Actual source code: umfpack.c

  1: #define PETSCMAT_DLL

  3: /* 
  4:         Provides an interface to the UMFPACKv4.3 sparse solver
  5: */

 7:  #include src/mat/impls/aij/seq/aij.h

 10: #include "umfpack.h"

 13: typedef struct {
 14:   void         *Symbolic, *Numeric;
 15:   double       Info[UMFPACK_INFO], Control[UMFPACK_CONTROL],*W;
 16:   int          *Wi,*ai,*aj,*perm_c;
 17:   PetscScalar  *av;
 18:   MatStructure flg;
 19:   PetscTruth   PetscMatOdering;

 21:   /* A few function pointers for inheritance */
 22:   PetscErrorCode (*MatDuplicate)(Mat,MatDuplicateOption,Mat*);
 23:   PetscErrorCode (*MatView)(Mat,PetscViewer);
 24:   PetscErrorCode (*MatAssemblyEnd)(Mat,MatAssemblyType);
 25:   PetscErrorCode (*MatLUFactorSymbolic)(Mat,IS,IS,MatFactorInfo*,Mat*);
 26:   PetscErrorCode (*MatDestroy)(Mat);
 27: 
 28:   /* Flag to clean up UMFPACK objects during Destroy */
 29:   PetscTruth CleanUpUMFPACK;
 30: } Mat_UMFPACK;

 32: EXTERN PetscErrorCode MatDuplicate_UMFPACK(Mat,MatDuplicateOption,Mat*);

 37: PetscErrorCode PETSCMAT_DLLEXPORT MatConvert_UMFPACK_SeqAIJ(Mat A,MatType type,MatReuse reuse,Mat *newmat)
 38: {
 39:   /* This routine is only called to convert an unfactored PETSc-UMFPACK matrix */
 40:   /* to its base PETSc type, so we will ignore 'MatType type'. */
 42:   Mat         B=*newmat;
 43:   Mat_UMFPACK *lu=(Mat_UMFPACK*)A->spptr;

 46:   if (reuse == MAT_INITIAL_MATRIX) {
 47:     MatDuplicate(A,MAT_COPY_VALUES,&B);
 48:   }
 49:   /* Reset the original function pointers */
 50:   B->ops->duplicate        = lu->MatDuplicate;
 51:   B->ops->view             = lu->MatView;
 52:   B->ops->assemblyend      = lu->MatAssemblyEnd;
 53:   B->ops->lufactorsymbolic = lu->MatLUFactorSymbolic;
 54:   B->ops->destroy          = lu->MatDestroy;
 55: 
 56:   PetscFree(lu);

 58:   PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_umfpack_C","",PETSC_NULL);
 59:   PetscObjectComposeFunction((PetscObject)B,"MatConvert_umfpack_seqaij_C","",PETSC_NULL);

 61:   PetscObjectChangeTypeName((PetscObject)B,MATSEQAIJ);
 62:   *newmat = B;
 63:   return(0);
 64: }

 69: PetscErrorCode MatDestroy_UMFPACK(Mat A)
 70: {
 72:   Mat_UMFPACK *lu=(Mat_UMFPACK*)A->spptr;

 75:   if (lu->CleanUpUMFPACK) {
 76:     umfpack_di_free_symbolic(&lu->Symbolic);
 77:     umfpack_di_free_numeric(&lu->Numeric);
 78:     PetscFree(lu->Wi);
 79:     PetscFree(lu->W);
 80:     if (lu->PetscMatOdering) {
 81:       PetscFree(lu->perm_c);
 82:     }
 83:   }
 84:   MatConvert_UMFPACK_SeqAIJ(A,MATSEQAIJ,MAT_REUSE_MATRIX,&A);
 85:   (*A->ops->destroy)(A);
 86:   return(0);
 87: }

 91: PetscErrorCode MatSolve_UMFPACK(Mat A,Vec b,Vec x)
 92: {
 93:   Mat_UMFPACK *lu = (Mat_UMFPACK*)A->spptr;
 94:   PetscScalar *av=lu->av,*ba,*xa;
 96:   int          *ai=lu->ai,*aj=lu->aj,status;
 97: 
 99:   /* solve Ax = b by umfpack_di_wsolve */
100:   /* ----------------------------------*/
101:   VecGetArray(b,&ba);
102:   VecGetArray(x,&xa);

104:   status = umfpack_di_wsolve(UMFPACK_At,ai,aj,av,xa,ba,lu->Numeric,lu->Control,lu->Info,lu->Wi,lu->W);
105:   umfpack_di_report_info(lu->Control, lu->Info);
106:   if (status < 0){
107:     umfpack_di_report_status(lu->Control, status);
108:     SETERRQ(PETSC_ERR_LIB,"umfpack_di_wsolve failed");
109:   }
110: 
111:   VecRestoreArray(b,&ba);
112:   VecRestoreArray(x,&xa);

114:   return(0);
115: }

119: PetscErrorCode MatLUFactorNumeric_UMFPACK(Mat A,MatFactorInfo *info,Mat *F)
120: {
121:   Mat_UMFPACK *lu=(Mat_UMFPACK*)(*F)->spptr;
123:   int         *ai=lu->ai,*aj=lu->aj,m=A->m,status;
124:   PetscScalar *av=lu->av;

127:   /* numeric factorization of A' */
128:   /* ----------------------------*/
129:   if (lu->flg == SAME_NONZERO_PATTERN && lu->Numeric){
130:       umfpack_di_free_numeric(&lu->Numeric);
131:   }
132:   status = umfpack_di_numeric (ai,aj,av,lu->Symbolic,&lu->Numeric,lu->Control,lu->Info);
133:   if (status < 0) SETERRQ(PETSC_ERR_LIB,"umfpack_di_numeric failed");
134:   /* report numeric factorization of A' when Control[PRL] > 3 */
135:   (void) umfpack_di_report_numeric (lu->Numeric, lu->Control);

137:   if (lu->flg == DIFFERENT_NONZERO_PATTERN){  /* first numeric factorization */
138:     /* allocate working space to be used by Solve */
139:     PetscMalloc(m * sizeof(int), &lu->Wi);
140:     PetscMalloc(5*m * sizeof(double), &lu->W);
141:   }
142:   lu->flg = SAME_NONZERO_PATTERN;
143:   lu->CleanUpUMFPACK = PETSC_TRUE;
144:   return(0);
145: }

147: /*
148:    Note the r permutation is ignored
149: */
152: PetscErrorCode MatLUFactorSymbolic_UMFPACK(Mat A,IS r,IS c,MatFactorInfo *info,Mat *F)
153: {
154:   Mat         B;
155:   Mat_SeqAIJ  *mat=(Mat_SeqAIJ*)A->data;
156:   Mat_UMFPACK *lu;
158:   int          m=A->m,n=A->n,*ai=mat->i,*aj=mat->j,status,*ra,idx;
159:   PetscScalar *av=mat->a;
160:   const char  *strategy[]={"AUTO","UNSYMMETRIC","SYMMETRIC","2BY2"},
161:               *scale[]={"NONE","SUM","MAX"};
162:   PetscTruth  flg;
163: 
165:   /* Create the factorization matrix F */
166:   MatCreate(A->comm,&B);
167:   MatSetSizes(B,PETSC_DECIDE,PETSC_DECIDE,m,n);
168:   MatSetType(B,A->type_name);
169:   MatSeqAIJSetPreallocation(B,0,PETSC_NULL);
170: 
171:   B->ops->lufactornumeric = MatLUFactorNumeric_UMFPACK;
172:   B->ops->solve           = MatSolve_UMFPACK;
173:   B->factor               = FACTOR_LU;
174:   B->assembled            = PETSC_TRUE;  /* required by -ksp_view */

176:   lu = (Mat_UMFPACK*)(B->spptr);
177: 
178:   /* initializations */
179:   /* ------------------------------------------------*/
180:   /* get the default control parameters */
181:   umfpack_di_defaults (lu->Control);
182:   lu->perm_c = PETSC_NULL;  /* use defaul UMFPACK col permutation */
183:   lu->Control[UMFPACK_IRSTEP] = 0; /* max num of iterative refinement steps to attempt */

185:   PetscOptionsBegin(A->comm,A->prefix,"UMFPACK Options","Mat");
186:   /* Control parameters used by reporting routiones */
187:   PetscOptionsReal("-mat_umfpack_prl","Control[UMFPACK_PRL]","None",lu->Control[UMFPACK_PRL],&lu->Control[UMFPACK_PRL],PETSC_NULL);

189:   /* Control parameters for symbolic factorization */
190:   PetscOptionsEList("-mat_umfpack_strategy","ordering and pivoting strategy","None",strategy,4,strategy[0],&idx,&flg);
191:   if (flg) {
192:     switch (idx){
193:     case 0: lu->Control[UMFPACK_STRATEGY] = UMFPACK_STRATEGY_AUTO; break;
194:     case 1: lu->Control[UMFPACK_STRATEGY] = UMFPACK_STRATEGY_UNSYMMETRIC; break;
195:     case 2: lu->Control[UMFPACK_STRATEGY] = UMFPACK_STRATEGY_SYMMETRIC; break;
196:     case 3: lu->Control[UMFPACK_STRATEGY] = UMFPACK_STRATEGY_2BY2; break;
197:     }
198:   }
199:   PetscOptionsReal("-mat_umfpack_dense_col","Control[UMFPACK_DENSE_COL]","None",lu->Control[UMFPACK_DENSE_COL],&lu->Control[UMFPACK_DENSE_COL],PETSC_NULL);
200:   PetscOptionsReal("-mat_umfpack_dense_row","Control[UMFPACK_DENSE_ROW]","None",lu->Control[UMFPACK_DENSE_ROW],&lu->Control[UMFPACK_DENSE_ROW],PETSC_NULL);
201:   PetscOptionsReal("-mat_umfpack_amd_dense","Control[UMFPACK_AMD_DENSE]","None",lu->Control[UMFPACK_AMD_DENSE],&lu->Control[UMFPACK_AMD_DENSE],PETSC_NULL);
202:   PetscOptionsReal("-mat_umfpack_block_size","Control[UMFPACK_BLOCK_SIZE]","None",lu->Control[UMFPACK_BLOCK_SIZE],&lu->Control[UMFPACK_BLOCK_SIZE],PETSC_NULL);
203:   PetscOptionsReal("-mat_umfpack_2by2_tolerance","Control[UMFPACK_2BY2_TOLERANCE]","None",lu->Control[UMFPACK_2BY2_TOLERANCE],&lu->Control[UMFPACK_2BY2_TOLERANCE],PETSC_NULL);
204:   PetscOptionsReal("-mat_umfpack_fixq","Control[UMFPACK_FIXQ]","None",lu->Control[UMFPACK_FIXQ],&lu->Control[UMFPACK_FIXQ],PETSC_NULL);
205:   PetscOptionsReal("-mat_umfpack_aggressive","Control[UMFPACK_AGGRESSIVE]","None",lu->Control[UMFPACK_AGGRESSIVE],&lu->Control[UMFPACK_AGGRESSIVE],PETSC_NULL);

207:   /* Control parameters used by numeric factorization */
208:   PetscOptionsReal("-mat_umfpack_pivot_tolerance","Control[UMFPACK_PIVOT_TOLERANCE]","None",lu->Control[UMFPACK_PIVOT_TOLERANCE],&lu->Control[UMFPACK_PIVOT_TOLERANCE],PETSC_NULL);
209:   PetscOptionsReal("-mat_umfpack_sym_pivot_tolerance","Control[UMFPACK_SYM_PIVOT_TOLERANCE]","None",lu->Control[UMFPACK_SYM_PIVOT_TOLERANCE],&lu->Control[UMFPACK_SYM_PIVOT_TOLERANCE],PETSC_NULL);
210:   PetscOptionsEList("-mat_umfpack_scale","Control[UMFPACK_SCALE]","None",scale,3,scale[0],&idx,&flg);
211:   if (flg) {
212:     switch (idx){
213:     case 0: lu->Control[UMFPACK_SCALE] = UMFPACK_SCALE_NONE; break;
214:     case 1: lu->Control[UMFPACK_SCALE] = UMFPACK_SCALE_SUM; break;
215:     case 2: lu->Control[UMFPACK_SCALE] = UMFPACK_SCALE_MAX; break;
216:     }
217:   }
218:   PetscOptionsReal("-mat_umfpack_alloc_init","Control[UMFPACK_ALLOC_INIT]","None",lu->Control[UMFPACK_ALLOC_INIT],&lu->Control[UMFPACK_ALLOC_INIT],PETSC_NULL);
219:   PetscOptionsReal("-mat_umfpack_front_alloc_init","Control[UMFPACK_FRONT_ALLOC_INIT]","None",lu->Control[UMFPACK_FRONT_ALLOC_INIT],&lu->Control[UMFPACK_ALLOC_INIT],PETSC_NULL);
220:   PetscOptionsReal("-mat_umfpack_droptol","Control[UMFPACK_DROPTOL]","None",lu->Control[UMFPACK_DROPTOL],&lu->Control[UMFPACK_DROPTOL],PETSC_NULL);

222:   /* Control parameters used by solve */
223:   PetscOptionsReal("-mat_umfpack_irstep","Control[UMFPACK_IRSTEP]","None",lu->Control[UMFPACK_IRSTEP],&lu->Control[UMFPACK_IRSTEP],PETSC_NULL);
224: 
225:   /* use Petsc mat ordering (note: size is for the transpose, and PETSc r = Umfpack perm_c) */
226:   PetscOptionsHasName(PETSC_NULL,"-pc_lu_mat_ordering_type",&lu->PetscMatOdering);
227:   if (lu->PetscMatOdering) {
228:     ISGetIndices(r,&ra);
229:     PetscMalloc(A->m*sizeof(int),&lu->perm_c);
230:     PetscMemcpy(lu->perm_c,ra,A->m*sizeof(int));
231:     ISRestoreIndices(r,&ra);
232:   }
233:   PetscOptionsEnd();

235:   /* print the control parameters */
236:   if( lu->Control[UMFPACK_PRL] > 1 ) umfpack_di_report_control (lu->Control);

238:   /* symbolic factorization of A' */
239:   /* ---------------------------------------------------------------------- */
240:   if (lu->PetscMatOdering) { /* use Petsc row ordering */
241:     status = umfpack_di_qsymbolic(n,m,ai,aj,av,lu->perm_c,&lu->Symbolic,lu->Control,lu->Info);
242:   } else { /* use Umfpack col ordering */
243:     status = umfpack_di_symbolic(n,m,ai,aj,av,&lu->Symbolic,lu->Control,lu->Info);
244:   }
245:   if (status < 0){
246:     umfpack_di_report_info(lu->Control, lu->Info);
247:     umfpack_di_report_status(lu->Control, status);
248:     SETERRQ(PETSC_ERR_LIB,"umfpack_di_symbolic failed");
249:   }
250:   /* report sumbolic factorization of A' when Control[PRL] > 3 */
251:   (void) umfpack_di_report_symbolic(lu->Symbolic, lu->Control);

253:   lu->flg = DIFFERENT_NONZERO_PATTERN;
254:   lu->ai  = ai;
255:   lu->aj  = aj;
256:   lu->av  = av;
257:   lu->CleanUpUMFPACK = PETSC_TRUE;
258:   *F = B;
259:   return(0);
260: }

264: PetscErrorCode MatAssemblyEnd_UMFPACK(Mat A,MatAssemblyType mode)
265: {
267:   Mat_UMFPACK *lu=(Mat_UMFPACK*)(A->spptr);

270:   (*lu->MatAssemblyEnd)(A,mode);
271:   lu->MatLUFactorSymbolic  = A->ops->lufactorsymbolic;
272:   A->ops->lufactorsymbolic = MatLUFactorSymbolic_UMFPACK;
273:   return(0);
274: }

276: /* used by -ksp_view */
279: PetscErrorCode MatFactorInfo_UMFPACK(Mat A,PetscViewer viewer)
280: {
281:   Mat_UMFPACK *lu= (Mat_UMFPACK*)A->spptr;

285:   /* check if matrix is UMFPACK type */
286:   if (A->ops->solve != MatSolve_UMFPACK) return(0);

288:   PetscViewerASCIIPrintf(viewer,"UMFPACK run parameters:\n");
289:   /* Control parameters used by reporting routiones */
290:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_PRL]: %g\n",lu->Control[UMFPACK_PRL]);

292:   /* Control parameters used by symbolic factorization */
293:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_STRATEGY]: %g\n",lu->Control[UMFPACK_STRATEGY]);
294:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_DENSE_COL]: %g\n",lu->Control[UMFPACK_DENSE_COL]);
295:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_DENSE_ROW]: %g\n",lu->Control[UMFPACK_DENSE_ROW]);
296:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_AMD_DENSE]: %g\n",lu->Control[UMFPACK_AMD_DENSE]);
297:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_BLOCK_SIZE]: %g\n",lu->Control[UMFPACK_BLOCK_SIZE]);
298:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_2BY2_TOLERANCE]: %g\n",lu->Control[UMFPACK_2BY2_TOLERANCE]);
299:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_FIXQ]: %g\n",lu->Control[UMFPACK_FIXQ]);
300:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_AGGRESSIVE]: %g\n",lu->Control[UMFPACK_AGGRESSIVE]);

302:   /* Control parameters used by numeric factorization */
303:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_PIVOT_TOLERANCE]: %g\n",lu->Control[UMFPACK_PIVOT_TOLERANCE]);
304:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_SYM_PIVOT_TOLERANCE]: %g\n",lu->Control[UMFPACK_SYM_PIVOT_TOLERANCE]);
305:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_SCALE]: %g\n",lu->Control[UMFPACK_SCALE]);
306:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_ALLOC_INIT]: %g\n",lu->Control[UMFPACK_ALLOC_INIT]);
307:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_DROPTOL]: %g\n",lu->Control[UMFPACK_DROPTOL]);

309:   /* Control parameters used by solve */
310:   PetscViewerASCIIPrintf(viewer,"  Control[UMFPACK_IRSTEP]: %g\n",lu->Control[UMFPACK_IRSTEP]);

312:   /* mat ordering */
313:   if(!lu->PetscMatOdering) PetscViewerASCIIPrintf(viewer,"  UMFPACK default matrix ordering is used (not the PETSc matrix ordering) \n");

315:   return(0);
316: }

320: PetscErrorCode MatView_UMFPACK(Mat A,PetscViewer viewer)
321: {
323:   PetscTruth        iascii;
324:   PetscViewerFormat format;
325:   Mat_UMFPACK       *lu=(Mat_UMFPACK*)(A->spptr);

328:   (*lu->MatView)(A,viewer);

330:   PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&iascii);
331:   if (iascii) {
332:     PetscViewerGetFormat(viewer,&format);
333:     if (format == PETSC_VIEWER_ASCII_FACTOR_INFO) {
334:       MatFactorInfo_UMFPACK(A,viewer);
335:     }
336:   }
337:   return(0);
338: }

343: PetscErrorCode PETSCMAT_DLLEXPORT MatConvert_SeqAIJ_UMFPACK(Mat A,MatType type,MatReuse reuse,Mat *newmat)
344: {
345:   /* This routine is only called to convert to MATUMFPACK */
346:   /* from MATSEQAIJ, so we will ignore 'MatType type'. */
348:   Mat         B=*newmat;
349:   Mat_UMFPACK *lu;

352:   if (reuse == MAT_INITIAL_MATRIX) {
353:     MatDuplicate(A,MAT_COPY_VALUES,&B);
354:   }

356:   PetscNew(Mat_UMFPACK,&lu);
357:   lu->MatDuplicate         = A->ops->duplicate;
358:   lu->MatView              = A->ops->view;
359:   lu->MatAssemblyEnd       = A->ops->assemblyend;
360:   lu->MatLUFactorSymbolic  = A->ops->lufactorsymbolic;
361:   lu->MatDestroy           = A->ops->destroy;
362:   lu->CleanUpUMFPACK       = PETSC_FALSE;

364:   B->spptr                 = (void*)lu;
365:   B->ops->duplicate        = MatDuplicate_UMFPACK;
366:   B->ops->view             = MatView_UMFPACK;
367:   B->ops->assemblyend      = MatAssemblyEnd_UMFPACK;
368:   B->ops->lufactorsymbolic = MatLUFactorSymbolic_UMFPACK;
369:   B->ops->destroy          = MatDestroy_UMFPACK;

371:   PetscObjectComposeFunctionDynamic((PetscObject)B,"MatConvert_seqaij_umfpack_C",
372:                                            "MatConvert_SeqAIJ_UMFPACK",MatConvert_SeqAIJ_UMFPACK);
373:   PetscObjectComposeFunctionDynamic((PetscObject)B,"MatConvert_umfpack_seqaij_C",
374:                                            "MatConvert_UMFPACK_SeqAIJ",MatConvert_UMFPACK_SeqAIJ);
375:   PetscLogInfo((0,"MatConvert_SeqAIJ_UMFPACK:Using UMFPACK for SeqAIJ LU factorization and solves.\n"));
376:   PetscObjectChangeTypeName((PetscObject)B,MATUMFPACK);
377:   *newmat = B;
378:   return(0);
379: }

384: PetscErrorCode MatDuplicate_UMFPACK(Mat A, MatDuplicateOption op, Mat *M)
385: {
387:   Mat_UMFPACK *lu=(Mat_UMFPACK*)A->spptr;

390:   (*lu->MatDuplicate)(A,op,M);
391:   PetscMemcpy((*M)->spptr,lu,sizeof(Mat_UMFPACK));
392:   return(0);
393: }

395: /*MC
396:   MATUMFPACK - MATUMFPACK = "umfpack" - A matrix type providing direct solvers (LU) for sequential matrices 
397:   via the external package UMFPACK.

399:   If UMFPACK is installed (see the manual for
400:   instructions on how to declare the existence of external packages),
401:   a matrix type can be constructed which invokes UMFPACK solvers.
402:   After calling MatCreate(...,A), simply call MatSetType(A,UMFPACK).
403:   This matrix type is only supported for double precision real.

405:   This matrix inherits from MATSEQAIJ.  As a result, MatSeqAIJSetPreallocation is 
406:   supported for this matrix type.  One can also call MatConvert for an inplace conversion to or from 
407:   the MATSEQAIJ type without data copy.

409:   Consult UMFPACK documentation for more information about the Control parameters
410:   which correspond to the options database keys below.

412:   Options Database Keys:
413: + -mat_type umfpack - sets the matrix type to "umfpack" during a call to MatSetFromOptions()
414: . -mat_umfpack_prl - UMFPACK print level: Control[UMFPACK_PRL]
415: . -mat_umfpack_dense_col <alpha_c> - UMFPACK dense column threshold: Control[UMFPACK_DENSE_COL]
416: . -mat_umfpack_block_size <bs> - UMFPACK block size for BLAS-Level 3 calls: Control[UMFPACK_BLOCK_SIZE]
417: . -mat_umfpack_pivot_tolerance <delta> - UMFPACK partial pivot tolerance: Control[UMFPACK_PIVOT_TOLERANCE]
418: . -mat_umfpack_alloc_init <delta> - UMFPACK factorized matrix allocation modifier: Control[UMFPACK_ALLOC_INIT]
419: - -mat_umfpack_irstep <maxit> - UMFPACK maximum number of iterative refinement steps: Control[UMFPACK_IRSTEP]

421:    Level: beginner

423: .seealso: PCLU
424: M*/

429: PetscErrorCode PETSCMAT_DLLEXPORT MatCreate_UMFPACK(Mat A)
430: {

434:   /* Change type name before calling MatSetType to force proper construction of SeqAIJ and UMFPACK types */
435:   PetscObjectChangeTypeName((PetscObject)A,MATUMFPACK);
436:   MatSetType(A,MATSEQAIJ);
437:   MatConvert_SeqAIJ_UMFPACK(A,MATUMFPACK,MAT_REUSE_MATRIX,&A);
438:   return(0);
439: }