Re: panning
Re: panning
- Subject: Re: panning
- From: James McCartney <email@hidden>
- Date: Mon, 6 May 2002 10:07:56 -0500
On Sunday, May 5, 2002, at 04:42 PM, Doug Wyatt wrote:
That's the correct algorithm ... you might find that the code will run
faster by doing two or more samples per loop iteration in order to give
the compiler more room for optimizing with instruction scheduling.
Be careful unrolling loops. Metrowerks' optimizer will unroll loops for
you and if you've unrolled them yourself you may wind up with doubly
unrolled loops including two separate pieces of code to deal with the
remainder, one you wrote and one inserted by the optimizer - a real
mess. Either let the optimizer unroll them or unroll them yourself and
use a pragma to turn off the optimizer's unrolling.
Also I recommend to not use Duff's device.
The switch's case labels put entry points into the unrolled loop, across
which the optimizer cannot reorder instructions. It is better to use one
unrolled loop and one for the remainder. Then the optimizer can freely
reorder instructions inside the unrolled loop.
--
--- james mccartney email@hidden <
http://www.audiosynth.com>
SuperCollider - a real time synthesis programming language for the
PowerMac.
<
ftp://www.audiosynth.com/pub/updates/SC2.2.14.sea.hqx>
_______________________________________________
coreaudio-api mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/coreaudio-api
Do not post admin requests to the list. They will be ignored.