What do cufft do different in computing the fft as opposed to MATLAB? I have an algorithm that uses several fft’s, which I’m converting to the GPU from MATLAB. But in one of the fft’s, when cufft and MATLAB gets the exact same inpu vector, they return completely different results. This is a forward fft, so no scaling have to be done after that. I am really confused and need your help to figure this out.