Home

Context Navigation

source: palm/trunk/SOURCE/pres.f90 @ 1587

Last change on this file since 1587 was 1576, checked in by raasch, 10 years ago
last commit documented
Property svn:keywords set to `Id`
File size: 24.2 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	! ------------------
[1343]	22	!
[1576]	23	!
[1321]	24	! Former revisions:
	25	! -----------------
	26	! $Id: pres.f90 1576 2015-03-27 10:23:30Z maronga $
	27	!
[1576]	28	! 1575 2015-03-27 09:56:27Z raasch
	29	! poismg_fast + respective module added, adjustments for psolver-queries
	30	!
[1343]	31	! 1342 2014-03-26 17:04:47Z kanani
	32	! REAL constants defined as wp-kind
	33	!
[1321]	34	! 1320 2014-03-20 08:40:49Z raasch
[1320]	35	! ONLY-attribute added to USE-statements,
	36	! kind-parameters added to all INTEGER and REAL declaration statements,
	37	! kinds are defined in new module kinds,
	38	! old module precision_kind is removed,
	39	! revision history before 2012 removed,
	40	! comment fields (!:) to be used for variable explanations added to
	41	! all variable declaration statements
[708]	42	!
[1319]	43	! 1318 2014-03-17 13:35:16Z raasch
	44	! module interfaces removed
	45	!
[1307]	46	! 1306 2014-03-13 14:30:59Z raasch
	47	! second argument removed from routine poisfft
	48	!
[1258]	49	! 1257 2013-11-08 15:18:40Z raasch
	50	! openacc loop and loop vector clauses removed, independent clauses added,
	51	! end parallel replaced by end parallel loop
	52	!
[1222]	53	! 1221 2013-09-10 08:59:13Z raasch
	54	! openACC porting of reduction operations, loops for calculating d are
	55	! using the rflags_s_inner multiply flag instead of the nzb_s_inner loop index
	56	!
[1213]	57	! 1212 2013-08-15 08:46:27Z raasch
	58	! call of poisfft_hybrid removed
	59	!
[1118]	60	! 1117 2013-03-27 11:15:36Z suehring
	61	! Bugfix in OpenMP parallelization.
	62	!
[1114]	63	! 1113 2013-03-10 02:48:14Z raasch
	64	! GPU-porting of several loops, some loops rearranged
	65	!
[1112]	66	! 1111 2013-03-08 23:54:10Z
	67	! openACC statements added,
	68	! ibc_p_b = 2 removed
	69	!
[1093]	70	! 1092 2013-02-02 11:24:22Z raasch
	71	! unused variables removed
	72	!
[1037]	73	! 1036 2012-10-22 13:43:42Z raasch
	74	! code put under GPL (PALM 3.9)
	75	!
[1004]	76	! 1003 2012-09-14 14:35:53Z raasch
	77	! adjustment of array tend for cases with unequal subdomain sizes removed
	78	!
[1]	79	! Revision 1.1 1997/07/24 11:24:44 raasch
	80	! Initial revision
	81	!
	82	!
	83	! Description:
	84	! ------------
	85	! Compute the divergence of the provisional velocity field. Solve the Poisson
	86	! equation for the perturbation pressure. Compute the final velocities using
	87	! this perturbation pressure. Compute the remaining divergence.
	88	!------------------------------------------------------------------------------!
	89
[1320]	90	USE arrays_3d, &
	91	ONLY: d, ddzu, ddzu_pres, ddzw, dzw, p, p_loc, tend, u, v, w
	92
	93	USE control_parameters, &
	94	ONLY: bc_lr_cyc, bc_ns_cyc, conserve_volume_flow, dt_3d, &
	95	gathered_size, ibc_p_b, ibc_p_t, intermediate_timestep_count, &
	96	mg_switch_to_pe0_level, on_device, outflow_l, outflow_n, &
	97	outflow_r, outflow_s, psolver, simulated_time, subdomain_size, &
	98	topography, volume_flow, volume_flow_area, volume_flow_initial
	99
	100	USE cpulog, &
	101	ONLY: cpu_log, log_point, log_point_s
	102
	103	USE grid_variables, &
	104	ONLY: ddx, ddy
	105
	106	USE indices, &
	107	ONLY: nbgp, ngp_2dh_outer, nx, nxl, nxlg, nxl_mg, nxr, nxrg, nxr_mg, &
	108	ny, nys, nysg, nys_mg, nyn, nyng, nyn_mg, nzb, nzb_s_inner, &
	109	nzb_u_inner, nzb_v_inner, nzb_w_inner, nzb_2d, nzt, nzt_mg, &
	110	rflags_s_inner
	111
	112	USE kinds
	113
[1]	114	USE pegrid
	115
[1320]	116	USE poisfft_mod, &
	117	ONLY: poisfft
	118
[1575]	119	USE poismg_mod
	120
[1320]	121	USE statistics, &
	122	ONLY: statistic_regions, sums_divnew_l, sums_divold_l, weight_pres, &
	123	weight_substep
	124
[1]	125	IMPLICIT NONE
	126
[1320]	127	INTEGER(iwp) :: i !:
	128	INTEGER(iwp) :: j !:
	129	INTEGER(iwp) :: k !:
[1]	130
[1320]	131	REAL(wp) :: ddt_3d !:
	132	REAL(wp) :: localsum !:
	133	REAL(wp) :: threadsum !:
	134	REAL(wp) :: d_weight_pres !:
[1]	135
[1320]	136	REAL(wp), DIMENSION(1:2) :: volume_flow_l !:
	137	REAL(wp), DIMENSION(1:2) :: volume_flow_offset !:
	138	REAL(wp), DIMENSION(1:nzt) :: w_l !:
	139	REAL(wp), DIMENSION(1:nzt) :: w_l_l !:
[1]	140
	141
	142	CALL cpu_log( log_point(8), 'pres', 'start' )
	143
[85]	144
[1342]	145	ddt_3d = 1.0_wp / dt_3d
	146	d_weight_pres = 1.0_wp / weight_pres(intermediate_timestep_count)
[85]	147
[1]	148	!
[707]	149	!-- Multigrid method expects array d to have one ghost layer.
	150	!--
[1575]	151	IF ( psolver(1:9) == 'multigrid' ) THEN
[667]	152
[1]	153	DEALLOCATE( d )
[667]	154	ALLOCATE( d(nzb:nzt+1,nys-1:nyn+1,nxl-1:nxr+1) )
[707]	155
	156	!
	157	!-- Since p is later used to hold the weighted average of the substeps, it
	158	!-- cannot be used in the iterative solver. Therefore, its initial value is
	159	!-- stored on p_loc, which is then iteratively advanced in every substep.
	160	IF ( intermediate_timestep_count == 1 ) THEN
	161	DO i = nxl-1, nxr+1
	162	DO j = nys-1, nyn+1
	163	DO k = nzb, nzt+1
	164	p_loc(k,j,i) = p(k,j,i)
	165	ENDDO
	166	ENDDO
	167	ENDDO
[667]	168	ENDIF
	169
[707]	170	ELSEIF ( psolver == 'sor' .AND. intermediate_timestep_count == 1 ) THEN
	171
	172	!
	173	!-- Since p is later used to hold the weighted average of the substeps, it
	174	!-- cannot be used in the iterative solver. Therefore, its initial value is
	175	!-- stored on p_loc, which is then iteratively advanced in every substep.
	176	p_loc = p
	177
[1]	178	ENDIF
	179
	180	!
[75]	181	!-- Conserve the volume flow at the outflow in case of non-cyclic lateral
	182	!-- boundary conditions
[106]	183	!-- WARNING: so far, this conservation does not work at the left/south
	184	!-- boundary if the topography at the inflow differs from that at the
	185	!-- outflow! For this case, volume_flow_area needs adjustment!
	186	!
	187	!-- Left/right
[709]	188	IF ( conserve_volume_flow .AND. ( outflow_l .OR. outflow_r ) ) THEN
[680]	189
[1342]	190	volume_flow(1) = 0.0_wp
	191	volume_flow_l(1) = 0.0_wp
[106]	192
	193	IF ( outflow_l ) THEN
	194	i = 0
	195	ELSEIF ( outflow_r ) THEN
	196	i = nx+1
	197	ENDIF
	198
	199	DO j = nys, nyn
	200	!
	201	!-- Sum up the volume flow through the south/north boundary
[709]	202	DO k = nzb_2d(j,i)+1, nzt
[667]	203	volume_flow_l(1) = volume_flow_l(1) + u(k,j,i) * dzw(k)
[106]	204	ENDDO
	205	ENDDO
	206
	207	#if defined( __parallel )
[680]	208	IF ( collective_wait ) CALL MPI_BARRIER( comm1dy, ierr )
[106]	209	CALL MPI_ALLREDUCE( volume_flow_l(1), volume_flow(1), 1, MPI_REAL, &
	210	MPI_SUM, comm1dy, ierr )
	211	#else
	212	volume_flow = volume_flow_l
	213	#endif
[709]	214	volume_flow_offset(1) = ( volume_flow_initial(1) - volume_flow(1) ) &
[106]	215	/ volume_flow_area(1)
	216
[667]	217	DO j = nysg, nyng
[709]	218	DO k = nzb_2d(j,i)+1, nzt
[106]	219	u(k,j,i) = u(k,j,i) + volume_flow_offset(1)
	220	ENDDO
	221	ENDDO
	222
	223	ENDIF
	224
	225	!
	226	!-- South/north
[709]	227	IF ( conserve_volume_flow .AND. ( outflow_n .OR. outflow_s ) ) THEN
[106]	228
[1342]	229	volume_flow(2) = 0.0_wp
	230	volume_flow_l(2) = 0.0_wp
[75]	231
[106]	232	IF ( outflow_s ) THEN
	233	j = 0
	234	ELSEIF ( outflow_n ) THEN
[75]	235	j = ny+1
[106]	236	ENDIF
	237
	238	DO i = nxl, nxr
[75]	239	!
[106]	240	!-- Sum up the volume flow through the south/north boundary
[709]	241	DO k = nzb_2d(j,i)+1, nzt
[667]	242	volume_flow_l(2) = volume_flow_l(2) + v(k,j,i) * dzw(k)
[75]	243	ENDDO
[106]	244	ENDDO
	245
[75]	246	#if defined( __parallel )
[680]	247	IF ( collective_wait ) CALL MPI_BARRIER( comm1dx, ierr )
[75]	248	CALL MPI_ALLREDUCE( volume_flow_l(2), volume_flow(2), 1, MPI_REAL, &
	249	MPI_SUM, comm1dx, ierr )
	250	#else
	251	volume_flow = volume_flow_l
	252	#endif
	253	volume_flow_offset(2) = ( volume_flow_initial(2) - volume_flow(2) ) &
[106]	254	/ volume_flow_area(2)
[75]	255
[667]	256	DO i = nxlg, nxrg
[709]	257	DO k = nzb_v_inner(j,i)+1, nzt
[106]	258	v(k,j,i) = v(k,j,i) + volume_flow_offset(2)
[75]	259	ENDDO
[106]	260	ENDDO
[75]	261
	262	ENDIF
	263
[76]	264	!
	265	!-- Remove mean vertical velocity
	266	IF ( ibc_p_b == 1 .AND. ibc_p_t == 1 ) THEN
[1342]	267	IF ( simulated_time > 0.0_wp ) THEN ! otherwise nzb_w_inner not yet known
	268	w_l = 0.0_wp; w_l_l = 0.0_wp
[76]	269	DO i = nxl, nxr
	270	DO j = nys, nyn
	271	DO k = nzb_w_inner(j,i)+1, nzt
	272	w_l_l(k) = w_l_l(k) + w(k,j,i)
	273	ENDDO
	274	ENDDO
	275	ENDDO
	276	#if defined( __parallel )
[622]	277	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
[709]	278	CALL MPI_ALLREDUCE( w_l_l(1), w_l(1), nzt, MPI_REAL, MPI_SUM, &
	279	comm2d, ierr )
[76]	280	#else
	281	w_l = w_l_l
	282	#endif
	283	DO k = 1, nzt
	284	w_l(k) = w_l(k) / ngp_2dh_outer(k,0)
	285	ENDDO
[667]	286	DO i = nxlg, nxrg
	287	DO j = nysg, nyng
[76]	288	DO k = nzb_w_inner(j,i)+1, nzt
	289	w(k,j,i) = w(k,j,i) - w_l(k)
	290	ENDDO
	291	ENDDO
	292	ENDDO
	293	ENDIF
	294	ENDIF
[75]	295
	296	!
[1]	297	!-- Compute the divergence of the provisional velocity field.
	298	CALL cpu_log( log_point_s(1), 'divergence', 'start' )
	299
[1575]	300	IF ( psolver(1:9) == 'multigrid' ) THEN
[1]	301	!$OMP PARALLEL DO SCHEDULE( STATIC )
	302	DO i = nxl-1, nxr+1
	303	DO j = nys-1, nyn+1
	304	DO k = nzb, nzt+1
[1342]	305	d(k,j,i) = 0.0_wp
[1]	306	ENDDO
	307	ENDDO
	308	ENDDO
	309	ELSE
	310	!$OMP PARALLEL DO SCHEDULE( STATIC )
[1111]	311	!$acc kernels present( d )
[1003]	312	DO i = nxl, nxr
	313	DO j = nys, nyn
	314	DO k = nzb+1, nzt
[1342]	315	d(k,j,i) = 0.0_wp
[1]	316	ENDDO
	317	ENDDO
	318	ENDDO
[1111]	319	!$acc end kernels
[1]	320	ENDIF
	321
[1342]	322	localsum = 0.0_wp
	323	threadsum = 0.0_wp
[1]	324
	325	#if defined( __ibm )
	326	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
	327	!$OMP DO SCHEDULE( STATIC )
	328	DO i = nxl, nxr
	329	DO j = nys, nyn
	330	DO k = nzb_s_inner(j,i)+1, nzt
[85]	331	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
	332	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
[673]	333	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k) ) * ddt_3d &
	334	* d_weight_pres
[1]	335	ENDDO
	336	!
	337	!-- Compute possible PE-sum of divergences for flow_statistics
	338	DO k = nzb_s_inner(j,i)+1, nzt
	339	threadsum = threadsum + ABS( d(k,j,i) )
	340	ENDDO
	341
	342	ENDDO
	343	ENDDO
	344
[707]	345	localsum = localsum + threadsum * dt_3d * &
	346	weight_pres(intermediate_timestep_count)
[693]	347
[1]	348	!$OMP END PARALLEL
	349	#else
	350
[1111]	351	!$OMP PARALLEL PRIVATE (i,j,k)
	352	!$OMP DO SCHEDULE( STATIC )
[1221]	353	!$acc kernels present( d, ddzw, rflags_s_inner, u, v, w )
	354	!$acc loop collapse( 3 )
[1111]	355	DO i = nxl, nxr
	356	DO j = nys, nyn
	357	DO k = 1, nzt
[1221]	358	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
	359	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
	360	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k) ) * ddt_3d &
	361	* d_weight_pres * rflags_s_inner(k,j,i)
[1]	362	ENDDO
	363	ENDDO
[1111]	364	ENDDO
	365	!$acc end kernels
	366	!$OMP END PARALLEL
[1]	367
	368	!
	369	!-- Compute possible PE-sum of divergences for flow_statistics
	370	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
	371	!$OMP DO SCHEDULE( STATIC )
[1221]	372	!$acc parallel loop collapse(3) present( d ) reduction(+:threadsum)
[1]	373	DO i = nxl, nxr
	374	DO j = nys, nyn
	375	DO k = nzb+1, nzt
	376	threadsum = threadsum + ABS( d(k,j,i) )
	377	ENDDO
	378	ENDDO
	379	ENDDO
[1257]	380	!$acc end parallel loop
[707]	381	localsum = localsum + threadsum * dt_3d * &
	382	weight_pres(intermediate_timestep_count)
[1]	383	!$OMP END PARALLEL
	384	#endif
	385
	386	!
	387	!-- For completeness, set the divergence sum of all statistic regions to those
	388	!-- of the total domain
	389	sums_divold_l(0:statistic_regions) = localsum
	390
	391	CALL cpu_log( log_point_s(1), 'divergence', 'stop' )
	392
	393	!
	394	!-- Compute the pressure perturbation solving the Poisson equation
[1575]	395	IF ( psolver == 'poisfft' ) THEN
[1]	396
	397	!
	398	!-- Solve Poisson equation via FFT and solution of tridiagonal matrices
[1575]	399	CALL poisfft( d )
[1212]	400
[1]	401	!
	402	!-- Store computed perturbation pressure and set boundary condition in
	403	!-- z-direction
	404	!$OMP PARALLEL DO
[1113]	405	!$acc kernels present( d, tend )
[1]	406	DO i = nxl, nxr
	407	DO j = nys, nyn
	408	DO k = nzb+1, nzt
	409	tend(k,j,i) = d(k,j,i)
	410	ENDDO
	411	ENDDO
	412	ENDDO
[1113]	413	!$acc end kernels
[1]	414
	415	!
	416	!-- Bottom boundary:
	417	!-- This condition is only required for internal output. The pressure
	418	!-- gradient (dp(nzb+1)-dp(nzb))/dz is not used anywhere else.
	419	IF ( ibc_p_b == 1 ) THEN
	420	!
	421	!-- Neumann (dp/dz = 0)
	422	!$OMP PARALLEL DO
[1113]	423	!$acc kernels present( nzb_s_inner, tend )
[667]	424	DO i = nxlg, nxrg
	425	DO j = nysg, nyng
[1]	426	tend(nzb_s_inner(j,i),j,i) = tend(nzb_s_inner(j,i)+1,j,i)
	427	ENDDO
	428	ENDDO
[1113]	429	!$acc end kernels
[1]	430
	431	ELSE
	432	!
	433	!-- Dirichlet
	434	!$OMP PARALLEL DO
[1113]	435	!$acc kernels present( tend )
[667]	436	DO i = nxlg, nxrg
	437	DO j = nysg, nyng
[1342]	438	tend(nzb_s_inner(j,i),j,i) = 0.0_wp
[1]	439	ENDDO
	440	ENDDO
[1113]	441	!$acc end kernels
[1]	442
	443	ENDIF
	444
	445	!
	446	!-- Top boundary
	447	IF ( ibc_p_t == 1 ) THEN
	448	!
	449	!-- Neumann
	450	!$OMP PARALLEL DO
[1113]	451	!$acc kernels present( tend )
[667]	452	DO i = nxlg, nxrg
	453	DO j = nysg, nyng
[1]	454	tend(nzt+1,j,i) = tend(nzt,j,i)
	455	ENDDO
	456	ENDDO
[1113]	457	!$acc end kernels
[1]	458
	459	ELSE
	460	!
	461	!-- Dirichlet
	462	!$OMP PARALLEL DO
[1113]	463	!$acc kernels present( tend )
[667]	464	DO i = nxlg, nxrg
	465	DO j = nysg, nyng
[1342]	466	tend(nzt+1,j,i) = 0.0_wp
[1]	467	ENDDO
	468	ENDDO
[1113]	469	!$acc end kernels
[1]	470
	471	ENDIF
	472
	473	!
	474	!-- Exchange boundaries for p
[1113]	475	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
	476	on_device = .TRUE. ! to be removed after complete porting
	477	ELSE ! of ghost point exchange
	478	!$acc update host( tend )
	479	ENDIF
[667]	480	CALL exchange_horiz( tend, nbgp )
[1113]	481	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
	482	on_device = .FALSE. ! to be removed after complete porting
	483	ELSE ! of ghost point exchange
	484	!$acc update device( tend )
	485	ENDIF
[1]	486
	487	ELSEIF ( psolver == 'sor' ) THEN
	488
	489	!
	490	!-- Solve Poisson equation for perturbation pressure using SOR-Red/Black
	491	!-- scheme
[707]	492	CALL sor( d, ddzu_pres, ddzw, p_loc )
	493	tend = p_loc
[1]	494
[1575]	495	ELSEIF ( psolver(1:9) == 'multigrid' ) THEN
[1]	496
	497	!
	498	!-- Solve Poisson equation for perturbation pressure using Multigrid scheme,
[667]	499	!-- array tend is used to store the residuals, logical exchange_mg is used
	500	!-- to discern data exchange in multigrid ( 1 ghostpoint ) and normal grid
	501	!-- ( nbgp ghost points ).
[778]	502
	503	!-- If the number of grid points of the gathered grid, which is collected
	504	!-- on PE0, is larger than the number of grid points of an PE, than array
	505	!-- tend will be enlarged.
	506	IF ( gathered_size > subdomain_size ) THEN
	507	DEALLOCATE( tend )
	508	ALLOCATE( tend(nzb:nzt_mg(mg_switch_to_pe0_level)+1,nys_mg( &
	509	mg_switch_to_pe0_level)-1:nyn_mg(mg_switch_to_pe0_level)+1,&
	510	nxl_mg(mg_switch_to_pe0_level)-1:nxr_mg( &
	511	mg_switch_to_pe0_level)+1) )
	512	ENDIF
	513
[1575]	514	IF ( psolver == 'multigrid' ) THEN
	515	CALL poismg( tend )
	516	ELSE
	517	CALL poismg_fast( tend )
	518	ENDIF
[707]	519
[778]	520	IF ( gathered_size > subdomain_size ) THEN
	521	DEALLOCATE( tend )
	522	ALLOCATE( tend(nzb:nzt+1,nysg:nyng,nxlg:nxrg) )
	523	ENDIF
	524
[1]	525	!
	526	!-- Restore perturbation pressure on tend because this array is used
	527	!-- further below to correct the velocity fields
[707]	528	DO i = nxl-1, nxr+1
	529	DO j = nys-1, nyn+1
	530	DO k = nzb, nzt+1
	531	tend(k,j,i) = p_loc(k,j,i)
	532	ENDDO
	533	ENDDO
	534	ENDDO
[667]	535
[1]	536	ENDIF
	537
	538	!
[707]	539	!-- Store perturbation pressure on array p, used for pressure data output.
	540	!-- Ghost layers are added in the output routines (except sor-method: see below)
	541	IF ( intermediate_timestep_count == 1 ) THEN
	542	!$OMP PARALLEL PRIVATE (i,j,k)
	543	!$OMP DO
[1113]	544	!$acc kernels present( p, tend, weight_substep )
[1257]	545	!$acc loop independent
[707]	546	DO i = nxl-1, nxr+1
[1257]	547	!$acc loop independent
[707]	548	DO j = nys-1, nyn+1
[1257]	549	!$acc loop independent
[707]	550	DO k = nzb, nzt+1
	551	p(k,j,i) = tend(k,j,i) * &
	552	weight_substep(intermediate_timestep_count)
[673]	553	ENDDO
[1]	554	ENDDO
[707]	555	ENDDO
[1113]	556	!$acc end kernels
[707]	557	!$OMP END PARALLEL
	558
	559	ELSE
	560	!$OMP PARALLEL PRIVATE (i,j,k)
	561	!$OMP DO
[1113]	562	!$acc kernels present( p, tend, weight_substep )
[1257]	563	!$acc loop independent
[707]	564	DO i = nxl-1, nxr+1
[1257]	565	!$acc loop independent
[707]	566	DO j = nys-1, nyn+1
[1257]	567	!$acc loop independent
[707]	568	DO k = nzb, nzt+1
	569	p(k,j,i) = p(k,j,i) + tend(k,j,i) * &
	570	weight_substep(intermediate_timestep_count)
[673]	571	ENDDO
	572	ENDDO
[707]	573	ENDDO
[1113]	574	!$acc end kernels
[707]	575	!$OMP END PARALLEL
	576
	577	ENDIF
[673]	578
[707]	579	!
	580	!-- SOR-method needs ghost layers for the next timestep
	581	IF ( psolver == 'sor' ) CALL exchange_horiz( p, nbgp )
[682]	582
[1]	583	!
	584	!-- Correction of the provisional velocities with the current perturbation
	585	!-- pressure just computed
[709]	586	IF ( conserve_volume_flow .AND. ( bc_lr_cyc .OR. bc_ns_cyc ) ) THEN
[1342]	587	volume_flow_l(1) = 0.0_wp
	588	volume_flow_l(2) = 0.0_wp
[1]	589	ENDIF
[707]	590
[1]	591	!$OMP PARALLEL PRIVATE (i,j,k)
	592	!$OMP DO
[1113]	593	!$acc kernels present( ddzu, nzb_u_inner, nzb_v_inner, nzb_w_inner, tend, u, v, w, weight_pres )
[1257]	594	!$acc loop independent
[673]	595	DO i = nxl, nxr
[1257]	596	!$acc loop independent
[1]	597	DO j = nys, nyn
[1257]	598	!$acc loop independent
[1113]	599	DO k = 1, nzt
	600	IF ( k > nzb_w_inner(j,i) ) THEN
	601	w(k,j,i) = w(k,j,i) - dt_3d * &
	602	( tend(k+1,j,i) - tend(k,j,i) ) * ddzu(k+1) * &
	603	weight_pres(intermediate_timestep_count)
	604	ENDIF
[1]	605	ENDDO
[1257]	606	!$acc loop independent
[1113]	607	DO k = 1, nzt
	608	IF ( k > nzb_u_inner(j,i) ) THEN
	609	u(k,j,i) = u(k,j,i) - dt_3d * &
	610	( tend(k,j,i) - tend(k,j,i-1) ) * ddx * &
	611	weight_pres(intermediate_timestep_count)
	612	ENDIF
[1]	613	ENDDO
[1257]	614	!$acc loop independent
[1113]	615	DO k = 1, nzt
	616	IF ( k > nzb_v_inner(j,i) ) THEN
	617	v(k,j,i) = v(k,j,i) - dt_3d * &
	618	( tend(k,j,i) - tend(k,j-1,i) ) * ddy * &
	619	weight_pres(intermediate_timestep_count)
	620	ENDIF
[673]	621	ENDDO
[1]	622
	623	ENDDO
	624	ENDDO
[1113]	625	!$acc end kernels
[1]	626	!$OMP END PARALLEL
[1113]	627
	628	!
	629	!-- Sum up the volume flow through the right and north boundary
	630	IF ( conserve_volume_flow .AND. bc_lr_cyc .AND. bc_ns_cyc .AND. &
	631	nxr == nx ) THEN
	632
	633	!$OMP PARALLEL PRIVATE (j,k)
	634	!$OMP DO
	635	DO j = nys, nyn
	636	!$OMP CRITICAL
	637	DO k = nzb_2d(j,nx) + 1, nzt
	638	volume_flow_l(1) = volume_flow_l(1) + u(k,j,nx) * dzw(k)
	639	ENDDO
	640	!$OMP END CRITICAL
	641	ENDDO
	642	!$OMP END PARALLEL
	643
	644	ENDIF
	645
	646	IF ( conserve_volume_flow .AND. bc_ns_cyc .AND. bc_lr_cyc .AND. &
	647	nyn == ny ) THEN
	648
	649	!$OMP PARALLEL PRIVATE (i,k)
	650	!$OMP DO
	651	DO i = nxl, nxr
	652	!$OMP CRITICAL
	653	DO k = nzb_2d(ny,i) + 1, nzt
	654	volume_flow_l(2) = volume_flow_l(2) + v(k,ny,i) * dzw(k)
	655	ENDDO
	656	!$OMP END CRITICAL
	657	ENDDO
	658	!$OMP END PARALLEL
	659
	660	ENDIF
[673]	661
[1]	662	!
	663	!-- Conserve the volume flow
[707]	664	IF ( conserve_volume_flow .AND. ( bc_lr_cyc .AND. bc_ns_cyc ) ) THEN
[1]	665
	666	#if defined( __parallel )
[622]	667	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
[1]	668	CALL MPI_ALLREDUCE( volume_flow_l(1), volume_flow(1), 2, MPI_REAL, &
	669	MPI_SUM, comm2d, ierr )
	670	#else
	671	volume_flow = volume_flow_l
	672	#endif
	673
	674	volume_flow_offset = ( volume_flow_initial - volume_flow ) / &
	675	volume_flow_area
	676
	677	!$OMP PARALLEL PRIVATE (i,j,k)
	678	!$OMP DO
	679	DO i = nxl, nxr
	680	DO j = nys, nyn
[667]	681	DO k = nzb_u_inner(j,i) + 1, nzt
	682	u(k,j,i) = u(k,j,i) + volume_flow_offset(1)
[719]	683	ENDDO
	684	DO k = nzb_v_inner(j,i) + 1, nzt
[667]	685	v(k,j,i) = v(k,j,i) + volume_flow_offset(2)
	686	ENDDO
[1]	687	ENDDO
	688	ENDDO
[667]	689
[1]	690	!$OMP END PARALLEL
	691
	692	ENDIF
	693
	694	!
	695	!-- Exchange of boundaries for the velocities
[1113]	696	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
	697	on_device = .TRUE. ! to be removed after complete porting
	698	ELSE ! of ghost point exchange
	699	!$acc update host( u, v, w )
	700	ENDIF
[667]	701	CALL exchange_horiz( u, nbgp )
	702	CALL exchange_horiz( v, nbgp )
	703	CALL exchange_horiz( w, nbgp )
[1113]	704	IF ( numprocs == 1 ) THEN ! workaround for single-core GPU runs
	705	on_device = .FALSE. ! to be removed after complete porting
	706	ELSE ! of ghost point exchange
	707	!$acc update device( u, v, w )
	708	ENDIF
[1]	709
	710	!
	711	!-- Compute the divergence of the corrected velocity field,
	712	!-- a possible PE-sum is computed in flow_statistics
	713	CALL cpu_log( log_point_s(1), 'divergence', 'start' )
[1342]	714	sums_divnew_l = 0.0_wp
[1]	715
	716	!
	717	!-- d must be reset to zero because it can contain nonzero values below the
	718	!-- topography
[1342]	719	IF ( topography /= 'flat' ) d = 0.0_wp
[1]	720
[1342]	721	localsum = 0.0_wp
	722	threadsum = 0.0_wp
[1]	723
	724	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
	725	!$OMP DO SCHEDULE( STATIC )
	726	#if defined( __ibm )
	727	DO i = nxl, nxr
	728	DO j = nys, nyn
	729	DO k = nzb_s_inner(j,i)+1, nzt
	730	d(k,j,i) = ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
	731	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
	732	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k)
	733	ENDDO
	734	DO k = nzb+1, nzt
	735	threadsum = threadsum + ABS( d(k,j,i) )
	736	ENDDO
	737	ENDDO
	738	ENDDO
	739	#else
[1221]	740	!$acc kernels present( d, ddzw, rflags_s_inner, u, v, w )
	741	!$acc loop collapse( 3 )
[1]	742	DO i = nxl, nxr
	743	DO j = nys, nyn
[1113]	744	DO k = 1, nzt
[1221]	745	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * ddx + &
	746	( v(k,j+1,i) - v(k,j,i) ) * ddy + &
	747	( w(k,j,i) - w(k-1,j,i) ) * ddzw(k) &
	748	) * rflags_s_inner(k,j,i)
[1113]	749	ENDDO
	750	ENDDO
	751	ENDDO
	752	!$acc end kernels
	753	!
	754	!-- Compute possible PE-sum of divergences for flow_statistics
[1221]	755	!$acc parallel loop collapse(3) present( d ) reduction(+:threadsum)
[1113]	756	DO i = nxl, nxr
	757	DO j = nys, nyn
[1221]	758	DO k = nzb+1, nzt
[1]	759	threadsum = threadsum + ABS( d(k,j,i) )
	760	ENDDO
	761	ENDDO
	762	ENDDO
[1257]	763	!$acc end parallel loop
[1]	764	#endif
[667]	765
[1]	766	localsum = localsum + threadsum
	767	!$OMP END PARALLEL
	768
	769	!
	770	!-- For completeness, set the divergence sum of all statistic regions to those
	771	!-- of the total domain
	772	sums_divnew_l(0:statistic_regions) = localsum
	773
	774	CALL cpu_log( log_point_s(1), 'divergence', 'stop' )
	775
	776	CALL cpu_log( log_point(8), 'pres', 'stop' )
	777
	778
	779	END SUBROUTINE pres

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

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |