Home

Context Navigation

source: palm/trunk/SOURCE/init_pegrid.f90 @ 161

Last change on this file since 161 was 151, checked in by raasch, 17 years ago
preliminary update for the turbulence recycling method
Property svn:keywords set to `Id`
File size: 32.3 KB

Line
1	SUBROUTINE init_pegrid
2
3	!------------------------------------------------------------------------------!
4	! Actual revisions:
5	! -----------------
6	! Collect on PE0 horizontal index bounds from all other PEs,
7	! broadcast the id of the inflow PE (using the respective communicator)
8	! TEST OUTPUT (TO BE REMOVED) logging mpi2 ierr values
9	!
10	! Former revisions:
11	! -----------------
12	! $Id: init_pegrid.f90 151 2008-03-07 13:42:18Z letzel $
13	!
14	! 114 2007-10-10 00:03:15Z raasch
15	! Allocation of wall flag arrays for multigrid solver
16	!
17	! 108 2007-08-24 15:10:38Z letzel
18	! Intercommunicator (comm_inter) and derived data type (type_xy) for
19	! coupled model runs created, assign coupling_mode_remote,
20	! indices nxlu and nysv are calculated (needed for non-cyclic boundary
21	! conditions)
22	!
23	! 82 2007-04-16 15:40:52Z raasch
24	! Cpp-directive lcmuk changed to intel_openmp_bug, setting of host on lcmuk by
25	! cpp-directive removed
26	!
27	! 75 2007-03-22 09:54:05Z raasch
28	! uxrp, vynp eliminated,
29	! dirichlet/neumann changed to dirichlet/radiation, etc.,
30	! poisfft_init is only called if fft-solver is switched on
31	!
32	! RCS Log replace by Id keyword, revision history cleaned up
33	!
34	! Revision 1.28 2006/04/26 13:23:32 raasch
35	! lcmuk does not understand the !$ comment so a cpp-directive is required
36	!
37	! Revision 1.1 1997/07/24 11:15:09 raasch
38	! Initial revision
39	!
40	!
41	! Description:
42	! ------------
43	! Determination of the virtual processor topology (if not prescribed by the
44	! user)and computation of the grid point number and array bounds of the local
45	! domains.
46	!------------------------------------------------------------------------------!
47
48	USE control_parameters
49	USE fft_xy
50	USE indices
51	USE pegrid
52	USE poisfft_mod
53	USE poisfft_hybrid_mod
54	USE statistics
55	USE transpose_indices
56
57
58	IMPLICIT NONE
59
60	INTEGER :: gathered_size, i, id_inflow_l, ind(5), j, k, &
61	maximum_grid_level_l, mg_switch_to_pe0_level_l, mg_levels_x, &
62	mg_levels_y, mg_levels_z, nnx_y, nnx_z, nny_x, nny_z, nnz_x, &
63	nnz_y, numproc_sqr, nx_total, nxl_l, nxr_l, nyn_l, nys_l, &
64	nzb_l, nzt_l, omp_get_num_threads, subdomain_size
65
66	INTEGER, DIMENSION(:), ALLOCATABLE :: ind_all, nxlf, nxrf, nynf, nysf
67
68	LOGICAL :: found
69
70	!
71	!-- Get the number of OpenMP threads
72	!$OMP PARALLEL
73	#if defined( __intel_openmp_bug )
74	threads_per_task = omp_get_num_threads()
75	#else
76	!$ threads_per_task = omp_get_num_threads()
77	#endif
78	!$OMP END PARALLEL
79
80
81	#if defined( __parallel )
82	!
83	!-- Determine the processor topology or check it, if prescribed by the user
84	IF ( npex == -1 .AND. npey == -1 ) THEN
85
86	!
87	!-- Automatic determination of the topology
88	!-- The default on SMP- and cluster-hosts is a 1d-decomposition along x
89	IF ( host(1:3) == 'ibm' .OR. host(1:3) == 'nec' .OR. &
90	host(1:2) == 'lc' .OR. host(1:3) == 'dec' ) THEN
91
92	pdims(1) = numprocs
93	pdims(2) = 1
94
95	ELSE
96
97	numproc_sqr = SQRT( REAL( numprocs ) )
98	pdims(1) = MAX( numproc_sqr , 1 )
99	DO WHILE ( MOD( numprocs , pdims(1) ) /= 0 )
100	pdims(1) = pdims(1) - 1
101	ENDDO
102	pdims(2) = numprocs / pdims(1)
103
104	ENDIF
105
106	ELSEIF ( npex /= -1 .AND. npey /= -1 ) THEN
107
108	!
109	!-- Prescribed by user. Number of processors on the prescribed topology
110	!-- must be equal to the number of PEs available to the job
111	IF ( ( npex * npey ) /= numprocs ) THEN
112	PRINT*, '+++ init_pegrid:'
113	PRINT, ' number of PEs of the prescribed topology (', npexnpey, &
114	') does not match the number of PEs available to the ', &
115	'job (', numprocs, ')'
116	CALL local_stop
117	ENDIF
118	pdims(1) = npex
119	pdims(2) = npey
120
121	ELSE
122	!
123	!-- If the processor topology is prescribed by the user, the number of
124	!-- PEs must be given in both directions
125	PRINT*, '+++ init_pegrid:'
126	PRINT*, ' if the processor topology is prescribed by the user, ', &
127	'both values of "npex" and "npey" must be given in the ', &
128	'NAMELIST-parameter file'
129	CALL local_stop
130
131	ENDIF
132
133	!
134	!-- The hybrid solver can only be used in case of a 1d-decomposition along x
135	IF ( pdims(2) /= 1 .AND. psolver == 'poisfft_hybrid' ) THEN
136	IF ( myid == 0 ) THEN
137	PRINT, '** init_pegrid: psolver = "poisfft_hybrid" can only be'
138	PRINT*, ' used in case of a 1d-decomposition along x'
139	ENDIF
140	ENDIF
141
142	!
143	!-- If necessary, set horizontal boundary conditions to non-cyclic
144	IF ( bc_lr /= 'cyclic' ) cyclic(1) = .FALSE.
145	IF ( bc_ns /= 'cyclic' ) cyclic(2) = .FALSE.
146
147	!
148	!-- Create the virtual processor grid
149	CALL MPI_CART_CREATE( comm_palm, ndim, pdims, cyclic, reorder, &
150	comm2d, ierr )
151	CALL MPI_COMM_RANK( comm2d, myid, ierr )
152	WRITE (myid_char,'(''_'',I4.4)') myid
153
154	CALL MPI_CART_COORDS( comm2d, myid, ndim, pcoord, ierr )
155	CALL MPI_CART_SHIFT( comm2d, 0, 1, pleft, pright, ierr )
156	CALL MPI_CART_SHIFT( comm2d, 1, 1, psouth, pnorth, ierr )
157
158	!
159	!-- Determine sub-topologies for transpositions
160	!-- Transposition from z to x:
161	remain_dims(1) = .TRUE.
162	remain_dims(2) = .FALSE.
163	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dx, ierr )
164	CALL MPI_COMM_RANK( comm1dx, myidx, ierr )
165	!
166	!-- Transposition from x to y
167	remain_dims(1) = .FALSE.
168	remain_dims(2) = .TRUE.
169	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dy, ierr )
170	CALL MPI_COMM_RANK( comm1dy, myidy, ierr )
171
172
173	!
174	!-- Find a grid (used for array d) which will match the transposition demands
175	IF ( grid_matching == 'strict' ) THEN
176
177	nxa = nx; nya = ny; nza = nz
178
179	ELSE
180
181	found = .FALSE.
182	xn: DO nxa = nx, 2*nx
183	!
184	!-- Meet conditions for nx
185	IF ( MOD( nxa+1, pdims(1) ) /= 0 .OR. &
186	MOD( nxa+1, pdims(2) ) /= 0 ) CYCLE xn
187
188	yn: DO nya = ny, 2*ny
189	!
190	!-- Meet conditions for ny
191	IF ( MOD( nya+1, pdims(2) ) /= 0 .OR. &
192	MOD( nya+1, pdims(1) ) /= 0 ) CYCLE yn
193
194
195	zn: DO nza = nz, 2*nz
196	!
197	!-- Meet conditions for nz
198	IF ( ( MOD( nza, pdims(1) ) /= 0 .AND. pdims(1) /= 1 .AND. &
199	pdims(2) /= 1 ) .OR. &
200	( MOD( nza, pdims(2) ) /= 0 .AND. dt_dosp /= 9999999.9 &
201	) ) THEN
202	CYCLE zn
203	ELSE
204	found = .TRUE.
205	EXIT xn
206	ENDIF
207
208	ENDDO zn
209
210	ENDDO yn
211
212	ENDDO xn
213
214	IF ( .NOT. found ) THEN
215	IF ( myid == 0 ) THEN
216	PRINT*,'+++ init_pegrid: no matching grid for transpositions found'
217	ENDIF
218	CALL local_stop
219	ENDIF
220
221	ENDIF
222
223	!
224	!-- Calculate array bounds in x-direction for every PE.
225	!-- The last PE along x may get less grid points than the others
226	ALLOCATE( nxlf(0:pdims(1)-1), nxrf(0:pdims(1)-1), nynf(0:pdims(2)-1), &
227	nysf(0:pdims(2)-1), nnx_pe(0:pdims(1)-1), nny_pe(0:pdims(2)-1) )
228
229	IF ( MOD( nxa+1 , pdims(1) ) /= 0 ) THEN
230	IF ( myid == 0 ) THEN
231	PRINT*,'+++ x-direction: gridpoint number (',nx+1,') is not an'
232	PRINT*,' integral divisor of the number of proces', &
233	&'sors (', pdims(1),')'
234	ENDIF
235	CALL local_stop
236	ELSE
237	nnx = ( nxa + 1 ) / pdims(1)
238	IF ( nnx*pdims(1) - ( nx + 1) > nnx ) THEN
239	IF ( myid == 0 ) THEN
240	PRINT*,'+++ x-direction: nx does not match the requirements ', &
241	'given by the number of PEs'
242	PRINT*,' used'
243	PRINT,' please use nx = ', nx - ( pdims(1) - ( nnxpdims(1) &
244	- ( nx + 1 ) ) ), ' instead of nx =', nx
245	ENDIF
246	CALL local_stop
247	ENDIF
248	ENDIF
249
250	!
251	!-- Left and right array bounds, number of gridpoints
252	DO i = 0, pdims(1)-1
253	nxlf(i) = i * nnx
254	nxrf(i) = ( i + 1 ) * nnx - 1
255	nnx_pe(i) = MIN( nx, nxrf(i) ) - nxlf(i) + 1
256	ENDDO
257
258	!
259	!-- Calculate array bounds in y-direction for every PE.
260	IF ( MOD( nya+1 , pdims(2) ) /= 0 ) THEN
261	IF ( myid == 0 ) THEN
262	PRINT*,'+++ y-direction: gridpoint number (',ny+1,') is not an'
263	PRINT*,' integral divisor of the number of proces', &
264	&'sors (', pdims(2),')'
265	ENDIF
266	CALL local_stop
267	ELSE
268	nny = ( nya + 1 ) / pdims(2)
269	IF ( nny*pdims(2) - ( ny + 1) > nny ) THEN
270	IF ( myid == 0 ) THEN
271	PRINT*,'+++ x-direction: nx does not match the requirements ', &
272	'given by the number of PEs'
273	PRINT*,' used'
274	PRINT,' please use nx = ', nx - ( pdims(1) - ( nnxpdims(1) &
275	- ( nx + 1 ) ) ), ' instead of nx =', nx
276	ENDIF
277	CALL local_stop
278	ENDIF
279	ENDIF
280
281	!
282	!-- South and north array bounds
283	DO j = 0, pdims(2)-1
284	nysf(j) = j * nny
285	nynf(j) = ( j + 1 ) * nny - 1
286	nny_pe(j) = MIN( ny, nynf(j) ) - nysf(j) + 1
287	ENDDO
288
289	!
290	!-- Local array bounds of the respective PEs
291	nxl = nxlf(pcoord(1))
292	nxra = nxrf(pcoord(1))
293	nxr = MIN( nx, nxra )
294	nys = nysf(pcoord(2))
295	nyna = nynf(pcoord(2))
296	nyn = MIN( ny, nyna )
297	nzb = 0
298	nzta = nza
299	nzt = MIN( nz, nzta )
300	nnz = nza
301
302	!
303	!-- Calculate array bounds and gridpoint numbers for the transposed arrays
304	!-- (needed in the pressure solver)
305	!-- For the transposed arrays, cyclic boundaries as well as top and bottom
306	!-- boundaries are omitted, because they are obstructive to the transposition
307
308	!
309	!-- 1. transposition z --> x
310	!-- This transposition is not neccessary in case of a 1d-decomposition along x,
311	!-- except that the uptream-spline method is switched on
312	IF ( pdims(2) /= 1 .OR. momentum_advec == 'ups-scheme' .OR. &
313	scalar_advec == 'ups-scheme' ) THEN
314
315	IF ( pdims(2) == 1 .AND. ( momentum_advec == 'ups-scheme' .OR. &
316	scalar_advec == 'ups-scheme' ) ) THEN
317	IF ( myid == 0 ) THEN
318	PRINT*,'+++ WARNING: init_pegrid: 1d-decomposition along x ', &
319	&'chosen but nz restrictions may occur'
320	PRINT*,' since ups-scheme is activated'
321	ENDIF
322	ENDIF
323	nys_x = nys
324	nyn_xa = nyna
325	nyn_x = nyn
326	nny_x = nny
327	IF ( MOD( nza , pdims(1) ) /= 0 ) THEN
328	IF ( myid == 0 ) THEN
329	PRINT*,'+++ transposition z --> x:'
330	PRINT*,' nz=',nz,' is not an integral divisior of pdims(1)=', &
331	&pdims(1)
332	ENDIF
333	CALL local_stop
334	ENDIF
335	nnz_x = nza / pdims(1)
336	nzb_x = 1 + myidx * nnz_x
337	nzt_xa = ( myidx + 1 ) * nnz_x
338	nzt_x = MIN( nzt, nzt_xa )
339
340	sendrecvcount_zx = nnx * nny * nnz_x
341
342	ENDIF
343
344	!
345	!-- 2. transposition x --> y
346	nnz_y = nnz_x
347	nzb_y = nzb_x
348	nzt_ya = nzt_xa
349	nzt_y = nzt_x
350	IF ( MOD( nxa+1 , pdims(2) ) /= 0 ) THEN
351	IF ( myid == 0 ) THEN
352	PRINT*,'+++ transposition x --> y:'
353	PRINT*,' nx+1=',nx+1,' is not an integral divisor of ',&
354	&'pdims(2)=',pdims(2)
355	ENDIF
356	CALL local_stop
357	ENDIF
358	nnx_y = (nxa+1) / pdims(2)
359	nxl_y = myidy * nnx_y
360	nxr_ya = ( myidy + 1 ) * nnx_y - 1
361	nxr_y = MIN( nx, nxr_ya )
362
363	sendrecvcount_xy = nnx_y * nny_x * nnz_y
364
365	!
366	!-- 3. transposition y --> z (ELSE: x --> y in case of 1D-decomposition
367	!-- along x)
368	IF ( pdims(2) /= 1 .OR. momentum_advec == 'ups-scheme' .OR. &
369	scalar_advec == 'ups-scheme' ) THEN
370	!
371	!-- y --> z
372	!-- This transposition is not neccessary in case of a 1d-decomposition
373	!-- along x, except that the uptream-spline method is switched on
374	nnx_z = nnx_y
375	nxl_z = nxl_y
376	nxr_za = nxr_ya
377	nxr_z = nxr_y
378	IF ( MOD( nya+1 , pdims(1) ) /= 0 ) THEN
379	IF ( myid == 0 ) THEN
380	PRINT*,'+++ Transposition y --> z:'
381	PRINT*,' ny+1=',ny+1,' is not an integral divisor of ',&
382	&'pdims(1)=',pdims(1)
383	ENDIF
384	CALL local_stop
385	ENDIF
386	nny_z = (nya+1) / pdims(1)
387	nys_z = myidx * nny_z
388	nyn_za = ( myidx + 1 ) * nny_z - 1
389	nyn_z = MIN( ny, nyn_za )
390
391	sendrecvcount_yz = nnx_y * nny_z * nnz_y
392
393	ELSE
394	!
395	!-- x --> y. This condition must be fulfilled for a 1D-decomposition along x
396	IF ( MOD( nya+1 , pdims(1) ) /= 0 ) THEN
397	IF ( myid == 0 ) THEN
398	PRINT*,'+++ Transposition x --> y:'
399	PRINT*,' ny+1=',ny+1,' is not an integral divisor of ',&
400	&'pdims(1)=',pdims(1)
401	ENDIF
402	CALL local_stop
403	ENDIF
404
405	ENDIF
406
407	!
408	!-- Indices for direct transpositions z --> y (used for calculating spectra)
409	IF ( dt_dosp /= 9999999.9 ) THEN
410	IF ( MOD( nza, pdims(2) ) /= 0 ) THEN
411	IF ( myid == 0 ) THEN
412	PRINT*,'+++ Direct transposition z --> y (needed for spectra):'
413	PRINT*,' nz=',nz,' is not an integral divisor of ',&
414	&'pdims(2)=',pdims(2)
415	ENDIF
416	CALL local_stop
417	ELSE
418	nxl_yd = nxl
419	nxr_yda = nxra
420	nxr_yd = nxr
421	nzb_yd = 1 + myidy * ( nza / pdims(2) )
422	nzt_yda = ( myidy + 1 ) * ( nza / pdims(2) )
423	nzt_yd = MIN( nzt, nzt_yda )
424
425	sendrecvcount_zyd = nnx * nny * ( nza / pdims(2) )
426	ENDIF
427	ENDIF
428
429	!
430	!-- Indices for direct transpositions y --> x (they are only possible in case
431	!-- of a 1d-decomposition along x)
432	IF ( pdims(2) == 1 ) THEN
433	nny_x = nny / pdims(1)
434	nys_x = myid * nny_x
435	nyn_xa = ( myid + 1 ) * nny_x - 1
436	nyn_x = MIN( ny, nyn_xa )
437	nzb_x = 1
438	nzt_xa = nza
439	nzt_x = nz
440	sendrecvcount_xy = nnx * nny_x * nza
441	ENDIF
442
443	!
444	!-- Indices for direct transpositions x --> y (they are only possible in case
445	!-- of a 1d-decomposition along y)
446	IF ( pdims(1) == 1 ) THEN
447	nnx_y = nnx / pdims(2)
448	nxl_y = myid * nnx_y
449	nxr_ya = ( myid + 1 ) * nnx_y - 1
450	nxr_y = MIN( nx, nxr_ya )
451	nzb_y = 1
452	nzt_ya = nza
453	nzt_y = nz
454	sendrecvcount_xy = nnx_y * nny * nza
455	ENDIF
456
457	!
458	!-- Arrays for storing the array bounds are needed any more
459	DEALLOCATE( nxlf , nxrf , nynf , nysf )
460
461	!
462	!-- Collect index bounds from other PEs (to be written to restart file later)
463	ALLOCATE( hor_index_bounds(4,0:numprocs-1) )
464
465	IF ( myid == 0 ) THEN
466
467	hor_index_bounds(1,0) = nxl
468	hor_index_bounds(2,0) = nxr
469	hor_index_bounds(3,0) = nys
470	hor_index_bounds(4,0) = nyn
471
472	!
473	!-- Receive data from all other PEs
474	DO i = 1, numprocs-1
475	CALL MPI_RECV( ibuf, 4, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, &
476	ierr )
477	hor_index_bounds(:,i) = ibuf(1:4)
478	ENDDO
479
480	ELSE
481	!
482	!-- Send index bounds to PE0
483	ibuf(1) = nxl
484	ibuf(2) = nxr
485	ibuf(3) = nys
486	ibuf(4) = nyn
487	CALL MPI_SEND( ibuf, 4, MPI_INTEGER, 0, myid, comm2d, ierr )
488
489	ENDIF
490
491	#if defined( __print )
492	!
493	!-- Control output
494	IF ( myid == 0 ) THEN
495	PRINT, ' processor topology *'
496	PRINT*, ' '
497	PRINT*, 'myid pcoord left right south north idx idy nxl: nxr',&
498	&' nys: nyn'
499	PRINT*, '------------------------------------------------------------',&
500	&'-----------'
501	WRITE (*,1000) 0, pcoord(1), pcoord(2), pleft, pright, psouth, pnorth, &
502	myidx, myidy, nxl, nxr, nys, nyn
503	1000 FORMAT (I4,2X,'(',I3,',',I3,')',3X,I4,2X,I4,3X,I4,2X,I4,2X,I3,1X,I3, &
504	2(2X,I4,':',I4))
505
506	!
507	!-- Receive data from the other PEs
508	DO i = 1,numprocs-1
509	CALL MPI_RECV( ibuf, 12, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, &
510	ierr )
511	WRITE (*,1000) i, ( ibuf(j) , j = 1,12 )
512	ENDDO
513	ELSE
514
515	!
516	!-- Send data to PE0
517	ibuf(1) = pcoord(1); ibuf(2) = pcoord(2); ibuf(3) = pleft
518	ibuf(4) = pright; ibuf(5) = psouth; ibuf(6) = pnorth; ibuf(7) = myidx
519	ibuf(8) = myidy; ibuf(9) = nxl; ibuf(10) = nxr; ibuf(11) = nys
520	ibuf(12) = nyn
521	CALL MPI_SEND( ibuf, 12, MPI_INTEGER, 0, myid, comm2d, ierr )
522	ENDIF
523	#endif
524
525	#if defined( __mpi2 )
526	!
527	!-- In case of coupled runs, get the port name on PE0 of the atmosphere model
528	!-- and pass it to PE0 of the ocean model
529	IF ( myid == 0 ) THEN
530
531	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
532
533	CALL MPI_OPEN_PORT( MPI_INFO_NULL, port_name, ierr )
534	!
535	!-- TEST OUTPUT (TO BE REMOVED)
536	WRITE(9,*) TRIM( coupling_mode ), &
537	', ierr after MPI_OPEN_PORT: ', ierr
538	CALL LOCAL_FLUSH( 9 )
539
540	CALL MPI_PUBLISH_NAME( 'palm_coupler', MPI_INFO_NULL, port_name, &
541	ierr )
542	!
543	!-- TEST OUTPUT (TO BE REMOVED)
544	WRITE(9,*) TRIM( coupling_mode ), &
545	', ierr after MPI_PUBLISH_NAME: ', ierr
546	CALL LOCAL_FLUSH( 9 )
547
548	!
549	!-- Write a flag file for the ocean model and the other atmosphere
550	!-- processes.
551	!-- There seems to be a bug in MPICH2 which causes hanging processes
552	!-- in case that execution of LOOKUP_NAME is continued too early
553	!-- (i.e. before the port has been created)
554	OPEN( 90, FILE='COUPLING_PORT_OPENED', FORM='FORMATTED' )
555	WRITE ( 90, '(''TRUE'')' )
556	CLOSE ( 90 )
557
558	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
559
560	!
561	!-- Continue only if the atmosphere model has created the port.
562	!-- There seems to be a bug in MPICH2 which causes hanging processes
563	!-- in case that execution of LOOKUP_NAME is continued too early
564	!-- (i.e. before the port has been created)
565	INQUIRE( FILE='COUPLING_PORT_OPENED', EXIST=found )
566	DO WHILE ( .NOT. found )
567	INQUIRE( FILE='COUPLING_PORT_OPENED', EXIST=found )
568	ENDDO
569
570	CALL MPI_LOOKUP_NAME( 'palm_coupler', MPI_INFO_NULL, port_name, ierr )
571	!
572	!-- TEST OUTPUT (TO BE REMOVED)
573	WRITE(9,*) TRIM( coupling_mode ), &
574	', ierr after MPI_LOOKUP_NAME: ', ierr
575	CALL LOCAL_FLUSH( 9 )
576
577
578	ENDIF
579
580	ENDIF
581
582	!
583	!-- In case of coupled runs, establish the connection between the atmosphere
584	!-- and the ocean model and define the intercommunicator (comm_inter)
585	CALL MPI_BARRIER( comm2d, ierr )
586	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
587
588	print*, '... before COMM_ACCEPT'
589	CALL MPI_COMM_ACCEPT( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD, &
590	comm_inter, ierr )
591	print*, '--- ierr = ', ierr
592	print*, '--- comm_inter atmosphere = ', comm_inter
593
594	coupling_mode_remote = 'ocean_to_atmosphere'
595
596	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
597
598	IF ( myid == 0 ) PRINT, '** read: ', port_name, ' ierr = ', ierr
599	print*, '... before COMM_CONNECT'
600	CALL MPI_COMM_CONNECT( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD, &
601	comm_inter, ierr )
602	print*, '--- ierr = ', ierr
603	print*, '--- comm_inter ocean = ', comm_inter
604
605	coupling_mode_remote = 'atmosphere_to_ocean'
606
607	ENDIF
608
609	!
610	!-- In case of coupled runs, create a new MPI derived datatype for the
611	!-- exchange of surface (xy) data .
612	!-- Gridpoint number for the exchange of ghost points (xy-plane)
613	ngp_xy = ( nxr - nxl + 3 ) * ( nyn - nys + 3 )
614
615	!
616	!-- Define a new MPI derived datatype for the exchange of ghost points in
617	!-- y-direction for 2D-arrays (line)
618	CALL MPI_TYPE_VECTOR( ngp_xy, 1, nzt-nzb+2, MPI_REAL, type_xy, ierr )
619	CALL MPI_TYPE_COMMIT( type_xy, ierr )
620	#endif
621
622	#else
623
624	!
625	!-- Array bounds when running on a single PE (respectively a non-parallel
626	!-- machine)
627	nxl = 0
628	nxr = nx
629	nxra = nx
630	nnx = nxr - nxl + 1
631	nys = 0
632	nyn = ny
633	nyna = ny
634	nny = nyn - nys + 1
635	nzb = 0
636	nzt = nz
637	nzta = nz
638	nnz = nz
639
640	ALLOCATE( hor_index_bounds(4,0:0) )
641	hor_index_bounds(1,0) = nxl
642	hor_index_bounds(2,0) = nxr
643	hor_index_bounds(3,0) = nys
644	hor_index_bounds(4,0) = nyn
645
646	!
647	!-- Array bounds for the pressure solver (in the parallel code, these bounds
648	!-- are the ones for the transposed arrays)
649	nys_x = nys
650	nyn_x = nyn
651	nyn_xa = nyn
652	nzb_x = nzb + 1
653	nzt_x = nzt
654	nzt_xa = nzt
655
656	nxl_y = nxl
657	nxr_y = nxr
658	nxr_ya = nxr
659	nzb_y = nzb + 1
660	nzt_y = nzt
661	nzt_ya = nzt
662
663	nxl_z = nxl
664	nxr_z = nxr
665	nxr_za = nxr
666	nys_z = nys
667	nyn_z = nyn
668	nyn_za = nyn
669
670	#endif
671
672	!
673	!-- Calculate number of grid levels necessary for the multigrid poisson solver
674	!-- as well as the gridpoint indices on each level
675	IF ( psolver == 'multigrid' ) THEN
676
677	!
678	!-- First calculate number of possible grid levels for the subdomains
679	mg_levels_x = 1
680	mg_levels_y = 1
681	mg_levels_z = 1
682
683	i = nnx
684	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
685	i = i / 2
686	mg_levels_x = mg_levels_x + 1
687	ENDDO
688
689	j = nny
690	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
691	j = j / 2
692	mg_levels_y = mg_levels_y + 1
693	ENDDO
694
695	k = nnz
696	DO WHILE ( MOD( k, 2 ) == 0 .AND. k /= 2 )
697	k = k / 2
698	mg_levels_z = mg_levels_z + 1
699	ENDDO
700
701	maximum_grid_level = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
702
703	!
704	!-- Find out, if the total domain allows more levels. These additional
705	!-- levels are processed on PE0 only.
706	IF ( numprocs > 1 ) THEN
707	IF ( mg_levels_z > MIN( mg_levels_x, mg_levels_y ) ) THEN
708	mg_switch_to_pe0_level_l = maximum_grid_level
709
710	mg_levels_x = 1
711	mg_levels_y = 1
712
713	i = nx+1
714	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
715	i = i / 2
716	mg_levels_x = mg_levels_x + 1
717	ENDDO
718
719	j = ny+1
720	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
721	j = j / 2
722	mg_levels_y = mg_levels_y + 1
723	ENDDO
724
725	maximum_grid_level_l = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
726
727	IF ( maximum_grid_level_l > mg_switch_to_pe0_level_l ) THEN
728	mg_switch_to_pe0_level_l = maximum_grid_level_l - &
729	mg_switch_to_pe0_level_l + 1
730	ELSE
731	mg_switch_to_pe0_level_l = 0
732	ENDIF
733	ELSE
734	mg_switch_to_pe0_level_l = 0
735	maximum_grid_level_l = maximum_grid_level
736	ENDIF
737
738	!
739	!-- Use switch level calculated above only if it is not pre-defined
740	!-- by user
741	IF ( mg_switch_to_pe0_level == 0 ) THEN
742
743	IF ( mg_switch_to_pe0_level_l /= 0 ) THEN
744	mg_switch_to_pe0_level = mg_switch_to_pe0_level_l
745	maximum_grid_level = maximum_grid_level_l
746	ENDIF
747
748	ELSE
749	!
750	!-- Check pre-defined value and reset to default, if neccessary
751	IF ( mg_switch_to_pe0_level < mg_switch_to_pe0_level_l .OR. &
752	mg_switch_to_pe0_level >= maximum_grid_level_l ) THEN
753	IF ( myid == 0 ) THEN
754	PRINT*, '+++ WARNING init_pegrid: mg_switch_to_pe0_level ', &
755	'out of range and reset to default (=0)'
756	ENDIF
757	mg_switch_to_pe0_level = 0
758	ELSE
759	!
760	!-- Use the largest number of possible levels anyway and recalculate
761	!-- the switch level to this largest number of possible values
762	maximum_grid_level = maximum_grid_level_l
763
764	ENDIF
765	ENDIF
766
767	ENDIF
768
769	ALLOCATE( grid_level_count(maximum_grid_level), &
770	nxl_mg(maximum_grid_level), nxr_mg(maximum_grid_level), &
771	nyn_mg(maximum_grid_level), nys_mg(maximum_grid_level), &
772	nzt_mg(maximum_grid_level) )
773
774	grid_level_count = 0
775	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzt_l = nzt
776
777	DO i = maximum_grid_level, 1 , -1
778
779	IF ( i == mg_switch_to_pe0_level ) THEN
780	#if defined( __parallel )
781	!
782	!-- Save the grid size of the subdomain at the switch level, because
783	!-- it is needed in poismg.
784	!-- Array bounds of the local subdomain grids are gathered on PE0
785	ind(1) = nxl_l; ind(2) = nxr_l
786	ind(3) = nys_l; ind(4) = nyn_l
787	ind(5) = nzt_l
788	ALLOCATE( ind_all(5*numprocs), mg_loc_ind(5,0:numprocs-1) )
789	CALL MPI_ALLGATHER( ind, 5, MPI_INTEGER, ind_all, 5, &
790	MPI_INTEGER, comm2d, ierr )
791	DO j = 0, numprocs-1
792	DO k = 1, 5
793	mg_loc_ind(k,j) = ind_all(k+j*5)
794	ENDDO
795	ENDDO
796	DEALLOCATE( ind_all )
797	!
798	!-- Calculate the grid size of the total domain gathered on PE0
799	nxr_l = ( nxr_l-nxl_l+1 ) * pdims(1) - 1
800	nxl_l = 0
801	nyn_l = ( nyn_l-nys_l+1 ) * pdims(2) - 1
802	nys_l = 0
803	!
804	!-- The size of this gathered array must not be larger than the
805	!-- array tend, which is used in the multigrid scheme as a temporary
806	!-- array
807	subdomain_size = ( nxr - nxl + 3 ) * ( nyn - nys + 3 ) * &
808	( nzt - nzb + 2 )
809	gathered_size = ( nxr_l - nxl_l + 3 ) * ( nyn_l - nys_l + 3 ) * &
810	( nzt_l - nzb + 2 )
811
812	IF ( gathered_size > subdomain_size ) THEN
813	IF ( myid == 0 ) THEN
814	PRINT*, '+++ init_pegrid: not enough memory for storing ', &
815	'gathered multigrid data on PE0'
816	ENDIF
817	CALL local_stop
818	ENDIF
819	#else
820	PRINT*, '+++ init_pegrid: multigrid gather/scatter impossible ', &
821	'in non parallel mode'
822	CALL local_stop
823	#endif
824	ENDIF
825
826	nxl_mg(i) = nxl_l
827	nxr_mg(i) = nxr_l
828	nys_mg(i) = nys_l
829	nyn_mg(i) = nyn_l
830	nzt_mg(i) = nzt_l
831
832	nxl_l = nxl_l / 2
833	nxr_l = nxr_l / 2
834	nys_l = nys_l / 2
835	nyn_l = nyn_l / 2
836	nzt_l = nzt_l / 2
837	ENDDO
838
839	ELSE
840
841	maximum_grid_level = 1
842
843	ENDIF
844
845	grid_level = maximum_grid_level
846
847	#if defined( __parallel )
848	!
849	!-- Gridpoint number for the exchange of ghost points (y-line for 2D-arrays)
850	ngp_y = nyn - nys + 1
851
852	!
853	!-- Define a new MPI derived datatype for the exchange of ghost points in
854	!-- y-direction for 2D-arrays (line)
855	CALL MPI_TYPE_VECTOR( nxr-nxl+3, 1, ngp_y+2, MPI_REAL, type_x, ierr )
856	CALL MPI_TYPE_COMMIT( type_x, ierr )
857	CALL MPI_TYPE_VECTOR( nxr-nxl+3, 1, ngp_y+2, MPI_INTEGER, type_x_int, ierr )
858	CALL MPI_TYPE_COMMIT( type_x_int, ierr )
859
860	!
861	!-- Calculate gridpoint numbers for the exchange of ghost points along x
862	!-- (yz-plane for 3D-arrays) and define MPI derived data type(s) for the
863	!-- exchange of ghost points in y-direction (xz-plane).
864	!-- Do these calculations for the model grid and (if necessary) also
865	!-- for the coarser grid levels used in the multigrid method
866	ALLOCATE ( ngp_yz(maximum_grid_level), type_xz(maximum_grid_level) )
867
868	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzb_l = nzb; nzt_l = nzt
869
870	DO i = maximum_grid_level, 1 , -1
871	ngp_yz(i) = (nzt_l - nzb_l + 2) * (nyn_l - nys_l + 3)
872
873	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+3, nzt_l-nzb_l+2, ngp_yz(i), &
874	MPI_REAL, type_xz(i), ierr )
875	CALL MPI_TYPE_COMMIT( type_xz(i), ierr )
876
877	nxl_l = nxl_l / 2
878	nxr_l = nxr_l / 2
879	nys_l = nys_l / 2
880	nyn_l = nyn_l / 2
881	nzt_l = nzt_l / 2
882	ENDDO
883	#endif
884
885	#if defined( __parallel )
886	!
887	!-- Setting of flags for inflow/outflow conditions in case of non-cyclic
888	!-- horizontal boundary conditions.
889	IF ( pleft == MPI_PROC_NULL ) THEN
890	IF ( bc_lr == 'dirichlet/radiation' ) THEN
891	inflow_l = .TRUE.
892	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
893	outflow_l = .TRUE.
894	ENDIF
895	ENDIF
896
897	IF ( pright == MPI_PROC_NULL ) THEN
898	IF ( bc_lr == 'dirichlet/radiation' ) THEN
899	outflow_r = .TRUE.
900	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
901	inflow_r = .TRUE.
902	ENDIF
903	ENDIF
904
905	IF ( psouth == MPI_PROC_NULL ) THEN
906	IF ( bc_ns == 'dirichlet/radiation' ) THEN
907	outflow_s = .TRUE.
908	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
909	inflow_s = .TRUE.
910	ENDIF
911	ENDIF
912
913	IF ( pnorth == MPI_PROC_NULL ) THEN
914	IF ( bc_ns == 'dirichlet/radiation' ) THEN
915	inflow_n = .TRUE.
916	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
917	outflow_n = .TRUE.
918	ENDIF
919	ENDIF
920
921	!
922	!-- Broadcast the id of the inflow PE
923	IF ( inflow_l ) THEN
924	id_inflow_l = myid
925	ELSE
926	id_inflow_l = 0
927	ENDIF
928	CALL MPI_ALLREDUCE( id_inflow_l, id_inflow, 1, MPI_INTEGER, MPI_SUM, &
929	comm1dx, ierr )
930
931	#else
932	IF ( bc_lr == 'dirichlet/radiation' ) THEN
933	inflow_l = .TRUE.
934	outflow_r = .TRUE.
935	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
936	outflow_l = .TRUE.
937	inflow_r = .TRUE.
938	ENDIF
939
940	IF ( bc_ns == 'dirichlet/radiation' ) THEN
941	inflow_n = .TRUE.
942	outflow_s = .TRUE.
943	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
944	outflow_n = .TRUE.
945	inflow_s = .TRUE.
946	ENDIF
947	#endif
948	!
949	!-- At the outflow, u or v, respectively, have to be calculated for one more
950	!-- grid point.
951	IF ( outflow_l ) THEN
952	nxlu = nxl + 1
953	ELSE
954	nxlu = nxl
955	ENDIF
956	IF ( outflow_s ) THEN
957	nysv = nys + 1
958	ELSE
959	nysv = nys
960	ENDIF
961
962	IF ( psolver == 'poisfft_hybrid' ) THEN
963	CALL poisfft_hybrid_ini
964	ELSEIF ( psolver == 'poisfft' ) THEN
965	CALL poisfft_init
966	ENDIF
967
968	!
969	!-- Allocate wall flag arrays used in the multigrid solver
970	IF ( psolver == 'multigrid' ) THEN
971
972	DO i = maximum_grid_level, 1, -1
973
974	SELECT CASE ( i )
975
976	CASE ( 1 )
977	ALLOCATE( wall_flags_1(nzb:nzt_mg(i)+1, &
978	nys_mg(i)-1:nyn_mg(i)+1, &
979	nxl_mg(i)-1:nxr_mg(i)+1) )
980
981	CASE ( 2 )
982	ALLOCATE( wall_flags_2(nzb:nzt_mg(i)+1, &
983	nys_mg(i)-1:nyn_mg(i)+1, &
984	nxl_mg(i)-1:nxr_mg(i)+1) )
985
986	CASE ( 3 )
987	ALLOCATE( wall_flags_3(nzb:nzt_mg(i)+1, &
988	nys_mg(i)-1:nyn_mg(i)+1, &
989	nxl_mg(i)-1:nxr_mg(i)+1) )
990
991	CASE ( 4 )
992	ALLOCATE( wall_flags_4(nzb:nzt_mg(i)+1, &
993	nys_mg(i)-1:nyn_mg(i)+1, &
994	nxl_mg(i)-1:nxr_mg(i)+1) )
995
996	CASE ( 5 )
997	ALLOCATE( wall_flags_5(nzb:nzt_mg(i)+1, &
998	nys_mg(i)-1:nyn_mg(i)+1, &
999	nxl_mg(i)-1:nxr_mg(i)+1) )
1000
1001	CASE ( 6 )
1002	ALLOCATE( wall_flags_6(nzb:nzt_mg(i)+1, &
1003	nys_mg(i)-1:nyn_mg(i)+1, &
1004	nxl_mg(i)-1:nxr_mg(i)+1) )
1005
1006	CASE ( 7 )
1007	ALLOCATE( wall_flags_7(nzb:nzt_mg(i)+1, &
1008	nys_mg(i)-1:nyn_mg(i)+1, &
1009	nxl_mg(i)-1:nxr_mg(i)+1) )
1010
1011	CASE ( 8 )
1012	ALLOCATE( wall_flags_8(nzb:nzt_mg(i)+1, &
1013	nys_mg(i)-1:nyn_mg(i)+1, &
1014	nxl_mg(i)-1:nxr_mg(i)+1) )
1015
1016	CASE ( 9 )
1017	ALLOCATE( wall_flags_9(nzb:nzt_mg(i)+1, &
1018	nys_mg(i)-1:nyn_mg(i)+1, &
1019	nxl_mg(i)-1:nxr_mg(i)+1) )
1020
1021	CASE ( 10 )
1022	ALLOCATE( wall_flags_10(nzb:nzt_mg(i)+1, &
1023	nys_mg(i)-1:nyn_mg(i)+1, &
1024	nxl_mg(i)-1:nxr_mg(i)+1) )
1025
1026	CASE DEFAULT
1027	IF ( myid == 0 ) PRINT*, '+++ init_pegrid: more than 10 ', &
1028	' multigrid levels'
1029	CALL local_stop
1030
1031	END SELECT
1032
1033	ENDDO
1034
1035	ENDIF
1036
1037	END SUBROUTINE init_pegrid

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

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |