1 #include <../src/mat/impls/aij/mpi/mpiaij.h> 2 /*@C 3 MatCreateMPIAIJMKL - Creates a sparse parallel matrix whose local 4 portions are stored as SEQAIJMKL matrices (a matrix class that inherits 5 from SEQAIJ but uses some operations provided by Intel MKL). The same 6 guidelines that apply to MPIAIJ matrices for preallocating the matrix 7 storage apply here as well. 8 9 Collective 10 11 Input Parameters: 12 + comm - MPI communicator 13 . m - number of local rows (or PETSC_DECIDE to have calculated if M is given) 14 This value should be the same as the local size used in creating the 15 y vector for the matrix-vector product y = Ax. 16 . n - This value should be the same as the local size used in creating the 17 x vector for the matrix-vector product y = Ax. (or PETSC_DECIDE to have 18 calculated if N is given) For square matrices n is almost always m. 19 . M - number of global rows (or PETSC_DETERMINE to have calculated if m is given) 20 . N - number of global columns (or PETSC_DETERMINE to have calculated if n is given) 21 . d_nz - number of nonzeros per row in DIAGONAL portion of local submatrix 22 (same value is used for all local rows) 23 . d_nnz - array containing the number of nonzeros in the various rows of the 24 DIAGONAL portion of the local submatrix (possibly different for each row) 25 or NULL, if d_nz is used to specify the nonzero structure. 26 The size of this array is equal to the number of local rows, i.e 'm'. 27 For matrices you plan to factor you must leave room for the diagonal entry and 28 put in the entry even if it is zero. 29 . o_nz - number of nonzeros per row in the OFF-DIAGONAL portion of local 30 submatrix (same value is used for all local rows). 31 - o_nnz - array containing the number of nonzeros in the various rows of the 32 OFF-DIAGONAL portion of the local submatrix (possibly different for 33 each row) or NULL, if o_nz is used to specify the nonzero 34 structure. The size of this array is equal to the number 35 of local rows, i.e 'm'. 36 37 Output Parameter: 38 . A - the matrix 39 40 Notes: 41 If the *_nnz parameter is given then the *_nz parameter is ignored 42 43 m,n,M,N parameters specify the size of the matrix, and its partitioning across 44 processors, while d_nz,d_nnz,o_nz,o_nnz parameters specify the approximate 45 storage requirements for this matrix. 46 47 If PETSC_DECIDE or PETSC_DETERMINE is used for a particular argument on one 48 processor than it must be used on all processors that share the object for 49 that argument. 50 51 The user MUST specify either the local or global matrix dimensions 52 (possibly both). 53 54 The parallel matrix is partitioned such that the first m0 rows belong to 55 process 0, the next m1 rows belong to process 1, the next m2 rows belong 56 to process 2 etc.. where m0,m1,m2... are the input parameter 'm'. 57 58 The DIAGONAL portion of the local submatrix of a processor can be defined 59 as the submatrix which is obtained by extraction the part corresponding 60 to the rows r1-r2 and columns r1-r2 of the global matrix, where r1 is the 61 first row that belongs to the processor, and r2 is the last row belonging 62 to the this processor. This is a square mxm matrix. The remaining portion 63 of the local submatrix (mxN) constitute the OFF-DIAGONAL portion. 64 65 If o_nnz, d_nnz are specified, then o_nz, and d_nz are ignored. 66 67 When calling this routine with a single process communicator, a matrix of 68 type SEQAIJMKL is returned. If a matrix of type MPIAIJMKL is desired 69 for this type of communicator, use the construction mechanism: 70 MatCreate(...,&A); MatSetType(A,MPIAIJMKL); MatMPIAIJSetPreallocation(A,...); 71 72 Options Database Keys: 73 . -mat_aijmkl_no_spmv2 - disables use of the SpMV2 inspector-executor routines 74 75 Level: intermediate 76 77 .seealso: MatCreate(), MatCreateSeqAIJMKL(), MatSetValues() 78 @*/ 79 PetscErrorCode MatCreateMPIAIJMKL(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) 80 { 81 PetscErrorCode ierr; 82 PetscMPIInt size; 83 84 PetscFunctionBegin; 85 ierr = MatCreate(comm,A);CHKERRQ(ierr); 86 ierr = MatSetSizes(*A,m,n,M,N);CHKERRQ(ierr); 87 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 88 if (size > 1) { 89 ierr = MatSetType(*A,MATMPIAIJMKL);CHKERRQ(ierr); 90 ierr = MatMPIAIJSetPreallocation(*A,d_nz,d_nnz,o_nz,o_nnz);CHKERRQ(ierr); 91 } else { 92 ierr = MatSetType(*A,MATSEQAIJMKL);CHKERRQ(ierr); 93 ierr = MatSeqAIJSetPreallocation(*A,d_nz,d_nnz);CHKERRQ(ierr); 94 } 95 PetscFunctionReturn(0); 96 } 97 98 PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJMKL(Mat,MatType,MatReuse,Mat*); 99 100 PetscErrorCode MatMPIAIJSetPreallocation_MPIAIJMKL(Mat B,PetscInt d_nz,const PetscInt d_nnz[],PetscInt o_nz,const PetscInt o_nnz[]) 101 { 102 Mat_MPIAIJ *b = (Mat_MPIAIJ*)B->data; 103 PetscErrorCode ierr; 104 105 PetscFunctionBegin; 106 ierr = MatMPIAIJSetPreallocation_MPIAIJ(B,d_nz,d_nnz,o_nz,o_nnz);CHKERRQ(ierr); 107 ierr = MatConvert_SeqAIJ_SeqAIJMKL(b->A, MATSEQAIJMKL, MAT_INPLACE_MATRIX, &b->A);CHKERRQ(ierr); 108 ierr = MatConvert_SeqAIJ_SeqAIJMKL(b->B, MATSEQAIJMKL, MAT_INPLACE_MATRIX, &b->B);CHKERRQ(ierr); 109 PetscFunctionReturn(0); 110 } 111 112 PETSC_INTERN PetscErrorCode MatConvert_MPIAIJ_MPIAIJMKL(Mat A,MatType type,MatReuse reuse,Mat *newmat) 113 { 114 PetscErrorCode ierr; 115 Mat B = *newmat; 116 117 PetscFunctionBegin; 118 if (reuse == MAT_INITIAL_MATRIX) { 119 ierr = MatDuplicate(A,MAT_COPY_VALUES,&B);CHKERRQ(ierr); 120 } 121 122 ierr = PetscObjectChangeTypeName((PetscObject) B, MATMPIAIJMKL);CHKERRQ(ierr); 123 ierr = PetscObjectComposeFunction((PetscObject)B,"MatMPIAIJSetPreallocation_C",MatMPIAIJSetPreallocation_MPIAIJMKL);CHKERRQ(ierr); 124 *newmat = B; 125 PetscFunctionReturn(0); 126 } 127 128 PETSC_EXTERN PetscErrorCode MatCreate_MPIAIJMKL(Mat A) 129 { 130 PetscErrorCode ierr; 131 132 PetscFunctionBegin; 133 ierr = MatSetType(A,MATMPIAIJ);CHKERRQ(ierr); 134 ierr = MatConvert_MPIAIJ_MPIAIJMKL(A,MATMPIAIJMKL,MAT_INPLACE_MATRIX,&A);CHKERRQ(ierr); 135 PetscFunctionReturn(0); 136 } 137 138 /*MC 139 MATAIJMKL - MATAIJMKL = "AIJMKL" - A matrix type to be used for sparse matrices. 140 141 This matrix type is identical to MATSEQAIJMKL when constructed with a single process communicator, 142 and MATMPIAIJMKL otherwise. As a result, for single process communicators, 143 MatSeqAIJSetPreallocation() is supported, and similarly MatMPIAIJSetPreallocation() is supported 144 for communicators controlling multiple processes. It is recommended that you call both of 145 the above preallocation routines for simplicity. 146 147 Options Database Keys: 148 . -mat_type aijmkl - sets the matrix type to "AIJMKL" during a call to MatSetFromOptions() 149 150 Level: beginner 151 152 .seealso: MatCreateMPIAIJMKL(), MATSEQAIJMKL, MATMPIAIJMKL 153 M*/ 154 155