Debian Codegraph @

gcc/testsuite/gcc.target/s390/vector/vperm-rev-z14.c

raw
1 /* Make sure that the reverse permute patterns are optimized
2 correctly. */
3 /* { dg-do run { target { s390*-*-* } } } */
4 /* { dg-options "-O2 -march=z14 -mzarch -fno-unroll-loops -save-temps" } */
5
6 /* { dg-final { scan-assembler-times "vpdi\t" 4 } } */
7 /* { dg-final { scan-assembler-times "verllg\t" 2 } } */
8
9 #include <assert.h>
10
11 __attribute__((noipa))
12 void reversel (long long *restrict a, long long *restrict b, int n)
13 {
14 for (int i = 0; i < n; i += 2)
15 {
16 a[i + 1] = b[i + 0];
17 a[i + 0] = b[i + 1];
18 }
19 }
20
21 __attribute__((noipa))
22 void reversed (double *restrict a, double *restrict b, int n)
23 {
24 for (int i = 0; i < n; i += 2)
25 {
26 a[i + 1] = b[i + 0];
27 a[i + 0] = b[i + 1];
28 }
29 }
30
31 __attribute__((noipa))
32 void reversei (unsigned int *restrict a, unsigned int *restrict b, int n)
33 {
34 for (int i = 0; i < n; i += 4)
35 {
36 a[i + 3] = b[i + 0];
37 a[i + 2] = b[i + 1];
38 a[i + 1] = b[i + 2];
39 a[i + 0] = b[i + 3];
40 }
41 }
42
43 __attribute__((noipa))
44 void reversef (float *restrict a, float *restrict b, int n)
45 {
46 for (int i = 0; i < n; i += 4)
47 {
48 a[i + 3] = b[i + 0];
49 a[i + 2] = b[i + 1];
50 a[i + 1] = b[i + 2];
51 a[i + 0] = b[i + 3];
52 }
53 }
54
55 int main()
56 {
57 const int n = 1024;
58 unsigned int u[n], u2[n];
59 long long l[n], l2[n];
60 double d[n], d2[n];
61 float f[n], f2[n];
62
63 for (int i = 0; i < n; i++)
64 {
65 u[i] = i;
66 l[i] = i;
67 d[i] = i;
68 f[i] = i;
69 u2[i] = i;
70 l2[i] = i;
71 d2[i] = i;
72 f2[i] = i;
73 }
74
75 reversei (u2, u, n);
76 reversel (l2, l, n);
77 reversed (d2, d, n);
78 reversef (f2, f, n);
79
80 for (int i = 0; i < n - 16; i++)
81 {
82 assert (u[i] == u2[i / (16 / sizeof (u[0])) * (16 / sizeof (u[0])) + 16 / sizeof (u[0]) - 1 - i % (16 / sizeof (u[0]))]);
83 assert (l[i] == l2[i / (16 / sizeof (l[0])) * (16 / sizeof (l[0])) + 16 / sizeof (l[0]) - 1 - i % (16 / sizeof (l[0]))]);
84 assert (d[i] == d2[i / (16 / sizeof (d[0])) * (16 / sizeof (d[0])) + 16 / sizeof (d[0]) - 1 - i % (16 / sizeof (d[0]))]);
85 assert (f[i] == f2[i / (16 / sizeof (f[0])) * (16 / sizeof (f[0])) + 16 / sizeof (f[0]) - 1 - i % (16 / sizeof (f[0]))]);
86 }
87 }