[gmpi] Re: Decision Time: 7.1.1

  • From: Matthew Xavier Mora <mxmora@xxxxxxxxx>
  • To: gmpi@xxxxxxxxxxxxx
  • Date: Fri, 13 Jun 2003 18:10:48 -0700

At 11:14 AM -0400 6/10/03, RonKuper@xxxxxxxxxxxx wrote: 



What about non-trival recursive DSP, such as multichannel delays with
feedback, or multichannel IIR EQs?  The claim for these is that the cost of
the DSP so far exceeds the cost of swizzling, that these processes might as
well be done on uninterleaved streams.

Answering this claim is best handled with a benchmark, I think.  If a modern
CPU can do a vector multiply or a vector add in a couple of cycles, and the
same CPU pays 20+ cycle penalties for cache miss, at what point does the
cost of swizzling become significant?

I don't have any first hand experience in this. Does anyone on the list?

I think designing a plugin API by worrying about how processor x works today is pretty short sided. In the not to distant future, processor y will be out. And what was fast on processor x may now be a performance problem on processor y.

Premature optimization someone said is the root of all evil. :-) Remember it is algorithm design that usually wins out over faster iron. Design the API so that it will work good on any processor and then profile and improve it later. By the time this API is finished (at the going rate of discussion), there will have been many new processors released. :-)



Matt



----------------------------------------------------------------------
Generalized Music Plugin Interface (GMPI) public discussion list
Participation in this list is contingent upon your abiding by the
following rules:  Please stay on topic.  You are responsible for your own
words.  Please respect your fellow subscribers.  Please do not
redistribute anyone else's words without their permission.

Archive: //www.freelists.org/archives/gmpi
Email gmpi-request@xxxxxxxxxxxxx w/ subject "unsubscribe" to unsubscribe

Other related posts:

  1. Home
  2. gmpi
  3. Archive
  4. 06-2003