[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20171030165828.u77shgkuhzawxu4g@hirez.programming.kicks-ass.net>
Date: Mon, 30 Oct 2017 17:58:28 +0100
From: Peter Zijlstra <peterz@...radead.org>
To: Christopher Lameter <cl@...ux.com>
Cc: cmetcalf@...lanox.com, linux-kernel@...r.kernel.org,
tglx@...utronix.de, torvalds@...ux-foundation.org, hpa@...or.com,
riel@...hat.com, mingo@...nel.org, efault@....de,
frederic@...nel.org, kernellwp@...il.com,
paulmck@...ux.vnet.ibm.com, lcapitulino@...hat.com,
linux-tip-commits@...r.kernel.org
Subject: Re: [tip:sched/core] sched/isolation: Document the isolcpus= flags
On Mon, Oct 30, 2017 at 11:30:59AM -0500, Christopher Lameter wrote:
> Historically cpusets were not used for cpu isolation. They were used to
> restrict applications threads to sets of cpus for performance reasons. And
> we are here dealing with individual processors.
The HPC workloads very much disabled load-balancing across most CPUs. If
you disable "sched_load_balance" the thing creates NULL sched_domains,
the exact thing isolcpus ends up doing.
This is something cpusets have done for a _long_ time, if not from the
very start.
Yes, you can also create smaller sched_domains which is useful for other
cases and you can even mix the lot, by creating a small set of
load-balanced CPUs for the system tasks while giving a bunch of
unbalanced CPUs to your application.
Powered by blists - more mailing lists