From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-pj1-x102f.google.com (mail-pj1-x102f.google.com [IPv6:2607:f8b0:4864:20::102f]) by sourceware.org (Postfix) with ESMTPS id E1F683858292 for ; Wed, 29 Jun 2022 19:31:05 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org E1F683858292 Received: by mail-pj1-x102f.google.com with SMTP id g16-20020a17090a7d1000b001ea9f820449so434592pjl.5 for ; Wed, 29 Jun 2022 12:31:05 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=5MRKTGfTLb4fn7fHbRuQupDIbkjWZdU4P9zzB9vkg6A=; b=0MgnnKGhIHUA+0bsvd80PGSorLnuWNX64FStJ23IXyhSiu1T2EdCSZoRoM7TLoGV4K 8h3hCRQ7832LeB4hvgIEnzxV5Gf5BPLBU+3ozgOLLMOmuHF2yWjXBM4ROopjNQI6tVVF AOzSMKdDQ+GsvUb1SxXLuwwy5vEwhuw2CLRdMYXskLvZP69BUg8Hv5Z+IF6Zi8rzAedC GDhcoVHdwqsPRvPW1cncQDWgaHyqJKsOBN2bPBs622e69BZZ6bE6AWTxWYGN8JhYE3ln WtTA2yAC4Hm3S/CzeeKS7hGhuBHh88RsAXlmtsVQLQFTosfRWxUsJ9EhtMmFcqoLwYzc NlpA== X-Gm-Message-State: AJIora/EGZBykMVOSaNh6RN2yaRCFHi4GPiBI4kDGO/Mel5MouFOQStt NEHkEdJKqahFONoFSR2Pp4BEsbO50RwhTGeAtR0= X-Google-Smtp-Source: AGRyM1uSsKP6RMNRm+Hb4tFTgDddceeot+jpD0OBKORWhzbXGjaPtjf9ikA1f9GruEqoIcY/qf/TQ70uwXfwTu/4yD4= X-Received: by 2002:a17:902:eb86:b0:16a:f36d:741a with SMTP id q6-20020a170902eb8600b0016af36d741amr10705922plg.149.1656531064626; Wed, 29 Jun 2022 12:31:04 -0700 (PDT) MIME-Version: 1.0 References: <20220628152735.17863-1-goldstein.w.n@gmail.com> <20220628152735.17863-3-goldstein.w.n@gmail.com> In-Reply-To: <20220628152735.17863-3-goldstein.w.n@gmail.com> From: "H.J. Lu" Date: Wed, 29 Jun 2022 12:30:28 -0700 Message-ID: Subject: Re: [PATCH v1 3/3] x86: Add support for building {w}memset{_chk} with explicit ISA level To: Noah Goldstein Cc: GNU C Library , "Carlos O'Donell" Content-Type: text/plain; charset="UTF-8" X-Spam-Status: No, score=-3025.0 required=5.0 tests=BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, FREEMAIL_FROM, GIT_PATCH_0, KAM_SHORT, RCVD_IN_DNSWL_NONE, SPF_HELO_NONE, SPF_PASS, TXREP, T_SCC_BODY_TEXT_LINE autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org X-BeenThere: libc-alpha@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Libc-alpha mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 29 Jun 2022 19:31:09 -0000 On Tue, Jun 28, 2022 at 8:27 AM Noah Goldstein wrote: > > 1. Refactor files so that all implementations are in the multiarch > directory > - Moved the implementation portion of memset sse2 from memset.S to > multiarch/memset-sse2.S > > - The non-multiarch file now only includes one of the > implementations in the multiarch directory based on the compiled > ISA level (only used for non-multiarch builds. Otherwise we go > through the ifunc selector). > > 2. Add ISA level build guards to different implementations. > - I.e memset-avx2-unaligned-erms.S which is ISA level 3 will only > build if compiled ISA level <= 3. Otherwise there is no reason > to include it as we will always use one of the ISA level 4 > implementations (memset-evex-unaligned-erms.S). > > 3. Add new multiarch/rtld-memset.S that just include the > non-multiarch memset.S which will in turn select the best > implementation based on the compiled ISA level. > > 4. Refactor the ifunc selector and ifunc implementation list to use > the ISA level aware wrapper macros that allow functions below the > compiled ISA level (with a guranteed replacement) to be skipped. > > Tested with and without multiarch on x86_64 for ISA levels: > {generic, x86-64-v2, x86-64-v3, x86-64-v4} > > And m32 with and without multiarch. > --- > sysdeps/x86_64/memset.S | 45 +---- > sysdeps/x86_64/multiarch/Makefile | 1 + > sysdeps/x86_64/multiarch/ifunc-impl-list.c | 165 +++++++++--------- > sysdeps/x86_64/multiarch/ifunc-memset.h | 45 ++--- > sysdeps/x86_64/multiarch/ifunc-wmemset.h | 21 ++- > .../multiarch/memset-avx2-unaligned-erms.S | 5 +- > .../multiarch/memset-avx512-no-vzeroupper.S | 4 +- > .../multiarch/memset-avx512-unaligned-erms.S | 12 +- > sysdeps/x86_64/multiarch/memset-erms.S | 25 +++ > .../multiarch/memset-evex-unaligned-erms.S | 12 +- > .../multiarch/memset-sse2-unaligned-erms.S | 57 ++++-- > .../multiarch/memset-vec-unaligned-erms.S | 23 --- > sysdeps/x86_64/multiarch/rtld-memset.S | 18 ++ > 13 files changed, 247 insertions(+), 186 deletions(-) > create mode 100644 sysdeps/x86_64/multiarch/memset-erms.S > create mode 100644 sysdeps/x86_64/multiarch/rtld-memset.S > > diff --git a/sysdeps/x86_64/memset.S b/sysdeps/x86_64/memset.S > index a6eea61a4d..f4e1bab601 100644 > --- a/sysdeps/x86_64/memset.S > +++ b/sysdeps/x86_64/memset.S > @@ -18,47 +18,18 @@ > . */ > > #include > -#define USE_WITH_SSE2 1 > > -#define VEC_SIZE 16 > -#define MOV_SIZE 3 > -#define RET_SIZE 1 > +#define MEMSET_SYMBOL(p,s) memset > +#define MEMSET_CHK_SYMBOL(p,s) p > > -#define VEC(i) xmm##i > -#define VMOVU movups > -#define VMOVA movaps > +#define WMEMSET_SYMBOL(p,s) __wmemset > +#define WMEMSET_CHK_SYMBOL(p,s) p > > -# define MEMSET_SET_VEC0_AND_SET_RETURN(d, r) \ > - movd d, %xmm0; \ > - movq r, %rax; \ > - punpcklbw %xmm0, %xmm0; \ > - punpcklwd %xmm0, %xmm0; \ > - pshufd $0, %xmm0, %xmm0 > +#define DEFAULT_IMPL_V1 "multiarch/memset-sse2-unaligned-erms.S" > +#define DEFAULT_IMPL_V3 "multiarch/memset-avx2-unaligned-erms.S" > +#define DEFAULT_IMPL_V4 "multiarch/memset-evex-unaligned-erms.S" > > -# define WMEMSET_SET_VEC0_AND_SET_RETURN(d, r) \ > - movd d, %xmm0; \ > - pshufd $0, %xmm0, %xmm0; \ > - movq r, %rax > - > -# define MEMSET_VDUP_TO_VEC0_HIGH() > -# define MEMSET_VDUP_TO_VEC0_LOW() > - > -# define WMEMSET_VDUP_TO_VEC0_HIGH() > -# define WMEMSET_VDUP_TO_VEC0_LOW() > - > -#define SECTION(p) p > - > -#ifndef MEMSET_SYMBOL > -# define MEMSET_CHK_SYMBOL(p,s) p > -# define MEMSET_SYMBOL(p,s) memset > -#endif > - > -#ifndef WMEMSET_SYMBOL > -# define WMEMSET_CHK_SYMBOL(p,s) p > -# define WMEMSET_SYMBOL(p,s) __wmemset > -#endif > - > -#include "multiarch/memset-vec-unaligned-erms.S" > +#include "isa-default-impl.h" > > libc_hidden_builtin_def (memset) > > diff --git a/sysdeps/x86_64/multiarch/Makefile b/sysdeps/x86_64/multiarch/Makefile > index 666ee4d5d6..0525cef3fe 100644 > --- a/sysdeps/x86_64/multiarch/Makefile > +++ b/sysdeps/x86_64/multiarch/Makefile > @@ -29,6 +29,7 @@ sysdep_routines += \ > memset-avx2-unaligned-erms-rtm \ > memset-avx512-no-vzeroupper \ > memset-avx512-unaligned-erms \ > + memset-erms \ This change should be the part of the other memset-erms patch. > memset-evex-unaligned-erms \ > memset-sse2-unaligned-erms \ > rawmemchr-avx2 \ > diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c > index 3ae4d49bee..4450e2c593 100644 > --- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c > +++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c > @@ -209,94 +209,97 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, > IFUNC_IMPL (i, name, __memset_chk, > IFUNC_IMPL_ADD (array, i, __memset_chk, 1, > __memset_chk_erms) > - IFUNC_IMPL_ADD (array, i, __memset_chk, 1, > - __memset_chk_sse2_unaligned) > - IFUNC_IMPL_ADD (array, i, __memset_chk, 1, > - __memset_chk_sse2_unaligned_erms) > - IFUNC_IMPL_ADD (array, i, __memset_chk, > + X86_IFUNC_IMPL_ADD_V4 (array, i, __memset_chk, > + (CPU_FEATURE_USABLE (AVX512VL) > + && CPU_FEATURE_USABLE (AVX512BW) > + && CPU_FEATURE_USABLE (BMI2)), > + __memset_chk_avx512_unaligned_erms) > + X86_IFUNC_IMPL_ADD_V4 (array, i, __memset_chk, > + (CPU_FEATURE_USABLE (AVX512VL) > + && CPU_FEATURE_USABLE (AVX512BW) > + && CPU_FEATURE_USABLE (BMI2)), > + __memset_chk_avx512_unaligned) > + X86_IFUNC_IMPL_ADD_V4 (array, i, __memset_chk, > + CPU_FEATURE_USABLE (AVX512F), > + __memset_chk_avx512_no_vzeroupper) > + X86_IFUNC_IMPL_ADD_V4 (array, i, __memset_chk, > + (CPU_FEATURE_USABLE (AVX512VL) > + && CPU_FEATURE_USABLE (AVX512BW) > + && CPU_FEATURE_USABLE (BMI2)), > + __memset_chk_evex_unaligned) > + X86_IFUNC_IMPL_ADD_V4 (array, i, __memset_chk, > + (CPU_FEATURE_USABLE (AVX512VL) > + && CPU_FEATURE_USABLE (AVX512BW) > + && CPU_FEATURE_USABLE (BMI2)), > + __memset_chk_evex_unaligned_erms) > + X86_IFUNC_IMPL_ADD_V3 (array, i, __memset_chk, > CPU_FEATURE_USABLE (AVX2), > __memset_chk_avx2_unaligned) > - IFUNC_IMPL_ADD (array, i, __memset_chk, > + X86_IFUNC_IMPL_ADD_V3 (array, i, __memset_chk, > CPU_FEATURE_USABLE (AVX2), > __memset_chk_avx2_unaligned_erms) > - IFUNC_IMPL_ADD (array, i, __memset_chk, > + X86_IFUNC_IMPL_ADD_V3 (array, i, __memset_chk, > (CPU_FEATURE_USABLE (AVX2) > && CPU_FEATURE_USABLE (RTM)), > __memset_chk_avx2_unaligned_rtm) > - IFUNC_IMPL_ADD (array, i, __memset_chk, > + X86_IFUNC_IMPL_ADD_V3 (array, i, __memset_chk, > (CPU_FEATURE_USABLE (AVX2) > && CPU_FEATURE_USABLE (RTM)), > __memset_chk_avx2_unaligned_erms_rtm) > - IFUNC_IMPL_ADD (array, i, __memset_chk, > + /* Can be lowered to V1 if a V2 implementation is added. */ > + X86_IFUNC_IMPL_ADD_V2 (array, i, __memset_chk, 1, > + __memset_chk_sse2_unaligned) > + X86_IFUNC_IMPL_ADD_V2 (array, i, __memset_chk, 1, > + __memset_chk_sse2_unaligned_erms) > + ) > +#endif > + > + /* Support sysdeps/x86_64/multiarch/memset.c. */ > + IFUNC_IMPL (i, name, memset, > + IFUNC_IMPL_ADD (array, i, memset, 1, > + __memset_erms) > + X86_IFUNC_IMPL_ADD_V4 (array, i, memset, > (CPU_FEATURE_USABLE (AVX512VL) > && CPU_FEATURE_USABLE (AVX512BW) > && CPU_FEATURE_USABLE (BMI2)), > - __memset_chk_evex_unaligned) > - IFUNC_IMPL_ADD (array, i, __memset_chk, > + __memset_avx512_unaligned_erms) > + X86_IFUNC_IMPL_ADD_V4 (array, i, memset, > (CPU_FEATURE_USABLE (AVX512VL) > && CPU_FEATURE_USABLE (AVX512BW) > && CPU_FEATURE_USABLE (BMI2)), > - __memset_chk_evex_unaligned_erms) > - IFUNC_IMPL_ADD (array, i, __memset_chk, > + __memset_avx512_unaligned) > + X86_IFUNC_IMPL_ADD_V4 (array, i, memset, > + CPU_FEATURE_USABLE (AVX512F), > + __memset_avx512_no_vzeroupper) > + X86_IFUNC_IMPL_ADD_V4 (array, i, memset, > (CPU_FEATURE_USABLE (AVX512VL) > && CPU_FEATURE_USABLE (AVX512BW) > && CPU_FEATURE_USABLE (BMI2)), > - __memset_chk_avx512_unaligned_erms) > - IFUNC_IMPL_ADD (array, i, __memset_chk, > + __memset_evex_unaligned) > + X86_IFUNC_IMPL_ADD_V4 (array, i, memset, > (CPU_FEATURE_USABLE (AVX512VL) > && CPU_FEATURE_USABLE (AVX512BW) > && CPU_FEATURE_USABLE (BMI2)), > - __memset_chk_avx512_unaligned) > - IFUNC_IMPL_ADD (array, i, __memset_chk, > - CPU_FEATURE_USABLE (AVX512F), > - __memset_chk_avx512_no_vzeroupper) > - ) > -#endif > - > - /* Support sysdeps/x86_64/multiarch/memset.c. */ > - IFUNC_IMPL (i, name, memset, > - IFUNC_IMPL_ADD (array, i, memset, 1, > - __memset_sse2_unaligned) > - IFUNC_IMPL_ADD (array, i, memset, 1, > - __memset_sse2_unaligned_erms) > - IFUNC_IMPL_ADD (array, i, memset, 1, __memset_erms) > - IFUNC_IMPL_ADD (array, i, memset, > + __memset_evex_unaligned_erms) > + X86_IFUNC_IMPL_ADD_V3 (array, i, memset, > CPU_FEATURE_USABLE (AVX2), > __memset_avx2_unaligned) > - IFUNC_IMPL_ADD (array, i, memset, > + X86_IFUNC_IMPL_ADD_V3 (array, i, memset, > CPU_FEATURE_USABLE (AVX2), > __memset_avx2_unaligned_erms) > - IFUNC_IMPL_ADD (array, i, memset, > + X86_IFUNC_IMPL_ADD_V3 (array, i, memset, > (CPU_FEATURE_USABLE (AVX2) > && CPU_FEATURE_USABLE (RTM)), > __memset_avx2_unaligned_rtm) > - IFUNC_IMPL_ADD (array, i, memset, > + X86_IFUNC_IMPL_ADD_V3 (array, i, memset, > (CPU_FEATURE_USABLE (AVX2) > && CPU_FEATURE_USABLE (RTM)), > __memset_avx2_unaligned_erms_rtm) > - IFUNC_IMPL_ADD (array, i, memset, > - (CPU_FEATURE_USABLE (AVX512VL) > - && CPU_FEATURE_USABLE (AVX512BW) > - && CPU_FEATURE_USABLE (BMI2)), > - __memset_evex_unaligned) > - IFUNC_IMPL_ADD (array, i, memset, > - (CPU_FEATURE_USABLE (AVX512VL) > - && CPU_FEATURE_USABLE (AVX512BW) > - && CPU_FEATURE_USABLE (BMI2)), > - __memset_evex_unaligned_erms) > - IFUNC_IMPL_ADD (array, i, memset, > - (CPU_FEATURE_USABLE (AVX512VL) > - && CPU_FEATURE_USABLE (AVX512BW) > - && CPU_FEATURE_USABLE (BMI2)), > - __memset_avx512_unaligned_erms) > - IFUNC_IMPL_ADD (array, i, memset, > - (CPU_FEATURE_USABLE (AVX512VL) > - && CPU_FEATURE_USABLE (AVX512BW) > - && CPU_FEATURE_USABLE (BMI2)), > - __memset_avx512_unaligned) > - IFUNC_IMPL_ADD (array, i, memset, > - CPU_FEATURE_USABLE (AVX512F), > - __memset_avx512_no_vzeroupper) > + /* Can be lowered to V1 if a V2 implementation is added. */ > + X86_IFUNC_IMPL_ADD_V2 (array, i, memset, 1, > + __memset_sse2_unaligned) > + X86_IFUNC_IMPL_ADD_V2 (array, i, memset, 1, > + __memset_sse2_unaligned_erms) > ) > > /* Support sysdeps/x86_64/multiarch/rawmemchr.c. */ > @@ -820,25 +823,26 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, > > /* Support sysdeps/x86_64/multiarch/wmemset.c. */ > IFUNC_IMPL (i, name, wmemset, > - IFUNC_IMPL_ADD (array, i, wmemset, 1, > - __wmemset_sse2_unaligned) > - IFUNC_IMPL_ADD (array, i, wmemset, > - CPU_FEATURE_USABLE (AVX2), > - __wmemset_avx2_unaligned) > - IFUNC_IMPL_ADD (array, i, wmemset, > - (CPU_FEATURE_USABLE (AVX2) > - && CPU_FEATURE_USABLE (RTM)), > - __wmemset_avx2_unaligned_rtm) > - IFUNC_IMPL_ADD (array, i, wmemset, > + X86_IFUNC_IMPL_ADD_V4 (array, i, wmemset, > (CPU_FEATURE_USABLE (AVX512VL) > && CPU_FEATURE_USABLE (AVX512BW) > && CPU_FEATURE_USABLE (BMI2)), > __wmemset_evex_unaligned) > - IFUNC_IMPL_ADD (array, i, wmemset, > + X86_IFUNC_IMPL_ADD_V4 (array, i, wmemset, > (CPU_FEATURE_USABLE (AVX512VL) > && CPU_FEATURE_USABLE (AVX512BW) > && CPU_FEATURE_USABLE (BMI2)), > - __wmemset_avx512_unaligned)) > + __wmemset_avx512_unaligned) > + X86_IFUNC_IMPL_ADD_V3 (array, i, wmemset, > + CPU_FEATURE_USABLE (AVX2), > + __wmemset_avx2_unaligned) > + X86_IFUNC_IMPL_ADD_V3 (array, i, wmemset, > + (CPU_FEATURE_USABLE (AVX2) > + && CPU_FEATURE_USABLE (RTM)), > + __wmemset_avx2_unaligned_rtm) > + /* Can be lowered to V1 if a V2 implementation is added. */ > + X86_IFUNC_IMPL_ADD_V2 (array, i, wmemset, 1, > + __wmemset_sse2_unaligned)) > > #ifdef SHARED > /* Support sysdeps/x86_64/multiarch/memcpy_chk.c. */ > @@ -1023,25 +1027,26 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, > #ifdef SHARED > /* Support sysdeps/x86_64/multiarch/wmemset_chk.c. */ > IFUNC_IMPL (i, name, __wmemset_chk, > - IFUNC_IMPL_ADD (array, i, __wmemset_chk, 1, > - __wmemset_chk_sse2_unaligned) > - IFUNC_IMPL_ADD (array, i, __wmemset_chk, > - CPU_FEATURE_USABLE (AVX2), > - __wmemset_chk_avx2_unaligned) > - IFUNC_IMPL_ADD (array, i, wmemset_chk, > - (CPU_FEATURE_USABLE (AVX2) > - && CPU_FEATURE_USABLE (RTM)), > - __wmemset_chk_avx2_unaligned_rtm) > - IFUNC_IMPL_ADD (array, i, __wmemset_chk, > + X86_IFUNC_IMPL_ADD_V4 (array, i, __wmemset_chk, > (CPU_FEATURE_USABLE (AVX512VL) > && CPU_FEATURE_USABLE (AVX512BW) > && CPU_FEATURE_USABLE (BMI2)), > __wmemset_chk_evex_unaligned) > - IFUNC_IMPL_ADD (array, i, __wmemset_chk, > + X86_IFUNC_IMPL_ADD_V4 (array, i, __wmemset_chk, > (CPU_FEATURE_USABLE (AVX512VL) > && CPU_FEATURE_USABLE (AVX512BW) > && CPU_FEATURE_USABLE (BMI2)), > - __wmemset_chk_avx512_unaligned)) > + __wmemset_chk_avx512_unaligned) > + X86_IFUNC_IMPL_ADD_V3 (array, i, __wmemset_chk, > + CPU_FEATURE_USABLE (AVX2), > + __wmemset_chk_avx2_unaligned) > + X86_IFUNC_IMPL_ADD_V3 (array, i, __wmemset_chk, > + (CPU_FEATURE_USABLE (AVX2) > + && CPU_FEATURE_USABLE (RTM)), > + __wmemset_chk_avx2_unaligned_rtm) > + /* Can be lowered to V1 if a V2 implementation is added. */ > + X86_IFUNC_IMPL_ADD_V2 (array, i, __wmemset_chk, 1, > + __wmemset_chk_sse2_unaligned)) > #endif > > return 0; > diff --git a/sysdeps/x86_64/multiarch/ifunc-memset.h b/sysdeps/x86_64/multiarch/ifunc-memset.h > index 64d179913c..ed514976aa 100644 > --- a/sysdeps/x86_64/multiarch/ifunc-memset.h > +++ b/sysdeps/x86_64/multiarch/ifunc-memset.h > @@ -20,10 +20,19 @@ > #include > > extern __typeof (REDIRECT_NAME) OPTIMIZE (erms) attribute_hidden; > -extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned) > + > +extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned) > attribute_hidden; > -extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned_erms) > +extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned_erms) > + attribute_hidden; > +extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_no_vzeroupper) > + attribute_hidden; > + > +extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned) > + attribute_hidden; > +extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned_erms) > attribute_hidden; > + > extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned) attribute_hidden; > extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned_erms) > attribute_hidden; > @@ -31,31 +40,26 @@ extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned_rtm) > attribute_hidden; > extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned_erms_rtm) > attribute_hidden; > -extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned) > - attribute_hidden; > -extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned_erms) > - attribute_hidden; > -extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned) > - attribute_hidden; > -extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned_erms) > + > +extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned) > attribute_hidden; > -extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_no_vzeroupper) > +extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned_erms) > attribute_hidden; > > static inline void * > IFUNC_SELECTOR (void) > { > - const struct cpu_features* cpu_features = __get_cpu_features (); > + const struct cpu_features *cpu_features = __get_cpu_features (); > > if (CPU_FEATURES_ARCH_P (cpu_features, Prefer_ERMS)) > return OPTIMIZE (erms); > > - if (CPU_FEATURE_USABLE_P (cpu_features, AVX512F) > + if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512F) > && !CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512)) > { > - if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL) > - && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW) > - && CPU_FEATURE_USABLE_P (cpu_features, BMI2)) > + if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512VL) > + && X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512BW) > + && X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, BMI2)) > { > if (CPU_FEATURE_USABLE_P (cpu_features, ERMS)) > return OPTIMIZE (avx512_unaligned_erms); > @@ -66,11 +70,11 @@ IFUNC_SELECTOR (void) > return OPTIMIZE (avx512_no_vzeroupper); > } > > - if (CPU_FEATURE_USABLE_P (cpu_features, AVX2)) > + if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX2)) > { > - if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL) > - && CPU_FEATURE_USABLE_P (cpu_features, AVX512BW) > - && CPU_FEATURE_USABLE_P (cpu_features, BMI2)) > + if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512VL) > + && X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512BW) > + && X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, BMI2)) > { > if (CPU_FEATURE_USABLE_P (cpu_features, ERMS)) > return OPTIMIZE (evex_unaligned_erms); > @@ -86,7 +90,8 @@ IFUNC_SELECTOR (void) > return OPTIMIZE (avx2_unaligned_rtm); > } > > - if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER)) > + if (X86_ISA_CPU_FEATURES_ARCH_P (cpu_features, > + Prefer_No_VZEROUPPER, !)) > { > if (CPU_FEATURE_USABLE_P (cpu_features, ERMS)) > return OPTIMIZE (avx2_unaligned_erms); > diff --git a/sysdeps/x86_64/multiarch/ifunc-wmemset.h b/sysdeps/x86_64/multiarch/ifunc-wmemset.h > index 87c48e2387..3810c719c6 100644 > --- a/sysdeps/x86_64/multiarch/ifunc-wmemset.h > +++ b/sysdeps/x86_64/multiarch/ifunc-wmemset.h > @@ -18,22 +18,26 @@ > > #include > > -extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned) attribute_hidden; > +extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned) attribute_hidden; > + > +extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned) attribute_hidden; > + > extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned) attribute_hidden; > extern __typeof (REDIRECT_NAME) OPTIMIZE (avx2_unaligned_rtm) > attribute_hidden; > -extern __typeof (REDIRECT_NAME) OPTIMIZE (evex_unaligned) attribute_hidden; > -extern __typeof (REDIRECT_NAME) OPTIMIZE (avx512_unaligned) attribute_hidden; > + > +extern __typeof (REDIRECT_NAME) OPTIMIZE (sse2_unaligned) attribute_hidden; > > static inline void * > IFUNC_SELECTOR (void) > { > - const struct cpu_features* cpu_features = __get_cpu_features (); > + const struct cpu_features *cpu_features = __get_cpu_features (); > > - if (CPU_FEATURE_USABLE_P (cpu_features, AVX2) > - && CPU_FEATURES_ARCH_P (cpu_features, AVX_Fast_Unaligned_Load)) > + if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX2) > + && X86_ISA_CPU_FEATURES_ARCH_P (cpu_features, > + AVX_Fast_Unaligned_Load, !)) > { > - if (CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)) > + if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)) > { > if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_AVX512)) > return OPTIMIZE (avx512_unaligned); > @@ -44,7 +48,8 @@ IFUNC_SELECTOR (void) > if (CPU_FEATURE_USABLE_P (cpu_features, RTM)) > return OPTIMIZE (avx2_unaligned_rtm); > > - if (!CPU_FEATURES_ARCH_P (cpu_features, Prefer_No_VZEROUPPER)) > + if (X86_ISA_CPU_FEATURES_ARCH_P (cpu_features, > + Prefer_No_VZEROUPPER, !)) > return OPTIMIZE (avx2_unaligned); > } > > diff --git a/sysdeps/x86_64/multiarch/memset-avx2-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-avx2-unaligned-erms.S > index c0bf2875d0..a9054a9122 100644 > --- a/sysdeps/x86_64/multiarch/memset-avx2-unaligned-erms.S > +++ b/sysdeps/x86_64/multiarch/memset-avx2-unaligned-erms.S > @@ -1,4 +1,7 @@ > -#if IS_IN (libc) > +#include > + > +#if ISA_SHOULD_BUILD (3) > + > # define USE_WITH_AVX2 1 > > # define VEC_SIZE 32 > diff --git a/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S b/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S > index c5be8f57ef..8cc9c16d73 100644 > --- a/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S > +++ b/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S > @@ -17,8 +17,10 @@ > . */ > > #include > +#include > + > +#if ISA_SHOULD_BUILD (4) > > -#if IS_IN (libc) > > #include "asm-syntax.h" > #ifndef MEMSET > diff --git a/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S > index 5241216a77..47623b8ee8 100644 > --- a/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S > +++ b/sysdeps/x86_64/multiarch/memset-avx512-unaligned-erms.S > @@ -1,4 +1,7 @@ > -#if IS_IN (libc) > +#include > + > +#if ISA_SHOULD_BUILD (4) > + > # define USE_WITH_AVX512 1 > > # define VEC_SIZE 64 > @@ -30,8 +33,15 @@ > # define WMEMSET_VDUP_TO_VEC0_LOW() > > # define SECTION(p) p##.evex512 > + > +#ifndef MEMSET_SYMBOL > # define MEMSET_SYMBOL(p,s) p##_avx512_##s > +#endif > +#ifndef WMEMSET_SYMBOL > # define WMEMSET_SYMBOL(p,s) p##_avx512_##s > +#endif > + > + > # define USE_LESS_VEC_MASK_STORE 1 > # include "memset-vec-unaligned-erms.S" > #endif > diff --git a/sysdeps/x86_64/multiarch/memset-erms.S b/sysdeps/x86_64/multiarch/memset-erms.S > new file mode 100644 > index 0000000000..1fce0c9fcc > --- /dev/null > +++ b/sysdeps/x86_64/multiarch/memset-erms.S > @@ -0,0 +1,25 @@ > +#include > + > +#if defined USE_MULTIARCH && IS_IN (libc) > + .text > +ENTRY (__memset_chk_erms) > + cmp %RDX_LP, %RCX_LP > + jb HIDDEN_JUMPTARGET (__chk_fail) > +END (__memset_chk_erms) > + > +/* Only used to measure performance of REP STOSB. */ > +ENTRY (__memset_erms) > + /* Skip zero length. */ > + test %RDX_LP, %RDX_LP > + jz L(stosb_return_zero) > + mov %RDX_LP, %RCX_LP > + movzbl %sil, %eax > + mov %RDI_LP, %RDX_LP > + rep stosb > + mov %RDX_LP, %RAX_LP > + ret > +L(stosb_return_zero): > + movq %rdi, %rax > + ret > +END (__memset_erms) > +#endif > diff --git a/sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S > index 6370021506..ac4b2d2d50 100644 > --- a/sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S > +++ b/sysdeps/x86_64/multiarch/memset-evex-unaligned-erms.S > @@ -1,4 +1,7 @@ > -#if IS_IN (libc) > +#include > + > +#if ISA_SHOULD_BUILD (4) > + > # define USE_WITH_EVEX 1 > > # define VEC_SIZE 32 > @@ -30,8 +33,15 @@ > # define WMEMSET_VDUP_TO_VEC0_LOW() > > # define SECTION(p) p##.evex > + > +#ifndef MEMSET_SYMBOL > # define MEMSET_SYMBOL(p,s) p##_evex_##s > +#endif > +#ifndef WMEMSET_SYMBOL > # define WMEMSET_SYMBOL(p,s) p##_evex_##s > +#endif > + > + > # define USE_LESS_VEC_MASK_STORE 1 > # include "memset-vec-unaligned-erms.S" > #endif > diff --git a/sysdeps/x86_64/multiarch/memset-sse2-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-sse2-unaligned-erms.S > index 3d92f6993a..44f9b8888b 100644 > --- a/sysdeps/x86_64/multiarch/memset-sse2-unaligned-erms.S > +++ b/sysdeps/x86_64/multiarch/memset-sse2-unaligned-erms.S > @@ -17,22 +17,51 @@ > License along with the GNU C Library; if not, see > . */ > > -#include > -#include > +#include > > -#if IS_IN (libc) > -# define MEMSET_SYMBOL(p,s) p##_sse2_##s > -# define WMEMSET_SYMBOL(p,s) p##_sse2_##s > +/* MINIMUM_X86_ISA_LEVEL <= 2 because there is no V2 implementation > + so we need this to build for ISA V2 builds. */ > +#if ISA_SHOULD_BUILD (2) > > -# ifdef SHARED > -# undef libc_hidden_builtin_def > -# define libc_hidden_builtin_def(name) > +# include > +# define USE_WITH_SSE2 1 > + > +# define VEC_SIZE 16 > +# define MOV_SIZE 3 > +# define RET_SIZE 1 > + > +# define VEC(i) xmm##i > +# define VMOVU movups > +# define VMOVA movaps > + > +# define MEMSET_SET_VEC0_AND_SET_RETURN(d, r) \ > + movd d, %xmm0; \ > + movq r, %rax; \ > + punpcklbw %xmm0, %xmm0; \ > + punpcklwd %xmm0, %xmm0; \ > + pshufd $0, %xmm0, %xmm0 > + > +# define WMEMSET_SET_VEC0_AND_SET_RETURN(d, r) \ > + movd d, %xmm0; \ > + pshufd $0, %xmm0, %xmm0; \ > + movq r, %rax > + > +# define MEMSET_VDUP_TO_VEC0_HIGH() > +# define MEMSET_VDUP_TO_VEC0_LOW() > + > +# define WMEMSET_VDUP_TO_VEC0_HIGH() > +# define WMEMSET_VDUP_TO_VEC0_LOW() > + > +# define SECTION(p) p > + > +# ifndef MEMSET_SYMBOL > +# define MEMSET_SYMBOL(p,s) p##_sse2_##s > # endif > > -# undef weak_alias > -# define weak_alias(original, alias) > -# undef strong_alias > -# define strong_alias(ignored1, ignored2) > -#endif > +# ifndef WMEMSET_SYMBOL > +# define WMEMSET_SYMBOL(p,s) p##_sse2_##s > +# endif > + > +# include "memset-vec-unaligned-erms.S" > > -#include > +#endif > diff --git a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S > index d98c613651..905d0fa464 100644 > --- a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S > +++ b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S > @@ -430,26 +430,3 @@ L(between_2_3): > #endif > ret > END (MEMSET_SYMBOL (__memset, unaligned_erms)) > - > -#if defined USE_MULTIARCH && IS_IN (libc) && VEC_SIZE == 16 > -ENTRY (__memset_chk_erms) > - cmp %RDX_LP, %RCX_LP > - jb HIDDEN_JUMPTARGET (__chk_fail) > -END (__memset_chk_erms) > - > -/* Only used to measure performance of REP STOSB. */ > -ENTRY (__memset_erms) > - /* Skip zero length. */ > - test %RDX_LP, %RDX_LP > - jz L(stosb_return_zero) > - mov %RDX_LP, %RCX_LP > - movzbl %sil, %eax > - mov %RDI_LP, %RDX_LP > - rep stosb > - mov %RDX_LP, %RAX_LP > - ret > -L(stosb_return_zero): > - movq %rdi, %rax > - ret > -END (__memset_erms) > -#endif > diff --git a/sysdeps/x86_64/multiarch/rtld-memset.S b/sysdeps/x86_64/multiarch/rtld-memset.S > new file mode 100644 > index 0000000000..d912bfa7cc > --- /dev/null > +++ b/sysdeps/x86_64/multiarch/rtld-memset.S > @@ -0,0 +1,18 @@ > +/* Copyright (C) 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 > + . */ > + > +#include "../memset.S" > -- > 2.34.1 > -- H.J.