Issues and pull requests are welcome.
Except for changes where the expected performance impact is negligible, contributors should run benchmarks both with and without their changes (for example by comparing their branch with main) and include the results in the Pull Request description.
Since some of the benchmarks are susceptible to runtime/system noise, it can be beneficial to increase the benchmark duration via the -benchtime flag.
A GitHub Action will automatically run all benchmarks and post them as a summary to the respective action run. If a benchmark exceeds the baseline by 1.5x, a warning comment will be generated on the PR. Refer to this workflow for more information.