Re: [PATCH V2 1/6] perf/x86/intel: Add Grand Ridge and Sierra Forest

[Peter Zijlstra]

On Thu, Jun 08, 2023 at 04:20:22PM +0000, Luck, Tony wrote:
> > Then I'm hoping their take-away is that random gibberish names don't
> > help anybody. The whole Intel naming scheme is impenetrable crap.
> 
> > > +#define INTEL_FAM6_ATOM_CRESTMONT_X  0xAF /* Sierra Forest */
> > > +#define INTEL_FAM6_ATOM_CRESTMONT    0xB6 /* Grand Ridge */
> 
> This just adds another layer of confusion. Sure, these two models are based
> on the same core. But giving the illusion that they are somehow the same will
> lead to tears before bedtime:
> 
> 1) They each took a snapshot of that core design on different dates, so there
>    are logic differences.
> 2) Feature fuses will be different
> 3) Microcode will be different
> 4) BIOS will be different
> 5) "uncore" is different, so anything implemented outside of the core
>     will be different.

All those things are true for INTEL_FAM6_SKYLAKE vs INTEL_FAM6_SKYLAKE_X
and all the other pre-hybrid desktop/server parts.

And we used to do the same with the previous ATOM things, see GOLDMONT /
GOLDMONT_D, TREMONT / TREMONT_D etc..

So why should this atom be treated differently? We get a server atom and
a client atom, yes they different in all the usual way, but they still
more similar to one another than to any other random chip we have.

In short, we used to have this for core parts, we used to have this for
atom parts, but now we magically need to break from it?

Anyway, let me do the rename and squish everything into a git tree.