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(a,b,c,d) klu_l_analyze((SuiteSparse_long)a,(SuiteSparse_long*)b,(SuiteSparse_long*)c,d) 16 #define klu_K_analyze_given(a,b,c,d,e,f) klu_l_analyze_given((SuiteSparse_long)a,(SuiteSparse_long*)b,(SuiteSparse_long*)c,(SuiteSparse_long*)d,(SuiteSparse_long*)e,f) 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(a,b,c,d,e) klu_zl_factor((SuiteSparse_long*)a,(SuiteSparse_long*)b,c,d,e); 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(a,b,c,d,e) klu_l_factor((SuiteSparse_long*)a,(SuiteSparse_long*)b,c,d,e); 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 EXTERN_C_BEGIN 80 #include <klu.h> 81 EXTERN_C_END 82 83 static const char *KluOrderingTypes[] = {"AMD","COLAMD"}; 84 static const char *scale[] ={"NONE","SUM","MAX"}; 85 86 typedef struct { 87 klu_K_common Common; 88 klu_K_symbolic *Symbolic; 89 klu_K_numeric *Numeric; 90 PetscInt *perm_c,*perm_r; 91 MatStructure flg; 92 PetscBool PetscMatOrdering; 93 PetscBool CleanUpKLU; 94 } Mat_KLU; 95 96 static PetscErrorCode MatDestroy_KLU(Mat A) 97 { 98 PetscErrorCode ierr; 99 Mat_KLU *lu=(Mat_KLU*)A->data; 100 101 PetscFunctionBegin; 102 if (lu->CleanUpKLU) { 103 klu_K_free_symbolic(&lu->Symbolic,&lu->Common); 104 klu_K_free_numeric(&lu->Numeric,&lu->Common); 105 ierr = PetscFree2(lu->perm_r,lu->perm_c);CHKERRQ(ierr); 106 } 107 ierr = PetscFree(A->data);CHKERRQ(ierr); 108 PetscFunctionReturn(0); 109 } 110 111 static PetscErrorCode MatSolveTranspose_KLU(Mat A,Vec b,Vec x) 112 { 113 Mat_KLU *lu = (Mat_KLU*)A->data; 114 PetscScalar *xa; 115 PetscErrorCode ierr; 116 PetscInt status; 117 118 PetscFunctionBegin; 119 /* KLU uses a column major format, solve Ax = b by klu_*_solve */ 120 /* ----------------------------------*/ 121 ierr = VecCopy(b,x); /* klu_solve stores the solution in rhs */ 122 ierr = VecGetArray(x,&xa); 123 status = klu_K_solve(lu->Symbolic,lu->Numeric,A->rmap->n,1,(PetscReal*)xa,&lu->Common); 124 if (status != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Solve failed"); 125 ierr = VecRestoreArray(x,&xa);CHKERRQ(ierr); 126 PetscFunctionReturn(0); 127 } 128 129 static PetscErrorCode MatSolve_KLU(Mat A,Vec b,Vec x) 130 { 131 Mat_KLU *lu = (Mat_KLU*)A->data; 132 PetscScalar *xa; 133 PetscErrorCode ierr; 134 PetscInt status; 135 136 PetscFunctionBegin; 137 /* KLU uses a column major format, solve A^Tx = b by klu_*_tsolve */ 138 /* ----------------------------------*/ 139 ierr = VecCopy(b,x); /* klu_solve stores the solution in rhs */ 140 ierr = VecGetArray(x,&xa); 141 #if defined(PETSC_USE_COMPLEX) 142 PetscInt conj_solve=1; 143 status = klu_K_tsolve(lu->Symbolic,lu->Numeric,A->rmap->n,1,(PetscReal*)xa,conj_solve,&lu->Common); /* conjugate solve */ 144 #else 145 status = klu_K_tsolve(lu->Symbolic,lu->Numeric,A->rmap->n,1,xa,&lu->Common); 146 #endif 147 if (status != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Solve failed"); 148 ierr = VecRestoreArray(x,&xa);CHKERRQ(ierr); 149 PetscFunctionReturn(0); 150 } 151 152 static PetscErrorCode MatLUFactorNumeric_KLU(Mat F,Mat A,const MatFactorInfo *info) 153 { 154 Mat_KLU *lu = (Mat_KLU*)(F)->data; 155 Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data; 156 PetscInt *ai = a->i,*aj=a->j; 157 PetscScalar *av = a->a; 158 159 PetscFunctionBegin; 160 /* numeric factorization of A' */ 161 /* ----------------------------*/ 162 163 if (lu->flg == SAME_NONZERO_PATTERN && lu->Numeric) { 164 klu_K_free_numeric(&lu->Numeric,&lu->Common); 165 } 166 lu->Numeric = klu_K_factor(ai,aj,(PetscReal*)av,lu->Symbolic,&lu->Common); 167 if (!lu->Numeric) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Numeric factorization failed"); 168 169 lu->flg = SAME_NONZERO_PATTERN; 170 lu->CleanUpKLU = PETSC_TRUE; 171 F->ops->solve = MatSolve_KLU; 172 F->ops->solvetranspose = MatSolveTranspose_KLU; 173 PetscFunctionReturn(0); 174 } 175 176 static PetscErrorCode MatLUFactorSymbolic_KLU(Mat F,Mat A,IS r,IS c,const MatFactorInfo *info) 177 { 178 Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data; 179 Mat_KLU *lu = (Mat_KLU*)(F->data); 180 PetscErrorCode ierr; 181 PetscInt i,*ai = a->i,*aj = a->j,m=A->rmap->n,n=A->cmap->n; 182 const PetscInt *ra,*ca; 183 184 PetscFunctionBegin; 185 if (lu->PetscMatOrdering) { 186 ierr = ISGetIndices(r,&ra);CHKERRQ(ierr); 187 ierr = ISGetIndices(c,&ca);CHKERRQ(ierr); 188 ierr = PetscMalloc2(m,&lu->perm_r,n,&lu->perm_c);CHKERRQ(ierr); 189 /* we cannot simply memcpy on 64 bit archs */ 190 for (i = 0; i < m; i++) lu->perm_r[i] = ra[i]; 191 for (i = 0; i < n; i++) lu->perm_c[i] = ca[i]; 192 ierr = ISRestoreIndices(r,&ra);CHKERRQ(ierr); 193 ierr = ISRestoreIndices(c,&ca);CHKERRQ(ierr); 194 } 195 196 /* symbolic factorization of A' */ 197 /* ---------------------------------------------------------------------- */ 198 if (r) { 199 lu->PetscMatOrdering = PETSC_TRUE; 200 lu->Symbolic = klu_K_analyze_given(n,ai,aj,lu->perm_c,lu->perm_r,&lu->Common); 201 } else { /* use klu internal ordering */ 202 lu->Symbolic = klu_K_analyze(n,ai,aj,&lu->Common); 203 } 204 if (!lu->Symbolic) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Symbolic Factorization failed"); 205 206 lu->flg = DIFFERENT_NONZERO_PATTERN; 207 lu->CleanUpKLU = PETSC_TRUE; 208 (F)->ops->lufactornumeric = MatLUFactorNumeric_KLU; 209 PetscFunctionReturn(0); 210 } 211 212 static PetscErrorCode MatView_Info_KLU(Mat A,PetscViewer viewer) 213 { 214 Mat_KLU *lu= (Mat_KLU*)A->data; 215 klu_K_numeric *Numeric=(klu_K_numeric*)lu->Numeric; 216 PetscErrorCode ierr; 217 218 PetscFunctionBegin; 219 ierr = PetscViewerASCIIPrintf(viewer,"KLU stats:\n");CHKERRQ(ierr); 220 ierr = PetscViewerASCIIPrintf(viewer," Number of diagonal blocks: %d\n",Numeric->nblocks); 221 ierr = PetscViewerASCIIPrintf(viewer," Total nonzeros=%d\n",Numeric->lnz+Numeric->unz);CHKERRQ(ierr); 222 ierr = PetscViewerASCIIPrintf(viewer,"KLU runtime parameters:\n");CHKERRQ(ierr); 223 /* Control parameters used by numeric factorization */ 224 ierr = PetscViewerASCIIPrintf(viewer," Partial pivoting tolerance: %g\n",lu->Common.tol);CHKERRQ(ierr); 225 /* BTF preordering */ 226 ierr = PetscViewerASCIIPrintf(viewer," BTF preordering enabled: %d\n",lu->Common.btf);CHKERRQ(ierr); 227 /* mat ordering */ 228 if (!lu->PetscMatOrdering) { 229 ierr = PetscViewerASCIIPrintf(viewer," Ordering: %s (not using the PETSc ordering)\n",KluOrderingTypes[(int)lu->Common.ordering]);CHKERRQ(ierr); 230 } 231 /* matrix row scaling */ 232 ierr = PetscViewerASCIIPrintf(viewer, " Matrix row scaling: %s\n",scale[(int)lu->Common.scale]);CHKERRQ(ierr); 233 PetscFunctionReturn(0); 234 } 235 236 static PetscErrorCode MatView_KLU(Mat A,PetscViewer viewer) 237 { 238 PetscErrorCode ierr; 239 PetscBool iascii; 240 PetscViewerFormat format; 241 242 PetscFunctionBegin; 243 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);CHKERRQ(ierr); 244 if (iascii) { 245 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 246 if (format == PETSC_VIEWER_ASCII_INFO) { 247 ierr = MatView_Info_KLU(A,viewer);CHKERRQ(ierr); 248 } 249 } 250 PetscFunctionReturn(0); 251 } 252 253 PetscErrorCode MatFactorGetSolverType_seqaij_klu(Mat A,MatSolverType *type) 254 { 255 PetscFunctionBegin; 256 *type = MATSOLVERKLU; 257 PetscFunctionReturn(0); 258 } 259 260 /*MC 261 MATSOLVERKLU = "klu" - A matrix type providing direct solvers (LU) for sequential matrices 262 via the external package KLU. 263 264 ./configure --download-suitesparse to install PETSc to use KLU 265 266 Use -pc_type lu -pc_factor_mat_solver_type klu to use this direct solver 267 268 Consult KLU documentation for more information on the options database keys below. 269 270 Options Database Keys: 271 + -mat_klu_pivot_tol <0.001> - Partial pivoting tolerance 272 . -mat_klu_use_btf <1> - Use BTF preordering 273 . -mat_klu_ordering <AMD> - KLU reordering scheme to reduce fill-in (choose one of) AMD COLAMD PETSC 274 - -mat_klu_row_scale <NONE> - Matrix row scaling (choose one of) NONE SUM MAX 275 276 Note: KLU is part of SuiteSparse http://faculty.cse.tamu.edu/davis/suitesparse.html 277 278 Level: beginner 279 280 .seealso: PCLU, MATSOLVERUMFPACK, MATSOLVERCHOLMOD, PCFactorSetMatSolverType(), MatSolverType 281 M*/ 282 283 PETSC_INTERN PetscErrorCode MatGetFactor_seqaij_klu(Mat A,MatFactorType ftype,Mat *F) 284 { 285 Mat B; 286 Mat_KLU *lu; 287 PetscErrorCode ierr; 288 PetscInt m=A->rmap->n,n=A->cmap->n,idx = 0,status; 289 PetscBool flg; 290 291 PetscFunctionBegin; 292 /* Create the factorization matrix F */ 293 ierr = MatCreate(PetscObjectComm((PetscObject)A),&B);CHKERRQ(ierr); 294 ierr = MatSetSizes(B,PETSC_DECIDE,PETSC_DECIDE,m,n);CHKERRQ(ierr); 295 ierr = PetscStrallocpy("klu",&((PetscObject)B)->type_name);CHKERRQ(ierr); 296 ierr = MatSetUp(B);CHKERRQ(ierr); 297 298 ierr = PetscNewLog(B,&lu);CHKERRQ(ierr); 299 300 B->data = lu; 301 B->ops->getinfo = MatGetInfo_External; 302 B->ops->lufactorsymbolic = MatLUFactorSymbolic_KLU; 303 B->ops->destroy = MatDestroy_KLU; 304 B->ops->view = MatView_KLU; 305 306 ierr = PetscObjectComposeFunction((PetscObject)B,"MatFactorGetSolverType_C",MatFactorGetSolverType_seqaij_klu);CHKERRQ(ierr); 307 308 B->factortype = MAT_FACTOR_LU; 309 B->assembled = PETSC_TRUE; /* required by -ksp_view */ 310 B->preallocated = PETSC_TRUE; 311 312 ierr = PetscFree(B->solvertype);CHKERRQ(ierr); 313 ierr = PetscStrallocpy(MATSOLVERKLU,&B->solvertype);CHKERRQ(ierr); 314 B->canuseordering = PETSC_TRUE; 315 ierr = PetscStrallocpy(MATORDERINGEXTERNAL,(char**)&B->preferredordering[MAT_FACTOR_LU]);CHKERRQ(ierr); 316 317 /* initializations */ 318 /* ------------------------------------------------*/ 319 /* get the default control parameters */ 320 status = klu_K_defaults(&lu->Common); 321 if (status <= 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_LIB,"KLU Initialization failed"); 322 323 lu->Common.scale = 0; /* No row scaling */ 324 325 ierr = PetscOptionsBegin(PetscObjectComm((PetscObject)A),((PetscObject)A)->prefix,"KLU Options","Mat");CHKERRQ(ierr); 326 /* Partial pivoting tolerance */ 327 ierr = PetscOptionsReal("-mat_klu_pivot_tol","Partial pivoting tolerance","None",lu->Common.tol,&lu->Common.tol,NULL);CHKERRQ(ierr); 328 /* BTF pre-ordering */ 329 ierr = PetscOptionsInt("-mat_klu_use_btf","Enable BTF preordering","None",(PetscInt)lu->Common.btf,(PetscInt*)&lu->Common.btf,NULL);CHKERRQ(ierr); 330 /* Matrix reordering */ 331 ierr = PetscOptionsEList("-mat_klu_ordering","Internal ordering method","None",KluOrderingTypes,sizeof(KluOrderingTypes)/sizeof(KluOrderingTypes[0]),KluOrderingTypes[0],&idx,&flg);CHKERRQ(ierr); 332 lu->Common.ordering = (int)idx; 333 /* Matrix row scaling */ 334 ierr = PetscOptionsEList("-mat_klu_row_scale","Matrix row scaling","None",scale,3,scale[0],&idx,&flg);CHKERRQ(ierr); 335 PetscOptionsEnd(); 336 *F = B; 337 PetscFunctionReturn(0); 338 } 339