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/aij/mpi/mpiaij.h> 7 8 #include <stdio.h> 9 #include <stdlib.h> 10 #include <math.h> 11 #include <mkl.h> 12 #include <mkl_cluster_sparse_solver.h> 13 14 /* 15 * Possible mkl_cpardiso phases that controls the execution of the solver. 16 * For more information check mkl_cpardiso manual. 17 */ 18 #define JOB_ANALYSIS 11 19 #define JOB_ANALYSIS_NUMERICAL_FACTORIZATION 12 20 #define JOB_ANALYSIS_NUMERICAL_FACTORIZATION_SOLVE_ITERATIVE_REFINEMENT 13 21 #define JOB_NUMERICAL_FACTORIZATION 22 22 #define JOB_NUMERICAL_FACTORIZATION_SOLVE_ITERATIVE_REFINEMENT 23 23 #define JOB_SOLVE_ITERATIVE_REFINEMENT 33 24 #define JOB_SOLVE_FORWARD_SUBSTITUTION 331 25 #define JOB_SOLVE_DIAGONAL_SUBSTITUTION 332 26 #define JOB_SOLVE_BACKWARD_SUBSTITUTION 333 27 #define JOB_RELEASE_OF_LU_MEMORY 0 28 #define JOB_RELEASE_OF_ALL_MEMORY -1 29 30 #define IPARM_SIZE 64 31 #define INT_TYPE MKL_INT 32 33 static const char *Err_MSG_CPardiso(int errNo){ 34 switch (errNo) { 35 case -1: 36 return "input inconsistent"; break; 37 case -2: 38 return "not enough memory"; break; 39 case -3: 40 return "reordering problem"; break; 41 case -4: 42 return "zero pivot, numerical factorization or iterative refinement problem"; break; 43 case -5: 44 return "unclassified (internal) error"; break; 45 case -6: 46 return "preordering failed (matrix types 11, 13 only)"; break; 47 case -7: 48 return "diagonal matrix problem"; break; 49 case -8: 50 return "32-bit integer overflow problem"; break; 51 case -9: 52 return "not enough memory for OOC"; break; 53 case -10: 54 return "problems with opening OOC temporary files"; break; 55 case -11: 56 return "read/write problems with the OOC data file"; break; 57 default : 58 return "unknown error"; 59 } 60 } 61 62 /* 63 * Internal data structure. 64 * For more information check mkl_cpardiso manual. 65 */ 66 67 typedef struct { 68 69 /* Configuration vector */ 70 INT_TYPE iparm[IPARM_SIZE]; 71 72 /* 73 * Internal mkl_cpardiso memory location. 74 * After the first call to mkl_cpardiso do not modify pt, as that could cause a serious memory leak. 75 */ 76 void *pt[IPARM_SIZE]; 77 78 MPI_Comm comm_mkl_cpardiso; 79 80 /* Basic mkl_cpardiso info*/ 81 INT_TYPE phase, maxfct, mnum, mtype, n, nrhs, msglvl, err; 82 83 /* Matrix structure */ 84 PetscScalar *a; 85 86 INT_TYPE *ia, *ja; 87 88 /* Number of non-zero elements */ 89 INT_TYPE nz; 90 91 /* Row permutaton vector*/ 92 INT_TYPE *perm; 93 94 /* Define is matrix preserve sparce structure. */ 95 MatStructure matstruc; 96 97 PetscErrorCode (*ConvertToTriples)(Mat, MatReuse, PetscInt*, PetscInt**, PetscInt**, PetscScalar**); 98 99 /* True if mkl_cpardiso function have been used. */ 100 PetscBool CleanUp; 101 } Mat_MKL_CPARDISO; 102 103 /* 104 * Copy the elements of matrix A. 105 * Input: 106 * - Mat A: MATSEQAIJ matrix 107 * - int shift: matrix index. 108 * - 0 for c representation 109 * - 1 for fortran representation 110 * - MatReuse reuse: 111 * - MAT_INITIAL_MATRIX: Create a new aij representation 112 * - MAT_REUSE_MATRIX: Reuse all aij representation and just change values 113 * Output: 114 * - int *nnz: Number of nonzero-elements. 