Mac OS X currently uses gettimeofday() for measuring interval time units. This is probably much slower than using UpTime(), or mach_absolute_time().

Ideally we should read the "time base" in PowerPC processors. We will need to make sure that threads running on different processors in a multiprocessor system have a consistent interval timer.

I think both UpTime() and mach_absolute_time() are using the CPU timebase. I'm not sure what we need to do to ensure consistency between CPUs.

I've been informed (on the darwin mailing list) that mach_absolute_time() will give consistent results on dual CPU systems. Things would be pretty hosed if it didn't.

While looking around for time-related bugs I came across this (very old) one which has already been fixed so I'm closing it as a duplicate of the relevant bug.