Using Floating-point in C++ - What Works, What Breaks, and Why

Explore the complexities of floating-point arithmetic in C++ through this comprehensive conference talk that demystifies when and why floating-point operations fail. Learn how different applications impose varying requirements on floating-point behavior, from performance-critical games to precision-demanding financial systems. Discover how floating-point operations are implemented across different C++ contexts and understand the counter-intuitive implications that follow. Examine how various environments balance correctness, predictability, and performance in subtly different ways. Analyze code examples that produce incorrect results or undefined behavior while remaining easily fixable, then progress to more complex calculations where IEEE-754, C++, and computational mathematics both restrict and assist your work. Gain practical insights on defining realistic expectations, structuring and testing floating-point code effectively, and understanding how libraries can help or hinder your efforts. Master the fundamental concepts needed to debug floating-point issues and understand the underlying mathematical and implementation reasons behind unexpected behavior in C++ floating-point operations.