We discussed whether to use float instead of double.
For fft, it may not be a good idea (because of variable multiplied by values large numbers of times). For the all other operation (symmetrization, scalar product, deconvolution per se) both are fine.
We discussed whether to use float instead of double. For fft, it may not be a good idea (because of variable multiplied by values large numbers of times). For the all other operation (symmetrization, scalar product, deconvolution per se) both are fine.