Re: [PATCH 1/2] x86,sched: Add support for frequency invariance

[Peter Zijlstra]

On Wed, Oct 02, 2019 at 02:26:44PM +0200, Giovanni Gherdovich wrote:
> On Tue, 2019-09-24 at 18:04 +0200, Peter Zijlstra wrote:
> > On Mon, Sep 09, 2019 at 04:42:15AM +0200, Giovanni Gherdovich wrote:
> > 
> > > +static void intel_set_cpu_max_freq(void)
> > > +{
> > > +	/*
> > > +	 * TODO: add support for:
> > > +	 *
> > > +	 * - Xeon Phi (KNM, KNL)
> > > +	 * - Xeon Gold/Platinum, Atom Goldmont/Goldmont Plus
> > > +	 * - Atom Silvermont
> > 
> > ISTR I had code for Atom.. what happened with that?
> 
> I'm being overly zealous and I wanted to get a Silvermont machine to test that
> code before sending.
> 
> The reason is that your code uses MSR_ATOM_CORE_RATIOS and
> MSR_ATOM_CORE_TURBO_RATIOS which are not documented in the SDM. I wanted to
> make sure those have the expected content on at least one machine before using
> them in my code. I have no doubt you, Srinivas and Len (who uses them in
> turbostat) have already checked but you know, more eyeballs.
> 
> I've talked to Len and Srinivas at LPC, they agreed that those two MSR may not
> have made it to the SDM but said the turbostat source code is the reference in
> this case.

Can you at least include the patch as RFC then? Perhaps other people,
who have hardware at hand, can then help test it.