Software Practices

Optimize Measured Hotspots

PERF-OPTIMIZE-MEASURED-HOTSPOTS

PerformanceVerificationChange Shape
Give feedback

Summary

Optimize code that measurement shows is on the important runtime path. This keeps review attention on changes whose user-visible impact justifies altering the code shape.

Rule

Optimize measured hotspots, not interesting code.

Why

Speeding up code that runs once, is not on the critical path, or is not visible to users usually wastes review attention. A tiny percentage win in a hot loop can matter more than a large percentage win in cold setup code.

Helps

  • Focuses optimization on code whose runtime contribution justifies changing its shape.

Limits

Small cleanup that also improves performance is fine when readability does not suffer. Call it a cleanup unless measurement shows performance is the reason to land it.

Agent Instruction

Optimize measured hotspots, not interesting code that runs once, is off the critical path, or is
invisible to users.

Mechanisms

Supported by profiling, flamegraphs, tracing, workload counters, benchmark selection, and PR notes that identify why the target is hot.

References