Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <934429dc-03d7-f7d4-8828-723e756f1411@efficios.com>
Date:   Wed, 26 Jul 2023 15:16:34 -0400
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.13.0
Subject: Re: [RFC PATCH 1/1] sched: Extend cpu idle state for 1ms
Content-Language: en-US
From:   Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
To:     Shrikanth Hegde <sshegde@linux.vnet.ibm.com>
Cc:     linux-kernel@vger.kernel.org, Ingo Molnar <mingo@redhat.com>,
        Valentin Schneider <vschneid@redhat.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Ben Segall <bsegall@google.com>, Mel Gorman <mgorman@suse.de>,
        Daniel Bristot de Oliveira <bristot@redhat.com>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        Juri Lelli <juri.lelli@redhat.com>,
        Swapnil Sapkal <Swapnil.Sapkal@amd.com>,
        Aaron Lu <aaron.lu@intel.com>, x86@kernel.org,
        Peter Zijlstra <peterz@infradead.org>,
        Srikar Dronamraju <srikar@linux.vnet.ibm.com>
References: <20230725193048.124796-1-mathieu.desnoyers@efficios.com>
 <69076f8e-191b-2e3e-d810-ea72d8ff18bb@linux.vnet.ibm.com>
 <79fa8a62-a74e-2623-9f03-1f1af85b6c07@efficios.com>
 <cab82676-27fd-b4e1-2cd8-3d8d26b44aa0@linux.vnet.ibm.com>
 <447f756c-9c79-f801-8257-a97cc8256efe@efficios.com>
In-Reply-To: <447f756c-9c79-f801-8257-a97cc8256efe@efficios.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 7/26/23 14:56, Mathieu Desnoyers wrote:
> On 7/26/23 13:40, Shrikanth Hegde wrote:
[...]
>> Do you have SMT here? What is the system utilization when you are running
>> this workload?
> 
> Yes, SMT is enabled, which brings the number of logical cpus to 384.

Here is an additional interesting data point with nosmt=force on
6.4.4:

hackbench -g 32 -f 20 --threads --pipe -l 480000 -s 100

baseline:                       90s
with idle-delay+nr_running<=4:  87s (3% speedup)

hackbench -g 16 -f 20 --threads --pipe -l 480000 -s 100

baseline:                       52s
with idle-delay+nr_running<=4:  32s (38% speedup)

So the impact of the patch appears to depend on how much the
system actually reaches idle, which does make sense.

Thanks,

Mathieu

-- 
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com