Minor improvements to alias.c:nonoverlapping_memrefs_p

[Richard Henderson]

On Tue, Nov 13, 2001 at 08:53:00PM -0500, Richard Kenner wrote:
> But why would that be considered as being in the frame?  You could pass
> a pointer to a global into a function so a dereference of the hard reg
> for the input could conflict with a global reference, unlike frame references.

My point exactly.  Your test is merely < LAST_VIRTUAL_REGNO, 
which includes *all* hard registers.


r~