From patchwork Mon Jul 15 08:22:28 2024 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Noah Goldstein X-Patchwork-Id: 1960472 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@legolas.ozlabs.org Authentication-Results: legolas.ozlabs.org; dkim=pass (2048-bit key; unprotected) header.d=gmail.com header.i=@gmail.com header.a=rsa-sha256 header.s=20230601 header.b=NHigtTEf; dkim-atps=neutral Authentication-Results: legolas.ozlabs.org; spf=pass (sender SPF authorized) smtp.mailfrom=sourceware.org (client-ip=8.43.85.97; helo=server2.sourceware.org; envelope-from=libc-alpha-bounces~incoming=patchwork.ozlabs.org@sourceware.org; receiver=patchwork.ozlabs.org) Received: from server2.sourceware.org (server2.sourceware.org [8.43.85.97]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature ECDSA (secp384r1) server-digest SHA384) (No client certificate requested) by legolas.ozlabs.org (Postfix) with ESMTPS id 4WMwFP0fhhz1xr4 for ; Mon, 15 Jul 2024 18:23:21 +1000 (AEST) Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id B311D386481D for ; Mon, 15 Jul 2024 08:23:18 +0000 (GMT) X-Original-To: libc-alpha@sourceware.org Delivered-To: libc-alpha@sourceware.org Received: from mail-io1-xd2c.google.com (mail-io1-xd2c.google.com [IPv6:2607:f8b0:4864:20::d2c]) by sourceware.org (Postfix) with ESMTPS id 36A40386481D for ; Mon, 15 Jul 2024 08:22:40 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org 36A40386481D Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=gmail.com ARC-Filter: OpenARC Filter v1.0.0 sourceware.org 36A40386481D Authentication-Results: server2.sourceware.org; arc=none smtp.remote-ip=2607:f8b0:4864:20::d2c ARC-Seal: i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1721031769; cv=none; b=D3KQOurv+Lwl6lKiwQwrTtYWC3JEeMFw/ANH+iBMOVWfOXflMB9F+PUjQjeKz0kfLo4pnt/F0yUsMXdriE7lKeNom8MevF9/bgDPNgcLpeRWm3v1MbT0h8mljaiMNwMqJDnF2HcJ4pcNX8u+eIj55o3xNss84K9oZ5Bn5HLsoaw= ARC-Message-Signature: i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1721031769; c=relaxed/simple; bh=KabFMIGhzWQiTmGPumu4xcxncvd2cVDvZcXCpq0Z4MM=; h=DKIM-Signature:From:To:Subject:Date:Message-Id:MIME-Version; b=Q2AKBJGSugLhfmcTD3nECzHgaJFJyxi+bF4N8Gi1MDxs96kiJzROekMX9y+g36lyOvjI9q042EuST4QbsyxJ5qViMJ9XSO83MXby0JoErdjV+Gjou6VGsm2F0pbPxG0/Q0QqILFU+JHEOX/7Q4lM3b6ewkbR11joF2HB8STYqHs= ARC-Authentication-Results: i=1; server2.sourceware.org Received: by mail-io1-xd2c.google.com with SMTP id ca18e2360f4ac-7fedb09357aso215605839f.2 for ; Mon, 15 Jul 2024 01:22:40 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20230601; t=1721031759; x=1721636559; darn=sourceware.org; h=content-transfer-encoding:mime-version:message-id:date:subject:cc :to:from:from:to:cc:subject:date:message-id:reply-to; bh=hNK8uiJztkmdUKYUq4qV1RYneKIDHs1/FsD8qGZmFQA=; b=NHigtTEfUKwCSCMZNFw2mkn4kwhfsFbmPi/Z1ULI2jXAFp1x8tOGEagbmYCEzMLPj2 9zPm6QyvOmcxO9DwfcifJBIL1WT7xMlpq2JoLscCTcM8aE1NfQ0s9e2JwhqT/Dq9r8qb QvUAZ0fiZmkjiSWiWF3d+8xPmAoNdy+zJiWrYyUQhsSvr+7PPnEwu/oN2HiivEO+UUK0 Lyqh9M8FIXacBb2p2C8pFhyfNEcIko9e+aeqGHLnhHJu+bxWlEpsOIpFhl+bYknfwfEJ I8alaF1HJafQ6Alc9iH6YNt/bYetngPFBcb4F8Pwy45sjuW3qF6cwK7kCyhdmMKJPmOS vPAA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1721031759; x=1721636559; h=content-transfer-encoding:mime-version:message-id:date:subject:cc :to:from:x-gm-message-state:from:to:cc:subject:date:message-id :reply-to; bh=hNK8uiJztkmdUKYUq4qV1RYneKIDHs1/FsD8qGZmFQA=; b=dAEyqfeUidqH0JA9yIBO3GUK4NEUvnh1wUSQ2fvKlIjyEXSghuOPPxQgDSoBIS2H7S CP6RXKBW+n3Y1oDUNAmR0ZkINYpdnzKcjs0g/eNvGCXQe23wCP9XbmxzgnttVkf6Plwk JGsQ8TG0TTrdWuFDZAyFJS6unTyYkL6/LAgSh/5QtFFD5EVFIPuirNPFQD1S/CHirujc 89+6XNvRBke6MSF1FQe6qpajoIP+zGt9US0h5aWdc8CLNmnJAt44F3UlwOYIekjRNcNo eYfE7d8LP+wU4mUdC0UFZaFBoPV/INcaI7xeWlIq6SBOjdBHJuyU5vQktPt5BX+tNC5A 6g9w== X-Gm-Message-State: AOJu0YyTMbdxGyYJ4AhNqyJadlpJUUxNRL84jHqWm47zPHKqd61Wj9Cs RAzbstO7d7BUyzXpbmFZUYrH4hNZsn9xWrT4o0D0S3Yb8TEJ30w6YWVQrg/hvYI= X-Google-Smtp-Source: AGHT+IGvzBiX05qmSPvwDKGdaLQWU69sY7O8v3N1hqBmFig1yLgVm0yJwBsuF1ktYh96gPLW6PvnGA== X-Received: by 2002:a05:6602:275a:b0:7eb:4b89:b706 with SMTP id ca18e2360f4ac-800005e08e7mr2039040339f.8.1721031758925; Mon, 15 Jul 2024 01:22:38 -0700 (PDT) Received: from noahgold-desk.sh.intel.com ([192.55.46.44]) by smtp.gmail.com with ESMTPSA id d2e1a72fcca58-70b7eca6a00sm3912408b3a.142.2024.07.15.01.22.37 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Mon, 15 Jul 2024 01:22:38 -0700 (PDT) From: Noah Goldstein To: libc-alpha@sourceware.org Cc: goldstein.w.n@gmail.com, hjl.tools@gmail.com Subject: [PATCH v1] x86: Disable non-temporal memset on Skylake Server Date: Mon, 15 Jul 2024 16:22:28 +0800 Message-Id: <20240715082228.715048-1-goldstein.w.n@gmail.com> X-Mailer: git-send-email 2.34.1 MIME-Version: 1.0 X-Spam-Status: No, score=-12.2 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 X-BeenThere: libc-alpha@sourceware.org X-Mailman-Version: 2.1.30 Precedence: list List-Id: Libc-alpha mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: libc-alpha-bounces~incoming=patchwork.ozlabs.org@sourceware.org The original commit enabling non-temporal memset on Skylake Server had erroneous benchmarks (actually done on ICX). Further benchmarks indicate non-temporal stores may in fact by a regression on Skylake Server. This commit may be over-cautious in some cases, but should avoid any regressions for 2.40. Tested using qemu on all x86_64 cpu arch supported by both qemu + GLIBC. --- sysdeps/x86/cpu-features.c | 11 +++++++++-- sysdeps/x86/cpu-tunables.c | 5 +++++ sysdeps/x86/dl-cacheinfo.h | 15 ++++++++------- .../cpu-features-preferred_feature_index_1.def | 1 + sysdeps/x86/tst-hwcap-tunables.c | 4 ++-- 5 files changed, 25 insertions(+), 11 deletions(-) diff --git a/sysdeps/x86/cpu-features.c b/sysdeps/x86/cpu-features.c index e501e084ef..f473fba216 100644 --- a/sysdeps/x86/cpu-features.c +++ b/sysdeps/x86/cpu-features.c @@ -870,11 +870,18 @@ init_cpu_features (struct cpu_features *cpu_features) /* Newer Bigcore microarch (larger non-temporal store threshold). */ - case INTEL_BIGCORE_SKYLAKE: - case INTEL_BIGCORE_KABYLAKE: case INTEL_BIGCORE_COMETLAKE: case INTEL_BIGCORE_SKYLAKE_AVX512: case INTEL_BIGCORE_CANNONLAKE: + /* Benchmarks indicate non-temporal memset is not + necessarily profitable on SKX (and in some cases much + worse). This is likely unique to SKX due its it unique + mesh interconnect (not present on ICX or BWD). Disable + non-temporal on all Skylake servers. */ + cpu_features->preferred[index_arch_Avoid_Non_Temporal_Memset] + |= bit_arch_Avoid_Non_Temporal_Memset; + case