[alsa-devel] [PATCH] ALSA: pcm_lib: avoid timing jitter in snd_pcm_read/write()

[David Dillow]

Raymond, I've asked you before to not trim cc lists. Believe it or not,
people may be involved in a conversation even if they are not on the
mailing list. In this case, Hans probably doesn't care, but you cannot
make that assumption.

On Mon, 2010-06-28 at 08:37 +0800, Raymond Yau wrote:
> 2010/6/28 David Dillow <dave at thedillows.org>
> > Perhaps documentation that recommends setting avail_min to 1 when using
> > read/write to avoid this issue would be helpful, in lieu of changing the
> > ALSA code or giving concrete guarantees.
> >
> > Dave
> >
> >
> Refer to
> http://article.gmane.org/gmane.linux.alsa.devel/69333/match=avail_min
>
> avail_min cannot be less than period size , however set_avail_min() did not
> return any error when the value is less than period size and this mislead
> the programmers to believe that they can set avail_min to 1

They used to be able to, and it would work. You likely cannot get
latency that low, but anyone that was expecting for that to magically
change the way hardware works is fooling themselves. avail_min is like
the timeout argument to poll() -- you're guaranteed at least that value,
but you may get something much larger.

I understand the rationale given for patch b0b7d0 to alsa-lib, but
limiting avail_min in this way removed a workaround for the muddled
snd_pcm_read/write() semantics. I suppose it is still possible, if
someone wants to color outside libasound -- the kernel still allows it.

> if you want low latency, you need to use small period size

Small period sizes are often needed for low latency -- 10ms is common
for VoIP -- but if you are stuck with a device that can only do two
periods per buffer, the current read/write code will cause unstable
audio with frequent overruns. You can add another round trip to the
kernel with snd_pcm_wait(), but that's suboptimal and should not be
required.

You can get reasonably good stability on limited devices with avail_min
== 1, or with a read/write that doesn't force you to wait two periods
when trying to work with one.

Dave