Maximising Throughput & Minimising Latency in High-Performance C++ Code

on YouTube ↗05m

" summary

Shows concrete data layout changes: pull hot fields into separate vectors, replace object-centric loops with struct of arrays vs array of structs transformations, and apply simple vectorized transformations to operate on contiguous memory. Demonstrates a hybrid pattern that keeps useful abstraction while reducing CPU cache misses and improving throughput, explains when the cache utilization wins justify extra implementation complexity, and gives practical before-and-after code patterns and tradeoffs for low-latency C++ systems.

AI-written summary · about

" chapters (9)

1. 0:00Intro: Why Performance Matters
2. 0:24Real-World Impact of Latency
3. 1:03Common C++ Data Modeling Patterns
4. 1:54Why Standard Object-Oriented Design Slows You Down
5. 2:22Pull-Based Transformation: Pros and Cons
6. 3:05A Hybrid Performance-First Approach
7. 3:54Optimizing Cache Usage with Vectorized Data
8. 4:36Takeaways on Data, Hardware, and Simplicity
9. 4:57Closing Thoughts

" related talks

• 01h 00mMoving Faster: Everyday efficiency in modern C++Alan Talbot · CppCon 2018
• 44mC++ Performance and OptimisationHubert Matthews · NDC TechTown 2017
• 01h 04mDesign for PerformanceFedor Pikus · CppCon 2018
• 35mAlgorithmic and microarchitecture optimizations of C++ applicationsAlexander Maslennikov · NDC TechTown 2020
• 57mThis Videogame Programmer Used the STL and You Will Never Guess What Happened NextMathieu Ropert · CppCon 2019
• 49mAlgorithmic and microarchitecture optimizations of C++ applicationsAlexander Maslennikov · C++ on Sea 2020
• 21mHPC Green Computing: Relationships Between Throughput and LatencyTullio Menga · Voxxed Days Milano 2019
• 58mRefresher on Containers, Algorithms and PerformanceVladimir Vishnevskii · C++ on Sea 2022