1 2 /* 3 Defines the basic matrix operations for sequential dense. 4 */ 5 6 #include <../src/mat/impls/dense/seq/dense.h> /*I "petscmat.h" I*/ 7 #include <petscblaslapack.h> 8 9 #include <../src/mat/impls/aij/seq/aij.h> 10 11 PetscErrorCode MatSeqDenseSymmetrize_Private(Mat A, PetscBool hermitian) 12 { 13 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 14 PetscInt j, k, n = A->rmap->n; 15 PetscScalar *v; 16 PetscErrorCode ierr; 17 18 PetscFunctionBegin; 19 if (A->rmap->n != A->cmap->n) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Cannot symmetrize a rectangular matrix"); 20 ierr = MatDenseGetArray(A,&v);CHKERRQ(ierr); 21 if (!hermitian) { 22 for (k=0;k<n;k++) { 23 for (j=k;j<n;j++) { 24 v[j*mat->lda + k] = v[k*mat->lda + j]; 25 } 26 } 27 } else { 28 for (k=0;k<n;k++) { 29 for (j=k;j<n;j++) { 30 v[j*mat->lda + k] = PetscConj(v[k*mat->lda + j]); 31 } 32 } 33 } 34 ierr = MatDenseRestoreArray(A,&v);CHKERRQ(ierr); 35 PetscFunctionReturn(0); 36 } 37 38 PETSC_EXTERN PetscErrorCode MatSeqDenseInvertFactors_Private(Mat A) 39 { 40 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 41 PetscErrorCode ierr; 42 PetscBLASInt info,n; 43 44 PetscFunctionBegin; 45 if (!A->rmap->n || !A->cmap->n) PetscFunctionReturn(0); 46 ierr = PetscBLASIntCast(A->cmap->n,&n);CHKERRQ(ierr); 47 if (A->factortype == MAT_FACTOR_LU) { 48 if (!mat->pivots) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Pivots not present"); 49 if (!mat->fwork) { 50 mat->lfwork = n; 51 ierr = PetscMalloc1(mat->lfwork,&mat->fwork);CHKERRQ(ierr); 52 ierr = PetscLogObjectMemory((PetscObject)A,mat->lfwork*sizeof(PetscBLASInt));CHKERRQ(ierr); 53 } 54 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 55 PetscStackCallBLAS("LAPACKgetri",LAPACKgetri_(&n,mat->v,&mat->lda,mat->pivots,mat->fwork,&mat->lfwork,&info)); 56 ierr = PetscFPTrapPop();CHKERRQ(ierr); 57 ierr = PetscLogFlops((1.0*A->cmap->n*A->cmap->n*A->cmap->n)/3.0);CHKERRQ(ierr); 58 } else if (A->factortype == MAT_FACTOR_CHOLESKY) { 59 if (A->spd) { 60 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 61 PetscStackCallBLAS("LAPACKpotri",LAPACKpotri_("L",&n,mat->v,&mat->lda,&info)); 62 ierr = PetscFPTrapPop();CHKERRQ(ierr); 63 ierr = MatSeqDenseSymmetrize_Private(A,PETSC_TRUE);CHKERRQ(ierr); 64 #if defined(PETSC_USE_COMPLEX) 65 } else if (A->hermitian) { 66 if (!mat->pivots) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Pivots not present"); 67 if (!mat->fwork) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Fwork not present"); 68 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 69 PetscStackCallBLAS("LAPACKhetri",LAPACKhetri_("L",&n,mat->v,&mat->lda,mat->pivots,mat->fwork,&info)); 70 ierr = PetscFPTrapPop();CHKERRQ(ierr); 71 ierr = MatSeqDenseSymmetrize_Private(A,PETSC_TRUE);CHKERRQ(ierr); 72 #endif 73 } else { /* symmetric case */ 74 if (!mat->pivots) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Pivots not present"); 75 if (!mat->fwork) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Fwork not present"); 76 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 77 PetscStackCallBLAS("LAPACKsytri",LAPACKsytri_("L",&n,mat->v,&mat->lda,mat->pivots,mat->fwork,&info)); 78 ierr = PetscFPTrapPop();CHKERRQ(ierr); 79 ierr = MatSeqDenseSymmetrize_Private(A,PETSC_FALSE);CHKERRQ(ierr); 80 } 81 if (info) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_MAT_CH_ZRPVT,"Bad Inversion: zero pivot in row %D",(PetscInt)info-1); 82 ierr = PetscLogFlops((1.0*A->cmap->n*A->cmap->n*A->cmap->n)/3.0);CHKERRQ(ierr); 83 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Matrix must be factored to solve"); 84 85 A->ops->solve = NULL; 86 A->ops->matsolve = NULL; 87 A->ops->solvetranspose = NULL; 88 A->ops->matsolvetranspose = NULL; 89 A->ops->solveadd = NULL; 90 A->ops->solvetransposeadd = NULL; 91 A->factortype = MAT_FACTOR_NONE; 92 ierr = PetscFree(A->solvertype);CHKERRQ(ierr); 93 PetscFunctionReturn(0); 94 } 95 96 PetscErrorCode MatZeroRowsColumns_SeqDense(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag,Vec x,Vec b) 97 { 98 PetscErrorCode ierr; 99 Mat_SeqDense *l = (Mat_SeqDense*)A->data; 100 PetscInt m = l->lda, n = A->cmap->n,r = A->rmap->n, i,j; 101 PetscScalar *slot,*bb,*v; 102 const PetscScalar *xx; 103 104 PetscFunctionBegin; 105 if (PetscDefined(USE_DEBUG)) { 106 for (i=0; i<N; i++) { 107 if (rows[i] < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row requested to be zeroed"); 108 if (rows[i] >= A->rmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row %D requested to be zeroed greater than or equal number of rows %D",rows[i],A->rmap->n); 109 if (rows[i] >= A->cmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Col %D requested to be zeroed greater than or equal number of cols %D",rows[i],A->cmap->n); 110 } 111 } 112 if (!N) PetscFunctionReturn(0); 113 114 /* fix right hand side if needed */ 115 if (x && b) { 116 Vec xt; 117 118 if (A->rmap->n != A->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only coded for square matrices"); 119 ierr = VecDuplicate(x,&xt);CHKERRQ(ierr); 120 ierr = VecCopy(x,xt);CHKERRQ(ierr); 121 ierr = VecScale(xt,-1.0);CHKERRQ(ierr); 122 ierr = MatMultAdd(A,xt,b,b);CHKERRQ(ierr); 123 ierr = VecDestroy(&xt);CHKERRQ(ierr); 124 ierr = VecGetArrayRead(x,&xx);CHKERRQ(ierr); 125 ierr = VecGetArray(b,&bb);CHKERRQ(ierr); 126 for (i=0; i<N; i++) bb[rows[i]] = diag*xx[rows[i]]; 127 ierr = VecRestoreArrayRead(x,&xx);CHKERRQ(ierr); 128 ierr = VecRestoreArray(b,&bb);CHKERRQ(ierr); 129 } 130 131 ierr = MatDenseGetArray(A,&v);CHKERRQ(ierr); 132 for (i=0; i<N; i++) { 133 slot = v + rows[i]*m; 134 ierr = PetscArrayzero(slot,r);CHKERRQ(ierr); 135 } 136 for (i=0; i<N; i++) { 137 slot = v + rows[i]; 138 for (j=0; j<n; j++) { *slot = 0.0; slot += m;} 139 } 140 if (diag != 0.0) { 141 if (A->rmap->n != A->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only coded for square matrices"); 142 for (i=0; i<N; i++) { 143 slot = v + (m+1)*rows[i]; 144 *slot = diag; 145 } 146 } 147 ierr = MatDenseRestoreArray(A,&v);CHKERRQ(ierr); 148 PetscFunctionReturn(0); 149 } 150 151 PetscErrorCode MatPtAPNumeric_SeqDense_SeqDense(Mat A,Mat P,Mat C) 152 { 153 Mat_SeqDense *c = (Mat_SeqDense*)(C->data); 154 PetscErrorCode ierr; 155 156 PetscFunctionBegin; 157 if (c->ptapwork) { 158 ierr = (*C->ops->matmultnumeric)(A,P,c->ptapwork);CHKERRQ(ierr); 159 ierr = (*C->ops->transposematmultnumeric)(P,c->ptapwork,C);CHKERRQ(ierr); 160 } else SETERRQ(PetscObjectComm((PetscObject)C),PETSC_ERR_SUP,"Must call MatPtAPSymbolic_SeqDense_SeqDense() first"); 161 PetscFunctionReturn(0); 162 } 163 164 PetscErrorCode MatPtAPSymbolic_SeqDense_SeqDense(Mat A,Mat P,PetscReal fill,Mat C) 165 { 166 Mat_SeqDense *c; 167 PetscBool cisdense; 168 PetscErrorCode ierr; 169 170 PetscFunctionBegin; 171 ierr = MatSetSizes(C,P->cmap->n,P->cmap->n,P->cmap->N,P->cmap->N);CHKERRQ(ierr); 172 ierr = PetscObjectTypeCompareAny((PetscObject)C,&cisdense,MATSEQDENSE,MATSEQDENSECUDA,"");CHKERRQ(ierr); 173 if (!cisdense) { 174 PetscBool flg; 175 176 ierr = PetscObjectTypeCompare((PetscObject)P,((PetscObject)A)->type_name,&flg);CHKERRQ(ierr); 177 ierr = MatSetType(C,flg ? ((PetscObject)A)->type_name : MATDENSE);CHKERRQ(ierr); 178 } 179 ierr = MatSetUp(C);CHKERRQ(ierr); 180 c = (Mat_SeqDense*)C->data; 181 ierr = MatCreate(PetscObjectComm((PetscObject)A),&c->ptapwork);CHKERRQ(ierr); 182 ierr = MatSetSizes(c->ptapwork,A->rmap->n,P->cmap->n,A->rmap->N,P->cmap->N);CHKERRQ(ierr); 183 ierr = MatSetType(c->ptapwork,((PetscObject)C)->type_name);CHKERRQ(ierr); 184 ierr = MatSetUp(c->ptapwork);CHKERRQ(ierr); 185 PetscFunctionReturn(0); 186 } 187 188 PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqDense(Mat A,MatType newtype,MatReuse reuse,Mat *newmat) 189 { 190 Mat B = NULL; 191 Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data; 192 Mat_SeqDense *b; 193 PetscErrorCode ierr; 194 PetscInt *ai=a->i,*aj=a->j,m=A->rmap->N,n=A->cmap->N,i; 195 MatScalar *av=a->a; 196 PetscBool isseqdense; 197 198 PetscFunctionBegin; 199 if (reuse == MAT_REUSE_MATRIX) { 200 ierr = PetscObjectTypeCompare((PetscObject)*newmat,MATSEQDENSE,&isseqdense);CHKERRQ(ierr); 201 if (!isseqdense) SETERRQ1(PetscObjectComm((PetscObject)*newmat),PETSC_ERR_USER,"Cannot reuse matrix of type %s",((PetscObject)(*newmat))->type_name); 202 } 203 if (reuse != MAT_REUSE_MATRIX) { 204 ierr = MatCreate(PetscObjectComm((PetscObject)A),&B);CHKERRQ(ierr); 205 ierr = MatSetSizes(B,m,n,m,n);CHKERRQ(ierr); 206 ierr = MatSetType(B,MATSEQDENSE);CHKERRQ(ierr); 207 ierr = MatSeqDenseSetPreallocation(B,NULL);CHKERRQ(ierr); 208 b = (Mat_SeqDense*)(B->data); 209 } else { 210 b = (Mat_SeqDense*)((*newmat)->data); 211 ierr = PetscArrayzero(b->v,m*n);CHKERRQ(ierr); 212 } 213 for (i=0; i<m; i++) { 214 PetscInt j; 215 for (j=0;j<ai[1]-ai[0];j++) { 216 b->v[*aj*m+i] = *av; 217 aj++; 218 av++; 219 } 220 ai++; 221 } 222 223 if (reuse == MAT_INPLACE_MATRIX) { 224 ierr = MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 225 ierr = MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 226 ierr = MatHeaderReplace(A,&B);CHKERRQ(ierr); 227 } else { 228 if (B) *newmat = B; 229 ierr = MatAssemblyBegin(*newmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 230 ierr = MatAssemblyEnd(*newmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 231 } 232 PetscFunctionReturn(0); 233 } 234 235 PETSC_INTERN PetscErrorCode MatConvert_SeqDense_SeqAIJ(Mat A, MatType newtype,MatReuse reuse,Mat *newmat) 236 { 237 Mat B; 238 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 239 PetscErrorCode ierr; 240 PetscInt i, j; 241 PetscInt *rows, *nnz; 242 MatScalar *aa = a->v, *vals; 243 244 PetscFunctionBegin; 245 ierr = MatCreate(PetscObjectComm((PetscObject)A),&B);CHKERRQ(ierr); 246 ierr = MatSetSizes(B,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);CHKERRQ(ierr); 247 ierr = MatSetType(B,MATSEQAIJ);CHKERRQ(ierr); 248 ierr = PetscCalloc3(A->rmap->n,&rows,A->rmap->n,&nnz,A->rmap->n,&vals);CHKERRQ(ierr); 249 for (j=0; j<A->cmap->n; j++) { 250 for (i=0; i<A->rmap->n; i++) if (aa[i] != 0.0 || i == j) ++nnz[i]; 251 aa += a->lda; 252 } 253 ierr = MatSeqAIJSetPreallocation(B,PETSC_DETERMINE,nnz);CHKERRQ(ierr); 254 aa = a->v; 255 for (j=0; j<A->cmap->n; j++) { 256 PetscInt numRows = 0; 257 for (i=0; i<A->rmap->n; i++) if (aa[i] != 0.0 || i == j) {rows[numRows] = i; vals[numRows++] = aa[i];} 258 ierr = MatSetValues(B,numRows,rows,1,&j,vals,INSERT_VALUES);CHKERRQ(ierr); 259 aa += a->lda; 260 } 261 ierr = PetscFree3(rows,nnz,vals);CHKERRQ(ierr); 262 ierr = MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 263 ierr = MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 264 265 if (reuse == MAT_INPLACE_MATRIX) { 266 ierr = MatHeaderReplace(A,&B);CHKERRQ(ierr); 267 } else { 268 *newmat = B; 269 } 270 PetscFunctionReturn(0); 271 } 272 273 PetscErrorCode MatAXPY_SeqDense(Mat Y,PetscScalar alpha,Mat X,MatStructure str) 274 { 275 Mat_SeqDense *x = (Mat_SeqDense*)X->data,*y = (Mat_SeqDense*)Y->data; 276 const PetscScalar *xv; 277 PetscScalar *yv; 278 PetscBLASInt N,m,ldax,lday,one = 1; 279 PetscErrorCode ierr; 280 281 PetscFunctionBegin; 282 ierr = MatDenseGetArrayRead(X,&xv);CHKERRQ(ierr); 283 ierr = MatDenseGetArray(Y,&yv);CHKERRQ(ierr); 284 ierr = PetscBLASIntCast(X->rmap->n*X->cmap->n,&N);CHKERRQ(ierr); 285 ierr = PetscBLASIntCast(X->rmap->n,&m);CHKERRQ(ierr); 286 ierr = PetscBLASIntCast(x->lda,&ldax);CHKERRQ(ierr); 287 ierr = PetscBLASIntCast(y->lda,&lday);CHKERRQ(ierr); 288 if (ldax>m || lday>m) { 289 PetscInt j; 290 291 for (j=0; j<X->cmap->n; j++) { 292 PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&m,&alpha,xv+j*ldax,&one,yv+j*lday,&one)); 293 } 294 } else { 295 PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&N,&alpha,xv,&one,yv,&one)); 296 } 297 ierr = MatDenseRestoreArrayRead(X,&xv);CHKERRQ(ierr); 298 ierr = MatDenseRestoreArray(Y,&yv);CHKERRQ(ierr); 299 ierr = PetscLogFlops(PetscMax(2*N-1,0));CHKERRQ(ierr); 300 PetscFunctionReturn(0); 301 } 302 303 static PetscErrorCode MatGetInfo_SeqDense(Mat A,MatInfoType flag,MatInfo *info) 304 { 305 PetscLogDouble N = A->rmap->n*A->cmap->n; 306 307 PetscFunctionBegin; 308 info->block_size = 1.0; 309 info->nz_allocated = N; 310 info->nz_used = N; 311 info->nz_unneeded = 0; 312 info->assemblies = A->num_ass; 313 info->mallocs = 0; 314 info->memory = ((PetscObject)A)->mem; 315 info->fill_ratio_given = 0; 316 info->fill_ratio_needed = 0; 317 info->factor_mallocs = 0; 318 PetscFunctionReturn(0); 319 } 320 321 PetscErrorCode MatScale_SeqDense(Mat A,PetscScalar alpha) 322 { 323 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 324 PetscScalar *v; 325 PetscErrorCode ierr; 326 PetscBLASInt one = 1,j,nz,lda; 327 328 PetscFunctionBegin; 329 ierr = MatDenseGetArray(A,&v);CHKERRQ(ierr); 330 ierr = PetscBLASIntCast(a->lda,&lda);CHKERRQ(ierr); 331 if (lda>A->rmap->n) { 332 ierr = PetscBLASIntCast(A->rmap->n,&nz);CHKERRQ(ierr); 333 for (j=0; j<A->cmap->n; j++) { 334 PetscStackCallBLAS("BLASscal",BLASscal_(&nz,&alpha,v+j*lda,&one)); 335 } 336 } else { 337 ierr = PetscBLASIntCast(A->rmap->n*A->cmap->n,&nz);CHKERRQ(ierr); 338 PetscStackCallBLAS("BLASscal",BLASscal_(&nz,&alpha,v,&one)); 339 } 340 ierr = PetscLogFlops(nz);CHKERRQ(ierr); 341 ierr = MatDenseRestoreArray(A,&v);CHKERRQ(ierr); 342 PetscFunctionReturn(0); 343 } 344 345 static PetscErrorCode MatIsHermitian_SeqDense(Mat A,PetscReal rtol,PetscBool *fl) 346 { 347 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 348 PetscInt i,j,m = A->rmap->n,N = a->lda; 349 const PetscScalar *v; 350 PetscErrorCode ierr; 351 352 PetscFunctionBegin; 353 *fl = PETSC_FALSE; 354 if (A->rmap->n != A->cmap->n) PetscFunctionReturn(0); 355 ierr = MatDenseGetArrayRead(A,&v);CHKERRQ(ierr); 356 for (i=0; i<m; i++) { 357 for (j=i; j<m; j++) { 358 if (PetscAbsScalar(v[i+j*N] - PetscConj(v[j+i*N])) > rtol) { 359 goto restore; 360 } 361 } 362 } 363 *fl = PETSC_TRUE; 364 restore: 365 ierr = MatDenseRestoreArrayRead(A,&v);CHKERRQ(ierr); 366 PetscFunctionReturn(0); 367 } 368 369 static PetscErrorCode MatIsSymmetric_SeqDense(Mat A,PetscReal rtol,PetscBool *fl) 370 { 371 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 372 PetscInt i,j,m = A->rmap->n,N = a->lda; 373 const PetscScalar *v; 374 PetscErrorCode ierr; 375 376 PetscFunctionBegin; 377 *fl = PETSC_FALSE; 378 if (A->rmap->n != A->cmap->n) PetscFunctionReturn(0); 379 ierr = MatDenseGetArrayRead(A,&v);CHKERRQ(ierr); 380 for (i=0; i<m; i++) { 381 for (j=i; j<m; j++) { 382 if (PetscAbsScalar(v[i+j*N] - v[j+i*N]) > rtol) { 383 goto restore; 384 } 385 } 386 } 387 *fl = PETSC_TRUE; 388 restore: 389 ierr = MatDenseRestoreArrayRead(A,&v);CHKERRQ(ierr); 390 PetscFunctionReturn(0); 391 } 392 393 PetscErrorCode MatDuplicateNoCreate_SeqDense(Mat newi,Mat A,MatDuplicateOption cpvalues) 394 { 395 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 396 PetscErrorCode ierr; 397 PetscInt lda = (PetscInt)mat->lda,j,m,nlda = lda; 398 399 PetscFunctionBegin; 400 ierr = PetscLayoutReference(A->rmap,&newi->rmap);CHKERRQ(ierr); 401 ierr = PetscLayoutReference(A->cmap,&newi->cmap);CHKERRQ(ierr); 402 if (cpvalues == MAT_SHARE_NONZERO_PATTERN) { /* propagate LDA */ 403 ierr = MatDenseSetLDA(newi,lda);CHKERRQ(ierr); 404 } 405 ierr = MatSeqDenseSetPreallocation(newi,NULL);CHKERRQ(ierr); 406 if (cpvalues == MAT_COPY_VALUES) { 407 const PetscScalar *av; 408 PetscScalar *v; 409 410 ierr = MatDenseGetArrayRead(A,&av);CHKERRQ(ierr); 411 ierr = MatDenseGetArray(newi,&v);CHKERRQ(ierr); 412 ierr = MatDenseGetLDA(newi,&nlda);CHKERRQ(ierr); 413 m = A->rmap->n; 414 if (lda>m || nlda>m) { 415 for (j=0; j<A->cmap->n; j++) { 416 ierr = PetscArraycpy(v+j*nlda,av+j*lda,m);CHKERRQ(ierr); 417 } 418 } else { 419 ierr = PetscArraycpy(v,av,A->rmap->n*A->cmap->n);CHKERRQ(ierr); 420 } 421 ierr = MatDenseRestoreArray(newi,&v);CHKERRQ(ierr); 422 ierr = MatDenseRestoreArrayRead(A,&av);CHKERRQ(ierr); 423 } 424 PetscFunctionReturn(0); 425 } 426 427 PetscErrorCode MatDuplicate_SeqDense(Mat A,MatDuplicateOption cpvalues,Mat *newmat) 428 { 429 PetscErrorCode ierr; 430 431 PetscFunctionBegin; 432 ierr = MatCreate(PetscObjectComm((PetscObject)A),newmat);CHKERRQ(ierr); 433 ierr = MatSetSizes(*newmat,A->rmap->n,A->cmap->n,A->rmap->n,A->cmap->n);CHKERRQ(ierr); 434 ierr = MatSetType(*newmat,((PetscObject)A)->type_name);CHKERRQ(ierr); 435 ierr = MatDuplicateNoCreate_SeqDense(*newmat,A,cpvalues);CHKERRQ(ierr); 436 PetscFunctionReturn(0); 437 } 438 439 static PetscErrorCode MatLUFactorNumeric_SeqDense(Mat fact,Mat A,const MatFactorInfo *info_dummy) 440 { 441 MatFactorInfo info; 442 PetscErrorCode ierr; 443 444 PetscFunctionBegin; 445 ierr = MatDuplicateNoCreate_SeqDense(fact,A,MAT_COPY_VALUES);CHKERRQ(ierr); 446 ierr = (*fact->ops->lufactor)(fact,0,0,&info);CHKERRQ(ierr); 447 PetscFunctionReturn(0); 448 } 449 450 static PetscErrorCode MatSolve_SeqDense(Mat A,Vec xx,Vec yy) 451 { 452 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 453 PetscErrorCode ierr; 454 const PetscScalar *x; 455 PetscScalar *y; 456 PetscBLASInt one = 1,info,m; 457 458 PetscFunctionBegin; 459 ierr = PetscBLASIntCast(A->rmap->n,&m);CHKERRQ(ierr); 460 ierr = VecGetArrayRead(xx,&x);CHKERRQ(ierr); 461 ierr = VecGetArray(yy,&y);CHKERRQ(ierr); 462 ierr = PetscArraycpy(y,x,A->rmap->n);CHKERRQ(ierr); 463 if (A->factortype == MAT_FACTOR_LU) { 464 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 465 PetscStackCallBLAS("LAPACKgetrs",LAPACKgetrs_("N",&m,&one,mat->v,&mat->lda,mat->pivots,y,&m,&info)); 466 ierr = PetscFPTrapPop();CHKERRQ(ierr); 467 if (info) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"GETRS - Bad solve"); 468 } else if (A->factortype == MAT_FACTOR_CHOLESKY) { 469 if (A->spd) { 470 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 471 PetscStackCallBLAS("LAPACKpotrs",LAPACKpotrs_("L",&m,&one,mat->v,&mat->lda,y,&m,&info)); 472 ierr = PetscFPTrapPop();CHKERRQ(ierr); 473 if (info) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"POTRS Bad solve"); 474 #if defined(PETSC_USE_COMPLEX) 475 } else if (A->hermitian) { 476 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 477 PetscStackCallBLAS("LAPACKhetrs",LAPACKhetrs_("L",&m,&one,mat->v,&mat->lda,mat->pivots,y,&m,&info)); 478 ierr = PetscFPTrapPop();CHKERRQ(ierr); 479 if (info) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"HETRS Bad solve"); 480 #endif 481 } else { /* symmetric case */ 482 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 483 PetscStackCallBLAS("LAPACKsytrs",LAPACKsytrs_("L",&m,&one,mat->v,&mat->lda,mat->pivots,y,&m,&info)); 484 ierr = PetscFPTrapPop();CHKERRQ(ierr); 485 if (info) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"SYTRS Bad solve"); 486 } 487 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Matrix must be factored to solve"); 488 ierr = VecRestoreArrayRead(xx,&x);CHKERRQ(ierr); 489 ierr = VecRestoreArray(yy,&y);CHKERRQ(ierr); 490 ierr = PetscLogFlops(2.0*A->cmap->n*A->cmap->n - A->cmap->n);CHKERRQ(ierr); 491 PetscFunctionReturn(0); 492 } 493 494 static PetscErrorCode MatMatSolve_SeqDense(Mat A,Mat B,Mat X) 495 { 496 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 497 PetscErrorCode ierr; 498 const PetscScalar *b; 499 PetscScalar *x; 500 PetscInt n; 501 PetscBLASInt nrhs,info,m; 502 503 PetscFunctionBegin; 504 ierr = PetscBLASIntCast(A->rmap->n,&m);CHKERRQ(ierr); 505 ierr = MatGetSize(B,NULL,&n);CHKERRQ(ierr); 506 ierr = PetscBLASIntCast(n,&nrhs);CHKERRQ(ierr); 507 ierr = MatDenseGetArrayRead(B,&b);CHKERRQ(ierr); 508 ierr = MatDenseGetArray(X,&x);CHKERRQ(ierr); 509 510 ierr = PetscArraycpy(x,b,m*nrhs);CHKERRQ(ierr); 511 512 if (A->factortype == MAT_FACTOR_LU) { 513 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 514 PetscStackCallBLAS("LAPACKgetrs",LAPACKgetrs_("N",&m,&nrhs,mat->v,&mat->lda,mat->pivots,x,&m,&info)); 515 ierr = PetscFPTrapPop();CHKERRQ(ierr); 516 if (info) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"GETRS - Bad solve"); 517 } else if (A->factortype == MAT_FACTOR_CHOLESKY) { 518 if (A->spd) { 519 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 520 PetscStackCallBLAS("LAPACKpotrs",LAPACKpotrs_("L",&m,&nrhs,mat->v,&mat->lda,x,&m,&info)); 521 ierr = PetscFPTrapPop();CHKERRQ(ierr); 522 if (info) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"POTRS Bad solve"); 523 #if defined(PETSC_USE_COMPLEX) 524 } else if (A->hermitian) { 525 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 526 PetscStackCallBLAS("LAPACKhetrs",LAPACKhetrs_("L",&m,&nrhs,mat->v,&mat->lda,mat->pivots,x,&m,&info)); 527 ierr = PetscFPTrapPop();CHKERRQ(ierr); 528 if (info) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"HETRS Bad solve"); 529 #endif 530 } else { /* symmetric case */ 531 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 532 PetscStackCallBLAS("LAPACKsytrs",LAPACKsytrs_("L",&m,&nrhs,mat->v,&mat->lda,mat->pivots,x,&m,&info)); 533 ierr = PetscFPTrapPop();CHKERRQ(ierr); 534 if (info) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"SYTRS Bad solve"); 535 } 536 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Matrix must be factored to solve"); 537 538 ierr = MatDenseRestoreArrayRead(B,&b);CHKERRQ(ierr); 539 ierr = MatDenseRestoreArray(X,&x);CHKERRQ(ierr); 540 ierr = PetscLogFlops(nrhs*(2.0*m*m - m));CHKERRQ(ierr); 541 PetscFunctionReturn(0); 542 } 543 544 static PetscErrorCode MatConjugate_SeqDense(Mat); 545 546 static PetscErrorCode MatSolveTranspose_SeqDense(Mat A,Vec xx,Vec yy) 547 { 548 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 549 PetscErrorCode ierr; 550 const PetscScalar *x; 551 PetscScalar *y; 552 PetscBLASInt one = 1,info,m; 553 554 PetscFunctionBegin; 555 ierr = PetscBLASIntCast(A->rmap->n,&m);CHKERRQ(ierr); 556 ierr = VecGetArrayRead(xx,&x);CHKERRQ(ierr); 557 ierr = VecGetArray(yy,&y);CHKERRQ(ierr); 558 ierr = PetscArraycpy(y,x,A->rmap->n);CHKERRQ(ierr); 559 if (A->factortype == MAT_FACTOR_LU) { 560 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 561 PetscStackCallBLAS("LAPACKgetrs",LAPACKgetrs_("T",&m,&one,mat->v,&mat->lda,mat->pivots,y,&m,&info)); 562 ierr = PetscFPTrapPop();CHKERRQ(ierr); 563 if (info) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"POTRS - Bad solve"); 564 } else if (A->factortype == MAT_FACTOR_CHOLESKY) { 565 if (A->spd) { 566 #if defined(PETSC_USE_COMPLEX) 567 ierr = MatConjugate_SeqDense(A);CHKERRQ(ierr); 568 #endif 569 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 570 PetscStackCallBLAS("LAPACKpotrs",LAPACKpotrs_("L",&m,&one,mat->v,&mat->lda,y,&m,&info)); 571 ierr = PetscFPTrapPop();CHKERRQ(ierr); 572 #if defined(PETSC_USE_COMPLEX) 573 ierr = MatConjugate_SeqDense(A);CHKERRQ(ierr); 574 #endif 575 if (info) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"POTRS Bad solve"); 576 #if defined(PETSC_USE_COMPLEX) 577 } else if (A->hermitian) { 578 ierr = MatConjugate_SeqDense(A);CHKERRQ(ierr); 579 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 580 PetscStackCallBLAS("LAPACKhetrs",LAPACKhetrs_("L",&m,&one,mat->v,&mat->lda,mat->pivots,y,&m,&info)); 581 ierr = PetscFPTrapPop();CHKERRQ(ierr); 582 ierr = MatConjugate_SeqDense(A);CHKERRQ(ierr); 583 #endif 584 } else { /* symmetric case */ 585 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 586 PetscStackCallBLAS("LAPACKsytrs",LAPACKsytrs_("L",&m,&one,mat->v,&mat->lda,mat->pivots,y,&m,&info)); 587 ierr = PetscFPTrapPop();CHKERRQ(ierr); 588 if (info) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"SYTRS Bad solve"); 589 } 590 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Matrix must be factored to solve"); 591 ierr = VecRestoreArrayRead(xx,&x);CHKERRQ(ierr); 592 ierr = VecRestoreArray(yy,&y);CHKERRQ(ierr); 593 ierr = PetscLogFlops(2.0*A->cmap->n*A->cmap->n - A->cmap->n);CHKERRQ(ierr); 594 PetscFunctionReturn(0); 595 } 596 597 /* ---------------------------------------------------------------*/ 598 /* COMMENT: I have chosen to hide row permutation in the pivots, 599 rather than put it in the Mat->row slot.*/ 600 PetscErrorCode MatLUFactor_SeqDense(Mat A,IS row,IS col,const MatFactorInfo *minfo) 601 { 602 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 603 PetscErrorCode ierr; 604 PetscBLASInt n,m,info; 605 606 PetscFunctionBegin; 607 ierr = PetscBLASIntCast(A->cmap->n,&n);CHKERRQ(ierr); 608 ierr = PetscBLASIntCast(A->rmap->n,&m);CHKERRQ(ierr); 609 if (!mat->pivots) { 610 ierr = PetscMalloc1(A->rmap->n,&mat->pivots);CHKERRQ(ierr); 611 ierr = PetscLogObjectMemory((PetscObject)A,A->rmap->n*sizeof(PetscBLASInt));CHKERRQ(ierr); 612 } 613 if (!A->rmap->n || !A->cmap->n) PetscFunctionReturn(0); 614 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 615 PetscStackCallBLAS("LAPACKgetrf",LAPACKgetrf_(&m,&n,mat->v,&mat->lda,mat->pivots,&info)); 616 ierr = PetscFPTrapPop();CHKERRQ(ierr); 617 618 if (info<0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"Bad argument to LU factorization"); 619 if (info>0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_MAT_LU_ZRPVT,"Bad LU factorization"); 620 621 A->ops->solve = MatSolve_SeqDense; 622 A->ops->matsolve = MatMatSolve_SeqDense; 623 A->ops->solvetranspose = MatSolveTranspose_SeqDense; 624 A->factortype = MAT_FACTOR_LU; 625 626 ierr = PetscFree(A->solvertype);CHKERRQ(ierr); 627 ierr = PetscStrallocpy(MATSOLVERPETSC,&A->solvertype);CHKERRQ(ierr); 628 629 ierr = PetscLogFlops((2.0*A->cmap->n*A->cmap->n*A->cmap->n)/3);CHKERRQ(ierr); 630 PetscFunctionReturn(0); 631 } 632 633 /* Cholesky as L*L^T or L*D*L^T and the symmetric/hermitian complex variants */ 634 PetscErrorCode MatCholeskyFactor_SeqDense(Mat A,IS perm,const MatFactorInfo *factinfo) 635 { 636 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 637 PetscErrorCode ierr; 638 PetscBLASInt info,n; 639 640 PetscFunctionBegin; 641 ierr = PetscBLASIntCast(A->cmap->n,&n);CHKERRQ(ierr); 642 if (!A->rmap->n || !A->cmap->n) PetscFunctionReturn(0); 643 if (A->spd) { 644 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 645 PetscStackCallBLAS("LAPACKpotrf",LAPACKpotrf_("L",&n,mat->v,&mat->lda,&info)); 646 ierr = PetscFPTrapPop();CHKERRQ(ierr); 647 #if defined(PETSC_USE_COMPLEX) 648 } else if (A->hermitian) { 649 if (!mat->pivots) { 650 ierr = PetscMalloc1(A->rmap->n,&mat->pivots);CHKERRQ(ierr); 651 ierr = PetscLogObjectMemory((PetscObject)A,A->rmap->n*sizeof(PetscBLASInt));CHKERRQ(ierr); 652 } 653 if (!mat->fwork) { 654 PetscScalar dummy; 655 656 mat->lfwork = -1; 657 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 658 PetscStackCallBLAS("LAPACKhetrf",LAPACKhetrf_("L",&n,mat->v,&mat->lda,mat->pivots,&dummy,&mat->lfwork,&info)); 659 ierr = PetscFPTrapPop();CHKERRQ(ierr); 660 mat->lfwork = (PetscInt)PetscRealPart(dummy); 661 ierr = PetscMalloc1(mat->lfwork,&mat->fwork);CHKERRQ(ierr); 662 ierr = PetscLogObjectMemory((PetscObject)A,mat->lfwork*sizeof(PetscBLASInt));CHKERRQ(ierr); 663 } 664 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 665 PetscStackCallBLAS("LAPACKhetrf",LAPACKhetrf_("L",&n,mat->v,&mat->lda,mat->pivots,mat->fwork,&mat->lfwork,&info)); 666 ierr = PetscFPTrapPop();CHKERRQ(ierr); 667 #endif 668 } else { /* symmetric case */ 669 if (!mat->pivots) { 670 ierr = PetscMalloc1(A->rmap->n,&mat->pivots);CHKERRQ(ierr); 671 ierr = PetscLogObjectMemory((PetscObject)A,A->rmap->n*sizeof(PetscBLASInt));CHKERRQ(ierr); 672 } 673 if (!mat->fwork) { 674 PetscScalar dummy; 675 676 mat->lfwork = -1; 677 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 678 PetscStackCallBLAS("LAPACKsytrf",LAPACKsytrf_("L",&n,mat->v,&mat->lda,mat->pivots,&dummy,&mat->lfwork,&info)); 679 ierr = PetscFPTrapPop();CHKERRQ(ierr); 680 mat->lfwork = (PetscInt)PetscRealPart(dummy); 681 ierr = PetscMalloc1(mat->lfwork,&mat->fwork);CHKERRQ(ierr); 682 ierr = PetscLogObjectMemory((PetscObject)A,mat->lfwork*sizeof(PetscBLASInt));CHKERRQ(ierr); 683 } 684 ierr = PetscFPTrapPush(PETSC_FP_TRAP_OFF);CHKERRQ(ierr); 685 PetscStackCallBLAS("LAPACKsytrf",LAPACKsytrf_("L",&n,mat->v,&mat->lda,mat->pivots,mat->fwork,&mat->lfwork,&info)); 686 ierr = PetscFPTrapPop();CHKERRQ(ierr); 687 } 688 if (info) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_MAT_CH_ZRPVT,"Bad factorization: zero pivot in row %D",(PetscInt)info-1); 689 690 A->ops->solve = MatSolve_SeqDense; 691 A->ops->matsolve = MatMatSolve_SeqDense; 692 A->ops->solvetranspose = MatSolveTranspose_SeqDense; 693 A->factortype = MAT_FACTOR_CHOLESKY; 694 695 ierr = PetscFree(A->solvertype);CHKERRQ(ierr); 696 ierr = PetscStrallocpy(MATSOLVERPETSC,&A->solvertype);CHKERRQ(ierr); 697 698 ierr = PetscLogFlops((1.0*A->cmap->n*A->cmap->n*A->cmap->n)/3.0);CHKERRQ(ierr); 699 PetscFunctionReturn(0); 700 } 701 702 PetscErrorCode MatCholeskyFactorNumeric_SeqDense(Mat fact,Mat A,const MatFactorInfo *info_dummy) 703 { 704 PetscErrorCode ierr; 705 MatFactorInfo info; 706 707 PetscFunctionBegin; 708 info.fill = 1.0; 709 710 ierr = MatDuplicateNoCreate_SeqDense(fact,A,MAT_COPY_VALUES);CHKERRQ(ierr); 711 ierr = (*fact->ops->choleskyfactor)(fact,0,&info);CHKERRQ(ierr); 712 PetscFunctionReturn(0); 713 } 714 715 PetscErrorCode MatCholeskyFactorSymbolic_SeqDense(Mat fact,Mat A,IS row,const MatFactorInfo *info) 716 { 717 PetscFunctionBegin; 718 fact->assembled = PETSC_TRUE; 719 fact->preallocated = PETSC_TRUE; 720 fact->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqDense; 721 fact->ops->solve = MatSolve_SeqDense; 722 fact->ops->matsolve = MatMatSolve_SeqDense; 723 fact->ops->solvetranspose = MatSolveTranspose_SeqDense; 724 PetscFunctionReturn(0); 725 } 726 727 PetscErrorCode MatLUFactorSymbolic_SeqDense(Mat fact,Mat A,IS row,IS col,const MatFactorInfo *info) 728 { 729 PetscFunctionBegin; 730 fact->preallocated = PETSC_TRUE; 731 fact->assembled = PETSC_TRUE; 732 fact->ops->lufactornumeric = MatLUFactorNumeric_SeqDense; 733 fact->ops->solve = MatSolve_SeqDense; 734 fact->ops->matsolve = MatMatSolve_SeqDense; 735 fact->ops->solvetranspose = MatSolveTranspose_SeqDense; 736 PetscFunctionReturn(0); 737 } 738 739 /* uses LAPACK */ 740 PETSC_INTERN PetscErrorCode MatGetFactor_seqdense_petsc(Mat A,MatFactorType ftype,Mat *fact) 741 { 742 PetscErrorCode ierr; 743 744 PetscFunctionBegin; 745 ierr = MatCreate(PetscObjectComm((PetscObject)A),fact);CHKERRQ(ierr); 746 ierr = MatSetSizes(*fact,A->rmap->n,A->cmap->n,A->rmap->n,A->cmap->n);CHKERRQ(ierr); 747 ierr = MatSetType(*fact,MATDENSE);CHKERRQ(ierr); 748 if (ftype == MAT_FACTOR_LU) { 749 (*fact)->ops->lufactorsymbolic = MatLUFactorSymbolic_SeqDense; 750 } else { 751 (*fact)->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SeqDense; 752 } 753 (*fact)->factortype = ftype; 754 755 ierr = PetscFree((*fact)->solvertype);CHKERRQ(ierr); 756 ierr = PetscStrallocpy(MATSOLVERPETSC,&(*fact)->solvertype);CHKERRQ(ierr); 757 PetscFunctionReturn(0); 758 } 759 760 /* ------------------------------------------------------------------*/ 761 static PetscErrorCode MatSOR_SeqDense(Mat A,Vec bb,PetscReal omega,MatSORType flag,PetscReal shift,PetscInt its,PetscInt lits,Vec xx) 762 { 763 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 764 PetscScalar *x,*v = mat->v,zero = 0.0,xt; 765 const PetscScalar *b; 766 PetscErrorCode ierr; 767 PetscInt m = A->rmap->n,i; 768 PetscBLASInt o = 1,bm; 769 770 PetscFunctionBegin; 771 #if defined(PETSC_HAVE_CUDA) 772 if (A->offloadmask == PETSC_OFFLOAD_GPU) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not implemented"); 773 #endif 774 if (shift == -1) shift = 0.0; /* negative shift indicates do not error on zero diagonal; this code never zeros on zero diagonal */ 775 ierr = PetscBLASIntCast(m,&bm);CHKERRQ(ierr); 776 if (flag & SOR_ZERO_INITIAL_GUESS) { 777 /* this is a hack fix, should have another version without the second BLASdotu */ 778 ierr = VecSet(xx,zero);CHKERRQ(ierr); 779 } 780 ierr = VecGetArray(xx,&x);CHKERRQ(ierr); 781 ierr = VecGetArrayRead(bb,&b);CHKERRQ(ierr); 782 its = its*lits; 783 if (its <= 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Relaxation requires global its %D and local its %D both positive",its,lits); 784 while (its--) { 785 if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) { 786 for (i=0; i<m; i++) { 787 PetscStackCallBLAS("BLASdotu",xt = b[i] - BLASdotu_(&bm,v+i,&bm,x,&o)); 788 x[i] = (1. - omega)*x[i] + omega*(xt+v[i + i*m]*x[i])/(v[i + i*m]+shift); 789 } 790 } 791 if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP) { 792 for (i=m-1; i>=0; i--) { 793 PetscStackCallBLAS("BLASdotu",xt = b[i] - BLASdotu_(&bm,v+i,&bm,x,&o)); 794 x[i] = (1. - omega)*x[i] + omega*(xt+v[i + i*m]*x[i])/(v[i + i*m]+shift); 795 } 796 } 797 } 798 ierr = VecRestoreArrayRead(bb,&b);CHKERRQ(ierr); 799 ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr); 800 PetscFunctionReturn(0); 801 } 802 803 /* -----------------------------------------------------------------*/ 804 PetscErrorCode MatMultTranspose_SeqDense(Mat A,Vec xx,Vec yy) 805 { 806 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 807 const PetscScalar *v = mat->v,*x; 808 PetscScalar *y; 809 PetscErrorCode ierr; 810 PetscBLASInt m, n,_One=1; 811 PetscScalar _DOne=1.0,_DZero=0.0; 812 813 PetscFunctionBegin; 814 ierr = PetscBLASIntCast(A->rmap->n,&m);CHKERRQ(ierr); 815 ierr = PetscBLASIntCast(A->cmap->n,&n);CHKERRQ(ierr); 816 ierr = VecGetArrayRead(xx,&x);CHKERRQ(ierr); 817 ierr = VecGetArrayWrite(yy,&y);CHKERRQ(ierr); 818 if (!A->rmap->n || !A->cmap->n) { 819 PetscBLASInt i; 820 for (i=0; i<n; i++) y[i] = 0.0; 821 } else { 822 PetscStackCallBLAS("BLASgemv",BLASgemv_("T",&m,&n,&_DOne,v,&mat->lda,x,&_One,&_DZero,y,&_One)); 823 ierr = PetscLogFlops(2.0*A->rmap->n*A->cmap->n - A->cmap->n);CHKERRQ(ierr); 824 } 825 ierr = VecRestoreArrayRead(xx,&x);CHKERRQ(ierr); 826 ierr = VecRestoreArrayWrite(yy,&y);CHKERRQ(ierr); 827 PetscFunctionReturn(0); 828 } 829 830 PetscErrorCode MatMult_SeqDense(Mat A,Vec xx,Vec yy) 831 { 832 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 833 PetscScalar *y,_DOne=1.0,_DZero=0.0; 834 PetscErrorCode ierr; 835 PetscBLASInt m, n, _One=1; 836 const PetscScalar *v = mat->v,*x; 837 838 PetscFunctionBegin; 839 ierr = PetscBLASIntCast(A->rmap->n,&m);CHKERRQ(ierr); 840 ierr = PetscBLASIntCast(A->cmap->n,&n);CHKERRQ(ierr); 841 ierr = VecGetArrayRead(xx,&x);CHKERRQ(ierr); 842 ierr = VecGetArrayWrite(yy,&y);CHKERRQ(ierr); 843 if (!A->rmap->n || !A->cmap->n) { 844 PetscBLASInt i; 845 for (i=0; i<m; i++) y[i] = 0.0; 846 } else { 847 PetscStackCallBLAS("BLASgemv",BLASgemv_("N",&m,&n,&_DOne,v,&(mat->lda),x,&_One,&_DZero,y,&_One)); 848 ierr = PetscLogFlops(2.0*A->rmap->n*A->cmap->n - A->rmap->n);CHKERRQ(ierr); 849 } 850 ierr = VecRestoreArrayRead(xx,&x);CHKERRQ(ierr); 851 ierr = VecRestoreArrayWrite(yy,&y);CHKERRQ(ierr); 852 PetscFunctionReturn(0); 853 } 854 855 PetscErrorCode MatMultAdd_SeqDense(Mat A,Vec xx,Vec zz,Vec yy) 856 { 857 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 858 const PetscScalar *v = mat->v,*x; 859 PetscScalar *y,_DOne=1.0; 860 PetscErrorCode ierr; 861 PetscBLASInt m, n, _One=1; 862 863 PetscFunctionBegin; 864 ierr = PetscBLASIntCast(A->rmap->n,&m);CHKERRQ(ierr); 865 ierr = PetscBLASIntCast(A->cmap->n,&n);CHKERRQ(ierr); 866 if (!A->rmap->n || !A->cmap->n) PetscFunctionReturn(0); 867 if (zz != yy) {ierr = VecCopy(zz,yy);CHKERRQ(ierr);} 868 ierr = VecGetArrayRead(xx,&x);CHKERRQ(ierr); 869 ierr = VecGetArray(yy,&y);CHKERRQ(ierr); 870 PetscStackCallBLAS("BLASgemv",BLASgemv_("N",&m,&n,&_DOne,v,&(mat->lda),x,&_One,&_DOne,y,&_One)); 871 ierr = VecRestoreArrayRead(xx,&x);CHKERRQ(ierr); 872 ierr = VecRestoreArray(yy,&y);CHKERRQ(ierr); 873 ierr = PetscLogFlops(2.0*A->rmap->n*A->cmap->n);CHKERRQ(ierr); 874 PetscFunctionReturn(0); 875 } 876 877 PetscErrorCode MatMultTransposeAdd_SeqDense(Mat A,Vec xx,Vec zz,Vec yy) 878 { 879 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 880 const PetscScalar *v = mat->v,*x; 881 PetscScalar *y; 882 PetscErrorCode ierr; 883 PetscBLASInt m, n, _One=1; 884 PetscScalar _DOne=1.0; 885 886 PetscFunctionBegin; 887 ierr = PetscBLASIntCast(A->rmap->n,&m);CHKERRQ(ierr); 888 ierr = PetscBLASIntCast(A->cmap->n,&n);CHKERRQ(ierr); 889 if (!A->rmap->n || !A->cmap->n) PetscFunctionReturn(0); 890 if (zz != yy) {ierr = VecCopy(zz,yy);CHKERRQ(ierr);} 891 ierr = VecGetArrayRead(xx,&x);CHKERRQ(ierr); 892 ierr = VecGetArray(yy,&y);CHKERRQ(ierr); 893 PetscStackCallBLAS("BLASgemv",BLASgemv_("T",&m,&n,&_DOne,v,&(mat->lda),x,&_One,&_DOne,y,&_One)); 894 ierr = VecRestoreArrayRead(xx,&x);CHKERRQ(ierr); 895 ierr = VecRestoreArray(yy,&y);CHKERRQ(ierr); 896 ierr = PetscLogFlops(2.0*A->rmap->n*A->cmap->n);CHKERRQ(ierr); 897 PetscFunctionReturn(0); 898 } 899 900 /* -----------------------------------------------------------------*/ 901 static PetscErrorCode MatGetRow_SeqDense(Mat A,PetscInt row,PetscInt *ncols,PetscInt **cols,PetscScalar **vals) 902 { 903 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 904 PetscErrorCode ierr; 905 PetscInt i; 906 907 PetscFunctionBegin; 908 *ncols = A->cmap->n; 909 if (cols) { 910 ierr = PetscMalloc1(A->cmap->n+1,cols);CHKERRQ(ierr); 911 for (i=0; i<A->cmap->n; i++) (*cols)[i] = i; 912 } 913 if (vals) { 914 const PetscScalar *v; 915 916 ierr = MatDenseGetArrayRead(A,&v);CHKERRQ(ierr); 917 ierr = PetscMalloc1(A->cmap->n+1,vals);CHKERRQ(ierr); 918 v += row; 919 for (i=0; i<A->cmap->n; i++) {(*vals)[i] = *v; v += mat->lda;} 920 ierr = MatDenseRestoreArrayRead(A,&v);CHKERRQ(ierr); 921 } 922 PetscFunctionReturn(0); 923 } 924 925 static PetscErrorCode MatRestoreRow_SeqDense(Mat A,PetscInt row,PetscInt *ncols,PetscInt **cols,PetscScalar **vals) 926 { 927 PetscErrorCode ierr; 928 929 PetscFunctionBegin; 930 if (cols) {ierr = PetscFree(*cols);CHKERRQ(ierr);} 931 if (vals) {ierr = PetscFree(*vals);CHKERRQ(ierr); } 932 PetscFunctionReturn(0); 933 } 934 /* ----------------------------------------------------------------*/ 935 static PetscErrorCode MatSetValues_SeqDense(Mat A,PetscInt m,const PetscInt indexm[],PetscInt n,const PetscInt indexn[],const PetscScalar v[],InsertMode addv) 936 { 937 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 938 PetscScalar *av; 939 PetscInt i,j,idx=0; 940 #if defined(PETSC_HAVE_CUDA) 941 PetscOffloadMask oldf; 942 #endif 943 PetscErrorCode ierr; 944 945 PetscFunctionBegin; 946 ierr = MatDenseGetArray(A,&av);CHKERRQ(ierr); 947 if (!mat->roworiented) { 948 if (addv == INSERT_VALUES) { 949 for (j=0; j<n; j++) { 950 if (indexn[j] < 0) {idx += m; continue;} 951 if (PetscUnlikelyDebug(indexn[j] >= A->cmap->n)) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",indexn[j],A->cmap->n-1); 952 for (i=0; i<m; i++) { 953 if (indexm[i] < 0) {idx++; continue;} 954 if (PetscUnlikelyDebug(indexm[i] >= A->rmap->n)) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",indexm[i],A->rmap->n-1); 955 av[indexn[j]*mat->lda + indexm[i]] = v[idx++]; 956 } 957 } 958 } else { 959 for (j=0; j<n; j++) { 960 if (indexn[j] < 0) {idx += m; continue;} 961 if (PetscUnlikelyDebug(indexn[j] >= A->cmap->n)) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",indexn[j],A->cmap->n-1); 962 for (i=0; i<m; i++) { 963 if (indexm[i] < 0) {idx++; continue;} 964 if (PetscUnlikelyDebug(indexm[i] >= A->rmap->n)) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",indexm[i],A->rmap->n-1); 965 av[indexn[j]*mat->lda + indexm[i]] += v[idx++]; 966 } 967 } 968 } 969 } else { 970 if (addv == INSERT_VALUES) { 971 for (i=0; i<m; i++) { 972 if (indexm[i] < 0) { idx += n; continue;} 973 if (PetscUnlikelyDebug(indexm[i] >= A->rmap->n)) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",indexm[i],A->rmap->n-1); 974 for (j=0; j<n; j++) { 975 if (indexn[j] < 0) { idx++; continue;} 976 if (PetscUnlikelyDebug(indexn[j] >= A->cmap->n)) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",indexn[j],A->cmap->n-1); 977 av[indexn[j]*mat->lda + indexm[i]] = v[idx++]; 978 } 979 } 980 } else { 981 for (i=0; i<m; i++) { 982 if (indexm[i] < 0) { idx += n; continue;} 983 if (PetscUnlikelyDebug(indexm[i] >= A->rmap->n)) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",indexm[i],A->rmap->n-1); 984 for (j=0; j<n; j++) { 985 if (indexn[j] < 0) { idx++; continue;} 986 if (PetscUnlikelyDebug(indexn[j] >= A->cmap->n)) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",indexn[j],A->cmap->n-1); 987 av[indexn[j]*mat->lda + indexm[i]] += v[idx++]; 988 } 989 } 990 } 991 } 992 /* hack to prevent unneeded copy to the GPU while returning the array */ 993 #if defined(PETSC_HAVE_CUDA) 994 oldf = A->offloadmask; 995 A->offloadmask = PETSC_OFFLOAD_GPU; 996 #endif 997 ierr = MatDenseRestoreArray(A,&av);CHKERRQ(ierr); 998 #if defined(PETSC_HAVE_CUDA) 999 A->offloadmask = (oldf == PETSC_OFFLOAD_UNALLOCATED ? PETSC_OFFLOAD_UNALLOCATED : PETSC_OFFLOAD_CPU); 1000 #endif 1001 PetscFunctionReturn(0); 1002 } 1003 1004 static PetscErrorCode MatGetValues_SeqDense(Mat A,PetscInt m,const PetscInt indexm[],PetscInt n,const PetscInt indexn[],PetscScalar v[]) 1005 { 1006 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 1007 const PetscScalar *vv; 1008 PetscInt i,j; 1009 PetscErrorCode ierr; 1010 1011 PetscFunctionBegin; 1012 ierr = MatDenseGetArrayRead(A,&vv);CHKERRQ(ierr); 1013 /* row-oriented output */ 1014 for (i=0; i<m; i++) { 1015 if (indexm[i] < 0) {v += n;continue;} 1016 if (indexm[i] >= A->rmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row %D requested larger than number rows %D",indexm[i],A->rmap->n); 1017 for (j=0; j<n; j++) { 1018 if (indexn[j] < 0) {v++; continue;} 1019 if (indexn[j] >= A->cmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column %D requested larger than number columns %D",indexn[j],A->cmap->n); 1020 *v++ = vv[indexn[j]*mat->lda + indexm[i]]; 1021 } 1022 } 1023 ierr = MatDenseRestoreArrayRead(A,&vv);CHKERRQ(ierr); 1024 PetscFunctionReturn(0); 1025 } 1026 1027 /* -----------------------------------------------------------------*/ 1028 1029 PetscErrorCode MatView_Dense_Binary(Mat mat,PetscViewer viewer) 1030 { 1031 PetscErrorCode ierr; 1032 PetscBool skipHeader; 1033 PetscViewerFormat format; 1034 PetscInt header[4],M,N,m,lda,i,j,k; 1035 const PetscScalar *v; 1036 PetscScalar *vwork; 1037 1038 PetscFunctionBegin; 1039 ierr = PetscViewerSetUp(viewer);CHKERRQ(ierr); 1040 ierr = PetscViewerBinaryGetSkipHeader(viewer,&skipHeader);CHKERRQ(ierr); 1041 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 1042 if (skipHeader) format = PETSC_VIEWER_NATIVE; 1043 1044 ierr = MatGetSize(mat,&M,&N);CHKERRQ(ierr); 1045 1046 /* write matrix header */ 1047 header[0] = MAT_FILE_CLASSID; header[1] = M; header[2] = N; 1048 header[3] = (format == PETSC_VIEWER_NATIVE) ? MATRIX_BINARY_FORMAT_DENSE : M*N; 1049 if (!skipHeader) {ierr = PetscViewerBinaryWrite(viewer,header,4,PETSC_INT);CHKERRQ(ierr);} 1050 1051 ierr = MatGetLocalSize(mat,&m,NULL);CHKERRQ(ierr); 1052 if (format != PETSC_VIEWER_NATIVE) { 1053 PetscInt nnz = m*N, *iwork; 1054 /* store row lengths for each row */ 1055 ierr = PetscMalloc1(nnz,&iwork);CHKERRQ(ierr); 1056 for (i=0; i<m; i++) iwork[i] = N; 1057 ierr = PetscViewerBinaryWriteAll(viewer,iwork,m,PETSC_DETERMINE,PETSC_DETERMINE,PETSC_INT);CHKERRQ(ierr); 1058 /* store column indices (zero start index) */ 1059 for (k=0, i=0; i<m; i++) 1060 for (j=0; j<N; j++, k++) 1061 iwork[k] = j; 1062 ierr = PetscViewerBinaryWriteAll(viewer,iwork,nnz,PETSC_DETERMINE,PETSC_DETERMINE,PETSC_INT);CHKERRQ(ierr); 1063 ierr = PetscFree(iwork);CHKERRQ(ierr); 1064 } 1065 /* store matrix values as a dense matrix in row major order */ 1066 ierr = PetscMalloc1(m*N,&vwork);CHKERRQ(ierr); 1067 ierr = MatDenseGetArrayRead(mat,&v);CHKERRQ(ierr); 1068 ierr = MatDenseGetLDA(mat,&lda);CHKERRQ(ierr); 1069 for (k=0, i=0; i<m; i++) 1070 for (j=0; j<N; j++, k++) 1071 vwork[k] = v[i+lda*j]; 1072 ierr = MatDenseRestoreArrayRead(mat,&v);CHKERRQ(ierr); 1073 ierr = PetscViewerBinaryWriteAll(viewer,vwork,m*N,PETSC_DETERMINE,PETSC_DETERMINE,PETSC_SCALAR);CHKERRQ(ierr); 1074 ierr = PetscFree(vwork);CHKERRQ(ierr); 1075 PetscFunctionReturn(0); 1076 } 1077 1078 PetscErrorCode MatLoad_Dense_Binary(Mat mat,PetscViewer viewer) 1079 { 1080 PetscErrorCode ierr; 1081 PetscBool skipHeader; 1082 PetscInt header[4],M,N,m,nz,lda,i,j,k; 1083 PetscInt rows,cols; 1084 PetscScalar *v,*vwork; 1085 1086 PetscFunctionBegin; 1087 ierr = PetscViewerSetUp(viewer);CHKERRQ(ierr); 1088 ierr = PetscViewerBinaryGetSkipHeader(viewer,&skipHeader);CHKERRQ(ierr); 1089 1090 if (!skipHeader) { 1091 ierr = PetscViewerBinaryRead(viewer,header,4,NULL,PETSC_INT);CHKERRQ(ierr); 1092 if (header[0] != MAT_FILE_CLASSID) SETERRQ(PetscObjectComm((PetscObject)viewer),PETSC_ERR_FILE_UNEXPECTED,"Not a matrix object in file"); 1093 M = header[1]; N = header[2]; 1094 if (M < 0) SETERRQ1(PetscObjectComm((PetscObject)viewer),PETSC_ERR_FILE_UNEXPECTED,"Matrix row size (%D) in file is negative",M); 1095 if (N < 0) SETERRQ1(PetscObjectComm((PetscObject)viewer),PETSC_ERR_FILE_UNEXPECTED,"Matrix column size (%D) in file is negative",N); 1096 nz = header[3]; 1097 if (nz != MATRIX_BINARY_FORMAT_DENSE && nz < 0) SETERRQ1(PetscObjectComm((PetscObject)viewer),PETSC_ERR_FILE_UNEXPECTED,"Unknown matrix format %D in file",nz); 1098 } else { 1099 ierr = MatGetSize(mat,&M,&N);CHKERRQ(ierr); 1100 if (M < 0 || N < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_USER,"Matrix binary file header was skipped, thus the user must specify the global sizes of input matrix"); 1101 nz = MATRIX_BINARY_FORMAT_DENSE; 1102 } 1103 1104 /* setup global sizes if not set */ 1105 if (mat->rmap->N < 0) mat->rmap->N = M; 1106 if (mat->cmap->N < 0) mat->cmap->N = N; 1107 ierr = MatSetUp(mat);CHKERRQ(ierr); 1108 /* check if global sizes are correct */ 1109 ierr = MatGetSize(mat,&rows,&cols);CHKERRQ(ierr); 1110 if (M != rows || N != cols) SETERRQ4(PetscObjectComm((PetscObject)viewer),PETSC_ERR_FILE_UNEXPECTED, "Matrix in file of different sizes (%d, %d) than the input matrix (%d, %d)",M,N,rows,cols); 1111 1112 ierr = MatGetSize(mat,NULL,&N);CHKERRQ(ierr); 1113 ierr = MatGetLocalSize(mat,&m,NULL);CHKERRQ(ierr); 1114 ierr = MatDenseGetArray(mat,&v);CHKERRQ(ierr); 1115 ierr = MatDenseGetLDA(mat,&lda);CHKERRQ(ierr); 1116 if (nz == MATRIX_BINARY_FORMAT_DENSE) { /* matrix in file is dense format */ 1117 PetscInt nnz = m*N; 1118 /* read in matrix values */ 1119 ierr = PetscMalloc1(nnz,&vwork);CHKERRQ(ierr); 1120 ierr = PetscViewerBinaryReadAll(viewer,vwork,nnz,PETSC_DETERMINE,PETSC_DETERMINE,PETSC_SCALAR);CHKERRQ(ierr); 1121 /* store values in column major order */ 1122 for (j=0; j<N; j++) 1123 for (i=0; i<m; i++) 1124 v[i+lda*j] = vwork[i*N+j]; 1125 ierr = PetscFree(vwork);CHKERRQ(ierr); 1126 } else { /* matrix in file is sparse format */ 1127 PetscInt nnz = 0, *rlens, *icols; 1128 /* read in row lengths */ 1129 ierr = PetscMalloc1(m,&rlens);CHKERRQ(ierr); 1130 ierr = PetscViewerBinaryReadAll(viewer,rlens,m,PETSC_DETERMINE,PETSC_DETERMINE,PETSC_INT);CHKERRQ(ierr); 1131 for (i=0; i<m; i++) nnz += rlens[i]; 1132 /* read in column indices and values */ 1133 ierr = PetscMalloc2(nnz,&icols,nnz,&vwork);CHKERRQ(ierr); 1134 ierr = PetscViewerBinaryReadAll(viewer,icols,nnz,PETSC_DETERMINE,PETSC_DETERMINE,PETSC_INT);CHKERRQ(ierr); 1135 ierr = PetscViewerBinaryReadAll(viewer,vwork,nnz,PETSC_DETERMINE,PETSC_DETERMINE,PETSC_SCALAR);CHKERRQ(ierr); 1136 /* store values in column major order */ 1137 for (k=0, i=0; i<m; i++) 1138 for (j=0; j<rlens[i]; j++, k++) 1139 v[i+lda*icols[k]] = vwork[k]; 1140 ierr = PetscFree(rlens);CHKERRQ(ierr); 1141 ierr = PetscFree2(icols,vwork);CHKERRQ(ierr); 1142 } 1143 ierr = MatDenseRestoreArray(mat,&v);CHKERRQ(ierr); 1144 ierr = MatAssemblyBegin(mat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1145 ierr = MatAssemblyEnd(mat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1146 PetscFunctionReturn(0); 1147 } 1148 1149 PetscErrorCode MatLoad_SeqDense(Mat newMat, PetscViewer viewer) 1150 { 1151 PetscBool isbinary, ishdf5; 1152 PetscErrorCode ierr; 1153 1154 PetscFunctionBegin; 1155 PetscValidHeaderSpecific(newMat,MAT_CLASSID,1); 1156 PetscValidHeaderSpecific(viewer,PETSC_VIEWER_CLASSID,2); 1157 /* force binary viewer to load .info file if it has not yet done so */ 1158 ierr = PetscViewerSetUp(viewer);CHKERRQ(ierr); 1159 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);CHKERRQ(ierr); 1160 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERHDF5, &ishdf5);CHKERRQ(ierr); 1161 if (isbinary) { 1162 ierr = MatLoad_Dense_Binary(newMat,viewer);CHKERRQ(ierr); 1163 } else if (ishdf5) { 1164 #if defined(PETSC_HAVE_HDF5) 1165 ierr = MatLoad_Dense_HDF5(newMat,viewer);CHKERRQ(ierr); 1166 #else 1167 SETERRQ(PetscObjectComm((PetscObject)newMat),PETSC_ERR_SUP,"HDF5 not supported in this build.\nPlease reconfigure using --download-hdf5"); 1168 #endif 1169 } else { 1170 SETERRQ2(PetscObjectComm((PetscObject)newMat),PETSC_ERR_SUP,"Viewer type %s not yet supported for reading %s matrices",((PetscObject)viewer)->type_name,((PetscObject)newMat)->type_name); 1171 } 1172 PetscFunctionReturn(0); 1173 } 1174 1175 static PetscErrorCode MatView_SeqDense_ASCII(Mat A,PetscViewer viewer) 1176 { 1177 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 1178 PetscErrorCode ierr; 1179 PetscInt i,j; 1180 const char *name; 1181 PetscScalar *v,*av; 1182 PetscViewerFormat format; 1183 #if defined(PETSC_USE_COMPLEX) 1184 PetscBool allreal = PETSC_TRUE; 1185 #endif 1186 1187 PetscFunctionBegin; 1188 ierr = MatDenseGetArrayRead(A,(const PetscScalar**)&av);CHKERRQ(ierr); 1189 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 1190 if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) { 1191 PetscFunctionReturn(0); /* do nothing for now */ 1192 } else if (format == PETSC_VIEWER_ASCII_COMMON) { 1193 ierr = PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);CHKERRQ(ierr); 1194 for (i=0; i<A->rmap->n; i++) { 1195 v = av + i; 1196 ierr = PetscViewerASCIIPrintf(viewer,"row %D:",i);CHKERRQ(ierr); 1197 for (j=0; j<A->cmap->n; j++) { 1198 #if defined(PETSC_USE_COMPLEX) 1199 if (PetscRealPart(*v) != 0.0 && PetscImaginaryPart(*v) != 0.0) { 1200 ierr = PetscViewerASCIIPrintf(viewer," (%D, %g + %g i) ",j,(double)PetscRealPart(*v),(double)PetscImaginaryPart(*v));CHKERRQ(ierr); 1201 } else if (PetscRealPart(*v)) { 1202 ierr = PetscViewerASCIIPrintf(viewer," (%D, %g) ",j,(double)PetscRealPart(*v));CHKERRQ(ierr); 1203 } 1204 #else 1205 if (*v) { 1206 ierr = PetscViewerASCIIPrintf(viewer," (%D, %g) ",j,(double)*v);CHKERRQ(ierr); 1207 } 1208 #endif 1209 v += a->lda; 1210 } 1211 ierr = PetscViewerASCIIPrintf(viewer,"\n");CHKERRQ(ierr); 1212 } 1213 ierr = PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);CHKERRQ(ierr); 1214 } else { 1215 ierr = PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);CHKERRQ(ierr); 1216 #if defined(PETSC_USE_COMPLEX) 1217 /* determine if matrix has all real values */ 1218 v = av; 1219 for (i=0; i<A->rmap->n*A->cmap->n; i++) { 1220 if (PetscImaginaryPart(v[i])) { allreal = PETSC_FALSE; break;} 1221 } 1222 #endif 1223 if (format == PETSC_VIEWER_ASCII_MATLAB) { 1224 ierr = PetscObjectGetName((PetscObject)A,&name);CHKERRQ(ierr); 1225 ierr = PetscViewerASCIIPrintf(viewer,"%% Size = %D %D \n",A->rmap->n,A->cmap->n);CHKERRQ(ierr); 1226 ierr = PetscViewerASCIIPrintf(viewer,"%s = zeros(%D,%D);\n",name,A->rmap->n,A->cmap->n);CHKERRQ(ierr); 1227 ierr = PetscViewerASCIIPrintf(viewer,"%s = [\n",name);CHKERRQ(ierr); 1228 } 1229 1230 for (i=0; i<A->rmap->n; i++) { 1231 v = av + i; 1232 for (j=0; j<A->cmap->n; j++) { 1233 #if defined(PETSC_USE_COMPLEX) 1234 if (allreal) { 1235 ierr = PetscViewerASCIIPrintf(viewer,"%18.16e ",(double)PetscRealPart(*v));CHKERRQ(ierr); 1236 } else { 1237 ierr = PetscViewerASCIIPrintf(viewer,"%18.16e + %18.16ei ",(double)PetscRealPart(*v),(double)PetscImaginaryPart(*v));CHKERRQ(ierr); 1238 } 1239 #else 1240 ierr = PetscViewerASCIIPrintf(viewer,"%18.16e ",(double)*v);CHKERRQ(ierr); 1241 #endif 1242 v += a->lda; 1243 } 1244 ierr = PetscViewerASCIIPrintf(viewer,"\n");CHKERRQ(ierr); 1245 } 1246 if (format == PETSC_VIEWER_ASCII_MATLAB) { 1247 ierr = PetscViewerASCIIPrintf(viewer,"];\n");CHKERRQ(ierr); 1248 } 1249 ierr = PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);CHKERRQ(ierr); 1250 } 1251 ierr = MatDenseRestoreArrayRead(A,(const PetscScalar**)&av);CHKERRQ(ierr); 1252 ierr = PetscViewerFlush(viewer);CHKERRQ(ierr); 1253 PetscFunctionReturn(0); 1254 } 1255 1256 #include <petscdraw.h> 1257 static PetscErrorCode MatView_SeqDense_Draw_Zoom(PetscDraw draw,void *Aa) 1258 { 1259 Mat A = (Mat) Aa; 1260 PetscErrorCode ierr; 1261 PetscInt m = A->rmap->n,n = A->cmap->n,i,j; 1262 int color = PETSC_DRAW_WHITE; 1263 const PetscScalar *v; 1264 PetscViewer viewer; 1265 PetscReal xl,yl,xr,yr,x_l,x_r,y_l,y_r; 1266 PetscViewerFormat format; 1267 1268 PetscFunctionBegin; 1269 ierr = PetscObjectQuery((PetscObject)A,"Zoomviewer",(PetscObject*)&viewer);CHKERRQ(ierr); 1270 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 1271 ierr = PetscDrawGetCoordinates(draw,&xl,&yl,&xr,&yr);CHKERRQ(ierr); 1272 1273 /* Loop over matrix elements drawing boxes */ 1274 ierr = MatDenseGetArrayRead(A,&v);CHKERRQ(ierr); 1275 if (format != PETSC_VIEWER_DRAW_CONTOUR) { 1276 ierr = PetscDrawCollectiveBegin(draw);CHKERRQ(ierr); 1277 /* Blue for negative and Red for positive */ 1278 for (j = 0; j < n; j++) { 1279 x_l = j; x_r = x_l + 1.0; 1280 for (i = 0; i < m; i++) { 1281 y_l = m - i - 1.0; 1282 y_r = y_l + 1.0; 1283 if (PetscRealPart(v[j*m+i]) > 0.) color = PETSC_DRAW_RED; 1284 else if (PetscRealPart(v[j*m+i]) < 0.) color = PETSC_DRAW_BLUE; 1285 else continue; 1286 ierr = PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);CHKERRQ(ierr); 1287 } 1288 } 1289 ierr = PetscDrawCollectiveEnd(draw);CHKERRQ(ierr); 1290 } else { 1291 /* use contour shading to indicate magnitude of values */ 1292 /* first determine max of all nonzero values */ 1293 PetscReal minv = 0.0, maxv = 0.0; 1294 PetscDraw popup; 1295 1296 for (i=0; i < m*n; i++) { 1297 if (PetscAbsScalar(v[i]) > maxv) maxv = PetscAbsScalar(v[i]); 1298 } 1299 if (minv >= maxv) maxv = minv + PETSC_SMALL; 1300 ierr = PetscDrawGetPopup(draw,&popup);CHKERRQ(ierr); 1301 ierr = PetscDrawScalePopup(popup,minv,maxv);CHKERRQ(ierr); 1302 1303 ierr = PetscDrawCollectiveBegin(draw);CHKERRQ(ierr); 1304 for (j=0; j<n; j++) { 1305 x_l = j; 1306 x_r = x_l + 1.0; 1307 for (i=0; i<m; i++) { 1308 y_l = m - i - 1.0; 1309 y_r = y_l + 1.0; 1310 color = PetscDrawRealToColor(PetscAbsScalar(v[j*m+i]),minv,maxv); 1311 ierr = PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);CHKERRQ(ierr); 1312 } 1313 } 1314 ierr = PetscDrawCollectiveEnd(draw);CHKERRQ(ierr); 1315 } 1316 ierr = MatDenseRestoreArrayRead(A,&v);CHKERRQ(ierr); 1317 PetscFunctionReturn(0); 1318 } 1319 1320 static PetscErrorCode MatView_SeqDense_Draw(Mat A,PetscViewer viewer) 1321 { 1322 PetscDraw draw; 1323 PetscBool isnull; 1324 PetscReal xr,yr,xl,yl,h,w; 1325 PetscErrorCode ierr; 1326 1327 PetscFunctionBegin; 1328 ierr = PetscViewerDrawGetDraw(viewer,0,&draw);CHKERRQ(ierr); 1329 ierr = PetscDrawIsNull(draw,&isnull);CHKERRQ(ierr); 1330 if (isnull) PetscFunctionReturn(0); 1331 1332 xr = A->cmap->n; yr = A->rmap->n; h = yr/10.0; w = xr/10.0; 1333 xr += w; yr += h; xl = -w; yl = -h; 1334 ierr = PetscDrawSetCoordinates(draw,xl,yl,xr,yr);CHKERRQ(ierr); 1335 ierr = PetscObjectCompose((PetscObject)A,"Zoomviewer",(PetscObject)viewer);CHKERRQ(ierr); 1336 ierr = PetscDrawZoom(draw,MatView_SeqDense_Draw_Zoom,A);CHKERRQ(ierr); 1337 ierr = PetscObjectCompose((PetscObject)A,"Zoomviewer",NULL);CHKERRQ(ierr); 1338 ierr = PetscDrawSave(draw);CHKERRQ(ierr); 1339 PetscFunctionReturn(0); 1340 } 1341 1342 PetscErrorCode MatView_SeqDense(Mat A,PetscViewer viewer) 1343 { 1344 PetscErrorCode ierr; 1345 PetscBool iascii,isbinary,isdraw; 1346 1347 PetscFunctionBegin; 1348 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);CHKERRQ(ierr); 1349 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);CHKERRQ(ierr); 1350 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);CHKERRQ(ierr); 1351 1352 if (iascii) { 1353 ierr = MatView_SeqDense_ASCII(A,viewer);CHKERRQ(ierr); 1354 } else if (isbinary) { 1355 ierr = MatView_Dense_Binary(A,viewer);CHKERRQ(ierr); 1356 } else if (isdraw) { 1357 ierr = MatView_SeqDense_Draw(A,viewer);CHKERRQ(ierr); 1358 } 1359 PetscFunctionReturn(0); 1360 } 1361 1362 static PetscErrorCode MatDensePlaceArray_SeqDense(Mat A,const PetscScalar *array) 1363 { 1364 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 1365 1366 PetscFunctionBegin; 1367 if (a->vecinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreColumnVec() first"); 1368 if (a->matinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreSubMatrix() first"); 1369 if (a->unplacedarray) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreArray() first"); 1370 a->unplacedarray = a->v; 1371 a->unplaced_user_alloc = a->user_alloc; 1372 a->v = (PetscScalar*) array; 1373 a->user_alloc = PETSC_TRUE; 1374 #if defined(PETSC_HAVE_CUDA) 1375 A->offloadmask = PETSC_OFFLOAD_CPU; 1376 #endif 1377 PetscFunctionReturn(0); 1378 } 1379 1380 static PetscErrorCode MatDenseResetArray_SeqDense(Mat A) 1381 { 1382 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 1383 1384 PetscFunctionBegin; 1385 if (a->vecinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreColumnVec() first"); 1386 if (a->matinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreSubMatrix() first"); 1387 a->v = a->unplacedarray; 1388 a->user_alloc = a->unplaced_user_alloc; 1389 a->unplacedarray = NULL; 1390 #if defined(PETSC_HAVE_CUDA) 1391 A->offloadmask = PETSC_OFFLOAD_CPU; 1392 #endif 1393 PetscFunctionReturn(0); 1394 } 1395 1396 static PetscErrorCode MatDenseReplaceArray_SeqDense(Mat A,const PetscScalar *array) 1397 { 1398 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 1399 PetscErrorCode ierr; 1400 1401 PetscFunctionBegin; 1402 if (a->vecinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreColumnVec() first"); 1403 if (a->matinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreSubMatrix() first"); 1404 if (!a->user_alloc) { ierr = PetscFree(a->v);CHKERRQ(ierr); } 1405 a->v = (PetscScalar*) array; 1406 a->user_alloc = PETSC_FALSE; 1407 #if defined(PETSC_HAVE_CUDA) 1408 A->offloadmask = PETSC_OFFLOAD_CPU; 1409 #endif 1410 PetscFunctionReturn(0); 1411 } 1412 1413 PetscErrorCode MatDestroy_SeqDense(Mat mat) 1414 { 1415 Mat_SeqDense *l = (Mat_SeqDense*)mat->data; 1416 PetscErrorCode ierr; 1417 1418 PetscFunctionBegin; 1419 #if defined(PETSC_USE_LOG) 1420 PetscLogObjectState((PetscObject)mat,"Rows %D Cols %D",mat->rmap->n,mat->cmap->n); 1421 #endif 1422 ierr = PetscFree(l->pivots);CHKERRQ(ierr); 1423 ierr = PetscFree(l->fwork);CHKERRQ(ierr); 1424 ierr = MatDestroy(&l->ptapwork);CHKERRQ(ierr); 1425 if (!l->user_alloc) {ierr = PetscFree(l->v);CHKERRQ(ierr);} 1426 if (!l->unplaced_user_alloc) {ierr = PetscFree(l->unplacedarray);CHKERRQ(ierr);} 1427 if (l->vecinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreColumnVec() first"); 1428 if (l->matinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreSubMatrix() first"); 1429 ierr = VecDestroy(&l->cvec);CHKERRQ(ierr); 1430 ierr = MatDestroy(&l->cmat);CHKERRQ(ierr); 1431 ierr = PetscFree(mat->data);CHKERRQ(ierr); 1432 1433 ierr = PetscObjectChangeTypeName((PetscObject)mat,0);CHKERRQ(ierr); 1434 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseGetLDA_C",NULL);CHKERRQ(ierr); 1435 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseSetLDA_C",NULL);CHKERRQ(ierr); 1436 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseGetArray_C",NULL);CHKERRQ(ierr); 1437 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseRestoreArray_C",NULL);CHKERRQ(ierr); 1438 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDensePlaceArray_C",NULL);CHKERRQ(ierr); 1439 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseResetArray_C",NULL);CHKERRQ(ierr); 1440 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseReplaceArray_C",NULL);CHKERRQ(ierr); 1441 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseGetArrayRead_C",NULL);CHKERRQ(ierr); 1442 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseRestoreArrayRead_C",NULL);CHKERRQ(ierr); 1443 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseGetArrayWrite_C",NULL);CHKERRQ(ierr); 1444 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseRestoreArrayWrite_C",NULL);CHKERRQ(ierr); 1445 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatConvert_seqdense_seqaij_C",NULL);CHKERRQ(ierr); 1446 #if defined(PETSC_HAVE_ELEMENTAL) 1447 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatConvert_seqdense_elemental_C",NULL);CHKERRQ(ierr); 1448 #endif 1449 #if defined(PETSC_HAVE_CUDA) 1450 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatConvert_seqdense_seqdensecuda_C",NULL);CHKERRQ(ierr); 1451 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatProductSetFromOptions_seqdensecuda_seqdensecuda_C",NULL);CHKERRQ(ierr); 1452 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatProductSetFromOptions_seqdensecuda_seqdense_C",NULL);CHKERRQ(ierr); 1453 #endif 1454 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatSeqDenseSetPreallocation_C",NULL);CHKERRQ(ierr); 1455 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatProductSetFromOptions_seqaij_seqdense_C",NULL);CHKERRQ(ierr); 1456 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatProductSetFromOptions_seqdense_seqdense_C",NULL);CHKERRQ(ierr); 1457 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatProductSetFromOptions_seqbaij_seqdense_C",NULL);CHKERRQ(ierr); 1458 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatProductSetFromOptions_seqsbaij_seqdense_C",NULL);CHKERRQ(ierr); 1459 1460 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseGetColumn_C",NULL);CHKERRQ(ierr); 1461 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseRestoreColumn_C",NULL);CHKERRQ(ierr); 1462 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseGetColumnVec_C",NULL);CHKERRQ(ierr); 1463 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseRestoreColumnVec_C",NULL);CHKERRQ(ierr); 1464 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseGetColumnVecRead_C",NULL);CHKERRQ(ierr); 1465 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseRestoreColumnVecRead_C",NULL);CHKERRQ(ierr); 1466 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseGetColumnVecWrite_C",NULL);CHKERRQ(ierr); 1467 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseRestoreColumnVecWrite_C",NULL);CHKERRQ(ierr); 1468 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseGetSubMatrix_C",NULL);CHKERRQ(ierr); 1469 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatDenseRestoreSubMatrix_C",NULL);CHKERRQ(ierr); 1470 PetscFunctionReturn(0); 1471 } 1472 1473 static PetscErrorCode MatTranspose_SeqDense(Mat A,MatReuse reuse,Mat *matout) 1474 { 1475 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 1476 PetscErrorCode ierr; 1477 PetscInt k,j,m,n,M; 1478 PetscScalar *v,tmp; 1479 1480 PetscFunctionBegin; 1481 m = A->rmap->n; M = mat->lda; n = A->cmap->n; 1482 if (reuse == MAT_INPLACE_MATRIX && m == n) { /* in place transpose */ 1483 ierr = MatDenseGetArray(A,&v);CHKERRQ(ierr); 1484 for (j=0; j<m; j++) { 1485 for (k=0; k<j; k++) { 1486 tmp = v[j + k*M]; 1487 v[j + k*M] = v[k + j*M]; 1488 v[k + j*M] = tmp; 1489 } 1490 } 1491 ierr = MatDenseRestoreArray(A,&v);CHKERRQ(ierr); 1492 } else { /* out-of-place transpose */ 1493 Mat tmat; 1494 Mat_SeqDense *tmatd; 1495 PetscScalar *v2; 1496 PetscInt M2; 1497 1498 if (reuse != MAT_REUSE_MATRIX) { 1499 ierr = MatCreate(PetscObjectComm((PetscObject)A),&tmat);CHKERRQ(ierr); 1500 ierr = MatSetSizes(tmat,A->cmap->n,A->rmap->n,A->cmap->n,A->rmap->n);CHKERRQ(ierr); 1501 ierr = MatSetType(tmat,((PetscObject)A)->type_name);CHKERRQ(ierr); 1502 ierr = MatSeqDenseSetPreallocation(tmat,NULL);CHKERRQ(ierr); 1503 } else tmat = *matout; 1504 1505 ierr = MatDenseGetArrayRead(A,(const PetscScalar**)&v);CHKERRQ(ierr); 1506 ierr = MatDenseGetArray(tmat,&v2);CHKERRQ(ierr); 1507 tmatd = (Mat_SeqDense*)tmat->data; 1508 M2 = tmatd->lda; 1509 for (j=0; j<n; j++) { 1510 for (k=0; k<m; k++) v2[j + k*M2] = v[k + j*M]; 1511 } 1512 ierr = MatDenseRestoreArray(tmat,&v2);CHKERRQ(ierr); 1513 ierr = MatDenseRestoreArrayRead(A,(const PetscScalar**)&v);CHKERRQ(ierr); 1514 ierr = MatAssemblyBegin(tmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1515 ierr = MatAssemblyEnd(tmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1516 if (reuse == MAT_INITIAL_MATRIX || reuse == MAT_REUSE_MATRIX) *matout = tmat; 1517 else { 1518 ierr = MatHeaderMerge(A,&tmat);CHKERRQ(ierr); 1519 } 1520 } 1521 PetscFunctionReturn(0); 1522 } 1523 1524 static PetscErrorCode MatEqual_SeqDense(Mat A1,Mat A2,PetscBool *flg) 1525 { 1526 Mat_SeqDense *mat1 = (Mat_SeqDense*)A1->data; 1527 Mat_SeqDense *mat2 = (Mat_SeqDense*)A2->data; 1528 PetscInt i; 1529 const PetscScalar *v1,*v2; 1530 PetscErrorCode ierr; 1531 1532 PetscFunctionBegin; 1533 if (A1->rmap->n != A2->rmap->n) {*flg = PETSC_FALSE; PetscFunctionReturn(0);} 1534 if (A1->cmap->n != A2->cmap->n) {*flg = PETSC_FALSE; PetscFunctionReturn(0);} 1535 ierr = MatDenseGetArrayRead(A1,&v1);CHKERRQ(ierr); 1536 ierr = MatDenseGetArrayRead(A2,&v2);CHKERRQ(ierr); 1537 for (i=0; i<A1->cmap->n; i++) { 1538 ierr = PetscArraycmp(v1,v2,A1->rmap->n,flg);CHKERRQ(ierr); 1539 if (*flg == PETSC_FALSE) PetscFunctionReturn(0); 1540 v1 += mat1->lda; 1541 v2 += mat2->lda; 1542 } 1543 ierr = MatDenseRestoreArrayRead(A1,&v1);CHKERRQ(ierr); 1544 ierr = MatDenseRestoreArrayRead(A2,&v2);CHKERRQ(ierr); 1545 *flg = PETSC_TRUE; 1546 PetscFunctionReturn(0); 1547 } 1548 1549 static PetscErrorCode MatGetDiagonal_SeqDense(Mat A,Vec v) 1550 { 1551 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 1552 PetscInt i,n,len; 1553 PetscScalar *x; 1554 const PetscScalar *vv; 1555 PetscErrorCode ierr; 1556 1557 PetscFunctionBegin; 1558 ierr = VecGetSize(v,&n);CHKERRQ(ierr); 1559 ierr = VecGetArray(v,&x);CHKERRQ(ierr); 1560 len = PetscMin(A->rmap->n,A->cmap->n); 1561 ierr = MatDenseGetArrayRead(A,&vv);CHKERRQ(ierr); 1562 if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming mat and vec"); 1563 for (i=0; i<len; i++) { 1564 x[i] = vv[i*mat->lda + i]; 1565 } 1566 ierr = MatDenseRestoreArrayRead(A,&vv);CHKERRQ(ierr); 1567 ierr = VecRestoreArray(v,&x);CHKERRQ(ierr); 1568 PetscFunctionReturn(0); 1569 } 1570 1571 static PetscErrorCode MatDiagonalScale_SeqDense(Mat A,Vec ll,Vec rr) 1572 { 1573 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 1574 const PetscScalar *l,*r; 1575 PetscScalar x,*v,*vv; 1576 PetscErrorCode ierr; 1577 PetscInt i,j,m = A->rmap->n,n = A->cmap->n; 1578 1579 PetscFunctionBegin; 1580 ierr = MatDenseGetArray(A,&vv);CHKERRQ(ierr); 1581 if (ll) { 1582 ierr = VecGetSize(ll,&m);CHKERRQ(ierr); 1583 ierr = VecGetArrayRead(ll,&l);CHKERRQ(ierr); 1584 if (m != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Left scaling vec wrong size"); 1585 for (i=0; i<m; i++) { 1586 x = l[i]; 1587 v = vv + i; 1588 for (j=0; j<n; j++) { (*v) *= x; v+= mat->lda;} 1589 } 1590 ierr = VecRestoreArrayRead(ll,&l);CHKERRQ(ierr); 1591 ierr = PetscLogFlops(1.0*n*m);CHKERRQ(ierr); 1592 } 1593 if (rr) { 1594 ierr = VecGetSize(rr,&n);CHKERRQ(ierr); 1595 ierr = VecGetArrayRead(rr,&r);CHKERRQ(ierr); 1596 if (n != A->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Right scaling vec wrong size"); 1597 for (i=0; i<n; i++) { 1598 x = r[i]; 1599 v = vv + i*mat->lda; 1600 for (j=0; j<m; j++) (*v++) *= x; 1601 } 1602 ierr = VecRestoreArrayRead(rr,&r);CHKERRQ(ierr); 1603 ierr = PetscLogFlops(1.0*n*m);CHKERRQ(ierr); 1604 } 1605 ierr = MatDenseRestoreArray(A,&vv);CHKERRQ(ierr); 1606 PetscFunctionReturn(0); 1607 } 1608 1609 PetscErrorCode MatNorm_SeqDense(Mat A,NormType type,PetscReal *nrm) 1610 { 1611 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 1612 PetscScalar *v,*vv; 1613 PetscReal sum = 0.0; 1614 PetscInt lda =mat->lda,m=A->rmap->n,i,j; 1615 PetscErrorCode ierr; 1616 1617 PetscFunctionBegin; 1618 ierr = MatDenseGetArrayRead(A,(const PetscScalar**)&vv);CHKERRQ(ierr); 1619 v = vv; 1620 if (type == NORM_FROBENIUS) { 1621 if (lda>m) { 1622 for (j=0; j<A->cmap->n; j++) { 1623 v = vv+j*lda; 1624 for (i=0; i<m; i++) { 1625 sum += PetscRealPart(PetscConj(*v)*(*v)); v++; 1626 } 1627 } 1628 } else { 1629 #if defined(PETSC_USE_REAL___FP16) 1630 PetscBLASInt one = 1,cnt = A->cmap->n*A->rmap->n; 1631 *nrm = BLASnrm2_(&cnt,v,&one); 1632 } 1633 #else 1634 for (i=0; i<A->cmap->n*A->rmap->n; i++) { 1635 sum += PetscRealPart(PetscConj(*v)*(*v)); v++; 1636 } 1637 } 1638 *nrm = PetscSqrtReal(sum); 1639 #endif 1640 ierr = PetscLogFlops(2.0*A->cmap->n*A->rmap->n);CHKERRQ(ierr); 1641 } else if (type == NORM_1) { 1642 *nrm = 0.0; 1643 for (j=0; j<A->cmap->n; j++) { 1644 v = vv + j*mat->lda; 1645 sum = 0.0; 1646 for (i=0; i<A->rmap->n; i++) { 1647 sum += PetscAbsScalar(*v); v++; 1648 } 1649 if (sum > *nrm) *nrm = sum; 1650 } 1651 ierr = PetscLogFlops(1.0*A->cmap->n*A->rmap->n);CHKERRQ(ierr); 1652 } else if (type == NORM_INFINITY) { 1653 *nrm = 0.0; 1654 for (j=0; j<A->rmap->n; j++) { 1655 v = vv + j; 1656 sum = 0.0; 1657 for (i=0; i<A->cmap->n; i++) { 1658 sum += PetscAbsScalar(*v); v += mat->lda; 1659 } 1660 if (sum > *nrm) *nrm = sum; 1661 } 1662 ierr = PetscLogFlops(1.0*A->cmap->n*A->rmap->n);CHKERRQ(ierr); 1663 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"No two norm"); 1664 ierr = MatDenseRestoreArrayRead(A,(const PetscScalar**)&vv);CHKERRQ(ierr); 1665 PetscFunctionReturn(0); 1666 } 1667 1668 static PetscErrorCode MatSetOption_SeqDense(Mat A,MatOption op,PetscBool flg) 1669 { 1670 Mat_SeqDense *aij = (Mat_SeqDense*)A->data; 1671 PetscErrorCode ierr; 1672 1673 PetscFunctionBegin; 1674 switch (op) { 1675 case MAT_ROW_ORIENTED: 1676 aij->roworiented = flg; 1677 break; 1678 case MAT_NEW_NONZERO_LOCATIONS: 1679 case MAT_NEW_NONZERO_LOCATION_ERR: 1680 case MAT_NEW_NONZERO_ALLOCATION_ERR: 1681 case MAT_NEW_DIAGONALS: 1682 case MAT_KEEP_NONZERO_PATTERN: 1683 case MAT_IGNORE_OFF_PROC_ENTRIES: 1684 case MAT_USE_HASH_TABLE: 1685 case MAT_IGNORE_ZERO_ENTRIES: 1686 case MAT_IGNORE_LOWER_TRIANGULAR: 1687 case MAT_SORTED_FULL: 1688 ierr = PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);CHKERRQ(ierr); 1689 break; 1690 case MAT_SPD: 1691 case MAT_SYMMETRIC: 1692 case MAT_STRUCTURALLY_SYMMETRIC: 1693 case MAT_HERMITIAN: 1694 case MAT_SYMMETRY_ETERNAL: 1695 /* These options are handled directly by MatSetOption() */ 1696 break; 1697 default: 1698 SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"unknown option %s",MatOptions[op]); 1699 } 1700 PetscFunctionReturn(0); 1701 } 1702 1703 static PetscErrorCode MatZeroEntries_SeqDense(Mat A) 1704 { 1705 Mat_SeqDense *l = (Mat_SeqDense*)A->data; 1706 PetscErrorCode ierr; 1707 PetscInt lda=l->lda,m=A->rmap->n,j; 1708 PetscScalar *v; 1709 1710 PetscFunctionBegin; 1711 ierr = MatDenseGetArray(A,&v);CHKERRQ(ierr); 1712 if (lda>m) { 1713 for (j=0; j<A->cmap->n; j++) { 1714 ierr = PetscArrayzero(v+j*lda,m);CHKERRQ(ierr); 1715 } 1716 } else { 1717 ierr = PetscArrayzero(v,A->rmap->n*A->cmap->n);CHKERRQ(ierr); 1718 } 1719 ierr = MatDenseRestoreArray(A,&v);CHKERRQ(ierr); 1720 PetscFunctionReturn(0); 1721 } 1722 1723 static PetscErrorCode MatZeroRows_SeqDense(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag,Vec x,Vec b) 1724 { 1725 PetscErrorCode ierr; 1726 Mat_SeqDense *l = (Mat_SeqDense*)A->data; 1727 PetscInt m = l->lda, n = A->cmap->n, i,j; 1728 PetscScalar *slot,*bb,*v; 1729 const PetscScalar *xx; 1730 1731 PetscFunctionBegin; 1732 if (PetscDefined(USE_DEBUG)) { 1733 for (i=0; i<N; i++) { 1734 if (rows[i] < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row requested to be zeroed"); 1735 if (rows[i] >= A->rmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row %D requested to be zeroed greater than or equal number of rows %D",rows[i],A->rmap->n); 1736 } 1737 } 1738 if (!N) PetscFunctionReturn(0); 1739 1740 /* fix right hand side if needed */ 1741 if (x && b) { 1742 ierr = VecGetArrayRead(x,&xx);CHKERRQ(ierr); 1743 ierr = VecGetArray(b,&bb);CHKERRQ(ierr); 1744 for (i=0; i<N; i++) bb[rows[i]] = diag*xx[rows[i]]; 1745 ierr = VecRestoreArrayRead(x,&xx);CHKERRQ(ierr); 1746 ierr = VecRestoreArray(b,&bb);CHKERRQ(ierr); 1747 } 1748 1749 ierr = MatDenseGetArray(A,&v);CHKERRQ(ierr); 1750 for (i=0; i<N; i++) { 1751 slot = v + rows[i]; 1752 for (j=0; j<n; j++) { *slot = 0.0; slot += m;} 1753 } 1754 if (diag != 0.0) { 1755 if (A->rmap->n != A->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only coded for square matrices"); 1756 for (i=0; i<N; i++) { 1757 slot = v + (m+1)*rows[i]; 1758 *slot = diag; 1759 } 1760 } 1761 ierr = MatDenseRestoreArray(A,&v);CHKERRQ(ierr); 1762 PetscFunctionReturn(0); 1763 } 1764 1765 static PetscErrorCode MatDenseGetLDA_SeqDense(Mat A,PetscInt *lda) 1766 { 1767 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 1768 1769 PetscFunctionBegin; 1770 *lda = mat->lda; 1771 PetscFunctionReturn(0); 1772 } 1773 1774 PetscErrorCode MatDenseGetArray_SeqDense(Mat A,PetscScalar **array) 1775 { 1776 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 1777 1778 PetscFunctionBegin; 1779 *array = mat->v; 1780 PetscFunctionReturn(0); 1781 } 1782 1783 PetscErrorCode MatDenseRestoreArray_SeqDense(Mat A,PetscScalar **array) 1784 { 1785 PetscFunctionBegin; 1786 *array = NULL; 1787 PetscFunctionReturn(0); 1788 } 1789 1790 /*@C 1791 MatDenseGetLDA - gets the leading dimension of the array returned from MatDenseGetArray() 1792 1793 Not collective 1794 1795 Input Parameter: 1796 . mat - a MATSEQDENSE or MATMPIDENSE matrix 1797 1798 Output Parameter: 1799 . lda - the leading dimension 1800 1801 Level: intermediate 1802 1803 .seealso: MatDenseGetArray(), MatDenseRestoreArray(), MatDenseGetArrayRead(), MatDenseRestoreArrayRead(), MatDenseSetLDA() 1804 @*/ 1805 PetscErrorCode MatDenseGetLDA(Mat A,PetscInt *lda) 1806 { 1807 PetscErrorCode ierr; 1808 1809 PetscFunctionBegin; 1810 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 1811 PetscValidPointer(lda,2); 1812 ierr = PetscUseMethod(A,"MatDenseGetLDA_C",(Mat,PetscInt*),(A,lda));CHKERRQ(ierr); 1813 PetscFunctionReturn(0); 1814 } 1815 1816 /*@C 1817 MatDenseSetLDA - Sets the leading dimension of the array used by the dense matrix 1818 1819 Not collective 1820 1821 Input Parameter: 1822 + mat - a MATSEQDENSE or MATMPIDENSE matrix 1823 - lda - the leading dimension 1824 1825 Level: intermediate 1826 1827 .seealso: MatDenseGetArray(), MatDenseRestoreArray(), MatDenseGetArrayRead(), MatDenseRestoreArrayRead(), MatDenseGetLDA() 1828 @*/ 1829 PetscErrorCode MatDenseSetLDA(Mat A,PetscInt lda) 1830 { 1831 PetscErrorCode ierr; 1832 1833 PetscFunctionBegin; 1834 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 1835 ierr = PetscTryMethod(A,"MatDenseSetLDA_C",(Mat,PetscInt),(A,lda));CHKERRQ(ierr); 1836 PetscFunctionReturn(0); 1837 } 1838 1839 /*@C 1840 MatDenseGetArray - gives read-write access to the array where the data for a dense matrix is stored 1841 1842 Logically Collective on Mat 1843 1844 Input Parameter: 1845 . mat - a dense matrix 1846 1847 Output Parameter: 1848 . array - pointer to the data 1849 1850 Level: intermediate 1851 1852 .seealso: MatDenseRestoreArray(), MatDenseGetArrayRead(), MatDenseRestoreArrayRead(), MatDenseGetArrayWrite(), MatDenseRestoreArrayWrite() 1853 @*/ 1854 PetscErrorCode MatDenseGetArray(Mat A,PetscScalar **array) 1855 { 1856 PetscErrorCode ierr; 1857 1858 PetscFunctionBegin; 1859 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 1860 PetscValidPointer(array,2); 1861 ierr = PetscUseMethod(A,"MatDenseGetArray_C",(Mat,PetscScalar**),(A,array));CHKERRQ(ierr); 1862 PetscFunctionReturn(0); 1863 } 1864 1865 /*@C 1866 MatDenseRestoreArray - returns access to the array where the data for a dense matrix is stored obtained by MatDenseGetArray() 1867 1868 Logically Collective on Mat 1869 1870 Input Parameters: 1871 + mat - a dense matrix 1872 - array - pointer to the data 1873 1874 Level: intermediate 1875 1876 .seealso: MatDenseGetArray(), MatDenseGetArrayRead(), MatDenseRestoreArrayRead(), MatDenseGetArrayWrite(), MatDenseRestoreArrayWrite() 1877 @*/ 1878 PetscErrorCode MatDenseRestoreArray(Mat A,PetscScalar **array) 1879 { 1880 PetscErrorCode ierr; 1881 1882 PetscFunctionBegin; 1883 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 1884 PetscValidPointer(array,2); 1885 ierr = PetscUseMethod(A,"MatDenseRestoreArray_C",(Mat,PetscScalar**),(A,array));CHKERRQ(ierr); 1886 ierr = PetscObjectStateIncrease((PetscObject)A);CHKERRQ(ierr); 1887 #if defined(PETSC_HAVE_CUDA) 1888 A->offloadmask = PETSC_OFFLOAD_CPU; 1889 #endif 1890 PetscFunctionReturn(0); 1891 } 1892 1893 /*@C 1894 MatDenseGetArrayRead - gives read-only access to the array where the data for a dense matrix is stored 1895 1896 Not Collective 1897 1898 Input Parameter: 1899 . mat - a dense matrix 1900 1901 Output Parameter: 1902 . array - pointer to the data 1903 1904 Level: intermediate 1905 1906 .seealso: MatDenseRestoreArrayRead(), MatDenseGetArray(), MatDenseRestoreArray(), MatDenseGetArrayWrite(), MatDenseRestoreArrayWrite() 1907 @*/ 1908 PetscErrorCode MatDenseGetArrayRead(Mat A,const PetscScalar **array) 1909 { 1910 PetscErrorCode ierr; 1911 1912 PetscFunctionBegin; 1913 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 1914 PetscValidPointer(array,2); 1915 ierr = PetscUseMethod(A,"MatDenseGetArrayRead_C",(Mat,const PetscScalar**),(A,array));CHKERRQ(ierr); 1916 PetscFunctionReturn(0); 1917 } 1918 1919 /*@C 1920 MatDenseRestoreArrayRead - returns access to the array where the data for a dense matrix is stored obtained by MatDenseGetArrayRead() 1921 1922 Not Collective 1923 1924 Input Parameters: 1925 + mat - a dense matrix 1926 - array - pointer to the data 1927 1928 Level: intermediate 1929 1930 .seealso: MatDenseGetArrayRead(), MatDenseGetArray(), MatDenseRestoreArray(), MatDenseGetArrayWrite(), MatDenseRestoreArrayWrite() 1931 @*/ 1932 PetscErrorCode MatDenseRestoreArrayRead(Mat A,const PetscScalar **array) 1933 { 1934 PetscErrorCode ierr; 1935 1936 PetscFunctionBegin; 1937 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 1938 PetscValidPointer(array,2); 1939 ierr = PetscUseMethod(A,"MatDenseRestoreArrayRead_C",(Mat,const PetscScalar**),(A,array));CHKERRQ(ierr); 1940 PetscFunctionReturn(0); 1941 } 1942 1943 /*@C 1944 MatDenseGetArrayWrite - gives write-only access to the array where the data for a dense matrix is stored 1945 1946 Not Collective 1947 1948 Input Parameter: 1949 . mat - a dense matrix 1950 1951 Output Parameter: 1952 . array - pointer to the data 1953 1954 Level: intermediate 1955 1956 .seealso: MatDenseRestoreArrayWrite(), MatDenseGetArray(), MatDenseRestoreArray(), MatDenseGetArrayRead(), MatDenseRestoreArrayRead() 1957 @*/ 1958 PetscErrorCode MatDenseGetArrayWrite(Mat A,PetscScalar **array) 1959 { 1960 PetscErrorCode ierr; 1961 1962 PetscFunctionBegin; 1963 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 1964 PetscValidPointer(array,2); 1965 ierr = PetscUseMethod(A,"MatDenseGetArrayWrite_C",(Mat,PetscScalar**),(A,array));CHKERRQ(ierr); 1966 PetscFunctionReturn(0); 1967 } 1968 1969 /*@C 1970 MatDenseRestoreArrayWrite - returns access to the array where the data for a dense matrix is stored obtained by MatDenseGetArrayWrite() 1971 1972 Not Collective 1973 1974 Input Parameters: 1975 + mat - a dense matrix 1976 - array - pointer to the data 1977 1978 Level: intermediate 1979 1980 .seealso: MatDenseGetArrayWrite(), MatDenseGetArray(), MatDenseRestoreArray(), MatDenseGetArrayRead(), MatDenseRestoreArrayRead() 1981 @*/ 1982 PetscErrorCode MatDenseRestoreArrayWrite(Mat A,PetscScalar **array) 1983 { 1984 PetscErrorCode ierr; 1985 1986 PetscFunctionBegin; 1987 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 1988 PetscValidPointer(array,2); 1989 ierr = PetscUseMethod(A,"MatDenseRestoreArrayWrite_C",(Mat,PetscScalar**),(A,array));CHKERRQ(ierr); 1990 ierr = PetscObjectStateIncrease((PetscObject)A);CHKERRQ(ierr); 1991 #if defined(PETSC_HAVE_CUDA) 1992 A->offloadmask = PETSC_OFFLOAD_CPU; 1993 #endif 1994 PetscFunctionReturn(0); 1995 } 1996 1997 static PetscErrorCode MatCreateSubMatrix_SeqDense(Mat A,IS isrow,IS iscol,PetscInt cs,MatReuse scall,Mat *B) 1998 { 1999 Mat_SeqDense *mat = (Mat_SeqDense*)A->data; 2000 PetscErrorCode ierr; 2001 PetscInt i,j,nrows,ncols,blda; 2002 const PetscInt *irow,*icol; 2003 PetscScalar *av,*bv,*v = mat->v; 2004 Mat newmat; 2005 2006 PetscFunctionBegin; 2007 ierr = ISGetIndices(isrow,&irow);CHKERRQ(ierr); 2008 ierr = ISGetIndices(iscol,&icol);CHKERRQ(ierr); 2009 ierr = ISGetLocalSize(isrow,&nrows);CHKERRQ(ierr); 2010 ierr = ISGetLocalSize(iscol,&ncols);CHKERRQ(ierr); 2011 2012 /* Check submatrixcall */ 2013 if (scall == MAT_REUSE_MATRIX) { 2014 PetscInt n_cols,n_rows; 2015 ierr = MatGetSize(*B,&n_rows,&n_cols);CHKERRQ(ierr); 2016 if (n_rows != nrows || n_cols != ncols) { 2017 /* resize the result matrix to match number of requested rows/columns */ 2018 ierr = MatSetSizes(*B,nrows,ncols,nrows,ncols);CHKERRQ(ierr); 2019 } 2020 newmat = *B; 2021 } else { 2022 /* Create and fill new matrix */ 2023 ierr = MatCreate(PetscObjectComm((PetscObject)A),&newmat);CHKERRQ(ierr); 2024 ierr = MatSetSizes(newmat,nrows,ncols,nrows,ncols);CHKERRQ(ierr); 2025 ierr = MatSetType(newmat,((PetscObject)A)->type_name);CHKERRQ(ierr); 2026 ierr = MatSeqDenseSetPreallocation(newmat,NULL);CHKERRQ(ierr); 2027 } 2028 2029 /* Now extract the data pointers and do the copy,column at a time */ 2030 ierr = MatDenseGetArray(newmat,&bv);CHKERRQ(ierr); 2031 ierr = MatDenseGetLDA(newmat,&blda);CHKERRQ(ierr); 2032 for (i=0; i<ncols; i++) { 2033 av = v + mat->lda*icol[i]; 2034 for (j=0; j<nrows; j++) bv[j] = av[irow[j]]; 2035 bv += blda; 2036 } 2037 ierr = MatDenseRestoreArray(newmat,&bv);CHKERRQ(ierr); 2038 2039 /* Assemble the matrices so that the correct flags are set */ 2040 ierr = MatAssemblyBegin(newmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2041 ierr = MatAssemblyEnd(newmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2042 2043 /* Free work space */ 2044 ierr = ISRestoreIndices(isrow,&irow);CHKERRQ(ierr); 2045 ierr = ISRestoreIndices(iscol,&icol);CHKERRQ(ierr); 2046 *B = newmat; 2047 PetscFunctionReturn(0); 2048 } 2049 2050 static PetscErrorCode MatCreateSubMatrices_SeqDense(Mat A,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *B[]) 2051 { 2052 PetscErrorCode ierr; 2053 PetscInt i; 2054 2055 PetscFunctionBegin; 2056 if (scall == MAT_INITIAL_MATRIX) { 2057 ierr = PetscCalloc1(n+1,B);CHKERRQ(ierr); 2058 } 2059 2060 for (i=0; i<n; i++) { 2061 ierr = MatCreateSubMatrix_SeqDense(A,irow[i],icol[i],PETSC_DECIDE,scall,&(*B)[i]);CHKERRQ(ierr); 2062 } 2063 PetscFunctionReturn(0); 2064 } 2065 2066 static PetscErrorCode MatAssemblyBegin_SeqDense(Mat mat,MatAssemblyType mode) 2067 { 2068 PetscFunctionBegin; 2069 PetscFunctionReturn(0); 2070 } 2071 2072 static PetscErrorCode MatAssemblyEnd_SeqDense(Mat mat,MatAssemblyType mode) 2073 { 2074 PetscFunctionBegin; 2075 PetscFunctionReturn(0); 2076 } 2077 2078 static PetscErrorCode MatCopy_SeqDense(Mat A,Mat B,MatStructure str) 2079 { 2080 Mat_SeqDense *a = (Mat_SeqDense*)A->data,*b = (Mat_SeqDense*)B->data; 2081 PetscErrorCode ierr; 2082 const PetscScalar *va; 2083 PetscScalar *vb; 2084 PetscInt lda1=a->lda,lda2=b->lda, m=A->rmap->n,n=A->cmap->n, j; 2085 2086 PetscFunctionBegin; 2087 /* If the two matrices don't have the same copy implementation, they aren't compatible for fast copy. */ 2088 if (A->ops->copy != B->ops->copy) { 2089 ierr = MatCopy_Basic(A,B,str);CHKERRQ(ierr); 2090 PetscFunctionReturn(0); 2091 } 2092 if (m != B->rmap->n || n != B->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"size(B) != size(A)"); 2093 ierr = MatDenseGetArrayRead(A,&va);CHKERRQ(ierr); 2094 ierr = MatDenseGetArray(B,&vb);CHKERRQ(ierr); 2095 if (lda1>m || lda2>m) { 2096 for (j=0; j<n; j++) { 2097 ierr = PetscArraycpy(vb+j*lda2,va+j*lda1,m);CHKERRQ(ierr); 2098 } 2099 } else { 2100 ierr = PetscArraycpy(vb,va,A->rmap->n*A->cmap->n);CHKERRQ(ierr); 2101 } 2102 ierr = MatDenseRestoreArray(B,&vb);CHKERRQ(ierr); 2103 ierr = MatDenseRestoreArrayRead(A,&va);CHKERRQ(ierr); 2104 ierr = MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2105 ierr = MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2106 PetscFunctionReturn(0); 2107 } 2108 2109 static PetscErrorCode MatSetUp_SeqDense(Mat A) 2110 { 2111 PetscErrorCode ierr; 2112 2113 PetscFunctionBegin; 2114 ierr = PetscLayoutSetUp(A->rmap);CHKERRQ(ierr); 2115 ierr = PetscLayoutSetUp(A->cmap);CHKERRQ(ierr); 2116 if (!A->preallocated) { 2117 ierr = MatSeqDenseSetPreallocation(A,0);CHKERRQ(ierr); 2118 } 2119 PetscFunctionReturn(0); 2120 } 2121 2122 static PetscErrorCode MatConjugate_SeqDense(Mat A) 2123 { 2124 PetscInt i,nz = A->rmap->n*A->cmap->n; 2125 PetscScalar *aa; 2126 PetscErrorCode ierr; 2127 2128 PetscFunctionBegin; 2129 ierr = MatDenseGetArray(A,&aa);CHKERRQ(ierr); 2130 for (i=0; i<nz; i++) aa[i] = PetscConj(aa[i]); 2131 ierr = MatDenseRestoreArray(A,&aa);CHKERRQ(ierr); 2132 PetscFunctionReturn(0); 2133 } 2134 2135 static PetscErrorCode MatRealPart_SeqDense(Mat A) 2136 { 2137 PetscInt i,nz = A->rmap->n*A->cmap->n; 2138 PetscScalar *aa; 2139 PetscErrorCode ierr; 2140 2141 PetscFunctionBegin; 2142 ierr = MatDenseGetArray(A,&aa);CHKERRQ(ierr); 2143 for (i=0; i<nz; i++) aa[i] = PetscRealPart(aa[i]); 2144 ierr = MatDenseRestoreArray(A,&aa);CHKERRQ(ierr); 2145 PetscFunctionReturn(0); 2146 } 2147 2148 static PetscErrorCode MatImaginaryPart_SeqDense(Mat A) 2149 { 2150 PetscInt i,nz = A->rmap->n*A->cmap->n; 2151 PetscScalar *aa; 2152 PetscErrorCode ierr; 2153 2154 PetscFunctionBegin; 2155 ierr = MatDenseGetArray(A,&aa);CHKERRQ(ierr); 2156 for (i=0; i<nz; i++) aa[i] = PetscImaginaryPart(aa[i]); 2157 ierr = MatDenseRestoreArray(A,&aa);CHKERRQ(ierr); 2158 PetscFunctionReturn(0); 2159 } 2160 2161 /* ----------------------------------------------------------------*/ 2162 PetscErrorCode MatMatMultSymbolic_SeqDense_SeqDense(Mat A,Mat B,PetscReal fill,Mat C) 2163 { 2164 PetscErrorCode ierr; 2165 PetscInt m=A->rmap->n,n=B->cmap->n; 2166 PetscBool cisdense; 2167 2168 PetscFunctionBegin; 2169 ierr = MatSetSizes(C,m,n,m,n);CHKERRQ(ierr); 2170 ierr = PetscObjectTypeCompareAny((PetscObject)C,&cisdense,MATSEQDENSE,MATSEQDENSECUDA,"");CHKERRQ(ierr); 2171 if (!cisdense) { 2172 PetscBool flg; 2173 2174 ierr = PetscObjectTypeCompare((PetscObject)B,((PetscObject)A)->type_name,&flg);CHKERRQ(ierr); 2175 ierr = MatSetType(C,flg ? ((PetscObject)A)->type_name : MATDENSE);CHKERRQ(ierr); 2176 } 2177 ierr = MatSetUp(C);CHKERRQ(ierr); 2178 PetscFunctionReturn(0); 2179 } 2180 2181 PetscErrorCode MatMatMultNumeric_SeqDense_SeqDense(Mat A,Mat B,Mat C) 2182 { 2183 Mat_SeqDense *a=(Mat_SeqDense*)A->data,*b=(Mat_SeqDense*)B->data,*c=(Mat_SeqDense*)C->data; 2184 PetscBLASInt m,n,k; 2185 const PetscScalar *av,*bv; 2186 PetscScalar *cv; 2187 PetscScalar _DOne=1.0,_DZero=0.0; 2188 PetscErrorCode ierr; 2189 2190 PetscFunctionBegin; 2191 ierr = PetscBLASIntCast(C->rmap->n,&m);CHKERRQ(ierr); 2192 ierr = PetscBLASIntCast(C->cmap->n,&n);CHKERRQ(ierr); 2193 ierr = PetscBLASIntCast(A->cmap->n,&k);CHKERRQ(ierr); 2194 if (!m || !n || !k) PetscFunctionReturn(0); 2195 ierr = MatDenseGetArrayRead(A,&av);CHKERRQ(ierr); 2196 ierr = MatDenseGetArrayRead(B,&bv);CHKERRQ(ierr); 2197 ierr = MatDenseGetArrayWrite(C,&cv);CHKERRQ(ierr); 2198 PetscStackCallBLAS("BLASgemm",BLASgemm_("N","N",&m,&n,&k,&_DOne,av,&a->lda,bv,&b->lda,&_DZero,cv,&c->lda)); 2199 ierr = PetscLogFlops(1.0*m*n*k + 1.0*m*n*(k-1));CHKERRQ(ierr); 2200 ierr = MatDenseRestoreArrayRead(A,&av);CHKERRQ(ierr); 2201 ierr = MatDenseRestoreArrayRead(B,&bv);CHKERRQ(ierr); 2202 ierr = MatDenseRestoreArrayWrite(C,&cv);CHKERRQ(ierr); 2203 PetscFunctionReturn(0); 2204 } 2205 2206 PetscErrorCode MatMatTransposeMultSymbolic_SeqDense_SeqDense(Mat A,Mat B,PetscReal fill,Mat C) 2207 { 2208 PetscErrorCode ierr; 2209 PetscInt m=A->rmap->n,n=B->rmap->n; 2210 PetscBool cisdense; 2211 2212 PetscFunctionBegin; 2213 ierr = MatSetSizes(C,m,n,m,n);CHKERRQ(ierr); 2214 ierr = PetscObjectTypeCompareAny((PetscObject)C,&cisdense,MATSEQDENSE,MATSEQDENSECUDA,"");CHKERRQ(ierr); 2215 if (!cisdense) { 2216 PetscBool flg; 2217 2218 ierr = PetscObjectTypeCompare((PetscObject)B,((PetscObject)A)->type_name,&flg);CHKERRQ(ierr); 2219 ierr = MatSetType(C,flg ? ((PetscObject)A)->type_name : MATDENSE);CHKERRQ(ierr); 2220 } 2221 ierr = MatSetUp(C);CHKERRQ(ierr); 2222 PetscFunctionReturn(0); 2223 } 2224 2225 PetscErrorCode MatMatTransposeMultNumeric_SeqDense_SeqDense(Mat A,Mat B,Mat C) 2226 { 2227 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2228 Mat_SeqDense *b = (Mat_SeqDense*)B->data; 2229 Mat_SeqDense *c = (Mat_SeqDense*)C->data; 2230 const PetscScalar *av,*bv; 2231 PetscScalar *cv; 2232 PetscBLASInt m,n,k; 2233 PetscScalar _DOne=1.0,_DZero=0.0; 2234 PetscErrorCode ierr; 2235 2236 PetscFunctionBegin; 2237 ierr = PetscBLASIntCast(C->rmap->n,&m);CHKERRQ(ierr); 2238 ierr = PetscBLASIntCast(C->cmap->n,&n);CHKERRQ(ierr); 2239 ierr = PetscBLASIntCast(A->cmap->n,&k);CHKERRQ(ierr); 2240 if (!m || !n || !k) PetscFunctionReturn(0); 2241 ierr = MatDenseGetArrayRead(A,&av);CHKERRQ(ierr); 2242 ierr = MatDenseGetArrayRead(B,&bv);CHKERRQ(ierr); 2243 ierr = MatDenseGetArrayWrite(C,&cv);CHKERRQ(ierr); 2244 PetscStackCallBLAS("BLASgemm",BLASgemm_("N","T",&m,&n,&k,&_DOne,av,&a->lda,bv,&b->lda,&_DZero,cv,&c->lda)); 2245 ierr = MatDenseRestoreArrayRead(A,&av);CHKERRQ(ierr); 2246 ierr = MatDenseRestoreArrayRead(B,&bv);CHKERRQ(ierr); 2247 ierr = MatDenseRestoreArrayWrite(C,&cv);CHKERRQ(ierr); 2248 ierr = PetscLogFlops(1.0*m*n*k + 1.0*m*n*(k-1));CHKERRQ(ierr); 2249 PetscFunctionReturn(0); 2250 } 2251 2252 PetscErrorCode MatTransposeMatMultSymbolic_SeqDense_SeqDense(Mat A,Mat B,PetscReal fill,Mat C) 2253 { 2254 PetscErrorCode ierr; 2255 PetscInt m=A->cmap->n,n=B->cmap->n; 2256 PetscBool cisdense; 2257 2258 PetscFunctionBegin; 2259 ierr = MatSetSizes(C,m,n,m,n);CHKERRQ(ierr); 2260 ierr = PetscObjectTypeCompareAny((PetscObject)C,&cisdense,MATSEQDENSE,MATSEQDENSECUDA,"");CHKERRQ(ierr); 2261 if (!cisdense) { 2262 PetscBool flg; 2263 2264 ierr = PetscObjectTypeCompare((PetscObject)B,((PetscObject)A)->type_name,&flg);CHKERRQ(ierr); 2265 ierr = MatSetType(C,flg ? ((PetscObject)A)->type_name : MATDENSE);CHKERRQ(ierr); 2266 } 2267 ierr = MatSetUp(C);CHKERRQ(ierr); 2268 PetscFunctionReturn(0); 2269 } 2270 2271 PetscErrorCode MatTransposeMatMultNumeric_SeqDense_SeqDense(Mat A,Mat B,Mat C) 2272 { 2273 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2274 Mat_SeqDense *b = (Mat_SeqDense*)B->data; 2275 Mat_SeqDense *c = (Mat_SeqDense*)C->data; 2276 const PetscScalar *av,*bv; 2277 PetscScalar *cv; 2278 PetscBLASInt m,n,k; 2279 PetscScalar _DOne=1.0,_DZero=0.0; 2280 PetscErrorCode ierr; 2281 2282 PetscFunctionBegin; 2283 ierr = PetscBLASIntCast(C->rmap->n,&m);CHKERRQ(ierr); 2284 ierr = PetscBLASIntCast(C->cmap->n,&n);CHKERRQ(ierr); 2285 ierr = PetscBLASIntCast(A->rmap->n,&k);CHKERRQ(ierr); 2286 if (!m || !n || !k) PetscFunctionReturn(0); 2287 ierr = MatDenseGetArrayRead(A,&av);CHKERRQ(ierr); 2288 ierr = MatDenseGetArrayRead(B,&bv);CHKERRQ(ierr); 2289 ierr = MatDenseGetArrayWrite(C,&cv);CHKERRQ(ierr); 2290 PetscStackCallBLAS("BLASgemm",BLASgemm_("T","N",&m,&n,&k,&_DOne,av,&a->lda,bv,&b->lda,&_DZero,cv,&c->lda)); 2291 ierr = MatDenseRestoreArrayRead(A,&av);CHKERRQ(ierr); 2292 ierr = MatDenseRestoreArrayRead(B,&bv);CHKERRQ(ierr); 2293 ierr = MatDenseRestoreArrayWrite(C,&cv);CHKERRQ(ierr); 2294 ierr = PetscLogFlops(1.0*m*n*k + 1.0*m*n*(k-1));CHKERRQ(ierr); 2295 PetscFunctionReturn(0); 2296 } 2297 2298 /* ----------------------------------------------- */ 2299 static PetscErrorCode MatProductSetFromOptions_SeqDense_AB(Mat C) 2300 { 2301 PetscFunctionBegin; 2302 C->ops->matmultsymbolic = MatMatMultSymbolic_SeqDense_SeqDense; 2303 C->ops->productsymbolic = MatProductSymbolic_AB; 2304 PetscFunctionReturn(0); 2305 } 2306 2307 static PetscErrorCode MatProductSetFromOptions_SeqDense_AtB(Mat C) 2308 { 2309 PetscFunctionBegin; 2310 C->ops->transposematmultsymbolic = MatTransposeMatMultSymbolic_SeqDense_SeqDense; 2311 C->ops->productsymbolic = MatProductSymbolic_AtB; 2312 PetscFunctionReturn(0); 2313 } 2314 2315 static PetscErrorCode MatProductSetFromOptions_SeqDense_ABt(Mat C) 2316 { 2317 PetscFunctionBegin; 2318 C->ops->mattransposemultsymbolic = MatMatTransposeMultSymbolic_SeqDense_SeqDense; 2319 C->ops->productsymbolic = MatProductSymbolic_ABt; 2320 PetscFunctionReturn(0); 2321 } 2322 2323 PETSC_INTERN PetscErrorCode MatProductSetFromOptions_SeqDense(Mat C) 2324 { 2325 PetscErrorCode ierr; 2326 Mat_Product *product = C->product; 2327 2328 PetscFunctionBegin; 2329 switch (product->type) { 2330 case MATPRODUCT_AB: 2331 ierr = MatProductSetFromOptions_SeqDense_AB(C);CHKERRQ(ierr); 2332 break; 2333 case MATPRODUCT_AtB: 2334 ierr = MatProductSetFromOptions_SeqDense_AtB(C);CHKERRQ(ierr); 2335 break; 2336 case MATPRODUCT_ABt: 2337 ierr = MatProductSetFromOptions_SeqDense_ABt(C);CHKERRQ(ierr); 2338 break; 2339 default: 2340 break; 2341 } 2342 PetscFunctionReturn(0); 2343 } 2344 /* ----------------------------------------------- */ 2345 2346 static PetscErrorCode MatGetRowMax_SeqDense(Mat A,Vec v,PetscInt idx[]) 2347 { 2348 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2349 PetscErrorCode ierr; 2350 PetscInt i,j,m = A->rmap->n,n = A->cmap->n,p; 2351 PetscScalar *x; 2352 const PetscScalar *aa; 2353 2354 PetscFunctionBegin; 2355 if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix"); 2356 ierr = VecGetArray(v,&x);CHKERRQ(ierr); 2357 ierr = VecGetLocalSize(v,&p);CHKERRQ(ierr); 2358 ierr = MatDenseGetArrayRead(A,&aa);CHKERRQ(ierr); 2359 if (p != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector"); 2360 for (i=0; i<m; i++) { 2361 x[i] = aa[i]; if (idx) idx[i] = 0; 2362 for (j=1; j<n; j++) { 2363 if (PetscRealPart(x[i]) < PetscRealPart(aa[i+a->lda*j])) {x[i] = aa[i + a->lda*j]; if (idx) idx[i] = j;} 2364 } 2365 } 2366 ierr = MatDenseRestoreArrayRead(A,&aa);CHKERRQ(ierr); 2367 ierr = VecRestoreArray(v,&x);CHKERRQ(ierr); 2368 PetscFunctionReturn(0); 2369 } 2370 2371 static PetscErrorCode MatGetRowMaxAbs_SeqDense(Mat A,Vec v,PetscInt idx[]) 2372 { 2373 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2374 PetscErrorCode ierr; 2375 PetscInt i,j,m = A->rmap->n,n = A->cmap->n,p; 2376 PetscScalar *x; 2377 PetscReal atmp; 2378 const PetscScalar *aa; 2379 2380 PetscFunctionBegin; 2381 if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix"); 2382 ierr = VecGetArray(v,&x);CHKERRQ(ierr); 2383 ierr = VecGetLocalSize(v,&p);CHKERRQ(ierr); 2384 ierr = MatDenseGetArrayRead(A,&aa);CHKERRQ(ierr); 2385 if (p != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector"); 2386 for (i=0; i<m; i++) { 2387 x[i] = PetscAbsScalar(aa[i]); 2388 for (j=1; j<n; j++) { 2389 atmp = PetscAbsScalar(aa[i+a->lda*j]); 2390 if (PetscAbsScalar(x[i]) < atmp) {x[i] = atmp; if (idx) idx[i] = j;} 2391 } 2392 } 2393 ierr = MatDenseRestoreArrayRead(A,&aa);CHKERRQ(ierr); 2394 ierr = VecRestoreArray(v,&x);CHKERRQ(ierr); 2395 PetscFunctionReturn(0); 2396 } 2397 2398 static PetscErrorCode MatGetRowMin_SeqDense(Mat A,Vec v,PetscInt idx[]) 2399 { 2400 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2401 PetscErrorCode ierr; 2402 PetscInt i,j,m = A->rmap->n,n = A->cmap->n,p; 2403 PetscScalar *x; 2404 const PetscScalar *aa; 2405 2406 PetscFunctionBegin; 2407 if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix"); 2408 ierr = MatDenseGetArrayRead(A,&aa);CHKERRQ(ierr); 2409 ierr = VecGetArray(v,&x);CHKERRQ(ierr); 2410 ierr = VecGetLocalSize(v,&p);CHKERRQ(ierr); 2411 if (p != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector"); 2412 for (i=0; i<m; i++) { 2413 x[i] = aa[i]; if (idx) idx[i] = 0; 2414 for (j=1; j<n; j++) { 2415 if (PetscRealPart(x[i]) > PetscRealPart(aa[i+a->lda*j])) {x[i] = aa[i + a->lda*j]; if (idx) idx[i] = j;} 2416 } 2417 } 2418 ierr = VecRestoreArray(v,&x);CHKERRQ(ierr); 2419 ierr = MatDenseRestoreArrayRead(A,&aa);CHKERRQ(ierr); 2420 PetscFunctionReturn(0); 2421 } 2422 2423 PetscErrorCode MatGetColumnVector_SeqDense(Mat A,Vec v,PetscInt col) 2424 { 2425 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2426 PetscErrorCode ierr; 2427 PetscScalar *x; 2428 const PetscScalar *aa; 2429 2430 PetscFunctionBegin; 2431 if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix"); 2432 ierr = MatDenseGetArrayRead(A,&aa);CHKERRQ(ierr); 2433 ierr = VecGetArray(v,&x);CHKERRQ(ierr); 2434 ierr = PetscArraycpy(x,aa+col*a->lda,A->rmap->n);CHKERRQ(ierr); 2435 ierr = VecRestoreArray(v,&x);CHKERRQ(ierr); 2436 ierr = MatDenseRestoreArrayRead(A,&aa);CHKERRQ(ierr); 2437 PetscFunctionReturn(0); 2438 } 2439 2440 PETSC_INTERN PetscErrorCode MatGetColumnNorms_SeqDense(Mat A,NormType type,PetscReal *norms) 2441 { 2442 PetscErrorCode ierr; 2443 PetscInt i,j,m,n; 2444 const PetscScalar *a; 2445 2446 PetscFunctionBegin; 2447 ierr = MatGetSize(A,&m,&n);CHKERRQ(ierr); 2448 ierr = PetscArrayzero(norms,n);CHKERRQ(ierr); 2449 ierr = MatDenseGetArrayRead(A,&a);CHKERRQ(ierr); 2450 if (type == NORM_2) { 2451 for (i=0; i<n; i++) { 2452 for (j=0; j<m; j++) { 2453 norms[i] += PetscAbsScalar(a[j]*a[j]); 2454 } 2455 a += m; 2456 } 2457 } else if (type == NORM_1) { 2458 for (i=0; i<n; i++) { 2459 for (j=0; j<m; j++) { 2460 norms[i] += PetscAbsScalar(a[j]); 2461 } 2462 a += m; 2463 } 2464 } else if (type == NORM_INFINITY) { 2465 for (i=0; i<n; i++) { 2466 for (j=0; j<m; j++) { 2467 norms[i] = PetscMax(PetscAbsScalar(a[j]),norms[i]); 2468 } 2469 a += m; 2470 } 2471 } else SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Unknown NormType"); 2472 ierr = MatDenseRestoreArrayRead(A,&a);CHKERRQ(ierr); 2473 if (type == NORM_2) { 2474 for (i=0; i<n; i++) norms[i] = PetscSqrtReal(norms[i]); 2475 } 2476 PetscFunctionReturn(0); 2477 } 2478 2479 static PetscErrorCode MatSetRandom_SeqDense(Mat x,PetscRandom rctx) 2480 { 2481 PetscErrorCode ierr; 2482 PetscScalar *a; 2483 PetscInt lda,m,n,i,j; 2484 2485 PetscFunctionBegin; 2486 ierr = MatGetSize(x,&m,&n);CHKERRQ(ierr); 2487 ierr = MatDenseGetLDA(x,&lda);CHKERRQ(ierr); 2488 ierr = MatDenseGetArray(x,&a);CHKERRQ(ierr); 2489 for (j=0; j<n; j++) { 2490 for (i=0; i<m; i++) { 2491 ierr = PetscRandomGetValue(rctx,a+j*lda+i);CHKERRQ(ierr); 2492 } 2493 } 2494 ierr = MatDenseRestoreArray(x,&a);CHKERRQ(ierr); 2495 PetscFunctionReturn(0); 2496 } 2497 2498 static PetscErrorCode MatMissingDiagonal_SeqDense(Mat A,PetscBool *missing,PetscInt *d) 2499 { 2500 PetscFunctionBegin; 2501 *missing = PETSC_FALSE; 2502 PetscFunctionReturn(0); 2503 } 2504 2505 /* vals is not const */ 2506 static PetscErrorCode MatDenseGetColumn_SeqDense(Mat A,PetscInt col,PetscScalar **vals) 2507 { 2508 PetscErrorCode ierr; 2509 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2510 PetscScalar *v; 2511 2512 PetscFunctionBegin; 2513 if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix"); 2514 ierr = MatDenseGetArray(A,&v);CHKERRQ(ierr); 2515 *vals = v+col*a->lda; 2516 ierr = MatDenseRestoreArray(A,&v);CHKERRQ(ierr); 2517 PetscFunctionReturn(0); 2518 } 2519 2520 static PetscErrorCode MatDenseRestoreColumn_SeqDense(Mat A,PetscScalar **vals) 2521 { 2522 PetscFunctionBegin; 2523 *vals = 0; /* user cannot accidently use the array later */ 2524 PetscFunctionReturn(0); 2525 } 2526 2527 /* -------------------------------------------------------------------*/ 2528 static struct _MatOps MatOps_Values = { MatSetValues_SeqDense, 2529 MatGetRow_SeqDense, 2530 MatRestoreRow_SeqDense, 2531 MatMult_SeqDense, 2532 /* 4*/ MatMultAdd_SeqDense, 2533 MatMultTranspose_SeqDense, 2534 MatMultTransposeAdd_SeqDense, 2535 0, 2536 0, 2537 0, 2538 /* 10*/ 0, 2539 MatLUFactor_SeqDense, 2540 MatCholeskyFactor_SeqDense, 2541 MatSOR_SeqDense, 2542 MatTranspose_SeqDense, 2543 /* 15*/ MatGetInfo_SeqDense, 2544 MatEqual_SeqDense, 2545 MatGetDiagonal_SeqDense, 2546 MatDiagonalScale_SeqDense, 2547 MatNorm_SeqDense, 2548 /* 20*/ MatAssemblyBegin_SeqDense, 2549 MatAssemblyEnd_SeqDense, 2550 MatSetOption_SeqDense, 2551 MatZeroEntries_SeqDense, 2552 /* 24*/ MatZeroRows_SeqDense, 2553 0, 2554 0, 2555 0, 2556 0, 2557 /* 29*/ MatSetUp_SeqDense, 2558 0, 2559 0, 2560 0, 2561 0, 2562 /* 34*/ MatDuplicate_SeqDense, 2563 0, 2564 0, 2565 0, 2566 0, 2567 /* 39*/ MatAXPY_SeqDense, 2568 MatCreateSubMatrices_SeqDense, 2569 0, 2570 MatGetValues_SeqDense, 2571 MatCopy_SeqDense, 2572 /* 44*/ MatGetRowMax_SeqDense, 2573 MatScale_SeqDense, 2574 MatShift_Basic, 2575 0, 2576 MatZeroRowsColumns_SeqDense, 2577 /* 49*/ MatSetRandom_SeqDense, 2578 0, 2579 0, 2580 0, 2581 0, 2582 /* 54*/ 0, 2583 0, 2584 0, 2585 0, 2586 0, 2587 /* 59*/ 0, 2588 MatDestroy_SeqDense, 2589 MatView_SeqDense, 2590 0, 2591 0, 2592 /* 64*/ 0, 2593 0, 2594 0, 2595 0, 2596 0, 2597 /* 69*/ MatGetRowMaxAbs_SeqDense, 2598 0, 2599 0, 2600 0, 2601 0, 2602 /* 74*/ 0, 2603 0, 2604 0, 2605 0, 2606 0, 2607 /* 79*/ 0, 2608 0, 2609 0, 2610 0, 2611 /* 83*/ MatLoad_SeqDense, 2612 MatIsSymmetric_SeqDense, 2613 MatIsHermitian_SeqDense, 2614 0, 2615 0, 2616 0, 2617 /* 89*/ 0, 2618 0, 2619 MatMatMultNumeric_SeqDense_SeqDense, 2620 0, 2621 0, 2622 /* 94*/ 0, 2623 0, 2624 0, 2625 MatMatTransposeMultNumeric_SeqDense_SeqDense, 2626 0, 2627 /* 99*/ MatProductSetFromOptions_SeqDense, 2628 0, 2629 0, 2630 MatConjugate_SeqDense, 2631 0, 2632 /*104*/ 0, 2633 MatRealPart_SeqDense, 2634 MatImaginaryPart_SeqDense, 2635 0, 2636 0, 2637 /*109*/ 0, 2638 0, 2639 MatGetRowMin_SeqDense, 2640 MatGetColumnVector_SeqDense, 2641 MatMissingDiagonal_SeqDense, 2642 /*114*/ 0, 2643 0, 2644 0, 2645 0, 2646 0, 2647 /*119*/ 0, 2648 0, 2649 0, 2650 0, 2651 0, 2652 /*124*/ 0, 2653 MatGetColumnNorms_SeqDense, 2654 0, 2655 0, 2656 0, 2657 /*129*/ 0, 2658 0, 2659 0, 2660 MatTransposeMatMultNumeric_SeqDense_SeqDense, 2661 0, 2662 /*134*/ 0, 2663 0, 2664 0, 2665 0, 2666 0, 2667 /*139*/ 0, 2668 0, 2669 0, 2670 0, 2671 0, 2672 MatCreateMPIMatConcatenateSeqMat_SeqDense, 2673 /*145*/ 0, 2674 0, 2675 0 2676 }; 2677 2678 /*@C 2679 MatCreateSeqDense - Creates a sequential dense matrix that 2680 is stored in column major order (the usual Fortran 77 manner). Many 2681 of the matrix operations use the BLAS and LAPACK routines. 2682 2683 Collective 2684 2685 Input Parameters: 2686 + comm - MPI communicator, set to PETSC_COMM_SELF 2687 . m - number of rows 2688 . n - number of columns 2689 - data - optional location of matrix data in column major order. Set data=NULL for PETSc 2690 to control all matrix memory allocation. 2691 2692 Output Parameter: 2693 . A - the matrix 2694 2695 Notes: 2696 The data input variable is intended primarily for Fortran programmers 2697 who wish to allocate their own matrix memory space. Most users should 2698 set data=NULL. 2699 2700 Level: intermediate 2701 2702 .seealso: MatCreate(), MatCreateDense(), MatSetValues() 2703 @*/ 2704 PetscErrorCode MatCreateSeqDense(MPI_Comm comm,PetscInt m,PetscInt n,PetscScalar *data,Mat *A) 2705 { 2706 PetscErrorCode ierr; 2707 2708 PetscFunctionBegin; 2709 ierr = MatCreate(comm,A);CHKERRQ(ierr); 2710 ierr = MatSetSizes(*A,m,n,m,n);CHKERRQ(ierr); 2711 ierr = MatSetType(*A,MATSEQDENSE);CHKERRQ(ierr); 2712 ierr = MatSeqDenseSetPreallocation(*A,data);CHKERRQ(ierr); 2713 PetscFunctionReturn(0); 2714 } 2715 2716 /*@C 2717 MatSeqDenseSetPreallocation - Sets the array used for storing the matrix elements 2718 2719 Collective 2720 2721 Input Parameters: 2722 + B - the matrix 2723 - data - the array (or NULL) 2724 2725 Notes: 2726 The data input variable is intended primarily for Fortran programmers 2727 who wish to allocate their own matrix memory space. Most users should 2728 need not call this routine. 2729 2730 Level: intermediate 2731 2732 .seealso: MatCreate(), MatCreateDense(), MatSetValues(), MatDenseSetLDA() 2733 2734 @*/ 2735 PetscErrorCode MatSeqDenseSetPreallocation(Mat B,PetscScalar data[]) 2736 { 2737 PetscErrorCode ierr; 2738 2739 PetscFunctionBegin; 2740 PetscValidHeaderSpecific(B,MAT_CLASSID,1); 2741 ierr = PetscTryMethod(B,"MatSeqDenseSetPreallocation_C",(Mat,PetscScalar[]),(B,data));CHKERRQ(ierr); 2742 PetscFunctionReturn(0); 2743 } 2744 2745 PetscErrorCode MatSeqDenseSetPreallocation_SeqDense(Mat B,PetscScalar *data) 2746 { 2747 Mat_SeqDense *b = (Mat_SeqDense*)B->data; 2748 PetscErrorCode ierr; 2749 2750 PetscFunctionBegin; 2751 B->preallocated = PETSC_TRUE; 2752 2753 ierr = PetscLayoutSetUp(B->rmap);CHKERRQ(ierr); 2754 ierr = PetscLayoutSetUp(B->cmap);CHKERRQ(ierr); 2755 2756 if (b->lda <= 0) b->lda = B->rmap->n; 2757 2758 ierr = PetscIntMultError(b->lda,B->cmap->n,NULL);CHKERRQ(ierr); 2759 if (!data) { /* petsc-allocated storage */ 2760 if (!b->user_alloc) { ierr = PetscFree(b->v);CHKERRQ(ierr); } 2761 ierr = PetscCalloc1((size_t)b->lda*B->cmap->n,&b->v);CHKERRQ(ierr); 2762 ierr = PetscLogObjectMemory((PetscObject)B,b->lda*B->cmap->n*sizeof(PetscScalar));CHKERRQ(ierr); 2763 2764 b->user_alloc = PETSC_FALSE; 2765 } else { /* user-allocated storage */ 2766 if (!b->user_alloc) { ierr = PetscFree(b->v);CHKERRQ(ierr); } 2767 b->v = data; 2768 b->user_alloc = PETSC_TRUE; 2769 } 2770 B->assembled = PETSC_TRUE; 2771 PetscFunctionReturn(0); 2772 } 2773 2774 #if defined(PETSC_HAVE_ELEMENTAL) 2775 PETSC_INTERN PetscErrorCode MatConvert_SeqDense_Elemental(Mat A, MatType newtype,MatReuse reuse,Mat *newmat) 2776 { 2777 Mat mat_elemental; 2778 PetscErrorCode ierr; 2779 const PetscScalar *array; 2780 PetscScalar *v_colwise; 2781 PetscInt M=A->rmap->N,N=A->cmap->N,i,j,k,*rows,*cols; 2782 2783 PetscFunctionBegin; 2784 ierr = PetscMalloc3(M*N,&v_colwise,M,&rows,N,&cols);CHKERRQ(ierr); 2785 ierr = MatDenseGetArrayRead(A,&array);CHKERRQ(ierr); 2786 /* convert column-wise array into row-wise v_colwise, see MatSetValues_Elemental() */ 2787 k = 0; 2788 for (j=0; j<N; j++) { 2789 cols[j] = j; 2790 for (i=0; i<M; i++) { 2791 v_colwise[j*M+i] = array[k++]; 2792 } 2793 } 2794 for (i=0; i<M; i++) { 2795 rows[i] = i; 2796 } 2797 ierr = MatDenseRestoreArrayRead(A,&array);CHKERRQ(ierr); 2798 2799 ierr = MatCreate(PetscObjectComm((PetscObject)A), &mat_elemental);CHKERRQ(ierr); 2800 ierr = MatSetSizes(mat_elemental,PETSC_DECIDE,PETSC_DECIDE,M,N);CHKERRQ(ierr); 2801 ierr = MatSetType(mat_elemental,MATELEMENTAL);CHKERRQ(ierr); 2802 ierr = MatSetUp(mat_elemental);CHKERRQ(ierr); 2803 2804 /* PETSc-Elemental interaface uses axpy for setting off-processor entries, only ADD_VALUES is allowed */ 2805 ierr = MatSetValues(mat_elemental,M,rows,N,cols,v_colwise,ADD_VALUES);CHKERRQ(ierr); 2806 ierr = MatAssemblyBegin(mat_elemental, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2807 ierr = MatAssemblyEnd(mat_elemental, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2808 ierr = PetscFree3(v_colwise,rows,cols);CHKERRQ(ierr); 2809 2810 if (reuse == MAT_INPLACE_MATRIX) { 2811 ierr = MatHeaderReplace(A,&mat_elemental);CHKERRQ(ierr); 2812 } else { 2813 *newmat = mat_elemental; 2814 } 2815 PetscFunctionReturn(0); 2816 } 2817 #endif 2818 2819 static PetscErrorCode MatDenseSetLDA_SeqDense(Mat B,PetscInt lda) 2820 { 2821 Mat_SeqDense *b = (Mat_SeqDense*)B->data; 2822 2823 PetscFunctionBegin; 2824 if (lda < B->rmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"LDA %D must be at least matrix dimension %D",lda,B->rmap->n); 2825 b->lda = lda; 2826 PetscFunctionReturn(0); 2827 } 2828 2829 PetscErrorCode MatCreateMPIMatConcatenateSeqMat_SeqDense(MPI_Comm comm,Mat inmat,PetscInt n,MatReuse scall,Mat *outmat) 2830 { 2831 PetscErrorCode ierr; 2832 PetscMPIInt size; 2833 2834 PetscFunctionBegin; 2835 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 2836 if (size == 1) { 2837 if (scall == MAT_INITIAL_MATRIX) { 2838 ierr = MatDuplicate(inmat,MAT_COPY_VALUES,outmat);CHKERRQ(ierr); 2839 } else { 2840 ierr = MatCopy(inmat,*outmat,SAME_NONZERO_PATTERN);CHKERRQ(ierr); 2841 } 2842 } else { 2843 ierr = MatCreateMPIMatConcatenateSeqMat_MPIDense(comm,inmat,n,scall,outmat);CHKERRQ(ierr); 2844 } 2845 PetscFunctionReturn(0); 2846 } 2847 2848 PetscErrorCode MatDenseGetColumnVec_SeqDense(Mat A,PetscInt col,Vec *v) 2849 { 2850 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2851 PetscErrorCode ierr; 2852 2853 PetscFunctionBegin; 2854 if (a->vecinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreColumnVec() first"); 2855 if (a->matinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreSubMatrix() first"); 2856 if (!a->cvec) { 2857 ierr = VecCreateSeqWithArray(PetscObjectComm((PetscObject)A),A->rmap->bs,A->rmap->n,NULL,&a->cvec);CHKERRQ(ierr); 2858 } 2859 a->vecinuse = col + 1; 2860 ierr = MatDenseGetArray(A,(PetscScalar**)&a->ptrinuse);CHKERRQ(ierr); 2861 ierr = VecPlaceArray(a->cvec,a->ptrinuse + (size_t)col * (size_t)a->lda);CHKERRQ(ierr); 2862 *v = a->cvec; 2863 PetscFunctionReturn(0); 2864 } 2865 2866 PetscErrorCode MatDenseRestoreColumnVec_SeqDense(Mat A,PetscInt col,Vec *v) 2867 { 2868 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2869 PetscErrorCode ierr; 2870 2871 PetscFunctionBegin; 2872 if (!a->vecinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseGetColumnVec() first"); 2873 if (!a->cvec) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Missing internal column vector"); 2874 a->vecinuse = 0; 2875 ierr = MatDenseRestoreArray(A,(PetscScalar**)&a->ptrinuse);CHKERRQ(ierr); 2876 ierr = VecResetArray(a->cvec);CHKERRQ(ierr); 2877 *v = NULL; 2878 PetscFunctionReturn(0); 2879 } 2880 2881 PetscErrorCode MatDenseGetColumnVecRead_SeqDense(Mat A,PetscInt col,Vec *v) 2882 { 2883 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2884 PetscErrorCode ierr; 2885 2886 PetscFunctionBegin; 2887 if (a->vecinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreColumnVec() first"); 2888 if (a->matinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreSubMatrix() first"); 2889 if (!a->cvec) { 2890 ierr = VecCreateSeqWithArray(PetscObjectComm((PetscObject)A),A->rmap->bs,A->rmap->n,NULL,&a->cvec);CHKERRQ(ierr); 2891 } 2892 a->vecinuse = col + 1; 2893 ierr = MatDenseGetArrayRead(A,&a->ptrinuse);CHKERRQ(ierr); 2894 ierr = VecPlaceArray(a->cvec,a->ptrinuse + (size_t)col * (size_t)a->lda);CHKERRQ(ierr); 2895 ierr = VecLockReadPush(a->cvec);CHKERRQ(ierr); 2896 *v = a->cvec; 2897 PetscFunctionReturn(0); 2898 } 2899 2900 PetscErrorCode MatDenseRestoreColumnVecRead_SeqDense(Mat A,PetscInt col,Vec *v) 2901 { 2902 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2903 PetscErrorCode ierr; 2904 2905 PetscFunctionBegin; 2906 if (!a->vecinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseGetColumnVec() first"); 2907 if (!a->cvec) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Missing internal column vector"); 2908 a->vecinuse = 0; 2909 ierr = MatDenseRestoreArrayRead(A,&a->ptrinuse);CHKERRQ(ierr); 2910 ierr = VecLockReadPop(a->cvec);CHKERRQ(ierr); 2911 ierr = VecResetArray(a->cvec);CHKERRQ(ierr); 2912 *v = NULL; 2913 PetscFunctionReturn(0); 2914 } 2915 2916 PetscErrorCode MatDenseGetColumnVecWrite_SeqDense(Mat A,PetscInt col,Vec *v) 2917 { 2918 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2919 PetscErrorCode ierr; 2920 2921 PetscFunctionBegin; 2922 if (a->vecinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreColumnVec() first"); 2923 if (a->matinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreSubMatrix() first"); 2924 if (!a->cvec) { 2925 ierr = VecCreateSeqWithArray(PetscObjectComm((PetscObject)A),A->rmap->bs,A->rmap->n,NULL,&a->cvec);CHKERRQ(ierr); 2926 } 2927 a->vecinuse = col + 1; 2928 ierr = MatDenseGetArrayWrite(A,(PetscScalar**)&a->ptrinuse);CHKERRQ(ierr); 2929 ierr = VecPlaceArray(a->cvec,a->ptrinuse + (size_t)col * (size_t)a->lda);CHKERRQ(ierr); 2930 *v = a->cvec; 2931 PetscFunctionReturn(0); 2932 } 2933 2934 PetscErrorCode MatDenseRestoreColumnVecWrite_SeqDense(Mat A,PetscInt col,Vec *v) 2935 { 2936 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2937 PetscErrorCode ierr; 2938 2939 PetscFunctionBegin; 2940 if (!a->vecinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseGetColumnVec() first"); 2941 if (!a->cvec) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Missing internal column vector"); 2942 a->vecinuse = 0; 2943 ierr = MatDenseRestoreArrayWrite(A,(PetscScalar**)&a->ptrinuse);CHKERRQ(ierr); 2944 ierr = VecResetArray(a->cvec);CHKERRQ(ierr); 2945 *v = NULL; 2946 PetscFunctionReturn(0); 2947 } 2948 2949 PetscErrorCode MatDenseGetSubMatrix_SeqDense(Mat A,PetscInt cbegin,PetscInt cend,Mat *v) 2950 { 2951 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2952 PetscErrorCode ierr; 2953 2954 PetscFunctionBegin; 2955 if (a->vecinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreColumnVec() first"); 2956 if (a->matinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseRestoreSubMatrix() first"); 2957 if (a->cmat && cend-cbegin != a->cmat->cmap->N) { 2958 ierr = MatDestroy(&a->cmat);CHKERRQ(ierr); 2959 } 2960 ierr = MatDenseGetArray(A,(PetscScalar**)&a->ptrinuse);CHKERRQ(ierr); 2961 if (!a->cmat) { 2962 ierr = MatCreateDense(PetscObjectComm((PetscObject)A),A->rmap->n,PETSC_DECIDE,A->rmap->N,cend-cbegin,(PetscScalar*)a->ptrinuse + (size_t)cbegin * (size_t)a->lda,&a->cmat);CHKERRQ(ierr); 2963 ierr = MatDenseSetLDA(a->cmat,a->lda);CHKERRQ(ierr); 2964 } else { 2965 ierr = MatDensePlaceArray(a->cmat,a->ptrinuse + (size_t)cbegin * (size_t)a->lda);CHKERRQ(ierr); 2966 } 2967 a->matinuse = cbegin + 1; 2968 *v = a->cmat; 2969 PetscFunctionReturn(0); 2970 } 2971 2972 PetscErrorCode MatDenseRestoreSubMatrix_SeqDense(Mat A,Mat *v) 2973 { 2974 Mat_SeqDense *a = (Mat_SeqDense*)A->data; 2975 PetscErrorCode ierr; 2976 2977 PetscFunctionBegin; 2978 if (!a->matinuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Need to call MatDenseGetSubMatrix() first"); 2979 if (!a->cmat) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Missing internal column matrix"); 2980 a->matinuse = 0; 2981 ierr = MatDenseRestoreArray(A,(PetscScalar**)&a->ptrinuse);CHKERRQ(ierr); 2982 ierr = MatDenseResetArray(a->cmat);CHKERRQ(ierr); 2983 *v = NULL; 2984 PetscFunctionReturn(0); 2985 } 2986 2987 /*MC 2988 MATSEQDENSE - MATSEQDENSE = "seqdense" - A matrix type to be used for sequential dense matrices. 2989 2990 Options Database Keys: 2991 . -mat_type seqdense - sets the matrix type to "seqdense" during a call to MatSetFromOptions() 2992 2993 Level: beginner 2994 2995 .seealso: MatCreateSeqDense() 2996 2997 M*/ 2998 PetscErrorCode MatCreate_SeqDense(Mat B) 2999 { 3000 Mat_SeqDense *b; 3001 PetscErrorCode ierr; 3002 PetscMPIInt size; 3003 3004 PetscFunctionBegin; 3005 ierr = MPI_Comm_size(PetscObjectComm((PetscObject)B),&size);CHKERRQ(ierr); 3006 if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Comm must be of size 1"); 3007 3008 ierr = PetscNewLog(B,&b);CHKERRQ(ierr); 3009 ierr = PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));CHKERRQ(ierr); 3010 B->data = (void*)b; 3011 3012 b->roworiented = PETSC_TRUE; 3013 3014 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseGetLDA_C",MatDenseGetLDA_SeqDense);CHKERRQ(ierr); 3015 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseSetLDA_C",MatDenseSetLDA_SeqDense);CHKERRQ(ierr); 3016 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseGetArray_C",MatDenseGetArray_SeqDense);CHKERRQ(ierr); 3017 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseRestoreArray_C",MatDenseRestoreArray_SeqDense);CHKERRQ(ierr); 3018 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDensePlaceArray_C",MatDensePlaceArray_SeqDense);CHKERRQ(ierr); 3019 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseResetArray_C",MatDenseResetArray_SeqDense);CHKERRQ(ierr); 3020 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseReplaceArray_C",MatDenseReplaceArray_SeqDense);CHKERRQ(ierr); 3021 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseGetArrayRead_C",MatDenseGetArray_SeqDense);CHKERRQ(ierr); 3022 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseRestoreArrayRead_C",MatDenseRestoreArray_SeqDense);CHKERRQ(ierr); 3023 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseGetArrayWrite_C",MatDenseGetArray_SeqDense);CHKERRQ(ierr); 3024 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseRestoreArrayWrite_C",MatDenseRestoreArray_SeqDense);CHKERRQ(ierr); 3025 ierr = PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqdense_seqaij_C",MatConvert_SeqDense_SeqAIJ);CHKERRQ(ierr); 3026 #if defined(PETSC_HAVE_ELEMENTAL) 3027 ierr = PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqdense_elemental_C",MatConvert_SeqDense_Elemental);CHKERRQ(ierr); 3028 #endif 3029 #if defined(PETSC_HAVE_CUDA) 3030 ierr = PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqdense_seqdensecuda_C",MatConvert_SeqDense_SeqDenseCUDA);CHKERRQ(ierr); 3031 ierr = PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqdensecuda_seqdensecuda_C",MatProductSetFromOptions_SeqDense);CHKERRQ(ierr); 3032 ierr = PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqdensecuda_seqdense_C",MatProductSetFromOptions_SeqDense);CHKERRQ(ierr); 3033 ierr = PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqdense_seqdensecuda_C",MatProductSetFromOptions_SeqDense);CHKERRQ(ierr); 3034 #endif 3035 ierr = PetscObjectComposeFunction((PetscObject)B,"MatSeqDenseSetPreallocation_C",MatSeqDenseSetPreallocation_SeqDense);CHKERRQ(ierr); 3036 ierr = PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqaij_seqdense_C",MatProductSetFromOptions_SeqAIJ_SeqDense);CHKERRQ(ierr); 3037 ierr = PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqdense_seqdense_C",MatProductSetFromOptions_SeqDense);CHKERRQ(ierr); 3038 ierr = PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqbaij_seqdense_C",MatProductSetFromOptions_SeqXBAIJ_SeqDense);CHKERRQ(ierr); 3039 ierr = PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqsbaij_seqdense_C",MatProductSetFromOptions_SeqXBAIJ_SeqDense);CHKERRQ(ierr); 3040 3041 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseGetColumn_C",MatDenseGetColumn_SeqDense);CHKERRQ(ierr); 3042 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseRestoreColumn_C",MatDenseRestoreColumn_SeqDense);CHKERRQ(ierr); 3043 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseGetColumnVec_C",MatDenseGetColumnVec_SeqDense);CHKERRQ(ierr); 3044 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseRestoreColumnVec_C",MatDenseRestoreColumnVec_SeqDense);CHKERRQ(ierr); 3045 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseGetColumnVecRead_C",MatDenseGetColumnVecRead_SeqDense);CHKERRQ(ierr); 3046 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseRestoreColumnVecRead_C",MatDenseRestoreColumnVecRead_SeqDense);CHKERRQ(ierr); 3047 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseGetColumnVecWrite_C",MatDenseGetColumnVecWrite_SeqDense);CHKERRQ(ierr); 3048 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseRestoreColumnVecWrite_C",MatDenseRestoreColumnVecWrite_SeqDense);CHKERRQ(ierr); 3049 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseGetSubMatrix_C",MatDenseGetSubMatrix_SeqDense);CHKERRQ(ierr); 3050 ierr = PetscObjectComposeFunction((PetscObject)B,"MatDenseRestoreSubMatrix_C",MatDenseRestoreSubMatrix_SeqDense);CHKERRQ(ierr); 3051 ierr = PetscObjectChangeTypeName((PetscObject)B,MATSEQDENSE);CHKERRQ(ierr); 3052 PetscFunctionReturn(0); 3053 } 3054 3055 /*@C 3056 MatDenseGetColumn - gives access to a column of a dense matrix. This is only the local part of the column. You MUST call MatDenseRestoreColumn() to avoid memory bleeding. 3057 3058 Not Collective 3059 3060 Input Parameters: 3061 + mat - a MATSEQDENSE or MATMPIDENSE matrix 3062 - col - column index 3063 3064 Output Parameter: 3065 . vals - pointer to the data 3066 3067 Level: intermediate 3068 3069 .seealso: MatDenseRestoreColumn() 3070 @*/ 3071 PetscErrorCode MatDenseGetColumn(Mat A,PetscInt col,PetscScalar **vals) 3072 { 3073 PetscErrorCode ierr; 3074 3075 PetscFunctionBegin; 3076 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 3077 PetscValidLogicalCollectiveInt(A,col,2); 3078 PetscValidPointer(vals,3); 3079 ierr = PetscUseMethod(A,"MatDenseGetColumn_C",(Mat,PetscInt,PetscScalar**),(A,col,vals));CHKERRQ(ierr); 3080 PetscFunctionReturn(0); 3081 } 3082 3083 /*@C 3084 MatDenseRestoreColumn - returns access to a column of a dense matrix which is returned by MatDenseGetColumn(). 3085 3086 Not Collective 3087 3088 Input Parameter: 3089 . mat - a MATSEQDENSE or MATMPIDENSE matrix 3090 3091 Output Parameter: 3092 . vals - pointer to the data 3093 3094 Level: intermediate 3095 3096 .seealso: MatDenseGetColumn() 3097 @*/ 3098 PetscErrorCode MatDenseRestoreColumn(Mat A,PetscScalar **vals) 3099 { 3100 PetscErrorCode ierr; 3101 3102 PetscFunctionBegin; 3103 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 3104 PetscValidPointer(vals,2); 3105 ierr = PetscUseMethod(A,"MatDenseRestoreColumn_C",(Mat,PetscScalar**),(A,vals));CHKERRQ(ierr); 3106 PetscFunctionReturn(0); 3107 } 3108 3109 /*@C 3110 MatDenseGetColumnVec - Gives read-write access to a column of a dense matrix, represented as a Vec. 3111 3112 Collective 3113 3114 Input Parameters: 3115 + mat - the Mat object 3116 - col - the column index 3117 3118 Output Parameter: 3119 . v - the vector 3120 3121 Notes: 3122 The vector is owned by PETSc. Users need to call MatDenseRestoreColumnVec() when the vector is no longer needed. 3123 Use MatDenseGetColumnVecRead() to obtain read-only access or MatDenseGetColumnVecWrite() for write-only access. 3124 3125 Level: intermediate 3126 3127 .seealso: MATDENSE, MATDENSECUDA, MatDenseGetColumnVecRead(), MatDenseGetColumnVecWrite(), MatDenseRestoreColumnVec(), MatDenseRestoreColumnVecRead(), MatDenseRestoreColumnVecWrite() 3128 @*/ 3129 PetscErrorCode MatDenseGetColumnVec(Mat A,PetscInt col,Vec *v) 3130 { 3131 PetscErrorCode ierr; 3132 3133 PetscFunctionBegin; 3134 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 3135 PetscValidType(A,1); 3136 PetscValidLogicalCollectiveInt(A,col,2); 3137 PetscValidPointer(v,3); 3138 if (!A->preallocated) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ORDER,"Matrix not preallocated"); 3139 if (col < 0 || col > A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Invalid col %D, should be in [0,%D)",col,A->cmap->N); 3140 ierr = PetscUseMethod(A,"MatDenseGetColumnVec_C",(Mat,PetscInt,Vec*),(A,col,v));CHKERRQ(ierr); 3141 PetscFunctionReturn(0); 3142 } 3143 3144 /*@C 3145 MatDenseRestoreColumnVec - Returns access to a column of a dense matrix obtained from MatDenseGetColumnVec(). 3146 3147 Collective 3148 3149 Input Parameters: 3150 + mat - the Mat object 3151 . col - the column index 3152 - v - the Vec object 3153 3154 Level: intermediate 3155 3156 .seealso: MATDENSE, MATDENSECUDA, MatDenseGetColumnVec(), MatDenseGetColumnVecRead(), MatDenseGetColumnVecWrite(), MatDenseRestoreColumnVecRead(), MatDenseRestoreColumnVecWrite() 3157 @*/ 3158 PetscErrorCode MatDenseRestoreColumnVec(Mat A,PetscInt col,Vec *v) 3159 { 3160 PetscErrorCode ierr; 3161 3162 PetscFunctionBegin; 3163 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 3164 PetscValidType(A,1); 3165 PetscValidLogicalCollectiveInt(A,col,2); 3166 PetscValidPointer(v,3); 3167 if (!A->preallocated) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ORDER,"Matrix not preallocated"); 3168 if (col < 0 || col > A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Invalid col %D, should be in [0,%D)",col,A->cmap->N); 3169 ierr = PetscUseMethod(A,"MatDenseRestoreColumnVec_C",(Mat,PetscInt,Vec*),(A,col,v));CHKERRQ(ierr); 3170 PetscFunctionReturn(0); 3171 } 3172 3173 /*@C 3174 MatDenseGetColumnVecRead - Gives read-only access to a column of a dense matrix, represented as a Vec. 3175 3176 Collective 3177 3178 Input Parameters: 3179 + mat - the Mat object 3180 - col - the column index 3181 3182 Output Parameter: 3183 . v - the vector 3184 3185 Notes: 3186 The vector is owned by PETSc and users cannot modify it. 3187 Users need to call MatDenseRestoreColumnVecRead() when the vector is no longer needed. 3188 Use MatDenseGetColumnVec() to obtain read-write access or MatDenseGetColumnVecWrite() for write-only access. 3189 3190 Level: intermediate 3191 3192 .seealso: MATDENSE, MATDENSECUDA, MatDenseGetColumnVec(), MatDenseGetColumnVecWrite(), MatDenseRestoreColumnVec(), MatDenseRestoreColumnVecRead(), MatDenseRestoreColumnVecWrite() 3193 @*/ 3194 PetscErrorCode MatDenseGetColumnVecRead(Mat A,PetscInt col,Vec *v) 3195 { 3196 PetscErrorCode ierr; 3197 3198 PetscFunctionBegin; 3199 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 3200 PetscValidType(A,1); 3201 PetscValidLogicalCollectiveInt(A,col,2); 3202 PetscValidPointer(v,3); 3203 if (!A->preallocated) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ORDER,"Matrix not preallocated"); 3204 if (col < 0 || col > A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Invalid col %D, should be in [0,%D)",col,A->cmap->N); 3205 ierr = PetscUseMethod(A,"MatDenseGetColumnVecRead_C",(Mat,PetscInt,Vec*),(A,col,v));CHKERRQ(ierr); 3206 PetscFunctionReturn(0); 3207 } 3208 3209 /*@C 3210 MatDenseRestoreColumnVecRead - Returns access to a column of a dense matrix obtained from MatDenseGetColumnVecRead(). 3211 3212 Collective 3213 3214 Input Parameters: 3215 + mat - the Mat object 3216 . col - the column index 3217 - v - the Vec object 3218 3219 Level: intermediate 3220 3221 .seealso: MATDENSE, MATDENSECUDA, MatDenseGetColumnVec(), MatDenseGetColumnVecRead(), MatDenseGetColumnVecWrite(), MatDenseRestoreColumnVec(), MatDenseRestoreColumnVecWrite() 3222 @*/ 3223 PetscErrorCode MatDenseRestoreColumnVecRead(Mat A,PetscInt col,Vec *v) 3224 { 3225 PetscErrorCode ierr; 3226 3227 PetscFunctionBegin; 3228 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 3229 PetscValidType(A,1); 3230 PetscValidLogicalCollectiveInt(A,col,2); 3231 PetscValidPointer(v,3); 3232 if (!A->preallocated) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ORDER,"Matrix not preallocated"); 3233 if (col < 0 || col > A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Invalid col %D, should be in [0,%D)",col,A->cmap->N); 3234 ierr = PetscUseMethod(A,"MatDenseRestoreColumnVecRead_C",(Mat,PetscInt,Vec*),(A,col,v));CHKERRQ(ierr); 3235 PetscFunctionReturn(0); 3236 } 3237 3238 /*@C 3239 MatDenseGetColumnVecWrite - Gives write-only access to a column of a dense matrix, represented as a Vec. 3240 3241 Collective 3242 3243 Input Parameters: 3244 + mat - the Mat object 3245 - col - the column index 3246 3247 Output Parameter: 3248 . v - the vector 3249 3250 Notes: 3251 The vector is owned by PETSc. Users need to call MatDenseRestoreColumnVecWrite() when the vector is no longer needed. 3252 Use MatDenseGetColumnVec() to obtain read-write access or MatDenseGetColumnVecRead() for read-only access. 3253 3254 Level: intermediate 3255 3256 .seealso: MATDENSE, MATDENSECUDA, MatDenseGetColumnVec(), MatDenseGetColumnVecRead(), MatDenseRestoreColumnVec(), MatDenseRestoreColumnVecRead(), MatDenseRestoreColumnVecWrite() 3257 @*/ 3258 PetscErrorCode MatDenseGetColumnVecWrite(Mat A,PetscInt col,Vec *v) 3259 { 3260 PetscErrorCode ierr; 3261 3262 PetscFunctionBegin; 3263 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 3264 PetscValidType(A,1); 3265 PetscValidLogicalCollectiveInt(A,col,2); 3266 PetscValidPointer(v,3); 3267 if (!A->preallocated) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ORDER,"Matrix not preallocated"); 3268 if (col < 0 || col > A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Invalid col %D, should be in [0,%D)",col,A->cmap->N); 3269 ierr = PetscUseMethod(A,"MatDenseGetColumnVecWrite_C",(Mat,PetscInt,Vec*),(A,col,v));CHKERRQ(ierr); 3270 PetscFunctionReturn(0); 3271 } 3272 3273 /*@C 3274 MatDenseRestoreColumnVecWrite - Returns access to a column of a dense matrix obtained from MatDenseGetColumnVecWrite(). 3275 3276 Collective 3277 3278 Input Parameters: 3279 + mat - the Mat object 3280 . col - the column index 3281 - v - the Vec object 3282 3283 Level: intermediate 3284 3285 .seealso: MATDENSE, MATDENSECUDA, MatDenseGetColumnVec(), MatDenseGetColumnVecRead(), MatDenseGetColumnVecWrite(), MatDenseRestoreColumnVec(), MatDenseRestoreColumnVecRead() 3286 @*/ 3287 PetscErrorCode MatDenseRestoreColumnVecWrite(Mat A,PetscInt col,Vec *v) 3288 { 3289 PetscErrorCode ierr; 3290 3291 PetscFunctionBegin; 3292 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 3293 PetscValidType(A,1); 3294 PetscValidLogicalCollectiveInt(A,col,2); 3295 PetscValidPointer(v,3); 3296 if (!A->preallocated) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ORDER,"Matrix not preallocated"); 3297 if (col < 0 || col > A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Invalid col %D, should be in [0,%D)",col,A->cmap->N); 3298 ierr = PetscUseMethod(A,"MatDenseRestoreColumnVecWrite_C",(Mat,PetscInt,Vec*),(A,col,v));CHKERRQ(ierr); 3299 PetscFunctionReturn(0); 3300 } 3301 3302 /*@C 3303 MatDenseGetSubMatrix - Gives access to a block of columns of a dense matrix, represented as a Mat. 3304 3305 Collective 3306 3307 Input Parameters: 3308 + mat - the Mat object 3309 . cbegin - the first index in the block 3310 - cend - the last index in the block 3311 3312 Output Parameter: 3313 . v - the matrix 3314 3315 Notes: 3316 The matrix is owned by PETSc. Users need to call MatDenseRestoreSubMatrix() when the matrix is no longer needed. 3317 3318 Level: intermediate 3319 3320 .seealso: MATDENSE, MATDENSECUDA, MatDenseGetColumnVec(), MatDenseRestoreColumnVec(), MatDenseRestoreSubMatrix() 3321 @*/ 3322 PetscErrorCode MatDenseGetSubMatrix(Mat A,PetscInt cbegin,PetscInt cend,Mat *v) 3323 { 3324 PetscErrorCode ierr; 3325 3326 PetscFunctionBegin; 3327 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 3328 PetscValidType(A,1); 3329 PetscValidLogicalCollectiveInt(A,cbegin,2); 3330 PetscValidLogicalCollectiveInt(A,cend,3); 3331 PetscValidPointer(v,4); 3332 if (!A->preallocated) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ORDER,"Matrix not preallocated"); 3333 if (cbegin < 0 || cbegin > A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Invalid cbegin %D, should be in [0,%D)",cbegin,A->cmap->N); 3334 if (cend < 0 || cend > A->cmap->N || cend <= cbegin) SETERRQ3(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Invalid cend %D, should be in [0,%D) and strictly greater than cbegin %D",cend,A->cmap->N,cbegin); 3335 ierr = PetscUseMethod(A,"MatDenseGetSubMatrix_C",(Mat,PetscInt,PetscInt,Mat*),(A,cbegin,cend,v));CHKERRQ(ierr); 3336 PetscFunctionReturn(0); 3337 } 3338 3339 /*@C 3340 MatDenseRestoreSubMatrix - Returns access to a block of columns of a dense matrix obtained from MatDenseGetSubMatrix(). 3341 3342 Collective 3343 3344 Input Parameters: 3345 + mat - the Mat object 3346 - v - the Mat object 3347 3348 Level: intermediate 3349 3350 .seealso: MATDENSE, MATDENSECUDA, MatDenseGetColumnVec(), MatDenseRestoreColumnVec(), MatDenseGetSubMatrix() 3351 @*/ 3352 PetscErrorCode MatDenseRestoreSubMatrix(Mat A,Mat *v) 3353 { 3354 PetscErrorCode ierr; 3355 3356 PetscFunctionBegin; 3357 PetscValidHeaderSpecific(A,MAT_CLASSID,1); 3358 PetscValidType(A,1); 3359 PetscValidPointer(v,2); 3360 if (!A->preallocated) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ORDER,"Matrix not preallocated"); 3361 ierr = PetscUseMethod(A,"MatDenseRestoreSubMatrix_C",(Mat,Mat*),(A,v));CHKERRQ(ierr); 3362 PetscFunctionReturn(0); 3363 } 3364