Without running JACK and with running JACK -p2048 44.1Khz and 96KHz,
without HR timer enabled and with HR timer enabled ALSA MIDI used to
make a thru connection was always ok when playing live, not only for the
PCI MIDI, but even for USB MIDI.
When playing the recordings from my last test by Qtractor, there wasn't
the issue anymore that FluidSynth and the internal Linux recorded
FluidSynth sometimes were out of sync. The MIDI and the audio track were
always in sync. I only tested it with system timer.
Anyway, there definitive is jitter for the recorded FluidSynth. It's
inaudible jitter, resp. it is audible when loop playing, because some
beats are played to early, because of jitter with negative delay. So
it's not an issue of bad timing, but because of notes that shouldn't be
played, when loop playing. It's still a recording from the last test,
not from today (note, I'm not writing about the hw MIDI at the moment).
There must have been something bad, that is ok today. I wondered myself,
because I never had this issues before.
I didn't test MIDI jitter for hw MIDI today. Anyway, ALSA MIDI for MIDI
thru is ok and also Qtractor always plays the MIDI and audio track in
sync.
For the important jitter tests I'll switch back to 64 Studio 3.0-beta3,
there might be something broken for 64 Studio 3.3 alpha or it was just
bad luck, anyway at least for FluidSynth with 3.0-beta3 there never were
issues.
Over the weekend I had no time to test MIDI as it was planned. It likely
will be the same for the next days.
Again there's hope that PCI MIDI jitter will be inaudible. Because even
the USB MIDI was ok for live playing, I suspect that the jitter for the
recordings is not regarding to ALSA MIDI. I wish I would have the time
to do some recordings right now.