[MPlayer-dev-eng] ALSA issue, r29549

[Uoti Urpala]

On Tue, 2009-09-29 at 17:12 +0200, Reimar Döffinger wrote:
> On Tue, Sep 29, 2009 at 05:51:41PM +0300, Uoti Urpala wrote:
> > On Tue, 2009-09-29 at 16:28 +0200, Reimar Döffinger wrote:
> > > Not sure about the deeper issue, but the following patch seems to fix it

> > If so have you tried
> > using oprofile to measure where the CPU use is?
> 
> Given that nobody succeeded in reproducing the issue, that is rather
> hard to do.

I interpreted your "seems to fix it" to mean that you had managed to
test the change on an affected machine. But I see you got an user on the
mplayer-users list to test it.

Based on the mails there I think rounding to just whole samples should
be tested (and not outburst like your patch does). One possibility is
that with multiple channels the data size gets large enough that max
unit size checks elsewhere trigger, that size is not a multiple of
sample per channel, and so alsa gets the data in multiple chunks which
do not individually consist of complete samples.