2Sum Algorithms
Time for yet another floating-point blog post! Since this is my third blog post of the year, I’d like to keep this one brief. This time, I’ll be discussing “2Sum” algorithms, an obsession of mine for a number of years now, consuming a large part of an internship I did at Intel during the summer of 2023. The classic 2Sum algorithm1, attributed to Knuth and Møller, computes the floating-point addition and error term of two floating-point numbers exactly.
-
Attribution comes from the “Handbook of Floating-Point Arithmetic” by Jean-Michel Muller et al. ↩