There are effective free open source FFT libraries for CUDA cards (along with BLAS implementations and other math goodies), so an FFT multiplication type extended precision arithmetic for CUDA cards seems practical, anyway. I suspect it is of no use for integers small enough for Karatsuba or Toom-Cook because of the copy time to and from the card, unless you are processing a large vector of them. The cards turn magical when you can do a huge number of identical operations in lock-step as separated by a fixed offset.
-----Original Message----- From: mpir-devel@googlegroups.com [mailto:mpir-devel@googlegroups.com] On Behalf Of Bill Hart Sent: Thursday, March 15, 2012 1:09 PM To: mpir-devel@googlegroups.com Subject: Re: [mpir-devel] Now that I am not on my mobile I can access the article and I see that they are comparing with GMP (though they don't say what version -- and it makes a big difference). [They also reference GMP as being written by the "GNU Open Source Community". The GNU people would not be thrilled about that. They don't use the phrase "Open Source" but "Free Software".] Anyhow, the GPU only seems to beat the CPUs when the multiplications are small, basically below the FFT range. Either way, the GPU is giving some speedup, assuming they compared with a recent GMP. So it is interesting. Bill. On 15 March 2012 19:46, Bill Hart <goodwillh...@googlemail.com> wrote: > Some of the techniques are probably very interesting, but it is hard > to be impressed by a six times speedup if the original library was > quite slow, which I bet it was. > > Thanks for pointing out the article. It will be interesting to follow > this development. > > Bill. > > > On Thursday, 15 March 2012, Dann Corbit <dcor...@connx.com> wrote: >> Very interesting article on multi-precision integer arithmetic ported >> to >> GPU: >> http://www.worldscinet.com/ppl/21/2103/free-access/S0129626411000266. >> pdf >> >> Not sure about the license, because it appears to be based on ARPREC >> and QD, which requires a fee for commercial use: >> http://crd-legacy.lbl.gov/~dhbailey/mpdist/ >> >> But I think at least it would be a good read because they talk about >> the obstacles and techniques that were necessary to get good performance. >> They saw up to 6x speedup for GTX 480 card verses core i7 870 general >> purpose CPU. Coincidentally, that is my exact hardware setup. >> >> -- >> You received this message because you are subscribed to the Google >> Groups "mpir-devel" group. >> To post to this group, send email to mpir-devel@googlegroups.com. >> To unsubscribe from this group, send email to >> mpir-devel+unsubscr...@googlegroups.com. >> For more options, visit this group at >> http://groups.google.com/group/mpir-devel?hl=en. >> >> -- You received this message because you are subscribed to the Google Groups "mpir-devel" group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en. -- You received this message because you are subscribed to the Google Groups "mpir-devel" group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
