Actual source code: matrix.c
1: /*
2: This is where the abstract matrix operations are defined
3: Portions of this code are under:
4: Copyright (c) 2022 Advanced Micro Devices, Inc. All rights reserved.
5: */
7: #include <petsc/private/matimpl.h>
8: #include <petsc/private/isimpl.h>
9: #include <petsc/private/vecimpl.h>
11: /* Logging support */
12: PetscClassId MAT_CLASSID;
13: PetscClassId MAT_COLORING_CLASSID;
14: PetscClassId MAT_FDCOLORING_CLASSID;
15: PetscClassId MAT_TRANSPOSECOLORING_CLASSID;
17: PetscLogEvent MAT_Mult, MAT_MultAdd, MAT_MultTranspose;
18: PetscLogEvent MAT_ADot, MAT_ANorm;
19: PetscLogEvent MAT_MultTransposeAdd, MAT_Solve, MAT_Solves, MAT_SolveAdd, MAT_SolveTranspose, MAT_MatSolve, MAT_MatTrSolve;
20: PetscLogEvent MAT_SolveTransposeAdd, MAT_SOR, MAT_ForwardSolve, MAT_BackwardSolve, MAT_LUFactor, MAT_LUFactorSymbolic;
21: PetscLogEvent MAT_LUFactorNumeric, MAT_CholeskyFactor, MAT_CholeskyFactorSymbolic, MAT_CholeskyFactorNumeric, MAT_ILUFactor;
22: PetscLogEvent MAT_ILUFactorSymbolic, MAT_ICCFactorSymbolic, MAT_Copy, MAT_Convert, MAT_Scale, MAT_AssemblyBegin;
23: PetscLogEvent MAT_QRFactorNumeric, MAT_QRFactorSymbolic, MAT_QRFactor;
24: PetscLogEvent MAT_AssemblyEnd, MAT_SetValues, MAT_GetValues, MAT_GetRow, MAT_GetRowIJ, MAT_CreateSubMats, MAT_GetOrdering, MAT_RedundantMat, MAT_GetSeqNonzeroStructure;
25: PetscLogEvent MAT_IncreaseOverlap, MAT_Partitioning, MAT_PartitioningND, MAT_Coarsen, MAT_ZeroEntries, MAT_Load, MAT_View, MAT_AXPY, MAT_FDColoringCreate;
26: PetscLogEvent MAT_FDColoringSetUp, MAT_FDColoringApply, MAT_Transpose, MAT_FDColoringFunction, MAT_CreateSubMat;
27: PetscLogEvent MAT_TransposeColoringCreate;
28: PetscLogEvent MAT_MatMult, MAT_MatMultSymbolic, MAT_MatMultNumeric;
29: PetscLogEvent MAT_PtAP, MAT_PtAPSymbolic, MAT_PtAPNumeric, MAT_RARt, MAT_RARtSymbolic, MAT_RARtNumeric;
30: PetscLogEvent MAT_MatTransposeMult, MAT_MatTransposeMultSymbolic, MAT_MatTransposeMultNumeric;
31: PetscLogEvent MAT_TransposeMatMult, MAT_TransposeMatMultSymbolic, MAT_TransposeMatMultNumeric;
32: PetscLogEvent MAT_MatMatMult, MAT_MatMatMultSymbolic, MAT_MatMatMultNumeric;
33: PetscLogEvent MAT_MultHermitianTranspose, MAT_MultHermitianTransposeAdd;
34: PetscLogEvent MAT_Getsymtransreduced, MAT_GetBrowsOfAcols;
35: PetscLogEvent MAT_GetBrowsOfAocols, MAT_Getlocalmat, MAT_Getlocalmatcondensed, MAT_Seqstompi, MAT_Seqstompinum, MAT_Seqstompisym;
36: PetscLogEvent MAT_GetMultiProcBlock;
37: PetscLogEvent MAT_CUSPARSECopyToGPU, MAT_CUSPARSECopyFromGPU, MAT_CUSPARSEGenerateTranspose, MAT_CUSPARSESolveAnalysis;
38: PetscLogEvent MAT_HIPSPARSECopyToGPU, MAT_HIPSPARSECopyFromGPU, MAT_HIPSPARSEGenerateTranspose, MAT_HIPSPARSESolveAnalysis;
39: PetscLogEvent MAT_PreallCOO, MAT_SetVCOO;
40: PetscLogEvent MAT_CreateGraph;
41: PetscLogEvent MAT_SetValuesBatch;
42: PetscLogEvent MAT_ViennaCLCopyToGPU;
43: PetscLogEvent MAT_CUDACopyToGPU, MAT_HIPCopyToGPU;
44: PetscLogEvent MAT_DenseCopyToGPU, MAT_DenseCopyFromGPU;
45: PetscLogEvent MAT_Merge, MAT_Residual, MAT_SetRandom;
46: PetscLogEvent MAT_FactorFactS, MAT_FactorInvS;
47: PetscLogEvent MATCOLORING_Apply, MATCOLORING_Comm, MATCOLORING_Local, MATCOLORING_ISCreate, MATCOLORING_SetUp, MATCOLORING_Weights;
48: PetscLogEvent MAT_H2Opus_Build, MAT_H2Opus_Compress, MAT_H2Opus_Orthog, MAT_H2Opus_LR;
50: const char *const MatFactorTypes[] = {"NONE", "LU", "CHOLESKY", "ILU", "ICC", "ILUDT", "QR", "MatFactorType", "MAT_FACTOR_", NULL};
52: /*@
53: MatSetRandom - Sets all components of a matrix to random numbers.
55: Logically Collective
57: Input Parameters:
58: + x - the matrix
59: - rctx - the `PetscRandom` object, formed by `PetscRandomCreate()`, or `NULL` and
60: it will create one internally.
62: Example:
63: .vb
64: PetscRandomCreate(PETSC_COMM_WORLD,&rctx);
65: MatSetRandom(x,rctx);
66: PetscRandomDestroy(rctx);
67: .ve
69: Level: intermediate
71: Notes:
72: For sparse matrices that have been preallocated but not been assembled, it randomly selects appropriate locations,
74: for sparse matrices that already have nonzero locations, it fills the locations with random numbers.
76: It generates an error if used on unassembled sparse matrices that have not been preallocated.
78: .seealso: [](ch_matrices), `Mat`, `PetscRandom`, `PetscRandomCreate()`, `MatZeroEntries()`, `MatSetValues()`, `PetscRandomDestroy()`
79: @*/
80: PetscErrorCode MatSetRandom(Mat x, PetscRandom rctx)
81: {
82: PetscRandom randObj = NULL;
84: PetscFunctionBegin;
88: MatCheckPreallocated(x, 1);
90: if (!rctx) {
91: MPI_Comm comm;
92: PetscCall(PetscObjectGetComm((PetscObject)x, &comm));
93: PetscCall(PetscRandomCreate(comm, &randObj));
94: PetscCall(PetscRandomSetType(randObj, x->defaultrandtype));
95: PetscCall(PetscRandomSetFromOptions(randObj));
96: rctx = randObj;
97: }
98: PetscCall(PetscLogEventBegin(MAT_SetRandom, x, rctx, 0, 0));
99: PetscUseTypeMethod(x, setrandom, rctx);
100: PetscCall(PetscLogEventEnd(MAT_SetRandom, x, rctx, 0, 0));
102: PetscCall(MatAssemblyBegin(x, MAT_FINAL_ASSEMBLY));
103: PetscCall(MatAssemblyEnd(x, MAT_FINAL_ASSEMBLY));
104: PetscCall(PetscRandomDestroy(&randObj));
105: PetscFunctionReturn(PETSC_SUCCESS);
106: }
108: /*@
109: MatCopyHashToXAIJ - copy hash table entries into an XAIJ matrix type
111: Logically Collective
113: Input Parameter:
114: . A - A matrix in unassembled, hash table form
116: Output Parameter:
117: . B - The XAIJ matrix. This can either be `A` or some matrix of equivalent size, e.g. obtained from `A` via `MatDuplicate()`
119: Example:
120: .vb
121: PetscCall(MatDuplicate(A, MAT_DO_NOT_COPY_VALUES, &B));
122: PetscCall(MatCopyHashToXAIJ(A, B));
123: .ve
125: Level: advanced
127: Notes:
128: If `B` is `A`, then the hash table data structure will be destroyed. `B` is assembled
130: .seealso: [](ch_matrices), `Mat`, `MAT_USE_HASH_TABLE`
131: @*/
132: PetscErrorCode MatCopyHashToXAIJ(Mat A, Mat B)
133: {
134: PetscFunctionBegin;
136: PetscUseTypeMethod(A, copyhashtoxaij, B);
137: PetscFunctionReturn(PETSC_SUCCESS);
138: }
140: /*@
141: MatFactorGetErrorZeroPivot - returns the pivot value that was determined to be zero and the row it occurred in
143: Logically Collective
145: Input Parameter:
146: . mat - the factored matrix
148: Output Parameters:
149: + pivot - the pivot value computed
150: - row - the row that the zero pivot occurred. This row value must be interpreted carefully due to row reorderings and which processes
151: the share the matrix
153: Level: advanced
155: Notes:
156: This routine does not work for factorizations done with external packages.
158: This routine should only be called if `MatGetFactorError()` returns a value of `MAT_FACTOR_NUMERIC_ZEROPIVOT`
160: This can also be called on non-factored matrices that come from, for example, matrices used in SOR.
162: .seealso: [](ch_matrices), `Mat`, `MatZeroEntries()`, `MatFactor()`, `MatGetFactor()`,
163: `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`, `MatFactorClearError()`,
164: `MAT_FACTOR_NUMERIC_ZEROPIVOT`
165: @*/
166: PetscErrorCode MatFactorGetErrorZeroPivot(Mat mat, PetscReal *pivot, PetscInt *row)
167: {
168: PetscFunctionBegin;
170: PetscAssertPointer(pivot, 2);
171: PetscAssertPointer(row, 3);
172: *pivot = mat->factorerror_zeropivot_value;
173: *row = mat->factorerror_zeropivot_row;
174: PetscFunctionReturn(PETSC_SUCCESS);
175: }
177: /*@
178: MatFactorGetError - gets the error code from a factorization
180: Logically Collective
182: Input Parameter:
183: . mat - the factored matrix
185: Output Parameter:
186: . err - the error code
188: Level: advanced
190: Note:
191: This can also be called on non-factored matrices that come from, for example, matrices used in SOR.
193: .seealso: [](ch_matrices), `Mat`, `MatZeroEntries()`, `MatFactor()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`,
194: `MatFactorClearError()`, `MatFactorGetErrorZeroPivot()`, `MatFactorError`
195: @*/
196: PetscErrorCode MatFactorGetError(Mat mat, MatFactorError *err)
197: {
198: PetscFunctionBegin;
200: PetscAssertPointer(err, 2);
201: *err = mat->factorerrortype;
202: PetscFunctionReturn(PETSC_SUCCESS);
203: }
205: /*@
206: MatFactorClearError - clears the error code in a factorization
208: Logically Collective
210: Input Parameter:
211: . mat - the factored matrix
213: Level: developer
215: Note:
216: This can also be called on non-factored matrices that come from, for example, matrices used in SOR.
218: .seealso: [](ch_matrices), `Mat`, `MatZeroEntries()`, `MatFactor()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`, `MatFactorGetError()`, `MatFactorGetErrorZeroPivot()`,
219: `MatGetErrorCode()`, `MatFactorError`
220: @*/
221: PetscErrorCode MatFactorClearError(Mat mat)
222: {
223: PetscFunctionBegin;
225: mat->factorerrortype = MAT_FACTOR_NOERROR;
226: mat->factorerror_zeropivot_value = 0.0;
227: mat->factorerror_zeropivot_row = 0;
228: PetscFunctionReturn(PETSC_SUCCESS);
229: }
231: PetscErrorCode MatFindNonzeroRowsOrCols_Basic(Mat mat, PetscBool cols, PetscReal tol, IS *nonzero)
232: {
233: Vec r, l;
234: const PetscScalar *al;
235: PetscInt i, nz, gnz, N, n, st;
237: PetscFunctionBegin;
238: PetscCall(MatCreateVecs(mat, &r, &l));
239: if (!cols) { /* nonzero rows */
240: PetscCall(MatGetOwnershipRange(mat, &st, NULL));
241: PetscCall(MatGetSize(mat, &N, NULL));
242: PetscCall(MatGetLocalSize(mat, &n, NULL));
243: PetscCall(VecSet(l, 0.0));
244: PetscCall(VecSetRandom(r, NULL));
245: PetscCall(MatMult(mat, r, l));
246: PetscCall(VecGetArrayRead(l, &al));
247: } else { /* nonzero columns */
248: PetscCall(MatGetOwnershipRangeColumn(mat, &st, NULL));
249: PetscCall(MatGetSize(mat, NULL, &N));
250: PetscCall(MatGetLocalSize(mat, NULL, &n));
251: PetscCall(VecSet(r, 0.0));
252: PetscCall(VecSetRandom(l, NULL));
253: PetscCall(MatMultTranspose(mat, l, r));
254: PetscCall(VecGetArrayRead(r, &al));
255: }
256: if (tol <= 0.0) {
257: for (i = 0, nz = 0; i < n; i++)
258: if (al[i] != 0.0) nz++;
259: } else {
260: for (i = 0, nz = 0; i < n; i++)
261: if (PetscAbsScalar(al[i]) > tol) nz++;
262: }
263: PetscCallMPI(MPIU_Allreduce(&nz, &gnz, 1, MPIU_INT, MPI_SUM, PetscObjectComm((PetscObject)mat)));
264: if (gnz != N) {
265: PetscInt *nzr;
266: PetscCall(PetscMalloc1(nz, &nzr));
267: if (nz) {
268: if (tol < 0) {
269: for (i = 0, nz = 0; i < n; i++)
270: if (al[i] != 0.0) nzr[nz++] = i + st;
271: } else {
272: for (i = 0, nz = 0; i < n; i++)
273: if (PetscAbsScalar(al[i]) > tol) nzr[nz++] = i + st;
274: }
275: }
276: PetscCall(ISCreateGeneral(PetscObjectComm((PetscObject)mat), nz, nzr, PETSC_OWN_POINTER, nonzero));
277: } else *nonzero = NULL;
278: if (!cols) { /* nonzero rows */
279: PetscCall(VecRestoreArrayRead(l, &al));
280: } else {
281: PetscCall(VecRestoreArrayRead(r, &al));
282: }
283: PetscCall(VecDestroy(&l));
284: PetscCall(VecDestroy(&r));
285: PetscFunctionReturn(PETSC_SUCCESS);
286: }
288: /*@
289: MatFindNonzeroRows - Locate all rows that are not completely zero in the matrix
291: Input Parameter:
292: . mat - the matrix
294: Output Parameter:
295: . keptrows - the rows that are not completely zero
297: Level: intermediate
299: Note:
300: `keptrows` is set to `NULL` if all rows are nonzero.
302: Developer Note:
303: If `keptrows` is not `NULL`, it must be sorted.
305: .seealso: [](ch_matrices), `Mat`, `MatFindZeroRows()`
306: @*/
307: PetscErrorCode MatFindNonzeroRows(Mat mat, IS *keptrows)
308: {
309: PetscFunctionBegin;
312: PetscAssertPointer(keptrows, 2);
313: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
314: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
315: if (mat->ops->findnonzerorows) PetscUseTypeMethod(mat, findnonzerorows, keptrows);
316: else PetscCall(MatFindNonzeroRowsOrCols_Basic(mat, PETSC_FALSE, 0.0, keptrows));
317: if (keptrows && *keptrows) PetscCall(ISSetInfo(*keptrows, IS_SORTED, IS_GLOBAL, PETSC_FALSE, PETSC_TRUE));
318: PetscFunctionReturn(PETSC_SUCCESS);
319: }
321: /*@
322: MatFindZeroRows - Locate all rows that are completely zero in the matrix
324: Input Parameter:
325: . mat - the matrix
327: Output Parameter:
328: . zerorows - the rows that are completely zero
330: Level: intermediate
332: Note:
333: `zerorows` is set to `NULL` if no rows are zero.
335: .seealso: [](ch_matrices), `Mat`, `MatFindNonzeroRows()`
336: @*/
337: PetscErrorCode MatFindZeroRows(Mat mat, IS *zerorows)
338: {
339: IS keptrows;
340: PetscInt m, n;
342: PetscFunctionBegin;
345: PetscAssertPointer(zerorows, 2);
346: PetscCall(MatFindNonzeroRows(mat, &keptrows));
347: /* MatFindNonzeroRows sets keptrows to NULL if there are no zero rows.
348: In keeping with this convention, we set zerorows to NULL if there are no zero
349: rows. */
350: if (keptrows == NULL) {
351: *zerorows = NULL;
352: } else {
353: PetscCall(MatGetOwnershipRange(mat, &m, &n));
354: PetscCall(ISComplement(keptrows, m, n, zerorows));
355: PetscCall(ISDestroy(&keptrows));
356: }
357: PetscFunctionReturn(PETSC_SUCCESS);
358: }
360: /*@
361: MatGetDiagonalBlock - Returns the part of the matrix associated with the on-process coupling
363: Not Collective
365: Input Parameter:
366: . A - the matrix
368: Output Parameter:
369: . a - the diagonal part (which is a SEQUENTIAL matrix)
371: Level: advanced
373: Notes:
374: See `MatCreateAIJ()` for more information on the "diagonal part" of the matrix.
376: Use caution, as the reference count on the returned matrix is not incremented and it is used as part of `A`'s normal operation.
378: .seealso: [](ch_matrices), `Mat`, `MatCreateAIJ()`, `MATAIJ`, `MATBAIJ`, `MATSBAIJ`
379: @*/
380: PetscErrorCode MatGetDiagonalBlock(Mat A, Mat *a)
381: {
382: PetscFunctionBegin;
385: PetscAssertPointer(a, 2);
386: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
387: if (A->ops->getdiagonalblock) PetscUseTypeMethod(A, getdiagonalblock, a);
388: else {
389: PetscMPIInt size;
391: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)A), &size));
392: PetscCheck(size == 1, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Not for parallel matrix type %s", ((PetscObject)A)->type_name);
393: *a = A;
394: }
395: PetscFunctionReturn(PETSC_SUCCESS);
396: }
398: /*@
399: MatGetTrace - Gets the trace of a matrix. The sum of the diagonal entries.
401: Collective
403: Input Parameter:
404: . mat - the matrix
406: Output Parameter:
407: . trace - the sum of the diagonal entries
409: Level: advanced
411: .seealso: [](ch_matrices), `Mat`
412: @*/
413: PetscErrorCode MatGetTrace(Mat mat, PetscScalar *trace)
414: {
415: Vec diag;
417: PetscFunctionBegin;
419: PetscAssertPointer(trace, 2);
420: PetscCall(MatCreateVecs(mat, &diag, NULL));
421: PetscCall(MatGetDiagonal(mat, diag));
422: PetscCall(VecSum(diag, trace));
423: PetscCall(VecDestroy(&diag));
424: PetscFunctionReturn(PETSC_SUCCESS);
425: }
427: /*@
428: MatRealPart - Zeros out the imaginary part of the matrix
430: Logically Collective
432: Input Parameter:
433: . mat - the matrix
435: Level: advanced
437: .seealso: [](ch_matrices), `Mat`, `MatImaginaryPart()`
438: @*/
439: PetscErrorCode MatRealPart(Mat mat)
440: {
441: PetscFunctionBegin;
444: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
445: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
446: MatCheckPreallocated(mat, 1);
447: PetscUseTypeMethod(mat, realpart);
448: PetscFunctionReturn(PETSC_SUCCESS);
449: }
451: /*@C
452: MatGetGhosts - Get the global indices of all ghost nodes defined by the sparse matrix
454: Collective
456: Input Parameter:
457: . mat - the matrix
459: Output Parameters:
460: + nghosts - number of ghosts (for `MATBAIJ` and `MATSBAIJ` matrices there is one ghost for each matrix block)
461: - ghosts - the global indices of the ghost points
463: Level: advanced
465: Note:
466: `nghosts` and `ghosts` are suitable to pass into `VecCreateGhost()` or `VecCreateGhostBlock()`
468: .seealso: [](ch_matrices), `Mat`, `VecCreateGhost()`, `VecCreateGhostBlock()`
469: @*/
470: PetscErrorCode MatGetGhosts(Mat mat, PetscInt *nghosts, const PetscInt *ghosts[])
471: {
472: PetscFunctionBegin;
475: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
476: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
477: if (mat->ops->getghosts) PetscUseTypeMethod(mat, getghosts, nghosts, ghosts);
478: else {
479: if (nghosts) *nghosts = 0;
480: if (ghosts) *ghosts = NULL;
481: }
482: PetscFunctionReturn(PETSC_SUCCESS);
483: }
485: /*@
486: MatImaginaryPart - Moves the imaginary part of the matrix to the real part and zeros the imaginary part
488: Logically Collective
490: Input Parameter:
491: . mat - the matrix
493: Level: advanced
495: .seealso: [](ch_matrices), `Mat`, `MatRealPart()`
496: @*/
497: PetscErrorCode MatImaginaryPart(Mat mat)
498: {
499: PetscFunctionBegin;
502: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
503: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
504: MatCheckPreallocated(mat, 1);
505: PetscUseTypeMethod(mat, imaginarypart);
506: PetscFunctionReturn(PETSC_SUCCESS);
507: }
509: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
510: /*@C
511: MatGetRow - Gets a row of a matrix. You MUST call `MatRestoreRow()`
512: for each row that you get to ensure that your application does
513: not bleed memory.
515: Not Collective
517: Input Parameters:
518: + mat - the matrix
519: - row - the row to get
521: Output Parameters:
522: + ncols - if not `NULL`, the number of nonzeros in `row`
523: . cols - if not `NULL`, the column numbers
524: - vals - if not `NULL`, the numerical values
526: Level: advanced
528: Notes:
529: This routine is provided for people who need to have direct access
530: to the structure of a matrix. We hope that we provide enough
531: high-level matrix routines that few users will need it.
533: `MatGetRow()` always returns 0-based column indices, regardless of
534: whether the internal representation is 0-based (default) or 1-based.
536: For better efficiency, set `cols` and/or `vals` to `NULL` if you do
537: not wish to extract these quantities.
539: The user can only examine the values extracted with `MatGetRow()`;
540: the values CANNOT be altered. To change the matrix entries, one
541: must use `MatSetValues()`.
543: You can only have one call to `MatGetRow()` outstanding for a particular
544: matrix at a time, per processor. `MatGetRow()` can only obtain rows
545: associated with the given processor, it cannot get rows from the
546: other processors; for that we suggest using `MatCreateSubMatrices()`, then
547: `MatGetRow()` on the submatrix. The row index passed to `MatGetRow()`
548: is in the global number of rows.
550: Use `MatGetRowIJ()` and `MatRestoreRowIJ()` to access all the local indices of the sparse matrix.
552: Use `MatSeqAIJGetArray()` and similar functions to access the numerical values for certain matrix types directly.
554: Fortran Note:
555: .vb
556: PetscInt, pointer :: cols(:)
557: PetscScalar, pointer :: vals(:)
558: .ve
560: .seealso: [](ch_matrices), `Mat`, `MatRestoreRow()`, `MatSetValues()`, `MatGetValues()`, `MatCreateSubMatrices()`, `MatGetDiagonal()`, `MatGetRowIJ()`, `MatRestoreRowIJ()`
561: @*/
562: PetscErrorCode MatGetRow(Mat mat, PetscInt row, PetscInt *ncols, const PetscInt *cols[], const PetscScalar *vals[])
563: {
564: PetscInt incols;
566: PetscFunctionBegin;
569: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
570: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
571: MatCheckPreallocated(mat, 1);
572: PetscCheck(row >= mat->rmap->rstart && row < mat->rmap->rend, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Only for local rows, %" PetscInt_FMT " not in [%" PetscInt_FMT ",%" PetscInt_FMT ")", row, mat->rmap->rstart, mat->rmap->rend);
573: PetscCall(PetscLogEventBegin(MAT_GetRow, mat, 0, 0, 0));
574: PetscUseTypeMethod(mat, getrow, row, &incols, (PetscInt **)cols, (PetscScalar **)vals);
575: if (ncols) *ncols = incols;
576: PetscCall(PetscLogEventEnd(MAT_GetRow, mat, 0, 0, 0));
577: PetscFunctionReturn(PETSC_SUCCESS);
578: }
580: /*@
581: MatConjugate - replaces the matrix values with their complex conjugates
583: Logically Collective
585: Input Parameter:
586: . mat - the matrix
588: Level: advanced
590: .seealso: [](ch_matrices), `Mat`, `MatRealPart()`, `MatImaginaryPart()`, `VecConjugate()`, `MatTranspose()`
591: @*/
592: PetscErrorCode MatConjugate(Mat mat)
593: {
594: PetscFunctionBegin;
596: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
597: if (PetscDefined(USE_COMPLEX) && !(mat->symmetric == PETSC_BOOL3_TRUE && mat->hermitian == PETSC_BOOL3_TRUE)) {
598: PetscUseTypeMethod(mat, conjugate);
599: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
600: }
601: PetscFunctionReturn(PETSC_SUCCESS);
602: }
604: /*@C
605: MatRestoreRow - Frees any temporary space allocated by `MatGetRow()`.
607: Not Collective
609: Input Parameters:
610: + mat - the matrix
611: . row - the row to get
612: . ncols - the number of nonzeros
613: . cols - the columns of the nonzeros
614: - vals - if nonzero the column values
616: Level: advanced
618: Notes:
619: This routine should be called after you have finished examining the entries.
621: This routine zeros out `ncols`, `cols`, and `vals`. This is to prevent accidental
622: us of the array after it has been restored. If you pass `NULL`, it will
623: not zero the pointers. Use of `cols` or `vals` after `MatRestoreRow()` is invalid.
625: Fortran Note:
626: .vb
627: PetscInt, pointer :: cols(:)
628: PetscScalar, pointer :: vals(:)
629: .ve
631: .seealso: [](ch_matrices), `Mat`, `MatGetRow()`
632: @*/
633: PetscErrorCode MatRestoreRow(Mat mat, PetscInt row, PetscInt *ncols, const PetscInt *cols[], const PetscScalar *vals[])
634: {
635: PetscFunctionBegin;
637: if (ncols) PetscAssertPointer(ncols, 3);
638: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
639: PetscTryTypeMethod(mat, restorerow, row, ncols, (PetscInt **)cols, (PetscScalar **)vals);
640: if (ncols) *ncols = 0;
641: if (cols) *cols = NULL;
642: if (vals) *vals = NULL;
643: PetscFunctionReturn(PETSC_SUCCESS);
644: }
646: /*@
647: MatGetRowUpperTriangular - Sets a flag to enable calls to `MatGetRow()` for matrix in `MATSBAIJ` format.
648: You should call `MatRestoreRowUpperTriangular()` after calling` MatGetRow()` and `MatRestoreRow()` to disable the flag.
650: Not Collective
652: Input Parameter:
653: . mat - the matrix
655: Level: advanced
657: Note:
658: The flag is to ensure that users are aware that `MatGetRow()` only provides the upper triangular part of the row for the matrices in `MATSBAIJ` format.
660: .seealso: [](ch_matrices), `Mat`, `MATSBAIJ`, `MatRestoreRowUpperTriangular()`
661: @*/
662: PetscErrorCode MatGetRowUpperTriangular(Mat mat)
663: {
664: PetscFunctionBegin;
667: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
668: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
669: MatCheckPreallocated(mat, 1);
670: PetscTryTypeMethod(mat, getrowuppertriangular);
671: PetscFunctionReturn(PETSC_SUCCESS);
672: }
674: /*@
675: MatRestoreRowUpperTriangular - Disable calls to `MatGetRow()` for matrix in `MATSBAIJ` format.
677: Not Collective
679: Input Parameter:
680: . mat - the matrix
682: Level: advanced
684: Note:
685: This routine should be called after you have finished calls to `MatGetRow()` and `MatRestoreRow()`.
687: .seealso: [](ch_matrices), `Mat`, `MATSBAIJ`, `MatGetRowUpperTriangular()`
688: @*/
689: PetscErrorCode MatRestoreRowUpperTriangular(Mat mat)
690: {
691: PetscFunctionBegin;
694: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
695: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
696: MatCheckPreallocated(mat, 1);
697: PetscTryTypeMethod(mat, restorerowuppertriangular);
698: PetscFunctionReturn(PETSC_SUCCESS);
699: }
701: /*@
702: MatSetOptionsPrefix - Sets the prefix used for searching for all
703: `Mat` options in the database.
705: Logically Collective
707: Input Parameters:
708: + A - the matrix
709: - prefix - the prefix to prepend to all option names
711: Level: advanced
713: Notes:
714: A hyphen (-) must NOT be given at the beginning of the prefix name.
715: The first character of all runtime options is AUTOMATICALLY the hyphen.
717: This is NOT used for options for the factorization of the matrix. Normally the
718: prefix is automatically passed in from the PC calling the factorization. To set
719: it directly use `MatSetOptionsPrefixFactor()`
721: .seealso: [](ch_matrices), `Mat`, `MatSetFromOptions()`, `MatSetOptionsPrefixFactor()`
722: @*/
723: PetscErrorCode MatSetOptionsPrefix(Mat A, const char prefix[])
724: {
725: PetscFunctionBegin;
727: PetscCall(PetscObjectSetOptionsPrefix((PetscObject)A, prefix));
728: PetscTryMethod(A, "MatSetOptionsPrefix_C", (Mat, const char[]), (A, prefix));
729: PetscFunctionReturn(PETSC_SUCCESS);
730: }
732: /*@
733: MatSetOptionsPrefixFactor - Sets the prefix used for searching for all matrix factor options in the database for
734: for matrices created with `MatGetFactor()`
736: Logically Collective
738: Input Parameters:
739: + A - the matrix
740: - prefix - the prefix to prepend to all option names for the factored matrix
742: Level: developer
744: Notes:
745: A hyphen (-) must NOT be given at the beginning of the prefix name.
746: The first character of all runtime options is AUTOMATICALLY the hyphen.
748: Normally the prefix is automatically passed in from the `PC` calling the factorization. To set
749: it directly when not using `KSP`/`PC` use `MatSetOptionsPrefixFactor()`
751: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSetFromOptions()`, `MatSetOptionsPrefix()`, `MatAppendOptionsPrefixFactor()`
752: @*/
753: PetscErrorCode MatSetOptionsPrefixFactor(Mat A, const char prefix[])
754: {
755: PetscFunctionBegin;
757: if (prefix) {
758: PetscAssertPointer(prefix, 2);
759: PetscCheck(prefix[0] != '-', PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONG, "Options prefix should not begin with a hyphen");
760: if (prefix != A->factorprefix) {
761: PetscCall(PetscFree(A->factorprefix));
762: PetscCall(PetscStrallocpy(prefix, &A->factorprefix));
763: }
764: } else PetscCall(PetscFree(A->factorprefix));
765: PetscFunctionReturn(PETSC_SUCCESS);
766: }
768: /*@
769: MatAppendOptionsPrefixFactor - Appends to the prefix used for searching for all matrix factor options in the database for
770: for matrices created with `MatGetFactor()`
772: Logically Collective
774: Input Parameters:
775: + A - the matrix
776: - prefix - the prefix to prepend to all option names for the factored matrix
778: Level: developer
780: Notes:
781: A hyphen (-) must NOT be given at the beginning of the prefix name.
782: The first character of all runtime options is AUTOMATICALLY the hyphen.
784: Normally the prefix is automatically passed in from the `PC` calling the factorization. To set
785: it directly when not using `KSP`/`PC` use `MatAppendOptionsPrefixFactor()`
787: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `PetscOptionsCreate()`, `PetscOptionsDestroy()`, `PetscObjectSetOptionsPrefix()`, `PetscObjectPrependOptionsPrefix()`,
788: `PetscObjectGetOptionsPrefix()`, `TSAppendOptionsPrefix()`, `SNESAppendOptionsPrefix()`, `KSPAppendOptionsPrefix()`, `MatSetOptionsPrefixFactor()`,
789: `MatSetOptionsPrefix()`
790: @*/
791: PetscErrorCode MatAppendOptionsPrefixFactor(Mat A, const char prefix[])
792: {
793: size_t len1, len2, new_len;
795: PetscFunctionBegin;
797: if (!prefix) PetscFunctionReturn(PETSC_SUCCESS);
798: if (!A->factorprefix) {
799: PetscCall(MatSetOptionsPrefixFactor(A, prefix));
800: PetscFunctionReturn(PETSC_SUCCESS);
801: }
802: PetscCheck(prefix[0] != '-', PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONG, "Options prefix should not begin with a hyphen");
804: PetscCall(PetscStrlen(A->factorprefix, &len1));
805: PetscCall(PetscStrlen(prefix, &len2));
806: new_len = len1 + len2 + 1;
807: PetscCall(PetscRealloc(new_len * sizeof(*A->factorprefix), &A->factorprefix));
808: PetscCall(PetscStrncpy(A->factorprefix + len1, prefix, len2 + 1));
809: PetscFunctionReturn(PETSC_SUCCESS);
810: }
812: /*@
813: MatAppendOptionsPrefix - Appends to the prefix used for searching for all
814: matrix options in the database.
816: Logically Collective
818: Input Parameters:
819: + A - the matrix
820: - prefix - the prefix to prepend to all option names
822: Level: advanced
824: Note:
825: A hyphen (-) must NOT be given at the beginning of the prefix name.
826: The first character of all runtime options is AUTOMATICALLY the hyphen.
828: .seealso: [](ch_matrices), `Mat`, `MatGetOptionsPrefix()`, `MatAppendOptionsPrefixFactor()`, `MatSetOptionsPrefix()`
829: @*/
830: PetscErrorCode MatAppendOptionsPrefix(Mat A, const char prefix[])
831: {
832: PetscFunctionBegin;
834: PetscCall(PetscObjectAppendOptionsPrefix((PetscObject)A, prefix));
835: PetscTryMethod(A, "MatAppendOptionsPrefix_C", (Mat, const char[]), (A, prefix));
836: PetscFunctionReturn(PETSC_SUCCESS);
837: }
839: /*@
840: MatGetOptionsPrefix - Gets the prefix used for searching for all
841: matrix options in the database.
843: Not Collective
845: Input Parameter:
846: . A - the matrix
848: Output Parameter:
849: . prefix - pointer to the prefix string used
851: Level: advanced
853: .seealso: [](ch_matrices), `Mat`, `MatAppendOptionsPrefix()`, `MatSetOptionsPrefix()`, `MatAppendOptionsPrefixFactor()`, `MatSetOptionsPrefixFactor()`
854: @*/
855: PetscErrorCode MatGetOptionsPrefix(Mat A, const char *prefix[])
856: {
857: PetscFunctionBegin;
859: PetscAssertPointer(prefix, 2);
860: PetscCall(PetscObjectGetOptionsPrefix((PetscObject)A, prefix));
861: PetscFunctionReturn(PETSC_SUCCESS);
862: }
864: /*@
865: MatGetState - Gets the state of a `Mat`. Same value as returned by `PetscObjectStateGet()`
867: Not Collective
869: Input Parameter:
870: . A - the matrix
872: Output Parameter:
873: . state - the object state
875: Level: advanced
877: Note:
878: Object state is an integer which gets increased every time
879: the object is changed. By saving and later querying the object state
880: one can determine whether information about the object is still current.
882: See `MatGetNonzeroState()` to determine if the nonzero structure of the matrix has changed.
884: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `PetscObjectStateGet()`, `MatGetNonzeroState()`
885: @*/
886: PetscErrorCode MatGetState(Mat A, PetscObjectState *state)
887: {
888: PetscFunctionBegin;
890: PetscAssertPointer(state, 2);
891: PetscCall(PetscObjectStateGet((PetscObject)A, state));
892: PetscFunctionReturn(PETSC_SUCCESS);
893: }
895: /*@
896: MatResetPreallocation - Reset matrix to use the original preallocation values provided by the user, for example with `MatXAIJSetPreallocation()`
898: Collective
900: Input Parameter:
901: . A - the matrix
903: Level: beginner
905: Notes:
906: After calling `MatAssemblyBegin()` and `MatAssemblyEnd()` with `MAT_FINAL_ASSEMBLY` the matrix data structures represent the nonzeros assigned to the
907: matrix. If that space is less than the preallocated space that extra preallocated space is no longer available to take on new values. `MatResetPreallocation()`
908: makes all of the preallocation space available
910: Current values in the matrix are lost in this call
912: Currently only supported for `MATAIJ` matrices.
914: .seealso: [](ch_matrices), `Mat`, `MatSeqAIJSetPreallocation()`, `MatMPIAIJSetPreallocation()`, `MatXAIJSetPreallocation()`
915: @*/
916: PetscErrorCode MatResetPreallocation(Mat A)
917: {
918: PetscFunctionBegin;
921: PetscUseMethod(A, "MatResetPreallocation_C", (Mat), (A));
922: PetscFunctionReturn(PETSC_SUCCESS);
923: }
925: /*@
926: MatResetHash - Reset the matrix so that it will use a hash table for the next round of `MatSetValues()` and `MatAssemblyBegin()`/`MatAssemblyEnd()`.
928: Collective
930: Input Parameter:
931: . A - the matrix
933: Level: intermediate
935: Notes:
936: The matrix will again delete the hash table data structures after following calls to `MatAssemblyBegin()`/`MatAssemblyEnd()` with `MAT_FINAL_ASSEMBLY`.
938: Currently only supported for `MATAIJ` matrices.
940: .seealso: [](ch_matrices), `Mat`, `MatResetPreallocation()`
941: @*/
942: PetscErrorCode MatResetHash(Mat A)
943: {
944: PetscFunctionBegin;
947: PetscCheck(A->insertmode == NOT_SET_VALUES, PETSC_COMM_SELF, PETSC_ERR_SUP, "Cannot reset to hash state after setting some values but not yet calling MatAssemblyBegin()/MatAssemblyEnd()");
948: if (A->num_ass == 0) PetscFunctionReturn(PETSC_SUCCESS);
949: PetscUseMethod(A, "MatResetHash_C", (Mat), (A));
950: /* These flags are used to determine whether certain setups occur */
951: A->was_assembled = PETSC_FALSE;
952: A->assembled = PETSC_FALSE;
953: /* Log that the state of this object has changed; this will help guarantee that preconditioners get re-setup */
954: PetscCall(PetscObjectStateIncrease((PetscObject)A));
955: PetscFunctionReturn(PETSC_SUCCESS);
956: }
958: /*@
959: MatSetUp - Sets up the internal matrix data structures for later use by the matrix
961: Collective
963: Input Parameter:
964: . A - the matrix
966: Level: advanced
968: Notes:
969: If the user has not set preallocation for this matrix then an efficient algorithm will be used for the first round of
970: setting values in the matrix.
972: This routine is called internally by other `Mat` functions when needed so rarely needs to be called by users
974: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatCreate()`, `MatDestroy()`, `MatXAIJSetPreallocation()`
975: @*/
976: PetscErrorCode MatSetUp(Mat A)
977: {
978: PetscFunctionBegin;
980: if (!((PetscObject)A)->type_name) {
981: PetscMPIInt size;
983: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)A), &size));
984: PetscCall(MatSetType(A, size == 1 ? MATSEQAIJ : MATMPIAIJ));
985: }
986: if (!A->preallocated) PetscTryTypeMethod(A, setup);
987: PetscCall(PetscLayoutSetUp(A->rmap));
988: PetscCall(PetscLayoutSetUp(A->cmap));
989: A->preallocated = PETSC_TRUE;
990: PetscFunctionReturn(PETSC_SUCCESS);
991: }
993: #if defined(PETSC_HAVE_SAWS)
994: #include <petscviewersaws.h>
995: #endif
997: /*
998: If threadsafety is on extraneous matrices may be printed
1000: This flag cannot be stored in the matrix because the original matrix in MatView() may assemble a new matrix which is passed into MatViewFromOptions()
1001: */
1002: #if !defined(PETSC_HAVE_THREADSAFETY)
1003: static PetscInt insidematview = 0;
1004: #endif
1006: /*@
1007: MatViewFromOptions - View properties of the matrix based on options set in the options database
1009: Collective
1011: Input Parameters:
1012: + A - the matrix
1013: . obj - optional additional object that provides the options prefix to use
1014: - name - command line option
1016: Options Database Key:
1017: . -name [viewertype][:...] - option name and values. See `PetscObjectViewFromOptions()` for the possible arguments
1019: Level: intermediate
1021: .seealso: [](ch_matrices), `Mat`, `MatView()`, `PetscObjectViewFromOptions()`, `MatCreate()`
1022: @*/
1023: PetscErrorCode MatViewFromOptions(Mat A, PetscObject obj, const char name[])
1024: {
1025: PetscFunctionBegin;
1027: #if !defined(PETSC_HAVE_THREADSAFETY)
1028: if (insidematview) PetscFunctionReturn(PETSC_SUCCESS);
1029: #endif
1030: PetscCall(PetscObjectViewFromOptions((PetscObject)A, obj, name));
1031: PetscFunctionReturn(PETSC_SUCCESS);
1032: }
1034: /*@
1035: MatView - display information about a matrix in a variety ways
1037: Collective on viewer
1039: Input Parameters:
1040: + mat - the matrix
1041: - viewer - visualization context
1043: Options Database Keys:
1044: + -mat_view ::ascii_info - Prints info on matrix at conclusion of `MatAssemblyEnd()`
1045: . -mat_view ::ascii_info_detail - Prints more detailed info
1046: . -mat_view - Prints matrix in ASCII format
1047: . -mat_view ::ascii_matlab - Prints matrix in MATLAB format
1048: . -mat_view draw - PetscDraws nonzero structure of matrix, using `MatView()` and `PetscDrawOpenX()`.
1049: . -display name - Sets display name (default is host)
1050: . -draw_pause sec - Sets number of seconds to pause after display
1051: . -mat_view socket - Sends matrix to socket, can be accessed from MATLAB (see Users-Manual: ch_matlab for details)
1052: . -viewer_socket_machine machine - -
1053: . -viewer_socket_port port - -
1054: . -mat_view binary - save matrix to file in binary format
1055: - -viewer_binary_filename name - -
1057: Level: beginner
1059: Notes:
1060: The available visualization contexts include
1061: + `PETSC_VIEWER_STDOUT_SELF` - for sequential matrices
1062: . `PETSC_VIEWER_STDOUT_WORLD` - for parallel matrices created on `PETSC_COMM_WORLD`
1063: . `PETSC_VIEWER_STDOUT_`(comm) - for matrices created on MPI communicator comm
1064: - `PETSC_VIEWER_DRAW_WORLD` - graphical display of nonzero structure
1066: The user can open alternative visualization contexts with
1067: + `PetscViewerASCIIOpen()` - Outputs matrix to a specified file
1068: . `PetscViewerBinaryOpen()` - Outputs matrix in binary to a specified file; corresponding input uses `MatLoad()`
1069: . `PetscViewerDrawOpen()` - Outputs nonzero matrix nonzero structure to an X window display
1070: - `PetscViewerSocketOpen()` - Outputs matrix to Socket viewer, `PETSCVIEWERSOCKET`. Only the `MATSEQDENSE` and `MATAIJ` types support this viewer.
1072: The user can call `PetscViewerPushFormat()` to specify the output
1073: format of ASCII printed objects (when using `PETSC_VIEWER_STDOUT_SELF`,
1074: `PETSC_VIEWER_STDOUT_WORLD` and `PetscViewerASCIIOpen()`). Available formats include
1075: + `PETSC_VIEWER_DEFAULT` - default, prints matrix contents
1076: . `PETSC_VIEWER_ASCII_MATLAB` - prints matrix contents in MATLAB format
1077: . `PETSC_VIEWER_ASCII_DENSE` - prints entire matrix including zeros
1078: . `PETSC_VIEWER_ASCII_COMMON` - prints matrix contents, using a sparse format common among all matrix types
1079: . `PETSC_VIEWER_ASCII_IMPL` - prints matrix contents, using an implementation-specific format (which is in many cases the same as the default)
1080: . `PETSC_VIEWER_ASCII_INFO` - prints basic information about the matrix size and structure (not the matrix entries)
1081: - `PETSC_VIEWER_ASCII_INFO_DETAIL` - prints more detailed information about the matrix nonzero structure (still not vector or matrix entries)
1083: The ASCII viewers are only recommended for small matrices on at most a moderate number of processes,
1084: the program will seemingly hang and take hours for larger matrices, for larger matrices one should use the binary format.
1086: In the debugger you can do "call MatView(mat,0)" to display the matrix. (The same holds for any PETSc object viewer).
1088: See the manual page for `MatLoad()` for the exact format of the binary file when the binary
1089: viewer is used.
1091: See share/petsc/matlab/PetscBinaryRead.m for a MATLAB code that can read in the binary file when the binary
1092: viewer is used and lib/petsc/bin/PetscBinaryIO.py for loading them into Python.
1094: One can use `-mat_view draw -draw_pause -1` to pause the graphical display of matrix nonzero structure,
1095: and then use the following mouse functions.
1096: .vb
1097: left mouse: zoom in
1098: middle mouse: zoom out
1099: right mouse: continue with the simulation
1100: .ve
1102: .seealso: [](ch_matrices), `Mat`, `PetscViewerPushFormat()`, `PetscViewerASCIIOpen()`, `PetscViewerDrawOpen()`, `PetscViewer`,
1103: `PetscViewerSocketOpen()`, `PetscViewerBinaryOpen()`, `MatLoad()`, `MatViewFromOptions()`
1104: @*/
1105: PetscErrorCode MatView(Mat mat, PetscViewer viewer)
1106: {
1107: PetscInt rows, cols, rbs, cbs;
1108: PetscBool isascii, isstring, issaws;
1109: PetscViewerFormat format;
1110: PetscMPIInt size;
1112: PetscFunctionBegin;
1115: if (!viewer) PetscCall(PetscViewerASCIIGetStdout(PetscObjectComm((PetscObject)mat), &viewer));
1118: PetscCall(PetscViewerGetFormat(viewer, &format));
1119: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)viewer), &size));
1120: if (size == 1 && format == PETSC_VIEWER_LOAD_BALANCE) PetscFunctionReturn(PETSC_SUCCESS);
1122: #if !defined(PETSC_HAVE_THREADSAFETY)
1123: insidematview++;
1124: #endif
1125: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERSTRING, &isstring));
1126: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &isascii));
1127: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERSAWS, &issaws));
1128: PetscCheck((isascii && (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL)) || !mat->factortype, PetscObjectComm((PetscObject)viewer), PETSC_ERR_ARG_WRONGSTATE, "No viewers for factored matrix except ASCII, info, or info_detail");
1130: PetscCall(PetscLogEventBegin(MAT_View, mat, viewer, 0, 0));
1131: if (isascii) {
1132: if (!mat->preallocated) {
1133: PetscCall(PetscViewerASCIIPrintf(viewer, "Matrix has not been preallocated yet\n"));
1134: #if !defined(PETSC_HAVE_THREADSAFETY)
1135: insidematview--;
1136: #endif
1137: PetscCall(PetscLogEventEnd(MAT_View, mat, viewer, 0, 0));
1138: PetscFunctionReturn(PETSC_SUCCESS);
1139: }
1140: if (!mat->assembled) {
1141: PetscCall(PetscViewerASCIIPrintf(viewer, "Matrix has not been assembled yet\n"));
1142: #if !defined(PETSC_HAVE_THREADSAFETY)
1143: insidematview--;
1144: #endif
1145: PetscCall(PetscLogEventEnd(MAT_View, mat, viewer, 0, 0));
1146: PetscFunctionReturn(PETSC_SUCCESS);
1147: }
1148: PetscCall(PetscObjectPrintClassNamePrefixType((PetscObject)mat, viewer));
1149: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
1150: MatNullSpace nullsp, transnullsp;
1152: PetscCall(PetscViewerASCIIPushTab(viewer));
1153: PetscCall(MatGetSize(mat, &rows, &cols));
1154: PetscCall(MatGetBlockSizes(mat, &rbs, &cbs));
1155: if (rbs != 1 || cbs != 1) {
1156: if (rbs != cbs) PetscCall(PetscViewerASCIIPrintf(viewer, "rows=%" PetscInt_FMT ", cols=%" PetscInt_FMT ", rbs=%" PetscInt_FMT ", cbs=%" PetscInt_FMT "%s\n", rows, cols, rbs, cbs, mat->bsizes ? " variable blocks set" : ""));
1157: else PetscCall(PetscViewerASCIIPrintf(viewer, "rows=%" PetscInt_FMT ", cols=%" PetscInt_FMT ", bs=%" PetscInt_FMT "%s\n", rows, cols, rbs, mat->bsizes ? " variable blocks set" : ""));
1158: } else PetscCall(PetscViewerASCIIPrintf(viewer, "rows=%" PetscInt_FMT ", cols=%" PetscInt_FMT "\n", rows, cols));
1159: if (mat->factortype) {
1160: MatSolverType solver;
1161: PetscCall(MatFactorGetSolverType(mat, &solver));
1162: PetscCall(PetscViewerASCIIPrintf(viewer, "package used to perform factorization: %s\n", solver));
1163: }
1164: if (mat->ops->getinfo) {
1165: PetscBool is_constant_or_diagonal;
1167: // Don't print nonzero information for constant or diagonal matrices, it just adds noise to the output
1168: PetscCall(PetscObjectTypeCompareAny((PetscObject)mat, &is_constant_or_diagonal, MATCONSTANTDIAGONAL, MATDIAGONAL, ""));
1169: if (!is_constant_or_diagonal) {
1170: MatInfo info;
1172: PetscCall(MatGetInfo(mat, MAT_GLOBAL_SUM, &info));
1173: PetscCall(PetscViewerASCIIPrintf(viewer, "total: nonzeros=%.f, allocated nonzeros=%.f\n", info.nz_used, info.nz_allocated));
1174: if (!mat->factortype) PetscCall(PetscViewerASCIIPrintf(viewer, "total number of mallocs used during MatSetValues calls=%" PetscInt_FMT "\n", (PetscInt)info.mallocs));
1175: }
1176: }
1177: PetscCall(MatGetNullSpace(mat, &nullsp));
1178: PetscCall(MatGetTransposeNullSpace(mat, &transnullsp));
1179: if (nullsp) PetscCall(PetscViewerASCIIPrintf(viewer, " has attached null space\n"));
1180: if (transnullsp && transnullsp != nullsp) PetscCall(PetscViewerASCIIPrintf(viewer, " has attached transposed null space\n"));
1181: PetscCall(MatGetNearNullSpace(mat, &nullsp));
1182: if (nullsp) PetscCall(PetscViewerASCIIPrintf(viewer, " has attached near null space\n"));
1183: PetscCall(PetscViewerASCIIPushTab(viewer));
1184: PetscCall(MatProductView(mat, viewer));
1185: PetscCall(PetscViewerASCIIPopTab(viewer));
1186: if (mat->bsizes && format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
1187: IS tmp;
1189: PetscCall(ISCreateGeneral(PetscObjectComm((PetscObject)viewer), mat->nblocks, mat->bsizes, PETSC_USE_POINTER, &tmp));
1190: PetscCall(PetscObjectSetName((PetscObject)tmp, "Block Sizes"));
1191: PetscCall(PetscViewerASCIIPushTab(viewer));
1192: PetscCall(ISView(tmp, viewer));
1193: PetscCall(PetscViewerASCIIPopTab(viewer));
1194: PetscCall(ISDestroy(&tmp));
1195: }
1196: }
1197: } else if (issaws) {
1198: #if defined(PETSC_HAVE_SAWS)
1199: PetscMPIInt rank;
1201: PetscCall(PetscObjectName((PetscObject)mat));
1202: PetscCallMPI(MPI_Comm_rank(PETSC_COMM_WORLD, &rank));
1203: if (!((PetscObject)mat)->amsmem && rank == 0) PetscCall(PetscObjectViewSAWs((PetscObject)mat, viewer));
1204: #endif
1205: } else if (isstring) {
1206: const char *type;
1207: PetscCall(MatGetType(mat, &type));
1208: PetscCall(PetscViewerStringSPrintf(viewer, " MatType: %-7.7s", type));
1209: PetscTryTypeMethod(mat, view, viewer);
1210: }
1211: if ((format == PETSC_VIEWER_NATIVE || format == PETSC_VIEWER_LOAD_BALANCE) && mat->ops->viewnative) {
1212: PetscCall(PetscViewerASCIIPushTab(viewer));
1213: PetscUseTypeMethod(mat, viewnative, viewer);
1214: PetscCall(PetscViewerASCIIPopTab(viewer));
1215: } else if (mat->ops->view) {
1216: PetscCall(PetscViewerASCIIPushTab(viewer));
1217: PetscUseTypeMethod(mat, view, viewer);
1218: PetscCall(PetscViewerASCIIPopTab(viewer));
1219: }
1220: if (isascii) {
1221: PetscCall(PetscViewerGetFormat(viewer, &format));
1222: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) PetscCall(PetscViewerASCIIPopTab(viewer));
1223: }
1224: PetscCall(PetscLogEventEnd(MAT_View, mat, viewer, 0, 0));
1225: #if !defined(PETSC_HAVE_THREADSAFETY)
1226: insidematview--;
1227: #endif
1228: PetscFunctionReturn(PETSC_SUCCESS);
1229: }
1231: #if defined(PETSC_USE_DEBUG)
1232: #include <../src/sys/totalview/tv_data_display.h>
1233: PETSC_UNUSED static int TV_display_type(const struct _p_Mat *mat)
1234: {
1235: TV_add_row("Local rows", "int", &mat->rmap->n);
1236: TV_add_row("Local columns", "int", &mat->cmap->n);
1237: TV_add_row("Global rows", "int", &mat->rmap->N);
1238: TV_add_row("Global columns", "int", &mat->cmap->N);
1239: TV_add_row("Typename", TV_ascii_string_type, ((PetscObject)mat)->type_name);
1240: return TV_format_OK;
1241: }
1242: #endif
1244: /*@
1245: MatLoad - Loads a matrix that has been stored in binary/HDF5 format
1246: with `MatView()`. The matrix format is determined from the options database.
1247: Generates a parallel MPI matrix if the communicator has more than one
1248: processor. The default matrix type is `MATAIJ`.
1250: Collective
1252: Input Parameters:
1253: + mat - the newly loaded matrix, this needs to have been created with `MatCreate()`
1254: or some related function before a call to `MatLoad()`
1255: - viewer - `PETSCVIEWERBINARY`/`PETSCVIEWERHDF5` file viewer
1257: Options Database Key:
1258: . -matload_block_size bs - set block size
1260: Level: beginner
1262: Notes:
1263: If the `Mat` type has not yet been given then `MATAIJ` is used, call `MatSetFromOptions()` on the
1264: `Mat` before calling this routine if you wish to set it from the options database.
1266: `MatLoad()` automatically loads into the options database any options
1267: given in the file filename.info where filename is the name of the file
1268: that was passed to the `PetscViewerBinaryOpen()`. The options in the info
1269: file will be ignored if you use the -viewer_binary_skip_info option.
1271: If the type or size of mat is not set before a call to `MatLoad()`, PETSc
1272: sets the default matrix type AIJ and sets the local and global sizes.
1273: If type and/or size is already set, then the same are used.
1275: In parallel, each processor can load a subset of rows (or the
1276: entire matrix). This routine is especially useful when a large
1277: matrix is stored on disk and only part of it is desired on each
1278: processor. For example, a parallel solver may access only some of
1279: the rows from each processor. The algorithm used here reads
1280: relatively small blocks of data rather than reading the entire
1281: matrix and then subsetting it.
1283: Viewer's `PetscViewerType` must be either `PETSCVIEWERBINARY` or `PETSCVIEWERHDF5`.
1284: Such viewer can be created using `PetscViewerBinaryOpen()` or `PetscViewerHDF5Open()`,
1285: or the sequence like
1286: .vb
1287: `PetscViewer` v;
1288: `PetscViewerCreate`(`PETSC_COMM_WORLD`,&v);
1289: `PetscViewerSetType`(v,`PETSCVIEWERBINARY`);
1290: `PetscViewerSetFromOptions`(v);
1291: `PetscViewerFileSetMode`(v,`FILE_MODE_READ`);
1292: `PetscViewerFileSetName`(v,"datafile");
1293: .ve
1294: The optional `PetscViewerSetFromOptions()` call allows overriding `PetscViewerSetType()` using the option
1295: .vb
1296: -viewer_type {binary, hdf5}
1297: .ve
1299: See the example src/ksp/ksp/tutorials/ex27.c with the first approach,
1300: and src/mat/tutorials/ex10.c with the second approach.
1302: In case of `PETSCVIEWERBINARY`, a native PETSc binary format is used. Each of the blocks
1303: is read onto MPI rank 0 and then shipped to its destination MPI rank, one after another.
1304: Multiple objects, both matrices and vectors, can be stored within the same file.
1305: Their `PetscObject` name is ignored; they are loaded in the order of their storage.
1307: Most users should not need to know the details of the binary storage
1308: format, since `MatLoad()` and `MatView()` completely hide these details.
1309: But for anyone who is interested, the standard binary matrix storage
1310: format is
1312: .vb
1313: PetscInt MAT_FILE_CLASSID
1314: PetscInt number of rows
1315: PetscInt number of columns
1316: PetscInt total number of nonzeros
1317: PetscInt *number nonzeros in each row
1318: PetscInt *column indices of all nonzeros (starting index is zero)
1319: PetscScalar *values of all nonzeros
1320: .ve
1321: If PETSc was not configured with `--with-64-bit-indices` then only `MATMPIAIJ` matrices with more than `PETSC_INT_MAX` non-zeros can be
1322: stored or loaded (each MPI process part of the matrix must have less than `PETSC_INT_MAX` nonzeros). Since the total nonzero count in this
1323: case will not fit in a (32-bit) `PetscInt` the value `PETSC_INT_MAX` is used for the header entry `total number of nonzeros`.
1325: PETSc automatically does the byte swapping for
1326: machines that store the bytes reversed. Thus if you write your own binary
1327: read/write routines you have to swap the bytes; see `PetscBinaryRead()`
1328: and `PetscBinaryWrite()` to see how this may be done.
1330: In case of `PETSCVIEWERHDF5`, a parallel HDF5 reader is used.
1331: Each processor's chunk is loaded independently by its owning MPI process.
1332: Multiple objects, both matrices and vectors, can be stored within the same file.
1333: They are looked up by their PetscObject name.
1335: As the MATLAB MAT-File Version 7.3 format is also a HDF5 flavor, we decided to use
1336: by default the same structure and naming of the AIJ arrays and column count
1337: within the HDF5 file. This means that a MAT file saved with -v7.3 flag, e.g.
1338: .vb
1339: save example.mat A b -v7.3
1340: .ve
1341: can be directly read by this routine (see Reference 1 for details).
1343: Depending on your MATLAB version, this format might be a default,
1344: otherwise you can set it as default in Preferences.
1346: Unless -nocompression flag is used to save the file in MATLAB,
1347: PETSc must be configured with ZLIB package.
1349: See also examples src/mat/tutorials/ex10.c and src/ksp/ksp/tutorials/ex27.c
1351: This reader currently supports only real `MATSEQAIJ`, `MATMPIAIJ`, `MATSEQDENSE` and `MATMPIDENSE` matrices for `PETSCVIEWERHDF5`
1353: Corresponding `MatView()` is not yet implemented.
1355: The loaded matrix is actually a transpose of the original one in MATLAB,
1356: unless you push `PETSC_VIEWER_HDF5_MAT` format (see examples above).
1357: With this format, matrix is automatically transposed by PETSc,
1358: unless the matrix is marked as SPD or symmetric
1359: (see `MatSetOption()`, `MAT_SPD`, `MAT_SYMMETRIC`).
1361: See MATLAB Documentation on `save()`, <https://www.mathworks.com/help/matlab/ref/save.html#btox10b-1-version>
1363: .seealso: [](ch_matrices), `Mat`, `PetscViewerBinaryOpen()`, `PetscViewerSetType()`, `MatView()`, `VecLoad()`
1364: @*/
1365: PetscErrorCode MatLoad(Mat mat, PetscViewer viewer)
1366: {
1367: PetscBool flg;
1369: PetscFunctionBegin;
1373: if (!((PetscObject)mat)->type_name) PetscCall(MatSetType(mat, MATAIJ));
1375: flg = PETSC_FALSE;
1376: PetscCall(PetscOptionsGetBool(((PetscObject)mat)->options, ((PetscObject)mat)->prefix, "-matload_symmetric", &flg, NULL));
1377: if (flg) {
1378: PetscCall(MatSetOption(mat, MAT_SYMMETRIC, PETSC_TRUE));
1379: PetscCall(MatSetOption(mat, MAT_SYMMETRY_ETERNAL, PETSC_TRUE));
1380: }
1381: flg = PETSC_FALSE;
1382: PetscCall(PetscOptionsGetBool(((PetscObject)mat)->options, ((PetscObject)mat)->prefix, "-matload_spd", &flg, NULL));
1383: if (flg) PetscCall(MatSetOption(mat, MAT_SPD, PETSC_TRUE));
1385: PetscCall(PetscLogEventBegin(MAT_Load, mat, viewer, 0, 0));
1386: PetscUseTypeMethod(mat, load, viewer);
1387: PetscCall(PetscLogEventEnd(MAT_Load, mat, viewer, 0, 0));
1388: PetscFunctionReturn(PETSC_SUCCESS);
1389: }
1391: static PetscErrorCode MatDestroy_Redundant(Mat_Redundant **redundant)
1392: {
1393: Mat_Redundant *redund = *redundant;
1395: PetscFunctionBegin;
1396: if (redund) {
1397: if (redund->matseq) { /* via MatCreateSubMatrices() */
1398: PetscCall(ISDestroy(&redund->isrow));
1399: PetscCall(ISDestroy(&redund->iscol));
1400: PetscCall(MatDestroySubMatrices(1, &redund->matseq));
1401: } else {
1402: PetscCall(PetscFree2(redund->send_rank, redund->recv_rank));
1403: PetscCall(PetscFree(redund->sbuf_j));
1404: PetscCall(PetscFree(redund->sbuf_a));
1405: for (PetscInt i = 0; i < redund->nrecvs; i++) {
1406: PetscCall(PetscFree(redund->rbuf_j[i]));
1407: PetscCall(PetscFree(redund->rbuf_a[i]));
1408: }
1409: PetscCall(PetscFree4(redund->sbuf_nz, redund->rbuf_nz, redund->rbuf_j, redund->rbuf_a));
1410: }
1412: PetscCall(PetscCommDestroy(&redund->subcomm));
1413: PetscCall(PetscFree(redund));
1414: }
1415: PetscFunctionReturn(PETSC_SUCCESS);
1416: }
1418: /*@
1419: MatDestroy - Frees space taken by a matrix.
1421: Collective
1423: Input Parameter:
1424: . A - the matrix
1426: Level: beginner
1428: Developer Note:
1429: Some special arrays of matrices are not destroyed in this routine but instead by the routines called by
1430: `MatDestroySubMatrices()`. Thus one must be sure that any changes here must also be made in those routines.
1431: `MatHeaderMerge()` and `MatHeaderReplace()` also manipulate the data in the `Mat` object and likely need changes
1432: if changes are needed here.
1434: .seealso: [](ch_matrices), `Mat`, `MatCreate()`
1435: @*/
1436: PetscErrorCode MatDestroy(Mat *A)
1437: {
1438: PetscFunctionBegin;
1439: if (!*A) PetscFunctionReturn(PETSC_SUCCESS);
1441: if (--((PetscObject)*A)->refct > 0) {
1442: *A = NULL;
1443: PetscFunctionReturn(PETSC_SUCCESS);
1444: }
1446: /* if memory was published with SAWs then destroy it */
1447: PetscCall(PetscObjectSAWsViewOff((PetscObject)*A));
1448: PetscTryTypeMethod(*A, destroy);
1450: PetscCall(PetscFree((*A)->factorprefix));
1451: PetscCall(PetscFree((*A)->defaultvectype));
1452: PetscCall(PetscFree((*A)->defaultrandtype));
1453: PetscCall(PetscFree((*A)->bsizes));
1454: PetscCall(PetscFree((*A)->solvertype));
1455: for (PetscInt i = 0; i < MAT_FACTOR_NUM_TYPES; i++) PetscCall(PetscFree((*A)->preferredordering[i]));
1456: if ((*A)->redundant && (*A)->redundant->matseq[0] == *A) (*A)->redundant->matseq[0] = NULL;
1457: PetscCall(MatDestroy_Redundant(&(*A)->redundant));
1458: PetscCall(MatProductClear(*A));
1459: PetscCall(MatNullSpaceDestroy(&(*A)->nullsp));
1460: PetscCall(MatNullSpaceDestroy(&(*A)->transnullsp));
1461: PetscCall(MatNullSpaceDestroy(&(*A)->nearnullsp));
1462: PetscCall(MatDestroy(&(*A)->schur));
1463: PetscCall(VecDestroy(&(*A)->dot_vec));
1464: PetscCall(PetscLayoutDestroy(&(*A)->rmap));
1465: PetscCall(PetscLayoutDestroy(&(*A)->cmap));
1466: PetscCall(PetscHeaderDestroy(A));
1467: PetscFunctionReturn(PETSC_SUCCESS);
1468: }
1470: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
1471: /*@
1472: MatSetValues - Inserts or adds a block of values into a matrix.
1473: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
1474: MUST be called after all calls to `MatSetValues()` have been completed.
1476: Not Collective
1478: Input Parameters:
1479: + mat - the matrix
1480: . m - the number of rows
1481: . idxm - the global indices of the rows
1482: . n - the number of columns
1483: . idxn - the global indices of the columns
1484: . v - a one-dimensional array that contains the values implicitly stored as a two-dimensional array, by default in row-major order.
1485: See `MAT_ROW_ORIENTED` in `MatSetOption()` for how to use column-major order.
1486: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
1488: Level: beginner
1490: Notes:
1491: Calls to `MatSetValues()` with the `INSERT_VALUES` and `ADD_VALUES`
1492: options cannot be mixed without intervening calls to the assembly
1493: routines.
1495: `MatSetValues()` uses 0-based row and column numbers in Fortran
1496: as well as in C.
1498: Negative indices may be passed in `idxm` and `idxn`, these rows and columns are simply ignored. This allows easily inserting element stiffness matrices
1499: with homogeneous Dirichlet boundary conditions that you don't want represented
1500: in the matrix.
1502: Efficiency Alert:
1503: The routine `MatSetValuesBlocked()` may offer much better efficiency
1504: for users of block sparse formats (`MATSEQBAIJ` and `MATMPIBAIJ`).
1506: Fortran Notes:
1507: If any of `idxm`, `idxn`, and `v` are scalars pass them using, for example,
1508: .vb
1509: call MatSetValues(mat, one, [idxm], one, [idxn], [v], INSERT_VALUES, ierr)
1510: .ve
1512: If `v` is a two-dimensional array make sure to first call `MatSetOption(mat, MAT_ROW_ORIENTED, PETSC_FALSE, ierr)` before using this function,
1513: otherwise the transpose of `v` will seemingly be inserted in the matrix, since Fortran passes two-dimensional arrays with column orientation.
1515: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1516: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`
1517: @*/
1518: PetscErrorCode MatSetValues(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], const PetscScalar v[], InsertMode addv)
1519: {
1520: PetscFunctionBeginHot;
1523: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1524: PetscAssertPointer(idxm, 3);
1525: PetscAssertPointer(idxn, 5);
1526: MatCheckPreallocated(mat, 1);
1528: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
1529: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
1531: if (PetscDefined(USE_DEBUG)) {
1532: PetscInt i, j;
1534: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
1535: if (v) {
1536: for (i = 0; i < m; i++) {
1537: for (j = 0; j < n; j++) {
1538: if (mat->erroriffailure && PetscIsInfOrNanScalar(v[i * n + j]))
1539: #if defined(PETSC_USE_COMPLEX)
1540: SETERRQ(PETSC_COMM_SELF, PETSC_ERR_FP, "Inserting %g+i%g at matrix entry (%" PetscInt_FMT ",%" PetscInt_FMT ")", (double)PetscRealPart(v[i * n + j]), (double)PetscImaginaryPart(v[i * n + j]), idxm[i], idxn[j]);
1541: #else
1542: SETERRQ(PETSC_COMM_SELF, PETSC_ERR_FP, "Inserting %g at matrix entry (%" PetscInt_FMT ",%" PetscInt_FMT ")", (double)v[i * n + j], idxm[i], idxn[j]);
1543: #endif
1544: }
1545: }
1546: }
1547: for (i = 0; i < m; i++) PetscCheck(idxm[i] < mat->rmap->N, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Cannot insert in row %" PetscInt_FMT ", maximum is %" PetscInt_FMT, idxm[i], mat->rmap->N - 1);
1548: for (i = 0; i < n; i++) PetscCheck(idxn[i] < mat->cmap->N, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Cannot insert in column %" PetscInt_FMT ", maximum is %" PetscInt_FMT, idxn[i], mat->cmap->N - 1);
1549: }
1551: if (mat->assembled) {
1552: mat->was_assembled = PETSC_TRUE;
1553: mat->assembled = PETSC_FALSE;
1554: }
1555: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
1556: PetscUseTypeMethod(mat, setvalues, m, idxm, n, idxn, v, addv);
1557: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
1558: PetscFunctionReturn(PETSC_SUCCESS);
1559: }
1561: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
1562: /*@
1563: MatSetValuesIS - Inserts or adds a block of values into a matrix using an `IS` to indicate the rows and columns
1564: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
1565: MUST be called after all calls to `MatSetValues()` have been completed.
1567: Not Collective
1569: Input Parameters:
1570: + mat - the matrix
1571: . ism - the rows to provide
1572: . isn - the columns to provide
1573: . v - a one-dimensional array that contains the values implicitly stored as a two-dimensional array, by default in row-major order.
1574: See `MAT_ROW_ORIENTED` in `MatSetOption()` for how to use column-major order.
1575: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
1577: Level: beginner
1579: Notes:
1580: By default, the values, `v`, are stored in row-major order. See `MAT_ROW_ORIENTED` in `MatSetOption()` for how to use column-major order.
1582: Calls to `MatSetValues()` with the `INSERT_VALUES` and `ADD_VALUES`
1583: options cannot be mixed without intervening calls to the assembly
1584: routines.
1586: `MatSetValues()` uses 0-based row and column numbers in Fortran
1587: as well as in C.
1589: Negative indices may be passed in `ism` and `isn`, these rows and columns are
1590: simply ignored. This allows easily inserting element stiffness matrices
1591: with homogeneous Dirichlet boundary conditions that you don't want represented
1592: in the matrix.
1594: Fortran Note:
1595: If `v` is a two-dimensional array make sure to first call `MatSetOption(mat, MAT_ROW_ORIENTED, PETSC_FALSE, ierr)` before using this function,
1596: otherwise the transpose of `v` will seemingly be inserted in the matrix, since Fortran passes two-dimensional arrays with column orientation.
1598: Efficiency Alert:
1599: The routine `MatSetValuesBlocked()` may offer much better efficiency
1600: for users of block sparse formats (`MATSEQBAIJ` and `MATMPIBAIJ`).
1602: This is currently not optimized for any particular `ISType`
1604: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatSetValues()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1605: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`
1606: @*/
1607: PetscErrorCode MatSetValuesIS(Mat mat, IS ism, IS isn, const PetscScalar v[], InsertMode addv)
1608: {
1609: PetscInt m, n;
1610: const PetscInt *rows, *cols;
1612: PetscFunctionBeginHot;
1614: PetscCall(ISGetIndices(ism, &rows));
1615: PetscCall(ISGetIndices(isn, &cols));
1616: PetscCall(ISGetLocalSize(ism, &m));
1617: PetscCall(ISGetLocalSize(isn, &n));
1618: PetscCall(MatSetValues(mat, m, rows, n, cols, v, addv));
1619: PetscCall(ISRestoreIndices(ism, &rows));
1620: PetscCall(ISRestoreIndices(isn, &cols));
1621: PetscFunctionReturn(PETSC_SUCCESS);
1622: }
1624: /*@
1625: MatSetValuesRowLocal - Inserts a row of nonzero values into a matrix
1627: Not Collective
1629: Input Parameters:
1630: + mat - the matrix
1631: . row - the row to set
1632: - v - a one-dimensional array that contains the values
1634: Level: intermediate
1636: Notes:
1637: Currently only supported for `MATAIJ`.
1639: All the nonzero values in `row` must be provided
1641: The matrix must have previously had its column indices set, likely by having been assembled.
1643: `row` must belong to this MPI process
1645: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1646: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`, `MatSetValues()`, `MatSetValuesRow()`, `MatSetLocalToGlobalMapping()`, `MATAIJ`
1647: @*/
1648: PetscErrorCode MatSetValuesRowLocal(Mat mat, PetscInt row, const PetscScalar v[])
1649: {
1650: PetscInt globalrow;
1652: PetscFunctionBegin;
1655: PetscAssertPointer(v, 3);
1656: PetscCall(ISLocalToGlobalMappingApply(mat->rmap->mapping, 1, &row, &globalrow));
1657: PetscCall(MatSetValuesRow(mat, globalrow, v));
1658: PetscFunctionReturn(PETSC_SUCCESS);
1659: }
1661: /*@
1662: MatSetValuesRow - Inserts a row of nonzero values into a matrix
1664: Not Collective
1666: Input Parameters:
1667: + mat - the matrix
1668: . row - the row to set
1669: - v - a one dimensional array of values
1671: Level: advanced
1673: Notes:
1674: Currently only supported for `MATAIJ`.
1676: All the nonzeros in `row` must be provided
1678: The matrix must have previously had its column indices set, likely by having been assembled.
1680: `row` must belong to this process
1682: .seealso: [](ch_matrices), `Mat`, `MatSetValues()`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1683: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`, `MATAIJ`
1684: @*/
1685: PetscErrorCode MatSetValuesRow(Mat mat, PetscInt row, const PetscScalar v[])
1686: {
1687: PetscFunctionBeginHot;
1690: MatCheckPreallocated(mat, 1);
1691: PetscAssertPointer(v, 3);
1692: PetscCheck(mat->insertmode != ADD_VALUES, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add and insert values");
1693: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
1694: mat->insertmode = INSERT_VALUES;
1696: if (mat->assembled) {
1697: mat->was_assembled = PETSC_TRUE;
1698: mat->assembled = PETSC_FALSE;
1699: }
1700: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
1701: PetscUseTypeMethod(mat, setvaluesrow, row, v);
1702: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
1703: PetscFunctionReturn(PETSC_SUCCESS);
1704: }
1706: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
1707: /*@
1708: MatSetValuesStencil - Inserts or adds a block of values into a matrix.
1709: Using structured grid indexing
1711: Not Collective
1713: Input Parameters:
1714: + mat - the matrix
1715: . m - number of rows being entered
1716: . idxm - grid coordinates (and component number when dof > 1) for matrix rows being entered
1717: . n - number of columns being entered
1718: . idxn - grid coordinates (and component number when dof > 1) for matrix columns being entered
1719: . v - a one-dimensional array that contains the values implicitly stored as a two-dimensional array, by default in row-major order.
1720: See `MAT_ROW_ORIENTED` in `MatSetOption()` for how to use column-major order.
1721: - addv - either `ADD_VALUES` to add to existing entries at that location or `INSERT_VALUES` to replace existing entries with new values
1723: Level: beginner
1725: Notes:
1726: By default the values, `v`, are row-oriented. See `MatSetOption()` for other options.
1728: Calls to `MatSetValuesStencil()` with the `INSERT_VALUES` and `ADD_VALUES`
1729: options cannot be mixed without intervening calls to the assembly
1730: routines.
1732: The grid coordinates are across the entire grid, not just the local portion
1734: `MatSetValuesStencil()` uses 0-based row and column numbers in Fortran
1735: as well as in C.
1737: For setting/accessing vector values via array coordinates you can use the `DMDAVecGetArray()` routine
1739: In order to use this routine you must either obtain the matrix with `DMCreateMatrix()`
1740: or call `MatSetLocalToGlobalMapping()` and `MatSetStencil()` first.
1742: The columns and rows in the stencil passed in MUST be contained within the
1743: ghost region of the given process as set with DMDACreateXXX() or `MatSetStencil()`. For example,
1744: if you create a `DMDA` with an overlap of one grid level and on a particular process its first
1745: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1746: first i index you can use in your column and row indices in `MatSetStencil()` is 5.
1748: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
1749: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
1750: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
1751: `DM_BOUNDARY_PERIODIC` boundary type.
1753: For indices that don't mean anything for your case (like the k index when working in 2d) or the c index when you have
1754: a single value per point) you can skip filling those indices.
1756: Inspired by the structured grid interface to the HYPRE package
1757: (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)
1759: Fortran Notes:
1760: If any of `idxm`, `idxn`, and `v` are scalars pass them using, for example,
1761: .vb
1762: call MatSetValuesStencil(mat, one, [idxm], one, [idxn], [v], INSERT_VALUES, ierr)
1763: .ve
1765: If `v` is a two-dimensional array make sure to first call `MatSetOption(mat, MAT_ROW_ORIENTED, PETSC_FALSE, ierr)` before using this function,
1766: otherwise the transpose of `v` will seemingly be inserted in the matrix, since Fortran passes two-dimensional arrays with column orientation.
1768: Efficiency Alert:
1769: The routine `MatSetValuesBlockedStencil()` may offer much better efficiency
1770: for users of block sparse formats (`MATSEQBAIJ` and `MATMPIBAIJ`).
1772: .seealso: [](ch_matrices), `Mat`, `DMDA`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1773: `MatSetValues()`, `MatSetValuesBlockedStencil()`, `MatSetStencil()`, `DMCreateMatrix()`, `DMDAVecGetArray()`, `MatStencil`
1774: @*/
1775: PetscErrorCode MatSetValuesStencil(Mat mat, PetscInt m, const MatStencil idxm[], PetscInt n, const MatStencil idxn[], const PetscScalar v[], InsertMode addv)
1776: {
1777: PetscInt buf[8192], *bufm = NULL, *bufn = NULL, *jdxm, *jdxn;
1778: PetscInt j, i, dim = mat->stencil.dim, *dims = mat->stencil.dims + 1, tmp;
1779: PetscInt *starts = mat->stencil.starts, *dxm = (PetscInt *)idxm, *dxn = (PetscInt *)idxn, sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1781: PetscFunctionBegin;
1782: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1785: PetscAssertPointer(idxm, 3);
1786: PetscAssertPointer(idxn, 5);
1788: if ((m + n) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
1789: jdxm = buf;
1790: jdxn = buf + m;
1791: } else {
1792: PetscCall(PetscMalloc2(m, &bufm, n, &bufn));
1793: jdxm = bufm;
1794: jdxn = bufn;
1795: }
1796: for (i = 0; i < m; i++) {
1797: for (j = 0; j < 3 - sdim; j++) dxm++;
1798: tmp = *dxm++ - starts[0];
1799: for (j = 0; j < dim - 1; j++) {
1800: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1801: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
1802: }
1803: if (mat->stencil.noc) dxm++;
1804: jdxm[i] = tmp;
1805: }
1806: for (i = 0; i < n; i++) {
1807: for (j = 0; j < 3 - sdim; j++) dxn++;
1808: tmp = *dxn++ - starts[0];
1809: for (j = 0; j < dim - 1; j++) {
1810: if ((*dxn++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1811: else tmp = tmp * dims[j] + *(dxn - 1) - starts[j + 1];
1812: }
1813: if (mat->stencil.noc) dxn++;
1814: jdxn[i] = tmp;
1815: }
1816: PetscCall(MatSetValuesLocal(mat, m, jdxm, n, jdxn, v, addv));
1817: PetscCall(PetscFree2(bufm, bufn));
1818: PetscFunctionReturn(PETSC_SUCCESS);
1819: }
1821: /*@
1822: MatSetValuesBlockedStencil - Inserts or adds a block of values into a matrix.
1823: Using structured grid indexing
1825: Not Collective
1827: Input Parameters:
1828: + mat - the matrix
1829: . m - number of rows being entered
1830: . idxm - grid coordinates for matrix rows being entered
1831: . n - number of columns being entered
1832: . idxn - grid coordinates for matrix columns being entered
1833: . v - a one-dimensional array that contains the values implicitly stored as a two-dimensional array, by default in row-major order.
1834: See `MAT_ROW_ORIENTED` in `MatSetOption()` for how to use column-major order.
1835: - addv - either `ADD_VALUES` to add to existing entries or `INSERT_VALUES` to replace existing entries with new values
1837: Level: beginner
1839: Notes:
1840: By default the values, `v`, are row-oriented and unsorted.
1841: See `MatSetOption()` for other options.
1843: Calls to `MatSetValuesBlockedStencil()` with the `INSERT_VALUES` and `ADD_VALUES`
1844: options cannot be mixed without intervening calls to the assembly
1845: routines.
1847: The grid coordinates are across the entire grid, not just the local portion
1849: `MatSetValuesBlockedStencil()` uses 0-based row and column numbers in Fortran
1850: as well as in C.
1852: For setting/accessing vector values via array coordinates you can use the `DMDAVecGetArray()` routine
1854: In order to use this routine you must either obtain the matrix with `DMCreateMatrix()`
1855: or call `MatSetBlockSize()`, `MatSetLocalToGlobalMapping()` and `MatSetStencil()` first.
1857: The columns and rows in the stencil passed in MUST be contained within the
1858: ghost region of the given process as set with DMDACreateXXX() or `MatSetStencil()`. For example,
1859: if you create a `DMDA` with an overlap of one grid level and on a particular process its first
1860: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1861: first i index you can use in your column and row indices in `MatSetStencil()` is 5.
1863: Negative indices may be passed in `idxm` and `idxn`, these rows and columns are
1864: simply ignored. This allows easily inserting element stiffness matrices
1865: with homogeneous Dirichlet boundary conditions that you don't want represented
1866: in the matrix.
1868: Inspired by the structured grid interface to the HYPRE package
1869: (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)
1871: Fortran Notes:
1872: If any of `idxm`, `idxn`, and `v` are scalars pass them using, for example,
1873: .vb
1874: call MatSetValuesBlockedStencil(mat, one, [idxm], one, [idxn], [v], INSERT_VALUES, ierr)
1875: .ve
1877: If `v` is a two-dimensional array make sure to first call `MatSetOption(mat, MAT_ROW_ORIENTED, PETSC_FALSE, ierr)` before using this function,
1878: otherwise the transpose of `v` will seemingly be inserted in the matrix, since Fortran passes two-dimensional arrays with column orientation.
1880: .seealso: [](ch_matrices), `Mat`, `DMDA`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1881: `MatSetValues()`, `MatSetValuesStencil()`, `MatSetStencil()`, `DMCreateMatrix()`, `DMDAVecGetArray()`, `MatStencil`,
1882: `MatSetBlockSize()`, `MatSetLocalToGlobalMapping()`
1883: @*/
1884: PetscErrorCode MatSetValuesBlockedStencil(Mat mat, PetscInt m, const MatStencil idxm[], PetscInt n, const MatStencil idxn[], const PetscScalar v[], InsertMode addv)
1885: {
1886: PetscInt buf[8192], *bufm = NULL, *bufn = NULL, *jdxm, *jdxn;
1887: PetscInt j, i, dim = mat->stencil.dim, *dims = mat->stencil.dims + 1, tmp;
1888: PetscInt *starts = mat->stencil.starts, *dxm = (PetscInt *)idxm, *dxn = (PetscInt *)idxn, sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1890: PetscFunctionBegin;
1891: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1894: PetscAssertPointer(idxm, 3);
1895: PetscAssertPointer(idxn, 5);
1896: PetscAssertPointer(v, 6);
1898: if ((m + n) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
1899: jdxm = buf;
1900: jdxn = buf + m;
1901: } else {
1902: PetscCall(PetscMalloc2(m, &bufm, n, &bufn));
1903: jdxm = bufm;
1904: jdxn = bufn;
1905: }
1906: for (i = 0; i < m; i++) {
1907: for (j = 0; j < 3 - sdim; j++) dxm++;
1908: tmp = *dxm++ - starts[0];
1909: for (j = 0; j < sdim - 1; j++) {
1910: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1911: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
1912: }
1913: dxm++;
1914: jdxm[i] = tmp;
1915: }
1916: for (i = 0; i < n; i++) {
1917: for (j = 0; j < 3 - sdim; j++) dxn++;
1918: tmp = *dxn++ - starts[0];
1919: for (j = 0; j < sdim - 1; j++) {
1920: if ((*dxn++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1921: else tmp = tmp * dims[j] + *(dxn - 1) - starts[j + 1];
1922: }
1923: dxn++;
1924: jdxn[i] = tmp;
1925: }
1926: PetscCall(MatSetValuesBlockedLocal(mat, m, jdxm, n, jdxn, v, addv));
1927: PetscCall(PetscFree2(bufm, bufn));
1928: PetscFunctionReturn(PETSC_SUCCESS);
1929: }
1931: /*@
1932: MatSetStencil - Sets the grid information for setting values into a matrix via
1933: `MatSetValuesStencil()`
1935: Not Collective
1937: Input Parameters:
1938: + mat - the matrix
1939: . dim - dimension of the grid 1, 2, or 3
1940: . dims - number of grid points in x, y, and z direction, including ghost points on your processor
1941: . starts - starting point of ghost nodes on your processor in x, y, and z direction
1942: - dof - number of degrees of freedom per node
1944: Level: beginner
1946: Notes:
1947: Inspired by the structured grid interface to the HYPRE package
1948: (www.llnl.gov/CASC/hyper)
1950: For matrices generated with `DMCreateMatrix()` this routine is automatically called and so not needed by the
1951: user.
1953: .seealso: [](ch_matrices), `Mat`, `MatStencil`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1954: `MatSetValues()`, `MatSetValuesBlockedStencil()`, `MatSetValuesStencil()`
1955: @*/
1956: PetscErrorCode MatSetStencil(Mat mat, PetscInt dim, const PetscInt dims[], const PetscInt starts[], PetscInt dof)
1957: {
1958: PetscFunctionBegin;
1960: PetscAssertPointer(dims, 3);
1961: PetscAssertPointer(starts, 4);
1963: mat->stencil.dim = dim + (dof > 1);
1964: for (PetscInt i = 0; i < dim; i++) {
1965: mat->stencil.dims[i] = dims[dim - i - 1]; /* copy the values in backwards */
1966: mat->stencil.starts[i] = starts[dim - i - 1];
1967: }
1968: mat->stencil.dims[dim] = dof;
1969: mat->stencil.starts[dim] = 0;
1970: mat->stencil.noc = (PetscBool)(dof == 1);
1971: PetscFunctionReturn(PETSC_SUCCESS);
1972: }
1974: /*@
1975: MatSetValuesBlocked - Inserts or adds a block of values into a matrix.
1977: Not Collective
1979: Input Parameters:
1980: + mat - the matrix
1981: . m - the number of block rows
1982: . idxm - the global block indices
1983: . n - the number of block columns
1984: . idxn - the global block indices
1985: . v - a one-dimensional array that contains the values implicitly stored as a two-dimensional array, by default in row-major order.
1986: See `MAT_ROW_ORIENTED` in `MatSetOption()` for how to use column-major order.
1987: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` replaces existing entries with new values
1989: Level: intermediate
1991: Notes:
1992: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call
1993: MatXXXXSetPreallocation() or `MatSetUp()` before using this routine.
1995: The `m` and `n` count the NUMBER of blocks in the row direction and column direction,
1996: NOT the total number of rows/columns; for example, if the block size is 2 and
1997: you are passing in values for rows 2,3,4,5 then `m` would be 2 (not 4).
1998: The values in `idxm` would be 1 2; that is the first index for each block divided by
1999: the block size.
2001: You must call `MatSetBlockSize()` when constructing this matrix (before
2002: preallocating it).
2004: By default, the values, `v`, are stored in row-major order. See `MAT_ROW_ORIENTED` in `MatSetOption()` for how to use column-major order.
2006: Calls to `MatSetValuesBlocked()` with the `INSERT_VALUES` and `ADD_VALUES`
2007: options cannot be mixed without intervening calls to the assembly
2008: routines.
2010: `MatSetValuesBlocked()` uses 0-based row and column numbers in Fortran
2011: as well as in C.
2013: Negative indices may be passed in `idxm` and `idxn`, these rows and columns are
2014: simply ignored. This allows easily inserting element stiffness matrices
2015: with homogeneous Dirichlet boundary conditions that you don't want represented
2016: in the matrix.
2018: Each time an entry is set within a sparse matrix via `MatSetValues()`,
2019: internal searching must be done to determine where to place the
2020: data in the matrix storage space. By instead inserting blocks of
2021: entries via `MatSetValuesBlocked()`, the overhead of matrix assembly is
2022: reduced.
2024: Example:
2025: .vb
2026: Suppose m=n=2 and block size(bs) = 2 The array is
2028: 1 2 | 3 4
2029: 5 6 | 7 8
2030: - - - | - - -
2031: 9 10 | 11 12
2032: 13 14 | 15 16
2034: v[] should be passed in like
2035: v[] = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]
2037: If you are not using row-oriented storage of v (that is you called MatSetOption(mat,MAT_ROW_ORIENTED,PETSC_FALSE)) then
2038: v[] = [1,5,9,13,2,6,10,14,3,7,11,15,4,8,12,16]
2039: .ve
2041: Fortran Notes:
2042: If any of `idmx`, `idxn`, and `v` are scalars pass them using, for example,
2043: .vb
2044: call MatSetValuesBlocked(mat, one, [idxm], one, [idxn], [v], INSERT_VALUES, ierr)
2045: .ve
2047: If `v` is a two-dimensional array make sure to first call `MatSetOption(mat, MAT_ROW_ORIENTED, PETSC_FALSE, ierr)` before using this function,
2048: otherwise the transpose of `v` will seemingly be inserted in the matrix, since Fortran passes two-dimensional arrays with column orientation.
2050: .seealso: [](ch_matrices), `Mat`, `MatSetBlockSize()`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetValuesBlockedLocal()`
2051: @*/
2052: PetscErrorCode MatSetValuesBlocked(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], const PetscScalar v[], InsertMode addv)
2053: {
2054: PetscFunctionBeginHot;
2057: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
2058: PetscAssertPointer(idxm, 3);
2059: PetscAssertPointer(idxn, 5);
2060: MatCheckPreallocated(mat, 1);
2061: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
2062: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
2063: if (PetscDefined(USE_DEBUG)) {
2064: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2065: PetscCheck(mat->ops->setvaluesblocked || mat->ops->setvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2066: }
2067: if (PetscDefined(USE_DEBUG)) {
2068: PetscInt rbs, cbs, M, N, i;
2069: PetscCall(MatGetBlockSizes(mat, &rbs, &cbs));
2070: PetscCall(MatGetSize(mat, &M, &N));
2071: for (i = 0; i < m; i++) PetscCheck(idxm[i] * rbs < M, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Row block %" PetscInt_FMT " contains an index %" PetscInt_FMT "*%" PetscInt_FMT " greater than row length %" PetscInt_FMT, i, idxm[i], rbs, M);
2072: for (i = 0; i < n; i++)
2073: PetscCheck(idxn[i] * cbs < N, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column block %" PetscInt_FMT " contains an index %" PetscInt_FMT "*%" PetscInt_FMT " greater than column length %" PetscInt_FMT, i, idxn[i], cbs, N);
2074: }
2075: if (mat->assembled) {
2076: mat->was_assembled = PETSC_TRUE;
2077: mat->assembled = PETSC_FALSE;
2078: }
2079: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
2080: if (mat->ops->setvaluesblocked) PetscUseTypeMethod(mat, setvaluesblocked, m, idxm, n, idxn, v, addv);
2081: else {
2082: PetscInt buf[8192], *bufr = NULL, *bufc = NULL, *iidxm, *iidxn;
2083: PetscInt i, j, bs, cbs;
2085: PetscCall(MatGetBlockSizes(mat, &bs, &cbs));
2086: if ((m * bs + n * cbs) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
2087: iidxm = buf;
2088: iidxn = buf + m * bs;
2089: } else {
2090: PetscCall(PetscMalloc2(m * bs, &bufr, n * cbs, &bufc));
2091: iidxm = bufr;
2092: iidxn = bufc;
2093: }
2094: for (i = 0; i < m; i++) {
2095: for (j = 0; j < bs; j++) iidxm[i * bs + j] = bs * idxm[i] + j;
2096: }
2097: if (m != n || bs != cbs || idxm != idxn) {
2098: for (i = 0; i < n; i++) {
2099: for (j = 0; j < cbs; j++) iidxn[i * cbs + j] = cbs * idxn[i] + j;
2100: }
2101: } else iidxn = iidxm;
2102: PetscCall(MatSetValues(mat, m * bs, iidxm, n * cbs, iidxn, v, addv));
2103: PetscCall(PetscFree2(bufr, bufc));
2104: }
2105: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
2106: PetscFunctionReturn(PETSC_SUCCESS);
2107: }
2109: /*@
2110: MatGetValues - Gets a block of local values from a matrix.
2112: Not Collective; can only return values that are owned by the give process
2114: Input Parameters:
2115: + mat - the matrix
2116: . v - a logically two-dimensional array for storing the values
2117: . m - the number of rows
2118: . idxm - the global indices of the rows
2119: . n - the number of columns
2120: - idxn - the global indices of the columns
2122: Level: advanced
2124: Notes:
2125: The user must allocate space (m*n `PetscScalar`s) for the values, `v`.
2126: The values, `v`, are then returned in a row-oriented format,
2127: analogous to that used by default in `MatSetValues()`.
2129: `MatGetValues()` uses 0-based row and column numbers in
2130: Fortran as well as in C.
2132: `MatGetValues()` requires that the matrix has been assembled
2133: with `MatAssemblyBegin()`/`MatAssemblyEnd()`. Thus, calls to
2134: `MatSetValues()` and `MatGetValues()` CANNOT be made in succession
2135: without intermediate matrix assembly.
2137: Negative row or column indices will be ignored and those locations in `v` will be
2138: left unchanged.
2140: For the standard row-based matrix formats, `idxm` can only contain rows owned by the requesting MPI process.
2141: That is, rows with global index greater than or equal to rstart and less than rend where rstart and rend are obtainable
2142: from `MatGetOwnershipRange`(mat,&rstart,&rend).
2144: .seealso: [](ch_matrices), `Mat`, `MatGetRow()`, `MatCreateSubMatrices()`, `MatSetValues()`, `MatGetOwnershipRange()`, `MatGetValuesLocal()`, `MatGetValue()`
2145: @*/
2146: PetscErrorCode MatGetValues(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], PetscScalar v[])
2147: {
2148: PetscFunctionBegin;
2151: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS);
2152: PetscAssertPointer(idxm, 3);
2153: PetscAssertPointer(idxn, 5);
2154: PetscAssertPointer(v, 6);
2155: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2156: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2157: MatCheckPreallocated(mat, 1);
2159: PetscCall(PetscLogEventBegin(MAT_GetValues, mat, 0, 0, 0));
2160: PetscUseTypeMethod(mat, getvalues, m, idxm, n, idxn, v);
2161: PetscCall(PetscLogEventEnd(MAT_GetValues, mat, 0, 0, 0));
2162: PetscFunctionReturn(PETSC_SUCCESS);
2163: }
2165: /*@
2166: MatGetValuesLocal - retrieves values from certain locations in a matrix using the local numbering of the indices
2167: defined previously by `MatSetLocalToGlobalMapping()`
2169: Not Collective
2171: Input Parameters:
2172: + mat - the matrix
2173: . nrow - number of rows
2174: . irow - the row local indices
2175: . ncol - number of columns
2176: - icol - the column local indices
2178: Output Parameter:
2179: . y - a one-dimensional array that contains the values implicitly stored as a two-dimensional array, by default in row-major order.
2180: See `MAT_ROW_ORIENTED` in `MatSetOption()` for how to use column-major order.
2182: Level: advanced
2184: Notes:
2185: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call `MatSetLocalToGlobalMapping()` before using this routine.
2187: This routine can only return values that are owned by the requesting MPI process. That is, for standard matrix formats, rows that, in the global numbering,
2188: are greater than or equal to rstart and less than rend where rstart and rend are obtainable from `MatGetOwnershipRange`(mat,&rstart,&rend). One can
2189: determine if the resulting global row associated with the local row r is owned by the requesting MPI process by applying the `ISLocalToGlobalMapping` set
2190: with `MatSetLocalToGlobalMapping()`.
2192: .seealso: [](ch_matrices), `Mat`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetLocalToGlobalMapping()`,
2193: `MatSetValuesLocal()`, `MatGetValues()`
2194: @*/
2195: PetscErrorCode MatGetValuesLocal(Mat mat, PetscInt nrow, const PetscInt irow[], PetscInt ncol, const PetscInt icol[], PetscScalar y[])
2196: {
2197: PetscFunctionBeginHot;
2200: MatCheckPreallocated(mat, 1);
2201: if (!nrow || !ncol) PetscFunctionReturn(PETSC_SUCCESS); /* no values to retrieve */
2202: PetscAssertPointer(irow, 3);
2203: PetscAssertPointer(icol, 5);
2204: if (PetscDefined(USE_DEBUG)) {
2205: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2206: PetscCheck(mat->ops->getvalueslocal || mat->ops->getvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2207: }
2208: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2209: PetscCall(PetscLogEventBegin(MAT_GetValues, mat, 0, 0, 0));
2210: if (mat->ops->getvalueslocal) PetscUseTypeMethod(mat, getvalueslocal, nrow, irow, ncol, icol, y);
2211: else {
2212: PetscInt buf[8192], *bufr = NULL, *bufc = NULL, *irowm, *icolm;
2213: if ((nrow + ncol) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
2214: irowm = buf;
2215: icolm = buf + nrow;
2216: } else {
2217: PetscCall(PetscMalloc2(nrow, &bufr, ncol, &bufc));
2218: irowm = bufr;
2219: icolm = bufc;
2220: }
2221: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "MatGetValuesLocal() cannot proceed without local-to-global row mapping (See MatSetLocalToGlobalMapping()).");
2222: PetscCheck(mat->cmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "MatGetValuesLocal() cannot proceed without local-to-global column mapping (See MatSetLocalToGlobalMapping()).");
2223: PetscCall(ISLocalToGlobalMappingApply(mat->rmap->mapping, nrow, irow, irowm));
2224: PetscCall(ISLocalToGlobalMappingApply(mat->cmap->mapping, ncol, icol, icolm));
2225: PetscCall(MatGetValues(mat, nrow, irowm, ncol, icolm, y));
2226: PetscCall(PetscFree2(bufr, bufc));
2227: }
2228: PetscCall(PetscLogEventEnd(MAT_GetValues, mat, 0, 0, 0));
2229: PetscFunctionReturn(PETSC_SUCCESS);
2230: }
2232: /*@
2233: MatSetValuesBatch - Adds (`ADD_VALUES`) many blocks of values into a matrix at once. The blocks must all be square and
2234: the same size. Currently, this can only be called once and creates the given matrix.
2236: Not Collective
2238: Input Parameters:
2239: + mat - the matrix
2240: . nb - the number of blocks
2241: . bs - the number of rows (and columns) in each block
2242: . rows - a concatenation of the rows for each block
2243: - v - a concatenation of logically two-dimensional arrays of values
2245: Level: advanced
2247: Notes:
2248: `MatSetPreallocationCOO()` and `MatSetValuesCOO()` may be a better way to provide the values
2250: In the future, we will extend this routine to handle rectangular blocks, and to allow multiple calls for a given matrix.
2252: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
2253: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`, `MatSetValues()`, `MatSetPreallocationCOO()`, `MatSetValuesCOO()`
2254: @*/
2255: PetscErrorCode MatSetValuesBatch(Mat mat, PetscInt nb, PetscInt bs, PetscInt rows[], const PetscScalar v[])
2256: {
2257: PetscFunctionBegin;
2260: PetscAssertPointer(rows, 4);
2261: PetscAssertPointer(v, 5);
2262: PetscAssert(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2264: PetscCall(PetscLogEventBegin(MAT_SetValuesBatch, mat, 0, 0, 0));
2265: if (mat->ops->setvaluesbatch) PetscUseTypeMethod(mat, setvaluesbatch, nb, bs, rows, v);
2266: else {
2267: for (PetscInt b = 0; b < nb; ++b) PetscCall(MatSetValues(mat, bs, &rows[b * bs], bs, &rows[b * bs], &v[b * bs * bs], ADD_VALUES));
2268: }
2269: PetscCall(PetscLogEventEnd(MAT_SetValuesBatch, mat, 0, 0, 0));
2270: PetscFunctionReturn(PETSC_SUCCESS);
2271: }
2273: /*@
2274: MatSetLocalToGlobalMapping - Sets a local-to-global numbering for use by
2275: the routine `MatSetValuesLocal()` to allow users to insert matrix entries
2276: using a local (per-processor) numbering.
2278: Not Collective
2280: Input Parameters:
2281: + x - the matrix
2282: . rmapping - row mapping created with `ISLocalToGlobalMappingCreate()` or `ISLocalToGlobalMappingCreateIS()`
2283: - cmapping - column mapping
2285: Level: intermediate
2287: Note:
2288: If the matrix is obtained with `DMCreateMatrix()` then this may already have been called on the matrix
2290: .seealso: [](ch_matrices), `Mat`, `DM`, `DMCreateMatrix()`, `MatGetLocalToGlobalMapping()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetValuesLocal()`, `MatGetValuesLocal()`
2291: @*/
2292: PetscErrorCode MatSetLocalToGlobalMapping(Mat x, ISLocalToGlobalMapping rmapping, ISLocalToGlobalMapping cmapping)
2293: {
2294: PetscFunctionBegin;
2299: if (x->ops->setlocaltoglobalmapping) PetscUseTypeMethod(x, setlocaltoglobalmapping, rmapping, cmapping);
2300: else {
2301: PetscCall(PetscLayoutSetISLocalToGlobalMapping(x->rmap, rmapping));
2302: PetscCall(PetscLayoutSetISLocalToGlobalMapping(x->cmap, cmapping));
2303: }
2304: PetscFunctionReturn(PETSC_SUCCESS);
2305: }
2307: /*@
2308: MatGetLocalToGlobalMapping - Gets the local-to-global numbering set by `MatSetLocalToGlobalMapping()`
2310: Not Collective
2312: Input Parameter:
2313: . A - the matrix
2315: Output Parameters:
2316: + rmapping - row mapping
2317: - cmapping - column mapping
2319: Level: advanced
2321: .seealso: [](ch_matrices), `Mat`, `MatSetLocalToGlobalMapping()`, `MatSetValuesLocal()`
2322: @*/
2323: PetscErrorCode MatGetLocalToGlobalMapping(Mat A, ISLocalToGlobalMapping *rmapping, ISLocalToGlobalMapping *cmapping)
2324: {
2325: PetscFunctionBegin;
2328: if (rmapping) {
2329: PetscAssertPointer(rmapping, 2);
2330: *rmapping = A->rmap->mapping;
2331: }
2332: if (cmapping) {
2333: PetscAssertPointer(cmapping, 3);
2334: *cmapping = A->cmap->mapping;
2335: }
2336: PetscFunctionReturn(PETSC_SUCCESS);
2337: }
2339: /*@
2340: MatSetLayouts - Sets the `PetscLayout` objects for rows and columns of a matrix
2342: Logically Collective
2344: Input Parameters:
2345: + A - the matrix
2346: . rmap - row layout
2347: - cmap - column layout
2349: Level: advanced
2351: Note:
2352: The `PetscLayout` objects are usually created automatically for the matrix so this routine rarely needs to be called.
2354: .seealso: [](ch_matrices), `Mat`, `PetscLayout`, `MatCreateVecs()`, `MatGetLocalToGlobalMapping()`, `MatGetLayouts()`
2355: @*/
2356: PetscErrorCode MatSetLayouts(Mat A, PetscLayout rmap, PetscLayout cmap)
2357: {
2358: PetscFunctionBegin;
2360: PetscCall(PetscLayoutReference(rmap, &A->rmap));
2361: PetscCall(PetscLayoutReference(cmap, &A->cmap));
2362: PetscFunctionReturn(PETSC_SUCCESS);
2363: }
2365: /*@
2366: MatGetLayouts - Gets the `PetscLayout` objects for rows and columns
2368: Not Collective
2370: Input Parameter:
2371: . A - the matrix
2373: Output Parameters:
2374: + rmap - row layout
2375: - cmap - column layout
2377: Level: advanced
2379: .seealso: [](ch_matrices), `Mat`, [Matrix Layouts](sec_matlayout), `PetscLayout`, `MatCreateVecs()`, `MatGetLocalToGlobalMapping()`, `MatSetLayouts()`
2380: @*/
2381: PetscErrorCode MatGetLayouts(Mat A, PetscLayout *rmap, PetscLayout *cmap)
2382: {
2383: PetscFunctionBegin;
2386: if (rmap) {
2387: PetscAssertPointer(rmap, 2);
2388: *rmap = A->rmap;
2389: }
2390: if (cmap) {
2391: PetscAssertPointer(cmap, 3);
2392: *cmap = A->cmap;
2393: }
2394: PetscFunctionReturn(PETSC_SUCCESS);
2395: }
2397: /*@
2398: MatSetValuesLocal - Inserts or adds values into certain locations of a matrix,
2399: using a local numbering of the rows and columns.
2401: Not Collective
2403: Input Parameters:
2404: + mat - the matrix
2405: . nrow - number of rows
2406: . irow - the row local indices
2407: . ncol - number of columns
2408: . icol - the column local indices
2409: . v - a one-dimensional array that contains the values implicitly stored as a two-dimensional array, by default in row-major order.
2410: See `MAT_ROW_ORIENTED` in `MatSetOption()` for how to use column-major order.
2411: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
2413: Level: intermediate
2415: Notes:
2416: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call `MatSetLocalToGlobalMapping()` before using this routine
2418: Calls to `MatSetValuesLocal()` with the `INSERT_VALUES` and `ADD_VALUES`
2419: options cannot be mixed without intervening calls to the assembly
2420: routines.
2422: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
2423: MUST be called after all calls to `MatSetValuesLocal()` have been completed.
2425: Fortran Notes:
2426: If any of `irow`, `icol`, and `v` are scalars pass them using, for example,
2427: .vb
2428: call MatSetValuesLocal(mat, one, [irow], one, [icol], [v], INSERT_VALUES, ierr)
2429: .ve
2431: If `v` is a two-dimensional array make sure to first call `MatSetOption(mat, MAT_ROW_ORIENTED, PETSC_FALSE, ierr)` before using this function,
2432: otherwise the transpose of `v` will seemingly be inserted in the matrix, since Fortran passes two-dimensional arrays with column orientation.
2434: .seealso: [](ch_matrices), `Mat`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetLocalToGlobalMapping()`,
2435: `MatGetValuesLocal()`
2436: @*/
2437: PetscErrorCode MatSetValuesLocal(Mat mat, PetscInt nrow, const PetscInt irow[], PetscInt ncol, const PetscInt icol[], const PetscScalar v[], InsertMode addv)
2438: {
2439: PetscFunctionBeginHot;
2442: MatCheckPreallocated(mat, 1);
2443: if (!nrow || !ncol) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
2444: PetscAssertPointer(irow, 3);
2445: PetscAssertPointer(icol, 5);
2446: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
2447: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
2448: if (PetscDefined(USE_DEBUG)) {
2449: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2450: PetscCheck(mat->ops->setvalueslocal || mat->ops->setvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2451: }
2453: if (mat->assembled) {
2454: mat->was_assembled = PETSC_TRUE;
2455: mat->assembled = PETSC_FALSE;
2456: }
2457: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
2458: if (mat->ops->setvalueslocal) PetscUseTypeMethod(mat, setvalueslocal, nrow, irow, ncol, icol, v, addv);
2459: else {
2460: PetscInt buf[8192], *bufr = NULL, *bufc = NULL;
2461: const PetscInt *irowm, *icolm;
2463: if ((!mat->rmap->mapping && !mat->cmap->mapping) || (nrow + ncol) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
2464: bufr = buf;
2465: bufc = buf + nrow;
2466: irowm = bufr;
2467: icolm = bufc;
2468: } else {
2469: PetscCall(PetscMalloc2(nrow, &bufr, ncol, &bufc));
2470: irowm = bufr;
2471: icolm = bufc;
2472: }
2473: if (mat->rmap->mapping) PetscCall(ISLocalToGlobalMappingApply(mat->rmap->mapping, nrow, irow, bufr));
2474: else irowm = irow;
2475: if (mat->cmap->mapping) {
2476: if (mat->cmap->mapping != mat->rmap->mapping || ncol != nrow || icol != irow) PetscCall(ISLocalToGlobalMappingApply(mat->cmap->mapping, ncol, icol, bufc));
2477: else icolm = irowm;
2478: } else icolm = icol;
2479: PetscCall(MatSetValues(mat, nrow, irowm, ncol, icolm, v, addv));
2480: if (bufr != buf) PetscCall(PetscFree2(bufr, bufc));
2481: }
2482: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
2483: PetscFunctionReturn(PETSC_SUCCESS);
2484: }
2486: /*@
2487: MatSetValuesBlockedLocal - Inserts or adds values into certain locations of a matrix,
2488: using a local ordering of the nodes a block at a time.
2490: Not Collective
2492: Input Parameters:
2493: + mat - the matrix
2494: . nrow - number of rows
2495: . irow - the row local indices
2496: . ncol - number of columns
2497: . icol - the column local indices
2498: . v - a one-dimensional array that contains the values implicitly stored as a two-dimensional array, by default in row-major order.
2499: See `MAT_ROW_ORIENTED` in `MatSetOption()` for how to use column-major order.
2500: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
2502: Level: intermediate
2504: Notes:
2505: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call `MatSetBlockSize()` and `MatSetLocalToGlobalMapping()`
2506: before using this routineBefore calling `MatSetValuesLocal()`, the user must first set the
2508: Calls to `MatSetValuesBlockedLocal()` with the `INSERT_VALUES` and `ADD_VALUES`
2509: options cannot be mixed without intervening calls to the assembly
2510: routines.
2512: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
2513: MUST be called after all calls to `MatSetValuesBlockedLocal()` have been completed.
2515: Fortran Notes:
2516: If any of `irow`, `icol`, and `v` are scalars pass them using, for example,
2517: .vb
2518: call MatSetValuesBlockedLocal(mat, one, [irow], one, [icol], [v], INSERT_VALUES, ierr)
2519: .ve
2521: If `v` is a two-dimensional array make sure to first call `MatSetOption(mat, MAT_ROW_ORIENTED, PETSC_FALSE, ierr)` before using this function,
2522: otherwise the transpose of `v` will seemingly be inserted in the matrix, since Fortran passes two-dimensional arrays with column orientation.
2524: .seealso: [](ch_matrices), `Mat`, `MatSetBlockSize()`, `MatSetLocalToGlobalMapping()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`,
2525: `MatSetValuesLocal()`, `MatSetValuesBlocked()`
2526: @*/
2527: PetscErrorCode MatSetValuesBlockedLocal(Mat mat, PetscInt nrow, const PetscInt irow[], PetscInt ncol, const PetscInt icol[], const PetscScalar v[], InsertMode addv)
2528: {
2529: PetscFunctionBeginHot;
2532: MatCheckPreallocated(mat, 1);
2533: if (!nrow || !ncol) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
2534: PetscAssertPointer(irow, 3);
2535: PetscAssertPointer(icol, 5);
2536: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
2537: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
2538: if (PetscDefined(USE_DEBUG)) {
2539: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2540: PetscCheck(mat->ops->setvaluesblockedlocal || mat->ops->setvaluesblocked || mat->ops->setvalueslocal || mat->ops->setvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2541: }
2543: if (mat->assembled) {
2544: mat->was_assembled = PETSC_TRUE;
2545: mat->assembled = PETSC_FALSE;
2546: }
2547: if (PetscUnlikelyDebug(mat->rmap->mapping)) { /* Condition on the mapping existing, because MatSetValuesBlockedLocal_IS does not require it to be set. */
2548: PetscInt irbs, rbs;
2549: PetscCall(MatGetBlockSizes(mat, &rbs, NULL));
2550: PetscCall(ISLocalToGlobalMappingGetBlockSize(mat->rmap->mapping, &irbs));
2551: PetscCheck(rbs == irbs, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Different row block sizes! mat %" PetscInt_FMT ", row l2g map %" PetscInt_FMT, rbs, irbs);
2552: }
2553: if (PetscUnlikelyDebug(mat->cmap->mapping)) {
2554: PetscInt icbs, cbs;
2555: PetscCall(MatGetBlockSizes(mat, NULL, &cbs));
2556: PetscCall(ISLocalToGlobalMappingGetBlockSize(mat->cmap->mapping, &icbs));
2557: PetscCheck(cbs == icbs, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Different col block sizes! mat %" PetscInt_FMT ", col l2g map %" PetscInt_FMT, cbs, icbs);
2558: }
2559: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
2560: if (mat->ops->setvaluesblockedlocal) PetscUseTypeMethod(mat, setvaluesblockedlocal, nrow, irow, ncol, icol, v, addv);
2561: else {
2562: PetscInt buf[8192], *bufr = NULL, *bufc = NULL;
2563: const PetscInt *irowm, *icolm;
2565: if ((!mat->rmap->mapping && !mat->cmap->mapping) || (nrow + ncol) <= ((PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf))) {
2566: bufr = buf;
2567: bufc = buf + nrow;
2568: irowm = bufr;
2569: icolm = bufc;
2570: } else {
2571: PetscCall(PetscMalloc2(nrow, &bufr, ncol, &bufc));
2572: irowm = bufr;
2573: icolm = bufc;
2574: }
2575: if (mat->rmap->mapping) PetscCall(ISLocalToGlobalMappingApplyBlock(mat->rmap->mapping, nrow, irow, bufr));
2576: else irowm = irow;
2577: if (mat->cmap->mapping) {
2578: if (mat->cmap->mapping != mat->rmap->mapping || ncol != nrow || icol != irow) PetscCall(ISLocalToGlobalMappingApplyBlock(mat->cmap->mapping, ncol, icol, bufc));
2579: else icolm = irowm;
2580: } else icolm = icol;
2581: PetscCall(MatSetValuesBlocked(mat, nrow, irowm, ncol, icolm, v, addv));
2582: if (bufr != buf) PetscCall(PetscFree2(bufr, bufc));
2583: }
2584: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
2585: PetscFunctionReturn(PETSC_SUCCESS);
2586: }
2588: /*@
2589: MatMultDiagonalBlock - Computes the matrix-vector product, $y = Dx$. Where `D` is defined by the inode or block structure of the diagonal
2591: Collective
2593: Input Parameters:
2594: + mat - the matrix
2595: - x - the vector to be multiplied
2597: Output Parameter:
2598: . y - the result
2600: Level: developer
2602: Note:
2603: The vectors `x` and `y` cannot be the same. I.e., one cannot
2604: call `MatMultDiagonalBlock`(A,y,y).
2606: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
2607: @*/
2608: PetscErrorCode MatMultDiagonalBlock(Mat mat, Vec x, Vec y)
2609: {
2610: PetscFunctionBegin;
2616: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2617: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2618: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2619: MatCheckPreallocated(mat, 1);
2621: PetscUseTypeMethod(mat, multdiagonalblock, x, y);
2622: PetscCall(PetscObjectStateIncrease((PetscObject)y));
2623: PetscFunctionReturn(PETSC_SUCCESS);
2624: }
2626: /*@
2627: MatMult - Computes the matrix-vector product, $y = Ax$.
2629: Neighbor-wise Collective
2631: Input Parameters:
2632: + mat - the matrix
2633: - x - the vector to be multiplied
2635: Output Parameter:
2636: . y - the result
2638: Level: beginner
2640: Note:
2641: The vectors `x` and `y` cannot be the same. I.e., one cannot
2642: call `MatMult`(A,y,y).
2644: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
2645: @*/
2646: PetscErrorCode MatMult(Mat mat, Vec x, Vec y)
2647: {
2648: PetscFunctionBegin;
2652: VecCheckAssembled(x);
2654: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2655: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2656: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2657: PetscCheck(mat->cmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, x->map->N);
2658: PetscCheck(mat->rmap->N == y->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec y: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, y->map->N);
2659: PetscCheck(mat->cmap->n == x->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->n, x->map->n);
2660: PetscCheck(mat->rmap->n == y->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec y: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->n, y->map->n);
2661: PetscCall(VecSetErrorIfLocked(y, 3));
2662: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(x, 2, PETSC_TRUE));
2663: MatCheckPreallocated(mat, 1);
2665: PetscCall(VecLockReadPush(x));
2666: PetscCall(PetscLogEventBegin(MAT_Mult, mat, x, y, 0));
2667: PetscUseTypeMethod(mat, mult, x, y);
2668: PetscCall(PetscLogEventEnd(MAT_Mult, mat, x, y, 0));
2669: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(y, 3, PETSC_FALSE));
2670: PetscCall(VecLockReadPop(x));
2671: PetscFunctionReturn(PETSC_SUCCESS);
2672: }
2674: /*@
2675: MatMultTranspose - Computes matrix transpose times a vector $y = A^T * x$.
2677: Neighbor-wise Collective
2679: Input Parameters:
2680: + mat - the matrix
2681: - x - the vector to be multiplied
2683: Output Parameter:
2684: . y - the result
2686: Level: beginner
2688: Notes:
2689: The vectors `x` and `y` cannot be the same. I.e., one cannot
2690: call `MatMultTranspose`(A,y,y).
2692: For complex numbers this does NOT compute the Hermitian (complex conjugate) transpose multiple,
2693: use `MatMultHermitianTranspose()`
2695: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatMultHermitianTranspose()`, `MatTranspose()`
2696: @*/
2697: PetscErrorCode MatMultTranspose(Mat mat, Vec x, Vec y)
2698: {
2699: PetscErrorCode (*op)(Mat, Vec, Vec) = NULL;
2701: PetscFunctionBegin;
2705: VecCheckAssembled(x);
2708: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2709: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2710: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2711: PetscCheck(mat->cmap->N == y->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec y: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, y->map->N);
2712: PetscCheck(mat->rmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, x->map->N);
2713: PetscCheck(mat->cmap->n == y->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec y: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->n, y->map->n);
2714: PetscCheck(mat->rmap->n == x->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->n, x->map->n);
2715: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(x, 2, PETSC_TRUE));
2716: MatCheckPreallocated(mat, 1);
2718: if (!mat->ops->multtranspose) {
2719: if (mat->symmetric == PETSC_BOOL3_TRUE && mat->ops->mult) op = mat->ops->mult;
2720: PetscCheck(op, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Matrix type %s does not have a multiply transpose defined or is symmetric and does not have a multiply defined", ((PetscObject)mat)->type_name);
2721: } else op = mat->ops->multtranspose;
2722: PetscCall(PetscLogEventBegin(MAT_MultTranspose, mat, x, y, 0));
2723: PetscCall(VecLockReadPush(x));
2724: PetscCall((*op)(mat, x, y));
2725: PetscCall(VecLockReadPop(x));
2726: PetscCall(PetscLogEventEnd(MAT_MultTranspose, mat, x, y, 0));
2727: PetscCall(PetscObjectStateIncrease((PetscObject)y));
2728: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(y, 3, PETSC_FALSE));
2729: PetscFunctionReturn(PETSC_SUCCESS);
2730: }
2732: /*@
2733: MatMultHermitianTranspose - Computes matrix Hermitian-transpose times a vector $y = A^H * x$.
2735: Neighbor-wise Collective
2737: Input Parameters:
2738: + mat - the matrix
2739: - x - the vector to be multiplied
2741: Output Parameter:
2742: . y - the result
2744: Level: beginner
2746: Notes:
2747: The vectors `x` and `y` cannot be the same. I.e., one cannot
2748: call `MatMultHermitianTranspose`(A,y,y).
2750: Also called the conjugate transpose, complex conjugate transpose, or adjoint.
2752: For real numbers `MatMultTranspose()` and `MatMultHermitianTranspose()` are identical.
2754: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatMultAdd()`, `MatMultHermitianTransposeAdd()`, `MatMultTranspose()`
2755: @*/
2756: PetscErrorCode MatMultHermitianTranspose(Mat mat, Vec x, Vec y)
2757: {
2758: PetscFunctionBegin;
2764: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2765: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2766: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2767: PetscCheck(mat->cmap->N == y->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec y: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, y->map->N);
2768: PetscCheck(mat->rmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, x->map->N);
2769: PetscCheck(mat->cmap->n == y->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec y: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->n, y->map->n);
2770: PetscCheck(mat->rmap->n == x->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->n, x->map->n);
2771: MatCheckPreallocated(mat, 1);
2773: PetscCall(PetscLogEventBegin(MAT_MultHermitianTranspose, mat, x, y, 0));
2774: #if defined(PETSC_USE_COMPLEX)
2775: if (mat->ops->multhermitiantranspose || (mat->hermitian == PETSC_BOOL3_TRUE && mat->ops->mult)) {
2776: PetscCall(VecLockReadPush(x));
2777: if (mat->ops->multhermitiantranspose) PetscUseTypeMethod(mat, multhermitiantranspose, x, y);
2778: else PetscUseTypeMethod(mat, mult, x, y);
2779: PetscCall(VecLockReadPop(x));
2780: } else {
2781: Vec w;
2782: PetscCall(VecDuplicate(x, &w));
2783: PetscCall(VecCopy(x, w));
2784: PetscCall(VecConjugate(w));
2785: PetscCall(MatMultTranspose(mat, w, y));
2786: PetscCall(VecDestroy(&w));
2787: PetscCall(VecConjugate(y));
2788: }
2789: PetscCall(PetscObjectStateIncrease((PetscObject)y));
2790: #else
2791: PetscCall(MatMultTranspose(mat, x, y));
2792: #endif
2793: PetscCall(PetscLogEventEnd(MAT_MultHermitianTranspose, mat, x, y, 0));
2794: PetscFunctionReturn(PETSC_SUCCESS);
2795: }
2797: /*@
2798: MatMultAdd - Computes $v3 = v2 + A * v1$.
2800: Neighbor-wise Collective
2802: Input Parameters:
2803: + mat - the matrix
2804: . v1 - the vector to be multiplied by `mat`
2805: - v2 - the vector to be added to the result
2807: Output Parameter:
2808: . v3 - the result
2810: Level: beginner
2812: Note:
2813: The vectors `v1` and `v3` cannot be the same. I.e., one cannot
2814: call `MatMultAdd`(A,v1,v2,v1).
