amd64/113111: Potentially wrong instructions will be produced for EM64T

[Astrodog]

On 5/29/07, NIIMI Satoshi <sa2c at sa2c.net> wrote:
>
> Mark Linimon wrote:
> > I'm confused, I thought "k8" was "AMD-only"?
>
> FreeBSD/amd64 runs on both AMD's AMD 64 CPUs (Opteron, Athlon 64, etc)
> and Intel's EM64T CPUs (Xeon, Core 2, etc).
>
>
I assume the issue here is the -march setting, when CPUTYPE isn't set for
lib32, in Makefile.inc1. The way I see it, this can be handled either in the
errata, or release notes, (set CPUTYPE/-march on EM64T), or the proposed
patch. I guess the real question is what's being offically supported, AMD64,
or both?

This isn't the only instruction difference between AMD64 and EM64T, so there
may be other things that need to be examined... but these things causing
issues would appear to be fairly rare, given the lack of AMD64 PRs related
to EM64T crashes or odd behavior on unknown instructions. There is also the
difference between early EM64T, and newer EM64T chips, so that might be
something to look at too.

--- Harrison Grundy