Floating-Point Error Handling in C++: What Actually Works

Part of

Computational Accuracy And Arithmetic Pitfalls