Re: [tip:sched/balancing] sched: Add comments to find_busiest_group() function

[Ingo Molnar <mingo@elte.hu>]

* Ray Lee <ray-lk@madrabbit.org> wrote:

> This commit says it's just adding comments, but...
> 
> On Wed, Mar 25, 2009 at 5:30 AM, Gautham R Shenoy <ego@in.ibm.com> wrote:
> > Commit-ID:  b7bb4c9bb01941fe8feb653f3410e7ed0c9bb786
> > Gitweb:     http://git.kernel.org/tip/b7bb4c9bb01941fe8feb653f3410e7ed0c9bb786
> > Author:     Gautham R Shenoy <ego@in.ibm.com>
> > AuthorDate: Wed, 25 Mar 2009 14:44:27 +0530
> > Committer:  Ingo Molnar <mingo@elte.hu>
> > CommitDate: Wed, 25 Mar 2009 13:28:30 +0100
> >
> > sched: Add comments to find_busiest_group() function
> >
> > Impact: cleanup
> >
> > Add /** style comments around find_busiest_group(). Also add a few
> > explanatory comments.
> 
> ...but there are actual code changes. Hard to know if you intended to
> do that and forgot to changelog it, or if it's an unexpected hunk that
> accidentally got included:
> 
> >        if (balance && !(*balance))
> >                goto ret;
> >
> > -       if (!sds.busiest || sds.this_load >= sds.max_load
> > -               || sds.busiest_nr_running == 0)
> > +       if (!sds.busiest || sds.busiest_nr_running == 0)
> > +               goto out_balanced;
> > +
> > +       if (sds.this_load >= sds.max_load)
> >                goto out_balanced;
> >
> >        sds.avg_load = (SCHED_LOAD_SCALE * sds.total_load) / sds.total_pwr;
> >
> > -       if (sds.this_load >= sds.avg_load ||
> > -                       100*sds.max_load <= sd->imbalance_pct * sds.this_load)
> > +       if (sds.this_load >= sds.avg_load)
> > +               goto out_balanced;
> > +
> > +       if (100 * sds.max_load <= sd->imbalance_pct * sds.this_load)
> >                goto out_balanced;
> >
> >        sds.busiest_load_per_task /= sds.busiest_nr_running;

yeah.

Note that it does not actually change the resulting logic, it splits 
out an over-long (and hard to read) series of conditions into an 
equivalent set of two if() statements. [the first one changes the 
order of two conditions - but that is harmless]

It indeed would have been nice to declare this in the changelog 
though.

	Ingo