Optimizing Go Code Without a Blindfold
summary
Shows concrete tools and numbers: raw benchmarks fluctuate up to 4–5% noise so tiny wins like 1.3% disappear, the workflow uses benchstat and benchcmd/benchcmp to collect multiple samples and report p‑values, and uses perf lock to cap CPU turbo so throttling stops inflating variance. Demonstrates compiler diagnostics (inlining, escape analysis, BCE) and a live change to a code generator that removes nested unused variables, which cuts a 2.7s codegen run by about 5% in practice.
AI-written summary · about
chapters (4)
related talks
14mOptimizing Go code without a blindfold
20mOptimizing Go code without a blindfold
40mTwo Go programs, three different profiling techniques, all in 40 minutes
51mObscure Go Optimisations
50mMADPIG: Measure and deepen performance in Go
55mTo Make Code Fast, All You Need Is… ☠️
45mBoost application performance with profile guided optimization
44mGo Profiling and Observability from Scratch