Actual source code: mpiaijperm.c

  1: #include <../src/mat/impls/aij/mpi/mpiaij.h>
  2: /*@C
  3:   MatCreateMPIAIJPERM - Creates a sparse parallel matrix whose local
  4:   portions are stored as `MATSEQAIJPERM` matrices (a matrix class that inherits
  5:   from SEQAIJ but includes some optimizations to allow more effective
  6:   vectorization).

  8:   Collective

 10:   Input Parameters:
 11: + comm  - MPI communicator
 12: . m     - number of local rows (or `PETSC_DECIDE` to have calculated if `M` is given)
 13:            This value should be the same as the local size used in creating the
 14:            y vector for the matrix-vector product y = Ax.
 15: . n     - This value should be the same as the local size used in creating the
 16:        x vector for the matrix-vector product y = Ax. (or PETSC_DECIDE to have
 17:        calculated if `N` is given) For square matrices `n` is almost always `m`.
 18: . M     - number of global rows (or `PETSC_DETERMINE` to have calculated if `m` is given)
 19: . N     - number of global columns (or `PETSC_DETERMINE` to have calculated if `n` is given)
 20: . d_nz  - number of nonzeros per row in DIAGONAL portion of local submatrix
 21:            (same value is used for all local rows)
 22: . d_nnz - array containing the number of nonzeros in the various rows of the
 23:            DIAGONAL portion of the local submatrix (possibly different for each row)
 24:            or `NULL`, if `d_nz` is used to specify the nonzero structure.
 25:            The size of this array is equal to the number of local rows, i.e `m`.
 26:            For matrices you plan to factor you must leave room for the diagonal entry and
 27:            put in the entry even if it is zero.
 28: . o_nz  - number of nonzeros per row in the OFF-DIAGONAL portion of local
 29:            submatrix (same value is used for all local rows).
 30: - o_nnz - array containing the number of nonzeros in the various rows of the
 31:            OFF-DIAGONAL portion of the local submatrix (possibly different for
 32:            each row) or `NULL`, if `o_nz` is used to specify the nonzero
 33:            structure. The size of this array is equal to the number
 34:            of local rows, i.e `m`.

 36:   Output Parameter:
 37: . A - the matrix

 39:   Options Database Keys:
 40: + -mat_no_inode            - Do not use inodes
 41: - -mat_inode_limit <limit> - Sets inode limit (max limit=5)

 43:   Level: intermediate

 45:   Notes:
 46:   If the *_nnz parameter is given then the *_nz parameter is ignored

 48:   `m`,`n`,`M`,`N` parameters specify the size of the matrix, and its partitioning across
 49:   processors, while `d_nz`,`d_nnz`,`o_nz`,`o_nnz` parameters specify the approximate
 50:   storage requirements for this matrix.

 52:   If `PETSC_DECIDE` or  `PETSC_DETERMINE` is used for a particular argument on one
 53:   processor than it must be used on all processors that share the object for
 54:   that argument.

 56:   The user MUST specify either the local or global matrix dimensions
 57:   (possibly both).

 59:   The parallel matrix is partitioned such that the first m0 rows belong to
 60:   process 0, the next m1 rows belong to process 1, the next m2 rows belong
 61:   to process 2 etc.. where m0,m1,m2... are the input parameter `m`.

 63:   The DIAGONAL portion of the local submatrix of a processor can be defined
 64:   as the submatrix which is obtained by extraction the part corresponding
 65:   to the rows r1-r2 and columns r1-r2 of the global matrix, where r1 is the
 66:   first row that belongs to the processor, and r2 is the last row belonging
 67:   to the this processor. This is a square mxm matrix. The remaining portion
 68:   of the local submatrix (mxN) constitute the OFF-DIAGONAL portion.

 70:   If `o_nnz`, `d_nnz` are specified, then `o_nz`, and `d_nz` are ignored.

 72:   When calling this routine with a single process communicator, a matrix of
 73:   type `MATSEQAIJPERM` is returned.  If a matrix of type `MATMPIAIJPERM` is desired
 74:   for this type of communicator, use the construction mechanism
 75: .vb
 76:    MatCreate(...,&A);
 77:    MatSetType(A,MPIAIJ);
 78:    MatMPIAIJSetPreallocation(A,...);
 79: .ve

 81:   By default, this format uses inodes (identical nodes) when possible.
 82:   We search for consecutive rows with the same nonzero structure, thereby
 83:   reusing matrix information to achieve increased efficiency.

 85: .seealso: [](ch_matrices), `Mat`, [Sparse Matrix Creation](sec_matsparse), `MATMPIAIJPERM`, `MatCreate()`, `MatCreateSeqAIJPERM()`, `MatSetValues()`
 86: @*/
 87: PetscErrorCode MatCreateMPIAIJPERM(MPI_Comm comm, PetscInt m, PetscInt n, PetscInt M, PetscInt N, PetscInt d_nz, const PetscInt d_nnz[], PetscInt o_nz, const PetscInt o_nnz[], Mat *A)
 88: {
 89:   PetscMPIInt size;

 91:   PetscFunctionBegin;
 92:   PetscCall(MatCreate(comm, A));
 93:   PetscCall(MatSetSizes(*A, m, n, M, N));
 94:   PetscCallMPI(MPI_Comm_size(comm, &size));
 95:   if (size > 1) {
 96:     PetscCall(MatSetType(*A, MATMPIAIJPERM));
 97:     PetscCall(MatMPIAIJSetPreallocation(*A, d_nz, d_nnz, o_nz, o_nnz));
 98:   } else {
 99:     PetscCall(MatSetType(*A, MATSEQAIJPERM));
100:     PetscCall(MatSeqAIJSetPreallocation(*A, d_nz, d_nnz));
101:   }
102:   PetscFunctionReturn(PETSC_SUCCESS);
103: }

105: static PetscErrorCode MatMPIAIJSetPreallocation_MPIAIJPERM(Mat B, PetscInt d_nz, const PetscInt d_nnz[], PetscInt o_nz, const PetscInt o_nnz[])
106: {
107:   Mat_MPIAIJ *b = (Mat_MPIAIJ *)B->data;

109:   PetscFunctionBegin;
110:   PetscCall(MatMPIAIJSetPreallocation_MPIAIJ(B, d_nz, d_nnz, o_nz, o_nnz));
111:   PetscCall(MatConvert_SeqAIJ_SeqAIJPERM(b->A, MATSEQAIJPERM, MAT_INPLACE_MATRIX, &b->A));
112:   PetscCall(MatConvert_SeqAIJ_SeqAIJPERM(b->B, MATSEQAIJPERM, MAT_INPLACE_MATRIX, &b->B));
113:   PetscFunctionReturn(PETSC_SUCCESS);
114: }

116: PETSC_INTERN PetscErrorCode MatConvert_MPIAIJ_MPIAIJPERM(Mat A, MatType type, MatReuse reuse, Mat *newmat)
117: {
118:   Mat B = *newmat;

120:   PetscFunctionBegin;
121:   if (reuse == MAT_INITIAL_MATRIX) PetscCall(MatDuplicate(A, MAT_COPY_VALUES, &B));

123:   PetscCall(PetscObjectChangeTypeName((PetscObject)B, MATMPIAIJPERM));
124:   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatMPIAIJSetPreallocation_C", MatMPIAIJSetPreallocation_MPIAIJPERM));
125:   *newmat = B;
126:   PetscFunctionReturn(PETSC_SUCCESS);
127: }

129: PETSC_EXTERN PetscErrorCode MatCreate_MPIAIJPERM(Mat A)
130: {
131:   PetscFunctionBegin;
132:   PetscCall(MatSetType(A, MATMPIAIJ));
133:   PetscCall(MatConvert_MPIAIJ_MPIAIJPERM(A, MATMPIAIJPERM, MAT_INPLACE_MATRIX, &A));
134:   PetscFunctionReturn(PETSC_SUCCESS);
135: }

137: /*MC
138:    MATAIJPERM - "AIJPERM" - A matrix type to be used for sparse matrices.

140:    This matrix type is identical to `MATSEQAIJPERM` when constructed with a single process communicator,
141:    and `MATMPIAIJPERM` otherwise.  As a result, for single process communicators,
142:   `MatSeqAIJSetPreallocation()` is supported, and similarly `MatMPIAIJSetPreallocation()` is supported
143:   for communicators controlling multiple processes.  It is recommended that you call both of
144:   the above preallocation routines for simplicity.

146:    Options Database Key:
147: . -mat_type aijperm - sets the matrix type to `MATAIJPERM`

149:   Level: beginner

151: .seealso: [](ch_matrices), `Mat`, `MatCreateMPIAIJPERM()`, `MATSEQAIJPERM`, `MATMPIAIJPERM`, `MATSEQAIJ`, `MATMPIAIJ`, `MATSEQAIJMKL`, `MATMPIAIJMKL`, `MATSEQAIJSELL`, `MATMPIAIJSELL`
152: M*/