Received-SPF: pass (google.com: domain of linux-kernel+bounces-133854-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
From: Hillf Danton <hdanton@sina.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: vincent.guittot@linaro.org,
	bsegall@google.com,
	kprateek.nayak@amd.com,
	efault@gmx.de,
	linux-kernel@vger.kernel.org
Subject: Re: [RFC][PATCH 10/10] sched/eevdf: Use sched_attr::sched_runtime to set request/slice suggestion
Date: Sat,  6 Apr 2024 16:16:53 +0800
Message-Id: <20240406081653.1161-1-hdanton@sina.com>
In-Reply-To: <20240405110010.934104715@infradead.org>
References: <20240405102754.435410987@infradead.org>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

On Fri, 05 Apr 2024 12:28:04 +0200 Peter Zijlstra <peterz@infradead.org>
> Allow applications to directly set a suggested request/slice length using
> sched_attr::sched_runtime.
> 
> The implementation clamps the value to: 0.1[ms] <= slice <= 100[ms]
> which is 1/10 the size of HZ=1000 and 10 times the size of HZ=100.
> 
Given HZ=100 for example, what is preventing applications of suggested
slice=0.5ms from running 5ms a tick? If slice is 90ms otoh, is tick able
to kick the curr that has been on cpu for 10ms off cpu, given
cfs_rq->nr_running > 1?

> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -984,7 +984,8 @@ static void update_deadline(struct cfs_r
>  	 * nice) while the request time r_i is determined by
>  	 * sysctl_sched_base_slice.
>  	 */
> -	se->slice = sysctl_sched_base_slice;
> +	if (!se->custom_slice)
> +		se->slice = sysctl_sched_base_slice;
>  
>  	/*
>  	 * EEVDF: vd_i = ve_i + r_i / w_i