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:   Mon, 13 Jan 2020 13:29:11 +0100
From:   Peter Zijlstra <peterz@infradead.org>
To:     "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc:     Wanpeng Li <kernellwp@gmail.com>,
        Paolo Bonzini <pbonzini@redhat.com>,
        LKML <linux-kernel@vger.kernel.org>, kvm <kvm@vger.kernel.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Marcelo Tosatti <mtosatti@redhat.com>,
        Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>,
        KarimAllah <karahmed@amazon.de>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        Ingo Molnar <mingo@kernel.org>,
        Ankur Arora <ankur.a.arora@oracle.com>,
        christopher.s.hall@intel.com, hubert.chrzaniuk@intel.com,
        len.brown@intel.com, thomas.lendacky@amd.com
Subject: Re: [PATCH RFC] sched/fair: Penalty the cfs task which executes
 mwait/hlt
Message-ID: <20200113122911.GE2827@hirez.programming.kicks-ass.net>
References: <1578448201-28218-1-git-send-email-wanpengli@tencent.com>
 <CANRm+Cx0LMK1b2mJiU7edCDoRfPfGLzY1Zqr5paBEPcWFFALhQ@mail.gmail.com>
 <20200113104314.GU2844@hirez.programming.kicks-ass.net>
 <2579281.NS3xOKR7ft@kreacher>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <2579281.NS3xOKR7ft@kreacher>
User-Agent: Mutt/1.10.1 (2018-07-13)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Mon, Jan 13, 2020 at 12:18:46PM +0100, Rafael J. Wysocki wrote:
> On Monday, January 13, 2020 11:43:14 AM CET Peter Zijlstra wrote:

> > Anyone, what will it take to get MPERF/TSC 'working' ?
> 
> The same thing that intel_pstate does.

But intel_pstate cheats, it has a FMS listing and possible 'interesting'
chips are excluded. For instance, Core2 has APERF/MPERF, but
intel_pstate does not support Core2.

Simlarly, intel_pstate does (obviously) not support AMD chips, even tho
those have APERF/MPERF.

Although I suppose Core2 doesn't have VMX and is therefore less
interesting, but then we'd need to gate the logic with something like:

	static_cpu_has(X86_FEATURE_APERFMPERF) &&
	(static_cpu_has(X86_FEATURE_VMX) || static_cpu_has(X86_FEATURE_SVM)

> Generally speaking, it shifts the mperf values by a number of positions
> depending on the CPU model, but that is 1 except for KNL.
> 
> See get_target_pstate().

I'm going to go out on a limb and guess that's the same KNL hack as
TurboStat has.

Is that really the only known case?