Home

Context Navigation

source: palm/trunk/SOURCE/timestep.f90 @ 4099

Last change on this file since 4099 was 3658, checked in by knoop, 6 years ago
OpenACC: flow_statistics partly ported to GPU
Property svn:keywords set to `Id`
File size: 17.2 KB

Rev	Line
[1682]	1	!> @file timestep.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	!
[258]	20	! Current revisions:
[866]	21	! ------------------
[1485]	22	!
[3049]	23	!
[1485]	24	! Former revisions:
	25	! -----------------
	26	! $Id: timestep.f90 3658 2019-01-07 20:28:54Z suehring $
[3634]	27	! OpenACC port for SPEC
	28	!
	29	! 3311 2018-10-05 12:34:56Z raasch
[3311]	30	! Stokes drift is regarded in timestep calculation
	31	!
	32	! 3274 2018-09-24 15:42:55Z knoop
[3274]	33	! Modularization of all bulk cloud physics code components
	34	!
	35	! 3241 2018-09-12 15:02:00Z raasch
[3241]	36	! unused variables removed
	37	!
	38	! 3120 2018-07-11 18:30:57Z gronemeier
[3120]	39	! Put ABS( km ) in computation of time step according to the diffusion criterion
	40	!
	41	! 3084 2018-06-19 15:30:55Z gronemeier
[3084]	42	! limit increase of dt_3d only in case of RANS mode
	43	!
	44	! 3083 2018-06-19 14:03:12Z gronemeier
[3083]	45	! limit dt_3d to be at maximum 2*old_dt; define old_dt at beginning of routine
	46	! Add km/kh_max
	47	!
	48	! 3049 2018-05-29 13:52:36Z Giersch
[3049]	49	! Error messages revised
	50	!
	51	! 3045 2018-05-28 07:55:41Z Giersch
[3045]	52	! Error message revised
	53	!
	54	! 2718 2018-01-02 08:49:38Z maronga
[2716]	55	! Corrected "Former revisions" section
	56	!
	57	! 2696 2017-12-14 17:12:51Z kanani
	58	! Change in file header (GPL part)
	59	!
	60	! 2365 2017-08-21 14:59:59Z kanani
[2365]	61	! Vertical grid nesting: Sync fine and coarse grid timestep (SadiqHuq)
	62	!
	63	! 2258 2017-06-08 07:55:13Z suehring
[2258]	64	! Bugfix, add pre-preprocessor directives to enable non-parrallel mode
[1485]	65	!
[2258]	66	! 2168 2017-03-06 13:08:38Z suehring
	67	!
[2131]	68	! 2130 2017-01-24 16:25:39Z raasch
	69	! bugfix: in case of nested runs the stop condition in case of too small
	70	! timesteps is communicated to all parent/child processes,
	71	! some formatting done
	72	!
[2119]	73	! 2118 2017-01-17 16:38:49Z raasch
	74	! OpenACC directives and related part of code removed
	75	!
[2001]	76	! 2000 2016-08-20 18:09:15Z knoop
	77	! Forced header and separation lines into 80 columns
	78	!
[1851]	79	! 1849 2016-04-08 11:33:18Z hoffmann
	80	! Adapted for modularization of microphysics
[1852]	81	!
[1683]	82	! 1682 2015-10-07 23:56:08Z knoop
	83	! Code annotations made doxygen readable
	84	!
[1485]	85	! 1484 2014-10-21 10:53:05Z kanani
[1484]	86	! Changes due to new module structure of the plant canopy model:
	87	! calculations and parameters related to the plant canopy model removed
	88	! (the limitation of the canopy drag, i.e. that the canopy drag itself should
	89	! not change the sign of the velocity components, is now assured for in the
	90	! calculation of the canopy tendency terms in subroutine plant_canopy_model)
[1343]	91	!
	92	! 1342 2014-03-26 17:04:47Z kanani
	93	! REAL constants defined as wp-kind
	94	!
[1323]	95	! 1322 2014-03-20 16:38:49Z raasch
	96	! REAL functions provided with KIND-attribute
	97	!
[1321]	98	! 1320 2014-03-20 08:40:49Z raasch
[1320]	99	! ONLY-attribute added to USE-statements,
	100	! kind-parameters added to all INTEGER and REAL declaration statements,
	101	! kinds are defined in new module kinds,
	102	! old module precision_kind is removed,
	103	! revision history before 2012 removed,
	104	! comment fields (!:) to be used for variable explanations added to
	105	! all variable declaration statements
[1321]	106	!
[1258]	107	! 1257 2013-11-08 15:18:40Z raasch
	108	! openacc porting
	109	! bugfix for calculation of advective timestep in case of vertically stretched
	110	! grids
	111	!
[1093]	112	! 1092 2013-02-02 11:24:22Z raasch
	113	! unused variables removed
	114	!
[1054]	115	! 1053 2012-11-13 17:11:03Z hoffmann
	116	! timestep is reduced in two-moment cloud scheme according to the maximum
	117	! terminal velocity of rain drops
	118	!
[1037]	119	! 1036 2012-10-22 13:43:42Z raasch
	120	! code put under GPL (PALM 3.9)
	121	!
[1002]	122	! 1001 2012-09-13 14:08:46Z raasch
	123	! all actions concerning leapfrog scheme removed
	124	!
[979]	125	! 978 2012-08-09 08:28:32Z fricke
	126	! restriction of the outflow damping layer in the diffusion criterion removed
	127	!
[867]	128	! 866 2012-03-28 06:44:41Z raasch
	129	! bugfix for timestep calculation in case of Galilei transformation,
	130	! special treatment in case of mirror velocity boundary condition removed
	131	!
[1]	132	! Revision 1.1 1997/08/11 06:26:19 raasch
	133	! Initial revision
	134	!
	135	!
	136	! Description:
	137	! ------------
[1682]	138	!> Compute the time step under consideration of the FCL and diffusion criterion.
[1]	139	!------------------------------------------------------------------------------!
[1682]	140	SUBROUTINE timestep
	141
[1]	142
[1320]	143	USE arrays_3d, &
[3311]	144	ONLY: dzu, dzw, kh, km, u, u_stokes_zu, v, v_stokes_zu, w
[1320]	145
	146	USE control_parameters, &
	147	ONLY: cfl_factor, coupling_mode, dt_3d, dt_fixed, dt_max, &
[3084]	148	galilei_transformation, old_dt, message_string, rans_mode, &
[1320]	149	stop_dt, terminate_coupled, terminate_coupled_remote, &
	150	timestep_reason, u_gtrans, use_ug_for_galilei_tr, v_gtrans
	151
	152	USE cpulog, &
	153	ONLY: cpu_log, log_point
	154
	155	USE grid_variables, &
	156	ONLY: dx, dx2, dy, dy2
	157
	158	USE indices, &
	159	ONLY: nxl, nxlg, nxr, nxrg, nyn, nyng, nys, nysg, nzb, nzt
	160
[1]	161	USE interfaces
[1320]	162
	163	USE kinds
	164
[3274]	165	USE bulk_cloud_model_mod, &
[1849]	166	ONLY: dt_precipitation
	167
[1]	168	USE pegrid
	169
[2130]	170	USE pmc_interface, &
	171	ONLY: nested_run
	172
[1320]	173	USE statistics, &
	174	ONLY: flow_statistics_called, hom, u_max, u_max_ijk, v_max, v_max_ijk,&
	175	w_max, w_max_ijk
	176
[2365]	177	USE vertical_nesting_mod, &
	178	ONLY: vnested, vnest_timestep_sync
	179
[1]	180	IMPLICIT NONE
	181
[1682]	182	INTEGER(iwp) :: i !<
	183	INTEGER(iwp) :: j !<
	184	INTEGER(iwp) :: k !<
[3083]	185	INTEGER(iwp) :: km_max_ijk(3) = -1 !< index values (i,j,k) of location where km_max occurs
	186	INTEGER(iwp) :: kh_max_ijk(3) = -1 !< index values (i,j,k) of location where kh_max occurs
[1]	187
[2130]	188	LOGICAL :: stop_dt_local !< local switch for controlling the time stepping
	189
[1682]	190	REAL(wp) :: div !<
	191	REAL(wp) :: dt_diff !<
	192	REAL(wp) :: dt_diff_l !<
	193	REAL(wp) :: dt_u !<
	194	REAL(wp) :: dt_u_l !<
	195	REAL(wp) :: dt_v !<
	196	REAL(wp) :: dt_v_l !<
	197	REAL(wp) :: dt_w !<
	198	REAL(wp) :: dt_w_l !<
[3083]	199	REAL(wp) :: km_max !< maximum of Km in entire domain
	200	REAL(wp) :: kh_max !< maximum of Kh in entire domain
[1682]	201	REAL(wp) :: u_gtrans_l !<
	202	REAL(wp) :: v_gtrans_l !<
[1320]	203
[1682]	204	REAL(wp), DIMENSION(2) :: uv_gtrans !<
	205	REAL(wp), DIMENSION(2) :: uv_gtrans_l !<
	206	REAL(wp), DIMENSION(3) :: reduce !<
	207	REAL(wp), DIMENSION(3) :: reduce_l !<
	208	REAL(wp), DIMENSION(nzb+1:nzt) :: dxyz2_min !<
[3634]	209	!$ACC DECLARE CREATE(dxyz2_min)
[1]	210
	211
	212	CALL cpu_log( log_point(12), 'calculate_timestep', 'start' )
[3658]	213
	214	!$ACC UPDATE &
	215	!$ACC HOST(u(nzb:nzt+1,nysg:nyng,nxlg:nxrg)) &
	216	!$ACC HOST(v(nzb:nzt+1,nysg:nyng,nxlg:nxrg)) &
	217	!$ACC HOST(w(nzb:nzt+1,nysg:nyng,nxlg:nxrg)) &
	218	!$ACC HOST(kh(nzb:nzt+1,nysg:nyng,nxlg:nxrg)) &
	219	!$ACC HOST(km(nzb:nzt+1,nysg:nyng,nxlg:nxrg))
	220
[3083]	221	!
	222	!-- Save former time step as reference
	223	old_dt = dt_3d
[1]	224
	225	!
	226	!-- In case of Galilei-transform not using the geostrophic wind as translation
	227	!-- velocity, compute the volume-averaged horizontal velocity components, which
	228	!-- will then be subtracted from the horizontal wind for the time step and
	229	!-- horizontal advection routines.
	230	IF ( galilei_transformation .AND. .NOT. use_ug_for_galilei_tr ) THEN
	231	IF ( flow_statistics_called ) THEN
	232	!
	233	!-- Horizontal averages already existent, just need to average them
	234	!-- vertically.
[1342]	235	u_gtrans = 0.0_wp
	236	v_gtrans = 0.0_wp
[1]	237	DO k = nzb+1, nzt
	238	u_gtrans = u_gtrans + hom(k,1,1,0)
	239	v_gtrans = v_gtrans + hom(k,1,2,0)
	240	ENDDO
[1322]	241	u_gtrans = u_gtrans / REAL( nzt - nzb, KIND=wp )
	242	v_gtrans = v_gtrans / REAL( nzt - nzb, KIND=wp )
[1]	243	ELSE
	244	!
	245	!-- Averaging over the entire model domain.
[1342]	246	u_gtrans_l = 0.0_wp
	247	v_gtrans_l = 0.0_wp
[1]	248	DO i = nxl, nxr
	249	DO j = nys, nyn
	250	DO k = nzb+1, nzt
[1257]	251	u_gtrans_l = u_gtrans_l + u(k,j,i)
	252	v_gtrans_l = v_gtrans_l + v(k,j,i)
[1]	253	ENDDO
	254	ENDDO
	255	ENDDO
[2130]	256	uv_gtrans_l(1) = u_gtrans_l / &
	257	REAL( (nxr-nxl+1)(nyn-nys+1)(nzt-nzb), KIND=wp )
	258	uv_gtrans_l(2) = v_gtrans_l / &
	259	REAL( (nxr-nxl+1)(nyn-nys+1)(nzt-nzb), KIND=wp )
[1]	260	#if defined( __parallel )
[622]	261	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
[2130]	262	CALL MPI_ALLREDUCE( uv_gtrans_l, uv_gtrans, 2, MPI_REAL, MPI_SUM, &
[1]	263	comm2d, ierr )
[1322]	264	u_gtrans = uv_gtrans(1) / REAL( numprocs, KIND=wp )
	265	v_gtrans = uv_gtrans(2) / REAL( numprocs, KIND=wp )
[1]	266	#else
	267	u_gtrans = uv_gtrans_l(1)
	268	v_gtrans = uv_gtrans_l(2)
	269	#endif
	270	ENDIF
	271	ENDIF
	272
[866]	273	!
[1257]	274	!-- Determine the maxima of the velocity components, including their
	275	!-- grid index positions.
[1320]	276	CALL global_min_max( nzb, nzt+1, nysg, nyng, nxlg, nxrg, u, 'abs', 0.0_wp, &
[866]	277	u_max, u_max_ijk )
[1320]	278	CALL global_min_max( nzb, nzt+1, nysg, nyng, nxlg, nxrg, v, 'abs', 0.0_wp, &
[866]	279	v_max, v_max_ijk )
[1320]	280	CALL global_min_max( nzb, nzt+1, nysg, nyng, nxlg, nxrg, w, 'abs', 0.0_wp, &
[866]	281	w_max, w_max_ijk )
	282
[1257]	283	IF ( .NOT. dt_fixed ) THEN
[866]	284	!
[1257]	285	!-- Variable time step:
	286	!-- Calculate the maximum time step according to the CFL-criterion,
	287	!-- individually for each velocity component
[1342]	288	dt_u_l = 999999.9_wp
	289	dt_v_l = 999999.9_wp
	290	dt_w_l = 999999.9_wp
[3634]	291	!$ACC PARALLEL LOOP COLLAPSE(3) PRIVATE(i,j,k) &
	292	!$ACC COPY(dt_u_l, dt_v_l, dt_w_l, u_stokes_zu, v_stokes_zu) &
	293	!$ACC REDUCTION(MIN: dt_u_l, dt_v_l, dt_w_l) &
	294	!$ACC PRESENT(u, v, w, dzu)
[1257]	295	DO i = nxl, nxr
	296	DO j = nys, nyn
	297	DO k = nzb+1, nzt
[2130]	298	dt_u_l = MIN( dt_u_l, ( dx / &
[3311]	299	( ABS( u(k,j,i) - u_gtrans + u_stokes_zu(k) ) &
	300	+ 1.0E-10_wp ) ) )
[2130]	301	dt_v_l = MIN( dt_v_l, ( dy / &
[3311]	302	( ABS( v(k,j,i) - v_gtrans + v_stokes_zu(k) ) &
	303	+ 1.0E-10_wp ) ) )
[2130]	304	dt_w_l = MIN( dt_w_l, ( dzu(k) / &
	305	( ABS( w(k,j,i) ) + 1.0E-10_wp ) ) )
[1257]	306	ENDDO
	307	ENDDO
	308	ENDDO
[1]	309
[1257]	310	#if defined( __parallel )
	311	reduce_l(1) = dt_u_l
	312	reduce_l(2) = dt_v_l
	313	reduce_l(3) = dt_w_l
	314	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
	315	CALL MPI_ALLREDUCE( reduce_l, reduce, 3, MPI_REAL, MPI_MIN, comm2d, ierr )
	316	dt_u = reduce(1)
	317	dt_v = reduce(2)
	318	dt_w = reduce(3)
	319	#else
	320	dt_u = dt_u_l
	321	dt_v = dt_v_l
	322	dt_w = dt_w_l
	323	#endif
	324
[1]	325	!
	326	!-- Compute time step according to the diffusion criterion.
[3120]	327	!-- First calculate minimum grid spacing which only depends on index k.
	328	!-- When using the dynamic subgrid model, negative km are possible.
[1342]	329	dt_diff_l = 999999.0_wp
[1]	330
[3634]	331	!$ACC PARALLEL LOOP PRESENT(dxyz2_min, dzw)
[1]	332	DO k = nzb+1, nzt
[1342]	333	dxyz2_min(k) = MIN( dx2, dy2, dzw(k)dzw(k) ) 0.125_wp
[1]	334	ENDDO
	335
[3241]	336	!$OMP PARALLEL private(i,j,k) reduction(MIN: dt_diff_l)
[2118]	337	!$OMP DO
[3634]	338	!$ACC PARALLEL LOOP COLLAPSE(3) PRIVATE(i,j,k) &
	339	!$ACC COPY(dt_diff_l) REDUCTION(MIN: dt_diff_l) &
	340	!$ACC PRESENT(dxyz2_min, kh, km)
[1]	341	DO i = nxl, nxr
	342	DO j = nys, nyn
	343	DO k = nzb+1, nzt
[2130]	344	dt_diff_l = MIN( dt_diff_l, dxyz2_min(k) / &
[3120]	345	( MAX( kh(k,j,i), ABS( km(k,j,i) ) ) &
	346	+ 1E-20_wp ) )
[1]	347	ENDDO
	348	ENDDO
	349	ENDDO
[2118]	350	!$OMP END PARALLEL
[1]	351	#if defined( __parallel )
[622]	352	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
[2130]	353	CALL MPI_ALLREDUCE( dt_diff_l, dt_diff, 1, MPI_REAL, MPI_MIN, comm2d, &
[1]	354	ierr )
	355	#else
	356	dt_diff = dt_diff_l
	357	#endif
	358
	359	!
[316]	360	!-- The time step is the minimum of the 3-4 components and the diffusion time
[1001]	361	!-- step minus a reduction (cfl_factor) to be on the safe side.
[3084]	362	!-- The time step must not exceed the maximum allowed value.
[2130]	363	dt_3d = cfl_factor * MIN( dt_diff, dt_u, dt_v, dt_w, dt_precipitation )
[3084]	364	dt_3d = MIN( dt_3d, dt_max )
	365	!
	366	!-- In RANS mode, the time step must not increase by more than a factor of 2
	367	IF ( rans_mode ) dt_3d = MIN( dt_3d, dt_max, 2.0_wp * old_dt )
[1]	368
	369	!
	370	!-- Remember the restricting time step criterion for later output.
[1484]	371	IF ( MIN( dt_u, dt_v, dt_w ) < dt_diff ) THEN
[1]	372	timestep_reason = 'A'
	373	ELSE
	374	timestep_reason = 'D'
	375	ENDIF
	376
	377	!
	378	!-- Set flag if the time step becomes too small.
[1342]	379	IF ( dt_3d < ( 0.00001_wp * dt_max ) ) THEN
[1]	380	stop_dt = .TRUE.
[108]	381
[3083]	382	!
	383	!-- Determine the maxima of the diffusion coefficients, including their
	384	!-- grid index positions.
	385	CALL global_min_max( nzb, nzt+1, nysg, nyng, nxlg, nxrg, km, 'abs', &
	386	0.0_wp, km_max, km_max_ijk )
	387	CALL global_min_max( nzb, nzt+1, nysg, nyng, nxlg, nxrg, kh, 'abs', &
	388	0.0_wp, kh_max, kh_max_ijk )
	389
[2130]	390	WRITE( message_string, * ) 'Time step has reached minimum limit.', &
[3046]	391	'&dt = ', dt_3d, ' s Simulation is terminated.', &
	392	'&old_dt = ', old_dt, ' s', &
	393	'&dt_u = ', dt_u, ' s', &
	394	'&dt_v = ', dt_v, ' s', &
	395	'&dt_w = ', dt_w, ' s', &
	396	'&dt_diff = ', dt_diff, ' s', &
[3083]	397	'&u_max = ', u_max, ' m/s k=', u_max_ijk(1), &
[2130]	398	' j=', u_max_ijk(2), ' i=', u_max_ijk(3), &
[3083]	399	'&v_max = ', v_max, ' m/s k=', v_max_ijk(1), &
[2130]	400	' j=', v_max_ijk(2), ' i=', v_max_ijk(3), &
[3083]	401	'&w_max = ', w_max, ' m/s k=', w_max_ijk(1), &
	402	' j=', w_max_ijk(2), ' i=', w_max_ijk(3), &
	403	'&km_max = ', km_max, ' m2/s2 k=', km_max_ijk(1), &
	404	' j=', km_max_ijk(2), ' i=', km_max_ijk(3), &
	405	'&kh_max = ', kh_max, ' m2/s2 k=', kh_max_ijk(1), &
	406	' j=', kh_max_ijk(2), ' i=', kh_max_ijk(3)
[258]	407	CALL message( 'timestep', 'PA0312', 0, 1, 0, 6, 0 )
[108]	408	!
	409	!-- In case of coupled runs inform the remote model of the termination
	410	!-- and its reason, provided the remote model has not already been
	411	!-- informed of another termination reason (terminate_coupled > 0) before.
[222]	412	#if defined( __parallel )
[108]	413	IF ( coupling_mode /= 'uncoupled' .AND. terminate_coupled == 0 ) THEN
	414	terminate_coupled = 2
[2130]	415	IF ( myid == 0 ) THEN
[667]	416	CALL MPI_SENDRECV( &
[2130]	417	terminate_coupled, 1, MPI_INTEGER, target_id, 0, &
	418	terminate_coupled_remote, 1, MPI_INTEGER, target_id, 0, &
[667]	419	comm_inter, status, ierr )
	420	ENDIF
[2130]	421	CALL MPI_BCAST( terminate_coupled_remote, 1, MPI_INTEGER, 0, &
	422	comm2d, ierr)
[108]	423	ENDIF
[222]	424	#endif
[1]	425	ENDIF
	426
	427	!
[2130]	428	!-- In case of nested runs all parent/child processes have to terminate if
	429	!-- one process has set the stop flag, i.e. they need to set the stop flag
	430	!-- too.
	431	IF ( nested_run ) THEN
	432	stop_dt_local = stop_dt
[2258]	433	#if defined( __parallel )
[2130]	434	CALL MPI_ALLREDUCE( stop_dt_local, stop_dt, 1, MPI_LOGICAL, MPI_LOR, &
	435	MPI_COMM_WORLD, ierr )
[2258]	436	#endif
[2130]	437	ENDIF
	438
	439	!
[1001]	440	!-- Ensure a smooth value (two significant digits) of the timestep.
[1342]	441	div = 1000.0_wp
[1001]	442	DO WHILE ( dt_3d < div )
[1342]	443	div = div / 10.0_wp
[1001]	444	ENDDO
[1342]	445	dt_3d = NINT( dt_3d * 100.0_wp / div ) * div / 100.0_wp
[1]	446
[1001]	447	ENDIF
[1]	448
[2365]	449	!
	450	!-- Vertical nesting: coarse and fine grid timestep has to be identical
	451	IF ( vnested ) CALL vnest_timestep_sync
	452
[1]	453	CALL cpu_log( log_point(12), 'calculate_timestep', 'stop' )
	454
	455	END SUBROUTINE timestep

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

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |