1 #if defined(PETSC_HAVE_LIBMKL_INTEL_ILP64) 2 #define MKL_ILP64 3 #endif 4 5 #include <../src/mat/impls/aij/seq/aij.h> /*I "petscmat.h" I*/ 6 #include <../src/mat/impls/dense/seq/dense.h> 7 8 #include <stdio.h> 9 #include <stdlib.h> 10 #include <math.h> 11 #include <mkl.h> 12 13 /* 14 * Possible mkl_pardiso phases that controls the execution of the solver. 15 * For more information check mkl_pardiso manual. 16 */ 17 #define JOB_ANALYSIS 11 18 #define JOB_ANALYSIS_NUMERICAL_FACTORIZATION 12 19 #define JOB_ANALYSIS_NUMERICAL_FACTORIZATION_SOLVE_ITERATIVE_REFINEMENT 13 20 #define JOB_NUMERICAL_FACTORIZATION 22 21 #define JOB_NUMERICAL_FACTORIZATION_SOLVE_ITERATIVE_REFINEMENT 23 22 #define JOB_SOLVE_ITERATIVE_REFINEMENT 33 23 #define JOB_SOLVE_FORWARD_SUBSTITUTION 331 24 #define JOB_SOLVE_DIAGONAL_SUBSTITUTION 332 25 #define JOB_SOLVE_BACKWARD_SUBSTITUTION 333 26 #define JOB_RELEASE_OF_LU_MEMORY 0 27 #define JOB_RELEASE_OF_ALL_MEMORY -1 28 29 #define IPARM_SIZE 64 30 31 #if defined(PETSC_USE_64BIT_INDICES) 32 #if defined(PETSC_HAVE_LIBMKL_INTEL_ILP64) 33 /* sizeof(MKL_INT) == sizeof(long long int) if ilp64*/ 34 #define INT_TYPE long long int 35 #define MKL_PARDISO pardiso 36 #define MKL_PARDISO_INIT pardisoinit 37 #else 38 #define INT_TYPE long long int 39 #define MKL_PARDISO pardiso_64 40 #define MKL_PARDISO_INIT pardiso_64init 41 #endif 42 #else 43 #define INT_TYPE int 44 #define MKL_PARDISO pardiso 45 #define MKL_PARDISO_INIT pardisoinit 46 #endif 47 48 49 /* 50 * Internal data structure. 51 * For more information check mkl_pardiso manual. 52 */ 53 typedef struct { 54 55 /* Configuration vector*/ 56 INT_TYPE iparm[IPARM_SIZE]; 57 58 /* 59 * Internal mkl_pardiso memory location. 60 * After the first call to mkl_pardiso do not modify pt, as that could cause a serious memory leak. 61 */ 62 void *pt[IPARM_SIZE]; 63 64 /* Basic mkl_pardiso info*/ 65 INT_TYPE phase, maxfct, mnum, mtype, n, nrhs, msglvl, err; 66 67 /* Matrix structure*/ 68 void *a; 69 INT_TYPE *ia, *ja; 70 71 /* Number of non-zero elements*/ 72 INT_TYPE nz; 73 74 /* Row permutaton vector*/ 75 INT_TYPE *perm; 76 77 /* Define if matrix preserves sparse structure.*/ 78 MatStructure matstruc; 79 80 /* True if mkl_pardiso function have been used.*/ 81 PetscBool CleanUp; 82 } Mat_MKL_PARDISO; 83 84 85 void pardiso_64init(void *pt, INT_TYPE *mtype, INT_TYPE iparm []) 86 { 87 int iparm_copy[IPARM_SIZE], mtype_copy, i; 88 89 mtype_copy = *mtype; 90 pardisoinit(pt, &mtype_copy, iparm_copy); 91 for(i = 0; i < IPARM_SIZE; i++){ 92 iparm[i] = iparm_copy[i]; 93 } 94 } 95 96 97 /* 98 * Copy the elements of matrix A. 99 * Input: 100 * - Mat A: MATSEQAIJ matrix 101 * - int shift: matrix index. 102 * - 0 for c representation 103 * - 1 for fortran representation 104 * - MatReuse reuse: 105 * - MAT_INITIAL_MATRIX: Create a new aij representation 106 * - MAT_REUSE_MATRIX: Reuse all aij representation and just change values 107 * Output: 108 * - int *nnz: Number of nonzero-elements. 109 * - int **r pointer to i index 110 * - int **c pointer to j elements 111 * - MATRIXTYPE **v: Non-zero elements 112 */ 113 #undef __FUNCT__ 114 #define __FUNCT__ "MatCopy_MKL_PARDISO" 115 PetscErrorCode MatCopy_MKL_PARDISO(Mat A, MatReuse reuse, INT_TYPE *nnz, INT_TYPE **r, INT_TYPE **c, void **v) 116 { 117 Mat_SeqAIJ *aa=(Mat_SeqAIJ*)A->data; 118 119 PetscFunctionBegin; 120 *v=aa->a; 121 if (reuse == MAT_INITIAL_MATRIX) { 122 *r = (INT_TYPE*)aa->i; 123 *c = (INT_TYPE*)aa->j; 124 *nnz = aa->nz; 125 } 126 PetscFunctionReturn(0); 127 } 128 129 /* 130 * Free memory for Mat_MKL_PARDISO structure and pointers to objects. 131 */ 132 #undef __FUNCT__ 133 #define __FUNCT__ "MatDestroy_MKL_PARDISO" 134 PetscErrorCode MatDestroy_MKL_PARDISO(Mat A) 135 { 136 Mat_MKL_PARDISO *mat_mkl_pardiso=(Mat_MKL_PARDISO*)A->spptr; 137 PetscErrorCode ierr; 138 139 PetscFunctionBegin; 140 /* Terminate instance, deallocate memories */ 141 if (mat_mkl_pardiso->CleanUp) { 142 mat_mkl_pardiso->phase = JOB_RELEASE_OF_ALL_MEMORY; 143 144 MKL_PARDISO (mat_mkl_pardiso->pt, 145 &mat_mkl_pardiso->maxfct, 146 &mat_mkl_pardiso->mnum, 147 &mat_mkl_pardiso->mtype, 148 &mat_mkl_pardiso->phase, 149 &mat_mkl_pardiso->n, 150 NULL, 151 NULL, 152 NULL, 153 mat_mkl_pardiso->perm, 154 &mat_mkl_pardiso->nrhs, 155 mat_mkl_pardiso->iparm, 156 &mat_mkl_pardiso->msglvl, 157 NULL, 158 NULL, 159 &mat_mkl_pardiso->err); 160 } 161 ierr = PetscFree(mat_mkl_pardiso->perm);CHKERRQ(ierr); 162 ierr = PetscFree(A->spptr);CHKERRQ(ierr); 163 164 /* clear composed functions */ 165 ierr = PetscObjectComposeFunction((PetscObject)A,"MatFactorGetSolverPackage_C",NULL);CHKERRQ(ierr); 166 ierr = PetscObjectComposeFunction((PetscObject)A,"MatMkl_PardisoSetCntl_C",NULL);CHKERRQ(ierr); 167 168 ierr = MatDestroy_SeqAIJ(A);CHKERRQ(ierr); 169 PetscFunctionReturn(0); 170 } 171 172 /* 173 * Computes Ax = b 174 */ 175 #undef __FUNCT__ 176 #define __FUNCT__ "MatSolve_MKL_PARDISO" 177 PetscErrorCode MatSolve_MKL_PARDISO(Mat A,Vec b,Vec x) 178 { 179 Mat_MKL_PARDISO *mat_mkl_pardiso=(Mat_MKL_PARDISO*)(A)->spptr; 180 PetscErrorCode ierr; 181 PetscScalar *xarray; 182 const PetscScalar *barray; 183 184 PetscFunctionBegin; 185 mat_mkl_pardiso->nrhs = 1; 186 ierr = VecGetArray(x,&xarray);CHKERRQ(ierr); 187 ierr = VecGetArrayRead(b,&barray);CHKERRQ(ierr); 188 189 /* solve phase */ 190 /*-------------*/ 191 mat_mkl_pardiso->phase = JOB_SOLVE_ITERATIVE_REFINEMENT; 192 MKL_PARDISO (mat_mkl_pardiso->pt, 193 &mat_mkl_pardiso->maxfct, 194 &mat_mkl_pardiso->mnum, 195 &mat_mkl_pardiso->mtype, 196 &mat_mkl_pardiso->phase, 197 &mat_mkl_pardiso->n, 198 mat_mkl_pardiso->a, 199 mat_mkl_pardiso->ia, 200 mat_mkl_pardiso->ja, 201 mat_mkl_pardiso->perm, 202 &mat_mkl_pardiso->nrhs, 203 mat_mkl_pardiso->iparm, 204 &mat_mkl_pardiso->msglvl, 205 (void*)barray, 206 (void*)xarray, 207 &mat_mkl_pardiso->err); 208 209 if (mat_mkl_pardiso->err < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MKL_PARDISO: err=%d. Please check manual\n",mat_mkl_pardiso->err); 210 ierr = VecRestoreArray(x,&xarray);CHKERRQ(ierr); 211 ierr = VecRestoreArrayRead(b,&barray);CHKERRQ(ierr); 212 mat_mkl_pardiso->CleanUp = PETSC_TRUE; 213 PetscFunctionReturn(0); 214 } 215 216 217 #undef __FUNCT__ 218 #define __FUNCT__ "MatSolveTranspose_MKL_PARDISO" 219 PetscErrorCode MatSolveTranspose_MKL_PARDISO(Mat A,Vec b,Vec x) 220 { 221 Mat_MKL_PARDISO *mat_mkl_pardiso=(Mat_MKL_PARDISO*)A->spptr; 222 PetscErrorCode ierr; 223 224 PetscFunctionBegin; 225 #if defined(PETSC_USE_COMPLEX) 226 mat_mkl_pardiso->iparm[12 - 1] = 1; 227 #else 228 mat_mkl_pardiso->iparm[12 - 1] = 2; 229 #endif 230 ierr = MatSolve_MKL_PARDISO(A,b,x);CHKERRQ(ierr); 231 mat_mkl_pardiso->iparm[12 - 1] = 0; 232 PetscFunctionReturn(0); 233 } 234 235 236 #undef __FUNCT__ 237 #define __FUNCT__ "MatMatSolve_MKL_PARDISO" 238 PetscErrorCode MatMatSolve_MKL_PARDISO(Mat A,Mat B,Mat X) 239 { 240 Mat_MKL_PARDISO *mat_mkl_pardiso=(Mat_MKL_PARDISO*)(A)->spptr; 241 PetscErrorCode ierr; 242 PetscScalar *barray, *xarray; 243 PetscBool flg; 244 245 PetscFunctionBegin; 246 ierr = PetscObjectTypeCompare((PetscObject)B,MATSEQDENSE,&flg);CHKERRQ(ierr); 247 if (!flg) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Matrix B must be MATSEQDENSE matrix"); 248 ierr = PetscObjectTypeCompare((PetscObject)X,MATSEQDENSE,&flg);CHKERRQ(ierr); 249 if (!flg) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Matrix X must be MATSEQDENSE matrix"); 250 251 ierr = MatGetSize(B,NULL,(PetscInt*)&mat_mkl_pardiso->nrhs);CHKERRQ(ierr); 252 253 if(mat_mkl_pardiso->nrhs > 0){ 254 ierr = MatDenseGetArray(B,&barray); 255 ierr = MatDenseGetArray(X,&xarray); 256 257 /* solve phase */ 258 /*-------------*/ 259 mat_mkl_pardiso->phase = JOB_SOLVE_ITERATIVE_REFINEMENT; 260 MKL_PARDISO (mat_mkl_pardiso->pt, 261 &mat_mkl_pardiso->maxfct, 262 &mat_mkl_pardiso->mnum, 263 &mat_mkl_pardiso->mtype, 264 &mat_mkl_pardiso->phase, 265 &mat_mkl_pardiso->n, 266 mat_mkl_pardiso->a, 267 mat_mkl_pardiso->ia, 268 mat_mkl_pardiso->ja, 269 mat_mkl_pardiso->perm, 270 &mat_mkl_pardiso->nrhs, 271 mat_mkl_pardiso->iparm, 272 &mat_mkl_pardiso->msglvl, 273 (void*)barray, 274 (void*)xarray, 275 &mat_mkl_pardiso->err); 276 if (mat_mkl_pardiso->err < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MKL_PARDISO: err=%d. Please check manual\n",mat_mkl_pardiso->err); 277 } 278 mat_mkl_pardiso->CleanUp = PETSC_TRUE; 279 PetscFunctionReturn(0); 280 } 281 282 /* 283 * LU Decomposition 284 */ 285 #undef __FUNCT__ 286 #define __FUNCT__ "MatFactorNumeric_MKL_PARDISO" 287 PetscErrorCode MatFactorNumeric_MKL_PARDISO(Mat F,Mat A,const MatFactorInfo *info) 288 { 289 Mat_MKL_PARDISO *mat_mkl_pardiso=(Mat_MKL_PARDISO*)(F)->spptr; 290 PetscErrorCode ierr; 291 292 /* numerical factorization phase */ 293 /*-------------------------------*/ 294 PetscFunctionBegin; 295 mat_mkl_pardiso->matstruc = SAME_NONZERO_PATTERN; 296 ierr = MatCopy_MKL_PARDISO(A, MAT_REUSE_MATRIX, &mat_mkl_pardiso->nz, &mat_mkl_pardiso->ia, &mat_mkl_pardiso->ja, &mat_mkl_pardiso->a);CHKERRQ(ierr); 297 298 /* numerical factorization phase */ 299 /*-------------------------------*/ 300 mat_mkl_pardiso->phase = JOB_NUMERICAL_FACTORIZATION; 301 MKL_PARDISO (mat_mkl_pardiso->pt, 302 &mat_mkl_pardiso->maxfct, 303 &mat_mkl_pardiso->mnum, 304 &mat_mkl_pardiso->mtype, 305 &mat_mkl_pardiso->phase, 306 &mat_mkl_pardiso->n, 307 mat_mkl_pardiso->a, 308 mat_mkl_pardiso->ia, 309 mat_mkl_pardiso->ja, 310 mat_mkl_pardiso->perm, 311 &mat_mkl_pardiso->nrhs, 312 mat_mkl_pardiso->iparm, 313 &mat_mkl_pardiso->msglvl, 314 NULL, 315 NULL, 316 &mat_mkl_pardiso->err); 317 if (mat_mkl_pardiso->err < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MKL_PARDISO: err=%d. Please check manual\n",mat_mkl_pardiso->err); 318 319 mat_mkl_pardiso->matstruc = SAME_NONZERO_PATTERN; 320 mat_mkl_pardiso->CleanUp = PETSC_TRUE; 321 PetscFunctionReturn(0); 322 } 323 324 /* Sets mkl_pardiso options from the options database */ 325 #undef __FUNCT__ 326 #define __FUNCT__ "PetscSetMKL_PARDISOFromOptions" 327 PetscErrorCode PetscSetMKL_PARDISOFromOptions(Mat F, Mat A) 328 { 329 Mat_MKL_PARDISO *mat_mkl_pardiso = (Mat_MKL_PARDISO*)F->spptr; 330 PetscErrorCode ierr; 331 PetscInt icntl; 332 PetscBool flg; 333 int pt[IPARM_SIZE], threads = 1; 334 335 PetscFunctionBegin; 336 ierr = PetscOptionsBegin(PetscObjectComm((PetscObject)A),((PetscObject)A)->prefix,"MKL_PARDISO Options","Mat");CHKERRQ(ierr); 337 ierr = PetscOptionsInt("-mat_mkl_pardiso_65","Number of threads to use","None",threads,&threads,&flg);CHKERRQ(ierr); 338 if (flg) mkl_set_num_threads(threads); 339 340 ierr = PetscOptionsInt("-mat_mkl_pardiso_66","Maximum number of factors with identical sparsity structure that must be kept in memory at the same time","None",mat_mkl_pardiso->maxfct,&icntl,&flg);CHKERRQ(ierr); 341 if (flg) mat_mkl_pardiso->maxfct = icntl; 342 343 ierr = PetscOptionsInt("-mat_mkl_pardiso_67","Indicates the actual matrix for the solution phase","None",mat_mkl_pardiso->mnum,&icntl,&flg);CHKERRQ(ierr); 344 if (flg) mat_mkl_pardiso->mnum = icntl; 345 346 ierr = PetscOptionsInt("-mat_mkl_pardiso_68","Message level information","None",mat_mkl_pardiso->msglvl,&icntl,&flg);CHKERRQ(ierr); 347 if (flg) mat_mkl_pardiso->msglvl = icntl; 348 349 ierr = PetscOptionsInt("-mat_mkl_pardiso_69","Defines the matrix type","None",mat_mkl_pardiso->mtype,&icntl,&flg);CHKERRQ(ierr); 350 if(flg){ 351 mat_mkl_pardiso->mtype = icntl; 352 MKL_PARDISO_INIT(&pt, &mat_mkl_pardiso->mtype, mat_mkl_pardiso->iparm); 353 #if defined(PETSC_USE_REAL_SINGLE) 354 mat_mkl_pardiso->iparm[27] = 1; 355 #else 356 mat_mkl_pardiso->iparm[27] = 0; 357 #endif 358 mat_mkl_pardiso->iparm[34] = 1; 359 } 360 ierr = PetscOptionsInt("-mat_mkl_pardiso_1","Use default values","None",mat_mkl_pardiso->iparm[0],&icntl,&flg);CHKERRQ(ierr); 361 362 if(flg && icntl != 0){ 363 ierr = PetscOptionsInt("-mat_mkl_pardiso_2","Fill-in reducing ordering for the input matrix","None",mat_mkl_pardiso->iparm[1],&icntl,&flg);CHKERRQ(ierr); 364 if (flg) mat_mkl_pardiso->iparm[1] = icntl; 365 366 ierr = PetscOptionsInt("-mat_mkl_pardiso_4","Preconditioned CGS/CG","None",mat_mkl_pardiso->iparm[3],&icntl,&flg);CHKERRQ(ierr); 367 if (flg) mat_mkl_pardiso->iparm[3] = icntl; 368 369 ierr = PetscOptionsInt("-mat_mkl_pardiso_5","User permutation","None",mat_mkl_pardiso->iparm[4],&icntl,&flg);CHKERRQ(ierr); 370 if (flg) mat_mkl_pardiso->iparm[4] = icntl; 371 372 ierr = PetscOptionsInt("-mat_mkl_pardiso_6","Write solution on x","None",mat_mkl_pardiso->iparm[5],&icntl,&flg);CHKERRQ(ierr); 373 if (flg) mat_mkl_pardiso->iparm[5] = icntl; 374 375 ierr = PetscOptionsInt("-mat_mkl_pardiso_8","Iterative refinement step","None",mat_mkl_pardiso->iparm[7],&icntl,&flg);CHKERRQ(ierr); 376 if (flg) mat_mkl_pardiso->iparm[7] = icntl; 377 378 ierr = PetscOptionsInt("-mat_mkl_pardiso_10","Pivoting perturbation","None",mat_mkl_pardiso->iparm[9],&icntl,&flg);CHKERRQ(ierr); 379 if (flg) mat_mkl_pardiso->iparm[9] = icntl; 380 381 ierr = PetscOptionsInt("-mat_mkl_pardiso_11","Scaling vectors","None",mat_mkl_pardiso->iparm[10],&icntl,&flg);CHKERRQ(ierr); 382 if (flg) mat_mkl_pardiso->iparm[10] = icntl; 383 384 ierr = PetscOptionsInt("-mat_mkl_pardiso_12","Solve with transposed or conjugate transposed matrix A","None",mat_mkl_pardiso->iparm[11],&icntl,&flg);CHKERRQ(ierr); 385 if (flg) mat_mkl_pardiso->iparm[11] = icntl; 386 387 ierr = PetscOptionsInt("-mat_mkl_pardiso_13","Improved accuracy using (non-) symmetric weighted matching","None",mat_mkl_pardiso->iparm[12],&icntl,&flg);CHKERRQ(ierr); 388 if (flg) mat_mkl_pardiso->iparm[12] = icntl; 389 390 ierr = PetscOptionsInt("-mat_mkl_pardiso_18","Numbers of non-zero elements","None",mat_mkl_pardiso->iparm[17],&icntl,&flg);CHKERRQ(ierr); 391 if (flg) mat_mkl_pardiso->iparm[17] = icntl; 392 393 ierr = PetscOptionsInt("-mat_mkl_pardiso_19","Report number of floating point operations","None",mat_mkl_pardiso->iparm[18],&icntl,&flg);CHKERRQ(ierr); 394 if (flg) mat_mkl_pardiso->iparm[18] = icntl; 395 396 ierr = PetscOptionsInt("-mat_mkl_pardiso_21","Pivoting for symmetric indefinite matrices","None",mat_mkl_pardiso->iparm[20],&icntl,&flg);CHKERRQ(ierr); 397 if (flg) mat_mkl_pardiso->iparm[20] = icntl; 398 399 ierr = PetscOptionsInt("-mat_mkl_pardiso_24","Parallel factorization control","None",mat_mkl_pardiso->iparm[23],&icntl,&flg);CHKERRQ(ierr); 400 if (flg) mat_mkl_pardiso->iparm[23] = icntl; 401 402 ierr = PetscOptionsInt("-mat_mkl_pardiso_25","Parallel forward/backward solve control","None",mat_mkl_pardiso->iparm[24],&icntl,&flg);CHKERRQ(ierr); 403 if (flg) mat_mkl_pardiso->iparm[24] = icntl; 404 405 ierr = PetscOptionsInt("-mat_mkl_pardiso_27","Matrix checker","None",mat_mkl_pardiso->iparm[26],&icntl,&flg);CHKERRQ(ierr); 406 if (flg) mat_mkl_pardiso->iparm[26] = icntl; 407 408 ierr = PetscOptionsInt("-mat_mkl_pardiso_31","Partial solve and computing selected components of the solution vectors","None",mat_mkl_pardiso->iparm[30],&icntl,&flg);CHKERRQ(ierr); 409 if (flg) mat_mkl_pardiso->iparm[30] = icntl; 410 411 ierr = PetscOptionsInt("-mat_mkl_pardiso_34","Optimal number of threads for conditional numerical reproducibility (CNR) mode","None",mat_mkl_pardiso->iparm[33],&icntl,&flg);CHKERRQ(ierr); 412 if (flg) mat_mkl_pardiso->iparm[33] = icntl; 413 414 ierr = PetscOptionsInt("-mat_mkl_pardiso_60","Intel MKL_PARDISO mode","None",mat_mkl_pardiso->iparm[59],&icntl,&flg);CHKERRQ(ierr); 415 if (flg) mat_mkl_pardiso->iparm[59] = icntl; 416 } 417 PetscOptionsEnd(); 418 PetscFunctionReturn(0); 419 } 420 421 #undef __FUNCT__ 422 #define __FUNCT__ "MatFactorMKL_PARDISOInitialize_Private" 423 PetscErrorCode MatFactorMKL_PARDISOInitialize_Private(Mat A, MatFactorType ftype, Mat_MKL_PARDISO *mat_mkl_pardiso) 424 { 425 PetscErrorCode ierr; 426 PetscInt i; 427 428 PetscFunctionBegin; 429 for ( i = 0; i < IPARM_SIZE; i++ ){ 430 mat_mkl_pardiso->iparm[i] = 0; 431 } 432 433 for ( i = 0; i < IPARM_SIZE; i++ ){ 434 mat_mkl_pardiso->pt[i] = 0; 435 } 436 437 /*Default options for both sym and unsym */ 438 mat_mkl_pardiso->iparm[ 0] = 1; /* Solver default parameters overriden with provided by iparm */ 439 mat_mkl_pardiso->iparm[ 1] = 2; /* Metis reordering */ 440 mat_mkl_pardiso->iparm[ 5] = 0; /* Write solution into x */ 441 mat_mkl_pardiso->iparm[ 7] = 2; /* Max number of iterative refinement steps */ 442 mat_mkl_pardiso->iparm[17] = -1; /* Output: Number of nonzeros in the factor LU */ 443 mat_mkl_pardiso->iparm[18] = -1; /* Output: Mflops for LU factorization */ 444 #if 0 445 mat_mkl_pardiso->iparm[23] = 1; /* Parallel factorization control*/ 446 #endif 447 mat_mkl_pardiso->iparm[34] = 1; /* Cluster Sparse Solver use C-style indexing for ia and ja arrays */ 448 mat_mkl_pardiso->iparm[39] = 0; /* Input: matrix/rhs/solution stored on master */ 449 450 mat_mkl_pardiso->CleanUp = PETSC_FALSE; 451 mat_mkl_pardiso->maxfct = 1; /* Maximum number of numerical factorizations. */ 452 mat_mkl_pardiso->mnum = 1; /* Which factorization to use. */ 453 mat_mkl_pardiso->msglvl = 0; /* 0: do not print 1: Print statistical information in file */ 454 mat_mkl_pardiso->phase = -1; 455 mat_mkl_pardiso->err = 0; 456 457 mat_mkl_pardiso->n = A->rmap->N; 458 mat_mkl_pardiso->nrhs = 1; 459 mat_mkl_pardiso->err = 0; 460 mat_mkl_pardiso->phase = -1; 461 462 if(ftype == MAT_FACTOR_LU){ 463 /*Default type for non-sym*/ 464 #if defined(PETSC_USE_COMPLEX) 465 mat_mkl_pardiso->mtype = 13; 466 #else 467 mat_mkl_pardiso->mtype = 11; 468 #endif 469 470 mat_mkl_pardiso->iparm[ 9] = 13; /* Perturb the pivot elements with 1E-13 */ 471 mat_mkl_pardiso->iparm[10] = 1; /* Use nonsymmetric permutation and scaling MPS */ 472 mat_mkl_pardiso->iparm[12] = 1; /* Switch on Maximum Weighted Matching algorithm (default for non-symmetric) */ 473 474 } else { 475 /*Default type for sym*/ 476 #if defined(PETSC_USE_COMPLEX) 477 mat_mkl_pardiso ->mtype = 3; 478 #else 479 mat_mkl_pardiso ->mtype = -2; 480 #endif 481 mat_mkl_pardiso->iparm[ 9] = 13; /* Perturb the pivot elements with 1E-13 */ 482 mat_mkl_pardiso->iparm[10] = 0; /* Use nonsymmetric permutation and scaling MPS */ 483 mat_mkl_pardiso->iparm[12] = 1; /* Switch on Maximum Weighted Matching algorithm (default for non-symmetric) */ 484 /* mat_mkl_pardiso->iparm[20] = 1; */ /* Apply 1x1 and 2x2 Bunch-Kaufman pivoting during the factorization process */ 485 #if defined(PETSC_USE_DEBUG) 486 mat_mkl_pardiso->iparm[26] = 1; /* Matrix checker */ 487 #endif 488 } 489 ierr = PetscMalloc1(A->rmap->N*sizeof(INT_TYPE), &mat_mkl_pardiso->perm);CHKERRQ(ierr); 490 for(i = 0; i < A->rmap->N; i++){ 491 mat_mkl_pardiso->perm[i] = 0; 492 } 493 PetscFunctionReturn(0); 494 } 495 496 /* 497 * Symbolic decomposition. Mkl_Pardiso analysis phase. 498 */ 499 #undef __FUNCT__ 500 #define __FUNCT__ "MatFactorSymbolic_AIJMKL_PARDISO_Private" 501 PetscErrorCode MatFactorSymbolic_AIJMKL_PARDISO_Private(Mat F,Mat A,const MatFactorInfo *info) 502 { 503 Mat_MKL_PARDISO *mat_mkl_pardiso = (Mat_MKL_PARDISO*)F->spptr; 504 PetscErrorCode ierr; 505 506 PetscFunctionBegin; 507 mat_mkl_pardiso->matstruc = DIFFERENT_NONZERO_PATTERN; 508 509 /* Set MKL_PARDISO options from the options database */ 510 ierr = PetscSetMKL_PARDISOFromOptions(F,A);CHKERRQ(ierr); 511 512 ierr = MatCopy_MKL_PARDISO(A, MAT_INITIAL_MATRIX, &mat_mkl_pardiso->nz, &mat_mkl_pardiso->ia, &mat_mkl_pardiso->ja, &mat_mkl_pardiso->a);CHKERRQ(ierr); 513 mat_mkl_pardiso->n = A->rmap->N; 514 515 /* analysis phase */ 516 /*----------------*/ 517 mat_mkl_pardiso->phase = JOB_ANALYSIS; 518 519 MKL_PARDISO (mat_mkl_pardiso->pt, 520 &mat_mkl_pardiso->maxfct, 521 &mat_mkl_pardiso->mnum, 522 &mat_mkl_pardiso->mtype, 523 &mat_mkl_pardiso->phase, 524 &mat_mkl_pardiso->n, 525 mat_mkl_pardiso->a, 526 mat_mkl_pardiso->ia, 527 mat_mkl_pardiso->ja, 528 mat_mkl_pardiso->perm, 529 &mat_mkl_pardiso->nrhs, 530 mat_mkl_pardiso->iparm, 531 &mat_mkl_pardiso->msglvl, 532 NULL, 533 NULL, 534 &mat_mkl_pardiso->err); 535 if (mat_mkl_pardiso->err < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MKL_PARDISO: err=%d\n. Please check manual",mat_mkl_pardiso->err); 536 537 mat_mkl_pardiso->CleanUp = PETSC_TRUE; 538 539 if(F->factortype == MAT_FACTOR_LU){ 540 F->ops->lufactornumeric = MatFactorNumeric_MKL_PARDISO; 541 } else { 542 F->ops->choleskyfactornumeric = MatFactorNumeric_MKL_PARDISO; 543 } 544 F->ops->solve = MatSolve_MKL_PARDISO; 545 F->ops->solvetranspose = MatSolveTranspose_MKL_PARDISO; 546 F->ops->matsolve = MatMatSolve_MKL_PARDISO; 547 PetscFunctionReturn(0); 548 } 549 550 #undef __FUNCT__ 551 #define __FUNCT__ "MatLUFactorSymbolic_AIJMKL_PARDISO" 552 PetscErrorCode MatLUFactorSymbolic_AIJMKL_PARDISO(Mat F,Mat A,IS r,IS c,const MatFactorInfo *info) 553 { 554 PetscErrorCode ierr; 555 556 PetscFunctionBegin; 557 ierr = MatFactorSymbolic_AIJMKL_PARDISO_Private(F, A, info);CHKERRQ(ierr); 558 PetscFunctionReturn(0); 559 } 560 561 #undef __FUNCT__ 562 #define __FUNCT__ "MatCholeskyFactorSymbolic_AIJMKL_PARDISO" 563 PetscErrorCode MatCholeskyFactorSymbolic_AIJMKL_PARDISO(Mat F,Mat A,IS r,const MatFactorInfo *info) 564 { 565 PetscErrorCode ierr; 566 567 PetscFunctionBegin; 568 ierr = MatFactorSymbolic_AIJMKL_PARDISO_Private(F, A, info);CHKERRQ(ierr); 569 PetscFunctionReturn(0); 570 } 571 572 #undef __FUNCT__ 573 #define __FUNCT__ "MatView_MKL_PARDISO" 574 PetscErrorCode MatView_MKL_PARDISO(Mat A, PetscViewer viewer) 575 { 576 PetscErrorCode ierr; 577 PetscBool iascii; 578 PetscViewerFormat format; 579 Mat_MKL_PARDISO *mat_mkl_pardiso=(Mat_MKL_PARDISO*)A->spptr; 580 PetscInt i; 581 582 PetscFunctionBegin; 583 /* check if matrix is mkl_pardiso type */ 584 if (A->ops->solve != MatSolve_MKL_PARDISO) PetscFunctionReturn(0); 585 586 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);CHKERRQ(ierr); 587 if (iascii) { 588 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 589 if (format == PETSC_VIEWER_ASCII_INFO) { 590 ierr = PetscViewerASCIIPrintf(viewer,"MKL_PARDISO run parameters:\n");CHKERRQ(ierr); 591 ierr = PetscViewerASCIIPrintf(viewer,"MKL_PARDISO phase: %d \n",mat_mkl_pardiso->phase);CHKERRQ(ierr); 592 for(i = 1; i <= 64; i++){ 593 ierr = PetscViewerASCIIPrintf(viewer,"MKL_PARDISO iparm[%d]: %d \n",i, mat_mkl_pardiso->iparm[i - 1]);CHKERRQ(ierr); 594 } 595 ierr = PetscViewerASCIIPrintf(viewer,"MKL_PARDISO maxfct: %d \n", mat_mkl_pardiso->maxfct);CHKERRQ(ierr); 596 ierr = PetscViewerASCIIPrintf(viewer,"MKL_PARDISO mnum: %d \n", mat_mkl_pardiso->mnum);CHKERRQ(ierr); 597 ierr = PetscViewerASCIIPrintf(viewer,"MKL_PARDISO mtype: %d \n", mat_mkl_pardiso->mtype);CHKERRQ(ierr); 598 ierr = PetscViewerASCIIPrintf(viewer,"MKL_PARDISO n: %d \n", mat_mkl_pardiso->n);CHKERRQ(ierr); 599 ierr = PetscViewerASCIIPrintf(viewer,"MKL_PARDISO nrhs: %d \n", mat_mkl_pardiso->nrhs);CHKERRQ(ierr); 600 ierr = PetscViewerASCIIPrintf(viewer,"MKL_PARDISO msglvl: %d \n", mat_mkl_pardiso->msglvl);CHKERRQ(ierr); 601 } 602 } 603 PetscFunctionReturn(0); 604 } 605 606 607 #undef __FUNCT__ 608 #define __FUNCT__ "MatGetInfo_MKL_PARDISO" 609 PetscErrorCode MatGetInfo_MKL_PARDISO(Mat A, MatInfoType flag, MatInfo *info) 610 { 611 Mat_MKL_PARDISO *mat_mkl_pardiso =(Mat_MKL_PARDISO*)A->spptr; 612 613 PetscFunctionBegin; 614 info->block_size = 1.0; 615 info->nz_allocated = mat_mkl_pardiso->nz + 0.0; 616 info->nz_unneeded = 0.0; 617 info->assemblies = 0.0; 618 info->mallocs = 0.0; 619 info->memory = 0.0; 620 info->fill_ratio_given = 0; 621 info->fill_ratio_needed = 0; 622 info->factor_mallocs = 0; 623 PetscFunctionReturn(0); 624 } 625 626 #undef __FUNCT__ 627 #define __FUNCT__ "MatMkl_PardisoSetCntl_MKL_PARDISO" 628 PetscErrorCode MatMkl_PardisoSetCntl_MKL_PARDISO(Mat F,PetscInt icntl,PetscInt ival) 629 { 630 Mat_MKL_PARDISO *mat_mkl_pardiso =(Mat_MKL_PARDISO*)F->spptr; 631 632 PetscFunctionBegin; 633 if(icntl <= 64){ 634 mat_mkl_pardiso->iparm[icntl - 1] = ival; 635 } else { 636 if(icntl == 65) 637 mkl_set_num_threads((int)ival); 638 else if(icntl == 66) 639 mat_mkl_pardiso->maxfct = ival; 640 else if(icntl == 67) 641 mat_mkl_pardiso->mnum = ival; 642 else if(icntl == 68) 643 mat_mkl_pardiso->msglvl = ival; 644 else if(icntl == 69){ 645 int pt[IPARM_SIZE]; 646 mat_mkl_pardiso->mtype = ival; 647 MKL_PARDISO_INIT(&pt, &mat_mkl_pardiso->mtype, mat_mkl_pardiso->iparm); 648 #if defined(PETSC_USE_REAL_SINGLE) 649 mat_mkl_pardiso->iparm[27] = 1; 650 #else 651 mat_mkl_pardiso->iparm[27] = 0; 652 #endif 653 mat_mkl_pardiso->iparm[34] = 1; 654 } 655 } 656 PetscFunctionReturn(0); 657 } 658 659 #undef __FUNCT__ 660 #define __FUNCT__ "MatMkl_PardisoSetCntl" 661 /*@ 662 MatMkl_PardisoSetCntl - Set Mkl_Pardiso parameters 663 664 Logically Collective on Mat 665 666 Input Parameters: 667 + F - the factored matrix obtained by calling MatGetFactor() 668 . icntl - index of Mkl_Pardiso parameter 669 - ival - value of Mkl_Pardiso parameter 670 671 Options Database: 672 . -mat_mkl_pardiso_<icntl> <ival> 673 674 Level: beginner 675 676 References: Mkl_Pardiso Users' Guide 677 678 .seealso: MatGetFactor() 679 @*/ 680 PetscErrorCode MatMkl_PardisoSetCntl(Mat F,PetscInt icntl,PetscInt ival) 681 { 682 PetscErrorCode ierr; 683 684 PetscFunctionBegin; 685 ierr = PetscTryMethod(F,"MatMkl_PardisoSetCntl_C",(Mat,PetscInt,PetscInt),(F,icntl,ival));CHKERRQ(ierr); 686 PetscFunctionReturn(0); 687 } 688 689 /*MC 690 MATSOLVERMKL_PARDISO - A matrix type providing direct solvers (LU) for 691 sequential matrices via the external package MKL_PARDISO. 692 693 Works with MATSEQAIJ matrices 694 695 Options Database Keys: 696 + -mat_mkl_pardiso_65 - Number of thrads to use 697 . -mat_mkl_pardiso_66 - Maximum number of factors with identical sparsity structure that must be kept in memory at the same time 698 . -mat_mkl_pardiso_67 - Indicates the actual matrix for the solution phase 699 . -mat_mkl_pardiso_68 - Message level information 700 . -mat_mkl_pardiso_69 - Defines the matrix type. IMPORTANT: When you set this flag, iparm parameters are going to be set to the default ones for the matrix type 701 . -mat_mkl_pardiso_1 - Use default values 702 . -mat_mkl_pardiso_2 - Fill-in reducing ordering for the input matrix 703 . -mat_mkl_pardiso_4 - Preconditioned CGS/CG 704 . -mat_mkl_pardiso_5 - User permutation 705 . -mat_mkl_pardiso_6 - Write solution on x 706 . -mat_mkl_pardiso_8 - Iterative refinement step 707 . -mat_mkl_pardiso_10 - Pivoting perturbation 708 . -mat_mkl_pardiso_11 - Scaling vectors 709 . -mat_mkl_pardiso_12 - Solve with transposed or conjugate transposed matrix A 710 . -mat_mkl_pardiso_13 - Improved accuracy using (non-) symmetric weighted matching 711 . -mat_mkl_pardiso_18 - Numbers of non-zero elements 712 . -mat_mkl_pardiso_19 - Report number of floating point operations 713 . -mat_mkl_pardiso_21 - Pivoting for symmetric indefinite matrices 714 . -mat_mkl_pardiso_24 - Parallel factorization control 715 . -mat_mkl_pardiso_25 - Parallel forward/backward solve control 716 . -mat_mkl_pardiso_27 - Matrix checker 717 . -mat_mkl_pardiso_31 - Partial solve and computing selected components of the solution vectors 718 . -mat_mkl_pardiso_34 - Optimal number of threads for conditional numerical reproducibility (CNR) mode 719 - -mat_mkl_pardiso_60 - Intel MKL_PARDISO mode 720 721 Level: beginner 722 723 For more information please check mkl_pardiso manual 724 725 .seealso: PCFactorSetMatSolverPackage(), MatSolverPackage 726 727 M*/ 728 #undef __FUNCT__ 729 #define __FUNCT__ "MatFactorGetSolverPackage_mkl_pardiso" 730 static PetscErrorCode MatFactorGetSolverPackage_mkl_pardiso(Mat A, const MatSolverPackage *type) 731 { 732 PetscFunctionBegin; 733 *type = MATSOLVERMKL_PARDISO; 734 PetscFunctionReturn(0); 735 } 736 737 /* MatGetFactor for Seq sbAIJ matrices */ 738 #undef __FUNCT__ 739 #define __FUNCT__ "MatGetFactor_sbaij_mkl_pardiso" 740 PETSC_EXTERN PetscErrorCode MatGetFactor_sbaij_mkl_pardiso(Mat A,MatFactorType ftype,Mat *F) 741 { 742 Mat B; 743 PetscErrorCode ierr; 744 Mat_MKL_PARDISO *mat_mkl_pardiso; 745 PetscBool isSeqSBAIJ; 746 PetscInt bs; 747 748 PetscFunctionBegin; 749 /* Create the factorization matrix */ 750 ierr = PetscObjectTypeCompare((PetscObject)A,MATSEQSBAIJ,&isSeqSBAIJ);CHKERRQ(ierr); 751 ierr = MatCreate(PetscObjectComm((PetscObject)A),&B);CHKERRQ(ierr); 752 ierr = MatSetSizes(B,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);CHKERRQ(ierr); 753 ierr = MatSetType(B,((PetscObject)A)->type_name);CHKERRQ(ierr); 754 ierr = MatSeqSBAIJSetPreallocation(B,1,0,NULL);CHKERRQ(ierr); 755 ierr = MatGetBlockSize(A,&bs); CHKERRQ(ierr); 756 757 if(bs != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrice MATSEQSBAIJ with block size other than 1 is not supported by Pardiso"); 758 if(ftype != MAT_FACTOR_CHOLESKY) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrice MATSEQAIJ should be used only with MAT_FACTOR_CHOLESKY."); 759 760 B->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_AIJMKL_PARDISO; 761 B->factortype = MAT_FACTOR_CHOLESKY; 762 B->ops->destroy = MatDestroy_MKL_PARDISO; 763 B->ops->view = MatView_MKL_PARDISO; 764 B->factortype = ftype; 765 B->ops->getinfo = MatGetInfo_MKL_PARDISO; 766 B->assembled = PETSC_TRUE; /* required by -ksp_view */ 767 768 ierr = PetscNewLog(B,&mat_mkl_pardiso);CHKERRQ(ierr); 769 B->spptr = mat_mkl_pardiso; 770 ierr = PetscObjectComposeFunction((PetscObject)B,"MatFactorGetSolverPackage_C",MatFactorGetSolverPackage_mkl_pardiso);CHKERRQ(ierr); 771 ierr = PetscObjectComposeFunction((PetscObject)B,"MatMkl_PardisoSetCntl_C",MatMkl_PardisoSetCntl_MKL_PARDISO);CHKERRQ(ierr); 772 ierr = MatFactorMKL_PARDISOInitialize_Private(A, ftype, mat_mkl_pardiso);CHKERRQ(ierr); 773 *F = B; 774 PetscFunctionReturn(0); 775 } 776 777 /* MatGetFactor for Seq AIJ matrices */ 778 #undef __FUNCT__ 779 #define __FUNCT__ "MatGetFactor_aij_mkl_pardiso" 780 PETSC_EXTERN PetscErrorCode MatGetFactor_aij_mkl_pardiso(Mat A,MatFactorType ftype,Mat *F) 781 { 782 Mat B; 783 PetscErrorCode ierr; 784 Mat_MKL_PARDISO *mat_mkl_pardiso; 785 PetscBool isSeqAIJ; 786 787 PetscFunctionBegin; 788 /* Create the factorization matrix */ 789 ierr = PetscObjectTypeCompare((PetscObject)A,MATSEQAIJ,&isSeqAIJ);CHKERRQ(ierr); 790 ierr = MatCreate(PetscObjectComm((PetscObject)A),&B);CHKERRQ(ierr); 791 ierr = MatSetSizes(B,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);CHKERRQ(ierr); 792 ierr = MatSetType(B,((PetscObject)A)->type_name);CHKERRQ(ierr); 793 ierr = MatSeqAIJSetPreallocation(B,0,NULL);CHKERRQ(ierr); 794 795 if(ftype != MAT_FACTOR_LU) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrice MATSEQAIJ should be used only with MAT_FACTOR_LU."); 796 797 B->ops->lufactorsymbolic = MatLUFactorSymbolic_AIJMKL_PARDISO; 798 B->factortype = MAT_FACTOR_LU; 799 B->ops->destroy = MatDestroy_MKL_PARDISO; 800 B->ops->view = MatView_MKL_PARDISO; 801 B->factortype = ftype; 802 B->ops->getinfo = MatGetInfo_MKL_PARDISO; 803 B->assembled = PETSC_TRUE; /* required by -ksp_view */ 804 805 ierr = PetscNewLog(B,&mat_mkl_pardiso);CHKERRQ(ierr); 806 B->spptr = mat_mkl_pardiso; 807 ierr = PetscObjectComposeFunction((PetscObject)B,"MatFactorGetSolverPackage_C",MatFactorGetSolverPackage_mkl_pardiso);CHKERRQ(ierr); 808 ierr = PetscObjectComposeFunction((PetscObject)B,"MatMkl_PardisoSetCntl_C",MatMkl_PardisoSetCntl_MKL_PARDISO);CHKERRQ(ierr); 809 ierr = MatFactorMKL_PARDISOInitialize_Private(A, ftype, mat_mkl_pardiso);CHKERRQ(ierr); 810 811 *F = B; 812 PetscFunctionReturn(0); 813 } 814 815 #undef __FUNCT__ 816 #define __FUNCT__ "MatSolverPackageRegister_MKL_Pardiso" 817 PETSC_EXTERN PetscErrorCode MatSolverPackageRegister_MKL_Pardiso(void) 818 { 819 PetscErrorCode ierr; 820 821 PetscFunctionBegin; 822 ierr = MatSolverPackageRegister(MATSOLVERMKL_PARDISO,MATSEQAIJ, MAT_FACTOR_LU, MatGetFactor_aij_mkl_pardiso );CHKERRQ(ierr); 823 ierr = MatSolverPackageRegister(MATSOLVERMKL_PARDISO,MATSEQSBAIJ, MAT_FACTOR_CHOLESKY,MatGetFactor_sbaij_mkl_pardiso);CHKERRQ(ierr); 824 PetscFunctionReturn(0); 825 } 826 827