1 2 #include <../src/mat/impls/aij/mpi/mpiaij.h> 3 #include <../src/mat/impls/baij/mpi/mpibaij.h> 4 #include <petsc/private/isimpl.h> 5 6 #undef __FUNCT__ 7 #define __FUNCT__ "MatFDColoringApply_BAIJ" 8 PetscErrorCode MatFDColoringApply_BAIJ(Mat J,MatFDColoring coloring,Vec x1,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,spidx,i,j; 13 PetscScalar dx=0.0,*w3_array,*dy_i,*dy=coloring->dy; 14 PetscScalar *vscale_array; 15 const PetscScalar *xx; 16 PetscReal epsilon=coloring->error_rel,umin=coloring->umin,unorm; 17 Vec w1=coloring->w1,w2=coloring->w2,w3,vscale=coloring->vscale; 18 void *fctx=coloring->fctx; 19 PetscInt ctype=coloring->ctype,nxloc,nrows_k; 20 PetscScalar *valaddr; 21 MatEntry *Jentry=coloring->matentry; 22 MatEntry2 *Jentry2=coloring->matentry2; 23 const PetscInt ncolors=coloring->ncolors,*ncolumns=coloring->ncolumns,*nrows=coloring->nrows; 24 PetscInt bs=J->rmap->bs; 25 26 PetscFunctionBegin; 27 /* (1) Set w1 = F(x1) */ 28 if (!coloring->fset) { 29 ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 30 ierr = (*f)(sctx,x1,w1,fctx);CHKERRQ(ierr); 31 ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 32 } else { 33 coloring->fset = PETSC_FALSE; 34 } 35 36 /* (2) Compute vscale = 1./dx - the local scale factors, including ghost points */ 37 ierr = VecGetLocalSize(x1,&nxloc);CHKERRQ(ierr); 38 if (coloring->htype[0] == 'w') { 39 /* vscale = dx is a constant scalar */ 40 ierr = VecNorm(x1,NORM_2,&unorm);CHKERRQ(ierr); 41 dx = 1.0/(PetscSqrtReal(1.0 + unorm)*epsilon); 42 } else { 43 ierr = VecGetArrayRead(x1,&xx);CHKERRQ(ierr); 44 ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr); 45 for (col=0; col<nxloc; col++) { 46 dx = xx[col]; 47 if (PetscAbsScalar(dx) < umin) { 48 if (PetscRealPart(dx) >= 0.0) dx = umin; 49 else if (PetscRealPart(dx) < 0.0 ) dx = -umin; 50 } 51 dx *= epsilon; 52 vscale_array[col] = 1.0/dx; 53 } 54 ierr = VecRestoreArrayRead(x1,&xx);CHKERRQ(ierr); 55 ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr); 56 } 57 if (ctype == IS_COLORING_GLOBAL && coloring->htype[0] == 'd') { 58 ierr = VecGhostUpdateBegin(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 59 ierr = VecGhostUpdateEnd(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 60 } 61 62 /* (3) Loop over each color */ 63 if (!coloring->w3) { 64 ierr = VecDuplicate(x1,&coloring->w3);CHKERRQ(ierr); 65 ierr = PetscLogObjectParent((PetscObject)coloring,(PetscObject)coloring->w3);CHKERRQ(ierr); 66 } 67 w3 = coloring->w3; 68 69 ierr = VecGetOwnershipRange(x1,&cstart,&cend);CHKERRQ(ierr); /* used by ghosted vscale */ 70 if (vscale) { 71 ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr); 72 } 73 nz = 0; 74 for (k=0; k<ncolors; k++) { 75 coloring->currentcolor = k; 76 77 /* 78 (3-1) Loop over each column associated with color 79 adding the perturbation to the vector w3 = x1 + dx. 80 */ 81 ierr = VecCopy(x1,w3);CHKERRQ(ierr); 82 dy_i = dy; 83 for (i=0; i<bs; i++) { /* Loop over a block of columns */ 84 ierr = VecGetArray(w3,&w3_array);CHKERRQ(ierr); 85 if (ctype == IS_COLORING_GLOBAL) w3_array -= cstart; /* shift pointer so global index can be used */ 86 if (coloring->htype[0] == 'w') { 87 for (l=0; l<ncolumns[k]; l++) { 88 col = i + bs*coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */ 89 w3_array[col] += 1.0/dx; 90 if (i) w3_array[col-1] -= 1.0/dx; /* resume original w3[col-1] */ 91 } 92 } else { /* htype == 'ds' */ 93 vscale_array -= cstart; /* shift pointer so global index can be used */ 94 for (l=0; l<ncolumns[k]; l++) { 95 col = i + bs*coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */ 96 w3_array[col] += 1.0/vscale_array[col]; 97 if (i) w3_array[col-1] -= 1.0/vscale_array[col-1]; /* resume original w3[col-1] */ 98 } 99 vscale_array += cstart; 100 } 101 if (ctype == IS_COLORING_GLOBAL) w3_array += cstart; 102 ierr = VecRestoreArray(w3,&w3_array);CHKERRQ(ierr); 103 104 /* 105 (3-2) Evaluate function at w3 = x1 + dx (here dx is a vector of perturbations) 106 w2 = F(x1 + dx) - F(x1) 107 */ 108 ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 109 ierr = VecPlaceArray(w2,dy_i);CHKERRQ(ierr); /* place w2 to the array dy_i */ 110 ierr = (*f)(sctx,w3,w2,fctx);CHKERRQ(ierr); 111 ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 112 ierr = VecAXPY(w2,-1.0,w1);CHKERRQ(ierr); 113 ierr = VecResetArray(w2);CHKERRQ(ierr); 114 dy_i += nxloc; /* points to dy+i*nxloc */ 115 } 116 117 /* 118 (3-3) Loop over rows of vector, putting results into Jacobian matrix 119 */ 120 nrows_k = nrows[k]; 121 if (coloring->htype[0] == 'w') { 122 for (l=0; l<nrows_k; l++) { 123 row = bs*Jentry2[nz].row; /* local row index */ 124 valaddr = Jentry2[nz++].valaddr; 125 spidx = 0; 126 dy_i = dy; 127 for (i=0; i<bs; i++) { /* column of the block */ 128 for (j=0; j<bs; j++) { /* row of the block */ 129 valaddr[spidx++] = dy_i[row+j]*dx; 130 } 131 dy_i += nxloc; /* points to dy+i*nxloc */ 132 } 133 } 134 } else { /* htype == 'ds' */ 135 for (l=0; l<nrows_k; l++) { 136 row = bs*Jentry[nz].row; /* local row index */ 137 col = bs*Jentry[nz].col; /* local column index */ 138 valaddr = Jentry[nz++].valaddr; 139 spidx = 0; 140 dy_i = dy; 141 for (i=0; i<bs; i++) { /* column of the block */ 142 for (j=0; j<bs; j++) { /* row of the block */ 143 valaddr[spidx++] = dy_i[row+j]*vscale_array[col+i]; 144 } 145 dy_i += nxloc; /* points to dy+i*nxloc */ 146 } 147 } 148 } 149 } 150 ierr = MatAssemblyBegin(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 151 ierr = MatAssemblyEnd(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 152 if (vscale) { 153 ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr); 154 } 155 156 coloring->currentcolor = -1; 157 PetscFunctionReturn(0); 158 } 159 160 #include <petscdm.h> 161 162 #undef __FUNCT__ 163 #define __FUNCT__ "MatFDColoringApply_AIJ" 164 PetscErrorCode MatFDColoringApply_AIJ(Mat J,MatFDColoring coloring,Vec x1,void *sctx) 165 { 166 PetscErrorCode (*f)(void*,Vec,Vec,void*) = (PetscErrorCode (*)(void*,Vec,Vec,void*))coloring->f; 167 PetscErrorCode ierr; 168 PetscInt k,cstart,cend,l,row,col,nz; 169 PetscScalar dx=0.0,*y,*w3_array; 170 const PetscScalar *xx; 171 PetscScalar *vscale_array; 172 PetscReal epsilon=coloring->error_rel,umin=coloring->umin,unorm; 173 Vec