Closed storopoli closed 2 years ago
Would pretty printing some outputs to a log be a solution or do we need something more fancy such as a static website or an automatic performance report comment? Static website is gonna be difficult since this repo already has Documenter in the gh-pages branch and automatic comment is gonna be annoying because bot comments are always annoying
Maybe we can add one issue about performance and automatically add comments to that issue with the benchmark results and a ref to the PR or to a commit on a branch. That way people can unsubscribe from the issue and not get notifications but still see results by clicking on the issue cross-ref. Also, the benchmark history is easily visible by scrolling through the issue comments.
EDIT: Nah. Nevermind. Logs make the most sense because they can easily be accessed from PRs and commits. LoopVectorization also just uses the logs as far as I can tell
We use BenchmarkCI in KernelFunctions: https://github.com/JuliaGaussianProcesses/KernelFunctions.jl/pull/399
Thanks for the nice improvements, @devmotion. It’ll be great if we can track performance regression for TuringGLM (e.g. using a notebook similar to tutorials), but that can be done in a separate PR.
Originally posted by @yebai in https://github.com/TuringLang/TuringGLM.jl/issues/45#issuecomment-1055540632