Jeremy wrote: >> Having not looked at the Prime95 code for 3 or 4 years, how ingrained is the >> FFT? How feasible would it be to take the FFT code and have it run on the >> GPU (assuming that someone has written a fairly fast GPGPU FFT already). >>
John wrote: >its massive amounts of hand tuned i686 assembler.... Sorry, I didn't make it clearer. I didn't mean porting the FFT code to GPGPU, I meant replacing the FFT/iFFT code with one written using CUDA or whatever GPGPU equivalent. I'm sure that a good FFT library already exists for GPGPU applications, so it should be a matter of leveraging the existing FFT libraries in the LL tests right? In principal, it seems like it should be fairly straightforward to do... If I recall correctly, Prime95 does some fancy stuff where the subtraction takes place in imaginary space or something, but in a very general sense isn't it an FFT->Convolve->iFFT loop? Assuming a double-precision GPU FFT is written, then you could at least take advantage of the GPU along with the CPUs with a minimal amount of effort. Even if the GPU FFT library isn't very efficient and gets half the speed of the theoretical output of the GPU -- well, that is still equivalent to a quad-core Xeon (assuming the 8-core Xeon quote from nVidia is accurate) for a fairly minimal programming effort. -Jeremy _______________________________________________ Prime mailing list [email protected] http://hogranch.com/mailman/listinfo/prime
