News for hackers. Stuff that matters.
Re:Improving Floating Point Accuracy: A Look at Sums by eru on Monday October 19, 2015 @08:51AM (#10412516)
Don't common ways to implement interval arithmetic massively overestimate the inaccuracy?
Re:Improving Floating Point Accuracy: A Look at Sums by rjmunro on Tuesday October 27, 2015 @01:47PM (#10459782)
Compared to what?

You can improve accuracy by optimising the way you do your calculations. A really trivial example: 10000 + .00001 - 10000 is less accurate than 10000 - 10000 + .00001. Without interval arithmetic, you can't easily measure this and decide which way round is best.