1 #include <../src/mat/impls/sell/mpi/mpisell.h> 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 PetscErrorCode MatFDColoringApply_BAIJ(Mat J,MatFDColoring coloring,Vec x1,void *sctx) 7 { 8 PetscErrorCode (*f)(void*,Vec,Vec,void*)=(PetscErrorCode (*)(void*,Vec,Vec,void*))coloring->f; 9 PetscErrorCode ierr; 10 PetscInt k,cstart,cend,l,row,col,nz,spidx,i,j; 11 PetscScalar dx=0.0,*w3_array,*dy_i,*dy=coloring->dy; 12 PetscScalar *vscale_array; 13 const PetscScalar *xx; 14 PetscReal epsilon=coloring->error_rel,umin=coloring->umin,unorm; 15 Vec w1=coloring->w1,w2=coloring->w2,w3,vscale=coloring->vscale; 16 void *fctx=coloring->fctx; 17 PetscInt ctype=coloring->ctype,nxloc,nrows_k; 18 PetscScalar *valaddr; 19 MatEntry *Jentry=coloring->matentry; 20 MatEntry2 *Jentry2=coloring->matentry2; 21 const PetscInt ncolors=coloring->ncolors,*ncolumns=coloring->ncolumns,*nrows=coloring->nrows; 22 PetscInt bs=J->rmap->bs; 23 24 PetscFunctionBegin; 25 ierr = VecBindToCPU(x1,PETSC_TRUE);CHKERRQ(ierr); 26 /* (1) Set w1 = F(x1) */ 27 if (!coloring->fset) { 28 ierr = PetscLogEventBegin(MAT_FDColoringFunction,coloring,0,0,0);CHKERRQ(ierr); 29 ierr = (*f)(sctx,x1,w1,fctx);CHKERRQ(ierr); 30 ierr = PetscLogEventEnd(MAT_FDColoringFunction,coloring,0,0,0);CHKERRQ(ierr); 31 } else { 32 coloring->fset = PETSC_FALSE; 33 } 34 35 /* (2) Compute vscale = 1./dx - the local scale factors, including ghost points */ 36 ierr = VecGetLocalSize(x1,&nxloc);CHKERRQ(ierr); 37 if (coloring->htype[0] == 'w') { 38 /* vscale = dx is a constant scalar */ 39 ierr = VecNorm(x1,NORM_2,&unorm);CHKERRQ(ierr); 40 dx = 1.0/(PetscSqrtReal(1.0 + unorm)*epsilon); 41 } else { 42 ierr = VecGetArrayRead(x1,&xx);CHKERRQ(ierr); 43 ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr); 44 for (col=0; col<nxloc; col++) { 45 dx = xx[col]; 46 if (PetscAbsScalar(dx) < umin) { 47 if (PetscRealPart(dx) >= 0.0) dx = umin; 48 else if (PetscRealPart(dx) < 0.0) dx = -umin; 49 } 50 dx *= epsilon; 51 vscale_array[col] = 1.0/dx; 52 } 53 ierr = VecRestoreArrayRead(x1,&xx);CHKERRQ(ierr); 54 ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr); 55 } 56 if (ctype == IS_COLORING_GLOBAL && coloring->htype[0] == 'd') { 57 ierr = VecGhostUpdateBegin(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 58 ierr = VecGhostUpdateEnd(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 59 } 60 61 /* (3) Loop over each color */ 62 if (!coloring->w3) { 63 ierr = VecDuplicate(x1,&coloring->w3);CHKERRQ(ierr); 64 /* Vec is used intensively in particular piece of scalar CPU code; won't benefit from bouncing back and forth to the GPU */ 65 ierr = VecBindToCPU(coloring->w3,PETSC_TRUE);CHKERRQ(ierr); 66 ierr = PetscLogObjectParent((PetscObject)coloring,(PetscObject)coloring->w3);CHKERRQ(ierr); 67 } 68 w3 = coloring->w3; 69 70 ierr = VecGetOwnershipRange(x1,&cstart,&cend);CHKERRQ(ierr); /* used by ghosted vscale */ 71 if (vscale) { 72 ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr); 73 } 74 nz = 0; 75 for (k=0; k<ncolors; k++) { 76 coloring->currentcolor = k; 77 78 /* 79 (3-1) Loop over each column associated with color 80 adding the perturbation to the vector w3 = x1 + dx. 81 */ 82 ierr = VecCopy(x1,w3);CHKERRQ(ierr); 83 dy_i = dy; 84 for (i=0; i<bs; i++) { /* Loop over a block of columns */ 85 ierr = VecGetArray(w3,&w3_array);CHKERRQ(ierr); 86 if (ctype == IS_COLORING_GLOBAL) w3_array -= cstart; /* shift pointer so global index can be used */ 87 if (coloring->htype[0] == 'w') { 88 for (l=0; l<ncolumns[k]; l++) { 89 col = i + bs*coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */ 90 w3_array[col] += 1.0/dx; 91 if (i) w3_array[col-1] -= 1.0/dx; /* resume original w3[col-1] */ 92 } 93 } else { /* htype == 'ds' */ 94 vscale_array -= cstart; /* shift pointer so global index can be used */ 95 for (l=0; l<ncolumns[k]; l++) { 96 col = i + bs*coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */ 97 w3_array[col] += 1.0/vscale_array[col]; 98 if (i) w3_array[col-1] -= 1.0/vscale_array[col-1]; /* resume original w3[col-1] */ 99 } 100 vscale_array += cstart; 101 } 102 if (ctype == IS_COLORING_GLOBAL) w3_array += cstart; 103 ierr = VecRestoreArray(w3,&w3_array);CHKERRQ(ierr); 104 105 /* 106 (3-2) Evaluate function at w3 = x1 + dx (here dx is a vector of perturbations) 107 w2 = F(x1 + dx) - F(x1) 108 */ 109 ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 110 ierr = VecPlaceArray(w2,dy_i);CHKERRQ(ierr); /* place w2 to the array dy_i */ 111 ierr = (*f)(sctx,w3,w2,fctx);CHKERRQ(ierr); 112 ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 113 ierr = VecAXPY(w2,-1.0,w1);CHKERRQ(ierr); 114 ierr = VecResetArray(w2);CHKERRQ(ierr); 115 dy_i += nxloc; /* points to dy+i*nxloc */ 116 } 117 118 /* 119 (3-3) Loop over rows of vector, putting results into Jacobian matrix 120 */ 121 nrows_k = nrows[k]; 122 if (coloring->htype[0] == 'w') { 123 for (l=0; l<nrows_k; l++) { 124 row = bs*Jentry2[nz].row; /* local row index */ 125 valaddr = Jentry2[nz++].valaddr; 126 spidx = 0; 127 dy_i = dy; 128 for (i=0; i<bs; i++) { /* column of the block */ 129 for (j=0; j<bs; j++) { /* row of the block */ 130 valaddr[spidx++] = dy_i[row+j]*dx; 131 } 132 dy_i += nxloc; /* points to dy+i*nxloc */ 133 } 134 } 135 } else { /* htype == 'ds' */ 136 for (l=0; l<nrows_k; l++) { 137 row = bs*Jentry[nz].row; /* local row index */ 138 col = bs*Jentry[nz].col; /* local column index */ 139 valaddr = Jentry[nz++].valaddr; 140 spidx = 0; 141 dy_i = dy; 142 for (i=0; i<bs; i++) { /* column of the block */ 143 for (j=0; j<bs; j++) { /* row of the block */ 144 valaddr[spidx++] = dy_i[row+j]*vscale_array[col+i]; 145 } 146 dy_i += nxloc; /* points to dy+i*nxloc */ 147 } 148 } 149 } 150 } 151 ierr = MatAssemblyBegin(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 152 ierr = MatAssemblyEnd(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 153 if (vscale) { 154 ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr); 155 } 156 157 coloring->currentcolor = -1; 158 ierr = VecBindToCPU(x1,PETSC_FALSE);CHKERRQ(ierr); 159 PetscFunctionReturn(0); 160 } 161 162 /* this is declared PETSC_EXTERN because it is used by MatFDColoringUseDM() which is in the DM library */ 163 PetscErrorCode MatFDColoringApply_AIJ(Mat J,MatFDColoring coloring,Vec x1,void *sctx) 164 { 165 PetscErrorCode (*f)(void*,Vec,Vec,void*) = (PetscErrorCode (*)(void*,Vec,Vec,void*))coloring->f; 166 PetscErrorCode ierr; 167 PetscInt k,cstart,cend,l,row,col,nz; 168 PetscScalar dx=0.0,*y,*w3_array; 169 const PetscScalar *xx; 170 PetscScalar *vscale_array; 171 PetscReal epsilon=coloring->error_rel,umin=coloring->umin,unorm; 172 Vec w1=coloring->w1,w2=coloring->w2,w3,vscale=coloring->vscale; 173 void *fctx=coloring->fctx; 174 ISColoringType ctype=coloring->ctype; 175 PetscInt nxloc,nrows_k; 176 MatEntry *Jentry=coloring->matentry; 177 MatEntry2 *Jentry2=coloring->matentry2; 178 const PetscInt ncolors=coloring->ncolors,*ncolumns=coloring->ncolumns,*nrows=coloring->nrows; 179 PetscBool alreadyboundtocpu; 180 181 PetscFunctionBegin; 182 ierr = VecBoundToCPU(x1,&alreadyboundtocpu);CHKERRQ(ierr); 183 ierr = VecBindToCPU(x1,PETSC_TRUE);CHKERRQ(ierr); 184 if ((ctype == IS_COLORING_LOCAL) && (J->ops->fdcoloringapply == MatFDColoringApply_AIJ)) SETERRQ(PetscObjectComm((PetscObject)J),PETSC_ERR_SUP,"Must call MatColoringUseDM() with IS_COLORING_LOCAL"); 185 /* (1) Set w1 = F(x1) */ 