Home

Context Navigation

source: palm/trunk/SOURCE/pres.f90 @ 1914

Last change on this file since 1914 was 1909, checked in by suehring, 8 years ago
last commit documented
Property svn:keywords set to `Id`
File size: 25.7 KB

Line
1	!> @file pres.f90
2	!--------------------------------------------------------------------------------!
3	! This file is part of PALM.
4	!
5	! PALM is free software: you can redistribute it and/or modify it under the terms
6	! of the GNU General Public License as published by the Free Software Foundation,
7	! either version 3 of the License, or (at your option) any later version.
8	!
9	! PALM is distributed in the hope that it will be useful, but WITHOUT ANY
10	! WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
11	! A PARTICULAR PURPOSE. See the GNU General Public License for more details.
12	!
13	! You should have received a copy of the GNU General Public License along with
14	! PALM. If not, see <http://www.gnu.org/licenses/>.
15	!
16	! Copyright 1997-2016 Leibniz Universitaet Hannover
17	!--------------------------------------------------------------------------------!
18	!
19	! Current revisions:
20	! ------------------
21	!
22	!
23	! Former revisions:
24	! -----------------
25	! $Id: pres.f90 1909 2016-05-25 17:23:33Z witha $
26	!
27	! 1908 2016-05-25 17:22:32Z suehring
28	! New divergence for output into RUN_CONTROL file is calculated only at last
29	! Runge-Kutta step
30	!
31	! 1845 2016-04-08 08:29:13Z raasch
32	! nzb_2d replace by nzb_u\|v_inner
33	!
34	! 1799 2016-04-05 08:35:55Z gronemeier
35	! Bugfix: excluded third dimension from horizontal volume flow calculation
36	!
37	! 1762 2016-02-25 12:31:13Z hellstea
38	! Introduction of nested domain feature
39	!
40	! 1682 2015-10-07 23:56:08Z knoop
41	! Code annotations made doxygen readable
42	!
43	! 1575 2015-03-27 09:56:27Z raasch
44	! poismg_fast + respective module added, adjustments for psolver-queries
45	!
46	! 1342 2014-03-26 17:04:47Z kanani
47	! REAL constants defined as wp-kind
48	!
49	! 1320 2014-03-20 08:40:49Z raasch
50	! ONLY-attribute added to USE-statements,
51	! kind-parameters added to all INTEGER and REAL declaration statements,
52	! kinds are defined in new module kinds,
53	! old module precision_kind is removed,
54	! revision history before 2012 removed,
55	! comment fields (!:) to be used for variable explanations added to
56	! all variable declaration statements
57	!
58	! 1318 2014-03-17 13:35:16Z raasch
59	! module interfaces removed
60	!
61	! 1306 2014-03-13 14:30:59Z raasch
62	! second argument removed from routine poisfft
63	!
64	! 1257 2013-11-08 15:18:40Z raasch
65	! openacc loop and loop vector clauses removed, independent clauses added,
66	! end parallel replaced by end parallel loop
67	!
68	! 1221 2013-09-10 08:59:13Z raasch
69	! openACC porting of reduction operations, loops for calculating d are
70	! using the rflags_s_inner multiply flag instead of the nzb_s_inner loop index
71	!
72	! 1212 2013-08-15 08:46:27Z raasch
73	! call of poisfft_hybrid removed
74	!
75	! 1117 2013-03-27 11:15:36Z suehring
76	! Bugfix in OpenMP parallelization.
77	!
78	! 1113 2013-03-10 02:48:14Z raasch
79	! GPU-porting of several loops, some loops rearranged
80	!
81	! 1111 2013-03-08 23:54:10Z
82	! openACC statements added,
83	! ibc_p_b = 2 removed
84	!
85	! 1092 2013-02-02 11:24:22Z raasch
86	! unused variables removed
87	!
88	! 1036 2012-10-22 13:43:42Z raasch
89	! code put under GPL (PALM 3.9)
90	!