INTEL_BIGCORE_SKYLAKE: + case INTEL_BIGCORE_KABYLAKE: case INTEL_BIGCORE_ICELAKE: case INTEL_BIGCORE_TIGERLAKE: case INTEL_BIGCORE_ROCKETLAKE: diff --git a/sysdeps/x86/cpu-tunables.c b/sysdeps/x86/cpu-tunables.c index 89da7a03da..ae9dcd6180 100644 --- a/sysdeps/x86/cpu-tunables.c +++ b/sysdeps/x86/cpu-tunables.c @@ -243,6 +243,11 @@ TUNABLE_CALLBACK (set_hwcaps) (tunable_val_t *valp) (n, cpu_features, MathVec_Prefer_No_AVX512, AVX512F, 24); } break; + case 25: + { + CHECK_GLIBC_IFUNC_PREFERRED_BOTH (n, cpu_features, + Avoid_Non_Temporal_Memset, 25); + } case 26: { CHECK_GLIBC_IFUNC_PREFERRED_NEED_BOTH diff --git a/sysdeps/x86/dl-cacheinfo.h b/sysdeps/x86/dl-cacheinfo.h index 5e77345a6e..a1c03b8903 100644 --- a/sysdeps/x86/dl-cacheinfo.h +++ b/sysdeps/x86/dl-cacheinfo.h @@ -991,13 +991,14 @@ dl_init_cacheinfo (struct cpu_features *cpu_features) /* Non-temporal stores are more performant on Intel and AMD hardware above non_temporal_threshold. Enable this for both Intel and AMD hardware. */ unsigned long int memset_non_temporal_threshold = SIZE_MAX; - if (cpu_features->basic.kind == arch_kind_intel - || cpu_features->basic.kind == arch_kind_amd) - memset_non_temporal_threshold = non_temporal_threshold; - - /* For AMD CPUs that support ERMS (Zen3+), REP MOVSB is in a lot of - cases slower than the vectorized path (and for some alignments, - it is really slow, check BZ #30994). */ + if (!CPU_FEATURES_ARCH_P (cpu_features, Avoid_Non_Temporal_Memset) + && (cpu_features->basic.kind == arch_kind_intel + || cpu_features->basic.kind == arch_kind_amd)) + memset_non_temporal_threshold = non_temporal_threshold; + + /* For AMD CPUs that support ERMS (Zen3+), REP MOVSB is in a lot of + cases slower than the vectorized path (and for some alignments, + it is really slow, check BZ #30994). */ if (cpu_features->basic.kind == arch_kind_amd) rep_movsb_threshold = non_temporal_threshold; diff --git a/sysdeps/x86/include/cpu-features-preferred_feature_index_1.def b/sysdeps/x86/include/cpu-features-preferred_feature_index_1.def index 85e7f54ec8..61bbbc2e89 100644 --- a/sysdeps/x86/include/cpu-features-preferred_feature_index_1.def +++ b/sysdeps/x86/include/cpu-features-preferred_feature_index_1.def @@ -33,3 +33,4 @@ BIT (Prefer_No_AVX512) BIT (MathVec_Prefer_No_AVX512) BIT (Prefer_FSRM) BIT (Avoid_Short_Distance_REP_MOVSB) +BIT (Avoid_Non_Temporal_Memset) diff --git a/sysdeps/x86/tst-hwcap-tunables.c b/sysdeps/x86/tst-hwcap-tunables.c index 8589a9fd66..94307283d7 100644 --- a/sysdeps/x86/tst-hwcap-tunables.c +++ b/sysdeps/x86/tst-hwcap-tunables.c @@ -60,7 +60,7 @@ static const struct test_t /* Disable everything. */ "-Prefer_ERMS,-Prefer_FSRM,-AVX,-AVX2,-AVX512F,-AVX512VL," "-SSE4_1,-SSE4_2,-SSSE3,-Fast_Unaligned_Load,-ERMS," - "-AVX_Fast_Unaligned_Load", + "-AVX_Fast_Unaligned_Load,-Avoid_Non_Temporal_Memset", test_1, array_length (test_1) }, @@ -68,7 +68,7 @@ static const struct test_t /* Same as before, but with some empty suboptions. */ ",-,-Prefer_ERMS,-Prefer_FSRM,-AVX,-AVX2,-AVX512F,-AVX512VL," "-SSE4_1,-SSE4_2,-SSSE3,-Fast_Unaligned_Load,,-," - "-ERMS,-AVX_Fast_Unaligned_Load,-,", + "-ERMS,-AVX_Fast_Unaligned_Load,-Avoid_Non_Temporal_Memset,-,", test_1, array_length (test_1) }