[PATCH] Fix wrong code generation for AMD Fam 11h CPUs in 32-bit mode

[Mikael Pettersson]

Torbjorn Granlund writes:
 > Thanks for a model bug report!
 > 
 > This is clearly a bug in GMP's config.guess.
 > 
 > Do you have any idea about how common this "K8.5" CPU core might be?  I
 > had completely missed its existence.  Were they only ever marketed as
 > "Turion", and only a fraction of the Turions used this core?

I suspect Fam 11h is somewhat common.  According to the Fam 11h revision
guide there are Athlons, Semprons, and Turions based on it, in both single
and dual-core configurarions (Turions only as dual-core).  I see no signs
of any Opterons being based on it though.

Turions have also been based on the older Fam 0Fh, and on the newer Fam 10h.

 > I am quite surprised by the nature of the 'make check' failures; if the
 > compiler is directed by GMP to use unimplemented instructions, I'd
 > really expect "Illegal instruction" traps, not miscomputations,
 > segfaults, and hangs!
 > 
 > And in 64-bit mode all works fine.  Even more strange.

Yes, I would have expected SIGILLs, but got none.  I even checked the first
10 or so core dumps, but apart from the first (explicit assert) they SEGVd
in plain integer code free of any fancy new instructions.

 > If you copy some of the failing GMP test cases to a K10 machine, do tey
 > work fine?

Yes, the binaries work fine on an Opteron 2427 (family 10h model 8).

/Mikael

P.S. Erik Hagersten just walked by and asked me to say Hi from him :-)