Re: Move ABS detection from fold-const.c to match.pd

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

On Sun, May 24, 2015 at 3:17 PM, Marc Glisse <marc.glisse@inria.fr> wrote:
>
> I forgot to mention I optimistically tried to write something like this:
>
> (match
>  (negated_value_for_comparison @0)
>  (negate @0))
> (match
>  (negated_value_for_comparison (negate @0))
>  @0)
> (match
>  (negated_value_for_comparison (minus @0 @1))
>  (if (!HONOR_SIGN_DEPENDENT_ROUNDING (type))
>   (minus @1 @0))
>
> without success. There is already a comment for logical_inverted_value about
> related limitations in genmatch.

Yeah, Prathamesh was working on inlining - not sure if that ended up
in sth usable?

+(match zerop integer_zerop)
+(match zerop real_zerop)

Would it also include fixed_zerop?  Note that with inlining implemented it would
duplicate the pattern for each match variant thus in this case adding a
tree.[ch] function zerop () might be better.

+   (simplify
+    (cnd (cmp @0 zerop) (convert?@2 @0) (negate@1 @2))
+    (if (cmp == EQ_EXPR || cmp == UNEQ_EXPR)
+     @1)
+    (if (cmp == NE_EXPR || cmp == LTGT_EXPR)
+     (non_lvalue @2))
+    (if (TYPE_SIGN (TREE_TYPE (@0)) == SIGNED /* implicit */
+        && TYPE_SIGN (type) == SIGNED
+        && element_precision (type) >= element_precision (TREE_TYPE (@0)))
+     (if (cmp == GE_EXPR || cmp == GT_EXPR
+         || (!flag_trapping_math && (cmp == UNGE_EXPR || cmp == UNGT_EXPR)))
+      (abs @2))
+     (if (cmp == LE_EXPR || cmp == LT_EXPR
+         || (!flag_trapping_math && (cmp == UNLE_EXPR || cmp == UNLT_EXPR)))
+      (negate (abs @2)))))
+   /* Now with the branches swapped.  */
+   (simplify
+    (cnd (cmp @0 zerop) (negate@1 (convert?@2 @0)) @2)

not obvious from a quick look - but would you be able to remove the
swapped branch
vairant if (cnd:c (cmp @0 zerop) X Y) would work by swapping X and Y?

The fold-const.c code doesn't seem to handle as many variants (esp.
the swapping?),
so maybe you can add a testcase that exercises some of the above on GIMPLE?

Thanks,
Richard.

>
> --
> Marc Glisse