Explore high-frequency trading (HFT) and low-latency trading systems in C++ through a comprehensive performance analysis in this 48-minute conference talk. Dive into micro-optimization techniques, examining the quirks in generated assembler across different compiler versions. Understand static branch prediction, its applications, and the intricacies of switch statements in a simple FIX-to-MIT/BIT trading system. Compare performance across various operating systems and challenge common beliefs about Spectre and Meltdown mitigations. Gain insights into hardware impacts, including AMD Bulldozer architecture, and learn how to leverage C++'s optimization capabilities for performance-critical systems in the financial trading domain.
A Performance Analysis of a Trading System over Compilers
Overview
Syllabus
Introduction
Background
Its not clever
Boost
Hardware
AMD bulldozer
Meltdown Spector
Example Code
GCC 447
GCC 863
ICC Intel
Log Scale
Static Branch Predictor
Misprediction Penalty
Builtin Expect
GCC
ICCClang
Expect
Performance
Switch Statements
GCC ICC
Ifelse statements
Break
ICC
Is it unstable
Does it matter
Taught by
NDC Conferences
