Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Fri, 15 Mar 2019 13:41:39 +0000
From:   Patrick Bellasi <patrick.bellasi@arm.com>
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     linux-kernel@vger.kernel.org, linux-pm@vger.kernel.org,
        linux-api@vger.kernel.org, Ingo Molnar <mingo@redhat.com>,
        Tejun Heo <tj@kernel.org>,
        "Rafael J . Wysocki" <rafael.j.wysocki@intel.com>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        Viresh Kumar <viresh.kumar@linaro.org>,
        Paul Turner <pjt@google.com>,
        Quentin Perret <quentin.perret@arm.com>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Morten Rasmussen <morten.rasmussen@arm.com>,
        Juri Lelli <juri.lelli@redhat.com>,
        Todd Kjos <tkjos@google.com>,
        Joel Fernandes <joelaf@google.com>,
        Steve Muckle <smuckle@google.com>,
        Suren Baghdasaryan <surenb@google.com>
Subject: Re: [PATCH v7 03/15] sched/core: uclamp: Add system default clamps
Message-ID: <20190315134139.hrw22zaut3shfs4d@e110439-lin>
References: <20190208100554.32196-1-patrick.bellasi@arm.com>
 <20190208100554.32196-4-patrick.bellasi@arm.com>
 <20190313143240.GH5922@hirez.programming.kicks-ass.net>
 <20190313170940.ngiafkmiijryhl2k@e110439-lin>
 <20190313201010.GU2482@worktop.programming.kicks-ass.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20190313201010.GU2482@worktop.programming.kicks-ass.net>
User-Agent: NeoMutt/20180716
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 13-Mar 21:10, Peter Zijlstra wrote:
> On Wed, Mar 13, 2019 at 05:09:40PM +0000, Patrick Bellasi wrote:
> 
> > Yes, that should be possible... will look into splitting this out in
> > v8 to have something like:
> > 
> > ---8<---
> > struct uclamp_req {
> > 	/* Clamp value "requested" by a scheduling entity */
> > 	unsigned int value		: bits_per(SCHED_CAPACITY_SCALE);
> > 	unsigned int bucket_id		: bits_per(UCLAMP_BUCKETS);
> > 	unsigned int active		: 1;
> > 	unsigned int user_defined	: 1;
> > }
> > 
> > struct uclamp_eff {
> > 	unsigned int value	        : bits_per(SCHED_CAPACITY_SCALE);
> > 	unsigned int bucket_id	        : bits_per(UCLAMP_BUCKETS);
> > }
> 
> No, have _1_ type. There is no point what so ever to splitting this.
> 
> Also, what's @user_defined about, I don't think I've seen that in the
> parent patch.

That's a flag added by one of the following patches, but with the
change you are suggesting below...

> > struct task_struct {
> >         // ...
> >         #ifdef CONFIG_UCLAMP_TASK
> >                 struct uclamp_req uclamp_req[UCLAMP_CNT];
> >                 struct uclamp_eff uclamp_eff[UCLAMP_CNT];
> 
> 		struct uclamp_se uclamp[UCLAMP_CNT];
> 		struct uclamp_se uclamp_req[UCLAMP_CNT];
> 
> Where the first is the very same introduced in patch #1, and leaving it
> in place avoids having to update the sites already using that (or start
> #1 with the _eff name to avoid having to change things around?).
> 
> >         #endif
> >         // ...
> > }
> > 
> > static inline struct uclamp_eff
> > uclamp_eff_get(struct task_struct *p, unsigned int clamp_id)
> > {
> >         struct uclamp_eff uc_eff = p->uclamp_eff[clamp_id];
> 
> just this ^, these lines seem like a superfluous duplication:
> 
> > 	uc_eff.bucket_id = p->uclamp_req[clamp_id].bucket_id;
> > 	uc_eff.value = p->uclamp_req[clamp_id].value;
> 
> 
> > 	if (unlikely(uc_eff.clamp_value > uclamp_default[clamp_id].value)) {
> > 		uc_eff.clamp_value = uclamp_default[clamp_id].value;
> > 		uc_eff.bucket_id = uclamp_default[clamp_id].bucket_id;
> 
> and:
> 
> 		uc = uclamp_default[clamp_id];

... with things like the above line it becomes quite tricky to exploit
the uclamp_se bitfield to track additional flags.

I'll try to remove the need for the "user_defined" flag, as long as we
have only the "active" you we can still manage to keep it in uclamp_se.

If instead we really need more flags, we will likely have to move them
into a separate bitfield. :/

> 
> > 	}
> > 
> >         return uc_eff;
> > }
> > 
> > static inline void
> > uclamp_eff_set(struct task_struct *p, unsigned int clamp_id)
> > {
> >         p->uclamp_eff[clamp_id] = uclamp_eff_get(p, clamp_id);
> > }
> > ---8<---
> > 
> > Is that what you mean ?
> 
> Getting there :-)

Yeah... let see :)


-- 
#include <best/regards.h>

Patrick Bellasi