On Wed, Feb 8, 2012 at 19:55, Ron Frazier (NTP) <[email protected]> wrote: > It doesn't appear to me that -M is related to process priority, but to > internal scheduler timing. I tinkered around with the -P parameter a > little, but couldn't detect any changes.
-P is a no-op for the Windows port. Windows doesn't use a simple numeric priority scheme. The NT microkernel underneath does, but Windows exposes priority class, which is process-wide, and thread priority deltas (above normal, below normal, etc). The ntpd Windows port has always ignored -P, changing that would be a backwards-incompatible change for little to no benefit. As previously described, ntpd on Windows always raises the priority class as much as possible, and sets different threads' relative priorities appropriately (unlike POSIX ntpd, Windows ntpd has also long used threads, one for I/O completions, and if interpolating a higher-resolution clock using the performance counter, another to gather correlations between the two oscillators involved, and transiently for DNS lookups). More recently, POSIX ntpd moved to using threads where available instead of a forked child for otherwise-blocking DNS queries. Cheers, Dave Hart _______________________________________________ questions mailing list [email protected] http://lists.ntp.org/listinfo/questions