115 * - int **r pointer to i index 116 * - int **c pointer to j elements 117 * - MATRIXTYPE **v: Non-zero elements 118 */ 119 #undef __FUNCT__ 120 #define __FUNCT__ "MatCopy_seqaij_seqaij_MKL_CPARDISO" 121 PetscErrorCode MatCopy_seqaij_seqaij_MKL_CPARDISO(Mat A, MatReuse reuse, PetscInt *nnz, PetscInt **r, PetscInt **c, PetscScalar **v) 122 { 123 Mat_SeqAIJ *aa=(Mat_SeqAIJ*)A->data; 124 125 PetscFunctionBegin; 126 *v=aa->a; 127 if (reuse == MAT_INITIAL_MATRIX) { 128 *r = (INT_TYPE*)aa->i; 129 *c = (INT_TYPE*)aa->j; 130 *nnz = aa->nz; 131 } 132 PetscFunctionReturn(0); 133 } 134 135 #undef __FUNCT__ 136 #define __FUNCT__ "MatConvertToTriples_mpiaij_mpiaij_MKL_CPARDISO" 137 PetscErrorCode MatConvertToTriples_mpiaij_mpiaij_MKL_CPARDISO(Mat A, MatReuse reuse, PetscInt *nnz, PetscInt **r, PetscInt **c, PetscScalar **v) 138 { 139 const PetscInt *ai, *aj, *bi, *bj,*garray,m=A->rmap->n,*ajj,*bjj; 140 PetscErrorCode ierr; 141 PetscInt rstart,nz,i,j,jj,irow,countA,countB; 142 PetscInt *row,*col; 143 const PetscScalar *av, *bv,*v1,*v2; 144 PetscScalar *val; 145 Mat_MPIAIJ *mat = (Mat_MPIAIJ*)A->data; 146 Mat_SeqAIJ *aa = (Mat_SeqAIJ*)(mat->A)->data; 147 Mat_SeqAIJ *bb = (Mat_SeqAIJ*)(mat->B)->data; 148 PetscInt nn, colA_start,jB,jcol; 149 150 PetscFunctionBegin; 151 ai=aa->i; aj=aa->j; bi=bb->i; bj=bb->j; rstart= A->rmap->rstart; 152 av=aa->a; bv=bb->a; 153 154 garray = mat->garray; 155 156 if (reuse == MAT_INITIAL_MATRIX) { 157 nz = aa->nz + bb->nz; 158 *nnz = nz; 159 ierr = PetscMalloc((nz*(sizeof(PetscInt)+sizeof(PetscScalar)) + (m+1)*sizeof(PetscInt)), &row);CHKERRQ(ierr); 160 col = row + m + 1; 161 val = (PetscScalar*)(col + nz); 162 *r = row; *c = col; *v = val; 163 row[0] = 0; 164 } else { 165 row = *r; col = *c; val = *v; 166 } 167 168 nz = 0; 169 for (i=0; i<m; i++) { 170 row[i] = nz; 171 countA = ai[i+1] - ai[i]; 172 countB = bi[i+1] - bi[i]; 173 ajj = aj + ai[i]; /* ptr to the beginning of this row */ 174 bjj = bj + bi[i]; 175 176 /* B part, smaller col index */ 177 colA_start = rstart + ajj[0]; /* the smallest global col index of A */ 178 jB = 0; 179 for (j=0; j<countB; j++) { 180 jcol = garray[bjj[j]]; 181 if (jcol > colA_start) { 182 jB = j; 183 break; 184 } 185 col[nz] = jcol; 186 val[nz++] = *bv++; 187 if (j==countB-1) jB = countB; 188 } 189 190 /* A part */ 191 for (j=0; j<countA; j++) { 192 col[nz] = rstart + ajj[j]; 193 val[nz++] = *av++; 194 } 195 196 /* B part, larger col index */ 197 for (j=jB; j<countB; j++) { 198 col[nz] = garray[bjj[j]]; 199 val[nz++] = *bv++; 200 } 201 } 202 row[m] = nz; 203 204 PetscFunctionReturn(0); 205 } 206 207 /* 208 * Free memory for Mat_MKL_CPARDISO structure and pointers to objects. 209 */ 210 #undef __FUNCT__ 211 #define __FUNCT__ "MatDestroy_MKL_CPARDISO" 212 PetscErrorCode MatDestroy_MKL_CPARDISO(Mat A) 213 { 214 Mat_MKL_CPARDISO *mat_mkl_cpardiso=(Mat_MKL_CPARDISO*)A->spptr; 215 PetscErrorCode ierr; 216 217 PetscFunctionBegin; 218 /* Terminate instance, deallocate memories */ 219 if (mat_mkl_cpardiso->CleanUp) { 220 mat_mkl_cpardiso->phase = JOB_RELEASE_OF_ALL_MEMORY; 221 222 cluster_sparse_solver ( 223 mat_mkl_cpardiso->pt, 224 &mat_mkl_cpardiso->maxfct, 225 &mat_mkl_cpardiso->mnum, 226 &mat_mkl_cpardiso->mtype, 227 &mat_mkl_cpardiso->phase, 228 &mat_mkl_cpardiso->n, 229 NULL, 230 NULL, 231 NULL, 232 mat_mkl_cpardiso->perm, 233 &mat_mkl_cpardiso->nrhs, 234 mat_mkl_cpardiso->iparm, 235 &mat_mkl_cpardiso->msglvl, 236 NULL, 237 NULL, 238 &mat_mkl_cpardiso->comm_mkl_cpardiso, 239 &mat_mkl_cpardiso->err); 240 } 241 ierr = PetscFree(A->spptr);CHKERRQ(ierr); 242 ierr = MPI_Comm_free(&(mat_mkl_cpardiso->comm_mkl_cpardiso));CHKERRQ(ierr); 243 244 /* clear composed functions */ 245 ierr = PetscObjectComposeFunction((PetscObject)A,"MatFactorGetSolverPackage_C",NULL);CHKERRQ(ierr); 246 ierr = PetscObjectComposeFunction((PetscObject)A,"MatMkl_CPardisoSetCntl_C",NULL);CHKERRQ(ierr); 247 PetscFunctionReturn(0); 248 } 249 250 /* 251 * Computes Ax = b 252 */ 253 #undef __FUNCT__ 254 #define __FUNCT__ "MatSolve_MKL_CPARDISO" 255 PetscErrorCode MatSolve_MKL_CPARDISO(Mat A,Vec b,Vec x) 256 { 257 Mat_MKL_CPARDISO *mat_mkl_cpardiso=(Mat_MKL_CPARDISO*)(A)->spptr; 258 PetscErrorCode ierr; 259 PetscScalar *xarray; 260 const PetscScalar *barray; 261 262 PetscFunctionBegin; 263 mat_mkl_cpardiso->nrhs = 1; 264 ierr = VecGetArray(x,&xarray);CHKERRQ(ierr); 265 ierr = VecGetArrayRead(b,&barray);CHKERRQ(ierr); 266 267 /* solve phase */ 268 /*-------------*/ 269 mat_mkl_cpardiso->phase = JOB_SOLVE_ITERATIVE_REFINEMENT; 270 cluster_sparse_solver ( 271 mat_mkl_cpardiso->pt, 272 &mat_mkl_cpardiso->maxfct, 273 &mat_mkl_cpardiso->mnum, 274 &mat_mkl_cpardiso->mtype, 275 &mat_mkl_cpardiso->phase, 276 &mat_mkl_cpardiso->n, 277 mat_mkl_cpardiso->a, 278 mat_mkl_cpardiso->ia, 279 mat_mkl_cpardiso->ja, 280 mat_mkl_cpardiso->perm, 281 &mat_mkl_cpardiso->nrhs, 282 mat_mkl_cpardiso->iparm, 283 &mat_mkl_cpardiso->msglvl, 284 (void*)barray, 285 (void*)xarray, 286 &mat_mkl_cpardiso->comm_mkl_cpardiso, 287 &mat_mkl_cpardiso->err); 288 289 if (mat_mkl_cpardiso->err < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MKL_CPARDISO: err=%d, msg = \"%s\". Please check manual\n",mat_mkl_cpardiso->err,Err_MSG_CPardiso(mat_mkl_cpardiso->err)); 290 291 ierr = VecRestoreArray(x,&xarray);CHKERRQ(ierr); 292 ierr = VecRestoreArrayRead(b,&barray);CHKERRQ(ierr); 293 mat_mkl_cpardiso->CleanUp = PETSC_TRUE; 294 PetscFunctionReturn(0); 295 } 296 297 #undef __FUNCT__ 298 #define __FUNCT__ "MatSolveTranspose_MKL_CPARDISO" 299 PetscErrorCode MatSolveTranspose_MKL_CPARDISO(Mat A,Vec b,Vec x) 300 { 301 Mat_MKL_CPARDISO *mat_mkl_cpardiso=(Mat_MKL_CPARDISO*)A->spptr; 302 PetscErrorCode ierr; 303 304 PetscFunctionBegin; 305 #if defined(PETSC_USE_COMPLEX) 306 mat_mkl_cpardiso->iparm[12 - 1] = 1; 307 #else 308 mat_mkl_cpardiso->iparm[12 - 1] = 2; 309 #endif 310 ierr = MatSolve_MKL_CPARDISO(A,b,x);CHKERRQ(ierr); 311 mat_mkl_cpardiso->iparm[12 - 1] = 0; 312 PetscFunctionReturn(0); 313 } 314 315 #undef __FUNCT__ 316 #define __FUNCT__ "MatMatSolve_MKL_CPARDISO" 317 PetscErrorCode MatMatSolve_MKL_CPARDISO(Mat A,Mat B,Mat X) 318 { 319 Mat_MKL_CPARDISO *mat_mkl_cpardiso=(Mat_MKL_CPARDISO*)(A)->spptr; 320 PetscErrorCode ierr; 321 PetscScalar *barray, *xarray; 322 PetscBool flg; 323 324 PetscFunctionBegin; 325 ierr = PetscObjectTypeCompare((PetscObject)B,MATSEQDENSE,&flg);CHKERRQ(ierr); 326 if (!flg) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Matrix B must be MATSEQDENSE matrix"); 327 ierr = PetscObjectTypeCompare((PetscObject)X,MATSEQDENSE,&flg);CHKERRQ(ierr); 328 if (!flg) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Matrix X must be MATSEQDENSE matrix"); 329 330 ierr = MatGetSize(B,NULL,(PetscInt*)&mat_mkl_cpardiso->nrhs);CHKERRQ(ierr); 331 332 if(mat_mkl_cpardiso->nrhs > 0){ 333 ierr = MatDenseGetArray(B,&barray); 334 ierr = MatDenseGetArray(X,&xarray); 335 336 /* solve phase */ 337 /*-------------*/ 338 mat_mkl_cpardiso->phase = JOB_SOLVE_ITERATIVE_REFINEMENT; 339 cluster_sparse_solver ( 340 mat_mkl_cpardiso->pt, 341 &mat_mkl_cpardiso->maxfct, 342 &mat_mkl_cpardiso->mnum, 343 &mat_mkl_cpardiso->mtype, 344 &mat_mkl_cpardiso->phase, 345 &mat_mkl_cpardiso->n, 346 mat_mkl_cpardiso->a, 347 mat_mkl_cpardiso->ia, 348 mat_mkl_cpardiso->ja, 349 mat_mkl_cpardiso->perm, 350 &mat_mkl_cpardiso->nrhs, 351 mat_mkl_cpardiso->iparm, 352 &mat_mkl_cpardiso->msglvl, 353 (void*)barray, 354 (void*)xarray, 355 &mat_mkl_cpardiso->comm_mkl_cpardiso, 356 &mat_mkl_cpardiso->err); 357 if (mat_mkl_cpardiso->err < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MKL_CPARDISO: err=%d, msg = \"%s\". Please check manual\n",mat_mkl_cpardiso->err,Err_MSG_CPardiso(mat_mkl_cpardiso->err)); 358 } 359 mat_mkl_cpardiso->CleanUp = PETSC_TRUE; 360 PetscFunctionReturn(0); 361 362 } 363 364 /* 365 * LU Decomposition 366 */ 367 #undef __FUNCT__ 368 #define __FUNCT__ "MatFactorNumeric_MKL_CPARDISO" 369 PetscErrorCode MatFactorNumeric_MKL_CPARDISO(Mat F,Mat A,const MatFactorInfo *info) 370 { 371 Mat_MKL_CPARDISO *mat_mkl_cpardiso=(Mat_MKL_CPARDISO*)(F)->spptr; 372 PetscErrorCode ierr; 373 374 /* numerical factorization phase */ 375 /*-------------------------------*/ 376 377 PetscFunctionBegin; 378 mat_mkl_cpardiso->matstruc = SAME_NONZERO_PATTERN; 379 ierr = (*mat_mkl_cpardiso->ConvertToTriples)(A, MAT_REUSE_MATRIX,&mat_mkl_cpardiso->nz,&mat_mkl_cpardiso->ia,&mat_mkl_cpardiso->ja,&mat_mkl_cpardiso->a);CHKERRQ(ierr); 380 381 /* numerical factorization phase */ 382 /*-------------------------------*/ 383 mat_mkl_cpardiso->phase = JOB_NUMERICAL_FACTORIZATION; 384 cluster_sparse_solver ( 385 mat_mkl_cpardiso->pt, 386 &mat_mkl_cpardiso->maxfct, 387 &mat_mkl_cpardiso->mnum, 388 &mat_mkl_cpardiso->mtype, 389 &mat_mkl_cpardiso->phase, 390 &mat_mkl_cpardiso->n, 391 mat_mkl_cpardiso->a, 392 mat_mkl_cpardiso->ia, 393 mat_mkl_cpardiso->ja, 394 mat_mkl_cpardiso->perm, 395 &mat_mkl_cpardiso->nrhs, 396 mat_mkl_cpardiso->iparm, 397 &mat_mkl_cpardiso->msglvl, 398 NULL, 399 NULL, 400 &mat_mkl_cpardiso->comm_mkl_cpardiso, 401 &mat_mkl_cpardiso->err); 402 if (mat_mkl_cpardiso->err < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MKL_CPARDISO: err=%d, msg = \"%s\". Please check manual\n",mat_mkl_cpardiso->err,Err_MSG_CPardiso(mat_mkl_cpardiso->err)); 403 404 mat_mkl_cpardiso->matstruc = SAME_NONZERO_PATTERN; 405 mat_mkl_cpardiso->CleanUp = PETSC_TRUE; 406 PetscFunctionReturn(0); 407 } 408 409 /* Sets mkl_cpardiso options from the options database */ 410 #undef __FUNCT__ 411 #define __FUNCT__ "PetscSetMKL_CPARDISOFromOptions" 412 PetscErrorCode PetscSetMKL_CPARDISOFromOptions(Mat F, Mat A) 413 { 414 Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO*)F->spptr; 415 PetscErrorCode ierr; 416 PetscInt icntl; 417 PetscBool flg; 418 int pt[IPARM_SIZE], threads; 419 420 PetscFunctionBegin; 421 ierr = PetscOptionsBegin(PetscObjectComm((PetscObject)A),((PetscObject)A)->prefix,"MKL_CPARDISO Options","Mat");CHKERRQ(ierr); 422 ierr = PetscOptionsInt("-mat_mkl_cpardiso_65", 423 "Number of threads to use", 424 "None", 425 threads, 426 &threads, 427 &flg);CHKERRQ(ierr); 428 if (flg) mkl_set_num_threads(threads); 429 430 ierr = PetscOptionsInt("-mat_mkl_cpardiso_66", 431 "Maximum number of factors with identical sparsity structure that must be kept in memory at the same time", 432 "None", 433 mat_mkl_cpardiso->maxfct, 434 &icntl, 435 &flg);CHKERRQ(ierr); 436 if (flg) mat_mkl_cpardiso->maxfct = icntl; 437 438 ierr = PetscOptionsInt("-mat_mkl_cpardiso_67", 439 "Indicates the actual matrix for the solution phase", 440 "None", 441 mat_mkl_cpardiso->mnum, 442 &icntl, 443 &flg);CHKERRQ(ierr); 444 if (flg) mat_mkl_cpardiso->mnum = icntl; 445 446 ierr = PetscOptionsInt("-mat_mkl_cpardiso_68", 447 "Message level information", 448 "None", 449 mat_mkl_cpardiso->msglvl, 450 &icntl, 451 &flg);CHKERRQ(ierr); 452 if (flg) mat_mkl_cpardiso->msglvl = icntl; 453 454 ierr = PetscOptionsInt("-mat_mkl_cpardiso_69", 455 "Defines the matrix type", 456 "None", 457 mat_mkl_cpardiso->mtype, 458 &icntl, 459 &flg);CHKERRQ(ierr); 460 if(flg){ 461 mat_mkl_cpardiso->mtype = icntl; 462 #if defined(PETSC_USE_REAL_SINGLE) 463 mat_mkl_cpardiso->iparm[27] = 1; 464 #else 465 mat_mkl_cpardiso->iparm[27] = 0; 466 #endif 467 mat_mkl_cpardiso->iparm[34] = 1; 468 } 469 ierr = PetscOptionsInt("-mat_mkl_cpardiso_1", 470 "Use default values", 471 "None", 472 mat_mkl_cpardiso->iparm[0], 473 &icntl, 474 &flg);CHKERRQ(ierr); 475 476 if(flg && icntl != 0){ 477 ierr = PetscOptionsInt("-mat_mkl_cpardiso_2", 478 "Fill-in reducing ordering for the input matrix", 479 "None", 480 mat_mkl_cpardiso->iparm[1], 481 &icntl, 482 &flg);CHKERRQ(ierr); 483 if (flg) mat_mkl_cpardiso->iparm[1] = icntl; 484 485 ierr = PetscOptionsInt("-mat_mkl_cpardiso_4", 486 "Preconditioned CGS/CG", 487 "None", 488 mat_mkl_cpardiso->iparm[3], 489 &icntl, 490 &flg);CHKERRQ(ierr); 491 if (flg) mat_mkl_cpardiso->iparm[3] = icntl; 492 493 ierr = PetscOptionsInt("-mat_mkl_cpardiso_5", 494 "User permutation", 495 "None", 496 mat_mkl_cpardiso->iparm[4], 497 &icntl, 498 &flg);CHKERRQ(ierr); 499 if (flg) mat_mkl_cpardiso->iparm[4] = icntl; 500 501 ierr = PetscOptionsInt("-mat_mkl_cpardiso_6", 502 "Write solution on x", 503 "None", 504 mat_mkl_cpardiso->iparm[5], 505 &icntl, 506 &flg);CHKERRQ(ierr); 507 if (flg) mat_mkl_cpardiso->iparm[5] = icntl; 508 509 ierr = PetscOptionsInt("-mat_mkl_cpardiso_8", 510 "Iterative refinement step", 511 "None", 512 mat_mkl_cpardiso->iparm[7], 513 &icntl, 514 &flg);CHKERRQ(ierr); 515 if (flg) mat_mkl_cpardiso->iparm[7] = icntl; 516 517 ierr = PetscOptionsInt("-mat_mkl_cpardiso_10", 518 "Pivoting perturbation", 519 "None", 520 mat_mkl_cpardiso->iparm[9], 521 &icntl, 522 &flg);CHKERRQ(ierr); 523 if (flg) mat_mkl_cpardiso->iparm[9] = icntl; 524 525 ierr = PetscOptionsInt("-mat_mkl_cpardiso_11", 526 "Scaling vectors", 527 "None", 528 mat_mkl_cpardiso->iparm[10], 529 &icntl, 530 &flg);CHKERRQ(ierr); 531 if (flg) mat_mkl_cpardiso->iparm[10] = icntl; 532 533 ierr = PetscOptionsInt("-mat_mkl_cpardiso_12", 534 "Solve with transposed or conjugate transposed matrix A", 535 "None", 536 mat_mkl_cpardiso->iparm[11], 537 &icntl, 538 &flg);CHKERRQ(ierr); 539 if (flg) mat_mkl_cpardiso->iparm[11] = icntl; 540 541 ierr = PetscOptionsInt("-mat_mkl_cpardiso_13", 542 "Improved accuracy using (non-) symmetric weighted matching", 543 "None", 544 mat_mkl_cpardiso->iparm[12], 545 &icntl, 546 &flg);CHKERRQ(ierr); 547 if (flg) mat_mkl_cpardiso->iparm[12] = icntl; 548 549 ierr = PetscOptionsInt("-mat_mkl_cpardiso_18", 550 "Numbers of non-zero elements", 551 "None", 552 mat_mkl_cpardiso->iparm[17], 553 &icntl, 554 &flg);CHKERRQ(ierr); 555 if (flg) mat_mkl_cpardiso->iparm[17] = icntl; 556 557 ierr = PetscOptionsInt("-mat_mkl_cpardiso_19", 558 "Report number of floating point operations", 559 "None", 560 mat_mkl_cpardiso->iparm[18], 561 &icntl, 562 &flg);CHKERRQ(ierr); 563 if (flg) mat_mkl_cpardiso->iparm[18] = icntl; 564 565 ierr = PetscOptionsInt("-mat_mkl_cpardiso_21", 566 "Pivoting for symmetric indefinite matrices", 567 "None", 568 mat_mkl_cpardiso->iparm[20], 569 &icntl, 570 &flg);CHKERRQ(ierr); 571 if (flg) mat_mkl_cpardiso->iparm[20] = icntl; 572 573 ierr = PetscOptionsInt("-mat_mkl_cpardiso_24", 574 "Parallel factorization control", 575 "None", 576 mat_mkl_cpardiso->iparm[23], 577 &icntl, 578 &flg);CHKERRQ(ierr); 579 if (flg) mat_mkl_cpardiso->iparm[23] = icntl; 580 581 ierr = PetscOptionsInt("-mat_mkl_cpardiso_25", 582 "Parallel forward/backward solve control", 583 "None", 584 mat_mkl_cpardiso->iparm[24], 585 &icntl, 586 &flg);CHKERRQ(ierr); 587 if (flg) mat_mkl_cpardiso->iparm[24] = icntl; 588 589 ierr = PetscOptionsInt("-mat_mkl_cpardiso_27", 590 "Matrix checker", 591 "None", 592 mat_mkl_cpardiso->iparm[26], 593 &icntl, 594 &flg);CHKERRQ(ierr); 595 if (flg) mat_mkl_cpardiso->iparm[26] = icntl; 596 597 ierr = PetscOptionsInt("-mat_mkl_cpardiso_31", 598 "Partial solve and computing selected components of the solution vectors", 599 "None", 600 mat_mkl_cpardiso->iparm[30], 601 &icntl, 602 &flg);CHKERRQ(ierr); 603 if (flg) mat_mkl_cpardiso->iparm[30] = icntl; 604 605 ierr = PetscOptionsInt("-mat_mkl_cpardiso_34", 606 "Optimal number of threads for conditional numerical reproducibility (CNR) mode", 607 "None", 608 mat_mkl_cpardiso->iparm[33], 609 &icntl, 610 &flg);CHKERRQ(ierr); 611 if (flg) mat_mkl_cpardiso->iparm[33] = icntl; 612 613 ierr = PetscOptionsInt("-mat_mkl_cpardiso_60", 614 "Intel MKL_CPARDISO mode", 615 "None", 616 mat_mkl_cpardiso->iparm[59], 617 &icntl, 618 &flg);CHKERRQ(ierr); 619 if (flg) mat_mkl_cpardiso->iparm[59] = icntl; 620 } 621 622 PetscOptionsEnd(); 623 PetscFunctionReturn(0); 624 } 625 626 #undef __FUNCT__ 627 #define __FUNCT__ "PetscInitialize_MKL_CPARDISO" 628 PetscErrorCode PetscInitialize_MKL_CPARDISO(Mat A, Mat_MKL_CPARDISO *mat_mkl_cpardiso) 629 { 630 PetscErrorCode ierr; 631 PetscInt i; 632 PetscMPIInt size; 633 634 PetscFunctionBegin; 635 636 ierr = MPI_Comm_dup(PetscObjectComm((PetscObject)A),&(mat_mkl_cpardiso->comm_mkl_cpardiso));CHKERRQ(ierr); 637 ierr = MPI_Comm_size(mat_mkl_cpardiso->comm_mkl_cpardiso, &size);CHKERRQ(ierr); 638 639 mat_mkl_cpardiso->CleanUp = PETSC_FALSE; 640 mat_mkl_cpardiso->maxfct = 1; 641 mat_mkl_cpardiso->mnum = 1; 642 mat_mkl_cpardiso->n = A->rmap->N; 643 mat_mkl_cpardiso->msglvl = 0; 644 mat_mkl_cpardiso->nrhs = 1; 645 mat_mkl_cpardiso->err = 0; 646 mat_mkl_cpardiso->phase = -1; 647 #if defined(PETSC_USE_COMPLEX) 648 mat_mkl_cpardiso->mtype = 13; 649 #else 650 mat_mkl_cpardiso->mtype = 11; 651 #endif 652 653 #if defined(PETSC_USE_REAL_SINGLE) 654 mat_mkl_cpardiso->iparm[27] = 1; 655 #else 656 mat_mkl_cpardiso->iparm[27] = 0; 657 #endif 658 659 mat_mkl_cpardiso->iparm[34] = 1; /* C style */ 660 661 mat_mkl_cpardiso->iparm[ 0] = 1; /* Solver default parameters overriden with provided by iparm */ 662 mat_mkl_cpardiso->iparm[ 1] = 2; /* Use METIS for fill-in reordering */ 663 mat_mkl_cpardiso->iparm[ 5] = 0; /* Write solution into x */ 664 mat_mkl_cpardiso->iparm[ 7] = 2; /* Max number of iterative refinement steps */ 665 mat_mkl_cpardiso->iparm[ 9] = 13; /* Perturb the pivot elements with 1E-13 */ 666 mat_mkl_cpardiso->iparm[10] = 1; /* Use nonsymmetric permutation and scaling MPS */ 667 mat_mkl_cpardiso->iparm[12] = 1; /* Switch on Maximum Weighted Matching algorithm (default for non-symmetric) */ 668 mat_mkl_cpardiso->iparm[17] = -1; /* Output: Number of nonzeros in the factor LU */ 669 mat_mkl_cpardiso->iparm[18] = -1; /* Output: Mflops for LU factorization */ 670 mat_mkl_cpardiso->iparm[26] = 1; /* Check input data for correctness */ 671 672 mat_mkl_cpardiso->iparm[39] = 0; 673 if (size > 1) { 674 mat_mkl_cpardiso->iparm[39] = 2; 675 mat_mkl_cpardiso->iparm[40] = A->rmap->rstart; 676 mat_mkl_cpardiso->iparm[41] = A->rmap->rend-1; 677 } 678 mat_mkl_cpardiso->perm = 0; 679 PetscFunctionReturn(0); 680 } 681 682 /* 683 * Symbolic decomposition. Mkl_Pardiso analysis phase. 684 */ 685 #undef __FUNCT__ 686 #define __FUNCT__ "MatLUFactorSymbolic_AIJMKL_CPARDISO" 687 PetscErrorCode MatLUFactorSymbolic_AIJMKL_CPARDISO(Mat F,Mat A,IS r,IS c,const MatFactorInfo *info) 688 { 689 Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO*)F->spptr; 690 PetscErrorCode ierr; 691 692 PetscFunctionBegin; 693 mat_mkl_cpardiso->matstruc = DIFFERENT_NONZERO_PATTERN; 694 695 /* Set MKL_CPARDISO options from the options database */ 696 ierr = PetscSetMKL_CPARDISOFromOptions(F,A);CHKERRQ(ierr); 697 698 ierr = (*mat_mkl_cpardiso->ConvertToTriples)(A,MAT_INITIAL_MATRIX,&mat_mkl_cpardiso->nz,&mat_mkl_cpardiso->ia,&mat_mkl_cpardiso->ja,&mat_mkl_cpardiso->a);CHKERRQ(ierr); 699 700 mat_mkl_cpardiso->n = A->rmap->N; 701 702 /* analysis phase */ 703 /*----------------*/ 704 mat_mkl_cpardiso->phase = JOB_ANALYSIS; 705 706 cluster_sparse_solver ( 707 mat_mkl_cpardiso->pt, 708 &mat_mkl_cpardiso->maxfct, 709 &mat_mkl_cpardiso->mnum, 710 &mat_mkl_cpardiso->mtype, 711 &mat_mkl_cpardiso->phase, 712 &mat_mkl_cpardiso->n, 713 mat_mkl_cpardiso->a, 714 mat_mkl_cpardiso->ia, 715 mat_mkl_cpardiso->ja, 716 mat_mkl_cpardiso->perm, 717 &mat_mkl_cpardiso->nrhs, 718 mat_mkl_cpardiso->iparm, 719 &mat_mkl_cpardiso->msglvl, 720 NULL, 721 NULL, 722 &mat_mkl_cpardiso->comm_mkl_cpardiso, 723 &mat_mkl_cpardiso->err); 724 725 if (mat_mkl_cpardiso->err < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error reported by MKL_CPARDISO: err=%d, msg = \"%s\". Please check manual\n",mat_mkl_cpardiso->err,Err_MSG_CPardiso(mat_mkl_cpardiso->err)); 726 727 mat_mkl_cpardiso->CleanUp = PETSC_TRUE; 728 F->ops->lufactornumeric = MatFactorNumeric_MKL_CPARDISO; 729 F->ops->solve = MatSolve_MKL_CPARDISO; 730 F->ops->solvetranspose = MatSolveTranspose_MKL_CPARDISO; 731 F->ops->matsolve = MatMatSolve_MKL_CPARDISO; 732 PetscFunctionReturn(0); 733 } 734 735 #undef __FUNCT__ 736 #define __FUNCT__ "MatView_MKL_CPARDISO" 737 PetscErrorCode MatView_MKL_CPARDISO(Mat A, PetscViewer viewer) 738 { 739 PetscErrorCode ierr; 740 PetscBool iascii; 741 PetscViewerFormat format; 742 Mat_MKL_CPARDISO *mat_mkl_cpardiso=(Mat_MKL_CPARDISO*)A->spptr; 743 PetscInt i; 744 745 PetscFunctionBegin; 746 /* check if matrix is mkl_cpardiso type */ 747 if (A->ops->solve != MatSolve_MKL_CPARDISO) PetscFunctionReturn(0); 748 749 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);CHKERRQ(ierr); 750 if (iascii) { 751 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 752 if (format == PETSC_VIEWER_ASCII_INFO) { 753 ierr = PetscViewerASCIIPrintf(viewer,"MKL_CPARDISO run parameters:\n");CHKERRQ(ierr); 754 ierr = PetscViewerASCIIPrintf(viewer,"MKL_CPARDISO phase: %d \n",mat_mkl_cpardiso->phase);CHKERRQ(ierr); 755 for(i = 1; i <= 64; i++){ 756 ierr = PetscViewerASCIIPrintf(viewer,"MKL_CPARDISO iparm[%d]: %d \n",i, mat_mkl_cpardiso->iparm[i - 1]);CHKERRQ(ierr); 757 } 758 ierr = PetscViewerASCIIPrintf(viewer,"MKL_CPARDISO maxfct: %d \n", mat_mkl_cpardiso->maxfct);CHKERRQ(ierr); 759 ierr = PetscViewerASCIIPrintf(viewer,"MKL_CPARDISO mnum: %d \n", mat_mkl_cpardiso->mnum);CHKERRQ(ierr); 760 ierr = PetscViewerASCIIPrintf(viewer,"MKL_CPARDISO mtype: %d \n", mat_mkl_cpardiso->mtype);CHKERRQ(ierr); 761 ierr = PetscViewerASCIIPrintf(viewer,"MKL_CPARDISO n: %d \n", mat_mkl_cpardiso->n);CHKERRQ(ierr); 762 ierr = PetscViewerASCIIPrintf(viewer,"MKL_CPARDISO nrhs: %d \n", mat_mkl_cpardiso->nrhs);CHKERRQ(ierr); 763 ierr = PetscViewerASCIIPrintf(viewer,"MKL_CPARDISO msglvl: %d \n", mat_mkl_cpardiso->msglvl);CHKERRQ(ierr); 764 } 765 } 766 PetscFunctionReturn(0); 767 } 768 769 #undef __FUNCT__ 770 #define __FUNCT__ "MatGetInfo_MKL_CPARDISO" 771 PetscErrorCode MatGetInfo_MKL_CPARDISO(Mat A, MatInfoType flag, MatInfo *info) 772 { 773 Mat_MKL_CPARDISO *mat_mkl_cpardiso =(Mat_MKL_CPARDISO*)A->spptr; 774 775 PetscFunctionBegin; 776 info->block_size = 1.0; 777 info->nz_allocated = mat_mkl_cpardiso->nz + 0.0; 778 info->nz_unneeded = 0.0; 779 info->assemblies = 0.0; 780 info->mallocs = 0.0; 781 info->memory = 0.0; 782 info->fill_ratio_given = 0; 783 info->fill_ratio_needed = 0; 784 info->factor_mallocs = 0; 785 PetscFunctionReturn(0); 786 } 787 788 #undef __FUNCT__ 789 #define __FUNCT__ "MatMkl_CPardisoSetCntl_MKL_CPARDISO" 790 PetscErrorCode MatMkl_CPardisoSetCntl_MKL_CPARDISO(Mat F,PetscInt icntl,PetscInt ival) 791 { 792 Mat_MKL_CPARDISO *mat_mkl_cpardiso =(Mat_MKL_CPARDISO*)F->spptr; 793 794 PetscFunctionBegin; 795 if(icntl <= 64){ 796 mat_mkl_cpardiso->iparm[icntl - 1] = ival; 797 } else { 798 if(icntl == 65) 799 mkl_set_num_threads((int)ival); 800 else if(icntl == 66) 801 mat_mkl_cpardiso->maxfct = ival; 802 else if(icntl == 67) 803 mat_mkl_cpardiso->mnum = ival; 804 else if(icntl == 68) 805 mat_mkl_cpardiso->msglvl = ival; 806 else if(icntl == 69){ 807 int pt[IPARM_SIZE]; 808 mat_mkl_cpardiso->mtype = ival; 809 #if defined(PETSC_USE_REAL_SINGLE) 810 mat_mkl_cpardiso->iparm[27] = 1; 811 #else 812 mat_mkl_cpardiso->iparm[27] = 0; 813 #endif 814 mat_mkl_cpardiso->iparm[34] = 1; 815 } 816 } 817 PetscFunctionReturn(0); 818 } 819 820 #undef __FUNCT__ 821 #define __FUNCT__ "MatMkl_CPardisoSetCntl" 822 /*@ 823 MatMkl_CPardisoSetCntl - Set Mkl_Pardiso parameters 824 825 Logically Collective on Mat 826 827 Input Parameters: 828 + F - the factored matrix obtained by calling MatGetFactor() 829 . icntl - index of Mkl_Pardiso parameter 830 - ival - value of Mkl_Pardiso parameter 831 832 Options Database: 833 . -mat_mkl_cpardiso_<icntl> <ival> 834 835 Level: beginner 836 837 References: Mkl_Pardiso Users' Guide 838 839 .seealso: MatGetFactor() 840 @*/ 841 PetscErrorCode MatMkl_CPardisoSetCntl(Mat F,PetscInt icntl,PetscInt ival) 842 { 843 PetscErrorCode ierr; 844 845 PetscFunctionBegin; 846 ierr = PetscTryMethod(F,"MatMkl_CPardisoSetCntl_C",(Mat,PetscInt,PetscInt),(F,icntl,ival));CHKERRQ(ierr); 847 PetscFunctionReturn(0); 848 } 849 850 #undef __FUNCT__ 851 #define __FUNCT__ "MatFactorGetSolverPackage_mkl_cpardiso" 852 static PetscErrorCode MatFactorGetSolverPackage_mkl_cpardiso(Mat A, const MatSolverPackage *type) 853 { 854 PetscFunctionBegin; 855 *type = MATSOLVERMKL_CPARDISO; 856 PetscFunctionReturn(0); 857 } 858 859 /* MatGetFactor for MPI AIJ matrices */ 860 #undef __FUNCT__ 861 #define __FUNCT__ "MatGetFactor_mpiaij_mkl_cpardiso" 862 PETSC_EXTERN PetscErrorCode MatGetFactor_mpiaij_mkl_cpardiso(Mat A,MatFactorType ftype,Mat *F) 863 { 864 Mat B; 865 PetscErrorCode ierr; 866 Mat_MKL_CPARDISO *mat_mkl_cpardiso; 867 PetscBool isSeqAIJ; 868 869 PetscFunctionBegin; 870 /* Create the factorization matrix */ 871 872 ierr = PetscObjectTypeCompare((PetscObject)A,MATSEQAIJ,&isSeqAIJ);CHKERRQ(ierr); 873 ierr = MatCreate(PetscObjectComm((PetscObject)A),&B);CHKERRQ(ierr); 874 ierr = MatSetSizes(B,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);CHKERRQ(ierr); 875 ierr = MatSetType(B,((PetscObject)A)->type_name);CHKERRQ(ierr); 876 877 ierr = PetscNewLog(B,&mat_mkl_cpardiso);CHKERRQ(ierr); 878 879 if (isSeqAIJ) { 880 ierr = MatSeqAIJSetPreallocation(B,0,NULL);CHKERRQ(ierr); 881 mat_mkl_cpardiso->ConvertToTriples = MatCopy_seqaij_seqaij_MKL_CPARDISO; 882 } else { 883 mat_mkl_cpardiso->ConvertToTriples = MatConvertToTriples_mpiaij_mpiaij_MKL_CPARDISO; 884 ierr = MatMPIAIJSetPreallocation(B,0,NULL,0,NULL);CHKERRQ(ierr); 885 } 886 887 B->ops->lufactorsymbolic = MatLUFactorSymbolic_AIJMKL_CPARDISO; 888 B->ops->destroy = MatDestroy_MKL_CPARDISO; 889 890 B->ops->view = MatView_MKL_CPARDISO; 891 B->ops->getinfo = MatGetInfo_MKL_CPARDISO; 892 893 B->factortype = ftype; 894 B->assembled = PETSC_TRUE; /* required by -ksp_view */ 895 896 B->spptr = mat_mkl_cpardiso; 897 898 ierr = PetscObjectComposeFunction((PetscObject)B,"MatFactorGetSolverPackage_C",MatFactorGetSolverPackage_mkl_cpardiso);CHKERRQ(ierr); 899 ierr = PetscObjectComposeFunction((PetscObject)B,"MatMkl_CPardisoSetCntl_C",MatMkl_CPardisoSetCntl_MKL_CPARDISO);CHKERRQ(ierr); 900 ierr = PetscInitialize_MKL_CPARDISO(A, mat_mkl_cpardiso);CHKERRQ(ierr); 901 902 *F = B; 903 PetscFunctionReturn(0); 904 } 905 906 #undef __FUNCT__ 907 #define __FUNCT__ "MatSolverPackageRegister_MKL_CPardiso" 908 PETSC_EXTERN PetscErrorCode MatSolverPackageRegister_MKL_CPardiso(void) 909 { 910 PetscErrorCode ierr; 911 912 PetscFunctionBegin; 913 ierr = MatSolverPackageRegister(MATSOLVERMKL_CPARDISO,MATMPIAIJ,MAT_FACTOR_LU,MatGetFactor_mpiaij_mkl_cpardiso);CHKERRQ(ierr); 914 ierr = MatSolverPackageRegister(MATSOLVERMKL_CPARDISO,MATSEQAIJ,MAT_FACTOR_LU,MatGetFactor_mpiaij_mkl_cpardiso);CHKERRQ(ierr); 915 PetscFunctionReturn(0); 916 } 917