New Optimization in Linux Scheduler: Merging cfs_rq and sched_entity for Dramatic Cache Hit Rate Improvement
Abstract On the Linux 6.15 test kernel, a series of patches submitted by Google engineer Zecheng Li significantly reduces pointer jumps and cache misses by merging the <span><span>cfs_rq</span></span> (Completely Fair Scheduler run queue) with the <span><span>sched_entity</span></span> (scheduling entity) structure and switching to a per-CPU allocator. In large-scale cgroup hierarchical scenarios, this patch reduces LLC (Last … Read more