Performance Optimization & Profiling

Measure before optimizing — gprof or perf (Linux) show where a program actually spends its time, rather than guessing.

High-value, C-specific things worth knowing once a real bottleneck is found:

• Compiler optimization flags (-O2, sometimes -O3) let the compiler restructure code for speed — always measure the optimized build, since debug builds (-O0) can be dramatically slower.
• Cache-friendly memory access — looping over a 2D array row-by-row (matching how it's actually laid out in memory) is often noticeably faster than column-by-column.
• Avoiding unnecessary malloc/free calls in a hot loop — allocation has real overhead; reuse a buffer instead of reallocating it every iteration where possible.

Common mistakes

• Benchmarking a debug build (-O0, with sanitizers enabled) and drawing conclusions about real-world performance from it.