[RFC] Tightening up the type system

[Diego Novillo]

On Thu, 2004-09-30 at 15:58, Geoffrey Keating wrote:
> Diego Novillo <dnovillo@redhat.com> writes:
> 
> > In fact, I would like to get to the point where we can simply call
> > 'gcc_assert (lang_hook.compatible_types_p (dest, orig))' when doing
> > propagation.  Ideally, all the necessary type conversions should be
> > exposed in the IL.
> 
> Should the validity of GIMPLE really depend on a language hook like
> this?  Surely it would make more sense to have a
> gimple_compatible_types_p that reflects what code that parses GIMPLE
> can actually handle (presumably a superset of what each language can
> generate).
>
Perhaps.  Today I was thinking something along the lines of using
alias_sets_conflict_p() as our type compatibility test, instead.  That
would loosen the restrictions and allow mixing things like const type *
and type *.

Right now, we have a hybrid model where we try to remove "useless" type
conversions, but then validate propagation opportunities using
compatible_types_p. 


Diego.