Re: [Alsa-user] Reproducible pops on some hardware

[Clemens L. <cla...@go...>]

Bryan Ischo wrote:
> OK I think I figured out a way:

This is how mplayer does it.

> Immediately after preparing a PCM device, and before doing any writes to
> it, I call snd_pcm_avail() and record that number (in my case, it is
> 16384) as "maxValidAvail".

This is the buffer size.

> Immediately before any call to snd_pcm_writei(), I again call
> snd_pcm_avail().  If the result is greater than maxValidAvail, I call
> snd_pcm_forward(pcm, snd_pcm_avail() - maxValidAvail).
>
> I *think* that what this accomplishes is to forward the write pointer
> within the ring buffer to the read pointer.

Yes.  (And the read pointer continues to move, so you'll have to write
really fast.)

> I don't really understand why snd_pcm_avail() goes greater than the ring
> buffer size when silence has been played out

It's the number of frames you have to write to completely fill up
the buffer.


Regards,
Clemens