| 1 |
/* { dg-do run } */ |
| 2 |
/* { dg-options "-O2" } */ |
| 3 |
/* { dg-additional-options "-std=c99" { target c } } */ |
| 4 |
/* { dg-additional-options "-msse2" { target sse2_runtime } } */ |
| 5 |
/* { dg-additional-options "-mavx" { target avx_runtime } } */ |
| 6 |
|
| 7 |
int u[1024], v[1024], w[1024], m; |
| 8 |
|
| 9 |
__attribute__((noinline, noclone)) void |
| 10 |
f1 (long a, long b) |
| 11 |
{ |
| 12 |
#pragma omp taskloop simd default(none) shared(u, v, w) nogroup |
| 13 |
for (long d = a; d < b; d++) |
| 14 |
u[d] = v[d] + w[d]; |
| 15 |
} |
| 16 |
|
| 17 |
__attribute__((noinline, noclone)) int |
| 18 |
f2 (long a, long b, long c) |
| 19 |
{ |
| 20 |
int d, e; |
| 21 |
#pragma omp taskloop simd default(none) shared(u, v, w) linear(d:1) linear(c:5) lastprivate(e) |
| 22 |
for (d = a; d < b; d++) |
| 23 |
{ |
| 24 |
u[d] = v[d] + w[d]; |
| 25 |
c = c + 5; |
| 26 |
e = c + 9; |
| 27 |
} |
| 28 |
return d + c + e; |
| 29 |
} |
| 30 |
|
| 31 |
__attribute__((noinline, noclone)) int |
| 32 |
f3 (long a, long b) |
| 33 |
{ |
| 34 |
int d; |
| 35 |
#pragma omp taskloop simd default(none) shared(u, v, w) |
| 36 |
for (d = a; d < b; d++) |
| 37 |
{ |
| 38 |
int *p = &d; |
| 39 |
u[d] = v[d] + w[d]; |
| 40 |
} |
| 41 |
return d; |
| 42 |
} |
| 43 |
|
| 44 |
__attribute__((noinline, noclone)) int |
| 45 |
f4 (long a, long b, long c, long d) |
| 46 |
{ |
| 47 |
int e, f, g; |
| 48 |
#pragma omp taskloop simd default(none) shared(u, v, w) collapse(2) lastprivate(g) |
| 49 |
for (e = a; e < b; e++) |
| 50 |
for (f = c; f < d; f++) |
| 51 |
{ |
| 52 |
int *p = &e; |
| 53 |
int *q = &f; |
| 54 |
int r = 32 * e + f; |
| 55 |
u[r] = v[r] + w[r]; |
| 56 |
g = r; |
| 57 |
} |
| 58 |
return e + f + g; |
| 59 |
} |
| 60 |
|
| 61 |
__attribute__((noinline, noclone)) int |
| 62 |
f5 (long a, long b, long c, long d) |
| 63 |
{ |
| 64 |
int e, f; |
| 65 |
#pragma omp taskloop simd default(none) shared(u, v, w) collapse(2) |
| 66 |
for (e = a; e < b; e++) |
| 67 |
for (f = c; f < d; f++) |
| 68 |
{ |
| 69 |
int r = 32 * e + f; |
| 70 |
u[r] = v[r] + w[r]; |
| 71 |
} |
| 72 |
return e + f; |
| 73 |
} |
| 74 |
|
| 75 |
int |
| 76 |
main () |
| 77 |
{ |
| 78 |
int i; |
| 79 |
for (i = 0; i < 1024; i++) |
| 80 |
{ |
| 81 |
v[i] = i; |
| 82 |
w[i] = i + 1; |
| 83 |
} |
| 84 |
#pragma omp parallel |
| 85 |
#pragma omp single |
| 86 |
f1 (0, 1024); |
| 87 |
for (i = 0; i < 1024; i++) |
| 88 |
if (u[i] != 2 * i + 1) |
| 89 |
__builtin_abort (); |
| 90 |
else |
| 91 |
{ |
| 92 |
v[i] = 1024 - i; |
| 93 |
w[i] = 512 - i; |
| 94 |
} |
| 95 |
#pragma omp parallel |
| 96 |
#pragma omp single |
| 97 |
m = f2 (2, 1022, 17); |
| 98 |
for (i = 0; i < 1024; i++) |
| 99 |
if ((i < 2 || i >= 1022) ? u[i] != 2 * i + 1 : u[i] != 1536 - 2 * i) |
| 100 |
__builtin_abort (); |
| 101 |
else |
| 102 |
{ |
| 103 |
v[i] = i; |
| 104 |
w[i] = i + 1; |
| 105 |
} |
| 106 |
if (m != 1022 + 2 * (1020 * 5 + 17) + 9) |
| 107 |
__builtin_abort (); |
| 108 |
#pragma omp parallel |
| 109 |
#pragma omp single |
| 110 |
m = f3 (0, 1024); |
| 111 |
for (i = 0; i < 1024; i++) |
| 112 |
if (u[i] != 2 * i + 1) |
| 113 |
__builtin_abort (); |
| 114 |
else |
| 115 |
{ |
| 116 |
v[i] = 1024 - i; |
| 117 |
w[i] = 512 - i; |
| 118 |
} |
| 119 |
if (m != 1024) |
| 120 |
__builtin_abort (); |
| 121 |
#pragma omp parallel |
| 122 |
#pragma omp single |
| 123 |
m = f4 (0, 32, 0, 32); |
| 124 |
for (i = 0; i < 1024; i++) |
| 125 |
if (u[i] != 1536 - 2 * i) |
| 126 |
__builtin_abort (); |
| 127 |
else |
| 128 |
{ |
| 129 |
v[i] = i; |
| 130 |
w[i] = i + 1; |
| 131 |
} |
| 132 |
if (m != 32 + 32 + 1023) |
| 133 |
__builtin_abort (); |
| 134 |
#pragma omp parallel |
| 135 |
#pragma omp single |
| 136 |
m = f5 (0, 32, 0, 32); |
| 137 |
for (i = 0; i < 1024; i++) |
| 138 |
if (u[i] != 2 * i + 1) |
| 139 |
__builtin_abort (); |
| 140 |
else |
| 141 |
{ |
| 142 |
v[i] = 1024 - i; |
| 143 |
w[i] = 512 - i; |
| 144 |
} |
| 145 |
if (m != 32 + 32) |
| 146 |
__builtin_abort (); |
| 147 |
return 0; |
| 148 |
} |