LWN: Comments on "Soft CPU affinity" https://lwn.net/Articles/792502/ This is a special feed containing comments posted to the individual LWN article titled "Soft CPU affinity". en-us Wed, 22 Oct 2025 17:45:14 +0000 Wed, 22 Oct 2025 17:45:14 +0000 https://www.rssboard.org/rss-specification lwn@lwn.net Soft CPU affinity https://lwn.net/Articles/793314/ https://lwn.net/Articles/793314/ post-factum <div class="FormattedComment"> Why not just offload such a decision to some smaller userspace daemon?<br> </div> Wed, 10 Jul 2019 11:33:09 +0000 Soft CPU affinity https://lwn.net/Articles/793278/ https://lwn.net/Articles/793278/ valarauca <div class="FormattedComment"> Isn't relocating memory &amp; changing NUMA masks already be part of this process?<br> <p> Presumably hot-plugging is already migrating memory prior to the CPU being removed, (or so I assume; without the on-cpu memory controller, the EC-DRAM will blank). Or is RAM kept live during a hot-plug?<br> <p> </div> Tue, 09 Jul 2019 19:56:28 +0000 Soft CPU affinity https://lwn.net/Articles/793155/ https://lwn.net/Articles/793155/ zlynx <div class="FormattedComment"> CPU hot-plug removal, with affine processes.<br> </div> Tue, 09 Jul 2019 01:07:56 +0000 Soft CPU affinity https://lwn.net/Articles/793008/ https://lwn.net/Articles/793008/ maxfragg <div class="FormattedComment"> Sounds like a hard sell.<br> On paper having a soft cpu affinity might sound useful, but in my experience, in most cases where people want to pin their tasks, they don't want any heuristics which might cause strange side-effects. Predicting when an application will break out of its soft affinity and how this will affect such a fine tuned system might cause more trouble than it is worth<br> </div> Mon, 08 Jul 2019 06:06:32 +0000