Float and integer arithmetic follow two different paradigms
Summary
The article argues that float and integer arithmetic follow different paradigms and that developers often mistakenly apply integer-safety patterns to floating-point math. It explains how integer overflow leads to undefined behavior while floating-point operations can produce NaN or Inf, which propagate through calculations. It critiques epsilon-based guards for division, provides code examples in C and Rust, discusses GLSL limitations, and advocates using isfinite checks to make code more robust.