Re: [PATCH V2 1/25] x86/cpufeatures: Enumerate Intel Hybrid Technology feature bit

[Liang, Kan]

On 3/10/2021 5:25 PM, Ricardo Neri wrote:
> On Wed, Mar 10, 2021 at 09:01:47PM +0100, Borislav Petkov wrote:
> > On Wed, Mar 10, 2021 at 11:46:44AM -0800, Ricardo Neri wrote:
> > > But this series provides the use case, right? Kan's patches handle PMU counters
> > > that may differ cross types of CPUs. In patch 2, get_hybrid_params()
> > > needs to check first if X86_FEATURE_HYBRID_CPU is enabled before
> > > querying the hybrid parameters. Otherwise, we would need to rely on the
> > > maximum level of CPUID, which may not be reliable.
> >
> > On Wed, Mar 10, 2021 at 11:33:54AM -0800, Srinivas Pandruvada wrote:
> > > We are working on changes to P-State driver for hybrid CPUs using this
> > > define. They are still work in progress.
> > > But this patch can be submitted later with our set of changes.
> >
> > Answering to both with a single mail:
> >
> > I don't have a problem with X86_FEATURE_HYBRID_CPU - I simply don't want
> > to show "hybrid_cpu" in /proc/cpuinfo unless there's a valid use case
> > for userspace to know that it is running on a hybrid CPU.
>
> Ah, I get your point now. You would like to see
>
> #define X86_FEATURE_HYBRID_CPU               (18*32+15) /* "" This part has CPUs of more than one type */
>
> Right? Now your first comment makes sense.
>
> Srinivas, Kan, I don't think we need to expose "hybrid_cpu" in
> /proc/cpuinfo, do we?

Right, Perf doesn't use the "hybrid_cpu" in /proc/cpuinfo.

Thanks,
Kan