On Fri, Oct 27, 2006 at 09:23:00AM -0600, James Ranson wrote:
> I ran across a program called "Brook GPU" which compiles programs to
> be run by the Graphics Processing Unit.  Specifically, a
> floating-point FFT was one of the included sample programs, so I
> thought it might be possible to implement a Mersenne tester in the
> Brook language. Then, one could use his graphics card to test
> exponents, even if he were already using his CPU for same.

Unfortunately, the GPU is only capable of doing single-precision math; GIMPS
needs double-precision. (One could increase the FFT length, or try to work
around the limits in some other way, but I'd believe you lose whatever
advantage you got in the first place.)

That's not to say that the GPU isn't useful for a whole lot of FFT-related
tasks; I'm currently doing underwater acoustics simulations (semester project
for school) on the GPU, for instance, where the accuracy is comparable and
the speed is _much_ (think three orders of magnitude) greater. It's just that
GIMPS' demands might be a bit different from many others'.

/* Steinar */
-- 
Homepage: http://www.sesse.net/
_______________________________________________
Prime mailing list
[email protected]
http://hogranch.com/mailman/listinfo/prime

Reply via email to