Re: [PATCH] Preheat CPU in benchtests

[Andi Kleen <andi at firstfloor dot org>]

> > 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.

It can be done with newer perf or for older kernels with simple-pmu
(http://halobates.de/simple-pmu.html)
I have some example code for perf if you want.

However it has the disadvantage that it doesn't affect the frequency
of parts of the CPU outside the CPU. So the memory controller
may still be running with low frequency etc., so the ratio
between core and other parts of the system can change.

So overall it's still better to set a fixed frequency.

-Andi