Jacob Wrote:
>All this is discussed extensively at
>http://www.mersenneforum.org/showthread.php?t=10275

Thanks for the link to that thread, but I don't see any mention of GPUs
capable of double-precision FP operations in that thread at all. It was all
talk of why current GPUs aren't capable of speeding up the current software.

Reading some more info on the GTX280, nVidia is claiming that their double
FP output is equal to an 8-core Xeon. If their double-precision really does
adhere to the IEEE spec and they aren't taking any shortcuts, I don't see
any reason why you couldn't port to a CUDA-based FFT. Seeing that GPUs are
now at the point of becoming feasible targets for LL testing, wouldn't it be
worthwhile to start looking into it? (Especially at the price/performance
point?)

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).



_______________________________________________
Prime mailing list
[email protected]
http://hogranch.com/mailman/listinfo/prime

Reply via email to