Home

Context Navigation

source: palm/trunk/SOURCE/pres.f90 @ 2042

Last change on this file since 2042 was 2038, checked in by knoop, 8 years ago
last commit documented
Property svn:keywords set to `Id`
File size: 27.9 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
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 1997-2016 Leibniz Universitaet Hannover
18	!------------------------------------------------------------------------------!
19	!
20	! Current revisions:
21	! ------------------
22	!
23	!
24	! Former revisions:
25	! -----------------
26	! $Id: pres.f90 2038 2016-10-26 11:16:56Z suehring $
27	!
28	! 2037 2016-10-26 11:15:40Z knoop
29	! Anelastic approximation implemented
30	!
31	! 2000 2016-08-20 18:09:15Z knoop
32	! Forced header and separation lines into 80 columns
33	!
34	! 1932 2016-06-10 12:09:21Z suehring
35	! Initial version of purely vertical nesting introduced.
36	!
37	! 1931 2016-06-10 12:06:59Z suehring
38	! Rename multigrid into multigrid_noopt and multigrid_fast into multigrid
39	!
40	! 1929 2016-06-09 16:25:25Z suehring
41	! Bugfix: weight_substep for initial call, replace by local variable
42	!
43	! 1918 2016-05-27 14:35:57Z raasch
44	! sum of divergence is also calculated when pres is called before the initial
45	! first time step,
46	! stearing is modified by using intermediate_timestep_count = 0 in order to
47	! determine if pres is called before the first initial timestep,
48	! bugfix: in case of Neumann conditions for pressure both at bottom and top,
49	! mean vertical velocity is also removed for the first time step
50	! bugfix for calculating divergences
51	!
52	! 1908 2016-05-25 17:22:32Z suehring
53	! New divergence for output into RUN_CONTROL file is calculated only at last
54	! Runge-Kutta step
55	!
56	! 1845 2016-04-08 08:29:13Z raasch
57	! nzb_2d replace by nzb_u\|v_inner
58	!
59	! 1799 2016-04-05 08:35:55Z gronemeier
60	! Bugfix: excluded third dimension from horizontal volume flow calculation
61	!
62	! 1762 2016-02-25 12:31:13Z hellstea
63	! Introduction of nested domain feature
64	!
65	! 1682 2015-10-07 23:56:08Z knoop
66	! Code annotations made doxygen readable
67	!
68	! 1575 2015-03-27 09:56:27Z raasch
69	! poismg_fast + respective module added, adjustments for psolver-queries
70	!
71	! 1342 2014-03-26 17:04:47Z kanani
72	! REAL constants defined as wp-kind
73	!
74	! 1320 2014-03-20 08:40:49Z raasch
75	! ONLY-attribute added to USE-statements,
76	! kind-parameters added to all INTEGER and REAL declaration statements,
77	! kinds are defined in new module kinds,
78	! old module precision_kind is removed,
79	! revision history before 2012 removed,
80	! comment fields (!:) to be used for variable explanations added to
81	! all variable declaration statements
82	!
83	! 1318 2014-03-17 13:35:16Z raasch
84	! module interfaces removed
85	!
86	! 1306 2014-03-13 14:30:59Z raasch
87	! second argument removed from routine poisfft
88	!
89	! 1257 2013-11-08 15:18:40Z raasch
90	! openacc loop and loop vector clauses removed, independent clauses added,
91	! end parallel replaced by end parallel loop
92	!
93	! 1221 2013-09-10 08:59:13Z raasch
94	! openACC porting of reduction operations, loops for calculating d are
95	! using the rflags_s_inner multiply flag instead of the nzb_s_inner loop index
96	!
97	! 1212 2013-08-15 08:46:27Z raasch
98	! call of poisfft_hybrid removed
99	!
100	! 1117 2013-03-27 11:15:36Z suehring
101	! Bugfix in OpenMP parallelization.
102	!
103	! 1113 2013-03-10 02:48:14Z raasch
104	! GPU-porting of several loops, some loops rearranged
105	!
106	! 1111 2013-03-08 23:54:10Z
107	! openACC statements added,
108	! ibc_p_b = 2 removed
109	!
110	! 1092 2013-02-02 11:24:22Z raasch
111	! unused variables removed
112	!
