1 #define PETSCMAT_DLL 2 3 /* 4 Defines a matrix-vector product for the MATMPIAIJCRL matrix class. 5 This class is derived from the MATMPIAIJ class and retains the 6 compressed row storage (aka Yale sparse matrix format) but augments 7 it with a column oriented storage that is more efficient for 8 matrix vector products on Vector machines. 9 10 CRL stands for constant row length (that is the same number of columns 11 is kept (padded with zeros) for each row of the sparse matrix. 12 13 See src/mat/impls/aij/seq/crl/crl.c for the sequential version 14 */ 15 16 #include "../src/mat/impls/aij/mpi/mpiaij.h" 17 #include "../src/mat/impls/aij/seq/crl/crl.h" 18 19 extern PetscErrorCode MatDestroy_MPIAIJ(Mat); 20 21 #undef __FUNCT__ 22 #define __FUNCT__ "MatDestroy_MPIAIJCRL" 23 PetscErrorCode MatDestroy_MPIAIJCRL(Mat A) 24 { 25 PetscErrorCode ierr; 26 Mat_AIJCRL *aijcrl = (Mat_AIJCRL *) A->spptr; 27 28 /* Free everything in the Mat_AIJCRL data structure. */ 29 ierr = PetscFree2(aijcrl->acols,aijcrl->icols);CHKERRQ(ierr); 30 if (aijcrl->fwork) { 31 ierr = VecDestroy(aijcrl->fwork);CHKERRQ(ierr); 32 } 33 if (aijcrl->xwork) { 34 ierr = VecDestroy(aijcrl->xwork);CHKERRQ(ierr); 35 } 36 ierr = PetscFree(aijcrl->array);CHKERRQ(ierr); 37 ierr = PetscFree(A->spptr);CHKERRQ(ierr); 38 39 ierr = PetscObjectChangeTypeName( (PetscObject)A, MATMPIAIJ);CHKERRQ(ierr); 40 ierr = MatDestroy_MPIAIJ(A);CHKERRQ(ierr); 41 PetscFunctionReturn(0); 42 } 43 44 #undef __FUNCT__ 45 #define __FUNCT__ "MPIAIJCRL_create_aijcrl" 46 PetscErrorCode MPIAIJCRL_create_aijcrl(Mat A) 47 { 48 Mat_MPIAIJ *a = (Mat_MPIAIJ *)(A)->data; 49 Mat_SeqAIJ *Aij = (Mat_SeqAIJ*)(a->A->data), *Bij = (Mat_SeqAIJ*)(a->B->data); 50 Mat_AIJCRL *aijcrl = (Mat_AIJCRL*) A->spptr; 51 PetscInt m = A->rmap->n; /* Number of rows in the matrix. */ 52 PetscInt nd = a->A->cmap->n; /* number of columns in diagonal portion */ 53 PetscInt *aj = Aij->j,*bj = Bij->j; /* From the CSR representation; points to the beginning of each row. */ 54 PetscInt i, j,rmax = 0,*icols, *ailen = Aij->ilen, *bilen = Bij->ilen; 55 PetscScalar *aa = Aij->a,*ba = Bij->a,*acols,*array; 56 PetscErrorCode ierr; 57 58 PetscFunctionBegin; 59 /* determine the row with the most columns */ 60 for (i=0; i<m; i++) { 61 rmax = PetscMax(rmax,ailen[i]+bilen[i]); 62 } 63 aijcrl->nz = Aij->nz+Bij->nz; 64 aijcrl->m = A->rmap->n; 65 aijcrl->rmax = rmax; 66 ierr = PetscFree2(aijcrl->acols,aijcrl->icols);CHKERRQ(ierr); 67 ierr = PetscMalloc2(rmax*m,PetscScalar,&aijcrl->acols,rmax*m,PetscInt,&aijcrl->icols);CHKERRQ(ierr); 68 acols = aijcrl->acols; 69 icols = aijcrl->icols; 70 for (i=0; i<m; i++) { 71 for (j=0; j<ailen[i]; j++) { 72 acols[j*m+i] = *aa++; 73 icols[j*m+i] = *aj++; 74 } 75 for (;j<ailen[i]+bilen[i]; j++) { 76 acols[j*m+i] = *ba++; 77 icols[j*m+i] = nd + *bj++; 78 } 79 for (;j<rmax; j++) { /* empty column entries */ 80 acols[j*m+i] = 0.0; 81 icols[j*m+i] = (j) ? icols[(j-1)*m+i] : 0; /* handle case where row is EMPTY */ 82 } 83 } 84 ierr = PetscInfo1(A,"Percentage of 0's introduced for vectorized multiply %g\n",1.0-((double)(aijcrl->nz))/((double)(rmax*m))); 85 86 ierr = PetscFree(aijcrl->array);CHKERRQ(ierr); 87 ierr = PetscMalloc((a->B->cmap->n+nd)*sizeof(PetscScalar),&array);CHKERRQ(ierr); 88 /* xwork array is actually B->n+nd long, but we define xwork this length so can copy into it */ 89 if (aijcrl->xwork) {ierr = VecDestroy(aijcrl->xwork);CHKERRQ(ierr);} 90 ierr = VecCreateMPIWithArray(((PetscObject)A)->comm,nd,PETSC_DECIDE,array,&aijcrl->xwork);CHKERRQ(ierr); 91 if (aijcrl->fwork) {ierr = VecDestroy(aijcrl->fwork);CHKERRQ(ierr);} 92 ierr = VecCreateSeqWithArray(PETSC_COMM_SELF,a->B->cmap->n,array+nd,&aijcrl->fwork);CHKERRQ(ierr); 93 aijcrl->array = array; 94 aijcrl->xscat = a->Mvctx; 95 PetscFunctionReturn(0); 96 } 97 98 extern PetscErrorCode MatAssemblyEnd_MPIAIJ(Mat,MatAssemblyType); 99 100 #undef __FUNCT__ 101 #define __FUNCT__ "MatAssemblyEnd_MPIAIJCRL" 102 PetscErrorCode MatAssemblyEnd_MPIAIJCRL(Mat A, MatAssemblyType mode) 103 { 104 PetscErrorCode ierr; 105 Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data; 106 Mat_SeqAIJ *Aij = (Mat_SeqAIJ*)(a->A->data), *Bij = (Mat_SeqAIJ*)(a->A->data); 107 108 PetscFunctionBegin; 109 Aij->inode.use = PETSC_FALSE; 110 Bij->inode.use = PETSC_FALSE; 111 ierr = MatAssemblyEnd_MPIAIJ(A,mode);CHKERRQ(ierr); 112 if (mode == MAT_FLUSH_ASSEMBLY) PetscFunctionReturn(0); 113 114 /* Now calculate the permutation and grouping information. */ 115 ierr = MPIAIJCRL_create_aijcrl(A);CHKERRQ(ierr); 116 PetscFunctionReturn(0); 117 } 118 119 extern PetscErrorCode MatMult_AIJCRL(Mat,Vec,Vec); 120 extern PetscErrorCode MatDuplicate_AIJCRL(Mat,MatDuplicateOption,Mat*); 121 122 /* MatConvert_MPIAIJ_MPIAIJCRL converts a MPIAIJ matrix into a 123 * MPIAIJCRL matrix. This routine is called by the MatCreate_MPIAIJCRL() 124 * routine, but can also be used to convert an assembled MPIAIJ matrix 125 * into a MPIAIJCRL one. */ 126 EXTERN_C_BEGIN 127 #undef __FUNCT__ 128 #define __FUNCT__ "MatConvert_MPIAIJ_MPIAIJCRL" 129 PetscErrorCode MatConvert_MPIAIJ_MPIAIJCRL(Mat A,const MatType type,MatReuse reuse,Mat *newmat) 130 { 131 PetscErrorCode ierr; 132 Mat B = *newmat; 133 Mat_AIJCRL *aijcrl; 134 135 PetscFunctionBegin; 136 if (reuse == MAT_INITIAL_MATRIX) { 137 ierr = MatDuplicate(A,MAT_COPY_VALUES,&B);CHKERRQ(ierr); 138 } 139 140 ierr = PetscNewLog(B,Mat_AIJCRL,&aijcrl);CHKERRQ(ierr); 141 B->spptr = (void *) aijcrl; 142 143 /* Set function pointers for methods that we inherit from AIJ but override. */ 144 B->ops->duplicate = MatDuplicate_AIJCRL; 145 B->ops->assemblyend = MatAssemblyEnd_MPIAIJCRL; 146 B->ops->destroy = MatDestroy_MPIAIJCRL; 147 B->ops->mult = MatMult_AIJCRL; 148 149 /* If A has already been assembled, compute the permutation. */ 150 if (A->assembled) { 151 ierr = MPIAIJCRL_create_aijcrl(B);CHKERRQ(ierr); 152 } 153 ierr = PetscObjectChangeTypeName((PetscObject)B,MATMPIAIJCRL);CHKERRQ(ierr); 154 *newmat = B; 155 PetscFunctionReturn(0); 156 } 157 EXTERN_C_END 158 159 160 #undef __FUNCT__ 161 #define __FUNCT__ "MatCreateMPIAIJCRL" 162 /*@C 163 MatCreateMPIAIJCRL - Creates a sparse matrix of type MPIAIJCRL. 164 This type inherits from AIJ, but stores some additional 165 information that is used to allow better vectorization of 166 the matrix-vector product. At the cost of increased storage, the AIJ formatted 167 matrix can be copied to a format in which pieces of the matrix are 168 stored in ELLPACK format, allowing the vectorized matrix multiply 169 routine to use stride-1 memory accesses. As with the AIJ type, it is 170 important to preallocate matrix storage in order to get good assembly 171 performance. 172 173 Collective on MPI_Comm 174 175 Input Parameters: 176 + comm - MPI communicator, set to PETSC_COMM_SELF 177 . m - number of rows 178 . n - number of columns 179 . nz - number of nonzeros per row (same for all rows) 180 - nnz - array containing the number of nonzeros in the various rows 181 (possibly different for each row) or PETSC_NULL 182 183 Output Parameter: 184 . A - the matrix 185 186 Notes: 187 If nnz is given then nz is ignored 188 189 Level: intermediate 190 191 .keywords: matrix, cray, sparse, parallel 192 193 .seealso: MatCreate(), MatCreateMPIAIJPERM(), MatSetValues() 194 @*/ 195 PetscErrorCode MatCreateMPIAIJCRL(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt nz,const PetscInt nnz[],PetscInt onz,const PetscInt onnz[],Mat *A) 196 { 197 PetscErrorCode ierr; 198 199 PetscFunctionBegin; 200 ierr = MatCreate(comm,A);CHKERRQ(ierr); 201 ierr = MatSetSizes(*A,m,n,m,n);CHKERRQ(ierr); 202 ierr = MatSetType(*A,MATMPIAIJCRL);CHKERRQ(ierr); 203 ierr = MatMPIAIJSetPreallocation_MPIAIJ(*A,nz,(PetscInt*)nnz,onz,(PetscInt*)onnz);CHKERRQ(ierr); 204 PetscFunctionReturn(0); 205 } 206 207 208 EXTERN_C_BEGIN 209 #undef __FUNCT__ 210 #define __FUNCT__ "MatCreate_MPIAIJCRL" 211 PetscErrorCode MatCreate_MPIAIJCRL(Mat A) 212 { 213 PetscErrorCode ierr; 214 215 PetscFunctionBegin; 216 ierr = MatSetType(A,MATMPIAIJ);CHKERRQ(ierr); 217 ierr = MatConvert_MPIAIJ_MPIAIJCRL(A,MATMPIAIJCRL,MAT_REUSE_MATRIX,&A);CHKERRQ(ierr); 218 PetscFunctionReturn(0); 219 } 220 EXTERN_C_END 221 222