[LAU] timing penalty?

I've been conducting a couple of experiments related to timing, and in
particular a rapid program change sequence, between Rosegarden, using
Alsamidi, through A2jmidid, to Linuxsampler.

1.) When i use alsamidi in both RG and LS, timing is reasonable, but i
get the occasional blip if too many PChanges happen in quick
succession. (And as a relative scale, we're talking about a sequence
of 1/64 notes, with a PC for each one, and additional volume CC
changes as well, at about 140 bpm.)

2.) When i use alsamidi in RG, and a2jmidid through to LS jackmidi,
timing has improved, but still the occasional missed PC.

So using a2jmidid seems to be a step forward, but my question is, is
there a 'penalty' of some sort for using it in bridging format, and if
so, what is it?

3.) When  i use jackmidi native apps, running Muse and LS, for
example, the timing is good, and i don't seem to miss fewer PChanges.

(Which doesn't mean i'm giving up on RG. It has too many essential
tools that help me workflow wise.)

Final Question.
If we use an RT kernel (which i am, being 2.6.29-RT in 64studio 3.0),
with 1024HZ (or higher) for timing, what can we do after that, in
jack, a2jmidid, RG, LS, etc.., to ensure that midi timing (Alsa or
Jack) is pristine, and the combination of apps has the best change of
coping with rapid, and high volume, midi data being sent?

I'm trying to finesse the midi timing and midi data management across
the board, and get it as good as possible. (It's already good. This is
about the....extra 10%.)

I'm definitely in favour of using jackmidi based on my modest and
subjective experience, but i'd like to stick with RG for the power of
it's toolset for workflow. (Plus a thousand other good reasons to use
it.)

So if anyone has ideas, and tips, on raising the midi data flow
precision in transistion between the above apps, and tips or compile
flags to do this at build time, or runtime, i'd appreciate some
insight.

I've applied the usual tweaks across all the apps, and have Jack
running at 128/2/48000, with no xruns during normal operation.

Alex.

64studio 64bit 3.0 (beta 3), on an AMD64 5600+ dual core, with 4GB ram.
M-Audio Delta 44 soundcard.

Parchment Studios (It started as a joke...)
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@email-addr-hidden
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
Received on Sun Apr 12 00:15:02 2009