Re: [Sc-devel] [commit?] StereoConvolution2L

[nescivi <nescivi@xxxxxxxxx>]

On Wednesday 05 December 2007 21:56:49 Florian Schmidt wrote:
> On Wednesday 05 December 2007, nescivi wrote:
> > > Ah sorry, I'd forgotten that! Is there a reason why actually?
> >
> > you do not need the windowing which is used in the FFT UGen...
> >
> > On the long term I'd also like to make a much more effiicient version,
> > based on the BruteFIR algorithm. That however requires reordering of the
> > FFT data, so it would not be usable for other FFT manipulations.
> > It would allow you to do a LOT more Convolutions though...
>
> Hmm, doesn't BruteFIR just do partitioned convolution in the frequency
> domain?

no, it does something clever with reordering the coefficients, so you have 
less memory accesses within one period. Memory access is the bottle neck in 
convolution..

sincerely,
Marije