On Thu, 16 Jul 2009 01:20:27 +0200 <hollunder(a)gmx.at> wrote: I usually don't do that, but here's another reply to myself, so others don't have that many issues: JACK_RT doesn't work at all, JACK seems to work, haven't tried OSS. It seems the presets are not included but a few examples are, you just need to locate and load them. I didn't push it because it's late but so far it ran fine.

On Wed, 15 Jul 2009, Paul Davis wrote: defining JACK enables support for a reference implementation of the JACK drivers for Zyn. It has a major shortcoming in that there is a big mutex lock happinging in the process() cycle that isn't quite RT-safe. defining JACK_RT enables the alternate JACK implementation that employs a ringbuffer to avoid The Big Lock. However, it has apparently never worked. -gabriel

On Fri, 17 Jul 2009 11:59:44 +0000 (GMT) Frank Kober <goemusic(a)yahoo.fr> wrote: True. I just tried and pushed and in normal use cases there's nothing to complain about. Changing patches doesn't cause it to disconnect from jack anymore. I often hear a slight crackle but that doesn't register as xrun. I pushed it hard using for example Folderol's HyperMatrix patch which is very demanding. With this, as well as some other patches, it crackles sometimes, without xruns. When you really push it over the limit it crackles heavily and produces xruns, but that's to be expected when there's no more cpu left to burn. Here's where I just saw that an rt kernel makes quite a difference: with stock 2.6.30 kernel I can play far less voices and once it gets near 100% cpu zynaddsubfx's jack backend is gone, forcing me to restart the application. So yeah, very good, the show stopping bugs seem to be mostly gone, it's usable again, which is actually just great :) Philipp

On Mon, 20 Jul 2009 23:03:36 +0000 (GMT) Frank Kober <goemusic(a)yahoo.fr> wrote: This is being very intensively worked on. There is a temporary fix, which cures the defect completely, but partially returns to the RT unsafe situation. I found it worth doing, for the sake of all the other benefits. In src/Output/JACKaudiooutput.cpp change the folowing lines: =================================================================== if (!pthread_mutex_trylock(&jackmaster->mutex)) { JACKhandlemidi(nframes); jackmaster->GetAudioOutSamples(nframes,jack_get_sample_rate(jackclient),outl,outr); pthread_mutex_unlock(&jackmaster->mutex); } else { memset(outl, 0, sizeof(jack_default_audio_sample_t) * nframes); memset(outr, 0, sizeof(jack_default_audio_sample_t) * nframes); } =================================================================== to ================================================================= pthread_mutex_trylock(&jackmaster->mutex); jackmaster->GetAudioOutSamples(nframes,jack_get_sample_rate(jackclient),outl,outr); pthread_mutex_unlock(&jackmaster->mutex); ================================================================= Also, an annoying little detail is that all the controllers are enabled by default except main volume (controller 7) which is, of course, the one I *always* want! The change below cures that, and I believe will be rolled into the next mantenance release. In src/Params/Controller.cpp Change: "volume.receive=0;" To: "volume.receive=1;" -- Will J Godfrey http://www.musically.me.uk