Re: ULE/SCHED_SMP diff for 7.0

From: Jeff Roberson <jroberson_at_chesapeake.net>
Date: Thu, 19 Jul 2007 01:56:53 -0700 (PDT)
On Thu, 19 Jul 2007, Dmitry Morozovsky wrote:

> On Wed, 18 Jul 2007, Jeff Roberson wrote:
>
> JR> > Works for me with MSI S420 lapton (core2duo T2400) on fresh -CURRENT/i386
> JR> > CPU: Genuine Intel(R) CPU           T2400  _at_ 1.83GHz (1833.41-MHz
> JR> > 686-class CPU)
> JR> >
> JR> > However, make -j3 buildworld buildkernel (trimmed GENERIC) time degraded
> JR> > by 7%:
> JR> >
> JR> > +------------------------------------------------------------------------------+
> JR> >    N           Min           Max        Median           Avg        Stddev
> JR> > x   5         36.33         36.77         36.45        36.486
> JR> > +   5         38.87         39.23         39.03        39.042
> JR> > Difference at 95.0% confidence
> JR> >        2.556 +/- 0.216322
> JR> >        7.00543% +/- 0.59289%
> JR> >        (Student's t, pooled s = 0.148324)
> JR> >
> JR> > before = kernel with SCHED_4BSD, after = SCHED_ULE with your patchset.
> JR> > Both
> JR> > are without WITNESS/INVARIANTS.
> JR>
> JR> Hi Dmitry,
> JR>
> JR> Can you test with http://people.freebsd.org/~jeff/sched_ule.c?  I believe
> JR> this will somewhat improve the situation with buildkernel.  I have tested
> JR> with my own core2duo laptop.  The first run after reboot is now about 7%
> JR> faster than before.  Subsequent runs are not improved as much.  Only 2-3%.
> JR> You can also try tuning sysctl kern.sched.steal_thresh down as low as 1 to
> JR> see how much this may help.  Unfortunately lower values tend to really hurt
> JR> other tests.
>
>
> Well, I'm a bit puzzled: with new sched_ule.c system+user time slightly
> decreases (within one minute), while real time goes up from 39 to 44 minutes!
> And most of the time I see 10-50% of idle in top.

Please try setting sysctl kern.sched.steal_thresh=2  This is the setting I 
intend to commit.  I also was only testing buildkernel which tends to run 
more processes concurrently as compared to buildworld.

Thanks,
Jeff

>
> Any hints?
>
> Sincerely,
> D.Marck                                     [DM5020, MCK-RIPE, DM3-RIPN]
> ------------------------------------------------------------------------
> *** Dmitry Morozovsky --- D.Marck --- Wild Woozle --- marck_at_rinet.ru ***
> ------------------------------------------------------------------------
> _______________________________________________
> freebsd-current_at_freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
>
Received on Thu Jul 19 2007 - 06:53:42 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:14 UTC