bisect-perf-regression

To find a GPU kernel performance regression, this Skill automates Git bisect by running user-defined benchmark commands across a commit range. It extracts performance metrics to pinpoint the exact commit responsible for the slowdown.

Git bisect works for performance issues by using a binary search algorithm across commits. This Skill automates executing your benchmark command at each step, measuring performance metrics, and narrowing down the commit range until it identifies the regression.

Yes, you can use a custom benchmark command to identify a code slowdown. You provide the specific command, such as './run_bench.sh', and the Skill executes it across the commit range to extract the performance metrics needed for regression analysis.

The best way to analyze a performance drop is to identify the offending commit and examine its code changes. This Skill reports the regression commit, analyzes its diff, and suggests potential root causes for the degradation.

Yes, you need to provide a specific commit range to start debugging performance regressions. You must specify a starting commit, such as a known good baseline, and an ending commit like HEAD to execute the automated bisect process.

Your application rendering slowdown is likely caused by a specific code change introducing a performance regression. This Skill automates finding the exact commit responsible by executing benchmarks across your commit history and analyzing the code diff.