Can C++ Data Oriented Design Be ONE MILLION Times Faster?

Explore how data-oriented design in C++ can achieve extraordinary performance improvements of up to one million times faster through this 54-minute conference talk from CppCon 2025. Learn how refactoring data into Entity-Component formats enables cleaner implementations where lambdas process ranges and SIMD operations exploit contiguous data layouts. Discover that performance bottlenecks aren't design flaws but rather performance opportunities created by hardware engineers, and understand how to unlock these opportunities by engineering code that aligns with system state. Examine the crucial role of processor statefulness, from instruction cache and registers to L1 cache and memory access patterns, in achieving hardware-friendly code performance. Go beyond hardware optimization to explore software statefulness and how carefully engineering the sequence of operations and algorithms to align with both hardware and software states can achieve unparalleled optimization. Apply George Polya's problem-solving process and heuristics to tackle complex design problems systematically. Study real-world examples from machine learning and mathematical finance that demonstrate how addressing both hardware and software considerations can improve performance by orders of magnitude, including six orders of magnitude improvements. Gain actionable techniques and insights into the exploration process that can lead to similar performance gains in your own C++ development work, with practical approaches to making code more ecologically sound and efficient.