[$] In search of faster this_cpu operations
A proposal to improve the performance of this_cpu operations in the Linux kernel was discussed at the 2026 Linux Storage, Filesystem, Memory Management, and BPF Summit. Yang Shi suggested reimplementing per-CPU variables to use the same address across all CPUs, which could enhance performance on various architectures. However, concerns about managing translation lookaside buffer entries and the potential impact on existing macros were raised during the discussion.
- ▪Yang Shi proposed a change to this_cpu operations to improve performance across different CPU architectures.
- ▪The new implementation would use per-CPU page tables to standardize variable addresses on all CPUs.
- ▪Performance benchmarks indicated a 13-18% reduction in system time on a 160-core Arm system.
Opening excerpt (first ~120 words) tap to expand
Please consider subscribing to LWN Subscriptions are the lifeblood of LWN.net. If you appreciate this content and would like to see more of it, your subscription will help to ensure that LWN continues to thrive. Please visit this page to join up and keep LWN on the net. By Jonathan CorbetMay 19, 2026 LSFMM+BPF The kernel's this_cpu operations are meant to speed access to per-CPU variables. They are more optimal on some CPUs than others, though. During a memory-management-track session at the 2026 Linux Storage, Filesystem, Memory Management, and BPF Summit, Yang Shi proposed a fundamental, and somewhat controversial, change to how these operations work in order to provide better performance on a wider range of architectures.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at LWN.net (Linux Weekly News).