30-year-old Xerox paper remains definitive guide to floating point math

Franc0Fernand0 · x · 2026-08-18

A tweet highlights a paper by David Goldberg from Xerox PARC, written over 30 years ago, as the best explanation for floating point arithmetic errors. The paper covers different floating-point formats, how rounding errors occur and how to avoid them, the IEEE 754 standard, and the messy reality of binary-decimal conversion.

Original post →

More from Research

Research channel →