Benchmarking Wild vs Mold

A deep dive investigates why Mold's newly updated linker benchmarks show Wild performing much worse relative to Mold than Wild's own August 4th benchmarks did. The author identifies key methodology differences: Mold's benchmarks delete the output file and use ext4 with --no-fork, while Wild's benchmarks historically ran on tmpfs without deleting the output and with forking enabled. After matching Mold's configuration, results become much closer, revealing that Wild currently lacks OS-specific optimizations (fallocate pre-allocation, hugepage mapping) that Mold uses for fast file creation on non-tmpfs filesystems - fixes already made for Wild's next release. The remaining gap is partly explained by Mold's recent 2.42.0/2.42.1 speed improvements and by thread count differences (Mold's Threadripper test used 32 threads vs Wild's 128).

Nguồn: https://davidlattimore.github.io/posts/2026/09/18/benchmarking-wild-vs-mold.html. 8 Sync News chỉ tóm tắt và dẫn link; bản quyền nội dung thuộc tác giả và nguồn gốc.