[haiku-development] Re: HDA
- From: pete.goodeve@xxxxxxxxxxxx
- To: haiku-development@xxxxxxxxxxxxx
- Date: Wed, 7 Nov 2012 00:11:01 -0800
On Wed, Nov 07, 2012 at 12:32:39AM -0500, Sean Collins wrote:
> Pete Goodeve wrote:
> >
> >There's just one glitch left for me... latency. Haiku HDA is using a
> >couple
> >of 10240-sample buffers, which adds up to about 1/10 sec. Way too slow
> >for the real-time stuff I want to do. This puzzles me a bit, because Sean
> >reported a while back on how pleased he was by HDA's very low latency!
> >It's certain;y not inherent, because Ubuntu on the same laptop has
> >unnoticable latency when driven from USB-MIDI.
> >
> >It's the next thing I have to look into, but auich on my previous machine
> >had a settings file where the buffers could be specified. Is there any
> >reason this wouldn't work for HDA too? (It might be preferable to have
> >this adjustable. There might be a trade-off between latency and
> >reliability on different machines.)
> >
>
> You will need to change the buffer sizes in the driver source files and
> recompile the driver, there is currently no way to adjust them, would be
> nice if you could do it with the media prefs
Yep -- this is what I intend to do. (Actually in the multi_audio code,
where a buffer-setting option is available.) Looks straightforward, but
I got diverted by the listsem glitch... (:-/) But did *you* do that to
get the short HDA latency? I thought you had it out of the box.
A Media Prefs option could be nice, I think. Might not be too hard, either,
as the rate is already settable there. Otherwise, I looked at how auich
does it (with a settings file), and it's again straightforward. Thinking
about it, the latter might be better, as otherwise it might be too easy to
mess up the buffers.
-- Pete --
Other related posts:
