Benjamin von Ardenne

Great, that solved it. Now it's time for me to change the documentation. I'll commit the changes later this week.

While changing the documentation and the examples accordingly I came across the problem, that the input is being altered by the `num.rfftinv` function but not by the `num.rfft` function. They...

Okay, I got the same altered results on my machine which is very curious. I will indeed as the FFTW mailing list. Stay tuned.

Okay, so I got an answer, that was quick. The answer was hidden in the documentation of the 1d c2r function (they did not bother to mention that in the...

Hey Francesco, sorry for the delay, holidays, work travel, presentations and such delayed my additional modifications to the documentations a bit. I have now tidied up every remaining bit of...

No problem. Concerning the tests: I have tested each function for its capabilities of accepting input and rerouting it to the underlying FFTW functions. Since all functions are just wraps...

I would like to replicate the bug in my Windows-VM but before I do that I would like to cross check with you, how you compile gsl-shell on windows? Are...

One more thing: The algorithm work for complex matrices only, so whenever real matrices are given as input I transform the matrix to a complex one with: ``` matrix.cnew(n1, n2,...

BTW, did some performance tests with numpy doing 1e5 2D FFTs on 16x16 matrices. Python is about 2.5 secs, the shell performs at 4.2 secs. The code is even less...

So, to be honest, after all I have read about FFT, the GSL version is not the best implementation and FFTW almost always supersedes it. It is very tempting to...