Carl Cook from Optiver delivers a presentation on low-latency C++ techniques used in high-frequency trading, emphasizing the importance of measurement and understanding compiler behavior over deep C++ knowledge. He covers various optimization strategies, including branch prediction, template-based configuration, lambda functions, object pooling, exception handling, and custom hash table implementations. Cook also discusses system tuning considerations like hyper-threading and core isolation, highlighting the need to minimize jitter and maintain cache efficiency. The presentation concludes with a discussion on accurate benchmarking methodologies using production-like environments and network packet analysis, followed by a Q&A session addressing topics such as Turbo Boost, L3 cache partitioning, profile-guided optimization, and the use of built-in expect for branch prediction.
Sign in to continue reading, translating and more.
Continue