113	! 1036 2012-10-22 13:43:42Z raasch
114	! code put under GPL (PALM 3.9)
115	!
116	! 1003 2012-09-14 14:35:53Z raasch
117	! adjustment of array tend for cases with unequal subdomain sizes removed
118	!
119	! Revision 1.1 1997/07/24 11:24:44 raasch
120	! Initial revision
121	!
122	!
123	! Description:
124	! ------------
125	!> Compute the divergence of the provisional velocity field. Solve the Poisson
126	!> equation for the perturbation pressure. Compute the final velocities using
127	!> this perturbation pressure. Compute the remaining divergence.
128	!------------------------------------------------------------------------------!
129	SUBROUTINE pres
130
131
132	USE arrays_3d, &
133	ONLY: d, ddzu, ddzu_pres, ddzw, dzw, p, p_loc, rho_air, rho_air_zw, &
134	tend, u, v, w
135
136	USE control_parameters, &
137	ONLY: bc_lr_cyc, bc_ns_cyc, conserve_volume_flow, dt_3d, &
138	gathered_size, ibc_p_b, ibc_p_t, intermediate_timestep_count, &
139	intermediate_timestep_count_max, mg_switch_to_pe0_level, &
140	nest_domain, on_device, outflow_l, outflow_n, outflow_r, &
141	outflow_s, psolver, subdomain_size, topography, volume_flow, &
142	volume_flow_area, volume_flow_initial
143
144	USE cpulog, &
145	ONLY: cpu_log, log_point, log_point_s
146
147	USE grid_variables, &
148	ONLY: ddx, ddy
149
150	USE indices, &
151	ONLY: nbgp, ngp_2dh_outer, nx, nxl, nxlg, nxl_mg, nxr, nxrg, nxr_mg, &
152	ny, nys, nysg, nys_mg, nyn, nyng, nyn_mg, nzb, nzb_s_inner, &
153	nzb_u_inner, nzb_v_inner, nzb_w_inner, nzt, nzt_mg, &
154	rflags_s_inner
155
156	USE kinds
157
158	USE pegrid
159
160	USE pmc_interface, &
161	ONLY: nesting_mode
162
163	USE poisfft_mod, &
164	ONLY: poisfft
165
166	USE poismg_mod
167
168	USE statistics, &
169	ONLY: statistic_regions, sums_divnew_l, sums_divold_l, weight_pres, &
170	weight_substep
171
172	IMPLICIT NONE
173
174	INTEGER(iwp) :: i !<
175	INTEGER(iwp) :: j !<
176	INTEGER(iwp) :: k !<
177
178	REAL(wp) :: ddt_3d !<
179	REAL(wp) :: d_weight_pres !<
180	REAL(wp) :: localsum !<
181	REAL(wp) :: threadsum !<
182	REAL(wp) :: weight_pres_l !<
183	REAL(wp) :: weight_substep_l !<
184
185	REAL(wp), DIMENSION(1:3) :: volume_flow_l !<
186	REAL(wp), DIMENSION(1:3) :: volume_flow_offset !<
187	REAL(wp), DIMENSION(1:nzt) :: w_l !<
188	REAL(wp), DIMENSION(1:nzt) :: w_l_l !<
189
190	LOGICAL :: nest_domain_nvn !<
191
192
193	CALL cpu_log( log_point(8), 'pres', 'start' )
194
195
196	!
197	!-- Calculate quantities to be used locally
198	ddt_3d = 1.0_wp / dt_3d
199	IF ( intermediate_timestep_count == 0 ) THEN
200	!
201	!-- If pres is called before initial time step
202	weight_pres_l = 1.0_wp
203	d_weight_pres = 1.0_wp
204	weight_substep_l = 1.0_wp
205	ELSE
206	weight_pres_l = weight_pres(intermediate_timestep_count)
207	d_weight_pres = 1.0_wp / weight_pres(intermediate_timestep_count)
208	weight_substep_l = weight_substep(intermediate_timestep_count)
209	ENDIF
210
211	!
212	!-- Multigrid method expects array d to have one ghost layer.
213	!--
214	IF ( psolver(1:9) == 'multigrid' ) THEN
215
216	DEALLOCATE( d )
217	ALLOCATE( d(nzb:nzt+1,nys-1:nyn+1,nxl-1:nxr+1) )
218
219	!
