Actual source code: pastix.c

  1: #define PETSCMAT_DLL

  3: /* 
  4:     Provides an interface to the PaStiX sparse solver
  5: */
 6:  #include ../src/mat/impls/aij/seq/aij.h
 7:  #include ../src/mat/impls/aij/mpi/mpiaij.h
 8:  #include ../src/mat/impls/sbaij/seq/sbaij.h
 9:  #include ../src/mat/impls/sbaij/mpi/mpisbaij.h

 11: #if defined(PETSC_HAVE_STDLIB_H)
 12: #include <stdlib.h>
 13: #endif
 14: #if defined(PETSC_HAVE_STRING_H)
 15: #include <string.h> 
 16: #endif

 19: #include "pastix.h"

 22: typedef struct Mat_Pastix_ {
 23:   pastix_data_t *pastix_data;              /* Pastix data storage structure                        */
 24:   MatStructure   matstruc;
 25:   PetscInt       n;                        /* Number of columns in the matrix                      */
 26:   PetscInt       *colptr;                  /* Index of first element of each column in row and val */
 27:   PetscInt       *row;                     /* Row of each element of the matrix                    */
 28:   PetscScalar    *val;                     /* Value of each element of the matrix                  */
 29:   PetscInt       *perm;                    /* Permutation tabular                                  */
 30:   PetscInt       *invp;                    /* Reverse permutation tabular                          */
 31:   PetscScalar    *rhs;                     /* Rhight-hand-side member                              */
 32:   PetscInt       rhsnbr;                   /* Rhight-hand-side number (must be 1)                  */
 33:   PetscInt       iparm[64];                /* Integer parameters                                   */
 34:   double         dparm[64];                /* Floating point parameters                            */
 35:   MPI_Comm       pastix_comm;              /* PaStiX MPI communicator                              */
 36:   PetscMPIInt    commRank;                 /* MPI rank                                             */
 37:   PetscMPIInt    commSize;                 /* MPI communicator size                                */
 38:   PetscTruth     CleanUpPastix;            /* Boolean indicating if we call PaStiX clean step      */
 39:   VecScatter     scat_rhs;
 40:   VecScatter     scat_sol;
 41:   Vec            b_seq;
 42:   PetscTruth     isAIJ;
 43:   PetscErrorCode (*MatDestroy)(Mat);
 44: } Mat_Pastix;

 46: EXTERN PetscErrorCode MatDuplicate_Pastix(Mat,MatDuplicateOption,Mat*);

 50: /* 
 51:    convert Petsc seqaij matrix to CSC: colptr[n], row[nz], val[nz] 

 53:   input: 
 54:     A       - matrix in seqaij or mpisbaij (bs=1) format
 55:     valOnly - FALSE: spaces are allocated and values are set for the CSC 
 56:               TRUE:  Only fill values
 57:   output:     
 58:     n       - Size of the matrix
 59:     colptr  - Index of first element of each column in row and val
 60:     row     - Row of each element of the matrix                   
 61:     values  - Value of each element of the matrix                 
 62:  */
 63: PetscErrorCode MatConvertToCSC(Mat A,PetscTruth valOnly,PetscInt *n,PetscInt **colptr,PetscInt **row,PetscScalar **values)
 64: {
 65:   Mat_SeqAIJ     *aa      = (Mat_SeqAIJ*)A->data;
 66:   PetscInt       *rowptr  = aa->i;
 67:   PetscInt       *col     = aa->j;
 68:   PetscScalar    *rvalues = aa->a;
 69:   PetscInt        m       = A->rmap->N;
 70:   PetscInt        nnz;
 71:   PetscInt        i,j, k;
 72:   PetscInt        base = 1;
 73:   PetscInt        idx;
 74:   PetscErrorCode  ierr;
 75:   PetscInt        colidx;
 76:   PetscInt       *colcount;
 77:   PetscTruth      isSBAIJ;
 78:   PetscTruth      isSeqSBAIJ;
 79:   PetscTruth      isMpiSBAIJ;
 80:   PetscTruth      isSym;

 83:   MatIsSymmetric(A,0.0,&isSym);
 84:   PetscTypeCompare((PetscObject)A,MATSBAIJ,&isSBAIJ);
 85:   PetscTypeCompare((PetscObject)A,MATSEQSBAIJ,&isSeqSBAIJ);
 86:   PetscTypeCompare((PetscObject)A,MATMPISBAIJ,&isMpiSBAIJ);

 88:   *n = A->cmap->N;

 90:   /* PaStiX only needs triangular matrix if matrix is symmetric 
 91:    */
 92:   if (isSym && !(isSBAIJ || isSeqSBAIJ || isMpiSBAIJ)) {
 93:     nnz = (aa->nz - *n)/2 + *n;
 94:   }
 95:   else {
 96:     nnz     = aa->nz;
 97:   }

 99:   if (!valOnly){
100:     PetscMalloc(((*n)+1) *sizeof(PetscInt)   ,colptr );
101:     PetscMalloc( nnz     *sizeof(PetscInt)   ,row);
102:     PetscMalloc( nnz     *sizeof(PetscScalar),values);

104:     if (isSBAIJ || isSeqSBAIJ || isMpiSBAIJ) {
105:         PetscMemcpy (*colptr, rowptr, ((*n)+1)*sizeof(PetscInt));
106:         for (i = 0; i < *n+1; i++)
107:           (*colptr)[i] += base;
108:         PetscMemcpy (*row, col, (nnz)*sizeof(PetscInt));
109:         for (i = 0; i < nnz; i++)
110:           (*row)[i] += base;
111:         PetscMemcpy (*values, rvalues, (nnz)*sizeof(PetscScalar));
112:     } else {
113:       PetscMalloc((*n)*sizeof(PetscInt)   ,&colcount);

115:       for (i = 0; i < m; i++) colcount[i] = 0;
116:       /* Fill-in colptr */
117:       for (i = 0; i < m; i++) {
118:         for (j = rowptr[i]; j < rowptr[i+1]; j++) {
119:           if (!isSym || col[j] <= i)  colcount[col[j]]++;
120:         }
121:       }
122: 
123:       (*colptr)[0] = base;
124:       for (j = 0; j < *n; j++) {
125:         (*colptr)[j+1] = (*colptr)[j] + colcount[j];
126:         /* in next loop we fill starting from (*colptr)[colidx] - base */
127:         colcount[j] = -base;
128:       }
129: 
130:       /* Fill-in rows and values */
131:       for (i = 0; i < m; i++) {
132:         for (j = rowptr[i]; j < rowptr[i+1]; j++) {
133:           if (!isSym || col[j] <= i) {
134:             colidx = col[j];
135:             idx    = (*colptr)[colidx] + colcount[colidx];
136:             (*row)[idx]    = i + base;
137:             (*values)[idx] = rvalues[j];
138:             colcount[colidx]++;
139:           }
140:         }
141:       }
142:       PetscFree(colcount);
143:     }
144:   } else {
145:     /* Fill-in only values */
146:     for (i = 0; i < m; i++) {
147:       for (j = rowptr[i]; j < rowptr[i+1]; j++) {
148:         colidx = col[j];
149:         if ((isSBAIJ || isSeqSBAIJ || isMpiSBAIJ) ||!isSym || col[j] <= i)
150:           {
151:             /* look for the value to fill */
152:             for (k = (*colptr)[colidx] - base; k < (*colptr)[colidx + 1] - base; k++) {
153:               if (((*row)[k]-base) == i) {
154:                 (*values)[k] = rvalues[j];
155:                 break;
156:               }
157:             }
158:             /* data structure of sparse matrix has changed */
159:             if (k == (*colptr)[colidx + 1] - base) SETERRQ1(PETSC_ERR_PLIB,"overflow on k %D",k);
160:           }
161:       }
162:     }
163:   }
164:   {
165:     PetscScalar *tmpvalues;
166:     PetscInt    *tmprows,*tmpcolptr;
167:     PetscMalloc3(nnz,PetscScalar,&tmpvalues,nnz,PetscInt,&tmprows,(*n+1),PetscInt,&tmpcolptr);
168:     if (sizeof(PetscScalar) != sizeof(pastix_float_t)) {
169:       SETERRQ2(PETSC_ERR_SUP,"sizeof(PetscScalar) %d != sizeof(pastix_float_t) %d",sizeof(PetscScalar),sizeof(pastix_float_t));
170:     }
171:     if (sizeof(PetscInt) != sizeof(pastix_int_t)) {
172:       SETERRQ2(PETSC_ERR_SUP,"sizeof(PetscInt) %d != sizeof(pastix_int_t) %d",sizeof(PetscInt),sizeof(pastix_int_t));
173:     }

175:     PetscMemcpy(tmpcolptr,*colptr,(*n+1)*sizeof(PetscInt));
176:     PetscMemcpy(tmprows,*row,nnz*sizeof(PetscInt));
177:     PetscMemcpy(tmpvalues,*values,nnz*sizeof(PetscScalar));
178:     PetscFree(*row);
179:     PetscFree(*values);

181:     pastix_checkMatrix(MPI_COMM_WORLD,API_VERBOSE_NO,((isSym != 0) ? API_SYM_YES : API_SYM_NO),API_YES,*n,&tmpcolptr,&tmprows,&tmpvalues,NULL,1);
182: 
183:     PetscMemcpy(*colptr,tmpcolptr,(*n+1)*sizeof(PetscInt));
184:     PetscMalloc(((*colptr)[*n]-1)*sizeof(PetscInt),row);
185:     PetscMemcpy(*row,tmprows,((*colptr)[*n]-1)*sizeof(PetscInt));
186:     PetscMalloc(((*colptr)[*n]-1)*sizeof(PetscScalar),values);
187:     PetscMemcpy(*values,tmpvalues,((*colptr)[*n]-1)*sizeof(PetscScalar));
188:     PetscFree3(tmpvalues,tmprows,tmpcolptr);
189:   }
190:   return(0);
191: }



197: /*
198:   Call clean step of PaStiX if lu->CleanUpPastix == true.
199:   Free the CSC matrix.
200:  */
201: PetscErrorCode MatDestroy_Pastix(Mat A)
202: {
203:   Mat_Pastix      *lu=(Mat_Pastix*)A->spptr;
204:   PetscErrorCode   ierr;
205:   PetscMPIInt      size=lu->commSize;

208:   if (lu->CleanUpPastix) {
209:     /* Terminate instance, deallocate memories */
210:     if (size > 1){
211:       VecScatterDestroy(lu->scat_rhs);
212:       VecDestroy(lu->b_seq);
213:       VecScatterDestroy(lu->scat_sol);
214:     }
215: 
216:     lu->iparm[IPARM_START_TASK]=API_TASK_CLEAN;
217:     lu->iparm[IPARM_END_TASK]=API_TASK_CLEAN;

219:     pastix((pastix_data_t **)&(lu->pastix_data),
220:                               lu->pastix_comm,
221:            (pastix_int_t)     lu->n,
222:            (pastix_int_t*)    lu->colptr,
223:            (pastix_int_t*)    lu->row,
224:            (pastix_float_t*)  lu->val,
225:            (pastix_int_t*)    lu->perm,
226:            (pastix_int_t*)    lu->invp,
227:            (pastix_float_t*)  lu->rhs,
228:            (pastix_int_t)     lu->rhsnbr,
229:            (pastix_int_t*)    lu->iparm,
230:                               lu->dparm);

232:     PetscFree(lu->colptr);
233:     PetscFree(lu->row);
234:     PetscFree(lu->val);
235:     PetscFree(lu->perm);
236:     PetscFree(lu->invp);
237:     MPI_Comm_free(&(lu->pastix_comm));
238:   }
239:   (lu->MatDestroy)(A);
240:   return(0);
241: }

245: /*
246:   Gather right-hand-side.
247:   Call for Solve step.
248:   Scatter solution.
249:  */
250: PetscErrorCode MatSolve_PaStiX(Mat A,Vec b,Vec x)
251: {
252:   Mat_Pastix     *lu=(Mat_Pastix*)A->spptr;
253:   PetscScalar    *array;
254:   Vec             x_seq;
255:   PetscErrorCode  ierr;

258:   lu->rhsnbr = 1;
259:   x_seq = lu->b_seq;
260:   if (lu->commSize > 1){
261:     /* PaStiX only supports centralized rhs. Scatter b into a seqential rhs vector */
262:     VecScatterBegin(lu->scat_rhs,b,x_seq,INSERT_VALUES,SCATTER_FORWARD);
263:     VecScatterEnd(lu->scat_rhs,b,x_seq,INSERT_VALUES,SCATTER_FORWARD);
264:     VecGetArray(x_seq,&array);
265:   } else {  /* size == 1 */
266:     VecCopy(b,x);
267:     VecGetArray(x,&array);
268:   }
269:   lu->rhs = array;
270:   if (lu->commSize == 1){
271:     VecRestoreArray(x,&array);
272:   } else {
273:     VecRestoreArray(x_seq,&array);
274:   }

276:   /* solve phase */
277:   /*-------------*/
278:   lu->iparm[IPARM_START_TASK] = API_TASK_SOLVE;
279:   lu->iparm[IPARM_END_TASK]   = API_TASK_REFINE;
280:   lu->iparm[IPARM_RHS_MAKING] = API_RHS_B;
281: 
282:   pastix((pastix_data_t **)&(lu->pastix_data),
283:          (MPI_Comm)         lu->pastix_comm,
284:          (pastix_int_t)     lu->n,
285:          (pastix_int_t*)    lu->colptr,
286:          (pastix_int_t*)    lu->row,
287:          (pastix_float_t*)  lu->val,
288:          (pastix_int_t*)    lu->perm,
289:          (pastix_int_t*)    lu->invp,
290:          (pastix_float_t*)  lu->rhs,
291:          (pastix_int_t)     lu->rhsnbr,
292:          (pastix_int_t*)    lu->iparm,
293:          (double*)          lu->dparm);
294: 
295:   if (lu->iparm[IPARM_ERROR_NUMBER] < 0) {
296:     SETERRQ1(PETSC_ERR_LIB,"Error reported by PaStiX in solve phase: lu->iparm[IPARM_ERROR_NUMBER] = %d\n",lu->iparm[IPARM_ERROR_NUMBER] );
297:   }

299:   if (lu->commSize == 1){
300:     VecRestoreArray(x,&(lu->rhs));
301:   } else {
302:     VecRestoreArray(x_seq,&(lu->rhs));
303:   }

305:   if (lu->commSize > 1) { /* convert PaStiX centralized solution to petsc mpi x */
306:     VecScatterBegin(lu->scat_sol,x_seq,x,INSERT_VALUES,SCATTER_FORWARD);
307:     VecScatterEnd(lu->scat_sol,x_seq,x,INSERT_VALUES,SCATTER_FORWARD);
308:   }
309:   return(0);
310: }

312: /*
313:   Numeric factorisation using PaStiX solver.

315:  */
318: PetscErrorCode MatFactorNumeric_PaStiX(Mat F,Mat A,const MatFactorInfo *info)
319: {
320:   Mat_Pastix    *lu =(Mat_Pastix*)(F)->spptr;
321:   Mat           *tseq;
322:   PetscErrorCode 0;
323:   PetscInt       icntl;
324:   PetscInt       M=A->rmap->N;
325:   PetscTruth     valOnly,flg, isSym;
326:   Mat            F_diag;
327:   IS             is_iden;
328:   Vec            b;
329:   IS             isrow;
330:   PetscTruth     isSeqAIJ,isSeqSBAIJ;

333: 
334:   PetscTypeCompare((PetscObject)A,MATSEQAIJ,&isSeqAIJ);
335:   PetscTypeCompare((PetscObject)A,MATSEQSBAIJ,&isSeqSBAIJ);
336:   if (lu->matstruc == DIFFERENT_NONZERO_PATTERN){
337:     (F)->ops->solve   = MatSolve_PaStiX;

339:     /* Initialize a PASTIX instance */
340:     MPI_Comm_dup(((PetscObject)A)->comm,&(lu->pastix_comm));
341:     MPI_Comm_rank(lu->pastix_comm, &lu->commRank);
342:     MPI_Comm_size(lu->pastix_comm, &lu->commSize);

344:     /* Set pastix options */
345:     lu->iparm[IPARM_MODIFY_PARAMETER] = API_NO;
346:     lu->iparm[IPARM_START_TASK]       = API_TASK_INIT;
347:     lu->iparm[IPARM_END_TASK]         = API_TASK_INIT;
348:     lu->rhsnbr = 1;

350:     /* Call to set default pastix options */
351:     pastix((pastix_data_t **)&(lu->pastix_data),
352:            (MPI_Comm)         lu->pastix_comm,
353:            (pastix_int_t)     lu->n,
354:            (pastix_int_t*)    lu->colptr,
355:            (pastix_int_t*)    lu->row,
356:            (pastix_float_t*)  lu->val,
357:            (pastix_int_t*)    lu->perm,
358:            (pastix_int_t*)    lu->invp,
359:            (pastix_float_t*)  lu->rhs,
360:            (pastix_int_t)     lu->rhsnbr,
361:            (pastix_int_t*)    lu->iparm,
362:            (double*)          lu->dparm);

364:     PetscOptionsBegin(((PetscObject)A)->comm,((PetscObject)A)->prefix,"PaStiX Options","Mat");

366:     icntl=-1;
367:     lu->iparm[IPARM_VERBOSE] = API_VERBOSE_NOT;
368:     PetscOptionsInt("-mat_pastix_verbose","iparm[IPARM_VERBOSE] : level of printing (0 to 2)","None",lu->iparm[IPARM_VERBOSE],&icntl,&flg);
369:     if ((flg && icntl > 0) || PetscLogPrintInfo) {
370:       lu->iparm[IPARM_VERBOSE] =  icntl;
371:     }
372:     icntl=-1;
373:     PetscOptionsInt("-mat_pastix_threadnbr","iparm[IPARM_THREAD_NBR] : Number of thread by MPI node","None",lu->iparm[IPARM_THREAD_NBR],&icntl,PETSC_NULL);
374:     if ((flg && icntl > 0)) {
375:       lu->iparm[IPARM_THREAD_NBR] = icntl;
376:     }
377:     PetscOptionsEnd();
378:     valOnly = PETSC_FALSE;
379:   }  else {
380:     valOnly = PETSC_TRUE;
381:   }

383:   lu->iparm[IPARM_MATRIX_VERIFICATION] = API_YES;

385:   /* convert mpi A to seq mat A */
386:   ISCreateStride(PETSC_COMM_SELF,M,0,1,&isrow);
387:   MatGetSubMatrices(A,1,&isrow,&isrow,MAT_INITIAL_MATRIX,&tseq);
388:   ISDestroy(isrow);

390:   MatConvertToCSC(*tseq,valOnly, &lu->n, &lu->colptr, &lu->row, &lu->val);
391:   MatIsSymmetric(*tseq,0.0,&isSym);
392:   MatDestroyMatrices(1,&tseq);

394:   PetscMalloc((lu->n)*sizeof(PetscInt)   ,&(lu->perm));
395:   PetscMalloc((lu->n)*sizeof(PetscInt)   ,&(lu->invp));

397:   if (isSym) {
398:     /* On symmetric matrix, LLT */
399:     lu->iparm[IPARM_SYM] = API_SYM_YES;
400:     lu->iparm[IPARM_FACTORIZATION] = API_FACT_LDLT;
401:   } else {
402:     /* On unsymmetric matrix, LU */
403:     lu->iparm[IPARM_SYM] = API_SYM_NO;
404:     lu->iparm[IPARM_FACTORIZATION] = API_FACT_LU;
405:   }
406: 
407:   /*----------------*/
408:   if (lu->matstruc == DIFFERENT_NONZERO_PATTERN){
409:     if (!(isSeqAIJ || isSeqSBAIJ)) {
410:       /* PaStiX only supports centralized rhs. Create scatter scat_rhs for repeated use in MatSolve() */
411:         VecCreateSeq(PETSC_COMM_SELF,A->cmap->N,&lu->b_seq);
412:         ISCreateStride(PETSC_COMM_SELF,A->cmap->N,0,1,&is_iden);
413:         VecCreate(((PetscObject)A)->comm,&b);
414:         VecSetSizes(b,A->rmap->n,PETSC_DECIDE);
415:         VecSetFromOptions(b);
416: 
417:         VecScatterCreate(b,is_iden,lu->b_seq,is_iden,&lu->scat_rhs);
418:         VecScatterCreate(lu->b_seq,is_iden,b,is_iden,&lu->scat_sol);
419:         ISDestroy(is_iden);
420:         VecDestroy(b);
421:     }
422:     lu->iparm[IPARM_START_TASK] = API_TASK_ORDERING;
423:     lu->iparm[IPARM_END_TASK]   = API_TASK_NUMFACT;

425:     pastix((pastix_data_t **)&(lu->pastix_data),
426:            (MPI_Comm)         lu->pastix_comm,
427:            (pastix_int_t)     lu->n,
428:            (pastix_int_t*)    lu->colptr,
429:            (pastix_int_t*)    lu->row,
430:            (pastix_float_t*)  lu->val,
431:            (pastix_int_t*)    lu->perm,
432:            (pastix_int_t*)    lu->invp,
433:            (pastix_float_t*)  lu->rhs,
434:            (pastix_int_t)     lu->rhsnbr,
435:            (pastix_int_t*)    lu->iparm,
436:            (double*)          lu->dparm);
437:     if (lu->iparm[IPARM_ERROR_NUMBER] < 0) {
438:       SETERRQ1(PETSC_ERR_LIB,"Error reported by PaStiX in analysis phase: iparm(IPARM_ERROR_NUMBER)=%d\n",lu->iparm[IPARM_ERROR_NUMBER]);
439:     }
440:   } else {
441:     lu->iparm[IPARM_START_TASK] = API_TASK_NUMFACT;
442:     lu->iparm[IPARM_END_TASK]   = API_TASK_NUMFACT;
443:     pastix((pastix_data_t **)&(lu->pastix_data),
444:            (MPI_Comm)         lu->pastix_comm,
445:            (pastix_int_t)     lu->n,
446:            (pastix_int_t*)    lu->colptr,
447:            (pastix_int_t*)    lu->row,
448:            (pastix_float_t*)  lu->val,
449:            (pastix_int_t*)    lu->perm,
450:            (pastix_int_t*)    lu->invp,
451:            (pastix_float_t*)  lu->rhs,
452:            (pastix_int_t)     lu->rhsnbr,
453:            (pastix_int_t*)    lu->iparm,
454:            (double*)          lu->dparm);

456:     if (lu->iparm[IPARM_ERROR_NUMBER] < 0) {
457:       SETERRQ1(PETSC_ERR_LIB,"Error reported by PaStiX in analysis phase: iparm(IPARM_ERROR_NUMBER)=%d\n",lu->iparm[IPARM_ERROR_NUMBER]);
458:     }
459:   }

461:   if (lu->commSize > 1){
462:     if ((F)->factor == MAT_FACTOR_LU){
463:       F_diag = ((Mat_MPIAIJ *)(F)->data)->A;
464:     } else {
465:       F_diag = ((Mat_MPISBAIJ *)(F)->data)->A;
466:     }
467:     F_diag->assembled = PETSC_TRUE;
468:   }
469:   (F)->assembled     = PETSC_TRUE;
470:   lu->matstruc       = SAME_NONZERO_PATTERN;
471:   lu->CleanUpPastix  = PETSC_TRUE;
472:   return(0);
473: }

475: /* Note the Petsc r and c permutations are ignored */
478: PetscErrorCode MatLUFactorSymbolic_AIJPASTIX(Mat F,Mat A,IS r,IS c,const MatFactorInfo *info)
479: {
480:   Mat_Pastix      *lu = (Mat_Pastix*)F->spptr;

483:   lu->iparm[IPARM_FACTORIZATION] = API_FACT_LU;
484:   lu->iparm[IPARM_SYM]           = API_SYM_YES;
485:   lu->matstruc                   = DIFFERENT_NONZERO_PATTERN;
486:   F->ops->lufactornumeric        = MatFactorNumeric_PaStiX;
487:   return(0);
488: }


491: /* Note the Petsc r permutation is ignored */
494: PetscErrorCode MatCholeskyFactorSymbolic_SBAIJPASTIX(Mat F,Mat A,IS r,const MatFactorInfo *info)
495: {
496:   Mat_Pastix      *lu = (Mat_Pastix*)(F)->spptr;

499:   lu->iparm[IPARM_FACTORIZATION]  = API_FACT_LLT;
500:   lu->iparm[IPARM_SYM]            = API_SYM_NO;
501:   lu->matstruc                    = DIFFERENT_NONZERO_PATTERN;
502:   (F)->ops->choleskyfactornumeric = MatFactorNumeric_PaStiX;
503:   return(0);
504: }

508: PetscErrorCode MatView_PaStiX(Mat A,PetscViewer viewer)
509: {
510:   PetscErrorCode    ierr;
511:   PetscTruth        iascii;
512:   PetscViewerFormat format;

515:   PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&iascii);
516:   if (iascii) {
517:     PetscViewerGetFormat(viewer,&format);
518:     if (format == PETSC_VIEWER_ASCII_INFO){
519:       Mat_Pastix      *lu=(Mat_Pastix*)A->spptr;

521:       PetscViewerASCIIPrintf(viewer,"PaStiX run parameters:\n");
522:       PetscViewerASCIIPrintf(viewer,"  Matrix type :                      %s \n",((lu->iparm[IPARM_SYM] == API_SYM_YES)?"Symmetric":"Unsymmetric"));
523:       PetscViewerASCIIPrintf(viewer,"  Level of printing (0,1,2):         %d \n",lu->iparm[IPARM_VERBOSE]);
524:       PetscViewerASCIIPrintf(viewer,"  Number of refinements iterations : %d \n",lu->iparm[IPARM_NBITER]);
525:       PetscPrintf(PETSC_COMM_SELF,"  Error :                        %g \n",lu->dparm[DPARM_RELATIVE_ERROR]);
526:     }
527:   }
528:   return(0);
529: }


532: /*MC
533:      MAT_SOLVER_PASTIX  - A solver package providing direct solvers (LU) for distributed
534:   and sequential matrices via the external package PaStiX.

536:   Use config/configure.py --download-pastix to have PETSc installed with PaStiX

538:   Options Database Keys:
539: + -mat_pastix_verbose   <0,1,2>   - print level
540: - -mat_pastix_threadnbr <integer> - Set the thread number by MPI task.

542:   Level: beginner

544: .seealso: PCFactorSetMatSolverPackage(), MatSolverPackage

546: M*/


551: PetscErrorCode MatGetInfo_PaStiX(Mat A,MatInfoType flag,MatInfo *info)
552: {
553:     Mat_Pastix  *lu =(Mat_Pastix*)A->spptr;

556:     info->block_size        = 1.0;
557:     info->nz_allocated      = lu->iparm[IPARM_NNZEROS];
558:     info->nz_used           = lu->iparm[IPARM_NNZEROS];
559:     info->nz_unneeded       = 0.0;
560:     info->assemblies        = 0.0;
561:     info->mallocs           = 0.0;
562:     info->memory            = 0.0;
563:     info->fill_ratio_given  = 0;
564:     info->fill_ratio_needed = 0;
565:     info->factor_mallocs    = 0;
566:     return(0);
567: }

572: PetscErrorCode MatFactorGetSolverPackage_pastix(Mat A,const MatSolverPackage *type)
573: {
575:   *type = MAT_SOLVER_PASTIX;
576:   return(0);
577: }

581: /*
582:     The seq and mpi versions of this function are the same 
583: */
586: PetscErrorCode MatGetFactor_seqaij_pastix(Mat A,MatFactorType ftype,Mat *F)
587: {
588:   Mat            B;
590:   Mat_Pastix    *pastix;

593:   if (ftype != MAT_FACTOR_LU) {
594:     SETERRQ(PETSC_ERR_SUP,"Cannot use PETSc AIJ matrices with PaStiX Cholesky, use SBAIJ matrix");
595:   }
596:   /* Create the factorization matrix */
597:   MatCreate(((PetscObject)A)->comm,&B);
598:   MatSetSizes(B,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);
599:   MatSetType(B,((PetscObject)A)->type_name);
600:   MatSeqAIJSetPreallocation(B,0,PETSC_NULL);

602:   B->ops->lufactorsymbolic = MatLUFactorSymbolic_AIJPASTIX;
603:   B->ops->view             = MatView_PaStiX;
604:   B->ops->getinfo          = MatGetInfo_PaStiX;
605:   PetscObjectComposeFunctionDynamic((PetscObject)B,"MatFactorGetSolverPackage_C","MatFactorGetSolverPackage_pastix", MatFactorGetSolverPackage_pastix);
606:   B->factor                = MAT_FACTOR_LU;

608:   PetscNewLog(B,Mat_Pastix,&pastix);
609:   pastix->CleanUpPastix             = PETSC_FALSE;
610:   pastix->isAIJ                     = PETSC_TRUE;
611:   pastix->scat_rhs                  = PETSC_NULL;
612:   pastix->scat_sol                  = PETSC_NULL;
613:   pastix->MatDestroy                = B->ops->destroy;
614:   B->ops->destroy                   = MatDestroy_Pastix;
615:   B->spptr                          = (void*)pastix;

617:   *F = B;
618:   return(0);
619: }


626: PetscErrorCode MatGetFactor_mpiaij_pastix(Mat A,MatFactorType ftype,Mat *F)
627: {
628:   Mat            B;
630:   Mat_Pastix    *pastix;

633:   if (ftype != MAT_FACTOR_LU) SETERRQ(PETSC_ERR_SUP,"Cannot use PETSc AIJ matrices with PaStiX Cholesky, use SBAIJ matrix");
634:   /* Create the factorization matrix */
635:   MatCreate(((PetscObject)A)->comm,&B);
636:   MatSetSizes(B,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);
637:   MatSetType(B,((PetscObject)A)->type_name);
638:   MatSeqAIJSetPreallocation(B,0,PETSC_NULL);
639:   MatMPIAIJSetPreallocation(B,0,PETSC_NULL,0,PETSC_NULL);

641:   B->ops->lufactorsymbolic = MatLUFactorSymbolic_AIJPASTIX;
642:   B->ops->view             = MatView_PaStiX;
643:   PetscObjectComposeFunctionDynamic((PetscObject)B,"MatFactorGetSolverPackage_C","MatFactorGetSolverPackage_pastix",MatFactorGetSolverPackage_pastix);
644:   B->factor                = MAT_FACTOR_LU;

646:   PetscNewLog(B,Mat_Pastix,&pastix);
647:   pastix->CleanUpPastix             = PETSC_FALSE;
648:   pastix->isAIJ                     = PETSC_TRUE;
649:   pastix->scat_rhs                  = PETSC_NULL;
650:   pastix->scat_sol                  = PETSC_NULL;
651:   pastix->MatDestroy                = B->ops->destroy;
652:   B->ops->destroy                  = MatDestroy_Pastix;
653:   B->spptr                         = (void*)pastix;

655:   *F = B;
656:   return(0);
657: }

663: PetscErrorCode MatGetFactor_seqsbaij_pastix(Mat A,MatFactorType ftype,Mat *F)
664: {
665:   Mat            B;
667:   Mat_Pastix    *pastix;

670:   if (ftype != MAT_FACTOR_CHOLESKY) {
671:     SETERRQ(PETSC_ERR_SUP,"Cannot use PETSc SBAIJ matrices with PaStiX LU, use AIJ matrix");
672:   }
673:   /* Create the factorization matrix */
674:   MatCreate(((PetscObject)A)->comm,&B);
675:   MatSetSizes(B,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);
676:   MatSetType(B,((PetscObject)A)->type_name);
677:   MatSeqSBAIJSetPreallocation(B,1,0,PETSC_NULL);
678:   MatMPISBAIJSetPreallocation(B,1,0,PETSC_NULL,0,PETSC_NULL);

680:   B->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SBAIJPASTIX;
681:   B->ops->view                   = MatView_PaStiX;
682:   PetscObjectComposeFunctionDynamic((PetscObject)B,"MatFactorGetSolverPackage_C","MatFactorGetSolverPackage_pastix",MatFactorGetSolverPackage_pastix);

684:   B->factor                      = MAT_FACTOR_CHOLESKY;

686:   PetscNewLog(B,Mat_Pastix,&pastix);
687:   pastix->CleanUpPastix             = PETSC_FALSE;
688:   pastix->isAIJ                     = PETSC_TRUE;
689:   pastix->scat_rhs                  = PETSC_NULL;
690:   pastix->scat_sol                  = PETSC_NULL;
691:   pastix->MatDestroy                = B->ops->destroy;
692:   B->ops->destroy                  = MatDestroy_Pastix;
693:   B->spptr                         = (void*)pastix;

695:   *F = B;
696:   return(0);
697: }

703: PetscErrorCode MatGetFactor_mpisbaij_pastix(Mat A,MatFactorType ftype,Mat *F)
704: {
705:   Mat            B;
707:   Mat_Pastix    *pastix;
708: 
710:   if (ftype != MAT_FACTOR_CHOLESKY) SETERRQ(PETSC_ERR_SUP,"Cannot use PETSc SBAIJ matrices with PaStiX LU, use AIJ matrix");

712:   /* Create the factorization matrix */
713:   MatCreate(((PetscObject)A)->comm,&B);
714:   MatSetSizes(B,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);
715:   MatSetType(B,((PetscObject)A)->type_name);
716:   MatSeqSBAIJSetPreallocation(B,1,0,PETSC_NULL);
717:   MatMPISBAIJSetPreallocation(B,1,0,PETSC_NULL,0,PETSC_NULL);

719:   B->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SBAIJPASTIX;
720:   B->ops->view                   = MatView_PaStiX;
721:   PetscObjectComposeFunctionDynamic((PetscObject)B,"MatFactorGetSolverPackage_C","MatFactorGetSolverPackage_pastix",MatFactorGetSolverPackage_pastix);
722:   B->factor                      = MAT_FACTOR_CHOLESKY;

724:   PetscNewLog(B,Mat_Pastix,&pastix);
725:   pastix->CleanUpPastix             = PETSC_FALSE;
726:   pastix->isAIJ                     = PETSC_TRUE;
727:   pastix->scat_rhs                  = PETSC_NULL;
728:   pastix->scat_sol                  = PETSC_NULL;
729:   pastix->MatDestroy                = B->ops->destroy;
730:   B->ops->destroy                   = MatDestroy_Pastix;
731:   B->spptr                          = (void*)pastix;

733:   *F = B;
734:   return(0);
735: }