| 1 |
! { dg-do run } |
| 2 |
|
| 3 |
! See also '../libgomp.oacc-c-c++-common/f-asyncwait-1.c'. |
| 4 |
|
| 5 |
! { dg-additional-options "--param=openacc-kernels=decompose" } */ |
| 6 |
|
| 7 |
! { dg-additional-options "-fopt-info-all-omp" } |
| 8 |
! { dg-additional-options "-foffload=-fopt-info-all-omp" } */ |
| 9 |
|
| 10 |
! { dg-additional-options "--param=openacc-privatization=noisy" } |
| 11 |
! { dg-additional-options "-foffload=--param=openacc-privatization=noisy" } |
| 12 |
! Prune a few: uninteresting, and potentially varying depending on GCC configuration (data types): |
| 13 |
! { dg-prune-output {note: variable 'D\.[0-9]+' declared in block isn't candidate for adjusting OpenACC privatization level: not addressable} } */ |
| 14 |
|
| 15 |
! It's only with Tcl 8.5 (released in 2007) that "the variable 'varName' |
| 16 |
! passed to 'incr' may be unset, and in that case, it will be set to [...]", |
| 17 |
! so to maintain compatibility with earlier Tcl releases, we manually |
| 18 |
! initialize counter variables: |
| 19 |
! { dg-line l_dummy[variable c_compute 0 c_loop_i 0] } |
| 20 |
! { dg-message "dummy" "" { target iN-VAl-Id } l_dummy } to avoid |
| 21 |
! "WARNING: dg-line var l_dummy defined, but not used". */ |
| 22 |
|
| 23 |
program asyncwait |
| 24 |
integer, parameter :: N = 64 |
| 25 |
real, allocatable :: a(:), b(:), c(:), d(:), e(:) |
| 26 |
integer i |
| 27 |
|
| 28 |
allocate (a(N)) |
| 29 |
allocate (b(N)) |
| 30 |
allocate (c(N)) |
| 31 |
allocate (d(N)) |
| 32 |
allocate (e(N)) |
| 33 |
|
| 34 |
a(:) = 3.0 |
| 35 |
b(:) = 0.0 |
| 36 |
|
| 37 |
!$acc data copy (a(1:N)) copy (b(1:N)) |
| 38 |
|
| 39 |
!$acc parallel async |
| 40 |
!$acc loop ! { dg-line l_loop_i[incr c_loop_i] } |
| 41 |
! { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } l_loop_i$c_loop_i } |
| 42 |
! { dg-optimized "assigned OpenACC gang vector loop parallelism" "" { target *-*-* } l_loop_i$c_loop_i } |
| 43 |
do i = 1, N |
| 44 |
b(i) = a(i) |
| 45 |
end do |
| 46 |
!$acc end parallel |
| 47 |
|
| 48 |
!$acc wait |
| 49 |
!$acc end data |
| 50 |
|
| 51 |
do i = 1, N |
| 52 |
if (a(i) .ne. 3.0) STOP 1 |
| 53 |
if (b(i) .ne. 3.0) STOP 2 |
| 54 |
end do |
| 55 |
|
| 56 |
a(:) = 2.0 |
| 57 |
b(:) = 0.0 |
| 58 |
|
| 59 |
!$acc data copy (a(1:N)) copy (b(1:N)) |
| 60 |
|
| 61 |
!$acc parallel async (1) |
| 62 |
!$acc loop ! { dg-line l_loop_i[incr c_loop_i] } |
| 63 |
! { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } l_loop_i$c_loop_i } |
| 64 |
! { dg-optimized "assigned OpenACC gang vector loop parallelism" "" { target *-*-* } l_loop_i$c_loop_i } |
| 65 |
do i = 1, N |
| 66 |
b(i) = a(i) |
| 67 |
end do |
| 68 |
!$acc end parallel |
| 69 |
|
| 70 |
!$acc wait (1) |
| 71 |
!$acc end data |
| 72 |
|
| 73 |
do i = 1, N |
| 74 |
if (a(i) .ne. 2.0) STOP 3 |
| 75 |
if (b(i) .ne. 2.0) STOP 4 |
| 76 |
end do |
| 77 |
|
| 78 |
a(:) = 3.0 |
| 79 |
b(:) = 0.0 |
| 80 |
c(:) = 0.0 |
| 81 |
d(:) = 0.0 |
| 82 |
|
| 83 |
!$acc data copy (a(1:N)) copy (b(1:N)) copy (c(1:N)) copy (d(1:N)) |
| 84 |
|
| 85 |
!$acc parallel async (1) |
| 86 |
do i = 1, N |
| 87 |
b(i) = (a(i) * a(i) * a(i)) / a(i) |
| 88 |
end do |
| 89 |
!$acc end parallel |
| 90 |
|
| 91 |
