[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20170906105131.gqjmaextmn3u6tj2@hirez.programming.kicks-ass.net>
Date: Wed, 6 Sep 2017 12:51:31 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Chris Wilson <chris@...is-wilson.co.uk>
Cc: Andy Lutomirski <luto@...nel.org>, Ingo Molnar <mingo@...nel.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: Abysmal scheduler performance in Linus' tree?
On Wed, Sep 06, 2017 at 12:44:20PM +0200, Peter Zijlstra wrote:
> > > + /* if this cache has capacity, come here */
> > > + if (this_stats.has_capacity && this_stats.nr_running < prev_stats.nr_running+1)
> > > + return true;
> >
> > This is still not working as intended, it should be
> >
> > if (this_stats.has_capacity && this_stats.nr_running+1 < prev_stats.nr_running)
> > return true;
> >
> > to fix the regression.
>
> Argh, you're quite right. Let me do a patch for that.
---
Subject: sched/fair: Fix wake_affine_llc() balance rules
From: Peter Zijlstra <peterz@...radead.org>
Date: Wed Sep 6 12:45:45 CEST 2017
Chris reported that the SMT balance rules got the +1 on the wrong
side, resulting in a bias towards the current LLC; which the
load-balancer would then try and undo.
Reported-by: Chris Wilson <chris@...is-wilson.co.uk>
Fixes: 90001d67be2f ("sched/fair: Fix wake_affine() for !NUMA_BALANCING")
Signed-off-by: Peter Zijlstra (Intel) <peterz@...radead.org>
---
kernel/sched/fair.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -5435,7 +5435,7 @@ wake_affine_llc(struct sched_domain *sd,
return false;
/* if this cache has capacity, come here */
- if (this_stats.has_capacity && this_stats.nr_running < prev_stats.nr_running+1)
+ if (this_stats.has_capacity && this_stats.nr_running+1 < prev_stats.nr_running)
return true;
/*
Powered by blists - more mailing lists