{"id":2222545,"url":"http://patchwork.ozlabs.org/api/1.1/patches/2222545/?format=json","web_url":"http://patchwork.ozlabs.org/project/glibc/patch/20260412174334.3111385-6-wangrui@loongson.cn/","project":{"id":41,"url":"http://patchwork.ozlabs.org/api/1.1/projects/41/?format=json","name":"GNU C Library","link_name":"glibc","list_id":"libc-alpha.sourceware.org","list_email":"libc-alpha@sourceware.org","web_url":"","scm_url":"","webscm_url":""},"msgid":"<20260412174334.3111385-6-wangrui@loongson.cn>","date":"2026-04-12T17:43:33","name":"[v9,5/6] loongarch: Enable THP-aligned load segments by default on 64-bit","commit_ref":null,"pull_url":null,"state":"new","archived":false,"hash":"63762563629e972f42b6e9895afc7581372fa4ef","submitter":{"id":85150,"url":"http://patchwork.ozlabs.org/api/1.1/people/85150/?format=json","name":"WANG Rui","email":"wangrui@loongson.cn"},"delegate":null,"mbox":"http://patchwork.ozlabs.org/project/glibc/patch/20260412174334.3111385-6-wangrui@loongson.cn/mbox/","series":[{"id":499622,"url":"http://patchwork.ozlabs.org/api/1.1/series/499622/?format=json","web_url":"http://patchwork.ozlabs.org/project/glibc/list/?series=499622","date":"2026-04-12T17:43:29","name":"elf: THP-aware load segment alignment","version":9,"mbox":"http://patchwork.ozlabs.org/series/499622/mbox/"}],"comments":"http://patchwork.ozlabs.org/api/patches/2222545/comments/","check":"pending","checks":"http://patchwork.ozlabs.org/api/patches/2222545/checks/","tags":{},"headers":{"Return-Path":"<libc-alpha-bounces~incoming=patchwork.ozlabs.org@sourceware.org>","X-Original-To":["incoming@patchwork.ozlabs.org","libc-alpha@sourceware.org"],"Delivered-To":["patchwork-incoming@legolas.ozlabs.org","libc-alpha@sourceware.org"],"Authentication-Results":["legolas.ozlabs.org;\n spf=pass (sender SPF authorized) smtp.mailfrom=sourceware.org\n (client-ip=2620:52:6:3111::32; helo=vm01.sourceware.org;\n envelope-from=libc-alpha-bounces~incoming=patchwork.ozlabs.org@sourceware.org;\n receiver=patchwork.ozlabs.org)","sourceware.org;\n dmarc=none (p=none dis=none) header.from=loongson.cn","sourceware.org; spf=pass smtp.mailfrom=loongson.cn","server2.sourceware.org;\n arc=none smtp.remote-ip=114.242.206.163"],"Received":["from vm01.sourceware.org (vm01.sourceware.org\n [IPv6:2620:52:6:3111::32])\n\t(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)\n\t key-exchange x25519 server-signature ECDSA (secp384r1) server-digest SHA384)\n\t(No client certificate requested)\n\tby legolas.ozlabs.org (Postfix) with ESMTPS id 4ftybb1RkDz1yCx\n\tfor <incoming@patchwork.ozlabs.org>; Mon, 13 Apr 2026 03:44:43 +1000 (AEST)","from vm01.sourceware.org (localhost [127.0.0.1])\n\tby sourceware.org (Postfix) with ESMTP id 9B72C4BA23D7\n\tfor <incoming@patchwork.ozlabs.org>; Sun, 12 Apr 2026 17:44:40 +0000 (GMT)","from mail.loongson.cn (mail.loongson.cn [114.242.206.163])\n by sourceware.org (Postfix) with ESMTP id 13BD14BA2E32\n for <libc-alpha@sourceware.org>; Sun, 12 Apr 2026 17:44:19 +0000 (GMT)","from loongson.cn (unknown [223.64.120.66])\n by gateway (Coremail) with SMTP id _____8AxCMLx2dtpCYUkAA--.2462S3;\n Mon, 13 Apr 2026 01:44:17 +0800 (CST)","from localhost (unknown [223.64.120.66])\n by front1 (Coremail) with SMTP id qMiowJBxrsLo2dtpo51rAA--.49052S4;\n Mon, 13 Apr 2026 01:44:15 +0800 (CST)"],"DKIM-Filter":["OpenDKIM Filter v2.11.0 sourceware.org 9B72C4BA23D7","OpenDKIM Filter v2.11.0 sourceware.org 13BD14BA2E32"],"DMARC-Filter":"OpenDMARC Filter v1.4.2 sourceware.org 13BD14BA2E32","ARC-Filter":"OpenARC Filter v1.0.0 sourceware.org 13BD14BA2E32","ARC-Seal":"i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1776015860; cv=none;\n b=VoyLM/pwWfngPJdInz0K25VpGyic5/l15CfCzBIlV7N/2xmJCp+3K4pc3jpulyJERD+PO1spR78PpIyj39kTmwHyI/ogzC9e8Bu6pfD4q/oy1noU4np5iUPTV1hpOQxpHJreklnujLNbYFBUvdUgX1NQ8XkL5EKUr+8IBKAec00=","ARC-Message-Signature":"i=1; a=rsa-sha256; d=sourceware.org; s=key;\n t=1776015860; c=relaxed/simple;\n bh=eZHFaLBd7oXRxrYYGmvQftnnh8LTLsZDZfT/vdpUyeQ=;\n h=From:To:Subject:Date:Message-ID:MIME-Version;\n b=BG728BFImlCWWyVuz7VoF3x+VrW4yGvW/LhOFr+luKLClab8s6NIYzqEpK4AqbXGnGqBcovrYY5Je5o4EywCXOLaRjT29wpoqRPh8PNyMNVb1/5z5GU7/N7c/7PrHUkhj8URi7ZulVW8/U/UZj8SAeUiC1Upv8nz1ONZlt/zB2w=","ARC-Authentication-Results":"i=1; server2.sourceware.org","From":"WANG Rui <wangrui@loongson.cn>","To":"libc-alpha@sourceware.org","Cc":"Adhemerval Zanella <adhemerval.zanella@linaro.org>,\n Dev Jain <dev.jain@arm.com>, Florian Weimer <fweimer@redhat.com>,\n Wilco Dijkstra <Wilco.Dijkstra@arm.com>, Xi Ruoyao <xry111@xry111.site>,\n WANG Xuerui <git@xen0n.name>, caiyinyu <caiyinyu@loongson.cn>,\n mengqinggang <mengqinggang@loongson.cn>,\n Huacai Chen <chenhuacai@kernel.org>, hjl.tools@gmail.com,\n WANG Rui <wangrui@loongson.cn>","Subject":"[PATCH v9 5/6] loongarch: Enable THP-aligned load segments by default\n on 64-bit","Date":"Mon, 13 Apr 2026 01:43:33 +0800","Message-ID":"<20260412174334.3111385-6-wangrui@loongson.cn>","X-Mailer":"git-send-email 2.53.0","In-Reply-To":"<20260412174334.3111385-1-wangrui@loongson.cn>","References":"<20260412174334.3111385-1-wangrui@loongson.cn>","MIME-Version":"1.0","Content-Transfer-Encoding":"8bit","X-CM-TRANSID":"qMiowJBxrsLo2dtpo51rAA--.49052S4","X-CM-SenderInfo":"pzdqw2txl6z05rqj20fqof0/","X-Coremail-Antispam":"1Uk129KBj93XoWxCw1UXw43KF1fGr4rtw1xXrc_yoWrWr1Upr\n WakFs5CF4rWr17C3yak3W7Z3W5WF4rCFs8Cr9xCw4UZryUJr1xZFsFyF1fJFy7J3yxGF4x\n uFnFq3WDuF1rAacCm3ZEXasCq-sJn29KB7ZKAUJUUUU7529EdanIXcx71UUUUU7KY7ZEXa\n sCq-sGcSsGvfJ3Ic02F40EFcxC0VAKzVAqx4xG6I80ebIjqfuFe4nvWSU5nxnvy29KBjDU\n 0xBIdaVrnRJUUUBYb4IE77IF4wAFF20E14v26r1j6r4UM7CY07I20VC2zVCF04k26cxKx2\n IYs7xG6rWj6s0DM7CIcVAFz4kK6r1q6r4UM28lY4IEw2IIxxk0rwA2F7IY1VAKz4vEj48v\n e4kI8wA2z4x0Y4vE2Ix0cI8IcVAFwI0_Xr0_Ar1l84ACjcxK6xIIjxv20xvEc7CjxVAFwI\n 0_Gr0_Cr1l84ACjcxK6I8E87Iv67AKxVW8JVWxJwA2z4x0Y4vEx4A2jsIEc7CjxVAFwI0_\n Gr0_Gr1UM2kKe7AKxVWUXVWUAwAS0I0E0xvYzxvE52x082IY62kv0487Mc804VCY07AIYI\n kI8VC2zVCFFI0UMc02F40EFcxC0VAKzVAqx4xG6I80ewAv7VC0I7IYx2IY67AKxVWrXVW3\n AwAv7VC2z280aVAFwI0_Gr0_Cr1lOx8S6xCaFVCjc4AY6r1j6r4UM4x0Y48IcxkI7VAKI4\n 8JMxkF7I0En4kS14v26r126r1DMxAIw28IcxkI7VAKI48JMxC20s026xCaFVCjc4AY6r1j\n 6r4UMxCIbckI1I0E14v26r1Y6r17MI8I3I0E5I8CrVAFwI0_Jr0_Jr4lx2IqxVCjr7xvwV\n AFwI0_JrI_JrWlx4CE17CEb7AF67AKxVWUtVW8ZwCIc40Y0x0EwIxGrwCI42IY6xIIjxv2\n 0xvE14v26ryj6F1UMIIF0xvE2Ix0cI8IcVCY1x0267AKxVW8JVWxJwCI42IY6xAIw20EY4\n v20xvaj40_Jr0_JF4lIxAIcVC2z280aVAFwI0_Gr0_Cr1lIxAIcVC2z280aVCY1x0267AK\n xVW8JVW8JrUvcSsGvfC2KfnxnUUI43ZEXa7IU8CzutUUUUU==","X-BeenThere":"libc-alpha@sourceware.org","X-Mailman-Version":"2.1.30","Precedence":"list","List-Id":"Libc-alpha mailing list <libc-alpha.sourceware.org>","List-Unsubscribe":"<https://sourceware.org/mailman/options/libc-alpha>,\n <mailto:libc-alpha-request@sourceware.org?subject=unsubscribe>","List-Archive":"<https://sourceware.org/pipermail/libc-alpha/>","List-Post":"<mailto:libc-alpha@sourceware.org>","List-Help":"<mailto:libc-alpha-request@sourceware.org?subject=help>","List-Subscribe":"<https://sourceware.org/mailman/listinfo/libc-alpha>,\n <mailto:libc-alpha-request@sourceware.org?subject=subscribe>","Errors-To":"libc-alpha-bounces~incoming=patchwork.ozlabs.org@sourceware.org"},"content":"On LoongArch64 Linux, aligning ELF load segments to Transparent Huge Page\n(THP) boundaries provides consistent performance benefits for large\nbinaries by reducing TLB pressure and improving instruction fetch\nefficiency.\n\nEnable THP-based load segment alignment by default on LoongArch64 by\nsetting `glibc.elf.thp=1` during startup. Define the default THP\npage size for load segment alignment on LoongArch64 as 32MB.\n\nThis allows the dynamic loader to apply THP-friendly alignment without\nrequiring the `glibc.elf.thp` tunable to be explicitly set.\n\nBenchmarks\n\nMachine: Loongson 3A6000 (LoongArch64)\nKernel: 6.18.13\n  CONFIG_READ_ONLY_THP_FOR_FS=y\n  CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS=y\n\nWorkload 1: building Cargo 1.93.0\nRustc: nightly-2026-02-26\n\n                Without patch        With patch\ninstructions    3,690,358,948,176    3,690,301,774,568\ncpu-cycles      4,233,025,766,760    4,035,866,635,741\nitlb-misses     9,708,829,532        2,700,014,717\ntime elapsed    302.40 s             289.68 s\n\nInstructions remain essentially unchanged. iTLB misses drop by about\n72%, reducing CPU cycles by about 4.7% and wall time by about 4.2%.\n\nWorkload 2: building Linux kernel v7.0-rc1\nLLVM: 21.1.8\n\n                Without patch        With patch\ninstructions    14,163,739,876,387   14,169,418,598,675\ncpu-cycles      19,231,890,317,741   16,851,494,928,181\nitlb-misses     91,142,010,440       90,779,245\ntime elapsed    1022.09 s            893.22 s\n\nInstructions remain roughly the same. iTLB misses drop from about 91B\nto about 90M (roughly 99.9% reduction), reducing CPU cycles by about\n12% and wall time by about 12.6%.\n\nReviewed-by: caiyinyu <caiyinyu@loongson.cn>\nSigned-off-by: WANG Rui <wangrui@loongson.cn>\n---\n .../unix/sysv/linux/loongarch/cpu-features.c  |  6 +++++\n .../loongarch/lp64/dl-map-segment-align.h     | 22 +++++++++++++++++++\n 2 files changed, 28 insertions(+)\n create mode 100644 sysdeps/unix/sysv/linux/loongarch/lp64/dl-map-segment-align.h","diff":"diff --git a/sysdeps/unix/sysv/linux/loongarch/cpu-features.c b/sysdeps/unix/sysv/linux/loongarch/cpu-features.c\nindex 0c125e64e2b..7ed54ff53e7 100644\n--- a/sysdeps/unix/sysv/linux/loongarch/cpu-features.c\n+++ b/sysdeps/unix/sysv/linux/loongarch/cpu-features.c\n@@ -27,4 +27,10 @@ init_cpu_features (struct cpu_features *cpu_features)\n   GLRO(dl_larch_cpu_features).hwcap = GLRO(dl_hwcap);\n   TUNABLE_GET (glibc, cpu, hwcaps, tunable_val_t *,\n \t       TUNABLE_CALLBACK (set_hwcaps));\n+\n+#ifdef __LP64__\n+  /* Enable THP-based load segment alignment by default on LoongArch64. */\n+  if (!TUNABLE_IS_INITIALIZED (glibc, elf, thp))\n+    TUNABLE_SET (glibc, elf, thp, 1);\n+#endif\n }\ndiff --git a/sysdeps/unix/sysv/linux/loongarch/lp64/dl-map-segment-align.h b/sysdeps/unix/sysv/linux/loongarch/lp64/dl-map-segment-align.h\nnew file mode 100644\nindex 00000000000..c51ee4ac47e\n--- /dev/null\n+++ b/sysdeps/unix/sysv/linux/loongarch/lp64/dl-map-segment-align.h\n@@ -0,0 +1,22 @@\n+/* _dl_map_segment_align.  LoongArch64 Linux version.\n+   Copyright (C) 2026 Free Software Foundation, Inc.\n+   Copyright The GNU Toolchain Authors.\n+   This file is part of the GNU C Library.\n+\n+   The GNU C Library is free software; you can redistribute it and/or\n+   modify it under the terms of the GNU Lesser General Public\n+   License as published by the Free Software Foundation; either\n+   version 2.1 of the License, or (at your option) any later version.\n+\n+   The GNU C Library is distributed in the hope that it will be useful,\n+   but WITHOUT ANY WARRANTY; without even the implied warranty of\n+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU\n+   Lesser General Public License for more details.\n+\n+   You should have received a copy of the GNU Lesser General Public\n+   License along with the GNU C Library; if not, see\n+   <https://www.gnu.org/licenses/>.  */\n+\n+#define DL_MAP_DEFAULT_THP_PAGESIZE (32 * 1024 * 1024)\n+\n+#include_next <dl-map-segment-align.h>\n","prefixes":["v9","5/6"]}