| 1 | SUBROUTINE transpose_xy( f_in, work, f_out ) |
|---|
| 2 | |
|---|
| 3 | !--------------------------------------------------------------------------------! |
|---|
| 4 | ! This file is part of PALM. |
|---|
| 5 | ! |
|---|
| 6 | ! PALM is free software: you can redistribute it and/or modify it under the terms |
|---|
| 7 | ! of the GNU General Public License as published by the Free Software Foundation, |
|---|
| 8 | ! either version 3 of the License, or (at your option) any later version. |
|---|
| 9 | ! |
|---|
| 10 | ! PALM is distributed in the hope that it will be useful, but WITHOUT ANY |
|---|
| 11 | ! WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR |
|---|
| 12 | ! A PARTICULAR PURPOSE. See the GNU General Public License for more details. |
|---|
| 13 | ! |
|---|
| 14 | ! You should have received a copy of the GNU General Public License along with |
|---|
| 15 | ! PALM. If not, see <http://www.gnu.org/licenses/>. |
|---|
| 16 | ! |
|---|
| 17 | ! Copyright 1997-2012 Leibniz University Hannover |
|---|
| 18 | !--------------------------------------------------------------------------------! |
|---|
| 19 | ! |
|---|
| 20 | ! Current revisions: |
|---|
| 21 | ! ----------------- |
|---|
| 22 | ! |
|---|
| 23 | ! |
|---|
| 24 | ! Former revisions: |
|---|
| 25 | ! ----------------- |
|---|
| 26 | ! $Id: transpose.f90 1107 2013-03-04 06:23:14Z raasch $ |
|---|
| 27 | ! |
|---|
| 28 | ! 1106 2013-03-04 05:31:38Z raasch |
|---|
| 29 | ! preprocessor lines rearranged so that routines can also be used in serial |
|---|
| 30 | ! (non-parallel) mode |
|---|
| 31 | ! |
|---|
| 32 | ! 1092 2013-02-02 11:24:22Z raasch |
|---|
| 33 | ! unused variables removed |
|---|
| 34 | ! |
|---|
| 35 | ! 1036 2012-10-22 13:43:42Z raasch |
|---|
| 36 | ! code put under GPL (PALM 3.9) |
|---|
| 37 | ! |
|---|
| 38 | ! 1003 2012-09-14 14:35:53Z raasch |
|---|
| 39 | ! indices nxa, nya, etc. replaced by nx, ny, etc. |
|---|
| 40 | ! |
|---|
| 41 | ! 683 2011-02-09 14:25:15Z raasch |
|---|
| 42 | ! openMP parallelization of transpositions for 2d-domain-decomposition |
|---|
| 43 | ! |
|---|
| 44 | ! 622 2010-12-10 08:08:13Z raasch |
|---|
| 45 | ! optional barriers included in order to speed up collective operations |
|---|
| 46 | ! |
|---|
| 47 | ! 164 2008-05-15 08:46:15Z raasch |
|---|
| 48 | ! f_inv changed from subroutine argument to automatic array in order to do |
|---|
| 49 | ! re-ordering from f_in to f_inv in one step, one array work is needed instead |
|---|
| 50 | ! of work1 and work2 |
|---|
| 51 | ! |
|---|
| 52 | ! February 2007 |
|---|
| 53 | ! RCS Log replace by Id keyword, revision history cleaned up |
|---|
| 54 | ! |
|---|
| 55 | ! Revision 1.2 2004/04/30 13:12:17 raasch |
|---|
| 56 | ! Switched from mpi_alltoallv to the simpler mpi_alltoall, |
|---|
| 57 | ! all former transpose-routine files collected in this file, enlarged |
|---|
| 58 | ! transposition arrays introduced |
|---|
| 59 | ! |
|---|
| 60 | ! Revision 1.1 2004/04/30 13:08:16 raasch |
|---|
| 61 | ! Initial revision (collection of former routines transpose_xy, transpose_xz, |
|---|
| 62 | ! transpose_yx, transpose_yz, transpose_zx, transpose_zy) |
|---|
| 63 | ! |
|---|
| 64 | ! Revision 1.1 1997/07/24 11:25:18 raasch |
|---|
| 65 | ! Initial revision |
|---|
| 66 | ! |
|---|
| 67 | ! |
|---|
| 68 | ! Description: |
|---|
| 69 | ! ------------ |
|---|
| 70 | ! Transposition of input array (f_in) from x to y. For the input array, all |
|---|
| 71 | ! elements along x reside on the same PE, while after transposition, all |
|---|
| 72 | ! elements along y reside on the same PE. |
|---|
| 73 | !------------------------------------------------------------------------------! |
|---|
| 74 | |
|---|
| 75 | USE cpulog |
|---|
| 76 | USE indices |
|---|
| 77 | USE interfaces |