91	! 1003 2012-09-14 14:35:53Z raasch
92	! adjustment of array tend for cases with unequal subdomain sizes removed
93	!
94	! Revision 1.1 1997/07/24 11:24:44 raasch
95	! Initial revision
96	!
97	!
98	! Description:
99	! ------------
100	!> Compute the divergence of the provisional velocity field. Solve the Poisson
101	!> equation for the perturbation pressure. Compute the final velocities using
102	!> this perturbation pressure. Compute the remaining divergence.
103	!------------------------------------------------------------------------------!
104	SUBROUTINE pres
105
106
107	USE arrays_3d, &
108	ONLY: d, ddzu, ddzu_pres, ddzw, dzw, p, p_loc, tend, u, v, w
109
110	USE control_parameters, &
111	ONLY: bc_lr_cyc, bc_ns_cyc, conserve_volume_flow, dt_3d, &
112	gathered_size, ibc_p_b, ibc_p_t, intermediate_timestep_count, &
113	intermediate_timestep_count_max, mg_switch_to_pe0_level, &
114	nest_domain, nest_bound_l, nest_bound_n, nest_bound_r, &
115	nest_bound_s, on_device, outflow_l, outflow_n, outflow_r, &
116	outflow_s, psolver, simulated_time, subdomain_size, topography, &
117	volume_flow, volume_flow_area, volume_flow_initial
118
119	USE cpulog, &
120	ONLY: cpu_log, log_point, log_point_s
121
122	USE grid_variables, &
123	ONLY: ddx, ddy
124
125	USE indices, &
126	ONLY: nbgp, ngp_2dh_outer, nx, nxl, nxlg, nxl_mg, nxr, nxrg, nxr_mg, &
127	ny, nys, nysg, nys_mg, nyn, nyng, nyn_mg, nzb, nzb_s_inner, &
128	nzb_u_inner, nzb_v_inner, nzb_w_inner, nzt, nzt_mg, &
129	rflags_s_inner
130
131	USE kinds
132
133	USE pegrid
134
135	USE poisfft_mod, &
136	ONLY: poisfft
137
138	USE poismg_mod
139
140	USE statistics, &
141	ONLY: statistic_regions, sums_divnew_l, sums_divold_l, weight_pres, &
142	weight_substep
143
144	IMPLICIT NONE
145
146	INTEGER(iwp) :: i !<
147	INTEGER(iwp) :: j !<
148	INTEGER(iwp) :: k !<
149
150	REAL(wp) :: ddt_3d !<
151	REAL(wp) :: localsum !<
152	REAL(wp) :: threadsum !<
153	REAL(wp) :: d_weight_pres !<
154
155	REAL(wp), DIMENSION(1:3) :: volume_flow_l !<
156	REAL(wp), DIMENSION(1:3) :: volume_flow_offset !<
157	REAL(wp), DIMENSION(1:nzt) :: w_l !<
158	REAL(wp), DIMENSION(1:nzt) :: w_l_l !<
159
160
161	CALL cpu_log( log_point(8), 'pres', 'start' )
162
163
164	ddt_3d = 1.0_wp / dt_3d
165	d_weight_pres = 1.0_wp / weight_pres(intermediate_timestep_count)
166
167	!
168	!-- Multigrid method expects array d to have one ghost layer.
169	!--
170	IF ( psolver(1:9) == 'multigrid' ) THEN
171
172	DEALLOCATE( d )
173	ALLOCATE( d(nzb:nzt+1,nys-1:nyn+1,nxl-1:nxr+1) )
174
175	!
176	!-- Since p is later used to hold the weighted average of the substeps, it
177	!-- cannot be used in the iterative solver. Therefore, its initial value is
178	!-- stored on p_loc, which is then iteratively advanced in every substep.
179	IF ( intermediate_timestep_count <= 1 ) THEN
180	DO i = nxl-1, nxr+1
181	DO j = nys-1, nyn+1
182	DO k = nzb, nzt+1
183	p_loc(k,j,i) = p(k,j,i)
184	ENDDO
185	ENDDO
186	ENDDO
187	ENDIF
188
189	ELSEIF ( psolver == 'sor' .AND. intermediate_timestep_count == 1 ) THEN
190
191	!
192	!-- Since p is later used to hold the weighted average of the substeps, it
193	!-- cannot be used in the iterative solver. Therefore, its initial value is
194	!-- stored on p_loc, which is then iteratively advanced in every substep.
195	p_loc = p
196
197	ENDIF
198
199	!
200	!-- Conserve the volume flow at the outflow in case of non-cyclic lateral
201	!-- boundary conditions
202	!-- WARNING: so far, this conservation does not work at the left/south
203	!-- boundary if the topography at the inflow differs from that at the
204	!-- outflow! For this case, volume_flow_area needs adjustment!
205	!
206	!-- Left/right
207	IF ( conserve_volume_flow .AND. ( outflow_l .OR. outflow_r ) ) THEN
208
209	volume_flow(1) = 0.0_wp
210	volume_flow_l(1) = 0.0_wp
211
212	IF ( outflow_l ) THEN
213	i = 0
214	ELSEIF ( outflow_r ) THEN
215	i = nx+1
216	ENDIF
217
218	DO j = nys, nyn
219	!
220	!-- Sum up the volume flow through the south/north boundary
221	DO k = nzb_u_inner(j,i)+1, nzt
222	volume_flow_l(1) = volume_flow_l(1) + u(k,j,i) * dzw(k)
223	ENDDO
224	ENDDO
225
226	#if defined( __parallel )
227	IF ( collective_wait ) CALL MPI_BARRIER( comm1dy, ierr )
228	CALL MPI_ALLREDUCE( volume_flow_l(1), volume_flow(1), 1, MPI_REAL, &
229	MPI_SUM, comm1dy, ierr )
230	#else
231	volume_flow = volume_flow_l
232	#endif
233	volume_flow_offset(1) = ( volume_flow_initial(1) - volume_flow(1) ) &
234	/ volume_flow_area(1)
235
236	DO j = nysg, nyng
237	DO k = nzb_u_inner(j,i)+1, nzt
238	u(k,j,i) = u(k,j,i) + volume_flow_offset(1)
239	ENDDO
240	ENDDO
241
242	ENDIF
243
244	!
245	!-- South/north
246	IF ( conserve_volume_flow .AND. ( outflow_n .OR. outflow_s ) ) THEN
247
248	volume_flow(2) = 0.0_wp
249	volume_flow_l(2) = 0.0_wp
250
251	IF ( outflow_s ) THEN
252	j = 0
253	ELSEIF ( outflow_n ) THEN
254	j = ny+1
255	ENDIF
256
257	DO i = nxl, nxr
258	!
259	!-- Sum up the volume flow through the south/north boundary
260	DO k = nzb_v_inner(j,i)+1, nzt
261	volume_flow_l(2) = volume_flow_l(2) + v(k,j,i) * dzw(k)
262	ENDDO
263	ENDDO
264
265	#if defined( __parallel )
266	IF ( collective_wait ) CALL MPI_BARRIER( comm1dx, ierr )
267	CALL MPI_ALLREDUCE( volume_flow_l(2), volume_flow(2), 1, MPI_REAL, &
268	MPI_SUM, comm1dx, ierr )
269	#else
270	volume_flow = volume_flow_l
271	#endif
272	volume_flow_offset(2) = ( volume_flow_initial(2) - volume_flow(2) ) &
273	/ volume_flow_area(2)
274
275	DO i = nxlg, nxrg
276	DO k = nzb_v_inner(j,i)+1, nzt
277	v(k,j,i) = v(k,j,i) + volume_flow_offset(2)
278	ENDDO
279	ENDDO
280
281	ENDIF
282
283	!
284	!-- Remove mean vertical velocity in case that Neumann conditions are
285	!-- used both at bottom and top boundary, and if not a nested domain
286	IF ( ibc_p_b == 1 .AND. ibc_p_t == 1 .AND. .NOT. nest_domain) THEN
287	IF ( simulated_time > 0.0_wp ) THEN ! otherwise nzb_w_inner not yet known
288	w_l = 0.0_wp; w_l_l = 0.0_wp
289	DO i = nxl, nxr
290	DO j = nys, nyn
291	DO k = nzb_w_inner(j,i)+1, nzt
292	w_l_l(k) = w_l_l(k) + w(k,j,i)
293	ENDDO
294	ENDDO
295	ENDDO
296	#if defined( __parallel )
297	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
298	CALL MPI_ALLREDUCE( w_l_l(1), w_l(1), nzt, MPI_REAL, MPI_SUM, &
299	comm2d, ierr )
300	#else
301	w_l = w_l_l
302	#endif
303	DO k = 1, nzt
304	w_l(k) = w_l(k) / ngp_2dh_outer(k,0)
305	ENDDO
306	DO i = nxlg, nxrg
307	DO j = nysg, nyng
308	DO k = nzb_w_inner(j,i)+1, nzt
309	w(k,j,i) = w(k,j,i) - w_l(k)
310	ENDDO
311	ENDDO
312	ENDDO
313	ENDIF
314	ENDIF
315
316	!
317	!-- Compute the divergence of the provisional velocity field.
318	CALL cpu_log( log_point_s(1), 'divergence', 'start' )
319
320	IF ( psolver(1:9) == 'multigrid' ) THEN
321	!$OMP PARALLEL DO SCHEDULE( STATIC )
322	DO i = nxl-1, nxr+1
323	DO j = nys-1, nyn+1
324	DO k = nzb, nzt+1
325	d(k,j,i) = 0.0_wp
326	ENDDO
327	ENDDO
328	ENDDO
329	ELSE
330	!$OMP PARALLEL DO SCHEDULE( STATIC )
331	!$acc kernels present( d )
332	DO i = nxl, nxr
333	DO j = nys, nyn
334	DO k = nzb+1, nzt
335	d(k,j,i) = 0.0_wp
336	ENDDO
337	ENDDO
338	ENDDO
339	!$acc end kernels
340	ENDIF
341
342	localsum = 0.0_wp
343	threadsum = 0.0_wp
344
345	#if defined( __ibm )
346	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
347	!$OMP DO SCHEDULE( STATIC )
348	DO i = nxl, nxr
349	DO j = nys, nyn
350	DO k = nzb_s_inner(j,i)+1, nzt
351	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
352	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
353	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k) ) * ddt_3d &
354	* d_weight_pres
355	ENDDO
356	!
357	!-- Compute possible PE-sum of divergences for flow_statistics
358	DO k = nzb_s_inner(j,i)+1, nzt
359	threadsum = threadsum + ABS( d(k,j,i) )
360	ENDDO
361
362	ENDDO
363	ENDDO
364
365	IF ( intermediate_timestep_count == intermediate_timestep_count_max ) THEN
366	localsum = localsum + threadsum * dt_3d * &
367	weight_pres(intermediate_timestep_count)
368	ENDIF
369	!$OMP END PARALLEL
370	#else
371
372	!$OMP PARALLEL PRIVATE (i,j,k)
373	!$OMP DO SCHEDULE( STATIC )
374	!$acc kernels present( d, ddzw, rflags_s_inner, u, v, w )
375	!$acc loop collapse( 3 )
376	DO i = nxl, nxr
377	DO j = nys, nyn
378	DO k = 1, nzt
379	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
380	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
381	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k) ) * ddt_3d &
382	* d_weight_pres * rflags_s_inner(k,j,i)
383	ENDDO
384	ENDDO
385	ENDDO
386	!$acc end kernels
387	!$OMP END PARALLEL
388
389	!
