| 1 | !> @file exchange_horiz.f90 |
|---|
| 2 | !------------------------------------------------------------------------------! |
|---|
| 3 | ! This file is part of the PALM model system. |
|---|
| 4 | ! |
|---|
| 5 | ! PALM is free software: you can redistribute it and/or modify it under the |
|---|
| 6 | ! terms of the GNU General Public License as published by the Free Software |
|---|
| 7 | ! Foundation, either version 3 of the License, or (at your option) any later |
|---|
| 8 | ! 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 1997-2018 Leibniz Universitaet Hannover |
|---|
| 18 | !------------------------------------------------------------------------------! |
|---|
| 19 | ! |
|---|
| 20 | ! Current revisions: |
|---|
| 21 | ! ----------------- |
|---|
| 22 | ! |
|---|
| 23 | ! |
|---|
| 24 | ! Former revisions: |
|---|
| 25 | ! ----------------- |
|---|
| 26 | ! $Id: exchange_horiz.f90 2718 2018-01-02 08:49:38Z Giersch $ |
|---|
| 27 | ! Corrected "Former revisions" section |
|---|
| 28 | ! |
|---|
| 29 | ! 2696 2017-12-14 17:12:51Z kanani |
|---|
| 30 | ! Change in file header (GPL part) |
|---|
| 31 | ! 3D-Integer exchange on multigrid level (MS) |
|---|
| 32 | ! |
|---|
| 33 | ! 2298 2017-06-29 09:28:18Z raasch |
|---|
| 34 | ! sendrecv_in_background related parts removed |
|---|
| 35 | ! |
|---|
| 36 | ! 2119 2017-01-17 16:51:50Z raasch |
|---|
| 37 | ! |
|---|
| 38 | ! 2118 2017-01-17 16:38:49Z raasch |
|---|
| 39 | ! OpenACC directives and related code removed |
|---|
| 40 | ! |
|---|
| 41 | ! 2000 2016-08-20 18:09:15Z knoop |
|---|
| 42 | ! Forced header and separation lines into 80 columns |
|---|
| 43 | ! |
|---|
| 44 | ! 1804 2016-04-05 16:30:18Z maronga |
|---|
| 45 | ! Removed code for parameter file check (__check) |
|---|
| 46 | ! |
|---|
| 47 | ! 1682 2015-10-07 23:56:08Z knoop |
|---|
| 48 | ! Code annotations made doxygen readable |
|---|
| 49 | ! |
|---|
| 50 | ! 1677 2015-10-02 13:25:23Z boeske |
|---|
| 51 | ! Added new routine for exchange of three-dimensional integer arrays |
|---|
| 52 | ! |
|---|
| 53 | ! 1569 2015-03-12 07:54:38Z raasch |
|---|
| 54 | ! bugfix in background communication part |
|---|
| 55 | ! |
|---|
| 56 | ! 1348 2014-03-27 18:01:03Z raasch |
|---|
| 57 | ! bugfix: on_device added to ONLY-list |
|---|
| 58 | ! |
|---|
| 59 | ! 1344 2014-03-26 17:33:09Z kanani |
|---|
| 60 | ! Added missing parameters to ONLY-attribute |
|---|
| 61 | ! |
|---|
| 62 | ! 1320 2014-03-20 08:40:49Z raasch |
|---|
| 63 | ! ONLY-attribute added to USE-statements, |
|---|
| 64 | ! kind-parameters added to all INTEGER and REAL declaration statements, |
|---|
| 65 | ! kinds are defined in new module kinds, |
|---|
| 66 | ! revision history before 2012 removed, |
|---|
| 67 | ! comment fields (!:) to be used for variable explanations added to |
|---|
| 68 | ! all variable declaration statements |
|---|
| 69 | ! |
|---|
| 70 | ! 1257 2013-11-08 15:18:40Z raasch |
|---|
| 71 | ! openacc loop and loop vector clauses removed |
|---|
| 72 | ! |
|---|
| 73 | ! 1128 2013-04-12 06:19:32Z raasch |
|---|
| 74 | ! modifications for asynchronous transfer, |
|---|
| 75 | ! local variables req, wait_stat are global now, and have been moved to module |
|---|
| 76 | ! pegrid |
|---|
| 77 | ! |
|---|
| 78 | ! 1113 2013-03-10 02:48:14Z raasch |
|---|
| 79 | ! GPU-porting for single-core (1PE) mode |
|---|
| 80 | ! |
|---|
| 81 | ! 1036 2012-10-22 13:43:42Z raasch |
|---|
| 82 | ! code put under GPL (PALM 3.9) |
|---|
| 83 | ! |
|---|
| 84 | ! 841 2012-02-28 12:29:49Z maronga |
|---|
| 85 | ! Excluded routine from compilation of namelist_file_check |
|---|
| 86 | ! |
|---|
| 87 | ! Revision 1.1 1997/07/24 11:13:29 raasch |
|---|
| 88 | ! Initial revision |
|---|
| 89 | ! |
|---|
| 90 | ! |
|---|
| 91 | ! Description: |
|---|
| 92 | ! ------------ |
|---|
| 93 | !> Exchange of lateral boundary values (parallel computers) and cyclic |
|---|
| 94 | !> lateral boundary conditions, respectively. |
|---|
| 95 | !------------------------------------------------------------------------------! |
|---|
| 96 | SUBROUTINE exchange_horiz( ar, nbgp_local) |
|---|
| 97 | |
|---|
| 98 | |
|---|
| 99 | USE control_parameters, & |
|---|
| 100 | ONLY: bc_lr, bc_lr_cyc, bc_ns, bc_ns_cyc, grid_level, & |
|---|
| 101 | mg_switch_to_pe0, synchronous_exchange |
|---|
| 102 | |
|---|
| 103 | USE cpulog, & |
|---|
| 104 | ONLY: cpu_log, log_point_s |
|---|
| 105 | |
|---|
| 106 | USE indices, & |
|---|
| 107 | ONLY: nxl, nxr, nyn, nys, nzb, nzt |
|---|
| 108 | |
|---|
| 109 | USE kinds |
|---|
| 110 | |
|---|
| 111 | USE pegrid |
|---|
| 112 | |
|---|
| 113 | IMPLICIT NONE |
|---|
| 114 | |
|---|
| 115 | |
|---|
| 116 | INTEGER(iwp) :: i !< |
|---|
| 117 | INTEGER(iwp) :: j !< |
|---|
| 118 | INTEGER(iwp) :: k !< |
|---|
| 119 | INTEGER(iwp) :: nbgp_local !< |
|---|
| 120 | |
|---|
| 121 | REAL(wp), DIMENSION(nzb:nzt+1,nys-nbgp_local:nyn+nbgp_local, & |
|---|
| 122 | nxl-nbgp_local:nxr+nbgp_local) :: ar !< |
|---|
| 123 | |
|---|
| 124 | |
|---|
| 125 | CALL cpu_log( log_point_s(2), 'exchange_horiz', 'start' ) |
|---|
| 126 | |
|---|
| 127 | #if defined( __parallel ) |
|---|
| 128 | |
|---|
| 129 | ! |
|---|
| 130 | !-- Exchange in x-direction of lateral boundaries |
|---|
| 131 | IF ( pdims(1) == 1 .OR. mg_switch_to_pe0 ) THEN |
|---|
| 132 | ! |
|---|
| 133 | !-- One-dimensional decomposition along y, boundary values can be exchanged |
|---|
| 134 | !-- within the PE memory |
|---|
| 135 | IF ( bc_lr_cyc ) THEN |
|---|
| 136 | ar(:,:,nxl-nbgp_local:nxl-1) = ar(:,:,nxr-nbgp_local+1:nxr) |
|---|
| 137 | ar(:,:,nxr+1:nxr+nbgp_local) = ar(:,:,nxl:nxl+nbgp_local-1) |
|---|
| 138 | ENDIF |
|---|
| 139 | |
|---|
| 140 | ELSE |
|---|
| 141 | |
|---|
| 142 | IF ( synchronous_exchange ) THEN |
|---|
| 143 | ! |
|---|
| 144 | !-- Send left boundary, receive right one (synchronous) |
|---|
| 145 | CALL MPI_SENDRECV( & |
|---|
| 146 | ar(nzb,nys-nbgp_local,nxl), 1, type_yz(grid_level), pleft, 0, & |
|---|
| 147 | ar(nzb,nys-nbgp_local,nxr+1), 1, type_yz(grid_level), pright, 0, & |
|---|
| 148 | comm2d, status, ierr ) |
|---|
| 149 | ! |
|---|
| 150 | !-- Send right boundary, receive left one (synchronous) |
|---|
| 151 | CALL MPI_SENDRECV( ar(nzb,nys-nbgp_local,nxr+1-nbgp_local), 1, & |
|---|
| 152 | type_yz(grid_level), pright, 1, & |
|---|
| 153 | ar(nzb,nys-nbgp_local,nxl-nbgp_local), 1, & |
|---|
| 154 | type_yz(grid_level), pleft, 1, & |
|---|
| 155 | comm2d, status, ierr ) |
|---|
| 156 | |
|---|
| 157 | ELSE |
|---|
| 158 | |
|---|
| 159 | ! |
|---|
| 160 | !-- Asynchroneous exchange |
|---|
| 161 | IF ( send_receive == 'lr' .OR. send_receive == 'al' ) THEN |
|---|
| 162 | |
|---|
| 163 | req(1:4) = 0 |
|---|
| 164 | req_count = 0 |
|---|
| 165 | ! |
|---|
| 166 | !-- Send left boundary, receive right one (asynchronous) |
|---|
| 167 | CALL MPI_ISEND( ar(nzb,nys-nbgp_local,nxl), 1, type_yz(grid_level), & |
|---|
| 168 | pleft, req_count, comm2d, req(req_count+1), ierr ) |
|---|
| 169 | CALL MPI_IRECV( ar(nzb,nys-nbgp_local,nxr+1), 1, type_yz(grid_level), & |
|---|
| 170 | pright, req_count, comm2d, req(req_count+2), ierr ) |
|---|
| 171 | ! |
|---|
| 172 | !-- Send right boundary, receive left one (asynchronous) |
|---|
| 173 | CALL MPI_ISEND( ar(nzb,nys-nbgp_local,nxr+1-nbgp_local), 1, & |
|---|
| 174 | type_yz(grid_level), pright, req_count+1, comm2d, & |
|---|
| 175 | req(req_count+3), ierr ) |
|---|
| 176 | CALL MPI_IRECV( ar(nzb,nys-nbgp_local,nxl-nbgp_local), 1, & |
|---|
| 177 | type_yz(grid_level), pleft, req_count+1, comm2d, & |
|---|
| 178 | req(req_count+4), ierr ) |
|---|
| 179 | |
|---|
| 180 | CALL MPI_WAITALL( 4, req, wait_stat, ierr ) |
|---|
| 181 | |
|---|
| 182 | ENDIF |
|---|
| 183 | |
|---|
| 184 | ENDIF |
|---|
| 185 | |
|---|
| 186 | ENDIF |
|---|
| 187 | |
|---|
| 188 | |
|---|
| 189 | IF ( pdims(2) == 1 .OR. mg_switch_to_pe0 ) THEN |
|---|
| 190 | ! |
|---|
| 191 | !-- One-dimensional decomposition along x, boundary values can be exchanged |
|---|
| 192 | !-- within the PE memory |
|---|
| 193 | IF ( bc_ns_cyc ) THEN |
|---|
| 194 | ar(:,nys-nbgp_local:nys-1,:) = ar(:,nyn-nbgp_local+1:nyn,:) |
|---|
| 195 | ar(:,nyn+1:nyn+nbgp_local,:) = ar(:,nys:nys+nbgp_local-1,:) |
|---|
| 196 | ENDIF |
|---|
| 197 | |
|---|
| 198 | ELSE |
|---|
| 199 | |
|---|
| 200 | IF ( synchronous_exchange ) THEN |
|---|
| 201 | ! |
|---|
| 202 | !-- Send front boundary, receive rear one (synchronous) |
|---|
| 203 | CALL MPI_SENDRECV( & |
|---|
| 204 | ar(nzb,nys,nxl-nbgp_local), 1, type_xz(grid_level), psouth, 0, & |
|---|
| 205 | ar(nzb,nyn+1,nxl-nbgp_local), 1, type_xz(grid_level), pnorth, 0, & |
|---|
| 206 | comm2d, status, ierr ) |
|---|
| 207 | ! |
|---|
| 208 | !-- Send rear boundary, receive front one (synchronous) |
|---|
| 209 | CALL MPI_SENDRECV( ar(nzb,nyn-nbgp_local+1,nxl-nbgp_local), 1, & |
|---|
| 210 | type_xz(grid_level), pnorth, 1, & |
|---|
| 211 | ar(nzb,nys-nbgp_local,nxl-nbgp_local), 1, & |
|---|
| 212 | type_xz(grid_level), psouth, 1, & |
|---|
| 213 | comm2d, status, ierr ) |
|---|
| 214 | |
|---|
| 215 | ELSE |
|---|
| 216 | |
|---|
| 217 | ! |
|---|
| 218 | !-- Asynchroneous exchange |
|---|
| 219 | IF ( send_receive == 'ns' .OR. send_receive == 'al' ) THEN |
|---|
| 220 | |
|---|
| 221 | req(1:4) = 0 |
|---|
| 222 | req_count = 0 |
|---|
| 223 | |
|---|
| 224 | ! |
|---|
| 225 | !-- Send front boundary, receive rear one (asynchronous) |
|---|
| 226 | CALL MPI_ISEND( ar(nzb,nys,nxl-nbgp_local), 1, type_xz(grid_level), & |
|---|
| 227 | psouth, req_count, comm2d, req(req_count+1), ierr ) |
|---|
| 228 | CALL MPI_IRECV( ar(nzb,nyn+1,nxl-nbgp_local), 1, type_xz(grid_level), & |
|---|
| 229 | pnorth, req_count, comm2d, req(req_count+2), ierr ) |
|---|
| 230 | ! |
|---|
| 231 | !-- Send rear boundary, receive front one (asynchronous) |
|---|
| 232 | CALL MPI_ISEND( ar(nzb,nyn-nbgp_local+1,nxl-nbgp_local), 1, & |
|---|
| 233 | type_xz(grid_level), pnorth, req_count+1, comm2d, & |
|---|
| 234 | req(req_count+3), ierr ) |
|---|
| 235 | CALL MPI_IRECV( ar(nzb,nys-nbgp_local,nxl-nbgp_local), 1, & |
|---|
| 236 | type_xz(grid_level), psouth, req_count+1, comm2d, & |
|---|
| 237 | req(req_count+4), ierr ) |
|---|
| 238 | |
|---|
| 239 | CALL MPI_WAITALL( 4, req, wait_stat, ierr ) |
|---|
| 240 | |
|---|
| 241 | ENDIF |
|---|
| 242 | |
|---|
| 243 | ENDIF |
|---|
| 244 | |
|---|
| 245 | ENDIF |
|---|
| 246 | |
|---|
| 247 | #else |
|---|
| 248 | |
|---|
| 249 | ! |
|---|
| 250 | !-- Lateral boundary conditions in the non-parallel case. |
|---|
| 251 | !-- Case dependent, because in GPU mode still not all arrays are on device. This |
|---|
| 252 | !-- workaround has to be removed later. Also, since PGI compiler 12.5 has problems |
|---|
| 253 | !-- with array syntax, explicit loops are used. |
|---|
| 254 | IF ( bc_lr == 'cyclic' ) THEN |
|---|
| 255 | ar(:,:,nxl-nbgp_local:nxl-1) = ar(:,:,nxr-nbgp_local+1:nxr) |
|---|
| 256 | ar(:,:,nxr+1:nxr+nbgp_local) = ar(:,:,nxl:nxl+nbgp_local-1) |
|---|
| 257 | ENDIF |
|---|
| 258 | |
|---|
| 259 | IF ( bc_ns == 'cyclic' ) THEN |
|---|
| 260 | ar(:,nys-nbgp_local:nys-1,:) = ar(:,nyn-nbgp_local+1:nyn,:) |
|---|
| 261 | ar(:,nyn+1:nyn+nbgp_local,:) = ar(:,nys:nys+nbgp_local-1,:) |
|---|
| 262 | ENDIF |
|---|
| 263 | |
|---|
| 264 | #endif |
|---|
| 265 | CALL cpu_log( log_point_s(2), 'exchange_horiz', 'stop' ) |
|---|
| 266 | |
|---|
| 267 | END SUBROUTINE exchange_horiz |
|---|
| 268 | |
|---|
| 269 | |
|---|
| 270 | !------------------------------------------------------------------------------! |
|---|
| 271 | ! Description: |
|---|
| 272 | ! ------------ |
|---|
| 273 | !> @todo Missing subroutine description. |
|---|
| 274 | !------------------------------------------------------------------------------! |
|---|
| 275 | SUBROUTINE exchange_horiz_int( ar, nys_l, nyn_l, nxl_l, nxr_l, nzt_l, nbgp_local) |
|---|
| 276 | |
|---|
| 277 | USE control_parameters, & |
|---|
| 278 | ONLY: bc_lr, bc_lr_cyc, bc_ns, bc_ns_cyc, grid_level |
|---|
| 279 | |
|---|
| 280 | USE indices, & |
|---|
| 281 | ONLY: nxl, nxr, nyn, nys, nzb, nzt |
|---|
| 282 | |
|---|
| 283 | USE kinds |
|---|
| 284 | |
|---|
| 285 | USE pegrid |
|---|
| 286 | |
|---|
| 287 | IMPLICIT NONE |
|---|
| 288 | |
|---|
| 289 | INTEGER(iwp) :: nxl_l !< local index bound at current grid level, left side |
|---|
| 290 | INTEGER(iwp) :: nxr_l !< local index bound at current grid level, right side |
|---|
| 291 | INTEGER(iwp) :: nyn_l !< local index bound at current grid level, north side |
|---|
| 292 | INTEGER(iwp) :: nys_l !< local index bound at current grid level, south side |
|---|
| 293 | INTEGER(iwp) :: nzt_l !< local index bound at current grid level, top |
|---|
| 294 | INTEGER(iwp) :: nbgp_local !< number of ghost points |
|---|
| 295 | |
|---|
| 296 | INTEGER(iwp), DIMENSION(nzb:nzt_l+1,nys_l-nbgp_local:nyn_l+nbgp_local, & |
|---|
| 297 | nxl_l-nbgp_local:nxr_l+nbgp_local) :: ar !< treated array |
|---|
| 298 | |
|---|
| 299 | |
|---|
| 300 | #if defined( __parallel ) |
|---|
| 301 | IF ( pdims(1) == 1 ) THEN |
|---|
| 302 | ! |
|---|
| 303 | !-- One-dimensional decomposition along y, boundary values can be exchanged |
|---|
| 304 | !-- within the PE memory |
|---|
| 305 | IF ( bc_lr_cyc ) THEN |
|---|
| 306 | ar(:,:,nxl_l-nbgp_local:nxl_l-1) = ar(:,:,nxr_l-nbgp_local+1:nxr_l) |
|---|
| 307 | ar(:,:,nxr_l+1:nxr_l+nbgp_local) = ar(:,:,nxl_l:nxl_l+nbgp_local-1) |
|---|
| 308 | ENDIF |
|---|
| 309 | ELSE |
|---|
| 310 | ! |