|---|
| 78 | USE pegrid |
|---|
| 79 | USE transpose_indices |
|---|
| 80 | |
|---|
| 81 | IMPLICIT NONE |
|---|
| 82 | |
|---|
| 83 | INTEGER :: i, j, k, l, m, ys |
|---|
| 84 | |
|---|
| 85 | REAL :: f_in(0:nx,nys_x:nyn_x,nzb_x:nzt_x), & |
|---|
| 86 | f_inv(nys_x:nyn_x,nzb_x:nzt_x,0:nx), & |
|---|
| 87 | f_out(0:ny,nxl_y:nxr_y,nzb_y:nzt_y), & |
|---|
| 88 | work(nnx*nny*nnz) |
|---|
| 89 | |
|---|
| 90 | ! |
|---|
| 91 | !-- Rearrange indices of input array in order to make data to be send |
|---|
| 92 | !-- by MPI contiguous |
|---|
| 93 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 94 | !$OMP DO |
|---|
| 95 | DO i = 0, nx |
|---|
| 96 | DO k = nzb_x, nzt_x |
|---|
| 97 | DO j = nys_x, nyn_x |
|---|
| 98 | f_inv(j,k,i) = f_in(i,j,k) |
|---|
| 99 | ENDDO |
|---|
| 100 | ENDDO |
|---|
| 101 | ENDDO |
|---|
| 102 | !$OMP END PARALLEL |
|---|
| 103 | |
|---|
| 104 | IF ( numprocs /= 1 ) THEN |
|---|
| 105 | |
|---|
| 106 | #if defined( __parallel ) |
|---|
| 107 | ! |
|---|
| 108 | !-- Transpose array |
|---|
| 109 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
|---|
| 110 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| 111 | CALL MPI_ALLTOALL( f_inv(nys_x,nzb_x,0), sendrecvcount_xy, MPI_REAL, & |
|---|
| 112 | work(1), sendrecvcount_xy, MPI_REAL, & |
|---|
| 113 | comm1dy, ierr ) |
|---|
| 114 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
|---|
| 115 | |
|---|
| 116 | ! |
|---|
| 117 | !-- Reorder transposed array |
|---|
| 118 | !$OMP PARALLEL PRIVATE ( i, j, k, l, m, ys ) |
|---|
| 119 | !$OMP DO |
|---|
| 120 | DO l = 0, pdims(2) - 1 |
|---|
| 121 | m = l * ( nxr_y - nxl_y + 1 ) * ( nzt_y - nzb_y + 1 ) * & |
|---|
| 122 | ( nyn_x - nys_x + 1 ) |
|---|
| 123 | ys = 0 + l * ( nyn_x - nys_x + 1 ) |
|---|
| 124 | DO i = nxl_y, nxr_y |
|---|
| 125 | DO k = nzb_y, nzt_y |
|---|
| 126 | DO j = ys, ys + nyn_x - nys_x |
|---|
| 127 | m = m + 1 |
|---|
| 128 | f_out(j,i,k) = work(m) |
|---|
| 129 | ENDDO |
|---|
| 130 | ENDDO |
|---|
| 131 | ENDDO |
|---|
| 132 | ENDDO |
|---|
| 133 | !$OMP END PARALLEL |
|---|
| 134 | #endif |
|---|
| 135 | |
|---|
| 136 | ELSE |
|---|
| 137 | |
|---|
| 138 | ! |
|---|
| 139 | !-- Reorder transposed array |
|---|
| 140 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 141 | !$OMP DO |
|---|
| 142 | DO k = nzb_y, nzt_y |
|---|
| 143 | DO i = nxl_y, nxr_y |
|---|
| 144 | DO j = 0, ny |
|---|
| 145 | f_out(j,i,k) = f_inv(j,k,i) |
|---|
| 146 | ENDDO |
|---|
| 147 | ENDDO |
|---|
| 148 | ENDDO |
|---|
| 149 | !$OMP END PARALLEL |
|---|
| 150 | |
|---|
| 151 | ENDIF |
|---|
| 152 | |
|---|
| 153 | END SUBROUTINE transpose_xy |
|---|
| 154 | |
|---|
| 155 | |
|---|
| 156 | SUBROUTINE transpose_xz( f_in, work, f_out ) |
|---|
| 157 | |
|---|
| 158 | !------------------------------------------------------------------------------! |
|---|
| 159 | ! Description: |
|---|
| 160 | ! ------------ |
|---|
| 161 | ! Transposition of input array (f_in) from x to z. For the input array, all |
|---|
| 162 | ! elements along x reside on the same PE, while after transposition, all |
|---|
| 163 | ! elements along z reside on the same PE. |
|---|
| 164 | !------------------------------------------------------------------------------! |
|---|
| 165 | |
|---|
| 166 | USE cpulog |
|---|
| 167 | USE indices |
|---|
| 168 | USE interfaces |
|---|
| 169 | USE pegrid |
|---|
| 170 | USE transpose_indices |
|---|
| 171 | |
|---|
| 172 | IMPLICIT NONE |
|---|
| 173 | |
|---|
| 174 | INTEGER :: i, j, k, l, m, xs |
|---|
| 175 | |
|---|
| 176 | REAL :: f_in(0:nx,nys_x:nyn_x,nzb_x:nzt_x), & |
|---|
| 177 | f_inv(nys:nyn,nxl:nxr,1:nz), & |
|---|
| 178 | f_out(1:nz,nys:nyn,nxl:nxr), & |
|---|
| 179 | work(nnx*nny*nnz) |
|---|
| 180 | |
|---|
| 181 | |
|---|
| 182 | ! |
|---|
| 183 | !-- If the PE grid is one-dimensional along y, the array has only to be |
|---|
| 184 | !-- reordered locally and therefore no transposition has to be done. |
|---|
| 185 | IF ( pdims(1) /= 1 ) THEN |
|---|
| 186 | |
|---|
| 187 | #if defined( __parallel ) |
|---|
| 188 | ! |
|---|
| 189 | !-- Reorder input array for transposition |
|---|
| 190 | !$OMP PARALLEL PRIVATE ( i, j, k, l, m, xs ) |
|---|
| 191 | !$OMP DO |
|---|
| 192 | DO l = 0, pdims(1) - 1 |
|---|
| 193 | m = l * ( nzt_x - nzb_x + 1 ) * nnx * ( nyn_x - nys_x + 1 ) |
|---|
| 194 | xs = 0 + l * nnx |
|---|
| 195 | DO k = nzb_x, nzt_x |
|---|
| 196 | DO i = xs, xs + nnx - 1 |
|---|
| 197 | DO j = nys_x, nyn_x |
|---|
| 198 | m = m + 1 |
|---|
| 199 | work(m) = f_in(i,j,k) |
|---|
| 200 | ENDDO |
|---|
| 201 | ENDDO |
|---|
| 202 | ENDDO |
|---|
| 203 | ENDDO |
|---|
| 204 | !$OMP END PARALLEL |
|---|
| 205 | |
|---|
| 206 | ! |
|---|
| 207 | !-- Transpose array |
|---|
| 208 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
|---|
| 209 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| 210 | CALL MPI_ALLTOALL( work(1), sendrecvcount_zx, MPI_REAL, & |
|---|
| 211 | f_inv(nys,nxl,1), sendrecvcount_zx, MPI_REAL, & |
|---|
| 212 | comm1dx, ierr ) |
|---|
| 213 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
|---|
| 214 | |
|---|
| 215 | ! |
|---|
| 216 | !-- Reorder transposed array in a way that the z index is in first position |
|---|
| 217 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 218 | !$OMP DO |
|---|
| 219 | DO k = 1, nz |
|---|
| 220 | DO i = nxl, nxr |
|---|
| 221 | DO j = nys, nyn |
|---|
| 222 | f_out(k,j,i) = f_inv(j,i,k) |
|---|
| 223 | ENDDO |
|---|
| 224 | ENDDO |
|---|
| 225 | ENDDO |
|---|
| 226 | !$OMP END PARALLEL |
|---|
| 227 | #endif |
|---|
| 228 | |
|---|
| 229 | ELSE |
|---|
| 230 | |
|---|
| 231 | ! |
|---|
| 232 | !-- Reorder the array in a way that the z index is in first position |
|---|
| 233 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 234 | !$OMP DO |
|---|
| 235 | DO i = nxl, nxr |
|---|
| 236 | DO j = nys, nyn |
|---|
| 237 | DO k = 1, nz |
|---|
| 238 | f_inv(j,i,k) = f_in(i,j,k) |
|---|
| 239 | ENDDO |
|---|
| 240 | ENDDO |
|---|
| 241 | ENDDO |
|---|
| 242 | !$OMP END PARALLEL |
|---|
| 243 | |
|---|
| 244 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 245 | !$OMP DO |
|---|
| 246 | DO k = 1, nz |
|---|
| 247 | DO i = nxl, nxr |
|---|
| 248 | DO j = nys, nyn |
|---|
| 249 | f_out(k,j,i) = f_inv(j,i,k) |
|---|
| 250 | ENDDO |
|---|
| 251 | ENDDO |
|---|
| 252 | ENDDO |
|---|
| 253 | !$OMP END PARALLEL |
|---|
| 254 | |
|---|
| 255 | ENDIF |
|---|
| 256 | |
|---|
| 257 | END SUBROUTINE transpose_xz |
|---|
| 258 | |
|---|
| 259 | |
|---|
| 260 | SUBROUTINE transpose_yx( f_in, work, f_out ) |
|---|
| 261 | |
|---|
| 262 | !------------------------------------------------------------------------------! |
|---|
| 263 | ! Description: |
|---|
| 264 | ! ------------ |
|---|
| 265 | ! Transposition of input array (f_in) from y to x. For the input array, all |
|---|
| 266 | ! elements along y reside on the same PE, while after transposition, all |
|---|
| 267 | ! elements along x reside on the same PE. |
|---|
| 268 | !------------------------------------------------------------------------------! |
|---|
| 269 | |
|---|
| 270 | USE cpulog |
|---|
| 271 | USE indices |
|---|
| 272 | USE interfaces |
|---|
| 273 | USE pegrid |
|---|
| 274 | USE transpose_indices |
|---|
| 275 | |
|---|
| 276 | IMPLICIT NONE |
|---|
| 277 | |
|---|
| 278 | INTEGER :: i, j, k, l, m, ys |
|---|
| 279 | |
|---|
| 280 | REAL :: f_in(0:ny,nxl_y:nxr_y,nzb_y:nzt_y), & |
|---|
| 281 | f_inv(nys_x:nyn_x,nzb_x:nzt_x,0:nx), & |
|---|
| 282 | f_out(0:nx,nys_x:nyn_x,nzb_x:nzt_x), & |
|---|
| 283 | work(nnx*nny*nnz) |
|---|
| 284 | |
|---|
| 285 | IF ( numprocs /= 1 ) THEN |
|---|
| 286 | |
|---|
| 287 | #if defined( __parallel ) |
|---|
| 288 | ! |
|---|
| 289 | !-- Reorder input array for transposition |
|---|
| 290 | !$OMP PARALLEL PRIVATE ( i, j, k, l, m, ys ) |
|---|
| 291 | !$OMP DO |
|---|
| 292 | DO l = 0, pdims(2) - 1 |
|---|
| 293 | m = l * ( nxr_y - nxl_y + 1 ) * ( nzt_y - nzb_y + 1 ) * & |
|---|
| 294 | ( nyn_x - nys_x + 1 ) |
|---|
| 295 | ys = 0 + l * ( nyn_x - nys_x + 1 ) |
|---|
| 296 | DO i = nxl_y, nxr_y |
|---|
| 297 | DO k = nzb_y, nzt_y |
|---|
| 298 | DO j = ys, ys + nyn_x - nys_x |
|---|
| 299 | m = m + 1 |
|---|
| 300 | work(m) = f_in(j,i,k) |
|---|
| 301 | ENDDO |
|---|
| 302 | ENDDO |
|---|
| 303 | ENDDO |
|---|
| 304 | ENDDO |
|---|
| 305 | !$OMP END PARALLEL |
|---|
| 306 | |
|---|
| 307 | ! |
|---|
| 308 | !-- Transpose array |
|---|
| 309 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
|---|
| 310 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| 311 | CALL MPI_ALLTOALL( work(1), sendrecvcount_xy, MPI_REAL, & |
|---|
| 312 | f_inv(nys_x,nzb_x,0), sendrecvcount_xy, MPI_REAL, & |
|---|
| 313 | comm1dy, ierr ) |
|---|
| 314 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
|---|
| 315 | #endif |
|---|
| 316 | |
|---|
| 317 | ELSE |
|---|
| 318 | |
|---|
| 319 | ! |
|---|
| 320 | !-- Reorder array f_in the same way as ALLTOALL did it |
|---|
| 321 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 322 | !$OMP DO |
|---|
| 323 | DO i = nxl_y, nxr_y |
|---|
| 324 | DO k = nzb_y, nzt_y |
|---|
| 325 | DO j = 0, ny |
|---|
| 326 | f_inv(j,k,i) = f_in(j,i,k) |
|---|
| 327 | ENDDO |
|---|
| 328 | ENDDO |
|---|
| 329 | ENDDO |
|---|
| 330 | !$OMP END PARALLEL |
|---|
| 331 | |
|---|
| 332 | ENDIF |
|---|
| 333 | |
|---|
| 334 | ! |
|---|
| 335 | !-- Reorder transposed array in a way that the x index is in first position |
|---|
| 336 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 337 | !$OMP DO |
|---|
| 338 | DO i = 0, nx |
|---|
| 339 | DO k = nzb_x, nzt_x |
|---|
| 340 | DO j = nys_x, nyn_x |
|---|
| 341 | f_out(i,j,k) = f_inv(j,k,i) |
|---|
| 342 | ENDDO |
|---|
| 343 | ENDDO |
|---|
| 344 | ENDDO |
|---|
| 345 | !$OMP END PARALLEL |
|---|
| 346 | |
|---|
| 347 | END SUBROUTINE transpose_yx |
|---|
| 348 | |
|---|
| 349 | |
|---|
| 350 | SUBROUTINE transpose_yxd( f_in, work, f_out ) |
|---|
| 351 | |
|---|
| 352 | !------------------------------------------------------------------------------! |
|---|
| 353 | ! Description: |
|---|
| 354 | ! ------------ |
|---|
| 355 | ! Transposition of input array (f_in) from y to x. For the input array, all |
|---|
| 356 | ! elements along y reside on the same PE, while after transposition, all |
|---|
| 357 | ! elements along x reside on the same PE. |
|---|
| 358 | ! This is a direct transposition for arrays with indices in regular order |
|---|
| 359 | ! (k,j,i) (cf. transpose_yx). |
|---|
| 360 | !------------------------------------------------------------------------------! |
|---|
| 361 | |
|---|
| 362 | USE cpulog |
|---|
| 363 | USE indices |
|---|
| 364 | USE interfaces |
|---|
| 365 | USE pegrid |
|---|
| 366 | USE transpose_indices |
|---|
| 367 | |
|---|
| 368 | IMPLICIT NONE |
|---|
| 369 | |
|---|
| 370 | INTEGER :: i, j, k, l, m, xs |
|---|
| 371 | |
|---|
| 372 | REAL :: f_in(1:nz,nys:nyn,nxl:nxr), f_inv(nxl:nxr,1:nz,nys:nyn), & |
|---|
| 373 | f_out(0:nx,nys_x:nyn_x,nzb_x:nzt_x), & |
|---|
| 374 | work(nnx*nny*nnz) |
|---|
| 375 | |
|---|
| 376 | #if defined( __parallel ) |
|---|
| 377 | |
|---|
| 378 | ! |
|---|
| 379 | !-- Rearrange indices of input array in order to make data to be send |
|---|
| 380 | !-- by MPI contiguous |
|---|
| 381 | DO k = 1, nz |
|---|
| 382 | DO j = nys, nyn |
|---|
| 383 | DO i = nxl, nxr |
|---|
| 384 | f_inv(i,k,j) = f_in(k,j,i) |
|---|
| 385 | ENDDO |
|---|
| 386 | ENDDO |
|---|
| 387 | ENDDO |
|---|
| 388 | |
|---|
| 389 | ! |
|---|
| 390 | !-- Transpose array |
|---|
| 391 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
|---|
| 392 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| 393 | CALL MPI_ALLTOALL( f_inv(nxl,1,nys), sendrecvcount_xy, MPI_REAL, & |
|---|
| 394 | work(1), sendrecvcount_xy, MPI_REAL, & |
|---|
| 395 | comm1dx, ierr ) |
|---|
| 396 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
|---|
| 397 | |
|---|
| 398 | ! |
|---|
| 399 | !-- Reorder transposed array |
|---|
| 400 | m = 0 |
|---|
| 401 | DO l = 0, pdims(1) - 1 |
|---|
| 402 | xs = 0 + l * nnx |
|---|
| 403 | DO j = nys_x, nyn_x |
|---|
| 404 | DO k = 1, nz |
|---|
| 405 | DO i = xs, xs + nnx - 1 |
|---|
| 406 | m = m + 1 |
|---|
| 407 | f_out(i,j,k) = work(m) |
|---|
| 408 | ENDDO |
|---|
| 409 | ENDDO |
|---|
| 410 | ENDDO |
|---|
| 411 | ENDDO |
|---|
| 412 | |
|---|
| 413 | #endif |
|---|
| 414 | |
|---|
| 415 | END SUBROUTINE transpose_yxd |
|---|
| 416 | |
|---|
| 417 | |
|---|
| 418 | SUBROUTINE transpose_yz( f_in, work, f_out ) |
|---|
| 419 | |
|---|
| 420 | !------------------------------------------------------------------------------! |
|---|
| 421 | ! Description: |
|---|
| 422 | ! ------------ |
|---|
| 423 | ! Transposition of input array (f_in) from y to z. For the input array, all |
|---|
| 424 | ! elements along y reside on the same PE, while after transposition, all |
|---|
| 425 | ! elements along z reside on the same PE. |
|---|
| 426 | !------------------------------------------------------------------------------! |
|---|
| 427 | |
|---|
| 428 | USE cpulog |
|---|
| 429 | USE indices |
|---|
| 430 | USE interfaces |
|---|
| 431 | USE pegrid |
|---|
| 432 | USE transpose_indices |
|---|
| 433 | |
|---|
| 434 | IMPLICIT NONE |
|---|
| 435 | |
|---|
| 436 | INTEGER :: i, j, k, l, m, zs |
|---|
| 437 | |
|---|
| 438 | REAL :: f_in(0:ny,nxl_y:nxr_y,nzb_y:nzt_y), & |
|---|
| 439 | f_inv(nxl_y:nxr_y,nzb_y:nzt_y,0:ny), & |
|---|
| 440 | f_out(nxl_z:nxr_z,nys_z:nyn_z,1:nz), & |
|---|
| 441 | work(nnx*nny*nnz) |
|---|
| 442 | |
|---|
| 443 | ! |
|---|
| 444 | !-- Rearrange indices of input array in order to make data to be send |
|---|
| 445 | !-- by MPI contiguous |
|---|
| 446 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 447 | !$OMP DO |
|---|
| 448 | DO j = 0, ny |
|---|
| 449 | DO k = nzb_y, nzt_y |
|---|
| 450 | DO i = nxl_y, nxr_y |
|---|
| 451 | f_inv(i,k,j) = f_in(j,i,k) |
|---|
| 452 | ENDDO |
|---|
| 453 | ENDDO |
|---|
| 454 | ENDDO |
|---|
| 455 | !$OMP END PARALLEL |
|---|
| 456 | |
|---|
| 457 | ! |
|---|
| 458 | !-- Move data to different array, because memory location of work1 is |
|---|
| 459 | !-- needed further below (work1 = work2). |
|---|
| 460 | !-- If the PE grid is one-dimensional along y, only local reordering |
|---|
| 461 | !-- of the data is necessary and no transposition has to be done. |
|---|
| 462 | IF ( pdims(1) == 1 ) THEN |
|---|
| 463 | |
|---|
| 464 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 465 | !$OMP DO |
|---|
| 466 | DO j = 0, ny |
|---|
| 467 | DO k = nzb_y, nzt_y |
|---|
| 468 | DO i = nxl_y, nxr_y |
|---|
| 469 | f_out(i,j,k) = f_inv(i,k,j) |
|---|
| 470 | ENDDO |
|---|
| 471 | ENDDO |
|---|
| 472 | ENDDO |
|---|
| 473 | !$OMP END PARALLEL |
|---|
| 474 | |
|---|
| 475 | ELSE |
|---|
| 476 | |
|---|
| 477 | #if defined( __parallel ) |
|---|
| 478 | ! |
|---|
| 479 | !-- Transpose array |
|---|
| 480 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
|---|
| 481 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| 482 | CALL MPI_ALLTOALL( f_inv(nxl_y,nzb_y,0), sendrecvcount_yz, MPI_REAL, & |
|---|
| 483 | work(1), sendrecvcount_yz, MPI_REAL, & |
|---|
| 484 | comm1dx, ierr ) |
|---|
| 485 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
|---|
| 486 | |
|---|
| 487 | ! |
|---|
| 488 | !-- Reorder transposed array |
|---|
| 489 | !$OMP PARALLEL PRIVATE ( i, j, k, l, m, zs ) |
|---|
| 490 | !$OMP DO |
|---|
| 491 | DO l = 0, pdims(1) - 1 |
|---|
| 492 | m = l * ( nyn_z - nys_z + 1 ) * ( nzt_y - nzb_y + 1 ) * & |
|---|
| 493 | ( nxr_z - nxl_z + 1 ) |
|---|
| 494 | zs = 1 + l * ( nzt_y - nzb_y + 1 ) |
|---|
| 495 | DO j = nys_z, nyn_z |
|---|
| 496 | DO k = zs, zs + nzt_y - nzb_y |
|---|
| 497 | DO i = nxl_z, nxr_z |
|---|
| 498 | m = m + 1 |
|---|
| 499 | f_out(i,j,k) = work(m) |
|---|
| 500 | ENDDO |
|---|
| 501 | ENDDO |
|---|
| 502 | ENDDO |
|---|
| 503 | ENDDO |
|---|
| 504 | !$OMP END PARALLEL |
|---|
| 505 | #endif |
|---|
| 506 | |
|---|
| 507 | ENDIF |
|---|
| 508 | |
|---|
| 509 | END SUBROUTINE transpose_yz |
|---|
| 510 | |
|---|
| 511 | |
|---|
| 512 | SUBROUTINE transpose_zx( f_in, work, f_out ) |
|---|
| 513 | |
|---|
| 514 | !------------------------------------------------------------------------------! |
|---|
| 515 | ! Description: |
|---|
| 516 | ! ------------ |
|---|
| 517 | ! Transposition of input array (f_in) from z to x. For the input array, all |
|---|
| 518 | ! elements along z reside on the same PE, while after transposition, all |
|---|
| 519 | ! elements along x reside on the same PE. |
|---|
| 520 | !------------------------------------------------------------------------------! |
|---|
| 521 | |
|---|
| 522 | USE cpulog |
|---|
| 523 | USE indices |
|---|
| 524 | USE interfaces |
|---|
| 525 | USE pegrid |
|---|
| 526 | USE transpose_indices |
|---|
| 527 | |
|---|
| 528 | IMPLICIT NONE |
|---|
| 529 | |
|---|
| 530 | INTEGER :: i, j, k, l, m, xs |
|---|
| 531 | |
|---|
| 532 | REAL :: f_in(1:nz,nys:nyn,nxl:nxr), f_out(0:nx,nys_x:nyn_x,nzb_x:nzt_x), & |
|---|
| 533 | work(nnx*nny*nnz) |
|---|
| 534 | |
|---|
| 535 | !$acc declare create ( f_inv ) |
|---|
| 536 | REAL :: f_inv(nys:nyn,nxl:nxr,1:nz) |
|---|
| 537 | |
|---|
| 538 | |
|---|
| 539 | ! |
|---|
| 540 | !-- Rearrange indices of input array in order to make data to be send |
|---|
| 541 | !-- by MPI contiguous |
|---|
| 542 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 543 | !$OMP DO |
|---|
| 544 | !$acc kernels present( f_in ) |
|---|
| 545 | !$acc loop |
|---|
| 546 | DO k = 1,nz |
|---|
| 547 | DO i = nxl, nxr |
|---|
| 548 | !$acc loop vector( 32 ) |
|---|
| 549 | DO j = nys, nyn |
|---|
| 550 | f_inv(j,i,k) = f_in(k,j,i) |
|---|
| 551 | ENDDO |
|---|
| 552 | ENDDO |
|---|
| 553 | ENDDO |
|---|
| 554 | !$OMP END PARALLEL |
|---|
| 555 | |
|---|
| 556 | ! |
|---|
| 557 | !-- Move data to different array, because memory location of work1 is |
|---|
| 558 | !-- needed further below (work1 = work2). |
|---|
| 559 | !-- If the PE grid is one-dimensional along y, only local reordering |
|---|
| 560 | !-- of the data is necessary and no transposition has to be done. |
|---|
| 561 | IF ( pdims(1) == 1 ) THEN |
|---|
| 562 | |
|---|
| 563 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 564 | !$OMP DO |
|---|
| 565 | !$acc kernels present( f_out ) |
|---|
| 566 | !$acc loop |
|---|
| 567 | DO k = 1, nz |
|---|
| 568 | DO i = nxl, nxr |
|---|
| 569 | !$acc loop vector( 32 ) |
|---|
| 570 | DO j = nys, nyn |
|---|
| 571 | f_out(i,j,k) = f_inv(j,i,k) |
|---|
| 572 | ENDDO |
|---|
| 573 | ENDDO |
|---|
| 574 | ENDDO |
|---|
| 575 | !$OMP END PARALLEL |
|---|
| 576 | |
|---|
| 577 | ELSE |
|---|
| 578 | |
|---|
| 579 | #if defined( __parallel ) |
|---|
| 580 | ! |
|---|
| 581 | !-- Transpose array |
|---|
| 582 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
|---|
| 583 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| 584 | CALL MPI_ALLTOALL( f_inv(nys,nxl,1), sendrecvcount_zx, MPI_REAL, & |
|---|
| 585 | work(1), sendrecvcount_zx, MPI_REAL, & |
|---|
| 586 | comm1dx, ierr ) |
|---|
| 587 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
|---|
| 588 | |
|---|
| 589 | ! |
|---|
| 590 | !-- Reorder transposed array |
|---|
| 591 | !$OMP PARALLEL PRIVATE ( i, j, k, l, m, xs ) |
|---|
| 592 | !$OMP DO |
|---|
| 593 | DO l = 0, pdims(1) - 1 |
|---|
| 594 | m = l * ( nzt_x - nzb_x + 1 ) * nnx * ( nyn_x - nys_x + 1 ) |
|---|
| 595 | xs = 0 + l * nnx |
|---|
| 596 | DO k = nzb_x, nzt_x |
|---|
| 597 | DO i = xs, xs + nnx - 1 |
|---|
| 598 | DO j = nys_x, nyn_x |
|---|
| 599 | m = m + 1 |
|---|
| 600 | f_out(i,j,k) = work(m) |
|---|
| 601 | ENDDO |
|---|
| 602 | ENDDO |
|---|
| 603 | ENDDO |
|---|
| 604 | ENDDO |
|---|
| 605 | !$OMP END PARALLEL |
|---|
| 606 | #endif |
|---|
| 607 | |
|---|
| 608 | ENDIF |
|---|
| 609 | |
|---|
| 610 | END SUBROUTINE transpose_zx |
|---|
| 611 | |
|---|
| 612 | |
|---|
| 613 | SUBROUTINE transpose_zy( f_in, work, f_out ) |
|---|
| 614 | |
|---|
| 615 | !------------------------------------------------------------------------------! |
|---|
| 616 | ! Description: |
|---|
| 617 | ! ------------ |
|---|
| 618 | ! Transposition of input array (f_in) from z to y. For the input array, all |
|---|
| 619 | ! elements along z reside on the same PE, while after transposition, all |
|---|
| 620 | ! elements along y reside on the same PE. |
|---|
| 621 | !------------------------------------------------------------------------------! |
|---|
| 622 | |
|---|
| 623 | USE cpulog |
|---|
| 624 | USE indices |
|---|
| 625 | USE interfaces |
|---|
| 626 | USE pegrid |
|---|
| 627 | USE transpose_indices |
|---|
| 628 | |
|---|
| 629 | IMPLICIT NONE |
|---|
| 630 | |
|---|
| 631 | INTEGER :: i, j, k, l, m, zs |
|---|
| 632 | |
|---|
| 633 | REAL :: f_in(nxl_z:nxr_z,nys_z:nyn_z,1:nz), & |
|---|
| 634 | f_inv(nxl_y:nxr_y,nzb_y:nzt_y,0:ny), & |
|---|
| 635 | f_out(0:ny,nxl_y:nxr_y,nzb_y:nzt_y), & |
|---|
| 636 | work(nnx*nny*nnz) |
|---|
| 637 | |
|---|
| 638 | ! |
|---|
| 639 | !-- If the PE grid is one-dimensional along y, the array has only to be |
|---|
| 640 | !-- reordered locally and therefore no transposition has to be done. |
|---|
| 641 | IF ( pdims(1) /= 1 ) THEN |
|---|
| 642 | |
|---|
| 643 | #if defined( __parallel ) |
|---|
| 644 | ! |
|---|
| 645 | !-- Reorder input array for transposition |
|---|
| 646 | !$OMP PARALLEL PRIVATE ( i, j, k, l, m, zs ) |
|---|
| 647 | !$OMP DO |
|---|
| 648 | DO l = 0, pdims(1) - 1 |
|---|
| 649 | m = l * ( nyn_z - nys_z + 1 ) * ( nzt_y - nzb_y + 1 ) * & |
|---|
| 650 | ( nxr_z - nxl_z + 1 ) |
|---|
| 651 | zs = 1 + l * ( nzt_y - nzb_y + 1 ) |
|---|
| 652 | DO j = nys_z, nyn_z |
|---|
| 653 | DO k = zs, zs + nzt_y - nzb_y |
|---|
| 654 | DO i = nxl_z, nxr_z |
|---|
| 655 | m = m + 1 |
|---|
| 656 | work(m) = f_in(i,j,k) |
|---|
| 657 | ENDDO |
|---|
| 658 | ENDDO |
|---|
| 659 | ENDDO |
|---|
| 660 | ENDDO |
|---|
| 661 | !$OMP END PARALLEL |
|---|
| 662 | |
|---|
| 663 | ! |
|---|
| 664 | !-- Transpose array |
|---|
| 665 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
|---|
| 666 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| 667 | CALL MPI_ALLTOALL( work(1), sendrecvcount_yz, MPI_REAL, & |
|---|
| 668 | f_inv(nxl_y,nzb_y,0), sendrecvcount_yz, MPI_REAL, & |
|---|
| 669 | comm1dx, ierr ) |
|---|
| 670 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
|---|
| 671 | #endif |
|---|
| 672 | |
|---|
| 673 | ELSE |
|---|
| 674 | ! |
|---|
| 675 | !-- Reorder the array in the same way like ALLTOALL did it |
|---|
| 676 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 677 | !$OMP DO |
|---|
| 678 | DO k = nzb_y, nzt_y |
|---|
| 679 | DO j = 0, ny |
|---|
| 680 | DO i = nxl_y, nxr_y |
|---|
| 681 | f_inv(i,k,j) = f_in(i,j,k) |
|---|
| 682 | ENDDO |
|---|
| 683 | ENDDO |
|---|
| 684 | ENDDO |
|---|
| 685 | !$OMP END PARALLEL |
