From: Noah Goldstein <goldstein.w.n@gmail.com>
To: "H.J. Lu" <hjl.tools@gmail.com>
Cc: GNU C Library <libc-alpha@sourceware.org>,
"Carlos O'Donell" <carlos@systemhalted.org>
Subject: Re: [PATCH v4 22/23] x86: Add EVEX optimized str{n}casecmp
Date: Fri, 25 Mar 2022 13:18:02 -0500 [thread overview]
Message-ID: <CAFUsyf+FsB7x-O-FB87vsxNF5XaXJ8+CxOr-sgaO+5PbXcH6qQ@mail.gmail.com> (raw)
In-Reply-To: <CAMe9rOp3wFrvLmM9PhtHWJis-i85QCrL17Oua8+xj=x2eSVxkg@mail.gmail.com>
On Fri, Mar 25, 2022 at 1:15 PM H.J. Lu <hjl.tools@gmail.com> wrote:
>
> On Thu, Mar 24, 2022 at 4:56 PM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
> >
> > geometric_mean(N=40) of all benchmarks EVEX / SSE42: .621
> >
> > All string/memory tests pass.
> > Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
> > ---
> > sysdeps/x86_64/multiarch/Makefile | 2 +
> > sysdeps/x86_64/multiarch/ifunc-impl-list.c | 16 +
> > sysdeps/x86_64/multiarch/ifunc-strcasecmp.h | 5 +
> > sysdeps/x86_64/multiarch/strcasecmp_l-evex.S | 23 ++
> > sysdeps/x86_64/multiarch/strcmp-evex.S | 290 ++++++++++++++++---
> > sysdeps/x86_64/multiarch/strncase_l-evex.S | 25 ++
> > 6 files changed, 321 insertions(+), 40 deletions(-)
> > create mode 100644 sysdeps/x86_64/multiarch/strcasecmp_l-evex.S
> > create mode 100644 sysdeps/x86_64/multiarch/strncase_l-evex.S
> >
> > diff --git a/sysdeps/x86_64/multiarch/Makefile b/sysdeps/x86_64/multiarch/Makefile
> > index 06e1848823..35d80dc2ff 100644
> > --- a/sysdeps/x86_64/multiarch/Makefile
> > +++ b/sysdeps/x86_64/multiarch/Makefile
> > @@ -57,6 +57,7 @@ sysdep_routines += \
> > strcasecmp_l-avx \
> > strcasecmp_l-avx2 \
> > strcasecmp_l-avx2-rtm \
> > + strcasecmp_l-evex \
> > strcasecmp_l-sse2 \
> > strcasecmp_l-sse4_2 \
> > strcasecmp_l-ssse3 \
> > @@ -97,6 +98,7 @@ sysdep_routines += \
> > strncase_l-avx \
> > strncase_l-avx2 \
> > strncase_l-avx2-rtm \
> > + strncase_l-evex \
> > strncase_l-sse2 \
> > strncase_l-sse4_2 \
> > strncase_l-ssse3 \
> > diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
> > index 3c556d07ac..f1a4d3dac2 100644
> > --- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c
> > +++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c
> > @@ -436,6 +436,10 @@ __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,
> > + (CPU_FEATURE_USABLE (AVX512VL)
> > + && CPU_FEATURE_USABLE (AVX512BW)),
> > + __strcasecmp_evex)
> > IFUNC_IMPL_ADD (array, i, strcasecmp,
> > CPU_FEATURE_USABLE (AVX2),
> > __strcasecmp_avx2)
> > @@ -456,6 +460,10 @@ __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_l,
> > + IFUNC_IMPL_ADD (array, i, strcasecmp,
> > + (CPU_FEATURE_USABLE (AVX512VL)
> > + && CPU_FEATURE_USABLE (AVX512BW)),
> > + __strcasecmp_l_evex)
> > IFUNC_IMPL_ADD (array, i, strcasecmp,
> > CPU_FEATURE_USABLE (AVX2),
> > __strcasecmp_l_avx2)
> > @@ -590,6 +598,10 @@ __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,
> > + IFUNC_IMPL_ADD (array, i, strncasecmp,
> > + (CPU_FEATURE_USABLE (AVX512VL)
> > + && CPU_FEATURE_USABLE (AVX512BW)),
> > + __strncasecmp_evex)
> > IFUNC_IMPL_ADD (array, i, strncasecmp,
> > CPU_FEATURE_USABLE (AVX2),
> > __strncasecmp_avx2)
> > @@ -611,6 +623,10 @@ __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,
> > + (CPU_FEATURE_USABLE (AVX512VL)
> > + && CPU_FEATURE_USABLE (AVX512BW)),
> > + __strncasecmp_l_evex)
> > IFUNC_IMPL_ADD (array, i, strncasecmp,
> > CPU_FEATURE_USABLE (AVX2),
> > __strncasecmp_l_avx2)
> > diff --git a/sysdeps/x86_64/multiarch/ifunc-strcasecmp.h b/sysdeps/x86_64/multiarch/ifunc-strcasecmp.h
> > index c4de111fd0..bf0d146e7f 100644
> > --- a/sysdeps/x86_64/multiarch/ifunc-strcasecmp.h
> > +++ b/sysdeps/x86_64/multiarch/ifunc-strcasecmp.h
> > @@ -25,6 +25,7 @@ extern __typeof (REDIRECT_NAME) OPTIMIZE (sse42) attribute_hidden;
> > extern __typeof (REDIRECT_NAME) OPTIMIZE (avx) attribute_hidden;
> > extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2) attribute_hidden;
> > extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_rtm) attribute_hidden;
> > +extern __typeof (REDIRECT_NAME) OPTIMIZE (evex) attribute_hidden;
> >
> > static inline void *
> > IFUNC_SELECTOR (void)
> > @@ -34,6 +35,10 @@ IFUNC_SELECTOR (void)
> > if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)
> > && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load))
> > {
> > + if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)
> > + && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW))
> > + return OPTIMIZE (evex);
> > +
> > if (CPU_FEATURE_USABLE_P (cpu_features, RTM))
> > return OPTIMIZE (avx2_rtm);
> >
> > diff --git a/sysdeps/x86_64/multiarch/strcasecmp_l-evex.S b/sysdeps/x86_64/multiarch/strcasecmp_l-evex.S
> > new file mode 100644
> > index 0000000000..58642db748
> > --- /dev/null
> > +++ b/sysdeps/x86_64/multiarch/strcasecmp_l-evex.S
> > @@ -0,0 +1,23 @@
> > +/* strcasecmp_l optimized with EVEX.
> > + Copyright (C) 2017-2022 Free Software Foundation, Inc.
> > + This file is part of the GNU C Library.
> > +
> > + The GNU C Library is free software; you can redistribute it and/or
> > + modify it under the terms of the GNU Lesser General Public
> > + License as published by the Free Software Foundation; either
> > + version 2.1 of the License, or (at your option) any later version.
> > +
> > + The GNU C Library is distributed in the hope that it will be useful,
> > + but WITHOUT ANY WARRANTY; without even the implied warranty of
> > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> > + Lesser General Public License for more details.
> > +
> > + You should have received a copy of the GNU Lesser General Public
> > + License along with the GNU C Library; if not, see
> > + <https://www.gnu.org/licenses/>. */
> > +
> > +#ifndef STRCMP
> > +# define STRCMP __strcasecmp_l_evex
> > +#endif
> > +#define USE_AS_STRCASECMP_L
> > +#include "strcmp-evex.S"
> > diff --git a/sysdeps/x86_64/multiarch/strcmp-evex.S b/sysdeps/x86_64/multiarch/strcmp-evex.S
> > index 56d8c118e4..2a5b3ce037 100644
> > --- a/sysdeps/x86_64/multiarch/strcmp-evex.S
> > +++ b/sysdeps/x86_64/multiarch/strcmp-evex.S
> > @@ -19,6 +19,9 @@
> > #if IS_IN (libc)
> >
> > # include <sysdep.h>
> > +# if defined USE_AS_STRCASECMP_L
> > +# include "locale-defines.h"
> > +# endif
> >
> > # ifndef STRCMP
> > # define STRCMP __strcmp_evex
> > @@ -34,19 +37,29 @@
> > # define VMOVA vmovdqa64
> >
> > # ifdef USE_AS_WCSCMP
> > -# define TESTEQ subl $0xff,
> > +# ifndef OVERFLOW_STRCMP
> > +# define OVERFLOW_STRCMP __wcscmp_evex
> > +# endif
> > +
> > +# define TESTEQ subl $0xff,
> > /* Compare packed dwords. */
> > # define VPCMP vpcmpd
> > # define VPMINU vpminud
> > # define VPTESTM vptestmd
> > +# define VPTESTNM vptestnmd
> > /* 1 dword char == 4 bytes. */
> > # define SIZE_OF_CHAR 4
> > # else
> > +# ifndef OVERFLOW_STRCMP
> > +# define OVERFLOW_STRCMP __strcmp_evex
> > +# endif
> > +
> > # define TESTEQ incl
> > /* Compare packed bytes. */
> > # define VPCMP vpcmpb
> > # define VPMINU vpminub
> > # define VPTESTM vptestmb
> > +# define VPTESTNM vptestnmb
> > /* 1 byte char == 1 byte. */
> > # define SIZE_OF_CHAR 1
> > # endif
> > @@ -73,11 +86,16 @@
> > # define VEC_OFFSET (-VEC_SIZE)
> > # endif
> >
> > -# define XMMZERO xmm16
> > # define XMM0 xmm17
> > # define XMM1 xmm18
> >
> > -# define YMMZERO ymm16
> > +# define XMM10 xmm27
> > +# define XMM11 xmm28
> > +# define XMM12 xmm29
> > +# define XMM13 xmm30
> > +# define XMM14 xmm31
> > +
> > +
> > # define YMM0 ymm17
> > # define YMM1 ymm18
> > # define YMM2 ymm19
> > @@ -89,6 +107,87 @@
> > # define YMM8 ymm25
> > # define YMM9 ymm26
> > # define YMM10 ymm27
> > +# define YMM11 ymm28
> > +# define YMM12 ymm29
> > +# define YMM13 ymm30
> > +# define YMM14 ymm31
> > +
> > +# ifdef USE_AS_STRCASECMP_L
> > +# define BYTE_LOOP_REG OFFSET_REG
> > +# else
> > +# define BYTE_LOOP_REG ecx
> > +# endif
> > +
> > +# ifdef USE_AS_STRCASECMP_L
> > +# ifdef USE_AS_STRNCMP
> > +# define STRCASECMP __strncasecmp_evex
> > +# define LOCALE_REG rcx
> > +# define LOCALE_REG_LP RCX_LP
> > +# define STRCASECMP_NONASCII __strncasecmp_l_nonascii
> > +# else
> > +# define STRCASECMP __strcasecmp_evex
> > +# define LOCALE_REG rdx
> > +# define LOCALE_REG_LP RDX_LP
> > +# define STRCASECMP_NONASCII __strcasecmp_l_nonascii
> > +# endif
> > +# endif
> > +
> > +# define LCASE_MIN_YMM %YMM12
> > +# define LCASE_MAX_YMM %YMM13
> > +# define CASE_ADD_YMM %YMM14
> > +
> > +# define LCASE_MIN_XMM %XMM12
> > +# define LCASE_MAX_XMM %XMM13
> > +# define CASE_ADD_XMM %XMM14
> > +
> > + /* NB: wcsncmp uses r11 but strcasecmp is never used in
> > + conjunction with wcscmp. */
> > +# define TOLOWER_BASE %r11
> > +
> > +# ifdef USE_AS_STRCASECMP_L
> > +# define _REG(x, y) x ## y
> > +# define REG(x, y) _REG(x, y)
> > +# define TOLOWER(reg1, reg2, ext) \
> > + vpsubb REG(LCASE_MIN_, ext), reg1, REG(%ext, 10); \
> > + vpsubb REG(LCASE_MIN_, ext), reg2, REG(%ext, 11); \
> > + vpcmpub $1, REG(LCASE_MAX_, ext), REG(%ext, 10), %k5; \
> > + vpcmpub $1, REG(LCASE_MAX_, ext), REG(%ext, 11), %k6; \
> > + vpaddb reg1, REG(CASE_ADD_, ext), reg1{%k5}; \
> > + vpaddb reg2, REG(CASE_ADD_, ext), reg2{%k6}
> > +
> > +# define TOLOWER_gpr(src, dst) movl (TOLOWER_BASE, src, 4), dst
> > +# define TOLOWER_YMM(...) TOLOWER(__VA_ARGS__, YMM)
> > +# define TOLOWER_XMM(...) TOLOWER(__VA_ARGS__, XMM)
> > +
> > +# define CMP_R1_R2(s1_reg, s2_reg, reg_out, ext) \
> > + TOLOWER (s1_reg, s2_reg, ext); \
> > + VPCMP $0, s1_reg, s2_reg, reg_out
> > +
> > +# define CMP_R1_S2(s1_reg, s2_mem, s2_reg, reg_out, ext) \
> > + VMOVU s2_mem, s2_reg; \
> > + CMP_R1_R2(s1_reg, s2_reg, reg_out, ext)
> > +
> > +# define CMP_R1_R2_YMM(...) CMP_R1_R2(__VA_ARGS__, YMM)
> > +# define CMP_R1_R2_XMM(...) CMP_R1_R2(__VA_ARGS__, XMM)
> > +
> > +# define CMP_R1_S2_YMM(...) CMP_R1_S2(__VA_ARGS__, YMM)
> > +# define CMP_R1_S2_XMM(...) CMP_R1_S2(__VA_ARGS__, XMM)
> > +
> > +# else
> > +# define TOLOWER_gpr(...)
> > +# define TOLOWER_YMM(...)
> > +# define TOLOWER_XMM(...)
> > +
> > +# define CMP_R1_R2_YMM(s1_reg, s2_reg, reg_out) \
> > + VPCMP $0, s2_reg, s1_reg, reg_out
> > +
> > +# define CMP_R1_R2_XMM(...) CMP_R1_R2_YMM(__VA_ARGS__)
> > +
> > +# define CMP_R1_S2_YMM(s1_reg, s2_mem, unused, reg_out) \
> > + VPCMP $0, s2_mem, s1_reg, reg_out
> > +
> > +# define CMP_R1_S2_XMM(...) CMP_R1_S2_YMM(__VA_ARGS__)
> > +# endif
> >
> > /* Warning!
> > wcscmp/wcsncmp have to use SIGNED comparison for elements.
> > @@ -112,8 +211,45 @@
> > returned. */
> >
> > .section .text.evex, "ax", @progbits
> > -ENTRY(STRCMP)
> > + .align 16
> > + .type STRCMP, @function
> > + .globl STRCMP
> > + .hidden STRCMP
> > +
> > +# ifdef USE_AS_STRCASECMP_L
> > +ENTRY (STRCASECMP)
> > + movq __libc_tsd_LOCALE@gottpoff(%rip), %rax
> > + mov %fs:(%rax), %LOCALE_REG_LP
> > +
> > + /* Either 1 or 5 bytes (dependeing if CET is enabled). */
> > + .p2align 4
> > +END (STRCASECMP)
> > + /* FALLTHROUGH to strcasecmp/strncasecmp_l. */
> > +# endif
> > +
> > + .p2align 4
> > +STRCMP:
> > + cfi_startproc
> > + _CET_ENDBR
> > + CALL_MCOUNT
> > +
> > +# if defined USE_AS_STRCASECMP_L
> > + /* We have to fall back on the C implementation for locales with
> > + encodings not matching ASCII for single bytes. */
> > +# if LOCALE_T___LOCALES != 0 || LC_CTYPE != 0
> > + mov LOCALE_T___LOCALES + LC_CTYPE * LP_SIZE(%LOCALE_REG), %RAX_LP
> > +# else
> > + mov (%LOCALE_REG), %RAX_LP
> > +# endif
> > + testl $1, LOCALE_DATA_VALUES + _NL_CTYPE_NONASCII_CASE * SIZEOF_VALUES(%rax)
> > + jne STRCASECMP_NONASCII
> > + leaq _nl_C_LC_CTYPE_tolower + 128 * 4(%rip), TOLOWER_BASE
> > +# endif
> > +
> > # ifdef USE_AS_STRNCMP
> > + /* Don't overwrite LOCALE_REG (rcx) until we have pass
> > + L(one_or_less). Otherwise we might use the wrong locale in
> > + the OVERFLOW_STRCMP (strcasecmp_l). */
> > # ifdef __ILP32__
> > /* Clear the upper 32 bits. */
> > movl %edx, %edx
> > @@ -125,6 +261,32 @@ ENTRY(STRCMP)
> > actually bound the buffer. */
> > jle L(one_or_less)
> > # endif
> > +
> > +# if defined USE_AS_STRCASECMP_L
> > + .section .rodata.cst32, "aM", @progbits, 32
> > + .align 32
> > +L(lcase_min):
> > + .quad 0x4141414141414141
> > + .quad 0x4141414141414141
> > + .quad 0x4141414141414141
> > + .quad 0x4141414141414141
> > +L(lcase_max):
> > + .quad 0x1a1a1a1a1a1a1a1a
> > + .quad 0x1a1a1a1a1a1a1a1a
> > + .quad 0x1a1a1a1a1a1a1a1a
> > + .quad 0x1a1a1a1a1a1a1a1a
> > +L(case_add):
> > + .quad 0x2020202020202020
> > + .quad 0x2020202020202020
> > + .quad 0x2020202020202020
> > + .quad 0x2020202020202020
> > + .previous
> > +
> > + vmovdqa64 L(lcase_min)(%rip), LCASE_MIN_YMM
> > + vmovdqa64 L(lcase_max)(%rip), LCASE_MAX_YMM
> > + vmovdqa64 L(case_add)(%rip), CASE_ADD_YMM
> > +# endif
> > +
> > movl %edi, %eax
> > orl %esi, %eax
> > /* Shift out the bits irrelivant to page boundary ([63:12]). */
> > @@ -139,7 +301,7 @@ L(no_page_cross):
> > VPTESTM %YMM0, %YMM0, %k2
> > /* Each bit cleared in K1 represents a mismatch or a null CHAR
> > in YMM0 and 32 bytes at (%rsi). */
> > - VPCMP $0, (%rsi), %YMM0, %k1{%k2}
> > + CMP_R1_S2_YMM (%YMM0, (%rsi), %YMM1, %k1){%k2}
> > kmovd %k1, %ecx
> > # ifdef USE_AS_STRNCMP
> > cmpq $CHAR_PER_VEC, %rdx
> > @@ -169,6 +331,8 @@ L(return_vec_0):
> > # else
> > movzbl (%rdi, %rcx), %eax
> > movzbl (%rsi, %rcx), %ecx
> > + TOLOWER_gpr (%rax, %eax)
> > + TOLOWER_gpr (%rcx, %ecx)
> > subl %ecx, %eax
> > # endif
> > L(ret0):
> > @@ -188,11 +352,15 @@ L(ret_zero):
> >
> > .p2align 4,, 5
> > L(one_or_less):
> > +# ifdef USE_AS_STRCASECMP_L
> > + /* Set locale argument for strcasecmp. */
> > + movq %LOCALE_REG, %rdx
> > +# endif
> > jb L(ret_zero)
> > -# ifdef USE_AS_WCSCMP
> > /* 'nbe' covers the case where length is negative (large
> > unsigned). */
> > - jnbe __wcscmp_evex
> > + jnbe OVERFLOW_STRCMP
> > +# ifdef USE_AS_WCSCMP
> > movl (%rdi), %edx
> > xorl %eax, %eax
> > cmpl (%rsi), %edx
> > @@ -201,11 +369,10 @@ L(one_or_less):
> > negl %eax
> > orl $1, %eax
> > # else
> > - /* 'nbe' covers the case where length is negative (large
> > - unsigned). */
> > - jnbe __strcmp_evex
> > movzbl (%rdi), %eax
> > movzbl (%rsi), %ecx
> > + TOLOWER_gpr (%rax, %eax)
> > + TOLOWER_gpr (%rcx, %ecx)
> > subl %ecx, %eax
> > # endif
> > L(ret1):
> > @@ -233,6 +400,8 @@ L(return_vec_1):
> > # else
> > movzbl VEC_SIZE(%rdi, %rcx), %eax
> > movzbl VEC_SIZE(%rsi, %rcx), %ecx
> > + TOLOWER_gpr (%rax, %eax)
> > + TOLOWER_gpr (%rcx, %ecx)
> > subl %ecx, %eax
> > # endif
> > L(ret2):
> > @@ -270,6 +439,8 @@ L(return_vec_2):
> > # else
> > movzbl (VEC_SIZE * 2)(%rdi, %rcx), %eax
> > movzbl (VEC_SIZE * 2)(%rsi, %rcx), %ecx
> > + TOLOWER_gpr (%rax, %eax)
> > + TOLOWER_gpr (%rcx, %ecx)
> > subl %ecx, %eax
> > # endif
> > L(ret3):
> > @@ -290,6 +461,8 @@ L(return_vec_3):
> > # else
> > movzbl (VEC_SIZE * 3)(%rdi, %rcx), %eax
> > movzbl (VEC_SIZE * 3)(%rsi, %rcx), %ecx
> > + TOLOWER_gpr (%rax, %eax)
> > + TOLOWER_gpr (%rcx, %ecx)
> > subl %ecx, %eax
> > # endif
> > L(ret4):
> > @@ -303,7 +476,7 @@ L(more_3x_vec):
> > /* Safe to compare 4x vectors. */
> > VMOVU (VEC_SIZE)(%rdi), %YMM0
> > VPTESTM %YMM0, %YMM0, %k2
> > - VPCMP $0, (VEC_SIZE)(%rsi), %YMM0, %k1{%k2}
> > + CMP_R1_S2_YMM (%YMM0, VEC_SIZE(%rsi), %YMM1, %k1){%k2}
> > kmovd %k1, %ecx
> > TESTEQ %ecx
> > jnz L(return_vec_1)
> > @@ -315,14 +488,14 @@ L(more_3x_vec):
> >
> > VMOVU (VEC_SIZE * 2)(%rdi), %YMM0
> > VPTESTM %YMM0, %YMM0, %k2
> > - VPCMP $0, (VEC_SIZE * 2)(%rsi), %YMM0, %k1{%k2}
> > + CMP_R1_S2_YMM (%YMM0, (VEC_SIZE * 2)(%rsi), %YMM1, %k1){%k2}
> > kmovd %k1, %ecx
> > TESTEQ %ecx
> > jnz L(return_vec_2)
> >
> > VMOVU (VEC_SIZE * 3)(%rdi), %YMM0
> > VPTESTM %YMM0, %YMM0, %k2
> > - VPCMP $0, (VEC_SIZE * 3)(%rsi), %YMM0, %k1{%k2}
> > + CMP_R1_S2_YMM (%YMM0, (VEC_SIZE * 3)(%rsi), %YMM1, %k1){%k2}
> > kmovd %k1, %ecx
> > TESTEQ %ecx
> > jnz L(return_vec_3)
> > @@ -381,7 +554,6 @@ L(prepare_loop_aligned):
> > subl %esi, %eax
> > andl $(PAGE_SIZE - 1), %eax
> >
> > - vpxorq %YMMZERO, %YMMZERO, %YMMZERO
> >
> > /* Loop 4x comparisons at a time. */
> > .p2align 4
> > @@ -413,22 +585,35 @@ L(loop_skip_page_cross_check):
> > /* A zero CHAR in YMM9 means that there is a null CHAR. */
> > VPMINU %YMM8, %YMM9, %YMM9
> >
> > - /* Each bit set in K1 represents a non-null CHAR in YMM8. */
> > + /* Each bit set in K1 represents a non-null CHAR in YMM9. */
> > VPTESTM %YMM9, %YMM9, %k1
> > -
> > +# ifndef USE_AS_STRCASECMP_L
> > vpxorq (VEC_SIZE * 0)(%rsi), %YMM0, %YMM1
> > vpxorq (VEC_SIZE * 1)(%rsi), %YMM2, %YMM3
> > vpxorq (VEC_SIZE * 2)(%rsi), %YMM4, %YMM5
> > /* Ternary logic to xor (VEC_SIZE * 3)(%rsi) with YMM6 while
> > oring with YMM1. Result is stored in YMM6. */
> > vpternlogd $0xde, (VEC_SIZE * 3)(%rsi), %YMM1, %YMM6
> > -
> > +# else
> > + VMOVU (VEC_SIZE * 0)(%rsi), %YMM1
> > + TOLOWER_YMM (%YMM0, %YMM1)
> > + VMOVU (VEC_SIZE * 1)(%rsi), %YMM3
> > + TOLOWER_YMM (%YMM2, %YMM3)
> > + VMOVU (VEC_SIZE * 2)(%rsi), %YMM5
> > + TOLOWER_YMM (%YMM4, %YMM5)
> > + VMOVU (VEC_SIZE * 3)(%rsi), %YMM7
> > + TOLOWER_YMM (%YMM6, %YMM7)
> > + vpxorq %YMM0, %YMM1, %YMM1
> > + vpxorq %YMM2, %YMM3, %YMM3
> > + vpxorq %YMM4, %YMM5, %YMM5
> > + vpternlogd $0xde, %YMM7, %YMM1, %YMM6
> > +# endif
> > /* Or together YMM3, YMM5, and YMM6. */
> > vpternlogd $0xfe, %YMM3, %YMM5, %YMM6
> >
> >
> > /* A non-zero CHAR in YMM6 represents a mismatch. */
> > - VPCMP $0, %YMMZERO, %YMM6, %k0{%k1}
> > + VPTESTNM %YMM6, %YMM6, %k0{%k1}
> > kmovd %k0, %LOOP_REG
> >
> > TESTEQ %LOOP_REG
> > @@ -437,13 +622,13 @@ L(loop_skip_page_cross_check):
> >
> > /* Find which VEC has the mismatch of end of string. */
> > VPTESTM %YMM0, %YMM0, %k1
> > - VPCMP $0, %YMMZERO, %YMM1, %k0{%k1}
> > + VPTESTNM %YMM1, %YMM1, %k0{%k1}
> > kmovd %k0, %ecx
> > TESTEQ %ecx
> > jnz L(return_vec_0_end)
> >
> > VPTESTM %YMM2, %YMM2, %k1
> > - VPCMP $0, %YMMZERO, %YMM3, %k0{%k1}
> > + VPTESTNM %YMM3, %YMM3, %k0{%k1}
> > kmovd %k0, %ecx
> > TESTEQ %ecx
> > jnz L(return_vec_1_end)
> > @@ -457,7 +642,7 @@ L(return_vec_2_3_end):
> > # endif
> >
> > VPTESTM %YMM4, %YMM4, %k1
> > - VPCMP $0, %YMMZERO, %YMM5, %k0{%k1}
> > + VPTESTNM %YMM5, %YMM5, %k0{%k1}
> > kmovd %k0, %ecx
> > TESTEQ %ecx
> > # if CHAR_PER_VEC <= 16
> > @@ -493,6 +678,8 @@ L(return_vec_3_end):
> > # else
> > movzbl (VEC_SIZE * 2)(%rdi, %LOOP_REG64), %eax
> > movzbl (VEC_SIZE * 2)(%rsi, %LOOP_REG64), %ecx
> > + TOLOWER_gpr (%rax, %eax)
> > + TOLOWER_gpr (%rcx, %ecx)
> > subl %ecx, %eax
> > xorl %r8d, %eax
> > subl %r8d, %eax
> > @@ -545,6 +732,8 @@ L(return_vec_0_end):
> > # else
> > movzbl (%rdi, %rcx), %eax
> > movzbl (%rsi, %rcx), %ecx
> > + TOLOWER_gpr (%rax, %eax)
> > + TOLOWER_gpr (%rcx, %ecx)
> > subl %ecx, %eax
> > /* Flip `eax` if `rdi` and `rsi` where swapped in page cross
> > logic. Subtract `r8d` after xor for zero case. */
> > @@ -569,6 +758,8 @@ L(return_vec_1_end):
> > # else
> > movzbl VEC_SIZE(%rdi, %rcx), %eax
> > movzbl VEC_SIZE(%rsi, %rcx), %ecx
> > + TOLOWER_gpr (%rax, %eax)
> > + TOLOWER_gpr (%rcx, %ecx)
> > subl %ecx, %eax
> > xorl %r8d, %eax
> > subl %r8d, %eax
> > @@ -598,7 +789,7 @@ L(page_cross_during_loop):
> >
> > VMOVA (%rdi), %YMM0
> > VPTESTM %YMM0, %YMM0, %k2
> > - VPCMP $0, (%rsi), %YMM0, %k1{%k2}
> > + CMP_R1_S2_YMM (%YMM0, (%rsi), %YMM1, %k1){%k2}
> > kmovd %k1, %ecx
> > TESTEQ %ecx
> > jnz L(return_vec_0_end)
> > @@ -619,8 +810,7 @@ L(less_1x_vec_till_page_cross):
> > been loaded earlier so must be valid. */
> > VMOVU -VEC_SIZE(%rdi, %rax), %YMM0
> > VPTESTM %YMM0, %YMM0, %k2
> > - VPCMP $0, -VEC_SIZE(%rsi, %rax), %YMM0, %k1{%k2}
> > -
> > + CMP_R1_S2_YMM (%YMM0, -VEC_SIZE(%rsi, %rax), %YMM1, %k1){%k2}
> > /* Mask of potentially valid bits. The lower bits can be out of
> > range comparisons (but safe regarding page crosses). */
> >
> > @@ -642,6 +832,8 @@ L(less_1x_vec_till_page_cross):
> >
> > # ifdef USE_AS_STRNCMP
> > # ifdef USE_AS_WCSCMP
> > + /* NB: strcasecmp not used with WCSCMP so this access to r11 is
> > + safe. */
> > movl %eax, %r11d
> > shrl $2, %r11d
> > cmpq %r11, %rdx
> > @@ -679,6 +871,8 @@ L(return_page_cross_cmp_mem):
> > # else
> > movzbl VEC_OFFSET(%rdi, %rcx), %eax
> > movzbl VEC_OFFSET(%rsi, %rcx), %ecx
> > + TOLOWER_gpr (%rax, %eax)
> > + TOLOWER_gpr (%rcx, %ecx)
> > subl %ecx, %eax
> > xorl %r8d, %eax
> > subl %r8d, %eax
> > @@ -709,7 +903,7 @@ L(more_2x_vec_till_page_cross):
> >
> > VMOVA VEC_SIZE(%rdi), %YMM0
> > VPTESTM %YMM0, %YMM0, %k2
> > - VPCMP $0, VEC_SIZE(%rsi), %YMM0, %k1{%k2}
> > + CMP_R1_S2_YMM (%YMM0, VEC_SIZE(%rsi), %YMM1, %k1){%k2}
> > kmovd %k1, %ecx
> > TESTEQ %ecx
> > jnz L(return_vec_1_end)
> > @@ -724,14 +918,14 @@ L(more_2x_vec_till_page_cross):
> > /* Safe to include comparisons from lower bytes. */
> > VMOVU -(VEC_SIZE * 2)(%rdi, %rax), %YMM0
> > VPTESTM %YMM0, %YMM0, %k2
> > - VPCMP $0, -(VEC_SIZE * 2)(%rsi, %rax), %YMM0, %k1{%k2}
> > + CMP_R1_S2_YMM (%YMM0, -(VEC_SIZE * 2)(%rsi, %rax), %YMM1, %k1){%k2}
> > kmovd %k1, %ecx
> > TESTEQ %ecx
> > jnz L(return_vec_page_cross_0)
> >
> > VMOVU -(VEC_SIZE * 1)(%rdi, %rax), %YMM0
> > VPTESTM %YMM0, %YMM0, %k2
> > - VPCMP $0, -(VEC_SIZE * 1)(%rsi, %rax), %YMM0, %k1{%k2}
> > + CMP_R1_S2_YMM (%YMM0, -(VEC_SIZE * 1)(%rsi, %rax), %YMM1, %k1){%k2}
> > kmovd %k1, %ecx
> > TESTEQ %ecx
> > jnz L(return_vec_page_cross_1)
> > @@ -740,6 +934,8 @@ L(more_2x_vec_till_page_cross):
> > /* Must check length here as length might proclude reading next
> > page. */
> > # ifdef USE_AS_WCSCMP
> > + /* NB: strcasecmp not used with WCSCMP so this access to r11 is
> > + safe. */
> > movl %eax, %r11d
> > shrl $2, %r11d
> > cmpq %r11, %rdx
> > @@ -754,12 +950,19 @@ L(more_2x_vec_till_page_cross):
> > VMOVA (VEC_SIZE * 3)(%rdi), %YMM6
> > VPMINU %YMM4, %YMM6, %YMM9
> > VPTESTM %YMM9, %YMM9, %k1
> > -
> > +# ifndef USE_AS_STRCASECMP_L
> > vpxorq (VEC_SIZE * 2)(%rsi), %YMM4, %YMM5
> > /* YMM6 = YMM5 | ((VEC_SIZE * 3)(%rsi) ^ YMM6). */
> > vpternlogd $0xde, (VEC_SIZE * 3)(%rsi), %YMM5, %YMM6
> > -
> > - VPCMP $0, %YMMZERO, %YMM6, %k0{%k1}
> > +# else
> > + VMOVU (VEC_SIZE * 2)(%rsi), %YMM5
> > + TOLOWER_YMM (%YMM4, %YMM5)
> > + VMOVU (VEC_SIZE * 3)(%rsi), %YMM7
> > + TOLOWER_YMM (%YMM6, %YMM7)
> > + vpxorq %YMM4, %YMM5, %YMM5
> > + vpternlogd $0xde, %YMM7, %YMM5, %YMM6
> > +# endif
> > + VPTESTNM %YMM6, %YMM6, %k0{%k1}
> > kmovd %k0, %LOOP_REG
> > TESTEQ %LOOP_REG
> > jnz L(return_vec_2_3_end)
> > @@ -815,6 +1018,8 @@ L(return_vec_page_cross_1):
> > # else
> > movzbl VEC_OFFSET(%rdi, %rcx), %eax
> > movzbl VEC_OFFSET(%rsi, %rcx), %ecx
> > + TOLOWER_gpr (%rax, %eax)
> > + TOLOWER_gpr (%rcx, %ecx)
> > subl %ecx, %eax
> > xorl %r8d, %eax
> > subl %r8d, %eax
> > @@ -871,7 +1076,7 @@ L(page_cross):
> > L(page_cross_loop):
> > VMOVU (%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %YMM0
> > VPTESTM %YMM0, %YMM0, %k2
> > - VPCMP $0, (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %YMM0, %k1{%k2}
> > + CMP_R1_S2_YMM (%YMM0, (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %YMM1, %k1){%k2}
> > kmovd %k1, %ecx
> > TESTEQ %ecx
> > jnz L(check_ret_vec_page_cross)
> > @@ -895,7 +1100,7 @@ L(page_cross_loop):
> > */
> > VMOVU (%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %YMM0
> > VPTESTM %YMM0, %YMM0, %k2
> > - VPCMP $0, (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %YMM0, %k1{%k2}
> > + CMP_R1_S2_YMM (%YMM0, (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %YMM1, %k1){%k2}
> >
> > kmovd %k1, %ecx
> > # ifdef USE_AS_STRNCMP
> > @@ -930,6 +1135,8 @@ L(ret_vec_page_cross_cont):
> > # else
> > movzbl (%rdi, %rcx, SIZE_OF_CHAR), %eax
> > movzbl (%rsi, %rcx, SIZE_OF_CHAR), %ecx
> > + TOLOWER_gpr (%rax, %eax)
> > + TOLOWER_gpr (%rcx, %ecx)
> > subl %ecx, %eax
> > xorl %r8d, %eax
> > subl %r8d, %eax
> > @@ -989,7 +1196,7 @@ L(less_1x_vec_till_page):
> > /* Use 16 byte comparison. */
> > vmovdqu (%rdi), %xmm0
> > VPTESTM %xmm0, %xmm0, %k2
> > - VPCMP $0, (%rsi), %xmm0, %k1{%k2}
> > + CMP_R1_S2_XMM (%xmm0, (%rsi), %xmm1, %k1){%k2}
> > kmovd %k1, %ecx
> > # ifdef USE_AS_WCSCMP
> > subl $0xf, %ecx
> > @@ -1009,7 +1216,7 @@ L(less_1x_vec_till_page):
> > # endif
> > vmovdqu (%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm0
> > VPTESTM %xmm0, %xmm0, %k2
> > - VPCMP $0, (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm0, %k1{%k2}
> > + CMP_R1_S2_XMM (%xmm0, (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm1, %k1){%k2}
> > kmovd %k1, %ecx
> > # ifdef USE_AS_WCSCMP
> > subl $0xf, %ecx
> > @@ -1048,7 +1255,7 @@ L(less_16_till_page):
> > vmovq (%rdi), %xmm0
> > vmovq (%rsi), %xmm1
> > VPTESTM %xmm0, %xmm0, %k2
> > - VPCMP $0, %xmm1, %xmm0, %k1{%k2}
> > + CMP_R1_R2_XMM (%xmm0, %xmm1, %k1){%k2}
> > kmovd %k1, %ecx
> > # ifdef USE_AS_WCSCMP
> > subl $0x3, %ecx
> > @@ -1068,7 +1275,7 @@ L(less_16_till_page):
> > vmovq (%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm0
> > vmovq (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm1
> > VPTESTM %xmm0, %xmm0, %k2
> > - VPCMP $0, %xmm1, %xmm0, %k1{%k2}
> > + CMP_R1_R2_XMM (%xmm0, %xmm1, %k1){%k2}
> > kmovd %k1, %ecx
> > # ifdef USE_AS_WCSCMP
> > subl $0x3, %ecx
> > @@ -1128,7 +1335,7 @@ L(ret_less_8_wcs):
> > vmovd (%rdi), %xmm0
> > vmovd (%rsi), %xmm1
> > VPTESTM %xmm0, %xmm0, %k2
> > - VPCMP $0, %xmm1, %xmm0, %k1{%k2}
> > + CMP_R1_R2_XMM (%xmm0, %xmm1, %k1){%k2}
> > kmovd %k1, %ecx
> > subl $0xf, %ecx
> > jnz L(check_ret_vec_page_cross)
> > @@ -1143,7 +1350,7 @@ L(ret_less_8_wcs):
> > vmovd (%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm0
> > vmovd (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm1
> > VPTESTM %xmm0, %xmm0, %k2
> > - VPCMP $0, %xmm1, %xmm0, %k1{%k2}
> > + CMP_R1_R2_XMM (%xmm0, %xmm1, %k1){%k2}
> > kmovd %k1, %ecx
> > subl $0xf, %ecx
> > jnz L(check_ret_vec_page_cross)
> > @@ -1176,7 +1383,9 @@ L(less_4_till_page):
> > L(less_4_loop):
> > movzbl (%rdi), %eax
> > movzbl (%rsi, %rdi), %ecx
> > - subl %ecx, %eax
> > + TOLOWER_gpr (%rax, %eax)
> > + TOLOWER_gpr (%rcx, %BYTE_LOOP_REG)
> > + subl %BYTE_LOOP_REG, %eax
> > jnz L(ret_less_4_loop)
> > testl %ecx, %ecx
> > jz L(ret_zero_4_loop)
> > @@ -1203,5 +1412,6 @@ L(ret_less_4_loop):
> > subl %r8d, %eax
> > ret
> > # endif
> > -END(STRCMP)
> > + cfi_endproc
> > + .size STRCMP, .-STRCMP
> > #endif
> > diff --git a/sysdeps/x86_64/multiarch/strncase_l-evex.S b/sysdeps/x86_64/multiarch/strncase_l-evex.S
> > new file mode 100644
> > index 0000000000..8a5af3695c
> > --- /dev/null
> > +++ b/sysdeps/x86_64/multiarch/strncase_l-evex.S
> > @@ -0,0 +1,25 @@
> > +/* strncasecmp_l optimized with EVEX.
> > + Copyright (C) 2017-2022 Free Software Foundation, Inc.
> > + This file is part of the GNU C Library.
> > +
> > + The GNU C Library is free software; you can redistribute it and/or
> > + modify it under the terms of the GNU Lesser General Public
> > + License as published by the Free Software Foundation; either
> > + version 2.1 of the License, or (at your option) any later version.
> > +
> > + The GNU C Library is distributed in the hope that it will be useful,
> > + but WITHOUT ANY WARRANTY; without even the implied warranty of
> > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> > + Lesser General Public License for more details.
> > +
> > + You should have received a copy of the GNU Lesser General Public
> > + License along with the GNU C Library; if not, see
> > + <https://www.gnu.org/licenses/>. */
> > +
> > +#ifndef STRCMP
> > +# define STRCMP __strncasecmp_l_evex
> > +#endif
> > +#define OVERFLOW_STRCMP __strcasecmp_l_evex
> > +#define USE_AS_STRCASECMP_L
> > +#define USE_AS_STRNCMP
> > +#include "strcmp-evex.S"
> > --
> > 2.25.1
> >
>
> LGTM.
>
> Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
Thanks, pushed the patchset.
>
> Thanks.
>
> --
> H.J.
next prev parent reply other threads:[~2022-03-25 18:18 UTC|newest]
Thread overview: 76+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-03-23 21:57 [PATCH v1 01/23] benchtests: Use json-lib in bench-strchr.c Noah Goldstein
2022-03-23 21:57 ` [PATCH v1 02/23] benchtests: Add random benchmark " Noah Goldstein
2022-03-24 18:44 ` H.J. Lu
2022-03-23 21:57 ` [PATCH v1 03/23] x86: Code cleanup in strchr-avx2 and comment justifying branch Noah Goldstein
2022-03-24 18:53 ` H.J. Lu
2022-03-24 19:20 ` Noah Goldstein
2022-03-24 19:36 ` H.J. Lu
2022-05-12 19:31 ` Sunil Pandey
2022-03-23 21:57 ` [PATCH v1 04/23] x86: Code cleanup in strchr-evex " Noah Goldstein
2022-03-24 18:54 ` H.J. Lu
2022-05-12 19:32 ` Sunil Pandey
2022-03-23 21:57 ` [PATCH v1 05/23] benchtests: Use json-lib in bench-strpbrk.c Noah Goldstein
2022-03-24 18:54 ` H.J. Lu
2022-03-23 21:57 ` [PATCH v1 06/23] benchtests: Use json-lib in bench-strspn.c Noah Goldstein
2022-03-24 18:54 ` H.J. Lu
2022-03-23 21:57 ` [PATCH v1 07/23] x86: Optimize strcspn and strpbrk in strcspn-c.c Noah Goldstein
2022-03-24 18:55 ` H.J. Lu
2022-05-12 19:34 ` Sunil Pandey
2022-03-23 21:57 ` [PATCH v1 08/23] x86: Optimize strspn in strspn-c.c Noah Goldstein
2022-03-24 18:56 ` H.J. Lu
2022-05-12 19:39 ` Sunil Pandey
2022-03-23 21:57 ` [PATCH v1 09/23] x86: Remove strcspn-sse2.S and use the generic implementation Noah Goldstein
2022-03-24 18:57 ` H.J. Lu
2022-05-12 19:40 ` Sunil Pandey
2022-03-23 21:57 ` [PATCH v1 10/23] x86: Remove strpbrk-sse2.S " Noah Goldstein
2022-03-24 18:57 ` H.J. Lu
2022-05-12 19:41 ` Sunil Pandey
2022-03-23 21:57 ` [PATCH v1 11/23] x86: Remove strspn-sse2.S " Noah Goldstein
2022-03-24 18:57 ` H.J. Lu
2022-05-12 19:42 ` Sunil Pandey
2022-03-23 21:57 ` [PATCH v1 12/23] x86: Fix fallback for wcsncmp_avx2 in strcmp-avx2.S [BZ #28896] Noah Goldstein
2022-03-24 18:59 ` H.J. Lu
2022-03-24 19:18 ` Noah Goldstein
2022-03-24 19:34 ` H.J. Lu
2022-03-24 19:39 ` Noah Goldstein
2022-03-24 20:50 ` [PATCH v2 12/31] " Noah Goldstein
2022-03-24 21:26 ` H.J. Lu
2022-03-24 21:43 ` Noah Goldstein
2022-03-24 21:58 ` H.J. Lu
2022-05-04 6:05 ` Sunil Pandey
2022-03-23 21:57 ` [PATCH v1 13/23] benchtests: Use json-lib in bench-strcasecmp.c Noah Goldstein
2022-03-24 19:00 ` H.J. Lu
2022-03-23 21:57 ` [PATCH v1 14/23] benchtests: Use json-lib in bench-strncasecmp.c Noah Goldstein
2022-03-24 19:00 ` H.J. Lu
2022-03-23 21:57 ` [PATCH v1 15/23] string: Expand page cross tests in test-strcasecmp.c Noah Goldstein
2022-03-24 19:01 ` H.J. Lu
2022-03-23 21:57 ` [PATCH v1 16/23] string: Expand page cross tests in test-strncasecmp.c Noah Goldstein
2022-03-24 19:01 ` H.J. Lu
2022-03-23 21:57 ` [PATCH v1 17/23] x86: Optimize str{n}casecmp TOLOWER logic in strcmp.S Noah Goldstein
2022-03-24 19:02 ` H.J. Lu
2022-05-12 19:44 ` Sunil Pandey
2022-03-23 21:57 ` [PATCH v1 18/23] x86: Optimize str{n}casecmp TOLOWER logic in strcmp-sse42.S Noah Goldstein
2022-03-24 19:02 ` H.J. Lu
2022-05-12 19:45 ` Sunil Pandey
2022-03-23 21:57 ` [PATCH v1 19/23] string: Expand page cross test cases in test-strcmp.c Noah Goldstein
2022-03-24 19:02 ` H.J. Lu
2022-03-23 21:57 ` [PATCH v1 20/23] string: Expand page cross test cases in test-strncmp.c Noah Goldstein
2022-03-24 19:02 ` H.J. Lu
2022-03-23 21:57 ` [PATCH v1 21/23] x86: Add AVX2 optimized str{n}casecmp Noah Goldstein
2022-03-24 19:03 ` H.J. Lu
2022-03-24 22:41 ` [PATCH v3 " Noah Goldstein
2022-03-24 22:41 ` [PATCH v3 22/23] x86: Add EVEX " Noah Goldstein
2022-03-24 23:56 ` [PATCH v4 21/23] x86: Add AVX2 " Noah Goldstein
2022-03-24 23:56 ` [PATCH v4 22/23] x86: Add EVEX " Noah Goldstein
2022-03-25 18:15 ` H.J. Lu
2022-03-25 18:18 ` Noah Goldstein [this message]
2022-05-12 19:47 ` Sunil Pandey
2022-05-12 19:52 ` Sunil Pandey
2022-03-25 18:14 ` [PATCH v4 21/23] x86: Add AVX2 " H.J. Lu
2022-05-12 19:52 ` Sunil Pandey
2022-03-23 21:57 ` [PATCH v1 22/23] x86: Add EVEX " Noah Goldstein
2022-03-24 19:04 ` H.J. Lu
2022-03-23 21:57 ` [PATCH v1 23/23] x86: Remove AVX str{n}casecmp Noah Goldstein
2022-03-24 19:04 ` H.J. Lu
2022-05-12 19:54 ` Sunil Pandey
2022-03-24 18:43 ` [PATCH v1 01/23] benchtests: Use json-lib in bench-strchr.c H.J. Lu
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=CAFUsyf+FsB7x-O-FB87vsxNF5XaXJ8+CxOr-sgaO+5PbXcH6qQ@mail.gmail.com \
--to=goldstein.w.n@gmail.com \
--cc=carlos@systemhalted.org \
--cc=hjl.tools@gmail.com \
--cc=libc-alpha@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: link
Be 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).