Measuring Qt Performance
summary
Shows how to get real numbers: uses Creator and QTestLib to wrap code in QTestLib QBenchmarkLoop, runs median repeats and verbose runs, and switches from timing to callgrind instruction counts when helpful. Demonstrates a measured microbenchmark (string concat) falling from ~0.058ms to ~0.056ms after changing the code and raising QTestLib’s minimum from 20ms to 50ms. Builds an atomic-based read-write lock prototype that beats QMutex in uncontended reads, warns that allocations and threading make results noisy and that validation (QCOMPARE) must guard against accidental optimization.
AI-written summary · about
chapters (8)
- 0:00Introduction to Qt performance
- 2:38Performance measurement basics
- 8:40Writing a Qt benchmark
- 15:28Benchmarking options and tools
- 23:02Profiling benchmarks
- 33:26Optimizing string concatenation
- 33:47Benchmarking Qt read-write locks
- 55:15Q&A on Qt performance tools
related talks
22mOptimizing Qt Application
01h 02mCreating Performant Qt Quick Apps
01h 11m“Benchmarking C++ Code"
59mCounting Nanoseconds Microbenchmarking C++ Code
37mImproving app performance with benchmarking
05mLightning Talk: Adventures in Benchmarking Timestamp Taking in Cpp
55mEffective QML
42mPerformance Matters