|---|
| 686 | |
|---|
| 687 | ENDIF |
|---|
| 688 | |
|---|
| 689 | ! |
|---|
| 690 | !-- Reorder transposed array in a way that the y index is in first position |
|---|
| 691 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
|---|
| 692 | !$OMP DO |
|---|
| 693 | DO k = nzb_y, nzt_y |
|---|
| 694 | DO i = nxl_y, nxr_y |
|---|
| 695 | DO j = 0, ny |
|---|
| 696 | f_out(j,i,k) = f_inv(i,k,j) |
|---|
| 697 | ENDDO |
|---|
| 698 | ENDDO |
|---|
| 699 | ENDDO |
|---|
| 700 | !$OMP END PARALLEL |
|---|
| 701 | |
|---|
| 702 | END SUBROUTINE transpose_zy |
|---|
| 703 | |
|---|
| 704 | |
|---|
| 705 | SUBROUTINE transpose_zyd( f_in, work, f_out ) |
|---|
| 706 | |
|---|
| 707 | !------------------------------------------------------------------------------! |
|---|
| 708 | ! Description: |
|---|
| 709 | ! ------------ |
|---|
| 710 | ! Transposition of input array (f_in) from z to y. For the input array, all |
|---|
| 711 | ! elements along z reside on the same PE, while after transposition, all |
|---|
| 712 | ! elements along y reside on the same PE. |
|---|
| 713 | ! This is a direct transposition for arrays with indices in regular order |
|---|
| 714 | ! (k,j,i) (cf. transpose_zy). |
|---|
| 715 | !------------------------------------------------------------------------------! |
|---|
| 716 | |
|---|
| 717 | USE cpulog |
|---|
| 718 | USE indices |
|---|
| 719 | USE interfaces |
|---|
| 720 | USE pegrid |
|---|
| 721 | USE transpose_indices |
|---|
| 722 | |
|---|
| 723 | IMPLICIT NONE |
|---|
| 724 | |
|---|
| 725 | INTEGER :: i, j, k, l, m, ys |
|---|
| 726 | |
|---|
| 727 | REAL :: f_in(1:nz,nys:nyn,nxl:nxr), f_inv(nys:nyn,nxl:nxr,1:nz), & |
|---|
| 728 | f_out(0:ny,nxl_yd:nxr_yd,nzb_yd:nzt_yd), & |
|---|
| 729 | work(nnx*nny*nnz) |
|---|
| 730 | |
|---|
| 731 | #if defined( __parallel ) |
|---|
| 732 | |
|---|
| 733 | ! |
|---|
| 734 | !-- Rearrange indices of input array in order to make data to be send |
|---|
| 735 | !-- by MPI contiguous |
|---|
| 736 | DO i = nxl, nxr |
|---|
| 737 | DO j = nys, nyn |
|---|
| 738 | DO k = 1, nz |
|---|
| 739 | f_inv(j,i,k) = f_in(k,j,i) |
|---|
| 740 | ENDDO |
|---|
| 741 | ENDDO |
|---|
| 742 | ENDDO |
|---|
| 743 | |
|---|
| 744 | ! |
|---|
| 745 | !-- Move data to different array, because memory location of work1 is |
|---|
| 746 | !-- needed further below (work1 = work2). |
|---|
| 747 | !-- If the PE grid is one-dimensional along x, only local reordering |
|---|
| 748 | !-- of the data is necessary and no transposition has to be done. |
|---|
| 749 | IF ( pdims(2) == 1 ) THEN |
|---|
| 750 | DO k = 1, nz |
|---|
| 751 | DO i = nxl, nxr |
|---|
| 752 | DO j = nys, nyn |
|---|
| 753 | f_out(j,i,k) = f_inv(j,i,k) |
|---|
| 754 | ENDDO |
|---|
| 755 | ENDDO |
|---|
| 756 | ENDDO |
|---|
| 757 | RETURN |
|---|
| 758 | ENDIF |
|---|
| 759 | |
|---|
| 760 | ! |
|---|
| 761 | !-- Transpose array |
|---|
| 762 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
|---|
| 763 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| 764 | CALL MPI_ALLTOALL( f_inv(nys,nxl,1), sendrecvcount_zyd, MPI_REAL, & |
|---|
| 765 | work(1), sendrecvcount_zyd, MPI_REAL, & |
|---|
| 766 | comm1dy, ierr ) |
|---|
| 767 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
|---|
| 768 | |
|---|
| 769 | ! |
|---|
| 770 | !-- Reorder transposed array |
|---|
| 771 | m = 0 |
|---|
| 772 | DO l = 0, pdims(2) - 1 |
|---|
| 773 | ys = 0 + l * nny |
|---|
| 774 | DO k = nzb_yd, nzt_yd |
|---|
| 775 | DO i = nxl_yd, nxr_yd |
|---|
| 776 | DO j = ys, ys + nny - 1 |
|---|
| 777 | m = m + 1 |
|---|
| 778 | f_out(j,i,k) = work(m) |
|---|
| 779 | ENDDO |
|---|
| 780 | ENDDO |
|---|
| 781 | ENDDO |
|---|
| 782 | ENDDO |
|---|
| 783 | |
|---|
| 784 | #endif |
|---|
| 785 | |
|---|
| 786 | END SUBROUTINE transpose_zyd |
|---|
