Optimising a small real-world C++ application

on YouTube ↗01h 24m

" summary

A configurable C++ converter for COBOL/BCD fixed‑format files becomes usable for disaster‑recovery by profiling and iterative change: uses strace, perf, valgrind/cachegrind and Compiler Explorer to find hotspots, replaces string comparisons with enums, removes runtime division by reading BCD nibbles directly, and switches to memory‑map via Boost iostreams to cut system calls. Measured surprises: mmap drops system time by ~25%, simple loop/formatting fixes give ~4× speedup, a 2‑byte lookup table adds ~10%, and nine incremental versions produce a working, much faster tool.

AI-written summary · about

" chapters (13)

1. 0:00Introduction to optimizing
2. 10:19Performance optimization strategy
3. 23:56Initial implementation approach
4. 30:26Using Valgrind cachegrind
5. 40:50Profiling with perf tools
6. 45:24Optimizing string handling
7. 51:59Compiler Explorer insights
8. 1:00:03Loop and branch optimizations
9. 1:08:42Memory mapping input file
10. 1:13:09Two-byte translation table
11. 1:17:31Template specialization attempt
12. 1:21:08Compile-time translation table
13. 1:22:58Final optimizations and summary

" related talks

• 24m(Ab)using Compiler ToolsRéka Kovács · CppCon 2019
• 35mAlgorithmic and microarchitecture optimizations of C++ applicationsAlexander Maslennikov · NDC TechTown 2020
• 45mPerformance Benchmarking with CeleroJohn Farrier · CppCon 2017
• 49mAlgorithmic and microarchitecture optimizations of C++ applicationsAlexander Maslennikov · C++ on Sea 2020
• 49mAn Engineering Approach to Optimising C++Dave Rowland · C++ on Sea 2023
• 57mLessons from 20 years of Debugging C++ CodeSebastian Theophil · ACCU 2023
• 01h 00mMoving Faster: Everyday efficiency in modern C++Alan Talbot · CppCon 2018
• 01h 30mTuning C++: Benchmarks, and CPUs, and Compilers! Oh My!Chandler Carruth · CppCon 2015