This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] Preheat CPU in benchtests
- From: OndÅej BÃlka <neleai at seznam dot cz>
- To: Andi Kleen <andi at firstfloor dot org>
- Cc: libc-alpha at sourceware dot org
- Date: Tue, 23 Apr 2013 17:17:26 +0200
- Subject: Re: [PATCH] Preheat CPU in benchtests
- References: <20130423061028 dot GA6257 at domone dot kolej dot mff dot cuni dot cz> <m27gjtwcmf dot fsf at firstfloor dot org>
On Tue, Apr 23, 2013 at 07:22:16AM -0700, Andi Kleen wrote:
> OndÅej BÃlka <neleai@seznam.cz> writes:
>
> > Benchmarks now are affected by cpu scaling when initialy run at low
> > frequency.
> >
> > Following benchmark runs nonsensial loop first to ensure that benchmark
> > are measured at maximal frequency. This greatly cuts time needed to
> > get accurate results.
>
> FWIW it's generally safer to disable frequency scaling explicitely
> through sysfs (but that needs root), as the reaction time of the
> p-state governour can be unpredictable.
Which needs root, so it would request typing password each time you run
automated benchmarks.
I consider for some time by CPU_CLK_UNHALTED performance counter.
However a documentation is lacking and I need it with low overhead.
>
> On Intel if you run like this too long Turbo may also stop turboing,
> as the temperature increases. I typically disable Turbo for micro
> benchmarks (set max pstate to be one less)
>
> -Andi
>
> --
> ak@linux.intel.com -- Speaking for myself only