Home

Context Navigation

source: palm/trunk/SOURCE/pres.f90 @ 1320

Last change on this file since 1320 was 1320, checked in by raasch, 10 years ago
ONLY-attribute added to USE-statements, kind-parameters added to all INTEGER and REAL declaration statements, kinds are defined in new module kinds, old module precision_kind is removed, revision history before 2012 removed, comment fields (!:) to be used for variable explanations added to all variable declaration statements
Property svn:keywords set to `Id`
File size: 23.8 KB

Rev	Line
[1]	1	SUBROUTINE pres
	2
[1036]	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	!
[1310]	17	! Copyright 1997-2014 Leibniz Universitaet Hannover
[1036]	18	!--------------------------------------------------------------------------------!
	19	!
[484]	20	! Current revisions:
[1212]	21	! ------------------
[1320]	22	! ONLY-attribute added to USE-statements,
	23	! kind-parameters added to all INTEGER and REAL declaration statements,
	24	! kinds are defined in new module kinds,
	25	! old module precision_kind is removed,
	26	! revision history before 2012 removed,
	27	! comment fields (!:) to be used for variable explanations added to
	28	! all variable declaration statements
[708]	29	!
	30	! Former revisions:
	31	! -----------------
	32	! $Id: pres.f90 1320 2014-03-20 08:40:49Z raasch $
	33	!
[1319]	34	! 1318 2014-03-17 13:35:16Z raasch
	35	! module interfaces removed
	36	!
[1307]	37	! 1306 2014-03-13 14:30:59Z raasch
	38	! second argument removed from routine poisfft
	39	!
[1258]	40	! 1257 2013-11-08 15:18:40Z raasch
	41	! openacc loop and loop vector clauses removed, independent clauses added,
	42	! end parallel replaced by end parallel loop
	43	!
[1222]	44	! 1221 2013-09-10 08:59:13Z raasch
	45	! openACC porting of reduction operations, loops for calculating d are
	46	! using the rflags_s_inner multiply flag instead of the nzb_s_inner loop index
	47	!
[1213]	48	! 1212 2013-08-15 08:46:27Z raasch
	49	! call of poisfft_hybrid removed
	50	!
[1118]	51	! 1117 2013-03-27 11:15:36Z suehring
	52	! Bugfix in OpenMP parallelization.
	53	!
[1114]	54	! 1113 2013-03-10 02:48:14Z raasch
	55	! GPU-porting of several loops, some loops rearranged
	56	!
[1112]	57	! 1111 2013-03-08 23:54:10Z
	58	! openACC statements added,
	59	! ibc_p_b = 2 removed
	60	!
[1093]	61	! 1092 2013-02-02 11:24:22Z raasch
	62	! unused variables removed
	63	!
[1037]	64	! 1036 2012-10-22 13:43:42Z raasch
	65	! code put under GPL (PALM 3.9)
	66	!
[1004]	67	! 1003 2012-09-14 14:35:53Z raasch
	68	! adjustment of array tend for cases with unequal subdomain sizes removed
	69	!
[1]	70	! Revision 1.1 1997/07/24 11:24:44 raasch
	71	! Initial revision
	72	!
	73	!
	74	! Description:
	75	! ------------
	76	! Compute the divergence of the provisional velocity field. Solve the Poisson
	77	! equation for the perturbation pressure. Compute the final velocities using
	78	! this perturbation pressure. Compute the remaining divergence.
	79	!------------------------------------------------------------------------------!
	80
[1320]	81	USE arrays_3d, &
	82	ONLY: d, ddzu, ddzu_pres, ddzw, dzw, p, p_loc, tend, u, v, w
	83
	84	USE control_parameters, &
	85	ONLY: bc_lr_cyc, bc_ns_cyc, conserve_volume_flow, dt_3d, &
	86	gathered_size, ibc_p_b, ibc_p_t, intermediate_timestep_count, &
	87	mg_switch_to_pe0_level, on_device, outflow_l, outflow_n, &
	88	outflow_r, outflow_s, psolver, simulated_time, subdomain_size, &
	89	topography, volume_flow, volume_flow_area, volume_flow_initial
	90
	91	USE cpulog, &
	92	ONLY: cpu_log, log_point, log_point_s
	93
	94	USE grid_variables, &
	95	ONLY: ddx, ddy
	96
	97	USE indices, &
	98	ONLY: nbgp, ngp_2dh_outer, nx, nxl, nxlg, nxl_mg, nxr, nxrg, nxr_mg, &
	99	ny, nys, nysg, nys_mg, nyn, nyng, nyn_mg, nzb, nzb_s_inner, &
	100	nzb_u_inner, nzb_v_inner, nzb_w_inner, nzb_2d, nzt, nzt_mg, &
	101	rflags_s_inner
	102
	103	USE kinds
	104
[1]	105	USE pegrid
	106
[1320]	107	USE poisfft_mod, &
	108	ONLY: poisfft
	109
	110	USE statistics, &
	111	ONLY: statistic_regions, sums_divnew_l, sums_divold_l, weight_pres, &
	112	weight_substep
	113
[1]	114	IMPLICIT NONE
	115
[1320]	116	INTEGER(iwp) :: i !:
	117	INTEGER(iwp) :: j !:
	118	INTEGER(iwp) :: k !:
[1]	119
[1320]	120	REAL(wp) :: ddt_3d !:
	121	REAL(wp) :: localsum !:
	122	REAL(wp) :: threadsum !:
	123	REAL(wp) :: d_weight_pres !:
[1]	124
[1320]	125	REAL(wp), DIMENSION(1:2) :: volume_flow_l !:
	126	REAL(wp), DIMENSION(1:2) :: volume_flow_offset !:
	127	REAL(wp), DIMENSION(1:nzt) :: w_l !:
	128	REAL(wp), DIMENSION(1:nzt) :: w_l_l !:
[1]	129
	130
	131	CALL cpu_log( log_point(8), 'pres', 'start' )
	132
[85]	133
	134	ddt_3d = 1.0 / dt_3d
[709]	135	d_weight_pres = 1.0 / weight_pres(intermediate_timestep_count)
[85]	136
[1]	137	!
[707]	138	!-- Multigrid method expects array d to have one ghost layer.
	139	!--
[1]	140	IF ( psolver == 'multigrid' ) THEN
[667]	141
[1]	142	DEALLOCATE( d )
[667]	143	ALLOCATE( d(nzb:nzt+1,nys-1:nyn+1,nxl-1:nxr+1) )
[707]	144
	145	!
	146	!-- Since p is later used to hold the weighted average of the substeps, it
	147	!-- cannot be used in the iterative solver. Therefore, its initial value is
	148	!-- stored on p_loc, which is then iteratively advanced in every substep.
	149	IF ( intermediate_timestep_count == 1 ) THEN
	150	DO i = nxl-1, nxr+1
	151	DO j = nys-1, nyn+1
	152	DO k = nzb, nzt+1
	153	p_loc(k,j,i) = p(k,j,i)
	154	ENDDO
	155	ENDDO
	156	ENDDO
[667]	157	ENDIF
	158
[707]	159	ELSEIF ( psolver == 'sor' .AND. intermediate_timestep_count == 1 ) THEN
	160
	161	!
	162	!-- Since p is later used to hold the weighted average of the substeps, it
	163	!-- cannot be used in the iterative solver. Therefore, its initial value is
	164	!-- stored on p_loc, which is then iteratively advanced in every substep.
	165	p_loc = p
	166
[1]	167	ENDIF
	168
	169	!
[75]	170	!-- Conserve the volume flow at the outflow in case of non-cyclic lateral
	171	!-- boundary conditions
[106]	172	!-- WARNING: so far, this conservation does not work at the left/south
	173	!-- boundary if the topography at the inflow differs from that at the
	174	!-- outflow! For this case, volume_flow_area needs adjustment!
	175	!
	176	!-- Left/right
[709]	177	IF ( conserve_volume_flow .AND. ( outflow_l .OR. outflow_r ) ) THEN
[680]	178
[106]	179	volume_flow(1) = 0.0
	180	volume_flow_l(1) = 0.0
	181
	182	IF ( outflow_l ) THEN
	183	i = 0
	184	ELSEIF ( outflow_r ) THEN
	185	i = nx+1
	186	ENDIF
	187
	188	DO j = nys, nyn
	189	!
	190	!-- Sum up the volume flow through the south/north boundary
[709]	191	DO k = nzb_2d(j,i)+1, nzt
[667]	192	volume_flow_l(1) = volume_flow_l(1) + u(k,j,i) * dzw(k)
[106]	193	ENDDO
	194	ENDDO
	195
	196	#if defined( __parallel )
[680]	197	IF ( collective_wait ) CALL MPI_BARRIER( comm1dy, ierr )
[106]	198	CALL MPI_ALLREDUCE( volume_flow_l(1), volume_flow(1), 1, MPI_REAL, &
	199	MPI_SUM, comm1dy, ierr )
	200	#else
	201	volume_flow = volume_flow_l
	202	#endif
[709]	203	volume_flow_offset(1) = ( volume_flow_initial(1) - volume_flow(1) ) &
[106]	204	/ volume_flow_area(1)
	205
[667]	206	DO j = nysg, nyng
[709]	207	DO k = nzb_2d(j,i)+1, nzt
[106]	208	u(k,j,i) = u(k,j,i) + volume_flow_offset(1)
	209	ENDDO
	210	ENDDO
	211
	212	ENDIF
	213
	214	!
	215	!-- South/north
[709]	216	IF ( conserve_volume_flow .AND. ( outflow_n .OR. outflow_s ) ) THEN
[106]	217
[75]	218	volume_flow(2) = 0.0
	219	volume_flow_l(2) = 0.0
	220
[106]	221	IF ( outflow_s ) THEN
	222	j = 0
	223	ELSEIF ( outflow_n ) THEN
[75]	224	j = ny+1
[106]	225	ENDIF
	226
	227	DO i = nxl, nxr
[75]	228	!
[106]	229	!-- Sum up the volume flow through the south/north boundary
[709]	230	DO k = nzb_2d(j,i)+1, nzt
[667]	231	volume_flow_l(2) = volume_flow_l(2) + v(k,j,i) * dzw(k)
[75]	232	ENDDO
[106]	233	ENDDO
	234
[75]	235	#if defined( __parallel )
[680]	236	IF ( collective_wait ) CALL MPI_BARRIER( comm1dx, ierr )
[75]	237	CALL MPI_ALLREDUCE( volume_flow_l(2), volume_flow(2), 1, MPI_REAL, &
	238	MPI_SUM, comm1dx, ierr )
	239	#else
	240	volume_flow = volume_flow_l
	241	#endif
	242	volume_flow_offset(2) = ( volume_flow_initial(2) - volume_flow(2) ) &
[106]	243	/ volume_flow_area(2)
[75]	244
[667]	245	DO i = nxlg, nxrg
[709]	246	DO k = nzb_v_inner(j,i)+1, nzt
[106]	247	v(k,j,i) = v(k,j,i) + volume_flow_offset(2)
[75]	248	ENDDO
[106]	249	ENDDO
[75]	250
	251	ENDIF
	252
[76]	253	!
	254	!-- Remove mean vertical velocity
	255	IF ( ibc_p_b == 1 .AND. ibc_p_t == 1 ) THEN
[709]	256	IF ( simulated_time > 0.0 ) THEN ! otherwise nzb_w_inner not yet known
[76]	257	w_l = 0.0; w_l_l = 0.0
	258	DO i = nxl, nxr
	259	DO j = nys, nyn
	260	DO k = nzb_w_inner(j,i)+1, nzt
	261	w_l_l(k) = w_l_l(k) + w(k,j,i)
	262	ENDDO
	263	ENDDO
	264	ENDDO
	265	#if defined( __parallel )
[622]	266	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
[709]	267	CALL MPI_ALLREDUCE( w_l_l(1), w_l(1), nzt, MPI_REAL, MPI_SUM, &
	268	comm2d, ierr )
[76]	269	#else
	270	w_l = w_l_l
	271	#endif
	272	DO k = 1, nzt
	273	w_l(k) = w_l(k) / ngp_2dh_outer(k,0)
	274	ENDDO
[667]	275	DO i = nxlg, nxrg
	276	DO j = nysg, nyng
[76]	277	DO k = nzb_w_inner(j,i)+1, nzt
	278	w(k,j,i) = w(k,j,i) - w_l(k)
	279	ENDDO
	280	ENDDO
	281	ENDDO
	282	ENDIF
	283	ENDIF
[75]	284
	285	!