w1=coloring->w1,w2=coloring->w2,w3,vscale=coloring->vscale; 174 void *fctx=coloring->fctx; 175 ISColoringType ctype=coloring->ctype; 176 PetscInt nxloc,nrows_k; 177 MatEntry *Jentry=coloring->matentry; 178 MatEntry2 *Jentry2=coloring->matentry2; 179 const PetscInt ncolors=coloring->ncolors,*ncolumns=coloring->ncolumns,*nrows=coloring->nrows; 180 181 PetscFunctionBegin; 182 if (ctype == IS_COLORING_LOCAL) { 183 Vec x1local; 184 DM dm; 185 ierr = MatGetDM(J,&dm);CHKERRQ(ierr); 186 if (!dm) SETERRQ(PetscObjectComm((PetscObject)J),PETSC_ERR_ARG_INCOMP,"IS_COLORING_LOCAL requires a DM"); 187 ierr = DMGetLocalVector(dm,&x1local);CHKERRQ(ierr); 188 ierr = DMGlobalToLocalBegin(dm,x1,INSERT_VALUES,x1local);CHKERRQ(ierr); 189 ierr = DMGlobalToLocalEnd(dm,x1,INSERT_VALUES,x1local);CHKERRQ(ierr); 190 x1 = x1local; 191 } 192 193 /* (1) Set w1 = F(x1) */ 194 if (!coloring->fset) { 195 ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 196 ierr = (*f)(sctx,x1,w1,fctx);CHKERRQ(ierr); 197 ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 198 } else { 199 coloring->fset = PETSC_FALSE; 200 } 201 202 /* (2) Compute vscale = 1./dx - the local scale factors, including ghost points */ 203 if (coloring->htype[0] == 'w') { 204 /* vscale = 1./dx is a constant scalar */ 205 ierr = VecNorm(x1,NORM_2,&unorm);CHKERRQ(ierr); 206 dx = 1.0/(PetscSqrtReal(1.0 + unorm)*epsilon); 207 } else { 208 ierr = VecGetLocalSize(x1,&nxloc);CHKERRQ(ierr); 209 ierr = VecGetArrayRead(x1,&xx);CHKERRQ(ierr); 210 ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr); 211 for (col=0; col<nxloc; col++) { 212 dx = xx[col]; 213 if (PetscAbsScalar(dx) < umin) { 214 if (PetscRealPart(dx) >= 0.0) dx = umin; 215 else if (PetscRealPart(dx) < 0.0 ) dx = -umin; 216 } 217 dx *= epsilon; 218 vscale_array[col] = 1.0/dx; 219 } 220 ierr = VecRestoreArrayRead(x1,&xx);CHKERRQ(ierr); 221 ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr); 222 } 223 if (ctype == IS_COLORING_GLOBAL && coloring->htype[0] == 'd') { 224 ierr = VecGhostUpdateBegin(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 225 ierr = VecGhostUpdateEnd(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 226 } 227 228 /* (3) Loop over each color */ 229 if (!coloring->w3) { 230 ierr = VecDuplicate(x1,&coloring->w3);CHKERRQ(ierr); 231 ierr = PetscLogObjectParent((PetscObject)coloring,(PetscObject)coloring->w3);CHKERRQ(ierr); 232 } 233 w3 = coloring->w3; 234 235 ierr = VecGetOwnershipRange(x1,&cstart,&cend);CHKERRQ(ierr); /* used by ghosted vscale */ 236 if (vscale) { 237 ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr); 238 } 239 nz = 0; 240 241 if (coloring->bcols > 1) { /* use blocked insertion of Jentry */ 242 PetscInt i,m=J->rmap->n,nbcols,bcols=coloring->bcols; 243 PetscScalar *dy=coloring->dy,*dy_k; 244 245 nbcols = 0; 246 for (k=0; k<ncolors; k+=bcols) { 247 coloring->currentcolor = k; 248 249 /* 250 (3-1) Loop over each column associated with color 251 adding the perturbation to the vector w3 = x1 + dx. 252 */ 253 254 dy_k = dy; 255 if (k + bcols > ncolors) bcols = ncolors - k; 256 for (i=0; i<bcols; i++) { 257 258 ierr = VecCopy(x1,w3);CHKERRQ(ierr); 259 ierr = VecGetArray(w3,&w3_array);CHKERRQ(ierr); 260 if (ctype == IS_COLORING_GLOBAL) w3_array -= cstart; /* shift pointer so global index can be used */ 261 if (coloring->htype[0] == 'w') { 262 for (l=0; l<ncolumns[k+i]; l++) { 263 col = coloring->columns[k+i][l]; /* local column (in global index!) of the matrix we are probing for */ 264 w3_array[col] += 1.0/dx; 265 } 266 } else { /* htype == 'ds' */ 267 vscale_array -= cstart; /* shift pointer so global index can be used */ 268 for (l=0; l<ncolumns[k+i]; l++) { 269 col = coloring->columns[k+i][l]; /* local column (in global index!) of the matrix we are probing for */ 270 w3_array[col] += 1.0/vscale_array[col]; 271 } 272 vscale_array += cstart; 273 } 274 if (ctype == IS_COLORING_GLOBAL) w3_array += cstart; 275 ierr = VecRestoreArray(w3,&w3_array);CHKERRQ(ierr); 276 277 /* 278 (3-2) Evaluate function at w3 = x1 + dx (here dx is a vector of perturbations) 279 w2 = F(x1 + dx) - F(x1) 280 */ 281 ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 282 ierr = VecPlaceArray(w2,dy_k);CHKERRQ(ierr); /* place w2 to the array dy_i */ 283 ierr = (*f)(sctx,w3,w2,fctx);CHKERRQ(ierr); 284 ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 285 ierr = VecAXPY(w2,-1.0,w1);CHKERRQ(ierr); 286 ierr = VecResetArray(w2);CHKERRQ(ierr); 287 dy_k += m; /* points to dy+i*nxloc */ 288 } 289 290 /* 291 (3-3) Loop over block rows of vector, putting results into Jacobian matrix 292 */ 293 nrows_k = nrows[nbcols++]; 294 ierr = VecGetArray(w2,&y);CHKERRQ(ierr); 295 296 if (coloring->htype[0] == 'w') { 297 for (l=0; l<nrows_k; l++) { 298 row = Jentry2[nz].row; /* local row index */ 299 *(Jentry2[nz++].valaddr) = dy[row]*dx; 300 } 301 } else { /* htype == 'ds' */ 302 for (l=0; l<nrows_k; l++) { 303 row = Jentry[nz].row; /* local row index */ 304 *(Jentry[nz].valaddr) = dy[row]*vscale_array[Jentry[nz].col]; 305 nz++; 306 } 307 } 308 ierr = VecRestoreArray(w2,&y);CHKERRQ(ierr); 309 } 310 } else { /* bcols == 1 */ 311 for (k=0; k<ncolors; k++) { 312 coloring->currentcolor = k; 313 314 /* 315 (3-1) Loop over each column associated with color 316 adding the perturbation to the vector w3 = x1 + dx. 317 */ 318 ierr = VecCopy(x1,w3);CHKERRQ(ierr); 319 ierr = VecGetArray(w3,&w3_array);CHKERRQ(ierr); 320 if (ctype == IS_COLORING_GLOBAL) w3_array -= cstart; /* shift pointer so global index can be used */ 321 if (coloring->htype[0] == 'w') { 322 for (l=0; l<ncolumns[k]; l++) { 323 col = coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */ 324 w3_array[col] += 1.0/dx; 325 } 326 } else { /* htype == 'ds' */ 327 vscale_array -= cstart; /* shift pointer so global index can be used */ 328 for (l=0; l<ncolumns[k]; l++) { 329 col = coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */ 330 w3_array[col] += 1.0/vscale_array[col]; 331 } 332 vscale_array += cstart; 333 } 334 if (ctype == IS_COLORING_GLOBAL) w3_array += cstart; 335 ierr = VecRestoreArray(w3,&w3_array);CHKERRQ(ierr); 336 337 /* 338 (3-2) Evaluate function at w3 = x1 + dx (here dx is a vector of perturbations) 339 w2 = F(x1 + dx) - F(x1) 340 */ 341 ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 342 ierr = (*f)(sctx,w3,w2,fctx);CHKERRQ(ierr); 343 ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 344 ierr = VecAXPY(w2,-1.0,w1);CHKERRQ(ierr); 345 346 /* 347 (3-3) Loop over rows of vector, putting results into Jacobian matrix 348 */ 349 nrows_k = nrows[k]; 350 ierr = VecGetArray(w2,&y);CHKERRQ(ierr); 351 if (coloring->htype[0] == 'w') { 352 for (l=0; l<nrows_k; l++) { 353 row = Jentry2[nz].row; /* local row index */ 354 *(Jentry2[nz++].valaddr) = y[row]*dx; 355 } 356 } else { /* htype == 'ds' */ 357 for (l=0; l<nrows_k; l++) { 358 row = Jentry[nz].row; /* local row index */ 359 *(Jentry[nz].valaddr) = y[row]*vscale_array[Jentry[nz].col]; 360 nz++; 361 } 362 } 363 ierr = VecRestoreArray(w2,&y);CHKERRQ(ierr); 364 } 365 } 366 367 ierr = MatAssemblyBegin(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 368 ierr = MatAssemblyEnd(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 369 if (vscale) { 370 ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr); 371 } 372 coloring->currentcolor = -1; 373 if (ctype == IS_COLORING_LOCAL) { 374 DM dm; 375 ierr = MatGetDM(J,&dm);CHKERRQ(ierr); 376 ierr = DMRestoreLocalVector(dm,&x1);CHKERRQ(ierr); 377 } 378 PetscFunctionReturn(0); 379 } 380 381 #undef __FUNCT__ 382 #define __FUNCT__ "MatFDColoringSetUp_MPIXAIJ" 383 PetscErrorCode MatFDColoringSetUp_MPIXAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c) 384 { 385 PetscErrorCode ierr; 386 PetscMPIInt size,*ncolsonproc,*disp,nn; 387 PetscInt i,n,nrows,nrows_i,j,k,m,ncols,col,*rowhit,cstart,cend,colb; 388 const PetscInt *is,*A_ci,*A_cj,*B_ci,*B_cj,*row=NULL,*ltog=NULL; 389 PetscInt nis=iscoloring->n,nctot,*cols; 390 IS *isa; 391 ISLocalToGlobalMapping map=mat->cmap->mapping; 392 PetscInt ctype=c->ctype,*spidxA,*spidxB,nz,bs,bs2,spidx; 393 Mat A,B; 394 PetscScalar *A_val,*B_val,**valaddrhit; 395 MatEntry *Jentry; 396 MatEntry2 *Jentry2; 397 PetscBool isBAIJ; 398 PetscInt bcols=c->bcols; 399 #if defined(PETSC_USE_CTABLE) 400 PetscTable colmap=NULL; 401 #else 402 PetscInt *colmap=NULL; /* local col number of off-diag col */ 403 #endif 404 405 PetscFunctionBegin; 406 if (ctype == IS_COLORING_LOCAL) { 407 if (!map) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_INCOMP,"When using ghosted differencing matrix must have local to global mapping provided with MatSetLocalToGlobalMapping"); 408 ierr = ISLocalToGlobalMappingGetIndices(map,<og);CHKERRQ(ierr); 409 } 410 411 ierr = MatGetBlockSize(mat,&bs);CHKERRQ(ierr); 412 ierr = PetscObjectTypeCompare((PetscObject)mat,MATMPIBAIJ,&isBAIJ);CHKERRQ(ierr); 413 if (isBAIJ) { 414 Mat_MPIBAIJ *baij=(Mat_MPIBAIJ*)mat->data; 415 Mat_SeqBAIJ *spA,*spB; 416 A = baij->A; spA = (Mat_SeqBAIJ*)A->data; A_val = spA->a; 417 B = baij->B; spB = (Mat_SeqBAIJ*)B->data; B_val = spB->a; 418 nz = spA->nz + spB->nz; /* total nonzero entries of mat */ 419 if (!baij->colmap) { 420 ierr = MatCreateColmap_MPIBAIJ_Private(mat);CHKERRQ(ierr); 421 } 422 colmap = baij->colmap; 423 ierr = MatGetColumnIJ_SeqBAIJ_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr); 424 ierr = MatGetColumnIJ_SeqBAIJ_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr); 425 426 if (ctype == IS_COLORING_GLOBAL && c->htype[0] == 'd') { /* create vscale for storing dx */ 427 PetscInt *garray; 428 ierr = PetscMalloc1(B->cmap->n,&garray);CHKERRQ(ierr); 429 for (i=0; i<baij->B->cmap->n/bs; i++) { 430 for (j=0; j<bs; j++) { 431 garray[i*bs+j] = bs*baij->garray[i]+j; 432 } 433 } 434 ierr = VecCreateGhost(PetscObjectComm((PetscObject)mat),mat->cmap->n,PETSC_DETERMINE,B->cmap->n,garray,&c->vscale);CHKERRQ(ierr); 435 ierr = PetscFree(garray);CHKERRQ(ierr); 436 } 437 } else { 438 Mat_MPIAIJ *aij=(Mat_MPIAIJ*)mat->data; 439 Mat_SeqAIJ *spA,*spB; 440 A = aij->A; spA = (Mat_SeqAIJ*)A->data; A_val = spA->a; 441 B = aij->B; spB = (Mat_SeqAIJ*)B->data; B_val = spB->a; 442 nz = spA->nz + spB->nz; /* total nonzero entries of mat */ 443 if (!aij->colmap) { 444 /* Allow access to data structures of local part of matrix 445 - creates aij->colmap which maps global column number to local number in part B */ 446 ierr = MatCreateColmap_MPIAIJ_Private(mat);CHKERRQ(ierr); 447 } 448 colmap = aij->colmap; 449 ierr = MatGetColumnIJ_SeqAIJ_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr); 450 ierr = MatGetColumnIJ_SeqAIJ_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr); 451 452 bs = 1; /* only bs=1 is supported for non MPIBAIJ matrix */ 453 454 if (ctype == IS_COLORING_GLOBAL && c->htype[0] == 'd') { /* create vscale for storing dx */ 455 ierr = VecCreateGhost(PetscObjectComm((PetscObject)mat),mat->cmap->n,PETSC_DETERMINE,B->cmap->n,aij->garray,&c->vscale);CHKERRQ(ierr); 456 } 457 } 458 459 m = mat->rmap->n/bs; 460 cstart = mat->cmap->rstart/bs; 461 cend = mat->cmap->rend/bs; 462 463 ierr = PetscMalloc1(nis,&c->ncolumns);CHKERRQ(ierr); 464 ierr = PetscMalloc1(nis,&c->columns);CHKERRQ(ierr); 465 ierr = PetscCalloc1(nis,&c->nrows);CHKERRQ(ierr); 466 ierr = PetscLogObjectMemory((PetscObject)c,3*nis*sizeof(PetscInt));CHKERRQ(ierr); 467 468 if (c->htype[0] == 'd') { 469 ierr = PetscMalloc1(nz,&Jentry);CHKERRQ(ierr); 470 ierr = PetscLogObjectMemory((PetscObject)c,nz*sizeof(MatEntry));CHKERRQ(ierr); 471 c->matentry = Jentry; 472 } else if (c->htype[0] == 'w') { 473 ierr = PetscMalloc1(nz,&Jentry2);CHKERRQ(ierr); 474 ierr = PetscLogObjectMemory((PetscObject)c,nz*sizeof(MatEntry2));CHKERRQ(ierr); 475 c->matentry2 = Jentry2; 476 } else SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"htype is not supported"); 477 478 ierr = PetscMalloc2(m+1,&rowhit,m+1,&valaddrhit);CHKERRQ(ierr); 479 nz = 0; 480 ierr = ISColoringGetIS(iscoloring,PETSC_IGNORE,&isa);CHKERRQ(ierr); 481 for (i=0; i<nis; i++) { /* for each local color */ 482 ierr = ISGetLocalSize(isa[i],&n);CHKERRQ(ierr); 483 ierr = ISGetIndices(isa[i],&is);CHKERRQ(ierr); 484 485 c->ncolumns[i] = n; /* local number of columns of this color on this process */ 486 if (n) { 487 ierr = PetscMalloc1(n,&c->columns[i]);CHKERRQ(ierr); 488 ierr = PetscLogObjectMemory((PetscObject)c,n*sizeof(PetscInt));CHKERRQ(ierr); 489 ierr = PetscMemcpy(c->columns[i],is,n*sizeof(PetscInt));CHKERRQ(ierr); 490 } else { 491 c->columns[i] = 0; 492 } 493 494 if (ctype == IS_COLORING_GLOBAL) { 495 /* Determine nctot, the total (parallel) number of columns of this color */ 496 ierr = MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);CHKERRQ(ierr); 497 ierr = PetscMalloc2(size,&ncolsonproc,size,&disp);CHKERRQ(ierr); 498 499 /* ncolsonproc[j]: local ncolumns on proc[j] of this color */ 500 ierr = PetscMPIIntCast(n,&nn);CHKERRQ(ierr); 501 ierr = MPI_Allgather(&nn,1,MPI_INT,ncolsonproc,1,MPI_INT,PetscObjectComm((PetscObject)mat));CHKERRQ(ierr); 502 nctot = 0; for (j=0; j<size; j++) nctot += ncolsonproc[j]; 503 if (!nctot) { 504 ierr = PetscInfo(mat,"Coloring of matrix has some unneeded colors with no corresponding rows\n");CHKERRQ(ierr); 505 } 506 507 disp[0] = 0; 508 for (j=1; j<size; j++) { 509 disp[j] = disp[j-1] + ncolsonproc[j-1]; 510 } 511 512 /* Get cols, the complete list of columns for this color on each process */ 513 ierr = PetscMalloc1(nctot+1,&cols);CHKERRQ(ierr); 514 ierr = MPI_Allgatherv((void*)is,n,MPIU_INT,cols,ncolsonproc,disp,MPIU_INT,PetscObjectComm((PetscObject)mat));CHKERRQ(ierr); 515 ierr = PetscFree2(ncolsonproc,disp);CHKERRQ(ierr); 516 } else if (ctype == IS_COLORING_LOCAL) { 517 /* Determine local number of columns of this color on this process, including ghost points */ 518 nctot = n; 519 ierr = PetscMalloc1(nctot+1,&cols);CHKERRQ(ierr); 520 ierr = PetscMemcpy(cols,is,n*sizeof(PetscInt));CHKERRQ(ierr); 521 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not provided for this MatFDColoring type"); 522 523 /* Mark all rows affect by these columns */ 524 ierr = PetscMemzero(rowhit,m*sizeof(PetscInt));CHKERRQ(ierr); 525 bs2 = bs*bs; 526 nrows_i = 0; 527 for (j=0; j<nctot; j++) { /* loop over columns*/ 528 if (ctype == IS_COLORING_LOCAL) { 529 col = ltog[cols[j]]; 530 } else { 531 col = cols[j]; 532 } 533 if (col >= cstart && col < cend) { /* column is in A, diagonal block of mat */ 534 row = A_cj + A_ci[col-cstart]; 535 nrows = A_ci[col-cstart+1] - A_ci[col-cstart]; 536 nrows_i += nrows; 537 /* loop over columns of A marking them in rowhit */ 538 for (k=0; k<nrows; k++) { 539 /* set valaddrhit for part A */ 540 spidx = bs2*spidxA[A_ci[col-cstart] + k]; 541 valaddrhit[*row] = &A_val[spidx]; 542 rowhit[*row++] = col - cstart + 1; /* local column index */ 543 } 544 } else { /* column is in B, off-diagonal block of mat */ 545 #if defined(PETSC_USE_CTABLE) 546 ierr = PetscTableFind(colmap,col+1,&colb);CHKERRQ(ierr); 547 colb--; 548 #else 549 colb = colmap[col] - 1; /* local column index */ 550 #endif 551 if (colb == -1) { 552 nrows = 0; 553 } else { 554 colb = colb/bs; 555 row = B_cj + B_ci[colb]; 556 nrows = B_ci[colb+1] - B_ci[colb]; 557 } 558 nrows_i += nrows; 559 /* loop over columns of B marking them in rowhit */ 560 for (k=0; k<nrows; k++) { 561 /* set valaddrhit for part B */ 562 spidx = bs2*spidxB[B_ci[colb] + k]; 563 valaddrhit[*row] = &B_val[spidx]; 564 rowhit[*row++] = colb + 1 + cend - cstart; /* local column index */ 565 } 566 } 567 } 568 c->nrows[i] = nrows_i; 569 570 if (c->htype[0] == 'd') { 571 for (j=0; j<m; j++) { 572 if (rowhit[j]) { 573 Jentry[nz].row = j; /* local row index */ 574 Jentry[nz].col = rowhit[j] - 1; /* local column index */ 575 Jentry[nz].valaddr = valaddrhit[j]; /* address of mat value for this entry */ 576 nz++; 577 } 578 } 579 } else { /* c->htype == 'wp' */ 580 for (j=0; j<m; j++) { 581 if (rowhit[j]) { 582 Jentry2[nz].row = j; /* local row index */ 583 Jentry2[nz].valaddr = valaddrhit[j]; /* address of mat value for this entry */ 584 nz++; 585 } 586 } 587 } 588 ierr = PetscFree(cols);CHKERRQ(ierr); 589 } 590 591 if (bcols > 1) { /* reorder Jentry for faster MatFDColoringApply() */ 592 ierr = MatFDColoringSetUpBlocked_AIJ_Private(mat,c,nz);CHKERRQ(ierr); 593 } 594 595 if (isBAIJ) { 596 ierr = MatRestoreColumnIJ_SeqBAIJ_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr); 597 ierr = MatRestoreColumnIJ_SeqBAIJ_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr); 598 ierr = PetscMalloc1(bs*mat->rmap->n,&c->dy);CHKERRQ(ierr); 599 } else { 600 ierr = MatRestoreColumnIJ_SeqAIJ_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr); 601 ierr = MatRestoreColumnIJ_SeqAIJ_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr); 602 } 603 604 ierr = ISColoringRestoreIS(iscoloring,&isa);CHKERRQ(ierr); 605 ierr = PetscFree2(rowhit,valaddrhit);CHKERRQ(ierr); 606 607 if (ctype == IS_COLORING_LOCAL) { 608 ierr = ISLocalToGlobalMappingRestoreIndices(map,<og);CHKERRQ(ierr); 609 } 610 ierr = PetscInfo3(c,"ncolors %D, brows %D and bcols %D are used.\n",c->ncolors,c->brows,c->bcols);CHKERRQ(ierr); 611 PetscFunctionReturn(0); 612 } 613 614 #undef __FUNCT__ 615 #define __FUNCT__ "MatFDColoringCreate_MPIXAIJ" 616 PetscErrorCode MatFDColoringCreate_MPIXAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c) 617 { 618 PetscErrorCode ierr; 619 PetscInt bs,nis=iscoloring->n,m=mat->rmap->n; 620 PetscBool isBAIJ; 621 622 PetscFunctionBegin; 623 /* set default brows and bcols for speedup inserting the dense matrix into sparse Jacobian; 624 bcols is chosen s.t. dy-array takes 50% of memory space as mat */ 625 ierr = MatGetBlockSize(mat,&bs);CHKERRQ(ierr); 626 ierr = PetscObjectTypeCompare((PetscObject)mat,MATMPIBAIJ,&isBAIJ);CHKERRQ(ierr); 627 if (isBAIJ || m == 0) { 628 c->brows = m; 629 c->bcols = 1; 630 } else { /* mpiaij matrix */ 631 /* bcols is chosen s.t. dy-array takes 50% of local memory space as mat */ 632 Mat_MPIAIJ *aij=(Mat_MPIAIJ*)mat->data; 633 Mat_SeqAIJ *spA,*spB; 634 Mat A,B; 635 PetscInt nz,brows,bcols; 636 PetscReal mem; 637 638 bs = 1; /* only bs=1 is supported for MPIAIJ matrix */ 639 640 A = aij->A; spA = (Mat_SeqAIJ*)A->data; 641 B = aij->B; spB = (Mat_SeqAIJ*)B->data; 642 nz = spA->nz + spB->nz; /* total local nonzero entries of mat */ 643 mem = nz*(sizeof(PetscScalar) + sizeof(PetscInt)) + 3*m*sizeof(PetscInt); 644 bcols = (PetscInt)(0.5*mem /(m*sizeof(PetscScalar))); 645 brows = 1000/bcols; 646 if (bcols > nis) bcols = nis; 647 if (brows == 0 || brows > m) brows = m; 648 c->brows = brows; 649 c->bcols = bcols; 650 } 651 652 c->M = mat->rmap->N/bs; /* set the global rows and columns and local rows */ 653 c->N = mat->cmap->N/bs; 654 c->m = mat->rmap->n/bs; 655 c->rstart = mat->rmap->rstart/bs; 656 c->ncolors = nis; 657 PetscFunctionReturn(0); 658 } 659