Pardon Rui :) Gustin and I are clueless because of HPET. Do you know any link? Or is there an understandable answer because of the hpet_max_user_freq issue? Just ignore this mail, if it's to hard to explain, sorry :S. I still try to optimize Qtractor for usage with external equipment.
Gustin Johnson wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Ralf Mardorf wrote: > >>>> I do know that MIDI jitter by Linux to equipment external the computer >>>> will decrease when using HR timer instead of System timer. I don't know >>>> the impact of the value for hpet_max_user_freq. >>>> >>>> >>> What is the difference? >>> >>> >> http://www.mail-archive.com/[email protected]/msg01458.html >> >> > To be more specific, define for me "System Timer" and "High Performance > Event Timer" (HPET) and how they are different. > > Interestingly Windows XP does not support HPET even if the hardware is > equipped with it. > AFAIK the system timer is using RTC and it's a programmable timer that interrupts the hardware periodically, while HPET won't interrupt the hardware in such a static way, because of the much higher resolution there will be less conflicts because of interrupts, dunno how this works. I guess to google for "tickless timer" might give some answers: http://www.google.de/#hl=de&ei=bJo2S7L5NdH5_Aar05iMCQ&sa=X&oi=spell&resnum=0&ct=result&cd=1&ved=0CBoQBSgA&q=tickless+timer&spell=1&fp=6e1a413bb204dd20 I didn't read the first link, but this one seems to be interesting and I guess I'll read it some day: http://kerneltrap.org/node/6750 Dunno if it's set up for my XP, but I once read a howto enable it for XP. It was just one short line to add somewhere. I guess you need service package 3. I searched, but I didn't found the link: http://www.google.de/#hl=de&q=HPET+XP+sp3&meta=&aq=f&oq=HPET+XP+sp3&fp=6e1a413bb204dd20 There's a discussion about this issue: http://forum.slysoft.com/archive/index.php/t-22236.html Somewhere is a thread or Wiki about HPET on different OSs and IIRC there is described how to enable it for XP. But if it should be enabled, it's unclear if ASIO will use it. >> I can't remember if I used 64 or 1024 for hpet_max_user_freq, but IIRC >> on LAD there was the advice to build the kernel with using 64 by >> default, even if I should have used 1024, 64 might be the better value. >> > > That does not seem like good advice if we are talking about the same > thing. > A shot in the dark: The frequency always is 1,000,000,000 Hz for HPET timers, while the maximal frequency for system timers is 1,000 Hz. It might be that hpet_max_user_freq is a value how precise an interrupt has to be forced. If I guess correctly ;), HR timers don't force it when there are no interrupts needed, while a normal timer always does this periodically. I can imagine that a value of 64 does mean, that there are some kind of ticks for the tickless system ;), e.g. ... this might be complete nonsense ... 1,000,000,000/64=15625000, while 1,000,000,000/1024=976562.5 ... thus the higher value, will cause a less precise resolution for sync. I also have a theory the other way around ;). We should ask this on LAD, instead of puzzling over this issue. _______________________________________________ 64studio-users mailing list [email protected] http://lists.64studio.com/mailman/listinfo/64studio-users
