fc875865208a0db31db39fd6d67b4d6a27ed3ed2
1 /* { dg-require-effective-target vect_int } */
2 /* { dg-additional-options "--param=vect-epilogues-nomask=0" } */
12 unsigned short X
[N
] __attribute__ ((__aligned__(__BIGGEST_ALIGNMENT__
)));
13 unsigned short Y
[N
] __attribute__ ((__aligned__(__BIGGEST_ALIGNMENT__
)));
15 /* short->short->int dot product.
16 Not detected as a dot-product pattern.
17 Requires support for non-widneing multiplication and widening-summation.
18 Vectorized with loop-aware SLP. */
19 __attribute__ ((noinline
)) unsigned int
20 foo1(int len
, int *result1
, int *result2
)
23 unsigned int res1
= 10, res2
= 20;
26 for (i
=0; i
<len
; i
++) {
27 prod
= X
[2*i
] * Y
[2*i
];
29 prod
= X
[2*i
+1] * Y
[2*i
+1];
41 unsigned int dot1
, dot2
;
51 foo1 (N
/2, &dot1
, &dot2
);
53 if (dot1
!= DOT1
|| dot2
!= DOT2
)
59 /* The initialization loop in main also gets vectorized. */
60 /* { dg-final { scan-tree-dump-times "vect_recog_dot_prod_pattern: detected" 1 "vect" { xfail *-*-* } } } */
61 /* { dg-final { scan-tree-dump-times "vectorized 1 loops" 2 "vect" { target { vect_short_mult && { vect_widen_sum_hi_to_si && vect_unpack } } } } } */
62 /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "vect" { xfail { vect_widen_sum_hi_to_si_pattern || { ! vect_unpack } } } } } */
63 /* Check we can elide permutes if SLP vectorizing the reduction. */
64 /* { dg-final { scan-tree-dump-times "VEC_PERM_EXPR" 0 "vect" { xfail { vect_widen_sum_hi_to_si_pattern || { ! vect_unpack } } } } } */