSE2), __memcpy_chk_sse2_unaligned) IFUNC_IMPL_ADD (array, i, __memcpy_chk, 1, __memcpy_chk_ia32)) /* Support sysdeps/i386/i686/multiarch/memcpy.S. */ IFUNC_IMPL (i, name, memcpy, - IFUNC_IMPL_ADD (array, i, memcpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, memcpy, CPU_FEATURE_USABLE (SSSE3), __memcpy_ssse3_rep) - IFUNC_IMPL_ADD (array, i, memcpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, memcpy, CPU_FEATURE_USABLE (SSSE3), __memcpy_ssse3) - IFUNC_IMPL_ADD (array, i, memcpy, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, memcpy, CPU_FEATURE_USABLE (SSE2), __memcpy_sse2_unaligned) IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_ia32)) /* Support sysdeps/i386/i686/multiarch/mempcpy_chk.S. */ IFUNC_IMPL (i, name, __mempcpy_chk, IFUNC_IMPL_ADD (array, i, __mempcpy_chk, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __mempcpy_chk_ssse3_rep) IFUNC_IMPL_ADD (array, i, __mempcpy_chk, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __mempcpy_chk_ssse3) IFUNC_IMPL_ADD (array, i, __mempcpy_chk, - HAS_CPU_FEATURE (SSE2), + CPU_FEATURE_USABLE (SSE2), __mempcpy_chk_sse2_unaligned) IFUNC_IMPL_ADD (array, i, __mempcpy_chk, 1, __mempcpy_chk_ia32)) /* Support sysdeps/i386/i686/multiarch/mempcpy.S. */ IFUNC_IMPL (i, name, mempcpy, - IFUNC_IMPL_ADD (array, i, mempcpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, mempcpy, CPU_FEATURE_USABLE (SSSE3), __mempcpy_ssse3_rep) - IFUNC_IMPL_ADD (array, i, mempcpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, mempcpy, CPU_FEATURE_USABLE (SSSE3), __mempcpy_ssse3) - IFUNC_IMPL_ADD (array, i, mempcpy, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, mempcpy, CPU_FEATURE_USABLE (SSE2), __mempcpy_sse2_unaligned) IFUNC_IMPL_ADD (array, i, mempcpy, 1, __mempcpy_ia32)) /* Support sysdeps/i386/i686/multiarch/strlen.S. */ IFUNC_IMPL (i, name, strlen, - IFUNC_IMPL_ADD (array, i, strlen, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, strlen, CPU_FEATURE_USABLE (SSE2), __strlen_sse2_bsf) - IFUNC_IMPL_ADD (array, i, strlen, HAS_CPU_FEATURE (SSE2), + IFUNC_IMPL_ADD (array, i, strlen, CPU_FEATURE_USABLE (SSE2), __strlen_sse2) IFUNC_IMPL_ADD (array, i, strlen, 1, __strlen_ia32)) /* Support sysdeps/i386/i686/multiarch/strncmp.S. */ IFUNC_IMPL (i, name, strncmp, - IFUNC_IMPL_ADD (array, i, strncmp, HAS_CPU_FEATURE (SSE4_2), + IFUNC_IMPL_ADD (array, i, strncmp, CPU_FEATURE_USABLE (SSE4_2), __strncmp_sse4_2) - IFUNC_IMPL_ADD (array, i, strncmp, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, strncmp, CPU_FEATURE_USABLE (SSSE3), __strncmp_ssse3) IFUNC_IMPL_ADD (array, i, strncmp, 1, __strncmp_ia32)) #endif diff --git a/sysdeps/i386/i686/multiarch/ifunc-memmove.h b/sysdeps/i386/i686/multiarch/ifunc-memmove.h index a590048..c05cb6d 100644 --- a/sysdeps/i386/i686/multiarch/ifunc-memmove.h +++ b/sysdeps/i386/i686/multiarch/ifunc-memmove.h @@ -33,7 +33,7 @@ IFUNC_SELECTOR (void) if (CPU_FEATURES_ARCH_P (cpu_features, Fast_Unaligned_Load)) return OPTIMIZE (sse2_unaligned); - if (CPU_FEATURES_CPU_P (cpu_features, SSSE3)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSSE3)) { if (CPU_FEATURES_ARCH_P (cpu_features, Fast_Rep_String)) return OPTIMIZE (ssse3_rep); diff --git a/sysdeps/i386/i686/multiarch/ifunc-memset.h b/sysdeps/i386/i686/multiarch/ifunc-memset.h index 14199c3..bead331 100644 --- a/sysdeps/i386/i686/multiarch/ifunc-memset.h +++ b/sysdeps/i386/i686/multiarch/ifunc-memset.h @@ -28,7 +28,7 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_CPU_P (cpu_features, SSE2)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE2)) { if (CPU_FEATURES_ARCH_P (cpu_features, Fast_Rep_String)) return OPTIMIZE (sse2_rep); diff --git a/sysdeps/i386/i686/multiarch/ifunc-sse2-bsf.h b/sysdeps/i386/i686/multiarch/ifunc-sse2-bsf.h index 8b6fa64..0d302a3 100644 --- a/sysdeps/i386/i686/multiarch/ifunc-sse2-bsf.h +++ b/sysdeps/i386/i686/multiarch/ifunc-sse2-bsf.h @@ -28,7 +28,7 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_CPU_P (cpu_features, SSE2)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE2)) { if (CPU_FEATURES_ARCH_P (cpu_features, Slow_BSF)) return OPTIMIZE (sse2); diff --git a/sysdeps/i386/i686/multiarch/ifunc-sse2-ssse3.h b/sysdeps/i386/i686/multiarch/ifunc-sse2-ssse3.h index 77b615e..c10ca4a 100644 --- a/sysdeps/i386/i686/multiarch/ifunc-sse2-ssse3.h +++ b/sysdeps/i386/i686/multiarch/ifunc-sse2-ssse3.h @@ -29,11 +29,11 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_CPU_P (cpu_features, SSE2) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE2) && CPU_FEATURES_ARCH_P (cpu_features, Fast_Rep_String)) return OPTIMIZE (sse2); - if (CPU_FEATURES_CPU_P (cpu_features, SSSE3)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSSE3)) return OPTIMIZE (ssse3); return OPTIMIZE (ia32); diff --git a/sysdeps/i386/i686/multiarch/ifunc-sse2.h b/sysdeps/i386/i686/multiarch/ifunc-sse2.h index c0dd85e..58794a2 100644 --- a/sysdeps/i386/i686/multiarch/ifunc-sse2.h +++ b/sysdeps/i386/i686/multiarch/ifunc-sse2.h @@ -27,7 +27,7 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_CPU_P (cpu_features, SSE2)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE2)) return OPTIMIZE (sse2); return OPTIMIZE (ia32); diff --git a/sysdeps/i386/i686/multiarch/ifunc-sse4_2.h b/sysdeps/i386/i686/multiarch/ifunc-sse4_2.h index b9b06d5..014be1d 100644 --- a/sysdeps/i386/i686/multiarch/ifunc-sse4_2.h +++ b/sysdeps/i386/i686/multiarch/ifunc-sse4_2.h @@ -27,7 +27,7 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_CPU_P (cpu_features, SSE4_2)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE4_2)) return OPTIMIZE (sse42); return OPTIMIZE (ia32); diff --git a/sysdeps/i386/i686/multiarch/ifunc-ssse3-sse4_2.h b/sysdeps/i386/i686/multiarch/ifunc-ssse3-sse4_2.h index b4074f3..39bfea9 100644 --- a/sysdeps/i386/i686/multiarch/ifunc-ssse3-sse4_2.h +++ b/sysdeps/i386/i686/multiarch/ifunc-ssse3-sse4_2.h @@ -29,10 +29,10 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_CPU_P (cpu_features, SSE4_2)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE4_2)) return OPTIMIZE (sse4_2); - if (CPU_FEATURES_CPU_P (cpu_features, SSSE3)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSSE3)) return OPTIMIZE (ssse3); return OPTIMIZE (ia32); diff --git a/sysdeps/i386/i686/multiarch/s_fma.c b/sysdeps/i386/i686/multiarch/s_fma.c index 90f649f..0729853 100644 --- a/sysdeps/i386/i686/multiarch/s_fma.c +++ b/sysdeps/i386/i686/multiarch/s_fma.c @@ -27,7 +27,7 @@ extern double __fma_ia32 (double x, double y, double z) attribute_hidden; extern double __fma_fma (double x, double y, double z) attribute_hidden; libm_ifunc (__fma, - HAS_ARCH_FEATURE (FMA_Usable) ? __fma_fma : __fma_ia32); + CPU_FEATURE_USABLE (FMA) ? __fma_fma : __fma_ia32); libm_alias_double (__fma, fma) #define __fma __fma_ia32 diff --git a/sysdeps/i386/i686/multiarch/s_fmaf.c b/sysdeps/i386/i686/multiarch/s_fmaf.c index 27757ec..20f965c 100644 --- a/sysdeps/i386/i686/multiarch/s_fmaf.c +++ b/sysdeps/i386/i686/multiarch/s_fmaf.c @@ -27,7 +27,7 @@ extern float __fmaf_ia32 (float x, float y, float z) attribute_hidden; extern float __fmaf_fma (float x, float y, float z) attribute_hidden; libm_ifunc (__fmaf, - HAS_ARCH_FEATURE (FMA_Usable) ? __fmaf_fma : __fmaf_ia32); + CPU_FEATURE_USABLE (FMA) ? __fmaf_fma : __fmaf_ia32); libm_alias_float (__fma, fma) #define __fmaf __fmaf_ia32 diff --git a/sysdeps/i386/i686/multiarch/wcscpy.c b/sysdeps/i386/i686/multiarch/wcscpy.c index 51347d7..f0038bc 100644 --- a/sysdeps/i386/i686/multiarch/wcscpy.c +++ b/sysdeps/i386/i686/multiarch/wcscpy.c @@ -34,7 +34,7 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_CPU_P (cpu_features, SSSE3)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSSE3)) return OPTIMIZE (ssse3); return OPTIMIZE (ia32); diff --git a/sysdeps/i386/setfpucw.c b/sysdeps/i386/setfpucw.c index 68f5b2e..c640a72 100644 --- a/sysdeps/i386/setfpucw.c +++ b/sysdeps/i386/setfpucw.c @@ -39,7 +39,7 @@ __setfpucw (fpu_control_t set) __asm__ ("fldcw %0" : : "m" (*&cw)); /* If the CPU supports SSE, we set the MXCSR as well. */ - if (HAS_CPU_FEATURE (SSE)) + if (CPU_FEATURE_USABLE (SSE)) { unsigned int xnew_exc; diff --git a/sysdeps/unix/sysv/linux/x86/elision-conf.c b/sysdeps/unix/sysv/linux/x86/elision-conf.c index b38b425..ecdb037 100644 --- a/sysdeps/unix/sysv/linux/x86/elision-conf.c +++ b/sysdeps/unix/sysv/linux/x86/elision-conf.c @@ -63,7 +63,7 @@ do_set_elision_enable (int32_t elision_enable) if __libc_enable_secure isn't enabled since elision_enable will be set according to the default, which is disabled. */ if (elision_enable == 1) - __pthread_force_elision = HAS_CPU_FEATURE (RTM) ? 1 : 0; + __pthread_force_elision = CPU_FEATURE_USABLE (RTM) ? 1 : 0; } /* The pthread->elision_enable tunable is 0 or 1 indicating that elision diff --git a/sysdeps/x86/cacheinfo.c b/sysdeps/x86/cacheinfo.c index 5366a37..217c21c 100644 --- a/sysdeps/x86/cacheinfo.c +++ b/sysdeps/x86/cacheinfo.c @@ -583,7 +583,7 @@ get_common_cache_info (long int *shared_ptr, unsigned int *threads_ptr, /* A value of 0 for the HTT bit indicates there is only a single logical processor. */ - if (HAS_CPU_FEATURE (HTT)) + if (CPU_FEATURE_USABLE (HTT)) { /* Figure out the number of logical threads that share the highest cache level. */ @@ -732,7 +732,7 @@ intel_bug_no_cache_info: /* Assume that all logical threads share the highest cache level. */ threads - = ((cpu_features->cpuid[COMMON_CPUID_INDEX_1].ebx + = ((cpu_features->features[COMMON_CPUID_INDEX_1].cpuid.ebx >> 16) & 0xff); } @@ -867,14 +867,14 @@ init_cacheinfo (void) unsigned int minimum_rep_movsb_threshold; /* NB: The default REP MOVSB threshold is 2048 * (VEC_SIZE / 16). */ unsigned int rep_movsb_threshold; - if (CPU_FEATURES_ARCH_P (cpu_features, AVX512F_Usable) - && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512)) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX512F) + && !CPU_FEATURE_PREFERRED_P (cpu_features, Prefer_No_AVX512)) { rep_movsb_threshold = 2048 * (64 / 16); minimum_rep_movsb_threshold = 64 * 8; } - else if (CPU_FEATURES_ARCH_P (cpu_features, - AVX_Fast_Unaligned_Load)) + else if (CPU_FEATURE_PREFERRED_P (cpu_features, + AVX_Fast_Unaligned_Load)) { rep_movsb_threshold = 2048 * (32 / 16); minimum_rep_movsb_threshold = 32 * 8; diff --git a/sysdeps/x86/cpu-features.c b/sysdeps/x86/cpu-features.c index c7673a2..4c24ba7 100644 --- a/sysdeps/x86/cpu-features.c +++ b/sysdeps/x86/cpu-features.c @@ -42,73 +42,109 @@ extern void TUNABLE_CALLBACK (set_x86_shstk) (tunable_val_t *) #endif static void -get_extended_indices (struct cpu_features *cpu_features) +update_usable (struct cpu_features *cpu_features) { - unsigned int eax, ebx, ecx, edx; - __cpuid (0x80000000, eax, ebx, ecx, edx); - if (eax >= 0x80000001) - __cpuid (0x80000001, - cpu_features->cpuid[COMMON_CPUID_INDEX_80000001].eax, - cpu_features->cpuid[COMMON_CPUID_INDEX_80000001].ebx, - cpu_features->cpuid[COMMON_CPUID_INDEX_80000001].ecx, - cpu_features->cpuid[COMMON_CPUID_INDEX_80000001].edx); - if (eax >= 0x80000007) - __cpuid (0x80000007, - cpu_features->cpuid[COMMON_CPUID_INDEX_80000007].eax, - cpu_features->cpuid[COMMON_CPUID_INDEX_80000007].ebx, - cpu_features->cpuid[COMMON_CPUID_INDEX_80000007].ecx, - cpu_features->cpuid[COMMON_CPUID_INDEX_80000007].edx); - if (eax >= 0x80000008) - __cpuid (0x80000008, - cpu_features->cpuid[COMMON_CPUID_INDEX_80000008].eax, - cpu_features->cpuid[COMMON_CPUID_INDEX_80000008].ebx, - cpu_features->cpuid[COMMON_CPUID_INDEX_80000008].ecx, - cpu_features->cpuid[COMMON_CPUID_INDEX_80000008].edx); -} - -static void -get_common_indices (struct cpu_features *cpu_features, - unsigned int *family, unsigned int *model, - unsigned int *extended_model, unsigned int *stepping) -{ - if (family) - { - unsigned int eax; - __cpuid (1, eax, cpu_features->cpuid[COMMON_CPUID_INDEX_1].ebx, - cpu_features->cpuid[COMMON_CPUID_INDEX_1].ecx, - cpu_features->cpuid[COMMON_CPUID_INDEX_1].edx); - cpu_features->cpuid[COMMON_CPUID_INDEX_1].eax = eax; - *family = (eax >> 8) & 0x0f; - *model = (eax >> 4) & 0x0f; - *extended_model = (eax >> 12) & 0xf0; - *stepping = eax & 0x0f; - if (*family == 0x0f) - { - *family += (eax >> 20) & 0xff; - *model += *extended_model; - } - } - - if (cpu_features->basic.max_cpuid >= 7) - { - __cpuid_count (7, 0, - cpu_features->cpuid[COMMON_CPUID_INDEX_7].eax, - cpu_features->cpuid[COMMON_CPUID_INDEX_7].ebx, - cpu_features->cpuid[COMMON_CPUID_INDEX_7].ecx, - cpu_features->cpuid[COMMON_CPUID_INDEX_7].edx); - __cpuid_count (7, 1, - cpu_features->cpuid[COMMON_CPUID_INDEX_7_ECX_1].eax, - cpu_features->cpuid[COMMON_CPUID_INDEX_7_ECX_1].ebx, - cpu_features->cpuid[COMMON_CPUID_INDEX_7_ECX_1].ecx, - cpu_features->cpuid[COMMON_CPUID_INDEX_7_ECX_1].edx); - } - - if (cpu_features->basic.max_cpuid >= 0xd) - __cpuid_count (0xd, 1, - cpu_features->cpuid[COMMON_CPUID_INDEX_D_ECX_1].eax, - cpu_features->cpuid[COMMON_CPUID_INDEX_D_ECX_1].ebx, - cpu_features->cpuid[COMMON_CPUID_INDEX_D_ECX_1].ecx, - cpu_features->cpuid[COMMON_CPUID_INDEX_D_ECX_1].edx); + /* Before COMMON_CPUID_INDEX_80000001, copy the cpuid array elements to + the usable array. */ + unsigned int i; + for (i = 0; i < COMMON_CPUID_INDEX_80000001; i++) + cpu_features->features[i].usable = cpu_features->features[i].cpuid; + + /* Before COMMON_CPUID_INDEX_80000001, clear the unknown usable bits + and the always zero bits. */ + CPU_FEATURE_UNSET (cpu_features, INDEX_1_ECX_16); + CPU_FEATURE_UNSET (cpu_features, INDEX_1_ECX_31); + CPU_FEATURE_UNSET (cpu_features, INDEX_1_EDX_10); + CPU_FEATURE_UNSET (cpu_features, INDEX_1_EDX_20); + CPU_FEATURE_UNSET (cpu_features, INDEX_1_EDX_30); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EBX_6); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EBX_22); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_ECX_13); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_ECX_15); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_ECX_16); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_ECX_23); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_ECX_24); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_ECX_26); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_0); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_1); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_5); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_6); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_7); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_9); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_11); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_12); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_13); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_17); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_19); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_21); + CPU_FEATURE_UNSET (cpu_features, INDEX_7_EDX_23); + + /* EAX/EBX from COMMON_CPUID_INDEX_1 and EAX from COMMON_CPUID_INDEX_7 + aren't used for CPU feature detection. */ + cpu_features->features[COMMON_CPUID_INDEX_1].usable.eax = 0; + cpu_features->features[COMMON_CPUID_INDEX_1].usable.ebx = 0; + cpu_features->features[COMMON_CPUID_INDEX_7].usable.eax = 0; + + /* Starting from COMMON_CPUID_INDEX_80000001, copy the cpuid bits to + usable bits. */ + CPU_FEATURE_SET_USABLE (cpu_features, LAHF64_SAHF64); + CPU_FEATURE_SET_USABLE (cpu_features, SVM); + CPU_FEATURE_SET_USABLE (cpu_features, LZCNT); + CPU_FEATURE_SET_USABLE (cpu_features, SSE4A); + CPU_FEATURE_SET_USABLE (cpu_features, PREFETCHW); + CPU_FEATURE_SET_USABLE (cpu_features, XOP); + CPU_FEATURE_SET_USABLE (cpu_features, LWP); + CPU_FEATURE_SET_USABLE (cpu_features, FMA4); + CPU_FEATURE_SET_USABLE (cpu_features, TBM); + CPU_FEATURE_SET_USABLE (cpu_features, SYSCALL_SYSRET); + CPU_FEATURE_SET_USABLE (cpu_features, NX); + CPU_FEATURE_SET_USABLE (cpu_features, PAGE1GB); + CPU_FEATURE_SET_USABLE (cpu_features, RDTSCP); + CPU_FEATURE_SET_USABLE (cpu_features, LM); + CPU_FEATURE_SET_USABLE (cpu_features, XSAVEOPT); + CPU_FEATURE_SET_USABLE (cpu_features, XSAVEC); + CPU_FEATURE_SET_USABLE (cpu_features, XGETBV_ECX_1); + CPU_FEATURE_SET_USABLE (cpu_features, XSAVES); + CPU_FEATURE_SET_USABLE (cpu_features, XFD); + CPU_FEATURE_SET_USABLE (cpu_features, INVARIANT_TSC); + CPU_FEATURE_SET_USABLE (cpu_features, WBNOINVD); + CPU_FEATURE_SET_USABLE (cpu_features, AVX512_BF16); + + /* MPX has been deprecated. */ + CPU_FEATURE_UNSET (cpu_features, MPX); + + /* Clear the usable bits which require OS support. */ + CPU_FEATURE_UNSET (cpu_features, FMA); + CPU_FEATURE_UNSET (cpu_features, AVX); + CPU_FEATURE_UNSET (cpu_features, F16C); + CPU_FEATURE_UNSET (cpu_features, AVX2); + CPU_FEATURE_UNSET (cpu_features, AVX512F); + CPU_FEATURE_UNSET (cpu_features, AVX512DQ); + CPU_FEATURE_UNSET (cpu_features, AVX512_IFMA); + CPU_FEATURE_UNSET (cpu_features, AVX512PF); + CPU_FEATURE_UNSET (cpu_features, AVX512ER); + CPU_FEATURE_UNSET (cpu_features, AVX512CD); + CPU_FEATURE_UNSET (cpu_features, AVX512BW); + CPU_FEATURE_UNSET (cpu_features, AVX512VL); + CPU_FEATURE_UNSET (cpu_features, AVX512_VBMI); + CPU_FEATURE_UNSET (cpu_features, PKU); + CPU_FEATURE_UNSET (cpu_features, AVX512_VBMI2); + CPU_FEATURE_UNSET (cpu_features, VAES); + CPU_FEATURE_UNSET (cpu_features, VPCLMULQDQ); + CPU_FEATURE_UNSET (cpu_features, AVX512_VNNI); + CPU_FEATURE_UNSET (cpu_features, AVX512_BITALG); + CPU_FEATURE_UNSET (cpu_features, AVX512_VPOPCNTDQ); + CPU_FEATURE_UNSET (cpu_features, AVX512_4VNNIW); + CPU_FEATURE_UNSET (cpu_features, AVX512_4FMAPS); + CPU_FEATURE_UNSET (cpu_features, AVX512_VP2INTERSECT); + CPU_FEATURE_UNSET (cpu_features, AMX_BF16); + CPU_FEATURE_UNSET (cpu_features, AMX_TILE); + CPU_FEATURE_UNSET (cpu_features, AMX_INT8); + CPU_FEATURE_UNSET (cpu_features, XOP); + CPU_FEATURE_UNSET (cpu_features, FMA4); + CPU_FEATURE_UNSET (cpu_features, XSAVEC); + CPU_FEATURE_UNSET (cpu_features, XFD); + CPU_FEATURE_UNSET (cpu_features, AVX512_BF16); /* Can we call xgetbv? */ if (CPU_FEATURES_CPU_P (cpu_features, OSXSAVE)) @@ -123,40 +159,28 @@ get_common_indices (struct cpu_features *cpu_features, /* Determine if AVX is usable. */ if (CPU_FEATURES_CPU_P (cpu_features, AVX)) { - cpu_features->usable[index_arch_AVX_Usable] - |= bit_arch_AVX_Usable; + CPU_FEATURE_SET (cpu_features, AVX); /* The following features depend on AVX being usable. */ /* Determine if AVX2 is usable. */ if (CPU_FEATURES_CPU_P (cpu_features, AVX2)) - { - cpu_features->usable[index_arch_AVX2_Usable] - |= bit_arch_AVX2_Usable; - - /* Unaligned load with 256-bit AVX registers are faster on - Intel/AMD processors with AVX2. */ - cpu_features->preferred[index_arch_AVX_Fast_Unaligned_Load] - |= bit_arch_AVX_Fast_Unaligned_Load; - } + { + CPU_FEATURE_SET (cpu_features, AVX2); + + /* Unaligned load with 256-bit AVX registers are faster + on Intel/AMD processors with AVX2. */ + cpu_features->preferred[index_arch_AVX_Fast_Unaligned_Load] + |= bit_arch_AVX_Fast_Unaligned_Load; + } /* Determine if FMA is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, FMA)) - cpu_features->usable[index_arch_FMA_Usable] - |= bit_arch_FMA_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, FMA); /* Determine if VAES is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, VAES)) - cpu_features->usable[index_arch_VAES_Usable] - |= bit_arch_VAES_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, VAES); /* Determine if VPCLMULQDQ is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, VPCLMULQDQ)) - cpu_features->usable[index_arch_VPCLMULQDQ_Usable] - |= bit_arch_VPCLMULQDQ_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, VPCLMULQDQ); /* Determine if XOP is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, XOP)) - cpu_features->usable[index_arch_XOP_Usable] - |= bit_arch_XOP_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, XOP); /* Determine if F16C is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, F16C)) - cpu_features->usable[index_arch_F16C_Usable] - |= bit_arch_F16C_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, F16C); } /* Check if OPMASK state, upper 256-bit of ZMM0-ZMM15 and @@ -168,73 +192,41 @@ get_common_indices (struct cpu_features *cpu_features, /* Determine if AVX512F is usable. */ if (CPU_FEATURES_CPU_P (cpu_features, AVX512F)) { - cpu_features->usable[index_arch_AVX512F_Usable] - |= bit_arch_AVX512F_Usable; + CPU_FEATURE_SET (cpu_features, AVX512F); /* Determine if AVX512CD is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512CD)) - cpu_features->usable[index_arch_AVX512CD_Usable] - |= bit_arch_AVX512CD_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512CD); /* Determine if AVX512ER is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512ER)) - cpu_features->usable[index_arch_AVX512ER_Usable] - |= bit_arch_AVX512ER_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512ER); /* Determine if AVX512PF is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512PF)) - cpu_features->usable[index_arch_AVX512PF_Usable] - |= bit_arch_AVX512PF_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512PF); /* Determine if AVX512VL is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512VL)) - cpu_features->usable[index_arch_AVX512VL_Usable] - |= bit_arch_AVX512VL_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512VL); /* Determine if AVX512DQ is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512DQ)) - cpu_features->usable[index_arch_AVX512DQ_Usable] - |= bit_arch_AVX512DQ_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512DQ); /* Determine if AVX512BW is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512BW)) - cpu_features->usable[index_arch_AVX512BW_Usable] - |= bit_arch_AVX512BW_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512BW); /* Determine if AVX512_4FMAPS is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512_4FMAPS)) - cpu_features->usable[index_arch_AVX512_4FMAPS_Usable] - |= bit_arch_AVX512_4FMAPS_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512_4FMAPS); /* Determine if AVX512_4VNNIW is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512_4VNNIW)) - cpu_features->usable[index_arch_AVX512_4VNNIW_Usable] - |= bit_arch_AVX512_4VNNIW_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512_4VNNIW); /* Determine if AVX512_BITALG is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512_BITALG)) - cpu_features->usable[index_arch_AVX512_BITALG_Usable] - |= bit_arch_AVX512_BITALG_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512_BITALG); /* Determine if AVX512_IFMA is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512_IFMA)) - cpu_features->usable[index_arch_AVX512_IFMA_Usable] - |= bit_arch_AVX512_IFMA_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512_IFMA); /* Determine if AVX512_VBMI is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512_VBMI)) - cpu_features->usable[index_arch_AVX512_VBMI_Usable] - |= bit_arch_AVX512_VBMI_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512_VBMI); /* Determine if AVX512_VBMI2 is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512_VBMI2)) - cpu_features->usable[index_arch_AVX512_VBMI2_Usable] - |= bit_arch_AVX512_VBMI2_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512_VBMI2); /* Determine if is AVX512_VNNI usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512_VNNI)) - cpu_features->usable[index_arch_AVX512_VNNI_Usable] - |= bit_arch_AVX512_VNNI_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512_VNNI); /* Determine if AVX512_VPOPCNTDQ is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512_VPOPCNTDQ)) - cpu_features->usable[index_arch_AVX512_VPOPCNTDQ_Usable] - |= bit_arch_AVX512_VPOPCNTDQ_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, + AVX512_VPOPCNTDQ); /* Determine if AVX512_VP2INTERSECT is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, - AVX512_VP2INTERSECT)) - cpu_features->usable[index_arch_AVX512_VP2INTERSECT_Usable] - |= bit_arch_AVX512_VP2INTERSECT_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, + AVX512_VP2INTERSECT); /* Determine if AVX512_BF16 is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AVX512_BF16)) - cpu_features->usable[index_arch_AVX512_BF16_Usable] - |= bit_arch_AVX512_BF16_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AVX512_BF16); } } } @@ -244,19 +236,17 @@ get_common_indices (struct cpu_features *cpu_features, == (bit_XTILECFG_state | bit_XTILEDATA_state)) { /* Determine if AMX_BF16 is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AMX_BF16)) - cpu_features->usable[index_arch_AMX_BF16_Usable] - |= bit_arch_AMX_BF16_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AMX_BF16); /* Determine if AMX_TILE is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AMX_TILE)) - cpu_features->usable[index_arch_AMX_TILE_Usable] - |= bit_arch_AMX_TILE_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AMX_TILE); /* Determine if AMX_INT8 is usable. */ - if (CPU_FEATURES_CPU_P (cpu_features, AMX_INT8)) - cpu_features->usable[index_arch_AMX_INT8_Usable] - |= bit_arch_AMX_INT8_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, AMX_INT8); } + + /* XFD is usable only when OSXSAVE is enabled. */ + CPU_FEATURE_SET_USABLE (cpu_features, XFD); + /* For _dl_runtime_resolve, set xsave_state_size to xsave area size + integer register save size and align it to 64 bytes. */ if (cpu_features->basic.max_cpuid >= 0xd) @@ -318,8 +308,7 @@ get_common_indices (struct cpu_features *cpu_features, { cpu_features->xsave_state_size = ALIGN_UP (size + STATE_SAVE_OFFSET, 64); - cpu_features->usable[index_arch_XSAVEC_Usable] - |= bit_arch_XSAVEC_Usable; + CPU_FEATURE_SET (cpu_features, XSAVEC); } } } @@ -328,8 +317,79 @@ get_common_indices (struct cpu_features *cpu_features, /* Determine if PKU is usable. */ if (CPU_FEATURES_CPU_P (cpu_features, OSPKE)) - cpu_features->usable[index_arch_PKU_Usable] - |= bit_arch_PKU_Usable; + CPU_FEATURE_SET (cpu_features, PKU); +} + +static void +get_extended_indices (struct cpu_features *cpu_features) +{ + unsigned int eax, ebx, ecx, edx; + __cpuid (0x80000000, eax, ebx, ecx, edx); + if (eax >= 0x80000001) + __cpuid (0x80000001, + cpu_features->features[COMMON_CPUID_INDEX_80000001].cpuid.eax, + cpu_features->features[COMMON_CPUID_INDEX_80000001].cpuid.ebx, + cpu_features->features[COMMON_CPUID_INDEX_80000001].cpuid.ecx, + cpu_features->features[COMMON_CPUID_INDEX_80000001].cpuid.edx); + if (eax >= 0x80000007) + __cpuid (0x80000007, + cpu_features->features[COMMON_CPUID_INDEX_80000007].cpuid.eax, + cpu_features->features[COMMON_CPUID_INDEX_80000007].cpuid.ebx, + cpu_features->features[COMMON_CPUID_INDEX_80000007].cpuid.ecx, + cpu_features->features[COMMON_CPUID_INDEX_80000007].cpuid.edx); + if (eax >= 0x80000008) + __cpuid (0x80000008, + cpu_features->features[COMMON_CPUID_INDEX_80000008].cpuid.eax, + cpu_features->features[COMMON_CPUID_INDEX_80000008].cpuid.ebx, + cpu_features->features[COMMON_CPUID_INDEX_80000008].cpuid.ecx, + cpu_features->features[COMMON_CPUID_INDEX_80000008].cpuid.edx); +} + +static void +get_common_indices (struct cpu_features *cpu_features, + unsigned int *family, unsigned int *model, + unsigned int *extended_model, unsigned int *stepping) +{ + if (family) + { + unsigned int eax; + __cpuid (1, eax, + cpu_features->features[COMMON_CPUID_INDEX_1].cpuid.ebx, + cpu_features->features[COMMON_CPUID_INDEX_1].cpuid.ecx, + cpu_features->features[COMMON_CPUID_INDEX_1].cpuid.edx); + cpu_features->features[COMMON_CPUID_INDEX_1].cpuid.eax = eax; + *family = (eax >> 8) & 0x0f; + *model = (eax >> 4) & 0x0f; + *extended_model = (eax >> 12) & 0xf0; + *stepping = eax & 0x0f; + if (*family == 0x0f) + { + *family += (eax >> 20) & 0xff; + *model += *extended_model; + } + } + + if (cpu_features->basic.max_cpuid >= 7) + { + __cpuid_count (7, 0, + cpu_features->features[COMMON_CPUID_INDEX_7].cpuid.eax, + cpu_features->features[COMMON_CPUID_INDEX_7].cpuid.ebx, + cpu_features->features[COMMON_CPUID_INDEX_7].cpuid.ecx, + cpu_features->features[COMMON_CPUID_INDEX_7].cpuid.edx); + __cpuid_count (7, 1, + cpu_features->features[COMMON_CPUID_INDEX_7_ECX_1].cpuid.eax, + cpu_features->features[COMMON_CPUID_INDEX_7_ECX_1].cpuid.ebx, + cpu_features->features[COMMON_CPUID_INDEX_7_ECX_1].cpuid.ecx, + cpu_features->features[COMMON_CPUID_INDEX_7_ECX_1].cpuid.edx); + } + + if (cpu_features->basic.max_cpuid >= 0xd) + __cpuid_count (0xd, 1, + cpu_features->features[COMMON_CPUID_INDEX_D_ECX_1].cpuid.eax, + cpu_features->features[COMMON_CPUID_INDEX_D_ECX_1].cpuid.ebx, + cpu_features->features[COMMON_CPUID_INDEX_D_ECX_1].cpuid.ecx, + cpu_features->features[COMMON_CPUID_INDEX_D_ECX_1].cpuid.edx); + } _Static_assert (((index_arch_Fast_Unaligned_Load @@ -353,8 +413,6 @@ init_cpu_features (struct cpu_features *cpu_features) unsigned int stepping = 0; enum cpu_features_kind kind; - cpu_features->usable_p = cpu_features->usable; - #if !HAS_CPUID if (__get_cpuid_max (0, 0) == 0) { @@ -377,6 +435,8 @@ init_cpu_features (struct cpu_features *cpu_features) get_extended_indices (cpu_features); + update_usable (cpu_features); + if (family == 0x06) { model += extended_model; @@ -473,7 +533,7 @@ init_cpu_features (struct cpu_features *cpu_features) with stepping >= 4) to avoid TSX on kernels that weren't updated with the latest microcode package (which disables broken feature by default). */ - cpu_features->cpuid[index_cpu_RTM].reg_RTM &= ~bit_cpu_RTM; + CPU_FEATURE_UNSET (cpu_features, RTM); break; } } @@ -502,15 +562,15 @@ init_cpu_features (struct cpu_features *cpu_features) get_extended_indices (cpu_features); - ecx = cpu_features->cpuid[COMMON_CPUID_INDEX_1].ecx; + update_usable (cpu_features); - if (HAS_ARCH_FEATURE (AVX_Usable)) + ecx = cpu_features->features[COMMON_CPUID_INDEX_1].cpuid.ecx; + + if (CPU_FEATURE_USABLE_P (cpu_features, AVX)) { /* Since the FMA4 bit is in COMMON_CPUID_INDEX_80000001 and FMA4 requires AVX, determine if FMA4 is usable here. */ - if (CPU_FEATURES_CPU_P (cpu_features, FMA4)) - cpu_features->usable[index_arch_FMA4_Usable] - |= bit_arch_FMA4_Usable; + CPU_FEATURE_SET_USABLE (cpu_features, FMA4); } if (family == 0x15) @@ -541,13 +601,15 @@ init_cpu_features (struct cpu_features *cpu_features) get_extended_indices (cpu_features); + update_usable (cpu_features); + model += extended_model; if (family == 0x6) { if (model == 0xf || model == 0x19) { - cpu_features->usable[index_arch_AVX_Usable] - &= ~(bit_arch_AVX_Usable | bit_arch_AVX2_Usable); + CPU_FEATURE_UNSET (cpu_features, AVX); + CPU_FEATURE_UNSET (cpu_features, AVX2); cpu_features->preferred[index_arch_Slow_SSE4_2] |= bit_arch_Slow_SSE4_2; @@ -560,8 +622,8 @@ init_cpu_features (struct cpu_features *cpu_features) { if (model == 0x1b) { - cpu_features->usable[index_arch_AVX_Usable] - &= ~(bit_arch_AVX_Usable | bit_arch_AVX2_Usable); + CPU_FEATURE_UNSET (cpu_features, AVX); + CPU_FEATURE_UNSET (cpu_features, AVX2); cpu_features->preferred[index_arch_Slow_SSE4_2] |= bit_arch_Slow_SSE4_2; @@ -571,8 +633,8 @@ init_cpu_features (struct cpu_features *cpu_features) } else if (model == 0x3b) { - cpu_features->usable[index_arch_AVX_Usable] - &= ~(bit_arch_AVX_Usable | bit_arch_AVX2_Usable); + CPU_FEATURE_UNSET (cpu_features, AVX); + CPU_FEATURE_UNSET (cpu_features, AVX2); cpu_features->preferred[index_arch_AVX_Fast_Unaligned_Load] &= ~bit_arch_AVX_Fast_Unaligned_Load; @@ -583,6 +645,7 @@ init_cpu_features (struct cpu_features *cpu_features) { kind = arch_kind_other; get_common_indices (cpu_features, NULL, NULL, NULL, NULL); + update_usable (cpu_features); } /* Support i586 if CX8 is available. */ @@ -629,31 +692,30 @@ no_cpuid: { const char *platform = NULL; - if (CPU_FEATURES_ARCH_P (cpu_features, AVX512F_Usable) - && CPU_FEATURES_CPU_P (cpu_features, AVX512CD)) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX512CD)) { - if (CPU_FEATURES_CPU_P (cpu_features, AVX512ER)) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX512ER)) { - if (CPU_FEATURES_CPU_P (cpu_features, AVX512PF)) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX512PF)) platform = "xeon_phi"; } else { - if (CPU_FEATURES_CPU_P (cpu_features, AVX512BW) - && CPU_FEATURES_CPU_P (cpu_features, AVX512DQ) - && CPU_FEATURES_CPU_P (cpu_features, AVX512VL)) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX512BW) + && CPU_FEATURE_USABLE_P (cpu_features, AVX512DQ) + && CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)) GLRO(dl_hwcap) |= HWCAP_X86_AVX512_1; } } if (platform == NULL - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable) - && CPU_FEATURES_ARCH_P (cpu_features, FMA_Usable) - && CPU_FEATURES_CPU_P (cpu_features, BMI1) - && CPU_FEATURES_CPU_P (cpu_features, BMI2) - && CPU_FEATURES_CPU_P (cpu_features, LZCNT) - && CPU_FEATURES_CPU_P (cpu_features, MOVBE) - && CPU_FEATURES_CPU_P (cpu_features, POPCNT)) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2) + && CPU_FEATURE_USABLE_P (cpu_features, FMA) + && CPU_FEATURE_USABLE_P (cpu_features, BMI1) + && CPU_FEATURE_USABLE_P (cpu_features, BMI2) + && CPU_FEATURE_USABLE_P (cpu_features, LZCNT) + && CPU_FEATURE_USABLE_P (cpu_features, MOVBE) + && CPU_FEATURE_USABLE_P (cpu_features, POPCNT)) platform = "haswell"; if (platform != NULL) @@ -661,7 +723,7 @@ no_cpuid: } #else GLRO(dl_hwcap) = 0; - if (CPU_FEATURES_CPU_P (cpu_features, SSE2)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE2)) GLRO(dl_hwcap) |= HWCAP_X86_SSE2; if (CPU_FEATURES_ARCH_P (cpu_features, I686)) @@ -696,9 +758,9 @@ no_cpuid: GLIBC_TUNABLES=glibc.cpu.hwcaps=-IBT,-SHSTK */ unsigned int cet_feature = 0; - if (!HAS_CPU_FEATURE (IBT)) + if (!CPU_FEATURE_USABLE (IBT)) cet_feature |= GNU_PROPERTY_X86_FEATURE_1_IBT; - if (!HAS_CPU_FEATURE (SHSTK)) + if (!CPU_FEATURE_USABLE (SHSTK)) cet_feature |= GNU_PROPERTY_X86_FEATURE_1_SHSTK; if (cet_feature) diff --git a/sysdeps/x86/cpu-features.h b/sysdeps/x86/cpu-features.h index 0383131..a0b9b91 100644 --- a/sysdeps/x86/cpu-features.h +++ b/sysdeps/x86/cpu-features.h @@ -20,15 +20,6 @@ enum { - /* The integer bit array index for the first set of usable feature - bits. */ - USABLE_FEATURE_INDEX_1 = 0, - /* The current maximum size of the feature integer bit array. */ - USABLE_FEATURE_INDEX_MAX -}; - -enum -{ /* The integer bit array index for the first set of preferred feature bits. */ PREFERRED_FEATURE_INDEX_1 = 0, @@ -57,6 +48,12 @@ struct cpuid_registers unsigned int edx; }; +struct cpuid_features +{ + struct cpuid_registers cpuid; + struct cpuid_registers usable; +}; + enum cpu_features_kind { arch_kind_unknown = 0, @@ -78,9 +75,7 @@ struct cpu_features_basic struct cpu_features { struct cpu_features_basic basic; - unsigned int *usable_p; - struct cpuid_registers cpuid[COMMON_CPUID_INDEX_MAX]; - unsigned int usable[USABLE_FEATURE_INDEX_MAX]; + struct cpuid_features features[COMMON_CPUID_INDEX_MAX]; unsigned int preferred[PREFERRED_FEATURE_INDEX_MAX]; /* The state size for XSAVEC or XSAVE. The type must be unsigned long int so that we use @@ -91,7 +86,7 @@ struct cpu_features unsigned long int xsave_state_size; /* The full state size for XSAVE when XSAVEC is disabled by - GLIBC_TUNABLES=glibc.cpu.hwcaps=-XSAVEC_Usable + GLIBC_TUNABLES=glibc.cpu.hwcaps=-XSAVEC */ unsigned int xsave_state_full_size; /* Data cache size for use in memory and string routines, typically @@ -114,117 +109,40 @@ extern const struct cpu_features *__get_cpu_features (void) __attribute__ ((const)); /* 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_##name[index_arch_##name] & (bit_arch_##name)) != 0) +#define CPU_FEATURE_CHECK_P(ptr, name, check) \ + ((ptr->features[index_cpu_##name].check.reg_##name \ + & bit_cpu_##name) != 0) +#define CPU_FEATURE_SET(ptr, name) \ + ptr->features[index_cpu_##name].usable.reg_##name |= bit_cpu_##name; +#define CPU_FEATURE_UNSET(ptr, name) \ + ptr->features[index_cpu_##name].usable.reg_##name &= ~bit_cpu_##name; +#define CPU_FEATURE_SET_USABLE(ptr, name) \ + ptr->features[index_cpu_##name].usable.reg_##name \ + |= ptr->features[index_cpu_##name].cpuid.reg_##name & bit_cpu_##name; +#define CPU_FEATURE_PREFERRED_P(ptr, name) \ + ((ptr->preferred[index_arch_##name] & bit_arch_##name) != 0) +#define CPU_FEATURE_CPU_P(ptr, name) \ + CPU_FEATURE_CHECK_P (ptr, name, cpuid) +#define CPU_FEATURE_USABLE_P(ptr, name) \ + CPU_FEATURE_CHECK_P (ptr, name, usable) /* HAS_CPU_FEATURE evaluates to true if CPU supports the feature. */ #define HAS_CPU_FEATURE(name) \ - CPU_FEATURES_CPU_P (__get_cpu_features (), name) -/* HAS_ARCH_FEATURE evaluates to true if we may use the feature at - runtime. */ -# define HAS_ARCH_FEATURE(name) \ - CPU_FEATURES_ARCH_P (__get_cpu_features (), name) + CPU_FEATURE_CPU_P (__get_cpu_features (), name) /* CPU_FEATURE_USABLE evaluates to true if the feature is usable. */ #define CPU_FEATURE_USABLE(name) \ - HAS_ARCH_FEATURE (name##_Usable) - -/* Architecture features. */ - -/* USABLE_FEATURE_INDEX_1. */ -#define bit_arch_AVX_Usable (1u << 0) -#define bit_arch_AVX2_Usable (1u << 1) -#define bit_arch_AVX512F_Usable (1u << 2) -#define bit_arch_AVX512CD_Usable (1u << 3) -#define bit_arch_AVX512ER_Usable (1u << 4) -#define bit_arch_AVX512PF_Usable (1u << 5) -#define bit_arch_AVX512VL_Usable (1u << 6) -#define bit_arch_AVX512DQ_Usable (1u << 7) -#define bit_arch_AVX512BW_Usable (1u << 8) -#define bit_arch_AVX512_4FMAPS_Usable (1u << 9) -#define bit_arch_AVX512_4VNNIW_Usable (1u << 10) -#define bit_arch_AVX512_BITALG_Usable (1u << 11) -#define bit_arch_AVX512_IFMA_Usable (1u << 12) -#define bit_arch_AVX512_VBMI_Usable (1u << 13) -#define bit_arch_AVX512_VBMI2_Usable (1u << 14) -#define bit_arch_AVX512_VNNI_Usable (1u << 15) -#define bit_arch_AVX512_VPOPCNTDQ_Usable (1u << 16) -#define bit_arch_FMA_Usable (1u << 17) -#define bit_arch_FMA4_Usable (1u << 18) -#define bit_arch_VAES_Usable (1u << 19) -#define bit_arch_VPCLMULQDQ_Usable (1u << 20) -#define bit_arch_XOP_Usable (1u << 21) -#define bit_arch_XSAVEC_Usable (1u << 22) -#define bit_arch_F16C_Usable (1u << 23) -#define bit_arch_AVX512_VP2INTERSECT_Usable (1u << 24) -#define bit_arch_AVX512_BF16_Usable (1u << 25) -#define bit_arch_PKU_Usable (1u << 26) -#define bit_arch_AMX_BF16_Usable (1u << 27) -#define bit_arch_AMX_TILE_Usable (1u << 28) -#define bit_arch_AMX_INT8_Usable (1u << 29) - -#define index_arch_AVX_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX2_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512F_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512CD_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512ER_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512PF_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512VL_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512BW_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512DQ_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512_4FMAPS_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512_4VNNIW_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512_BITALG_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512_IFMA_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512_VBMI_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512_VBMI2_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512_VNNI_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512_VPOPCNTDQ_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_FMA_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_FMA4_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_VAES_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_VPCLMULQDQ_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_XOP_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_XSAVEC_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_F16C_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512_VP2INTERSECT_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AVX512_BF16_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_PKU_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AMX_BF16_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AMX_TILE_Usable USABLE_FEATURE_INDEX_1 -#define index_arch_AMX_INT8_Usable USABLE_FEATURE_INDEX_1 - -#define feature_AVX_Usable usable -#define feature_AVX2_Usable usable -#define feature_AVX512F_Usable usable -#define feature_AVX512CD_Usable usable -#define feature_AVX512ER_Usable usable -#define feature_AVX512PF_Usable usable -#define feature_AVX512VL_Usable usable -#define feature_AVX512BW_Usable usable -#define feature_AVX512DQ_Usable usable -#define feature_AVX512_4FMAPS_Usable usable -#define feature_AVX512_4VNNIW_Usable usable -#define feature_AVX512_BITALG_Usable usable -#define