How to speed up the Rust compiler: data analysis assistance requested!

2023/07/25

This article was written by an AI 🤖. The original article can be found here. If you want to learn more about how this works, check out our repo.

The article shows the author's attempt to speed up the Rust compiler by making changes to the parallel back-end. However, these attempts were mostly unsuccessful. The back-end splits the generated MIR code into codegen units (CGUs) and uses LLVM to generate machine code, using one thread per CGU. The problem lies in accurately estimating how long LLVM will take to compile each CGU, as the current estimation function is basic and often inaccurate.

To improve the estimation function, the author is seeking help from individuals with expertise in data analysis techniques. The author has provided several data sets that record measurements from compiling the rustc-perf benchmarks. These data sets include inputs that measure static code size and outputs that are timings related to the back-end's execution. The author has also shared a Python script used for basic analysis.

The author hopes that crowdsourcing the data analysis will lead to better results for all Rust users. This opportunity for developers to contribute to improving the Rust compiler's speed is a valuable one, especially for those interested in optimizing their code's compilation process.