From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from nikam.ms.mff.cuni.cz (nikam.ms.mff.cuni.cz [195.113.20.16]) by sourceware.org (Postfix) with ESMTPS id 09CFB3858CDA for ; Tue, 10 Jan 2023 04:31:35 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org 09CFB3858CDA Authentication-Results: sourceware.org; dmarc=fail (p=none dis=none) header.from=ucw.cz Authentication-Results: sourceware.org; spf=none smtp.mailfrom=kam.mff.cuni.cz Received: by nikam.ms.mff.cuni.cz (Postfix, from userid 16202) id 342CB284ACA; Tue, 10 Jan 2023 05:31:34 +0100 (CET) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=ucw.cz; s=gen1; t=1673325094; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:mime-version:mime-version:content-type:content-type; bh=o5IYj+qpV2IwSUrjCBIPQ/qZXP2CT4l+YBOVTUAQ0JY=; b=CQxx/RtybHWYGNNwQOIsCtcek2bbn6OgsXMvEhEaC95We+KzxDw0BFZrbPiH3qjEIxkycf mCylJUTq2FrmMgEVH382c7iNj0E6612ujqodf2Y5lM34ZNLnKbTg0uQaCMRIulxRgl8wLe /RfueEjLg9hJIwkztLPcYdNLNIANOMk= Date: Tue, 10 Jan 2023 05:31:34 +0100 From: Jan Hubicka To: gcc-patches@gcc.gnu.org Subject: More znver4 x86-tune flags Message-ID: MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline X-Spam-Status: No, score=-11.3 required=5.0 tests=BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,GIT_PATCH_0,HEADER_FROM_DIFFERENT_DOMAINS,RCVD_IN_MSPIKE_H3,RCVD_IN_MSPIKE_WL,SPF_HELO_NONE,SPF_NONE,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: 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) /* 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. */