Actual source code: superlu_dist.c

petsc-3.13.5 2020-09-01
Report Typos and Errors
  1: /*
  2:         Provides an interface to the SuperLU_DIST sparse solver
  3: */

  5:  #include <../src/mat/impls/aij/seq/aij.h>
  6:  #include <../src/mat/impls/aij/mpi/mpiaij.h>
  7: #include <petscpkg_version.h>

  9: EXTERN_C_BEGIN
 10: #if defined(PETSC_USE_COMPLEX)
 11: #include <superlu_zdefs.h>
 12: #if PETSC_PKG_SUPERLU_DIST_VERSION_GE(6,3,0)
 13: #define LUstructInit zLUstructInit
 14: #define ScalePermstructInit zScalePermstructInit
 15: #define ScalePermstructFree zScalePermstructFree
 16: #define LUstructFree zLUstructFree
 17: #define Destroy_LU zDestroy_LU
 18: #define ScalePermstruct_t zScalePermstruct_t
 19: #define LUstruct_t zLUstruct_t
 20: #define SOLVEstruct_t zSOLVEstruct_t
 21: #endif
 22: #else
 23: #include <superlu_ddefs.h>
 24: #if PETSC_PKG_SUPERLU_DIST_VERSION_GE(6,3,0)
 25: #define LUstructInit dLUstructInit
 26: #define ScalePermstructInit dScalePermstructInit
 27: #define ScalePermstructFree dScalePermstructFree
 28: #define LUstructFree dLUstructFree
 29: #define Destroy_LU dDestroy_LU
 30: #define ScalePermstruct_t dScalePermstruct_t
 31: #define LUstruct_t dLUstruct_t
 32: #define SOLVEstruct_t dSOLVEstruct_t
 33: #endif
 34: #endif
 35: EXTERN_C_END

 37: typedef struct {
 38:   int_t                  nprow,npcol,*row,*col;
 39:   gridinfo_t             grid;
 40:   superlu_dist_options_t options;
 41:   SuperMatrix            A_sup;
 42:   ScalePermstruct_t      ScalePermstruct;
 43:   LUstruct_t             LUstruct;
 44:   int                    StatPrint;
 45:   SOLVEstruct_t          SOLVEstruct;
 46:   fact_t                 FactPattern;
 47:   MPI_Comm               comm_superlu;
 48: #if defined(PETSC_USE_COMPLEX)
 49:   doublecomplex          *val;
 50: #else
 51:   double                 *val;
 52: #endif
 53:   PetscBool              matsolve_iscalled,matmatsolve_iscalled;
 54:   PetscBool              CleanUpSuperLU_Dist;  /* Flag to clean up (non-global) SuperLU objects during Destroy */
 55: } Mat_SuperLU_DIST;


 58: PetscErrorCode MatSuperluDistGetDiagU_SuperLU_DIST(Mat F,PetscScalar *diagU)
 59: {
 60:   Mat_SuperLU_DIST *lu = (Mat_SuperLU_DIST*)F->data;

 63: #if defined(PETSC_USE_COMPLEX)
 64:   PetscStackCall("SuperLU_DIST:pzGetDiagU",pzGetDiagU(F->rmap->N,&lu->LUstruct,&lu->grid,(doublecomplex*)diagU));
 65: #else
 66:   PetscStackCall("SuperLU_DIST:pdGetDiagU",pdGetDiagU(F->rmap->N,&lu->LUstruct,&lu->grid,diagU));
 67: #endif
 68:   return(0);
 69: }

 71: PetscErrorCode MatSuperluDistGetDiagU(Mat F,PetscScalar *diagU)
 72: {

 77:   PetscTryMethod(F,"MatSuperluDistGetDiagU_C",(Mat,PetscScalar*),(F,diagU));
 78:   return(0);
 79: }

 81: static PetscErrorCode MatDestroy_SuperLU_DIST(Mat A)
 82: {
 83:   PetscErrorCode   ierr;
 84:   Mat_SuperLU_DIST *lu = (Mat_SuperLU_DIST*)A->data;

 87:   if (lu->CleanUpSuperLU_Dist) {
 88:     /* Deallocate SuperLU_DIST storage */
 89:     PetscStackCall("SuperLU_DIST:Destroy_CompRowLoc_Matrix_dist",Destroy_CompRowLoc_Matrix_dist(&lu->A_sup));
 90:     if (lu->options.SolveInitialized) {
 91: #if defined(PETSC_USE_COMPLEX)
 92:       PetscStackCall("SuperLU_DIST:zSolveFinalize",zSolveFinalize(&lu->options, &lu->SOLVEstruct));
 93: #else
 94:       PetscStackCall("SuperLU_DIST:dSolveFinalize",dSolveFinalize(&lu->options, &lu->SOLVEstruct));
 95: #endif
 96:     }
 97:     PetscStackCall("SuperLU_DIST:Destroy_LU",Destroy_LU(A->cmap->N, &lu->grid, &lu->LUstruct));
 98:     PetscStackCall("SuperLU_DIST:ScalePermstructFree",ScalePermstructFree(&lu->ScalePermstruct));
 99:     PetscStackCall("SuperLU_DIST:LUstructFree",LUstructFree(&lu->LUstruct));

101:     /* Release the SuperLU_DIST process grid. */
102:     PetscStackCall("SuperLU_DIST:superlu_gridexit",superlu_gridexit(&lu->grid));
103:     MPI_Comm_free(&(lu->comm_superlu));
104:   }
105:   PetscFree(A->data);
106:   /* clear composed functions */
107:   PetscObjectComposeFunction((PetscObject)A,"MatFactorGetSolverType_C",NULL);
108:   PetscObjectComposeFunction((PetscObject)A,"MatSuperluDistGetDiagU_C",NULL);

110:   return(0);
111: }

113: static PetscErrorCode MatSolve_SuperLU_DIST(Mat A,Vec b_mpi,Vec x)
114: {
115:   Mat_SuperLU_DIST *lu = (Mat_SuperLU_DIST*)A->data;
116:   PetscErrorCode   ierr;
117:   PetscInt         m=A->rmap->n;
118:   SuperLUStat_t    stat;
119:   double           berr[1];
120:   PetscScalar      *bptr=NULL;
121:   int              info; /* SuperLU_Dist info code is ALWAYS an int, even with long long indices */
122:   static PetscBool cite = PETSC_FALSE;

125:   if (lu->options.Fact != FACTORED) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"SuperLU_DIST options.Fact must equal FACTORED");
126:   PetscCitationsRegister("@article{lidemmel03,\n  author = {Xiaoye S. Li and James W. Demmel},\n  title = {{SuperLU_DIST}: A Scalable Distributed-Memory Sparse Direct\n           Solver for Unsymmetric Linear Systems},\n  journal = {ACM Trans. Mathematical Software},\n  volume = {29},\n  number = {2},\n  pages = {110-140},\n  year = 2003\n}\n",&cite);

128:   if (lu->options.SolveInitialized && !lu->matsolve_iscalled) {
129:     /* see comments in MatMatSolve() */
130: #if defined(PETSC_USE_COMPLEX)
131:     PetscStackCall("SuperLU_DIST:zSolveFinalize",zSolveFinalize(&lu->options, &lu->SOLVEstruct));
132: #else
133:     PetscStackCall("SuperLU_DIST:dSolveFinalize",dSolveFinalize(&lu->options, &lu->SOLVEstruct));
134: #endif
135:     lu->options.SolveInitialized = NO;
136:   }
137:   VecCopy(b_mpi,x);
138:   VecGetArray(x,&bptr);

140:   PetscStackCall("SuperLU_DIST:PStatInit",PStatInit(&stat));        /* Initialize the statistics variables. */
141: #if defined(PETSC_USE_COMPLEX)
142:   PetscStackCall("SuperLU_DIST:pzgssvx",pzgssvx(&lu->options,&lu->A_sup,&lu->ScalePermstruct,(doublecomplex*)bptr,m,1,&lu->grid,&lu->LUstruct,&lu->SOLVEstruct,berr,&stat,&info));
143: #else
144:   PetscStackCall("SuperLU_DIST:pdgssvx",pdgssvx(&lu->options,&lu->A_sup,&lu->ScalePermstruct,bptr,m,1,&lu->grid,&lu->LUstruct,&lu->SOLVEstruct,berr,&stat,&info));
145: #endif
146:   if (info) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"pdgssvx fails, info: %d\n",info);

148:   if (lu->options.PrintStat) PetscStackCall("SuperLU_DIST:PStatPrint",PStatPrint(&lu->options, &stat, &lu->grid));  /* Print the statistics. */
149:   PetscStackCall("SuperLU_DIST:PStatFree",PStatFree(&stat));

151:   VecRestoreArray(x,&bptr);
152:   lu->matsolve_iscalled    = PETSC_TRUE;
153:   lu->matmatsolve_iscalled = PETSC_FALSE;
154:   return(0);
155: }

157: static PetscErrorCode MatMatSolve_SuperLU_DIST(Mat A,Mat B_mpi,Mat X)
158: {
159:   Mat_SuperLU_DIST *lu = (Mat_SuperLU_DIST*)A->data;
160:   PetscErrorCode   ierr;
161:   PetscInt         m=A->rmap->n,nrhs;
162:   SuperLUStat_t    stat;
163:   double           berr[1];
164:   PetscScalar      *bptr;
165:   int              info; /* SuperLU_Dist info code is ALWAYS an int, even with long long indices */
166:   PetscBool        flg;

169:   if (lu->options.Fact != FACTORED) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"SuperLU_DIST options.Fact must equal FACTORED");
170:   PetscObjectTypeCompareAny((PetscObject)B_mpi,&flg,MATSEQDENSE,MATMPIDENSE,NULL);
171:   if (!flg) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Matrix B must be MATDENSE matrix");
172:   if (X != B_mpi) {
173:     PetscObjectTypeCompareAny((PetscObject)X,&flg,MATSEQDENSE,MATMPIDENSE,NULL);
174:     if (!flg) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Matrix X must be MATDENSE matrix");
175:   }

177:   if (lu->options.SolveInitialized && !lu->matmatsolve_iscalled) {
178:     /* communication pattern of SOLVEstruct is unlikely created for matmatsolve,
179:        thus destroy it and create a new SOLVEstruct.
180:        Otherwise it may result in memory corruption or incorrect solution
181:        See src/mat/tests/ex125.c */
182: #if defined(PETSC_USE_COMPLEX)
183:     PetscStackCall("SuperLU_DIST:zSolveFinalize",zSolveFinalize(&lu->options, &lu->SOLVEstruct));
184: #else
185:     PetscStackCall("SuperLU_DIST:dSolveFinalize",dSolveFinalize(&lu->options, &lu->SOLVEstruct));
186: #endif
187:     lu->options.SolveInitialized = NO;
188:   }
189:   if (X != B_mpi) {
190:     MatCopy(B_mpi,X,SAME_NONZERO_PATTERN);
191:   }

193:   MatGetSize(B_mpi,NULL,&nrhs);

195:   PetscStackCall("SuperLU_DIST:PStatInit",PStatInit(&stat));        /* Initialize the statistics variables. */
196:   MatDenseGetArray(X,&bptr);

198: #if defined(PETSC_USE_COMPLEX)
199:   PetscStackCall("SuperLU_DIST:pzgssvx",pzgssvx(&lu->options,&lu->A_sup,&lu->ScalePermstruct,(doublecomplex*)bptr,m,nrhs,&lu->grid, &lu->LUstruct,&lu->SOLVEstruct,berr,&stat,&info));
200: #else
201:   PetscStackCall("SuperLU_DIST:pdgssvx",pdgssvx(&lu->options,&lu->A_sup,&lu->ScalePermstruct,bptr,m,nrhs,&lu->grid,&lu->LUstruct,&lu->SOLVEstruct,berr,&stat,&info));
202: #endif

204:   if (info) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"pdgssvx fails, info: %d\n",info);
205:   MatDenseRestoreArray(X,&bptr);

207:   if (lu->options.PrintStat) PetscStackCall("SuperLU_DIST:PStatPrint",PStatPrint(&lu->options, &stat, &lu->grid));  /* Print the statistics. */
208:   PetscStackCall("SuperLU_DIST:PStatFree",PStatFree(&stat));
209:   lu->matsolve_iscalled    = PETSC_FALSE;
210:   lu->matmatsolve_iscalled = PETSC_TRUE;
211:   return(0);
212: }

214: /*
215:   input:
216:    F:        numeric Cholesky factor
217:   output:
218:    nneg:     total number of negative pivots
219:    nzero:    total number of zero pivots
220:    npos:     (global dimension of F) - nneg - nzero
221: */
222: static PetscErrorCode MatGetInertia_SuperLU_DIST(Mat F,PetscInt *nneg,PetscInt *nzero,PetscInt *npos)
223: {
224:   PetscErrorCode   ierr;
225:   Mat_SuperLU_DIST *lu = (Mat_SuperLU_DIST*)F->data;
226:   PetscScalar      *diagU=NULL;
227:   PetscInt         M,i,neg=0,zero=0,pos=0;
228:   PetscReal        r;

231:   if (!F->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Matrix factor F is not assembled");
232:   if (lu->options.RowPerm != NOROWPERM) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Must set NOROWPERM");
233:   MatGetSize(F,&M,NULL);
234:   PetscMalloc1(M,&diagU);
235:   MatSuperluDistGetDiagU(F,diagU);
236:   for (i=0; i<M; i++) {
237: #if defined(PETSC_USE_COMPLEX)
238:     r = PetscImaginaryPart(diagU[i])/10.0;
239:     if (r< -PETSC_MACHINE_EPSILON || r>PETSC_MACHINE_EPSILON) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"diagU[%d]=%g + i %g is non-real",i,PetscRealPart(diagU[i]),r*10.0);
240:     r = PetscRealPart(diagU[i]);
241: #else
242:     r = diagU[i];
243: #endif
244:     if (r > 0) {
245:       pos++;
246:     } else if (r < 0) {
247:       neg++;
248:     } else zero++;
249:   }

251:   PetscFree(diagU);
252:   if (nneg)  *nneg  = neg;
253:   if (nzero) *nzero = zero;
254:   if (npos)  *npos  = pos;
255:   return(0);
256: }

258: static PetscErrorCode MatLUFactorNumeric_SuperLU_DIST(Mat F,Mat A,const MatFactorInfo *info)
259: {
260:   Mat_SuperLU_DIST  *lu = (Mat_SuperLU_DIST*)F->data;
261:   Mat               Aloc;
262:   const PetscScalar *av;
263:   const PetscInt    *ai=NULL,*aj=NULL;
264:   PetscInt          nz,dummy;
265:   int               sinfo;   /* SuperLU_Dist info flag is always an int even with long long indices */
266:   SuperLUStat_t     stat;
267:   double            *berr=0;
268:   PetscBool         ismpiaij,isseqaij,flg;
269:   PetscErrorCode    ierr;

272:   PetscObjectBaseTypeCompare((PetscObject)A,MATSEQAIJ,&isseqaij);
273:   PetscObjectBaseTypeCompare((PetscObject)A,MATMPIAIJ,&ismpiaij);
274:   if (ismpiaij) {
275:     MatMPIAIJGetLocalMat(A,MAT_INITIAL_MATRIX,&Aloc);
276:   } else if (isseqaij) {
277:     PetscObjectReference((PetscObject)A);
278:     Aloc = A;
279:   } else SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Not for type %s",((PetscObject)A)->type_name);

281:   MatGetRowIJ(Aloc,0,PETSC_FALSE,PETSC_FALSE,&dummy,&ai,&aj,&flg);
282:   if (!flg) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"GetRowIJ failed");
283:   MatSeqAIJGetArrayRead(Aloc,&av);
284:   nz   = ai[Aloc->rmap->n];

286:   /* Allocations for A_sup */
287:   if (lu->options.Fact == DOFACT) { /* first numeric factorization */
288: #if defined(PETSC_USE_COMPLEX)
289:     PetscStackCall("SuperLU_DIST:zallocateA_dist",zallocateA_dist(Aloc->rmap->n, nz, &lu->val, &lu->col, &lu->row));
290: #else
291:     PetscStackCall("SuperLU_DIST:dallocateA_dist",dallocateA_dist(Aloc->rmap->n, nz, &lu->val, &lu->col, &lu->row));
292: #endif
293:   } else { /* successive numeric factorization, sparsity pattern and perm_c are reused. */
294:     if (lu->FactPattern == SamePattern_SameRowPerm) {
295:       lu->options.Fact = SamePattern_SameRowPerm; /* matrix has similar numerical values */
296:     } else if (lu->FactPattern == SamePattern) {
297:       PetscStackCall("SuperLU_DIST:Destroy_LU",Destroy_LU(A->rmap->N, &lu->grid, &lu->LUstruct)); /* Deallocate L and U matrices. */
298:       lu->options.Fact = SamePattern;
299:     } else if (lu->FactPattern == DOFACT) {
300:       PetscStackCall("SuperLU_DIST:Destroy_CompRowLoc_Matrix_dist",Destroy_CompRowLoc_Matrix_dist(&lu->A_sup));
301:       PetscStackCall("SuperLU_DIST:Destroy_LU",Destroy_LU(A->rmap->N, &lu->grid, &lu->LUstruct));
302:       lu->options.Fact = DOFACT;

304: #if defined(PETSC_USE_COMPLEX)
305:       PetscStackCall("SuperLU_DIST:zallocateA_dist",zallocateA_dist(Aloc->rmap->n, nz, &lu->val, &lu->col, &lu->row));
306: #else
307:       PetscStackCall("SuperLU_DIST:dallocateA_dist",dallocateA_dist(Aloc->rmap->n, nz, &lu->val, &lu->col, &lu->row));
308: #endif
309:     } else {
310:       SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"options.Fact must be one of SamePattern SamePattern_SameRowPerm DOFACT");
311:     }
312:   }

314:   /* Copy AIJ matrix to superlu_dist matrix */
315:   PetscArraycpy(lu->row,ai,Aloc->rmap->n+1);
316:   PetscArraycpy(lu->col,aj,nz);
317:   PetscArraycpy(lu->val,av,nz);
318:   MatRestoreRowIJ(Aloc,0,PETSC_FALSE,PETSC_FALSE,&dummy,&ai,&aj,&flg);
319:   if (!flg) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"RestoreRowIJ failed");
320:   MatSeqAIJRestoreArrayRead(Aloc,&av);
321:   MatDestroy(&Aloc);

323:   /* Create and setup A_sup */
324:   if (lu->options.Fact == DOFACT) {
325: #if defined(PETSC_USE_COMPLEX)
326:     PetscStackCall("SuperLU_DIST:zCreate_CompRowLoc_Matrix_dist",zCreate_CompRowLoc_Matrix_dist(&lu->A_sup, A->rmap->N, A->cmap->N, nz, A->rmap->n, A->rmap->rstart, lu->val, lu->col, lu->row, SLU_NR_loc, SLU_Z, SLU_GE));
327: #else
328:     PetscStackCall("SuperLU_DIST:dCreate_CompRowLoc_Matrix_dist",dCreate_CompRowLoc_Matrix_dist(&lu->A_sup, A->rmap->N, A->cmap->N, nz, A->rmap->n, A->rmap->rstart, lu->val, lu->col, lu->row, SLU_NR_loc, SLU_D, SLU_GE));
329: #endif
330:   }

332:   /* Factor the matrix. */
333:   PetscStackCall("SuperLU_DIST:PStatInit",PStatInit(&stat));   /* Initialize the statistics variables. */
334: #if defined(PETSC_USE_COMPLEX)
335:   PetscStackCall("SuperLU_DIST:pzgssvx",pzgssvx(&lu->options, &lu->A_sup, &lu->ScalePermstruct, 0, A->rmap->n, 0, &lu->grid, &lu->LUstruct, &lu->SOLVEstruct, berr, &stat, &sinfo));
336: #else
337:   PetscStackCall("SuperLU_DIST:pdgssvx",pdgssvx(&lu->options, &lu->A_sup, &lu->ScalePermstruct, 0, A->rmap->n, 0, &lu->grid, &lu->LUstruct, &lu->SOLVEstruct, berr, &stat, &sinfo));
338: #endif

340:   if (sinfo > 0) {
341:     if (A->erroriffailure) {
342:       SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_MAT_LU_ZRPVT,"Zero pivot in row %D",sinfo);
343:     } else {
344:       if (sinfo <= lu->A_sup.ncol) {
345:         F->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
346:         PetscInfo1(F,"U(i,i) is exactly zero, i= %D\n",sinfo);
347:       } else if (sinfo > lu->A_sup.ncol) {
348:         /*
349:          number of bytes allocated when memory allocation
350:          failure occurred, plus A->ncol.
351:          */
352:         F->factorerrortype = MAT_FACTOR_OUTMEMORY;
353:         PetscInfo1(F,"Number of bytes allocated when memory allocation fails %D\n",sinfo);
354:       }
355:     }
356:   } else if (sinfo < 0) {
357:     SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB, "info = %D, argument in p*gssvx() had an illegal value", sinfo);
358:   }

360:   if (lu->options.PrintStat) {
361:     PetscStackCall("SuperLU_DIST:PStatPrint",PStatPrint(&lu->options, &stat, &lu->grid));  /* Print the statistics. */
362:   }
363:   PetscStackCall("SuperLU_DIST:PStatFree",PStatFree(&stat));
364:   F->assembled     = PETSC_TRUE;
365:   F->preallocated  = PETSC_TRUE;
366:   lu->options.Fact = FACTORED; /* The factored form of A is supplied. Local option used by this func. only */
367:   return(0);
368: }

370: /* Note the Petsc r and c permutations are ignored */
371: static PetscErrorCode MatLUFactorSymbolic_SuperLU_DIST(Mat F,Mat A,IS r,IS c,const MatFactorInfo *info)
372: {
373:   Mat_SuperLU_DIST *lu = (Mat_SuperLU_DIST*)F->data;
374:   PetscInt         M   = A->rmap->N,N=A->cmap->N;

377:   /* Initialize the SuperLU process grid. */
378:   PetscStackCall("SuperLU_DIST:superlu_gridinit",superlu_gridinit(lu->comm_superlu, lu->nprow, lu->npcol, &lu->grid));

380:   /* Initialize ScalePermstruct and LUstruct. */
381:   PetscStackCall("SuperLU_DIST:ScalePermstructInit",ScalePermstructInit(M, N, &lu->ScalePermstruct));
382:   PetscStackCall("SuperLU_DIST:LUstructInit",LUstructInit(N, &lu->LUstruct));
383:   F->ops->lufactornumeric = MatLUFactorNumeric_SuperLU_DIST;
384:   F->ops->solve           = MatSolve_SuperLU_DIST;
385:   F->ops->matsolve        = MatMatSolve_SuperLU_DIST;
386:   F->ops->getinertia      = NULL;

388:   if (A->symmetric || A->hermitian) F->ops->getinertia = MatGetInertia_SuperLU_DIST;
389:   lu->CleanUpSuperLU_Dist = PETSC_TRUE;
390:   return(0);
391: }

393: static PetscErrorCode MatCholeskyFactorSymbolic_SuperLU_DIST(Mat F,Mat A,IS r,const MatFactorInfo *info)
394: {

398:   MatLUFactorSymbolic_SuperLU_DIST(F,A,r,r,info);
399:   F->ops->choleskyfactornumeric = MatLUFactorNumeric_SuperLU_DIST;
400:   return(0);
401: }

403: static PetscErrorCode MatFactorGetSolverType_aij_superlu_dist(Mat A,MatSolverType *type)
404: {
406:   *type = MATSOLVERSUPERLU_DIST;
407:   return(0);
408: }

410: static PetscErrorCode MatView_Info_SuperLU_DIST(Mat A,PetscViewer viewer)
411: {
412:   Mat_SuperLU_DIST       *lu=(Mat_SuperLU_DIST*)A->data;
413:   superlu_dist_options_t options;
414:   PetscErrorCode         ierr;

417:   /* check if matrix is superlu_dist type */
418:   if (A->ops->solve != MatSolve_SuperLU_DIST) return(0);

420:   options = lu->options;
421:   PetscViewerASCIIPrintf(viewer,"SuperLU_DIST run parameters:\n");
422:   PetscViewerASCIIPrintf(viewer,"  Process grid nprow %D x npcol %D \n",lu->nprow,lu->npcol);
423:   PetscViewerASCIIPrintf(viewer,"  Equilibrate matrix %s \n",PetscBools[options.Equil != NO]);
424:   PetscViewerASCIIPrintf(viewer,"  Replace tiny pivots %s \n",PetscBools[options.ReplaceTinyPivot != NO]);
425:   PetscViewerASCIIPrintf(viewer,"  Use iterative refinement %s \n",PetscBools[options.IterRefine == SLU_DOUBLE]);
426:   PetscViewerASCIIPrintf(viewer,"  Processors in row %d col partition %d \n",lu->nprow,lu->npcol);

428:   switch (options.RowPerm) {
429:   case NOROWPERM:
430:     PetscViewerASCIIPrintf(viewer,"  Row permutation NOROWPERM\n");
431:     break;
432:   case LargeDiag_MC64:
433:     PetscViewerASCIIPrintf(viewer,"  Row permutation LargeDiag_MC64\n");
434:     break;
435:   case LargeDiag_AWPM:
436:     PetscViewerASCIIPrintf(viewer,"  Row permutation LargeDiag_AWPM\n");
437:     break;
438:   case MY_PERMR:
439:     PetscViewerASCIIPrintf(viewer,"  Row permutation MY_PERMR\n");
440:     break;
441:   default:
442:     SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown column permutation");
443:   }

445:   switch (options.ColPerm) {
446:   case NATURAL:
447:     PetscViewerASCIIPrintf(viewer,"  Column permutation NATURAL\n");
448:     break;
449:   case MMD_AT_PLUS_A:
450:     PetscViewerASCIIPrintf(viewer,"  Column permutation MMD_AT_PLUS_A\n");
451:     break;
452:   case MMD_ATA:
453:     PetscViewerASCIIPrintf(viewer,"  Column permutation MMD_ATA\n");
454:     break;
455:   /*  Even though this is called METIS, the SuperLU_DIST code sets this by default if PARMETIS is defined, not METIS */
456:   case METIS_AT_PLUS_A:
457:     PetscViewerASCIIPrintf(viewer,"  Column permutation METIS_AT_PLUS_A\n");
458:     break;
459:   case PARMETIS:
460:     PetscViewerASCIIPrintf(viewer,"  Column permutation PARMETIS\n");
461:     break;
462:   default:
463:     SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown column permutation");
464:   }

466:   PetscViewerASCIIPrintf(viewer,"  Parallel symbolic factorization %s \n",PetscBools[options.ParSymbFact != NO]);

468:   if (lu->FactPattern == SamePattern) {
469:     PetscViewerASCIIPrintf(viewer,"  Repeated factorization SamePattern\n");
470:   } else if (lu->FactPattern == SamePattern_SameRowPerm) {
471:     PetscViewerASCIIPrintf(viewer,"  Repeated factorization SamePattern_SameRowPerm\n");
472:   } else if (lu->FactPattern == DOFACT) {
473:     PetscViewerASCIIPrintf(viewer,"  Repeated factorization DOFACT\n");
474:   } else {
475:     SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown factorization pattern");
476:   }
477:   return(0);
478: }

480: static PetscErrorCode MatView_SuperLU_DIST(Mat A,PetscViewer viewer)
481: {
482:   PetscErrorCode    ierr;
483:   PetscBool         iascii;
484:   PetscViewerFormat format;

487:   PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
488:   if (iascii) {
489:     PetscViewerGetFormat(viewer,&format);
490:     if (format == PETSC_VIEWER_ASCII_INFO) {
491:       MatView_Info_SuperLU_DIST(A,viewer);
492:     }
493:   }
494:   return(0);
495: }

497: static PetscErrorCode MatGetFactor_aij_superlu_dist(Mat A,MatFactorType ftype,Mat *F)
498: {
499:   Mat                    B;
500:   Mat_SuperLU_DIST       *lu;
501:   PetscErrorCode         ierr;
502:   PetscInt               M=A->rmap->N,N=A->cmap->N,indx;
503:   PetscMPIInt            size;
504:   superlu_dist_options_t options;
505:   PetscBool              flg;
506:   const char             *colperm[]     = {"NATURAL","MMD_AT_PLUS_A","MMD_ATA","METIS_AT_PLUS_A","PARMETIS"};
507:   const char             *rowperm[]     = {"NOROWPERM","LargeDiag_MC64","LargeDiag_AWPM","MY_PERMR"};
508:   const char             *factPattern[] = {"SamePattern","SamePattern_SameRowPerm","DOFACT"};
509:   PetscBool              set;

512:   /* Create the factorization matrix */
513:   MatCreate(PetscObjectComm((PetscObject)A),&B);
514:   MatSetSizes(B,A->rmap->n,A->cmap->n,M,N);
515:   PetscStrallocpy("superlu_dist",&((PetscObject)B)->type_name);
516:   MatSetUp(B);
517:   B->ops->getinfo = MatGetInfo_External;
518:   B->ops->view    = MatView_SuperLU_DIST;
519:   B->ops->destroy = MatDestroy_SuperLU_DIST;

521:   /* Set the default input options:
522:      options.Fact              = DOFACT;
523:      options.Equil             = YES;
524:      options.ParSymbFact       = NO;
525:      options.ColPerm           = METIS_AT_PLUS_A;
526:      options.RowPerm           = LargeDiag_MC64;
527:      options.ReplaceTinyPivot  = YES;
528:      options.IterRefine        = DOUBLE;
529:      options.Trans             = NOTRANS;
530:      options.SolveInitialized  = NO; -hold the communication pattern used MatSolve() and MatMatSolve()
531:      options.RefineInitialized = NO;
532:      options.PrintStat         = YES;
533:      options.SymPattern        = NO;
534:   */
535:   set_default_options_dist(&options);

537:   if (ftype == MAT_FACTOR_LU) {
538:     B->factortype = MAT_FACTOR_LU;
539:     B->ops->lufactorsymbolic = MatLUFactorSymbolic_SuperLU_DIST;
540:   } else {
541:     B->factortype = MAT_FACTOR_CHOLESKY;
542:     B->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SuperLU_DIST;
543:     options.SymPattern = YES;
544:   }

546:   /* set solvertype */
547:   PetscFree(B->solvertype);
548:   PetscStrallocpy(MATSOLVERSUPERLU_DIST,&B->solvertype);

550:   PetscNewLog(B,&lu);
551:   B->data = lu;

553:   MPI_Comm_dup(PetscObjectComm((PetscObject)A),&(lu->comm_superlu));
554:   MPI_Comm_size(PetscObjectComm((PetscObject)A),&size);
555:   /* Default num of process columns and rows */
556:   lu->nprow = (int_t) (0.5 + PetscSqrtReal((PetscReal)size));
557:   if (!lu->nprow) lu->nprow = 1;
558:   while (lu->nprow > 0) {
559:     lu->npcol = (int_t) (size/lu->nprow);
560:     if (size == lu->nprow * lu->npcol) break;
561:     lu->nprow--;
562:   }

564:   PetscOptionsBegin(PetscObjectComm((PetscObject)A),((PetscObject)A)->prefix,"SuperLU_Dist Options","Mat");
565:   PetscOptionsInt("-mat_superlu_dist_r","Number rows in processor partition","None",lu->nprow,(PetscInt*)&lu->nprow,NULL);
566:   PetscOptionsInt("-mat_superlu_dist_c","Number columns in processor partition","None",lu->npcol,(PetscInt*)&lu->npcol,NULL);
567:   if (size != lu->nprow * lu->npcol) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Number of processes %d must equal to nprow %d * npcol %d",size,lu->nprow,lu->npcol);

569:   PetscOptionsBool("-mat_superlu_dist_equil","Equilibrate matrix","None",options.Equil ? PETSC_TRUE : PETSC_FALSE,&flg,&set);
570:   if (set && !flg) options.Equil = NO;

572:   PetscOptionsEList("-mat_superlu_dist_rowperm","Row permutation","None",rowperm,4,rowperm[1],&indx,&flg);
573:   if (flg) {
574:     switch (indx) {
575:     case 0:
576:       options.RowPerm = NOROWPERM;
577:       break;
578:     case 1:
579:       options.RowPerm = LargeDiag_MC64;
580:       break;
581:     case 2:
582:       options.RowPerm = LargeDiag_AWPM;
583:       break;
584:     case 3:
585:       options.RowPerm = MY_PERMR;
586:       break;
587:     default:
588:       SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown row permutation");
589:     }
590:   }

592:   PetscOptionsEList("-mat_superlu_dist_colperm","Column permutation","None",colperm,5,colperm[3],&indx,&flg);
593:   if (flg) {
594:     switch (indx) {
595:     case 0:
596:       options.ColPerm = NATURAL;
597:       break;
598:     case 1:
599:       options.ColPerm = MMD_AT_PLUS_A;
600:       break;
601:     case 2:
602:       options.ColPerm = MMD_ATA;
603:       break;
604:     case 3:
605:       options.ColPerm = METIS_AT_PLUS_A;
606:       break;
607:     case 4:
608:       options.ColPerm = PARMETIS;   /* only works for np>1 */
609:       break;
610:     default:
611:       SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown column permutation");
612:     }
613:   }

615:   options.ReplaceTinyPivot = NO;
616:   PetscOptionsBool("-mat_superlu_dist_replacetinypivot","Replace tiny pivots","None",options.ReplaceTinyPivot ? PETSC_TRUE : PETSC_FALSE,&flg,&set);
617:   if (set && flg) options.ReplaceTinyPivot = YES;

619:   options.ParSymbFact = NO;
620:   PetscOptionsBool("-mat_superlu_dist_parsymbfact","Parallel symbolic factorization","None",PETSC_FALSE,&flg,&set);
621:   if (set && flg && size>1) {
622: #if defined(PETSC_HAVE_PARMETIS)
623:     options.ParSymbFact = YES;
624:     options.ColPerm     = PARMETIS;   /* in v2.2, PARMETIS is forced for ParSymbFact regardless of user ordering setting */
625: #else
626:     SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"parsymbfact needs PARMETIS");
627: #endif
628:   }

630:   lu->FactPattern = SamePattern;
631:   PetscOptionsEList("-mat_superlu_dist_fact","Sparsity pattern for repeated matrix factorization","None",factPattern,3,factPattern[0],&indx,&flg);
632:   if (flg) {
633:     switch (indx) {
634:     case 0:
635:       lu->FactPattern = SamePattern;
636:       break;
637:     case 1:
638:       lu->FactPattern = SamePattern_SameRowPerm;
639:       break;
640:     case 2:
641:       lu->FactPattern = DOFACT;
642:       break;
643:     }
644:   }

646:   options.IterRefine = NOREFINE;
647:   PetscOptionsBool("-mat_superlu_dist_iterrefine","Use iterative refinement","None",options.IterRefine == NOREFINE ? PETSC_FALSE : PETSC_TRUE ,&flg,&set);
648:   if (set) {
649:     if (flg) options.IterRefine = SLU_DOUBLE;
650:     else options.IterRefine = NOREFINE;
651:   }

653:   if (PetscLogPrintInfo) options.PrintStat = YES;
654:   else options.PrintStat = NO;
655:   PetscOptionsBool("-mat_superlu_dist_statprint","Print factorization information","None",(PetscBool)options.PrintStat,(PetscBool*)&options.PrintStat,NULL);
656:   PetscOptionsEnd();

658:   lu->options              = options;
659:   lu->options.Fact         = DOFACT;
660:   lu->matsolve_iscalled    = PETSC_FALSE;
661:   lu->matmatsolve_iscalled = PETSC_FALSE;

663:   PetscObjectComposeFunction((PetscObject)B,"MatFactorGetSolverType_C",MatFactorGetSolverType_aij_superlu_dist);
664:   PetscObjectComposeFunction((PetscObject)B,"MatSuperluDistGetDiagU_C",MatSuperluDistGetDiagU_SuperLU_DIST);

666:   *F = B;
667:   return(0);
668: }

670: PETSC_EXTERN PetscErrorCode MatSolverTypeRegister_SuperLU_DIST(void)
671: {
674:   MatSolverTypeRegister(MATSOLVERSUPERLU_DIST,MATMPIAIJ,MAT_FACTOR_LU,MatGetFactor_aij_superlu_dist);
675:   MatSolverTypeRegister(MATSOLVERSUPERLU_DIST,MATSEQAIJ,MAT_FACTOR_LU,MatGetFactor_aij_superlu_dist);
676:   MatSolverTypeRegister(MATSOLVERSUPERLU_DIST,MATMPIAIJ,MAT_FACTOR_CHOLESKY,MatGetFactor_aij_superlu_dist);
677:   MatSolverTypeRegister(MATSOLVERSUPERLU_DIST,MATSEQAIJ,MAT_FACTOR_CHOLESKY,MatGetFactor_aij_superlu_dist);
678:   return(0);
679: }

681: /*MC
682:   MATSOLVERSUPERLU_DIST - Parallel direct solver package for LU factorization

684:   Use ./configure --download-superlu_dist --download-parmetis --download-metis --download-ptscotch  to have PETSc installed with SuperLU_DIST

686:   Use -pc_type lu -pc_factor_mat_solver_type superlu_dist to use this direct solver

688:    Works with AIJ matrices

690:   Options Database Keys:
691: + -mat_superlu_dist_r <n> - number of rows in processor partition
692: . -mat_superlu_dist_c <n> - number of columns in processor partition
693: . -mat_superlu_dist_equil - equilibrate the matrix
694: . -mat_superlu_dist_rowperm <NOROWPERM,LargeDiag_MC64,LargeDiag_AWPM,MY_PERMR> - row permutation
695: . -mat_superlu_dist_colperm <NATURAL,MMD_AT_PLUS_A,MMD_ATA,METIS_AT_PLUS_A,PARMETIS> - column permutation
696: . -mat_superlu_dist_replacetinypivot - replace tiny pivots
697: . -mat_superlu_dist_fact <SamePattern> - (choose one of) SamePattern SamePattern_SameRowPerm DOFACT
698: . -mat_superlu_dist_iterrefine - use iterative refinement
699: - -mat_superlu_dist_statprint - print factorization information

701:    Level: beginner

703: .seealso: PCLU

705: .seealso: PCFactorSetMatSolverType(), MatSolverType

707: M*/