| 1 |
int a[256]; |
| 2 |
|
| 3 |
__attribute__((noinline, noclone)) int |
| 4 |
f1 (int i) |
| 5 |
{ |
| 6 |
#pragma omp parallel for linear (i: 4) |
| 7 |
for (int j = 16; j < 64; j++) |
| 8 |
{ |
| 9 |
a[i] = j; |
| 10 |
i += 4; |
| 11 |
} |
| 12 |
return i; |
| 13 |
} |
| 14 |
|
| 15 |
__attribute__((noinline, noclone)) short int & |
| 16 |
f2 (short int &i, char k) |
| 17 |
{ |
| 18 |
#pragma omp parallel for linear (i: k + 1) |
| 19 |
for (long j = 16; j < 64; j++) |
| 20 |
{ |
| 21 |
a[i] = j; |
| 22 |
i += 4; |
| 23 |
} |
| 24 |
return i; |
| 25 |
} |
| 26 |
|
| 27 |
template <typename T> |
| 28 |
__attribute__((noinline, noclone)) T |
| 29 |
f3 (T i, T k) |
| 30 |
{ |
| 31 |
#pragma omp parallel for linear (i: k) |
| 32 |
for (short j = 16; j < 64; j++) |
| 33 |
{ |
| 34 |
a[i] = j; |
| 35 |
i += 4; |
| 36 |
} |
| 37 |
return i; |
| 38 |
} |
| 39 |
|
| 40 |
template <typename T> |
| 41 |
__attribute__((noinline, noclone)) T & |
| 42 |
f4 (T &i) |
| 43 |
{ |
| 44 |
#pragma omp parallel for linear (i: 4) schedule(static, 3) |
| 45 |
for (int j = 16; j < 64; j++) |
| 46 |
{ |
| 47 |
a[i] = j; |
| 48 |
i += 4; |
| 49 |
} |
| 50 |
return i; |
| 51 |
} |
| 52 |
|
| 53 |
__attribute__((noinline, noclone)) short int |
| 54 |
f5 (short int i, char &k) |
| 55 |
{ |
| 56 |
#pragma omp parallel for linear (i: k + 1) schedule(static, 5) |
| 57 |
for (long j = 16; j < 64; j++) |
| 58 |
{ |
| 59 |
a[i] = j; |
| 60 |
i += 4; |
| 61 |
} |
| 62 |
return i; |
| 63 |
} |
| 64 |
|
| 65 |
template <int N> |
| 66 |
__attribute__((noinline, noclone)) long long int |
| 67 |
f6 (long long int i, long long int k) |
| 68 |
{ |
| 69 |
#pragma omp parallel for linear (i: k) schedule(static, 7) |
| 70 |
for (short j = 16; j < 64; j++) |
| 71 |
{ |
| 72 |
a[i] = j; |
| 73 |
i += 4; |
| 74 |
} |
| 75 |
return i; |
| 76 |
} |
| 77 |
|
| 78 |
__attribute__((noinline, noclone)) int |
| 79 |
f7 (int &i) |
| 80 |
{ |
| 81 |
#pragma omp parallel for linear (i: 4) schedule(dynamic, 3) |
| 82 |
for (int j = 16; j < 64; j++) |
| 83 |
{ |
| 84 |
a[i] = j; |
| 85 |
i += 4; |
| 86 |
} |
| 87 |
return i; |
| 88 |
} |
| 89 |
|
| 90 |
__attribute__((noinline, noclone)) short int |
| 91 |
f8 (short int i, char k) |
| 92 |
{ |
| 93 |
#pragma omp parallel for linear (i: k + 1) schedule(dynamic, 5) |
| 94 |
for (long j = 16; j < 64; j++) |
| 95 |
{ |
| 96 |
a[i] = j; |
| 97 |
i += 4; |
| 98 |
} |
| 99 |
return i; |
| 100 |
} |
| 101 |
|
| 102 |
__attribute__((noinline, noclone)) long long int |
| 103 |
f9 (long long int i, long long int k) |
| 104 |
{ |
| 105 |
#pragma omp parallel for linear (i: k) schedule(dynamic, 7) |
| 106 |
for (short j = 16; j < 64; j++) |
| 107 |
{ |
| 108 |
a[i] = j; |
| 109 |
i += 4; |
| 110 |
} |
| 111 |
return i; |
| 112 |
} |
| 113 |
|
| 114 |
template <typename T> |
| 115 |
__attribute__((noinline, noclone)) T & |
| 116 |
f10 (T &i, long &step) |
| 117 |
{ |
| 118 |
#pragma omp parallel for linear (i: 4) |
| 119 |
for (int j = 16; j < 112; j += step) |
| 120 |
{ |
| 121 |
a[i] = j / 2 + 8; |
| 122 |
i += 4; |
| 123 |
} |
| 124 |
return i; |
| 125 |
} |
| 126 |
|
| 127 |
__attribute__((noinline, noclone)) short int |
| 128 |
f11 (short int i, char k, char step) |
| 129 |
{ |
| 130 |
#pragma omp parallel for linear (i: k + 1) |
| 131 |
for (long j = 16; j < 112; j += step) |
| 132 |
{ |
| 133 |
a[i] = j / 2 + 8; |
| 134 |
i += 4; |
| 135 |
} |
| 136 |
return i; |
| 137 |
} |
| 138 |
|
| 139 |
__attribute__((noinline, noclone)) long long int |
| 140 |
f12 (long long int i, long long int k, int step) |
| 141 |
{ |
| 142 |
