CPUTYPE and friends, from 'make.conf' benchmark

[Oliver Pinter]

On 12/5/11, rank1seeker at gmail.com <rank1seeker at gmail.com> wrote:
> I've took machine and installed binary FreeBSD(amd64 8.2-p4 GENERIC) on it.
> Then I've installed 'benchmarks/unixbench' port.
>
> So everything is a default generic binary install ('make.conf' empty - no
> CPU optimization flags)
>
> After running: '# time unixbench', final score was:
>     394.2
> Completed in 22.8 min
>
>
> Then I've recompiled everything, from src, world and kernel, with
> 'make.conf':
> --
> CPUTYPE?=core2
> CFLAGS+=-march=native
> NO_CPU_CFLAGS=yes
> COPTFLAGS+=-march=native
> NO_CPU_COPTFLAGS=yes
> --
>
> After reboot, I've run: '# time unixbench', final score was:
>     313.5
> Completed in 26.7 min
>
> I'm getting worse result, with optimized FreeBSD's binaries?!?
> How come?

What is the test environment?

All tests running after a reboot? (if not, then do...)

Run all test in single user, when freebsd reboots, after the system
become up, than we have a running bgfsck (fixme).

And run the tests in multiple time, example with make.conf setting 5X
and without make.conf setting 5X.

>
>
> Domagoj Smolčić
> _______________________________________________
> freebsd-hackers at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-hackers
> To unsubscribe, send any mail to "freebsd-hackers-unsubscribe at freebsd.org"
>