1 2 #include <../src/mat/impls/aij/mpi/mpiaij.h> 3 4 extern PetscErrorCode MatCreateColmap_MPIAIJ_Private(Mat); 5 6 #undef __FUNCT__ 7 #define __FUNCT__ "MatFDColoringApply_AIJ_new" 8 PetscErrorCode MatFDColoringApply_AIJ_new(Mat J,MatFDColoring coloring,Vec x1,MatStructure *flag,void *sctx) 9 { 10 PetscErrorCode (*f)(void*,Vec,Vec,void*) = (PetscErrorCode (*)(void*,Vec,Vec,void*))coloring->f; 11 PetscErrorCode ierr; 12 PetscInt k,cstart,cend,l,row,col,nz; 13 PetscScalar dx=0.0,*y,*xx,*w3_array; 14 PetscScalar *vscale_array; 15 PetscReal epsilon=coloring->error_rel,umin=coloring->umin,unorm; 16 Vec w1=coloring->w1,w2=coloring->w2,w3,vscale=coloring->vscale; 17 void *fctx=coloring->fctx; 18 PetscBool flg=PETSC_FALSE; 19 PetscInt ctype=coloring->ctype,nxloc,nrows_k; 20 Mat_MPIAIJ *aij=(Mat_MPIAIJ*)J->data; 21 MatEntry *Jentry=coloring->matentry; 22 const PetscInt ncolors=coloring->ncolors,*ncolumns=coloring->ncolumns,*nrows=coloring->nrows; 23 24 PetscFunctionBegin; 25 ierr = MatSetUnfactored(J);CHKERRQ(ierr); 26 ierr = PetscOptionsGetBool(NULL,"-mat_fd_coloring_dont_rezero",&flg,NULL);CHKERRQ(ierr); 27 if (flg) { 28 ierr = PetscInfo(coloring,"Not calling MatZeroEntries()\n");CHKERRQ(ierr); 29 } else { 30 PetscBool assembled; 31 ierr = MatAssembled(J,&assembled);CHKERRQ(ierr); 32 if (assembled) { 33 ierr = MatZeroEntries(J);CHKERRQ(ierr); 34 } 35 } 36 37 /* create vscale for storing dx */ 38 if (!vscale) { 39 if (ctype == IS_COLORING_GLOBAL && coloring->htype[0] == 'd') { 40 ierr = VecCreateGhost(PetscObjectComm((PetscObject)J),J->cmap->n,PETSC_DETERMINE,aij->B->cmap->n,aij->garray,&vscale);CHKERRQ(ierr); 41 } else if (ctype == IS_COLORING_GHOSTED) { 42 ierr = VecDuplicate(x1,&vscale);CHKERRQ(ierr); 43 } 44 coloring->vscale = vscale; 45 } 46 47 /* (1) Set w1 = F(x1) */ 48 if (!coloring->fset) { 49 ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 50 ierr = (*f)(sctx,x1,w1,fctx);CHKERRQ(ierr); 51 ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 52 } else { 53 coloring->fset = PETSC_FALSE; 54 } 55 56 /* (2) Compute vscale = 1./dx - the local scale factors, including ghost points */ 57 if (coloring->htype[0] == 'w') { 58 /* vscale = dx is a constant scalar */ 59 ierr = VecNorm(x1,NORM_2,&unorm);CHKERRQ(ierr); 60 dx = 1.0/(PetscSqrtReal(1.0 + unorm)*epsilon); 61 } else { 62 ierr = VecGetLocalSize(x1,&nxloc);CHKERRQ(ierr); 63 ierr = VecGetArray(x1,&xx);CHKERRQ(ierr); 64 ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr); 65 for (col=0; col<nxloc; col++) { 66 dx = xx[col]; 67 if (PetscAbsScalar(dx) < umin) { 68 if (PetscRealPart(dx) >= 0.0) dx = umin; 69 else if (PetscRealPart(dx) < 0.0 ) dx = -umin; 70 } 71 dx *= epsilon; 72 vscale_array[col] = 1.0/dx; 73 } 74 ierr = VecRestoreArray(x1,&xx);CHKERRQ(ierr); 75 ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr); 76 } 77 if (ctype == IS_COLORING_GLOBAL && coloring->htype[0] != 'w') { 78 ierr = VecGhostUpdateBegin(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 79 ierr = VecGhostUpdateEnd(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 80 } 81 82 /* (3) Loop over each color */ 83 if (!coloring->w3) { 84 ierr = VecDuplicate(x1,&coloring->w3);CHKERRQ(ierr); 85 ierr = PetscLogObjectParent((PetscObject)coloring,(PetscObject)coloring->w3);CHKERRQ(ierr); 86 } 87 w3 = coloring->w3; 88 89 ierr = VecGetOwnershipRange(x1,&cstart,&cend);CHKERRQ(ierr); /* used by ghosted vscale */ 90 if (vscale) { 91 ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr); 92 } 93 nz = 0; 94 for (k=0; k<ncolors; k++) { 95 coloring->currentcolor = k; 96 97 /* 98 (3-1) Loop over each column associated with color 99 adding the perturbation to the vector w3 = x1 + dx. 100 */ 101 ierr = VecCopy(x1,w3);CHKERRQ(ierr); 102 ierr = VecGetArray(w3,&w3_array);CHKERRQ(ierr); 103 if (ctype == IS_COLORING_GLOBAL) w3_array -= cstart; /* shift pointer so global index can be used */ 104 if (coloring->htype[0] == 'w') { 105 for (l=0; l<ncolumns[k]; l++) { 106 col = coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */ 107 w3_array[col] += 1.0/dx; 108 } 109 } else { /* htype == 'ds' */ 110 vscale_array -= cstart; /* shift pointer so global index can be used */ 111 for (l=0; l<ncolumns[k]; l++) { 112 col = coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */ 113 w3_array[col] += 1.0/vscale_array[col]; 114 } 115 vscale_array += cstart; 116 } 117 if (ctype == IS_COLORING_GLOBAL) w3_array += cstart; 118 ierr = VecRestoreArray(w3,&w3_array);CHKERRQ(ierr); 119 120 /* 121 (3-2) Evaluate function at w3 = x1 + dx (here dx is a vector of perturbations) 122 w2 = F(x1 + dx) - F(x1) 123 */ 124 ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 125 ierr = (*f)(sctx,w3,w2,fctx);CHKERRQ(ierr); 126 ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 127 ierr = VecAXPY(w2,-1.0,w1);CHKERRQ(ierr); 128 129 /* 130 (3-3) Loop over rows of vector, putting results into Jacobian matrix 131 */ 132 nrows_k = nrows[k]; 133 ierr = VecGetArray(w2,&y);CHKERRQ(ierr); 134 if (coloring->htype[0] == 'w') { 135 for (l=0; l<nrows_k; l++) { 136 row = Jentry[nz].row; /* local row index */ 137 *(Jentry[nz++].valaddr) = y[row]*dx; 138 } 139 } else { /* htype == 'ds' */ 140 for (l=0; l<nrows_k; l++) { 141 row = Jentry[nz].row; /* local row index */ 142 *(Jentry[nz].valaddr) = y[row]*vscale_array[Jentry[nz].col]; 143 nz++; 144 } 145 } 146 ierr = VecRestoreArray(w2,&y);CHKERRQ(ierr); 147 } 148 ierr = MatAssemblyBegin(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 149 ierr = MatAssemblyEnd(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 150 if (vscale) { 151 ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr); 152 } 153 154 coloring->currentcolor = -1; 155 PetscFunctionReturn(0); 156 } 157 158 #undef __FUNCT__ 159 #define __FUNCT__ "MatFDColoringCreate_MPIAIJ_new" 160 PetscErrorCode MatFDColoringCreate_MPIAIJ_new(Mat mat,ISColoring iscoloring,MatFDColoring c) 161 { 162 Mat_MPIAIJ *aij=(Mat_MPIAIJ*)mat->data; 163 PetscErrorCode ierr; 164 PetscMPIInt size,*ncolsonproc,*disp,nn; 165 PetscInt i,n,nrows,nrows_i,j,k,m,ncols,col; 166 const PetscInt *is,*A_ci,*A_cj,*B_ci,*B_cj,*row = NULL,*ltog=NULL; 167 PetscInt nis=iscoloring->n,nctot,*cols; 168 PetscInt *rowhit,cstart,cend,colb; 169 IS *isa; 170 ISLocalToGlobalMapping map=mat->cmap->mapping; 171 PetscInt ctype=c->ctype; 172 Mat A=aij->A,B=aij->B; 173 Mat_SeqAIJ *spA=(Mat_SeqAIJ*)A->data,*spB=(Mat_SeqAIJ*)B->data; 174 PetscScalar *A_val=spA->a,*B_val=spB->a; 175 PetscInt spidx; 176 PetscInt *spidxA,*spidxB,nz; 177 PetscScalar **valaddrhit; 178 MatEntry *Jentry; 179 180 PetscFunctionBegin; 181 if (ctype == IS_COLORING_GHOSTED) { 182 if (!map) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_INCOMP,"When using ghosted differencing matrix must have local to global mapping provided with MatSetLocalToGlobalMapping"); 183 ierr = ISLocalToGlobalMappingGetIndices(map,<og);CHKERRQ(ierr); 184 } 185 186 m = mat->rmap->n; 187 cstart = mat->cmap->rstart; 188 cend = mat->cmap->rend; 189 c->M = mat->rmap->N; /* set the global rows and columns and local rows */ 190 c->N = mat->cmap->N; 191 c->m = m; 192 c->rstart = mat->rmap->rstart; 193 194 c->ncolors = nis; 195 ierr = PetscMalloc(nis*sizeof(PetscInt),&c->ncolumns);CHKERRQ(ierr); 196 ierr = PetscMalloc(nis*sizeof(PetscInt*),&c->columns);CHKERRQ(ierr); 197 ierr = PetscMalloc(nis*sizeof(PetscInt),&c->nrows);CHKERRQ(ierr); 198 ierr = PetscLogObjectMemory((PetscObject)c,3*nis*sizeof(PetscInt));CHKERRQ(ierr); 199 200 nz = spA->nz + spB->nz; /* total nonzero entries of mat */ 201 ierr = PetscMalloc(nz*sizeof(MatEntry),&Jentry);CHKERRQ(ierr); 202 ierr = PetscLogObjectMemory((PetscObject)c,nz*sizeof(MatEntry));CHKERRQ(ierr); 203 c->matentry = Jentry; 204 205 /* Allow access to data structures of local part of matrix 206 - creates aij->colmap which maps global column number to local number in part B */ 207 if (!aij->colmap) { 208 ierr = MatCreateColmap_MPIAIJ_Private(mat);CHKERRQ(ierr); 209 } 210 ierr = MatGetColumnIJ_SeqAIJ_Color(aij->A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr); 211 ierr = MatGetColumnIJ_SeqAIJ_Color(aij->B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr); 212 213 ierr = PetscMalloc2(m+1,PetscInt,&rowhit,m+1,PetscScalar*,&valaddrhit);CHKERRQ(ierr); 214 nz = 0; 215 ierr = ISColoringGetIS(iscoloring,PETSC_IGNORE,&isa);CHKERRQ(ierr); 216 for (i=0; i<nis; i++) { /* for each local color */ 217 ierr = ISGetLocalSize(isa[i],&n);CHKERRQ(ierr); 218 ierr = ISGetIndices(isa[i],&is);CHKERRQ(ierr); 219 220 c->ncolumns[i] = n; /* local number of columns of this color on this process */ 221 if (n) { 222 ierr = PetscMalloc(n*sizeof(PetscInt),&c->columns[i]);CHKERRQ(ierr); 223 ierr = PetscLogObjectMemory((PetscObject)c,n*sizeof(PetscInt));CHKERRQ(ierr); 224 ierr = PetscMemcpy(c->columns[i],is,n*sizeof(PetscInt));CHKERRQ(ierr); 225 /* convert global column indices to local ones! */ 226 227 } else { 228 c->columns[i] = 0; 229 } 230 231 if (ctype == IS_COLORING_GLOBAL) { 232 /* Determine nctot, the total (parallel) number of columns of this color */ 233 ierr = MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);CHKERRQ(ierr); 234 ierr = PetscMalloc2(size,PetscMPIInt,&ncolsonproc,size,PetscMPIInt,&disp);CHKERRQ(ierr); 235 236 /* ncolsonproc[j]: local ncolumns on proc[j] of this color */ 237 ierr = PetscMPIIntCast(n,&nn);CHKERRQ(ierr); 238 ierr = MPI_Allgather(&nn,1,MPI_INT,ncolsonproc,1,MPI_INT,PetscObjectComm((PetscObject)mat));CHKERRQ(ierr); 239 nctot = 0; for (j=0; j<size; j++) nctot += ncolsonproc[j]; 240 if (!nctot) { 241 ierr = PetscInfo(mat,"Coloring of matrix has some unneeded colors with no corresponding rows\n");CHKERRQ(ierr); 242 } 243 244 disp[0] = 0; 245 for (j=1; j<size; j++) { 246 disp[j] = disp[j-1] + ncolsonproc[j-1]; 247 } 248 249 /* Get cols, the complete list of columns for this color on each process */ 250 ierr = PetscMalloc((nctot+1)*sizeof(PetscInt),&cols);CHKERRQ(ierr); 251 ierr = MPI_Allgatherv((void*)is,n,MPIU_INT,cols,ncolsonproc,disp,MPIU_INT,PetscObjectComm((PetscObject)mat));CHKERRQ(ierr); 252 ierr = PetscFree2(ncolsonproc,disp);CHKERRQ(ierr); 253 } else if (ctype == IS_COLORING_GHOSTED) { 254 /* Determine local number of columns of this color on this process, including ghost points */ 255 nctot = n; 256 ierr = PetscMalloc((nctot+1)*sizeof(PetscInt),&cols);CHKERRQ(ierr); 257 ierr = PetscMemcpy(cols,is,n*sizeof(PetscInt));CHKERRQ(ierr); 258 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not provided for this MatFDColoring type"); 259 260 /* Mark all rows affect by these columns */ 261 ierr = PetscMemzero(rowhit,m*sizeof(PetscInt));CHKERRQ(ierr); 262 nrows_i = 0; 263 for (j=0; j<nctot; j++) { /* loop over columns*/ 264 if (ctype == IS_COLORING_GHOSTED) { 265 col = ltog[cols[j]]; 266 } else { 267 col = cols[j]; 268 } 269 if (col >= cstart && col < cend) { /* column is in diagonal block of matrix A */ 270 row = A_cj + A_ci[col-cstart]; 271 nrows = A_ci[col-cstart+1] - A_ci[col-cstart]; 272 nrows_i += nrows; 273 /* loop over columns of A marking them in rowhit */ 274 for (k=0; k<nrows; k++) { 275 /* set valaddrhit for part A */ 276 spidx = spidxA[A_ci[col-cstart] + k]; 277 valaddrhit[*row] = &A_val[spidx]; 278 rowhit[*row++] = col - cstart + 1; /* local column index */ 279 } 280 } else { /* column is in off-diagonal block of matrix B */ 281 #if defined(PETSC_USE_CTABLE) 282 ierr = PetscTableFind(aij->colmap,col+1,&colb);CHKERRQ(ierr); 283 colb--; 284 #else 285 colb = aij->colmap[col] - 1; /* local column index */ 286 #endif 287 if (colb == -1) { 288 nrows = 0; 289 } else { 290 row = B_cj + B_ci[colb]; 291 nrows = B_ci[colb+1] - B_ci[colb]; 292 } 293 nrows_i += nrows; 294 /* loop over columns of B marking them in rowhit */ 295 for (k=0; k<nrows; k++) { 296 /* set valaddrhit for part B */ 297 spidx = spidxB[B_ci[colb] + k]; 298 valaddrhit[*row] = &B_val[spidx]; 299 rowhit[*row++] = colb + 1 + cend - cstart; /* local column index */ 300 } 301 } 302 } 303 c->nrows[i] = nrows_i; 304 305 for (j=0; j<m; j++) { 306 if (rowhit[j]) { 307 Jentry[nz].row = j; /* local row index */ 308 Jentry[nz].col = rowhit[j] - 1; /* local column index */ 309 Jentry[nz].valaddr = valaddrhit[j]; /* address of mat value for this entry */ 310 nz++; 311 } 312 } 313 ierr = PetscFree(cols);CHKERRQ(ierr); 314 } 315 ierr = ISColoringRestoreIS(iscoloring,&isa);CHKERRQ(ierr); 316 if (nz != spA->nz + spB->nz) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"nz %d != mat->nz %d\n",nz,spA->nz+spB->nz); 317 318 ierr = PetscFree2(rowhit,valaddrhit);CHKERRQ(ierr); 319 ierr = MatRestoreColumnIJ_SeqAIJ_Color(aij->A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr); 320 ierr = MatRestoreColumnIJ_SeqAIJ_Color(aij->B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr); 321 if (ctype == IS_COLORING_GHOSTED) { 322 ierr = ISLocalToGlobalMappingRestoreIndices(map,<og);CHKERRQ(ierr); 323 } 324 325 mat->ops->fdcoloringapply = MatFDColoringApply_AIJ_new; 326 PetscFunctionReturn(0); 327 } 328 329 /*------------------------------------------------------*/ 330 #undef __FUNCT__ 331 #define __FUNCT__ "MatFDColoringCreate_MPIAIJ" 332 PetscErrorCode MatFDColoringCreate_MPIAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c) 333 { 334 Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data; 335 PetscErrorCode ierr; 336 PetscMPIInt size,*ncolsonproc,*disp,nn; 337 PetscInt i,n,nrows,j,k,m,ncols,col; 338 const PetscInt *is,*A_ci,*A_cj,*B_ci,*B_cj,*rows = 0,*ltog; 339 PetscInt nis = iscoloring->n,nctot,*cols; 340 PetscInt *rowhit,M,cstart,cend,colb; 341 PetscInt *columnsforrow,l; 342 IS *isa; 343 PetscBool done,flg; 344 ISLocalToGlobalMapping map = mat->cmap->mapping; 345 PetscInt ctype=c->ctype; 346 PetscBool new_impl=PETSC_FALSE; 347 348 PetscFunctionBegin; 349 ierr = PetscOptionsName("-new","using new impls","",&new_impl);CHKERRQ(ierr); 350 if (new_impl){ 351 ierr = MatFDColoringCreate_MPIAIJ_new(mat,iscoloring,c);CHKERRQ(ierr); 352 PetscFunctionReturn(0); 353 } 354 if (ctype == IS_COLORING_GHOSTED && !map) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_INCOMP,"When using ghosted differencing matrix must have local to global mapping provided with MatSetLocalToGlobalMapping"); 355 356 if (map) {ierr = ISLocalToGlobalMappingGetIndices(map,<og);CHKERRQ(ierr);} 357 else ltog = NULL; 358 ierr = ISColoringGetIS(iscoloring,PETSC_IGNORE,&isa);CHKERRQ(ierr); 359 360 M = mat->rmap->n; 361 cstart = mat->cmap->rstart; 362 cend = mat->cmap->rend; 363 c->M = mat->rmap->N; /* set the global rows and columns and local rows */ 364 c->N = mat->cmap->N; 365 c->m = mat->rmap->n; 366 c->rstart = mat->rmap->rstart; 367 368 c->ncolors = nis; 369 ierr = PetscMalloc(nis*sizeof(PetscInt),&c->ncolumns);CHKERRQ(ierr); 370 ierr = PetscMalloc(nis*sizeof(PetscInt*),&c->columns);CHKERRQ(ierr); 371 ierr = PetscMalloc(nis*sizeof(PetscInt),&c->nrows);CHKERRQ(ierr); 372 ierr = PetscMalloc(nis*sizeof(PetscInt*),&c->rows);CHKERRQ(ierr); 373 ierr = PetscMalloc(nis*sizeof(PetscInt*),&c->columnsforrow);CHKERRQ(ierr); 374 ierr = PetscLogObjectMemory((PetscObject)c,5*nis*sizeof(PetscInt));CHKERRQ(ierr); 375 376 /* Allow access to data structures of local part of matrix */ 377 if (!aij->colmap) { 378 ierr = MatCreateColmap_MPIAIJ_Private(mat);CHKERRQ(ierr); 379 } 380 ierr = MatGetColumnIJ(aij->A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&done);CHKERRQ(ierr); 381 ierr = MatGetColumnIJ(aij->B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&done);CHKERRQ(ierr); 382 383 ierr = PetscMalloc((M+1)*sizeof(PetscInt),&rowhit);CHKERRQ(ierr); 384 ierr = PetscMalloc((M+1)*sizeof(PetscInt),&columnsforrow);CHKERRQ(ierr); 385 386 for (i=0; i<nis; i++) { 387 ierr = ISGetLocalSize(isa[i],&n);CHKERRQ(ierr); 388 ierr = ISGetIndices(isa[i],&is);CHKERRQ(ierr); 389 390 c->ncolumns[i] = n; /* local number of columns of this color on this process */ 391 if (n) { 392 ierr = PetscMalloc(n*sizeof(PetscInt),&c->columns[i]);CHKERRQ(ierr); 393 ierr = PetscLogObjectMemory((PetscObject)c,n*sizeof(PetscInt));CHKERRQ(ierr); 394 ierr = PetscMemcpy(c->columns[i],is,n*sizeof(PetscInt));CHKERRQ(ierr); 395 } else { 396 c->columns[i] = 0; 397 } 398 399 if (ctype == IS_COLORING_GLOBAL) { 400 /* Determine the total (parallel) number of columns of this color */ 401 ierr = MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);CHKERRQ(ierr); 402 ierr = PetscMalloc2(size,PetscMPIInt,&ncolsonproc,size,PetscMPIInt,&disp);CHKERRQ(ierr); 403 404 ierr = PetscMPIIntCast(n,&nn);CHKERRQ(ierr); 405 ierr = MPI_Allgather(&nn,1,MPI_INT,ncolsonproc,1,MPI_INT,PetscObjectComm((PetscObject)mat));CHKERRQ(ierr); 406 nctot = 0; for (j=0; j<size; j++) nctot += ncolsonproc[j]; 407 if (!nctot) { 408 ierr = PetscInfo(mat,"Coloring of matrix has some unneeded colors with no corresponding rows\n");CHKERRQ(ierr); 409 } 410 411 disp[0] = 0; 412 for (j=1; j<size; j++) { 413 disp[j] = disp[j-1] + ncolsonproc[j-1]; 414 } 415 416 /* Get complete list of columns for color on each processor */ 417 ierr = PetscMalloc((nctot+1)*sizeof(PetscInt),&cols);CHKERRQ(ierr); 418 ierr = MPI_Allgatherv((void*)is,n,MPIU_INT,cols,ncolsonproc,disp,MPIU_INT,PetscObjectComm((PetscObject)mat));CHKERRQ(ierr); 419 ierr = PetscFree2(ncolsonproc,disp);CHKERRQ(ierr); 420 } else if (ctype == IS_COLORING_GHOSTED) { 421 /* Determine local number of columns of this color on this process, including ghost points */ 422 nctot = n; 423 ierr = PetscMalloc((nctot+1)*sizeof(PetscInt),&cols);CHKERRQ(ierr); 424 ierr = PetscMemcpy(cols,is,n*sizeof(PetscInt));CHKERRQ(ierr); 425 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not provided for this MatFDColoring type"); 426 427 /* 428 Mark all rows affect by these columns 429 */ 430 /* Temporary option to allow for debugging/testing */ 431 flg = PETSC_FALSE; 432 ierr = PetscOptionsGetBool(NULL,"-matfdcoloring_slow",&flg,NULL);CHKERRQ(ierr); 433 if (!flg) { /*-----------------------------------------------------------------------------*/ 434 /* crude, fast version */ 435 ierr = PetscMemzero(rowhit,M*sizeof(PetscInt));CHKERRQ(ierr); 436 /* loop over columns*/ 437 for (j=0; j<nctot; j++) { 438 if (ctype == IS_COLORING_GHOSTED) { 439 col = ltog[cols[j]]; 440 } else { 441 col = cols[j]; 442 } 443 if (col >= cstart && col < cend) { 444 /* column is in diagonal block of matrix */ 445 rows = A_cj + A_ci[col-cstart]; 446 m = A_ci[col-cstart+1] - A_ci[col-cstart]; 447 } else { 448 #if defined(PETSC_USE_CTABLE) 449 ierr = PetscTableFind(aij->colmap,col+1,&colb);CHKERRQ(ierr); 450 colb--; 451 #else 452 colb = aij->colmap[col] - 1; 453 #endif 454 if (colb == -1) { 455 m = 0; 456 } else { 457 rows = B_cj + B_ci[colb]; 458 m = B_ci[colb+1] - B_ci[colb]; 459 } 460 } 461 /* loop over columns marking them in rowhit */ 462 for (k=0; k<m; k++) { 463 rowhit[*rows++] = col + 1; 464 } 465 } 466 467 /* count the number of hits */ 468 nrows = 0; 469 for (j=0; j<M; j++) { 470 if (rowhit[j]) nrows++; 471 } 472 c->nrows[i] = nrows; 473 ierr = PetscMalloc((nrows+1)*sizeof(PetscInt),&c->rows[i]);CHKERRQ(ierr); 474 ierr = PetscMalloc((nrows+1)*sizeof(PetscInt),&c->columnsforrow[i]);CHKERRQ(ierr); 475 ierr = PetscLogObjectMemory((PetscObject)c,2*(nrows+1)*sizeof(PetscInt));CHKERRQ(ierr); 476 nrows = 0; 477 for (j=0; j<M; j++) { 478 if (rowhit[j]) { 479 c->rows[i][nrows] = j; /* local row index */ 480 c->columnsforrow[i][nrows] = rowhit[j] - 1; /* global column index */ 481 nrows++; 482 } 483 } 484 } else { /*-------------------------------------------------------------------------------*/ 485 /* slow version, using rowhit as a linked list */ 486 PetscInt currentcol,fm,mfm; 487 rowhit[M] = M; 488 nrows = 0; 489 /* loop over columns*/ 490 for (j=0; j<nctot; j++) { 491 if (ctype == IS_COLORING_GHOSTED) { 492 col = ltog[cols[j]]; 493 } else { 494 col = cols[j]; 495 } 496 if (col >= cstart && col < cend) { 497 /* column is in diagonal block of matrix */ 498 rows = A_cj + A_ci[col-cstart]; 499 m = A_ci[col-cstart+1] - A_ci[col-cstart]; 500 } else { 501 #if defined(PETSC_USE_CTABLE) 502 ierr = PetscTableFind(aij->colmap,col+1,&colb);CHKERRQ(ierr); 503 colb--; 504 #else 505 colb = aij->colmap[col] - 1; 506 #endif 507 if (colb == -1) { 508 m = 0; 509 } else { 510 rows = B_cj + B_ci[colb]; 511 m = B_ci[colb+1] - B_ci[colb]; 512 } 513 } 514 515 /* loop over columns marking them in rowhit */ 516 fm = M; /* fm points to first entry in linked list */ 517 for (k=0; k<m; k++) { 518 currentcol = *rows++; 519 /* is it already in the list? */ 520 do { 521 mfm = fm; 522 fm = rowhit[fm]; 523 } while (fm < currentcol); 524 /* not in list so add it */ 525 if (fm != currentcol) { 526 nrows++; 527 columnsforrow[currentcol] = col; 528 /* next three lines insert new entry into linked list */ 529 rowhit[mfm] = currentcol; 530 rowhit[currentcol] = fm; 531 fm = currentcol; 532 /* fm points to present position in list since we know the columns are sorted */ 533 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Invalid coloring of matrix detected"); 534 } 535 } 536 c->nrows[i] = nrows; 537 538 ierr = PetscMalloc((nrows+1)*sizeof(PetscInt),&c->rows[i]);CHKERRQ(ierr); 539 ierr = PetscMalloc((nrows+1)*sizeof(PetscInt),&c->columnsforrow[i]);CHKERRQ(ierr); 540 ierr = PetscLogObjectMemory((PetscObject)c,(nrows+1)*sizeof(PetscInt));CHKERRQ(ierr); 541 /* now store the linked list of rows into c->rows[i] */ 542 nrows = 0; 543 fm = rowhit[M]; 544 do { 545 c->rows[i][nrows] = fm; 546 c->columnsforrow[i][nrows++] = columnsforrow[fm]; 547 fm = rowhit[fm]; 548 } while (fm < M); 549 } /* ---------------------------------------------------------------------------------------*/ 550 ierr = PetscFree(cols);CHKERRQ(ierr); 551 } 552 553 /* Optimize by adding the vscale, and scaleforrow[][] fields */ 554 /* 555 vscale will contain the "diagonal" on processor scalings followed by the off processor 556 */ 557 if (ctype == IS_COLORING_GLOBAL) { 558 ierr = VecCreateGhost(PetscObjectComm((PetscObject)mat),aij->A->rmap->n,PETSC_DETERMINE,aij->B->cmap->n,aij->garray,&c->vscale);CHKERRQ(ierr); 559 ierr = PetscMalloc(c->ncolors*sizeof(PetscInt*),&c->vscaleforrow);CHKERRQ(ierr); 560 for (k=0; k<c->ncolors; k++) { 561 ierr = PetscMalloc((c->nrows[k]+1)*sizeof(PetscInt),&c->vscaleforrow[k]);CHKERRQ(ierr); 562 for (l=0; l<c->nrows[k]; l++) { 563 col = c->columnsforrow[k][l]; 564 if (col >= cstart && col < cend) { 565 /* column is in diagonal block of matrix */ 566 colb = col - cstart; 567 } else { 568 /* column is in "off-processor" part */ 569 #if defined(PETSC_USE_CTABLE) 570 ierr = PetscTableFind(aij->colmap,col+1,&colb);CHKERRQ(ierr); 571 colb--; 572 #else 573 colb = aij->colmap[col] - 1; 574 #endif 575 colb += cend - cstart; 576 } 577 c->vscaleforrow[k][l] = colb; 578 } 579 } 580 } else if (ctype == IS_COLORING_GHOSTED) { 581 /* Get gtol mapping */ 582 PetscInt N = mat->cmap->N,nlocal,*gtol; 583 ierr = PetscMalloc((N+1)*sizeof(PetscInt),>ol);CHKERRQ(ierr); 584 for (i=0; i<N; i++) gtol[i] = -1; 585 ierr = ISLocalToGlobalMappingGetSize(map,&nlocal);CHKERRQ(ierr); 586 for (i=0; i<nlocal; i++) gtol[ltog[i]] = i; 587 588 c->vscale = 0; /* will be created in MatFDColoringApply() */ 589 ierr = PetscMalloc(c->ncolors*sizeof(PetscInt*),&c->vscaleforrow);CHKERRQ(ierr); 590 for (k=0; k<c->ncolors; k++) { 591 ierr = PetscMalloc((c->nrows[k]+1)*sizeof(PetscInt),&c->vscaleforrow[k]);CHKERRQ(ierr); 592 for (l=0; l<c->nrows[k]; l++) { 593 col = c->columnsforrow[k][l]; /* global column index */ 594 c->vscaleforrow[k][l] = gtol[col]; /* local column index */ 595 } 596 } 597 ierr = PetscFree(gtol);CHKERRQ(ierr); 598 } 599 ierr = ISColoringRestoreIS(iscoloring,&isa);CHKERRQ(ierr); 600 601 ierr = PetscFree(rowhit);CHKERRQ(ierr); 602 ierr = PetscFree(columnsforrow);CHKERRQ(ierr); 603 ierr = MatRestoreColumnIJ(aij->A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&done);CHKERRQ(ierr); 604 ierr = MatRestoreColumnIJ(aij->B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&done);CHKERRQ(ierr); 605 if (map) {ierr = ISLocalToGlobalMappingRestoreIndices(map,<og);CHKERRQ(ierr);} 606 PetscFunctionReturn(0); 607 } 608 609 610 611 612 613 614