Re: [RFCv5 PATCH 40/46] sched/cpufreq_sched: compute freq_new based on capacity_orig_of()

From	Steve Muckle <steve.muckle@linaro.org>
Newsgroups	linux.kernel
Subject	Re: [RFCv5 PATCH 40/46] sched/cpufreq_sched: compute freq_new based on capacity_orig_of()
Date	2015-10-13 21:50 +0200
Message-ID	<qjdKb-80D-61@gated-at.bofh.it> (permalink)
References	(1 earlier) <pJGg2-Pr-35@gated-at.bofh.it> <pXPMv-2Vm-37@gated-at.bofh.it> <pXXqF-5FD-1@gated-at.bofh.it> <pYehP-4Ic-13@gated-at.bofh.it> <pYryq-737-7@gated-at.bofh.it>
Organization	linux.* mail to news gateway

Show all headers | View raw

On 08/17/2015 05:19 AM, Juri Lelli wrote:
>> Nah, just maybe: (capacity << SCHED_CAPACITY_SHIFT) / capacity_orig_of()
>> > such that you don't have to export that knowledge to this thing.
>> > 
> Oh, right. I guess we can just go with something like:
> 
>  req_cap = get_cpu_usage(cpu) * capacity_margin / capacity_orig_of(cpu);
> 
> on fair.c side and switch back to
> 
>  freq_new = capacity * policy->max >> SCHED_CAPACITY_SHIFT;
> 
> on cpufreq_sched.c side. That saves us exporting capacity_orig_of().

Another good reason to make this change - I believe the current code
will not support a cpufreq policy covering CPUs of different types (such
as a b.L system in a single frequency domain).

The max requested capacity is calculated using the absolute/global
capacity values requested from each CPU. A small CPU needing a higher
OPP may get ignored because of a request from a big CPU for a lower OPP,
because the big CPU @ lower OPP shows up as the max capacity request.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Back to linux.kernel | Previous | Next | Find similar | Unroll thread

Thread

Re: [RFCv5 PATCH 40/46] sched/cpufreq_sched: compute freq_new based  on capacity_orig_of() Steve Muckle <steve.muckle@linaro.org> - 2015-10-13 21:50 +0200

csiph-web