Profiling with perf: where the CPU goes

perf is the main CPU profiler on Linux. It samples the processor through a hardware counter (the PMU) every N milliseconds and records stacks. The result: where function X spends the most time. load-average tells you how busy the system is; perf tells you on what exactly.

In this lesson you go through the minimal workflow:

1. Start a CPU-bound load
2. perf record takes samples
3. perf report shows the top functions
4. You learn how to read the result

⚠️ Caveat: PMU sampling on a VM (the OrbStack VM) can be limited: some events may be unavailable. Basic CPU-time profiling works.