Recognize the absolute-difference idiom whose selector compares the
operands directly (a > b ? a - b : b - a), optionally widened, as
+/- abs (A - B).

The same-width case is not folded under -fsanitize=signed-integer-overflow
so the rewrite does not move the overflow the sanitizer reports.

        PR tree-optimization/50856

gcc/ChangeLog:

        * match.pd (A CMP B ? A - B : B - A -> +/- abs (A - B)): New
        patterns.

gcc/testsuite/ChangeLog:

        * gcc.dg/tree-ssa/pr50856.c: New test.
        * g++.target/aarch64/pr50856.C: New test.

Signed-off-by: Eikansh Gupta <[email protected]>
---
 gcc/match.pd                               | 33 +++++++++++++++++
 gcc/testsuite/g++.target/aarch64/pr50856.C | 26 ++++++++++++++
 gcc/testsuite/gcc.dg/tree-ssa/pr50856.c    | 41 ++++++++++++++++++++++
 3 files changed, 100 insertions(+)
 create mode 100644 gcc/testsuite/g++.target/aarch64/pr50856.C
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr50856.c

diff --git a/gcc/match.pd b/gcc/match.pd
index a7cec25dbad..7f480e43f7b 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -7154,6 +7154,39 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
       (convert (negate (absu:utype @0))))
       (negate (abs @0)))))
  )
+ /* A >/>= B ? A - B : B - A -> abs (A - B)
+    with the operands optionally widened first.  */
+ (for cmp (gt ge)
+  (simplify
+   (cnd (cmp @0 @1)
+       (minus@4 (convert1?@2 @0) (convert2?@3 @1))
+       (minus @3 @2))
+   (if (ANY_INTEGRAL_TYPE_P (type)
+       && !TYPE_UNSIGNED (type)
+       && types_match (TREE_TYPE (@0), TREE_TYPE (@1))
+       && (element_precision (@0) < element_precision (type)
+           || (types_match (TREE_TYPE (@0), type)
+               && TYPE_OVERFLOW_UNDEFINED (type)
+               /* Hoisting A - B out of the selector could move the
+                  overflow the sanitizer reports.  */
+               && !sanitize_flags_p (SANITIZE_SI_OVERFLOW)))
+       && (!VECTOR_TYPE_P (type)
+           || target_supports_op_p (type, ABS_EXPR, optab_vector)))
+    (abs @4))))
+ /* A </<= B ? A - B : B - A -> -abs (A - B).  Widened operands only.  */
+ (for cmp (lt le)
+  (simplify
+   (cnd (cmp @0 @1)
+       (minus@4 (convert1?@2 @0) (convert2?@3 @1))
+       (minus @3 @2))
+   (if (ANY_INTEGRAL_TYPE_P (type)
+       && !TYPE_UNSIGNED (type)
+       && types_match (TREE_TYPE (@0), TREE_TYPE (@1))
+       && element_precision (@0) < element_precision (type)
+       && (!VECTOR_TYPE_P (type)
+           || (target_supports_op_p (type, ABS_EXPR, optab_vector)
+               && target_supports_op_p (type, NEGATE_EXPR, optab_vector))))
+    (negate (abs @4)))))
 )
 
 /* -(type)!A -> (type)A - 1.  */
diff --git a/gcc/testsuite/g++.target/aarch64/pr50856.C 
b/gcc/testsuite/g++.target/aarch64/pr50856.C
new file mode 100644
index 00000000000..f10f16f3834
--- /dev/null
+++ b/gcc/testsuite/g++.target/aarch64/pr50856.C
@@ -0,0 +1,26 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+
+/* vector a CMP b ? a - b : b - a -> abs(a - b).  Folds when the
+   target has a vector abs.  */
+
+typedef short v8hi __attribute__ ((vector_size (16)));
+typedef int v4si __attribute__ ((vector_size (16)));
+
+v8hi abd_h (v8hi a, v8hi b)
+{
+  return a > b ? a - b : b - a;
+}
+
+v4si abd_s (v4si a, v4si b)
+{
+  return a >= b ? a - b : b - a;
+}
+
+/* Same-width -abs must not fold.  */
+v8hi nabd_h (v8hi a, v8hi b)
+{
+  return a < b ? a - b : b - a;
+}
+
+/* { dg-final { scan-tree-dump-times "ABS_EXPR" 2 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr50856.c 
b/gcc/testsuite/gcc.dg/tree-ssa/pr50856.c
new file mode 100644
index 00000000000..3c0513efbbd
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr50856.c
@@ -0,0 +1,41 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+
+/* a CMP b ? a - b : b - a -> +/- abs (a - b).  */
+
+int gt_w (unsigned char a, unsigned char b)
+{
+  int x = a, y = b;
+  return a > b ? x - y : y - x;
+}
+
+int ge_w (unsigned char a, unsigned char b)
+{
+  int x = a, y = b;
+  return a >= b ? x - y : y - x;
+}
+
+int lt_w (unsigned char a, unsigned char b)
+{
+  int x = a, y = b;
+  return a < b ? x - y : y - x;
+}
+
+int le_w (unsigned char a, unsigned char b)
+{
+  int x = a, y = b;
+  return a <= b ? x - y : y - x;
+}
+
+int gt_i (int a, int b)
+{
+  return a > b ? a - b : b - a;
+}
+
+/* Same-width -abs must not fold.  */
+int lt_i (int a, int b)
+{
+  return a < b ? a - b : b - a;
+}
+
+/* { dg-final { scan-tree-dump-times "ABS_EXPR" 5 "optimized" } } */
-- 
2.34.1

Reply via email to