[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1462952635.3961.15.camel@suse.de>
Date: Wed, 11 May 2016 09:43:55 +0200
From: Mike Galbraith <mgalbraith@...e.de>
To: Yuyang Du <yuyang.du@...el.com>,
Peter Zijlstra <peterz@...radead.org>
Cc: mingo@...nel.org, linux-kernel@...r.kernel.org, clm@...com,
matt@...eblueprint.co.uk, tglx@...utronix.de, fweisbec@...il.com
Subject: Re: [RFC][PATCH 5/7] sched: Rewrite select_idle_siblings()
On Wed, 2016-05-11 at 07:42 +0800, Yuyang Du wrote:
> Do you have any suggestion about doing other part of
> select_task_rq_fair?
Conjure up a decent metric for placing tasks in tasty hot L2? That
would be really lovely to have (/me pondering page faults..).
We used to have an avg_overlap heuristic, which would let us stack high
speed ping-pong like things such that they could really smoke. It was
too fragile though, so had to go away.
-Mike
Powered by blists - more mailing lists