[1]	286	!-- Compute the divergence of the provisional velocity field.
	287	CALL cpu_log( log_point_s(1), 'divergence', 'start' )
	288
	289	IF ( psolver == 'multigrid' ) THEN
	290	!$OMP PARALLEL DO SCHEDULE( STATIC )
	291	DO i = nxl-1, nxr+1
	292	DO j = nys-1, nyn+1
	293	DO k = nzb, nzt+1
	294	d(k,j,i) = 0.0
	295	ENDDO
	296	ENDDO
	297	ENDDO
	298	ELSE
	299	!$OMP PARALLEL DO SCHEDULE( STATIC )
[1111]	300	!$acc kernels present( d )
[1003]	301	DO i = nxl, nxr
	302	DO j = nys, nyn
	303	DO k = nzb+1, nzt
[1]	304	d(k,j,i) = 0.0
	305	ENDDO
	306	ENDDO
	307	ENDDO
[1111]	308	!$acc end kernels
[1]	309	ENDIF
	310
	311	localsum = 0.0
	312	threadsum = 0.0
	313
	314	#if defined( __ibm )
	315	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
	316	!$OMP DO SCHEDULE( STATIC )
	317	DO i = nxl, nxr
	318	DO j = nys, nyn
	319	DO k = nzb_s_inner(j,i)+1, nzt
[85]	320	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
	321	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
[673]	322	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k) ) * ddt_3d &
	323	* d_weight_pres
[1]	324	ENDDO
	325	!
	326	!-- Compute possible PE-sum of divergences for flow_statistics
	327	DO k = nzb_s_inner(j,i)+1, nzt
	328	threadsum = threadsum + ABS( d(k,j,i) )
	329	ENDDO
	330
	331	ENDDO
	332	ENDDO
	333
[707]	334	localsum = localsum + threadsum * dt_3d * &
	335	weight_pres(intermediate_timestep_count)
[693]	336
[1]	337	!$OMP END PARALLEL
	338	#else
	339
[1111]	340	!$OMP PARALLEL PRIVATE (i,j,k)
	341	!$OMP DO SCHEDULE( STATIC )
[1221]	342	!$acc kernels present( d, ddzw, rflags_s_inner, u, v, w )
	343	!$acc loop collapse( 3 )
[1111]	344	DO i = nxl, nxr
	345	DO j = nys, nyn
	346	DO k = 1, nzt
[1221]	347	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
	348	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
	349	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k) ) * ddt_3d &
	350	* d_weight_pres * rflags_s_inner(k,j,i)
[1]	351	ENDDO
	352	ENDDO
[1111]	353	ENDDO
	354	!$acc end kernels
	355	!$OMP END PARALLEL
[1]	356
	357	!
	358	!-- Compute possible PE-sum of divergences for flow_statistics
	359	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
	360	!$OMP DO SCHEDULE( STATIC )
[1221]	361	!$acc parallel loop collapse(3) present( d ) reduction(+:threadsum)
[1]	362	DO i = nxl, nxr
	363	DO j = nys, nyn
	364	DO k = nzb+1, nzt
	365	threadsum = threadsum + ABS( d(k,j,i) )
	366	ENDDO
	367	ENDDO
	368	ENDDO
[1257]	369	!$acc end parallel loop
[707]	370	localsum = localsum + threadsum * dt_3d * &
	371	weight_pres(intermediate_timestep_count)
[1]	372	!$OMP END PARALLEL
	373	#endif
	374
	375	!
	376	!-- For completeness, set the divergence sum of all statistic regions to those
	377	!-- of the total domain
	378	sums_divold_l(0:statistic_regions) = localsum
	379
	380	CALL cpu_log( log_point_s(1), 'divergence', 'stop' )
	381
	382	!
	383	!-- Compute the pressure perturbation solving the Poisson equation
	384	IF ( psolver(1:7) == 'poisfft' ) THEN
	385
	386	!
	387	!-- Solve Poisson equation via FFT and solution of tridiagonal matrices
	388	IF ( psolver == 'poisfft' ) THEN
[1212]	389
[1306]	390	CALL poisfft( d )
[1113]	391
[1]	392	ENDIF
	393
	394	!
	395	!-- Store computed perturbation pressure and set boundary condition in
	396	!-- z-direction
	397	!$OMP PARALLEL DO
[1113]	398	!$acc kernels present( d, tend )
[1]	399	DO i = nxl, nxr
	400	DO j = nys, nyn
	401	DO k = nzb+1, nzt
	402	tend(k,j,i) = d(k,j,i)
	403	ENDDO
	404	ENDDO
	405	ENDDO
[1113]	406	!$acc end kernels
[1]	407
	408	!
	409	!-- Bottom boundary:
	410	!-- This condition is only required for internal output. The pressure
	411	!-- gradient (dp(nzb+1)-dp(nzb))/dz is not used anywhere else.
	412	IF ( ibc_p_b == 1 ) THEN
	413	!
	414	!-- Neumann (dp/dz = 0)
	415	!$OMP PARALLEL DO
[1113]	416	!$acc kernels present( nzb_s_inner, tend )
[667]	417	DO i = nxlg, nxrg
	418	DO j = nysg, nyng
[1]	419	tend(nzb_s_inner(j,i),j,i) = tend(nzb_s_inner(j,i)+1,j,i)
	420	ENDDO
	421	ENDDO
[1113]	422	!$acc end kernels
[1]	423
	424	ELSE
	425	!
	426	!-- Dirichlet
	427	!$OMP PARALLEL DO
[1113]	428	!$acc kernels present( tend )
[667]	429	DO i = nxlg, nxrg
	430	DO j = nysg, nyng
[1]	431	tend(nzb_s_inner(j,i),j,i) = 0.0
	432	ENDDO
	433	ENDDO
[1113]	434	!$acc end kernels
[1]	435
	436	ENDIF
	437
	438	!
	439	!-- Top boundary
	440	IF ( ibc_p_t == 1 ) THEN
	441	!
	442	!-- Neumann
	443	!$OMP PARALLEL DO
[1113]	444	!$acc kernels present( tend )
[667]	445	DO i = nxlg, nxrg
	446	DO j = nysg, nyng
[1]	447	tend(nzt+1,j,i) = tend(nzt,j,i)
	448	ENDDO
	449	ENDDO
[1113]	450	!$acc end kernels
[1]	451
	452	ELSE
	453	!
	454	!-- Dirichlet
	455	!$OMP PARALLEL DO
[1113]	456	!$acc kernels present( tend )
[667]	457	DO i = nxlg, nxrg
	458	DO j = nysg, nyng
[1]	459	tend(nzt+1,j,i) = 0.0
	460	ENDDO
	461	ENDDO
[1113]	462	!$acc end kernels
[1]	463
	464	ENDIF
	465
	466	!
	467	!-- Exchange boundaries for p
[1113]	468	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
	469	on_device = .TRUE. ! to be removed after complete porting
	470	ELSE ! of ghost point exchange
	471	!$acc update host( tend )
	472	ENDIF
[667]	473	CALL exchange_horiz( tend, nbgp )
[1113]	474	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
	475	on_device = .FALSE. ! to be removed after complete porting
	476	ELSE ! of ghost point exchange
	477	!$acc update device( tend )
	478	ENDIF
[1]	479
	480	ELSEIF ( psolver == 'sor' ) THEN
	481
	482	!
	483	!-- Solve Poisson equation for perturbation pressure using SOR-Red/Black
	484	!-- scheme
[707]	485	CALL sor( d, ddzu_pres, ddzw, p_loc )
	486	tend = p_loc
[1]	487
	488	ELSEIF ( psolver == 'multigrid' ) THEN
	489
	490	!
	491	!-- Solve Poisson equation for perturbation pressure using Multigrid scheme,
[667]	492	!-- array tend is used to store the residuals, logical exchange_mg is used
	493	!-- to discern data exchange in multigrid ( 1 ghostpoint ) and normal grid
	494	!-- ( nbgp ghost points ).
[778]	495
	496	!-- If the number of grid points of the gathered grid, which is collected
	497	!-- on PE0, is larger than the number of grid points of an PE, than array
	498	!-- tend will be enlarged.
	499	IF ( gathered_size > subdomain_size ) THEN
	500	DEALLOCATE( tend )
	501	ALLOCATE( tend(nzb:nzt_mg(mg_switch_to_pe0_level)+1,nys_mg( &
	502	mg_switch_to_pe0_level)-1:nyn_mg(mg_switch_to_pe0_level)+1,&
	503	nxl_mg(mg_switch_to_pe0_level)-1:nxr_mg( &
	504	mg_switch_to_pe0_level)+1) )
	505	ENDIF
	506
[1]	507	CALL poismg( tend )
[707]	508
[778]	509	IF ( gathered_size > subdomain_size ) THEN
	510	DEALLOCATE( tend )
	511	ALLOCATE( tend(nzb:nzt+1,nysg:nyng,nxlg:nxrg) )
	512	ENDIF
	513
[1]	514	!
	515	!-- Restore perturbation pressure on tend because this array is used
	516	!-- further below to correct the velocity fields
[707]	517	DO i = nxl-1, nxr+1
	518	DO j = nys-1, nyn+1
	519	DO k = nzb, nzt+1
	520	tend(k,j,i) = p_loc(k,j,i)
	521	ENDDO
	522	ENDDO
	523	ENDDO
[667]	524
[1]	525	ENDIF
	526
	527	!
[707]	528	!-- Store perturbation pressure on array p, used for pressure data output.
	529	!-- Ghost layers are added in the output routines (except sor-method: see below)
	530	IF ( intermediate_timestep_count == 1 ) THEN
	531	!$OMP PARALLEL PRIVATE (i,j,k)
	532	!$OMP DO
[1113]	533	!$acc kernels present( p, tend, weight_substep )
[1257]	534	!$acc loop independent
[707]	535	DO i = nxl-1, nxr+1
[1257]	536	!$acc loop independent
[707]	537	DO j = nys-1, nyn+1
[1257]	538	!$acc loop independent
[707]	539	DO k = nzb, nzt+1
	540	p(k,j,i) = tend(k,j,i) * &
	541	weight_substep(intermediate_timestep_count)
[673]	542	ENDDO
[1]	543	ENDDO
[707]	544	ENDDO
[1113]	545	!$acc end kernels
[707]	546	!$OMP END PARALLEL
	547
	548	ELSE
	549	!$OMP PARALLEL PRIVATE (i,j,k)
	550	!$OMP DO
[1113]	551	!$acc kernels present( p, tend, weight_substep )
[1257]	552	!$acc loop independent
[707]	553	DO i = nxl-1, nxr+1
[1257]	554	!$acc loop independent
[707]	555	DO j = nys-1, nyn+1
[1257]	556	!$acc loop independent
[707]	557	DO k = nzb, nzt+1
	558	p(k,j,i) = p(k,j,i) + tend(k,j,i) * &
	559	weight_substep(intermediate_timestep_count)
[673]	560	ENDDO
	561	ENDDO
[707]	562	ENDDO
[1113]	563	!$acc end kernels
[707]	564	!$OMP END PARALLEL
	565
	566	ENDIF
[673]	567
[707]	568	!
	569	!-- SOR-method needs ghost layers for the next timestep
	570	IF ( psolver == 'sor' ) CALL exchange_horiz( p, nbgp )
[682]	571
[1]	572	!
	573	!-- Correction of the provisional velocities with the current perturbation
	574	!-- pressure just computed
[709]	575	IF ( conserve_volume_flow .AND. ( bc_lr_cyc .OR. bc_ns_cyc ) ) THEN
[1]	576	volume_flow_l(1) = 0.0
	577	volume_flow_l(2) = 0.0
	578	ENDIF
[707]	579
[1]	580	!$OMP PARALLEL PRIVATE (i,j,k)
	581	!$OMP DO
[1113]	582	!$acc kernels present( ddzu, nzb_u_inner, nzb_v_inner, nzb_w_inner, tend, u, v, w, weight_pres )
[1257]	583	!$acc loop independent
[673]	584	DO i = nxl, nxr
[1257]	585	!$acc loop independent
[1]	586	DO j = nys, nyn
[1257]	587	!$acc loop independent
[1113]	588	DO k = 1, nzt
	589	IF ( k > nzb_w_inner(j,i) ) THEN
	590	w(k,j,i) = w(k,j,i) - dt_3d * &
	591	( tend(k+1,j,i) - tend(k,j,i) ) * ddzu(k+1) * &
	592	weight_pres(intermediate_timestep_count)
	593	ENDIF
[1]	594	ENDDO
[1257]	595	!$acc loop independent
[1113]	596	DO k = 1, nzt
	597	IF ( k > nzb_u_inner(j,i) ) THEN
	598	u(k,j,i) = u(k,j,i) - dt_3d * &
	599	( tend(k,j,i) - tend(k,j,i-1) ) * ddx * &
	600	weight_pres(intermediate_timestep_count)
	601	ENDIF
[1]	602	ENDDO
[1257]	603	!$acc loop independent
[1113]	604	DO k = 1, nzt
	605	IF ( k > nzb_v_inner(j,i) ) THEN
	606	v(k,j,i) = v(k,j,i) - dt_3d * &
	607	( tend(k,j,i) - tend(k,j-1,i) ) * ddy * &
	608	weight_pres(intermediate_timestep_count)
	609	ENDIF
[673]	610	ENDDO
[1]	611
	612	ENDDO
	613	ENDDO
[1113]	614	!$acc end kernels
[1]	615	!$OMP END PARALLEL
[1113]	616
	617	!
	618	!-- Sum up the volume flow through the right and north boundary
	619	IF ( conserve_volume_flow .AND. bc_lr_cyc .AND. bc_ns_cyc .AND. &
	620	nxr == nx ) THEN
	621
	622	!$OMP PARALLEL PRIVATE (j,k)
	623	!$OMP DO
	624	DO j = nys, nyn
	625	!$OMP CRITICAL
	626	DO k = nzb_2d(j,nx) + 1, nzt
	627	volume_flow_l(1) = volume_flow_l(1) + u(k,j,nx) * dzw(k)
	628	ENDDO
	629	!$OMP END CRITICAL
	630	ENDDO
	631	!$OMP END PARALLEL
	632
	633	ENDIF
	634
	635	IF ( conserve_volume_flow .AND. bc_ns_cyc .AND. bc_lr_cyc .AND. &
	636	nyn == ny ) THEN
	637
	638	!$OMP PARALLEL PRIVATE (i,k)
	639	!$OMP DO
	640	DO i = nxl, nxr
	641	!$OMP CRITICAL
	642	DO k = nzb_2d(ny,i) + 1, nzt
	643	volume_flow_l(2) = volume_flow_l(2) + v(k,ny,i) * dzw(k)
	644	ENDDO
	645	!$OMP END CRITICAL
	646	ENDDO
	647	!$OMP END PARALLEL
	648
	649	ENDIF
[673]	650
[1]	651	!
	652	!-- Conserve the volume flow
[707]	653	IF ( conserve_volume_flow .AND. ( bc_lr_cyc .AND. bc_ns_cyc ) ) THEN
[1]	654
	655	#if defined( __parallel )
[622]	656	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
[1]	657	CALL MPI_ALLREDUCE( volume_flow_l(1), volume_flow(1), 2, MPI_REAL, &
	658	MPI_SUM, comm2d, ierr )
	659	#else
	660	volume_flow = volume_flow_l
	661	#endif
	662
	663	volume_flow_offset = ( volume_flow_initial - volume_flow ) / &
	664	volume_flow_area
	665
	666	!$OMP PARALLEL PRIVATE (i,j,k)
	667	!$OMP DO
	668	DO i = nxl, nxr
	669	DO j = nys, nyn
[667]	670	DO k = nzb_u_inner(j,i) + 1, nzt
	671	u(k,j,i) = u(k,j,i) + volume_flow_offset(1)
[719]	672	ENDDO
	673	DO k = nzb_v_inner(j,i) + 1, nzt
[667]	674	v(k,j,i) = v(k,j,i) + volume_flow_offset(2)
	675	ENDDO
[1]	676	ENDDO
	677	ENDDO
[667]	678
[1]	679	!$OMP END PARALLEL
	680
	681	ENDIF
	682
	683	!
	684	!-- Exchange of boundaries for the velocities
[1113]	685	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
	686	on_device = .TRUE. ! to be removed after complete porting
	687	ELSE ! of ghost point exchange
	688	!$acc update host( u, v, w )
	689	ENDIF
[667]	690	CALL exchange_horiz( u, nbgp )
	691	CALL exchange_horiz( v, nbgp )
	692	CALL exchange_horiz( w, nbgp )
[1113]	693	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
	694	on_device = .FALSE. ! to be removed after complete porting
	695	ELSE ! of ghost point exchange
	696	!$acc update device( u, v, w )
	697	ENDIF
[1]	698
	699	!
	700	!-- Compute the divergence of the corrected velocity field,
	701	!-- a possible PE-sum is computed in flow_statistics
	702	CALL cpu_log( log_point_s(1), 'divergence', 'start' )
	703	sums_divnew_l = 0.0
	704
	705	!
	706	!-- d must be reset to zero because it can contain nonzero values below the
	707	!-- topography
	708	IF ( topography /= 'flat' ) d = 0.0
	709
	710	localsum = 0.0
	711	threadsum = 0.0
	712
	713	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
	714	!$OMP DO SCHEDULE( STATIC )
	715	#if defined( __ibm )
	716	DO i = nxl, nxr
	717	DO j = nys, nyn
	718	DO k = nzb_s_inner(j,i)+1, nzt
	719	d(k,j,i) = ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
	720	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
	721	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k)
	722	ENDDO
	723	DO k = nzb+1, nzt
	724	threadsum = threadsum + ABS( d(k,j,i) )
	725	ENDDO
	726	ENDDO
	727	ENDDO
	728	#else
[1221]	729	!$acc kernels present( d, ddzw, rflags_s_inner, u, v, w )
	730	!$acc loop collapse( 3 )
[1]	731	DO i = nxl, nxr
	732	DO j = nys, nyn
[1113]	733	DO k = 1, nzt
[1221]	734	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
	735	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
	736	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k) &
	737	) * rflags_s_inner(k,j,i)
[1113]	738	ENDDO
	739	ENDDO
	740	ENDDO
	741	!$acc end kernels
	742	!
	743	!-- Compute possible PE-sum of divergences for flow_statistics
[1221]	744	!$acc parallel loop collapse(3) present( d ) reduction(+:threadsum)
[1113]	745	DO i = nxl, nxr
	746	DO j = nys, nyn
[1221]	747	DO k = nzb+1, nzt
[1]	748	threadsum = threadsum + ABS( d(k,j,i) )
	749	ENDDO
	750	ENDDO
	751	ENDDO
[1257]	752	!$acc end parallel loop
[1]	753	#endif
[667]	754
[1]	755	localsum = localsum + threadsum
	756	!$OMP END PARALLEL
	757
	758	!
	759	!-- For completeness, set the divergence sum of all statistic regions to those
	760	!-- of the total domain
	761	sums_divnew_l(0:statistic_regions) = localsum
	762
	763	CALL cpu_log( log_point_s(1), 'divergence', 'stop' )
	764
	765	CALL cpu_log( log_point(8), 'pres', 'stop' )
	766
	767
	768	END SUBROUTINE pres

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

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |