On Wed, Feb 17, 2016 at 10:09 PM, Michael Ellerman <m...@ellerman.id.au> wrote:
> On Wed, 2016-02-17 at 16:07 +1100, Michael Neuling wrote: > > > Add a cputable entry for POWER9. More code is required to actually > > boot and run on a POWER9 but this gets the base piece in which we can > > start building on. > > > > Copies over from POWER8 except for: > > - Adds a new CPU_FTR_ARCH_30 bit to start hanging new architecture > > ARCH thirty? > > Would CPU_FTR_ARCH_3 read better? > > Or CPU_FTR_ARCH_3_00 ? The user visible version flags all have the pattern ARCH_X_XX while the in-kernel flags use ARCH_XXX. It should probably be CPU_FTR_ARCH_300 for consistency with the other kernel flags. > +#define COMMON_USER_POWER9 (COMMON_USER_PPC64 | PPC_FEATURE_ARCH_2_06 > |\ > > + PPC_FEATURE_SMT | > PPC_FEATURE_ICACHE_SNOOP | \ > > + PPC_FEATURE_TRUE_LE | \ > > + PPC_FEATURE_PSERIES_PERFMON_COMPAT) > > That looks like it's == COMMON_USER_POWER8. > > > +#define COMMON_USER2_POWER9 (PPC_FEATURE2_ARCH_2_07 | \ > > + PPC_FEATURE2_HTM_COMP | \ > > + PPC_FEATURE2_HTM_NOSC_COMP | \ > > + PPC_FEATURE2_DSCR | \ > > + PPC_FEATURE2_ISEL | PPC_FEATURE2_TAR | \ > > + PPC_FEATURE2_VEC_CRYPTO | \ > > + PPC_FEATURE2_ARCH_3_00 | \ > > + PPC_FEATURE2_HAS_IEEE128) > > And this could be COMMON_USER_POWER8 + ARCH_3 + HAS_IEEE128 I think? It could be, but similarly the POWER8 flags could also be POWER7 + some. I think they're separate so flags can be easily removed if need be, but I'm not sure how useful that is.
_______________________________________________ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
