From: Jens R. <pli...@ya...> - 2005-09-23 12:01:25
|
Hi Luis, I am now at work (nothing really to do right now), and I do not have access to my machine with PlanetCCRMA installed on it, so I cannot tell you the IRQ´s my machine is using, and which ALSA version is running on that machine. But what I know about my PlanetCCRMA installation is the following: - I am using the "other" kernel PlanetCCRMA offers for download (the kernel with "much better low latency performance", with the realtime preemption patch), which can be installed by "apt-get install planetccrma-core-edge" - I have "upgraded" both Jack and MusE. PlanetCCRMA comes with Jack 0.99.x and Muse 0.7.1, and I implemented Jack 0.100.0 (official release) and MusE 0.7.2pre2. If you want, I can tell you how I did that. All other software packages are existing in their latest versions (like ALSA), whenever I apply an "apt-get update" and "apt-get dist-upgrade". - I do not have any USB devices connected to my computer. > You still will get this stalled no-sound event-piling behaviour every > now and then, but less frequently. I do not use big soundfonts themselves, like FluidR3, but extracts of FluidR3. I use Smurf / Swami to cut out of FluidR3 the instruments that I need and save them in smaller individual sound files, at the time being for drums, bass, guitar (the latter shodddy, by the way: does anybody know good sounding free electric guitar soundfonts?). When I use these smaller soundfonts in MusE, the size of the occupied RAM is much smaller than with the original FluidR3 sound file. And the song gets loaded quicker ... :) > Anyway, I found out that if you start MusE with a priority higher > than jack, the chances that the software will behave as expected are more. In my PlanetCCRMA installation, Jacks start with prio 10, and Muse with 50 (=default value). I never tested other values, works nearly always fine for me. > [kompoza@membrillo ~]$ /etc/init.d/rtirq status > > PID CLS RTPRIO NI PRI %CPU STAT COMMAND > 303 FF 80 - 120 0.0 S< IRQ 8 [.........N/ 0] rtc > 1794 FF 70 - 110 0.0 S< IRQ 17 [........../ 0] > uhci_hcd:usb > 1815 FF 69 - 109 0.0 S< IRQ 18 [........../ 0] > yenta, uhci_hcd:usb > 1848 FF 68 - 108 0.3 S< IRQ 19 [........../ 0] > yenta, uhci_hcd:usb > 1707 FF 67 - 107 0.2 S< IRQ 23 [........../ 0] > uhci_hcd:usb1, ehci_hcd:usb > 1502 FF 60 - 100 0.0 S< IRQ 21 [.........N/ 0] > Intel > ICH > 1614 FF 59 - 99 0.0 S< IRQ 22 [.........N/ 0] > Intel ICH Modem, ohci > 398 FF 50 - 90 0.0 S< IRQ 1 [.........N/ 3] i > 15 FF 49 - 89 0.0 S< IRQ 9 [........../ 8] > acpi > 321 FF 49 - 89 0.0 S< IRQ 12 [.........N/ 0] i > 358 FF 46 - 86 0.0 S< IRQ 14 [........../ 2] ide > 1337 FF 44 - 84 0.0 S< IRQ 6 > 2640 FF 37 - 77 0.0 S< IRQ 4 > 2641 FF 36 - 76 0.0 S< IRQ 3 > 2662 FF 35 - 75 0.0 S< IRQ 7 [..P......./ 0] > parport > 2947 FF 34 - 74 0.0 S< IRQ 16 [........../ 0] eth I am not familiar with USB devices and the question, which IRQ´s USB devices should have, but did´nt I read somewhere that one should avoid USB devices if one wants to use them with real time capabilities? (Maybe I am saying total nonsense here). I could be that your USB devices run under a very low IRQ priority, lower than your hard drive. So, if you could, you should try a PCI sound card, and assign an IRQ with a reasonable high priority, see: http://sourceforge.net/mailarchive/forum.php?thread_id=1079993&forum_id=7073 http://www-128.ibm.com/developerworks/library/l-hw2.html > [kompoza@membrillo ~]$ /usr/bin/jackd -R -P65 -dalsa -dhw:2 -r44100 > -p256 -n3& Currently I have two machines with each a PlanetCCRMA installation, machine 1 has a Soundblaster AWE 32, the other one a SoundBlaster Live 5.1, and on both machines I use qjackctl to start Jack. The ideal settings in qjackctl seem to differ from soundcard to soundcard, so I guess that I cannot give you an advice about the best parameters for starting jack. But, as I think I can remember, the buffer size is 512 on both machines, on yours it is 256. And I think somebody else wrote via this mailing list that one should avoid buffer sizes < 512, at least with regards to MusE. Maybe you should increase that one to 512. > [kompoza@membrillo ~]$muse -P 70& I let MusE start with prio 50, and jack with prio 10. > Mmmmm, now that I see this... if you start MusE with the -D flag you > see > > watchdog set to SCHED_FIFO priority 99 > Thread <Midi> set to SCHED_FIFO priority 70 > > I wonder how this fits with my IRQ priorities. I have to make some > tests... I am not familiar yet with the SCHED_FIFO priorities, should I have a look for you at home which SCHED_FIFO´s my machines have? > Anyway, I also found out that when working with MusE it is wise to > start jackd from a console. I cannot confirm this. > If you start it from qjackctl and MusE > watchdog kicks in, it will bring down your whole X session. I upgraded Jack from 0.99.36 to 0.100.0 (official release), trying to prevent "MusE killing x server on loading tune...", see the mail from Robert Jonsson, Mon, 27 Jun 2005 21:28:23 +0200, via the MusE mailing list. > I am also not certain of who is it to be blamed, MusE or CCRMA, but > it is true that other sequencers with not so mythical but more flowery > names don't give so much trouble. I did not try Rosegarden yet in my PlanetCCRMA installations ... but compared to my very first MusE installation in my "Linux from Scratch" partition, MusE and Jack (Xruns) run reasonably fine and stable in the two PlanetCCRMA installations. Regards, Jens ___________________________________________________________ Gesendet von Yahoo! Mail - Jetzt mit 1GB Speicher kostenlos - Hier anmelden: http://mail.yahoo.de |