[LAU] Testing JACK and PA latency

[Patrick Shirkey]

For comparison taking PA out of the graph:

  jack_iodelay (out) -> ecasound (in) -> ecasound (out) -> system (out) ->
system (in) -> jack_iodelay (in)


I see stable performance with the following result after a few seconds of
"burn in".

 218.905 frames      4.561 ms total roundtrip latency
	extra loopback latency: 26 frames
	use 13 for the backend arguments -I and -O ?? Inv

However after switching to the browser and writing this email then
switching back again I see the following data:


 33114.906 frames    689.894 ms total roundtrip latency
	extra loopback latency: 32922 frames
	use 16461 for the backend arguments -I and -O ??


Switching back to the browser to write some more I see the a stable
4.561ms again and that has stayed stable after a couple of switches
between browser/console. After more test the latency is definitely being
affected by xorg and switching between apps but not every time.

With only jack_iodelay running I see the cpu load hovering between 30 to
40%. With ecasound added to the graph I see cpu load hovering between 30
to 50%.

Ecasound has not reported any underruns during this test. With PA included
I had 32 underruns over night (approx 6 hours of running).

The things I am now trying to rule out or trace:

- Potential priority conflicts between jack and xorg (I'm running gnome3)
- What causes erratic results with PA in the graph


Suggestions welcome.


--
Patrick Shirkey
Boost Hardware Ltd