On Mon, Jan 31, 2005 at 09:51:55AM +0000, Steve Harris wrote: > On Mon, Jan 31, 2005 at 10:15:56 +0100, Alfons Adriaensen wrote: > > On Sun, Jan 30, 2005 at 06:55:45PM -0600, Jack O'Quin wrote: > > > > > > > In the case of JAMin it was dictated by our desire to use FFTW rather > > > than writing our own FFT and trying to figure out how to divide the > > > computation into relatively equal parts. > > > > For Jamin, you have two channels, each one needs a Fwd and an Inv FFT, > > so that would already be 4 approx. equal parts (assuming the FFT dominates > > the work to be done). > > Actually there is on fwd, and 3 inv per channel, each xover band is a > seperate inv FFT.
So that should enable you to divide the work into 8 pieces. (down to period_size 32 already, and counting...) -- FA