!$acc parallel async (1) |
| 92 |
do i = 1, N |
| 93 |
c(i) = (a(i) * 4) / a(i) |
| 94 |
end do |
| 95 |
!$acc end parallel |
| 96 |
|
| 97 |
!$acc parallel async (1) |
| 98 |
!$acc loop ! { dg-line l_loop_i[incr c_loop_i] } |
| 99 |
! { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } l_loop_i$c_loop_i } |
| 100 |
! { dg-optimized "assigned OpenACC gang vector loop parallelism" "" { target *-*-* } l_loop_i$c_loop_i } |
| 101 |
do i = 1, N |
| 102 |
d(i) = ((a(i) * a(i) + a(i)) / a(i)) - a(i) |
| 103 |
end do |
| 104 |
!$acc end parallel |
| 105 |
|
| 106 |
!$acc wait (1) |
| 107 |
!$acc end data |
| 108 |
|
| 109 |
do i = 1, N |
| 110 |
if (a(i) .ne. 3.0) STOP 5 |
| 111 |
if (b(i) .ne. 9.0) STOP 6 |
| 112 |
if (c(i) .ne. 4.0) STOP 7 |
| 113 |
if (d(i) .ne. 1.0) STOP 8 |
| 114 |
end do |
| 115 |
|
| 116 |
a(:) = 2.0 |
| 117 |
b(:) = 0.0 |
| 118 |
c(:) = 0.0 |
| 119 |
d(:) = 0.0 |
| 120 |
e(:) = 0.0 |
| 121 |
|
| 122 |
!$acc data copy (a(1:N), b(1:N), c(1:N), d(1:N), e(1:N)) |
| 123 |
|
| 124 |
!$acc parallel async (1) |
| 125 |
do i = 1, N |
| 126 |
b(i) = (a(i) * a(i) * a(i)) / a(i) |
| 127 |
end do |
| 128 |
!$acc end parallel |
| 129 |
|
| 130 |
!$acc parallel async (1) |
| 131 |
!$acc loop ! { dg-line l_loop_i[incr c_loop_i] } |
| 132 |
! { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } l_loop_i$c_loop_i } |
| 133 |
! { dg-optimized "assigned OpenACC gang vector loop parallelism" "" { target *-*-* } l_loop_i$c_loop_i } |
| 134 |
do i = 1, N |
| 135 |
c(i) = (a(i) * 4) / a(i) |
| 136 |
end do |
| 137 |
!$acc end parallel |
| 138 |
|
| 139 |
!$acc parallel async (1) |
| 140 |
!$acc loop ! { dg-line l_loop_i[incr c_loop_i] } |
| 141 |
! { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } l_loop_i$c_loop_i } |
| 142 |
! { dg-optimized "assigned OpenACC gang vector loop parallelism" "" { target *-*-* } l_loop_i$c_loop_i } |
| 143 |
do i = 1, N |
| 144 |
d(i) = ((a(i) * a(i) + a(i)) / a(i)) - a(i) |
| 145 |
end do |
| 146 |
!$acc end parallel |
| 147 |
|
| 148 |
!$acc parallel wait (1) async (1) |
| 149 |
!$acc loop ! { dg-line l_loop_i[incr c_loop_i] } |
| 150 |
! { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } l_loop_i$c_loop_i } |
| 151 |
! { dg-optimized "assigned OpenACC gang vector loop parallelism" "" { target *-*-* } l_loop_i$c_loop_i } |
| 152 |
do i = 1, N |
| 153 |
e(i) = a(i) + b(i) + c(i) + d(i) |
| 154 |
end do |
| 155 |
!$acc end parallel |
| 156 |
|
| 157 |
!$acc wait (1) |
| 158 |
!$acc end data |
| 159 |
|
| 160 |
do i = 1, N |
| 161 |
if (a(i) .ne. 2.0) STOP 9 |
| 162 |
if (b(i) .ne. 4.0) STOP 10 |
| 163 |
if (c(i) .ne. 4.0) STOP 11 |
| 164 |
if (d(i) .ne. 1.0) STOP 12 |
| 165 |
if (e(i) .ne. 11.0) STOP 13 |
| 166 |
end do |
| 167 |
|
| 168 |
a(:) = 3.0 |
| 169 |
b(:) = 0.0 |
| 170 |
|
| 171 |
!$acc data copy (a(1:N)) copy (b(1:N)) |
| 172 |
|
| 173 |
!$acc kernels async |
| 174 |
!$acc loop ! { dg-line l_loop_i[incr c_loop_i] } |
| 175 |
! { dg-note {forwarded loop nest in OpenACC 'kernels' region to 'parloops' for analysis} {} { target *-*-* } l_loop_i$c_loop_i } |
| 176 |
! { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } l_loop_i$c_loop_i } |
| 177 |
! { dg-optimized "assigned OpenACC seq loop parallelism" "" { target *-*-* } l_loop_i$c_loop_i } |
| 178 |
do i = 1, N |
| 179 |
b(i) = a(i) |
| 180 |
end do |
| 181 |
!$acc end kernels |
| 182 |
|
| 183 |
!$acc wait |
| 184 |
!$acc end data |
| 185 |
|
| 186 |
do i = 1, N |
| 187 |
if (a(i) .ne. 3.0) STOP 14 |
| 188 |
if (b(i) .ne. 3.0) STOP 15 |
| 189 |
end do |
| 190 |
|
| 191 |
a(:) = 2.0 |
| 192 |
b(:) = 0.0 |
| 193 |
|
| 194 |
!$acc data copy (a(1:N)) copy (b(1:N)) |
| 195 |
|
| 196 |
!$acc kernels async (1) |
| 197 |
!$acc loop ! { dg-line l_loop_i[incr c_loop_i] } |
| 198 |
! { dg-note {forwarded loop nest in OpenACC 'kernels' region to 'parloops' for analysis} {} { target *-*-* } l_loop_i$c_loop_i } |
| 199 |
! { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } l_loop_i$c_loop_i } |
| 200 |
! { dg-optimized "assigned OpenACC seq loop parallelism" "" { target *-*-* } l_loop_i$c_loop_i } |
| 201 |
do i = 1, N |
| 202 |
b(i) = a(i) |
| 203 |
end do |
| 204 |
!$acc end kernels |
| 205 |
|
| 206 |
!$acc wait (1) |
| 207 |
!$acc end data |
| 208 |
|
| 209 |
do i = 1, N |
| 210 |
if (a(i) .ne. 2.0) STOP 16 |
| 211 |
if (b(i) .ne. 2.0) STOP 17 |
| 212 |
end do |
| 213 |
|
| 214 |
a(:) = 3.0 |
| 215 |
b(:) = 0.0 |
| 216 |
c(:) = 0.0 |
| 217 |
d(:) = 0.0 |
| 218 |
|
| 219 |
!$acc data copy (a(1:N)) copy (b(1:N)) copy (c(1:N)) copy (d(1:N)) |
| 220 |
|
| 221 |
!$acc kernels async (1) ! { dg-line l_compute[incr c_compute] } |
| 222 |
! { dg-note {OpenACC 'kernels' decomposition: variable 'i' in 'copy' clause requested to be made addressable} {} { target *-*-* } l_compute$c_compute } |
| 223 |
! { dg-note {variable 'i' made addressable} {} { target *-*-* } l_compute$c_compute } */ |
| 224 |
! { dg-optimized "assigned OpenACC seq loop parallelism" "" { target *-*-* } l_compute$c_compute } |
| 225 |
! { dg-note {beginning 'parloops' part in OpenACC 'kernels' region} {} { target *-*-* } .+1 } |
| 226 |
do i = 1, N |
| 227 |
b(i) = (a(i) * a(i) * a(i)) / a(i) |
| 228 |
end do |
| 229 |
!$acc end kernels |
| 230 |
|
| 231 |
!$acc kernels async (1) ! { dg-line l_compute[incr c_compute] } |
| 232 |
! { dg-note {OpenACC 'kernels' decomposition: variable 'i' in 'copy' clause requested to be made addressable} {} { target *-*-* } l_compute$c_compute } |
| 233 |
! { dg-note {variable 'i' already made addressable} {} { target *-*-* } l_compute$c_compute } */ |
| 234 |
! { dg-optimized "assigned OpenACC seq loop parallelism" "" { target *-*-* } l_compute$c_compute } |
| 235 |
! { dg-note {beginning 'parloops' part in OpenACC 'kernels' region} {} { target *-*-* } .+1 } |
| 236 |
do i = 1, N |
| 237 |
c(i) = (a(i) * 4) / a(i) |
| 238 |
end do |
| 239 |
!$acc end kernels |
| 240 |
|
| 241 |
!$acc kernels async (1) |
| 242 |
!$acc loop ! { dg-line l_loop_i[incr c_loop_i] } |
| 243 |
! { dg-note {forwarded loop nest in OpenACC 'kernels' region to 'parloops' for analysis} {} { target *-*-* } l_loop_i$c_loop_i } |
| 244 |
! { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } l_loop_i$c_loop_i } |
| 245 |
! { dg-optimized "assigned OpenACC seq loop parallelism" "" { target *-*-* } l_loop_i$c_loop_i } |
| 246 |
do i = 1, N |
| 247 |
d(i) = ((a(i) * a(i) + a(i)) / a(i)) - a(i) |
| 248 |
end do |
| 249 |
