Date: Dec 17, 2010 4:58 PM
Author: Alan
Subject: Re: IIR Filters - Matlab's Filter - Speed Improvements

"Royi Avital" <RoyiREMOVEAvital@yahoo.com> wrote in message <iegial$psl$1@fred.mathworks.com>...
> "Bruno Luong" <b.luong@fogale.findmycountry> wrote in message <ie1svc$p6r$1@fred.mathworks.com>...
> > "Royi Avital" <RoyiREMOVEAvital@yahoo.com> wrote in message <ie1mqg$ld0$1@fred.mathworks.com>...
> > > As I said, I'm trying to speed up the operation of a simple IIR filter. That's it.
> > >
> > > I was wondering if someone has any experience with that.
> > > Is it possible to beat Matlab's "filter" command with a MEX file?
> > > Anyone has ever managed to beat the filter command using other Matlab's built in functions?

> >
> > Hardly. There is no evidence to think Matlab filter is sloppy.
> >
> > Bruno

>
> No one has ever tried to improve it using MEX?
> How about using the GPU?


Doesn't Matlab's filter function implement the difference equation directly? If so, it might be faster to use fft and the convolution theorem, depending on the filter and signal lengths, and how often you need to recompute the ffts.