Use snd_pcm_avail_delay() in pa_alsa_safe_delay() so that we can check the delay
value against the avail value and patch it up when it looks invalid. Only do
this for capture.
Move the code to start the capture and the smoother closer together to improve
smoother accuracy.
Rework things to look more like the alsa sink where the device is started in
only one place.
Currently when rewinding alsa, a fixed value of 256 bytes is used,
which represents 1.33ms @ 48kHz (2ch, 16bit). This is typically fine
and due to DMA constraints we would not want to rewind less than this.
However with more demanding sample specs, (e.g. 8ch 192kHz 32bit)
256 bytes is likely not sufficient, so calculate what 1.33ms would
be and use which ever value is bigger.
Discussed with David Henningsson and Pierre-Louis Bossart here:
http://thread.gmane.org/gmane.comp.audio.pulseaudio.general/7286
Rewinding the ring buffer completely causes audible issues with DMAs.
Previous solution didn't work with tsched=0, and used tsched_watermark
for guardband, which isn't linked to hardware and could become really high
if underflows occurred.
Added separate parameter that can be tuned to hardware limitations and size
of DMA bursts.
The smoother is paused when the device is suspended but never resumed on
unsuspend. Pass the paused = FALSE flag to the pa_smoother_reset() call to make
it unpause when unsuspending. This patch improves source timings quite a bit.
This is required to when playing on a52: device, rewind is broken
in those plugins.
Credits to Michael Rans <mcarans@yahoo.co.uk> for finding this
workaround, and Tanu Kaskinen <tanuk@iki.fi> for providing
valuable feedback.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@intel.com>
This is not 100% ideal as we have not way to tie specific boosts to specific
inputs and this particular chipset (as noted in #772) appears to
support just that.
For the time being incorporate it into the normal boost logic.
See http://pulseaudio.org/ticket/772
In virtual machines sound card clocks and OS scheduling tend to become
unreliable, adding various 'uneven' latencies. The adaptive algorithm
that handles drop-outs does not handle it this well: in contrast to
drop-outs on real machines that are evenly distributed, small and can
easily be encountered via the adpative algorithms, drop-outs in VMs tend
to happen abruptly, and massively, which is not easy to counter.
This patch simply disables timer based scheduling in VMs reverting to
classic IO based scheduling. This should help make PA perform better in
VMs.
https://bugzilla.redhat.com/show_bug.cgi?id=532775
If an element does not control some channels assume they are 0dB in
comparison to the other elements, i.e. do not influence the volume at
all. Previously we were assuming they were as high as the highest of the
channels we do control.
- As discussed on alsa-devel it's probably better to initialize the
buffer size first, followed by the period size. If that fails try the
other way round. If that fails try to configure only buffer size. If
that fails try to configure only period size. Finally, try to
configure neither.
- Don't require integral periods anymore.
Both of these changes should help improving compatibility with various
weirder sound devices, such as TV cards.
