Messages in this thread | | | Subject | Re: [RFC PATCH 1/3] x86/cpu: Unify CPU family, model, stepping calculation | From | Paolo Bonzini <> | Date | Wed, 18 Nov 2015 12:35:24 +0100 |
| |
On 18/11/2015 12:28, Borislav Petkov wrote: >> On 14/11/2015 11:37, Borislav Petkov wrote: >>> > > vendor = x86_vendor(); >>> > > - family = x86_family(); >>> > > + family = x86_family_cpuid(); >> > >> > What about renaming x86_vendor() so that this looks like >> > >> > - vendor = x86_vendor(); >> > - family = x86_family(); >> > + vendor = x86_cpuid_vendor(); >> > + family = x86_cpuid_family(); > > The idea is that x86_family_cpuid() gives the family *after* having > executed CPUID while x86_family() only computes the family from a > supplied CPUID_1_EAX. I.e., the last saves us the CPUID call.
Yes, exactly. I'm suggesting that the same applies to x86_vendor(). I also prefer x86_cpuid_* to x86_*_cpuid because, once you add two functions in the same family it's nice that they share a prefix.
Paolo
> Hmm, maybe I should make that more clear ...
| |