Linux 7.4 Starts Taking In Patches To Speed Up Kernel Builds

The Linux 7.4 kernel has begun receiving patches from Arm engineer Lorenzo Stoakes aimed at speeding up the build process. In full, this work is said to cut kernel compile times quite substantially.

Where it came from

The effort has been underway since early September. Its starting point was a curious observation: an AI brought in to tackle the problem produced plenty of, to put it mildly, hideous code — yet it helped pinpoint the major bottlenecks standing in the way of faster compilation. The patches clear up numerous single-threaded chokepoints and other issues, for both the GCC and the LLVM/Clang compilers.

By the figures cited, the complete set of changes makes kernel builds roughly 36% faster and incremental builds nearly 70% faster. The patch series was previously examined in detail in a dedicated Phoronix piece about approaching a ten-second kernel build.

What has landed

As part of the Kbuild changes in Linux 7.4, an initial set of improvements has been merged. These include avoiding allocating .modinfo in vmlinux, speeding up kallsyms, eliminating unnecessary sorting, emitting relocations only on the final vmlinux link, re-implementing mksysmap in C, caching more cc-option and ld-option results, and using pigz where available for faster Gzip compression.

The author notes that once Linux 7.4-rc1 is tagged, he will run fresh build benchmarks to gauge the impact of the changes merged so far.

Source: Phoronix