News for hackers. Stuff that matters.
Re:Intel Underestimates Error Bounds by 1.3 Quintillion by nroets on Friday October 10, 2014 @03:49AM (#8436677)
This is a storm in a tea cup. "The absolute error ... was fairly constant at about 4e-21, which is quite small. For many purposes this will not matter"

Good programmers know that most floating point calculations introduce small errors and find ways to make their programs cope.

Re:Intel Underestimates Error Bounds by 1.3 Quintillion by lvh on Friday October 10, 2014 @03:54AM (#8436689)
That's pretty bad for a double that you're about to do a bunch of operations with. It's pretty easy to throw numerical precision away.