On Thu, 8 Mar 2007 14:03:22 +0000
Andrew Doran <ad@netbsd.org> wrote:

> On Thu, Mar 08, 2007 at 07:45:52AM -0500, Matthew Mondor wrote:
> 
> > $ time ./clocktest1 1000000
> >     8.35s real     0.52s user     7.81s system
> > $ time ./clocktest2 1000000
> >     8.50s real     0.52s user     7.96s system
> > $ time ./clocktest3 1000000
> >     3.40s real     0.29s user     3.09s system
> 
> Those numbers don't make sense to me. What CPU and what release were
> they done on? From an Athlon MP @ 1GHz running 4.99.13:
> 
> $ time ./clocktest1 1000000
>     1.36s real     0.17s user     1.13s system
> $ time ./clocktest2 1000000
>     1.35s real     0.08s user     1.23s system
> $ time ./clocktest3 1000000
>     4.15s real     0.19s user     2.55s system

More numbers compiled with -march=i686 -O2 on 3.1 on IBM Intel 866MHz
desktop:

$ time ./clocktest1 1000000 
    1.32s real     0.20s user     1.12s system
$ time ./clocktest2 1000000 
    1.29s real     0.16s user     1.12s system
(no kern.hardclock_ticks on 3.1)

And again tests on the previous 900MHz laptop but now compiled with
above CFLAGS:

$ time ./clocktest1 1000000 
    7.47s real     0.50s user     6.95s system
$ time ./clocktest2 1000000 
    7.48s real     0.32s user     7.14s system
$ time ./clocktest3 1000000 
    3.43s real     0.25s user     3.15s system

So the theoretically faster CPU doesn't appear to be with -current but I
unfortunately can't test with 3.1 on that laptop now without some
trouble.

-- 
Matt