Home

Context Navigation

source: palm/trunk/SOURCE/pres.f90 @ 1966

Last change on this file since 1966 was 1933, checked in by hellstea, 8 years ago
last commit documented
Property svn:keywords set to `Id`
File size: 27.4 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 1933 2016-06-13 07:12:51Z maronga $
26	!
27	! 1932 2016-06-10 12:09:21Z suehring
28	! Initial version of purely vertical nesting introduced.
29	!
30	! 1931 2016-06-10 12:06:59Z suehring
31	! Rename multigrid into multigrid_noopt and multigrid_fast into multigrid
32	!
33	! 1929 2016-06-09 16:25:25Z suehring
34	! Bugfix: weight_substep for initial call, replace by local variable
35	!
36	! 1918 2016-05-27 14:35:57Z raasch
37	! sum of divergence is also calculated when pres is called before the initial
38	! first time step,
39	! stearing is modified by using intermediate_timestep_count = 0 in order to
40	! determine if pres is called before the first initial timestep,
41	! bugfix: in case of Neumann conditions for pressure both at bottom and top,
42	! mean vertical velocity is also removed for the first time step
43	! bugfix for calculating divergences
44	!
45	! 1908 2016-05-25 17:22:32Z suehring
46	! New divergence for output into RUN_CONTROL file is calculated only at last
47	! Runge-Kutta step
48	!
49	! 1845 2016-04-08 08:29:13Z raasch
50	! nzb_2d replace by nzb_u\|v_inner
51	!
52	! 1799 2016-04-05 08:35:55Z gronemeier
53	! Bugfix: excluded third dimension from horizontal volume flow calculation
54	!
55	! 1762 2016-02-25 12:31:13Z hellstea
56	! Introduction of nested domain feature
57	!
58	! 1682 2015-10-07 23:56:08Z knoop
59	! Code annotations made doxygen readable
60	!
61	! 1575 2015-03-27 09:56:27Z raasch
62	! poismg_fast + respective module added, adjustments for psolver-queries
63	!
64	! 1342 2014-03-26 17:04:47Z kanani
65	! REAL constants defined as wp-kind
66	!
67	! 1320 2014-03-20 08:40:49Z raasch
68	! ONLY-attribute added to USE-statements,
69	! kind-parameters added to all INTEGER and REAL declaration statements,
70	! kinds are defined in new module kinds,
71	! old module precision_kind is removed,
72	! revision history before 2012 removed,
73	! comment fields (!:) to be used for variable explanations added to
74	! all variable declaration statements
75	!
76	! 1318 2014-03-17 13:35:16Z raasch
77	! module interfaces removed
78	!
79	! 1306 2014-03-13 14:30:59Z raasch
80	! second argument removed from routine poisfft
81	!
82	! 1257 2013-11-08 15:18:40Z raasch
83	! openacc loop and loop vector clauses removed, independent clauses added,
84	! end parallel replaced by end parallel loop
85	!
86	! 1221 2013-09-10 08:59:13Z raasch
87	! openACC porting of reduction operations, loops for calculating d are
88	! using the rflags_s_inner multiply flag instead of the nzb_s_inner loop index
89	!
90	! 1212 2013-08-15 08:46:27Z raasch
91	! call of poisfft_hybrid removed
92	!
93	! 1117 2013-03-27 11:15:36Z suehring
94	! Bugfix in OpenMP parallelization.
95	!
96	! 1113 2013-03-10 02:48:14Z raasch
97	! GPU-porting of several loops, some loops rearranged
98	!
99	! 1111 2013-03-08 23:54:10Z
100	! openACC statements added,
101	! ibc_p_b = 2 removed
102	!
103	! 1092 2013-02-02 11:24:22Z raasch
104	! unused variables removed
105	!
106	! 1036 2012-10-22 13:43:42Z raasch
107	! code put under GPL (PALM 3.9)
108	!
109	! 1003 2012-09-14 14:35:53Z raasch
110	! adjustment of array tend for cases with unequal subdomain sizes removed
111	!
112	! Revision 1.1 1997/07/24 11:24:44 raasch
113	! Initial revision
114	!
115	!
116	! Description:
117	! ------------
118	!> Compute the divergence of the provisional velocity field. Solve the Poisson
119	!> equation for the perturbation pressure. Compute the final velocities using
120	!> this perturbation pressure. Compute the remaining divergence.
121	!------------------------------------------------------------------------------!
122	SUBROUTINE pres
123
124
125	USE arrays_3d, &
126	ONLY: d, ddzu, ddzu_pres, ddzw, dzw, p, p_loc, tend, u, v, w
127
128	USE control_parameters, &
129	ONLY: bc_lr_cyc, bc_ns_cyc, conserve_volume_flow, dt_3d, &
130	gathered_size, ibc_p_b, ibc_p_t, intermediate_timestep_count, &
131	intermediate_timestep_count_max, mg_switch_to_pe0_level, &
132	nest_domain, on_device, outflow_l, outflow_n, outflow_r, &
133	outflow_s, psolver, subdomain_size, topography, volume_flow, &
134	volume_flow_area, volume_flow_initial
135
136	USE cpulog, &
137	ONLY: cpu_log, log_point, log_point_s
138
139	USE grid_variables, &
140	ONLY: ddx, ddy
141
142	USE indices, &
143	ONLY: nbgp, ngp_2dh_outer, nx, nxl, nxlg, nxl_mg, nxr, nxrg, nxr_mg, &
144	ny, nys, nysg, nys_mg, nyn, nyng, nyn_mg, nzb, nzb_s_inner, &
145	nzb_u_inner, nzb_v_inner, nzb_w_inner, nzt, nzt_mg, &
146	rflags_s_inner
147
148	USE kinds
149
150	USE pegrid
151
152	USE pmc_interface, &
153	ONLY: nesting_mode
154
155	USE poisfft_mod, &
156	ONLY: poisfft
157
158	USE poismg_mod
159
160	USE statistics, &
161	ONLY: statistic_regions, sums_divnew_l, sums_divold_l, weight_pres, &
162	weight_substep
163
164	IMPLICIT NONE
165
166	INTEGER(iwp) :: i !<
167	INTEGER(iwp) :: j !<
168	INTEGER(iwp) :: k !<
169
170	REAL(wp) :: ddt_3d !<
171	REAL(wp) :: d_weight_pres !<
172	REAL(wp) :: localsum !<
173	REAL(wp) :: threadsum !<
174	REAL(wp) :: weight_pres_l !<
175	REAL(wp) :: weight_substep_l !<
176
177	REAL(wp), DIMENSION(1:3) :: volume_flow_l !<
178	REAL(wp), DIMENSION(1:3) :: volume_flow_offset !<
179	REAL(wp), DIMENSION(1:nzt) :: w_l !<
180	REAL(wp), DIMENSION(1:nzt) :: w_l_l !<
181
182	LOGICAL :: nest_domain_nvn !<
183
184
185	CALL cpu_log( log_point(8), 'pres', 'start' )
186
187
188	!
189	!-- Calculate quantities to be used locally
190	ddt_3d = 1.0_wp / dt_3d
191	IF ( intermediate_timestep_count == 0 ) THEN
192	!
193	!-- If pres is called before initial time step
194	weight_pres_l = 1.0_wp
195	d_weight_pres = 1.0_wp
196	weight_substep_l = 1.0_wp
197	ELSE
198	weight_pres_l = weight_pres(intermediate_timestep_count)
199	d_weight_pres = 1.0_wp / weight_pres(intermediate_timestep_count)
200	weight_substep_l = weight_substep(intermediate_timestep_count)
201	ENDIF
202
203	!
204	!-- Multigrid method expects array d to have one ghost layer.
205	!--
206	IF ( psolver(1:9) == 'multigrid' ) THEN
207
208	DEALLOCATE( d )
209	ALLOCATE( d(nzb:nzt+1,nys-1:nyn+1,nxl-1:nxr+1) )
210
211	!
212	!-- Since p is later used to hold the weighted average of the substeps, it
213	!-- cannot be used in the iterative solver. Therefore, its initial value is
214	!-- stored on p_loc, which is then iteratively advanced in every substep.
215	IF ( intermediate_timestep_count <= 1 ) THEN
216	DO i = nxl-1, nxr+1
217	DO j = nys-1, nyn+1
218	DO k = nzb, nzt+1
219	p_loc(k,j,i) = p(k,j,i)
220	ENDDO
221	ENDDO
222	ENDDO
223	ENDIF
224
225	ELSEIF ( psolver == 'sor' .AND. intermediate_timestep_count <= 1 ) THEN
226
227	!
228	!-- Since p is later used to hold the weighted average of the substeps, it
229	!-- cannot be used in the iterative solver. Therefore, its initial value is
230	!-- stored on p_loc, which is then iteratively advanced in every substep.
231	p_loc = p
232
233	ENDIF
234
235	!
236	!-- Conserve the volume flow at the outflow in case of non-cyclic lateral
237	!-- boundary conditions
238	!-- WARNING: so far, this conservation does not work at the left/south
239	!-- boundary if the topography at the inflow differs from that at the
240	!-- outflow! For this case, volume_flow_area needs adjustment!
241	!
242	!-- Left/right
243	IF ( conserve_volume_flow .AND. ( outflow_l .OR. outflow_r ) ) THEN
244
245	volume_flow(1) = 0.0_wp
246	volume_flow_l(1) = 0.0_wp
247
248	IF ( outflow_l ) THEN
249	i = 0
250	ELSEIF ( outflow_r ) THEN
251	i = nx+1
252	ENDIF
253
254	DO j = nys, nyn
255	!
256	!-- Sum up the volume flow through the south/north boundary
257	DO k = nzb_u_inner(j,i)+1, nzt
258	volume_flow_l(1) = volume_flow_l(1) + u(k,j,i) * dzw(k)
259	ENDDO
260	ENDDO
261
262	#if defined( __parallel )
263	IF ( collective_wait ) CALL MPI_BARRIER( comm1dy, ierr )
264	CALL MPI_ALLREDUCE( volume_flow_l(1), volume_flow(1), 1, MPI_REAL, &
265	MPI_SUM, comm1dy, ierr )
266	#else
267	volume_flow = volume_flow_l
268	#endif
269	volume_flow_offset(1) = ( volume_flow_initial(1) - volume_flow(1) ) &
270	/ volume_flow_area(1)
271
272	DO j = nysg, nyng
273	DO k = nzb_u_inner(j,i)+1, nzt
274	u(k,j,i) = u(k,j,i) + volume_flow_offset(1)
275	ENDDO
276	ENDDO
277
278	ENDIF
279
280	!
281	!-- South/north
282	IF ( conserve_volume_flow .AND. ( outflow_n .OR. outflow_s ) ) THEN
283
284	volume_flow(2) = 0.0_wp
285	volume_flow_l(2) = 0.0_wp
286
287	IF ( outflow_s ) THEN
288	j = 0
289	ELSEIF ( outflow_n ) THEN
290	j = ny+1
291	ENDIF
292
293	DO i = nxl, nxr
294	!
295	!-- Sum up the volume flow through the south/north boundary
296	DO k = nzb_v_inner(j,i)+1, nzt
297	volume_flow_l(2) = volume_flow_l(2) + v(k,j,i) * dzw(k)
298	ENDDO
299	ENDDO
300
301	#if defined( __parallel )
302	IF ( collective_wait ) CALL MPI_BARRIER( comm1dx, ierr )
303	CALL MPI_ALLREDUCE( volume_flow_l(2), volume_flow(2), 1, MPI_REAL, &
304	MPI_SUM, comm1dx, ierr )
305	#else
306	volume_flow = volume_flow_l
307	#endif
308	volume_flow_offset(2) = ( volume_flow_initial(2) - volume_flow(2) ) &
309	/ volume_flow_area(2)
310
311	DO i = nxlg, nxrg
312	DO k = nzb_v_inner(j,i)+1, nzt
313	v(k,j,i) = v(k,j,i) + volume_flow_offset(2)
314	ENDDO
315	ENDDO
316
317	ENDIF
318
319	!
320	!-- Remove mean vertical velocity in case that Neumann conditions are
321	!-- used both at bottom and top boundary, and if not a nested domain in a
322	!-- normal nesting run. In case of vertical nesting, this must be done.
323	!-- Therefore an auxiliary logical variable nest_domain_nvn is used here, and
324	!-- nvn stands for non-vertical nesting.
325	!-- This cannot be done before the first initial time step because ngp_2dh_outer
326	!-- is not yet known then.
327	nest_domain_nvn = nest_domain
328	IF ( nest_domain .AND. nesting_mode == 'vertical' ) THEN
329	nest_domain_nvn = .FALSE.
330	ENDIF
331
332	IF ( ibc_p_b == 1 .AND. ibc_p_t == 1 .AND. &
333	.NOT. nest_domain_nvn .AND. intermediate_timestep_count /= 0 ) &
334	THEN
335	w_l = 0.0_wp; w_l_l = 0.0_wp
336	DO i = nxl, nxr
337	DO j = nys, nyn
338	DO k = nzb_w_inner(j,i)+1, nzt
339	w_l_l(k) = w_l_l(k) + w(k,j,i)
340	ENDDO
341	ENDDO
342	ENDDO
343	#if defined( __parallel )
344	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
345	CALL MPI_ALLREDUCE( w_l_l(1), w_l(1), nzt, MPI_REAL, MPI_SUM, &
346	comm2d, ierr )
347	#else
348	w_l = w_l_l
349	#endif
350	DO k = 1, nzt
351	w_l(k) = w_l(k) / ngp_2dh_outer(k,0)
352	ENDDO
353	DO i = nxlg, nxrg
354	DO j = nysg, nyng
355	DO k = nzb_w_inner(j,i)+1, nzt
356	w(k,j,i) = w(k,j,i) - w_l(k)
357	ENDDO
358	ENDDO
359	ENDDO
360	ENDIF
361
362	!
363	!-- Compute the divergence of the provisional velocity field.
364	CALL cpu_log( log_point_s(1), 'divergence', 'start' )
365
366	IF ( psolver(1:9) == 'multigrid' ) THEN
367	!$OMP PARALLEL DO SCHEDULE( STATIC )
368	DO i = nxl-1, nxr+1
369	DO j = nys-1, nyn+1
370	DO k = nzb, nzt+1
371	d(k,j,i) = 0.0_wp
372	ENDDO
373	ENDDO
374	ENDDO
375	ELSE
376	!$OMP PARALLEL DO SCHEDULE( STATIC )
377	!$acc kernels present( d )
378	DO i = nxl, nxr
379	DO j = nys, nyn
380	DO k = nzb+1, nzt
381	d(k,j,i) = 0.0_wp
382	ENDDO
383	ENDDO
384	ENDDO
385	!$acc end kernels
386	ENDIF
387
388	localsum = 0.0_wp
389	threadsum = 0.0_wp
390
391	#if defined( __ibm )
392	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
393	!$OMP DO SCHEDULE( STATIC )
394	DO i = nxl, nxr
395	DO j = nys, nyn
396	DO k = nzb_s_inner(j,i)+1, nzt
397	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
398	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
399	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k) ) * ddt_3d &
400	* d_weight_pres
401	ENDDO
402	!
403	!-- Compute possible PE-sum of divergences for flow_statistics
404	DO k = nzb_s_inner(j,i)+1, nzt
405	threadsum = threadsum + ABS( d(k,j,i) )
406	ENDDO
407
408	ENDDO
409	ENDDO
410
411	IF ( intermediate_timestep_count == intermediate_timestep_count_max .OR. &
412	intermediate_timestep_count == 0 ) THEN
413	localsum = localsum + threadsum * dt_3d * weight_pres_l
414	ENDIF
415	!$OMP END PARALLEL
416	#else
417
418	!$OMP PARALLEL PRIVATE (i,j,k)
419	!$OMP DO SCHEDULE( STATIC )
420	!$acc kernels present( d, ddzw, rflags_s_inner, u, v, w )
421	!$acc loop collapse( 3 )
422	DO i = nxl, nxr
423	DO j = nys, nyn
424	DO k = 1, nzt
425	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
426	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
427	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k) ) * ddt_3d &
428	* d_weight_pres * rflags_s_inner(k,j,i)
429	ENDDO
430	ENDDO
431	ENDDO
432	!$acc end kernels
433	!$OMP END PARALLEL
434
435	!
436	!-- Compute possible PE-sum of divergences for flow_statistics. Carry out
437	!-- computation only at last Runge-Kutta substep.
438	IF ( intermediate_timestep_count == intermediate_timestep_count_max .OR. &
439	intermediate_timestep_count == 0 ) THEN
440	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
441	!$OMP DO SCHEDULE( STATIC )
442	!$acc parallel loop collapse(3) present( d ) reduction(+:threadsum)
443	DO i = nxl, nxr
444	DO j = nys, nyn
445	DO k = nzb+1, nzt
446	threadsum = threadsum + ABS( d(k,j,i) )
447	ENDDO
448	ENDDO
449	ENDDO
450	!$acc end parallel loop
451	localsum = localsum + threadsum * dt_3d * weight_pres_l
452	!$OMP END PARALLEL
453	ENDIF
454	#endif
455
456	!
457	!-- For completeness, set the divergence sum of all statistic regions to those
458	!-- of the total domain
459	IF ( intermediate_timestep_count == intermediate_timestep_count_max .OR. &
460	intermediate_timestep_count == 0 ) THEN
461	sums_divold_l(0:statistic_regions) = localsum
462	ENDIF
463
464	CALL cpu_log( log_point_s(1), 'divergence', 'stop' )
465
466	!
467	!-- Compute the pressure perturbation solving the Poisson equation
468	IF ( psolver == 'poisfft' ) THEN
469
470	!
471	!-- Solve Poisson equation via FFT and solution of tridiagonal matrices
472	CALL poisfft( d )
473
474	!
475	!-- Store computed perturbation pressure and set boundary condition in
476	!-- z-direction
477	!$OMP PARALLEL DO
478	!$acc kernels present( d, tend )
479	DO i = nxl, nxr
480	DO j = nys, nyn
481	DO k = nzb+1, nzt
482	tend(k,j,i) = d(k,j,i)
483	ENDDO
484	ENDDO
485	ENDDO
486	!$acc end kernels
487
488	!
489	!-- Bottom boundary:
490	!-- This condition is only required for internal output. The pressure
491	!-- gradient (dp(nzb+1)-dp(nzb))/dz is not used anywhere else.
492	IF ( ibc_p_b == 1 ) THEN
493	!
494	!-- Neumann (dp/dz = 0)
495	!$OMP PARALLEL DO
496	!$acc kernels present( nzb_s_inner, tend )
497	DO i = nxlg, nxrg
498	DO j = nysg, nyng
499	tend(nzb_s_inner(j,i),j,i) = tend(nzb_s_inner(j,i)+1,j,i)
500	ENDDO
501	ENDDO
502	!$acc end kernels
503
504	ELSE
505	!
506	!-- Dirichlet
507	!$OMP PARALLEL DO
508	!$acc kernels present( tend )
509	DO i = nxlg, nxrg
510	DO j = nysg, nyng
511	tend(nzb_s_inner(j,i),j,i) = 0.0_wp
512	ENDDO
513	ENDDO
514	!$acc end kernels
515
516	ENDIF
517
518	!
519	!-- Top boundary
520	IF ( ibc_p_t == 1 ) THEN
521	!
522	!-- Neumann
523	!$OMP PARALLEL DO
524	!$acc kernels present( tend )
525	DO i = nxlg, nxrg
526	DO j = nysg, nyng
527	tend(nzt+1,j,i) = tend(nzt,j,i)
528	ENDDO
529	ENDDO
530	!$acc end kernels
531
532	ELSE
533	!
534	!-- Dirichlet
535	!$OMP PARALLEL DO
536	!$acc kernels present( tend )
537	DO i = nxlg, nxrg
538	DO j = nysg, nyng
539	tend(nzt+1,j,i) = 0.0_wp
540	ENDDO
541	ENDDO
542	!$acc end kernels
543
544	ENDIF
545
546	!
547	!-- Exchange boundaries for p
548	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
549	on_device = .TRUE. ! to be removed after complete porting
550	ELSE ! of ghost point exchange
551	!$acc update host( tend )
552	ENDIF
553	CALL exchange_horiz( tend, nbgp )
554	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
555	on_device = .FALSE. ! to be removed after complete porting
556	ELSE ! of ghost point exchange
557	!$acc update device( tend )
558	ENDIF
559
560	ELSEIF ( psolver == 'sor' ) THEN
561
562	!
563	!-- Solve Poisson equation for perturbation pressure using SOR-Red/Black
564	!-- scheme
565	CALL sor( d, ddzu_pres, ddzw, p_loc )
566	tend = p_loc
567
568	ELSEIF ( psolver(1:9) == 'multigrid' ) THEN
569
570	!
571	!-- Solve Poisson equation for perturbation pressure using Multigrid scheme,
572	!-- array tend is used to store the residuals, logical exchange_mg is used
573	!-- to discern data exchange in multigrid ( 1 ghostpoint ) and normal grid
574	!-- ( nbgp ghost points ).
575
576	!-- If the number of grid points of the gathered grid, which is collected
577	!-- on PE0, is larger than the number of grid points of an PE, than array
578	!-- tend will be enlarged.
579	IF ( gathered_size > subdomain_size ) THEN
580	DEALLOCATE( tend )
581	ALLOCATE( tend(nzb:nzt_mg(mg_switch_to_pe0_level)+1,nys_mg( &
582	mg_switch_to_pe0_level)-1:nyn_mg(mg_switch_to_pe0_level)+1,&
583	nxl_mg(mg_switch_to_pe0_level)-1:nxr_mg( &
584	mg_switch_to_pe0_level)+1) )
585	ENDIF
586
587	IF ( psolver == 'multigrid' ) THEN
588	CALL poismg( tend )
589	ELSE
590	CALL poismg_noopt( tend )
591	ENDIF
592
593	IF ( gathered_size > subdomain_size ) THEN
594	DEALLOCATE( tend )
595	ALLOCATE( tend(nzb:nzt+1,nysg:nyng,nxlg:nxrg) )
596	ENDIF
597
598	!
599	!-- Restore perturbation pressure on tend because this array is used
600	!-- further below to correct the velocity fields
601	DO i = nxl-1, nxr+1
602	DO j = nys-1, nyn+1
603	DO k = nzb, nzt+1
604	tend(k,j,i) = p_loc(k,j,i)
605	ENDDO
606	ENDDO
607	ENDDO
608
609	ENDIF
610
611	!
612	!-- Store perturbation pressure on array p, used for pressure data output.
613	!-- Ghost layers are added in the output routines (except sor-method: see below)
614	IF ( intermediate_timestep_count <= 1 ) THEN
615	!$OMP PARALLEL PRIVATE (i,j,k)
616	!$OMP DO
617	!$acc kernels present( p, tend, weight_substep_l )
618	!$acc loop independent
619	DO i = nxl-1, nxr+1
620	!$acc loop independent
621	DO j = nys-1, nyn+1
622	!$acc loop independent
623	DO k = nzb, nzt+1
624	p(k,j,i) = tend(k,j,i) * &
625	weight_substep_l
626	ENDDO
627	ENDDO
628	ENDDO
629	!$acc end kernels
630	!$OMP END PARALLEL
631
632	ELSEIF ( intermediate_timestep_count > 1 ) THEN
633	!$OMP PARALLEL PRIVATE (i,j,k)
634	!$OMP DO
635	!$acc kernels present( p, tend, weight_substep_l )
636	!$acc loop independent
637	DO i = nxl-1, nxr+1
638	!$acc loop independent
639	DO j = nys-1, nyn+1
640	!$acc loop independent
641	DO k = nzb, nzt+1
642	p(k,j,i) = p(k,j,i) + tend(k,j,i) * &
643	weight_substep_l
644	ENDDO
645	ENDDO
646	ENDDO
647	!$acc end kernels
648	!$OMP END PARALLEL
649
650	ENDIF
651
652	!
653	!-- SOR-method needs ghost layers for the next timestep
654	IF ( psolver == 'sor' ) CALL exchange_horiz( p, nbgp )
655
656	!
657	!-- Correction of the provisional velocities with the current perturbation
658	!-- pressure just computed
659	IF ( conserve_volume_flow .AND. ( bc_lr_cyc .OR. bc_ns_cyc ) ) THEN
660	volume_flow_l(1) = 0.0_wp
661	volume_flow_l(2) = 0.0_wp
662	ENDIF
663
664	!$OMP PARALLEL PRIVATE (i,j,k)
665	!$OMP DO
666	!$acc kernels present( ddzu, nzb_u_inner, nzb_v_inner, nzb_w_inner, tend, u, v, w )
667	!$acc loop independent
668	DO i = nxl, nxr
669	!$acc loop independent
670	DO j = nys, nyn
671	!$acc loop independent
672	DO k = 1, nzt
673	IF ( k > nzb_w_inner(j,i) ) THEN
674	w(k,j,i) = w(k,j,i) - dt_3d * &
675	( tend(k+1,j,i) - tend(k,j,i) ) * ddzu(k+1) * &
676	weight_pres_l
677	ENDIF
678	ENDDO
679	!$acc loop independent
680	DO k = 1, nzt
681	IF ( k > nzb_u_inner(j,i) ) THEN
682	u(k,j,i) = u(k,j,i) - dt_3d * &
683	( tend(k,j,i) - tend(k,j,i-1) ) * ddx * &
684	weight_pres_l
685	ENDIF
686	ENDDO
687	!$acc loop independent
688	DO k = 1, nzt
689	IF ( k > nzb_v_inner(j,i) ) THEN
690	v(k,j,i) = v(k,j,i) - dt_3d * &
691	( tend(k,j,i) - tend(k,j-1,i) ) * ddy * &
692	weight_pres_l
693	ENDIF
694	ENDDO
695
696	ENDDO
697	ENDDO
698	!$acc end kernels
699	!$OMP END PARALLEL
700
701	!
702	!-- Sum up the volume flow through the right and north boundary
703	IF ( conserve_volume_flow .AND. bc_lr_cyc .AND. bc_ns_cyc .AND. &
704	nxr == nx ) THEN
705
706	!$OMP PARALLEL PRIVATE (j,k)
707	!$OMP DO
708	DO j = nys, nyn
709	!$OMP CRITICAL
710	DO k = nzb_u_inner(j,nx) + 1, nzt
711	volume_flow_l(1) = volume_flow_l(1) + u(k,j,nx) * dzw(k)
712	ENDDO
713	!$OMP END CRITICAL
714	ENDDO
715	!$OMP END PARALLEL
716
717	ENDIF
718
719	IF ( conserve_volume_flow .AND. bc_ns_cyc .AND. bc_lr_cyc .AND. &
720	nyn == ny ) THEN
721
722	!$OMP PARALLEL PRIVATE (i,k)
723	!$OMP DO
724	DO i = nxl, nxr
725	!$OMP CRITICAL
726	DO k = nzb_v_inner(ny,i) + 1, nzt
727	volume_flow_l(2) = volume_flow_l(2) + v(k,ny,i) * dzw(k)
728	ENDDO
729	!$OMP END CRITICAL
730	ENDDO
731	!$OMP END PARALLEL
732
733	ENDIF
734
735	!
736	!-- Conserve the volume flow
737	IF ( conserve_volume_flow .AND. ( bc_lr_cyc .AND. bc_ns_cyc ) ) THEN
738
739	#if defined( __parallel )
740	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
741	CALL MPI_ALLREDUCE( volume_flow_l(1), volume_flow(1), 2, MPI_REAL, &
742	MPI_SUM, comm2d, ierr )
743	#else
744	volume_flow = volume_flow_l
745	#endif
746
747	volume_flow_offset(1:2) = ( volume_flow_initial(1:2) - volume_flow(1:2) ) / &
748	volume_flow_area(1:2)
749
750	!$OMP PARALLEL PRIVATE (i,j,k)
751	!$OMP DO
752	DO i = nxl, nxr
753	DO j = nys, nyn
754	DO k = nzb_u_inner(j,i) + 1, nzt
755	u(k,j,i) = u(k,j,i) + volume_flow_offset(1)
756	ENDDO
757	DO k = nzb_v_inner(j,i) + 1, nzt
758	v(k,j,i) = v(k,j,i) + volume_flow_offset(2)
759	ENDDO
760	ENDDO
761	ENDDO
762
763	!$OMP END PARALLEL
764
765	ENDIF
766
767	!
768	!-- Exchange of boundaries for the velocities
769	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
770	on_device = .TRUE. ! to be removed after complete porting
771	ELSE ! of ghost point exchange
772	!$acc update host( u, v, w )
773	ENDIF
774	CALL exchange_horiz( u, nbgp )
775	CALL exchange_horiz( v, nbgp )
776	CALL exchange_horiz( w, nbgp )
777	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
778	on_device = .FALSE. ! to be removed after complete porting
779	ELSE ! of ghost point exchange
780	!$acc update device( u, v, w )
781	ENDIF
782
783	!
784	!-- Compute the divergence of the corrected velocity field,
785	!-- A possible PE-sum is computed in flow_statistics. Carry out computation
786	!-- only at last Runge-Kutta step.
787	IF ( intermediate_timestep_count == intermediate_timestep_count_max .OR. &
788	intermediate_timestep_count == 0 ) THEN
789	CALL cpu_log( log_point_s(1), 'divergence', 'start' )
790	sums_divnew_l = 0.0_wp
791
792	!
793	!-- d must be reset to zero because it can contain nonzero values below the
794	!-- topography
795	IF ( topography /= 'flat' ) d = 0.0_wp
796
797	localsum = 0.0_wp
798	threadsum = 0.0_wp
799
800	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
801	!$OMP DO SCHEDULE( STATIC )
802	#if defined( __ibm )
803	DO i = nxl, nxr
804	DO j = nys, nyn
805	DO k = nzb_s_inner(j,i)+1, nzt
806	d(k,j,i) = ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
807	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
808	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k)
809	ENDDO
810	DO k = nzb+1, nzt
811	threadsum = threadsum + ABS( d(k,j,i) )
812	ENDDO
813	ENDDO
814	ENDDO
815	#else
816	!$acc kernels present( d, ddzw, rflags_s_inner, u, v, w )
817	!$acc loop collapse( 3 )
818	DO i = nxl, nxr
819	DO j = nys, nyn
820	DO k = 1, nzt
821	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
822	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
823	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k) &
824	) * rflags_s_inner(k,j,i)
825	ENDDO
826	ENDDO
827	ENDDO
828	!$acc end kernels
829	!
830	!-- Compute possible PE-sum of divergences for flow_statistics
831	!$acc parallel loop collapse(3) present( d ) reduction(+:threadsum)
832	DO i = nxl, nxr
833	DO j = nys, nyn
834	DO k = nzb+1, nzt
835	threadsum = threadsum + ABS( d(k,j,i) )
836	ENDDO
837	ENDDO
838	ENDDO
839	!$acc end parallel loop
840	#endif
841
842	localsum = localsum + threadsum
843	!$OMP END PARALLEL
844
845	!
846	!-- For completeness, set the divergence sum of all statistic regions to those
847	!-- of the total domain
848	sums_divnew_l(0:statistic_regions) = localsum
849
850	CALL cpu_log( log_point_s(1), 'divergence', 'stop' )
851
852	ENDIF
853
854	CALL cpu_log( log_point(8), 'pres', 'stop' )
855
856
857	END SUBROUTINE pres

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

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |