21 Jan
2004
21 Jan
'04
7:41 a.m.
At Wed, 21 Jan 2004 03:01:19 -0800,
Andrew Morton wrote:
a few months ago before 2.6.0 was released, i tested the audio latency
with different kernels. the result was shown in the internal
conference in SUSE, and i totally forgot to release the data until now
:)
maybe it will show you other aspects.
here you can find the slides
http://www.alsa-project.org/~iwai/audio-latency.pdf
some notes:
- the red line of the graph is the buffer size - when the plot
overcomes this line, the buffer underrun occurs, aka sound
dropout.
- performed on Athlon 2200+ / 256MB / SB live
- the kernel tested at that time was slightly old, 2.6.0-test10,
so the result with the latest kernel may differ.
- suse 2.4 kernel runs with HZ=1000 and O(1) scheduler
--
Takashi Iwai <tiwai(a)suse.de> ALSA Developer - www.alsa-project.org
I just wanted to elaborate abit; -R in the
argument line indicates that jack
should be running SCHED_FIFO, and since there are no specific settings for
period-size and number of periods, jack is probably defaulting to 2-periods
and 1024 samples period size. 2.4 works rather reliably with period-sizes
quite a bit smaller than that. I think 256 is commonly used, with good
hardware+drivers I suppose you could go even lower.
I don't know what these numbers mean...
What does this transalate to in interval-in-milliseconds, and after how
many milliseconds delay will an underrun occur?
And what else is happening on the system at the time?