Super-Callifragilistic-Expealido-phat-cious!! finally, we can do real MIDI sequencing with Linux: while (1) { ... usleep (usecs_till_next_event); /* we're know its the right time, instead of being rounded to the default HZ */ deliver_events (); } actually, i think you need a different sleep()-alike to do this, but the point remains the same. i just hope that this change reflects the whole enchilada (reprogramming the APIC) rather than just the POSIX clock API. is there any sign of this?

Paul Davis writes: Is this really new ? I was porting some C++ thread classes (originally developed for Solaris) to 2.4.19 this week. The ITC mechanism uses pthread_cond_timedwait(), which takes a struct timespec * referring to an absolute time, the same you can get with clock_gettime(). I had the impression this gave me almost microsecond resolution timing (but I did not look closesly at the precision at the time). -- Fons Adriaensen ALCATEL SPACE

Steve Harris writes: I know... but when I ran the unit tests I didn't look to closely at the accuracy, only to happy that it was working at all :-). I'll very this next week. -- FA

Paul Davis writes: Seems to be 20 ms on my unpatched standard SuSE 2.4.19. I imagine this can be lowered by modifying the kernel config ? Yes, in my first test I used something like 150 microseconds, and the machine froze for 15 seconds (100000 iterations). No problem with that. TNX for the info. -- FA