AI: Lot of "Hideous" Code but Found Bottlenecks for Faster Linux Compilation

ARM engineer Lorenzo Stoakes posted 23 patches to the Linux kernel mailing list that use an unnamed large language model to identify and fix build bottlenecks, achieving up to 36% faster full kernel builds, 70% faster incremental builds, and ~90% faster noop builds. Stoakes audited and rewrote much of the LLM-generated code, which he described as 'hideous,' and each commit carries an 'Assisted-by' tag.

AI Made A Lot Of "Hideous" Code But Found Major Bottlenecks For Faster Linux Compilation An ARM Linux engineer leveraged AI/LLMs to help uncover some improvements to "significant speed up" the Linux kernel build times. In the process the LLM used generated "a lot of code, much of it hideous" but in the end was used to find some very significant time savings for faster kernel builds via more task parallelization in different areas of the kernel build process. ARM engineer and longtime Linux developer Lorenzo Stoakes posted the set of 23 patches today to " Lorenzo did note that an unnamed large language model LLM was used for uncovering these bottlenecks and addressing them, to which he then audited and evaluated the code personally. Lorenzo wrote: "An LLM was used to first determine where the bottlenecks were then to figure out how to improve them. The patches work to parallelize a lot of single-threaded tasks and improving the efficiency of the build process. Areas such as Kbuild, kallsyms, modpost, objtool, mksysmap, and the Rust build system were all affected by this work. In the end the build time improvements are very significant across a spectrum of hardware and different kernel configurations: More benchmark results and details can be found via ARM engineer and longtime Linux developer Lorenzo Stoakes posted the set of 23 patches today to " significantly speed up kernel builds ." Fundamentally it's to address the significant time spent stuck in single-threaded bottlenecks throughout the kernel build process. In the end he found a kernel build with all modules enabled was around 36% faster, incremental kernel builds up to 70% faster, and noop builds up to ~90% faster. Lorenzo did note that an unnamed large language model LLM was used for uncovering these bottlenecks and addressing them, to which he then audited and evaluated the code personally. Lorenzo wrote: "An LLM was used to first determine where the bottlenecks were then to figure out how to improve them. It generated a lot of code, much of it hideous. I extensively audited and rewrote a lot of it, and heavily edited commit messages, the cover letter and comments.x The LLM has also orchestrated build runs, testing, debugging and analysis. I have manually checked for correctness in both build and running kernels generated with this series applied. Performance improvements were also verified manually. Since an LLM was used extensively, each commit carries an Assisted-by tag." The patches work to parallelize a lot of single-threaded tasks and improving the efficiency of the build process. Areas such as Kbuild, kallsyms, modpost, objtool, mksysmap, and the Rust build system were all affected by this work. In the end the build time improvements are very significant across a spectrum of hardware and different kernel configurations: More benchmark results and details can be found via