FFT again



?iga Lenar?i? wrote:

> Maxima is a very old piece of code with a lot of different  
> contributors
> working on a fairly opaque code base. There is understandable  
> resistance
> to changing things which work, even if they provide speedups, because
> the goal is to not give wrong answers, and it can be hard to determine
> what the effect of changes will be.

Of course. By my tests results are exactly the same ( 'lmax' on the  
absolute of difference gives 0.0, for random complex data), but I  
can't prove that, can I?

Even if there were any problems, I'd be willing to spend my time to  
correct it and in the worst case, the commit can be undone. Since  
fft.lisp is rather selfcontained, there's not much more to it, than  
loading the new fft.lisp and comparing results. The results are the  
same.

Regards,
Ziga