186 if (!coloring->fset) { 187 ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 188 ierr = (*f)(sctx,x1,w1,fctx);CHKERRQ(ierr); 189 ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 190 } else { 191 coloring->fset = PETSC_FALSE; 192 } 193 194 /* (2) Compute vscale = 1./dx - the local scale factors, including ghost points */ 195 if (coloring->htype[0] == 'w') { 196 /* vscale = 1./dx is a constant scalar */ 197 ierr = VecNorm(x1,NORM_2,&unorm);CHKERRQ(ierr); 198 dx = 1.0/(PetscSqrtReal(1.0 + unorm)*epsilon); 199 } else { 200 ierr = VecGetLocalSize(x1,&nxloc);CHKERRQ(ierr); 201 ierr = VecGetArrayRead(x1,&xx);CHKERRQ(ierr); 202 ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr); 203 for (col=0; col<nxloc; col++) { 204 dx = xx[col]; 205 if (PetscAbsScalar(dx) < umin) { 206 if (PetscRealPart(dx) >= 0.0) dx = umin; 207 else if (PetscRealPart(dx) < 0.0) dx = -umin; 208 } 209 dx *= epsilon; 210 vscale_array[col] = 1.0/dx; 211 } 212 ierr = VecRestoreArrayRead(x1,&xx);CHKERRQ(ierr); 213 ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr); 214 } 215 if (ctype == IS_COLORING_GLOBAL && coloring->htype[0] == 'd') { 216 ierr = VecGhostUpdateBegin(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 217 ierr = VecGhostUpdateEnd(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 218 } 219 220 /* (3) Loop over each color */ 221 if (!coloring->w3) { 222 ierr = VecDuplicate(x1,&coloring->w3);CHKERRQ(ierr); 223 ierr = PetscLogObjectParent((PetscObject)coloring,(PetscObject)coloring->w3);CHKERRQ(ierr); 224 } 225 w3 = coloring->w3; 226 227 ierr = VecGetOwnershipRange(x1,&cstart,&cend);CHKERRQ(ierr); /* used by ghosted vscale */ 228 if (vscale) { 229 ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr); 230 } 231 nz = 0; 232 233 if (coloring->bcols > 1) { /* use blocked insertion of Jentry */ 234 PetscInt i,m=J->rmap->n,nbcols,bcols=coloring->bcols; 235 PetscScalar *dy=coloring->dy,*dy_k; 236 237 nbcols = 0; 238 for (k=0; k<ncolors; k+=bcols) { 239 240 /* 241 (3-1) Loop over each column associated with color 242 adding the perturbation to the vector w3 = x1 + dx. 243 */ 244 245 dy_k = dy; 246 if (k + bcols > ncolors) bcols = ncolors - k; 247 for (i=0; i<bcols; i++) { 248 coloring->currentcolor = k+i; 249 250 ierr = VecCopy(x1,w3);CHKERRQ(ierr); 251 ierr = VecGetArray(w3,&w3_array);CHKERRQ(ierr); 252 if (ctype == IS_COLORING_GLOBAL) w3_array -= cstart; /* shift pointer so global index can be used */ 253 if (coloring->htype[0] == 'w') { 254 for (l=0; l<ncolumns[k+i]; l++) { 255 col = coloring->columns[k+i][l]; /* local column (in global index!) of the matrix we are probing for */ 256 w3_array[col] += 1.0/dx; 257 } 258 } else { /* htype == 'ds' */ 259 vscale_array -= cstart; /* shift pointer so global index can be used */ 260 for (l=0; l<ncolumns[k+i]; l++) { 261 col = coloring->columns[k+i][l]; /* local column (in global index!) of the matrix we are probing for */ 262 w3_array[col] += 1.0/vscale_array[col]; 263 } 264 vscale_array += cstart; 265 } 266 if (ctype == IS_COLORING_GLOBAL) w3_array += cstart; 267 ierr = VecRestoreArray(w3,&w3_array);CHKERRQ(ierr); 268 269 /* 270 (3-2) Evaluate function at w3 = x1 + dx (here dx is a vector of perturbations) 271 w2 = F(x1 + dx) - F(x1) 272 */ 273 ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 274 ierr = VecPlaceArray(w2,dy_k);CHKERRQ(ierr); /* place w2 to the array dy_i */ 275 ierr = (*f)(sctx,w3,w2,fctx);CHKERRQ(ierr); 276 ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 277 ierr = VecAXPY(w2,-1.0,w1);CHKERRQ(ierr); 278 ierr = VecResetArray(w2);CHKERRQ(ierr); 279 dy_k += m; /* points to dy+i*nxloc */ 280 } 281 282 /* 283 (3-3) Loop over block rows of vector, putting results into Jacobian matrix 284 */ 285 nrows_k = nrows[nbcols++]; 286 287 if (coloring->htype[0] == 'w') { 288 for (l=0; l<nrows_k; l++) { 289 row = Jentry2[nz].row; /* local row index */ 290 *(Jentry2[nz++].valaddr) = dy[row]*dx; 291 } 292 } else { /* htype == 'ds' */ 293 for (l=0; l<nrows_k; l++) { 294 row = Jentry[nz].row; /* local row index */ 295 *(Jentry[nz].valaddr) = dy[row]*vscale_array[Jentry[nz].col]; 296 nz++; 297 } 298 } 299 } 300 } else { /* bcols == 1 */ 301 for (k=0; k<ncolors; k++) { 302 coloring->currentcolor = k; 303 304 /* 305 (3-1) Loop over each column associated with color 306 adding the perturbation to the vector w3 = x1 + dx. 307 */ 308 ierr = VecCopy(x1,w3);CHKERRQ(ierr); 309 ierr = VecGetArray(w3,&w3_array);CHKERRQ(ierr); 310 if (ctype == IS_COLORING_GLOBAL) w3_array -= cstart; /* shift pointer so global index can be used */ 311 if (coloring->htype[0] == 'w') { 312 for (l=0; l<ncolumns[k]; l++) { 313 col = coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */ 314 w3_array[col] += 1.0/dx; 315 } 316 } else { /* htype == 'ds' */ 317 vscale_array -= cstart; /* shift pointer so global index can be used */ 318 for (l=0; l<ncolumns[k]; l++) { 319 col = coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */ 320 w3_array[col] += 1.0/vscale_array[col]; 321 } 322 vscale_array += cstart; 323 } 324 if (ctype == IS_COLORING_GLOBAL) w3_array += cstart; 325 ierr = VecRestoreArray(w3,&w3_array);CHKERRQ(ierr); 326 327 /* 328 (3-2) Evaluate function at w3 = x1 + dx (here dx is a vector of perturbations) 329 w2 = F(x1 + dx) - F(x1) 330 */ 331 ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 332 ierr = (*f)(sctx,w3,w2,fctx);CHKERRQ(ierr); 333 ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr); 334 ierr = VecAXPY(w2,-1.0,w1);CHKERRQ(ierr); 335 336 /* 337 (3-3) Loop over rows of vector, putting results into Jacobian matrix 338 */ 339 nrows_k = nrows[k]; 340 ierr = VecGetArray(w2,&y);CHKERRQ(ierr); 341 if (coloring->htype[0] == 'w') { 342 for (l=0; l<nrows_k; l++) { 343 row = Jentry2[nz].row; /* local row index */ 344 *(Jentry2[nz++].valaddr) = y[row]*dx; 345 } 346 } else { /* htype == 'ds' */ 347 for (l=0; l<nrows_k; l++) { 348 row = Jentry[nz].row; /* local row index */ 349 *(Jentry[nz].valaddr) = y[row]*vscale_array[Jentry[nz].col]; 350 nz++; 351 } 352 } 353 ierr = VecRestoreArray(w2,&y);CHKERRQ(ierr); 354 } 355 } 356 357 #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) 358 if (J->offloadmask != PETSC_OFFLOAD_UNALLOCATED) J->offloadmask = PETSC_OFFLOAD_CPU; 359 #endif 360 ierr = MatAssemblyBegin(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 361 ierr = MatAssemblyEnd(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 362 if (vscale) { 363 ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr); 364 } 365 coloring->currentcolor = -1; 366 if (!alreadyboundtocpu) {ierr = VecBindToCPU(x1,PETSC_FALSE);CHKERRQ(ierr);} 367 PetscFunctionReturn(0); 368 } 369 370 PetscErrorCode MatFDColoringSetUp_MPIXAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c) 371 { 372 PetscErrorCode ierr; 373 PetscMPIInt size,*ncolsonproc,*disp,nn; 374 PetscInt i,n,nrows,nrows_i,j,k,m,ncols,col,*rowhit,cstart,cend,colb; 375 const PetscInt *is,*A_ci,*A_cj,*B_ci,*B_cj,*row=NULL,*ltog=NULL; 376 PetscInt nis=iscoloring->n,nctot,*cols,tmp = 0; 377 ISLocalToGlobalMapping map=mat->cmap->mapping; 378 PetscInt ctype=c->ctype,*spidxA,*spidxB,nz,bs,bs2,spidx; 379 Mat A,B; 380 PetscScalar *A_val,*B_val,**valaddrhit; 381 MatEntry *Jentry; 382 MatEntry2 *Jentry2; 383 PetscBool isBAIJ,isSELL; 384 PetscInt bcols=c->bcols; 385 #if defined(PETSC_USE_CTABLE) 386 PetscTable colmap=NULL; 387 #else 388 PetscInt *colmap=NULL; /* local col number of off-diag col */ 389 #endif 390 391 PetscFunctionBegin; 392 if (ctype == IS_COLORING_LOCAL) { 393 if (!map) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_INCOMP,"When using ghosted differencing matrix must have local to global mapping provided with MatSetLocalToGlobalMapping"); 394 ierr = ISLocalToGlobalMappingGetIndices(map,<og);CHKERRQ(ierr); 395 } 396 397 ierr = MatGetBlockSize(mat,&bs);CHKERRQ(ierr); 398 ierr = PetscObjectBaseTypeCompare((PetscObject)mat,MATMPIBAIJ,&isBAIJ);CHKERRQ(ierr); 399 ierr = PetscObjectTypeCompare((PetscObject)mat,MATMPISELL,&isSELL);CHKERRQ(ierr); 400 if (isBAIJ) { 401 Mat_MPIBAIJ *baij=(Mat_MPIBAIJ*)mat->data; 402 Mat_SeqBAIJ *spA,*spB; 403 A = baij->A; spA = (Mat_SeqBAIJ*)A->data; A_val = spA->a; 404 B = baij->B; spB = (Mat_SeqBAIJ*)B->data; B_val = spB->a; 405 nz = spA->nz + spB->nz; /* total nonzero entries of mat */ 406 if (!baij->colmap) { 407 ierr = MatCreateColmap_MPIBAIJ_Private(mat);CHKERRQ(ierr); 408 } 409 colmap = baij->colmap; 410 ierr = MatGetColumnIJ_SeqBAIJ_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr); 411 ierr = MatGetColumnIJ_SeqBAIJ_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr); 412 413 if (ctype == IS_COLORING_GLOBAL && c->htype[0] == 'd') { /* create vscale for storing dx */ 414 PetscInt *garray; 415 ierr = PetscMalloc1(B->cmap->n,&garray);CHKERRQ(ierr); 416 for (i=0; i<baij->B->cmap->n/bs; i++) { 417 for (j=0; j<bs; j++) { 418 garray[i*bs+j] = bs*baij->garray[i]+j; 419 } 420 } 421 ierr = VecCreateGhost(PetscObjectComm((PetscObject)mat),mat->cmap->n,PETSC_DETERMINE,B->cmap->n,garray,&c->vscale);CHKERRQ(ierr); 422 ierr = VecBindToCPU(c->vscale,PETSC_TRUE);CHKERRQ(ierr); 423 ierr = PetscFree(garray);CHKERRQ(ierr); 424 } 425 } else if (isSELL) { 426 Mat_MPISELL *sell=(Mat_MPISELL*)mat->data; 427 Mat_SeqSELL *spA,*spB; 428 A = sell->A; spA = (Mat_SeqSELL*)A->data; A_val = spA->val; 429 B = sell->B; spB = (Mat_SeqSELL*)B->data; B_val = spB->val; 430 nz = spA->nz + spB->nz; /* total nonzero entries of mat */ 431 if (!sell->colmap) { 432 /* Allow access to data structures of local part of matrix 433 - creates aij->colmap which maps global column number to local number in part B */ 434 ierr = MatCreateColmap_MPISELL_Private(mat);CHKERRQ(ierr); 435 } 436 colmap = sell->colmap; 437 ierr = MatGetColumnIJ_SeqSELL_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr); 438 ierr = MatGetColumnIJ_SeqSELL_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr); 439 440 bs = 1; /* only bs=1 is supported for non MPIBAIJ matrix */ 441 442 if (ctype == IS_COLORING_GLOBAL && c->htype[0] == 'd') { /* create vscale for storing dx */ 443 ierr = VecCreateGhost(PetscObjectComm((PetscObject)mat),mat->cmap->n,PETSC_DETERMINE,B->cmap->n,sell->garray,&c->vscale);CHKERRQ(ierr); 444 ierr = VecBindToCPU(c->vscale,PETSC_TRUE);CHKERRQ(ierr); 445 } 446 } else { 447 Mat_MPIAIJ *aij=(Mat_MPIAIJ*)mat->data; 448 Mat_SeqAIJ *spA,*spB; 449 A = aij->A; spA = (Mat_SeqAIJ*)A->data; A_val = spA->a; 450 B = aij->B; spB = (Mat_SeqAIJ*)B->data; B_val = spB->a; 451 nz = spA->nz + spB->nz; /* total nonzero entries of mat */ 452 if (!aij->colmap) { 453 /* Allow access to data structures of local part of matrix 454 - creates aij->colmap which maps global column number to local number in part B */ 455 ierr = MatCreateColmap_MPIAIJ_Private(mat);CHKERRQ(ierr); 456 } 457 colmap = aij->colmap; 458 ierr = MatGetColumnIJ_SeqAIJ_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr); 459 ierr = MatGetColumnIJ_SeqAIJ_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr); 460 461 bs = 1; /* only bs=1 is supported for non MPIBAIJ matrix */ 