xref: /petsc/src/mat/impls/aij/mpi/superlu_dist/superlu_dist.c (revision ad5dfce3ff083d44a135ec532239c0cf546dc749)
1 /*
2         Provides an interface to the SuperLU_DIST sparse solver
3 */
4 
5 #include <../src/mat/impls/aij/seq/aij.h>
6 #include <../src/mat/impls/aij/mpi/mpiaij.h>
7 #include <petscpkg_version.h>
8 
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
36 
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;
56 
57 PetscErrorCode MatSuperluDistGetDiagU_SuperLU_DIST(Mat F,PetscScalar *diagU)
58 {
59   Mat_SuperLU_DIST *lu = (Mat_SuperLU_DIST*)F->data;
60 
61   PetscFunctionBegin;
62 #if defined(PETSC_USE_COMPLEX)
63   PetscStackCall("SuperLU_DIST:pzGetDiagU",pzGetDiagU(F->rmap->N,&lu->LUstruct,&lu->grid,(doublecomplex*)diagU));
64 #else
65   PetscStackCall("SuperLU_DIST:pdGetDiagU",pdGetDiagU(F->rmap->N,&lu->LUstruct,&lu->grid,diagU));
66 #endif
67   PetscFunctionReturn(0);
68 }
69 
70 PetscErrorCode MatSuperluDistGetDiagU(Mat F,PetscScalar *diagU)
71 {
72   PetscErrorCode ierr;
73 
74   PetscFunctionBegin;
75   PetscValidHeaderSpecific(F,MAT_CLASSID,1);
76   ierr = PetscTryMethod(F,"MatSuperluDistGetDiagU_C",(Mat,PetscScalar*),(F,diagU));CHKERRQ(ierr);
77   PetscFunctionReturn(0);
78 }
79 
80 /*  This allows reusing the Superlu_DIST communicator and grid when only a single SuperLU_DIST matrix is used at a time */
81 typedef struct {
82   MPI_Comm   comm;
83   PetscBool  busy;
84   gridinfo_t grid;
85 } PetscSuperLU_DIST;
86 static PetscMPIInt Petsc_Superlu_dist_keyval = MPI_KEYVAL_INVALID;
87 
88 PETSC_EXTERN PetscMPIInt MPIAPI Petsc_Superlu_dist_keyval_Delete_Fn(MPI_Comm comm,PetscMPIInt keyval,void *attr_val,void *extra_state)
89 {
90   PetscErrorCode    ierr;
91   PetscSuperLU_DIST *context = (PetscSuperLU_DIST *) attr_val;
92 
93   PetscFunctionBegin;
94   if (keyval != Petsc_Superlu_dist_keyval) SETERRMPI(PETSC_COMM_SELF,PETSC_ERR_ARG_CORRUPT,"Unexpected keyval");
95   ierr = PetscInfo(NULL,"Removing Petsc_Superlu_dist_keyval attribute from communicator that is being freed\n");CHKERRQ(ierr);
96   PetscStackCall("SuperLU_DIST:superlu_gridexit",superlu_gridexit(&context->grid));
97   ierr = MPI_Comm_free(&context->comm);CHKERRMPI(ierr);
98   ierr = PetscFree(context);CHKERRQ(ierr);
99   PetscFunctionReturn(MPI_SUCCESS);
100 }
101 
102 /*
103    Performs MPI_Comm_free_keyval() on Petsc_Superlu_dist_keyval but keeps the global variable for
104    users who do not destroy all PETSc objects before PetscFinalize().
105 
106    The value Petsc_Superlu_dist_keyval is retained so that Petsc_Superlu_dist_keyval_Delete_Fn()
107    can still check that the keyval associated with the MPI communicator is correct when the MPI
108    communicator is destroyed.
109 
110    This is called in PetscFinalize()
111 */
112 static PetscErrorCode Petsc_Superlu_dist_keyval_free(void)
113 {
114   PetscErrorCode ierr;
115   PetscMPIInt    Petsc_Superlu_dist_keyval_temp = Petsc_Superlu_dist_keyval;
116 
117   PetscFunctionBegin;
118   ierr = PetscInfo(NULL,"Freeing Petsc_Superlu_dist_keyval\n");CHKERRQ(ierr);
119   ierr = MPI_Comm_free_keyval(&Petsc_Superlu_dist_keyval_temp);CHKERRMPI(ierr);
120   PetscFunctionReturn(0);
121 }
122 
123 static PetscErrorCode MatDestroy_SuperLU_DIST(Mat A)
124 {
125   PetscErrorCode   ierr;
126   Mat_SuperLU_DIST *lu = (Mat_SuperLU_DIST*)A->data;
127 
128   PetscFunctionBegin;
129   if (lu->CleanUpSuperLU_Dist) {
130     /* Deallocate SuperLU_DIST storage */
131     PetscStackCall("SuperLU_DIST:Destroy_CompRowLoc_Matrix_dist",Destroy_CompRowLoc_Matrix_dist(&lu->A_sup));
132     if (lu->options.SolveInitialized) {
133 #if defined(PETSC_USE_COMPLEX)
134       PetscStackCall("SuperLU_DIST:zSolveFinalize",zSolveFinalize(&lu->options, &lu->SOLVEstruct));
135 #else
136       PetscStackCall("SuperLU_DIST:dSolveFinalize",dSolveFinalize(&lu->options, &lu->SOLVEstruct));
137 #endif
138     }
139     PetscStackCall("SuperLU_DIST:Destroy_LU",Destroy_LU(A->cmap->N, &lu->grid, &lu->LUstruct));
140     PetscStackCall("SuperLU_DIST:ScalePermstructFree",ScalePermstructFree(&lu->ScalePermstruct));
141     PetscStackCall("SuperLU_DIST:LUstructFree",LUstructFree(&lu->LUstruct));
142 
143     /* Release the SuperLU_DIST process grid. Only if the matrix has its own copy, this is it is not in the communicator context */
144     if (lu->comm_superlu) {
145       PetscStackCall("SuperLU_DIST:superlu_gridexit",superlu_gridexit(&lu->grid));
146     }
147   }
148   /*
149    * We always need to release the communicator that was created in MatGetFactor_aij_superlu_dist.
150    * lu->CleanUpSuperLU_Dist was turned on in MatLUFactorSymbolic_SuperLU_DIST. There are some use
151    * cases where we only create a matrix but do not solve mat. In these cases, lu->CleanUpSuperLU_Dist
152    * is off, and the communicator was not released or marked as "not busy " in the old code.
153    * Here we try to release comm regardless.
154   */
155   if (lu->comm_superlu) {
156     ierr = MPI_Comm_free(&(lu->comm_superlu));CHKERRMPI(ierr);
157   } else {
158     PetscSuperLU_DIST *context;
159     MPI_Comm          comm;
160     PetscMPIInt       flg;
161 
162     ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr);
163     ierr = MPI_Comm_get_attr(comm,Petsc_Superlu_dist_keyval,&context,&flg);CHKERRMPI(ierr);
164     if (!flg) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Communicator does not have expected Petsc_Superlu_dist_keyval attribute");
165     context->busy = PETSC_FALSE;
166   }
167 
168   ierr = PetscFree(A->data);CHKERRQ(ierr);
169   /* clear composed functions */
170   ierr = PetscObjectComposeFunction((PetscObject)A,"MatFactorGetSolverType_C",NULL);CHKERRQ(ierr);
171   ierr = PetscObjectComposeFunction((PetscObject)A,"MatSuperluDistGetDiagU_C",NULL);CHKERRQ(ierr);
172 
173   PetscFunctionReturn(0);
174 }
175 
176 static PetscErrorCode MatSolve_SuperLU_DIST(Mat A,Vec b_mpi,Vec x)
177 {
178   Mat_SuperLU_DIST *lu = (Mat_SuperLU_DIST*)A->data;
179   PetscErrorCode   ierr;
180   PetscInt         m=A->rmap->n;
181   SuperLUStat_t    stat;
182   double           berr[1];
183   PetscScalar      *bptr=NULL;
184   int              info; /* SuperLU_Dist info code is ALWAYS an int, even with long long indices */
185   static PetscBool cite = PETSC_FALSE;
186 
187   PetscFunctionBegin;
188   if (lu->options.Fact != FACTORED) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"SuperLU_DIST options.Fact must equal FACTORED");
189   ierr = 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);CHKERRQ(ierr);
190 
191   if (lu->options.SolveInitialized && !lu->matsolve_iscalled) {
192     /* see comments in MatMatSolve() */
193 #if defined(PETSC_USE_COMPLEX)
194     PetscStackCall("SuperLU_DIST:zSolveFinalize",zSolveFinalize(&lu->options, &lu->SOLVEstruct));
195 #else
196     PetscStackCall("SuperLU_DIST:dSolveFinalize",dSolveFinalize(&lu->options, &lu->SOLVEstruct));
197 #endif
198     lu->options.SolveInitialized = NO;
199   }
200   ierr = VecCopy(b_mpi,x);CHKERRQ(ierr);
201   ierr = VecGetArray(x,&bptr);CHKERRQ(ierr);
202 
203   PetscStackCall("SuperLU_DIST:PStatInit",PStatInit(&stat));        /* Initialize the statistics variables. */
204 #if defined(PETSC_USE_COMPLEX)
205   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));
206 #else
207   PetscStackCall("SuperLU_DIST:pdgssvx",pdgssvx(&lu->options,&lu->A_sup,&lu->ScalePermstruct,bptr,m,1,&lu->grid,&lu->LUstruct,&lu->SOLVEstruct,berr,&stat,&info));
208 #endif
209   if (info) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"pdgssvx fails, info: %d\n",info);
210 
211   if (lu->options.PrintStat) PetscStackCall("SuperLU_DIST:PStatPrint",PStatPrint(&lu->options, &stat, &lu->grid));  /* Print the statistics. */
212   PetscStackCall("SuperLU_DIST:PStatFree",PStatFree(&stat));
213 
214   ierr = VecRestoreArray(x,&bptr);CHKERRQ(ierr);
215   lu->matsolve_iscalled    = PETSC_TRUE;
216   lu->matmatsolve_iscalled = PETSC_FALSE;
217   PetscFunctionReturn(0);
218 }
219 
220 static PetscErrorCode MatMatSolve_SuperLU_DIST(Mat A,Mat B_mpi,Mat X)
221 {
222   Mat_SuperLU_DIST *lu = (Mat_SuperLU_DIST*)A->data;
223   PetscErrorCode   ierr;
224   PetscInt         m=A->rmap->n,nrhs;
225   SuperLUStat_t    stat;
226   double           berr[1];
227   PetscScalar      *bptr;
228   int              info; /* SuperLU_Dist info code is ALWAYS an int, even with long long indices */
229   PetscBool        flg;
230 
231   PetscFunctionBegin;
232   if (lu->options.Fact != FACTORED) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"SuperLU_DIST options.Fact must equal FACTORED");
233   ierr = PetscObjectTypeCompareAny((PetscObject)B_mpi,&flg,MATSEQDENSE,MATMPIDENSE,NULL);CHKERRQ(ierr);
234   if (!flg) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Matrix B must be MATDENSE matrix");
235   if (X != B_mpi) {
236     ierr = PetscObjectTypeCompareAny((PetscObject)X,&flg,MATSEQDENSE,MATMPIDENSE,NULL);CHKERRQ(ierr);
237     if (!flg) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Matrix X must be MATDENSE matrix");
238   }
239 
240   if (lu->options.SolveInitialized && !lu->matmatsolve_iscalled) {
241     /* communication pattern of SOLVEstruct is unlikely created for matmatsolve,
242        thus destroy it and create a new SOLVEstruct.
243        Otherwise it may result in memory corruption or incorrect solution
244        See src/mat/tests/ex125.c */
245 #if defined(PETSC_USE_COMPLEX)
246     PetscStackCall("SuperLU_DIST:zSolveFinalize",zSolveFinalize(&lu->options, &lu->SOLVEstruct));
247 #else
248     PetscStackCall("SuperLU_DIST:dSolveFinalize",dSolveFinalize(&lu->options, &lu->SOLVEstruct));
249 #endif
250     lu->options.SolveInitialized = NO;
251   }
252   if (X != B_mpi) {
253     ierr = MatCopy(B_mpi,X,SAME_NONZERO_PATTERN);CHKERRQ(ierr);
254   }
255 
256   ierr = MatGetSize(B_mpi,NULL,&nrhs);CHKERRQ(ierr);
257 
258   PetscStackCall("SuperLU_DIST:PStatInit",PStatInit(&stat));        /* Initialize the statistics variables. */
259   ierr = MatDenseGetArray(X,&bptr);CHKERRQ(ierr);
260 
261 #if defined(PETSC_USE_COMPLEX)
262   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));
263 #else
264   PetscStackCall("SuperLU_DIST:pdgssvx",pdgssvx(&lu->options,&lu->A_sup,&lu->ScalePermstruct,bptr,m,nrhs,&lu->grid,&lu->LUstruct,&lu->SOLVEstruct,berr,&stat,&info));
265 #endif
266 
267   if (info) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"pdgssvx fails, info: %d\n",info);
268   ierr = MatDenseRestoreArray(X,&bptr);CHKERRQ(ierr);
269 
270   if (lu->options.PrintStat) PetscStackCall("SuperLU_DIST:PStatPrint",PStatPrint(&lu->options, &stat, &lu->grid));  /* Print the statistics. */
271   PetscStackCall("SuperLU_DIST:PStatFree",PStatFree(&stat));
272   lu->matsolve_iscalled    = PETSC_FALSE;
273   lu->matmatsolve_iscalled = PETSC_TRUE;
274   PetscFunctionReturn(0);
275 }
276 
277 /*
278   input:
279    F:        numeric Cholesky factor
280   output:
281    nneg:     total number of negative pivots
282    nzero:    total number of zero pivots
283    npos:     (global dimension of F) - nneg - nzero
284 */
285 static PetscErrorCode MatGetInertia_SuperLU_DIST(Mat F,PetscInt *nneg,PetscInt *nzero,PetscInt *npos)
286 {
287   PetscErrorCode   ierr;
288   Mat_SuperLU_DIST *lu = (Mat_SuperLU_DIST*)F->data;
289   PetscScalar      *diagU=NULL;
290   PetscInt         M,i,neg=0,zero=0,pos=0;
291   PetscReal        r;
292 
293   PetscFunctionBegin;
294   if (!F->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Matrix factor F is not assembled");
295   if (lu->options.RowPerm != NOROWPERM) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Must set NOROWPERM");
296   ierr = MatGetSize(F,&M,NULL);CHKERRQ(ierr);
297   ierr = PetscMalloc1(M,&diagU);CHKERRQ(ierr);
298   ierr = MatSuperluDistGetDiagU(F,diagU);CHKERRQ(ierr);
299   for (i=0; i<M; i++) {
300 #if defined(PETSC_USE_COMPLEX)
301     r = PetscImaginaryPart(diagU[i])/10.0;
302     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);
303     r = PetscRealPart(diagU[i]);
304 #else
305     r = diagU[i];
306 #endif
307     if (r > 0) {
308       pos++;
309     } else if (r < 0) {
310       neg++;
311     } else zero++;
312   }
313 
314   ierr = PetscFree(diagU);CHKERRQ(ierr);
315   if (nneg)  *nneg  = neg;
316   if (nzero) *nzero = zero;
317   if (npos)  *npos  = pos;
318   PetscFunctionReturn(0);
319 }
320 
321 static PetscErrorCode MatLUFactorNumeric_SuperLU_DIST(Mat F,Mat A,const MatFactorInfo *info)
322 {
323   Mat_SuperLU_DIST  *lu = (Mat_SuperLU_DIST*)F->data;
324   Mat               Aloc;
325   const PetscScalar *av;
326   const PetscInt    *ai=NULL,*aj=NULL;
327   PetscInt          nz,dummy;
328   int               sinfo;   /* SuperLU_Dist info flag is always an int even with long long indices */
329   SuperLUStat_t     stat;
330   double            *berr=0;
331   PetscBool         ismpiaij,isseqaij,flg;
332   PetscErrorCode    ierr;
333 
334   PetscFunctionBegin;
335   ierr = PetscObjectBaseTypeCompare((PetscObject)A,MATSEQAIJ,&isseqaij);CHKERRQ(ierr);
336   ierr = PetscObjectBaseTypeCompare((PetscObject)A,MATMPIAIJ,&ismpiaij);CHKERRQ(ierr);
337   if (ismpiaij) {
338     ierr = MatMPIAIJGetLocalMat(A,MAT_INITIAL_MATRIX,&Aloc);CHKERRQ(ierr);
339   } else if (isseqaij) {
340     ierr = PetscObjectReference((PetscObject)A);CHKERRQ(ierr);
341     Aloc = A;
342   } else SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Not for type %s",((PetscObject)A)->type_name);
343 
344   ierr = MatGetRowIJ(Aloc,0,PETSC_FALSE,PETSC_FALSE,&dummy,&ai,&aj,&flg);CHKERRQ(ierr);
345   if (!flg) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"GetRowIJ failed");
346   ierr = MatSeqAIJGetArrayRead(Aloc,&av);CHKERRQ(ierr);
347   nz   = ai[Aloc->rmap->n];
348 
349   /* Allocations for A_sup */
350   if (lu->options.Fact == DOFACT) { /* first numeric factorization */
351 #if defined(PETSC_USE_COMPLEX)
352     PetscStackCall("SuperLU_DIST:zallocateA_dist",zallocateA_dist(Aloc->rmap->n, nz, &lu->val, &lu->col, &lu->row));
353 #else
354     PetscStackCall("SuperLU_DIST:dallocateA_dist",dallocateA_dist(Aloc->rmap->n, nz, &lu->val, &lu->col, &lu->row));
355 #endif
356   } else { /* successive numeric factorization, sparsity pattern and perm_c are reused. */
357     if (lu->FactPattern == SamePattern_SameRowPerm) {
358       lu->options.Fact = SamePattern_SameRowPerm; /* matrix has similar numerical values */
359     } else if (lu->FactPattern == SamePattern) {
360       PetscStackCall("SuperLU_DIST:Destroy_LU",Destroy_LU(A->rmap->N, &lu->grid, &lu->LUstruct)); /* Deallocate L and U matrices. */
361       lu->options.Fact = SamePattern;
362     } else if (lu->FactPattern == DOFACT) {
363       PetscStackCall("SuperLU_DIST:Destroy_CompRowLoc_Matrix_dist",Destroy_CompRowLoc_Matrix_dist(&lu->A_sup));
364       PetscStackCall("SuperLU_DIST:Destroy_LU",Destroy_LU(A->rmap->N, &lu->grid, &lu->LUstruct));
365       lu->options.Fact = DOFACT;
366 
367 #if defined(PETSC_USE_COMPLEX)
368       PetscStackCall("SuperLU_DIST:zallocateA_dist",zallocateA_dist(Aloc->rmap->n, nz, &lu->val, &lu->col, &lu->row));
369 #else
370       PetscStackCall("SuperLU_DIST:dallocateA_dist",dallocateA_dist(Aloc->rmap->n, nz, &lu->val, &lu->col, &lu->row));
371 #endif
372     } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"options.Fact must be one of SamePattern SamePattern_SameRowPerm DOFACT");
373   }
374 
375   /* Copy AIJ matrix to superlu_dist matrix */
376   ierr = PetscArraycpy(lu->row,ai,Aloc->rmap->n+1);CHKERRQ(ierr);
377   ierr = PetscArraycpy(lu->col,aj,nz);CHKERRQ(ierr);
378   ierr = PetscArraycpy(lu->val,av,nz);CHKERRQ(ierr);
379   ierr = MatRestoreRowIJ(Aloc,0,PETSC_FALSE,PETSC_FALSE,&dummy,&ai,&aj,&flg);CHKERRQ(ierr);
380   if (!flg) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"RestoreRowIJ failed");
381   ierr = MatSeqAIJRestoreArrayRead(Aloc,&av);CHKERRQ(ierr);
382   ierr = MatDestroy(&Aloc);CHKERRQ(ierr);
383 
384   /* Create and setup A_sup */
385   if (lu->options.Fact == DOFACT) {
386 #if defined(PETSC_USE_COMPLEX)
387     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));
388 #else
389     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));
390 #endif
391   }
392 
393   /* Factor the matrix. */
394   PetscStackCall("SuperLU_DIST:PStatInit",PStatInit(&stat));   /* Initialize the statistics variables. */
395 #if defined(PETSC_USE_COMPLEX)
396   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));
397 #else
398   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));
399 #endif
400 
401   if (sinfo > 0) {
402     if (A->erroriffailure) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_MAT_LU_ZRPVT,"Zero pivot in row %D",sinfo);
403     else {
404       if (sinfo <= lu->A_sup.ncol) {
405         F->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
406         ierr = PetscInfo1(F,"U(i,i) is exactly zero, i= %D\n",sinfo);CHKERRQ(ierr);
407       } else if (sinfo > lu->A_sup.ncol) {
408         /*
409          number of bytes allocated when memory allocation
410          failure occurred, plus A->ncol.
411          */
412         F->factorerrortype = MAT_FACTOR_OUTMEMORY;
413         ierr = PetscInfo1(F,"Number of bytes allocated when memory allocation fails %D\n",sinfo);CHKERRQ(ierr);
414       }
415     }
416   } else if (sinfo < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB, "info = %D, argument in p*gssvx() had an illegal value", sinfo);
417 
418   if (lu->options.PrintStat) {
419     PetscStackCall("SuperLU_DIST:PStatPrint",PStatPrint(&lu->options, &stat, &lu->grid));  /* Print the statistics. */
420   }
421   PetscStackCall("SuperLU_DIST:PStatFree",PStatFree(&stat));
422   F->assembled     = PETSC_TRUE;
423   F->preallocated  = PETSC_TRUE;
424   lu->options.Fact = FACTORED; /* The factored form of A is supplied. Local option used by this func. only */
425   PetscFunctionReturn(0);
426 }
427 
428 /* Note the Petsc r and c permutations are ignored */
429 static PetscErrorCode MatLUFactorSymbolic_SuperLU_DIST(Mat F,Mat A,IS r,IS c,const MatFactorInfo *info)
430 {
431   Mat_SuperLU_DIST *lu = (Mat_SuperLU_DIST*)F->data;
432   PetscInt         M   = A->rmap->N,N=A->cmap->N;
433 
434   PetscFunctionBegin;
435   /* Initialize ScalePermstruct and LUstruct. */
436   PetscStackCall("SuperLU_DIST:ScalePermstructInit",ScalePermstructInit(M, N, &lu->ScalePermstruct));
437   PetscStackCall("SuperLU_DIST:LUstructInit",LUstructInit(N, &lu->LUstruct));
438   F->ops->lufactornumeric = MatLUFactorNumeric_SuperLU_DIST;
439   F->ops->solve           = MatSolve_SuperLU_DIST;
440   F->ops->matsolve        = MatMatSolve_SuperLU_DIST;
441   F->ops->getinertia      = NULL;
442 
443   if (A->symmetric || A->hermitian) F->ops->getinertia = MatGetInertia_SuperLU_DIST;
444   lu->CleanUpSuperLU_Dist = PETSC_TRUE;
445   PetscFunctionReturn(0);
446 }
447 
448 static PetscErrorCode MatCholeskyFactorSymbolic_SuperLU_DIST(Mat F,Mat A,IS r,const MatFactorInfo *info)
449 {
450   PetscErrorCode ierr;
451 
452   PetscFunctionBegin;
453   ierr = MatLUFactorSymbolic_SuperLU_DIST(F,A,r,r,info);CHKERRQ(ierr);
454   F->ops->choleskyfactornumeric = MatLUFactorNumeric_SuperLU_DIST;
455   PetscFunctionReturn(0);
456 }
457 
458 static PetscErrorCode MatFactorGetSolverType_aij_superlu_dist(Mat A,MatSolverType *type)
459 {
460   PetscFunctionBegin;
461   *type = MATSOLVERSUPERLU_DIST;
462   PetscFunctionReturn(0);
463 }
464 
465 static PetscErrorCode MatView_Info_SuperLU_DIST(Mat A,PetscViewer viewer)
466 {
467   Mat_SuperLU_DIST       *lu=(Mat_SuperLU_DIST*)A->data;
468   superlu_dist_options_t options;
469   PetscErrorCode         ierr;
470 
471   PetscFunctionBegin;
472   /* check if matrix is superlu_dist type */
473   if (A->ops->solve != MatSolve_SuperLU_DIST) PetscFunctionReturn(0);
474 
475   options = lu->options;
476   ierr    = PetscViewerASCIIPrintf(viewer,"SuperLU_DIST run parameters:\n");CHKERRQ(ierr);
477   ierr    = PetscViewerASCIIPrintf(viewer,"  Process grid nprow %D x npcol %D \n",lu->nprow,lu->npcol);CHKERRQ(ierr);
478   ierr    = PetscViewerASCIIPrintf(viewer,"  Equilibrate matrix %s \n",PetscBools[options.Equil != NO]);CHKERRQ(ierr);
479   ierr    = PetscViewerASCIIPrintf(viewer,"  Replace tiny pivots %s \n",PetscBools[options.ReplaceTinyPivot != NO]);CHKERRQ(ierr);
480   ierr    = PetscViewerASCIIPrintf(viewer,"  Use iterative refinement %s \n",PetscBools[options.IterRefine == SLU_DOUBLE]);CHKERRQ(ierr);
481   ierr    = PetscViewerASCIIPrintf(viewer,"  Processors in row %d col partition %d \n",lu->nprow,lu->npcol);CHKERRQ(ierr);
482 
483   switch (options.RowPerm) {
484   case NOROWPERM:
485     ierr = PetscViewerASCIIPrintf(viewer,"  Row permutation NOROWPERM\n");CHKERRQ(ierr);
486     break;
487   case LargeDiag_MC64:
488     ierr = PetscViewerASCIIPrintf(viewer,"  Row permutation LargeDiag_MC64\n");CHKERRQ(ierr);
489     break;
490   case LargeDiag_AWPM:
491     ierr = PetscViewerASCIIPrintf(viewer,"  Row permutation LargeDiag_AWPM\n");CHKERRQ(ierr);
492     break;
493   case MY_PERMR:
494     ierr = PetscViewerASCIIPrintf(viewer,"  Row permutation MY_PERMR\n");CHKERRQ(ierr);
495     break;
496   default:
497     SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown column permutation");
498   }
499 
500   switch (options.ColPerm) {
501   case NATURAL:
502     ierr = PetscViewerASCIIPrintf(viewer,"  Column permutation NATURAL\n");CHKERRQ(ierr);
503     break;
504   case MMD_AT_PLUS_A:
505     ierr = PetscViewerASCIIPrintf(viewer,"  Column permutation MMD_AT_PLUS_A\n");CHKERRQ(ierr);
506     break;
507   case MMD_ATA:
508     ierr = PetscViewerASCIIPrintf(viewer,"  Column permutation MMD_ATA\n");CHKERRQ(ierr);
509     break;
510   /*  Even though this is called METIS, the SuperLU_DIST code sets this by default if PARMETIS is defined, not METIS */
511   case METIS_AT_PLUS_A:
512     ierr = PetscViewerASCIIPrintf(viewer,"  Column permutation METIS_AT_PLUS_A\n");CHKERRQ(ierr);
513     break;
514   case PARMETIS:
515     ierr = PetscViewerASCIIPrintf(viewer,"  Column permutation PARMETIS\n");CHKERRQ(ierr);
516     break;
517   default:
518     SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown column permutation");
519   }
520 
521   ierr = PetscViewerASCIIPrintf(viewer,"  Parallel symbolic factorization %s \n",PetscBools[options.ParSymbFact != NO]);CHKERRQ(ierr);
522 
523   if (lu->FactPattern == SamePattern) {
524     ierr = PetscViewerASCIIPrintf(viewer,"  Repeated factorization SamePattern\n");CHKERRQ(ierr);
525   } else if (lu->FactPattern == SamePattern_SameRowPerm) {
526     ierr = PetscViewerASCIIPrintf(viewer,"  Repeated factorization SamePattern_SameRowPerm\n");CHKERRQ(ierr);
527   } else if (lu->FactPattern == DOFACT) {
528     ierr = PetscViewerASCIIPrintf(viewer,"  Repeated factorization DOFACT\n");CHKERRQ(ierr);
529   } else {
530     SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown factorization pattern");
531   }
532   PetscFunctionReturn(0);
533 }
534 
535 static PetscErrorCode MatView_SuperLU_DIST(Mat A,PetscViewer viewer)
536 {
537   PetscErrorCode    ierr;
538   PetscBool         iascii;
539   PetscViewerFormat format;
540 
541   PetscFunctionBegin;
542   ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);CHKERRQ(ierr);
543   if (iascii) {
544     ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr);
545     if (format == PETSC_VIEWER_ASCII_INFO) {
546       ierr = MatView_Info_SuperLU_DIST(A,viewer);CHKERRQ(ierr);
547     }
548   }
549   PetscFunctionReturn(0);
550 }
551 
552 static PetscErrorCode MatGetFactor_aij_superlu_dist(Mat A,MatFactorType ftype,Mat *F)
553 {
554   Mat                    B;
555   Mat_SuperLU_DIST       *lu;
556   PetscErrorCode         ierr;
557   PetscInt               M=A->rmap->N,N=A->cmap->N,indx;
558   PetscMPIInt            size;
559   superlu_dist_options_t options;
560   PetscBool              flg;
561   const char             *colperm[]     = {"NATURAL","MMD_AT_PLUS_A","MMD_ATA","METIS_AT_PLUS_A","PARMETIS"};
562   const char             *rowperm[]     = {"NOROWPERM","LargeDiag_MC64","LargeDiag_AWPM","MY_PERMR"};
563   const char             *factPattern[] = {"SamePattern","SamePattern_SameRowPerm","DOFACT"};
564   PetscBool              set;
565 
566   PetscFunctionBegin;
567   /* Create the factorization matrix */
568   ierr = MatCreate(PetscObjectComm((PetscObject)A),&B);CHKERRQ(ierr);
569   ierr = MatSetSizes(B,A->rmap->n,A->cmap->n,M,N);CHKERRQ(ierr);
570   ierr = PetscStrallocpy("superlu_dist",&((PetscObject)B)->type_name);CHKERRQ(ierr);
571   ierr = MatSetUp(B);CHKERRQ(ierr);
572   B->ops->getinfo = MatGetInfo_External;
573   B->ops->view    = MatView_SuperLU_DIST;
574   B->ops->destroy = MatDestroy_SuperLU_DIST;
575 
576   /* Set the default input options:
577      options.Fact              = DOFACT;
578      options.Equil             = YES;
579      options.ParSymbFact       = NO;
580      options.ColPerm           = METIS_AT_PLUS_A;
581      options.RowPerm           = LargeDiag_MC64;
582      options.ReplaceTinyPivot  = YES;
583      options.IterRefine        = DOUBLE;
584      options.Trans             = NOTRANS;
585      options.SolveInitialized  = NO; -hold the communication pattern used MatSolve() and MatMatSolve()
586      options.RefineInitialized = NO;
587      options.PrintStat         = YES;
588      options.SymPattern        = NO;
589   */
590   set_default_options_dist(&options);
591 
592   B->trivialsymbolic = PETSC_TRUE;
593   if (ftype == MAT_FACTOR_LU) {
594     B->factortype = MAT_FACTOR_LU;
595     B->ops->lufactorsymbolic = MatLUFactorSymbolic_SuperLU_DIST;
596   } else {
597     B->factortype = MAT_FACTOR_CHOLESKY;
598     B->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SuperLU_DIST;
599     options.SymPattern = YES;
600   }
601 
602   /* set solvertype */
603   ierr = PetscFree(B->solvertype);CHKERRQ(ierr);
604   ierr = PetscStrallocpy(MATSOLVERSUPERLU_DIST,&B->solvertype);CHKERRQ(ierr);
605 
606   ierr    = PetscNewLog(B,&lu);CHKERRQ(ierr);
607   B->data = lu;
608   ierr = MPI_Comm_size(PetscObjectComm((PetscObject)A),&size);CHKERRMPI(ierr);
609 
610   {
611     PetscMPIInt       flg;
612     MPI_Comm          comm;
613     PetscSuperLU_DIST *context = NULL;
614 
615     ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr);
616     if (Petsc_Superlu_dist_keyval == MPI_KEYVAL_INVALID) {
617       ierr = MPI_Comm_create_keyval(MPI_COMM_NULL_COPY_FN,Petsc_Superlu_dist_keyval_Delete_Fn,&Petsc_Superlu_dist_keyval,(void*)0);CHKERRMPI(ierr);
618       ierr = PetscRegisterFinalize(Petsc_Superlu_dist_keyval_free);CHKERRQ(ierr);
619     }
620     ierr = MPI_Comm_get_attr(comm,Petsc_Superlu_dist_keyval,&context,&flg);CHKERRMPI(ierr);
621     if (!flg || context->busy) {
622       if (!flg) {
623         ierr = PetscNew(&context);CHKERRQ(ierr);
624         context->busy = PETSC_TRUE;
625         ierr = MPI_Comm_dup(comm,&context->comm);CHKERRMPI(ierr);
626         ierr = MPI_Comm_set_attr(comm,Petsc_Superlu_dist_keyval,context);CHKERRMPI(ierr);
627       } else {
628         ierr = MPI_Comm_dup(comm,&lu->comm_superlu);CHKERRMPI(ierr);
629       }
630 
631       /* Default num of process columns and rows */
632       lu->nprow = (int_t) (0.5 + PetscSqrtReal((PetscReal)size));
633       if (!lu->nprow) lu->nprow = 1;
634       while (lu->nprow > 0) {
635         lu->npcol = (int_t) (size/lu->nprow);
636         if (size == lu->nprow * lu->npcol) break;
637         lu->nprow--;
638       }
639       ierr = PetscOptionsBegin(PetscObjectComm((PetscObject)A),((PetscObject)A)->prefix,"SuperLU_Dist Options","Mat");CHKERRQ(ierr);
640       ierr = PetscOptionsInt("-mat_superlu_dist_r","Number rows in processor partition","None",lu->nprow,(PetscInt*)&lu->nprow,NULL);CHKERRQ(ierr);
641       ierr = PetscOptionsInt("-mat_superlu_dist_c","Number columns in processor partition","None",lu->npcol,(PetscInt*)&lu->npcol,NULL);CHKERRQ(ierr);
642       ierr = PetscOptionsEnd();CHKERRQ(ierr);
643       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);
644       PetscStackCall("SuperLU_DIST:superlu_gridinit",superlu_gridinit(context ? context->comm : lu->comm_superlu, lu->nprow, lu->npcol, &lu->grid));
645       if (context) context->grid = lu->grid;
646       ierr = PetscInfo(NULL,"Duplicating a communicator for SuperLU_DIST and calling superlu_gridinit()\n");
647       if (!flg) {
648         ierr = PetscInfo(NULL,"Storing communicator and SuperLU_DIST grid in communicator attribute\n");
649       } else {
650         ierr = PetscInfo(NULL,"Communicator attribute already in use so not saving communicator and SuperLU_DIST grid in communicator attribute \n");
651       }
652     } else {
653       ierr = PetscInfo(NULL,"Reusing communicator and superlu_gridinit() for SuperLU_DIST from communicator attribute.");
654       context->busy = PETSC_TRUE;
655       lu->grid      = context->grid;
656     }
657   }
658 
659   ierr = PetscOptionsBegin(PetscObjectComm((PetscObject)A),((PetscObject)A)->prefix,"SuperLU_Dist Options","Mat");CHKERRQ(ierr);
660   ierr = PetscOptionsBool("-mat_superlu_dist_equil","Equilibrate matrix","None",options.Equil ? PETSC_TRUE : PETSC_FALSE,&flg,&set);CHKERRQ(ierr);
661   if (set && !flg) options.Equil = NO;
662 
663   ierr = PetscOptionsEList("-mat_superlu_dist_rowperm","Row permutation","None",rowperm,4,rowperm[1],&indx,&flg);CHKERRQ(ierr);
664   if (flg) {
665     switch (indx) {
666     case 0:
667       options.RowPerm = NOROWPERM;
668       break;
669     case 1:
670       options.RowPerm = LargeDiag_MC64;
671       break;
672     case 2:
673       options.RowPerm = LargeDiag_AWPM;
674       break;
675     case 3:
676       options.RowPerm = MY_PERMR;
677       break;
678     default:
679       SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown row permutation");
680     }
681   }
682 
683   ierr = PetscOptionsEList("-mat_superlu_dist_colperm","Column permutation","None",colperm,5,colperm[3],&indx,&flg);CHKERRQ(ierr);
684   if (flg) {
685     switch (indx) {
686     case 0:
687       options.ColPerm = NATURAL;
688       break;
689     case 1:
690       options.ColPerm = MMD_AT_PLUS_A;
691       break;
692     case 2:
693       options.ColPerm = MMD_ATA;
694       break;
695     case 3:
696       options.ColPerm = METIS_AT_PLUS_A;
697       break;
698     case 4:
699       options.ColPerm = PARMETIS;   /* only works for np>1 */
700       break;
701     default:
702       SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown column permutation");
703     }
704   }
705 
706   options.ReplaceTinyPivot = NO;
707   ierr = PetscOptionsBool("-mat_superlu_dist_replacetinypivot","Replace tiny pivots","None",options.ReplaceTinyPivot ? PETSC_TRUE : PETSC_FALSE,&flg,&set);CHKERRQ(ierr);
708   if (set && flg) options.ReplaceTinyPivot = YES;
709 
710   options.ParSymbFact = NO;
711   ierr = PetscOptionsBool("-mat_superlu_dist_parsymbfact","Parallel symbolic factorization","None",PETSC_FALSE,&flg,&set);CHKERRQ(ierr);
712   if (set && flg && size>1) {
713 #if defined(PETSC_HAVE_PARMETIS)
714     options.ParSymbFact = YES;
715     options.ColPerm     = PARMETIS;   /* in v2.2, PARMETIS is forced for ParSymbFact regardless of user ordering setting */
716 #else
717     SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"parsymbfact needs PARMETIS");
718 #endif
719   }
720 
721   lu->FactPattern = SamePattern;
722   ierr = PetscOptionsEList("-mat_superlu_dist_fact","Sparsity pattern for repeated matrix factorization","None",factPattern,3,factPattern[0],&indx,&flg);CHKERRQ(ierr);
723   if (flg) {
724     switch (indx) {
725     case 0:
726       lu->FactPattern = SamePattern;
727       break;
728     case 1:
729       lu->FactPattern = SamePattern_SameRowPerm;
730       break;
731     case 2:
732       lu->FactPattern = DOFACT;
733       break;
734     }
735   }
736 
737   options.IterRefine = NOREFINE;
738   ierr               = PetscOptionsBool("-mat_superlu_dist_iterrefine","Use iterative refinement","None",options.IterRefine == NOREFINE ? PETSC_FALSE : PETSC_TRUE ,&flg,&set);CHKERRQ(ierr);
739   if (set) {
740     if (flg) options.IterRefine = SLU_DOUBLE;
741     else options.IterRefine = NOREFINE;
742   }
743 
744   if (PetscLogPrintInfo) options.PrintStat = YES;
745   else options.PrintStat = NO;
746   ierr = PetscOptionsBool("-mat_superlu_dist_statprint","Print factorization information","None",(PetscBool)options.PrintStat,(PetscBool*)&options.PrintStat,NULL);CHKERRQ(ierr);
747   ierr = PetscOptionsEnd();CHKERRQ(ierr);
748 
749   lu->options              = options;
750   lu->options.Fact         = DOFACT;
751   lu->matsolve_iscalled    = PETSC_FALSE;
752   lu->matmatsolve_iscalled = PETSC_FALSE;
753 
754   ierr = PetscObjectComposeFunction((PetscObject)B,"MatFactorGetSolverType_C",MatFactorGetSolverType_aij_superlu_dist);CHKERRQ(ierr);
755   ierr = PetscObjectComposeFunction((PetscObject)B,"MatSuperluDistGetDiagU_C",MatSuperluDistGetDiagU_SuperLU_DIST);CHKERRQ(ierr);
756 
757   *F = B;
758   PetscFunctionReturn(0);
759 }
760 
761 PETSC_EXTERN PetscErrorCode MatSolverTypeRegister_SuperLU_DIST(void)
762 {
763   PetscErrorCode ierr;
764   PetscFunctionBegin;
765   ierr = MatSolverTypeRegister(MATSOLVERSUPERLU_DIST,MATMPIAIJ,MAT_FACTOR_LU,MatGetFactor_aij_superlu_dist);CHKERRQ(ierr);
766   ierr = MatSolverTypeRegister(MATSOLVERSUPERLU_DIST,MATSEQAIJ,MAT_FACTOR_LU,MatGetFactor_aij_superlu_dist);CHKERRQ(ierr);
767   ierr = MatSolverTypeRegister(MATSOLVERSUPERLU_DIST,MATMPIAIJ,MAT_FACTOR_CHOLESKY,MatGetFactor_aij_superlu_dist);CHKERRQ(ierr);
768   ierr = MatSolverTypeRegister(MATSOLVERSUPERLU_DIST,MATSEQAIJ,MAT_FACTOR_CHOLESKY,MatGetFactor_aij_superlu_dist);CHKERRQ(ierr);
769   PetscFunctionReturn(0);
770 }
771 
772 /*MC
773   MATSOLVERSUPERLU_DIST - Parallel direct solver package for LU factorization
774 
775   Use ./configure --download-superlu_dist --download-parmetis --download-metis --download-ptscotch  to have PETSc installed with SuperLU_DIST
776 
777   Use -pc_type lu -pc_factor_mat_solver_type superlu_dist to use this direct solver
778 
779    Works with AIJ matrices
780 
781   Options Database Keys:
782 + -mat_superlu_dist_r <n> - number of rows in processor partition
783 . -mat_superlu_dist_c <n> - number of columns in processor partition
784 . -mat_superlu_dist_equil - equilibrate the matrix
785 . -mat_superlu_dist_rowperm <NOROWPERM,LargeDiag_MC64,LargeDiag_AWPM,MY_PERMR> - row permutation
786 . -mat_superlu_dist_colperm <NATURAL,MMD_AT_PLUS_A,MMD_ATA,METIS_AT_PLUS_A,PARMETIS> - column permutation
787 . -mat_superlu_dist_replacetinypivot - replace tiny pivots
788 . -mat_superlu_dist_fact <SamePattern> - (choose one of) SamePattern SamePattern_SameRowPerm DOFACT
789 . -mat_superlu_dist_iterrefine - use iterative refinement
790 - -mat_superlu_dist_statprint - print factorization information
791 
792    Level: beginner
793 
794 .seealso: PCLU
795 
796 .seealso: PCFactorSetMatSolverType(), MatSolverType
797 
798 M*/
799