Re: [PATCH 5/8 v2] middle-end: Add cltz_complement idiom recognition

[Richard Biener <richard.guenther@gmail.com>]

On Thu, Jan 19, 2023 at 10:19 AM Jan-Benedict Glaw <jbglaw@lug-owl.de> wrote:
>
> On Thu, 2022-12-22 17:42:16 +0000, Andrew Carlotti via Gcc-patches <gcc-patches@gcc.gnu.org> wrote:
> > New patch below, bootstrapped and regression tested on
> > aarch64-unknown-linux-gnu and x86_64-pc-linux-gnu - ok to merge?
>
> > diff --git a/gcc/tree-ssa-loop-niter.cc b/gcc/tree-ssa-loop-niter.cc
> > index fece876099c1687569d6351e7d2416ea6acae5b5..ce2441f2a6dbdf2d8fe42755d5d1abd8a631bb5c 100644
> > --- a/gcc/tree-ssa-loop-niter.cc
> > +++ b/gcc/tree-ssa-loop-niter.cc
> > @@ -42,6 +42,7 @@ along with GCC; see the file COPYING3.  If not see
> >  #include "tree-chrec.h"
> >  #include "tree-scalar-evolution.h"
> >  #include "tree-dfa.h"
> > +#include "internal-fn.h"
> >  #include "gimple-range.h"
> >
> >
> > @@ -2198,6 +2199,224 @@ number_of_iterations_popcount (loop_p loop, edge exit,
> >    return true;
> >  }
> >
> > +/* Return an expression that counts the leading/trailing zeroes of src.
> > +
> > +   If define_at_zero is true, then the built expression will be defined to
> > +   return the precision of src when src == 0 (using either a conditional
> > +   expression or a suitable internal function).
> > +   Otherwise, we can elide the conditional expression and let src = 0 invoke
> > +   undefined behaviour.  */
> > +
> > +static tree
> > +build_cltz_expr (tree src, bool leading, bool define_at_zero)
> > +{
> [...]
> > +
> > +  tree call;
> > +  if (use_ifn)
> > +    {
> > +      call = build_call_expr_internal_loc (UNKNOWN_LOCATION, ifn,
> > +                                        integer_type_node, 1, src);
> > +      int val;
> > +      scalar_int_mode mode = SCALAR_INT_TYPE_MODE (utype);
>          ^^^^^^^^^^^^^^^^^^^^
>
> This will give us a new unused variable warning.

I wonder if hardening the defaults.h macros like

#define CLZ_DEFINED_VALUE_AT_ZERO(MODE, VALUE)  (((MODE), (VALUE)), 0)

fixes that and makes sense (also to avoid losing side-effects for the arguments)

Richard.


> > +      int optab_defined_at_zero
> > +     = leading ? CLZ_DEFINED_VALUE_AT_ZERO (mode, val)
> > +               : CTZ_DEFINED_VALUE_AT_ZERO (mode, val);
> > +      if (define_at_zero && !(optab_defined_at_zero == 2 && val == prec))
> > +     {
> > +       tree is_zero = fold_build2 (NE_EXPR, boolean_type_node, src,
> > +                                   build_zero_cst (TREE_TYPE (src)));
> > +       call = fold_build3(COND_EXPR, integer_type_node, is_zero, call,
> > +                          build_int_cst (integer_type_node, prec));
> > +     }
> > +    }
>
> MfG, JBG
>
> --