Profile a program with stax (live CPU stacks, off-CPU waits, target/GPU spans, annotated disassembly) to find where time actually goes. Use whenever you need to measure performance, find hot functions, or answer "why is this slow" instead of guessing. Covers building with the right debug symbols, record → wait → flame → top → annotate, the flame-before-top rule, on-CPU vs off-CPU, and where the full reference lives.
2026-06-30