feature_AVX512_IFMA_Usable usable -#define feature_AVX512_VBMI_Usable usable -#define feature_AVX512_VBMI2_Usable usable -#define feature_AVX512_VNNI_Usable usable -#define feature_AVX512_VPOPCNTDQ_Usable usable -#define feature_FMA_Usable usable -#define feature_FMA4_Usable usable -#define feature_VAES_Usable usable -#define feature_VPCLMULQDQ_Usable usable -#define feature_XOP_Usable usable -#define feature_XSAVEC_Usable usable -#define feature_F16C_Usable usable -#define feature_AVX512_VP2INTERSECT_Usable usable -#define feature_AVX512_BF16_Usable usable -#define feature_PKU_Usable usable -#define feature_AMX_BF16_Usable usable -#define feature_AMX_TILE_Usable usable -#define feature_AMX_INT8_Usable usable + CPU_FEATURE_USABLE_P (__get_cpu_features (), name) +/* CPU_FEATURE_PREFER evaluates to true if we prefer the feature at + runtime. */ +#define CPU_FEATURE_PREFERRED(name) \ + CPU_FEATURE_PREFERRED_P(__get_cpu_features (), name) + +#define CPU_FEATURES_CPU_P(ptr, name) \ + CPU_FEATURE_CPU_P (ptr, name) +#define CPU_FEATURES_ARCH_P(ptr, name) \ + CPU_FEATURE_PREFERRED_P (ptr, name) +#define HAS_ARCH_FEATURE(name) \ + CPU_FEATURE_PREFERRED (name) /* CPU features. */ @@ -247,6 +165,7 @@ extern const struct cpu_features *__get_cpu_features (void) #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) @@ -261,6 +180,7 @@ extern const struct cpu_features *__get_cpu_features (void) #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) @@ -273,6 +193,7 @@ extern const struct cpu_features *__get_cpu_features (void) #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) @@ -282,6 +203,7 @@ extern const struct cpu_features *__get_cpu_features (void) #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) @@ -291,6 +213,7 @@ extern const struct cpu_features *__get_cpu_features (void) #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) /* COMMON_CPUID_INDEX_7. */ @@ -302,12 +225,14 @@ extern const struct cpu_features *__get_cpu_features (void) #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_PQM (1u << 12) +#define bit_cpu_DEPR_FPU_CS_DS (1u << 13) #define bit_cpu_MPX (1u << 14) #define bit_cpu_PQE (1u << 15) #define bit_cpu_AVX512F (1u << 16) @@ -316,6 +241,7 @@ extern const struct cpu_features *__get_cpu_features (void) #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) @@ -340,9 +266,17 @@ extern const struct cpu_features *__get_cpu_features (void) #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_INDEX_7_ECX_23 (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) @@ -350,17 +284,30 @@ extern const struct cpu_features *__get_cpu_features (void) #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_INDEX_7_EDX_5 (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_INDEX_7_EDX_11 (1u << 11) +#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_INDEX_7_EDX_23 (1u << 23) #define bit_cpu_AMX_TILE (1u << 24) #define bit_cpu_AMX_INT8 (1u << 25) #define bit_cpu_IBRS_IBPB (1u << 26) @@ -433,6 +380,7 @@ extern const struct cpu_features *__get_cpu_features (void) #define index_cpu_CMPXCHG16B COMMON_CPUID_INDEX_1 #define index_cpu_XTPRUPDCTRL COMMON_CPUID_INDEX_1 #define index_cpu_PDCM COMMON_CPUID_INDEX_1 +#define index_cpu_INDEX_1_ECX_16 COMMON_CPUID_INDEX_1 #define index_cpu_PCID COMMON_CPUID_INDEX_1 #define index_cpu_DCA COMMON_CPUID_INDEX_1 #define index_cpu_SSE4_1 COMMON_CPUID_INDEX_1 @@ -447,6 +395,7 @@ extern const struct cpu_features *__get_cpu_features (void) #define index_cpu_AVX COMMON_CPUID_INDEX_1 #define index_cpu_F16C COMMON_CPUID_INDEX_1 #define index_cpu_RDRAND COMMON_CPUID_INDEX_1 +#define index_cpu_INDEX_1_ECX_31 COMMON_CPUID_INDEX_1 /* ECX. */ #define index_cpu_FPU COMMON_CPUID_INDEX_1 @@ -459,6 +408,7 @@ extern const struct cpu_features *__get_cpu_features (void) #define index_cpu_MCE COMMON_CPUID_INDEX_1 #define index_cpu_CX8 COMMON_CPUID_INDEX_1 #define index_cpu_APIC COMMON_CPUID_INDEX_1 +#define index_cpu_INDEX_1_EDX_10 COMMON_CPUID_INDEX_1 #define index_cpu_SEP COMMON_CPUID_INDEX_1 #define index_cpu_MTRR COMMON_CPUID_INDEX_1 #define index_cpu_PGE COMMON_CPUID_INDEX_1 @@ -468,6 +418,7 @@ extern const struct cpu_features *__get_cpu_features (void) #define index_cpu_PSE_36 COMMON_CPUID_INDEX_1 #define index_cpu_PSN COMMON_CPUID_INDEX_1 #define index_cpu_CLFSH COMMON_CPUID_INDEX_1 +#define index_cpu_INDEX_1_EDX_20 COMMON_CPUID_INDEX_1 #define index_cpu_DS COMMON_CPUID_INDEX_1 #define index_cpu_ACPI COMMON_CPUID_INDEX_1 #define index_cpu_MMX COMMON_CPUID_INDEX_1 @@ -477,6 +428,7 @@ extern const struct cpu_features *__get_cpu_features (void) #define index_cpu_SS COMMON_CPUID_INDEX_1 #define index_cpu_HTT COMMON_CPUID_INDEX_1 #define index_cpu_TM COMMON_CPUID_INDEX_1 +#define index_cpu_INDEX_1_EDX_30 COMMON_CPUID_INDEX_1 #define index_cpu_PBE COMMON_CPUID_INDEX_1 /* COMMON_CPUID_INDEX_7. */ @@ -488,12 +440,14 @@ extern const struct cpu_features *__get_cpu_features (void) #define index_cpu_BMI1 COMMON_CPUID_INDEX_7 #define index_cpu_HLE COMMON_CPUID_INDEX_7 #define index_cpu_AVX2 COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EBX_6 COMMON_CPUID_INDEX_7 #define index_cpu_SMEP COMMON_CPUID_INDEX_7 #define index_cpu_BMI2 COMMON_CPUID_INDEX_7 #define index_cpu_ERMS COMMON_CPUID_INDEX_7 #define index_cpu_INVPCID COMMON_CPUID_INDEX_7 #define index_cpu_RTM COMMON_CPUID_INDEX_7 #define index_cpu_PQM COMMON_CPUID_INDEX_7 +#define index_cpu_DEPR_FPU_CS_DS COMMON_CPUID_INDEX_7 #define index_cpu_MPX COMMON_CPUID_INDEX_7 #define index_cpu_PQE COMMON_CPUID_INDEX_7 #define index_cpu_AVX512F COMMON_CPUID_INDEX_7 @@ -502,6 +456,7 @@ extern const struct cpu_features *__get_cpu_features (void) #define index_cpu_ADX COMMON_CPUID_INDEX_7 #define index_cpu_SMAP COMMON_CPUID_INDEX_7 #define index_cpu_AVX512_IFMA COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EBX_22 COMMON_CPUID_INDEX_7 #define index_cpu_CLFLUSHOPT COMMON_CPUID_INDEX_7 #define index_cpu_CLWB COMMON_CPUID_INDEX_7 #define index_cpu_TRACE COMMON_CPUID_INDEX_7 @@ -526,9 +481,15 @@ extern const struct cpu_features *__get_cpu_features (void) #define index_cpu_VPCLMULQDQ COMMON_CPUID_INDEX_7 #define index_cpu_AVX512_VNNI COMMON_CPUID_INDEX_7 #define index_cpu_AVX512_BITALG COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_ECX_13 COMMON_CPUID_INDEX_7 #define index_cpu_AVX512_VPOPCNTDQ COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_ECX_15 COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_ECX_16 COMMON_CPUID_INDEX_7 #define index_cpu_RDPID COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_ECX_23 COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_ECX_24 COMMON_CPUID_INDEX_7 #define index_cpu_CLDEMOTE COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_ECX_26 COMMON_CPUID_INDEX_7 #define index_cpu_MOVDIRI COMMON_CPUID_INDEX_7 #define index_cpu_MOVDIR64B COMMON_CPUID_INDEX_7 #define index_cpu_ENQCMD COMMON_CPUID_INDEX_7 @@ -536,17 +497,30 @@ extern const struct cpu_features *__get_cpu_features (void) #define index_cpu_PKS COMMON_CPUID_INDEX_7 /* EDX. */ +#define index_cpu_INDEX_7_EDX_0 COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EDX_1 COMMON_CPUID_INDEX_7 #define index_cpu_AVX512_4VNNIW COMMON_CPUID_INDEX_7 #define index_cpu_AVX512_4FMAPS COMMON_CPUID_INDEX_7 #define index_cpu_FSRM COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EDX_5 COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EDX_6 COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EDX_7 COMMON_CPUID_INDEX_7 #define index_cpu_AVX512_VP2INTERSECT COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EDX_9 COMMON_CPUID_INDEX_7 #define index_cpu_MD_CLEAR COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EDX_11 COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EDX_12 COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EDX_13 COMMON_CPUID_INDEX_7 #define index_cpu_SERIALIZE COMMON_CPUID_INDEX_7 #define index_cpu_HYBRID COMMON_CPUID_INDEX_7 #define index_cpu_TSXLDTRK COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EDX_17 COMMON_CPUID_INDEX_7 #define index_cpu_PCONFIG COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EDX_19 COMMON_CPUID_INDEX_7 #define index_cpu_IBT COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EDX_21 COMMON_CPUID_INDEX_7 #define index_cpu_AMX_BF16 COMMON_CPUID_INDEX_7 +#define index_cpu_INDEX_7_EDX_23 COMMON_CPUID_INDEX_7 #define index_cpu_AMX_TILE COMMON_CPUID_INDEX_7 #define index_cpu_AMX_INT8 COMMON_CPUID_INDEX_7 #define index_cpu_IBRS_IBPB COMMON_CPUID_INDEX_7 @@ -619,6 +593,7 @@ extern const struct cpu_features *__get_cpu_features (void) #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 @@ -633,6 +608,7 @@ extern const struct cpu_features *__get_cpu_features (void) #define reg_AVX ecx #define reg_F16C ecx #define reg_RDRAND ecx +#define reg_INDEX_1_ECX_31 ecx /* EDX. */ #define reg_FPU edx @@ -645,6 +621,7 @@ extern const struct cpu_features *__get_cpu_features (void) #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 @@ -654,6 +631,7 @@ extern const struct cpu_features *__get_cpu_features (void) #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 @@ -663,6 +641,7 @@ extern const struct cpu_features *__get_cpu_features (void) #define reg_SS edx #define reg_HTT edx #define reg_TM edx +#define reg_INDEX_1_EDX_30 edx #define reg_PBE edx /* COMMON_CPUID_INDEX_7. */ @@ -675,11 +654,13 @@ extern const struct cpu_features *__get_cpu_features (void) #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_PQM ebx +#define reg_DEPR_FPU_CS_DS ebx #define reg_MPX ebx #define reg_PQE ebx #define reg_AVX512F ebx @@ -688,6 +669,7 @@ extern const struct cpu_features *__get_cpu_features (void) #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 @@ -712,9 +694,15 @@ extern const struct cpu_features *__get_cpu_features (void) #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_INDEX_7_ECX_23 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 @@ -722,17 +710,30 @@ extern const struct cpu_features *__get_cpu_features (void) #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_INDEX_7_EDX_5 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_INDEX_7_EDX_11 edx +#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_INDEX_7_EDX_23 edx #define reg_AMX_TILE edx #define reg_AMX_INT8 edx #define reg_IBRS_IBPB edx @@ -821,23 +822,6 @@ extern const struct cpu_features *__get_cpu_features (void) #define index_arch_MathVec_Prefer_No_AVX512 PREFERRED_FEATURE_INDEX_1 #define index_arch_Prefer_FSRM PREFERRED_FEATURE_INDEX_1 -#define feature_Fast_Rep_String preferred -#define feature_Fast_Copy_Backward preferred -#define feature_Slow_BSF preferred -#define feature_Fast_Unaligned_Load preferred -#define feature_Prefer_PMINUB_for_stringop preferred -#define feature_Fast_Unaligned_Copy preferred -#define feature_I586 preferred -#define feature_I686 preferred -#define feature_Slow_SSE4_2 preferred -#define feature_AVX_Fast_Unaligned_Load preferred -#define feature_Prefer_MAP_32BIT_EXEC preferred -#define feature_Prefer_No_VZEROUPPER preferred -#define feature_Prefer_ERMS preferred -#define feature_Prefer_No_AVX512 preferred -#define feature_MathVec_Prefer_No_AVX512 preferred -#define feature_Prefer_FSRM preferred - /* XCR0 Feature flags. */ #define bit_XMM_state (1u << 1) #define bit_YMM_state (1u << 2) @@ -851,8 +835,6 @@ extern const struct cpu_features *__get_cpu_features (void) /* Unused for x86. */ # define INIT_ARCH() # define __get_cpu_features() (&GLRO(dl_x86_cpu_features)) -# define x86_get_cpuid_registers(i) \ - (&(GLRO(dl_x86_cpu_features).cpuid[i])) # endif #ifdef __x86_64__ diff --git a/sysdeps/x86/cpu-tunables.c b/sysdeps/x86/cpu-tunables.c index 666ec57..588bbf9 100644 --- a/sysdeps/x86/cpu-tunables.c +++ b/sysdeps/x86/cpu-tunables.c @@ -43,66 +43,45 @@ extern __typeof (memcmp) DEFAULT_MEMCMP; _Static_assert (sizeof (#name) - 1 == len, #name " != " #len); \ if (!DEFAULT_MEMCMP (f, #name, len)) \ { \ - cpu_features->cpuid[index_cpu_##name].reg_##name \ - &= ~bit_cpu_##name; \ + CPU_FEATURE_UNSET (cpu_features, name) \ break; \ } -/* Disable an ARCH feature NAME. We don't enable an ARCH feature which - isn't available. */ -# define CHECK_GLIBC_IFUNC_ARCH_OFF(f, cpu_features, name, len) \ +/* Disable a preferred feature NAME. We don't enable a preferred feature + which isn't available. */ +# define CHECK_GLIBC_IFUNC_PREFERRED_OFF(f, cpu_features, name, len) \ _Static_assert (sizeof (#name) - 1 == len, #name " != " #len); \ if (!DEFAULT_MEMCMP (f, #name, len)) \ { \ - cpu_features->feature_##name[index_arch_##name] \ + cpu_features->preferred[index_arch_##name] \ &= ~bit_arch_##name; \ break; \ } -/* Enable/disable an ARCH feature NAME. */ -# define CHECK_GLIBC_IFUNC_ARCH_BOTH(f, cpu_features, name, disable, \ - len) \ +/* Enable/disable a preferred feature NAME. */ +# define CHECK_GLIBC_IFUNC_PREFERRED_BOTH(f, cpu_features, name, \ + disable, len) \ _Static_assert (sizeof (#name) - 1 == len, #name " != " #len); \ if (!DEFAULT_MEMCMP (f, #name, len)) \ { \ if (disable) \ - cpu_features->feature_##name[index_arch_##name] \ - &= ~bit_arch_##name; \ + cpu_features->preferred[index_arch_##name] &= ~bit_arch_##name; \ else \ - cpu_features->feature_##name[index_arch_##name] \ - |= bit_arch_##name; \ + cpu_features->preferred[index_arch_##name] |= bit_arch_##name; \ break; \ } -/* Enable/disable an ARCH feature NAME. Enable an ARCH feature only - if the ARCH feature NEED is also enabled. */ -# define CHECK_GLIBC_IFUNC_ARCH_NEED_ARCH_BOTH(f, cpu_features, name, \ +/* Enable/disable a preferred feature NAME. Enable a preferred feature + only if the feature NEED is usable. */ +# define CHECK_GLIBC_IFUNC_PREFERRED_NEED_BOTH(f, cpu_features, name, \ need, disable, len) \ _Static_assert (sizeof (#name) - 1 == len, #name " != " #len); \ if (!DEFAULT_MEMCMP (f, #name, len)) \ { \ if (disable) \ - cpu_features->feature_##name[index_arch_##name] \ - &= ~bit_arch_##name; \ - else if (CPU_FEATURES_ARCH_P (cpu_features, need)) \ - cpu_features->feature_##name[index_arch_##name] \ - |= bit_arch_##name; \ - break; \ - } - -/* Enable/disable an ARCH feature NAME. Enable an ARCH feature only - if the CPU feature NEED is also enabled. */ -# define CHECK_GLIBC_IFUNC_ARCH_NEED_CPU_BOTH(f, cpu_features, name, \ - need, disable, len) \ - _Static_assert (sizeof (#name) - 1 == len, #name " != " #len); \ - if (!DEFAULT_MEMCMP (f, #name, len)) \ - { \ - if (disable) \ - cpu_features->feature_##name[index_arch_##name] \ - &= ~bit_arch_##name; \ - else if (CPU_FEATURES_CPU_P (cpu_features, need)) \ - cpu_features->feature_##name[index_arch_##name] \ - |= bit_arch_##name; \ + cpu_features->preferred[index_arch_##name] &= ~bit_arch_##name; \ + else if (CPU_FEATURE_USABLE_P (cpu_features, need)) \ + cpu_features->preferred[index_arch_##name] |= bit_arch_##name; \ break; \ } @@ -178,8 +157,8 @@ TUNABLE_CALLBACK (set_hwcaps) (tunable_val_t *valp) CHECK_GLIBC_IFUNC_CPU_OFF (n, cpu_features, ERMS, 4); CHECK_GLIBC_IFUNC_CPU_OFF (n, cpu_features, FMA4, 4); CHECK_GLIBC_IFUNC_CPU_OFF (n, cpu_features, SSE2, 4); - CHECK_GLIBC_IFUNC_ARCH_OFF (n, cpu_features, I586, 4); - CHECK_GLIBC_IFUNC_ARCH_OFF (n, cpu_features, I686, 4); + CHECK_GLIBC_IFUNC_PREFERRED_OFF (n, cpu_features, I586, 4); + CHECK_GLIBC_IFUNC_PREFERRED_OFF (n, cpu_features, I686, 4); } break; case 5: @@ -197,6 +176,13 @@ TUNABLE_CALLBACK (set_hwcaps) (tunable_val_t *valp) CHECK_GLIBC_IFUNC_CPU_OFF (n, cpu_features, POPCNT, 6); CHECK_GLIBC_IFUNC_CPU_OFF (n, cpu_features, SSE4_1, 6); CHECK_GLIBC_IFUNC_CPU_OFF (n, cpu_features, SSE4_2, 6); + if (!DEFAULT_MEMCMP (n, "XSAVEC", 6)) + { + /* Update xsave_state_size to XSAVE state size. */ + cpu_features->xsave_state_size + = cpu_features->xsave_state_full_size; + CPU_FEATURE_UNSET (cpu_features, XSAVEC); + } } break; case 7: @@ -216,115 +202,85 @@ TUNABLE_CALLBACK (set_hwcaps) (tunable_val_t *valp) CHECK_GLIBC_IFUNC_CPU_OFF (n, cpu_features, AVX512PF, 8); CHECK_GLIBC_IFUNC_CPU_OFF (n, cpu_features, AVX512VL, 8); } - CHECK_GLIBC_IFUNC_ARCH_BOTH (n, cpu_features, Slow_BSF, - disable, 8); - break; - case 10: - if (disable) - { - CHECK_GLIBC_IFUNC_ARCH_OFF (n, cpu_features, AVX_Usable, - 10); - CHECK_GLIBC_IFUNC_ARCH_OFF (n, cpu_features, FMA_Usable, - 10); - } + CHECK_GLIBC_IFUNC_PREFERRED_BOTH (n, cpu_features, Slow_BSF, + disable, 8); break; case 11: - if (disable) { - CHECK_GLIBC_IFUNC_ARCH_OFF (n, cpu_features, AVX2_Usable, - 11); - CHECK_GLIBC_IFUNC_ARCH_OFF (n, cpu_features, FMA4_Usable, - 11); - } - CHECK_GLIBC_IFUNC_ARCH_BOTH (n, cpu_features, Prefer_ERMS, - disable, 11); - CHECK_GLIBC_IFUNC_ARCH_NEED_CPU_BOTH (n, cpu_features, - Slow_SSE4_2, SSE4_2, + CHECK_GLIBC_IFUNC_PREFERRED_BOTH (n, cpu_features, + Prefer_ERMS, disable, 11); - CHECK_GLIBC_IFUNC_ARCH_BOTH (n, cpu_features, Prefer_FSRM, - disable, 11); - break; - case 13: - if (disable) - { - /* Update xsave_state_size to XSAVE state size. */ - cpu_features->xsave_state_size - = cpu_features->xsave_state_full_size; - CHECK_GLIBC_IFUNC_ARCH_OFF (n, cpu_features, - XSAVEC_Usable, 13); - } - break; - case 14: - if (disable) - { - CHECK_GLIBC_IFUNC_ARCH_OFF (n, cpu_features, - AVX512F_Usable, 14); + CHECK_GLIBC_IFUNC_PREFERRED_BOTH (n, cpu_features, + Prefer_FSRM, + disable, 11); + CHECK_GLIBC_IFUNC_PREFERRED_NEED_BOTH (n, cpu_features, + Slow_SSE4_2, + SSE4_2, + disable, 11); } break; case 15: - if (disable) { - CHECK_GLIBC_IFUNC_ARCH_OFF (n, cpu_features, - AVX512DQ_Usable, 15); + CHECK_GLIBC_IFUNC_PREFERRED_BOTH (n, cpu_features, + Fast_Rep_String, + disable, 15); } - CHECK_GLIBC_IFUNC_ARCH_BOTH (n, cpu_features, Fast_Rep_String, - disable, 15); break; case 16: { - CHECK_GLIBC_IFUNC_ARCH_NEED_ARCH_BOTH - (n, cpu_features, Prefer_No_AVX512, AVX512F_Usable, + CHECK_GLIBC_IFUNC_PREFERRED_NEED_BOTH + (n, cpu_features, Prefer_No_AVX512, AVX512F, disable, 16); } break; case 18: { - CHECK_GLIBC_IFUNC_ARCH_BOTH (n, cpu_features, - Fast_Copy_Backward, disable, - 18); + CHECK_GLIBC_IFUNC_PREFERRED_BOTH (n, cpu_features, + Fast_Copy_Backward, + disable, 18); } break; case 19: { - CHECK_GLIBC_IFUNC_ARCH_BOTH (n, cpu_features, - Fast_Unaligned_Load, disable, - 19); - CHECK_GLIBC_IFUNC_ARCH_BOTH (n, cpu_features, - Fast_Unaligned_Copy, disable, - 19); + CHECK_GLIBC_IFUNC_PREFERRED_BOTH (n, cpu_features, + Fast_Unaligned_Load, + disable, 19); + CHECK_GLIBC_IFUNC_PREFERRED_BOTH (n, cpu_features, + Fast_Unaligned_Copy, + disable, 19); } break; case 20: { - CHECK_GLIBC_IFUNC_ARCH_NEED_ARCH_BOTH - (n, cpu_features, Prefer_No_VZEROUPPER, AVX_Usable, - disable, 20); + CHECK_GLIBC_IFUNC_PREFERRED_NEED_BOTH + (n, cpu_features, Prefer_No_VZEROUPPER, AVX, disable, + 20); } break; case 21: { - CHECK_GLIBC_IFUNC_ARCH_BOTH (n, cpu_features, - Prefer_MAP_32BIT_EXEC, disable, - 21); + CHECK_GLIBC_IFUNC_PREFERRED_BOTH (n, cpu_features, + Prefer_MAP_32BIT_EXEC, + disable, 21); } break; case 23: { - CHECK_GLIBC_IFUNC_ARCH_NEED_ARCH_BOTH - (n, cpu_features, AVX_Fast_Unaligned_Load, AVX_Usable, + CHECK_GLIBC_IFUNC_PREFERRED_NEED_BOTH + (n, cpu_features, AVX_Fast_Unaligned_Load, AVX, disable, 23); } break; case 24: { - CHECK_GLIBC_IFUNC_ARCH_NEED_ARCH_BOTH - (n, cpu_features, MathVec_Prefer_No_AVX512, - AVX512F_Usable, disable, 24); + CHECK_GLIBC_IFUNC_PREFERRED_NEED_BOTH + (n, cpu_features, MathVec_Prefer_No_AVX512, AVX512F, + disable, 24); } break; case 26: { - CHECK_GLIBC_IFUNC_ARCH_NEED_CPU_BOTH + CHECK_GLIBC_IFUNC_PREFERRED_NEED_BOTH (n, cpu_features, Prefer_PMINUB_for_stringop, SSE2, disable, 26); } diff --git a/sysdeps/x86/dl-cet.c b/sysdeps/x86/dl-cet.c index 5524b66..03572f7 100644 --- a/sysdeps/x86/dl-cet.c +++ b/sysdeps/x86/dl-cet.c @@ -74,10 +74,10 @@ dl_cet_check (struct link_map *m, const char *program) GLIBC_TUNABLES=glibc.cpu.hwcaps=-IBT,-SHSTK */ - enable_ibt &= (HAS_CPU_FEATURE (IBT) + enable_ibt &= (CPU_FEATURE_USABLE (IBT) && (enable_ibt_type == cet_always_on || (m->l_cet & lc_ibt) != 0)); - enable_shstk &= (HAS_CPU_FEATURE (SHSTK) + enable_shstk &= (CPU_FEATURE_USABLE (SHSTK) && (enable_shstk_type == cet_always_on || (m->l_cet & lc_shstk) != 0)); } diff --git a/sysdeps/x86/tst-get-cpu-features.c b/sysdeps/x86/tst-get-cpu-features.c index 2cff2e8..080c58e7 100644 --- a/sysdeps/x86/tst-get-cpu-features.c +++ b/sysdeps/x86/tst-get-cpu-features.c @@ -139,6 +139,7 @@ do_test (void) CHECK_CPU_FEATURE (INVPCID); CHECK_CPU_FEATURE (RTM); CHECK_CPU_FEATURE (PQM); + CHECK_CPU_FEATURE (DEPR_FPU_CS_DS); CHECK_CPU_FEATURE (MPX); CHECK_CPU_FEATURE (PQE); CHECK_CPU_FEATURE (AVX512F); @@ -220,35 +221,156 @@ do_test (void) CHECK_CPU_FEATURE (AVX512_BF16); printf ("Usable CPU features:\n"); + CHECK_CPU_FEATURE_USABLE (SSE3); + CHECK_CPU_FEATURE_USABLE (PCLMULQDQ); + CHECK_CPU_FEATURE_USABLE (DTES64); + CHECK_CPU_FEATURE_USABLE (MONITOR); + CHECK_CPU_FEATURE_USABLE (DS_CPL); + CHECK_CPU_FEATURE_USABLE (VMX); + CHECK_CPU_FEATURE_USABLE (SMX); + CHECK_CPU_FEATURE_USABLE (EST); + CHECK_CPU_FEATURE_USABLE (TM2); + CHECK_CPU_FEATURE_USABLE (SSSE3); + CHECK_CPU_FEATURE_USABLE (CNXT_ID); + CHECK_CPU_FEATURE_USABLE (SDBG); CHECK_CPU_FEATURE_USABLE (FMA); + CHECK_CPU_FEATURE_USABLE (CMPXCHG16B); + CHECK_CPU_FEATURE_USABLE (XTPRUPDCTRL); + CHECK_CPU_FEATURE_USABLE (PDCM); + CHECK_CPU_FEATURE_USABLE (PCID); + CHECK_CPU_FEATURE_USABLE (DCA); + CHECK_CPU_FEATURE_USABLE (SSE4_1); + CHECK_CPU_FEATURE_USABLE (SSE4_2); + CHECK_CPU_FEATURE_USABLE (X2APIC); + CHECK_CPU_FEATURE_USABLE (MOVBE); + CHECK_CPU_FEATURE_USABLE (POPCNT); + CHECK_CPU_FEATURE_USABLE (TSC_DEADLINE); + CHECK_CPU_FEATURE_USABLE (AES); + CHECK_CPU_FEATURE_USABLE (XSAVE); + CHECK_CPU_FEATURE_USABLE (OSXSAVE); CHECK_CPU_FEATURE_USABLE (AVX); CHECK_CPU_FEATURE_USABLE (F16C); + CHECK_CPU_FEATURE_USABLE (RDRAND); + CHECK_CPU_FEATURE_USABLE (FPU); + CHECK_CPU_FEATURE_USABLE (VME); + CHECK_CPU_FEATURE_USABLE (DE); + CHECK_CPU_FEATURE_USABLE (PSE); + CHECK_CPU_FEATURE_USABLE (TSC); + CHECK_CPU_FEATURE_USABLE (MSR); + CHECK_CPU_FEATURE_USABLE (PAE); + CHECK_CPU_FEATURE_USABLE (MCE); + CHECK_CPU_FEATURE_USABLE (CX8); + CHECK_CPU_FEATURE_USABLE (APIC); + CHECK_CPU_FEATURE_USABLE (SEP); + CHECK_CPU_FEATURE_USABLE (MTRR); + CHECK_CPU_FEATURE_USABLE (PGE); + CHECK_CPU_FEATURE_USABLE (MCA); + CHECK_CPU_FEATURE_USABLE (CMOV); + CHECK_CPU_FEATURE_USABLE (PAT); + CHECK_CPU_FEATURE_USABLE (PSE_36); + CHECK_CPU_FEATURE_USABLE (PSN); + CHECK_CPU_FEATURE_USABLE (CLFSH); + CHECK_CPU_FEATURE_USABLE (DS); + CHECK_CPU_FEATURE_USABLE (ACPI); + CHECK_CPU_FEATURE_USABLE (MMX); + CHECK_CPU_FEATURE_USABLE (FXSR); + CHECK_CPU_FEATURE_USABLE (SSE); + CHECK_CPU_FEATURE_USABLE (SSE2); + CHECK_CPU_FEATURE_USABLE (SS); + CHECK_CPU_FEATURE_USABLE (HTT); + CHECK_CPU_FEATURE_USABLE (TM); + CHECK_CPU_FEATURE_USABLE (PBE); + CHECK_CPU_FEATURE_USABLE (FSGSBASE); + CHECK_CPU_FEATURE_USABLE (TSC_ADJUST); + CHECK_CPU_FEATURE_USABLE (SGX); + CHECK_CPU_FEATURE_USABLE (BMI1); + CHECK_CPU_FEATURE_USABLE (HLE); CHECK_CPU_FEATURE_USABLE (AVX2); + CHECK_CPU_FEATURE_USABLE (SMEP); + CHECK_CPU_FEATURE_USABLE (BMI2); + CHECK_CPU_FEATURE_USABLE (ERMS); + CHECK_CPU_FEATURE_USABLE (INVPCID); + CHECK_CPU_FEATURE_USABLE (RTM); + CHECK_CPU_FEATURE_USABLE (PQM); + CHECK_CPU_FEATURE_USABLE (DEPR_FPU_CS_DS); + CHECK_CPU_FEATURE_USABLE (MPX); + CHECK_CPU_FEATURE_USABLE (PQE); CHECK_CPU_FEATURE_USABLE (AVX512F); CHECK_CPU_FEATURE_USABLE (AVX512DQ); + CHECK_CPU_FEATURE_USABLE (RDSEED); + CHECK_CPU_FEATURE_USABLE (ADX); + CHECK_CPU_FEATURE_USABLE (SMAP); CHECK_CPU_FEATURE_USABLE (AVX512_IFMA); + CHECK_CPU_FEATURE_USABLE (CLFLUSHOPT); + CHECK_CPU_FEATURE_USABLE (CLWB); + CHECK_CPU_FEATURE_USABLE (TRACE); CHECK_CPU_FEATURE_USABLE (AVX512PF); CHECK_CPU_FEATURE_USABLE (AVX512ER); CHECK_CPU_FEATURE_USABLE (AVX512CD); + CHECK_CPU_FEATURE_USABLE (SHA); CHECK_CPU_FEATURE_USABLE (AVX512BW); CHECK_CPU_FEATURE_USABLE (AVX512VL); + CHECK_CPU_FEATURE_USABLE (PREFETCHWT1); CHECK_CPU_FEATURE_USABLE (AVX512_VBMI); + CHECK_CPU_FEATURE_USABLE (UMIP); CHECK_CPU_FEATURE_USABLE (PKU); + CHECK_CPU_FEATURE_USABLE (OSPKE); + CHECK_CPU_FEATURE_USABLE (WAITPKG); CHECK_CPU_FEATURE_USABLE (AVX512_VBMI2); + CHECK_CPU_FEATURE_USABLE (SHSTK); + CHECK_CPU_FEATURE_USABLE (GFNI); CHECK_CPU_FEATURE_USABLE (VAES); CHECK_CPU_FEATURE_USABLE (VPCLMULQDQ); CHECK_CPU_FEATURE_USABLE (AVX512_VNNI); CHECK_CPU_FEATURE_USABLE (AVX512_BITALG); CHECK_CPU_FEATURE_USABLE (AVX512_VPOPCNTDQ); + CHECK_CPU_FEATURE_USABLE (RDPID); + CHECK_CPU_FEATURE_USABLE (CLDEMOTE); + CHECK_CPU_FEATURE_USABLE (MOVDIRI); + CHECK_CPU_FEATURE_USABLE (MOVDIR64B); + CHECK_CPU_FEATURE_USABLE (ENQCMD); + CHECK_CPU_FEATURE_USABLE (SGX_LC); + CHECK_CPU_FEATURE_USABLE (PKS); CHECK_CPU_FEATURE_USABLE (AVX512_4VNNIW); CHECK_CPU_FEATURE_USABLE (AVX512_4FMAPS); + CHECK_CPU_FEATURE_USABLE (FSRM); CHECK_CPU_FEATURE_USABLE (AVX512_VP2INTERSECT); + CHECK_CPU_FEATURE_USABLE (MD_CLEAR); + CHECK_CPU_FEATURE_USABLE (SERIALIZE); + CHECK_CPU_FEATURE_USABLE (HYBRID); + CHECK_CPU_FEATURE_USABLE (TSXLDTRK); + CHECK_CPU_FEATURE_USABLE (PCONFIG); + CHECK_CPU_FEATURE_USABLE (IBT); CHECK_CPU_FEATURE_USABLE (AMX_BF16); CHECK_CPU_FEATURE_USABLE (AMX_TILE); CHECK_CPU_FEATURE_USABLE (AMX_INT8); + CHECK_CPU_FEATURE_USABLE (IBRS_IBPB); + CHECK_CPU_FEATURE_USABLE (STIBP); + CHECK_CPU_FEATURE_USABLE (L1D_FLUSH); + CHECK_CPU_FEATURE_USABLE (ARCH_CAPABILITIES); + CHECK_CPU_FEATURE_USABLE (CORE_CAPABILITIES); + CHECK_CPU_FEATURE_USABLE (SSBD); + CHECK_CPU_FEATURE_USABLE (LAHF64_SAHF64); + CHECK_CPU_FEATURE_USABLE (SVM); + CHECK_CPU_FEATURE_USABLE (LZCNT); + CHECK_CPU_FEATURE_USABLE (SSE4A); + CHECK_CPU_FEATURE_USABLE (PREFETCHW); CHECK_CPU_FEATURE_USABLE (XOP); + CHECK_CPU_FEATURE_USABLE (LWP); CHECK_CPU_FEATURE_USABLE (FMA4); + CHECK_CPU_FEATURE_USABLE (TBM); + CHECK_CPU_FEATURE_USABLE (SYSCALL_SYSRET); + CHECK_CPU_FEATURE_USABLE (NX); + CHECK_CPU_FEATURE_USABLE (PAGE1GB); + CHECK_CPU_FEATURE_USABLE (RDTSCP); + CHECK_CPU_FEATURE_USABLE (LM); + CHECK_CPU_FEATURE_USABLE (XSAVEOPT); CHECK_CPU_FEATURE_USABLE (XSAVEC); + CHECK_CPU_FEATURE_USABLE (XGETBV_ECX_1); + CHECK_CPU_FEATURE_USABLE (XSAVES); + CHECK_CPU_FEATURE_USABLE (XFD); + CHECK_CPU_FEATURE_USABLE (INVARIANT_TSC); + CHECK_CPU_FEATURE_USABLE (WBNOINVD); CHECK_CPU_FEATURE_USABLE (AVX512_BF16); return 0; diff --git a/sysdeps/x86_64/Makefile b/sysdeps/x86_64/Makefile index e3bb45d..42b97c5 100644 --- a/sysdeps/x86_64/Makefile +++ b/sysdeps/x86_64/Makefile @@ -57,7 +57,7 @@ modules-names += x86_64/tst-x86_64mod-1 LDFLAGS-tst-x86_64mod-1.so = -Wl,-soname,tst-x86_64mod-1.so ifneq (no,$(have-tunables)) # Test the state size for XSAVE when XSAVEC is disabled. -tst-x86_64-1-ENV = GLIBC_TUNABLES=glibc.cpu.hwcaps=-XSAVEC_Usable +tst-x86_64-1-ENV = GLIBC_TUNABLES=glibc.cpu.hwcaps=-XSAVEC endif $(objpfx)tst-x86_64-1: $(objpfx)x86_64/tst-x86_64mod-1.so @@ -71,10 +71,10 @@ CFLAGS-tst-platformmod-2.c = -mno-avx LDFLAGS-tst-platformmod-2.so = -Wl,-soname,tst-platformmod-2.so $(objpfx)tst-platform-1: $(objpfx)tst-platformmod-1.so $(objpfx)tst-platform-1.out: $(objpfx)x86_64/tst-platformmod-2.so -# Turn off AVX512F_Usable and AVX2_Usable so that GLRO(dl_platform) is +# Turn off AVX512F and AVX2 so that GLRO(dl_platform) is # always set to x86_64. tst-platform-1-ENV = LD_PRELOAD=$(objpfx)\$$PLATFORM/tst-platformmod-2.so \ - GLIBC_TUNABLES=glibc.cpu.hwcaps=-AVX512F_Usable,-AVX2_Usable + GLIBC_TUNABLES=glibc.cpu.hwcaps=-AVX512F,-AVX2 endif tests += tst-audit3 tst-audit4 tst-audit5 tst-audit6 tst-audit7 \ diff --git a/sysdeps/x86_64/dl-machine.h b/sysdeps/x86_64/dl-machine.h index 8e9baff..ca73d8f 100644 --- a/sysdeps/x86_64/dl-machine.h +++ b/sysdeps/x86_64/dl-machine.h @@ -99,9 +99,9 @@ elf_machine_runtime_setup (struct link_map *l, int lazy, int profile) end in this function. */ if (__glibc_unlikely (profile)) { - if (HAS_ARCH_FEATURE (AVX512F_Usable)) + if (CPU_FEATURE_USABLE (AVX512F)) *(ElfW(Addr) *) (got + 2) = (ElfW(Addr)) &_dl_runtime_profile_avx512; - else if (HAS_ARCH_FEATURE (AVX_Usable)) + else if (CPU_FEATURE_USABLE (AVX)) *(ElfW(Addr) *) (got + 2) = (ElfW(Addr)) &_dl_runtime_profile_avx; else *(ElfW(Addr) *) (got + 2) = (ElfW(Addr)) &_dl_runtime_profile_sse; @@ -119,7 +119,7 @@ elf_machine_runtime_setup (struct link_map *l, int lazy, int profile) the resolved address. */ if (GLRO(dl_x86_cpu_features).xsave_state_size != 0) *(ElfW(Addr) *) (got + 2) - = (HAS_ARCH_FEATURE (XSAVEC_Usable) + = (CPU_FEATURE_USABLE (XSAVEC) ? (ElfW(Addr)) &_dl_runtime_resolve_xsavec : (ElfW(Addr)) &_dl_runtime_resolve_xsave); else diff --git a/sysdeps/x86_64/fpu/math-tests-arch.h b/sysdeps/x86_64/fpu/math-tests-arch.h index 435ddad..33ea763 100644 --- a/sysdeps/x86_64/fpu/math-tests-arch.h +++ b/sysdeps/x86_64/fpu/math-tests-arch.h @@ -24,7 +24,7 @@ # define CHECK_ARCH_EXT \ do \ { \ - if (!HAS_ARCH_FEATURE (AVX_Usable)) return; \ + if (!CPU_FEATURE_USABLE (AVX)) return; \ } \ while (0) @@ -34,7 +34,7 @@ # define CHECK_ARCH_EXT \ do \ { \ - if (!HAS_ARCH_FEATURE (AVX2_Usable)) return; \ + if (!CPU_FEATURE_USABLE (AVX2)) return; \ } \ while (0) @@ -44,7 +44,7 @@ # define CHECK_ARCH_EXT \ do \ { \ - if (!HAS_ARCH_FEATURE (AVX512F_Usable)) return; \ + if (!CPU_FEATURE_USABLE (AVX512F)) return; \ } \ while (0) diff --git a/sysdeps/x86_64/fpu/multiarch/ifunc-avx-fma4.h b/sysdeps/x86_64/fpu/multiarch/ifunc-avx-fma4.h index 86835ee..95fe2f4 100644 --- a/sysdeps/x86_64/fpu/multiarch/ifunc-avx-fma4.h +++ b/sysdeps/x86_64/fpu/multiarch/ifunc-avx-fma4.h @@ -29,14 +29,14 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_ARCH_P (cpu_features, FMA_Usable) - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)) + if (CPU_FEATURE_USABLE_P (cpu_features, FMA) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2)) return OPTIMIZE (fma); - if (CPU_FEATURES_ARCH_P (cpu_features, FMA4_Usable)) + if (CPU_FEATURE_USABLE_P (cpu_features, FMA4)) return OPTIMIZE (fma4); - if (CPU_FEATURES_ARCH_P (cpu_features, AVX_Usable)) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX)) return OPTIMIZE (avx); return OPTIMIZE (sse2); diff --git a/sysdeps/x86_64/fpu/multiarch/ifunc-fma.h b/sysdeps/x86_64/fpu/multiarch/ifunc-fma.h index 2242d97..0a25a44 100644 --- a/sysdeps/x86_64/fpu/multiarch/ifunc-fma.h +++ b/sysdeps/x86_64/fpu/multiarch/ifunc-fma.h @@ -26,8 +26,8 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_ARCH_P (cpu_features, FMA_Usable) - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)) + if (CPU_FEATURE_USABLE_P (cpu_features, FMA) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2)) return OPTIMIZE (fma); return OPTIMIZE (sse2); diff --git a/sysdeps/x86_64/fpu/multiarch/ifunc-fma4.h b/sysdeps/x86_64/fpu/multiarch/ifunc-fma4.h index 03adf86..7659758 100644 --- a/sysdeps/x86_64/fpu/multiarch/ifunc-fma4.h +++ b/sysdeps/x86_64/fpu/multiarch/ifunc-fma4.h @@ -28,11 +28,11 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_ARCH_P (cpu_features, FMA_Usable) - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)) + if (CPU_FEATURE_USABLE_P (cpu_features, FMA) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2)) return OPTIMIZE (fma); - if (CPU_FEATURES_ARCH_P (cpu_features, FMA4_Usable)) + if (CPU_FEATURE_USABLE_P (cpu_features, FMA)) return OPTIMIZE (fma4); return OPTIMIZE (sse2); diff --git a/sysdeps/x86_64/fpu/multiarch/ifunc-mathvec-avx2.h b/sysdeps/x86_64/fpu/multiarch/ifunc-mathvec-avx2.h index 9c5c6f1..2655e55 100644 --- a/sysdeps/x86_64/fpu/multiarch/ifunc-mathvec-avx2.h +++ b/sysdeps/x86_64/fpu/multiarch/ifunc-mathvec-avx2.h @@ -31,8 +31,8 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_ARCH_P (cpu_features, FMA_Usable) - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)) + if (CPU_FEATURE_USABLE_P (cpu_features, FMA) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2)) return OPTIMIZE (avx2); return OPTIMIZE (sse_wrapper); diff --git a/sysdeps/x86_64/fpu/multiarch/ifunc-mathvec-avx512.h b/sysdeps/x86_64/fpu/multiarch/ifunc-mathvec-avx512.h index 70e22c5..5f83265 100644 --- a/sysdeps/x86_64/fpu/multiarch/ifunc-mathvec-avx512.h +++ b/sysdeps/x86_64/fpu/multiarch/ifunc-mathvec-avx512.h @@ -34,10 +34,10 @@ IFUNC_SELECTOR (void) if (!CPU_FEATURES_ARCH_P (cpu_features, MathVec_Prefer_No_AVX512)) { - if (CPU_FEATURES_ARCH_P (cpu_features, AVX512DQ_Usable)) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX512DQ)) return OPTIMIZE (skx); - if (CPU_FEATURES_ARCH_P (cpu_features, AVX512F_Usable)) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX512F)) return OPTIMIZE (knl); } diff --git a/sysdeps/x86_64/fpu/multiarch/ifunc-mathvec-sse4_1.h b/sysdeps/x86_64/fpu/multiarch/ifunc-mathvec-sse4_1.h index 63005c0..7240e55 100644 --- a/sysdeps/x86_64/fpu/multiarch/ifunc-mathvec-sse4_1.h +++ b/sysdeps/x86_64/fpu/multiarch/ifunc-mathvec-sse4_1.h @@ -31,7 +31,7 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_CPU_P (cpu_features, SSE4_1)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE4_1)) return OPTIMIZE (sse4); return OPTIMIZE (sse2); diff --git a/sysdeps/x86_64/fpu/multiarch/ifunc-sse4_1.h b/sysdeps/x86_64/fpu/multiarch/ifunc-sse4_1.h index 7f26215..e5d8a6f 100644 --- a/sysdeps/x86_64/fpu/multiarch/ifunc-sse4_1.h +++ b/sysdeps/x86_64/fpu/multiarch/ifunc-sse4_1.h @@ -26,7 +26,7 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_CPU_P (cpu_features, SSE4_1)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE4_1)) return OPTIMIZE (sse41); return OPTIMIZE (c); diff --git a/sysdeps/x86_64/fpu/multiarch/s_fma.c b/sysdeps/x86_64/fpu/multiarch/s_fma.c index 9992a1e..0d8c0b7 100644 --- a/sysdeps/x86_64/fpu/multiarch/s_fma.c +++ b/sysdeps/x86_64/fpu/multiarch/s_fma.c @@ -41,8 +41,8 @@ __fma_fma4 (double x, double y, double z) } -libm_ifunc (__fma, HAS_ARCH_FEATURE (FMA_Usable) - ? __fma_fma3 : (HAS_ARCH_FEATURE (FMA4_Usable) +libm_ifunc (__fma, CPU_FEATURE_USABLE (FMA) + ? __fma_fma3 : (CPU_FEATURE_USABLE (FMA4) ? __fma_fma4 : __fma_sse2)); libm_alias_double (__fma, fma) diff --git a/sysdeps/x86_64/fpu/multiarch/s_fmaf.c b/sysdeps/x86_64/fpu/multiarch/s_fmaf.c index 4cbcf1f..c01e5a2 100644 --- a/sysdeps/x86_64/fpu/multiarch/s_fmaf.c +++ b/sysdeps/x86_64/fpu/multiarch/s_fmaf.c @@ -40,8 +40,8 @@ __fmaf_fma4 (float x, float y, float z) } -libm_ifunc (__fmaf, HAS_ARCH_FEATURE (FMA_Usable) - ? __fmaf_fma3 : (HAS_ARCH_FEATURE (FMA4_Usable) +libm_ifunc (__fmaf, CPU_FEATURE_USABLE (FMA) + ? __fmaf_fma3 : (CPU_FEATURE_USABLE (FMA4) ? __fmaf_fma4 : __fmaf_sse2)); libm_alias_float (__fma, fma) diff --git a/sysdeps/x86_64/multiarch/ifunc-avx2.h b/sysdeps/x86_64/multiarch/ifunc-avx2.h index 69f3039..f4e311d 100644 --- a/sysdeps/x86_64/multiarch/ifunc-avx2.h +++ b/sysdeps/x86_64/multiarch/ifunc-avx2.h @@ -28,7 +28,7 @@ IFUNC_SELECTOR (void) const struct cpu_features* cpu_features = __get_cpu_features (); if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER) - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2) && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load)) return OPTIMIZE (avx2); diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c index ce7eb1e..f93ec39 100644 --- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c +++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c @@ -41,19 +41,19 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/memchr.c. */ IFUNC_IMPL (i, name, memchr, IFUNC_IMPL_ADD (array, i, memchr, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __memchr_avx2) IFUNC_IMPL_ADD (array, i, memchr, 1, __memchr_sse2)) /* Support sysdeps/x86_64/multiarch/memcmp.c. */ IFUNC_IMPL (i, name, memcmp, IFUNC_IMPL_ADD (array, i, memcmp, - (HAS_ARCH_FEATURE (AVX2_Usable) - && HAS_CPU_FEATURE (MOVBE)), + (CPU_FEATURE_USABLE (AVX2) + && CPU_FEATURE_USABLE (MOVBE)), __memcmp_avx2_movbe) - IFUNC_IMPL_ADD (array, i, memcmp, HAS_CPU_FEATURE (SSE4_1), + IFUNC_IMPL_ADD (array, i, memcmp, CPU_FEATURE_USABLE (SSE4_1), __memcmp_sse4_1) - IFUNC_IMPL_ADD (array, i, memcmp, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, memcmp, CPU_FEATURE_USABLE (SSSE3), __memcmp_ssse3) IFUNC_IMPL_ADD (array, i, memcmp, 1, __memcmp_sse2)) @@ -61,25 +61,25 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/memmove_chk.c. */ IFUNC_IMPL (i, name, __memmove_chk, IFUNC_IMPL_ADD (array, i, __memmove_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memmove_chk_avx512_no_vzeroupper) IFUNC_IMPL_ADD (array, i, __memmove_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memmove_chk_avx512_unaligned) IFUNC_IMPL_ADD (array, i, __memmove_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memmove_chk_avx512_unaligned_erms) IFUNC_IMPL_ADD (array, i, __memmove_chk, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __memmove_chk_avx_unaligned) IFUNC_IMPL_ADD (array, i, __memmove_chk, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __memmove_chk_avx_unaligned_erms) IFUNC_IMPL_ADD (array, i, __memmove_chk, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __memmove_chk_ssse3_back) IFUNC_IMPL_ADD (array, i, __memmove_chk, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __memmove_chk_ssse3) IFUNC_IMPL_ADD (array, i, __memmove_chk, 1, __memmove_chk_sse2_unaligned) @@ -92,23 +92,23 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/memmove.c. */ IFUNC_IMPL (i, name, memmove, IFUNC_IMPL_ADD (array, i, memmove, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __memmove_avx_unaligned) IFUNC_IMPL_ADD (array, i, memmove, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __memmove_avx_unaligned_erms) IFUNC_IMPL_ADD (array, i, memmove, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memmove_avx512_no_vzeroupper) IFUNC_IMPL_ADD (array, i, memmove, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memmove_avx512_unaligned) IFUNC_IMPL_ADD (array, i, memmove, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memmove_avx512_unaligned_erms) - IFUNC_IMPL_ADD (array, i, memmove, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, memmove, CPU_FEATURE_USABLE (SSSE3), __memmove_ssse3_back) - IFUNC_IMPL_ADD (array, i, memmove, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, memmove, CPU_FEATURE_USABLE (SSSE3), __memmove_ssse3) IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_erms) IFUNC_IMPL_ADD (array, i, memmove, 1, @@ -119,7 +119,7 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/memrchr.c. */ IFUNC_IMPL (i, name, memrchr, IFUNC_IMPL_ADD (array, i, memrchr, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __memrchr_avx2) IFUNC_IMPL_ADD (array, i, memrchr, 1, __memrchr_sse2)) @@ -133,19 +133,19 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, IFUNC_IMPL_ADD (array, i, __memset_chk, 1, __memset_chk_sse2_unaligned_erms) IFUNC_IMPL_ADD (array, i, __memset_chk, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __memset_chk_avx2_unaligned) IFUNC_IMPL_ADD (array, i, __memset_chk, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __memset_chk_avx2_unaligned_erms) IFUNC_IMPL_ADD (array, i, __memset_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memset_chk_avx512_unaligned_erms) IFUNC_IMPL_ADD (array, i, __memset_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memset_chk_avx512_unaligned) IFUNC_IMPL_ADD (array, i, __memset_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memset_chk_avx512_no_vzeroupper) ) #endif @@ -158,48 +158,48 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, __memset_sse2_unaligned_erms) IFUNC_IMPL_ADD (array, i, memset, 1, __memset_erms) IFUNC_IMPL_ADD (array, i, memset, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __memset_avx2_unaligned) IFUNC_IMPL_ADD (array, i, memset, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __memset_avx2_unaligned_erms) IFUNC_IMPL_ADD (array, i, memset, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memset_avx512_unaligned_erms) IFUNC_IMPL_ADD (array, i, memset, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memset_avx512_unaligned) IFUNC_IMPL_ADD (array, i, memset, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memset_avx512_no_vzeroupper) ) /* Support sysdeps/x86_64/multiarch/rawmemchr.c. */ IFUNC_IMPL (i, name, rawmemchr, IFUNC_IMPL_ADD (array, i, rawmemchr, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __rawmemchr_avx2) IFUNC_IMPL_ADD (array, i, rawmemchr, 1, __rawmemchr_sse2)) /* Support sysdeps/x86_64/multiarch/strlen.c. */ IFUNC_IMPL (i, name, strlen, IFUNC_IMPL_ADD (array, i, strlen, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __strlen_avx2) IFUNC_IMPL_ADD (array, i, strlen, 1, __strlen_sse2)) /* Support sysdeps/x86_64/multiarch/strnlen.c. */ IFUNC_IMPL (i, name, strnlen, IFUNC_IMPL_ADD (array, i, strnlen, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __strnlen_avx2) IFUNC_IMPL_ADD (array, i, strnlen, 