Home

Context Navigation

source: palm/trunk/SOURCE/init_pegrid.f90 @ 1809

Last change on this file since 1809 was 1805, checked in by maronga, 8 years ago
last commit documented
Property svn:keywords set to `Id`
File size: 42.9 KB

Line
1	!> @file init_pegrid.f90
2	!--------------------------------------------------------------------------------!
3	! This file is part of PALM.
4	!
5	! PALM is free software: you can redistribute it and/or modify it under the terms
6	! of the GNU General Public License as published by the Free Software Foundation,
7	! either version 3 of the License, or (at your option) any later version.
8	!
9	! PALM is distributed in the hope that it will be useful, but WITHOUT ANY
10	! WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
11	! A PARTICULAR PURPOSE. See the GNU General Public License for more details.
12	!
13	! You should have received a copy of the GNU General Public License along with
14	! PALM. If not, see <http://www.gnu.org/licenses/>.
15	!
16	! Copyright 1997-2014 Leibniz Universitaet Hannover
17	!--------------------------------------------------------------------------------!
18	!
19	! Current revisions:
20	! ------------------
21	!
22	!
23	! Former revisions:
24	! -----------------
25	! $Id: init_pegrid.f90 1805 2016-04-05 16:32:34Z raasch $
26	!
27	! 1804 2016-04-05 16:30:18Z maronga
28	! Removed code for parameter file check (__check)
29	!
30	! 1779 2016-03-03 08:01:28Z raasch
31	! changes regarding nested domain removed: virtual PE grid will be automatically
32	! calculated for nested runs too
33	!
34	! 1764 2016-02-28 12:45:19Z raasch
35	! cpp-statements for nesting removed
36	!
37	! 1762 2016-02-25 12:31:13Z hellstea
38	! Introduction of nested domain feature
39	!
40	! 1682 2015-10-07 23:56:08Z knoop
41	! Code annotations made doxygen readable
42	!
43	! 1677 2015-10-02 13:25:23Z boeske
44	! New MPI-data types for exchange of 3D integer arrays.
45	!
46	! 1575 2015-03-27 09:56:27Z raasch
47	! adjustments for psolver-queries, calculation of ngp_xz added
48	!
49	! 1565 2015-03-09 20:59:31Z suehring
50	! Refine if-clause for setting nbgp.
51	!
52	! 1557 2015-03-05 16:43:04Z suehring
53	! Adjustment for monotonic limiter
54	!
55	! 1468 2014-09-24 14:06:57Z maronga
56	! Adapted for use on up to 6-digit processor cores
57	!
58	! 1435 2014-07-21 10:37:02Z keck
59	! bugfix: added missing parameter coupling_mode_remote to ONLY-attribute
60	!
61	! 1402 2014-05-09 14:25:13Z raasch
62	! location messages modified
63	!
64	! 1384 2014-05-02 14:31:06Z raasch
65	! location messages added
66	!
67	! 1353 2014-04-08 15:21:23Z heinze
68	! REAL constants provided with KIND-attribute
69	!
70	! 1322 2014-03-20 16:38:49Z raasch
71	! REAL functions provided with KIND-attribute
72	!
73	! 1320 2014-03-20 08:40:49Z raasch
74	! ONLY-attribute added to USE-statements,
75	! kind-parameters added to all INTEGER and REAL declaration statements,
76	! kinds are defined in new module kinds,
77	! revision history before 2012 removed,
78	! comment fields (!:) to be used for variable explanations added to
79	! all variable declaration statements
80	!
81	! 1304 2014-03-12 10:29:42Z raasch
82	! bugfix: single core MPI runs missed some settings of transpose indices
83	!
84	! 1212 2013-08-15 08:46:27Z raasch
85	! error message for poisfft_hybrid removed
86	!
87	! 1159 2013-05-21 11:58:22Z fricke
88	! dirichlet/neumann and neumann/dirichlet removed
89	!
90	! 1139 2013-04-18 07:25:03Z raasch
91	! bugfix for calculating the id of the PE carrying the recycling plane
92	!
93	! 1111 2013-03-08 23:54:10Z raasch
94	! initialization of poisfft moved to module poisfft
95	!
96	! 1092 2013-02-02 11:24:22Z raasch
97	! unused variables removed
98	!
99	! 1056 2012-11-16 15:28:04Z raasch
100	! Indices for arrays n.._mg start from zero due to definition of arrays f2 and
101	! p2 as automatic arrays in recursive subroutine next_mg_level
102	!
103	! 1041 2012-11-06 02:36:29Z raasch
104	! a 2d virtual processor topology is used by default for all machines
105	!
106	! 1036 2012-10-22 13:43:42Z raasch
107	! code put under GPL (PALM 3.9)
108	!
109	! 1003 2012-09-14 14:35:53Z raasch
110	! subdomains must have identical size (grid matching = "match" removed)
111	!
112	! 1001 2012-09-13 14:08:46Z raasch
113	! all actions concerning upstream-spline-method removed
114	!
115	! 978 2012-08-09 08:28:32Z fricke
116	! dirichlet/neumann and neumann/dirichlet added
117	! nxlu and nysv are also calculated for inflow boundary
118	!
119	! 809 2012-01-30 13:32:58Z maronga
120	! Bugfix: replaced .AND. and .NOT. with && and ! in the preprocessor directives
121	!
122	! 807 2012-01-25 11:53:51Z maronga
123	! New cpp directive "__check" implemented which is used by check_namelist_files
124	!
125	! Revision 1.1 1997/07/24 11:15:09 raasch
126	! Initial revision
127	!
128	!
129	! Description:
130	! ------------
131	!> Determination of the virtual processor topology (if not prescribed by the
132	!> user)and computation of the grid point number and array bounds of the local
133	!> domains.
134	!------------------------------------------------------------------------------!
