From mboxrd@z Thu Jan 1 00:00:00 1970 From: Patrick Bellasi Subject: Re: [PATCH v8 04/16] sched/core: uclamp: Add system default clamps Date: Thu, 9 May 2019 10:10:57 +0100 Message-ID: <20190509091057.ckef2ley4eswyzds@e110439-lin> References: <20190402104153.25404-1-patrick.bellasi@arm.com> <20190402104153.25404-5-patrick.bellasi@arm.com> <20190508190733.GC32547@worktop.programming.kicks-ass.net> <20190508191529.GA26813@worktop.programming.kicks-ass.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20190508191529.GA26813@worktop.programming.kicks-ass.net> Sender: linux-kernel-owner@vger.kernel.org To: Peter Zijlstra Cc: linux-kernel@vger.kernel.org, linux-pm@vger.kernel.org, linux-api@vger.kernel.org, Ingo Molnar , Tejun Heo , "Rafael J . Wysocki" , Vincent Guittot , Viresh Kumar , Paul Turner , Quentin Perret , Dietmar Eggemann , Morten Rasmussen , Juri Lelli , Todd Kjos , Joel Fernandes , Steve Muckle , Suren Baghdasaryan List-Id: linux-api@vger.kernel.org On 08-May 21:15, Peter Zijlstra wrote: > On Wed, May 08, 2019 at 09:07:33PM +0200, Peter Zijlstra wrote: > > On Tue, Apr 02, 2019 at 11:41:40AM +0100, Patrick Bellasi wrote: > > > +static inline struct uclamp_se > > > +uclamp_eff_get(struct task_struct *p, unsigned int clamp_id) > > > +{ > > > + struct uclamp_se uc_req = p->uclamp_req[clamp_id]; > > > + struct uclamp_se uc_max = uclamp_default[clamp_id]; > > > + > > > + /* System default restrictions always apply */ > > > + if (unlikely(uc_req.value > uc_max.value)) > > > + return uc_max; > > > + > > > + return uc_req; > > > +} > > > + > > > +static inline unsigned int > > > +uclamp_eff_bucket_id(struct task_struct *p, unsigned int clamp_id) > > > +{ > > > + struct uclamp_se uc_eff; > > > + > > > + /* Task currently refcounted: use back-annotated (effective) bucket */ > > > + if (p->uclamp[clamp_id].active) > > > + return p->uclamp[clamp_id].bucket_id; > > > + > > > + uc_eff = uclamp_eff_get(p, clamp_id); > > > + > > > + return uc_eff.bucket_id; > > > +} > > > + > > > +unsigned int uclamp_eff_value(struct task_struct *p, unsigned int clamp_id) > > > +{ > > > + struct uclamp_se uc_eff; > > > + > > > + /* Task currently refcounted: use back-annotated (effective) value */ > > > + if (p->uclamp[clamp_id].active) > > > + return p->uclamp[clamp_id].value; > > > + > > > + uc_eff = uclamp_eff_get(p, clamp_id); > > > + > > > + return uc_eff.value; > > > +} > > > > This is 'wrong' because: > > > > uclamp_eff_value(p,id) := uclamp_eff(p,id).value > > Clearly I means to say the above does not hold with the given > implementation, while the naming would suggest it does. Not sure to completely get your point... AFAIU, what you call uclamp_eff(p, id).value is the "value" member of the struct returned by uclamp_eff_get(p,id), which is back annotate by uclamp_rq_inc_id(p, rq, id) in: p->uclamp[clamp_id].value when a task becomes RUNNABLE. > > Which seems to suggest the uclamp_eff_*() functions want another name. That function returns the effective value of a task, which is either: 1. the back annotated value for a RUNNABLE task or 2. the aggregation of task-specific, system-default and cgroup values for a non RUNNABLE task. > > Also, suppose the above would be true; does GCC really generate better > > code for the LHS compared to the RHS? It generate "sane" code which implements the above logic and allows to know that whenever we call uclamp_eff_value(p,id) we get the most updated effective value for a task, independently from its {!}RUNNABLE state. I would keep the function but, since Suren also complained also about the name... perhaps I should come up with a better name? Proposals? -- #include Patrick Bellasi