1 | SUBROUTINE poismg( r ) |
---|
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 1997-2014 Leibniz Universitaet Hannover |
---|
18 | !--------------------------------------------------------------------------------! |
---|
19 | ! |
---|
20 | ! Attention: Loop unrolling and cache optimization in SOR-Red/Black method |
---|
21 | ! still does not give the expected speedup! Further work required. |
---|
22 | ! |
---|
23 | ! Current revisions: |
---|
24 | ! ----------------- |
---|
25 | ! |
---|
26 | ! |
---|
27 | ! Former revisions: |
---|
28 | ! ----------------- |
---|
29 | ! $Id: poismg.f90 1323 2014-03-20 17:09:54Z maronga $ |
---|
30 | ! |
---|
31 | ! 1322 2014-03-20 16:38:49Z raasch |
---|
32 | ! REAL constants defined as wp-kind |
---|
33 | ! |
---|
34 | ! 1320 2014-03-20 08:40:49Z raasch |
---|
35 | ! ONLY-attribute added to USE-statements, |
---|
36 | ! kind-parameters added to all INTEGER and REAL declaration statements, |
---|
37 | ! kinds are defined in new module kinds, |
---|
38 | ! old module precision_kind is removed, |
---|
39 | ! revision history before 2012 removed, |
---|
40 | ! comment fields (!:) to be used for variable explanations added to |
---|
41 | ! all variable declaration statements |
---|
42 | ! |
---|
43 | ! 1318 2014-03-17 13:35:16Z raasch |
---|
44 | ! module interfaces removed |
---|
45 | ! |
---|
46 | ! 1159 2013-05-21 11:58:22Z fricke |
---|
47 | ! bc_lr/ns_dirneu/neudir removed |
---|
48 | ! |
---|
49 | ! 1092 2013-02-02 11:24:22Z raasch |
---|
50 | ! unused variables removed |
---|
51 | ! |
---|
52 | ! 1056 2012-11-16 15:28:04Z raasch |
---|
53 | ! Bugfix: all ghost points have to be used for allocating p3 |
---|
54 | ! arrays p2, f2, and f2_l changed from allocatable to automatic |
---|
55 | ! |
---|
56 | ! 1036 2012-10-22 13:43:42Z raasch |
---|
57 | ! code put under GPL (PALM 3.9) |
---|
58 | ! |
---|
59 | ! 996 2012-09-07 10:41:47Z raasch |
---|
60 | ! little reformatting |
---|
61 | ! |
---|
62 | ! 978 2012-08-09 08:28:32Z fricke |
---|
63 | ! bc_lr/ns_dirneu/neudir added |
---|
64 | ! |
---|
65 | ! 880 2012-04-13 06:28:59Z raasch |
---|
66 | ! Bugfix: preprocessor statements for parallel execution added |
---|
67 | ! |
---|
68 | ! 778 2011-11-07 14:18:25Z fricke |
---|
69 | ! Allocation of p3 changes when multigrid is used and the collected field on PE0 |
---|
70 | ! has more grid points than the subdomain of an PE. |
---|
71 | ! |
---|
72 | ! 707 2011-03-29 11:39:40Z raasch |
---|
73 | ! p_loc is used instead of p in the main routine (poismg). |
---|
74 | ! On coarse grid levels, gathered data are identically processed on all PEs |
---|
75 | ! (before, on PE0 only), so that the subsequent scattering of data is not |
---|
76 | ! neccessary any more. |
---|
77 | ! bc_lr/ns replaced by bc_lr/ns_cyc/dirrad/raddir |
---|
78 | ! Bugfix: bottom (nzb) and top (nzt+1) boundary conditions set in routines |
---|
79 | ! resid and restrict. They were missed before which may have led to |
---|
80 | ! unpredictable results. |
---|
81 | ! |
---|
82 | ! 667 2010-12-23 12:06:00Z suehring/gryschka |
---|
83 | ! Calls of exchange_horiz are modified. |
---|
84 | ! |
---|
85 | ! 622 2010-12-10 08:08:13Z raasch |
---|
86 | ! optional barriers included in order to speed up collective operations |
---|
87 | ! |
---|
88 | ! 257 2009-03-11 15:17:42Z heinze |
---|
89 | ! Output of messages replaced by message handling routine. |
---|
90 | ! |
---|
91 | ! 181 2008-07-30 07:07:47Z raasch |
---|
92 | ! Bugfix: grid_level+1 has to be used in restrict for flags-array |
---|
93 | ! |
---|
94 | ! 114 2007-10-10 00:03:15Z raasch |
---|
95 | ! Boundary conditions at walls are implicitly set using flag arrays. Only |
---|
96 | ! Neumann BC is allowed. Upper walls are still not realized. |
---|
97 | ! Bottom and top BCs for array f_mg in restrict removed because boundary |
---|
98 | ! values are not needed (right hand side of SOR iteration). |
---|
99 | ! |
---|
100 | ! 75 2007-03-22 09:54:05Z raasch |
---|
101 | ! 2nd+3rd argument removed from exchange horiz |
---|
102 | ! |
---|
103 | ! RCS Log replace by Id keyword, revision history cleaned up |
---|
104 | ! |
---|
105 | ! Revision 1.6 2005/03/26 20:55:54 raasch |
---|
106 | ! Implementation of non-cyclic (Neumann) horizontal boundary conditions, |
---|
107 | ! routine prolong simplified (one call of exchange_horiz spared) |
---|
108 | ! |
---|
109 | ! Revision 1.1 2001/07/20 13:10:51 raasch |
---|
110 | ! Initial revision |
---|
111 | ! |
---|
112 | ! |
---|
113 | ! Description: |
---|
114 | ! ------------ |
---|
115 | ! Solves the Poisson equation for the perturbation pressure with a multigrid |
---|
116 | ! V- or W-Cycle scheme. |
---|
117 | ! |
---|
118 | ! This multigrid method was originally developed for PALM by Joerg Uhlenbrock, |
---|
119 | ! September 2000 - July 2001. |
---|
120 | !------------------------------------------------------------------------------! |
---|
121 | |
---|
122 | USE arrays_3d, & |
---|
123 | ONLY: d, p_loc |
---|
124 | |
---|
125 | USE control_parameters, & |
---|
126 | ONLY: gathered_size, grid_level, grid_level_count, & |
---|
127 | maximum_grid_level, message_string, mgcycles, mg_cycles, & |
---|
128 | mg_switch_to_pe0_level, residual_limit, subdomain_size |
---|
129 | |
---|
130 | USE cpulog, & |
---|
131 | ONLY: cpu_log, log_point_s |
---|
132 | |
---|
133 | USE indices, & |
---|
134 | ONLY: nxl, nxlg, nxl_mg, nxr, nxrg, nxr_mg, nys, nysg, nys_mg, nyn, & |
---|
135 | nyng, nyn_mg, nzb, nzt, nzt_mg |
---|
136 | |
---|
137 | USE kinds |
---|
138 | |
---|
139 | USE pegrid |
---|
140 | |
---|
141 | IMPLICIT NONE |
---|
142 | |
---|
143 | REAL(wp) :: maxerror !: |
---|
144 | REAL(wp) :: maximum_mgcycles !: |
---|
145 | REAL(wp) :: residual_norm !: |
---|
146 | |
---|
147 | REAL(wp), DIMENSION(nzb:nzt+1,nys-1:nyn+1,nxl-1:nxr+1) :: r !: |
---|
148 | |
---|
149 | REAL(wp), DIMENSION(:,:,:), ALLOCATABLE :: p3 !: |
---|
150 | |
---|
151 | |
---|
152 | CALL cpu_log( log_point_s(29), 'poismg', 'start' ) |
---|
153 | ! |
---|
154 | !-- Initialize arrays and variables used in this subroutine |
---|
155 | |
---|
156 | !-- If the number of grid points of the gathered grid, which is collected |
---|
157 | !-- on PE0, is larger than the number of grid points of an PE, than array |
---|
158 | !-- p3 will be enlarged. |
---|
159 | IF ( gathered_size > subdomain_size ) THEN |
---|
160 | ALLOCATE( p3(nzb:nzt_mg(mg_switch_to_pe0_level)+1,nys_mg( & |
---|
161 | mg_switch_to_pe0_level)-1:nyn_mg(mg_switch_to_pe0_level)+1,& |
---|
162 | nxl_mg(mg_switch_to_pe0_level)-1:nxr_mg( & |
---|
163 | mg_switch_to_pe0_level)+1) ) |
---|
164 | ELSE |
---|
165 | ALLOCATE ( p3(nzb:nzt+1,nysg:nyng,nxlg:nxrg) ) |
---|
166 | ENDIF |
---|
167 | |
---|
168 | p3 = 0.0 |
---|
169 | |
---|
170 | ! |
---|
171 | !-- Ghost boundaries have to be added to divergence array. |
---|
172 | !-- Exchange routine needs to know the grid level! |
---|
173 | grid_level = maximum_grid_level |
---|
174 | CALL exchange_horiz( d, 1) |
---|
175 | d(nzb,:,:) = d(nzb+1,:,:) |
---|
176 | |
---|
177 | ! |
---|
178 | !-- Initiation of the multigrid scheme. Does n cycles until the |
---|
179 | !-- residual is smaller than the given limit. The accuracy of the solution |
---|
180 | !-- of the poisson equation will increase with the number of cycles. |
---|
181 | !-- If the number of cycles is preset by the user, this number will be |
---|
182 | !-- carried out regardless of the accuracy. |
---|
183 | grid_level_count = 0 |
---|
184 | mgcycles = 0 |
---|
185 | IF ( mg_cycles == -1 ) THEN |
---|
186 | maximum_mgcycles = 0 |
---|
187 | residual_norm = 1.0 |
---|
188 | ELSE |
---|
189 | maximum_mgcycles = mg_cycles |
---|
190 | residual_norm = 0.0 |
---|
191 | ENDIF |
---|
192 | |
---|
193 | DO WHILE ( residual_norm > residual_limit .OR. & |
---|
194 | mgcycles < maximum_mgcycles ) |
---|
195 | |
---|
196 | CALL next_mg_level( d, p_loc, p3, r) |
---|
197 | |
---|
198 | ! |
---|
199 | !-- Calculate the residual if the user has not preset the number of |
---|
200 | !-- cycles to be performed |
---|
201 | IF ( maximum_mgcycles == 0 ) THEN |
---|
202 | CALL resid( d, p_loc, r ) |
---|
203 | maxerror = SUM( r(nzb+1:nzt,nys:nyn,nxl:nxr)**2 ) |
---|
204 | |
---|
205 | #if defined( __parallel ) |
---|
206 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
---|
207 | CALL MPI_ALLREDUCE( maxerror, residual_norm, 1, MPI_REAL, MPI_SUM, & |
---|
208 | comm2d, ierr) |
---|
209 | #else |
---|
210 | residual_norm = maxerror |
---|
211 | #endif |
---|
212 | residual_norm = SQRT( residual_norm ) |
---|
213 | ENDIF |
---|
214 | |
---|
215 | mgcycles = mgcycles + 1 |
---|
216 | |
---|
217 | ! |
---|
218 | !-- If the user has not limited the number of cycles, stop the run in case |
---|
219 | !-- of insufficient convergence |
---|
220 | IF ( mgcycles > 1000 .AND. mg_cycles == -1 ) THEN |
---|
221 | message_string = 'no sufficient convergence within 1000 cycles' |
---|
222 | CALL message( 'poismg', 'PA0283', 1, 2, 0, 6, 0 ) |
---|
223 | ENDIF |
---|
224 | |
---|
225 | ENDDO |
---|
226 | |
---|
227 | DEALLOCATE( p3 ) |
---|
228 | |
---|
229 | ! |
---|
230 | !-- Unset the grid level. Variable is used to determine the MPI datatypes for |
---|
231 | !-- ghost point exchange |
---|
232 | grid_level = 0 |
---|
233 | |
---|
234 | CALL cpu_log( log_point_s(29), 'poismg', 'stop' ) |
---|
235 | |
---|
236 | END SUBROUTINE poismg |
---|
237 | |
---|
238 | |
---|
239 | |
---|
240 | SUBROUTINE resid( f_mg, p_mg, r ) |
---|
241 | |
---|
242 | !------------------------------------------------------------------------------! |
---|
243 | ! Description: |
---|
244 | ! ------------ |
---|
245 | ! Computes the residual of the perturbation pressure. |
---|
246 | !------------------------------------------------------------------------------! |
---|
247 | |
---|
248 | USE arrays_3d, & |
---|
249 | ONLY: f1_mg, f2_mg, f3_mg |
---|
250 | |
---|
251 | USE control_parameters, & |
---|
252 | ONLY: bc_lr_cyc, bc_ns_cyc, grid_level, ibc_p_b, ibc_p_t, inflow_l, & |
---|
253 | inflow_n, inflow_r, inflow_s, outflow_l, outflow_n, outflow_r, & |
---|
254 | outflow_s |
---|
255 | |
---|
256 | USE grid_variables, & |
---|
257 | ONLY: ddx2_mg, ddy2_mg |
---|
258 | |
---|
259 | USE indices, & |
---|
260 | ONLY: flags, wall_flags_1, wall_flags_2, wall_flags_3, wall_flags_4, & |
---|
261 | wall_flags_5, wall_flags_6, wall_flags_7, wall_flags_8, & |
---|
262 | wall_flags_9, wall_flags_10, nxl_mg, nxr_mg, nys_mg, nyn_mg, & |
---|
263 | nzb, nzt_mg |
---|
264 | |
---|
265 | USE kinds |
---|
266 | |
---|
267 | IMPLICIT NONE |
---|
268 | |
---|
269 | INTEGER(iwp) :: i |
---|
270 | INTEGER(iwp) :: j |
---|
271 | INTEGER(iwp) :: k |
---|
272 | INTEGER(iwp) :: l |
---|
273 | |
---|
274 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
275 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
276 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: f_mg !: |
---|
277 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
278 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
279 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: p_mg !: |
---|
280 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
281 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
282 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: r !: |
---|
283 | |
---|
284 | ! |
---|
285 | !-- Calculate the residual |
---|
286 | l = grid_level |
---|
287 | |
---|
288 | ! |
---|
289 | !-- Choose flag array of this level |
---|
290 | SELECT CASE ( l ) |
---|
291 | CASE ( 1 ) |
---|
292 | flags => wall_flags_1 |
---|
293 | CASE ( 2 ) |
---|
294 | flags => wall_flags_2 |
---|
295 | CASE ( 3 ) |
---|
296 | flags => wall_flags_3 |
---|
297 | CASE ( 4 ) |
---|
298 | flags => wall_flags_4 |
---|
299 | CASE ( 5 ) |
---|
300 | flags => wall_flags_5 |
---|
301 | CASE ( 6 ) |
---|
302 | flags => wall_flags_6 |
---|
303 | CASE ( 7 ) |
---|
304 | flags => wall_flags_7 |
---|
305 | CASE ( 8 ) |
---|
306 | flags => wall_flags_8 |
---|
307 | CASE ( 9 ) |
---|
308 | flags => wall_flags_9 |
---|
309 | CASE ( 10 ) |
---|
310 | flags => wall_flags_10 |
---|
311 | END SELECT |
---|
312 | |
---|
313 | !$OMP PARALLEL PRIVATE (i,j,k) |
---|
314 | !$OMP DO |
---|
315 | DO i = nxl_mg(l), nxr_mg(l) |
---|
316 | DO j = nys_mg(l), nyn_mg(l) |
---|
317 | DO k = nzb+1, nzt_mg(l) |
---|
318 | r(k,j,i) = f_mg(k,j,i) & |
---|
319 | - ddx2_mg(l) * & |
---|
320 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
321 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
322 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
323 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
324 | - ddy2_mg(l) * & |
---|
325 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
326 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
327 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
328 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
329 | - f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
330 | - f3_mg(k,l) * & |
---|
331 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
332 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
333 | + f1_mg(k,l) * p_mg(k,j,i) |
---|
334 | ! |
---|
335 | !-- Residual within topography should be zero |
---|
336 | r(k,j,i) = r(k,j,i) * ( 1.0 - IBITS( flags(k,j,i), 6, 1 ) ) |
---|
337 | ENDDO |
---|
338 | ENDDO |
---|
339 | ENDDO |
---|
340 | !$OMP END PARALLEL |
---|
341 | |
---|
342 | ! |
---|
343 | !-- Horizontal boundary conditions |
---|
344 | CALL exchange_horiz( r, 1) |
---|
345 | |
---|
346 | IF ( .NOT. bc_lr_cyc ) THEN |
---|
347 | IF ( inflow_l .OR. outflow_l ) r(:,:,nxl_mg(l)-1) = r(:,:,nxl_mg(l)) |
---|
348 | IF ( inflow_r .OR. outflow_r ) r(:,:,nxr_mg(l)+1) = r(:,:,nxr_mg(l)) |
---|
349 | ENDIF |
---|
350 | |
---|
351 | IF ( .NOT. bc_ns_cyc ) THEN |
---|
352 | IF ( inflow_n .OR. outflow_n ) r(:,nyn_mg(l)+1,:) = r(:,nyn_mg(l),:) |
---|
353 | IF ( inflow_s .OR. outflow_s ) r(:,nys_mg(l)-1,:) = r(:,nys_mg(l),:) |
---|
354 | ENDIF |
---|
355 | |
---|
356 | ! |
---|
357 | !-- Boundary conditions at bottom and top of the domain. |
---|
358 | !-- These points are not handled by the above loop. Points may be within |
---|
359 | !-- buildings, but that doesn't matter. |
---|
360 | IF ( ibc_p_b == 1 ) THEN |
---|
361 | r(nzb,:,: ) = r(nzb+1,:,:) |
---|
362 | ELSE |
---|
363 | r(nzb,:,: ) = 0.0 |
---|
364 | ENDIF |
---|
365 | |
---|
366 | IF ( ibc_p_t == 1 ) THEN |
---|
367 | r(nzt_mg(l)+1,:,: ) = r(nzt_mg(l),:,:) |
---|
368 | ELSE |
---|
369 | r(nzt_mg(l)+1,:,: ) = 0.0 |
---|
370 | ENDIF |
---|
371 | |
---|
372 | |
---|
373 | END SUBROUTINE resid |
---|
374 | |
---|
375 | |
---|
376 | |
---|
377 | SUBROUTINE restrict( f_mg, r ) |
---|
378 | |
---|
379 | !------------------------------------------------------------------------------! |
---|
380 | ! Description: |
---|
381 | ! ------------ |
---|
382 | ! Interpolates the residual on the next coarser grid with "full weighting" |
---|
383 | ! scheme |
---|
384 | !------------------------------------------------------------------------------! |
---|
385 | |
---|
386 | USE control_parameters, & |
---|
387 | ONLY: bc_lr_cyc, bc_ns_cyc, grid_level, ibc_p_b, ibc_p_t, inflow_l, & |
---|
388 | inflow_n, inflow_r, inflow_s, outflow_l, outflow_n, outflow_r, & |
---|
389 | outflow_s |
---|
390 | |
---|
391 | USE indices, & |
---|
392 | ONLY: flags, wall_flags_1, wall_flags_2, wall_flags_3, wall_flags_4, & |
---|
393 | wall_flags_5, wall_flags_6, wall_flags_7, wall_flags_8, & |
---|
394 | wall_flags_9, wall_flags_10, nxl_mg, nxr_mg, nys_mg, nyn_mg, & |
---|
395 | nzb, nzt_mg |
---|
396 | |
---|
397 | USE kinds |
---|
398 | |
---|
399 | IMPLICIT NONE |
---|
400 | |
---|
401 | INTEGER(iwp) :: i !: |
---|
402 | INTEGER(iwp) :: ic !: |
---|
403 | INTEGER(iwp) :: j !: |
---|
404 | INTEGER(iwp) :: jc !: |
---|
405 | INTEGER(iwp) :: k !: |
---|
406 | INTEGER(iwp) :: kc !: |
---|
407 | INTEGER(iwp) :: l !: |
---|
408 | |
---|
409 | REAL(wp) :: rkjim !: |
---|
410 | REAL(wp) :: rkjip !: |
---|
411 | REAL(wp) :: rkjmi !: |
---|
412 | REAL(wp) :: rkjmim !: |
---|
413 | REAL(wp) :: rkjmip !: |
---|
414 | REAL(wp) :: rkjpi !: |
---|
415 | REAL(wp) :: rkjpim !: |
---|
416 | REAL(wp) :: rkjpip !: |
---|
417 | REAL(wp) :: rkmji !: |
---|
418 | REAL(wp) :: rkmjim !: |
---|
419 | REAL(wp) :: rkmjip !: |
---|
420 | REAL(wp) :: rkmjmi !: |
---|
421 | REAL(wp) :: rkmjmim !: |
---|
422 | REAL(wp) :: rkmjmip !: |
---|
423 | REAL(wp) :: rkmjpi !: |
---|
424 | REAL(wp) :: rkmjpim !: |
---|
425 | REAL(wp) :: rkmjpip !: |
---|
426 | |
---|
427 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
428 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
429 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: f_mg !: |
---|
430 | |
---|
431 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level+1)+1, & |
---|
432 | nys_mg(grid_level+1)-1:nyn_mg(grid_level+1)+1, & |
---|
433 | nxl_mg(grid_level+1)-1:nxr_mg(grid_level+1)+1) :: r !: |
---|
434 | |
---|
435 | ! |
---|
436 | !-- Interpolate the residual |
---|
437 | l = grid_level |
---|
438 | |
---|
439 | ! |
---|
440 | !-- Choose flag array of the upper level |
---|
441 | SELECT CASE ( l+1 ) |
---|
442 | CASE ( 1 ) |
---|
443 | flags => wall_flags_1 |
---|
444 | CASE ( 2 ) |
---|
445 | flags => wall_flags_2 |
---|
446 | CASE ( 3 ) |
---|
447 | flags => wall_flags_3 |
---|
448 | CASE ( 4 ) |
---|
449 | flags => wall_flags_4 |
---|
450 | CASE ( 5 ) |
---|
451 | flags => wall_flags_5 |
---|
452 | CASE ( 6 ) |
---|
453 | flags => wall_flags_6 |
---|
454 | CASE ( 7 ) |
---|
455 | flags => wall_flags_7 |
---|
456 | CASE ( 8 ) |
---|
457 | flags => wall_flags_8 |
---|
458 | CASE ( 9 ) |
---|
459 | flags => wall_flags_9 |
---|
460 | CASE ( 10 ) |
---|
461 | flags => wall_flags_10 |
---|
462 | END SELECT |
---|
463 | |
---|
464 | !$OMP PARALLEL PRIVATE (i,j,k,ic,jc,kc) |
---|
465 | !$OMP DO |
---|
466 | DO ic = nxl_mg(l), nxr_mg(l) |
---|
467 | i = 2*ic |
---|
468 | DO jc = nys_mg(l), nyn_mg(l) |
---|
469 | j = 2*jc |
---|
470 | DO kc = nzb+1, nzt_mg(l) |
---|
471 | k = 2*kc-1 |
---|
472 | ! |
---|
473 | !-- Use implicit Neumann BCs if the respective gridpoint is inside |
---|
474 | !-- the building |
---|
475 | rkjim = r(k,j,i-1) + IBITS( flags(k,j,i-1), 6, 1 ) * & |
---|
476 | ( r(k,j,i) - r(k,j,i-1) ) |
---|
477 | rkjip = r(k,j,i+1) + IBITS( flags(k,j,i+1), 6, 1 ) * & |
---|
478 | ( r(k,j,i) - r(k,j,i+1) ) |
---|
479 | rkjpi = r(k,j+1,i) + IBITS( flags(k,j+1,i), 6, 1 ) * & |
---|
480 | ( r(k,j,i) - r(k,j+1,i) ) |
---|
481 | rkjmi = r(k,j-1,i) + IBITS( flags(k,j-1,i), 6, 1 ) * & |
---|
482 | ( r(k,j,i) - r(k,j-1,i) ) |
---|
483 | rkjmim = r(k,j-1,i-1) + IBITS( flags(k,j-1,i-1), 6, 1 ) * & |
---|
484 | ( r(k,j,i) - r(k,j-1,i-1) ) |
---|
485 | rkjpim = r(k,j+1,i-1) + IBITS( flags(k,j+1,i-1), 6, 1 ) * & |
---|
486 | ( r(k,j,i) - r(k,j+1,i-1) ) |
---|
487 | rkjmip = r(k,j-1,i+1) + IBITS( flags(k,j-1,i+1), 6, 1 ) * & |
---|
488 | ( r(k,j,i) - r(k,j-1,i+1) ) |
---|
489 | rkjpip = r(k,j+1,i+1) + IBITS( flags(k,j+1,i+1), 6, 1 ) * & |
---|
490 | ( r(k,j,i) - r(k,j+1,i+1) ) |
---|
491 | rkmji = r(k-1,j,i) + IBITS( flags(k-1,j,i), 6, 1 ) * & |
---|
492 | ( r(k,j,i) - r(k-1,j,i) ) |
---|
493 | rkmjim = r(k-1,j,i-1) + IBITS( flags(k-1,j,i-1), 6, 1 ) * & |
---|
494 | ( r(k,j,i) - r(k-1,j,i-1) ) |
---|
495 | rkmjip = r(k-1,j,i+1) + IBITS( flags(k-1,j,i+1), 6, 1 ) * & |
---|
496 | ( r(k,j,i) - r(k-1,j,i+1) ) |
---|
497 | rkmjpi = r(k-1,j+1,i) + IBITS( flags(k-1,j+1,i), 6, 1 ) * & |
---|
498 | ( r(k,j,i) - r(k-1,j+1,i) ) |
---|
499 | rkmjmi = r(k-1,j-1,i) + IBITS( flags(k-1,j-1,i), 6, 1 ) * & |
---|
500 | ( r(k,j,i) - r(k-1,j-1,i) ) |
---|
501 | rkmjmim = r(k-1,j-1,i-1) + IBITS( flags(k-1,j-1,i-1), 6, 1 ) * & |
---|
502 | ( r(k,j,i) - r(k-1,j-1,i-1) ) |
---|
503 | rkmjpim = r(k-1,j+1,i-1) + IBITS( flags(k-1,j+1,i-1), 6, 1 ) * & |
---|
504 | ( r(k,j,i) - r(k-1,j+1,i-1) ) |
---|
505 | rkmjmip = r(k-1,j-1,i+1) + IBITS( flags(k-1,j-1,i+1), 6, 1 ) * & |
---|
506 | ( r(k,j,i) - r(k-1,j-1,i+1) ) |
---|
507 | rkmjpip = r(k-1,j+1,i+1) + IBITS( flags(k-1,j+1,i+1), 6, 1 ) * & |
---|
508 | ( r(k,j,i) - r(k-1,j+1,i+1) ) |
---|
509 | |
---|
510 | f_mg(kc,jc,ic) = 1.0 / 64.0_wp * ( & |
---|
511 | 8.0 * r(k,j,i) & |
---|
512 | + 4.0 * ( rkjim + rkjip + & |
---|
513 | rkjpi + rkjmi ) & |
---|
514 | + 2.0 * ( rkjmim + rkjpim + & |
---|
515 | rkjmip + rkjpip ) & |
---|
516 | + 4.0 * rkmji & |
---|
517 | + 2.0 * ( rkmjim + rkmjim + & |
---|
518 | rkmjpi + rkmjmi ) & |
---|
519 | + ( rkmjmim + rkmjpim + & |
---|
520 | rkmjmip + rkmjpip ) & |
---|
521 | + 4.0 * r(k+1,j,i) & |
---|
522 | + 2.0 * ( r(k+1,j,i-1) + r(k+1,j,i+1) + & |
---|
523 | r(k+1,j+1,i) + r(k+1,j-1,i) ) & |
---|
524 | + ( r(k+1,j-1,i-1) + r(k+1,j+1,i-1) + & |
---|
525 | r(k+1,j-1,i+1) + r(k+1,j+1,i+1) ) & |
---|
526 | ) |
---|
527 | |
---|
528 | ! f_mg(kc,jc,ic) = 1.0 / 64.0_wp * ( & |
---|
529 | ! 8.0 * r(k,j,i) & |
---|
530 | ! + 4.0 * ( r(k,j,i-1) + r(k,j,i+1) + & |
---|
531 | ! r(k,j+1,i) + r(k,j-1,i) ) & |
---|
532 | ! + 2.0 * ( r(k,j-1,i-1) + r(k,j+1,i-1) + & |
---|
533 | ! r(k,j-1,i+1) + r(k,j+1,i+1) ) & |
---|
534 | ! + 4.0 * r(k-1,j,i) & |
---|
535 | ! + 2.0 * ( r(k-1,j,i-1) + r(k-1,j,i+1) + & |
---|
536 | ! r(k-1,j+1,i) + r(k-1,j-1,i) ) & |
---|
537 | ! + ( r(k-1,j-1,i-1) + r(k-1,j+1,i-1) + & |
---|
538 | ! r(k-1,j-1,i+1) + r(k-1,j+1,i+1) ) & |
---|
539 | ! + 4.0 * r(k+1,j,i) & |
---|
540 | ! + 2.0 * ( r(k+1,j,i-1) + r(k+1,j,i+1) + & |
---|
541 | ! r(k+1,j+1,i) + r(k+1,j-1,i) ) & |
---|
542 | ! + ( r(k+1,j-1,i-1) + r(k+1,j+1,i-1) + & |
---|
543 | ! r(k+1,j-1,i+1) + r(k+1,j+1,i+1) ) & |
---|
544 | ! ) |
---|
545 | ENDDO |
---|
546 | ENDDO |
---|
547 | ENDDO |
---|
548 | !$OMP END PARALLEL |
---|
549 | |
---|
550 | ! |
---|
551 | !-- Horizontal boundary conditions |
---|
552 | CALL exchange_horiz( f_mg, 1) |
---|
553 | |
---|
554 | IF ( .NOT. bc_lr_cyc ) THEN |
---|
555 | IF (inflow_l .OR. outflow_l) f_mg(:,:,nxl_mg(l)-1) = f_mg(:,:,nxl_mg(l)) |
---|
556 | IF (inflow_r .OR. outflow_r) f_mg(:,:,nxr_mg(l)+1) = f_mg(:,:,nxr_mg(l)) |
---|
557 | ENDIF |
---|
558 | |
---|
559 | IF ( .NOT. bc_ns_cyc ) THEN |
---|
560 | IF (inflow_n .OR. outflow_n) f_mg(:,nyn_mg(l)+1,:) = f_mg(:,nyn_mg(l),:) |
---|
561 | IF (inflow_s .OR. outflow_s) f_mg(:,nys_mg(l)-1,:) = f_mg(:,nys_mg(l),:) |
---|
562 | ENDIF |
---|
563 | |
---|
564 | ! |
---|
565 | !-- Boundary conditions at bottom and top of the domain. |
---|
566 | !-- These points are not handled by the above loop. Points may be within |
---|
567 | !-- buildings, but that doesn't matter. |
---|
568 | IF ( ibc_p_b == 1 ) THEN |
---|
569 | f_mg(nzb,:,: ) = f_mg(nzb+1,:,:) |
---|
570 | ELSE |
---|
571 | f_mg(nzb,:,: ) = 0.0 |
---|
572 | ENDIF |
---|
573 | |
---|
574 | IF ( ibc_p_t == 1 ) THEN |
---|
575 | f_mg(nzt_mg(l)+1,:,: ) = f_mg(nzt_mg(l),:,:) |
---|
576 | ELSE |
---|
577 | f_mg(nzt_mg(l)+1,:,: ) = 0.0 |
---|
578 | ENDIF |
---|
579 | |
---|
580 | |
---|
581 | END SUBROUTINE restrict |
---|
582 | |
---|
583 | |
---|
584 | |
---|
585 | SUBROUTINE prolong( p, temp ) |
---|
586 | |
---|
587 | !------------------------------------------------------------------------------! |
---|
588 | ! Description: |
---|
589 | ! ------------ |
---|
590 | ! Interpolates the correction of the perturbation pressure |
---|
591 | ! to the next finer grid. |
---|
592 | !------------------------------------------------------------------------------! |
---|
593 | |
---|
594 | USE control_parameters, & |
---|
595 | ONLY: bc_lr_cyc, bc_ns_cyc, grid_level, ibc_p_b, ibc_p_t, inflow_l, & |
---|
596 | inflow_n, inflow_r, inflow_s, outflow_l, outflow_n, outflow_r, & |
---|
597 | outflow_s |
---|
598 | |
---|
599 | USE indices, & |
---|
600 | ONLY: nxl_mg, nxr_mg, nys_mg, nyn_mg, nzb, nzt_mg |
---|
601 | |
---|
602 | USE kinds |
---|
603 | |
---|
604 | IMPLICIT NONE |
---|
605 | |
---|
606 | INTEGER(iwp) :: i !: |
---|
607 | INTEGER(iwp) :: j !: |
---|
608 | INTEGER(iwp) :: k !: |
---|
609 | INTEGER(iwp) :: l !: |
---|
610 | |
---|
611 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level-1)+1, & |
---|
612 | nys_mg(grid_level-1)-1:nyn_mg(grid_level-1)+1, & |
---|
613 | nxl_mg(grid_level-1)-1:nxr_mg(grid_level-1)+1 ) :: p !: |
---|
614 | |
---|
615 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
616 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
617 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: temp !: |
---|
618 | |
---|
619 | |
---|
620 | ! |
---|
621 | !-- First, store elements of the coarser grid on the next finer grid |
---|
622 | l = grid_level |
---|
623 | |
---|
624 | !$OMP PARALLEL PRIVATE (i,j,k) |
---|
625 | !$OMP DO |
---|
626 | DO i = nxl_mg(l-1), nxr_mg(l-1) |
---|
627 | DO j = nys_mg(l-1), nyn_mg(l-1) |
---|
628 | !CDIR NODEP |
---|
629 | DO k = nzb+1, nzt_mg(l-1) |
---|
630 | ! |
---|
631 | !-- Points of the coarse grid are directly stored on the next finer |
---|
632 | !-- grid |
---|
633 | temp(2*k-1,2*j,2*i) = p(k,j,i) |
---|
634 | ! |
---|
635 | !-- Points between two coarse-grid points |
---|
636 | temp(2*k-1,2*j,2*i+1) = 0.5 * ( p(k,j,i) + p(k,j,i+1) ) |
---|
637 | temp(2*k-1,2*j+1,2*i) = 0.5 * ( p(k,j,i) + p(k,j+1,i) ) |
---|
638 | temp(2*k,2*j,2*i) = 0.5 * ( p(k,j,i) + p(k+1,j,i) ) |
---|
639 | ! |
---|
640 | !-- Points in the center of the planes stretched by four points |
---|
641 | !-- of the coarse grid cube |
---|
642 | temp(2*k-1,2*j+1,2*i+1) = 0.25 * ( p(k,j,i) + p(k,j,i+1) + & |
---|
643 | p(k,j+1,i) + p(k,j+1,i+1) ) |
---|
644 | temp(2*k,2*j,2*i+1) = 0.25 * ( p(k,j,i) + p(k,j,i+1) + & |
---|
645 | p(k+1,j,i) + p(k+1,j,i+1) ) |
---|
646 | temp(2*k,2*j+1,2*i) = 0.25 * ( p(k,j,i) + p(k,j+1,i) + & |
---|
647 | p(k+1,j,i) + p(k+1,j+1,i) ) |
---|
648 | ! |
---|
649 | !-- Points in the middle of coarse grid cube |
---|
650 | temp(2*k,2*j+1,2*i+1) = 0.125 * ( p(k,j,i) + p(k,j,i+1) + & |
---|
651 | p(k,j+1,i) + p(k,j+1,i+1) + & |
---|
652 | p(k+1,j,i) + p(k+1,j,i+1) + & |
---|
653 | p(k+1,j+1,i) + p(k+1,j+1,i+1) ) |
---|
654 | ENDDO |
---|
655 | ENDDO |
---|
656 | ENDDO |
---|
657 | !$OMP END PARALLEL |
---|
658 | |
---|
659 | ! |
---|
660 | !-- Horizontal boundary conditions |
---|
661 | CALL exchange_horiz( temp, 1) |
---|
662 | |
---|
663 | IF ( .NOT. bc_lr_cyc ) THEN |
---|
664 | IF (inflow_l .OR. outflow_l) temp(:,:,nxl_mg(l)-1) = temp(:,:,nxl_mg(l)) |
---|
665 | IF (inflow_r .OR. outflow_r) temp(:,:,nxr_mg(l)+1) = temp(:,:,nxr_mg(l)) |
---|
666 | ENDIF |
---|
667 | |
---|
668 | IF ( .NOT. bc_ns_cyc ) THEN |
---|
669 | IF (inflow_n .OR. outflow_n) temp(:,nyn_mg(l)+1,:) = temp(:,nyn_mg(l),:) |
---|
670 | IF (inflow_s .OR. outflow_s) temp(:,nys_mg(l)-1,:) = temp(:,nys_mg(l),:) |
---|
671 | ENDIF |
---|
672 | |
---|
673 | ! |
---|
674 | !-- Bottom and top boundary conditions |
---|
675 | IF ( ibc_p_b == 1 ) THEN |
---|
676 | temp(nzb,:,: ) = temp(nzb+1,:,:) |
---|
677 | ELSE |
---|
678 | temp(nzb,:,: ) = 0.0 |
---|
679 | ENDIF |
---|
680 | |
---|
681 | IF ( ibc_p_t == 1 ) THEN |
---|
682 | temp(nzt_mg(l)+1,:,: ) = temp(nzt_mg(l),:,:) |
---|
683 | ELSE |
---|
684 | temp(nzt_mg(l)+1,:,: ) = 0.0 |
---|
685 | ENDIF |
---|
686 | |
---|
687 | |
---|
688 | END SUBROUTINE prolong |
---|
689 | |
---|
690 | |
---|
691 | SUBROUTINE redblack( f_mg, p_mg ) |
---|
692 | |
---|
693 | !------------------------------------------------------------------------------! |
---|
694 | ! Description: |
---|
695 | ! ------------ |
---|
696 | ! Relaxation method for the multigrid scheme. A Gauss-Seidel iteration with |
---|
697 | ! 3D-Red-Black decomposition (GS-RB) is used. |
---|
698 | !------------------------------------------------------------------------------! |
---|
699 | |
---|
700 | USE arrays_3d, & |
---|
701 | ONLY: f1_mg, f2_mg, f3_mg |
---|
702 | |
---|
703 | USE control_parameters, & |
---|
704 | ONLY: bc_lr_cyc, bc_ns_cyc, grid_level, ibc_p_b, ibc_p_t, inflow_l, & |
---|
705 | inflow_n, inflow_r, inflow_s, ngsrb, outflow_l, outflow_n, & |
---|
706 | outflow_r, outflow_s |
---|
707 | |
---|
708 | USE cpulog, & |
---|
709 | ONLY: cpu_log, log_point_s |
---|
710 | |
---|
711 | USE grid_variables, & |
---|
712 | ONLY: ddx2_mg, ddy2_mg |
---|
713 | |
---|
714 | USE indices, & |
---|
715 | ONLY: flags, wall_flags_1, wall_flags_2, wall_flags_3, wall_flags_4, & |
---|
716 | wall_flags_5, wall_flags_6, wall_flags_7, wall_flags_8, & |
---|
717 | wall_flags_9, wall_flags_10, nxl_mg, nxr_mg, nys_mg, nyn_mg, & |
---|
718 | nzb, nzt_mg |
---|
719 | |
---|
720 | USE kinds |
---|
721 | |
---|
722 | IMPLICIT NONE |
---|
723 | |
---|
724 | INTEGER(iwp) :: color !: |
---|
725 | INTEGER(iwp) :: i !: |
---|
726 | INTEGER(iwp) :: ic !: |
---|
727 | INTEGER(iwp) :: j !: |
---|
728 | INTEGER(iwp) :: jc !: |
---|
729 | INTEGER(iwp) :: jj !: |
---|
730 | INTEGER(iwp) :: k !: |
---|
731 | INTEGER(iwp) :: l !: |
---|
732 | INTEGER(iwp) :: n !: |
---|
733 | |
---|
734 | LOGICAL :: unroll !: |
---|
735 | |
---|
736 | REAL(wp) :: wall_left !: |
---|
737 | REAL(wp) :: wall_north !: |
---|
738 | REAL(wp) :: wall_right !: |
---|
739 | REAL(wp) :: wall_south !: |
---|
740 | REAL(wp) :: wall_total !: |
---|
741 | REAL(wp) :: wall_top !: |
---|
742 | |
---|
743 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
744 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
745 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: f_mg !: |
---|
746 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
747 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
748 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: p_mg !: |
---|
749 | |
---|
750 | l = grid_level |
---|
751 | |
---|
752 | ! |
---|
753 | !-- Choose flag array of this level |
---|
754 | SELECT CASE ( l ) |
---|
755 | CASE ( 1 ) |
---|
756 | flags => wall_flags_1 |
---|
757 | CASE ( 2 ) |
---|
758 | flags => wall_flags_2 |
---|
759 | CASE ( 3 ) |
---|
760 | flags => wall_flags_3 |
---|
761 | CASE ( 4 ) |
---|
762 | flags => wall_flags_4 |
---|
763 | CASE ( 5 ) |
---|
764 | flags => wall_flags_5 |
---|
765 | CASE ( 6 ) |
---|
766 | flags => wall_flags_6 |
---|
767 | CASE ( 7 ) |
---|
768 | flags => wall_flags_7 |
---|
769 | CASE ( 8 ) |
---|
770 | flags => wall_flags_8 |
---|
771 | CASE ( 9 ) |
---|
772 | flags => wall_flags_9 |
---|
773 | CASE ( 10 ) |
---|
774 | flags => wall_flags_10 |
---|
775 | END SELECT |
---|
776 | |
---|
777 | unroll = ( MOD( nyn_mg(l)-nys_mg(l)+1, 4 ) == 0 .AND. & |
---|
778 | MOD( nxr_mg(l)-nxl_mg(l)+1, 2 ) == 0 ) |
---|
779 | |
---|
780 | DO n = 1, ngsrb |
---|
781 | |
---|
782 | DO color = 1, 2 |
---|
783 | |
---|
784 | IF ( .NOT. unroll ) THEN |
---|
785 | |
---|
786 | CALL cpu_log( log_point_s(36), 'redblack_no_unroll', 'start' ) |
---|
787 | |
---|
788 | ! |
---|
789 | !-- Without unrolling of loops, no cache optimization |
---|
790 | DO i = nxl_mg(l), nxr_mg(l), 2 |
---|
791 | DO j = nys_mg(l) + 2 - color, nyn_mg(l), 2 |
---|
792 | DO k = nzb+1, nzt_mg(l), 2 |
---|
793 | ! p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
794 | ! ddx2_mg(l) * ( p_mg(k,j,i+1) + p_mg(k,j,i-1) ) & |
---|
795 | ! + ddy2_mg(l) * ( p_mg(k,j+1,i) + p_mg(k,j-1,i) ) & |
---|
796 | ! + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
797 | ! + f3_mg(k,l) * p_mg(k-1,j,i) - f_mg(k,j,i) & |
---|
798 | ! ) |
---|
799 | |
---|
800 | p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
801 | ddx2_mg(l) * & |
---|
802 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
803 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
804 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
805 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
806 | + ddy2_mg(l) * & |
---|
807 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
808 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
809 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
810 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
811 | + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
812 | + f3_mg(k,l) * & |
---|
813 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
814 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
815 | - f_mg(k,j,i) ) |
---|
816 | ENDDO |
---|
817 | ENDDO |
---|
818 | ENDDO |
---|
819 | |
---|
820 | DO i = nxl_mg(l)+1, nxr_mg(l), 2 |
---|
821 | DO j = nys_mg(l) + (color-1), nyn_mg(l), 2 |
---|
822 | DO k = nzb+1, nzt_mg(l), 2 |
---|
823 | p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
824 | ddx2_mg(l) * & |
---|
825 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
826 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
827 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
828 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
829 | + ddy2_mg(l) * & |
---|
830 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
831 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
832 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
833 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
834 | + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
835 | + f3_mg(k,l) * & |
---|
836 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
837 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
838 | - f_mg(k,j,i) ) |
---|
839 | ENDDO |
---|
840 | ENDDO |
---|
841 | ENDDO |
---|
842 | |
---|
843 | DO i = nxl_mg(l), nxr_mg(l), 2 |
---|
844 | DO j = nys_mg(l) + (color-1), nyn_mg(l), 2 |
---|
845 | DO k = nzb+2, nzt_mg(l), 2 |
---|
846 | p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
847 | ddx2_mg(l) * & |
---|
848 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
849 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
850 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
851 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
852 | + ddy2_mg(l) * & |
---|
853 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
854 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
855 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
856 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
857 | + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
858 | + f3_mg(k,l) * & |
---|
859 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
860 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
861 | - f_mg(k,j,i) ) |
---|
862 | ENDDO |
---|
863 | ENDDO |
---|
864 | ENDDO |
---|
865 | |
---|
866 | DO i = nxl_mg(l)+1, nxr_mg(l), 2 |
---|
867 | DO j = nys_mg(l) + 2 - color, nyn_mg(l), 2 |
---|
868 | DO k = nzb+2, nzt_mg(l), 2 |
---|
869 | p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
870 | ddx2_mg(l) * & |
---|
871 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
872 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
873 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
874 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
875 | + ddy2_mg(l) * & |
---|
876 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
877 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
878 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
879 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
880 | + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
881 | + f3_mg(k,l) * & |
---|
882 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
883 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
884 | - f_mg(k,j,i) ) |
---|
885 | ENDDO |
---|
886 | ENDDO |
---|
887 | ENDDO |
---|
888 | CALL cpu_log( log_point_s(36), 'redblack_no_unroll', 'stop' ) |
---|
889 | |
---|
890 | ELSE |
---|
891 | |
---|
892 | ! |
---|
893 | !-- Loop unrolling along y, only one i loop for better cache use |
---|
894 | CALL cpu_log( log_point_s(38), 'redblack_unroll', 'start' ) |
---|
895 | DO ic = nxl_mg(l), nxr_mg(l), 2 |
---|
896 | DO jc = nys_mg(l), nyn_mg(l), 4 |
---|
897 | i = ic |
---|
898 | jj = jc+2-color |
---|
899 | DO k = nzb+1, nzt_mg(l), 2 |
---|
900 | j = jj |
---|
901 | p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
902 | ddx2_mg(l) * & |
---|
903 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
904 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
905 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
906 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
907 | + ddy2_mg(l) * & |
---|
908 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
909 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
910 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
911 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
912 | + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
913 | + f3_mg(k,l) * & |
---|
914 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
915 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
916 | - f_mg(k,j,i) ) |
---|
917 | j = jj+2 |
---|
918 | p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
919 | ddx2_mg(l) * & |
---|
920 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
921 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
922 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
923 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
924 | + ddy2_mg(l) * & |
---|
925 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
926 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
927 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
928 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
929 | + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
930 | + f3_mg(k,l) * & |
---|
931 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
932 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
933 | - f_mg(k,j,i) ) |
---|
934 | ENDDO |
---|
935 | |
---|
936 | i = ic+1 |
---|
937 | jj = jc+color-1 |
---|
938 | DO k = nzb+1, nzt_mg(l), 2 |
---|
939 | j =jj |
---|
940 | p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
941 | ddx2_mg(l) * & |
---|
942 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
943 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
944 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
945 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
946 | + ddy2_mg(l) * & |
---|
947 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
948 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
949 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
950 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
951 | + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
952 | + f3_mg(k,l) * & |
---|
953 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
954 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
955 | - f_mg(k,j,i) ) |
---|
956 | j = jj+2 |
---|
957 | p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
958 | ddx2_mg(l) * & |
---|
959 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
960 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
961 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
962 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
963 | + ddy2_mg(l) * & |
---|
964 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
965 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
966 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
967 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
968 | + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
969 | + f3_mg(k,l) * & |
---|
970 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
971 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
972 | - f_mg(k,j,i) ) |
---|
973 | ENDDO |
---|
974 | |
---|
975 | i = ic |
---|
976 | jj = jc+color-1 |
---|
977 | DO k = nzb+2, nzt_mg(l), 2 |
---|
978 | j =jj |
---|
979 | p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
980 | ddx2_mg(l) * & |
---|
981 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
982 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
983 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
984 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
985 | + ddy2_mg(l) * & |
---|
986 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
987 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
988 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
989 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
990 | + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
991 | + f3_mg(k,l) * & |
---|
992 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
993 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
994 | - f_mg(k,j,i) ) |
---|
995 | j = jj+2 |
---|
996 | p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
997 | ddx2_mg(l) * & |
---|
998 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
999 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
1000 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
1001 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
1002 | + ddy2_mg(l) * & |
---|
1003 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
1004 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
1005 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
1006 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
1007 | + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
1008 | + f3_mg(k,l) * & |
---|
1009 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
1010 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
1011 | - f_mg(k,j,i) ) |
---|
1012 | ENDDO |
---|
1013 | |
---|
1014 | i = ic+1 |
---|
1015 | jj = jc+2-color |
---|
1016 | DO k = nzb+2, nzt_mg(l), 2 |
---|
1017 | j =jj |
---|
1018 | p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
1019 | ddx2_mg(l) * & |
---|
1020 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
1021 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
1022 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
1023 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
1024 | + ddy2_mg(l) * & |
---|
1025 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
1026 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
1027 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
1028 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
1029 | + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
1030 | + f3_mg(k,l) * & |
---|
1031 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
1032 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
1033 | - f_mg(k,j,i) ) |
---|
1034 | j = jj+2 |
---|
1035 | p_mg(k,j,i) = 1.0 / f1_mg(k,l) * ( & |
---|
1036 | ddx2_mg(l) * & |
---|
1037 | ( p_mg(k,j,i+1) + IBITS( flags(k,j,i), 5, 1 ) * & |
---|
1038 | ( p_mg(k,j,i) - p_mg(k,j,i+1) ) + & |
---|
1039 | p_mg(k,j,i-1) + IBITS( flags(k,j,i), 4, 1 ) * & |
---|
1040 | ( p_mg(k,j,i) - p_mg(k,j,i-1) ) ) & |
---|
1041 | + ddy2_mg(l) * & |
---|
1042 | ( p_mg(k,j+1,i) + IBITS( flags(k,j,i), 3, 1 ) * & |
---|
1043 | ( p_mg(k,j,i) - p_mg(k,j+1,i) ) + & |
---|
1044 | p_mg(k,j-1,i) + IBITS( flags(k,j,i), 2, 1 ) * & |
---|
1045 | ( p_mg(k,j,i) - p_mg(k,j-1,i) ) ) & |
---|
1046 | + f2_mg(k,l) * p_mg(k+1,j,i) & |
---|
1047 | + f3_mg(k,l) * & |
---|
1048 | ( p_mg(k-1,j,i) + IBITS( flags(k,j,i), 0, 1 ) * & |
---|
1049 | ( p_mg(k,j,i) - p_mg(k-1,j,i) ) ) & |
---|
1050 | - f_mg(k,j,i) ) |
---|
1051 | ENDDO |
---|
1052 | |
---|
1053 | ENDDO |
---|
1054 | ENDDO |
---|
1055 | CALL cpu_log( log_point_s(38), 'redblack_unroll', 'stop' ) |
---|
1056 | |
---|
1057 | ENDIF |
---|
1058 | |
---|
1059 | ! |
---|
1060 | !-- Horizontal boundary conditions |
---|
1061 | CALL exchange_horiz( p_mg, 1 ) |
---|
1062 | |
---|
1063 | IF ( .NOT. bc_lr_cyc ) THEN |
---|
1064 | IF ( inflow_l .OR. outflow_l ) THEN |
---|
1065 | p_mg(:,:,nxl_mg(l)-1) = p_mg(:,:,nxl_mg(l)) |
---|
1066 | ENDIF |
---|
1067 | IF ( inflow_r .OR. outflow_r ) THEN |
---|
1068 | p_mg(:,:,nxr_mg(l)+1) = p_mg(:,:,nxr_mg(l)) |
---|
1069 | ENDIF |
---|
1070 | ENDIF |
---|
1071 | |
---|
1072 | IF ( .NOT. bc_ns_cyc ) THEN |
---|
1073 | IF ( inflow_n .OR. outflow_n ) THEN |
---|
1074 | p_mg(:,nyn_mg(l)+1,:) = p_mg(:,nyn_mg(l),:) |
---|
1075 | ENDIF |
---|
1076 | IF ( inflow_s .OR. outflow_s ) THEN |
---|
1077 | p_mg(:,nys_mg(l)-1,:) = p_mg(:,nys_mg(l),:) |
---|
1078 | ENDIF |
---|
1079 | ENDIF |
---|
1080 | |
---|
1081 | ! |
---|
1082 | !-- Bottom and top boundary conditions |
---|
1083 | IF ( ibc_p_b == 1 ) THEN |
---|
1084 | p_mg(nzb,:,: ) = p_mg(nzb+1,:,:) |
---|
1085 | ELSE |
---|
1086 | p_mg(nzb,:,: ) = 0.0 |
---|
1087 | ENDIF |
---|
1088 | |
---|
1089 | IF ( ibc_p_t == 1 ) THEN |
---|
1090 | p_mg(nzt_mg(l)+1,:,: ) = p_mg(nzt_mg(l),:,:) |
---|
1091 | ELSE |
---|
1092 | p_mg(nzt_mg(l)+1,:,: ) = 0.0 |
---|
1093 | ENDIF |
---|
1094 | |
---|
1095 | ENDDO |
---|
1096 | |
---|
1097 | ENDDO |
---|
1098 | |
---|
1099 | ! |
---|
1100 | !-- Set pressure within topography and at the topography surfaces |
---|
1101 | !$OMP PARALLEL PRIVATE (i,j,k,wall_left,wall_north,wall_right,wall_south,wall_top,wall_total) |
---|
1102 | !$OMP DO |
---|
1103 | DO i = nxl_mg(l), nxr_mg(l) |
---|
1104 | DO j = nys_mg(l), nyn_mg(l) |
---|
1105 | DO k = nzb, nzt_mg(l) |
---|
1106 | ! |
---|
1107 | !-- First, set pressure inside topography to zero |
---|
1108 | p_mg(k,j,i) = p_mg(k,j,i) * ( 1.0 - IBITS( flags(k,j,i), 6, 1 ) ) |
---|
1109 | ! |
---|
1110 | !-- Second, determine if the gridpoint inside topography is adjacent |
---|
1111 | !-- to a wall and set its value to a value given by the average of |
---|
1112 | !-- those values obtained from Neumann boundary condition |
---|
1113 | wall_left = IBITS( flags(k,j,i-1), 5, 1 ) |
---|
1114 | wall_right = IBITS( flags(k,j,i+1), 4, 1 ) |
---|
1115 | wall_south = IBITS( flags(k,j-1,i), 3, 1 ) |
---|
1116 | wall_north = IBITS( flags(k,j+1,i), 2, 1 ) |
---|
1117 | wall_top = IBITS( flags(k+1,j,i), 0, 1 ) |
---|
1118 | wall_total = wall_left + wall_right + wall_south + wall_north + & |
---|
1119 | wall_top |
---|
1120 | |
---|
1121 | IF ( wall_total > 0.0 ) THEN |
---|
1122 | p_mg(k,j,i) = 1.0 / wall_total * & |
---|
1123 | ( wall_left * p_mg(k,j,i-1) + & |
---|
1124 | wall_right * p_mg(k,j,i+1) + & |
---|
1125 | wall_south * p_mg(k,j-1,i) + & |
---|
1126 | wall_north * p_mg(k,j+1,i) + & |
---|
1127 | wall_top * p_mg(k+1,j,i) ) |
---|
1128 | ENDIF |
---|
1129 | ENDDO |
---|
1130 | ENDDO |
---|
1131 | ENDDO |
---|
1132 | !$OMP END PARALLEL |
---|
1133 | |
---|
1134 | ! |
---|
1135 | !-- One more time horizontal boundary conditions |
---|
1136 | CALL exchange_horiz( p_mg, 1) |
---|
1137 | |
---|
1138 | |
---|
1139 | END SUBROUTINE redblack |
---|
1140 | |
---|
1141 | |
---|
1142 | |
---|
1143 | SUBROUTINE mg_gather( f2, f2_sub ) |
---|
1144 | |
---|
1145 | USE control_parameters, & |
---|
1146 | ONLY: grid_level |
---|
1147 | |
---|
1148 | USE cpulog, & |
---|
1149 | ONLY: cpu_log, log_point_s |
---|
1150 | |
---|
1151 | USE indices, & |
---|
1152 | ONLY: mg_loc_ind, nxl_mg, nxr_mg, nys_mg, nyn_mg, nzb, nzt_mg |
---|
1153 | |
---|
1154 | USE kinds |
---|
1155 | |
---|
1156 | USE pegrid |
---|
1157 | |
---|
1158 | IMPLICIT NONE |
---|
1159 | |
---|
1160 | INTEGER(iwp) :: i !: |
---|
1161 | INTEGER(iwp) :: il !: |
---|
1162 | INTEGER(iwp) :: ir !: |
---|
1163 | INTEGER(iwp) :: j !: |
---|
1164 | INTEGER(iwp) :: jn !: |
---|
1165 | INTEGER(iwp) :: js !: |
---|
1166 | INTEGER(iwp) :: k !: |
---|
1167 | INTEGER(iwp) :: nwords !: |
---|
1168 | |
---|
1169 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
1170 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
1171 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: f2 !: |
---|
1172 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
1173 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
1174 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: f2_l !: |
---|
1175 | |
---|
1176 | REAL(wp), DIMENSION(nzb:mg_loc_ind(5,myid)+1, & |
---|
1177 | mg_loc_ind(3,myid)-1:mg_loc_ind(4,myid)+1, & |
---|
1178 | mg_loc_ind(1,myid)-1:mg_loc_ind(2,myid)+1) :: f2_sub !: |
---|
1179 | |
---|
1180 | |
---|
1181 | #if defined( __parallel ) |
---|
1182 | CALL cpu_log( log_point_s(34), 'mg_gather', 'start' ) |
---|
1183 | |
---|
1184 | f2_l = 0.0 |
---|
1185 | |
---|
1186 | ! |
---|
1187 | !-- Store the local subdomain array on the total array |
---|
1188 | js = mg_loc_ind(3,myid) |
---|
1189 | IF ( south_border_pe ) js = js - 1 |
---|
1190 | jn = mg_loc_ind(4,myid) |
---|
1191 | IF ( north_border_pe ) jn = jn + 1 |
---|
1192 | il = mg_loc_ind(1,myid) |
---|
1193 | IF ( left_border_pe ) il = il - 1 |
---|
1194 | ir = mg_loc_ind(2,myid) |
---|
1195 | IF ( right_border_pe ) ir = ir + 1 |
---|
1196 | DO i = il, ir |
---|
1197 | DO j = js, jn |
---|
1198 | DO k = nzb, nzt_mg(grid_level)+1 |
---|
1199 | f2_l(k,j,i) = f2_sub(k,j,i) |
---|
1200 | ENDDO |
---|
1201 | ENDDO |
---|
1202 | ENDDO |
---|
1203 | |
---|
1204 | ! |
---|
1205 | !-- Find out the number of array elements of the total array |
---|
1206 | nwords = SIZE( f2 ) |
---|
1207 | |
---|
1208 | ! |
---|
1209 | !-- Gather subdomain data from all PEs |
---|
1210 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
---|
1211 | CALL MPI_ALLREDUCE( f2_l(nzb,nys_mg(grid_level)-1,nxl_mg(grid_level)-1), & |
---|
1212 | f2(nzb,nys_mg(grid_level)-1,nxl_mg(grid_level)-1), & |
---|
1213 | nwords, MPI_REAL, MPI_SUM, comm2d, ierr ) |
---|
1214 | |
---|
1215 | CALL cpu_log( log_point_s(34), 'mg_gather', 'stop' ) |
---|
1216 | #endif |
---|
1217 | |
---|
1218 | END SUBROUTINE mg_gather |
---|
1219 | |
---|
1220 | |
---|
1221 | |
---|
1222 | SUBROUTINE mg_scatter( p2, p2_sub ) |
---|
1223 | ! |
---|
1224 | !-- TODO: It may be possible to improve the speed of this routine by using |
---|
1225 | !-- non-blocking communication |
---|
1226 | |
---|
1227 | USE control_parameters, & |
---|
1228 | ONLY: grid_level |
---|
1229 | |
---|
1230 | USE cpulog, & |
---|
1231 | ONLY: cpu_log, log_point_s |
---|
1232 | |
---|
1233 | USE indices, & |
---|
1234 | ONLY: mg_loc_ind, nxl_mg, nxr_mg, nys_mg, nyn_mg, nzb, nzt_mg |
---|
1235 | |
---|
1236 | USE kinds |
---|
1237 | |
---|
1238 | USE pegrid |
---|
1239 | |
---|
1240 | IMPLICIT NONE |
---|
1241 | |
---|
1242 | INTEGER(iwp) :: nwords !: |
---|
1243 | |
---|
1244 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level-1)+1, & |
---|
1245 | nys_mg(grid_level-1)-1:nyn_mg(grid_level-1)+1, & |
---|
1246 | nxl_mg(grid_level-1)-1:nxr_mg(grid_level-1)+1) :: p2 !: |
---|
1247 | |
---|
1248 | REAL(wp), DIMENSION(nzb:mg_loc_ind(5,myid)+1, & |
---|
1249 | mg_loc_ind(3,myid)-1:mg_loc_ind(4,myid)+1, & |
---|
1250 | mg_loc_ind(1,myid)-1:mg_loc_ind(2,myid)+1) :: p2_sub !: |
---|
1251 | |
---|
1252 | ! |
---|
1253 | !-- Find out the number of array elements of the subdomain array |
---|
1254 | nwords = SIZE( p2_sub ) |
---|
1255 | |
---|
1256 | #if defined( __parallel ) |
---|
1257 | CALL cpu_log( log_point_s(35), 'mg_scatter', 'start' ) |
---|
1258 | |
---|
1259 | p2_sub = p2(:,mg_loc_ind(3,myid)-1:mg_loc_ind(4,myid)+1, & |
---|
1260 | mg_loc_ind(1,myid)-1:mg_loc_ind(2,myid)+1) |
---|
1261 | |
---|
1262 | CALL cpu_log( log_point_s(35), 'mg_scatter', 'stop' ) |
---|
1263 | #endif |
---|
1264 | |
---|
1265 | END SUBROUTINE mg_scatter |
---|
1266 | |
---|
1267 | |
---|
1268 | |
---|
1269 | RECURSIVE SUBROUTINE next_mg_level( f_mg, p_mg, p3, r ) |
---|
1270 | |
---|
1271 | !------------------------------------------------------------------------------! |
---|
1272 | ! Description: |
---|
1273 | ! ------------ |
---|
1274 | ! This is where the multigrid technique takes place. V- and W- Cycle are |
---|
1275 | ! implemented and steered by the parameter "gamma". Parameter "nue" determines |
---|
1276 | ! the convergence of the multigrid iterative solution. There are nue times |
---|
1277 | ! RB-GS iterations. It should be set to "1" or "2", considering the time effort |
---|
1278 | ! one would like to invest. Last choice shows a very good converging factor, |
---|
1279 | ! but leads to an increase in computing time. |
---|
1280 | !------------------------------------------------------------------------------! |
---|
1281 | |
---|
1282 | USE control_parameters, & |
---|
1283 | ONLY: bc_lr_dirrad, bc_lr_raddir, bc_ns_dirrad, bc_ns_raddir, & |
---|
1284 | gamma_mg, grid_level, grid_level_count, ibc_p_b, ibc_p_t, & |
---|
1285 | inflow_l, inflow_n, inflow_r, inflow_s, maximum_grid_level, & |
---|
1286 | mg_switch_to_pe0_level, mg_switch_to_pe0, ngsrb, outflow_l, & |
---|
1287 | outflow_n, outflow_r, outflow_s |
---|
1288 | |
---|
1289 | |
---|
1290 | USE indices, & |
---|
1291 | ONLY: mg_loc_ind, nxl, nxl_mg, nxr, nxr_mg, nys, nys_mg, nyn, & |
---|
1292 | nyn_mg, nzb, nzt, nzt_mg |
---|
1293 | |
---|
1294 | USE kinds |
---|
1295 | |
---|
1296 | USE pegrid |
---|
1297 | |
---|
1298 | IMPLICIT NONE |
---|
1299 | |
---|
1300 | INTEGER(iwp) :: i !: |
---|
1301 | INTEGER(iwp) :: j !: |
---|
1302 | INTEGER(iwp) :: k !: |
---|
1303 | INTEGER(iwp) :: nxl_mg_save !: |
---|
1304 | INTEGER(iwp) :: nxr_mg_save !: |
---|
1305 | INTEGER(iwp) :: nyn_mg_save !: |
---|
1306 | INTEGER(iwp) :: nys_mg_save !: |
---|
1307 | INTEGER(iwp) :: nzt_mg_save !: |
---|
1308 | |
---|
1309 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
1310 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
1311 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: f_mg !: |
---|
1312 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
1313 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
1314 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: p_mg !: |
---|
1315 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
1316 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
1317 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: p3 !: |
---|
1318 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level)+1, & |
---|
1319 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
1320 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) :: r !: |
---|
1321 | |
---|
1322 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level-1)+1, & |
---|
1323 | nys_mg(grid_level-1)-1:nyn_mg(grid_level-1)+1, & |
---|
1324 | nxl_mg(grid_level-1)-1:nxr_mg(grid_level-1)+1) :: f2 !: |
---|
1325 | REAL(wp), DIMENSION(nzb:nzt_mg(grid_level-1)+1, & |
---|
1326 | nys_mg(grid_level-1)-1:nyn_mg(grid_level-1)+1, & |
---|
1327 | nxl_mg(grid_level-1)-1:nxr_mg(grid_level-1)+1) :: p2 !: |
---|
1328 | |
---|
1329 | REAL(wp), DIMENSION(:,:,:), ALLOCATABLE :: f2_sub !: |
---|
1330 | REAL(wp), DIMENSION(:,:,:), ALLOCATABLE :: p2_sub !: |
---|
1331 | |
---|
1332 | ! |
---|
1333 | !-- Restriction to the coarsest grid |
---|
1334 | 10 IF ( grid_level == 1 ) THEN |
---|
1335 | |
---|
1336 | ! |
---|
1337 | !-- Solution on the coarsest grid. Double the number of Gauss-Seidel |
---|
1338 | !-- iterations in order to get a more accurate solution. |
---|
1339 | ngsrb = 2 * ngsrb |
---|
1340 | |
---|
1341 | CALL redblack( f_mg, p_mg ) |
---|
1342 | |
---|
1343 | ngsrb = ngsrb / 2 |
---|
1344 | |
---|
1345 | |
---|
1346 | ELSEIF ( grid_level /= 1 ) THEN |
---|
1347 | |
---|
1348 | grid_level_count(grid_level) = grid_level_count(grid_level) + 1 |
---|
1349 | |
---|
1350 | ! |
---|
1351 | !-- Solution on the actual grid level |
---|
1352 | CALL redblack( f_mg, p_mg ) |
---|
1353 | |
---|
1354 | ! |
---|
1355 | !-- Determination of the actual residual |
---|
1356 | CALL resid( f_mg, p_mg, r ) |
---|
1357 | |
---|
1358 | ! |
---|
1359 | !-- Restriction of the residual (finer grid values!) to the next coarser |
---|
1360 | !-- grid. Therefore, the grid level has to be decremented now. nxl..nzt have |
---|
1361 | !-- to be set to the coarse grid values, because these variables are needed |
---|
1362 | !-- for the exchange of ghost points in routine exchange_horiz |
---|
1363 | grid_level = grid_level - 1 |
---|
1364 | nxl = nxl_mg(grid_level) |
---|
1365 | nys = nys_mg(grid_level) |
---|
1366 | nxr = nxr_mg(grid_level) |
---|
1367 | nyn = nyn_mg(grid_level) |
---|
1368 | nzt = nzt_mg(grid_level) |
---|
1369 | |
---|
1370 | IF ( grid_level == mg_switch_to_pe0_level ) THEN |
---|
1371 | |
---|
1372 | ! |
---|
1373 | !-- From this level on, calculations are done on PE0 only. |
---|
1374 | !-- First, carry out restriction on the subdomain. |
---|
1375 | !-- Therefore, indices of the level have to be changed to subdomain values |
---|
1376 | !-- in between (otherwise, the restrict routine would expect |
---|
1377 | !-- the gathered array) |
---|
1378 | |
---|
1379 | nxl_mg_save = nxl_mg(grid_level) |
---|
1380 | nxr_mg_save = nxr_mg(grid_level) |
---|
1381 | nys_mg_save = nys_mg(grid_level) |
---|
1382 | nyn_mg_save = nyn_mg(grid_level) |
---|
1383 | nzt_mg_save = nzt_mg(grid_level) |
---|
1384 | nxl_mg(grid_level) = mg_loc_ind(1,myid) |
---|
1385 | nxr_mg(grid_level) = mg_loc_ind(2,myid) |
---|
1386 | nys_mg(grid_level) = mg_loc_ind(3,myid) |
---|
1387 | nyn_mg(grid_level) = mg_loc_ind(4,myid) |
---|
1388 | nzt_mg(grid_level) = mg_loc_ind(5,myid) |
---|
1389 | nxl = mg_loc_ind(1,myid) |
---|
1390 | nxr = mg_loc_ind(2,myid) |
---|
1391 | nys = mg_loc_ind(3,myid) |
---|
1392 | nyn = mg_loc_ind(4,myid) |
---|
1393 | nzt = mg_loc_ind(5,myid) |
---|
1394 | |
---|
1395 | ALLOCATE( f2_sub(nzb:nzt_mg(grid_level)+1, & |
---|
1396 | nys_mg(grid_level)-1:nyn_mg(grid_level)+1, & |
---|
1397 | nxl_mg(grid_level)-1:nxr_mg(grid_level)+1) ) |
---|
1398 | |
---|
1399 | CALL restrict( f2_sub, r ) |
---|
1400 | |
---|
1401 | ! |
---|
1402 | !-- Restore the correct indices of this level |
---|
1403 | nxl_mg(grid_level) = nxl_mg_save |
---|
1404 | nxr_mg(grid_level) = nxr_mg_save |
---|
1405 | nys_mg(grid_level) = nys_mg_save |
---|
1406 | nyn_mg(grid_level) = nyn_mg_save |
---|
1407 | nzt_mg(grid_level) = nzt_mg_save |
---|
1408 | nxl = nxl_mg(grid_level) |
---|
1409 | nxr = nxr_mg(grid_level) |
---|
1410 | nys = nys_mg(grid_level) |
---|
1411 | nyn = nyn_mg(grid_level) |
---|
1412 | nzt = nzt_mg(grid_level) |
---|
1413 | ! |
---|
1414 | !-- Gather all arrays from the subdomains on PE0 |
---|
1415 | CALL mg_gather( f2, f2_sub ) |
---|
1416 | |
---|
1417 | ! |
---|
1418 | !-- Set switch for routine exchange_horiz, that no ghostpoint exchange |
---|
1419 | !-- has to be carried out from now on |
---|
1420 | mg_switch_to_pe0 = .TRUE. |
---|
1421 | |
---|
1422 | ! |
---|
1423 | !-- In case of non-cyclic lateral boundary conditions, both in- and |
---|
1424 | !-- outflow conditions have to be used on all PEs after the switch, |
---|
1425 | !-- because then they have the total domain. |
---|
1426 | IF ( bc_lr_dirrad ) THEN |
---|
1427 | inflow_l = .TRUE. |
---|
1428 | inflow_r = .FALSE. |
---|
1429 | outflow_l = .FALSE. |
---|
1430 | outflow_r = .TRUE. |
---|
1431 | ELSEIF ( bc_lr_raddir ) THEN |
---|
1432 | inflow_l = .FALSE. |
---|
1433 | inflow_r = .TRUE. |
---|
1434 | outflow_l = .TRUE. |
---|
1435 | outflow_r = .FALSE. |
---|
1436 | ENDIF |
---|
1437 | |
---|
1438 | IF ( bc_ns_dirrad ) THEN |
---|
1439 | inflow_n = .TRUE. |
---|
1440 | inflow_s = .FALSE. |
---|
1441 | outflow_n = .FALSE. |
---|
1442 | outflow_s = .TRUE. |
---|
1443 | ELSEIF ( bc_ns_raddir ) THEN |
---|
1444 | inflow_n = .FALSE. |
---|
1445 | inflow_s = .TRUE. |
---|
1446 | outflow_n = .TRUE. |
---|
1447 | outflow_s = .FALSE. |
---|
1448 | ENDIF |
---|
1449 | |
---|
1450 | DEALLOCATE( f2_sub ) |
---|
1451 | |
---|
1452 | ELSE |
---|
1453 | |
---|
1454 | CALL restrict( f2, r ) |
---|
1455 | |
---|
1456 | ENDIF |
---|
1457 | |
---|
1458 | p2 = 0.0 |
---|
1459 | |
---|
1460 | ! |
---|
1461 | !-- Repeat the same procedure till the coarsest grid is reached |
---|
1462 | CALL next_mg_level( f2, p2, p3, r ) |
---|
1463 | |
---|
1464 | ENDIF |
---|
1465 | |
---|
1466 | ! |
---|
1467 | !-- Now follows the prolongation |
---|
1468 | IF ( grid_level >= 2 ) THEN |
---|
1469 | |
---|
1470 | ! |
---|
1471 | !-- Prolongation of the new residual. The values are transferred |
---|
1472 | !-- from the coarse to the next finer grid. |
---|
1473 | IF ( grid_level == mg_switch_to_pe0_level+1 ) THEN |
---|
1474 | |
---|
1475 | #if defined( __parallel ) |
---|
1476 | ! |
---|
1477 | !-- At this level, the new residual first has to be scattered from |
---|
1478 | !-- PE0 to the other PEs |
---|
1479 | ALLOCATE( p2_sub(nzb:mg_loc_ind(5,myid)+1, & |
---|
1480 | mg_loc_ind(3,myid)-1:mg_loc_ind(4,myid)+1, & |
---|
1481 | mg_loc_ind(1,myid)-1:mg_loc_ind(2,myid)+1) ) |
---|
1482 | |
---|
1483 | CALL mg_scatter( p2, p2_sub ) |
---|
1484 | |
---|
1485 | ! |
---|
1486 | !-- Therefore, indices of the previous level have to be changed to |
---|
1487 | !-- subdomain values in between (otherwise, the prolong routine would |
---|
1488 | !-- expect the gathered array) |
---|
1489 | nxl_mg_save = nxl_mg(grid_level-1) |
---|
1490 | nxr_mg_save = nxr_mg(grid_level-1) |
---|
1491 | nys_mg_save = nys_mg(grid_level-1) |
---|
1492 | nyn_mg_save = nyn_mg(grid_level-1) |
---|
1493 | nzt_mg_save = nzt_mg(grid_level-1) |
---|
1494 | nxl_mg(grid_level-1) = mg_loc_ind(1,myid) |
---|
1495 | nxr_mg(grid_level-1) = mg_loc_ind(2,myid) |
---|
1496 | nys_mg(grid_level-1) = mg_loc_ind(3,myid) |
---|
1497 | nyn_mg(grid_level-1) = mg_loc_ind(4,myid) |
---|
1498 | nzt_mg(grid_level-1) = mg_loc_ind(5,myid) |
---|
1499 | |
---|
1500 | ! |
---|
1501 | !-- Set switch for routine exchange_horiz, that ghostpoint exchange |
---|
1502 | !-- has to be carried again out from now on |
---|
1503 | mg_switch_to_pe0 = .FALSE. |
---|
1504 | |
---|
1505 | ! |
---|
1506 | !-- For non-cyclic lateral boundary conditions, restore the |
---|
1507 | !-- in-/outflow conditions |
---|
1508 | inflow_l = .FALSE.; inflow_r = .FALSE. |
---|
1509 | inflow_n = .FALSE.; inflow_s = .FALSE. |
---|
1510 | outflow_l = .FALSE.; outflow_r = .FALSE. |
---|
1511 | outflow_n = .FALSE.; outflow_s = .FALSE. |
---|
1512 | |
---|
1513 | IF ( pleft == MPI_PROC_NULL ) THEN |
---|
1514 | IF ( bc_lr_dirrad ) THEN |
---|
1515 | inflow_l = .TRUE. |
---|
1516 | ELSEIF ( bc_lr_raddir ) THEN |
---|
1517 | outflow_l = .TRUE. |
---|
1518 | ENDIF |
---|
1519 | ENDIF |
---|
1520 | |
---|
1521 | IF ( pright == MPI_PROC_NULL ) THEN |
---|
1522 | IF ( bc_lr_dirrad ) THEN |
---|
1523 | outflow_r = .TRUE. |
---|
1524 | ELSEIF ( bc_lr_raddir ) THEN |
---|
1525 | inflow_r = .TRUE. |
---|
1526 | ENDIF |
---|
1527 | ENDIF |
---|
1528 | |
---|
1529 | IF ( psouth == MPI_PROC_NULL ) THEN |
---|
1530 | IF ( bc_ns_dirrad ) THEN |
---|
1531 | outflow_s = .TRUE. |
---|
1532 | ELSEIF ( bc_ns_raddir ) THEN |
---|
1533 | inflow_s = .TRUE. |
---|
1534 | ENDIF |
---|
1535 | ENDIF |
---|
1536 | |
---|
1537 | IF ( pnorth == MPI_PROC_NULL ) THEN |
---|
1538 | IF ( bc_ns_dirrad ) THEN |
---|
1539 | inflow_n = .TRUE. |
---|
1540 | ELSEIF ( bc_ns_raddir ) THEN |
---|
1541 | outflow_n = .TRUE. |
---|
1542 | ENDIF |
---|
1543 | ENDIF |
---|
1544 | |
---|
1545 | CALL prolong( p2_sub, p3 ) |
---|
1546 | |
---|
1547 | ! |
---|
1548 | !-- Restore the correct indices of the previous level |
---|
1549 | nxl_mg(grid_level-1) = nxl_mg_save |
---|
1550 | nxr_mg(grid_level-1) = nxr_mg_save |
---|
1551 | nys_mg(grid_level-1) = nys_mg_save |
---|
1552 | nyn_mg(grid_level-1) = nyn_mg_save |
---|
1553 | nzt_mg(grid_level-1) = nzt_mg_save |
---|
1554 | |
---|
1555 | DEALLOCATE( p2_sub ) |
---|
1556 | #endif |
---|
1557 | |
---|
1558 | ELSE |
---|
1559 | |
---|
1560 | CALL prolong( p2, p3 ) |
---|
1561 | |
---|
1562 | ENDIF |
---|
1563 | |
---|
1564 | ! |
---|
1565 | !-- Computation of the new pressure correction. Therefore, |
---|
1566 | !-- values from prior grids are added up automatically stage by stage. |
---|
1567 | DO i = nxl_mg(grid_level)-1, nxr_mg(grid_level)+1 |
---|
1568 | DO j = nys_mg(grid_level)-1, nyn_mg(grid_level)+1 |
---|
1569 | DO k = nzb, nzt_mg(grid_level)+1 |
---|
1570 | p_mg(k,j,i) = p_mg(k,j,i) + p3(k,j,i) |
---|
1571 | ENDDO |
---|
1572 | ENDDO |
---|
1573 | ENDDO |
---|
1574 | |
---|
1575 | ! |
---|
1576 | !-- Relaxation of the new solution |
---|
1577 | CALL redblack( f_mg, p_mg ) |
---|
1578 | |
---|
1579 | ENDIF |
---|
1580 | |
---|
1581 | |
---|
1582 | ! |
---|
1583 | !-- The following few lines serve the steering of the multigrid scheme |
---|
1584 | IF ( grid_level == maximum_grid_level ) THEN |
---|
1585 | |
---|
1586 | GOTO 20 |
---|
1587 | |
---|
1588 | ELSEIF ( grid_level /= maximum_grid_level .AND. grid_level /= 1 .AND. & |
---|
1589 | grid_level_count(grid_level) /= gamma_mg ) THEN |
---|
1590 | |
---|
1591 | GOTO 10 |
---|
1592 | |
---|
1593 | ENDIF |
---|
1594 | |
---|
1595 | ! |
---|
1596 | !-- Reset counter for the next call of poismg |
---|
1597 | grid_level_count(grid_level) = 0 |
---|
1598 | |
---|
1599 | ! |
---|
1600 | !-- Continue with the next finer level. nxl..nzt have to be |
---|
1601 | !-- set to the finer grid values, because these variables are needed for the |
---|
1602 | !-- exchange of ghost points in routine exchange_horiz |
---|
1603 | grid_level = grid_level + 1 |
---|
1604 | nxl = nxl_mg(grid_level) |
---|
1605 | nxr = nxr_mg(grid_level) |
---|
1606 | nys = nys_mg(grid_level) |
---|
1607 | nyn = nyn_mg(grid_level) |
---|
1608 | nzt = nzt_mg(grid_level) |
---|
1609 | |
---|
1610 | 20 CONTINUE |
---|
1611 | |
---|
1612 | END SUBROUTINE next_mg_level |
---|