bisect-perf-regression

Automate Git bisect to locate the commit introducing GPU kernel performance regressions.

260|104|Updated Nov 23, 2025

Search Tags:#automation #git #regression #benchmark #gpu #bisect

Ready to run

npx skills add https://github.com/ROCm/FlyDSL --skill bisect-perf-regression-rocm

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: bisect-perf-regression
Source: https://github.com/ROCm/FlyDSL/tree/main/.claude/skills/bisect-perf-regression
Command: npx skills add https://github.com/ROCm/FlyDSL --skill bisect-perf-regression-rocm

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Automate the identification of the exact Git commit that introduces GPU kernel performance regressions by performing an automated git bisect across a commit range and benchmark command.

Core Features & Use Cases

Automated Git bisect over a commit range to locate the regression.
Executes a user-provided benchmark command on each candidate commit and extracts a numeric performance metric.
Generates a comprehensive regression report including the offending commit, diff, and a root-cause hypothesis.
Supports optional build steps, multiple benchmark runs with median aggregation, and environment-safe stashing when needed.
Useful for continuous integration and performance regression debugging in GPU kernel pipelines.

Quick Start

Bisect a repository to locate the exact commit that introduces a GPU kernel performance regression using a benchmark command.

Frequently Asked Questions about bisect-perf-regression

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I find the exact commit that introduced a GPU kernel performance regression?▼

To find the exact commit that introduced a GPU kernel performance regression, you can automate a git bisect across a given commit range using a benchmark command. The process checks out commits, executes repeated benchmark runs with median aggregation, and reports the offending commit with its diff.

What is the best way to automate git bisect for benchmark testing?▼

Automating git bisect for benchmark testing involves executing a user-provided benchmark command on each candidate commit to extract a numeric performance metric. The automated process applies threshold-based classification to identify regressions, optionally running build steps and stashing environment changes safely.

Can I use git bisect to locate GPU kernel regressions if my codebase requires a build step?▼

Yes, you can locate GPU kernel regressions with a build step by specifying optional build commands during the automated git bisect. The process safely handles environment-safe stashing and executes the required build before running repeated benchmark checks on each commit.

Does automated regression reporting include the code diff for the offending commit?▼

Yes, automated regression reporting includes the code diff for the offending commit. The generated comprehensive report identifies the exact commit that introduced the regression and provides a root-cause hypothesis based on the benchmark results.

How do you classify performance regressions during an automated git bisect?▼

Performance regressions during an automated git bisect are classified using threshold-based criteria applied to a numeric performance metric. The process extracts this metric by executing a user-provided benchmark command and aggregating multiple benchmark runs using the median.

When do I need automated git bisect for GPU kernel performance monitoring?▼

You need automated git bisect for GPU kernel performance monitoring when debugging continuous integration pipelines or identifying the exact commit that introduces a performance regression. It applies to any codebase with GPU kernels where automated benchmarking is possible.