From: Christopher Sean Morrison <brlcad@ma...> - 2011-10-27 02:46:02
Thanks for your libfft patches you posted in August, but I was wondering if you saw my reply? If you don't have it, I had questions about the impact, benefits, and motivation behind the patch.
Clearly a benefit, it's nice to see how the code reduces by using the fftw3 library but I'd be a little concerned about the performance impact. Our "unrolled" library has been a dominant speed king for several decades, even if the convolutions were only for fixed sizes. How does it compare with your changes? Do you use our FFT tools?
I'm also very curious how in all that refactoring, you happened to come across bn_cx_div() as a replacement for cdiv() in fftfast.c? Impressive. ;)
Get latest updates about Open Source Projects, Conferences and News.