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