Home

Context Navigation

source: palm/trunk/SOURCE/init_pegrid.f90 @ 4185

Last change on this file since 4185 was 4182, checked in by scharf, 5 years ago
corrected "Former revisions" section minor formatting in "Former revisions" section added "Author" section
Property svn:keywords set to `Id`
File size: 49.6 KB

Rev	Line
[1682]	1	!> @file init_pegrid.f90
[2000]	2	!------------------------------------------------------------------------------!
[2696]	3	! This file is part of the PALM model system.
[1036]	4	!
[2000]	5	! PALM is free software: you can redistribute it and/or modify it under the
	6	! terms of the GNU General Public License as published by the Free Software
	7	! Foundation, either version 3 of the License, or (at your option) any later
	8	! version.
[1036]	9	!
	10	! PALM is distributed in the hope that it will be useful, but WITHOUT ANY
	11	! WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
	12	! A PARTICULAR PURPOSE. See the GNU General Public License for more details.
	13	!
	14	! You should have received a copy of the GNU General Public License along with
	15	! PALM. If not, see <http://www.gnu.org/licenses/>.
	16	!
[3655]	17	! Copyright 1997-2019 Leibniz Universitaet Hannover
[2000]	18	!------------------------------------------------------------------------------!
[1036]	19	!
[254]	20	! Current revisions:
[1322]	21	! ------------------
[2201]	22	!
[3553]	23	!
[1321]	24	! Former revisions:
	25	! -----------------
	26	! $Id: init_pegrid.f90 4182 2019-08-22 15:20:23Z oliver.maas $
[4182]	27	! Corrected "Former revisions" section
	28	!
	29	! 4045 2019-06-21 10:58:47Z raasch
[4045]	30	! bugfix: kind attribute added to nint function to allow for large integers which may appear in
	31	! case of default recycling width and small grid spacings
	32	!
	33	! 3999 2019-05-23 16:09:37Z suehring
[3999]	34	! Spend 3 ghost points also in case of pw-scheme when nesting is applied
	35	!
	36	! 3897 2019-04-15 11:51:14Z suehring
[3897]	37	! Minor revision of multigrid check; give warning instead of an abort.
	38	!
	39	! 3890 2019-04-12 15:59:20Z suehring
[3890]	40	! Check if grid coarsening is possible on subdomain, in order to avoid that
	41	! multigrid approach effectively reduces to a Gauss-Seidel scheme.
	42	!
	43	! 3885 2019-04-11 11:29:34Z kanani
[3885]	44	! Changes related to global restructuring of location messages and introduction
	45	! of additional debug messages
	46	!
	47	! 3884 2019-04-10 13:31:55Z Giersch
[3884]	48	! id_recycling is only calculated in case of tubulent inflow
	49	!
	50	! 3761 2019-02-25 15:31:42Z raasch
[3761]	51	! unused variable removed
	52	!
	53	! 3655 2019-01-07 16:51:22Z knoop
[3553]	54	! variables documented
[2259]	55	!
[4182]	56	! Revision 1.1 1997/07/24 11:15:09 raasch
	57	! Initial revision
	58	!
	59	!
[1]	60	! Description:
	61	! ------------
[1682]	62	!> Determination of the virtual processor topology (if not prescribed by the
	63	!> user)and computation of the grid point number and array bounds of the local
	64	!> domains.
[2696]	65	!> @todo: remove MPI-data types for 2D exchange on coarse multigrid level (not
	66	!> used any more)
[1]	67	!------------------------------------------------------------------------------!
[1682]	68	SUBROUTINE init_pegrid
	69
[1]	70
[1320]	71	USE control_parameters, &
[3182]	72	ONLY: bc_dirichlet_l, bc_dirichlet_n, bc_dirichlet_r, bc_dirichlet_s, &
	73	bc_lr, bc_ns, bc_radiation_l, bc_radiation_n, bc_radiation_r, &
[3241]	74	bc_radiation_s, coupling_mode, coupling_topology, gathered_size,&
	75	grid_level, grid_level_count, maximum_grid_level, &
[3761]	76	message_string, mg_switch_to_pe0_level, momentum_advec, &
[3182]	77	psolver, outflow_source_plane, recycling_width, scalar_advec, &
[3884]	78	subdomain_size, turbulent_inflow, turbulent_outflow, y_shift
[1]	79
[1320]	80	USE grid_variables, &
	81	ONLY: dx
	82
	83	USE indices, &
	84	ONLY: mg_loc_ind, nbgp, nnx, nny, nnz, nx, nx_a, nx_o, nxl, nxl_mg, &
	85	nxlu, nxr, nxr_mg, ny, ny_a, ny_o, nyn, nyn_mg, nys, nys_mg, &
	86	nysv, nz, nzb, nzt, nzt_mg, wall_flags_1, wall_flags_2, &
	87	wall_flags_3, wall_flags_4, wall_flags_5, wall_flags_6, &
	88	wall_flags_7, wall_flags_8, wall_flags_9, wall_flags_10
[1]	89
[1320]	90	USE kinds
	91
	92	USE pegrid
[3999]	93
	94	USE pmc_interface, &
	95	ONLY: nested_run
[2238]	96
[1833]	97	USE spectra_mod, &
[1922]	98	ONLY: calculate_spectra, dt_dosp
[1833]	99
[2259]	100	USE synthetic_turbulence_generator_mod, &
[2938]	101	ONLY: id_stg_left, id_stg_north, id_stg_right, id_stg_south, &
	102	use_syn_turb_gen
[2259]	103
[1320]	104	USE transpose_indices, &
	105	ONLY: nxl_y, nxl_yd, nxl_z, nxr_y, nxr_yd, nxr_z, nyn_x, nyn_z, nys_x,&
	106	nys_z, nzb_x, nzb_y, nzb_yd, nzt_x, nzt_yd, nzt_y
[667]	107
[2365]	108	USE vertical_nesting_mod, &
	109	ONLY: vnested, vnest_init_pegrid_domain, vnest_init_pegrid_rank
	110
[1]	111	IMPLICIT NONE
	112
[3552]	113	INTEGER(iwp) :: i !< running index over number of processors or number of multigrid level
	114	INTEGER(iwp) :: id_inflow_l !< ID indicating processors located at the left inflow boundary
[2050]	115	INTEGER(iwp) :: id_outflow_l !< local value of id_outflow
	116	INTEGER(iwp) :: id_outflow_source_l !< local value of id_outflow_source
[3552]	117	INTEGER(iwp) :: id_recycling_l !< ID indicating processors located at the recycling plane
[2938]	118	INTEGER(iwp) :: id_stg_left_l !< left lateral boundary local core id in case of turbulence generator
	119	INTEGER(iwp) :: id_stg_north_l !< north lateral boundary local core id in case of turbulence generator
	120	INTEGER(iwp) :: id_stg_right_l !< right lateral boundary local core id in case of turbulence generator
	121	INTEGER(iwp) :: id_stg_south_l !< south lateral boundary local core id in case of turbulence generator
[3552]	122	INTEGER(iwp) :: ind(5) !< array containing the subdomain bounds
	123	INTEGER(iwp) :: j !< running index, used for various loops
	124	INTEGER(iwp) :: k !< number of vertical grid points in different multigrid level
	125	INTEGER(iwp) :: maximum_grid_level_l !< maximum number of grid level without switching to PE 0
	126	INTEGER(iwp) :: mg_levels_x !< maximum number of grid level allowed along x-direction
	127	INTEGER(iwp) :: mg_levels_y !< maximum number of grid level allowed along y-direction
	128	INTEGER(iwp) :: mg_levels_z !< maximum number of grid level allowed along z-direction
	129	INTEGER(iwp) :: mg_switch_to_pe0_level_l !< maximum number of grid level with switching to PE 0
	130	INTEGER(iwp) :: nnx_y !< quotient of number of grid points along x-direction and number of PEs used along y-direction
	131	INTEGER(iwp) :: nny_x !< quotient of number of grid points along y-direction and number of PEs used along x-direction
	132	INTEGER(iwp) :: nny_z !< quotient of number of grid points along y-direction and number of PEs used along x-direction
	133	INTEGER(iwp) :: nnz_x !< quotient of number of grid points along z-direction and number of PEs used along x-direction
	134	INTEGER(iwp) :: nnz_y !< quotient of number of grid points along z-direction and number of PEs used along x-direction
	135	INTEGER(iwp) :: numproc_sqr !< square root of the number of processors
	136	INTEGER(iwp) :: nxl_l !< lower index bound along x-direction on subdomain and different multigrid level
	137	INTEGER(iwp) :: nxr_l !< upper index bound along x-direction on subdomain and different multigrid level
	138	INTEGER(iwp) :: nyn_l !< lower index bound along y-direction on subdomain and different multigrid level
	139	INTEGER(iwp) :: nys_l !< upper index bound along y-direction on subdomain and different multigrid level
	140	INTEGER(iwp) :: nzb_l !< lower index bound along z-direction on subdomain and different multigrid level
	141	INTEGER(iwp) :: nzt_l !< upper index bound along z-direction on subdomain and different multigrid level
	142	!$ INTEGER(iwp) :: omp_get_num_threads !< number of OpenMP threads
[1]	143
[3552]	144	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: ind_all !< dummy array containing index bounds on subdomain, used for gathering
	145	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nxlf !< lower index bound allong x-direction for every PE
	146	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nxrf !< upper index bound allong x-direction for every PE
	147	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nynf !< lower index bound allong y-direction for every PE
	148	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nysf !< lower index bound allong y-direction for every PE
[1]	149
[3552]	150	INTEGER(iwp), DIMENSION(2) :: pdims_remote !< number of PEs used for coupled model (only in atmospher-ocean coupling)
[2372]	151	INTEGER(iwp) :: lcoord(2) !< PE coordinates of left neighbor along x and y
	152	INTEGER(iwp) :: rcoord(2) !< PE coordinates of right neighbor along x and y
[667]	153
[1]	154	!
	155	!-- Get the number of OpenMP threads
	156	!$OMP PARALLEL
	157	!$ threads_per_task = omp_get_num_threads()
	158	!$OMP END PARALLEL
	159
	160
	161	#if defined( __parallel )
[667]	162
[3885]	163	CALL location_message( 'creating virtual PE grids + MPI derived data types', 'start' )
[1764]	164
[1]	165	!
[2177]	166	!-- Determine the processor topology or check it, if prescribed by the user
[1779]	167	IF ( npex == -1 .AND. npey == -1 ) THEN
[1]	168
	169	!
[2177]	170	!-- Automatic determination of the topology
[1779]	171	numproc_sqr = SQRT( REAL( numprocs, KIND=wp ) )
	172	pdims(1) = MAX( numproc_sqr , 1 )
[2180]	173	DO WHILE ( MOD( numprocs , pdims(1) ) /= 0 )
[1779]	174	pdims(1) = pdims(1) - 1
	175	ENDDO
[2180]	176	pdims(2) = numprocs / pdims(1)
[1]	177
[1779]	178	ELSEIF ( npex /= -1 .AND. npey /= -1 ) THEN
[1]	179
	180	!
[1779]	181	!-- Prescribed by user. Number of processors on the prescribed topology
	182	!-- must be equal to the number of PEs available to the job
	183	IF ( ( npex * npey ) /= numprocs ) THEN
[3045]	184	WRITE( message_string, * ) 'number of PEs of the prescribed ', &
[3046]	185	'topology (', npex*npey,') does not match & the number of ', &
[1779]	186	'PEs available to the job (', numprocs, ')'
	187	CALL message( 'init_pegrid', 'PA0221', 1, 2, 0, 6, 0 )
	188	ENDIF
	189	pdims(1) = npex
	190	pdims(2) = npey
[1]	191
[1779]	192	ELSE
[1]	193	!
[1779]	194	!-- If the processor topology is prescribed by the user, the number of
	195	!-- PEs must be given in both directions
[3045]	196	message_string = 'if the processor topology is prescribed by th' // &
[3046]	197	'e user & both values of "npex" and "npey" must be given' // &
[1779]	198	' in the &NAMELIST-parameter file'
	199	CALL message( 'init_pegrid', 'PA0222', 1, 2, 0, 6, 0 )
[1]	200
	201	ENDIF
	202
	203	!
	204	!-- If necessary, set horizontal boundary conditions to non-cyclic
[722]	205	IF ( bc_lr /= 'cyclic' ) cyclic(1) = .FALSE.
	206	IF ( bc_ns /= 'cyclic' ) cyclic(2) = .FALSE.
[1]	207
[807]	208
[1]	209	!
	210	!-- Create the virtual processor grid
	211	CALL MPI_CART_CREATE( comm_palm, ndim, pdims, cyclic, reorder, &
	212	comm2d, ierr )
	213	CALL MPI_COMM_RANK( comm2d, myid, ierr )
[1468]	214	WRITE (myid_char,'(''_'',I6.6)') myid
[1]	215
	216	CALL MPI_CART_COORDS( comm2d, myid, ndim, pcoord, ierr )
	217	CALL MPI_CART_SHIFT( comm2d, 0, 1, pleft, pright, ierr )
	218	CALL MPI_CART_SHIFT( comm2d, 1, 1, psouth, pnorth, ierr )
	219	!
[2372]	220	!-- In case of cyclic boundary conditions, a y-shift at the boundaries in
	221	!-- x-direction can be introduced via parameter y_shift. The shift is done
	222	!-- by modifying the processor grid in such a way that processors located
	223	!-- at the x-boundary communicate across it to processors with y-coordinate
	224	!-- shifted by y_shift relative to their own. This feature can not be used
	225	!-- in combination with an fft pressure solver. It has been implemented to
	226	!-- counter the effect of streak structures in case of cyclic boundary
	227	!-- conditions. For a description of these see Munters
	228	!-- (2016; dx.doi.org/10.1063/1.4941912)
	229	!--
	230	!-- Get coordinates of left and right neighbor on PE grid
	231	IF ( y_shift /= 0 ) THEN
	232
	233	IF ( bc_lr /= 'cyclic' .OR. bc_ns /= 'cyclic' ) THEN
	234	message_string = 'y_shift /= 0 is only allowed for cyclic ' // &
	235	'boundary conditions in both directions '
	236	CALL message( 'check_parameters', 'PA0467', 1, 2, 0, 6, 0 )
	237	ENDIF
	238	IF ( TRIM( psolver ) /= 'multigrid' .AND. &
	239	TRIM( psolver ) /= 'multigrid_noopt') &
	240	THEN
	241	message_string = 'y_shift /= 0 requires a multigrid pressure solver '
	242	CALL message( 'check_parameters', 'PA0468', 1, 2, 0, 6, 0 )
	243	ENDIF
	244
	245	CALL MPI_CART_COORDS( comm2d, pright, ndim, rcoord, ierr )
	246	CALL MPI_CART_COORDS( comm2d, pleft, ndim, lcoord, ierr )
	247
	248	!
	249	!-- If the x(y)-coordinate of the right (left) neighbor is smaller (greater)
	250	!-- than that of the calling process, then the calling process is located on
	251	!-- the right (left) boundary of the processor grid. In that case,
	252	!-- the y-coordinate of that neighbor is increased (decreased) by y_shift.
	253	!-- The rank of the process with that coordinate is then inquired and the
	254	!-- neighbor rank for MPI_SENDRECV, pright (pleft) is set to it.
	255	!-- In this way, the calling process receives a new right (left) neighbor
	256	!-- for all future MPI_SENDRECV calls. That neighbor has a y-coordinate
	257	!-- of y+(-)y_shift, where y is the original right (left) neighbor's
	258	!-- y-coordinate. The modulo-operation ensures that if the neighbor's
	259	!-- y-coordinate exceeds the grid-boundary, it will be relocated to
	260	!-- the opposite part of the grid cyclicly.
	261	IF ( rcoord(1) < pcoord(1) ) THEN
	262	rcoord(2) = MODULO( rcoord(2) + y_shift, pdims(2) )
	263	CALL MPI_CART_RANK( comm2d, rcoord, pright, ierr )
	264	ENDIF
	265
	266	IF ( lcoord(1) > pcoord(1) ) THEN
	267	lcoord(2) = MODULO( lcoord(2) - y_shift, pdims(2) )
	268	CALL MPI_CART_RANK( comm2d, lcoord, pleft, ierr )
	269	ENDIF
	270	ENDIF
	271	!
[2365]	272	!-- Vertical nesting: store four lists that identify partner ranks to exchange
	273	!-- data
	274	IF ( vnested ) CALL vnest_init_pegrid_rank
	275
	276	!
[1]	277	!-- Determine sub-topologies for transpositions
	278	!-- Transposition from z to x:
	279	remain_dims(1) = .TRUE.
	280	remain_dims(2) = .FALSE.
