[alsa-devel] [PATCH - dmix v3 0/1] pcm: dmix: Align slave_hw_ptr to slave period boundary

Takashi Iwai tiwai at suse.de
Tue Nov 6 15:27:10 CET 2018


On Tue, 06 Nov 2018 13:48:08 +0100,
Timo Wischer wrote:
> 
> On 10/30/18 14:21, Takashi Iwai wrote:
> > Yes, and is this correct?  Suppose you start a stream at the position
> > one sample before the next period (psize * N + (psize-1)). Then
> > slave_hw_ptr is psize * N.  At the next moment, the dpcm->hw.ptr
> > reaches to psize * (N+1), and snd_pcm_dmix_sync_ptr() gets called.
> > Then slave_hw_ptr will be updated to psize * (N+1) although only one
> > sample has been processed?
> 
> I have created a spreadsheet to simulate the behavior of dmix (see
> attachment). I used this sheet to test different corner cases with
> different implementations.
> Column B-D describes 3 corner cases with the original
> implementation. Column E-H describes the corner cases with patch v2
> applied and column I-L describes the corner cases with patch v3.
> With patch v3 I was not able to find a corner case which would
> fail. All our internal audio test are also passing fine.
> 
> The corner cases are described in the form
> N*period+period-1
> (N+1)period+1
> (N+2)period+1
> 
> The first line describes the value of dmix->spcm->hw.ptr when
> snd_pcm_dmix_start() will be called.
> The second line describes the hw_ptr when poll() returns for the first
> time and the third line describes the hw_ptr when poll() returns for
> the second time.
> 
> In column B is the issue case described when the patch is not
> applied. snd_pcm_avail() returns 2 frames only (B27) but the buffer
> only contains 3 frames (B18).
> Therefore up to 5 frames are available (a buffer size of 8 frames is
> choosen).
> 
> The issue case of patch v2 is shown in column H. snd_pcm_avail()
> returns 7 frames (H46) but the buffer contains still 3 frames
> (H37). Therefore there are only 5 frames free to overwrite.

Hrm, sorry, it's still not clear to me what you've tested and what
these items represent.

> Do you see any corner cases which I missed or any other drawbacks?

I guess that the biggest issue is the understanding of PCM period
wakeup; let me cite the description of a part of your patch:

 "But since 3 snd_pcm_period_elapsed(), 3 periods should be available
  and it should have been able to write.
  If rsnd_pointer during the start was 0x120 which is 3 periods
  then 0x248 - 0x120 =  128 it could go on with write."

This assumption can't be applied.

However, the current implementation of dmix is designed to achieve the
lowest latency by putting the data at the exact position being played
back now.  This, of course, doesn't guarantee the period wakeup = one
period free to fill like the above.  So it's the design choice.

The current code has a workaround for the case of nperiods <= 2, by
setting the initial slave_appl_ptr to the next period start.  This
guarantees that the period wakeup = one period to be filled.  But its
cost is the start latency; the playback doesn't start immediately but
wait until the next period start.


Takashi


More information about the Alsa-devel mailing list