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;
From:   "Rafael J. Wysocki" <rjw@rjwysocki.net>
To:     Linux PM <linux-pm@vger.kernel.org>,
        Peter Zijlstra <peterz@infradead.org>,
        Anna-Maria Behnsen <anna-maria@linutronix.de>
Cc:     LKML <linux-kernel@vger.kernel.org>,
        Frederic Weisbecker <frederic@kernel.org>,
        Kajetan Puchalski <kajetan.puchalski@arm.com>
Subject: [RFC/RFT][PATCH v1 0/2] cpuidle: teo: Do not check timers unconditionally every time
Date:   Tue, 01 Aug 2023 21:35:15 +0200
Message-ID: <4511619.LvFx2qVVIh@kreacher>
MIME-Version: 1.0
Content-Transfer-Encoding: 7Bit
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi Folks,

This is on top of the fixes series posted previously:

https://lore.kernel.org/linux-pm/4515817.LvFx2qVVIh@kreacher/

(I'll put it all into one git branch tomorrow).

I started to play with the idea described here

https://lore.kernel.org/linux-pm/CAJZ5v0hQh2Pg_uXxj8KBRw3oLS1WdsU+rUafBAAq7dRdbRwYSA@mail.gmail.com/

and this is the result.

Note that this is completely experimental, even though it doesn't kill any of
the test boxes I've run it on.

Patch [1/2] moves the tick_nohz_get_sleep_length() call in teo_select() after
a preliminary idle state selection based on statistics and patch [2/2] adds
checks to avoid it completely if the idle state selected so far is shallow
enough.

I would appreciate checking if this actually makes any difference.

Thanks!