Home

Context Navigation

source: palm/trunk/SOURCE/init_pegrid.f90 @ 1822

Last change on this file since 1822 was 1818, checked in by maronga, 8 years ago
last commit documented / copyright update
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 1997-2016 Leibniz Universitaet Hannover
17	!--------------------------------------------------------------------------------!
18	!
19	! Current revisions:
20	! ------------------
21	!
22	!
23	! Former revisions:
24	! -----------------
25	! $Id: init_pegrid.f90 1818 2016-04-06 15:53:27Z hoffmann $
26	!
27	! 1815 2016-04-06 13:49:59Z raasch
28	! cpp-directives for intel openmp bug removed
29	!
30	! 1804 2016-04-05 16:30:18Z maronga
31	! Removed code for parameter file check (__check)
32	!
33	! 1779 2016-03-03 08:01:28Z raasch
34	! changes regarding nested domain removed: virtual PE grid will be automatically
35	! calculated for nested runs too
36	!
37	! 1764 2016-02-28 12:45:19Z raasch
38	! cpp-statements for nesting removed
39	!
40	! 1762 2016-02-25 12:31:13Z hellstea
41	! Introduction of nested domain feature
42	!
43	! 1682 2015-10-07 23:56:08Z knoop
44	! Code annotations made doxygen readable
45	!
46	! 1677 2015-10-02 13:25:23Z boeske
47	! New MPI-data types for exchange of 3D integer arrays.
48	!
49	! 1575 2015-03-27 09:56:27Z raasch
50	! adjustments for psolver-queries, calculation of ngp_xz added
51	!
52	! 1565 2015-03-09 20:59:31Z suehring
53	! Refine if-clause for setting nbgp.
54	!
55	! 1557 2015-03-05 16:43:04Z suehring
56	! Adjustment for monotonic limiter
57	!
58	! 1468 2014-09-24 14:06:57Z maronga
59	! Adapted for use on up to 6-digit processor cores
60	!
61	! 1435 2014-07-21 10:37:02Z keck
62	! bugfix: added missing parameter coupling_mode_remote to ONLY-attribute
63	!
64	! 1402 2014-05-09 14:25:13Z raasch
65	! location messages modified
66	!
67	! 1384 2014-05-02 14:31:06Z raasch
68	! location messages added
69	!
70	! 1353 2014-04-08 15:21:23Z heinze
71	! REAL constants provided with KIND-attribute
72	!
73	! 1322 2014-03-20 16:38:49Z raasch
74	! REAL functions provided with KIND-attribute
75	!
76	! 1320 2014-03-20 08:40:49Z raasch
77	! ONLY-attribute added to USE-statements,
78	! kind-parameters added to all INTEGER and REAL declaration statements,
79	! kinds are defined in new module kinds,
80	! revision history before 2012 removed,
81	! comment fields (!:) to be used for variable explanations added to
82	! all variable declaration statements
83	!
84	! 1304 2014-03-12 10:29:42Z raasch
85	! bugfix: single core MPI runs missed some settings of transpose indices
86	!
87	! 1212 2013-08-15 08:46:27Z raasch
88	! error message for poisfft_hybrid removed
89	!
90	! 1159 2013-05-21 11:58:22Z fricke
91	! dirichlet/neumann and neumann/dirichlet removed
92	!
93	! 1139 2013-04-18 07:25:03Z raasch
94	! bugfix for calculating the id of the PE carrying the recycling plane
95	!
96	! 1111 2013-03-08 23:54:10Z raasch
97	! initialization of poisfft moved to module poisfft
98	!
99	! 1092 2013-02-02 11:24:22Z raasch
100	! unused variables removed
101	!
102	! 1056 2012-11-16 15:28:04Z raasch
103	! Indices for arrays n.._mg start from zero due to definition of arrays f2 and
104	! p2 as automatic arrays in recursive subroutine next_mg_level
105	!
106	! 1041 2012-11-06 02:36:29Z raasch
107	! a 2d virtual processor topology is used by default for all machines
108	!
109	! 1036 2012-10-22 13:43:42Z raasch
110	! code put under GPL (PALM 3.9)
111	!
112	! 1003 2012-09-14 14:35:53Z raasch
113	! subdomains must have identical size (grid matching = "match" removed)
114	!
115	! 1001 2012-09-13 14:08:46Z raasch
116	! all actions concerning upstream-spline-method removed
117	!
118	! 978 2012-08-09 08:28:32Z fricke
119	! dirichlet/neumann and neumann/dirichlet added
120	! nxlu and nysv are also calculated for inflow boundary
121	!
122	! 809 2012-01-30 13:32:58Z maronga
123	! Bugfix: replaced .AND. and .NOT. with && and ! in the preprocessor directives
124	!
125	! 807 2012-01-25 11:53:51Z maronga
126	! New cpp directive "__check" implemented which is used by check_namelist_files
127	!
128	! Revision 1.1 1997/07/24 11:15:09 raasch
129	! Initial revision
130	!
131	!
132	! Description:
133	! ------------
134	!> Determination of the virtual processor topology (if not prescribed by the
135	!> user)and computation of the grid point number and array bounds of the local
136	!> domains.
137	!------------------------------------------------------------------------------!
138	SUBROUTINE init_pegrid
139
140
141	USE control_parameters, &
142	ONLY: bc_lr, bc_ns, coupling_mode, coupling_mode_remote, &
143	coupling_topology, dt_dosp, gathered_size, grid_level, &
144	grid_level_count, host, inflow_l, inflow_n, inflow_r, inflow_s, &
145	io_blocks, io_group, maximum_grid_level, &
146	maximum_parallel_io_streams, message_string, &
147	mg_switch_to_pe0_level, momentum_advec, nest_bound_l, &
148	nest_bound_n, nest_bound_r, nest_bound_s, neutral, psolver, &
149	outflow_l, outflow_n, outflow_r, outflow_s, recycling_width, &
150	scalar_advec, subdomain_size
151
152	USE grid_variables, &
153	ONLY: dx
154
155	USE indices, &
156	ONLY: mg_loc_ind, nbgp, nnx, nny, nnz, nx, nx_a, nx_o, nxl, nxl_mg, &
157	nxlu, nxr, nxr_mg, ny, ny_a, ny_o, nyn, nyn_mg, nys, nys_mg, &
158	nysv, nz, nzb, nzt, nzt_mg, wall_flags_1, wall_flags_2, &
159	wall_flags_3, wall_flags_4, wall_flags_5, wall_flags_6, &
160	wall_flags_7, wall_flags_8, wall_flags_9, wall_flags_10
161
162	USE kinds
163
164	USE pegrid
165
166	USE transpose_indices, &
167	ONLY: nxl_y, nxl_yd, nxl_z, nxr_y, nxr_yd, nxr_z, nyn_x, nyn_z, nys_x,&
168	nys_z, nzb_x, nzb_y, nzb_yd, nzt_x, nzt_yd, nzt_y
169
170	IMPLICIT NONE
171
172	INTEGER(iwp) :: i !<
173	INTEGER(iwp) :: id_inflow_l !<
174	INTEGER(iwp) :: id_recycling_l !<
175	INTEGER(iwp) :: ind(5) !<
176	INTEGER(iwp) :: j !<
177	INTEGER(iwp) :: k !<
178	INTEGER(iwp) :: maximum_grid_level_l !<
179	INTEGER(iwp) :: mg_levels_x !<
180	INTEGER(iwp) :: mg_levels_y !<
181	INTEGER(iwp) :: mg_levels_z !<
182	INTEGER(iwp) :: mg_switch_to_pe0_level_l !<
183	INTEGER(iwp) :: nnx_y !<
184	INTEGER(iwp) :: nnx_z !<
185	INTEGER(iwp) :: nny_x !<
186	INTEGER(iwp) :: nny_z !<
187	INTEGER(iwp) :: nnz_x !<
188	INTEGER(iwp) :: nnz_y !<
189	INTEGER(iwp) :: numproc_sqr !<
190	INTEGER(iwp) :: nxl_l !<
191	INTEGER(iwp) :: nxr_l !<
192	INTEGER(iwp) :: nyn_l !<
193	INTEGER(iwp) :: nys_l !<
194	INTEGER(iwp) :: nzb_l !<
195	INTEGER(iwp) :: nzt_l !<
196	INTEGER(iwp) :: omp_get_num_threads !<
197
198	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: ind_all !<
199	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nxlf !<
200	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nxrf !<
201	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nynf !<
202	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nysf !<
203
204	INTEGER(iwp), DIMENSION(2) :: pdims_remote !<
205
206	#if defined( __mpi2 )
207	LOGICAL :: found !<
208	#endif
209
210	!
211	!-- Get the number of OpenMP threads
212	!$OMP PARALLEL
213	!$ threads_per_task = omp_get_num_threads()
214	!$OMP END PARALLEL
215
216
217	#if defined( __parallel )
218
219	CALL location_message( 'creating virtual PE grids + MPI derived data types', &
220	.FALSE. )
221
222	!
223	!-- Determine the processor topology or check it, if prescribed by the user
224	IF ( npex == -1 .AND. npey == -1 ) THEN
225
226	!
227	!-- Automatic determination of the topology
228	numproc_sqr = SQRT( REAL( numprocs, KIND=wp ) )
229	pdims(1) = MAX( numproc_sqr , 1 )
230	DO WHILE ( MOD( numprocs , pdims(1) ) /= 0 )
231	pdims(1) = pdims(1) - 1
232	ENDDO
233	pdims(2) = numprocs / pdims(1)
234
235	ELSEIF ( npex /= -1 .AND. npey /= -1 ) THEN
236
237	!
238	!-- Prescribed by user. Number of processors on the prescribed topology
239	!-- must be equal to the number of PEs available to the job
240	IF ( ( npex * npey ) /= numprocs ) THEN
241	WRITE( message_string, * ) 'number of PEs of the prescribed ', &
242	'topology (', npex*npey,') does not match & the number of ', &
243	'PEs available to the job (', numprocs, ')'
244	CALL message( 'init_pegrid', 'PA0221', 1, 2, 0, 6, 0 )
245	ENDIF
246	pdims(1) = npex
247	pdims(2) = npey
248
249	ELSE
250	!
251	!-- If the processor topology is prescribed by the user, the number of
252	!-- PEs must be given in both directions
253	message_string = 'if the processor topology is prescribed by th' // &
254	'e user& both values of "npex" and "npey" must be given' // &
255	' in the &NAMELIST-parameter file'
256	CALL message( 'init_pegrid', 'PA0222', 1, 2, 0, 6, 0 )
257
258	ENDIF
259
260	!
261	!-- For communication speedup, set barriers in front of collective
262	!-- communications by default on SGI-type systems
263	IF ( host(3:5) == 'sgi' ) collective_wait = .TRUE.
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	!-- Create the virtual processor grid
273	CALL MPI_CART_CREATE( comm_palm, ndim, pdims, cyclic, reorder, &
274	comm2d, ierr )
275	CALL MPI_COMM_RANK( comm2d, myid, ierr )
276	WRITE (myid_char,'(''_'',I6.6)') myid
277
278	CALL MPI_CART_COORDS( comm2d, myid, ndim, pcoord, ierr )
279	CALL MPI_CART_SHIFT( comm2d, 0, 1, pleft, pright, ierr )
280	CALL MPI_CART_SHIFT( comm2d, 1, 1, psouth, pnorth, ierr )
281
282	!
283	!-- Determine sub-topologies for transpositions
284	!-- Transposition from z to x:
285	remain_dims(1) = .TRUE.
286	remain_dims(2) = .FALSE.
287	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dx, ierr )
288	CALL MPI_COMM_RANK( comm1dx, myidx, ierr )
289	!
290	!-- Transposition from x to y
291	remain_dims(1) = .FALSE.
292	remain_dims(2) = .TRUE.
293	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dy, ierr )
294	CALL MPI_COMM_RANK( comm1dy, myidy, ierr )
295
296
297	!
298	!-- Calculate array bounds along x-direction for every PE.
299	ALLOCATE( nxlf(0:pdims(1)-1), nxrf(0:pdims(1)-1), nynf(0:pdims(2)-1), &
300	nysf(0:pdims(2)-1) )
301
302	IF ( MOD( nx+1 , pdims(1) ) /= 0 ) THEN
303	WRITE( message_string, * ) 'x-direction: gridpoint number (',nx+1,') ',&
304	'is not an& integral divisor of the number ', &
305	'processors (', pdims(1),')'
306	CALL message( 'init_pegrid', 'PA0225', 1, 2, 0, 6, 0 )
307	ELSE
308	nnx = ( nx + 1 ) / pdims(1)
309	IF ( nnx*pdims(1) - ( nx + 1) > nnx ) THEN
310	WRITE( message_string, * ) 'x-direction: nx does not match the', &
311	'requirements given by the number of PEs &used', &
312	'& please use nx = ', nx - ( pdims(1) - ( nnx*pdims(1) &
313	- ( nx + 1 ) ) ), ' instead of nx =', nx
314	CALL message( 'init_pegrid', 'PA0226', 1, 2, 0, 6, 0 )
315	ENDIF
316	ENDIF
317
318	!
319	!-- Left and right array bounds, number of gridpoints
320	DO i = 0, pdims(1)-1
321	nxlf(i) = i * nnx
322	nxrf(i) = ( i + 1 ) * nnx - 1
323	ENDDO
324
325	!
326	!-- Calculate array bounds in y-direction for every PE.
327	IF ( MOD( ny+1 , pdims(2) ) /= 0 ) THEN
328	WRITE( message_string, * ) 'y-direction: gridpoint number (',ny+1,') ', &
329	'is not an& integral divisor of the number of', &
330	'processors (', pdims(2),')'
331	CALL message( 'init_pegrid', 'PA0227', 1, 2, 0, 6, 0 )
332	ELSE
333	nny = ( ny + 1 ) / pdims(2)
334	IF ( nny*pdims(2) - ( ny + 1) > nny ) THEN
335	WRITE( message_string, * ) 'y-direction: ny does not match the', &
336	'requirements given by the number of PEs &used ', &
337	'& please use ny = ', ny - ( pdims(2) - ( nnx*pdims(2) &
338	- ( ny + 1 ) ) ), ' instead of ny =', ny
339	CALL message( 'init_pegrid', 'PA0228', 1, 2, 0, 6, 0 )
340	ENDIF
341	ENDIF
342
343	!
344	!-- South and north array bounds
345	DO j = 0, pdims(2)-1
346	nysf(j) = j * nny
347	nynf(j) = ( j + 1 ) * nny - 1
348	ENDDO
349
350	!
351	!-- Local array bounds of the respective PEs
352	nxl = nxlf(pcoord(1))
353	nxr = nxrf(pcoord(1))
354	nys = nysf(pcoord(2))
355	nyn = nynf(pcoord(2))
356	nzb = 0
357	nzt = nz
358	nnz = nz
359
360	!
361	!-- Set switches to define if the PE is situated at the border of the virtual
362	!-- processor grid
363	IF ( nxl == 0 ) left_border_pe = .TRUE.
364	IF ( nxr == nx ) right_border_pe = .TRUE.
365	IF ( nys == 0 ) south_border_pe = .TRUE.
366	IF ( nyn == ny ) north_border_pe = .TRUE.
367
368	!
369	!-- Calculate array bounds and gridpoint numbers for the transposed arrays
370	!-- (needed in the pressure solver)
371	!-- For the transposed arrays, cyclic boundaries as well as top and bottom
372	!-- boundaries are omitted, because they are obstructive to the transposition
373
374	!
375	!-- 1. transposition z --> x
376	!-- This transposition is not neccessary in case of a 1d-decomposition along x
377	nys_x = nys
378	nyn_x = nyn
379	nny_x = nny
380	nnz_x = nz / pdims(1)
381	nzb_x = 1 + myidx * nnz_x
382	nzt_x = ( myidx + 1 ) * nnz_x
383	sendrecvcount_zx = nnx * nny * nnz_x
384
385	IF ( pdims(2) /= 1 ) THEN
386	IF ( MOD( nz , pdims(1) ) /= 0 ) THEN
387	WRITE( message_string, * ) 'transposition z --> x:', &
388	'&nz=',nz,' is not an integral divisior of pdims(1)=', &
389	pdims(1)
390	CALL message( 'init_pegrid', 'PA0230', 1, 2, 0, 6, 0 )
391	ENDIF
392	ENDIF
393
394	!
395	!-- 2. transposition x --> y
396	nnz_y = nnz_x
397	nzb_y = nzb_x
398	nzt_y = nzt_x
399	IF ( MOD( nx+1 , pdims(2) ) /= 0 ) THEN
400	WRITE( message_string, * ) 'transposition x --> y:', &
401	'&nx+1=',nx+1,' is not an integral divisor of ',&
402	'pdims(2)=',pdims(2)
403	CALL message( 'init_pegrid', 'PA0231', 1, 2, 0, 6, 0 )
404	ENDIF
405	nnx_y = (nx+1) / pdims(2)
406	nxl_y = myidy * nnx_y
407	nxr_y = ( myidy + 1 ) * nnx_y - 1
408	sendrecvcount_xy = nnx_y * nny_x * nnz_y
409
410	!
411	!-- 3. transposition y --> z (ELSE: x --> y in case of 1D-decomposition
412	!-- along x)
413	nnx_z = nnx_y
414	nxl_z = nxl_y
415	nxr_z = nxr_y
416	nny_z = (ny+1) / pdims(1)
417	nys_z = myidx * nny_z
418	nyn_z = ( myidx + 1 ) * nny_z - 1
419	sendrecvcount_yz = nnx_y * nny_z * nnz_y
420
421	IF ( pdims(2) /= 1 ) THEN
422	!
423	!-- y --> z
424	!-- This transposition is not neccessary in case of a 1d-decomposition
425	!-- along x, except that the uptream-spline method is switched on
426	IF ( MOD( ny+1 , pdims(1) ) /= 0 ) THEN
427	WRITE( message_string, * ) 'transposition y --> z:', &
428	'& ny+1=',ny+1,' is not an integral divisor of ',&
429	'pdims(1)=',pdims(1)
430	CALL message( 'init_pegrid', 'PA0232', 1, 2, 0, 6, 0 )
431	ENDIF
432
433	ELSE
434	!
435	!-- x --> y. This condition must be fulfilled for a 1D-decomposition along x
436	IF ( MOD( ny+1 , pdims(1) ) /= 0 ) THEN
437	WRITE( message_string, * ) 'transposition x --> y:', &
438	'& ny+1=',ny+1,' is not an integral divisor of ',&
439	'pdims(1)=',pdims(1)
440	CALL message( 'init_pegrid', 'PA0233', 1, 2, 0, 6, 0 )
441	ENDIF
442
443	ENDIF
444
445	!
446	!-- Indices for direct transpositions z --> y (used for calculating spectra)
447	IF ( dt_dosp /= 9999999.9_wp ) THEN
448	IF ( MOD( nz, pdims(2) ) /= 0 ) THEN
449	WRITE( message_string, * ) 'direct transposition z --> y (needed ', &
450	'for spectra):& nz=',nz,' is not an integral divisor of ',&
451	'pdims(2)=',pdims(2)
452	CALL message( 'init_pegrid', 'PA0234', 1, 2, 0, 6, 0 )
453	ELSE
454	nxl_yd = nxl
455	nxr_yd = nxr
456	nzb_yd = 1 + myidy * ( nz / pdims(2) )
457	nzt_yd = ( myidy + 1 ) * ( nz / pdims(2) )
458	sendrecvcount_zyd = nnx * nny * ( nz / pdims(2) )
459	ENDIF
460	ENDIF
461
462	!
463	!-- Indices for direct transpositions y --> x (they are only possible in case
464	!-- of a 1d-decomposition along x)
465	IF ( pdims(2) == 1 ) THEN
466	nny_x = nny / pdims(1)
467	nys_x = myid * nny_x
468	nyn_x = ( myid + 1 ) * nny_x - 1
469	nzb_x = 1
470	nzt_x = nz
471	sendrecvcount_xy = nnx * nny_x * nz
472	ENDIF
473
474	!
475	!-- Indices for direct transpositions x --> y (they are only possible in case
476	!-- of a 1d-decomposition along y)
477	IF ( pdims(1) == 1 ) THEN
478	nnx_y = nnx / pdims(2)
479	nxl_y = myid * nnx_y
480	nxr_y = ( myid + 1 ) * nnx_y - 1
481	nzb_y = 1
482	nzt_y = nz
483	sendrecvcount_xy = nnx_y * nny * nz
484	ENDIF
485
486	!
487	!-- Arrays for storing the array bounds are needed any more
488	DEALLOCATE( nxlf , nxrf , nynf , nysf )
489
490
491	!
492	!-- Collect index bounds from other PEs (to be written to restart file later)
493	ALLOCATE( hor_index_bounds(4,0:numprocs-1) )
494
495	IF ( myid == 0 ) THEN
496
497	hor_index_bounds(1,0) = nxl
498	hor_index_bounds(2,0) = nxr
499	hor_index_bounds(3,0) = nys
500	hor_index_bounds(4,0) = nyn
501
502	!
503	!-- Receive data from all other PEs
504	DO i = 1, numprocs-1
505	CALL MPI_RECV( ibuf, 4, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, &
506	ierr )
507	hor_index_bounds(:,i) = ibuf(1:4)
508	ENDDO
509
510	ELSE
511	!
512	!-- Send index bounds to PE0
513	ibuf(1) = nxl
514	ibuf(2) = nxr
515	ibuf(3) = nys
516	ibuf(4) = nyn
517	CALL MPI_SEND( ibuf, 4, MPI_INTEGER, 0, myid, comm2d, ierr )
518
519	ENDIF
520
521
522	#if defined( __print )
523	!
524	!-- Control output
525	IF ( myid == 0 ) THEN
526	PRINT, ' processor topology *'
527	PRINT*, ' '
528	PRINT*, 'myid pcoord left right south north idx idy nxl: nxr',&
529	&' nys: nyn'
530	PRINT*, '------------------------------------------------------------',&
531	&'-----------'
532	WRITE (*,1000) 0, pcoord(1), pcoord(2), pleft, pright, psouth, pnorth, &
533	myidx, myidy, nxl, nxr, nys, nyn
534	1000 FORMAT (I4,2X,'(',I3,',',I3,')',3X,I4,2X,I4,3X,I4,2X,I4,2X,I3,1X,I3, &
535	2(2X,I4,':',I4))
536
537	!
538	!-- Receive data from the other PEs
539	DO i = 1,numprocs-1
540	CALL MPI_RECV( ibuf, 12, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, &
541	ierr )
542	WRITE (*,1000) i, ( ibuf(j) , j = 1,12 )
543	ENDDO
544	ELSE
545
546	!
547	!-- Send data to PE0
548	ibuf(1) = pcoord(1); ibuf(2) = pcoord(2); ibuf(3) = pleft
549	ibuf(4) = pright; ibuf(5) = psouth; ibuf(6) = pnorth; ibuf(7) = myidx
550	ibuf(8) = myidy; ibuf(9) = nxl; ibuf(10) = nxr; ibuf(11) = nys
551	ibuf(12) = nyn
552	CALL MPI_SEND( ibuf, 12, MPI_INTEGER, 0, myid, comm2d, ierr )
553	ENDIF
554	#endif
555
556	#if defined( __parallel )
557	#if defined( __mpi2 )
558	!
559	!-- In case of coupled runs, get the port name on PE0 of the atmosphere model
560	!-- and pass it to PE0 of the ocean model
561	IF ( myid == 0 ) THEN
562
563	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
564
565	CALL MPI_OPEN_PORT( MPI_INFO_NULL, port_name, ierr )
566
567	CALL MPI_PUBLISH_NAME( 'palm_coupler', MPI_INFO_NULL, port_name, &
568	ierr )
569
570	!
571	!-- Write a flag file for the ocean model and the other atmosphere
572	!-- processes.
573	!-- There seems to be a bug in MPICH2 which causes hanging processes
574	!-- in case that execution of LOOKUP_NAME is continued too early
575	!-- (i.e. before the port has been created)
576	OPEN( 90, FILE='COUPLING_PORT_OPENED', FORM='FORMATTED' )
577	WRITE ( 90, '(''TRUE'')' )
578	CLOSE ( 90 )
579
580	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
581
582	!
583	!-- Continue only if the atmosphere model has created the port.
584	!-- There seems to be a bug in MPICH2 which causes hanging processes
585	!-- in case that execution of LOOKUP_NAME is continued too early
586	!-- (i.e. before the port has been created)
587	INQUIRE( FILE='COUPLING_PORT_OPENED', EXIST=found )
588	DO WHILE ( .NOT. found )
589	INQUIRE( FILE='COUPLING_PORT_OPENED', EXIST=found )
590	ENDDO
591
592	CALL MPI_LOOKUP_NAME( 'palm_coupler', MPI_INFO_NULL, port_name, ierr )
593
594	ENDIF
595
596	ENDIF
597
598	!
599	!-- In case of coupled runs, establish the connection between the atmosphere
600	!-- and the ocean model and define the intercommunicator (comm_inter)
601	CALL MPI_BARRIER( comm2d, ierr )
602	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
603
604	CALL MPI_COMM_ACCEPT( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD, &
605	comm_inter, ierr )
606	coupling_mode_remote = 'ocean_to_atmosphere'
607
608	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
609
610	CALL MPI_COMM_CONNECT( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD, &
611	comm_inter, ierr )
612	coupling_mode_remote = 'atmosphere_to_ocean'
613
614	ENDIF
615	#endif
616
617	!
618	!-- Determine the number of ghost point layers
619	IF ( ( scalar_advec == 'ws-scheme' .AND. .NOT. neutral ) .OR. &
620	scalar_advec == 'ws-scheme-mono' .OR. &
621	momentum_advec == 'ws-scheme' ) THEN
622	nbgp = 3
623	ELSE
624	nbgp = 1
625	ENDIF
626
627	!
628	!-- Create a new MPI derived datatype for the exchange of surface (xy) data,
629	!-- which is needed for coupled atmosphere-ocean runs.
630	!-- First, calculate number of grid points of an xy-plane.
631	ngp_xy = ( nxr - nxl + 1 + 2 * nbgp ) * ( nyn - nys + 1 + 2 * nbgp )
632	CALL MPI_TYPE_VECTOR( ngp_xy, 1, nzt-nzb+2, MPI_REAL, type_xy, ierr )
633	CALL MPI_TYPE_COMMIT( type_xy, ierr )
634
635	IF ( TRIM( coupling_mode ) /= 'uncoupled' ) THEN
636
637	!
638	!-- Pass the number of grid points of the atmosphere model to
639	!-- the ocean model and vice versa
640	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
641
642	nx_a = nx
643	ny_a = ny
644
645	IF ( myid == 0 ) THEN
646
647	CALL MPI_SEND( nx_a, 1, MPI_INTEGER, numprocs, 1, comm_inter, &
648	ierr )
649	CALL MPI_SEND( ny_a, 1, MPI_INTEGER, numprocs, 2, comm_inter, &
650	ierr )
651	CALL MPI_SEND( pdims, 2, MPI_INTEGER, numprocs, 3, comm_inter, &
652	ierr )
653	CALL MPI_RECV( nx_o, 1, MPI_INTEGER, numprocs, 4, comm_inter, &
654	status, ierr )
655	CALL MPI_RECV( ny_o, 1, MPI_INTEGER, numprocs, 5, comm_inter, &
656	status, ierr )
657	CALL MPI_RECV( pdims_remote, 2, MPI_INTEGER, numprocs, 6, &
658	comm_inter, status, ierr )
659	ENDIF
660
661	CALL MPI_BCAST( nx_o, 1, MPI_INTEGER, 0, comm2d, ierr )
662	CALL MPI_BCAST( ny_o, 1, MPI_INTEGER, 0, comm2d, ierr )
663	CALL MPI_BCAST( pdims_remote, 2, MPI_INTEGER, 0, comm2d, ierr )
664
665	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
666
667	nx_o = nx
668	ny_o = ny
669
670	IF ( myid == 0 ) THEN
671
672	CALL MPI_RECV( nx_a, 1, MPI_INTEGER, 0, 1, comm_inter, status, &
673	ierr )
674	CALL MPI_RECV( ny_a, 1, MPI_INTEGER, 0, 2, comm_inter, status, &
675	ierr )
676	CALL MPI_RECV( pdims_remote, 2, MPI_INTEGER, 0, 3, comm_inter, &
677	status, ierr )
678	CALL MPI_SEND( nx_o, 1, MPI_INTEGER, 0, 4, comm_inter, ierr )
679	CALL MPI_SEND( ny_o, 1, MPI_INTEGER, 0, 5, comm_inter, ierr )
680	CALL MPI_SEND( pdims, 2, MPI_INTEGER, 0, 6, comm_inter, ierr )
681	ENDIF
682
683	CALL MPI_BCAST( nx_a, 1, MPI_INTEGER, 0, comm2d, ierr)
684	CALL MPI_BCAST( ny_a, 1, MPI_INTEGER, 0, comm2d, ierr)
685	CALL MPI_BCAST( pdims_remote, 2, MPI_INTEGER, 0, comm2d, ierr)
686
687	ENDIF
688
689	ngp_a = ( nx_a+1 + 2 * nbgp ) * ( ny_a+1 + 2 * nbgp )
690	ngp_o = ( nx_o+1 + 2 * nbgp ) * ( ny_o+1 + 2 * nbgp )
691
692	!
693	!-- Determine if the horizontal grid and the number of PEs in ocean and
694	!-- atmosphere is same or not
695	IF ( nx_o == nx_a .AND. ny_o == ny_a .AND. &
696	pdims(1) == pdims_remote(1) .AND. pdims(2) == pdims_remote(2) ) &
697	THEN
698	coupling_topology = 0
699	ELSE
700	coupling_topology = 1
701	ENDIF
702
703	!
704	!-- Determine the target PEs for the exchange between ocean and
705	!-- atmosphere (comm2d)
706	IF ( coupling_topology == 0 ) THEN
707	!
708	!-- In case of identical topologies, every atmosphere PE has exactly one
709	!-- ocean PE counterpart and vice versa
710	IF ( TRIM( coupling_mode ) == 'atmosphere_to_ocean' ) THEN
711	target_id = myid + numprocs
712	ELSE
713	target_id = myid
714	ENDIF
715
716	ELSE
717	!
718	!-- In case of nonequivalent topology in ocean and atmosphere only for
719	!-- PE0 in ocean and PE0 in atmosphere a target_id is needed, since
720	!-- data echxchange between ocean and atmosphere will be done only
721	!-- between these PEs.
722	IF ( myid == 0 ) THEN
723
724	IF ( TRIM( coupling_mode ) == 'atmosphere_to_ocean' ) THEN
725	target_id = numprocs
726	ELSE
727	target_id = 0
728	ENDIF
729
730	ENDIF
731
732	ENDIF
733
734	ENDIF
735
736
737	#endif
738
739	#else
740
741	!
742	!-- Array bounds when running on a single PE (respectively a non-parallel
743	!-- machine)
744	nxl = 0
745	nxr = nx
746	nnx = nxr - nxl + 1
747	nys = 0
748	nyn = ny
749	nny = nyn - nys + 1
750	nzb = 0
751	nzt = nz
752	nnz = nz
753
754	ALLOCATE( hor_index_bounds(4,0:0) )
755	hor_index_bounds(1,0) = nxl
756	hor_index_bounds(2,0) = nxr
757	hor_index_bounds(3,0) = nys
758	hor_index_bounds(4,0) = nyn
759
760	!
761	!-- Array bounds for the pressure solver (in the parallel code, these bounds
762	!-- are the ones for the transposed arrays)
763	nys_x = nys
764	nyn_x = nyn
765	nzb_x = nzb + 1
766	nzt_x = nzt
767
768	nxl_y = nxl
769	nxr_y = nxr
770	nzb_y = nzb + 1
771	nzt_y = nzt
772
773	nxl_z = nxl
774	nxr_z = nxr
775	nys_z = nys
776	nyn_z = nyn
777
778	#endif
779
780	!
781	!-- Calculate number of grid levels necessary for the multigrid poisson solver
782	!-- as well as the gridpoint indices on each level
783	IF ( psolver(1:9) == 'multigrid' ) THEN
784
785	!
786	!-- First calculate number of possible grid levels for the subdomains
787	mg_levels_x = 1
788	mg_levels_y = 1
789	mg_levels_z = 1
790
791	i = nnx
792	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
793	i = i / 2
794	mg_levels_x = mg_levels_x + 1
795	ENDDO
796
797	j = nny
798	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
799	j = j / 2
800	mg_levels_y = mg_levels_y + 1
801	ENDDO
802
803	k = nz ! do not use nnz because it might be > nz due to transposition
804	! requirements
805	DO WHILE ( MOD( k, 2 ) == 0 .AND. k /= 2 )
806	k = k / 2
807	mg_levels_z = mg_levels_z + 1
808	ENDDO
809
810	maximum_grid_level = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
811
812	!
813	!-- Find out, if the total domain allows more levels. These additional
814	!-- levels are identically processed on all PEs.
815	IF ( numprocs > 1 .AND. mg_switch_to_pe0_level /= -1 ) THEN
816
817	IF ( mg_levels_z > MIN( mg_levels_x, mg_levels_y ) ) THEN
818
819	mg_switch_to_pe0_level_l = maximum_grid_level
820
821	mg_levels_x = 1
822	mg_levels_y = 1
823
824	i = nx+1
825	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
826	i = i / 2
827	mg_levels_x = mg_levels_x + 1
828	ENDDO
829
830	j = ny+1
831	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
832	j = j / 2
833	mg_levels_y = mg_levels_y + 1
834	ENDDO
835
836	maximum_grid_level_l = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
837
838	IF ( maximum_grid_level_l > mg_switch_to_pe0_level_l ) THEN
839	mg_switch_to_pe0_level_l = maximum_grid_level_l - &
840	mg_switch_to_pe0_level_l + 1
841	ELSE
842	mg_switch_to_pe0_level_l = 0
843	ENDIF
844
845	ELSE
846	mg_switch_to_pe0_level_l = 0
847	maximum_grid_level_l = maximum_grid_level
848
849	ENDIF
850
851	!
852	!-- Use switch level calculated above only if it is not pre-defined
853	!-- by user
854	IF ( mg_switch_to_pe0_level == 0 ) THEN
855	IF ( mg_switch_to_pe0_level_l /= 0 ) THEN
856	mg_switch_to_pe0_level = mg_switch_to_pe0_level_l
857	maximum_grid_level = maximum_grid_level_l
858	ENDIF
859
860	ELSE
861	!
862	!-- Check pre-defined value and reset to default, if neccessary
863	IF ( mg_switch_to_pe0_level < mg_switch_to_pe0_level_l .OR. &
864	mg_switch_to_pe0_level >= maximum_grid_level_l ) THEN
865	message_string = 'mg_switch_to_pe0_level ' // &
866	'out of range and reset to default (=0)'
867	CALL message( 'init_pegrid', 'PA0235', 0, 1, 0, 6, 0 )
868	mg_switch_to_pe0_level = 0
869	ELSE
870	!
871	!-- Use the largest number of possible levels anyway and recalculate
872	!-- the switch level to this largest number of possible values
873	maximum_grid_level = maximum_grid_level_l
874
875	ENDIF
876
877	ENDIF
878
879	ENDIF
880
881	ALLOCATE( grid_level_count(maximum_grid_level), &
882	nxl_mg(0:maximum_grid_level), nxr_mg(0:maximum_grid_level), &
883	nyn_mg(0:maximum_grid_level), nys_mg(0:maximum_grid_level), &
884	nzt_mg(0:maximum_grid_level) )
885
886	grid_level_count = 0
887	!
888	!-- Index zero required as dummy due to definition of arrays f2 and p2 in
889	!-- recursive subroutine next_mg_level
890	nxl_mg(0) = 0; nxr_mg(0) = 0; nyn_mg(0) = 0; nys_mg(0) = 0; nzt_mg(0) = 0
891
892	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzt_l = nzt
893
894	DO i = maximum_grid_level, 1 , -1
895
896	IF ( i == mg_switch_to_pe0_level ) THEN
897	#if defined( __parallel )
898	!
899	!-- Save the grid size of the subdomain at the switch level, because
900	!-- it is needed in poismg.
901	ind(1) = nxl_l; ind(2) = nxr_l
902	ind(3) = nys_l; ind(4) = nyn_l
903	ind(5) = nzt_l
904	ALLOCATE( ind_all(5*numprocs), mg_loc_ind(5,0:numprocs-1) )
905	CALL MPI_ALLGATHER( ind, 5, MPI_INTEGER, ind_all, 5, &
906	MPI_INTEGER, comm2d, ierr )
907	DO j = 0, numprocs-1
908	DO k = 1, 5
909	mg_loc_ind(k,j) = ind_all(k+j*5)
910	ENDDO
911	ENDDO
912	DEALLOCATE( ind_all )
913	!
914	!-- Calculate the grid size of the total domain
915	nxr_l = ( nxr_l-nxl_l+1 ) * pdims(1) - 1
916	nxl_l = 0
917	nyn_l = ( nyn_l-nys_l+1 ) * pdims(2) - 1
918	nys_l = 0
919	!
920	!-- The size of this gathered array must not be larger than the
921	!-- array tend, which is used in the multigrid scheme as a temporary
922	!-- array. Therefore the subdomain size of an PE is calculated and
923	!-- the size of the gathered grid. These values are used in
924	!-- routines pres and poismg
925	subdomain_size = ( nxr - nxl + 2 * nbgp + 1 ) * &
926	( nyn - nys + 2 * nbgp + 1 ) * ( nzt - nzb + 2 )
927	gathered_size = ( nxr_l - nxl_l + 3 ) * ( nyn_l - nys_l + 3 ) * &
928	( nzt_l - nzb + 2 )
929
930	#else
931	message_string = 'multigrid gather/scatter impossible ' // &
932	'in non parallel mode'
933	CALL message( 'init_pegrid', 'PA0237', 1, 2, 0, 6, 0 )
934	#endif
935	ENDIF
936
937	nxl_mg(i) = nxl_l
938	nxr_mg(i) = nxr_l
939	nys_mg(i) = nys_l
940	nyn_mg(i) = nyn_l
941	nzt_mg(i) = nzt_l
942
943	nxl_l = nxl_l / 2
944	nxr_l = nxr_l / 2
945	nys_l = nys_l / 2
946	nyn_l = nyn_l / 2
947	nzt_l = nzt_l / 2
948
949	ENDDO
950
951	!
952	!-- Temporary problem: Currently calculation of maxerror iin routine poismg crashes
953	!-- if grid data are collected on PE0 already on the finest grid level.
954	!-- To be solved later.
955	IF ( maximum_grid_level == mg_switch_to_pe0_level ) THEN
956	message_string = 'grid coarsening on subdomain level cannot be performed'
957	CALL message( 'poismg', 'PA0236', 1, 2, 0, 6, 0 )
958	ENDIF
959
960	ELSE
961
962	maximum_grid_level = 0
963
964	ENDIF
965
966	!
967	!-- Default level 0 tells exchange_horiz that all ghost planes have to be
968	!-- exchanged. grid_level is adjusted in poismg, where only one ghost plane
969	!-- is required.
970	grid_level = 0
971
972	#if defined( __parallel )
973	!
974	!-- Gridpoint number for the exchange of ghost points (y-line for 2D-arrays)
975	ngp_y = nyn - nys + 1 + 2 * nbgp
976
977	!
978	!-- Define new MPI derived datatypes for the exchange of ghost points in
979	!-- x- and y-direction for 2D-arrays (line)
980	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_REAL, type_x, &
981	ierr )
982	CALL MPI_TYPE_COMMIT( type_x, ierr )
983	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_INTEGER, &
984	type_x_int, ierr )
985	CALL MPI_TYPE_COMMIT( type_x_int, ierr )
986
987	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_REAL, type_y, ierr )
988	CALL MPI_TYPE_COMMIT( type_y, ierr )
989	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_INTEGER, type_y_int, ierr )
990	CALL MPI_TYPE_COMMIT( type_y_int, ierr )
991
992
993	!
994	!-- Calculate gridpoint numbers for the exchange of ghost points along x
995	!-- (yz-plane for 3D-arrays) and define MPI derived data type(s) for the
996	!-- exchange of ghost points in y-direction (xz-plane).
997	!-- Do these calculations for the model grid and (if necessary) also
998	!-- for the coarser grid levels used in the multigrid method
999	ALLOCATE ( ngp_xz(0:maximum_grid_level), ngp_yz(0:maximum_grid_level), &
1000	type_xz(0:maximum_grid_level), type_yz(0:maximum_grid_level) )
1001
1002	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzb_l = nzb; nzt_l = nzt
1003
1004	!
1005	!-- Discern between the model grid, which needs nbgp ghost points and
1006	!-- grid levels for the multigrid scheme. In the latter case only one
1007	!-- ghost point is necessary.
1008	!-- First definition of MPI-datatypes for exchange of ghost layers on normal
1009	!-- grid. The following loop is needed for data exchange in poismg.f90.
1010	!
1011	!-- Determine number of grid points of yz-layer for exchange
1012	ngp_yz(0) = (nzt - nzb + 2) * (nyn - nys + 1 + 2 * nbgp)
1013
1014	!
1015	!-- Define an MPI-datatype for the exchange of left/right boundaries.
1016	!-- Although data are contiguous in physical memory (which does not
1017	!-- necessarily require an MPI-derived datatype), the data exchange between
1018	!-- left and right PE's using the MPI-derived type is 10% faster than without.
1019	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2nbgp, nbgp(nzt-nzb+2), ngp_yz(0), &
1020	MPI_REAL, type_xz(0), ierr )
1021	CALL MPI_TYPE_COMMIT( type_xz(0), ierr )
1022
1023	CALL MPI_TYPE_VECTOR( nbgp, ngp_yz(0), ngp_yz(0), MPI_REAL, type_yz(0), &
1024	ierr )
1025	CALL MPI_TYPE_COMMIT( type_yz(0), ierr )
1026
1027	!
1028	!-- Definition of MPI-datatypes for multigrid method (coarser level grids)
1029	IF ( psolver(1:9) == 'multigrid' ) THEN
1030	!
1031	!-- Definition of MPI-datatyoe as above, but only 1 ghost level is used
1032	DO i = maximum_grid_level, 1 , -1
1033
1034	ngp_xz(i) = (nzt_l - nzb_l + 2) * (nxr_l - nxl_l + 3)
1035	ngp_yz(i) = (nzt_l - nzb_l + 2) * (nyn_l - nys_l + 3)
1036
1037	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+3, nzt_l-nzb_l+2, ngp_yz(i), &
1038	MPI_REAL, type_xz(i), ierr )
1039	CALL MPI_TYPE_COMMIT( type_xz(i), ierr )
1040
1041	CALL MPI_TYPE_VECTOR( 1, ngp_yz(i), ngp_yz(i), MPI_REAL, type_yz(i), &
1042	ierr )
1043	CALL MPI_TYPE_COMMIT( type_yz(i), ierr )
1044
1045	nxl_l = nxl_l / 2
1046	nxr_l = nxr_l / 2
1047	nys_l = nys_l / 2
1048	nyn_l = nyn_l / 2
1049	nzt_l = nzt_l / 2
1050
1051	ENDDO
1052
1053	ENDIF
1054	!
1055	!-- Define data types for exchange of 3D Integer arrays.
1056	ngp_yz_int = (nzt - nzb + 2) * (nyn - nys + 1 + 2 * nbgp)
1057
1058	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2nbgp, nbgp(nzt-nzb+2), ngp_yz_int, &
1059	MPI_INTEGER, type_xz_int, ierr )
1060	CALL MPI_TYPE_COMMIT( type_xz_int, ierr )
1061
1062	CALL MPI_TYPE_VECTOR( nbgp, ngp_yz_int, ngp_yz_int, MPI_INTEGER, type_yz_int, &
1063	ierr )
1064	CALL MPI_TYPE_COMMIT( type_yz_int, ierr )
1065
1066	#endif
1067
1068	#if defined( __parallel )
1069	!
1070	!-- Setting of flags for inflow/outflow/nesting conditions in case of non-cyclic
1071	!-- horizontal boundary conditions.
1072	IF ( pleft == MPI_PROC_NULL ) THEN
1073	IF ( bc_lr == 'dirichlet/radiation' ) THEN
1074	inflow_l = .TRUE.
1075	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
1076	outflow_l = .TRUE.
1077	ELSEIF ( bc_lr == 'nested' ) THEN
1078	nest_bound_l = .TRUE.
1079	ENDIF
1080	ENDIF
1081
1082	IF ( pright == MPI_PROC_NULL ) THEN
1083	IF ( bc_lr == 'dirichlet/radiation' ) THEN
1084	outflow_r = .TRUE.
1085	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
1086	inflow_r = .TRUE.
1087	ELSEIF ( bc_lr == 'nested' ) THEN
1088	nest_bound_r = .TRUE.
1089	ENDIF
1090	ENDIF
1091
1092	IF ( psouth == MPI_PROC_NULL ) THEN
1093	IF ( bc_ns == 'dirichlet/radiation' ) THEN
1094	outflow_s = .TRUE.
1095	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
1096	inflow_s = .TRUE.
1097	ELSEIF ( bc_ns == 'nested' ) THEN
1098	nest_bound_s = .TRUE.
1099	ENDIF
1100	ENDIF
1101
1102	IF ( pnorth == MPI_PROC_NULL ) THEN
1103	IF ( bc_ns == 'dirichlet/radiation' ) THEN
1104	inflow_n = .TRUE.
1105	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
1106	outflow_n = .TRUE.
1107	ELSEIF ( bc_ns == 'nested' ) THEN
1108	nest_bound_n = .TRUE.
1109	ENDIF
1110	ENDIF
1111
1112	!
1113	!-- Broadcast the id of the inflow PE
1114	IF ( inflow_l ) THEN
1115	id_inflow_l = myidx
1116	ELSE
1117	id_inflow_l = 0
1118	ENDIF
1119	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1120	CALL MPI_ALLREDUCE( id_inflow_l, id_inflow, 1, MPI_INTEGER, MPI_SUM, &
1121	comm1dx, ierr )
1122
1123	!
1124	!-- Broadcast the id of the recycling plane
1125	!-- WARNING: needs to be adjusted in case of inflows other than from left side!
1126	IF ( NINT( recycling_width / dx ) >= nxl .AND. &
1127	NINT( recycling_width / dx ) <= nxr ) THEN
1128	id_recycling_l = myidx
1129	ELSE
1130	id_recycling_l = 0
1131	ENDIF
1132	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1133	CALL MPI_ALLREDUCE( id_recycling_l, id_recycling, 1, MPI_INTEGER, MPI_SUM, &
1134	comm1dx, ierr )
1135
1136	CALL location_message( 'finished', .TRUE. )
1137
1138	#else
1139	IF ( bc_lr == 'dirichlet/radiation' ) THEN
1140	inflow_l = .TRUE.
1141	outflow_r = .TRUE.
1142	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
1143	outflow_l = .TRUE.
1144	inflow_r = .TRUE.
1145	ENDIF
1146
1147	IF ( bc_ns == 'dirichlet/radiation' ) THEN
1148	inflow_n = .TRUE.
1149	outflow_s = .TRUE.
1150	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
1151	outflow_n = .TRUE.
1152	inflow_s = .TRUE.
1153	ENDIF
1154	#endif
1155
1156	!
1157	!-- At the inflow or outflow, u or v, respectively, have to be calculated for
1158	!-- one more grid point.
1159	IF ( inflow_l .OR. outflow_l .OR. nest_bound_l ) THEN
1160	nxlu = nxl + 1
1161	ELSE
1162	nxlu = nxl
1163	ENDIF
1164	IF ( inflow_s .OR. outflow_s .OR. nest_bound_s ) THEN
1165	nysv = nys + 1
1166	ELSE
1167	nysv = nys
1168	ENDIF
1169
1170	!
1171	!-- Allocate wall flag arrays used in the multigrid solver
1172	IF ( psolver(1:9) == 'multigrid' ) THEN
1173
1174	DO i = maximum_grid_level, 1, -1
1175
1176	SELECT CASE ( i )
1177
1178	CASE ( 1 )
1179	ALLOCATE( wall_flags_1(nzb:nzt_mg(i)+1, &
1180	nys_mg(i)-1:nyn_mg(i)+1, &
1181	nxl_mg(i)-1:nxr_mg(i)+1) )
1182
1183	CASE ( 2 )
1184	ALLOCATE( wall_flags_2(nzb:nzt_mg(i)+1, &
1185	nys_mg(i)-1:nyn_mg(i)+1, &
1186	nxl_mg(i)-1:nxr_mg(i)+1) )
1187
1188	CASE ( 3 )
1189	ALLOCATE( wall_flags_3(nzb:nzt_mg(i)+1, &
1190	nys_mg(i)-1:nyn_mg(i)+1, &
1191	nxl_mg(i)-1:nxr_mg(i)+1) )
1192
1193	CASE ( 4 )
1194	ALLOCATE( wall_flags_4(nzb:nzt_mg(i)+1, &
1195	nys_mg(i)-1:nyn_mg(i)+1, &
1196	nxl_mg(i)-1:nxr_mg(i)+1) )
1197
1198	CASE ( 5 )
1199	ALLOCATE( wall_flags_5(nzb:nzt_mg(i)+1, &
1200	nys_mg(i)-1:nyn_mg(i)+1, &
1201	nxl_mg(i)-1:nxr_mg(i)+1) )
1202
1203	CASE ( 6 )
1204	ALLOCATE( wall_flags_6(nzb:nzt_mg(i)+1, &
1205	nys_mg(i)-1:nyn_mg(i)+1, &
1206	nxl_mg(i)-1:nxr_mg(i)+1) )
1207
1208	CASE ( 7 )
1209	ALLOCATE( wall_flags_7(nzb:nzt_mg(i)+1, &
1210	nys_mg(i)-1:nyn_mg(i)+1, &
1211	nxl_mg(i)-1:nxr_mg(i)+1) )
1212
1213	CASE ( 8 )
1214	ALLOCATE( wall_flags_8(nzb:nzt_mg(i)+1, &
1215	nys_mg(i)-1:nyn_mg(i)+1, &
1216	nxl_mg(i)-1:nxr_mg(i)+1) )
1217
1218	CASE ( 9 )
1219	ALLOCATE( wall_flags_9(nzb:nzt_mg(i)+1, &
1220	nys_mg(i)-1:nyn_mg(i)+1, &
1221	nxl_mg(i)-1:nxr_mg(i)+1) )
1222
1223	CASE ( 10 )
1224	ALLOCATE( wall_flags_10(nzb:nzt_mg(i)+1, &
1225	nys_mg(i)-1:nyn_mg(i)+1, &
1226	nxl_mg(i)-1:nxr_mg(i)+1) )
1227
1228	CASE DEFAULT
1229	message_string = 'more than 10 multigrid levels'
1230	CALL message( 'init_pegrid', 'PA0238', 1, 2, 0, 6, 0 )
1231
1232	END SELECT
1233
1234	ENDDO
1235
1236	ENDIF
1237
1238	!
1239	!-- Calculate the number of groups into which parallel I/O is split.
1240	!-- The default for files which are opened by all PEs (or where each
1241	!-- PE opens his own independent file) is, that all PEs are doing input/output
1242	!-- in parallel at the same time. This might cause performance or even more
1243	!-- severe problems depending on the configuration of the underlying file
1244	!-- system.
1245	!-- First, set the default:
1246	IF ( maximum_parallel_io_streams == -1 .OR. &
1247	maximum_parallel_io_streams > numprocs ) THEN
1248	maximum_parallel_io_streams = numprocs
1249	ENDIF
1250
1251	!
1252	!-- Now calculate the number of io_blocks and the io_group to which the
1253	!-- respective PE belongs. I/O of the groups is done in serial, but in parallel
1254	!-- for all PEs belonging to the same group. A preliminary setting with myid
1255	!-- based on MPI_COMM_WORLD has been done in parin.
1256	io_blocks = numprocs / maximum_parallel_io_streams
1257	io_group = MOD( myid+1, io_blocks )
1258
1259
1260	END SUBROUTINE init_pegrid

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

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |