[llvm-dev] Making Clang/LLVM faster using code layout optimizations

[Maksim Panchenko via llvm-dev]

We’d like to share how Clang could be made up to 15% faster using code layout optimizations implemented in BOLT – our open-source LLVM-based post-link optimizer. These improvements are achieved on top of PGO and LTO. We thought the results and the practical value will be interesting to the LLVM community. The detailed step-by-step guide of optimizing Clang is available at:

https://github.com/facebookincubator/BOLT/blob/master/docs/OptimizingClang.md

While we think that there is room for improvement in LLVM’s code layout algorithm, we strongly believe that, to achieve the maximum benefits for large applications, a post-link optimizer is needed. We would like to know what people think about including BOLT into LLVM family of tools.

More details about BOLT are available in our arXiv.org paper at: https://arxiv.org/pdf/1807.06735.pdf

Maksim & Rafael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20181016/4dd3e47a/attachment.html>