Home

Context Navigation

source: palm/trunk/SOURCE/pres.f90 @ 3947

Last change on this file since 3947 was 3849, checked in by knoop, 6 years ago
Bugfix: added proper OpenACC support to disturb_field
Property svn:keywords set to `Id`
File size: 33.5 KB

Rev	Line
[1682]	1	!> @file pres.f90
[2000]	2	!------------------------------------------------------------------------------!
[2696]	3	! This file is part of the PALM model system.
[1036]	4	!
[2000]	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.
[1036]	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	!
[3655]	17	! Copyright 1997-2019 Leibniz Universitaet Hannover
[2000]	18	!------------------------------------------------------------------------------!
[1036]	19	!
[484]	20	! Current revisions:
[1212]	21	! ------------------
[1932]	22	!
[3183]	23	!
[1919]	24	! Former revisions:
	25	! -----------------
[3016]	26	! $Id: pres.f90 3849 2019-04-01 16:35:16Z hellstea $
[3634]	27	! OpenACC port for SPEC
	28	!
	29	! 3347 2018-10-15 14:21:08Z suehring
[3347]	30	! Bugfixes in offline nesting.
	31	! Add comment.
	32	!
	33	! 3341 2018-10-15 10:31:27Z suehring
[3183]	34	! Rename variables for boundary flags and nesting
	35	!
	36	! 3182 2018-07-27 13:36:03Z suehring
[3016]	37	! Dollar sign added before Id
	38	!
	39	! 2696 2017-12-14 17:12:51Z kanani
[3013]	40	! To avoid jumps while plotting w-profiles w level nzt+1 is set to w level nzt
	41	! after velocity modifications through the pressure solver were carried out
	42	!
	43	! 2696 2017-12-14 17:12:51Z kanani
[2716]	44	! Corrected "Former revisions" section
[2696]	45	!
[2716]	46	! 2696 2017-12-14 17:12:51Z kanani
	47	! Change in file header (GPL part)
	48	! poismg_noopt modularized (MS)
	49	!
[2696]	50	! 2298 2017-06-29 09:28:18Z raasch
[2298]	51	! comment changed + some formatting
	52	!
	53	! 2233 2017-05-30 18:08:54Z suehring
[1919]	54	!
[2233]	55	! 2232 2017-05-30 17:47:52Z suehring
	56	! Adjustments to new topography and surface concept
	57	!
[2119]	58	! 2118 2017-01-17 16:38:49Z raasch
	59	! OpenACC directives and related code removed
	60	!
[2074]	61	! 2073 2016-11-30 14:34:05Z raasch
	62	! openmp bugfix for calculation of new divergence
	63	!
[2038]	64	! 2037 2016-10-26 11:15:40Z knoop
	65	! Anelastic approximation implemented
	66	!
[2001]	67	! 2000 2016-08-20 18:09:15Z knoop
	68	! Forced header and separation lines into 80 columns
	69	!
[1933]	70	! 1932 2016-06-10 12:09:21Z suehring
	71	! Initial version of purely vertical nesting introduced.
	72	!
[1932]	73	! 1931 2016-06-10 12:06:59Z suehring
	74	! Rename multigrid into multigrid_noopt and multigrid_fast into multigrid
	75	!
[1930]	76	! 1929 2016-06-09 16:25:25Z suehring
	77	! Bugfix: weight_substep for initial call, replace by local variable
	78	!
[1919]	79	! 1918 2016-05-27 14:35:57Z raasch
[1918]	80	! sum of divergence is also calculated when pres is called before the initial
	81	! first time step,
	82	! stearing is modified by using intermediate_timestep_count = 0 in order to
	83	! determine if pres is called before the first initial timestep,
	84	! bugfix: in case of Neumann conditions for pressure both at bottom and top,
	85	! mean vertical velocity is also removed for the first time step
	86	! bugfix for calculating divergences
[1321]	87	!
[1909]	88	! 1908 2016-05-25 17:22:32Z suehring
	89	! New divergence for output into RUN_CONTROL file is calculated only at last
	90	! Runge-Kutta step
	91	!
[1846]	92	! 1845 2016-04-08 08:29:13Z raasch
	93	! nzb_2d replace by nzb_u\|v_inner
	94	!
[1800]	95	! 1799 2016-04-05 08:35:55Z gronemeier
	96	! Bugfix: excluded third dimension from horizontal volume flow calculation
	97	!
[1763]	98	! 1762 2016-02-25 12:31:13Z hellstea
	99	! Introduction of nested domain feature
	100	!
[1683]	101	! 1682 2015-10-07 23:56:08Z knoop
	102	! Code annotations made doxygen readable
	103	!
[1576]	104	! 1575 2015-03-27 09:56:27Z raasch
	105	! poismg_fast + respective module added, adjustments for psolver-queries
	106	!
[1343]	107	! 1342 2014-03-26 17:04:47Z kanani
	108	! REAL constants defined as wp-kind
	109	!
[1321]	110	! 1320 2014-03-20 08:40:49Z raasch
[1320]	111	! ONLY-attribute added to USE-statements,
	112	! kind-parameters added to all INTEGER and REAL declaration statements,
	113	! kinds are defined in new module kinds,
	114	! old module precision_kind is removed,
	115	! revision history before 2012 removed,
	116	! comment fields (!:) to be used for variable explanations added to
	117	! all variable declaration statements
[708]	118	!
[1319]	119	! 1318 2014-03-17 13:35:16Z raasch
	120	! module interfaces removed
	121	!
[1307]	122	! 1306 2014-03-13 14:30:59Z raasch
	123	! second argument removed from routine poisfft
	124	!
[1258]	125	! 1257 2013-11-08 15:18:40Z raasch
	126	! openacc loop and loop vector clauses removed, independent clauses added,
	127	! end parallel replaced by end parallel loop
	128	!
[1222]	129	! 1221 2013-09-10 08:59:13Z raasch
	130	! openACC porting of reduction operations, loops for calculating d are
	131	! using the rflags_s_inner multiply flag instead of the nzb_s_inner loop index
	132	!
[1213]	133	! 1212 2013-08-15 08:46:27Z raasch
	134	! call of poisfft_hybrid removed
	135	!
[1118]	136	! 1117 2013-03-27 11:15:36Z suehring
	137	! Bugfix in OpenMP parallelization.
	138	!
[1114]	139	! 1113 2013-03-10 02:48:14Z raasch
	140	! GPU-porting of several loops, some loops rearranged
	141	!
[1112]	142	! 1111 2013-03-08 23:54:10Z
	143	! openACC statements added,
	144	! ibc_p_b = 2 removed
	145	!
[1093]	146	! 1092 2013-02-02 11:24:22Z raasch
	147	! unused variables removed
	148	!
[1037]	149	! 1036 2012-10-22 13:43:42Z raasch
	150	! code put under GPL (PALM 3.9)
	151	!
[1004]	152	! 1003 2012-09-14 14:35:53Z raasch
	153	! adjustment of array tend for cases with unequal subdomain sizes removed
	154	!
[1]	155	! Revision 1.1 1997/07/24 11:24:44 raasch
	156	! Initial revision
	157	!
	158	!
	159	! Description:
	160	! ------------
[1682]	161	!> Compute the divergence of the provisional velocity field. Solve the Poisson
	162	!> equation for the perturbation pressure. Compute the final velocities using
	163	!> this perturbation pressure. Compute the remaining divergence.
[1]	164	!------------------------------------------------------------------------------!
[1682]	165	SUBROUTINE pres
	166
[1]	167
[1320]	168	USE arrays_3d, &
[2037]	169	ONLY: d, ddzu, ddzu_pres, ddzw, dzw, p, p_loc, rho_air, rho_air_zw, &
	170	tend, u, v, w
[1320]	171
	172	USE control_parameters, &
[3182]	173	ONLY: bc_lr_cyc, bc_ns_cyc, bc_radiation_l, bc_radiation_n, &
	174	bc_radiation_r, bc_radiation_s, child_domain, &
	175	conserve_volume_flow, coupling_mode, &
[3012]	176	dt_3d, gathered_size, ibc_p_b, ibc_p_t, &
	177	intermediate_timestep_count, intermediate_timestep_count_max, &
[3347]	178	mg_switch_to_pe0_level, nesting_offline, &
	179	psolver, subdomain_size, &
[3182]	180	topography, volume_flow, volume_flow_area, volume_flow_initial
[1320]	181
	182	USE cpulog, &
	183	ONLY: cpu_log, log_point, log_point_s
	184
	185	USE grid_variables, &
	186	ONLY: ddx, ddy
	187
	188	USE indices, &
	189	ONLY: nbgp, ngp_2dh_outer, nx, nxl, nxlg, nxl_mg, nxr, nxrg, nxr_mg, &
[2232]	190	ny, nys, nysg, nys_mg, nyn, nyng, nyn_mg, nzb, nzt, nzt_mg, &
	191	wall_flags_0
[1320]	192
	193	USE kinds
	194
[1]	195	USE pegrid
[1933]	196
	197	USE pmc_interface, &
	198	ONLY: nesting_mode
[1]	199
[1320]	200	USE poisfft_mod, &
	201	ONLY: poisfft
	202
[1575]	203	USE poismg_mod
	204
[2696]	205	USE poismg_noopt_mod
	206
[1320]	207	USE statistics, &
	208	ONLY: statistic_regions, sums_divnew_l, sums_divold_l, weight_pres, &
	209	weight_substep
	210
[2232]	211	USE surface_mod, &
	212	ONLY : bc_h
	213
[1]	214	IMPLICIT NONE
	215
[1682]	216	INTEGER(iwp) :: i !<
	217	INTEGER(iwp) :: j !<
	218	INTEGER(iwp) :: k !<
[2232]	219	INTEGER(iwp) :: m !<
[1]	220
[1682]	221	REAL(wp) :: ddt_3d !<
[1918]	222	REAL(wp) :: d_weight_pres !<
[1682]	223	REAL(wp) :: localsum !<
	224	REAL(wp) :: threadsum !<
[1918]	225	REAL(wp) :: weight_pres_l !<
[1929]	226	REAL(wp) :: weight_substep_l !<
[1]	227
[1762]	228	REAL(wp), DIMENSION(1:3) :: volume_flow_l !<
	229	REAL(wp), DIMENSION(1:3) :: volume_flow_offset !<
[1682]	230	REAL(wp), DIMENSION(1:nzt) :: w_l !<
	231	REAL(wp), DIMENSION(1:nzt) :: w_l_l !<
[1]	232
[3182]	233	LOGICAL :: child_domain_nvn !<
[1]	234
[1933]	235
[1]	236	CALL cpu_log( log_point(8), 'pres', 'start' )
	237
[1918]	238	!
	239	!-- Calculate quantities to be used locally
[1342]	240	ddt_3d = 1.0_wp / dt_3d
[1918]	241	IF ( intermediate_timestep_count == 0 ) THEN
	242	!
	243	!-- If pres is called before initial time step
[1929]	244	weight_pres_l = 1.0_wp
	245	d_weight_pres = 1.0_wp
	246	weight_substep_l = 1.0_wp
[1918]	247	ELSE
[1929]	248	weight_pres_l = weight_pres(intermediate_timestep_count)
	249	d_weight_pres = 1.0_wp / weight_pres(intermediate_timestep_count)
	250	weight_substep_l = weight_substep(intermediate_timestep_count)
[1918]	251	ENDIF
[85]	252
[1]	253	!
[707]	254	!-- Multigrid method expects array d to have one ghost layer.
	255	!--
[1575]	256	IF ( psolver(1:9) == 'multigrid' ) THEN
[667]	257
[1]	258	DEALLOCATE( d )
[667]	259	ALLOCATE( d(nzb:nzt+1,nys-1:nyn+1,nxl-1:nxr+1) )
[707]	260
	261	!
	262	!-- Since p is later used to hold the weighted average of the substeps, it
	263	!-- cannot be used in the iterative solver. Therefore, its initial value is
	264	!-- stored on p_loc, which is then iteratively advanced in every substep.
[1762]	265	IF ( intermediate_timestep_count <= 1 ) THEN
[707]	266	DO i = nxl-1, nxr+1
	267	DO j = nys-1, nyn+1
	268	DO k = nzb, nzt+1
	269	p_loc(k,j,i) = p(k,j,i)
	270	ENDDO
	271	ENDDO
	272	ENDDO
[667]	273	ENDIF
	274
[1918]	275	ELSEIF ( psolver == 'sor' .AND. intermediate_timestep_count <= 1 ) THEN
[707]	276
	277	!
	278	!-- Since p is later used to hold the weighted average of the substeps, it
	279	!-- cannot be used in the iterative solver. Therefore, its initial value is
	280	!-- stored on p_loc, which is then iteratively advanced in every substep.
	281	p_loc = p
	282
[1]	283	ENDIF
	284
	285	!
[75]	286	!-- Conserve the volume flow at the outflow in case of non-cyclic lateral
	287	!-- boundary conditions
[106]	288	!-- WARNING: so far, this conservation does not work at the left/south
	289	!-- boundary if the topography at the inflow differs from that at the
	290	!-- outflow! For this case, volume_flow_area needs adjustment!
	291	!
	292	!-- Left/right
[3182]	293	IF ( conserve_volume_flow .AND. ( bc_radiation_l .OR. &
	294	bc_radiation_r ) ) THEN
[680]	295
[1342]	296	volume_flow(1) = 0.0_wp
	297	volume_flow_l(1) = 0.0_wp
[106]	298
[3182]	299	IF ( bc_radiation_l ) THEN
[106]	300	i = 0
[3182]	301	ELSEIF ( bc_radiation_r ) THEN
[106]	302	i = nx+1
	303	ENDIF
	304
	305	DO j = nys, nyn
	306	!
	307	!-- Sum up the volume flow through the south/north boundary
[2232]	308	DO k = nzb+1, nzt
	309	volume_flow_l(1) = volume_flow_l(1) + u(k,j,i) * dzw(k) &
	310	* MERGE( 1.0_wp, 0.0_wp, &
	311	BTEST( wall_flags_0(k,j,i), 1 ) &
	312	)
[106]	313	ENDDO
	314	ENDDO
	315
	316	#if defined( __parallel )
[680]	317	IF ( collective_wait ) CALL MPI_BARRIER( comm1dy, ierr )
[106]	318	CALL MPI_ALLREDUCE( volume_flow_l(1), volume_flow(1), 1, MPI_REAL, &
	319	MPI_SUM, comm1dy, ierr )
	320	#else
	321	volume_flow = volume_flow_l
	322	#endif
[709]	323	volume_flow_offset(1) = ( volume_flow_initial(1) - volume_flow(1) ) &
[106]	324	/ volume_flow_area(1)
	325
[667]	326	DO j = nysg, nyng
[2232]	327	DO k = nzb+1, nzt
	328	u(k,j,i) = u(k,j,i) + volume_flow_offset(1) &
	329	* MERGE( 1.0_wp, 0.0_wp, &
	330	BTEST( wall_flags_0(k,j,i), 1 ) &
	331	)
[106]	332	ENDDO
	333	ENDDO
	334
	335	ENDIF
	336
	337	!
	338	!-- South/north
[3182]	339	IF ( conserve_volume_flow .AND. ( bc_radiation_n .OR. bc_radiation_s ) ) THEN
[106]	340
[1342]	341	volume_flow(2) = 0.0_wp
	342	volume_flow_l(2) = 0.0_wp
[75]	343
[3182]	344	IF ( bc_radiation_s ) THEN
[106]	345	j = 0
[3182]	346	ELSEIF ( bc_radiation_n ) THEN
[75]	347	j = ny+1
[106]	348	ENDIF
	349
	350	DO i = nxl, nxr
[75]	351	!
[106]	352	!-- Sum up the volume flow through the south/north boundary
[2232]	353	DO k = nzb+1, nzt
	354	volume_flow_l(2) = volume_flow_l(2) + v(k,j,i) * dzw(k) &
	355	* MERGE( 1.0_wp, 0.0_wp, &
	356	BTEST( wall_flags_0(k,j,i), 2 ) &
	357	)
[75]	358	ENDDO
[106]	359	ENDDO
	360
[75]	361	#if defined( __parallel )
[680]	362	IF ( collective_wait ) CALL MPI_BARRIER( comm1dx, ierr )
[75]	363	CALL MPI_ALLREDUCE( volume_flow_l(2), volume_flow(2), 1, MPI_REAL, &
	364	MPI_SUM, comm1dx, ierr )
	365	#else
	366	volume_flow = volume_flow_l
	367	#endif
	368	volume_flow_offset(2) = ( volume_flow_initial(2) - volume_flow(2) ) &
[106]	369	/ volume_flow_area(2)
[75]	370
[667]	371	DO i = nxlg, nxrg
[2232]	372	DO k = nzb+1, nzt
	373	v(k,j,i) = v(k,j,i) + volume_flow_offset(2) &
	374	* MERGE( 1.0_wp, 0.0_wp, &
	375	BTEST( wall_flags_0(k,j,i), 2 ) &
	376	)
[75]	377	ENDDO
[106]	378	ENDDO
[75]	379
	380	ENDIF
	381
[76]	382	!
[1762]	383	!-- Remove mean vertical velocity in case that Neumann conditions are
[1933]	384	!-- used both at bottom and top boundary, and if not a nested domain in a
	385	!-- normal nesting run. In case of vertical nesting, this must be done.
[3182]	386	!-- Therefore an auxiliary logical variable child_domain_nvn is used here, and
[1933]	387	!-- nvn stands for non-vertical nesting.
[1918]	388	!-- This cannot be done before the first initial time step because ngp_2dh_outer
	389	!-- is not yet known then.
[3182]	390	child_domain_nvn = child_domain
	391	IF ( child_domain .AND. nesting_mode == 'vertical' ) THEN
	392	child_domain_nvn = .FALSE.
[1933]	393	ENDIF
	394
[3347]	395	IF ( ibc_p_b == 1 .AND. ibc_p_t == 1 .AND. .NOT. nesting_offline .AND. &
[3182]	396	.NOT. child_domain_nvn .AND. intermediate_timestep_count /= 0 ) &
[1918]	397	THEN
	398	w_l = 0.0_wp; w_l_l = 0.0_wp
	399	DO i = nxl, nxr
	400	DO j = nys, nyn
[2232]	401	DO k = nzb+1, nzt
	402	w_l_l(k) = w_l_l(k) + w(k,j,i) &
	403	* MERGE( 1.0_wp, 0.0_wp, &
	404	BTEST( wall_flags_0(k,j,i), 3 ) &
	405	)
[76]	406	ENDDO
	407	ENDDO
[1918]	408	ENDDO
[76]	409	#if defined( __parallel )
[1918]	410	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
	411	CALL MPI_ALLREDUCE( w_l_l(1), w_l(1), nzt, MPI_REAL, MPI_SUM, &
	412	comm2d, ierr )
[76]	413	#else
[1918]	414	w_l = w_l_l
[76]	415	#endif
[1918]	416	DO k = 1, nzt
	417	w_l(k) = w_l(k) / ngp_2dh_outer(k,0)
	418	ENDDO
	419	DO i = nxlg, nxrg
	420	DO j = nysg, nyng
[2232]	421	DO k = nzb+1, nzt
	422	w(k,j,i) = w(k,j,i) - w_l(k) &
	423	* MERGE( 1.0_wp, 0.0_wp, &
	424	BTEST( wall_flags_0(k,j,i), 3 ) &
	425	)
[76]	426	ENDDO
	427	ENDDO
[1918]	428	ENDDO
[76]	429	ENDIF
[75]	430
	431	!
[1]	432	!-- Compute the divergence of the provisional velocity field.
	433	CALL cpu_log( log_point_s(1), 'divergence', 'start' )
	434
[1575]	435	IF ( psolver(1:9) == 'multigrid' ) THEN
[2232]	436	!$OMP PARALLEL DO SCHEDULE( STATIC ) PRIVATE (i,j,k)
[1]	437	DO i = nxl-1, nxr+1
	438	DO j = nys-1, nyn+1
	439	DO k = nzb, nzt+1
[1342]	440	d(k,j,i) = 0.0_wp
[1]	441	ENDDO
	442	ENDDO
	443	ENDDO
	444	ELSE
[2232]	445	!$OMP PARALLEL DO SCHEDULE( STATIC ) PRIVATE (i,j,k)
[3634]	446	!$ACC PARALLEL LOOP COLLAPSE(3) PRIVATE(i, j, k) &
	447	!$ACC PRESENT(d)
[1003]	448	DO i = nxl, nxr
	449	DO j = nys, nyn
	450	DO k = nzb+1, nzt
[1342]	451	d(k,j,i) = 0.0_wp
[1]	452	ENDDO
	453	ENDDO
	454	ENDDO
	455	ENDIF
	456
[1342]	457	localsum = 0.0_wp
	458	threadsum = 0.0_wp
[1]	459
	460	#if defined( __ibm )
	461	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
	462	!$OMP DO SCHEDULE( STATIC )
	463	DO i = nxl, nxr
	464	DO j = nys, nyn
[2232]	465	DO k = nzb+1, nzt
[2037]	466	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * rho_air(k) * ddx + &
	467	( v(k,j+1,i) - v(k,j,i) ) * rho_air(k) * ddy + &
	468	( w(k,j,i) * rho_air_zw(k) - &
	469	w(k-1,j,i) * rho_air_zw(k-1) ) * ddzw(k) &
[2232]	470	) * ddt_3d * d_weight_pres &
	471	* MERGE( 1.0_wp, 0.0_wp, &
	472	BTEST( wall_flags_0(k,j,i), 0 ) &
	473	)
[1]	474	ENDDO
	475	!
	476	!-- Compute possible PE-sum of divergences for flow_statistics
[2232]	477	DO k = nzb+1, nzt
	478	threadsum = threadsum + ABS( d(k,j,i) ) &
	479	* MERGE( 1.0_wp, 0.0_wp, &
	480	BTEST( wall_flags_0(k,j,i), 0 ) &
	481	)
[1]	482	ENDDO
	483
	484	ENDDO
	485	ENDDO
	486
[1918]	487	IF ( intermediate_timestep_count == intermediate_timestep_count_max .OR. &
	488	intermediate_timestep_count == 0 ) THEN
	489	localsum = localsum + threadsum * dt_3d * weight_pres_l
[1908]	490	ENDIF
[1]	491	!$OMP END PARALLEL
	492	#else
	493
[1111]	494	!$OMP PARALLEL PRIVATE (i,j,k)
	495	!$OMP DO SCHEDULE( STATIC )
[3634]	496	!$ACC PARALLEL LOOP COLLAPSE(3) PRIVATE(i, j, k) &
	497	!$ACC PRESENT(u, v, w, rho_air, rho_air_zw, ddzw, wall_flags_0) &
	498	!$ACC PRESENT(d)
[1111]	499	DO i = nxl, nxr
	500	DO j = nys, nyn
	501	DO k = 1, nzt
[2037]	502	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * rho_air(k) * ddx + &
	503	( v(k,j+1,i) - v(k,j,i) ) * rho_air(k) * ddy + &
	504	( w(k,j,i) * rho_air_zw(k) - &
	505	w(k-1,j,i) * rho_air_zw(k-1) ) * ddzw(k) &
[2232]	506	) * ddt_3d * d_weight_pres &
	507	* MERGE( 1.0_wp, 0.0_wp, &
	508	BTEST( wall_flags_0(k,j,i), 0 ) &
	509	)
[1]	510	ENDDO
	511	ENDDO
[1111]	512	ENDDO
	513	!$OMP END PARALLEL
[1]	514
	515	!
[1908]	516	!-- Compute possible PE-sum of divergences for flow_statistics. Carry out
	517	!-- computation only at last Runge-Kutta substep.
[1918]	518	IF ( intermediate_timestep_count == intermediate_timestep_count_max .OR. &
	519	intermediate_timestep_count == 0 ) THEN
[1908]	520	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
	521	!$OMP DO SCHEDULE( STATIC )
[3634]	522	!$ACC PARALLEL LOOP COLLAPSE(3) PRIVATE(i,j,k) &
	523	!$ACC REDUCTION(+:threadsum) COPY(threadsum) &
	524	!$ACC PRESENT(d)
[1908]	525	DO i = nxl, nxr
	526	DO j = nys, nyn
	527	DO k = nzb+1, nzt
	528	threadsum = threadsum + ABS( d(k,j,i) )
	529	ENDDO
[1]	530	ENDDO
	531	ENDDO
[1918]	532	localsum = localsum + threadsum * dt_3d * weight_pres_l
[1908]	533	!$OMP END PARALLEL
	534	ENDIF
[1]	535	#endif
	536
	537	!
	538	!-- For completeness, set the divergence sum of all statistic regions to those
	539	!-- of the total domain
[1918]	540	IF ( intermediate_timestep_count == intermediate_timestep_count_max .OR. &
	541	intermediate_timestep_count == 0 ) THEN
[1908]	542	sums_divold_l(0:statistic_regions) = localsum
[1918]	543	ENDIF
[1]	544
	545	CALL cpu_log( log_point_s(1), 'divergence', 'stop' )
	546
	547	!
	548	!-- Compute the pressure perturbation solving the Poisson equation
[1575]	549	IF ( psolver == 'poisfft' ) THEN
[1]	550
	551	!
	552	!-- Solve Poisson equation via FFT and solution of tridiagonal matrices
[1575]	553	CALL poisfft( d )
[1212]	554
[1]	555	!
	556	!-- Store computed perturbation pressure and set boundary condition in
	557	!-- z-direction
[2232]	558	!$OMP PARALLEL DO PRIVATE (i,j,k)
[3634]	559	!$ACC PARALLEL LOOP COLLAPSE(3) PRIVATE(i, j, k) &
	560	!$ACC PRESENT(d, tend)
[1]	561	DO i = nxl, nxr
	562	DO j = nys, nyn
	563	DO k = nzb+1, nzt
	564	tend(k,j,i) = d(k,j,i)
	565	ENDDO
	566	ENDDO
	567	ENDDO
	568
	569	!
	570	!-- Bottom boundary:
	571	!-- This condition is only required for internal output. The pressure
	572	!-- gradient (dp(nzb+1)-dp(nzb))/dz is not used anywhere else.
	573	IF ( ibc_p_b == 1 ) THEN
	574	!
[2232]	575	!-- Neumann (dp/dz = 0). Using surfae data type, first for non-natural
	576	!-- surfaces, then for natural and urban surfaces
	577	!-- Upward facing
	578	!$OMP PARALLEL DO PRIVATE( i, j, k )
[3634]	579	!$ACC PARALLEL LOOP PRIVATE(i, j, k) &
	580	!$ACC PRESENT(bc_h, tend)
[2232]	581	DO m = 1, bc_h(0)%ns
[2298]	582	i = bc_h(0)%i(m)
	583	j = bc_h(0)%j(m)
	584	k = bc_h(0)%k(m)
[2232]	585	tend(k-1,j,i) = tend(k,j,i)
[1]	586	ENDDO
[2232]	587	!
	588	!-- Downward facing
	589	!$OMP PARALLEL DO PRIVATE( i, j, k )
[3634]	590	!$ACC PARALLEL LOOP PRIVATE(i, j, k) &
	591	!$ACC PRESENT(bc_h, tend)
[2232]	592	DO m = 1, bc_h(1)%ns
[2298]	593	i = bc_h(1)%i(m)
	594	j = bc_h(1)%j(m)
	595	k = bc_h(1)%k(m)
[2232]	596	tend(k+1,j,i) = tend(k,j,i)
	597	ENDDO
[1]	598
	599	ELSE
	600	!
[2232]	601	!-- Dirichlet. Using surface data type, first for non-natural
	602	!-- surfaces, then for natural and urban surfaces
	603	!-- Upward facing
	604	!$OMP PARALLEL DO PRIVATE( i, j, k )
	605	DO m = 1, bc_h(0)%ns
[2298]	606	i = bc_h(0)%i(m)
	607	j = bc_h(0)%j(m)
	608	k = bc_h(0)%k(m)
[2232]	609	tend(k-1,j,i) = 0.0_wp
[1]	610	ENDDO
[2232]	611	!
	612	!-- Downward facing
	613	!$OMP PARALLEL DO PRIVATE( i, j, k )
	614	DO m = 1, bc_h(1)%ns
[2298]	615	i = bc_h(1)%i(m)
	616	j = bc_h(1)%j(m)
	617	k = bc_h(1)%k(m)
[2232]	618	tend(k+1,j,i) = 0.0_wp
	619	ENDDO
[1]	620
	621	ENDIF
	622
	623	!
	624	!-- Top boundary
	625	IF ( ibc_p_t == 1 ) THEN
	626	!
	627	!-- Neumann
[2232]	628	!$OMP PARALLEL DO PRIVATE (i,j,k)
[667]	629	DO i = nxlg, nxrg
	630	DO j = nysg, nyng
[1]	631	tend(nzt+1,j,i) = tend(nzt,j,i)
	632	ENDDO
	633	ENDDO
	634
	635	ELSE
	636	!
	637	!-- Dirichlet
[2232]	638	!$OMP PARALLEL DO PRIVATE (i,j,k)
[3634]	639	!$ACC PARALLEL LOOP COLLAPSE(2) PRIVATE(i, j) &
	640	!$ACC PRESENT(tend)
[667]	641	DO i = nxlg, nxrg
	642	DO j = nysg, nyng
[1342]	643	tend(nzt+1,j,i) = 0.0_wp
[1]	644	ENDDO
	645	ENDDO
	646
	647	ENDIF
	648
	649	!
	650	!-- Exchange boundaries for p
[667]	651	CALL exchange_horiz( tend, nbgp )
[1]	652
	653	ELSEIF ( psolver == 'sor' ) THEN
	654
	655	!
	656	!-- Solve Poisson equation for perturbation pressure using SOR-Red/Black
	657	!-- scheme
[707]	658	CALL sor( d, ddzu_pres, ddzw, p_loc )
	659	tend = p_loc
[1]	660
[1575]	661	ELSEIF ( psolver(1:9) == 'multigrid' ) THEN
[1]	662
	663	!
	664	!-- Solve Poisson equation for perturbation pressure using Multigrid scheme,
[2298]	665	!-- array tend is used to store the residuals.
[778]	666
	667	!-- If the number of grid points of the gathered grid, which is collected
	668	!-- on PE0, is larger than the number of grid points of an PE, than array
	669	!-- tend will be enlarged.
	670	IF ( gathered_size > subdomain_size ) THEN
	671	DEALLOCATE( tend )
	672	ALLOCATE( tend(nzb:nzt_mg(mg_switch_to_pe0_level)+1,nys_mg( &
	673	mg_switch_to_pe0_level)-1:nyn_mg(mg_switch_to_pe0_level)+1,&
	674	nxl_mg(mg_switch_to_pe0_level)-1:nxr_mg( &
	675	mg_switch_to_pe0_level)+1) )
	676	ENDIF
	677
[1575]	678	IF ( psolver == 'multigrid' ) THEN
	679	CALL poismg( tend )
	680	ELSE
[1931]	681	CALL poismg_noopt( tend )
[1575]	682	ENDIF
[707]	683
[778]	684	IF ( gathered_size > subdomain_size ) THEN
	685	DEALLOCATE( tend )
	686	ALLOCATE( tend(nzb:nzt+1,nysg:nyng,nxlg:nxrg) )
	687	ENDIF
	688
[1]	689	!
	690	!-- Restore perturbation pressure on tend because this array is used
	691	!-- further below to correct the velocity fields
[707]	692	DO i = nxl-1, nxr+1
	693	DO j = nys-1, nyn+1
	694	DO k = nzb, nzt+1
	695	tend(k,j,i) = p_loc(k,j,i)
	696	ENDDO
	697	ENDDO
	698	ENDDO
[667]	699
[1]	700	ENDIF
	701
	702	!
[707]	703	!-- Store perturbation pressure on array p, used for pressure data output.
	704	!-- Ghost layers are added in the output routines (except sor-method: see below)
[1918]	705	IF ( intermediate_timestep_count <= 1 ) THEN
[707]	706	!$OMP PARALLEL PRIVATE (i,j,k)
	707	!$OMP DO
[3634]	708	!$ACC PARALLEL LOOP COLLAPSE(3) PRIVATE(i, j, k) &
	709	!$ACC PRESENT(p, tend)
[707]	710	DO i = nxl-1, nxr+1
	711	DO j = nys-1, nyn+1
	712	DO k = nzb, nzt+1
	713	p(k,j,i) = tend(k,j,i) * &
[1929]	714	weight_substep_l
[673]	715	ENDDO
[1]	716	ENDDO
[707]	717	ENDDO
	718	!$OMP END PARALLEL
	719
[1762]	720	ELSEIF ( intermediate_timestep_count > 1 ) THEN
[707]	721	!$OMP PARALLEL PRIVATE (i,j,k)
	722	!$OMP DO
[3634]	723	!$ACC PARALLEL LOOP COLLAPSE(3) PRIVATE(i, j, k) &
	724	!$ACC PRESENT(p, tend)
[707]	725	DO i = nxl-1, nxr+1
	726	DO j = nys-1, nyn+1
	727	DO k = nzb, nzt+1
	728	p(k,j,i) = p(k,j,i) + tend(k,j,i) * &
[1929]	729	weight_substep_l
[673]	730	ENDDO
	731	ENDDO
[707]	732	ENDDO
	733	!$OMP END PARALLEL
	734
	735	ENDIF
[673]	736
[707]	737	!
	738	!-- SOR-method needs ghost layers for the next timestep
	739	IF ( psolver == 'sor' ) CALL exchange_horiz( p, nbgp )
[682]	740
[1]	741	!
	742	!-- Correction of the provisional velocities with the current perturbation
	743	!-- pressure just computed
[709]	744	IF ( conserve_volume_flow .AND. ( bc_lr_cyc .OR. bc_ns_cyc ) ) THEN
[1342]	745	volume_flow_l(1) = 0.0_wp
	746	volume_flow_l(2) = 0.0_wp
[1]	747	ENDIF
[3347]	748	!
	749	!-- Add pressure gradients to the velocity components. Note, the loops are
	750	!-- running over the entire model domain, even though, in case of non-cyclic
	751	!-- boundaries u- and v-component are not prognostic at i=0 and j=0,
	752	!-- respectiveley. However, in case of Dirichlet boundary conditions for the
	753	!-- velocities, zero-gradient conditions for the pressure are set, so that
	754	!-- no modification is imposed at the boundaries.
[1]	755	!$OMP PARALLEL PRIVATE (i,j,k)
	756	!$OMP DO
[3634]	757	!$ACC PARALLEL LOOP COLLAPSE(2) PRIVATE(i, j, k) &
	758	!$ACC PRESENT(u, v, w, tend, ddzu, wall_flags_0)
[673]	759	DO i = nxl, nxr
[1]	760	DO j = nys, nyn
[2118]	761
[2232]	762	DO k = nzb+1, nzt
	763	w(k,j,i) = w(k,j,i) - dt_3d * &
	764	( tend(k+1,j,i) - tend(k,j,i) ) * ddzu(k+1) &
	765	* weight_pres_l &
	766	* MERGE( 1.0_wp, 0.0_wp, &
	767	BTEST( wall_flags_0(k,j,i), 3 ) &
	768	)
[1]	769	ENDDO
[2118]	770
[2232]	771	DO k = nzb+1, nzt
	772	u(k,j,i) = u(k,j,i) - dt_3d * &
	773	( tend(k,j,i) - tend(k,j,i-1) ) * ddx &
	774	* weight_pres_l &
	775	* MERGE( 1.0_wp, 0.0_wp, &
	776	BTEST( wall_flags_0(k,j,i), 1 ) &
	777	)
[1]	778	ENDDO
[2118]	779
[2232]	780	DO k = nzb+1, nzt
	781	v(k,j,i) = v(k,j,i) - dt_3d * &
	782	( tend(k,j,i) - tend(k,j-1,i) ) * ddy &
	783	* weight_pres_l &
	784	* MERGE( 1.0_wp, 0.0_wp, &
	785	BTEST( wall_flags_0(k,j,i), 2 ) &
	786	)
[673]	787	ENDDO
[1]	788
	789	ENDDO
	790	ENDDO
	791	!$OMP END PARALLEL
[1113]	792
	793	!
[3012]	794	!-- The vertical velocity is not set to zero at nzt + 1 for nested domains
	795	!-- Instead it is set to the values of nzt (see routine vnest_boundary_conds
	796	!-- or pmci_interp_tril_t) BEFORE calling the pressure solver. To avoid jumps
	797	!-- while plotting profiles w at the top has to be set to the values in the
	798	!-- height nzt after above modifications. Hint: w level nzt+1 does not impact
	799	!-- results.
[3182]	800	IF ( child_domain .OR. coupling_mode == 'vnested_fine' ) THEN
[3012]	801	w(nzt+1,:,:) = w(nzt,:,:)
	802	ENDIF
	803
	804	!
[1113]	805	!-- Sum up the volume flow through the right and north boundary
[2232]	806	IF ( conserve_volume_flow .AND. bc_lr_cyc .AND. bc_ns_cyc .AND. &
[1113]	807	nxr == nx ) THEN
	808
	809	!$OMP PARALLEL PRIVATE (j,k)
	810	!$OMP DO
	811	DO j = nys, nyn
	812	!$OMP CRITICAL
[2232]	813	DO k = nzb+1, nzt
	814	volume_flow_l(1) = volume_flow_l(1) + u(k,j,nxr) * dzw(k) &
	815	* MERGE( 1.0_wp, 0.0_wp, &
	816	BTEST( wall_flags_0(k,j,nxr), 1 )&
	817	)
[1113]	818	ENDDO
	819	!$OMP END CRITICAL
	820	ENDDO
	821	!$OMP END PARALLEL
	822
	823	ENDIF
	824
[2232]	825	IF ( conserve_volume_flow .AND. bc_ns_cyc .AND. bc_lr_cyc .AND. &
[1113]	826	nyn == ny ) THEN
	827
	828	!$OMP PARALLEL PRIVATE (i,k)
	829	!$OMP DO
	830	DO i = nxl, nxr
	831	!$OMP CRITICAL
[2232]	832	DO k = nzb+1, nzt
	833	volume_flow_l(2) = volume_flow_l(2) + v(k,nyn,i) * dzw(k) &
	834	* MERGE( 1.0_wp, 0.0_wp, &
	835	BTEST( wall_flags_0(k,nyn,i), 2 )&
	836	)
[1113]	837	ENDDO
	838	!$OMP END CRITICAL
	839	ENDDO
	840	!$OMP END PARALLEL
	841
	842	ENDIF
[673]	843
[1]	844	!
	845	!-- Conserve the volume flow
[707]	846	IF ( conserve_volume_flow .AND. ( bc_lr_cyc .AND. bc_ns_cyc ) ) THEN
[1]	847
	848	#if defined( __parallel )
[622]	849	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
[1]	850	CALL MPI_ALLREDUCE( volume_flow_l(1), volume_flow(1), 2, MPI_REAL, &
	851	MPI_SUM, comm2d, ierr )
	852	#else
	853	volume_flow = volume_flow_l
	854	#endif
	855
[1799]	856	volume_flow_offset(1:2) = ( volume_flow_initial(1:2) - volume_flow(1:2) ) / &
	857	volume_flow_area(1:2)
[1]	858
	859	!$OMP PARALLEL PRIVATE (i,j,k)
	860	!$OMP DO
	861	DO i = nxl, nxr
	862	DO j = nys, nyn
[2232]	863	DO k = nzb+1, nzt
	864	u(k,j,i) = u(k,j,i) + volume_flow_offset(1) &
	865	* MERGE( 1.0_wp, 0.0_wp, &
	866	BTEST( wall_flags_0(k,j,i), 1 ) &
	867	)
[719]	868	ENDDO
[2232]	869	DO k = nzb+1, nzt
	870	v(k,j,i) = v(k,j,i) + volume_flow_offset(2) &
	871	* MERGE( 1.0_wp, 0.0_wp, &
	872	BTEST( wall_flags_0(k,j,i), 2 ) &
	873	)
[667]	874	ENDDO
[1]	875	ENDDO
	876	ENDDO
[667]	877
[1]	878	!$OMP END PARALLEL
	879
	880	ENDIF
	881
	882	!
	883	!-- Exchange of boundaries for the velocities
[667]	884	CALL exchange_horiz( u, nbgp )
	885	CALL exchange_horiz( v, nbgp )
	886	CALL exchange_horiz( w, nbgp )
[1]	887
	888	!
	889	!-- Compute the divergence of the corrected velocity field,
[1908]	890	!-- A possible PE-sum is computed in flow_statistics. Carry out computation
	891	!-- only at last Runge-Kutta step.
[1918]	892	IF ( intermediate_timestep_count == intermediate_timestep_count_max .OR. &
	893	intermediate_timestep_count == 0 ) THEN
[1908]	894	CALL cpu_log( log_point_s(1), 'divergence', 'start' )
	895	sums_divnew_l = 0.0_wp
[1]	896
	897	!
[1908]	898	!-- d must be reset to zero because it can contain nonzero values below the
	899	!-- topography
	900	IF ( topography /= 'flat' ) d = 0.0_wp
[1]	901
[1908]	902	localsum = 0.0_wp
	903	threadsum = 0.0_wp
[1]	904
[1908]	905	!$OMP PARALLEL PRIVATE (i,j,k) FIRSTPRIVATE(threadsum) REDUCTION(+:localsum)
[2073]	906	#if defined( __ibm )
[1908]	907	!$OMP DO SCHEDULE( STATIC )
	908	DO i = nxl, nxr
	909	DO j = nys, nyn
[2232]	910	DO k = nzb+1, nzt
	911	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * rho_air(k) * ddx + &
	912	( v(k,j+1,i) - v(k,j,i) ) * rho_air(k) * ddy + &
	913	( w(k,j,i) * rho_air_zw(k) - &
	914	w(k-1,j,i) * rho_air_zw(k-1) ) * ddzw(k) &
	915	) * MERGE( 1.0_wp, 0.0_wp, &
	916	BTEST( wall_flags_0(k,j,i), 0 ) &
	917	)
[1908]	918	ENDDO
	919	DO k = nzb+1, nzt
	920	threadsum = threadsum + ABS( d(k,j,i) )
	921	ENDDO
[1]	922	ENDDO
	923	ENDDO
	924	#else
[2073]	925	!$OMP DO SCHEDULE( STATIC )
[3634]	926	!$ACC PARALLEL LOOP COLLAPSE(3) PRIVATE(i, j, k) &
	927	!$ACC PRESENT(u, v, w, rho_air, rho_air_zw, ddzw, wall_flags_0) &
	928	!$ACC PRESENT(d)
[1908]	929	DO i = nxl, nxr
	930	DO j = nys, nyn
[2232]	931	DO k = nzb+1, nzt
[2037]	932	d(k,j,i) = ( ( u(k,j,i+1) - u(k,j,i) ) * rho_air(k) * ddx + &
	933	( v(k,j+1,i) - v(k,j,i) ) * rho_air(k) * ddy + &
	934	( w(k,j,i) * rho_air_zw(k) - &
	935	w(k-1,j,i) * rho_air_zw(k-1) ) * ddzw(k) &
[2232]	936	) * MERGE( 1.0_wp, 0.0_wp, &
	937	BTEST( wall_flags_0(k,j,i), 0 ) &
	938	)
[1908]	939	ENDDO
[1113]	940	ENDDO
	941	ENDDO
	942	!
[1908]	943	!-- Compute possible PE-sum of divergences for flow_statistics
[2073]	944	!$OMP DO SCHEDULE( STATIC )
[3634]	945	!$ACC PARALLEL LOOP COLLAPSE(3) PRIVATE(i, j, k) &
	946	!$ACC REDUCTION(+:threadsum) COPY(threadsum) &
	947	!$ACC PRESENT(d)
[1908]	948	DO i = nxl, nxr
	949	DO j = nys, nyn
	950	DO k = nzb+1, nzt
	951	threadsum = threadsum + ABS( d(k,j,i) )
	952	ENDDO
[1]	953	ENDDO
	954	ENDDO
	955	#endif
[667]	956
[1908]	957	localsum = localsum + threadsum
	958	!$OMP END PARALLEL
[1]	959
	960	!
[1908]	961	!-- For completeness, set the divergence sum of all statistic regions to those
	962	!-- of the total domain
	963	sums_divnew_l(0:statistic_regions) = localsum
[1]	964
[1908]	965	CALL cpu_log( log_point_s(1), 'divergence', 'stop' )
[1]	966
[1908]	967	ENDIF
	968
[1]	969	CALL cpu_log( log_point(8), 'pres', 'stop' )
	970
	971	END SUBROUTINE pres

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

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |