xref: /petsc/src/mat/impls/aij/seq/klu/klu.c (revision 609bdbee21ea3be08735c64dbe00a9ab27759925)
1 
2 /*
3    Provides an interface to the KLUv1.2 sparse solver
4 
5    When build with PETSC_USE_64BIT_INDICES this will use SuiteSparse_long as the
6    integer type in KLU, otherwise it will use int. This means
7    all integers in this file are simply declared as PetscInt. Also it means
8    that KLU SuiteSparse_long version MUST be built with 64 bit integers when used.
9 
10 */
11 #include <../src/mat/impls/aij/seq/aij.h>
12 
13 #if defined(PETSC_USE_64BIT_INDICES)
14 #define klu_K_defaults                klu_l_defaults
15 #define klu_K_analyze                 klu_l_analyze
16 #define klu_K_analyze_given           klu_l_analyze_given
17 #define klu_K_free_symbolic           klu_l_free_symbolic
18 #define klu_K_free_numeric            klu_l_free_numeric
19 #define klu_K_common                  klu_l_common
20 #define klu_K_symbolic                klu_l_symbolic
21 #define klu_K_numeric                 klu_l_numeric
22 #if defined(PETSC_USE_COMPLEX)
23 #define klu_K_factor                  klu_zl_factor
24 #define klu_K_solve                   klu_zl_solve
25 #define klu_K_tsolve                  klu_zl_tsolve
26 #define klu_K_refactor                klu_zl_refactor
27 #define klu_K_sort                    klu_zl_sort
28 #define klu_K_flops                   klu_zl_flops
29 #define klu_K_rgrowth                 klu_zl_rgrowth
30 #define klu_K_condest                 klu_zl_condest
31 #define klu_K_rcond                   klu_zl_rcond
32 #define klu_K_scale                   klu_zl_scale
33 #else
34 #define klu_K_factor                  klu_l_factor
35 #define klu_K_solve                   klu_l_solve
36 #define klu_K_tsolve                  klu_l_tsolve
37 #define klu_K_refactor                klu_l_refactor
38 #define klu_K_sort                    klu_l_sort
39 #define klu_K_flops                   klu_l_flops
40 #define klu_K_rgrowth                 klu_l_rgrowth
41 #define klu_K_condest                 klu_l_condest
42 #define klu_K_rcond                   klu_l_rcond
43 #define klu_K_scale                   klu_l_scale
44 #endif
45 #else
46 #define klu_K_defaults                klu_defaults
47 #define klu_K_analyze                 klu_analyze
48 #define klu_K_analyze_given           klu_analyze_given
49 #define klu_K_free_symbolic           klu_free_symbolic
50 #define klu_K_free_numeric            klu_free_numeric
51 #define klu_K_common                  klu_common
52 #define klu_K_symbolic                klu_symbolic
53 #define klu_K_numeric                 klu_numeric
54 #if defined(PETSC_USE_COMPLEX)
55 #define klu_K_factor                  klu_z_factor
56 #define klu_K_solve                   klu_z_solve
57 #define klu_K_tsolve                  klu_z_tsolve
58 #define klu_K_refactor                klu_z_refactor
59 #define klu_K_sort                    klu_z_sort
60 #define klu_K_flops                   klu_z_flops
61 #define klu_K_rgrowth                 klu_z_rgrowth
62 #define klu_K_condest                 klu_z_condest
63 #define klu_K_rcond                   klu_z_rcond
64 #define klu_K_scale                   klu_z_scale
65 #else
66 #define klu_K_factor                  klu_factor
67 #define klu_K_solve                   klu_solve
68 #define klu_K_tsolve                  klu_tsolve
69 #define klu_K_refactor                klu_refactor
70 #define klu_K_sort                    klu_sort
71 #define klu_K_flops                   klu_flops
72 #define klu_K_rgrowth                 klu_rgrowth
73 #define klu_K_condest                 klu_condest
74 #define klu_K_rcond                   klu_rcond
75 #define klu_K_scale                   klu_scale
76 #endif
77 #endif
78 
79 #define SuiteSparse_long long long
80 #define SuiteSparse_long_max LONG_LONG_MAX
81 #define SuiteSparse_long_id "%lld"
82 
83 EXTERN_C_BEGIN
84 #include <klu.h>
85 EXTERN_C_END
86 
87 static const char *KluOrderingTypes[] = {"AMD","COLAMD","PETSC"};
88 static const char *scale[] ={"NONE","SUM","MAX"};
89 
90 typedef struct {
91   klu_K_common   Common;
92   klu_K_symbolic *Symbolic;
93   klu_K_numeric  *Numeric;
94   PetscInt       *perm_c,*perm_r;
95   MatStructure   flg;
96   PetscBool      PetscMatOrdering;
97   PetscBool      CleanUpKLU;
98 } Mat_KLU;
99 
100 static PetscErrorCode MatDestroy_KLU(Mat A)
101 {
102   PetscErrorCode ierr;
103   Mat_KLU    *lu=(Mat_KLU*)A->data;
104 
105   PetscFunctionBegin;
106   if (lu->CleanUpKLU) {
107     klu_K_free_symbolic(&lu->Symbolic,&lu->Common);
108     klu_K_free_numeric(&lu->Numeric,&lu->Common);
109     ierr = PetscFree2(lu->perm_r,lu->perm_c);CHKERRQ(ierr);
110   }
111   ierr = PetscFree(A->data);CHKERRQ(ierr);
112   PetscFunctionReturn(0);
113 }
114 
115 static PetscErrorCode MatSolveTranspose_KLU(Mat A,Vec b,Vec x)
116 {
117   Mat_KLU       *lu = (Mat_KLU*)A->data;
118   PetscScalar    *xa;
119   PetscErrorCode ierr;
120   PetscInt       status;
121 
122   PetscFunctionBegin;
123   /* KLU uses a column major format, solve Ax = b by klu_*_solve */
124   /* ----------------------------------*/
125   ierr = VecCopy(b,x); /* klu_solve stores the solution in rhs */
126   ierr = VecGetArray(x,&xa);
127   status = klu_K_solve(lu->Symbolic,lu->Numeric,A->rmap->n,1,(PetscReal*)xa,&lu->Common);
128   if (status != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Solve failed");
129   ierr = VecRestoreArray(x,&xa);CHKERRQ(ierr);
130   PetscFunctionReturn(0);
131 }
132 
133 static PetscErrorCode MatSolve_KLU(Mat A,Vec b,Vec x)
134 {
135   Mat_KLU       *lu = (Mat_KLU*)A->data;
136   PetscScalar    *xa;
137   PetscErrorCode ierr;
138   PetscInt       status;
139 
140   PetscFunctionBegin;
141   /* KLU uses a column major format, solve A^Tx = b by klu_*_tsolve */
142   /* ----------------------------------*/
143   ierr = VecCopy(b,x); /* klu_solve stores the solution in rhs */
144   ierr = VecGetArray(x,&xa);
145 #if defined(PETSC_USE_COMPLEX)
146   PetscInt conj_solve=1;
147   status = klu_K_tsolve(lu->Symbolic,lu->Numeric,A->rmap->n,1,(PetscReal*)xa,conj_solve,&lu->Common); /* conjugate solve */
148 #else
149   status = klu_K_tsolve(lu->Symbolic,lu->Numeric,A->rmap->n,1,xa,&lu->Common);
150 #endif
151   if (status != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Solve failed");
152   ierr = VecRestoreArray(x,&xa);CHKERRQ(ierr);
153   PetscFunctionReturn(0);
154 }
155 
156 static PetscErrorCode MatLUFactorNumeric_KLU(Mat F,Mat A,const MatFactorInfo *info)
157 {
158   Mat_KLU        *lu = (Mat_KLU*)(F)->data;
159   Mat_SeqAIJ     *a  = (Mat_SeqAIJ*)A->data;
160   PetscInt       *ai = a->i,*aj=a->j;
161   PetscScalar    *av = a->a;
162 
163   PetscFunctionBegin;
164   /* numeric factorization of A' */
165   /* ----------------------------*/
166 
167   if (lu->flg == SAME_NONZERO_PATTERN && lu->Numeric) {
168     klu_K_free_numeric(&lu->Numeric,&lu->Common);
169   }
170   lu->Numeric = klu_K_factor(ai,aj,(PetscReal*)av,lu->Symbolic,&lu->Common);
171   if(!lu->Numeric) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Numeric factorization failed");
172 
173   lu->flg                = SAME_NONZERO_PATTERN;
174   lu->CleanUpKLU         = PETSC_TRUE;
175   F->ops->solve          = MatSolve_KLU;
176   F->ops->solvetranspose = MatSolveTranspose_KLU;
177   PetscFunctionReturn(0);
178 }
179 
180 static PetscErrorCode MatLUFactorSymbolic_KLU(Mat F,Mat A,IS r,IS c,const MatFactorInfo *info)
181 {
182   Mat_SeqAIJ     *a  = (Mat_SeqAIJ*)A->data;
183   Mat_KLU       *lu = (Mat_KLU*)(F->data);
184   PetscErrorCode ierr;
185   PetscInt       i,*ai = a->i,*aj = a->j,m=A->rmap->n,n=A->cmap->n;
186   const PetscInt *ra,*ca;
187 
188   PetscFunctionBegin;
189   if (lu->PetscMatOrdering) {
190     ierr = ISGetIndices(r,&ra);CHKERRQ(ierr);
191     ierr = ISGetIndices(c,&ca);CHKERRQ(ierr);
192     ierr = PetscMalloc2(m,&lu->perm_r,n,&lu->perm_c);CHKERRQ(ierr);
193     /* we cannot simply memcpy on 64 bit archs */
194     for (i = 0; i < m; i++) lu->perm_r[i] = ra[i];
195     for (i = 0; i < n; i++) lu->perm_c[i] = ca[i];
196     ierr = ISRestoreIndices(r,&ra);CHKERRQ(ierr);
197     ierr = ISRestoreIndices(c,&ca);CHKERRQ(ierr);
198   }
199 
200   /* symbolic factorization of A' */
201   /* ---------------------------------------------------------------------- */
202   if (lu->PetscMatOrdering) { /* use Petsc ordering */
203     lu->Symbolic = klu_K_analyze_given(n,ai,aj,lu->perm_c,lu->perm_r,&lu->Common);
204   } else { /* use klu internal ordering */
205     lu->Symbolic = klu_K_analyze(n,ai,aj,&lu->Common);
206   }
207   if (!lu->Symbolic) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Symbolic Factorization failed");
208 
209   lu->flg                   = DIFFERENT_NONZERO_PATTERN;
210   lu->CleanUpKLU            = PETSC_TRUE;
211   (F)->ops->lufactornumeric = MatLUFactorNumeric_KLU;
212   PetscFunctionReturn(0);
213 }
214 
215 static PetscErrorCode MatFactorInfo_KLU(Mat A,PetscViewer viewer)
216 {
217   Mat_KLU       *lu= (Mat_KLU*)A->data;
218   klu_K_numeric *Numeric=(klu_K_numeric*)lu->Numeric;
219   PetscErrorCode ierr;
220 
221   PetscFunctionBegin;
222   /* check if matrix is KLU type */
223   if (A->ops->solve != MatSolve_KLU) PetscFunctionReturn(0);
224 
225   ierr = PetscViewerASCIIPrintf(viewer,"KLU stats:\n");CHKERRQ(ierr);
226   ierr = PetscViewerASCIIPrintf(viewer,"  Number of diagonal blocks: %d\n",Numeric->nblocks);
227   ierr = PetscViewerASCIIPrintf(viewer,"  Total nonzeros=%d\n",Numeric->lnz+Numeric->unz);CHKERRQ(ierr);
228 
229   ierr = PetscViewerASCIIPrintf(viewer,"KLU runtime parameters:\n");CHKERRQ(ierr);
230 
231   /* Control parameters used by numeric factorization */
232   ierr = PetscViewerASCIIPrintf(viewer,"  Partial pivoting tolerance: %g\n",lu->Common.tol);CHKERRQ(ierr);
233   /* BTF preordering */
234   ierr = PetscViewerASCIIPrintf(viewer,"  BTF preordering enabled: %d\n",lu->Common.btf);CHKERRQ(ierr);
235   /* mat ordering */
236   if (!lu->PetscMatOrdering) {
237     ierr = PetscViewerASCIIPrintf(viewer,"  Ordering: %s (not using the PETSc ordering)\n",KluOrderingTypes[(int)lu->Common.ordering]);CHKERRQ(ierr);
238   } else {
239     ierr = PetscViewerASCIIPrintf(viewer,"  Using PETSc ordering\n");CHKERRQ(ierr);
240   }
241   /* matrix row scaling */
242   ierr = PetscViewerASCIIPrintf(viewer, "  Matrix row scaling: %s\n",scale[(int)lu->Common.scale]);CHKERRQ(ierr);
243   PetscFunctionReturn(0);
244 }
245 
246 static PetscErrorCode MatView_KLU(Mat A,PetscViewer viewer)
247 {
248   PetscErrorCode    ierr;
249   PetscBool         iascii;
250   PetscViewerFormat format;
251 
252   PetscFunctionBegin;
253   ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);CHKERRQ(ierr);
254   if (iascii) {
255     ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr);
256     if (format == PETSC_VIEWER_ASCII_INFO) {
257       ierr = MatFactorInfo_KLU(A,viewer);CHKERRQ(ierr);
258     }
259   }
260   PetscFunctionReturn(0);
261 }
262 
263 PetscErrorCode MatFactorGetSolverPackage_seqaij_klu(Mat A,const MatSolverPackage *type)
264 {
265   PetscFunctionBegin;
266   *type = MATSOLVERKLU;
267   PetscFunctionReturn(0);
268 }
269 
270 
271 /*MC
272   MATSOLVERKLU = "klu" - A matrix type providing direct solvers (LU) for sequential matrices
273   via the external package KLU.
274 
275   ./configure --download-suitesparse to install PETSc to use KLU
276 
277   Use -pc_type lu -pc_factor_mat_solver_package klu to us this direct solver
278 
279   Consult KLU documentation for more information on the options database keys below.
280 
281   Options Database Keys:
282 + -mat_klu_pivot_tol <0.001>                  - Partial pivoting tolerance
283 . -mat_klu_use_btf <1>                        - Use BTF preordering
284 . -mat_klu_ordering <AMD>                     - KLU reordering scheme to reduce fill-in (choose one of) AMD COLAMD PETSC
285 - -mat_klu_row_scale <NONE>                   - Matrix row scaling (choose one of) NONE SUM MAX
286 
287    Note: KLU is part of SuiteSparse http://faculty.cse.tamu.edu/davis/suitesparse.html
288 
289    Level: beginner
290 
291 .seealso: PCLU, MATSOLVERUMFPACK, MATSOLVERCHOLMOD, PCFactorSetMatSolverPackage(), MatSolverPackage
292 M*/
293 
294 PETSC_INTERN PetscErrorCode MatGetFactor_seqaij_klu(Mat A,MatFactorType ftype,Mat *F)
295 {
296   Mat            B;
297   Mat_KLU       *lu;
298   PetscErrorCode ierr;
299   PetscInt       m=A->rmap->n,n=A->cmap->n,idx,status;
300   PetscBool      flg;
301 
302   PetscFunctionBegin;
303   /* Create the factorization matrix F */
304   ierr = MatCreate(PetscObjectComm((PetscObject)A),&B);CHKERRQ(ierr);
305   ierr = MatSetSizes(B,PETSC_DECIDE,PETSC_DECIDE,m,n);CHKERRQ(ierr);
306   ierr = PetscStrallocpy("klu",&((PetscObject)B)->type_name);CHKERRQ(ierr);
307   ierr = MatSetUp(B);CHKERRQ(ierr);
308 
309   ierr = PetscNewLog(B,&lu);CHKERRQ(ierr);
310 
311   B->data                  = lu;
312   B->ops->getinfo          = MatGetInfo_External;
313   B->ops->lufactorsymbolic = MatLUFactorSymbolic_KLU;
314   B->ops->destroy          = MatDestroy_KLU;
315   B->ops->view             = MatView_KLU;
316 
317   ierr = PetscObjectComposeFunction((PetscObject)B,"MatFactorGetSolverPackage_C",MatFactorGetSolverPackage_seqaij_klu);CHKERRQ(ierr);
318 
319   B->factortype   = MAT_FACTOR_LU;
320   B->assembled    = PETSC_TRUE;           /* required by -ksp_view */
321   B->preallocated = PETSC_TRUE;
322 
323   ierr = PetscFree(B->solvertype);CHKERRQ(ierr);
324   ierr = PetscStrallocpy(MATSOLVERKLU,&B->solvertype);CHKERRQ(ierr);
325 
326   /* initializations */
327   /* ------------------------------------------------*/
328   /* get the default control parameters */
329   status = klu_K_defaults(&lu->Common);
330   if(status <= 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Initialization failed");
331 
332   lu->Common.scale = 0; /* No row scaling */
333 
334   ierr = PetscOptionsBegin(PetscObjectComm((PetscObject)A),((PetscObject)A)->prefix,"KLU Options","Mat");CHKERRQ(ierr);
335   /* Partial pivoting tolerance */
336   ierr = PetscOptionsReal("-mat_klu_pivot_tol","Partial pivoting tolerance","None",lu->Common.tol,&lu->Common.tol,NULL);CHKERRQ(ierr);
337   /* BTF pre-ordering */
338   ierr = PetscOptionsInt("-mat_klu_use_btf","Enable BTF preordering","None",lu->Common.btf,&lu->Common.btf,NULL);CHKERRQ(ierr);
339   /* Matrix reordering */
340   ierr = PetscOptionsEList("-mat_klu_ordering","Internal ordering method","None",KluOrderingTypes,sizeof(KluOrderingTypes)/sizeof(KluOrderingTypes[0]),KluOrderingTypes[0],&idx,&flg);CHKERRQ(ierr);
341   if (flg) {
342     if ((int)idx == 2) lu->PetscMatOrdering = PETSC_TRUE;   /* use Petsc mat ordering (note: size is for the transpose, and PETSc r = Klu perm_c) */
343     else lu->Common.ordering = (int)idx;
344   }
345   /* Matrix row scaling */
346   ierr = PetscOptionsEList("-mat_klu_row_scale","Matrix row scaling","None",scale,3,scale[0],&idx,&flg);CHKERRQ(ierr);
347   PetscOptionsEnd();
348   *F = B;
349   PetscFunctionReturn(0);
350 }
351