1 2 /* 3 Provides an interface to the MUMPS sparse solver 4 */ 5 6 #include <../src/mat/impls/aij/mpi/mpiaij.h> /*I "petscmat.h" I*/ 7 #include <../src/mat/impls/sbaij/mpi/mpisbaij.h> 8 9 EXTERN_C_BEGIN 10 #if defined(PETSC_USE_COMPLEX) 11 #include <zmumps_c.h> 12 #else 13 #include <dmumps_c.h> 14 #endif 15 EXTERN_C_END 16 #define JOB_INIT -1 17 #define JOB_FACTSYMBOLIC 1 18 #define JOB_FACTNUMERIC 2 19 #define JOB_SOLVE 3 20 #define JOB_END -2 21 22 23 /* macros s.t. indices match MUMPS documentation */ 24 #define ICNTL(I) icntl[(I)-1] 25 #define CNTL(I) cntl[(I)-1] 26 #define INFOG(I) infog[(I)-1] 27 #define INFO(I) info[(I)-1] 28 #define RINFOG(I) rinfog[(I)-1] 29 #define RINFO(I) rinfo[(I)-1] 30 31 typedef struct { 32 #if defined(PETSC_USE_COMPLEX) 33 ZMUMPS_STRUC_C id; 34 #else 35 DMUMPS_STRUC_C id; 36 #endif 37 MatStructure matstruc; 38 PetscMPIInt myid,size; 39 PetscInt *irn,*jcn,nz,sym,nSolve; 40 PetscScalar *val; 41 MPI_Comm comm_mumps; 42 VecScatter scat_rhs, scat_sol; 43 PetscBool isAIJ,CleanUpMUMPS; 44 Vec b_seq,x_seq; 45 PetscErrorCode (*Destroy)(Mat); 46 PetscErrorCode (*ConvertToTriples)(Mat, int, MatReuse, int*, int**, int**, PetscScalar**); 47 } Mat_MUMPS; 48 49 extern PetscErrorCode MatDuplicate_MUMPS(Mat,MatDuplicateOption,Mat*); 50 51 52 /* MatConvertToTriples_A_B */ 53 /*convert Petsc matrix to triples: row[nz], col[nz], val[nz] */ 54 /* 55 input: 56 A - matrix in aij,baij or sbaij (bs=1) format 57 shift - 0: C style output triple; 1: Fortran style output triple. 58 reuse - MAT_INITIAL_MATRIX: spaces are allocated and values are set for the triple 59 MAT_REUSE_MATRIX: only the values in v array are updated 60 output: 61 nnz - dim of r, c, and v (number of local nonzero entries of A) 62 r, c, v - row and col index, matrix values (matrix triples) 63 */ 64 65 #undef __FUNCT__ 66 #define __FUNCT__ "MatConvertToTriples_seqaij_seqaij" 67 PetscErrorCode MatConvertToTriples_seqaij_seqaij(Mat A,int shift,MatReuse reuse,int *nnz,int **r, int **c, PetscScalar **v) 68 { 69 const PetscInt *ai,*aj,*ajj,M=A->rmap->n; 70 PetscInt nz,rnz,i,j; 71 PetscErrorCode ierr; 72 PetscInt *row,*col; 73 Mat_SeqAIJ *aa=(Mat_SeqAIJ*)A->data; 74 75 PetscFunctionBegin; 76 *v=aa->a; 77 if (reuse == MAT_INITIAL_MATRIX){ 78 nz = aa->nz; ai = aa->i; aj = aa->j; 79 *nnz = nz; 80 ierr = PetscMalloc(2*nz*sizeof(PetscInt), &row);CHKERRQ(ierr); 81 col = row + nz; 82 83 nz = 0; 84 for(i=0; i<M; i++) { 85 rnz = ai[i+1] - ai[i]; 86 ajj = aj + ai[i]; 87 for(j=0; j<rnz; j++) { 88 row[nz] = i+shift; col[nz++] = ajj[j] + shift; 89 } 90 } 91 *r = row; *c = col; 92 } 93 PetscFunctionReturn(0); 94 } 95 96 #undef __FUNCT__ 97 #define __FUNCT__ "MatConvertToTriples_seqbaij_seqaij" 98 PetscErrorCode MatConvertToTriples_seqbaij_seqaij(Mat A,int shift,MatReuse reuse,int *nnz,int **r, int **c, PetscScalar **v) 99 { 100 Mat_SeqBAIJ *aa=(Mat_SeqBAIJ*)A->data; 101 const PetscInt *ai,*aj,*ajj,bs=A->rmap->bs,bs2=aa->bs2,M=A->rmap->N/bs; 102 PetscInt nz,idx=0,rnz,i,j,k,m; 103 PetscErrorCode ierr; 104 PetscInt *row,*col; 105 106 PetscFunctionBegin; 107 *v = aa->a; 108 if (reuse == MAT_INITIAL_MATRIX){ 109 ai = aa->i; aj = aa->j; 110 nz = bs2*aa->nz; 111 *nnz = nz; 112 ierr = PetscMalloc(2*nz*sizeof(PetscInt), &row);CHKERRQ(ierr); 113 col = row + nz; 114 115 for(i=0; i<M; i++) { 116 ajj = aj + ai[i]; 117 rnz = ai[i+1] - ai[i]; 118 for(k=0; k<rnz; k++) { 119 for(j=0; j<bs; j++) { 120 for(m=0; m<bs; m++) { 121 row[idx] = i*bs + m + shift; 122 col[idx++] = bs*(ajj[k]) + j + shift; 123 } 124 } 125 } 126 } 127 *r = row; *c = col; 128 } 129 PetscFunctionReturn(0); 130 } 131 132 #undef __FUNCT__ 133 #define __FUNCT__ "MatConvertToTriples_seqsbaij_seqsbaij" 134 PetscErrorCode MatConvertToTriples_seqsbaij_seqsbaij(Mat A,int shift,MatReuse reuse,int *nnz,int **r, int **c, PetscScalar **v) 135 { 136 const PetscInt *ai, *aj,*ajj,M=A->rmap->n; 137 PetscInt nz,rnz,i,j; 138 PetscErrorCode ierr; 139 PetscInt *row,*col; 140 Mat_SeqSBAIJ *aa=(Mat_SeqSBAIJ*)A->data; 141 142 PetscFunctionBegin; 143 *v = aa->a; 144 if (reuse == MAT_INITIAL_MATRIX){ 145 nz = aa->nz;ai=aa->i; aj=aa->j;*v=aa->a; 146 *nnz = nz; 147 ierr = PetscMalloc(2*nz*sizeof(PetscInt), &row);CHKERRQ(ierr); 148 col = row + nz; 149 150 nz = 0; 151 for(i=0; i<M; i++) { 152 rnz = ai[i+1] - ai[i]; 153 ajj = aj + ai[i]; 154 for(j=0; j<rnz; j++) { 155 row[nz] = i+shift; col[nz++] = ajj[j] + shift; 156 } 157 } 158 *r = row; *c = col; 159 } 160 PetscFunctionReturn(0); 161 } 162 163 #undef __FUNCT__ 164 #define __FUNCT__ "MatConvertToTriples_seqaij_seqsbaij" 165 PetscErrorCode MatConvertToTriples_seqaij_seqsbaij(Mat A,int shift,MatReuse reuse,int *nnz,int **r, int **c, PetscScalar **v) 166 { 167 const PetscInt *ai,*aj,*ajj,*adiag,M=A->rmap->n; 168 PetscInt nz,rnz,i,j; 169 const PetscScalar *av,*v1; 170 PetscScalar *val; 171 PetscErrorCode ierr; 172 PetscInt *row,*col; 173 Mat_SeqSBAIJ *aa=(Mat_SeqSBAIJ*)A->data; 174 175 PetscFunctionBegin; 176 ai=aa->i; aj=aa->j;av=aa->a; 177 adiag=aa->diag; 178 if (reuse == MAT_INITIAL_MATRIX){ 179 nz = M + (aa->nz-M)/2; 180 *nnz = nz; 181 ierr = PetscMalloc((2*nz*sizeof(PetscInt)+nz*sizeof(PetscScalar)), &row);CHKERRQ(ierr); 182 col = row + nz; 183 val = (PetscScalar*)(col + nz); 184 185 nz = 0; 186 for(i=0; i<M; i++) { 187 rnz = ai[i+1] - adiag[i]; 188 ajj = aj + adiag[i]; 189 v1 = av + adiag[i]; 190 for(j=0; j<rnz; j++) { 191 row[nz] = i+shift; col[nz] = ajj[j] + shift; val[nz++] = v1[j]; 192 } 193 } 194 *r = row; *c = col; *v = val; 195 } else { 196 nz = 0; val = *v; 197 for(i=0; i <M; i++) { 198 rnz = ai[i+1] - adiag[i]; 199 ajj = aj + adiag[i]; 200 v1 = av + adiag[i]; 201 for(j=0; j<rnz; j++) { 202 val[nz++] = v1[j]; 203 } 204 } 205 } 206 PetscFunctionReturn(0); 207 } 208 209 #undef __FUNCT__ 210 #define __FUNCT__ "MatConvertToTriples_mpisbaij_mpisbaij" 211 PetscErrorCode MatConvertToTriples_mpisbaij_mpisbaij(Mat A,int shift,MatReuse reuse,int *nnz,int **r, int **c, PetscScalar **v) 212 { 213 const PetscInt *ai, *aj, *bi, *bj,*garray,m=A->rmap->n,*ajj,*bjj; 214 PetscErrorCode ierr; 215 PetscInt rstart,nz,i,j,jj,irow,countA,countB; 216 PetscInt *row,*col; 217 const PetscScalar *av, *bv,*v1,*v2; 218 PetscScalar *val; 219 Mat_MPISBAIJ *mat = (Mat_MPISBAIJ*)A->data; 220 Mat_SeqSBAIJ *aa=(Mat_SeqSBAIJ*)(mat->A)->data; 221 Mat_SeqBAIJ *bb=(Mat_SeqBAIJ*)(mat->B)->data; 222 223 PetscFunctionBegin; 224 ai=aa->i; aj=aa->j; bi=bb->i; bj=bb->j; rstart= A->rmap->rstart; 225 garray = mat->garray; 226 av=aa->a; bv=bb->a; 227 228 if (reuse == MAT_INITIAL_MATRIX){ 229 nz = aa->nz + bb->nz; 230 *nnz = nz; 231 ierr = PetscMalloc((2*nz*sizeof(PetscInt)+nz*sizeof(PetscScalar)), &row);CHKERRQ(ierr); 232 col = row + nz; 233 val = (PetscScalar*)(col + nz); 234 235 *r = row; *c = col; *v = val; 236 } else { 237 row = *r; col = *c; val = *v; 238 } 239 240 jj = 0; irow = rstart; 241 for ( i=0; i<m; i++ ) { 242 ajj = aj + ai[i]; /* ptr to the beginning of this row */ 243 countA = ai[i+1] - ai[i]; 244 countB = bi[i+1] - bi[i]; 245 bjj = bj + bi[i]; 246 v1 = av + ai[i]; 247 v2 = bv + bi[i]; 248 249 /* A-part */ 250 for (j=0; j<countA; j++){ 251 if (reuse == MAT_INITIAL_MATRIX) { 252 row[jj] = irow + shift; col[jj] = rstart + ajj[j] + shift; 253 } 254 val[jj++] = v1[j]; 255 } 256 257 /* B-part */ 258 for(j=0; j < countB; j++){ 259 if (reuse == MAT_INITIAL_MATRIX) { 260 row[jj] = irow + shift; col[jj] = garray[bjj[j]] + shift; 261 } 262 val[jj++] = v2[j]; 263 } 264 irow++; 265 } 266 PetscFunctionReturn(0); 267 } 268 269 #undef __FUNCT__ 270 #define __FUNCT__ "MatConvertToTriples_mpiaij_mpiaij" 271 PetscErrorCode MatConvertToTriples_mpiaij_mpiaij(Mat A,int shift,MatReuse reuse,int *nnz,int **r, int **c, PetscScalar **v) 272 { 273 const PetscInt *ai, *aj, *bi, *bj,*garray,m=A->rmap->n,*ajj,*bjj; 274 PetscErrorCode ierr; 275 PetscInt rstart,nz,i,j,jj,irow,countA,countB; 276 PetscInt *row,*col; 277 const PetscScalar *av, *bv,*v1,*v2; 278 PetscScalar *val; 279 Mat_MPIAIJ *mat = (Mat_MPIAIJ*)A->data; 280 Mat_SeqAIJ *aa=(Mat_SeqAIJ*)(mat->A)->data; 281 Mat_SeqAIJ *bb=(Mat_SeqAIJ*)(mat->B)->data; 282 283 PetscFunctionBegin; 284 ai=aa->i; aj=aa->j; bi=bb->i; bj=bb->j; rstart= A->rmap->rstart; 285 garray = mat->garray; 286 av=aa->a; bv=bb->a; 287 288 if (reuse == MAT_INITIAL_MATRIX){ 289 nz = aa->nz + bb->nz; 290 *nnz = nz; 291 ierr = PetscMalloc((2*nz*sizeof(PetscInt)+nz*sizeof(PetscScalar)), &row);CHKERRQ(ierr); 292 col = row + nz; 293 val = (PetscScalar*)(col + nz); 294 295 *r = row; *c = col; *v = val; 296 } else { 297 row = *r; col = *c; val = *v; 298 } 299 300 jj = 0; irow = rstart; 301 for ( i=0; i<m; i++ ) { 302 ajj = aj + ai[i]; /* ptr to the beginning of this row */ 303 countA = ai[i+1] - ai[i]; 304 countB = bi[i+1] - bi[i]; 305 bjj = bj + bi[i]; 306 v1 = av + ai[i]; 307 v2 = bv + bi[i]; 308 309 /* A-part */ 310 for (j=0; j<countA; j++){ 311 if (reuse == MAT_INITIAL_MATRIX){ 312 row[jj] = irow + shift; col[jj] = rstart + ajj[j] + shift; 313 } 314 val[jj++] = v1[j]; 315 } 316 317 /* B-part */ 318 for(j=0; j < countB; j++){ 319 if (reuse == MAT_INITIAL_MATRIX){ 320 row[jj] = irow + shift; col[jj] = garray[bjj[j]] + shift; 321 } 322 val[jj++] = v2[j]; 323 } 324 irow++; 325 } 326 PetscFunctionReturn(0); 327 } 328 329 #undef __FUNCT__ 330 #define __FUNCT__ "MatConvertToTriples_mpibaij_mpiaij" 331 PetscErrorCode MatConvertToTriples_mpibaij_mpiaij(Mat A,int shift,MatReuse reuse,int *nnz,int **r, int **c, PetscScalar **v) 332 { 333 Mat_MPIBAIJ *mat = (Mat_MPIBAIJ*)A->data; 334 Mat_SeqBAIJ *aa=(Mat_SeqBAIJ*)(mat->A)->data; 335 Mat_SeqBAIJ *bb=(Mat_SeqBAIJ*)(mat->B)->data; 336 const PetscInt *ai = aa->i, *bi = bb->i, *aj = aa->j, *bj = bb->j,*ajj, *bjj; 337 const PetscInt *garray = mat->garray,mbs=mat->mbs,rstart=A->rmap->rstart; 338 const PetscInt bs = A->rmap->bs,bs2=mat->bs2; 339 PetscErrorCode ierr; 340 PetscInt nz,i,j,k,n,jj,irow,countA,countB,idx; 341 PetscInt *row,*col; 342 const PetscScalar *av=aa->a, *bv=bb->a,*v1,*v2; 343 PetscScalar *val; 344 345 PetscFunctionBegin; 346 347 if (reuse == MAT_INITIAL_MATRIX) { 348 nz = bs2*(aa->nz + bb->nz); 349 *nnz = nz; 350 ierr = PetscMalloc((2*nz*sizeof(PetscInt)+nz*sizeof(PetscScalar)), &row);CHKERRQ(ierr); 351 col = row + nz; 352 val = (PetscScalar*)(col + nz); 353 354 *r = row; *c = col; *v = val; 355 } else { 356 row = *r; col = *c; val = *v; 357 } 358 359 jj = 0; irow = rstart; 360 for ( i=0; i<mbs; i++ ) { 361 countA = ai[i+1] - ai[i]; 362 countB = bi[i+1] - bi[i]; 363 ajj = aj + ai[i]; 364 bjj = bj + bi[i]; 365 v1 = av + bs2*ai[i]; 366 v2 = bv + bs2*bi[i]; 367 368 idx = 0; 369 /* A-part */ 370 for (k=0; k<countA; k++){ 371 for (j=0; j<bs; j++) { 372 for (n=0; n<bs; n++) { 373 if (reuse == MAT_INITIAL_MATRIX){ 374 row[jj] = irow + n + shift; 375 col[jj] = rstart + bs*ajj[k] + j + shift; 376 } 377 val[jj++] = v1[idx++]; 378 } 379 } 380 } 381 382 idx = 0; 383 /* B-part */ 384 for(k=0; k<countB; k++){ 385 for (j=0; j<bs; j++) { 386 for (n=0; n<bs; n++) { 387 if (reuse == MAT_INITIAL_MATRIX){ 388 row[jj] = irow + n + shift; 389 col[jj] = bs*garray[bjj[k]] + j + shift; 390 } 391 val[jj++] = v2[idx++]; 392 } 393 } 394 } 395 irow += bs; 396 } 397 PetscFunctionReturn(0); 398 } 399 400 #undef __FUNCT__ 401 #define __FUNCT__ "MatConvertToTriples_mpiaij_mpisbaij" 402 PetscErrorCode MatConvertToTriples_mpiaij_mpisbaij(Mat A,int shift,MatReuse reuse,int *nnz,int **r, int **c, PetscScalar **v) 403 { 404 const PetscInt *ai, *aj,*adiag, *bi, *bj,*garray,m=A->rmap->n,*ajj,*bjj; 405 PetscErrorCode ierr; 406 PetscInt rstart,nz,nza,nzb,i,j,jj,irow,countA,countB; 407 PetscInt *row,*col; 408 const PetscScalar *av, *bv,*v1,*v2; 409 PetscScalar *val; 410 Mat_MPIAIJ *mat = (Mat_MPIAIJ*)A->data; 411 Mat_SeqAIJ *aa=(Mat_SeqAIJ*)(mat->A)->data; 412 Mat_SeqAIJ *bb=(Mat_SeqAIJ*)(mat->B)->data; 413 414 PetscFunctionBegin; 415 ai=aa->i; aj=aa->j; adiag=aa->diag; 416 bi=bb->i; bj=bb->j; garray = mat->garray; 417 av=aa->a; bv=bb->a; 418 rstart = A->rmap->rstart; 419 420 if (reuse == MAT_INITIAL_MATRIX) { 421 nza = 0; /* num of upper triangular entries in mat->A, including diagonals */ 422 nzb = 0; /* num of upper triangular entries in mat->B */ 423 for(i=0; i<m; i++){ 424 nza += (ai[i+1] - adiag[i]); 425 countB = bi[i+1] - bi[i]; 426 bjj = bj + bi[i]; 427 for (j=0; j<countB; j++){ 428 if (garray[bjj[j]] > rstart) nzb++; 429 } 430 } 431 432 nz = nza + nzb; /* total nz of upper triangular part of mat */ 433 *nnz = nz; 434 ierr = PetscMalloc((2*nz*sizeof(PetscInt)+nz*sizeof(PetscScalar)), &row);CHKERRQ(ierr); 435 col = row + nz; 436 val = (PetscScalar*)(col + nz); 437 438 *r = row; *c = col; *v = val; 439 } else { 440 row = *r; col = *c; val = *v; 441 } 442 443 jj = 0; irow = rstart; 444 for ( i=0; i<m; i++ ) { 445 ajj = aj + adiag[i]; /* ptr to the beginning of the diagonal of this row */ 446 v1 = av + adiag[i]; 447 countA = ai[i+1] - adiag[i]; 448 countB = bi[i+1] - bi[i]; 449 bjj = bj + bi[i]; 450 v2 = bv + bi[i]; 451 452 /* A-part */ 453 for (j=0; j<countA; j++){ 454 if (reuse == MAT_INITIAL_MATRIX) { 455 row[jj] = irow + shift; col[jj] = rstart + ajj[j] + shift; 456 } 457 val[jj++] = v1[j]; 458 } 459 460 /* B-part */ 461 for(j=0; j < countB; j++){ 462 if (garray[bjj[j]] > rstart) { 463 if (reuse == MAT_INITIAL_MATRIX) { 464 row[jj] = irow + shift; col[jj] = garray[bjj[j]] + shift; 465 } 466 val[jj++] = v2[j]; 467 } 468 } 469 irow++; 470 } 471 PetscFunctionReturn(0); 472 } 473 474 #undef __FUNCT__ 475 #define __FUNCT__ "MatDestroy_MUMPS" 476 PetscErrorCode MatDestroy_MUMPS(Mat A) 477 { 478 Mat_MUMPS *lu=(Mat_MUMPS*)A->spptr; 479 PetscErrorCode ierr; 480 481 PetscFunctionBegin; 482 if (lu && lu->CleanUpMUMPS) { 483 /* Terminate instance, deallocate memories */ 484 ierr = PetscFree2(lu->id.sol_loc,lu->id.isol_loc);CHKERRQ(ierr); 485 ierr = VecScatterDestroy(&lu->scat_rhs);CHKERRQ(ierr); 486 ierr = VecDestroy(&lu->b_seq);CHKERRQ(ierr); 487 ierr = VecScatterDestroy(&lu->scat_sol);CHKERRQ(ierr); 488 ierr = VecDestroy(&lu->x_seq);CHKERRQ(ierr); 489 ierr=PetscFree(lu->id.perm_in);CHKERRQ(ierr); 490 ierr = PetscFree(lu->irn);CHKERRQ(ierr); 491 lu->id.job=JOB_END; 492 #if defined(PETSC_USE_COMPLEX) 493 zmumps_c(&lu->id); 494 #else 495 dmumps_c(&lu->id); 496 #endif 497 ierr = MPI_Comm_free(&(lu->comm_mumps));CHKERRQ(ierr); 498 } 499 if (lu && lu->Destroy) { 500 ierr = (lu->Destroy)(A);CHKERRQ(ierr); 501 } 502 ierr = PetscFree(A->spptr);CHKERRQ(ierr); 503 504 /* clear composed functions */ 505 ierr = PetscObjectComposeFunctionDynamic((PetscObject)A,"MatFactorGetSolverPackage_C","",PETSC_NULL);CHKERRQ(ierr); 506 ierr = PetscObjectComposeFunctionDynamic((PetscObject)A,"MatMumpsSetIcntl_C","",PETSC_NULL);CHKERRQ(ierr); 507 PetscFunctionReturn(0); 508 } 509 510 #undef __FUNCT__ 511 #define __FUNCT__ "MatSolve_MUMPS" 512 PetscErrorCode MatSolve_MUMPS(Mat A,Vec b,Vec x) 513 { 514 Mat_MUMPS *lu=(Mat_MUMPS*)A->spptr; 515 PetscScalar *array; 516 Vec b_seq; 517 IS is_iden,is_petsc; 518 PetscErrorCode ierr; 519 PetscInt i; 520 521 PetscFunctionBegin; 522 lu->id.nrhs = 1; 523 b_seq = lu->b_seq; 524 if (lu->size > 1){ 525 /* MUMPS only supports centralized rhs. Scatter b into a seqential rhs vector */ 526 ierr = VecScatterBegin(lu->scat_rhs,b,b_seq,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 527 ierr = VecScatterEnd(lu->scat_rhs,b,b_seq,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 528 if (!lu->myid) {ierr = VecGetArray(b_seq,&array);CHKERRQ(ierr);} 529 } else { /* size == 1 */ 530 ierr = VecCopy(b,x);CHKERRQ(ierr); 531 ierr = VecGetArray(x,&array);CHKERRQ(ierr); 532 } 533 if (!lu->myid) { /* define rhs on the host */ 534 lu->id.nrhs = 1; 535 #if defined(PETSC_USE_COMPLEX) 536 lu->id.rhs = (mumps_double_complex*)array; 537 #else 538 lu->id.rhs = array; 539 #endif 540 } 541 542 /* solve phase */ 543 /*-------------*/ 544 lu->id.job = JOB_SOLVE; 545 #if defined(PETSC_USE_COMPLEX) 546 zmumps_c(&lu->id); 547 #else 548 dmumps_c(&lu->id); 549 #endif 550 if (lu->id.INFOG(1) < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MUMPS in solve phase: INFOG(1)=%d\n",lu->id.INFOG(1)); 551 552 if (lu->size > 1) { /* convert mumps distributed solution to petsc mpi x */ 553 if (!lu->nSolve){ /* create scatter scat_sol */ 554 ierr = ISCreateStride(PETSC_COMM_SELF,lu->id.lsol_loc,0,1,&is_iden);CHKERRQ(ierr); /* from */ 555 for (i=0; i<lu->id.lsol_loc; i++){ 556 lu->id.isol_loc[i] -= 1; /* change Fortran style to C style */ 557 } 558 ierr = ISCreateGeneral(PETSC_COMM_SELF,lu->id.lsol_loc,lu->id.isol_loc,PETSC_COPY_VALUES,&is_petsc);CHKERRQ(ierr); /* to */ 559 ierr = VecScatterCreate(lu->x_seq,is_iden,x,is_petsc,&lu->scat_sol);CHKERRQ(ierr); 560 ierr = ISDestroy(&is_iden);CHKERRQ(ierr); 561 ierr = ISDestroy(&is_petsc);CHKERRQ(ierr); 562 } 563 ierr = VecScatterBegin(lu->scat_sol,lu->x_seq,x,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 564 ierr = VecScatterEnd(lu->scat_sol,lu->x_seq,x,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 565 } 566 lu->nSolve++; 567 PetscFunctionReturn(0); 568 } 569 570 #undef __FUNCT__ 571 #define __FUNCT__ "MatSolveTranspose_MUMPS" 572 PetscErrorCode MatSolveTranspose_MUMPS(Mat A,Vec b,Vec x) 573 { 574 Mat_MUMPS *lu=(Mat_MUMPS*)A->spptr; 575 PetscErrorCode ierr; 576 577 PetscFunctionBegin; 578 lu->id.ICNTL(9) = 0; 579 ierr = MatSolve_MUMPS(A,b,x);CHKERRQ(ierr); 580 lu->id.ICNTL(9) = 1; 581 PetscFunctionReturn(0); 582 } 583 584 #undef __FUNCT__ 585 #define __FUNCT__ "MatMatSolve_MUMPS" 586 PetscErrorCode MatMatSolve_MUMPS(Mat A,Mat B,Mat X) 587 { 588 PetscErrorCode ierr; 589 PetscBool flg; 590 591 PetscFunctionBegin; 592 ierr = PetscObjectTypeCompareAny((PetscObject)B,&flg,MATSEQDENSE,MATMPIDENSE,PETSC_NULL);CHKERRQ(ierr); 593 if (!flg) SETERRQ(((PetscObject)A)->comm,PETSC_ERR_ARG_WRONG,"Matrix B must be MATDENSE matrix"); 594 ierr = PetscObjectTypeCompareAny((PetscObject)X,&flg,MATSEQDENSE,MATMPIDENSE,PETSC_NULL);CHKERRQ(ierr); 595 if (!flg) SETERRQ(((PetscObject)A)->comm,PETSC_ERR_ARG_WRONG,"Matrix X must be MATDENSE matrix"); SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"MatMatSolve_MUMPS() is not implemented yet"); 596 PetscFunctionReturn(0); 597 } 598 599 #if !defined(PETSC_USE_COMPLEX) 600 /* 601 input: 602 F: numeric factor 603 output: 604 nneg: total number of negative pivots 605 nzero: 0 606 npos: (global dimension of F) - nneg 607 */ 608 609 #undef __FUNCT__ 610 #define __FUNCT__ "MatGetInertia_SBAIJMUMPS" 611 PetscErrorCode MatGetInertia_SBAIJMUMPS(Mat F,int *nneg,int *nzero,int *npos) 612 { 613 Mat_MUMPS *lu =(Mat_MUMPS*)F->spptr; 614 PetscErrorCode ierr; 615 PetscMPIInt size; 616 617 PetscFunctionBegin; 618 ierr = MPI_Comm_size(((PetscObject)F)->comm,&size);CHKERRQ(ierr); 619 /* MUMPS 4.3.1 calls ScaLAPACK when ICNTL(13)=0 (default), which does not offer the possibility to compute the inertia of a dense matrix. Set ICNTL(13)=1 to skip ScaLAPACK */ 620 if (size > 1 && lu->id.ICNTL(13) != 1) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"ICNTL(13)=%d. -mat_mumps_icntl_13 must be set as 1 for correct global matrix inertia\n",lu->id.INFOG(13)); 621 if (nneg){ 622 if (!lu->myid){ 623 *nneg = lu->id.INFOG(12); 624 } 625 ierr = MPI_Bcast(nneg,1,MPI_INT,0,lu->comm_mumps);CHKERRQ(ierr); 626 } 627 if (nzero) *nzero = 0; 628 if (npos) *npos = F->rmap->N - (*nneg); 629 PetscFunctionReturn(0); 630 } 631 #endif /* !defined(PETSC_USE_COMPLEX) */ 632 633 #undef __FUNCT__ 634 #define __FUNCT__ "MatFactorNumeric_MUMPS" 635 PetscErrorCode MatFactorNumeric_MUMPS(Mat F,Mat A,const MatFactorInfo *info) 636 { 637 Mat_MUMPS *lu =(Mat_MUMPS*)(F)->spptr; 638 PetscErrorCode ierr; 639 Mat F_diag; 640 PetscBool isMPIAIJ; 641 642 PetscFunctionBegin; 643 ierr = (*lu->ConvertToTriples)(A, 1, MAT_REUSE_MATRIX, &lu->nz, &lu->irn, &lu->jcn, &lu->val);CHKERRQ(ierr); 644 645 /* numerical factorization phase */ 646 /*-------------------------------*/ 647 lu->id.job = JOB_FACTNUMERIC; 648 if(!lu->id.ICNTL(18)) { 649 if (!lu->myid) { 650 #if defined(PETSC_USE_COMPLEX) 651 lu->id.a = (mumps_double_complex*)lu->val; 652 #else 653 lu->id.a = lu->val; 654 #endif 655 } 656 } else { 657 #if defined(PETSC_USE_COMPLEX) 658 lu->id.a_loc = (mumps_double_complex*)lu->val; 659 #else 660 lu->id.a_loc = lu->val; 661 #endif 662 } 663 #if defined(PETSC_USE_COMPLEX) 664 zmumps_c(&lu->id); 665 #else 666 dmumps_c(&lu->id); 667 #endif 668 if (lu->id.INFOG(1) < 0) { 669 if (lu->id.INFO(1) == -13) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MUMPS in numerical factorization phase: Cannot allocate required memory %d megabytes\n",lu->id.INFO(2)); 670 else SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MUMPS in numerical factorization phase: INFO(1)=%d, INFO(2)=%d\n",lu->id.INFO(1),lu->id.INFO(2)); 671 } 672 if (!lu->myid && lu->id.ICNTL(16) > 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB," lu->id.ICNTL(16):=%d\n",lu->id.INFOG(16)); 673 674 if (lu->size > 1){ 675 ierr = PetscObjectTypeCompare((PetscObject)A,MATMPIAIJ,&isMPIAIJ);CHKERRQ(ierr); 676 if(isMPIAIJ) { 677 F_diag = ((Mat_MPIAIJ *)(F)->data)->A; 678 } else { 679 F_diag = ((Mat_MPISBAIJ *)(F)->data)->A; 680 } 681 F_diag->assembled = PETSC_TRUE; 682 if (lu->nSolve){ 683 ierr = VecScatterDestroy(&lu->scat_sol);CHKERRQ(ierr); 684 ierr = PetscFree2(lu->id.sol_loc,lu->id.isol_loc);CHKERRQ(ierr); 685 ierr = VecDestroy(&lu->x_seq);CHKERRQ(ierr); 686 } 687 } 688 (F)->assembled = PETSC_TRUE; 689 lu->matstruc = SAME_NONZERO_PATTERN; 690 lu->CleanUpMUMPS = PETSC_TRUE; 691 lu->nSolve = 0; 692 693 if (lu->size > 1){ 694 /* distributed solution */ 695 if (!lu->nSolve){ 696 /* Create x_seq=sol_loc for repeated use */ 697 PetscInt lsol_loc; 698 PetscScalar *sol_loc; 699 lsol_loc = lu->id.INFO(23); /* length of sol_loc */ 700 ierr = PetscMalloc2(lsol_loc,PetscScalar,&sol_loc,lsol_loc,PetscInt,&lu->id.isol_loc);CHKERRQ(ierr); 701 lu->id.lsol_loc = lsol_loc; 702 #if defined(PETSC_USE_COMPLEX) 703 lu->id.sol_loc = (mumps_double_complex*)sol_loc; 704 #else 705 lu->id.sol_loc = sol_loc; 706 #endif 707 ierr = VecCreateSeqWithArray(PETSC_COMM_SELF,1,lsol_loc,sol_loc,&lu->x_seq);CHKERRQ(ierr); 708 } 709 } 710 PetscFunctionReturn(0); 711 } 712 713 /* Sets MUMPS options from the options database */ 714 #undef __FUNCT__ 715 #define __FUNCT__ "PetscSetMUMPSFromOptions" 716 PetscErrorCode PetscSetMUMPSFromOptions(Mat F, Mat A) 717 { 718 Mat_MUMPS *mumps = (Mat_MUMPS*)F->spptr; 719 PetscErrorCode ierr; 720 PetscInt icntl; 721 PetscBool flg; 722 723 PetscFunctionBegin; 724 ierr = PetscOptionsBegin(((PetscObject)A)->comm,((PetscObject)A)->prefix,"MUMPS Options","Mat");CHKERRQ(ierr); 725 ierr = PetscOptionsInt("-mat_mumps_icntl_1","ICNTL(1): output stream for error messages","None",mumps->id.ICNTL(1),&icntl,&flg);CHKERRQ(ierr); 726 if (flg) mumps->id.ICNTL(1) = icntl; 727 ierr = PetscOptionsInt("-mat_mumps_icntl_2","ICNTL(2): output stream for diagnostic printing, statistics, and warning","None",mumps->id.ICNTL(2),&icntl,&flg);CHKERRQ(ierr); 728 if (flg) mumps->id.ICNTL(2) = icntl; 729 ierr = PetscOptionsInt("-mat_mumps_icntl_3","ICNTL(3): output stream for global information, collected on the host","None",mumps->id.ICNTL(3),&icntl,&flg);CHKERRQ(ierr); 730 if (flg) mumps->id.ICNTL(3) = icntl; 731 732 ierr = PetscOptionsInt("-mat_mumps_icntl_4","ICNTL(4): level of printing (0 to 4)","None",mumps->id.ICNTL(4),&icntl,&flg);CHKERRQ(ierr); 733 if (flg) mumps->id.ICNTL(4) = icntl; 734 if (mumps->id.ICNTL(4) || PetscLogPrintInfo ) mumps->id.ICNTL(3) = 6; /* resume MUMPS default id.ICNTL(3) = 6 */ 735 736 ierr = PetscOptionsInt("-mat_mumps_icntl_6","ICNTL(6): permuting and/or scaling the matrix (0 to 7)","None",mumps->id.ICNTL(6),&icntl,&flg);CHKERRQ(ierr); 737 if (flg) mumps->id.ICNTL(6) = icntl; 738 739 ierr = PetscOptionsInt("-mat_mumps_icntl_7","ICNTL(7): matrix ordering (0 to 7). 3=Scotch, 4=PORD, 5=Metis","None",mumps->id.ICNTL(7),&icntl,&flg);CHKERRQ(ierr); 740 if (flg) { 741 if (icntl== 1 && mumps->size > 1){ 742 SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"pivot order be set by the user in PERM_IN -- not supported by the PETSc/MUMPS interface\n"); 743 } else { 744 mumps->id.ICNTL(7) = icntl; 745 } 746 } 747 748 ierr = PetscOptionsInt("-mat_mumps_icntl_8","ICNTL(8): scaling strategy (-2 to 8 or 77)","None",mumps->id.ICNTL(8),&mumps->id.ICNTL(8),PETSC_NULL);CHKERRQ(ierr); 749 ierr = PetscOptionsInt("-mat_mumps_icntl_10","ICNTL(10): max num of refinements","None",mumps->id.ICNTL(10),&mumps->id.ICNTL(10),PETSC_NULL);CHKERRQ(ierr); 750 ierr = PetscOptionsInt("-mat_mumps_icntl_11","ICNTL(11): statistics related to the linear system solved (via -ksp_view)","None",mumps->id.ICNTL(11),&mumps->id.ICNTL(11),PETSC_NULL);CHKERRQ(ierr); 751 ierr = PetscOptionsInt("-mat_mumps_icntl_12","ICNTL(12): efficiency control: defines the ordering strategy with scaling constraints (0 to 3)","None",mumps->id.ICNTL(12),&mumps->id.ICNTL(12),PETSC_NULL);CHKERRQ(ierr); 752 ierr = PetscOptionsInt("-mat_mumps_icntl_13","ICNTL(13): efficiency control: with or without ScaLAPACK","None",mumps->id.ICNTL(13),&mumps->id.ICNTL(13),PETSC_NULL);CHKERRQ(ierr); 753 ierr = PetscOptionsInt("-mat_mumps_icntl_14","ICNTL(14): percentage of estimated workspace increase","None",mumps->id.ICNTL(14),&mumps->id.ICNTL(14),PETSC_NULL);CHKERRQ(ierr); 754 ierr = PetscOptionsInt("-mat_mumps_icntl_19","ICNTL(19): Schur complement","None",mumps->id.ICNTL(19),&mumps->id.ICNTL(19),PETSC_NULL);CHKERRQ(ierr); 755 756 ierr = PetscOptionsInt("-mat_mumps_icntl_22","ICNTL(22): in-core/out-of-core facility (0 or 1)","None",mumps->id.ICNTL(22),&mumps->id.ICNTL(22),PETSC_NULL);CHKERRQ(ierr); 757 ierr = PetscOptionsInt("-mat_mumps_icntl_23","ICNTL(23): max size of the working memory (MB) that can allocate per processor","None",mumps->id.ICNTL(23),&mumps->id.ICNTL(23),PETSC_NULL);CHKERRQ(ierr); 758 ierr = PetscOptionsInt("-mat_mumps_icntl_24","ICNTL(24): detection of null pivot rows (0 or 1)","None",mumps->id.ICNTL(24),&mumps->id.ICNTL(24),PETSC_NULL);CHKERRQ(ierr); 759 if (mumps->id.ICNTL(24)){ 760 mumps->id.ICNTL(13) = 1; /* turn-off ScaLAPACK to help with the correct detection of null pivots */ 761 } 762 763 ierr = PetscOptionsInt("-mat_mumps_icntl_25","ICNTL(25): computation of a null space basis","None",mumps->id.ICNTL(25),&mumps->id.ICNTL(25),PETSC_NULL);CHKERRQ(ierr); 764 ierr = PetscOptionsInt("-mat_mumps_icntl_26","ICNTL(26): Schur options for right-hand side or solution vector","None",mumps->id.ICNTL(26),&mumps->id.ICNTL(26),PETSC_NULL);CHKERRQ(ierr); 765 ierr = PetscOptionsInt("-mat_mumps_icntl_27","ICNTL(27): experimental parameter","None",mumps->id.ICNTL(27),&mumps->id.ICNTL(27),PETSC_NULL);CHKERRQ(ierr); 766 ierr = PetscOptionsInt("-mat_mumps_icntl_28","ICNTL(28): use 1 for sequential analysis and ictnl(7) ordering, or 2 for parallel analysis and ictnl(29) ordering","None",mumps->id.ICNTL(28),&mumps->id.ICNTL(28),PETSC_NULL);CHKERRQ(ierr); 767 ierr = PetscOptionsInt("-mat_mumps_icntl_29","ICNTL(29): parallel ordering 1 = ptscotch 2 = parmetis","None",mumps->id.ICNTL(29),&mumps->id.ICNTL(29),PETSC_NULL);CHKERRQ(ierr); 768 ierr = PetscOptionsInt("-mat_mumps_icntl_30","ICNTL(30): compute user-specified set of entries in inv(A)","None",mumps->id.ICNTL(30),&mumps->id.ICNTL(30),PETSC_NULL);CHKERRQ(ierr); 769 ierr = PetscOptionsInt("-mat_mumps_icntl_31","ICNTL(31): factors can be discarded in the solve phase","None",mumps->id.ICNTL(31),&mumps->id.ICNTL(31),PETSC_NULL);CHKERRQ(ierr); 770 ierr = PetscOptionsInt("-mat_mumps_icntl_33","ICNTL(33): compute determinant","None",mumps->id.ICNTL(33),&mumps->id.ICNTL(33),PETSC_NULL);CHKERRQ(ierr); 771 772 ierr = PetscOptionsReal("-mat_mumps_cntl_1","CNTL(1): relative pivoting threshold","None",mumps->id.CNTL(1),&mumps->id.CNTL(1),PETSC_NULL);CHKERRQ(ierr); 773 ierr = PetscOptionsReal("-mat_mumps_cntl_2","CNTL(2): stopping criterion of refinement","None",mumps->id.CNTL(2),&mumps->id.CNTL(2),PETSC_NULL);CHKERRQ(ierr); 774 ierr = PetscOptionsReal("-mat_mumps_cntl_3","CNTL(3): absolute pivoting threshold","None",mumps->id.CNTL(3),&mumps->id.CNTL(3),PETSC_NULL);CHKERRQ(ierr); 775 ierr = PetscOptionsReal("-mat_mumps_cntl_4","CNTL(4): value for static pivoting","None",mumps->id.CNTL(4),&mumps->id.CNTL(4),PETSC_NULL);CHKERRQ(ierr); 776 ierr = PetscOptionsReal("-mat_mumps_cntl_5","CNTL(5): fixation for null pivots","None",mumps->id.CNTL(5),&mumps->id.CNTL(5),PETSC_NULL);CHKERRQ(ierr); 777 778 ierr = PetscOptionsString("-mat_mumps_ooc_tmpdir", "out of core directory", "None", mumps->id.ooc_tmpdir, mumps->id.ooc_tmpdir, 256, PETSC_NULL); 779 PetscOptionsEnd(); 780 PetscFunctionReturn(0); 781 } 782 783 #undef __FUNCT__ 784 #define __FUNCT__ "PetscInitializeMUMPS" 785 PetscErrorCode PetscInitializeMUMPS(Mat A,Mat_MUMPS* mumps) 786 { 787 PetscErrorCode ierr; 788 789 PetscFunctionBegin; 790 ierr = MPI_Comm_rank(((PetscObject)A)->comm, &mumps->myid); 791 ierr = MPI_Comm_size(((PetscObject)A)->comm,&mumps->size);CHKERRQ(ierr); 792 ierr = MPI_Comm_dup(((PetscObject)A)->comm,&(mumps->comm_mumps));CHKERRQ(ierr); 793 mumps->id.comm_fortran = MPI_Comm_c2f(mumps->comm_mumps); 794 795 mumps->id.job = JOB_INIT; 796 mumps->id.par = 1; /* host participates factorizaton and solve */ 797 mumps->id.sym = mumps->sym; 798 #if defined(PETSC_USE_COMPLEX) 799 zmumps_c(&mumps->id); 800 #else 801 dmumps_c(&mumps->id); 802 #endif 803 804 mumps->CleanUpMUMPS = PETSC_FALSE; 805 mumps->scat_rhs = PETSC_NULL; 806 mumps->scat_sol = PETSC_NULL; 807 mumps->nSolve = 0; 808 809 /* set PETSc-MUMPS default options - override MUMPS default */ 810 mumps->id.ICNTL(3) = 0; 811 mumps->id.ICNTL(4) = 0; 812 if (mumps->size == 1){ 813 mumps->id.ICNTL(18) = 0; /* centralized assembled matrix input */ 814 } else { 815 mumps->id.ICNTL(18) = 3; /* distributed assembled matrix input */ 816 mumps->id.ICNTL(21) = 1; /* distributed solution */ 817 } 818 PetscFunctionReturn(0); 819 } 820 821 /* Note the Petsc r and c permutations are ignored */ 822 #undef __FUNCT__ 823 #define __FUNCT__ "MatLUFactorSymbolic_AIJMUMPS" 824 PetscErrorCode MatLUFactorSymbolic_AIJMUMPS(Mat F,Mat A,IS r,IS c,const MatFactorInfo *info) 825 { 826 Mat_MUMPS *lu = (Mat_MUMPS*)F->spptr; 827 PetscErrorCode ierr; 828 Vec b; 829 IS is_iden; 830 const PetscInt M = A->rmap->N; 831 832 PetscFunctionBegin; 833 lu->matstruc = DIFFERENT_NONZERO_PATTERN; 834 835 /* Set MUMPS options from the options database */ 836 ierr = PetscSetMUMPSFromOptions(F,A);CHKERRQ(ierr); 837 838 ierr = (*lu->ConvertToTriples)(A, 1, MAT_INITIAL_MATRIX, &lu->nz, &lu->irn, &lu->jcn, &lu->val);CHKERRQ(ierr); 839 840 /* analysis phase */ 841 /*----------------*/ 842 lu->id.job = JOB_FACTSYMBOLIC; 843 lu->id.n = M; 844 switch (lu->id.ICNTL(18)){ 845 case 0: /* centralized assembled matrix input */ 846 if (!lu->myid) { 847 lu->id.nz =lu->nz; lu->id.irn=lu->irn; lu->id.jcn=lu->jcn; 848 if (lu->id.ICNTL(6)>1){ 849 #if defined(PETSC_USE_COMPLEX) 850 lu->id.a = (mumps_double_complex*)lu->val; 851 #else 852 lu->id.a = lu->val; 853 #endif 854 } 855 if (lu->id.ICNTL(7) == 1){ /* use user-provide matrix ordering */ 856 if (!lu->myid) { 857 const PetscInt *idx; 858 PetscInt i,*perm_in; 859 ierr = PetscMalloc(M*sizeof(PetscInt),&perm_in);CHKERRQ(ierr); 860 ierr = ISGetIndices(r,&idx);CHKERRQ(ierr); 861 lu->id.perm_in = perm_in; 862 for (i=0; i<M; i++) perm_in[i] = idx[i]+1; /* perm_in[]: start from 1, not 0! */ 863 ierr = ISRestoreIndices(r,&idx);CHKERRQ(ierr); 864 } 865 } 866 } 867 break; 868 case 3: /* distributed assembled matrix input (size>1) */ 869 lu->id.nz_loc = lu->nz; 870 lu->id.irn_loc=lu->irn; lu->id.jcn_loc=lu->jcn; 871 if (lu->id.ICNTL(6)>1) { 872 #if defined(PETSC_USE_COMPLEX) 873 lu->id.a_loc = (mumps_double_complex*)lu->val; 874 #else 875 lu->id.a_loc = lu->val; 876 #endif 877 } 878 /* MUMPS only supports centralized rhs. Create scatter scat_rhs for repeated use in MatSolve() */ 879 if (!lu->myid){ 880 ierr = VecCreateSeq(PETSC_COMM_SELF,A->cmap->N,&lu->b_seq);CHKERRQ(ierr); 881 ierr = ISCreateStride(PETSC_COMM_SELF,A->cmap->N,0,1,&is_iden);CHKERRQ(ierr); 882 } else { 883 ierr = VecCreateSeq(PETSC_COMM_SELF,0,&lu->b_seq);CHKERRQ(ierr); 884 ierr = ISCreateStride(PETSC_COMM_SELF,0,0,1,&is_iden);CHKERRQ(ierr); 885 } 886 ierr = VecCreate(((PetscObject)A)->comm,&b);CHKERRQ(ierr); 887 ierr = VecSetSizes(b,A->rmap->n,PETSC_DECIDE);CHKERRQ(ierr); 888 ierr = VecSetFromOptions(b);CHKERRQ(ierr); 889 890 ierr = VecScatterCreate(b,is_iden,lu->b_seq,is_iden,&lu->scat_rhs);CHKERRQ(ierr); 891 ierr = ISDestroy(&is_iden);CHKERRQ(ierr); 892 ierr = VecDestroy(&b);CHKERRQ(ierr); 893 break; 894 } 895 #if defined(PETSC_USE_COMPLEX) 896 zmumps_c(&lu->id); 897 #else 898 dmumps_c(&lu->id); 899 #endif 900 if (lu->id.INFOG(1) < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MUMPS in analysis phase: INFOG(1)=%d\n",lu->id.INFOG(1)); 901 902 F->ops->lufactornumeric = MatFactorNumeric_MUMPS; 903 F->ops->solve = MatSolve_MUMPS; 904 F->ops->solvetranspose = MatSolveTranspose_MUMPS; 905 F->ops->matsolve = MatMatSolve_MUMPS; 906 PetscFunctionReturn(0); 907 } 908 909 /* Note the Petsc r and c permutations are ignored */ 910 #undef __FUNCT__ 911 #define __FUNCT__ "MatLUFactorSymbolic_BAIJMUMPS" 912 PetscErrorCode MatLUFactorSymbolic_BAIJMUMPS(Mat F,Mat A,IS r,IS c,const MatFactorInfo *info) 913 { 914 915 Mat_MUMPS *lu = (Mat_MUMPS*)F->spptr; 916 PetscErrorCode ierr; 917 Vec b; 918 IS is_iden; 919 const PetscInt M = A->rmap->N; 920 921 PetscFunctionBegin; 922 lu->matstruc = DIFFERENT_NONZERO_PATTERN; 923 924 /* Set MUMPS options from the options database */ 925 ierr = PetscSetMUMPSFromOptions(F,A);CHKERRQ(ierr); 926 927 ierr = (*lu->ConvertToTriples)(A, 1, MAT_INITIAL_MATRIX, &lu->nz, &lu->irn, &lu->jcn, &lu->val);CHKERRQ(ierr); 928 929 /* analysis phase */ 930 /*----------------*/ 931 lu->id.job = JOB_FACTSYMBOLIC; 932 lu->id.n = M; 933 switch (lu->id.ICNTL(18)){ 934 case 0: /* centralized assembled matrix input */ 935 if (!lu->myid) { 936 lu->id.nz =lu->nz; lu->id.irn=lu->irn; lu->id.jcn=lu->jcn; 937 if (lu->id.ICNTL(6)>1){ 938 #if defined(PETSC_USE_COMPLEX) 939 lu->id.a = (mumps_double_complex*)lu->val; 940 #else 941 lu->id.a = lu->val; 942 #endif 943 } 944 } 945 break; 946 case 3: /* distributed assembled matrix input (size>1) */ 947 lu->id.nz_loc = lu->nz; 948 lu->id.irn_loc=lu->irn; lu->id.jcn_loc=lu->jcn; 949 if (lu->id.ICNTL(6)>1) { 950 #if defined(PETSC_USE_COMPLEX) 951 lu->id.a_loc = (mumps_double_complex*)lu->val; 952 #else 953 lu->id.a_loc = lu->val; 954 #endif 955 } 956 /* MUMPS only supports centralized rhs. Create scatter scat_rhs for repeated use in MatSolve() */ 957 if (!lu->myid){ 958 ierr = VecCreateSeq(PETSC_COMM_SELF,A->cmap->N,&lu->b_seq);CHKERRQ(ierr); 959 ierr = ISCreateStride(PETSC_COMM_SELF,A->cmap->N,0,1,&is_iden);CHKERRQ(ierr); 960 } else { 961 ierr = VecCreateSeq(PETSC_COMM_SELF,0,&lu->b_seq);CHKERRQ(ierr); 962 ierr = ISCreateStride(PETSC_COMM_SELF,0,0,1,&is_iden);CHKERRQ(ierr); 963 } 964 ierr = VecCreate(((PetscObject)A)->comm,&b);CHKERRQ(ierr); 965 ierr = VecSetSizes(b,A->rmap->n,PETSC_DECIDE);CHKERRQ(ierr); 966 ierr = VecSetFromOptions(b);CHKERRQ(ierr); 967 968 ierr = VecScatterCreate(b,is_iden,lu->b_seq,is_iden,&lu->scat_rhs);CHKERRQ(ierr); 969 ierr = ISDestroy(&is_iden);CHKERRQ(ierr); 970 ierr = VecDestroy(&b);CHKERRQ(ierr); 971 break; 972 } 973 #if defined(PETSC_USE_COMPLEX) 974 zmumps_c(&lu->id); 975 #else 976 dmumps_c(&lu->id); 977 #endif 978 if (lu->id.INFOG(1) < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MUMPS in analysis phase: INFOG(1)=%d\n",lu->id.INFOG(1)); 979 980 F->ops->lufactornumeric = MatFactorNumeric_MUMPS; 981 F->ops->solve = MatSolve_MUMPS; 982 F->ops->solvetranspose = MatSolveTranspose_MUMPS; 983 PetscFunctionReturn(0); 984 } 985 986 /* Note the Petsc r permutation and factor info are ignored */ 987 #undef __FUNCT__ 988 #define __FUNCT__ "MatCholeskyFactorSymbolic_MUMPS" 989 PetscErrorCode MatCholeskyFactorSymbolic_MUMPS(Mat F,Mat A,IS r,const MatFactorInfo *info) 990 { 991 Mat_MUMPS *lu = (Mat_MUMPS*)F->spptr; 992 PetscErrorCode ierr; 993 Vec b; 994 IS is_iden; 995 const PetscInt M = A->rmap->N; 996 997 PetscFunctionBegin; 998 lu->matstruc = DIFFERENT_NONZERO_PATTERN; 999 1000 /* Set MUMPS options from the options database */ 1001 ierr = PetscSetMUMPSFromOptions(F,A);CHKERRQ(ierr); 1002 1003 ierr = (*lu->ConvertToTriples)(A, 1 , MAT_INITIAL_MATRIX, &lu->nz, &lu->irn, &lu->jcn, &lu->val);CHKERRQ(ierr); 1004 1005 /* analysis phase */ 1006 /*----------------*/ 1007 lu->id.job = JOB_FACTSYMBOLIC; 1008 lu->id.n = M; 1009 switch (lu->id.ICNTL(18)){ 1010 case 0: /* centralized assembled matrix input */ 1011 if (!lu->myid) { 1012 lu->id.nz =lu->nz; lu->id.irn=lu->irn; lu->id.jcn=lu->jcn; 1013 if (lu->id.ICNTL(6)>1){ 1014 #if defined(PETSC_USE_COMPLEX) 1015 lu->id.a = (mumps_double_complex*)lu->val; 1016 #else 1017 lu->id.a = lu->val; 1018 #endif 1019 } 1020 } 1021 break; 1022 case 3: /* distributed assembled matrix input (size>1) */ 1023 lu->id.nz_loc = lu->nz; 1024 lu->id.irn_loc=lu->irn; lu->id.jcn_loc=lu->jcn; 1025 if (lu->id.ICNTL(6)>1) { 1026 #if defined(PETSC_USE_COMPLEX) 1027 lu->id.a_loc = (mumps_double_complex*)lu->val; 1028 #else 1029 lu->id.a_loc = lu->val; 1030 #endif 1031 } 1032 /* MUMPS only supports centralized rhs. Create scatter scat_rhs for repeated use in MatSolve() */ 1033 if (!lu->myid){ 1034 ierr = VecCreateSeq(PETSC_COMM_SELF,A->cmap->N,&lu->b_seq);CHKERRQ(ierr); 1035 ierr = ISCreateStride(PETSC_COMM_SELF,A->cmap->N,0,1,&is_iden);CHKERRQ(ierr); 1036 } else { 1037 ierr = VecCreateSeq(PETSC_COMM_SELF,0,&lu->b_seq);CHKERRQ(ierr); 1038 ierr = ISCreateStride(PETSC_COMM_SELF,0,0,1,&is_iden);CHKERRQ(ierr); 1039 } 1040 ierr = VecCreate(((PetscObject)A)->comm,&b);CHKERRQ(ierr); 1041 ierr = VecSetSizes(b,A->rmap->n,PETSC_DECIDE);CHKERRQ(ierr); 1042 ierr = VecSetFromOptions(b);CHKERRQ(ierr); 1043 1044 ierr = VecScatterCreate(b,is_iden,lu->b_seq,is_iden,&lu->scat_rhs);CHKERRQ(ierr); 1045 ierr = ISDestroy(&is_iden);CHKERRQ(ierr); 1046 ierr = VecDestroy(&b);CHKERRQ(ierr); 1047 break; 1048 } 1049 #if defined(PETSC_USE_COMPLEX) 1050 zmumps_c(&lu->id); 1051 #else 1052 dmumps_c(&lu->id); 1053 #endif 1054 if (lu->id.INFOG(1) < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MUMPS in analysis phase: INFOG(1)=%d\n",lu->id.INFOG(1)); 1055 1056 F->ops->choleskyfactornumeric = MatFactorNumeric_MUMPS; 1057 F->ops->solve = MatSolve_MUMPS; 1058 F->ops->solvetranspose = MatSolve_MUMPS; 1059 #if !defined(PETSC_USE_COMPLEX) 1060 F->ops->getinertia = MatGetInertia_SBAIJMUMPS; 1061 #else 1062 F->ops->getinertia = PETSC_NULL; 1063 #endif 1064 PetscFunctionReturn(0); 1065 } 1066 1067 #undef __FUNCT__ 1068 #define __FUNCT__ "MatView_MUMPS" 1069 PetscErrorCode MatView_MUMPS(Mat A,PetscViewer viewer) 1070 { 1071 PetscErrorCode ierr; 1072 PetscBool iascii; 1073 PetscViewerFormat format; 1074 Mat_MUMPS *lu=(Mat_MUMPS*)A->spptr; 1075 1076 PetscFunctionBegin; 1077 /* check if matrix is mumps type */ 1078 if (A->ops->solve != MatSolve_MUMPS) PetscFunctionReturn(0); 1079 1080 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);CHKERRQ(ierr); 1081 if (iascii) { 1082 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 1083 if (format == PETSC_VIEWER_ASCII_INFO){ 1084 ierr = PetscViewerASCIIPrintf(viewer,"MUMPS run parameters:\n");CHKERRQ(ierr); 1085 ierr = PetscViewerASCIIPrintf(viewer," SYM (matrix type): %d \n",lu->id.sym);CHKERRQ(ierr); 1086 ierr = PetscViewerASCIIPrintf(viewer," PAR (host participation): %d \n",lu->id.par);CHKERRQ(ierr); 1087 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(1) (output for error): %d \n",lu->id.ICNTL(1));CHKERRQ(ierr); 1088 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(2) (output of diagnostic msg): %d \n",lu->id.ICNTL(2));CHKERRQ(ierr); 1089 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(3) (output for global info): %d \n",lu->id.ICNTL(3));CHKERRQ(ierr); 1090 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(4) (level of printing): %d \n",lu->id.ICNTL(4));CHKERRQ(ierr); 1091 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(5) (input mat struct): %d \n",lu->id.ICNTL(5));CHKERRQ(ierr); 1092 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(6) (matrix prescaling): %d \n",lu->id.ICNTL(6));CHKERRQ(ierr); 1093 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(7) (sequentia matrix ordering):%d \n",lu->id.ICNTL(7));CHKERRQ(ierr); 1094 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(8) (scalling strategy): %d \n",lu->id.ICNTL(8));CHKERRQ(ierr); 1095 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(10) (max num of refinements): %d \n",lu->id.ICNTL(10));CHKERRQ(ierr); 1096 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(11) (error analysis): %d \n",lu->id.ICNTL(11));CHKERRQ(ierr); 1097 if (lu->id.ICNTL(11)>0) { 1098 ierr = PetscViewerASCIIPrintf(viewer," RINFOG(4) (inf norm of input mat): %g\n",lu->id.RINFOG(4));CHKERRQ(ierr); 1099 ierr = PetscViewerASCIIPrintf(viewer," RINFOG(5) (inf norm of solution): %g\n",lu->id.RINFOG(5));CHKERRQ(ierr); 1100 ierr = PetscViewerASCIIPrintf(viewer," RINFOG(6) (inf norm of residual): %g\n",lu->id.RINFOG(6));CHKERRQ(ierr); 1101 ierr = PetscViewerASCIIPrintf(viewer," RINFOG(7),RINFOG(8) (backward error est): %g, %g\n",lu->id.RINFOG(7),lu->id.RINFOG(8));CHKERRQ(ierr); 1102 ierr = PetscViewerASCIIPrintf(viewer," RINFOG(9) (error estimate): %g \n",lu->id.RINFOG(9));CHKERRQ(ierr); 1103 ierr = PetscViewerASCIIPrintf(viewer," RINFOG(10),RINFOG(11)(condition numbers): %g, %g\n",lu->id.RINFOG(10),lu->id.RINFOG(11));CHKERRQ(ierr); 1104 } 1105 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(12) (efficiency control): %d \n",lu->id.ICNTL(12));CHKERRQ(ierr); 1106 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(13) (efficiency control): %d \n",lu->id.ICNTL(13));CHKERRQ(ierr); 1107 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(14) (percentage of estimated workspace increase): %d \n",lu->id.ICNTL(14));CHKERRQ(ierr); 1108 /* ICNTL(15-17) not used */ 1109 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(18) (input mat struct): %d \n",lu->id.ICNTL(18));CHKERRQ(ierr); 1110 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(19) (Shur complement info): %d \n",lu->id.ICNTL(19));CHKERRQ(ierr); 1111 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(20) (rhs sparse pattern): %d \n",lu->id.ICNTL(20));CHKERRQ(ierr); 1112 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(21) (solution struct): %d \n",lu->id.ICNTL(21));CHKERRQ(ierr); 1113 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(22) (in-core/out-of-core facility): %d \n",lu->id.ICNTL(22));CHKERRQ(ierr); 1114 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(23) (max size of memory can be allocated locally):%d \n",lu->id.ICNTL(23));CHKERRQ(ierr); 1115 1116 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(24) (detection of null pivot rows): %d \n",lu->id.ICNTL(24));CHKERRQ(ierr); 1117 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(25) (computation of a null space basis): %d \n",lu->id.ICNTL(25));CHKERRQ(ierr); 1118 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(26) (Schur options for rhs or solution): %d \n",lu->id.ICNTL(26));CHKERRQ(ierr); 1119 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(27) (experimental parameter): %d \n",lu->id.ICNTL(27));CHKERRQ(ierr); 1120 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(28) (use parallel or sequential ordering): %d \n",lu->id.ICNTL(28));CHKERRQ(ierr); 1121 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(29) (parallel ordering): %d \n",lu->id.ICNTL(29));CHKERRQ(ierr); 1122 1123 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(30) (user-specified set of entries in inv(A)): %d \n",lu->id.ICNTL(30));CHKERRQ(ierr); 1124 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(31) (factors is discarded in the solve phase): %d \n",lu->id.ICNTL(31));CHKERRQ(ierr); 1125 ierr = PetscViewerASCIIPrintf(viewer," ICNTL(33) (compute determinant): %d \n",lu->id.ICNTL(33));CHKERRQ(ierr); 1126 1127 ierr = PetscViewerASCIIPrintf(viewer," CNTL(1) (relative pivoting threshold): %g \n",lu->id.CNTL(1));CHKERRQ(ierr); 1128 ierr = PetscViewerASCIIPrintf(viewer," CNTL(2) (stopping criterion of refinement): %g \n",lu->id.CNTL(2));CHKERRQ(ierr); 1129 ierr = PetscViewerASCIIPrintf(viewer," CNTL(3) (absolute pivoting threshold): %g \n",lu->id.CNTL(3));CHKERRQ(ierr); 1130 ierr = PetscViewerASCIIPrintf(viewer," CNTL(4) (value of static pivoting): %g \n",lu->id.CNTL(4));CHKERRQ(ierr); 1131 ierr = PetscViewerASCIIPrintf(viewer," CNTL(5) (fixation for null pivots): %g \n",lu->id.CNTL(5));CHKERRQ(ierr); 1132 1133 /* infomation local to each processor */ 1134 ierr = PetscViewerASCIIPrintf(viewer, " RINFO(1) (local estimated flops for the elimination after analysis): \n");CHKERRQ(ierr); 1135 ierr = PetscViewerASCIISynchronizedAllow(viewer,PETSC_TRUE);CHKERRQ(ierr); 1136 ierr = PetscViewerASCIISynchronizedPrintf(viewer," [%d] %g \n",lu->myid,lu->id.RINFO(1));CHKERRQ(ierr); 1137 ierr = PetscViewerFlush(viewer); 1138 ierr = PetscViewerASCIIPrintf(viewer, " RINFO(2) (local estimated flops for the assembly after factorization): \n");CHKERRQ(ierr); 1139 ierr = PetscViewerASCIISynchronizedPrintf(viewer," [%d] %g \n",lu->myid,lu->id.RINFO(2));CHKERRQ(ierr); 1140 ierr = PetscViewerFlush(viewer); 1141 ierr = PetscViewerASCIIPrintf(viewer, " RINFO(3) (local estimated flops for the elimination after factorization): \n");CHKERRQ(ierr); 1142 ierr = PetscViewerASCIISynchronizedPrintf(viewer," [%d] %g \n",lu->myid,lu->id.RINFO(3));CHKERRQ(ierr); 1143 ierr = PetscViewerFlush(viewer); 1144 1145 ierr = PetscViewerASCIIPrintf(viewer, " INFO(15) (estimated size of (in MB) MUMPS internal data for running numerical factorization): \n");CHKERRQ(ierr); 1146 ierr = PetscViewerASCIISynchronizedPrintf(viewer," [%d] %d \n",lu->myid,lu->id.INFO(15));CHKERRQ(ierr); 1147 ierr = PetscViewerFlush(viewer); 1148 1149 ierr = PetscViewerASCIIPrintf(viewer, " INFO(16) (size of (in MB) MUMPS internal data used during numerical factorization): \n");CHKERRQ(ierr); 1150 ierr = PetscViewerASCIISynchronizedPrintf(viewer," [%d] %d \n",lu->myid,lu->id.INFO(16));CHKERRQ(ierr); 1151 ierr = PetscViewerFlush(viewer); 1152 1153 ierr = PetscViewerASCIIPrintf(viewer, " INFO(23) (num of pivots eliminated on this processor after factorization): \n");CHKERRQ(ierr); 1154 ierr = PetscViewerASCIISynchronizedPrintf(viewer," [%d] %d \n",lu->myid,lu->id.INFO(23));CHKERRQ(ierr); 1155 ierr = PetscViewerFlush(viewer); 1156 ierr = PetscViewerASCIISynchronizedAllow(viewer,PETSC_FALSE);CHKERRQ(ierr); 1157 1158 if (!lu->myid){ /* information from the host */ 1159 ierr = PetscViewerASCIIPrintf(viewer," RINFOG(1) (global estimated flops for the elimination after analysis): %g \n",lu->id.RINFOG(1));CHKERRQ(ierr); 1160 ierr = PetscViewerASCIIPrintf(viewer," RINFOG(2) (global estimated flops for the assembly after factorization): %g \n",lu->id.RINFOG(2));CHKERRQ(ierr); 1161 ierr = PetscViewerASCIIPrintf(viewer," RINFOG(3) (global estimated flops for the elimination after factorization): %g \n",lu->id.RINFOG(3));CHKERRQ(ierr); 1162 ierr = PetscViewerASCIIPrintf(viewer," (RINFOG(12) RINFOG(13))*2^INFOG(34) (determinant): (%g,%g)*(2^%d)\n",lu->id.RINFOG(12),lu->id.RINFOG(13),lu->id.INFOG(34));CHKERRQ(ierr); 1163 1164 ierr = PetscViewerASCIIPrintf(viewer," INFOG(3) (estimated real workspace for factors on all processors after analysis): %d \n",lu->id.INFOG(3));CHKERRQ(ierr); 1165 ierr = PetscViewerASCIIPrintf(viewer," INFOG(4) (estimated integer workspace for factors on all processors after analysis): %d \n",lu->id.INFOG(4));CHKERRQ(ierr); 1166 ierr = PetscViewerASCIIPrintf(viewer," INFOG(5) (estimated maximum front size in the complete tree): %d \n",lu->id.INFOG(5));CHKERRQ(ierr); 1167 ierr = PetscViewerASCIIPrintf(viewer," INFOG(6) (number of nodes in the complete tree): %d \n",lu->id.INFOG(6));CHKERRQ(ierr); 1168 ierr = PetscViewerASCIIPrintf(viewer," INFOG(7) (ordering option effectively use after analysis): %d \n",lu->id.INFOG(7));CHKERRQ(ierr); 1169 ierr = PetscViewerASCIIPrintf(viewer," INFOG(8) (structural symmetry in percent of the permuted matrix after analysis): %d \n",lu->id.INFOG(8));CHKERRQ(ierr); 1170 ierr = PetscViewerASCIIPrintf(viewer," INFOG(9) (total real/complex workspace to store the matrix factors after factorization): %d \n",lu->id.INFOG(9));CHKERRQ(ierr); 1171 ierr = PetscViewerASCIIPrintf(viewer," INFOG(10) (total integer space store the matrix factors after factorization): %d \n",lu->id.INFOG(10));CHKERRQ(ierr); 1172 ierr = PetscViewerASCIIPrintf(viewer," INFOG(11) (order of largest frontal matrix after factorization): %d \n",lu->id.INFOG(11));CHKERRQ(ierr); 1173 ierr = PetscViewerASCIIPrintf(viewer," INFOG(12) (number of off-diagonal pivots): %d \n",lu->id.INFOG(12));CHKERRQ(ierr); 1174 ierr = PetscViewerASCIIPrintf(viewer," INFOG(13) (number of delayed pivots after factorization): %d \n",lu->id.INFOG(13));CHKERRQ(ierr); 1175 ierr = PetscViewerASCIIPrintf(viewer," INFOG(14) (number of memory compress after factorization): %d \n",lu->id.INFOG(14));CHKERRQ(ierr); 1176 ierr = PetscViewerASCIIPrintf(viewer," INFOG(15) (number of steps of iterative refinement after solution): %d \n",lu->id.INFOG(15));CHKERRQ(ierr); 1177 ierr = PetscViewerASCIIPrintf(viewer," INFOG(16) (estimated size (in MB) of all MUMPS internal data for factorization after analysis: value on the most memory consuming processor): %d \n",lu->id.INFOG(16));CHKERRQ(ierr); 1178 ierr = PetscViewerASCIIPrintf(viewer," INFOG(17) (estimated size of all MUMPS internal data for factorization after analysis: sum over all processors): %d \n",lu->id.INFOG(17));CHKERRQ(ierr); 1179 ierr = PetscViewerASCIIPrintf(viewer," INFOG(18) (size of all MUMPS internal data allocated during factorization: value on the most memory consuming processor): %d \n",lu->id.INFOG(18));CHKERRQ(ierr); 1180 ierr = PetscViewerASCIIPrintf(viewer," INFOG(19) (size of all MUMPS internal data allocated during factorization: sum over all processors): %d \n",lu->id.INFOG(19));CHKERRQ(ierr); 1181 ierr = PetscViewerASCIIPrintf(viewer," INFOG(20) (estimated number of entries in the factors): %d \n",lu->id.INFOG(20));CHKERRQ(ierr); 1182 ierr = PetscViewerASCIIPrintf(viewer," INFOG(21) (size in MB of memory effectively used during factorization - value on the most memory consuming processor): %d \n",lu->id.INFOG(21));CHKERRQ(ierr); 1183 ierr = PetscViewerASCIIPrintf(viewer," INFOG(22) (size in MB of memory effectively used during factorization - sum over all processors): %d \n",lu->id.INFOG(22));CHKERRQ(ierr); 1184 ierr = PetscViewerASCIIPrintf(viewer," INFOG(23) (after analysis: value of ICNTL(6) effectively used): %d \n",lu->id.INFOG(23));CHKERRQ(ierr); 1185 ierr = PetscViewerASCIIPrintf(viewer," INFOG(24) (after analysis: value of ICNTL(12) effectively used): %d \n",lu->id.INFOG(24));CHKERRQ(ierr); 1186 ierr = PetscViewerASCIIPrintf(viewer," INFOG(25) (after factorization: number of pivots modified by static pivoting): %d \n",lu->id.INFOG(25));CHKERRQ(ierr); 1187 } 1188 } 1189 } 1190 PetscFunctionReturn(0); 1191 } 1192 1193 #undef __FUNCT__ 1194 #define __FUNCT__ "MatGetInfo_MUMPS" 1195 PetscErrorCode MatGetInfo_MUMPS(Mat A,MatInfoType flag,MatInfo *info) 1196 { 1197 Mat_MUMPS *mumps =(Mat_MUMPS*)A->spptr; 1198 1199 PetscFunctionBegin; 1200 info->block_size = 1.0; 1201 info->nz_allocated = mumps->id.INFOG(20); 1202 info->nz_used = mumps->id.INFOG(20); 1203 info->nz_unneeded = 0.0; 1204 info->assemblies = 0.0; 1205 info->mallocs = 0.0; 1206 info->memory = 0.0; 1207 info->fill_ratio_given = 0; 1208 info->fill_ratio_needed = 0; 1209 info->factor_mallocs = 0; 1210 PetscFunctionReturn(0); 1211 } 1212 1213 /* -------------------------------------------------------------------------------------------*/ 1214 #undef __FUNCT__ 1215 #define __FUNCT__ "MatMumpsSetIcntl_MUMPS" 1216 PetscErrorCode MatMumpsSetIcntl_MUMPS(Mat F,PetscInt icntl,PetscInt ival) 1217 { 1218 Mat_MUMPS *lu =(Mat_MUMPS*)F->spptr; 1219 1220 PetscFunctionBegin; 1221 lu->id.ICNTL(icntl) = ival; 1222 PetscFunctionReturn(0); 1223 } 1224 1225 #undef __FUNCT__ 1226 #define __FUNCT__ "MatMumpsSetIcntl" 1227 /*@ 1228 MatMumpsSetIcntl - Set MUMPS parameter ICNTL() 1229 1230 Logically Collective on Mat 1231 1232 Input Parameters: 1233 + F - the factored matrix obtained by calling MatGetFactor() from PETSc-MUMPS interface 1234 . icntl - index of MUMPS parameter array ICNTL() 1235 - ival - value of MUMPS ICNTL(icntl) 1236 1237 Options Database: 1238 . -mat_mumps_icntl_<icntl> <ival> 1239 1240 Level: beginner 1241 1242 References: MUMPS Users' Guide 1243 1244 .seealso: MatGetFactor() 1245 @*/ 1246 PetscErrorCode MatMumpsSetIcntl(Mat F,PetscInt icntl,PetscInt ival) 1247 { 1248 PetscErrorCode ierr; 1249 1250 PetscFunctionBegin; 1251 PetscValidLogicalCollectiveInt(F,icntl,2); 1252 PetscValidLogicalCollectiveInt(F,ival,3); 1253 ierr = PetscTryMethod(F,"MatMumpsSetIcntl_C",(Mat,PetscInt,PetscInt),(F,icntl,ival));CHKERRQ(ierr); 1254 PetscFunctionReturn(0); 1255 } 1256 1257 /*MC 1258 MATSOLVERMUMPS - A matrix type providing direct solvers (LU and Cholesky) for 1259 distributed and sequential matrices via the external package MUMPS. 1260 1261 Works with MATAIJ and MATSBAIJ matrices 1262 1263 Options Database Keys: 1264 + -mat_mumps_icntl_4 <0,...,4> - print level 1265 . -mat_mumps_icntl_6 <0,...,7> - matrix prescaling options (see MUMPS User's Guide) 1266 . -mat_mumps_icntl_7 <0,...,7> - matrix orderings (see MUMPS User's Guidec) 1267 . -mat_mumps_icntl_9 <1,2> - A or A^T x=b to be solved: 1 denotes A, 2 denotes A^T 1268 . -mat_mumps_icntl_10 <n> - maximum number of iterative refinements 1269 . -mat_mumps_icntl_11 <n> - error analysis, a positive value returns statistics during -ksp_view 1270 . -mat_mumps_icntl_12 <n> - efficiency control (see MUMPS User's Guide) 1271 . -mat_mumps_icntl_13 <n> - efficiency control (see MUMPS User's Guide) 1272 . -mat_mumps_icntl_14 <n> - efficiency control (see MUMPS User's Guide) 1273 . -mat_mumps_icntl_15 <n> - efficiency control (see MUMPS User's Guide) 1274 . -mat_mumps_cntl_1 <delta> - relative pivoting threshold 1275 . -mat_mumps_cntl_2 <tol> - stopping criterion for refinement 1276 - -mat_mumps_cntl_3 <adelta> - absolute pivoting threshold 1277 1278 Level: beginner 1279 1280 .seealso: PCFactorSetMatSolverPackage(), MatSolverPackage 1281 1282 M*/ 1283 1284 EXTERN_C_BEGIN 1285 #undef __FUNCT__ 1286 #define __FUNCT__ "MatFactorGetSolverPackage_mumps" 1287 PetscErrorCode MatFactorGetSolverPackage_mumps(Mat A,const MatSolverPackage *type) 1288 { 1289 PetscFunctionBegin; 1290 *type = MATSOLVERMUMPS; 1291 PetscFunctionReturn(0); 1292 } 1293 EXTERN_C_END 1294 1295 EXTERN_C_BEGIN 1296 /* MatGetFactor for Seq and MPI AIJ matrices */ 1297 #undef __FUNCT__ 1298 #define __FUNCT__ "MatGetFactor_aij_mumps" 1299 PetscErrorCode MatGetFactor_aij_mumps(Mat A,MatFactorType ftype,Mat *F) 1300 { 1301 Mat B; 1302 PetscErrorCode ierr; 1303 Mat_MUMPS *mumps; 1304 PetscBool isSeqAIJ; 1305 1306 PetscFunctionBegin; 1307 /* Create the factorization matrix */ 1308 ierr = PetscObjectTypeCompare((PetscObject)A,MATSEQAIJ,&isSeqAIJ);CHKERRQ(ierr); 1309 ierr = MatCreate(((PetscObject)A)->comm,&B);CHKERRQ(ierr); 1310 ierr = MatSetSizes(B,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);CHKERRQ(ierr); 1311 ierr = MatSetType(B,((PetscObject)A)->type_name);CHKERRQ(ierr); 1312 if (isSeqAIJ) { 1313 ierr = MatSeqAIJSetPreallocation(B,0,PETSC_NULL);CHKERRQ(ierr); 1314 } else { 1315 ierr = MatMPIAIJSetPreallocation(B,0,PETSC_NULL,0,PETSC_NULL);CHKERRQ(ierr); 1316 } 1317 1318 ierr = PetscNewLog(B,Mat_MUMPS,&mumps);CHKERRQ(ierr); 1319 B->ops->view = MatView_MUMPS; 1320 B->ops->getinfo = MatGetInfo_MUMPS; 1321 ierr = PetscObjectComposeFunctionDynamic((PetscObject)B,"MatFactorGetSolverPackage_C","MatFactorGetSolverPackage_mumps",MatFactorGetSolverPackage_mumps);CHKERRQ(ierr); 1322 ierr = PetscObjectComposeFunctionDynamic((PetscObject)B,"MatMumpsSetIcntl_C","MatMumpsSetIcntl_MUMPS",MatMumpsSetIcntl_MUMPS);CHKERRQ(ierr); 1323 if (ftype == MAT_FACTOR_LU) { 1324 B->ops->lufactorsymbolic = MatLUFactorSymbolic_AIJMUMPS; 1325 B->factortype = MAT_FACTOR_LU; 1326 if (isSeqAIJ) mumps->ConvertToTriples = MatConvertToTriples_seqaij_seqaij; 1327 else mumps->ConvertToTriples = MatConvertToTriples_mpiaij_mpiaij; 1328 mumps->sym = 0; 1329 } else { 1330 B->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_MUMPS; 1331 B->factortype = MAT_FACTOR_CHOLESKY; 1332 if (isSeqAIJ) mumps->ConvertToTriples = MatConvertToTriples_seqaij_seqsbaij; 1333 else mumps->ConvertToTriples = MatConvertToTriples_mpiaij_mpisbaij; 1334 if (A->spd_set && A->spd) mumps->sym = 1; 1335 else mumps->sym = 2; 1336 } 1337 1338 mumps->isAIJ = PETSC_TRUE; 1339 mumps->Destroy = B->ops->destroy; 1340 B->ops->destroy = MatDestroy_MUMPS; 1341 B->spptr = (void*)mumps; 1342 ierr = PetscInitializeMUMPS(A,mumps);CHKERRQ(ierr); 1343 1344 *F = B; 1345 PetscFunctionReturn(0); 1346 } 1347 EXTERN_C_END 1348 1349 1350 EXTERN_C_BEGIN 1351 /* MatGetFactor for Seq and MPI SBAIJ matrices */ 1352 #undef __FUNCT__ 1353 #define __FUNCT__ "MatGetFactor_sbaij_mumps" 1354 PetscErrorCode MatGetFactor_sbaij_mumps(Mat A,MatFactorType ftype,Mat *F) 1355 { 1356 Mat B; 1357 PetscErrorCode ierr; 1358 Mat_MUMPS *mumps; 1359 PetscBool isSeqSBAIJ; 1360 1361 PetscFunctionBegin; 1362 if (ftype != MAT_FACTOR_CHOLESKY) SETERRQ(((PetscObject)A)->comm,PETSC_ERR_SUP,"Cannot use PETSc SBAIJ matrices with MUMPS LU, use AIJ matrix"); 1363 if(A->rmap->bs > 1) SETERRQ(((PetscObject)A)->comm,PETSC_ERR_SUP,"Cannot use PETSc SBAIJ matrices with block size > 1 with MUMPS Cholesky, use AIJ matrix instead"); 1364 ierr = PetscObjectTypeCompare((PetscObject)A,MATSEQSBAIJ,&isSeqSBAIJ);CHKERRQ(ierr); 1365 /* Create the factorization matrix */ 1366 ierr = MatCreate(((PetscObject)A)->comm,&B);CHKERRQ(ierr); 1367 ierr = MatSetSizes(B,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);CHKERRQ(ierr); 1368 ierr = MatSetType(B,((PetscObject)A)->type_name);CHKERRQ(ierr); 1369 ierr = PetscNewLog(B,Mat_MUMPS,&mumps);CHKERRQ(ierr); 1370 if (isSeqSBAIJ) { 1371 ierr = MatSeqSBAIJSetPreallocation(B,1,0,PETSC_NULL);CHKERRQ(ierr); 1372 mumps->ConvertToTriples = MatConvertToTriples_seqsbaij_seqsbaij; 1373 } else { 1374 ierr = MatMPISBAIJSetPreallocation(B,1,0,PETSC_NULL,0,PETSC_NULL);CHKERRQ(ierr); 1375 mumps->ConvertToTriples = MatConvertToTriples_mpisbaij_mpisbaij; 1376 } 1377 1378 B->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_MUMPS; 1379 B->ops->view = MatView_MUMPS; 1380 ierr = PetscObjectComposeFunctionDynamic((PetscObject)B,"MatFactorGetSolverPackage_C","MatFactorGetSolverPackage_mumps",MatFactorGetSolverPackage_mumps);CHKERRQ(ierr); 1381 ierr = PetscObjectComposeFunctionDynamic((PetscObject)B,"MatMumpsSetIcntl_C","MatMumpsSetIcntl",MatMumpsSetIcntl);CHKERRQ(ierr); 1382 B->factortype = MAT_FACTOR_CHOLESKY; 1383 if (A->spd_set && A->spd) mumps->sym = 1; 1384 else mumps->sym = 2; 1385 1386 mumps->isAIJ = PETSC_FALSE; 1387 mumps->Destroy = B->ops->destroy; 1388 B->ops->destroy = MatDestroy_MUMPS; 1389 B->spptr = (void*)mumps; 1390 ierr = PetscInitializeMUMPS(A,mumps);CHKERRQ(ierr); 1391 1392 *F = B; 1393 PetscFunctionReturn(0); 1394 } 1395 EXTERN_C_END 1396 1397 EXTERN_C_BEGIN 1398 #undef __FUNCT__ 1399 #define __FUNCT__ "MatGetFactor_baij_mumps" 1400 PetscErrorCode MatGetFactor_baij_mumps(Mat A,MatFactorType ftype,Mat *F) 1401 { 1402 Mat B; 1403 PetscErrorCode ierr; 1404 Mat_MUMPS *mumps; 1405 PetscBool isSeqBAIJ; 1406 1407 PetscFunctionBegin; 1408 /* Create the factorization matrix */ 1409 ierr = PetscObjectTypeCompare((PetscObject)A,MATSEQBAIJ,&isSeqBAIJ);CHKERRQ(ierr); 1410 ierr = MatCreate(((PetscObject)A)->comm,&B);CHKERRQ(ierr); 1411 ierr = MatSetSizes(B,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);CHKERRQ(ierr); 1412 ierr = MatSetType(B,((PetscObject)A)->type_name);CHKERRQ(ierr); 1413 if (isSeqBAIJ) { 1414 ierr = MatSeqBAIJSetPreallocation(B,A->rmap->bs,0,PETSC_NULL);CHKERRQ(ierr); 1415 } else { 1416 ierr = MatMPIBAIJSetPreallocation(B,A->rmap->bs,0,PETSC_NULL,0,PETSC_NULL);CHKERRQ(ierr); 1417 } 1418 1419 ierr = PetscNewLog(B,Mat_MUMPS,&mumps);CHKERRQ(ierr); 1420 if (ftype == MAT_FACTOR_LU) { 1421 B->ops->lufactorsymbolic = MatLUFactorSymbolic_BAIJMUMPS; 1422 B->factortype = MAT_FACTOR_LU; 1423 if (isSeqBAIJ) mumps->ConvertToTriples = MatConvertToTriples_seqbaij_seqaij; 1424 else mumps->ConvertToTriples = MatConvertToTriples_mpibaij_mpiaij; 1425 mumps->sym = 0; 1426 } else { 1427 SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Cannot use PETSc BAIJ matrices with MUMPS Cholesky, use SBAIJ or AIJ matrix instead\n"); 1428 } 1429 1430 B->ops->view = MatView_MUMPS; 1431 ierr = PetscObjectComposeFunctionDynamic((PetscObject)B,"MatFactorGetSolverPackage_C","MatFactorGetSolverPackage_mumps",MatFactorGetSolverPackage_mumps);CHKERRQ(ierr); 1432 ierr = PetscObjectComposeFunctionDynamic((PetscObject)B,"MatMumpsSetIcntl_C","MatMumpsSetIcntl_MUMPS",MatMumpsSetIcntl_MUMPS);CHKERRQ(ierr); 1433 1434 mumps->isAIJ = PETSC_TRUE; 1435 mumps->Destroy = B->ops->destroy; 1436 B->ops->destroy = MatDestroy_MUMPS; 1437 B->spptr = (void*)mumps; 1438 ierr = PetscInitializeMUMPS(A,mumps);CHKERRQ(ierr); 1439 1440 *F = B; 1441 PetscFunctionReturn(0); 1442 } 1443 EXTERN_C_END 1444 1445