public inbox for glibc-cvs@sourceware.org
help / color / mirror / Atom feed
* [glibc/release/2.28/master] x86: Set Prefer_No_VZEROUPPER and add Prefer_AVX2_STRCMP
@ 2022-01-27 22:40 H.J. Lu
0 siblings, 0 replies; only message in thread
From: H.J. Lu @ 2022-01-27 22:40 UTC (permalink / raw)
To: glibc-cvs
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=b2e59f87ae10e5874bb150bd87f3b92b47b1090e
commit b2e59f87ae10e5874bb150bd87f3b92b47b1090e
Author: H.J. Lu <hjl.tools@gmail.com>
Date: Fri Feb 26 05:36:59 2021 -0800
x86: Set Prefer_No_VZEROUPPER and add Prefer_AVX2_STRCMP
1. Set Prefer_No_VZEROUPPER if RTM is usable to avoid RTM abort triggered
by VZEROUPPER inside a transactionally executing RTM region.
2. Since to compare 2 32-byte strings, 256-bit EVEX strcmp requires 2
loads, 3 VPCMPs and 2 KORDs while AVX2 strcmp requires 1 load, 2 VPCMPEQs,
1 VPMINU and 1 VPMOVMSKB, AVX2 strcmp is faster than EVEX strcmp. Add
Prefer_AVX2_STRCMP to prefer AVX2 strcmp family functions.
(cherry picked from commit 1da50d4bda07f04135dca39f40e79fc9eabed1f8)
Diff:
---
sysdeps/x86/cpu-features.c | 20 ++++++++++++++++++--
sysdeps/x86/cpu-features.h | 2 ++
sysdeps/x86/cpu-tunables.c | 3 +++
3 files changed, 23 insertions(+), 2 deletions(-)
diff --git a/sysdeps/x86/cpu-features.c b/sysdeps/x86/cpu-features.c
index 4695ac80d4..5ddc8eebf4 100644
--- a/sysdeps/x86/cpu-features.c
+++ b/sysdeps/x86/cpu-features.c
@@ -347,8 +347,24 @@ init_cpu_features (struct cpu_features *cpu_features)
cpu_features->feature[index_arch_Prefer_No_VZEROUPPER]
|= bit_arch_Prefer_No_VZEROUPPER;
else
- cpu_features->feature[index_arch_Prefer_No_AVX512]
- |= bit_arch_Prefer_No_AVX512;
+ {
+ cpu_features->feature[index_arch_Prefer_No_AVX512]
+ |= bit_arch_Prefer_No_AVX512;
+
+ /* Avoid RTM abort triggered by VZEROUPPER inside a
+ transactionally executing RTM region. */
+ if (CPU_FEATURES_CPU_P (cpu_features, RTM))
+ cpu_features->feature[index_arch_Prefer_No_VZEROUPPER]
+ |= bit_arch_Prefer_No_VZEROUPPER;
+
+ /* Since to compare 2 32-byte strings, 256-bit EVEX strcmp
+ requires 2 loads, 3 VPCMPs and 2 KORDs while AVX2 strcmp
+ requires 1 load, 2 VPCMPEQs, 1 VPMINU and 1 VPMOVMSKB,
+ AVX2 strcmp is faster than EVEX strcmp. */
+ if (CPU_FEATURES_ARCH_P (cpu_features, AVX2_Usable))
+ cpu_features->feature[index_arch_Prefer_AVX2_STRCMP]
+ |= bit_arch_Prefer_AVX2_STRCMP;
+ }
}
/* This spells out "AuthenticAMD". */
else if (ebx == 0x68747541 && ecx == 0x444d4163 && edx == 0x69746e65)
diff --git a/sysdeps/x86/cpu-features.h b/sysdeps/x86/cpu-features.h
index 347a4b118d..1f60296e7b 100644
--- a/sysdeps/x86/cpu-features.h
+++ b/sysdeps/x86/cpu-features.h
@@ -41,6 +41,7 @@
#define bit_arch_MathVec_Prefer_No_AVX512 (1 << 21)
#define bit_arch_XSAVEC_Usable (1 << 22)
#define bit_arch_Prefer_FSRM (1 << 23)
+#define bit_arch_Prefer_AVX2_STRCMP (1 << 24)
/* CPUID Feature flags. */
@@ -266,6 +267,7 @@ extern const struct cpu_features *__get_cpu_features (void)
# 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
+# define index_arch_Prefer_AVX2_STRCMP FEATURE_INDEX_1
#endif /* !__ASSEMBLER__ */
diff --git a/sysdeps/x86/cpu-tunables.c b/sysdeps/x86/cpu-tunables.c
index 69155a8f44..824ee76b1a 100644
--- a/sysdeps/x86/cpu-tunables.c
+++ b/sysdeps/x86/cpu-tunables.c
@@ -282,6 +282,9 @@ TUNABLE_CALLBACK (set_hwcaps) (tunable_val_t *valp)
CHECK_GLIBC_IFUNC_ARCH_BOTH (n, cpu_features,
Fast_Copy_Backward, disable,
18);
+ CHECK_GLIBC_IFUNC_ARCH_NEED_ARCH_BOTH
+ (n, cpu_features, Prefer_AVX2_STRCMP, AVX2_Usable,
+ disable, 18);
}
break;
case 19:
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2022-01-27 22:40 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-01-27 22:40 [glibc/release/2.28/master] x86: Set Prefer_No_VZEROUPPER and add Prefer_AVX2_STRCMP H.J. Lu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).