As discussed in the PR, the reduction code isn't able to cope with type
promotions/demotions in the reduction computation, so if we recognize an
over-widening pattern that has vect_reduction_def type, we most likely make
it non-vectorizable.
2021-02-02 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/98848
* tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
* gcc.dg/vect/pr98848.c: New test.
* gcc.dg/vect/pr92205.c: Remove xfail.
return d;
}
-/* { dg-final { scan-tree-dump "vectorized 1 loops" "vect" { xfail *-*-* } } } */
+/* { dg-final { scan-tree-dump "vectorized 1 loops" "vect" { target { vect_unpack && { ! vect_no_bitwise } } } } } */
--- /dev/null
+/* PR tree-optimization/98848 */
+/* { dg-do compile } */
+/* { dg-require-effective-target vect_int } */
+
+short a[9000];
+
+int
+foo (void)
+{
+ int b = a[0];
+ int i;
+ for (i = 1; i < 9000; i ++)
+ if (a[i] < b)
+ b = a[i];
+ return b;
+}
+
+/* { dg-final { scan-tree-dump-times "vectorized 1 loop" 1 "vect" { xfail { vect_no_int_add || vect_no_int_min_max } } } } */
tree type = TREE_TYPE (lhs);
tree_code code = gimple_assign_rhs_code (last_stmt);
+ /* Punt for reductions where we don't handle the type conversions. */
+ if (STMT_VINFO_DEF_TYPE (last_stmt_info) == vect_reduction_def)
+ return NULL;
+
/* Keep the first operand of a COND_EXPR as-is: only the other two
operands are interesting. */
unsigned int first_op = (code == COND_EXPR ? 2 : 1);