Torvalds Submits His First AI-Aided Kernel Patch: Concurrent File Reading Speed Soars Nearly Threefold
2 hour ago / Read about 0 minute
Author:小编   

Linux creator Linus Torvalds has established a novel experimental branch, dubbed 'noref-filemap-read-with-llm', within his personal Git repository. This marks the inaugural kernel patch series he has crafted with the aid of Anthropic's AI coding assistant, Claude Code. The objective is to bolster the scalability of page reference counting and tackle the persistent challenge of multi-core CPU contention that arises during multi-threaded, high-frequency, yet low-volume reads of identical files within Linux memory management.

Practical testing reveals that when eight threads read the same file concurrently, the read throughput experiences a nearly threefold surge. Simultaneously, the kernel clock cycles expended per read operation plummet by 70%. Presently, the patch remains in the experimental phase and has yet to be integrated into the official Linux 7.4 mainline.

Recently, Torvalds has also harnessed large language models to refactor legacy code, unearthing three bugs that had eluded detection within the Linux system for a staggering 13 years. This signifies that AI-assisted kernel development has evolved from merely enhancing efficiency to uncovering deeply concealed issues.