What problem does it solve? Go developers often draw wrong performance conclusions from single benchmark runs, noisy CI environments, or misread pprof profiles. This Skill provides a rigorous measurement methodology covering benchmark authoring, statistical comparison, profiling, and regression gating so optimization decisions rest on trustworthy data. ## Core Features & Use Cases - Benchmark Authoring: Write benchmarks using Go 1.24+ b.Loop(), memory tracking with b.ReportAllocs(), custom metrics via b.ReportMetric(), and table-driven sub-benchmarks. - Statistical Comparison: Compare before/after runs with benchstat, interpret p-values and confidence intervals, interleave runs to eliminate systematic bias, and avoid p-hacking traps. - Profiling & Diagnostics: Generate CPU, memory, and execution trace profiles from benchmarks, analyze escape analysis and inlining decisions with compiler flags, and investigate production issues with Prometheus runtime metrics. - CI Regression Detection: Gate pull requests with benchdiff, cob, or gobenchdata, mitigate noisy-neighbor variance on shared runners, and tune self-hosted runners for reproducible results. - Use Case: After optimizing a JSON parser, run go test -bench=. -count=10 on both versions, compare with benchstat, and paste the statistically significant results into the commit message to document the improvement. ## Quick Start Ask the AI to write a benchmark for your Go function using b.Loop() and then compare the optimized version against the baseline with benchstat.