1, __strnlen_sse2)) /* Support sysdeps/x86_64/multiarch/stpncpy.c. */ IFUNC_IMPL (i, name, stpncpy, - IFUNC_IMPL_ADD (array, i, stpncpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, stpncpy, CPU_FEATURE_USABLE (SSSE3), __stpncpy_ssse3) - IFUNC_IMPL_ADD (array, i, stpncpy, HAS_ARCH_FEATURE (AVX2_Usable), + IFUNC_IMPL_ADD (array, i, stpncpy, CPU_FEATURE_USABLE (AVX2), __stpncpy_avx2) IFUNC_IMPL_ADD (array, i, stpncpy, 1, __stpncpy_sse2_unaligned) @@ -207,9 +207,9 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/stpcpy.c. */ IFUNC_IMPL (i, name, stpcpy, - IFUNC_IMPL_ADD (array, i, stpcpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, stpcpy, CPU_FEATURE_USABLE (SSSE3), __stpcpy_ssse3) - IFUNC_IMPL_ADD (array, i, stpcpy, HAS_ARCH_FEATURE (AVX2_Usable), + IFUNC_IMPL_ADD (array, i, stpcpy, CPU_FEATURE_USABLE (AVX2), __stpcpy_avx2) IFUNC_IMPL_ADD (array, i, stpcpy, 1, __stpcpy_sse2_unaligned) IFUNC_IMPL_ADD (array, i, stpcpy, 1, __stpcpy_sse2)) @@ -217,35 +217,35 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/strcasecmp_l.c. */ IFUNC_IMPL (i, name, strcasecmp, IFUNC_IMPL_ADD (array, i, strcasecmp, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __strcasecmp_avx) IFUNC_IMPL_ADD (array, i, strcasecmp, - HAS_CPU_FEATURE (SSE4_2), + CPU_FEATURE_USABLE (SSE4_2), __strcasecmp_sse42) IFUNC_IMPL_ADD (array, i, strcasecmp, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __strcasecmp_ssse3) IFUNC_IMPL_ADD (array, i, strcasecmp, 1, __strcasecmp_sse2)) /* Support sysdeps/x86_64/multiarch/strcasecmp_l.c. */ IFUNC_IMPL (i, name, strcasecmp_l, IFUNC_IMPL_ADD (array, i, strcasecmp_l, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __strcasecmp_l_avx) IFUNC_IMPL_ADD (array, i, strcasecmp_l, - HAS_CPU_FEATURE (SSE4_2), + CPU_FEATURE_USABLE (SSE4_2), __strcasecmp_l_sse42) IFUNC_IMPL_ADD (array, i, strcasecmp_l, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __strcasecmp_l_ssse3) IFUNC_IMPL_ADD (array, i, strcasecmp_l, 1, __strcasecmp_l_sse2)) /* Support sysdeps/x86_64/multiarch/strcat.c. */ IFUNC_IMPL (i, name, strcat, - IFUNC_IMPL_ADD (array, i, strcat, HAS_ARCH_FEATURE (AVX2_Usable), + IFUNC_IMPL_ADD (array, i, strcat, CPU_FEATURE_USABLE (AVX2), __strcat_avx2) - IFUNC_IMPL_ADD (array, i, strcat, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, strcat, CPU_FEATURE_USABLE (SSSE3), __strcat_ssse3) IFUNC_IMPL_ADD (array, i, strcat, 1, __strcat_sse2_unaligned) IFUNC_IMPL_ADD (array, i, strcat, 1, __strcat_sse2)) @@ -253,7 +253,7 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/strchr.c. */ IFUNC_IMPL (i, name, strchr, IFUNC_IMPL_ADD (array, i, strchr, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __strchr_avx2) IFUNC_IMPL_ADD (array, i, strchr, 1, __strchr_sse2_no_bsf) IFUNC_IMPL_ADD (array, i, strchr, 1, __strchr_sse2)) @@ -261,54 +261,54 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/strchrnul.c. */ IFUNC_IMPL (i, name, strchrnul, IFUNC_IMPL_ADD (array, i, strchrnul, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __strchrnul_avx2) IFUNC_IMPL_ADD (array, i, strchrnul, 1, __strchrnul_sse2)) /* Support sysdeps/x86_64/multiarch/strrchr.c. */ IFUNC_IMPL (i, name, strrchr, IFUNC_IMPL_ADD (array, i, strrchr, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __strrchr_avx2) IFUNC_IMPL_ADD (array, i, strrchr, 1, __strrchr_sse2)) /* Support sysdeps/x86_64/multiarch/strcmp.c. */ IFUNC_IMPL (i, name, strcmp, IFUNC_IMPL_ADD (array, i, strcmp, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __strcmp_avx2) - IFUNC_IMPL_ADD (array, i, strcmp, HAS_CPU_FEATURE (SSE4_2), + IFUNC_IMPL_ADD (array, i, strcmp, CPU_FEATURE_USABLE (SSE4_2), __strcmp_sse42) - IFUNC_IMPL_ADD (array, i, strcmp, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, strcmp, CPU_FEATURE_USABLE (SSSE3), __strcmp_ssse3) IFUNC_IMPL_ADD (array, i, strcmp, 1, __strcmp_sse2_unaligned) IFUNC_IMPL_ADD (array, i, strcmp, 1, __strcmp_sse2)) /* Support sysdeps/x86_64/multiarch/strcpy.c. */ IFUNC_IMPL (i, name, strcpy, - IFUNC_IMPL_ADD (array, i, strcpy, HAS_ARCH_FEATURE (AVX2_Usable), + IFUNC_IMPL_ADD (array, i, strcpy, CPU_FEATURE_USABLE (AVX2), __strcpy_avx2) - IFUNC_IMPL_ADD (array, i, strcpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, strcpy, CPU_FEATURE_USABLE (SSSE3), __strcpy_ssse3) IFUNC_IMPL_ADD (array, i, strcpy, 1, __strcpy_sse2_unaligned) IFUNC_IMPL_ADD (array, i, strcpy, 1, __strcpy_sse2)) /* Support sysdeps/x86_64/multiarch/strcspn.c. */ IFUNC_IMPL (i, name, strcspn, - IFUNC_IMPL_ADD (array, i, strcspn, HAS_CPU_FEATURE (SSE4_2), + IFUNC_IMPL_ADD (array, i, strcspn, CPU_FEATURE_USABLE (SSE4_2), __strcspn_sse42) IFUNC_IMPL_ADD (array, i, strcspn, 1, __strcspn_sse2)) /* Support sysdeps/x86_64/multiarch/strncase_l.c. */ IFUNC_IMPL (i, name, strncasecmp, IFUNC_IMPL_ADD (array, i, strncasecmp, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __strncasecmp_avx) IFUNC_IMPL_ADD (array, i, strncasecmp, - HAS_CPU_FEATURE (SSE4_2), + CPU_FEATURE_USABLE (SSE4_2), __strncasecmp_sse42) IFUNC_IMPL_ADD (array, i, strncasecmp, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __strncasecmp_ssse3) IFUNC_IMPL_ADD (array, i, strncasecmp, 1, __strncasecmp_sse2)) @@ -316,22 +316,22 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/strncase_l.c. */ IFUNC_IMPL (i, name, strncasecmp_l, IFUNC_IMPL_ADD (array, i, strncasecmp_l, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __strncasecmp_l_avx) IFUNC_IMPL_ADD (array, i, strncasecmp_l, - HAS_CPU_FEATURE (SSE4_2), + CPU_FEATURE_USABLE (SSE4_2), __strncasecmp_l_sse42) IFUNC_IMPL_ADD (array, i, strncasecmp_l, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __strncasecmp_l_ssse3) IFUNC_IMPL_ADD (array, i, strncasecmp_l, 1, __strncasecmp_l_sse2)) /* Support sysdeps/x86_64/multiarch/strncat.c. */ IFUNC_IMPL (i, name, strncat, - IFUNC_IMPL_ADD (array, i, strncat, HAS_ARCH_FEATURE (AVX2_Usable), + IFUNC_IMPL_ADD (array, i, strncat, CPU_FEATURE_USABLE (AVX2), __strncat_avx2) - IFUNC_IMPL_ADD (array, i, strncat, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, strncat, CPU_FEATURE_USABLE (SSSE3), __strncat_ssse3) IFUNC_IMPL_ADD (array, i, strncat, 1, __strncat_sse2_unaligned) @@ -339,9 +339,9 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/strncpy.c. */ IFUNC_IMPL (i, name, strncpy, - IFUNC_IMPL_ADD (array, i, strncpy, HAS_ARCH_FEATURE (AVX2_Usable), + IFUNC_IMPL_ADD (array, i, strncpy, CPU_FEATURE_USABLE (AVX2), __strncpy_avx2) - IFUNC_IMPL_ADD (array, i, strncpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, strncpy, CPU_FEATURE_USABLE (SSSE3), __strncpy_ssse3) IFUNC_IMPL_ADD (array, i, strncpy, 1, __strncpy_sse2_unaligned) @@ -349,14 +349,14 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/strpbrk.c. */ IFUNC_IMPL (i, name, strpbrk, - IFUNC_IMPL_ADD (array, i, strpbrk, HAS_CPU_FEATURE (SSE4_2), + IFUNC_IMPL_ADD (array, i, strpbrk, CPU_FEATURE_USABLE (SSE4_2), __strpbrk_sse42) IFUNC_IMPL_ADD (array, i, strpbrk, 1, __strpbrk_sse2)) /* Support sysdeps/x86_64/multiarch/strspn.c. */ IFUNC_IMPL (i, name, strspn, - IFUNC_IMPL_ADD (array, i, strspn, HAS_CPU_FEATURE (SSE4_2), + IFUNC_IMPL_ADD (array, i, strspn, CPU_FEATURE_USABLE (SSE4_2), __strspn_sse42) IFUNC_IMPL_ADD (array, i, strspn, 1, __strspn_sse2)) @@ -368,70 +368,70 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/wcschr.c. */ IFUNC_IMPL (i, name, wcschr, IFUNC_IMPL_ADD (array, i, wcschr, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __wcschr_avx2) IFUNC_IMPL_ADD (array, i, wcschr, 1, __wcschr_sse2)) /* Support sysdeps/x86_64/multiarch/wcsrchr.c. */ IFUNC_IMPL (i, name, wcsrchr, IFUNC_IMPL_ADD (array, i, wcsrchr, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __wcsrchr_avx2) IFUNC_IMPL_ADD (array, i, wcsrchr, 1, __wcsrchr_sse2)) /* Support sysdeps/x86_64/multiarch/wcscmp.c. */ IFUNC_IMPL (i, name, wcscmp, IFUNC_IMPL_ADD (array, i, wcscmp, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __wcscmp_avx2) IFUNC_IMPL_ADD (array, i, wcscmp, 1, __wcscmp_sse2)) /* Support sysdeps/x86_64/multiarch/wcsncmp.c. */ IFUNC_IMPL (i, name, wcsncmp, IFUNC_IMPL_ADD (array, i, wcsncmp, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __wcsncmp_avx2) IFUNC_IMPL_ADD (array, i, wcsncmp, 1, __wcsncmp_sse2)) /* Support sysdeps/x86_64/multiarch/wcscpy.c. */ IFUNC_IMPL (i, name, wcscpy, - IFUNC_IMPL_ADD (array, i, wcscpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, wcscpy, CPU_FEATURE_USABLE (SSSE3), __wcscpy_ssse3) IFUNC_IMPL_ADD (array, i, wcscpy, 1, __wcscpy_sse2)) /* Support sysdeps/x86_64/multiarch/wcslen.c. */ IFUNC_IMPL (i, name, wcslen, IFUNC_IMPL_ADD (array, i, wcslen, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __wcslen_avx2) IFUNC_IMPL_ADD (array, i, wcslen, 1, __wcslen_sse2)) /* Support sysdeps/x86_64/multiarch/wcsnlen.c. */ IFUNC_IMPL (i, name, wcsnlen, IFUNC_IMPL_ADD (array, i, wcsnlen, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __wcsnlen_avx2) IFUNC_IMPL_ADD (array, i, wcsnlen, - HAS_CPU_FEATURE (SSE4_1), + CPU_FEATURE_USABLE (SSE4_1), __wcsnlen_sse4_1) IFUNC_IMPL_ADD (array, i, wcsnlen, 1, __wcsnlen_sse2)) /* Support sysdeps/x86_64/multiarch/wmemchr.c. */ IFUNC_IMPL (i, name, wmemchr, IFUNC_IMPL_ADD (array, i, wmemchr, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __wmemchr_avx2) IFUNC_IMPL_ADD (array, i, wmemchr, 1, __wmemchr_sse2)) /* Support sysdeps/x86_64/multiarch/wmemcmp.c. */ IFUNC_IMPL (i, name, wmemcmp, IFUNC_IMPL_ADD (array, i, wmemcmp, - (HAS_ARCH_FEATURE (AVX2_Usable) - && HAS_CPU_FEATURE (MOVBE)), + (CPU_FEATURE_USABLE (AVX2) + && CPU_FEATURE_USABLE (MOVBE)), __wmemcmp_avx2_movbe) - IFUNC_IMPL_ADD (array, i, wmemcmp, HAS_CPU_FEATURE (SSE4_1), + IFUNC_IMPL_ADD (array, i, wmemcmp, CPU_FEATURE_USABLE (SSE4_1), __wmemcmp_sse4_1) - IFUNC_IMPL_ADD (array, i, wmemcmp, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, wmemcmp, CPU_FEATURE_USABLE (SSSE3), __wmemcmp_ssse3) IFUNC_IMPL_ADD (array, i, wmemcmp, 1, __wmemcmp_sse2)) @@ -440,35 +440,35 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, IFUNC_IMPL_ADD (array, i, wmemset, 1, __wmemset_sse2_unaligned) IFUNC_IMPL_ADD (array, i, wmemset, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __wmemset_avx2_unaligned) IFUNC_IMPL_ADD (array, i, wmemset, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __wmemset_avx512_unaligned)) #ifdef SHARED /* Support sysdeps/x86_64/multiarch/memcpy_chk.c. */ IFUNC_IMPL (i, name, __memcpy_chk, IFUNC_IMPL_ADD (array, i, __memcpy_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memcpy_chk_avx512_no_vzeroupper) IFUNC_IMPL_ADD (array, i, __memcpy_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memcpy_chk_avx512_unaligned) IFUNC_IMPL_ADD (array, i, __memcpy_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memcpy_chk_avx512_unaligned_erms) IFUNC_IMPL_ADD (array, i, __memcpy_chk, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __memcpy_chk_avx_unaligned) IFUNC_IMPL_ADD (array, i, __memcpy_chk, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __memcpy_chk_avx_unaligned_erms) IFUNC_IMPL_ADD (array, i, __memcpy_chk, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __memcpy_chk_ssse3_back) IFUNC_IMPL_ADD (array, i, __memcpy_chk, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __memcpy_chk_ssse3) IFUNC_IMPL_ADD (array, i, __memcpy_chk, 1, __memcpy_chk_sse2_unaligned) @@ -481,23 +481,23 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/memcpy.c. */ IFUNC_IMPL (i, name, memcpy, IFUNC_IMPL_ADD (array, i, memcpy, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __memcpy_avx_unaligned) IFUNC_IMPL_ADD (array, i, memcpy, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __memcpy_avx_unaligned_erms) - IFUNC_IMPL_ADD (array, i, memcpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, memcpy, CPU_FEATURE_USABLE (SSSE3), __memcpy_ssse3_back) - IFUNC_IMPL_ADD (array, i, memcpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, memcpy, CPU_FEATURE_USABLE (SSSE3), __memcpy_ssse3) IFUNC_IMPL_ADD (array, i, memcpy, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memcpy_avx512_no_vzeroupper) IFUNC_IMPL_ADD (array, i, memcpy, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memcpy_avx512_unaligned) IFUNC_IMPL_ADD (array, i, memcpy, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __memcpy_avx512_unaligned_erms) IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_sse2_unaligned) IFUNC_IMPL_ADD (array, i, memcpy, 1, @@ -508,25 +508,25 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/mempcpy_chk.c. */ IFUNC_IMPL (i, name, __mempcpy_chk, IFUNC_IMPL_ADD (array, i, __mempcpy_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __mempcpy_chk_avx512_no_vzeroupper) IFUNC_IMPL_ADD (array, i, __mempcpy_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __mempcpy_chk_avx512_unaligned) IFUNC_IMPL_ADD (array, i, __mempcpy_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __mempcpy_chk_avx512_unaligned_erms) IFUNC_IMPL_ADD (array, i, __mempcpy_chk, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __mempcpy_chk_avx_unaligned) IFUNC_IMPL_ADD (array, i, __mempcpy_chk, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __mempcpy_chk_avx_unaligned_erms) IFUNC_IMPL_ADD (array, i, __mempcpy_chk, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __mempcpy_chk_ssse3_back) IFUNC_IMPL_ADD (array, i, __mempcpy_chk, - HAS_CPU_FEATURE (SSSE3), + CPU_FEATURE_USABLE (SSSE3), __mempcpy_chk_ssse3) IFUNC_IMPL_ADD (array, i, __mempcpy_chk, 1, __mempcpy_chk_sse2_unaligned) @@ -539,23 +539,23 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/mempcpy.c. */ IFUNC_IMPL (i, name, mempcpy, IFUNC_IMPL_ADD (array, i, mempcpy, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __mempcpy_avx512_no_vzeroupper) IFUNC_IMPL_ADD (array, i, mempcpy, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __mempcpy_avx512_unaligned) IFUNC_IMPL_ADD (array, i, mempcpy, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __mempcpy_avx512_unaligned_erms) IFUNC_IMPL_ADD (array, i, mempcpy, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __mempcpy_avx_unaligned) IFUNC_IMPL_ADD (array, i, mempcpy, - HAS_ARCH_FEATURE (AVX_Usable), + CPU_FEATURE_USABLE (AVX), __mempcpy_avx_unaligned_erms) - IFUNC_IMPL_ADD (array, i, mempcpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, mempcpy, CPU_FEATURE_USABLE (SSSE3), __mempcpy_ssse3_back) - IFUNC_IMPL_ADD (array, i, mempcpy, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, mempcpy, CPU_FEATURE_USABLE (SSSE3), __mempcpy_ssse3) IFUNC_IMPL_ADD (array, i, mempcpy, 1, __mempcpy_sse2_unaligned) @@ -566,11 +566,11 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/x86_64/multiarch/strncmp.c. */ IFUNC_IMPL (i, name, strncmp, IFUNC_IMPL_ADD (array, i, strncmp, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __strncmp_avx2) - IFUNC_IMPL_ADD (array, i, strncmp, HAS_CPU_FEATURE (SSE4_2), + IFUNC_IMPL_ADD (array, i, strncmp, CPU_FEATURE_USABLE (SSE4_2), __strncmp_sse42) - IFUNC_IMPL_ADD (array, i, strncmp, HAS_CPU_FEATURE (SSSE3), + IFUNC_IMPL_ADD (array, i, strncmp, CPU_FEATURE_USABLE (SSSE3), __strncmp_ssse3) IFUNC_IMPL_ADD (array, i, strncmp, 1, __strncmp_sse2)) @@ -580,10 +580,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, IFUNC_IMPL_ADD (array, i, __wmemset_chk, 1, __wmemset_chk_sse2_unaligned) IFUNC_IMPL_ADD (array, i, __wmemset_chk, - HAS_ARCH_FEATURE (AVX2_Usable), + CPU_FEATURE_USABLE (AVX2), __wmemset_chk_avx2_unaligned) IFUNC_IMPL_ADD (array, i, __wmemset_chk, - HAS_ARCH_FEATURE (AVX512F_Usable), + CPU_FEATURE_USABLE (AVX512F), __wmemset_chk_avx512_unaligned)) #endif diff --git a/sysdeps/x86_64/multiarch/ifunc-memcmp.h b/sysdeps/x86_64/multiarch/ifunc-memcmp.h index c14db39..0e21b3a 100644 --- a/sysdeps/x86_64/multiarch/ifunc-memcmp.h +++ b/sysdeps/x86_64/multiarch/ifunc-memcmp.h @@ -30,15 +30,15 @@ IFUNC_SELECTOR (void) const struct cpu_features* cpu_features = __get_cpu_features (); if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER) - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable) - && CPU_FEATURES_CPU_P (cpu_features, MOVBE) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2) + && CPU_FEATURE_USABLE_P (cpu_features, MOVBE) && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load)) return OPTIMIZE (avx2_movbe); - if (CPU_FEATURES_CPU_P (cpu_features, SSE4_1)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE4_1)) return OPTIMIZE (sse4_1); - if (CPU_FEATURES_CPU_P (cpu_features, SSSE3)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSSE3)) return OPTIMIZE (ssse3); return OPTIMIZE (sse2); diff --git a/sysdeps/x86_64/multiarch/ifunc-memmove.h b/sysdeps/x86_64/multiarch/ifunc-memmove.h index 81673d2..9ada03a 100644 --- a/sysdeps/x86_64/multiarch/ifunc-memmove.h +++ b/sysdeps/x86_64/multiarch/ifunc-memmove.h @@ -45,13 +45,13 @@ IFUNC_SELECTOR (void) || CPU_FEATURES_ARCH_P (cpu_features, Prefer_FSRM)) return OPTIMIZE (erms); - if (CPU_FEATURES_ARCH_P (cpu_features, AVX512F_Usable) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX512F) && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512)) { if (CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER)) return OPTIMIZE (avx512_no_vzeroupper); - if (CPU_FEATURES_CPU_P (cpu_features, ERMS)) + if (CPU_FEATURE_USABLE_P (cpu_features, ERMS)) return OPTIMIZE (avx512_unaligned_erms); return OPTIMIZE (avx512_unaligned); @@ -59,16 +59,16 @@ IFUNC_SELECTOR (void) if (CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load)) { - if (CPU_FEATURES_CPU_P (cpu_features, ERMS)) + if (CPU_FEATURE_USABLE_P (cpu_features, ERMS)) return OPTIMIZE (avx_unaligned_erms); return OPTIMIZE (avx_unaligned); } - if (!CPU_FEATURES_CPU_P (cpu_features, SSSE3) + if (!CPU_FEATURE_USABLE_P (cpu_features, SSSE3) || CPU_FEATURES_ARCH_P (cpu_features, Fast_Unaligned_Copy)) { - if (CPU_FEATURES_CPU_P (cpu_features, ERMS)) + if (CPU_FEATURE_USABLE_P (cpu_features, ERMS)) return OPTIMIZE (sse2_unaligned_erms); return OPTIMIZE (sse2_unaligned); diff --git a/sysdeps/x86_64/multiarch/ifunc-memset.h b/sysdeps/x86_64/multiarch/ifunc-memset.h index d690293..f52613d 100644 --- a/sysdeps/x86_64/multiarch/ifunc-memset.h +++ b/sysdeps/x86_64/multiarch/ifunc-memset.h @@ -42,27 +42,27 @@ IFUNC_SELECTOR (void) if (CPU_FEATURES_ARCH_P (cpu_features, Prefer_ERMS)) return OPTIMIZE (erms); - if (CPU_FEATURES_ARCH_P (cpu_features, AVX512F_Usable) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX512F) && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512)) { if (CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER)) return OPTIMIZE (avx512_no_vzeroupper); - if (CPU_FEATURES_CPU_P (cpu_features, ERMS)) + if (CPU_FEATURE_USABLE_P (cpu_features, ERMS)) return OPTIMIZE (avx512_unaligned_erms); return OPTIMIZE (avx512_unaligned); } - if (CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable)) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)) { - if (CPU_FEATURES_CPU_P (cpu_features, ERMS)) + if (CPU_FEATURE_USABLE_P (cpu_features, ERMS)) return OPTIMIZE (avx2_unaligned_erms); else return OPTIMIZE (avx2_unaligned); } - if (CPU_FEATURES_CPU_P (cpu_features, ERMS)) + if (CPU_FEATURE_USABLE_P (cpu_features, ERMS)) return OPTIMIZE (sse2_unaligned_erms); return OPTIMIZE (sse2_unaligned); diff --git a/sysdeps/x86_64/multiarch/ifunc-sse4_2.h b/sysdeps/x86_64/multiarch/ifunc-sse4_2.h index 082179c..cbf1838 100644 --- a/sysdeps/x86_64/multiarch/ifunc-sse4_2.h +++ b/sysdeps/x86_64/multiarch/ifunc-sse4_2.h @@ -27,7 +27,7 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_CPU_P (cpu_features, SSE4_2)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE4_2)) return OPTIMIZE (sse42); return OPTIMIZE (sse2); diff --git a/sysdeps/x86_64/multiarch/ifunc-strcasecmp.h b/sysdeps/x86_64/multiarch/ifunc-strcasecmp.h index f349ee7..0818333 100644 --- a/sysdeps/x86_64/multiarch/ifunc-strcasecmp.h +++ b/sysdeps/x86_64/multiarch/ifunc-strcasecmp.h @@ -29,14 +29,14 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_ARCH_P (cpu_features, AVX_Usable)) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX)) return OPTIMIZE (avx); - if (CPU_FEATURES_CPU_P (cpu_features, SSE4_2) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE4_2) && !CPU_FEATURES_ARCH_P (cpu_features, Slow_SSE4_2)) return OPTIMIZE (sse42); - if (CPU_FEATURES_CPU_P (cpu_features, SSSE3)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSSE3)) return OPTIMIZE (ssse3); return OPTIMIZE (sse2); diff --git a/sysdeps/x86_64/multiarch/ifunc-strcpy.h b/sysdeps/x86_64/multiarch/ifunc-strcpy.h index ae4f451..63b0dc0 100644 --- a/sysdeps/x86_64/multiarch/ifunc-strcpy.h +++ b/sysdeps/x86_64/multiarch/ifunc-strcpy.h @@ -32,14 +32,14 @@ IFUNC_SELECTOR (void) const struct cpu_features* cpu_features = __get_cpu_features (); if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER) - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2) && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load)) return OPTIMIZE (avx2); if (CPU_FEATURES_ARCH_P (cpu_features, Fast_Unaligned_Load)) return OPTIMIZE (sse2_unaligned); - if (CPU_FEATURES_CPU_P (cpu_features, SSSE3)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSSE3)) return OPTIMIZE (ssse3); return OPTIMIZE (sse2); diff --git a/sysdeps/x86_64/multiarch/ifunc-wmemset.h b/sysdeps/x86_64/multiarch/ifunc-wmemset.h index 583f631..8cfce56 100644 --- a/sysdeps/x86_64/multiarch/ifunc-wmemset.h +++ b/sysdeps/x86_64/multiarch/ifunc-wmemset.h @@ -28,10 +28,10 @@ IFUNC_SELECTOR (void) const struct cpu_features* cpu_features = __get_cpu_features (); if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER) - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2) && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load)) { - if (CPU_FEATURES_ARCH_P (cpu_features, AVX512F_Usable) + if (CPU_FEATURE_USABLE_P (cpu_features, AVX512F) && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512)) return OPTIMIZE (avx512_unaligned); else diff --git a/sysdeps/x86_64/multiarch/sched_cpucount.c b/sysdeps/x86_64/multiarch/sched_cpucount.c index 686fe07..074c663 100644 --- a/sysdeps/x86_64/multiarch/sched_cpucount.c +++ b/sysdeps/x86_64/multiarch/sched_cpucount.c @@ -33,4 +33,4 @@ #undef __sched_cpucount libc_ifunc (__sched_cpucount, - HAS_CPU_FEATURE (POPCNT) ? popcount_cpucount : generic_cpucount); + CPU_FEATURE_USABLE (POPCNT) ? popcount_cpucount : generic_cpucount); diff --git a/sysdeps/x86_64/multiarch/strchr.c b/sysdeps/x86_64/multiarch/strchr.c index f27980d..8df4609 100644 --- a/sysdeps/x86_64/multiarch/strchr.c +++ b/sysdeps/x86_64/multiarch/strchr.c @@ -36,7 +36,7 @@ IFUNC_SELECTOR (void) const struct cpu_features* cpu_features = __get_cpu_features (); if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER) - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2) && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load)) return OPTIMIZE (avx2); diff --git a/sysdeps/x86_64/multiarch/strcmp.c b/sysdeps/x86_64/multiarch/strcmp.c index 4db7332..16ae72a 100644 --- a/sysdeps/x86_64/multiarch/strcmp.c +++ b/sysdeps/x86_64/multiarch/strcmp.c @@ -37,14 +37,14 @@ IFUNC_SELECTOR (void) const struct cpu_features* cpu_features = __get_cpu_features (); if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER) - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2) && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load)) return OPTIMIZE (avx2); if (CPU_FEATURES_ARCH_P (cpu_features, Fast_Unaligned_Load)) return OPTIMIZE (sse2_unaligned); - if (CPU_FEATURES_CPU_P (cpu_features, SSSE3)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSSE3)) return OPTIMIZE (ssse3); return OPTIMIZE (sse2); diff --git a/sysdeps/x86_64/multiarch/strncmp.c b/sysdeps/x86_64/multiarch/strncmp.c index 6b63b0a..3c94b3f 100644 --- a/sysdeps/x86_64/multiarch/strncmp.c +++ b/sysdeps/x86_64/multiarch/strncmp.c @@ -37,15 +37,15 @@ IFUNC_SELECTOR (void) const struct cpu_features* cpu_features = __get_cpu_features (); if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER) - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2) && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load)) return OPTIMIZE (avx2); - if (CPU_FEATURES_CPU_P (cpu_features, SSE4_2) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE4_2) && !CPU_FEATURES_ARCH_P (cpu_features, Slow_SSE4_2)) return OPTIMIZE (sse42); - if (CPU_FEATURES_CPU_P (cpu_features, SSSE3)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSSE3)) return OPTIMIZE (ssse3); return OPTIMIZE (sse2); diff --git a/sysdeps/x86_64/multiarch/test-multiarch.c b/sysdeps/x86_64/multiarch/test-multiarch.c index 317373c..2782803 100644 --- a/sysdeps/x86_64/multiarch/test-multiarch.c +++ b/sysdeps/x86_64/multiarch/test-multiarch.c @@ -75,18 +75,18 @@ do_test (int argc, char **argv) int fails; get_cpuinfo (); - fails = check_proc ("avx", HAS_ARCH_FEATURE (AVX_Usable), - "HAS_ARCH_FEATURE (AVX_Usable)"); - fails += check_proc ("fma4", HAS_ARCH_FEATURE (FMA4_Usable), - "HAS_ARCH_FEATURE (FMA4_Usable)"); - fails += check_proc ("sse4_2", HAS_CPU_FEATURE (SSE4_2), - "HAS_CPU_FEATURE (SSE4_2)"); - fails += check_proc ("sse4_1", HAS_CPU_FEATURE (SSE4_1) - , "HAS_CPU_FEATURE (SSE4_1)"); - fails += check_proc ("ssse3", HAS_CPU_FEATURE (SSSE3), - "HAS_CPU_FEATURE (SSSE3)"); - fails += check_proc ("popcnt", HAS_CPU_FEATURE (POPCNT), - "HAS_CPU_FEATURE (POPCNT)"); + fails = check_proc ("avx", CPU_FEATURE_USABLE (AVX), + "CPU_FEATURE_USABLE (AVX)"); + fails += check_proc ("fma4", CPU_FEATURE_USABLE (FMA4), + "CPU_FEATURE_USABLE (FMA4)"); + fails += check_proc ("sse4_2", CPU_FEATURE_USABLE (SSE4_2), + "CPU_FEATURE_USABLE (SSE4_2)"); + fails += check_proc ("sse4_1", CPU_FEATURE_USABLE (SSE4_1) + , "CPU_FEATURE_USABLE (SSE4_1)"); + fails += check_proc ("ssse3", CPU_FEATURE_USABLE (SSSE3), + "CPU_FEATURE_USABLE (SSSE3)"); + fails += check_proc ("popcnt", CPU_FEATURE_USABLE (POPCNT), + "CPU_FEATURE_USABLE (POPCNT)"); printf ("%d differences between /proc/cpuinfo and glibc code.\n", fails); diff --git a/sysdeps/x86_64/multiarch/wcscpy.c b/sysdeps/x86_64/multiarch/wcscpy.c index 0dd2a9a..e08536c 100644 --- a/sysdeps/x86_64/multiarch/wcscpy.c +++ b/sysdeps/x86_64/multiarch/wcscpy.c @@ -34,7 +34,7 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_CPU_P (cpu_features, SSSE3)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSSE3)) return OPTIMIZE (ssse3); return OPTIMIZE (sse2); diff --git a/sysdeps/x86_64/multiarch/wcsnlen.c b/sysdeps/x86_64/multiarch/wcsnlen.c index 8c1fc1a..52e7e5d 100644 --- a/sysdeps/x86_64/multiarch/wcsnlen.c +++ b/sysdeps/x86_64/multiarch/wcsnlen.c @@ -36,11 +36,11 @@ IFUNC_SELECTOR (void) const struct cpu_features* cpu_features = __get_cpu_features (); if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER) - && CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable) + && CPU_FEATURE_USABLE_P (cpu_features, AVX2) && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load)) return OPTIMIZE (avx2); - if (CPU_FEATURES_CPU_P (cpu_features, SSE4_1)) + if (CPU_FEATURE_USABLE_P (cpu_features, SSE4_1)) return OPTIMIZE (sse4_1); return OPTIMIZE (sse2); |