public inbox for glibc-cvs@sourceware.org help / color / mirror / Atom feed
From: H.J. Lu <hjl@sourceware.org> To: glibc-cvs@sourceware.org Subject: [glibc/release/2.30/master] x86: Set Prefer_No_VZEROUPPER and add Prefer_AVX2_STRCMP Date: Thu, 27 Jan 2022 20:46:42 +0000 (GMT) [thread overview] Message-ID: <20220127204642.A7F0E3850424@sourceware.org> (raw) https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=293a787c88328bdad56b2cbe5d3b4b87a362bf5d commit 293a787c88328bdad56b2cbe5d3b4b87a362bf5d 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 4bab154913..a4d1eacbe7 100644 --- a/sysdeps/x86/cpu-features.c +++ b/sysdeps/x86/cpu-features.c @@ -424,8 +424,24 @@ 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" or "HygonGenuine". */ else if ((ebx == 0x68747541 && ecx == 0x444d4163 && edx == 0x69746e65) diff --git a/sysdeps/x86/cpu-features.h b/sysdeps/x86/cpu-features.h index 03a9b2a5e9..ca2924bd95 100644 --- a/sysdeps/x86/cpu-features.h +++ b/sysdeps/x86/cpu-features.h @@ -897,6 +897,7 @@ extern const struct cpu_features *__get_cpu_features (void) #define bit_arch_Prefer_FSRM (1u << 13) #define bit_arch_Prefer_No_AVX512 (1u << 14) #define bit_arch_MathVec_Prefer_No_AVX512 (1u << 15) +#define bit_arch_Prefer_AVX2_STRCMP (1u << 16) #define index_arch_Fast_Rep_String FEATURE_INDEX_2 #define index_arch_Fast_Copy_Backward FEATURE_INDEX_2 @@ -914,6 +915,7 @@ extern const struct cpu_features *__get_cpu_features (void) #define index_arch_Prefer_No_AVX512 FEATURE_INDEX_2 #define index_arch_MathVec_Prefer_No_AVX512 FEATURE_INDEX_2 #define index_arch_Prefer_FSRM FEATURE_INDEX_2 +#define index_arch_Prefer_AVX2_STRCMP FEATURE_INDEX_2 /* XCR0 Feature flags. */ #define bit_XMM_state (1u << 1) diff --git a/sysdeps/x86/cpu-tunables.c b/sysdeps/x86/cpu-tunables.c index 2cb315111b..d4d5e450a6 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:
reply other threads:[~2022-01-27 20:46 UTC|newest] Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20220127204642.A7F0E3850424@sourceware.org \ --to=hjl@sourceware.org \ --cc=glibc-cvs@sourceware.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
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).