Would it be possible to enable cuFFT to use streams? As far as I am aware as of Cuda 2.1 the cuFFT library only uses stream 0. My application is a multi-channel audio convolution plugin where I do a load, FFT, mul / add, store; everything but the FFT can be done asynchronously and I would like to get the whole lot into streams to help optimise the workflow.
The cuFFT source is no longer available I see (http://forums.nvidia.com/index.php?showtopic=59101), but even when it was it didn’t seem to be complete enough for me to make the changes myself. So it would be excellent if either the full source could be published so I can modify it myself; or if this is not possible, I would like to suggest that the stream functionality be added in to the cufftExec* function calls. If it’s not technically possible for some reason, it would be interesting to know why.