On Tue, Nov 07, 2006 at 07:23:51PM +0100, Arjan van de Ven wrote:
> is it just me imagining things, or is the RHEL5 HZ value going to be a
> regression for userspace; the current rhel5 kernel behavior is that if
> userspace asks for a 1ms delay, it gets like a 7ms delay. In RHEL4, the
> max delay was 2ms...
actually, a bit worse it seems.
$ cat /etc/redhat-release
Red Hat Enterprise Linux AS release 4 (Nahant Update 3)
$ ./a.out
Unacceptable long delay; asked for 1000 usec, got 3208 usec
(similar runs are +/- <100 usecs
meanwhile on an FC6 box running a similar vintage kernel..
(don't have a rhel5 testbox handy right now) sees a lot more variance.
Anything between 3000us to 8229us. Interesting.
Dave
_______________________________________________
rhelv5-beta-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/rhelv5-beta-list