135	SUBROUTINE init_pegrid
136
137
138	USE control_parameters, &
139	ONLY: bc_lr, bc_ns, coupling_mode, coupling_mode_remote, &
140	coupling_topology, dt_dosp, gathered_size, grid_level, &
141	grid_level_count, host, inflow_l, inflow_n, inflow_r, inflow_s, &
142	io_blocks, io_group, maximum_grid_level, &
143	maximum_parallel_io_streams, message_string, &
144	mg_switch_to_pe0_level, momentum_advec, nest_bound_l, &
145	nest_bound_n, nest_bound_r, nest_bound_s, neutral, psolver, &
146	outflow_l, outflow_n, outflow_r, outflow_s, recycling_width, &
147	scalar_advec, subdomain_size
148
149	USE grid_variables, &
150	ONLY: dx
151
152	USE indices, &
153	ONLY: mg_loc_ind, nbgp, nnx, nny, nnz, nx, nx_a, nx_o, nxl, nxl_mg, &
154	nxlu, nxr, nxr_mg, ny, ny_a, ny_o, nyn, nyn_mg, nys, nys_mg, &
155	nysv, nz, nzb, nzt, nzt_mg, wall_flags_1, wall_flags_2, &
156	wall_flags_3, wall_flags_4, wall_flags_5, wall_flags_6, &
157	wall_flags_7, wall_flags_8, wall_flags_9, wall_flags_10
158
159	USE kinds
160
161	USE pegrid
162
163	USE transpose_indices, &
164	ONLY: nxl_y, nxl_yd, nxl_z, nxr_y, nxr_yd, nxr_z, nyn_x, nyn_z, nys_x,&
165	nys_z, nzb_x, nzb_y, nzb_yd, nzt_x, nzt_yd, nzt_y
166
167	IMPLICIT NONE
168
169	INTEGER(iwp) :: i !<
170	INTEGER(iwp) :: id_inflow_l !<
171	INTEGER(iwp) :: id_recycling_l !<
172	INTEGER(iwp) :: ind(5) !<
173	INTEGER(iwp) :: j !<
174	INTEGER(iwp) :: k !<
175	INTEGER(iwp) :: maximum_grid_level_l !<
176	INTEGER(iwp) :: mg_levels_x !<
177	INTEGER(iwp) :: mg_levels_y !<
178	INTEGER(iwp) :: mg_levels_z !<
179	INTEGER(iwp) :: mg_switch_to_pe0_level_l !<
180	INTEGER(iwp) :: nnx_y !<
181	INTEGER(iwp) :: nnx_z !<
182	INTEGER(iwp) :: nny_x !<
183	INTEGER(iwp) :: nny_z !<
184	INTEGER(iwp) :: nnz_x !<
185	INTEGER(iwp) :: nnz_y !<
186	INTEGER(iwp) :: numproc_sqr !<
187	INTEGER(iwp) :: nxl_l !<
188	INTEGER(iwp) :: nxr_l !<
189	INTEGER(iwp) :: nyn_l !<
190	INTEGER(iwp) :: nys_l !<
191	INTEGER(iwp) :: nzb_l !<
192	INTEGER(iwp) :: nzt_l !<
193	INTEGER(iwp) :: omp_get_num_threads !<
194
195	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: ind_all !<
196	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nxlf !<
197	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nxrf !<
198	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nynf !<
199	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nysf !<
200
201	INTEGER(iwp), DIMENSION(2) :: pdims_remote !<
202
203	#if defined( __mpi2 )
204	LOGICAL :: found !<
205	#endif
206
207	!
208	!-- Get the number of OpenMP threads
209	!$OMP PARALLEL
210	#if defined( __intel_openmp_bug )
211	threads_per_task = omp_get_num_threads()
212	#else
213	!$ threads_per_task = omp_get_num_threads()
214	#endif
215	!$OMP END PARALLEL
216
217
218	#if defined( __parallel )
219
220	CALL location_message( 'creating virtual PE grids + MPI derived data types', &
221	.FALSE. )
222
223	!
224	!-- Determine the processor topology or check it, if prescribed by the user
225	IF ( npex == -1 .AND. npey == -1 ) THEN
226
227	!
228	!-- Automatic determination of the topology
229	numproc_sqr = SQRT( REAL( numprocs, KIND=wp ) )
230	pdims(1) = MAX( numproc_sqr , 1 )
231	DO WHILE ( MOD( numprocs , pdims(1) ) /= 0 )
232	pdims(1) = pdims(1) - 1
233	ENDDO
234	pdims(2) = numprocs / pdims(1)
235
236	ELSEIF ( npex /= -1 .AND. npey /= -1 ) THEN
237
238	!
239	!-- Prescribed by user. Number of processors on the prescribed topology
240	!-- must be equal to the number of PEs available to the job
241	IF ( ( npex * npey ) /= numprocs ) THEN
242	WRITE( message_string, * ) 'number of PEs of the prescribed ', &
243	'topology (', npex*npey,') does not match & the number of ', &
244	'PEs available to the job (', numprocs, ')'
245	CALL message( 'init_pegrid', 'PA0221', 1, 2, 0, 6, 0 )
246	ENDIF
247	pdims(1) = npex
248	pdims(2) = npey
249
250	ELSE
251	!
252	!-- If the processor topology is prescribed by the user, the number of
253	!-- PEs must be given in both directions
254	message_string = 'if the processor topology is prescribed by th' // &
255	'e user& both values of "npex" and "npey" must be given' // &
256	' in the &NAMELIST-parameter file'
257	CALL message( 'init_pegrid', 'PA0222', 1, 2, 0, 6, 0 )
258
259	ENDIF
260
261	!
262	!-- For communication speedup, set barriers in front of collective
263	!-- communications by default on SGI-type systems
264	IF ( host(3:5) == 'sgi' ) collective_wait = .TRUE.
265
266	!
267	!-- If necessary, set horizontal boundary conditions to non-cyclic
268	IF ( bc_lr /= 'cyclic' ) cyclic(1) = .FALSE.
269	IF ( bc_ns /= 'cyclic' ) cyclic(2) = .FALSE.
270
271
272	!
273	!-- Create the virtual processor grid
274	CALL MPI_CART_CREATE( comm_palm, ndim, pdims, cyclic, reorder, &
275	comm2d, ierr )
276	CALL MPI_COMM_RANK( comm2d, myid, ierr )
277	WRITE (myid_char,'(''_'',I6.6)') myid
278
279	CALL MPI_CART_COORDS( comm2d, myid, ndim, pcoord, ierr )
280	CALL MPI_CART_SHIFT( comm2d, 0, 1, pleft, pright, ierr )
281	CALL MPI_CART_SHIFT( comm2d, 1, 1, psouth, pnorth, ierr )
282
283	!
284	!-- Determine sub-topologies for transpositions
285	!-- Transposition from z to x:
286	remain_dims(1) = .TRUE.
287	remain_dims(2) = .FALSE.
288	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dx, ierr )
289	CALL MPI_COMM_RANK( comm1dx, myidx, ierr )
290	!
291	!-- Transposition from x to y
292	remain_dims(1) = .FALSE.
293	remain_dims(2) = .TRUE.
294	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dy, ierr )
295	CALL MPI_COMM_RANK( comm1dy, myidy, ierr )
296
297
298	!
299	!-- Calculate array bounds along x-direction for every PE.
