xref: /petsc/src/mat/impls/aij/seq/superlu/superlu.c (revision 834855d6effb0d027771461c8e947ee1ce5a1e17)
1 /*
2      This file implements a subclass of the SeqAIJ matrix class that uses
3      the SuperLU sparse solver.
4 */
5 
6 /*
7      Defines the data structure for the base matrix type (SeqAIJ)
8 */
9 #include <../src/mat/impls/aij/seq/aij.h> /*I "petscmat.h" I*/
10 
11 /*
12      SuperLU include files
13 */
14 EXTERN_C_BEGIN
15 #if defined(PETSC_USE_COMPLEX)
16   #if defined(PETSC_USE_REAL_SINGLE)
17     #include <slu_cdefs.h>
18   #else
19     #include <slu_zdefs.h>
20   #endif
21 #else
22   #if defined(PETSC_USE_REAL_SINGLE)
23     #include <slu_sdefs.h>
24   #else
25     #include <slu_ddefs.h>
26   #endif
27 #endif
28 #include <slu_util.h>
29 EXTERN_C_END
30 
31 /*
32      This is the data that defines the SuperLU factored matrix type
33 */
34 typedef struct {
35   SuperMatrix       A, L, U, B, X;
36   superlu_options_t options;
37   PetscInt         *perm_c; /* column permutation vector */
38   PetscInt         *perm_r; /* row permutations from partial pivoting */
39   PetscInt         *etree;
40   PetscReal        *R, *C;
41   char              equed[1];
42   PetscInt          lwork;
43   void             *work;
44   PetscReal         rpg, rcond;
45   mem_usage_t       mem_usage;
46   MatStructure      flg;
47   SuperLUStat_t     stat;
48   Mat               A_dup;
49   PetscScalar      *rhs_dup;
50   GlobalLU_t        Glu;
51   PetscBool         needconversion;
52 
53   /* Flag to clean up (non-global) SuperLU objects during Destroy */
54   PetscBool CleanUpSuperLU;
55 } Mat_SuperLU;
56 
57 /*
58     Utility function
59 */
MatView_Info_SuperLU(Mat A,PetscViewer viewer)60 static PetscErrorCode MatView_Info_SuperLU(Mat A, PetscViewer viewer)
61 {
62   Mat_SuperLU      *lu = (Mat_SuperLU *)A->data;
63   superlu_options_t options;
64 
65   PetscFunctionBegin;
66   options = lu->options;
67 
68   PetscCall(PetscViewerASCIIPrintf(viewer, "SuperLU run parameters:\n"));
69   PetscCall(PetscViewerASCIIPrintf(viewer, "  Equil: %s\n", (options.Equil != NO) ? "YES" : "NO"));
70   PetscCall(PetscViewerASCIIPrintf(viewer, "  ColPerm: %" PetscInt_FMT "\n", options.ColPerm));
71   PetscCall(PetscViewerASCIIPrintf(viewer, "  IterRefine: %" PetscInt_FMT "\n", options.IterRefine));
72   PetscCall(PetscViewerASCIIPrintf(viewer, "  SymmetricMode: %s\n", (options.SymmetricMode != NO) ? "YES" : "NO"));
73   PetscCall(PetscViewerASCIIPrintf(viewer, "  DiagPivotThresh: %g\n", options.DiagPivotThresh));
74   PetscCall(PetscViewerASCIIPrintf(viewer, "  PivotGrowth: %s\n", (options.PivotGrowth != NO) ? "YES" : "NO"));
75   PetscCall(PetscViewerASCIIPrintf(viewer, "  ConditionNumber: %s\n", (options.ConditionNumber != NO) ? "YES" : "NO"));
76   PetscCall(PetscViewerASCIIPrintf(viewer, "  RowPerm: %" PetscInt_FMT "\n", options.RowPerm));
77   PetscCall(PetscViewerASCIIPrintf(viewer, "  ReplaceTinyPivot: %s\n", (options.ReplaceTinyPivot != NO) ? "YES" : "NO"));
78   PetscCall(PetscViewerASCIIPrintf(viewer, "  PrintStat: %s\n", (options.PrintStat != NO) ? "YES" : "NO"));
79   PetscCall(PetscViewerASCIIPrintf(viewer, "  lwork: %" PetscInt_FMT "\n", lu->lwork));
80   if (A->factortype == MAT_FACTOR_ILU) {
81     PetscCall(PetscViewerASCIIPrintf(viewer, "  ILU_DropTol: %g\n", options.ILU_DropTol));
82     PetscCall(PetscViewerASCIIPrintf(viewer, "  ILU_FillTol: %g\n", options.ILU_FillTol));
83     PetscCall(PetscViewerASCIIPrintf(viewer, "  ILU_FillFactor: %g\n", options.ILU_FillFactor));
84     PetscCall(PetscViewerASCIIPrintf(viewer, "  ILU_DropRule: %" PetscInt_FMT "\n", options.ILU_DropRule));
85     PetscCall(PetscViewerASCIIPrintf(viewer, "  ILU_Norm: %" PetscInt_FMT "\n", options.ILU_Norm));
86     PetscCall(PetscViewerASCIIPrintf(viewer, "  ILU_MILU: %" PetscInt_FMT "\n", options.ILU_MILU));
87   }
88   PetscFunctionReturn(PETSC_SUCCESS);
89 }
90 
MatSolve_SuperLU_Private(Mat A,Vec b,Vec x)91 static PetscErrorCode MatSolve_SuperLU_Private(Mat A, Vec b, Vec x)
92 {
93   Mat_SuperLU       *lu = (Mat_SuperLU *)A->data;
94   const PetscScalar *barray;
95   PetscScalar       *xarray;
96   PetscInt           info, i, n;
97   PetscReal          ferr, berr;
98   static PetscBool   cite = PETSC_FALSE;
99 
100   PetscFunctionBegin;
101   if (lu->lwork == -1) PetscFunctionReturn(PETSC_SUCCESS);
102   PetscCall(PetscCitationsRegister("@article{superlu99,\n  author  = {James W. Demmel and Stanley C. Eisenstat and\n             John R. Gilbert and Xiaoye S. Li and Joseph W. H. Liu},\n  title = {A supernodal approach to sparse partial "
103                                    "pivoting},\n  journal = {SIAM J. Matrix Analysis and Applications},\n  year = {1999},\n  volume  = {20},\n  number = {3},\n  pages = {720-755}\n}\n",
104                                    &cite));
105 
106   PetscCall(VecGetLocalSize(x, &n));
107   lu->B.ncol = 1; /* Set the number of right-hand side */
108   if (lu->options.Equil && !lu->rhs_dup) {
109     /* superlu overwrites b when Equil is used, thus create rhs_dup to keep user's b unchanged */
110     PetscCall(PetscMalloc1(n, &lu->rhs_dup));
111   }
112   if (lu->options.Equil) {
113     /* Copy b into rsh_dup */
114     PetscCall(VecGetArrayRead(b, &barray));
115     PetscCall(PetscArraycpy(lu->rhs_dup, barray, n));
116     PetscCall(VecRestoreArrayRead(b, &barray));
117     barray = lu->rhs_dup;
118   } else {
119     PetscCall(VecGetArrayRead(b, &barray));
120   }
121   PetscCall(VecGetArray(x, &xarray));
122 
123 #if defined(PETSC_USE_COMPLEX)
124   #if defined(PETSC_USE_REAL_SINGLE)
125   ((DNformat *)lu->B.Store)->nzval = (singlecomplex *)barray;
126   ((DNformat *)lu->X.Store)->nzval = (singlecomplex *)xarray;
127   #else
128   ((DNformat *)lu->B.Store)->nzval = (doublecomplex *)barray;
129   ((DNformat *)lu->X.Store)->nzval = (doublecomplex *)xarray;
130   #endif
131 #else
132   ((DNformat *)lu->B.Store)->nzval = (void *)barray;
133   ((DNformat *)lu->X.Store)->nzval = xarray;
134 #endif
135 
136   lu->options.Fact = FACTORED; /* Indicate the factored form of A is supplied. */
137   if (A->factortype == MAT_FACTOR_LU) {
138 #if defined(PETSC_USE_COMPLEX)
139   #if defined(PETSC_USE_REAL_SINGLE)
140     PetscStackCallExternalVoid("SuperLU:cgssvx", cgssvx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &ferr, &berr, &lu->Glu, &lu->mem_usage, &lu->stat, &info));
141   #else
142     PetscStackCallExternalVoid("SuperLU:zgssvx", zgssvx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &ferr, &berr, &lu->Glu, &lu->mem_usage, &lu->stat, &info));
143   #endif
144 #else
145   #if defined(PETSC_USE_REAL_SINGLE)
146     PetscStackCallExternalVoid("SuperLU:sgssvx", sgssvx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &ferr, &berr, &lu->Glu, &lu->mem_usage, &lu->stat, &info));
147   #else
148     PetscStackCallExternalVoid("SuperLU:dgssvx", dgssvx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &ferr, &berr, &lu->Glu, &lu->mem_usage, &lu->stat, &info));
149   #endif
150 #endif
151   } else if (A->factortype == MAT_FACTOR_ILU) {
152 #if defined(PETSC_USE_COMPLEX)
153   #if defined(PETSC_USE_REAL_SINGLE)
154     PetscStackCallExternalVoid("SuperLU:cgsisx", cgsisx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &lu->Glu, &lu->mem_usage, &lu->stat, &info));
155   #else
156     PetscStackCallExternalVoid("SuperLU:zgsisx", zgsisx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &lu->Glu, &lu->mem_usage, &lu->stat, &info));
157   #endif
158 #else
159   #if defined(PETSC_USE_REAL_SINGLE)
160     PetscStackCallExternalVoid("SuperLU:sgsisx", sgsisx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &lu->Glu, &lu->mem_usage, &lu->stat, &info));
161   #else
162     PetscStackCallExternalVoid("SuperLU:dgsisx", dgsisx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &lu->Glu, &lu->mem_usage, &lu->stat, &info));
163   #endif
164 #endif
165   } else SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "Factor type not supported");
166   if (!lu->options.Equil) PetscCall(VecRestoreArrayRead(b, &barray));
167   PetscCall(VecRestoreArray(x, &xarray));
168 
169   if (!info || info == lu->A.ncol + 1) {
170     if (lu->options.IterRefine) {
171       PetscCall(PetscPrintf(PETSC_COMM_SELF, "Iterative Refinement:\n"));
172       PetscCall(PetscPrintf(PETSC_COMM_SELF, "  %8s%8s%16s%16s\n", "rhs", "Steps", "FERR", "BERR"));
173       for (i = 0; i < 1; ++i) PetscCall(PetscPrintf(PETSC_COMM_SELF, "  %8d%8d%16e%16e\n", i + 1, lu->stat.RefineSteps, ferr, berr));
174     }
175   } else if (info > 0) {
176     if (lu->lwork == -1) {
177       PetscCall(PetscPrintf(PETSC_COMM_SELF, "  ** Estimated memory: %" PetscInt_FMT " bytes\n", info - lu->A.ncol));
178     } else {
179       PetscCall(PetscPrintf(PETSC_COMM_SELF, "  Warning: gssvx() returns info %" PetscInt_FMT "\n", info));
180     }
181   } else PetscCheck(info >= 0, PETSC_COMM_SELF, PETSC_ERR_LIB, "info = %" PetscInt_FMT ", the %" PetscInt_FMT "-th argument in gssvx() had an illegal value", info, -info);
182 
183   if (lu->options.PrintStat) {
184     PetscCall(PetscPrintf(PETSC_COMM_SELF, "MatSolve__SuperLU():\n"));
185     PetscStackCallExternalVoid("SuperLU:StatPrint", StatPrint(&lu->stat));
186   }
187   PetscFunctionReturn(PETSC_SUCCESS);
188 }
189 
MatSolve_SuperLU(Mat A,Vec b,Vec x)190 static PetscErrorCode MatSolve_SuperLU(Mat A, Vec b, Vec x)
191 {
192   Mat_SuperLU *lu = (Mat_SuperLU *)A->data;
193   trans_t      oldOption;
194 
195   PetscFunctionBegin;
196   PetscCall(VecFlag(x, A->factorerrortype));
197   if (A->factorerrortype) {
198     PetscCall(PetscInfo(A, "MatSolve is called with singular matrix factor, skip\n"));
199     PetscFunctionReturn(PETSC_SUCCESS);
200   }
201 
202   oldOption         = lu->options.Trans;
203   lu->options.Trans = TRANS;
204   PetscCall(MatSolve_SuperLU_Private(A, b, x));
205   lu->options.Trans = oldOption;
206   PetscFunctionReturn(PETSC_SUCCESS);
207 }
208 
MatSolveTranspose_SuperLU(Mat A,Vec b,Vec x)209 static PetscErrorCode MatSolveTranspose_SuperLU(Mat A, Vec b, Vec x)
210 {
211   Mat_SuperLU *lu = (Mat_SuperLU *)A->data;
212   trans_t      oldOption;
213 
214   PetscFunctionBegin;
215   PetscCall(VecFlag(x, A->factorerrortype));
216   if (A->factorerrortype) {
217     PetscCall(PetscInfo(A, "MatSolve is called with singular matrix factor, skip\n"));
218     PetscFunctionReturn(PETSC_SUCCESS);
219   }
220 
221   oldOption         = lu->options.Trans;
222   lu->options.Trans = NOTRANS;
223   PetscCall(MatSolve_SuperLU_Private(A, b, x));
224   lu->options.Trans = oldOption;
225   PetscFunctionReturn(PETSC_SUCCESS);
226 }
227 
MatLUFactorNumeric_SuperLU(Mat F,Mat A,const MatFactorInfo * info)228 static PetscErrorCode MatLUFactorNumeric_SuperLU(Mat F, Mat A, const MatFactorInfo *info)
229 {
230   Mat_SuperLU *lu = (Mat_SuperLU *)F->data;
231   Mat_SeqAIJ  *aa;
232   PetscInt     sinfo;
233   PetscReal    ferr, berr;
234   NCformat    *Ustore;
235   SCformat    *Lstore;
236 
237   PetscFunctionBegin;
238   if (lu->flg == SAME_NONZERO_PATTERN) { /* successive numerical factorization */
239     lu->options.Fact = SamePattern;
240     /* Ref: ~SuperLU_3.0/EXAMPLE/dlinsolx2.c */
241     Destroy_SuperMatrix_Store(&lu->A);
242     if (lu->A_dup) PetscCall(MatCopy_SeqAIJ(A, lu->A_dup, SAME_NONZERO_PATTERN));
243 
244     if (lu->lwork >= 0) {
245       PetscStackCallExternalVoid("SuperLU:Destroy_SuperNode_Matrix", Destroy_SuperNode_Matrix(&lu->L));
246       PetscStackCallExternalVoid("SuperLU:Destroy_CompCol_Matrix", Destroy_CompCol_Matrix(&lu->U));
247       lu->options.Fact = SamePattern;
248     }
249   }
250 
251   /* Create the SuperMatrix for lu->A=A^T:
252        Since SuperLU likes column-oriented matrices,we pass it the transpose,
253        and then solve A^T X = B in MatSolve(). */
254   if (lu->A_dup) {
255     aa = (Mat_SeqAIJ *)lu->A_dup->data;
256   } else {
257     aa = (Mat_SeqAIJ *)A->data;
258   }
259 #if defined(PETSC_USE_COMPLEX)
260   #if defined(PETSC_USE_REAL_SINGLE)
261   PetscStackCallExternalVoid("SuperLU:cCreate_CompCol_Matrix", cCreate_CompCol_Matrix(&lu->A, A->cmap->n, A->rmap->n, aa->nz, (singlecomplex *)aa->a, aa->j, aa->i, SLU_NC, SLU_C, SLU_GE));
262   #else
263   PetscStackCallExternalVoid("SuperLU:zCreate_CompCol_Matrix", zCreate_CompCol_Matrix(&lu->A, A->cmap->n, A->rmap->n, aa->nz, (doublecomplex *)aa->a, aa->j, aa->i, SLU_NC, SLU_Z, SLU_GE));
264   #endif
265 #else
266   #if defined(PETSC_USE_REAL_SINGLE)
267   PetscStackCallExternalVoid("SuperLU:sCreate_CompCol_Matrix", sCreate_CompCol_Matrix(&lu->A, A->cmap->n, A->rmap->n, aa->nz, aa->a, aa->j, aa->i, SLU_NC, SLU_S, SLU_GE));
268   #else
269   PetscStackCallExternalVoid("SuperLU:dCreate_CompCol_Matrix", dCreate_CompCol_Matrix(&lu->A, A->cmap->n, A->rmap->n, aa->nz, aa->a, aa->j, aa->i, SLU_NC, SLU_D, SLU_GE));
270   #endif
271 #endif
272 
273   /* Numerical factorization */
274   lu->B.ncol = 0; /* Indicate not to solve the system */
275   if (F->factortype == MAT_FACTOR_LU) {
276 #if defined(PETSC_USE_COMPLEX)
277   #if defined(PETSC_USE_REAL_SINGLE)
278     PetscStackCallExternalVoid("SuperLU:cgssvx", cgssvx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &ferr, &berr, &lu->Glu, &lu->mem_usage, &lu->stat, &sinfo));
279   #else
280     PetscStackCallExternalVoid("SuperLU:zgssvx", zgssvx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &ferr, &berr, &lu->Glu, &lu->mem_usage, &lu->stat, &sinfo));
281   #endif
282 #else
283   #if defined(PETSC_USE_REAL_SINGLE)
284     PetscStackCallExternalVoid("SuperLU:sgssvx", sgssvx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &ferr, &berr, &lu->Glu, &lu->mem_usage, &lu->stat, &sinfo));
285   #else
286     PetscStackCallExternalVoid("SuperLU:dgssvx", dgssvx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &ferr, &berr, &lu->Glu, &lu->mem_usage, &lu->stat, &sinfo));
287   #endif
288 #endif
289   } else if (F->factortype == MAT_FACTOR_ILU) {
290     /* Compute the incomplete factorization, condition number and pivot growth */
291 #if defined(PETSC_USE_COMPLEX)
292   #if defined(PETSC_USE_REAL_SINGLE)
293     PetscStackCallExternalVoid("SuperLU:cgsisx", cgsisx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &lu->Glu, &lu->mem_usage, &lu->stat, &sinfo));
294   #else
295     PetscStackCallExternalVoid("SuperLU:zgsisx", zgsisx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &lu->Glu, &lu->mem_usage, &lu->stat, &sinfo));
296   #endif
297 #else
298   #if defined(PETSC_USE_REAL_SINGLE)
299     PetscStackCallExternalVoid("SuperLU:sgsisx", sgsisx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &lu->Glu, &lu->mem_usage, &lu->stat, &sinfo));
300   #else
301     PetscStackCallExternalVoid("SuperLU:dgsisx", dgsisx(&lu->options, &lu->A, lu->perm_c, lu->perm_r, lu->etree, lu->equed, lu->R, lu->C, &lu->L, &lu->U, lu->work, lu->lwork, &lu->B, &lu->X, &lu->rpg, &lu->rcond, &lu->Glu, &lu->mem_usage, &lu->stat, &sinfo));
302   #endif
303 #endif
304   } else SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "Factor type not supported");
305   if (!sinfo || sinfo == lu->A.ncol + 1) {
306     if (lu->options.PivotGrowth) PetscCall(PetscPrintf(PETSC_COMM_SELF, "  Recip. pivot growth = %e\n", lu->rpg));
307     if (lu->options.ConditionNumber) PetscCall(PetscPrintf(PETSC_COMM_SELF, "  Recip. condition number = %e\n", lu->rcond));
308   } else if (sinfo > 0) {
309     if (A->erroriffailure) {
310       SETERRQ(PETSC_COMM_SELF, PETSC_ERR_MAT_LU_ZRPVT, "Zero pivot in row %" PetscInt_FMT, sinfo);
311     } else {
312       if (sinfo <= lu->A.ncol) {
313         if (lu->options.ILU_FillTol == 0.0) F->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
314         PetscCall(PetscInfo(F, "Number of zero pivots %" PetscInt_FMT ", ILU_FillTol %g\n", sinfo, lu->options.ILU_FillTol));
315       } else if (sinfo == lu->A.ncol + 1) {
316         /*
317          U is nonsingular, but RCOND is less than machine
318                       precision, meaning that the matrix is singular to
319                       working precision. Nevertheless, the solution and
320                       error bounds are computed because there are a number
321                       of situations where the computed solution can be more
322                       accurate than the value of RCOND would suggest.
323          */
324         PetscCall(PetscInfo(F, "Matrix factor U is nonsingular, but is singular to working precision. The solution is computed. info %" PetscInt_FMT "\n", sinfo));
325       } else { /* sinfo > lu->A.ncol + 1 */
326         F->factorerrortype = MAT_FACTOR_OUTMEMORY;
327         PetscCall(PetscInfo(F, "Number of bytes allocated when memory allocation fails %" PetscInt_FMT "\n", sinfo));
328       }
329     }
330   } else SETERRQ(PETSC_COMM_SELF, PETSC_ERR_LIB, "info = %" PetscInt_FMT ", the %" PetscInt_FMT "-th argument in gssvx() had an illegal value", sinfo, -sinfo);
331 
332   if (lu->options.PrintStat) {
333     PetscCall(PetscPrintf(PETSC_COMM_SELF, "MatLUFactorNumeric_SuperLU():\n"));
334     PetscStackCallExternalVoid("SuperLU:StatPrint", StatPrint(&lu->stat));
335     Lstore = (SCformat *)lu->L.Store;
336     Ustore = (NCformat *)lu->U.Store;
337     PetscCall(PetscPrintf(PETSC_COMM_SELF, "  No of nonzeros in factor L = %" PetscInt_FMT "\n", Lstore->nnz));
338     PetscCall(PetscPrintf(PETSC_COMM_SELF, "  No of nonzeros in factor U = %" PetscInt_FMT "\n", Ustore->nnz));
339     PetscCall(PetscPrintf(PETSC_COMM_SELF, "  No of nonzeros in L+U = %" PetscInt_FMT "\n", Lstore->nnz + Ustore->nnz - lu->A.ncol));
340     PetscCall(PetscPrintf(PETSC_COMM_SELF, "  L\\U MB %.3f\ttotal MB needed %.3f\n", lu->mem_usage.for_lu / 1e6, lu->mem_usage.total_needed / 1e6));
341   }
342 
343   lu->flg                = SAME_NONZERO_PATTERN;
344   F->ops->solve          = MatSolve_SuperLU;
345   F->ops->solvetranspose = MatSolveTranspose_SuperLU;
346   F->ops->matsolve       = NULL;
347   PetscFunctionReturn(PETSC_SUCCESS);
348 }
349 
MatDestroy_SuperLU(Mat A)350 static PetscErrorCode MatDestroy_SuperLU(Mat A)
351 {
352   Mat_SuperLU *lu = (Mat_SuperLU *)A->data;
353 
354   PetscFunctionBegin;
355   if (lu->CleanUpSuperLU) { /* Free the SuperLU datastructures */
356     PetscStackCallExternalVoid("SuperLU:Destroy_SuperMatrix_Store", Destroy_SuperMatrix_Store(&lu->A));
357     if (lu->lwork >= 0) {
358       PetscStackCallExternalVoid("SuperLU:Destroy_SuperNode_Matrix", Destroy_SuperNode_Matrix(&lu->L));
359       PetscStackCallExternalVoid("SuperLU:Destroy_CompCol_Matrix", Destroy_CompCol_Matrix(&lu->U));
360     }
361   }
362   PetscStackCallExternalVoid("SuperLU:Destroy_SuperMatrix_Store", Destroy_SuperMatrix_Store(&lu->B));
363   PetscStackCallExternalVoid("SuperLU:Destroy_SuperMatrix_Store", Destroy_SuperMatrix_Store(&lu->X));
364   PetscStackCallExternalVoid("SuperLU:StatFree", StatFree(&lu->stat));
365   PetscCall(PetscFree(lu->etree));
366   PetscCall(PetscFree(lu->perm_r));
367   PetscCall(PetscFree(lu->perm_c));
368   PetscCall(PetscFree(lu->R));
369   PetscCall(PetscFree(lu->C));
370   PetscCall(PetscFree(lu->rhs_dup));
371   PetscCall(MatDestroy(&lu->A_dup));
372   PetscCall(PetscFree(A->data));
373 
374   /* clear composed functions */
375   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatFactorGetSolverType_C", NULL));
376   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatSuperluSetILUDropTol_C", NULL));
377   PetscFunctionReturn(PETSC_SUCCESS);
378 }
379 
MatView_SuperLU(Mat A,PetscViewer viewer)380 static PetscErrorCode MatView_SuperLU(Mat A, PetscViewer viewer)
381 {
382   PetscBool         isascii;
383   PetscViewerFormat format;
384 
385   PetscFunctionBegin;
386   PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &isascii));
387   if (isascii) {
388     PetscCall(PetscViewerGetFormat(viewer, &format));
389     if (format == PETSC_VIEWER_ASCII_INFO) PetscCall(MatView_Info_SuperLU(A, viewer));
390   }
391   PetscFunctionReturn(PETSC_SUCCESS);
392 }
393 
MatLUFactorSymbolic_SuperLU(Mat F,Mat A,IS r,IS c,const MatFactorInfo * info)394 static PetscErrorCode MatLUFactorSymbolic_SuperLU(Mat F, Mat A, IS r, IS c, const MatFactorInfo *info)
395 {
396   Mat_SuperLU *lu = (Mat_SuperLU *)F->data;
397   PetscInt     indx;
398   PetscBool    flg, set;
399   PetscReal    real_input;
400   const char  *colperm[]    = {"NATURAL", "MMD_ATA", "MMD_AT_PLUS_A", "COLAMD"}; /* MY_PERMC - not supported by the PETSc interface yet */
401   const char  *iterrefine[] = {"NOREFINE", "SINGLE", "DOUBLE", "EXTRA"};
402   const char  *rowperm[]    = {"NOROWPERM", "LargeDiag"}; /* MY_PERMC - not supported by the PETSc interface yet */
403 
404   PetscFunctionBegin;
405   /* Set options to F */
406   PetscOptionsBegin(PetscObjectComm((PetscObject)F), ((PetscObject)F)->prefix, "SuperLU Options", "Mat");
407   PetscCall(PetscOptionsBool("-mat_superlu_equil", "Equil", "None", (PetscBool)lu->options.Equil, (PetscBool *)&lu->options.Equil, NULL));
408   PetscCall(PetscOptionsEList("-mat_superlu_colperm", "ColPerm", "None", colperm, 4, colperm[3], &indx, &flg));
409   if (flg) lu->options.ColPerm = (colperm_t)indx;
410   PetscCall(PetscOptionsEList("-mat_superlu_iterrefine", "IterRefine", "None", iterrefine, 4, iterrefine[0], &indx, &flg));
411   if (flg) lu->options.IterRefine = (IterRefine_t)indx;
412   PetscCall(PetscOptionsBool("-mat_superlu_symmetricmode", "SymmetricMode", "None", (PetscBool)lu->options.SymmetricMode, &flg, &set));
413   if (set && flg) lu->options.SymmetricMode = YES;
414   PetscCall(PetscOptionsReal("-mat_superlu_diagpivotthresh", "DiagPivotThresh", "None", lu->options.DiagPivotThresh, &real_input, &flg));
415   if (flg) lu->options.DiagPivotThresh = (double)real_input;
416   PetscCall(PetscOptionsBool("-mat_superlu_pivotgrowth", "PivotGrowth", "None", (PetscBool)lu->options.PivotGrowth, &flg, &set));
417   if (set && flg) lu->options.PivotGrowth = YES;
418   PetscCall(PetscOptionsBool("-mat_superlu_conditionnumber", "ConditionNumber", "None", (PetscBool)lu->options.ConditionNumber, &flg, &set));
419   if (set && flg) lu->options.ConditionNumber = YES;
420   PetscCall(PetscOptionsEList("-mat_superlu_rowperm", "rowperm", "None", rowperm, 2, rowperm[lu->options.RowPerm], &indx, &flg));
421   if (flg) lu->options.RowPerm = (rowperm_t)indx;
422   PetscCall(PetscOptionsBool("-mat_superlu_replacetinypivot", "ReplaceTinyPivot", "None", (PetscBool)lu->options.ReplaceTinyPivot, &flg, &set));
423   if (set && flg) lu->options.ReplaceTinyPivot = YES;
424   PetscCall(PetscOptionsBool("-mat_superlu_printstat", "PrintStat", "None", (PetscBool)lu->options.PrintStat, &flg, &set));
425   if (set && flg) lu->options.PrintStat = YES;
426   PetscCall(PetscOptionsInt("-mat_superlu_lwork", "size of work array in bytes used by factorization", "None", lu->lwork, &lu->lwork, NULL));
427   if (lu->lwork > 0) {
428     /* lwork is in bytes, hence PetscMalloc() is used here, not PetscMalloc1()*/
429     PetscCall(PetscMalloc(lu->lwork, &lu->work));
430   } else if (lu->lwork != 0 && lu->lwork != -1) {
431     PetscCall(PetscPrintf(PETSC_COMM_SELF, "   Warning: lwork %" PetscInt_FMT " is not supported by SUPERLU. The default lwork=0 is used.\n", lu->lwork));
432     lu->lwork = 0;
433   }
434   /* ilu options */
435   PetscCall(PetscOptionsReal("-mat_superlu_ilu_droptol", "ILU_DropTol", "None", lu->options.ILU_DropTol, &real_input, &flg));
436   if (flg) lu->options.ILU_DropTol = (double)real_input;
437   PetscCall(PetscOptionsReal("-mat_superlu_ilu_filltol", "ILU_FillTol", "None", lu->options.ILU_FillTol, &real_input, &flg));
438   if (flg) lu->options.ILU_FillTol = (double)real_input;
439   PetscCall(PetscOptionsReal("-mat_superlu_ilu_fillfactor", "ILU_FillFactor", "None", lu->options.ILU_FillFactor, &real_input, &flg));
440   if (flg) lu->options.ILU_FillFactor = (double)real_input;
441   PetscCall(PetscOptionsInt("-mat_superlu_ilu_droprull", "ILU_DropRule", "None", lu->options.ILU_DropRule, &lu->options.ILU_DropRule, NULL));
442   PetscCall(PetscOptionsInt("-mat_superlu_ilu_norm", "ILU_Norm", "None", lu->options.ILU_Norm, &indx, &flg));
443   if (flg) lu->options.ILU_Norm = (norm_t)indx;
444   PetscCall(PetscOptionsInt("-mat_superlu_ilu_milu", "ILU_MILU", "None", lu->options.ILU_MILU, &indx, &flg));
445   if (flg) lu->options.ILU_MILU = (milu_t)indx;
446   PetscOptionsEnd();
447 
448   lu->flg                 = DIFFERENT_NONZERO_PATTERN;
449   lu->CleanUpSuperLU      = PETSC_TRUE;
450   F->ops->lufactornumeric = MatLUFactorNumeric_SuperLU;
451 
452   /* if we are here, the nonzero pattern has changed unless the user explicitly called MatLUFactorSymbolic */
453   PetscCall(MatDestroy(&lu->A_dup));
454   if (lu->needconversion) PetscCall(MatConvert(A, MATSEQAIJ, MAT_INITIAL_MATRIX, &lu->A_dup));
455   if (lu->options.Equil == YES && !lu->A_dup) { /* superlu overwrites input matrix and rhs when Equil is used, thus create A_dup to keep user's A unchanged */
456     PetscCall(MatDuplicate_SeqAIJ(A, MAT_COPY_VALUES, &lu->A_dup));
457   }
458   PetscFunctionReturn(PETSC_SUCCESS);
459 }
460 
MatSuperluSetILUDropTol_SuperLU(Mat F,PetscReal dtol)461 static PetscErrorCode MatSuperluSetILUDropTol_SuperLU(Mat F, PetscReal dtol)
462 {
463   Mat_SuperLU *lu = (Mat_SuperLU *)F->data;
464 
465   PetscFunctionBegin;
466   lu->options.ILU_DropTol = dtol;
467   PetscFunctionReturn(PETSC_SUCCESS);
468 }
469 
470 /*@
471   MatSuperluSetILUDropTol - Set SuperLU <https://portal.nersc.gov/project/sparse/superlu/superlu_ug.pdf> ILU drop tolerance
472 
473   Logically Collective
474 
475   Input Parameters:
476 + F    - the factored matrix obtained by calling `MatGetFactor()`
477 - dtol - drop tolerance
478 
479   Options Database Key:
480 . -mat_superlu_ilu_droptol <dtol> - the drop tolerance
481 
482   Level: beginner
483 
484 .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MATSOLVERSUPERLU`
485 @*/
MatSuperluSetILUDropTol(Mat F,PetscReal dtol)486 PetscErrorCode MatSuperluSetILUDropTol(Mat F, PetscReal dtol)
487 {
488   PetscFunctionBegin;
489   PetscValidHeaderSpecific(F, MAT_CLASSID, 1);
490   PetscValidLogicalCollectiveReal(F, dtol, 2);
491   PetscTryMethod(F, "MatSuperluSetILUDropTol_C", (Mat, PetscReal), (F, dtol));
492   PetscFunctionReturn(PETSC_SUCCESS);
493 }
494 
MatFactorGetSolverType_seqaij_superlu(Mat A,MatSolverType * type)495 static PetscErrorCode MatFactorGetSolverType_seqaij_superlu(Mat A, MatSolverType *type)
496 {
497   PetscFunctionBegin;
498   *type = MATSOLVERSUPERLU;
499   PetscFunctionReturn(PETSC_SUCCESS);
500 }
501 
502 /*MC
503   MATSOLVERSUPERLU = "superlu" - A solver package providing solvers LU and ILU for sequential matrices
504   via the external package SuperLU <https://portal.nersc.gov/project/sparse/superlu/superlu_ug.pdf>
505 
506   Use `./configure --download-superlu` to have PETSc installed with SuperLU
507 
508   Use `-pc_type lu` `-pc_factor_mat_solver_type superlu` to use this direct solver
509 
510   Options Database Keys:
511 + -mat_superlu_equil <FALSE>            - Equil (None)
512 . -mat_superlu_colperm <COLAMD>         - (choose one of) `NATURAL`, `MMD_ATA MMD_AT_PLUS_A`, `COLAMD`
513 . -mat_superlu_iterrefine <NOREFINE>    - (choose one of) `NOREFINE`, `SINGLE`, `DOUBLE`, `EXTRA`
514 . -mat_superlu_symmetricmode: <FALSE>   - SymmetricMode (None)
515 . -mat_superlu_diagpivotthresh <1>      - DiagPivotThresh (None)
516 . -mat_superlu_pivotgrowth <FALSE>      - PivotGrowth (None)
517 . -mat_superlu_conditionnumber <FALSE>  - ConditionNumber (None)
518 . -mat_superlu_rowperm <NOROWPERM>      - (choose one of) `NOROWPERM`, `LargeDiag`
519 . -mat_superlu_replacetinypivot <FALSE> - ReplaceTinyPivot (None)
520 . -mat_superlu_printstat <FALSE>        - PrintStat (None)
521 . -mat_superlu_lwork <0>                - size of work array in bytes used by factorization (None)
522 . -mat_superlu_ilu_droptol <0>          - ILU_DropTol (None)
523 . -mat_superlu_ilu_filltol <0>          - ILU_FillTol (None)
524 . -mat_superlu_ilu_fillfactor <0>       - ILU_FillFactor (None)
525 . -mat_superlu_ilu_droprull <0>         - ILU_DropRule (None)
526 . -mat_superlu_ilu_norm <0>             - ILU_Norm (None)
527 - -mat_superlu_ilu_milu <0>             - ILU_MILU (None)
528 
529    Level: beginner
530 
531    Notes:
532    Do not confuse this with `MATSOLVERSUPERLU_DIST` which is for parallel sparse solves
533 
534    Cannot use ordering provided by PETSc, provides its own.
535 
536 .seealso: [](ch_matrices), `Mat`, `PCLU`, `PCILU`, `MATSOLVERSUPERLU_DIST`, `MATSOLVERMUMPS`, `PCFactorSetMatSolverType()`, `MatSolverType`
537 M*/
538 
MatGetFactor_seqaij_superlu(Mat A,MatFactorType ftype,Mat * F)539 static PetscErrorCode MatGetFactor_seqaij_superlu(Mat A, MatFactorType ftype, Mat *F)
540 {
541   Mat          B;
542   Mat_SuperLU *lu;
543   PetscInt     m = A->rmap->n, n = A->cmap->n;
544 
545   PetscFunctionBegin;
546   PetscCall(MatCreate(PetscObjectComm((PetscObject)A), &B));
547   PetscCall(MatSetSizes(B, A->rmap->n, A->cmap->n, PETSC_DETERMINE, PETSC_DETERMINE));
548   PetscCall(PetscStrallocpy("superlu", &((PetscObject)B)->type_name));
549   PetscCall(MatSetUp(B));
550   B->trivialsymbolic = PETSC_TRUE;
551   PetscCheck(ftype == MAT_FACTOR_LU || ftype == MAT_FACTOR_ILU, PETSC_COMM_SELF, PETSC_ERR_SUP, "Factor type not supported");
552   B->ops->lufactorsymbolic  = MatLUFactorSymbolic_SuperLU;
553   B->ops->ilufactorsymbolic = MatLUFactorSymbolic_SuperLU;
554 
555   PetscCall(PetscFree(B->solvertype));
556   PetscCall(PetscStrallocpy(MATSOLVERSUPERLU, &B->solvertype));
557 
558   B->ops->getinfo = MatGetInfo_External;
559   B->ops->destroy = MatDestroy_SuperLU;
560   B->ops->view    = MatView_SuperLU;
561   B->factortype   = ftype;
562   B->assembled    = PETSC_TRUE; /* required by -ksp_view */
563   B->preallocated = PETSC_TRUE;
564 
565   PetscCall(PetscNew(&lu));
566 
567   if (ftype == MAT_FACTOR_LU) {
568     set_default_options(&lu->options);
569     /* Comments from SuperLU_4.0/SRC/dgssvx.c:
570       "Whether or not the system will be equilibrated depends on the
571        scaling of the matrix A, but if equilibration is used, A is
572        overwritten by diag(R)*A*diag(C) and B by diag(R)*B
573        (if options->Trans=NOTRANS) or diag(C)*B (if options->Trans = TRANS or CONJ)."
574      We set 'options.Equil = NO' as default because additional space is needed for it.
575     */
576     lu->options.Equil = NO;
577   } else if (ftype == MAT_FACTOR_ILU) {
578     /* Set the default input options of ilu: */
579     PetscStackCallExternalVoid("SuperLU:ilu_set_default_options", ilu_set_default_options(&lu->options));
580   }
581   lu->options.PrintStat = NO;
582 
583   /* Initialize the statistics variables. */
584   PetscStackCallExternalVoid("SuperLU:StatInit", StatInit(&lu->stat));
585   lu->lwork = 0; /* allocate space internally by system malloc */
586 
587   /* Allocate spaces (notice sizes are for the transpose) */
588   PetscCall(PetscMalloc1(m, &lu->etree));
589   PetscCall(PetscMalloc1(n, &lu->perm_r));
590   PetscCall(PetscMalloc1(m, &lu->perm_c));
591   PetscCall(PetscMalloc1(n, &lu->R));
592   PetscCall(PetscMalloc1(m, &lu->C));
593 
594   /* create rhs and solution x without allocate space for .Store */
595 #if defined(PETSC_USE_COMPLEX)
596   #if defined(PETSC_USE_REAL_SINGLE)
597   PetscStackCallExternalVoid("SuperLU:cCreate_Dense_Matrix(", cCreate_Dense_Matrix(&lu->B, m, 1, NULL, m, SLU_DN, SLU_C, SLU_GE));
598   PetscStackCallExternalVoid("SuperLU:cCreate_Dense_Matrix(", cCreate_Dense_Matrix(&lu->X, m, 1, NULL, m, SLU_DN, SLU_C, SLU_GE));
599   #else
600   PetscStackCallExternalVoid("SuperLU:zCreate_Dense_Matrix", zCreate_Dense_Matrix(&lu->B, m, 1, NULL, m, SLU_DN, SLU_Z, SLU_GE));
601   PetscStackCallExternalVoid("SuperLU:zCreate_Dense_Matrix", zCreate_Dense_Matrix(&lu->X, m, 1, NULL, m, SLU_DN, SLU_Z, SLU_GE));
602   #endif
603 #else
604   #if defined(PETSC_USE_REAL_SINGLE)
605   PetscStackCallExternalVoid("SuperLU:sCreate_Dense_Matrix", sCreate_Dense_Matrix(&lu->B, m, 1, NULL, m, SLU_DN, SLU_S, SLU_GE));
606   PetscStackCallExternalVoid("SuperLU:sCreate_Dense_Matrix", sCreate_Dense_Matrix(&lu->X, m, 1, NULL, m, SLU_DN, SLU_S, SLU_GE));
607   #else
608   PetscStackCallExternalVoid("SuperLU:dCreate_Dense_Matrix", dCreate_Dense_Matrix(&lu->B, m, 1, NULL, m, SLU_DN, SLU_D, SLU_GE));
609   PetscStackCallExternalVoid("SuperLU:dCreate_Dense_Matrix", dCreate_Dense_Matrix(&lu->X, m, 1, NULL, m, SLU_DN, SLU_D, SLU_GE));
610   #endif
611 #endif
612 
613   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatFactorGetSolverType_C", MatFactorGetSolverType_seqaij_superlu));
614   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatSuperluSetILUDropTol_C", MatSuperluSetILUDropTol_SuperLU));
615   B->data = lu;
616 
617   *F = B;
618   PetscFunctionReturn(PETSC_SUCCESS);
619 }
620 
MatGetFactor_seqsell_superlu(Mat A,MatFactorType ftype,Mat * F)621 static PetscErrorCode MatGetFactor_seqsell_superlu(Mat A, MatFactorType ftype, Mat *F)
622 {
623   Mat_SuperLU *lu;
624 
625   PetscFunctionBegin;
626   PetscCall(MatGetFactor_seqaij_superlu(A, ftype, F));
627   lu                 = (Mat_SuperLU *)((*F)->data);
628   lu->needconversion = PETSC_TRUE;
629   PetscFunctionReturn(PETSC_SUCCESS);
630 }
631 
MatSolverTypeRegister_SuperLU(void)632 PETSC_INTERN PetscErrorCode MatSolverTypeRegister_SuperLU(void)
633 {
634   PetscFunctionBegin;
635   PetscCall(MatSolverTypeRegister(MATSOLVERSUPERLU, MATSEQAIJ, MAT_FACTOR_LU, MatGetFactor_seqaij_superlu));
636   PetscCall(MatSolverTypeRegister(MATSOLVERSUPERLU, MATSEQAIJ, MAT_FACTOR_ILU, MatGetFactor_seqaij_superlu));
637   PetscCall(MatSolverTypeRegister(MATSOLVERSUPERLU, MATSEQSELL, MAT_FACTOR_LU, MatGetFactor_seqsell_superlu));
638   PetscCall(MatSolverTypeRegister(MATSOLVERSUPERLU, MATSEQSELL, MAT_FACTOR_ILU, MatGetFactor_seqsell_superlu));
639   PetscFunctionReturn(PETSC_SUCCESS);
640 }
641