Re: [linux-audio-dev] jack in a live situation (LinuxSampler has this problem too)

Paul Davis wrote: Apparently it doesn't. I'm using jack-audio-connection-kit-0.98.0-2.rh90.ccrma and I have the same problems. LinuxSampler is producing short CPU peaks when lots of voices gets fired up and sometimes jack still kicks it out. Very annoying and if one wanted to use LinuxSampler live it would be impossible because you would never have the guarantee that jack does not kick it out, restarting jack and clients on stage is not an option :) So please jack developers check this issue and fix it in the next update it will be very useful for anyone wanting to use jack clients like softsynths,samplers, FXes etc in a live situation. A sporadic (let's say one each 10min or so) fragment xrun would be barely audible and would allow the live player to continue playing going almost unnoticed rather than telling the audience "I'm sorry jack kicked my synth wait a little bit so that I restart it so you can continue enjoying my performance :) ) LinuxSampler's jack callback is completely free from mutexes and bad syscalls, it could be improved by eg limiting the number of voices that can be started/released per audio fragment, that way the CPU spikes are lowered but as said you would never have the guarantee that jack does not kick it out. Sometimes when the OS is loaded (due to heavy disk I/O etc) it could delay the execution of jack clients too causing it getting kicked because jack thinks the client took too much to execute. LinuxSampler needs to work with small fragmentsizes, 64 - 128 frames so that you can play it lively so the kicking problem is more severe than in ardour where you can easily go with higher fragmentsizes (256-1024) without the user complaining too much. cheers, Benno http://www.linuxsampler.org