lkml.org 
[lkml]   [2015]   [Mar]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [RFCv3 PATCH 14/48] arm: Frequency invariant scheduler load-tracking support
On Mon, Mar 23, 2015 at 01:39:44PM +0000, Peter Zijlstra wrote:
> On Wed, Feb 04, 2015 at 06:30:51PM +0000, Morten Rasmussen wrote:
> > +/* cpufreq callback function setting current cpu frequency */
> > +void arch_scale_set_curr_freq(int cpu, unsigned long freq)
> > +{
> > + atomic_long_set(&per_cpu(cpu_curr_freq, cpu), freq);
> > +}
> > +
> > +/* cpufreq callback function setting max cpu frequency */
> > +void arch_scale_set_max_freq(int cpu, unsigned long freq)
> > +{
> > + atomic_long_set(&per_cpu(cpu_max_freq, cpu), freq);
> > +}
> > +
> > +unsigned long arch_scale_freq_capacity(struct sched_domain *sd, int cpu)
> > +{
> > + unsigned long curr = atomic_long_read(&per_cpu(cpu_curr_freq, cpu));
> > + unsigned long max = atomic_long_read(&per_cpu(cpu_max_freq, cpu));
> > +
> > + if (!curr || !max)
> > + return SCHED_CAPACITY_SCALE;
> > +
> > + return (curr * SCHED_CAPACITY_SCALE) / max;
> > +}
>
> so I've no idea how many cycles an (integer) division takes on ARM; but
> doesn't it make sense to do this division (once) in
> arch_scale_set_curr_freq() instead of every time we need the result?

It does. In fact I have already prepared that fix for v4. Integer
division is expensive ARM and we call this function a lot.


\
 
 \ /
  Last update: 2015-03-24 11:01    [W:0.087 / U:0.900 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site