Home

Context Navigation

init_pegrid.f90 @ 4894

Last change on this file since 4894 was 4893, checked in by raasch, 4 years ago
revised output of surface data via MPI-IO for better performance
Property svn:keywords set to `Id`
File size: 53.5 KB

Line
1	!> @file init_pegrid.f90
2	!--------------------------------------------------------------------------------------------------!
3	! This file is part of the PALM model system.
4	!
5	! PALM is free software: you can redistribute it and/or modify it under the terms of the GNU General
6	! Public License as published by the Free Software Foundation, either version 3 of the License, or
7	! (at your option) any later version.
8	!
9	! PALM is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the
10	! implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
11	! Public License for more details.
12	!
13	! You should have received a copy of the GNU General Public License along with PALM. If not, see
14	! <http://www.gnu.org/licenses/>.
15	!
16	! Current revisions:
17	! ------------------
18	!
19	!
20	! Former revisions:
21	! -----------------
22	! $Id: init_pegrid.f90 4893 2021-03-02 16:39:14Z raasch $
23	! replaced use_syn_turb_gen by syn_turb_gen
24	!
25	! 4648 2020-08-25 07:52:08Z raasch
26	! file re-formatted to follow the PALM coding standard
27	!
28	! 4564 2020-06-12 14:03:36Z raasch
29	! Vertical nesting method of Huq et al. (2019) removed
30	!
31	! 4461 2020-03-12 16:51:59Z raasch
32	! communicator configurations for four virtual pe grids defined
33	!
34	! 4444 2020-03-05 15:59:50Z raasch
35	! bugfix: cpp-directives for serial mode added
36	!
37	! 4360 2020-01-07 11:25:50Z suehring
38	! changed message PA0467
39	!
40	! 4264 2019-10-15 16:00:23Z scharf
41	! corrected error message string
42	!
43	! 4241 2019-09-27 06:32:47Z raasch
44	! Check added to ensure that subdomain grid has at least the size as given by the number of ghost
45	! points
46	!
47	! 4182 2019-08-22 15:20:23Z scharf
48	! Corrected "Former revisions" section
49	!
50	! 4045 2019-06-21 10:58:47Z raasch
51	! bugfix: kind attribute added to nint function to allow for large integers which may appear in case
52	! of default recycling width and small grid spacings
53	!
54	! 3999 2019-05-23 16:09:37Z suehring
55	! Spend 3 ghost points also in case of pw-scheme when nesting is applied
56	!
57	! 3897 2019-04-15 11:51:14Z suehring
58	! Minor revision of multigrid check; give warning instead of an abort.
59	!
60	! 3890 2019-04-12 15:59:20Z suehring
61	! Check if grid coarsening is possible on subdomain, in order to avoid that multigrid approach
62	! effectively reduces to a Gauss-Seidel scheme.
63	!
64	! 3885 2019-04-11 11:29:34Z kanani
65	! Changes related to global restructuring of location messages and introduction of additional debug
66	! messages
67	!
68	! 3884 2019-04-10 13:31:55Z Giersch
69	! id_recycling is only calculated in case of tubulent inflow
70	!
71	! 3761 2019-02-25 15:31:42Z raasch
72	! unused variable removed
73	!
74	! 3655 2019-01-07 16:51:22Z knoop
75	! variables documented
76	!
77	! Revision 1.1 1997/07/24 11:15:09 raasch
78	! Initial revision
79	!
80	!
81	! Description:
82	! ------------
83	!> Determination of the virtual processor topology (if not prescribed by the user) and computation
84	!> of the grid point number and array bounds of the local domains.
85	!> @todo: remove MPI-data types for 2D exchange on coarse multigrid level (not used any more)
86	!--------------------------------------------------------------------------------------------------!
87	SUBROUTINE init_pegrid
88
89
90	USE control_parameters, &
91	ONLY: bc_dirichlet_l, bc_dirichlet_n, bc_dirichlet_r, bc_dirichlet_s, bc_lr, bc_ns, &
92	bc_radiation_l, bc_radiation_n, bc_radiation_r, bc_radiation_s, grid_level, &
93	grid_level_count, maximum_grid_level, message_string, mg_switch_to_pe0_level, psolver
94
95
96	#if defined( __parallel )
97	USE control_parameters, &
98	ONLY: coupling_mode, coupling_topology, gathered_size, momentum_advec, &
99	outflow_source_plane, recycling_width, scalar_advec, subdomain_size, &
100	syn_turb_gen, turbulent_inflow, turbulent_outflow, y_shift
101
102	USE grid_variables, &
103	ONLY: dx
104	#endif
105
106	USE indices, &
107	ONLY: nnx, nny, nnz, nx, nxl, nxl_mg, nxlu, nxr, nxr_mg, ny, nyn, nyn_mg, nys, nys_mg, &
108	nysv, nz, nzb, nzt, nzt_mg, wall_flags_1, wall_flags_2, wall_flags_3, wall_flags_4, &
109	wall_flags_5, wall_flags_6, wall_flags_7, wall_flags_8, wall_flags_9, wall_flags_10
110
111	#if defined( __parallel )
112	USE indices, &
113	ONLY: mg_loc_ind, nbgp, nx_a, nx_o, ny_a, ny_o
114	#endif
115
116	USE kinds
117
118	USE pegrid
119
120	#if defined( __parallel )
121	USE pmc_interface, &
122	ONLY: nested_run
123
124	USE spectra_mod, &
125	ONLY: calculate_spectra
126
127	USE synthetic_turbulence_generator_mod, &
128	ONLY: id_stg_left, id_stg_north, id_stg_right, id_stg_south
129	#endif
130
131	USE transpose_indices, &
132	ONLY: nxl_y, nxl_z, nxr_y, nxr_z, nyn_x, nyn_z, nys_x, nys_z, nzb_x, nzb_y, nzt_x, nzt_y
133
134	#if defined( __parallel )
135	USE transpose_indices, &
136	ONLY: nxl_yd, nxr_yd, nzb_yd, nzt_yd
137	#endif
138
139	IMPLICIT NONE
140
141	INTEGER(iwp) :: i !< running index over number of processors or number of multigrid level
142	#if defined( __parallel )
143	INTEGER(iwp) :: id_inflow_l !< ID indicating processors located at the left inflow boundary
144	INTEGER(iwp) :: id_outflow_l !< local value of id_outflow
145	INTEGER(iwp) :: id_outflow_source_l !< local value of id_outflow_source
146	INTEGER(iwp) :: id_recycling_l !< ID indicating processors located at the recycling plane
147	INTEGER(iwp) :: id_stg_left_l !< left lateral boundary local core id in case of turbulence generator
148	INTEGER(iwp) :: id_stg_north_l !< north lateral boundary local core id in case of turbulence generator
149	INTEGER(iwp) :: id_stg_right_l !< right lateral boundary local core id in case of turbulence generator
150	INTEGER(iwp) :: id_stg_south_l !< south lateral boundary local core id in case of turbulence generator
151	INTEGER(iwp) :: ind(5) !< array containing the subdomain bounds
152	#endif
153	INTEGER(iwp) :: j !< running index, used for various loops
154	INTEGER(iwp) :: k !< number of vertical grid points in different multigrid level
155	INTEGER(iwp) :: maximum_grid_level_l !< maximum number of grid level without switching to PE 0
156	INTEGER(iwp) :: mg_levels_x !< maximum number of grid level allowed along x-direction
157	INTEGER(iwp) :: mg_levels_y !< maximum number of grid level allowed along y-direction
158	INTEGER(iwp) :: mg_levels_z !< maximum number of grid level allowed along z-direction
159	INTEGER(iwp) :: mg_switch_to_pe0_level_l !< maximum number of grid level with switching to PE 0
160	#if defined( __parallel )
161	INTEGER(iwp) :: nnx_y !< quotient of number of grid points along x-direction and number of PEs used along y-direction
162	INTEGER(iwp) :: nny_x !< quotient of number of grid points along y-direction and number of PEs used along x-direction
163	INTEGER(iwp) :: nny_z !< quotient of number of grid points along y-direction and number of PEs used along x-direction
164	INTEGER(iwp) :: nnz_x !< quotient of number of grid points along z-direction and number of PEs used along x-direction
165	INTEGER(iwp) :: nnz_y !< quotient of number of grid points along z-direction and number of PEs used along x-direction
166	INTEGER(iwp) :: numproc_sqr !< square root of the number of processors
167	#endif
168	INTEGER(iwp) :: nxl_l !< lower index bound along x-direction on subdomain and different multigrid level
169	INTEGER(iwp) :: nxr_l !< upper index bound along x-direction on subdomain and different multigrid level
170	INTEGER(iwp) :: nyn_l !< lower index bound along y-direction on subdomain and different multigrid level
171	INTEGER(iwp) :: nys_l !< upper index bound along y-direction on subdomain and different multigrid level
172	#if defined( __parallel )
173	INTEGER(iwp) :: nzb_l !< lower index bound along z-direction on subdomain and different multigrid level
174	#endif
175	INTEGER(iwp) :: nzt_l !< upper index bound along z-direction on subdomain and different multigrid level
176	!$ INTEGER(iwp) :: omp_get_num_threads !< number of OpenMP threads
177
178	#if defined( __parallel )
179	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: ind_all !< dummy array containing index bounds on subdomain, used for gathering
180	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nxlf !< lower index bound allong x-direction for every PE
181	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nxrf !< upper index bound allong x-direction for every PE
182	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nynf !< lower index bound allong y-direction for every PE
183	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nysf !< lower index bound allong y-direction for every PE
184
185	INTEGER(iwp), DIMENSION(2) :: pdims_remote !< number of PEs used for coupled model (only in atmospher-ocean coupling)
186	INTEGER(iwp) :: lcoord(2) !< PE coordinates of left neighbor along x and y
187	INTEGER(iwp) :: rcoord(2) !< PE coordinates of right neighbor along x and y
188	#endif
189
190	!
191	!-- Get the number of OpenMP threads
192	!$OMP PARALLEL
193	!$ threads_per_task = omp_get_num_threads()
194	!$OMP END PARALLEL
195
196
197	#if defined( __parallel )
198
199	CALL location_message( 'creating virtual PE grids + MPI derived data types', 'start' )
200
201	!
202	!-- Determine the processor topology or check it, if prescribed by the user
203	IF ( npex == -1 .AND. npey == -1 ) THEN
204
205	!
206	!-- Automatic determination of the topology
207	numproc_sqr = SQRT( REAL( numprocs, KIND=wp ) )
208	pdims(1) = MAX( numproc_sqr , 1 )
209	DO WHILE ( MOD( numprocs , pdims(1) ) /= 0 )
210	pdims(1) = pdims(1) - 1
211	ENDDO
212	pdims(2) = numprocs / pdims(1)
213
214	ELSEIF ( npex /= -1 .AND. npey /= -1 ) THEN
215
216	!
217	!-- Prescribed by user. Number of processors on the prescribed topology must be equal to the
218	!-- number of PEs available to the job
219	IF ( ( npex * npey ) /= numprocs ) THEN
220	WRITE( message_string, * ) 'number of PEs of the prescribed ', 'topology (', npex*npey, &
221	') does not match & the number of ', &
222	'PEs available to the job (', numprocs, ')'
223	CALL message( 'init_pegrid', 'PA0221', 1, 2, 0, 6, 0 )
224	ENDIF
225	pdims(1) = npex
226	pdims(2) = npey
227
228	ELSE
229	!
230	!-- If the processor topology is prescribed by the user, the number of
231	!-- PEs must be given in both directions
232	message_string = 'if the processor topology is prescribed by th' // &
233	'e user & both values of "npex" and "npey" must be given' // &
234	' in the &NAMELIST-parameter file'
235	CALL message( 'init_pegrid', 'PA0222', 1, 2, 0, 6, 0 )
236
237	ENDIF
238
239	!
240	!-- Create four default MPI communicators for the 2d virtual PE grid. One of them will be used as
241	!-- the main communicator for this run, while others might be used for specific quantities like
242	!-- aerosol, chemical species, or passive scalars), if their horizontal boundary conditions shall
243	!-- be different from those of the other quantities (e.g. non-cyclic conditions for aerosols, and
244	!-- cyclic conditions for all others).
245	DO i = 1, 4
246
247	IF ( i == 1 ) cyclic = (/ .TRUE., .TRUE. /) ! cyclic along x and y
248	IF ( i == 2 ) cyclic = (/ .TRUE., .FALSE. /) ! cyclic along x
249	IF ( i == 3 ) cyclic = (/ .FALSE., .TRUE. /) ! cyllic along y
250	IF ( i == 4 ) cyclic = (/ .FALSE., .FALSE. /) ! non-cyclic
251
252	CALL MPI_CART_CREATE( comm_palm, ndim, pdims, cyclic, reorder, &
253	communicator_configurations(i)%mpi_communicator, ierr )
254
255	CALL MPI_CART_SHIFT( communicator_configurations(i)%mpi_communicator, 0, 1, &
256	communicator_configurations(i)%pleft, &
257	communicator_configurations(i)%pright, ierr )
258
259	CALL MPI_CART_SHIFT( communicator_configurations(i)%mpi_communicator, 1, 1, &
260	communicator_configurations(i)%psouth, &
261	communicator_configurations(i)%pnorth, ierr )
262
263	ENDDO
264
265	!
266	!-- If necessary, set horizontal boundary conditions to non-cyclic
267	IF ( bc_lr /= 'cyclic' ) cyclic(1) = .FALSE.
268	IF ( bc_ns /= 'cyclic' ) cyclic(2) = .FALSE.
269
270
271	!
272	!-- Set the main communicator (virtual pe grid) for this run
273	IF ( bc_lr == 'cyclic' .AND. bc_ns == 'cyclic' ) i = 1
274	IF ( bc_lr == 'cyclic' .AND. bc_ns /= 'cyclic' ) i = 2
275	IF ( bc_lr /= 'cyclic' .AND. bc_ns == 'cyclic' ) i = 3
276	IF ( bc_lr /= 'cyclic' .AND. bc_ns /= 'cyclic' ) i = 4
277
278	comm2d = communicator_configurations(i)%mpi_communicator
279	pleft = communicator_configurations(i)%pleft
280	pright = communicator_configurations(i)%pright
281	psouth = communicator_configurations(i)%psouth
282	pnorth = communicator_configurations(i)%pnorth
283
284	!
285	!-- Set rank and coordinates of the main communicator
286	CALL MPI_COMM_RANK( comm2d, myid, ierr )
287	WRITE (myid_char,'(''_'',I6.6)') myid
288
289	CALL MPI_CART_COORDS( comm2d, myid, ndim, pcoord, ierr )
290
291	!
292	!-- In case of cyclic boundary conditions, a y-shift at the boundaries in x-direction can be
293	!-- introduced via parameter y_shift. The shift is done by modifying the processor grid in such a
294	!-- way that processors located at the x-boundary communicate across it to processors with
295	!-- y-coordinate shifted by y_shift relative to their own. This feature can not be used in
296	!-- combination with an fft pressure solver. It has been implemented to counter the effect of streak
297	!-- structures in case of cyclic boundary conditions. For a description of these see Munters
298	!-- (2016; dx.doi.org/10.1063/1.4941912)
299	!--
300	!-- Get coordinates of left and right neighbor on PE grid
301	IF ( y_shift /= 0 ) THEN
302	IF ( bc_lr == 'cyclic' ) THEN
303	IF ( TRIM( psolver ) /= 'multigrid' .AND. TRIM( psolver ) /= 'multigrid_noopt') THEN
304	message_string = 'y_shift /= 0 requires a multigrid pressure solver '
305	CALL message( 'check_parameters', 'PA0468', 1, 2, 0, 6, 0 )
306	ENDIF
307
308	CALL MPI_CART_COORDS( comm2d, pright, ndim, rcoord, ierr )
309	CALL MPI_CART_COORDS( comm2d, pleft, ndim, lcoord, ierr )
310
311	!
