Re: Format mismatch
Re: Format mismatch
- Subject: Re: Format mismatch
- From: kelly jacklin <email@hidden>
- Date: Mon, 4 Oct 2004 09:12:49 -0700
On Oct 4, 2004, at 8:54 AM, Mark's Studio wrote:
My problem is that my processing is altivec unrolled by 8 so im
skipping samples,is there a smart way to buffer the data and only call
my altivec function when there is enough frames ?
As for the SRC at the end, im not sure it's the best solution since
the processing can't compensate for the artifacts introduced by the
SRC.
The point is arguable, but regardless as an AU, you cannot control
where the host does SRC, so you must be prepared for this situation in
real host apps.
Furthermore there are devices that actually have non-altivec-grained
buffer frame counts (for example certain video/audio muxed devices,
which have a 160{1,2} cadence samples per frame, and pull one buffer
per frame). You need to be able to handle non-altivec-grained
buffers...
The way this is typically handled is for the altivec code to run
through loops for the altivec-grained (or even greater, if your
processing is unrolled) part of the buffer, and then run through a
scalar version for the remaining samples. Buffering the data would
involve artificially introducing latency into your AU (which you must
report correctly, for hosts to be able to compensate properly for the
latency you are imposing)...
Assuming you will always be presented with altivec-grained buffers will
fail with shipping hosts and drivers today.
kelly
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden