What Every Computer Scientist Should Know About Floating-Point Arithmetic

According to this post on reddit, Oracle has broken the links to "What Every Computer Scientist Should Know About Floating-Point Arithmetic", originally published in the March 1991 issue of the ACM's Computing Surveys. Oddly enough, that is the only issue of Computing Surveys that I own, having stolen a copy from someone at UTCS for just that article.

In any case, lest it be lost entirely in the depths of Oracle's documentation, I grabbed a copy of "What Every Computer Scientist Should Know About Floating-Point Arithmetic" without any permission whatsoever simply because everyone (not just computer scientists) who deals with floating-point numbers really ought to have some kind of grasp of the stuff in this paper.

(Of course, if Oracle sends me a cease-and-desist, I'll probably collapse like a soggy waffle.)

Check out Don't Store That in a Float for a good discussion about why you should know what you should know about floating-point arithmetic.

Orthogonality

This is a story about how bad code gets written by someone who knows better.