1 2 #include <../src/mat/impls/baij/seq/baij.h> 3 #include <../src/mat/impls/sbaij/seq/sbaij.h> 4 #include <petsc/private/kernels/blockinvert.h> 5 #include <petscis.h> 6 7 /* 8 input: 9 F -- numeric factor 10 output: 11 nneg, nzero, npos: matrix inertia 12 */ 13 14 #undef __FUNCT__ 15 #define __FUNCT__ "MatGetInertia_SeqSBAIJ" 16 PetscErrorCode MatGetInertia_SeqSBAIJ(Mat F,PetscInt *nneig,PetscInt *nzero,PetscInt *npos) 17 { 18 Mat_SeqSBAIJ *fact_ptr = (Mat_SeqSBAIJ*)F->data; 19 MatScalar *dd = fact_ptr->a; 20 PetscInt mbs =fact_ptr->mbs,bs=F->rmap->bs,i,nneig_tmp,npos_tmp,*fi = fact_ptr->diag; 21 22 PetscFunctionBegin; 23 if (bs != 1) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"No support for bs: %D >1 yet",bs); 24 nneig_tmp = 0; npos_tmp = 0; 25 for (i=0; i<mbs; i++) { 26 if (PetscRealPart(dd[*fi]) > 0.0) npos_tmp++; 27 else if (PetscRealPart(dd[*fi]) < 0.0) nneig_tmp++; 28 fi++; 29 } 30 if (nneig) *nneig = nneig_tmp; 31 if (npos) *npos = npos_tmp; 32 if (nzero) *nzero = mbs - nneig_tmp - npos_tmp; 33 PetscFunctionReturn(0); 34 } 35 36 /* 37 Symbolic U^T*D*U factorization for SBAIJ format. Modified from SSF of YSMP. 38 Use Modified Sparse Row (MSR) storage for u and ju. See page 85, "Iterative Methods ..." by Saad. 39 */ 40 #undef __FUNCT__ 41 #define __FUNCT__ "MatCholeskyFactorSymbolic_SeqSBAIJ_MSR" 42 PetscErrorCode MatCholeskyFactorSymbolic_SeqSBAIJ_MSR(Mat F,Mat A,IS perm,const MatFactorInfo *info) 43 { 44 Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data,*b; 45 PetscErrorCode ierr; 46 const PetscInt *rip,*ai,*aj; 47 PetscInt i,mbs = a->mbs,*jutmp,bs = A->rmap->bs,bs2=a->bs2; 48 PetscInt m,reallocs = 0,prow; 49 PetscInt *jl,*q,jmin,jmax,juidx,nzk,qm,*iu,*ju,k,j,vj,umax,maxadd; 50 PetscReal f = info->fill; 51 PetscBool perm_identity; 52 53 PetscFunctionBegin; 54 /* check whether perm is the identity mapping */ 55 ierr = ISIdentity(perm,&perm_identity);CHKERRQ(ierr); 56 ierr = ISGetIndices(perm,&rip);CHKERRQ(ierr); 57 58 if (perm_identity) { /* without permutation */ 59 a->permute = PETSC_FALSE; 60 61 ai = a->i; aj = a->j; 62 } else { /* non-trivial permutation */ 63 a->permute = PETSC_TRUE; 64 65 ierr = MatReorderingSeqSBAIJ(A,perm);CHKERRQ(ierr); 66 67 ai = a->inew; aj = a->jnew; 68 } 69 70 /* initialization */ 71 ierr = PetscMalloc1(mbs+1,&iu);CHKERRQ(ierr); 72 umax = (PetscInt)(f*ai[mbs] + 1); umax += mbs + 1; 73 ierr = PetscMalloc1(umax,&ju);CHKERRQ(ierr); 74 iu[0] = mbs+1; 75 juidx = mbs + 1; /* index for ju */ 76 /* jl linked list for pivot row -- linked list for col index */ 77 ierr = PetscMalloc2(mbs,&jl,mbs,&q);CHKERRQ(ierr); 78 for (i=0; i<mbs; i++) { 79 jl[i] = mbs; 80 q[i] = 0; 81 } 82 83 /* for each row k */ 84 for (k=0; k<mbs; k++) { 85 for (i=0; i<mbs; i++) q[i] = 0; /* to be removed! */ 86 nzk = 0; /* num. of nz blocks in k-th block row with diagonal block excluded */ 87 q[k] = mbs; 88 /* initialize nonzero structure of k-th row to row rip[k] of A */ 89 jmin = ai[rip[k]] +1; /* exclude diag[k] */ 90 jmax = ai[rip[k]+1]; 91 for (j=jmin; j<jmax; j++) { 92 vj = rip[aj[j]]; /* col. value */ 93 if (vj > k) { 94 qm = k; 95 do { 96 m = qm; qm = q[m]; 97 } while (qm < vj); 98 if (qm == vj) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Duplicate entry in A\n"); 99 nzk++; 100 q[m] = vj; 101 q[vj] = qm; 102 } /* if (vj > k) */ 103 } /* for (j=jmin; j<jmax; j++) */ 104 105 /* modify nonzero structure of k-th row by computing fill-in 106 for each row i to be merged in */ 107 prow = k; 108 prow = jl[prow]; /* next pivot row (== mbs for symbolic factorization) */ 109 110 while (prow < k) { 111 /* merge row prow into k-th row */ 112 jmin = iu[prow] + 1; jmax = iu[prow+1]; 113 qm = k; 114 for (j=jmin; j<jmax; j++) { 115 vj = ju[j]; 116 do { 117 m = qm; qm = q[m]; 118 } while (qm < vj); 119 if (qm != vj) { 120 nzk++; q[m] = vj; q[vj] = qm; qm = vj; 121 } 122 } 123 prow = jl[prow]; /* next pivot row */ 124 } 125 126 /* add k to row list for first nonzero element in k-th row */ 127 if (nzk > 0) { 128 i = q[k]; /* col value of first nonzero element in U(k, k+1:mbs-1) */ 129 jl[k] = jl[i]; jl[i] = k; 130 } 131 iu[k+1] = iu[k] + nzk; 132 133 /* allocate more space to ju if needed */ 134 if (iu[k+1] > umax) { 135 /* estimate how much additional space we will need */ 136 /* use the strategy suggested by David Hysom <hysom@perch-t.icase.edu> */ 137 /* just double the memory each time */ 138 maxadd = umax; 139 if (maxadd < nzk) maxadd = (mbs-k)*(nzk+1)/2; 140 umax += maxadd; 141 142 /* allocate a longer ju */ 143 ierr = PetscMalloc1(umax,&jutmp);CHKERRQ(ierr); 144 ierr = PetscMemcpy(jutmp,ju,iu[k]*sizeof(PetscInt));CHKERRQ(ierr); 145 ierr = PetscFree(ju);CHKERRQ(ierr); 146 ju = jutmp; 147 reallocs++; /* count how many times we realloc */ 148 } 149 150 /* save nonzero structure of k-th row in ju */ 151 i=k; 152 while (nzk--) { 153 i = q[i]; 154 ju[juidx++] = i; 155 } 156 } 157 158 #if defined(PETSC_USE_INFO) 159 if (ai[mbs] != 0) { 160 PetscReal af = ((PetscReal)iu[mbs])/((PetscReal)ai[mbs]); 161 ierr = PetscInfo3(A,"Reallocs %D Fill ratio:given %g needed %g\n",reallocs,(double)f,(double)af);CHKERRQ(ierr); 162 ierr = PetscInfo1(A,"Run with -pc_factor_fill %g or use \n",(double)af);CHKERRQ(ierr); 163 ierr = PetscInfo1(A,"PCFactorSetFill(pc,%g);\n",(double)af);CHKERRQ(ierr); 164 ierr = PetscInfo(A,"for best performance.\n");CHKERRQ(ierr); 165 } else { 166 ierr = PetscInfo(A,"Empty matrix.\n");CHKERRQ(ierr); 167 } 168 #endif 169 170 ierr = ISRestoreIndices(perm,&rip);CHKERRQ(ierr); 171 ierr = PetscFree2(jl,q);CHKERRQ(ierr); 172 173 /* put together the new matrix */ 174 ierr = MatSeqSBAIJSetPreallocation_SeqSBAIJ(F,bs,MAT_SKIP_ALLOCATION,NULL);CHKERRQ(ierr); 175 176 /* ierr = PetscLogObjectParent((PetscObject)B,(PetscObject)iperm);CHKERRQ(ierr); */ 177 b = (Mat_SeqSBAIJ*)(F)->data; 178 b->singlemalloc = PETSC_FALSE; 179 b->free_a = PETSC_TRUE; 180 b->free_ij = PETSC_TRUE; 181 182 ierr = PetscMalloc1((iu[mbs]+1)*bs2,&b->a);CHKERRQ(ierr); 183 b->j = ju; 184 b->i = iu; 185 b->diag = 0; 186 b->ilen = 0; 187 b->imax = 0; 188 b->row = perm; 189 190 b->pivotinblocks = PETSC_FALSE; /* need to get from MatFactorInfo */ 191 192 ierr = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr); 193 194 b->icol = perm; 195 ierr = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr); 196 ierr = PetscMalloc1(bs*mbs+bs,&b->solve_work);CHKERRQ(ierr); 197 /* In b structure: Free imax, ilen, old a, old j. 198 Allocate idnew, solve_work, new a, new j */ 199 ierr = PetscLogObjectMemory((PetscObject)F,(iu[mbs]-mbs)*(sizeof(PetscInt)+sizeof(MatScalar)));CHKERRQ(ierr); 200 b->maxnz = b->nz = iu[mbs]; 201 202 (F)->info.factor_mallocs = reallocs; 203 (F)->info.fill_ratio_given = f; 204 if (ai[mbs] != 0) { 205 (F)->info.fill_ratio_needed = ((PetscReal)iu[mbs])/((PetscReal)ai[mbs]); 206 } else { 207 (F)->info.fill_ratio_needed = 0.0; 208 } 209 ierr = MatSeqSBAIJSetNumericFactorization_inplace(F,perm_identity);CHKERRQ(ierr); 210 PetscFunctionReturn(0); 211 } 212 /* 213 Symbolic U^T*D*U factorization for SBAIJ format. 214 See MatICCFactorSymbolic_SeqAIJ() for description of its data structure. 215 */ 216 #include <petscbt.h> 217 #include <../src/mat/utils/freespace.h> 218 #undef __FUNCT__ 219 #define __FUNCT__ "MatCholeskyFactorSymbolic_SeqSBAIJ" 220 PetscErrorCode MatCholeskyFactorSymbolic_SeqSBAIJ(Mat fact,Mat A,IS perm,const MatFactorInfo *info) 221 { 222 Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data; 223 Mat_SeqSBAIJ *b; 224 PetscErrorCode ierr; 225 PetscBool perm_identity,missing; 226 PetscReal fill = info->fill; 227 const PetscInt *rip,*ai=a->i,*aj=a->j; 228 PetscInt i,mbs=a->mbs,bs=A->rmap->bs,reallocs=0,prow; 229 PetscInt *jl,jmin,jmax,nzk,*ui,k,j,*il,nextprow; 230 PetscInt nlnk,*lnk,ncols,*cols,*uj,**ui_ptr,*uj_ptr,*udiag; 231 PetscFreeSpaceList free_space=NULL,current_space=NULL; 232 PetscBT lnkbt; 233 234 PetscFunctionBegin; 235 if (A->rmap->n != A->cmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Must be square matrix, rows %D columns %D",A->rmap->n,A->cmap->n); 236 ierr = MatMissingDiagonal(A,&missing,&i);CHKERRQ(ierr); 237 if (missing) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Matrix is missing diagonal entry %D",i); 238 if (bs > 1) { 239 ierr = MatCholeskyFactorSymbolic_SeqSBAIJ_inplace(fact,A,perm,info);CHKERRQ(ierr); 240 PetscFunctionReturn(0); 241 } 242 243 /* check whether perm is the identity mapping */ 244 ierr = ISIdentity(perm,&perm_identity);CHKERRQ(ierr); 245 if (!perm_identity) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix reordering is not supported for sbaij matrix. Use aij format"); 246 a->permute = PETSC_FALSE; 247 ierr = ISGetIndices(perm,&rip);CHKERRQ(ierr); 248 249 /* initialization */ 250 ierr = PetscMalloc1(mbs+1,&ui);CHKERRQ(ierr); 251 ierr = PetscMalloc1(mbs+1,&udiag);CHKERRQ(ierr); 252 ui[0] = 0; 253 254 /* jl: linked list for storing indices of the pivot rows 255 il: il[i] points to the 1st nonzero entry of U(i,k:mbs-1) */ 256 ierr = PetscMalloc4(mbs,&ui_ptr,mbs,&il,mbs,&jl,mbs,&cols);CHKERRQ(ierr); 257 for (i=0; i<mbs; i++) { 258 jl[i] = mbs; il[i] = 0; 259 } 260 261 /* create and initialize a linked list for storing column indices of the active row k */ 262 nlnk = mbs + 1; 263 ierr = PetscLLCreate(mbs,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr); 264 265 /* initial FreeSpace size is fill*(ai[mbs]+1) */ 266 ierr = PetscFreeSpaceGet((PetscInt)(fill*(ai[mbs]+1)),&free_space);CHKERRQ(ierr); 267 current_space = free_space; 268 269 for (k=0; k<mbs; k++) { /* for each active row k */ 270 /* initialize lnk by the column indices of row rip[k] of A */ 271 nzk = 0; 272 ncols = ai[k+1] - ai[k]; 273 if (!ncols) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_MAT_CH_ZRPVT,"Empty row %D in matrix ",k); 274 for (j=0; j<ncols; j++) { 275 i = *(aj + ai[k] + j); 276 cols[j] = i; 277 } 278 ierr = PetscLLAdd(ncols,cols,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr); 279 nzk += nlnk; 280 281 /* update lnk by computing fill-in for each pivot row to be merged in */ 282 prow = jl[k]; /* 1st pivot row */ 283 284 while (prow < k) { 285 nextprow = jl[prow]; 286 /* merge prow into k-th row */ 287 jmin = il[prow] + 1; /* index of the 2nd nzero entry in U(prow,k:mbs-1) */ 288 jmax = ui[prow+1]; 289 ncols = jmax-jmin; 290 uj_ptr = ui_ptr[prow] + jmin - ui[prow]; /* points to the 2nd nzero entry in U(prow,k:mbs-1) */ 291 ierr = PetscLLAddSorted(ncols,uj_ptr,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr); 292 nzk += nlnk; 293 294 /* update il and jl for prow */ 295 if (jmin < jmax) { 296 il[prow] = jmin; 297 j = *uj_ptr; jl[prow] = jl[j]; jl[j] = prow; 298 } 299 prow = nextprow; 300 } 301 302 /* if free space is not available, make more free space */ 303 if (current_space->local_remaining<nzk) { 304 i = mbs - k + 1; /* num of unfactored rows */ 305 i *= PetscMin(nzk, i-1); /* i*nzk, i*(i-1): estimated and max additional space needed */ 306 ierr = PetscFreeSpaceGet(i,¤t_space);CHKERRQ(ierr); 307 reallocs++; 308 } 309 310 /* copy data into free space, then initialize lnk */ 311 ierr = PetscLLClean(mbs,mbs,nzk,lnk,current_space->array,lnkbt);CHKERRQ(ierr); 312 313 /* add the k-th row into il and jl */ 314 if (nzk > 1) { 315 i = current_space->array[1]; /* col value of the first nonzero element in U(k, k+1:mbs-1) */ 316 jl[k] = jl[i]; jl[i] = k; 317 il[k] = ui[k] + 1; 318 } 319 ui_ptr[k] = current_space->array; 320 321 current_space->array += nzk; 322 current_space->local_used += nzk; 323 current_space->local_remaining -= nzk; 324 325 ui[k+1] = ui[k] + nzk; 326 } 327 328 ierr = ISRestoreIndices(perm,&rip);CHKERRQ(ierr); 329 ierr = PetscFree4(ui_ptr,il,jl,cols);CHKERRQ(ierr); 330 331 /* destroy list of free space and other temporary array(s) */ 332 ierr = PetscMalloc1(ui[mbs]+1,&uj);CHKERRQ(ierr); 333 ierr = PetscFreeSpaceContiguous_Cholesky(&free_space,uj,mbs,ui,udiag);CHKERRQ(ierr); /* store matrix factor */ 334 ierr = PetscLLDestroy(lnk,lnkbt);CHKERRQ(ierr); 335 336 /* put together the new matrix in MATSEQSBAIJ format */ 337 ierr = MatSeqSBAIJSetPreallocation_SeqSBAIJ(fact,bs,MAT_SKIP_ALLOCATION,NULL);CHKERRQ(ierr); 338 339 b = (Mat_SeqSBAIJ*)fact->data; 340 b->singlemalloc = PETSC_FALSE; 341 b->free_a = PETSC_TRUE; 342 b->free_ij = PETSC_TRUE; 343 344 ierr = PetscMalloc1(ui[mbs]+1,&b->a);CHKERRQ(ierr); 345 346 b->j = uj; 347 b->i = ui; 348 b->diag = udiag; 349 b->free_diag = PETSC_TRUE; 350 b->ilen = 0; 351 b->imax = 0; 352 b->row = perm; 353 b->icol = perm; 354 355 ierr = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr); 356 ierr = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr); 357 358 b->pivotinblocks = PETSC_FALSE; /* need to get from MatFactorInfo */ 359 360 ierr = PetscMalloc1(mbs+1,&b->solve_work);CHKERRQ(ierr); 361 ierr = PetscLogObjectMemory((PetscObject)fact,ui[mbs]*(sizeof(PetscInt)+sizeof(MatScalar)));CHKERRQ(ierr); 362 363 b->maxnz = b->nz = ui[mbs]; 364 365 fact->info.factor_mallocs = reallocs; 366 fact->info.fill_ratio_given = fill; 367 if (ai[mbs] != 0) { 368 fact->info.fill_ratio_needed = ((PetscReal)ui[mbs])/ai[mbs]; 369 } else { 370 fact->info.fill_ratio_needed = 0.0; 371 } 372 #if defined(PETSC_USE_INFO) 373 if (ai[mbs] != 0) { 374 PetscReal af = fact->info.fill_ratio_needed; 375 ierr = PetscInfo3(A,"Reallocs %D Fill ratio:given %g needed %g\n",reallocs,(double)fill,(double)af);CHKERRQ(ierr); 376 ierr = PetscInfo1(A,"Run with -pc_factor_fill %g or use \n",(double)af);CHKERRQ(ierr); 377 ierr = PetscInfo1(A,"PCFactorSetFill(pc,%g) for best performance.\n",(double)af);CHKERRQ(ierr); 378 } else { 379 ierr = PetscInfo(A,"Empty matrix.\n");CHKERRQ(ierr); 380 } 381 #endif 382 fact->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering; 383 PetscFunctionReturn(0); 384 } 385 386 #undef __FUNCT__ 387 #define __FUNCT__ "MatCholeskyFactorSymbolic_SeqSBAIJ_inplace" 388 PetscErrorCode MatCholeskyFactorSymbolic_SeqSBAIJ_inplace(Mat fact,Mat A,IS perm,const MatFactorInfo *info) 389 { 390 Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data; 391 Mat_SeqSBAIJ *b; 392 PetscErrorCode ierr; 393 PetscBool perm_identity,missing; 394 PetscReal fill = info->fill; 395 const PetscInt *rip,*ai,*aj; 396 PetscInt i,mbs=a->mbs,bs=A->rmap->bs,reallocs=0,prow,d; 397 PetscInt *jl,jmin,jmax,nzk,*ui,k,j,*il,nextprow; 398 PetscInt nlnk,*lnk,ncols,*cols,*uj,**ui_ptr,*uj_ptr; 399 PetscFreeSpaceList free_space=NULL,current_space=NULL; 400 PetscBT lnkbt; 401 402 PetscFunctionBegin; 403 ierr = MatMissingDiagonal(A,&missing,&d);CHKERRQ(ierr); 404 if (missing) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Matrix is missing diagonal entry %D",d); 405 406 /* 407 This code originally uses Modified Sparse Row (MSR) storage 408 (see page 85, "Iterative Methods ..." by Saad) for the output matrix B - bad choise! 409 Then it is rewritten so the factor B takes seqsbaij format. However the associated 410 MatCholeskyFactorNumeric_() have not been modified for the cases of bs>1 or !perm_identity, 411 thus the original code in MSR format is still used for these cases. 412 The code below should replace MatCholeskyFactorSymbolic_SeqSBAIJ_MSR() whenever 413 MatCholeskyFactorNumeric_() is modified for using sbaij symbolic factor. 414 */ 415 if (bs > 1) { 416 ierr = MatCholeskyFactorSymbolic_SeqSBAIJ_MSR(fact,A,perm,info);CHKERRQ(ierr); 417 PetscFunctionReturn(0); 418 } 419 420 /* check whether perm is the identity mapping */ 421 ierr = ISIdentity(perm,&perm_identity);CHKERRQ(ierr); 422 423 if (perm_identity) { 424 a->permute = PETSC_FALSE; 425 426 ai = a->i; aj = a->j; 427 } else { 428 SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix reordering is not supported for sbaij matrix. Use aij format"); 429 #if 0 430 /* There are bugs for reordeing. Needs further work. 431 MatReordering for sbaij cannot be efficient. User should use aij formt! */ 432 a->permute = PETSC_TRUE; 433 434 ierr = MatReorderingSeqSBAIJ(A,perm);CHKERRQ(ierr); 435 ai = a->inew; aj = a->jnew; 436 #endif 437 } 438 ierr = ISGetIndices(perm,&rip);CHKERRQ(ierr); 439 440 /* initialization */ 441 ierr = PetscMalloc1(mbs+1,&ui);CHKERRQ(ierr); 442 ui[0] = 0; 443 444 /* jl: linked list for storing indices of the pivot rows 445 il: il[i] points to the 1st nonzero entry of U(i,k:mbs-1) */ 446 ierr = PetscMalloc4(mbs,&ui_ptr,mbs,&il,mbs,&jl,mbs,&cols);CHKERRQ(ierr); 447 for (i=0; i<mbs; i++) { 448 jl[i] = mbs; il[i] = 0; 449 } 450 451 /* create and initialize a linked list for storing column indices of the active row k */ 452 nlnk = mbs + 1; 453 ierr = PetscLLCreate(mbs,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr); 454 455 /* initial FreeSpace size is fill*(ai[mbs]+1) */ 456 ierr = PetscFreeSpaceGet((PetscInt)(fill*(ai[mbs]+1)),&free_space);CHKERRQ(ierr); 457 current_space = free_space; 458 459 for (k=0; k<mbs; k++) { /* for each active row k */ 460 /* initialize lnk by the column indices of row rip[k] of A */ 461 nzk = 0; 462 ncols = ai[rip[k]+1] - ai[rip[k]]; 463 for (j=0; j<ncols; j++) { 464 i = *(aj + ai[rip[k]] + j); 465 cols[j] = rip[i]; 466 } 467 ierr = PetscLLAdd(ncols,cols,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr); 468 nzk += nlnk; 469 470 /* update lnk by computing fill-in for each pivot row to be merged in */ 471 prow = jl[k]; /* 1st pivot row */ 472 473 while (prow < k) { 474 nextprow = jl[prow]; 475 /* merge prow into k-th row */ 476 jmin = il[prow] + 1; /* index of the 2nd nzero entry in U(prow,k:mbs-1) */ 477 jmax = ui[prow+1]; 478 ncols = jmax-jmin; 479 uj_ptr = ui_ptr[prow] + jmin - ui[prow]; /* points to the 2nd nzero entry in U(prow,k:mbs-1) */ 480 ierr = PetscLLAddSorted(ncols,uj_ptr,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr); 481 nzk += nlnk; 482 483 /* update il and jl for prow */ 484 if (jmin < jmax) { 485 il[prow] = jmin; 486 487 j = *uj_ptr; jl[prow] = jl[j]; jl[j] = prow; 488 } 489 prow = nextprow; 490 } 491 492 /* if free space is not available, make more free space */ 493 if (current_space->local_remaining<nzk) { 494 i = mbs - k + 1; /* num of unfactored rows */ 495 i = PetscMin(i*nzk, i*(i-1)); /* i*nzk, i*(i-1): estimated and max additional space needed */ 496 ierr = PetscFreeSpaceGet(i,¤t_space);CHKERRQ(ierr); 497 reallocs++; 498 } 499 500 /* copy data into free space, then initialize lnk */ 501 ierr = PetscLLClean(mbs,mbs,nzk,lnk,current_space->array,lnkbt);CHKERRQ(ierr); 502 503 /* add the k-th row into il and jl */ 504 if (nzk-1 > 0) { 505 i = current_space->array[1]; /* col value of the first nonzero element in U(k, k+1:mbs-1) */ 506 jl[k] = jl[i]; jl[i] = k; 507 il[k] = ui[k] + 1; 508 } 509 ui_ptr[k] = current_space->array; 510 511 current_space->array += nzk; 512 current_space->local_used += nzk; 513 current_space->local_remaining -= nzk; 514 515 ui[k+1] = ui[k] + nzk; 516 } 517 518 ierr = ISRestoreIndices(perm,&rip);CHKERRQ(ierr); 519 ierr = PetscFree4(ui_ptr,il,jl,cols);CHKERRQ(ierr); 520 521 /* destroy list of free space and other temporary array(s) */ 522 ierr = PetscMalloc1(ui[mbs]+1,&uj);CHKERRQ(ierr); 523 ierr = PetscFreeSpaceContiguous(&free_space,uj);CHKERRQ(ierr); 524 ierr = PetscLLDestroy(lnk,lnkbt);CHKERRQ(ierr); 525 526 /* put together the new matrix in MATSEQSBAIJ format */ 527 ierr = MatSeqSBAIJSetPreallocation_SeqSBAIJ(fact,bs,MAT_SKIP_ALLOCATION,NULL);CHKERRQ(ierr); 528 529 b = (Mat_SeqSBAIJ*)fact->data; 530 b->singlemalloc = PETSC_FALSE; 531 b->free_a = PETSC_TRUE; 532 b->free_ij = PETSC_TRUE; 533 534 ierr = PetscMalloc1(ui[mbs]+1,&b->a);CHKERRQ(ierr); 535 536 b->j = uj; 537 b->i = ui; 538 b->diag = 0; 539 b->ilen = 0; 540 b->imax = 0; 541 b->row = perm; 542 543 b->pivotinblocks = PETSC_FALSE; /* need to get from MatFactorInfo */ 544 545 ierr = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr); 546 b->icol = perm; 547 ierr = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr); 548 ierr = PetscMalloc1(mbs+1,&b->solve_work);CHKERRQ(ierr); 549 ierr = PetscLogObjectMemory((PetscObject)fact,(ui[mbs]-mbs)*(sizeof(PetscInt)+sizeof(MatScalar)));CHKERRQ(ierr); 550 b->maxnz = b->nz = ui[mbs]; 551 552 fact->info.factor_mallocs = reallocs; 553 fact->info.fill_ratio_given = fill; 554 if (ai[mbs] != 0) { 555 fact->info.fill_ratio_needed = ((PetscReal)ui[mbs])/ai[mbs]; 556 } else { 557 fact->info.fill_ratio_needed = 0.0; 558 } 559 #if defined(PETSC_USE_INFO) 560 if (ai[mbs] != 0) { 561 PetscReal af = fact->info.fill_ratio_needed; 562 ierr = PetscInfo3(A,"Reallocs %D Fill ratio:given %g needed %g\n",reallocs,(double)fill,(double)af);CHKERRQ(ierr); 563 ierr = PetscInfo1(A,"Run with -pc_factor_fill %g or use \n",(double)af);CHKERRQ(ierr); 564 ierr = PetscInfo1(A,"PCFactorSetFill(pc,%g) for best performance.\n",(double)af);CHKERRQ(ierr); 565 } else { 566 ierr = PetscInfo(A,"Empty matrix.