kernel-microbenchmark

By vllm-project. Updated September 3, 2026. Categories: Debugging.

Build, debug, and interpret vLLM GPU kernel microbenchmarks for CUDA, Triton, and CuteDSL, including CUPTI timing, correctness checks, generated-code inspection, multi-GPU measurements, and SOL sanity checks.

View repository on GitHub

SKILL.md

The description above is the indexed summary. With JavaScript enabled, this page also loads the SKILL.md text from GitHub.