Are there any plans for double precision support for FFT?

I’m testing at the moment on CUDA using the single precision FFT library and it soon starts to go wild as, I guess, the errors start to accumulate for large numbers of points. Using FFTW on the host I don’t get this problem. When is double precision FFT support due?

thanks
Mark

Unfortunately, we still do not have plan to support double precision CuFFT library in the near future.

Khanh

This is not true, support for double precision FFT is in the roadmap.

That’s good to hear. Can you give me some timescales.

thanks

Mark

I am also interested in double precision fft. When will it be available?