390	!-- Compute possible PE-sum of divergences for flow_statistics. Carry out
391	!-- computation only at last Runge-Kutta substep.
392	IF ( intermediate_timestep_count == intermediate_timestep_count_max ) THEN
393	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
394	!$OMP DO SCHEDULE( STATIC )
395	!$acc parallel loop collapse(3) present( d ) reduction(+:threadsum)
396	DO i = nxl, nxr
397	DO j = nys, nyn
398	DO k = nzb+1, nzt
399	threadsum = threadsum + ABS( d(k,j,i) )
400	ENDDO
401	ENDDO
402	ENDDO
403	!$acc end parallel loop
404	localsum = localsum + threadsum * dt_3d * &
405	weight_pres(intermediate_timestep_count)
406	!$OMP END PARALLEL
407	ENDIF
408	#endif
409
410	!
411	!-- For completeness, set the divergence sum of all statistic regions to those
412	!-- of the total domain
413	IF ( intermediate_timestep_count == intermediate_timestep_count_max ) &
414	sums_divold_l(0:statistic_regions) = localsum
415
416	CALL cpu_log( log_point_s(1), 'divergence', 'stop' )
417
418	!
419	!-- Compute the pressure perturbation solving the Poisson equation
420	IF ( psolver == 'poisfft' ) THEN
421
422	!
423	!-- Solve Poisson equation via FFT and solution of tridiagonal matrices
424	CALL poisfft( d )
425
426	!
427	!-- Store computed perturbation pressure and set boundary condition in
428	!-- z-direction
429	!$OMP PARALLEL DO
430	!$acc kernels present( d, tend )
431	DO i = nxl, nxr
432	DO j = nys, nyn
433	DO k = nzb+1, nzt
434	tend(k,j,i) = d(k,j,i)
435	ENDDO
436	ENDDO
437	ENDDO
438	!$acc end kernels
439
440	!
441	!-- Bottom boundary:
442	!-- This condition is only required for internal output. The pressure
443	!-- gradient (dp(nzb+1)-dp(nzb))/dz is not used anywhere else.
444	IF ( ibc_p_b == 1 ) THEN
445	!
446	!-- Neumann (dp/dz = 0)
447	!$OMP PARALLEL DO
448	!$acc kernels present( nzb_s_inner, tend )
449	DO i = nxlg, nxrg
450	DO j = nysg, nyng
451	tend(nzb_s_inner(j,i),j,i) = tend(nzb_s_inner(j,i)+1,j,i)
452	ENDDO
453	ENDDO
454	!$acc end kernels
455
456	ELSE
457	!
458	!-- Dirichlet
459	!$OMP PARALLEL DO
460	!$acc kernels present( tend )
461	DO i = nxlg, nxrg
462	DO j = nysg, nyng
463	tend(nzb_s_inner(j,i),j,i) = 0.0_wp
464	ENDDO
465	ENDDO
466	!$acc end kernels
467
468	ENDIF
469
470	!
471	!-- Top boundary
472	IF ( ibc_p_t == 1 ) THEN
473	!
474	!-- Neumann
475	!$OMP PARALLEL DO
476	!$acc kernels present( tend )
477	DO i = nxlg, nxrg
478	DO j = nysg, nyng
479	tend(nzt+1,j,i) = tend(nzt,j,i)
480	ENDDO
481	ENDDO
482	!$acc end kernels
483
484	ELSE
485	!
486	!-- Dirichlet
487	!$OMP PARALLEL DO
488	!$acc kernels present( tend )
489	DO i = nxlg, nxrg
490	DO j = nysg, nyng
491	tend(nzt+1,j,i) = 0.0_wp
492	ENDDO
493	ENDDO
494	!$acc end kernels
495
496	ENDIF
497
498	!
499	!-- Exchange boundaries for p
500	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
501	on_device = .TRUE. ! to be removed after complete porting
502	ELSE ! of ghost point exchange
503	!$acc update host( tend )
504	ENDIF
505	CALL exchange_horiz( tend, nbgp )
506	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
507	on_device = .FALSE. ! to be removed after complete porting
508	ELSE ! of ghost point exchange
509	!$acc update device( tend )
510	ENDIF
511
512	ELSEIF ( psolver == 'sor' ) THEN
513
514	!
515	!-- Solve Poisson equation for perturbation pressure using SOR-Red/Black
516	!-- scheme
517	CALL sor( d, ddzu_pres, ddzw, p_loc )
518	tend = p_loc
519
520	ELSEIF ( psolver(1:9) == 'multigrid' ) THEN
521
522	!
523	!-- Solve Poisson equation for perturbation pressure using Multigrid scheme,
524	!-- array tend is used to store the residuals, logical exchange_mg is used
525	!-- to discern data exchange in multigrid ( 1 ghostpoint ) and normal grid
526	!-- ( nbgp ghost points ).
527
528	!-- If the number of grid points of the gathered grid, which is collected
529	!-- on PE0, is larger than the number of grid points of an PE, than array
530	!-- tend will be enlarged.
531	IF ( gathered_size > subdomain_size ) THEN
532	DEALLOCATE( tend )
533	ALLOCATE( tend(nzb:nzt_mg(mg_switch_to_pe0_level)+1,nys_mg( &
534	mg_switch_to_pe0_level)-1:nyn_mg(mg_switch_to_pe0_level)+1,&
535	nxl_mg(mg_switch_to_pe0_level)-1:nxr_mg( &
536	mg_switch_to_pe0_level)+1) )
537	ENDIF
538
539	IF ( psolver == 'multigrid' ) THEN
540	CALL poismg( tend )
541	ELSE
542	CALL poismg_fast( tend )
543	ENDIF
544
545	IF ( gathered_size > subdomain_size ) THEN
546	DEALLOCATE( tend )
547	ALLOCATE( tend(nzb:nzt+1,nysg:nyng,nxlg:nxrg) )
548	ENDIF
549
550	!
551	!-- Restore perturbation pressure on tend because this array is used
552	!-- further below to correct the velocity fields
553	DO i = nxl-1, nxr+1
554	DO j = nys-1, nyn+1
555	DO k = nzb, nzt+1
556	tend(k,j,i) = p_loc(k,j,i)
557	ENDDO
558	ENDDO
559	ENDDO
560
561	ENDIF
562
563	!
564	!-- Store perturbation pressure on array p, used for pressure data output.
565	!-- Ghost layers are added in the output routines (except sor-method: see below)
566	IF ( intermediate_timestep_count == 1 ) THEN
567	!$OMP PARALLEL PRIVATE (i,j,k)
568	!$OMP DO
569	!$acc kernels present( p, tend, weight_substep )
570	!$acc loop independent
571	DO i = nxl-1, nxr+1
572	!$acc loop independent
573	DO j = nys-1, nyn+1
574	!$acc loop independent
575	DO k = nzb, nzt+1
576	p(k,j,i) = tend(k,j,i) * &
577	weight_substep(intermediate_timestep_count)
578	ENDDO
579	ENDDO
580	ENDDO
581	!$acc end kernels
582	!$OMP END PARALLEL
583
584	ELSEIF ( intermediate_timestep_count > 1 ) THEN
585	!$OMP PARALLEL PRIVATE (i,j,k)
586	!$OMP DO
587	!$acc kernels present( p, tend, weight_substep )
588	!$acc loop independent
589	DO i = nxl-1, nxr+1
590	!$acc loop independent
591	DO j = nys-1, nyn+1
592	!$acc loop independent
593	DO k = nzb, nzt+1
594	p(k,j,i) = p(k,j,i) + tend(k,j,i) * &
595	weight_substep(intermediate_timestep_count)
596	ENDDO
597	ENDDO
598	ENDDO
599	!$acc end kernels
600	!$OMP END PARALLEL
601
602	ENDIF
603
604	!
