Received-SPF: pass (google.com: domain of linux-kernel+bounces-205591-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Message-ID: <c11acac7-085a-4041-a1f3-8b4f46e4b691@arm.com>
Date: Fri, 7 Jun 2024 10:14:38 +0200
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH 4/6] cpuidle: teo: Increase minimum time to stop tick
To: Christian Loehle <christian.loehle@arm.com>, linux-pm@vger.kernel.org,
 linux-kernel@vger.kernel.org, rafael@kernel.org
Cc: vincent.guittot@linaro.org, qyousef@layalina.io, peterz@infradead.org,
 daniel.lezcano@linaro.org, anna-maria@linutronix.de,
 kajetan.puchalski@arm.com, lukasz.luba@arm.com
References: <20240606090050.327614-1-christian.loehle@arm.com>
 <20240606090050.327614-5-christian.loehle@arm.com>
From: Dietmar Eggemann <dietmar.eggemann@arm.com>
Content-Language: en-US
In-Reply-To: <20240606090050.327614-5-christian.loehle@arm.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit

On 06/06/2024 11:00, Christian Loehle wrote:
> Since stopping the tick isn't free, add at least some minor constant
> (1ms) for the threshold to stop the tick.

Sounds pretty arbitrary to me? 'duration_ns' is either based on
target_residency_ns or tick_nohz_get_sleep_length() or even set to
TICK_NSEC/2. Does adding 1ms makes sense to all these cases? But then
why 1ms?

> Signed-off-by: Christian Loehle <christian.loehle@arm.com>
> ---
>  drivers/cpuidle/governors/teo.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/cpuidle/governors/teo.c b/drivers/cpuidle/governors/teo.c
> index 216d34747e3b..ca9422bbd8db 100644
> --- a/drivers/cpuidle/governors/teo.c
> +++ b/drivers/cpuidle/governors/teo.c
> @@ -622,10 +622,10 @@ static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
>  	/*
>  	 * Allow the tick to be stopped unless the selected state is a polling
>  	 * one or the expected idle duration is shorter than the tick period
> -	 * length.
> +	 * length plus some constant (1ms) to account for stopping it.
>  	 */
>  	if ((!(drv->states[idx].flags & CPUIDLE_FLAG_POLLING) &&
> -	    duration_ns >= TICK_NSEC) || tick_nohz_tick_stopped())
> +	    duration_ns > NSEC_PER_MSEC + TICK_NSEC) || tick_nohz_tick_stopped())
>  		return idx;
>  
>  out_tick_state: