mpp_domains_mod

Domain decomposition and domain update for message-passing codes. More...

Data Types
interface	check_data_size
	Private interface for internal usage, compares two sizes. More...
type	contact_type
	Type used to represent the contact between tiles. More...
type	domain1d
	One dimensional domain used to manage shared data access between pes. More...
type	domain1d_spec
	A private type used to specify index limits for a domain decomposition. More...
type	domain2d
	The domain2D type contains all the necessary information to define the global, compute and data domains of each task, as well as the PE associated with the task. The PEs from which remote data may be acquired to update the data domain are also contained in a linked list of neighbours. More...
type	domain2d_spec
	Private type to specify multiple index limits and pe information for a 2D domain. More...
type	domain_axis_spec
	Used to specify index limits along an axis of a domain. More...
type	domaincommunicator2d
	Used for sending domain data between pe's. More...
type	domainug
	Domain information for managing data on unstructured grids. More...
type	index_type
	index bounds for use in nestSpec More...
interface	mpp_broadcast_domain
	Broadcasts domain to every pe. Only useful outside the context of it's own pelist. More...
interface	mpp_check_field
	Parallel checking between two ensembles which run on different set pes at the same time There are two forms for the mpp_check_field call. The 2D version is generally to be used and 3D version is built by repeated calls to the 2D version. Example usage: More...
interface	mpp_complete_do_update
	Private interface used for non blocking updates. More...
interface	mpp_complete_group_update
	Completes a pending non-blocking group update Must follow a call to mpp_start_group_update. More...
interface	mpp_complete_update_domains
	Must be used after a call to mpp_start_update_domains in order to complete a nonblocking domain update. See mpp_start_update_domains for more info. More...
interface	mpp_copy_domain
	Copy 1D or 2D domain. More...
interface	mpp_create_group_update
	Constructor for the mpp_group_update_type which is then used with mpp_start_group_update. More...
interface	mpp_deallocate_domain
	Deallocate given 1D or 2D domain. More...
interface	mpp_define_domains
	Set up a domain decomposition. More...
interface	mpp_define_layout
	Retrieve layout associated with a domain decomposition. Given a global 2D domain and the number of divisions in the decomposition ndivs (usually the PE count unless some domains are masked) this calls returns a 2D domain layout. By default, mpp_define_layout will attempt to divide the 2D index space into domains that maintain the aspect ratio of the global domain. If this cannot be done, the algorithm favours domains that are longer in x than y, a preference that could improve vector performance. Example usage: More...
interface	mpp_define_null_domain
	Defines a nullified 1D or 2D domain. More...
interface	mpp_do_check
	Private interface to updates data domain of 3D field whose computational domains have been computed. More...
interface	mpp_do_get_boundary
interface	mpp_do_get_boundary_ad
interface	mpp_do_global_field
	Private helper interface used by mpp_global_field. More...
interface	mpp_do_global_field_ad
interface	mpp_do_group_update
interface	mpp_do_redistribute
interface	mpp_do_update
	Private interface used for mpp_update_domains. More...
interface	mpp_do_update_ad
	Passes a data field from a unstructured grid to an structured grid Example usage: More...
interface	mpp_do_update_nest_coarse
	Used by mpp_update_nest_coarse to perform domain updates. More...
interface	mpp_do_update_nest_fine
interface	mpp_get_boundary
	Get the boundary data for symmetric domain when the data is at C, E, or N-cell center. mpp_get_boundary is used to get the boundary data for symmetric domain when the data is at C, E, or N-cell center. For cubic grid, the data should always at C-cell center. Example usage: More...
interface	mpp_get_boundary_ad
interface	mpp_get_compute_domain
	These routines retrieve the axis specifications associated with the compute domains. The domain is a derived type with private elements. These routines retrieve the axis specifications associated with the compute domains The 2D version of these is a simple extension of 1D. Example usage: More...
interface	mpp_get_compute_domains
	Retrieve the entire array of compute domain extents associated with a decomposition. More...
interface	mpp_get_data_domain
	These routines retrieve the axis specifications associated with the data domains. The domain is a derived type with private elements. These routines retrieve the axis specifications associated with the data domains. The 2D version of these is a simple extension of 1D. Example usage: More...
interface	mpp_get_domain_extents
interface	mpp_get_f2c_index
	Get the index of the data passed from fine grid to coarse grid. Example usage: More...
interface	mpp_get_global_domain
	These routines retrieve the axis specifications associated with the global domains. The domain is a derived type with private elements. These routines retrieve the axis specifications associated with the global domains. The 2D version of these is a simple extension of 1D. Example usage: More...
interface	mpp_get_global_domains
interface	mpp_get_layout
	Retrieve layout associated with a domain decomposition The 1D version of this call returns the number of divisions that was assigned to this decomposition axis. The 2D version of this call returns an array of dimension 2 holding the results on two axes. Example usage: More...
interface	mpp_get_memory_domain
	These routines retrieve the axis specifications associated with the memory domains. The domain is a derived type with private elements. These routines retrieve the axis specifications associated with the memory domains. The 2D version of these is a simple extension of 1D. Example usage: More...
interface	mpp_get_neighbor_pe
	Retrieve PE number of a neighboring domain. More...
interface	mpp_get_pelist
	Retrieve list of PEs associated with a domain decomposition. The 1D version of this call returns an array of the PEs assigned to this 1D domain decomposition. In addition the optional argument pos may be used to retrieve the 0-based position of the domain local to the calling PE, i.e., domain%list(pos)%pe is the local PE, as returned by mpp_pe(). The 2D version of this call is identical to 1D version. More...
interface	mpp_global_field
	Fill in a global array from domain-decomposed arrays. More...
interface	mpp_global_field_ad
interface	mpp_global_field_ug
	Same functionality as mpp_global_field but for unstructured domains. More...
interface	mpp_global_max
	Global max of domain-decomposed arrays. mpp_global_max is used to get the maximum value of a domain-decomposed array on each PE. MPP_TYPE_can be of type integer or real; of 4-byte or 8-byte kind; of rank up to 5. The dimension of locus must equal the rank of field. All PEs in a domain decomposition must call mpp_global_max, and each will have the result upon exit. The function mpp_global_min, with an identical syntax. is also available. More...
interface	mpp_global_min
	Global min of domain-decomposed arrays. mpp_global_min is used to get the minimum value of a domain-decomposed array on each PE. MPP_TYPE_can be of type integer or real; of 4-byte or 8-byte kind; of rank up to 5. The dimension of locus must equal the rank of field. All PEs in a domain decomposition must call mpp_global_min, and each will have the result upon exit. The function mpp_global_max, with an identical syntax. is also available. More...
interface	mpp_global_sum
	Global sum of domain-decomposed arrays. mpp_global_sum is used to get the sum of a domain-decomposed array on each PE. MPP_TYPE_ can be of type integer, complex, or real; of 4-byte or 8-byte kind; of rank up to 5. More...
interface	mpp_global_sum_ad
interface	mpp_global_sum_tl
type	mpp_group_update_type
	used for updates on a group More...
interface	mpp_modify_domain
	Modifies the extents (compute, data and global) of a given domain. More...
interface	mpp_nullify_domain_list
	Nullify domain list. This interface is needed in mpp_domains_test. 1-D case can be added in if needed. Example usage: More...
interface	mpp_pass_sg_to_ug
	Passes data from a structured grid to an unstructured grid Example usage: More...
interface	mpp_pass_ug_to_sg
	Passes a data field from a structured grid to an unstructured grid Example usage: More...
interface	mpp_redistribute
	Reorganization of distributed global arrays. mpp_redistribute is used to reorganize a distributed array. MPP_TYPE_can be of type integer, complex, or real; of 4-byte or 8-byte kind; of rank up to 5. Example usage: call mpp_redistribute( domain_in, field_in, domain_out, field_out ) More...
interface	mpp_reset_group_update_field
interface	mpp_set_compute_domain
	These routines set the axis specifications associated with the compute domains. The domain is a derived type with private elements. These routines set the axis specifications associated with the compute domains The 2D version of these is a simple extension of 1D. Example usage: More...
interface	mpp_set_data_domain
	These routines set the axis specifications associated with the data domains. The domain is a derived type with private elements. These routines set the axis specifications associated with the data domains. The 2D version of these is a simple extension of 1D. Example usage: More...
interface	mpp_set_global_domain
	These routines set the axis specifications associated with the global domains. The domain is a derived type with private elements. These routines set the axis specifications associated with the global domains. The 2D version of these is a simple extension of 1D. Example usage: More...
interface	mpp_start_do_update
	Private interface used for non blocking updates. More...
interface	mpp_start_group_update
	Starts non-blocking group update Must be followed up with a call to mpp_complete_group_update mpp_group_update_type can be created with mpp_create_group_update. More...
interface	mpp_start_update_domains
	Interface to start halo updates mpp_start_update_domains is used to start a halo update of a domain-decomposed array on each PE. MPP_TYPE_ can be of type complex, integer, logical or real; of 4-byte or 8-byte kind; of rank up to 5. The vector version (with two input data fields) is only present for \ereal types. \empp_start_update_domains must be paired together with \empp_complete_update_domains. In mpp_start_update_domains, a buffer will be pre-post to receive (non-blocking) the data and data on computational domain will be packed and sent (non-blocking send) to other processor. In mpp_complete_update_domains, buffer will be unpacked to fill the halo and mpp_sync_self will be called to to ensure communication safe at the last call of mpp_complete_update_domains. Each mpp_update_domains can be replaced by the combination of mpp_start_update_domains and mpp_complete_update_domains. The arguments in mpp_start_update_domains and mpp_complete_update_domains should be the exact the same as in mpp_update_domains to be replaced except no optional argument "complete". The following are examples on how to replace mpp_update_domains with mpp_start_update_domains/mpp_complete_update_domains. More...
interface	mpp_update_domains
	Performs halo updates for a given domain. More...
interface	mpp_update_domains_ad
	Similar to mpp_update_domains , updates adjoint domains. More...
interface	mpp_update_nest_coarse
	Pass the data from fine grid to fill the buffer to be ready to be interpolated onto coarse grid. Example usage: More...
interface	mpp_update_nest_fine
	Pass the data from coarse grid to fill the buffer to be ready to be interpolated onto fine grid. Example usage: More...
type	nest_domain_type
	domain with nested fine and course tiles More...
type	nest_level_type
	Private type to hold data for each level of nesting. More...
type	nestspec
	Used to specify bounds and index information for nested tiles as a linked list. More...
type	nonblock_type
	Used for nonblocking data transfer. More...
interface	operator(.eq.)
	Equality/inequality operators for domaintypes. More...
interface	operator(.ne.)
type	overlap_type
	Type for overlapping data. More...
type	overlapspec
	Private type for overlap specifications. More...
type	tile_type
	Upper and lower x and y bounds for a tile. More...
type	unstruct_axis_spec
	Private type for axis specification data for an unstructured grid. More...
type	unstruct_domain_spec
	Private type for axis specification data for an unstructured domain. More...
type	unstruct_overlap_type
	Private type. More...
type	unstruct_pass_type
	Private type. More...

Functions/Subroutines
subroutine	add_check_overlap (overlap_out, overlap_in)
	this routine adds the overlap_in into overlap_out
subroutine	add_update_overlap (overlap_out, overlap_in)
subroutine	allocate_check_overlap (overlap, count)
subroutine	allocate_nest_overlap (overlap, count)
subroutine	allocate_update_overlap (overlap, count)
subroutine	apply_cyclic_offset (lstart, lend, offset, gstart, gend, gsize)
	add offset to the index
subroutine	check_alignment (is, ie, js, je, isg, ieg, jsg, jeg, alignment)
subroutine	check_data_size_1d (module, str1, size1, str2, size2)
subroutine	check_data_size_2d (module, str1, isize1, jsize1, str2, isize2, jsize2)
subroutine	check_message_size (domain, update, send, recv, position)
subroutine	check_overlap_pe_order (domain, overlap, name)
subroutine	compute_overlap_coarse_to_fine (nest_domain, overlap, extra_halo, position, name)
subroutine	compute_overlap_fine_to_coarse (nest_domain, overlap, position, name)
	This routine will compute the send and recv information between overlapped nesting region. The data is assumed on T-cell center.
subroutine	compute_overlap_sg2ug (UG_domain, SG_domain)
subroutine	compute_overlap_ug2sg (UG_domain)
subroutine	compute_overlaps (domain, position, update, check, ishift, jshift, x_cyclic_offset, y_cyclic_offset, whalo, ehalo, shalo, nhalo)
	Computes remote domain overlaps. More...
subroutine	compute_overlaps_fold_east (domain, position, ishift, jshift)
	computes remote domain overlaps assumes only one in each direction will calculate the overlapping for T,E,C,N-cell seperately. here assume fold-east and y-cyclic boundary condition
subroutine	compute_overlaps_fold_south (domain, position, ishift, jshift)
	Computes remote domain overlaps assumes only one in each direction will calculate the overlapping for T,E,C,N-cell seperately.
subroutine	compute_overlaps_fold_west (domain, position, ishift, jshift)
	Computes remote domain overlaps assumes only one in each direction will calculate the overlapping for T,E,C,N-cell seperately.
subroutine	convert_index_back (domain, ishift, jshift, rotate, is_in, ie_in, js_in, je_in, is_out, ie_out, js_out, je_out)
integer function	convert_index_to_coarse (domain, ishift, jshift, tile_coarse, istart_coarse, iend_coarse, jstart_coarse, jend_coarse, ntiles_coarse, tile_in, is_in, ie_in, js_in, je_in, is_out, ie_out, js_out, je_out, rotate_out)
integer function	convert_index_to_nest (domain, ishift, jshift, tile_coarse, istart_coarse, iend_coarse, jstart_coarse, jend_coarse, ntiles_coarse, tile_in, is_in, ie_in, js_in, je_in, is_out, ie_out, js_out, je_out, rotate_out)
	This routine will convert the global coarse grid index to nest grid index.
subroutine	copy_nest_overlap (overlap_out, overlap_in)
subroutine	deallocate_comm (d_comm)
subroutine	deallocate_domain2d_local (domain)
subroutine	deallocate_nest_overlap (overlap)
subroutine	deallocate_overlap_type (overlap)
subroutine	deallocate_overlapspec (overlap)
subroutine	deallocate_unstruct_overlap_type (overlap)
subroutine	deallocate_unstruct_pass_type (domain)
subroutine	debug_message_size (overlap, name)
subroutine	define_contact_point (domain, position, num_contact, tile1, tile2, align1, align2, refine1, refine2, istart1, iend1, jstart1, jend1, istart2, iend2, jstart2, jend2, isgList, iegList, jsgList, jegList)
	compute the overlapping between tiles for the T-cell. More...
subroutine	define_nest_level_type (nest_domain, x_refine, y_refine, extra_halo)
subroutine	expand_check_overlap_list (overlaplist, npes)
subroutine	expand_update_overlap_list (overlapList, npes)
subroutine	fill_contact (Contact, tile, is1, ie1, js1, je1, is2, ie2, js2, je2, align1, align2, refine1, refine2)
	always fill the contact according to index order.
subroutine	fill_corner_contact (eCont, sCont, wCont, nCont, isg, ieg, jsg, jeg, numR, numS, tileRecv, tileSend, is1Recv, ie1Recv, js1Recv, je1Recv, is2Recv, ie2Recv, js2Recv, je2Recv, is1Send, ie1Send, js1Send, je1Send, is2Send, ie2Send, js2Send, je2Send, align1Recv, align2Recv, align1Send, align2Send, whalo, ehalo, shalo, nhalo, tileMe)
subroutine	fill_overlap (overlap, domain, m, is, ie, js, je, isc, iec, jsc, jec, isg, ieg, jsg, jeg, dir, reverse, symmetry)
subroutine	fill_overlap_recv_fold (overlap, domain, m, is, ie, js, je, isd, ied, jsd, jed, isg, ieg, dir, ishift, position, ioff, middle, symmetry)
subroutine	fill_overlap_recv_nofold (overlap, domain, m, is, ie, js, je, isd, ied, jsd, jed, isg, ieg, dir, ioff, is_cyclic, folded, symmetry)
subroutine	fill_overlap_send_fold (overlap, domain, m, is, ie, js, je, isc, iec, jsc, jec, isg, ieg, dir, ishift, position, ioff, middle, symmetry)
subroutine	fill_overlap_send_nofold (overlap, domain, m, is, ie, js, je, isc, iec, jsc, jec, isg, ieg, dir, ioff, is_cyclic, folded, symmetry)
integer function	find_index (array, index_data, start_pos)
integer function	find_key (key, sorted, insert)
subroutine	free_comm (domain_id, l_addr, l_addr2)
subroutine	get_coarse_index (rotate, is, ie, js, je, iadd, jadd, is_c, ie_c, js_c, je_c)
type(domaincommunicator2d) function, pointer	get_comm (domain_id, l_addr, l_addr2)
subroutine	get_fold_index_east (jsg, jeg, ieg, jshift, position, is, ie, js, je)
subroutine	get_fold_index_north (isg, ieg, jeg, ishift, position, is, ie, js, je)
subroutine	get_fold_index_south (isg, ieg, jsg, ishift, position, is, ie, js, je)
subroutine	get_fold_index_west (jsg, jeg, isg, jshift, position, is, ie, js, je)
integer function	get_nest_vector_recv (nest_domain, update_x, update_y, ind_x, ind_y, start_pos, pelist)
integer function	get_nest_vector_send (nest_domain, update_x, update_y, ind_x, ind_y, start_pos, pelist)
subroutine	get_nnest (domain, num_nest, tile_coarse, istart_coarse, iend_coarse, jstart_coarse, jend_coarse, x_refine, y_refine, nnest, t_coarse, ncross_coarse, rotate_coarse, is_coarse, ie_coarse, js_coarse, je_coarse, is_fine, ie_fine, js_fine, je_fine)
subroutine	init_index_type (indexData)
subroutine	init_overlap_type (overlap)
subroutine	insert_check_overlap (overlap, pe, tileMe, dir, rotation, is, ie, js, je)
subroutine	insert_nest_overlap (overlap, pe, is, ie, js, je, dir, rotation)
subroutine	insert_overlap_type (overlap, pe, tileMe, tileNbr, is, ie, js, je, dir, rotation, from_contact)
subroutine	insert_update_overlap (overlap, pe, is1, ie1, js1, je1, is2, ie2, js2, je2, dir, reverse, symmetry)
subroutine	mpp_broadcast_domain_ug (domain)
	Broadcast domain (useful only outside the context of its own pelist)
subroutine	mpp_compute_block_extent (isg, ieg, ndivs, ibegin, iend)
	Computes the extents of a grid block. More...
subroutine	mpp_compute_extent (isg, ieg, ndivs, ibegin, iend, extent)
	Computes extents for a grid decomposition with the given indices and divisions.
subroutine	mpp_deallocate_domain1d (domain)
subroutine	mpp_deallocate_domain2d (domain)
subroutine	mpp_deallocate_domainug (domain)
subroutine	mpp_define_domains1d (global_indices, ndivs, domain, pelist, flags, halo, extent, maskmap, memory_size, begin_halo, end_halo)
	Define data and computational domains on a 1D set of data (isg:ieg) and assign them to PEs. More...
subroutine	mpp_define_domains2d (global_indices, layout, domain, pelist, xflags, yflags, xhalo, yhalo, xextent, yextent, maskmap, name, symmetry, memory_size, whalo, ehalo, shalo, nhalo, is_mosaic, tile_count, tile_id, complete, x_cyclic_offset, y_cyclic_offset)
	Define 2D data and computational domain on global rectilinear cartesian domain (isg:ieg,jsg:jeg) and assign them to PEs. More...
subroutine	mpp_define_io_domain (domain, io_layout)
	Define the layout for IO pe's for the given domain. More...
subroutine	mpp_define_layout2d (global_indices, ndivs, layout)
subroutine	mpp_define_mosaic (global_indices, layout, domain, num_tile, num_contact, tile1, tile2, istart1, iend1, jstart1, jend1, istart2, iend2, jstart2, jend2, pe_start, pe_end, pelist, whalo, ehalo, shalo, nhalo, xextent, yextent, maskmap, name, memory_size, symmetry, xflags, yflags, tile_id)
	Defines a domain for mosaic tile grids. More...
subroutine	mpp_define_mosaic_pelist (sizes, pe_start, pe_end, pelist, costpertile)
	Defines a pelist for use with mosaic tiles. More...
subroutine	mpp_define_nest_domains (nest_domain, domain, num_nest, nest_level, tile_fine, tile_coarse, istart_coarse, icount_coarse, jstart_coarse, jcount_coarse, npes_nest_tile, x_refine, y_refine, extra_halo, name)
	Set up a domain to pass data between aligned coarse and fine grid of nested model. More...
subroutine	mpp_define_null_domain1d (domain)
subroutine	mpp_define_null_domain2d (domain)
subroutine	mpp_define_null_ug_domain (domain)
subroutine	mpp_define_unstruct_domain (UG_domain, SG_domain, npts_tile, grid_nlev, ndivs, npes_io_group, grid_index, name)
logical(l8_kind) function	mpp_domain_ug_is_tile_root_pe (domain)
logical function	mpp_domainug_eq (a, b)
	Overload the .eq. for UG.
logical function	mpp_domainug_ne (a, b)
	Overload the .ne. for UG.
subroutine	mpp_get_c2f_index (nest_domain, is_fine, ie_fine, js_fine, je_fine, is_coarse, ie_coarse, js_coarse, je_coarse, dir, nest_level, position)
	Get the index of the data passed from coarse grid to fine grid. More...
subroutine	mpp_get_f2c_index_coarse (nest_domain, is_coarse, ie_coarse, js_coarse, je_coarse, nest_level, position)
subroutine	mpp_get_f2c_index_fine (nest_domain, is_coarse, ie_coarse, js_coarse, je_coarse, is_fine, ie_fine, js_fine, je_fine, nest_level, position)
integer(i4_kind) function	mpp_get_io_domain_ug_layout (domain)
type(domain2d) function, pointer	mpp_get_nest_coarse_domain (nest_domain, nest_level)
type(domain2d) function, pointer	mpp_get_nest_fine_domain (nest_domain, nest_level)
integer function	mpp_get_nest_fine_npes (nest_domain, nest_level)
subroutine	mpp_get_nest_fine_pelist (nest_domain, nest_level, pelist)
integer function	mpp_get_nest_npes (nest_domain, nest_level)
subroutine	mpp_get_nest_pelist (nest_domain, nest_level, pelist)
subroutine	mpp_get_ug_compute_domain (domain, begin, end, size)
subroutine	mpp_get_ug_compute_domains (domain, begin, end, size)
subroutine	mpp_get_ug_domain_grid_index (domain, grid_index)
integer function	mpp_get_ug_domain_npes (domain)
integer function	mpp_get_ug_domain_ntiles (domain)
subroutine	mpp_get_ug_domain_pelist (domain, pelist)
integer function	mpp_get_ug_domain_tile_id (domain)
subroutine	mpp_get_ug_domain_tile_list (domain, tiles)
subroutine	mpp_get_ug_domain_tile_pe_inf (domain, root_pe, npes, pelist)
subroutine	mpp_get_ug_domains_index (domain, begin, end)
subroutine	mpp_get_ug_global_domain (domain, begin, end, size)
type(domainug) function, pointer	mpp_get_ug_io_domain (domain)
subroutine	mpp_get_ug_sg_domain (UG_domain, SG_domain)
subroutine	mpp_global_field_free_comm (domain, l_addr, ksize, l_addr2, flags)
type(domaincommunicator2d) function, pointer	mpp_global_field_init_comm (domain, l_addr, isize_g, jsize_g, isize_l, jsize_l, ksize, l_addr2, flags, position)
	initializes a DomainCommunicator2D type for use in mpp_global_field
logical function	mpp_is_nest_coarse (nest_domain, nest_level)
logical function	mpp_is_nest_fine (nest_domain, nest_level)
subroutine	mpp_modify_domain1d (domain_in, domain_out, cbegin, cend, gbegin, gend, hbegin, hend)
	Modifies the exents of a domain. More...
subroutine	mpp_modify_domain2d (domain_in, domain_out, isc, iec, jsc, jec, isg, ieg, jsg, jeg, whalo, ehalo, shalo, nhalo)
logical function	mpp_mosaic_defined ()
	Accessor function for value of mosaic_defined.
subroutine	mpp_redistribute_free_comm (domain_in, l_addr, domain_out, l_addr2, ksize, lsize)
type(domaincommunicator2d) function, pointer	mpp_redistribute_init_comm (domain_in, l_addrs_in, domain_out, l_addrs_out, isize_in, jsize_in, ksize_in, isize_out, jsize_out, ksize_out)
subroutine	mpp_shift_nest_domains (nest_domain, domain, delta_i_coarse, delta_j_coarse, extra_halo)
	Based on mpp_define_nest_domains, but just resets positioning of nest Modifies the parent/coarse start and end indices of the nest location Computes new overlaps of nest PEs on parent PEs Ramstrom/HRD Moving Nest. More...
subroutine	pop_key (sorted, idx, n_idx, key_idx)
subroutine	print_nest_overlap (overlap, msg)
integer function	push_key (sorted, idx, n_idx, insert, key, ival)
type(nestspec) function, pointer	search_c2f_nest_overlap (nest_domain, nest_level, extra_halo, position)
type(nestspec) function, pointer	search_f2c_nest_overlap (nest_domain, nest_level, position)
subroutine	set_bound_overlap (domain, position)
	set up the overlapping for boundary if the domain is symmetry.
subroutine	set_check_overlap (domain, position)
	set up the overlapping for boundary check if the domain is symmetry. The check will be done on current pe for east boundary for E-cell, north boundary for N-cell, East and North boundary for C-cell
subroutine	set_contact_point (domain, position)
	this routine sets the overlapping between tiles for E,C,N-cell based on T-cell overlapping
subroutine	set_domain_comm_inf (update)
integer(i8_kind) function	set_domain_id (d_id, ksize, flags, gtype, position, whalo, ehalo, shalo, nhalo)
subroutine	set_overlaps (domain, overlap_in, overlap_out, whalo_out, ehalo_out, shalo_out, nhalo_out)
	this routine sets up the overlapping for mpp_update_domains for arbitrary halo update. should be the halo size defined in mpp_define_domains. xhalo_out, yhalo_out should not be exactly the same as xhalo_in, yhalo_in currently we didn't consider about tripolar grid situation, because in the folded north region, the overlapping is specified through list of points, not through rectangular. But will return back to solve this problem in the future.
subroutine	set_single_overlap (overlap_in, overlap_out, isoff, ieoff, jsoff, jeoff, index, dir, rotation)

Variables
integer, save	a2_sort_len =0
	length sorted memory list
integer, save	a_sort_len =0
	length sorted memory list
integer(i8_kind), parameter	addr2_base = 65536_i8_kind
	= 0x0000000000010000
integer, dimension(-1:max_addrs2), save	addrs2_idx =-9999
	index of addr2 associated with d_comm
integer(i8_kind), dimension(max_addrs2), save	addrs2_sorted =-9999
	list of sorted local addresses
integer, dimension(-1:max_addrs), save	addrs_idx =-9999
	index of address associated with d_comm
integer(i8_kind), dimension(max_addrs), save	addrs_sorted =-9999
	list of sorted local addresses
logical	complete_group_update_on = .false.
logical	complete_update = .false.
integer	current_id_update = 0
type(domaincommunicator2d), dimension(:), allocatable, target, save	d_comm
	domain communicators
integer, dimension(-1:max_fields), save	d_comm_idx =-9999
	index of d_comm associated with sorted addresses
integer, save	dc_sort_len =0
	length sorted comm keys (=num active communicators)
integer(i8_kind), dimension(max_fields), save	dckey_sorted =-9999
	list of sorted local addresses
logical	debug = .FALSE.
logical	debug_message_passing = .false.
	Will check the consistency on the boundary between processor/tile when updating domain for symmetric domain and check the consistency on the north folded edge.
character(len=32)	debug_update_domain = "none"
	namelist interface More...
integer	debug_update_level = NO_CHECK
logical	domain_clocks_on =.FALSE.
integer(i8_kind)	domain_cnt =0
logical	efp_sum_overflow_check = .false.
	If .true., always do overflow_check when doing EFP bitwise mpp_global_sum.
integer, parameter	field_s = 0
integer, parameter	field_x = 1
integer, parameter	field_y = 2
integer	group_pack_clock =0
integer	group_recv_clock =0
integer	group_send_clock =0
integer	group_unpk_clock =0
integer	group_update_buffer_pos = 0
integer	group_wait_clock =0
integer(i8_kind), parameter	gt_base = 256_i8_kind
integer, save	i_sort_len =0
	length sorted domain ids list
integer, dimension(-1:max_dom_ids), save	ids_idx =-9999
	index of d_comm associated with sorted addesses
integer(i8_kind), dimension(max_dom_ids), save	ids_sorted =-9999
	list of sorted domain identifiers
integer(i8_kind), parameter	ke_base = 281474976710656_i8_kind
integer, parameter	max_addrs =512
integer, parameter	max_addrs2 =128
integer, parameter	max_dom_ids =128
integer, parameter	max_fields =1024
integer, parameter	max_nonblock_update = 100
integer, parameter	maxlist = 100
integer, parameter	maxoverlap = 200
logical	module_is_initialized = .false.
logical	mosaic_defined = .false.
integer	mpp_domains_stack_hwm =0
integer	mpp_domains_stack_size =0
integer, save	n_addrs =0
	number of memory addresses used
integer, save	n_addrs2 =0
	number of memory addresses used
integer, save	n_comm =0
	number of communicators used
integer, save	n_ids =0
	number of domain ids used (=i_sort_len; domain ids are never removed)
integer	nest_pack_clock =0
integer	nest_recv_clock =0
integer	nest_send_clock =0
integer	nest_unpk_clock =0
integer	nest_wait_clock =0
integer, parameter	no_check = -1
integer	nonblock_buffer_pos = 0
type(nonblock_type), dimension(:), allocatable	nonblock_data
integer	nonblock_group_buffer_pos = 0
integer	nonblock_group_pack_clock =0
integer	nonblock_group_recv_clock =0
integer	nonblock_group_send_clock =0
integer	nonblock_group_unpk_clock =0
integer	nonblock_group_wait_clock =0
integer	nthread_control_loop = 8
	Determine the loop order for packing and unpacking. When number of threads is greater than nthread_control_loop, the k-loop will be moved outside and combined with number of pack and unpack. When the number of threads is less than or equal to nthread_control_loop, the k-loop is moved inside, but still outside, of j,i loop.
type(domain1d), save, public	null_domain1d
type(domain2d), save, public	null_domain2d
type(domainug), save, public	null_domainug
integer	num_nonblock_group_update = 0
integer	num_update = 0
integer	pack_clock =0
integer	pe
integer	recv_clock =0
integer	recv_clock_nonblock =0
integer	send_clock =0
integer	send_pack_clock_nonblock =0
logical	start_update = .true.
integer	unpk_clock =0
integer	unpk_clock_nonblock =0
logical	use_alltoallw = .false.
logical	verbose =.FALSE.
integer	wait_clock =0
integer	wait_clock_nonblock =0
subroutine	mpp_domains_init (flags)
	Initialize domain decomp package. More...
subroutine	init_nonblock_type (nonblock_obj)
	Initialize domain decomp package. More...
subroutine	mpp_domains_exit ()
	Exit mpp_domains_mod. Serves no particular purpose, but is provided should you require to re-initialize mpp_domains_mod, for some odd reason.
subroutine	mpp_check_field_3d (field_in, pelist1, pelist2, domain, mesg, w_halo, s_halo, e_halo, n_halo, force_abort, position)
	This routine is used to do parallel checking for 3d data between n and m pe. The comparison is is done on pelist2. When size of pelist2 is 1, we can check the halo; otherwise, halo can not be checked. More...
subroutine	mpp_check_field_2d (field_in, pelist1, pelist2, domain, mesg, w_halo, s_halo, e_halo, n_halo, force_abort, position)
	This routine is used to do parallel checking for 2d data between n and m pe. The comparison is is done on pelist2. When size of pelist2 is 1, we can check the halo; otherwise, halo can not be checked. More...
subroutine	mpp_check_field_2d_type1 (field_in, pelist1, pelist2, domain, mesg, w_halo, s_halo, e_halo, n_halo, force_abort)
	This routine is used to check field between running on 1 pe (pelist2) and n pe(pelist1). The need_to_be_checked data is sent to the pelist2 and All the comparison is done on pelist2. More...
subroutine	mpp_check_field_2d_type2 (field_in, pelist1, pelist2, domain, mesg, force_abort)
	This routine is used to check field between running on m pe (root pe) and n pe. This routine can not check halo. More...
subroutine logical	mpp_broadcast_domain_1 (domain)
	broadcast domain (useful only outside the context of its own pelist)
subroutine	mpp_broadcast_domain_nest_coarse (domain, tile_coarse)
	Broadcast nested domain (useful only outside the context of its own pelist)
subroutine	mpp_domains_set_stack_size (n)
	Set user stack size. More...
logical function	mpp_domain1d_eq (a, b)
	Set user stack size. More...
logical function	mpp_domain1d_ne (a, b)
	Set user stack size. More...
logical function	mpp_domain2d_eq (a, b)
	Set user stack size. More...
logical function	mpp_domain2d_ne (a, b)
	Set user stack size. More...
subroutine	mpp_get_compute_domain1d (domain, begin, end, size, max_size, is_global)
	Set user stack size. More...
subroutine	mpp_get_data_domain1d (domain, begin, end, size, max_size, is_global)
	Set user stack size. More...
subroutine	mpp_get_global_domain1d (domain, begin, end, size, max_size)
	Set user stack size. More...
subroutine	mpp_get_memory_domain1d (domain, begin, end, size, max_size, is_global)
	Set user stack size. More...
subroutine	mpp_get_compute_domain2d (domain, xbegin, xend, ybegin, yend, xsize, xmax_size, ysize, ymax_size, x_is_global, y_is_global, tile_count, position)
	Set user stack size. More...
subroutine	mpp_get_data_domain2d (domain, xbegin, xend, ybegin, yend, xsize, xmax_size, ysize, ymax_size, x_is_global, y_is_global, tile_count, position)
	Set user stack size. More...
subroutine	mpp_get_global_domain2d (domain, xbegin, xend, ybegin, yend, xsize, xmax_size, ysize, ymax_size, tile_count, position)
	Set user stack size. More...
subroutine	mpp_get_memory_domain2d (domain, xbegin, xend, ybegin, yend, xsize, xmax_size, ysize, ymax_size, x_is_global, y_is_global, position)
	Set user stack size. More...
subroutine	mpp_set_super_grid_indices (grid)
	Modifies the indices in the domain_axis_spec type to those of the supergrid. More...
subroutine	mpp_create_super_grid_domain (domain)
	Modifies the indices of the input domain to create the supergrid domain. More...
subroutine	mpp_set_compute_domain1d (domain, begin, end, size, is_global)
	Set user stack size. More...
subroutine	mpp_set_compute_domain2d (domain, xbegin, xend, ybegin, yend, xsize, ysize, x_is_global, y_is_global, tile_count)
	Set user stack size. More...
subroutine	mpp_set_data_domain1d (domain, begin, end, size, is_global)
	Set user stack size. More...
subroutine	mpp_set_data_domain2d (domain, xbegin, xend, ybegin, yend, xsize, ysize, x_is_global, y_is_global, tile_count)
	Set user stack size. More...
subroutine	mpp_set_global_domain1d (domain, begin, end, size)
	Set user stack size. More...
subroutine	mpp_set_global_domain2d (domain, xbegin, xend, ybegin, yend, xsize, ysize, tile_count)
	Set user stack size. More...
subroutine	mpp_get_domain_components (domain, x, y, tile_count)
	Retrieve 1D components of 2D decomposition. More...
subroutine	mpp_get_compute_domains1d (domain, begin, end, size)
	Set user stack size. More...
subroutine	mpp_get_compute_domains2d (domain, xbegin, xend, xsize, ybegin, yend, ysize, position)
	Set user stack size. More...
subroutine	mpp_get_global_domains1d (domain, begin, end, size)
	Set user stack size. More...
subroutine	mpp_get_global_domains2d (domain, xbegin, xend, xsize, ybegin, yend, ysize, position)
	Set user stack size. More...
subroutine	mpp_get_domain_extents1d (domain, xextent, yextent)
	Set user stack size. More...
subroutine	mpp_get_domain_extents2d (domain, xextent, yextent)
	This will return xextent and yextent for each tile.
integer function	mpp_get_domain_pe (domain)
	Set user stack size. More...
integer function	mpp_get_domain_tile_root_pe (domain)
	Set user stack size. More...
integer function	mpp_get_domain_tile_commid (domain)
	Set user stack size. More...
integer function	mpp_get_domain_commid (domain)
	Set user stack size. More...
type(domain2d) function, pointer	mpp_get_io_domain (domain)
	Set user stack size. More...
subroutine	mpp_get_pelist1d (domain, pelist, pos)
	Set user stack size. More...
subroutine	mpp_get_pelist2d (domain, pelist, pos)
	Set user stack size. More...
subroutine	mpp_get_layout1d (domain, layout)
	Set user stack size. More...
subroutine	mpp_get_layout2d (domain, layout)
	Set user stack size. More...
subroutine	mpp_get_domain_shift (domain, ishift, jshift, position)
	Returns the shift value in x and y-direction according to domain position.. More...
subroutine	mpp_get_neighbor_pe_1d (domain, direction, pe)
	Return PE to the righ/left of this PE-domain.
subroutine	mpp_get_neighbor_pe_2d (domain, direction, pe)
	Return PE North/South/East/West of this PE-domain. direction must be NORTH, SOUTH, EAST or WEST.
subroutine	nullify_domain2d_list (domain)
	Set user stack size. More...
logical function	mpp_domain_is_symmetry (domain)
	Set user stack size. More...
logical function	mpp_domain_is_initialized (domain)
	Set user stack size. More...
logical function	domain_update_is_needed (domain, whalo, ehalo, shalo, nhalo)
	Set user stack size. More...
type(overlapspec) function, pointer	search_update_overlap (domain, whalo, ehalo, shalo, nhalo, position)
	this routine found the domain has the same halo size with the input whalo, ehalo,
type(overlapspec) function, pointer	search_check_overlap (domain, position)
	this routine finds the check at certain position
type(overlapspec) function, pointer	search_bound_overlap (domain, position)
	This routine finds the bound at certain position.
integer function, dimension(size(domain%tile_id(:)))	mpp_get_tile_id (domain)
	Returns the tile_id on current pe.
subroutine	mpp_get_tile_list (domain, tiles)
	Return the tile_id on current pelist. one-tile-per-pe is assumed.
integer function	mpp_get_ntile_count (domain)
	Returns number of tiles in mosaic.
integer function	mpp_get_current_ntile (domain)
	Returns number of tile on current pe.
logical function	mpp_domain_is_tile_root_pe (domain)
	Returns if current pe is the root pe of the tile, if number of tiles on current pe is greater than 1, will return true, if isc==isg and jsc==jsg also will return true, otherwise false will be returned.
integer function	mpp_get_tile_npes (domain)
	Returns number of processors used on current tile.
subroutine	mpp_get_tile_pelist (domain, pelist)
	Get the processors list used on current tile.
subroutine	mpp_get_tile_compute_domains (domain, xbegin, xend, ybegin, yend, position)
	Set user stack size. More...
integer function	mpp_get_num_overlap (domain, action, p, position)
	Set user stack size. More...
subroutine	mpp_get_update_size (domain, nsend, nrecv, position)
	Set user stack size. More...
subroutine	mpp_get_update_pelist (domain, action, pelist, position)
	Set user stack size. More...
subroutine	mpp_get_overlap (domain, action, p, is, ie, js, je, dir, rot, position)
	Set user stack size. More...
character(len=name_length) function	mpp_get_domain_name (domain)
	Set user stack size. More...
integer function	mpp_get_domain_root_pe (domain)
	Set user stack size. More...
integer function	mpp_get_domain_npes (domain)
	Set user stack size. More...
subroutine	mpp_get_domain_pelist (domain, pelist)
	Set user stack size. More...
integer function, dimension(2)	mpp_get_io_domain_layout (domain)
	Set user stack size. More...
integer function	get_rank_send (domain, overlap_x, overlap_y, rank_x, rank_y, ind_x, ind_y)
	Set user stack size. More...
integer function	get_rank_recv (domain, overlap_x, overlap_y, rank_x, rank_y, ind_x, ind_y)
	Set user stack size. More...
integer function	get_vector_recv (domain, update_x, update_y, ind_x, ind_y, start_pos, pelist)
	Set user stack size. More...
integer function	get_vector_send (domain, update_x, update_y, ind_x, ind_y, start_pos, pelist)
	Set user stack size. More...
integer function	get_rank_unpack (domain, overlap_x, overlap_y, rank_x, rank_y, ind_x, ind_y)
	Set user stack size. More...
integer function	get_mesgsize (overlap, do_dir)
	Set user stack size. More...
subroutine	mpp_set_domain_symmetry (domain, symmetry)
	Set user stack size. More...
recursive subroutine	mpp_copy_domain1d (domain_in, domain_out)
	Copies input 1d domain to the output 1d domain. More...
subroutine	mpp_copy_domain2d (domain_in, domain_out)
	Copies input 2d domain to the output 2d domain. More...
subroutine	mpp_copy_domain2d_spec (domain2D_spec_in, domain2d_spec_out)
	Copies input 2d domain spec to the output 2d domain spec. More...
subroutine	mpp_copy_domain1d_spec (domain1D_spec_in, domain1D_spec_out)
	Copies input 1d domain spec to the output 1d domain spec. More...
subroutine	mpp_copy_domain_axis_spec (domain_axis_spec_in, domain_axis_spec_out)
	Copies input domain_axis_spec to the output domain_axis_spec. More...
subroutine	set_group_update (group, domain)
	Set user stack size. More...
subroutine	mpp_clear_group_update (group)
	Set user stack size. More...
logical function	mpp_group_update_initialized (group)
	Set user stack size. More...
logical function	mpp_group_update_is_set (group)
	Set user stack size. More...

Detailed Description

Domain decomposition and domain update for message-passing codes.

Instantiates a layout with the given indices and divisions.

Author

V. Balaji SGI/GFDL Princeton University

A set of simple calls for domain decomposition and domain updates on rectilinear grids. It requires the module mpp.F90, upon which it is built.
Scalable implementations of finite-difference codes are generally based on decomposing the model domain into subdomains that are distributed among processors. These domains will then be obliged to exchange data at their boundaries if data dependencies are merely nearest-neighbour, or may need to acquire information from the global domain if there are extended data dependencies, as in the spectral transform. The domain decomposition is a key operation in the development of parallel codes.

mpp_domains_mod provides a domain decomposition and domain update API for rectilinear grids, built on top of the mpp_mod API for message passing. Features of mpp_domains_mod include:

Simple, minimal API, with free access to underlying API for more complicated stuff.

Design toward typical use in climate/weather CFD codes.

[Domains]

It is assumed that domain decomposition will mainly be in 2 horizontal dimensions, which will in general be the two fastest-varying indices. There is a separate implementation of 1D decomposition on the fastest-varying index, and 1D decomposition on the second index, treated as a special case of 2D decomposition, is also possible. We define domain as the grid associated with a task. We define the compute domain as the set of gridpoints that are computed by a task, and the data domain as the set of points that are required by the task for the calculation. There can in general be more than 1 task per PE, though often the number of domains is the same as the processor count. We define the global domain as the global computational domain of the entire model (i.e, the same as the computational domain if run on a single processor). 2D domains are defined using a derived type domain2D, constructed as follows (see comments in code for more details).

 type, public :: domain_axis_spec
   private
   integer :: begin, end, size, max_size
   logical :: is_global
 end type domain_axis_spec

 type, public :: domain1D
   private
   type(domain_axis_spec) :: compute, data, global, active
   logical :: mustputb, mustgetb, mustputf, mustgetf, folded
   type(domain1D), pointer, dimension(:) :: list
   integer :: pe  ! pe to which the domain is assigned
   integer :: pos
 end type domain1D

 type, public :: domain2D
   private
   type(domain1D) :: x
   type(domain1D) :: y
   type(domain2D), pointer, dimension(:) :: list
   integer :: pe ! PE to which this domain is assigned
   integer :: pos
 end type domain2D

 type(domain1D), public :: NULL_DOMAIN1D
 type(domain2D), public :: NULL_DOMAIN2D

---

Data Type Documentation

mpp_domains_mod::check_data_size

interface mpp_domains_mod::check_data_size

Private interface for internal usage, compares two sizes.

Definition at line 2335 of file mpp_domains.F90.

Private Member Functions
	check_data_size_1d
	check_data_size_2d

mpp_domains_mod::contact_type

type mpp_domains_mod::contact_type

Type used to represent the contact between tiles.

Note

This type will only be used in mpp_domains_define.inc

Definition at line 417 of file mpp_domains.F90.

Collaboration diagram for contact_type:

Private Attributes
integer, dimension(:), pointer	align1 =>NULL()
integer, dimension(:), pointer	align2 =>NULL()
	alignment of me and neighbor
integer, dimension(:), pointer	ie1 =>NULL()
	i-index of current tile repsenting contact
integer, dimension(:), pointer	ie2 =>NULL()
	i-index of neighbor tile repsenting contact
integer, dimension(:), pointer	is1 =>NULL()
integer, dimension(:), pointer	is2 =>NULL()
integer, dimension(:), pointer	je1 =>NULL()
	j-index of current tile repsenting contact
integer, dimension(:), pointer	je2 =>NULL()
	j-index of neighbor tile repsenting contact
integer, dimension(:), pointer	js1 =>NULL()
integer, dimension(:), pointer	js2 =>NULL()
integer	ncontact
	number of neighbor tile.
real, dimension(:), pointer	refine1 =>NULL()
real, dimension(:), pointer	refine2 =>NULL()
integer, dimension(:), pointer	tile =>NULL()
	neighbor tile

mpp_domains_mod::domain1d

type mpp_domains_mod::domain1d

One dimensional domain used to manage shared data access between pes.

Definition at line 631 of file mpp_domains.F90.

Collaboration diagram for domain1d:

Private Attributes
type(domain_axis_spec)	compute
	index limits for compute domain
logical	cyclic
	true if domain is cyclic
type(domain_axis_spec)	domain_data
	index limits for data domain
type(domain_axis_spec)	global
	index limits for global domain
integer	goffset
	needed for global sum
type(domain1d), dimension(:), pointer	list =>NULL()
	list of each pe's domains
integer	loffset
	needed for global sum
type(domain_axis_spec)	memory
	index limits for memory domain
integer	pe
	PE to which this domain is assigned.
integer	pos
	position of this PE within link list, i.e domainlist(pos)pe = pe

mpp_domains_mod::domain1d_spec

type mpp_domains_mod::domain1d_spec

A private type used to specify index limits for a domain decomposition.

Definition at line 298 of file mpp_domains.F90.

Collaboration diagram for domain1d_spec:

Private Attributes
type(domain_axis_spec)	compute
type(domain_axis_spec)	global
integer	pos

mpp_domains_mod::domain2d

type mpp_domains_mod::domain2d

The domain2D type contains all the necessary information to define the global, compute and data domains of each task, as well as the PE associated with the task. The PEs from which remote data may be acquired to update the data domain are also contained in a linked list of neighbours.

Domain types of higher rank can be constructed from type domain1D typically we only need 1 and 2D, but could need higher (e.g 3D LES) some elements are repeated below if they are needed once per domain, not once per axis

Definition at line 367 of file mpp_domains.F90.

Collaboration diagram for domain2d:

Private Attributes
type(overlapspec), pointer	bound_c => NULL()
	send information for getting boundary value for symmetry domain.
type(overlapspec), pointer	bound_e => NULL()
	send information for getting boundary value for symmetry domain.
type(overlapspec), pointer	bound_n => NULL()
	send information for getting boundary value for symmetry domain.
type(overlapspec), pointer	check_c => NULL()
	send and recv information for boundary consistency check of C-cell
type(overlapspec), pointer	check_e => NULL()
	send and recv information for boundary consistency check of E-cell
type(overlapspec), pointer	check_n => NULL()
	send and recv information for boundary consistency check of N-cell
integer	comm_id
	MPI communicator for the mosaic.
integer	ehalo
	halo size in x-direction
integer	fold
integer(i8_kind)	id
logical	initialized =.FALSE.
	indicate if the overlapping is computed or not.
type(domain2d), pointer	io_domain => NULL()
	domain for IO, will be set through calling mpp_set_io_domain ( this will be changed).
integer, dimension(2)	io_layout
	io_layout, will be set through mpp_define_io_domain default = domain layout
type(domain2d_spec), dimension(:), pointer	list => NULL()
	domain decomposition on pe list
integer	max_ntile_pe
	maximum value in the pelist of number of tiles on each pe.
character(len=name_length)	name ='unnamed'
	name of the domain, default is "unspecified"
integer	ncontacts
	number of contact region within mosaic.
integer	nhalo
	halo size in y-direction
integer	ntiles
	number of tiles within mosaic
integer	pe
	PE to which this domain is assigned.
integer, dimension(:,:), pointer	pearray => NULL()
	pe of each layout position
integer	pos
	position of this PE within link list
logical	rotated_ninety
	indicate if any contact rotate NINETY or MINUS_NINETY
integer	shalo
logical	symmetry
	indicate the domain is symmetric or non-symmetric.
integer	tile_comm_id
	MPI communicator for this tile of domain.
integer, dimension(:), pointer	tile_id => NULL()
	tile id of each tile on current processor
integer, dimension(:), pointer	tile_id_all => NULL()
	tile id of all the tiles of domain
integer	tile_root_pe
	root pe of current tile.
type(tile_type), dimension(:), pointer	tilelist => NULL()
	store tile information
type(overlapspec), pointer	update_c => NULL()
	send and recv information for halo update of C-cell.
type(overlapspec), pointer	update_e => NULL()
	send and recv information for halo update of E-cell.
type(overlapspec), pointer	update_n => NULL()
	send and recv information for halo update of N-cell.
type(overlapspec), pointer	update_t => NULL()
	send and recv information for halo update of T-cell.
integer	whalo
type(domain1d), dimension(:), pointer	x => NULL()
	x-direction domain decomposition
type(domain1d), dimension(:), pointer	y => NULL()
	y-direction domain decomposition

mpp_domains_mod::domain2d_spec

type mpp_domains_mod::domain2d_spec

Private type to specify multiple index limits and pe information for a 2D domain.

Definition at line 307 of file mpp_domains.F90.

Collaboration diagram for domain2d_spec:

Private Attributes
integer	pe
	PE to which this domain is assigned.
integer	pos
	position of this PE within link list
integer, dimension(:), pointer	tile_id => NULL()
	tile id of each tile
integer	tile_root_pe
	root pe of tile.
type(domain1d_spec), dimension(:), pointer	x => NULL()
	x-direction domain decomposition
type(domain1d_spec), dimension(:), pointer	y => NULL()
	y-direction domain decomposition

mpp_domains_mod::domain_axis_spec

type mpp_domains_mod::domain_axis_spec

Used to specify index limits along an axis of a domain.

Definition at line 287 of file mpp_domains.F90.

Collaboration diagram for domain_axis_spec:

Private Attributes
integer	begin
	start of domain axis
integer	end
	end of domain axis
logical	is_global
	.true. if domain axis extent covers global domain
integer	max_size
	max size in set
integer	size
	size of domain axis

mpp_domains_mod::domaincommunicator2d

type mpp_domains_mod::domaincommunicator2d

Used for sending domain data between pe's.

Definition at line 498 of file mpp_domains.F90.

Collaboration diagram for domaincommunicator2d:

Private Attributes
integer, dimension(:), allocatable	cfrom_pe
integer, dimension(:), allocatable	cto_pe
type(domain2d), pointer	domain =>NULL()
type(domain2d), pointer	domain_in =>NULL()
type(domain2d), pointer	domain_out =>NULL()
integer	gf_ioff =0
integer	gf_joff =0
integer(i8_kind)	id =-9999
logical	initialized =.false.
integer	isize =0
integer	isize_in =0
integer	isize_max =0
integer	isize_out =0
integer, dimension(:), allocatable	isizer
integer	jsize =0
integer	jsize_in =0
integer	jsize_max =0
integer	jsize_out =0
integer, dimension(:), allocatable	jsizer
integer	ke =0
integer(i8_kind)	l_addr =-9999
integer(i8_kind)	l_addrx =-9999
integer(i8_kind)	l_addry =-9999
integer	position
	data location. T, E, C, or N.
logical, dimension(:), allocatable	r_do_buf
integer, dimension(:), allocatable	r_msize
type(overlapspec), dimension(:,:,:,:), pointer	recv => NULL()
integer, dimension(:,:), allocatable	recvie
integer, dimension(:,:), allocatable	recvis
integer, dimension(:,:), allocatable	recvje
integer, dimension(:,:), allocatable	recvjs
integer(i8_kind), dimension(:), allocatable	rem_addr
integer(i8_kind), dimension(:,:), allocatable	rem_addrl
integer(i8_kind), dimension(:,:), allocatable	rem_addrlx
integer(i8_kind), dimension(:,:), allocatable	rem_addrly
integer(i8_kind), dimension(:), allocatable	rem_addrx
integer(i8_kind), dimension(:), allocatable	rem_addry
integer	rlist_size =0
logical, dimension(:), allocatable	s_do_buf
integer, dimension(:), allocatable	s_msize
type(overlapspec), dimension(:,:,:,:), pointer	send => NULL()
integer, dimension(:,:), allocatable	sendie
integer, dimension(:,:), allocatable	sendis
integer, dimension(:,:), allocatable	sendisr
integer, dimension(:,:), allocatable	sendje
integer, dimension(:,:), allocatable	sendjs
integer, dimension(:,:), allocatable	sendjsr
integer	slist_size =0

mpp_domains_mod::domainug

type mpp_domains_mod::domainug

Domain information for managing data on unstructured grids.

Definition at line 266 of file mpp_domains.F90.

Collaboration diagram for domainug:

Private Attributes
type(unstruct_axis_spec)	compute
type(unstruct_axis_spec)	global
	axis specifications
integer, dimension(:), pointer	grid_index => NULL()
	index of grid on current pe
type(domainug), pointer	io_domain =>NULL()
integer(i4_kind)	io_layout
type(unstruct_domain_spec), dimension(:), pointer	list =>NULL()
integer	npes_io_group
integer	ntiles
integer	pe
integer	pos
type(unstruct_pass_type)	sg2ug
type(domain2d), pointer	sg_domain => NULL()
integer	tile_id
integer	tile_npes
integer	tile_root_pe
type(unstruct_pass_type)	ug2sg

mpp_domains_mod::index_type

type mpp_domains_mod::index_type

index bounds for use in nestSpec

Definition at line 431 of file mpp_domains.F90.

Collaboration diagram for index_type:

Private Attributes
integer	ie_me
integer	ie_you
integer	is_me
integer	is_you
integer	je_me
integer	je_you
integer	js_me
integer	js_you

mpp_domains_mod::mpp_broadcast_domain

interface mpp_domains_mod::mpp_broadcast_domain

Broadcasts domain to every pe. Only useful outside the context of it's own pelist.

Example usage: call mpp_broadcast_domain(domain) call mpp_broadcast_domain(domain_in, domain_out) call mpp_broadcast_domain(domain, tile_coarse) ! nested domains

Definition at line 1505 of file mpp_domains.F90.

Private Member Functions
	mpp_broadcast_domain_1
	mpp_broadcast_domain_2
	mpp_broadcast_domain_nest_coarse
	mpp_broadcast_domain_nest_fine
	mpp_broadcast_domain_ug

mpp_domains_mod::mpp_check_field

interface mpp_domains_mod::mpp_check_field

Parallel checking between two ensembles which run on different set pes at the same time
There are two forms for the mpp_check_field call. The 2D version is generally to be used and 3D version is built by repeated calls to the 2D version.

Example usage:

call mpp_check_field(field_in, pelist1, pelist2, domain, mesg, &
                           w_halo, s_halo, e_halo, n_halo, force_abort  )

Parameters

field_in	Field to be checked
domain	Domain of current pe
mesg	Message to be printed out
w_halo	Halo size to be checked, default is 0
s_halo	Halo size to be checked, default is 0
e_halo	Halo size to be checked, default is 0
n_halo	Halo size to be checked, default is 0
force_abort	When true, abort program when any difference found. Default is false.

Definition at line 1751 of file mpp_domains.F90.

Private Member Functions
	mpp_check_field_2d
	mpp_check_field_3d

mpp_domains_mod::mpp_complete_do_update

interface mpp_domains_mod::mpp_complete_do_update

Private interface used for non blocking updates.

Definition at line 1294 of file mpp_domains.F90.

Private Member Functions
	mpp_complete_do_update_i4_3d
	mpp_complete_do_update_i8_3d
	mpp_complete_do_update_r4_3d
	mpp_complete_do_update_r4_3dv
	mpp_complete_do_update_r8_3d
	mpp_complete_do_update_r8_3dv

mpp_domains_mod::mpp_complete_group_update

interface mpp_domains_mod::mpp_complete_group_update

Completes a pending non-blocking group update Must follow a call to mpp_start_group_update.

Parameters

[in,out]	type(mpp_group_update_type)	group
[in,out]	type(domain2D)	domain
[in]	d_type	data type

Definition at line 1354 of file mpp_domains.F90.

Private Member Functions
	mpp_complete_group_update_r4
	mpp_complete_group_update_r8

mpp_domains_mod::mpp_complete_update_domains

interface mpp_domains_mod::mpp_complete_update_domains

Must be used after a call to mpp_start_update_domains in order to complete a nonblocking domain update. See mpp_start_update_domains for more info.

Definition at line 1236 of file mpp_domains.F90.

Private Member Functions
	mpp_complete_update_domain2d_i4_2d
	mpp_complete_update_domain2d_i4_3d
	mpp_complete_update_domain2d_i4_4d
	mpp_complete_update_domain2d_i4_5d
	mpp_complete_update_domain2d_i8_2d
	mpp_complete_update_domain2d_i8_3d
	mpp_complete_update_domain2d_i8_4d
	mpp_complete_update_domain2d_i8_5d
	mpp_complete_update_domain2d_r4_2d
	mpp_complete_update_domain2d_r4_2dv
	mpp_complete_update_domain2d_r4_3d
	mpp_complete_update_domain2d_r4_3dv
	mpp_complete_update_domain2d_r4_4d
	mpp_complete_update_domain2d_r4_4dv
	mpp_complete_update_domain2d_r4_5d
	mpp_complete_update_domain2d_r4_5dv
	mpp_complete_update_domain2d_r8_2d
	mpp_complete_update_domain2d_r8_2dv
	mpp_complete_update_domain2d_r8_3d
	mpp_complete_update_domain2d_r8_3dv
	mpp_complete_update_domain2d_r8_4d
	mpp_complete_update_domain2d_r8_4dv
	mpp_complete_update_domain2d_r8_5d
	mpp_complete_update_domain2d_r8_5dv

mpp_domains_mod::mpp_copy_domain

interface mpp_domains_mod::mpp_copy_domain

Copy 1D or 2D domain.

Parameters

domain_in	Input domain to get read
domain_out	Output domain to get written to

Definition at line 912 of file mpp_domains.F90.

Private Member Functions
	mpp_copy_domain1d
	mpp_copy_domain2d

mpp_domains_mod::mpp_create_group_update

interface mpp_domains_mod::mpp_create_group_update

Constructor for the mpp_group_update_type which is then used with mpp_start_group_update.

Parameters

Definition at line 1314 of file mpp_domains.F90.

Private Member Functions
	mpp_create_group_update_r4_2d
	mpp_create_group_update_r4_2dv
	mpp_create_group_update_r4_3d
	mpp_create_group_update_r4_3dv
	mpp_create_group_update_r4_4d
	mpp_create_group_update_r4_4dv
	mpp_create_group_update_r8_2d
	mpp_create_group_update_r8_2dv
	mpp_create_group_update_r8_3d
	mpp_create_group_update_r8_3dv
	mpp_create_group_update_r8_4d
	mpp_create_group_update_r8_4dv

mpp_domains_mod::mpp_deallocate_domain

interface mpp_domains_mod::mpp_deallocate_domain

Deallocate given 1D or 2D domain.

Parameters

domain

an allocated domain1D or domain2D

Definition at line 919 of file mpp_domains.F90.

Private Member Functions
	mpp_deallocate_domain1d
	mpp_deallocate_domain2d

mpp_domains_mod::mpp_define_domains

interface mpp_domains_mod::mpp_define_domains

Set up a domain decomposition.

There are two forms for the mpp_define_domains call. The 2D version is generally to be used but is built by repeated calls to the 1D version, also provided.

Example usage:

               call mpp_define_domains( global_indices, ndivs, domain, &
                              pelist, flags, halo, extent, maskmap )
               call mpp_define_domains( global_indices, layout, domain, pelist, &
                              xflags, yflags, xhalo, yhalo,           &
                              xextent, yextent, maskmap, name )

Parameters

	global_indices	Defines the global domain.
	ndivs	The number of domain divisions required.
[in,out]	domain	Holds the resulting domain decomposition.
	pelist	List of PEs to which the domains are to be assigned.
	flags	An optional flag to pass additional information about the desired domain topology. Useful flags in a 1D decomposition include GLOBAL_DATA_DOMAIN and CYCLIC_GLOBAL_DOMAIN. Flags are integers: multiple flags may be added together. The flag values are public parameters available by use association.
	halo	Width of the halo.
	extent	Normally mpp_define_domains attempts an even division of the global domain across ndivs domains. The extent array can be used by the user to pass a custom domain division. The extent array has ndivs elements and holds the compute domain widths, which should add up to cover the global domain exactly.
	maskmap	Some divisions may be masked (maskmap=.FALSE.) to exclude them from the computation (e.g for ocean model domains that are all land). The maskmap array is dimensioned ndivs and contains .TRUE. values for any domain that must be included in the computation (default all). The pelist array length should match the number of domains included in the computation.

Example usage:

call mpp_define_domains( (/1,100/), 10, domain, &
                         flags=global_data_domain+cyclic_global_domain, halo=2 )

defines 10 compute domains spanning the range [1,100] of the global domain. The compute domains are non-overlapping blocks of 10. All the data domains are global, and with a halo of 2 span the range [-1:102]. And since the global domain has been declared to be cyclic, domain(9)next => domain(0) and domain(0)prev => domain(9). A field is allocated on the data domain, and computations proceed on the compute domain. A call to mpp_update_domains would fill in the values in the halo region:

call mpp_get_data_domain( domain, isd, ied ) !returns -1 and 102
call mpp_get_compute_domain( domain, is, ie ) !returns (1,10) on PE 0 ...
allocate( a(isd:ied) )
do i = is,ie
  a(i) = <perform computations>
end do
call mpp_update_domains( a, domain )

The call to mpp_update_domainsfills in the regions outside the compute domain. Since the global domain is cyclic, the values at i=(-1,0) are the same as at i=(99,100); and i=(101,102) are the same as i=(1,2).

The 2D version is just an extension of this syntax to two dimensions.

The 2D version of the above should generally be used in codes, including 1D-decomposed ones, if there is a possibility of future evolution toward 2D decomposition. The arguments are similar to the 1D case, except that now we have optional arguments flags, halo, extent and maskmap along two axes.

flags can now take an additional possible value to fold one or more edges. This is done by using flags FOLD_WEST_EDGE, FOLD_EAST_EDGE, FOLD_SOUTH_EDGE or FOLD_NORTH_EDGE. When a fold exists (e.g cylindrical domain), vector fields reverse sign upon crossing the fold. This parity reversal is performed only in the vector version of mpp_update_domains. In addition, shift operations may need to be applied to vector fields on staggered grids, also described in the vector interface to mpp_update_domains.

name is the name associated with the decomposition, e.g 'Ocean model'. If this argument is present, mpp_define_domains will print the domain decomposition generated to stdlog.

Examples: call mpp_define_domains( (/1,100,1,100/), (/2,2/), domain, xhalo=1 ) will create the following domain layout:

domain	domain(1)	domain(2)	domain(3)	domain(4)
Compute domain	1,50,1,50	51,100,1,50	1,50,51,100	51,100,51,100
Data domain	0,51,1,50	50,101,1,50	0,51,51,100	50,101,51,100

Again, we allocate arrays on the data domain, perform computations on the compute domain, and call mpp_update_domains to update the halo region.

If we wished to perfom a 1D decomposition along Y on the same global domain, we could use:

               call mpp_define_domains( (/1,100,1,100/), layout=(/4,1/), domain, xhalo=1 )

This will create the following domain layout:

domain	domain(1)	domain(2)	domain(3)	domain(4)
Compute domain	1,100,1,25	1,100,26,50	1,100,51,75	1,100,76,100
Data domain	0,101,1,25	0,101,26,50	0,101,51,75	1,101,76,100

Definition at line 891 of file mpp_domains.F90.

Private Member Functions
	mpp_define_domains1d
	mpp_define_domains2d

mpp_domains_mod::mpp_define_layout

interface mpp_domains_mod::mpp_define_layout

Retrieve layout associated with a domain decomposition. Given a global 2D domain and the number of divisions in the decomposition ndivs (usually the PE count unless some domains are masked) this calls returns a 2D domain layout. By default, mpp_define_layout will attempt to divide the 2D index space into domains that maintain the aspect ratio of the global domain. If this cannot be done, the algorithm favours domains that are longer in x than y, a preference that could improve vector performance.
Example usage:

call mpp_define_layout( global_indices, ndivs, layout )

Definition at line 773 of file mpp_domains.F90.

Private Member Functions
	mpp_define_layout2d

mpp_domains_mod::mpp_define_null_domain

interface mpp_domains_mod::mpp_define_null_domain

Defines a nullified 1D or 2D domain.

Example usage:

call mpp_define_null_domain(domain)

Definition at line 903 of file mpp_domains.F90.

Private Member Functions
	mpp_define_null_domain1d
	mpp_define_null_domain2d

mpp_domains_mod::mpp_do_check

interface mpp_domains_mod::mpp_do_check

Private interface to updates data domain of 3D field whose computational domains have been computed.

Definition at line 1555 of file mpp_domains.F90.

Private Member Functions
	mpp_do_check_c4_3d
	mpp_do_check_c8_3d
	mpp_do_check_i4_3d
	mpp_do_check_i8_3d
	mpp_do_check_r4_3d
	mpp_do_check_r4_3dv
	mpp_do_check_r8_3d
	mpp_do_check_r8_3dv

mpp_domains_mod::mpp_do_get_boundary

interface mpp_domains_mod::mpp_do_get_boundary

Definition at line 1656 of file mpp_domains.F90.

Private Member Functions
	mpp_do_get_boundary_r4_3d
	mpp_do_get_boundary_r4_3dv
	mpp_do_get_boundary_r8_3d
	mpp_do_get_boundary_r8_3dv

mpp_domains_mod::mpp_do_get_boundary_ad

interface mpp_domains_mod::mpp_do_get_boundary_ad

Definition at line 1664 of file mpp_domains.F90.

Private Member Functions
	mpp_do_get_boundary_ad_r4_3d
	mpp_do_get_boundary_ad_r4_3dv
	mpp_do_get_boundary_ad_r8_3d
	mpp_do_get_boundary_ad_r8_3dv

mpp_domains_mod::mpp_do_global_field

interface mpp_domains_mod::mpp_do_global_field

Private helper interface used by mpp_global_field.

Definition at line 1865 of file mpp_domains.F90.

Private Member Functions
	mpp_do_global_field2d_c4_3d
	mpp_do_global_field2d_c8_3d
	mpp_do_global_field2d_i4_3d
	mpp_do_global_field2d_i8_3d
	mpp_do_global_field2d_l4_3d
	mpp_do_global_field2d_l8_3d
	mpp_do_global_field2d_r4_3d
	mpp_do_global_field2d_r8_3d

mpp_domains_mod::mpp_do_global_field_ad

interface mpp_domains_mod::mpp_do_global_field_ad

Definition at line 1917 of file mpp_domains.F90.

Private Member Functions
	mpp_do_global_field2d_c4_3d_ad
	mpp_do_global_field2d_c8_3d_ad
	mpp_do_global_field2d_i4_3d_ad
	mpp_do_global_field2d_i8_3d_ad
	mpp_do_global_field2d_l4_3d_ad
	mpp_do_global_field2d_l8_3d_ad
	mpp_do_global_field2d_r4_3d_ad
	mpp_do_global_field2d_r8_3d_ad

mpp_domains_mod::mpp_do_group_update

interface mpp_domains_mod::mpp_do_group_update

Definition at line 1330 of file mpp_domains.F90.

Private Member Functions
	mpp_do_group_update_r4
	mpp_do_group_update_r8

mpp_domains_mod::mpp_do_redistribute

interface mpp_domains_mod::mpp_do_redistribute

Definition at line 1718 of file mpp_domains.F90.

Private Member Functions
	mpp_do_redistribute_c4_3d
	mpp_do_redistribute_c8_3d
	mpp_do_redistribute_i4_3d
	mpp_do_redistribute_i8_3d
	mpp_do_redistribute_l4_3d
	mpp_do_redistribute_l8_3d
	mpp_do_redistribute_r4_3d
	mpp_do_redistribute_r8_3d

mpp_domains_mod::mpp_do_update

interface mpp_domains_mod::mpp_do_update

Private interface used for mpp_update_domains.

Definition at line 1539 of file mpp_domains.F90.

Private Member Functions
	mpp_do_update_c4_3d
	mpp_do_update_c8_3d
	mpp_do_update_i4_3d
	mpp_do_update_i8_3d
	mpp_do_update_r4_3d
	mpp_do_update_r4_3dv
	mpp_do_update_r8_3d
	mpp_do_update_r8_3dv

mpp_domains_mod::mpp_do_update_ad

interface mpp_domains_mod::mpp_do_update_ad

Passes a data field from a unstructured grid to an structured grid
Example usage:

       call mpp_pass_UG_to_SG(UG_domain, field_UG, field_SG)

Definition at line 1608 of file mpp_domains.F90.

Private Member Functions
	mpp_do_update_ad_r4_3d
	mpp_do_update_ad_r4_3dv
	mpp_do_update_ad_r8_3d
	mpp_do_update_ad_r8_3dv

mpp_domains_mod::mpp_do_update_nest_coarse

interface mpp_domains_mod::mpp_do_update_nest_coarse

Used by mpp_update_nest_coarse to perform domain updates.

Definition at line 1471 of file mpp_domains.F90.

Private Member Functions
	mpp_do_update_nest_coarse_i4_3d
	mpp_do_update_nest_coarse_i8_3d
	mpp_do_update_nest_coarse_r4_3d
	mpp_do_update_nest_coarse_r4_3dv
	mpp_do_update_nest_coarse_r8_3d
	mpp_do_update_nest_coarse_r8_3dv

mpp_domains_mod::mpp_do_update_nest_fine

interface mpp_domains_mod::mpp_do_update_nest_fine

Definition at line 1415 of file mpp_domains.F90.

Private Member Functions
	mpp_do_update_nest_fine_i4_3d
	mpp_do_update_nest_fine_i8_3d
	mpp_do_update_nest_fine_r4_3d
	mpp_do_update_nest_fine_r4_3dv
	mpp_do_update_nest_fine_r8_3d
	mpp_do_update_nest_fine_r8_3dv

mpp_domains_mod::mpp_get_boundary

interface mpp_domains_mod::mpp_get_boundary

Get the boundary data for symmetric domain when the data is at C, E, or N-cell center.
mpp_get_boundary is used to get the boundary data for symmetric domain when the data is at C, E, or N-cell center. For cubic grid, the data should always at C-cell center.
Example usage:

               call mpp_get_boundary(domain, field, ebuffer, sbuffer, wbuffer, nbuffer)

Get boundary information from domain and field and store in buffers

Definition at line 1624 of file mpp_domains.F90.

Private Member Functions
	mpp_get_boundary_r4_2d
	mpp_get_boundary_r4_2dv
	mpp_get_boundary_r4_3d
	mpp_get_boundary_r4_3dv
	mpp_get_boundary_r8_2d
	mpp_get_boundary_r8_2dv
	mpp_get_boundary_r8_3d
	mpp_get_boundary_r8_3dv

mpp_domains_mod::mpp_get_boundary_ad

interface mpp_domains_mod::mpp_get_boundary_ad

Definition at line 1644 of file mpp_domains.F90.

Private Member Functions
	mpp_get_boundary_ad_r4_2d
	mpp_get_boundary_ad_r4_2dv
	mpp_get_boundary_ad_r4_3d
	mpp_get_boundary_ad_r4_3dv
	mpp_get_boundary_ad_r8_2d
	mpp_get_boundary_ad_r8_2dv
	mpp_get_boundary_ad_r8_3d
	mpp_get_boundary_ad_r8_3dv

mpp_domains_mod::mpp_get_compute_domain

interface mpp_domains_mod::mpp_get_compute_domain

These routines retrieve the axis specifications associated with the compute domains. The domain is a derived type with private elements. These routines retrieve the axis specifications associated with the compute domains The 2D version of these is a simple extension of 1D.
Example usage:

       call mpp_get_compute_domain(domain_1D, is, ie)
       call mpp_get_compute_domain(domain_2D, is, ie, js, je)

Definition at line 2192 of file mpp_domains.F90.

Private Member Functions
	mpp_get_compute_domain1d
	mpp_get_compute_domain2d

mpp_domains_mod::mpp_get_compute_domains

interface mpp_domains_mod::mpp_get_compute_domains

Retrieve the entire array of compute domain extents associated with a decomposition.

Parameters

	domain	2D domain
[out]	xbegin,ybegin	x and y domain starting indices
[out]	xsize,ysize	x and y domain sizes Example usage: call mpp_get_compute_domains( domain, xbegin, xend, xsize, & ybegin, yend, ysize )

Definition at line 2207 of file mpp_domains.F90.

Private Member Functions
	mpp_get_compute_domains1d
	mpp_get_compute_domains2d

mpp_domains_mod::mpp_get_data_domain

interface mpp_domains_mod::mpp_get_data_domain

These routines retrieve the axis specifications associated with the data domains. The domain is a derived type with private elements. These routines retrieve the axis specifications associated with the data domains. The 2D version of these is a simple extension of 1D.
Example usage:

               call mpp_get_data_domain(domain_1d, isd, ied)
               call mpp_get_data_domain(domain_2d, isd, ied, jsd, jed)

Definition at line 2227 of file mpp_domains.F90.

Private Member Functions
	mpp_get_data_domain1d
	mpp_get_data_domain2d

mpp_domains_mod::mpp_get_domain_extents

interface mpp_domains_mod::mpp_get_domain_extents

Definition at line 2261 of file mpp_domains.F90.

Private Member Functions
	mpp_get_domain_extents1d
	mpp_get_domain_extents2d

mpp_domains_mod::mpp_get_f2c_index

interface mpp_domains_mod::mpp_get_f2c_index

Get the index of the data passed from fine grid to coarse grid.
Example usage:

       call mpp_get_F2C_index(nest_domain, is_coarse, ie_coarse, js_coarse, je_coarse,
                       is_fine, ie_fine, js_fine, je_fine, nest_level, position)

Definition at line 1492 of file mpp_domains.F90.

Private Member Functions
	mpp_get_f2c_index_coarse
	mpp_get_f2c_index_fine

mpp_domains_mod::mpp_get_global_domain

interface mpp_domains_mod::mpp_get_global_domain

These routines retrieve the axis specifications associated with the global domains. The domain is a derived type with private elements. These routines retrieve the axis specifications associated with the global domains. The 2D version of these is a simple extension of 1D.
Example usage:

               call mpp_get_global_domain(domain_1d, isg, ieg)
               call mpp_get_global_domain(domain_2d, isg, ieg, jsg, jeg)

Definition at line 2241 of file mpp_domains.F90.

Private Member Functions
	mpp_get_global_domain1d
	mpp_get_global_domain2d

mpp_domains_mod::mpp_get_global_domains

interface mpp_domains_mod::mpp_get_global_domains

Definition at line 2213 of file mpp_domains.F90.

Private Member Functions
	mpp_get_global_domains1d
	mpp_get_global_domains2d

mpp_domains_mod::mpp_get_layout

interface mpp_domains_mod::mpp_get_layout

Retrieve layout associated with a domain decomposition The 1D version of this call returns the number of divisions that was assigned to this decomposition axis. The 2D version of this call returns an array of dimension 2 holding the results on two axes.
Example usage:

               call mpp_get_layout( domain, layout )

Definition at line 2329 of file mpp_domains.F90.

Private Member Functions
	mpp_get_layout1d
	mpp_get_layout2d

mpp_domains_mod::mpp_get_memory_domain

interface mpp_domains_mod::mpp_get_memory_domain

These routines retrieve the axis specifications associated with the memory domains. The domain is a derived type with private elements. These routines retrieve the axis specifications associated with the memory domains. The 2D version of these is a simple extension of 1D.
Example usage:

               call mpp_get_memory_domain(domain_1d, ism, iem)
               call mpp_get_memory_domain(domain_2d, ism, iem, jsm, jem)

Definition at line 2255 of file mpp_domains.F90.

Private Member Functions
	mpp_get_memory_domain1d
	mpp_get_memory_domain2d

mpp_domains_mod::mpp_get_neighbor_pe

interface mpp_domains_mod::mpp_get_neighbor_pe

Retrieve PE number of a neighboring domain.

Given a 1-D or 2-D domain decomposition, this call allows users to retrieve the PE number of an adjacent PE-domain while taking into account that the domain may have holes (masked) and/or have cyclic boundary conditions and/or a folded edge. Which PE-domain will be retrived will depend on "direction": +1 (right) or -1 (left) for a 1-D domain decomposition and either NORTH, SOUTH, EAST, WEST, NORTH_EAST, SOUTH_EAST, SOUTH_WEST, or NORTH_WEST for a 2-D decomposition. If no neighboring domain exists (masked domain), then the returned "pe" value will be set to NULL_PE.

Example usage:

               call mpp_get_neighbor_pe( domain1d, direction=+1   , pe)

Set pe to the neighbor pe number that is to the right of the current pe

               call mpp_get_neighbor_pe( domain2d, direction=NORTH, pe)

Get neighbor pe number that's above/north of the current pe

Definition at line 2147 of file mpp_domains.F90.

Private Member Functions
	mpp_get_neighbor_pe_1d
	mpp_get_neighbor_pe_2d

mpp_domains_mod::mpp_get_pelist

interface mpp_domains_mod::mpp_get_pelist

Retrieve list of PEs associated with a domain decomposition. The 1D version of this call returns an array of the PEs assigned to this 1D domain decomposition. In addition the optional argument pos may be used to retrieve the 0-based position of the domain local to the calling PE, i.e., domain%list(pos)%pe is the local PE, as returned by mpp_pe(). The 2D version of this call is identical to 1D version.

Definition at line 2316 of file mpp_domains.F90.

Private Member Functions
	mpp_get_pelist1d
	mpp_get_pelist2d

mpp_domains_mod::mpp_global_field

interface mpp_domains_mod::mpp_global_field

Fill in a global array from domain-decomposed arrays.

mpp_global_field is used to get an entire domain-decomposed array on each PE. MPP_TYPE_ can be of type complex, integer, logical or real; of 4-byte or 8-byte kind; of rank up to 5.

All PEs in a domain decomposition must call mpp_global_field, and each will have a complete global field at the end. Please note that a global array of rank 3 or higher could occupy a lot of memory.

Parameters

	domain	2D domain
	local	Data dimensioned on either the compute or data domains of 'domain'
[out]	global	output data dimensioned on the corresponding global domain
	flags	can be either XONLY or YONLY parameters to specify a globalization on one axis only

Example usage:

call mpp_global_field( domain, local, global, flags )

Definition at line 1784 of file mpp_domains.F90.

Private Member Functions
	mpp_global_field2d_c4_2d
	mpp_global_field2d_c4_3d
	mpp_global_field2d_c4_4d
	mpp_global_field2d_c4_5d
	mpp_global_field2d_c8_2d
	mpp_global_field2d_c8_3d
	mpp_global_field2d_c8_4d
	mpp_global_field2d_c8_5d
	mpp_global_field2d_i4_2d
	mpp_global_field2d_i4_3d
	mpp_global_field2d_i4_4d
	mpp_global_field2d_i4_5d
	mpp_global_field2d_i8_2d
	mpp_global_field2d_i8_3d
	mpp_global_field2d_i8_4d
	mpp_global_field2d_i8_5d
	mpp_global_field2d_l4_2d
	mpp_global_field2d_l4_3d
	mpp_global_field2d_l4_4d
	mpp_global_field2d_l4_5d
	mpp_global_field2d_l8_2d
	mpp_global_field2d_l8_3d
	mpp_global_field2d_l8_4d
	mpp_global_field2d_l8_5d
	mpp_global_field2d_r4_2d
	mpp_global_field2d_r4_3d
	mpp_global_field2d_r4_4d
	mpp_global_field2d_r4_5d
	mpp_global_field2d_r8_2d
	mpp_global_field2d_r8_3d
	mpp_global_field2d_r8_4d
	mpp_global_field2d_r8_5d

mpp_domains_mod::mpp_global_field_ad

interface mpp_domains_mod::mpp_global_field_ad

Definition at line 1824 of file mpp_domains.F90.

Private Member Functions
	mpp_global_field2d_c4_2d_ad
	mpp_global_field2d_c4_3d_ad
	mpp_global_field2d_c4_4d_ad
	mpp_global_field2d_c4_5d_ad
	mpp_global_field2d_c8_2d_ad
	mpp_global_field2d_c8_3d_ad
	mpp_global_field2d_c8_4d_ad
	mpp_global_field2d_c8_5d_ad
	mpp_global_field2d_i4_2d_ad
	mpp_global_field2d_i4_3d_ad
	mpp_global_field2d_i4_4d_ad
	mpp_global_field2d_i4_5d_ad
	mpp_global_field2d_i8_2d_ad
	mpp_global_field2d_i8_3d_ad
	mpp_global_field2d_i8_4d_ad
	mpp_global_field2d_i8_5d_ad
	mpp_global_field2d_l4_2d_ad
	mpp_global_field2d_l4_3d_ad
	mpp_global_field2d_l4_4d_ad
	mpp_global_field2d_l4_5d_ad
	mpp_global_field2d_l8_2d_ad
	mpp_global_field2d_l8_3d_ad
	mpp_global_field2d_l8_4d_ad
	mpp_global_field2d_l8_5d_ad
	mpp_global_field2d_r4_2d_ad
	mpp_global_field2d_r4_3d_ad
	mpp_global_field2d_r4_4d_ad
	mpp_global_field2d_r4_5d_ad
	mpp_global_field2d_r8_2d_ad
	mpp_global_field2d_r8_3d_ad
	mpp_global_field2d_r8_4d_ad
	mpp_global_field2d_r8_5d_ad

mpp_domains_mod::mpp_global_field_ug

interface mpp_domains_mod::mpp_global_field_ug

Same functionality as mpp_global_field but for unstructured domains.

Definition at line 1897 of file mpp_domains.F90.

Private Member Functions
	mpp_global_field2d_ug_i4_2d
	mpp_global_field2d_ug_i4_3d
	mpp_global_field2d_ug_i4_4d
	mpp_global_field2d_ug_i4_5d
	mpp_global_field2d_ug_i8_2d
	mpp_global_field2d_ug_i8_3d
	mpp_global_field2d_ug_i8_4d
	mpp_global_field2d_ug_i8_5d
	mpp_global_field2d_ug_r4_2d
	mpp_global_field2d_ug_r4_3d
	mpp_global_field2d_ug_r4_4d
	mpp_global_field2d_ug_r4_5d
	mpp_global_field2d_ug_r8_2d
	mpp_global_field2d_ug_r8_3d
	mpp_global_field2d_ug_r8_4d
	mpp_global_field2d_ug_r8_5d

mpp_domains_mod::mpp_global_max

interface mpp_domains_mod::mpp_global_max

Global max of domain-decomposed arrays.
mpp_global_max is used to get the maximum value of a domain-decomposed array on each PE. MPP_TYPE_can be of type integer or real; of 4-byte or 8-byte kind; of rank up to 5. The dimension of locus must equal the rank of field.

All PEs in a domain decomposition must call mpp_global_max, and each will have the result upon exit. The function mpp_global_min, with an identical syntax. is also available.

Parameters

domain	2D domain
field	field data dimensioned on either the compute or data domains of 'domain'
locus	If present, van be used to retrieve the location of the maximum

Example usage: mpp_global_max( domain, field, locus )

Definition at line 1949 of file mpp_domains.F90.

Private Member Functions
	mpp_global_max_i4_2d
	mpp_global_max_i4_3d
	mpp_global_max_i4_4d
	mpp_global_max_i4_5d
	mpp_global_max_i8_2d
	mpp_global_max_i8_3d
	mpp_global_max_i8_4d
	mpp_global_max_i8_5d
	mpp_global_max_r4_2d
	mpp_global_max_r4_3d
	mpp_global_max_r4_4d
	mpp_global_max_r4_5d
	mpp_global_max_r8_2d
	mpp_global_max_r8_3d
	mpp_global_max_r8_4d
	mpp_global_max_r8_5d

mpp_domains_mod::mpp_global_min

interface mpp_domains_mod::mpp_global_min

Global min of domain-decomposed arrays.
mpp_global_min is used to get the minimum value of a domain-decomposed array on each PE. MPP_TYPE_can be of type integer or real; of 4-byte or 8-byte kind; of rank up to 5. The dimension of locus must equal the rank of field.

All PEs in a domain decomposition must call mpp_global_min, and each will have the result upon exit. The function mpp_global_max, with an identical syntax. is also available.

Parameters

domain	2D domain
field	field data dimensioned on either the compute or data domains of 'domain'
locus	If present, van be used to retrieve the location of the minimum

Example usage: mpp_global_min( domain, field, locus )

Definition at line 1985 of file mpp_domains.F90.

Private Member Functions
	mpp_global_min_i4_2d
	mpp_global_min_i4_3d
	mpp_global_min_i4_4d
	mpp_global_min_i4_5d
	mpp_global_min_i8_2d
	mpp_global_min_i8_3d
	mpp_global_min_i8_4d
	mpp_global_min_i8_5d
	mpp_global_min_r4_2d
	mpp_global_min_r4_3d
	mpp_global_min_r4_4d
	mpp_global_min_r4_5d
	mpp_global_min_r8_2d
	mpp_global_min_r8_3d
	mpp_global_min_r8_4d
	mpp_global_min_r8_5d

mpp_domains_mod::mpp_global_sum

interface mpp_domains_mod::mpp_global_sum

Global sum of domain-decomposed arrays.
mpp_global_sum is used to get the sum of a domain-decomposed array on each PE. MPP_TYPE_ can be of type integer, complex, or real; of 4-byte or 8-byte kind; of rank up to 5.

Parameters

domain	2D domain
field	field data dimensioned on either the compute or data domain of 'domain'
flags	If present must have the value BITWISE_EXACT_SUM. This produces a sum that is guaranteed to produce the identical result irrespective of how the domain is decomposed. This method does the sum first along the ranks beyond 2, and then calls mpp_global_field to produce a global 2D array which is then summed. The default method, which is considerably faster, does a local sum followed by mpp_sum across the domain decomposition.

Example usage: call mpp_global_sum( domain, field, flags )

Note

All PEs in a domain decomposition must call mpp_global_sum, and each will have the result upon exit.

Definition at line 2023 of file mpp_domains.F90.

Private Member Functions
	mpp_global_sum_c4_2d
	mpp_global_sum_c4_3d
	mpp_global_sum_c4_4d
	mpp_global_sum_c4_5d
	mpp_global_sum_c8_2d
	mpp_global_sum_c8_3d
	mpp_global_sum_c8_4d
	mpp_global_sum_c8_5d
	mpp_global_sum_i4_2d
	mpp_global_sum_i4_3d
	mpp_global_sum_i4_4d
	mpp_global_sum_i4_5d
	mpp_global_sum_i8_2d
	mpp_global_sum_i8_3d
	mpp_global_sum_i8_4d
	mpp_global_sum_i8_5d
	mpp_global_sum_r4_2d
	mpp_global_sum_r4_3d
	mpp_global_sum_r4_4d
	mpp_global_sum_r4_5d
	mpp_global_sum_r8_2d
	mpp_global_sum_r8_3d
	mpp_global_sum_r8_4d
	mpp_global_sum_r8_5d

mpp_domains_mod::mpp_global_sum_ad

interface mpp_domains_mod::mpp_global_sum_ad

Definition at line 2090 of file mpp_domains.F90.

Private Member Functions
	mpp_global_sum_ad_c4_2d
	mpp_global_sum_ad_c4_3d
	mpp_global_sum_ad_c4_4d
	mpp_global_sum_ad_c4_5d
	mpp_global_sum_ad_c8_2d
	mpp_global_sum_ad_c8_3d
	mpp_global_sum_ad_c8_4d
	mpp_global_sum_ad_c8_5d
	mpp_global_sum_ad_i4_2d
	mpp_global_sum_ad_i4_3d
	mpp_global_sum_ad_i4_4d
	mpp_global_sum_ad_i4_5d
	mpp_global_sum_ad_i8_2d
	mpp_global_sum_ad_i8_3d
	mpp_global_sum_ad_i8_4d
	mpp_global_sum_ad_i8_5d
	mpp_global_sum_ad_r4_2d
	mpp_global_sum_ad_r4_3d
	mpp_global_sum_ad_r4_4d
	mpp_global_sum_ad_r4_5d
	mpp_global_sum_ad_r8_2d
	mpp_global_sum_ad_r8_3d
	mpp_global_sum_ad_r8_4d
	mpp_global_sum_ad_r8_5d

mpp_domains_mod::mpp_global_sum_tl

interface mpp_domains_mod::mpp_global_sum_tl

Definition at line 2056 of file mpp_domains.F90.

Private Member Functions
	mpp_global_sum_tl_c4_2d
	mpp_global_sum_tl_c4_3d
	mpp_global_sum_tl_c4_4d
	mpp_global_sum_tl_c4_5d
	mpp_global_sum_tl_c8_2d
	mpp_global_sum_tl_c8_3d
	mpp_global_sum_tl_c8_4d
	mpp_global_sum_tl_c8_5d
	mpp_global_sum_tl_i4_2d
	mpp_global_sum_tl_i4_3d
	mpp_global_sum_tl_i4_4d
	mpp_global_sum_tl_i4_5d
	mpp_global_sum_tl_i8_2d
	mpp_global_sum_tl_i8_3d
	mpp_global_sum_tl_i8_4d
	mpp_global_sum_tl_i8_5d
	mpp_global_sum_tl_r4_2d
	mpp_global_sum_tl_r4_3d
	mpp_global_sum_tl_r4_4d
	mpp_global_sum_tl_r4_5d
	mpp_global_sum_tl_r8_2d
	mpp_global_sum_tl_r8_3d
	mpp_global_sum_tl_r8_4d
	mpp_global_sum_tl_r8_5d

mpp_domains_mod::mpp_group_update_type

type mpp_domains_mod::mpp_group_update_type

used for updates on a group

Definition at line 575 of file mpp_domains.F90.

Collaboration diagram for mpp_group_update_type:

Private Attributes
integer(i8_kind), dimension(max_domain_fields)	addrs_s
integer(i8_kind), dimension(max_domain_fields)	addrs_x
integer(i8_kind), dimension(max_domain_fields)	addrs_y
integer, dimension(maxoverlap)	buffer_pos_recv
integer, dimension(maxoverlap)	buffer_pos_send
integer	buffer_start_pos = -1
integer	ehalo_s =0
integer	ehalo_v =0
integer	flags_s =0
integer	flags_v =0
integer, dimension(maxoverlap)	from_pe
integer	gridtype =0
integer	ie_s =0
integer	ie_x =0
integer	ie_y =0
logical	initialized = .FALSE.
integer	is_s =0
integer	is_x =0
integer	is_y =0
integer	isize_s =0
integer	isize_x =0
integer	isize_y =0
integer	je_s =0
integer	je_x =0
integer	je_y =0
integer	js_s =0
integer	js_x =0
integer	js_y =0
integer	jsize_s =0
integer	jsize_x =0
integer	jsize_y =0
logical	k_loop_inside = .TRUE.
integer	ksize_s =1
integer	ksize_v =1
integer	nhalo_s =0
integer	nhalo_v =0
logical	nonsym_edge = .FALSE.
integer	npack =0
integer	nrecv =0
integer	nscalar = 0
integer	nsend =0
integer	nunpack =0
integer	nvector = 0
integer, dimension(maxoverlap)	pack_buffer_pos
integer, dimension(maxoverlap)	pack_ie
integer, dimension(maxoverlap)	pack_is
integer, dimension(maxoverlap)	pack_je
integer, dimension(maxoverlap)	pack_js
integer, dimension(maxoverlap)	pack_rotation
integer, dimension(maxoverlap)	pack_size
integer, dimension(maxoverlap)	pack_type
integer	position =0
logical, dimension(8)	recv_s
integer, dimension(maxoverlap)	recv_size
logical, dimension(8)	recv_x
logical, dimension(8)	recv_y
integer, dimension(max_request)	request_recv
integer, dimension(max_request)	request_send
integer	reset_index_s = 0
integer	reset_index_v = 0
integer, dimension(maxoverlap)	send_size
integer	shalo_s =0
integer	shalo_v =0
integer, dimension(maxoverlap)	to_pe
integer	tot_msgsize = 0
integer, dimension(max_request)	type_recv
integer, dimension(maxoverlap)	unpack_buffer_pos
integer, dimension(maxoverlap)	unpack_ie
integer, dimension(maxoverlap)	unpack_is
integer, dimension(maxoverlap)	unpack_je
integer, dimension(maxoverlap)	unpack_js
integer, dimension(maxoverlap)	unpack_rotation
integer, dimension(maxoverlap)	unpack_size
integer, dimension(maxoverlap)	unpack_type
integer	whalo_s =0
integer	whalo_v =0

mpp_domains_mod::mpp_modify_domain

interface mpp_domains_mod::mpp_modify_domain

Modifies the extents (compute, data and global) of a given domain.

Definition at line 926 of file mpp_domains.F90.

Private Member Functions
	mpp_modify_domain1d
	mpp_modify_domain2d

mpp_domains_mod::mpp_nullify_domain_list

interface mpp_domains_mod::mpp_nullify_domain_list

Nullify domain list. This interface is needed in mpp_domains_test. 1-D case can be added in if needed.
Example usage:

               call mpp_nullify_domain_list(domain)

Definition at line 2346 of file mpp_domains.F90.

Private Member Functions
	nullify_domain2d_list

mpp_domains_mod::mpp_pass_sg_to_ug

interface mpp_domains_mod::mpp_pass_sg_to_ug

Passes data from a structured grid to an unstructured grid
Example usage:

       call mpp_pass_SG_to_UG(domain, sg_data, ug_data)

Definition at line 1575 of file mpp_domains.F90.

Private Member Functions
	mpp_pass_sg_to_ug_i4_2d
	mpp_pass_sg_to_ug_i4_3d
	mpp_pass_sg_to_ug_l4_2d
	mpp_pass_sg_to_ug_l4_3d
	mpp_pass_sg_to_ug_r4_2d
	mpp_pass_sg_to_ug_r4_3d
	mpp_pass_sg_to_ug_r8_2d
	mpp_pass_sg_to_ug_r8_3d

mpp_domains_mod::mpp_pass_ug_to_sg

interface mpp_domains_mod::mpp_pass_ug_to_sg

Passes a data field from a structured grid to an unstructured grid
Example usage:

       call mpp_pass_SG_to_UG(SG_domain, field_SG, field_UG)

Definition at line 1591 of file mpp_domains.F90.

Private Member Functions
	mpp_pass_ug_to_sg_i4_2d
	mpp_pass_ug_to_sg_i4_3d
	mpp_pass_ug_to_sg_l4_2d
	mpp_pass_ug_to_sg_l4_3d
	mpp_pass_ug_to_sg_r4_2d
	mpp_pass_ug_to_sg_r4_3d
	mpp_pass_ug_to_sg_r8_2d
	mpp_pass_ug_to_sg_r8_3d

mpp_domains_mod::mpp_redistribute

interface mpp_domains_mod::mpp_redistribute

Reorganization of distributed global arrays.
mpp_redistribute is used to reorganize a distributed array. MPP_TYPE_can be of type integer, complex, or real; of 4-byte or 8-byte kind; of rank up to 5.
Example usage: call mpp_redistribute( domain_in, field_in, domain_out, field_out )

Definition at line 1678 of file mpp_domains.F90.

Private Member Functions
	mpp_redistribute_c4_2d
	mpp_redistribute_c4_3d
	mpp_redistribute_c4_4d
	mpp_redistribute_c4_5d
	mpp_redistribute_c8_2d
	mpp_redistribute_c8_3d
	mpp_redistribute_c8_4d
	mpp_redistribute_c8_5d
	mpp_redistribute_i4_2d
	mpp_redistribute_i4_3d
	mpp_redistribute_i4_4d
	mpp_redistribute_i4_5d
	mpp_redistribute_i8_2d
	mpp_redistribute_i8_3d
	mpp_redistribute_i8_4d
	mpp_redistribute_i8_5d
	mpp_redistribute_r4_2d
	mpp_redistribute_r4_3d
	mpp_redistribute_r4_4d
	mpp_redistribute_r4_5d
	mpp_redistribute_r8_2d
	mpp_redistribute_r8_3d
	mpp_redistribute_r8_4d
	mpp_redistribute_r8_5d

mpp_domains_mod::mpp_reset_group_update_field

interface mpp_domains_mod::mpp_reset_group_update_field

Definition at line 1360 of file mpp_domains.F90.

Private Member Functions
	mpp_reset_group_update_field_r4_2d
	mpp_reset_group_update_field_r4_2dv
	mpp_reset_group_update_field_r4_3d
	mpp_reset_group_update_field_r4_3dv
	mpp_reset_group_update_field_r4_4d
	mpp_reset_group_update_field_r4_4dv
	mpp_reset_group_update_field_r8_2d
	mpp_reset_group_update_field_r8_2dv
	mpp_reset_group_update_field_r8_3d
	mpp_reset_group_update_field_r8_3dv
	mpp_reset_group_update_field_r8_4d
	mpp_reset_group_update_field_r8_4dv

mpp_domains_mod::mpp_set_compute_domain

interface mpp_domains_mod::mpp_set_compute_domain

These routines set the axis specifications associated with the compute domains. The domain is a derived type with private elements. These routines set the axis specifications associated with the compute domains The 2D version of these is a simple extension of 1D.
Example usage:

               call mpp_get_data_domain(domain_1d, isd, ied)
               call mpp_get_data_domain(domain_2d, isd, ied, jsd, jed)

Definition at line 2275 of file mpp_domains.F90.

Private Member Functions
	mpp_set_compute_domain1d
	mpp_set_compute_domain2d

mpp_domains_mod::mpp_set_data_domain

interface mpp_domains_mod::mpp_set_data_domain

These routines set the axis specifications associated with the data domains. The domain is a derived type with private elements. These routines set the axis specifications associated with the data domains. The 2D version of these is a simple extension of 1D.
Example usage:

               call mpp_set_data_domain(domain_1d, isd, ied)
               call mpp_set_data_domain(domain_2d, isd, ied, jsd, jed)

Definition at line 2289 of file mpp_domains.F90.

Private Member Functions
	mpp_set_data_domain1d
	mpp_set_data_domain2d

mpp_domains_mod::mpp_set_global_domain

interface mpp_domains_mod::mpp_set_global_domain

These routines set the axis specifications associated with the global domains. The domain is a derived type with private elements. These routines set the axis specifications associated with the global domains. The 2D version of these is a simple extension of 1D.
Example usage:

               call mpp_set_global_domain(domain_1d, isg, ieg)
               call mpp_set_global_domain(domain_2d, isg, ieg, jsg, jeg)

Definition at line 2303 of file mpp_domains.F90.

Private Member Functions
	mpp_set_global_domain1d
	mpp_set_global_domain2d

mpp_domains_mod::mpp_start_do_update

interface mpp_domains_mod::mpp_start_do_update

Private interface used for non blocking updates.

Definition at line 1277 of file mpp_domains.F90.

Private Member Functions
	mpp_start_do_update_i4_3d
	mpp_start_do_update_i8_3d
	mpp_start_do_update_r4_3d
	mpp_start_do_update_r4_3dv
	mpp_start_do_update_r8_3d
	mpp_start_do_update_r8_3dv

mpp_domains_mod::mpp_start_group_update

interface mpp_domains_mod::mpp_start_group_update

Starts non-blocking group update Must be followed up with a call to mpp_complete_group_update mpp_group_update_type can be created with mpp_create_group_update.

Parameters

[in,out]	type(mpp_group_update_type)	group type created for group update
[in,out]	type(domain2D)	domain to update

Definition at line 1342 of file mpp_domains.F90.

Private Member Functions
	mpp_start_group_update_r4
	mpp_start_group_update_r8

mpp_domains_mod::mpp_start_update_domains

interface mpp_domains_mod::mpp_start_update_domains

Interface to start halo updates mpp_start_update_domains is used to start a halo update of a domain-decomposed array on each PE. MPP_TYPE_ can be of type complex, integer, logical or real; of 4-byte or 8-byte kind; of rank up to 5. The vector version (with two input data fields) is only present for \ereal types.

\empp_start_update_domains must be paired together with \empp_complete_update_domains. In mpp_start_update_domains, a buffer will be pre-post to receive (non-blocking) the data and data on computational domain will be packed and sent (non-blocking send) to other processor. In mpp_complete_update_domains, buffer will be unpacked to fill the halo and mpp_sync_self will be called to to ensure communication safe at the last call of mpp_complete_update_domains.

Each mpp_update_domains can be replaced by the combination of mpp_start_update_domains and mpp_complete_update_domains. The arguments in mpp_start_update_domains and mpp_complete_update_domains should be the exact the same as in mpp_update_domains to be replaced except no optional argument "complete". The following are examples on how to replace mpp_update_domains with mpp_start_update_domains/mpp_complete_update_domains.

Example 1: Replace one scalar mpp_update_domains.

Replace

call mpp_update_domains(data, domain, flags=update_flags)

with

id_update = mpp_start_update_domains(data, domain, flags=update_flags)
...( doing some computation )
call mpp_complete_update_domains(id_update, data, domain, flags=update_flags)

Example 2: Replace group scalar mpp_update_domains

Replace

call mpp_update_domains(data_1, domain, flags=update_flags, complete=.false.)
.... ( other n-2 call mpp_update_domains with complete = .false. )
call mpp_update_domains(data_n, domain, flags=update_flags, complete=.true. )

With

id_up_1 = mpp_start_update_domains(data_1, domain, flags=update_flags)
.... ( other n-2 call mpp_start_update_domains )
id_up_n = mpp_start_update_domains(data_n, domain, flags=update_flags)

..... ( doing some computation )

call mpp_complete_update_domains(id_up_1, data_1, domain, flags=update_flags)
.... ( other n-2 call mpp_complete_update_domains )
call mpp_complete_update_domains(id_up_n, data_n, domain, flags=update_flags)

Example 3: Replace group CGRID_NE vector, mpp_update_domains

Replace

call mpp_update_domains(u_1, v_1, domain, flags=update_flgs, gridtype=CGRID_NE, complete=.false.)
.... ( other n-2 call mpp_update_domains with complete = .false. )
call mpp_update_domains(u_1, v_1, domain, flags=update_flags, gridtype=CGRID_NE, complete=.true. )

with

id_up_1 = mpp_start_update_domains(u_1, v_1, domain, flags=update_flags, gridtype=CGRID_NE)
.... ( other n-2 call mpp_start_update_domains )
id_up_n = mpp_start_update_domains(u_n, v_n, domain, flags=update_flags, gridtype=CGRID_NE)

..... ( doing some computation )

call mpp_complete_update_domains(id_up_1, u_1, v_1, domain, flags=update_flags, gridtype=CGRID_NE)
.... ( other n-2 call mpp_complete_update_domains )
call mpp_complete_update_domains(id_up_n, u_n, v_n, domain, flags=update_flags, gridtype=CGRID_NE)

For 2D domain updates, if there are halos present along both x and y, we can choose to update one only, by specifying flags=XUPDATE or flags=YUPDATE. In addition, one-sided updates can be performed by setting flags to any combination of WUPDATE, EUPDATE, SUPDATE and NUPDATE, to update the west, east, north and south halos respectively. Any combination of halos may be used by adding the requisite flags, e.g: flags=XUPDATE+SUPDATE or flags=EUPDATE+WUPDATE+SUPDATE will update the east, west and south halos.

If a call to mpp_start_update_domains/mpp_complete_update_domains involves at least one E-W halo and one N-S halo, the corners involved will also be updated, i.e, in the example above, the SE and SW corners will be updated.

If flags is not supplied, that is equivalent to flags=XUPDATE+YUPDATE.

The vector version is passed the x and y components of a vector field in tandem, and both are updated upon return. They are passed together to treat parity issues on various grids. For example, on a cubic sphere projection, the x and y components may be interchanged when passing from an equatorial cube face to a polar face. For grids with folds, vector components change sign on crossing the fold. Paired scalar quantities can also be passed with the vector version if flags=SCALAR_PAIR, in which case components are appropriately interchanged, but signs are not.

Special treatment at boundaries such as folds is also required for staggered grids. The following types of staggered grids are recognized:
1) AGRID: values are at grid centers.
2) BGRID_NE: vector fields are at the NE vertex of a grid cell, i.e: the array elements u(i,j) and v(i,j) are actually at (i+½,j+½) with respect to the grid centers.
3) BGRID_SW: vector fields are at the SW vertex of a grid cell, i.e., the array elements u(i,j) and v(i,j) are actually at (i-½,j-½) with respect to the grid centers.
4) CGRID_NE: vector fields are at the N and E faces of a grid cell, i.e: the array elements u(i,j) and v(i,j) are actually at (i+½,j) and (i,j+½) with respect to the grid centers.
5) CGRID_SW: vector fields are at the S and W faces of a grid cell, i.e: the array elements u(i,j) and v(i,j) are actually at (i-½,j) and (i,j-½) with respect to the grid centers.

The gridtypes listed above are all available by use association as integer parameters. If vector fields are at staggered locations, the optional argument gridtype must be appropriately set for correct treatment at boundaries.
It is safe to apply vector field updates to the appropriate arrays irrespective of the domain topology: if the topology requires no special treatment of vector fields, specifying gridtype will do no harm.

mpp_start_update_domains/mpp_complete_update_domains internally buffers the data being sent and received into single messages for efficiency. A turnable internal buffer area in memory is provided for this purpose by mpp_domains_mod. The size of this buffer area can be set by the user by calling mpp_domains_set_stack_size.
Example usage:

            call mpp_start_update_domains( field, domain, flags )
            call mpp_complete_update_domains( field, domain, flags )

Definition at line 1193 of file mpp_domains.F90.

Private Member Functions
	mpp_start_update_domain2d_i4_2d
	mpp_start_update_domain2d_i4_3d
	mpp_start_update_domain2d_i4_4d
	mpp_start_update_domain2d_i4_5d
	mpp_start_update_domain2d_i8_2d
	mpp_start_update_domain2d_i8_3d
	mpp_start_update_domain2d_i8_4d
	mpp_start_update_domain2d_i8_5d
	mpp_start_update_domain2d_r4_2d
	mpp_start_update_domain2d_r4_2dv
	mpp_start_update_domain2d_r4_3d
	mpp_start_update_domain2d_r4_3dv
	mpp_start_update_domain2d_r4_4d
	mpp_start_update_domain2d_r4_4dv
	mpp_start_update_domain2d_r4_5d
	mpp_start_update_domain2d_r4_5dv
	mpp_start_update_domain2d_r8_2d
	mpp_start_update_domain2d_r8_2dv
	mpp_start_update_domain2d_r8_3d
	mpp_start_update_domain2d_r8_3dv
	mpp_start_update_domain2d_r8_4d
	mpp_start_update_domain2d_r8_4dv
	mpp_start_update_domain2d_r8_5d
	mpp_start_update_domain2d_r8_5dv

mpp_domains_mod::mpp_update_domains

interface mpp_domains_mod::mpp_update_domains

Performs halo updates for a given domain.

Used to perform a halo update of a domain-decomposed array on each PE. MPP_TYPE can be of type complex, integer, logical or real of 4-byte or 8-byte kind; of rank up to 5. The vector version (with two input data fields) is only present for real types. For 2D domain updates, if there are halos present along both x and y, we can choose to update one only, by specifying flags=XUPDATE or flags=YUPDATE. In addition, one-sided updates can be performed by setting flags to any combination of WUPDATE, EUPDATE, SUPDATE and NUPDATE to update the west, east, north and south halos respectively. Any combination of halos may be used by adding the requisite flags, e.g: flags=XUPDATE+SUPDATE or flags=EUPDATE+WUPDATE+SUPDATE will update the east, west and south halos.

If a call to mpp_update_domains involves at least one E-W halo and one N-S halo, the corners involved will also be updated, i.e, in the example above, the SE and SW corners will be updated.
If flags is not supplied, that is equivalent to flags=XUPDATE+YUPDATE.

The vector version is passed the x and y components of a vector field in tandem, and both are updated upon return. They are passed together to treat parity issues on various grids. For example, on a cubic sphere projection, the x y components may be interchanged when passing from an equatorial cube face to a polar face. For grids with folds, vector components change sign on crossing the fold. Paired scalar quantities can also be passed with the vector version if flags=SCALAR_PAIR, in which case components are appropriately interchanged, but signs are not.

Special treatment at boundaries such as folds is also required for staggered grids. The following types of staggered grids are recognized:

1) AGRID: values are at grid centers.
2) BGRID_NE: vector fields are at the NE vertex of a grid cell, i.e: the array elements u(i,j)and v(i,j)are actually at (i,j;) with respect to the grid centers.
3) BGRID_SW: vector fields are at the SW vertex of a grid cell, i.e: the array elements u(i,j) and v(i,j) are actually at (i;,j;) with respect to the grid centers
4) CGRID_NE: vector fields are at the N and E faces of a grid cell, i.e: the array elements u(i,j) and v(i,j) are actually at (i;,j) and (i,j+½) with respect to the grid centers.
5) CGRID_SW: vector fields are at the S and W faces of a grid cell, i.e: the array elements u(i,j)and v(i,j) are actually at (i;,j) and (i,j;) with respect to the grid centers.

The gridtypes listed above are all available by use association as integer parameters. The scalar version of mpp_update_domains assumes that the values of a scalar field are always at AGRID locations, and no special boundary treatment is required. If vector fields are at staggered locations, the optional argument gridtype must be appropriately set for correct treatment at boundaries.
It is safe to apply vector field updates to the appropriate arrays irrespective of the domain topology: if the topology requires no special treatment of vector fields, specifying gridtype will do no harm.

mpp_update_domains internally buffers the date being sent and received into single messages for efficiency. A turnable internal buffer area in memory is provided for this purpose by mpp_domains_mod. The size of this buffer area can be set by the user by calling mpp_domains mpp_domains_set_stack_size.

Example usage: call mpp_update_domains( field, domain, flags ) Update a 1D domain for the given field. call mpp_update_domains( fieldx, fieldy, domain, flags, gridtype ) Update a 2D domain for the given fields.

Definition at line 1012 of file mpp_domains.F90.

Private Member Functions
	mpp_update_domain2d_i4_2d
	mpp_update_domain2d_i4_3d
	mpp_update_domain2d_i4_4d
	mpp_update_domain2d_i4_5d
	mpp_update_domain2d_i8_2d
	mpp_update_domain2d_i8_3d
	mpp_update_domain2d_i8_4d
	mpp_update_domain2d_i8_5d
	mpp_update_domain2d_r4_2d
	mpp_update_domain2d_r4_2dv
	mpp_update_domain2d_r4_3d
	mpp_update_domain2d_r4_3dv
	mpp_update_domain2d_r4_4d
	mpp_update_domain2d_r4_4dv
	mpp_update_domain2d_r4_5d
	mpp_update_domain2d_r4_5dv
	mpp_update_domain2d_r8_2d
	mpp_update_domain2d_r8_2dv
	mpp_update_domain2d_r8_3d
	mpp_update_domain2d_r8_3dv
	mpp_update_domain2d_r8_4d
	mpp_update_domain2d_r8_4dv
	mpp_update_domain2d_r8_5d
	mpp_update_domain2d_r8_5dv

mpp_domains_mod::mpp_update_domains_ad

interface mpp_domains_mod::mpp_update_domains_ad

Similar to mpp_update_domains , updates adjoint domains.

Definition at line 1518 of file mpp_domains.F90.

Private Member Functions
	mpp_update_domains_ad_2d_r4_2d
	mpp_update_domains_ad_2d_r4_2dv
	mpp_update_domains_ad_2d_r4_3d
	mpp_update_domains_ad_2d_r4_3dv
	mpp_update_domains_ad_2d_r4_4d
	mpp_update_domains_ad_2d_r4_4dv
	mpp_update_domains_ad_2d_r4_5d
	mpp_update_domains_ad_2d_r4_5dv
	mpp_update_domains_ad_2d_r8_2d
	mpp_update_domains_ad_2d_r8_2dv
	mpp_update_domains_ad_2d_r8_3d
	mpp_update_domains_ad_2d_r8_3dv
	mpp_update_domains_ad_2d_r8_4d
	mpp_update_domains_ad_2d_r8_4dv
	mpp_update_domains_ad_2d_r8_5d
	mpp_update_domains_ad_2d_r8_5dv

mpp_domains_mod::mpp_update_nest_coarse

interface mpp_domains_mod::mpp_update_nest_coarse

Pass the data from fine grid to fill the buffer to be ready to be interpolated onto coarse grid.
Example usage:

          call mpp_update_nest_coarse(field, nest_domain, field_out, nest_level, complete,
                            position, name, tile_count)

Definition at line 1437 of file mpp_domains.F90.

Private Member Functions
	mpp_update_nest_coarse_i4_2d
	mpp_update_nest_coarse_i4_3d
	mpp_update_nest_coarse_i4_4d
	mpp_update_nest_coarse_i8_2d
	mpp_update_nest_coarse_i8_3d
	mpp_update_nest_coarse_i8_4d
	mpp_update_nest_coarse_r4_2d
	mpp_update_nest_coarse_r4_2dv
	mpp_update_nest_coarse_r4_3d
	mpp_update_nest_coarse_r4_3dv
	mpp_update_nest_coarse_r4_4d
	mpp_update_nest_coarse_r4_4dv
	mpp_update_nest_coarse_r8_2d
	mpp_update_nest_coarse_r8_2dv
	mpp_update_nest_coarse_r8_3d
	mpp_update_nest_coarse_r8_3dv
	mpp_update_nest_coarse_r8_4d
	mpp_update_nest_coarse_r8_4dv

mpp_domains_mod::mpp_update_nest_fine

interface mpp_domains_mod::mpp_update_nest_fine

Pass the data from coarse grid to fill the buffer to be ready to be interpolated onto fine grid.
Example usage:

           call mpp_update_nest_fine(field, nest_domain, wbuffer, ebuffer, sbuffer,
                       nbuffer, nest_level, flags, complete, position, extra_halo, name,
                       tile_count)

Definition at line 1383 of file mpp_domains.F90.

Private Member Functions
	mpp_update_nest_fine_i4_2d
	mpp_update_nest_fine_i4_3d
	mpp_update_nest_fine_i4_4d
	mpp_update_nest_fine_i8_2d
	mpp_update_nest_fine_i8_3d
	mpp_update_nest_fine_i8_4d
	mpp_update_nest_fine_r4_2d
	mpp_update_nest_fine_r4_2dv
	mpp_update_nest_fine_r4_3d
	mpp_update_nest_fine_r4_3dv
	mpp_update_nest_fine_r4_4d
	mpp_update_nest_fine_r4_4dv
	mpp_update_nest_fine_r8_2d
	mpp_update_nest_fine_r8_2dv
	mpp_update_nest_fine_r8_3d
	mpp_update_nest_fine_r8_3dv
	mpp_update_nest_fine_r8_4d
	mpp_update_nest_fine_r8_4dv

mpp_domains_mod::nest_domain_type

type mpp_domains_mod::nest_domain_type

domain with nested fine and course tiles

Definition at line 455 of file mpp_domains.F90.

Collaboration diagram for nest_domain_type:

Private Attributes
integer, dimension(:), pointer	iend_coarse
integer, dimension(:), pointer	iend_fine
integer, dimension(:), pointer	istart_coarse
integer, dimension(:), pointer	istart_fine
integer, dimension(:), pointer	jend_coarse
integer, dimension(:), pointer	jend_fine
integer, dimension(:), pointer	jstart_coarse
integer, dimension(:), pointer	jstart_fine
character(len=name_length)	name
type(nest_level_type), dimension(:), pointer	nest => NULL()
integer, dimension(:), pointer	nest_level => NULL()
	Added for moving nest functionality.
integer	num_level
integer	num_nest
integer, dimension(:), pointer	tile_coarse
integer, dimension(:), pointer	tile_fine

mpp_domains_mod::nest_level_type

type mpp_domains_mod::nest_level_type

Private type to hold data for each level of nesting.

Definition at line 468 of file mpp_domains.F90.

Collaboration diagram for nest_level_type:

Private Attributes
type(nestspec), pointer	c2f_c => NULL()
type(nestspec), pointer	c2f_e => NULL()
type(nestspec), pointer	c2f_n => NULL()
type(nestspec), pointer	c2f_t => NULL()
type(domain2d), pointer	domain_coarse => NULL()
type(domain2d), pointer	domain_fine => NULL()
type(nestspec), pointer	f2c_c => NULL()
type(nestspec), pointer	f2c_e => NULL()
type(nestspec), pointer	f2c_n => NULL()
type(nestspec), pointer	f2c_t => NULL()
integer, dimension(:), pointer	iend_coarse
integer, dimension(:), pointer	iend_fine
logical	is_coarse
logical	is_coarse_pe
logical	is_fine
logical	is_fine_pe
integer, dimension(:), pointer	istart_coarse
integer, dimension(:), pointer	istart_fine
integer, dimension(:), pointer	jend_coarse
integer, dimension(:), pointer	jend_fine
integer, dimension(:), pointer	jstart_coarse
integer, dimension(:), pointer	jstart_fine
integer, dimension(:), pointer	my_nest_id
integer	my_num_nest
integer	num_nest
logical	on_level
integer, dimension(:), pointer	pelist => NULL()
integer, dimension(:), pointer	pelist_coarse => NULL()
integer, dimension(:), pointer	pelist_fine => NULL()
integer, dimension(:), pointer	tile_coarse
integer, dimension(:), pointer	tile_fine
integer	x_refine
integer	y_refine

mpp_domains_mod::nestspec

type mpp_domains_mod::nestspec

Used to specify bounds and index information for nested tiles as a linked list.

Definition at line 438 of file mpp_domains.F90.

Collaboration diagram for nestspec:

Private Attributes
type(index_type)	center
type(index_type)	east
integer	extra_halo
type(nestspec), pointer	next => NULL()
type(index_type)	north
integer	nrecv
integer	nsend
type(overlap_type), dimension(:), pointer	recv => NULL()
type(overlap_type), dimension(:), pointer	send => NULL()
type(index_type)	south
type(index_type)	west
integer	xbegin
integer	xbegin_c
integer	xbegin_f
integer	xend
integer	xend_c
integer	xend_f
integer	xsize_c
integer	ybegin
integer	ybegin_c
integer	ybegin_f
integer	yend
integer	yend_c
integer	yend_f
integer	ysize_c

mpp_domains_mod::nonblock_type

type mpp_domains_mod::nonblock_type

Used for nonblocking data transfer.

Definition at line 548 of file mpp_domains.F90.

Collaboration diagram for nonblock_type:

Private Attributes
integer, dimension(max_request)	buffer_pos_recv
integer, dimension(max_request)	buffer_pos_send
integer(i8_kind), dimension(max_domain_fields)	field_addrs
integer(i8_kind), dimension(max_domain_fields)	field_addrs2
integer	nfields
integer	recv_msgsize
integer	recv_pos
integer, dimension(max_request)	request_recv
integer	request_recv_count
integer, dimension(max_request)	request_send
integer	request_send_count
integer	send_msgsize
integer	send_pos
integer, dimension(max_request)	size_recv
integer, dimension(max_request)	type_recv
integer	update_ehalo
integer	update_flags
integer	update_gridtype
integer	update_nhalo
integer	update_position
integer	update_shalo
integer	update_whalo

mpp_domains_mod::operator(.eq.)

interface mpp_domains_mod::operator(.eq.)

Equality/inequality operators for domaintypes.

The module provides public operators to check for equality/inequality of domaintypes, e.g:

       type(domain1D) :: a, b
       type(domain2D) :: c, d
       ...
       if( a.NE.b )then
       ...
       end if
       if( c==d )then
       ...
       end if

Domains are considered equal if and only if the start and end indices of each of their component global, data and compute domains are equal.

Definition at line 2170 of file mpp_domains.F90.

Private Member Functions
	mpp_domain1d_eq
	mpp_domain2d_eq
	mpp_domainug_eq

mpp_domains_mod::operator(.ne.)

interface mpp_domains_mod::operator(.ne.)

Definition at line 2177 of file mpp_domains.F90.

Private Member Functions
	mpp_domain1d_ne
	mpp_domain2d_ne
	mpp_domainug_ne

mpp_domains_mod::overlap_type

type mpp_domains_mod::overlap_type

Type for overlapping data.

Definition at line 319 of file mpp_domains.F90.

Collaboration diagram for overlap_type:

Private Attributes
integer	count = 0
	number of overlapping
integer, dimension(:), pointer	dir => NULL()
	direction ( value 1,2,3,4 = E,S,W,N)
logical, dimension(:), pointer	from_contact => NULL()
	indicate if the overlap is computed from define_contact_overlap
integer, dimension(:), pointer	ie => NULL()
	ending i-index
integer, dimension(:), pointer	index => NULL()
	for refinement
integer, dimension(:), pointer	is => NULL()
	starting i-index
integer, dimension(:), pointer	je => NULL()
	ending j-index
integer, dimension(:), pointer	js => NULL()
	starting j-index
integer, dimension(:), pointer	msgsize => NULL()
	overlapping msgsize to be sent or received
integer	pe
integer, dimension(:), pointer	rotation => NULL()
	rotation angle.
integer	start_pos
	start position in the buffer
integer, dimension(:), pointer	tileme => NULL()
	my tile id for this overlap
integer, dimension(:), pointer	tilenbr => NULL()
	neighbor tile id for this overlap
integer	totsize
	all message size

mpp_domains_mod::overlapspec

type mpp_domains_mod::overlapspec

Private type for overlap specifications.

Definition at line 341 of file mpp_domains.F90.

Collaboration diagram for overlapspec:

Private Attributes
integer	ehalo
type(overlapspec), pointer	next => NULL()
integer	nhalo
	halo size
integer	nrecv
integer	nsend
type(overlap_type), dimension(:), pointer	recv => NULL()
integer	recvsize
type(overlap_type), dimension(:), pointer	send => NULL()
integer	sendsize
integer	shalo
integer	whalo
integer	xbegin
integer	xend
integer	ybegin
integer	yend

mpp_domains_mod::tile_type

type mpp_domains_mod::tile_type

Upper and lower x and y bounds for a tile.

Definition at line 354 of file mpp_domains.F90.

Collaboration diagram for tile_type:

Private Attributes
integer	xbegin
integer	xend
integer	ybegin
integer	yend

mpp_domains_mod::unstruct_axis_spec

type mpp_domains_mod::unstruct_axis_spec

Private type for axis specification data for an unstructured grid.

Definition at line 229 of file mpp_domains.F90.

Collaboration diagram for unstruct_axis_spec:

Private Attributes
integer	begin
integer	begin_index
integer	end
integer	end_index
integer	max_size
integer	size

mpp_domains_mod::unstruct_domain_spec

type mpp_domains_mod::unstruct_domain_spec

Private type for axis specification data for an unstructured domain.

Definition at line 237 of file mpp_domains.F90.

Collaboration diagram for unstruct_domain_spec:

Private Attributes
type(unstruct_axis_spec)	compute
integer	pe
integer	pos
integer	tile_id

mpp_domains_mod::unstruct_overlap_type

type mpp_domains_mod::unstruct_overlap_type

Private type.

Definition at line 247 of file mpp_domains.F90.

Collaboration diagram for unstruct_overlap_type:

Private Attributes
integer	count = 0
integer, dimension(:), pointer	i =>NULL()
integer, dimension(:), pointer	j =>NULL()
integer	pe

mpp_domains_mod::unstruct_pass_type

type mpp_domains_mod::unstruct_pass_type

Private type.

Definition at line 257 of file mpp_domains.F90.

Collaboration diagram for unstruct_pass_type:

Private Attributes
integer	nrecv
integer	nsend
type(unstruct_overlap_type), dimension(:), pointer	recv =>NULL()
type(unstruct_overlap_type), dimension(:), pointer	send =>NULL()

Function/Subroutine Documentation

compute_overlaps()

subroutine compute_overlaps	(	type(domain2d), intent(inout)	domain,
		integer, intent(in)	position,
		type(overlapspec), intent(inout), pointer	update,
		type(overlapspec), intent(inout), pointer	check,
		integer, intent(in)	ishift,
		integer, intent(in)	jshift,
		integer, intent(in)	x_cyclic_offset,
		integer, intent(in)	y_cyclic_offset,
		integer, intent(in)	whalo,
		integer, intent(in)	ehalo,
		integer, intent(in)	shalo,
		integer, intent(in)	nhalo
	)

Computes remote domain overlaps.

Assumes only one in each direction will calculate the overlapping for T,E,C,N-cell seperately.

Definition at line 1594 of file mpp_domains_define.inc.

define_contact_point()

subroutine define_contact_point	(	type(domain2d), intent(inout)	domain,
		integer, intent(in)	position,
		integer, intent(in)	num_contact,
		integer, dimension(:), intent(in)	tile1,
		integer, dimension(:), intent(in)	tile2,
		integer, dimension(:), intent(in)	align1,
		integer, dimension(:), intent(in)	align2,
		real, dimension(:), intent(in)	refine1,
		real, dimension(:), intent(in)	refine2,
		integer, dimension(:), intent(in)	istart1,
		integer, dimension(:), intent(in)	iend1,
		integer, dimension(:), intent(in)	jstart1,
		integer, dimension(:), intent(in)	jend1,
		integer, dimension(:), intent(in)	istart2,
		integer, dimension(:), intent(in)	iend2,
		integer, dimension(:), intent(in)	jstart2,
		integer, dimension(:), intent(in)	jend2,
		integer, dimension(:), intent(in)	isgList,
		integer, dimension(:), intent(in)	iegList,
		integer, dimension(:), intent(in)	jsgList,
		integer, dimension(:), intent(in)	jegList
	)

compute the overlapping between tiles for the T-cell.

Parameters

[in]	num_contact	number of contact regions
[in]	tile2	tile number
[in]	align2	align direction of contact region
[in]	refine2	refinement between tiles
[in]	iend1	i-index in tile_1 of contact region
[in]	jend1	j-index in tile_1 of contact region
[in]	iend2	i-index in tile_2 of contact region
[in]	jend2	j-index in tile_2 of contact region
[in]	ieglist	i-global domain of each tile
[in]	jeglist	j-global domain of each tile

Definition at line 5291 of file mpp_domains_define.inc.

define_nest_level_type()

subroutine define_nest_level_type	(	type(nest_level_type), intent(inout)	nest_domain,
		integer, intent(in)	x_refine,
		integer, intent(in)	y_refine,
		integer, intent(in)	extra_halo
	)

Parameters

[in,out]	nest_domain	nest domain to be defined
[in]	extra_halo	halo value
[in]	y_refine	x and y refinements

Definition at line 465 of file mpp_define_nest_domains.inc.

domain_update_is_needed()

logical function domain_update_is_needed	(	type(domain2d), intent(in)	domain,
		integer, intent(in)	whalo,
		integer, intent(in)	ehalo,
		integer, intent(in)	shalo,
		integer, intent(in)	nhalo
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1002 of file mpp_domains_util.inc.

get_mesgsize()

integer function get_mesgsize	(	type(overlap_type), intent(in)	overlap,
		logical, dimension(:), intent(in)	do_dir
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1715 of file mpp_domains_util.inc.

get_rank_recv()

integer function get_rank_recv	(	type(domain2d), intent(in)	domain,
		type(overlapspec), intent(in)	overlap_x,
		type(overlapspec), intent(in)	overlap_y,
		integer, intent(out)	rank_x,
		integer, intent(out)	rank_y,
		integer, intent(out)	ind_x,
		integer, intent(out)	ind_y
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1523 of file mpp_domains_util.inc.

get_rank_send()

integer function get_rank_send	(	type(domain2d), intent(in)	domain,
		type(overlapspec), intent(in)	overlap_x,
		type(overlapspec), intent(in)	overlap_y,
		integer, intent(out)	rank_x,
		integer, intent(out)	rank_y,
		integer, intent(out)	ind_x,
		integer, intent(out)	ind_y
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1496 of file mpp_domains_util.inc.

get_rank_unpack()

integer function get_rank_unpack	(	type(domain2d), intent(in)	domain,
		type(overlapspec), intent(in)	overlap_x,
		type(overlapspec), intent(in)	overlap_y,
		integer, intent(out)	rank_x,
		integer, intent(out)	rank_y,
		integer, intent(out)	ind_x,
		integer, intent(out)	ind_y
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1687 of file mpp_domains_util.inc.

get_vector_recv()

integer function get_vector_recv	(	type(domain2d), intent(in)	domain,
		type(overlapspec), intent(in)	update_x,
		type(overlapspec), intent(in)	update_y,
		integer, dimension(:), intent(out)	ind_x,
		integer, dimension(:), intent(out)	ind_y,
		integer, dimension(:), intent(out)	start_pos,
		integer, dimension(:), intent(out)	pelist
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1553 of file mpp_domains_util.inc.

get_vector_send()

integer function get_vector_send	(	type(domain2d), intent(in)	domain,
		type(overlapspec), intent(in)	update_x,
		type(overlapspec), intent(in)	update_y,
		integer, dimension(:), intent(out)	ind_x,
		integer, dimension(:), intent(out)	ind_y,
		integer, dimension(:), intent(out)	start_pos,
		integer, dimension(:), intent(out)	pelist
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1619 of file mpp_domains_util.inc.

init_nonblock_type()

subroutine init_nonblock_type

(

type(nonblock_type), intent(inout)

nonblock_obj

)

Initialize domain decomp package.

Called to initialize the mpp_domains_mod package. flags can be set to MPP_VERBOSE to have mpp_domains_mod keep you informed of what it's up to. MPP_DEBUG returns even more information for debugging.

mpp_domains_init will call mpp_init, to make sure mpp_mod is initialized. (Repeated calls to mpp_init do no harm, so don't worry if you already called it).

Definition at line 125 of file mpp_domains_misc.inc.

mpp_check_field_2d()

subroutine mpp_check_field_2d	(	real, dimension(:,:), intent(in)	field_in,
		integer, dimension(:), intent(in)	pelist1,
		integer, dimension(:), intent(in)	pelist2,
		type(domain2d), intent(in)	domain,
		character(len=*), intent(in)	mesg,
		integer, intent(in), optional	w_halo,
		integer, intent(in), optional	s_halo,
		integer, intent(in), optional	e_halo,
		integer, intent(in), optional	n_halo,
		logical, intent(in), optional	force_abort,
		integer, intent(in), optional	position
	)

This routine is used to do parallel checking for 2d data between n and m pe. The comparison is is done on pelist2. When size of pelist2 is 1, we can check the halo; otherwise, halo can not be checked.

Parameters

[in]	field_in	field to be checked
[in]	pelist2	pe list for the two groups
[in]	domain	domain for each pe
[in]	mesg	message to be printed out if differences found
[in]	n_halo	halo size for west, south, east and north
[in]	force_abort	when true, call mpp_error if any difference found. default value is false.
[in]	position	when domain is symmetry, only value = CENTER is implemented.

Definition at line 218 of file mpp_domains_misc.inc.

mpp_check_field_2d_type1()

subroutine mpp_check_field_2d_type1	(	real, dimension(:,:), intent(in)	field_in,
		integer, dimension(:), intent(in)	pelist1,
		integer, dimension(:), intent(in)	pelist2,
		type(domain2d), intent(in)	domain,
		character(len=*), intent(in)	mesg,
		integer, intent(in), optional	w_halo,
		integer, intent(in), optional	s_halo,
		integer, intent(in), optional	e_halo,
		integer, intent(in), optional	n_halo,
		logical, intent(in), optional	force_abort
	)

This routine is used to check field between running on 1 pe (pelist2) and n pe(pelist1). The need_to_be_checked data is sent to the pelist2 and All the comparison is done on pelist2.

Parameters

[in]	field_in	field to be checked
[in]	pelist2	pe list for the two groups
[in]	domain	domain for each pe
[in]	mesg	message to be printed out if differences found
[in]	n_halo	halo size for west, south, east and north
[in]	force_abort	when, call mpp_error if any difference found. default value is false.

Definition at line 257 of file mpp_domains_misc.inc.

mpp_check_field_2d_type2()

subroutine mpp_check_field_2d_type2	(	real, dimension(:,:), intent(in)	field_in,
		integer, dimension(:), intent(in)	pelist1,
		integer, dimension(:), intent(in)	pelist2,
		type(domain2d), intent(in)	domain,
		character(len=*), intent(in)	mesg,
		logical, intent(in), optional	force_abort
	)

This routine is used to check field between running on m pe (root pe) and n pe. This routine can not check halo.

Parameters

[in]

force_abort

when, call mpp_error if any difference found. default value is false.

Definition at line 388 of file mpp_domains_misc.inc.

mpp_check_field_3d()

subroutine mpp_check_field_3d	(	real, dimension(:,:,:), intent(in)	field_in,
		integer, dimension(:), intent(in)	pelist1,
		integer, dimension(:), intent(in)	pelist2,
		type(domain2d), intent(in)	domain,
		character(len=*), intent(in)	mesg,
		integer, intent(in), optional	w_halo,
		integer, intent(in), optional	s_halo,
		integer, intent(in), optional	e_halo,
		integer, intent(in), optional	n_halo,
		logical, intent(in), optional	force_abort,
		integer, intent(in), optional	position
	)

This routine is used to do parallel checking for 3d data between n and m pe. The comparison is is done on pelist2. When size of pelist2 is 1, we can check the halo; otherwise, halo can not be checked.

Parameters

[in]	field_in	field to be checked
[in]	pelist2	pe list for the two groups
[in]	domain	domain for each pe
[in]	mesg	message to be printed out if differences found
[in]	n_halo	halo size for west, south, east and north
[in]	force_abort	when true, call mpp_error if any difference found. default value is false.
[in]	position	when domain is symmetry, only value = CENTER is implemented.

Definition at line 185 of file mpp_domains_misc.inc.

mpp_clear_group_update()

subroutine mpp_clear_group_update

(

type(mpp_group_update_type), intent(inout)

group

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 2374 of file mpp_domains_util.inc.

mpp_compute_block_extent()

subroutine mpp_compute_block_extent	(	integer, intent(in)	isg,
		integer, intent(in)	ieg,
		integer, intent(in)	ndivs,
		integer, dimension(:), intent(out)	ibegin,
		integer, dimension(:), intent(out)	iend
	)

Computes the extents of a grid block.

Tis implementation is different from mpp_compute_extents The last block might have most points

Definition at line 161 of file mpp_domains_define.inc.

mpp_copy_domain1d()

recursive subroutine mpp_copy_domain1d	(	type(domain1d), intent(in)	domain_in,
		type(domain1d), intent(inout)	domain_out
	)

Copies input 1d domain to the output 1d domain.

Parameters

[in]	domain_in	Input domain
[in,out]	domain_out	Output domain

Definition at line 1741 of file mpp_domains_util.inc.

mpp_copy_domain1d_spec()

subroutine mpp_copy_domain1d_spec	(	type(domain1d_spec), intent(in)	domain1D_spec_in,
		type(domain1d_spec), intent(out)	domain1D_spec_out
	)

Copies input 1d domain spec to the output 1d domain spec.

Parameters

[in]	domain1d_spec_in	Input
[out]	domain1d_spec_out	Output

Definition at line 1896 of file mpp_domains_util.inc.

mpp_copy_domain2d()

subroutine mpp_copy_domain2d	(	type(domain2d), intent(in)	domain_in,
		type(domain2d), intent(inout)	domain_out
	)

Copies input 2d domain to the output 2d domain.

Parameters

[in]	domain_in	Input domain
[in,out]	domain_out	Output domain

Definition at line 1773 of file mpp_domains_util.inc.

mpp_copy_domain2d_spec()

subroutine mpp_copy_domain2d_spec	(	type(domain2d_spec), intent(in)	domain2D_spec_in,
		type(domain2d_spec), intent(out)	domain2d_spec_out
	)

Copies input 2d domain spec to the output 2d domain spec.

Parameters

[in]	domain2d_spec_in	Input
[out]	domain2d_spec_out	Output

Definition at line 1850 of file mpp_domains_util.inc.

mpp_copy_domain_axis_spec()

subroutine mpp_copy_domain_axis_spec	(	type(domain_axis_spec), intent(in)	domain_axis_spec_in,
		type(domain_axis_spec), intent(out)	domain_axis_spec_out
	)

Copies input domain_axis_spec to the output domain_axis_spec.

Parameters

[in]	domain_axis_spec_in	Input
[out]	domain_axis_spec_out	Output

Definition at line 1907 of file mpp_domains_util.inc.

mpp_create_super_grid_domain()

subroutine mpp_create_super_grid_domain

(

type(domain2d), intent(inout)

domain

)

Modifies the indices of the input domain to create the supergrid domain.

This is an example of how to use mpp_create_super_grid_domain

call mpp_copy_domain(domain_in, domain_out)
call super_grid_domain(domain_out)

domain_in is the original domain, domain_out is the domain with the supergrid indices.

Parameters

[in,out]

domain

Input domain

Definition at line 293 of file mpp_domains_util.inc.

mpp_define_domains1d()

subroutine mpp_define_domains1d	(	integer, dimension(:), intent(in)	global_indices,
		integer, intent(in)	ndivs,
		type(domain1d), intent(inout)	domain,
		integer, dimension(0:), intent(in), optional	pelist,
		integer, intent(in), optional	flags,
		integer, intent(in), optional	halo,
		integer, dimension(0:), intent(in), optional	extent,
		logical, dimension(0:), intent(in), optional	maskmap,
		integer, intent(in), optional	memory_size,
		integer, intent(in), optional	begin_halo,
		integer, intent(in), optional	end_halo
	)

Define data and computational domains on a 1D set of data (isg:ieg) and assign them to PEs.

Parameters

[in]	global_indices	(/ isg, ieg /) gives the extent of global domain
[in]	ndivs	number of divisions of domain: even divisions unless extent is present.
[in,out]	domain	the returned domain1D; declared inout so that existing links, if any, can be nullified
[in]	pelist	list of PEs to which domains are to be assigned (default 0...npes-1); size of pelist must correspond to number of mask=.TRUE. divisions
[in]	halo	flags define whether compute and data domains are global (undecomposed) and whether the global domain has periodic boundaries. halo defines halo width (currently the same on both sides)
[in]	extent	array extent; defines width of each division (used for non-uniform domain decomp, for e.g load-balancing)
[in]	maskmap	a division whose maskmap=.FALSE. is not assigned to any domain. By default we assume decomposition of compute and data domains, non-periodic boundaries, no halo, as close to uniform extents as the input parameters permit

Definition at line 281 of file mpp_domains_define.inc.

mpp_define_domains2d()

subroutine mpp_define_domains2d	(	integer, dimension(:), intent(in)	global_indices,
		integer, dimension(:), intent(in)	layout,
		type(domain2d), intent(inout)	domain,
		integer, dimension(0:), intent(in), optional	pelist,
		integer, intent(in), optional	xflags,
		integer, intent(in), optional	yflags,
		integer, intent(in), optional	xhalo,
		integer, intent(in), optional	yhalo,
		integer, dimension(0:), intent(in), optional	xextent,
		integer, dimension(0:), intent(in), optional	yextent,
		logical, dimension(0:,0:), intent(in), optional	maskmap,
		character(len=*), intent(in), optional	name,
		logical, intent(in), optional	symmetry,
		integer, dimension(:), intent(in), optional	memory_size,
		integer, intent(in), optional	whalo,
		integer, intent(in), optional	ehalo,
		integer, intent(in), optional	shalo,
		integer, intent(in), optional	nhalo,
		logical, intent(in), optional	is_mosaic,
		integer, intent(in), optional	tile_count,
		integer, intent(in), optional	tile_id,
		logical, intent(in), optional	complete,
		integer, intent(in), optional	x_cyclic_offset,
		integer, intent(in), optional	y_cyclic_offset
	)

Define 2D data and computational domain on global rectilinear cartesian domain (isg:ieg,jsg:jeg) and assign them to PEs.

Parameters

[in]	global_indices	(/ isg, ieg, jsg, jeg /)
[in]	layout	pe layout
[in,out]	domain	2D domain decomposition to define
[in]	pelist	current pelist to run on
[in]	yflags	directional flag
[in]	yhalo	halo sizes for x and y indices
[in]	is_mosaic	indicate if calling mpp_define_domains from mpp_define_mosaic.
[in]	nhalo	halo size for West, East, South and North direction. if whalo and ehalo is not present, will take the value of xhalo if shalo and nhalo is not present, will take the value of yhalo
[in]	tile_count	tile number on current pe, default value is 1. this is for the situation that multiple tiles on one processor
[in]	tile_id	tile id
[in]	complete	true indicate mpp_define_domain is completed for mosaic definition.
[in]	x_cyclic_offset	offset for x-cyclic boundary condition, (0,j) = (ni, mod(j+x_cyclic_offset,nj)) (ni+1, j)=(1 ,mod(j+nj-x_cyclic_offset,nj))
[in]	y_cyclic_offset	offset for y-cyclic boundary condition (i,0) = (mod(i+y_cyclic_offset,ni), nj)) (i,nj+1) =(mod(mod(i+ni-y_cyclic_offset,ni), 1) )

Definition at line 608 of file mpp_domains_define.inc.

mpp_define_io_domain()

subroutine mpp_define_io_domain	(	type(domain2d), intent(inout)	domain,
		integer, dimension(2), intent(in)	io_layout
	)

Define the layout for IO pe's for the given domain.

Parameters

[in,out]	domain	Input 2D domain
[in]	io_layout	2 value io pe layout to define

Definition at line 457 of file mpp_domains_define.inc.

mpp_define_layout2d()

subroutine mpp_define_layout2d	(	integer, dimension(:), intent(in)	global_indices,
		integer, intent(in)	ndivs,
		integer, dimension(:), intent(out)	layout
	)

Parameters

[in]	global_indices	(/ isg, ieg, jsg, jeg /); Defines the global domain.
[in]	ndivs	number of divisions to divide global domain

Definition at line 27 of file mpp_domains_define.inc.

mpp_define_mosaic()

subroutine mpp_define_mosaic	(	integer, dimension(:,:), intent(in)	global_indices,
		integer, dimension(:,:), intent(in)	layout,
		type(domain2d), intent(inout)	domain,
		integer, intent(in)	num_tile,
		integer, intent(in)	num_contact,
		integer, dimension(:), intent(in)	tile1,
		integer, dimension(:), intent(in)	tile2,
		integer, dimension(:), intent(in)	istart1,
		integer, dimension(:), intent(in)	iend1,
		integer, dimension(:), intent(in)	jstart1,
		integer, dimension(:), intent(in)	jend1,
		integer, dimension(:), intent(in)	istart2,
		integer, dimension(:), intent(in)	iend2,
		integer, dimension(:), intent(in)	jstart2,
		integer, dimension(:), intent(in)	jend2,
		integer, dimension(:), intent(in)	pe_start,
		integer, dimension(:), intent(in)	pe_end,
		integer, dimension(:), intent(in), optional	pelist,
		integer, intent(in), optional	whalo,
		integer, intent(in), optional	ehalo,
		integer, intent(in), optional	shalo,
		integer, intent(in), optional	nhalo,
		integer, dimension(:,:), intent(in), optional	xextent,
		integer, dimension(:,:), intent(in), optional	yextent,
		logical, dimension(:,:,:), intent(in), optional	maskmap,
		character(len=*), intent(in), optional	name,
		integer, dimension(2), intent(in), optional	memory_size,
		logical, intent(in), optional	symmetry,
		integer, intent(in), optional	xflags,
		integer, intent(in), optional	yflags,
		integer, dimension(:), intent(in), optional	tile_id
	)

Defines a domain for mosaic tile grids.

Parameters

[in]	num_tile	number of tiles in the mosaic
[in]	num_contact	number of contact region between tiles.
[in]	tile2	tile number
[in]	iend1	i-index in tile_1 of contact region
[in]	jend1	j-index in tile_1 of contact region
[in]	iend2	i-index in tile_2 of contact region
[in]	jend2	j-index in tile_2 of contact region
[in]	pe_start	start pe of the pelist used in each tile
[in]	pe_end	end pe of the pelist used in each tile
[in]	pelist	list of processors used in mosaic
[in]	tile_id	tile_id of each tile in the mosaic

Definition at line 1201 of file mpp_domains_define.inc.

mpp_define_mosaic_pelist()

subroutine mpp_define_mosaic_pelist	(	integer, dimension(:), intent(in)	sizes,
		integer, dimension(:), intent(inout)	pe_start,
		integer, dimension(:), intent(inout)	pe_end,
		integer, dimension(:), intent(in), optional	pelist,
		integer, dimension(:), intent(in), optional	costpertile
	)

Defines a pelist for use with mosaic tiles.

Note

The following routine may need to revised to improve the capability. It is very hard to make it balance for all the situation. Hopefully some smart idea will come up someday.

Definition at line 62 of file mpp_domains_define.inc.

mpp_define_nest_domains()

subroutine mpp_define_nest_domains	(	type(nest_domain_type), intent(inout)	nest_domain,
		type(domain2d), intent(in), target	domain,
		integer, intent(in)	num_nest,
		integer, dimension(:), intent(in)	nest_level,
		integer, dimension(:), intent(in)	tile_fine,
		integer, dimension(:), intent(in)	tile_coarse,
		integer, dimension(:), intent(in)	istart_coarse,
		integer, dimension(:), intent(in)	icount_coarse,
		integer, dimension(:), intent(in)	jstart_coarse,
		integer, dimension(:), intent(in)	jcount_coarse,
		integer, dimension(:), intent(in)	npes_nest_tile,
		integer, dimension(:), intent(in)	x_refine,
		integer, dimension(:), intent(in)	y_refine,
		integer, intent(in), optional	extra_halo,
		character(len=*), intent(in), optional	name
	)

Set up a domain to pass data between aligned coarse and fine grid of nested model.

Set up a domain to pass data between aligned coarse and fine grid of a nested model. Supports multiple and telescoping nests. A telescoping nest is defined as a nest within a nest. Nest domains may span multiple tiles, but cannot contain a coarse-grid, cube corner. Concurrent nesting is the only supported mechanism, i.e. coarse and fine grid are on individual, non-overlapping, processor lists. Coarse and fine grid domain need to be defined before calling mpp_define_nest_domains. An mpp_broadcast is needed to broadcast both fine and coarse grid domain onto all processors.

mpp_update_nest_coarse is used to pass data from fine grid to coarse grid computing domain. mpp_update_nest_fine is used to pass data from coarse grid to fine grid halo. You may call mpp_get_C2F_index before calling mpp_update_nest_fine to get the index for passing data from coarse to fine. You may call mpp_get_F2C_index before calling mpp_update_nest_coarse to get the index for passing data from coarse to fine.

Note

The following tests for nesting of regular lat-lon grids upon a cubed-sphere grid are done in test_mpp_domains:
a) a first-level nest spanning multiple cubed-sphere faces (tiles 1, 2, & 4)
b) a first-level nest wholly contained within tile 3
c) a second-level nest contained within the nest mentioned in a)
Tests are done for data at T, E, C, N-cell center.

Below is an example to pass data between fine and coarse grid (More details on how to use the nesting domain update are available in routine test_update_nest_domain of test_fms/mpp/test_mpp_domains.F90.

 if( concurrent ) then
   call mpp_broadcast_domain(domain_fine)
   call mpp_broadcast_domain(domain_coarse)
 endif
 
 call mpp_define_nest_domains(nest_domain,domain,num_nest,nest_level(1:num_nest), &
                             tile_fine(1:num_nest), tile_coarse(1:num_nest), &
                              istart_coarse(1:num_nest), icount_coarse(1:num_nest), &
                              jstart_coarse(1:num_nest), jcount_coarse(1:num_nest), &
                              npes_nest_tile, x_refine(1:num_nest), y_refine(1:num_nest), &
                              extra_halo=extra_halo, name="nest_domain")
 
 call mpp_get_c2f_index(nest_domain, isw_f, iew_f, jsw_f, jew_f, isw_c, iew_c, jsw_c, jew_c, west, level)
 call mpp_get_c2f_index(nest_domain, ise_f, iee_f, jse_f, jee_f, ise_c, iee_c, jse_c, jee_c, east, level)
 call mpp_get_c2f_index(nest_domain, iss_f, ies_f, jss_f, jes_f, iss_c, ies_c, jss_c, jes_c, south, level)
 call mpp_get_c2f_index(nest_domain, isn_f, ien_f, jsn_f, jen_f, isn_c, ien_c, jsn_c, jen_c, north, level)
 
 allocate(wbuffer(isw_c:iew_c, jsw_c:jew_c,nz))
 allocate(ebuffer(ise_c:iee_c, jse_c:jee_c,nz))
 allocate(sbuffer(iss_c:ies_c, jss_c:jes_c,nz))
 allocate(nbuffer(isn_c:ien_c, jsn_c:jen_c,nz))
 call mpp_update_nest_fine(x, nest_domain, wbuffer, sbuffer, ebuffer, nbuffer)
 
 call mpp_get_f2c_index(nest_domain, is_c, ie_c, js_c, je_c, is_f, ie_f, js_f, je_f, nest_level=level)
 allocate(buffer(is_f:ie_f, js_f:je_f,nz))
 call mpp_update_nest_coarse(x, nest_domain, buffer)
@note currently the contact will be limited to overlap contact.
@param [in,out] nest_domain holds the information to pass data
 between nest and parent grids.
@param [in] domain domain for the grid defined in the current pelist
@param [in] num_nest number of nests
@param [in] nest_level array containing the nest level for each nest
(>1 implies a telescoping nest)
@param [in] tile_coarse array containing tile number of the
 nest grid(monotonically increasing starting with 7),
 array containing tile number of the parent grid corresponding
 to the lower left corner of a given nest
@param [in] jcount_coarse start
 array containing index in the parent grid of the lower left corner of a given
 nest, count: array containing span of the nest on the parent grid
@param [in] npes_nest_tile array containing number of pes to allocated
 to each defined tile
@param [in] y_refine array containing refinement ratio
 for each nest
@param [in] extra_halo extra halo for passing data from coarse grid to fine grid.
 default is 0 and currently only support extra_halo = 0.
@param [in] name name of the nest domain

Definition at line 95 of file mpp_define_nest_domains.inc.

mpp_define_unstruct_domain()

subroutine mpp_define_unstruct_domain	(	type(domainug), intent(inout)	UG_domain,
		type(domain2d), intent(in), target	SG_domain,
		integer, dimension(:), intent(in)	npts_tile,
		integer, dimension(:), intent(in)	grid_nlev,
		integer, intent(in)	ndivs,
		integer, intent(in)	npes_io_group,
		integer, dimension(:), intent(in)	grid_index,
		character(len=*), intent(in), optional	name
	)

Parameters

[in]	npts_tile	number of unstructured points on each tile
[in]	grid_nlev	number of levels in each unstructured grid.
[in]	npes_io_group	number of processors in a io group. Only pe with same tile_id in the same group

Definition at line 25 of file mpp_unstruct_domain.inc.

mpp_domain1d_eq()

logical function mpp_domain1d_eq	(	type(domain1d), intent(in)	a,
		type(domain1d), intent(in)	b
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 60 of file mpp_domains_util.inc.

mpp_domain1d_ne()

logical function mpp_domain1d_ne	(	type(domain1d), intent(in)	a,
		type(domain1d), intent(in)	b
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 78 of file mpp_domains_util.inc.

mpp_domain2d_eq()

logical function mpp_domain2d_eq	(	type(domain2d), intent(in)	a,
		type(domain2d), intent(in)	b
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 86 of file mpp_domains_util.inc.

mpp_domain2d_ne()

logical function mpp_domain2d_ne	(	type(domain2d), intent(in)	a,
		type(domain2d), intent(in)	b
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 110 of file mpp_domains_util.inc.

mpp_domain_is_initialized()

logical function mpp_domain_is_initialized

(

type(domain2d), intent(in)

domain

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 988 of file mpp_domains_util.inc.

mpp_domain_is_symmetry()

logical function mpp_domain_is_symmetry

(

type(domain2d), intent(in)

domain

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 978 of file mpp_domains_util.inc.

mpp_domains_init()

subroutine mpp_domains_init

(

integer, intent(in), optional

flags

)

Initialize domain decomp package.

Called to initialize the mpp_domains_mod package. flags can be set to MPP_VERBOSE to have mpp_domains_mod keep you informed of what it's up to. MPP_DEBUG returns even more information for debugging.

mpp_domains_init will call mpp_init, to make sure mpp_mod is initialized. (Repeated calls to mpp_init do no harm, so don't worry if you already called it).

Definition at line 44 of file mpp_domains_misc.inc.

mpp_domains_set_stack_size()

subroutine mpp_domains_set_stack_size

(

integer, intent(in)

n

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 35 of file mpp_domains_util.inc.

mpp_get_c2f_index()

subroutine mpp_get_c2f_index	(	type(nest_domain_type), intent(in)	nest_domain,
		integer, intent(out)	is_fine,
		integer, intent(out)	ie_fine,
		integer, intent(out)	js_fine,
		integer, intent(out)	je_fine,
		integer, intent(out)	is_coarse,
		integer, intent(out)	ie_coarse,
		integer, intent(out)	js_coarse,
		integer, intent(out)	je_coarse,
		integer, intent(in)	dir,
		integer, intent(in)	nest_level,
		integer, intent(in), optional	position
	)

Get the index of the data passed from coarse grid to fine grid.

Get the index of the data passed from coarse grid to fine grid.

Example usage:

call mpp_get_c2f_index(nest_domain, is_fine, ie_fine, js_fine, je_fine,
                           is_coarse, ie_coarse, js_coarse, je_coarse, dir,
                           nest_level, position)

Parameters

[in]	nest_domain	holds the information to pass data between fine and coarse grids
[out]	je_fine	index in the fine grid of the nested region
[out]	je_coarse	index in the coarse grid of the nested region
[in]	nest_level	direction of the halo update. Its value should be WEST, EAST, SOUTH or NORTH.; level of the nest (> 1 implies a telescoping nest)
[in]	position	Cell position. It value should be CENTER, EAST, CORNER, or NORTH.

Definition at line 1638 of file mpp_define_nest_domains.inc.

mpp_get_compute_domain1d()

subroutine mpp_get_compute_domain1d	(	type(domain1d), intent(in)	domain,
		integer, intent(out), optional	begin,
		integer, intent(out), optional	end,
		integer, intent(out), optional	size,
		integer, intent(out), optional	max_size,
		logical, intent(out), optional	is_global
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 124 of file mpp_domains_util.inc.

mpp_get_compute_domain2d()

subroutine mpp_get_compute_domain2d	(	type(domain2d), intent(in)	domain,
		integer, intent(out), optional	xbegin,
		integer, intent(out), optional	xend,
		integer, intent(out), optional	ybegin,
		integer, intent(out), optional	yend,
		integer, intent(out), optional	xsize,
		integer, intent(out), optional	xmax_size,
		integer, intent(out), optional	ysize,
		integer, intent(out), optional	ymax_size,
		logical, intent(out), optional	x_is_global,
		logical, intent(out), optional	y_is_global,
		integer, intent(in), optional	tile_count,
		integer, intent(in), optional	position
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 178 of file mpp_domains_util.inc.

mpp_get_compute_domains1d()

subroutine mpp_get_compute_domains1d	(	type(domain1d), intent(in)	domain,
		integer, dimension(:), intent(out), optional	begin,
		integer, dimension(:), intent(out), optional	end,
		integer, dimension(:), intent(out), optional	size
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 445 of file mpp_domains_util.inc.

mpp_get_compute_domains2d()

subroutine mpp_get_compute_domains2d	(	type(domain2d), intent(in)	domain,
		integer, dimension(:), intent(out), optional	xbegin,
		integer, dimension(:), intent(out), optional	xend,
		integer, dimension(:), intent(out), optional	xsize,
		integer, dimension(:), intent(out), optional	ybegin,
		integer, dimension(:), intent(out), optional	yend,
		integer, dimension(:), intent(out), optional	ysize,
		integer, intent(in), optional	position
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 471 of file mpp_domains_util.inc.

mpp_get_data_domain1d()

subroutine mpp_get_data_domain1d	(	type(domain1d), intent(in)	domain,
		integer, intent(out), optional	begin,
		integer, intent(out), optional	end,
		integer, intent(out), optional	size,
		integer, intent(out), optional	max_size,
		logical, intent(out), optional	is_global
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 138 of file mpp_domains_util.inc.

mpp_get_data_domain2d()

subroutine mpp_get_data_domain2d	(	type(domain2d), intent(in)	domain,
		integer, intent(out), optional	xbegin,
		integer, intent(out), optional	xend,
		integer, intent(out), optional	ybegin,
		integer, intent(out), optional	yend,
		integer, intent(out), optional	xsize,
		integer, intent(out), optional	xmax_size,
		integer, intent(out), optional	ysize,
		integer, intent(out), optional	ymax_size,
		logical, intent(out), optional	x_is_global,
		logical, intent(out), optional	y_is_global,
		integer, intent(in), optional	tile_count,
		integer, intent(in), optional	position
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 203 of file mpp_domains_util.inc.

mpp_get_domain_commid()

integer function mpp_get_domain_commid

(

integer, intent(in)

domain

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 702 of file mpp_domains_util.inc.

mpp_get_domain_components()

subroutine mpp_get_domain_components	(	type(domain2d), intent(in)	domain,
		type(domain1d), intent(inout), optional	x,
		type(domain1d), intent(inout), optional	y,
		integer, intent(in), optional	tile_count
	)

Retrieve 1D components of 2D decomposition.

It is sometime necessary to have direct recourse to the domain1D types that compose a domain2D object. This call retrieves them.

call mpp_get_domain_components( domain, x, y )

Definition at line 431 of file mpp_domains_util.inc.

mpp_get_domain_extents1d()

subroutine mpp_get_domain_extents1d	(	type(domain2d), intent(in)	domain,
		integer, dimension(0:), intent(inout)	xextent,
		integer, dimension(0:), intent(inout)	yextent
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 617 of file mpp_domains_util.inc.

mpp_get_domain_name()

character(len=name_length) function mpp_get_domain_name

(

type(domain2d), intent(in)

domain

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1441 of file mpp_domains_util.inc.

mpp_get_domain_npes()

integer function mpp_get_domain_npes

(

type(domain2d), intent(in)

domain

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1458 of file mpp_domains_util.inc.

mpp_get_domain_pe()

integer function mpp_get_domain_pe

(

type(domain2d), intent(in)

domain

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 674 of file mpp_domains_util.inc.

mpp_get_domain_pelist()

subroutine mpp_get_domain_pelist	(	type(domain2d), intent(in)	domain,
		integer, dimension(:), intent(out)	pelist
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1469 of file mpp_domains_util.inc.

mpp_get_domain_root_pe()

integer function mpp_get_domain_root_pe

(

type(domain2d), intent(in)

domain

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1450 of file mpp_domains_util.inc.

mpp_get_domain_shift()

subroutine mpp_get_domain_shift	(	type(domain2d), intent(in)	domain,
		integer, intent(out)	ishift,
		integer, intent(out)	jshift,
		integer, intent(in), optional	position
	)

Returns the shift value in x and y-direction according to domain position..

When domain is symmetry, one extra point maybe needed in x- and/or y-direction. This routine will return the shift value based on the position

call mpp_get_domain_shift( domain, ishift, jshift, position )

Parameters

[out]	jshift	return value will be 0 or 1.
[in]	position	position of data. Its value can be CENTER, EAST, NORTH or CORNER.

Definition at line 811 of file mpp_domains_util.inc.

mpp_get_domain_tile_commid()

integer function mpp_get_domain_tile_commid

(

type(domain2d), intent(in)

domain

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 693 of file mpp_domains_util.inc.

mpp_get_domain_tile_root_pe()

integer function mpp_get_domain_tile_root_pe

(

type(domain2d), intent(in)

domain

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 684 of file mpp_domains_util.inc.

mpp_get_f2c_index_coarse()

subroutine mpp_get_f2c_index_coarse	(	type(nest_domain_type), intent(in)	nest_domain,
		integer, intent(out)	is_coarse,
		integer, intent(out)	ie_coarse,
		integer, intent(out)	js_coarse,
		integer, intent(out)	je_coarse,
		integer, intent(in)	nest_level,
		integer, intent(in), optional	position
	)

Parameters

[in]	nest_domain	Holds the information to pass data between fine and coarse grid.
[out]	je_coarse	index in the fine grid of the nested region
[in]	nest_level	level of the nest (> 1 implies a telescoping nest)
[in]	position	Cell position. It value should be CENTER, EAST, CORNER, or NORTH.

Definition at line 1768 of file mpp_define_nest_domains.inc.

mpp_get_f2c_index_fine()

subroutine mpp_get_f2c_index_fine	(	type(nest_domain_type), intent(in)	nest_domain,
		integer, intent(out)	is_coarse,
		integer, intent(out)	ie_coarse,
		integer, intent(out)	js_coarse,
		integer, intent(out)	je_coarse,
		integer, intent(out)	is_fine,
		integer, intent(out)	ie_fine,
		integer, intent(out)	js_fine,
		integer, intent(out)	je_fine,
		integer, intent(in)	nest_level,
		integer, intent(in), optional	position
	)

Parameters

[in]	nest_domain	Holds the information to pass data between fine and coarse grid.
[out]	je_fine	index in the fine grid of the nested region
[out]	je_coarse	index in the coarse grid of the nested region
[in]	nest_level	level of the nest (> 1 implies a telescoping nest)
[in]	position	Cell position. It value should be CENTER, EAST, CORNER, or NORTH.

Definition at line 1719 of file mpp_define_nest_domains.inc.

mpp_get_global_domain1d()

subroutine mpp_get_global_domain1d	(	type(domain1d), intent(in)	domain,
		integer, intent(out), optional	begin,
		integer, intent(out), optional	end,
		integer, intent(out), optional	size,
		integer, intent(out), optional	max_size
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 152 of file mpp_domains_util.inc.

mpp_get_global_domain2d()

subroutine mpp_get_global_domain2d	(	type(domain2d), intent(in)	domain,
		integer, intent(out), optional	xbegin,
		integer, intent(out), optional	xend,
		integer, intent(out), optional	ybegin,
		integer, intent(out), optional	yend,
		integer, intent(out), optional	xsize,
		integer, intent(out), optional	xmax_size,
		integer, intent(out), optional	ysize,
		integer, intent(out), optional	ymax_size,
		integer, intent(in), optional	tile_count,
		integer, intent(in), optional	position
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 228 of file mpp_domains_util.inc.

mpp_get_global_domains1d()

subroutine mpp_get_global_domains1d	(	type(domain1d), intent(in)	domain,
		integer, dimension(:), intent(out), optional	begin,
		integer, dimension(:), intent(out), optional	end,
		integer, dimension(:), intent(out), optional	size
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 530 of file mpp_domains_util.inc.

mpp_get_global_domains2d()

subroutine mpp_get_global_domains2d	(	type(domain2d), intent(in)	domain,
		integer, dimension(:), intent(out), optional	xbegin,
		integer, dimension(:), intent(out), optional	xend,
		integer, dimension(:), intent(out), optional	xsize,
		integer, dimension(:), intent(out), optional	ybegin,
		integer, dimension(:), intent(out), optional	yend,
		integer, dimension(:), intent(out), optional	ysize,
		integer, intent(in), optional	position
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 557 of file mpp_domains_util.inc.

mpp_get_io_domain()

type(domain2d) function, pointer mpp_get_io_domain

(

type(domain2d), intent(in)

domain

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 711 of file mpp_domains_util.inc.

mpp_get_io_domain_layout()

integer function, dimension(2) mpp_get_io_domain_layout

(

type(domain2d), intent(in)

domain

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1487 of file mpp_domains_util.inc.

mpp_get_layout1d()

subroutine mpp_get_layout1d	(	type(domain1d), intent(in)	domain,
		integer, intent(out)	layout
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 773 of file mpp_domains_util.inc.

mpp_get_layout2d()

subroutine mpp_get_layout2d	(	type(domain2d), intent(in)	domain,
		integer, dimension(2), intent(out)	layout
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 789 of file mpp_domains_util.inc.

mpp_get_memory_domain1d()

subroutine mpp_get_memory_domain1d	(	type(domain1d), intent(in)	domain,
		integer, intent(out), optional	begin,
		integer, intent(out), optional	end,
		integer, intent(out), optional	size,
		integer, intent(out), optional	max_size,
		logical, intent(out), optional	is_global
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 164 of file mpp_domains_util.inc.

mpp_get_memory_domain2d()

subroutine mpp_get_memory_domain2d	(	type(domain2d), intent(in)	domain,
		integer, intent(out), optional	xbegin,
		integer, intent(out), optional	xend,
		integer, intent(out), optional	ybegin,
		integer, intent(out), optional	yend,
		integer, intent(out), optional	xsize,
		integer, intent(out), optional	xmax_size,
		integer, intent(out), optional	ysize,
		integer, intent(out), optional	ymax_size,
		logical, intent(out), optional	x_is_global,
		logical, intent(out), optional	y_is_global,
		integer, intent(in), optional	position
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 252 of file mpp_domains_util.inc.

mpp_get_num_overlap()

integer function mpp_get_num_overlap	(	type(domain2d), intent(in)	domain,
		integer, intent(in)	action,
		integer, intent(in)	p,
		integer, intent(in), optional	position
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1279 of file mpp_domains_util.inc.

mpp_get_overlap()

subroutine mpp_get_overlap	(	type(domain2d), intent(in)	domain,
		integer, intent(in)	action,
		integer, intent(in)	p,
		integer, dimension(:), intent(out)	is,
		integer, dimension(:), intent(out)	ie,
		integer, dimension(:), intent(out)	js,
		integer, dimension(:), intent(out)	je,
		integer, dimension(:), intent(out)	dir,
		integer, dimension(:), intent(out)	rot,
		integer, intent(in), optional	position
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1388 of file mpp_domains_util.inc.

mpp_get_pelist1d()

subroutine mpp_get_pelist1d	(	type(domain1d), intent(in)	domain,
		integer, dimension(:), intent(out)	pelist,
		integer, intent(out), optional	pos
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 729 of file mpp_domains_util.inc.

mpp_get_pelist2d()

subroutine mpp_get_pelist2d	(	type(domain2d), intent(in)	domain,
		integer, dimension(:), intent(out)	pelist,
		integer, intent(out), optional	pos
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 753 of file mpp_domains_util.inc.

mpp_get_tile_compute_domains()

subroutine mpp_get_tile_compute_domains	(	type(domain2d), intent(in)	domain,
		integer, dimension(:), intent(out)	xbegin,
		integer, dimension(:), intent(out)	xend,
		integer, dimension(:), intent(out)	ybegin,
		integer, dimension(:), intent(out)	yend,
		integer, intent(in), optional	position
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1236 of file mpp_domains_util.inc.

mpp_get_update_pelist()

subroutine mpp_get_update_pelist	(	type(domain2d), intent(in)	domain,
		integer, intent(in)	action,
		integer, dimension(:), intent(inout)	pelist,
		integer, intent(in), optional	position
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1346 of file mpp_domains_util.inc.

mpp_get_update_size()

subroutine mpp_get_update_size	(	type(domain2d), intent(in)	domain,
		integer, intent(out)	nsend,
		integer, intent(out)	nrecv,
		integer, intent(in), optional	position
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1318 of file mpp_domains_util.inc.

mpp_group_update_initialized()

logical function mpp_group_update_initialized

(

type(mpp_group_update_type), intent(in)

group

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 2388 of file mpp_domains_util.inc.

mpp_group_update_is_set()

logical function mpp_group_update_is_set

(

type(mpp_group_update_type), intent(in)

group

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 2397 of file mpp_domains_util.inc.

mpp_modify_domain1d()

subroutine mpp_modify_domain1d	(	type(domain1d), intent(in)	domain_in,
		type(domain1d), intent(inout)	domain_out,
		integer, intent(in), optional	cbegin,
		integer, intent(in), optional	cend,
		integer, intent(in), optional	gbegin,
		integer, intent(in), optional	gend,
		integer, intent(in), optional	hbegin,
		integer, intent(in), optional	hend
	)

Modifies the exents of a domain.

Parameters

[in]	domain_in	The source domain.
[in,out]	domain_out	The returned domain.
[in]	hend	halo size
[in]	cend	Axis specifications associated with the compute domain of the returned 1D domain.
[in]	gend	Axis specifications associated with the global domain of the returned 1D domain.

Definition at line 7545 of file mpp_domains_define.inc.

mpp_modify_domain2d()

subroutine mpp_modify_domain2d	(	type(domain2d), intent(in)	domain_in,
		type(domain2d), intent(inout)	domain_out,
		integer, intent(in), optional	isc,
		integer, intent(in), optional	iec,
		integer, intent(in), optional	jsc,
		integer, intent(in), optional	jec,
		integer, intent(in), optional	isg,
		integer, intent(in), optional	ieg,
		integer, intent(in), optional	jsg,
		integer, intent(in), optional	jeg,
		integer, intent(in), optional	whalo,
		integer, intent(in), optional	ehalo,
		integer, intent(in), optional	shalo,
		integer, intent(in), optional	nhalo
	)

Parameters

[in]	domain_in	The source domain.
[in,out]	domain_out	The returned domain.
[in]	jec	Zonal and meridional axis specifications associated with the global domain of the returned 2D domain.
[in]	jeg	Zonal axis specifications associated with the global domain of the returned 2D domain.
[in]	nhalo	halo size in x- and y- directions

Definition at line 7581 of file mpp_domains_define.inc.

mpp_set_compute_domain1d()

subroutine mpp_set_compute_domain1d	(	type(domain1d), intent(inout)	domain,
		integer, intent(in), optional	begin,
		integer, intent(in), optional	end,
		integer, intent(in), optional	size,
		logical, intent(in), optional	is_global
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 338 of file mpp_domains_util.inc.

mpp_set_compute_domain2d()

subroutine mpp_set_compute_domain2d	(	type(domain2d), intent(inout)	domain,
		integer, intent(in), optional	xbegin,
		integer, intent(in), optional	xend,
		integer, intent(in), optional	ybegin,
		integer, intent(in), optional	yend,
		integer, intent(in), optional	xsize,
		integer, intent(in), optional	ysize,
		logical, intent(in), optional	x_is_global,
		logical, intent(in), optional	y_is_global,
		integer, intent(in), optional	tile_count
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 351 of file mpp_domains_util.inc.

mpp_set_data_domain1d()

subroutine mpp_set_data_domain1d	(	type(domain1d), intent(inout)	domain,
		integer, intent(in), optional	begin,
		integer, intent(in), optional	end,
		integer, intent(in), optional	size,
		logical, intent(in), optional	is_global
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 368 of file mpp_domains_util.inc.

mpp_set_data_domain2d()

subroutine mpp_set_data_domain2d	(	type(domain2d), intent(inout)	domain,
		integer, intent(in), optional	xbegin,
		integer, intent(in), optional	xend,
		integer, intent(in), optional	ybegin,
		integer, intent(in), optional	yend,
		integer, intent(in), optional	xsize,
		integer, intent(in), optional	ysize,
		logical, intent(in), optional	x_is_global,
		logical, intent(in), optional	y_is_global,
		integer, intent(in), optional	tile_count
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 381 of file mpp_domains_util.inc.

mpp_set_domain_symmetry()

subroutine mpp_set_domain_symmetry	(	type(domain2d), intent(inout)	domain,
		logical, intent(in)	symmetry
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1732 of file mpp_domains_util.inc.

mpp_set_global_domain1d()

subroutine mpp_set_global_domain1d	(	type(domain1d), intent(inout)	domain,
		integer, intent(in), optional	begin,
		integer, intent(in), optional	end,
		integer, intent(in), optional	size
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 398 of file mpp_domains_util.inc.

mpp_set_global_domain2d()

subroutine mpp_set_global_domain2d	(	type(domain2d), intent(inout)	domain,
		integer, intent(in), optional	xbegin,
		integer, intent(in), optional	xend,
		integer, intent(in), optional	ybegin,
		integer, intent(in), optional	yend,
		integer, intent(in), optional	xsize,
		integer, intent(in), optional	ysize,
		integer, intent(in), optional	tile_count
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 409 of file mpp_domains_util.inc.

mpp_set_super_grid_indices()

subroutine mpp_set_super_grid_indices

(

type(domain_axis_spec), intent(inout)

grid

)

Modifies the indices in the domain_axis_spec type to those of the supergrid.

Parameters

[in,out]

grid

domain_axis_spec type

Definition at line 276 of file mpp_domains_util.inc.

mpp_shift_nest_domains()

subroutine mpp_shift_nest_domains	(	type(nest_domain_type), intent(inout)	nest_domain,
		type(domain2d), intent(in), target	domain,
		integer, dimension(:), intent(in)	delta_i_coarse,
		integer, dimension(:), intent(in)	delta_j_coarse,
		integer, intent(in), optional	extra_halo
	)

Based on mpp_define_nest_domains, but just resets positioning of nest Modifies the parent/coarse start and end indices of the nest location Computes new overlaps of nest PEs on parent PEs Ramstrom/HRD Moving Nest.

Parameters

[in,out]	nest_domain	holds the information to pass data between nest and parent grids.
[in]	domain	domain for the grid defined in the current pelist
[in]	delta_i_coarse	Array of deltas of coarse grid in y direction
[in]	delta_j_coarse	Array of deltas of coarse grid in y direction
[in]	extra_halo	Extra halo size

Definition at line 387 of file mpp_define_nest_domains.inc.

nullify_domain2d_list()

subroutine nullify_domain2d_list

(

type(domain2d), intent(inout)

domain

)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 970 of file mpp_domains_util.inc.

set_group_update()

subroutine set_group_update	(	type(mpp_group_update_type), intent(inout)	group,
		type(domain2d), intent(inout)	domain
	)

Set user stack size.

This sets the size of an array that is used for internal storage by mpp_domains. This array is used, for instance, to buffer the data sent and received in halo updates.
This call has implied global synchronization. It should be placed somewhere where all PEs can call it.

Definition at line 1919 of file mpp_domains_util.inc.

Variable Documentation

debug_update_domain

character(len=32) debug_update_domain = "none"

private

namelist interface

when debug_update_domain = none, no debug will be done. When debug_update_domain is set to fatal, the run will be exited with fatal error message When debug_update_domain is set to warning, the run will output warning message. When debug update_domain is set to note, the run will output some note message.

Definition at line 730 of file mpp_domains.F90.