[PD] IEMlib and compilation optimised for P4s

David NG McCallum david2003 at mentalfloss.ca
Fri Nov 7 10:37:03 CET 2003


I was using one of the bp filters, which I think is a wrapper for the 
[filter~] object, so not the FIR~.

I was just trying to build an EQ to replace a VST plugin I'm using right 
now. The problem was that my performance patch is heavily based on the 
use of delays and whenever the feedback got really small the filters in 
the EQ that it was ultimately going out just started freaking out.

<sigh> Back to VST...

. David McCallum
.  Music wants to be free
. http://mentalfloss.ca/sintheta

Nicholas Mariette wrote:
> Hi David,
> I've recently had the same experience of unexpected high-CPU usage from FIR~
> in Iemlib (under WinXP).
> Which Iemlib filter have you had this experience with?
> I found that I can run a maximum of about 8 FIR filters of 200samples length
> on a P4 2GHz.  This does seem a little low for such a short FIR.
> I tried recompiling Iemlib with MSVC++6.0 but I got no performance
> improvement.  I haven't yet looked at the code or tried special compilers
> like the Intel compiler.
> However, I have found that Ben Saylor's partitioned convolution external is
> somewhat more efficient for the same FIR filters.  If you're referring to
> FIR filters, I'd try that external which can be found here.
> http://home.gci.net/~pamsaylor/ben/partconv~-0.1.tar.gz
> And I'd be interested to hear of others' experiences with FIRs in PD.
> Anyone?
> Nick

More information about the Pd-list mailing list