	281	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dx, ierr )
	282	CALL MPI_COMM_RANK( comm1dx, myidx, ierr )
	283	!
	284	!-- Transposition from x to y
	285	remain_dims(1) = .FALSE.
	286	remain_dims(2) = .TRUE.
	287	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dy, ierr )
	288	CALL MPI_COMM_RANK( comm1dy, myidy, ierr )
	289
	290
	291	!
[1003]	292	!-- Calculate array bounds along x-direction for every PE.
[3045]	293	ALLOCATE( nxlf(0:pdims(1)-1), nxrf(0:pdims(1)-1), nynf(0:pdims(2)-1), &
[1003]	294	nysf(0:pdims(2)-1) )
[1]	295
[1003]	296	IF ( MOD( nx+1 , pdims(1) ) /= 0 ) THEN
[3045]	297	WRITE( message_string, * ) 'x-direction: gridpoint number (',nx+1,') ', &
[3046]	298	'is not an& integral divisor of the number ', &
[3045]	299	'of processors (', pdims(1),')'
[254]	300	CALL message( 'init_pegrid', 'PA0225', 1, 2, 0, 6, 0 )
[1]	301	ELSE
[1003]	302	nnx = ( nx + 1 ) / pdims(1)
[1]	303	ENDIF
	304
	305	!
	306	!-- Left and right array bounds, number of gridpoints
	307	DO i = 0, pdims(1)-1
	308	nxlf(i) = i * nnx
	309	nxrf(i) = ( i + 1 ) * nnx - 1
	310	ENDDO
	311
	312	!
	313	!-- Calculate array bounds in y-direction for every PE.
[1003]	314	IF ( MOD( ny+1 , pdims(2) ) /= 0 ) THEN
[274]	315	WRITE( message_string, * ) 'y-direction: gridpoint number (',ny+1,') ', &
[3046]	316	'is not an& integral divisor of the number of', &
[274]	317	'processors (', pdims(2),')'
[254]	318	CALL message( 'init_pegrid', 'PA0227', 1, 2, 0, 6, 0 )
[1]	319	ELSE
[1003]	320	nny = ( ny + 1 ) / pdims(2)
[1]	321	ENDIF
	322
	323	!
	324	!-- South and north array bounds
	325	DO j = 0, pdims(2)-1
	326	nysf(j) = j * nny
	327	nynf(j) = ( j + 1 ) * nny - 1
	328	ENDDO
	329
	330	!
	331	!-- Local array bounds of the respective PEs
[1003]	332	nxl = nxlf(pcoord(1))
	333	nxr = nxrf(pcoord(1))
	334	nys = nysf(pcoord(2))
	335	nyn = nynf(pcoord(2))
	336	nzb = 0
	337	nzt = nz
	338	nnz = nz
[1]	339
	340	!
[707]	341	!-- Set switches to define if the PE is situated at the border of the virtual
	342	!-- processor grid
	343	IF ( nxl == 0 ) left_border_pe = .TRUE.
	344	IF ( nxr == nx ) right_border_pe = .TRUE.
	345	IF ( nys == 0 ) south_border_pe = .TRUE.
	346	IF ( nyn == ny ) north_border_pe = .TRUE.
	347
	348	!
[1]	349	!-- Calculate array bounds and gridpoint numbers for the transposed arrays
	350	!-- (needed in the pressure solver)
	351	!-- For the transposed arrays, cyclic boundaries as well as top and bottom
	352	!-- boundaries are omitted, because they are obstructive to the transposition
	353
	354	!
	355	!-- 1. transposition z --> x
[1001]	356	!-- This transposition is not neccessary in case of a 1d-decomposition along x
[2938]	357	IF ( psolver == 'poisfft' .OR. calculate_spectra ) THEN
[1304]	358
[1922]	359	IF ( pdims(2) /= 1 ) THEN
	360	IF ( MOD( nz , pdims(1) ) /= 0 ) THEN
	361	WRITE( message_string, * ) 'transposition z --> x:', &
[3046]	362	'& nz=',nz,' is not an integral divisior of pdims(1)=', &
[274]	363	pdims(1)
[1922]	364	CALL message( 'init_pegrid', 'PA0230', 1, 2, 0, 6, 0 )
	365	ENDIF
[1]	366	ENDIF
[1922]	367
	368	nys_x = nys
	369	nyn_x = nyn
	370	nny_x = nny
	371	nnz_x = nz / pdims(1)
	372	nzb_x = 1 + myidx * nnz_x
	373	nzt_x = ( myidx + 1 ) * nnz_x
	374	sendrecvcount_zx = nnx * nny * nnz_x
	375
[1]	376	ENDIF
	377
[1922]	378
	379	IF ( psolver == 'poisfft' ) THEN
[1]	380	!
[1922]	381	!-- 2. transposition x --> y
	382	IF ( MOD( nx+1 , pdims(2) ) /= 0 ) THEN
	383	WRITE( message_string, * ) 'transposition x --> y:', &
[3046]	384	'& nx+1=',nx+1,' is not an integral divisor of ', &
[1922]	385	'pdims(2)=',pdims(2)
	386	CALL message( 'init_pegrid', 'PA0231', 1, 2, 0, 6, 0 )
	387	ENDIF
[1]	388
[1922]	389	nnz_y = nnz_x
	390	nzb_y = nzb_x
	391	nzt_y = nzt_x
	392	nnx_y = (nx+1) / pdims(2)
	393	nxl_y = myidy * nnx_y
	394	nxr_y = ( myidy + 1 ) * nnx_y - 1
	395	sendrecvcount_xy = nnx_y * nny_x * nnz_y
[1]	396	!
[1922]	397	!-- 3. transposition y --> z
	398	!-- (ELSE: x --> y in case of 1D-decomposition along x)
	399	nxl_z = nxl_y
	400	nxr_z = nxr_y
	401	nny_z = (ny+1) / pdims(1)
	402	nys_z = myidx * nny_z
	403	nyn_z = ( myidx + 1 ) * nny_z - 1
	404	sendrecvcount_yz = nnx_y * nny_z * nnz_y
[1304]	405
[1922]	406	IF ( pdims(2) /= 1 ) THEN
[1]	407	!
[1922]	408	!-- y --> z
	409	!-- This transposition is not neccessary in case of a 1d-decomposition
	410	!-- along x, except that the uptream-spline method is switched on
	411	IF ( MOD( ny+1 , pdims(1) ) /= 0 ) THEN
	412	WRITE( message_string, * ) 'transposition y --> z:', &
[3046]	413	'& ny+1=',ny+1,' is not an integral divisor of',&
[1922]	414	' pdims(1)=',pdims(1)
	415	CALL message( 'init_pegrid', 'PA0232', 1, 2, 0, 6, 0 )
	416	ENDIF
[1]	417
[1922]	418	ELSE
[1]	419	!
[1922]	420	!-- x --> y
	421	!-- This condition must be fulfilled for a 1D-decomposition along x
	422	IF ( MOD( ny+1 , pdims(1) ) /= 0 ) THEN
	423	WRITE( message_string, * ) 'transposition x --> y:', &
[3046]	424	'& ny+1=',ny+1,' is not an integral divisor of',&
[1922]	425	' pdims(1)=',pdims(1)
	426	CALL message( 'init_pegrid', 'PA0233', 1, 2, 0, 6, 0 )
	427	ENDIF
	428
[1]	429	ENDIF
	430
	431	ENDIF
	432
	433	!
	434	!-- Indices for direct transpositions z --> y (used for calculating spectra)
[1922]	435	IF ( calculate_spectra ) THEN
[1003]	436	IF ( MOD( nz, pdims(2) ) /= 0 ) THEN
[1922]	437	WRITE( message_string, * ) 'direct transposition z --> y (needed ', &
[3045]	438	'for spectra): nz=',nz,' is not an integral divisor of ', &
[274]	439	'pdims(2)=',pdims(2)
[254]	440	CALL message( 'init_pegrid', 'PA0234', 1, 2, 0, 6, 0 )
[1]	441	ELSE
[1003]	442	nxl_yd = nxl
	443	nxr_yd = nxr
	444	nzb_yd = 1 + myidy * ( nz / pdims(2) )
	445	nzt_yd = ( myidy + 1 ) * ( nz / pdims(2) )
	446	sendrecvcount_zyd = nnx * nny * ( nz / pdims(2) )
[1]	447	ENDIF
	448	ENDIF
	449
[1922]	450	IF ( psolver == 'poisfft' .OR. calculate_spectra ) THEN
[1]	451	!
[1922]	452	!-- Indices for direct transpositions y --> x
	453	!-- (they are only possible in case of a 1d-decomposition along x)
	454	IF ( pdims(2) == 1 ) THEN
	455	nny_x = nny / pdims(1)
	456	nys_x = myid * nny_x
	457	nyn_x = ( myid + 1 ) * nny_x - 1
	458	nzb_x = 1
	459	nzt_x = nz
	460	sendrecvcount_xy = nnx * nny_x * nz
	461	ENDIF
	462
[1]	463	ENDIF
	464
[1922]	465	IF ( psolver == 'poisfft' ) THEN
[1]	466	!
[1922]	467	!-- Indices for direct transpositions x --> y
	468	!-- (they are only possible in case of a 1d-decomposition along y)
	469	IF ( pdims(1) == 1 ) THEN
	470	nnx_y = nnx / pdims(2)
	471	nxl_y = myid * nnx_y
	472	nxr_y = ( myid + 1 ) * nnx_y - 1
	473	nzb_y = 1
	474	nzt_y = nz
	475	sendrecvcount_xy = nnx_y * nny * nz
	476	ENDIF
	477
[1]	478	ENDIF
	479
	480	!
	481	!-- Arrays for storing the array bounds are needed any more
	482	DEALLOCATE( nxlf , nxrf , nynf , nysf )
	483
[807]	484
[145]	485	!
	486	!-- Collect index bounds from other PEs (to be written to restart file later)
	487	ALLOCATE( hor_index_bounds(4,0:numprocs-1) )
	488
	489	IF ( myid == 0 ) THEN
	490
	491	hor_index_bounds(1,0) = nxl
	492	hor_index_bounds(2,0) = nxr
	493	hor_index_bounds(3,0) = nys
	494	hor_index_bounds(4,0) = nyn
	495
	496	!
	497	!-- Receive data from all other PEs
	498	DO i = 1, numprocs-1
	499	CALL MPI_RECV( ibuf, 4, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, &
	500	ierr )
	501	hor_index_bounds(:,i) = ibuf(1:4)
	502	ENDDO
	503
	504	ELSE
	505	!
	506	!-- Send index bounds to PE0
	507	ibuf(1) = nxl
	508	ibuf(2) = nxr
	509	ibuf(3) = nys
	510	ibuf(4) = nyn
	511	CALL MPI_SEND( ibuf, 4, MPI_INTEGER, 0, myid, comm2d, ierr )
	512
	513	ENDIF
	514
[807]	515
[1]	516	#if defined( __print )
	517	!
	518	!-- Control output
	519	IF ( myid == 0 ) THEN
	520	PRINT, ' processor topology *'
	521	PRINT*, ' '
	522	PRINT*, 'myid pcoord left right south north idx idy nxl: nxr',&
	523	&' nys: nyn'
	524	PRINT*, '------------------------------------------------------------',&
	525	&'-----------'
	526	WRITE (*,1000) 0, pcoord(1), pcoord(2), pleft, pright, psouth, pnorth, &
	527	myidx, myidy, nxl, nxr, nys, nyn
	528	1000 FORMAT (I4,2X,'(',I3,',',I3,')',3X,I4,2X,I4,3X,I4,2X,I4,2X,I3,1X,I3, &
	529	2(2X,I4,':',I4))
	530
	531	!
[108]	532	!-- Receive data from the other PEs
[1]	533	DO i = 1,numprocs-1
	534	CALL MPI_RECV( ibuf, 12, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, &
	535	ierr )
	536	WRITE (*,1000) i, ( ibuf(j) , j = 1,12 )
	537	ENDDO
	538	ELSE
	539
	540	!
	541	!-- Send data to PE0
	542	ibuf(1) = pcoord(1); ibuf(2) = pcoord(2); ibuf(3) = pleft
	543	ibuf(4) = pright; ibuf(5) = psouth; ibuf(6) = pnorth; ibuf(7) = myidx
	544	ibuf(8) = myidy; ibuf(9) = nxl; ibuf(10) = nxr; ibuf(11) = nys
	545	ibuf(12) = nyn
	546	CALL MPI_SEND( ibuf, 12, MPI_INTEGER, 0, myid, comm2d, ierr )
	547	ENDIF
	548	#endif
	549
[667]	550	!
[709]	551	!-- Determine the number of ghost point layers
[3347]	552	IF ( scalar_advec == 'ws-scheme' .OR. &
[3999]	553	momentum_advec == 'ws-scheme' .OR. nested_run ) THEN
[667]	554	nbgp = 3
	555	ELSE
	556	nbgp = 1
[709]	557	ENDIF
[667]	558
[102]	559	!
[709]	560	!-- Create a new MPI derived datatype for the exchange of surface (xy) data,
	561	!-- which is needed for coupled atmosphere-ocean runs.
	562	!-- First, calculate number of grid points of an xy-plane.
[667]	563	ngp_xy = ( nxr - nxl + 1 + 2 * nbgp ) * ( nyn - nys + 1 + 2 * nbgp )
[102]	564	CALL MPI_TYPE_VECTOR( ngp_xy, 1, nzt-nzb+2, MPI_REAL, type_xy, ierr )
	565	CALL MPI_TYPE_COMMIT( type_xy, ierr )
[667]	566
[2365]	567	IF ( TRIM( coupling_mode ) /= 'uncoupled' .AND. .NOT. vnested ) THEN
[667]	568
	569	!
	570	!-- Pass the number of grid points of the atmosphere model to
	571	!-- the ocean model and vice versa
	572	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
	573
	574	nx_a = nx
	575	ny_a = ny
	576
[709]	577	IF ( myid == 0 ) THEN
	578
	579	CALL MPI_SEND( nx_a, 1, MPI_INTEGER, numprocs, 1, comm_inter, &
	580	ierr )
	581	CALL MPI_SEND( ny_a, 1, MPI_INTEGER, numprocs, 2, comm_inter, &
	582	ierr )
	583	CALL MPI_SEND( pdims, 2, MPI_INTEGER, numprocs, 3, comm_inter, &
	584	ierr )
	585	CALL MPI_RECV( nx_o, 1, MPI_INTEGER, numprocs, 4, comm_inter, &
	586	status, ierr )
	587	CALL MPI_RECV( ny_o, 1, MPI_INTEGER, numprocs, 5, comm_inter, &
	588	status, ierr )
	589	CALL MPI_RECV( pdims_remote, 2, MPI_INTEGER, numprocs, 6, &
[667]	590	comm_inter, status, ierr )
	591	ENDIF
	592
[709]	593	CALL MPI_BCAST( nx_o, 1, MPI_INTEGER, 0, comm2d, ierr )
	594	CALL MPI_BCAST( ny_o, 1, MPI_INTEGER, 0, comm2d, ierr )
	595	CALL MPI_BCAST( pdims_remote, 2, MPI_INTEGER, 0, comm2d, ierr )
[667]	596
	597	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
	598
	599	nx_o = nx
	600	ny_o = ny
	601
	602	IF ( myid == 0 ) THEN
[709]	603
	604	CALL MPI_RECV( nx_a, 1, MPI_INTEGER, 0, 1, comm_inter, status, &
	605	ierr )
	606	CALL MPI_RECV( ny_a, 1, MPI_INTEGER, 0, 2, comm_inter, status, &
	607	ierr )
	608	CALL MPI_RECV( pdims_remote, 2, MPI_INTEGER, 0, 3, comm_inter, &
	609	status, ierr )
	610	CALL MPI_SEND( nx_o, 1, MPI_INTEGER, 0, 4, comm_inter, ierr )
	611	CALL MPI_SEND( ny_o, 1, MPI_INTEGER, 0, 5, comm_inter, ierr )
	612	CALL MPI_SEND( pdims, 2, MPI_INTEGER, 0, 6, comm_inter, ierr )
[667]	613	ENDIF
	614
	615	CALL MPI_BCAST( nx_a, 1, MPI_INTEGER, 0, comm2d, ierr)
	616	CALL MPI_BCAST( ny_a, 1, MPI_INTEGER, 0, comm2d, ierr)
	617	CALL MPI_BCAST( pdims_remote, 2, MPI_INTEGER, 0, comm2d, ierr)
	618
	619	ENDIF
	620
[709]	621	ngp_a = ( nx_a+1 + 2 * nbgp ) * ( ny_a+1 + 2 * nbgp )
	622	ngp_o = ( nx_o+1 + 2 * nbgp ) * ( ny_o+1 + 2 * nbgp )
[667]	623
	624	!
[709]	625	!-- Determine if the horizontal grid and the number of PEs in ocean and
	626	!-- atmosphere is same or not
	627	IF ( nx_o == nx_a .AND. ny_o == ny_a .AND. &
[667]	628	pdims(1) == pdims_remote(1) .AND. pdims(2) == pdims_remote(2) ) &
	629	THEN
	630	coupling_topology = 0
	631	ELSE
	632	coupling_topology = 1
	633	ENDIF
	634
	635	!
	636	!-- Determine the target PEs for the exchange between ocean and
	637	!-- atmosphere (comm2d)
[709]	638	IF ( coupling_topology == 0 ) THEN
	639	!
	640	!-- In case of identical topologies, every atmosphere PE has exactly one
	641	!-- ocean PE counterpart and vice versa
	642	IF ( TRIM( coupling_mode ) == 'atmosphere_to_ocean' ) THEN
[667]	643	target_id = myid + numprocs
	644	ELSE
	645	target_id = myid
	646	ENDIF
	647
	648	ELSE
	649	!
	650	!-- In case of nonequivalent topology in ocean and atmosphere only for
	651	!-- PE0 in ocean and PE0 in atmosphere a target_id is needed, since
[709]	652	!-- data echxchange between ocean and atmosphere will be done only
	653	!-- between these PEs.
	654	IF ( myid == 0 ) THEN
	655
	656	IF ( TRIM( coupling_mode ) == 'atmosphere_to_ocean' ) THEN
[667]	657	target_id = numprocs
	658	ELSE
	659	target_id = 0
	660	ENDIF
[709]	661
[667]	662	ENDIF
[709]	663
[667]	664	ENDIF
	665
	666	ENDIF
	667
[2365]	668	!
	669	!-- Store partner grid point co-ordinates as lists.
	670	!-- Create custom MPI vector datatypes for contiguous data transfer
	671	IF ( vnested ) CALL vnest_init_pegrid_domain
[667]	672
[1]	673	#else
	674
	675	!
	676	!-- Array bounds when running on a single PE (respectively a non-parallel
	677	!-- machine)
[1003]	678	nxl = 0
	679	nxr = nx
	680	nnx = nxr - nxl + 1
	681	nys = 0
	682	nyn = ny
	683	nny = nyn - nys + 1
	684	nzb = 0
	685	nzt = nz
	686	nnz = nz
[1]	687
[145]	688	ALLOCATE( hor_index_bounds(4,0:0) )
	689	hor_index_bounds(1,0) = nxl
	690	hor_index_bounds(2,0) = nxr
	691	hor_index_bounds(3,0) = nys
	692	hor_index_bounds(4,0) = nyn
	693
[1]	694	!
	695	!-- Array bounds for the pressure solver (in the parallel code, these bounds
	696	!-- are the ones for the transposed arrays)
[1003]	697	nys_x = nys
	698	nyn_x = nyn
	699	nzb_x = nzb + 1
	700	nzt_x = nzt
[1]	701
[1003]	702	nxl_y = nxl
	703	nxr_y = nxr
	704	nzb_y = nzb + 1
	705	nzt_y = nzt
[1]	706
[1003]	707	nxl_z = nxl
	708	nxr_z = nxr
	709	nys_z = nys
	710	nyn_z = nyn
[1]	711
	712	#endif
	713
	714	!
	715	!-- Calculate number of grid levels necessary for the multigrid poisson solver
	716	!-- as well as the gridpoint indices on each level
[1575]	717	IF ( psolver(1:9) == 'multigrid' ) THEN
[1]	718
	719	!
	720	!-- First calculate number of possible grid levels for the subdomains
	721	mg_levels_x = 1
	722	mg_levels_y = 1
	723	mg_levels_z = 1
	724
	725	i = nnx
	726	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
	727	i = i / 2
	728	mg_levels_x = mg_levels_x + 1
	729	ENDDO
	730
	731	j = nny
	732	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
	733	j = j / 2
	734	mg_levels_y = mg_levels_y + 1
	735	ENDDO
	736
[181]	737	k = nz ! do not use nnz because it might be > nz due to transposition
	738	! requirements
[1]	739	DO WHILE ( MOD( k, 2 ) == 0 .AND. k /= 2 )
	740	k = k / 2
	741	mg_levels_z = mg_levels_z + 1
	742	ENDDO
[2197]	743	!
	744	!-- The optimized MG-solver does not allow odd values for nz at the coarsest
	745	!-- grid level
	746	IF ( TRIM( psolver ) /= 'multigrid_noopt' ) THEN
	747	IF ( MOD( k, 2 ) /= 0 ) mg_levels_z = mg_levels_z - 1
[3057]	748	!
	749	!-- An odd value of nz does not work. The finest level must have an even
	750	!-- value.
	751	IF ( mg_levels_z == 0 ) THEN
	752	message_string = 'optimized multigrid method requires nz to be even'
[3058]	753	CALL message( 'init_pegrid', 'PA0495', 1, 2, 0, 6, 0 )
[3057]	754	ENDIF
[2197]	755	ENDIF
[1]	756
	757	maximum_grid_level = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
[3890]	758	!
	759	!-- Check if subdomain sizes prevents any coarsening.
	760	!-- This case, the maximum number of grid levels is 1, i.e. effectively
	761	!-- a Gauss-Seidel scheme is applied rather than a multigrid approach.
[3897]	762	!-- Give a warning in this case.
	763	IF ( maximum_grid_level == 1 .AND. mg_switch_to_pe0_level == -1 ) THEN
	764	message_string = 'No grid coarsening possible, multigrid ' // &
	765	'approach effectively reduces to a Gauss-Seidel ' //&
	766	'scheme.'
	767
	768	CALL message( 'poismg', 'PA0648', 0, 1, 0, 6, 0 )
[3890]	769	ENDIF
[1]	770
	771	!
	772	!-- Find out, if the total domain allows more levels. These additional
[709]	773	!-- levels are identically processed on all PEs.
[197]	774	IF ( numprocs > 1 .AND. mg_switch_to_pe0_level /= -1 ) THEN
[709]	775
[1]	776	IF ( mg_levels_z > MIN( mg_levels_x, mg_levels_y ) ) THEN
[709]	777
[1]	778	mg_switch_to_pe0_level_l = maximum_grid_level
	779
	780	mg_levels_x = 1
	781	mg_levels_y = 1
	782
	783	i = nx+1
	784	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
	785	i = i / 2
	786	mg_levels_x = mg_levels_x + 1
	787	ENDDO
	788
	789	j = ny+1
	790	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
	791	j = j / 2
	792	mg_levels_y = mg_levels_y + 1
	793	ENDDO
	794
	795	maximum_grid_level_l = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
	796
	797	IF ( maximum_grid_level_l > mg_switch_to_pe0_level_l ) THEN
	798	mg_switch_to_pe0_level_l = maximum_grid_level_l - &
	799	mg_switch_to_pe0_level_l + 1
	800	ELSE
	801	mg_switch_to_pe0_level_l = 0
	802	ENDIF
[709]	803
[1]	804	ELSE
[3057]	805
[1]	806	mg_switch_to_pe0_level_l = 0
	807	maximum_grid_level_l = maximum_grid_level
[709]	808
[1]	809	ENDIF
	810
	811	!
	812	!-- Use switch level calculated above only if it is not pre-defined
	813	!-- by user
	814	IF ( mg_switch_to_pe0_level == 0 ) THEN
	815	IF ( mg_switch_to_pe0_level_l /= 0 ) THEN
	816	mg_switch_to_pe0_level = mg_switch_to_pe0_level_l
	817	maximum_grid_level = maximum_grid_level_l
	818	ENDIF
	819
	820	ELSE
	821	!
	822	!-- Check pre-defined value and reset to default, if neccessary
[3045]	823	IF ( mg_switch_to_pe0_level < mg_switch_to_pe0_level_l .OR. &
[1]	824	mg_switch_to_pe0_level >= maximum_grid_level_l ) THEN
[3045]	825	message_string = 'mg_switch_to_pe0_level ' // &
[2271]	826	'out of range and reset to 0'
[254]	827	CALL message( 'init_pegrid', 'PA0235', 0, 1, 0, 6, 0 )
[1]	828	mg_switch_to_pe0_level = 0
	829	ELSE
	830	!
	831	!-- Use the largest number of possible levels anyway and recalculate
	832	!-- the switch level to this largest number of possible values
	833	maximum_grid_level = maximum_grid_level_l
	834
	835	ENDIF
[709]	836
[1]	837	ENDIF
	838
	839	ENDIF
	840
[1056]	841	ALLOCATE( grid_level_count(maximum_grid_level), &
	842	nxl_mg(0:maximum_grid_level), nxr_mg(0:maximum_grid_level), &
	843	nyn_mg(0:maximum_grid_level), nys_mg(0:maximum_grid_level), &
	844	nzt_mg(0:maximum_grid_level) )
[1]	845
	846	grid_level_count = 0
[1056]	847	!
	848	!-- Index zero required as dummy due to definition of arrays f2 and p2 in
	849	!-- recursive subroutine next_mg_level
	850	nxl_mg(0) = 0; nxr_mg(0) = 0; nyn_mg(0) = 0; nys_mg(0) = 0; nzt_mg(0) = 0
[778]	851
[1]	852	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzt_l = nzt
	853
	854	DO i = maximum_grid_level, 1 , -1
	855
	856	IF ( i == mg_switch_to_pe0_level ) THEN
[1804]	857	#if defined( __parallel )
[1]	858	!
	859	!-- Save the grid size of the subdomain at the switch level, because
	860	!-- it is needed in poismg.
	861	ind(1) = nxl_l; ind(2) = nxr_l
	862	ind(3) = nys_l; ind(4) = nyn_l
	863	ind(5) = nzt_l
	864	ALLOCATE( ind_all(5*numprocs), mg_loc_ind(5,0:numprocs-1) )
	865	CALL MPI_ALLGATHER( ind, 5, MPI_INTEGER, ind_all, 5, &
	866	MPI_INTEGER, comm2d, ierr )
	867	DO j = 0, numprocs-1
	868	DO k = 1, 5
	869	mg_loc_ind(k,j) = ind_all(k+j*5)
	870	ENDDO
	871	ENDDO
	872	DEALLOCATE( ind_all )
	873	!
[709]	874	!-- Calculate the grid size of the total domain
[1]	875	nxr_l = ( nxr_l-nxl_l+1 ) * pdims(1) - 1
	876	nxl_l = 0
	877	nyn_l = ( nyn_l-nys_l+1 ) * pdims(2) - 1
	878	nys_l = 0
	879	!
	880	!-- The size of this gathered array must not be larger than the
	881	!-- array tend, which is used in the multigrid scheme as a temporary
[778]	882	!-- array. Therefore the subdomain size of an PE is calculated and
	883	!-- the size of the gathered grid. These values are used in
	884	!-- routines pres and poismg
	885	subdomain_size = ( nxr - nxl + 2 * nbgp + 1 ) * &
	886	( nyn - nys + 2 * nbgp + 1 ) * ( nzt - nzb + 2 )
[3045]	887	gathered_size = ( nxr_l - nxl_l + 3 ) * ( nyn_l - nys_l + 3 ) * &
[1]	888	( nzt_l - nzb + 2 )
	889
[1804]	890	#else
[3045]	891	message_string = 'multigrid gather/scatter impossible ' // &
[1]	892	'in non parallel mode'
[254]	893	CALL message( 'init_pegrid', 'PA0237', 1, 2, 0, 6, 0 )
[1]	894	#endif
	895	ENDIF
	896
	897	nxl_mg(i) = nxl_l
	898	nxr_mg(i) = nxr_l
	899	nys_mg(i) = nys_l
	900	nyn_mg(i) = nyn_l
	901	nzt_mg(i) = nzt_l
	902
	903	nxl_l = nxl_l / 2
	904	nxr_l = nxr_l / 2
	905	nys_l = nys_l / 2
	906	nyn_l = nyn_l / 2
	907	nzt_l = nzt_l / 2
[3890]	908
[1]	909	ENDDO
	910
[780]	911	!
[3045]	912	!-- Temporary problem: Currently calculation of maxerror in routine poismg crashes
[780]	913	!-- if grid data are collected on PE0 already on the finest grid level.
	914	!-- To be solved later.
	915	IF ( maximum_grid_level == mg_switch_to_pe0_level ) THEN
	916	message_string = 'grid coarsening on subdomain level cannot be performed'
	917	CALL message( 'poismg', 'PA0236', 1, 2, 0, 6, 0 )
	918	ENDIF
	919
[1]	920	ELSE
	921
[667]	922	maximum_grid_level = 0
[1]	923
	924	ENDIF
	925
[722]	926	!
	927	!-- Default level 0 tells exchange_horiz that all ghost planes have to be
	928	!-- exchanged. grid_level is adjusted in poismg, where only one ghost plane
	929	!-- is required.
	930	grid_level = 0
[1]	931
[1804]	932	#if defined( __parallel )
[1]	933	!
	934	!-- Gridpoint number for the exchange of ghost points (y-line for 2D-arrays)
[667]	935	ngp_y = nyn - nys + 1 + 2 * nbgp
[1]	936
	937	!
[709]	938	!-- Define new MPI derived datatypes for the exchange of ghost points in
	939	!-- x- and y-direction for 2D-arrays (line)
[1968]	940	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_REAL, type_x, &
[709]	941	ierr )
[1]	942	CALL MPI_TYPE_COMMIT( type_x, ierr )
	943
[667]	944	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_REAL, type_y, ierr )
	945	CALL MPI_TYPE_COMMIT( type_y, ierr )
[1968]	946	!
	947	!-- Define new MPI derived datatypes for the exchange of ghost points in
[3542]	948	!-- x- and y-direction for 2D-INTEGER arrays (line) - on normal grid.
	949	!-- Define types for 32-bit and 8-bit Integer. The 8-bit Integer are only
	950	!-- required on normal grid, while 32-bit Integer may be also required on
	951	!-- coarser grid level in case of multigrid solver.
	952	!
	953	!-- 8-bit Integer
	954	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_BYTE, &
	955	type_x_byte, ierr )
	956	CALL MPI_TYPE_COMMIT( type_x_byte, ierr )
	957
	958	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_BYTE, &
	959	type_y_byte, ierr )
	960	CALL MPI_TYPE_COMMIT( type_y_byte, ierr )
	961	!
	962	!-- 32-bit Integer
[1968]	963	ALLOCATE( type_x_int(0:maximum_grid_level), &
	964	type_y_int(0:maximum_grid_level) )
[3542]	965
[1968]	966	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_INTEGER, &
	967	type_x_int(0), ierr )
	968	CALL MPI_TYPE_COMMIT( type_x_int(0), ierr )
[667]	969
[1968]	970	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_INTEGER, type_y_int(0), ierr )
	971	CALL MPI_TYPE_COMMIT( type_y_int(0), ierr )
[1]	972	!
	973	!-- Calculate gridpoint numbers for the exchange of ghost points along x
	974	!-- (yz-plane for 3D-arrays) and define MPI derived data type(s) for the
	975	!-- exchange of ghost points in y-direction (xz-plane).
	976	!-- Do these calculations for the model grid and (if necessary) also
	977	!-- for the coarser grid levels used in the multigrid method
[2696]	978	ALLOCATE ( ngp_xz(0:maximum_grid_level), &
	979	ngp_xz_int(0:maximum_grid_level), &
	980	ngp_yz(0:maximum_grid_level), &
	981	ngp_yz_int(0:maximum_grid_level), &
	982	type_xz(0:maximum_grid_level), &
	983	type_xz_int(0:maximum_grid_level), &
	984	type_yz(0:maximum_grid_level), &
	985	type_yz_int(0:maximum_grid_level) )
[1]	986
	987	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzb_l = nzb; nzt_l = nzt
[709]	988
[667]	989	!
	990	!-- Discern between the model grid, which needs nbgp ghost points and
	991	!-- grid levels for the multigrid scheme. In the latter case only one
	992	!-- ghost point is necessary.
[709]	993	!-- First definition of MPI-datatypes for exchange of ghost layers on normal
[667]	994	!-- grid. The following loop is needed for data exchange in poismg.f90.
	995	!
	996	!-- Determine number of grid points of yz-layer for exchange
	997	ngp_yz(0) = (nzt - nzb + 2) * (nyn - nys + 1 + 2 * nbgp)
[709]	998
[667]	999	!
[709]	1000	!-- Define an MPI-datatype for the exchange of left/right boundaries.
	1001	!-- Although data are contiguous in physical memory (which does not
	1002	!-- necessarily require an MPI-derived datatype), the data exchange between
	1003	!-- left and right PE's using the MPI-derived type is 10% faster than without.
[667]	1004	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2nbgp, nbgp(nzt-nzb+2), ngp_yz(0), &
[709]	1005	MPI_REAL, type_xz(0), ierr )
[667]	1006	CALL MPI_TYPE_COMMIT( type_xz(0), ierr )
[1]	1007
[709]	1008	CALL MPI_TYPE_VECTOR( nbgp, ngp_yz(0), ngp_yz(0), MPI_REAL, type_yz(0), &
	1009	ierr )
[667]	1010	CALL MPI_TYPE_COMMIT( type_yz(0), ierr )
[709]	1011
[667]	1012	!
[2696]	1013	!-- Define data types for exchange of 3D Integer arrays.
	1014	ngp_yz_int(0) = (nzt - nzb + 2) * (nyn - nys + 1 + 2 * nbgp)
	1015
	1016	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2nbgp, nbgp(nzt-nzb+2), ngp_yz_int(0), &
	1017	MPI_INTEGER, type_xz_int(0), ierr )
	1018	CALL MPI_TYPE_COMMIT( type_xz_int(0), ierr )
	1019
	1020	CALL MPI_TYPE_VECTOR( nbgp, ngp_yz_int(0), ngp_yz_int(0), MPI_INTEGER, &
	1021	type_yz_int(0), ierr )
	1022	CALL MPI_TYPE_COMMIT( type_yz_int(0), ierr )
	1023
	1024	!
[709]	1025	!-- Definition of MPI-datatypes for multigrid method (coarser level grids)
[1575]	1026	IF ( psolver(1:9) == 'multigrid' ) THEN
[667]	1027	!
[709]	1028	!-- Definition of MPI-datatyoe as above, but only 1 ghost level is used
	1029	DO i = maximum_grid_level, 1 , -1
[1968]	1030	!
[2696]	1031	!-- For 3D-exchange on different multigrid level, one ghost point for
	1032	!-- REAL arrays, two ghost points for INTEGER arrays
[1575]	1033	ngp_xz(i) = (nzt_l - nzb_l + 2) * (nxr_l - nxl_l + 3)
[667]	1034	ngp_yz(i) = (nzt_l - nzb_l + 2) * (nyn_l - nys_l + 3)
	1035
[2696]	1036	ngp_xz_int(i) = (nzt_l - nzb_l + 2) * (nxr_l - nxl_l + 3)
	1037	ngp_yz_int(i) = (nzt_l - nzb_l + 2) * (nyn_l - nys_l + 3)
	1038	!
	1039	!-- MPI data type for REAL arrays, for xz-layers
	1040	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+3, nzt_l-nzb_l+2, ngp_yz(i), &
[709]	1041	MPI_REAL, type_xz(i), ierr )
[667]	1042	CALL MPI_TYPE_COMMIT( type_xz(i), ierr )
[1]	1043
[2696]	1044	!
	1045	!-- MPI data type for INTEGER arrays, for xz-layers
	1046	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+3, nzt_l-nzb_l+2, ngp_yz_int(i), &
	1047	MPI_INTEGER, type_xz_int(i), ierr )
	1048	CALL MPI_TYPE_COMMIT( type_xz_int(i), ierr )
	1049
	1050	!
	1051	!-- MPI data type for REAL arrays, for yz-layers
[709]	1052	CALL MPI_TYPE_VECTOR( 1, ngp_yz(i), ngp_yz(i), MPI_REAL, type_yz(i), &
	1053	ierr )
[667]	1054	CALL MPI_TYPE_COMMIT( type_yz(i), ierr )
[2696]	1055	!
	1056	!-- MPI data type for INTEGER arrays, for yz-layers
	1057	CALL MPI_TYPE_VECTOR( 1, ngp_yz_int(i), ngp_yz_int(i), MPI_INTEGER, &
	1058	type_yz_int(i), ierr )
	1059	CALL MPI_TYPE_COMMIT( type_yz_int(i), ierr )
[667]	1060
[1968]	1061
	1062	!-- For 2D-exchange of INTEGER arrays on coarser grid level, where 2 ghost
[3542]	1063	!-- points need to be exchanged. Only required for 32-bit Integer arrays.
	1064	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+5, 2, nyn_l-nys_l+5, MPI_INTEGER, &
[1968]	1065	type_x_int(i), ierr )
	1066	CALL MPI_TYPE_COMMIT( type_x_int(i), ierr )
	1067
	1068
[3542]	1069	CALL MPI_TYPE_VECTOR( 2, nyn_l-nys_l+5, nyn_l-nys_l+5, MPI_INTEGER, &
[1968]	1070	type_y_int(i), ierr )
	1071	CALL MPI_TYPE_COMMIT( type_y_int(i), ierr )
	1072
[667]	1073	nxl_l = nxl_l / 2
	1074	nxr_l = nxr_l / 2
	1075	nys_l = nys_l / 2
	1076	nyn_l = nyn_l / 2
	1077	nzt_l = nzt_l / 2
[709]	1078
[667]	1079	ENDDO
[709]	1080
	1081	ENDIF
[1677]	1082
[1]	1083	#endif
	1084
[1804]	1085	#if defined( __parallel )
[1]	1086	!
[1933]	1087	!-- Setting of flags for inflow/outflow/nesting conditions.
[1964]	1088	IF ( pleft == MPI_PROC_NULL ) THEN
[3182]	1089	IF ( bc_lr == 'dirichlet/radiation' .OR. bc_lr == 'nested' .OR. &
	1090	bc_lr == 'nesting_offline' ) THEN
	1091	bc_dirichlet_l = .TRUE.
[1964]	1092	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
[3182]	1093	bc_radiation_l = .TRUE.
[1]	1094	ENDIF
	1095	ENDIF
[1933]	1096
[1964]	1097	IF ( pright == MPI_PROC_NULL ) THEN
	1098	IF ( bc_lr == 'dirichlet/radiation' ) THEN
[3182]	1099	bc_radiation_r = .TRUE.
	1100	ELSEIF ( bc_lr == 'radiation/dirichlet' .OR. bc_lr == 'nested' .OR. &
	1101	bc_lr == 'nesting_offline' ) THEN
	1102	bc_dirichlet_r = .TRUE.
[1]	1103	ENDIF
	1104	ENDIF
	1105
[1964]	1106	IF ( psouth == MPI_PROC_NULL ) THEN
	1107	IF ( bc_ns == 'dirichlet/radiation' ) THEN
[3182]	1108	bc_radiation_s = .TRUE.
	1109	ELSEIF ( bc_ns == 'radiation/dirichlet' .OR. bc_ns == 'nested' .OR. &
	1110	bc_ns == 'nesting_offline' ) THEN
	1111	bc_dirichlet_s = .TRUE.
[1]	1112	ENDIF
	1113	ENDIF
	1114
[1964]	1115	IF ( pnorth == MPI_PROC_NULL ) THEN
[3182]	1116	IF ( bc_ns == 'dirichlet/radiation' .OR. bc_ns == 'nested' .OR. &
	1117	bc_ns == 'nesting_offline' ) THEN
	1118	bc_dirichlet_n = .TRUE.
[1964]	1119	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
[3182]	1120	bc_radiation_n = .TRUE.
[1]	1121	ENDIF
	1122	ENDIF
[2938]	1123	!
	1124	!-- In case of synthetic turbulence geneartor determine ids.
	1125	!-- Please note, if no forcing or nesting is applied, the generator is applied
	1126	!-- only at the left lateral boundary.
	1127	IF ( use_syn_turb_gen ) THEN
[3182]	1128	IF ( bc_dirichlet_l ) THEN
[2938]	1129	id_stg_left_l = myidx
	1130	ELSE
	1131	id_stg_left_l = 0
	1132	ENDIF
[3182]	1133	IF ( bc_dirichlet_r ) THEN
[2938]	1134	id_stg_right_l = myidx
	1135	ELSE
	1136	id_stg_right_l = 0
	1137	ENDIF
[3182]	1138	IF ( bc_dirichlet_s ) THEN
[2938]	1139	id_stg_south_l = myidy
	1140	ELSE
	1141	id_stg_south_l = 0
	1142	ENDIF
[3182]	1143	IF ( bc_dirichlet_n ) THEN
[2938]	1144	id_stg_north_l = myidy
	1145	ELSE
	1146	id_stg_north_l = 0
	1147	ENDIF
[1968]	1148
[2938]	1149	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
	1150	CALL MPI_ALLREDUCE( id_stg_left_l, id_stg_left, 1, MPI_INTEGER, &
	1151	MPI_SUM, comm1dx, ierr )
	1152
	1153	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
	1154	CALL MPI_ALLREDUCE( id_stg_right_l, id_stg_right, 1, MPI_INTEGER, &
	1155	MPI_SUM, comm1dx, ierr )
	1156
	1157	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
	1158	CALL MPI_ALLREDUCE( id_stg_south_l, id_stg_south, 1, MPI_INTEGER, &
	1159	MPI_SUM, comm1dy, ierr )
	1160
	1161	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
	1162	CALL MPI_ALLREDUCE( id_stg_north_l, id_stg_north, 1, MPI_INTEGER, &
	1163	MPI_SUM, comm1dy, ierr )
	1164
	1165	ENDIF
	1166
[151]	1167	!
	1168	!-- Broadcast the id of the inflow PE
[3182]	1169	IF ( bc_dirichlet_l ) THEN
[163]	1170	id_inflow_l = myidx
[151]	1171	ELSE
	1172	id_inflow_l = 0
	1173	ENDIF
[622]	1174	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
[151]	1175	CALL MPI_ALLREDUCE( id_inflow_l, id_inflow, 1, MPI_INTEGER, MPI_SUM, &
	1176	comm1dx, ierr )
	1177
[163]	1178	!
	1179	!-- Broadcast the id of the recycling plane
	1180	!-- WARNING: needs to be adjusted in case of inflows other than from left side!
[3884]	1181	IF ( turbulent_inflow ) THEN
	1182
[4045]	1183	IF ( NINT( recycling_width / dx, KIND=idp ) >= nxl .AND. &
	1184	NINT( recycling_width / dx, KIND=idp ) <= nxr ) THEN
[3884]	1185	id_recycling_l = myidx
	1186	ELSE
	1187	id_recycling_l = 0
	1188	ENDIF
	1189	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
	1190	CALL MPI_ALLREDUCE( id_recycling_l, id_recycling, 1, MPI_INTEGER, MPI_SUM, &
	1191	comm1dx, ierr )
	1192
[163]	1193	ENDIF
	1194
[2050]	1195	!
	1196	!-- Broadcast the id of the outflow PE and outflow-source plane
	1197	IF ( turbulent_outflow ) THEN
	1198
[3182]	1199	IF ( bc_radiation_r ) THEN
[2050]	1200	id_outflow_l = myidx
	1201	ELSE
	1202	id_outflow_l = 0
	1203	ENDIF
	1204	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
	1205	CALL MPI_ALLREDUCE( id_outflow_l, id_outflow, 1, MPI_INTEGER, MPI_SUM, &
	1206	comm1dx, ierr )
	1207
	1208	IF ( NINT( outflow_source_plane / dx ) >= nxl .AND. &
	1209	NINT( outflow_source_plane / dx ) <= nxr ) THEN
	1210	id_outflow_source_l = myidx
	1211	ELSE
	1212	id_outflow_source_l = 0
	1213	ENDIF
	1214	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
	1215	CALL MPI_ALLREDUCE( id_outflow_source_l, id_outflow_source, 1, &
	1216	MPI_INTEGER, MPI_SUM, comm1dx, ierr )
	1217
	1218	ENDIF
	1219
[3885]	1220	CALL location_message( 'creating virtual PE grids + MPI derived data types', 'finished' )
[1384]	1221
[1804]	1222	#else
[1159]	1223	IF ( bc_lr == 'dirichlet/radiation' ) THEN
[3182]	1224	bc_dirichlet_l = .TRUE.
	1225	bc_radiation_r = .TRUE.
[1159]	1226	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
[3182]	1227	bc_radiation_l = .TRUE.
	1228	bc_dirichlet_r = .TRUE.
[1]	1229	ENDIF
	1230
[1159]	1231	IF ( bc_ns == 'dirichlet/radiation' ) THEN
[3182]	1232	bc_dirichlet_n = .TRUE.
	1233	bc_radiation_s = .TRUE.
[1159]	1234	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
[3182]	1235	bc_radiation_n = .TRUE.
	1236	bc_dirichlet_s = .TRUE.
[1]	1237	ENDIF
	1238	#endif
[807]	1239
[106]	1240	!
[978]	1241	!-- At the inflow or outflow, u or v, respectively, have to be calculated for
	1242	!-- one more grid point.
[3182]	1243	IF ( bc_dirichlet_l .OR. bc_radiation_l ) THEN
[106]	1244	nxlu = nxl + 1
	1245	ELSE
	1246	nxlu = nxl
	1247	ENDIF
[3182]	1248	IF ( bc_dirichlet_s .OR. bc_radiation_s ) THEN
[106]	1249	nysv = nys + 1
	1250	ELSE
	1251	nysv = nys
	1252	ENDIF
[1]	1253
[114]	1254	!
	1255	!-- Allocate wall flag arrays used in the multigrid solver
[1575]	1256	IF ( psolver(1:9) == 'multigrid' ) THEN
[114]	1257
	1258	DO i = maximum_grid_level, 1, -1
	1259
	1260	SELECT CASE ( i )
	1261
	1262	CASE ( 1 )
	1263	ALLOCATE( wall_flags_1(nzb:nzt_mg(i)+1, &
	1264	nys_mg(i)-1:nyn_mg(i)+1, &
	1265	nxl_mg(i)-1:nxr_mg(i)+1) )
	1266
	1267	CASE ( 2 )
	1268	ALLOCATE( wall_flags_2(nzb:nzt_mg(i)+1, &
	1269	nys_mg(i)-1:nyn_mg(i)+1, &
	1270	nxl_mg(i)-1:nxr_mg(i)+1) )
	1271
	1272	CASE ( 3 )
	1273	ALLOCATE( wall_flags_3(nzb:nzt_mg(i)+1, &
	1274	nys_mg(i)-1:nyn_mg(i)+1, &
	1275	nxl_mg(i)-1:nxr_mg(i)+1) )
	1276
	1277	CASE ( 4 )
	1278	ALLOCATE( wall_flags_4(nzb:nzt_mg(i)+1, &
	1279	nys_mg(i)-1:nyn_mg(i)+1, &
	1280	nxl_mg(i)-1:nxr_mg(i)+1) )
	1281
	1282	CASE ( 5 )
	1283	ALLOCATE( wall_flags_5(nzb:nzt_mg(i)+1, &
	1284	nys_mg(i)-1:nyn_mg(i)+1, &
	1285	nxl_mg(i)-1:nxr_mg(i)+1) )
	1286
	1287	CASE ( 6 )
	1288	ALLOCATE( wall_flags_6(nzb:nzt_mg(i)+1, &
	1289	nys_mg(i)-1:nyn_mg(i)+1, &
	1290	nxl_mg(i)-1:nxr_mg(i)+1) )
	1291
	1292	CASE ( 7 )
	1293	ALLOCATE( wall_flags_7(nzb:nzt_mg(i)+1, &
	1294	nys_mg(i)-1:nyn_mg(i)+1, &
	1295	nxl_mg(i)-1:nxr_mg(i)+1) )
	1296
	1297	CASE ( 8 )
	1298	ALLOCATE( wall_flags_8(nzb:nzt_mg(i)+1, &
	1299	nys_mg(i)-1:nyn_mg(i)+1, &
	1300	nxl_mg(i)-1:nxr_mg(i)+1) )
	1301
	1302	CASE ( 9 )
	1303	ALLOCATE( wall_flags_9(nzb:nzt_mg(i)+1, &
	1304	nys_mg(i)-1:nyn_mg(i)+1, &
	1305	nxl_mg(i)-1:nxr_mg(i)+1) )
	1306
	1307	CASE ( 10 )
	1308	ALLOCATE( wall_flags_10(nzb:nzt_mg(i)+1, &
	1309	nys_mg(i)-1:nyn_mg(i)+1, &
	1310	nxl_mg(i)-1:nxr_mg(i)+1) )
	1311
	1312	CASE DEFAULT
[254]	1313	message_string = 'more than 10 multigrid levels'
	1314	CALL message( 'init_pegrid', 'PA0238', 1, 2, 0, 6, 0 )
[114]	1315
	1316	END SELECT
	1317
	1318	ENDDO
	1319
	1320	ENDIF
	1321
[1]	1322	END SUBROUTINE init_pegrid

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

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |