From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [63.128.21.124]) by sourceware.org (Postfix) with ESMTP id 87A63385041A for ; Thu, 4 Feb 2021 10:28:30 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.3.2 sourceware.org 87A63385041A Received: from mimecast-mx01.redhat.com (mimecast-mx01.redhat.com [209.132.183.4]) (Using TLS) by relay.mimecast.com with ESMTP id us-mta-416-tF4fMhBcPk6VNQjNEjTFyw-1; Thu, 04 Feb 2021 05:28:27 -0500 X-MC-Unique: tF4fMhBcPk6VNQjNEjTFyw-1 Received: from smtp.corp.redhat.com (int-mx03.intmail.prod.int.phx2.redhat.com [10.5.11.13]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mimecast-mx01.redhat.com (Postfix) with ESMTPS id 4AF151800D41; Thu, 4 Feb 2021 10:28:26 +0000 (UTC) Received: from oldenburg.str.redhat.com (ovpn-112-116.ams2.redhat.com [10.36.112.116]) by smtp.corp.redhat.com (Postfix) with ESMTPS id 3EEE26F968; Thu, 4 Feb 2021 10:28:25 +0000 (UTC) From: Florian Weimer To: Daniel Black Cc: libc-alpha@sourceware.org, aarcange@redhat.com Subject: Re: ld.so - text and .data application segments to huge pages References: Date: Thu, 04 Feb 2021 11:28:30 +0100 In-Reply-To: (Daniel Black's message of "Thu, 4 Feb 2021 18:58:34 +1100") Message-ID: <87o8h0m8j5.fsf@oldenburg.str.redhat.com> User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/27.1 (gnu/linux) MIME-Version: 1.0 X-Scanned-By: MIMEDefang 2.79 on 10.5.11.13 X-Mimecast-Spam-Score: 0 X-Mimecast-Originator: redhat.com Content-Type: text/plain X-Spam-Status: No, score=-6.6 required=5.0 tests=BAYES_00, DKIMWL_WL_HIGH, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, RCVD_IN_DNSWL_LOW, RCVD_IN_MSPIKE_H4, RCVD_IN_MSPIKE_WL, SPF_HELO_NONE, SPF_PASS, TXREP autolearn=ham autolearn_force=no version=3.4.2 X-Spam-Checker-Version: SpamAssassin 3.4.2 (2018-09-13) 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: Thu, 04 Feb 2021 10:28:31 -0000 * Daniel Black: > Hi, > > I received this exceptionally large (slightly convoluted) patch to > MariaDB to move the .text and .data sections to huge pages in the > MariaDB issue https://jira.mariadb.org/browse/MDEV-24051. > > The author Dmitriy Philimonov (Huawei) claimed great gains of "9% in > TPS speedup. iTLB && dTLB misses reduced in 38 times and 5 times > respectively. Latency becomes lower, the jitter in the TPS reduced > significantly", so it's definitely beneficial. > > Obviously maintaining such a large loader work around in a userspace > application is going to be a burden on the skill set of developers we > have available. > > How complex would it be for the glibc elf loader to use huge pages for > data segments on load? (extending _dl_map_segments with an environment > variable control like LD_HUGEPAGES I assume). > > Is there any required kernel work to implement this feature? > > Would such a concept be supported by the developer community here? > > If so I'm happy to write up a feature request (https://sourceware.org/bugzilla). > > Anyone off hand interested in writing it? I think the kernel already supports this via transparent hugepages, if you increase the segment alignment via those link editor flags (as in the proposed patch). Cc:ing Andrea for additional comments. Thanks, Florian -- Red Hat GmbH, https://de.redhat.com/ , Registered seat: Grasbrunn, Commercial register: Amtsgericht Muenchen, HRB 153243, Managing Directors: Charles Cachera, Brian Klemm, Laurie Krebs, Michael O'Neill