Cache Aware Scheduling Merged For Linux 7.2 For Boosting Modern Intel & AMD CPUs
The scheduler updates were merged this morning for the Linux 7.2 kernel and it's exciting. Cache Aware Scheduling has finally been merged! This is a win for especially modern Intel and AMD processors with multiple last level caches (LLCs).
Last month I noted that it looked like Cache Aware Scheduling would finally land for Linux 7.2 and as of today it's been realized. The code is now merged and exposed via the CONFIG_SCHED_CACHE Kconfig option. This has been in development for more than a year to benefit today's increasing core counts and bigger, more complex cache layouts.
Cache Aware Scheduling on Linux tries to ensure that tasks sharing data are co-located to the same last level cache (LLC) domain for ensuring better cache locality and reducing cache misses/bouncing. I have benchmarked earlier versions of the patches and found great performance on AMD EPYC CPUs and Xeon 6 benefiting nicely too. Last month I ran some fresh benchmarks in Cache Aware Scheduling Shows Nice Wins For AMD Zen 5 On PostgreSQL, Valkey, Network Performance. I'll have more Cache Aware Scheduling benchmarks from the Linux 7.2 kernel now that it's finally merged.
Kudos to the Intel engineers pushing this work for more than one year and getting it across the finish line.
The scheduling merge for Linux 7.2 also brought improvements to SD_ASYM_CPUCAPACITY scheduling with SMT awareness, optimizing cfs_rq and sched_entity allocation for better data locality, Optimized Donor Migration for Proxy Execution, and various other fixes and scheduling improvements.
See the scheduler pull request for the full list of the now-merged changes for Linux 7.2.
Last month I noted that it looked like Cache Aware Scheduling would finally land for Linux 7.2 and as of today it's been realized. The code is now merged and exposed via the CONFIG_SCHED_CACHE Kconfig option. This has been in development for more than a year to benefit today's increasing core counts and bigger, more complex cache layouts.
Cache Aware Scheduling on Linux tries to ensure that tasks sharing data are co-located to the same last level cache (LLC) domain for ensuring better cache locality and reducing cache misses/bouncing. I have benchmarked earlier versions of the patches and found great performance on AMD EPYC CPUs and Xeon 6 benefiting nicely too. Last month I ran some fresh benchmarks in Cache Aware Scheduling Shows Nice Wins For AMD Zen 5 On PostgreSQL, Valkey, Network Performance. I'll have more Cache Aware Scheduling benchmarks from the Linux 7.2 kernel now that it's finally merged.
👁 Cache Aware Scheduling sysfs
Kudos to the Intel engineers pushing this work for more than one year and getting it across the finish line.
The scheduling merge for Linux 7.2 also brought improvements to SD_ASYM_CPUCAPACITY scheduling with SMT awareness, optimizing cfs_rq and sched_entity allocation for better data locality, Optimized Donor Migration for Proxy Execution, and various other fixes and scheduling improvements.
See the scheduler pull request for the full list of the now-merged changes for Linux 7.2.