|---|
| 311 | !-- Send left boundary, receive right one (synchronous) |
|---|
| 312 | CALL MPI_SENDRECV( & |
|---|
| 313 | ar(nzb,nys_l-nbgp_local,nxl_l), 1, type_yz_int(grid_level), pleft, 0,& |
|---|
| 314 | ar(nzb,nys_l-nbgp_local,nxr_l+1), 1, type_yz_int(grid_level), pright, 0,& |
|---|
| 315 | comm2d, status, ierr ) |
|---|
| 316 | ! |
|---|
| 317 | !-- Send right boundary, receive left one (synchronous) |
|---|
| 318 | CALL MPI_SENDRECV( & |
|---|
| 319 | ar(nzb,nys_l-nbgp_local,nxr_l+1-nbgp_local), 1, type_yz_int(grid_level),& |
|---|
| 320 | pright, 1, & |
|---|
| 321 | ar(nzb,nys_l-nbgp_local,nxl_l-nbgp_local), 1, type_yz_int(grid_level),& |
|---|
| 322 | pleft, 1, & |
|---|
| 323 | comm2d, status, ierr ) |
|---|
| 324 | ENDIF |
|---|
| 325 | |
|---|
| 326 | |
|---|
| 327 | IF ( pdims(2) == 1 ) THEN |
|---|
| 328 | ! |
|---|
| 329 | !-- One-dimensional decomposition along x, boundary values can be exchanged |
|---|
| 330 | !-- within the PE memory |
|---|
| 331 | IF ( bc_ns_cyc ) THEN |
|---|
| 332 | ar(:,nys_l-nbgp_local:nys_l-1,:) = ar(:,nyn_l-nbgp_local+1:nyn_l,:) |
|---|
| 333 | ar(:,nyn_l+1:nyn_l+nbgp_local,:) = ar(:,nys_l:nys_l+nbgp_local-1,:) |
|---|
| 334 | ENDIF |
|---|
| 335 | |
|---|
| 336 | ELSE |
|---|
| 337 | |
|---|
| 338 | ! |
|---|
| 339 | !-- Send front boundary, receive rear one (synchronous) |
|---|
| 340 | CALL MPI_SENDRECV( & |
|---|
| 341 | ar(nzb,nys_l,nxl_l-nbgp_local), 1, type_xz_int(grid_level), psouth, 0,& |
|---|
| 342 | ar(nzb,nyn_l+1,nxl_l-nbgp_local), 1, type_xz_int(grid_level), pnorth, 0,& |
|---|
| 343 | comm2d, status, ierr ) |
|---|
| 344 | ! |
|---|
| 345 | !-- Send rear boundary, receive front one (synchronous) |
|---|
| 346 | CALL MPI_SENDRECV( ar(nzb,nyn_l-nbgp_local+1,nxl_l-nbgp_local), 1, & |
|---|
| 347 | type_xz_int(grid_level), pnorth, 1, & |
|---|
| 348 | ar(nzb,nys_l-nbgp_local,nxl_l-nbgp_local), 1, & |
|---|
| 349 | type_xz_int(grid_level), psouth, 1, & |
|---|
| 350 | comm2d, status, ierr ) |
|---|
| 351 | |
|---|
| 352 | ENDIF |
|---|
| 353 | |
|---|
| 354 | #else |
|---|
| 355 | |
|---|
| 356 | IF ( bc_lr == 'cyclic' ) THEN |
|---|
| 357 | ar(:,:,nxl_l-nbgp_local:nxl_l-1) = ar(:,:,nxr_l-nbgp_local+1:nxr_l) |
|---|
| 358 | ar(:,:,nxr_l+1:nxr_l+nbgp_local) = ar(:,:,nxl_l:nxl_l+nbgp_local-1) |
|---|
| 359 | ENDIF |
|---|
| 360 | |
|---|
| 361 | IF ( bc_ns == 'cyclic' ) THEN |
|---|
| 362 | ar(:,nys_l-nbgp_local:nys_l-1,:) = ar(:,nyn_l-nbgp_local+1:nyn_l,:) |
|---|
| 363 | ar(:,nyn_l+1:nyn_l+nbgp_local,:) = ar(:,nys_l:nys_l+nbgp_local-1,:) |
|---|
| 364 | ENDIF |
|---|
| 365 | |
|---|
| 366 | #endif |
|---|
| 367 | |
|---|
| 368 | |
|---|
| 369 | END SUBROUTINE exchange_horiz_int |
|---|
