Bug 102334

I don't have those commit ids available, but that date does coincide with the rc1 merge. Is that true? Just checking the merge base would be useful to narrow it down between i915 vs the world.

Ok, between v4.12 and drm-tip I see a drop from 6 to 4.5 on my bxt. Should poke Martin about ezbench bisection...

First bisection gives:

commit 3fed382b46baac83703130fe4cd3d9147f427fb9
Author: Rik van Riel <riel@redhat.com>
Date:   Fri Jun 23 12:55:29 2017 -0400

    sched/numa: Implement NUMA node level wake_affine()
    
    Since select_idle_sibling() can place a task anywhere on a socket,
    comparing loads between individual CPU cores makes no real sense
    for deciding whether to do an affine wakeup across sockets, either.
    
    Instead, compare the load between the sockets in a similar way the
    load balancer and the numa balancing code do.
    
    Signed-off-by: Rik van Riel <riel@redhat.com>
    Cc: Linus Torvalds <torvalds@linux-foundation.org>
    Cc: Mel Gorman <mgorman@suse.de>
    Cc: Mike Galbraith <efault@gmx.de>
    Cc: Peter Zijlstra <peterz@infradead.org>
    Cc: Thomas Gleixner <tglx@linutronix.de>
    Cc: jhladky@redhat.com
    Cc: linux-kernel@vger.kernel.org
    Link: http://lkml.kernel.org/r/20170623165530.22514-4-riel@redhat.com
    Signed-off-by: Ingo Molnar <mingo@kernel.org>

Created attachment 133763 [details] [review]
Fix wake_affine regression

Playing around to restore the old code; gives us the spread across cores required to make OglMultithread happier.

Ok, found a version by PeterZ: 20170801121912.fnykqlq3r5jcbtn2@hirez.programming.kicks-ass.net

Gave it a cursory test and it appears to have fixed this particular regression, so pushed to topic/core-for-CI for wider testing:

commit 367a3aa35e7bed10a74bed139f6728916f5d1508
Author: Peter Zijlstra <peterz@infradead.org>
Date:   Tue Aug 1 23:43:07 2017 +0200

    sched/fair: Fix wake_affine() for !NUMA_BALANCING

It should only affect !llc platforms, but the logic is slightly different than the original prior to

commit 3fed382b46baac83703130fe4cd3d9147f427fb9
Author: Rik van Riel <riel@redhat.com>
Date:   Fri Jun 23 12:55:29 2017 -0400

    sched/numa: Implement NUMA node level wake_affine()

so testing definitely required.

Part 1 upstream:

commit 90001d67be2fa2acbe3510d1f64fa6533efa30ef
Author: Peter Zijlstra <peterz@infradead.org>
Date:   Mon Jul 31 17:50:05 2017 +0200

    sched/fair: Fix wake_affine() for !NUMA_BALANCING

Needs another fix which is pending.

            Commit-ID:  a731ebe6f17bff9e7ca12ef227f9da4d5bdf8425
            Gitweb:     http://git.kernel.org/tip/a731ebe6f17bff9e7ca12ef227f9da4d5bdf8425
            Author:     Peter Zijlstra <peterz@infradead.org>
            AuthorDate: Wed, 6 Sep 2017 12:51:31 +0200
            Committer:  Ingo Molnar <mingo@kernel.org>
            CommitDate: Thu, 7 Sep 2017 09:29:31 +0200
            
            sched/fair: Fix wake_affine_llc() balancing rules
            
            Chris Wilson 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@chris-wilson.co.uk>
            Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
            Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
            Cc: Andy Lutomirski <luto@kernel.org>
            Cc: Linus Torvalds <torvalds@linux-foundation.org>
            Cc: Mike Galbraith <efault@gmx.de>
            Cc: Peter Zijlstra <peterz@infradead.org>
            Cc: Thomas Gleixner <tglx@linutronix.de>
            Cc: linux-kernel@vger.kernel.org
            Fixes: 90001d67be2f ("sched/fair: Fix wake_affine() for !NUMA_BALANCING")
            Link: http://lkml.kernel.org/r/20170906105131.gqjmaextmn3u6tj2@hirez.programming.kicks-ass.net
            Signed-off-by: Ingo Molnar <mingo@kernel.org>

There have been *two*, equally large BXT improvements to SynMark Multithread performance.

First one, which already more than fixed the earlier regression was between:
650ffdaee4 at 2017-08-24 14:30:13 UTC
4823f19b97 at 2017-08-25 15:55:28 UTC

and the second was between:
587d276952 at 2017-09-06 17:54:04 UTC
2300dd9c44 at 2017-09-07 17:00:01 UTC

I assume latter to be the above referred commit.  That actually improved Multithread performance slightly also on several other platforms.

While I can now verify this, I'm wondering a bit about the first commit...

Thanks!

Closing old verified.