300	ALLOCATE( nxlf(0:pdims(1)-1), nxrf(0:pdims(1)-1), nynf(0:pdims(2)-1), &
301	nysf(0:pdims(2)-1) )
302
303	IF ( MOD( nx+1 , pdims(1) ) /= 0 ) THEN
304	WRITE( message_string, * ) 'x-direction: gridpoint number (',nx+1,') ',&
305	'is not an& integral divisor of the number ', &
306	'processors (', pdims(1),')'
307	CALL message( 'init_pegrid', 'PA0225', 1, 2, 0, 6, 0 )
308	ELSE
309	nnx = ( nx + 1 ) / pdims(1)
310	IF ( nnx*pdims(1) - ( nx + 1) > nnx ) THEN
311	WRITE( message_string, * ) 'x-direction: nx does not match the', &
312	'requirements given by the number of PEs &used', &
313	'& please use nx = ', nx - ( pdims(1) - ( nnx*pdims(1) &
314	- ( nx + 1 ) ) ), ' instead of nx =', nx
315	CALL message( 'init_pegrid', 'PA0226', 1, 2, 0, 6, 0 )
316	ENDIF
317	ENDIF
318
319	!
320	!-- Left and right array bounds, number of gridpoints
321	DO i = 0, pdims(1)-1
322	nxlf(i) = i * nnx
323	nxrf(i) = ( i + 1 ) * nnx - 1
324	ENDDO
325
326	!
327	!-- Calculate array bounds in y-direction for every PE.
328	IF ( MOD( ny+1 , pdims(2) ) /= 0 ) THEN
329	WRITE( message_string, * ) 'y-direction: gridpoint number (',ny+1,') ', &
330	'is not an& integral divisor of the number of', &
331	'processors (', pdims(2),')'
332	CALL message( 'init_pegrid', 'PA0227', 1, 2, 0, 6, 0 )
333	ELSE
334	nny = ( ny + 1 ) / pdims(2)
335	IF ( nny*pdims(2) - ( ny + 1) > nny ) THEN
336	WRITE( message_string, * ) 'y-direction: ny does not match the', &
337	'requirements given by the number of PEs &used ', &
338	'& please use ny = ', ny - ( pdims(2) - ( nnx*pdims(2) &
339	- ( ny + 1 ) ) ), ' instead of ny =', ny
340	CALL message( 'init_pegrid', 'PA0228', 1, 2, 0, 6, 0 )
341	ENDIF
342	ENDIF
343
344	!
345	!-- South and north array bounds
346	DO j = 0, pdims(2)-1
347	nysf(j) = j * nny
348	nynf(j) = ( j + 1 ) * nny - 1
349	ENDDO
350
351	!
352	!-- Local array bounds of the respective PEs
353	nxl = nxlf(pcoord(1))
354	nxr = nxrf(pcoord(1))
355	nys = nysf(pcoord(2))
356	nyn = nynf(pcoord(2))
357	nzb = 0
358	nzt = nz
359	nnz = nz
360
361	!
362	!-- Set switches to define if the PE is situated at the border of the virtual
363	!-- processor grid
364	IF ( nxl == 0 ) left_border_pe = .TRUE.
365	IF ( nxr == nx ) right_border_pe = .TRUE.
366	IF ( nys == 0 ) south_border_pe = .TRUE.
367	IF ( nyn == ny ) north_border_pe = .TRUE.
368
369	!
370	!-- Calculate array bounds and gridpoint numbers for the transposed arrays
371	!-- (needed in the pressure solver)
372	!-- For the transposed arrays, cyclic boundaries as well as top and bottom
373	!-- boundaries are omitted, because they are obstructive to the transposition
374
375	!
376	!-- 1. transposition z --> x
377	!-- This transposition is not neccessary in case of a 1d-decomposition along x
378	nys_x = nys
379	nyn_x = nyn
380	nny_x = nny
381	nnz_x = nz / pdims(1)
382	nzb_x = 1 + myidx * nnz_x
383	nzt_x = ( myidx + 1 ) * nnz_x
384	sendrecvcount_zx = nnx * nny * nnz_x
385
386	IF ( pdims(2) /= 1 ) THEN
387	IF ( MOD( nz , pdims(1) ) /= 0 ) THEN
388	WRITE( message_string, * ) 'transposition z --> x:', &
389	'&nz=',nz,' is not an integral divisior of pdims(1)=', &
390	pdims(1)
391	CALL message( 'init_pegrid', 'PA0230', 1, 2, 0, 6, 0 )
392	ENDIF
393	ENDIF
394
395	!
396	!-- 2. transposition x --> y
397	nnz_y = nnz_x
398	nzb_y = nzb_x
399	nzt_y = nzt_x
400	IF ( MOD( nx+1 , pdims(2) ) /= 0 ) THEN
401	WRITE( message_string, * ) 'transposition x --> y:', &
402	'&nx+1=',nx+1,' is not an integral divisor of ',&
403	'pdims(2)=',pdims(2)
404	CALL message( 'init_pegrid', 'PA0231', 1, 2, 0, 6, 0 )
405	ENDIF
406	nnx_y = (nx+1) / pdims(2)
407	nxl_y = myidy * nnx_y
408	nxr_y = ( myidy + 1 ) * nnx_y - 1
409	sendrecvcount_xy = nnx_y * nny_x * nnz_y
410
411	!
412	!-- 3. transposition y --> z (ELSE: x --> y in case of 1D-decomposition
413	!-- along x)
414	nnx_z = nnx_y
415	nxl_z = nxl_y
416	nxr_z = nxr_y
417	nny_z = (ny+1) / pdims(1)
418	nys_z = myidx * nny_z
419	nyn_z = ( myidx + 1 ) * nny_z - 1
420	sendrecvcount_yz = nnx_y * nny_z * nnz_y
421
422	IF ( pdims(2) /= 1 ) THEN
423	!
424	!-- y --> z
425	!-- This transposition is not neccessary in case of a 1d-decomposition
426	!-- along x, except that the uptream-spline method is switched on
427	IF ( MOD( ny+1 , pdims(1) ) /= 0 ) THEN
428	WRITE( message_string, * ) 'transposition y --> z:', &
429	'& ny+1=',ny+1,' is not an integral divisor of ',&
430	'pdims(1)=',pdims(1)
431	CALL message( 'init_pegrid', 'PA0232', 1, 2, 0, 6, 0 )
432	ENDIF
433
434	ELSE
435	!
436	!-- x --> y. This condition must be fulfilled for a 1D-decomposition along x
437	IF ( MOD( ny+1 , pdims(1) ) /= 0 ) THEN
438	WRITE( message_string, * ) 'transposition x --> y:', &
439	'& ny+1=',ny+1,' is not an integral divisor of ',&
440	'pdims(1)=',pdims(1)
441	CALL message( 'init_pegrid', 'PA0233', 1, 2, 0, 6, 0 )
442	ENDIF
443
444	ENDIF
445
446	!
447	!-- Indices for direct transpositions z --> y (used for calculating spectra)
448	IF ( dt_dosp /= 9999999.9_wp ) THEN
449	IF ( MOD( nz, pdims(2) ) /= 0 ) THEN
450	WRITE( message_string, * ) 'direct transposition z --> y (needed ', &
451	'for spectra):& nz=',nz,' is not an integral divisor of ',&
452	'pdims(2)=',pdims(2)
453	CALL message( 'init_pegrid', 'PA0234', 1, 2, 0, 6, 0 )
454	ELSE
455	nxl_yd = nxl
456	nxr_yd = nxr
457	nzb_yd = 1 + myidy * ( nz / pdims(2) )
458	nzt_yd = ( myidy + 1 ) * ( nz / pdims(2) )
459	sendrecvcount_zyd = nnx * nny * ( nz / pdims(2) )
460	ENDIF
461	ENDIF
462
463	!
464	!-- Indices for direct transpositions y --> x (they are only possible in case
465	!-- of a 1d-decomposition along x)
466	IF ( pdims(2) == 1 ) THEN
467	nny_x = nny / pdims(1)
468	nys_x = myid * nny_x
469	nyn_x = ( myid + 1 ) * nny_x - 1
470	nzb_x = 1
471	nzt_x = nz
472	sendrecvcount_xy = nnx * nny_x * nz
473	ENDIF
474
475	!
476	!-- Indices for direct transpositions x --> y (they are only possible in case
477	!-- of a 1d-decomposition along y)
478	IF ( pdims(1) == 1 ) THEN
479	nnx_y = nnx / pdims(2)
480	nxl_y = myid * nnx_y
481	nxr_y = ( myid + 1 ) * nnx_y - 1
482	nzb_y = 1
483	nzt_y = nz
484	sendrecvcount_xy = nnx_y * nny * nz
485	ENDIF
486
487	!
488	!-- Arrays for storing the array bounds are needed any more
489	DEALLOCATE( nxlf , nxrf , nynf , nysf )
490
491
492	!
493	!-- Collect index bounds from other PEs (to be written to restart file later)
494	ALLOCATE( hor_index_bounds(4,0:numprocs-1) )
495
496	IF ( myid == 0 ) THEN
497
498	hor_index_bounds(1,0) = nxl
499	hor_index_bounds(2,0) = nxr
500	hor_index_bounds(3,0) = nys
501	hor_index_bounds(4,0) = nyn
502
503	!
504	!-- Receive data from all other PEs
505	DO i = 1, numprocs-1
506	CALL MPI_RECV( ibuf, 4, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, &
507	ierr )
508	hor_index_bounds(:,i) = ibuf(1:4)
509	ENDDO
510
511	ELSE
512	!
513	!-- Send index bounds to PE0
514	ibuf(1) = nxl
515	ibuf(2) = nxr
516	ibuf(3) = nys
517	ibuf(4) = nyn
518	CALL MPI_SEND( ibuf, 4, MPI_INTEGER, 0, myid, comm2d, ierr )
519
520	ENDIF
521
522
523	#if defined( __print )
524	!
525	!-- Control output
526	IF ( myid == 0 ) THEN
527	PRINT, ' processor topology *'
528	PRINT*, ' '
529	PRINT*, 'myid pcoord left right south north idx idy nxl: nxr',&
530	&' nys: nyn'
531	PRINT*, '------------------------------------------------------------',&
532	&'-----------'
533	WRITE (*,1000) 0, pcoord(1), pcoord(2), pleft, pright, psouth, pnorth, &
534	myidx, myidy, nxl, nxr, nys, nyn
535	1000 FORMAT (I4,2X,'(',I3,',',I3,')',3X,I4,2X,I4,3X,I4,2X,I4,2X,I3,1X,I3, &
536	2(2X,I4,':',I4))
537
538	!
539	!-- Receive data from the other PEs
540	DO i = 1,numprocs-1
541	CALL MPI_RECV( ibuf, 12, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, &
542	ierr )
543	WRITE (*,1000) i, ( ibuf(j) , j = 1,12 )
544	ENDDO
545	ELSE
546
547	!
548	!-- Send data to PE0
549	ibuf(1) = pcoord(1); ibuf(2) = pcoord(2); ibuf(3) = pleft
550	ibuf(4) = pright; ibuf(5) = psouth; ibuf(6) = pnorth; ibuf(7) = myidx
551	ibuf(8) = myidy; ibuf(9) = nxl; ibuf(10) = nxr; ibuf(11) = nys
552	ibuf(12) = nyn
553	CALL MPI_SEND( ibuf, 12, MPI_INTEGER, 0, myid, comm2d, ierr )
554	ENDIF
555	#endif
556
557	#if defined( __parallel )
558	#if defined( __mpi2 )
559	!
560	!-- In case of coupled runs, get the port name on PE0 of the atmosphere model
561	!-- and pass it to PE0 of the ocean model
562	IF ( myid == 0 ) THEN
563
564	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
565
566	CALL MPI_OPEN_PORT( MPI_INFO_NULL, port_name, ierr )
567
568	CALL MPI_PUBLISH_NAME( 'palm_coupler', MPI_INFO_NULL, port_name, &
569	ierr )
570
571	!
572	!-- Write a flag file for the ocean model and the other atmosphere
573	!-- processes.
574	!-- There seems to be a bug in MPICH2 which causes hanging processes
575	!-- in case that execution of LOOKUP_NAME is continued too early
576	!-- (i.e. before the port has been created)
577	OPEN( 90, FILE='COUPLING_PORT_OPENED', FORM='FORMATTED' )
578	WRITE ( 90, '(''TRUE'')' )
579	CLOSE ( 90 )
580
581	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
582
583	!
584	!-- Continue only if the atmosphere model has created the port.
585	!-- There seems to be a bug in MPICH2 which causes hanging processes
586	!-- in case that execution of LOOKUP_NAME is continued too early
587	!-- (i.e. before the port has been created)
588	INQUIRE( FILE='COUPLING_PORT_OPENED', EXIST=found )
589	DO WHILE ( .NOT. found )
590	INQUIRE( FILE='COUPLING_PORT_OPENED', EXIST=found )
591	ENDDO
592
593	CALL MPI_LOOKUP_NAME( 'palm_coupler', MPI_INFO_NULL, port_name, ierr )
594
595	ENDIF
596
597	ENDIF
598
599	!
600	!-- In case of coupled runs, establish the connection between the atmosphere
601	!-- and the ocean model and define the intercommunicator (comm_inter)
602	CALL MPI_BARRIER( comm2d, ierr )
603	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
604
605	CALL MPI_COMM_ACCEPT( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD, &
606	comm_inter, ierr )
607	coupling_mode_remote = 'ocean_to_atmosphere'
608
609	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
610
611	CALL MPI_COMM_CONNECT( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD, &
612	comm_inter, ierr )
613	coupling_mode_remote = 'atmosphere_to_ocean'
614
615	ENDIF
616	#endif
617
618	!
619	!-- Determine the number of ghost point layers
620	IF ( ( scalar_advec == 'ws-scheme' .AND. .NOT. neutral ) .OR. &
621	scalar_advec == 'ws-scheme-mono' .OR. &
622	momentum_advec == 'ws-scheme' ) THEN
623	nbgp = 3
624	ELSE
625	nbgp = 1
626	ENDIF
627
628	!
629	!-- Create a new MPI derived datatype for the exchange of surface (xy) data,
630	!-- which is needed for coupled atmosphere-ocean runs.
631	!-- First, calculate number of grid points of an xy-plane.
632	ngp_xy = ( nxr - nxl + 1 + 2 * nbgp ) * ( nyn - nys + 1 + 2 * nbgp )
633	CALL MPI_TYPE_VECTOR( ngp_xy, 1, nzt-nzb+2, MPI_REAL, type_xy, ierr )
634	CALL MPI_TYPE_COMMIT( type_xy, ierr )
635
636	IF ( TRIM( coupling_mode ) /= 'uncoupled' ) THEN
637
638	!
639	!-- Pass the number of grid points of the atmosphere model to
640	!-- the ocean model and vice versa
641	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
642
643	nx_a = nx
644	ny_a = ny
645
646	IF ( myid == 0 ) THEN
647
648	CALL MPI_SEND( nx_a, 1, MPI_INTEGER, numprocs, 1, comm_inter, &
649	ierr )
650	CALL MPI_SEND( ny_a, 1, MPI_INTEGER, numprocs, 2, comm_inter, &
651	ierr )
652	CALL MPI_SEND( pdims, 2, MPI_INTEGER, numprocs, 3, comm_inter, &
653	ierr )
654	CALL MPI_RECV( nx_o, 1, MPI_INTEGER, numprocs, 4, comm_inter, &
655	status, ierr )
656	CALL MPI_RECV( ny_o, 1, MPI_INTEGER, numprocs, 5, comm_inter, &
657	status, ierr )
658	CALL MPI_RECV( pdims_remote, 2, MPI_INTEGER, numprocs, 6, &
659	comm_inter, status, ierr )
660	ENDIF
661
662	CALL MPI_BCAST( nx_o, 1, MPI_INTEGER, 0, comm2d, ierr )
663	CALL MPI_BCAST( ny_o, 1, MPI_INTEGER, 0, comm2d, ierr )
664	CALL MPI_BCAST( pdims_remote, 2, MPI_INTEGER, 0, comm2d, ierr )
665
666	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
667
668	nx_o = nx
669	ny_o = ny
670
671	IF ( myid == 0 ) THEN
672
673	CALL MPI_RECV( nx_a, 1, MPI_INTEGER, 0, 1, comm_inter, status, &
674	ierr )
675	CALL MPI_RECV( ny_a, 1, MPI_INTEGER, 0, 2, comm_inter, status, &
676	ierr )
677	CALL MPI_RECV( pdims_remote, 2, MPI_INTEGER, 0, 3, comm_inter, &
678	status, ierr )
679	CALL MPI_SEND( nx_o, 1, MPI_INTEGER, 0, 4, comm_inter, ierr )
680	CALL MPI_SEND( ny_o, 1, MPI_INTEGER, 0, 5, comm_inter, ierr )
681	CALL MPI_SEND( pdims, 2, MPI_INTEGER, 0, 6, comm_inter, ierr )
682	ENDIF
683
684	CALL MPI_BCAST( nx_a, 1, MPI_INTEGER, 0, comm2d, ierr)
685	CALL MPI_BCAST( ny_a, 1, MPI_INTEGER, 0, comm2d, ierr)
686	CALL MPI_BCAST( pdims_remote, 2, MPI_INTEGER, 0, comm2d, ierr)
687
688	ENDIF
689
690	ngp_a = ( nx_a+1 + 2 * nbgp ) * ( ny_a+1 + 2 * nbgp )
691	ngp_o = ( nx_o+1 + 2 * nbgp ) * ( ny_o+1 + 2 * nbgp )
692
693	!
694	!-- Determine if the horizontal grid and the number of PEs in ocean and
695	!-- atmosphere is same or not
696	IF ( nx_o == nx_a .AND. ny_o == ny_a .AND. &
697	pdims(1) == pdims_remote(1) .AND. pdims(2) == pdims_remote(2) ) &
698	THEN
699	coupling_topology = 0
700	ELSE
701	coupling_topology = 1
702	ENDIF
703
704	!
705	!-- Determine the target PEs for the exchange between ocean and
706	!-- atmosphere (comm2d)
707	IF ( coupling_topology == 0 ) THEN
708	!
709	!-- In case of identical topologies, every atmosphere PE has exactly one
710	!-- ocean PE counterpart and vice versa
711	IF ( TRIM( coupling_mode ) == 'atmosphere_to_ocean' ) THEN
712	target_id = myid + numprocs
713	ELSE
714	target_id = myid
715	ENDIF
716
717	ELSE
718	!
719	!-- In case of nonequivalent topology in ocean and atmosphere only for
720	!-- PE0 in ocean and PE0 in atmosphere a target_id is needed, since
721	!-- data echxchange between ocean and atmosphere will be done only
722	!-- between these PEs.
723	IF ( myid == 0 ) THEN
724
725	IF ( TRIM( coupling_mode ) == 'atmosphere_to_ocean' ) THEN
726	target_id = numprocs
727	ELSE
728	target_id = 0
729	ENDIF
730
731	ENDIF
732
733	ENDIF
734
735	ENDIF
736
737
738	#endif
739
740	#else
741
742	!
743	!-- Array bounds when running on a single PE (respectively a non-parallel
744	!-- machine)
745	nxl = 0
746	nxr = nx
747	nnx = nxr - nxl + 1
748	nys = 0
749	nyn = ny
750	nny = nyn - nys + 1
751	nzb = 0
752	nzt = nz
753	nnz = nz
754
755	ALLOCATE( hor_index_bounds(4,0:0) )
756	hor_index_bounds(1,0) = nxl
757	hor_index_bounds(2,0) = nxr
758	hor_index_bounds(3,0) = nys
759	hor_index_bounds(4,0) = nyn
760
761	!
762	!-- Array bounds for the pressure solver (in the parallel code, these bounds
763	!-- are the ones for the transposed arrays)
764	nys_x = nys
765	nyn_x = nyn
766	nzb_x = nzb + 1
767	nzt_x = nzt
768
769	nxl_y = nxl
770	nxr_y = nxr
771	nzb_y = nzb + 1
772	nzt_y = nzt
773
774	nxl_z = nxl
775	nxr_z = nxr
776	nys_z = nys
777	nyn_z = nyn
778
779	#endif
780
781	!
782	!-- Calculate number of grid levels necessary for the multigrid poisson solver
783	!-- as well as the gridpoint indices on each level
784	IF ( psolver(1:9) == 'multigrid' ) THEN
785
786	!
787	!-- First calculate number of possible grid levels for the subdomains
788	mg_levels_x = 1
789	mg_levels_y = 1
790	mg_levels_z = 1
791
792	i = nnx
793	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
794	i = i / 2
795	mg_levels_x = mg_levels_x + 1
796	ENDDO
797
798	j = nny
799	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
800	j = j / 2
801	mg_levels_y = mg_levels_y + 1
802	ENDDO
803
804	k = nz ! do not use nnz because it might be > nz due to transposition
805	! requirements
806	DO WHILE ( MOD( k, 2 ) == 0 .AND. k /= 2 )
807	k = k / 2
808	mg_levels_z = mg_levels_z + 1
809	ENDDO
810
811	maximum_grid_level = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
812
813	!
814	!-- Find out, if the total domain allows more levels. These additional
815	!-- levels are identically processed on all PEs.
816	IF ( numprocs > 1 .AND. mg_switch_to_pe0_level /= -1 ) THEN
817
818	IF ( mg_levels_z > MIN( mg_levels_x, mg_levels_y ) ) THEN
819
820	mg_switch_to_pe0_level_l = maximum_grid_level
821
822	mg_levels_x = 1
823	mg_levels_y = 1
824
825	i = nx+1
826	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
827	i = i / 2
828	mg_levels_x = mg_levels_x + 1
829	ENDDO
830
831	j = ny+1
832	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
833	j = j / 2
834	mg_levels_y = mg_levels_y + 1
835	ENDDO
836
837	maximum_grid_level_l = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
838
839	IF ( maximum_grid_level_l > mg_switch_to_pe0_level_l ) THEN
840	mg_switch_to_pe0_level_l = maximum_grid_level_l - &
841	mg_switch_to_pe0_level_l + 1
842	ELSE
843	mg_switch_to_pe0_level_l = 0
844	ENDIF
845
846	ELSE
847	mg_switch_to_pe0_level_l = 0
848	maximum_grid_level_l = maximum_grid_level
849
850	ENDIF
851
852	!
853	!-- Use switch level calculated above only if it is not pre-defined
854	!-- by user
855	IF ( mg_switch_to_pe0_level == 0 ) THEN
856	IF ( mg_switch_to_pe0_level_l /= 0 ) THEN
857	mg_switch_to_pe0_level = mg_switch_to_pe0_level_l
858	maximum_grid_level = maximum_grid_level_l
859	ENDIF
860
861	ELSE
862	!
863	!-- Check pre-defined value and reset to default, if neccessary
864	IF ( mg_switch_to_pe0_level < mg_switch_to_pe0_level_l .OR. &
865	mg_switch_to_pe0_level >= maximum_grid_level_l ) THEN
866	message_string = 'mg_switch_to_pe0_level ' // &
867	'out of range and reset to default (=0)'
868	CALL message( 'init_pegrid', 'PA0235', 0, 1, 0, 6, 0 )
869	mg_switch_to_pe0_level = 0
870	ELSE
871	!
872	!-- Use the largest number of possible levels anyway and recalculate
873	!-- the switch level to this largest number of possible values
874	maximum_grid_level = maximum_grid_level_l
875
876	ENDIF
877
878	ENDIF
879
880	ENDIF
881
882	ALLOCATE( grid_level_count(maximum_grid_level), &
883	nxl_mg(0:maximum_grid_level), nxr_mg(0:maximum_grid_level), &
884	nyn_mg(0:maximum_grid_level), nys_mg(0:maximum_grid_level), &
885	nzt_mg(0:maximum_grid_level) )
886
887	grid_level_count = 0
888	!
889	!-- Index zero required as dummy due to definition of arrays f2 and p2 in
890	!-- recursive subroutine next_mg_level
891	nxl_mg(0) = 0; nxr_mg(0) = 0; nyn_mg(0) = 0; nys_mg(0) = 0; nzt_mg(0) = 0
892
893	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzt_l = nzt
894
895	DO i = maximum_grid_level, 1 , -1
896
897	IF ( i == mg_switch_to_pe0_level ) THEN
898	#if defined( __parallel )
899	!
900	!-- Save the grid size of the subdomain at the switch level, because
901	!-- it is needed in poismg.
902	ind(1) = nxl_l; ind(2) = nxr_l
903	ind(3) = nys_l; ind(4) = nyn_l
904	ind(5) = nzt_l
905	ALLOCATE( ind_all(5*numprocs), mg_loc_ind(5,0:numprocs-1) )
906	CALL MPI_ALLGATHER( ind, 5, MPI_INTEGER, ind_all, 5, &
907	MPI_INTEGER, comm2d, ierr )
908	DO j = 0, numprocs-1
909	DO k = 1, 5
910	mg_loc_ind(k,j) = ind_all(k+j*5)
911	ENDDO
912	ENDDO
913	DEALLOCATE( ind_all )
914	!
915	!-- Calculate the grid size of the total domain
916	nxr_l = ( nxr_l-nxl_l+1 ) * pdims(1) - 1
917	nxl_l = 0
918	nyn_l = ( nyn_l-nys_l+1 ) * pdims(2) - 1
919	nys_l = 0
920	!
921	!-- The size of this gathered array must not be larger than the
922	!-- array tend, which is used in the multigrid scheme as a temporary
923	!-- array. Therefore the subdomain size of an PE is calculated and
924	!-- the size of the gathered grid. These values are used in
925	!-- routines pres and poismg
926	subdomain_size = ( nxr - nxl + 2 * nbgp + 1 ) * &
927	( nyn - nys + 2 * nbgp + 1 ) * ( nzt - nzb + 2 )
928	gathered_size = ( nxr_l - nxl_l + 3 ) * ( nyn_l - nys_l + 3 ) * &
929	( nzt_l - nzb + 2 )
930
931	#else
932	message_string = 'multigrid gather/scatter impossible ' // &
933	'in non parallel mode'
934	CALL message( 'init_pegrid', 'PA0237', 1, 2, 0, 6, 0 )
935	#endif
936	ENDIF
937
938	nxl_mg(i) = nxl_l
939	nxr_mg(i) = nxr_l
940	nys_mg(i) = nys_l
941	nyn_mg(i) = nyn_l
942	nzt_mg(i) = nzt_l
943
944	nxl_l = nxl_l / 2
945	nxr_l = nxr_l / 2
946	nys_l = nys_l / 2
947	nyn_l = nyn_l / 2
948	nzt_l = nzt_l / 2
949
950	ENDDO
951
952	!
953	!-- Temporary problem: Currently calculation of maxerror iin routine poismg crashes
954	!-- if grid data are collected on PE0 already on the finest grid level.
955	!-- To be solved later.
956	IF ( maximum_grid_level == mg_switch_to_pe0_level ) THEN
957	message_string = 'grid coarsening on subdomain level cannot be performed'
958	CALL message( 'poismg', 'PA0236', 1, 2, 0, 6, 0 )
959	ENDIF
960
961	ELSE
962
963	maximum_grid_level = 0
964
965	ENDIF
966
967	!
968	!-- Default level 0 tells exchange_horiz that all ghost planes have to be
969	!-- exchanged. grid_level is adjusted in poismg, where only one ghost plane
970	!-- is required.
971	grid_level = 0
972
973	#if defined( __parallel )
974	!
975	!-- Gridpoint number for the exchange of ghost points (y-line for 2D-arrays)
976	ngp_y = nyn - nys + 1 + 2 * nbgp
977
978	!
979	!-- Define new MPI derived datatypes for the exchange of ghost points in
980	!-- x- and y-direction for 2D-arrays (line)
981	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_REAL, type_x, &
982	ierr )
983	CALL MPI_TYPE_COMMIT( type_x, ierr )
984	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_INTEGER, &
985	type_x_int, ierr )
986	CALL MPI_TYPE_COMMIT( type_x_int, ierr )
987
988	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_REAL, type_y, ierr )
989	CALL MPI_TYPE_COMMIT( type_y, ierr )
990	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_INTEGER, type_y_int, ierr )
991	CALL MPI_TYPE_COMMIT( type_y_int, ierr )
992
993
994	!
995	!-- Calculate gridpoint numbers for the exchange of ghost points along x
996	!-- (yz-plane for 3D-arrays) and define MPI derived data type(s) for the
997	!-- exchange of ghost points in y-direction (xz-plane).
998	!-- Do these calculations for the model grid and (if necessary) also
999	!-- for the coarser grid levels used in the multigrid method
1000	ALLOCATE ( ngp_xz(0:maximum_grid_level), ngp_yz(0:maximum_grid_level), &
1001	type_xz(0:maximum_grid_level), type_yz(0:maximum_grid_level) )
1002
1003	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzb_l = nzb; nzt_l = nzt
1004
1005	!
1006	!-- Discern between the model grid, which needs nbgp ghost points and
1007	!-- grid levels for the multigrid scheme. In the latter case only one
1008	!-- ghost point is necessary.
1009	!-- First definition of MPI-datatypes for exchange of ghost layers on normal
1010	!-- grid. The following loop is needed for data exchange in poismg.f90.
1011	!
1012	!-- Determine number of grid points of yz-layer for exchange
1013	ngp_yz(0) = (nzt - nzb + 2) * (nyn - nys + 1 + 2 * nbgp)
1014
1015	!
1016	!-- Define an MPI-datatype for the exchange of left/right boundaries.
1017	!-- Although data are contiguous in physical memory (which does not
1018	!-- necessarily require an MPI-derived datatype), the data exchange between
1019	!-- left and right PE's using the MPI-derived type is 10% faster than without.
1020	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2nbgp, nbgp(nzt-nzb+2), ngp_yz(0), &
1021	MPI_REAL, type_xz(0), ierr )
1022	CALL MPI_TYPE_COMMIT( type_xz(0), ierr )
1023
1024	CALL MPI_TYPE_VECTOR( nbgp, ngp_yz(0), ngp_yz(0), MPI_REAL, type_yz(0), &
1025	ierr )
1026	CALL MPI_TYPE_COMMIT( type_yz(0), ierr )
1027
1028	!
1029	!-- Definition of MPI-datatypes for multigrid method (coarser level grids)
1030	IF ( psolver(1:9) == 'multigrid' ) THEN
1031	!
1032	!-- Definition of MPI-datatyoe as above, but only 1 ghost level is used
1033	DO i = maximum_grid_level, 1 , -1
1034
1035	ngp_xz(i) = (nzt_l - nzb_l + 2) * (nxr_l - nxl_l + 3)
1036	ngp_yz(i) = (nzt_l - nzb_l + 2) * (nyn_l - nys_l + 3)
1037
1038	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+3, nzt_l-nzb_l+2, ngp_yz(i), &
1039	MPI_REAL, type_xz(i), ierr )
1040	CALL MPI_TYPE_COMMIT( type_xz(i), ierr )
1041
1042	CALL MPI_TYPE_VECTOR( 1, ngp_yz(i), ngp_yz(i), MPI_REAL, type_yz(i), &
1043	ierr )
1044	CALL MPI_TYPE_COMMIT( type_yz(i), ierr )
1045
1046	nxl_l = nxl_l / 2
1047	nxr_l = nxr_l / 2
1048	nys_l = nys_l / 2
1049	nyn_l = nyn_l / 2
1050	nzt_l = nzt_l / 2
1051
1052	ENDDO
1053
1054	ENDIF
1055	!
1056	!-- Define data types for exchange of 3D Integer arrays.
1057	ngp_yz_int = (nzt - nzb + 2) * (nyn - nys + 1 + 2 * nbgp)
1058
1059	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2nbgp, nbgp(nzt-nzb+2), ngp_yz_int, &
1060	MPI_INTEGER, type_xz_int, ierr )
1061	CALL MPI_TYPE_COMMIT( type_xz_int, ierr )
1062
1063	CALL MPI_TYPE_VECTOR( nbgp, ngp_yz_int, ngp_yz_int, MPI_INTEGER, type_yz_int, &
1064	ierr )
1065	CALL MPI_TYPE_COMMIT( type_yz_int, ierr )
1066
1067	#endif
1068
1069	#if defined( __parallel )
1070	!
1071	!-- Setting of flags for inflow/outflow/nesting conditions in case of non-cyclic
1072	!-- horizontal boundary conditions.
1073	IF ( pleft == MPI_PROC_NULL ) THEN
1074	IF ( bc_lr == 'dirichlet/radiation' ) THEN
1075	inflow_l = .TRUE.
1076	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
1077	outflow_l = .TRUE.
1078	ELSEIF ( bc_lr == 'nested' ) THEN
1079	nest_bound_l = .TRUE.
1080	ENDIF
1081	ENDIF
1082
1083	IF ( pright == MPI_PROC_NULL ) THEN
1084	IF ( bc_lr == 'dirichlet/radiation' ) THEN
1085	outflow_r = .TRUE.
1086	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
1087	inflow_r = .TRUE.
1088	ELSEIF ( bc_lr == 'nested' ) THEN
1089	nest_bound_r = .TRUE.
1090	ENDIF
1091	ENDIF
1092
1093	IF ( psouth == MPI_PROC_NULL ) THEN
1094	IF ( bc_ns == 'dirichlet/radiation' ) THEN
1095	outflow_s = .TRUE.
1096	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
1097	inflow_s = .TRUE.
1098	ELSEIF ( bc_ns == 'nested' ) THEN
1099	nest_bound_s = .TRUE.
1100	ENDIF
1101	ENDIF
1102
1103	IF ( pnorth == MPI_PROC_NULL ) THEN
1104	IF ( bc_ns == 'dirichlet/radiation' ) THEN
1105	inflow_n = .TRUE.
1106	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
1107	outflow_n = .TRUE.
1108	ELSEIF ( bc_ns == 'nested' ) THEN
1109	nest_bound_n = .TRUE.
1110	ENDIF
1111	ENDIF
1112
1113	!
1114	!-- Broadcast the id of the inflow PE
1115	IF ( inflow_l ) THEN
1116	id_inflow_l = myidx
1117	ELSE
1118	id_inflow_l = 0
1119	ENDIF
1120	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1121	CALL MPI_ALLREDUCE( id_inflow_l, id_inflow, 1, MPI_INTEGER, MPI_SUM, &
1122	comm1dx, ierr )
1123
1124	!
1125	!-- Broadcast the id of the recycling plane
1126	!-- WARNING: needs to be adjusted in case of inflows other than from left side!
1127	IF ( NINT( recycling_width / dx ) >= nxl .AND. &
1128	NINT( recycling_width / dx ) <= nxr ) THEN
1129	id_recycling_l = myidx
1130	ELSE
1131	id_recycling_l = 0
1132	ENDIF
1133	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1134	CALL MPI_ALLREDUCE( id_recycling_l, id_recycling, 1, MPI_INTEGER, MPI_SUM, &
1135	comm1dx, ierr )
1136
1137	CALL location_message( 'finished', .TRUE. )
1138
1139	#else
1140	IF ( bc_lr == 'dirichlet/radiation' ) THEN
1141	inflow_l = .TRUE.
1142	outflow_r = .TRUE.
1143	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
1144	outflow_l = .TRUE.
1145	inflow_r = .TRUE.
1146	ENDIF
1147
1148	IF ( bc_ns == 'dirichlet/radiation' ) THEN
1149	inflow_n = .TRUE.
1150	outflow_s = .TRUE.
1151	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
1152	outflow_n = .TRUE.
1153	inflow_s = .TRUE.
1154	ENDIF
1155	#endif
1156
1157	!
1158	!-- At the inflow or outflow, u or v, respectively, have to be calculated for
1159	!-- one more grid point.
1160	IF ( inflow_l .OR. outflow_l .OR. nest_bound_l ) THEN
1161	nxlu = nxl + 1
1162	ELSE
1163	nxlu = nxl
1164	ENDIF
1165	IF ( inflow_s .OR. outflow_s .OR. nest_bound_s ) THEN
1166	nysv = nys + 1
1167	ELSE
1168	nysv = nys
1169	ENDIF
1170
1171	!
1172	!-- Allocate wall flag arrays used in the multigrid solver
1173	IF ( psolver(1:9) == 'multigrid' ) THEN
1174
1175	DO i = maximum_grid_level, 1, -1
1176
1177	SELECT CASE ( i )
1178
1179	CASE ( 1 )
1180	ALLOCATE( wall_flags_1(nzb:nzt_mg(i)+1, &
1181	nys_mg(i)-1:nyn_mg(i)+1, &
1182	nxl_mg(i)-1:nxr_mg(i)+1) )
1183
1184	CASE ( 2 )
1185	ALLOCATE( wall_flags_2(nzb:nzt_mg(i)+1, &
1186	nys_mg(i)-1:nyn_mg(i)+1, &
1187	nxl_mg(i)-1:nxr_mg(i)+1) )
1188
1189	CASE ( 3 )
1190	ALLOCATE( wall_flags_3(nzb:nzt_mg(i)+1, &
1191	nys_mg(i)-1:nyn_mg(i)+1, &
1192	nxl_mg(i)-1:nxr_mg(i)+1) )
1193
1194	CASE ( 4 )
1195	ALLOCATE( wall_flags_4(nzb:nzt_mg(i)+1, &
1196	nys_mg(i)-1:nyn_mg(i)+1, &
1197	nxl_mg(i)-1:nxr_mg(i)+1) )
1198
1199	CASE ( 5 )
1200	ALLOCATE( wall_flags_5(nzb:nzt_mg(i)+1, &
1201	nys_mg(i)-1:nyn_mg(i)+1, &
1202	nxl_mg(i)-1:nxr_mg(i)+1) )
1203
1204	CASE ( 6 )
1205	ALLOCATE( wall_flags_6(nzb:nzt_mg(i)+1, &
1206	nys_mg(i)-1:nyn_mg(i)+1, &
1207	nxl_mg(i)-1:nxr_mg(i)+1) )
1208
1209	CASE ( 7 )
1210	ALLOCATE( wall_flags_7(nzb:nzt_mg(i)+1, &
1211	nys_mg(i)-1:nyn_mg(i)+1, &
1212	nxl_mg(i)-1:nxr_mg(i)+1) )
1213
1214	CASE ( 8 )
1215	ALLOCATE( wall_flags_8(nzb:nzt_mg(i)+1, &
1216	nys_mg(i)-1:nyn_mg(i)+1, &
1217	nxl_mg(i)-1:nxr_mg(i)+1) )
1218
1219	CASE ( 9 )
1220	ALLOCATE( wall_flags_9(nzb:nzt_mg(i)+1, &
1221	nys_mg(i)-1:nyn_mg(i)+1, &
1222	nxl_mg(i)-1:nxr_mg(i)+1) )
1223
1224	CASE ( 10 )
1225	ALLOCATE( wall_flags_10(nzb:nzt_mg(i)+1, &
1226	nys_mg(i)-1:nyn_mg(i)+1, &
1227	nxl_mg(i)-1:nxr_mg(i)+1) )
1228
1229	CASE DEFAULT
1230	message_string = 'more than 10 multigrid levels'
1231	CALL message( 'init_pegrid', 'PA0238', 1, 2, 0, 6, 0 )
1232
1233	END SELECT
1234
1235	ENDDO
1236
1237	ENDIF
1238
1239	!
1240	!-- Calculate the number of groups into which parallel I/O is split.
1241	!-- The default for files which are opened by all PEs (or where each
1242	!-- PE opens his own independent file) is, that all PEs are doing input/output
1243	!-- in parallel at the same time. This might cause performance or even more
1244	!-- severe problems depending on the configuration of the underlying file
1245	!-- system.
1246	!-- First, set the default:
1247	IF ( maximum_parallel_io_streams == -1 .OR. &
1248	maximum_parallel_io_streams > numprocs ) THEN
1249	maximum_parallel_io_streams = numprocs
1250	ENDIF
1251
1252	!
1253	!-- Now calculate the number of io_blocks and the io_group to which the
1254	!-- respective PE belongs. I/O of the groups is done in serial, but in parallel
1255	!-- for all PEs belonging to the same group. A preliminary setting with myid
1256	!-- based on MPI_COMM_WORLD has been done in parin.
1257	io_blocks = numprocs / maximum_parallel_io_streams
1258	io_group = MOD( myid+1, io_blocks )
1259
1260
1261	END SUBROUTINE init_pegrid

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

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |