Message ID | 20180521173740.GA12107@intel.com |
---|---|
State | New |
Headers | show |
Series | x86-64: Check Prefer_FSRM in ifunc-memmove.h | expand |
On Mon, May 21, 2018 at 10:37 AM, H.J. Lu <hongjiu.lu@intel.com> wrote: > Although the Fast Short REP MOVSB (FSRM) implementations of memmove, > memcpy and mempcpy aren't used by the current processors, this patch > adds Prefer_FSRM check in ifunc-memmove.h so that they can be used in > the future. > > * sysdeps/x86/cpu-features.h (bit_arch_Prefer_FSRM): New. > (index_arch_Prefer_FSRM): Likewise. > * sysdeps/x86/cpu-tunables.c (TUNABLE_CALLBACK (set_hwcaps)): > Also check Prefer_FSRM. > * sysdeps/x86_64/multiarch/ifunc-memmove.h (IFUNC_SELECTOR): > Also return OPTIMIZE (erms) for Prefer_FSRM. This is the patch I am going to check in shortly.
diff --git a/sysdeps/x86/cpu-features.h b/sysdeps/x86/cpu-features.h index 2088bd73ee..624e681e96 100644 --- a/sysdeps/x86/cpu-features.h +++ b/sysdeps/x86/cpu-features.h @@ -40,6 +40,7 @@ #define bit_arch_Prefer_No_AVX512 (1 << 20) #define bit_arch_MathVec_Prefer_No_AVX512 (1 << 21) #define bit_arch_XSAVEC_Usable (1 << 22) +#define bit_arch_Prefer_FSRM (1 << 23) /* CPUID Feature flags. */ @@ -264,6 +265,7 @@ extern const struct cpu_features *__get_cpu_features (void) # define index_arch_Prefer_No_AVX512 FEATURE_INDEX_1 # define index_arch_MathVec_Prefer_No_AVX512 FEATURE_INDEX_1 # define index_arch_XSAVEC_Usable FEATURE_INDEX_1 +# define index_arch_Prefer_FSRM FEATURE_INDEX_1 #endif /* !__ASSEMBLER__ */ diff --git a/sysdeps/x86/cpu-tunables.c b/sysdeps/x86/cpu-tunables.c index a21a615ec2..af761dcbbc 100644 --- a/sysdeps/x86/cpu-tunables.c +++ b/sysdeps/x86/cpu-tunables.c @@ -241,6 +241,8 @@ TUNABLE_CALLBACK (set_hwcaps) (tunable_val_t *valp) CHECK_GLIBC_IFUNC_ARCH_NEED_CPU_BOTH (n, cpu_features, Slow_SSE4_2, SSE4_2, disable, 11); + CHECK_GLIBC_IFUNC_ARCH_BOTH (n, cpu_features, Prefer_FSRM, + disable, 11); break; case 13: if (disable) diff --git a/sysdeps/x86_64/multiarch/ifunc-memmove.h b/sysdeps/x86_64/multiarch/ifunc-memmove.h index a2ffba0531..5b1eb1c92c 100644 --- a/sysdeps/x86_64/multiarch/ifunc-memmove.h +++ b/sysdeps/x86_64/multiarch/ifunc-memmove.h @@ -41,7 +41,8 @@ IFUNC_SELECTOR (void) { const struct cpu_features* cpu_features = __get_cpu_features (); - if (CPU_FEATURES_ARCH_P (cpu_features, Prefer_ERMS)) + if (CPU_FEATURES_ARCH_P (cpu_features, Prefer_ERMS) + || CPU_FEATURES_ARCH_P (cpu_features, Prefer_FSRM)) return OPTIMIZE (erms); if (CPU_FEATURES_ARCH_P (cpu_features, AVX512F_Usable)