Part II: A Performance Analysis of a Trading System over Compliers
summary
A simple FIX-to-MIT/BIT translator handles socket I/O, parses prices and quantities, and is used to measure full round-trip times over 100k orders; micro‑optimizations cut mean round‑trip from ~85 μs to ~50–55 μs but hardware choice drops it further to ~22 μs. Compiler behavior varies: GCC 5.3 shows odd regressions, GCC 7/8 significantly slower than GCC 6, and clang compiles ~7× faster but generates similarly poor code. OS and Spectre/Meltdown mitigations can add tens of microseconds (e.g. CentOS 6.9 ~90→125 μs).
AI-written summary · about
chapters (6)
- 0:00Start
- 33:09Hardware
- 36:14Spectra and Meltdown
- 38:03Centos
- 44:45Mitigations for Spectrum Meltdown
- 46:09References
related talks
01h 05mThe Speed Game: Automated Trading Systems in C++
01h 00mJust-in-Time Compilation
44mCompilation Speedup Using C++ Modules: A Case Study
38mOptimized String Processing in RISC-V: How Toolchain Improvements Can Boost Performance
01h 04mCPU design effects
51mBoosting Software Efficiency in C++: 100% Performance Improvement in a C++ Embedded System
01h 09mComputer Architecture, C++, and High Performance
01h 00mIntroduction to Hardware Efficiency in Cpp