xref: /petsc/src/mat/impls/aij/seq/klu/klu.c (revision 4fc747eaadbeca11629f314a99edccbc2ed7b3d3)
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 #undef __FUNCT__
101 #define __FUNCT__ "MatDestroy_KLU"
102 static PetscErrorCode MatDestroy_KLU(Mat A)
103 {
104   PetscErrorCode ierr;
105   Mat_KLU    *lu=(Mat_KLU*)A->data;
106 
107   PetscFunctionBegin;
108   if (lu->CleanUpKLU) {
109     klu_K_free_symbolic(&lu->Symbolic,&lu->Common);
110     klu_K_free_numeric(&lu->Numeric,&lu->Common);
111     ierr = PetscFree2(lu->perm_r,lu->perm_c);CHKERRQ(ierr);
112   }
113   ierr = PetscFree(A->data);CHKERRQ(ierr);
114   PetscFunctionReturn(0);
115 }
116 
117 #undef __FUNCT__
118 #define __FUNCT__ "MatSolveTranspose_KLU"
119 static PetscErrorCode MatSolveTranspose_KLU(Mat A,Vec b,Vec x)
120 {
121   Mat_KLU       *lu = (Mat_KLU*)A->data;
122   PetscScalar    *xa;
123   PetscErrorCode ierr;
124   PetscInt       status;
125 
126   PetscFunctionBegin;
127   /* KLU uses a column major format, solve Ax = b by klu_*_solve */
128   /* ----------------------------------*/
129   ierr = VecCopy(b,x); /* klu_solve stores the solution in rhs */
130   ierr = VecGetArray(x,&xa);
131   status = klu_K_solve(lu->Symbolic,lu->Numeric,A->rmap->n,1,(PetscReal*)xa,&lu->Common);
132   if (status != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Solve failed");
133   ierr = VecRestoreArray(x,&xa);CHKERRQ(ierr);
134   PetscFunctionReturn(0);
135 }
136 
137 #undef __FUNCT__
138 #define __FUNCT__ "MatSolve_KLU"
139 static PetscErrorCode MatSolve_KLU(Mat A,Vec b,Vec x)
140 {
141   Mat_KLU       *lu = (Mat_KLU*)A->data;
142   PetscScalar    *xa;
143   PetscErrorCode ierr;
144   PetscInt       status;
145 
146   PetscFunctionBegin;
147   /* KLU uses a column major format, solve A^Tx = b by klu_*_tsolve */
148   /* ----------------------------------*/
149   ierr = VecCopy(b,x); /* klu_solve stores the solution in rhs */
150   ierr = VecGetArray(x,&xa);
151 #if defined(PETSC_USE_COMPLEX)
152   PetscInt conj_solve=1;
153   status = klu_K_tsolve(lu->Symbolic,lu->Numeric,A->rmap->n,1,(PetscReal*)xa,conj_solve,&lu->Common); /* conjugate solve */
154 #else
155   status = klu_K_tsolve(lu->Symbolic,lu->Numeric,A->rmap->n,1,xa,&lu->Common);
156 #endif
157   if (status != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Solve failed");
158   ierr = VecRestoreArray(x,&xa);CHKERRQ(ierr);
159   PetscFunctionReturn(0);
160 }
161 
162 #undef __FUNCT__
163 #define __FUNCT__ "MatLUFactorNumeric_KLU"
164 static PetscErrorCode MatLUFactorNumeric_KLU(Mat F,Mat A,const MatFactorInfo *info)
165 {
166   Mat_KLU        *lu = (Mat_KLU*)(F)->data;
167   Mat_SeqAIJ     *a  = (Mat_SeqAIJ*)A->data;
168   PetscInt       *ai = a->i,*aj=a->j;
169   PetscScalar    *av = a->a;
170 
171   PetscFunctionBegin;
172   /* numeric factorization of A' */
173   /* ----------------------------*/
174 
175   if (lu->flg == SAME_NONZERO_PATTERN && lu->Numeric) {
176     klu_K_free_numeric(&lu->Numeric,&lu->Common);
177   }
178   lu->Numeric = klu_K_factor(ai,aj,(PetscReal*)av,lu->Symbolic,&lu->Common);
179   if(!lu->Numeric) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Numeric factorization failed");
180 
181   lu->flg                = SAME_NONZERO_PATTERN;
182   lu->CleanUpKLU         = PETSC_TRUE;
183   F->ops->solve          = MatSolve_KLU;
184   F->ops->solvetranspose = MatSolveTranspose_KLU;
185   PetscFunctionReturn(0);
186 }
187 
188 #undef __FUNCT__
189 #define __FUNCT__ "MatLUFactorSymbolic_KLU"
190 static PetscErrorCode MatLUFactorSymbolic_KLU(Mat F,Mat A,IS r,IS c,const MatFactorInfo *info)
191 {
192   Mat_SeqAIJ     *a  = (Mat_SeqAIJ*)A->data;
193   Mat_KLU       *lu = (Mat_KLU*)(F->data);
194   PetscErrorCode ierr;
195   PetscInt       i,*ai = a->i,*aj = a->j,m=A->rmap->n,n=A->cmap->n;
196   const PetscInt *ra,*ca;
197 
198   PetscFunctionBegin;
199   if (lu->PetscMatOrdering) {
200     ierr = ISGetIndices(r,&ra);CHKERRQ(ierr);
201     ierr = ISGetIndices(c,&ca);CHKERRQ(ierr);
202     ierr = PetscMalloc2(m,&lu->perm_r,n,&lu->perm_c);CHKERRQ(ierr);
203     /* we cannot simply memcpy on 64 bit archs */
204     for (i = 0; i < m; i++) lu->perm_r[i] = ra[i];
205     for (i = 0; i < n; i++) lu->perm_c[i] = ca[i];
206     ierr = ISRestoreIndices(r,&ra);CHKERRQ(ierr);
207     ierr = ISRestoreIndices(c,&ca);CHKERRQ(ierr);
208   }
209 
210   /* symbolic factorization of A' */
211   /* ---------------------------------------------------------------------- */
212   if (lu->PetscMatOrdering) { /* use Petsc ordering */
213     lu->Symbolic = klu_K_analyze_given(n,ai,aj,lu->perm_c,lu->perm_r,&lu->Common);
214   } else { /* use klu internal ordering */
215     lu->Symbolic = klu_K_analyze(n,ai,aj,&lu->Common);
216   }
217   if (!lu->Symbolic) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Symbolic Factorization failed");
218 
219   lu->flg                   = DIFFERENT_NONZERO_PATTERN;
220   lu->CleanUpKLU            = PETSC_TRUE;
221   (F)->ops->lufactornumeric = MatLUFactorNumeric_KLU;
222   PetscFunctionReturn(0);
223 }
224 
225 #undef __FUNCT__
226 #define __FUNCT__ "MatFactorInfo_KLU"
227 static PetscErrorCode MatFactorInfo_KLU(Mat A,PetscViewer viewer)
228 {
229   Mat_KLU       *lu= (Mat_KLU*)A->data;
230   klu_K_numeric *Numeric=(klu_K_numeric*)lu->Numeric;
231   PetscErrorCode ierr;
232 
233   PetscFunctionBegin;
234   /* check if matrix is KLU type */
235   if (A->ops->solve != MatSolve_KLU) PetscFunctionReturn(0);
236 
237   ierr = PetscViewerASCIIPrintf(viewer,"KLU stats:\n");CHKERRQ(ierr);
238   ierr = PetscViewerASCIIPrintf(viewer,"  Number of diagonal blocks: %d\n",Numeric->nblocks);
239   ierr = PetscViewerASCIIPrintf(viewer,"  Total nonzeros=%d\n",Numeric->lnz+Numeric->unz);CHKERRQ(ierr);
240 
241   ierr = PetscViewerASCIIPrintf(viewer,"KLU runtime parameters:\n");CHKERRQ(ierr);
242 
243   /* Control parameters used by numeric factorization */
244   ierr = PetscViewerASCIIPrintf(viewer,"  Partial pivoting tolerance: %g\n",lu->Common.tol);CHKERRQ(ierr);
245   /* BTF preordering */
246   ierr = PetscViewerASCIIPrintf(viewer,"  BTF preordering enabled: %d\n",lu->Common.btf);CHKERRQ(ierr);
247   /* mat ordering */
248   if (!lu->PetscMatOrdering) {
249     ierr = PetscViewerASCIIPrintf(viewer,"  Ordering: %s (not using the PETSc ordering)\n",KluOrderingTypes[(int)lu->Common.ordering]);CHKERRQ(ierr);
250   } else {
251     ierr = PetscViewerASCIIPrintf(viewer,"  Using PETSc ordering\n");CHKERRQ(ierr);
252   }
253   /* matrix row scaling */
254   ierr = PetscViewerASCIIPrintf(viewer, "  Matrix row scaling: %s\n",scale[(int)lu->Common.scale]);CHKERRQ(ierr);
255   PetscFunctionReturn(0);
256 }
257 
258 #undef __FUNCT__
259 #define __FUNCT__ "MatView_KLU"
260 static PetscErrorCode MatView_KLU(Mat A,PetscViewer viewer)
261 {
262   PetscErrorCode    ierr;
263   PetscBool         iascii;
264   PetscViewerFormat format;
265 
266   PetscFunctionBegin;
267   ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);CHKERRQ(ierr);
268   if (iascii) {
269     ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr);
270     if (format == PETSC_VIEWER_ASCII_INFO) {
271       ierr = MatFactorInfo_KLU(A,viewer);CHKERRQ(ierr);
272     }
273   }
274   PetscFunctionReturn(0);
275 }
276 
277 #undef __FUNCT__
278 #define __FUNCT__ "MatFactorGetSolverPackage_seqaij_klu"
279 PetscErrorCode MatFactorGetSolverPackage_seqaij_klu(Mat A,const MatSolverPackage *type)
280 {
281   PetscFunctionBegin;
282   *type = MATSOLVERKLU;
283   PetscFunctionReturn(0);
284 }
285 
286 
287 /*MC
288   MATSOLVERKLU = "klu" - A matrix type providing direct solvers (LU) for sequential matrices
289   via the external package KLU.
290 
291   ./configure --download-suitesparse to install PETSc to use KLU
292 
293   Use -pc_type lu -pc_factor_mat_solver_package klu to us this direct solver
294 
295   Consult KLU documentation for more information on the options database keys below.
296 
297   Options Database Keys:
298 + -mat_klu_pivot_tol <0.001>                  - Partial pivoting tolerance
299 . -mat_klu_use_btf <1>                        - Use BTF preordering
300 . -mat_klu_ordering <AMD>                     - KLU reordering scheme to reduce fill-in (choose one of) AMD COLAMD PETSC
301 - -mat_klu_row_scale <NONE>                   - Matrix row scaling (choose one of) NONE SUM MAX
302 
303    Note: KLU is part of SuiteSparse http://faculty.cse.tamu.edu/davis/suitesparse.html
304 
305    Level: beginner
306 
307 .seealso: PCLU, MATSOLVERUMFPACK, MATSOLVERCHOLMOD, PCFactorSetMatSolverPackage(), MatSolverPackage
308 M*/
309 
310 #undef __FUNCT__
311 #define __FUNCT__ "MatGetFactor_seqaij_klu"
312 PETSC_INTERN PetscErrorCode MatGetFactor_seqaij_klu(Mat A,MatFactorType ftype,Mat *F)
313 {
314   Mat            B;
315   Mat_KLU       *lu;
316   PetscErrorCode ierr;
317   PetscInt       m=A->rmap->n,n=A->cmap->n,idx,status;
318   PetscBool      flg;
319 
320   PetscFunctionBegin;
321   /* Create the factorization matrix F */
322   ierr = MatCreate(PetscObjectComm((PetscObject)A),&B);CHKERRQ(ierr);
323   ierr = MatSetSizes(B,PETSC_DECIDE,PETSC_DECIDE,m,n);CHKERRQ(ierr);
324   ierr = PetscStrallocpy("klu",&((PetscObject)B)->type_name);CHKERRQ(ierr);
325   ierr = MatSetUp(B);CHKERRQ(ierr);
326 
327   ierr = PetscNewLog(B,&lu);CHKERRQ(ierr);
328 
329   B->data                  = lu;
330   B->ops->getinfo          = MatGetInfo_External;
331   B->ops->lufactorsymbolic = MatLUFactorSymbolic_KLU;
332   B->ops->destroy          = MatDestroy_KLU;
333   B->ops->view             = MatView_KLU;
334 
335   ierr = PetscObjectComposeFunction((PetscObject)B,"MatFactorGetSolverPackage_C",MatFactorGetSolverPackage_seqaij_klu);CHKERRQ(ierr);
336 
337   B->factortype   = MAT_FACTOR_LU;
338   B->assembled    = PETSC_TRUE;           /* required by -ksp_view */
339   B->preallocated = PETSC_TRUE;
340 
341   ierr = PetscFree(B->solvertype);CHKERRQ(ierr);
342   ierr = PetscStrallocpy(MATSOLVERKLU,&B->solvertype);CHKERRQ(ierr);
343 
344   /* initializations */
345   /* ------------------------------------------------*/
346   /* get the default control parameters */
347   status = klu_K_defaults(&lu->Common);
348   if(status <= 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Initialization failed");
349 
350   lu->Common.scale = 0; /* No row scaling */
351 
352   ierr = PetscOptionsBegin(PetscObjectComm((PetscObject)A),((PetscObject)A)->prefix,"KLU Options","Mat");CHKERRQ(ierr);
353   /* Partial pivoting tolerance */
354   ierr = PetscOptionsReal("-mat_klu_pivot_tol","Partial pivoting tolerance","None",lu->Common.tol,&lu->Common.tol,NULL);CHKERRQ(ierr);
355   /* BTF pre-ordering */
356   ierr = PetscOptionsInt("-mat_klu_use_btf","Enable BTF preordering","None",lu->Common.btf,&lu->Common.btf,NULL);CHKERRQ(ierr);
357   /* Matrix reordering */
358   ierr = PetscOptionsEList("-mat_klu_ordering","Internal ordering method","None",KluOrderingTypes,sizeof(KluOrderingTypes)/sizeof(KluOrderingTypes[0]),KluOrderingTypes[0],&idx,&flg);CHKERRQ(ierr);
359   if (flg) {
360     if ((int)idx == 2) lu->PetscMatOrdering = PETSC_TRUE;   /* use Petsc mat ordering (note: size is for the transpose, and PETSc r = Klu perm_c) */
361     else lu->Common.ordering = (int)idx;
362   }
363   /* Matrix row scaling */
364   ierr = PetscOptionsEList("-mat_klu_row_scale","Matrix row scaling","None",scale,3,scale[0],&idx,&flg);CHKERRQ(ierr);
365   PetscOptionsEnd();
366   *F = B;
367   PetscFunctionReturn(0);
368 }
369