PERFORMANCE Signal 335
Mold's Linker Benchmarks Show Substantial Performance Advantage Over Wild
↗
Illustration only Photo by Indra Projects on Unsplash
Comments
The benchmarking results indicate significant performance discrepancies between Mold and Wild linkers. Understanding the causes of these differences is crucial for engineers choosing a linker for their projects, as it affects build efficiency and resource usage.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Mold's benchmarks were run on different hardware and filesystem configurations compared to Wild's.
The performance differences are largely influenced by how output files are managed during the linking process.
Upcoming updates to Wild may address some of the performance gaps identified in recent benchmarks.
THE READ
What the cluster adds up to.
The recent benchmarking results have revealed that Mold outperforms Wild in several scenarios. Mold's benchmarks were conducted using a 64 core Threadripper and an Apple M1 Ultra, while Wild's tests were limited to a 16 core Ryzen machine. This disparity in hardware capabilities likely contributes to the performance differences observed.
A critical factor affecting performance is the handling of output files. Mold's approach of deleting output files between runs contrasts with Wild's previous practice of using existing files on tmpfs. This can lead to significant discrepancies in linking times, especially as many users may not operate on tmpfs, making Mold's methodology more applicable to real-world scenarios.
The benchmarks also highlight that Wild's performance can improve by adopting Mold's techniques, such as using ext4 and the --no-fork option. These changes are expected to be incorporated into future releases of Wild, which could help close the performance gap between the two linkers.
Written by elseif from the cluster below · checked for specifics the sources never containedTHE CLUSTER