30 Jan
2013
30 Jan
'13
11:44 p.m.
On Wed, Jan 30, 2013 at 07:14:11PM +0100, Robin Gareus wrote:
This works reliably; no x-runs:
jackd -d alsa -d hw:2 -r 48000 -p 64 -n 3
jack-delay measures: 403.949 frames 8.416 ms
Occasional rare x-runs with
jackd --sync -d alsa -d hw:2 -r 48000 -p 64 -n 2
roundtrip latency: 274.945 frames 5.728 ms
That's odd. The difference should be one period, 64 frames.
But it is 129 frames.
Now we're closing in...
jackd --sync -d alsa -d hw:2 -r 48000 -p 32 -n 2
roundtrip latency: 175.953 frames 3.666 ms
x-runs depend on system load. jack_capture still runs x-runs free for >
1 hour. Idle DSP load is only 5.6% with the CPU in low-power (1GHz). Yet
jack_delay alone bumps DSP load up to 20%.
The % load for jack_delay shouldn't depend much on the period size,
it just processes period_size frames each time.
If it does there could be significant wakeup jitter - jack measures
the worst case load IIRC.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)