25 Oct
2003
25 Oct
'03
12:33 p.m.
lördagen den 25 oktober 2003 15.28 skrev Mathias Lundgren:
....
I've run hydrogen feeding it midi via MusE and
jack as audio output. When
using -p 512 it sounds OK, even though it's possible to hear that the
timing isn't accurate. When increasing the period, -p 2048, it's really
horrible. I've glanced at the midi input thread sources in hydrogen and
noticed there are a few mutex thingies there. I've tried increasing the
real time priority of the midi thread, with no luck. I guess jack's
callback function (audio processing) and midi input thread are locking each
other out, and when the buffer gets bigger it results in greater
inaccuracy. When running hydrogen with alsa output in the same scenario as
above, there's no timing problems (but then I can't use it together with
any other apps). The above scenario applies to Zynadd as well.
Correcting myself. I've run both zynadd and hydrogen fed by muse, with muse
running without audio at all, or jack output. zynadd & hydrogen using Oss.
Zynadd runs like a clock in both cases (muse w/without jackaudio so the
problem shouldn't be MusE's midi-feeding), when it comes to sync, but
hydrogen doesn't. The inaccuracies in hydrogen are even worse with jack with
large period than with hydrogen using the Oss-driver, but there are
inaccuracies even when running hydrogen with Oss, when fed midi from muse.
/Mathias