Path: csiph.com!weretis.net!feeder8.news.weretis.net!fu-berlin.de!bofh.it!news.nic.it!robomod From: Marc Haber Newsgroups: linux.debian.kernel Subject: Re: CONFIG_PREEMPT_DYNAMIC=y? Date: Tue, 10 Oct 2023 15:50:01 +0200 Message-ID: References: X-Original-To: Emanuele Rocca X-Mailbox-Line: From debian-kernel-request@lists.debian.org Tue Oct 10 13:44:20 2023 Old-Return-Path: X-Amavis-Spam-Status: No, score=-6.9 tagged_above=-10000 required=5.3 tests=[BAYES_00=-2, FOURLA=0.1, LDO_WHITELIST=-5] autolearn=ham autolearn_force=no X-Policyd-Weight: using cached result; rate:hard: -5.5 MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline User-Agent: Mutt/2.2.12 (2023-09-09) X-Mailing-List: archive/latest/140334 List-ID: List-URL: List-Archive: https://lists.debian.org/msgid-search/ZSVVIi9ZRssRY9-X@torres.zugschlus.de Approved: robomod@news.nic.it Lines: 48 Organization: linux.* mail to news gateway Sender: robomod@news.nic.it X-Original-Cc: Johannes Schauer Marin Rodrigues , debian-kernel@lists.debian.org X-Original-Date: Tue, 10 Oct 2023 15:44:02 +0200 X-Original-Message-ID: X-Original-References: <169678202078.3029203.13684704010302829243@localhost> <3679117.hVnToU1rA3@bagend> <169693286797.3453145.7694745520755158332@localhost> Xref: csiph.com linux.debian.kernel:80643 On Tue, Oct 10, 2023 at 12:32:22PM +0200, Emanuele Rocca wrote: > On 2023-10-10 12:14, Johannes Schauer Marin Rodrigues wrote: > > Quoting Emanuele Rocca (2023-10-10 12:10:07) > > > Setting CONFIG_HAVE_PREEMPT_DYNAMIC=y on arm64 is likely safe for us too, but > > > we may want to run some benchmark first to see if there are any noticeable > > > slowdowns. > > > > I have a couple of arm64 boards here that I can run benchmarks on. What would > > you recommend to run to test the effect of CONFIG_HAVE_PREEMPT_DYNAMIC=y? > > I'm not sure what the best benchmark would be, but maybe you could start > with something generic like timing a kernel build with and without > PREEMPT_DYNAMIC and throwing 'hey' at your favorite web server? > https://github.com/rakyll/hey Whatever you do, check whether having a serial console still works with that setting. I had trouble with that in kernel 6.5, had some fruitless discussions on the LKML and developer interest abruptly died when I found out that the bug I encountered could be worked around by having those settings: [1/4998]mh@fan:~ $ grep PREEMPT /boot/config-6.5.6-zgws1 CONFIG_PREEMPT_BUILD=y # CONFIG_PREEMPT_NONE is not set CONFIG_PREEMPT_VOLUNTARY=y # CONFIG_PREEMPT is not set CONFIG_PREEMPT_COUNT=y CONFIG_PREEMPTION=y CONFIG_PREEMPT_DYNAMIC=y CONFIG_PREEMPT_RCU=y CONFIG_HAVE_PREEMPT_DYNAMIC=y CONFIG_HAVE_PREEMPT_DYNAMIC_CALL=y CONFIG_PREEMPT_NOTIFIERS=y CONFIG_DRM_I915_PREEMPT_TIMEOUT=640 CONFIG_DRM_I915_PREEMPT_TIMEOUT_COMPUTE=7500 # CONFIG_DEBUG_PREEMPT is not set # CONFIG_PREEMPT_TRACER is not set # CONFIG_PREEMPTIRQ_DELAY_TEST is not set Greetings Marc -- ----------------------------------------------------------------------------- Marc Haber | "I don't trust Computers. They | Mailadresse im Header Leimen, Germany | lose things." Winona Ryder | Fon: *49 6224 1600402 Nordisch by Nature | How to make an American Quilt | Fax: *49 6224 1600421