Overengineering is the root of all evil
summary
Claims concrete wins: measured microbenchmarks with BenchmarkDotNet show a clean OOP version is ~20% faster than LINQ but a table-driven rewrite runs in 15% of baseline time (about 7× faster), simple vectorization using Vector<T> yields ~25× improvement, and adding parallelism hits ~100× on four cores. The talk demos code before/after, explains vectorization and thread splitting, and gives exact ratios and practical patterns to reproduce — worth watching if numbers and runnable tools decide you.
AI-written summary · about
chapters (16)
- 0:00Welcome
- 1:04Hardware speed improvements
- 4:21Software slower despite hardware
- 5:03Premature optimization
- 8:12Necessary performance work
- 10:19Sports app startup case study
- 12:27BenchmarkDotNet performance
- 15:26Clean code vs performance
- 20:00Polymorphism area example
- 24:54Switch vs clean code performance
- 27:48Table-driven optimization
- 29:14Baseline results
- 31:20Vectorization gains
- 35:52Parallelization scaling
- 38:00Key takeaways
- 40:10Q&A