#pragma omp parallel for linear (i: k) |
| 143 |
for (short j = 16; j < 112; j += step) |
| 144 |
{ |
| 145 |
a[i] = j / 2 + 8; |
| 146 |
i += 4; |
| 147 |
} |
| 148 |
return i; |
| 149 |
} |
| 150 |
|
| 151 |
__attribute__((noinline, noclone)) int |
| 152 |
f13 (int &i, long long int step) |
| 153 |
{ |
| 154 |
#pragma omp parallel for linear (i: 4) schedule(static, 3) |
| 155 |
for (int j = 16; j < 112; j += step) |
| 156 |
{ |
| 157 |
a[i] = j / 2 + 8; |
| 158 |
i += 4; |
| 159 |
} |
| 160 |
return i; |
| 161 |
} |
| 162 |
|
| 163 |
__attribute__((noinline, noclone)) short int |
| 164 |
f14 (short int &i, char &k, int &step) |
| 165 |
{ |
| 166 |
#pragma omp parallel for linear (i: k + 1) schedule(static, 5) |
| 167 |
for (long j = 16; j < 112; j += step) |
| 168 |
{ |
| 169 |
a[i] = j / 2 + 8; |
| 170 |
i += 4; |
| 171 |
} |
| 172 |
return i; |
| 173 |
} |
| 174 |
|
| 175 |
template <int N> |
| 176 |
__attribute__((noinline, noclone)) long long int |
| 177 |
f15 (long long int i, long long int k, long int step) |
| 178 |
{ |
| 179 |
#pragma omp parallel for linear (i: k) schedule(static, 7) |
| 180 |
for (short j = 16; j < 112; j += step) |
| 181 |
{ |
| 182 |
a[i] = j / 2 + 8; |
| 183 |
i += 4; |
| 184 |
} |
| 185 |
return i; |
| 186 |
} |
| 187 |
|
| 188 |
__attribute__((noinline, noclone)) int |
| 189 |
f16 (int i, long long int step) |
| 190 |
{ |
| 191 |
#pragma omp parallel for linear (i: 4) schedule(dynamic, 3) |
| 192 |
for (int j = 16; j < 112; j += step) |
| 193 |
{ |
| 194 |
a[i] = j / 2 + 8; |
| 195 |
i += 4; |
| 196 |
} |
| 197 |
return i; |
| 198 |
} |
| 199 |
|
| 200 |
__attribute__((noinline, noclone)) short int |
| 201 |
f17 (short int i, char k, int step) |
| 202 |
{ |
| 203 |
#pragma omp parallel for linear (i: k + 1) schedule(dynamic, 5) |
| 204 |
for (long j = 16; j < 112; j += step) |
| 205 |
{ |
| 206 |
a[i] = j / 2 + 8; |
| 207 |
i += 4; |
| 208 |
} |
| 209 |
return i; |
| 210 |
} |
| 211 |
|
| 212 |
template <typename T> |
| 213 |
__attribute__((noinline, noclone)) T |
| 214 |
f18 (T i, T k, long int step) |
| 215 |
{ |
| 216 |
#pragma omp parallel for linear (i: k) schedule(dynamic, 7) |
| 217 |
for (short j = 16; j < 112; j += step) |
| 218 |
{ |
| 219 |
a[i] = j / 2 + 8; |
| 220 |
i += 4; |
| 221 |
} |
| 222 |
return i; |
| 223 |
} |
| 224 |
|
| 225 |
int |
| 226 |
main () |
| 227 |
{ |
| 228 |
#define TEST(x) \ |
| 229 |
if (x != 8 + 48 * 4) \ |
| 230 |
__builtin_abort (); \ |
| 231 |
for (int i = 0; i < 256; i++) \ |
| 232 |
if (a[i] != (((i & 3) == 0 && i >= 8 \ |
| 233 |
&& i < 8 + 48 * 4) \ |
| 234 |
? ((i - 8) / 4) + 16 : 0)) \ |
| 235 |
__builtin_abort (); \ |
| 236 |
__builtin_memset (a, 0, sizeof (a)) |
| 237 |
TEST (f1 (8)); |
| 238 |
short int vs = 8; |
| 239 |
TEST (f2 (vs, 3)); |
| 240 |
TEST (f3 (8LL, 4LL)); |
| 241 |
int vi = 8; |
| 242 |
TEST (f4 (vi)); |
| 243 |
char vk = 3; |
| 244 |
TEST (f5 (8, vk)); |
| 245 |
TEST (f6<7> (8LL, 4LL)); |
| 246 |
vi = 8; |
| 247 |
TEST (f7 (vi)); |
| 248 |
TEST (f8 (8, 3)); |
| 249 |
TEST (f9 (8LL, 4LL)); |
| 250 |
vi = 8; |
| 251 |
long vl = 2; |
| 252 |
TEST (f10 (vi, vl)); |
| 253 |
TEST (f11 (8, 3, 2)); |
| 254 |
TEST (f12 (8LL, 4LL, 2)); |
| 255 |
vi = 8; |
| 256 |
TEST (f13 (vi, 2)); |
| 257 |
vs = 8; |
| 258 |
vk = 3; |
| 259 |
vi = 2; |
| 260 |
TEST (f14 (vs, vk, vi)); |
| 261 |
TEST (f15<9> (8LL, 4LL, 2)); |
| 262 |
TEST (f16 (8, 2)); |
| 263 |
TEST (f17 (8, 3, 2)); |
| 264 |
long long int vll1 = 8LL; |
| 265 |
long long int vll2 = 4LL; |
| 266 |
TEST (f18<long long int &> (vll1, vll2, 2)); |
| 267 |
return 0; |
| 268 |
} |