2816: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMult()`, `MatMultTransposeAdd()`
2817: @*/
2818: PetscErrorCode MatMultAdd(Mat mat, Vec v1, Vec v2, Vec v3)
2819: {
2820: PetscFunctionBegin;
2827: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2828: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2829: PetscCheck(mat->cmap->N == v1->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec v1: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, v1->map->N);
2830: /* PetscCheck(mat->rmap->N == v2->map->N,PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec v2: global dim %" PetscInt_FMT " %" PetscInt_FMT,mat->rmap->N,v2->map->N);
2831: PetscCheck(mat->rmap->N == v3->map->N,PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec v3: global dim %" PetscInt_FMT " %" PetscInt_FMT,mat->rmap->N,v3->map->N); */
2832: PetscCheck(mat->rmap->n == v3->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec v3: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->n, v3->map->n);
2833: PetscCheck(mat->rmap->n == v2->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec v2: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->n, v2->map->n);
2834: PetscCheck(v1 != v3, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "v1 and v3 must be different vectors");
2835: MatCheckPreallocated(mat, 1);
2837: PetscCall(PetscLogEventBegin(MAT_MultAdd, mat, v1, v2, v3));
2838: PetscCall(VecLockReadPush(v1));
2839: PetscUseTypeMethod(mat, multadd, v1, v2, v3);
2840: PetscCall(VecLockReadPop(v1));
2841: PetscCall(PetscLogEventEnd(MAT_MultAdd, mat, v1, v2, v3));
2842: PetscCall(PetscObjectStateIncrease((PetscObject)v3));
2843: PetscFunctionReturn(PETSC_SUCCESS);
2844: }
2846: /*@
2847: MatMultTransposeAdd - Computes $v3 = v2 + A^T * v1$.
2849: Neighbor-wise Collective
2851: Input Parameters:
2852: + mat - the matrix
2853: . v1 - the vector to be multiplied by the transpose of the matrix
2854: - v2 - the vector to be added to the result
2856: Output Parameter:
2857: . v3 - the result
2859: Level: beginner
2861: Note:
2862: The vectors `v1` and `v3` cannot be the same. I.e., one cannot
2863: call `MatMultTransposeAdd`(A,v1,v2,v1).
2865: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMult()`
2866: @*/
2867: PetscErrorCode MatMultTransposeAdd(Mat mat, Vec v1, Vec v2, Vec v3)
2868: {
2869: PetscErrorCode (*op)(Mat, Vec, Vec, Vec) = (!mat->ops->multtransposeadd && mat->symmetric) ? mat->ops->multadd : mat->ops->multtransposeadd;
2871: PetscFunctionBegin;
2878: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2879: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2880: PetscCheck(mat->rmap->N == v1->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec v1: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, v1->map->N);
2881: PetscCheck(mat->cmap->N == v2->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec v2: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, v2->map->N);
2882: PetscCheck(mat->cmap->N == v3->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec v3: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, v3->map->N);
2883: PetscCheck(v1 != v3, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "v1 and v3 must be different vectors");
2884: PetscCheck(op, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2885: MatCheckPreallocated(mat, 1);
2887: PetscCall(PetscLogEventBegin(MAT_MultTransposeAdd, mat, v1, v2, v3));
2888: PetscCall(VecLockReadPush(v1));
2889: PetscCall((*op)(mat, v1, v2, v3));
2890: PetscCall(VecLockReadPop(v1));
2891: PetscCall(PetscLogEventEnd(MAT_MultTransposeAdd, mat, v1, v2, v3));
2892: PetscCall(PetscObjectStateIncrease((PetscObject)v3));
2893: PetscFunctionReturn(PETSC_SUCCESS);
2894: }
2896: /*@
2897: MatMultHermitianTransposeAdd - Computes $v3 = v2 + A^H * v1$.
2899: Neighbor-wise Collective
2901: Input Parameters:
2902: + mat - the matrix
2903: . v1 - the vector to be multiplied by the Hermitian transpose
2904: - v2 - the vector to be added to the result
2906: Output Parameter:
2907: . v3 - the result
2909: Level: beginner
2911: Note:
2912: The vectors `v1` and `v3` cannot be the same. I.e., one cannot
2913: call `MatMultHermitianTransposeAdd`(A,v1,v2,v1).
2915: .seealso: [](ch_matrices), `Mat`, `MatMultHermitianTranspose()`, `MatMultTranspose()`, `MatMultAdd()`, `MatMult()`
2916: @*/
2917: PetscErrorCode MatMultHermitianTransposeAdd(Mat mat, Vec v1, Vec v2, Vec v3)
2918: {
2919: PetscFunctionBegin;
2926: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2927: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2928: PetscCheck(v1 != v3, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "v1 and v3 must be different vectors");
2929: PetscCheck(mat->rmap->N == v1->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec v1: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, v1->map->N);
2930: PetscCheck(mat->cmap->N == v2->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec v2: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, v2->map->N);
2931: PetscCheck(mat->cmap->N == v3->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec v3: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, v3->map->N);
2932: MatCheckPreallocated(mat, 1);
2934: PetscCall(PetscLogEventBegin(MAT_MultHermitianTransposeAdd, mat, v1, v2, v3));
2935: PetscCall(VecLockReadPush(v1));
2936: if (mat->ops->multhermitiantransposeadd) PetscUseTypeMethod(mat, multhermitiantransposeadd, v1, v2, v3);
2937: else {
2938: Vec w, z;
2939: PetscCall(VecDuplicate(v1, &w));
2940: PetscCall(VecCopy(v1, w));
2941: PetscCall(VecConjugate(w));
2942: PetscCall(VecDuplicate(v3, &z));
2943: PetscCall(MatMultTranspose(mat, w, z));
2944: PetscCall(VecDestroy(&w));
2945: PetscCall(VecConjugate(z));
2946: if (v2 != v3) PetscCall(VecWAXPY(v3, 1.0, v2, z));
2947: else PetscCall(VecAXPY(v3, 1.0, z));
2948: PetscCall(VecDestroy(&z));
2949: }
2950: PetscCall(VecLockReadPop(v1));
2951: PetscCall(PetscLogEventEnd(MAT_MultHermitianTransposeAdd, mat, v1, v2, v3));
2952: PetscCall(PetscObjectStateIncrease((PetscObject)v3));
2953: PetscFunctionReturn(PETSC_SUCCESS);
2954: }
2956: PetscErrorCode MatADot_Default(Mat mat, Vec x, Vec y, PetscScalar *val)
2957: {
2958: PetscFunctionBegin;
2959: if (!mat->dot_vec) PetscCall(MatCreateVecs(mat, &mat->dot_vec, NULL));
2960: PetscCall(MatMult(mat, x, mat->dot_vec));
2961: PetscCall(VecDot(mat->dot_vec, y, val));
2962: PetscFunctionReturn(PETSC_SUCCESS);
2963: }
2965: PetscErrorCode MatANorm_Default(Mat mat, Vec x, PetscReal *val)
2966: {
2967: PetscScalar sval;
2969: PetscFunctionBegin;
2970: PetscCall(MatADot_Default(mat, x, x, &sval));
2971: PetscCheck(PetscRealPart(sval) >= 0.0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "Matrix argument is not positive definite");
2972: PetscCheck(PetscAbsReal(PetscImaginaryPart(sval)) < 100 * PETSC_MACHINE_EPSILON, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "Matrix argument is not Hermitian");
2973: *val = PetscSqrtReal(PetscRealPart(sval));
2974: PetscFunctionReturn(PETSC_SUCCESS);
2975: }
2977: /*@
2978: MatADot - Computes the inner product with respect to a matrix, i.e., $(x, y)_A = y^H A x$ where $A$ is symmetric (Hermitian when using complex)
2979: positive definite.
2981: Collective
2983: Input Parameters:
2984: + mat - matrix used to define the inner product
2985: . x - first vector
2986: - y - second vector
2988: Output Parameter:
2989: . val - the dot product with respect to `A`
2991: Level: intermediate
2993: Note:
2994: For complex vectors, `MatADot()` computes
2995: $$
2996: val = (x,y)_A = y^H A x,
2997: $$
2998: where $y^H$ denotes the conjugate transpose of `y`. Note that this corresponds to the "mathematicians" complex
2999: inner product where the SECOND argument gets the complex conjugate.
3001: .seealso: [](ch_matrices), `Mat`, `MatANorm()`, `VecDot()`, `VecNorm()`, `MatMult()`, `MatMultAdd()`, `MatMultTransposeAdd()`
3002: @*/
3003: PetscErrorCode MatADot(Mat mat, Vec x, Vec y, PetscScalar *val)
3004: {
3005: PetscFunctionBegin;
3009: VecCheckAssembled(x);
3011: VecCheckAssembled(y);
3014: PetscAssertPointer(val, 4);
3015: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3016: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3017: PetscCheck(mat->cmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, x->map->N);
3018: PetscCheck(mat->rmap->N == y->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec y: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, y->map->N);
3019: PetscCheck(mat->cmap->n == x->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->n, x->map->n);
3020: PetscCheck(mat->rmap->n == y->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec y: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->n, y->map->n);
3021: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(x, 2, PETSC_TRUE));
3022: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(y, 3, PETSC_TRUE));
3023: MatCheckPreallocated(mat, 1);
3025: PetscCall(VecLockReadPush(x));
3026: PetscCall(VecLockReadPush(y));
3027: PetscCall(PetscLogEventBegin(MAT_ADot, mat, x, y, 0));
3028: PetscUseTypeMethod(mat, adot, x, y, val);
3029: PetscCall(PetscLogEventEnd(MAT_ADot, mat, x, y, 0));
3030: PetscCall(VecLockReadPop(y));
3031: PetscCall(VecLockReadPop(x));
3032: PetscFunctionReturn(PETSC_SUCCESS);
3033: }
3035: /*@
3036: MatANorm - Computes the norm with respect to a matrix, i.e., $(x, x)_A^{1/2} = (x^H A x)^{1/2}$ where $A$ is symmetric (Hermitian when using complex)
3037: positive definite.
3039: Collective
3041: Input Parameters:
3042: + mat - matrix used to define norm
3043: - x - the vector to compute the norm of
3045: Output Parameter:
3046: . val - the norm with respect to `A`
3048: Level: intermediate
3050: Note:
3051: For complex vectors, `MatANorm()` computes
3052: $$
3053: val = (x,x)_A^{1/2} = (x^H A x)^{1/2},
3054: $$
3055: where $x^H$ denotes the conjugate transpose of `x`.
3057: .seealso: [](ch_matrices), `Mat`, `MatADot()`, `VecDot()`, `VecNorm()`, `MatMult()`, `MatMultAdd()`, `MatMultTransposeAdd()`
3058: @*/
3059: PetscErrorCode MatANorm(Mat mat, Vec x, PetscReal *val)
3060: {
3061: PetscFunctionBegin;
3065: VecCheckAssembled(x);
3067: PetscAssertPointer(val, 3);
3068: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3069: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3070: PetscCheck(mat->cmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, x->map->N);
3071: PetscCheck(mat->rmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, x->map->N);
3072: PetscCheck(mat->cmap->n == x->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->n, x->map->n);
3073: PetscCheck(mat->rmap->n == x->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->n, x->map->n);
3074: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(x, 2, PETSC_TRUE));
3075: MatCheckPreallocated(mat, 1);
3077: PetscCall(VecLockReadPush(x));
3078: PetscCall(PetscLogEventBegin(MAT_ANorm, mat, x, 0, 0));
3079: PetscUseTypeMethod(mat, anorm, x, val);
3080: PetscCall(PetscLogEventEnd(MAT_ANorm, mat, x, 0, 0));
3081: PetscCall(VecLockReadPop(x));
3082: PetscFunctionReturn(PETSC_SUCCESS);
3083: }
3085: /*@
3086: MatGetFactorType - gets the type of factorization a matrix is
3088: Not Collective
3090: Input Parameter:
3091: . mat - the matrix
3093: Output Parameter:
3094: . t - the type, one of `MAT_FACTOR_NONE`, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ILU`, `MAT_FACTOR_ICC,MAT_FACTOR_ILUDT`, `MAT_FACTOR_QR`
3096: Level: intermediate
3098: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorType`, `MatGetFactor()`, `MatSetFactorType()`, `MAT_FACTOR_NONE`, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ILU`,
3099: `MAT_FACTOR_ICC`, `MAT_FACTOR_ILUDT`, `MAT_FACTOR_QR`
3100: @*/
3101: PetscErrorCode MatGetFactorType(Mat mat, MatFactorType *t)
3102: {
3103: PetscFunctionBegin;
3106: PetscAssertPointer(t, 2);
3107: *t = mat->factortype;
3108: PetscFunctionReturn(PETSC_SUCCESS);
3109: }
3111: /*@
3112: MatSetFactorType - sets the type of factorization a matrix is
3114: Logically Collective
3116: Input Parameters:
3117: + mat - the matrix
3118: - t - the type, one of `MAT_FACTOR_NONE`, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ILU`, `MAT_FACTOR_ICC,MAT_FACTOR_ILUDT`, `MAT_FACTOR_QR`
3120: Level: intermediate
3122: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorType`, `MatGetFactor()`, `MatGetFactorType()`, `MAT_FACTOR_NONE`, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ILU`,
3123: `MAT_FACTOR_ICC`, `MAT_FACTOR_ILUDT`, `MAT_FACTOR_QR`
3124: @*/
3125: PetscErrorCode MatSetFactorType(Mat mat, MatFactorType t)
3126: {
3127: PetscFunctionBegin;
3130: mat->factortype = t;
3131: PetscFunctionReturn(PETSC_SUCCESS);
3132: }
3134: /*@
3135: MatGetInfo - Returns information about matrix storage (number of
3136: nonzeros, memory, etc.).
3138: Collective if `MAT_GLOBAL_MAX` or `MAT_GLOBAL_SUM` is used as the flag
3140: Input Parameters:
3141: + mat - the matrix
3142: - flag - flag indicating the type of parameters to be returned (`MAT_LOCAL` - local matrix, `MAT_GLOBAL_MAX` - maximum over all processors, `MAT_GLOBAL_SUM` - sum over all processors)
3144: Output Parameter:
3145: . info - matrix information context
3147: Options Database Key:
3148: . -mat_view ::ascii_info - print matrix info to `PETSC_STDOUT`
3150: Level: intermediate
3152: Notes:
3153: The `MatInfo` context contains a variety of matrix data, including
3154: number of nonzeros allocated and used, number of mallocs during
3155: matrix assembly, etc. Additional information for factored matrices
3156: is provided (such as the fill ratio, number of mallocs during
3157: factorization, etc.).
3159: Example:
3160: See the file ${PETSC_DIR}/include/petscmat.h for a complete list of
3161: data within the `MatInfo` context. For example,
3162: .vb
3163: MatInfo info;
3164: Mat A;
3165: double mal, nz_a, nz_u;
3167: MatGetInfo(A, MAT_LOCAL, &info);
3168: mal = info.mallocs;
3169: nz_a = info.nz_allocated;
3170: .ve
3172: .seealso: [](ch_matrices), `Mat`, `MatInfo`, `MatStashGetInfo()`
3173: @*/
3174: PetscErrorCode MatGetInfo(Mat mat, MatInfoType flag, MatInfo *info)
3175: {
3176: PetscFunctionBegin;
3179: PetscAssertPointer(info, 3);
3180: MatCheckPreallocated(mat, 1);
3181: PetscUseTypeMethod(mat, getinfo, flag, info);
3182: PetscFunctionReturn(PETSC_SUCCESS);
3183: }
3185: /*
3186: This is used by external packages where it is not easy to get the info from the actual
3187: matrix factorization.
3188: */
3189: PetscErrorCode MatGetInfo_External(Mat A, MatInfoType flag, MatInfo *info)
3190: {
3191: PetscFunctionBegin;
3192: PetscCall(PetscMemzero(info, sizeof(MatInfo)));
3193: PetscFunctionReturn(PETSC_SUCCESS);
3194: }
3196: /*@
3197: MatLUFactor - Performs in-place LU factorization of matrix.
3199: Collective
3201: Input Parameters:
3202: + mat - the matrix
3203: . row - row permutation
3204: . col - column permutation
3205: - info - options for factorization, includes
3206: .vb
3207: fill - expected fill as ratio of original fill.
3208: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3209: Run with the option -info to determine an optimal value to use
3210: .ve
3212: Level: developer
3214: Notes:
3215: Most users should employ the `KSP` interface for linear solvers
3216: instead of working directly with matrix algebra routines such as this.
3217: See, e.g., `KSPCreate()`.
3219: This changes the state of the matrix to a factored matrix; it cannot be used
3220: for example with `MatSetValues()` unless one first calls `MatSetUnfactored()`.
3222: This is really in-place only for dense matrices, the preferred approach is to use `MatGetFactor()`, `MatLUFactorSymbolic()`, and `MatLUFactorNumeric()`
3223: when not using `KSP`.
3225: Fortran Note:
3226: A valid (non-null) `info` argument must be provided
3228: .seealso: [](ch_matrices), [Matrix Factorization](sec_matfactor), `Mat`, `MatFactorType`, `MatLUFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`,
3229: `MatGetOrdering()`, `MatSetUnfactored()`, `MatFactorInfo`, `MatGetFactor()`
3230: @*/
3231: PetscErrorCode MatLUFactor(Mat mat, IS row, IS col, const MatFactorInfo *info)
3232: {
3233: MatFactorInfo tinfo;
3235: PetscFunctionBegin;
3239: if (info) PetscAssertPointer(info, 4);
3241: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3242: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3243: MatCheckPreallocated(mat, 1);
3244: if (!info) {
3245: PetscCall(MatFactorInfoInitialize(&tinfo));
3246: info = &tinfo;
3247: }
3249: PetscCall(PetscLogEventBegin(MAT_LUFactor, mat, row, col, 0));
3250: PetscUseTypeMethod(mat, lufactor, row, col, info);
3251: PetscCall(PetscLogEventEnd(MAT_LUFactor, mat, row, col, 0));
3252: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3253: PetscFunctionReturn(PETSC_SUCCESS);
3254: }
3256: /*@
3257: MatILUFactor - Performs in-place ILU factorization of matrix.
3259: Collective
3261: Input Parameters:
3262: + mat - the matrix
3263: . row - row permutation
3264: . col - column permutation
3265: - info - structure containing
3266: .vb
3267: levels - number of levels of fill.
3268: expected fill - as ratio of original fill.
3269: 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
3270: missing diagonal entries)
3271: .ve
3273: Level: developer
3275: Notes:
3276: Most users should employ the `KSP` interface for linear solvers
3277: instead of working directly with matrix algebra routines such as this.
3278: See, e.g., `KSPCreate()`.
3280: Probably really in-place only when level of fill is zero, otherwise allocates
3281: new space to store factored matrix and deletes previous memory. The preferred approach is to use `MatGetFactor()`, `MatILUFactorSymbolic()`, and `MatLUFactorNumeric()`
3282: when not using `KSP`.
3284: Fortran Note:
3285: A valid (non-null) `info` argument must be provided
3287: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatILUFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`, `MatFactorInfo`
3288: @*/
3289: PetscErrorCode MatILUFactor(Mat mat, IS row, IS col, const MatFactorInfo *info)
3290: {
3291: PetscFunctionBegin;
3295: PetscAssertPointer(info, 4);
3297: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "matrix must be square");
3298: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3299: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3300: MatCheckPreallocated(mat, 1);
3302: PetscCall(PetscLogEventBegin(MAT_ILUFactor, mat, row, col, 0));
3303: PetscUseTypeMethod(mat, ilufactor, row, col, info);
3304: PetscCall(PetscLogEventEnd(MAT_ILUFactor, mat, row, col, 0));
3305: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3306: PetscFunctionReturn(PETSC_SUCCESS);
3307: }
3309: /*@
3310: MatLUFactorSymbolic - Performs symbolic LU factorization of matrix.
3311: Call this routine before calling `MatLUFactorNumeric()` and after `MatGetFactor()`.
3313: Collective
3315: Input Parameters:
3316: + fact - the factor matrix obtained with `MatGetFactor()`
3317: . mat - the matrix
3318: . row - the row permutation
3319: . col - the column permutation
3320: - info - options for factorization, includes
3321: .vb
3322: fill - expected fill as ratio of original fill. Run with the option -info to determine an optimal value to use
3323: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3324: .ve
3326: Level: developer
3328: Notes:
3329: See [Matrix Factorization](sec_matfactor) for additional information about factorizations
3331: Most users should employ the simplified `KSP` interface for linear solvers
3332: instead of working directly with matrix algebra routines such as this.
3333: See, e.g., `KSPCreate()`.
3335: Fortran Note:
3336: A valid (non-null) `info` argument must be provided
3338: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatLUFactor()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`, `MatFactorInfo`, `MatFactorInfoInitialize()`
3339: @*/
3340: PetscErrorCode MatLUFactorSymbolic(Mat fact, Mat mat, IS row, IS col, const MatFactorInfo *info)
3341: {
3342: MatFactorInfo tinfo;
3344: PetscFunctionBegin;
3349: if (info) PetscAssertPointer(info, 5);
3352: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3353: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3354: MatCheckPreallocated(mat, 2);
3355: if (!info) {
3356: PetscCall(MatFactorInfoInitialize(&tinfo));
3357: info = &tinfo;
3358: }
3360: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_LUFactorSymbolic, mat, row, col, 0));
3361: PetscUseTypeMethod(fact, lufactorsymbolic, mat, row, col, info);
3362: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_LUFactorSymbolic, mat, row, col, 0));
3363: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3364: PetscFunctionReturn(PETSC_SUCCESS);
3365: }
3367: /*@
3368: MatLUFactorNumeric - Performs numeric LU factorization of a matrix.
3369: Call this routine after first calling `MatLUFactorSymbolic()` and `MatGetFactor()`.
3371: Collective
3373: Input Parameters:
3374: + fact - the factor matrix obtained with `MatGetFactor()`
3375: . mat - the matrix
3376: - info - options for factorization
3378: Level: developer
3380: Notes:
3381: See `MatLUFactor()` for in-place factorization. See
3382: `MatCholeskyFactorNumeric()` for the symmetric, positive definite case.
3384: Most users should employ the `KSP` interface for linear solvers
3385: instead of working directly with matrix algebra routines such as this.
3386: See, e.g., `KSPCreate()`.
3388: Fortran Note:
3389: A valid (non-null) `info` argument must be provided
3391: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatFactorInfo`, `MatLUFactorSymbolic()`, `MatLUFactor()`, `MatCholeskyFactor()`
3392: @*/
3393: PetscErrorCode MatLUFactorNumeric(Mat fact, Mat mat, const MatFactorInfo *info)
3394: {
3395: MatFactorInfo tinfo;
3397: PetscFunctionBegin;
3402: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3403: PetscCheck(mat->rmap->N == (fact)->rmap->N && mat->cmap->N == (fact)->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Mat fact: global dimensions are different %" PetscInt_FMT " should = %" PetscInt_FMT " %" PetscInt_FMT " should = %" PetscInt_FMT,
3404: mat->rmap->N, (fact)->rmap->N, mat->cmap->N, (fact)->cmap->N);
3406: MatCheckPreallocated(mat, 2);
3407: if (!info) {
3408: PetscCall(MatFactorInfoInitialize(&tinfo));
3409: info = &tinfo;
3410: }
3412: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_LUFactorNumeric, mat, fact, 0, 0));
3413: else PetscCall(PetscLogEventBegin(MAT_LUFactor, mat, fact, 0, 0));
3414: PetscUseTypeMethod(fact, lufactornumeric, mat, info);
3415: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_LUFactorNumeric, mat, fact, 0, 0));
3416: else PetscCall(PetscLogEventEnd(MAT_LUFactor, mat, fact, 0, 0));
3417: PetscCall(MatViewFromOptions(fact, NULL, "-mat_factor_view"));
3418: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3419: PetscFunctionReturn(PETSC_SUCCESS);
3420: }
3422: /*@
3423: MatCholeskyFactor - Performs in-place Cholesky factorization of a
3424: symmetric matrix.
3426: Collective
3428: Input Parameters:
3429: + mat - the matrix
3430: . perm - row and column permutations
3431: - info - expected fill as ratio of original fill
3433: Level: developer
3435: Notes:
3436: See `MatLUFactor()` for the nonsymmetric case. See also `MatGetFactor()`,
3437: `MatCholeskyFactorSymbolic()`, and `MatCholeskyFactorNumeric()`.
3439: Most users should employ the `KSP` interface for linear solvers
3440: instead of working directly with matrix algebra routines such as this.
3441: See, e.g., `KSPCreate()`.
3443: Fortran Note:
3444: A valid (non-null) `info` argument must be provided
3446: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatFactorInfo`, `MatLUFactor()`, `MatCholeskyFactorSymbolic()`, `MatCholeskyFactorNumeric()`,
3447: `MatGetOrdering()`
3448: @*/
3449: PetscErrorCode MatCholeskyFactor(Mat mat, IS perm, const MatFactorInfo *info)
3450: {
3451: MatFactorInfo tinfo;
3453: PetscFunctionBegin;
3456: if (info) PetscAssertPointer(info, 3);
3458: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "Matrix must be square");
3459: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3460: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3461: MatCheckPreallocated(mat, 1);
3462: if (!info) {
3463: PetscCall(MatFactorInfoInitialize(&tinfo));
3464: info = &tinfo;
3465: }
3467: PetscCall(PetscLogEventBegin(MAT_CholeskyFactor, mat, perm, 0, 0));
3468: PetscUseTypeMethod(mat, choleskyfactor, perm, info);
3469: PetscCall(PetscLogEventEnd(MAT_CholeskyFactor, mat, perm, 0, 0));
3470: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3471: PetscFunctionReturn(PETSC_SUCCESS);
3472: }
3474: /*@
3475: MatCholeskyFactorSymbolic - Performs symbolic Cholesky factorization
3476: of a symmetric matrix.
3478: Collective
3480: Input Parameters:
3481: + fact - the factor matrix obtained with `MatGetFactor()`
3482: . mat - the matrix
3483: . perm - row and column permutations
3484: - info - options for factorization, includes
3485: .vb
3486: fill - expected fill as ratio of original fill.
3487: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3488: Run with the option -info to determine an optimal value to use
3489: .ve
3491: Level: developer
3493: Notes:
3494: See `MatLUFactorSymbolic()` for the nonsymmetric case. See also
3495: `MatCholeskyFactor()` and `MatCholeskyFactorNumeric()`.
3497: Most users should employ the `KSP` interface for linear solvers
3498: instead of working directly with matrix algebra routines such as this.
3499: See, e.g., `KSPCreate()`.
3501: Fortran Note:
3502: A valid (non-null) `info` argument must be provided
3504: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactor()`, `MatCholeskyFactorNumeric()`,
3505: `MatGetOrdering()`
3506: @*/
3507: PetscErrorCode MatCholeskyFactorSymbolic(Mat fact, Mat mat, IS perm, const MatFactorInfo *info)
3508: {
3509: MatFactorInfo tinfo;
3511: PetscFunctionBegin;
3515: if (info) PetscAssertPointer(info, 4);
3518: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "Matrix must be square");
3519: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3520: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3521: MatCheckPreallocated(mat, 2);
3522: if (!info) {
3523: PetscCall(MatFactorInfoInitialize(&tinfo));
3524: info = &tinfo;
3525: }
3527: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_CholeskyFactorSymbolic, mat, perm, 0, 0));
3528: PetscUseTypeMethod(fact, choleskyfactorsymbolic, mat, perm, info);
3529: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_CholeskyFactorSymbolic, mat, perm, 0, 0));
3530: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3531: PetscFunctionReturn(PETSC_SUCCESS);
3532: }
3534: /*@
3535: MatCholeskyFactorNumeric - Performs numeric Cholesky factorization
3536: of a symmetric matrix. Call this routine after first calling `MatGetFactor()` and
3537: `MatCholeskyFactorSymbolic()`.
3539: Collective
3541: Input Parameters:
3542: + fact - the factor matrix obtained with `MatGetFactor()`, where the factored values are stored
3543: . mat - the initial matrix that is to be factored
3544: - info - options for factorization
3546: Level: developer
3548: Note:
3549: Most users should employ the `KSP` interface for linear solvers
3550: instead of working directly with matrix algebra routines such as this.
3551: See, e.g., `KSPCreate()`.
3553: Fortran Note:
3554: A valid (non-null) `info` argument must be provided
3556: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatCholeskyFactorSymbolic()`, `MatCholeskyFactor()`, `MatLUFactorNumeric()`
3557: @*/
3558: PetscErrorCode MatCholeskyFactorNumeric(Mat fact, Mat mat, const MatFactorInfo *info)
3559: {
3560: MatFactorInfo tinfo;
3562: PetscFunctionBegin;
3567: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3568: PetscCheck(mat->rmap->N == (fact)->rmap->N && mat->cmap->N == (fact)->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Mat fact: global dim %" PetscInt_FMT " should = %" PetscInt_FMT " %" PetscInt_FMT " should = %" PetscInt_FMT,
3569: mat->rmap->N, (fact)->rmap->N, mat->cmap->N, (fact)->cmap->N);
3570: MatCheckPreallocated(mat, 2);
3571: if (!info) {
3572: PetscCall(MatFactorInfoInitialize(&tinfo));
3573: info = &tinfo;
3574: }
3576: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_CholeskyFactorNumeric, mat, fact, 0, 0));
3577: else PetscCall(PetscLogEventBegin(MAT_CholeskyFactor, mat, fact, 0, 0));
3578: PetscUseTypeMethod(fact, choleskyfactornumeric, mat, info);
3579: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_CholeskyFactorNumeric, mat, fact, 0, 0));
3580: else PetscCall(PetscLogEventEnd(MAT_CholeskyFactor, mat, fact, 0, 0));
3581: PetscCall(MatViewFromOptions(fact, NULL, "-mat_factor_view"));
3582: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3583: PetscFunctionReturn(PETSC_SUCCESS);
3584: }
3586: /*@
3587: MatQRFactor - Performs in-place QR factorization of matrix.
3589: Collective
3591: Input Parameters:
3592: + mat - the matrix
3593: . col - column permutation
3594: - info - options for factorization, includes
3595: .vb
3596: fill - expected fill as ratio of original fill.
3597: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3598: Run with the option -info to determine an optimal value to use
3599: .ve
3601: Level: developer
3603: Notes:
3604: Most users should employ the `KSP` interface for linear solvers
3605: instead of working directly with matrix algebra routines such as this.
3606: See, e.g., `KSPCreate()`.
3608: This changes the state of the matrix to a factored matrix; it cannot be used
3609: for example with `MatSetValues()` unless one first calls `MatSetUnfactored()`.
3611: Fortran Note:
3612: A valid (non-null) `info` argument must be provided
3614: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatQRFactorSymbolic()`, `MatQRFactorNumeric()`, `MatLUFactor()`,
3615: `MatSetUnfactored()`
3616: @*/
3617: PetscErrorCode MatQRFactor(Mat mat, IS col, const MatFactorInfo *info)
3618: {
3619: PetscFunctionBegin;
3622: if (info) PetscAssertPointer(info, 3);
3624: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3625: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3626: MatCheckPreallocated(mat, 1);
3627: PetscCall(PetscLogEventBegin(MAT_QRFactor, mat, col, 0, 0));
3628: PetscUseMethod(mat, "MatQRFactor_C", (Mat, IS, const MatFactorInfo *), (mat, col, info));
3629: PetscCall(PetscLogEventEnd(MAT_QRFactor, mat, col, 0, 0));
3630: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3631: PetscFunctionReturn(PETSC_SUCCESS);
3632: }
3634: /*@
3635: MatQRFactorSymbolic - Performs symbolic QR factorization of matrix.
3636: Call this routine after `MatGetFactor()` but before calling `MatQRFactorNumeric()`.
3638: Collective
3640: Input Parameters:
3641: + fact - the factor matrix obtained with `MatGetFactor()`
3642: . mat - the matrix
3643: . col - column permutation
3644: - info - options for factorization, includes
3645: .vb
3646: fill - expected fill as ratio of original fill.
3647: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3648: Run with the option -info to determine an optimal value to use
3649: .ve
3651: Level: developer
3653: Note:
3654: Most users should employ the `KSP` interface for linear solvers
3655: instead of working directly with matrix algebra routines such as this.
3656: See, e.g., `KSPCreate()`.
3658: Fortran Note:
3659: A valid (non-null) `info` argument must be provided
3661: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatFactorInfo`, `MatQRFactor()`, `MatQRFactorNumeric()`, `MatLUFactor()`, `MatFactorInfoInitialize()`
3662: @*/
3663: PetscErrorCode MatQRFactorSymbolic(Mat fact, Mat mat, IS col, const MatFactorInfo *info)
3664: {
3665: MatFactorInfo tinfo;
3667: PetscFunctionBegin;
3671: if (info) PetscAssertPointer(info, 4);
3674: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3675: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3676: MatCheckPreallocated(mat, 2);
3677: if (!info) {
3678: PetscCall(MatFactorInfoInitialize(&tinfo));
3679: info = &tinfo;
3680: }
3682: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_QRFactorSymbolic, fact, mat, col, 0));
3683: PetscUseMethod(fact, "MatQRFactorSymbolic_C", (Mat, Mat, IS, const MatFactorInfo *), (fact, mat, col, info));
3684: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_QRFactorSymbolic, fact, mat, col, 0));
3685: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3686: PetscFunctionReturn(PETSC_SUCCESS);
3687: }
3689: /*@
3690: MatQRFactorNumeric - Performs numeric QR factorization of a matrix.
3691: Call this routine after first calling `MatGetFactor()`, and `MatQRFactorSymbolic()`.
3693: Collective
3695: Input Parameters:
3696: + fact - the factor matrix obtained with `MatGetFactor()`
3697: . mat - the matrix
3698: - info - options for factorization
3700: Level: developer
3702: Notes:
3703: See `MatQRFactor()` for in-place factorization.
3705: Most users should employ the `KSP` interface for linear solvers
3706: instead of working directly with matrix algebra routines such as this.
3707: See, e.g., `KSPCreate()`.
3709: Fortran Note:
3710: A valid (non-null) `info` argument must be provided
3712: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatQRFactor()`, `MatQRFactorSymbolic()`, `MatLUFactor()`
3713: @*/
3714: PetscErrorCode MatQRFactorNumeric(Mat fact, Mat mat, const MatFactorInfo *info)
3715: {
3716: MatFactorInfo tinfo;
3718: PetscFunctionBegin;
3723: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3724: PetscCheck(mat->rmap->N == fact->rmap->N && mat->cmap->N == fact->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Mat fact: global dimensions are different %" PetscInt_FMT " should = %" PetscInt_FMT " %" PetscInt_FMT " should = %" PetscInt_FMT,
3725: mat->rmap->N, (fact)->rmap->N, mat->cmap->N, (fact)->cmap->N);
3727: MatCheckPreallocated(mat, 2);
3728: if (!info) {
3729: PetscCall(MatFactorInfoInitialize(&tinfo));
3730: info = &tinfo;
3731: }
3733: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_QRFactorNumeric, mat, fact, 0, 0));
3734: else PetscCall(PetscLogEventBegin(MAT_QRFactor, mat, fact, 0, 0));
3735: PetscUseMethod(fact, "MatQRFactorNumeric_C", (Mat, Mat, const MatFactorInfo *), (fact, mat, info));
3736: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_QRFactorNumeric, mat, fact, 0, 0));
3737: else PetscCall(PetscLogEventEnd(MAT_QRFactor, mat, fact, 0, 0));
3738: PetscCall(MatViewFromOptions(fact, NULL, "-mat_factor_view"));
3739: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3740: PetscFunctionReturn(PETSC_SUCCESS);
3741: }
3743: /*@
3744: MatSolve - Solves $A x = b$, given a factored matrix.
3746: Neighbor-wise Collective
3748: Input Parameters:
3749: + mat - the factored matrix
3750: - b - the right-hand-side vector
3752: Output Parameter:
3753: . x - the result vector
3755: Level: developer
3757: Notes:
3758: The vectors `b` and `x` cannot be the same. I.e., one cannot
3759: call `MatSolve`(A,x,x).
3761: Most users should employ the `KSP` interface for linear solvers
3762: instead of working directly with matrix algebra routines such as this.
3763: See, e.g., `KSPCreate()`.
3765: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatLUFactor()`, `MatSolveAdd()`, `MatSolveTranspose()`, `MatSolveTransposeAdd()`
3766: @*/
3767: PetscErrorCode MatSolve(Mat mat, Vec b, Vec x)
3768: {
3769: PetscFunctionBegin;
3774: PetscCheckSameComm(mat, 1, b, 2);
3775: PetscCheckSameComm(mat, 1, x, 3);
3776: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3777: PetscCheck(mat->cmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, x->map->N);
3778: PetscCheck(mat->rmap->N == b->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec b: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, b->map->N);
3779: PetscCheck(mat->rmap->n == b->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec b: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->n, b->map->n);
3780: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3781: MatCheckPreallocated(mat, 1);
3783: PetscCall(PetscLogEventBegin(MAT_Solve, mat, b, x, 0));
3784: PetscCall(VecFlag(x, mat->factorerrortype));
3785: if (mat->factorerrortype) PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
3786: else PetscUseTypeMethod(mat, solve, b, x);
3787: PetscCall(PetscLogEventEnd(MAT_Solve, mat, b, x, 0));
3788: PetscCall(PetscObjectStateIncrease((PetscObject)x));
3789: PetscFunctionReturn(PETSC_SUCCESS);
3790: }
3792: static PetscErrorCode MatMatSolve_Basic(Mat A, Mat B, Mat X, PetscBool trans)
3793: {
3794: Vec b, x;
3795: PetscInt N, i;
3796: PetscErrorCode (*f)(Mat, Vec, Vec);
3797: PetscBool Abound, Bneedconv = PETSC_FALSE, Xneedconv = PETSC_FALSE;
3799: PetscFunctionBegin;
3800: if (A->factorerrortype) {
3801: PetscCall(PetscInfo(A, "MatFactorError %d\n", A->factorerrortype));
3802: PetscCall(MatSetInf(X));
3803: PetscFunctionReturn(PETSC_SUCCESS);
3804: }
3805: f = (!trans || (!A->ops->solvetranspose && A->symmetric)) ? A->ops->solve : A->ops->solvetranspose;
3806: PetscCheck(f, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Mat type %s", ((PetscObject)A)->type_name);
3807: PetscCall(MatBoundToCPU(A, &Abound));
3808: if (!Abound) {
3809: PetscCall(PetscObjectTypeCompareAny((PetscObject)B, &Bneedconv, MATSEQDENSE, MATMPIDENSE, ""));
3810: PetscCall(PetscObjectTypeCompareAny((PetscObject)X, &Xneedconv, MATSEQDENSE, MATMPIDENSE, ""));
3811: }
3812: #if PetscDefined(HAVE_CUDA)
3813: if (Bneedconv) PetscCall(MatConvert(B, MATDENSECUDA, MAT_INPLACE_MATRIX, &B));
3814: if (Xneedconv) PetscCall(MatConvert(X, MATDENSECUDA, MAT_INPLACE_MATRIX, &X));
3815: #elif PetscDefined(HAVE_HIP)
3816: if (Bneedconv) PetscCall(MatConvert(B, MATDENSEHIP, MAT_INPLACE_MATRIX, &B));
3817: if (Xneedconv) PetscCall(MatConvert(X, MATDENSEHIP, MAT_INPLACE_MATRIX, &X));
3818: #endif
3819: PetscCall(MatGetSize(B, NULL, &N));
3820: for (i = 0; i < N; i++) {
3821: PetscCall(MatDenseGetColumnVecRead(B, i, &b));
3822: PetscCall(MatDenseGetColumnVecWrite(X, i, &x));
3823: PetscCall((*f)(A, b, x));
3824: PetscCall(MatDenseRestoreColumnVecWrite(X, i, &x));
3825: PetscCall(MatDenseRestoreColumnVecRead(B, i, &b));
3826: }
3827: if (Bneedconv) PetscCall(MatConvert(B, MATDENSE, MAT_INPLACE_MATRIX, &B));
3828: if (Xneedconv) PetscCall(MatConvert(X, MATDENSE, MAT_INPLACE_MATRIX, &X));
3829: PetscFunctionReturn(PETSC_SUCCESS);
3830: }
3832: /*@
3833: MatMatSolve - Solves $A X = B$, given a factored matrix.
3835: Neighbor-wise Collective
3837: Input Parameters:
3838: + A - the factored matrix
3839: - B - the right-hand-side matrix `MATDENSE` (or sparse `MATAIJ`-- when using MUMPS)
3841: Output Parameter:
3842: . X - the result matrix (dense matrix)
3844: Level: developer
3846: Note:
3847: If `B` is a `MATDENSE` matrix then one can call `MatMatSolve`(A,B,B) except with `MATSOLVERMKL_CPARDISO`;
3848: otherwise, `B` and `X` cannot be the same.
3850: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSolve()`, `MatMatSolveTranspose()`, `MatLUFactor()`, `MatCholeskyFactor()`
3851: @*/
3852: PetscErrorCode MatMatSolve(Mat A, Mat B, Mat X)
3853: {
3854: PetscFunctionBegin;
3859: PetscCheckSameComm(A, 1, B, 2);
3860: PetscCheckSameComm(A, 1, X, 3);
3861: PetscCheck(A->cmap->N == X->rmap->N, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_SIZ, "Mat A,Mat X: global dim %" PetscInt_FMT " %" PetscInt_FMT, A->cmap->N, X->rmap->N);
3862: PetscCheck(A->rmap->N == B->rmap->N, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_SIZ, "Mat A,Mat B: global dim %" PetscInt_FMT " %" PetscInt_FMT, A->rmap->N, B->rmap->N);
3863: PetscCheck(X->cmap->N == B->cmap->N, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_SIZ, "Solution matrix must have same number of columns as rhs matrix");
3864: if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3865: MatCheckPreallocated(A, 1);
3867: PetscCall(PetscLogEventBegin(MAT_MatSolve, A, B, X, 0));
3868: if (!A->ops->matsolve) {
3869: PetscCall(PetscInfo(A, "Mat type %s using basic MatMatSolve\n", ((PetscObject)A)->type_name));
3870: PetscCall(MatMatSolve_Basic(A, B, X, PETSC_FALSE));
3871: } else PetscUseTypeMethod(A, matsolve, B, X);
3872: PetscCall(PetscLogEventEnd(MAT_MatSolve, A, B, X, 0));
3873: PetscCall(PetscObjectStateIncrease((PetscObject)X));
3874: PetscFunctionReturn(PETSC_SUCCESS);
3875: }
3877: /*@
3878: MatMatSolveTranspose - Solves $A^T X = B $, given a factored matrix.
3880: Neighbor-wise Collective
3882: Input Parameters:
3883: + A - the factored matrix
3884: - B - the right-hand-side matrix (`MATDENSE` matrix)
3886: Output Parameter:
3887: . X - the result matrix (dense matrix)
3889: Level: developer
3891: Note:
3892: The matrices `B` and `X` cannot be the same. I.e., one cannot
3893: call `MatMatSolveTranspose`(A,X,X).
3895: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSolveTranspose()`, `MatMatSolve()`, `MatLUFactor()`, `MatCholeskyFactor()`
3896: @*/
3897: PetscErrorCode MatMatSolveTranspose(Mat A, Mat B, Mat X)
3898: {
3899: PetscFunctionBegin;
3904: PetscCheckSameComm(A, 1, B, 2);
3905: PetscCheckSameComm(A, 1, X, 3);
3906: PetscCheck(X != B, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_IDN, "X and B must be different matrices");
3907: PetscCheck(A->cmap->N == X->rmap->N, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_SIZ, "Mat A,Mat X: global dim %" PetscInt_FMT " %" PetscInt_FMT, A->cmap->N, X->rmap->N);
3908: PetscCheck(A->rmap->N == B->rmap->N, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_SIZ, "Mat A,Mat B: global dim %" PetscInt_FMT " %" PetscInt_FMT, A->rmap->N, B->rmap->N);
3909: PetscCheck(A->rmap->n == B->rmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat A,Mat B: local dim %" PetscInt_FMT " %" PetscInt_FMT, A->rmap->n, B->rmap->n);
3910: PetscCheck(X->cmap->N >= B->cmap->N, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Solution matrix must have same number of columns as rhs matrix");
3911: if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3912: MatCheckPreallocated(A, 1);
3914: PetscCall(PetscLogEventBegin(MAT_MatSolve, A, B, X, 0));
3915: if (!A->ops->matsolvetranspose) {
3916: PetscCall(PetscInfo(A, "Mat type %s using basic MatMatSolveTranspose\n", ((PetscObject)A)->type_name));
3917: PetscCall(MatMatSolve_Basic(A, B, X, PETSC_TRUE));
3918: } else PetscUseTypeMethod(A, matsolvetranspose, B, X);
3919: PetscCall(PetscLogEventEnd(MAT_MatSolve, A, B, X, 0));
3920: PetscCall(PetscObjectStateIncrease((PetscObject)X));
3921: PetscFunctionReturn(PETSC_SUCCESS);
3922: }
3924: /*@
3925: MatMatTransposeSolve - Solves $A X = B^T$, given a factored matrix.
3927: Neighbor-wise Collective
3929: Input Parameters:
3930: + A - the factored matrix
3931: - Bt - the transpose of right-hand-side matrix as a `MATDENSE`
3933: Output Parameter:
3934: . X - the result matrix (dense matrix)
3936: Level: developer
3938: Note:
3939: For MUMPS, it only supports centralized sparse compressed column format on the host processor for right-hand side matrix. User must create `Bt` in sparse compressed row
3940: format on the host processor and call `MatMatTransposeSolve()` to implement MUMPS' `MatMatSolve()`.
3942: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatMatSolve()`, `MatMatSolveTranspose()`, `MatLUFactor()`, `MatCholeskyFactor()`
3943: @*/
3944: PetscErrorCode MatMatTransposeSolve(Mat A, Mat Bt, Mat X)
3945: {
3946: PetscFunctionBegin;
3951: PetscCheckSameComm(A, 1, Bt, 2);
3952: PetscCheckSameComm(A, 1, X, 3);
3954: PetscCheck(X != Bt, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_IDN, "X and B must be different matrices");
3955: PetscCheck(A->cmap->N == X->rmap->N, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_SIZ, "Mat A,Mat X: global dim %" PetscInt_FMT " %" PetscInt_FMT, A->cmap->N, X->rmap->N);
3956: PetscCheck(A->rmap->N == Bt->cmap->N, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_SIZ, "Mat A,Mat Bt: global dim %" PetscInt_FMT " %" PetscInt_FMT, A->rmap->N, Bt->cmap->N);
3957: PetscCheck(X->cmap->N >= Bt->rmap->N, PetscObjectComm((PetscObject)X), PETSC_ERR_ARG_SIZ, "Solution matrix must have same number of columns as row number of the rhs matrix");
3958: if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3959: PetscCheck(A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
3960: MatCheckPreallocated(A, 1);
3962: PetscCall(PetscLogEventBegin(MAT_MatTrSolve, A, Bt, X, 0));
3963: PetscUseTypeMethod(A, mattransposesolve, Bt, X);
3964: PetscCall(PetscLogEventEnd(MAT_MatTrSolve, A, Bt, X, 0));
3965: PetscCall(PetscObjectStateIncrease((PetscObject)X));
3966: PetscFunctionReturn(PETSC_SUCCESS);
3967: }
3969: /*@
3970: MatForwardSolve - Solves $ L x = b $, given a factored matrix, $A = LU $, or
3971: $U^T*D^(1/2) x = b$, given a factored symmetric matrix, $A = U^T*D*U$,
3973: Neighbor-wise Collective
3975: Input Parameters:
3976: + mat - the factored matrix
3977: - b - the right-hand-side vector
3979: Output Parameter:
3980: . x - the result vector
3982: Level: developer
3984: Notes:
3985: `MatSolve()` should be used for most applications, as it performs
3986: a forward solve followed by a backward solve.
3988: The vectors `b` and `x` cannot be the same, i.e., one cannot
3989: call `MatForwardSolve`(A,x,x).
3991: For matrix in `MATSEQBAIJ` format with block size larger than 1,
3992: the diagonal blocks are not implemented as $D = D^(1/2) * D^(1/2)$ yet.
3993: `MatForwardSolve()` solves $U^T*D y = b$, and
3994: `MatBackwardSolve()` solves $U x = y$.
3995: Thus they do not provide a symmetric preconditioner.
3997: .seealso: [](ch_matrices), `Mat`, `MatBackwardSolve()`, `MatGetFactor()`, `MatSolve()`
3998: @*/
3999: PetscErrorCode MatForwardSolve(Mat mat, Vec b, Vec x)
4000: {
4001: PetscFunctionBegin;
4006: PetscCheckSameComm(mat, 1, b, 2);
4007: PetscCheckSameComm(mat, 1, x, 3);
4008: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
4009: PetscCheck(mat->cmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, x->map->N);
4010: PetscCheck(mat->rmap->N == b->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec b: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, b->map->N);
4011: PetscCheck(mat->rmap->n == b->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec b: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->n, b->map->n);
4012: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
4013: MatCheckPreallocated(mat, 1);
4015: PetscCall(PetscLogEventBegin(MAT_ForwardSolve, mat, b, x, 0));
4016: PetscUseTypeMethod(mat, forwardsolve, b, x);
4017: PetscCall(PetscLogEventEnd(MAT_ForwardSolve, mat, b, x, 0));
4018: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4019: PetscFunctionReturn(PETSC_SUCCESS);
4020: }
4022: /*@
4023: MatBackwardSolve - Solves $U x = b$, given a factored matrix, $A = LU$.
4024: $D^(1/2) U x = b$, given a factored symmetric matrix, $A = U^T*D*U$,
4026: Neighbor-wise Collective
4028: Input Parameters:
4029: + mat - the factored matrix
4030: - b - the right-hand-side vector
4032: Output Parameter:
4033: . x - the result vector
4035: Level: developer
4037: Notes:
4038: `MatSolve()` should be used for most applications, as it performs
4039: a forward solve followed by a backward solve.
4041: The vectors `b` and `x` cannot be the same. I.e., one cannot
4042: call `MatBackwardSolve`(A,x,x).
4044: For matrix in `MATSEQBAIJ` format with block size larger than 1,
4045: the diagonal blocks are not implemented as $D = D^(1/2) * D^(1/2)$ yet.
4046: `MatForwardSolve()` solves $U^T*D y = b$, and
4047: `MatBackwardSolve()` solves $U x = y$.
4048: Thus they do not provide a symmetric preconditioner.
4050: .seealso: [](ch_matrices), `Mat`, `MatForwardSolve()`, `MatGetFactor()`, `MatSolve()`
4051: @*/
4052: PetscErrorCode MatBackwardSolve(Mat mat, Vec b, Vec x)
4053: {
4054: PetscFunctionBegin;
4059: PetscCheckSameComm(mat, 1, b, 2);
4060: PetscCheckSameComm(mat, 1, x, 3);
4061: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
4062: PetscCheck(mat->cmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, x->map->N);
4063: PetscCheck(mat->rmap->N == b->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec b: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, b->map->N);
4064: PetscCheck(mat->rmap->n == b->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec b: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->n, b->map->n);
4065: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
4066: MatCheckPreallocated(mat, 1);
4068: PetscCall(PetscLogEventBegin(MAT_BackwardSolve, mat, b, x, 0));
4069: PetscUseTypeMethod(mat, backwardsolve, b, x);
4070: PetscCall(PetscLogEventEnd(MAT_BackwardSolve, mat, b, x, 0));
4071: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4072: PetscFunctionReturn(PETSC_SUCCESS);
4073: }
4075: /*@
4076: MatSolveAdd - Computes $x = y + A^{-1}*b$, given a factored matrix.
4078: Neighbor-wise Collective
4080: Input Parameters:
4081: + mat - the factored matrix
4082: . b - the right-hand-side vector
4083: - y - the vector to be added to
4085: Output Parameter:
4086: . x - the result vector
4088: Level: developer
4090: Note:
4091: The vectors `b` and `x` cannot be the same. I.e., one cannot
4092: call `MatSolveAdd`(A,x,y,x).
4094: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatSolve()`, `MatGetFactor()`, `MatSolveTranspose()`, `MatSolveTransposeAdd()`
4095: @*/
4096: PetscErrorCode MatSolveAdd(Mat mat, Vec b, Vec y, Vec x)
4097: {
4098: PetscScalar one = 1.0;
4099: Vec tmp;
4101: PetscFunctionBegin;
4107: PetscCheckSameComm(mat, 1, b, 2);
4108: PetscCheckSameComm(mat, 1, y, 3);
4109: PetscCheckSameComm(mat, 1, x, 4);
4110: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
4111: PetscCheck(mat->cmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, x->map->N);
4112: PetscCheck(mat->rmap->N == b->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec b: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, b->map->N);
4113: PetscCheck(mat->rmap->N == y->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec y: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, y->map->N);
4114: PetscCheck(mat->rmap->n == b->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec b: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->n, b->map->n);
4115: PetscCheck(x->map->n == y->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Vec x,Vec y: local dim %" PetscInt_FMT " %" PetscInt_FMT, x->map->n, y->map->n);
4116: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
4117: MatCheckPreallocated(mat, 1);
4119: PetscCall(PetscLogEventBegin(MAT_SolveAdd, mat, b, x, y));
4120: PetscCall(VecFlag(x, mat->factorerrortype));
4121: if (mat->factorerrortype) {
4122: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
4123: } else if (mat->ops->solveadd) {
4124: PetscUseTypeMethod(mat, solveadd, b, y, x);
4125: } else {
4126: /* do the solve then the add manually */
4127: if (x != y) {
4128: PetscCall(MatSolve(mat, b, x));
4129: PetscCall(VecAXPY(x, one, y));
4130: } else {
4131: PetscCall(VecDuplicate(x, &tmp));
4132: PetscCall(VecCopy(x, tmp));
4133: PetscCall(MatSolve(mat, b, x));
4134: PetscCall(VecAXPY(x, one, tmp));
4135: PetscCall(VecDestroy(&tmp));
4136: }
4137: }
4138: PetscCall(PetscLogEventEnd(MAT_SolveAdd, mat, b, x, y));
4139: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4140: PetscFunctionReturn(PETSC_SUCCESS);
4141: }
4143: /*@
4144: MatSolveTranspose - Solves $A^T x = b$, given a factored matrix.
4146: Neighbor-wise Collective
4148: Input Parameters:
4149: + mat - the factored matrix
4150: - b - the right-hand-side vector
4152: Output Parameter:
4153: . x - the result vector
4155: Level: developer
4157: Notes:
4158: The vectors `b` and `x` cannot be the same. I.e., one cannot
4159: call `MatSolveTranspose`(A,x,x).
4161: Most users should employ the `KSP` interface for linear solvers
4162: instead of working directly with matrix algebra routines such as this.
4163: See, e.g., `KSPCreate()`.
4165: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `KSP`, `MatSolve()`, `MatSolveAdd()`, `MatSolveTransposeAdd()`
4166: @*/
4167: PetscErrorCode MatSolveTranspose(Mat mat, Vec b, Vec x)
4168: {
4169: PetscErrorCode (*f)(Mat, Vec, Vec) = (!mat->ops->solvetranspose && mat->symmetric) ? mat->ops->solve : mat->ops->solvetranspose;
4171: PetscFunctionBegin;
4176: PetscCheckSameComm(mat, 1, b, 2);
4177: PetscCheckSameComm(mat, 1, x, 3);
4178: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
4179: PetscCheck(mat->rmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, x->map->N);
4180: PetscCheck(mat->cmap->N == b->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec b: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, b->map->N);
4181: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
4182: MatCheckPreallocated(mat, 1);
4183: PetscCall(PetscLogEventBegin(MAT_SolveTranspose, mat, b, x, 0));
4184: PetscCall(VecFlag(x, mat->factorerrortype));
4185: if (mat->factorerrortype) {
4186: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
4187: } else {
4188: PetscCheck(f, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Matrix type %s", ((PetscObject)mat)->type_name);
4189: PetscCall((*f)(mat, b, x));
4190: }
4191: PetscCall(PetscLogEventEnd(MAT_SolveTranspose, mat, b, x, 0));
4192: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4193: PetscFunctionReturn(PETSC_SUCCESS);
4194: }
4196: /*@
4197: MatSolveTransposeAdd - Computes $x = y + A^{-T} b$
4198: factored matrix.
4200: Neighbor-wise Collective
4202: Input Parameters:
4203: + mat - the factored matrix
4204: . b - the right-hand-side vector
4205: - y - the vector to be added to
4207: Output Parameter:
4208: . x - the result vector
4210: Level: developer
4212: Note:
4213: The vectors `b` and `x` cannot be the same. I.e., one cannot
4214: call `MatSolveTransposeAdd`(A,x,y,x).
4216: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatSolve()`, `MatSolveAdd()`, `MatSolveTranspose()`
4217: @*/
4218: PetscErrorCode MatSolveTransposeAdd(Mat mat, Vec b, Vec y, Vec x)
4219: {
4220: PetscScalar one = 1.0;
4221: Vec tmp;
4222: PetscErrorCode (*f)(Mat, Vec, Vec, Vec) = (!mat->ops->solvetransposeadd && mat->symmetric) ? mat->ops->solveadd : mat->ops->solvetransposeadd;
4224: PetscFunctionBegin;
4230: PetscCheckSameComm(mat, 1, b, 2);
4231: PetscCheckSameComm(mat, 1, y, 3);
4232: PetscCheckSameComm(mat, 1, x, 4);
4233: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
4234: PetscCheck(mat->rmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, x->map->N);
4235: PetscCheck(mat->cmap->N == b->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec b: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, b->map->N);
4236: PetscCheck(mat->cmap->N == y->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec y: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, y->map->N);
4237: PetscCheck(x->map->n == y->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Vec x,Vec y: local dim %" PetscInt_FMT " %" PetscInt_FMT, x->map->n, y->map->n);
4238: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
4239: MatCheckPreallocated(mat, 1);
4241: PetscCall(PetscLogEventBegin(MAT_SolveTransposeAdd, mat, b, x, y));
4242: PetscCall(VecFlag(x, mat->factorerrortype));
4243: if (mat->factorerrortype) {
4244: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
4245: } else if (f) {
4246: PetscCall((*f)(mat, b, y, x));
4247: } else {
4248: /* do the solve then the add manually */
4249: if (x != y) {
4250: PetscCall(MatSolveTranspose(mat, b, x));
4251: PetscCall(VecAXPY(x, one, y));
4252: } else {
4253: PetscCall(VecDuplicate(x, &tmp));
4254: PetscCall(VecCopy(x, tmp));
4255: PetscCall(MatSolveTranspose(mat, b, x));
4256: PetscCall(VecAXPY(x, one, tmp));
4257: PetscCall(VecDestroy(&tmp));
4258: }
4259: }
4260: PetscCall(PetscLogEventEnd(MAT_SolveTransposeAdd, mat, b, x, y));
4261: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4262: PetscFunctionReturn(PETSC_SUCCESS);
4263: }
4265: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
4266: /*@
4267: MatSOR - Computes relaxation (SOR, Gauss-Seidel) sweeps.
4269: Neighbor-wise Collective
4271: Input Parameters:
4272: + mat - the matrix
4273: . b - the right-hand side
4274: . omega - the relaxation factor
4275: . flag - flag indicating the type of SOR (see below)
4276: . shift - diagonal shift
4277: . its - the number of iterations
4278: - lits - the number of local iterations
4280: Output Parameter:
4281: . x - the solution (can contain an initial guess, use option `SOR_ZERO_INITIAL_GUESS` to indicate no guess)
4283: SOR Flags:
4284: + `SOR_FORWARD_SWEEP` - forward SOR
4285: . `SOR_BACKWARD_SWEEP` - backward SOR
4286: . `SOR_SYMMETRIC_SWEEP` - SSOR (symmetric SOR)
4287: . `SOR_LOCAL_FORWARD_SWEEP` - local forward SOR
4288: . `SOR_LOCAL_BACKWARD_SWEEP` - local forward SOR
4289: . `SOR_LOCAL_SYMMETRIC_SWEEP` - local SSOR
4290: . `SOR_EISENSTAT` - SOR with Eisenstat trick
4291: . `SOR_APPLY_UPPER`, `SOR_APPLY_LOWER` - applies upper/lower triangular part of matrix to vector (with `omega`)
4292: - `SOR_ZERO_INITIAL_GUESS` - zero initial guess
4294: Level: developer
4296: Notes:
4297: `SOR_LOCAL_FORWARD_SWEEP`, `SOR_LOCAL_BACKWARD_SWEEP`, and
4298: `SOR_LOCAL_SYMMETRIC_SWEEP` perform separate independent smoothings
4299: on each processor.
4301: Application programmers will not generally use `MatSOR()` directly,
4302: but instead will employ `PCSOR` or `PCEISENSTAT`
4304: For `MATBAIJ`, `MATSBAIJ`, and `MATAIJ` matrices with inodes, this does a block SOR smoothing, otherwise it does a pointwise smoothing.
4305: For `MATAIJ` matrices with inodes, the block sizes are determined by the inode sizes, not the block size set with `MatSetBlockSize()`
4307: Vectors `x` and `b` CANNOT be the same
4309: The flags are implemented as bitwise inclusive or operations.
4310: For example, use (`SOR_ZERO_INITIAL_GUESS` | `SOR_SYMMETRIC_SWEEP`)
4311: to specify a zero initial guess for SSOR.
4313: Developer Note:
4314: We should add block SOR support for `MATAIJ` matrices with block size set to greater than one and no inodes
4316: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `KSP`, `PC`, `MatGetFactor()`
4317: @*/
4318: PetscErrorCode MatSOR(Mat mat, Vec b, PetscReal omega, MatSORType flag, PetscReal shift, PetscInt its, PetscInt lits, Vec x)
4319: {
4320: PetscFunctionBegin;
4325: PetscCheckSameComm(mat, 1, b, 2);
4326: PetscCheckSameComm(mat, 1, x, 8);
4327: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4328: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4329: PetscCheck(mat->cmap->N == x->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec x: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->cmap->N, x->map->N);
4330: PetscCheck(mat->rmap->N == b->map->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_SIZ, "Mat mat,Vec b: global dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->N, b->map->N);
4331: PetscCheck(mat->rmap->n == b->map->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Mat mat,Vec b: local dim %" PetscInt_FMT " %" PetscInt_FMT, mat->rmap->n, b->map->n);
4332: PetscCheck(its > 0, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Relaxation requires global its %" PetscInt_FMT " positive", its);
4333: PetscCheck(lits > 0, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Relaxation requires local its %" PetscInt_FMT " positive", lits);
4334: PetscCheck(b != x, PETSC_COMM_SELF, PETSC_ERR_ARG_IDN, "b and x vector cannot be the same");
4336: MatCheckPreallocated(mat, 1);
4337: PetscCall(PetscLogEventBegin(MAT_SOR, mat, b, x, 0));
4338: PetscUseTypeMethod(mat, sor, b, omega, flag, shift, its, lits, x);
4339: PetscCall(PetscLogEventEnd(MAT_SOR, mat, b, x, 0));
4340: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4341: PetscFunctionReturn(PETSC_SUCCESS);
4342: }
4344: /*
4345: Default matrix copy routine.
4346: */
4347: PetscErrorCode MatCopy_Basic(Mat A, Mat B, MatStructure str)
4348: {
4349: PetscInt i, rstart = 0, rend = 0, nz;
4350: const PetscInt *cwork;
4351: const PetscScalar *vwork;
4353: PetscFunctionBegin;
4354: if (B->assembled) PetscCall(MatZeroEntries(B));
4355: if (str == SAME_NONZERO_PATTERN) {
4356: PetscCall(MatGetOwnershipRange(A, &rstart, &rend));
4357: for (i = rstart; i < rend; i++) {
4358: PetscCall(MatGetRow(A, i, &nz, &cwork, &vwork));
4359: PetscCall(MatSetValues(B, 1, &i, nz, cwork, vwork, INSERT_VALUES));
4360: PetscCall(MatRestoreRow(A, i, &nz, &cwork, &vwork));
4361: }
4362: } else {
4363: PetscCall(MatAYPX(B, 0.0, A, str));
4364: }
4365: PetscCall(MatAssemblyBegin(B, MAT_FINAL_ASSEMBLY));
4366: PetscCall(MatAssemblyEnd(B, MAT_FINAL_ASSEMBLY));
4367: PetscFunctionReturn(PETSC_SUCCESS);
4368: }
4370: /*@
4371: MatCopy - Copies a matrix to another matrix.
4373: Collective
4375: Input Parameters:
4376: + A - the matrix
4377: - str - `SAME_NONZERO_PATTERN` or `DIFFERENT_NONZERO_PATTERN`
4379: Output Parameter:
4380: . B - where the copy is put
4382: Level: intermediate
4384: Notes:
4385: If you use `SAME_NONZERO_PATTERN`, then the two matrices must have the same nonzero pattern or the routine will crash.
4387: `MatCopy()` copies the matrix entries of a matrix to another existing
4388: matrix (after first zeroing the second matrix). A related routine is
4389: `MatConvert()`, which first creates a new matrix and then copies the data.
4391: .seealso: [](ch_matrices), `Mat`, `MatConvert()`, `MatDuplicate()`
4392: @*/
4393: PetscErrorCode MatCopy(Mat A, Mat B, MatStructure str)
4394: {
4395: PetscInt i;
4397: PetscFunctionBegin;
4402: PetscCheckSameComm(A, 1, B, 2);
4403: MatCheckPreallocated(B, 2);
4404: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4405: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4406: PetscCheck(A->rmap->N == B->rmap->N && A->cmap->N == B->cmap->N, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_SIZ, "Mat A,Mat B: global dim (%" PetscInt_FMT ",%" PetscInt_FMT ") (%" PetscInt_FMT ",%" PetscInt_FMT ")", A->rmap->N, B->rmap->N,
4407: A->cmap->N, B->cmap->N);
4408: MatCheckPreallocated(A, 1);
4409: if (A == B) PetscFunctionReturn(PETSC_SUCCESS);
4411: PetscCall(PetscLogEventBegin(MAT_Copy, A, B, 0, 0));
4412: if (A->ops->copy) PetscUseTypeMethod(A, copy, B, str);
4413: else PetscCall(MatCopy_Basic(A, B, str));
4415: B->stencil.dim = A->stencil.dim;
4416: B->stencil.noc = A->stencil.noc;
4417: for (i = 0; i <= A->stencil.dim + (A->stencil.noc ? 0 : -1); i++) {
4418: B->stencil.dims[i] = A->stencil.dims[i];
4419: B->stencil.starts[i] = A->stencil.starts[i];
4420: }
4422: PetscCall(PetscLogEventEnd(MAT_Copy, A, B, 0, 0));
4423: PetscCall(PetscObjectStateIncrease((PetscObject)B));
4424: PetscFunctionReturn(PETSC_SUCCESS);
4425: }
4427: /*@
4428: MatConvert - Converts a matrix to another matrix, either of the same
4429: or different type.
4431: Collective
4433: Input Parameters:
4434: + mat - the matrix
4435: . newtype - new matrix type. Use `MATSAME` to create a new matrix of the
4436: same type as the original matrix.
4437: - reuse - denotes if the destination matrix is to be created or reused.
4438: Use `MAT_INPLACE_MATRIX` for inplace conversion (that is when you want the input `Mat` to be changed to contain the matrix in the new format), otherwise use
4439: `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX` (can only be used after the first call was made with `MAT_INITIAL_MATRIX`, causes the matrix space in M to be reused).
4441: Output Parameter:
4442: . M - pointer to place new matrix
4444: Level: intermediate
4446: Notes:
4447: `MatConvert()` first creates a new matrix and then copies the data from
4448: the first matrix. A related routine is `MatCopy()`, which copies the matrix
4449: entries of one matrix to another already existing matrix context.
4451: Cannot be used to convert a sequential matrix to parallel or parallel to sequential,
4452: the MPI communicator of the generated matrix is always the same as the communicator
4453: of the input matrix.
4455: .seealso: [](ch_matrices), `Mat`, `MatCopy()`, `MatDuplicate()`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`
4456: @*/
4457: PetscErrorCode MatConvert(Mat mat, MatType newtype, MatReuse reuse, Mat *M)
4458: {
4459: PetscBool sametype, issame, flg;
4460: PetscBool3 issymmetric, ishermitian, isspd;
4461: char convname[256], mtype[256];
4462: Mat B;
4464: PetscFunctionBegin;
4467: PetscAssertPointer(M, 4);
4468: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4469: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4470: MatCheckPreallocated(mat, 1);
4472: PetscCall(PetscOptionsGetString(((PetscObject)mat)->options, ((PetscObject)mat)->prefix, "-matconvert_type", mtype, sizeof(mtype), &flg));
4473: if (flg) newtype = mtype;
4475: PetscCall(PetscObjectTypeCompare((PetscObject)mat, newtype, &sametype));
4476: PetscCall(PetscStrcmp(newtype, "same", &issame));
4477: PetscCheck(!(reuse == MAT_INPLACE_MATRIX) || !(mat != *M), PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "MAT_INPLACE_MATRIX requires same input and output matrix");
4478: if (reuse == MAT_REUSE_MATRIX) {
4480: PetscCheck(mat != *M, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "MAT_REUSE_MATRIX means reuse matrix in final argument, perhaps you mean MAT_INPLACE_MATRIX");
4481: }
4483: if ((reuse == MAT_INPLACE_MATRIX) && (issame || sametype)) {
4484: PetscCall(PetscInfo(mat, "Early return for inplace %s %d %d\n", ((PetscObject)mat)->type_name, sametype, issame));
4485: PetscFunctionReturn(PETSC_SUCCESS);
4486: }
4488: /* Cache Mat options because some converters use MatHeaderReplace() */
4489: issymmetric = mat->symmetric;
4490: ishermitian = mat->hermitian;
4491: isspd = mat->spd;
4493: if ((sametype || issame) && (reuse == MAT_INITIAL_MATRIX) && mat->ops->duplicate) {
4494: PetscCall(PetscInfo(mat, "Calling duplicate for initial matrix %s %d %d\n", ((PetscObject)mat)->type_name, sametype, issame));
4495: PetscUseTypeMethod(mat, duplicate, MAT_COPY_VALUES, M);
4496: } else {
4497: PetscErrorCode (*conv)(Mat, MatType, MatReuse, Mat *) = NULL;
4498: const char *prefix[3] = {"seq", "mpi", ""};
4499: PetscInt i;
4500: /*
4501: Order of precedence:
4502: 0) See if newtype is a superclass of the current matrix.
4503: 1) See if a specialized converter is known to the current matrix.
4504: 2) See if a specialized converter is known to the desired matrix class.
4505: 3) See if a good general converter is registered for the desired class
4506: (as of 6/27/03 only MATMPIADJ falls into this category).
4507: 4) See if a good general converter is known for the current matrix.
4508: 5) Use a really basic converter.
4509: */
4511: /* 0) See if newtype is a superclass of the current matrix.
4512: i.e mat is mpiaij and newtype is aij */
4513: for (i = 0; i < (PetscInt)PETSC_STATIC_ARRAY_LENGTH(prefix); i++) {
4514: PetscCall(PetscStrncpy(convname, prefix[i], sizeof(convname)));
4515: PetscCall(PetscStrlcat(convname, newtype, sizeof(convname)));
4516: PetscCall(PetscStrcmp(convname, ((PetscObject)mat)->type_name, &flg));
4517: PetscCall(PetscInfo(mat, "Check superclass %s %s -> %d\n", convname, ((PetscObject)mat)->type_name, flg));
4518: if (flg) {
4519: if (reuse == MAT_INPLACE_MATRIX) {
4520: PetscCall(PetscInfo(mat, "Early return\n"));
4521: PetscFunctionReturn(PETSC_SUCCESS);
4522: } else if (reuse == MAT_INITIAL_MATRIX && mat->ops->duplicate) {
4523: PetscCall(PetscInfo(mat, "Calling MatDuplicate\n"));
4524: PetscUseTypeMethod(mat, duplicate, MAT_COPY_VALUES, M);
4525: PetscFunctionReturn(PETSC_SUCCESS);
4526: } else if (reuse == MAT_REUSE_MATRIX && mat->ops->copy) {
4527: PetscCall(PetscInfo(mat, "Calling MatCopy\n"));
4528: PetscCall(MatCopy(mat, *M, SAME_NONZERO_PATTERN));
4529: PetscFunctionReturn(PETSC_SUCCESS);
4530: }
4531: }
4532: }
4533: /* 1) See if a specialized converter is known to the current matrix and the desired class */
4534: for (i = 0; i < (PetscInt)PETSC_STATIC_ARRAY_LENGTH(prefix); i++) {
4535: PetscCall(PetscStrncpy(convname, "MatConvert_", sizeof(convname)));
4536: PetscCall(PetscStrlcat(convname, ((PetscObject)mat)->type_name, sizeof(convname)));
4537: PetscCall(PetscStrlcat(convname, "_", sizeof(convname)));
4538: PetscCall(PetscStrlcat(convname, prefix[i], sizeof(convname)));
4539: PetscCall(PetscStrlcat(convname, issame ? ((PetscObject)mat)->type_name : newtype, sizeof(convname)));
4540: PetscCall(PetscStrlcat(convname, "_C", sizeof(convname)));
4541: PetscCall(PetscObjectQueryFunction((PetscObject)mat, convname, &conv));
4542: PetscCall(PetscInfo(mat, "Check specialized (1) %s (%s) -> %d\n", convname, ((PetscObject)mat)->type_name, !!conv));
4543: if (conv) goto foundconv;
4544: }
4546: /* 2) See if a specialized converter is known to the desired matrix class. */
4547: PetscCall(MatCreate(PetscObjectComm((PetscObject)mat), &B));
4548: PetscCall(MatSetSizes(B, mat->rmap->n, mat->cmap->n, mat->rmap->N, mat->cmap->N));
4549: PetscCall(MatSetType(B, newtype));
4550: for (i = 0; i < (PetscInt)PETSC_STATIC_ARRAY_LENGTH(prefix); i++) {
4551: PetscCall(PetscStrncpy(convname, "MatConvert_", sizeof(convname)));
4552: PetscCall(PetscStrlcat(convname, ((PetscObject)mat)->type_name, sizeof(convname)));
4553: PetscCall(PetscStrlcat(convname, "_", sizeof(convname)));
4554: PetscCall(PetscStrlcat(convname, prefix[i], sizeof(convname)));
4555: PetscCall(PetscStrlcat(convname, newtype, sizeof(convname)));
4556: PetscCall(PetscStrlcat(convname, "_C", sizeof(convname)));
4557: PetscCall(PetscObjectQueryFunction((PetscObject)B, convname, &conv));
4558: PetscCall(PetscInfo(mat, "Check specialized (2) %s (%s) -> %d\n", convname, ((PetscObject)B)->type_name, !!conv));
4559: if (conv) {
4560: PetscCall(MatDestroy(&B));
4561: goto foundconv;
4562: }
4563: }
4565: /* 3) See if a good general converter is registered for the desired class */
4566: conv = B->ops->convertfrom;
4567: PetscCall(PetscInfo(mat, "Check convertfrom (%s) -> %d\n", ((PetscObject)B)->type_name, !!conv));
4568: PetscCall(MatDestroy(&B));
4569: if (conv) goto foundconv;
4571: /* 4) See if a good general converter is known for the current matrix */
4572: if (mat->ops->convert) conv = mat->ops->convert;
4573: PetscCall(PetscInfo(mat, "Check general convert (%s) -> %d\n", ((PetscObject)mat)->type_name, !!conv));
4574: if (conv) goto foundconv;
4576: /* 5) Use a really basic converter. */
4577: PetscCall(PetscInfo(mat, "Using MatConvert_Basic\n"));
4578: conv = MatConvert_Basic;
4580: foundconv:
4581: PetscCall(PetscLogEventBegin(MAT_Convert, mat, 0, 0, 0));
4582: PetscCall((*conv)(mat, newtype, reuse, M));
4583: if (mat->rmap->mapping && mat->cmap->mapping && !(*M)->rmap->mapping && !(*M)->cmap->mapping) {
4584: /* the block sizes must be same if the mappings are copied over */
4585: (*M)->rmap->bs = mat->rmap->bs;
4586: (*M)->cmap->bs = mat->cmap->bs;
4587: PetscCall(PetscObjectReference((PetscObject)mat->rmap->mapping));
4588: PetscCall(PetscObjectReference((PetscObject)mat->cmap->mapping));
4589: (*M)->rmap->mapping = mat->rmap->mapping;
4590: (*M)->cmap->mapping = mat->cmap->mapping;
4591: }
4592: (*M)->stencil.dim = mat->stencil.dim;
4593: (*M)->stencil.noc = mat->stencil.noc;
4594: for (i = 0; i <= mat->stencil.dim + (mat->stencil.noc ? 0 : -1); i++) {
4595: (*M)->stencil.dims[i] = mat->stencil.dims[i];
4596: (*M)->stencil.starts[i] = mat->stencil.starts[i];
4597: }
4598: PetscCall(PetscLogEventEnd(MAT_Convert, mat, 0, 0, 0));
4599: }
4600: PetscCall(PetscObjectStateIncrease((PetscObject)*M));
4602: /* Reset Mat options */
4603: if (issymmetric != PETSC_BOOL3_UNKNOWN) PetscCall(MatSetOption(*M, MAT_SYMMETRIC, PetscBool3ToBool(issymmetric)));
4604: if (ishermitian != PETSC_BOOL3_UNKNOWN) PetscCall(MatSetOption(*M, MAT_HERMITIAN, PetscBool3ToBool(ishermitian)));
4605: if (isspd != PETSC_BOOL3_UNKNOWN) PetscCall(MatSetOption(*M, MAT_SPD, PetscBool3ToBool(isspd)));
4606: PetscFunctionReturn(PETSC_SUCCESS);
4607: }
4609: /*@
4610: MatFactorGetSolverType - Returns name of the package providing the factorization routines
4612: Not Collective
4614: Input Parameter:
4615: . mat - the matrix, must be a factored matrix
4617: Output Parameter:
4618: . type - the string name of the package (do not free this string)
4620: Level: intermediate
4622: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSolverType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`
4623: @*/
4624: PetscErrorCode MatFactorGetSolverType(Mat mat, MatSolverType *type)
4625: {
4626: PetscErrorCode (*conv)(Mat, MatSolverType *);
4628: PetscFunctionBegin;
4631: PetscAssertPointer(type, 2);
4632: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Only for factored matrix");
4633: PetscCall(PetscObjectQueryFunction((PetscObject)mat, "MatFactorGetSolverType_C", &conv));
4634: if (conv) PetscCall((*conv)(mat, type));
4635: else *type = MATSOLVERPETSC;
4636: PetscFunctionReturn(PETSC_SUCCESS);
4637: }
4639: typedef struct _MatSolverTypeForSpecifcType *MatSolverTypeForSpecifcType;
4640: struct _MatSolverTypeForSpecifcType {
4641: MatType mtype;
4642: /* no entry for MAT_FACTOR_NONE */
4643: PetscErrorCode (*createfactor[MAT_FACTOR_NUM_TYPES - 1])(Mat, MatFactorType, Mat *);
4644: MatSolverTypeForSpecifcType next;
4645: };
4647: typedef struct _MatSolverTypeHolder *MatSolverTypeHolder;
4648: struct _MatSolverTypeHolder {
4649: char *name;
4650: MatSolverTypeForSpecifcType handlers;
4651: MatSolverTypeHolder next;
4652: };
4654: static MatSolverTypeHolder MatSolverTypeHolders = NULL;
4656: /*@C
4657: MatSolverTypeRegister - Registers a `MatSolverType` that works for a particular matrix type
4659: Logically Collective, No Fortran Support
4661: Input Parameters:
4662: + package - name of the package, for example `petsc` or `superlu`
4663: . mtype - the matrix type that works with this package
4664: . ftype - the type of factorization supported by the package
4665: - createfactor - routine that will create the factored matrix ready to be used
4667: Level: developer
4669: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorGetSolverType()`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`,
4670: `MatGetFactor()`
4671: @*/
4672: PetscErrorCode MatSolverTypeRegister(MatSolverType package, MatType mtype, MatFactorType ftype, PetscErrorCode (*createfactor)(Mat, MatFactorType, Mat *))
4673: {
4674: MatSolverTypeHolder next = MatSolverTypeHolders, prev = NULL;
4675: PetscBool flg;
4676: MatSolverTypeForSpecifcType inext, iprev = NULL;
4678: PetscFunctionBegin;
4679: PetscCall(MatInitializePackage());
4680: if (!next) {
4681: PetscCall(PetscNew(&MatSolverTypeHolders));
4682: PetscCall(PetscStrallocpy(package, &MatSolverTypeHolders->name));
4683: PetscCall(PetscNew(&MatSolverTypeHolders->handlers));
4684: PetscCall(PetscStrallocpy(mtype, (char **)&MatSolverTypeHolders->handlers->mtype));
4685: MatSolverTypeHolders->handlers->createfactor[(int)ftype - 1] = createfactor;
4686: PetscFunctionReturn(PETSC_SUCCESS);
4687: }
4688: while (next) {
4689: PetscCall(PetscStrcasecmp(package, next->name, &flg));
4690: if (flg) {
4691: PetscCheck(next->handlers, PETSC_COMM_SELF, PETSC_ERR_PLIB, "MatSolverTypeHolder is missing handlers");
4692: inext = next->handlers;
4693: while (inext) {
4694: PetscCall(PetscStrcasecmp(mtype, inext->mtype, &flg));
4695: if (flg) {
4696: inext->createfactor[(int)ftype - 1] = createfactor;
4697: PetscFunctionReturn(PETSC_SUCCESS);
4698: }
4699: iprev = inext;
4700: inext = inext->next;
4701: }
4702: PetscCall(PetscNew(&iprev->next));
4703: PetscCall(PetscStrallocpy(mtype, (char **)&iprev->next->mtype));
4704: iprev->next->createfactor[(int)ftype - 1] = createfactor;
4705: PetscFunctionReturn(PETSC_SUCCESS);
4706: }
4707: prev = next;
4708: next = next->next;
4709: }
4710: PetscCall(PetscNew(&prev->next));
4711: PetscCall(PetscStrallocpy(package, &prev->next->name));
4712: PetscCall(PetscNew(&prev->next->handlers));
4713: PetscCall(PetscStrallocpy(mtype, (char **)&prev->next->handlers->mtype));
4714: prev->next->handlers->createfactor[(int)ftype - 1] = createfactor;
4715: PetscFunctionReturn(PETSC_SUCCESS);
4716: }
4718: /*@C
4719: MatSolverTypeGet - Gets the function that creates the factor matrix if it exist
4721: Input Parameters:
4722: + type - name of the package, for example `petsc` or `superlu`, if this is `NULL`, then the first result that satisfies the other criteria is returned
4723: . ftype - the type of factorization supported by the type
4724: - mtype - the matrix type that works with this type
4726: Output Parameters:
4727: + foundtype - `PETSC_TRUE` if the type was registered
4728: . foundmtype - `PETSC_TRUE` if the type supports the requested mtype
4729: - createfactor - routine that will create the factored matrix ready to be used or `NULL` if not found
4731: Calling sequence of `createfactor`:
4732: + A - the matrix providing the factor matrix
4733: . ftype - the `MatFactorType` of the factor requested
4734: - B - the new factor matrix that responds to MatXXFactorSymbolic,Numeric() functions, such as `MatLUFactorSymbolic()`
4736: Level: developer
4738: Note:
4739: When `type` is `NULL` the available functions are searched for based on the order of the calls to `MatSolverTypeRegister()` in `MatInitializePackage()`.
4740: Since different PETSc configurations may have different external solvers, seemingly identical runs with different PETSc configurations may use a different solver.
4741: For example if one configuration had `--download-mumps` while a different one had `--download-superlu_dist`.
4743: .seealso: [](ch_matrices), `Mat`, `MatFactorType`, `MatType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatSolverTypeRegister()`, `MatGetFactor()`,
4744: `MatInitializePackage()`
4745: @*/
4746: PetscErrorCode MatSolverTypeGet(MatSolverType type, MatType mtype, MatFactorType ftype, PetscBool *foundtype, PetscBool *foundmtype, PetscErrorCode (**createfactor)(Mat A, MatFactorType ftype, Mat *B))
4747: {
4748: MatSolverTypeHolder next = MatSolverTypeHolders;
4749: PetscBool flg;
4750: MatSolverTypeForSpecifcType inext;
4752: PetscFunctionBegin;
4753: if (foundtype) *foundtype = PETSC_FALSE;
4754: if (foundmtype) *foundmtype = PETSC_FALSE;
4755: if (createfactor) *createfactor = NULL;
4757: if (type) {
4758: while (next) {
4759: PetscCall(PetscStrcasecmp(type, next->name, &flg));
4760: if (flg) {
4761: if (foundtype) *foundtype = PETSC_TRUE;
4762: inext = next->handlers;
4763: while (inext) {
4764: PetscCall(PetscStrbeginswith(mtype, inext->mtype, &flg));
4765: if (flg) {
4766: if (foundmtype) *foundmtype = PETSC_TRUE;
4767: if (createfactor) *createfactor = inext->createfactor[(int)ftype - 1];
4768: PetscFunctionReturn(PETSC_SUCCESS);
4769: }
4770: inext = inext->next;
4771: }
4772: }
4773: next = next->next;
4774: }
4775: } else {
4776: while (next) {
4777: inext = next->handlers;
4778: while (inext) {
4779: PetscCall(PetscStrcmp(mtype, inext->mtype, &flg));
4780: if (flg && inext->createfactor[(int)ftype - 1]) {
4781: if (foundtype) *foundtype = PETSC_TRUE;
4782: if (foundmtype) *foundmtype = PETSC_TRUE;
4783: if (createfactor) *createfactor = inext->createfactor[(int)ftype - 1];
4784: PetscFunctionReturn(PETSC_SUCCESS);
4785: }
4786: inext = inext->next;
4787: }
4788: next = next->next;
4789: }
4790: /* try with base classes inext->mtype */
4791: next = MatSolverTypeHolders;
4792: while (next) {
4793: inext = next->handlers;
4794: while (inext) {
4795: PetscCall(PetscStrbeginswith(mtype, inext->mtype, &flg));
4796: if (flg && inext->createfactor[(int)ftype - 1]) {
4797: if (foundtype) *foundtype = PETSC_TRUE;
4798: if (foundmtype) *foundmtype = PETSC_TRUE;
4799: if (createfactor) *createfactor = inext->createfactor[(int)ftype - 1];
4800: PetscFunctionReturn(PETSC_SUCCESS);
4801: }
4802: inext = inext->next;
4803: }
4804: next = next->next;
4805: }
4806: }
4807: PetscFunctionReturn(PETSC_SUCCESS);
4808: }
4810: PetscErrorCode MatSolverTypeDestroy(void)
4811: {
4812: MatSolverTypeHolder next = MatSolverTypeHolders, prev;
4813: MatSolverTypeForSpecifcType inext, iprev;
4815: PetscFunctionBegin;
4816: while (next) {
4817: PetscCall(PetscFree(next->name));
4818: inext = next->handlers;
4819: while (inext) {
4820: PetscCall(PetscFree(inext->mtype));
4821: iprev = inext;
4822: inext = inext->next;
4823: PetscCall(PetscFree(iprev));
4824: }
4825: prev = next;
4826: next = next->next;
4827: PetscCall(PetscFree(prev));
4828: }
4829: MatSolverTypeHolders = NULL;
4830: PetscFunctionReturn(PETSC_SUCCESS);
4831: }
4833: /*@
4834: MatFactorGetCanUseOrdering - Indicates if the factorization can use the ordering provided in `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`
4836: Logically Collective
4838: Input Parameter:
4839: . mat - the matrix
4841: Output Parameter:
4842: . flg - `PETSC_TRUE` if uses the ordering
4844: Level: developer
4846: Note:
4847: Most internal PETSc factorizations use the ordering passed to the factorization routine but external
4848: packages do not, thus we want to skip generating the ordering when it is not needed or used.
4850: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`
4851: @*/
4852: PetscErrorCode MatFactorGetCanUseOrdering(Mat mat, PetscBool *flg)
4853: {
4854: PetscFunctionBegin;
4855: *flg = mat->canuseordering;
4856: PetscFunctionReturn(PETSC_SUCCESS);
4857: }
4859: /*@
4860: MatFactorGetPreferredOrdering - The preferred ordering for a particular matrix factor object
4862: Logically Collective
4864: Input Parameters:
4865: + mat - the matrix obtained with `MatGetFactor()`
4866: - ftype - the factorization type to be used
4868: Output Parameter:
4869: . otype - the preferred ordering type
4871: Level: developer
4873: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorType`, `MatOrderingType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`
4874: @*/
4875: PetscErrorCode MatFactorGetPreferredOrdering(Mat mat, MatFactorType ftype, MatOrderingType *otype)
4876: {
4877: PetscFunctionBegin;
4878: *otype = mat->preferredordering[ftype];
4879: PetscCheck(*otype, PETSC_COMM_SELF, PETSC_ERR_PLIB, "MatFactor did not have a preferred ordering");
4880: PetscFunctionReturn(PETSC_SUCCESS);
4881: }
4883: /*@
4884: MatGetFactor - Returns a matrix suitable to calls to routines such as `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`, `MatILUFactorSymbolic()`,
4885: `MatICCFactorSymbolic()`, `MatLUFactorNumeric()`, and `MatCholeskyFactorNumeric()`
4887: Collective
4889: Input Parameters:
4890: + mat - the matrix
4891: . type - name of solver type, for example, `superlu_dist`, `petsc` (to use PETSc's solver if it is available), if this is `NULL`, then the first result that satisfies
4892: the other criteria is returned
4893: - ftype - factor type, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`
4895: Output Parameter:
4896: . f - the factor matrix used with MatXXFactorSymbolic,Numeric() calls. Can be `NULL` in some cases, see notes below.
4898: Options Database Keys:
4899: + -pc_factor_mat_solver_type type - choose the type at run time. When using `KSP` solvers
4900: . -pc_factor_mat_factor_on_host (true|false) - do matrix factorization on host (with device matrices). Default is doing it on device
4901: - -pc_factor_mat_solve_on_host (true|false) - do matrix solve on host (with device matrices). Default is doing it on device
4903: Level: intermediate
4905: Notes:
4906: Some of the packages, such as MUMPS, have options for controlling the factorization, these are in the form `-prefix_mat_packagename_packageoption`
4907: (for example, `-mat_mumps_icntl_6 1`) where `prefix` is normally set automatically from the calling `KSP`/`PC`. If `MatGetFactor()` is called directly,
4908: without using a `PC`, one can set the prefix by
4909: calling `MatSetOptionsPrefixFactor()` on the originating matrix or `MatSetOptionsPrefix()` on the resulting factor matrix.
4911: Some PETSc matrix formats have alternative solvers available that are provided by alternative packages
4912: such as PaStiX, SuperLU_DIST, MUMPS etc. PETSc must have been configured to use the external solver,
4913: using the corresponding `./configure` option such as `--download-package` or `--with-package-dir`.
4915: When `type` is `NULL` the available results are searched for based on the order of the calls to `MatSolverTypeRegister()` in `MatInitializePackage()`.
4916: Since different PETSc configurations may have different external solvers, seemingly identical runs with different PETSc configurations may use a different solver.
4917: For example if one configuration had `--download-mumps` while a different one had `--download-superlu_dist`.
4919: The return matrix can be `NULL` if the requested factorization is not available, since some combinations of matrix types and factorization
4920: types registered with `MatSolverTypeRegister()` cannot be fully tested if not at runtime.
4922: Developer Note:
4923: This should actually be called `MatCreateFactor()` since it creates a new factor object
4925: The `MatGetFactor()` implementations should not be accessing the PETSc options database or making other decisions about solver options,
4926: that should be delayed until the later operations. This is to ensure the correct options prefix has been set in the factor matrix.
4928: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `KSP`, `MatSolverType`, `MatFactorType`, `MatCopy()`, `MatDuplicate()`,
4929: `MatGetFactorAvailable()`, `MatFactorGetCanUseOrdering()`, `MatSolverTypeRegister()`, `MatSolverTypeGet()`,
4930: `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`, `MatInitializePackage()`,
4931: `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`, `MatILUFactorSymbolic()`,
4932: `MatICCFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactorNumeric()`
4933: @*/
4934: PetscErrorCode MatGetFactor(Mat mat, MatSolverType type, MatFactorType ftype, Mat *f)
4935: {
4936: PetscBool foundtype, foundmtype, shell, hasop = PETSC_FALSE;
4937: PetscErrorCode (*conv)(Mat, MatFactorType, Mat *);
4939: PetscFunctionBegin;
4943: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4944: MatCheckPreallocated(mat, 1);
4946: PetscCall(MatIsShell(mat, &shell));
4947: if (shell) PetscCall(MatHasOperation(mat, MATOP_GET_FACTOR, &hasop));
4948: if (hasop) {
4949: PetscUseTypeMethod(mat, getfactor, type, ftype, f);
4950: PetscFunctionReturn(PETSC_SUCCESS);
4951: }
4953: PetscCall(MatSolverTypeGet(type, ((PetscObject)mat)->type_name, ftype, &foundtype, &foundmtype, &conv));
4954: if (!foundtype) {
4955: if (type) {
4956: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_MISSING_FACTOR, "Could not locate solver type %s for factorization type %s and matrix type %s. Perhaps you must ./configure with --download-%s", type, MatFactorTypes[ftype],
4957: ((PetscObject)mat)->type_name, type);
4958: } else {
4959: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_MISSING_FACTOR, "Could not locate a solver type for factorization type %s and matrix type %s.", MatFactorTypes[ftype], ((PetscObject)mat)->type_name);
4960: }
4961: }
4962: PetscCheck(foundmtype, PetscObjectComm((PetscObject)mat), PETSC_ERR_MISSING_FACTOR, "MatSolverType %s does not support matrix type %s", type, ((PetscObject)mat)->type_name);
4963: PetscCheck(conv, PetscObjectComm((PetscObject)mat), PETSC_ERR_MISSING_FACTOR, "MatSolverType %s does not support factorization type %s for matrix type %s", type, MatFactorTypes[ftype], ((PetscObject)mat)->type_name);
4965: PetscCall((*conv)(mat, ftype, f));
4966: if (mat->factorprefix) PetscCall(MatSetOptionsPrefix(*f, mat->factorprefix));
4967: PetscFunctionReturn(PETSC_SUCCESS);
4968: }
4970: /*@
4971: MatGetFactorAvailable - Returns a flag if matrix supports particular type and factor type
4973: Not Collective
4975: Input Parameters:
4976: + mat - the matrix
4977: . type - name of solver type, for example, `superlu`, `petsc` (to use PETSc's default)
4978: - ftype - factor type, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`
4980: Output Parameter:
4981: . flg - PETSC_TRUE if the factorization is available
4983: Level: intermediate
4985: Notes:
4986: Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
4987: such as pastix, superlu, mumps etc.
4989: PETSc must have been ./configure to use the external solver, using the option --download-package
4991: Developer Note:
4992: This should actually be called `MatCreateFactorAvailable()` since `MatGetFactor()` creates a new factor object
4994: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatSolverType`, `MatFactorType`, `MatGetFactor()`, `MatCopy()`, `MatDuplicate()`, `MatSolverTypeRegister()`,
4995: `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`, `MatSolverTypeGet()`
4996: @*/
4997: PetscErrorCode MatGetFactorAvailable(Mat mat, MatSolverType type, MatFactorType ftype, PetscBool *flg)
4998: {
4999: PetscErrorCode (*gconv)(Mat, MatFactorType, Mat *);
5001: PetscFunctionBegin;
5003: PetscAssertPointer(flg, 4);
5005: *flg = PETSC_FALSE;
5006: if (!((PetscObject)mat)->type_name) PetscFunctionReturn(PETSC_SUCCESS);
5008: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5009: MatCheckPreallocated(mat, 1);
5011: PetscCall(MatSolverTypeGet(type, ((PetscObject)mat)->type_name, ftype, NULL, NULL, &gconv));
5012: *flg = gconv ? PETSC_TRUE : PETSC_FALSE;
5013: PetscFunctionReturn(PETSC_SUCCESS);
5014: }
5016: /*@
5017: MatDuplicate - Duplicates a matrix including the non-zero structure.
5019: Collective
5021: Input Parameters:
5022: + mat - the matrix
5023: - op - One of `MAT_DO_NOT_COPY_VALUES`, `MAT_COPY_VALUES`, or `MAT_SHARE_NONZERO_PATTERN`.
5024: See the manual page for `MatDuplicateOption()` for an explanation of these options.
5026: Output Parameter:
5027: . M - pointer to place new matrix
5029: Level: intermediate
5031: Notes:
5032: You cannot change the nonzero pattern for the parent or child matrix later if you use `MAT_SHARE_NONZERO_PATTERN`.
5034: If `op` is not `MAT_COPY_VALUES` the numerical values in the new matrix are zeroed.
5036: May be called with an unassembled input `Mat` if `MAT_DO_NOT_COPY_VALUES` is used, in which case the output `Mat` is unassembled as well.
5038: When original mat is a product of matrix operation, e.g., an output of `MatMatMult()` or `MatCreateSubMatrix()`, only the matrix data structure of `mat`
5039: is duplicated and the internal data structures created for the reuse of previous matrix operations are not duplicated.
5040: User should not use `MatDuplicate()` to create new matrix `M` if `M` is intended to be reused as the product of matrix operation.
5042: .seealso: [](ch_matrices), `Mat`, `MatCopy()`, `MatConvert()`, `MatDuplicateOption`
5043: @*/
5044: PetscErrorCode MatDuplicate(Mat mat, MatDuplicateOption op, Mat *M)
5045: {
5046: Mat B;
5047: VecType vtype;
5048: PetscInt i;
5049: PetscObject dm, container_h, container_d;
5050: PetscErrorCodeFn *viewf;
5052: PetscFunctionBegin;
5055: PetscAssertPointer(M, 3);
5056: PetscCheck(op != MAT_COPY_VALUES || mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "MAT_COPY_VALUES not allowed for unassembled matrix");
5057: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5058: MatCheckPreallocated(mat, 1);
5060: PetscCall(PetscLogEventBegin(MAT_Convert, mat, 0, 0, 0));
5061: PetscUseTypeMethod(mat, duplicate, op, M);
5062: PetscCall(PetscLogEventEnd(MAT_Convert, mat, 0, 0, 0));
5063: B = *M;
5065: PetscCall(MatGetOperation(mat, MATOP_VIEW, &viewf));
5066: if (viewf) PetscCall(MatSetOperation(B, MATOP_VIEW, viewf));
5067: PetscCall(MatGetVecType(mat, &vtype));
5068: PetscCall(MatSetVecType(B, vtype));
5070: B->stencil.dim = mat->stencil.dim;
5071: B->stencil.noc = mat->stencil.noc;
5072: for (i = 0; i <= mat->stencil.dim + (mat->stencil.noc ? 0 : -1); i++) {
5073: B->stencil.dims[i] = mat->stencil.dims[i];
5074: B->stencil.starts[i] = mat->stencil.starts[i];
5075: }
5077: B->nooffproczerorows = mat->nooffproczerorows;
5078: B->nooffprocentries = mat->nooffprocentries;
5080: PetscCall(PetscObjectQuery((PetscObject)mat, "__PETSc_dm", &dm));
5081: if (dm) PetscCall(PetscObjectCompose((PetscObject)B, "__PETSc_dm", dm));
5082: PetscCall(PetscObjectQuery((PetscObject)mat, "__PETSc_MatCOOStruct_Host", &container_h));
5083: if (container_h) PetscCall(PetscObjectCompose((PetscObject)B, "__PETSc_MatCOOStruct_Host", container_h));
5084: PetscCall(PetscObjectQuery((PetscObject)mat, "__PETSc_MatCOOStruct_Device", &container_d));
5085: if (container_d) PetscCall(PetscObjectCompose((PetscObject)B, "__PETSc_MatCOOStruct_Device", container_d));
5086: if (op == MAT_COPY_VALUES) PetscCall(MatPropagateSymmetryOptions(mat, B));
5087: PetscCall(PetscObjectStateIncrease((PetscObject)B));
5088: PetscFunctionReturn(PETSC_SUCCESS);
5089: }
5091: /*@
5092: MatGetDiagonal - Gets the diagonal of a matrix as a `Vec`
5094: Logically Collective
5096: Input Parameter:
5097: . mat - the matrix
5099: Output Parameter:
5100: . v - the diagonal of the matrix
5102: Level: intermediate
5104: Note:
5105: If `mat` has local sizes `n` x `m`, this routine fills the first `ndiag = min(n, m)` entries
5106: of `v` with the diagonal values. Thus `v` must have local size of at least `ndiag`. If `v`
5107: is larger than `ndiag`, the values of the remaining entries are unspecified.
5109: Currently only correct in parallel for square matrices.
5111: .seealso: [](ch_matrices), `Mat`, `Vec`, `MatGetRow()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMaxAbs()`
5112: @*/
5113: PetscErrorCode MatGetDiagonal(Mat mat, Vec v)
5114: {
5115: PetscFunctionBegin;
5119: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5120: MatCheckPreallocated(mat, 1);
5121: if (PetscDefined(USE_DEBUG)) {
5122: PetscInt nv, row, col, ndiag;
5124: PetscCall(VecGetLocalSize(v, &nv));
5125: PetscCall(MatGetLocalSize(mat, &row, &col));
5126: ndiag = PetscMin(row, col);
5127: PetscCheck(nv >= ndiag, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Nonconforming Mat and Vec. Vec local size %" PetscInt_FMT " < Mat local diagonal length %" PetscInt_FMT, nv, ndiag);
5128: }
5130: PetscUseTypeMethod(mat, getdiagonal, v);
5131: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5132: PetscFunctionReturn(PETSC_SUCCESS);
5133: }
5135: /*@
5136: MatGetRowMin - Gets the minimum value (of the real part) of each
5137: row of the matrix
5139: Logically Collective
5141: Input Parameter:
5142: . mat - the matrix
5144: Output Parameters:
5145: + v - the vector for storing the maximums
5146: - idx - the indices of the column found for each row (optional, pass `NULL` if not needed)
5148: Level: intermediate
5150: Note:
5151: The result of this call are the same as if one converted the matrix to dense format
5152: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
5154: This code is only implemented for a couple of matrix formats.
5156: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMaxAbs()`, `MatGetRowMinAbs()`,
5157: `MatGetRowMax()`
5158: @*/
5159: PetscErrorCode MatGetRowMin(Mat mat, Vec v, PetscInt idx[])
5160: {
5161: PetscFunctionBegin;
5165: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5167: if (!mat->cmap->N) {
5168: PetscCall(VecSet(v, PETSC_MAX_REAL));
5169: if (idx) {
5170: PetscInt i, m = mat->rmap->n;
5171: for (i = 0; i < m; i++) idx[i] = -1;
5172: }
5173: } else {
5174: MatCheckPreallocated(mat, 1);
5175: }
5176: PetscUseTypeMethod(mat, getrowmin, v, idx);
5177: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5178: PetscFunctionReturn(PETSC_SUCCESS);
5179: }
5181: /*@
5182: MatGetRowMinAbs - Gets the minimum value (in absolute value) of each
5183: row of the matrix
5185: Logically Collective
5187: Input Parameter:
5188: . mat - the matrix
5190: Output Parameters:
5191: + v - the vector for storing the minimums
5192: - idx - the indices of the column found for each row (or `NULL` if not needed)
5194: Level: intermediate
5196: Notes:
5197: if a row is completely empty or has only 0.0 values, then the `idx` value for that
5198: row is 0 (the first column).
5200: This code is only implemented for a couple of matrix formats.
5202: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMax()`, `MatGetRowMaxAbs()`, `MatGetRowMin()`
5203: @*/
5204: PetscErrorCode MatGetRowMinAbs(Mat mat, Vec v, PetscInt idx[])
5205: {
5206: PetscFunctionBegin;
5210: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5211: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5213: if (!mat->cmap->N) {
5214: PetscCall(VecSet(v, 0.0));
5215: if (idx) {
5216: PetscInt i, m = mat->rmap->n;
5217: for (i = 0; i < m; i++) idx[i] = -1;
5218: }
5219: } else {
5220: MatCheckPreallocated(mat, 1);
5221: if (idx) PetscCall(PetscArrayzero(idx, mat->rmap->n));
5222: PetscUseTypeMethod(mat, getrowminabs, v, idx);
5223: }
5224: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5225: PetscFunctionReturn(PETSC_SUCCESS);
5226: }
5228: /*@
5229: MatGetRowMax - Gets the maximum value (of the real part) of each
5230: row of the matrix
5232: Logically Collective
5234: Input Parameter:
5235: . mat - the matrix
5237: Output Parameters:
5238: + v - the vector for storing the maximums
5239: - idx - the indices of the column found for each row (optional, otherwise pass `NULL`)
5241: Level: intermediate
5243: Notes:
5244: The result of this call are the same as if one converted the matrix to dense format
5245: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
5247: This code is only implemented for a couple of matrix formats.
5249: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMaxAbs()`, `MatGetRowMin()`, `MatGetRowMinAbs()`
5250: @*/
5251: PetscErrorCode MatGetRowMax(Mat mat, Vec v, PetscInt idx[])
5252: {
5253: PetscFunctionBegin;
5257: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5259: if (!mat->cmap->N) {
5260: PetscCall(VecSet(v, PETSC_MIN_REAL));
5261: if (idx) {
5262: PetscInt i, m = mat->rmap->n;
5263: for (i = 0; i < m; i++) idx[i] = -1;
5264: }
5265: } else {
5266: MatCheckPreallocated(mat, 1);
5267: PetscUseTypeMethod(mat, getrowmax, v, idx);
5268: }
5269: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5270: PetscFunctionReturn(PETSC_SUCCESS);
5271: }
5273: /*@
5274: MatGetRowMaxAbs - Gets the maximum value (in absolute value) of each
5275: row of the matrix
5277: Logically Collective
5279: Input Parameter:
5280: . mat - the matrix
5282: Output Parameters:
5283: + v - the vector for storing the maximums
5284: - idx - the indices of the column found for each row (or `NULL` if not needed)
5286: Level: intermediate
5288: Notes:
5289: if a row is completely empty or has only 0.0 values, then the `idx` value for that
5290: row is 0 (the first column).
5292: This code is only implemented for a couple of matrix formats.
5294: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowSum()`, `MatGetRowMin()`, `MatGetRowMinAbs()`
5295: @*/
5296: PetscErrorCode MatGetRowMaxAbs(Mat mat, Vec v, PetscInt idx[])
5297: {
5298: PetscFunctionBegin;
5302: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5304: if (!mat->cmap->N) {
5305: PetscCall(VecSet(v, 0.0));
5306: if (idx) {
5307: PetscInt i, m = mat->rmap->n;
5308: for (i = 0; i < m; i++) idx[i] = -1;
5309: }
5310: } else {
5311: MatCheckPreallocated(mat, 1);
5312: if (idx) PetscCall(PetscArrayzero(idx, mat->rmap->n));
5313: PetscUseTypeMethod(mat, getrowmaxabs, v, idx);
5314: }
5315: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5316: PetscFunctionReturn(PETSC_SUCCESS);
5317: }
5319: /*@
5320: MatGetRowSumAbs - Gets the sum value (in absolute value) of each row of the matrix
5322: Logically Collective
5324: Input Parameter:
5325: . mat - the matrix
5327: Output Parameter:
5328: . v - the vector for storing the sum
5330: Level: intermediate
5332: This code is only implemented for a couple of matrix formats.
5334: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMax()`, `MatGetRowMin()`, `MatGetRowMinAbs()`
5335: @*/
5336: PetscErrorCode MatGetRowSumAbs(Mat mat, Vec v)
5337: {
5338: PetscFunctionBegin;
5342: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5344: if (!mat->cmap->N) PetscCall(VecSet(v, 0.0));
5345: else {
5346: MatCheckPreallocated(mat, 1);
5347: PetscUseTypeMethod(mat, getrowsumabs, v);
5348: }
5349: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5350: PetscFunctionReturn(PETSC_SUCCESS);
5351: }
5353: /*@
5354: MatGetRowSum - Gets the sum of each row of the matrix
5356: Logically or Neighborhood Collective
5358: Input Parameter:
5359: . mat - the matrix
5361: Output Parameter:
5362: . v - the vector for storing the sum of rows
5364: Level: intermediate
5366: Note:
5367: This code is slow since it is not currently specialized for different formats
5369: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMax()`, `MatGetRowMin()`, `MatGetRowMaxAbs()`, `MatGetRowMinAbs()`, `MatGetRowSumAbs()`
5370: @*/
5371: PetscErrorCode MatGetRowSum(Mat mat, Vec v)
5372: {
5373: Vec ones;
5375: PetscFunctionBegin;
5379: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5380: MatCheckPreallocated(mat, 1);
5381: PetscCall(MatCreateVecs(mat, &ones, NULL));
5382: PetscCall(VecSet(ones, 1.));
5383: PetscCall(MatMult(mat, ones, v));
5384: PetscCall(VecDestroy(&ones));
5385: PetscFunctionReturn(PETSC_SUCCESS);
5386: }
5388: /*@
5389: MatTransposeSetPrecursor - Set the matrix from which the second matrix will receive numerical transpose data with a call to `MatTranspose`(A,`MAT_REUSE_MATRIX`,&B)
5390: when B was not obtained with `MatTranspose`(A,`MAT_INITIAL_MATRIX`,&B)
5392: Collective
5394: Input Parameter:
5395: . mat - the matrix to provide the transpose
5397: Output Parameter:
5398: . B - the matrix to contain the transpose; it MUST have the nonzero structure of the transpose of A or the code will crash or generate incorrect results
5400: Level: advanced
5402: Note:
5403: Normally the use of `MatTranspose`(A, `MAT_REUSE_MATRIX`, &B) requires that `B` was obtained with a call to `MatTranspose`(A, `MAT_INITIAL_MATRIX`, &B). This
5404: routine allows bypassing that call.
5406: .seealso: [](ch_matrices), `Mat`, `MatTransposeSymbolic()`, `MatTranspose()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`
5407: @*/
5408: PetscErrorCode MatTransposeSetPrecursor(Mat mat, Mat B)
5409: {
5410: MatParentState *rb = NULL;
5412: PetscFunctionBegin;
5413: PetscCall(PetscNew(&rb));
5414: rb->id = ((PetscObject)mat)->id;
5415: rb->state = 0;
5416: PetscCall(MatGetNonzeroState(mat, &rb->nonzerostate));
5417: PetscCall(PetscObjectContainerCompose((PetscObject)B, "MatTransposeParent", rb, PetscCtxDestroyDefault));
5418: PetscFunctionReturn(PETSC_SUCCESS);
5419: }
5421: static PetscErrorCode MatTranspose_Private(Mat mat, MatReuse reuse, Mat *B, PetscBool conjugate)
5422: {
5423: PetscContainer rB = NULL;
5424: MatParentState *rb = NULL;
5425: PetscErrorCode (*f)(Mat, MatReuse, Mat *) = NULL;
5427: PetscFunctionBegin;
5430: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5431: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5432: PetscCheck(reuse != MAT_INPLACE_MATRIX || mat == *B, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "MAT_INPLACE_MATRIX requires last matrix to match first");
5433: PetscCheck(reuse != MAT_REUSE_MATRIX || mat != *B, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Perhaps you mean MAT_INPLACE_MATRIX");
5434: MatCheckPreallocated(mat, 1);
5435: if (reuse == MAT_REUSE_MATRIX) {
5436: PetscCall(PetscObjectQuery((PetscObject)*B, "MatTransposeParent", (PetscObject *)&rB));
5437: PetscCheck(rB, PetscObjectComm((PetscObject)*B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from call to MatTranspose(). Suggest MatTransposeSetPrecursor().");
5438: PetscCall(PetscContainerGetPointer(rB, &rb));
5439: PetscCheck(rb->id == ((PetscObject)mat)->id, PetscObjectComm((PetscObject)*B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from input matrix");
5440: if (rb->state == ((PetscObject)mat)->state) PetscFunctionReturn(PETSC_SUCCESS);
5441: }
5443: if (conjugate) {
5444: f = mat->ops->hermitiantranspose;
5445: if (f) PetscCall((*f)(mat, reuse, B));
5446: }
5447: if (!f && !(reuse == MAT_INPLACE_MATRIX && mat->hermitian == PETSC_BOOL3_TRUE && conjugate)) {
5448: PetscCall(PetscLogEventBegin(MAT_Transpose, mat, 0, 0, 0));
5449: if (reuse != MAT_INPLACE_MATRIX || mat->symmetric != PETSC_BOOL3_TRUE) {
5450: PetscUseTypeMethod(mat, transpose, reuse, B);
5451: PetscCall(PetscObjectStateIncrease((PetscObject)*B));
5452: }
5453: PetscCall(PetscLogEventEnd(MAT_Transpose, mat, 0, 0, 0));
5454: if (conjugate) PetscCall(MatConjugate(*B));
5455: }
5457: if (reuse == MAT_INITIAL_MATRIX) PetscCall(MatTransposeSetPrecursor(mat, *B));
5458: if (reuse != MAT_INPLACE_MATRIX) {
5459: PetscCall(PetscObjectQuery((PetscObject)*B, "MatTransposeParent", (PetscObject *)&rB));
5460: PetscCall(PetscContainerGetPointer(rB, &rb));
5461: rb->state = ((PetscObject)mat)->state;
5462: rb->nonzerostate = mat->nonzerostate;
5463: }
5464: PetscFunctionReturn(PETSC_SUCCESS);
5465: }
5467: /*@
5468: MatTranspose - Computes the transpose of a matrix, either in-place or out-of-place.
5470: Collective
5472: Input Parameters:
5473: + mat - the matrix to transpose
5474: - reuse - either `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, or `MAT_INPLACE_MATRIX`
5476: Output Parameter:
5477: . B - the transpose of the matrix
5479: Level: intermediate
5481: Notes:
5482: If you use `MAT_INPLACE_MATRIX` then you must pass in `&mat` for `B`
5484: `MAT_REUSE_MATRIX` uses the `B` matrix obtained from a previous call to this function with `MAT_INITIAL_MATRIX` to store the transpose. If you already have a matrix to contain the
5485: transpose, call `MatTransposeSetPrecursor(mat, B)` before calling this routine.
5487: If the nonzero structure of `mat` changed from the previous call to this function with the same matrices an error will be generated for some matrix types.
5489: Consider using `MatCreateTranspose()` instead if you only need a matrix that behaves like the transpose but don't need the storage to be changed.
5490: For example, the result of `MatCreateTranspose()` will compute the transpose of the given matrix times a vector for matrix-vector products computed with `MatMult()`.
5492: If `mat` is unchanged from the last call this function returns immediately without recomputing the result
5494: If you only need the symbolic transpose of a matrix, and not the numerical values, use `MatTransposeSymbolic()`
5496: .seealso: [](ch_matrices), `Mat`, `MatTransposeSetPrecursor()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`,
5497: `MatTransposeSymbolic()`, `MatCreateTranspose()`
5498: @*/
5499: PetscErrorCode MatTranspose(Mat mat, MatReuse reuse, Mat *B)
5500: {
5501: PetscFunctionBegin;
5502: PetscCall(MatTranspose_Private(mat, reuse, B, PETSC_FALSE));
5503: PetscFunctionReturn(PETSC_SUCCESS);
5504: }
5506: /*@
5507: MatTransposeSymbolic - Computes the symbolic part of the transpose of a matrix.
5509: Collective
5511: Input Parameter:
5512: . A - the matrix to transpose
5514: Output Parameter:
5515: . B - the transpose. This is a complete matrix but the numerical portion is invalid. One can call `MatTranspose`(A,`MAT_REUSE_MATRIX`,&B) to compute the
5516: numerical portion.
5518: Level: intermediate
5520: Note:
5521: This is not supported for many matrix types, use `MatTranspose()` in those cases
5523: .seealso: [](ch_matrices), `Mat`, `MatTransposeSetPrecursor()`, `MatTranspose()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`
5524: @*/
5525: PetscErrorCode MatTransposeSymbolic(Mat A, Mat *B)
5526: {
5527: PetscFunctionBegin;
5530: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5531: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5532: PetscCall(PetscLogEventBegin(MAT_Transpose, A, 0, 0, 0));
5533: PetscUseTypeMethod(A, transposesymbolic, B);
5534: PetscCall(PetscLogEventEnd(MAT_Transpose, A, 0, 0, 0));
5536: PetscCall(MatTransposeSetPrecursor(A, *B));
5537: PetscFunctionReturn(PETSC_SUCCESS);
5538: }
5540: PetscErrorCode MatTransposeCheckNonzeroState_Private(Mat A, Mat B)
5541: {
5542: PetscContainer rB;
5543: MatParentState *rb;
5545: PetscFunctionBegin;
5548: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5549: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5550: PetscCall(PetscObjectQuery((PetscObject)B, "MatTransposeParent", (PetscObject *)&rB));
5551: PetscCheck(rB, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from call to MatTranspose()");
5552: PetscCall(PetscContainerGetPointer(rB, &rb));
5553: PetscCheck(rb->id == ((PetscObject)A)->id, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from input matrix");
5554: PetscCheck(rb->nonzerostate == A->nonzerostate, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONGSTATE, "Reuse matrix has changed nonzero structure");
5555: PetscFunctionReturn(PETSC_SUCCESS);
5556: }
5558: /*@
5559: MatIsTranspose - Test whether a matrix is another one's transpose,
5560: or its own, in which case it tests symmetry.
5562: Collective
5564: Input Parameters:
5565: + A - the matrix to test
5566: . B - the matrix to test against, this can equal the first parameter
5567: - tol - tolerance, differences between entries smaller than this are counted as zero
5569: Output Parameter:
5570: . flg - the result
5572: Level: intermediate
5574: Notes:
5575: The sequential algorithm has a running time of the order of the number of nonzeros; the parallel
5576: test involves parallel copies of the block off-diagonal parts of the matrix.
5578: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatIsSymmetric()`, `MatIsHermitian()`
5579: @*/
5580: PetscErrorCode MatIsTranspose(Mat A, Mat B, PetscReal tol, PetscBool *flg)
5581: {
5582: PetscErrorCode (*f)(Mat, Mat, PetscReal, PetscBool *), (*g)(Mat, Mat, PetscReal, PetscBool *);
5584: PetscFunctionBegin;
5587: PetscAssertPointer(flg, 4);
5588: PetscCall(PetscObjectQueryFunction((PetscObject)A, "MatIsTranspose_C", &f));
5589: PetscCall(PetscObjectQueryFunction((PetscObject)B, "MatIsTranspose_C", &g));
5590: *flg = PETSC_FALSE;
5591: if (f && g) {
5592: PetscCheck(f == g, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_NOTSAMETYPE, "Matrices do not have the same comparator for symmetry test");
5593: PetscCall((*f)(A, B, tol, flg));
5594: } else {
5595: MatType mattype;
5597: PetscCall(MatGetType(f ? B : A, &mattype));
5598: SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "Matrix of type %s does not support checking for transpose", mattype);
5599: }
5600: PetscFunctionReturn(PETSC_SUCCESS);
5601: }
5603: /*@
5604: MatHermitianTranspose - Computes an in-place or out-of-place Hermitian transpose of a matrix in complex conjugate.
5606: Collective
5608: Input Parameters:
5609: + mat - the matrix to transpose and complex conjugate
5610: - reuse - either `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, or `MAT_INPLACE_MATRIX`
5612: Output Parameter:
5613: . B - the Hermitian transpose
5615: Level: intermediate
5617: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`
5618: @*/
5619: PetscErrorCode MatHermitianTranspose(Mat mat, MatReuse reuse, Mat *B)
5620: {
5621: PetscFunctionBegin;
5622: PetscCall(MatTranspose_Private(mat, reuse, B, PETSC_TRUE));
5623: PetscFunctionReturn(PETSC_SUCCESS);
5624: }
5626: /*@
5627: MatIsHermitianTranspose - Test whether a matrix is another one's Hermitian transpose,
5629: Collective
5631: Input Parameters:
5632: + A - the matrix to test
5633: . B - the matrix to test against, this can equal the first parameter
5634: - tol - tolerance, differences between entries smaller than this are counted as zero
5636: Output Parameter:
5637: . flg - the result
5639: Level: intermediate
5641: Notes:
5642: Only available for `MATAIJ` matrices.
5644: The sequential algorithm
5645: has a running time of the order of the number of nonzeros; the parallel
5646: test involves parallel copies of the block off-diagonal parts of the matrix.
5648: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatIsSymmetric()`, `MatIsHermitian()`, `MatIsTranspose()`
5649: @*/
5650: PetscErrorCode MatIsHermitianTranspose(Mat A, Mat B, PetscReal tol, PetscBool *flg)
5651: {
5652: PetscErrorCode (*f)(Mat, Mat, PetscReal, PetscBool *), (*g)(Mat, Mat, PetscReal, PetscBool *);
5654: PetscFunctionBegin;
5657: PetscAssertPointer(flg, 4);
5658: PetscCall(PetscObjectQueryFunction((PetscObject)A, "MatIsHermitianTranspose_C", &f));
5659: PetscCall(PetscObjectQueryFunction((PetscObject)B, "MatIsHermitianTranspose_C", &g));
5660: if (f && g) {
5661: PetscCheck(f == g, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_NOTSAMETYPE, "Matrices do not have the same comparator for Hermitian test");
5662: PetscCall((*f)(A, B, tol, flg));
5663: } else {
5664: MatType mattype;
5666: PetscCall(MatGetType(f ? B : A, &mattype));
5667: SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "Matrix of type %s does not support checking for Hermitian transpose", mattype);
5668: }
5669: PetscFunctionReturn(PETSC_SUCCESS);
5670: }
5672: /*@
5673: MatPermute - Creates a new matrix with rows and columns permuted from the
5674: original.
5676: Collective
5678: Input Parameters:
5679: + mat - the matrix to permute
5680: . row - row permutation, each processor supplies only the permutation for its rows
5681: - col - column permutation, each processor supplies only the permutation for its columns
5683: Output Parameter:
5684: . B - the permuted matrix
5686: Level: advanced
5688: Note:
5689: The index sets map from row/col of permuted matrix to row/col of original matrix.
5690: The index sets should be on the same communicator as mat and have the same local sizes.
5692: Developer Note:
5693: If you want to implement `MatPermute()` for a matrix type, and your approach doesn't
5694: exploit the fact that row and col are permutations, consider implementing the
5695: more general `MatCreateSubMatrix()` instead.
5697: .seealso: [](ch_matrices), `Mat`, `MatGetOrdering()`, `ISAllGather()`, `MatCreateSubMatrix()`
5698: @*/
5699: PetscErrorCode MatPermute(Mat mat, IS row, IS col, Mat *B)
5700: {
5701: PetscFunctionBegin;
5706: PetscAssertPointer(B, 4);
5707: PetscCheckSameComm(mat, 1, row, 2);
5708: if (row != col) PetscCheckSameComm(row, 2, col, 3);
5709: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5710: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5711: PetscCheck(mat->ops->permute || mat->ops->createsubmatrix, PETSC_COMM_SELF, PETSC_ERR_SUP, "MatPermute not available for Mat type %s", ((PetscObject)mat)->type_name);
5712: MatCheckPreallocated(mat, 1);
5714: if (mat->ops->permute) {
5715: PetscUseTypeMethod(mat, permute, row, col, B);
5716: PetscCall(PetscObjectStateIncrease((PetscObject)*B));
5717: } else {
5718: PetscCall(MatCreateSubMatrix(mat, row, col, MAT_INITIAL_MATRIX, B));
5719: }
5720: PetscFunctionReturn(PETSC_SUCCESS);
5721: }
5723: /*@
5724: MatEqual - Compares two matrices.
5726: Collective
5728: Input Parameters:
5729: + A - the first matrix
5730: - B - the second matrix
5732: Output Parameter:
5733: . flg - `PETSC_TRUE` if the matrices are equal; `PETSC_FALSE` otherwise.
5735: Level: intermediate
5737: Note:
5738: If either of the matrix is "matrix-free", meaning the matrix entries are not stored explicitly then equality is determined by comparing
5739: the results of several matrix-vector product using randomly created vectors, see `MatMultEqual()`.
5741: .seealso: [](ch_matrices), `Mat`, `MatMultEqual()`
5742: @*/
5743: PetscErrorCode MatEqual(Mat A, Mat B, PetscBool *flg)
5744: {
5745: PetscFunctionBegin;
5750: PetscAssertPointer(flg, 3);
5751: PetscCheckSameComm(A, 1, B, 2);
5752: MatCheckPreallocated(A, 1);
5753: MatCheckPreallocated(B, 2);
5754: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5755: PetscCheck(B->assembled, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5756: PetscCheck(A->rmap->N == B->rmap->N && A->cmap->N == B->cmap->N, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_SIZ, "Mat A,Mat B: global dim %" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT, A->rmap->N, B->rmap->N, A->cmap->N,
5757: B->cmap->N);
5758: if (A->ops->equal && A->ops->equal == B->ops->equal) PetscUseTypeMethod(A, equal, B, flg);
5759: else PetscCall(MatMultEqual(A, B, 10, flg));
5760: PetscFunctionReturn(PETSC_SUCCESS);
5761: }
5763: /*@
5764: MatDiagonalScale - Scales a matrix on the left and right by diagonal
5765: matrices that are stored as vectors. Either of the two scaling
5766: matrices can be `NULL`.
5768: Collective
5770: Input Parameters:
5771: + mat - the matrix to be scaled
5772: . l - the left scaling vector (or `NULL`)
5773: - r - the right scaling vector (or `NULL`)
5775: Level: intermediate
5777: Note:
5778: `MatDiagonalScale()` computes $A = LAR$, where
5779: L = a diagonal matrix (stored as a vector), R = a diagonal matrix (stored as a vector)
5780: The L scales the rows of the matrix, the R scales the columns of the matrix.
5782: .seealso: [](ch_matrices), `Mat`, `MatScale()`, `MatShift()`, `MatDiagonalSet()`
5783: @*/
5784: PetscErrorCode MatDiagonalScale(Mat mat, Vec l, Vec r)
5785: {
5786: PetscBool flg = PETSC_FALSE;
5788: PetscFunctionBegin;
5791: if (l) {
5793: PetscCheckSameComm(mat, 1, l, 2);
5794: }
5795: if (r) {
5797: PetscCheckSameComm(mat, 1, r, 3);
5798: }
5799: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5800: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5801: MatCheckPreallocated(mat, 1);
5802: if (!l && !r) PetscFunctionReturn(PETSC_SUCCESS);
5804: PetscCall(PetscLogEventBegin(MAT_Scale, mat, 0, 0, 0));
5805: PetscUseTypeMethod(mat, diagonalscale, l, r);
5806: PetscCall(PetscLogEventEnd(MAT_Scale, mat, 0, 0, 0));
5807: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
5808: if (l != r && (PetscBool3ToBool(mat->symmetric) || PetscBool3ToBool(mat->hermitian))) {
5809: if (!PetscDefined(USE_COMPLEX) || PetscBool3ToBool(mat->symmetric)) {
5810: if (l && r) PetscCall(VecEqual(l, r, &flg));
5811: if (!flg) {
5812: PetscCall(PetscObjectTypeCompareAny((PetscObject)mat, &flg, MATSEQSBAIJ, MATMPISBAIJ, ""));
5813: PetscCheck(!flg, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "For symmetric format, left and right scaling vectors must be the same");
5814: mat->symmetric = mat->spd = PETSC_BOOL3_FALSE;
5815: if (!PetscDefined(USE_COMPLEX)) mat->hermitian = PETSC_BOOL3_FALSE;
5816: else mat->hermitian = PETSC_BOOL3_UNKNOWN;
5817: }
5818: }
5819: if (PetscDefined(USE_COMPLEX) && PetscBool3ToBool(mat->hermitian)) {
5820: flg = PETSC_FALSE;
5821: if (l && r) {
5822: Vec conjugate;
5824: PetscCall(VecDuplicate(l, &conjugate));
5825: PetscCall(VecCopy(l, conjugate));
5826: PetscCall(VecConjugate(conjugate));
5827: PetscCall(VecEqual(conjugate, r, &flg));
5828: PetscCall(VecDestroy(&conjugate));
5829: }
5830: if (!flg) {
5831: PetscCall(PetscObjectTypeCompareAny((PetscObject)mat, &flg, MATSEQSBAIJ, MATMPISBAIJ, ""));
5832: PetscCheck(!flg, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "For symmetric format and Hermitian matrix, left and right scaling vectors must be conjugate one of the other");
5833: mat->hermitian = PETSC_BOOL3_FALSE;
5834: mat->symmetric = mat->spd = PETSC_BOOL3_UNKNOWN;
5835: }
5836: }
5837: }
5838: PetscFunctionReturn(PETSC_SUCCESS);
5839: }
5841: /*@
5842: MatScale - Scales all elements of a matrix by a given number.
5844: Logically Collective
5846: Input Parameters:
5847: + mat - the matrix to be scaled
5848: - a - the scaling value
5850: Level: intermediate
5852: .seealso: [](ch_matrices), `Mat`, `MatDiagonalScale()`
5853: @*/
5854: PetscErrorCode MatScale(Mat mat, PetscScalar a)
5855: {
5856: PetscFunctionBegin;
5859: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5860: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5862: MatCheckPreallocated(mat, 1);
5864: PetscCall(PetscLogEventBegin(MAT_Scale, mat, 0, 0, 0));
5865: if (a != (PetscScalar)1.0) {
5866: PetscUseTypeMethod(mat, scale, a);
5867: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
5868: }
5869: PetscCall(PetscLogEventEnd(MAT_Scale, mat, 0, 0, 0));
5870: PetscFunctionReturn(PETSC_SUCCESS);
5871: }
5873: /*@
5874: MatNorm - Calculates various norms of a matrix.
5876: Collective
5878: Input Parameters:
5879: + mat - the matrix
5880: - type - the type of norm, `NORM_1`, `NORM_FROBENIUS`, `NORM_INFINITY`
5882: Output Parameter:
5883: . nrm - the resulting norm
5885: Level: intermediate
5887: .seealso: [](ch_matrices), `Mat`
5888: @*/
5889: PetscErrorCode MatNorm(Mat mat, NormType type, PetscReal *nrm)
5890: {
5891: PetscFunctionBegin;
5894: PetscAssertPointer(nrm, 3);
5896: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5897: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5898: MatCheckPreallocated(mat, 1);
5900: PetscUseTypeMethod(mat, norm, type, nrm);
5901: PetscFunctionReturn(PETSC_SUCCESS);
5902: }
5904: /*
5905: This variable is used to prevent counting of MatAssemblyBegin() that
5906: are called from within a MatAssemblyEnd().
5907: */
5908: static PetscInt MatAssemblyEnd_InUse = 0;
5909: /*@
5910: MatAssemblyBegin - Begins assembling the matrix. This routine should
5911: be called after completing all calls to `MatSetValues()`.
5913: Collective
5915: Input Parameters:
5916: + mat - the matrix
5917: - type - type of assembly, either `MAT_FLUSH_ASSEMBLY` or `MAT_FINAL_ASSEMBLY`
5919: Level: beginner
5921: Notes:
5922: `MatSetValues()` generally caches the values that belong to other MPI processes. The matrix is ready to
5923: use only after `MatAssemblyBegin()` and `MatAssemblyEnd()` have been called.
5925: Use `MAT_FLUSH_ASSEMBLY` when switching between `ADD_VALUES` and `INSERT_VALUES`
5926: in `MatSetValues()`; use `MAT_FINAL_ASSEMBLY` for the final assembly before
5927: using the matrix.
5929: ALL processes that share a matrix MUST call `MatAssemblyBegin()` and `MatAssemblyEnd()` the SAME NUMBER of times, and each time with the
5930: same flag of `MAT_FLUSH_ASSEMBLY` or `MAT_FINAL_ASSEMBLY` for all processes. Thus you CANNOT locally change from `ADD_VALUES` to `INSERT_VALUES`, that is
5931: a global collective operation requiring all processes that share the matrix.
5933: Space for preallocated nonzeros that is not filled by a call to `MatSetValues()` or a related routine are compressed
5934: out by assembly. If you intend to use that extra space on a subsequent assembly, be sure to insert explicit zeros
5935: before `MAT_FINAL_ASSEMBLY` so the space is not compressed out.
5937: .seealso: [](ch_matrices), `Mat`, `MatAssemblyEnd()`, `MatSetValues()`, `MatAssembled()`
5938: @*/
5939: PetscErrorCode MatAssemblyBegin(Mat mat, MatAssemblyType type)
5940: {
5941: PetscFunctionBegin;
5944: MatCheckPreallocated(mat, 1);
5945: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix. Did you forget to call MatSetUnfactored()?");
5946: if (mat->assembled) {
5947: mat->was_assembled = PETSC_TRUE;
5948: mat->assembled = PETSC_FALSE;
5949: }
5951: if (!MatAssemblyEnd_InUse) {
5952: PetscCall(PetscLogEventBegin(MAT_AssemblyBegin, mat, 0, 0, 0));
5953: PetscTryTypeMethod(mat, assemblybegin, type);
5954: PetscCall(PetscLogEventEnd(MAT_AssemblyBegin, mat, 0, 0, 0));
5955: } else PetscTryTypeMethod(mat, assemblybegin, type);
5956: PetscFunctionReturn(PETSC_SUCCESS);
5957: }
5959: /*@
5960: MatAssembled - Indicates if a matrix has been assembled and is ready for
5961: use; for example, in matrix-vector product.
5963: Not Collective
5965: Input Parameter:
5966: . mat - the matrix
5968: Output Parameter:
5969: . assembled - `PETSC_TRUE` or `PETSC_FALSE`
5971: Level: advanced
5973: .seealso: [](ch_matrices), `Mat`, `MatAssemblyEnd()`, `MatSetValues()`, `MatAssemblyBegin()`
5974: @*/
5975: PetscErrorCode MatAssembled(Mat mat, PetscBool *assembled)
5976: {
5977: PetscFunctionBegin;
5979: PetscAssertPointer(assembled, 2);
5980: *assembled = mat->assembled;
5981: PetscFunctionReturn(PETSC_SUCCESS);
5982: }
5984: /*@
5985: MatAssemblyEnd - Completes assembling the matrix. This routine should
5986: be called after `MatAssemblyBegin()`.
5988: Collective
5990: Input Parameters:
5991: + mat - the matrix
5992: - type - type of assembly, either `MAT_FLUSH_ASSEMBLY` or `MAT_FINAL_ASSEMBLY`
5994: Options Database Key:
5995: . -mat_view [viewertype][:...] - option name and values. See `MatViewFromOptions()`/`PetscObjectViewFromOptions()` for the possible arguments
5997: Level: beginner
5999: .seealso: [](ch_matrices), `Mat`, `MatAssemblyBegin()`, `MatSetValues()`, `PetscDrawOpenX()`, `PetscDrawCreate()`, `MatView()`, `MatAssembled()`, `PetscViewerSocketOpen()`,
6000: `MatViewFromOptions()`, `PetscObjectViewFromOptions()`
6001: @*/
6002: PetscErrorCode MatAssemblyEnd(Mat mat, MatAssemblyType type)
6003: {
6004: static PetscInt inassm = 0;
6005: PetscBool flg = PETSC_FALSE;
6007: PetscFunctionBegin;
6011: inassm++;
6012: MatAssemblyEnd_InUse++;
6013: if (MatAssemblyEnd_InUse == 1) { /* Do the logging only the first time through */
6014: PetscCall(PetscLogEventBegin(MAT_AssemblyEnd, mat, 0, 0, 0));
6015: PetscTryTypeMethod(mat, assemblyend, type);
6016: PetscCall(PetscLogEventEnd(MAT_AssemblyEnd, mat, 0, 0, 0));
6017: } else PetscTryTypeMethod(mat, assemblyend, type);
6019: /* Flush assembly is not a true assembly */
6020: if (type != MAT_FLUSH_ASSEMBLY) {
6021: if (mat->num_ass) {
6022: if (!mat->symmetry_eternal) {
6023: mat->symmetric = PETSC_BOOL3_UNKNOWN;
6024: mat->hermitian = PETSC_BOOL3_UNKNOWN;
6025: }
6026: if (!mat->structural_symmetry_eternal && mat->ass_nonzerostate != mat->nonzerostate) mat->structurally_symmetric = PETSC_BOOL3_UNKNOWN;
6027: if (!mat->spd_eternal) mat->spd = PETSC_BOOL3_UNKNOWN;
6028: }
6029: mat->num_ass++;
6030: mat->assembled = PETSC_TRUE;
6031: mat->ass_nonzerostate = mat->nonzerostate;
6032: }
6034: mat->insertmode = NOT_SET_VALUES;
6035: MatAssemblyEnd_InUse--;
6036: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6037: if (inassm == 1 && type != MAT_FLUSH_ASSEMBLY) {
6038: PetscCall(MatViewFromOptions(mat, NULL, "-mat_view"));
6040: if (mat->checksymmetryonassembly) {
6041: PetscCall(MatIsSymmetric(mat, mat->checksymmetrytol, &flg));
6042: if (flg) {
6043: PetscCall(PetscPrintf(PetscObjectComm((PetscObject)mat), "Matrix is symmetric (tolerance %g)\n", (double)mat->checksymmetrytol));
6044: } else {
6045: PetscCall(PetscPrintf(PetscObjectComm((PetscObject)mat), "Matrix is not symmetric (tolerance %g)\n", (double)mat->checksymmetrytol));
6046: }
6047: }
6048: if (mat->nullsp && mat->checknullspaceonassembly) PetscCall(MatNullSpaceTest(mat->nullsp, mat, NULL));
6049: }
6050: inassm--;
6051: PetscFunctionReturn(PETSC_SUCCESS);
6052: }
6054: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
6055: /*@
6056: MatSetOption - Sets a parameter option for a matrix. Some options
6057: may be specific to certain storage formats. Some options
6058: determine how values will be inserted (or added). Sorted,
6059: row-oriented input will generally assemble the fastest. The default
6060: is row-oriented.
6062: Logically Collective for certain operations, such as `MAT_SPD`, not collective for `MAT_ROW_ORIENTED`, see `MatOption`
6064: Input Parameters:
6065: + mat - the matrix
6066: . op - the option, one of those listed below (and possibly others),
6067: - flg - turn the option on (`PETSC_TRUE`) or off (`PETSC_FALSE`)
6069: Options Describing Matrix Structure:
6070: + `MAT_SPD` - symmetric positive definite
6071: . `MAT_SYMMETRIC` - symmetric in terms of both structure and value
6072: . `MAT_HERMITIAN` - transpose is the complex conjugation
6073: . `MAT_STRUCTURALLY_SYMMETRIC` - symmetric nonzero structure
6074: . `MAT_SYMMETRY_ETERNAL` - indicates the symmetry (or Hermitian structure) or its absence will persist through any changes to the matrix
6075: . `MAT_STRUCTURAL_SYMMETRY_ETERNAL` - indicates the structural symmetry or its absence will persist through any changes to the matrix
6076: . `MAT_SPD_ETERNAL` - indicates the value of `MAT_SPD` (true or false) will persist through any changes to the matrix
6078: These are not really options of the matrix, they are knowledge about the structure of the matrix that users may provide so that they
6079: do not need to be computed (usually at a high cost)
6081: Options For Use with `MatSetValues()`:
6082: Insert a logically dense subblock, which can be
6083: . `MAT_ROW_ORIENTED` - row-oriented (default)
6085: These options reflect the data you pass in with `MatSetValues()`; it has
6086: nothing to do with how the data is stored internally in the matrix
6087: data structure.
6089: When (re)assembling a matrix, we can restrict the input for
6090: efficiency/debugging purposes. These options include
6091: . `MAT_NEW_NONZERO_LOCATIONS` - additional insertions will be allowed if they generate a new nonzero (slow)
6092: . `MAT_FORCE_DIAGONAL_ENTRIES` - forces diagonal entries to be allocated
6093: . `MAT_IGNORE_OFF_PROC_ENTRIES` - drops off-processor entries
6094: . `MAT_NEW_NONZERO_LOCATION_ERR` - generates an error for new matrix entry
6095: . `MAT_USE_HASH_TABLE` - uses a hash table to speed up matrix assembly
6096: . `MAT_NO_OFF_PROC_ENTRIES` - you know each process will only set values for its own rows, will generate an error if
6097: any process sets values for another process. This avoids all reductions in the MatAssembly routines and thus improves
6098: performance for very large process counts.
6099: - `MAT_SUBSET_OFF_PROC_ENTRIES` - you know that the first assembly after setting this flag will set a superset
6100: of the off-process entries required for all subsequent assemblies. This avoids a rendezvous step in the MatAssembly
6101: functions, instead sending only neighbor messages.
6103: Level: intermediate
6105: Notes:
6106: Except for `MAT_UNUSED_NONZERO_LOCATION_ERR` and `MAT_ROW_ORIENTED` all processes that share the matrix must pass the same value in flg!
6108: Some options are relevant only for particular matrix types and
6109: are thus ignored by others. Other options are not supported by
6110: certain matrix types and will generate an error message if set.
6112: If using Fortran to compute a matrix, one may need to
6113: use the column-oriented option (or convert to the row-oriented
6114: format).
6116: `MAT_NEW_NONZERO_LOCATIONS` set to `PETSC_FALSE` indicates that any add or insertion
6117: that would generate a new entry in the nonzero structure is instead
6118: ignored. Thus, if memory has not already been allocated for this particular
6119: data, then the insertion is ignored. For dense matrices, in which
6120: the entire array is allocated, no entries are ever ignored.
6121: Set after the first `MatAssemblyEnd()`. If this option is set, then the `MatAssemblyBegin()`/`MatAssemblyEnd()` processes has one less global reduction
6123: `MAT_NEW_NONZERO_LOCATION_ERR` set to PETSC_TRUE indicates that any add or insertion
6124: that would generate a new entry in the nonzero structure instead produces
6125: an error. (Currently supported for `MATAIJ` and `MATBAIJ` formats only.) If this option is set, then the `MatAssemblyBegin()`/`MatAssemblyEnd()` processes has one less global reduction
6127: `MAT_NEW_NONZERO_ALLOCATION_ERR` set to `PETSC_TRUE` indicates that any add or insertion
6128: that would generate a new entry that has not been preallocated will
6129: instead produce an error. (Currently supported for `MATAIJ` and `MATBAIJ` formats
6130: only.) This is a useful flag when debugging matrix memory preallocation.
6131: If this option is set, then the `MatAssemblyBegin()`/`MatAssemblyEnd()` processes has one less global reduction
6133: `MAT_IGNORE_OFF_PROC_ENTRIES` set to `PETSC_TRUE` indicates entries destined for
6134: other processors should be dropped, rather than stashed.
6135: This is useful if you know that the "owning" processor is also
6136: always generating the correct matrix entries, so that PETSc need
6137: not transfer duplicate entries generated on another processor.
6139: `MAT_USE_HASH_TABLE` indicates that a hash table be used to improve the
6140: searches during matrix assembly. When this flag is set, the hash table
6141: is created during the first matrix assembly. This hash table is
6142: used the next time through, during `MatSetValues()`/`MatSetValuesBlocked()`
6143: to improve the searching of indices. `MAT_NEW_NONZERO_LOCATIONS` flag
6144: should be used with `MAT_USE_HASH_TABLE` flag. This option is currently
6145: supported by `MATMPIBAIJ` format only.
6147: `MAT_KEEP_NONZERO_PATTERN` indicates when `MatZeroRows()` is called the zeroed entries
6148: are kept in the nonzero structure. This flag is not used for `MatZeroRowsColumns()`
6150: `MAT_IGNORE_ZERO_ENTRIES` - for `MATAIJ` and `MATIS` matrices this will stop zero values from creating
6151: a zero location in the matrix
6153: `MAT_USE_INODES` - indicates using inode version of the code - works with `MATAIJ` matrix types
6155: `MAT_NO_OFF_PROC_ZERO_ROWS` - you know each process will only zero its own rows. This avoids all reductions in the
6156: zero row routines and thus improves performance for very large process counts.
6158: `MAT_IGNORE_LOWER_TRIANGULAR` - For `MATSBAIJ` matrices will ignore any insertions you make in the lower triangular
6159: part of the matrix (since they should match the upper triangular part).
6161: `MAT_SORTED_FULL` - each process provides exactly its local rows; all column indices for a given row are passed in a
6162: single call to `MatSetValues()`, preallocation is perfect, row-oriented, `INSERT_VALUES` is used. Common
6163: with finite difference schemes with non-periodic boundary conditions.
6165: Developer Note:
6166: `MAT_SYMMETRY_ETERNAL`, `MAT_STRUCTURAL_SYMMETRY_ETERNAL`, and `MAT_SPD_ETERNAL` are used by `MatAssemblyEnd()` and in other
6167: places where otherwise the value of `MAT_SYMMETRIC`, `MAT_STRUCTURALLY_SYMMETRIC` or `MAT_SPD` would need to be changed back
6168: to `PETSC_BOOL3_UNKNOWN` because the matrix values had changed so the code cannot be certain that the related property had
6169: not changed.
6171: .seealso: [](ch_matrices), `MatOption`, `Mat`, `MatGetOption()`
6172: @*/
6173: PetscErrorCode MatSetOption(Mat mat, MatOption op, PetscBool flg)
6174: {
6175: PetscFunctionBegin;
6177: if (op > 0) {
6180: }
6182: PetscCheck(((int)op) > MAT_OPTION_MIN && ((int)op) < MAT_OPTION_MAX, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Options %d is out of range", (int)op);
6184: switch (op) {
6185: case MAT_FORCE_DIAGONAL_ENTRIES:
6186: mat->force_diagonals = flg;
6187: PetscFunctionReturn(PETSC_SUCCESS);
6188: case MAT_NO_OFF_PROC_ENTRIES:
6189: mat->nooffprocentries = flg;
6190: PetscFunctionReturn(PETSC_SUCCESS);
6191: case MAT_SUBSET_OFF_PROC_ENTRIES:
6192: mat->assembly_subset = flg;
6193: if (!mat->assembly_subset) { /* See the same logic in VecAssembly wrt VEC_SUBSET_OFF_PROC_ENTRIES */
6194: #if !defined(PETSC_HAVE_MPIUNI)
6195: PetscCall(MatStashScatterDestroy_BTS(&mat->stash));
6196: #endif
6197: mat->stash.first_assembly_done = PETSC_FALSE;
6198: }
6199: PetscFunctionReturn(PETSC_SUCCESS);
6200: case MAT_NO_OFF_PROC_ZERO_ROWS:
6201: mat->nooffproczerorows = flg;
6202: PetscFunctionReturn(PETSC_SUCCESS);
6203: case MAT_SPD:
6204: if (flg) {
6205: mat->spd = PETSC_BOOL3_TRUE;
6206: mat->symmetric = PETSC_BOOL3_TRUE;
6207: mat->structurally_symmetric = PETSC_BOOL3_TRUE;
6208: #if !defined(PETSC_USE_COMPLEX)
6209: mat->hermitian = PETSC_BOOL3_TRUE;
6210: #endif
6211: } else {
6212: mat->spd = PETSC_BOOL3_FALSE;
6213: }
6214: break;
6215: case MAT_SYMMETRIC:
6216: mat->symmetric = PetscBoolToBool3(flg);
6217: if (flg) mat->structurally_symmetric = PETSC_BOOL3_TRUE;
6218: #if !defined(PETSC_USE_COMPLEX)
6219: mat->hermitian = PetscBoolToBool3(flg);
6220: #endif
6221: break;
6222: case MAT_HERMITIAN:
6223: mat->hermitian = PetscBoolToBool3(flg);
6224: if (flg) mat->structurally_symmetric = PETSC_BOOL3_TRUE;
6225: #if !defined(PETSC_USE_COMPLEX)
6226: mat->symmetric = PetscBoolToBool3(flg);
6227: #endif
6228: break;
6229: case MAT_STRUCTURALLY_SYMMETRIC:
6230: mat->structurally_symmetric = PetscBoolToBool3(flg);
6231: break;
6232: case MAT_SYMMETRY_ETERNAL:
6233: PetscCheck(mat->symmetric != PETSC_BOOL3_UNKNOWN, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Cannot set MAT_SYMMETRY_ETERNAL without first setting MAT_SYMMETRIC to true or false");
6234: mat->symmetry_eternal = flg;
6235: if (flg) mat->structural_symmetry_eternal = PETSC_TRUE;
6236: break;
6237: case MAT_STRUCTURAL_SYMMETRY_ETERNAL:
6238: PetscCheck(mat->structurally_symmetric != PETSC_BOOL3_UNKNOWN, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Cannot set MAT_STRUCTURAL_SYMMETRY_ETERNAL without first setting MAT_STRUCTURALLY_SYMMETRIC to true or false");
6239: mat->structural_symmetry_eternal = flg;
6240: break;
6241: case MAT_SPD_ETERNAL:
6242: PetscCheck(mat->spd != PETSC_BOOL3_UNKNOWN, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Cannot set MAT_SPD_ETERNAL without first setting MAT_SPD to true or false");
6243: mat->spd_eternal = flg;
6244: if (flg) {
6245: mat->structural_symmetry_eternal = PETSC_TRUE;
6246: mat->symmetry_eternal = PETSC_TRUE;
6247: }
6248: break;
6249: case MAT_STRUCTURE_ONLY:
6250: mat->structure_only = flg;
6251: break;
6252: case MAT_SORTED_FULL:
6253: mat->sortedfull = flg;
6254: break;
6255: default:
6256: break;
6257: }
6258: PetscTryTypeMethod(mat, setoption, op, flg);
6259: PetscFunctionReturn(PETSC_SUCCESS);
6260: }
6262: /*@
6263: MatGetOption - Gets a parameter option that has been set for a matrix.
6265: Logically Collective
6267: Input Parameters:
6268: + mat - the matrix
6269: - op - the option, this only responds to certain options, check the code for which ones
6271: Output Parameter:
6272: . flg - turn the option on (`PETSC_TRUE`) or off (`PETSC_FALSE`)
6274: Level: intermediate
6276: Notes:
6277: Can only be called after `MatSetSizes()` and `MatSetType()` have been set.
6279: Certain option values may be unknown, for those use the routines `MatIsSymmetric()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, or
6280: `MatIsSymmetricKnown()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetricKnown()`
6282: .seealso: [](ch_matrices), `Mat`, `MatOption`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`,
6283: `MatIsSymmetricKnown()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetricKnown()`
6284: @*/
6285: PetscErrorCode MatGetOption(Mat mat, MatOption op, PetscBool *flg)
6286: {
6287: PetscFunctionBegin;
6291: PetscCheck(((int)op) > MAT_OPTION_MIN && ((int)op) < MAT_OPTION_MAX, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Options %d is out of range", (int)op);
6292: PetscCheck(((PetscObject)mat)->type_name, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_TYPENOTSET, "Cannot get options until type and size have been set, see MatSetType() and MatSetSizes()");
6294: switch (op) {
6295: case MAT_NO_OFF_PROC_ENTRIES:
6296: *flg = mat->nooffprocentries;
6297: break;
6298: case MAT_NO_OFF_PROC_ZERO_ROWS:
6299: *flg = mat->nooffproczerorows;
6300: break;
6301: case MAT_SYMMETRIC:
6302: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsSymmetric() or MatIsSymmetricKnown()");
6303: break;
6304: case MAT_HERMITIAN:
6305: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsHermitian() or MatIsHermitianKnown()");
6306: break;
6307: case MAT_STRUCTURALLY_SYMMETRIC:
6308: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsStructurallySymmetric() or MatIsStructurallySymmetricKnown()");
6309: break;
6310: case MAT_SPD:
6311: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsSPDKnown()");
6312: break;
6313: case MAT_SYMMETRY_ETERNAL:
6314: *flg = mat->symmetry_eternal;
6315: break;
6316: case MAT_STRUCTURAL_SYMMETRY_ETERNAL:
6317: *flg = mat->symmetry_eternal;
6318: break;
6319: default:
6320: break;
6321: }
6322: PetscFunctionReturn(PETSC_SUCCESS);
6323: }
6325: /*@
6326: MatZeroEntries - Zeros all entries of a matrix. For sparse matrices
6327: this routine retains the old nonzero structure.
6329: Logically Collective
6331: Input Parameter:
6332: . mat - the matrix
6334: Level: intermediate
6336: Note:
6337: If the matrix was not preallocated then a default, likely poor preallocation will be set in the matrix, so this should be called after the preallocation phase.
6338: See the Performance chapter of the users manual for information on preallocating matrices.
6340: .seealso: [](ch_matrices), `Mat`, `MatZeroRows()`, `MatZeroRowsColumns()`
6341: @*/
6342: PetscErrorCode MatZeroEntries(Mat mat)
6343: {
6344: PetscFunctionBegin;
6347: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6348: PetscCheck(mat->insertmode == NOT_SET_VALUES, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for matrices where you have set values but not yet assembled");
6349: MatCheckPreallocated(mat, 1);
6351: PetscCall(PetscLogEventBegin(MAT_ZeroEntries, mat, 0, 0, 0));
6352: PetscUseTypeMethod(mat, zeroentries);
6353: PetscCall(PetscLogEventEnd(MAT_ZeroEntries, mat, 0, 0, 0));
6354: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6355: PetscFunctionReturn(PETSC_SUCCESS);
6356: }
6358: /*@
6359: MatZeroRowsColumns - Zeros all entries (except possibly the main diagonal)
6360: of a set of rows and columns of a matrix.
6362: Collective
6364: Input Parameters:
6365: + mat - the matrix
6366: . numRows - the number of rows/columns to zero
6367: . rows - the global row indices
6368: . diag - value put in the diagonal of the eliminated rows
6369: . x - optional vector of the solution for zeroed rows (other entries in vector are not used), these must be set before this call
6370: - b - optional vector of the right-hand side, that will be adjusted by provided solution entries
6372: Level: intermediate
6374: Notes:
6375: This routine, along with `MatZeroRows()`, is typically used to eliminate known Dirichlet boundary conditions from a linear system.
6377: For each zeroed row, the value of the corresponding `b` is set to diag times the value of the corresponding `x`.
6378: The other entries of `b` will be adjusted by the known values of `x` times the corresponding matrix entries in the columns that are being eliminated
6380: If the resulting linear system is to be solved with `KSP` then one can (but does not have to) call `KSPSetInitialGuessNonzero()` to allow the
6381: Krylov method to take advantage of the known solution on the zeroed rows.
6383: For the parallel case, all processes that share the matrix (i.e.,
6384: those in the communicator used for matrix creation) MUST call this
6385: routine, regardless of whether any rows being zeroed are owned by
6386: them.
6388: Unlike `MatZeroRows()`, this ignores the `MAT_KEEP_NONZERO_PATTERN` option value set with `MatSetOption()`, it merely zeros those entries in the matrix, but never
6389: removes them from the nonzero pattern. The nonzero pattern of the matrix can still change if a nonzero needs to be inserted on a diagonal entry that was previously
6390: missing.
6392: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
6393: list only rows local to itself).
6395: The option `MAT_NO_OFF_PROC_ZERO_ROWS` does not apply to this routine.
6397: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRows()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6398: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6399: @*/
6400: PetscErrorCode MatZeroRowsColumns(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6401: {
6402: PetscFunctionBegin;
6405: if (numRows) PetscAssertPointer(rows, 3);
6406: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6407: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6408: MatCheckPreallocated(mat, 1);
6410: PetscUseTypeMethod(mat, zerorowscolumns, numRows, rows, diag, x, b);
6411: PetscCall(MatViewFromOptions(mat, NULL, "-mat_view"));
6412: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6413: PetscFunctionReturn(PETSC_SUCCESS);
6414: }
6416: /*@
6417: MatZeroRowsColumnsIS - Zeros all entries (except possibly the main diagonal)
6418: of a set of rows and columns of a matrix.
6420: Collective
6422: Input Parameters:
6423: + mat - the matrix
6424: . is - the rows to zero
6425: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6426: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6427: - b - optional vector of right-hand side, that will be adjusted by provided solution
6429: Level: intermediate
6431: Note:
6432: See `MatZeroRowsColumns()` for details on how this routine operates.
6434: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6435: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRows()`, `MatZeroRowsColumnsStencil()`
6436: @*/
6437: PetscErrorCode MatZeroRowsColumnsIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6438: {
6439: PetscInt numRows;
6440: const PetscInt *rows;
6442: PetscFunctionBegin;
6447: PetscCall(ISGetLocalSize(is, &numRows));
6448: PetscCall(ISGetIndices(is, &rows));
6449: PetscCall(MatZeroRowsColumns(mat, numRows, rows, diag, x, b));
6450: PetscCall(ISRestoreIndices(is, &rows));
6451: PetscFunctionReturn(PETSC_SUCCESS);
6452: }
6454: /*@
6455: MatZeroRows - Zeros all entries (except possibly the main diagonal)
6456: of a set of rows of a matrix.
6458: Collective
6460: Input Parameters:
6461: + mat - the matrix
6462: . numRows - the number of rows to zero
6463: . rows - the global row indices
6464: . diag - value put in the diagonal of the zeroed rows
6465: . x - optional vector of solutions for zeroed rows (other entries in vector are not used), these must be set before this call
6466: - b - optional vector of right-hand side, that will be adjusted by provided solution entries
6468: Level: intermediate
6470: Notes:
6471: This routine, along with `MatZeroRowsColumns()`, is typically used to eliminate known Dirichlet boundary conditions from a linear system.
6473: For each zeroed row, the value of the corresponding `b` is set to `diag` times the value of the corresponding `x`.
6475: If the resulting linear system is to be solved with `KSP` then one can (but does not have to) call `KSPSetInitialGuessNonzero()` to allow the
6476: Krylov method to take advantage of the known solution on the zeroed rows.
6478: May be followed by using a `PC` of type `PCREDISTRIBUTE` to solve the reduced problem (`PCDISTRIBUTE` completely eliminates the zeroed rows and their corresponding columns)
6479: from the matrix.
6481: Unlike `MatZeroRowsColumns()` for the `MATAIJ` and `MATBAIJ` matrix formats this removes the old nonzero structure, from the eliminated rows of the matrix
6482: but does not release memory. Because of this removal matrix-vector products with the adjusted matrix will be a bit faster. For the dense
6483: formats this does not alter the nonzero structure.
6485: If the option `MatSetOption`(mat,`MAT_KEEP_NONZERO_PATTERN`,`PETSC_TRUE`) the nonzero structure
6486: of the matrix is not changed the values are
6487: merely zeroed.
6489: The user can set a value in the diagonal entry (or for the `MATAIJ` format
6490: formats can optionally remove the main diagonal entry from the
6491: nonzero structure as well, by passing 0.0 as the final argument).
6493: For the parallel case, all processes that share the matrix (i.e.,
6494: those in the communicator used for matrix creation) MUST call this
6495: routine, regardless of whether any rows being zeroed are owned by
6496: them.
6498: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
6499: list only rows local to itself).
6501: You can call `MatSetOption`(mat,`MAT_NO_OFF_PROC_ZERO_ROWS`,`PETSC_TRUE`) if each process indicates only rows it
6502: owns that are to be zeroed. This saves a global synchronization in the implementation.
6504: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6505: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`, `PCREDISTRIBUTE`, `MAT_KEEP_NONZERO_PATTERN`
6506: @*/
6507: PetscErrorCode MatZeroRows(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6508: {
6509: PetscFunctionBegin;
6512: if (numRows) PetscAssertPointer(rows, 3);
6513: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6514: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6515: MatCheckPreallocated(mat, 1);
6517: PetscUseTypeMethod(mat, zerorows, numRows, rows, diag, x, b);
6518: PetscCall(MatViewFromOptions(mat, NULL, "-mat_view"));
6519: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6520: PetscFunctionReturn(PETSC_SUCCESS);
6521: }
6523: /*@
6524: MatZeroRowsIS - Zeros all entries (except possibly the main diagonal)
6525: of a set of rows of a matrix indicated by an `IS`
6527: Collective
6529: Input Parameters:
6530: + mat - the matrix
6531: . is - index set, `IS`, of rows to remove (if `NULL` then no row is removed)
6532: . diag - value put in all diagonals of eliminated rows
6533: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6534: - b - optional vector of right-hand side, that will be adjusted by provided solution
6536: Level: intermediate
6538: Note:
6539: See `MatZeroRows()` for details on how this routine operates.
6541: .seealso: [](ch_matrices), `Mat`, `MatZeroRows()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6542: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`, `IS`
6543: @*/
6544: PetscErrorCode MatZeroRowsIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6545: {
6546: PetscInt numRows = 0;
6547: const PetscInt *rows = NULL;
6549: PetscFunctionBegin;
6552: if (is) {
6554: PetscCall(ISGetLocalSize(is, &numRows));
6555: PetscCall(ISGetIndices(is, &rows));
6556: }
6557: PetscCall(MatZeroRows(mat, numRows, rows, diag, x, b));
6558: if (is) PetscCall(ISRestoreIndices(is, &rows));
6559: PetscFunctionReturn(PETSC_SUCCESS);
6560: }
6562: /*@
6563: MatZeroRowsStencil - Zeros all entries (except possibly the main diagonal)
6564: of a set of rows of a matrix indicated by a `MatStencil`. These rows must be local to the process.
6566: Collective
6568: Input Parameters:
6569: + mat - the matrix
6570: . numRows - the number of rows to remove
6571: . rows - the grid coordinates (and component number when dof > 1) for matrix rows indicated by an array of `MatStencil`
6572: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6573: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6574: - b - optional vector of right-hand side, that will be adjusted by provided solution
6576: Level: intermediate
6578: Notes:
6579: See `MatZeroRows()` for details on how this routine operates.
6581: The grid coordinates are across the entire grid, not just the local portion
6583: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
6584: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
6585: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
6586: `DM_BOUNDARY_PERIODIC` boundary type.
6588: For indices that don't mean anything for your case (like the `k` index when working in 2d) or the `c` index when you have
6589: a single value per point) you can skip filling those indices.
6591: Fortran Note:
6592: `idxm` and `idxn` should be declared as
6593: .vb
6594: MatStencil idxm(4, m)
6595: .ve
6596: and the values inserted using
6597: .vb
6598: idxm(MatStencil_i, 1) = i
6599: idxm(MatStencil_j, 1) = j
6600: idxm(MatStencil_k, 1) = k
6601: idxm(MatStencil_c, 1) = c
6602: etc
6603: .ve
6605: .seealso: [](ch_matrices), `Mat`, `MatStencil`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRows()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6606: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6607: @*/
6608: PetscErrorCode MatZeroRowsStencil(Mat mat, PetscInt numRows, const MatStencil rows[], PetscScalar diag, Vec x, Vec b)
6609: {
6610: PetscInt dim = mat->stencil.dim;
6611: PetscInt sdim = dim - (1 - (PetscInt)mat->stencil.noc);
6612: PetscInt *dims = mat->stencil.dims + 1;
6613: PetscInt *starts = mat->stencil.starts;
6614: PetscInt *dxm = (PetscInt *)rows;
6615: PetscInt *jdxm, i, j, tmp, numNewRows = 0;
6617: PetscFunctionBegin;
6620: if (numRows) PetscAssertPointer(rows, 3);
6622: PetscCall(PetscMalloc1(numRows, &jdxm));
6623: for (i = 0; i < numRows; ++i) {
6624: /* Skip unused dimensions (they are ordered k, j, i, c) */
6625: for (j = 0; j < 3 - sdim; ++j) dxm++;
6626: /* Local index in X dir */
6627: tmp = *dxm++ - starts[0];
6628: /* Loop over remaining dimensions */
6629: for (j = 0; j < dim - 1; ++j) {
6630: /* If nonlocal, set index to be negative */
6631: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = PETSC_INT_MIN;
6632: /* Update local index */
6633: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
6634: }
6635: /* Skip component slot if necessary */
6636: if (mat->stencil.noc) dxm++;
6637: /* Local row number */
6638: if (tmp >= 0) jdxm[numNewRows++] = tmp;
6639: }
6640: PetscCall(MatZeroRowsLocal(mat, numNewRows, jdxm, diag, x, b));
6641: PetscCall(PetscFree(jdxm));
6642: PetscFunctionReturn(PETSC_SUCCESS);
6643: }
6645: /*@
6646: MatZeroRowsColumnsStencil - Zeros all row and column entries (except possibly the main diagonal)
6647: of a set of rows and columns of a matrix.
6649: Collective
6651: Input Parameters:
6652: + mat - the matrix
6653: . numRows - the number of rows/columns to remove
6654: . rows - the grid coordinates (and component number when dof > 1) for matrix rows
6655: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6656: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6657: - b - optional vector of right-hand side, that will be adjusted by provided solution
6659: Level: intermediate
6661: Notes:
6662: See `MatZeroRowsColumns()` for details on how this routine operates.
6664: The grid coordinates are across the entire grid, not just the local portion
6666: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
6667: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
6668: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
6669: `DM_BOUNDARY_PERIODIC` boundary type.
6671: For indices that don't mean anything for your case (like the `k` index when working in 2d) or the `c` index when you have
6672: a single value per point) you can skip filling those indices.
6674: Fortran Note:
6675: `idxm` and `idxn` should be declared as
6676: .vb
6677: MatStencil idxm(4, m)
6678: .ve
6679: and the values inserted using
6680: .vb
6681: idxm(MatStencil_i, 1) = i
6682: idxm(MatStencil_j, 1) = j
6683: idxm(MatStencil_k, 1) = k
6684: idxm(MatStencil_c, 1) = c
6685: etc
6686: .ve
6688: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6689: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRows()`
6690: @*/
6691: PetscErrorCode MatZeroRowsColumnsStencil(Mat mat, PetscInt numRows, const MatStencil rows[], PetscScalar diag, Vec x, Vec b)
6692: {
6693: PetscInt dim = mat->stencil.dim;
6694: PetscInt sdim = dim - (1 - (PetscInt)mat->stencil.noc);
6695: PetscInt *dims = mat->stencil.dims + 1;
6696: PetscInt *starts = mat->stencil.starts;
6697: PetscInt *dxm = (PetscInt *)rows;
6698: PetscInt *jdxm, i, j, tmp, numNewRows = 0;
6700: PetscFunctionBegin;
6703: if (numRows) PetscAssertPointer(rows, 3);
6705: PetscCall(PetscMalloc1(numRows, &jdxm));
6706: for (i = 0; i < numRows; ++i) {
6707: /* Skip unused dimensions (they are ordered k, j, i, c) */
6708: for (j = 0; j < 3 - sdim; ++j) dxm++;
6709: /* Local index in X dir */
6710: tmp = *dxm++ - starts[0];
6711: /* Loop over remaining dimensions */
6712: for (j = 0; j < dim - 1; ++j) {
6713: /* If nonlocal, set index to be negative */
6714: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = PETSC_INT_MIN;
6715: /* Update local index */
6716: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
6717: }
6718: /* Skip component slot if necessary */
6719: if (mat->stencil.noc) dxm++;
6720: /* Local row number */
6721: if (tmp >= 0) jdxm[numNewRows++] = tmp;
6722: }
6723: PetscCall(MatZeroRowsColumnsLocal(mat, numNewRows, jdxm, diag, x, b));
6724: PetscCall(PetscFree(jdxm));
6725: PetscFunctionReturn(PETSC_SUCCESS);
6726: }
6728: /*@
6729: MatZeroRowsLocal - Zeros all entries (except possibly the main diagonal)
6730: of a set of rows of a matrix; using local numbering of rows.
6732: Collective
6734: Input Parameters:
6735: + mat - the matrix
6736: . numRows - the number of rows to remove
6737: . rows - the local row indices
6738: . diag - value put in all diagonals of eliminated rows
6739: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6740: - b - optional vector of right-hand side, that will be adjusted by provided solution
6742: Level: intermediate
6744: Notes:
6745: Before calling `MatZeroRowsLocal()`, the user must first set the
6746: local-to-global mapping by calling MatSetLocalToGlobalMapping(), this is often already set for matrices obtained with `DMCreateMatrix()`.
6748: See `MatZeroRows()` for details on how this routine operates.
6750: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRows()`, `MatSetOption()`,
6751: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6752: @*/
6753: PetscErrorCode MatZeroRowsLocal(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6754: {
6755: PetscFunctionBegin;
6758: if (numRows) PetscAssertPointer(rows, 3);
6759: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6760: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6761: MatCheckPreallocated(mat, 1);
6763: if (mat->ops->zerorowslocal) {
6764: PetscUseTypeMethod(mat, zerorowslocal, numRows, rows, diag, x, b);
6765: } else {
6766: IS is, newis;
6767: PetscInt *newRows, nl = 0;
6769: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Need to provide local to global mapping to matrix first");
6770: PetscCall(ISCreateGeneral(PETSC_COMM_SELF, numRows, rows, PETSC_USE_POINTER, &is));
6771: PetscCall(ISLocalToGlobalMappingApplyIS(mat->rmap->mapping, is, &newis));
6772: PetscCall(ISGetIndices(newis, (const PetscInt **)&newRows));
6773: for (PetscInt i = 0; i < numRows; i++)
6774: if (newRows[i] > -1) newRows[nl++] = newRows[i];
6775: PetscUseTypeMethod(mat, zerorows, nl, newRows, diag, x, b);
6776: PetscCall(ISRestoreIndices(newis, (const PetscInt **)&newRows));
6777: PetscCall(ISDestroy(&newis));
6778: PetscCall(ISDestroy(&is));
6779: }
6780: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6781: PetscFunctionReturn(PETSC_SUCCESS);
6782: }
6784: /*@
6785: MatZeroRowsLocalIS - Zeros all entries (except possibly the main diagonal)
6786: of a set of rows of a matrix; using local numbering of rows.
6788: Collective
6790: Input Parameters:
6791: + mat - the matrix
6792: . is - index set of rows to remove
6793: . diag - value put in all diagonals of eliminated rows
6794: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6795: - b - optional vector of right-hand side, that will be adjusted by provided solution
6797: Level: intermediate
6799: Notes:
6800: Before calling `MatZeroRowsLocalIS()`, the user must first set the
6801: local-to-global mapping by calling `MatSetLocalToGlobalMapping()`, this is often already set for matrices obtained with `DMCreateMatrix()`.
6803: See `MatZeroRows()` for details on how this routine operates.
6805: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRows()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6806: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6807: @*/
6808: PetscErrorCode MatZeroRowsLocalIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6809: {
6810: PetscInt numRows;
6811: const PetscInt *rows;
6813: PetscFunctionBegin;
6817: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6818: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6819: MatCheckPreallocated(mat, 1);
6821: PetscCall(ISGetLocalSize(is, &numRows));
6822: PetscCall(ISGetIndices(is, &rows));
6823: PetscCall(MatZeroRowsLocal(mat, numRows, rows, diag, x, b));
6824: PetscCall(ISRestoreIndices(is, &rows));
6825: PetscFunctionReturn(PETSC_SUCCESS);
6826: }
6828: /*@
6829: MatZeroRowsColumnsLocal - Zeros all entries (except possibly the main diagonal)
6830: of a set of rows and columns of a matrix; using local numbering of rows.
6832: Collective
6834: Input Parameters:
6835: + mat - the matrix
6836: . numRows - the number of rows to remove
6837: . rows - the global row indices
6838: . diag - value put in all diagonals of eliminated rows
6839: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6840: - b - optional vector of right-hand side, that will be adjusted by provided solution
6842: Level: intermediate
6844: Notes:
6845: Before calling `MatZeroRowsColumnsLocal()`, the user must first set the
6846: local-to-global mapping by calling `MatSetLocalToGlobalMapping()`, this is often already set for matrices obtained with `DMCreateMatrix()`.
6848: See `MatZeroRowsColumns()` for details on how this routine operates.
6850: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6851: `MatZeroRows()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6852: @*/
6853: PetscErrorCode MatZeroRowsColumnsLocal(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6854: {
6855: PetscFunctionBegin;
6858: if (numRows) PetscAssertPointer(rows, 3);
6859: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6860: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6861: MatCheckPreallocated(mat, 1);
6863: if (mat->ops->zerorowscolumnslocal) {
6864: PetscUseTypeMethod(mat, zerorowscolumnslocal, numRows, rows, diag, x, b);
6865: } else {
6866: IS is, newis;
6867: PetscInt *newRows, nl = 0;
6869: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Need to provide local to global mapping to matrix first");
6870: PetscCall(ISCreateGeneral(PETSC_COMM_SELF, numRows, rows, PETSC_USE_POINTER, &is));
6871: PetscCall(ISLocalToGlobalMappingApplyIS(mat->rmap->mapping, is, &newis));
6872: PetscCall(ISGetIndices(newis, (const PetscInt **)&newRows));
6873: for (PetscInt i = 0; i < numRows; i++)
6874: if (newRows[i] > -1) newRows[nl++] = newRows[i];
6875: PetscUseTypeMethod(mat, zerorowscolumns, nl, newRows, diag, x, b);
6876: PetscCall(ISRestoreIndices(newis, (const PetscInt **)&newRows));
6877: PetscCall(ISDestroy(&newis));
6878: PetscCall(ISDestroy(&is));
6879: }
6880: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6881: PetscFunctionReturn(PETSC_SUCCESS);
6882: }
6884: /*@
6885: MatZeroRowsColumnsLocalIS - Zeros all entries (except possibly the main diagonal)
6886: of a set of rows and columns of a matrix; using local numbering of rows.
6888: Collective
6890: Input Parameters:
6891: + mat - the matrix
6892: . is - index set of rows to remove
6893: . diag - value put in all diagonals of eliminated rows
6894: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6895: - b - optional vector of right-hand side, that will be adjusted by provided solution
6897: Level: intermediate
6899: Notes:
6900: Before calling `MatZeroRowsColumnsLocalIS()`, the user must first set the
6901: local-to-global mapping by calling `MatSetLocalToGlobalMapping()`, this is often already set for matrices obtained with `DMCreateMatrix()`.
6903: See `MatZeroRowsColumns()` for details on how this routine operates.
6905: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6906: `MatZeroRowsColumnsLocal()`, `MatZeroRows()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6907: @*/
6908: PetscErrorCode MatZeroRowsColumnsLocalIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6909: {
6910: PetscInt numRows;
6911: const PetscInt *rows;
6913: PetscFunctionBegin;
6917: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6918: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6919: MatCheckPreallocated(mat, 1);
6921: PetscCall(ISGetLocalSize(is, &numRows));
6922: PetscCall(ISGetIndices(is, &rows));
6923: PetscCall(MatZeroRowsColumnsLocal(mat, numRows, rows, diag, x, b));
6924: PetscCall(ISRestoreIndices(is, &rows));
6925: PetscFunctionReturn(PETSC_SUCCESS);
6926: }
6928: /*@
6929: MatGetSize - Returns the numbers of rows and columns in a matrix.
6931: Not Collective
6933: Input Parameter:
6934: . mat - the matrix
6936: Output Parameters:
6937: + m - the number of global rows
6938: - n - the number of global columns
6940: Level: beginner
6942: Note:
6943: Both output parameters can be `NULL` on input.
6945: .seealso: [](ch_matrices), `Mat`, `MatSetSizes()`, `MatGetLocalSize()`
6946: @*/
6947: PetscErrorCode MatGetSize(Mat mat, PetscInt *m, PetscInt *n)
6948: {
6949: PetscFunctionBegin;
6951: if (m) *m = mat->rmap->N;
6952: if (n) *n = mat->cmap->N;
6953: PetscFunctionReturn(PETSC_SUCCESS);
6954: }
6956: /*@
6957: MatGetLocalSize - For most matrix formats, excluding `MATELEMENTAL` and `MATSCALAPACK`, Returns the number of local rows and local columns
6958: of a matrix. For all matrices this is the local size of the left and right vectors as returned by `MatCreateVecs()`.
6960: Not Collective
6962: Input Parameter:
6963: . mat - the matrix
6965: Output Parameters:
6966: + m - the number of local rows, use `NULL` to not obtain this value
6967: - n - the number of local columns, use `NULL` to not obtain this value
6969: Level: beginner
6971: .seealso: [](ch_matrices), `Mat`, `MatSetSizes()`, `MatGetSize()`
6972: @*/
6973: PetscErrorCode MatGetLocalSize(Mat mat, PetscInt *m, PetscInt *n)
6974: {
6975: PetscFunctionBegin;
6977: if (m) PetscAssertPointer(m, 2);
6978: if (n) PetscAssertPointer(n, 3);
6979: if (m) *m = mat->rmap->n;
6980: if (n) *n = mat->cmap->n;
6981: PetscFunctionReturn(PETSC_SUCCESS);
6982: }
6984: /*@
6985: MatGetOwnershipRangeColumn - Returns the range of matrix columns associated with rows of a
6986: vector one multiplies this matrix by that are owned by this processor.
6988: Not Collective, unless matrix has not been allocated, then collective
6990: Input Parameter:
6991: . mat - the matrix
6993: Output Parameters:
6994: + m - the global index of the first local column, use `NULL` to not obtain this value
6995: - n - one more than the global index of the last local column, use `NULL` to not obtain this value
6997: Level: developer
6999: Notes:
7000: If the `Mat` was obtained from a `DM` with `DMCreateMatrix()`, then the range values are determined by the specific `DM`.
7002: If the `Mat` was created directly the range values are determined by the local size passed to `MatSetSizes()` or `MatCreateAIJ()`.
7003: If `PETSC_DECIDE` was passed as the local size, then the vector uses default values for the range using `PetscSplitOwnership()`.
7005: For certain `DM`, such as `DMDA`, it is better to use `DM` specific routines, such as `DMDAGetGhostCorners()`, to determine
7006: the local values in the matrix.
7008: Returns the columns of the "diagonal block" for most sparse matrix formats. See [Matrix
7009: Layouts](sec_matlayout) for details on matrix layouts.
7011: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRanges()`, `MatGetOwnershipRangesColumn()`, `PetscLayout`,
7012: `MatSetSizes()`, `MatCreateAIJ()`, `DMDAGetGhostCorners()`, `DM`
7013: @*/
7014: PetscErrorCode MatGetOwnershipRangeColumn(Mat mat, PetscInt *m, PetscInt *n)
7015: {
7016: PetscFunctionBegin;
7019: if (m) PetscAssertPointer(m, 2);
7020: if (n) PetscAssertPointer(n, 3);
7021: MatCheckPreallocated(mat, 1);
7022: if (m) *m = mat->cmap->rstart;
7023: if (n) *n = mat->cmap->rend;
7024: PetscFunctionReturn(PETSC_SUCCESS);
7025: }
7027: /*@
7028: MatGetOwnershipRange - For matrices that own values by row, excludes `MATELEMENTAL` and `MATSCALAPACK`, returns the range of matrix rows owned by
7029: this MPI process.
7031: Not Collective
7033: Input Parameter:
7034: . mat - the matrix
7036: Output Parameters:
7037: + m - the global index of the first local row, use `NULL` to not obtain this value
7038: - n - one more than the global index of the last local row, use `NULL` to not obtain this value
7040: Level: beginner
7042: Notes:
7043: If the `Mat` was obtained from a `DM` with `DMCreateMatrix()`, then the range values are determined by the specific `DM`.
7045: If the `Mat` was created directly the range values are determined by the local size passed to `MatSetSizes()` or `MatCreateAIJ()`.
7046: If `PETSC_DECIDE` was passed as the local size, then the vector uses default values for the range using `PetscSplitOwnership()`.
7048: For certain `DM`, such as `DMDA`, it is better to use `DM` specific routines, such as `DMDAGetGhostCorners()`, to determine
7049: the local values in the matrix.
7051: The high argument is one more than the last element stored locally.
7053: For all matrices it returns the range of matrix rows associated with rows of a vector that
7054: would contain the result of a matrix vector product with this matrix. See [Matrix
7055: Layouts](sec_matlayout) for details on matrix layouts.
7057: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRanges()`, `MatGetOwnershipRangeColumn()`, `MatGetOwnershipRangesColumn()`, `PetscSplitOwnership()`,
7058: `PetscSplitOwnershipBlock()`, `PetscLayout`, `MatSetSizes()`, `MatCreateAIJ()`, `DMDAGetGhostCorners()`, `DM`
7059: @*/
7060: PetscErrorCode MatGetOwnershipRange(Mat mat, PetscInt *m, PetscInt *n)
7061: {
7062: PetscFunctionBegin;
7065: if (m) PetscAssertPointer(m, 2);
7066: if (n) PetscAssertPointer(n, 3);
7067: MatCheckPreallocated(mat, 1);
7068: if (m) *m = mat->rmap->rstart;
7069: if (n) *n = mat->rmap->rend;
7070: PetscFunctionReturn(PETSC_SUCCESS);
7071: }
7073: /*@C
7074: MatGetOwnershipRanges - For matrices that own values by row, excludes `MATELEMENTAL` and
7075: `MATSCALAPACK`, returns the range of matrix rows owned by each process.
7077: Not Collective, unless matrix has not been allocated
7079: Input Parameter:
7080: . mat - the matrix
7082: Output Parameter:
7083: . ranges - start of each processors portion plus one more than the total length at the end, of length `size` + 1
7084: where `size` is the number of MPI processes used by `mat`
7086: Level: beginner
7088: Notes:
7089: If the `Mat` was obtained from a `DM` with `DMCreateMatrix()`, then the range values are determined by the specific `DM`.
7091: If the `Mat` was created directly the range values are determined by the local size passed to `MatSetSizes()` or `MatCreateAIJ()`.
7092: If `PETSC_DECIDE` was passed as the local size, then the vector uses default values for the range using `PetscSplitOwnership()`.
7094: For certain `DM`, such as `DMDA`, it is better to use `DM` specific routines, such as `DMDAGetGhostCorners()`, to determine
7095: the local values in the matrix.
7097: For all matrices it returns the ranges of matrix rows associated with rows of a vector that
7098: would contain the result of a matrix vector product with this matrix. See [Matrix
7099: Layouts](sec_matlayout) for details on matrix layouts.
7101: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`, `MatGetOwnershipRangesColumn()`, `PetscLayout`,
7102: `PetscSplitOwnership()`, `PetscSplitOwnershipBlock()`, `MatSetSizes()`, `MatCreateAIJ()`,
7103: `DMDAGetGhostCorners()`, `DM`
7104: @*/
7105: PetscErrorCode MatGetOwnershipRanges(Mat mat, const PetscInt *ranges[])
7106: {
7107: PetscFunctionBegin;
7110: MatCheckPreallocated(mat, 1);
7111: PetscCall(PetscLayoutGetRanges(mat->rmap, ranges));
7112: PetscFunctionReturn(PETSC_SUCCESS);
7113: }
7115: /*@C
7116: MatGetOwnershipRangesColumn - Returns the ranges of matrix columns associated with rows of a
7117: vector one multiplies this vector by that are owned by each processor.
7119: Not Collective, unless matrix has not been allocated
7121: Input Parameter:
7122: . mat - the matrix
7124: Output Parameter:
7125: . ranges - start of each processors portion plus one more than the total length at the end
7127: Level: beginner
7129: Notes:
7130: If the `Mat` was obtained from a `DM` with `DMCreateMatrix()`, then the range values are determined by the specific `DM`.
7132: If the `Mat` was created directly the range values are determined by the local size passed to `MatSetSizes()` or `MatCreateAIJ()`.
7133: If `PETSC_DECIDE` was passed as the local size, then the vector uses default values for the range using `PetscSplitOwnership()`.
7135: For certain `DM`, such as `DMDA`, it is better to use `DM` specific routines, such as `DMDAGetGhostCorners()`, to determine
7136: the local values in the matrix.
7138: Returns the columns of the "diagonal blocks", for most sparse matrix formats. See [Matrix
7139: Layouts](sec_matlayout) for details on matrix layouts.
7141: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`, `MatGetOwnershipRanges()`,
7142: `PetscSplitOwnership()`, `PetscSplitOwnershipBlock()`, `PetscLayout`, `MatSetSizes()`, `MatCreateAIJ()`,
7143: `DMDAGetGhostCorners()`, `DM`
7144: @*/
7145: PetscErrorCode MatGetOwnershipRangesColumn(Mat mat, const PetscInt *ranges[])
7146: {
7147: PetscFunctionBegin;
7150: MatCheckPreallocated(mat, 1);
7151: PetscCall(PetscLayoutGetRanges(mat->cmap, ranges));
7152: PetscFunctionReturn(PETSC_SUCCESS);
7153: }
7155: /*@
7156: MatGetOwnershipIS - Get row and column ownership of a matrices' values as index sets.
7158: Not Collective
7160: Input Parameter:
7161: . A - matrix
7163: Output Parameters:
7164: + rows - rows in which this process owns elements, , use `NULL` to not obtain this value
7165: - cols - columns in which this process owns elements, use `NULL` to not obtain this value
7167: Level: intermediate
7169: Note:
7170: You should call `ISDestroy()` on the returned `IS`
7172: For most matrices, excluding `MATELEMENTAL` and `MATSCALAPACK`, this corresponds to values
7173: returned by `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`. For `MATELEMENTAL` and
7174: `MATSCALAPACK` the ownership is more complicated. See [Matrix Layouts](sec_matlayout) for
7175: details on matrix layouts.
7177: .seealso: [](ch_matrices), `IS`, `Mat`, `MatGetOwnershipRanges()`, `MatSetValues()`, `MATELEMENTAL`, `MATSCALAPACK`
7178: @*/
7179: PetscErrorCode MatGetOwnershipIS(Mat A, IS *rows, IS *cols)
7180: {
7181: PetscErrorCode (*f)(Mat, IS *, IS *);
7183: PetscFunctionBegin;
7186: MatCheckPreallocated(A, 1);
7187: PetscCall(PetscObjectQueryFunction((PetscObject)A, "MatGetOwnershipIS_C", &f));
7188: if (f) {
7189: PetscCall((*f)(A, rows, cols));
7190: } else { /* Create a standard row-based partition, each process is responsible for ALL columns in their row block */
7191: if (rows) PetscCall(ISCreateStride(PETSC_COMM_SELF, A->rmap->n, A->rmap->rstart, 1, rows));
7192: if (cols) PetscCall(ISCreateStride(PETSC_COMM_SELF, A->cmap->N, 0, 1, cols));
7193: }
7194: PetscFunctionReturn(PETSC_SUCCESS);
7195: }
7197: /*@
7198: MatILUFactorSymbolic - Performs symbolic ILU factorization of a matrix obtained with `MatGetFactor()`
7199: Uses levels of fill only, not drop tolerance. Use `MatLUFactorNumeric()`
7200: to complete the factorization.
7202: Collective
7204: Input Parameters:
7205: + fact - the factorized matrix obtained with `MatGetFactor()`
7206: . mat - the matrix
7207: . row - row permutation
7208: . col - column permutation
7209: - info - structure containing
7210: .vb
7211: levels - number of levels of fill.
7212: expected fill - as ratio of original fill.
7213: 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
7214: missing diagonal entries)
7215: .ve
7217: Level: developer
7219: Notes:
7220: See [Matrix Factorization](sec_matfactor) for additional information.
7222: Most users should employ the `KSP` interface for linear solvers
7223: instead of working directly with matrix algebra routines such as this.
7224: See, e.g., `KSPCreate()`.
7226: Uses the definition of level of fill as in Y. Saad, {cite}`saad2003`
7228: Fortran Note:
7229: A valid (non-null) `info` argument must be provided
7231: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`,
7232: `MatGetOrdering()`, `MatFactorInfo`
7233: @*/
7234: PetscErrorCode MatILUFactorSymbolic(Mat fact, Mat mat, IS row, IS col, const MatFactorInfo *info)
7235: {
7236: PetscFunctionBegin;
7241: PetscAssertPointer(info, 5);
7242: PetscAssertPointer(fact, 1);
7243: PetscCheck(info->levels >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Levels of fill negative %" PetscInt_FMT, (PetscInt)info->levels);
7244: PetscCheck(info->fill >= 1.0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Expected fill less than 1.0 %g", (double)info->fill);
7245: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7246: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7247: MatCheckPreallocated(mat, 2);
7249: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_ILUFactorSymbolic, mat, row, col, 0));
7250: PetscUseTypeMethod(fact, ilufactorsymbolic, mat, row, col, info);
7251: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_ILUFactorSymbolic, mat, row, col, 0));
7252: PetscFunctionReturn(PETSC_SUCCESS);
7253: }
7255: /*@
7256: MatICCFactorSymbolic - Performs symbolic incomplete
7257: Cholesky factorization for a symmetric matrix. Use
7258: `MatCholeskyFactorNumeric()` to complete the factorization.
7260: Collective
7262: Input Parameters:
7263: + fact - the factorized matrix obtained with `MatGetFactor()`
7264: . mat - the matrix to be factored
7265: . perm - row and column permutation
7266: - info - structure containing
7267: .vb
7268: levels - number of levels of fill.
7269: expected fill - as ratio of original fill.
7270: .ve
7272: Level: developer
7274: Notes:
7275: Most users should employ the `KSP` interface for linear solvers
7276: instead of working directly with matrix algebra routines such as this.
7277: See, e.g., `KSPCreate()`.
7279: This uses the definition of level of fill as in Y. Saad {cite}`saad2003`
7281: Fortran Note:
7282: A valid (non-null) `info` argument must be provided
7284: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatCholeskyFactorNumeric()`, `MatCholeskyFactor()`, `MatFactorInfo`
7285: @*/
7286: PetscErrorCode MatICCFactorSymbolic(Mat fact, Mat mat, IS perm, const MatFactorInfo *info)
7287: {
7288: PetscFunctionBegin;
7292: PetscAssertPointer(info, 4);
7293: PetscAssertPointer(fact, 1);
7294: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7295: PetscCheck(info->levels >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Levels negative %" PetscInt_FMT, (PetscInt)info->levels);
7296: PetscCheck(info->fill >= 1.0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Expected fill less than 1.0 %g", (double)info->fill);
7297: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7298: MatCheckPreallocated(mat, 2);
7300: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_ICCFactorSymbolic, mat, perm, 0, 0));
7301: PetscUseTypeMethod(fact, iccfactorsymbolic, mat, perm, info);
7302: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_ICCFactorSymbolic, mat, perm, 0, 0));
7303: PetscFunctionReturn(PETSC_SUCCESS);
7304: }
7306: /*@C
7307: MatCreateSubMatrices - Extracts several submatrices from a matrix. If submat
7308: points to an array of valid matrices, they may be reused to store the new
7309: submatrices.
7311: Collective
7313: Input Parameters:
7314: + mat - the matrix
7315: . n - the number of submatrixes to be extracted (on this processor, may be zero)
7316: . irow - index set of rows to extract
7317: . icol - index set of columns to extract
7318: - scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
7320: Output Parameter:
7321: . submat - the array of submatrices
7323: Level: advanced
7325: Notes:
7326: `MatCreateSubMatrices()` can extract ONLY sequential submatrices
7327: (from both sequential and parallel matrices). Use `MatCreateSubMatrix()`
7328: to extract a parallel submatrix.
7330: Some matrix types place restrictions on the row and column
7331: indices, such as that they be sorted or that they be equal to each other.
7333: The index sets may not have duplicate entries.
7335: When extracting submatrices from a parallel matrix, each processor can
7336: form a different submatrix by setting the rows and columns of its
7337: individual index sets according to the local submatrix desired.
7339: When finished using the submatrices, the user should destroy
7340: them with `MatDestroySubMatrices()`.
7342: `MAT_REUSE_MATRIX` can only be used when the nonzero structure of the
7343: original matrix has not changed from that last call to `MatCreateSubMatrices()`.
7345: This routine creates the matrices in submat; you should NOT create them before
7346: calling it. It also allocates the array of matrix pointers submat.
7348: For `MATBAIJ` matrices the index sets must respect the block structure, that is if they
7349: request one row/column in a block, they must request all rows/columns that are in
7350: that block. For example, if the block size is 2 you cannot request just row 0 and
7351: column 0.
7353: Fortran Note:
7354: .vb
7355: Mat, pointer :: submat(:)
7356: .ve
7358: .seealso: [](ch_matrices), `Mat`, `MatDestroySubMatrices()`, `MatCreateSubMatrix()`, `MatGetRow()`, `MatGetDiagonal()`, `MatReuse`
7359: @*/
7360: PetscErrorCode MatCreateSubMatrices(Mat mat, PetscInt n, const IS irow[], const IS icol[], MatReuse scall, Mat *submat[])
7361: {
7362: PetscInt i;
7363: PetscBool eq;
7365: PetscFunctionBegin;
7368: if (n) {
7369: PetscAssertPointer(irow, 3);
7371: PetscAssertPointer(icol, 4);
7373: }
7374: PetscAssertPointer(submat, 6);
7375: if (n && scall == MAT_REUSE_MATRIX) {
7376: PetscAssertPointer(*submat, 6);
7378: }
7379: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7380: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7381: MatCheckPreallocated(mat, 1);
7382: PetscCall(PetscLogEventBegin(MAT_CreateSubMats, mat, 0, 0, 0));
7383: PetscUseTypeMethod(mat, createsubmatrices, n, irow, icol, scall, submat);
7384: PetscCall(PetscLogEventEnd(MAT_CreateSubMats, mat, 0, 0, 0));
7385: for (i = 0; i < n; i++) {
7386: (*submat)[i]->factortype = MAT_FACTOR_NONE; /* in case in place factorization was previously done on submatrix */
7387: PetscCall(ISEqualUnsorted(irow[i], icol[i], &eq));
7388: if (eq) PetscCall(MatPropagateSymmetryOptions(mat, (*submat)[i]));
7389: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
7390: if (mat->boundtocpu && mat->bindingpropagates) {
7391: PetscCall(MatBindToCPU((*submat)[i], PETSC_TRUE));
7392: PetscCall(MatSetBindingPropagates((*submat)[i], PETSC_TRUE));
7393: }
7394: #endif
7395: }
7396: PetscFunctionReturn(PETSC_SUCCESS);
7397: }
7399: /*@C
7400: MatCreateSubMatricesMPI - Extracts MPI submatrices across a sub communicator of `mat` (by pairs of `IS` that may live on subcomms).
7402: Collective
7404: Input Parameters:
7405: + mat - the matrix
7406: . n - the number of submatrixes to be extracted
7407: . irow - index set of rows to extract
7408: . icol - index set of columns to extract
7409: - scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
7411: Output Parameter:
7412: . submat - the array of submatrices
7414: Level: advanced
7416: Note:
7417: This is used by `PCGASM`
7419: .seealso: [](ch_matrices), `Mat`, `PCGASM`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRow()`, `MatGetDiagonal()`, `MatReuse`
7420: @*/
7421: PetscErrorCode MatCreateSubMatricesMPI(Mat mat, PetscInt n, const IS irow[], const IS icol[], MatReuse scall, Mat *submat[])
7422: {
7423: PetscInt i;
7424: PetscBool eq;
7426: PetscFunctionBegin;
7429: if (n) {
7430: PetscAssertPointer(irow, 3);
7432: PetscAssertPointer(icol, 4);
7434: }
7435: PetscAssertPointer(submat, 6);
7436: if (n && scall == MAT_REUSE_MATRIX) {
7437: PetscAssertPointer(*submat, 6);
7439: }
7440: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7441: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7442: MatCheckPreallocated(mat, 1);
7444: PetscCall(PetscLogEventBegin(MAT_CreateSubMats, mat, 0, 0, 0));
7445: PetscUseTypeMethod(mat, createsubmatricesmpi, n, irow, icol, scall, submat);
7446: PetscCall(PetscLogEventEnd(MAT_CreateSubMats, mat, 0, 0, 0));
7447: for (i = 0; i < n; i++) {
7448: PetscCall(ISEqualUnsorted(irow[i], icol[i], &eq));
7449: if (eq) PetscCall(MatPropagateSymmetryOptions(mat, (*submat)[i]));
7450: }
7451: PetscFunctionReturn(PETSC_SUCCESS);
7452: }
7454: /*@C
7455: MatDestroyMatrices - Destroys an array of matrices
7457: Collective
7459: Input Parameters:
7460: + n - the number of local matrices
7461: - mat - the matrices (this is a pointer to the array of matrices)
7463: Level: advanced
7465: Notes:
7466: Frees not only the matrices, but also the array that contains the matrices
7468: For matrices obtained with `MatCreateSubMatrices()` use `MatDestroySubMatrices()`
7470: .seealso: [](ch_matrices), `Mat`, `MatCreateSubMatrices()`, `MatDestroySubMatrices()`
7471: @*/
7472: PetscErrorCode MatDestroyMatrices(PetscInt n, Mat *mat[])
7473: {
7474: PetscInt i;
7476: PetscFunctionBegin;
7477: if (!*mat) PetscFunctionReturn(PETSC_SUCCESS);
7478: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Trying to destroy negative number of matrices %" PetscInt_FMT, n);
7479: PetscAssertPointer(mat, 2);
7481: for (i = 0; i < n; i++) PetscCall(MatDestroy(&(*mat)[i]));
7483: /* memory is allocated even if n = 0 */
7484: PetscCall(PetscFree(*mat));
7485: PetscFunctionReturn(PETSC_SUCCESS);
7486: }
7488: /*@C
7489: MatDestroySubMatrices - Destroys a set of matrices obtained with `MatCreateSubMatrices()`.
7491: Collective
7493: Input Parameters:
7494: + n - the number of local matrices
7495: - mat - the matrices (this is a pointer to the array of matrices, to match the calling sequence of `MatCreateSubMatrices()`)
7497: Level: advanced
7499: Note:
7500: Frees not only the matrices, but also the array that contains the matrices
7502: .seealso: [](ch_matrices), `Mat`, `MatCreateSubMatrices()`, `MatDestroyMatrices()`
7503: @*/
7504: PetscErrorCode MatDestroySubMatrices(PetscInt n, Mat *mat[])
7505: {
7506: Mat mat0;
7508: PetscFunctionBegin;
7509: if (!*mat) PetscFunctionReturn(PETSC_SUCCESS);
7510: /* mat[] is an array of length n+1, see MatCreateSubMatrices_xxx() */
7511: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Trying to destroy negative number of matrices %" PetscInt_FMT, n);
7512: PetscAssertPointer(mat, 2);
7514: mat0 = (*mat)[0];
7515: if (mat0 && mat0->ops->destroysubmatrices) {
7516: PetscCall((*mat0->ops->destroysubmatrices)(n, mat));
7517: } else {
7518: PetscCall(MatDestroyMatrices(n, mat));
7519: }
7520: PetscFunctionReturn(PETSC_SUCCESS);
7521: }
7523: /*@
7524: MatGetSeqNonzeroStructure - Extracts the nonzero structure from a matrix and stores it, in its entirety, on each process
7526: Collective
7528: Input Parameter:
7529: . mat - the matrix
7531: Output Parameter:
7532: . matstruct - the sequential matrix with the nonzero structure of `mat`
7534: Level: developer
7536: .seealso: [](ch_matrices), `Mat`, `MatDestroySeqNonzeroStructure()`, `MatCreateSubMatrices()`, `MatDestroyMatrices()`
7537: @*/
7538: PetscErrorCode MatGetSeqNonzeroStructure(Mat mat, Mat *matstruct)
7539: {
7540: PetscFunctionBegin;
7542: PetscAssertPointer(matstruct, 2);
7545: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7546: MatCheckPreallocated(mat, 1);
7548: PetscCall(PetscLogEventBegin(MAT_GetSeqNonzeroStructure, mat, 0, 0, 0));
7549: PetscUseTypeMethod(mat, getseqnonzerostructure, matstruct);
7550: PetscCall(PetscLogEventEnd(MAT_GetSeqNonzeroStructure, mat, 0, 0, 0));
7551: PetscFunctionReturn(PETSC_SUCCESS);
7552: }
7554: /*@C
7555: MatDestroySeqNonzeroStructure - Destroys matrix obtained with `MatGetSeqNonzeroStructure()`.
7557: Collective
7559: Input Parameter:
7560: . mat - the matrix
7562: Level: advanced
7564: Note:
7565: This is not needed, one can just call `MatDestroy()`
7567: .seealso: [](ch_matrices), `Mat`, `MatGetSeqNonzeroStructure()`
7568: @*/
7569: PetscErrorCode MatDestroySeqNonzeroStructure(Mat *mat)
7570: {
7571: PetscFunctionBegin;
7572: PetscAssertPointer(mat, 1);
7573: PetscCall(MatDestroy(mat));
7574: PetscFunctionReturn(PETSC_SUCCESS);
7575: }
7577: /*@
7578: MatIncreaseOverlap - Given a set of submatrices indicated by index sets,
7579: replaces the index sets by larger ones that represent submatrices with
7580: additional overlap.
7582: Collective
7584: Input Parameters:
7585: + mat - the matrix
7586: . n - the number of index sets
7587: . is - the array of index sets (these index sets will changed during the call)
7588: - ov - the additional overlap requested
7590: Options Database Key:
7591: . -mat_increase_overlap_scalable - use a scalable algorithm to compute the overlap (supported by MPIAIJ matrix)
7593: Level: developer
7595: Note:
7596: The computed overlap preserves the matrix block sizes when the blocks are square.
7597: That is: if a matrix nonzero for a given block would increase the overlap all columns associated with
7598: that block are included in the overlap regardless of whether each specific column would increase the overlap.
7600: .seealso: [](ch_matrices), `Mat`, `PCASM`, `MatSetBlockSize()`, `MatIncreaseOverlapSplit()`, `MatCreateSubMatrices()`
7601: @*/
7602: PetscErrorCode MatIncreaseOverlap(Mat mat, PetscInt n, IS is[], PetscInt ov)
7603: {
7604: PetscInt i, bs, cbs;
7606: PetscFunctionBegin;
7610: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Must have one or more domains, you have %" PetscInt_FMT, n);
7611: if (n) {
7612: PetscAssertPointer(is, 3);
7614: }
7615: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7616: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7617: MatCheckPreallocated(mat, 1);
7619: if (!ov || !n) PetscFunctionReturn(PETSC_SUCCESS);
7620: PetscCall(PetscLogEventBegin(MAT_IncreaseOverlap, mat, 0, 0, 0));
7621: PetscUseTypeMethod(mat, increaseoverlap, n, is, ov);
7622: PetscCall(PetscLogEventEnd(MAT_IncreaseOverlap, mat, 0, 0, 0));
7623: PetscCall(MatGetBlockSizes(mat, &bs, &cbs));
7624: if (bs == cbs) {
7625: for (i = 0; i < n; i++) PetscCall(ISSetBlockSize(is[i], bs));
7626: }
7627: PetscFunctionReturn(PETSC_SUCCESS);
7628: }
7630: PetscErrorCode MatIncreaseOverlapSplit_Single(Mat, IS *, PetscInt);
7632: /*@
7633: MatIncreaseOverlapSplit - Given a set of submatrices indicated by index sets across
7634: a sub communicator, replaces the index sets by larger ones that represent submatrices with
7635: additional overlap.
7637: Collective
7639: Input Parameters:
7640: + mat - the matrix
7641: . n - the number of index sets
7642: . is - the array of index sets (these index sets will changed during the call)
7643: - ov - the additional overlap requested
7645: ` Options Database Key:
7646: . -mat_increase_overlap_scalable - use a scalable algorithm to compute the overlap (supported by MPIAIJ matrix)
7648: Level: developer
7650: .seealso: [](ch_matrices), `Mat`, `MatCreateSubMatrices()`, `MatIncreaseOverlap()`
7651: @*/
7652: PetscErrorCode MatIncreaseOverlapSplit(Mat mat, PetscInt n, IS is[], PetscInt ov)
7653: {
7654: PetscInt i;
7656: PetscFunctionBegin;
7659: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Must have one or more domains, you have %" PetscInt_FMT, n);
7660: if (n) {
7661: PetscAssertPointer(is, 3);
7663: }
7664: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7665: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7666: MatCheckPreallocated(mat, 1);
7667: if (!ov) PetscFunctionReturn(PETSC_SUCCESS);
7668: PetscCall(PetscLogEventBegin(MAT_IncreaseOverlap, mat, 0, 0, 0));
7669: for (i = 0; i < n; i++) PetscCall(MatIncreaseOverlapSplit_Single(mat, &is[i], ov));
7670: PetscCall(PetscLogEventEnd(MAT_IncreaseOverlap, mat, 0, 0, 0));
7671: PetscFunctionReturn(PETSC_SUCCESS);
7672: }
7674: /*@
7675: MatGetBlockSize - Returns the matrix block size.
7677: Not Collective
7679: Input Parameter:
7680: . mat - the matrix
7682: Output Parameter:
7683: . bs - block size
7685: Level: intermediate
7687: Notes:
7688: Block row formats are `MATBAIJ` and `MATSBAIJ` ALWAYS have square block storage in the matrix.
7690: If the block size has not been set yet this routine returns 1.
7692: .seealso: [](ch_matrices), `Mat`, `MATBAIJ`, `MATSBAIJ`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSizes()`
7693: @*/
7694: PetscErrorCode MatGetBlockSize(Mat mat, PetscInt *bs)
7695: {
7696: PetscFunctionBegin;
7698: PetscAssertPointer(bs, 2);
7699: *bs = mat->rmap->bs;
7700: PetscFunctionReturn(PETSC_SUCCESS);
7701: }
7703: /*@
7704: MatGetBlockSizes - Returns the matrix block row and column sizes.
7706: Not Collective
7708: Input Parameter:
7709: . mat - the matrix
7711: Output Parameters:
7712: + rbs - row block size
7713: - cbs - column block size
7715: Level: intermediate
7717: Notes:
7718: Block row formats are `MATBAIJ` and `MATSBAIJ` ALWAYS have square block storage in the matrix.
7719: If you pass a different block size for the columns than the rows, the row block size determines the square block storage.
7721: If a block size has not been set yet this routine returns 1.
7723: .seealso: [](ch_matrices), `Mat`, `MATBAIJ`, `MATSBAIJ`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSize()`, `MatSetBlockSizes()`
7724: @*/
7725: PetscErrorCode MatGetBlockSizes(Mat mat, PetscInt *rbs, PetscInt *cbs)
7726: {
7727: PetscFunctionBegin;
7729: if (rbs) PetscAssertPointer(rbs, 2);
7730: if (cbs) PetscAssertPointer(cbs, 3);
7731: if (rbs) *rbs = mat->rmap->bs;
7732: if (cbs) *cbs = mat->cmap->bs;
7733: PetscFunctionReturn(PETSC_SUCCESS);
7734: }
7736: /*@
7737: MatSetBlockSize - Sets the matrix block size.
7739: Logically Collective
7741: Input Parameters:
7742: + mat - the matrix
7743: - bs - block size
7745: Level: intermediate
7747: Notes:
7748: Block row formats are `MATBAIJ` and `MATSBAIJ` formats ALWAYS have square block storage in the matrix.
7749: This must be called before `MatSetUp()` or MatXXXSetPreallocation() (or will default to 1) and the block size cannot be changed later.
7751: For `MATAIJ` matrix format, this function can be called at a later stage, provided that the specified block size
7752: is compatible with the matrix local sizes.
7754: .seealso: [](ch_matrices), `Mat`, `MATBAIJ`, `MATSBAIJ`, `MATAIJ`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`, `MatGetBlockSizes()`
7755: @*/
7756: PetscErrorCode MatSetBlockSize(Mat mat, PetscInt bs)
7757: {
7758: PetscFunctionBegin;
7761: PetscCall(MatSetBlockSizes(mat, bs, bs));
7762: PetscFunctionReturn(PETSC_SUCCESS);
7763: }
7765: typedef struct {
7766: PetscInt n;
7767: IS *is;
7768: Mat *mat;
7769: PetscObjectState nonzerostate;
7770: Mat C;
7771: } EnvelopeData;
7773: static PetscErrorCode EnvelopeDataDestroy(PetscCtxRt ptr)
7774: {
7775: EnvelopeData *edata = *(EnvelopeData **)ptr;
7777: PetscFunctionBegin;
7778: for (PetscInt i = 0; i < edata->n; i++) PetscCall(ISDestroy(&edata->is[i]));
7779: PetscCall(PetscFree(edata->is));
7780: PetscCall(PetscFree(edata));
7781: PetscFunctionReturn(PETSC_SUCCESS);
7782: }
7784: /*@
7785: MatComputeVariableBlockEnvelope - Given a matrix whose nonzeros are in blocks along the diagonal this computes and stores
7786: the sizes of these blocks in the matrix. An individual block may lie over several processes.
7788: Collective
7790: Input Parameter:
7791: . mat - the matrix
7793: Level: intermediate
7795: Notes:
7796: There can be zeros within the blocks
7798: The blocks can overlap between processes, including laying on more than two processes
7800: .seealso: [](ch_matrices), `Mat`, `MatInvertVariableBlockEnvelope()`, `MatSetVariableBlockSizes()`
7801: @*/
7802: PetscErrorCode MatComputeVariableBlockEnvelope(Mat mat)
7803: {
7804: PetscInt n, *sizes, *starts, i = 0, env = 0, tbs = 0, lblocks = 0, rstart, II, ln = 0, cnt = 0, cstart, cend;
7805: PetscInt *diag, *odiag, sc;
7806: VecScatter scatter;
7807: PetscScalar *seqv;
7808: const PetscScalar *parv;
7809: const PetscInt *ia, *ja;
7810: PetscBool set, flag, done;
7811: Mat AA = mat, A;
7812: MPI_Comm comm;
7813: PetscMPIInt rank, size, tag;
7814: MPI_Status status;
7815: PetscContainer container;
7816: EnvelopeData *edata;
7817: Vec seq, par;
7818: IS isglobal;
7820: PetscFunctionBegin;
7822: PetscCall(MatIsSymmetricKnown(mat, &set, &flag));
7823: if (!set || !flag) {
7824: /* TODO: only needs nonzero structure of transpose */
7825: PetscCall(MatTranspose(mat, MAT_INITIAL_MATRIX, &AA));
7826: PetscCall(MatAXPY(AA, 1.0, mat, DIFFERENT_NONZERO_PATTERN));
7827: }
7828: PetscCall(MatAIJGetLocalMat(AA, &A));
7829: PetscCall(MatGetRowIJ(A, 0, PETSC_FALSE, PETSC_FALSE, &n, &ia, &ja, &done));
7830: PetscCheck(done, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Unable to get IJ structure from matrix");
7832: PetscCall(MatGetLocalSize(mat, &n, NULL));
7833: PetscCall(PetscObjectGetNewTag((PetscObject)mat, &tag));
7834: PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
7835: PetscCallMPI(MPI_Comm_size(comm, &size));
7836: PetscCallMPI(MPI_Comm_rank(comm, &rank));
7838: PetscCall(PetscMalloc2(n, &sizes, n, &starts));
7840: if (rank > 0) {
7841: PetscCallMPI(MPI_Recv(&env, 1, MPIU_INT, rank - 1, tag, comm, &status));
7842: PetscCallMPI(MPI_Recv(&tbs, 1, MPIU_INT, rank - 1, tag, comm, &status));
7843: }
7844: PetscCall(MatGetOwnershipRange(mat, &rstart, NULL));
7845: for (i = 0; i < n; i++) {
7846: env = PetscMax(env, ja[ia[i + 1] - 1]);
7847: II = rstart + i;
7848: if (env == II) {
7849: starts[lblocks] = tbs;
7850: sizes[lblocks++] = 1 + II - tbs;
7851: tbs = 1 + II;
7852: }
7853: }
7854: if (rank < size - 1) {
7855: PetscCallMPI(MPI_Send(&env, 1, MPIU_INT, rank + 1, tag, comm));
7856: PetscCallMPI(MPI_Send(&tbs, 1, MPIU_INT, rank + 1, tag, comm));
7857: }
7859: PetscCall(MatRestoreRowIJ(A, 0, PETSC_FALSE, PETSC_FALSE, &n, &ia, &ja, &done));
7860: if (!set || !flag) PetscCall(MatDestroy(&AA));
7861: PetscCall(MatDestroy(&A));
7863: PetscCall(PetscNew(&edata));
7864: PetscCall(MatGetNonzeroState(mat, &edata->nonzerostate));
7865: edata->n = lblocks;
7866: /* create IS needed for extracting blocks from the original matrix */
7867: PetscCall(PetscMalloc1(lblocks, &edata->is));
7868: for (PetscInt i = 0; i < lblocks; i++) PetscCall(ISCreateStride(PETSC_COMM_SELF, sizes[i], starts[i], 1, &edata->is[i]));
7870: /* Create the resulting inverse matrix nonzero structure with preallocation information */
7871: PetscCall(MatCreate(PetscObjectComm((PetscObject)mat), &edata->C));
7872: PetscCall(MatSetSizes(edata->C, mat->rmap->n, mat->cmap->n, mat->rmap->N, mat->cmap->N));
7873: PetscCall(MatSetBlockSizesFromMats(edata->C, mat, mat));
7874: PetscCall(MatSetType(edata->C, MATAIJ));
7876: /* Communicate the start and end of each row, from each block to the correct rank */
7877: /* TODO: Use PetscSF instead of VecScatter */
7878: for (PetscInt i = 0; i < lblocks; i++) ln += sizes[i];
7879: PetscCall(VecCreateSeq(PETSC_COMM_SELF, 2 * ln, &seq));
7880: PetscCall(VecGetArrayWrite(seq, &seqv));
7881: for (PetscInt i = 0; i < lblocks; i++) {
7882: for (PetscInt j = 0; j < sizes[i]; j++) {
7883: seqv[cnt] = starts[i];
7884: seqv[cnt + 1] = starts[i] + sizes[i];
7885: cnt += 2;
7886: }
7887: }
7888: PetscCall(VecRestoreArrayWrite(seq, &seqv));
7889: PetscCallMPI(MPI_Scan(&cnt, &sc, 1, MPIU_INT, MPI_SUM, PetscObjectComm((PetscObject)mat)));
7890: sc -= cnt;
7891: PetscCall(VecCreateMPI(PetscObjectComm((PetscObject)mat), 2 * mat->rmap->n, 2 * mat->rmap->N, &par));
7892: PetscCall(ISCreateStride(PETSC_COMM_SELF, cnt, sc, 1, &isglobal));
7893: PetscCall(VecScatterCreate(seq, NULL, par, isglobal, &scatter));
7894: PetscCall(ISDestroy(&isglobal));
7895: PetscCall(VecScatterBegin(scatter, seq, par, INSERT_VALUES, SCATTER_FORWARD));
7896: PetscCall(VecScatterEnd(scatter, seq, par, INSERT_VALUES, SCATTER_FORWARD));
7897: PetscCall(VecScatterDestroy(&scatter));
7898: PetscCall(VecDestroy(&seq));
7899: PetscCall(MatGetOwnershipRangeColumn(mat, &cstart, &cend));
7900: PetscCall(PetscMalloc2(mat->rmap->n, &diag, mat->rmap->n, &odiag));
7901: PetscCall(VecGetArrayRead(par, &parv));
7902: cnt = 0;
7903: PetscCall(MatGetSize(mat, NULL, &n));
7904: for (PetscInt i = 0; i < mat->rmap->n; i++) {
7905: PetscInt start, end, d = 0, od = 0;
7907: start = (PetscInt)PetscRealPart(parv[cnt]);
7908: end = (PetscInt)PetscRealPart(parv[cnt + 1]);
7909: cnt += 2;
7911: if (start < cstart) {
7912: od += cstart - start + n - cend;
7913: d += cend - cstart;
7914: } else if (start < cend) {
7915: od += n - cend;
7916: d += cend - start;
7917: } else od += n - start;
7918: if (end <= cstart) {
7919: od -= cstart - end + n - cend;
7920: d -= cend - cstart;
7921: } else if (end < cend) {
7922: od -= n - cend;
7923: d -= cend - end;
7924: } else od -= n - end;
7926: odiag[i] = od;
7927: diag[i] = d;
7928: }
7929: PetscCall(VecRestoreArrayRead(par, &parv));
7930: PetscCall(VecDestroy(&par));
7931: PetscCall(MatXAIJSetPreallocation(edata->C, mat->rmap->bs, diag, odiag, NULL, NULL));
7932: PetscCall(PetscFree2(diag, odiag));
7933: PetscCall(PetscFree2(sizes, starts));
7935: PetscCall(PetscContainerCreate(PETSC_COMM_SELF, &container));
7936: PetscCall(PetscContainerSetPointer(container, edata));
7937: PetscCall(PetscContainerSetCtxDestroy(container, EnvelopeDataDestroy));
7938: PetscCall(PetscObjectCompose((PetscObject)mat, "EnvelopeData", (PetscObject)container));
7939: PetscCall(PetscObjectDereference((PetscObject)container));
7940: PetscFunctionReturn(PETSC_SUCCESS);
7941: }
7943: /*@
7944: MatInvertVariableBlockEnvelope - set matrix C to be the inverted block diagonal of matrix A
7946: Collective
7948: Input Parameters:
7949: + A - the matrix
7950: - reuse - indicates if the `C` matrix was obtained from a previous call to this routine
7952: Output Parameter:
7953: . C - matrix with inverted block diagonal of `A`
7955: Level: advanced
7957: Note:
7958: For efficiency the matrix `A` should have all the nonzero entries clustered in smallish blocks along the diagonal.
7960: .seealso: [](ch_matrices), `Mat`, `MatInvertBlockDiagonal()`, `MatComputeBlockDiagonal()`
7961: @*/
7962: PetscErrorCode MatInvertVariableBlockEnvelope(Mat A, MatReuse reuse, Mat *C)
7963: {
7964: PetscContainer container;
7965: EnvelopeData *edata;
7966: PetscObjectState nonzerostate;
7968: PetscFunctionBegin;
7969: PetscCall(PetscObjectQuery((PetscObject)A, "EnvelopeData", (PetscObject *)&container));
7970: if (!container) {
7971: PetscCall(MatComputeVariableBlockEnvelope(A));
7972: PetscCall(PetscObjectQuery((PetscObject)A, "EnvelopeData", (PetscObject *)&container));
7973: }
7974: PetscCall(PetscContainerGetPointer(container, &edata));
7975: PetscCall(MatGetNonzeroState(A, &nonzerostate));
7976: PetscCheck(nonzerostate <= edata->nonzerostate, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Cannot handle changes to matrix nonzero structure");
7977: PetscCheck(reuse != MAT_REUSE_MATRIX || *C == edata->C, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "C matrix must be the same as previously output");
7979: PetscCall(MatCreateSubMatrices(A, edata->n, edata->is, edata->is, MAT_INITIAL_MATRIX, &edata->mat));
7980: *C = edata->C;
7982: for (PetscInt i = 0; i < edata->n; i++) {
7983: Mat D;
7984: PetscScalar *dvalues;
7986: PetscCall(MatConvert(edata->mat[i], MATSEQDENSE, MAT_INITIAL_MATRIX, &D));
7987: PetscCall(MatSetOption(*C, MAT_ROW_ORIENTED, PETSC_FALSE));
7988: PetscCall(MatSeqDenseInvert(D));
7989: PetscCall(MatDenseGetArray(D, &dvalues));
7990: PetscCall(MatSetValuesIS(*C, edata->is[i], edata->is[i], dvalues, INSERT_VALUES));
7991: PetscCall(MatDestroy(&D));
7992: }
7993: PetscCall(MatDestroySubMatrices(edata->n, &edata->mat));
7994: PetscCall(MatAssemblyBegin(*C, MAT_FINAL_ASSEMBLY));
7995: PetscCall(MatAssemblyEnd(*C, MAT_FINAL_ASSEMBLY));
7996: PetscFunctionReturn(PETSC_SUCCESS);
7997: }
7999: /*@
8000: MatSetVariableBlockSizes - Sets diagonal point-blocks of the matrix that need not be of the same size
8002: Not Collective
8004: Input Parameters:
8005: + mat - the matrix
8006: . nblocks - the number of blocks on this process, each block can only exist on a single process
8007: - bsizes - the block sizes
8009: Level: intermediate
8011: Notes:
8012: Currently used by `PCVPBJACOBI` for `MATAIJ` matrices
8014: Each variable point-block set of degrees of freedom must live on a single MPI process. That is a point block cannot straddle two MPI processes.
8016: .seealso: [](ch_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`, `MatGetBlockSizes()`, `MatGetVariableBlockSizes()`,
8017: `MatComputeVariableBlockEnvelope()`, `PCVPBJACOBI`
8018: @*/
8019: PetscErrorCode MatSetVariableBlockSizes(Mat mat, PetscInt nblocks, const PetscInt bsizes[])
8020: {
8021: PetscInt ncnt = 0, nlocal;
8023: PetscFunctionBegin;
8025: PetscCall(MatGetLocalSize(mat, &nlocal, NULL));
8026: PetscCheck(nblocks >= 0 && nblocks <= nlocal, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Number of local blocks %" PetscInt_FMT " is not in [0, %" PetscInt_FMT "]", nblocks, nlocal);
8027: for (PetscInt i = 0; i < nblocks; i++) ncnt += bsizes[i];
8028: PetscCheck(ncnt == nlocal, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Sum of local block sizes %" PetscInt_FMT " does not equal local size of matrix %" PetscInt_FMT, ncnt, nlocal);
8029: PetscCall(PetscFree(mat->bsizes));
8030: mat->nblocks = nblocks;
8031: PetscCall(PetscMalloc1(nblocks, &mat->bsizes));
8032: PetscCall(PetscArraycpy(mat->bsizes, bsizes, nblocks));
8033: PetscFunctionReturn(PETSC_SUCCESS);
8034: }
8036: /*@C
8037: MatGetVariableBlockSizes - Gets a diagonal blocks of the matrix that need not be of the same size
8039: Not Collective; No Fortran Support
8041: Input Parameter:
8042: . mat - the matrix
8044: Output Parameters:
8045: + nblocks - the number of blocks on this process
8046: - bsizes - the block sizes
8048: Level: intermediate
8050: .seealso: [](ch_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`, `MatGetBlockSizes()`, `MatSetVariableBlockSizes()`, `MatComputeVariableBlockEnvelope()`
8051: @*/
8052: PetscErrorCode MatGetVariableBlockSizes(Mat mat, PetscInt *nblocks, const PetscInt *bsizes[])
8053: {
8054: PetscFunctionBegin;
8056: if (nblocks) *nblocks = mat->nblocks;
8057: if (bsizes) *bsizes = mat->bsizes;
8058: PetscFunctionReturn(PETSC_SUCCESS);
8059: }
8061: /*@
8062: MatSelectVariableBlockSizes - When creating a submatrix, pass on the variable block sizes
8064: Not Collective
8066: Input Parameter:
8067: + subA - the submatrix
8068: . A - the original matrix
8069: - isrow - The `IS` of selected rows for the submatrix, must be sorted
8071: Level: developer
8073: Notes:
8074: If the index set is not sorted or contains off-process entries, this function will do nothing.
8076: .seealso: [](ch_matrices), `Mat`, `MatSetVariableBlockSizes()`, `MatComputeVariableBlockEnvelope()`
8077: @*/
8078: PetscErrorCode MatSelectVariableBlockSizes(Mat subA, Mat A, IS isrow)
8079: {
8080: const PetscInt *rows;
8081: PetscInt n, rStart, rEnd, Nb = 0;
8082: PetscBool flg = A->bsizes ? PETSC_TRUE : PETSC_FALSE;
8084: PetscFunctionBegin;
8085: // The code for block size extraction does not support an unsorted IS
8086: if (flg) PetscCall(ISSorted(isrow, &flg));
8087: // We don't support originally off-diagonal blocks
8088: if (flg) {
8089: PetscCall(MatGetOwnershipRange(A, &rStart, &rEnd));
8090: PetscCall(ISGetLocalSize(isrow, &n));
8091: PetscCall(ISGetIndices(isrow, &rows));
8092: for (PetscInt i = 0; i < n && flg; ++i) {
8093: if (rows[i] < rStart || rows[i] >= rEnd) flg = PETSC_FALSE;
8094: }
8095: PetscCall(ISRestoreIndices(isrow, &rows));
8096: }
8097: // quiet return if we can't extract block size
8098: PetscCallMPI(MPIU_Allreduce(MPI_IN_PLACE, &flg, 1, MPI_C_BOOL, MPI_LAND, PetscObjectComm((PetscObject)subA)));
8099: if (!flg) PetscFunctionReturn(PETSC_SUCCESS);
8101: // extract block sizes
8102: PetscCall(ISGetIndices(isrow, &rows));
8103: for (PetscInt b = 0, gr = rStart, i = 0; b < A->nblocks; ++b) {
8104: PetscBool occupied = PETSC_FALSE;
8106: for (PetscInt br = 0; br < A->bsizes[b]; ++br) {
8107: const PetscInt row = gr + br;
8109: if (i == n) break;
8110: if (rows[i] == row) {
8111: occupied = PETSC_TRUE;
8112: ++i;
8113: }
8114: while (i < n && rows[i] < row) ++i;
8115: }
8116: gr += A->bsizes[b];
8117: if (occupied) ++Nb;
8118: }
8119: subA->nblocks = Nb;
8120: PetscCall(PetscFree(subA->bsizes));
8121: PetscCall(PetscMalloc1(subA->nblocks, &subA->bsizes));
8122: PetscInt sb = 0;
8123: for (PetscInt b = 0, gr = rStart, i = 0; b < A->nblocks; ++b) {
8124: if (sb < subA->nblocks) subA->bsizes[sb] = 0;
8125: for (PetscInt br = 0; br < A->bsizes[b]; ++br) {
8126: const PetscInt row = gr + br;
8128: if (i == n) break;
8129: if (rows[i] == row) {
8130: ++subA->bsizes[sb];
8131: ++i;
8132: }
8133: while (i < n && rows[i] < row) ++i;
8134: }
8135: gr += A->bsizes[b];
8136: if (sb < subA->nblocks && subA->bsizes[sb]) ++sb;
8137: }
8138: PetscCheck(sb == subA->nblocks, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Invalid number of blocks %" PetscInt_FMT " != %" PetscInt_FMT, sb, subA->nblocks);
8139: PetscInt nlocal, ncnt = 0;
8140: PetscCall(MatGetLocalSize(subA, &nlocal, NULL));
8141: PetscCheck(subA->nblocks >= 0 && subA->nblocks <= nlocal, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Number of local blocks %" PetscInt_FMT " is not in [0, %" PetscInt_FMT "]", subA->nblocks, nlocal);
8142: for (PetscInt i = 0; i < subA->nblocks; i++) ncnt += subA->bsizes[i];
8143: PetscCheck(ncnt == nlocal, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Sum of local block sizes %" PetscInt_FMT " does not equal local size of matrix %" PetscInt_FMT, ncnt, nlocal);
8144: PetscCall(ISRestoreIndices(isrow, &rows));
8145: PetscFunctionReturn(PETSC_SUCCESS);
8146: }
8148: /*@
8149: MatSetBlockSizes - Sets the matrix block row and column sizes.
8151: Logically Collective
8153: Input Parameters:
8154: + mat - the matrix
8155: . rbs - row block size
8156: - cbs - column block size
8158: Level: intermediate
8160: Notes:
8161: Block row formats are `MATBAIJ` and `MATSBAIJ`. These formats ALWAYS have square block storage in the matrix.
8162: If you pass a different block size for the columns than the rows, the row block size determines the square block storage.
8163: This must be called before `MatSetUp()` or MatXXXSetPreallocation() (or will default to 1) and the block size cannot be changed later.
8165: For `MATAIJ` matrix this function can be called at a later stage, provided that the specified block sizes
8166: are compatible with the matrix local sizes.
8168: The row and column block size determine the blocksize of the "row" and "column" vectors returned by `MatCreateVecs()`.
8170: .seealso: [](ch_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSize()`, `MatGetBlockSizes()`
8171: @*/
8172: PetscErrorCode MatSetBlockSizes(Mat mat, PetscInt rbs, PetscInt cbs)
8173: {
8174: PetscFunctionBegin;
8178: PetscTryTypeMethod(mat, setblocksizes, rbs, cbs);
8179: if (mat->rmap->refcnt) {
8180: ISLocalToGlobalMapping l2g = NULL;
8181: PetscLayout nmap = NULL;
8183: PetscCall(PetscLayoutDuplicate(mat->rmap, &nmap));
8184: if (mat->rmap->mapping) PetscCall(ISLocalToGlobalMappingDuplicate(mat->rmap->mapping, &l2g));
8185: PetscCall(PetscLayoutDestroy(&mat->rmap));
8186: mat->rmap = nmap;
8187: mat->rmap->mapping = l2g;
8188: }
8189: if (mat->cmap->refcnt) {
8190: ISLocalToGlobalMapping l2g = NULL;
8191: PetscLayout nmap = NULL;
8193: PetscCall(PetscLayoutDuplicate(mat->cmap, &nmap));
8194: if (mat->cmap->mapping) PetscCall(ISLocalToGlobalMappingDuplicate(mat->cmap->mapping, &l2g));
8195: PetscCall(PetscLayoutDestroy(&mat->cmap));
8196: mat->cmap = nmap;
8197: mat->cmap->mapping = l2g;
8198: }
8199: PetscCall(PetscLayoutSetBlockSize(mat->rmap, rbs));
8200: PetscCall(PetscLayoutSetBlockSize(mat->cmap, cbs));
8201: PetscFunctionReturn(PETSC_SUCCESS);
8202: }
8204: /*@
8205: MatSetBlockSizesFromMats - Sets the matrix block row and column sizes to match a pair of matrices
8207: Logically Collective
8209: Input Parameters:
8210: + mat - the matrix
8211: . fromRow - matrix from which to copy row block size
8212: - fromCol - matrix from which to copy column block size (can be same as `fromRow`)
8214: Level: developer
8216: .seealso: [](ch_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`
8217: @*/
8218: PetscErrorCode MatSetBlockSizesFromMats(Mat mat, Mat fromRow, Mat fromCol)
8219: {
8220: PetscFunctionBegin;
8224: PetscTryTypeMethod(mat, setblocksizes, fromRow->rmap->bs, fromCol->cmap->bs);
8225: PetscCall(PetscLayoutSetBlockSize(mat->rmap, fromRow->rmap->bs));
8226: PetscCall(PetscLayoutSetBlockSize(mat->cmap, fromCol->cmap->bs));
8227: PetscFunctionReturn(PETSC_SUCCESS);
8228: }
8230: /*@
8231: MatResidual - Default routine to calculate the residual r = b - Ax
8233: Collective
8235: Input Parameters:
8236: + mat - the matrix
8237: . b - the right-hand-side
8238: - x - the approximate solution
8240: Output Parameter:
8241: . r - location to store the residual
8243: Level: developer
8245: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatMultAdd()`, `PCMGSetResidual()`
8246: @*/
8247: PetscErrorCode MatResidual(Mat mat, Vec b, Vec x, Vec r)
8248: {
8249: PetscFunctionBegin;
8255: MatCheckPreallocated(mat, 1);
8256: PetscCall(PetscLogEventBegin(MAT_Residual, mat, 0, 0, 0));
8257: if (!mat->ops->residual) {
8258: PetscCall(MatMult(mat, x, r));
8259: PetscCall(VecAYPX(r, -1.0, b));
8260: } else {
8261: PetscUseTypeMethod(mat, residual, b, x, r);
8262: }
8263: PetscCall(PetscLogEventEnd(MAT_Residual, mat, 0, 0, 0));
8264: PetscFunctionReturn(PETSC_SUCCESS);
8265: }
8267: /*@C
8268: MatGetRowIJ - Returns the compressed row storage i and j indices for the local rows of a sparse matrix
8270: Collective
8272: Input Parameters:
8273: + mat - the matrix
8274: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
8275: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
8276: - inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8277: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8278: always used.
8280: Output Parameters:
8281: + n - number of local rows in the (possibly compressed) matrix, use `NULL` if not needed
8282: . ia - the row pointers; that is ia[0] = 0, ia[row] = ia[row-1] + number of elements in that row of the matrix, use `NULL` if not needed
8283: . ja - the column indices, use `NULL` if not needed
8284: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
8285: are responsible for handling the case when done == `PETSC_FALSE` and ia and ja are not set
8287: Level: developer
8289: Notes:
8290: You CANNOT change any of the ia[] or ja[] values.
8292: Use `MatRestoreRowIJ()` when you are finished accessing the ia[] and ja[] values.
8294: Fortran Notes:
8295: Use
8296: .vb
8297: PetscInt, pointer :: ia(:),ja(:)
8298: call MatGetRowIJ(mat,shift,symmetric,inodecompressed,n,ia,ja,done,ierr)
8299: ! Access the ith and jth entries via ia(i) and ja(j)
8300: .ve
8302: .seealso: [](ch_matrices), `Mat`, `MATAIJ`, `MatGetColumnIJ()`, `MatRestoreRowIJ()`, `MatSeqAIJGetArray()`
8303: @*/
8304: PetscErrorCode MatGetRowIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
8305: {
8306: PetscFunctionBegin;
8309: if (n) PetscAssertPointer(n, 5);
8310: if (ia) PetscAssertPointer(ia, 6);
8311: if (ja) PetscAssertPointer(ja, 7);
8312: if (done) PetscAssertPointer(done, 8);
8313: MatCheckPreallocated(mat, 1);
8314: if (!mat->ops->getrowij && done) *done = PETSC_FALSE;
8315: else {
8316: if (done) *done = PETSC_TRUE;
8317: PetscCall(PetscLogEventBegin(MAT_GetRowIJ, mat, 0, 0, 0));
8318: PetscUseTypeMethod(mat, getrowij, shift, symmetric, inodecompressed, n, ia, ja, done);
8319: PetscCall(PetscLogEventEnd(MAT_GetRowIJ, mat, 0, 0, 0));
8320: }
8321: PetscFunctionReturn(PETSC_SUCCESS);
8322: }
8324: /*@C
8325: MatGetColumnIJ - Returns the compressed column storage i and j indices for sequential matrices.
8327: Collective
8329: Input Parameters:
8330: + mat - the matrix
8331: . shift - 1 or zero indicating we want the indices starting at 0 or 1
8332: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be
8333: symmetrized
8334: - inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8335: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8336: always used.
8338: Output Parameters:
8339: + n - number of columns in the (possibly compressed) matrix
8340: . ia - the column pointers; that is ia[0] = 0, ia[col] = i[col-1] + number of elements in that col of the matrix
8341: . ja - the row indices
8342: - done - `PETSC_TRUE` or `PETSC_FALSE`, indicating whether the values have been returned
8344: Level: developer
8346: .seealso: [](ch_matrices), `Mat`, `MatGetRowIJ()`, `MatRestoreColumnIJ()`
8347: @*/
8348: PetscErrorCode MatGetColumnIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
8349: {
8350: PetscFunctionBegin;
8353: PetscAssertPointer(n, 5);
8354: if (ia) PetscAssertPointer(ia, 6);
8355: if (ja) PetscAssertPointer(ja, 7);
8356: PetscAssertPointer(done, 8);
8357: MatCheckPreallocated(mat, 1);
8358: if (!mat->ops->getcolumnij) *done = PETSC_FALSE;
8359: else {
8360: *done = PETSC_TRUE;
8361: PetscUseTypeMethod(mat, getcolumnij, shift, symmetric, inodecompressed, n, ia, ja, done);
8362: }
8363: PetscFunctionReturn(PETSC_SUCCESS);
8364: }
8366: /*@C
8367: MatRestoreRowIJ - Call after you are completed with the ia,ja indices obtained with `MatGetRowIJ()`.
8369: Collective
8371: Input Parameters:
8372: + mat - the matrix
8373: . shift - 1 or zero indicating we want the indices starting at 0 or 1
8374: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
8375: . inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8376: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8377: always used.
8378: . n - size of (possibly compressed) matrix
8379: . ia - the row pointers
8380: - ja - the column indices
8382: Output Parameter:
8383: . done - `PETSC_TRUE` or `PETSC_FALSE` indicated that the values have been returned
8385: Level: developer
8387: Note:
8388: This routine zeros out `n`, `ia`, and `ja`. This is to prevent accidental
8389: us of the array after it has been restored. If you pass `NULL`, it will
8390: not zero the pointers. Use of ia or ja after `MatRestoreRowIJ()` is invalid.
8392: .seealso: [](ch_matrices), `Mat`, `MatGetRowIJ()`, `MatRestoreColumnIJ()`
8393: @*/
8394: PetscErrorCode MatRestoreRowIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
8395: {
8396: PetscFunctionBegin;
8399: if (ia) PetscAssertPointer(ia, 6);
8400: if (ja) PetscAssertPointer(ja, 7);
8401: if (done) PetscAssertPointer(done, 8);
8402: MatCheckPreallocated(mat, 1);
8404: if (!mat->ops->restorerowij && done) *done = PETSC_FALSE;
8405: else {
8406: if (done) *done = PETSC_TRUE;
8407: PetscUseTypeMethod(mat, restorerowij, shift, symmetric, inodecompressed, n, ia, ja, done);
8408: if (n) *n = 0;
8409: if (ia) *ia = NULL;
8410: if (ja) *ja = NULL;
8411: }
8412: PetscFunctionReturn(PETSC_SUCCESS);
8413: }
8415: /*@C
8416: MatRestoreColumnIJ - Call after you are completed with the ia,ja indices obtained with `MatGetColumnIJ()`.
8418: Collective
8420: Input Parameters:
8421: + mat - the matrix
8422: . shift - 1 or zero indicating we want the indices starting at 0 or 1
8423: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
8424: - inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8425: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8426: always used.
8428: Output Parameters:
8429: + n - size of (possibly compressed) matrix
8430: . ia - the column pointers
8431: . ja - the row indices
8432: - done - `PETSC_TRUE` or `PETSC_FALSE` indicated that the values have been returned
8434: Level: developer
8436: .seealso: [](ch_matrices), `Mat`, `MatGetColumnIJ()`, `MatRestoreRowIJ()`
8437: @*/
8438: PetscErrorCode MatRestoreColumnIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
8439: {
8440: PetscFunctionBegin;
8443: if (ia) PetscAssertPointer(ia, 6);
8444: if (ja) PetscAssertPointer(ja, 7);
8445: PetscAssertPointer(done, 8);
8446: MatCheckPreallocated(mat, 1);
8448: if (!mat->ops->restorecolumnij) *done = PETSC_FALSE;
8449: else {
8450: *done = PETSC_TRUE;
8451: PetscUseTypeMethod(mat, restorecolumnij, shift, symmetric, inodecompressed, n, ia, ja, done);
8452: if (n) *n = 0;
8453: if (ia) *ia = NULL;
8454: if (ja) *ja = NULL;
8455: }
8456: PetscFunctionReturn(PETSC_SUCCESS);
8457: }
8459: /*@
8460: MatColoringPatch - Used inside matrix coloring routines that use `MatGetRowIJ()` and/or
8461: `MatGetColumnIJ()`.
8463: Collective
8465: Input Parameters:
8466: + mat - the matrix
8467: . ncolors - maximum color value
8468: . n - number of entries in colorarray
8469: - colorarray - array indicating color for each column
8471: Output Parameter:
8472: . iscoloring - coloring generated using colorarray information
8474: Level: developer
8476: .seealso: [](ch_matrices), `Mat`, `MatGetRowIJ()`, `MatGetColumnIJ()`
8477: @*/
8478: PetscErrorCode MatColoringPatch(Mat mat, PetscInt ncolors, PetscInt n, ISColoringValue colorarray[], ISColoring *iscoloring)
8479: {
8480: PetscFunctionBegin;
8483: PetscAssertPointer(colorarray, 4);
8484: PetscAssertPointer(iscoloring, 5);
8485: MatCheckPreallocated(mat, 1);
8487: if (!mat->ops->coloringpatch) {
8488: PetscCall(ISColoringCreate(PetscObjectComm((PetscObject)mat), ncolors, n, colorarray, PETSC_OWN_POINTER, iscoloring));
8489: } else {
8490: PetscUseTypeMethod(mat, coloringpatch, ncolors, n, colorarray, iscoloring);
8491: }
8492: PetscFunctionReturn(PETSC_SUCCESS);
8493: }
8495: /*@
8496: MatSetUnfactored - Resets a factored matrix to be treated as unfactored.
8498: Logically Collective
8500: Input Parameter:
8501: . mat - the factored matrix to be reset
8503: Level: developer
8505: Notes:
8506: This routine should be used only with factored matrices formed by in-place
8507: factorization via ILU(0) (or by in-place LU factorization for the `MATSEQDENSE`
8508: format). This option can save memory, for example, when solving nonlinear
8509: systems with a matrix-free Newton-Krylov method and a matrix-based, in-place
8510: ILU(0) preconditioner.
8512: One can specify in-place ILU(0) factorization by calling
8513: .vb
8514: PCType(pc,PCILU);
8515: PCFactorSeUseInPlace(pc);
8516: .ve
8517: or by using the options -pc_type ilu -pc_factor_in_place
8519: In-place factorization ILU(0) can also be used as a local
8520: solver for the blocks within the block Jacobi or additive Schwarz
8521: methods (runtime option: -sub_pc_factor_in_place). See Users-Manual: ch_pc
8522: for details on setting local solver options.
8524: Most users should employ the `KSP` interface for linear solvers
8525: instead of working directly with matrix algebra routines such as this.
8526: See, e.g., `KSPCreate()`.
8528: .seealso: [](ch_matrices), `Mat`, `PCFactorSetUseInPlace()`, `PCFactorGetUseInPlace()`
8529: @*/
8530: PetscErrorCode MatSetUnfactored(Mat mat)
8531: {
8532: PetscFunctionBegin;
8535: MatCheckPreallocated(mat, 1);
8536: mat->factortype = MAT_FACTOR_NONE;
8537: if (!mat->ops->setunfactored) PetscFunctionReturn(PETSC_SUCCESS);
8538: PetscUseTypeMethod(mat, setunfactored);
8539: PetscFunctionReturn(PETSC_SUCCESS);
8540: }
8542: /*@
8543: MatCreateSubMatrix - Gets a single submatrix on the same number of processors
8544: as the original matrix.
8546: Collective
8548: Input Parameters:
8549: + mat - the original matrix
8550: . isrow - parallel `IS` containing the rows this processor should obtain
8551: . iscol - parallel `IS` containing all columns you wish to keep. Each process should list the columns that will be in IT's "diagonal part" in the new matrix.
8552: - cll - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
8554: Output Parameter:
8555: . newmat - the new submatrix, of the same type as the original matrix
8557: Level: advanced
8559: Notes:
8560: The submatrix will be able to be multiplied with vectors using the same layout as `iscol`.
8562: Some matrix types place restrictions on the row and column indices, such
8563: as that they be sorted or that they be equal to each other. For `MATBAIJ` and `MATSBAIJ` matrices the indices must include all rows/columns of a block;
8564: for example, if the block size is 3 one cannot select the 0 and 2 rows without selecting the 1 row.
8566: The index sets may not have duplicate entries.
8568: The first time this is called you should use a cll of `MAT_INITIAL_MATRIX`,
8569: the `MatCreateSubMatrix()` routine will create the newmat for you. Any additional calls
8570: to this routine with a mat of the same nonzero structure and with a call of `MAT_REUSE_MATRIX`
8571: will reuse the matrix generated the first time. You should call `MatDestroy()` on `newmat` when
8572: you are finished using it.
8574: The communicator of the newly obtained matrix is ALWAYS the same as the communicator of
8575: the input matrix.
8577: If `iscol` is `NULL` then all columns are obtained (not supported in Fortran).
8579: If `isrow` and `iscol` have a nontrivial block-size, then the resulting matrix has this block-size as well. This feature
8580: is used by `PCFIELDSPLIT` to allow easy nesting of its use.
8582: Example usage:
8583: Consider the following 8x8 matrix with 34 non-zero values, that is
8584: assembled across 3 processors. Let's assume that proc0 owns 3 rows,
8585: proc1 owns 3 rows, proc2 owns 2 rows. This division can be shown
8586: as follows
8587: .vb
8588: 1 2 0 | 0 3 0 | 0 4
8589: Proc0 0 5 6 | 7 0 0 | 8 0
8590: 9 0 10 | 11 0 0 | 12 0
8591: -------------------------------------
8592: 13 0 14 | 15 16 17 | 0 0
8593: Proc1 0 18 0 | 19 20 21 | 0 0
8594: 0 0 0 | 22 23 0 | 24 0
8595: -------------------------------------
8596: Proc2 25 26 27 | 0 0 28 | 29 0
8597: 30 0 0 | 31 32 33 | 0 34
8598: .ve
8600: Suppose `isrow` = [0 1 | 4 | 6 7] and `iscol` = [1 2 | 3 4 5 | 6]. The resulting submatrix is
8602: .vb
8603: 2 0 | 0 3 0 | 0
8604: Proc0 5 6 | 7 0 0 | 8
8605: -------------------------------
8606: Proc1 18 0 | 19 20 21 | 0
8607: -------------------------------
8608: Proc2 26 27 | 0 0 28 | 29
8609: 0 0 | 31 32 33 | 0
8610: .ve
8612: .seealso: [](ch_matrices), `Mat`, `MatCreateSubMatrices()`, `MatCreateSubMatricesMPI()`, `MatCreateSubMatrixVirtual()`, `MatSubMatrixVirtualUpdate()`
8613: @*/
8614: PetscErrorCode MatCreateSubMatrix(Mat mat, IS isrow, IS iscol, MatReuse cll, Mat *newmat)
8615: {
8616: PetscMPIInt size;
8617: Mat *local;
8618: IS iscoltmp;
8619: PetscBool flg;
8621: PetscFunctionBegin;
8625: PetscAssertPointer(newmat, 5);
8628: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
8629: PetscCheck(cll != MAT_IGNORE_MATRIX, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Cannot use MAT_IGNORE_MATRIX");
8630: PetscCheck(cll != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Cannot use MAT_INPLACE_MATRIX");
8632: MatCheckPreallocated(mat, 1);
8633: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
8635: if (!iscol || isrow == iscol) {
8636: PetscBool stride;
8637: PetscMPIInt grabentirematrix = 0, grab;
8638: PetscCall(PetscObjectTypeCompare((PetscObject)isrow, ISSTRIDE, &stride));
8639: if (stride) {
8640: PetscInt first, step, n, rstart, rend;
8641: PetscCall(ISStrideGetInfo(isrow, &first, &step));
8642: if (step == 1) {
8643: PetscCall(MatGetOwnershipRange(mat, &rstart, &rend));
8644: if (rstart == first) {
8645: PetscCall(ISGetLocalSize(isrow, &n));
8646: if (n == rend - rstart) grabentirematrix = 1;
8647: }
8648: }
8649: }
8650: PetscCallMPI(MPIU_Allreduce(&grabentirematrix, &grab, 1, MPI_INT, MPI_MIN, PetscObjectComm((PetscObject)mat)));
8651: if (grab) {
8652: PetscCall(PetscInfo(mat, "Getting entire matrix as submatrix\n"));
8653: if (cll == MAT_INITIAL_MATRIX) {
8654: *newmat = mat;
8655: PetscCall(PetscObjectReference((PetscObject)mat));
8656: }
8657: PetscFunctionReturn(PETSC_SUCCESS);
8658: }
8659: }
8661: if (!iscol) {
8662: PetscCall(ISCreateStride(PetscObjectComm((PetscObject)mat), mat->cmap->n, mat->cmap->rstart, 1, &iscoltmp));
8663: } else {
8664: iscoltmp = iscol;
8665: }
8667: /* if original matrix is on just one processor then use submatrix generated */
8668: if (mat->ops->createsubmatrices && !mat->ops->createsubmatrix && size == 1 && cll == MAT_REUSE_MATRIX) {
8669: PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscoltmp, MAT_REUSE_MATRIX, &newmat));
8670: goto setproperties;
8671: } else if (mat->ops->createsubmatrices && !mat->ops->createsubmatrix && size == 1) {
8672: PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscoltmp, MAT_INITIAL_MATRIX, &local));
8673: *newmat = *local;
8674: PetscCall(PetscFree(local));
8675: goto setproperties;
8676: } else if (!mat->ops->createsubmatrix) {
8677: /* Create a new matrix type that implements the operation using the full matrix */
8678: PetscCall(PetscLogEventBegin(MAT_CreateSubMat, mat, 0, 0, 0));
8679: switch (cll) {
8680: case MAT_INITIAL_MATRIX:
8681: PetscCall(MatCreateSubMatrixVirtual(mat, isrow, iscoltmp, newmat));
8682: break;
8683: case MAT_REUSE_MATRIX:
8684: PetscCall(MatSubMatrixVirtualUpdate(*newmat, mat, isrow, iscoltmp));
8685: break;
8686: default:
8687: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Invalid MatReuse, must be either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX");
8688: }
8689: PetscCall(PetscLogEventEnd(MAT_CreateSubMat, mat, 0, 0, 0));
8690: goto setproperties;
8691: }
8693: PetscCall(PetscLogEventBegin(MAT_CreateSubMat, mat, 0, 0, 0));
8694: PetscUseTypeMethod(mat, createsubmatrix, isrow, iscoltmp, cll, newmat);
8695: PetscCall(PetscLogEventEnd(MAT_CreateSubMat, mat, 0, 0, 0));
8697: setproperties:
8698: if ((*newmat)->symmetric == PETSC_BOOL3_UNKNOWN && (*newmat)->structurally_symmetric == PETSC_BOOL3_UNKNOWN && (*newmat)->spd == PETSC_BOOL3_UNKNOWN && (*newmat)->hermitian == PETSC_BOOL3_UNKNOWN) {
8699: PetscCall(ISEqualUnsorted(isrow, iscoltmp, &flg));
8700: if (flg) PetscCall(MatPropagateSymmetryOptions(mat, *newmat));
8701: }
8702: if (!iscol) PetscCall(ISDestroy(&iscoltmp));
8703: if (*newmat && cll == MAT_INITIAL_MATRIX) PetscCall(PetscObjectStateIncrease((PetscObject)*newmat));
8704: if (!iscol || isrow == iscol) PetscCall(MatSelectVariableBlockSizes(*newmat, mat, isrow));
8705: PetscFunctionReturn(PETSC_SUCCESS);
8706: }
8708: /*@
8709: MatPropagateSymmetryOptions - Propagates symmetry options set on a matrix to another matrix
8711: Not Collective
8713: Input Parameters:
8714: + A - the matrix we wish to propagate options from
8715: - B - the matrix we wish to propagate options to
8717: Level: beginner
8719: Note:
8720: Propagates the options associated to `MAT_SYMMETRY_ETERNAL`, `MAT_STRUCTURALLY_SYMMETRIC`, `MAT_HERMITIAN`, `MAT_SPD`, `MAT_SYMMETRIC`, and `MAT_STRUCTURAL_SYMMETRY_ETERNAL`
8722: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatIsSymmetricKnown()`, `MatIsSPDKnown()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetricKnown()`
8723: @*/
8724: PetscErrorCode MatPropagateSymmetryOptions(Mat A, Mat B)
8725: {
8726: PetscFunctionBegin;
8729: B->symmetry_eternal = A->symmetry_eternal;
8730: B->structural_symmetry_eternal = A->structural_symmetry_eternal;
8731: B->symmetric = A->symmetric;
8732: B->structurally_symmetric = A->structurally_symmetric;
8733: B->spd = A->spd;
8734: B->hermitian = A->hermitian;
8735: PetscFunctionReturn(PETSC_SUCCESS);
8736: }
8738: /*@
8739: MatStashSetInitialSize - sets the sizes of the matrix stash, that is
8740: used during the assembly process to store values that belong to
8741: other processors.
8743: Not Collective
8745: Input Parameters:
8746: + mat - the matrix
8747: . size - the initial size of the stash.
8748: - bsize - the initial size of the block-stash(if used).
8750: Options Database Keys:
8751: + -matstash_initial_size size or size0,size1,...,sizep-1 - set initial size
8752: - -matstash_block_initial_size bsize or bsize0,bsize1,...,bsizep-1 - set initial block size
8754: Level: intermediate
8756: Notes:
8757: The block-stash is used for values set with `MatSetValuesBlocked()` while
8758: the stash is used for values set with `MatSetValues()`
8760: Run with the option -info and look for output of the form
8761: MatAssemblyBegin_MPIXXX:Stash has MM entries, uses nn mallocs.
8762: to determine the appropriate value, MM, to use for size and
8763: MatAssemblyBegin_MPIXXX:Block-Stash has BMM entries, uses nn mallocs.
8764: to determine the value, BMM to use for bsize
8766: .seealso: [](ch_matrices), `MatAssemblyBegin()`, `MatAssemblyEnd()`, `Mat`, `MatStashGetInfo()`
8767: @*/
8768: PetscErrorCode MatStashSetInitialSize(Mat mat, PetscInt size, PetscInt bsize)
8769: {
8770: PetscFunctionBegin;
8773: PetscCall(MatStashSetInitialSize_Private(&mat->stash, size));
8774: PetscCall(MatStashSetInitialSize_Private(&mat->bstash, bsize));
8775: PetscFunctionReturn(PETSC_SUCCESS);
8776: }
8778: /*@
8779: MatInterpolateAdd - $w = y + A*x$ or $A^T*x$ depending on the shape of
8780: the matrix
8782: Neighbor-wise Collective
8784: Input Parameters:
8785: + A - the matrix
8786: . x - the vector to be multiplied by the interpolation operator
8787: - y - the vector to be added to the result
8789: Output Parameter:
8790: . w - the resulting vector
8792: Level: intermediate
8794: Notes:
8795: `w` may be the same vector as `y`.
8797: This allows one to use either the restriction or interpolation (its transpose)
8798: matrix to do the interpolation
8800: .seealso: [](ch_matrices), `Mat`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatRestrict()`, `PCMG`
8801: @*/
8802: PetscErrorCode MatInterpolateAdd(Mat A, Vec x, Vec y, Vec w)
8803: {
8804: PetscInt M, N, Ny;
8806: PetscFunctionBegin;
8811: PetscCall(MatGetSize(A, &M, &N));
8812: PetscCall(VecGetSize(y, &Ny));
8813: if (M == Ny) PetscCall(MatMultAdd(A, x, y, w));
8814: else PetscCall(MatMultTransposeAdd(A, x, y, w));
8815: PetscFunctionReturn(PETSC_SUCCESS);
8816: }
8818: /*@
8819: MatInterpolate - $y = A*x$ or $A^T*x$ depending on the shape of
8820: the matrix
8822: Neighbor-wise Collective
8824: Input Parameters:
8825: + A - the matrix
8826: - x - the vector to be interpolated
8828: Output Parameter:
8829: . y - the resulting vector
8831: Level: intermediate
8833: Note:
8834: This allows one to use either the restriction or interpolation (its transpose)
8835: matrix to do the interpolation
8837: .seealso: [](ch_matrices), `Mat`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatRestrict()`, `PCMG`
8838: @*/
8839: PetscErrorCode MatInterpolate(Mat A, Vec x, Vec y)
8840: {
8841: PetscInt M, N, Ny;
8843: PetscFunctionBegin;
8847: PetscCall(MatGetSize(A, &M, &N));
8848: PetscCall(VecGetSize(y, &Ny));
8849: if (M == Ny) PetscCall(MatMult(A, x, y));
8850: else PetscCall(MatMultTranspose(A, x, y));
8851: PetscFunctionReturn(PETSC_SUCCESS);
8852: }
8854: /*@
8855: MatRestrict - $y = A*x$ or $A^T*x$
8857: Neighbor-wise Collective
8859: Input Parameters:
8860: + A - the matrix
8861: - x - the vector to be restricted
8863: Output Parameter:
8864: . y - the resulting vector
8866: Level: intermediate
8868: Note:
8869: This allows one to use either the restriction or interpolation (its transpose)
8870: matrix to do the restriction
8872: .seealso: [](ch_matrices), `Mat`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatInterpolate()`, `PCMG`
8873: @*/
8874: PetscErrorCode MatRestrict(Mat A, Vec x, Vec y)
8875: {
8876: PetscInt M, N, Nx;
8878: PetscFunctionBegin;
8882: PetscCall(MatGetSize(A, &M, &N));
8883: PetscCall(VecGetSize(x, &Nx));
8884: if (M == Nx) PetscCall(MatMultTranspose(A, x, y));
8885: else PetscCall(MatMult(A, x, y));
8886: PetscFunctionReturn(PETSC_SUCCESS);
8887: }
8889: /*@
8890: MatMatInterpolateAdd - $Y = W + A*X$ or $W + A^T*X$ depending on the shape of `A`
8892: Neighbor-wise Collective
8894: Input Parameters:
8895: + A - the matrix
8896: . x - the input dense matrix to be multiplied
8897: - w - the input dense matrix to be added to the result
8899: Output Parameter:
8900: . y - the output dense matrix
8902: Level: intermediate
8904: Note:
8905: This allows one to use either the restriction or interpolation (its transpose)
8906: matrix to do the interpolation. `y` matrix can be reused if already created with the proper sizes,
8907: otherwise it will be recreated. `y` must be initialized to `NULL` if not supplied.
8909: .seealso: [](ch_matrices), `Mat`, `MatInterpolateAdd()`, `MatMatInterpolate()`, `MatMatRestrict()`, `PCMG`
8910: @*/
8911: PetscErrorCode MatMatInterpolateAdd(Mat A, Mat x, Mat w, Mat *y)
8912: {
8913: PetscInt M, N, Mx, Nx, Mo, My = 0, Ny = 0;
8914: PetscBool trans = PETSC_TRUE;
8915: MatReuse reuse = MAT_INITIAL_MATRIX;
8917: PetscFunctionBegin;
8923: PetscCall(MatGetSize(A, &M, &N));
8924: PetscCall(MatGetSize(x, &Mx, &Nx));
8925: if (N == Mx) trans = PETSC_FALSE;
8926: else PetscCheck(M == Mx, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Size mismatch: A %" PetscInt_FMT "x%" PetscInt_FMT ", X %" PetscInt_FMT "x%" PetscInt_FMT, M, N, Mx, Nx);
8927: Mo = trans ? N : M;
8928: if (*y) {
8929: PetscCall(MatGetSize(*y, &My, &Ny));
8930: if (Mo == My && Nx == Ny) reuse = MAT_REUSE_MATRIX;
8931: else {
8932: PetscCheck(w || *y != w, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Cannot reuse y and w, size mismatch: A %" PetscInt_FMT "x%" PetscInt_FMT ", X %" PetscInt_FMT "x%" PetscInt_FMT ", Y %" PetscInt_FMT "x%" PetscInt_FMT, M, N, Mx, Nx, My, Ny);
8933: PetscCall(MatDestroy(y));
8934: }
8935: }
8937: if (w && *y == w) { /* this is to minimize changes in PCMG */
8938: PetscBool flg;
8940: PetscCall(PetscObjectQuery((PetscObject)*y, "__MatMatIntAdd_w", (PetscObject *)&w));
8941: if (w) {
8942: PetscInt My, Ny, Mw, Nw;
8944: PetscCall(PetscObjectTypeCompare((PetscObject)*y, ((PetscObject)w)->type_name, &flg));
8945: PetscCall(MatGetSize(*y, &My, &Ny));
8946: PetscCall(MatGetSize(w, &Mw, &Nw));
8947: if (!flg || My != Mw || Ny != Nw) w = NULL;
8948: }
8949: if (!w) {
8950: PetscCall(MatDuplicate(*y, MAT_COPY_VALUES, &w));
8951: PetscCall(PetscObjectCompose((PetscObject)*y, "__MatMatIntAdd_w", (PetscObject)w));
8952: PetscCall(PetscObjectDereference((PetscObject)w));
8953: } else PetscCall(MatCopy(*y, w, UNKNOWN_NONZERO_PATTERN));
8954: }
8955: if (!trans) PetscCall(MatMatMult(A, x, reuse, PETSC_DETERMINE, y));
8956: else PetscCall(MatTransposeMatMult(A, x, reuse, PETSC_DETERMINE, y));
8957: if (w) PetscCall(MatAXPY(*y, 1.0, w, UNKNOWN_NONZERO_PATTERN));
8958: PetscFunctionReturn(PETSC_SUCCESS);
8959: }
8961: /*@
8962: MatMatInterpolate - $Y = A*X$ or $A^T*X$ depending on the shape of `A`
8964: Neighbor-wise Collective
8966: Input Parameters:
8967: + A - the matrix
8968: - x - the input dense matrix
8970: Output Parameter:
8971: . y - the output dense matrix
8973: Level: intermediate
8975: Note:
8976: This allows one to use either the restriction or interpolation (its transpose)
8977: matrix to do the interpolation. `y` matrix can be reused if already created with the proper sizes,
8978: otherwise it will be recreated. `y` must be initialized to `NULL` if not supplied.
8980: .seealso: [](ch_matrices), `Mat`, `MatInterpolate()`, `MatRestrict()`, `MatMatRestrict()`, `PCMG`
8981: @*/
8982: PetscErrorCode MatMatInterpolate(Mat A, Mat x, Mat *y)
8983: {
8984: PetscFunctionBegin;
8985: PetscCall(MatMatInterpolateAdd(A, x, NULL, y));
8986: PetscFunctionReturn(PETSC_SUCCESS);
8987: }
8989: /*@
8990: MatMatRestrict - $Y = A*X$ or $A^T*X$ depending on the shape of `A`
8992: Neighbor-wise Collective
8994: Input Parameters:
8995: + A - the matrix
8996: - x - the input dense matrix
8998: Output Parameter:
8999: . y - the output dense matrix
9001: Level: intermediate
9003: Note:
9004: This allows one to use either the restriction or interpolation (its transpose)
9005: matrix to do the restriction. `y` matrix can be reused if already created with the proper sizes,
9006: otherwise it will be recreated. `y` must be initialized to `NULL` if not supplied.
9008: .seealso: [](ch_matrices), `Mat`, `MatRestrict()`, `MatInterpolate()`, `MatMatInterpolate()`, `PCMG`
9009: @*/
9010: PetscErrorCode MatMatRestrict(Mat A, Mat x, Mat *y)
9011: {
9012: PetscFunctionBegin;
9013: PetscCall(MatMatInterpolateAdd(A, x, NULL, y));
9014: PetscFunctionReturn(PETSC_SUCCESS);
9015: }
9017: /*@
9018: MatGetNullSpace - retrieves the null space of a matrix.
9020: Logically Collective
9022: Input Parameters:
9023: + mat - the matrix
9024: - nullsp - the null space object
9026: Level: developer
9028: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatSetNullSpace()`, `MatNullSpace`
9029: @*/
9030: PetscErrorCode MatGetNullSpace(Mat mat, MatNullSpace *nullsp)
9031: {
9032: PetscFunctionBegin;
9034: PetscAssertPointer(nullsp, 2);
9035: *nullsp = (mat->symmetric == PETSC_BOOL3_TRUE && !mat->nullsp) ? mat->transnullsp : mat->nullsp;
9036: PetscFunctionReturn(PETSC_SUCCESS);
9037: }
9039: /*@C
9040: MatGetNullSpaces - gets the null spaces, transpose null spaces, and near null spaces from an array of matrices
9042: Logically Collective
9044: Input Parameters:
9045: + n - the number of matrices
9046: - mat - the array of matrices
9048: Output Parameters:
9049: . nullsp - an array of null spaces, `NULL` for each matrix that does not have a null space, length 3 * `n`
9051: Level: developer
9053: Note:
9054: Call `MatRestoreNullspaces()` to provide these to another array of matrices
9056: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatSetTransposeNullSpace()`, `MatGetTransposeNullSpace()`,
9057: `MatNullSpaceRemove()`, `MatRestoreNullSpaces()`
9058: @*/
9059: PetscErrorCode MatGetNullSpaces(PetscInt n, Mat mat[], MatNullSpace *nullsp[])
9060: {
9061: PetscFunctionBegin;
9062: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Number of matrices %" PetscInt_FMT " must be non-negative", n);
9063: PetscAssertPointer(mat, 2);
9064: PetscAssertPointer(nullsp, 3);
9066: PetscCall(PetscCalloc1(3 * n, nullsp));
9067: for (PetscInt i = 0; i < n; i++) {
9069: (*nullsp)[i] = mat[i]->nullsp;
9070: PetscCall(PetscObjectReference((PetscObject)(*nullsp)[i]));
9071: (*nullsp)[n + i] = mat[i]->nearnullsp;
9072: PetscCall(PetscObjectReference((PetscObject)(*nullsp)[n + i]));
9073: (*nullsp)[2 * n + i] = mat[i]->transnullsp;
9074: PetscCall(PetscObjectReference((PetscObject)(*nullsp)[2 * n + i]));
9075: }
9076: PetscFunctionReturn(PETSC_SUCCESS);
9077: }
9079: /*@C
9080: MatRestoreNullSpaces - sets the null spaces, transpose null spaces, and near null spaces obtained with `MatGetNullSpaces()` for an array of matrices
9082: Logically Collective
9084: Input Parameters:
9085: + n - the number of matrices
9086: . mat - the array of matrices
9087: - nullsp - an array of null spaces
9089: Level: developer
9091: Note:
9092: Call `MatGetNullSpaces()` to create `nullsp`
9094: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatSetTransposeNullSpace()`, `MatGetTransposeNullSpace()`,
9095: `MatNullSpaceRemove()`, `MatGetNullSpaces()`
9096: @*/
9097: PetscErrorCode MatRestoreNullSpaces(PetscInt n, Mat mat[], MatNullSpace *nullsp[])
9098: {
9099: PetscFunctionBegin;
9100: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Number of matrices %" PetscInt_FMT " must be non-negative", n);
9101: PetscAssertPointer(mat, 2);
9102: PetscAssertPointer(nullsp, 3);
9103: PetscAssertPointer(*nullsp, 3);
9105: for (PetscInt i = 0; i < n; i++) {
9107: PetscCall(MatSetNullSpace(mat[i], (*nullsp)[i]));
9108: PetscCall(PetscObjectDereference((PetscObject)(*nullsp)[i]));
9109: PetscCall(MatSetNearNullSpace(mat[i], (*nullsp)[n + i]));
9110: PetscCall(PetscObjectDereference((PetscObject)(*nullsp)[n + i]));
9111: PetscCall(MatSetTransposeNullSpace(mat[i], (*nullsp)[2 * n + i]));
9112: PetscCall(PetscObjectDereference((PetscObject)(*nullsp)[2 * n + i]));
9113: }
9114: PetscCall(PetscFree(*nullsp));
9115: PetscFunctionReturn(PETSC_SUCCESS);
9116: }
9118: /*@
9119: MatSetNullSpace - attaches a null space to a matrix.
9121: Logically Collective
9123: Input Parameters:
9124: + mat - the matrix
9125: - nullsp - the null space object
9127: Level: advanced
9129: Notes:
9130: This null space is used by the `KSP` linear solvers to solve singular systems.
9132: Overwrites any previous null space that may have been attached. You can remove the null space from the matrix object by calling this routine with an nullsp of `NULL`
9134: For inconsistent singular systems (linear systems where the right-hand side is not in the range of the operator) the `KSP` residuals will not converge
9135: to zero but the linear system will still be solved in a least squares sense.
9137: The fundamental theorem of linear algebra (Gilbert Strang, Introduction to Applied Mathematics, page 72) states that
9138: the domain of a matrix $A$ (from $R^n$ to $R^m$ ($m$ rows, $n$ columns) $R^n$ = the direct sum of the null space of $A$, $n(A)$, plus the range of $A^T$, $R(A^T)$.
9139: Similarly $R^m$ = direct sum $n(A^T) + R(A)$. Hence the linear system $A x = b$ has a solution only if $b$ in $R(A)$ (or correspondingly $b$ is orthogonal to
9140: $n(A^T))$ and if $x$ is a solution then $x + \alpha n(A)$ is a solution for any $\alpha$. The minimum norm solution is orthogonal to $n(A)$. For problems without a solution
9141: the solution that minimizes the norm of the residual (the least squares solution) can be obtained by solving $A x = \hat{b}$ where $\hat{b}$ is $b$ orthogonalized to the $n(A^T)$.
9142: This $\hat{b}$ can be obtained by calling `MatNullSpaceRemove()` with the null space of the transpose of the matrix.
9144: If the matrix is known to be symmetric because it is an `MATSBAIJ` matrix or one has called
9145: `MatSetOption`(mat,`MAT_SYMMETRIC` or possibly `MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`); this
9146: routine also automatically calls `MatSetTransposeNullSpace()`.
9148: The user should call `MatNullSpaceDestroy()`.
9150: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatSetTransposeNullSpace()`, `MatGetTransposeNullSpace()`, `MatNullSpaceRemove()`,
9151: `KSPSetPCSide()`
9152: @*/
9153: PetscErrorCode MatSetNullSpace(Mat mat, MatNullSpace nullsp)
9154: {
9155: PetscFunctionBegin;
9158: PetscCall(PetscObjectReference((PetscObject)nullsp));
9159: PetscCall(MatNullSpaceDestroy(&mat->nullsp));
9160: mat->nullsp = nullsp;
9161: if (mat->symmetric == PETSC_BOOL3_TRUE) PetscCall(MatSetTransposeNullSpace(mat, nullsp));
9162: PetscFunctionReturn(PETSC_SUCCESS);
9163: }
9165: /*@
9166: MatGetTransposeNullSpace - retrieves the null space of the transpose of a matrix.
9168: Logically Collective
9170: Input Parameters:
9171: + mat - the matrix
9172: - nullsp - the null space object
9174: Level: developer
9176: .seealso: [](ch_matrices), `Mat`, `MatNullSpace`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatSetTransposeNullSpace()`, `MatSetNullSpace()`, `MatGetNullSpace()`
9177: @*/
9178: PetscErrorCode MatGetTransposeNullSpace(Mat mat, MatNullSpace *nullsp)
9179: {
9180: PetscFunctionBegin;
9183: PetscAssertPointer(nullsp, 2);
9184: *nullsp = (mat->symmetric == PETSC_BOOL3_TRUE && !mat->transnullsp) ? mat->nullsp : mat->transnullsp;
9185: PetscFunctionReturn(PETSC_SUCCESS);
9186: }
9188: /*@
9189: MatSetTransposeNullSpace - attaches the null space of a transpose of a matrix to the matrix
9191: Logically Collective
9193: Input Parameters:
9194: + mat - the matrix
9195: - nullsp - the null space object
9197: Level: advanced
9199: Notes:
9200: This allows solving singular linear systems defined by the transpose of the matrix using `KSP` solvers with left preconditioning.
9202: See `MatSetNullSpace()`
9204: .seealso: [](ch_matrices), `Mat`, `MatNullSpace`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatSetNullSpace()`, `MatGetTransposeNullSpace()`, `MatNullSpaceRemove()`, `KSPSetPCSide()`
9205: @*/
9206: PetscErrorCode MatSetTransposeNullSpace(Mat mat, MatNullSpace nullsp)
9207: {
9208: PetscFunctionBegin;
9211: PetscCall(PetscObjectReference((PetscObject)nullsp));
9212: PetscCall(MatNullSpaceDestroy(&mat->transnullsp));
9213: mat->transnullsp = nullsp;
9214: PetscFunctionReturn(PETSC_SUCCESS);
9215: }
9217: /*@
9218: MatSetNearNullSpace - attaches a null space to a matrix, which is often the null space (rigid body modes) of the operator without boundary conditions
9219: This null space will be used to provide near null space vectors to a multigrid preconditioner built from this matrix.
9221: Logically Collective
9223: Input Parameters:
9224: + mat - the matrix
9225: - nullsp - the null space object
9227: Level: advanced
9229: Notes:
9230: Overwrites any previous near null space that may have been attached
9232: You can remove the null space by calling this routine with an `nullsp` of `NULL`
9234: .seealso: [](ch_matrices), `Mat`, `MatNullSpace`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNullSpace()`, `MatNullSpaceCreateRigidBody()`, `MatGetNearNullSpace()`
9235: @*/
9236: PetscErrorCode MatSetNearNullSpace(Mat mat, MatNullSpace nullsp)
9237: {
9238: PetscFunctionBegin;
9242: MatCheckPreallocated(mat, 1);
9243: PetscCall(PetscObjectReference((PetscObject)nullsp));
9244: PetscCall(MatNullSpaceDestroy(&mat->nearnullsp));
9245: mat->nearnullsp = nullsp;
9246: PetscFunctionReturn(PETSC_SUCCESS);
9247: }
9249: /*@
9250: MatGetNearNullSpace - Get null space attached with `MatSetNearNullSpace()`
9252: Not Collective
9254: Input Parameter:
9255: . mat - the matrix
9257: Output Parameter:
9258: . nullsp - the null space object, `NULL` if not set
9260: Level: advanced
9262: .seealso: [](ch_matrices), `Mat`, `MatNullSpace`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatNullSpaceCreate()`
9263: @*/
9264: PetscErrorCode MatGetNearNullSpace(Mat mat, MatNullSpace *nullsp)
9265: {
9266: PetscFunctionBegin;
9269: PetscAssertPointer(nullsp, 2);
9270: MatCheckPreallocated(mat, 1);
9271: *nullsp = mat->nearnullsp;
9272: PetscFunctionReturn(PETSC_SUCCESS);
9273: }
9275: /*@
9276: MatICCFactor - Performs in-place incomplete Cholesky factorization of matrix.
9278: Collective
9280: Input Parameters:
9281: + mat - the matrix
9282: . row - row/column permutation
9283: - info - information on desired factorization process
9285: Level: developer
9287: Notes:
9288: Probably really in-place only when level of fill is zero, otherwise allocates
9289: new space to store factored matrix and deletes previous memory.
9291: Most users should employ the `KSP` interface for linear solvers
9292: instead of working directly with matrix algebra routines such as this.
9293: See, e.g., `KSPCreate()`.
9295: Fortran Note:
9296: A valid (non-null) `info` argument must be provided
9298: .seealso: [](ch_matrices), `Mat`, `MatFactorInfo`, `MatGetFactor()`, `MatICCFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`
9299: @*/
9300: PetscErrorCode MatICCFactor(Mat mat, IS row, const MatFactorInfo *info)
9301: {
9302: PetscFunctionBegin;
9306: PetscAssertPointer(info, 3);
9307: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "matrix must be square");
9308: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
9309: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
9310: MatCheckPreallocated(mat, 1);
9311: PetscUseTypeMethod(mat, iccfactor, row, info);
9312: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
9313: PetscFunctionReturn(PETSC_SUCCESS);
9314: }
9316: /*@
9317: MatDiagonalScaleLocal - Scales columns of a matrix given the scaling values including the
9318: ghosted ones.
9320: Not Collective
9322: Input Parameters:
9323: + mat - the matrix
9324: - diag - the diagonal values, including ghost ones
9326: Level: developer
9328: Notes:
9329: Works only for `MATMPIAIJ` and `MATMPIBAIJ` matrices
9331: This allows one to avoid during communication to perform the scaling that must be done with `MatDiagonalScale()`
9333: .seealso: [](ch_matrices), `Mat`, `MatDiagonalScale()`
9334: @*/
9335: PetscErrorCode MatDiagonalScaleLocal(Mat mat, Vec diag)
9336: {
9337: PetscMPIInt size;
9339: PetscFunctionBegin;
9344: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Matrix must be already assembled");
9345: PetscCall(PetscLogEventBegin(MAT_Scale, mat, 0, 0, 0));
9346: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
9347: if (size == 1) {
9348: PetscInt n, m;
9349: PetscCall(VecGetSize(diag, &n));
9350: PetscCall(MatGetSize(mat, NULL, &m));
9351: PetscCheck(m == n, PETSC_COMM_SELF, PETSC_ERR_SUP, "Only supported for sequential matrices when no ghost points/periodic conditions");
9352: PetscCall(MatDiagonalScale(mat, NULL, diag));
9353: } else PetscUseMethod(mat, "MatDiagonalScaleLocal_C", (Mat, Vec), (mat, diag));
9354: PetscCall(PetscLogEventEnd(MAT_Scale, mat, 0, 0, 0));
9355: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
9356: PetscFunctionReturn(PETSC_SUCCESS);
9357: }
9359: /*@
9360: MatGetInertia - Gets the inertia from a factored matrix
9362: Collective
9364: Input Parameter:
9365: . mat - the matrix
9367: Output Parameters:
9368: + nneg - number of negative eigenvalues
9369: . nzero - number of zero eigenvalues
9370: - npos - number of positive eigenvalues
9372: Level: advanced
9374: Note:
9375: Matrix must have been factored by `MatCholeskyFactor()`
9377: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatCholeskyFactor()`
9378: @*/
9379: PetscErrorCode MatGetInertia(Mat mat, PetscInt *nneg, PetscInt *nzero, PetscInt *npos)
9380: {
9381: PetscFunctionBegin;
9384: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
9385: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Numeric factor mat is not assembled");
9386: PetscUseTypeMethod(mat, getinertia, nneg, nzero, npos);
9387: PetscFunctionReturn(PETSC_SUCCESS);
9388: }
9390: /*@C
9391: MatSolves - Solves $A x = b$, given a factored matrix, for a collection of vectors
9393: Neighbor-wise Collective
9395: Input Parameters:
9396: + mat - the factored matrix obtained with `MatGetFactor()`
9397: - b - the right-hand-side vectors
9399: Output Parameter:
9400: . x - the result vectors
9402: Level: developer
9404: Note:
9405: The vectors `b` and `x` cannot be the same. I.e., one cannot
9406: call `MatSolves`(A,x,x).
9408: .seealso: [](ch_matrices), `Mat`, `Vecs`, `MatSolveAdd()`, `MatSolveTranspose()`, `MatSolveTransposeAdd()`, `MatSolve()`
9409: @*/
9410: PetscErrorCode MatSolves(Mat mat, Vecs b, Vecs x)
9411: {
9412: PetscFunctionBegin;
9415: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
9416: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
9417: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
9419: MatCheckPreallocated(mat, 1);
9420: PetscCall(PetscLogEventBegin(MAT_Solves, mat, 0, 0, 0));
9421: PetscUseTypeMethod(mat, solves, b, x);
9422: PetscCall(PetscLogEventEnd(MAT_Solves, mat, 0, 0, 0));
9423: PetscFunctionReturn(PETSC_SUCCESS);
9424: }
9426: /*@
9427: MatIsSymmetric - Test whether a matrix is symmetric
9429: Collective
9431: Input Parameters:
9432: + A - the matrix to test
9433: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact transpose)
9435: Output Parameter:
9436: . flg - the result
9438: Level: intermediate
9440: Notes:
9441: For real numbers `MatIsSymmetric()` and `MatIsHermitian()` return identical results
9443: If the matrix does not yet know if it is symmetric or not this can be an expensive operation, also available `MatIsSymmetricKnown()`
9445: One can declare that a matrix is symmetric with `MatSetOption`(mat,`MAT_SYMMETRIC`,`PETSC_TRUE`) and if it is known to remain symmetric
9446: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9448: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetricKnown()`,
9449: `MAT_SYMMETRIC`, `MAT_SYMMETRY_ETERNAL`
9450: @*/
9451: PetscErrorCode MatIsSymmetric(Mat A, PetscReal tol, PetscBool *flg)
9452: {
9453: PetscFunctionBegin;
9455: PetscAssertPointer(flg, 3);
9456: if (A->symmetric != PETSC_BOOL3_UNKNOWN && !tol) *flg = PetscBool3ToBool(A->symmetric);
9457: else {
9458: if (A->ops->issymmetric) PetscUseTypeMethod(A, issymmetric, tol, flg);
9459: else PetscCall(MatIsTranspose(A, A, tol, flg));
9460: if (!tol) PetscCall(MatSetOption(A, MAT_SYMMETRIC, *flg));
9461: }
9462: PetscFunctionReturn(PETSC_SUCCESS);
9463: }
9465: /*@
9466: MatIsHermitian - Test whether a matrix is Hermitian
9468: Collective
9470: Input Parameters:
9471: + A - the matrix to test
9472: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact Hermitian)
9474: Output Parameter:
9475: . flg - the result
9477: Level: intermediate
9479: Notes:
9480: For real numbers `MatIsSymmetric()` and `MatIsHermitian()` return identical results
9482: If the matrix does not yet know if it is Hermitian or not this can be an expensive operation, also available `MatIsHermitianKnown()`
9484: One can declare that a matrix is Hermitian with `MatSetOption`(mat,`MAT_HERMITIAN`,`PETSC_TRUE`) and if it is known to remain Hermitian
9485: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMEMTRY_ETERNAL`,`PETSC_TRUE`)
9487: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetric()`, `MatSetOption()`,
9488: `MatIsSymmetricKnown()`, `MatIsSymmetric()`, `MAT_HERMITIAN`, `MAT_SYMMETRY_ETERNAL`
9489: @*/
9490: PetscErrorCode MatIsHermitian(Mat A, PetscReal tol, PetscBool *flg)
9491: {
9492: PetscFunctionBegin;
9494: PetscAssertPointer(flg, 3);
9495: if (A->hermitian != PETSC_BOOL3_UNKNOWN && !tol) *flg = PetscBool3ToBool(A->hermitian);
9496: else {
9497: if (A->ops->ishermitian) PetscUseTypeMethod(A, ishermitian, tol, flg);
9498: else PetscCall(MatIsHermitianTranspose(A, A, tol, flg));
9499: if (!tol) PetscCall(MatSetOption(A, MAT_HERMITIAN, *flg));
9500: }
9501: PetscFunctionReturn(PETSC_SUCCESS);
9502: }
9504: /*@
9505: MatIsSymmetricKnown - Checks if a matrix knows if it is symmetric or not and its symmetric state
9507: Not Collective
9509: Input Parameter:
9510: . A - the matrix to check
9512: Output Parameters:
9513: + set - `PETSC_TRUE` if the matrix knows its symmetry state (this tells you if the next flag is valid)
9514: - flg - the result (only valid if set is `PETSC_TRUE`)
9516: Level: advanced
9518: Notes:
9519: Does not check the matrix values directly, so this may return unknown (set = `PETSC_FALSE`). Use `MatIsSymmetric()`
9520: if you want it explicitly checked
9522: One can declare that a matrix is symmetric with `MatSetOption`(mat,`MAT_SYMMETRIC`,`PETSC_TRUE`) and if it is known to remain symmetric
9523: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9525: .seealso: [](ch_matrices), `Mat`, `MAT_SYMMETRY_ETERNAL`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitianKnown()`
9526: @*/
9527: PetscErrorCode MatIsSymmetricKnown(Mat A, PetscBool *set, PetscBool *flg)
9528: {
9529: PetscFunctionBegin;
9531: PetscAssertPointer(set, 2);
9532: PetscAssertPointer(flg, 3);
9533: if (A->symmetric != PETSC_BOOL3_UNKNOWN) {
9534: *set = PETSC_TRUE;
9535: *flg = PetscBool3ToBool(A->symmetric);
9536: } else *set = PETSC_FALSE;
9537: PetscFunctionReturn(PETSC_SUCCESS);
9538: }
9540: /*@
9541: MatIsSPDKnown - Checks if a matrix knows if it is symmetric positive definite or not and its symmetric positive definite state
9543: Not Collective
9545: Input Parameter:
9546: . A - the matrix to check
9548: Output Parameters:
9549: + set - `PETSC_TRUE` if the matrix knows its symmetric positive definite state (this tells you if the next flag is valid)
9550: - flg - the result (only valid if set is `PETSC_TRUE`)
9552: Level: advanced
9554: Notes:
9555: Does not check the matrix values directly, so this may return unknown (set = `PETSC_FALSE`).
9557: One can declare that a matrix is SPD with `MatSetOption`(mat,`MAT_SPD`,`PETSC_TRUE`) and if it is known to remain SPD
9558: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SPD_ETERNAL`,`PETSC_TRUE`)
9560: .seealso: [](ch_matrices), `Mat`, `MAT_SPD_ETERNAL`, `MAT_SPD`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitianKnown()`
9561: @*/
9562: PetscErrorCode MatIsSPDKnown(Mat A, PetscBool *set, PetscBool *flg)
9563: {
9564: PetscFunctionBegin;
9566: PetscAssertPointer(set, 2);
9567: PetscAssertPointer(flg, 3);
9568: if (A->spd != PETSC_BOOL3_UNKNOWN) {
9569: *set = PETSC_TRUE;
9570: *flg = PetscBool3ToBool(A->spd);
9571: } else *set = PETSC_FALSE;
9572: PetscFunctionReturn(PETSC_SUCCESS);
9573: }
9575: /*@
9576: MatIsHermitianKnown - Checks if a matrix knows if it is Hermitian or not and its Hermitian state
9578: Not Collective
9580: Input Parameter:
9581: . A - the matrix to check
9583: Output Parameters:
9584: + set - `PETSC_TRUE` if the matrix knows its Hermitian state (this tells you if the next flag is valid)
9585: - flg - the result (only valid if set is `PETSC_TRUE`)
9587: Level: advanced
9589: Notes:
9590: Does not check the matrix values directly, so this may return unknown (set = `PETSC_FALSE`). Use `MatIsHermitian()`
9591: if you want it explicitly checked
9593: One can declare that a matrix is Hermitian with `MatSetOption`(mat,`MAT_HERMITIAN`,`PETSC_TRUE`) and if it is known to remain Hermitian
9594: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9596: .seealso: [](ch_matrices), `Mat`, `MAT_SYMMETRY_ETERNAL`, `MAT_HERMITIAN`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`
9597: @*/
9598: PetscErrorCode MatIsHermitianKnown(Mat A, PetscBool *set, PetscBool *flg)
9599: {
9600: PetscFunctionBegin;
9602: PetscAssertPointer(set, 2);
9603: PetscAssertPointer(flg, 3);
9604: if (A->hermitian != PETSC_BOOL3_UNKNOWN) {
9605: *set = PETSC_TRUE;
9606: *flg = PetscBool3ToBool(A->hermitian);
9607: } else *set = PETSC_FALSE;
9608: PetscFunctionReturn(PETSC_SUCCESS);
9609: }
9611: /*@
9612: MatIsStructurallySymmetric - Test whether a matrix is structurally symmetric
9614: Collective
9616: Input Parameter:
9617: . A - the matrix to test
9619: Output Parameter:
9620: . flg - the result
9622: Level: intermediate
9624: Notes:
9625: If the matrix does yet know it is structurally symmetric this can be an expensive operation, also available `MatIsStructurallySymmetricKnown()`
9627: One can declare that a matrix is structurally symmetric with `MatSetOption`(mat,`MAT_STRUCTURALLY_SYMMETRIC`,`PETSC_TRUE`) and if it is known to remain structurally
9628: symmetric after changes to the matrices values one can call `MatSetOption`(mat,`MAT_STRUCTURAL_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9630: .seealso: [](ch_matrices), `Mat`, `MAT_STRUCTURALLY_SYMMETRIC`, `MAT_STRUCTURAL_SYMMETRY_ETERNAL`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsSymmetric()`, `MatSetOption()`, `MatIsStructurallySymmetricKnown()`
9631: @*/
9632: PetscErrorCode MatIsStructurallySymmetric(Mat A, PetscBool *flg)
9633: {
9634: PetscFunctionBegin;
9636: PetscAssertPointer(flg, 2);
9637: if (A->structurally_symmetric != PETSC_BOOL3_UNKNOWN) *flg = PetscBool3ToBool(A->structurally_symmetric);
9638: else {
9639: PetscUseTypeMethod(A, isstructurallysymmetric, flg);
9640: PetscCall(MatSetOption(A, MAT_STRUCTURALLY_SYMMETRIC, *flg));
9641: }
9642: PetscFunctionReturn(PETSC_SUCCESS);
9643: }
9645: /*@
9646: MatIsStructurallySymmetricKnown - Checks if a matrix knows if it is structurally symmetric or not and its structurally symmetric state
9648: Not Collective
9650: Input Parameter:
9651: . A - the matrix to check
9653: Output Parameters:
9654: + set - PETSC_TRUE if the matrix knows its structurally symmetric state (this tells you if the next flag is valid)
9655: - flg - the result (only valid if set is PETSC_TRUE)
9657: Level: advanced
9659: Notes:
9660: One can declare that a matrix is structurally symmetric with `MatSetOption`(mat,`MAT_STRUCTURALLY_SYMMETRIC`,`PETSC_TRUE`) and if it is known to remain structurally
9661: symmetric after changes to the matrices values one can call `MatSetOption`(mat,`MAT_STRUCTURAL_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9663: Use `MatIsStructurallySymmetric()` to explicitly check if a matrix is structurally symmetric (this is an expensive operation)
9665: .seealso: [](ch_matrices), `Mat`, `MAT_STRUCTURALLY_SYMMETRIC`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitianKnown()`
9666: @*/
9667: PetscErrorCode MatIsStructurallySymmetricKnown(Mat A, PetscBool *set, PetscBool *flg)
9668: {
9669: PetscFunctionBegin;
9671: PetscAssertPointer(set, 2);
9672: PetscAssertPointer(flg, 3);
9673: if (A->structurally_symmetric != PETSC_BOOL3_UNKNOWN) {
9674: *set = PETSC_TRUE;
9675: *flg = PetscBool3ToBool(A->structurally_symmetric);
9676: } else *set = PETSC_FALSE;
9677: PetscFunctionReturn(PETSC_SUCCESS);
9678: }
9680: /*@
9681: MatStashGetInfo - Gets how many values are currently in the matrix stash, i.e. need
9682: to be communicated to other processors during the `MatAssemblyBegin()`/`MatAssemblyEnd()` process
9684: Not Collective
9686: Input Parameter:
9687: . mat - the matrix
9689: Output Parameters:
9690: + nstash - the size of the stash
9691: . reallocs - the number of additional mallocs incurred.
9692: . bnstash - the size of the block stash
9693: - breallocs - the number of additional mallocs incurred.in the block stash
9695: Level: advanced
9697: .seealso: [](ch_matrices), `MatAssemblyBegin()`, `MatAssemblyEnd()`, `Mat`, `MatStashSetInitialSize()`
9698: @*/
9699: PetscErrorCode MatStashGetInfo(Mat mat, PetscInt *nstash, PetscInt *reallocs, PetscInt *bnstash, PetscInt *breallocs)
9700: {
9701: PetscFunctionBegin;
9702: PetscCall(MatStashGetInfo_Private(&mat->stash, nstash, reallocs));
9703: PetscCall(MatStashGetInfo_Private(&mat->bstash, bnstash, breallocs));
9704: PetscFunctionReturn(PETSC_SUCCESS);
9705: }
9707: /*@
9708: MatCreateVecs - Get vector(s) compatible with the matrix, i.e. with the same
9709: parallel layout, `PetscLayout` for rows and columns
9711: Collective
9713: Input Parameter:
9714: . mat - the matrix
9716: Output Parameters:
9717: + right - (optional) vector that the matrix can be multiplied against
9718: - left - (optional) vector that the matrix vector product can be stored in
9720: Options Database Key:
9721: . -mat_vec_type type - set the `VecType` of the created vectors during `MatSetFromOptions()`
9723: Level: advanced
9725: Notes:
9726: The blocksize of the returned vectors is determined by the row and column block sizes set with `MatSetBlockSizes()` or the single blocksize (same for both) set by `MatSetBlockSize()`.
9728: The `VecType` of the created vectors is determined by the `MatType` of `mat`. This can be overridden by using `MatSetVecType()` or the option `-mat_vec_type`.
9730: These are new vectors which are not owned by the `mat`, they should be destroyed with `VecDestroy()` when no longer needed.
9732: PETSc `Vec` always have all zero entries when created with `MatCreateVecs()` until routines such as `VecSet()` or `VecSetValues()`
9733: are used to change the values. There is no reason to call `VecZeroEntries()` after creation.
9735: .seealso: [](ch_matrices), `Mat`, `Vec`, `VecCreate()`, `VecDestroy()`, `DMCreateGlobalVector()`, `MatSetVecType()`
9736: @*/
9737: PetscErrorCode MatCreateVecs(Mat mat, Vec *right, Vec *left)
9738: {
9739: PetscFunctionBegin;
9742: if (mat->ops->getvecs) {
9743: PetscUseTypeMethod(mat, getvecs, right, left);
9744: } else {
9745: if (right) {
9746: PetscCheck(mat->cmap->n >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "PetscLayout for columns not yet setup");
9747: PetscCall(VecCreateWithLayout_Private(mat->cmap, right));
9748: PetscCall(VecSetType(*right, mat->defaultvectype));
9749: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
9750: if (mat->boundtocpu && mat->bindingpropagates) {
9751: PetscCall(VecSetBindingPropagates(*right, PETSC_TRUE));
9752: PetscCall(VecBindToCPU(*right, PETSC_TRUE));
9753: }
9754: #endif
9755: }
9756: if (left) {
9757: PetscCheck(mat->rmap->n >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "PetscLayout for rows not yet setup");
9758: PetscCall(VecCreateWithLayout_Private(mat->rmap, left));
9759: PetscCall(VecSetType(*left, mat->defaultvectype));
9760: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
9761: if (mat->boundtocpu && mat->bindingpropagates) {
9762: PetscCall(VecSetBindingPropagates(*left, PETSC_TRUE));
9763: PetscCall(VecBindToCPU(*left, PETSC_TRUE));
9764: }
9765: #endif
9766: }
9767: }
9768: PetscFunctionReturn(PETSC_SUCCESS);
9769: }
9771: /*@
9772: MatFactorInfoInitialize - Initializes a `MatFactorInfo` data structure
9773: with default values.
9775: Not Collective
9777: Input Parameter:
9778: . info - the `MatFactorInfo` data structure
9780: Level: developer
9782: Notes:
9783: The solvers are generally used through the `KSP` and `PC` objects, for example
9784: `PCLU`, `PCILU`, `PCCHOLESKY`, `PCICC`
9786: Once the data structure is initialized one may change certain entries as desired for the particular factorization to be performed
9788: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorInfo`
9789: @*/
9790: PetscErrorCode MatFactorInfoInitialize(MatFactorInfo *info)
9791: {
9792: PetscFunctionBegin;
9793: PetscCall(PetscMemzero(info, sizeof(MatFactorInfo)));
9794: PetscFunctionReturn(PETSC_SUCCESS);
9795: }
9797: /*@
9798: MatFactorSetSchurIS - Set indices corresponding to the Schur complement you wish to have computed
9800: Collective
9802: Input Parameters:
9803: + mat - the factored matrix
9804: - is - the index set defining the Schur indices (0-based)
9806: Level: advanced
9808: Notes:
9809: Call `MatFactorSolveSchurComplement()` or `MatFactorSolveSchurComplementTranspose()` after this call to solve a Schur complement system.
9811: You can call `MatFactorGetSchurComplement()` or `MatFactorCreateSchurComplement()` after this call.
9813: This functionality is only supported for `MATSOLVERMUMPS` and `MATSOLVERMKL_PARDISO`
9815: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorGetSchurComplement()`, `MatFactorRestoreSchurComplement()`, `MatFactorCreateSchurComplement()`, `MatFactorSolveSchurComplement()`,
9816: `MatFactorSolveSchurComplementTranspose()`, `MATSOLVERMUMPS`, `MATSOLVERMKL_PARDISO`
9817: @*/
9818: PetscErrorCode MatFactorSetSchurIS(Mat mat, IS is)
9819: {
9820: PetscErrorCode (*f)(Mat, IS);
9822: PetscFunctionBegin;
9827: PetscCheckSameComm(mat, 1, is, 2);
9828: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Only for factored matrix");
9829: PetscCall(PetscObjectQueryFunction((PetscObject)mat, "MatFactorSetSchurIS_C", &f));
9830: PetscCheck(f, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "The selected MatSolverType does not support Schur complement computation. You should use MATSOLVERMUMPS or MATSOLVERMKL_PARDISO");
9831: PetscCall(MatDestroy(&mat->schur));
9832: PetscCall((*f)(mat, is));
9833: PetscCheck(mat->schur, PetscObjectComm((PetscObject)mat), PETSC_ERR_PLIB, "Schur complement has not been created");
9834: PetscFunctionReturn(PETSC_SUCCESS);
9835: }
9837: /*@
9838: MatFactorCreateSchurComplement - Create a Schur complement matrix object using Schur data computed during the factorization step
9840: Logically Collective
9842: Input Parameters:
9843: + F - the factored matrix obtained by calling `MatGetFactor()`
9844: . S - location where to return the Schur complement, can be `NULL`
9845: - status - the status of the Schur complement matrix, can be `NULL`
9847: Level: advanced
9849: Notes:
9850: You must call `MatFactorSetSchurIS()` before calling this routine.
9852: This functionality is only supported for `MATSOLVERMUMPS` and `MATSOLVERMKL_PARDISO`
9854: The routine provides a copy of the Schur matrix stored within the solver data structures.
9855: The caller must destroy the object when it is no longer needed.
9856: If `MatFactorInvertSchurComplement()` has been called, the routine gets back the inverse.
9858: Use `MatFactorGetSchurComplement()` to get access to the Schur complement matrix inside the factored matrix instead of making a copy of it (which this function does)
9860: See `MatCreateSchurComplement()` or `MatGetSchurComplement()` for ways to create virtual or approximate Schur complements.
9862: Developer Note:
9863: The reason this routine exists is because the representation of the Schur complement within the factor matrix may be different than a standard PETSc
9864: matrix representation and we normally do not want to use the time or memory to make a copy as a regular PETSc matrix.
9866: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorGetSchurComplement()`, `MatFactorSchurStatus`, `MATSOLVERMUMPS`, `MATSOLVERMKL_PARDISO`
9867: @*/
9868: PetscErrorCode MatFactorCreateSchurComplement(Mat F, Mat *S, MatFactorSchurStatus *status)
9869: {
9870: PetscFunctionBegin;
9872: if (S) PetscAssertPointer(S, 2);
9873: if (status) PetscAssertPointer(status, 3);
9874: if (S) {
9875: PetscErrorCode (*f)(Mat, Mat *);
9877: PetscCall(PetscObjectQueryFunction((PetscObject)F, "MatFactorCreateSchurComplement_C", &f));
9878: if (f) PetscCall((*f)(F, S));
9879: else PetscCall(MatDuplicate(F->schur, MAT_COPY_VALUES, S));
9880: }
9881: if (status) *status = F->schur_status;
9882: PetscFunctionReturn(PETSC_SUCCESS);
9883: }
9885: /*@
9886: MatFactorGetSchurComplement - Gets access to a Schur complement matrix using the current Schur data within a factored matrix
9888: Logically Collective
9890: Input Parameters:
9891: + F - the factored matrix obtained by calling `MatGetFactor()`
9892: . S - location where to return the Schur complement, can be `NULL`
9893: - status - the status of the Schur complement matrix, can be `NULL`
9895: Level: advanced
9897: Notes:
9898: You must call `MatFactorSetSchurIS()` before calling this routine.
9900: Schur complement mode is currently implemented for sequential matrices with factor type of `MATSOLVERMUMPS`
9902: The routine returns a the Schur Complement stored within the data structures of the solver.
9904: If `MatFactorInvertSchurComplement()` has previously been called, the returned matrix is actually the inverse of the Schur complement.
9906: The returned matrix should not be destroyed; the caller should call `MatFactorRestoreSchurComplement()` when the object is no longer needed.
9908: Use `MatFactorCreateSchurComplement()` to create a copy of the Schur complement matrix that is within a factored matrix
9910: See `MatCreateSchurComplement()` or `MatGetSchurComplement()` for ways to create virtual or approximate Schur complements.
9912: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorRestoreSchurComplement()`, `MatFactorCreateSchurComplement()`, `MatFactorSchurStatus`
9913: @*/
9914: PetscErrorCode MatFactorGetSchurComplement(Mat F, Mat *S, MatFactorSchurStatus *status)
9915: {
9916: PetscFunctionBegin;
9918: if (S) {
9919: PetscAssertPointer(S, 2);
9920: *S = F->schur;
9921: }
9922: if (status) {
9923: PetscAssertPointer(status, 3);
9924: *status = F->schur_status;
9925: }
9926: PetscFunctionReturn(PETSC_SUCCESS);
9927: }
9929: static PetscErrorCode MatFactorUpdateSchurStatus_Private(Mat F)
9930: {
9931: Mat S = F->schur;
9933: PetscFunctionBegin;
9934: switch (F->schur_status) {
9935: case MAT_FACTOR_SCHUR_UNFACTORED: // fall-through
9936: case MAT_FACTOR_SCHUR_INVERTED:
9937: if (S) {
9938: S->ops->solve = NULL;
9939: S->ops->matsolve = NULL;
9940: S->ops->solvetranspose = NULL;
9941: S->ops->matsolvetranspose = NULL;
9942: S->ops->solveadd = NULL;
9943: S->ops->solvetransposeadd = NULL;
9944: S->factortype = MAT_FACTOR_NONE;
9945: PetscCall(PetscFree(S->solvertype));
9946: }
9947: case MAT_FACTOR_SCHUR_FACTORED: // fall-through
9948: break;
9949: default:
9950: SETERRQ(PetscObjectComm((PetscObject)F), PETSC_ERR_SUP, "Unhandled MatFactorSchurStatus %d", F->schur_status);
9951: }
9952: PetscFunctionReturn(PETSC_SUCCESS);
9953: }
9955: /*@
9956: MatFactorRestoreSchurComplement - Restore the Schur complement matrix object obtained from a call to `MatFactorGetSchurComplement()`
9958: Logically Collective
9960: Input Parameters:
9961: + F - the factored matrix obtained by calling `MatGetFactor()`
9962: . S - location where the Schur complement is stored
9963: - status - the status of the Schur complement matrix (see `MatFactorSchurStatus`)
9965: Level: advanced
9967: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorCreateSchurComplement()`, `MatFactorSchurStatus`
9968: @*/
9969: PetscErrorCode MatFactorRestoreSchurComplement(Mat F, Mat *S, MatFactorSchurStatus status)
9970: {
9971: PetscFunctionBegin;
9973: if (S) {
9975: *S = NULL;
9976: }
9977: F->schur_status = status;
9978: PetscCall(MatFactorUpdateSchurStatus_Private(F));
9979: PetscFunctionReturn(PETSC_SUCCESS);
9980: }
9982: /*@
9983: MatFactorSolveSchurComplementTranspose - Solve the transpose of the Schur complement system computed during the factorization step
9985: Logically Collective
9987: Input Parameters:
9988: + F - the factored matrix obtained by calling `MatGetFactor()`
9989: . rhs - location where the right-hand side of the Schur complement system is stored
9990: - sol - location where the solution of the Schur complement system has to be returned
9992: Level: advanced
9994: Notes:
9995: The sizes of the vectors should match the size of the Schur complement
9997: Must be called after `MatFactorSetSchurIS()`
9999: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorSolveSchurComplement()`
10000: @*/
10001: PetscErrorCode MatFactorSolveSchurComplementTranspose(Mat F, Vec rhs, Vec sol)
10002: {
10003: PetscFunctionBegin;
10010: PetscCheckSameComm(F, 1, rhs, 2);
10011: PetscCheckSameComm(F, 1, sol, 3);
10012: PetscCall(MatFactorFactorizeSchurComplement(F));
10013: switch (F->schur_status) {
10014: case MAT_FACTOR_SCHUR_FACTORED:
10015: PetscCall(MatSolveTranspose(F->schur, rhs, sol));
10016: break;
10017: case MAT_FACTOR_SCHUR_INVERTED:
10018: PetscCall(MatMultTranspose(F->schur, rhs, sol));
10019: break;
10020: default:
10021: SETERRQ(PetscObjectComm((PetscObject)F), PETSC_ERR_SUP, "Unhandled MatFactorSchurStatus %d", F->schur_status);
10022: }
10023: PetscFunctionReturn(PETSC_SUCCESS);
10024: }
10026: /*@
10027: MatFactorSolveSchurComplement - Solve the Schur complement system computed during the factorization step
10029: Logically Collective
10031: Input Parameters:
10032: + F - the factored matrix obtained by calling `MatGetFactor()`
10033: . rhs - location where the right-hand side of the Schur complement system is stored
10034: - sol - location where the solution of the Schur complement system has to be returned
10036: Level: advanced
10038: Notes:
10039: The sizes of the vectors should match the size of the Schur complement
10041: Must be called after `MatFactorSetSchurIS()`
10043: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorSolveSchurComplementTranspose()`
10044: @*/
10045: PetscErrorCode MatFactorSolveSchurComplement(Mat F, Vec rhs, Vec sol)
10046: {
10047: PetscFunctionBegin;
10054: PetscCheckSameComm(F, 1, rhs, 2);
10055: PetscCheckSameComm(F, 1, sol, 3);
10056: PetscCall(MatFactorFactorizeSchurComplement(F));
10057: switch (F->schur_status) {
10058: case MAT_FACTOR_SCHUR_FACTORED:
10059: PetscCall(MatSolve(F->schur, rhs, sol));
10060: break;
10061: case MAT_FACTOR_SCHUR_INVERTED:
10062: PetscCall(MatMult(F->schur, rhs, sol));
10063: break;
10064: default:
10065: SETERRQ(PetscObjectComm((PetscObject)F), PETSC_ERR_SUP, "Unhandled MatFactorSchurStatus %d", F->schur_status);
10066: }
10067: PetscFunctionReturn(PETSC_SUCCESS);
10068: }
10070: PETSC_SINGLE_LIBRARY_INTERN PetscErrorCode MatSeqDenseInvertFactors_Private(Mat);
10071: #if PetscDefined(HAVE_CUDA)
10072: PETSC_SINGLE_LIBRARY_INTERN PetscErrorCode MatSeqDenseCUDAInvertFactors_Internal(Mat);
10073: #endif
10075: /* Schur status updated in the interface */
10076: static PetscErrorCode MatFactorInvertSchurComplement_Private(Mat F)
10077: {
10078: Mat S = F->schur;
10080: PetscFunctionBegin;
10081: if (S) {
10082: PetscMPIInt size;
10083: PetscBool isdense, isdensecuda;
10085: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)S), &size));
10086: PetscCheck(size <= 1, PetscObjectComm((PetscObject)S), PETSC_ERR_SUP, "Not yet implemented");
10087: PetscCall(PetscObjectTypeCompare((PetscObject)S, MATSEQDENSE, &isdense));
10088: PetscCall(PetscObjectTypeCompare((PetscObject)S, MATSEQDENSECUDA, &isdensecuda));
10089: PetscCheck(isdense || isdensecuda, PetscObjectComm((PetscObject)S), PETSC_ERR_SUP, "Not implemented for type %s", ((PetscObject)S)->type_name);
10090: PetscCall(PetscLogEventBegin(MAT_FactorInvS, F, 0, 0, 0));
10091: if (isdense) {
10092: PetscCall(MatSeqDenseInvertFactors_Private(S));
10093: } else if (isdensecuda) {
10094: #if defined(PETSC_HAVE_CUDA)
10095: PetscCall(MatSeqDenseCUDAInvertFactors_Internal(S));
10096: #endif
10097: }
10098: // HIP??????????????
10099: PetscCall(PetscLogEventEnd(MAT_FactorInvS, F, 0, 0, 0));
10100: }
10101: PetscFunctionReturn(PETSC_SUCCESS);
10102: }
10104: /*@
10105: MatFactorInvertSchurComplement - Invert the Schur complement matrix computed during the factorization step
10107: Logically Collective
10109: Input Parameter:
10110: . F - the factored matrix obtained by calling `MatGetFactor()`
10112: Level: advanced
10114: Notes:
10115: Must be called after `MatFactorSetSchurIS()`.
10117: Call `MatFactorGetSchurComplement()` or `MatFactorCreateSchurComplement()` AFTER this call to actually compute the inverse and get access to it.
10119: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorGetSchurComplement()`, `MatFactorCreateSchurComplement()`
10120: @*/
10121: PetscErrorCode MatFactorInvertSchurComplement(Mat F)
10122: {
10123: PetscFunctionBegin;
10126: if (F->schur_status == MAT_FACTOR_SCHUR_INVERTED) PetscFunctionReturn(PETSC_SUCCESS);
10127: PetscCall(MatFactorFactorizeSchurComplement(F));
10128: PetscCall(MatFactorInvertSchurComplement_Private(F));
10129: F->schur_status = MAT_FACTOR_SCHUR_INVERTED;
10130: PetscFunctionReturn(PETSC_SUCCESS);
10131: }
10133: /*@
10134: MatFactorFactorizeSchurComplement - Factorize the Schur complement matrix computed during the factorization step
10136: Logically Collective
10138: Input Parameter:
10139: . F - the factored matrix obtained by calling `MatGetFactor()`
10141: Level: advanced
10143: Note:
10144: Must be called after `MatFactorSetSchurIS()`
10146: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorInvertSchurComplement()`
10147: @*/
10148: PetscErrorCode MatFactorFactorizeSchurComplement(Mat F)
10149: {
10150: MatFactorInfo info;
10152: PetscFunctionBegin;
10155: if (F->schur_status == MAT_FACTOR_SCHUR_INVERTED || F->schur_status == MAT_FACTOR_SCHUR_FACTORED) PetscFunctionReturn(PETSC_SUCCESS);
10156: PetscCall(PetscLogEventBegin(MAT_FactorFactS, F, 0, 0, 0));
10157: PetscCall(PetscMemzero(&info, sizeof(MatFactorInfo)));
10158: if (F->factortype == MAT_FACTOR_CHOLESKY) { /* LDL^t regarded as Cholesky */
10159: PetscCall(MatCholeskyFactor(F->schur, NULL, &info));
10160: } else {
10161: PetscCall(MatLUFactor(F->schur, NULL, NULL, &info));
10162: }
10163: PetscCall(PetscLogEventEnd(MAT_FactorFactS, F, 0, 0, 0));
10164: F->schur_status = MAT_FACTOR_SCHUR_FACTORED;
10165: PetscFunctionReturn(PETSC_SUCCESS);
10166: }
10168: /*@
10169: MatPtAP - Creates the matrix product $C = P^T * A * P$
10171: Neighbor-wise Collective
10173: Input Parameters:
10174: + A - the matrix
10175: . P - the projection matrix
10176: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10177: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(P)), use `PETSC_DETERMINE` or `PETSC_CURRENT` if you do not have a good estimate
10178: if the result is a dense matrix this is irrelevant
10180: Output Parameter:
10181: . C - the product matrix
10183: Level: intermediate
10185: Notes:
10186: `C` will be created and must be destroyed by the user with `MatDestroy()`.
10188: This is a convenience routine that wraps the use of the `MatProductCreate()` with a `MatProductType` of `MATPRODUCT_PtAP`
10189: functionality into a single function call. For more involved matrix-matrix operations see `MatProductCreate()`.
10191: The deprecated `PETSC_DEFAULT` in `fill` also means use the current value
10193: Developer Note:
10194: For matrix types without special implementation the function fallbacks to `MatMatMult()` followed by `MatTransposeMatMult()`.
10196: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MatMatMult()`, `MatRARt()`
10197: @*/
10198: PetscErrorCode MatPtAP(Mat A, Mat P, MatReuse scall, PetscReal fill, Mat *C)
10199: {
10200: PetscFunctionBegin;
10201: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*C, 5);
10202: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
10204: if (scall == MAT_INITIAL_MATRIX) {
10205: PetscCall(MatProductCreate(A, P, NULL, C));
10206: PetscCall(MatProductSetType(*C, MATPRODUCT_PtAP));
10207: PetscCall(MatProductSetAlgorithm(*C, "default"));
10208: PetscCall(MatProductSetFill(*C, fill));
10210: (*C)->product->api_user = PETSC_TRUE;
10211: PetscCall(MatProductSetFromOptions(*C));
10212: PetscCheck((*C)->ops->productsymbolic, PetscObjectComm((PetscObject)*C), PETSC_ERR_SUP, "MatProduct %s not supported for A %s and P %s", MatProductTypes[MATPRODUCT_PtAP], ((PetscObject)A)->type_name, ((PetscObject)P)->type_name);
10213: PetscCall(MatProductSymbolic(*C));
10214: } else { /* scall == MAT_REUSE_MATRIX */
10215: PetscCall(MatProductReplaceMats(A, P, NULL, *C));
10216: }
10218: PetscCall(MatProductNumeric(*C));
10219: if (A->symmetric == PETSC_BOOL3_TRUE) {
10220: PetscCall(MatSetOption(*C, MAT_SYMMETRIC, PETSC_TRUE));
10221: (*C)->spd = A->spd;
10222: }
10223: PetscFunctionReturn(PETSC_SUCCESS);
10224: }
10226: /*@
10227: MatRARt - Creates the matrix product $C = R * A * R^T$
10229: Neighbor-wise Collective
10231: Input Parameters:
10232: + A - the matrix
10233: . R - the projection matrix
10234: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10235: - fill - expected fill as ratio of nnz(C)/nnz(A), use `PETSC_DETERMINE` or `PETSC_CURRENT` if you do not have a good estimate
10236: if the result is a dense matrix this is irrelevant
10238: Output Parameter:
10239: . C - the product matrix
10241: Level: intermediate
10243: Notes:
10244: `C` will be created and must be destroyed by the user with `MatDestroy()`.
10246: This is a convenience routine that wraps the use of the `MatProductCreate()` with a `MatProductType` of `MATPRODUCT_RARt`
10247: functionality into a single function call. For more involved matrix-matrix operations see `MatProductCreate()`.
10249: This routine is currently only implemented for pairs of `MATAIJ` matrices and classes
10250: which inherit from `MATAIJ`. Due to PETSc sparse matrix block row distribution among processes,
10251: the parallel `MatRARt()` is implemented computing the explicit transpose of `R`, which can be very expensive.
10252: We recommend using `MatPtAP()` when possible.
10254: The deprecated `PETSC_DEFAULT` in `fill` also means use the current value
10256: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MatMatMult()`, `MatPtAP()`
10257: @*/
10258: PetscErrorCode MatRARt(Mat A, Mat R, MatReuse scall, PetscReal fill, Mat *C)
10259: {
10260: PetscFunctionBegin;
10261: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*C, 5);
10262: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
10264: if (scall == MAT_INITIAL_MATRIX) {
10265: PetscCall(MatProductCreate(A, R, NULL, C));
10266: PetscCall(MatProductSetType(*C, MATPRODUCT_RARt));
10267: PetscCall(MatProductSetAlgorithm(*C, "default"));
10268: PetscCall(MatProductSetFill(*C, fill));
10270: (*C)->product->api_user = PETSC_TRUE;
10271: PetscCall(MatProductSetFromOptions(*C));
10272: PetscCheck((*C)->ops->productsymbolic, PetscObjectComm((PetscObject)*C), PETSC_ERR_SUP, "MatProduct %s not supported for A %s and R %s", MatProductTypes[MATPRODUCT_RARt], ((PetscObject)A)->type_name, ((PetscObject)R)->type_name);
10273: PetscCall(MatProductSymbolic(*C));
10274: } else { /* scall == MAT_REUSE_MATRIX */
10275: PetscCall(MatProductReplaceMats(A, R, NULL, *C));
10276: }
10278: PetscCall(MatProductNumeric(*C));
10279: if (A->symmetric == PETSC_BOOL3_TRUE) PetscCall(MatSetOption(*C, MAT_SYMMETRIC, PETSC_TRUE));
10280: PetscFunctionReturn(PETSC_SUCCESS);
10281: }
10283: static PetscErrorCode MatProduct_Private(Mat A, Mat B, MatReuse scall, PetscReal fill, MatProductType ptype, Mat *C)
10284: {
10285: PetscBool flg = PETSC_TRUE;
10287: PetscFunctionBegin;
10288: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "MAT_INPLACE_MATRIX product not supported");
10289: if (scall == MAT_INITIAL_MATRIX) {
10290: PetscCall(PetscInfo(A, "Calling MatProduct API with MAT_INITIAL_MATRIX and product type %s\n", MatProductTypes[ptype]));
10291: PetscCall(MatProductCreate(A, B, NULL, C));
10292: PetscCall(MatProductSetAlgorithm(*C, MATPRODUCTALGORITHMDEFAULT));
10293: PetscCall(MatProductSetFill(*C, fill));
10294: } else { /* scall == MAT_REUSE_MATRIX */
10295: Mat_Product *product = (*C)->product;
10297: PetscCall(PetscObjectBaseTypeCompareAny((PetscObject)*C, &flg, MATSEQDENSE, MATMPIDENSE, ""));
10298: if (flg && product && product->type != ptype) {
10299: PetscCall(MatProductClear(*C));
10300: product = NULL;
10301: }
10302: PetscCall(PetscInfo(A, "Calling MatProduct API with MAT_REUSE_MATRIX %s product present and product type %s\n", product ? "with" : "without", MatProductTypes[ptype]));
10303: if (!product) { /* user provide the dense matrix *C without calling MatProductCreate() or reusing it from previous calls */
10304: PetscCheck(flg, PetscObjectComm((PetscObject)*C), PETSC_ERR_SUP, "Call MatProductCreate() first");
10305: PetscCall(MatProductCreate_Private(A, B, NULL, *C));
10306: product = (*C)->product;
10307: product->fill = fill;
10308: product->clear = PETSC_TRUE;
10309: } else { /* user may change input matrices A or B when MAT_REUSE_MATRIX */
10310: flg = PETSC_FALSE;
10311: PetscCall(MatProductReplaceMats(A, B, NULL, *C));
10312: }
10313: }
10314: if (flg) {
10315: (*C)->product->api_user = PETSC_TRUE;
10316: PetscCall(MatProductSetType(*C, ptype));
10317: PetscCall(MatProductSetFromOptions(*C));
10318: PetscCall(MatProductSymbolic(*C));
10319: }
10320: PetscCall(MatProductNumeric(*C));
10321: PetscFunctionReturn(PETSC_SUCCESS);
10322: }
10324: /*@
10325: MatMatMult - Performs matrix-matrix multiplication $ C=A*B $.
10327: Neighbor-wise Collective
10329: Input Parameters:
10330: + A - the left matrix
10331: . B - the right matrix
10332: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10333: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use `PETSC_DETERMINE` or `PETSC_CURRENT` if you do not have a good estimate
10334: if the result is a dense matrix this is irrelevant
10336: Output Parameter:
10337: . C - the product matrix
10339: Notes:
10340: Unless scall is `MAT_REUSE_MATRIX` C will be created.
10342: `MAT_REUSE_MATRIX` can only be used if the matrices A and B have the same nonzero pattern as in the previous call and C was obtained from a previous
10343: call to this function with `MAT_INITIAL_MATRIX`.
10345: To determine the correct fill value, run with `-info` and search for the string "Fill ratio" to see the value actually needed.
10347: In the special case where matrix `B` (and hence `C`) are dense you can create the correctly sized matrix `C` yourself and then call this routine with `MAT_REUSE_MATRIX`,
10348: rather than first having `MatMatMult()` create it for you. You can NEVER do this if the matrix `C` is sparse.
10350: The deprecated `PETSC_DEFAULT` in `fill` also means use the current value
10352: This is a convenience routine that wraps the use of the `MatProductCreate()` with a `MatProductType` of `MATPRODUCT_AB`
10353: functionality into a single function call. For more involved matrix-matrix operations see `MatProductCreate()`.
10355: Example of Usage:
10356: .vb
10357: MatProductCreate(A,B,NULL,&C);
10358: MatProductSetType(C,MATPRODUCT_AB);
10359: MatProductSymbolic(C);
10360: MatProductNumeric(C); // compute C=A * B
10361: MatProductReplaceMats(A1,B1,NULL,C); // compute C=A1 * B1
10362: MatProductNumeric(C);
10363: MatProductReplaceMats(A2,NULL,NULL,C); // compute C=A2 * B1
10364: MatProductNumeric(C);
10365: .ve
10367: Level: intermediate
10369: .seealso: [](ch_matrices), `Mat`, `MatProductType`, `MATPRODUCT_AB`, `MatTransposeMatMult()`, `MatMatTransposeMult()`, `MatPtAP()`, `MatProductCreate()`, `MatProductSymbolic()`, `MatProductReplaceMats()`, `MatProductNumeric()`
10370: @*/
10371: PetscErrorCode MatMatMult(Mat A, Mat B, MatReuse scall, PetscReal fill, Mat *C)
10372: {
10373: PetscFunctionBegin;
10374: PetscCall(MatProduct_Private(A, B, scall, fill, MATPRODUCT_AB, C));
10375: PetscFunctionReturn(PETSC_SUCCESS);
10376: }
10378: /*@
10379: MatMatTransposeMult - Performs matrix-matrix multiplication $C = A*B^T$.
10381: Neighbor-wise Collective
10383: Input Parameters:
10384: + A - the left matrix
10385: . B - the right matrix
10386: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10387: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use `PETSC_DETERMINE` or `PETSC_CURRENT` if not known
10389: Output Parameter:
10390: . C - the product matrix
10392: Options Database Key:
10393: . -matmattransmult_mpidense_mpidense_via {allgatherv,cyclic} - Choose between algorithms for `MATMPIDENSE` matrices: the
10394: first redundantly copies the transposed `B` matrix on each process and requires O(log P) communication complexity;
10395: the second never stores more than one portion of the `B` matrix at a time but requires O(P) communication complexity.
10397: Level: intermediate
10399: Notes:
10400: C will be created if `MAT_INITIAL_MATRIX` and must be destroyed by the user with `MatDestroy()`.
10402: `MAT_REUSE_MATRIX` can only be used if the matrices A and B have the same nonzero pattern as in the previous call
10404: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
10405: actually needed.
10407: This routine is currently only implemented for pairs of `MATSEQAIJ` matrices, for the `MATSEQDENSE` class,
10408: and for pairs of `MATMPIDENSE` matrices.
10410: This is a convenience routine that wraps the use of the `MatProductCreate()` with a `MatProductType` of `MATPRODUCT_ABt`
10411: functionality into a single function call. For more involved matrix-matrix operations see `MatProductCreate()`.
10413: The deprecated `PETSC_DEFAULT` in `fill` also means use the current value
10415: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MATPRODUCT_ABt`, `MatMatMult()`, `MatTransposeMatMult()`, `MatPtAP()`, `MatProductAlgorithm`, `MatProductType`
10416: @*/
10417: PetscErrorCode MatMatTransposeMult(Mat A, Mat B, MatReuse scall, PetscReal fill, Mat *C)
10418: {
10419: PetscFunctionBegin;
10420: PetscCall(MatProduct_Private(A, B, scall, fill, MATPRODUCT_ABt, C));
10421: if (A == B) PetscCall(MatSetOption(*C, MAT_SYMMETRIC, PETSC_TRUE));
10422: PetscFunctionReturn(PETSC_SUCCESS);
10423: }
10425: /*@
10426: MatTransposeMatMult - Performs matrix-matrix multiplication $C = A^T*B$.
10428: Neighbor-wise Collective
10430: Input Parameters:
10431: + A - the left matrix
10432: . B - the right matrix
10433: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10434: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use `PETSC_DETERMINE` or `PETSC_CURRENT` if not known
10436: Output Parameter:
10437: . C - the product matrix
10439: Level: intermediate
10441: Notes:
10442: `C` will be created if `MAT_INITIAL_MATRIX` and must be destroyed by the user with `MatDestroy()`.
10444: `MAT_REUSE_MATRIX` can only be used if `A` and `B` have the same nonzero pattern as in the previous call.
10446: This is a convenience routine that wraps the use of `MatProductCreate()` with a `MatProductType` of `MATPRODUCT_AtB`
10447: functionality into a single function call. For more involved matrix-matrix operations see `MatProductCreate()`.
10449: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
10450: actually needed.
10452: This routine is currently implemented for pairs of `MATAIJ` matrices and pairs of `MATSEQDENSE` matrices and classes
10453: which inherit from `MATSEQAIJ`. `C` will be of the same type as the input matrices.
10455: The deprecated `PETSC_DEFAULT` in `fill` also means use the current value
10457: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MATPRODUCT_AtB`, `MatMatMult()`, `MatMatTransposeMult()`, `MatPtAP()`
10458: @*/
10459: PetscErrorCode MatTransposeMatMult(Mat A, Mat B, MatReuse scall, PetscReal fill, Mat *C)
10460: {
10461: PetscFunctionBegin;
10462: PetscCall(MatProduct_Private(A, B, scall, fill, MATPRODUCT_AtB, C));
10463: PetscFunctionReturn(PETSC_SUCCESS);
10464: }
10466: /*@
10467: MatMatMatMult - Performs matrix-matrix-matrix multiplication D=A*B*C.
10469: Neighbor-wise Collective
10471: Input Parameters:
10472: + A - the left matrix
10473: . B - the middle matrix
10474: . C - the right matrix
10475: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10476: - fill - expected fill as ratio of nnz(D)/(nnz(A) + nnz(B)+nnz(C)), use `PETSC_DETERMINE` or `PETSC_CURRENT` if you do not have a good estimate
10477: if the result is a dense matrix this is irrelevant
10479: Output Parameter:
10480: . D - the product matrix
10482: Level: intermediate
10484: Notes:
10485: Unless `scall` is `MAT_REUSE_MATRIX` `D` will be created.
10487: `MAT_REUSE_MATRIX` can only be used if the matrices `A`, `B`, and `C` have the same nonzero pattern as in the previous call
10489: This is a convenience routine that wraps the use of the `MatProductCreate()` with a `MatProductType` of `MATPRODUCT_ABC`
10490: functionality into a single function call. For more involved matrix-matrix operations see `MatProductCreate()`.
10492: To determine the correct fill value, run with `-info` and search for the string "Fill ratio" to see the value
10493: actually needed.
10495: If you have many matrices with the same non-zero structure to multiply, you
10496: should use `MAT_REUSE_MATRIX` in all calls but the first
10498: The deprecated `PETSC_DEFAULT` in `fill` also means use the current value
10500: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MATPRODUCT_ABC`, `MatMatMult`, `MatPtAP()`, `MatMatTransposeMult()`, `MatTransposeMatMult()`
10501: @*/
10502: PetscErrorCode MatMatMatMult(Mat A, Mat B, Mat C, MatReuse scall, PetscReal fill, Mat *D)
10503: {
10504: PetscFunctionBegin;
10505: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*D, 6);
10506: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
10508: if (scall == MAT_INITIAL_MATRIX) {
10509: PetscCall(MatProductCreate(A, B, C, D));
10510: PetscCall(MatProductSetType(*D, MATPRODUCT_ABC));
10511: PetscCall(MatProductSetAlgorithm(*D, "default"));
10512: PetscCall(MatProductSetFill(*D, fill));
10514: (*D)->product->api_user = PETSC_TRUE;
10515: PetscCall(MatProductSetFromOptions(*D));
10516: PetscCheck((*D)->ops->productsymbolic, PetscObjectComm((PetscObject)*D), PETSC_ERR_SUP, "MatProduct %s not supported for A %s, B %s and C %s", MatProductTypes[MATPRODUCT_ABC], ((PetscObject)A)->type_name, ((PetscObject)B)->type_name,
10517: ((PetscObject)C)->type_name);
10518: PetscCall(MatProductSymbolic(*D));
10519: } else { /* user may change input matrices when REUSE */
10520: PetscCall(MatProductReplaceMats(A, B, C, *D));
10521: }
10522: PetscCall(MatProductNumeric(*D));
10523: PetscFunctionReturn(PETSC_SUCCESS);
10524: }
10526: /*@
10527: MatCreateRedundantMatrix - Create redundant matrices and put them into processors of subcommunicators.
10529: Collective
10531: Input Parameters:
10532: + mat - the matrix
10533: . nsubcomm - the number of subcommunicators (= number of redundant parallel or sequential matrices)
10534: . subcomm - MPI communicator split from the communicator where mat resides in (or `MPI_COMM_NULL` if nsubcomm is used)
10535: - reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10537: Output Parameter:
10538: . matredundant - redundant matrix
10540: Level: advanced
10542: Notes:
10543: `MAT_REUSE_MATRIX` can only be used when the nonzero structure of the
10544: original matrix has not changed from that last call to `MatCreateRedundantMatrix()`.
10546: This routine creates the duplicated matrices in the subcommunicators; you should NOT create them before
10547: calling it.
10549: `PetscSubcommCreate()` can be used to manage the creation of the subcomm but need not be.
10551: .seealso: [](ch_matrices), `Mat`, `MatDestroy()`, `PetscSubcommCreate()`, `PetscSubcomm`
10552: @*/
10553: PetscErrorCode MatCreateRedundantMatrix(Mat mat, PetscInt nsubcomm, MPI_Comm subcomm, MatReuse reuse, Mat *matredundant)
10554: {
10555: MPI_Comm comm;
10556: PetscMPIInt size;
10557: PetscInt mloc_sub, nloc_sub, rstart, rend, M = mat->rmap->N, N = mat->cmap->N, bs = mat->rmap->bs;
10558: Mat_Redundant *redund = NULL;
10559: PetscSubcomm psubcomm = NULL;
10560: MPI_Comm subcomm_in = subcomm;
10561: Mat *matseq;
10562: IS isrow, iscol;
10563: PetscBool newsubcomm = PETSC_FALSE;
10565: PetscFunctionBegin;
10567: if (nsubcomm && reuse == MAT_REUSE_MATRIX) {
10568: PetscAssertPointer(*matredundant, 5);
10570: }
10572: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
10573: if (size == 1 || nsubcomm == 1) {
10574: if (reuse == MAT_INITIAL_MATRIX) {
10575: PetscCall(MatDuplicate(mat, MAT_COPY_VALUES, matredundant));
10576: } else {
10577: PetscCheck(*matredundant != mat, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "MAT_REUSE_MATRIX means reuse the matrix passed in as the final argument, not the original matrix");
10578: PetscCall(MatCopy(mat, *matredundant, SAME_NONZERO_PATTERN));
10579: }
10580: PetscFunctionReturn(PETSC_SUCCESS);
10581: }
10583: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10584: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10585: MatCheckPreallocated(mat, 1);
10587: PetscCall(PetscLogEventBegin(MAT_RedundantMat, mat, 0, 0, 0));
10588: if (subcomm_in == MPI_COMM_NULL && reuse == MAT_INITIAL_MATRIX) { /* get subcomm if user does not provide subcomm */
10589: /* create psubcomm, then get subcomm */
10590: PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
10591: PetscCallMPI(MPI_Comm_size(comm, &size));
10592: PetscCheck(nsubcomm >= 1 && nsubcomm <= size, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "nsubcomm must between 1 and %d", size);
10594: PetscCall(PetscSubcommCreate(comm, &psubcomm));
10595: PetscCall(PetscSubcommSetNumber(psubcomm, nsubcomm));
10596: PetscCall(PetscSubcommSetType(psubcomm, PETSC_SUBCOMM_CONTIGUOUS));
10597: PetscCall(PetscSubcommSetFromOptions(psubcomm));
10598: PetscCall(PetscCommDuplicate(PetscSubcommChild(psubcomm), &subcomm, NULL));
10599: newsubcomm = PETSC_TRUE;
10600: PetscCall(PetscSubcommDestroy(&psubcomm));
10601: }
10603: /* get isrow, iscol and a local sequential matrix matseq[0] */
10604: if (reuse == MAT_INITIAL_MATRIX) {
10605: mloc_sub = PETSC_DECIDE;
10606: nloc_sub = PETSC_DECIDE;
10607: if (bs < 1) {
10608: PetscCall(PetscSplitOwnership(subcomm, &mloc_sub, &M));
10609: PetscCall(PetscSplitOwnership(subcomm, &nloc_sub, &N));
10610: } else {
10611: PetscCall(PetscSplitOwnershipBlock(subcomm, bs, &mloc_sub, &M));
10612: PetscCall(PetscSplitOwnershipBlock(subcomm, bs, &nloc_sub, &N));
10613: }
10614: PetscCallMPI(MPI_Scan(&mloc_sub, &rend, 1, MPIU_INT, MPI_SUM, subcomm));
10615: rstart = rend - mloc_sub;
10616: PetscCall(ISCreateStride(PETSC_COMM_SELF, mloc_sub, rstart, 1, &isrow));
10617: PetscCall(ISCreateStride(PETSC_COMM_SELF, N, 0, 1, &iscol));
10618: PetscCall(ISSetIdentity(iscol));
10619: } else { /* reuse == MAT_REUSE_MATRIX */
10620: PetscCheck(*matredundant != mat, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "MAT_REUSE_MATRIX means reuse the matrix passed in as the final argument, not the original matrix");
10621: /* retrieve subcomm */
10622: PetscCall(PetscObjectGetComm((PetscObject)*matredundant, &subcomm));
10623: redund = (*matredundant)->redundant;
10624: isrow = redund->isrow;
10625: iscol = redund->iscol;
10626: matseq = redund->matseq;
10627: }
10628: PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscol, reuse, &matseq));
10630: /* get matredundant over subcomm */
10631: if (reuse == MAT_INITIAL_MATRIX) {
10632: PetscCall(MatCreateMPIMatConcatenateSeqMat(subcomm, matseq[0], nloc_sub, reuse, matredundant));
10634: /* create a supporting struct and attach it to C for reuse */
10635: PetscCall(PetscNew(&redund));
10636: (*matredundant)->redundant = redund;
10637: redund->isrow = isrow;
10638: redund->iscol = iscol;
10639: redund->matseq = matseq;
10640: if (newsubcomm) {
10641: redund->subcomm = subcomm;
10642: } else {
10643: redund->subcomm = MPI_COMM_NULL;
10644: }
10645: } else {
10646: PetscCall(MatCreateMPIMatConcatenateSeqMat(subcomm, matseq[0], PETSC_DECIDE, reuse, matredundant));
10647: }
10648: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
10649: if (matseq[0]->boundtocpu && matseq[0]->bindingpropagates) {
10650: PetscCall(MatBindToCPU(*matredundant, PETSC_TRUE));
10651: PetscCall(MatSetBindingPropagates(*matredundant, PETSC_TRUE));
10652: }
10653: #endif
10654: PetscCall(PetscLogEventEnd(MAT_RedundantMat, mat, 0, 0, 0));
10655: PetscFunctionReturn(PETSC_SUCCESS);
10656: }
10658: /*@C
10659: MatGetMultiProcBlock - Create multiple 'parallel submatrices' from
10660: a given `Mat`. Each submatrix can span multiple procs.
10662: Collective
10664: Input Parameters:
10665: + mat - the matrix
10666: . subComm - the sub communicator obtained as if by `MPI_Comm_split(PetscObjectComm((PetscObject)mat))`
10667: - scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10669: Output Parameter:
10670: . subMat - parallel sub-matrices each spanning a given `subcomm`
10672: Level: advanced
10674: Notes:
10675: The submatrix partition across processors is dictated by `subComm` a
10676: communicator obtained by `MPI_comm_split()` or via `PetscSubcommCreate()`. The `subComm`
10677: is not restricted to be grouped with consecutive original MPI processes.
10679: Due the `MPI_Comm_split()` usage, the parallel layout of the submatrices
10680: map directly to the layout of the original matrix [wrt the local
10681: row,col partitioning]. So the original 'DiagonalMat' naturally maps
10682: into the 'DiagonalMat' of the `subMat`, hence it is used directly from
10683: the `subMat`. However the offDiagMat looses some columns - and this is
10684: reconstructed with `MatSetValues()`
10686: This is used by `PCBJACOBI` when a single block spans multiple MPI processes.
10688: .seealso: [](ch_matrices), `Mat`, `MatCreateRedundantMatrix()`, `MatCreateSubMatrices()`, `PCBJACOBI`
10689: @*/
10690: PetscErrorCode MatGetMultiProcBlock(Mat mat, MPI_Comm subComm, MatReuse scall, Mat *subMat)
10691: {
10692: PetscMPIInt commsize, subCommSize;
10694: PetscFunctionBegin;
10695: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &commsize));
10696: PetscCallMPI(MPI_Comm_size(subComm, &subCommSize));
10697: PetscCheck(subCommSize <= commsize, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "CommSize %d < SubCommZize %d", commsize, subCommSize);
10699: PetscCheck(scall != MAT_REUSE_MATRIX || *subMat != mat, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "MAT_REUSE_MATRIX means reuse the matrix passed in as the final argument, not the original matrix");
10700: PetscCall(PetscLogEventBegin(MAT_GetMultiProcBlock, mat, 0, 0, 0));
10701: PetscUseTypeMethod(mat, getmultiprocblock, subComm, scall, subMat);
10702: PetscCall(PetscLogEventEnd(MAT_GetMultiProcBlock, mat, 0, 0, 0));
10703: PetscFunctionReturn(PETSC_SUCCESS);
10704: }
10706: /*@
10707: MatGetLocalSubMatrix - Gets a reference to a submatrix specified in local numbering
10709: Not Collective
10711: Input Parameters:
10712: + mat - matrix to extract local submatrix from
10713: . isrow - local row indices for submatrix
10714: - iscol - local column indices for submatrix
10716: Output Parameter:
10717: . submat - the submatrix
10719: Level: intermediate
10721: Notes:
10722: `submat` should be disposed of with `MatRestoreLocalSubMatrix()`.
10724: Depending on the format of `mat`, the returned `submat` may not implement `MatMult()`. Its communicator may be
10725: the same as `mat`, it may be `PETSC_COMM_SELF`, or some other sub-communictor of `mat`'s.
10727: `submat` always implements `MatSetValuesLocal()`. If `isrow` and `iscol` have the same block size, then
10728: `MatSetValuesBlockedLocal()` will also be implemented.
10730: `mat` must have had a `ISLocalToGlobalMapping` provided to it with `MatSetLocalToGlobalMapping()`.
10731: Matrices obtained with `DMCreateMatrix()` generally already have the local to global mapping provided.
10733: .seealso: [](ch_matrices), `Mat`, `MatRestoreLocalSubMatrix()`, `MatCreateLocalRef()`, `MatSetLocalToGlobalMapping()`
10734: @*/
10735: PetscErrorCode MatGetLocalSubMatrix(Mat mat, IS isrow, IS iscol, Mat *submat)
10736: {
10737: PetscFunctionBegin;
10741: PetscCheckSameComm(isrow, 2, iscol, 3);
10742: PetscAssertPointer(submat, 4);
10743: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Matrix must have local to global mapping provided before this call");
10745: if (mat->ops->getlocalsubmatrix) {
10746: PetscUseTypeMethod(mat, getlocalsubmatrix, isrow, iscol, submat);
10747: } else {
10748: PetscCall(MatCreateLocalRef(mat, isrow, iscol, submat));
10749: }
10750: (*submat)->assembled = mat->assembled;
10751: PetscFunctionReturn(PETSC_SUCCESS);
10752: }
10754: /*@
10755: MatRestoreLocalSubMatrix - Restores a reference to a submatrix specified in local numbering obtained with `MatGetLocalSubMatrix()`
10757: Not Collective
10759: Input Parameters:
10760: + mat - matrix to extract local submatrix from
10761: . isrow - local row indices for submatrix
10762: . iscol - local column indices for submatrix
10763: - submat - the submatrix
10765: Level: intermediate
10767: .seealso: [](ch_matrices), `Mat`, `MatGetLocalSubMatrix()`
10768: @*/
10769: PetscErrorCode MatRestoreLocalSubMatrix(Mat mat, IS isrow, IS iscol, Mat *submat)
10770: {
10771: PetscFunctionBegin;
10775: PetscCheckSameComm(isrow, 2, iscol, 3);
10776: PetscAssertPointer(submat, 4);
10779: if (mat->ops->restorelocalsubmatrix) {
10780: PetscUseTypeMethod(mat, restorelocalsubmatrix, isrow, iscol, submat);
10781: } else {
10782: PetscCall(MatDestroy(submat));
10783: }
10784: *submat = NULL;
10785: PetscFunctionReturn(PETSC_SUCCESS);
10786: }
10788: /*@
10789: MatFindZeroDiagonals - Finds all the rows of a matrix that have zero or no diagonal entry in the matrix
10791: Collective
10793: Input Parameter:
10794: . mat - the matrix
10796: Output Parameter:
10797: . is - if any rows have zero diagonals this contains the list of them
10799: Level: developer
10801: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
10802: @*/
10803: PetscErrorCode MatFindZeroDiagonals(Mat mat, IS *is)
10804: {
10805: PetscFunctionBegin;
10808: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10809: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10811: if (!mat->ops->findzerodiagonals) {
10812: Vec diag;
10813: const PetscScalar *a;
10814: PetscInt *rows;
10815: PetscInt rStart, rEnd, r, nrow = 0;
10817: PetscCall(MatCreateVecs(mat, &diag, NULL));
10818: PetscCall(MatGetDiagonal(mat, diag));
10819: PetscCall(MatGetOwnershipRange(mat, &rStart, &rEnd));
10820: PetscCall(VecGetArrayRead(diag, &a));
10821: for (r = 0; r < rEnd - rStart; ++r)
10822: if (a[r] == 0.0) ++nrow;
10823: PetscCall(PetscMalloc1(nrow, &rows));
10824: nrow = 0;
10825: for (r = 0; r < rEnd - rStart; ++r)
10826: if (a[r] == 0.0) rows[nrow++] = r + rStart;
10827: PetscCall(VecRestoreArrayRead(diag, &a));
10828: PetscCall(VecDestroy(&diag));
10829: PetscCall(ISCreateGeneral(PetscObjectComm((PetscObject)mat), nrow, rows, PETSC_OWN_POINTER, is));
10830: } else {
10831: PetscUseTypeMethod(mat, findzerodiagonals, is);
10832: }
10833: PetscFunctionReturn(PETSC_SUCCESS);
10834: }
10836: /*@
10837: MatFindOffBlockDiagonalEntries - Finds all the rows of a matrix that have entries outside of the main diagonal block (defined by the matrix block size)
10839: Collective
10841: Input Parameter:
10842: . mat - the matrix
10844: Output Parameter:
10845: . is - contains the list of rows with off block diagonal entries
10847: Level: developer
10849: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
10850: @*/
10851: PetscErrorCode MatFindOffBlockDiagonalEntries(Mat mat, IS *is)
10852: {
10853: PetscFunctionBegin;
10856: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10857: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10859: PetscUseTypeMethod(mat, findoffblockdiagonalentries, is);
10860: PetscFunctionReturn(PETSC_SUCCESS);
10861: }
10863: /*@C
10864: MatInvertBlockDiagonal - Inverts the block diagonal entries.
10866: Collective; No Fortran Support
10868: Input Parameter:
10869: . mat - the matrix
10871: Output Parameter:
10872: . values - the block inverses in column major order (FORTRAN-like)
10874: Level: advanced
10876: Notes:
10877: The size of the blocks is determined by the block size of the matrix.
10879: The blocks never overlap between two MPI processes, use `MatInvertVariableBlockEnvelope()` for that case
10881: The blocks all have the same size, use `MatInvertVariableBlockDiagonal()` for variable block size
10883: .seealso: [](ch_matrices), `Mat`, `MatInvertVariableBlockEnvelope()`, `MatInvertBlockDiagonalMat()`
10884: @*/
10885: PetscErrorCode MatInvertBlockDiagonal(Mat mat, const PetscScalar *values[])
10886: {
10887: PetscFunctionBegin;
10889: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10890: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10891: PetscUseTypeMethod(mat, invertblockdiagonal, values);
10892: PetscFunctionReturn(PETSC_SUCCESS);
10893: }
10895: /*@
10896: MatInvertVariableBlockDiagonal - Inverts the point block diagonal entries.
10898: Collective; No Fortran Support
10900: Input Parameters:
10901: + mat - the matrix
10902: . nblocks - the number of blocks on the process, set with `MatSetVariableBlockSizes()`
10903: - bsizes - the size of each block on the process, set with `MatSetVariableBlockSizes()`
10905: Output Parameter:
10906: . values - the block inverses in column major order (FORTRAN-like)
10908: Level: advanced
10910: Notes:
10911: Use `MatInvertBlockDiagonal()` if all blocks have the same size
10913: The blocks never overlap between two MPI processes, use `MatInvertVariableBlockEnvelope()` for that case
10915: .seealso: [](ch_matrices), `Mat`, `MatInvertBlockDiagonal()`, `MatSetVariableBlockSizes()`, `MatInvertVariableBlockEnvelope()`
10916: @*/
10917: PetscErrorCode MatInvertVariableBlockDiagonal(Mat mat, PetscInt nblocks, const PetscInt bsizes[], PetscScalar values[])
10918: {
10919: PetscFunctionBegin;
10921: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10922: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10923: PetscUseTypeMethod(mat, invertvariableblockdiagonal, nblocks, bsizes, values);
10924: PetscFunctionReturn(PETSC_SUCCESS);
10925: }
10927: /*@
10928: MatInvertBlockDiagonalMat - set the values of matrix C to be the inverted block diagonal of matrix A
10930: Collective
10932: Input Parameters:
10933: + A - the matrix
10934: - C - matrix with inverted block diagonal of `A`. This matrix should be created and may have its type set.
10936: Level: advanced
10938: Note:
10939: The blocksize of the matrix is used to determine the blocks on the diagonal of `C`
10941: .seealso: [](ch_matrices), `Mat`, `MatInvertBlockDiagonal()`
10942: @*/
10943: PetscErrorCode MatInvertBlockDiagonalMat(Mat A, Mat C)
10944: {
10945: const PetscScalar *vals;
10946: PetscInt *dnnz;
10947: PetscInt m, rstart, rend, bs, i, j;
10949: PetscFunctionBegin;
10950: PetscCall(MatInvertBlockDiagonal(A, &vals));
10951: PetscCall(MatGetBlockSize(A, &bs));
10952: PetscCall(MatGetLocalSize(A, &m, NULL));
10953: PetscCall(MatSetLayouts(C, A->rmap, A->cmap));
10954: PetscCall(MatSetBlockSizes(C, A->rmap->bs, A->cmap->bs));
10955: PetscCall(PetscMalloc1(m / bs, &dnnz));
10956: for (j = 0; j < m / bs; j++) dnnz[j] = 1;
10957: PetscCall(MatXAIJSetPreallocation(C, bs, dnnz, NULL, NULL, NULL));
10958: PetscCall(PetscFree(dnnz));
10959: PetscCall(MatGetOwnershipRange(C, &rstart, &rend));
10960: PetscCall(MatSetOption(C, MAT_ROW_ORIENTED, PETSC_FALSE));
10961: for (i = rstart / bs; i < rend / bs; i++) PetscCall(MatSetValuesBlocked(C, 1, &i, 1, &i, &vals[(i - rstart / bs) * bs * bs], INSERT_VALUES));
10962: PetscCall(MatSetOption(C, MAT_NO_OFF_PROC_ENTRIES, PETSC_TRUE));
10963: PetscCall(MatAssemblyBegin(C, MAT_FINAL_ASSEMBLY));
10964: PetscCall(MatAssemblyEnd(C, MAT_FINAL_ASSEMBLY));
10965: PetscCall(MatSetOption(C, MAT_NO_OFF_PROC_ENTRIES, PETSC_FALSE));
10966: PetscCall(MatSetOption(C, MAT_ROW_ORIENTED, PETSC_TRUE));
10967: PetscFunctionReturn(PETSC_SUCCESS);
10968: }
10970: /*@
10971: MatTransposeColoringDestroy - Destroys a coloring context for matrix product $C = A*B^T$ that was created
10972: via `MatTransposeColoringCreate()`.
10974: Collective
10976: Input Parameter:
10977: . c - coloring context
10979: Level: intermediate
10981: .seealso: [](ch_matrices), `Mat`, `MatTransposeColoringCreate()`
10982: @*/
10983: PetscErrorCode MatTransposeColoringDestroy(MatTransposeColoring *c)
10984: {
10985: MatTransposeColoring matcolor = *c;
10987: PetscFunctionBegin;
10988: if (!matcolor) PetscFunctionReturn(PETSC_SUCCESS);
10989: if (--((PetscObject)matcolor)->refct > 0) {
10990: matcolor = NULL;
10991: PetscFunctionReturn(PETSC_SUCCESS);
10992: }
10994: PetscCall(PetscFree3(matcolor->ncolumns, matcolor->nrows, matcolor->colorforrow));
10995: PetscCall(PetscFree(matcolor->rows));
10996: PetscCall(PetscFree(matcolor->den2sp));
10997: PetscCall(PetscFree(matcolor->colorforcol));
10998: PetscCall(PetscFree(matcolor->columns));
10999: if (matcolor->brows > 0) PetscCall(PetscFree(matcolor->lstart));
11000: PetscCall(PetscHeaderDestroy(c));
11001: PetscFunctionReturn(PETSC_SUCCESS);
11002: }
11004: /*@
11005: MatTransColoringApplySpToDen - Given a symbolic matrix product $C = A*B^T$ for which
11006: a `MatTransposeColoring` context has been created, computes a dense $B^T$ by applying
11007: `MatTransposeColoring` to sparse `B`.
11009: Collective
11011: Input Parameters:
11012: + coloring - coloring context created with `MatTransposeColoringCreate()`
11013: - B - sparse matrix
11015: Output Parameter:
11016: . Btdense - dense matrix $B^T$
11018: Level: developer
11020: Note:
11021: These are used internally for some implementations of `MatRARt()`
11023: .seealso: [](ch_matrices), `Mat`, `MatTransposeColoringCreate()`, `MatTransposeColoringDestroy()`, `MatTransColoringApplyDenToSp()`
11024: @*/
11025: PetscErrorCode MatTransColoringApplySpToDen(MatTransposeColoring coloring, Mat B, Mat Btdense)
11026: {
11027: PetscFunctionBegin;
11032: PetscCall((*B->ops->transcoloringapplysptoden)(coloring, B, Btdense));
11033: PetscFunctionReturn(PETSC_SUCCESS);
11034: }
11036: /*@
11037: MatTransColoringApplyDenToSp - Given a symbolic matrix product $C_{sp} = A*B^T$ for which
11038: a `MatTransposeColoring` context has been created and a dense matrix $C_{den} = A*B^T_{dense}$
11039: in which `B^T_{dens}` is obtained from `MatTransColoringApplySpToDen()`, recover sparse matrix
11040: $C_{sp}$ from $C_{den}$.
11042: Collective
11044: Input Parameters:
11045: + matcoloring - coloring context created with `MatTransposeColoringCreate()`
11046: - Cden - matrix product of a sparse matrix and a dense matrix Btdense
11048: Output Parameter:
11049: . Csp - sparse matrix
11051: Level: developer
11053: Note:
11054: These are used internally for some implementations of `MatRARt()`
11056: .seealso: [](ch_matrices), `Mat`, `MatTransposeColoringCreate()`, `MatTransposeColoringDestroy()`, `MatTransColoringApplySpToDen()`
11057: @*/
11058: PetscErrorCode MatTransColoringApplyDenToSp(MatTransposeColoring matcoloring, Mat Cden, Mat Csp)
11059: {
11060: PetscFunctionBegin;
11065: PetscCall((*Csp->ops->transcoloringapplydentosp)(matcoloring, Cden, Csp));
11066: PetscCall(MatAssemblyBegin(Csp, MAT_FINAL_ASSEMBLY));
11067: PetscCall(MatAssemblyEnd(Csp, MAT_FINAL_ASSEMBLY));
11068: PetscFunctionReturn(PETSC_SUCCESS);
11069: }
11071: /*@
11072: MatTransposeColoringCreate - Creates a matrix coloring context for the matrix product $C = A*B^T$.
11074: Collective
11076: Input Parameters:
11077: + mat - the matrix product C
11078: - iscoloring - the coloring of the matrix; usually obtained with `MatColoringCreate()` or `DMCreateColoring()`
11080: Output Parameter:
11081: . color - the new coloring context
11083: Level: intermediate
11085: .seealso: [](ch_matrices), `Mat`, `MatTransposeColoringDestroy()`, `MatTransColoringApplySpToDen()`,
11086: `MatTransColoringApplyDenToSp()`
11087: @*/
11088: PetscErrorCode MatTransposeColoringCreate(Mat mat, ISColoring iscoloring, MatTransposeColoring *color)
11089: {
11090: MatTransposeColoring c;
11091: MPI_Comm comm;
11093: PetscFunctionBegin;
11094: PetscAssertPointer(color, 3);
11096: PetscCall(PetscLogEventBegin(MAT_TransposeColoringCreate, mat, 0, 0, 0));
11097: PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
11098: PetscCall(PetscHeaderCreate(c, MAT_TRANSPOSECOLORING_CLASSID, "MatTransposeColoring", "Matrix product C=A*B^T via coloring", "Mat", comm, MatTransposeColoringDestroy, NULL));
11099: c->ctype = iscoloring->ctype;
11100: PetscUseTypeMethod(mat, transposecoloringcreate, iscoloring, c);
11101: *color = c;
11102: PetscCall(PetscLogEventEnd(MAT_TransposeColoringCreate, mat, 0, 0, 0));
11103: PetscFunctionReturn(PETSC_SUCCESS);
11104: }
11106: /*@
11107: MatGetNonzeroState - Returns a 64-bit integer representing the current state of nonzeros in the matrix. If the
11108: matrix has had new nonzero locations added to (or removed from) the matrix since the previous call, the value will be larger.
11110: Not Collective
11112: Input Parameter:
11113: . mat - the matrix
11115: Output Parameter:
11116: . state - the current state
11118: Level: intermediate
11120: Notes:
11121: You can only compare states from two different calls to the SAME matrix, you cannot compare calls between
11122: different matrices
11124: Use `PetscObjectStateGet()` to check for changes to the numerical values in a matrix
11126: Use the result of `PetscObjectGetId()` to compare if a previously checked matrix is the same as the current matrix, do not compare object pointers.
11128: .seealso: [](ch_matrices), `Mat`, `PetscObjectStateGet()`, `PetscObjectGetId()`
11129: @*/
11130: PetscErrorCode MatGetNonzeroState(Mat mat, PetscObjectState *state)
11131: {
11132: PetscFunctionBegin;
11134: *state = mat->nonzerostate;
11135: PetscFunctionReturn(PETSC_SUCCESS);
11136: }
11138: /*@
11139: MatCreateMPIMatConcatenateSeqMat - Creates a single large PETSc matrix by concatenating sequential
11140: matrices from each processor
11142: Collective
11144: Input Parameters:
11145: + comm - the communicators the parallel matrix will live on
11146: . seqmat - the input sequential matrices
11147: . n - number of local columns (or `PETSC_DECIDE`)
11148: - reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
11150: Output Parameter:
11151: . mpimat - the parallel matrix generated
11153: Level: developer
11155: Note:
11156: The number of columns of the matrix in EACH processor MUST be the same.
11158: .seealso: [](ch_matrices), `Mat`
11159: @*/
11160: PetscErrorCode MatCreateMPIMatConcatenateSeqMat(MPI_Comm comm, Mat seqmat, PetscInt n, MatReuse reuse, Mat *mpimat)
11161: {
11162: PetscMPIInt size;
11164: PetscFunctionBegin;
11165: PetscCallMPI(MPI_Comm_size(comm, &size));
11166: if (size == 1) {
11167: if (reuse == MAT_INITIAL_MATRIX) {
11168: PetscCall(MatDuplicate(seqmat, MAT_COPY_VALUES, mpimat));
11169: } else {
11170: PetscCall(MatCopy(seqmat, *mpimat, SAME_NONZERO_PATTERN));
11171: }
11172: PetscFunctionReturn(PETSC_SUCCESS);
11173: }
11175: PetscCheck(reuse != MAT_REUSE_MATRIX || seqmat != *mpimat, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "MAT_REUSE_MATRIX means reuse the matrix passed in as the final argument, not the original matrix");
11177: PetscCall(PetscLogEventBegin(MAT_Merge, seqmat, 0, 0, 0));
11178: PetscCall((*seqmat->ops->creatempimatconcatenateseqmat)(comm, seqmat, n, reuse, mpimat));
11179: PetscCall(PetscLogEventEnd(MAT_Merge, seqmat, 0, 0, 0));
11180: PetscFunctionReturn(PETSC_SUCCESS);
11181: }
11183: /*@
11184: MatSubdomainsCreateCoalesce - Creates index subdomains by coalescing adjacent MPI processes' ownership ranges.
11186: Collective
11188: Input Parameters:
11189: + A - the matrix to create subdomains from
11190: - N - requested number of subdomains
11192: Output Parameters:
11193: + n - number of subdomains resulting on this MPI process
11194: - iss - `IS` list with indices of subdomains on this MPI process
11196: Level: advanced
11198: Note:
11199: The number of subdomains must be smaller than the communicator size
11201: .seealso: [](ch_matrices), `Mat`, `IS`
11202: @*/
11203: PetscErrorCode MatSubdomainsCreateCoalesce(Mat A, PetscInt N, PetscInt *n, IS *iss[])
11204: {
11205: MPI_Comm comm, subcomm;
11206: PetscMPIInt size, rank, color;
11207: PetscInt rstart, rend, k;
11209: PetscFunctionBegin;
11210: PetscCall(PetscObjectGetComm((PetscObject)A, &comm));
11211: PetscCallMPI(MPI_Comm_size(comm, &size));
11212: PetscCallMPI(MPI_Comm_rank(comm, &rank));
11213: PetscCheck(N >= 1 && N < size, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "number of subdomains must be > 0 and < %d, got N = %" PetscInt_FMT, size, N);
11214: *n = 1;
11215: k = size / N + (size % N > 0); /* There are up to k ranks to a color */
11216: color = rank / k;
11217: PetscCallMPI(MPI_Comm_split(comm, color, rank, &subcomm));
11218: PetscCall(PetscMalloc1(1, iss));
11219: PetscCall(MatGetOwnershipRange(A, &rstart, &rend));
11220: PetscCall(ISCreateStride(subcomm, rend - rstart, rstart, 1, iss[0]));
11221: PetscCallMPI(MPI_Comm_free(&subcomm));
11222: PetscFunctionReturn(PETSC_SUCCESS);
11223: }
11225: /*@
11226: MatGalerkin - Constructs the coarse grid problem matrix via Galerkin projection.
11228: If the interpolation and restriction operators are the same, uses `MatPtAP()`.
11229: If they are not the same, uses `MatMatMatMult()`.
11231: Once the coarse grid problem is constructed, correct for interpolation operators
11232: that are not of full rank, which can legitimately happen in the case of non-nested
11233: geometric multigrid.
11235: Input Parameters:
11236: + restrct - restriction operator
11237: . dA - fine grid matrix
11238: . interpolate - interpolation operator
11239: . reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
11240: - fill - expected fill, use `PETSC_DETERMINE` or `PETSC_DETERMINE` if you do not have a good estimate
11242: Output Parameter:
11243: . A - the Galerkin coarse matrix
11245: Options Database Key:
11246: . -pc_mg_galerkin (both|pmat|mat|none) - for what matrices the Galerkin process should be used
11248: Level: developer
11250: Note:
11251: The deprecated `PETSC_DEFAULT` in `fill` also means use the current value
11253: .seealso: [](ch_matrices), `Mat`, `MatPtAP()`, `MatMatMatMult()`
11254: @*/
11255: PetscErrorCode MatGalerkin(Mat restrct, Mat dA, Mat interpolate, MatReuse reuse, PetscReal fill, Mat *A)
11256: {
11257: IS zerorows;
11258: Vec diag;
11260: PetscFunctionBegin;
11261: PetscCheck(reuse != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
11262: /* Construct the coarse grid matrix */
11263: if (interpolate == restrct) {
11264: PetscCall(MatPtAP(dA, interpolate, reuse, fill, A));
11265: } else {
11266: PetscCall(MatMatMatMult(restrct, dA, interpolate, reuse, fill, A));
11267: }
11269: /* If the interpolation matrix is not of full rank, A will have zero rows.
11270: This can legitimately happen in the case of non-nested geometric multigrid.
11271: In that event, we set the rows of the matrix to the rows of the identity,
11272: ignoring the equations (as the RHS will also be zero). */
11274: PetscCall(MatFindZeroRows(*A, &zerorows));
11276: if (zerorows != NULL) { /* if there are any zero rows */
11277: PetscCall(MatCreateVecs(*A, &diag, NULL));
11278: PetscCall(MatGetDiagonal(*A, diag));
11279: PetscCall(VecISSet(diag, zerorows, 1.0));
11280: PetscCall(MatDiagonalSet(*A, diag, INSERT_VALUES));
11281: PetscCall(VecDestroy(&diag));
11282: PetscCall(ISDestroy(&zerorows));
11283: }
11284: PetscFunctionReturn(PETSC_SUCCESS);
11285: }
11287: /*@C
11288: MatSetOperation - Allows user to set a matrix operation for any matrix type
11290: Logically Collective
11292: Input Parameters:
11293: + mat - the matrix
11294: . op - the name of the operation
11295: - f - the function that provides the operation
11297: Level: developer
11299: Example Usage:
11300: .vb
11301: extern PetscErrorCode usermult(Mat, Vec, Vec);
11303: PetscCall(MatCreateXXX(comm, ..., &A));
11304: PetscCall(MatSetOperation(A, MATOP_MULT, (PetscErrorCodeFn *)usermult));
11305: .ve
11307: Notes:
11308: See the file `include/petscmat.h` for a complete list of matrix
11309: operations, which all have the form MATOP_<OPERATION>, where
11310: <OPERATION> is the name (in all capital letters) of the
11311: user interface routine (e.g., `MatMult()` -> `MATOP_MULT`).
11313: All user-provided functions (except for `MATOP_DESTROY`) should have the same calling
11314: sequence as the usual matrix interface routines, since they
11315: are intended to be accessed via the usual matrix interface
11316: routines, e.g.,
11317: .vb
11318: MatMult(Mat, Vec, Vec) -> usermult(Mat, Vec, Vec)
11319: .ve
11321: In particular each function MUST return `PETSC_SUCCESS` on success and
11322: nonzero on failure.
11324: This routine is distinct from `MatShellSetOperation()` in that it can be called on any matrix type.
11326: .seealso: [](ch_matrices), `Mat`, `MatGetOperation()`, `MatCreateShell()`, `MatShellSetContext()`, `MatShellSetOperation()`
11327: @*/
11328: PetscErrorCode MatSetOperation(Mat mat, MatOperation op, PetscErrorCodeFn *f)
11329: {
11330: PetscFunctionBegin;
11332: if (op == MATOP_VIEW && !mat->ops->viewnative && f != (PetscErrorCodeFn *)mat->ops->view) mat->ops->viewnative = mat->ops->view;
11333: (((PetscErrorCodeFn **)mat->ops)[op]) = f;
11334: PetscFunctionReturn(PETSC_SUCCESS);
11335: }
11337: /*@C
11338: MatGetOperation - Gets a matrix operation for any matrix type.
11340: Not Collective
11342: Input Parameters:
11343: + mat - the matrix
11344: - op - the name of the operation
11346: Output Parameter:
11347: . f - the function that provides the operation
11349: Level: developer
11351: Example Usage:
11352: .vb
11353: PetscErrorCode (*usermult)(Mat, Vec, Vec);
11355: MatGetOperation(A, MATOP_MULT, (PetscErrorCodeFn **)&usermult);
11356: .ve
11358: Notes:
11359: See the file `include/petscmat.h` for a complete list of matrix
11360: operations, which all have the form MATOP_<OPERATION>, where
11361: <OPERATION> is the name (in all capital letters) of the
11362: user interface routine (e.g., `MatMult()` -> `MATOP_MULT`).
11364: This routine is distinct from `MatShellGetOperation()` in that it can be called on any matrix type.
11366: .seealso: [](ch_matrices), `Mat`, `MatSetOperation()`, `MatCreateShell()`, `MatShellGetContext()`, `MatShellGetOperation()`
11367: @*/
11368: PetscErrorCode MatGetOperation(Mat mat, MatOperation op, PetscErrorCodeFn **f)
11369: {
11370: PetscFunctionBegin;
11372: *f = (((PetscErrorCodeFn **)mat->ops)[op]);
11373: PetscFunctionReturn(PETSC_SUCCESS);
11374: }
11376: /*@
11377: MatHasOperation - Determines whether the given matrix supports the particular operation.
11379: Not Collective
11381: Input Parameters:
11382: + mat - the matrix
11383: - op - the operation, for example, `MATOP_GET_DIAGONAL`
11385: Output Parameter:
11386: . has - either `PETSC_TRUE` or `PETSC_FALSE`
11388: Level: advanced
11390: Note:
11391: See `MatSetOperation()` for additional discussion on naming convention and usage of `op`.
11393: .seealso: [](ch_matrices), `Mat`, `MatCreateShell()`, `MatGetOperation()`, `MatSetOperation()`
11394: @*/
11395: PetscErrorCode MatHasOperation(Mat mat, MatOperation op, PetscBool *has)
11396: {
11397: PetscFunctionBegin;
11399: PetscAssertPointer(has, 3);
11400: if (mat->ops->hasoperation) {
11401: PetscUseTypeMethod(mat, hasoperation, op, has);
11402: } else {
11403: if (((void **)mat->ops)[op]) *has = PETSC_TRUE;
11404: else {
11405: *has = PETSC_FALSE;
11406: if (op == MATOP_CREATE_SUBMATRIX) {
11407: PetscMPIInt size;
11409: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
11410: if (size == 1) PetscCall(MatHasOperation(mat, MATOP_CREATE_SUBMATRICES, has));
11411: }
11412: }
11413: }
11414: PetscFunctionReturn(PETSC_SUCCESS);
11415: }
11417: /*@
11418: MatHasCongruentLayouts - Determines whether the rows and columns layouts of the matrix are congruent
11420: Collective
11422: Input Parameter:
11423: . mat - the matrix
11425: Output Parameter:
11426: . cong - either `PETSC_TRUE` or `PETSC_FALSE`
11428: Level: beginner
11430: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatSetSizes()`, `PetscLayout`
11431: @*/
11432: PetscErrorCode MatHasCongruentLayouts(Mat mat, PetscBool *cong)
11433: {
11434: PetscFunctionBegin;
11437: PetscAssertPointer(cong, 2);
11438: if (!mat->rmap || !mat->cmap) {
11439: *cong = mat->rmap == mat->cmap ? PETSC_TRUE : PETSC_FALSE;
11440: PetscFunctionReturn(PETSC_SUCCESS);
11441: }
11442: if (mat->congruentlayouts == PETSC_DECIDE) { /* first time we compare rows and cols layouts */
11443: PetscCall(PetscLayoutSetUp(mat->rmap));
11444: PetscCall(PetscLayoutSetUp(mat->cmap));
11445: PetscCall(PetscLayoutCompare(mat->rmap, mat->cmap, cong));
11446: if (*cong) mat->congruentlayouts = 1;
11447: else mat->congruentlayouts = 0;
11448: } else *cong = mat->congruentlayouts ? PETSC_TRUE : PETSC_FALSE;
11449: PetscFunctionReturn(PETSC_SUCCESS);
11450: }
11452: PetscErrorCode MatSetInf(Mat A)
11453: {
11454: PetscFunctionBegin;
11455: PetscUseTypeMethod(A, setinf);
11456: PetscFunctionReturn(PETSC_SUCCESS);
11457: }
11459: /*@
11460: MatCreateGraph - create a scalar matrix (that is a matrix with one vertex for each block vertex in the original matrix), for use in graph algorithms
11461: and possibly removes small values from the graph structure.
11463: Collective
11465: Input Parameters:
11466: + A - the matrix
11467: . sym - `PETSC_TRUE` indicates that the graph should be symmetrized
11468: . scale - `PETSC_TRUE` indicates that the graph edge weights should be symmetrically scaled with the diagonal entry
11469: . filter - filter value - < 0: does nothing; == 0: removes only 0.0 entries; otherwise: removes entries with abs(entries) <= value
11470: . num_idx - size of `index` array
11471: - index - array of block indices to use for graph strength of connection weight
11473: Output Parameter:
11474: . graph - the resulting graph
11476: Level: advanced
11478: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `PCGAMG`
11479: @*/
11480: PetscErrorCode MatCreateGraph(Mat A, PetscBool sym, PetscBool scale, PetscReal filter, PetscInt num_idx, PetscInt index[], Mat *graph)
11481: {
11482: PetscFunctionBegin;
11486: PetscAssertPointer(graph, 7);
11487: PetscCall(PetscLogEventBegin(MAT_CreateGraph, A, 0, 0, 0));
11488: PetscUseTypeMethod(A, creategraph, sym, scale, filter, num_idx, index, graph);
11489: PetscCall(PetscLogEventEnd(MAT_CreateGraph, A, 0, 0, 0));
11490: PetscFunctionReturn(PETSC_SUCCESS);
11491: }
11493: /*@
11494: MatEliminateZeros - eliminate the nondiagonal zero entries in place from the nonzero structure of a sparse `Mat` in place,
11495: meaning the same memory is used for the matrix, and no new memory is allocated.
11497: Collective
11499: Input Parameters:
11500: + A - the matrix
11501: - keep - if for a given row of `A`, the diagonal coefficient is zero, indicates whether it should be left in the structure or eliminated as well
11503: Level: intermediate
11505: Developer Note:
11506: The entries in the sparse matrix data structure are shifted to fill in the unneeded locations in the data. Thus the end
11507: of the arrays in the data structure are unneeded.
11509: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatCreateGraph()`, `MatFilter()`
11510: @*/
11511: PetscErrorCode MatEliminateZeros(Mat A, PetscBool keep)
11512: {
11513: PetscFunctionBegin;
11515: PetscUseTypeMethod(A, eliminatezeros, keep);
11516: PetscFunctionReturn(PETSC_SUCCESS);
11517: }
11519: /*@C
11520: MatGetCurrentMemType - Get the memory location of the matrix
11522: Not Collective, but the result will be the same on all MPI processes
11524: Input Parameter:
11525: . A - the matrix whose memory type we are checking
11527: Output Parameter:
11528: . m - the memory type
11530: Level: intermediate
11532: .seealso: [](ch_matrices), `Mat`, `MatBoundToCPU()`, `PetscMemType`
11533: @*/
11534: PetscErrorCode MatGetCurrentMemType(Mat A, PetscMemType *m)
11535: {
11536: PetscFunctionBegin;
11538: PetscAssertPointer(m, 2);
11539: if (A->ops->getcurrentmemtype) PetscUseTypeMethod(A, getcurrentmemtype, m);
11540: else *m = PETSC_MEMTYPE_HOST;
11541: PetscFunctionReturn(PETSC_SUCCESS);
11542: }