605	!-- SOR-method needs ghost layers for the next timestep
606	IF ( psolver == 'sor' ) CALL exchange_horiz( p, nbgp )
607
608	!
609	!-- Correction of the provisional velocities with the current perturbation
610	!-- pressure just computed
611	IF ( conserve_volume_flow .AND. ( bc_lr_cyc .OR. bc_ns_cyc ) ) THEN
612	volume_flow_l(1) = 0.0_wp
613	volume_flow_l(2) = 0.0_wp
614	ENDIF
615
616	!$OMP PARALLEL PRIVATE (i,j,k)
617	!$OMP DO
618	!$acc kernels present( ddzu, nzb_u_inner, nzb_v_inner, nzb_w_inner, tend, u, v, w, weight_pres )
619	!$acc loop independent
620	DO i = nxl, nxr
621	!$acc loop independent
622	DO j = nys, nyn
623	!$acc loop independent
624	DO k = 1, nzt
625	IF ( k > nzb_w_inner(j,i) ) THEN
626	w(k,j,i) = w(k,j,i) - dt_3d * &
627	( tend(k+1,j,i) - tend(k,j,i) ) * ddzu(k+1) * &
628	weight_pres(intermediate_timestep_count)
629	ENDIF
630	ENDDO
631	!$acc loop independent
632	DO k = 1, nzt
633	IF ( k > nzb_u_inner(j,i) ) THEN
634	u(k,j,i) = u(k,j,i) - dt_3d * &
635	( tend(k,j,i) - tend(k,j,i-1) ) * ddx * &
636	weight_pres(intermediate_timestep_count)
637	ENDIF
638	ENDDO
639	!$acc loop independent
640	DO k = 1, nzt
641	IF ( k > nzb_v_inner(j,i) ) THEN
642	v(k,j,i) = v(k,j,i) - dt_3d * &
643	( tend(k,j,i) - tend(k,j-1,i) ) * ddy * &
644	weight_pres(intermediate_timestep_count)
645	ENDIF
646	ENDDO
647
648	ENDDO
649	ENDDO
650	!$acc end kernels
651	!$OMP END PARALLEL
652
653	!
654	!-- Sum up the volume flow through the right and north boundary
655	IF ( conserve_volume_flow .AND. bc_lr_cyc .AND. bc_ns_cyc .AND. &
656	nxr == nx ) THEN
657
658	!$OMP PARALLEL PRIVATE (j,k)
659	!$OMP DO
660	DO j = nys, nyn
661	!$OMP CRITICAL
662	DO k = nzb_u_inner(j,nx) + 1, nzt
663	volume_flow_l(1) = volume_flow_l(1) + u(k,j,nx) * dzw(k)
664	ENDDO
665	!$OMP END CRITICAL
666	ENDDO
667	!$OMP END PARALLEL
668
669	ENDIF
670
671	IF ( conserve_volume_flow .AND. bc_ns_cyc .AND. bc_lr_cyc .AND. &
672	nyn == ny ) THEN
673
674	!$OMP PARALLEL PRIVATE (i,k)
675	!$OMP DO
676	DO i = nxl, nxr
677	!$OMP CRITICAL
678	DO k = nzb_v_inner(ny,i) + 1, nzt
679	volume_flow_l(2) = volume_flow_l(2) + v(k,ny,i) * dzw(k)
680	ENDDO
681	!$OMP END CRITICAL
682	ENDDO
683	!$OMP END PARALLEL
684
685	ENDIF
686
687	!
