Python Performance Optimization

Measure first. Optimize the bottleneck that matters. Keep correctness tests around every performance change.

For advanced patterns such as NumPy vectorization, caching, memory management, parallelism, async I/O, database optimization, and benchmarking tools, read references/advanced-patterns.md.

Intake

Collect:

• workload and input size;
• target metric: latency, throughput, memory, CPU, I/O;
• current measurement and desired target;
• environment and hardware;
• correctness tests or golden outputs.

Profiling Order