Actual source code: factor.c
1: #include <../src/ksp/pc/impls/factor/factor.h>
2: #include <petsc/private/matimpl.h>
4: /*
5: If an ordering is not yet set and the matrix is available determine a default ordering
6: */
7: PetscErrorCode PCFactorSetDefaultOrdering_Factor(PC pc)
8: {
9: PetscBool foundmtype, flg, destroy = PETSC_FALSE;
10: const char *prefix;
12: PetscFunctionBegin;
13: if (pc->pmat) {
14: PetscCall(PCGetOptionsPrefix(pc, &prefix));
15: PetscCall(MatSetOptionsPrefixFactor(pc->pmat, prefix));
16: PC_Factor *fact = (PC_Factor *)pc->data;
17: PetscCall(MatSolverTypeGet(fact->solvertype, ((PetscObject)pc->pmat)->type_name, fact->factortype, NULL, &foundmtype, NULL));
18: if (foundmtype) {
19: if (!fact->fact) {
20: /* factored matrix is not present at this point, we want to create it during PCSetUp.
21: Since this can be called from setfromoptions, we destroy it when we are done with it */
22: PetscCall(MatGetFactor(pc->pmat, fact->solvertype, fact->factortype, &fact->fact));
23: destroy = PETSC_TRUE;
24: }
25: if (!fact->fact) PetscFunctionReturn(PETSC_SUCCESS);
26: if (!fact->fact->assembled) {
27: PetscCall(PetscStrcmp(fact->solvertype, fact->fact->solvertype, &flg));
28: if (!flg) {
29: Mat B;
30: PetscCall(MatGetFactor(pc->pmat, fact->solvertype, fact->factortype, &B));
31: PetscCall(MatHeaderReplace(fact->fact, &B));
32: }
33: }
34: if (!fact->ordering) {
35: PetscBool canuseordering;
36: MatOrderingType otype;
38: PetscCall(MatFactorGetCanUseOrdering(fact->fact, &canuseordering));
39: if (canuseordering) {
40: PetscCall(MatFactorGetPreferredOrdering(fact->fact, fact->factortype, &otype));
41: } else otype = MATORDERINGEXTERNAL;
42: PetscCall(PetscStrallocpy(otype, (char **)&fact->ordering));
43: }
44: if (destroy) PetscCall(MatDestroy(&fact->fact));
45: }
46: }
47: PetscFunctionReturn(PETSC_SUCCESS);
48: }
50: static PetscErrorCode PCFactorSetReuseOrdering_Factor(PC pc, PetscBool flag)
51: {
52: PC_Factor *lu = (PC_Factor *)pc->data;
54: PetscFunctionBegin;
55: lu->reuseordering = flag;
56: PetscFunctionReturn(PETSC_SUCCESS);
57: }
59: static PetscErrorCode PCFactorSetReuseFill_Factor(PC pc, PetscBool flag)
60: {
61: PC_Factor *lu = (PC_Factor *)pc->data;
63: PetscFunctionBegin;
64: lu->reusefill = flag;
65: PetscFunctionReturn(PETSC_SUCCESS);
66: }
68: static PetscErrorCode PCFactorSetUseInPlace_Factor(PC pc, PetscBool flg)
69: {
70: PC_Factor *dir = (PC_Factor *)pc->data;
72: PetscFunctionBegin;
73: dir->inplace = flg;
74: PetscFunctionReturn(PETSC_SUCCESS);
75: }
77: static PetscErrorCode PCFactorGetUseInPlace_Factor(PC pc, PetscBool *flg)
78: {
79: PC_Factor *dir = (PC_Factor *)pc->data;
81: PetscFunctionBegin;
82: *flg = dir->inplace;
83: PetscFunctionReturn(PETSC_SUCCESS);
84: }
86: /*@
87: PCFactorSetUpMatSolverType - Can be called after `KSPSetOperators()` or `PCSetOperators()`, causes `MatGetFactor()` to be called so then one may
88: set the options for that particular factorization object.
90: Input Parameter:
91: . pc - the preconditioner context
93: Note:
94: After you have called this function (which has to be after the `KSPSetOperators()` or `PCSetOperators()`) you can call `PCFactorGetMatrix()` and then set factor options on that matrix.
95: This function raises an error if the requested combination of solver package and matrix type is not supported.
97: Level: intermediate
99: .seealso: [](ch_ksp), `PCCHOLESKY`, `PCLU`, `PCFactorSetMatSolverType()`, `PCFactorGetMatrix()`
100: @*/
101: PetscErrorCode PCFactorSetUpMatSolverType(PC pc)
102: {
103: PetscFunctionBegin;
105: PetscTryMethod(pc, "PCFactorSetUpMatSolverType_C", (PC), (pc));
106: PetscFunctionReturn(PETSC_SUCCESS);
107: }
109: /*@
110: PCFactorSetZeroPivot - Sets the size at which smaller pivots are declared to be zero
112: Logically Collective
114: Input Parameters:
115: + pc - the preconditioner context
116: - zero - all pivots smaller than this will be considered zero
118: Options Database Key:
119: . -pc_factor_zeropivot <zero> - Sets tolerance for what is considered a zero pivot
121: Level: intermediate
123: .seealso: [](ch_ksp), `PCCHOLESKY`, `PCLU`, `PCFactorSetShiftType()`, `PCFactorSetShiftAmount()`
124: @*/
125: PetscErrorCode PCFactorSetZeroPivot(PC pc, PetscReal zero)
126: {
127: PetscFunctionBegin;
130: PetscTryMethod(pc, "PCFactorSetZeroPivot_C", (PC, PetscReal), (pc, zero));
131: PetscFunctionReturn(PETSC_SUCCESS);
132: }
134: /*@
135: PCFactorSetShiftType - adds a particular type of quantity to the diagonal of the matrix during
136: numerical factorization, thus the matrix has nonzero pivots
138: Logically Collective
140: Input Parameters:
141: + pc - the preconditioner context
142: - shifttype - type of shift; one of `MAT_SHIFT_NONE`, `MAT_SHIFT_NONZERO`, `MAT_SHIFT_POSITIVE_DEFINITE`, `MAT_SHIFT_INBLOCKS`
144: Options Database Key:
145: . -pc_factor_shift_type <shifttype> - Sets shift type; use '-help' for a list of available types
147: Level: intermediate
149: .seealso: [](ch_ksp), `PCCHOLESKY`, `PCLU`, `PCFactorSetZeroPivot()`, `PCFactorSetShiftAmount()`
150: @*/
151: PetscErrorCode PCFactorSetShiftType(PC pc, MatFactorShiftType shifttype)
152: {
153: PetscFunctionBegin;
156: PetscTryMethod(pc, "PCFactorSetShiftType_C", (PC, MatFactorShiftType), (pc, shifttype));
157: PetscFunctionReturn(PETSC_SUCCESS);
158: }
160: /*@
161: PCFactorSetShiftAmount - adds a quantity to the diagonal of the matrix during
162: numerical factorization, thus the matrix has nonzero pivots
164: Logically Collective
166: Input Parameters:
167: + pc - the preconditioner context
168: - shiftamount - amount of shift or `PETSC_DECIDE` for the default
170: Options Database Key:
171: . -pc_factor_shift_amount <shiftamount> - Sets shift amount or -1 for the default
173: Level: intermediate
175: .seealso: [](ch_ksp), `PCCHOLESKY`, `PCLU`, `PCFactorSetZeroPivot()`, `PCFactorSetShiftType()`
176: @*/
177: PetscErrorCode PCFactorSetShiftAmount(PC pc, PetscReal shiftamount)
178: {
179: PetscFunctionBegin;
182: PetscTryMethod(pc, "PCFactorSetShiftAmount_C", (PC, PetscReal), (pc, shiftamount));
183: PetscFunctionReturn(PETSC_SUCCESS);
184: }
186: /*@
187: PCFactorSetDropTolerance - The preconditioner will use an `PCILU`
188: based on a drop tolerance.
190: Logically Collective
192: Input Parameters:
193: + pc - the preconditioner context
194: . dt - the drop tolerance, try from 1.e-10 to .1
195: . dtcol - tolerance for column pivot, good values [0.1 to 0.01]
196: - maxrowcount - the max number of nonzeros allowed in a row, best value
197: depends on the number of nonzeros in row of original matrix
199: Options Database Key:
200: . -pc_factor_drop_tolerance <dt,dtcol,maxrowcount> - Sets drop tolerance
202: Level: intermediate
204: Note:
205: There are NO default values for the 3 parameters, you must set them with reasonable values for your
206: matrix. We don't know how to compute reasonable values.
208: .seealso: [](ch_ksp), `PCILU`
209: @*/
210: PetscErrorCode PCFactorSetDropTolerance(PC pc, PetscReal dt, PetscReal dtcol, PetscInt maxrowcount)
211: {
212: PetscFunctionBegin;
216: PetscTryMethod(pc, "PCFactorSetDropTolerance_C", (PC, PetscReal, PetscReal, PetscInt), (pc, dt, dtcol, maxrowcount));
217: PetscFunctionReturn(PETSC_SUCCESS);
218: }
220: /*@
221: PCFactorGetZeroPivot - Gets the tolerance used to define a zero privot
223: Not Collective
225: Input Parameter:
226: . pc - the preconditioner context
228: Output Parameter:
229: . pivot - the tolerance
231: Level: intermediate
233: .seealso: [](ch_ksp), `PCLU`, `PCCHOLESKY`, `PCFactorSetZeroPivot()`
234: @*/
235: PetscErrorCode PCFactorGetZeroPivot(PC pc, PetscReal *pivot)
236: {
237: PetscFunctionBegin;
239: PetscUseMethod(pc, "PCFactorGetZeroPivot_C", (PC, PetscReal *), (pc, pivot));
240: PetscFunctionReturn(PETSC_SUCCESS);
241: }
243: /*@
244: PCFactorGetShiftAmount - Gets the tolerance used to define a zero privot
246: Not Collective
248: Input Parameter:
249: . pc - the preconditioner context
251: Output Parameter:
252: . shift - how much to shift the diagonal entry
254: Level: intermediate
256: .seealso: [](ch_ksp), `PCLU`, `PCCHOLESKY`, `PCFactorSetShiftAmount()`, `PCFactorSetShiftType()`, `PCFactorGetShiftType()`
257: @*/
258: PetscErrorCode PCFactorGetShiftAmount(PC pc, PetscReal *shift)
259: {
260: PetscFunctionBegin;
262: PetscUseMethod(pc, "PCFactorGetShiftAmount_C", (PC, PetscReal *), (pc, shift));
263: PetscFunctionReturn(PETSC_SUCCESS);
264: }
266: /*@
267: PCFactorGetShiftType - Gets the type of shift, if any, done when a zero pivot is detected
269: Not Collective
271: Input Parameter:
272: . pc - the preconditioner context
274: Output Parameter:
275: . type - one of `MAT_SHIFT_NONE`, `MAT_SHIFT_NONZERO`, `MAT_SHIFT_POSITIVE_DEFINITE`, or `MAT_SHIFT_INBLOCKS`
277: Level: intermediate
279: .seealso: [](ch_ksp), `PCLU`, `PCCHOLESKY`, `PCFactorSetShiftType()`, `MatFactorShiftType`, `PCFactorSetShiftAmount()`, `PCFactorGetShiftAmount()`
280: @*/
281: PetscErrorCode PCFactorGetShiftType(PC pc, MatFactorShiftType *type)
282: {
283: PetscFunctionBegin;
285: PetscUseMethod(pc, "PCFactorGetShiftType_C", (PC, MatFactorShiftType *), (pc, type));
286: PetscFunctionReturn(PETSC_SUCCESS);
287: }
289: /*@
290: PCFactorGetLevels - Gets the number of levels of fill to use.
292: Logically Collective
294: Input Parameter:
295: . pc - the preconditioner context
297: Output Parameter:
298: . levels - number of levels of fill
300: Level: intermediate
302: .seealso: [](ch_ksp), `PCILU`, `PCICC`, `PCFactorSetLevels()`
303: @*/
304: PetscErrorCode PCFactorGetLevels(PC pc, PetscInt *levels)
305: {
306: PetscFunctionBegin;
308: PetscUseMethod(pc, "PCFactorGetLevels_C", (PC, PetscInt *), (pc, levels));
309: PetscFunctionReturn(PETSC_SUCCESS);
310: }
312: /*@
313: PCFactorSetLevels - Sets the number of levels of fill to use.
315: Logically Collective
317: Input Parameters:
318: + pc - the preconditioner context
319: - levels - number of levels of fill
321: Options Database Key:
322: . -pc_factor_levels <levels> - Sets fill level
324: Level: intermediate
326: .seealso: [](ch_ksp), `PCILU`, `PCICC`, `PCFactorGetLevels()`
327: @*/
328: PetscErrorCode PCFactorSetLevels(PC pc, PetscInt levels)
329: {
330: PetscFunctionBegin;
332: PetscCheck(levels >= 0, PetscObjectComm((PetscObject)pc), PETSC_ERR_ARG_OUTOFRANGE, "negative levels");
334: PetscTryMethod(pc, "PCFactorSetLevels_C", (PC, PetscInt), (pc, levels));
335: PetscFunctionReturn(PETSC_SUCCESS);
336: }
338: /*@
339: PCFactorSetAllowDiagonalFill - Causes all diagonal matrix entries to be
340: treated as level 0 fill even if there is no non-zero location.
342: Logically Collective
344: Input Parameters:
345: + pc - the preconditioner context
346: - flg - `PETSC_TRUE` to turn on, `PETSC_FALSE` to turn off
348: Options Database Key:
349: . -pc_factor_diagonal_fill <bool> - allow the diagonal fill
351: Note:
352: Does not apply with 0 fill.
354: Level: intermediate
356: .seealso: [](ch_ksp), `PCILU`, `PCICC`, `PCFactorGetAllowDiagonalFill()`
357: @*/
358: PetscErrorCode PCFactorSetAllowDiagonalFill(PC pc, PetscBool flg)
359: {
360: PetscFunctionBegin;
362: PetscTryMethod(pc, "PCFactorSetAllowDiagonalFill_C", (PC, PetscBool), (pc, flg));
363: PetscFunctionReturn(PETSC_SUCCESS);
364: }
366: /*@
367: PCFactorGetAllowDiagonalFill - Determines if all diagonal matrix entries are
368: treated as level 0 fill even if there is no non-zero location.
370: Logically Collective
372: Input Parameter:
373: . pc - the preconditioner context
375: Output Parameter:
376: . flg - `PETSC_TRUE` to turn on, `PETSC_FALSE` to turn off
378: Note:
379: Does not apply with 0 fill.
381: Level: intermediate
383: .seealso: [](ch_ksp), `PCILU`, `PCICC`, `PCFactorSetAllowDiagonalFill()`
384: @*/
385: PetscErrorCode PCFactorGetAllowDiagonalFill(PC pc, PetscBool *flg)
386: {
387: PetscFunctionBegin;
389: PetscUseMethod(pc, "PCFactorGetAllowDiagonalFill_C", (PC, PetscBool *), (pc, flg));
390: PetscFunctionReturn(PETSC_SUCCESS);
391: }
393: /*@
394: PCFactorReorderForNonzeroDiagonal - reorders rows/columns of matrix to remove zeros from diagonal
396: Logically Collective
398: Input Parameters:
399: + pc - the preconditioner context
400: - rtol - diagonal entries smaller than this in absolute value are considered zero
402: Options Database Key:
403: . -pc_factor_nonzeros_along_diagonal <tol> - perform the reordering with the given tolerance
405: Level: intermediate
407: .seealso: [](ch_ksp), `PCILU`, `PCICC`, `PCFactorSetFill()`, `PCFactorSetShiftAmount()`, `PCFactorSetZeroPivot()`, `MatReorderForNonzeroDiagonal()`
408: @*/
409: PetscErrorCode PCFactorReorderForNonzeroDiagonal(PC pc, PetscReal rtol)
410: {
411: PetscFunctionBegin;
414: PetscTryMethod(pc, "PCFactorReorderForNonzeroDiagonal_C", (PC, PetscReal), (pc, rtol));
415: PetscFunctionReturn(PETSC_SUCCESS);
416: }
418: /*@
419: PCFactorSetMatSolverType - sets the solver package that is used to perform the factorization
421: Logically Collective
423: Input Parameters:
424: + pc - the preconditioner context
425: - stype - for example, `MATSOLVERSUPERLU`, `MATSOLVERSUPERLU_DIST`, `MATSOLVERMUMPS`
427: Options Database Key:
428: . -pc_factor_mat_solver_type <stype> - petsc, superlu, superlu_dist, mumps, cusparse
430: Level: intermediate
432: Note:
433: The default type is set by searching for available types based on the order of the calls to `MatSolverTypeRegister()` in `MatInitializePackage()`.
434: Since different PETSc configurations may have different external solvers, seemingly identical runs with different PETSc configurations may use a different solver.
435: For example if one configuration had --download-mumps while a different one had --download-superlu_dist.
437: .seealso: [](ch_ksp), `PCLU`, `PCCHOLESKY`, `MatGetFactor()`, `MatSolverType`, `PCFactorGetMatSolverType()`, `MatSolverTypeRegister()`,
438: `MatInitializePackage()`, `MATSOLVERSUPERLU`, `MATSOLVERSUPERLU_DIST`, `MATSOLVERMUMPS`, `MatSolverTypeGet()`
439: @*/
440: PetscErrorCode PCFactorSetMatSolverType(PC pc, MatSolverType stype)
441: {
442: PetscFunctionBegin;
444: PetscTryMethod(pc, "PCFactorSetMatSolverType_C", (PC, MatSolverType), (pc, stype));
445: PetscFunctionReturn(PETSC_SUCCESS);
446: }
448: /*@
449: PCFactorGetMatSolverType - gets the solver package that is used to perform the factorization
451: Not Collective
453: Input Parameter:
454: . pc - the preconditioner context
456: Output Parameter:
457: . stype - for example, `MATSOLVERSUPERLU`, `MATSOLVERSUPERLU_DIST`, `MATSOLVERMUMPS`
459: Level: intermediate
461: .seealso: [](ch_ksp), `PCLU`, `PCCHOLESKY`, `MatGetFactor()`, `MatSolverType`, `MATSOLVERSUPERLU`,
462: `MATSOLVERSUPERLU_DIST`, `MATSOLVERMUMPS`
463: @*/
464: PetscErrorCode PCFactorGetMatSolverType(PC pc, MatSolverType *stype)
465: {
466: PetscErrorCode (*f)(PC, MatSolverType *);
468: PetscFunctionBegin;
470: PetscAssertPointer(stype, 2);
471: PetscCall(PetscObjectQueryFunction((PetscObject)pc, "PCFactorGetMatSolverType_C", &f));
472: if (f) PetscCall((*f)(pc, stype));
473: else *stype = NULL;
474: PetscFunctionReturn(PETSC_SUCCESS);
475: }
477: /*@
478: PCFactorSetFill - Indicate the amount of fill you expect in the factored matrix,
479: fill = number nonzeros in factor/number nonzeros in original matrix.
481: Not Collective, each process can expect a different amount of fill
483: Input Parameters:
484: + pc - the preconditioner context
485: - fill - amount of expected fill
487: Options Database Key:
488: . -pc_factor_fill <fill> - Sets fill amount
490: Level: intermediate
492: Notes:
493: For sparse matrix factorizations it is difficult to predict how much
494: fill to expect. By running with the option -info PETSc will print the
495: actual amount of fill used; allowing you to set the value accurately for
496: future runs. Default PETSc uses a value of 5.0
498: This is ignored for most solver packages
500: This parameter has NOTHING to do with the levels-of-fill of ILU(). That is set with `PCFactorSetLevels()` or -pc_factor_levels.
502: .seealso: [](ch_ksp), `PCLU`, `PCCHOLESKY`, `PCILU`, `PCICC`, `PCFactorSetReuseFill()`
503: @*/
504: PetscErrorCode PCFactorSetFill(PC pc, PetscReal fill)
505: {
506: PetscFunctionBegin;
508: PetscCheck(fill >= 1.0, PetscObjectComm((PetscObject)pc), PETSC_ERR_ARG_OUTOFRANGE, "Fill factor cannot be less than 1.0");
509: PetscTryMethod(pc, "PCFactorSetFill_C", (PC, PetscReal), (pc, fill));
510: PetscFunctionReturn(PETSC_SUCCESS);
511: }
513: /*@
514: PCFactorSetUseInPlace - Tells the preconditioner to do an in-place factorization.
516: Logically Collective
518: Input Parameters:
519: + pc - the preconditioner context
520: - flg - `PETSC_TRUE` to enable, `PETSC_FALSE` to disable
522: Options Database Key:
523: . -pc_factor_in_place <true,false> - Activate/deactivate in-place factorization
525: Note:
526: For dense matrices, this enables the solution of much larger problems.
527: For sparse matrices the factorization cannot be done truly in-place
528: so this does not save memory during the factorization, but after the matrix
529: is factored, the original unfactored matrix is freed, thus recovering that
530: space. For ICC(0) and ILU(0) with the default natural ordering the factorization is done efficiently in-place.
532: `PCFactorSetUseInplace()` can only be used with the `KSP` method `KSPPREONLY` or when
533: a different matrix is provided for the multiply and the preconditioner in
534: a call to `KSPSetOperators()`.
535: This is because the Krylov space methods require an application of the
536: matrix multiplication, which is not possible here because the matrix has
537: been factored in-place, replacing the original matrix.
539: Level: intermediate
541: .seealso: [](ch_ksp), `PC`, `Mat`, `PCLU`, `PCCHOLESKY`, `PCILU`, `PCICC`, `PCFactorGetUseInPlace()`
542: @*/
543: PetscErrorCode PCFactorSetUseInPlace(PC pc, PetscBool flg)
544: {
545: PetscFunctionBegin;
547: PetscTryMethod(pc, "PCFactorSetUseInPlace_C", (PC, PetscBool), (pc, flg));
548: PetscFunctionReturn(PETSC_SUCCESS);
549: }
551: /*@
552: PCFactorGetUseInPlace - Determines if an in-place factorization is being used.
554: Logically Collective
556: Input Parameter:
557: . pc - the preconditioner context
559: Output Parameter:
560: . flg - `PETSC_TRUE` to enable, `PETSC_FALSE` to disable
562: Level: intermediate
564: .seealso: [](ch_ksp), `PCLU`, `PCCHOLESKY`, `PCILU`, `PCICC`, `PCFactorSetUseInPlace()`
565: @*/
566: PetscErrorCode PCFactorGetUseInPlace(PC pc, PetscBool *flg)
567: {
568: PetscFunctionBegin;
570: PetscUseMethod(pc, "PCFactorGetUseInPlace_C", (PC, PetscBool *), (pc, flg));
571: PetscFunctionReturn(PETSC_SUCCESS);
572: }
574: /*@
575: PCFactorSetMatOrderingType - Sets the ordering routine (to reduce fill) to
576: be used in the `PCLU`, `PCCHOLESKY`, `PCILU`, or `PCICC` preconditioners
578: Logically Collective
580: Input Parameters:
581: + pc - the preconditioner context
582: - ordering - the matrix ordering name, for example, `MATORDERINGND` or `MATORDERINGRCM`
584: Options Database Key:
585: . -pc_factor_mat_ordering_type <nd,rcm,...,external> - Sets ordering routine
587: Level: intermediate
589: Notes:
590: Nested dissection is used by default for some of PETSc's sparse matrix formats
592: For `PCCHOLESKY` and `PCICC` and the `MATSBAIJ` format the only reordering available is natural since only the upper half of the matrix is stored
593: and reordering this matrix is very expensive.
595: You can use a `MATSEQAIJ` matrix with Cholesky and ICC and use any ordering.
597: `MATORDERINGEXTERNAL` means PETSc will not compute an ordering and the package will use its own ordering, usable with `MATSOLVERCHOLMOD`, `MATSOLVERUMFPACK`, and others.
599: .seealso: [](ch_ksp), `PCLU`, `PCCHOLESKY`, `PCILU`, `PCICC`, `MatOrderingType`, `MATORDERINGEXTERNAL`, `MATORDERINGND`, `MATORDERINGRCM`
600: @*/
601: PetscErrorCode PCFactorSetMatOrderingType(PC pc, MatOrderingType ordering)
602: {
603: PetscFunctionBegin;
605: PetscTryMethod(pc, "PCFactorSetMatOrderingType_C", (PC, MatOrderingType), (pc, ordering));
606: PetscFunctionReturn(PETSC_SUCCESS);
607: }
609: /*@
610: PCFactorSetColumnPivot - Determines when column pivoting is done during matrix factorization.
611: For PETSc dense matrices column pivoting is always done, for PETSc sparse matrices
612: it is never done. For the MATLAB and `MATSOLVERSUPERLU` factorization this is used.
614: Logically Collective
616: Input Parameters:
617: + pc - the preconditioner context
618: - dtcol - 0.0 implies no pivoting, 1.0 complete pivoting (slower, requires more memory but more stable)
620: Options Database Key:
621: . -pc_factor_pivoting <dtcol> - perform the pivoting with the given tolerance
623: Level: intermediate
625: .seealso: [](ch_ksp), `PCLU`, `PCCHOLESKY`, `PCILU`, `PCICC`, `PCILUSetMatOrdering()`, `PCFactorSetPivotInBlocks()`
626: @*/
627: PetscErrorCode PCFactorSetColumnPivot(PC pc, PetscReal dtcol)
628: {
629: PetscFunctionBegin;
632: PetscTryMethod(pc, "PCFactorSetColumnPivot_C", (PC, PetscReal), (pc, dtcol));
633: PetscFunctionReturn(PETSC_SUCCESS);
634: }
636: /*@
637: PCFactorSetPivotInBlocks - Determines if pivoting is done while factoring each block
638: with `MATBAIJ` or `MATSBAIJ` matrices
640: Logically Collective
642: Input Parameters:
643: + pc - the preconditioner context
644: - pivot - `PETSC_TRUE` or `PETSC_FALSE`
646: Options Database Key:
647: . -pc_factor_pivot_in_blocks <true,false> - Pivot inside matrix dense blocks for `MATBAIJ` and `MATSBAIJ`
649: Level: intermediate
651: .seealso: [](ch_ksp), `PCLU`, `PCCHOLESKY`, `PCILU`, `PCICC`, `PCILUSetMatOrdering()`, `PCFactorSetColumnPivot()`
652: @*/
653: PetscErrorCode PCFactorSetPivotInBlocks(PC pc, PetscBool pivot)
654: {
655: PetscFunctionBegin;
658: PetscTryMethod(pc, "PCFactorSetPivotInBlocks_C", (PC, PetscBool), (pc, pivot));
659: PetscFunctionReturn(PETSC_SUCCESS);
660: }
662: /*@
663: PCFactorSetReuseFill - When matrices with different nonzero structure are factored,
664: this causes later ones to use the fill ratio computed in the initial factorization.
666: Logically Collective
668: Input Parameters:
669: + pc - the preconditioner context
670: - flag - `PETSC_TRUE` to reuse else `PETSC_FALSE`
672: Options Database Key:
673: . -pc_factor_reuse_fill - Activates `PCFactorSetReuseFill()`
675: Level: intermediate
677: .seealso: [](ch_ksp), `PCLU`, `PCCHOLESKY`, `PCILU`, `PCICC`, `PCFactorSetReuseOrdering()`, `PCFactorSetFill()`
678: @*/
679: PetscErrorCode PCFactorSetReuseFill(PC pc, PetscBool flag)
680: {
681: PetscFunctionBegin;
684: PetscTryMethod(pc, "PCFactorSetReuseFill_C", (PC, PetscBool), (pc, flag));
685: PetscFunctionReturn(PETSC_SUCCESS);
686: }
688: PetscErrorCode PCFactorInitialize(PC pc, MatFactorType ftype)
689: {
690: PC_Factor *fact = (PC_Factor *)pc->data;
692: PetscFunctionBegin;
693: PetscCall(MatFactorInfoInitialize(&fact->info));
694: fact->factortype = ftype;
695: fact->info.shifttype = (PetscReal)MAT_SHIFT_NONE;
696: fact->info.shiftamount = 100.0 * PETSC_MACHINE_EPSILON;
697: fact->info.zeropivot = 100.0 * PETSC_MACHINE_EPSILON;
698: fact->info.pivotinblocks = 1.0;
699: pc->ops->getfactoredmatrix = PCFactorGetMatrix_Factor;
701: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetZeroPivot_C", PCFactorSetZeroPivot_Factor));
702: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetZeroPivot_C", PCFactorGetZeroPivot_Factor));
703: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetShiftType_C", PCFactorSetShiftType_Factor));
704: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetShiftType_C", PCFactorGetShiftType_Factor));
705: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetShiftAmount_C", PCFactorSetShiftAmount_Factor));
706: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetShiftAmount_C", PCFactorGetShiftAmount_Factor));
707: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetMatSolverType_C", PCFactorGetMatSolverType_Factor));
708: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetMatSolverType_C", PCFactorSetMatSolverType_Factor));
709: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetUpMatSolverType_C", PCFactorSetUpMatSolverType_Factor));
710: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetFill_C", PCFactorSetFill_Factor));
711: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetMatOrderingType_C", PCFactorSetMatOrderingType_Factor));
712: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetLevels_C", PCFactorSetLevels_Factor));
713: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetLevels_C", PCFactorGetLevels_Factor));
714: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetAllowDiagonalFill_C", PCFactorSetAllowDiagonalFill_Factor));
715: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetAllowDiagonalFill_C", PCFactorGetAllowDiagonalFill_Factor));
716: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetPivotInBlocks_C", PCFactorSetPivotInBlocks_Factor));
717: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetUseInPlace_C", PCFactorSetUseInPlace_Factor));
718: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetUseInPlace_C", PCFactorGetUseInPlace_Factor));
719: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetReuseOrdering_C", PCFactorSetReuseOrdering_Factor));
720: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetReuseFill_C", PCFactorSetReuseFill_Factor));
721: PetscFunctionReturn(PETSC_SUCCESS);
722: }
724: PetscErrorCode PCFactorClearComposedFunctions(PC pc)
725: {
726: PetscFunctionBegin;
727: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetZeroPivot_C", NULL));
728: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetZeroPivot_C", NULL));
729: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetShiftType_C", NULL));
730: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetShiftType_C", NULL));
731: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetShiftAmount_C", NULL));
732: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetShiftAmount_C", NULL));
733: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetMatSolverType_C", NULL));
734: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetMatSolverType_C", NULL));
735: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetUpMatSolverType_C", NULL));
736: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetFill_C", NULL));
737: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetMatOrderingType_C", NULL));
738: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetLevels_C", NULL));
739: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetLevels_C", NULL));
740: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetAllowDiagonalFill_C", NULL));
741: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetAllowDiagonalFill_C", NULL));
742: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetPivotInBlocks_C", NULL));
743: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetUseInPlace_C", NULL));
744: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorGetUseInPlace_C", NULL));
745: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetReuseOrdering_C", NULL));
746: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetReuseFill_C", NULL));
747: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorReorderForNonzeroDiagonal_C", NULL));
748: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCFactorSetDropTolerance_C", NULL));
749: PetscFunctionReturn(PETSC_SUCCESS);
750: }