This is the mail archive of the
mailing list for the GCC project.
Re: [middle-end] PR c/29092, vector types not compatible when they should be
- From: Roger Sayle <roger at eyesopen dot com>
- To: Andrew Pinski <pinskia at physics dot uc dot edu>
- Cc: Paolo Bonzini <paolo dot bonzini at lu dot unisi dot ch>, GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Thu, 28 Sep 2006 21:47:14 -0600 (MDT)
- Subject: Re: [middle-end] PR c/29092, vector types not compatible when they should be
On Thu, 28 Sep 2006, Andrew Pinski wrote:
> > On Wed, 27 Sep 2006, Paolo Bonzini wrote:
> > http://gcc.gnu.org/ml/gcc-patches/2006-09/msg01215.html
> > > 2006-09-25 Paolo Bonzini <email@example.com>
> > >
> > > PR c/29092
> > > * tree.c (make_vector_type): Memoize types.
> > > (used_vector_types): New.
> > Hmm. I'm not really happy with the way that you're using the front-end
> > specific types_compatible_p langhook to decide whether to create a new
> > VECTOR_TYPE node or not. The middle-end, and particulary new type
> > construction in tree.c has/needs a stricter type system...
> Shouldn't this be more like what we do for arrays? Note I have not
> looked into how we deal with arrays but I know that we can do
> make_array_type and have it return the same array type twice.
Good catch! It does look like this is a job for type_hash_canon,
but the current vector main variant building/hashing code in
make_vector_type might need a rewrite.