!$acc end kernels |
| 250 |
|
| 251 |
!$acc wait (1) |
| 252 |
!$acc end data |
| 253 |
|
| 254 |
do i = 1, N |
| 255 |
if (a(i) .ne. 3.0) STOP 18 |
| 256 |
if (b(i) .ne. 9.0) STOP 19 |
| 257 |
if (c(i) .ne. 4.0) STOP 20 |
| 258 |
if (d(i) .ne. 1.0) STOP 21 |
| 259 |
end do |
| 260 |
|
| 261 |
a(:) = 2.0 |
| 262 |
b(:) = 0.0 |
| 263 |
c(:) = 0.0 |
| 264 |
d(:) = 0.0 |
| 265 |
e(:) = 0.0 |
| 266 |
|
| 267 |
!$acc data copy (a(1:N), b(1:N), c(1:N), d(1:N), e(1:N)) |
| 268 |
|
| 269 |
!$acc kernels async (1) ! { dg-line l_compute[incr c_compute] } |
| 270 |
! { dg-note {OpenACC 'kernels' decomposition: variable 'i' in 'copy' clause requested to be made addressable} {} { target *-*-* } l_compute$c_compute } |
| 271 |
! { dg-note {variable 'i' already made addressable} {} { target *-*-* } l_compute$c_compute } */ |
| 272 |
! { dg-optimized "assigned OpenACC seq loop parallelism" "" { target *-*-* } l_compute$c_compute } |
| 273 |
! { dg-note {beginning 'parloops' part in OpenACC 'kernels' region} {} { target *-*-* } .+1 } |
| 274 |
do i = 1, N |
| 275 |
b(i) = (a(i) * a(i) * a(i)) / a(i) |
| 276 |
end do |
| 277 |
!$acc end kernels |
| 278 |
|
| 279 |
!$acc kernels async (1) |
| 280 |
!$acc loop ! { dg-line l_loop_i[incr c_loop_i] } |
| 281 |
! { dg-note {forwarded loop nest in OpenACC 'kernels' region to 'parloops' for analysis} {} { target *-*-* } l_loop_i$c_loop_i } |
| 282 |
! { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } l_loop_i$c_loop_i } |
| 283 |
! { dg-optimized "assigned OpenACC seq loop parallelism" "" { target *-*-* } l_loop_i$c_loop_i } |
| 284 |
do i = 1, N |
| 285 |
c(i) = (a(i) * 4) / a(i) |
| 286 |
end do |
| 287 |
!$acc end kernels |
| 288 |
|
| 289 |
!$acc kernels async (1) |
| 290 |
!$acc loop ! { dg-line l_loop_i[incr c_loop_i] } |
| 291 |
! { dg-note {forwarded loop nest in OpenACC 'kernels' region to 'parloops' for analysis} {} { target *-*-* } l_loop_i$c_loop_i } |
| 292 |
! { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } l_loop_i$c_loop_i } |
| 293 |
! { dg-optimized "assigned OpenACC seq loop parallelism" "" { target *-*-* } l_loop_i$c_loop_i } |
| 294 |
do i = 1, N |
| 295 |
d(i) = ((a(i) * a(i) + a(i)) / a(i)) - a(i) |
| 296 |
end do |
| 297 |
!$acc end kernels |
| 298 |
|
| 299 |
!$acc kernels wait (1) async (1) |
| 300 |
!$acc loop ! { dg-line l_loop_i[incr c_loop_i] } |
| 301 |
! { dg-note {forwarded loop nest in OpenACC 'kernels' region to 'parloops' for analysis} {} { target *-*-* } l_loop_i$c_loop_i } |
| 302 |
! { dg-note {variable 'i' in 'private' clause isn't candidate for adjusting OpenACC privatization level: not addressable} "" { target *-*-* } l_loop_i$c_loop_i } |
| 303 |
! { dg-optimized "assigned OpenACC seq loop parallelism" "" { target *-*-* } l_loop_i$c_loop_i } |
| 304 |
do i = 1, N |
| 305 |
e(i) = a(i) + b(i) + c(i) + d(i) |
| 306 |
end do |
| 307 |
!$acc end kernels |
| 308 |
|
| 309 |
!$acc wait (1) |
| 310 |
!$acc end data |
| 311 |
|
| 312 |
do i = 1, N |
| 313 |
if (a(i) .ne. 2.0) STOP 22 |
| 314 |
if (b(i) .ne. 4.0) STOP 23 |
| 315 |
if (c(i) .ne. 4.0) STOP 24 |
| 316 |
if (d(i) .ne. 1.0) STOP 25 |
| 317 |
if (e(i) .ne. 11.0) STOP 26 |
| 318 |
end do |
| 319 |
end program asyncwait |