I've not had time to screw around with your test program, but I've noticed that the clock skews are almost exactly one second back. This is can be compensated for pretty easily, and the results _look_ appropriate.
Curious, but that's life. Maybe I'm grabbing it right before it's updated the second? I'll look at rdtsc as a solution, too. -DMZ On Wed, 2005-11-09 at 10:59 -0500, Rob wrote: > On Tue, Nov 08, 2005 at 03:23:19PM -0500, David Zakar wrote: > > Does anyone have a good workaround? This is utterly killing some latency > > testing I'm doing, since I can't trust the results. > > FWIW, if you are doing packet latency testing, use the timestamps you get > from capturing through libpcap instead: gettimeofday is too inaccurate > for <10ms timings. > > - Rob > . >
