News for hackers. Stuff that matters.
Re:Intel Underestimates Error Bounds by 1.3 Quintillion by jhallenworld on Friday October 10, 2014 @08:19AM (#8437321)
Related: here is a very good article and comments on estimating transcendental functions: http://www.coranac.com/2009/07/sines/

I'm annoyed that one of the links in the comments is broken: there was a page where you entered your function and it provided a chebyshev polynomial to generate it to a particular precision.

EDIT: well here is another one: http://goddard.net.nz/files/js/chebyshev/cheby.html

Re:Intel Underestimates Error Bounds by 1.3 Quintillion by edwintorok on Friday October 10, 2014 @08:48AM (#8437400)
There is some interesting info here as well on how to compute most of libm (including transcedental functions) in a correctly-rounded/precise way: http://lipforge.ens-lyon.fr/frs/download.php/153/crlibm-1.0b...
Re:Intel Underestimates Error Bounds by 1.3 Quintillion by jhallenworld on Friday October 10, 2014 @09:05AM (#8437443)
Also, the old standard reference for this was "Computer Approximations" by Hart and Cheney