[patch] fold-const.c: Bring fold_negate_const() from tree-ssa.

Kazu Hirata kazu@cs.umass.edu
Fri Mar 19 18:22:00 GMT 2004


Hi Jeff,

>  >If this is OK, I am planning to bring three more functions (and its
>  >children) one by one, namely
>  >
>  >  fold_abs_const(),
>  >  fold_relational_const(), and
>  >  fold_relational_hi_lo().
> I would strongly recommand against fold_relational_hi_lo.  It's got some
> subtle problems.

OK.

> I would also suggest you speak with Richard Henderson who was starting to
> work through the redesign of our folders.  

The only clean-ups I've have in mind so far are use of temporary
variables

  http://gcc.gnu.org/ml/gcc-patches/2004-03/msg01598.html

and possibly introduction of fold_build(), somewhat analogous to
simplify_gen_...(), instead of fold (build (...)).  Doing so may save
some temporaries when further folding is possible.  Of course, I won't
submit the patch for this if the effect is too small, code gets too
complicated, or the patch conflicts with long-term goals of tree-ssa,
etc.

Kazu Hirata



More information about the Gcc-patches mailing list