Re: Calibrating Buffer Size and Count

On Saturday, May 29, 2010 9:28:34 pm Kip Warner wrote:
> On Fri, 2010-05-28 at 13:29 -0700, Chris Robinson wrote:
> > So, for instance, you would generate 4 to 8 buffers, and fill them each
> > with frequency/4*channels*bits/8 bytes (where frequency, channels, and
> > bits are taken from the stream data format that's given to OpenAL):
> 
> The 4 in the divisor, is that the number of buffers one chose to
> generate?

It's actually how much of a second, per buffer, to fill (ie. 1/4th of a second 
in each buffer; 2 would be half, and 8 would be 1/8th of a second, etc). With 
a one-second long buffer queue, it's the same as the number of buffers to 
generate.

> How do I determine the number of buffers to use for a stream? Do you
> reckon to just hardcode it at, say, 8, and instead leave the size of the
> buffers themselves as the parameter to fiddle with?

It's pretty arbitrary. Generally speaking, it's probably not a problem to 
hard-code the buffer count (eg. 4 or 8 or so), as long as you make the buffer 
size flexible depending on the stream format.

What I tend to do, when simplicity is called for, is hard-code the buffers to 
3 or 4, and specify a buffer size of 16384 sample frames (which is multiplied 
by 2 or 4, depending on the decoded format, to get the byte size). This should 
work fine as long as the stream doesn't use a really high sample rate (like 
96khz or more).

> > int framesize = stream->channels * stream->bits / 8;
> > BytesPerBuffer = stream->Freq / 4 * framesize;
> 
> The 4 again in the divisor is the number of buffers chosen for the
> stream?

It's the same as above. How much of a second's worth of data to fill each 
buffer.

> PS If you remember, please cc me so I get it faster than the batch mode
> dispatches.

Sure. Sorry about that.
_______________________________________________
Openal mailing list
[email protected]
http://opensource.creative.com/mailman/listinfo/openal