312	!-- If the x(y)-coordinate of the right (left) neighbor is smaller (greater) than that of the
313	!-- calling process, then the calling process is located on the right (left) boundary of the
314	!-- processor grid. In that case, the y-coordinate of that neighbor is increased (decreased)
315	!-- by y_shift.
316	!-- The rank of the process with that coordinate is then inquired and the neighbor rank for
317	!-- MPI_SENDRECV, pright (pleft) is set to it.
318	!-- In this way, the calling process receives a new right (left) neighbor for all future
319	!-- MPI_SENDRECV calls. That neighbor has a y-coordinate of y+(-)y_shift, where y is the
320	!-- original right (left) neighbor's y-coordinate. The modulo-operation ensures that if the
321	!-- neighbor's y-coordinate exceeds the grid-boundary, it will be relocated to the opposite
322	!-- part of the grid cyclicly.
323	IF ( rcoord(1) < pcoord(1) ) THEN
324	rcoord(2) = MODULO( rcoord(2) + y_shift, pdims(2) )
325	CALL MPI_CART_RANK( comm2d, rcoord, pright, ierr )
326	ENDIF
327
328	IF ( lcoord(1) > pcoord(1) ) THEN
329	lcoord(2) = MODULO( lcoord(2) - y_shift, pdims(2) )
330	CALL MPI_CART_RANK( comm2d, lcoord, pleft, ierr )
331	ENDIF
332
333	ELSE
334	!
335	!-- y-shift for non-cyclic boundary conditions is only implemented
336	!-- for the turbulence recycling method in inflow_turbulence.f90
337	IF ( .NOT. turbulent_inflow ) THEN
338	message_string = 'y_shift /= 0 is only allowed for cyclic ' // &
339	'boundary conditions in both directions ' // &
340	'or with turbulent_inflow == .TRUE.'
341	CALL message( 'check_parameters', 'PA0467', 1, 2, 0, 6, 0 )
342	ENDIF
343	ENDIF
344	ENDIF
345
346	!
347	!-- Determine sub-topologies for transpositions
348	!-- Transposition from z to x:
349	remain_dims(1) = .TRUE.
350	remain_dims(2) = .FALSE.
351	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dx, ierr )
352	CALL MPI_COMM_RANK( comm1dx, myidx, ierr )
353	!
354	!-- Transposition from x to y
355	remain_dims(1) = .FALSE.
356	remain_dims(2) = .TRUE.
357	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dy, ierr )
358	CALL MPI_COMM_RANK( comm1dy, myidy, ierr )
359
360	!
361	!-- Calculate array bounds along x-direction for every PE.
362	ALLOCATE( nxlf(0:pdims(1)-1), nxrf(0:pdims(1)-1), nynf(0:pdims(2)-1), nysf(0:pdims(2)-1) )
363
364	IF ( MOD( nx+1 , pdims(1) ) /= 0 ) THEN
365	WRITE( message_string, * ) 'x-direction: gridpoint number (' ,nx+1, ') ', &
366	'is not an& integral multiple of the number', &
367	' of processors (', pdims(1), ')'
368	CALL message( 'init_pegrid', 'PA0225', 1, 2, 0, 6, 0 )
369	ELSE
370	nnx = ( nx + 1 ) / pdims(1)
371	ENDIF
372
373	!
374	!-- Left and right array bounds, number of gridpoints
375	DO i = 0, pdims(1)-1
376	nxlf(i) = i * nnx
377	nxrf(i) = ( i + 1 ) * nnx - 1
378	ENDDO
379
380	!
381	!-- Calculate array bounds in y-direction for every PE.
382	IF ( MOD( ny+1 , pdims(2) ) /= 0 ) THEN
383	WRITE( message_string, * ) 'y-direction: gridpoint number (', ny+1, ') ', &
384	'is not an& integral multiple of the number', &
385	' of processors (', pdims(2), ')'
386	CALL message( 'init_pegrid', 'PA0227', 1, 2, 0, 6, 0 )
387	ELSE
388	nny = ( ny + 1 ) / pdims(2)
389	ENDIF
390
391	!
392	!-- South and north array bounds
393	DO j = 0, pdims(2)-1
394	nysf(j) = j * nny
395	nynf(j) = ( j + 1 ) * nny - 1
396	ENDDO
397
398	!
399	!-- Local array bounds of the respective PEs
400	nxl = nxlf(pcoord(1))
401	nxr = nxrf(pcoord(1))
402	nys = nysf(pcoord(2))
403	nyn = nynf(pcoord(2))
404	nzb = 0
405	nzt = nz
406	nnz = nz
407
408	!
409	!-- Set switches to define if the PE is situated at the border of the virtual processor grid
410	IF ( nxl == 0 ) left_border_pe = .TRUE.
411	IF ( nxr == nx ) right_border_pe = .TRUE.
412	IF ( nys == 0 ) south_border_pe = .TRUE.
413	IF ( nyn == ny ) north_border_pe = .TRUE.
414
415	!
416	!-- Calculate array bounds and gridpoint numbers for the transposed arrays (needed in the pressure
417	!-- solver)
418	!-- For the transposed arrays, cyclic boundaries as well as top and bottom boundaries are omitted,
419	!-- because they are obstructive to the transposition
420
421	!
422	!-- 1. transposition z --> x
423	!-- This transposition is not neccessary in case of a 1d-decomposition along x
424	IF ( psolver == 'poisfft' .OR. calculate_spectra ) THEN
425
426	IF ( pdims(2) /= 1 ) THEN
427	IF ( MOD( nz , pdims(1) ) /= 0 ) THEN
428	WRITE( message_string, * ) 'transposition z --> x:& ', &
429	'nz=', nz, ' is not an integral multiple ', &
430	'of pdims(1)=', pdims(1)
431	CALL message( 'init_pegrid', 'PA0230', 1, 2, 0, 6, 0 )
432	ENDIF
433	ENDIF
434
435	nys_x = nys
436	nyn_x = nyn
437	nny_x = nny
438	nnz_x = nz / pdims(1)
439	nzb_x = 1 + myidx * nnz_x
440	nzt_x = ( myidx + 1 ) * nnz_x
441	sendrecvcount_zx = nnx * nny * nnz_x
442
443	ENDIF
444
445
446	IF ( psolver == 'poisfft' ) THEN
447	!
448	!-- 2. transposition x --> y
449	IF ( MOD( nx+1 , pdims(2) ) /= 0 ) THEN
450	WRITE( message_string, * ) 'transposition x --> y:& ', &
451	'nx+1=', nx+1, ' is not an integral ', &
452	'multiple of pdims(2)=', pdims(2)
453	CALL message( 'init_pegrid', 'PA0231', 1, 2, 0, 6, 0 )
454	ENDIF
455
456	nnz_y = nnz_x
457	nzb_y = nzb_x
458	nzt_y = nzt_x
459	nnx_y = (nx+1) / pdims(2)
460	nxl_y = myidy * nnx_y
461	nxr_y = ( myidy + 1 ) * nnx_y - 1
462	sendrecvcount_xy = nnx_y * nny_x * nnz_y
463	!
464	!-- 3. transposition y --> z
465	!-- (ELSE: x --> y in case of 1D-decomposition along x)
466	nxl_z = nxl_y
467	nxr_z = nxr_y
468	nny_z = (ny+1) / pdims(1)
469	nys_z = myidx * nny_z
470	nyn_z = ( myidx + 1 ) * nny_z - 1
471	sendrecvcount_yz = nnx_y * nny_z * nnz_y
472
473	IF ( pdims(2) /= 1 ) THEN
474	!
475	!-- y --> z
476	!-- This transposition is not neccessary in case of a 1d-decomposition
477	!-- along x, except that the uptream-spline method is switched on
478	IF ( MOD( ny+1 , pdims(1) ) /= 0 ) THEN
479	WRITE( message_string, * ) 'transposition y --> z:& ', &
480	'ny+1=', ny+1, ' is not an integral ', &
481	'multiple of pdims(1)=', pdims(1)
482	CALL message( 'init_pegrid', 'PA0232', 1, 2, 0, 6, 0 )
483	ENDIF
484
485	ELSE
486	!
487	!-- x --> y
488	!-- This condition must be fulfilled for a 1D-decomposition along x
489	IF ( MOD( ny+1 , pdims(1) ) /= 0 ) THEN
490	WRITE( message_string, * ) 'transposition x --> y:& ', &
491	'ny+1=', ny+1, ' is not an integral ', &
492	'multiple of pdims(1)=', pdims(1)
493	CALL message( 'init_pegrid', 'PA0233', 1, 2, 0, 6, 0 )
494	ENDIF
495
496	ENDIF
497
498	ENDIF
499
500	!
501	!-- Indices for direct transpositions z --> y (used for calculating spectra)
502	IF ( calculate_spectra ) THEN
503	IF ( MOD( nz, pdims(2) ) /= 0 ) THEN
504	WRITE( message_string, * ) 'direct transposition z --> y (needed ', &
505	'for spectra):& nz=', nz, ' is not an ', &
506	'integral multiple of pdims(2)=', pdims(2)
507	CALL message( 'init_pegrid', 'PA0234', 1, 2, 0, 6, 0 )
508	ELSE
509	nxl_yd = nxl
510	nxr_yd = nxr
511	nzb_yd = 1 + myidy * ( nz / pdims(2) )
512	nzt_yd = ( myidy + 1 ) * ( nz / pdims(2) )
513	sendrecvcount_zyd = nnx * nny * ( nz / pdims(2) )
514	ENDIF
515	ENDIF
516
517	IF ( psolver == 'poisfft' .OR. calculate_spectra ) THEN
518	!
519	!-- Indices for direct transpositions y --> x
520	!-- (they are only possible in case of a 1d-decomposition along x)
521	IF ( pdims(2) == 1 ) THEN
522	nny_x = nny / pdims(1)
523	nys_x = myid * nny_x
524	nyn_x = ( myid + 1 ) * nny_x - 1
525	nzb_x = 1
526	nzt_x = nz
527	sendrecvcount_xy = nnx * nny_x * nz
528	ENDIF
529
530	ENDIF
531
532	IF ( psolver == 'poisfft' ) THEN
533	!
534	!-- Indices for direct transpositions x --> y
535	!-- (they are only possible in case of a 1d-decomposition along y)
536	IF ( pdims(1) == 1 ) THEN
537	nnx_y = nnx / pdims(2)
538	nxl_y = myid * nnx_y
539	nxr_y = ( myid + 1 ) * nnx_y - 1
540	nzb_y = 1
541	nzt_y = nz
542	sendrecvcount_xy = nnx_y * nny * nz
543	ENDIF
544
545	ENDIF
546
547	!
548	!-- Arrays for storing the array bounds are needed any more
549	DEALLOCATE( nxlf , nxrf , nynf , nysf )
550
551
552	!
553	!-- Collect index bounds from other PEs (to be written to restart file later)
554	ALLOCATE( hor_index_bounds(4,0:numprocs-1) )
555
556	IF ( myid == 0 ) THEN
557
558	hor_index_bounds(1,0) = nxl
559	hor_index_bounds(2,0) = nxr
560	hor_index_bounds(3,0) = nys
561	hor_index_bounds(4,0) = nyn
562
563	!
564	!-- Receive data from all other PEs
565	DO i = 1, numprocs-1
566	CALL MPI_RECV( ibuf, 4, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, ierr )
567	hor_index_bounds(:,i) = ibuf(1:4)
568	ENDDO
569
570	ELSE
571	!
572	!-- Send index bounds to PE0
573	ibuf(1) = nxl
574	ibuf(2) = nxr
575	ibuf(3) = nys
576	ibuf(4) = nyn
577	CALL MPI_SEND( ibuf, 4, MPI_INTEGER, 0, myid, comm2d, ierr )
578
579	ENDIF
580
581
582	#if defined( __print )
583	!
584	!-- Control output
585	IF ( myid == 0 ) THEN
586	PRINT, ' processor topology *'
587	PRINT*, ' '
588	PRINT*, 'myid pcoord left right south north idx idy nxl: nxr',' nys: nyn'
589	PRINT*, '------------------------------------------------------------','-----------'
590	WRITE (*,1000) 0, pcoord(1), pcoord(2), pleft, pright, psouth, pnorth, myidx, myidy, nxl, &
591	nxr, nys, nyn
592	1000 FORMAT (I4,2X,'(',I3,',',I3,')',3X,I4,2X,I4,3X,I4,2X,I4,2X,I3,1X,I3,2(2X,I4,':',I4))
593
594	!
595	!-- Receive data from the other PEs
596	DO i = 1,numprocs-1
597	CALL MPI_RECV( ibuf, 12, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, ierr )
598	WRITE (*,1000) i, ( ibuf(j) , j = 1,12 )
599	ENDDO
600	ELSE
601
602	!
603	!-- Send data to PE0
604	ibuf(1) = pcoord(1); ibuf(2) = pcoord(2); ibuf(3) = pleft
605	ibuf(4) = pright; ibuf(5) = psouth; ibuf(6) = pnorth; ibuf(7) = myidx
606	ibuf(8) = myidy; ibuf(9) = nxl; ibuf(10) = nxr; ibuf(11) = nys
607	ibuf(12) = nyn
608	CALL MPI_SEND( ibuf, 12, MPI_INTEGER, 0, myid, comm2d, ierr )
609	ENDIF
610	#endif
611
612	!
613	!-- Determine the number of ghost point layers
614	IF ( scalar_advec == 'ws-scheme' .OR. &
615	momentum_advec == 'ws-scheme' .OR. nested_run ) THEN
616	nbgp = 3
617	ELSE
618	nbgp = 1
619	ENDIF
620
621	!
622	!-- Check that the number of computational grid points is not smaller than the number of ghost
623	!-- points.
624	IF ( nnx < nbgp ) THEN
625	WRITE( message_string, * ) 'number of subdomain grid points along x (', nnx, ') is smaller',&
626	'than the number of ghost points (', nbgp, ')'
627	CALL message( 'init_pegrid', 'PA0682', 1, 2, 0, 6, 0 )
628	ENDIF
629	IF ( nny < nbgp ) THEN
630	WRITE( message_string, * ) 'number of subdomain grid points along y (', nny, ') is smaller',&
631	'than the number of ghost points (', nbgp, ')'
632	CALL message( 'init_pegrid', 'PA0683', 1, 2, 0, 6, 0 )
633	ENDIF
634
635	!
636	!-- Create a new MPI derived datatype for the exchange of surface (xy) data, which is needed for
637	!-- coupled atmosphere-ocean runs.
638	!-- First, calculate number of grid points of an xy-plane.
639	ngp_xy = ( nxr - nxl + 1 + 2 * nbgp ) * ( nyn - nys + 1 + 2 * nbgp )
640	CALL MPI_TYPE_VECTOR( ngp_xy, 1, nzt-nzb+2, MPI_REAL, type_xy, ierr )
641	CALL MPI_TYPE_COMMIT( type_xy, ierr )
642
643	IF ( TRIM( coupling_mode ) /= 'uncoupled' ) THEN
644
645	!
646	!-- Pass the number of grid points of the atmosphere model to
647	!-- the ocean model and vice versa
648	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
649
650	nx_a = nx
651	ny_a = ny
652
653	IF ( myid == 0 ) THEN
654
655	CALL MPI_SEND( nx_a, 1, MPI_INTEGER, numprocs, 1, comm_inter, ierr )
656	CALL MPI_SEND( ny_a, 1, MPI_INTEGER, numprocs, 2, comm_inter, ierr )
657	CALL MPI_SEND( pdims, 2, MPI_INTEGER, numprocs, 3, comm_inter, ierr )
658	CALL MPI_RECV( nx_o, 1, MPI_INTEGER, numprocs, 4, comm_inter, status, ierr )
659	CALL MPI_RECV( ny_o, 1, MPI_INTEGER, numprocs, 5, comm_inter, status, ierr )
660	CALL MPI_RECV( pdims_remote, 2, MPI_INTEGER, numprocs, 6, comm_inter, status, ierr )
661	ENDIF
662
663	CALL MPI_BCAST( nx_o, 1, MPI_INTEGER, 0, comm2d, ierr )
664	CALL MPI_BCAST( ny_o, 1, MPI_INTEGER, 0, comm2d, ierr )
665	CALL MPI_BCAST( pdims_remote, 2, MPI_INTEGER, 0, comm2d, ierr )
666
667	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
668
669	nx_o = nx
670	ny_o = ny
671
672	IF ( myid == 0 ) THEN
673
674	CALL MPI_RECV( nx_a, 1, MPI_INTEGER, 0, 1, comm_inter, status, ierr )
675	CALL MPI_RECV( ny_a, 1, MPI_INTEGER, 0, 2, comm_inter, status, ierr )
676	CALL MPI_RECV( pdims_remote, 2, MPI_INTEGER, 0, 3, comm_inter, status, ierr )
677	CALL MPI_SEND( nx_o, 1, MPI_INTEGER, 0, 4, comm_inter, ierr )
678	CALL MPI_SEND( ny_o, 1, MPI_INTEGER, 0, 5, comm_inter, ierr )
679	CALL MPI_SEND( pdims, 2, MPI_INTEGER, 0, 6, comm_inter, ierr )
680	ENDIF
681
682	CALL MPI_BCAST( nx_a, 1, MPI_INTEGER, 0, comm2d, ierr)
683	CALL MPI_BCAST( ny_a, 1, MPI_INTEGER, 0, comm2d, ierr)
684	CALL MPI_BCAST( pdims_remote, 2, MPI_INTEGER, 0, comm2d, ierr)
685
686	ENDIF
687
688	ngp_a = ( nx_a+1 + 2 * nbgp ) * ( ny_a+1 + 2 * nbgp )
689	ngp_o = ( nx_o+1 + 2 * nbgp ) * ( ny_o+1 + 2 * nbgp )
690
691	!
692	!-- Determine if the horizontal grid and the number of PEs in ocean and atmosphere is same or not.
693	IF ( nx_o == nx_a .AND. ny_o == ny_a .AND. &
694	pdims(1) == pdims_remote(1) .AND. pdims(2) == pdims_remote(2) ) THEN
695	coupling_topology = 0
696	ELSE
697	coupling_topology = 1
698	ENDIF
699
700	!
701	!-- Determine the target PEs for the exchange between ocean and atmosphere (comm2d)
702	IF ( coupling_topology == 0 ) THEN
703	!
704	!-- In case of identical topologies, every atmosphere PE has exactly one ocean PE counterpart
705	!-- and vice versa
706	IF ( TRIM( coupling_mode ) == 'atmosphere_to_ocean' ) THEN
707	target_id = myid + numprocs
708	ELSE
709	target_id = myid
710	ENDIF
711
712	ELSE
713	!
714	!-- In case of nonequivalent topology in ocean and atmosphere only for PE0 in ocean and PE0 in
715	!-- atmosphere a target_id is needed, since data echxchange between ocean and atmosphere will
716	!-- be done only between these PEs.
717	IF ( myid == 0 ) THEN
718
719	IF ( TRIM( coupling_mode ) == 'atmosphere_to_ocean' ) THEN
720	target_id = numprocs
721	ELSE
722	target_id = 0
723	ENDIF
724
725	ENDIF
726
727	ENDIF
728
729	ENDIF
730
731	#else
732
733	!
734	!-- Array bounds when running on a single PE (respectively a non-parallel machine)
735	nxl = 0
736	nxr = nx
737	nnx = nxr - nxl + 1
738	nys = 0
739	nyn = ny
740	nny = nyn - nys + 1
741	nzb = 0
742	nzt = nz
743	nnz = nz
744
745	ALLOCATE( hor_index_bounds(4,0:0) )
746	hor_index_bounds(1,0) = nxl
747	hor_index_bounds(2,0) = nxr
748	hor_index_bounds(3,0) = nys
749	hor_index_bounds(4,0) = nyn
750
751	!
752	!-- Array bounds for the pressure solver (in the parallel code, these bounds are the ones for the
753	!-- transposed arrays)
754	nys_x = nys
755	nyn_x = nyn
756	nzb_x = nzb + 1
757	nzt_x = nzt
758
759	nxl_y = nxl
760	nxr_y = nxr
761	nzb_y = nzb + 1
762	nzt_y = nzt
763
764	nxl_z = nxl
765	nxr_z = nxr
766	nys_z = nys
767	nyn_z = nyn
768
769	#endif
770
771	!
772	!-- Calculate number of grid levels necessary for the multigrid poisson solver as well as the
773	!-- gridpoint indices on each level
774	IF ( psolver(1:9) == 'multigrid' ) THEN
775
776	!
777	!-- First calculate number of possible grid levels for the subdomains
778	mg_levels_x = 1
779	mg_levels_y = 1
780	mg_levels_z = 1
781
782	i = nnx
783	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
784	i = i / 2
785	mg_levels_x = mg_levels_x + 1
786	ENDDO
787
788	j = nny
789	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
790	j = j / 2
791	mg_levels_y = mg_levels_y + 1
792	ENDDO
793
794	k = nz ! do not use nnz because it might be > nz due to transposition
795	! requirements
796	DO WHILE ( MOD( k, 2 ) == 0 .AND. k /= 2 )
797	k = k / 2
798	mg_levels_z = mg_levels_z + 1
799	ENDDO
800	!
801	!-- The optimized MG-solver does not allow odd values for nz at the coarsest grid level
802	IF ( TRIM( psolver ) /= 'multigrid_noopt' ) THEN
803	IF ( MOD( k, 2 ) /= 0 ) mg_levels_z = mg_levels_z - 1
804	!
805	!-- An odd value of nz does not work. The finest level must have an even value.
806	IF ( mg_levels_z == 0 ) THEN
807	message_string = 'optimized multigrid method requires nz to be even'
808	CALL message( 'init_pegrid', 'PA0495', 1, 2, 0, 6, 0 )
809	ENDIF
810	ENDIF
811
812	maximum_grid_level = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
813	!
814	!-- Check if subdomain sizes prevents any coarsening.
815	!-- This case, the maximum number of grid levels is 1, i.e. effectively a Gauss-Seidel scheme is
816	!-- applied rather than a multigrid approach.
817	!-- Give a warning in this case.
818	IF ( maximum_grid_level == 1 .AND. mg_switch_to_pe0_level == -1 ) THEN
819	message_string = 'No grid coarsening possible, multigrid ' // &
820	'approach effectively reduces to a Gauss-Seidel ' // &
821	'scheme.'
822
823	CALL message( 'poismg', 'PA0648', 0, 1, 0, 6, 0 )
824	ENDIF
825
826	!
827	!-- Find out, if the total domain allows more levels. These additional levels are identically
828	!-- processed on all PEs.
829	IF ( numprocs > 1 .AND. mg_switch_to_pe0_level /= -1 ) THEN
830
831	IF ( mg_levels_z > MIN( mg_levels_x, mg_levels_y ) ) THEN
832
833	mg_switch_to_pe0_level_l = maximum_grid_level
834
835	mg_levels_x = 1
836	mg_levels_y = 1
837
838	i = nx+1
839	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
840	i = i / 2
841	mg_levels_x = mg_levels_x + 1
842	ENDDO
843
844	j = ny+1
845	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
846	j = j / 2
847	mg_levels_y = mg_levels_y + 1
848	ENDDO
849
850	maximum_grid_level_l = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
851
852	IF ( maximum_grid_level_l > mg_switch_to_pe0_level_l ) THEN
853	mg_switch_to_pe0_level_l = maximum_grid_level_l - mg_switch_to_pe0_level_l + 1
854	ELSE
855	mg_switch_to_pe0_level_l = 0
856	ENDIF
857
858	ELSE
859
860	mg_switch_to_pe0_level_l = 0
861	maximum_grid_level_l = maximum_grid_level
862
863	ENDIF
864
865	!
866	!-- Use switch level calculated above only if it is not pre-defined by user
867	IF ( mg_switch_to_pe0_level == 0 ) THEN
868	IF ( mg_switch_to_pe0_level_l /= 0 ) THEN
869	mg_switch_to_pe0_level = mg_switch_to_pe0_level_l
870	maximum_grid_level = maximum_grid_level_l
871	ENDIF
872
873	ELSE
874	!
875	!-- Check pre-defined value and reset to default, if neccessary
876	IF ( mg_switch_to_pe0_level < mg_switch_to_pe0_level_l .OR. &
877	mg_switch_to_pe0_level >= maximum_grid_level_l ) THEN
878	message_string = 'mg_switch_to_pe0_level ' // &
879	'out of range and reset to 0'
880	CALL message( 'init_pegrid', 'PA0235', 0, 1, 0, 6, 0 )
881	mg_switch_to_pe0_level = 0
882	ELSE
883	!
884	!-- Use the largest number of possible levels anyway and recalculate the switch level to
885	!-- this largest number of possible values
886	maximum_grid_level = maximum_grid_level_l
887
888	ENDIF
889
890	ENDIF
891
892	ENDIF
893
894	ALLOCATE( grid_level_count(maximum_grid_level), &
895	nxl_mg(0:maximum_grid_level), nxr_mg(0:maximum_grid_level), &
896	nyn_mg(0:maximum_grid_level), nys_mg(0:maximum_grid_level), &
897	nzt_mg(0:maximum_grid_level) )
898
899	grid_level_count = 0
900	!
901	!-- Index zero required as dummy due to definition of arrays f2 and p2 in recursive subroutine
902	!-- next_mg_level
903	nxl_mg(0) = 0; nxr_mg(0) = 0; nyn_mg(0) = 0; nys_mg(0) = 0; nzt_mg(0) = 0
904
905	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzt_l = nzt
906
907	DO i = maximum_grid_level, 1 , -1
908
909	IF ( i == mg_switch_to_pe0_level ) THEN
910	#if defined( __parallel )
911	!
912	!-- Save the grid size of the subdomain at the switch level, because it is needed in poismg.
913	ind(1) = nxl_l; ind(2) = nxr_l
914	ind(3) = nys_l; ind(4) = nyn_l
915	ind(5) = nzt_l
916	ALLOCATE( ind_all(5*numprocs), mg_loc_ind(5,0:numprocs-1) )
917	CALL MPI_ALLGATHER( ind, 5, MPI_INTEGER, ind_all, 5, &
918	MPI_INTEGER, comm2d, ierr )
919	DO j = 0, numprocs-1
920	DO k = 1, 5
921	mg_loc_ind(k,j) = ind_all(k+j*5)
922	ENDDO
923	ENDDO
924	DEALLOCATE( ind_all )
925	!
926	!-- Calculate the grid size of the total domain
927	nxr_l = ( nxr_l-nxl_l+1 ) * pdims(1) - 1
928	nxl_l = 0
929	nyn_l = ( nyn_l-nys_l+1 ) * pdims(2) - 1
930	nys_l = 0
931	!
932	!-- The size of this gathered array must not be larger than the array tend, which is used
933	!-- in the multigrid scheme as a temporary array. Therefore the subdomain size of an PE is
934	!-- calculated and the size of the gathered grid. These values are used in routines pres
935	!-- and poismg.
936	subdomain_size = ( nxr - nxl + 2 * nbgp + 1 ) * &
937	( nyn - nys + 2 * nbgp + 1 ) * ( nzt - nzb + 2 )
938	gathered_size = ( nxr_l - nxl_l + 3 ) * ( nyn_l - nys_l + 3 ) * ( nzt_l - nzb + 2 )
939
940	#else
941	message_string = 'multigrid gather/scatter impossible ' // &
942	'in non parallel mode'
943	CALL message( 'init_pegrid', 'PA0237', 1, 2, 0, 6, 0 )
944	#endif
945	ENDIF
946
947	nxl_mg(i) = nxl_l
948	nxr_mg(i) = nxr_l
949	nys_mg(i) = nys_l
950	nyn_mg(i) = nyn_l
951	nzt_mg(i) = nzt_l
952
953	nxl_l = nxl_l / 2
954	nxr_l = nxr_l / 2
955	nys_l = nys_l / 2
956	nyn_l = nyn_l / 2
957	nzt_l = nzt_l / 2
958
959	ENDDO
960
961	!
962	!-- Temporary problem: Currently calculation of maxerror in routine poismg crashes if grid data
963	!-- are collected on PE0 already on the finest grid level.
964	!-- To be solved later.
965	IF ( maximum_grid_level == mg_switch_to_pe0_level ) THEN
966	message_string = 'grid coarsening on subdomain level cannot be performed'
967	CALL message( 'poismg', 'PA0236', 1, 2, 0, 6, 0 )
968	ENDIF
969
970	ELSE
971
972	maximum_grid_level = 0
973
974	ENDIF
975
976	!
977	!-- Default level 0 tells exchange_horiz that all ghost planes have to be exchanged. grid_level is
978	!-- adjusted in poismg, where only one ghost plane is required.
979	grid_level = 0
980
981	#if defined( __parallel )
982	!
983	!-- Gridpoint number for the exchange of ghost points (y-line for 2D-arrays)
984	ngp_y = nyn - nys + 1 + 2 * nbgp
985
986	!
987	!-- Define new MPI derived datatypes for the exchange of ghost points in x- and y-direction for
988	!-- 2D-arrays (line)
989	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_REAL, type_x, ierr )
990	CALL MPI_TYPE_COMMIT( type_x, ierr )
991
992	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_REAL, type_y, ierr )
993	CALL MPI_TYPE_COMMIT( type_y, ierr )
994	!
995	!-- Define new MPI derived datatypes for the exchange of ghost points in x- and y-direction for
996	!-- 2D-INTEGER arrays (line) - on normal grid.
997	!-- Define types for 32-bit and 8-bit Integer. The 8-bit Integer are only required on normal grid,
998	!-- while 32-bit Integer may be also required on coarser grid level in case of multigrid solver.
999	!
1000	!-- 8-bit Integer
1001	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_BYTE, type_x_byte, ierr )
1002	CALL MPI_TYPE_COMMIT( type_x_byte, ierr )
1003
1004	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_BYTE, type_y_byte, ierr )
1005	CALL MPI_TYPE_COMMIT( type_y_byte, ierr )
1006	!
1007	!-- 32-bit Integer
1008	ALLOCATE( type_x_int(0:maximum_grid_level), type_y_int(0:maximum_grid_level) )
1009
1010	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_INTEGER, type_x_int(0), ierr )
1011	CALL MPI_TYPE_COMMIT( type_x_int(0), ierr )
1012
1013	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_INTEGER, type_y_int(0), ierr )
1014	CALL MPI_TYPE_COMMIT( type_y_int(0), ierr )
1015	!
1016	!-- Calculate gridpoint numbers for the exchange of ghost points along x (yz-plane for 3D-arrays)
1017	!-- and define MPI derived data type(s) for the exchange of ghost points in y-direction (xz-plane).
1018	!-- Do these calculations for the model grid and (if necessary) also for the coarser grid levels
1019	!-- used in the multigrid method
1020	ALLOCATE ( ngp_xz(0:maximum_grid_level), &
1021	ngp_xz_int(0:maximum_grid_level), &
1022	ngp_yz(0:maximum_grid_level), &
1023	ngp_yz_int(0:maximum_grid_level), &
1024	type_xz(0:maximum_grid_level), &
1025	type_xz_int(0:maximum_grid_level), &
1026	type_yz(0:maximum_grid_level), &
1027	type_yz_int(0:maximum_grid_level) )
1028
1029	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzb_l = nzb; nzt_l = nzt
1030
1031	!
1032	!-- Discern between the model grid, which needs nbgp ghost points and grid levels for the multigrid
1033	!-- scheme. In the latter case only one ghost point is necessary.
1034	!-- First definition of MPI-datatypes for exchange of ghost layers on normal grid. The following
1035	!-- loop is needed for data exchange in poismg.f90.
1036	!
1037	!-- Determine number of grid points of yz-layer for exchange
1038	ngp_yz(0) = (nzt - nzb + 2) * (nyn - nys + 1 + 2 * nbgp)
1039
1040	!
1041	!-- Define an MPI-datatype for the exchange of left/right boundaries.
1042	!-- Although data are contiguous in physical memory (which does not necessarily require an
1043	!-- MPI-derived datatype), the data exchange between left and right PE's using the MPI-derived type
1044	!-- is 10% faster than without.
1045	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2nbgp, nbgp(nzt-nzb+2), ngp_yz(0), MPI_REAL, type_xz(0), &
1046	ierr )
1047	CALL MPI_TYPE_COMMIT( type_xz(0), ierr )
1048
1049	CALL MPI_TYPE_VECTOR( nbgp, ngp_yz(0), ngp_yz(0), MPI_REAL, type_yz(0), ierr )
1050	CALL MPI_TYPE_COMMIT( type_yz(0), ierr )
1051
1052	!
1053	!-- Define data types for exchange of 3D Integer arrays.
1054	ngp_yz_int(0) = (nzt - nzb + 2) * (nyn - nys + 1 + 2 * nbgp)
1055
1056	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2nbgp, nbgp(nzt-nzb+2), ngp_yz_int(0), MPI_INTEGER, &
1057	type_xz_int(0), ierr )
1058	CALL MPI_TYPE_COMMIT( type_xz_int(0), ierr )
1059
1060	CALL MPI_TYPE_VECTOR( nbgp, ngp_yz_int(0), ngp_yz_int(0), MPI_INTEGER, type_yz_int(0), ierr )
1061	CALL MPI_TYPE_COMMIT( type_yz_int(0), ierr )
1062
1063	!
1064	!-- Definition of MPI-datatypes for multigrid method (coarser level grids)
1065	IF ( psolver(1:9) == 'multigrid' ) THEN
1066	!
1067	!-- Definition of MPI-datatyoe as above, but only 1 ghost level is used
1068	DO i = maximum_grid_level, 1 , -1
1069	!
1070	!-- For 3D-exchange on different multigrid level, one ghost point for REAL arrays, two ghost
1071	!-- points for INTEGER arrays
1072	ngp_xz(i) = (nzt_l - nzb_l + 2) * (nxr_l - nxl_l + 3)
1073	ngp_yz(i) = (nzt_l - nzb_l + 2) * (nyn_l - nys_l + 3)
1074
1075	ngp_xz_int(i) = (nzt_l - nzb_l + 2) * (nxr_l - nxl_l + 3)
1076	ngp_yz_int(i) = (nzt_l - nzb_l + 2) * (nyn_l - nys_l + 3)
1077	!
1078	!-- MPI data type for REAL arrays, for xz-layers
1079	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+3, nzt_l-nzb_l+2, ngp_yz(i), MPI_REAL, type_xz(i), &
1080	ierr )
1081	CALL MPI_TYPE_COMMIT( type_xz(i), ierr )
1082
1083	!
1084	!-- MPI data type for INTEGER arrays, for xz-layers
1085	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+3, nzt_l-nzb_l+2, ngp_yz_int(i), MPI_INTEGER, &
1086	type_xz_int(i), ierr )
1087	CALL MPI_TYPE_COMMIT( type_xz_int(i), ierr )
1088
1089	!
1090	!-- MPI data type for REAL arrays, for yz-layers
1091	CALL MPI_TYPE_VECTOR( 1, ngp_yz(i), ngp_yz(i), MPI_REAL, type_yz(i), ierr )
1092	CALL MPI_TYPE_COMMIT( type_yz(i), ierr )
1093	!
1094	!-- MPI data type for INTEGER arrays, for yz-layers
1095	CALL MPI_TYPE_VECTOR( 1, ngp_yz_int(i), ngp_yz_int(i), MPI_INTEGER, type_yz_int(i), ierr )
1096	CALL MPI_TYPE_COMMIT( type_yz_int(i), ierr )
1097
1098
1099	!-- For 2D-exchange of INTEGER arrays on coarser grid level, where 2 ghost points need to be
1100	!-- exchanged. Only required for 32-bit Integer arrays.
1101	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+5, 2, nyn_l-nys_l+5, MPI_INTEGER, type_x_int(i), ierr )
1102	CALL MPI_TYPE_COMMIT( type_x_int(i), ierr )
1103
1104
1105	CALL MPI_TYPE_VECTOR( 2, nyn_l-nys_l+5, nyn_l-nys_l+5, MPI_INTEGER, type_y_int(i), ierr )
1106	CALL MPI_TYPE_COMMIT( type_y_int(i), ierr )
1107
1108	nxl_l = nxl_l / 2
1109	nxr_l = nxr_l / 2
1110	nys_l = nys_l / 2
1111	nyn_l = nyn_l / 2
1112	nzt_l = nzt_l / 2
1113
1114	ENDDO
1115
1116	ENDIF
1117
1118	#endif
1119
1120	#if defined( __parallel )
1121	!
1122	!-- Setting of flags for inflow/outflow/nesting conditions.
1123	IF ( pleft == MPI_PROC_NULL ) THEN
1124	IF ( bc_lr == 'dirichlet/radiation' .OR. bc_lr == 'nested' .OR. &
1125	bc_lr == 'nesting_offline' ) THEN
1126	bc_dirichlet_l = .TRUE.
1127	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
1128	bc_radiation_l = .TRUE.
1129	ENDIF
1130	ENDIF
1131
1132	IF ( pright == MPI_PROC_NULL ) THEN
1133	IF ( bc_lr == 'dirichlet/radiation' ) THEN
1134	bc_radiation_r = .TRUE.
1135	ELSEIF ( bc_lr == 'radiation/dirichlet' .OR. bc_lr == 'nested' .OR. &
1136	bc_lr == 'nesting_offline' ) THEN
1137	bc_dirichlet_r = .TRUE.
1138	ENDIF
1139	ENDIF
1140
1141	IF ( psouth == MPI_PROC_NULL ) THEN
1142	IF ( bc_ns == 'dirichlet/radiation' ) THEN
1143	bc_radiation_s = .TRUE.
1144	ELSEIF ( bc_ns == 'radiation/dirichlet' .OR. bc_ns == 'nested' .OR. &
1145	bc_ns == 'nesting_offline' ) THEN
1146	bc_dirichlet_s = .TRUE.
1147	ENDIF
1148	ENDIF
1149
1150	IF ( pnorth == MPI_PROC_NULL ) THEN
1151	IF ( bc_ns == 'dirichlet/radiation' .OR. bc_ns == 'nested' .OR. &
1152	bc_ns == 'nesting_offline' ) THEN
1153	bc_dirichlet_n = .TRUE.
1154	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
1155	bc_radiation_n = .TRUE.
1156	ENDIF
1157	ENDIF
1158	!
1159	!-- In case of synthetic turbulence geneartor determine ids.
1160	!-- Please note, if no forcing or nesting is applied, the generator is applied only at the left
1161	!-- lateral boundary.
1162	IF ( syn_turb_gen ) THEN
1163	IF ( bc_dirichlet_l ) THEN
1164	id_stg_left_l = myidx
1165	ELSE
1166	id_stg_left_l = 0
1167	ENDIF
1168	IF ( bc_dirichlet_r ) THEN
1169	id_stg_right_l = myidx
1170	ELSE
1171	id_stg_right_l = 0
1172	ENDIF
1173	IF ( bc_dirichlet_s ) THEN
1174	id_stg_south_l = myidy
1175	ELSE
1176	id_stg_south_l = 0
1177	ENDIF
1178	IF ( bc_dirichlet_n ) THEN
1179	id_stg_north_l = myidy
1180	ELSE
1181	id_stg_north_l = 0
1182	ENDIF
1183
1184	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1185	CALL MPI_ALLREDUCE( id_stg_left_l, id_stg_left, 1, MPI_INTEGER, MPI_SUM, comm1dx, ierr )
1186
1187	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1188	CALL MPI_ALLREDUCE( id_stg_right_l, id_stg_right, 1, MPI_INTEGER, MPI_SUM, comm1dx, ierr )
1189
1190	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1191	CALL MPI_ALLREDUCE( id_stg_south_l, id_stg_south, 1, MPI_INTEGER, MPI_SUM, comm1dy, ierr )
1192
1193	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1194	CALL MPI_ALLREDUCE( id_stg_north_l, id_stg_north, 1, MPI_INTEGER, MPI_SUM, comm1dy, ierr )
1195
1196	ENDIF
1197
1198	!
1199	!-- Broadcast the id of the inflow PE
1200	IF ( bc_dirichlet_l ) THEN
1201	id_inflow_l = myidx
1202	ELSE
1203	id_inflow_l = 0
1204	ENDIF
1205	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1206	CALL MPI_ALLREDUCE( id_inflow_l, id_inflow, 1, MPI_INTEGER, MPI_SUM, comm1dx, ierr )
1207
1208	!
1209	!-- Broadcast the id of the recycling plane
1210	!-- WARNING: needs to be adjusted in case of inflows other than from left side!
1211	IF ( turbulent_inflow ) THEN
1212
1213	IF ( NINT( recycling_width / dx, KIND=idp ) >= nxl .AND. &
1214	NINT( recycling_width / dx, KIND=idp ) <= nxr ) THEN
1215	id_recycling_l = myidx
1216	ELSE
1217	id_recycling_l = 0
1218	ENDIF
1219	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1220	CALL MPI_ALLREDUCE( id_recycling_l, id_recycling, 1, MPI_INTEGER, MPI_SUM, comm1dx, ierr )
1221
1222	ENDIF
1223
1224	!
1225	!-- Broadcast the id of the outflow PE and outflow-source plane
1226	IF ( turbulent_outflow ) THEN
1227
1228	IF ( bc_radiation_r ) THEN
1229	id_outflow_l = myidx
1230	ELSE
1231	id_outflow_l = 0
1232	ENDIF
1233	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1234	CALL MPI_ALLREDUCE( id_outflow_l, id_outflow, 1, MPI_INTEGER, MPI_SUM, &
1235	comm1dx, ierr )
1236
1237	IF ( NINT( outflow_source_plane / dx ) >= nxl .AND. &
1238	NINT( outflow_source_plane / dx ) <= nxr ) THEN
1239	id_outflow_source_l = myidx
1240	ELSE
1241	id_outflow_source_l = 0
1242	ENDIF
1243	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1244	CALL MPI_ALLREDUCE( id_outflow_source_l, id_outflow_source, 1, MPI_INTEGER, MPI_SUM, &
1245	comm1dx, ierr )
1246
1247	ENDIF
1248
1249	CALL location_message( 'creating virtual PE grids + MPI derived data types', 'finished' )
1250
1251	#else
1252	IF ( bc_lr == 'dirichlet/radiation' ) THEN
1253	bc_dirichlet_l = .TRUE.
1254	bc_radiation_r = .TRUE.
1255	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
1256	bc_radiation_l = .TRUE.
1257	bc_dirichlet_r = .TRUE.
1258	ENDIF
1259
1260	IF ( bc_ns == 'dirichlet/radiation' ) THEN
1261	bc_dirichlet_n = .TRUE.
1262	bc_radiation_s = .TRUE.
1263	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
1264	bc_radiation_n = .TRUE.
1265	bc_dirichlet_s = .TRUE.
1266	ENDIF
1267	#endif
1268
1269	!
1270	!-- At the inflow or outflow, u or v, respectively, have to be calculated for one more grid point.
1271	IF ( bc_dirichlet_l .OR. bc_radiation_l ) THEN
1272	nxlu = nxl + 1
1273	ELSE
1274	nxlu = nxl
1275	ENDIF
1276	IF ( bc_dirichlet_s .OR. bc_radiation_s ) THEN
1277	nysv = nys + 1
1278	ELSE
1279	nysv = nys
1280	ENDIF
1281
1282	!
1283	!-- Allocate wall flag arrays used in the multigrid solver
1284	IF ( psolver(1:9) == 'multigrid' ) THEN
1285
1286	DO i = maximum_grid_level, 1, -1
1287
1288	SELECT CASE ( i )
1289
1290	CASE ( 1 )
1291	ALLOCATE( wall_flags_1(nzb:nzt_mg(i)+1, &
1292	nys_mg(i)-1:nyn_mg(i)+1, &
1293	nxl_mg(i)-1:nxr_mg(i)+1) )
1294
1295	CASE ( 2 )
1296	ALLOCATE( wall_flags_2(nzb:nzt_mg(i)+1, &
1297	nys_mg(i)-1:nyn_mg(i)+1, &
1298	nxl_mg(i)-1:nxr_mg(i)+1) )
1299
1300	CASE ( 3 )
1301	ALLOCATE( wall_flags_3(nzb:nzt_mg(i)+1, &
1302	nys_mg(i)-1:nyn_mg(i)+1, &
1303	nxl_mg(i)-1:nxr_mg(i)+1) )
1304
1305	CASE ( 4 )
1306	ALLOCATE( wall_flags_4(nzb:nzt_mg(i)+1, &
1307	nys_mg(i)-1:nyn_mg(i)+1, &
1308	nxl_mg(i)-1:nxr_mg(i)+1) )
1309
1310	CASE ( 5 )
1311	ALLOCATE( wall_flags_5(nzb:nzt_mg(i)+1, &
1312	nys_mg(i)-1:nyn_mg(i)+1, &
1313	nxl_mg(i)-1:nxr_mg(i)+1) )
1314
1315	CASE ( 6 )
1316	ALLOCATE( wall_flags_6(nzb:nzt_mg(i)+1, &
1317	nys_mg(i)-1:nyn_mg(i)+1, &
1318	nxl_mg(i)-1:nxr_mg(i)+1) )
1319
1320	CASE ( 7 )
1321	ALLOCATE( wall_flags_7(nzb:nzt_mg(i)+1, &
1322	nys_mg(i)-1:nyn_mg(i)+1, &
1323	nxl_mg(i)-1:nxr_mg(i)+1) )
1324
1325	CASE ( 8 )
1326	ALLOCATE( wall_flags_8(nzb:nzt_mg(i)+1, &
1327	nys_mg(i)-1:nyn_mg(i)+1, &
1328	nxl_mg(i)-1:nxr_mg(i)+1) )
1329
1330	CASE ( 9 )
1331	ALLOCATE( wall_flags_9(nzb:nzt_mg(i)+1, &
1332	nys_mg(i)-1:nyn_mg(i)+1, &
1333	nxl_mg(i)-1:nxr_mg(i)+1) )
1334
1335	CASE ( 10 )
1336	ALLOCATE( wall_flags_10(nzb:nzt_mg(i)+1, &
1337	nys_mg(i)-1:nyn_mg(i)+1, &
1338	nxl_mg(i)-1:nxr_mg(i)+1) )
1339
1340	CASE DEFAULT
1341	message_string = 'more than 10 multigrid levels'
1342	CALL message( 'init_pegrid', 'PA0238', 1, 2, 0, 6, 0 )
1343
1344	END SELECT
1345
1346	ENDDO
1347
1348	ENDIF
1349
1350	END SUBROUTINE init_pegrid

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

Context Navigation

source: palm/trunk/SOURCE/init_pegrid.f90 @ 4894

Download in other formats: