Re: [RFC PATCH v1 0/3] Scaled statistics using APERF/MPERF in x86

From: Arjan van de Ven
Date: Tue May 27 2008 - 12:42:34 EST


On Tue, 27 May 2008 19:34:40 +0530
Vaidyanathan Srinivasan <svaidy@xxxxxxxxxxxxxxxxxx> wrote:
>
> What we are proposing is a scaled time value that is scaled to the
> current CPU capacity. If the scaled utilisation is 50% when the CPU
> is at 100% capacity, it is expected to remain at 50% even if the CPU's
> capacity is dropped to 50%, while the traditional utilisation value
> will be 100%.

When you use the word "capacity" I cringe ;(

>
> The problem in the above two cases is that we had assumed that the
> maximum CPU capacity is 100% at normal capacity (without IDA).
>
> If the CPU is at half the maximum frequency, then scaled stats should
> show 50%.

see frequency != capacity.
It's about more than frequency. It's about how much cache you have
available too. If you run single threaded on a dual core cpu, you have
100% of the cache, but the cpu is 50% idle. But that doesn't mean that
when you double the load, you actually get 2x the performance. So
you're not at 50% of capacity!



> The point I am trying to make is whether scaling should be done
> relative to CPUs designed maximum capacity or maximum capacity under
> current constraints is to be discussed.

now you're back at "capacity".. we were at frequency before ;(


>
> Case A:
> ------
>
> Scaled stats is stats relative to maximum designed capacity including
> IDA

you don't know what that is though.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/