C++ Mythbusters
summary
Claims tested and tooling recommended focus on measurable results and concrete fixes: printf-family uses global locale and a mutex so replacing it with libfmt yields big wins (Blender sees a 4X speedup on mac and ~20X on Windows), C++20 coroutines ship but need library support, std::regex compiles and runs slowly so use ctre or other fast regex libs, std::optional can silently force copies unless APIs match and monadic pipelines avoid that, std::move is a cast not an action and sinking-by-value changes ownership and performance.
AI-written summary · about
chapters (18)
- 0:00Talk overview
- 1:25C++ community challenges
- 5:48Origins and goals
- 7:33Evaluating C++ myths
- 10:03printf performance pitfalls
- 13:19C++ tooling ecosystem
- 18:35Regex performance
- 22:12Optional performance
- 27:58Using std::optional
- 34:26Debunking std::move myth
- 39:23Const reference vs sync
- 45:52Move semantics vs Rust
- 47:36Costs of excessive const
- 50:08Encapsulation and getters
- 52:16Data-oriented design
- 55:33Iterators vs ranges debate
- 57:15New C++ features
- 59:19Future topics
related talks
56mThrowing Tools at Ranges
01h 00mBack to Basics: (Range) Algorithms in C++
01h 02mUsing Modern C++ to Revive an Old Design
01h 01mGrowing Pains: Lessons from seeing C++ codebases mature
01h 05mBetter C++ ranges
41mOverengineering is the root of all evil
55mC++ Concurrency TS 2 Use Cases and Future Direction
01h 00mDesign Idioms from an Alternate Universe