Performance matters - new tricks for old dogs
summary
Claims concrete performance wins and working prototypes live on pb.org: PFreorderRays implements swap/move-and-pop range erasure and reports averages like range erasure 12,000× faster and singular erase up to millions× faster for big elements; plf_list with chunked allocation shows large wins (sorting 77%–91% faster, clear millions× faster for trivial types); pfq and pfstack adapt block sizes and benchmark ~10–65% and ~80% faster respectively; sectors gives real O(1) operator[] across growing multiple blocks and the prototypes work.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:02PF reorder Rays evolution
- 6:48PF list explained
- 19:19PF IND sort
- 30:25PFQ faster adaptive Q
- 34:19PF stack
- 36:52Sectors and random access
- 47:46Time complexity reconsidered
- 57:40Summary
related talks
01h 00mMoving Faster: Everyday efficiency in modern C++
01h 04mDesign for Performance
51mDesigning for Performance
01h 05mPerformance Optimization in Software Development - Being Friendly to Your Hardware
40mLudicrous Speed: Designing For Performance on the JVM
46mHigh Performance .NET
42mPractical Optimizations
01h 29mWhen Nanoseconds Matter: Ultrafast Trading Systems in C++