On 28/01/2023 09:03, Volker Rümelin wrote:
Am 22.01.23 um 19:13 schrieb Mark Cave-Ayland:
On 15/01/2023 13:45, Volker Rümelin wrote:
Am 15.01.23 um 14:08 schrieb Volker Rümelin:
Ccing a few more people who might be interested in this patch series.
@Mark:
After this patch series, the code in your out of tree ASC audio device (and a few
in tree audio devices) could be simplified. write_audio() and the loops calling
write_audio() could be removed.
Hi Volker,
I know we have discussed this in a separate thread off-list, but this is
fantastic!
Just out of interest, if the available bytes wraps the circular buffer will the
audio core call the audio callback twice to maximise the ability of the guest to
generate samples before the next audio timer? Or does that not make much difference
in practice?
Hi Mark,
I guess with circular buffer you refer to the mixing engine buffer. The audio system
calls the callback once on every audio timer event. If the available bytes wrap the
mixing engine ringbuffer, the audio_pcm_sw_resample_out() function uses two writes to
write all available bytes. Compared to the unpatched version, nothing has changed in
this regard. Of course the audio frontend devices are still free to write 'avail'
bytes with multiple calls to AUD_write().
With best regards,
Volker
Yes that makes sense, thanks for confirming this. I'm sorry that I'm not familiar
enough with the audio side to do a proper review but obviously the A-B still stands
and I would certainly be keen to see this merged.
ATB,
Mark.
