On Wed, 2024-07-03 at 12:01 +0530, Dhananjay Ugwekar wrote: > Hi Rui, > > On 7/3/2024 9:37 AM, Zhang, Rui wrote: > > On Tue, 2024-07-02 at 15:50 +0530, Dhananjay Ugwekar wrote: > > > > > > > > > > > For Intel products, we have > > > > 1. Casecadelake-AP which has multi-die per package and has per- > > > > die > > > > RAPL > > > > MSRs > > > > 2. all other platforms which has single-die per package, so its > > > > RAPL > > > > MSRs can be considered as either package-scope or die-scope > > > > This applies to Thermal MSRs as well. > > > > > > > > so for these MSRs, we can treat them as > > > > 1. always die-scope for all existing platforms > > > > or > > > > 2. package-scope with the exception of Casecadelake-ap > > > > And current kernel code follows rule 1. > > > > > > > > I propose we switch to rule 2 for these code because rule 1 can > > > > be > > > > broke on future multi-die systems (This is already true for > > > > Thermal > > > > MSRs). > > > > > > I have a doubt about this, won't the future Intel multi-die > > > systems > > > have die-scope for the RAPL PMU like Casecadelake-AP? > > > > For future multi-die systems that I know, the RAPL is still package > > scope > > I think in that case we can go with rule 2, it would be future proof > for Intel systems. If you agree, I can make the change in next > version. > > Something like below?, > > -#define rapl_pmu_is_pkg_scope() \ > - (boot_cpu_data.x86_vendor == X86_VENDOR_AMD || > \ > > > - boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) > > +#define rapl_pmu_is_die_scope() \ > + (boot_cpu_data.x86_model_id == CASCADELAKE) > sounds good to me. Just a reminder that using boot_cpu_data.vfm is a better choice here.
And it would be great to get Peter' view on this. thanks, rui > Regards, > Dhananjay > > but it is just lucky that RAPL control is not exposed via the > > MSRs so rule 1 is not actually broke for RAPL PMU (while it is > > indeed > > broken for other drivers like thermal). > > > > In short, if we stick with rule 1, the RAPL PMU still works. > > Switching> to rule 2 to be consistent with the other drivers is > > also a choice IMV.> > > thanks, > > rui > > > > > > If yes, then rule 1 above seems better. > > > > > > Regards, > > > Dhananjay > > > > > > > > > > > In this sense, I think it is okay to call it pkg level rapl for > > > > both > > > > Intel and AMD. > > > > > > > > thanks, > > > > rui > > > > > > >
