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 SuperLU - an external package providing director solvers for MPIAIJ 14 matrices. 15 16 Notes: PETSc does not natively support PCLU for MPIAIJ 17 matrices. However, if SuperLU_dist is installed (see the manual for 18 instructions on how to declare the existence of external packages), 19 it can be specified as the solver to be used by a call to 20 MatUseSuperLU_DIST_MPIAIJ(). This is only supported for 21 double precision real. 22 23 Options Database Keys: 24 + -mat_aij_superlu_dist_r <n> : number of rows in processor partition 25 . -mat_aij_superlu_dist_c <n> : number of columns in processor partition 26 . -mat_aij_superlu_dist_matinput 0|1 : matrix input mode; 0=global, 1=distributed 27 . -mat_aij_superlu_dist_equil :, equilibrate the matrix 28 . -mat_aij_superlu_dist_rowperm LargeDiag|NATURAL : row permutation 29 . -mat_aij_superlu_dist_colperm MMD_AT_PLUS_A|MMD_ATA|COLAMD|NATURAL : column permutation 30 . -mat_aij_superlu_dist_replacetinypivot : replace tiny pivots 31 . -mat_aij_superlu_dist_iterrefine : use iterative refinement 32 - -mat_aij_superlu_dist_statprint : print factorization information 33 34 .seealso: MatUseSuperLU_DIST_MPIAIJ(), PCLU 35 M*/ 36 37 EXTERN_C_BEGIN 38 #if defined(PETSC_USE_COMPLEX) 39 #include "superlu_zdefs.h" 40 #else 41 #include "superlu_ddefs.h" 42 #endif 43 EXTERN_C_END 44 45 typedef enum { GLOBAL,DISTRIBUTED 46 } SuperLU_MatInputMode; 47 48 typedef struct { 49 int_t nprow,npcol,*row,*col; 50 gridinfo_t grid; 51 superlu_options_t options; 52 SuperMatrix A_sup; 53 ScalePermstruct_t ScalePermstruct; 54 LUstruct_t LUstruct; 55 int StatPrint; 56 int MatInputMode; 57 SOLVEstruct_t SOLVEstruct; 58 MatStructure flg; 59 MPI_Comm comm_superlu; 60 #if defined(PETSC_USE_COMPLEX) 61 doublecomplex *val; 62 #else 63 double *val; 64 #endif 65 66 /* A few function pointers for inheritance */ 67 int (*MatView)(Mat,PetscViewer); 68 int (*MatAssemblyEnd)(Mat,MatAssemblyType); 69 int (*MatDestroy)(Mat); 70 71 /* Flag to clean up (non-global) SuperLU objects during Destroy */ 72 PetscTruth CleanUpSuperLUDist; 73 } Mat_MPIAIJ_SuperLU_DIST; 74 75 #undef __FUNCT__ 76 #define __FUNCT__ "MatDestroy_MPIAIJ_SuperLU_DIST" 77 int MatDestroy_MPIAIJ_SuperLU_DIST(Mat A) 78 { 79 Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data; 80 Mat_MPIAIJ_SuperLU_DIST *lu = (Mat_MPIAIJ_SuperLU_DIST*)A->spptr; 81 int ierr, size=a->size,(*destroy)(Mat); 82 83 PetscFunctionBegin; 84 if (lu->CleanUpSuperLUDist) { 85 /* Deallocate SuperLU_DIST storage */ 86 if (lu->MatInputMode == GLOBAL) { 87 Destroy_CompCol_Matrix_dist(&lu->A_sup); 88 } else { 89 Destroy_CompRowLoc_Matrix_dist(&lu->A_sup); 90 if ( lu->options.SolveInitialized ) { 91 #if defined(PETSC_USE_COMPLEX) 92 zSolveFinalize(&lu->options, &lu->SOLVEstruct); 93 #else 94 dSolveFinalize(&lu->options, &lu->SOLVEstruct); 95 #endif 96 } 97 } 98 Destroy_LU(A->N, &lu->grid, &lu->LUstruct); 99 ScalePermstructFree(&lu->ScalePermstruct); 100 LUstructFree(&lu->LUstruct); 101 102 /* Release the SuperLU_DIST process grid. */ 103 superlu_gridexit(&lu->grid); 104 105 ierr = MPI_Comm_free(&(lu->comm_superlu));CHKERRQ(ierr); 106 } 107 destroy = lu->MatDestroy; 108 ierr = PetscFree(lu);CHKERRQ(ierr); 109 ierr = (*destroy)(A);CHKERRQ(ierr); 110 111 PetscFunctionReturn(0); 112 } 113 114 #undef __FUNCT__ 115 #define __FUNCT__ "MatView_MPIAIJ_Spooles_DIST" 116 int MatView_MPIAIJ_SuperLU_DIST(Mat A,PetscViewer viewer) 117 { 118 int ierr; 119 PetscTruth isascii; 120 PetscViewerFormat format; 121 Mat_MPIAIJ_SuperLU_DIST *lu=(Mat_MPIAIJ_SuperLU_DIST*)(A->spptr); 122 123 PetscFunctionBegin; 124 ierr = (*lu->MatView)(A,viewer);CHKERRQ(ierr); 125 126 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&isascii);CHKERRQ(ierr); 127 if (isascii) { 128 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 129 if (format == PETSC_VIEWER_ASCII_FACTOR_INFO) { 130 ierr = MatMPIAIJFactorInfo_SuperLu(A,viewer);CHKERRQ(ierr); 131 } 132 } 133 PetscFunctionReturn(0); 134 } 135 136 #undef __FUNCT__ 137 #define __FUNCT__ "MatAssemblyEnd_MPIAIJ_SuperLU_DIST" 138 int MatAssemblyEnd_MPIAIJ_SuperLU_DIST(Mat A,MatAssemblyType mode) { 139 int ierr; 140 Mat_MPIAIJ_SuperLU_DIST *lu=(Mat_MPIAIJ_SuperLU_DIST*)(A->spptr); 141 142 PetscFunctionBegin; 143 ierr = (*lu->MatAssemblyEnd)(A,mode);CHKERRQ(ierr); 144 ierr = MatUseSuperLU_DIST_MPIAIJ(A);CHKERRQ(ierr); 145 PetscFunctionReturn(0); 146 } 147 148 #undef __FUNCT__ 149 #define __FUNCT__ "MatSolve_MPIAIJ_SuperLU_DIST" 150 int MatSolve_MPIAIJ_SuperLU_DIST(Mat A,Vec b_mpi,Vec x) 151 { 152 Mat_MPIAIJ *aa = (Mat_MPIAIJ*)A->data; 153 Mat_MPIAIJ_SuperLU_DIST *lu = (Mat_MPIAIJ_SuperLU_DIST*)A->spptr; 154 int ierr, size=aa->size; 155 int m=A->M, N=A->N; 156 SuperLUStat_t stat; 157 double berr[1]; 158 PetscScalar *bptr; 159 int info, nrhs=1; 160 Vec x_seq; 161 IS iden; 162 VecScatter scat; 163 PetscLogDouble time0,time,time_min,time_max; 164 165 PetscFunctionBegin; 166 if (size > 1) { 167 if (lu->MatInputMode == GLOBAL) { /* global mat input, convert b to x_seq */ 168 ierr = VecCreateSeq(PETSC_COMM_SELF,N,&x_seq);CHKERRQ(ierr); 169 ierr = ISCreateStride(PETSC_COMM_SELF,N,0,1,&iden);CHKERRQ(ierr); 170 ierr = VecScatterCreate(b_mpi,iden,x_seq,iden,&scat);CHKERRQ(ierr); 171 ierr = ISDestroy(iden);CHKERRQ(ierr); 172 173 ierr = VecScatterBegin(b_mpi,x_seq,INSERT_VALUES,SCATTER_FORWARD,scat);CHKERRQ(ierr); 174 ierr = VecScatterEnd(b_mpi,x_seq,INSERT_VALUES,SCATTER_FORWARD,scat);CHKERRQ(ierr); 175 ierr = VecGetArray(x_seq,&bptr);CHKERRQ(ierr); 176 } else { /* distributed mat input */ 177 ierr = VecCopy(b_mpi,x);CHKERRQ(ierr); 178 ierr = VecGetArray(x,&bptr);CHKERRQ(ierr); 179 } 180 } else { /* size == 1 */ 181 ierr = VecCopy(b_mpi,x);CHKERRQ(ierr); 182 ierr = VecGetArray(x,&bptr);CHKERRQ(ierr); 183 } 184 185 lu->options.Fact = FACTORED; /* The factored form of A is supplied. Local option used by this func. only.*/ 186 187 PStatInit(&stat); /* Initialize the statistics variables. */ 188 if (lu->StatPrint) { 189 ierr = MPI_Barrier(A->comm);CHKERRQ(ierr); /* to be removed */ 190 ierr = PetscGetTime(&time0);CHKERRQ(ierr); /* to be removed */ 191 } 192 if (lu->MatInputMode == GLOBAL) { 193 #if defined(PETSC_USE_COMPLEX) 194 pzgssvx_ABglobal(&lu->options, &lu->A_sup, &lu->ScalePermstruct,(doublecomplex*)bptr, m, nrhs, 195 &lu->grid, &lu->LUstruct, berr, &stat, &info); 196 #else 197 pdgssvx_ABglobal(&lu->options, &lu->A_sup, &lu->ScalePermstruct,bptr, m, nrhs, 198 &lu->grid, &lu->LUstruct, berr, &stat, &info); 199 #endif 200 } else { /* distributed mat input */ 201 #if defined(PETSC_USE_COMPLEX) 202 pzgssvx(&lu->options, &lu->A_sup, &lu->ScalePermstruct, (doublecomplex*)bptr, A->M, nrhs, &lu->grid, 203 &lu->LUstruct, &lu->SOLVEstruct, berr, &stat, &info); 204 if (info) SETERRQ1(1,"pzgssvx fails, info: %d\n",info); 205 #else 206 pdgssvx(&lu->options, &lu->A_sup, &lu->ScalePermstruct, bptr, A->M, nrhs, &lu->grid, 207 &lu->LUstruct, &lu->SOLVEstruct, berr, &stat, &info); 208 if (info) SETERRQ1(1,"pdgssvx fails, info: %d\n",info); 209 #endif 210 } 211 if (lu->StatPrint) { 212 ierr = PetscGetTime(&time);CHKERRQ(ierr); /* to be removed */ 213 PStatPrint(&lu->options, &stat, &lu->grid); /* Print the statistics. */ 214 } 215 PStatFree(&stat); 216 217 if (size > 1) { 218 if (lu->MatInputMode == GLOBAL){ /* convert seq x to mpi x */ 219 ierr = VecRestoreArray(x_seq,&bptr);CHKERRQ(ierr); 220 ierr = VecScatterBegin(x_seq,x,INSERT_VALUES,SCATTER_REVERSE,scat);CHKERRQ(ierr); 221 ierr = VecScatterEnd(x_seq,x,INSERT_VALUES,SCATTER_REVERSE,scat);CHKERRQ(ierr); 222 ierr = VecScatterDestroy(scat);CHKERRQ(ierr); 223 ierr = VecDestroy(x_seq);CHKERRQ(ierr); 224 } else { 225 ierr = VecRestoreArray(x,&bptr);CHKERRQ(ierr); 226 } 227 } else { 228 ierr = VecRestoreArray(x,&bptr);CHKERRQ(ierr); 229 } 230 if (lu->StatPrint) { 231 time0 = time - time0; 232 ierr = MPI_Reduce(&time0,&time_max,1,MPI_DOUBLE,MPI_MAX,0,A->comm);CHKERRQ(ierr); 233 ierr = MPI_Reduce(&time0,&time_min,1,MPI_DOUBLE,MPI_MIN,0,A->comm);CHKERRQ(ierr); 234 ierr = MPI_Reduce(&time0,&time,1,MPI_DOUBLE,MPI_SUM,0,A->comm);CHKERRQ(ierr); 235 time = time/size; /* average time */ 236 ierr = PetscPrintf(A->comm, " Time for superlu_dist solve (max/min/avg): %g / %g / %g\n\n",time_max,time_min,time);CHKERRQ(ierr); 237 } 238 PetscFunctionReturn(0); 239 } 240 241 #undef __FUNCT__ 242 #define __FUNCT__ "MatLUFactorNumeric_MPIAIJ_SuperLU_DIST" 243 int MatLUFactorNumeric_MPIAIJ_SuperLU_DIST(Mat A,Mat *F) 244 { 245 Mat_MPIAIJ *fac = (Mat_MPIAIJ*)(*F)->data,*mat; 246 Mat *tseq,A_seq = PETSC_NULL; 247 Mat_SeqAIJ *aa,*bb; 248 Mat_MPIAIJ_SuperLU_DIST *lu = (Mat_MPIAIJ_SuperLU_DIST*)(*F)->spptr; 249 int M=A->M,N=A->N,info,ierr,size=fac->size,i,*ai,*aj,*bi,*bj,nz,rstart,*garray, 250 m=A->m, irow,colA_start,j,jcol,jB,countA,countB,*bjj,*ajj; 251 SuperLUStat_t stat; 252 double *berr=0; 253 IS isrow; 254 PetscLogDouble time0[2],time[2],time_min[2],time_max[2]; 255 #if defined(PETSC_USE_COMPLEX) 256 doublecomplex *av, *bv; 257 #else 258 double *av, *bv; 259 #endif 260 261 PetscFunctionBegin; 262 if (lu->StatPrint) { 263 ierr = MPI_Barrier(A->comm);CHKERRQ(ierr); 264 ierr = PetscGetTime(&time0[0]);CHKERRQ(ierr); 265 } 266 267 if (lu->MatInputMode == GLOBAL) { /* global mat input */ 268 if (size > 1) { /* convert mpi A to seq mat A */ 269 ierr = ISCreateStride(PETSC_COMM_SELF,M,0,1,&isrow); CHKERRQ(ierr); 270 ierr = MatGetSubMatrices(A,1,&isrow,&isrow,MAT_INITIAL_MATRIX,&tseq); CHKERRQ(ierr); 271 ierr = ISDestroy(isrow);CHKERRQ(ierr); 272 273 A_seq = *tseq; 274 ierr = PetscFree(tseq);CHKERRQ(ierr); 275 aa = (Mat_SeqAIJ*)A_seq->data; 276 } else { 277 aa = (Mat_SeqAIJ*)A->data; 278 } 279 280 /* Allocate storage, then convert Petsc NR matrix to SuperLU_DIST NC */ 281 if (lu->flg == DIFFERENT_NONZERO_PATTERN) {/* first numeric factorization */ 282 #if defined(PETSC_USE_COMPLEX) 283 zallocateA_dist(N, aa->nz, &lu->val, &lu->col, &lu->row); 284 #else 285 dallocateA_dist(N, aa->nz, &lu->val, &lu->col, &lu->row); 286 #endif 287 } else { /* successive numeric factorization, sparsity pattern is reused. */ 288 Destroy_CompCol_Matrix_dist(&lu->A_sup); 289 Destroy_LU(N, &lu->grid, &lu->LUstruct); 290 lu->options.Fact = SamePattern; 291 } 292 #if defined(PETSC_USE_COMPLEX) 293 zCompRow_to_CompCol_dist(M,N,aa->nz,(doublecomplex*)aa->a,aa->j,aa->i,&lu->val,&lu->col, &lu->row); 294 #else 295 dCompRow_to_CompCol_dist(M,N,aa->nz,aa->a,aa->j,aa->i,&lu->val, &lu->col, &lu->row); 296 #endif 297 298 /* Create compressed column matrix A_sup. */ 299 #if defined(PETSC_USE_COMPLEX) 300 zCreate_CompCol_Matrix_dist(&lu->A_sup, M, N, aa->nz, lu->val, lu->col, lu->row, SLU_NC, SLU_Z, SLU_GE); 301 #else 302 dCreate_CompCol_Matrix_dist(&lu->A_sup, M, N, aa->nz, lu->val, lu->col, lu->row, SLU_NC, SLU_D, SLU_GE); 303 #endif 304 } else { /* distributed mat input */ 305 mat = (Mat_MPIAIJ*)A->data; 306 aa=(Mat_SeqAIJ*)(mat->A)->data; 307 bb=(Mat_SeqAIJ*)(mat->B)->data; 308 ai=aa->i; aj=aa->j; 309 bi=bb->i; bj=bb->j; 310 #if defined(PETSC_USE_COMPLEX) 311 av=(doublecomplex*)aa->a; 312 bv=(doublecomplex*)bb->a; 313 #else 314 av=aa->a; 315 bv=bb->a; 316 #endif 317 rstart = mat->rstart; 318 nz = aa->nz + bb->nz; 319 garray = mat->garray; 320 rstart = mat->rstart; 321 322 if (lu->flg == DIFFERENT_NONZERO_PATTERN) {/* first numeric factorization */ 323 #if defined(PETSC_USE_COMPLEX) 324 zallocateA_dist(m, nz, &lu->val, &lu->col, &lu->row); 325 #else 326 dallocateA_dist(m, nz, &lu->val, &lu->col, &lu->row); 327 #endif 328 } else { /* successive numeric factorization, sparsity pattern and perm_c are reused. */ 329 /* Destroy_CompRowLoc_Matrix_dist(&lu->A_sup); */ /* crash! */ 330 Destroy_LU(N, &lu->grid, &lu->LUstruct); 331 lu->options.Fact = SamePattern; 332 } 333 nz = 0; jB = 0; irow = mat->rstart; 334 for ( i=0; i<m; i++ ) { 335 lu->row[i] = nz; 336 countA = ai[i+1] - ai[i]; 337 countB = bi[i+1] - bi[i]; 338 ajj = aj + ai[i]; /* ptr to the beginning of this row */ 339 bjj = bj + bi[i]; 340 341 /* B part, smaller col index */ 342 colA_start = mat->rstart + ajj[0]; /* the smallest global col index of A */ 343 for (j=0; j<countB; j++){ 344 jcol = garray[bjj[j]]; 345 if (jcol > colA_start) { 346 jB = j; 347 break; 348 } 349 lu->col[nz] = jcol; 350 lu->val[nz++] = *bv++; 351 if (j==countB-1) jB = countB; 352 } 353 354 /* A part */ 355 for (j=0; j<countA; j++){ 356 lu->col[nz] = mat->rstart + ajj[j]; 357 lu->val[nz++] = *av++; 358 } 359 360 /* B part, larger col index */ 361 for (j=jB; j<countB; j++){ 362 lu->col[nz] = garray[bjj[j]]; 363 lu->val[nz++] = *bv++; 364 } 365 } 366 lu->row[m] = nz; 367 #if defined(PETSC_USE_COMPLEX) 368 zCreate_CompRowLoc_Matrix_dist(&lu->A_sup, M, N, nz, m, rstart, 369 lu->val, lu->col, lu->row, SLU_NR_loc, SLU_Z, SLU_GE); 370 #else 371 dCreate_CompRowLoc_Matrix_dist(&lu->A_sup, M, N, nz, m, rstart, 372 lu->val, lu->col, lu->row, SLU_NR_loc, SLU_D, SLU_GE); 373 #endif 374 } 375 if (lu->StatPrint) { 376 ierr = PetscGetTime(&time[0]);CHKERRQ(ierr); 377 time0[0] = time[0] - time0[0]; 378 } 379 380 /* Factor the matrix. */ 381 PStatInit(&stat); /* Initialize the statistics variables. */ 382 383 if (lu->StatPrint) { 384 ierr = MPI_Barrier(A->comm);CHKERRQ(ierr); 385 ierr = PetscGetTime(&time0[1]);CHKERRQ(ierr); 386 } 387 388 if (lu->MatInputMode == GLOBAL) { /* global mat input */ 389 #if defined(PETSC_USE_COMPLEX) 390 pzgssvx_ABglobal(&lu->options, &lu->A_sup, &lu->ScalePermstruct, 0, M, 0, 391 &lu->grid, &lu->LUstruct, berr, &stat, &info); 392 #else 393 pdgssvx_ABglobal(&lu->options, &lu->A_sup, &lu->ScalePermstruct, 0, M, 0, 394 &lu->grid, &lu->LUstruct, berr, &stat, &info); 395 #endif 396 } else { /* distributed mat input */ 397 #if defined(PETSC_USE_COMPLEX) 398 pzgssvx(&lu->options, &lu->A_sup, &lu->ScalePermstruct, 0, M, 0, &lu->grid, 399 &lu->LUstruct, &lu->SOLVEstruct, berr, &stat, &info); 400 if (info) SETERRQ1(1,"pzgssvx fails, info: %d\n",info); 401 #else 402 pdgssvx(&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,"pdgssvx fails, info: %d\n",info); 405 #endif 406 } 407 if (lu->StatPrint) { 408 ierr = PetscGetTime(&time[1]);CHKERRQ(ierr); /* to be removed */ 409 time0[1] = time[1] - time0[1]; 410 if (lu->StatPrint) PStatPrint(&lu->options, &stat, &lu->grid); /* Print the statistics. */ 411 } 412 PStatFree(&stat); 413 414 if (lu->MatInputMode == GLOBAL && size > 1){ 415 ierr = MatDestroy(A_seq);CHKERRQ(ierr); 416 } 417 418 if (lu->StatPrint) { 419 ierr = MPI_Reduce(time0,time_max,2,MPI_DOUBLE,MPI_MAX,0,A->comm); 420 ierr = MPI_Reduce(time0,time_min,2,MPI_DOUBLE,MPI_MIN,0,A->comm); 421 ierr = MPI_Reduce(time0,time,2,MPI_DOUBLE,MPI_SUM,0,A->comm); 422 for (i=0; i<2; i++) time[i] = time[i]/size; /* average time */ 423 ierr = PetscPrintf(A->comm, " Time for mat conversion (max/min/avg): %g / %g / %g\n",time_max[0],time_min[0],time[0]); 424 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]); 425 } 426 (*F)->assembled = PETSC_TRUE; 427 lu->flg = SAME_NONZERO_PATTERN; 428 PetscFunctionReturn(0); 429 } 430 431 /* Note the Petsc r and c permutations are ignored */ 432 #undef __FUNCT__ 433 #define __FUNCT__ "MatLUFactorSymbolic_MPIAIJ_SuperLU_DIST" 434 int MatLUFactorSymbolic_MPIAIJ_SuperLU_DIST(Mat A,IS r,IS c,MatFactorInfo *info,Mat *F) 435 { 436 Mat B; 437 Mat_MPIAIJ_SuperLU_DIST *lu; 438 int ierr,M=A->M,N=A->N,size; 439 superlu_options_t options; 440 char buff[32]; 441 PetscTruth flg; 442 char *ptype[] = {"MMD_AT_PLUS_A","NATURAL","MMD_ATA","COLAMD"}; 443 char *prtype[] = {"LargeDiag","NATURAL"}; 444 PetscFunctionBegin; 445 446 /* Create the factorization matrix */ 447 ierr = MatCreate(A->comm,A->m,A->n,M,N,&B);CHKERRQ(ierr); 448 ierr = MatSetType(B,MATSUPERLUDIST);CHKERRQ(ierr); 449 ierr = MatSeqAIJSetPreallocation(B,0,PETSC_NULL); 450 ierr = MatMPIAIJSetPreallocation(B,0,PETSC_NULL,0,PETSC_NULL);CHKERRQ(ierr); 451 452 B->ops->solve = MatSolve_MPIAIJ_SuperLU_DIST; 453 B->factor = FACTOR_LU; 454 455 lu = (Mat_MPIAIJ_SuperLU_DIST*)(B->spptr); 456 457 /* Set the input options */ 458 set_default_options(&options); 459 lu->MatInputMode = GLOBAL; 460 ierr = MPI_Comm_dup(A->comm,&(lu->comm_superlu));CHKERRQ(ierr); 461 462 ierr = MPI_Comm_size(A->comm,&size);CHKERRQ(ierr); 463 lu->nprow = size/2; /* Default process rows. */ 464 if (lu->nprow == 0) lu->nprow = 1; 465 lu->npcol = size/lu->nprow; /* Default process columns. */ 466 467 ierr = PetscOptionsBegin(A->comm,A->prefix,"SuperLU_Dist Options","Mat");CHKERRQ(ierr); 468 469 ierr = PetscOptionsInt("-mat_aij_superlu_dist_r","Number rows in processor partition","None",lu->nprow,&lu->nprow,PETSC_NULL);CHKERRQ(ierr); 470 ierr = PetscOptionsInt("-mat_aij_superlu_dist_c","Number columns in processor partition","None",lu->npcol,&lu->npcol,PETSC_NULL);CHKERRQ(ierr); 471 if (size != lu->nprow * lu->npcol) SETERRQ(1,"Number of processes should be equal to nprow*npcol"); 472 473 ierr = PetscOptionsInt("-mat_aij_superlu_dist_matinput","Matrix input mode (0: GLOBAL; 1: DISTRIBUTED)","None",lu->MatInputMode,&lu->MatInputMode,PETSC_NULL);CHKERRQ(ierr); 474 if(lu->MatInputMode == DISTRIBUTED && size == 1) lu->MatInputMode = GLOBAL; 475 476 ierr = PetscOptionsLogical("-mat_aij_superlu_dist_equil","Equilibrate matrix","None",PETSC_TRUE,&flg,0);CHKERRQ(ierr); 477 if (!flg) { 478 options.Equil = NO; 479 } 480 481 ierr = PetscOptionsEList("-mat_aij_superlu_dist_rowperm","Row permutation","None",prtype,2,prtype[0],buff,32,&flg);CHKERRQ(ierr); 482 while (flg) { 483 ierr = PetscStrcmp(buff,"LargeDiag",&flg);CHKERRQ(ierr); 484 if (flg) { 485 options.RowPerm = LargeDiag; 486 break; 487 } 488 ierr = PetscStrcmp(buff,"NATURAL",&flg);CHKERRQ(ierr); 489 if (flg) { 490 options.RowPerm = NOROWPERM; 491 break; 492 } 493 SETERRQ1(1,"Unknown row permutation %s",buff); 494 } 495 496 ierr = PetscOptionsEList("-mat_aij_superlu_dist_colperm","Column permutation","None",ptype,4,ptype[0],buff,32,&flg);CHKERRQ(ierr); 497 while (flg) { 498 ierr = PetscStrcmp(buff,"MMD_AT_PLUS_A",&flg);CHKERRQ(ierr); 499 if (flg) { 500 options.ColPerm = MMD_AT_PLUS_A; 501 break; 502 } 503 ierr = PetscStrcmp(buff,"NATURAL",&flg);CHKERRQ(ierr); 504 if (flg) { 505 options.ColPerm = NATURAL; 506 break; 507 } 508 ierr = PetscStrcmp(buff,"MMD_ATA",&flg);CHKERRQ(ierr); 509 if (flg) { 510 options.ColPerm = MMD_ATA; 511 break; 512 } 513 ierr = PetscStrcmp(buff,"COLAMD",&flg);CHKERRQ(ierr); 514 if (flg) { 515 options.ColPerm = COLAMD; 516 break; 517 } 518 SETERRQ1(1,"Unknown column permutation %s",buff); 519 } 520 521 ierr = PetscOptionsLogical("-mat_aij_superlu_dist_replacetinypivot","Replace tiny pivots","None",PETSC_TRUE,&flg,0);CHKERRQ(ierr); 522 if (!flg) { 523 options.ReplaceTinyPivot = NO; 524 } 525 526 options.IterRefine = NOREFINE; 527 ierr = PetscOptionsLogical("-mat_aij_superlu_dist_iterrefine","Use iterative refinement","None",PETSC_FALSE,&flg,0);CHKERRQ(ierr); 528 if (flg) { 529 options.IterRefine = DOUBLE; 530 } 531 532 if (PetscLogPrintInfo) { 533 lu->StatPrint = (int)PETSC_TRUE; 534 } else { 535 lu->StatPrint = (int)PETSC_FALSE; 536 } 537 ierr = PetscOptionsLogical("-mat_aij_superlu_dist_statprint","Print factorization information","None", 538 (PetscTruth)lu->StatPrint,(PetscTruth*)&lu->StatPrint,0);CHKERRQ(ierr); 539 PetscOptionsEnd(); 540 541 /* Initialize the SuperLU process grid. */ 542 superlu_gridinit(lu->comm_superlu, lu->nprow, lu->npcol, &lu->grid); 543 544 /* Initialize ScalePermstruct and LUstruct. */ 545 ScalePermstructInit(M, N, &lu->ScalePermstruct); 546 LUstructInit(M, N, &lu->LUstruct); 547 548 lu->options = options; 549 lu->flg = DIFFERENT_NONZERO_PATTERN; 550 lu->CleanUpSuperLUDist = PETSC_TRUE; 551 *F = B; 552 PetscFunctionReturn(0); 553 } 554 555 #undef __FUNCT__ 556 #define __FUNCT__ "MatUseSuperLU_DIST_MPIAIJ" 557 /*@ 558 MatUseSuperLU_DIST_MPIAIJ -- Let the distributed SuperLU package 559 be the solver used by PCLU 560 561 Input Parameter: 562 . A - the matrix 563 564 Level: intermediate 565 566 .seealso: SuperLU, PCSetType(), PCLU 567 @*/ 568 int MatUseSuperLU_DIST_MPIAIJ(Mat A) 569 { 570 PetscFunctionBegin; 571 A->ops->lufactorsymbolic = MatLUFactorSymbolic_MPIAIJ_SuperLU_DIST; 572 A->ops->lufactornumeric = MatLUFactorNumeric_MPIAIJ_SuperLU_DIST; 573 PetscFunctionReturn(0); 574 } 575 576 #undef __FUNCT__ 577 #define __FUNCT__ "MatMPIAIJFactorInfo_SuperLu" 578 int MatMPIAIJFactorInfo_SuperLu(Mat A,PetscViewer viewer) 579 { 580 Mat_MPIAIJ_SuperLU_DIST *lu= (Mat_MPIAIJ_SuperLU_DIST*)A->spptr; 581 superlu_options_t options; 582 int ierr; 583 char *colperm; 584 585 PetscFunctionBegin; 586 /* check if matrix is superlu_dist type */ 587 if (A->ops->solve != MatSolve_MPIAIJ_SuperLU_DIST) PetscFunctionReturn(0); 588 589 options = lu->options; 590 ierr = PetscViewerASCIIPrintf(viewer,"SuperLU_DIST run parameters:\n");CHKERRQ(ierr); 591 ierr = PetscViewerASCIIPrintf(viewer," Equilibrate matrix %s \n",(options.Equil != NO) ? "true": "false");CHKERRQ(ierr); 592 ierr = PetscViewerASCIIPrintf(viewer," Replace tiny pivots %s \n",(options.ReplaceTinyPivot != NO) ? "true": "false");CHKERRQ(ierr); 593 ierr = PetscViewerASCIIPrintf(viewer," Use iterative refinement %s \n",(options.IterRefine == DOUBLE) ? "true": "false");CHKERRQ(ierr); 594 ierr = PetscViewerASCIIPrintf(viewer," Processors in row %d col partition %d \n",lu->nprow,lu->npcol);CHKERRQ(ierr); 595 ierr = PetscViewerASCIIPrintf(viewer," Row permutation %s \n",(options.RowPerm == NOROWPERM) ? "NATURAL": "LargeDiag");CHKERRQ(ierr); 596 if (options.ColPerm == NATURAL) { 597 colperm = "NATURAL"; 598 } else if (options.ColPerm == MMD_AT_PLUS_A) { 599 colperm = "MMD_AT_PLUS_A"; 600 } else if (options.ColPerm == MMD_ATA) { 601 colperm = "MMD_ATA"; 602 } else if (options.ColPerm == COLAMD) { 603 colperm = "COLAMD"; 604 } else { 605 SETERRQ(1,"Unknown column permutation"); 606 } 607 ierr = PetscViewerASCIIPrintf(viewer," Column permutation %s \n",colperm);CHKERRQ(ierr); 608 PetscFunctionReturn(0); 609 } 610 611 EXTERN_C_BEGIN 612 #undef __FUNCT__ 613 #define __FUNCT__ "MatCreate_MPIAIJ_SuperLU_DIST" 614 int MatCreate_MPIAIJ_SuperLU_DIST(Mat A) { 615 int ierr,size; 616 MPI_Comm comm; 617 Mat_MPIAIJ_SuperLU_DIST *lu; 618 619 PetscFunctionBegin; 620 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 621 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr);CHKERRQ(ierr); 622 if (size == 1) { 623 ierr = MatSetType(A,MATSEQAIJ);CHKERRQ(ierr); 624 } else { 625 ierr = MatSetType(A,MATMPIAIJ);CHKERRQ(ierr); 626 } 627 ierr = MatUseSuperLU_DIST_MPIAIJ(A);CHKERRQ(ierr); 628 629 ierr = PetscNew(Mat_MPIAIJ_SuperLU_DIST,&lu);CHKERRQ(ierr); 630 lu->MatView = A->ops->view; 631 lu->MatAssemblyEnd = A->ops->assemblyend; 632 lu->MatDestroy = A->ops->destroy; 633 lu->CleanUpSuperLUDist = PETSC_FALSE; 634 A->spptr = (void*)lu; 635 A->ops->view = MatView_MPIAIJ_SuperLU_DIST; 636 A->ops->assemblyend = MatAssemblyEnd_MPIAIJ_SuperLU_DIST; 637 A->ops->destroy = MatDestroy_MPIAIJ_SuperLU_DIST; 638 PetscFunctionReturn(0); 639 } 640 EXTERN_C_END 641 642 EXTERN_C_BEGIN 643 #undef __FUNCT__ 644 #define __FUNCT__ "MatLoad_MPIAIJ_SuperLU_DIST" 645 int MatLoad_MPIAIJ_SuperLU_DIST(PetscViewer viewer,MatType type,Mat *A) { 646 int ierr,size,(*r)(PetscViewer,MatType,Mat*); 647 MPI_Comm comm; 648 649 PetscFunctionBegin; 650 ierr = PetscObjectGetComm((PetscObject)viewer,&comm);CHKERRQ(ierr); 651 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 652 if (size == 1) { 653 ierr = PetscFListFind(comm,MatLoadList,MATSEQAIJ,(void(**)(void))&r);CHKERRQ(ierr); 654 } else { 655 ierr = PetscFListFind(comm,MatLoadList,MATMPIAIJ,(void(**)(void))&r);CHKERRQ(ierr); 656 } 657 ierr = (*r)(viewer,type,A);CHKERRQ(ierr); 658 PetscFunctionReturn(0); 659 } 660 EXTERN_C_END 661