1 /*$Id: superlu_DIST.c,v 1.10 2001/08/15 15:56:50 bsmith Exp $*/ 2 /* 3 Provides an interface to the SuperLU_DIST_2.0 sparse solver 4 */ 5 6 #include "src/mat/impls/aij/seq/aij.h" 7 #include "src/mat/impls/aij/mpi/mpiaij.h" 8 #if defined(PETSC_HAVE_STDLIB_H) /* This is to get arround weird problem with SuperLU on cray */ 9 #include "stdlib.h" 10 #endif 11 12 /*MC 13 MATSUPERLU_DIST - a matrix type providing direct solvers for parallel matrices 14 via the external package SuperLU_DIST. 15 16 If SuperLU_DIST is installed (see the manual for 17 instructions on how to declare the existence of external packages), 18 a matrix type can be constructed which invokes SuperLU_DIST solvers. 19 After calling MatCreate(...,A), simply call MatSetType(A,MATSUPERLU_DIST). 20 This matrix type is only supported for double precision real. 21 22 This matrix inherits from MATSEQAIJ when constructed with a single process communicator, 23 and from MATMPIAIJ otherwise. 24 25 Options Database Keys: 26 + -mat_type superlu_dist 27 . -mat_superlu_dist_r <n> : number of rows in processor partition 28 . -mat_superlu_dist_c <n> : number of columns in processor partition 29 . -mat_superlu_dist_matinput 0|1 : matrix input mode; 0=global, 1=distributed 30 . -mat_superlu_dist_equil :, equilibrate the matrix 31 . -mat_superlu_dist_rowperm LargeDiag|NATURAL : row permutation 32 . -mat_superlu_dist_colperm MMD_AT_PLUS_A|MMD_ATA|COLAMD|NATURAL : column permutation 33 . -mat_superlu_dist_replacetinypivot : replace tiny pivots 34 . -mat_superlu_dist_iterrefine : use iterative refinement 35 - -mat_superlu_dist_statprint : print factorization information 36 37 .seealso: PCLU 38 M*/ 39 40 EXTERN_C_BEGIN 41 #if defined(PETSC_USE_COMPLEX) 42 #include "superlu_zdefs.h" 43 #else 44 #include "superlu_ddefs.h" 45 #endif 46 EXTERN_C_END 47 48 typedef enum { GLOBAL,DISTRIBUTED 49 } SuperLU_MatInputMode; 50 51 typedef struct { 52 int_t nprow,npcol,*row,*col; 53 gridinfo_t grid; 54 superlu_options_t options; 55 SuperMatrix A_sup; 56 ScalePermstruct_t ScalePermstruct; 57 LUstruct_t LUstruct; 58 int StatPrint; 59 int MatInputMode; 60 SOLVEstruct_t SOLVEstruct; 61 MatStructure flg; 62 MPI_Comm comm_superlu; 63 #if defined(PETSC_USE_COMPLEX) 64 doublecomplex *val; 65 #else 66 double *val; 67 #endif 68 69 /* A few function pointers for inheritance */ 70 int (*MatView)(Mat,PetscViewer); 71 int (*MatAssemblyEnd)(Mat,MatAssemblyType); 72 int (*MatDestroy)(Mat); 73 74 /* Flag to clean up (non-global) SuperLU objects during Destroy */ 75 PetscTruth CleanUpSuperLUDist; 76 } Mat_MPIAIJ_SuperLU_DIST; 77 78 #undef __FUNCT__ 79 #define __FUNCT__ "MatDestroy_MPIAIJ_SuperLU_DIST" 80 int MatDestroy_MPIAIJ_SuperLU_DIST(Mat A) 81 { 82 Mat_MPIAIJ_SuperLU_DIST *lu = (Mat_MPIAIJ_SuperLU_DIST*)A->spptr; 83 int ierr,(*destroy)(Mat); 84 85 PetscFunctionBegin; 86 if (lu->CleanUpSuperLUDist) { 87 /* Deallocate SuperLU_DIST storage */ 88 if (lu->MatInputMode == GLOBAL) { 89 Destroy_CompCol_Matrix_dist(&lu->A_sup); 90 } else { 91 Destroy_CompRowLoc_Matrix_dist(&lu->A_sup); 92 if ( lu->options.SolveInitialized ) { 93 #if defined(PETSC_USE_COMPLEX) 94 zSolveFinalize(&lu->options, &lu->SOLVEstruct); 95 #else 96 dSolveFinalize(&lu->options, &lu->SOLVEstruct); 97 #endif 98 } 99 } 100 Destroy_LU(A->N, &lu->grid, &lu->LUstruct); 101 ScalePermstructFree(&lu->ScalePermstruct); 102 LUstructFree(&lu->LUstruct); 103 104 /* Release the SuperLU_DIST process grid. */ 105 superlu_gridexit(&lu->grid); 106 107 ierr = MPI_Comm_free(&(lu->comm_superlu));CHKERRQ(ierr); 108 } 109 destroy = lu->MatDestroy; 110 ierr = PetscFree(lu);CHKERRQ(ierr); 111 ierr = (*destroy)(A);CHKERRQ(ierr); 112 113 PetscFunctionReturn(0); 114 } 115 116 extern int MatMPIAIJFactorInfo_SuperLu(Mat A,PetscViewer viewer); 117 118 #undef __FUNCT__ 119 #define __FUNCT__ "MatView_MPIAIJ_Spooles_DIST" 120 int MatView_MPIAIJ_SuperLU_DIST(Mat A,PetscViewer viewer) 121 { 122 int ierr; 123 PetscTruth isascii; 124 PetscViewerFormat format; 125 Mat_MPIAIJ_SuperLU_DIST *lu=(Mat_MPIAIJ_SuperLU_DIST*)(A->spptr); 126 127 PetscFunctionBegin; 128 ierr = (*lu->MatView)(A,viewer);CHKERRQ(ierr); 129 130 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&isascii);CHKERRQ(ierr); 131 if (isascii) { 132 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 133 if (format == PETSC_VIEWER_ASCII_FACTOR_INFO) { 134 ierr = MatMPIAIJFactorInfo_SuperLu(A,viewer);CHKERRQ(ierr); 135 } 136 } 137 PetscFunctionReturn(0); 138 } 139 140 #undef __FUNCT__ 141 #define __FUNCT__ "MatAssemblyEnd_MPIAIJ_SuperLU_DIST" 142 int MatAssemblyEnd_MPIAIJ_SuperLU_DIST(Mat A,MatAssemblyType mode) { 143 int ierr; 144 Mat_MPIAIJ_SuperLU_DIST *lu=(Mat_MPIAIJ_SuperLU_DIST*)(A->spptr); 145 146 PetscFunctionBegin; 147 ierr = (*lu->MatAssemblyEnd)(A,mode);CHKERRQ(ierr); 148 ierr = MatUseSuperLU_DIST_MPIAIJ(A);CHKERRQ(ierr); 149 PetscFunctionReturn(0); 150 } 151 152 #undef __FUNCT__ 153 #define __FUNCT__ "MatSolve_MPIAIJ_SuperLU_DIST" 154 int MatSolve_MPIAIJ_SuperLU_DIST(Mat A,Vec b_mpi,Vec x) 155 { 156 Mat_MPIAIJ *aa = (Mat_MPIAIJ*)A->data; 157 Mat_MPIAIJ_SuperLU_DIST *lu = (Mat_MPIAIJ_SuperLU_DIST*)A->spptr; 158 int ierr, size=aa->size; 159 int m=A->M, N=A->N; 160 SuperLUStat_t stat; 161 double berr[1]; 162 PetscScalar *bptr; 163 int info, nrhs=1; 164 Vec x_seq; 165 IS iden; 166 VecScatter scat; 167 PetscLogDouble time0,time,time_min,time_max; 168 169 PetscFunctionBegin; 170 if (size > 1) { 171 if (lu->MatInputMode == GLOBAL) { /* global mat input, convert b to x_seq */ 172 ierr = VecCreateSeq(PETSC_COMM_SELF,N,&x_seq);CHKERRQ(ierr); 173 ierr = ISCreateStride(PETSC_COMM_SELF,N,0,1,&iden);CHKERRQ(ierr); 174 ierr = VecScatterCreate(b_mpi,iden,x_seq,iden,&scat);CHKERRQ(ierr); 175 ierr = ISDestroy(iden);CHKERRQ(ierr); 176 177 ierr = VecScatterBegin(b_mpi,x_seq,INSERT_VALUES,SCATTER_FORWARD,scat);CHKERRQ(ierr); 178 ierr = VecScatterEnd(b_mpi,x_seq,INSERT_VALUES,SCATTER_FORWARD,scat);CHKERRQ(ierr); 179 ierr = VecGetArray(x_seq,&bptr);CHKERRQ(ierr); 180 } else { /* distributed mat input */ 181 ierr = VecCopy(b_mpi,x);CHKERRQ(ierr); 182 ierr = VecGetArray(x,&bptr);CHKERRQ(ierr); 183 } 184 } else { /* size == 1 */ 185 ierr = VecCopy(b_mpi,x);CHKERRQ(ierr); 186 ierr = VecGetArray(x,&bptr);CHKERRQ(ierr); 187 } 188 189 lu->options.Fact = FACTORED; /* The factored form of A is supplied. Local option used by this func. only.*/ 190 191 PStatInit(&stat); /* Initialize the statistics variables. */ 192 if (lu->StatPrint) { 193 ierr = MPI_Barrier(A->comm);CHKERRQ(ierr); /* to be removed */ 194 ierr = PetscGetTime(&time0);CHKERRQ(ierr); /* to be removed */ 195 } 196 if (lu->MatInputMode == GLOBAL) { 197 #if defined(PETSC_USE_COMPLEX) 198 pzgssvx_ABglobal(&lu->options, &lu->A_sup, &lu->ScalePermstruct,(doublecomplex*)bptr, m, nrhs, 199 &lu->grid, &lu->LUstruct, berr, &stat, &info); 200 #else 201 pdgssvx_ABglobal(&lu->options, &lu->A_sup, &lu->ScalePermstruct,bptr, m, nrhs, 202 &lu->grid, &lu->LUstruct, berr, &stat, &info); 203 #endif 204 } else { /* distributed mat input */ 205 #if defined(PETSC_USE_COMPLEX) 206 pzgssvx(&lu->options, &lu->A_sup, &lu->ScalePermstruct, (doublecomplex*)bptr, A->M, nrhs, &lu->grid, 207 &lu->LUstruct, &lu->SOLVEstruct, berr, &stat, &info); 208 if (info) SETERRQ1(1,"pzgssvx fails, info: %d\n",info); 209 #else 210 pdgssvx(&lu->options, &lu->A_sup, &lu->ScalePermstruct, bptr, A->M, nrhs, &lu->grid, 211 &lu->LUstruct, &lu->SOLVEstruct, berr, &stat, &info); 212 if (info) SETERRQ1(1,"pdgssvx fails, info: %d\n",info); 213 #endif 214 } 215 if (lu->StatPrint) { 216 ierr = PetscGetTime(&time);CHKERRQ(ierr); /* to be removed */ 217 PStatPrint(&lu->options, &stat, &lu->grid); /* Print the statistics. */ 218 } 219 PStatFree(&stat); 220 221 if (size > 1) { 222 if (lu->MatInputMode == GLOBAL){ /* convert seq x to mpi x */ 223 ierr = VecRestoreArray(x_seq,&bptr);CHKERRQ(ierr); 224 ierr = VecScatterBegin(x_seq,x,INSERT_VALUES,SCATTER_REVERSE,scat);CHKERRQ(ierr); 225 ierr = VecScatterEnd(x_seq,x,INSERT_VALUES,SCATTER_REVERSE,scat);CHKERRQ(ierr); 226 ierr = VecScatterDestroy(scat);CHKERRQ(ierr); 227 ierr = VecDestroy(x_seq);CHKERRQ(ierr); 228 } else { 229 ierr = VecRestoreArray(x,&bptr);CHKERRQ(ierr); 230 } 231 } else { 232 ierr = VecRestoreArray(x,&bptr);CHKERRQ(ierr); 233 } 234 if (lu->StatPrint) { 235 time0 = time - time0; 236 ierr = MPI_Reduce(&time0,&time_max,1,MPI_DOUBLE,MPI_MAX,0,A->comm);CHKERRQ(ierr); 237 ierr = MPI_Reduce(&time0,&time_min,1,MPI_DOUBLE,MPI_MIN,0,A->comm);CHKERRQ(ierr); 238 ierr = MPI_Reduce(&time0,&time,1,MPI_DOUBLE,MPI_SUM,0,A->comm);CHKERRQ(ierr); 239 time = time/size; /* average time */ 240 ierr = PetscPrintf(A->comm, " Time for superlu_dist solve (max/min/avg): %g / %g / %g\n\n",time_max,time_min,time);CHKERRQ(ierr); 241 } 242 PetscFunctionReturn(0); 243 } 244 245 #undef __FUNCT__ 246 #define __FUNCT__ "MatLUFactorNumeric_MPIAIJ_SuperLU_DIST" 247 int MatLUFactorNumeric_MPIAIJ_SuperLU_DIST(Mat A,Mat *F) 248 { 249 Mat_MPIAIJ *fac = (Mat_MPIAIJ*)(*F)->data,*mat; 250 Mat *tseq,A_seq = PETSC_NULL; 251 Mat_SeqAIJ *aa,*bb; 252 Mat_MPIAIJ_SuperLU_DIST *lu = (Mat_MPIAIJ_SuperLU_DIST*)(*F)->spptr; 253 int M=A->M,N=A->N,info,ierr,size=fac->size,i,*ai,*aj,*bi,*bj,nz,rstart,*garray, 254 m=A->m, irow,colA_start,j,jcol,jB,countA,countB,*bjj,*ajj; 255 SuperLUStat_t stat; 256 double *berr=0; 257 IS isrow; 258 PetscLogDouble time0[2],time[2],time_min[2],time_max[2]; 259 #if defined(PETSC_USE_COMPLEX) 260 doublecomplex *av, *bv; 261 #else 262 double *av, *bv; 263 #endif 264 265 PetscFunctionBegin; 266 if (lu->StatPrint) { 267 ierr = MPI_Barrier(A->comm);CHKERRQ(ierr); 268 ierr = PetscGetTime(&time0[0]);CHKERRQ(ierr); 269 } 270 271 if (lu->MatInputMode == GLOBAL) { /* global mat input */ 272 if (size > 1) { /* convert mpi A to seq mat A */ 273 ierr = ISCreateStride(PETSC_COMM_SELF,M,0,1,&isrow); CHKERRQ(ierr); 274 ierr = MatGetSubMatrices(A,1,&isrow,&isrow,MAT_INITIAL_MATRIX,&tseq); CHKERRQ(ierr); 275 ierr = ISDestroy(isrow);CHKERRQ(ierr); 276 277 A_seq = *tseq; 278 ierr = PetscFree(tseq);CHKERRQ(ierr); 279 aa = (Mat_SeqAIJ*)A_seq->data; 280 } else { 281 aa = (Mat_SeqAIJ*)A->data; 282 } 283 284 /* Allocate storage, then convert Petsc NR matrix to SuperLU_DIST NC */ 285 if (lu->flg == DIFFERENT_NONZERO_PATTERN) {/* first numeric factorization */ 286 #if defined(PETSC_USE_COMPLEX) 287 zallocateA_dist(N, aa->nz, &lu->val, &lu->col, &lu->row); 288 #else 289 dallocateA_dist(N, aa->nz, &lu->val, &lu->col, &lu->row); 290 #endif 291 } else { /* successive numeric factorization, sparsity pattern is reused. */ 292 Destroy_CompCol_Matrix_dist(&lu->A_sup); 293 Destroy_LU(N, &lu->grid, &lu->LUstruct); 294 lu->options.Fact = SamePattern; 295 } 296 #if defined(PETSC_USE_COMPLEX) 297 zCompRow_to_CompCol_dist(M,N,aa->nz,(doublecomplex*)aa->a,aa->j,aa->i,&lu->val,&lu->col, &lu->row); 298 #else 299 dCompRow_to_CompCol_dist(M,N,aa->nz,aa->a,aa->j,aa->i,&lu->val, &lu->col, &lu->row); 300 #endif 301 302 /* Create compressed column matrix A_sup. */ 303 #if defined(PETSC_USE_COMPLEX) 304 zCreate_CompCol_Matrix_dist(&lu->A_sup, M, N, aa->nz, lu->val, lu->col, lu->row, SLU_NC, SLU_Z, SLU_GE); 305 #else 306 dCreate_CompCol_Matrix_dist(&lu->A_sup, M, N, aa->nz, lu->val, lu->col, lu->row, SLU_NC, SLU_D, SLU_GE); 307 #endif 308 } else { /* distributed mat input */ 309 mat = (Mat_MPIAIJ*)A->data; 310 aa=(Mat_SeqAIJ*)(mat->A)->data; 311 bb=(Mat_SeqAIJ*)(mat->B)->data; 312 ai=aa->i; aj=aa->j; 313 bi=bb->i; bj=bb->j; 314 #if defined(PETSC_USE_COMPLEX) 315 av=(doublecomplex*)aa->a; 316 bv=(doublecomplex*)bb->a; 317 #else 318 av=aa->a; 319 bv=bb->a; 320 #endif 321 rstart = mat->rstart; 322 nz = aa->nz + bb->nz; 323 garray = mat->garray; 324 rstart = mat->rstart; 325 326 if (lu->flg == DIFFERENT_NONZERO_PATTERN) {/* first numeric factorization */ 327 #if defined(PETSC_USE_COMPLEX) 328 zallocateA_dist(m, nz, &lu->val, &lu->col, &lu->row); 329 #else 330 dallocateA_dist(m, nz, &lu->val, &lu->col, &lu->row); 331 #endif 332 } else { /* successive numeric factorization, sparsity pattern and perm_c are reused. */ 333 /* Destroy_CompRowLoc_Matrix_dist(&lu->A_sup); */ /* crash! */ 334 Destroy_LU(N, &lu->grid, &lu->LUstruct); 335 lu->options.Fact = SamePattern; 336 } 337 nz = 0; jB = 0; irow = mat->rstart; 338 for ( i=0; i<m; i++ ) { 339 lu->row[i] = nz; 340 countA = ai[i+1] - ai[i]; 341 countB = bi[i+1] - bi[i]; 342 ajj = aj + ai[i]; /* ptr to the beginning of this row */ 343 bjj = bj + bi[i]; 344 345 /* B part, smaller col index */ 346 colA_start = mat->rstart + ajj[0]; /* the smallest global col index of A */ 347 for (j=0; j<countB; j++){ 348 jcol = garray[bjj[j]]; 349 if (jcol > colA_start) { 350 jB = j; 351 break; 352 } 353 lu->col[nz] = jcol; 354 lu->val[nz++] = *bv++; 355 if (j==countB-1) jB = countB; 356 } 357 358 /* A part */ 359 for (j=0; j<countA; j++){ 360 lu->col[nz] = mat->rstart + ajj[j]; 361 lu->val[nz++] = *av++; 362 } 363 364 /* B part, larger col index */ 365 for (j=jB; j<countB; j++){ 366 lu->col[nz] = garray[bjj[j]]; 367 lu->val[nz++] = *bv++; 368 } 369 } 370 lu->row[m] = nz; 371 #if defined(PETSC_USE_COMPLEX) 372 zCreate_CompRowLoc_Matrix_dist(&lu->A_sup, M, N, nz, m, rstart, 373 lu->val, lu->col, lu->row, SLU_NR_loc, SLU_Z, SLU_GE); 374 #else 375 dCreate_CompRowLoc_Matrix_dist(&lu->A_sup, M, N, nz, m, rstart, 376 lu->val, lu->col, lu->row, SLU_NR_loc, SLU_D, SLU_GE); 377 #endif 378 } 379 if (lu->StatPrint) { 380 ierr = PetscGetTime(&time[0]);CHKERRQ(ierr); 381 time0[0] = time[0] - time0[0]; 382 } 383 384 /* Factor the matrix. */ 385 PStatInit(&stat); /* Initialize the statistics variables. */ 386 387 if (lu->StatPrint) { 388 ierr = MPI_Barrier(A->comm);CHKERRQ(ierr); 389 ierr = PetscGetTime(&time0[1]);CHKERRQ(ierr); 390 } 391 392 if (lu->MatInputMode == GLOBAL) { /* global mat input */ 393 #if defined(PETSC_USE_COMPLEX) 394 pzgssvx_ABglobal(&lu->options, &lu->A_sup, &lu->ScalePermstruct, 0, M, 0, 395 &lu->grid, &lu->LUstruct, berr, &stat, &info); 396 #else 397 pdgssvx_ABglobal(&lu->options, &lu->A_sup, &lu->ScalePermstruct, 0, M, 0, 398 &lu->grid, &lu->LUstruct, berr, &stat, &info); 399 #endif 400 } else { /* distributed mat input */ 401 #if defined(PETSC_USE_COMPLEX) 402 pzgssvx(&lu->options, &lu->A_sup, &lu->ScalePermstruct, 0, M, 0, &lu->grid, 403 &lu->LUstruct, &lu->SOLVEstruct, berr, &stat, &info); 404 if (info) SETERRQ1(1,"pzgssvx fails, info: %d\n",info); 405 #else 406 pdgssvx(&lu->options, &lu->A_sup, &lu->ScalePermstruct, 0, M, 0, &lu->grid, 407 &lu->LUstruct, &lu->SOLVEstruct, berr, &stat, &info); 408 if (info) SETERRQ1(1,"pdgssvx fails, info: %d\n",info); 409 #endif 410 } 411 if (lu->StatPrint) { 412 ierr = PetscGetTime(&time[1]);CHKERRQ(ierr); /* to be removed */ 413 time0[1] = time[1] - time0[1]; 414 if (lu->StatPrint) PStatPrint(&lu->options, &stat, &lu->grid); /* Print the statistics. */ 415 } 416 PStatFree(&stat); 417 418 if (lu->MatInputMode == GLOBAL && size > 1){ 419 ierr = MatDestroy(A_seq);CHKERRQ(ierr); 420 } 421 422 if (lu->StatPrint) { 423 ierr = MPI_Reduce(time0,time_max,2,MPI_DOUBLE,MPI_MAX,0,A->comm); 424 ierr = MPI_Reduce(time0,time_min,2,MPI_DOUBLE,MPI_MIN,0,A->comm); 425 ierr = MPI_Reduce(time0,time,2,MPI_DOUBLE,MPI_SUM,0,A->comm); 426 for (i=0; i<2; i++) time[i] = time[i]/size; /* average time */ 427 ierr = PetscPrintf(A->comm, " Time for mat conversion (max/min/avg): %g / %g / %g\n",time_max[0],time_min[0],time[0]); 428 ierr = PetscPrintf(A->comm, " Time for superlu_dist fact (max/min/avg): %g / %g / %g\n\n",time_max[1],time_min[1],time[1]); 429 } 430 (*F)->assembled = PETSC_TRUE; 431 lu->flg = SAME_NONZERO_PATTERN; 432 PetscFunctionReturn(0); 433 } 434 435 /* Note the Petsc r and c permutations are ignored */ 436 #undef __FUNCT__ 437 #define __FUNCT__ "MatLUFactorSymbolic_MPIAIJ_SuperLU_DIST" 438 int MatLUFactorSymbolic_MPIAIJ_SuperLU_DIST(Mat A,IS r,IS c,MatFactorInfo *info,Mat *F) 439 { 440 Mat B; 441 Mat_MPIAIJ_SuperLU_DIST *lu; 442 int ierr,M=A->M,N=A->N,size; 443 superlu_options_t options; 444 char buff[32]; 445 PetscTruth flg; 446 char *ptype[] = {"MMD_AT_PLUS_A","NATURAL","MMD_ATA","COLAMD"}; 447 char *prtype[] = {"LargeDiag","NATURAL"}; 448 PetscFunctionBegin; 449 450 /* Create the factorization matrix */ 451 ierr = MatCreate(A->comm,A->m,A->n,M,N,&B);CHKERRQ(ierr); 452 ierr = MatSetType(B,MATSUPERLU_DIST);CHKERRQ(ierr); 453 ierr = MatSeqAIJSetPreallocation(B,0,PETSC_NULL); 454 ierr = MatMPIAIJSetPreallocation(B,0,PETSC_NULL,0,PETSC_NULL);CHKERRQ(ierr); 455 456 B->ops->solve = MatSolve_MPIAIJ_SuperLU_DIST; 457 B->factor = FACTOR_LU; 458 459 lu = (Mat_MPIAIJ_SuperLU_DIST*)(B->spptr); 460 461 /* Set the input options */ 462 set_default_options(&options); 463 lu->MatInputMode = GLOBAL; 464 ierr = MPI_Comm_dup(A->comm,&(lu->comm_superlu));CHKERRQ(ierr); 465 466 ierr = MPI_Comm_size(A->comm,&size);CHKERRQ(ierr); 467 lu->nprow = size/2; /* Default process rows. */ 468 if (lu->nprow == 0) lu->nprow = 1; 469 lu->npcol = size/lu->nprow; /* Default process columns. */ 470 471 ierr = PetscOptionsBegin(A->comm,A->prefix,"SuperLU_Dist Options","Mat");CHKERRQ(ierr); 472 473 ierr = PetscOptionsInt("-mat_superlu_dist_r","Number rows in processor partition","None",lu->nprow,&lu->nprow,PETSC_NULL);CHKERRQ(ierr); 474 ierr = PetscOptionsInt("-mat_superlu_dist_c","Number columns in processor partition","None",lu->npcol,&lu->npcol,PETSC_NULL);CHKERRQ(ierr); 475 if (size != lu->nprow * lu->npcol) SETERRQ(1,"Number of processes should be equal to nprow*npcol"); 476 477 ierr = PetscOptionsInt("-mat_superlu_dist_matinput","Matrix input mode (0: GLOBAL; 1: DISTRIBUTED)","None",lu->MatInputMode,&lu->MatInputMode,PETSC_NULL);CHKERRQ(ierr); 478 if(lu->MatInputMode == DISTRIBUTED && size == 1) lu->MatInputMode = GLOBAL; 479 480 ierr = PetscOptionsLogical("-mat_superlu_dist_equil","Equilibrate matrix","None",PETSC_TRUE,&flg,0);CHKERRQ(ierr); 481 if (!flg) { 482 options.Equil = NO; 483 } 484 485 ierr = PetscOptionsEList("-mat_superlu_dist_rowperm","Row permutation","None",prtype,2,prtype[0],buff,32,&flg);CHKERRQ(ierr); 486 while (flg) { 487 ierr = PetscStrcmp(buff,"LargeDiag",&flg);CHKERRQ(ierr); 488 if (flg) { 489 options.RowPerm = LargeDiag; 490 break; 491 } 492 ierr = PetscStrcmp(buff,"NATURAL",&flg);CHKERRQ(ierr); 493 if (flg) { 494 options.RowPerm = NOROWPERM; 495 break; 496 } 497 SETERRQ1(1,"Unknown row permutation %s",buff); 498 } 499 500 ierr = PetscOptionsEList("-mat_superlu_dist_colperm","Column permutation","None",ptype,4,ptype[0],buff,32,&flg);CHKERRQ(ierr); 501 while (flg) { 502 ierr = PetscStrcmp(buff,"MMD_AT_PLUS_A",&flg);CHKERRQ(ierr); 503 if (flg) { 504 options.ColPerm = MMD_AT_PLUS_A; 505 break; 506 } 507 ierr = PetscStrcmp(buff,"NATURAL",&flg);CHKERRQ(ierr); 508 if (flg) { 509 options.ColPerm = NATURAL; 510 break; 511 } 512 ierr = PetscStrcmp(buff,"MMD_ATA",&flg);CHKERRQ(ierr); 513 if (flg) { 514 options.ColPerm = MMD_ATA; 515 break; 516 } 517 ierr = PetscStrcmp(buff,"COLAMD",&flg);CHKERRQ(ierr); 518 if (flg) { 519 options.ColPerm = COLAMD; 520 break; 521 } 522 SETERRQ1(1,"Unknown column permutation %s",buff); 523 } 524 525 ierr = PetscOptionsLogical("-mat_superlu_dist_replacetinypivot","Replace tiny pivots","None",PETSC_TRUE,&flg,0);CHKERRQ(ierr); 526 if (!flg) { 527 options.ReplaceTinyPivot = NO; 528 } 529 530 options.IterRefine = NOREFINE; 531 ierr = PetscOptionsLogical("-mat_superlu_dist_iterrefine","Use iterative refinement","None",PETSC_FALSE,&flg,0);CHKERRQ(ierr); 532 if (flg) { 533 options.IterRefine = DOUBLE; 534 } 535 536 if (PetscLogPrintInfo) { 537 lu->StatPrint = (int)PETSC_TRUE; 538 } else { 539 lu->StatPrint = (int)PETSC_FALSE; 540 } 541 ierr = PetscOptionsLogical("-mat_superlu_dist_statprint","Print factorization information","None", 542 (PetscTruth)lu->StatPrint,(PetscTruth*)&lu->StatPrint,0);CHKERRQ(ierr); 543 PetscOptionsEnd(); 544 545 /* Initialize the SuperLU process grid. */ 546 superlu_gridinit(lu->comm_superlu, lu->nprow, lu->npcol, &lu->grid); 547 548 /* Initialize ScalePermstruct and LUstruct. */ 549 ScalePermstructInit(M, N, &lu->ScalePermstruct); 550 LUstructInit(M, N, &lu->LUstruct); 551 552 lu->options = options; 553 lu->flg = DIFFERENT_NONZERO_PATTERN; 554 lu->CleanUpSuperLUDist = PETSC_TRUE; 555 *F = B; 556 PetscFunctionReturn(0); 557 } 558 559 #undef __FUNCT__ 560 #define __FUNCT__ "MatUseSuperLU_DIST_MPIAIJ" 561 int MatUseSuperLU_DIST_MPIAIJ(Mat A) 562 { 563 PetscFunctionBegin; 564 A->ops->lufactorsymbolic = MatLUFactorSymbolic_MPIAIJ_SuperLU_DIST; 565 A->ops->lufactornumeric = MatLUFactorNumeric_MPIAIJ_SuperLU_DIST; 566 PetscFunctionReturn(0); 567 } 568 569 #undef __FUNCT__ 570 #define __FUNCT__ "MatMPIAIJFactorInfo_SuperLu" 571 int MatMPIAIJFactorInfo_SuperLu(Mat A,PetscViewer viewer) 572 { 573 Mat_MPIAIJ_SuperLU_DIST *lu= (Mat_MPIAIJ_SuperLU_DIST*)A->spptr; 574 superlu_options_t options; 575 int ierr; 576 char *colperm; 577 578 PetscFunctionBegin; 579 /* check if matrix is superlu_dist type */ 580 if (A->ops->solve != MatSolve_MPIAIJ_SuperLU_DIST) PetscFunctionReturn(0); 581 582 options = lu->options; 583 ierr = PetscViewerASCIIPrintf(viewer,"SuperLU_DIST run parameters:\n");CHKERRQ(ierr); 584 ierr = PetscViewerASCIIPrintf(viewer," Equilibrate matrix %s \n",(options.Equil != NO) ? "true": "false");CHKERRQ(ierr); 585 ierr = PetscViewerASCIIPrintf(viewer," Replace tiny pivots %s \n",(options.ReplaceTinyPivot != NO) ? "true": "false");CHKERRQ(ierr); 586 ierr = PetscViewerASCIIPrintf(viewer," Use iterative refinement %s \n",(options.IterRefine == DOUBLE) ? "true": "false");CHKERRQ(ierr); 587 ierr = PetscViewerASCIIPrintf(viewer," Processors in row %d col partition %d \n",lu->nprow,lu->npcol);CHKERRQ(ierr); 588 ierr = PetscViewerASCIIPrintf(viewer," Row permutation %s \n",(options.RowPerm == NOROWPERM) ? "NATURAL": "LargeDiag");CHKERRQ(ierr); 589 if (options.ColPerm == NATURAL) { 590 colperm = "NATURAL"; 591 } else if (options.ColPerm == MMD_AT_PLUS_A) { 592 colperm = "MMD_AT_PLUS_A"; 593 } else if (options.ColPerm == MMD_ATA) { 594 colperm = "MMD_ATA"; 595 } else if (options.ColPerm == COLAMD) { 596 colperm = "COLAMD"; 597 } else { 598 SETERRQ(1,"Unknown column permutation"); 599 } 600 ierr = PetscViewerASCIIPrintf(viewer," Column permutation %s \n",colperm);CHKERRQ(ierr); 601 PetscFunctionReturn(0); 602 } 603 604 EXTERN_C_BEGIN 605 #undef __FUNCT__ 606 #define __FUNCT__ "MatCreate_MPIAIJ_SuperLU_DIST" 607 int MatCreate_MPIAIJ_SuperLU_DIST(Mat A) { 608 int ierr,size; 609 MPI_Comm comm; 610 Mat_MPIAIJ_SuperLU_DIST *lu; 611 612 PetscFunctionBegin; 613 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 614 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr);CHKERRQ(ierr); 615 if (size == 1) { 616 ierr = MatSetType(A,MATSEQAIJ);CHKERRQ(ierr); 617 } else { 618 ierr = MatSetType(A,MATMPIAIJ);CHKERRQ(ierr); 619 } 620 ierr = MatUseSuperLU_DIST_MPIAIJ(A);CHKERRQ(ierr); 621 622 ierr = PetscNew(Mat_MPIAIJ_SuperLU_DIST,&lu);CHKERRQ(ierr); 623 lu->MatView = A->ops->view; 624 lu->MatAssemblyEnd = A->ops->assemblyend; 625 lu->MatDestroy = A->ops->destroy; 626 lu->CleanUpSuperLUDist = PETSC_FALSE; 627 A->spptr = (void*)lu; 628 A->ops->view = MatView_MPIAIJ_SuperLU_DIST; 629 A->ops->assemblyend = MatAssemblyEnd_MPIAIJ_SuperLU_DIST; 630 A->ops->destroy = MatDestroy_MPIAIJ_SuperLU_DIST; 631 PetscFunctionReturn(0); 632 } 633 EXTERN_C_END 634 635 EXTERN_C_BEGIN 636 #undef __FUNCT__ 637 #define __FUNCT__ "MatLoad_MPIAIJ_SuperLU_DIST" 638 int MatLoad_MPIAIJ_SuperLU_DIST(PetscViewer viewer,MatType type,Mat *A) { 639 int ierr,size,(*r)(PetscViewer,MatType,Mat*); 640 MPI_Comm comm; 641 642 PetscFunctionBegin; 643 ierr = PetscObjectGetComm((PetscObject)viewer,&comm);CHKERRQ(ierr); 644 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 645 if (size == 1) { 646 ierr = PetscFListFind(comm,MatLoadList,MATSEQAIJ,(void(**)(void))&r);CHKERRQ(ierr); 647 } else { 648 ierr = PetscFListFind(comm,MatLoadList,MATMPIAIJ,(void(**)(void))&r);CHKERRQ(ierr); 649 } 650 ierr = (*r)(viewer,type,A);CHKERRQ(ierr); 651 PetscFunctionReturn(0); 652 } 653 EXTERN_C_END 654