Open JayKickliter opened 10 years ago
Only single precision is supported at the moment, so --enable-single currently doesn't do anything.
Thanks. Do you know if double-precision is in the works? I see you've contributed to the library, is there any kind of roadmap for the project? I'm looking for a non-GPL alternative to FFTW. I really don't need all the bells and whistles of FFTW, but the ability to operate on both 32 and 64 bit would be terrific.
BTW: In the little test I just did (calling FFTS from Julia), FFTS is significantly faster than Julia's default FFTW interface: 0.7 ms vs 1.2 ms. Not news to you but great work.
I don't think there is a roadmap as such. From the discussions I've had with Anthony AVX(2) has higher priority, though double precision has been brought up a few times.
For now I think the recommended solution is to use sfft, which is the precursor to this library. The only caveat to using that library is that the transform size must be known at compile time, as it does not use runtime code gen like ffts does.
When running
./configure
with--enable-single
enabled, does that preclude the ability to perform an FFT on a double precision data? I'm sorry if this is a stupid question, but I can't find an explanation of the options anywhere.