The FFT-class should also support double precision via template parameter.
Is there any chance that this will be implemented?
(Otherwise I would have to make myself a minimal solution, but I would like to save me the time and double work
This would be really nice to have.
This would be nice not just for the FFTs, but for the juce Convolution too.
I am finding that large Impulse Responses with only 3 or 4 digits of actual resolution due to the floats processing to be really quite limiting. Okay for reverb yes, but not for high quality filtering.
And some FIRs, in combination with the partitioning of the zero or low latency mode of the convolution, result in large positive values being added to large negative values, which generates a disproportionate amount of noise and instability that can only be cured by using doubles.