\n");CHKERRQ(ierr); 567 } 568 #endif 569 ierr = MatSeqSBAIJSetNumericFactorization_inplace(fact,perm_identity);CHKERRQ(ierr); 570 PetscFunctionReturn(0); 571 } 572 573 #undef __FUNCT__ 574 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_N" 575 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_N(Mat C,Mat A,const MatFactorInfo *info) 576 { 577 Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data,*b = (Mat_SeqSBAIJ*)C->data; 578 IS perm = b->row; 579 PetscErrorCode ierr; 580 const PetscInt *ai,*aj,*perm_ptr,mbs=a->mbs,*bi=b->i,*bj=b->j; 581 PetscInt i,j; 582 PetscInt *a2anew,k,k1,jmin,jmax,*jl,*il,vj,nexti,ili; 583 PetscInt bs =A->rmap->bs,bs2 = a->bs2; 584 MatScalar *ba = b->a,*aa,*ap,*dk,*uik; 585 MatScalar *u,*diag,*rtmp,*rtmp_ptr; 586 MatScalar *work; 587 PetscInt *pivots; 588 589 PetscFunctionBegin; 590 /* initialization */ 591 ierr = PetscCalloc1(bs2*mbs,&rtmp);CHKERRQ(ierr); 592 ierr = PetscMalloc2(mbs,&il,mbs,&jl);CHKERRQ(ierr); 593 for (i=0; i<mbs; i++) { 594 jl[i] = mbs; il[0] = 0; 595 } 596 ierr = PetscMalloc3(bs2,&dk,bs2,&uik,bs,&work);CHKERRQ(ierr); 597 ierr = PetscMalloc1(bs,&pivots);CHKERRQ(ierr); 598 599 ierr = ISGetIndices(perm,&perm_ptr);CHKERRQ(ierr); 600 601 /* check permutation */ 602 if (!a->permute) { 603 ai = a->i; aj = a->j; aa = a->a; 604 } else { 605 ai = a->inew; aj = a->jnew; 606 ierr = PetscMalloc1(bs2*ai[mbs],&aa);CHKERRQ(ierr); 607 ierr = PetscMemcpy(aa,a->a,bs2*ai[mbs]*sizeof(MatScalar));CHKERRQ(ierr); 608 ierr = PetscMalloc1(ai[mbs],&a2anew);CHKERRQ(ierr); 609 ierr = PetscMemcpy(a2anew,a->a2anew,(ai[mbs])*sizeof(PetscInt));CHKERRQ(ierr); 610 611 for (i=0; i<mbs; i++) { 612 jmin = ai[i]; jmax = ai[i+1]; 613 for (j=jmin; j<jmax; j++) { 614 while (a2anew[j] != j) { 615 k = a2anew[j]; a2anew[j] = a2anew[k]; a2anew[k] = k; 616 for (k1=0; k1<bs2; k1++) { 617 dk[k1] = aa[k*bs2+k1]; 618 aa[k*bs2+k1] = aa[j*bs2+k1]; 619 aa[j*bs2+k1] = dk[k1]; 620 } 621 } 622 /* transform columnoriented blocks that lie in the lower triangle to roworiented blocks */ 623 if (i > aj[j]) { 624 /* printf("change orientation, row: %d, col: %d\n",i,aj[j]); */ 625 ap = aa + j*bs2; /* ptr to the beginning of j-th block of aa */ 626 for (k=0; k<bs2; k++) dk[k] = ap[k]; /* dk <- j-th block of aa */ 627 for (k=0; k<bs; k++) { /* j-th block of aa <- dk^T */ 628 for (k1=0; k1<bs; k1++) *ap++ = dk[k + bs*k1]; 629 } 630 } 631 } 632 } 633 ierr = PetscFree(a2anew);CHKERRQ(ierr); 634 } 635 636 /* for each row k */ 637 for (k = 0; k<mbs; k++) { 638 639 /*initialize k-th row with elements nonzero in row perm(k) of A */ 640 jmin = ai[perm_ptr[k]]; jmax = ai[perm_ptr[k]+1]; 641 642 ap = aa + jmin*bs2; 643 for (j = jmin; j < jmax; j++) { 644 vj = perm_ptr[aj[j]]; /* block col. index */ 645 rtmp_ptr = rtmp + vj*bs2; 646 for (i=0; i<bs2; i++) *rtmp_ptr++ = *ap++; 647 } 648 649 /* modify k-th row by adding in those rows i with U(i,k) != 0 */ 650 ierr = PetscMemcpy(dk,rtmp+k*bs2,bs2*sizeof(MatScalar));CHKERRQ(ierr); 651 i = jl[k]; /* first row to be added to k_th row */ 652 653 while (i < k) { 654 nexti = jl[i]; /* next row to be added to k_th row */ 655 656 /* compute multiplier */ 657 ili = il[i]; /* index of first nonzero element in U(i,k:bms-1) */ 658 659 /* uik = -inv(Di)*U_bar(i,k) */ 660 diag = ba + i*bs2; 661 u = ba + ili*bs2; 662 ierr = PetscMemzero(uik,bs2*sizeof(MatScalar));CHKERRQ(ierr); 663 PetscKernel_A_gets_A_minus_B_times_C(bs,uik,diag,u); 664 665 /* update D(k) += -U(i,k)^T * U_bar(i,k) */ 666 PetscKernel_A_gets_A_plus_Btranspose_times_C(bs,dk,uik,u); 667 ierr = PetscLogFlops(4.0*bs*bs2);CHKERRQ(ierr); 668 669 /* update -U(i,k) */ 670 ierr = PetscMemcpy(ba+ili*bs2,uik,bs2*sizeof(MatScalar));CHKERRQ(ierr); 671 672 /* add multiple of row i to k-th row ... */ 673 jmin = ili + 1; jmax = bi[i+1]; 674 if (jmin < jmax) { 675 for (j=jmin; j<jmax; j++) { 676 /* rtmp += -U(i,k)^T * U_bar(i,j) */ 677 rtmp_ptr = rtmp + bj[j]*bs2; 678 u = ba + j*bs2; 679 PetscKernel_A_gets_A_plus_Btranspose_times_C(bs,rtmp_ptr,uik,u); 680 } 681 ierr = PetscLogFlops(2.0*bs*bs2*(jmax-jmin));CHKERRQ(ierr); 682 683 /* ... add i to row list for next nonzero entry */ 684 il[i] = jmin; /* update il(i) in column k+1, ... mbs-1 */ 685 j = bj[jmin]; 686 jl[i] = jl[j]; jl[j] = i; /* update jl */ 687 } 688 i = nexti; 689 } 690 691 /* save nonzero entries in k-th row of U ... */ 692 693 /* invert diagonal block */ 694 diag = ba+k*bs2; 695 ierr = PetscMemcpy(diag,dk,bs2*sizeof(MatScalar));CHKERRQ(ierr); 696 ierr = PetscKernel_A_gets_inverse_A(bs,diag,pivots,work);CHKERRQ(ierr); 697 698 jmin = bi[k]; jmax = bi[k+1]; 699 if (jmin < jmax) { 700 for (j=jmin; j<jmax; j++) { 701 vj = bj[j]; /* block col. index of U */ 702 u = ba + j*bs2; 703 rtmp_ptr = rtmp + vj*bs2; 704 for (k1=0; k1<bs2; k1++) { 705 *u++ = *rtmp_ptr; 706 *rtmp_ptr++ = 0.0; 707 } 708 } 709 710 /* ... add k to row list for first nonzero entry in k-th row */ 711 il[k] = jmin; 712 i = bj[jmin]; 713 jl[k] = jl[i]; jl[i] = k; 714 } 715 } 716 717 ierr = PetscFree(rtmp);CHKERRQ(ierr); 718 ierr = PetscFree2(il,jl);CHKERRQ(ierr); 719 ierr = PetscFree3(dk,uik,work);CHKERRQ(ierr); 720 ierr = PetscFree(pivots);CHKERRQ(ierr); 721 if (a->permute) { 722 ierr = PetscFree(aa);CHKERRQ(ierr); 723 } 724 725 ierr = ISRestoreIndices(perm,&perm_ptr);CHKERRQ(ierr); 726 727 C->ops->solve = MatSolve_SeqSBAIJ_N_inplace; 728 C->ops->solvetranspose = MatSolve_SeqSBAIJ_N_inplace; 729 C->ops->forwardsolve = MatForwardSolve_SeqSBAIJ_N_inplace; 730 C->ops->backwardsolve = MatBackwardSolve_SeqSBAIJ_N_inplace; 731 732 C->assembled = PETSC_TRUE; 733 C->preallocated = PETSC_TRUE; 734 735 ierr = PetscLogFlops(1.3333*bs*bs2*b->mbs);CHKERRQ(ierr); /* from inverting diagonal blocks */ 736 PetscFunctionReturn(0); 737 } 738 739 #undef __FUNCT__ 740 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_N_NaturalOrdering" 741 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_N_NaturalOrdering(Mat C,Mat A,const MatFactorInfo *info) 742 { 743 Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data,*b = (Mat_SeqSBAIJ*)C->data; 744 PetscErrorCode ierr; 745 PetscInt i,j,mbs=a->mbs,*bi=b->i,*bj=b->j; 746 PetscInt *ai,*aj,k,k1,jmin,jmax,*jl,*il,vj,nexti,ili; 747 PetscInt bs =A->rmap->bs,bs2 = a->bs2; 748 MatScalar *ba = b->a,*aa,*ap,*dk,*uik; 749 MatScalar *u,*diag,*rtmp,*rtmp_ptr; 750 MatScalar *work; 751 PetscInt *pivots; 752 753 PetscFunctionBegin; 754 ierr = PetscCalloc1(bs2*mbs,&rtmp);CHKERRQ(ierr); 755 ierr = PetscMalloc2(mbs,&il,mbs,&jl);CHKERRQ(ierr); 756 for (i=0; i<mbs; i++) { 757 jl[i] = mbs; il[0] = 0; 758 } 759 ierr = PetscMalloc3(bs2,&dk,bs2,&uik,bs,&work);CHKERRQ(ierr); 760 ierr = PetscMalloc1(bs,&pivots);CHKERRQ(ierr); 761 762 ai = a->i; aj = a->j; aa = a->a; 763 764 /* for each row k */ 765 for (k = 0; k<mbs; k++) { 766 767 /*initialize k-th row with elements nonzero in row k of A */ 768 jmin = ai[k]; jmax = ai[k+1]; 769 ap = aa + jmin*bs2; 770 for (j = jmin; j < jmax; j++) { 771 vj = aj[j]; /* block col. index */ 772 rtmp_ptr = rtmp + vj*bs2; 773 for (i=0; i<bs2; i++) *rtmp_ptr++ = *ap++; 774 } 775 776 /* modify k-th row by adding in those rows i with U(i,k) != 0 */ 777 ierr = PetscMemcpy(dk,rtmp+k*bs2,bs2*sizeof(MatScalar));CHKERRQ(ierr); 778 i = jl[k]; /* first row to be added to k_th row */ 779 780 while (i < k) { 781 nexti = jl[i]; /* next row to be added to k_th row */ 782 783 /* compute multiplier */ 784 ili = il[i]; /* index of first nonzero element in U(i,k:bms-1) */ 785 786 /* uik = -inv(Di)*U_bar(i,k) */ 787 diag = ba + i*bs2; 788 u = ba + ili*bs2; 789 ierr = PetscMemzero(uik,bs2*sizeof(MatScalar));CHKERRQ(ierr); 790 PetscKernel_A_gets_A_minus_B_times_C(bs,uik,diag,u); 791 792 /* update D(k) += -U(i,k)^T * U_bar(i,k) */ 793 PetscKernel_A_gets_A_plus_Btranspose_times_C(bs,dk,uik,u); 794 ierr = PetscLogFlops(2.0*bs*bs2);CHKERRQ(ierr); 795 796 /* update -U(i,k) */ 797 ierr = PetscMemcpy(ba+ili*bs2,uik,bs2*sizeof(MatScalar));CHKERRQ(ierr); 798 799 /* add multiple of row i to k-th row ... */ 800 jmin = ili + 1; jmax = bi[i+1]; 801 if (jmin < jmax) { 802 for (j=jmin; j<jmax; j++) { 803 /* rtmp += -U(i,k)^T * U_bar(i,j) */ 804 rtmp_ptr = rtmp + bj[j]*bs2; 805 u = ba + j*bs2; 806 PetscKernel_A_gets_A_plus_Btranspose_times_C(bs,rtmp_ptr,uik,u); 807 } 808 ierr = PetscLogFlops(2.0*bs*bs2*(jmax-jmin));CHKERRQ(ierr); 809 810 /* ... add i to row list for next nonzero entry */ 811 il[i] = jmin; /* update il(i) in column k+1, ... mbs-1 */ 812 j = bj[jmin]; 813 jl[i] = jl[j]; jl[j] = i; /* update jl */ 814 } 815 i = nexti; 816 } 817 818 /* save nonzero entries in k-th row of U ... */ 819 820 /* invert diagonal block */ 821 diag = ba+k*bs2; 822 ierr = PetscMemcpy(diag,dk,bs2*sizeof(MatScalar));CHKERRQ(ierr); 823 ierr = PetscKernel_A_gets_inverse_A(bs,diag,pivots,work);CHKERRQ(ierr); 824 825 jmin = bi[k]; jmax = bi[k+1]; 826 if (jmin < jmax) { 827 for (j=jmin; j<jmax; j++) { 828 vj = bj[j]; /* block col. index of U */ 829 u = ba + j*bs2; 830 rtmp_ptr = rtmp + vj*bs2; 831 for (k1=0; k1<bs2; k1++) { 832 *u++ = *rtmp_ptr; 833 *rtmp_ptr++ = 0.0; 834 } 835 } 836 837 /* ... add k to row list for first nonzero entry in k-th row */ 838 il[k] = jmin; 839 i = bj[jmin]; 840 jl[k] = jl[i]; jl[i] = k; 841 } 842 } 843 844 ierr = PetscFree(rtmp);CHKERRQ(ierr); 845 ierr = PetscFree2(il,jl);CHKERRQ(ierr); 846 ierr = PetscFree3(dk,uik,work);CHKERRQ(ierr); 847 ierr = PetscFree(pivots);CHKERRQ(ierr); 848 849 C->ops->solve = MatSolve_SeqSBAIJ_N_NaturalOrdering_inplace; 850 C->ops->solvetranspose = MatSolve_SeqSBAIJ_N_NaturalOrdering_inplace; 851 C->ops->forwardsolve = MatForwardSolve_SeqSBAIJ_N_NaturalOrdering_inplace; 852 C->ops->backwardsolve = MatBackwardSolve_SeqSBAIJ_N_NaturalOrdering_inplace; 853 C->assembled = PETSC_TRUE; 854 C->preallocated = PETSC_TRUE; 855 856 ierr = PetscLogFlops(1.3333*bs*bs2*b->mbs);CHKERRQ(ierr); /* from inverting diagonal blocks */ 857 PetscFunctionReturn(0); 858 } 859 860 /* 861 Numeric U^T*D*U factorization for SBAIJ format. Modified from SNF of YSMP. 862 Version for blocks 2 by 2. 863 */ 864 #undef __FUNCT__ 865 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_2" 866 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_2(Mat C,Mat A,const MatFactorInfo *info) 867 { 868 Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data,*b = (Mat_SeqSBAIJ*)C->data; 869 IS perm = b->row; 870 PetscErrorCode ierr; 871 const PetscInt *ai,*aj,*perm_ptr; 872 PetscInt i,j,mbs=a->mbs,*bi=b->i,*bj=b->j; 873 PetscInt *a2anew,k,k1,jmin,jmax,*jl,*il,vj,nexti,ili; 874 MatScalar *ba = b->a,*aa,*ap; 875 MatScalar *u,*diag,*rtmp,*rtmp_ptr,dk[4],uik[4]; 876 PetscReal shift = info->shiftamount; 877 878 PetscFunctionBegin; 879 /* initialization */ 880 /* il and jl record the first nonzero element in each row of the accessing 881 window U(0:k, k:mbs-1). 882 jl: list of rows to be added to uneliminated rows 883 i>= k: jl(i) is the first row to be added to row i 884 i< k: jl(i) is the row following row i in some list of rows 885 jl(i) = mbs indicates the end of a list 886 il(i): points to the first nonzero element in columns k,...,mbs-1 of 887 row i of U */ 888 ierr = PetscCalloc1(4*mbs,&rtmp);CHKERRQ(ierr); 889 ierr = PetscMalloc2(mbs,&il,mbs,&jl);CHKERRQ(ierr); 890 for (i=0; i<mbs; i++) { 891 jl[i] = mbs; il[0] = 0; 892 } 893 ierr = ISGetIndices(perm,&perm_ptr);CHKERRQ(ierr); 894 895 /* check permutation */ 896 if (!a->permute) { 897 ai = a->i; aj = a->j; aa = a->a; 898 } else { 899 ai = a->inew; aj = a->jnew; 900 ierr = PetscMalloc1(4*ai[mbs],&aa);CHKERRQ(ierr); 901 ierr = PetscMemcpy(aa,a->a,4*ai[mbs]*sizeof(MatScalar));CHKERRQ(ierr); 902 ierr = PetscMalloc1(ai[mbs],&a2anew);CHKERRQ(ierr); 903 ierr = PetscMemcpy(a2anew,a->a2anew,(ai[mbs])*sizeof(PetscInt));CHKERRQ(ierr); 904 905 for (i=0; i<mbs; i++) { 906 jmin = ai[i]; jmax = ai[i+1]; 907 for (j=jmin; j<jmax; j++) { 908 while (a2anew[j] != j) { 909 k = a2anew[j]; a2anew[j] = a2anew[k]; a2anew[k] = k; 910 for (k1=0; k1<4; k1++) { 911 dk[k1] = aa[k*4+k1]; 912 aa[k*4+k1] = aa[j*4+k1]; 913 aa[j*4+k1] = dk[k1]; 914 } 915 } 916 /* transform columnoriented blocks that lie in the lower triangle to roworiented blocks */ 917 if (i > aj[j]) { 918 /* printf("change orientation, row: %d, col: %d\n",i,aj[j]); */ 919 ap = aa + j*4; /* ptr to the beginning of the block */ 920 dk[1] = ap[1]; /* swap ap[1] and ap[2] */ 921 ap[1] = ap[2]; 922 ap[2] = dk[1]; 923 } 924 } 925 } 926 ierr = PetscFree(a2anew);CHKERRQ(ierr); 927 } 928 929 /* for each row k */ 930 for (k = 0; k<mbs; k++) { 931 932 /*initialize k-th row with elements nonzero in row perm(k) of A */ 933 jmin = ai[perm_ptr[k]]; jmax = ai[perm_ptr[k]+1]; 934 ap = aa + jmin*4; 935 for (j = jmin; j < jmax; j++) { 936 vj = perm_ptr[aj[j]]; /* block col. index */ 937 rtmp_ptr = rtmp + vj*4; 938 for (i=0; i<4; i++) *rtmp_ptr++ = *ap++; 939 } 940 941 /* modify k-th row by adding in those rows i with U(i,k) != 0 */ 942 ierr = PetscMemcpy(dk,rtmp+k*4,4*sizeof(MatScalar));CHKERRQ(ierr); 943 i = jl[k]; /* first row to be added to k_th row */ 944 945 while (i < k) { 946 nexti = jl[i]; /* next row to be added to k_th row */ 947 948 /* compute multiplier */ 949 ili = il[i]; /* index of first nonzero element in U(i,k:bms-1) */ 950 951 /* uik = -inv(Di)*U_bar(i,k): - ba[ili]*ba[i] */ 952 diag = ba + i*4; 953 u = ba + ili*4; 954 uik[0] = -(diag[0]*u[0] + diag[2]*u[1]); 955 uik[1] = -(diag[1]*u[0] + diag[3]*u[1]); 956 uik[2] = -(diag[0]*u[2] + diag[2]*u[3]); 957 uik[3] = -(diag[1]*u[2] + diag[3]*u[3]); 958 959 /* update D(k) += -U(i,k)^T * U_bar(i,k): dk += uik*ba[ili] */ 960 dk[0] += uik[0]*u[0] + uik[1]*u[1]; 961 dk[1] += uik[2]*u[0] + uik[3]*u[1]; 962 dk[2] += uik[0]*u[2] + uik[1]*u[3]; 963 dk[3] += uik[2]*u[2] + uik[3]*u[3]; 964 965 ierr = PetscLogFlops(16.0*2.0);CHKERRQ(ierr); 966 967 /* update -U(i,k): ba[ili] = uik */ 968 ierr = PetscMemcpy(ba+ili*4,uik,4*sizeof(MatScalar));CHKERRQ(ierr); 969 970 /* add multiple of row i to k-th row ... */ 971 jmin = ili + 1; jmax = bi[i+1]; 972 if (jmin < jmax) { 973 for (j=jmin; j<jmax; j++) { 974 /* rtmp += -U(i,k)^T * U_bar(i,j): rtmp[bj[j]] += uik*ba[j]; */ 975 rtmp_ptr = rtmp + bj[j]*4; 976 u = ba + j*4; 977 rtmp_ptr[0] += uik[0]*u[0] + uik[1]*u[1]; 978 rtmp_ptr[1] += uik[2]*u[0] + uik[3]*u[1]; 979 rtmp_ptr[2] += uik[0]*u[2] + uik[1]*u[3]; 980 rtmp_ptr[3] += uik[2]*u[2] + uik[3]*u[3]; 981 } 982 ierr = PetscLogFlops(16.0*(jmax-jmin));CHKERRQ(ierr); 983 984 /* ... add i to row list for next nonzero entry */ 985 il[i] = jmin; /* update il(i) in column k+1, ... mbs-1 */ 986 j = bj[jmin]; 987 jl[i] = jl[j]; jl[j] = i; /* update jl */ 988 } 989 i = nexti; 990 } 991 992 /* save nonzero entries in k-th row of U ... */ 993 994 /* invert diagonal block */ 995 diag = ba+k*4; 996 ierr = PetscMemcpy(diag,dk,4*sizeof(MatScalar));CHKERRQ(ierr); 997 ierr = PetscKernel_A_gets_inverse_A_2(diag,shift);CHKERRQ(ierr); 998 999 jmin = bi[k]; jmax = bi[k+1]; 1000 if (jmin < jmax) { 1001 for (j=jmin; j<jmax; j++) { 1002 vj = bj[j]; /* block col. index of U */ 1003 u = ba + j*4; 1004 rtmp_ptr = rtmp + vj*4; 1005 for (k1=0; k1<4; k1++) { 1006 *u++ = *rtmp_ptr; 1007 *rtmp_ptr++ = 0.0; 1008 } 1009 } 1010 1011 /* ... add k to row list for first nonzero entry in k-th row */ 1012 il[k] = jmin; 1013 i = bj[jmin]; 1014 jl[k] = jl[i]; jl[i] = k; 1015 } 1016 } 1017 1018 ierr = PetscFree(rtmp);CHKERRQ(ierr); 1019 ierr = PetscFree2(il,jl);CHKERRQ(ierr); 1020 if (a->permute) { 1021 ierr = PetscFree(aa);CHKERRQ(ierr); 1022 } 1023 ierr = ISRestoreIndices(perm,&perm_ptr);CHKERRQ(ierr); 1024 1025 C->ops->solve = MatSolve_SeqSBAIJ_2_inplace; 1026 C->ops->solvetranspose = MatSolve_SeqSBAIJ_2_inplace; 1027 C->assembled = PETSC_TRUE; 1028 C->preallocated = PETSC_TRUE; 1029 1030 ierr = PetscLogFlops(1.3333*8*b->mbs);CHKERRQ(ierr); /* from inverting diagonal blocks */ 1031 PetscFunctionReturn(0); 1032 } 1033 1034 /* 1035 Version for when blocks are 2 by 2 Using natural ordering 1036 */ 1037 #undef __FUNCT__ 1038 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_2_NaturalOrdering" 1039 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_2_NaturalOrdering(Mat C,Mat A,const MatFactorInfo *info) 1040 { 1041 Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data,*b = (Mat_SeqSBAIJ*)C->data; 1042 PetscErrorCode ierr; 1043 PetscInt i,j,mbs=a->mbs,*bi=b->i,*bj=b->j; 1044 PetscInt *ai,*aj,k,k1,jmin,jmax,*jl,*il,vj,nexti,ili; 1045 MatScalar *ba = b->a,*aa,*ap,dk[8],uik[8]; 1046 MatScalar *u,*diag,*rtmp,*rtmp_ptr; 1047 PetscReal shift = info->shiftamount; 1048 1049 PetscFunctionBegin; 1050 /* initialization */ 1051 /* il and jl record the first nonzero element in each row of the accessing 1052 window U(0:k, k:mbs-1). 1053 jl: list of rows to be added to uneliminated rows 1054 i>= k: jl(i) is the first row to be added to row i 1055 i< k: jl(i) is the row following row i in some list of rows 1056 jl(i) = mbs indicates the end of a list 1057 il(i): points to the first nonzero element in columns k,...,mbs-1 of 1058 row i of U */ 1059 ierr = PetscCalloc1(4*mbs,&rtmp);CHKERRQ(ierr); 1060 ierr = PetscMalloc2(mbs,&il,mbs,&jl);CHKERRQ(ierr); 1061 for (i=0; i<mbs; i++) { 1062 jl[i] = mbs; il[0] = 0; 1063 } 1064 ai = a->i; aj = a->j; aa = a->a; 1065 1066 /* for each row k */ 1067 for (k = 0; k<mbs; k++) { 1068 1069 /*initialize k-th row with elements nonzero in row k of A */ 1070 jmin = ai[k]; jmax = ai[k+1]; 1071 ap = aa + jmin*4; 1072 for (j = jmin; j < jmax; j++) { 1073 vj = aj[j]; /* block col. index */ 1074 rtmp_ptr = rtmp + vj*4; 1075 for (i=0; i<4; i++) *rtmp_ptr++ = *ap++; 1076 } 1077 1078 /* modify k-th row by adding in those rows i with U(i,k) != 0 */ 1079 ierr = PetscMemcpy(dk,rtmp+k*4,4*sizeof(MatScalar));CHKERRQ(ierr); 1080 i = jl[k]; /* first row to be added to k_th row */ 1081 1082 while (i < k) { 1083 nexti = jl[i]; /* next row to be added to k_th row */ 1084 1085 /* compute multiplier */ 1086 ili = il[i]; /* index of first nonzero element in U(i,k:bms-1) */ 1087 1088 /* uik = -inv(Di)*U_bar(i,k): - ba[ili]*ba[i] */ 1089 diag = ba + i*4; 1090 u = ba + ili*4; 1091 uik[0] = -(diag[0]*u[0] + diag[2]*u[1]); 1092 uik[1] = -(diag[1]*u[0] + diag[3]*u[1]); 1093 uik[2] = -(diag[0]*u[2] + diag[2]*u[3]); 1094 uik[3] = -(diag[1]*u[2] + diag[3]*u[3]); 1095 1096 /* update D(k) += -U(i,k)^T * U_bar(i,k): dk += uik*ba[ili] */ 1097 dk[0] += uik[0]*u[0] + uik[1]*u[1]; 1098 dk[1] += uik[2]*u[0] + uik[3]*u[1]; 1099 dk[2] += uik[0]*u[2] + uik[1]*u[3]; 1100 dk[3] += uik[2]*u[2] + uik[3]*u[3]; 1101 1102 ierr = PetscLogFlops(16.0*2.0);CHKERRQ(ierr); 1103 1104 /* update -U(i,k): ba[ili] = uik */ 1105 ierr = PetscMemcpy(ba+ili*4,uik,4*sizeof(MatScalar));CHKERRQ(ierr); 1106 1107 /* add multiple of row i to k-th row ... */ 1108 jmin = ili + 1; jmax = bi[i+1]; 1109 if (jmin < jmax) { 1110 for (j=jmin; j<jmax; j++) { 1111 /* rtmp += -U(i,k)^T * U_bar(i,j): rtmp[bj[j]] += uik*ba[j]; */ 1112 rtmp_ptr = rtmp + bj[j]*4; 1113 u = ba + j*4; 1114 rtmp_ptr[0] += uik[0]*u[0] + uik[1]*u[1]; 1115 rtmp_ptr[1] += uik[2]*u[0] + uik[3]*u[1]; 1116 rtmp_ptr[2] += uik[0]*u[2] + uik[1]*u[3]; 1117 rtmp_ptr[3] += uik[2]*u[2] + uik[3]*u[3]; 1118 } 1119 ierr = PetscLogFlops(16.0*(jmax-jmin));CHKERRQ(ierr); 1120 1121 /* ... add i to row list for next nonzero entry */ 1122 il[i] = jmin; /* update il(i) in column k+1, ... mbs-1 */ 1123 j = bj[jmin]; 1124 jl[i] = jl[j]; jl[j] = i; /* update jl */ 1125 } 1126 i = nexti; 1127 } 1128 1129 /* save nonzero entries in k-th row of U ... */ 1130 1131 /* invert diagonal block */ 1132 diag = ba+k*4; 1133 ierr = PetscMemcpy(diag,dk,4*sizeof(MatScalar));CHKERRQ(ierr); 1134 ierr = PetscKernel_A_gets_inverse_A_2(diag,shift);CHKERRQ(ierr); 1135 1136 jmin = bi[k]; jmax = bi[k+1]; 1137 if (jmin < jmax) { 1138 for (j=jmin; j<jmax; j++) { 1139 vj = bj[j]; /* block col. index of U */ 1140 u = ba + j*4; 1141 rtmp_ptr = rtmp + vj*4; 1142 for (k1=0; k1<4; k1++) { 1143 *u++ = *rtmp_ptr; 1144 *rtmp_ptr++ = 0.0; 1145 } 1146 } 1147 1148 /* ... add k to row list for first nonzero entry in k-th row */ 1149 il[k] = jmin; 1150 i = bj[jmin]; 1151 jl[k] = jl[i]; jl[i] = k; 1152 } 1153 } 1154 1155 ierr = PetscFree(rtmp);CHKERRQ(ierr); 1156 ierr = PetscFree2(il,jl);CHKERRQ(ierr); 1157 1158 C->ops->solve = MatSolve_SeqSBAIJ_2_NaturalOrdering_inplace; 1159 C->ops->solvetranspose = MatSolve_SeqSBAIJ_2_NaturalOrdering_inplace; 1160 C->ops->forwardsolve = MatForwardSolve_SeqSBAIJ_2_NaturalOrdering_inplace; 1161 C->ops->backwardsolve = MatBackwardSolve_SeqSBAIJ_2_NaturalOrdering_inplace; 1162 C->assembled = PETSC_TRUE; 1163 C->preallocated = PETSC_TRUE; 1164 1165 ierr = PetscLogFlops(1.3333*8*b->mbs);CHKERRQ(ierr); /* from inverting diagonal blocks */ 1166 PetscFunctionReturn(0); 1167 } 1168 1169 /* 1170 Numeric U^T*D*U factorization for SBAIJ format. 1171 Version for blocks are 1 by 1. 1172 */ 1173 #undef __FUNCT__ 1174 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_1_inplace" 1175 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_1_inplace(Mat C,Mat A,const MatFactorInfo *info) 1176 { 1177 Mat_SeqSBAIJ *a=(Mat_SeqSBAIJ*)A->data,*b=(Mat_SeqSBAIJ*)C->data; 1178 IS ip=b->row; 1179 PetscErrorCode ierr; 1180 const PetscInt *ai,*aj,*rip; 1181 PetscInt *a2anew,i,j,mbs=a->mbs,*bi=b->i,*bj=b->j,*bcol; 1182 PetscInt k,jmin,jmax,*jl,*il,col,nexti,ili,nz; 1183 MatScalar *rtmp,*ba=b->a,*bval,*aa,dk,uikdi; 1184 PetscReal rs; 1185 FactorShiftCtx sctx; 1186 1187 PetscFunctionBegin; 1188 /* MatPivotSetUp(): initialize shift context sctx */ 1189 ierr = PetscMemzero(&sctx,sizeof(FactorShiftCtx));CHKERRQ(ierr); 1190 1191 ierr = ISGetIndices(ip,&rip);CHKERRQ(ierr); 1192 if (!a->permute) { 1193 ai = a->i; aj = a->j; aa = a->a; 1194 } else { 1195 ai = a->inew; aj = a->jnew; 1196 nz = ai[mbs]; 1197 ierr = PetscMalloc1(nz,&aa);CHKERRQ(ierr); 1198 a2anew = a->a2anew; 1199 bval = a->a; 1200 for (j=0; j<nz; j++) { 1201 aa[a2anew[j]] = *(bval++); 1202 } 1203 } 1204 1205 /* initialization */ 1206 /* il and jl record the first nonzero element in each row of the accessing 1207 window U(0:k, k:mbs-1). 1208 jl: list of rows to be added to uneliminated rows 1209 i>= k: jl(i) is the first row to be added to row i 1210 i< k: jl(i) is the row following row i in some list of rows 1211 jl(i) = mbs indicates the end of a list 1212 il(i): points to the first nonzero element in columns k,...,mbs-1 of 1213 row i of U */ 1214 ierr = PetscMalloc3(mbs,&rtmp,mbs,&il,mbs,&jl);CHKERRQ(ierr); 1215 1216 do { 1217 sctx.newshift = PETSC_FALSE; 1218 for (i=0; i<mbs; i++) { 1219 rtmp[i] = 0.0; jl[i] = mbs; il[0] = 0; 1220 } 1221 1222 for (k = 0; k<mbs; k++) { 1223 /*initialize k-th row by the perm[k]-th row of A */ 1224 jmin = ai[rip[k]]; jmax = ai[rip[k]+1]; 1225 bval = ba + bi[k]; 1226 for (j = jmin; j < jmax; j++) { 1227 col = rip[aj[j]]; 1228 rtmp[col] = aa[j]; 1229 *bval++ = 0.0; /* for in-place factorization */ 1230 } 1231 1232 /* shift the diagonal of the matrix */ 1233 if (sctx.nshift) rtmp[k] += sctx.shift_amount; 1234 1235 /* modify k-th row by adding in those rows i with U(i,k)!=0 */ 1236 dk = rtmp[k]; 1237 i = jl[k]; /* first row to be added to k_th row */ 1238 1239 while (i < k) { 1240 nexti = jl[i]; /* next row to be added to k_th row */ 1241 1242 /* compute multiplier, update diag(k) and U(i,k) */ 1243 ili = il[i]; /* index of first nonzero element in U(i,k:bms-1) */ 1244 uikdi = -ba[ili]*ba[bi[i]]; /* diagonal(k) */ 1245 dk += uikdi*ba[ili]; 1246 ba[ili] = uikdi; /* -U(i,k) */ 1247 1248 /* add multiple of row i to k-th row */ 1249 jmin = ili + 1; jmax = bi[i+1]; 1250 if (jmin < jmax) { 1251 for (j=jmin; j<jmax; j++) rtmp[bj[j]] += uikdi*ba[j]; 1252 ierr = PetscLogFlops(2.0*(jmax-jmin));CHKERRQ(ierr); 1253 1254 /* update il and jl for row i */ 1255 il[i] = jmin; 1256 j = bj[jmin]; jl[i] = jl[j]; jl[j] = i; 1257 } 1258 i = nexti; 1259 } 1260 1261 /* shift the diagonals when zero pivot is detected */ 1262 /* compute rs=sum of abs(off-diagonal) */ 1263 rs = 0.0; 1264 jmin = bi[k]+1; 1265 nz = bi[k+1] - jmin; 1266 if (nz) { 1267 bcol = bj + jmin; 1268 while (nz--) { 1269 rs += PetscAbsScalar(rtmp[*bcol]); 1270 bcol++; 1271 } 1272 } 1273 1274 sctx.rs = rs; 1275 sctx.pv = dk; 1276 ierr = MatPivotCheck(A,info,&sctx,k);CHKERRQ(ierr); 1277 if (sctx.newshift) break; /* sctx.shift_amount is updated */ 1278 dk = sctx.pv; 1279 1280 /* copy data into U(k,:) */ 1281 ba[bi[k]] = 1.0/dk; /* U(k,k) */ 1282 jmin = bi[k]+1; jmax = bi[k+1]; 1283 if (jmin < jmax) { 1284 for (j=jmin; j<jmax; j++) { 1285 col = bj[j]; ba[j] = rtmp[col]; rtmp[col] = 0.0; 1286 } 1287 /* add the k-th row into il and jl */ 1288 il[k] = jmin; 1289 i = bj[jmin]; jl[k] = jl[i]; jl[i] = k; 1290 } 1291 } 1292 } while (sctx.newshift); 1293 ierr = PetscFree3(rtmp,il,jl);CHKERRQ(ierr); 1294 if (a->permute) {ierr = PetscFree(aa);CHKERRQ(ierr);} 1295 1296 ierr = ISRestoreIndices(ip,&rip);CHKERRQ(ierr); 1297 1298 C->ops->solve = MatSolve_SeqSBAIJ_1_inplace; 1299 C->ops->solves = MatSolves_SeqSBAIJ_1_inplace; 1300 C->ops->solvetranspose = MatSolve_SeqSBAIJ_1_inplace; 1301 C->ops->forwardsolve = MatForwardSolve_SeqSBAIJ_1_inplace; 1302 C->ops->backwardsolve = MatBackwardSolve_SeqSBAIJ_1_inplace; 1303 C->assembled = PETSC_TRUE; 1304 C->preallocated = PETSC_TRUE; 1305 1306 ierr = PetscLogFlops(C->rmap->N);CHKERRQ(ierr); 1307 if (sctx.nshift) { 1308 if (info->shifttype == (PetscReal)MAT_SHIFT_NONZERO) { 1309 ierr = PetscInfo2(A,"number of shiftnz tries %D, shift_amount %g\n",sctx.nshift,(double)sctx.shift_amount);CHKERRQ(ierr); 1310 } else if (info->shifttype == (PetscReal)MAT_SHIFT_POSITIVE_DEFINITE) { 1311 ierr = PetscInfo2(A,"number of shiftpd tries %D, shift_amount %g\n",sctx.nshift,(double)sctx.shift_amount);CHKERRQ(ierr); 1312 } 1313 } 1314 PetscFunctionReturn(0); 1315 } 1316 1317 /* 1318 Version for when blocks are 1 by 1 Using natural ordering under new datastructure 1319 Modified from MatCholeskyFactorNumeric_SeqAIJ() 1320 */ 1321 #undef __FUNCT__ 1322 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering" 1323 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering(Mat B,Mat A,const MatFactorInfo *info) 1324 { 1325 Mat_SeqSBAIJ *a=(Mat_SeqSBAIJ*)A->data; 1326 Mat_SeqSBAIJ *b=(Mat_SeqSBAIJ*)B->data; 1327 PetscErrorCode ierr; 1328 PetscInt i,j,mbs=A->rmap->n,*bi=b->i,*bj=b->j,*bdiag=b->diag,*bjtmp; 1329 PetscInt *ai=a->i,*aj=a->j,*ajtmp; 1330 PetscInt k,jmin,jmax,*c2r,*il,col,nexti,ili,nz; 1331 MatScalar *rtmp,*ba=b->a,*bval,*aa=a->a,dk,uikdi; 1332 FactorShiftCtx sctx; 1333 PetscReal rs; 1334 MatScalar d,*v; 1335 1336 PetscFunctionBegin; 1337 ierr = PetscMalloc3(mbs,&rtmp,mbs,&il,mbs,&c2r);CHKERRQ(ierr); 1338 1339 /* MatPivotSetUp(): initialize shift context sctx */ 1340 ierr = PetscMemzero(&sctx,sizeof(FactorShiftCtx));CHKERRQ(ierr); 1341 1342 if (info->shifttype == (PetscReal)MAT_SHIFT_POSITIVE_DEFINITE) { /* set sctx.shift_top=max{rs} */ 1343 sctx.shift_top = info->zeropivot; 1344 1345 ierr = PetscMemzero(rtmp,mbs*sizeof(MatScalar));CHKERRQ(ierr); 1346 1347 for (i=0; i<mbs; i++) { 1348 /* calculate sum(|aij|)-RealPart(aii), amt of shift needed for this row */ 1349 d = (aa)[a->diag[i]]; 1350 rtmp[i] += -PetscRealPart(d); /* diagonal entry */ 1351 ajtmp = aj + ai[i] + 1; /* exclude diagonal */ 1352 v = aa + ai[i] + 1; 1353 nz = ai[i+1] - ai[i] - 1; 1354 for (j=0; j<nz; j++) { 1355 rtmp[i] += PetscAbsScalar(v[j]); 1356 rtmp[ajtmp[j]] += PetscAbsScalar(v[j]); 1357 } 1358 if (PetscRealPart(rtmp[i]) > sctx.shift_top) sctx.shift_top = PetscRealPart(rtmp[i]); 1359 } 1360 sctx.shift_top *= 1.1; 1361 sctx.nshift_max = 5; 1362 sctx.shift_lo = 0.; 1363 sctx.shift_hi = 1.; 1364 } 1365 1366 /* allocate working arrays 1367 c2r: linked list, keep track of pivot rows for a given column. c2r[col]: head of the list for a given col 1368 il: for active k row, il[i] gives the index of the 1st nonzero entry in U[i,k:n-1] in bj and ba arrays 1369 */ 1370 do { 1371 sctx.newshift = PETSC_FALSE; 1372 1373 for (i=0; i<mbs; i++) c2r[i] = mbs; 1374 if (mbs) il[0] = 0; 1375 1376 for (k = 0; k<mbs; k++) { 1377 /* zero rtmp */ 1378 nz = bi[k+1] - bi[k]; 1379 bjtmp = bj + bi[k]; 1380 for (j=0; j<nz; j++) rtmp[bjtmp[j]] = 0.0; 1381 1382 /* load in initial unfactored row */ 1383 bval = ba + bi[k]; 1384 jmin = ai[k]; jmax = ai[k+1]; 1385 for (j = jmin; j < jmax; j++) { 1386 col = aj[j]; 1387 rtmp[col] = aa[j]; 1388 *bval++ = 0.0; /* for in-place factorization */ 1389 } 1390 /* shift the diagonal of the matrix: ZeropivotApply() */ 1391 rtmp[k] += sctx.shift_amount; /* shift the diagonal of the matrix */ 1392 1393 /* modify k-th row by adding in those rows i with U(i,k)!=0 */ 1394 dk = rtmp[k]; 1395 i = c2r[k]; /* first row to be added to k_th row */ 1396 1397 while (i < k) { 1398 nexti = c2r[i]; /* next row to be added to k_th row */ 1399 1400 /* compute multiplier, update diag(k) and U(i,k) */ 1401 ili = il[i]; /* index of first nonzero element in U(i,k:bms-1) */ 1402 uikdi = -ba[ili]*ba[bdiag[i]]; /* diagonal(k) */ 1403 dk += uikdi*ba[ili]; /* update diag[k] */ 1404 ba[ili] = uikdi; /* -U(i,k) */ 1405 1406 /* add multiple of row i to k-th row */ 1407 jmin = ili + 1; jmax = bi[i+1]; 1408 if (jmin < jmax) { 1409 for (j=jmin; j<jmax; j++) rtmp[bj[j]] += uikdi*ba[j]; 1410 /* update il and c2r for row i */ 1411 il[i] = jmin; 1412 j = bj[jmin]; c2r[i] = c2r[j]; c2r[j] = i; 1413 } 1414 i = nexti; 1415 } 1416 1417 /* copy data into U(k,:) */ 1418 rs = 0.0; 1419 jmin = bi[k]; jmax = bi[k+1]-1; 1420 if (jmin < jmax) { 1421 for (j=jmin; j<jmax; j++) { 1422 col = bj[j]; ba[j] = rtmp[col]; rs += PetscAbsScalar(ba[j]); 1423 } 1424 /* add the k-th row into il and c2r */ 1425 il[k] = jmin; 1426 i = bj[jmin]; c2r[k] = c2r[i]; c2r[i] = k; 1427 } 1428 1429 sctx.rs = rs; 1430 sctx.pv = dk; 1431 ierr = MatPivotCheck(A,info,&sctx,k);CHKERRQ(ierr); 1432 if (sctx.newshift) break; 1433 dk = sctx.pv; 1434 1435 ba[bdiag[k]] = 1.0/dk; /* U(k,k) */ 1436 } 1437 } while (sctx.newshift); 1438 1439 ierr = PetscFree3(rtmp,il,c2r);CHKERRQ(ierr); 1440 1441 B->ops->solve = MatSolve_SeqSBAIJ_1_NaturalOrdering; 1442 B->ops->solves = MatSolves_SeqSBAIJ_1; 1443 B->ops->solvetranspose = MatSolve_SeqSBAIJ_1_NaturalOrdering; 1444 B->ops->forwardsolve = MatForwardSolve_SeqSBAIJ_1_NaturalOrdering; 1445 B->ops->backwardsolve = MatBackwardSolve_SeqSBAIJ_1_NaturalOrdering; 1446 1447 B->assembled = PETSC_TRUE; 1448 B->preallocated = PETSC_TRUE; 1449 1450 ierr = PetscLogFlops(B->rmap->n);CHKERRQ(ierr); 1451 1452 /* MatPivotView() */ 1453 if (sctx.nshift) { 1454 if (info->shifttype == (PetscReal)MAT_SHIFT_POSITIVE_DEFINITE) { 1455 ierr = PetscInfo4(A,"number of shift_pd tries %D, shift_amount %g, diagonal shifted up by %e fraction top_value %e\n",sctx.nshift,(double)sctx.shift_amount,(double)sctx.shift_fraction,(double)sctx.shift_top);CHKERRQ(ierr); 1456 } else if (info->shifttype == (PetscReal)MAT_SHIFT_NONZERO) { 1457 ierr = PetscInfo2(A,"number of shift_nz tries %D, shift_amount %g\n",sctx.nshift,(double)sctx.shift_amount);CHKERRQ(ierr); 1458 } else if (info->shifttype == (PetscReal)MAT_SHIFT_INBLOCKS) { 1459 ierr = PetscInfo2(A,"number of shift_inblocks applied %D, each shift_amount %g\n",sctx.nshift,(double)info->shiftamount);CHKERRQ(ierr); 1460 } 1461 } 1462 PetscFunctionReturn(0); 1463 } 1464 1465 #undef __FUNCT__ 1466 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering_inplace" 1467 PetscErrorCode MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering_inplace(Mat C,Mat A,const MatFactorInfo *info) 1468 { 1469 Mat_SeqSBAIJ *a=(Mat_SeqSBAIJ*)A->data,*b=(Mat_SeqSBAIJ*)C->data; 1470 PetscErrorCode ierr; 1471 PetscInt i,j,mbs = a->mbs; 1472 PetscInt *ai=a->i,*aj=a->j,*bi=b->i,*bj=b->j; 1473 PetscInt k,jmin,*jl,*il,nexti,ili,*acol,*bcol,nz; 1474 MatScalar *rtmp,*ba=b->a,*aa=a->a,dk,uikdi,*aval,*bval; 1475 PetscReal rs; 1476 FactorShiftCtx sctx; 1477 1478 PetscFunctionBegin; 1479 /* MatPivotSetUp(): initialize shift context sctx */ 1480 ierr = PetscMemzero(&sctx,sizeof(FactorShiftCtx));CHKERRQ(ierr); 1481 1482 /* initialization */ 1483 /* il and jl record the first nonzero element in each row of the accessing 1484 window U(0:k, k:mbs-1). 1485 jl: list of rows to be added to uneliminated rows 1486 i>= k: jl(i) is the first row to be added to row i 1487 i< k: jl(i) is the row following row i in some list of rows 1488 jl(i) = mbs indicates the end of a list 1489 il(i): points to the first nonzero element in U(i,k:mbs-1) 1490 */ 1491 ierr = PetscMalloc1(mbs,&rtmp);CHKERRQ(ierr); 1492 ierr = PetscMalloc2(mbs,&il,mbs,&jl);CHKERRQ(ierr); 1493 1494 do { 1495 sctx.newshift = PETSC_FALSE; 1496 for (i=0; i<mbs; i++) { 1497 rtmp[i] = 0.0; jl[i] = mbs; il[0] = 0; 1498 } 1499 1500 for (k = 0; k<mbs; k++) { 1501 /*initialize k-th row with elements nonzero in row perm(k) of A */ 1502 nz = ai[k+1] - ai[k]; 1503 acol = aj + ai[k]; 1504 aval = aa + ai[k]; 1505 bval = ba + bi[k]; 1506 while (nz--) { 1507 rtmp[*acol++] = *aval++; 1508 *bval++ = 0.0; /* for in-place factorization */ 1509 } 1510 1511 /* shift the diagonal of the matrix */ 1512 if (sctx.nshift) rtmp[k] += sctx.shift_amount; 1513 1514 /* modify k-th row by adding in those rows i with U(i,k)!=0 */ 1515 dk = rtmp[k]; 1516 i = jl[k]; /* first row to be added to k_th row */ 1517 1518 while (i < k) { 1519 nexti = jl[i]; /* next row to be added to k_th row */ 1520 /* compute multiplier, update D(k) and U(i,k) */ 1521 ili = il[i]; /* index of first nonzero element in U(i,k:bms-1) */ 1522 uikdi = -ba[ili]*ba[bi[i]]; 1523 dk += uikdi*ba[ili]; 1524 ba[ili] = uikdi; /* -U(i,k) */ 1525 1526 /* add multiple of row i to k-th row ... */ 1527 jmin = ili + 1; 1528 nz = bi[i+1] - jmin; 1529 if (nz > 0) { 1530 bcol = bj + jmin; 1531 bval = ba + jmin; 1532 ierr = PetscLogFlops(2.0*nz);CHKERRQ(ierr); 1533 while (nz--) rtmp[*bcol++] += uikdi*(*bval++); 1534 1535 /* update il and jl for i-th row */ 1536 il[i] = jmin; 1537 j = bj[jmin]; jl[i] = jl[j]; jl[j] = i; 1538 } 1539 i = nexti; 1540 } 1541 1542 /* shift the diagonals when zero pivot is detected */ 1543 /* compute rs=sum of abs(off-diagonal) */ 1544 rs = 0.0; 1545 jmin = bi[k]+1; 1546 nz = bi[k+1] - jmin; 1547 if (nz) { 1548 bcol = bj + jmin; 1549 while (nz--) { 1550 rs += PetscAbsScalar(rtmp[*bcol]); 1551 bcol++; 1552 } 1553 } 1554 1555 sctx.rs = rs; 1556 sctx.pv = dk; 1557 ierr = MatPivotCheck(A,info,&sctx,k);CHKERRQ(ierr); 1558 if (sctx.newshift) break; /* sctx.shift_amount is updated */ 1559 dk = sctx.pv; 1560 1561 /* copy data into U(k,:) */ 1562 ba[bi[k]] = 1.0/dk; 1563 jmin = bi[k]+1; 1564 nz = bi[k+1] - jmin; 1565 if (nz) { 1566 bcol = bj + jmin; 1567 bval = ba + jmin; 1568 while (nz--) { 1569 *bval++ = rtmp[*bcol]; 1570 rtmp[*bcol++] = 0.0; 1571 } 1572 /* add k-th row into il and jl */ 1573 il[k] = jmin; 1574 i = bj[jmin]; jl[k] = jl[i]; jl[i] = k; 1575 } 1576 } /* end of for (k = 0; k<mbs; k++) */ 1577 } while (sctx.newshift); 1578 ierr = PetscFree(rtmp);CHKERRQ(ierr); 1579 ierr = PetscFree2(il,jl);CHKERRQ(ierr); 1580 1581 C->ops->solve = MatSolve_SeqSBAIJ_1_NaturalOrdering_inplace; 1582 C->ops->solves = MatSolves_SeqSBAIJ_1_inplace; 1583 C->ops->solvetranspose = MatSolve_SeqSBAIJ_1_NaturalOrdering_inplace; 1584 C->ops->forwardsolve = MatForwardSolve_SeqSBAIJ_1_NaturalOrdering_inplace; 1585 C->ops->backwardsolve = MatBackwardSolve_SeqSBAIJ_1_NaturalOrdering_inplace; 1586 1587 C->assembled = PETSC_TRUE; 1588 C->preallocated = PETSC_TRUE; 1589 1590 ierr = PetscLogFlops(C->rmap->N);CHKERRQ(ierr); 1591 if (sctx.nshift) { 1592 if (info->shifttype == (PetscReal)MAT_SHIFT_NONZERO) { 1593 ierr = PetscInfo2(A,"number of shiftnz tries %D, shift_amount %g\n",sctx.nshift,(double)sctx.shift_amount);CHKERRQ(ierr); 1594 } else if (info->shifttype == (PetscReal)MAT_SHIFT_POSITIVE_DEFINITE) { 1595 ierr = PetscInfo2(A,"number of shiftpd tries %D, shift_amount %g\n",sctx.nshift,(double)sctx.shift_amount);CHKERRQ(ierr); 1596 } 1597 } 1598 PetscFunctionReturn(0); 1599 } 1600 1601 #undef __FUNCT__ 1602 #define __FUNCT__ "MatCholeskyFactor_SeqSBAIJ" 1603 PetscErrorCode MatCholeskyFactor_SeqSBAIJ(Mat A,IS perm,const MatFactorInfo *info) 1604 { 1605 PetscErrorCode ierr; 1606 Mat C; 1607 1608 PetscFunctionBegin; 1609 ierr = MatGetFactor(A,"petsc",MAT_FACTOR_CHOLESKY,&C);CHKERRQ(ierr); 1610 ierr = MatCholeskyFactorSymbolic(C,A,perm,info);CHKERRQ(ierr); 1611 ierr = MatCholeskyFactorNumeric(C,A,info);CHKERRQ(ierr); 1612 1613 A->ops->solve = C->ops->solve; 1614 A->ops->solvetranspose = C->ops->solvetranspose; 1615 1616 ierr = MatHeaderMerge(A,C);CHKERRQ(ierr); 1617 PetscFunctionReturn(0); 1618 } 1619 1620 1621