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;
From:   "Doug Smythies" <dsmythies@telus.net>
To:     "'Rafael J. Wysocki'" <rjw@rjwysocki.net>
Cc:     "'Rafael J. Wysocki'" <rafael@kernel.org>,
        "'Viresh Kumar'" <viresh.kumar@linaro.org>,
        "'Srinivas Pandruvada'" <srinivas.pandruvada@linux.intel.com>,
        "'Zhang Rui'" <rui.zhang@intel.com>,
        "'LKML'" <linux-kernel@vger.kernel.org>,
        "'Linux PM'" <linux-pm@vger.kernel.org>
References: <13269660.K2JYd4sGFX@kreacher> <11312387.r5AVKgp8zO@kreacher>
In-Reply-To: <11312387.r5AVKgp8zO@kreacher>
Subject: RE: [PATCH v3 0/4] cpufreq: intel_pstate: Handle powersave governor correctly in the passive mode with HWP
Date:   Tue, 10 Nov 2020 13:37:00 -0800
Message-ID: <000d01d6b7a9$a1abdf80$e5039e80$@net>
MIME-Version: 1.0
Content-Type: text/plain;
        charset="us-ascii"
Content-Transfer-Encoding: 7bit
Content-Language: en-ca
Thread-Index: Ada3huWgkL5djedHQomhtWyn688yUAAH3TGw
Precedence: bulk

On 2020.11.10 09:22 Rafael J. Wysocki wrote:
> On Monday, November 9, 2020 5:49:49 PM CET Rafael J. Wysocki wrote:
>>
>> Even after the changes made very recently, the handling of the powersave
>> governor is not exactly as expected when intel_pstate operates in the
>> "passive" mode with HWP enabled.
>>
>> Namely, in that case HWP is not limited to the policy min frequency, but it
>> can scale the frequency up to the policy max limit and it cannot be constrained
>> currently, because there are no provisions for that in the framework.
>>
>> To address that, patches [1-3/4] add a new governor flag to indicate that this
>> governor wants the target frequency to be set to the exact value passed to the
>> driver, if possible, and change the powersave and performance governors to have
>> that flag set.
>>
>> The last patch makes intel_pstate take that flag into account when programming
>> the HWP Request MSR.
> 
> The v3 simply uses different names for the new governor flags.

Thank you.

I tested v2, with positive results, as reported for v1. I do not have time to
re-test v3.

My input is to also default this flag to be set for the userspace and ondemand governors.

userspace: I tested with and without this flag set, and the flag is needed if
the user expects the scaling_setspeed to be enforced.
Disclaimer: I don't normally actually use the userspace governor.

ondemand: from my tests, the ondemand response more closely mimics acpi-ondemand with the flag set.
Power consumption has been better for the limited testing done.
However, it is also a function of work/sleep frequency for periodic workflows and a function of
INTEL_CPUFREQ_TRANSITION_DELAY_HWP. I am saying that my ability to support the suggestion to default
to setting the flag is a little weak.

... Doug