I want to run audio FFT transformations on a GPU using, possibly, OpenCL. What are the best optimization strategies for:
- converting audio signals to FFT;
- transfer them to the graphics card;
- compute those
- return the FFT into audio signal domain
if possible, i would like to do such a thing:
- with minimum latency possible;
- in real time;
- using really big fft window sizes;
- using really high sample rates and bit depth