462 463 if (ctype == IS_COLORING_GLOBAL && c->htype[0] == 'd') { /* create vscale for storing dx */ 464 ierr = VecCreateGhost(PetscObjectComm((PetscObject)mat),mat->cmap->n,PETSC_DETERMINE,B->cmap->n,aij->garray,&c->vscale);CHKERRQ(ierr); 465 ierr = VecBindToCPU(c->vscale,PETSC_TRUE);CHKERRQ(ierr); 466 } 467 } 468 469 m = mat->rmap->n/bs; 470 cstart = mat->cmap->rstart/bs; 471 cend = mat->cmap->rend/bs; 472 473 ierr = PetscMalloc2(nis,&c->ncolumns,nis,&c->columns);CHKERRQ(ierr); 474 ierr = PetscMalloc1(nis,&c->nrows);CHKERRQ(ierr); 475 ierr = PetscLogObjectMemory((PetscObject)c,3*nis*sizeof(PetscInt));CHKERRQ(ierr); 476 477 if (c->htype[0] == 'd') { 478 ierr = PetscMalloc1(nz,&Jentry);CHKERRQ(ierr); 479 ierr = PetscLogObjectMemory((PetscObject)c,nz*sizeof(MatEntry));CHKERRQ(ierr); 480 c->matentry = Jentry; 481 } else if (c->htype[0] == 'w') { 482 ierr = PetscMalloc1(nz,&Jentry2);CHKERRQ(ierr); 483 ierr = PetscLogObjectMemory((PetscObject)c,nz*sizeof(MatEntry2));CHKERRQ(ierr); 484 c->matentry2 = Jentry2; 485 } else SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"htype is not supported"); 486 487 ierr = PetscMalloc2(m+1,&rowhit,m+1,&valaddrhit);CHKERRQ(ierr); 488 nz = 0; 489 ierr = ISColoringGetIS(iscoloring,PETSC_OWN_POINTER, PETSC_IGNORE,&c->isa);CHKERRQ(ierr); 490 491 if (ctype == IS_COLORING_GLOBAL) { 492 ierr = MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);CHKERRMPI(ierr); 493 ierr = PetscMalloc2(size,&ncolsonproc,size,&disp);CHKERRQ(ierr); 494 } 495 496 for (i=0; i<nis; i++) { /* for each local color */ 497 ierr = ISGetLocalSize(c->isa[i],&n);CHKERRQ(ierr); 498 ierr = ISGetIndices(c->isa[i],&is);CHKERRQ(ierr); 499 500 c->ncolumns[i] = n; /* local number of columns of this color on this process */ 501 c->columns[i] = (PetscInt*)is; 502 503 if (ctype == IS_COLORING_GLOBAL) { 504 /* Determine nctot, the total (parallel) number of columns of this color */ 505 /* ncolsonproc[j]: local ncolumns on proc[j] of this color */ 506 ierr = PetscMPIIntCast(n,&nn);CHKERRQ(ierr); 507 ierr = MPI_Allgather(&nn,1,MPI_INT,ncolsonproc,1,MPI_INT,PetscObjectComm((PetscObject)mat));CHKERRMPI(ierr); 508 nctot = 0; for (j=0; j<size; j++) nctot += ncolsonproc[j]; 509 if (!nctot) { 510 ierr = PetscInfo(mat,"Coloring of matrix has some unneeded colors with no corresponding rows\n");CHKERRQ(ierr); 511 } 512 513 disp[0] = 0; 514 for (j=1; j<size; j++) { 515 disp[j] = disp[j-1] + ncolsonproc[j-1]; 516 } 517 518 /* Get cols, the complete list of columns for this color on each process */ 519 ierr = PetscMalloc1(nctot+1,&cols);CHKERRQ(ierr); 520 ierr = MPI_Allgatherv((void*)is,n,MPIU_INT,cols,ncolsonproc,disp,MPIU_INT,PetscObjectComm((PetscObject)mat));CHKERRMPI(ierr); 521 } else if (ctype == IS_COLORING_LOCAL) { 522 /* Determine local number of columns of this color on this process, including ghost points */ 523 nctot = n; 524 cols = (PetscInt*)is; 525 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not provided for this MatFDColoring type"); 526 527 /* Mark all rows affect by these columns */ 528 ierr = PetscArrayzero(rowhit,m);CHKERRQ(ierr); 529 bs2 = bs*bs; 530 nrows_i = 0; 531 for (j=0; j<nctot; j++) { /* loop over columns*/ 532 if (ctype == IS_COLORING_LOCAL) { 533 col = ltog[cols[j]]; 534 } else { 535 col = cols[j]; 536 } 537 if (col >= cstart && col < cend) { /* column is in A, diagonal block of mat */ 538 tmp = A_ci[col-cstart]; 539 row = A_cj + tmp; 540 nrows = A_ci[col-cstart+1] - tmp; 541 nrows_i += nrows; 542 543 /* loop over columns of A marking them in rowhit */ 544 for (k=0; k<nrows; k++) { 545 /* set valaddrhit for part A */ 546 spidx = bs2*spidxA[tmp + k]; 547 valaddrhit[*row] = &A_val[spidx]; 548 rowhit[*row++] = col - cstart + 1; /* local column index */ 549 } 550 } else { /* column is in B, off-diagonal block of mat */ 551 #if defined(PETSC_USE_CTABLE) 552 ierr = PetscTableFind(colmap,col+1,&colb);CHKERRQ(ierr); 553 colb--; 554 #else 555 colb = colmap[col] - 1; /* local column index */ 556 #endif 557 if (colb == -1) { 558 nrows = 0; 559 } else { 560 colb = colb/bs; 561 tmp = B_ci[colb]; 562 row = B_cj + tmp; 563 nrows = B_ci[colb+1] - tmp; 564 } 565 nrows_i += nrows; 566 /* loop over columns of B marking them in rowhit */ 567 for (k=0; k<nrows; k++) { 568 /* set valaddrhit for part B */ 569 spidx = bs2*spidxB[tmp + k]; 570 valaddrhit[*row] = &B_val[spidx]; 571 rowhit[*row++] = colb + 1 + cend - cstart; /* local column index */ 572 } 573 } 574 } 575 c->nrows[i] = nrows_i; 576 577 if (c->htype[0] == 'd') { 578 for (j=0; j<m; j++) { 579 if (rowhit[j]) { 580 Jentry[nz].row = j; /* local row index */ 581 Jentry[nz].col = rowhit[j] - 1; /* local column index */ 582 Jentry[nz].valaddr = valaddrhit[j]; /* address of mat value for this entry */ 583 nz++; 584 } 585 } 586 } else { /* c->htype == 'wp' */ 587 for (j=0; j<m; j++) { 588 if (rowhit[j]) { 589 Jentry2[nz].row = j; /* local row index */ 590 Jentry2[nz].valaddr = valaddrhit[j]; /* address of mat value for this entry */ 591 nz++; 592 } 593 } 594 } 595 if (ctype == IS_COLORING_GLOBAL) { 596 ierr = PetscFree(cols);CHKERRQ(ierr); 597 } 598 } 599 if (ctype == IS_COLORING_GLOBAL) { 600 ierr = PetscFree2(ncolsonproc,disp);CHKERRQ(ierr); 601 } 602 603 if (bcols > 1) { /* reorder Jentry for faster MatFDColoringApply() */ 604 ierr = MatFDColoringSetUpBlocked_AIJ_Private(mat,c,nz);CHKERRQ(ierr); 605 } 606 607 if (isBAIJ) { 608 ierr = MatRestoreColumnIJ_SeqBAIJ_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr); 609 ierr = MatRestoreColumnIJ_SeqBAIJ_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr); 610 ierr = PetscMalloc1(bs*mat->rmap->n,&c->dy);CHKERRQ(ierr); 611 } else if (isSELL) { 612 ierr = MatRestoreColumnIJ_SeqSELL_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr); 613 ierr = MatRestoreColumnIJ_SeqSELL_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr); 614 } else { 615 ierr = MatRestoreColumnIJ_SeqAIJ_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr); 616 ierr = MatRestoreColumnIJ_SeqAIJ_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr); 617 } 618 619 ierr = ISColoringRestoreIS(iscoloring,PETSC_OWN_POINTER,&c->isa);CHKERRQ(ierr); 620 ierr = PetscFree2(rowhit,valaddrhit);CHKERRQ(ierr); 621 622 if (ctype == IS_COLORING_LOCAL) { 623 ierr = ISLocalToGlobalMappingRestoreIndices(map,<og);CHKERRQ(ierr); 624 } 625 ierr = PetscInfo3(c,"ncolors %D, brows %D and bcols %D are used.\n",c->ncolors,c->brows,c->bcols);CHKERRQ(ierr); 626 PetscFunctionReturn(0); 627 } 628 629 PetscErrorCode MatFDColoringCreate_MPIXAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c) 630 { 631 PetscErrorCode ierr; 632 PetscInt bs,nis=iscoloring->n,m=mat->rmap->n; 633 PetscBool isBAIJ,isSELL; 634 635 PetscFunctionBegin; 636 /* set default brows and bcols for speedup inserting the dense matrix into sparse Jacobian; 637 bcols is chosen s.t. dy-array takes 50% of memory space as mat */ 638 ierr = MatGetBlockSize(mat,&bs);CHKERRQ(ierr); 639 ierr = PetscObjectBaseTypeCompare((PetscObject)mat,MATMPIBAIJ,&isBAIJ);CHKERRQ(ierr); 640 ierr = PetscObjectTypeCompare((PetscObject)mat,MATMPISELL,&isSELL);CHKERRQ(ierr); 641 if (isBAIJ || m == 0) { 642 c->brows = m; 643 c->bcols = 1; 644 } else if (isSELL) { 645 /* bcols is chosen s.t. dy-array takes 50% of local memory space as mat */ 646 Mat_MPISELL *sell=(Mat_MPISELL*)mat->data; 647 Mat_SeqSELL *spA,*spB; 648 Mat A,B; 649 PetscInt nz,brows,bcols; 650 PetscReal mem; 651 652 bs = 1; /* only bs=1 is supported for MPISELL matrix */ 653 654 A = sell->A; spA = (Mat_SeqSELL*)A->data; 655 B = sell->B; spB = (Mat_SeqSELL*)B->data; 656 nz = spA->nz + spB->nz; /* total local nonzero entries of mat */ 657 mem = nz*(sizeof(PetscScalar) + sizeof(PetscInt)) + 3*m*sizeof(PetscInt); 658 bcols = (PetscInt)(0.5*mem /(m*sizeof(PetscScalar))); 659 brows = 1000/bcols; 660 if (bcols > nis) bcols = nis; 661 if (brows == 0 || brows > m) brows = m; 662 c->brows = brows; 663 c->bcols = bcols; 664 } else { /* mpiaij matrix */ 665 /* bcols is chosen s.t. dy-array takes 50% of local memory space as mat */ 666 Mat_MPIAIJ *aij=(Mat_MPIAIJ*)mat->data; 667 Mat_SeqAIJ *spA,*spB; 668 Mat A,B; 669 PetscInt nz,brows,bcols; 670 PetscReal mem; 671 672 bs = 1; /* only bs=1 is supported for MPIAIJ matrix */ 673 674 A = aij->A; spA = (Mat_SeqAIJ*)A->data; 675 B = aij->B; spB = (Mat_SeqAIJ*)B->data; 676 nz = spA->nz + spB->nz; /* total local nonzero entries of mat */ 677 mem = nz*(sizeof(PetscScalar) + sizeof(PetscInt)) + 3*m*sizeof(PetscInt); 678 bcols = (PetscInt)(0.5*mem /(m*sizeof(PetscScalar))); 679 brows = 1000/bcols; 680 if (bcols > nis) bcols = nis; 681 if (brows == 0 || brows > m) brows = m; 682 c->brows = brows; 683 c->bcols = bcols; 684 } 685 686 c->M = mat->rmap->N/bs; /* set the global rows and columns and local rows */ 687 c->N = mat->cmap->N/bs; 688 c->m = mat->rmap->n/bs; 689 c->rstart = mat->rmap->rstart/bs; 690 c->ncolors = nis; 691 PetscFunctionReturn(0); 692 } 693 694 /*@C 695 696 MatFDColoringSetValues - takes a matrix in compressed color format and enters the matrix into a PETSc Mat 697 698 Collective on J 699 700 Input Parameters: 701 + J - the sparse matrix 702 . coloring - created with MatFDColoringCreate() and a local coloring 703 - y - column major storage of matrix values with one color of values per column, the number of rows of y should match 704 the number of local rows of J and the number of columns is the number of colors. 705 706 Level: intermediate 707 708 Notes: the matrix in compressed color format may come from an Automatic Differentiation code 709 710 The code will be slightly faster if MatFDColoringSetBlockSize(coloring,PETSC_DEFAULT,nc); is called immediately after creating the coloring 711 712 .seealso: MatFDColoringCreate(), ISColoring, ISColoringCreate(), ISColoringSetType(), IS_COLORING_LOCAL, MatFDColoringSetBlockSize() 713 714 @*/ 715 PetscErrorCode MatFDColoringSetValues(Mat J,MatFDColoring coloring,const PetscScalar *y) 716 { 717 PetscErrorCode ierr; 718 MatEntry2 *Jentry2; 719 PetscInt row,i,nrows_k,l,ncolors,nz = 0,bcols,nbcols = 0; 720 const PetscInt *nrows; 721 PetscBool eq; 722 723 PetscFunctionBegin; 724 PetscValidHeaderSpecific(J,MAT_CLASSID,1); 725 PetscValidHeaderSpecific(coloring,MAT_FDCOLORING_CLASSID,2); 726 ierr = PetscObjectCompareId((PetscObject)J,coloring->matid,&eq);CHKERRQ(ierr); 727 if (!eq) SETERRQ(PetscObjectComm((PetscObject)J),PETSC_ERR_ARG_WRONG,"Matrix used with MatFDColoringSetValues() must be that used with MatFDColoringCreate()"); 728 Jentry2 = coloring->matentry2; 729 nrows = coloring->nrows; 730 ncolors = coloring->ncolors; 731 bcols = coloring->bcols; 732 733 for (i=0; i<ncolors; i+=bcols) { 734 nrows_k = nrows[nbcols++]; 735 for (l=0; l<nrows_k; l++) { 736 row = Jentry2[nz].row; /* local row index */ 737 *(Jentry2[nz++].valaddr) = y[row]; 738 } 739 y += bcols*coloring->m; 740 } 741 ierr = MatAssemblyBegin(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 742 ierr = MatAssemblyEnd(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 743 PetscFunctionReturn(0); 744 } 745