220	!-- Since p is later used to hold the weighted average of the substeps, it
221	!-- cannot be used in the iterative solver. Therefore, its initial value is
222	!-- stored on p_loc, which is then iteratively advanced in every substep.
223	IF ( intermediate_timestep_count <= 1 ) THEN
224	DO i = nxl-1, nxr+1
225	DO j = nys-1, nyn+1
226	DO k = nzb, nzt+1
227	p_loc(k,j,i) = p(k,j,i)
228	ENDDO
229	ENDDO
230	ENDDO
231	ENDIF
232
233	ELSEIF ( psolver == 'sor' .AND. intermediate_timestep_count <= 1 ) THEN
234
235	!
236	!-- Since p is later used to hold the weighted average of the substeps, it
237	!-- cannot be used in the iterative solver. Therefore, its initial value is
238	!-- stored on p_loc, which is then iteratively advanced in every substep.
239	p_loc = p
240
241	ENDIF
242
243	!
244	!-- Conserve the volume flow at the outflow in case of non-cyclic lateral
245	!-- boundary conditions
246	!-- WARNING: so far, this conservation does not work at the left/south
247	!-- boundary if the topography at the inflow differs from that at the
248	!-- outflow! For this case, volume_flow_area needs adjustment!
249	!
250	!-- Left/right
251	IF ( conserve_volume_flow .AND. ( outflow_l .OR. outflow_r ) ) THEN
252
253	volume_flow(1) = 0.0_wp
254	volume_flow_l(1) = 0.0_wp
255
256	IF ( outflow_l ) THEN
257	i = 0
258	ELSEIF ( outflow_r ) THEN
259	i = nx+1
260	ENDIF
261
262	DO j = nys, nyn
263	!
264	!-- Sum up the volume flow through the south/north boundary
265	DO k = nzb_u_inner(j,i)+1, nzt
266	volume_flow_l(1) = volume_flow_l(1) + u(k,j,i) * dzw(k)
267	ENDDO
268	ENDDO
269
270	#if defined( __parallel )
271	IF ( collective_wait ) CALL MPI_BARRIER( comm1dy, ierr )
272	CALL MPI_ALLREDUCE( volume_flow_l(1), volume_flow(1), 1, MPI_REAL, &
273	MPI_SUM, comm1dy, ierr )
274	#else
275	volume_flow = volume_flow_l
276	#endif
277	volume_flow_offset(1) = ( volume_flow_initial(1) - volume_flow(1) ) &
278	/ volume_flow_area(1)
279
280	DO j = nysg, nyng
281	DO k = nzb_u_inner(j,i)+1, nzt
282	u(k,j,i) = u(k,j,i) + volume_flow_offset(1)
283	ENDDO
284	ENDDO
285
286	ENDIF
287
288	!
289	!-- South/north
290	IF ( conserve_volume_flow .AND. ( outflow_n .OR. outflow_s ) ) THEN
291
292	volume_flow(2) = 0.0_wp
293	volume_flow_l(2) = 0.0_wp
294
295	IF ( outflow_s ) THEN
296	j = 0
297	ELSEIF ( outflow_n ) THEN
298	j = ny+1
299	ENDIF
300
301	DO i = nxl, nxr
302	!
303	!-- Sum up the volume flow through the south/north boundary
304	DO k = nzb_v_inner(j,i)+1, nzt
305	volume_flow_l(2) = volume_flow_l(2) + v(k,j,i) * dzw(k)
306	ENDDO
307	ENDDO
308
309	#if defined( __parallel )
310	IF ( collective_wait ) CALL MPI_BARRIER( comm1dx, ierr )
311	CALL MPI_ALLREDUCE( volume_flow_l(2), volume_flow(2), 1, MPI_REAL, &
312	MPI_SUM, comm1dx, ierr )
313	#else
314	volume_flow = volume_flow_l
315	#endif
316	volume_flow_offset(2) = ( volume_flow_initial(2) - volume_flow(2) ) &
317	/ volume_flow_area(2)
318
319	DO i = nxlg, nxrg
320	DO k = nzb_v_inner(j,i)+1, nzt
321	v(k,j,i) = v(k,j,i) + volume_flow_offset(2)
322	ENDDO
323	ENDDO
324
325	ENDIF
326
327	!
