Wild链接器项目对Mold最近发布的基准测试结果进行了分析，发现两个项目的测试配置存在显著差异1。Mold的测试采用ext4文件系统、删除输出文件、使用--no-fork参数，而Wild历史测试则使用tmpfs文件系统、保留输出文件、默认fork行为1。通过采用Mold相同的测试配置，Wild团队在M1 Ultra架构上成功复现了Mold的基准测试结果1。
Mold在2.42.0和2.42.1版本获得了主要性能提升，这两个版本发布于Wild最后一次基准测试（2026年8月4日）之后1。Mold的最新基准测试于2026年8月28日发布1。Wild团队表示已实现fallocate文件预分配和hugepages映射技术，相关优化将在下一个版本发布1。
The Wild linker project has scrutinized recent benchmark results published by Mold, which appeared to show Wild's performance significantly lagging behind its competitor 1. Upon investigation, the Wild team discovered substantial differences in how the two projects configured their tests, including variations in output file handling, filesystem types, and startup parameters 1.
When Wild researchers replicated Mold's test configuration—using ext4 filesystem, deleting output files, and employing the --no-fork parameter—they successfully reproduced comparable results on M1 Ultra architecture 1. By contrast, Wild's previous benchmarks from August 4, 2026, had utilized tmpfs filesystem, retained output files, and relied on default fork behavior 1. Mold's most recent benchmarks were released on August 28, 2026, revealing that versions 2.42.0 and 2.42.1 achieved substantial performance improvements—releases that came after Wild's last published benchmark 1.
The Wild team has indicated plans to implement additional optimization techniques, including fallocate file preallocation and hugepages mapping, which are scheduled for inclusion in an upcoming release 1.
评论
还没有评论，欢迎留下第一条。