1 2 #include <../src/mat/impls/aij/seq/aij.h> 3 #include <../src/mat/impls/baij/seq/baij.h> 4 5 /* 6 This routine is shared by SeqAIJ and SeqBAIJ matrices, 7 since it operators only on the nonzero structure of the elements or blocks. 8 */ 9 #undef __FUNCT__ 10 #define __FUNCT__ "MatFDColoringCreate_SeqXAIJ" 11 PetscErrorCode MatFDColoringCreate_SeqXAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c) 12 { 13 PetscErrorCode ierr; 14 PetscInt bs,nis=iscoloring->n; 15 PetscBool isBAIJ; 16 17 PetscFunctionBegin; 18 /* set default brows and bcols for speedup inserting the dense matrix into sparse Jacobian */ 19 ierr = MatGetBlockSize(mat,&bs);CHKERRQ(ierr); 20 ierr = PetscObjectTypeCompare((PetscObject)mat,MATSEQBAIJ,&isBAIJ);CHKERRQ(ierr); 21 if (isBAIJ) { /* brows and bcols will not be used */ 22 c->brows = 0; 23 c->bcols = 0; 24 } else { /* seqaij matrix */ 25 /* bcols is chosen s.t. dy-array takes 50% of memory space as mat */ 26 Mat_SeqAIJ *spA = (Mat_SeqAIJ*)mat->data; 27 PetscReal mem; 28 PetscInt nz,brows,bcols,m=mat->rmap->n; 29 30 bs = 1; /* only bs=1 is supported for SeqAIJ matrix */ 31 32 nz = spA->nz; 33 mem = nz*(sizeof(PetscScalar) + sizeof(PetscInt)) + 3*m*sizeof(PetscInt); 34 bcols = (PetscInt)(0.5*mem /(m*sizeof(PetscScalar))); 35 brows = 1000/bcols; 36 if (bcols > nis) bcols = nis; 37 if (brows == 0 || brows > m) brows = m; 38 c->brows = brows; 39 c->bcols = bcols; 40 } 41 42 c->M = mat->rmap->N/bs; /* set total rows, columns and local rows */ 43 c->N = mat->cmap->N/bs; 44 c->m = mat->rmap->N/bs; 45 c->rstart = 0; 46 c->ncolors = nis; 47 c->ctype = IS_COLORING_GHOSTED; 48 PetscFunctionReturn(0); 49 } 50 51 #undef __FUNCT__ 52 #define __FUNCT__ "MatFDColoringSetUp_SeqXAIJ" 53 PetscErrorCode MatFDColoringSetUp_SeqXAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c) 54 { 55 PetscErrorCode ierr; 56 PetscInt i,n,nrows,mbs=c->m,j,k,m,ncols,col,nis=iscoloring->n,*rowhit,bs,bs2,*spidx,nz; 57 const PetscInt *is,*row,*ci,*cj; 58 IS *isa; 59 PetscBool isBAIJ; 60 PetscScalar *A_val,**valaddrhit; 61 MatEntry *Jentry; 62 PetscInt *color_start; 63 PetscInt bcols=c->bcols; 64 65 PetscFunctionBegin; 66 ierr = ISColoringGetIS(iscoloring,PETSC_IGNORE,&isa);CHKERRQ(ierr); 67 68 ierr = MatGetBlockSize(mat,&bs);CHKERRQ(ierr); 69 ierr = PetscObjectTypeCompare((PetscObject)mat,MATSEQBAIJ,&isBAIJ);CHKERRQ(ierr); 70 if (isBAIJ) { 71 Mat_SeqBAIJ *spA = (Mat_SeqBAIJ*)mat->data; 72 A_val = spA->a; 73 nz = spA->nz; 74 } else { 75 Mat_SeqAIJ *spA = (Mat_SeqAIJ*)mat->data; 76 A_val = spA->a; 77 nz = spA->nz; 78 bs = 1; /* only bs=1 is supported for SeqAIJ matrix */ 79 } 80 81 ierr = PetscMalloc(nis*sizeof(PetscInt),&c->ncolumns);CHKERRQ(ierr); 82 ierr = PetscMalloc(nis*sizeof(PetscInt*),&c->columns);CHKERRQ(ierr); 83 ierr = PetscMalloc(nis*sizeof(PetscInt),&c->nrows);CHKERRQ(ierr); 84 ierr = PetscLogObjectMemory((PetscObject)c,3*nis*sizeof(PetscInt));CHKERRQ(ierr); 85 86 ierr = PetscMalloc(nz*sizeof(MatEntry),&Jentry);CHKERRQ(ierr); 87 ierr = PetscLogObjectMemory((PetscObject)c,nz*sizeof(MatEntry));CHKERRQ(ierr); 88 c->matentry = Jentry; 89 90 if (isBAIJ) { 91 ierr = MatGetColumnIJ_SeqBAIJ_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);CHKERRQ(ierr); 92 } else { 93 ierr = MatGetColumnIJ_SeqAIJ_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);CHKERRQ(ierr); 94 } 95 96 ierr = PetscMalloc3(c->m,PetscInt,&rowhit,c->m,PetscScalar*,&valaddrhit,nis+1,PetscInt,&color_start);CHKERRQ(ierr); 97 ierr = PetscMemzero(rowhit,c->m*sizeof(PetscInt));CHKERRQ(ierr); 98 99 nz = 0; 100 for (i=0; i<nis; i++) { /* loop over colors */ 101 color_start[i] = nz; 102 103 ierr = ISGetLocalSize(isa[i],&n);CHKERRQ(ierr); 104 ierr = ISGetIndices(isa[i],&is);CHKERRQ(ierr); 105 106 c->ncolumns[i] = n; 107 if (n) { 108 ierr = PetscMalloc(n*sizeof(PetscInt),&c->columns[i]);CHKERRQ(ierr); 109 ierr = PetscLogObjectMemory((PetscObject)c,n*sizeof(PetscInt));CHKERRQ(ierr); 110 ierr = PetscMemcpy(c->columns[i],is,n*sizeof(PetscInt));CHKERRQ(ierr); 111 } else { 112 c->columns[i] = 0; 113 } 114 115 /* fast, crude version requires O(N*N) work */ 116 bs2 = bs*bs; 117 nrows = 0; 118 for (j=0; j<n; j++) { /* loop over columns */ 119 col = is[j]; 120 row = cj + ci[col]; 121 m = ci[col+1] - ci[col]; 122 nrows += m; 123 for (k=0; k<m; k++) { /* loop over columns marking them in rowhit */ 124 rowhit[*row] = col + 1; 125 valaddrhit[*row++] = &A_val[bs2*spidx[ci[col] + k]]; 126 } 127 } 128 c->nrows[i] = nrows; /* total num of rows for this color */ 129 130 for (j=0; j<mbs; j++) { /* loop over rows */ 131 if (rowhit[j]) { 132 Jentry[nz].row = j; /* local row index */ 133 Jentry[nz].col = rowhit[j] - 1; /* local column index */ 134 Jentry[nz].valaddr = valaddrhit[j]; /* address of mat value for this entry */ 135 nz++; 136 rowhit[j] = 0.0; /* zero rowhit for reuse */ 137 } 138 } 139 ierr = ISRestoreIndices(isa[i],&is);CHKERRQ(ierr); 140 } 141 color_start[nis] = nz; 142 143 /*---------- reorder Jentry for faster MatFDColoringApply() ------------*/ 144 if (!isBAIJ && bcols > 1) { 145 PetscInt nbcols=0,brows=c->brows,*row_start,*nrows_new,nz_new,row_end; 146 MatEntry *Jentry_new; 147 148 m = mbs; 149 if (brows < 1 || brows > m) brows = m; 150 151 ierr = PetscMalloc(nz*sizeof(MatEntry),&Jentry_new);CHKERRQ(ierr); 152 ierr = PetscMalloc(bcols*sizeof(PetscInt),&row_start);CHKERRQ(ierr); 153 ierr = PetscMalloc(nis*sizeof(PetscInt),&nrows_new);CHKERRQ(ierr); 154 155 nz_new = 0; 156 for (i=0; i<nis; i+=bcols) { /* loop over colors */ 157 if (i + bcols > nis) bcols = nis - i; 158 159 row_end = brows; 160 if (row_end > mbs) row_end = mbs; 161 for (j=0; j<bcols; j++) row_start[j] = 0; 162 while (row_end <= mbs) { /* loop over block rows */ 163 for (j=0; j<bcols; j++) { /* loop over block columns */ 164 nrows = c->nrows[i+j]; 165 nz = color_start[i+j]; 166 while (row_start[j] < nrows) { 167 if (Jentry[nz].row >= row_end) { 168 color_start[i+j] = nz; 169 break; 170 } else { /* copy Jentry[nz] to Jentry_new[nz_new] */ 171 Jentry_new[nz_new].row = Jentry[nz].row + j*mbs; /* index in dy-array */ 172 Jentry_new[nz_new].col = Jentry[nz].col; 173 Jentry_new[nz_new].valaddr = Jentry[nz].valaddr; 174 nz_new++; nz++; row_start[j]++; 175 } 176 } 177 } 178 if (row_end == mbs) break; 179 row_end += brows; 180 if (row_end > mbs) row_end = mbs; 181 } 182 nrows_new[nbcols++] = nz_new; 183 } 184 for (i=nbcols-1; i>0; i--) nrows_new[i] -= nrows_new[i-1]; 185 ierr = PetscFree(c->nrows);CHKERRQ(ierr); 186 c->nrows = nrows_new; 187 188 ierr = PetscFree(Jentry);CHKERRQ(ierr); 189 c->matentry = Jentry_new; 190 ierr = PetscMalloc(c->bcols*mat->rmap->n*sizeof(PetscScalar),&c->dy);CHKERRQ(ierr); 191 ierr = PetscFree(row_start);CHKERRQ(ierr); 192 } 193 /*---------------------------------------*/ 194 195 if (isBAIJ) { 196 ierr = MatRestoreColumnIJ_SeqBAIJ_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);CHKERRQ(ierr); 197 ierr = PetscMalloc(bs*mat->rmap->n*sizeof(PetscScalar),&c->dy);CHKERRQ(ierr); 198 } else { 199 ierr = MatRestoreColumnIJ_SeqAIJ_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);CHKERRQ(ierr); 200 } 201 ierr = PetscFree3(rowhit,valaddrhit,color_start);CHKERRQ(ierr); 202 ierr = ISColoringRestoreIS(iscoloring,&isa);CHKERRQ(ierr); 203 204 ierr = VecCreateGhost(PetscObjectComm((PetscObject)mat),mat->rmap->n,PETSC_DETERMINE,0,NULL,&c->vscale);CHKERRQ(ierr); 205 PetscFunctionReturn(0); 206 } 207