From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-yb1-xb31.google.com (mail-yb1-xb31.google.com [IPv6:2607:f8b0:4864:20::b31]) by sourceware.org (Postfix) with ESMTPS id 656F53858CDA for ; Tue, 10 Jan 2023 07:41:11 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org 656F53858CDA Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=gmail.com Received: by mail-yb1-xb31.google.com with SMTP id 203so10937825yby.10 for ; Mon, 09 Jan 2023 23:41:11 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112; h=cc:to:subject:message-id:date:from:in-reply-to:references :mime-version:from:to:cc:subject:date:message-id:reply-to; bh=bY34Do7pMDcAEriKV8fKdJEO7QwXx44Soi7qq2jialI=; b=A6WgtMZB/JdaGITsEfAsw6wR7SaruC6J2+7KsMqTrfyzxkm+8KS3IItdPJFudhfuFf lN62XEl9rGLvhIXP4p4DhRTYbaFN4khBnkKuvr6q1ziHuz7WQ/3qNgOpN7m36PTJEyQh 8HWvwK7K48AEHy1uo8fw/KMejNOWlAxE7/SMshIdUgjDQTRok2CyqNJZFURNocN+QR/7 kQToLtG13mvAhat+agDxH3WktvnAFn3yuDtaoTwUU6XFiqlBtnSv20X5wkAyZnMJZPCk vfngrHpyUgVYIwQkORT6bGpd3gPs1HPuLBAeJtJMC0wdedefqNSzxXL5n5ZgZ3MlxSQr xwFA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=cc:to:subject:message-id:date:from:in-reply-to:references :mime-version:x-gm-message-state:from:to:cc:subject:date:message-id :reply-to; bh=bY34Do7pMDcAEriKV8fKdJEO7QwXx44Soi7qq2jialI=; b=pJsP+cfKB73xqbfVn2k87to9XSBheamrjnzyvUUnYNDzmVmm8kzY5QyapHTLzRxM9S 6NkXusXrwwGRgmCwxPzWAFZM0GsRQDqtmmcsWHuyzC/t1AesMxkKEJLVTdi30EuWmc8o ERN1hyKwKqpFNlCriLGF9l9u0Z2eaROKZHsady521yZF5H1QaW64XuF30+aIsLGhvmtZ hbhqP4t43STX6NnZ7Ppjk41YFwyD20ZmsezBMI98fAVv1/Iuqre69YdQdmZC+fuqnhjG 0tt++XO/wQJ/LC9pPK34dgUtQY1l7RphnqqvipUKr5KrknipJwsy/czRV46dZrmLFk9F zANw== X-Gm-Message-State: AFqh2krYf0OmjNeITNtGcbpPFzsUW4tUkLCRx9917MwG3pOD4R/i0ACH ZAaeKmHLDpYbvuCFZv46a1JkHLSaLVbDtG0KltZSNciFFS9eQw== X-Google-Smtp-Source: AMrXdXtiG1G/8HHpKRvfSs7alZvQHIYR1DzPEWnD4fXMkuDs1j8C/1taqdytWBC3nZB7GosNsJYbHosm+QwvfVP2f78= X-Received: by 2002:a25:1604:0:b0:796:d8a1:79b4 with SMTP id 4-20020a251604000000b00796d8a179b4mr5094350ybw.296.1673336470643; Mon, 09 Jan 2023 23:41:10 -0800 (PST) MIME-Version: 1.0 References: In-Reply-To: From: Hongtao Liu Date: Tue, 10 Jan 2023 15:40:59 +0800 Message-ID: Subject: Re: More znver4 x86-tune flags To: Jan Hubicka Cc: gcc-patches@gcc.gnu.org Content-Type: text/plain; charset="UTF-8" X-Spam-Status: No, score=-7.5 required=5.0 tests=BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,FREEMAIL_FROM,GIT_PATCH_0,RCVD_IN_DNSWL_NONE,SPF_HELO_NONE,SPF_PASS,TXREP autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org List-Id: On Tue, Jan 10, 2023 at 12:32 PM Jan Hubicka via Gcc-patches wrote: > > > Hi, > this patch adds more tunes for zen4: > - new tunes for avx512 scater instructions. > In micro benchmarks these seems consistent loss compared to open-coded coe > - disable use of gather for zen4 > While these are win for a micro benchmarks (based on TSVC), enabling gather > is a loss for parest. So for now it seems safe to keep it off. > - disable pass to avoid FMA chains for znver4 since fmadd was optimized and does not seem > to cause regressions. > > Bootstrapped/regtested x86_64. > Honza > > * i386.cc (ix86_vectorize_builtin_scatter): Guard scatter by TARGET_USE_SCATTER. > * i386.h (TARGET_USE_SCATTER_2PARTS, TARGET_USE_SCATTER_4PARTS, TARGET_USE_SCATTER): New macros. > * x86-tune.def (TARGET_USE_SCATTER_2PARTS, TARGET_USE_SCATTER_4PARTS, TARGET_USE_SCATTER): New tunes. > (X86_TUNE_AVOID_256FMA_CHAINS, X86_TUNE_AVOID_512FMA_CHAINS): Disable for znver4. > (X86_TUNE_USE_GATHER): Disable for zen4. > diff --git a/gcc/config/i386/i386.cc b/gcc/config/i386/i386.cc > index de978d19063..9fb69f6c174 100644 > --- a/gcc/config/i386/i386.cc > +++ b/gcc/config/i386/i386.cc > @@ -19051,6 +19051,13 @@ ix86_vectorize_builtin_scatter (const_tree vectype, > if (!TARGET_AVX512F) > return NULL_TREE; > > + if (known_eq (TYPE_VECTOR_SUBPARTS (vectype), 2u) > + ? !TARGET_USE_SCATTER_2PARTS > + : (known_eq (TYPE_VECTOR_SUBPARTS (vectype), 4u) > + ? !TARGET_USE_SCATTER_4PARTS > + : !TARGET_USE_SCATTER)) > + return NULL_TREE; > + > if ((TREE_CODE (index_type) != INTEGER_TYPE > && !POINTER_TYPE_P (index_type)) > || (TYPE_MODE (index_type) != SImode > diff --git a/gcc/config/i386/i386.h b/gcc/config/i386/i386.h > index e6a603ed31a..cd7ed19e29c 100644 > --- a/gcc/config/i386/i386.h > +++ b/gcc/config/i386/i386.h > @@ -397,10 +397,16 @@ extern unsigned char ix86_tune_features[X86_TUNE_LAST]; > ix86_tune_features[X86_TUNE_AVOID_4BYTE_PREFIXES] > #define TARGET_USE_GATHER_2PARTS \ > ix86_tune_features[X86_TUNE_USE_GATHER_2PARTS] > +#define TARGET_USE_SCATTER_2PARTS \ > + ix86_tune_features[X86_TUNE_USE_SCATTER_2PARTS] > #define TARGET_USE_GATHER_4PARTS \ > ix86_tune_features[X86_TUNE_USE_GATHER_4PARTS] > +#define TARGET_USE_SCATTER_4PARTS \ > + ix86_tune_features[X86_TUNE_USE_SCATTER_4PARTS] > #define TARGET_USE_GATHER \ > ix86_tune_features[X86_TUNE_USE_GATHER] > +#define TARGET_USE_SCATTER \ > + ix86_tune_features[X86_TUNE_USE_SCATTER] > #define TARGET_FUSE_CMP_AND_BRANCH_32 \ > ix86_tune_features[X86_TUNE_FUSE_CMP_AND_BRANCH_32] > #define TARGET_FUSE_CMP_AND_BRANCH_64 \ > diff --git a/gcc/config/i386/x86-tune.def b/gcc/config/i386/x86-tune.def > index fae3b650434..7e9c7244fc0 100644 > --- a/gcc/config/i386/x86-tune.def > +++ b/gcc/config/i386/x86-tune.def > @@ -483,28 +483,43 @@ DEF_TUNE (X86_TUNE_AVOID_4BYTE_PREFIXES, "avoid_4byte_prefixes", > DEF_TUNE (X86_TUNE_USE_GATHER_2PARTS, "use_gather_2parts", > ~(m_ZNVER1 | m_ZNVER2 | m_ZNVER3 | m_ZNVER4 | m_ALDERLAKE | m_CORE_ATOM | m_GENERIC)) > > +/* X86_TUNE_USE_SCATTER_2PARTS: Use scater instructions for vectors with 2 > + elements. */ > +DEF_TUNE (X86_TUNE_USE_SCATTER_2PARTS, "use_scatter_2parts", > + ~(m_ZNVER4 | m_GENERIC)) > + > /* X86_TUNE_USE_GATHER_4PARTS: Use gather instructions for vectors with 4 > elements. */ > DEF_TUNE (X86_TUNE_USE_GATHER_4PARTS, "use_gather_4parts", > ~(m_ZNVER1 | m_ZNVER2 | m_ZNVER3 | m_ZNVER4 | m_ALDERLAKE | m_CORE_ATOM | m_GENERIC)) > > +/* X86_TUNE_USE_SCATTER_4PARTS: Use scater instructions for vectors with 4 > + elements. */ > +DEF_TUNE (X86_TUNE_USE_SCATTER_4PARTS, "use_scatter_4parts", > + ~(m_ZNVER4 | m_GENERIC)) > + > /* X86_TUNE_USE_GATHER: Use gather instructions for vectors with 8 or more > elements. */ > DEF_TUNE (X86_TUNE_USE_GATHER, "use_gather", > - ~(m_ZNVER1 | m_ZNVER2 | m_ALDERLAKE | m_CORE_ATOM | m_GENERIC)) > + ~(m_ZNVER1 | m_ZNVER2 | m_ZNVER4 | m_ALDERLAKE | m_CORE_ATOM | m_GENERIC)) > + > +/* X86_TUNE_USE_SCATTER: Use scater instructions for vectors with 8 or more > + elements. */ > +DEF_TUNE (X86_TUNE_USE_SCATTER, "use_scatter", > + ~(m_ZNVER4 | m_GENERIC)) > > /* X86_TUNE_AVOID_128FMA_CHAINS: Avoid creating loops with tight 128bit or > smaller FMA chain. */ > -DEF_TUNE (X86_TUNE_AVOID_128FMA_CHAINS, "avoid_fma_chains", m_ZNVER) > +DEF_TUNE (X86_TUNE_AVOID_128FMA_CHAINS, "avoid_fma_chains", m_ZNVER1 | m_ZNVER2 | m_ZNVER3) According to comments, it's *256bit or smaller*, so shouldn't avoid_fma_chains be implied by avoid_fma256_chains. > > /* X86_TUNE_AVOID_256FMA_CHAINS: Avoid creating loops with tight 256bit or > smaller FMA chain. */ > -DEF_TUNE (X86_TUNE_AVOID_256FMA_CHAINS, "avoid_fma256_chains", m_ZNVER2 | m_ZNVER3 | m_ZNVER4 > +DEF_TUNE (X86_TUNE_AVOID_256FMA_CHAINS, "avoid_fma256_chains", m_ZNVER2 | m_ZNVER3 > | m_ALDERLAKE | m_SAPPHIRERAPIDS | m_CORE_ATOM) > > /* X86_TUNE_AVOID_512FMA_CHAINS: Avoid creating loops with tight 512bit or > smaller FMA chain. */ > -DEF_TUNE (X86_TUNE_AVOID_512FMA_CHAINS, "avoid_fma512_chains", m_ZNVER4) > +DEF_TUNE (X86_TUNE_AVOID_512FMA_CHAINS, "avoid_fma512_chains", m_NONE) > > /* X86_TUNE_V2DF_REDUCTION_PREFER_PHADDPD: Prefer haddpd > for v2df vector reduction. */ -- BR, Hongtao