688	!-- Conserve the volume flow
689	IF ( conserve_volume_flow .AND. ( bc_lr_cyc .AND. bc_ns_cyc ) ) THEN
690
691	#if defined( __parallel )
692	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
693	CALL MPI_ALLREDUCE( volume_flow_l(1), volume_flow(1), 2, MPI_REAL, &
694	MPI_SUM, comm2d, ierr )
695	#else
696	volume_flow = volume_flow_l
697	#endif
698
699	volume_flow_offset(1:2) = ( volume_flow_initial(1:2) - volume_flow(1:2) ) / &
700	volume_flow_area(1:2)
701
702	!$OMP PARALLEL PRIVATE (i,j,k)
703	!$OMP DO
704	DO i = nxl, nxr
705	DO j = nys, nyn
706	DO k = nzb_u_inner(j,i) + 1, nzt
707	u(k,j,i) = u(k,j,i) + volume_flow_offset(1)
708	ENDDO
709	DO k = nzb_v_inner(j,i) + 1, nzt
710	v(k,j,i) = v(k,j,i) + volume_flow_offset(2)
711	ENDDO
712	ENDDO
713	ENDDO
714
715	!$OMP END PARALLEL
716
717	ENDIF
718
719	!
720	!-- Exchange of boundaries for the velocities
721	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
722	on_device = .TRUE. ! to be removed after complete porting
723	ELSE ! of ghost point exchange
724	!$acc update host( u, v, w )
725	ENDIF
726	CALL exchange_horiz( u, nbgp )
727	CALL exchange_horiz( v, nbgp )
728	CALL exchange_horiz( w, nbgp )
729	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
730	on_device = .FALSE. ! to be removed after complete porting
731	ELSE ! of ghost point exchange
732	!$acc update device( u, v, w )
733	ENDIF
734
735	!
736	!-- Compute the divergence of the corrected velocity field,
737	!-- A possible PE-sum is computed in flow_statistics. Carry out computation
738	!-- only at last Runge-Kutta step.
739	IF ( intermediate_timestep_count == intermediate_timestep_count_max ) THEN
740	CALL cpu_log( log_point_s(1), 'divergence', 'start' )
741	sums_divnew_l = 0.0_wp
742
743	!
744	!-- d must be reset to zero because it can contain nonzero values below the
745	!-- topography
746	IF ( topography /= 'flat' ) d = 0.0_wp
747
748	localsum = 0.0_wp
749	threadsum = 0.0_wp
750
751	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
752	!$OMP DO SCHEDULE( STATIC )
753	#if defined( __ibm )
754	DO i = nxl, nxr
755	DO j = nys, nyn
756	DO k = nzb_s_inner(j,i)+1, nzt
757	d(k,j,i) = ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
758	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
759	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k)
760	ENDDO
761	DO k = nzb+1, nzt
762	threadsum = threadsum + ABS( d(k,j,i) )
763	ENDDO
764	ENDDO
765	ENDDO
766	#else
767	!$acc kernels present( d, ddzw, rflags_s_inner, u, v, w )
768	!$acc loop collapse( 3 )
769	DO i = nxl, nxr
770	DO j = nys, nyn
771	DO k = 1, nzt
772	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
773	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
774	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k) &
775	) * rflags_s_inner(k,j,i)
776	ENDDO
777	ENDDO
778	ENDDO
779	!$acc end kernels
780	!
781	!-- Compute possible PE-sum of divergences for flow_statistics
782	!$acc parallel loop collapse(3) present( d ) reduction(+:threadsum)
783	DO i = nxl, nxr
784	DO j = nys, nyn
785	DO k = nzb+1, nzt
786	threadsum = threadsum + ABS( d(k,j,i) )
787	ENDDO
788	ENDDO
789	ENDDO
790	!$acc end parallel loop
791	#endif
792
793	localsum = localsum + threadsum
794	!$OMP END PARALLEL
795
796	!
797	!-- For completeness, set the divergence sum of all statistic regions to those
798	!-- of the total domain
799	sums_divnew_l(0:statistic_regions) = localsum
800
801	CALL cpu_log( log_point_s(1), 'divergence', 'stop' )
802
803	ENDIF
804
805	CALL cpu_log( log_point(8), 'pres', 'stop' )
806
807
808	END SUBROUTINE pres

Note: See TracBrowser for help on using the repository browser.

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |