Home

Context Navigation

source: palm/trunk/SOURCE/init_pegrid.f90 @ 190

Last change on this file since 190 was 181, checked in by raasch, 16 years ago
bugfixes + adjustments for SGI ICE system
Property svn:keywords set to `Id`
File size: 33.2 KB

Line
1	SUBROUTINE init_pegrid
2
3	!------------------------------------------------------------------------------!
4	! Actual revisions:
5	! -----------------
6	! ATTENTION: nnz_x undefined problem still has to be solved!!!!!!!!
7	! nz is used instead nnz for calculating mg-levels
8	! Collect on PE0 horizontal index bounds from all other PEs,
9	! broadcast the id of the inflow PE (using the respective communicator)
10	! TEST OUTPUT (TO BE REMOVED) logging mpi2 ierr values
11	!
12	! Former revisions:
13	! -----------------
14	! $Id: init_pegrid.f90 181 2008-07-30 07:07:47Z letzel $
15	!
16	! 114 2007-10-10 00:03:15Z raasch
17	! Allocation of wall flag arrays for multigrid solver
18	!
19	! 108 2007-08-24 15:10:38Z letzel
20	! Intercommunicator (comm_inter) and derived data type (type_xy) for
21	! coupled model runs created, assign coupling_mode_remote,
22	! indices nxlu and nysv are calculated (needed for non-cyclic boundary
23	! conditions)
24	!
25	! 82 2007-04-16 15:40:52Z raasch
26	! Cpp-directive lcmuk changed to intel_openmp_bug, setting of host on lcmuk by
27	! cpp-directive removed
28	!
29	! 75 2007-03-22 09:54:05Z raasch
30	! uxrp, vynp eliminated,
31	! dirichlet/neumann changed to dirichlet/radiation, etc.,
32	! poisfft_init is only called if fft-solver is switched on
33	!
34	! RCS Log replace by Id keyword, revision history cleaned up
35	!
36	! Revision 1.28 2006/04/26 13:23:32 raasch
37	! lcmuk does not understand the !$ comment so a cpp-directive is required
38	!
39	! Revision 1.1 1997/07/24 11:15:09 raasch
40	! Initial revision
41	!
42	!
43	! Description:
44	! ------------
45	! Determination of the virtual processor topology (if not prescribed by the
46	! user)and computation of the grid point number and array bounds of the local
47	! domains.
48	!------------------------------------------------------------------------------!
49
50	USE control_parameters
51	USE fft_xy
52	USE grid_variables
53	USE indices
54	USE pegrid
55	USE poisfft_mod
56	USE poisfft_hybrid_mod
57	USE statistics
58	USE transpose_indices
59
60
61	IMPLICIT NONE
62
63	INTEGER :: gathered_size, i, id_inflow_l, id_recycling_l, ind(5), j, k, &
64	maximum_grid_level_l, mg_switch_to_pe0_level_l, mg_levels_x, &
65	mg_levels_y, mg_levels_z, nnx_y, nnx_z, nny_x, nny_z, nnz_x, &
66	nnz_y, numproc_sqr, nx_total, nxl_l, nxr_l, nyn_l, nys_l, &
67	nzb_l, nzt_l, omp_get_num_threads, subdomain_size
68
69	INTEGER, DIMENSION(:), ALLOCATABLE :: ind_all, nxlf, nxrf, nynf, nysf
70
71	LOGICAL :: found
72
73	!
74	!-- Get the number of OpenMP threads
75	!$OMP PARALLEL
76	#if defined( __intel_openmp_bug )
77	threads_per_task = omp_get_num_threads()
78	#else
79	!$ threads_per_task = omp_get_num_threads()
80	#endif
81	!$OMP END PARALLEL
82
83
84	#if defined( __parallel )
85	!
86	!-- Determine the processor topology or check it, if prescribed by the user
87	IF ( npex == -1 .AND. npey == -1 ) THEN
88
89	!
90	!-- Automatic determination of the topology
91	!-- The default on SMP- and cluster-hosts is a 1d-decomposition along x
92	IF ( host(1:3) == 'ibm' .OR. host(1:3) == 'nec' .OR. &
93	host(1:2) == 'lc' .OR. host(1:3) == 'dec' ) THEN
94
95	pdims(1) = numprocs
96	pdims(2) = 1
97
98	ELSE
99
100	numproc_sqr = SQRT( REAL( numprocs ) )
101	pdims(1) = MAX( numproc_sqr , 1 )
102	DO WHILE ( MOD( numprocs , pdims(1) ) /= 0 )
103	pdims(1) = pdims(1) - 1
104	ENDDO
105	pdims(2) = numprocs / pdims(1)
106
107	ENDIF
108
109	ELSEIF ( npex /= -1 .AND. npey /= -1 ) THEN
110
111	!
112	!-- Prescribed by user. Number of processors on the prescribed topology
113	!-- must be equal to the number of PEs available to the job
114	IF ( ( npex * npey ) /= numprocs ) THEN
115	PRINT*, '+++ init_pegrid:'
116	PRINT, ' number of PEs of the prescribed topology (', npexnpey, &
117	') does not match the number of PEs available to the ', &
118	'job (', numprocs, ')'
119	CALL local_stop
120	ENDIF
121	pdims(1) = npex
122	pdims(2) = npey
123
124	ELSE
125	!
126	!-- If the processor topology is prescribed by the user, the number of
127	!-- PEs must be given in both directions
128	PRINT*, '+++ init_pegrid:'
129	PRINT*, ' if the processor topology is prescribed by the user, ', &
130	'both values of "npex" and "npey" must be given in the ', &
131	'NAMELIST-parameter file'
132	CALL local_stop
133
134	ENDIF
135
136	!
137	!-- The hybrid solver can only be used in case of a 1d-decomposition along x
138	IF ( pdims(2) /= 1 .AND. psolver == 'poisfft_hybrid' ) THEN
139	IF ( myid == 0 ) THEN
140	PRINT, '** init_pegrid: psolver = "poisfft_hybrid" can only be'
141	PRINT*, ' used in case of a 1d-decomposition along x'
142	ENDIF
143	ENDIF
144
145	!
146	!-- If necessary, set horizontal boundary conditions to non-cyclic
147	IF ( bc_lr /= 'cyclic' ) cyclic(1) = .FALSE.
148	IF ( bc_ns /= 'cyclic' ) cyclic(2) = .FALSE.
149
150	!
151	!-- Create the virtual processor grid
152	CALL MPI_CART_CREATE( comm_palm, ndim, pdims, cyclic, reorder, &
153	comm2d, ierr )
154	CALL MPI_COMM_RANK( comm2d, myid, ierr )
155	WRITE (myid_char,'(''_'',I4.4)') myid
156
157	CALL MPI_CART_COORDS( comm2d, myid, ndim, pcoord, ierr )
158	CALL MPI_CART_SHIFT( comm2d, 0, 1, pleft, pright, ierr )
159	CALL MPI_CART_SHIFT( comm2d, 1, 1, psouth, pnorth, ierr )
160
161	!
162	!-- Determine sub-topologies for transpositions
163	!-- Transposition from z to x:
164	remain_dims(1) = .TRUE.
165	remain_dims(2) = .FALSE.
166	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dx, ierr )
167	CALL MPI_COMM_RANK( comm1dx, myidx, ierr )
168	!
169	!-- Transposition from x to y
170	remain_dims(1) = .FALSE.
171	remain_dims(2) = .TRUE.
172	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dy, ierr )
173	CALL MPI_COMM_RANK( comm1dy, myidy, ierr )
174
175
176	!
177	!-- Find a grid (used for array d) which will match the transposition demands
178	IF ( grid_matching == 'strict' ) THEN
179
180	nxa = nx; nya = ny; nza = nz
181
182	ELSE
183
184	found = .FALSE.
185	xn: DO nxa = nx, 2*nx
186	!
187	!-- Meet conditions for nx
188	IF ( MOD( nxa+1, pdims(1) ) /= 0 .OR. &
189	MOD( nxa+1, pdims(2) ) /= 0 ) CYCLE xn
190
191	yn: DO nya = ny, 2*ny
192	!
193	!-- Meet conditions for ny
194	IF ( MOD( nya+1, pdims(2) ) /= 0 .OR. &
195	MOD( nya+1, pdims(1) ) /= 0 ) CYCLE yn
196
197
198	zn: DO nza = nz, 2*nz
199	!
200	!-- Meet conditions for nz
201	IF ( ( MOD( nza, pdims(1) ) /= 0 .AND. pdims(1) /= 1 .AND. &
202	pdims(2) /= 1 ) .OR. &
203	( MOD( nza, pdims(2) ) /= 0 .AND. dt_dosp /= 9999999.9 &
204	) ) THEN
205	CYCLE zn
206	ELSE
207	found = .TRUE.
208	EXIT xn
209	ENDIF
210
211	ENDDO zn
212
213	ENDDO yn
214
215	ENDDO xn
216
217	IF ( .NOT. found ) THEN
218	IF ( myid == 0 ) THEN
219	PRINT*,'+++ init_pegrid: no matching grid for transpositions found'
220	ENDIF
221	CALL local_stop
222	ENDIF
223
224	ENDIF
225
226	!
227	!-- Calculate array bounds in x-direction for every PE.
228	!-- The last PE along x may get less grid points than the others
229	ALLOCATE( nxlf(0:pdims(1)-1), nxrf(0:pdims(1)-1), nynf(0:pdims(2)-1), &
230	nysf(0:pdims(2)-1), nnx_pe(0:pdims(1)-1), nny_pe(0:pdims(2)-1) )
231
232	IF ( MOD( nxa+1 , pdims(1) ) /= 0 ) THEN
233	IF ( myid == 0 ) THEN
234	PRINT*,'+++ x-direction: gridpoint number (',nx+1,') is not an'
235	PRINT*,' integral divisor of the number of proces', &
236	&'sors (', pdims(1),')'
237	ENDIF
238	CALL local_stop
239	ELSE
240	nnx = ( nxa + 1 ) / pdims(1)
241	IF ( nnx*pdims(1) - ( nx + 1) > nnx ) THEN
242	IF ( myid == 0 ) THEN
243	PRINT*,'+++ x-direction: nx does not match the requirements ', &
244	'given by the number of PEs'
245	PRINT*,' used'
246	PRINT,' please use nx = ', nx - ( pdims(1) - ( nnxpdims(1) &
247	- ( nx + 1 ) ) ), ' instead of nx =', nx
248	ENDIF
249	CALL local_stop
250	ENDIF
251	ENDIF
252
253	!
254	!-- Left and right array bounds, number of gridpoints
255	DO i = 0, pdims(1)-1
256	nxlf(i) = i * nnx
257	nxrf(i) = ( i + 1 ) * nnx - 1
258	nnx_pe(i) = MIN( nx, nxrf(i) ) - nxlf(i) + 1
259	ENDDO
260
261	!
262	!-- Calculate array bounds in y-direction for every PE.
263	IF ( MOD( nya+1 , pdims(2) ) /= 0 ) THEN
264	IF ( myid == 0 ) THEN
265	PRINT*,'+++ y-direction: gridpoint number (',ny+1,') is not an'
266	PRINT*,' integral divisor of the number of proces', &
267	&'sors (', pdims(2),')'
268	ENDIF
269	CALL local_stop
270	ELSE
271	nny = ( nya + 1 ) / pdims(2)
272	IF ( nny*pdims(2) - ( ny + 1) > nny ) THEN
273	IF ( myid == 0 ) THEN
274	PRINT*,'+++ x-direction: nx does not match the requirements ', &
275	'given by the number of PEs'
276	PRINT*,' used'
277	PRINT,' please use nx = ', nx - ( pdims(1) - ( nnxpdims(1) &
278	- ( nx + 1 ) ) ), ' instead of nx =', nx
279	ENDIF
280	CALL local_stop
281	ENDIF
282	ENDIF
283
284	!
285	!-- South and north array bounds
286	DO j = 0, pdims(2)-1
287	nysf(j) = j * nny
288	nynf(j) = ( j + 1 ) * nny - 1
289	nny_pe(j) = MIN( ny, nynf(j) ) - nysf(j) + 1
290	ENDDO
291
292	!
293	!-- Local array bounds of the respective PEs
294	nxl = nxlf(pcoord(1))
295	nxra = nxrf(pcoord(1))
296	nxr = MIN( nx, nxra )
297	nys = nysf(pcoord(2))
298	nyna = nynf(pcoord(2))
299	nyn = MIN( ny, nyna )
300	nzb = 0
301	nzta = nza
302	nzt = MIN( nz, nzta )
303	nnz = nza
304
305	!
306	!-- Calculate array bounds and gridpoint numbers for the transposed arrays
307	!-- (needed in the pressure solver)
308	!-- For the transposed arrays, cyclic boundaries as well as top and bottom
309	!-- boundaries are omitted, because they are obstructive to the transposition
310
311	!
312	!-- 1. transposition z --> x
313	!-- This transposition is not neccessary in case of a 1d-decomposition along x,
314	!-- except that the uptream-spline method is switched on
315	IF ( pdims(2) /= 1 .OR. momentum_advec == 'ups-scheme' .OR. &
316	scalar_advec == 'ups-scheme' ) THEN
317
318	IF ( pdims(2) == 1 .AND. ( momentum_advec == 'ups-scheme' .OR. &
319	scalar_advec == 'ups-scheme' ) ) THEN
320	IF ( myid == 0 ) THEN
321	PRINT*,'+++ WARNING: init_pegrid: 1d-decomposition along x ', &
322	&'chosen but nz restrictions may occur'
323	PRINT*,' since ups-scheme is activated'
324	ENDIF
325	ENDIF
326	nys_x = nys
327	nyn_xa = nyna
328	nyn_x = nyn
329	nny_x = nny
330	IF ( MOD( nza , pdims(1) ) /= 0 ) THEN
331	IF ( myid == 0 ) THEN
332	PRINT*,'+++ transposition z --> x:'
333	PRINT*,' nz=',nz,' is not an integral divisior of pdims(1)=', &
334	&pdims(1)
335	ENDIF
336	CALL local_stop
337	ENDIF
338	nnz_x = nza / pdims(1)
339	nzb_x = 1 + myidx * nnz_x
340	nzt_xa = ( myidx + 1 ) * nnz_x
341	nzt_x = MIN( nzt, nzt_xa )
342
343	sendrecvcount_zx = nnx * nny * nnz_x
344
345	ELSE
346	!
347	!--- Setting of dummy values because otherwise variables are undefined in
348	!--- the next step x --> y
349	!--- WARNING: This case has still to be clarified!!!!!!!!!!!!
350	nnz_x = 1
351	nzb_x = 1
352	nzt_xa = 1
353	nzt_x = 1
354	nny_x = nny
355
356	ENDIF
357
358	!
359	!-- 2. transposition x --> y
360	nnz_y = nnz_x
361	nzb_y = nzb_x
362	nzt_ya = nzt_xa
363	nzt_y = nzt_x
364	IF ( MOD( nxa+1 , pdims(2) ) /= 0 ) THEN
365	IF ( myid == 0 ) THEN
366	PRINT*,'+++ transposition x --> y:'
367	PRINT*,' nx+1=',nx+1,' is not an integral divisor of ',&
368	&'pdims(2)=',pdims(2)
369	ENDIF
370	CALL local_stop
371	ENDIF
372	nnx_y = (nxa+1) / pdims(2)
373	nxl_y = myidy * nnx_y
374	nxr_ya = ( myidy + 1 ) * nnx_y - 1
375	nxr_y = MIN( nx, nxr_ya )
376
377	sendrecvcount_xy = nnx_y * nny_x * nnz_y
378
379	!
380	!-- 3. transposition y --> z (ELSE: x --> y in case of 1D-decomposition
381	!-- along x)
382	IF ( pdims(2) /= 1 .OR. momentum_advec == 'ups-scheme' .OR. &
383	scalar_advec == 'ups-scheme' ) THEN
384	!
385	!-- y --> z
386	!-- This transposition is not neccessary in case of a 1d-decomposition
387	!-- along x, except that the uptream-spline method is switched on
388	nnx_z = nnx_y
389	nxl_z = nxl_y
390	nxr_za = nxr_ya
391	nxr_z = nxr_y
392	IF ( MOD( nya+1 , pdims(1) ) /= 0 ) THEN
393	IF ( myid == 0 ) THEN
394	PRINT*,'+++ Transposition y --> z:'
395	PRINT*,' ny+1=',ny+1,' is not an integral divisor of ',&
396	&'pdims(1)=',pdims(1)
397	ENDIF
398	CALL local_stop
399	ENDIF
400	nny_z = (nya+1) / pdims(1)
401	nys_z = myidx * nny_z
402	nyn_za = ( myidx + 1 ) * nny_z - 1
403	nyn_z = MIN( ny, nyn_za )
404
405	sendrecvcount_yz = nnx_y * nny_z * nnz_y
406
407	ELSE
408	!
409	!-- x --> y. This condition must be fulfilled for a 1D-decomposition along x
410	IF ( MOD( nya+1 , pdims(1) ) /= 0 ) THEN
411	IF ( myid == 0 ) THEN
412	PRINT*,'+++ Transposition x --> y:'
413	PRINT*,' ny+1=',ny+1,' is not an integral divisor of ',&
414	&'pdims(1)=',pdims(1)
415	ENDIF
416	CALL local_stop
417	ENDIF
418
419	ENDIF
420
421	!
422	!-- Indices for direct transpositions z --> y (used for calculating spectra)
423	IF ( dt_dosp /= 9999999.9 ) THEN
424	IF ( MOD( nza, pdims(2) ) /= 0 ) THEN
425	IF ( myid == 0 ) THEN
426	PRINT*,'+++ Direct transposition z --> y (needed for spectra):'
427	PRINT*,' nz=',nz,' is not an integral divisor of ',&
428	&'pdims(2)=',pdims(2)
429	ENDIF
430	CALL local_stop
431	ELSE
432	nxl_yd = nxl
433	nxr_yda = nxra
434	nxr_yd = nxr
435	nzb_yd = 1 + myidy * ( nza / pdims(2) )
436	nzt_yda = ( myidy + 1 ) * ( nza / pdims(2) )
437	nzt_yd = MIN( nzt, nzt_yda )
438
439	sendrecvcount_zyd = nnx * nny * ( nza / pdims(2) )
440	ENDIF
441	ENDIF
442
443	!
444	!-- Indices for direct transpositions y --> x (they are only possible in case
445	!-- of a 1d-decomposition along x)
446	IF ( pdims(2) == 1 ) THEN
447	nny_x = nny / pdims(1)
448	nys_x = myid * nny_x
449	nyn_xa = ( myid + 1 ) * nny_x - 1
450	nyn_x = MIN( ny, nyn_xa )
451	nzb_x = 1
452	nzt_xa = nza
453	nzt_x = nz
454	sendrecvcount_xy = nnx * nny_x * nza
455	ENDIF
456
457	!
458	!-- Indices for direct transpositions x --> y (they are only possible in case
459	!-- of a 1d-decomposition along y)
460	IF ( pdims(1) == 1 ) THEN
461	nnx_y = nnx / pdims(2)
462	nxl_y = myid * nnx_y
463	nxr_ya = ( myid + 1 ) * nnx_y - 1
464	nxr_y = MIN( nx, nxr_ya )
465	nzb_y = 1
466	nzt_ya = nza
467	nzt_y = nz
468	sendrecvcount_xy = nnx_y * nny * nza
469	ENDIF
470
471	!
472	!-- Arrays for storing the array bounds are needed any more
473	DEALLOCATE( nxlf , nxrf , nynf , nysf )
474
475	!
476	!-- Collect index bounds from other PEs (to be written to restart file later)
477	ALLOCATE( hor_index_bounds(4,0:numprocs-1) )
478
479	IF ( myid == 0 ) THEN
480
481	hor_index_bounds(1,0) = nxl
482	hor_index_bounds(2,0) = nxr
483	hor_index_bounds(3,0) = nys
484	hor_index_bounds(4,0) = nyn
485
486	!
487	!-- Receive data from all other PEs
488	DO i = 1, numprocs-1
489	CALL MPI_RECV( ibuf, 4, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, &
490	ierr )
491	hor_index_bounds(:,i) = ibuf(1:4)
492	ENDDO
493
494	ELSE
495	!
496	!-- Send index bounds to PE0
497	ibuf(1) = nxl
498	ibuf(2) = nxr
499	ibuf(3) = nys
500	ibuf(4) = nyn
501	CALL MPI_SEND( ibuf, 4, MPI_INTEGER, 0, myid, comm2d, ierr )
502
503	ENDIF
504
505	#if defined( __print )
506	!
507	!-- Control output
508	IF ( myid == 0 ) THEN
509	PRINT, ' processor topology *'
510	PRINT*, ' '
511	PRINT*, 'myid pcoord left right south north idx idy nxl: nxr',&
512	&' nys: nyn'
513	PRINT*, '------------------------------------------------------------',&
514	&'-----------'
515	WRITE (*,1000) 0, pcoord(1), pcoord(2), pleft, pright, psouth, pnorth, &
516	myidx, myidy, nxl, nxr, nys, nyn
517	1000 FORMAT (I4,2X,'(',I3,',',I3,')',3X,I4,2X,I4,3X,I4,2X,I4,2X,I3,1X,I3, &
518	2(2X,I4,':',I4))
519
520	!
521	!-- Receive data from the other PEs
522	DO i = 1,numprocs-1
523	CALL MPI_RECV( ibuf, 12, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, &
524	ierr )
525	WRITE (*,1000) i, ( ibuf(j) , j = 1,12 )
526	ENDDO
527	ELSE
528
529	!
530	!-- Send data to PE0
531	ibuf(1) = pcoord(1); ibuf(2) = pcoord(2); ibuf(3) = pleft
532	ibuf(4) = pright; ibuf(5) = psouth; ibuf(6) = pnorth; ibuf(7) = myidx
533	ibuf(8) = myidy; ibuf(9) = nxl; ibuf(10) = nxr; ibuf(11) = nys
534	ibuf(12) = nyn
535	CALL MPI_SEND( ibuf, 12, MPI_INTEGER, 0, myid, comm2d, ierr )
536	ENDIF
537	#endif
538
539	#if defined( __mpi2 )
540	!
541	!-- In case of coupled runs, get the port name on PE0 of the atmosphere model
542	!-- and pass it to PE0 of the ocean model
543	IF ( myid == 0 ) THEN
544
545	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
546
547	CALL MPI_OPEN_PORT( MPI_INFO_NULL, port_name, ierr )
548	!
549	!-- TEST OUTPUT (TO BE REMOVED)
550	WRITE(9,*) TRIM( coupling_mode ), &
551	', ierr after MPI_OPEN_PORT: ', ierr
552	CALL LOCAL_FLUSH( 9 )
553
554	CALL MPI_PUBLISH_NAME( 'palm_coupler', MPI_INFO_NULL, port_name, &
555	ierr )
556	!
557	!-- TEST OUTPUT (TO BE REMOVED)
558	WRITE(9,*) TRIM( coupling_mode ), &
559	', ierr after MPI_PUBLISH_NAME: ', ierr
560	CALL LOCAL_FLUSH( 9 )
561
562	!
563	!-- Write a flag file for the ocean model and the other atmosphere
564	!-- processes.
565	!-- There seems to be a bug in MPICH2 which causes hanging processes
566	!-- in case that execution of LOOKUP_NAME is continued too early
567	!-- (i.e. before the port has been created)
568	OPEN( 90, FILE='COUPLING_PORT_OPENED', FORM='FORMATTED' )
569	WRITE ( 90, '(''TRUE'')' )
570	CLOSE ( 90 )
571
572	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
573
574	!
575	!-- Continue only if the atmosphere model has created the port.
576	!-- There seems to be a bug in MPICH2 which causes hanging processes
577	!-- in case that execution of LOOKUP_NAME is continued too early
578	!-- (i.e. before the port has been created)
579	INQUIRE( FILE='COUPLING_PORT_OPENED', EXIST=found )
580	DO WHILE ( .NOT. found )
581	INQUIRE( FILE='COUPLING_PORT_OPENED', EXIST=found )
582	ENDDO
583
584	CALL MPI_LOOKUP_NAME( 'palm_coupler', MPI_INFO_NULL, port_name, ierr )
585	!
586	!-- TEST OUTPUT (TO BE REMOVED)
587	WRITE(9,*) TRIM( coupling_mode ), &
588	', ierr after MPI_LOOKUP_NAME: ', ierr
589	CALL LOCAL_FLUSH( 9 )
590
591
592	ENDIF
593
594	ENDIF
595
596	!
597	!-- In case of coupled runs, establish the connection between the atmosphere
598	!-- and the ocean model and define the intercommunicator (comm_inter)
599	CALL MPI_BARRIER( comm2d, ierr )
600	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
601
602	print*, '... before COMM_ACCEPT'
603	CALL MPI_COMM_ACCEPT( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD, &
604	comm_inter, ierr )
605	print*, '--- ierr = ', ierr
606	print*, '--- comm_inter atmosphere = ', comm_inter
607
608	coupling_mode_remote = 'ocean_to_atmosphere'
609
610	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
611
612	IF ( myid == 0 ) PRINT, '** read: ', port_name, ' ierr = ', ierr
613	print*, '... before COMM_CONNECT'
614	CALL MPI_COMM_CONNECT( port_name, MPI_INFO_NULL, 0, MPI_COMM_WORLD, &
615	comm_inter, ierr )
616	print*, '--- ierr = ', ierr
617	print*, '--- comm_inter ocean = ', comm_inter
618
619	coupling_mode_remote = 'atmosphere_to_ocean'
620
621	ENDIF
622
623	!
624	!-- In case of coupled runs, create a new MPI derived datatype for the
625	!-- exchange of surface (xy) data .
626	!-- Gridpoint number for the exchange of ghost points (xy-plane)
627	ngp_xy = ( nxr - nxl + 3 ) * ( nyn - nys + 3 )
628
629	!
630	!-- Define a new MPI derived datatype for the exchange of ghost points in
631	!-- y-direction for 2D-arrays (line)
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	#endif
635
636	#else
637
638	!
639	!-- Array bounds when running on a single PE (respectively a non-parallel
640	!-- machine)
641	nxl = 0
642	nxr = nx
643	nxra = nx
644	nnx = nxr - nxl + 1
645	nys = 0
646	nyn = ny
647	nyna = ny
648	nny = nyn - nys + 1
649	nzb = 0
650	nzt = nz
651	nzta = nz
652	nnz = nz
653
654	ALLOCATE( hor_index_bounds(4,0:0) )
655	hor_index_bounds(1,0) = nxl
656	hor_index_bounds(2,0) = nxr
657	hor_index_bounds(3,0) = nys
658	hor_index_bounds(4,0) = nyn
659
660	!
661	!-- Array bounds for the pressure solver (in the parallel code, these bounds
662	!-- are the ones for the transposed arrays)
663	nys_x = nys
664	nyn_x = nyn
665	nyn_xa = nyn
666	nzb_x = nzb + 1
667	nzt_x = nzt
668	nzt_xa = nzt
669
670	nxl_y = nxl
671	nxr_y = nxr
672	nxr_ya = nxr
673	nzb_y = nzb + 1
674	nzt_y = nzt
675	nzt_ya = nzt
676
677	nxl_z = nxl
678	nxr_z = nxr
679	nxr_za = nxr
680	nys_z = nys
681	nyn_z = nyn
682	nyn_za = nyn
683
684	#endif
685
686	!
687	!-- Calculate number of grid levels necessary for the multigrid poisson solver
688	!-- as well as the gridpoint indices on each level
689	IF ( psolver == 'multigrid' ) THEN
690
691	!
692	!-- First calculate number of possible grid levels for the subdomains
693	mg_levels_x = 1
694	mg_levels_y = 1
695	mg_levels_z = 1
696
697	i = nnx
698	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
699	i = i / 2
700	mg_levels_x = mg_levels_x + 1
701	ENDDO
702
703	j = nny
704	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
705	j = j / 2
706	mg_levels_y = mg_levels_y + 1
707	ENDDO
708
709	k = nz ! do not use nnz because it might be > nz due to transposition
710	! requirements
711	DO WHILE ( MOD( k, 2 ) == 0 .AND. k /= 2 )
712	k = k / 2
713	mg_levels_z = mg_levels_z + 1
714	ENDDO
715
716	maximum_grid_level = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
717
718	!
719	!-- Find out, if the total domain allows more levels. These additional
720	!-- levels are processed on PE0 only.
721	IF ( numprocs > 1 ) THEN
722	IF ( mg_levels_z > MIN( mg_levels_x, mg_levels_y ) ) THEN
723	mg_switch_to_pe0_level_l = maximum_grid_level
724
725	mg_levels_x = 1
726	mg_levels_y = 1
727
728	i = nx+1
729	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
730	i = i / 2
731	mg_levels_x = mg_levels_x + 1
732	ENDDO
733
734	j = ny+1
735	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
736	j = j / 2
737	mg_levels_y = mg_levels_y + 1
738	ENDDO
739
740	maximum_grid_level_l = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
741
742	IF ( maximum_grid_level_l > mg_switch_to_pe0_level_l ) THEN
743	mg_switch_to_pe0_level_l = maximum_grid_level_l - &
744	mg_switch_to_pe0_level_l + 1
745	ELSE
746	mg_switch_to_pe0_level_l = 0
747	ENDIF
748	ELSE
749	mg_switch_to_pe0_level_l = 0
750	maximum_grid_level_l = maximum_grid_level
751	ENDIF
752
753	!
754	!-- Use switch level calculated above only if it is not pre-defined
755	!-- by user
756	IF ( mg_switch_to_pe0_level == 0 ) THEN
757
758	IF ( mg_switch_to_pe0_level_l /= 0 ) THEN
759	mg_switch_to_pe0_level = mg_switch_to_pe0_level_l
760	maximum_grid_level = maximum_grid_level_l
761	ENDIF
762
763	ELSE
764	!
765	!-- Check pre-defined value and reset to default, if neccessary
766	IF ( mg_switch_to_pe0_level < mg_switch_to_pe0_level_l .OR. &
767	mg_switch_to_pe0_level >= maximum_grid_level_l ) THEN
768	IF ( myid == 0 ) THEN
769	PRINT*, '+++ WARNING init_pegrid: mg_switch_to_pe0_level ', &
770	'out of range and reset to default (=0)'
771	ENDIF
772	mg_switch_to_pe0_level = 0
773	ELSE
774	!
775	!-- Use the largest number of possible levels anyway and recalculate
776	!-- the switch level to this largest number of possible values
777	maximum_grid_level = maximum_grid_level_l
778
779	ENDIF
780	ENDIF
781
782	ENDIF
783
784	ALLOCATE( grid_level_count(maximum_grid_level), &
785	nxl_mg(maximum_grid_level), nxr_mg(maximum_grid_level), &
786	nyn_mg(maximum_grid_level), nys_mg(maximum_grid_level), &
787	nzt_mg(maximum_grid_level) )
788
789	grid_level_count = 0
790	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzt_l = nzt
791
792	DO i = maximum_grid_level, 1 , -1
793
794	IF ( i == mg_switch_to_pe0_level ) THEN
795	#if defined( __parallel )
796	!
797	!-- Save the grid size of the subdomain at the switch level, because
798	!-- it is needed in poismg.
799	!-- Array bounds of the local subdomain grids are gathered on PE0
800	ind(1) = nxl_l; ind(2) = nxr_l
801	ind(3) = nys_l; ind(4) = nyn_l
802	ind(5) = nzt_l
803	ALLOCATE( ind_all(5*numprocs), mg_loc_ind(5,0:numprocs-1) )
804	CALL MPI_ALLGATHER( ind, 5, MPI_INTEGER, ind_all, 5, &
805	MPI_INTEGER, comm2d, ierr )
806	DO j = 0, numprocs-1
807	DO k = 1, 5
808	mg_loc_ind(k,j) = ind_all(k+j*5)
809	ENDDO
810	ENDDO
811	DEALLOCATE( ind_all )
812	!
813	!-- Calculate the grid size of the total domain gathered on PE0
814	nxr_l = ( nxr_l-nxl_l+1 ) * pdims(1) - 1
815	nxl_l = 0
816	nyn_l = ( nyn_l-nys_l+1 ) * pdims(2) - 1
817	nys_l = 0
818	!
819	!-- The size of this gathered array must not be larger than the
820	!-- array tend, which is used in the multigrid scheme as a temporary
821	!-- array
822	subdomain_size = ( nxr - nxl + 3 ) * ( nyn - nys + 3 ) * &
823	( nzt - nzb + 2 )
824	gathered_size = ( nxr_l - nxl_l + 3 ) * ( nyn_l - nys_l + 3 ) * &
825	( nzt_l - nzb + 2 )
826
827	IF ( gathered_size > subdomain_size ) THEN
828	IF ( myid == 0 ) THEN
829	PRINT*, '+++ init_pegrid: not enough memory for storing ', &
830	'gathered multigrid data on PE0'
831	ENDIF
832	CALL local_stop
833	ENDIF
834	#else
835	PRINT*, '+++ init_pegrid: multigrid gather/scatter impossible ', &
836	'in non parallel mode'
837	CALL local_stop
838	#endif
839	ENDIF
840
841	nxl_mg(i) = nxl_l
842	nxr_mg(i) = nxr_l
843	nys_mg(i) = nys_l
844	nyn_mg(i) = nyn_l
845	nzt_mg(i) = nzt_l
846
847	nxl_l = nxl_l / 2
848	nxr_l = nxr_l / 2
849	nys_l = nys_l / 2
850	nyn_l = nyn_l / 2
851	nzt_l = nzt_l / 2
852	ENDDO
853
854	ELSE
855
856	maximum_grid_level = 1
857
858	ENDIF
859
860	grid_level = maximum_grid_level
861
862	#if defined( __parallel )
863	!
864	!-- Gridpoint number for the exchange of ghost points (y-line for 2D-arrays)
865	ngp_y = nyn - nys + 1
866
867	!
868	!-- Define a new MPI derived datatype for the exchange of ghost points in
869	!-- y-direction for 2D-arrays (line)
870	CALL MPI_TYPE_VECTOR( nxr-nxl+3, 1, ngp_y+2, MPI_REAL, type_x, ierr )
871	CALL MPI_TYPE_COMMIT( type_x, ierr )
872	CALL MPI_TYPE_VECTOR( nxr-nxl+3, 1, ngp_y+2, MPI_INTEGER, type_x_int, ierr )
873	CALL MPI_TYPE_COMMIT( type_x_int, ierr )
874
875	!
876	!-- Calculate gridpoint numbers for the exchange of ghost points along x
877	!-- (yz-plane for 3D-arrays) and define MPI derived data type(s) for the
878	!-- exchange of ghost points in y-direction (xz-plane).
879	!-- Do these calculations for the model grid and (if necessary) also
880	!-- for the coarser grid levels used in the multigrid method
881	ALLOCATE ( ngp_yz(maximum_grid_level), type_xz(maximum_grid_level) )
882
883	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzb_l = nzb; nzt_l = nzt
884
885	DO i = maximum_grid_level, 1 , -1
886	ngp_yz(i) = (nzt_l - nzb_l + 2) * (nyn_l - nys_l + 3)
887
888	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+3, nzt_l-nzb_l+2, ngp_yz(i), &
889	MPI_REAL, type_xz(i), ierr )
890	CALL MPI_TYPE_COMMIT( type_xz(i), ierr )
891
892	nxl_l = nxl_l / 2
893	nxr_l = nxr_l / 2
894	nys_l = nys_l / 2
895	nyn_l = nyn_l / 2
896	nzt_l = nzt_l / 2
897	ENDDO
898	#endif
899
900	#if defined( __parallel )
901	!
902	!-- Setting of flags for inflow/outflow conditions in case of non-cyclic
903	!-- horizontal boundary conditions.
904	IF ( pleft == MPI_PROC_NULL ) THEN
905	IF ( bc_lr == 'dirichlet/radiation' ) THEN
906	inflow_l = .TRUE.
907	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
908	outflow_l = .TRUE.
909	ENDIF
910	ENDIF
911
912	IF ( pright == MPI_PROC_NULL ) THEN
913	IF ( bc_lr == 'dirichlet/radiation' ) THEN
914	outflow_r = .TRUE.
915	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
916	inflow_r = .TRUE.
917	ENDIF
918	ENDIF
919
920	IF ( psouth == MPI_PROC_NULL ) THEN
921	IF ( bc_ns == 'dirichlet/radiation' ) THEN
922	outflow_s = .TRUE.
923	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
924	inflow_s = .TRUE.
925	ENDIF
926	ENDIF
927
928	IF ( pnorth == MPI_PROC_NULL ) THEN
929	IF ( bc_ns == 'dirichlet/radiation' ) THEN
930	inflow_n = .TRUE.
931	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
932	outflow_n = .TRUE.
933	ENDIF
934	ENDIF
935
936	!
937	!-- Broadcast the id of the inflow PE
938	IF ( inflow_l ) THEN
939	id_inflow_l = myidx
940	ELSE
941	id_inflow_l = 0
942	ENDIF
943	CALL MPI_ALLREDUCE( id_inflow_l, id_inflow, 1, MPI_INTEGER, MPI_SUM, &
944	comm1dx, ierr )
945
946	!
947	!-- Broadcast the id of the recycling plane
948	!-- WARNING: needs to be adjusted in case of inflows other than from left side!
949	IF ( ( recycling_width / dx ) >= nxl .AND. ( recycling_width / dx ) <= nxr ) &
950	THEN
951	id_recycling_l = myidx
952	ELSE
953	id_recycling_l = 0
954	ENDIF
955	CALL MPI_ALLREDUCE( id_recycling_l, id_recycling, 1, MPI_INTEGER, MPI_SUM, &
956	comm1dx, ierr )
957
958	#else
959	IF ( bc_lr == 'dirichlet/radiation' ) THEN
960	inflow_l = .TRUE.
961	outflow_r = .TRUE.
962	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
963	outflow_l = .TRUE.
964	inflow_r = .TRUE.
965	ENDIF
966
967	IF ( bc_ns == 'dirichlet/radiation' ) THEN
968	inflow_n = .TRUE.
969	outflow_s = .TRUE.
970	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
971	outflow_n = .TRUE.
972	inflow_s = .TRUE.
973	ENDIF
974	#endif
975	!
976	!-- At the outflow, u or v, respectively, have to be calculated for one more
977	!-- grid point.
978	IF ( outflow_l ) THEN
979	nxlu = nxl + 1
980	ELSE
981	nxlu = nxl
982	ENDIF
983	IF ( outflow_s ) THEN
984	nysv = nys + 1
985	ELSE
986	nysv = nys
987	ENDIF
988
989	IF ( psolver == 'poisfft_hybrid' ) THEN
990	CALL poisfft_hybrid_ini
991	ELSEIF ( psolver == 'poisfft' ) THEN
992	CALL poisfft_init
993	ENDIF
994
995	!
996	!-- Allocate wall flag arrays used in the multigrid solver
997	IF ( psolver == 'multigrid' ) THEN
998
999	DO i = maximum_grid_level, 1, -1
1000
1001	SELECT CASE ( i )
1002
1003	CASE ( 1 )
1004	ALLOCATE( wall_flags_1(nzb:nzt_mg(i)+1, &
1005	nys_mg(i)-1:nyn_mg(i)+1, &
1006	nxl_mg(i)-1:nxr_mg(i)+1) )
1007
1008	CASE ( 2 )
1009	ALLOCATE( wall_flags_2(nzb:nzt_mg(i)+1, &
1010	nys_mg(i)-1:nyn_mg(i)+1, &
1011	nxl_mg(i)-1:nxr_mg(i)+1) )
1012
1013	CASE ( 3 )
1014	ALLOCATE( wall_flags_3(nzb:nzt_mg(i)+1, &
1015	nys_mg(i)-1:nyn_mg(i)+1, &
1016	nxl_mg(i)-1:nxr_mg(i)+1) )
1017
1018	CASE ( 4 )
1019	ALLOCATE( wall_flags_4(nzb:nzt_mg(i)+1, &
1020	nys_mg(i)-1:nyn_mg(i)+1, &
1021	nxl_mg(i)-1:nxr_mg(i)+1) )
1022
1023	CASE ( 5 )
1024	ALLOCATE( wall_flags_5(nzb:nzt_mg(i)+1, &
1025	nys_mg(i)-1:nyn_mg(i)+1, &
1026	nxl_mg(i)-1:nxr_mg(i)+1) )
1027
1028	CASE ( 6 )
1029	ALLOCATE( wall_flags_6(nzb:nzt_mg(i)+1, &
1030	nys_mg(i)-1:nyn_mg(i)+1, &
1031	nxl_mg(i)-1:nxr_mg(i)+1) )
1032
1033	CASE ( 7 )
1034	ALLOCATE( wall_flags_7(nzb:nzt_mg(i)+1, &
1035	nys_mg(i)-1:nyn_mg(i)+1, &
1036	nxl_mg(i)-1:nxr_mg(i)+1) )
1037
1038	CASE ( 8 )
1039	ALLOCATE( wall_flags_8(nzb:nzt_mg(i)+1, &
1040	nys_mg(i)-1:nyn_mg(i)+1, &
1041	nxl_mg(i)-1:nxr_mg(i)+1) )
1042
1043	CASE ( 9 )
1044	ALLOCATE( wall_flags_9(nzb:nzt_mg(i)+1, &
1045	nys_mg(i)-1:nyn_mg(i)+1, &
1046	nxl_mg(i)-1:nxr_mg(i)+1) )
1047
1048	CASE ( 10 )
1049	ALLOCATE( wall_flags_10(nzb:nzt_mg(i)+1, &
1050	nys_mg(i)-1:nyn_mg(i)+1, &
1051	nxl_mg(i)-1:nxr_mg(i)+1) )
1052
1053	CASE DEFAULT
1054	IF ( myid == 0 ) PRINT*, '+++ init_pegrid: more than 10 ', &
1055	' multigrid levels'
1056	CALL local_stop
1057
1058	END SELECT
1059
1060	ENDDO
1061
1062	ENDIF
1063
1064	END SUBROUTINE init_pegrid

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

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |