Actual source code: mkl_cpardiso.c
1: #include <petscsys.h>
2: #include <../src/mat/impls/aij/mpi/mpiaij.h>
3: #include <../src/mat/impls/sbaij/mpi/mpisbaij.h>
5: #if defined(PETSC_HAVE_MKL_INTEL_ILP64)
6: #define MKL_ILP64
7: #endif
8: #include <mkl.h>
9: #include <mkl_cluster_sparse_solver.h>
11: /*
12: * Possible mkl_cpardiso phases that controls the execution of the solver.
13: * For more information check mkl_cpardiso manual.
14: */
15: #define JOB_ANALYSIS 11
16: #define JOB_ANALYSIS_NUMERICAL_FACTORIZATION 12
17: #define JOB_ANALYSIS_NUMERICAL_FACTORIZATION_SOLVE_ITERATIVE_REFINEMENT 13
18: #define JOB_NUMERICAL_FACTORIZATION 22
19: #define JOB_NUMERICAL_FACTORIZATION_SOLVE_ITERATIVE_REFINEMENT 23
20: #define JOB_SOLVE_ITERATIVE_REFINEMENT 33
21: #define JOB_SOLVE_FORWARD_SUBSTITUTION 331
22: #define JOB_SOLVE_DIAGONAL_SUBSTITUTION 332
23: #define JOB_SOLVE_BACKWARD_SUBSTITUTION 333
24: #define JOB_RELEASE_OF_LU_MEMORY 0
25: #define JOB_RELEASE_OF_ALL_MEMORY -1
27: #define IPARM_SIZE 64
28: #define INT_TYPE MKL_INT
30: static const char *Err_MSG_CPardiso(int errNo)
31: {
32: switch (errNo) {
33: case -1:
34: return "input inconsistent";
35: break;
36: case -2:
37: return "not enough memory";
38: break;
39: case -3:
40: return "reordering problem";
41: break;
42: case -4:
43: return "zero pivot, numerical factorization or iterative refinement problem";
44: break;
45: case -5:
46: return "unclassified (internal) error";
47: break;
48: case -6:
49: return "preordering failed (matrix types 11, 13 only)";
50: break;
51: case -7:
52: return "diagonal matrix problem";
53: break;
54: case -8:
55: return "32-bit integer overflow problem";
56: break;
57: case -9:
58: return "not enough memory for OOC";
59: break;
60: case -10:
61: return "problems with opening OOC temporary files";
62: break;
63: case -11:
64: return "read/write problems with the OOC data file";
65: break;
66: default:
67: return "unknown error";
68: }
69: }
71: /*
72: * Internal data structure.
73: * For more information check mkl_cpardiso manual.
74: */
76: typedef struct {
77: /* Configuration vector */
78: INT_TYPE iparm[IPARM_SIZE];
80: /*
81: * Internal mkl_cpardiso memory location.
82: * After the first call to mkl_cpardiso do not modify pt, as that could cause a serious memory leak.
83: */
84: void *pt[IPARM_SIZE];
86: MPI_Fint comm_mkl_cpardiso;
88: /* Basic mkl_cpardiso info*/
89: INT_TYPE phase, maxfct, mnum, mtype, n, nrhs, msglvl, err;
91: /* Matrix structure */
92: PetscScalar *a;
94: INT_TYPE *ia, *ja;
96: /* Number of non-zero elements */
97: INT_TYPE nz;
99: /* Row permutaton vector*/
100: INT_TYPE *perm;
102: /* Define is matrix preserve sparse structure. */
103: MatStructure matstruc;
105: PetscErrorCode (*ConvertToTriples)(Mat, MatReuse, PetscInt *, PetscInt **, PetscInt **, PetscScalar **);
107: /* True if mkl_cpardiso function have been used. */
108: PetscBool CleanUp;
109: } Mat_MKL_CPARDISO;
111: /*
112: * Copy the elements of matrix A.
113: * Input:
114: * - Mat A: MATSEQAIJ matrix
115: * - int shift: matrix index.
116: * - 0 for c representation
117: * - 1 for fortran representation
118: * - MatReuse reuse:
119: * - MAT_INITIAL_MATRIX: Create a new aij representation
120: * - MAT_REUSE_MATRIX: Reuse all aij representation and just change values
121: * Output:
122: * - int *nnz: Number of nonzero-elements.
123: * - int **r pointer to i index
124: * - int **c pointer to j elements
125: * - MATRIXTYPE **v: Non-zero elements
126: */
127: static PetscErrorCode MatCopy_seqaij_seqaij_MKL_CPARDISO(Mat A, MatReuse reuse, PetscInt *nnz, PetscInt **r, PetscInt **c, PetscScalar **v)
128: {
129: Mat_SeqAIJ *aa = (Mat_SeqAIJ *)A->data;
131: PetscFunctionBegin;
132: *v = aa->a;
133: if (reuse == MAT_INITIAL_MATRIX) {
134: *r = (INT_TYPE *)aa->i;
135: *c = (INT_TYPE *)aa->j;
136: *nnz = aa->nz;
137: }
138: PetscFunctionReturn(PETSC_SUCCESS);
139: }
141: static PetscErrorCode MatConvertToTriples_mpiaij_mpiaij_MKL_CPARDISO(Mat A, MatReuse reuse, PetscInt *nnz, PetscInt **r, PetscInt **c, PetscScalar **v)
142: {
143: const PetscInt *ai, *aj, *bi, *bj, *garray, m = A->rmap->n, *ajj, *bjj;
144: PetscInt rstart, nz, i, j, countA, countB;
145: PetscInt *row, *col;
146: const PetscScalar *av, *bv;
147: PetscScalar *val;
148: Mat_MPIAIJ *mat = (Mat_MPIAIJ *)A->data;
149: Mat_SeqAIJ *aa = (Mat_SeqAIJ *)mat->A->data;
150: Mat_SeqAIJ *bb = (Mat_SeqAIJ *)mat->B->data;
151: PetscInt colA_start, jB, jcol;
153: PetscFunctionBegin;
154: ai = aa->i;
155: aj = aa->j;
156: bi = bb->i;
157: bj = bb->j;
158: rstart = A->rmap->rstart;
159: av = aa->a;
160: bv = bb->a;
162: garray = mat->garray;
164: if (reuse == MAT_INITIAL_MATRIX) {
165: nz = aa->nz + bb->nz;
166: *nnz = nz;
167: PetscCall(PetscMalloc3(m + 1, &row, nz, &col, nz, &val));
168: *r = row;
169: *c = col;
170: *v = val;
171: } else {
172: row = *r;
173: col = *c;
174: val = *v;
175: }
177: nz = 0;
178: for (i = 0; i < m; i++) {
179: row[i] = nz;
180: countA = ai[i + 1] - ai[i];
181: countB = bi[i + 1] - bi[i];
182: ajj = aj + ai[i]; /* ptr to the beginning of this row */
183: bjj = bj + bi[i];
185: /* B part, smaller col index */
186: colA_start = rstart + ajj[0]; /* the smallest global col index of A */
187: jB = 0;
188: for (j = 0; j < countB; j++) {
189: jcol = garray[bjj[j]];
190: if (jcol > colA_start) break;
191: col[nz] = jcol;
192: val[nz++] = *bv++;
193: }
194: jB = j;
196: /* A part */
197: for (j = 0; j < countA; j++) {
198: col[nz] = rstart + ajj[j];
199: val[nz++] = *av++;
200: }
202: /* B part, larger col index */
203: for (j = jB; j < countB; j++) {
204: col[nz] = garray[bjj[j]];
205: val[nz++] = *bv++;
206: }
207: }
208: row[m] = nz;
209: PetscFunctionReturn(PETSC_SUCCESS);
210: }
212: static PetscErrorCode MatConvertToTriples_mpibaij_mpibaij_MKL_CPARDISO(Mat A, MatReuse reuse, PetscInt *nnz, PetscInt **r, PetscInt **c, PetscScalar **v)
213: {
214: const PetscInt *ai, *aj, *bi, *bj, *garray, bs = A->rmap->bs, bs2 = bs * bs, m = A->rmap->n / bs, *ajj, *bjj;
215: PetscInt rstart, nz, i, j, countA, countB;
216: PetscInt *row, *col;
217: const PetscScalar *av, *bv;
218: PetscScalar *val;
219: Mat_MPIBAIJ *mat = (Mat_MPIBAIJ *)A->data;
220: Mat_SeqBAIJ *aa = (Mat_SeqBAIJ *)mat->A->data;
221: Mat_SeqBAIJ *bb = (Mat_SeqBAIJ *)mat->B->data;
222: PetscInt colA_start, jB, jcol;
224: PetscFunctionBegin;
225: ai = aa->i;
226: aj = aa->j;
227: bi = bb->i;
228: bj = bb->j;
229: rstart = A->rmap->rstart / bs;
230: av = aa->a;
231: bv = bb->a;
233: garray = mat->garray;
235: if (reuse == MAT_INITIAL_MATRIX) {
236: nz = aa->nz + bb->nz;
237: *nnz = nz;
238: PetscCall(PetscMalloc3(m + 1, &row, nz, &col, nz * bs2, &val));
239: *r = row;
240: *c = col;
241: *v = val;
242: } else {
243: row = *r;
244: col = *c;
245: val = *v;
246: }
248: nz = 0;
249: for (i = 0; i < m; i++) {
250: row[i] = nz + 1;
251: countA = ai[i + 1] - ai[i];
252: countB = bi[i + 1] - bi[i];
253: ajj = aj + ai[i]; /* ptr to the beginning of this row */
254: bjj = bj + bi[i];
256: /* B part, smaller col index */
257: colA_start = rstart + (countA > 0 ? ajj[0] : 0); /* the smallest global col index of A */
258: jB = 0;
259: for (j = 0; j < countB; j++) {
260: jcol = garray[bjj[j]];
261: if (jcol > colA_start) break;
262: col[nz++] = jcol + 1;
263: }
264: jB = j;
265: PetscCall(PetscArraycpy(val, bv, jB * bs2));
266: val += jB * bs2;
267: bv += jB * bs2;
269: /* A part */
270: for (j = 0; j < countA; j++) col[nz++] = rstart + ajj[j] + 1;
271: PetscCall(PetscArraycpy(val, av, countA * bs2));
272: val += countA * bs2;
273: av += countA * bs2;
275: /* B part, larger col index */
276: for (j = jB; j < countB; j++) col[nz++] = garray[bjj[j]] + 1;
277: PetscCall(PetscArraycpy(val, bv, (countB - jB) * bs2));
278: val += (countB - jB) * bs2;
279: bv += (countB - jB) * bs2;
280: }
281: row[m] = nz + 1;
282: PetscFunctionReturn(PETSC_SUCCESS);
283: }
285: static PetscErrorCode MatConvertToTriples_mpisbaij_mpisbaij_MKL_CPARDISO(Mat A, MatReuse reuse, PetscInt *nnz, PetscInt **r, PetscInt **c, PetscScalar **v)
286: {
287: const PetscInt *ai, *aj, *bi, *bj, *garray, bs = A->rmap->bs, bs2 = bs * bs, m = A->rmap->n / bs, *ajj, *bjj;
288: PetscInt rstart, nz, i, j, countA, countB;
289: PetscInt *row, *col;
290: const PetscScalar *av, *bv;
291: PetscScalar *val;
292: Mat_MPISBAIJ *mat = (Mat_MPISBAIJ *)A->data;
293: Mat_SeqSBAIJ *aa = (Mat_SeqSBAIJ *)mat->A->data;
294: Mat_SeqBAIJ *bb = (Mat_SeqBAIJ *)mat->B->data;
296: PetscFunctionBegin;
297: ai = aa->i;
298: aj = aa->j;
299: bi = bb->i;
300: bj = bb->j;
301: rstart = A->rmap->rstart / bs;
302: av = aa->a;
303: bv = bb->a;
305: garray = mat->garray;
307: if (reuse == MAT_INITIAL_MATRIX) {
308: nz = aa->nz + bb->nz;
309: *nnz = nz;
310: PetscCall(PetscMalloc3(m + 1, &row, nz, &col, nz * bs2, &val));
311: *r = row;
312: *c = col;
313: *v = val;
314: } else {
315: row = *r;
316: col = *c;
317: val = *v;
318: }
320: nz = 0;
321: for (i = 0; i < m; i++) {
322: row[i] = nz + 1;
323: countA = ai[i + 1] - ai[i];
324: countB = bi[i + 1] - bi[i];
325: ajj = aj + ai[i]; /* ptr to the beginning of this row */
326: bjj = bj + bi[i];
328: /* A part */
329: for (j = 0; j < countA; j++) col[nz++] = rstart + ajj[j] + 1;
330: PetscCall(PetscArraycpy(val, av, countA * bs2));
331: val += countA * bs2;
332: av += countA * bs2;
334: /* B part, larger col index */
335: for (j = 0; j < countB; j++) col[nz++] = garray[bjj[j]] + 1;
336: PetscCall(PetscArraycpy(val, bv, countB * bs2));
337: val += countB * bs2;
338: bv += countB * bs2;
339: }
340: row[m] = nz + 1;
341: PetscFunctionReturn(PETSC_SUCCESS);
342: }
344: /*
345: * Free memory for Mat_MKL_CPARDISO structure and pointers to objects.
346: */
347: static PetscErrorCode MatDestroy_MKL_CPARDISO(Mat A)
348: {
349: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)A->data;
350: MPI_Comm comm;
352: PetscFunctionBegin;
353: /* Terminate instance, deallocate memories */
354: if (mat_mkl_cpardiso->CleanUp) {
355: mat_mkl_cpardiso->phase = JOB_RELEASE_OF_ALL_MEMORY;
357: cluster_sparse_solver(mat_mkl_cpardiso->pt, &mat_mkl_cpardiso->maxfct, &mat_mkl_cpardiso->mnum, &mat_mkl_cpardiso->mtype, &mat_mkl_cpardiso->phase, &mat_mkl_cpardiso->n, NULL, NULL, NULL, mat_mkl_cpardiso->perm, &mat_mkl_cpardiso->nrhs,
358: mat_mkl_cpardiso->iparm, &mat_mkl_cpardiso->msglvl, NULL, NULL, &mat_mkl_cpardiso->comm_mkl_cpardiso, (PetscInt *)&mat_mkl_cpardiso->err);
359: }
361: if (mat_mkl_cpardiso->ConvertToTriples != MatCopy_seqaij_seqaij_MKL_CPARDISO) PetscCall(PetscFree3(mat_mkl_cpardiso->ia, mat_mkl_cpardiso->ja, mat_mkl_cpardiso->a));
362: comm = MPI_Comm_f2c(mat_mkl_cpardiso->comm_mkl_cpardiso);
363: PetscCallMPI(MPI_Comm_free(&comm));
364: PetscCall(PetscFree(A->data));
366: /* clear composed functions */
367: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatFactorGetSolverType_C", NULL));
368: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatMkl_CPardisoSetCntl_C", NULL));
369: PetscFunctionReturn(PETSC_SUCCESS);
370: }
372: /*
373: * Computes Ax = b
374: */
375: static PetscErrorCode MatSolve_MKL_CPARDISO(Mat A, Vec b, Vec x)
376: {
377: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)A->data;
378: PetscScalar *xarray;
379: const PetscScalar *barray;
381: PetscFunctionBegin;
382: mat_mkl_cpardiso->nrhs = 1;
383: PetscCall(VecGetArray(x, &xarray));
384: PetscCall(VecGetArrayRead(b, &barray));
386: /* solve phase */
387: mat_mkl_cpardiso->phase = JOB_SOLVE_ITERATIVE_REFINEMENT;
388: cluster_sparse_solver(mat_mkl_cpardiso->pt, &mat_mkl_cpardiso->maxfct, &mat_mkl_cpardiso->mnum, &mat_mkl_cpardiso->mtype, &mat_mkl_cpardiso->phase, &mat_mkl_cpardiso->n, mat_mkl_cpardiso->a, mat_mkl_cpardiso->ia, mat_mkl_cpardiso->ja,
389: mat_mkl_cpardiso->perm, &mat_mkl_cpardiso->nrhs, mat_mkl_cpardiso->iparm, &mat_mkl_cpardiso->msglvl, (void *)barray, (void *)xarray, &mat_mkl_cpardiso->comm_mkl_cpardiso, (PetscInt *)&mat_mkl_cpardiso->err);
391: PetscCheck(mat_mkl_cpardiso->err >= 0, PETSC_COMM_SELF, PETSC_ERR_LIB, "Error reported by MKL Cluster PARDISO: err=%d, msg = \"%s\". Please check manual", mat_mkl_cpardiso->err, Err_MSG_CPardiso(mat_mkl_cpardiso->err));
393: PetscCall(VecRestoreArray(x, &xarray));
394: PetscCall(VecRestoreArrayRead(b, &barray));
395: mat_mkl_cpardiso->CleanUp = PETSC_TRUE;
396: PetscFunctionReturn(PETSC_SUCCESS);
397: }
399: static PetscErrorCode MatForwardSolve_MKL_CPARDISO(Mat A, Vec b, Vec x)
400: {
401: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)A->data;
402: PetscScalar *xarray;
403: const PetscScalar *barray;
405: PetscFunctionBegin;
406: mat_mkl_cpardiso->nrhs = 1;
407: PetscCall(VecGetArray(x, &xarray));
408: PetscCall(VecGetArrayRead(b, &barray));
410: /* solve phase */
411: mat_mkl_cpardiso->phase = JOB_SOLVE_FORWARD_SUBSTITUTION;
412: cluster_sparse_solver(mat_mkl_cpardiso->pt, &mat_mkl_cpardiso->maxfct, &mat_mkl_cpardiso->mnum, &mat_mkl_cpardiso->mtype, &mat_mkl_cpardiso->phase, &mat_mkl_cpardiso->n, mat_mkl_cpardiso->a, mat_mkl_cpardiso->ia, mat_mkl_cpardiso->ja,
413: mat_mkl_cpardiso->perm, &mat_mkl_cpardiso->nrhs, mat_mkl_cpardiso->iparm, &mat_mkl_cpardiso->msglvl, (void *)barray, (void *)xarray, &mat_mkl_cpardiso->comm_mkl_cpardiso, (PetscInt *)&mat_mkl_cpardiso->err);
415: PetscCheck(mat_mkl_cpardiso->err >= 0, PETSC_COMM_SELF, PETSC_ERR_LIB, "Error reported by MKL Cluster PARDISO: err=%d, msg = \"%s\". Please check manual", mat_mkl_cpardiso->err, Err_MSG_CPardiso(mat_mkl_cpardiso->err));
417: PetscCall(VecRestoreArray(x, &xarray));
418: PetscCall(VecRestoreArrayRead(b, &barray));
419: mat_mkl_cpardiso->CleanUp = PETSC_TRUE;
420: PetscFunctionReturn(PETSC_SUCCESS);
421: }
423: static PetscErrorCode MatBackwardSolve_MKL_CPARDISO(Mat A, Vec b, Vec x)
424: {
425: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)A->data;
426: PetscScalar *xarray;
427: const PetscScalar *barray;
429: PetscFunctionBegin;
430: mat_mkl_cpardiso->nrhs = 1;
431: PetscCall(VecGetArray(x, &xarray));
432: PetscCall(VecGetArrayRead(b, &barray));
434: /* solve phase */
435: mat_mkl_cpardiso->phase = JOB_SOLVE_BACKWARD_SUBSTITUTION;
436: cluster_sparse_solver(mat_mkl_cpardiso->pt, &mat_mkl_cpardiso->maxfct, &mat_mkl_cpardiso->mnum, &mat_mkl_cpardiso->mtype, &mat_mkl_cpardiso->phase, &mat_mkl_cpardiso->n, mat_mkl_cpardiso->a, mat_mkl_cpardiso->ia, mat_mkl_cpardiso->ja,
437: mat_mkl_cpardiso->perm, &mat_mkl_cpardiso->nrhs, mat_mkl_cpardiso->iparm, &mat_mkl_cpardiso->msglvl, (void *)barray, (void *)xarray, &mat_mkl_cpardiso->comm_mkl_cpardiso, (PetscInt *)&mat_mkl_cpardiso->err);
439: PetscCheck(mat_mkl_cpardiso->err >= 0, PETSC_COMM_SELF, PETSC_ERR_LIB, "Error reported by MKL Cluster PARDISO: err=%d, msg = \"%s\". Please check manual", mat_mkl_cpardiso->err, Err_MSG_CPardiso(mat_mkl_cpardiso->err));
441: PetscCall(VecRestoreArray(x, &xarray));
442: PetscCall(VecRestoreArrayRead(b, &barray));
443: mat_mkl_cpardiso->CleanUp = PETSC_TRUE;
444: PetscFunctionReturn(PETSC_SUCCESS);
445: }
447: static PetscErrorCode MatSolveTranspose_MKL_CPARDISO(Mat A, Vec b, Vec x)
448: {
449: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)A->data;
451: PetscFunctionBegin;
452: #if defined(PETSC_USE_COMPLEX)
453: mat_mkl_cpardiso->iparm[12 - 1] = 1;
454: #else
455: mat_mkl_cpardiso->iparm[12 - 1] = 2;
456: #endif
457: PetscCall(MatSolve_MKL_CPARDISO(A, b, x));
458: mat_mkl_cpardiso->iparm[12 - 1] = 0;
459: PetscFunctionReturn(PETSC_SUCCESS);
460: }
462: static PetscErrorCode MatMatSolve_MKL_CPARDISO(Mat A, Mat B, Mat X)
463: {
464: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)A->data;
465: PetscScalar *xarray;
466: const PetscScalar *barray;
468: PetscFunctionBegin;
469: PetscCall(MatGetSize(B, NULL, (PetscInt *)&mat_mkl_cpardiso->nrhs));
471: if (mat_mkl_cpardiso->nrhs > 0) {
472: PetscCall(MatDenseGetArrayRead(B, &barray));
473: PetscCall(MatDenseGetArray(X, &xarray));
475: PetscCheck(barray != xarray, PETSC_COMM_SELF, PETSC_ERR_SUP, "B and X cannot share the same memory location");
477: /* solve phase */
478: mat_mkl_cpardiso->phase = JOB_SOLVE_ITERATIVE_REFINEMENT;
479: cluster_sparse_solver(mat_mkl_cpardiso->pt, &mat_mkl_cpardiso->maxfct, &mat_mkl_cpardiso->mnum, &mat_mkl_cpardiso->mtype, &mat_mkl_cpardiso->phase, &mat_mkl_cpardiso->n, mat_mkl_cpardiso->a, mat_mkl_cpardiso->ia, mat_mkl_cpardiso->ja,
480: mat_mkl_cpardiso->perm, &mat_mkl_cpardiso->nrhs, mat_mkl_cpardiso->iparm, &mat_mkl_cpardiso->msglvl, (void *)barray, (void *)xarray, &mat_mkl_cpardiso->comm_mkl_cpardiso, (PetscInt *)&mat_mkl_cpardiso->err);
481: PetscCheck(mat_mkl_cpardiso->err >= 0, PETSC_COMM_SELF, PETSC_ERR_LIB, "Error reported by MKL Cluster PARDISO: err=%d, msg = \"%s\". Please check manual", mat_mkl_cpardiso->err, Err_MSG_CPardiso(mat_mkl_cpardiso->err));
482: PetscCall(MatDenseRestoreArrayRead(B, &barray));
483: PetscCall(MatDenseRestoreArray(X, &xarray));
484: }
485: mat_mkl_cpardiso->CleanUp = PETSC_TRUE;
486: PetscFunctionReturn(PETSC_SUCCESS);
487: }
489: /*
490: * LU Decomposition
491: */
492: static PetscErrorCode MatFactorNumeric_MKL_CPARDISO(Mat F, Mat A, const MatFactorInfo *info)
493: {
494: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)F->data;
496: PetscFunctionBegin;
497: mat_mkl_cpardiso->matstruc = SAME_NONZERO_PATTERN;
498: PetscCall((*mat_mkl_cpardiso->ConvertToTriples)(A, MAT_REUSE_MATRIX, &mat_mkl_cpardiso->nz, &mat_mkl_cpardiso->ia, &mat_mkl_cpardiso->ja, &mat_mkl_cpardiso->a));
500: mat_mkl_cpardiso->phase = JOB_NUMERICAL_FACTORIZATION;
501: cluster_sparse_solver(mat_mkl_cpardiso->pt, &mat_mkl_cpardiso->maxfct, &mat_mkl_cpardiso->mnum, &mat_mkl_cpardiso->mtype, &mat_mkl_cpardiso->phase, &mat_mkl_cpardiso->n, mat_mkl_cpardiso->a, mat_mkl_cpardiso->ia, mat_mkl_cpardiso->ja,
502: mat_mkl_cpardiso->perm, &mat_mkl_cpardiso->nrhs, mat_mkl_cpardiso->iparm, &mat_mkl_cpardiso->msglvl, NULL, NULL, &mat_mkl_cpardiso->comm_mkl_cpardiso, &mat_mkl_cpardiso->err);
503: PetscCheck(mat_mkl_cpardiso->err >= 0, PETSC_COMM_SELF, PETSC_ERR_LIB, "Error reported by MKL Cluster PARDISO: err=%d, msg = \"%s\". Please check manual", mat_mkl_cpardiso->err, Err_MSG_CPardiso(mat_mkl_cpardiso->err));
505: mat_mkl_cpardiso->matstruc = SAME_NONZERO_PATTERN;
506: mat_mkl_cpardiso->CleanUp = PETSC_TRUE;
507: PetscFunctionReturn(PETSC_SUCCESS);
508: }
510: /* Sets mkl_cpardiso options from the options database */
511: static PetscErrorCode MatSetFromOptions_MKL_CPARDISO(Mat F, Mat A)
512: {
513: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)F->data;
514: PetscInt icntl, threads;
515: PetscBool flg;
517: PetscFunctionBegin;
518: PetscOptionsBegin(PetscObjectComm((PetscObject)F), ((PetscObject)F)->prefix, "MKL Cluster PARDISO Options", "Mat");
519: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_65", "Suggested number of threads to use within MKL Cluster PARDISO", "None", threads, &threads, &flg));
520: if (flg) mkl_set_num_threads((int)threads);
522: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_66", "Maximum number of factors with identical sparsity structure that must be kept in memory at the same time", "None", mat_mkl_cpardiso->maxfct, &icntl, &flg));
523: if (flg) mat_mkl_cpardiso->maxfct = icntl;
525: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_67", "Indicates the actual matrix for the solution phase", "None", mat_mkl_cpardiso->mnum, &icntl, &flg));
526: if (flg) mat_mkl_cpardiso->mnum = icntl;
528: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_68", "Message level information", "None", mat_mkl_cpardiso->msglvl, &icntl, &flg));
529: if (flg) mat_mkl_cpardiso->msglvl = icntl;
531: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_69", "Defines the matrix type", "None", mat_mkl_cpardiso->mtype, &icntl, &flg));
532: if (flg) mat_mkl_cpardiso->mtype = icntl;
533: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_1", "Use default values", "None", mat_mkl_cpardiso->iparm[0], &icntl, &flg));
535: if (flg && icntl != 0) {
536: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_2", "Fill-in reducing ordering for the input matrix", "None", mat_mkl_cpardiso->iparm[1], &icntl, &flg));
537: if (flg) mat_mkl_cpardiso->iparm[1] = icntl;
539: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_4", "Preconditioned CGS/CG", "None", mat_mkl_cpardiso->iparm[3], &icntl, &flg));
540: if (flg) mat_mkl_cpardiso->iparm[3] = icntl;
542: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_5", "User permutation", "None", mat_mkl_cpardiso->iparm[4], &icntl, &flg));
543: if (flg) mat_mkl_cpardiso->iparm[4] = icntl;
545: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_6", "Write solution on x", "None", mat_mkl_cpardiso->iparm[5], &icntl, &flg));
546: if (flg) mat_mkl_cpardiso->iparm[5] = icntl;
548: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_8", "Iterative refinement step", "None", mat_mkl_cpardiso->iparm[7], &icntl, &flg));
549: if (flg) mat_mkl_cpardiso->iparm[7] = icntl;
551: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_10", "Pivoting perturbation", "None", mat_mkl_cpardiso->iparm[9], &icntl, &flg));
552: if (flg) mat_mkl_cpardiso->iparm[9] = icntl;
554: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_11", "Scaling vectors", "None", mat_mkl_cpardiso->iparm[10], &icntl, &flg));
555: if (flg) mat_mkl_cpardiso->iparm[10] = icntl;
557: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_12", "Solve with transposed or conjugate transposed matrix A", "None", mat_mkl_cpardiso->iparm[11], &icntl, &flg));
558: if (flg) mat_mkl_cpardiso->iparm[11] = icntl;
560: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_13", "Improved accuracy using (non-) symmetric weighted matching", "None", mat_mkl_cpardiso->iparm[12], &icntl, &flg));
561: if (flg) mat_mkl_cpardiso->iparm[12] = icntl;
563: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_18", "Numbers of non-zero elements", "None", mat_mkl_cpardiso->iparm[17], &icntl, &flg));
564: if (flg) mat_mkl_cpardiso->iparm[17] = icntl;
566: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_19", "Report number of floating point operations", "None", mat_mkl_cpardiso->iparm[18], &icntl, &flg));
567: if (flg) mat_mkl_cpardiso->iparm[18] = icntl;
569: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_21", "Pivoting for symmetric indefinite matrices", "None", mat_mkl_cpardiso->iparm[20], &icntl, &flg));
570: if (flg) mat_mkl_cpardiso->iparm[20] = icntl;
572: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_24", "Parallel factorization control", "None", mat_mkl_cpardiso->iparm[23], &icntl, &flg));
573: if (flg) mat_mkl_cpardiso->iparm[23] = icntl;
575: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_25", "Parallel forward/backward solve control", "None", mat_mkl_cpardiso->iparm[24], &icntl, &flg));
576: if (flg) mat_mkl_cpardiso->iparm[24] = icntl;
578: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_27", "Matrix checker", "None", mat_mkl_cpardiso->iparm[26], &icntl, &flg));
579: if (flg) mat_mkl_cpardiso->iparm[26] = icntl;
581: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_31", "Partial solve and computing selected components of the solution vectors", "None", mat_mkl_cpardiso->iparm[30], &icntl, &flg));
582: if (flg) mat_mkl_cpardiso->iparm[30] = icntl;
584: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_34", "Optimal number of threads for conditional numerical reproducibility (CNR) mode", "None", mat_mkl_cpardiso->iparm[33], &icntl, &flg));
585: if (flg) mat_mkl_cpardiso->iparm[33] = icntl;
587: PetscCall(PetscOptionsInt("-mat_mkl_cpardiso_60", "Intel MKL Cluster PARDISO mode", "None", mat_mkl_cpardiso->iparm[59], &icntl, &flg));
588: if (flg) mat_mkl_cpardiso->iparm[59] = icntl;
589: }
591: PetscOptionsEnd();
592: PetscFunctionReturn(PETSC_SUCCESS);
593: }
595: static PetscErrorCode PetscInitialize_MKL_CPARDISO(Mat A, Mat_MKL_CPARDISO *mat_mkl_cpardiso)
596: {
597: PetscInt bs;
598: PetscBool match;
599: PetscMPIInt size;
600: MPI_Comm comm;
602: PetscFunctionBegin;
603: PetscCallMPI(MPI_Comm_dup(PetscObjectComm((PetscObject)A), &comm));
604: PetscCallMPI(MPI_Comm_size(comm, &size));
605: mat_mkl_cpardiso->comm_mkl_cpardiso = MPI_Comm_c2f(comm);
607: mat_mkl_cpardiso->CleanUp = PETSC_FALSE;
608: mat_mkl_cpardiso->maxfct = 1;
609: mat_mkl_cpardiso->mnum = 1;
610: mat_mkl_cpardiso->n = A->rmap->N;
611: if (mat_mkl_cpardiso->iparm[36]) mat_mkl_cpardiso->n /= mat_mkl_cpardiso->iparm[36];
612: mat_mkl_cpardiso->msglvl = 0;
613: mat_mkl_cpardiso->nrhs = 1;
614: mat_mkl_cpardiso->err = 0;
615: mat_mkl_cpardiso->phase = -1;
616: #if defined(PETSC_USE_COMPLEX)
617: mat_mkl_cpardiso->mtype = 13;
618: #else
619: mat_mkl_cpardiso->mtype = 11;
620: #endif
622: #if defined(PETSC_USE_REAL_SINGLE)
623: mat_mkl_cpardiso->iparm[27] = 1;
624: #else
625: mat_mkl_cpardiso->iparm[27] = 0;
626: #endif
628: mat_mkl_cpardiso->iparm[0] = 1; /* Solver default parameters overridden with provided by iparm */
629: mat_mkl_cpardiso->iparm[1] = 2; /* Use METIS for fill-in reordering */
630: mat_mkl_cpardiso->iparm[5] = 0; /* Write solution into x */
631: mat_mkl_cpardiso->iparm[7] = 2; /* Max number of iterative refinement steps */
632: mat_mkl_cpardiso->iparm[9] = 13; /* Perturb the pivot elements with 1E-13 */
633: mat_mkl_cpardiso->iparm[10] = 1; /* Use nonsymmetric permutation and scaling MPS */
634: mat_mkl_cpardiso->iparm[12] = 1; /* Switch on Maximum Weighted Matching algorithm (default for non-symmetric) */
635: mat_mkl_cpardiso->iparm[17] = -1; /* Output: Number of nonzeros in the factor LU */
636: mat_mkl_cpardiso->iparm[18] = -1; /* Output: Mflops for LU factorization */
637: mat_mkl_cpardiso->iparm[26] = 1; /* Check input data for correctness */
639: mat_mkl_cpardiso->iparm[39] = 0;
640: if (size > 1) {
641: mat_mkl_cpardiso->iparm[39] = 2;
642: mat_mkl_cpardiso->iparm[40] = A->rmap->rstart;
643: mat_mkl_cpardiso->iparm[41] = A->rmap->rend - 1;
644: }
645: PetscCall(PetscObjectTypeCompareAny((PetscObject)A, &match, MATMPIBAIJ, MATMPISBAIJ, ""));
646: if (match) {
647: PetscCall(MatGetBlockSize(A, &bs));
648: mat_mkl_cpardiso->iparm[36] = bs;
649: mat_mkl_cpardiso->iparm[40] /= bs;
650: mat_mkl_cpardiso->iparm[41] /= bs;
651: mat_mkl_cpardiso->iparm[40]++;
652: mat_mkl_cpardiso->iparm[41]++;
653: mat_mkl_cpardiso->iparm[34] = 0; /* Fortran style */
654: } else {
655: mat_mkl_cpardiso->iparm[34] = 1; /* C style */
656: }
658: mat_mkl_cpardiso->perm = 0;
659: PetscFunctionReturn(PETSC_SUCCESS);
660: }
662: /*
663: * Symbolic decomposition. Mkl_Pardiso analysis phase.
664: */
665: static PetscErrorCode MatLUFactorSymbolic_AIJMKL_CPARDISO(Mat F, Mat A, IS r, IS c, const MatFactorInfo *info)
666: {
667: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)F->data;
669: PetscFunctionBegin;
670: mat_mkl_cpardiso->matstruc = DIFFERENT_NONZERO_PATTERN;
672: /* Set MKL_CPARDISO options from the options database */
673: PetscCall(MatSetFromOptions_MKL_CPARDISO(F, A));
674: PetscCall((*mat_mkl_cpardiso->ConvertToTriples)(A, MAT_INITIAL_MATRIX, &mat_mkl_cpardiso->nz, &mat_mkl_cpardiso->ia, &mat_mkl_cpardiso->ja, &mat_mkl_cpardiso->a));
676: mat_mkl_cpardiso->n = A->rmap->N;
677: if (mat_mkl_cpardiso->iparm[36]) mat_mkl_cpardiso->n /= mat_mkl_cpardiso->iparm[36];
679: /* analysis phase */
680: mat_mkl_cpardiso->phase = JOB_ANALYSIS;
682: cluster_sparse_solver(mat_mkl_cpardiso->pt, &mat_mkl_cpardiso->maxfct, &mat_mkl_cpardiso->mnum, &mat_mkl_cpardiso->mtype, &mat_mkl_cpardiso->phase, &mat_mkl_cpardiso->n, mat_mkl_cpardiso->a, mat_mkl_cpardiso->ia, mat_mkl_cpardiso->ja,
683: mat_mkl_cpardiso->perm, &mat_mkl_cpardiso->nrhs, mat_mkl_cpardiso->iparm, &mat_mkl_cpardiso->msglvl, NULL, NULL, &mat_mkl_cpardiso->comm_mkl_cpardiso, (PetscInt *)&mat_mkl_cpardiso->err);
685: PetscCheck(mat_mkl_cpardiso->err >= 0, PETSC_COMM_SELF, PETSC_ERR_LIB, "Error reported by MKL Cluster PARDISO: err=%d, msg = \"%s\".Check manual", mat_mkl_cpardiso->err, Err_MSG_CPardiso(mat_mkl_cpardiso->err));
687: mat_mkl_cpardiso->CleanUp = PETSC_TRUE;
688: F->ops->lufactornumeric = MatFactorNumeric_MKL_CPARDISO;
689: F->ops->solve = MatSolve_MKL_CPARDISO;
690: F->ops->forwardsolve = MatForwardSolve_MKL_CPARDISO;
691: F->ops->backwardsolve = MatBackwardSolve_MKL_CPARDISO;
692: F->ops->solvetranspose = MatSolveTranspose_MKL_CPARDISO;
693: F->ops->matsolve = MatMatSolve_MKL_CPARDISO;
694: PetscFunctionReturn(PETSC_SUCCESS);
695: }
697: static PetscErrorCode MatCholeskyFactorSymbolic_AIJMKL_CPARDISO(Mat F, Mat A, IS perm, const MatFactorInfo *info)
698: {
699: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)F->data;
701: PetscFunctionBegin;
702: mat_mkl_cpardiso->matstruc = DIFFERENT_NONZERO_PATTERN;
704: /* Set MKL_CPARDISO options from the options database */
705: PetscCall(MatSetFromOptions_MKL_CPARDISO(F, A));
706: PetscCall((*mat_mkl_cpardiso->ConvertToTriples)(A, MAT_INITIAL_MATRIX, &mat_mkl_cpardiso->nz, &mat_mkl_cpardiso->ia, &mat_mkl_cpardiso->ja, &mat_mkl_cpardiso->a));
708: mat_mkl_cpardiso->n = A->rmap->N;
709: if (mat_mkl_cpardiso->iparm[36]) mat_mkl_cpardiso->n /= mat_mkl_cpardiso->iparm[36];
710: PetscCheck(!PetscDefined(USE_COMPLEX), PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "No support for PARDISO CHOLESKY with complex scalars! Use MAT_FACTOR_LU instead");
711: if (A->spd == PETSC_BOOL3_TRUE) mat_mkl_cpardiso->mtype = 2;
712: else mat_mkl_cpardiso->mtype = -2;
714: /* analysis phase */
715: mat_mkl_cpardiso->phase = JOB_ANALYSIS;
717: cluster_sparse_solver(mat_mkl_cpardiso->pt, &mat_mkl_cpardiso->maxfct, &mat_mkl_cpardiso->mnum, &mat_mkl_cpardiso->mtype, &mat_mkl_cpardiso->phase, &mat_mkl_cpardiso->n, mat_mkl_cpardiso->a, mat_mkl_cpardiso->ia, mat_mkl_cpardiso->ja,
718: mat_mkl_cpardiso->perm, &mat_mkl_cpardiso->nrhs, mat_mkl_cpardiso->iparm, &mat_mkl_cpardiso->msglvl, NULL, NULL, &mat_mkl_cpardiso->comm_mkl_cpardiso, (PetscInt *)&mat_mkl_cpardiso->err);
720: PetscCheck(mat_mkl_cpardiso->err >= 0, PETSC_COMM_SELF, PETSC_ERR_LIB, "Error reported by MKL Cluster PARDISO: err=%d, msg = \"%s\".Check manual", mat_mkl_cpardiso->err, Err_MSG_CPardiso(mat_mkl_cpardiso->err));
722: mat_mkl_cpardiso->CleanUp = PETSC_TRUE;
723: F->ops->choleskyfactornumeric = MatFactorNumeric_MKL_CPARDISO;
724: F->ops->solve = MatSolve_MKL_CPARDISO;
725: F->ops->solvetranspose = MatSolveTranspose_MKL_CPARDISO;
726: F->ops->matsolve = MatMatSolve_MKL_CPARDISO;
727: if (A->spd == PETSC_BOOL3_TRUE) {
728: F->ops->forwardsolve = MatForwardSolve_MKL_CPARDISO;
729: F->ops->backwardsolve = MatBackwardSolve_MKL_CPARDISO;
730: }
731: PetscFunctionReturn(PETSC_SUCCESS);
732: }
734: static PetscErrorCode MatView_MKL_CPARDISO(Mat A, PetscViewer viewer)
735: {
736: PetscBool iascii;
737: PetscViewerFormat format;
738: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)A->data;
739: PetscInt i;
741: PetscFunctionBegin;
742: /* check if matrix is mkl_cpardiso type */
743: if (A->ops->solve != MatSolve_MKL_CPARDISO) PetscFunctionReturn(PETSC_SUCCESS);
745: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &iascii));
746: if (iascii) {
747: PetscCall(PetscViewerGetFormat(viewer, &format));
748: if (format == PETSC_VIEWER_ASCII_INFO) {
749: PetscCall(PetscViewerASCIIPrintf(viewer, "MKL Cluster PARDISO run parameters:\n"));
750: PetscCall(PetscViewerASCIIPrintf(viewer, "MKL Cluster PARDISO phase: %d \n", mat_mkl_cpardiso->phase));
751: for (i = 1; i <= 64; i++) PetscCall(PetscViewerASCIIPrintf(viewer, "MKL Cluster PARDISO iparm[%d]: %d \n", i, mat_mkl_cpardiso->iparm[i - 1]));
752: PetscCall(PetscViewerASCIIPrintf(viewer, "MKL Cluster PARDISO maxfct: %d \n", mat_mkl_cpardiso->maxfct));
753: PetscCall(PetscViewerASCIIPrintf(viewer, "MKL Cluster PARDISO mnum: %d \n", mat_mkl_cpardiso->mnum));
754: PetscCall(PetscViewerASCIIPrintf(viewer, "MKL Cluster PARDISO mtype: %d \n", mat_mkl_cpardiso->mtype));
755: PetscCall(PetscViewerASCIIPrintf(viewer, "MKL Cluster PARDISO n: %d \n", mat_mkl_cpardiso->n));
756: PetscCall(PetscViewerASCIIPrintf(viewer, "MKL Cluster PARDISO nrhs: %d \n", mat_mkl_cpardiso->nrhs));
757: PetscCall(PetscViewerASCIIPrintf(viewer, "MKL Cluster PARDISO msglvl: %d \n", mat_mkl_cpardiso->msglvl));
758: }
759: }
760: PetscFunctionReturn(PETSC_SUCCESS);
761: }
763: static PetscErrorCode MatGetInfo_MKL_CPARDISO(Mat A, MatInfoType flag, MatInfo *info)
764: {
765: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)A->data;
767: PetscFunctionBegin;
768: info->block_size = 1.0;
769: info->nz_allocated = mat_mkl_cpardiso->nz + 0.0;
770: info->nz_unneeded = 0.0;
771: info->assemblies = 0.0;
772: info->mallocs = 0.0;
773: info->memory = 0.0;
774: info->fill_ratio_given = 0;
775: info->fill_ratio_needed = 0;
776: info->factor_mallocs = 0;
777: PetscFunctionReturn(PETSC_SUCCESS);
778: }
780: static PetscErrorCode MatMkl_CPardisoSetCntl_MKL_CPARDISO(Mat F, PetscInt icntl, PetscInt ival)
781: {
782: Mat_MKL_CPARDISO *mat_mkl_cpardiso = (Mat_MKL_CPARDISO *)F->data;
784: PetscFunctionBegin;
785: if (icntl <= 64) {
786: mat_mkl_cpardiso->iparm[icntl - 1] = ival;
787: } else {
788: if (icntl == 65) mkl_set_num_threads((int)ival);
789: else if (icntl == 66) mat_mkl_cpardiso->maxfct = ival;
790: else if (icntl == 67) mat_mkl_cpardiso->mnum = ival;
791: else if (icntl == 68) mat_mkl_cpardiso->msglvl = ival;
792: else if (icntl == 69) mat_mkl_cpardiso->mtype = ival;
793: }
794: PetscFunctionReturn(PETSC_SUCCESS);
795: }
797: /*@
798: MatMkl_CPardisoSetCntl - Set MKL Cluster PARDISO parameters
799: <https://www.intel.com/content/www/us/en/docs/onemkl/developer-reference-c/2023-2/onemkl-pardiso-parallel-direct-sparse-solver-iface.html>
801: Logically Collective
803: Input Parameters:
804: + F - the factored matrix obtained by calling `MatGetFactor()`
805: . icntl - index of MKL Cluster PARDISO parameter
806: - ival - value of MKL Cluster PARDISO parameter
808: Options Database Key:
809: . -mat_mkl_cpardiso_<icntl> <ival> - set the option numbered icntl to ival
811: Level: intermediate
813: Note:
814: This routine cannot be used if you are solving the linear system with `TS`, `SNES`, or `KSP`, only if you directly call `MatGetFactor()` so use the options
815: database approach when working with `TS`, `SNES`, or `KSP`. See `MATSOLVERMKL_CPARDISO` for the options
817: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MATMPIAIJ`, `MATSOLVERMKL_CPARDISO`
818: @*/
819: PetscErrorCode MatMkl_CPardisoSetCntl(Mat F, PetscInt icntl, PetscInt ival)
820: {
821: PetscFunctionBegin;
822: PetscTryMethod(F, "MatMkl_CPardisoSetCntl_C", (Mat, PetscInt, PetscInt), (F, icntl, ival));
823: PetscFunctionReturn(PETSC_SUCCESS);
824: }
826: /*MC
827: MATSOLVERMKL_CPARDISO - A matrix type providing direct solvers (LU) for parallel matrices via the external package MKL Cluster PARDISO
828: <https://www.intel.com/content/www/us/en/docs/onemkl/developer-reference-c/2023-2/onemkl-pardiso-parallel-direct-sparse-solver-iface.html>
830: Works with `MATMPIAIJ` matrices
832: Use `-pc_type lu` `-pc_factor_mat_solver_type mkl_cpardiso` to use this direct solver
834: Options Database Keys:
835: + -mat_mkl_cpardiso_65 - Suggested number of threads to use within MKL Cluster PARDISO
836: . -mat_mkl_cpardiso_66 - Maximum number of factors with identical sparsity structure that must be kept in memory at the same time
837: . -mat_mkl_cpardiso_67 - Indicates the actual matrix for the solution phase
838: . -mat_mkl_cpardiso_68 - Message level information, use 1 to get detailed information on the solver options
839: . -mat_mkl_cpardiso_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
840: . -mat_mkl_cpardiso_1 - Use default values
841: . -mat_mkl_cpardiso_2 - Fill-in reducing ordering for the input matrix
842: . -mat_mkl_cpardiso_4 - Preconditioned CGS/CG
843: . -mat_mkl_cpardiso_5 - User permutation
844: . -mat_mkl_cpardiso_6 - Write solution on x
845: . -mat_mkl_cpardiso_8 - Iterative refinement step
846: . -mat_mkl_cpardiso_10 - Pivoting perturbation
847: . -mat_mkl_cpardiso_11 - Scaling vectors
848: . -mat_mkl_cpardiso_12 - Solve with transposed or conjugate transposed matrix A
849: . -mat_mkl_cpardiso_13 - Improved accuracy using (non-) symmetric weighted matching
850: . -mat_mkl_cpardiso_18 - Numbers of non-zero elements
851: . -mat_mkl_cpardiso_19 - Report number of floating point operations
852: . -mat_mkl_cpardiso_21 - Pivoting for symmetric indefinite matrices
853: . -mat_mkl_cpardiso_24 - Parallel factorization control
854: . -mat_mkl_cpardiso_25 - Parallel forward/backward solve control
855: . -mat_mkl_cpardiso_27 - Matrix checker
856: . -mat_mkl_cpardiso_31 - Partial solve and computing selected components of the solution vectors
857: . -mat_mkl_cpardiso_34 - Optimal number of threads for conditional numerical reproducibility (CNR) mode
858: - -mat_mkl_cpardiso_60 - Intel MKL Cluster PARDISO mode
860: Level: beginner
862: Notes:
863: Use `-mat_mkl_cpardiso_68 1` to display the number of threads the solver is using. MKL does not provide a way to directly access this
864: information.
866: For more information on the options check
867: <https://www.intel.com/content/www/us/en/docs/onemkl/developer-reference-c/2023-2/onemkl-pardiso-parallel-direct-sparse-solver-iface.html>
869: .seealso: [](ch_matrices), `Mat`, `PCFactorSetMatSolverType()`, `MatSolverType`, `MatMkl_CPardisoSetCntl()`, `MatGetFactor()`, `MATSOLVERMKL_PARDISO`
870: M*/
872: static PetscErrorCode MatFactorGetSolverType_mkl_cpardiso(Mat A, MatSolverType *type)
873: {
874: PetscFunctionBegin;
875: *type = MATSOLVERMKL_CPARDISO;
876: PetscFunctionReturn(PETSC_SUCCESS);
877: }
879: /* MatGetFactor for MPI AIJ matrices */
880: static PetscErrorCode MatGetFactor_mpiaij_mkl_cpardiso(Mat A, MatFactorType ftype, Mat *F)
881: {
882: Mat B;
883: Mat_MKL_CPARDISO *mat_mkl_cpardiso;
884: PetscBool isSeqAIJ, isMPIBAIJ, isMPISBAIJ;
886: PetscFunctionBegin;
887: /* Create the factorization matrix */
889: PetscCall(PetscObjectTypeCompare((PetscObject)A, MATSEQAIJ, &isSeqAIJ));
890: PetscCall(PetscObjectTypeCompare((PetscObject)A, MATMPIBAIJ, &isMPIBAIJ));
891: PetscCall(PetscObjectTypeCompare((PetscObject)A, MATMPISBAIJ, &isMPISBAIJ));
892: PetscCall(MatCreate(PetscObjectComm((PetscObject)A), &B));
893: PetscCall(MatSetSizes(B, A->rmap->n, A->cmap->n, A->rmap->N, A->cmap->N));
894: PetscCall(PetscStrallocpy("mkl_cpardiso", &((PetscObject)B)->type_name));
895: PetscCall(MatSetUp(B));
897: PetscCall(PetscNew(&mat_mkl_cpardiso));
899: if (isSeqAIJ) mat_mkl_cpardiso->ConvertToTriples = MatCopy_seqaij_seqaij_MKL_CPARDISO;
900: else if (isMPIBAIJ) mat_mkl_cpardiso->ConvertToTriples = MatConvertToTriples_mpibaij_mpibaij_MKL_CPARDISO;
901: else if (isMPISBAIJ) mat_mkl_cpardiso->ConvertToTriples = MatConvertToTriples_mpisbaij_mpisbaij_MKL_CPARDISO;
902: else mat_mkl_cpardiso->ConvertToTriples = MatConvertToTriples_mpiaij_mpiaij_MKL_CPARDISO;
904: if (ftype == MAT_FACTOR_LU) B->ops->lufactorsymbolic = MatLUFactorSymbolic_AIJMKL_CPARDISO;
905: else B->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_AIJMKL_CPARDISO;
906: B->ops->destroy = MatDestroy_MKL_CPARDISO;
908: B->ops->view = MatView_MKL_CPARDISO;
909: B->ops->getinfo = MatGetInfo_MKL_CPARDISO;
911: B->factortype = ftype;
912: B->assembled = PETSC_TRUE; /* required by -ksp_view */
914: B->data = mat_mkl_cpardiso;
916: /* set solvertype */
917: PetscCall(PetscFree(B->solvertype));
918: PetscCall(PetscStrallocpy(MATSOLVERMKL_CPARDISO, &B->solvertype));
920: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatFactorGetSolverType_C", MatFactorGetSolverType_mkl_cpardiso));
921: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatMkl_CPardisoSetCntl_C", MatMkl_CPardisoSetCntl_MKL_CPARDISO));
922: PetscCall(PetscInitialize_MKL_CPARDISO(A, mat_mkl_cpardiso));
924: *F = B;
925: PetscFunctionReturn(PETSC_SUCCESS);
926: }
928: PETSC_EXTERN PetscErrorCode MatSolverTypeRegister_MKL_CPardiso(void)
929: {
930: PetscFunctionBegin;
931: PetscCall(MatSolverTypeRegister(MATSOLVERMKL_CPARDISO, MATMPIAIJ, MAT_FACTOR_LU, MatGetFactor_mpiaij_mkl_cpardiso));
932: PetscCall(MatSolverTypeRegister(MATSOLVERMKL_CPARDISO, MATSEQAIJ, MAT_FACTOR_LU, MatGetFactor_mpiaij_mkl_cpardiso));
933: PetscCall(MatSolverTypeRegister(MATSOLVERMKL_CPARDISO, MATMPIBAIJ, MAT_FACTOR_LU, MatGetFactor_mpiaij_mkl_cpardiso));
934: PetscCall(MatSolverTypeRegister(MATSOLVERMKL_CPARDISO, MATMPISBAIJ, MAT_FACTOR_CHOLESKY, MatGetFactor_mpiaij_mkl_cpardiso));
935: PetscFunctionReturn(PETSC_SUCCESS);
936: }