This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: AltiVec [RFH] Overload resolution of vec_subs intrinsic fails


On Sat, Apr 10, 2004 at 11:32:18AM +0200, Paolo Bonzini wrote:
> > No.  I mentioned at the time that we should probably recurse
> > on compatibility.
> 
> I have this (and the removal of unneeded vector modes) implemented in
> tree-ssa.

Ah.

> Anyway, unsigned types are compatible with signed ones, aren't them?

Um, yes.

So, really, the bug is the altivec code trying to use "compatibility"
to implement something else.  I wonder if all of this can be 
re-implemented somehow; perhaps something like

#define vec_addc(x, y)						\
  ( (__typeof(x)) __builtin_altivec_addc(x, y) )

Assuming that the two arguments must be of the same type, and the type
of the result matches the type of the arguments, we can do all of the
type checking and insn selection in the compiler.



r~


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]