328	!-- Remove mean vertical velocity in case that Neumann conditions are
329	!-- used both at bottom and top boundary, and if not a nested domain in a
330	!-- normal nesting run. In case of vertical nesting, this must be done.
331	!-- Therefore an auxiliary logical variable nest_domain_nvn is used here, and
332	!-- nvn stands for non-vertical nesting.
333	!-- This cannot be done before the first initial time step because ngp_2dh_outer
334	!-- is not yet known then.
335	nest_domain_nvn = nest_domain
336	IF ( nest_domain .AND. nesting_mode == 'vertical' ) THEN
337	nest_domain_nvn = .FALSE.
338	ENDIF
339
340	IF ( ibc_p_b == 1 .AND. ibc_p_t == 1 .AND. &
341	.NOT. nest_domain_nvn .AND. intermediate_timestep_count /= 0 ) &
342	THEN
343	w_l = 0.0_wp; w_l_l = 0.0_wp
344	DO i = nxl, nxr
345	DO j = nys, nyn
346	DO k = nzb_w_inner(j,i)+1, nzt
347	w_l_l(k) = w_l_l(k) + w(k,j,i)
348	ENDDO
349	ENDDO
350	ENDDO
351	#if defined( __parallel )
352	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
353	CALL MPI_ALLREDUCE( w_l_l(1), w_l(1), nzt, MPI_REAL, MPI_SUM, &
354	comm2d, ierr )
355	#else
356	w_l = w_l_l
357	#endif
358	DO k = 1, nzt
359	w_l(k) = w_l(k) / ngp_2dh_outer(k,0)
360	ENDDO
361	DO i = nxlg, nxrg
362	DO j = nysg, nyng
363	DO k = nzb_w_inner(j,i)+1, nzt
364	w(k,j,i) = w(k,j,i) - w_l(k)
365	ENDDO
366	ENDDO
367	ENDDO
368	ENDIF
369
370	!
371	!-- Compute the divergence of the provisional velocity field.
372	CALL cpu_log( log_point_s(1), 'divergence', 'start' )
373
374	IF ( psolver(1:9) == 'multigrid' ) THEN
375	!$OMP PARALLEL DO SCHEDULE( STATIC )
376	DO i = nxl-1, nxr+1
377	DO j = nys-1, nyn+1
378	DO k = nzb, nzt+1
379	d(k,j,i) = 0.0_wp
380	ENDDO
381	ENDDO
382	ENDDO
383	ELSE
384	!$OMP PARALLEL DO SCHEDULE( STATIC )
385	!$acc kernels present( d )
386	DO i = nxl, nxr
387	DO j = nys, nyn
388	DO k = nzb+1, nzt
389	d(k,j,i) = 0.0_wp
390	ENDDO
391	ENDDO
392	ENDDO
393	!$acc end kernels
394	ENDIF
395
396	localsum = 0.0_wp
397	threadsum = 0.0_wp
398
399	#if defined( __ibm )
400	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
401	!$OMP DO SCHEDULE( STATIC )
402	DO i = nxl, nxr
403	DO j = nys, nyn
404	DO k = nzb_s_inner(j,i)+1, nzt
405	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * rho_air(k) * ddx + &
406	( v(k,j+1,i) - v(k,j,i) ) * rho_air(k) * ddy + &
407	( w(k,j,i) * rho_air_zw(k) - &
408	w(k-1,j,i) * rho_air_zw(k-1) ) * ddzw(k) &
409	) * ddt_3d * d_weight_pres
410	ENDDO
411	!
412	!-- Compute possible PE-sum of divergences for flow_statistics
413	DO k = nzb_s_inner(j,i)+1, nzt
414	threadsum = threadsum + ABS( d(k,j,i) )
415	ENDDO
416
417	ENDDO
418	ENDDO
419
420	IF ( intermediate_timestep_count == intermediate_timestep_count_max .OR. &
421	intermediate_timestep_count == 0 ) THEN
422	localsum = localsum + threadsum * dt_3d * weight_pres_l
423	ENDIF
424	!$OMP END PARALLEL
425	#else
426
427	!$OMP PARALLEL PRIVATE (i,j,k)
428	!$OMP DO SCHEDULE( STATIC )
429	!$acc kernels present( d, ddzw, rflags_s_inner, u, v, w )
430	!$acc loop collapse( 3 )
431	DO i = nxl, nxr
432	DO j = nys, nyn
433	DO k = 1, nzt
434	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * rho_air(k) * ddx + &
435	( v(k,j+1,i) - v(k,j,i) ) * rho_air(k) * ddy + &
436	( w(k,j,i) * rho_air_zw(k) - &
437	w(k-1,j,i) * rho_air_zw(k-1) ) * ddzw(k) &
438	) * ddt_3d * d_weight_pres * rflags_s_inner(k,j,i)
439	ENDDO
440	ENDDO
441	ENDDO
442	!$acc end kernels
443	!$OMP END PARALLEL
444
445	!
446	!-- Compute possible PE-sum of divergences for flow_statistics. Carry out
447	!-- computation only at last Runge-Kutta substep.
448	IF ( intermediate_timestep_count == intermediate_timestep_count_max .OR. &
449	intermediate_timestep_count == 0 ) THEN
450	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
451	!$OMP DO SCHEDULE( STATIC )
452	!$acc parallel loop collapse(3) present( d ) reduction(+:threadsum)
453	DO i = nxl, nxr
454	DO j = nys, nyn
455	DO k = nzb+1, nzt
456	threadsum = threadsum + ABS( d(k,j,i) )
457	ENDDO
458	ENDDO
459	ENDDO
460	!$acc end parallel loop
461	localsum = localsum + threadsum * dt_3d * weight_pres_l
462	!$OMP END PARALLEL
463	ENDIF
464	#endif
465
466	!
467	!-- For completeness, set the divergence sum of all statistic regions to those
468	!-- of the total domain
469	IF ( intermediate_timestep_count == intermediate_timestep_count_max .OR. &
470	intermediate_timestep_count == 0 ) THEN
471	sums_divold_l(0:statistic_regions) = localsum
472	ENDIF
473
474	CALL cpu_log( log_point_s(1), 'divergence', 'stop' )
475
476	!
477	!-- Compute the pressure perturbation solving the Poisson equation
478	IF ( psolver == 'poisfft' ) THEN
479
480	!
481	!-- Solve Poisson equation via FFT and solution of tridiagonal matrices
482	CALL poisfft( d )
483
484	!
485	!-- Store computed perturbation pressure and set boundary condition in
486	!-- z-direction
487	!$OMP PARALLEL DO
488	!$acc kernels present( d, tend )
489	DO i = nxl, nxr
490	DO j = nys, nyn
491	DO k = nzb+1, nzt
492	tend(k,j,i) = d(k,j,i)
493	ENDDO
494	ENDDO
495	ENDDO
496	!$acc end kernels
497
498	!
499	!-- Bottom boundary:
500	!-- This condition is only required for internal output. The pressure
501	!-- gradient (dp(nzb+1)-dp(nzb))/dz is not used anywhere else.
502	IF ( ibc_p_b == 1 ) THEN
503	!
504	!-- Neumann (dp/dz = 0)
505	!$OMP PARALLEL DO
506	!$acc kernels present( nzb_s_inner, tend )
507	DO i = nxlg, nxrg
508	DO j = nysg, nyng
509	tend(nzb_s_inner(j,i),j,i) = tend(nzb_s_inner(j,i)+1,j,i)
510	ENDDO
511	ENDDO
512	!$acc end kernels
513
514	ELSE
515	!
516	!-- Dirichlet
517	!$OMP PARALLEL DO
518	!$acc kernels present( tend )
519	DO i = nxlg, nxrg
520	DO j = nysg, nyng
521	tend(nzb_s_inner(j,i),j,i) = 0.0_wp
522	ENDDO
523	ENDDO
524	!$acc end kernels
525
526	ENDIF
527
528	!
529	!-- Top boundary
530	IF ( ibc_p_t == 1 ) THEN
531	!
532	!-- Neumann
533	!$OMP PARALLEL DO
534	!$acc kernels present( tend )
535	DO i = nxlg, nxrg
536	DO j = nysg, nyng
537	tend(nzt+1,j,i) = tend(nzt,j,i)
538	ENDDO
539	ENDDO
540	!$acc end kernels
541
542	ELSE
543	!
544	!-- Dirichlet
545	!$OMP PARALLEL DO
546	!$acc kernels present( tend )
547	DO i = nxlg, nxrg
548	DO j = nysg, nyng
549	tend(nzt+1,j,i) = 0.0_wp
550	ENDDO
551	ENDDO
552	!$acc end kernels
553
554	ENDIF
555
556	!
557	!-- Exchange boundaries for p
558	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
559	on_device = .TRUE. ! to be removed after complete porting
560	ELSE ! of ghost point exchange
561	!$acc update host( tend )
562	ENDIF
563	CALL exchange_horiz( tend, nbgp )
564	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
565	on_device = .FALSE. ! to be removed after complete porting
566	ELSE ! of ghost point exchange
567	!$acc update device( tend )
568	ENDIF
569
570	ELSEIF ( psolver == 'sor' ) THEN
571
572	!
573	!-- Solve Poisson equation for perturbation pressure using SOR-Red/Black
574	!-- scheme
575	CALL sor( d, ddzu_pres, ddzw, p_loc )
576	tend = p_loc
577
578	ELSEIF ( psolver(1:9) == 'multigrid' ) THEN
579
580	!
581	!-- Solve Poisson equation for perturbation pressure using Multigrid scheme,
582	!-- array tend is used to store the residuals, logical exchange_mg is used
583	!-- to discern data exchange in multigrid ( 1 ghostpoint ) and normal grid
584	!-- ( nbgp ghost points ).
585
586	!-- If the number of grid points of the gathered grid, which is collected
587	!-- on PE0, is larger than the number of grid points of an PE, than array
588	!-- tend will be enlarged.
589	IF ( gathered_size > subdomain_size ) THEN
590	DEALLOCATE( tend )
591	ALLOCATE( tend(nzb:nzt_mg(mg_switch_to_pe0_level)+1,nys_mg( &
592	mg_switch_to_pe0_level)-1:nyn_mg(mg_switch_to_pe0_level)+1,&
593	nxl_mg(mg_switch_to_pe0_level)-1:nxr_mg( &
594	mg_switch_to_pe0_level)+1) )
595	ENDIF
596
597	IF ( psolver == 'multigrid' ) THEN
598	CALL poismg( tend )
599	ELSE
600	CALL poismg_noopt( tend )
601	ENDIF
602
603	IF ( gathered_size > subdomain_size ) THEN
604	DEALLOCATE( tend )
605	ALLOCATE( tend(nzb:nzt+1,nysg:nyng,nxlg:nxrg) )
606	ENDIF
607
608	!
609	!-- Restore perturbation pressure on tend because this array is used
610	!-- further below to correct the velocity fields
611	DO i = nxl-1, nxr+1
612	DO j = nys-1, nyn+1
613	DO k = nzb, nzt+1
614	tend(k,j,i) = p_loc(k,j,i)
615	ENDDO
616	ENDDO
617	ENDDO
618
619	ENDIF
620
621	!
622	!-- Store perturbation pressure on array p, used for pressure data output.
623	!-- Ghost layers are added in the output routines (except sor-method: see below)
624	IF ( intermediate_timestep_count <= 1 ) THEN
625	!$OMP PARALLEL PRIVATE (i,j,k)
626	!$OMP DO
627	!$acc kernels present( p, tend, weight_substep_l )
628	!$acc loop independent
629	DO i = nxl-1, nxr+1
630	!$acc loop independent
631	DO j = nys-1, nyn+1
632	!$acc loop independent
633	DO k = nzb, nzt+1
634	p(k,j,i) = tend(k,j,i) * &
635	weight_substep_l
636	ENDDO
637	ENDDO
638	ENDDO
639	!$acc end kernels
640	!$OMP END PARALLEL
641
642	ELSEIF ( intermediate_timestep_count > 1 ) THEN
643	!$OMP PARALLEL PRIVATE (i,j,k)
644	!$OMP DO
645	!$acc kernels present( p, tend, weight_substep_l )
646	!$acc loop independent
647	DO i = nxl-1, nxr+1
648	!$acc loop independent
649	DO j = nys-1, nyn+1
650	!$acc loop independent
651	DO k = nzb, nzt+1
652	p(k,j,i) = p(k,j,i) + tend(k,j,i) * &
653	weight_substep_l
654	ENDDO
655	ENDDO
656	ENDDO
657	!$acc end kernels
658	!$OMP END PARALLEL
659
660	ENDIF
661
662	!
663	!-- SOR-method needs ghost layers for the next timestep
664	IF ( psolver == 'sor' ) CALL exchange_horiz( p, nbgp )
665
666	!
667	!-- Correction of the provisional velocities with the current perturbation
668	!-- pressure just computed
669	IF ( conserve_volume_flow .AND. ( bc_lr_cyc .OR. bc_ns_cyc ) ) THEN
670	volume_flow_l(1) = 0.0_wp
671	volume_flow_l(2) = 0.0_wp
672	ENDIF
673
674	!$OMP PARALLEL PRIVATE (i,j,k)
675	!$OMP DO
676	!$acc kernels present( ddzu, nzb_u_inner, nzb_v_inner, nzb_w_inner, tend, u, v, w )
677	!$acc loop independent
678	DO i = nxl, nxr
679	!$acc loop independent
680	DO j = nys, nyn
681	!$acc loop independent
682	DO k = 1, nzt
683	IF ( k > nzb_w_inner(j,i) ) THEN
684	w(k,j,i) = w(k,j,i) - dt_3d * &
685	( tend(k+1,j,i) - tend(k,j,i) ) * ddzu(k+1) * &
686	weight_pres_l
687	ENDIF
688	ENDDO
689	!$acc loop independent
690	DO k = 1, nzt
691	IF ( k > nzb_u_inner(j,i) ) THEN
692	u(k,j,i) = u(k,j,i) - dt_3d * &
693	( tend(k,j,i) - tend(k,j,i-1) ) * ddx * &
694	weight_pres_l
695	ENDIF
696	ENDDO
697	!$acc loop independent
698	DO k = 1, nzt
699	IF ( k > nzb_v_inner(j,i) ) THEN
700	v(k,j,i) = v(k,j,i) - dt_3d * &
701	( tend(k,j,i) - tend(k,j-1,i) ) * ddy * &
702	weight_pres_l
703	ENDIF
704	ENDDO
705
706	ENDDO
707	ENDDO
708	!$acc end kernels
709	!$OMP END PARALLEL
710
711	!
712	!-- Sum up the volume flow through the right and north boundary
713	IF ( conserve_volume_flow .AND. bc_lr_cyc .AND. bc_ns_cyc .AND. &
714	nxr == nx ) THEN
715
716	!$OMP PARALLEL PRIVATE (j,k)
717	!$OMP DO
718	DO j = nys, nyn
719	!$OMP CRITICAL
720	DO k = nzb_u_inner(j,nx) + 1, nzt
721	volume_flow_l(1) = volume_flow_l(1) + u(k,j,nx) * dzw(k)
722	ENDDO
723	!$OMP END CRITICAL
724	ENDDO
725	!$OMP END PARALLEL
726
727	ENDIF
728
729	IF ( conserve_volume_flow .AND. bc_ns_cyc .AND. bc_lr_cyc .AND. &
730	nyn == ny ) THEN
731
732	!$OMP PARALLEL PRIVATE (i,k)
733	!$OMP DO
734	DO i = nxl, nxr
735	!$OMP CRITICAL
736	DO k = nzb_v_inner(ny,i) + 1, nzt
737	volume_flow_l(2) = volume_flow_l(2) + v(k,ny,i) * dzw(k)
738	ENDDO
739	!$OMP END CRITICAL
740	ENDDO
741	!$OMP END PARALLEL
742
743	ENDIF
744
745	!
746	!-- Conserve the volume flow
747	IF ( conserve_volume_flow .AND. ( bc_lr_cyc .AND. bc_ns_cyc ) ) THEN
748
749	#if defined( __parallel )
750	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
751	CALL MPI_ALLREDUCE( volume_flow_l(1), volume_flow(1), 2, MPI_REAL, &
752	MPI_SUM, comm2d, ierr )
753	#else
754	volume_flow = volume_flow_l
755	#endif
756
757	volume_flow_offset(1:2) = ( volume_flow_initial(1:2) - volume_flow(1:2) ) / &
758	volume_flow_area(1:2)
759
760	!$OMP PARALLEL PRIVATE (i,j,k)
761	!$OMP DO
762	DO i = nxl, nxr
763	DO j = nys, nyn
764	DO k = nzb_u_inner(j,i) + 1, nzt
765	u(k,j,i) = u(k,j,i) + volume_flow_offset(1)
766	ENDDO
767	DO k = nzb_v_inner(j,i) + 1, nzt
768	v(k,j,i) = v(k,j,i) + volume_flow_offset(2)
769	ENDDO
770	ENDDO
771	ENDDO
772
773	!$OMP END PARALLEL
774
775	ENDIF
776
777	!
