OpenCL replacement for CUDA fft Library?

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello,

I would like to port my application from CUDA to OpenCL, but OpenCL doesn’t have a builtin fft library. Will there be a replacement available someday?

Regards,

Nils

2

Here’s my vote for an OpenCL FFT!

I know that Apple and NVIDIA have each developed a FFT on OpenCL, and haven’t released them. The reason for that is that they are internal implementations that aren’t as well-written as they could be. In the Apple case, they don’t want to release a FFT library that isn’t ready so people don’t develop against it preventing them from changing later.

So, if anyone has a FFT that they’ve implemented please post it! Especially if it’s 1024 R2C, I could really use that :)