Re: pg_test_timing tool for EXPLAIN ANALYZE overhead

On Wed, Feb 22, 2012 at 18:44, Greg Smith <greg(at)2ndquadrant(dot)com> wrote:
> As far as I've been able to tell, there aren't any issues unique to Windows
> there.  Multiple cores can have their TSC results get out of sync on Windows
> for the same reason they do on Linux systems, and there's also the same
> frequency/temperature issues.

Not on recent Linux kernel versions. Linux automatically detects when
the TSC is unstable (due to power management or out-of-sync
cores/sockets) and automatically falls back to the more expensive HPET
or ACPI methods.

e.g:
% dmesg |grep -i tsc
[ 0.000000] Fast TSC calibration using PIT
[ 0.164075] checking TSC synchronization [CPU#0 -> CPU#1]: passed.
[ 0.197062] Switching to clocksource tsc
[ 0.260960] Marking TSC unstable due to TSC halts in idle

Whether these tests cover 100% of the possible conditions, and whether
the detection has race conditions or not, I don't know.

Regards,
Marti