Re: [PATCH 09/10] perf/x86: Remove intel_excl_states::init_state

[Peter Zijlstra <peterz@infradead.org>]

On Thu, May 21, 2015 at 06:39:29AM -0700, Stephane Eranian wrote:
> On Thu, May 21, 2015 at 4:17 AM, Peter Zijlstra <peterz@infradead.org> wrote:
> > For some obscure reason intel_{start,stop}_scheduling() copy the HT
> > state to an intermediate array. This would make sense if we ever were
> > to make changes to it which we'd have to discard.
> >
> > Except we don't. By the time we call intel_commit_scheduling() we're;
> > as the name implies; committed to them. We'll never back out.
> >
> Well, I remember there was a reason why this state was introduced. I thought
> we could do it without initially but had to add it to solve a problem.
> We do backtrack
> in the scheduling algorithm because of the tincremental transaction between the
> generic layer and the x86 layer. I'll have to remember the exact case where this
> happens.
> 
> Now, it may be that because you changed the logic of the xl vs. xlo, this saved
> state is not needed anymore. I'll have to look at that some more.

Another hint its pointless is that stop does an unconditional copy back.

So no matter what we did, all changes are always published.