1 /* { dg-require-effective-target vect_int } */
2
3 #include <stdarg.h>
4 #include "tree-vect.h"
5
6 #define M00 100
7 #define M10 216
8 #define M20 23
9 #define M01 1322
10 #define M11 13
11 #define M21 27271
12 #define M02 74
13 #define M12 191
14 #define M22 500
15
16 #define N 16
17
foo(unsigned int * __restrict__ pInput,unsigned int * __restrict__ pOutput)18 void foo (unsigned int *__restrict__ pInput, unsigned int *__restrict__ pOutput)
19 {
20 unsigned int i, a, b, c;
21
22 for (i = 0; i < N / 3; i++)
23 {
24 a = *pInput++;
25 b = *pInput++;
26 c = *pInput++;
27
28 *pOutput++ = M00 * a + M01 * b + M02 * c;
29 *pOutput++ = M10 * a + M11 * b + M12 * c;
30 *pOutput++ = M20 * a + M21 * b + M22 * c;
31 }
32 }
33
main(int argc,const char * argv[])34 int main (int argc, const char* argv[])
35 {
36 unsigned int input[N], output[N], i;
37 unsigned int check_results[N] = {1470, 395, 28271, 5958, 1655, 111653, 10446, 2915, 195035, 14934, 4175, 278417, 19422, 5435, 361799, 0};
38
39 check_vect ();
40
41 for (i = 0; i < N; i++)
42 {
43 input[i] = i%256;
44 output[i] = 0;
45 __asm__ volatile ("");
46 }
47
48 foo (input, output);
49
50 for (i = 0; i < N; i++)
51 {
52 if (output[i] != check_results[i])
53 abort ();
54 __asm__ volatile ("");
55 }
56
57 return 0;
58 }
59
60 /* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { target vect_perm } } } */
61 /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "vect" { target { vect_perm && {! vect_load_lanes } } } } } */
62 /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "vect" { target vect_load_lanes } } } */
63 /* { dg-final { scan-tree-dump "note: Built SLP cancelled: can use load/store-lanes" "vect" { target { vect_perm && vect_load_lanes } } } } */
64 /* { dg-final { scan-tree-dump "LOAD_LANES" "vect" { target vect_load_lanes } } } */
65 /* { dg-final { scan-tree-dump "STORE_LANES" "vect" { target vect_load_lanes } } } */
66
67