778	!-- Exchange of boundaries for the velocities
779	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
780	on_device = .TRUE. ! to be removed after complete porting
781	ELSE ! of ghost point exchange
782	!$acc update host( u, v, w )
783	ENDIF
784	CALL exchange_horiz( u, nbgp )
785	CALL exchange_horiz( v, nbgp )
786	CALL exchange_horiz( w, nbgp )
787	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
788	on_device = .FALSE. ! to be removed after complete porting
789	ELSE ! of ghost point exchange
790	!$acc update device( u, v, w )
791	ENDIF
792
793	!
794	!-- Compute the divergence of the corrected velocity field,
795	!-- A possible PE-sum is computed in flow_statistics. Carry out computation
796	!-- only at last Runge-Kutta step.
797	IF ( intermediate_timestep_count == intermediate_timestep_count_max .OR. &
798	intermediate_timestep_count == 0 ) THEN
799	CALL cpu_log( log_point_s(1), 'divergence', 'start' )
800	sums_divnew_l = 0.0_wp
801
802	!
803	!-- d must be reset to zero because it can contain nonzero values below the
804	!-- topography
805	IF ( topography /= 'flat' ) d = 0.0_wp
806
807	localsum = 0.0_wp
808	threadsum = 0.0_wp
809
810	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
811	!$OMP DO SCHEDULE( STATIC )
812	#if defined( __ibm )
813	DO i = nxl, nxr
814	DO j = nys, nyn
815	DO k = nzb_s_inner(j,i)+1, nzt
816	d(k,j,i) = ( u(k,j,i+1) - u(k,j,i) ) * rho_air(k) * ddx + &
817	( v(k,j+1,i) - v(k,j,i) ) * rho_air(k) * ddy + &
818	( w(k,j,i) * rho_air_zw(k) - &
819	w(k-1,j,i) * rho_air_zw(k-1) ) * ddzw(k)
820	ENDDO
821	DO k = nzb+1, nzt
822	threadsum = threadsum + ABS( d(k,j,i) )
823	ENDDO
824	ENDDO
825	ENDDO
826	#else
827	!$acc kernels present( d, ddzw, rflags_s_inner, u, v, w )
828	!$acc loop collapse( 3 )
829	DO i = nxl, nxr
830	DO j = nys, nyn
831	DO k = 1, nzt
832	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * rho_air(k) * ddx + &
833	( v(k,j+1,i) - v(k,j,i) ) * rho_air(k) * ddy + &
834	( w(k,j,i) * rho_air_zw(k) - &
835	w(k-1,j,i) * rho_air_zw(k-1) ) * ddzw(k) &
836	) * rflags_s_inner(k,j,i)
837	ENDDO
838	ENDDO
839	ENDDO
840	!$acc end kernels
841	!
842	!-- Compute possible PE-sum of divergences for flow_statistics
843	!$acc parallel loop collapse(3) present( d ) reduction(+:threadsum)
844	DO i = nxl, nxr
845	DO j = nys, nyn
846	DO k = nzb+1, nzt
847	threadsum = threadsum + ABS( d(k,j,i) )
848	ENDDO
849	ENDDO
850	ENDDO
851	!$acc end parallel loop
852	#endif
853
854	localsum = localsum + threadsum
855	!$OMP END PARALLEL
856
857	!
858	!-- For completeness, set the divergence sum of all statistic regions to those
859	!-- of the total domain
860	sums_divnew_l(0:statistic_regions) = localsum
861
862	CALL cpu_log( log_point_s(1), 'divergence', 'stop' )
863
864	ENDIF
865
866	CALL cpu_log( log_point(8), 'pres', 'stop' )
867
868
869	END SUBROUTINE pres

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

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |