This GitHub app provides support for a streamlined benchmarking utility that maintains a persistent, dynamic dashboard of kernel performance across various backends, including Wave and IREE. It allows automation to trigger benchmarking CI runs after tuning requests as well as pull requests and merges to the Wave repository.

• triggers benchmarking runs after pull requests and merges to Wave repo
• pulls latest results from benchmark workflow artifacts
  - after each pull request, merge, and nightly
  - visualizes performance improvements between versions and changes across different backends (iree, wave, hipblaslt, ck, etc.) and kernel types (gemm, attention, conv, etc.)
• allows users to select kernels for tuning which automatically triggers a tuning workflow and saves optimal configurations for future runs
• persists a history of kernel performance over time between versions and changes