Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20201016163634.857573-1-wvw@google.com>
In-Reply-To: <20201016163634.857573-1-wvw@google.com>
From:   "Rafael J. Wysocki" <rafael@kernel.org>
Date:   Fri, 16 Oct 2020 19:01:42 +0200
Message-ID: <CAJZ5v0in0N6USFqvuLcccpirahj=oOki67+Lq9_5i_q7Ep7T6g@mail.gmail.com>
Subject: Re: [PATCH] sched: cpufreq_schedutil: maintain raw cache when next_f
 is not changed
To:     Wei Wang <wvw@google.com>
Cc:     wei.vince.wang@gmail.com, Viresh Kumar <viresh.kumar@linaro.org>,
        Quentin Perret <qperret@google.com>,
        "Rafael J. Wysocki" <rjw@rjwysocki.net>,
        Ingo Molnar <mingo@redhat.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Juri Lelli <juri.lelli@redhat.com>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Ben Segall <bsegall@google.com>, Mel Gorman <mgorman@suse.de>,
        Daniel Bristot de Oliveira <bristot@redhat.com>,
        Linux PM <linux-pm@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Oct 16, 2020 at 6:36 PM Wei Wang <wvw@google.com> wrote:
>
> Currently, raw cache will be reset when next_f is changed after
> get_next_freq for correctness. However, it may introduce more
> cycles. This patch changes it to maintain the cached value instead of
> dropping it.

IMV you need to be more specific about why this helps.

> This is adapted from https://android-review.googlesource.com/1352810/
>
> Signed-off-by: Wei Wang <wvw@google.com>
> ---
>  kernel/sched/cpufreq_schedutil.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
> index 5ae7b4e6e8d6..ae3ae7fcd027 100644
> --- a/kernel/sched/cpufreq_schedutil.c
> +++ b/kernel/sched/cpufreq_schedutil.c
> @@ -31,6 +31,7 @@ struct sugov_policy {
>         s64                     freq_update_delay_ns;
>         unsigned int            next_freq;
>         unsigned int            cached_raw_freq;
> +       unsigned int            prev_cached_raw_freq;
>
>         /* The next fields are only needed if fast switch cannot be used: */
>         struct                  irq_work irq_work;
> @@ -165,6 +166,7 @@ static unsigned int get_next_freq(struct sugov_policy *sg_policy,
>                 return sg_policy->next_freq;
>
>         sg_policy->need_freq_update = false;
> +       sg_policy->prev_cached_raw_freq = sg_policy->cached_raw_freq;
>         sg_policy->cached_raw_freq = freq;
>         return cpufreq_driver_resolve_freq(policy, freq);
>  }
> @@ -464,8 +466,8 @@ static void sugov_update_single(struct update_util_data *hook, u64 time,
>         if (busy && next_f < sg_policy->next_freq) {
>                 next_f = sg_policy->next_freq;
>
> -               /* Reset cached freq as next_freq has changed */
> -               sg_policy->cached_raw_freq = 0;
> +               /* Restore cached freq as next_freq has changed */
> +               sg_policy->cached_raw_freq = sg_policy->prev_cached_raw_freq;
>         }
>
>         /*
> @@ -828,6 +830,7 @@ static int sugov_start(struct cpufreq_policy *policy)
>         sg_policy->limits_changed               = false;
>         sg_policy->need_freq_update             = false;
>         sg_policy->cached_raw_freq              = 0;
> +       sg_policy->prev_cached_raw_freq         = 0;
>
>         for_each_cpu(cpu, policy->cpus) {
>                 struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);
> --
> 2.29.0.rc1.297.gfa9743e501-goog
>