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_MultTransposeAdd, MAT_Solve, MAT_Solves, MAT_SolveAdd, MAT_SolveTranspose, MAT_MatSolve, MAT_MatTrSolve;
19: PetscLogEvent MAT_SolveTransposeAdd, MAT_SOR, MAT_ForwardSolve, MAT_BackwardSolve, MAT_LUFactor, MAT_LUFactorSymbolic;
20: PetscLogEvent MAT_LUFactorNumeric, MAT_CholeskyFactor, MAT_CholeskyFactorSymbolic, MAT_CholeskyFactorNumeric, MAT_ILUFactor;
21: PetscLogEvent MAT_ILUFactorSymbolic, MAT_ICCFactorSymbolic, MAT_Copy, MAT_Convert, MAT_Scale, MAT_AssemblyBegin;
22: PetscLogEvent MAT_QRFactorNumeric, MAT_QRFactorSymbolic, MAT_QRFactor;
23: PetscLogEvent MAT_AssemblyEnd, MAT_SetValues, MAT_GetValues, MAT_GetRow, MAT_GetRowIJ, MAT_CreateSubMats, MAT_GetOrdering, MAT_RedundantMat, MAT_GetSeqNonzeroStructure;
24: PetscLogEvent MAT_IncreaseOverlap, MAT_Partitioning, MAT_PartitioningND, MAT_Coarsen, MAT_ZeroEntries, MAT_Load, MAT_View, MAT_AXPY, MAT_FDColoringCreate;
25: PetscLogEvent MAT_FDColoringSetUp, MAT_FDColoringApply, MAT_Transpose, MAT_FDColoringFunction, MAT_CreateSubMat;
26: PetscLogEvent MAT_TransposeColoringCreate;
27: PetscLogEvent MAT_MatMult, MAT_MatMultSymbolic, MAT_MatMultNumeric;
28: PetscLogEvent MAT_PtAP, MAT_PtAPSymbolic, MAT_PtAPNumeric, MAT_RARt, MAT_RARtSymbolic, MAT_RARtNumeric;
29: PetscLogEvent MAT_MatTransposeMult, MAT_MatTransposeMultSymbolic, MAT_MatTransposeMultNumeric;
30: PetscLogEvent MAT_TransposeMatMult, MAT_TransposeMatMultSymbolic, MAT_TransposeMatMultNumeric;
31: PetscLogEvent MAT_MatMatMult, MAT_MatMatMultSymbolic, MAT_MatMatMultNumeric;
32: PetscLogEvent MAT_MultHermitianTranspose, MAT_MultHermitianTransposeAdd;
33: PetscLogEvent MAT_Getsymtransreduced, MAT_GetBrowsOfAcols;
34: PetscLogEvent MAT_GetBrowsOfAocols, MAT_Getlocalmat, MAT_Getlocalmatcondensed, MAT_Seqstompi, MAT_Seqstompinum, MAT_Seqstompisym;
35: PetscLogEvent MAT_GetMultiProcBlock;
36: PetscLogEvent MAT_CUSPARSECopyToGPU, MAT_CUSPARSECopyFromGPU, MAT_CUSPARSEGenerateTranspose, MAT_CUSPARSESolveAnalysis;
37: PetscLogEvent MAT_HIPSPARSECopyToGPU, MAT_HIPSPARSECopyFromGPU, MAT_HIPSPARSEGenerateTranspose, MAT_HIPSPARSESolveAnalysis;
38: PetscLogEvent MAT_PreallCOO, MAT_SetVCOO;
39: PetscLogEvent MAT_SetValuesBatch;
40: PetscLogEvent MAT_ViennaCLCopyToGPU;
41: PetscLogEvent MAT_CUDACopyToGPU;
42: PetscLogEvent MAT_DenseCopyToGPU, MAT_DenseCopyFromGPU;
43: PetscLogEvent MAT_Merge, MAT_Residual, MAT_SetRandom;
44: PetscLogEvent MAT_FactorFactS, MAT_FactorInvS;
45: PetscLogEvent MATCOLORING_Apply, MATCOLORING_Comm, MATCOLORING_Local, MATCOLORING_ISCreate, MATCOLORING_SetUp, MATCOLORING_Weights;
46: PetscLogEvent MAT_H2Opus_Build, MAT_H2Opus_Compress, MAT_H2Opus_Orthog, MAT_H2Opus_LR;
48: const char *const MatFactorTypes[] = {"NONE", "LU", "CHOLESKY", "ILU", "ICC", "ILUDT", "QR", "MatFactorType", "MAT_FACTOR_", NULL};
50: /*@
51: MatSetRandom - Sets all components of a matrix to random numbers.
53: Logically Collective
55: Input Parameters:
56: + x - the matrix
57: - rctx - the `PetscRandom` object, formed by `PetscRandomCreate()`, or `NULL` and
58: it will create one internally.
60: Example:
61: .vb
62: PetscRandomCreate(PETSC_COMM_WORLD,&rctx);
63: MatSetRandom(x,rctx);
64: PetscRandomDestroy(rctx);
65: .ve
67: Level: intermediate
69: Notes:
70: For sparse matrices that have been preallocated but not been assembled, it randomly selects appropriate locations,
72: for sparse matrices that already have nonzero locations, it fills the locations with random numbers.
74: It generates an error if used on unassembled sparse matrices that have not been preallocated.
76: .seealso: [](ch_matrices), `Mat`, `PetscRandom`, `PetscRandomCreate()`, `MatZeroEntries()`, `MatSetValues()`, `PetscRandomDestroy()`
77: @*/
78: PetscErrorCode MatSetRandom(Mat x, PetscRandom rctx)
79: {
80: PetscRandom randObj = NULL;
82: PetscFunctionBegin;
86: MatCheckPreallocated(x, 1);
88: if (!rctx) {
89: MPI_Comm comm;
90: PetscCall(PetscObjectGetComm((PetscObject)x, &comm));
91: PetscCall(PetscRandomCreate(comm, &randObj));
92: PetscCall(PetscRandomSetType(randObj, x->defaultrandtype));
93: PetscCall(PetscRandomSetFromOptions(randObj));
94: rctx = randObj;
95: }
96: PetscCall(PetscLogEventBegin(MAT_SetRandom, x, rctx, 0, 0));
97: PetscUseTypeMethod(x, setrandom, rctx);
98: PetscCall(PetscLogEventEnd(MAT_SetRandom, x, rctx, 0, 0));
100: PetscCall(MatAssemblyBegin(x, MAT_FINAL_ASSEMBLY));
101: PetscCall(MatAssemblyEnd(x, MAT_FINAL_ASSEMBLY));
102: PetscCall(PetscRandomDestroy(&randObj));
103: PetscFunctionReturn(PETSC_SUCCESS);
104: }
106: /*@
107: MatFactorGetErrorZeroPivot - returns the pivot value that was determined to be zero and the row it occurred in
109: Logically Collective
111: Input Parameter:
112: . mat - the factored matrix
114: Output Parameters:
115: + pivot - the pivot value computed
116: - row - the row that the zero pivot occurred. This row value must be interpreted carefully due to row reorderings and which processes
117: the share the matrix
119: Level: advanced
121: Notes:
122: This routine does not work for factorizations done with external packages.
124: This routine should only be called if `MatGetFactorError()` returns a value of `MAT_FACTOR_NUMERIC_ZEROPIVOT`
126: This can also be called on non-factored matrices that come from, for example, matrices used in SOR.
128: .seealso: [](ch_matrices), `Mat`, `MatZeroEntries()`, `MatFactor()`, `MatGetFactor()`,
129: `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`, `MatFactorClearError()`,
130: `MAT_FACTOR_NUMERIC_ZEROPIVOT`
131: @*/
132: PetscErrorCode MatFactorGetErrorZeroPivot(Mat mat, PetscReal *pivot, PetscInt *row)
133: {
134: PetscFunctionBegin;
136: PetscAssertPointer(pivot, 2);
137: PetscAssertPointer(row, 3);
138: *pivot = mat->factorerror_zeropivot_value;
139: *row = mat->factorerror_zeropivot_row;
140: PetscFunctionReturn(PETSC_SUCCESS);
141: }
143: /*@
144: MatFactorGetError - gets the error code from a factorization
146: Logically Collective
148: Input Parameter:
149: . mat - the factored matrix
151: Output Parameter:
152: . err - the error code
154: Level: advanced
156: Note:
157: This can also be called on non-factored matrices that come from, for example, matrices used in SOR.
159: .seealso: [](ch_matrices), `Mat`, `MatZeroEntries()`, `MatFactor()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`,
160: `MatFactorClearError()`, `MatFactorGetErrorZeroPivot()`, `MatFactorError`
161: @*/
162: PetscErrorCode MatFactorGetError(Mat mat, MatFactorError *err)
163: {
164: PetscFunctionBegin;
166: PetscAssertPointer(err, 2);
167: *err = mat->factorerrortype;
168: PetscFunctionReturn(PETSC_SUCCESS);
169: }
171: /*@
172: MatFactorClearError - clears the error code in a factorization
174: Logically Collective
176: Input Parameter:
177: . mat - the factored matrix
179: Level: developer
181: Note:
182: This can also be called on non-factored matrices that come from, for example, matrices used in SOR.
184: .seealso: [](ch_matrices), `Mat`, `MatZeroEntries()`, `MatFactor()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`, `MatFactorGetError()`, `MatFactorGetErrorZeroPivot()`,
185: `MatGetErrorCode()`, `MatFactorError`
186: @*/
187: PetscErrorCode MatFactorClearError(Mat mat)
188: {
189: PetscFunctionBegin;
191: mat->factorerrortype = MAT_FACTOR_NOERROR;
192: mat->factorerror_zeropivot_value = 0.0;
193: mat->factorerror_zeropivot_row = 0;
194: PetscFunctionReturn(PETSC_SUCCESS);
195: }
197: PETSC_INTERN PetscErrorCode MatFindNonzeroRowsOrCols_Basic(Mat mat, PetscBool cols, PetscReal tol, IS *nonzero)
198: {
199: Vec r, l;
200: const PetscScalar *al;
201: PetscInt i, nz, gnz, N, n;
203: PetscFunctionBegin;
204: PetscCall(MatCreateVecs(mat, &r, &l));
205: if (!cols) { /* nonzero rows */
206: PetscCall(MatGetSize(mat, &N, NULL));
207: PetscCall(MatGetLocalSize(mat, &n, NULL));
208: PetscCall(VecSet(l, 0.0));
209: PetscCall(VecSetRandom(r, NULL));
210: PetscCall(MatMult(mat, r, l));
211: PetscCall(VecGetArrayRead(l, &al));
212: } else { /* nonzero columns */
213: PetscCall(MatGetSize(mat, NULL, &N));
214: PetscCall(MatGetLocalSize(mat, NULL, &n));
215: PetscCall(VecSet(r, 0.0));
216: PetscCall(VecSetRandom(l, NULL));
217: PetscCall(MatMultTranspose(mat, l, r));
218: PetscCall(VecGetArrayRead(r, &al));
219: }
220: if (tol <= 0.0) {
221: for (i = 0, nz = 0; i < n; i++)
222: if (al[i] != 0.0) nz++;
223: } else {
224: for (i = 0, nz = 0; i < n; i++)
225: if (PetscAbsScalar(al[i]) > tol) nz++;
226: }
227: PetscCall(MPIU_Allreduce(&nz, &gnz, 1, MPIU_INT, MPI_SUM, PetscObjectComm((PetscObject)mat)));
228: if (gnz != N) {
229: PetscInt *nzr;
230: PetscCall(PetscMalloc1(nz, &nzr));
231: if (nz) {
232: if (tol < 0) {
233: for (i = 0, nz = 0; i < n; i++)
234: if (al[i] != 0.0) nzr[nz++] = i;
235: } else {
236: for (i = 0, nz = 0; i < n; i++)
237: if (PetscAbsScalar(al[i]) > tol) nzr[nz++] = i;
238: }
239: }
240: PetscCall(ISCreateGeneral(PetscObjectComm((PetscObject)mat), nz, nzr, PETSC_OWN_POINTER, nonzero));
241: } else *nonzero = NULL;
242: if (!cols) { /* nonzero rows */
243: PetscCall(VecRestoreArrayRead(l, &al));
244: } else {
245: PetscCall(VecRestoreArrayRead(r, &al));
246: }
247: PetscCall(VecDestroy(&l));
248: PetscCall(VecDestroy(&r));
249: PetscFunctionReturn(PETSC_SUCCESS);
250: }
252: /*@
253: MatFindNonzeroRows - Locate all rows that are not completely zero in the matrix
255: Input Parameter:
256: . mat - the matrix
258: Output Parameter:
259: . keptrows - the rows that are not completely zero
261: Level: intermediate
263: Note:
264: `keptrows` is set to `NULL` if all rows are nonzero.
266: .seealso: [](ch_matrices), `Mat`, `MatFindZeroRows()`
267: @*/
268: PetscErrorCode MatFindNonzeroRows(Mat mat, IS *keptrows)
269: {
270: PetscFunctionBegin;
273: PetscAssertPointer(keptrows, 2);
274: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
275: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
276: if (mat->ops->findnonzerorows) PetscUseTypeMethod(mat, findnonzerorows, keptrows);
277: else PetscCall(MatFindNonzeroRowsOrCols_Basic(mat, PETSC_FALSE, 0.0, keptrows));
278: PetscFunctionReturn(PETSC_SUCCESS);
279: }
281: /*@
282: MatFindZeroRows - Locate all rows that are completely zero in the matrix
284: Input Parameter:
285: . mat - the matrix
287: Output Parameter:
288: . zerorows - the rows that are completely zero
290: Level: intermediate
292: Note:
293: `zerorows` is set to `NULL` if no rows are zero.
295: .seealso: [](ch_matrices), `Mat`, `MatFindNonzeroRows()`
296: @*/
297: PetscErrorCode MatFindZeroRows(Mat mat, IS *zerorows)
298: {
299: IS keptrows;
300: PetscInt m, n;
302: PetscFunctionBegin;
305: PetscAssertPointer(zerorows, 2);
306: PetscCall(MatFindNonzeroRows(mat, &keptrows));
307: /* MatFindNonzeroRows sets keptrows to NULL if there are no zero rows.
308: In keeping with this convention, we set zerorows to NULL if there are no zero
309: rows. */
310: if (keptrows == NULL) {
311: *zerorows = NULL;
312: } else {
313: PetscCall(MatGetOwnershipRange(mat, &m, &n));
314: PetscCall(ISComplement(keptrows, m, n, zerorows));
315: PetscCall(ISDestroy(&keptrows));
316: }
317: PetscFunctionReturn(PETSC_SUCCESS);
318: }
320: /*@
321: MatGetDiagonalBlock - Returns the part of the matrix associated with the on-process coupling
323: Not Collective
325: Input Parameter:
326: . A - the matrix
328: Output Parameter:
329: . a - the diagonal part (which is a SEQUENTIAL matrix)
331: Level: advanced
333: Notes:
334: See `MatCreateAIJ()` for more information on the "diagonal part" of the matrix.
336: Use caution, as the reference count on the returned matrix is not incremented and it is used as part of `A`'s normal operation.
338: .seealso: [](ch_matrices), `Mat`, `MatCreateAIJ()`, `MATAIJ`, `MATBAIJ`, `MATSBAIJ`
339: @*/
340: PetscErrorCode MatGetDiagonalBlock(Mat A, Mat *a)
341: {
342: PetscFunctionBegin;
345: PetscAssertPointer(a, 2);
346: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
347: if (A->ops->getdiagonalblock) PetscUseTypeMethod(A, getdiagonalblock, a);
348: else {
349: PetscMPIInt size;
351: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)A), &size));
352: PetscCheck(size == 1, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Not for parallel matrix type %s", ((PetscObject)A)->type_name);
353: *a = A;
354: }
355: PetscFunctionReturn(PETSC_SUCCESS);
356: }
358: /*@
359: MatGetTrace - Gets the trace of a matrix. The sum of the diagonal entries.
361: Collective
363: Input Parameter:
364: . mat - the matrix
366: Output Parameter:
367: . trace - the sum of the diagonal entries
369: Level: advanced
371: .seealso: [](ch_matrices), `Mat`
372: @*/
373: PetscErrorCode MatGetTrace(Mat mat, PetscScalar *trace)
374: {
375: Vec diag;
377: PetscFunctionBegin;
379: PetscAssertPointer(trace, 2);
380: PetscCall(MatCreateVecs(mat, &diag, NULL));
381: PetscCall(MatGetDiagonal(mat, diag));
382: PetscCall(VecSum(diag, trace));
383: PetscCall(VecDestroy(&diag));
384: PetscFunctionReturn(PETSC_SUCCESS);
385: }
387: /*@
388: MatRealPart - Zeros out the imaginary part of the matrix
390: Logically Collective
392: Input Parameter:
393: . mat - the matrix
395: Level: advanced
397: .seealso: [](ch_matrices), `Mat`, `MatImaginaryPart()`
398: @*/
399: PetscErrorCode MatRealPart(Mat mat)
400: {
401: PetscFunctionBegin;
404: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
405: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
406: MatCheckPreallocated(mat, 1);
407: PetscUseTypeMethod(mat, realpart);
408: PetscFunctionReturn(PETSC_SUCCESS);
409: }
411: /*@C
412: MatGetGhosts - Get the global indices of all ghost nodes defined by the sparse matrix
414: Collective
416: Input Parameter:
417: . mat - the matrix
419: Output Parameters:
420: + nghosts - number of ghosts (for `MATBAIJ` and `MATSBAIJ` matrices there is one ghost for each matrix block)
421: - ghosts - the global indices of the ghost points
423: Level: advanced
425: Note:
426: `nghosts` and `ghosts` are suitable to pass into `VecCreateGhost()` or `VecCreateGhostBlock()`
428: .seealso: [](ch_matrices), `Mat`, `VecCreateGhost()`, `VecCreateGhostBlock()`
429: @*/
430: PetscErrorCode MatGetGhosts(Mat mat, PetscInt *nghosts, const PetscInt *ghosts[])
431: {
432: PetscFunctionBegin;
435: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
436: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
437: if (mat->ops->getghosts) PetscUseTypeMethod(mat, getghosts, nghosts, ghosts);
438: else {
439: if (nghosts) *nghosts = 0;
440: if (ghosts) *ghosts = NULL;
441: }
442: PetscFunctionReturn(PETSC_SUCCESS);
443: }
445: /*@
446: MatImaginaryPart - Moves the imaginary part of the matrix to the real part and zeros the imaginary part
448: Logically Collective
450: Input Parameter:
451: . mat - the matrix
453: Level: advanced
455: .seealso: [](ch_matrices), `Mat`, `MatRealPart()`
456: @*/
457: PetscErrorCode MatImaginaryPart(Mat mat)
458: {
459: PetscFunctionBegin;
462: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
463: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
464: MatCheckPreallocated(mat, 1);
465: PetscUseTypeMethod(mat, imaginarypart);
466: PetscFunctionReturn(PETSC_SUCCESS);
467: }
469: /*@
470: MatMissingDiagonal - Determine if sparse matrix is missing a diagonal entry (or block entry for `MATBAIJ` and `MATSBAIJ` matrices) in the nonzero structure
472: Not Collective
474: Input Parameter:
475: . mat - the matrix
477: Output Parameters:
478: + missing - is any diagonal entry missing
479: - dd - first diagonal entry that is missing (optional) on this process
481: Level: advanced
483: Note:
484: This does not return diagonal entries that are in the nonzero structure but happen to have a zero numerical value
486: .seealso: [](ch_matrices), `Mat`
487: @*/
488: PetscErrorCode MatMissingDiagonal(Mat mat, PetscBool *missing, PetscInt *dd)
489: {
490: PetscFunctionBegin;
493: PetscAssertPointer(missing, 2);
494: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix %s", ((PetscObject)mat)->type_name);
495: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
496: PetscUseTypeMethod(mat, missingdiagonal, missing, dd);
497: PetscFunctionReturn(PETSC_SUCCESS);
498: }
500: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
501: /*@C
502: MatGetRow - Gets a row of a matrix. You MUST call `MatRestoreRow()`
503: for each row that you get to ensure that your application does
504: not bleed memory.
506: Not Collective
508: Input Parameters:
509: + mat - the matrix
510: - row - the row to get
512: Output Parameters:
513: + ncols - if not `NULL`, the number of nonzeros in `row`
514: . cols - if not `NULL`, the column numbers
515: - vals - if not `NULL`, the numerical values
517: Level: advanced
519: Notes:
520: This routine is provided for people who need to have direct access
521: to the structure of a matrix. We hope that we provide enough
522: high-level matrix routines that few users will need it.
524: `MatGetRow()` always returns 0-based column indices, regardless of
525: whether the internal representation is 0-based (default) or 1-based.
527: For better efficiency, set `cols` and/or `vals` to `NULL` if you do
528: not wish to extract these quantities.
530: The user can only examine the values extracted with `MatGetRow()`;
531: the values CANNOT be altered. To change the matrix entries, one
532: must use `MatSetValues()`.
534: You can only have one call to `MatGetRow()` outstanding for a particular
535: matrix at a time, per processor. `MatGetRow()` can only obtain rows
536: associated with the given processor, it cannot get rows from the
537: other processors; for that we suggest using `MatCreateSubMatrices()`, then
538: `MatGetRow()` on the submatrix. The row index passed to `MatGetRow()`
539: is in the global number of rows.
541: Use `MatGetRowIJ()` and `MatRestoreRowIJ()` to access all the local indices of the sparse matrix.
543: Use `MatSeqAIJGetArray()` and similar functions to access the numerical values for certain matrix types directly.
545: Fortran Note:
546: The calling sequence is
547: .vb
548: MatGetRow(matrix,row,ncols,cols,values,ierr)
549: Mat matrix (input)
550: integer row (input)
551: integer ncols (output)
552: integer cols(maxcols) (output)
553: double precision (or double complex) values(maxcols) output
554: .ve
555: where maxcols >= maximum nonzeros in any row of the matrix.
557: .seealso: [](ch_matrices), `Mat`, `MatRestoreRow()`, `MatSetValues()`, `MatGetValues()`, `MatCreateSubMatrices()`, `MatGetDiagonal()`, `MatGetRowIJ()`, `MatRestoreRowIJ()`
558: @*/
559: PetscErrorCode MatGetRow(Mat mat, PetscInt row, PetscInt *ncols, const PetscInt *cols[], const PetscScalar *vals[])
560: {
561: PetscInt incols;
563: PetscFunctionBegin;
566: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
567: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
568: MatCheckPreallocated(mat, 1);
569: 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);
570: PetscCall(PetscLogEventBegin(MAT_GetRow, mat, 0, 0, 0));
571: PetscUseTypeMethod(mat, getrow, row, &incols, (PetscInt **)cols, (PetscScalar **)vals);
572: if (ncols) *ncols = incols;
573: PetscCall(PetscLogEventEnd(MAT_GetRow, mat, 0, 0, 0));
574: PetscFunctionReturn(PETSC_SUCCESS);
575: }
577: /*@
578: MatConjugate - replaces the matrix values with their complex conjugates
580: Logically Collective
582: Input Parameter:
583: . mat - the matrix
585: Level: advanced
587: .seealso: [](ch_matrices), `Mat`, `MatRealPart()`, `MatImaginaryPart()`, `VecConjugate()`, `MatTranspose()`
588: @*/
589: PetscErrorCode MatConjugate(Mat mat)
590: {
591: PetscFunctionBegin;
593: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
594: if (PetscDefined(USE_COMPLEX) && mat->hermitian != PETSC_BOOL3_TRUE) {
595: PetscUseTypeMethod(mat, conjugate);
596: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
597: }
598: PetscFunctionReturn(PETSC_SUCCESS);
599: }
601: /*@C
602: MatRestoreRow - Frees any temporary space allocated by `MatGetRow()`.
604: Not Collective
606: Input Parameters:
607: + mat - the matrix
608: . row - the row to get
609: . ncols - the number of nonzeros
610: . cols - the columns of the nonzeros
611: - vals - if nonzero the column values
613: Level: advanced
615: Notes:
616: This routine should be called after you have finished examining the entries.
618: This routine zeros out `ncols`, `cols`, and `vals`. This is to prevent accidental
619: us of the array after it has been restored. If you pass `NULL`, it will
620: not zero the pointers. Use of `cols` or `vals` after `MatRestoreRow()` is invalid.
622: Fortran Notes:
623: The calling sequence is
624: .vb
625: MatRestoreRow(matrix,row,ncols,cols,values,ierr)
626: Mat matrix (input)
627: integer row (input)
628: integer ncols (output)
629: integer cols(maxcols) (output)
630: double precision (or double complex) values(maxcols) output
631: .ve
632: Where maxcols >= maximum nonzeros in any row of the matrix.
634: In Fortran `MatRestoreRow()` MUST be called after `MatGetRow()`
635: before another call to `MatGetRow()` can be made.
637: .seealso: [](ch_matrices), `Mat`, `MatGetRow()`
638: @*/
639: PetscErrorCode MatRestoreRow(Mat mat, PetscInt row, PetscInt *ncols, const PetscInt *cols[], const PetscScalar *vals[])
640: {
641: PetscFunctionBegin;
643: if (ncols) PetscAssertPointer(ncols, 3);
644: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
645: if (!mat->ops->restorerow) PetscFunctionReturn(PETSC_SUCCESS);
646: PetscUseTypeMethod(mat, restorerow, row, ncols, (PetscInt **)cols, (PetscScalar **)vals);
647: if (ncols) *ncols = 0;
648: if (cols) *cols = NULL;
649: if (vals) *vals = NULL;
650: PetscFunctionReturn(PETSC_SUCCESS);
651: }
653: /*@
654: MatGetRowUpperTriangular - Sets a flag to enable calls to `MatGetRow()` for matrix in `MATSBAIJ` format.
655: You should call `MatRestoreRowUpperTriangular()` after calling` MatGetRow()` and `MatRestoreRow()` to disable the flag.
657: Not Collective
659: Input Parameter:
660: . mat - the matrix
662: Level: advanced
664: Note:
665: 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.
667: .seealso: [](ch_matrices), `Mat`, `MATSBAIJ`, `MatRestoreRowUpperTriangular()`
668: @*/
669: PetscErrorCode MatGetRowUpperTriangular(Mat mat)
670: {
671: PetscFunctionBegin;
674: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
675: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
676: MatCheckPreallocated(mat, 1);
677: if (!mat->ops->getrowuppertriangular) PetscFunctionReturn(PETSC_SUCCESS);
678: PetscUseTypeMethod(mat, getrowuppertriangular);
679: PetscFunctionReturn(PETSC_SUCCESS);
680: }
682: /*@
683: MatRestoreRowUpperTriangular - Disable calls to `MatGetRow()` for matrix in `MATSBAIJ` format.
685: Not Collective
687: Input Parameter:
688: . mat - the matrix
690: Level: advanced
692: Note:
693: This routine should be called after you have finished calls to `MatGetRow()` and `MatRestoreRow()`.
695: .seealso: [](ch_matrices), `Mat`, `MATSBAIJ`, `MatGetRowUpperTriangular()`
696: @*/
697: PetscErrorCode MatRestoreRowUpperTriangular(Mat mat)
698: {
699: PetscFunctionBegin;
702: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
703: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
704: MatCheckPreallocated(mat, 1);
705: if (!mat->ops->restorerowuppertriangular) PetscFunctionReturn(PETSC_SUCCESS);
706: PetscUseTypeMethod(mat, restorerowuppertriangular);
707: PetscFunctionReturn(PETSC_SUCCESS);
708: }
710: /*@C
711: MatSetOptionsPrefix - Sets the prefix used for searching for all
712: `Mat` options in the database.
714: Logically Collective
716: Input Parameters:
717: + A - the matrix
718: - prefix - the prefix to prepend to all option names
720: Level: advanced
722: Notes:
723: A hyphen (-) must NOT be given at the beginning of the prefix name.
724: The first character of all runtime options is AUTOMATICALLY the hyphen.
726: This is NOT used for options for the factorization of the matrix. Normally the
727: prefix is automatically passed in from the PC calling the factorization. To set
728: it directly use `MatSetOptionsPrefixFactor()`
730: .seealso: [](ch_matrices), `Mat`, `MatSetFromOptions()`, `MatSetOptionsPrefixFactor()`
731: @*/
732: PetscErrorCode MatSetOptionsPrefix(Mat A, const char prefix[])
733: {
734: PetscFunctionBegin;
736: PetscCall(PetscObjectSetOptionsPrefix((PetscObject)A, prefix));
737: PetscFunctionReturn(PETSC_SUCCESS);
738: }
740: /*@C
741: MatSetOptionsPrefixFactor - Sets the prefix used for searching for all matrix factor options in the database for
742: for matrices created with `MatGetFactor()`
744: Logically Collective
746: Input Parameters:
747: + A - the matrix
748: - prefix - the prefix to prepend to all option names for the factored matrix
750: Level: developer
752: Notes:
753: A hyphen (-) must NOT be given at the beginning of the prefix name.
754: The first character of all runtime options is AUTOMATICALLY the hyphen.
756: Normally the prefix is automatically passed in from the `PC` calling the factorization. To set
757: it directly when not using `KSP`/`PC` use `MatSetOptionsPrefixFactor()`
759: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSetFromOptions()`, `MatSetOptionsPrefix()`, `MatAppendOptionsPrefixFactor()`
760: @*/
761: PetscErrorCode MatSetOptionsPrefixFactor(Mat A, const char prefix[])
762: {
763: PetscFunctionBegin;
765: if (prefix) {
766: PetscAssertPointer(prefix, 2);
767: PetscCheck(prefix[0] != '-', PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONG, "Options prefix should not begin with a hyphen");
768: if (prefix != A->factorprefix) {
769: PetscCall(PetscFree(A->factorprefix));
770: PetscCall(PetscStrallocpy(prefix, &A->factorprefix));
771: }
772: } else PetscCall(PetscFree(A->factorprefix));
773: PetscFunctionReturn(PETSC_SUCCESS);
774: }
776: /*@C
777: MatAppendOptionsPrefixFactor - Appends to the prefix used for searching for all matrix factor options in the database for
778: for matrices created with `MatGetFactor()`
780: Logically Collective
782: Input Parameters:
783: + A - the matrix
784: - prefix - the prefix to prepend to all option names for the factored matrix
786: Level: developer
788: Notes:
789: A hyphen (-) must NOT be given at the beginning of the prefix name.
790: The first character of all runtime options is AUTOMATICALLY the hyphen.
792: Normally the prefix is automatically passed in from the `PC` calling the factorization. To set
793: it directly when not using `KSP`/`PC` use `MatAppendOptionsPrefixFactor()`
795: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `PetscOptionsCreate()`, `PetscOptionsDestroy()`, `PetscObjectSetOptionsPrefix()`, `PetscObjectPrependOptionsPrefix()`,
796: `PetscObjectGetOptionsPrefix()`, `TSAppendOptionsPrefix()`, `SNESAppendOptionsPrefix()`, `KSPAppendOptionsPrefix()`, `MatSetOptionsPrefixFactor()`,
797: `MatSetOptionsPrefix()`
798: @*/
799: PetscErrorCode MatAppendOptionsPrefixFactor(Mat A, const char prefix[])
800: {
801: size_t len1, len2, new_len;
803: PetscFunctionBegin;
805: if (!prefix) PetscFunctionReturn(PETSC_SUCCESS);
806: if (!A->factorprefix) {
807: PetscCall(MatSetOptionsPrefixFactor(A, prefix));
808: PetscFunctionReturn(PETSC_SUCCESS);
809: }
810: PetscCheck(prefix[0] != '-', PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONG, "Options prefix should not begin with a hyphen");
812: PetscCall(PetscStrlen(A->factorprefix, &len1));
813: PetscCall(PetscStrlen(prefix, &len2));
814: new_len = len1 + len2 + 1;
815: PetscCall(PetscRealloc(new_len * sizeof(*(A->factorprefix)), &A->factorprefix));
816: PetscCall(PetscStrncpy(A->factorprefix + len1, prefix, len2 + 1));
817: PetscFunctionReturn(PETSC_SUCCESS);
818: }
820: /*@C
821: MatAppendOptionsPrefix - Appends to the prefix used for searching for all
822: matrix options in the database.
824: Logically Collective
826: Input Parameters:
827: + A - the matrix
828: - prefix - the prefix to prepend to all option names
830: Level: advanced
832: Note:
833: A hyphen (-) must NOT be given at the beginning of the prefix name.
834: The first character of all runtime options is AUTOMATICALLY the hyphen.
836: .seealso: [](ch_matrices), `Mat`, `MatGetOptionsPrefix()`, `MatAppendOptionsPrefixFactor()`, `MatSetOptionsPrefix()`
837: @*/
838: PetscErrorCode MatAppendOptionsPrefix(Mat A, const char prefix[])
839: {
840: PetscFunctionBegin;
842: PetscCall(PetscObjectAppendOptionsPrefix((PetscObject)A, prefix));
843: PetscFunctionReturn(PETSC_SUCCESS);
844: }
846: /*@C
847: MatGetOptionsPrefix - Gets the prefix used for searching for all
848: matrix options in the database.
850: Not Collective
852: Input Parameter:
853: . A - the matrix
855: Output Parameter:
856: . prefix - pointer to the prefix string used
858: Level: advanced
860: Fortran Note:
861: The user should pass in a string `prefix` of
862: sufficient length to hold the prefix.
864: .seealso: [](ch_matrices), `Mat`, `MatAppendOptionsPrefix()`, `MatSetOptionsPrefix()`, `MatAppendOptionsPrefixFactor()`, `MatSetOptionsPrefixFactor()`
865: @*/
866: PetscErrorCode MatGetOptionsPrefix(Mat A, const char *prefix[])
867: {
868: PetscFunctionBegin;
870: PetscAssertPointer(prefix, 2);
871: PetscCall(PetscObjectGetOptionsPrefix((PetscObject)A, prefix));
872: PetscFunctionReturn(PETSC_SUCCESS);
873: }
875: /*@
876: MatResetPreallocation - Reset matrix to use the original nonzero pattern provided by the user.
878: Collective
880: Input Parameter:
881: . A - the matrix
883: Level: beginner
885: Notes:
886: The allocated memory will be shrunk after calling `MatAssemblyBegin()` and `MatAssemblyEnd()` with `MAT_FINAL_ASSEMBLY`.
888: Users can reset the preallocation to access the original memory.
890: Currently only supported for `MATAIJ` matrices.
892: .seealso: [](ch_matrices), `Mat`, `MatSeqAIJSetPreallocation()`, `MatMPIAIJSetPreallocation()`, `MatXAIJSetPreallocation()`
893: @*/
894: PetscErrorCode MatResetPreallocation(Mat A)
895: {
896: PetscFunctionBegin;
899: PetscCheck(A->insertmode == NOT_SET_VALUES, PETSC_COMM_SELF, PETSC_ERR_SUP, "Cannot reset preallocation after setting some values but not yet calling MatAssemblyBegin()/MatAsssemblyEnd()");
900: if (A->num_ass == 0) PetscFunctionReturn(PETSC_SUCCESS);
901: PetscUseMethod(A, "MatResetPreallocation_C", (Mat), (A));
902: PetscFunctionReturn(PETSC_SUCCESS);
903: }
905: /*@
906: MatSetUp - Sets up the internal matrix data structures for later use.
908: Collective
910: Input Parameter:
911: . A - the matrix
913: Level: intermediate
915: Notes:
916: If the user has not set preallocation for this matrix then an efficient algorithm will be used for the first round of
917: setting values in the matrix.
919: This routine is called internally by other matrix functions when needed so rarely needs to be called by users
921: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatCreate()`, `MatDestroy()`, `MatXAIJSetPreallocation()`
922: @*/
923: PetscErrorCode MatSetUp(Mat A)
924: {
925: PetscFunctionBegin;
927: if (!((PetscObject)A)->type_name) {
928: PetscMPIInt size;
930: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)A), &size));
931: PetscCall(MatSetType(A, size == 1 ? MATSEQAIJ : MATMPIAIJ));
932: }
933: if (!A->preallocated) PetscTryTypeMethod(A, setup);
934: PetscCall(PetscLayoutSetUp(A->rmap));
935: PetscCall(PetscLayoutSetUp(A->cmap));
936: A->preallocated = PETSC_TRUE;
937: PetscFunctionReturn(PETSC_SUCCESS);
938: }
940: #if defined(PETSC_HAVE_SAWS)
941: #include <petscviewersaws.h>
942: #endif
944: /*
945: If threadsafety is on extraneous matrices may be printed
947: This flag cannot be stored in the matrix because the original matrix in MatView() may assemble a new matrix which is passed into MatViewFromOptions()
948: */
949: #if !defined(PETSC_HAVE_THREADSAFETY)
950: static PetscInt insidematview = 0;
951: #endif
953: /*@C
954: MatViewFromOptions - View properties of the matrix based on options set in the options database
956: Collective
958: Input Parameters:
959: + A - the matrix
960: . obj - optional additional object that provides the options prefix to use
961: - name - command line option
963: Options Database Key:
964: . -mat_view [viewertype]:... - the viewer and its options
966: Level: intermediate
968: Note:
969: .vb
970: If no value is provided ascii:stdout is used
971: ascii[:[filename][:[format][:append]]] defaults to stdout - format can be one of ascii_info, ascii_info_detail, or ascii_matlab,
972: for example ascii::ascii_info prints just the information about the object not all details
973: unless :append is given filename opens in write mode, overwriting what was already there
974: binary[:[filename][:[format][:append]]] defaults to the file binaryoutput
975: draw[:drawtype[:filename]] for example, draw:tikz, draw:tikz:figure.tex or draw:x
976: socket[:port] defaults to the standard output port
977: saws[:communicatorname] publishes object to the Scientific Application Webserver (SAWs)
978: .ve
980: .seealso: [](ch_matrices), `Mat`, `MatView()`, `PetscObjectViewFromOptions()`, `MatCreate()`
981: @*/
982: PetscErrorCode MatViewFromOptions(Mat A, PetscObject obj, const char name[])
983: {
984: PetscFunctionBegin;
986: #if !defined(PETSC_HAVE_THREADSAFETY)
987: if (insidematview) PetscFunctionReturn(PETSC_SUCCESS);
988: #endif
989: PetscCall(PetscObjectViewFromOptions((PetscObject)A, obj, name));
990: PetscFunctionReturn(PETSC_SUCCESS);
991: }
993: /*@C
994: MatView - display information about a matrix in a variety ways
996: Collective
998: Input Parameters:
999: + mat - the matrix
1000: - viewer - visualization context
1002: Options Database Keys:
1003: + -mat_view ::ascii_info - Prints info on matrix at conclusion of `MatAssemblyEnd()`
1004: . -mat_view ::ascii_info_detail - Prints more detailed info
1005: . -mat_view - Prints matrix in ASCII format
1006: . -mat_view ::ascii_matlab - Prints matrix in MATLAB format
1007: . -mat_view draw - PetscDraws nonzero structure of matrix, using `MatView()` and `PetscDrawOpenX()`.
1008: . -display <name> - Sets display name (default is host)
1009: . -draw_pause <sec> - Sets number of seconds to pause after display
1010: . -mat_view socket - Sends matrix to socket, can be accessed from MATLAB (see Users-Manual: ch_matlab for details)
1011: . -viewer_socket_machine <machine> - -
1012: . -viewer_socket_port <port> - -
1013: . -mat_view binary - save matrix to file in binary format
1014: - -viewer_binary_filename <name> - -
1016: Level: beginner
1018: Notes:
1019: The available visualization contexts include
1020: + `PETSC_VIEWER_STDOUT_SELF` - for sequential matrices
1021: . `PETSC_VIEWER_STDOUT_WORLD` - for parallel matrices created on `PETSC_COMM_WORLD`
1022: . `PETSC_VIEWER_STDOUT_`(comm) - for matrices created on MPI communicator comm
1023: - `PETSC_VIEWER_DRAW_WORLD` - graphical display of nonzero structure
1025: The user can open alternative visualization contexts with
1026: + `PetscViewerASCIIOpen()` - Outputs matrix to a specified file
1027: . `PetscViewerBinaryOpen()` - Outputs matrix in binary to a
1028: specified file; corresponding input uses `MatLoad()`
1029: . `PetscViewerDrawOpen()` - Outputs nonzero matrix structure to
1030: an X window display
1031: - `PetscViewerSocketOpen()` - Outputs matrix to Socket viewer.
1032: Currently only the `MATSEQDENSE` and `MATAIJ`
1033: matrix types support the Socket viewer.
1035: The user can call `PetscViewerPushFormat()` to specify the output
1036: format of ASCII printed objects (when using `PETSC_VIEWER_STDOUT_SELF`,
1037: `PETSC_VIEWER_STDOUT_WORLD` and `PetscViewerASCIIOpen()`). Available formats include
1038: + `PETSC_VIEWER_DEFAULT` - default, prints matrix contents
1039: . `PETSC_VIEWER_ASCII_MATLAB` - prints matrix contents in MATLAB format
1040: . `PETSC_VIEWER_ASCII_DENSE` - prints entire matrix including zeros
1041: . `PETSC_VIEWER_ASCII_COMMON` - prints matrix contents, using a sparse
1042: format common among all matrix types
1043: . `PETSC_VIEWER_ASCII_IMPL` - prints matrix contents, using an implementation-specific
1044: format (which is in many cases the same as the default)
1045: . `PETSC_VIEWER_ASCII_INFO` - prints basic information about the matrix
1046: size and structure (not the matrix entries)
1047: - `PETSC_VIEWER_ASCII_INFO_DETAIL` - prints more detailed information about
1048: the matrix structure
1050: The ASCII viewers are only recommended for small matrices on at most a moderate number of processes,
1051: the program will seemingly hang and take hours for larger matrices, for larger matrices one should use the binary format.
1053: In the debugger you can do "call MatView(mat,0)" to display the matrix. (The same holds for any PETSc object viewer).
1055: See the manual page for `MatLoad()` for the exact format of the binary file when the binary
1056: viewer is used.
1058: See share/petsc/matlab/PetscBinaryRead.m for a MATLAB code that can read in the binary file when the binary
1059: viewer is used and lib/petsc/bin/PetscBinaryIO.py for loading them into Python.
1061: One can use '-mat_view draw -draw_pause -1' to pause the graphical display of matrix nonzero structure,
1062: and then use the following mouse functions.
1063: .vb
1064: left mouse: zoom in
1065: middle mouse: zoom out
1066: right mouse: continue with the simulation
1067: .ve
1069: .seealso: [](ch_matrices), `Mat`, `PetscViewerPushFormat()`, `PetscViewerASCIIOpen()`, `PetscViewerDrawOpen()`, `PetscViewer`,
1070: `PetscViewerSocketOpen()`, `PetscViewerBinaryOpen()`, `MatLoad()`, `MatViewFromOptions()`
1071: @*/
1072: PetscErrorCode MatView(Mat mat, PetscViewer viewer)
1073: {
1074: PetscInt rows, cols, rbs, cbs;
1075: PetscBool isascii, isstring, issaws;
1076: PetscViewerFormat format;
1077: PetscMPIInt size;
1079: PetscFunctionBegin;
1082: if (!viewer) PetscCall(PetscViewerASCIIGetStdout(PetscObjectComm((PetscObject)mat), &viewer));
1084: PetscCheckSameComm(mat, 1, viewer, 2);
1086: PetscCall(PetscViewerGetFormat(viewer, &format));
1087: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
1088: if (size == 1 && format == PETSC_VIEWER_LOAD_BALANCE) PetscFunctionReturn(PETSC_SUCCESS);
1090: #if !defined(PETSC_HAVE_THREADSAFETY)
1091: insidematview++;
1092: #endif
1093: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERSTRING, &isstring));
1094: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &isascii));
1095: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERSAWS, &issaws));
1096: PetscCheck((isascii && (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL)) || !mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "No viewers for factored matrix except ASCII, info, or info_detail");
1098: PetscCall(PetscLogEventBegin(MAT_View, mat, viewer, 0, 0));
1099: if (isascii) {
1100: if (!mat->preallocated) {
1101: PetscCall(PetscViewerASCIIPrintf(viewer, "Matrix has not been preallocated yet\n"));
1102: #if !defined(PETSC_HAVE_THREADSAFETY)
1103: insidematview--;
1104: #endif
1105: PetscCall(PetscLogEventEnd(MAT_View, mat, viewer, 0, 0));
1106: PetscFunctionReturn(PETSC_SUCCESS);
1107: }
1108: if (!mat->assembled) {
1109: PetscCall(PetscViewerASCIIPrintf(viewer, "Matrix has not been assembled yet\n"));
1110: #if !defined(PETSC_HAVE_THREADSAFETY)
1111: insidematview--;
1112: #endif
1113: PetscCall(PetscLogEventEnd(MAT_View, mat, viewer, 0, 0));
1114: PetscFunctionReturn(PETSC_SUCCESS);
1115: }
1116: PetscCall(PetscObjectPrintClassNamePrefixType((PetscObject)mat, viewer));
1117: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
1118: MatNullSpace nullsp, transnullsp;
1120: PetscCall(PetscViewerASCIIPushTab(viewer));
1121: PetscCall(MatGetSize(mat, &rows, &cols));
1122: PetscCall(MatGetBlockSizes(mat, &rbs, &cbs));
1123: if (rbs != 1 || cbs != 1) {
1124: if (rbs != cbs) PetscCall(PetscViewerASCIIPrintf(viewer, "rows=%" PetscInt_FMT ", cols=%" PetscInt_FMT ", rbs=%" PetscInt_FMT ", cbs=%" PetscInt_FMT "\n", rows, cols, rbs, cbs));
1125: else PetscCall(PetscViewerASCIIPrintf(viewer, "rows=%" PetscInt_FMT ", cols=%" PetscInt_FMT ", bs=%" PetscInt_FMT "\n", rows, cols, rbs));
1126: } else PetscCall(PetscViewerASCIIPrintf(viewer, "rows=%" PetscInt_FMT ", cols=%" PetscInt_FMT "\n", rows, cols));
1127: if (mat->factortype) {
1128: MatSolverType solver;
1129: PetscCall(MatFactorGetSolverType(mat, &solver));
1130: PetscCall(PetscViewerASCIIPrintf(viewer, "package used to perform factorization: %s\n", solver));
1131: }
1132: if (mat->ops->getinfo) {
1133: MatInfo info;
1134: PetscCall(MatGetInfo(mat, MAT_GLOBAL_SUM, &info));
1135: PetscCall(PetscViewerASCIIPrintf(viewer, "total: nonzeros=%.f, allocated nonzeros=%.f\n", info.nz_used, info.nz_allocated));
1136: if (!mat->factortype) PetscCall(PetscViewerASCIIPrintf(viewer, "total number of mallocs used during MatSetValues calls=%" PetscInt_FMT "\n", (PetscInt)info.mallocs));
1137: }
1138: PetscCall(MatGetNullSpace(mat, &nullsp));
1139: PetscCall(MatGetTransposeNullSpace(mat, &transnullsp));
1140: if (nullsp) PetscCall(PetscViewerASCIIPrintf(viewer, " has attached null space\n"));
1141: if (transnullsp && transnullsp != nullsp) PetscCall(PetscViewerASCIIPrintf(viewer, " has attached transposed null space\n"));
1142: PetscCall(MatGetNearNullSpace(mat, &nullsp));
1143: if (nullsp) PetscCall(PetscViewerASCIIPrintf(viewer, " has attached near null space\n"));
1144: PetscCall(PetscViewerASCIIPushTab(viewer));
1145: PetscCall(MatProductView(mat, viewer));
1146: PetscCall(PetscViewerASCIIPopTab(viewer));
1147: }
1148: } else if (issaws) {
1149: #if defined(PETSC_HAVE_SAWS)
1150: PetscMPIInt rank;
1152: PetscCall(PetscObjectName((PetscObject)mat));
1153: PetscCallMPI(MPI_Comm_rank(PETSC_COMM_WORLD, &rank));
1154: if (!((PetscObject)mat)->amsmem && rank == 0) PetscCall(PetscObjectViewSAWs((PetscObject)mat, viewer));
1155: #endif
1156: } else if (isstring) {
1157: const char *type;
1158: PetscCall(MatGetType(mat, &type));
1159: PetscCall(PetscViewerStringSPrintf(viewer, " MatType: %-7.7s", type));
1160: PetscTryTypeMethod(mat, view, viewer);
1161: }
1162: if ((format == PETSC_VIEWER_NATIVE || format == PETSC_VIEWER_LOAD_BALANCE) && mat->ops->viewnative) {
1163: PetscCall(PetscViewerASCIIPushTab(viewer));
1164: PetscUseTypeMethod(mat, viewnative, viewer);
1165: PetscCall(PetscViewerASCIIPopTab(viewer));
1166: } else if (mat->ops->view) {
1167: PetscCall(PetscViewerASCIIPushTab(viewer));
1168: PetscUseTypeMethod(mat, view, viewer);
1169: PetscCall(PetscViewerASCIIPopTab(viewer));
1170: }
1171: if (isascii) {
1172: PetscCall(PetscViewerGetFormat(viewer, &format));
1173: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) PetscCall(PetscViewerASCIIPopTab(viewer));
1174: }
1175: PetscCall(PetscLogEventEnd(MAT_View, mat, viewer, 0, 0));
1176: #if !defined(PETSC_HAVE_THREADSAFETY)
1177: insidematview--;
1178: #endif
1179: PetscFunctionReturn(PETSC_SUCCESS);
1180: }
1182: #if defined(PETSC_USE_DEBUG)
1183: #include <../src/sys/totalview/tv_data_display.h>
1184: PETSC_UNUSED static int TV_display_type(const struct _p_Mat *mat)
1185: {
1186: TV_add_row("Local rows", "int", &mat->rmap->n);
1187: TV_add_row("Local columns", "int", &mat->cmap->n);
1188: TV_add_row("Global rows", "int", &mat->rmap->N);
1189: TV_add_row("Global columns", "int", &mat->cmap->N);
1190: TV_add_row("Typename", TV_ascii_string_type, ((PetscObject)mat)->type_name);
1191: return TV_format_OK;
1192: }
1193: #endif
1195: /*@C
1196: MatLoad - Loads a matrix that has been stored in binary/HDF5 format
1197: with `MatView()`. The matrix format is determined from the options database.
1198: Generates a parallel MPI matrix if the communicator has more than one
1199: processor. The default matrix type is `MATAIJ`.
1201: Collective
1203: Input Parameters:
1204: + mat - the newly loaded matrix, this needs to have been created with `MatCreate()`
1205: or some related function before a call to `MatLoad()`
1206: - viewer - `PETSCVIEWERBINARY`/`PETSCVIEWERHDF5` file viewer
1208: Options Database Key:
1209: . -matload_block_size <bs> - set block size
1211: Level: beginner
1213: Notes:
1214: If the `Mat` type has not yet been given then `MATAIJ` is used, call `MatSetFromOptions()` on the
1215: `Mat` before calling this routine if you wish to set it from the options database.
1217: `MatLoad()` automatically loads into the options database any options
1218: given in the file filename.info where filename is the name of the file
1219: that was passed to the `PetscViewerBinaryOpen()`. The options in the info
1220: file will be ignored if you use the -viewer_binary_skip_info option.
1222: If the type or size of mat is not set before a call to `MatLoad()`, PETSc
1223: sets the default matrix type AIJ and sets the local and global sizes.
1224: If type and/or size is already set, then the same are used.
1226: In parallel, each processor can load a subset of rows (or the
1227: entire matrix). This routine is especially useful when a large
1228: matrix is stored on disk and only part of it is desired on each
1229: processor. For example, a parallel solver may access only some of
1230: the rows from each processor. The algorithm used here reads
1231: relatively small blocks of data rather than reading the entire
1232: matrix and then subsetting it.
1234: Viewer's `PetscViewerType` must be either `PETSCVIEWERBINARY` or `PETSCVIEWERHDF5`.
1235: Such viewer can be created using `PetscViewerBinaryOpen()` or `PetscViewerHDF5Open()`,
1236: or the sequence like
1237: .vb
1238: `PetscViewer` v;
1239: `PetscViewerCreate`(`PETSC_COMM_WORLD`,&v);
1240: `PetscViewerSetType`(v,`PETSCVIEWERBINARY`);
1241: `PetscViewerSetFromOptions`(v);
1242: `PetscViewerFileSetMode`(v,`FILE_MODE_READ`);
1243: `PetscViewerFileSetName`(v,"datafile");
1244: .ve
1245: The optional `PetscViewerSetFromOptions()` call allows overriding `PetscViewerSetType()` using the option
1246: $ -viewer_type {binary, hdf5}
1248: See the example src/ksp/ksp/tutorials/ex27.c with the first approach,
1249: and src/mat/tutorials/ex10.c with the second approach.
1251: In case of `PETSCVIEWERBINARY`, a native PETSc binary format is used. Each of the blocks
1252: is read onto MPI rank 0 and then shipped to its destination MPI rank, one after another.
1253: Multiple objects, both matrices and vectors, can be stored within the same file.
1254: Their `PetscObject` name is ignored; they are loaded in the order of their storage.
1256: Most users should not need to know the details of the binary storage
1257: format, since `MatLoad()` and `MatView()` completely hide these details.
1258: But for anyone who is interested, the standard binary matrix storage
1259: format is
1261: .vb
1262: PetscInt MAT_FILE_CLASSID
1263: PetscInt number of rows
1264: PetscInt number of columns
1265: PetscInt total number of nonzeros
1266: PetscInt *number nonzeros in each row
1267: PetscInt *column indices of all nonzeros (starting index is zero)
1268: PetscScalar *values of all nonzeros
1269: .ve
1270: If PETSc was not configured with `--with-64-bit-indices` then only `MATMPIAIJ` matrices with more than `PETSC_INT_MAX` non-zeros can be
1271: 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
1272: case will not fit in a (32-bit) `PetscInt` the value `PETSC_INT_MAX` is used for the header entry `total number of nonzeros`.
1274: PETSc automatically does the byte swapping for
1275: machines that store the bytes reversed. Thus if you write your own binary
1276: read/write routines you have to swap the bytes; see `PetscBinaryRead()`
1277: and `PetscBinaryWrite()` to see how this may be done.
1279: In case of `PETSCVIEWERHDF5`, a parallel HDF5 reader is used.
1280: Each processor's chunk is loaded independently by its owning MPI process.
1281: Multiple objects, both matrices and vectors, can be stored within the same file.
1282: They are looked up by their PetscObject name.
1284: As the MATLAB MAT-File Version 7.3 format is also a HDF5 flavor, we decided to use
1285: by default the same structure and naming of the AIJ arrays and column count
1286: within the HDF5 file. This means that a MAT file saved with -v7.3 flag, e.g.
1287: $ save example.mat A b -v7.3
1288: can be directly read by this routine (see Reference 1 for details).
1290: Depending on your MATLAB version, this format might be a default,
1291: otherwise you can set it as default in Preferences.
1293: Unless -nocompression flag is used to save the file in MATLAB,
1294: PETSc must be configured with ZLIB package.
1296: See also examples src/mat/tutorials/ex10.c and src/ksp/ksp/tutorials/ex27.c
1298: This reader currently supports only real `MATSEQAIJ`, `MATMPIAIJ`, `MATSEQDENSE` and `MATMPIDENSE` matrices for `PETSCVIEWERHDF5`
1300: Corresponding `MatView()` is not yet implemented.
1302: The loaded matrix is actually a transpose of the original one in MATLAB,
1303: unless you push `PETSC_VIEWER_HDF5_MAT` format (see examples above).
1304: With this format, matrix is automatically transposed by PETSc,
1305: unless the matrix is marked as SPD or symmetric
1306: (see `MatSetOption()`, `MAT_SPD`, `MAT_SYMMETRIC`).
1308: See MATLAB Documentation on `save()`, <https://www.mathworks.com/help/matlab/ref/save.html#btox10b-1-version>
1310: .seealso: [](ch_matrices), `Mat`, `PetscViewerBinaryOpen()`, `PetscViewerSetType()`, `MatView()`, `VecLoad()`
1311: @*/
1312: PetscErrorCode MatLoad(Mat mat, PetscViewer viewer)
1313: {
1314: PetscBool flg;
1316: PetscFunctionBegin;
1320: if (!((PetscObject)mat)->type_name) PetscCall(MatSetType(mat, MATAIJ));
1322: flg = PETSC_FALSE;
1323: PetscCall(PetscOptionsGetBool(((PetscObject)mat)->options, ((PetscObject)mat)->prefix, "-matload_symmetric", &flg, NULL));
1324: if (flg) {
1325: PetscCall(MatSetOption(mat, MAT_SYMMETRIC, PETSC_TRUE));
1326: PetscCall(MatSetOption(mat, MAT_SYMMETRY_ETERNAL, PETSC_TRUE));
1327: }
1328: flg = PETSC_FALSE;
1329: PetscCall(PetscOptionsGetBool(((PetscObject)mat)->options, ((PetscObject)mat)->prefix, "-matload_spd", &flg, NULL));
1330: if (flg) PetscCall(MatSetOption(mat, MAT_SPD, PETSC_TRUE));
1332: PetscCall(PetscLogEventBegin(MAT_Load, mat, viewer, 0, 0));
1333: PetscUseTypeMethod(mat, load, viewer);
1334: PetscCall(PetscLogEventEnd(MAT_Load, mat, viewer, 0, 0));
1335: PetscFunctionReturn(PETSC_SUCCESS);
1336: }
1338: static PetscErrorCode MatDestroy_Redundant(Mat_Redundant **redundant)
1339: {
1340: Mat_Redundant *redund = *redundant;
1342: PetscFunctionBegin;
1343: if (redund) {
1344: if (redund->matseq) { /* via MatCreateSubMatrices() */
1345: PetscCall(ISDestroy(&redund->isrow));
1346: PetscCall(ISDestroy(&redund->iscol));
1347: PetscCall(MatDestroySubMatrices(1, &redund->matseq));
1348: } else {
1349: PetscCall(PetscFree2(redund->send_rank, redund->recv_rank));
1350: PetscCall(PetscFree(redund->sbuf_j));
1351: PetscCall(PetscFree(redund->sbuf_a));
1352: for (PetscInt i = 0; i < redund->nrecvs; i++) {
1353: PetscCall(PetscFree(redund->rbuf_j[i]));
1354: PetscCall(PetscFree(redund->rbuf_a[i]));
1355: }
1356: PetscCall(PetscFree4(redund->sbuf_nz, redund->rbuf_nz, redund->rbuf_j, redund->rbuf_a));
1357: }
1359: if (redund->subcomm) PetscCall(PetscCommDestroy(&redund->subcomm));
1360: PetscCall(PetscFree(redund));
1361: }
1362: PetscFunctionReturn(PETSC_SUCCESS);
1363: }
1365: /*@C
1366: MatDestroy - Frees space taken by a matrix.
1368: Collective
1370: Input Parameter:
1371: . A - the matrix
1373: Level: beginner
1375: Developer Note:
1376: Some special arrays of matrices are not destroyed in this routine but instead by the routines called by
1377: `MatDestroySubMatrices()`. Thus one must be sure that any changes here must also be made in those routines.
1378: `MatHeaderMerge()` and `MatHeaderReplace()` also manipulate the data in the `Mat` object and likely need changes
1379: if changes are needed here.
1381: .seealso: [](ch_matrices), `Mat`, `MatCreate()`
1382: @*/
1383: PetscErrorCode MatDestroy(Mat *A)
1384: {
1385: PetscFunctionBegin;
1386: if (!*A) PetscFunctionReturn(PETSC_SUCCESS);
1388: if (--((PetscObject)(*A))->refct > 0) {
1389: *A = NULL;
1390: PetscFunctionReturn(PETSC_SUCCESS);
1391: }
1393: /* if memory was published with SAWs then destroy it */
1394: PetscCall(PetscObjectSAWsViewOff((PetscObject)*A));
1395: PetscTryTypeMethod((*A), destroy);
1397: PetscCall(PetscFree((*A)->factorprefix));
1398: PetscCall(PetscFree((*A)->defaultvectype));
1399: PetscCall(PetscFree((*A)->defaultrandtype));
1400: PetscCall(PetscFree((*A)->bsizes));
1401: PetscCall(PetscFree((*A)->solvertype));
1402: for (PetscInt i = 0; i < MAT_FACTOR_NUM_TYPES; i++) PetscCall(PetscFree((*A)->preferredordering[i]));
1403: if ((*A)->redundant && (*A)->redundant->matseq[0] == *A) (*A)->redundant->matseq[0] = NULL;
1404: PetscCall(MatDestroy_Redundant(&(*A)->redundant));
1405: PetscCall(MatProductClear(*A));
1406: PetscCall(MatNullSpaceDestroy(&(*A)->nullsp));
1407: PetscCall(MatNullSpaceDestroy(&(*A)->transnullsp));
1408: PetscCall(MatNullSpaceDestroy(&(*A)->nearnullsp));
1409: PetscCall(MatDestroy(&(*A)->schur));
1410: PetscCall(PetscLayoutDestroy(&(*A)->rmap));
1411: PetscCall(PetscLayoutDestroy(&(*A)->cmap));
1412: PetscCall(PetscHeaderDestroy(A));
1413: PetscFunctionReturn(PETSC_SUCCESS);
1414: }
1416: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
1417: /*@C
1418: MatSetValues - Inserts or adds a block of values into a matrix.
1419: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
1420: MUST be called after all calls to `MatSetValues()` have been completed.
1422: Not Collective
1424: Input Parameters:
1425: + mat - the matrix
1426: . v - a logically two-dimensional array of values
1427: . m - the number of rows
1428: . idxm - the global indices of the rows
1429: . n - the number of columns
1430: . idxn - the global indices of the columns
1431: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
1433: Level: beginner
1435: Notes:
1436: By default the values, `v`, are stored row-oriented. See `MatSetOption()` for other options.
1438: Calls to `MatSetValues()` with the `INSERT_VALUES` and `ADD_VALUES`
1439: options cannot be mixed without intervening calls to the assembly
1440: routines.
1442: `MatSetValues()` uses 0-based row and column numbers in Fortran
1443: as well as in C.
1445: Negative indices may be passed in `idxm` and `idxn`, these rows and columns are
1446: simply ignored. This allows easily inserting element stiffness matrices
1447: with homogeneous Dirichlet boundary conditions that you don't want represented
1448: in the matrix.
1450: Efficiency Alert:
1451: The routine `MatSetValuesBlocked()` may offer much better efficiency
1452: for users of block sparse formats (`MATSEQBAIJ` and `MATMPIBAIJ`).
1454: Developer Note:
1455: This is labeled with C so does not automatically generate Fortran stubs and interfaces
1456: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
1458: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1459: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`
1460: @*/
1461: PetscErrorCode MatSetValues(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], const PetscScalar v[], InsertMode addv)
1462: {
1463: PetscFunctionBeginHot;
1466: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1467: PetscAssertPointer(idxm, 3);
1468: PetscAssertPointer(idxn, 5);
1469: MatCheckPreallocated(mat, 1);
1471: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
1472: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
1474: if (PetscDefined(USE_DEBUG)) {
1475: PetscInt i, j;
1477: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
1478: for (i = 0; i < m; i++) {
1479: for (j = 0; j < n; j++) {
1480: if (mat->erroriffailure && PetscIsInfOrNanScalar(v[i * n + j]))
1481: #if defined(PETSC_USE_COMPLEX)
1482: 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]);
1483: #else
1484: 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]);
1485: #endif
1486: }
1487: }
1488: 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);
1489: 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);
1490: }
1492: if (mat->assembled) {
1493: mat->was_assembled = PETSC_TRUE;
1494: mat->assembled = PETSC_FALSE;
1495: }
1496: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
1497: PetscUseTypeMethod(mat, setvalues, m, idxm, n, idxn, v, addv);
1498: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
1499: PetscFunctionReturn(PETSC_SUCCESS);
1500: }
1502: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
1503: /*@C
1504: MatSetValuesIS - Inserts or adds a block of values into a matrix using an `IS` to indicate the rows and columns
1505: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
1506: MUST be called after all calls to `MatSetValues()` have been completed.
1508: Not Collective
1510: Input Parameters:
1511: + mat - the matrix
1512: . v - a logically two-dimensional array of values
1513: . ism - the rows to provide
1514: . isn - the columns to provide
1515: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
1517: Level: beginner
1519: Notes:
1520: By default the values, `v`, are stored row-oriented. See `MatSetOption()` for other options.
1522: Calls to `MatSetValues()` with the `INSERT_VALUES` and `ADD_VALUES`
1523: options cannot be mixed without intervening calls to the assembly
1524: routines.
1526: `MatSetValues()` uses 0-based row and column numbers in Fortran
1527: as well as in C.
1529: Negative indices may be passed in `ism` and `isn`, these rows and columns are
1530: simply ignored. This allows easily inserting element stiffness matrices
1531: with homogeneous Dirichlet boundary conditions that you don't want represented
1532: in the matrix.
1534: Efficiency Alert:
1535: The routine `MatSetValuesBlocked()` may offer much better efficiency
1536: for users of block sparse formats (`MATSEQBAIJ` and `MATMPIBAIJ`).
1538: This is currently not optimized for any particular `ISType`
1540: Developer Note:
1541: This is labeled with C so does not automatically generate Fortran stubs and interfaces
1542: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
1544: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatSetValues()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1545: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`
1546: @*/
1547: PetscErrorCode MatSetValuesIS(Mat mat, IS ism, IS isn, const PetscScalar v[], InsertMode addv)
1548: {
1549: PetscInt m, n;
1550: const PetscInt *rows, *cols;
1552: PetscFunctionBeginHot;
1554: PetscCall(ISGetIndices(ism, &rows));
1555: PetscCall(ISGetIndices(isn, &cols));
1556: PetscCall(ISGetLocalSize(ism, &m));
1557: PetscCall(ISGetLocalSize(isn, &n));
1558: PetscCall(MatSetValues(mat, m, rows, n, cols, v, addv));
1559: PetscCall(ISRestoreIndices(ism, &rows));
1560: PetscCall(ISRestoreIndices(isn, &cols));
1561: PetscFunctionReturn(PETSC_SUCCESS);
1562: }
1564: /*@
1565: MatSetValuesRowLocal - Inserts a row (block row for `MATBAIJ` matrices) of nonzero
1566: values into a matrix
1568: Not Collective
1570: Input Parameters:
1571: + mat - the matrix
1572: . row - the (block) row to set
1573: - v - a logically two-dimensional array of values
1575: Level: intermediate
1577: Notes:
1578: The values, `v`, are column-oriented (for the block version) and sorted
1580: All the nonzero values in `row` must be provided
1582: The matrix must have previously had its column indices set, likely by having been assembled.
1584: `row` must belong to this MPI process
1586: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1587: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`, `MatSetValues()`, `MatSetValuesRow()`, `MatSetLocalToGlobalMapping()`
1588: @*/
1589: PetscErrorCode MatSetValuesRowLocal(Mat mat, PetscInt row, const PetscScalar v[])
1590: {
1591: PetscInt globalrow;
1593: PetscFunctionBegin;
1596: PetscAssertPointer(v, 3);
1597: PetscCall(ISLocalToGlobalMappingApply(mat->rmap->mapping, 1, &row, &globalrow));
1598: PetscCall(MatSetValuesRow(mat, globalrow, v));
1599: PetscFunctionReturn(PETSC_SUCCESS);
1600: }
1602: /*@
1603: MatSetValuesRow - Inserts a row (block row for `MATBAIJ` matrices) of nonzero
1604: values into a matrix
1606: Not Collective
1608: Input Parameters:
1609: + mat - the matrix
1610: . row - the (block) row to set
1611: - v - a logically two-dimensional (column major) array of values for block matrices with blocksize larger than one, otherwise a one dimensional array of values
1613: Level: advanced
1615: Notes:
1616: The values, `v`, are column-oriented for the block version.
1618: All the nonzeros in `row` must be provided
1620: THE MATRIX MUST HAVE PREVIOUSLY HAD ITS COLUMN INDICES SET. IT IS RARE THAT THIS ROUTINE IS USED, usually `MatSetValues()` is used.
1622: `row` must belong to this process
1624: .seealso: [](ch_matrices), `Mat`, `MatSetValues()`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1625: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`
1626: @*/
1627: PetscErrorCode MatSetValuesRow(Mat mat, PetscInt row, const PetscScalar v[])
1628: {
1629: PetscFunctionBeginHot;
1632: MatCheckPreallocated(mat, 1);
1633: PetscAssertPointer(v, 3);
1634: PetscCheck(mat->insertmode != ADD_VALUES, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add and insert values");
1635: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
1636: mat->insertmode = INSERT_VALUES;
1638: if (mat->assembled) {
1639: mat->was_assembled = PETSC_TRUE;
1640: mat->assembled = PETSC_FALSE;
1641: }
1642: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
1643: PetscUseTypeMethod(mat, setvaluesrow, row, v);
1644: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
1645: PetscFunctionReturn(PETSC_SUCCESS);
1646: }
1648: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
1649: /*@
1650: MatSetValuesStencil - Inserts or adds a block of values into a matrix.
1651: Using structured grid indexing
1653: Not Collective
1655: Input Parameters:
1656: + mat - the matrix
1657: . m - number of rows being entered
1658: . idxm - grid coordinates (and component number when dof > 1) for matrix rows being entered
1659: . n - number of columns being entered
1660: . idxn - grid coordinates (and component number when dof > 1) for matrix columns being entered
1661: . v - a logically two-dimensional array of values
1662: - addv - either `ADD_VALUES` to add to existing entries at that location or `INSERT_VALUES` to replace existing entries with new values
1664: Level: beginner
1666: Notes:
1667: By default the values, `v`, are row-oriented. See `MatSetOption()` for other options.
1669: Calls to `MatSetValuesStencil()` with the `INSERT_VALUES` and `ADD_VALUES`
1670: options cannot be mixed without intervening calls to the assembly
1671: routines.
1673: The grid coordinates are across the entire grid, not just the local portion
1675: `MatSetValuesStencil()` uses 0-based row and column numbers in Fortran
1676: as well as in C.
1678: For setting/accessing vector values via array coordinates you can use the `DMDAVecGetArray()` routine
1680: In order to use this routine you must either obtain the matrix with `DMCreateMatrix()`
1681: or call `MatSetLocalToGlobalMapping()` and `MatSetStencil()` first.
1683: The columns and rows in the stencil passed in MUST be contained within the
1684: ghost region of the given process as set with DMDACreateXXX() or `MatSetStencil()`. For example,
1685: if you create a `DMDA` with an overlap of one grid level and on a particular process its first
1686: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1687: first i index you can use in your column and row indices in `MatSetStencil()` is 5.
1689: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
1690: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
1691: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
1692: `DM_BOUNDARY_PERIODIC` boundary type.
1694: 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
1695: a single value per point) you can skip filling those indices.
1697: Inspired by the structured grid interface to the HYPRE package
1698: (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)
1700: Efficiency Alert:
1701: The routine `MatSetValuesBlockedStencil()` may offer much better efficiency
1702: for users of block sparse formats (`MATSEQBAIJ` and `MATMPIBAIJ`).
1704: Fortran Note:
1705: `idxm` and `idxn` should be declared as
1706: $ MatStencil idxm(4,m),idxn(4,n)
1707: and the values inserted using
1708: .vb
1709: idxm(MatStencil_i,1) = i
1710: idxm(MatStencil_j,1) = j
1711: idxm(MatStencil_k,1) = k
1712: idxm(MatStencil_c,1) = c
1713: etc
1714: .ve
1716: .seealso: [](ch_matrices), `Mat`, `DMDA`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`
1717: `MatSetValues()`, `MatSetValuesBlockedStencil()`, `MatSetStencil()`, `DMCreateMatrix()`, `DMDAVecGetArray()`, `MatStencil`
1718: @*/
1719: PetscErrorCode MatSetValuesStencil(Mat mat, PetscInt m, const MatStencil idxm[], PetscInt n, const MatStencil idxn[], const PetscScalar v[], InsertMode addv)
1720: {
1721: PetscInt buf[8192], *bufm = NULL, *bufn = NULL, *jdxm, *jdxn;
1722: PetscInt j, i, dim = mat->stencil.dim, *dims = mat->stencil.dims + 1, tmp;
1723: PetscInt *starts = mat->stencil.starts, *dxm = (PetscInt *)idxm, *dxn = (PetscInt *)idxn, sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1725: PetscFunctionBegin;
1726: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1729: PetscAssertPointer(idxm, 3);
1730: PetscAssertPointer(idxn, 5);
1732: if ((m + n) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
1733: jdxm = buf;
1734: jdxn = buf + m;
1735: } else {
1736: PetscCall(PetscMalloc2(m, &bufm, n, &bufn));
1737: jdxm = bufm;
1738: jdxn = bufn;
1739: }
1740: for (i = 0; i < m; i++) {
1741: for (j = 0; j < 3 - sdim; j++) dxm++;
1742: tmp = *dxm++ - starts[0];
1743: for (j = 0; j < dim - 1; j++) {
1744: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1745: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
1746: }
1747: if (mat->stencil.noc) dxm++;
1748: jdxm[i] = tmp;
1749: }
1750: for (i = 0; i < n; i++) {
1751: for (j = 0; j < 3 - sdim; j++) dxn++;
1752: tmp = *dxn++ - starts[0];
1753: for (j = 0; j < dim - 1; j++) {
1754: if ((*dxn++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1755: else tmp = tmp * dims[j] + *(dxn - 1) - starts[j + 1];
1756: }
1757: if (mat->stencil.noc) dxn++;
1758: jdxn[i] = tmp;
1759: }
1760: PetscCall(MatSetValuesLocal(mat, m, jdxm, n, jdxn, v, addv));
1761: PetscCall(PetscFree2(bufm, bufn));
1762: PetscFunctionReturn(PETSC_SUCCESS);
1763: }
1765: /*@
1766: MatSetValuesBlockedStencil - Inserts or adds a block of values into a matrix.
1767: Using structured grid indexing
1769: Not Collective
1771: Input Parameters:
1772: + mat - the matrix
1773: . m - number of rows being entered
1774: . idxm - grid coordinates for matrix rows being entered
1775: . n - number of columns being entered
1776: . idxn - grid coordinates for matrix columns being entered
1777: . v - a logically two-dimensional array of values
1778: - addv - either `ADD_VALUES` to add to existing entries or `INSERT_VALUES` to replace existing entries with new values
1780: Level: beginner
1782: Notes:
1783: By default the values, `v`, are row-oriented and unsorted.
1784: See `MatSetOption()` for other options.
1786: Calls to `MatSetValuesBlockedStencil()` with the `INSERT_VALUES` and `ADD_VALUES`
1787: options cannot be mixed without intervening calls to the assembly
1788: routines.
1790: The grid coordinates are across the entire grid, not just the local portion
1792: `MatSetValuesBlockedStencil()` uses 0-based row and column numbers in Fortran
1793: as well as in C.
1795: For setting/accessing vector values via array coordinates you can use the `DMDAVecGetArray()` routine
1797: In order to use this routine you must either obtain the matrix with `DMCreateMatrix()`
1798: or call `MatSetBlockSize()`, `MatSetLocalToGlobalMapping()` and `MatSetStencil()` first.
1800: The columns and rows in the stencil passed in MUST be contained within the
1801: ghost region of the given process as set with DMDACreateXXX() or `MatSetStencil()`. For example,
1802: if you create a `DMDA` with an overlap of one grid level and on a particular process its first
1803: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1804: first i index you can use in your column and row indices in `MatSetStencil()` is 5.
1806: Negative indices may be passed in idxm and idxn, these rows and columns are
1807: simply ignored. This allows easily inserting element stiffness matrices
1808: with homogeneous Dirichlet boundary conditions that you don't want represented
1809: in the matrix.
1811: Inspired by the structured grid interface to the HYPRE package
1812: (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)
1814: Fortran Note:
1815: `idxm` and `idxn` should be declared as
1816: $ MatStencil idxm(4,m),idxn(4,n)
1817: and the values inserted using
1818: .vb
1819: idxm(MatStencil_i,1) = i
1820: idxm(MatStencil_j,1) = j
1821: idxm(MatStencil_k,1) = k
1822: etc
1823: .ve
1825: .seealso: [](ch_matrices), `Mat`, `DMDA`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`
1826: `MatSetValues()`, `MatSetValuesStencil()`, `MatSetStencil()`, `DMCreateMatrix()`, `DMDAVecGetArray()`, `MatStencil`,
1827: `MatSetBlockSize()`, `MatSetLocalToGlobalMapping()`
1828: @*/
1829: PetscErrorCode MatSetValuesBlockedStencil(Mat mat, PetscInt m, const MatStencil idxm[], PetscInt n, const MatStencil idxn[], const PetscScalar v[], InsertMode addv)
1830: {
1831: PetscInt buf[8192], *bufm = NULL, *bufn = NULL, *jdxm, *jdxn;
1832: PetscInt j, i, dim = mat->stencil.dim, *dims = mat->stencil.dims + 1, tmp;
1833: PetscInt *starts = mat->stencil.starts, *dxm = (PetscInt *)idxm, *dxn = (PetscInt *)idxn, sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1835: PetscFunctionBegin;
1836: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1839: PetscAssertPointer(idxm, 3);
1840: PetscAssertPointer(idxn, 5);
1841: PetscAssertPointer(v, 6);
1843: if ((m + n) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
1844: jdxm = buf;
1845: jdxn = buf + m;
1846: } else {
1847: PetscCall(PetscMalloc2(m, &bufm, n, &bufn));
1848: jdxm = bufm;
1849: jdxn = bufn;
1850: }
1851: for (i = 0; i < m; i++) {
1852: for (j = 0; j < 3 - sdim; j++) dxm++;
1853: tmp = *dxm++ - starts[0];
1854: for (j = 0; j < sdim - 1; j++) {
1855: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1856: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
1857: }
1858: dxm++;
1859: jdxm[i] = tmp;
1860: }
1861: for (i = 0; i < n; i++) {
1862: for (j = 0; j < 3 - sdim; j++) dxn++;
1863: tmp = *dxn++ - starts[0];
1864: for (j = 0; j < sdim - 1; j++) {
1865: if ((*dxn++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1866: else tmp = tmp * dims[j] + *(dxn - 1) - starts[j + 1];
1867: }
1868: dxn++;
1869: jdxn[i] = tmp;
1870: }
1871: PetscCall(MatSetValuesBlockedLocal(mat, m, jdxm, n, jdxn, v, addv));
1872: PetscCall(PetscFree2(bufm, bufn));
1873: PetscFunctionReturn(PETSC_SUCCESS);
1874: }
1876: /*@
1877: MatSetStencil - Sets the grid information for setting values into a matrix via
1878: `MatSetValuesStencil()`
1880: Not Collective
1882: Input Parameters:
1883: + mat - the matrix
1884: . dim - dimension of the grid 1, 2, or 3
1885: . dims - number of grid points in x, y, and z direction, including ghost points on your processor
1886: . starts - starting point of ghost nodes on your processor in x, y, and z direction
1887: - dof - number of degrees of freedom per node
1889: Level: beginner
1891: Notes:
1892: Inspired by the structured grid interface to the HYPRE package
1893: (www.llnl.gov/CASC/hyper)
1895: For matrices generated with `DMCreateMatrix()` this routine is automatically called and so not needed by the
1896: user.
1898: .seealso: [](ch_matrices), `Mat`, `MatStencil`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`
1899: `MatSetValues()`, `MatSetValuesBlockedStencil()`, `MatSetValuesStencil()`
1900: @*/
1901: PetscErrorCode MatSetStencil(Mat mat, PetscInt dim, const PetscInt dims[], const PetscInt starts[], PetscInt dof)
1902: {
1903: PetscFunctionBegin;
1905: PetscAssertPointer(dims, 3);
1906: PetscAssertPointer(starts, 4);
1908: mat->stencil.dim = dim + (dof > 1);
1909: for (PetscInt i = 0; i < dim; i++) {
1910: mat->stencil.dims[i] = dims[dim - i - 1]; /* copy the values in backwards */
1911: mat->stencil.starts[i] = starts[dim - i - 1];
1912: }
1913: mat->stencil.dims[dim] = dof;
1914: mat->stencil.starts[dim] = 0;
1915: mat->stencil.noc = (PetscBool)(dof == 1);
1916: PetscFunctionReturn(PETSC_SUCCESS);
1917: }
1919: /*@C
1920: MatSetValuesBlocked - Inserts or adds a block of values into a matrix.
1922: Not Collective
1924: Input Parameters:
1925: + mat - the matrix
1926: . v - a logically two-dimensional array of values
1927: . m - the number of block rows
1928: . idxm - the global block indices
1929: . n - the number of block columns
1930: . idxn - the global block indices
1931: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` replaces existing entries with new values
1933: Level: intermediate
1935: Notes:
1936: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call
1937: MatXXXXSetPreallocation() or `MatSetUp()` before using this routine.
1939: The `m` and `n` count the NUMBER of blocks in the row direction and column direction,
1940: NOT the total number of rows/columns; for example, if the block size is 2 and
1941: you are passing in values for rows 2,3,4,5 then m would be 2 (not 4).
1942: The values in idxm would be 1 2; that is the first index for each block divided by
1943: the block size.
1945: You must call `MatSetBlockSize()` when constructing this matrix (before
1946: preallocating it).
1948: By default the values, `v`, are row-oriented, so the layout of
1949: `v` is the same as for `MatSetValues()`. See `MatSetOption()` for other options.
1951: Calls to `MatSetValuesBlocked()` with the `INSERT_VALUES` and `ADD_VALUES`
1952: options cannot be mixed without intervening calls to the assembly
1953: routines.
1955: `MatSetValuesBlocked()` uses 0-based row and column numbers in Fortran
1956: as well as in C.
1958: Negative indices may be passed in `idxm` and `idxn`, these rows and columns are
1959: simply ignored. This allows easily inserting element stiffness matrices
1960: with homogeneous Dirichlet boundary conditions that you don't want represented
1961: in the matrix.
1963: Each time an entry is set within a sparse matrix via `MatSetValues()`,
1964: internal searching must be done to determine where to place the
1965: data in the matrix storage space. By instead inserting blocks of
1966: entries via `MatSetValuesBlocked()`, the overhead of matrix assembly is
1967: reduced.
1969: Example:
1970: .vb
1971: Suppose m=n=2 and block size(bs) = 2 The array is
1973: 1 2 | 3 4
1974: 5 6 | 7 8
1975: - - - | - - -
1976: 9 10 | 11 12
1977: 13 14 | 15 16
1979: v[] should be passed in like
1980: v[] = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]
1982: If you are not using row oriented storage of v (that is you called MatSetOption(mat,MAT_ROW_ORIENTED,PETSC_FALSE)) then
1983: v[] = [1,5,9,13,2,6,10,14,3,7,11,15,4,8,12,16]
1984: .ve
1986: .seealso: [](ch_matrices), `Mat`, `MatSetBlockSize()`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetValuesBlockedLocal()`
1987: @*/
1988: PetscErrorCode MatSetValuesBlocked(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], const PetscScalar v[], InsertMode addv)
1989: {
1990: PetscFunctionBeginHot;
1993: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1994: PetscAssertPointer(idxm, 3);
1995: PetscAssertPointer(idxn, 5);
1996: MatCheckPreallocated(mat, 1);
1997: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
1998: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
1999: if (PetscDefined(USE_DEBUG)) {
2000: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2001: PetscCheck(mat->ops->setvaluesblocked || mat->ops->setvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2002: }
2003: if (PetscDefined(USE_DEBUG)) {
2004: PetscInt rbs, cbs, M, N, i;
2005: PetscCall(MatGetBlockSizes(mat, &rbs, &cbs));
2006: PetscCall(MatGetSize(mat, &M, &N));
2007: for (i = 0; i < m; i++) PetscCheck(idxm[i] * rbs < M, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Row block index %" PetscInt_FMT " (index %" PetscInt_FMT ") greater than row length %" PetscInt_FMT, i, idxm[i], M);
2008: for (i = 0; i < n; i++) PetscCheck(idxn[i] * cbs < N, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column block index %" PetscInt_FMT " (index %" PetscInt_FMT ") great than column length %" PetscInt_FMT, i, idxn[i], N);
2009: }
2010: if (mat->assembled) {
2011: mat->was_assembled = PETSC_TRUE;
2012: mat->assembled = PETSC_FALSE;
2013: }
2014: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
2015: if (mat->ops->setvaluesblocked) {
2016: PetscUseTypeMethod(mat, setvaluesblocked, m, idxm, n, idxn, v, addv);
2017: } else {
2018: PetscInt buf[8192], *bufr = NULL, *bufc = NULL, *iidxm, *iidxn;
2019: PetscInt i, j, bs, cbs;
2021: PetscCall(MatGetBlockSizes(mat, &bs, &cbs));
2022: if ((m * bs + n * cbs) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
2023: iidxm = buf;
2024: iidxn = buf + m * bs;
2025: } else {
2026: PetscCall(PetscMalloc2(m * bs, &bufr, n * cbs, &bufc));
2027: iidxm = bufr;
2028: iidxn = bufc;
2029: }
2030: for (i = 0; i < m; i++) {
2031: for (j = 0; j < bs; j++) iidxm[i * bs + j] = bs * idxm[i] + j;
2032: }
2033: if (m != n || bs != cbs || idxm != idxn) {
2034: for (i = 0; i < n; i++) {
2035: for (j = 0; j < cbs; j++) iidxn[i * cbs + j] = cbs * idxn[i] + j;
2036: }
2037: } else iidxn = iidxm;
2038: PetscCall(MatSetValues(mat, m * bs, iidxm, n * cbs, iidxn, v, addv));
2039: PetscCall(PetscFree2(bufr, bufc));
2040: }
2041: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
2042: PetscFunctionReturn(PETSC_SUCCESS);
2043: }
2045: /*@C
2046: MatGetValues - Gets a block of local values from a matrix.
2048: Not Collective; can only return values that are owned by the give process
2050: Input Parameters:
2051: + mat - the matrix
2052: . v - a logically two-dimensional array for storing the values
2053: . m - the number of rows
2054: . idxm - the global indices of the rows
2055: . n - the number of columns
2056: - idxn - the global indices of the columns
2058: Level: advanced
2060: Notes:
2061: The user must allocate space (m*n `PetscScalar`s) for the values, `v`.
2062: The values, `v`, are then returned in a row-oriented format,
2063: analogous to that used by default in `MatSetValues()`.
2065: `MatGetValues()` uses 0-based row and column numbers in
2066: Fortran as well as in C.
2068: `MatGetValues()` requires that the matrix has been assembled
2069: with `MatAssemblyBegin()`/`MatAssemblyEnd()`. Thus, calls to
2070: `MatSetValues()` and `MatGetValues()` CANNOT be made in succession
2071: without intermediate matrix assembly.
2073: Negative row or column indices will be ignored and those locations in `v` will be
2074: left unchanged.
2076: For the standard row-based matrix formats, `idxm` can only contain rows owned by the requesting MPI process.
2077: That is, rows with global index greater than or equal to rstart and less than rend where rstart and rend are obtainable
2078: from `MatGetOwnershipRange`(mat,&rstart,&rend).
2080: .seealso: [](ch_matrices), `Mat`, `MatGetRow()`, `MatCreateSubMatrices()`, `MatSetValues()`, `MatGetOwnershipRange()`, `MatGetValuesLocal()`, `MatGetValue()`
2081: @*/
2082: PetscErrorCode MatGetValues(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], PetscScalar v[])
2083: {
2084: PetscFunctionBegin;
2087: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS);
2088: PetscAssertPointer(idxm, 3);
2089: PetscAssertPointer(idxn, 5);
2090: PetscAssertPointer(v, 6);
2091: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2092: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2093: MatCheckPreallocated(mat, 1);
2095: PetscCall(PetscLogEventBegin(MAT_GetValues, mat, 0, 0, 0));
2096: PetscUseTypeMethod(mat, getvalues, m, idxm, n, idxn, v);
2097: PetscCall(PetscLogEventEnd(MAT_GetValues, mat, 0, 0, 0));
2098: PetscFunctionReturn(PETSC_SUCCESS);
2099: }
2101: /*@C
2102: MatGetValuesLocal - retrieves values from certain locations in a matrix using the local numbering of the indices
2103: defined previously by `MatSetLocalToGlobalMapping()`
2105: Not Collective
2107: Input Parameters:
2108: + mat - the matrix
2109: . nrow - number of rows
2110: . irow - the row local indices
2111: . ncol - number of columns
2112: - icol - the column local indices
2114: Output Parameter:
2115: . y - a logically two-dimensional array of values
2117: Level: advanced
2119: Notes:
2120: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call `MatSetLocalToGlobalMapping()` before using this routine.
2122: 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,
2123: are greater than or equal to rstart and less than rend where rstart and rend are obtainable from `MatGetOwnershipRange`(mat,&rstart,&rend). One can
2124: determine if the resulting global row associated with the local row r is owned by the requesting MPI process by applying the `ISLocalToGlobalMapping` set
2125: with `MatSetLocalToGlobalMapping()`.
2127: Developer Note:
2128: This is labelled with C so does not automatically generate Fortran stubs and interfaces
2129: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2131: .seealso: [](ch_matrices), `Mat`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetLocalToGlobalMapping()`,
2132: `MatSetValuesLocal()`, `MatGetValues()`
2133: @*/
2134: PetscErrorCode MatGetValuesLocal(Mat mat, PetscInt nrow, const PetscInt irow[], PetscInt ncol, const PetscInt icol[], PetscScalar y[])
2135: {
2136: PetscFunctionBeginHot;
2139: MatCheckPreallocated(mat, 1);
2140: if (!nrow || !ncol) PetscFunctionReturn(PETSC_SUCCESS); /* no values to retrieve */
2141: PetscAssertPointer(irow, 3);
2142: PetscAssertPointer(icol, 5);
2143: if (PetscDefined(USE_DEBUG)) {
2144: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2145: PetscCheck(mat->ops->getvalueslocal || mat->ops->getvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2146: }
2147: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2148: PetscCall(PetscLogEventBegin(MAT_GetValues, mat, 0, 0, 0));
2149: if (mat->ops->getvalueslocal) PetscUseTypeMethod(mat, getvalueslocal, nrow, irow, ncol, icol, y);
2150: else {
2151: PetscInt buf[8192], *bufr = NULL, *bufc = NULL, *irowm, *icolm;
2152: if ((nrow + ncol) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
2153: irowm = buf;
2154: icolm = buf + nrow;
2155: } else {
2156: PetscCall(PetscMalloc2(nrow, &bufr, ncol, &bufc));
2157: irowm = bufr;
2158: icolm = bufc;
2159: }
2160: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "MatGetValuesLocal() cannot proceed without local-to-global row mapping (See MatSetLocalToGlobalMapping()).");
2161: PetscCheck(mat->cmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "MatGetValuesLocal() cannot proceed without local-to-global column mapping (See MatSetLocalToGlobalMapping()).");
2162: PetscCall(ISLocalToGlobalMappingApply(mat->rmap->mapping, nrow, irow, irowm));
2163: PetscCall(ISLocalToGlobalMappingApply(mat->cmap->mapping, ncol, icol, icolm));
2164: PetscCall(MatGetValues(mat, nrow, irowm, ncol, icolm, y));
2165: PetscCall(PetscFree2(bufr, bufc));
2166: }
2167: PetscCall(PetscLogEventEnd(MAT_GetValues, mat, 0, 0, 0));
2168: PetscFunctionReturn(PETSC_SUCCESS);
2169: }
2171: /*@
2172: MatSetValuesBatch - Adds (`ADD_VALUES`) many blocks of values into a matrix at once. The blocks must all be square and
2173: the same size. Currently, this can only be called once and creates the given matrix.
2175: Not Collective
2177: Input Parameters:
2178: + mat - the matrix
2179: . nb - the number of blocks
2180: . bs - the number of rows (and columns) in each block
2181: . rows - a concatenation of the rows for each block
2182: - v - a concatenation of logically two-dimensional arrays of values
2184: Level: advanced
2186: Notes:
2187: `MatSetPreallocationCOO()` and `MatSetValuesCOO()` may be a better way to provide the values
2189: In the future, we will extend this routine to handle rectangular blocks, and to allow multiple calls for a given matrix.
2191: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
2192: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`, `MatSetValues()`, `MatSetPreallocationCOO()`, `MatSetValuesCOO()`
2193: @*/
2194: PetscErrorCode MatSetValuesBatch(Mat mat, PetscInt nb, PetscInt bs, PetscInt rows[], const PetscScalar v[])
2195: {
2196: PetscFunctionBegin;
2199: PetscAssertPointer(rows, 4);
2200: PetscAssertPointer(v, 5);
2201: PetscAssert(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2203: PetscCall(PetscLogEventBegin(MAT_SetValuesBatch, mat, 0, 0, 0));
2204: if (mat->ops->setvaluesbatch) PetscUseTypeMethod(mat, setvaluesbatch, nb, bs, rows, v);
2205: else {
2206: for (PetscInt b = 0; b < nb; ++b) PetscCall(MatSetValues(mat, bs, &rows[b * bs], bs, &rows[b * bs], &v[b * bs * bs], ADD_VALUES));
2207: }
2208: PetscCall(PetscLogEventEnd(MAT_SetValuesBatch, mat, 0, 0, 0));
2209: PetscFunctionReturn(PETSC_SUCCESS);
2210: }
2212: /*@
2213: MatSetLocalToGlobalMapping - Sets a local-to-global numbering for use by
2214: the routine `MatSetValuesLocal()` to allow users to insert matrix entries
2215: using a local (per-processor) numbering.
2217: Not Collective
2219: Input Parameters:
2220: + x - the matrix
2221: . rmapping - row mapping created with `ISLocalToGlobalMappingCreate()` or `ISLocalToGlobalMappingCreateIS()`
2222: - cmapping - column mapping
2224: Level: intermediate
2226: Note:
2227: If the matrix is obtained with `DMCreateMatrix()` then this may already have been called on the matrix
2229: .seealso: [](ch_matrices), `Mat`, `DM`, `DMCreateMatrix()`, `MatGetLocalToGlobalMapping()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetValuesLocal()`, `MatGetValuesLocal()`
2230: @*/
2231: PetscErrorCode MatSetLocalToGlobalMapping(Mat x, ISLocalToGlobalMapping rmapping, ISLocalToGlobalMapping cmapping)
2232: {
2233: PetscFunctionBegin;
2238: if (x->ops->setlocaltoglobalmapping) PetscUseTypeMethod(x, setlocaltoglobalmapping, rmapping, cmapping);
2239: else {
2240: PetscCall(PetscLayoutSetISLocalToGlobalMapping(x->rmap, rmapping));
2241: PetscCall(PetscLayoutSetISLocalToGlobalMapping(x->cmap, cmapping));
2242: }
2243: PetscFunctionReturn(PETSC_SUCCESS);
2244: }
2246: /*@
2247: MatGetLocalToGlobalMapping - Gets the local-to-global numbering set by `MatSetLocalToGlobalMapping()`
2249: Not Collective
2251: Input Parameter:
2252: . A - the matrix
2254: Output Parameters:
2255: + rmapping - row mapping
2256: - cmapping - column mapping
2258: Level: advanced
2260: .seealso: [](ch_matrices), `Mat`, `MatSetLocalToGlobalMapping()`, `MatSetValuesLocal()`
2261: @*/
2262: PetscErrorCode MatGetLocalToGlobalMapping(Mat A, ISLocalToGlobalMapping *rmapping, ISLocalToGlobalMapping *cmapping)
2263: {
2264: PetscFunctionBegin;
2267: if (rmapping) {
2268: PetscAssertPointer(rmapping, 2);
2269: *rmapping = A->rmap->mapping;
2270: }
2271: if (cmapping) {
2272: PetscAssertPointer(cmapping, 3);
2273: *cmapping = A->cmap->mapping;
2274: }
2275: PetscFunctionReturn(PETSC_SUCCESS);
2276: }
2278: /*@
2279: MatSetLayouts - Sets the `PetscLayout` objects for rows and columns of a matrix
2281: Logically Collective
2283: Input Parameters:
2284: + A - the matrix
2285: . rmap - row layout
2286: - cmap - column layout
2288: Level: advanced
2290: Note:
2291: The `PetscLayout` objects are usually created automatically for the matrix so this routine rarely needs to be called.
2293: .seealso: [](ch_matrices), `Mat`, `PetscLayout`, `MatCreateVecs()`, `MatGetLocalToGlobalMapping()`, `MatGetLayouts()`
2294: @*/
2295: PetscErrorCode MatSetLayouts(Mat A, PetscLayout rmap, PetscLayout cmap)
2296: {
2297: PetscFunctionBegin;
2299: PetscCall(PetscLayoutReference(rmap, &A->rmap));
2300: PetscCall(PetscLayoutReference(cmap, &A->cmap));
2301: PetscFunctionReturn(PETSC_SUCCESS);
2302: }
2304: /*@
2305: MatGetLayouts - Gets the `PetscLayout` objects for rows and columns
2307: Not Collective
2309: Input Parameter:
2310: . A - the matrix
2312: Output Parameters:
2313: + rmap - row layout
2314: - cmap - column layout
2316: Level: advanced
2318: .seealso: [](ch_matrices), `Mat`, [Matrix Layouts](sec_matlayout), `PetscLayout`, `MatCreateVecs()`, `MatGetLocalToGlobalMapping()`, `MatSetLayouts()`
2319: @*/
2320: PetscErrorCode MatGetLayouts(Mat A, PetscLayout *rmap, PetscLayout *cmap)
2321: {
2322: PetscFunctionBegin;
2325: if (rmap) {
2326: PetscAssertPointer(rmap, 2);
2327: *rmap = A->rmap;
2328: }
2329: if (cmap) {
2330: PetscAssertPointer(cmap, 3);
2331: *cmap = A->cmap;
2332: }
2333: PetscFunctionReturn(PETSC_SUCCESS);
2334: }
2336: /*@C
2337: MatSetValuesLocal - Inserts or adds values into certain locations of a matrix,
2338: using a local numbering of the rows and columns.
2340: Not Collective
2342: Input Parameters:
2343: + mat - the matrix
2344: . nrow - number of rows
2345: . irow - the row local indices
2346: . ncol - number of columns
2347: . icol - the column local indices
2348: . y - a logically two-dimensional array of values
2349: - addv - either `INSERT_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
2351: Level: intermediate
2353: Notes:
2354: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call `MatSetLocalToGlobalMapping()` before using this routine
2356: Calls to `MatSetValuesLocal()` with the `INSERT_VALUES` and `ADD_VALUES`
2357: options cannot be mixed without intervening calls to the assembly
2358: routines.
2360: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
2361: MUST be called after all calls to `MatSetValuesLocal()` have been completed.
2363: Developer Note:
2364: This is labeled with C so does not automatically generate Fortran stubs and interfaces
2365: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2367: .seealso: [](ch_matrices), `Mat`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetLocalToGlobalMapping()`,
2368: `MatGetValuesLocal()`
2369: @*/
2370: PetscErrorCode MatSetValuesLocal(Mat mat, PetscInt nrow, const PetscInt irow[], PetscInt ncol, const PetscInt icol[], const PetscScalar y[], InsertMode addv)
2371: {
2372: PetscFunctionBeginHot;
2375: MatCheckPreallocated(mat, 1);
2376: if (!nrow || !ncol) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
2377: PetscAssertPointer(irow, 3);
2378: PetscAssertPointer(icol, 5);
2379: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
2380: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
2381: if (PetscDefined(USE_DEBUG)) {
2382: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2383: PetscCheck(mat->ops->setvalueslocal || mat->ops->setvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2384: }
2386: if (mat->assembled) {
2387: mat->was_assembled = PETSC_TRUE;
2388: mat->assembled = PETSC_FALSE;
2389: }
2390: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
2391: if (mat->ops->setvalueslocal) PetscUseTypeMethod(mat, setvalueslocal, nrow, irow, ncol, icol, y, addv);
2392: else {
2393: PetscInt buf[8192], *bufr = NULL, *bufc = NULL;
2394: const PetscInt *irowm, *icolm;
2396: if ((!mat->rmap->mapping && !mat->cmap->mapping) || (nrow + ncol) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
2397: bufr = buf;
2398: bufc = buf + nrow;
2399: irowm = bufr;
2400: icolm = bufc;
2401: } else {
2402: PetscCall(PetscMalloc2(nrow, &bufr, ncol, &bufc));
2403: irowm = bufr;
2404: icolm = bufc;
2405: }
2406: if (mat->rmap->mapping) PetscCall(ISLocalToGlobalMappingApply(mat->rmap->mapping, nrow, irow, bufr));
2407: else irowm = irow;
2408: if (mat->cmap->mapping) {
2409: if (mat->cmap->mapping != mat->rmap->mapping || ncol != nrow || icol != irow) {
2410: PetscCall(ISLocalToGlobalMappingApply(mat->cmap->mapping, ncol, icol, bufc));
2411: } else icolm = irowm;
2412: } else icolm = icol;
2413: PetscCall(MatSetValues(mat, nrow, irowm, ncol, icolm, y, addv));
2414: if (bufr != buf) PetscCall(PetscFree2(bufr, bufc));
2415: }
2416: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
2417: PetscFunctionReturn(PETSC_SUCCESS);
2418: }
2420: /*@C
2421: MatSetValuesBlockedLocal - Inserts or adds values into certain locations of a matrix,
2422: using a local ordering of the nodes a block at a time.
2424: Not Collective
2426: Input Parameters:
2427: + mat - the matrix
2428: . nrow - number of rows
2429: . irow - the row local indices
2430: . ncol - number of columns
2431: . icol - the column local indices
2432: . y - a logically two-dimensional array of values
2433: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
2435: Level: intermediate
2437: Notes:
2438: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call `MatSetBlockSize()` and `MatSetLocalToGlobalMapping()`
2439: before using this routineBefore calling `MatSetValuesLocal()`, the user must first set the
2441: Calls to `MatSetValuesBlockedLocal()` with the `INSERT_VALUES` and `ADD_VALUES`
2442: options cannot be mixed without intervening calls to the assembly
2443: routines.
2445: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
2446: MUST be called after all calls to `MatSetValuesBlockedLocal()` have been completed.
2448: Developer Note:
2449: This is labeled with C so does not automatically generate Fortran stubs and interfaces
2450: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2452: .seealso: [](ch_matrices), `Mat`, `MatSetBlockSize()`, `MatSetLocalToGlobalMapping()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`,
2453: `MatSetValuesLocal()`, `MatSetValuesBlocked()`
2454: @*/
2455: PetscErrorCode MatSetValuesBlockedLocal(Mat mat, PetscInt nrow, const PetscInt irow[], PetscInt ncol, const PetscInt icol[], const PetscScalar y[], InsertMode addv)
2456: {
2457: PetscFunctionBeginHot;
2460: MatCheckPreallocated(mat, 1);
2461: if (!nrow || !ncol) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
2462: PetscAssertPointer(irow, 3);
2463: PetscAssertPointer(icol, 5);
2464: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
2465: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
2466: if (PetscDefined(USE_DEBUG)) {
2467: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2468: 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);
2469: }
2471: if (mat->assembled) {
2472: mat->was_assembled = PETSC_TRUE;
2473: mat->assembled = PETSC_FALSE;
2474: }
2475: if (PetscUnlikelyDebug(mat->rmap->mapping)) { /* Condition on the mapping existing, because MatSetValuesBlockedLocal_IS does not require it to be set. */
2476: PetscInt irbs, rbs;
2477: PetscCall(MatGetBlockSizes(mat, &rbs, NULL));
2478: PetscCall(ISLocalToGlobalMappingGetBlockSize(mat->rmap->mapping, &irbs));
2479: PetscCheck(rbs == irbs, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Different row block sizes! mat %" PetscInt_FMT ", row l2g map %" PetscInt_FMT, rbs, irbs);
2480: }
2481: if (PetscUnlikelyDebug(mat->cmap->mapping)) {
2482: PetscInt icbs, cbs;
2483: PetscCall(MatGetBlockSizes(mat, NULL, &cbs));
2484: PetscCall(ISLocalToGlobalMappingGetBlockSize(mat->cmap->mapping, &icbs));
2485: PetscCheck(cbs == icbs, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Different col block sizes! mat %" PetscInt_FMT ", col l2g map %" PetscInt_FMT, cbs, icbs);
2486: }
2487: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
2488: if (mat->ops->setvaluesblockedlocal) PetscUseTypeMethod(mat, setvaluesblockedlocal, nrow, irow, ncol, icol, y, addv);
2489: else {
2490: PetscInt buf[8192], *bufr = NULL, *bufc = NULL;
2491: const PetscInt *irowm, *icolm;
2493: if ((!mat->rmap->mapping && !mat->cmap->mapping) || (nrow + ncol) <= ((PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf))) {
2494: bufr = buf;
2495: bufc = buf + nrow;
2496: irowm = bufr;
2497: icolm = bufc;
2498: } else {
2499: PetscCall(PetscMalloc2(nrow, &bufr, ncol, &bufc));
2500: irowm = bufr;
2501: icolm = bufc;
2502: }
2503: if (mat->rmap->mapping) PetscCall(ISLocalToGlobalMappingApplyBlock(mat->rmap->mapping, nrow, irow, bufr));
2504: else irowm = irow;
2505: if (mat->cmap->mapping) {
2506: if (mat->cmap->mapping != mat->rmap->mapping || ncol != nrow || icol != irow) {
2507: PetscCall(ISLocalToGlobalMappingApplyBlock(mat->cmap->mapping, ncol, icol, bufc));
2508: } else icolm = irowm;
2509: } else icolm = icol;
2510: PetscCall(MatSetValuesBlocked(mat, nrow, irowm, ncol, icolm, y, addv));
2511: if (bufr != buf) PetscCall(PetscFree2(bufr, bufc));
2512: }
2513: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
2514: PetscFunctionReturn(PETSC_SUCCESS);
2515: }
2517: /*@
2518: MatMultDiagonalBlock - Computes the matrix-vector product, $y = Dx$. Where `D` is defined by the inode or block structure of the diagonal
2520: Collective
2522: Input Parameters:
2523: + mat - the matrix
2524: - x - the vector to be multiplied
2526: Output Parameter:
2527: . y - the result
2529: Level: developer
2531: Note:
2532: The vectors `x` and `y` cannot be the same. I.e., one cannot
2533: call `MatMultDiagonalBlock`(A,y,y).
2535: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
2536: @*/
2537: PetscErrorCode MatMultDiagonalBlock(Mat mat, Vec x, Vec y)
2538: {
2539: PetscFunctionBegin;
2545: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2546: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2547: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2548: MatCheckPreallocated(mat, 1);
2550: PetscUseTypeMethod(mat, multdiagonalblock, x, y);
2551: PetscCall(PetscObjectStateIncrease((PetscObject)y));
2552: PetscFunctionReturn(PETSC_SUCCESS);
2553: }
2555: /*@
2556: MatMult - Computes the matrix-vector product, $y = Ax$.
2558: Neighbor-wise Collective
2560: Input Parameters:
2561: + mat - the matrix
2562: - x - the vector to be multiplied
2564: Output Parameter:
2565: . y - the result
2567: Level: beginner
2569: Note:
2570: The vectors `x` and `y` cannot be the same. I.e., one cannot
2571: call `MatMult`(A,y,y).
2573: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
2574: @*/
2575: PetscErrorCode MatMult(Mat mat, Vec x, Vec y)
2576: {
2577: PetscFunctionBegin;
2581: VecCheckAssembled(x);
2583: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2584: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2585: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2586: 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);
2587: 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);
2588: 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);
2589: 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);
2590: PetscCall(VecSetErrorIfLocked(y, 3));
2591: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(x, 2, PETSC_TRUE));
2592: MatCheckPreallocated(mat, 1);
2594: PetscCall(VecLockReadPush(x));
2595: PetscCall(PetscLogEventBegin(MAT_Mult, mat, x, y, 0));
2596: PetscUseTypeMethod(mat, mult, x, y);
2597: PetscCall(PetscLogEventEnd(MAT_Mult, mat, x, y, 0));
2598: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(y, 3, PETSC_FALSE));
2599: PetscCall(VecLockReadPop(x));
2600: PetscFunctionReturn(PETSC_SUCCESS);
2601: }
2603: /*@
2604: MatMultTranspose - Computes matrix transpose times a vector $y = A^T * x$.
2606: Neighbor-wise Collective
2608: Input Parameters:
2609: + mat - the matrix
2610: - x - the vector to be multiplied
2612: Output Parameter:
2613: . y - the result
2615: Level: beginner
2617: Notes:
2618: The vectors `x` and `y` cannot be the same. I.e., one cannot
2619: call `MatMultTranspose`(A,y,y).
2621: For complex numbers this does NOT compute the Hermitian (complex conjugate) transpose multiple,
2622: use `MatMultHermitianTranspose()`
2624: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatMultHermitianTranspose()`, `MatTranspose()`
2625: @*/
2626: PetscErrorCode MatMultTranspose(Mat mat, Vec x, Vec y)
2627: {
2628: PetscErrorCode (*op)(Mat, Vec, Vec) = NULL;
2630: PetscFunctionBegin;
2634: VecCheckAssembled(x);
2637: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2638: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2639: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2640: 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);
2641: 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);
2642: 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);
2643: 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);
2644: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(x, 2, PETSC_TRUE));
2645: MatCheckPreallocated(mat, 1);
2647: if (!mat->ops->multtranspose) {
2648: if (mat->symmetric == PETSC_BOOL3_TRUE && mat->ops->mult) op = mat->ops->mult;
2649: 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);
2650: } else op = mat->ops->multtranspose;
2651: PetscCall(PetscLogEventBegin(MAT_MultTranspose, mat, x, y, 0));
2652: PetscCall(VecLockReadPush(x));
2653: PetscCall((*op)(mat, x, y));
2654: PetscCall(VecLockReadPop(x));
2655: PetscCall(PetscLogEventEnd(MAT_MultTranspose, mat, x, y, 0));
2656: PetscCall(PetscObjectStateIncrease((PetscObject)y));
2657: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(y, 3, PETSC_FALSE));
2658: PetscFunctionReturn(PETSC_SUCCESS);
2659: }
2661: /*@
2662: MatMultHermitianTranspose - Computes matrix Hermitian-transpose times a vector $y = A^H * x$.
2664: Neighbor-wise Collective
2666: Input Parameters:
2667: + mat - the matrix
2668: - x - the vector to be multiplied
2670: Output Parameter:
2671: . y - the result
2673: Level: beginner
2675: Notes:
2676: The vectors `x` and `y` cannot be the same. I.e., one cannot
2677: call `MatMultHermitianTranspose`(A,y,y).
2679: Also called the conjugate transpose, complex conjugate transpose, or adjoint.
2681: For real numbers `MatMultTranspose()` and `MatMultHermitianTranspose()` are identical.
2683: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatMultAdd()`, `MatMultHermitianTransposeAdd()`, `MatMultTranspose()`
2684: @*/
2685: PetscErrorCode MatMultHermitianTranspose(Mat mat, Vec x, Vec y)
2686: {
2687: PetscFunctionBegin;
2693: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2694: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2695: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2696: 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);
2697: 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);
2698: 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);
2699: 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);
2700: MatCheckPreallocated(mat, 1);
2702: PetscCall(PetscLogEventBegin(MAT_MultHermitianTranspose, mat, x, y, 0));
2703: #if defined(PETSC_USE_COMPLEX)
2704: if (mat->ops->multhermitiantranspose || (mat->hermitian == PETSC_BOOL3_TRUE && mat->ops->mult)) {
2705: PetscCall(VecLockReadPush(x));
2706: if (mat->ops->multhermitiantranspose) PetscUseTypeMethod(mat, multhermitiantranspose, x, y);
2707: else PetscUseTypeMethod(mat, mult, x, y);
2708: PetscCall(VecLockReadPop(x));
2709: } else {
2710: Vec w;
2711: PetscCall(VecDuplicate(x, &w));
2712: PetscCall(VecCopy(x, w));
2713: PetscCall(VecConjugate(w));
2714: PetscCall(MatMultTranspose(mat, w, y));
2715: PetscCall(VecDestroy(&w));
2716: PetscCall(VecConjugate(y));
2717: }
2718: PetscCall(PetscObjectStateIncrease((PetscObject)y));
2719: #else
2720: PetscCall(MatMultTranspose(mat, x, y));
2721: #endif
2722: PetscCall(PetscLogEventEnd(MAT_MultHermitianTranspose, mat, x, y, 0));
2723: PetscFunctionReturn(PETSC_SUCCESS);
2724: }
2726: /*@
2727: MatMultAdd - Computes $v3 = v2 + A * v1$.
2729: Neighbor-wise Collective
2731: Input Parameters:
2732: + mat - the matrix
2733: . v1 - the vector to be multiplied by `mat`
2734: - v2 - the vector to be added to the result
2736: Output Parameter:
2737: . v3 - the result
2739: Level: beginner
2741: Note:
2742: The vectors `v1` and `v3` cannot be the same. I.e., one cannot
2743: call `MatMultAdd`(A,v1,v2,v1).
2745: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMult()`, `MatMultTransposeAdd()`
2746: @*/
2747: PetscErrorCode MatMultAdd(Mat mat, Vec v1, Vec v2, Vec v3)
2748: {
2749: PetscFunctionBegin;
2756: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2757: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2758: 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);
2759: /* 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);
2760: 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); */
2761: 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);
2762: 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);
2763: PetscCheck(v1 != v3, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "v1 and v3 must be different vectors");
2764: MatCheckPreallocated(mat, 1);
2766: PetscCall(PetscLogEventBegin(MAT_MultAdd, mat, v1, v2, v3));
2767: PetscCall(VecLockReadPush(v1));
2768: PetscUseTypeMethod(mat, multadd, v1, v2, v3);
2769: PetscCall(VecLockReadPop(v1));
2770: PetscCall(PetscLogEventEnd(MAT_MultAdd, mat, v1, v2, v3));
2771: PetscCall(PetscObjectStateIncrease((PetscObject)v3));
2772: PetscFunctionReturn(PETSC_SUCCESS);
2773: }
2775: /*@
2776: MatMultTransposeAdd - Computes $v3 = v2 + A^T * v1$.
2778: Neighbor-wise Collective
2780: Input Parameters:
2781: + mat - the matrix
2782: . v1 - the vector to be multiplied by the transpose of the matrix
2783: - v2 - the vector to be added to the result
2785: Output Parameter:
2786: . v3 - the result
2788: Level: beginner
2790: Note:
2791: The vectors `v1` and `v3` cannot be the same. I.e., one cannot
2792: call `MatMultTransposeAdd`(A,v1,v2,v1).
2794: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMult()`
2795: @*/
2796: PetscErrorCode MatMultTransposeAdd(Mat mat, Vec v1, Vec v2, Vec v3)
2797: {
2798: PetscErrorCode (*op)(Mat, Vec, Vec, Vec) = (!mat->ops->multtransposeadd && mat->symmetric) ? mat->ops->multadd : mat->ops->multtransposeadd;
2800: PetscFunctionBegin;
2807: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2808: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2809: 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);
2810: 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);
2811: 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);
2812: PetscCheck(v1 != v3, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "v1 and v3 must be different vectors");
2813: PetscCheck(op, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2814: MatCheckPreallocated(mat, 1);
2816: PetscCall(PetscLogEventBegin(MAT_MultTransposeAdd, mat, v1, v2, v3));
2817: PetscCall(VecLockReadPush(v1));
2818: PetscCall((*op)(mat, v1, v2, v3));
2819: PetscCall(VecLockReadPop(v1));
2820: PetscCall(PetscLogEventEnd(MAT_MultTransposeAdd, mat, v1, v2, v3));
2821: PetscCall(PetscObjectStateIncrease((PetscObject)v3));
2822: PetscFunctionReturn(PETSC_SUCCESS);
2823: }
2825: /*@
2826: MatMultHermitianTransposeAdd - Computes $v3 = v2 + A^H * v1$.
2828: Neighbor-wise Collective
2830: Input Parameters:
2831: + mat - the matrix
2832: . v1 - the vector to be multiplied by the Hermitian transpose
2833: - v2 - the vector to be added to the result
2835: Output Parameter:
2836: . v3 - the result
2838: Level: beginner
2840: Note:
2841: The vectors `v1` and `v3` cannot be the same. I.e., one cannot
2842: call `MatMultHermitianTransposeAdd`(A,v1,v2,v1).
2844: .seealso: [](ch_matrices), `Mat`, `MatMultHermitianTranspose()`, `MatMultTranspose()`, `MatMultAdd()`, `MatMult()`
2845: @*/
2846: PetscErrorCode MatMultHermitianTransposeAdd(Mat mat, Vec v1, Vec v2, Vec v3)
2847: {
2848: PetscFunctionBegin;
2855: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2856: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2857: PetscCheck(v1 != v3, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "v1 and v3 must be different vectors");
2858: 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);
2859: 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);
2860: 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);
2861: MatCheckPreallocated(mat, 1);
2863: PetscCall(PetscLogEventBegin(MAT_MultHermitianTransposeAdd, mat, v1, v2, v3));
2864: PetscCall(VecLockReadPush(v1));
2865: if (mat->ops->multhermitiantransposeadd) PetscUseTypeMethod(mat, multhermitiantransposeadd, v1, v2, v3);
2866: else {
2867: Vec w, z;
2868: PetscCall(VecDuplicate(v1, &w));
2869: PetscCall(VecCopy(v1, w));
2870: PetscCall(VecConjugate(w));
2871: PetscCall(VecDuplicate(v3, &z));
2872: PetscCall(MatMultTranspose(mat, w, z));
2873: PetscCall(VecDestroy(&w));
2874: PetscCall(VecConjugate(z));
2875: if (v2 != v3) {
2876: PetscCall(VecWAXPY(v3, 1.0, v2, z));
2877: } else {
2878: PetscCall(VecAXPY(v3, 1.0, z));
2879: }
2880: PetscCall(VecDestroy(&z));
2881: }
2882: PetscCall(VecLockReadPop(v1));
2883: PetscCall(PetscLogEventEnd(MAT_MultHermitianTransposeAdd, mat, v1, v2, v3));
2884: PetscCall(PetscObjectStateIncrease((PetscObject)v3));
2885: PetscFunctionReturn(PETSC_SUCCESS);
2886: }
2888: /*@C
2889: MatGetFactorType - gets the type of factorization a matrix is
2891: Not Collective
2893: Input Parameter:
2894: . mat - the matrix
2896: Output Parameter:
2897: . 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`
2899: Level: intermediate
2901: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorType`, `MatGetFactor()`, `MatSetFactorType()`, `MAT_FACTOR_NONE`, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ILU`,
2902: `MAT_FACTOR_ICC`,`MAT_FACTOR_ILUDT`, `MAT_FACTOR_QR`
2903: @*/
2904: PetscErrorCode MatGetFactorType(Mat mat, MatFactorType *t)
2905: {
2906: PetscFunctionBegin;
2909: PetscAssertPointer(t, 2);
2910: *t = mat->factortype;
2911: PetscFunctionReturn(PETSC_SUCCESS);
2912: }
2914: /*@C
2915: MatSetFactorType - sets the type of factorization a matrix is
2917: Logically Collective
2919: Input Parameters:
2920: + mat - the matrix
2921: - 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`
2923: Level: intermediate
2925: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorType`, `MatGetFactor()`, `MatGetFactorType()`, `MAT_FACTOR_NONE`, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ILU`,
2926: `MAT_FACTOR_ICC`,`MAT_FACTOR_ILUDT`, `MAT_FACTOR_QR`
2927: @*/
2928: PetscErrorCode MatSetFactorType(Mat mat, MatFactorType t)
2929: {
2930: PetscFunctionBegin;
2933: mat->factortype = t;
2934: PetscFunctionReturn(PETSC_SUCCESS);
2935: }
2937: /*@C
2938: MatGetInfo - Returns information about matrix storage (number of
2939: nonzeros, memory, etc.).
2941: Collective if `MAT_GLOBAL_MAX` or `MAT_GLOBAL_SUM` is used as the flag
2943: Input Parameters:
2944: + mat - the matrix
2945: - 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)
2947: Output Parameter:
2948: . info - matrix information context
2950: Options Database Key:
2951: . -mat_view ::ascii_info - print matrix info to `PETSC_STDOUT`
2953: Notes:
2954: The `MatInfo` context contains a variety of matrix data, including
2955: number of nonzeros allocated and used, number of mallocs during
2956: matrix assembly, etc. Additional information for factored matrices
2957: is provided (such as the fill ratio, number of mallocs during
2958: factorization, etc.).
2960: Example:
2961: See the file ${PETSC_DIR}/include/petscmat.h for a complete list of
2962: data within the MatInfo context. For example,
2963: .vb
2964: MatInfo info;
2965: Mat A;
2966: double mal, nz_a, nz_u;
2968: MatGetInfo(A, MAT_LOCAL, &info);
2969: mal = info.mallocs;
2970: nz_a = info.nz_allocated;
2971: .ve
2973: Fortran users should declare info as a double precision
2974: array of dimension `MAT_INFO_SIZE`, and then extract the parameters
2975: of interest. See the file ${PETSC_DIR}/include/petsc/finclude/petscmat.h
2976: a complete list of parameter names.
2977: .vb
2978: double precision info(MAT_INFO_SIZE)
2979: double precision mal, nz_a
2980: Mat A
2981: integer ierr
2983: call MatGetInfo(A, MAT_LOCAL, info, ierr)
2984: mal = info(MAT_INFO_MALLOCS)
2985: nz_a = info(MAT_INFO_NZ_ALLOCATED)
2986: .ve
2988: Level: intermediate
2990: Developer Note:
2991: The Fortran interface is not autogenerated as the
2992: interface definition cannot be generated correctly [due to `MatInfo` argument]
2994: .seealso: [](ch_matrices), `Mat`, `MatInfo`, `MatStashGetInfo()`
2995: @*/
2996: PetscErrorCode MatGetInfo(Mat mat, MatInfoType flag, MatInfo *info)
2997: {
2998: PetscFunctionBegin;
3001: PetscAssertPointer(info, 3);
3002: MatCheckPreallocated(mat, 1);
3003: PetscUseTypeMethod(mat, getinfo, flag, info);
3004: PetscFunctionReturn(PETSC_SUCCESS);
3005: }
3007: /*
3008: This is used by external packages where it is not easy to get the info from the actual
3009: matrix factorization.
3010: */
3011: PetscErrorCode MatGetInfo_External(Mat A, MatInfoType flag, MatInfo *info)
3012: {
3013: PetscFunctionBegin;
3014: PetscCall(PetscMemzero(info, sizeof(MatInfo)));
3015: PetscFunctionReturn(PETSC_SUCCESS);
3016: }
3018: /*@C
3019: MatLUFactor - Performs in-place LU factorization of matrix.
3021: Collective
3023: Input Parameters:
3024: + mat - the matrix
3025: . row - row permutation
3026: . col - column permutation
3027: - info - options for factorization, includes
3028: .vb
3029: fill - expected fill as ratio of original fill.
3030: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3031: Run with the option -info to determine an optimal value to use
3032: .ve
3034: Level: developer
3036: Notes:
3037: Most users should employ the `KSP` interface for linear solvers
3038: instead of working directly with matrix algebra routines such as this.
3039: See, e.g., `KSPCreate()`.
3041: This changes the state of the matrix to a factored matrix; it cannot be used
3042: for example with `MatSetValues()` unless one first calls `MatSetUnfactored()`.
3044: This is really in-place only for dense matrices, the preferred approach is to use `MatGetFactor()`, `MatLUFactorSymbolic()`, and `MatLUFactorNumeric()`
3045: when not using `KSP`.
3047: Developer Note:
3048: The Fortran interface is not autogenerated as the
3049: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3051: .seealso: [](ch_matrices), [Matrix Factorization](sec_matfactor), `Mat`, `MatFactorType`, `MatLUFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`,
3052: `MatGetOrdering()`, `MatSetUnfactored()`, `MatFactorInfo`, `MatGetFactor()`
3053: @*/
3054: PetscErrorCode MatLUFactor(Mat mat, IS row, IS col, const MatFactorInfo *info)
3055: {
3056: MatFactorInfo tinfo;
3058: PetscFunctionBegin;
3062: if (info) PetscAssertPointer(info, 4);
3064: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3065: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3066: MatCheckPreallocated(mat, 1);
3067: if (!info) {
3068: PetscCall(MatFactorInfoInitialize(&tinfo));
3069: info = &tinfo;
3070: }
3072: PetscCall(PetscLogEventBegin(MAT_LUFactor, mat, row, col, 0));
3073: PetscUseTypeMethod(mat, lufactor, row, col, info);
3074: PetscCall(PetscLogEventEnd(MAT_LUFactor, mat, row, col, 0));
3075: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3076: PetscFunctionReturn(PETSC_SUCCESS);
3077: }
3079: /*@C
3080: MatILUFactor - Performs in-place ILU factorization of matrix.
3082: Collective
3084: Input Parameters:
3085: + mat - the matrix
3086: . row - row permutation
3087: . col - column permutation
3088: - info - structure containing
3089: .vb
3090: levels - number of levels of fill.
3091: expected fill - as ratio of original fill.
3092: 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
3093: missing diagonal entries)
3094: .ve
3096: Level: developer
3098: Notes:
3099: Most users should employ the `KSP` interface for linear solvers
3100: instead of working directly with matrix algebra routines such as this.
3101: See, e.g., `KSPCreate()`.
3103: Probably really in-place only when level of fill is zero, otherwise allocates
3104: new space to store factored matrix and deletes previous memory. The preferred approach is to use `MatGetFactor()`, `MatILUFactorSymbolic()`, and `MatILUFactorNumeric()`
3105: when not using `KSP`.
3107: Developer Note:
3108: The Fortran interface is not autogenerated as the
3109: interface definition cannot be generated correctly [due to MatFactorInfo]
3111: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatILUFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`, `MatFactorInfo`
3112: @*/
3113: PetscErrorCode MatILUFactor(Mat mat, IS row, IS col, const MatFactorInfo *info)
3114: {
3115: PetscFunctionBegin;
3119: PetscAssertPointer(info, 4);
3121: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "matrix must be square");
3122: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3123: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3124: MatCheckPreallocated(mat, 1);
3126: PetscCall(PetscLogEventBegin(MAT_ILUFactor, mat, row, col, 0));
3127: PetscUseTypeMethod(mat, ilufactor, row, col, info);
3128: PetscCall(PetscLogEventEnd(MAT_ILUFactor, mat, row, col, 0));
3129: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3130: PetscFunctionReturn(PETSC_SUCCESS);
3131: }
3133: /*@C
3134: MatLUFactorSymbolic - Performs symbolic LU factorization of matrix.
3135: Call this routine before calling `MatLUFactorNumeric()` and after `MatGetFactor()`.
3137: Collective
3139: Input Parameters:
3140: + fact - the factor matrix obtained with `MatGetFactor()`
3141: . mat - the matrix
3142: . row - the row permutation
3143: . col - the column permutation
3144: - info - options for factorization, includes
3145: .vb
3146: fill - expected fill as ratio of original fill. Run with the option -info to determine an optimal value to use
3147: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3148: .ve
3150: Level: developer
3152: Notes:
3153: See [Matrix Factorization](sec_matfactor) for additional information about factorizations
3155: Most users should employ the simplified `KSP` interface for linear solvers
3156: instead of working directly with matrix algebra routines such as this.
3157: See, e.g., `KSPCreate()`.
3159: Developer Note:
3160: The Fortran interface is not autogenerated as the
3161: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3163: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatLUFactor()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`, `MatFactorInfo`, `MatFactorInfoInitialize()`
3164: @*/
3165: PetscErrorCode MatLUFactorSymbolic(Mat fact, Mat mat, IS row, IS col, const MatFactorInfo *info)
3166: {
3167: MatFactorInfo tinfo;
3169: PetscFunctionBegin;
3174: if (info) PetscAssertPointer(info, 5);
3177: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3178: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3179: MatCheckPreallocated(mat, 2);
3180: if (!info) {
3181: PetscCall(MatFactorInfoInitialize(&tinfo));
3182: info = &tinfo;
3183: }
3185: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_LUFactorSymbolic, mat, row, col, 0));
3186: PetscUseTypeMethod(fact, lufactorsymbolic, mat, row, col, info);
3187: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_LUFactorSymbolic, mat, row, col, 0));
3188: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3189: PetscFunctionReturn(PETSC_SUCCESS);
3190: }
3192: /*@C
3193: MatLUFactorNumeric - Performs numeric LU factorization of a matrix.
3194: Call this routine after first calling `MatLUFactorSymbolic()` and `MatGetFactor()`.
3196: Collective
3198: Input Parameters:
3199: + fact - the factor matrix obtained with `MatGetFactor()`
3200: . mat - the matrix
3201: - info - options for factorization
3203: Level: developer
3205: Notes:
3206: See `MatLUFactor()` for in-place factorization. See
3207: `MatCholeskyFactorNumeric()` for the symmetric, positive definite case.
3209: Most users should employ the `KSP` interface for linear solvers
3210: instead of working directly with matrix algebra routines such as this.
3211: See, e.g., `KSPCreate()`.
3213: Developer Note:
3214: The Fortran interface is not autogenerated as the
3215: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3217: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatFactorInfo`, `MatLUFactorSymbolic()`, `MatLUFactor()`, `MatCholeskyFactor()`
3218: @*/
3219: PetscErrorCode MatLUFactorNumeric(Mat fact, Mat mat, const MatFactorInfo *info)
3220: {
3221: MatFactorInfo tinfo;
3223: PetscFunctionBegin;
3228: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3229: 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,
3230: mat->rmap->N, (fact)->rmap->N, mat->cmap->N, (fact)->cmap->N);
3232: MatCheckPreallocated(mat, 2);
3233: if (!info) {
3234: PetscCall(MatFactorInfoInitialize(&tinfo));
3235: info = &tinfo;
3236: }
3238: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_LUFactorNumeric, mat, fact, 0, 0));
3239: else PetscCall(PetscLogEventBegin(MAT_LUFactor, mat, fact, 0, 0));
3240: PetscUseTypeMethod(fact, lufactornumeric, mat, info);
3241: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_LUFactorNumeric, mat, fact, 0, 0));
3242: else PetscCall(PetscLogEventEnd(MAT_LUFactor, mat, fact, 0, 0));
3243: PetscCall(MatViewFromOptions(fact, NULL, "-mat_factor_view"));
3244: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3245: PetscFunctionReturn(PETSC_SUCCESS);
3246: }
3248: /*@C
3249: MatCholeskyFactor - Performs in-place Cholesky factorization of a
3250: symmetric matrix.
3252: Collective
3254: Input Parameters:
3255: + mat - the matrix
3256: . perm - row and column permutations
3257: - info - expected fill as ratio of original fill
3259: Level: developer
3261: Notes:
3262: See `MatLUFactor()` for the nonsymmetric case. See also `MatGetFactor()`,
3263: `MatCholeskyFactorSymbolic()`, and `MatCholeskyFactorNumeric()`.
3265: Most users should employ the `KSP` interface for linear solvers
3266: instead of working directly with matrix algebra routines such as this.
3267: See, e.g., `KSPCreate()`.
3269: Developer Note:
3270: The Fortran interface is not autogenerated as the
3271: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3273: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatFactorInfo`, `MatLUFactor()`, `MatCholeskyFactorSymbolic()`, `MatCholeskyFactorNumeric()`
3274: `MatGetOrdering()`
3275: @*/
3276: PetscErrorCode MatCholeskyFactor(Mat mat, IS perm, const MatFactorInfo *info)
3277: {
3278: MatFactorInfo tinfo;
3280: PetscFunctionBegin;
3283: if (info) PetscAssertPointer(info, 3);
3285: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "Matrix must be square");
3286: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3287: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3288: MatCheckPreallocated(mat, 1);
3289: if (!info) {
3290: PetscCall(MatFactorInfoInitialize(&tinfo));
3291: info = &tinfo;
3292: }
3294: PetscCall(PetscLogEventBegin(MAT_CholeskyFactor, mat, perm, 0, 0));
3295: PetscUseTypeMethod(mat, choleskyfactor, perm, info);
3296: PetscCall(PetscLogEventEnd(MAT_CholeskyFactor, mat, perm, 0, 0));
3297: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3298: PetscFunctionReturn(PETSC_SUCCESS);
3299: }
3301: /*@C
3302: MatCholeskyFactorSymbolic - Performs symbolic Cholesky factorization
3303: of a symmetric matrix.
3305: Collective
3307: Input Parameters:
3308: + fact - the factor matrix obtained with `MatGetFactor()`
3309: . mat - the matrix
3310: . perm - row and column permutations
3311: - info - options for factorization, includes
3312: .vb
3313: fill - expected fill as ratio of original fill.
3314: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3315: Run with the option -info to determine an optimal value to use
3316: .ve
3318: Level: developer
3320: Notes:
3321: See `MatLUFactorSymbolic()` for the nonsymmetric case. See also
3322: `MatCholeskyFactor()` and `MatCholeskyFactorNumeric()`.
3324: Most users should employ the `KSP` interface for linear solvers
3325: instead of working directly with matrix algebra routines such as this.
3326: See, e.g., `KSPCreate()`.
3328: Developer Note:
3329: The Fortran interface is not autogenerated as the
3330: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3332: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactor()`, `MatCholeskyFactorNumeric()`
3333: `MatGetOrdering()`
3334: @*/
3335: PetscErrorCode MatCholeskyFactorSymbolic(Mat fact, Mat mat, IS perm, const MatFactorInfo *info)
3336: {
3337: MatFactorInfo tinfo;
3339: PetscFunctionBegin;
3343: if (info) PetscAssertPointer(info, 4);
3346: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "Matrix must be square");
3347: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3348: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3349: MatCheckPreallocated(mat, 2);
3350: if (!info) {
3351: PetscCall(MatFactorInfoInitialize(&tinfo));
3352: info = &tinfo;
3353: }
3355: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_CholeskyFactorSymbolic, mat, perm, 0, 0));
3356: PetscUseTypeMethod(fact, choleskyfactorsymbolic, mat, perm, info);
3357: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_CholeskyFactorSymbolic, mat, perm, 0, 0));
3358: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3359: PetscFunctionReturn(PETSC_SUCCESS);
3360: }
3362: /*@C
3363: MatCholeskyFactorNumeric - Performs numeric Cholesky factorization
3364: of a symmetric matrix. Call this routine after first calling `MatGetFactor()` and
3365: `MatCholeskyFactorSymbolic()`.
3367: Collective
3369: Input Parameters:
3370: + fact - the factor matrix obtained with `MatGetFactor()`, where the factored values are stored
3371: . mat - the initial matrix that is to be factored
3372: - info - options for factorization
3374: Level: developer
3376: Note:
3377: Most users should employ the `KSP` interface for linear solvers
3378: instead of working directly with matrix algebra routines such as this.
3379: See, e.g., `KSPCreate()`.
3381: Developer Note:
3382: The Fortran interface is not autogenerated as the
3383: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3385: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatCholeskyFactorSymbolic()`, `MatCholeskyFactor()`, `MatLUFactorNumeric()`
3386: @*/
3387: PetscErrorCode MatCholeskyFactorNumeric(Mat fact, Mat mat, const MatFactorInfo *info)
3388: {
3389: MatFactorInfo tinfo;
3391: PetscFunctionBegin;
3396: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3397: 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,
3398: mat->rmap->N, (fact)->rmap->N, mat->cmap->N, (fact)->cmap->N);
3399: MatCheckPreallocated(mat, 2);
3400: if (!info) {
3401: PetscCall(MatFactorInfoInitialize(&tinfo));
3402: info = &tinfo;
3403: }
3405: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_CholeskyFactorNumeric, mat, fact, 0, 0));
3406: else PetscCall(PetscLogEventBegin(MAT_CholeskyFactor, mat, fact, 0, 0));
3407: PetscUseTypeMethod(fact, choleskyfactornumeric, mat, info);
3408: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_CholeskyFactorNumeric, mat, fact, 0, 0));
3409: else PetscCall(PetscLogEventEnd(MAT_CholeskyFactor, mat, fact, 0, 0));
3410: PetscCall(MatViewFromOptions(fact, NULL, "-mat_factor_view"));
3411: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3412: PetscFunctionReturn(PETSC_SUCCESS);
3413: }
3415: /*@
3416: MatQRFactor - Performs in-place QR factorization of matrix.
3418: Collective
3420: Input Parameters:
3421: + mat - the matrix
3422: . col - column permutation
3423: - info - options for factorization, includes
3424: .vb
3425: fill - expected fill as ratio of original fill.
3426: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3427: Run with the option -info to determine an optimal value to use
3428: .ve
3430: Level: developer
3432: Notes:
3433: Most users should employ the `KSP` interface for linear solvers
3434: instead of working directly with matrix algebra routines such as this.
3435: See, e.g., `KSPCreate()`.
3437: This changes the state of the matrix to a factored matrix; it cannot be used
3438: for example with `MatSetValues()` unless one first calls `MatSetUnfactored()`.
3440: Developer Note:
3441: The Fortran interface is not autogenerated as the
3442: interface definition cannot be generated correctly [due to MatFactorInfo]
3444: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatQRFactorSymbolic()`, `MatQRFactorNumeric()`, `MatLUFactor()`,
3445: `MatSetUnfactored()`
3446: @*/
3447: PetscErrorCode MatQRFactor(Mat mat, IS col, const MatFactorInfo *info)
3448: {
3449: PetscFunctionBegin;
3452: if (info) PetscAssertPointer(info, 3);
3454: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3455: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3456: MatCheckPreallocated(mat, 1);
3457: PetscCall(PetscLogEventBegin(MAT_QRFactor, mat, col, 0, 0));
3458: PetscUseMethod(mat, "MatQRFactor_C", (Mat, IS, const MatFactorInfo *), (mat, col, info));
3459: PetscCall(PetscLogEventEnd(MAT_QRFactor, mat, col, 0, 0));
3460: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3461: PetscFunctionReturn(PETSC_SUCCESS);
3462: }
3464: /*@
3465: MatQRFactorSymbolic - Performs symbolic QR factorization of matrix.
3466: Call this routine after `MatGetFactor()` but before calling `MatQRFactorNumeric()`.
3468: Collective
3470: Input Parameters:
3471: + fact - the factor matrix obtained with `MatGetFactor()`
3472: . mat - the matrix
3473: . col - column permutation
3474: - info - options for factorization, includes
3475: .vb
3476: fill - expected fill as ratio of original fill.
3477: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3478: Run with the option -info to determine an optimal value to use
3479: .ve
3481: Level: developer
3483: Note:
3484: Most users should employ the `KSP` interface for linear solvers
3485: instead of working directly with matrix algebra routines such as this.
3486: See, e.g., `KSPCreate()`.
3488: Developer Note:
3489: The Fortran interface is not autogenerated as the
3490: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3492: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatFactorInfo`, `MatQRFactor()`, `MatQRFactorNumeric()`, `MatLUFactor()`, `MatFactorInfoInitialize()`
3493: @*/
3494: PetscErrorCode MatQRFactorSymbolic(Mat fact, Mat mat, IS col, const MatFactorInfo *info)
3495: {
3496: MatFactorInfo tinfo;
3498: PetscFunctionBegin;
3502: if (info) PetscAssertPointer(info, 4);
3505: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3506: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3507: MatCheckPreallocated(mat, 2);
3508: if (!info) {
3509: PetscCall(MatFactorInfoInitialize(&tinfo));
3510: info = &tinfo;
3511: }
3513: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_QRFactorSymbolic, fact, mat, col, 0));
3514: PetscUseMethod(fact, "MatQRFactorSymbolic_C", (Mat, Mat, IS, const MatFactorInfo *), (fact, mat, col, info));
3515: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_QRFactorSymbolic, fact, mat, col, 0));
3516: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3517: PetscFunctionReturn(PETSC_SUCCESS);
3518: }
3520: /*@
3521: MatQRFactorNumeric - Performs numeric QR factorization of a matrix.
3522: Call this routine after first calling `MatGetFactor()`, and `MatQRFactorSymbolic()`.
3524: Collective
3526: Input Parameters:
3527: + fact - the factor matrix obtained with `MatGetFactor()`
3528: . mat - the matrix
3529: - info - options for factorization
3531: Level: developer
3533: Notes:
3534: See `MatQRFactor()` for in-place factorization.
3536: Most users should employ the `KSP` interface for linear solvers
3537: instead of working directly with matrix algebra routines such as this.
3538: See, e.g., `KSPCreate()`.
3540: Developer Note:
3541: The Fortran interface is not autogenerated as the
3542: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3544: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatQRFactor()`, `MatQRFactorSymbolic()`, `MatLUFactor()`
3545: @*/
3546: PetscErrorCode MatQRFactorNumeric(Mat fact, Mat mat, const MatFactorInfo *info)
3547: {
3548: MatFactorInfo tinfo;
3550: PetscFunctionBegin;
3555: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3556: 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,
3557: mat->rmap->N, (fact)->rmap->N, mat->cmap->N, (fact)->cmap->N);
3559: MatCheckPreallocated(mat, 2);
3560: if (!info) {
3561: PetscCall(MatFactorInfoInitialize(&tinfo));
3562: info = &tinfo;
3563: }
3565: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_QRFactorNumeric, mat, fact, 0, 0));
3566: else PetscCall(PetscLogEventBegin(MAT_QRFactor, mat, fact, 0, 0));
3567: PetscUseMethod(fact, "MatQRFactorNumeric_C", (Mat, Mat, const MatFactorInfo *), (fact, mat, info));
3568: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_QRFactorNumeric, mat, fact, 0, 0));
3569: else PetscCall(PetscLogEventEnd(MAT_QRFactor, mat, fact, 0, 0));
3570: PetscCall(MatViewFromOptions(fact, NULL, "-mat_factor_view"));
3571: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3572: PetscFunctionReturn(PETSC_SUCCESS);
3573: }
3575: /*@
3576: MatSolve - Solves $A x = b$, given a factored matrix.
3578: Neighbor-wise Collective
3580: Input Parameters:
3581: + mat - the factored matrix
3582: - b - the right-hand-side vector
3584: Output Parameter:
3585: . x - the result vector
3587: Level: developer
3589: Notes:
3590: The vectors `b` and `x` cannot be the same. I.e., one cannot
3591: call `MatSolve`(A,x,x).
3593: Most users should employ the `KSP` interface for linear solvers
3594: instead of working directly with matrix algebra routines such as this.
3595: See, e.g., `KSPCreate()`.
3597: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatLUFactor()`, `MatSolveAdd()`, `MatSolveTranspose()`, `MatSolveTransposeAdd()`
3598: @*/
3599: PetscErrorCode MatSolve(Mat mat, Vec b, Vec x)
3600: {
3601: PetscFunctionBegin;
3606: PetscCheckSameComm(mat, 1, b, 2);
3607: PetscCheckSameComm(mat, 1, x, 3);
3608: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3609: 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);
3610: 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);
3611: 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);
3612: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3613: MatCheckPreallocated(mat, 1);
3615: PetscCall(PetscLogEventBegin(MAT_Solve, mat, b, x, 0));
3616: if (mat->factorerrortype) {
3617: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
3618: PetscCall(VecSetInf(x));
3619: } else PetscUseTypeMethod(mat, solve, b, x);
3620: PetscCall(PetscLogEventEnd(MAT_Solve, mat, b, x, 0));
3621: PetscCall(PetscObjectStateIncrease((PetscObject)x));
3622: PetscFunctionReturn(PETSC_SUCCESS);
3623: }
3625: static PetscErrorCode MatMatSolve_Basic(Mat A, Mat B, Mat X, PetscBool trans)
3626: {
3627: Vec b, x;
3628: PetscInt N, i;
3629: PetscErrorCode (*f)(Mat, Vec, Vec);
3630: PetscBool Abound, Bneedconv = PETSC_FALSE, Xneedconv = PETSC_FALSE;
3632: PetscFunctionBegin;
3633: if (A->factorerrortype) {
3634: PetscCall(PetscInfo(A, "MatFactorError %d\n", A->factorerrortype));
3635: PetscCall(MatSetInf(X));
3636: PetscFunctionReturn(PETSC_SUCCESS);
3637: }
3638: f = (!trans || (!A->ops->solvetranspose && A->symmetric)) ? A->ops->solve : A->ops->solvetranspose;
3639: PetscCheck(f, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Mat type %s", ((PetscObject)A)->type_name);
3640: PetscCall(MatBoundToCPU(A, &Abound));
3641: if (!Abound) {
3642: PetscCall(PetscObjectTypeCompareAny((PetscObject)B, &Bneedconv, MATSEQDENSE, MATMPIDENSE, ""));
3643: PetscCall(PetscObjectTypeCompareAny((PetscObject)X, &Xneedconv, MATSEQDENSE, MATMPIDENSE, ""));
3644: }
3645: #if PetscDefined(HAVE_CUDA)
3646: if (Bneedconv) PetscCall(MatConvert(B, MATDENSECUDA, MAT_INPLACE_MATRIX, &B));
3647: if (Xneedconv) PetscCall(MatConvert(X, MATDENSECUDA, MAT_INPLACE_MATRIX, &X));
3648: #elif PetscDefined(HAVE_HIP)
3649: if (Bneedconv) PetscCall(MatConvert(B, MATDENSEHIP, MAT_INPLACE_MATRIX, &B));
3650: if (Xneedconv) PetscCall(MatConvert(X, MATDENSEHIP, MAT_INPLACE_MATRIX, &X));
3651: #endif
3652: PetscCall(MatGetSize(B, NULL, &N));
3653: for (i = 0; i < N; i++) {
3654: PetscCall(MatDenseGetColumnVecRead(B, i, &b));
3655: PetscCall(MatDenseGetColumnVecWrite(X, i, &x));
3656: PetscCall((*f)(A, b, x));
3657: PetscCall(MatDenseRestoreColumnVecWrite(X, i, &x));
3658: PetscCall(MatDenseRestoreColumnVecRead(B, i, &b));
3659: }
3660: if (Bneedconv) PetscCall(MatConvert(B, MATDENSE, MAT_INPLACE_MATRIX, &B));
3661: if (Xneedconv) PetscCall(MatConvert(X, MATDENSE, MAT_INPLACE_MATRIX, &X));
3662: PetscFunctionReturn(PETSC_SUCCESS);
3663: }
3665: /*@
3666: MatMatSolve - Solves $A X = B$, given a factored matrix.
3668: Neighbor-wise Collective
3670: Input Parameters:
3671: + A - the factored matrix
3672: - B - the right-hand-side matrix `MATDENSE` (or sparse `MATAIJ`-- when using MUMPS)
3674: Output Parameter:
3675: . X - the result matrix (dense matrix)
3677: Level: developer
3679: Note:
3680: If `B` is a `MATDENSE` matrix then one can call `MatMatSolve`(A,B,B) except with `MATSOLVERMKL_CPARDISO`;
3681: otherwise, `B` and `X` cannot be the same.
3683: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSolve()`, `MatMatSolveTranspose()`, `MatLUFactor()`, `MatCholeskyFactor()`
3684: @*/
3685: PetscErrorCode MatMatSolve(Mat A, Mat B, Mat X)
3686: {
3687: PetscFunctionBegin;
3692: PetscCheckSameComm(A, 1, B, 2);
3693: PetscCheckSameComm(A, 1, X, 3);
3694: 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);
3695: 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);
3696: 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");
3697: if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3698: MatCheckPreallocated(A, 1);
3700: PetscCall(PetscLogEventBegin(MAT_MatSolve, A, B, X, 0));
3701: if (!A->ops->matsolve) {
3702: PetscCall(PetscInfo(A, "Mat type %s using basic MatMatSolve\n", ((PetscObject)A)->type_name));
3703: PetscCall(MatMatSolve_Basic(A, B, X, PETSC_FALSE));
3704: } else PetscUseTypeMethod(A, matsolve, B, X);
3705: PetscCall(PetscLogEventEnd(MAT_MatSolve, A, B, X, 0));
3706: PetscCall(PetscObjectStateIncrease((PetscObject)X));
3707: PetscFunctionReturn(PETSC_SUCCESS);
3708: }
3710: /*@
3711: MatMatSolveTranspose - Solves $A^T X = B $, given a factored matrix.
3713: Neighbor-wise Collective
3715: Input Parameters:
3716: + A - the factored matrix
3717: - B - the right-hand-side matrix (`MATDENSE` matrix)
3719: Output Parameter:
3720: . X - the result matrix (dense matrix)
3722: Level: developer
3724: Note:
3725: The matrices `B` and `X` cannot be the same. I.e., one cannot
3726: call `MatMatSolveTranspose`(A,X,X).
3728: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSolveTranspose()`, `MatMatSolve()`, `MatLUFactor()`, `MatCholeskyFactor()`
3729: @*/
3730: PetscErrorCode MatMatSolveTranspose(Mat A, Mat B, Mat X)
3731: {
3732: PetscFunctionBegin;
3737: PetscCheckSameComm(A, 1, B, 2);
3738: PetscCheckSameComm(A, 1, X, 3);
3739: PetscCheck(X != B, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_IDN, "X and B must be different matrices");
3740: 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);
3741: 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);
3742: 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);
3743: 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");
3744: if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3745: MatCheckPreallocated(A, 1);
3747: PetscCall(PetscLogEventBegin(MAT_MatSolve, A, B, X, 0));
3748: if (!A->ops->matsolvetranspose) {
3749: PetscCall(PetscInfo(A, "Mat type %s using basic MatMatSolveTranspose\n", ((PetscObject)A)->type_name));
3750: PetscCall(MatMatSolve_Basic(A, B, X, PETSC_TRUE));
3751: } else PetscUseTypeMethod(A, matsolvetranspose, B, X);
3752: PetscCall(PetscLogEventEnd(MAT_MatSolve, A, B, X, 0));
3753: PetscCall(PetscObjectStateIncrease((PetscObject)X));
3754: PetscFunctionReturn(PETSC_SUCCESS);
3755: }
3757: /*@
3758: MatMatTransposeSolve - Solves $A X = B^T$, given a factored matrix.
3760: Neighbor-wise Collective
3762: Input Parameters:
3763: + A - the factored matrix
3764: - Bt - the transpose of right-hand-side matrix as a `MATDENSE`
3766: Output Parameter:
3767: . X - the result matrix (dense matrix)
3769: Level: developer
3771: Note:
3772: 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
3773: format on the host processor and call `MatMatTransposeSolve()` to implement MUMPS' `MatMatSolve()`.
3775: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatMatSolve()`, `MatMatSolveTranspose()`, `MatLUFactor()`, `MatCholeskyFactor()`
3776: @*/
3777: PetscErrorCode MatMatTransposeSolve(Mat A, Mat Bt, Mat X)
3778: {
3779: PetscFunctionBegin;
3784: PetscCheckSameComm(A, 1, Bt, 2);
3785: PetscCheckSameComm(A, 1, X, 3);
3787: PetscCheck(X != Bt, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_IDN, "X and B must be different matrices");
3788: 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);
3789: 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);
3790: 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");
3791: if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3792: PetscCheck(A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
3793: MatCheckPreallocated(A, 1);
3795: PetscCall(PetscLogEventBegin(MAT_MatTrSolve, A, Bt, X, 0));
3796: PetscUseTypeMethod(A, mattransposesolve, Bt, X);
3797: PetscCall(PetscLogEventEnd(MAT_MatTrSolve, A, Bt, X, 0));
3798: PetscCall(PetscObjectStateIncrease((PetscObject)X));
3799: PetscFunctionReturn(PETSC_SUCCESS);
3800: }
3802: /*@
3803: MatForwardSolve - Solves $ L x = b $, given a factored matrix, $A = LU $, or
3804: $U^T*D^(1/2) x = b$, given a factored symmetric matrix, $A = U^T*D*U$,
3806: Neighbor-wise Collective
3808: Input Parameters:
3809: + mat - the factored matrix
3810: - b - the right-hand-side vector
3812: Output Parameter:
3813: . x - the result vector
3815: Level: developer
3817: Notes:
3818: `MatSolve()` should be used for most applications, as it performs
3819: a forward solve followed by a backward solve.
3821: The vectors `b` and `x` cannot be the same, i.e., one cannot
3822: call `MatForwardSolve`(A,x,x).
3824: For matrix in `MATSEQBAIJ` format with block size larger than 1,
3825: the diagonal blocks are not implemented as $D = D^(1/2) * D^(1/2)$ yet.
3826: `MatForwardSolve()` solves $U^T*D y = b$, and
3827: `MatBackwardSolve()` solves $U x = y$.
3828: Thus they do not provide a symmetric preconditioner.
3830: .seealso: [](ch_matrices), `Mat`, `MatBackwardSolve()`, `MatGetFactor()`, `MatSolve()`
3831: @*/
3832: PetscErrorCode MatForwardSolve(Mat mat, Vec b, Vec x)
3833: {
3834: PetscFunctionBegin;
3839: PetscCheckSameComm(mat, 1, b, 2);
3840: PetscCheckSameComm(mat, 1, x, 3);
3841: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3842: 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);
3843: 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);
3844: 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);
3845: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3846: MatCheckPreallocated(mat, 1);
3848: PetscCall(PetscLogEventBegin(MAT_ForwardSolve, mat, b, x, 0));
3849: PetscUseTypeMethod(mat, forwardsolve, b, x);
3850: PetscCall(PetscLogEventEnd(MAT_ForwardSolve, mat, b, x, 0));
3851: PetscCall(PetscObjectStateIncrease((PetscObject)x));
3852: PetscFunctionReturn(PETSC_SUCCESS);
3853: }
3855: /*@
3856: MatBackwardSolve - Solves $U x = b$, given a factored matrix, $A = LU$.
3857: $D^(1/2) U x = b$, given a factored symmetric matrix, $A = U^T*D*U$,
3859: Neighbor-wise Collective
3861: Input Parameters:
3862: + mat - the factored matrix
3863: - b - the right-hand-side vector
3865: Output Parameter:
3866: . x - the result vector
3868: Level: developer
3870: Notes:
3871: `MatSolve()` should be used for most applications, as it performs
3872: a forward solve followed by a backward solve.
3874: The vectors `b` and `x` cannot be the same. I.e., one cannot
3875: call `MatBackwardSolve`(A,x,x).
3877: For matrix in `MATSEQBAIJ` format with block size larger than 1,
3878: the diagonal blocks are not implemented as $D = D^(1/2) * D^(1/2)$ yet.
3879: `MatForwardSolve()` solves $U^T*D y = b$, and
3880: `MatBackwardSolve()` solves $U x = y$.
3881: Thus they do not provide a symmetric preconditioner.
3883: .seealso: [](ch_matrices), `Mat`, `MatForwardSolve()`, `MatGetFactor()`, `MatSolve()`
3884: @*/
3885: PetscErrorCode MatBackwardSolve(Mat mat, Vec b, Vec x)
3886: {
3887: PetscFunctionBegin;
3892: PetscCheckSameComm(mat, 1, b, 2);
3893: PetscCheckSameComm(mat, 1, x, 3);
3894: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3895: 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);
3896: 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);
3897: 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);
3898: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3899: MatCheckPreallocated(mat, 1);
3901: PetscCall(PetscLogEventBegin(MAT_BackwardSolve, mat, b, x, 0));
3902: PetscUseTypeMethod(mat, backwardsolve, b, x);
3903: PetscCall(PetscLogEventEnd(MAT_BackwardSolve, mat, b, x, 0));
3904: PetscCall(PetscObjectStateIncrease((PetscObject)x));
3905: PetscFunctionReturn(PETSC_SUCCESS);
3906: }
3908: /*@
3909: MatSolveAdd - Computes $x = y + A^{-1}*b$, given a factored matrix.
3911: Neighbor-wise Collective
3913: Input Parameters:
3914: + mat - the factored matrix
3915: . b - the right-hand-side vector
3916: - y - the vector to be added to
3918: Output Parameter:
3919: . x - the result vector
3921: Level: developer
3923: Note:
3924: The vectors `b` and `x` cannot be the same. I.e., one cannot
3925: call `MatSolveAdd`(A,x,y,x).
3927: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatSolve()`, `MatGetFactor()`, `MatSolveTranspose()`, `MatSolveTransposeAdd()`
3928: @*/
3929: PetscErrorCode MatSolveAdd(Mat mat, Vec b, Vec y, Vec x)
3930: {
3931: PetscScalar one = 1.0;
3932: Vec tmp;
3934: PetscFunctionBegin;
3940: PetscCheckSameComm(mat, 1, b, 2);
3941: PetscCheckSameComm(mat, 1, y, 3);
3942: PetscCheckSameComm(mat, 1, x, 4);
3943: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3944: 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);
3945: 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);
3946: 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);
3947: 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);
3948: 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);
3949: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3950: MatCheckPreallocated(mat, 1);
3952: PetscCall(PetscLogEventBegin(MAT_SolveAdd, mat, b, x, y));
3953: if (mat->factorerrortype) {
3954: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
3955: PetscCall(VecSetInf(x));
3956: } else if (mat->ops->solveadd) {
3957: PetscUseTypeMethod(mat, solveadd, b, y, x);
3958: } else {
3959: /* do the solve then the add manually */
3960: if (x != y) {
3961: PetscCall(MatSolve(mat, b, x));
3962: PetscCall(VecAXPY(x, one, y));
3963: } else {
3964: PetscCall(VecDuplicate(x, &tmp));
3965: PetscCall(VecCopy(x, tmp));
3966: PetscCall(MatSolve(mat, b, x));
3967: PetscCall(VecAXPY(x, one, tmp));
3968: PetscCall(VecDestroy(&tmp));
3969: }
3970: }
3971: PetscCall(PetscLogEventEnd(MAT_SolveAdd, mat, b, x, y));
3972: PetscCall(PetscObjectStateIncrease((PetscObject)x));
3973: PetscFunctionReturn(PETSC_SUCCESS);
3974: }
3976: /*@
3977: MatSolveTranspose - Solves $A^T x = b$, given a factored matrix.
3979: Neighbor-wise Collective
3981: Input Parameters:
3982: + mat - the factored matrix
3983: - b - the right-hand-side vector
3985: Output Parameter:
3986: . x - the result vector
3988: Level: developer
3990: Notes:
3991: The vectors `b` and `x` cannot be the same. I.e., one cannot
3992: call `MatSolveTranspose`(A,x,x).
3994: Most users should employ the `KSP` interface for linear solvers
3995: instead of working directly with matrix algebra routines such as this.
3996: See, e.g., `KSPCreate()`.
3998: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `KSP`, `MatSolve()`, `MatSolveAdd()`, `MatSolveTransposeAdd()`
3999: @*/
4000: PetscErrorCode MatSolveTranspose(Mat mat, Vec b, Vec x)
4001: {
4002: PetscErrorCode (*f)(Mat, Vec, Vec) = (!mat->ops->solvetranspose && mat->symmetric) ? mat->ops->solve : mat->ops->solvetranspose;
4004: PetscFunctionBegin;
4009: PetscCheckSameComm(mat, 1, b, 2);
4010: PetscCheckSameComm(mat, 1, x, 3);
4011: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
4012: 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);
4013: 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);
4014: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
4015: MatCheckPreallocated(mat, 1);
4016: PetscCall(PetscLogEventBegin(MAT_SolveTranspose, mat, b, x, 0));
4017: if (mat->factorerrortype) {
4018: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
4019: PetscCall(VecSetInf(x));
4020: } else {
4021: PetscCheck(f, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Matrix type %s", ((PetscObject)mat)->type_name);
4022: PetscCall((*f)(mat, b, x));
4023: }
4024: PetscCall(PetscLogEventEnd(MAT_SolveTranspose, mat, b, x, 0));
4025: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4026: PetscFunctionReturn(PETSC_SUCCESS);
4027: }
4029: /*@
4030: MatSolveTransposeAdd - Computes $x = y + A^{-T} b$
4031: factored matrix.
4033: Neighbor-wise Collective
4035: Input Parameters:
4036: + mat - the factored matrix
4037: . b - the right-hand-side vector
4038: - y - the vector to be added to
4040: Output Parameter:
4041: . x - the result vector
4043: Level: developer
4045: Note:
4046: The vectors `b` and `x` cannot be the same. I.e., one cannot
4047: call `MatSolveTransposeAdd`(A,x,y,x).
4049: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatSolve()`, `MatSolveAdd()`, `MatSolveTranspose()`
4050: @*/
4051: PetscErrorCode MatSolveTransposeAdd(Mat mat, Vec b, Vec y, Vec x)
4052: {
4053: PetscScalar one = 1.0;
4054: Vec tmp;
4055: PetscErrorCode (*f)(Mat, Vec, Vec, Vec) = (!mat->ops->solvetransposeadd && mat->symmetric) ? mat->ops->solveadd : mat->ops->solvetransposeadd;
4057: PetscFunctionBegin;
4063: PetscCheckSameComm(mat, 1, b, 2);
4064: PetscCheckSameComm(mat, 1, y, 3);
4065: PetscCheckSameComm(mat, 1, x, 4);
4066: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
4067: 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);
4068: 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);
4069: 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);
4070: 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);
4071: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
4072: MatCheckPreallocated(mat, 1);
4074: PetscCall(PetscLogEventBegin(MAT_SolveTransposeAdd, mat, b, x, y));
4075: if (mat->factorerrortype) {
4076: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
4077: PetscCall(VecSetInf(x));
4078: } else if (f) {
4079: PetscCall((*f)(mat, b, y, x));
4080: } else {
4081: /* do the solve then the add manually */
4082: if (x != y) {
4083: PetscCall(MatSolveTranspose(mat, b, x));
4084: PetscCall(VecAXPY(x, one, y));
4085: } else {
4086: PetscCall(VecDuplicate(x, &tmp));
4087: PetscCall(VecCopy(x, tmp));
4088: PetscCall(MatSolveTranspose(mat, b, x));
4089: PetscCall(VecAXPY(x, one, tmp));
4090: PetscCall(VecDestroy(&tmp));
4091: }
4092: }
4093: PetscCall(PetscLogEventEnd(MAT_SolveTransposeAdd, mat, b, x, y));
4094: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4095: PetscFunctionReturn(PETSC_SUCCESS);
4096: }
4098: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
4099: /*@
4100: MatSOR - Computes relaxation (SOR, Gauss-Seidel) sweeps.
4102: Neighbor-wise Collective
4104: Input Parameters:
4105: + mat - the matrix
4106: . b - the right hand side
4107: . omega - the relaxation factor
4108: . flag - flag indicating the type of SOR (see below)
4109: . shift - diagonal shift
4110: . its - the number of iterations
4111: - lits - the number of local iterations
4113: Output Parameter:
4114: . x - the solution (can contain an initial guess, use option `SOR_ZERO_INITIAL_GUESS` to indicate no guess)
4116: SOR Flags:
4117: + `SOR_FORWARD_SWEEP` - forward SOR
4118: . `SOR_BACKWARD_SWEEP` - backward SOR
4119: . `SOR_SYMMETRIC_SWEEP` - SSOR (symmetric SOR)
4120: . `SOR_LOCAL_FORWARD_SWEEP` - local forward SOR
4121: . `SOR_LOCAL_BACKWARD_SWEEP` - local forward SOR
4122: . `SOR_LOCAL_SYMMETRIC_SWEEP` - local SSOR
4123: . `SOR_EISENSTAT` - SOR with Eisenstat trick
4124: . `SOR_APPLY_UPPER`, `SOR_APPLY_LOWER` - applies
4125: upper/lower triangular part of matrix to
4126: vector (with omega)
4127: - `SOR_ZERO_INITIAL_GUESS` - zero initial guess
4129: Level: developer
4131: Notes:
4132: `SOR_LOCAL_FORWARD_SWEEP`, `SOR_LOCAL_BACKWARD_SWEEP`, and
4133: `SOR_LOCAL_SYMMETRIC_SWEEP` perform separate independent smoothings
4134: on each processor.
4136: Application programmers will not generally use `MatSOR()` directly,
4137: but instead will employ the `KSP`/`PC` interface.
4139: For `MATBAIJ`, `MATSBAIJ`, and `MATAIJ` matrices with Inodes this does a block SOR smoothing, otherwise it does a pointwise smoothing
4141: Most users should employ the `KSP` interface for linear solvers
4142: instead of working directly with matrix algebra routines such as this.
4143: See, e.g., `KSPCreate()`.
4145: Vectors `x` and `b` CANNOT be the same
4147: The flags are implemented as bitwise inclusive or operations.
4148: For example, use (`SOR_ZERO_INITIAL_GUESS` | `SOR_SYMMETRIC_SWEEP`)
4149: to specify a zero initial guess for SSOR.
4151: Developer Note:
4152: We should add block SOR support for `MATAIJ` matrices with block size set to great than one and no inodes
4154: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `KSP`, `PC`, `MatGetFactor()`
4155: @*/
4156: PetscErrorCode MatSOR(Mat mat, Vec b, PetscReal omega, MatSORType flag, PetscReal shift, PetscInt its, PetscInt lits, Vec x)
4157: {
4158: PetscFunctionBegin;
4163: PetscCheckSameComm(mat, 1, b, 2);
4164: PetscCheckSameComm(mat, 1, x, 8);
4165: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4166: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4167: 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);
4168: 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);
4169: 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);
4170: PetscCheck(its > 0, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Relaxation requires global its %" PetscInt_FMT " positive", its);
4171: PetscCheck(lits > 0, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Relaxation requires local its %" PetscInt_FMT " positive", lits);
4172: PetscCheck(b != x, PETSC_COMM_SELF, PETSC_ERR_ARG_IDN, "b and x vector cannot be the same");
4174: MatCheckPreallocated(mat, 1);
4175: PetscCall(PetscLogEventBegin(MAT_SOR, mat, b, x, 0));
4176: PetscUseTypeMethod(mat, sor, b, omega, flag, shift, its, lits, x);
4177: PetscCall(PetscLogEventEnd(MAT_SOR, mat, b, x, 0));
4178: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4179: PetscFunctionReturn(PETSC_SUCCESS);
4180: }
4182: /*
4183: Default matrix copy routine.
4184: */
4185: PetscErrorCode MatCopy_Basic(Mat A, Mat B, MatStructure str)
4186: {
4187: PetscInt i, rstart = 0, rend = 0, nz;
4188: const PetscInt *cwork;
4189: const PetscScalar *vwork;
4191: PetscFunctionBegin;
4192: if (B->assembled) PetscCall(MatZeroEntries(B));
4193: if (str == SAME_NONZERO_PATTERN) {
4194: PetscCall(MatGetOwnershipRange(A, &rstart, &rend));
4195: for (i = rstart; i < rend; i++) {
4196: PetscCall(MatGetRow(A, i, &nz, &cwork, &vwork));
4197: PetscCall(MatSetValues(B, 1, &i, nz, cwork, vwork, INSERT_VALUES));
4198: PetscCall(MatRestoreRow(A, i, &nz, &cwork, &vwork));
4199: }
4200: } else {
4201: PetscCall(MatAYPX(B, 0.0, A, str));
4202: }
4203: PetscCall(MatAssemblyBegin(B, MAT_FINAL_ASSEMBLY));
4204: PetscCall(MatAssemblyEnd(B, MAT_FINAL_ASSEMBLY));
4205: PetscFunctionReturn(PETSC_SUCCESS);
4206: }
4208: /*@
4209: MatCopy - Copies a matrix to another matrix.
4211: Collective
4213: Input Parameters:
4214: + A - the matrix
4215: - str - `SAME_NONZERO_PATTERN` or `DIFFERENT_NONZERO_PATTERN`
4217: Output Parameter:
4218: . B - where the copy is put
4220: Level: intermediate
4222: Notes:
4223: If you use `SAME_NONZERO_PATTERN` then the two matrices must have the same nonzero pattern or the routine will crash.
4225: `MatCopy()` copies the matrix entries of a matrix to another existing
4226: matrix (after first zeroing the second matrix). A related routine is
4227: `MatConvert()`, which first creates a new matrix and then copies the data.
4229: .seealso: [](ch_matrices), `Mat`, `MatConvert()`, `MatDuplicate()`
4230: @*/
4231: PetscErrorCode MatCopy(Mat A, Mat B, MatStructure str)
4232: {
4233: PetscInt i;
4235: PetscFunctionBegin;
4240: PetscCheckSameComm(A, 1, B, 2);
4241: MatCheckPreallocated(B, 2);
4242: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4243: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4244: 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,
4245: A->cmap->N, B->cmap->N);
4246: MatCheckPreallocated(A, 1);
4247: if (A == B) PetscFunctionReturn(PETSC_SUCCESS);
4249: PetscCall(PetscLogEventBegin(MAT_Copy, A, B, 0, 0));
4250: if (A->ops->copy) PetscUseTypeMethod(A, copy, B, str);
4251: else PetscCall(MatCopy_Basic(A, B, str));
4253: B->stencil.dim = A->stencil.dim;
4254: B->stencil.noc = A->stencil.noc;
4255: for (i = 0; i <= A->stencil.dim + (A->stencil.noc ? 0 : -1); i++) {
4256: B->stencil.dims[i] = A->stencil.dims[i];
4257: B->stencil.starts[i] = A->stencil.starts[i];
4258: }
4260: PetscCall(PetscLogEventEnd(MAT_Copy, A, B, 0, 0));
4261: PetscCall(PetscObjectStateIncrease((PetscObject)B));
4262: PetscFunctionReturn(PETSC_SUCCESS);
4263: }
4265: /*@C
4266: MatConvert - Converts a matrix to another matrix, either of the same
4267: or different type.
4269: Collective
4271: Input Parameters:
4272: + mat - the matrix
4273: . newtype - new matrix type. Use `MATSAME` to create a new matrix of the
4274: same type as the original matrix.
4275: - reuse - denotes if the destination matrix is to be created or reused.
4276: 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
4277: `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).
4279: Output Parameter:
4280: . M - pointer to place new matrix
4282: Level: intermediate
4284: Notes:
4285: `MatConvert()` first creates a new matrix and then copies the data from
4286: the first matrix. A related routine is `MatCopy()`, which copies the matrix
4287: entries of one matrix to another already existing matrix context.
4289: Cannot be used to convert a sequential matrix to parallel or parallel to sequential,
4290: the MPI communicator of the generated matrix is always the same as the communicator
4291: of the input matrix.
4293: .seealso: [](ch_matrices), `Mat`, `MatCopy()`, `MatDuplicate()`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`
4294: @*/
4295: PetscErrorCode MatConvert(Mat mat, MatType newtype, MatReuse reuse, Mat *M)
4296: {
4297: PetscBool sametype, issame, flg;
4298: PetscBool3 issymmetric, ishermitian;
4299: char convname[256], mtype[256];
4300: Mat B;
4302: PetscFunctionBegin;
4305: PetscAssertPointer(M, 4);
4306: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4307: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4308: MatCheckPreallocated(mat, 1);
4310: PetscCall(PetscOptionsGetString(((PetscObject)mat)->options, ((PetscObject)mat)->prefix, "-matconvert_type", mtype, sizeof(mtype), &flg));
4311: if (flg) newtype = mtype;
4313: PetscCall(PetscObjectTypeCompare((PetscObject)mat, newtype, &sametype));
4314: PetscCall(PetscStrcmp(newtype, "same", &issame));
4315: PetscCheck(!(reuse == MAT_INPLACE_MATRIX) || !(mat != *M), PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "MAT_INPLACE_MATRIX requires same input and output matrix");
4316: if (reuse == MAT_REUSE_MATRIX) {
4318: PetscCheck(mat != *M, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "MAT_REUSE_MATRIX means reuse matrix in final argument, perhaps you mean MAT_INPLACE_MATRIX");
4319: }
4321: if ((reuse == MAT_INPLACE_MATRIX) && (issame || sametype)) {
4322: PetscCall(PetscInfo(mat, "Early return for inplace %s %d %d\n", ((PetscObject)mat)->type_name, sametype, issame));
4323: PetscFunctionReturn(PETSC_SUCCESS);
4324: }
4326: /* Cache Mat options because some converters use MatHeaderReplace */
4327: issymmetric = mat->symmetric;
4328: ishermitian = mat->hermitian;
4330: if ((sametype || issame) && (reuse == MAT_INITIAL_MATRIX) && mat->ops->duplicate) {
4331: PetscCall(PetscInfo(mat, "Calling duplicate for initial matrix %s %d %d\n", ((PetscObject)mat)->type_name, sametype, issame));
4332: PetscUseTypeMethod(mat, duplicate, MAT_COPY_VALUES, M);
4333: } else {
4334: PetscErrorCode (*conv)(Mat, MatType, MatReuse, Mat *) = NULL;
4335: const char *prefix[3] = {"seq", "mpi", ""};
4336: PetscInt i;
4337: /*
4338: Order of precedence:
4339: 0) See if newtype is a superclass of the current matrix.
4340: 1) See if a specialized converter is known to the current matrix.
4341: 2) See if a specialized converter is known to the desired matrix class.
4342: 3) See if a good general converter is registered for the desired class
4343: (as of 6/27/03 only MATMPIADJ falls into this category).
4344: 4) See if a good general converter is known for the current matrix.
4345: 5) Use a really basic converter.
4346: */
4348: /* 0) See if newtype is a superclass of the current matrix.
4349: i.e mat is mpiaij and newtype is aij */
4350: for (i = 0; i < 2; i++) {
4351: PetscCall(PetscStrncpy(convname, prefix[i], sizeof(convname)));
4352: PetscCall(PetscStrlcat(convname, newtype, sizeof(convname)));
4353: PetscCall(PetscStrcmp(convname, ((PetscObject)mat)->type_name, &flg));
4354: PetscCall(PetscInfo(mat, "Check superclass %s %s -> %d\n", convname, ((PetscObject)mat)->type_name, flg));
4355: if (flg) {
4356: if (reuse == MAT_INPLACE_MATRIX) {
4357: PetscCall(PetscInfo(mat, "Early return\n"));
4358: PetscFunctionReturn(PETSC_SUCCESS);
4359: } else if (reuse == MAT_INITIAL_MATRIX && mat->ops->duplicate) {
4360: PetscCall(PetscInfo(mat, "Calling MatDuplicate\n"));
4361: PetscUseTypeMethod(mat, duplicate, MAT_COPY_VALUES, M);
4362: PetscFunctionReturn(PETSC_SUCCESS);
4363: } else if (reuse == MAT_REUSE_MATRIX && mat->ops->copy) {
4364: PetscCall(PetscInfo(mat, "Calling MatCopy\n"));
4365: PetscCall(MatCopy(mat, *M, SAME_NONZERO_PATTERN));
4366: PetscFunctionReturn(PETSC_SUCCESS);
4367: }
4368: }
4369: }
4370: /* 1) See if a specialized converter is known to the current matrix and the desired class */
4371: for (i = 0; i < 3; i++) {
4372: PetscCall(PetscStrncpy(convname, "MatConvert_", sizeof(convname)));
4373: PetscCall(PetscStrlcat(convname, ((PetscObject)mat)->type_name, sizeof(convname)));
4374: PetscCall(PetscStrlcat(convname, "_", sizeof(convname)));
4375: PetscCall(PetscStrlcat(convname, prefix[i], sizeof(convname)));
4376: PetscCall(PetscStrlcat(convname, issame ? ((PetscObject)mat)->type_name : newtype, sizeof(convname)));
4377: PetscCall(PetscStrlcat(convname, "_C", sizeof(convname)));
4378: PetscCall(PetscObjectQueryFunction((PetscObject)mat, convname, &conv));
4379: PetscCall(PetscInfo(mat, "Check specialized (1) %s (%s) -> %d\n", convname, ((PetscObject)mat)->type_name, !!conv));
4380: if (conv) goto foundconv;
4381: }
4383: /* 2) See if a specialized converter is known to the desired matrix class. */
4384: PetscCall(MatCreate(PetscObjectComm((PetscObject)mat), &B));
4385: PetscCall(MatSetSizes(B, mat->rmap->n, mat->cmap->n, mat->rmap->N, mat->cmap->N));
4386: PetscCall(MatSetType(B, newtype));
4387: for (i = 0; i < 3; i++) {
4388: PetscCall(PetscStrncpy(convname, "MatConvert_", sizeof(convname)));
4389: PetscCall(PetscStrlcat(convname, ((PetscObject)mat)->type_name, sizeof(convname)));
4390: PetscCall(PetscStrlcat(convname, "_", sizeof(convname)));
4391: PetscCall(PetscStrlcat(convname, prefix[i], sizeof(convname)));
4392: PetscCall(PetscStrlcat(convname, newtype, sizeof(convname)));
4393: PetscCall(PetscStrlcat(convname, "_C", sizeof(convname)));
4394: PetscCall(PetscObjectQueryFunction((PetscObject)B, convname, &conv));
4395: PetscCall(PetscInfo(mat, "Check specialized (2) %s (%s) -> %d\n", convname, ((PetscObject)B)->type_name, !!conv));
4396: if (conv) {
4397: PetscCall(MatDestroy(&B));
4398: goto foundconv;
4399: }
4400: }
4402: /* 3) See if a good general converter is registered for the desired class */
4403: conv = B->ops->convertfrom;
4404: PetscCall(PetscInfo(mat, "Check convertfrom (%s) -> %d\n", ((PetscObject)B)->type_name, !!conv));
4405: PetscCall(MatDestroy(&B));
4406: if (conv) goto foundconv;
4408: /* 4) See if a good general converter is known for the current matrix */
4409: if (mat->ops->convert) conv = mat->ops->convert;
4410: PetscCall(PetscInfo(mat, "Check general convert (%s) -> %d\n", ((PetscObject)mat)->type_name, !!conv));
4411: if (conv) goto foundconv;
4413: /* 5) Use a really basic converter. */
4414: PetscCall(PetscInfo(mat, "Using MatConvert_Basic\n"));
4415: conv = MatConvert_Basic;
4417: foundconv:
4418: PetscCall(PetscLogEventBegin(MAT_Convert, mat, 0, 0, 0));
4419: PetscCall((*conv)(mat, newtype, reuse, M));
4420: if (mat->rmap->mapping && mat->cmap->mapping && !(*M)->rmap->mapping && !(*M)->cmap->mapping) {
4421: /* the block sizes must be same if the mappings are copied over */
4422: (*M)->rmap->bs = mat->rmap->bs;
4423: (*M)->cmap->bs = mat->cmap->bs;
4424: PetscCall(PetscObjectReference((PetscObject)mat->rmap->mapping));
4425: PetscCall(PetscObjectReference((PetscObject)mat->cmap->mapping));
4426: (*M)->rmap->mapping = mat->rmap->mapping;
4427: (*M)->cmap->mapping = mat->cmap->mapping;
4428: }
4429: (*M)->stencil.dim = mat->stencil.dim;
4430: (*M)->stencil.noc = mat->stencil.noc;
4431: for (i = 0; i <= mat->stencil.dim + (mat->stencil.noc ? 0 : -1); i++) {
4432: (*M)->stencil.dims[i] = mat->stencil.dims[i];
4433: (*M)->stencil.starts[i] = mat->stencil.starts[i];
4434: }
4435: PetscCall(PetscLogEventEnd(MAT_Convert, mat, 0, 0, 0));
4436: }
4437: PetscCall(PetscObjectStateIncrease((PetscObject)*M));
4439: /* Copy Mat options */
4440: if (issymmetric == PETSC_BOOL3_TRUE) PetscCall(MatSetOption(*M, MAT_SYMMETRIC, PETSC_TRUE));
4441: else if (issymmetric == PETSC_BOOL3_FALSE) PetscCall(MatSetOption(*M, MAT_SYMMETRIC, PETSC_FALSE));
4442: if (ishermitian == PETSC_BOOL3_TRUE) PetscCall(MatSetOption(*M, MAT_HERMITIAN, PETSC_TRUE));
4443: else if (ishermitian == PETSC_BOOL3_FALSE) PetscCall(MatSetOption(*M, MAT_HERMITIAN, PETSC_FALSE));
4444: PetscFunctionReturn(PETSC_SUCCESS);
4445: }
4447: /*@C
4448: MatFactorGetSolverType - Returns name of the package providing the factorization routines
4450: Not Collective
4452: Input Parameter:
4453: . mat - the matrix, must be a factored matrix
4455: Output Parameter:
4456: . type - the string name of the package (do not free this string)
4458: Level: intermediate
4460: Fortran Note:
4461: Pass in an empty string and the package name will be copied into it. Make sure the string is long enough.
4463: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSolverType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`
4464: @*/
4465: PetscErrorCode MatFactorGetSolverType(Mat mat, MatSolverType *type)
4466: {
4467: PetscErrorCode (*conv)(Mat, MatSolverType *);
4469: PetscFunctionBegin;
4472: PetscAssertPointer(type, 2);
4473: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Only for factored matrix");
4474: PetscCall(PetscObjectQueryFunction((PetscObject)mat, "MatFactorGetSolverType_C", &conv));
4475: if (conv) PetscCall((*conv)(mat, type));
4476: else *type = MATSOLVERPETSC;
4477: PetscFunctionReturn(PETSC_SUCCESS);
4478: }
4480: typedef struct _MatSolverTypeForSpecifcType *MatSolverTypeForSpecifcType;
4481: struct _MatSolverTypeForSpecifcType {
4482: MatType mtype;
4483: /* no entry for MAT_FACTOR_NONE */
4484: PetscErrorCode (*createfactor[MAT_FACTOR_NUM_TYPES - 1])(Mat, MatFactorType, Mat *);
4485: MatSolverTypeForSpecifcType next;
4486: };
4488: typedef struct _MatSolverTypeHolder *MatSolverTypeHolder;
4489: struct _MatSolverTypeHolder {
4490: char *name;
4491: MatSolverTypeForSpecifcType handlers;
4492: MatSolverTypeHolder next;
4493: };
4495: static MatSolverTypeHolder MatSolverTypeHolders = NULL;
4497: /*@C
4498: MatSolverTypeRegister - Registers a `MatSolverType` that works for a particular matrix type
4500: Input Parameters:
4501: + package - name of the package, for example petsc or superlu
4502: . mtype - the matrix type that works with this package
4503: . ftype - the type of factorization supported by the package
4504: - createfactor - routine that will create the factored matrix ready to be used
4506: Level: developer
4508: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorGetSolverType()`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`,
4509: `MatGetFactor()`
4510: @*/
4511: PetscErrorCode MatSolverTypeRegister(MatSolverType package, MatType mtype, MatFactorType ftype, PetscErrorCode (*createfactor)(Mat, MatFactorType, Mat *))
4512: {
4513: MatSolverTypeHolder next = MatSolverTypeHolders, prev = NULL;
4514: PetscBool flg;
4515: MatSolverTypeForSpecifcType inext, iprev = NULL;
4517: PetscFunctionBegin;
4518: PetscCall(MatInitializePackage());
4519: if (!next) {
4520: PetscCall(PetscNew(&MatSolverTypeHolders));
4521: PetscCall(PetscStrallocpy(package, &MatSolverTypeHolders->name));
4522: PetscCall(PetscNew(&MatSolverTypeHolders->handlers));
4523: PetscCall(PetscStrallocpy(mtype, (char **)&MatSolverTypeHolders->handlers->mtype));
4524: MatSolverTypeHolders->handlers->createfactor[(int)ftype - 1] = createfactor;
4525: PetscFunctionReturn(PETSC_SUCCESS);
4526: }
4527: while (next) {
4528: PetscCall(PetscStrcasecmp(package, next->name, &flg));
4529: if (flg) {
4530: PetscCheck(next->handlers, PETSC_COMM_SELF, PETSC_ERR_PLIB, "MatSolverTypeHolder is missing handlers");
4531: inext = next->handlers;
4532: while (inext) {
4533: PetscCall(PetscStrcasecmp(mtype, inext->mtype, &flg));
4534: if (flg) {
4535: inext->createfactor[(int)ftype - 1] = createfactor;
4536: PetscFunctionReturn(PETSC_SUCCESS);
4537: }
4538: iprev = inext;
4539: inext = inext->next;
4540: }
4541: PetscCall(PetscNew(&iprev->next));
4542: PetscCall(PetscStrallocpy(mtype, (char **)&iprev->next->mtype));
4543: iprev->next->createfactor[(int)ftype - 1] = createfactor;
4544: PetscFunctionReturn(PETSC_SUCCESS);
4545: }
4546: prev = next;
4547: next = next->next;
4548: }
4549: PetscCall(PetscNew(&prev->next));
4550: PetscCall(PetscStrallocpy(package, &prev->next->name));
4551: PetscCall(PetscNew(&prev->next->handlers));
4552: PetscCall(PetscStrallocpy(mtype, (char **)&prev->next->handlers->mtype));
4553: prev->next->handlers->createfactor[(int)ftype - 1] = createfactor;
4554: PetscFunctionReturn(PETSC_SUCCESS);
4555: }
4557: /*@C
4558: MatSolverTypeGet - Gets the function that creates the factor matrix if it exist
4560: Input Parameters:
4561: + 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
4562: . ftype - the type of factorization supported by the type
4563: - mtype - the matrix type that works with this type
4565: Output Parameters:
4566: + foundtype - `PETSC_TRUE` if the type was registered
4567: . foundmtype - `PETSC_TRUE` if the type supports the requested mtype
4568: - createfactor - routine that will create the factored matrix ready to be used or `NULL` if not found
4570: Calling sequence of `createfactor`:
4571: + A - the matrix providing the factor matrix
4572: . mtype - the `MatType` of the factor requested
4573: - B - the new factor matrix that responds to MatXXFactorSymbolic,Numeric() functions, such as `MatLUFactorSymbolic()`
4575: Level: developer
4577: Note:
4578: When `type` is `NULL` the available functions are searched for based on the order of the calls to `MatSolverTypeRegister()` in `MatInitializePackage()`.
4579: Since different PETSc configurations may have different external solvers, seemingly identical runs with different PETSc configurations may use a different solver.
4580: For example if one configuration had --download-mumps while a different one had --download-superlu_dist.
4582: .seealso: [](ch_matrices), `Mat`, `MatFactorType`, `MatType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatSolverTypeRegister()`, `MatGetFactor()`,
4583: `MatInitializePackage()`
4584: @*/
4585: PetscErrorCode MatSolverTypeGet(MatSolverType type, MatType mtype, MatFactorType ftype, PetscBool *foundtype, PetscBool *foundmtype, PetscErrorCode (**createfactor)(Mat A, MatFactorType mtype, Mat *B))
4586: {
4587: MatSolverTypeHolder next = MatSolverTypeHolders;
4588: PetscBool flg;
4589: MatSolverTypeForSpecifcType inext;
4591: PetscFunctionBegin;
4592: if (foundtype) *foundtype = PETSC_FALSE;
4593: if (foundmtype) *foundmtype = PETSC_FALSE;
4594: if (createfactor) *createfactor = NULL;
4596: if (type) {
4597: while (next) {
4598: PetscCall(PetscStrcasecmp(type, next->name, &flg));
4599: if (flg) {
4600: if (foundtype) *foundtype = PETSC_TRUE;
4601: inext = next->handlers;
4602: while (inext) {
4603: PetscCall(PetscStrbeginswith(mtype, inext->mtype, &flg));
4604: if (flg) {
4605: if (foundmtype) *foundmtype = PETSC_TRUE;
4606: if (createfactor) *createfactor = inext->createfactor[(int)ftype - 1];
4607: PetscFunctionReturn(PETSC_SUCCESS);
4608: }
4609: inext = inext->next;
4610: }
4611: }
4612: next = next->next;
4613: }
4614: } else {
4615: while (next) {
4616: inext = next->handlers;
4617: while (inext) {
4618: PetscCall(PetscStrcmp(mtype, inext->mtype, &flg));
4619: if (flg && inext->createfactor[(int)ftype - 1]) {
4620: if (foundtype) *foundtype = PETSC_TRUE;
4621: if (foundmtype) *foundmtype = PETSC_TRUE;
4622: if (createfactor) *createfactor = inext->createfactor[(int)ftype - 1];
4623: PetscFunctionReturn(PETSC_SUCCESS);
4624: }
4625: inext = inext->next;
4626: }
4627: next = next->next;
4628: }
4629: /* try with base classes inext->mtype */
4630: next = MatSolverTypeHolders;
4631: while (next) {
4632: inext = next->handlers;
4633: while (inext) {
4634: PetscCall(PetscStrbeginswith(mtype, inext->mtype, &flg));
4635: if (flg && inext->createfactor[(int)ftype - 1]) {
4636: if (foundtype) *foundtype = PETSC_TRUE;
4637: if (foundmtype) *foundmtype = PETSC_TRUE;
4638: if (createfactor) *createfactor = inext->createfactor[(int)ftype - 1];
4639: PetscFunctionReturn(PETSC_SUCCESS);
4640: }
4641: inext = inext->next;
4642: }
4643: next = next->next;
4644: }
4645: }
4646: PetscFunctionReturn(PETSC_SUCCESS);
4647: }
4649: PetscErrorCode MatSolverTypeDestroy(void)
4650: {
4651: MatSolverTypeHolder next = MatSolverTypeHolders, prev;
4652: MatSolverTypeForSpecifcType inext, iprev;
4654: PetscFunctionBegin;
4655: while (next) {
4656: PetscCall(PetscFree(next->name));
4657: inext = next->handlers;
4658: while (inext) {
4659: PetscCall(PetscFree(inext->mtype));
4660: iprev = inext;
4661: inext = inext->next;
4662: PetscCall(PetscFree(iprev));
4663: }
4664: prev = next;
4665: next = next->next;
4666: PetscCall(PetscFree(prev));
4667: }
4668: MatSolverTypeHolders = NULL;
4669: PetscFunctionReturn(PETSC_SUCCESS);
4670: }
4672: /*@C
4673: MatFactorGetCanUseOrdering - Indicates if the factorization can use the ordering provided in `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`
4675: Logically Collective
4677: Input Parameter:
4678: . mat - the matrix
4680: Output Parameter:
4681: . flg - `PETSC_TRUE` if uses the ordering
4683: Level: developer
4685: Note:
4686: Most internal PETSc factorizations use the ordering passed to the factorization routine but external
4687: packages do not, thus we want to skip generating the ordering when it is not needed or used.
4689: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`
4690: @*/
4691: PetscErrorCode MatFactorGetCanUseOrdering(Mat mat, PetscBool *flg)
4692: {
4693: PetscFunctionBegin;
4694: *flg = mat->canuseordering;
4695: PetscFunctionReturn(PETSC_SUCCESS);
4696: }
4698: /*@C
4699: MatFactorGetPreferredOrdering - The preferred ordering for a particular matrix factor object
4701: Logically Collective
4703: Input Parameters:
4704: + mat - the matrix obtained with `MatGetFactor()`
4705: - ftype - the factorization type to be used
4707: Output Parameter:
4708: . otype - the preferred ordering type
4710: Level: developer
4712: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorType`, `MatOrderingType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`
4713: @*/
4714: PetscErrorCode MatFactorGetPreferredOrdering(Mat mat, MatFactorType ftype, MatOrderingType *otype)
4715: {
4716: PetscFunctionBegin;
4717: *otype = mat->preferredordering[ftype];
4718: PetscCheck(*otype, PETSC_COMM_SELF, PETSC_ERR_PLIB, "MatFactor did not have a preferred ordering");
4719: PetscFunctionReturn(PETSC_SUCCESS);
4720: }
4722: /*@C
4723: MatGetFactor - Returns a matrix suitable to calls to MatXXFactorSymbolic,Numeric()
4725: Collective
4727: Input Parameters:
4728: + mat - the matrix
4729: . type - name of solver type, for example, superlu, petsc (to use PETSc's solver if it is available), if this is 'NULL' then the first result that satisfies
4730: the other criteria is returned
4731: - ftype - factor type, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`
4733: Output Parameter:
4734: . f - the factor matrix used with MatXXFactorSymbolic,Numeric() calls. Can be `NULL` in some cases, see notes below.
4736: Options Database Keys:
4737: + -pc_factor_mat_solver_type <type> - choose the type at run time. When using `KSP` solvers
4738: - -mat_factor_bind_factorization <host, device> - Where to do matrix factorization? Default is device (might consume more device memory.
4739: One can choose host to save device memory). Currently only supported with `MATSEQAIJCUSPARSE` matrices.
4741: Level: intermediate
4743: Notes:
4744: The return matrix can be `NULL` if the requested factorization is not available, since some combinations of matrix types and factorization
4745: types registered with `MatSolverTypeRegister()` cannot be fully tested if not at runtime.
4747: Users usually access the factorization solvers via `KSP`
4749: Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
4750: such as pastix, superlu, mumps etc. PETSc must have been ./configure to use the external solver, using the option --download-package or --with-package-dir
4752: When `type` is `NULL` the available results are searched for based on the order of the calls to `MatSolverTypeRegister()` in `MatInitializePackage()`.
4753: Since different PETSc configurations may have different external solvers, seemingly identical runs with different PETSc configurations may use a different solver.
4754: For example if one configuration had --download-mumps while a different one had --download-superlu_dist.
4756: Some of the packages have options for controlling the factorization, these are in the form -prefix_mat_packagename_packageoption
4757: where prefix is normally obtained from the calling `KSP`/`PC`. If `MatGetFactor()` is called directly one can set
4758: call `MatSetOptionsPrefixFactor()` on the originating matrix or `MatSetOptionsPrefix()` on the resulting factor matrix.
4760: Developer Note:
4761: This should actually be called `MatCreateFactor()` since it creates a new factor object
4763: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `KSP`, `MatSolverType`, `MatFactorType`, `MatCopy()`, `MatDuplicate()`,
4764: `MatGetFactorAvailable()`, `MatFactorGetCanUseOrdering()`, `MatSolverTypeRegister()`, `MatSolverTypeGet()`
4765: `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`, `MatInitializePackage()`
4766: @*/
4767: PetscErrorCode MatGetFactor(Mat mat, MatSolverType type, MatFactorType ftype, Mat *f)
4768: {
4769: PetscBool foundtype, foundmtype;
4770: PetscErrorCode (*conv)(Mat, MatFactorType, Mat *);
4772: PetscFunctionBegin;
4776: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4777: MatCheckPreallocated(mat, 1);
4779: PetscCall(MatSolverTypeGet(type, ((PetscObject)mat)->type_name, ftype, &foundtype, &foundmtype, &conv));
4780: if (!foundtype) {
4781: if (type) {
4782: 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],
4783: ((PetscObject)mat)->type_name, type);
4784: } else {
4785: 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);
4786: }
4787: }
4788: PetscCheck(foundmtype, PetscObjectComm((PetscObject)mat), PETSC_ERR_MISSING_FACTOR, "MatSolverType %s does not support matrix type %s", type, ((PetscObject)mat)->type_name);
4789: 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);
4791: PetscCall((*conv)(mat, ftype, f));
4792: if (mat->factorprefix) PetscCall(MatSetOptionsPrefix(*f, mat->factorprefix));
4793: PetscFunctionReturn(PETSC_SUCCESS);
4794: }
4796: /*@C
4797: MatGetFactorAvailable - Returns a a flag if matrix supports particular type and factor type
4799: Not Collective
4801: Input Parameters:
4802: + mat - the matrix
4803: . type - name of solver type, for example, superlu, petsc (to use PETSc's default)
4804: - ftype - factor type, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`
4806: Output Parameter:
4807: . flg - PETSC_TRUE if the factorization is available
4809: Level: intermediate
4811: Notes:
4812: Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
4813: such as pastix, superlu, mumps etc.
4815: PETSc must have been ./configure to use the external solver, using the option --download-package
4817: Developer Note:
4818: This should actually be called `MatCreateFactorAvailable()` since `MatGetFactor()` creates a new factor object
4820: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatSolverType`, `MatFactorType`, `MatGetFactor()`, `MatCopy()`, `MatDuplicate()`, `MatSolverTypeRegister()`,
4821: `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`, `MatSolverTypeGet()`
4822: @*/
4823: PetscErrorCode MatGetFactorAvailable(Mat mat, MatSolverType type, MatFactorType ftype, PetscBool *flg)
4824: {
4825: PetscErrorCode (*gconv)(Mat, MatFactorType, Mat *);
4827: PetscFunctionBegin;
4830: PetscAssertPointer(flg, 4);
4832: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4833: MatCheckPreallocated(mat, 1);
4835: PetscCall(MatSolverTypeGet(type, ((PetscObject)mat)->type_name, ftype, NULL, NULL, &gconv));
4836: *flg = gconv ? PETSC_TRUE : PETSC_FALSE;
4837: PetscFunctionReturn(PETSC_SUCCESS);
4838: }
4840: /*@
4841: MatDuplicate - Duplicates a matrix including the non-zero structure.
4843: Collective
4845: Input Parameters:
4846: + mat - the matrix
4847: - op - One of `MAT_DO_NOT_COPY_VALUES`, `MAT_COPY_VALUES`, or `MAT_SHARE_NONZERO_PATTERN`.
4848: See the manual page for `MatDuplicateOption()` for an explanation of these options.
4850: Output Parameter:
4851: . M - pointer to place new matrix
4853: Level: intermediate
4855: Notes:
4856: You cannot change the nonzero pattern for the parent or child matrix later if you use `MAT_SHARE_NONZERO_PATTERN`.
4858: If `op` is not `MAT_COPY_VALUES` the numerical values in the new matrix are zeroed.
4860: 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.
4862: When original mat is a product of matrix operation, e.g., an output of `MatMatMult()` or `MatCreateSubMatrix()`, only the matrix data structure of `mat`
4863: is duplicated and the internal data structures created for the reuse of previous matrix operations are not duplicated.
4864: User should not use `MatDuplicate()` to create new matrix `M` if `M` is intended to be reused as the product of matrix operation.
4866: .seealso: [](ch_matrices), `Mat`, `MatCopy()`, `MatConvert()`, `MatDuplicateOption`
4867: @*/
4868: PetscErrorCode MatDuplicate(Mat mat, MatDuplicateOption op, Mat *M)
4869: {
4870: Mat B;
4871: VecType vtype;
4872: PetscInt i;
4873: PetscObject dm, container_h, container_d;
4874: void (*viewf)(void);
4876: PetscFunctionBegin;
4879: PetscAssertPointer(M, 3);
4880: PetscCheck(op != MAT_COPY_VALUES || mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "MAT_COPY_VALUES not allowed for unassembled matrix");
4881: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4882: MatCheckPreallocated(mat, 1);
4884: *M = NULL;
4885: PetscCall(PetscLogEventBegin(MAT_Convert, mat, 0, 0, 0));
4886: PetscUseTypeMethod(mat, duplicate, op, M);
4887: PetscCall(PetscLogEventEnd(MAT_Convert, mat, 0, 0, 0));
4888: B = *M;
4890: PetscCall(MatGetOperation(mat, MATOP_VIEW, &viewf));
4891: if (viewf) PetscCall(MatSetOperation(B, MATOP_VIEW, viewf));
4892: PetscCall(MatGetVecType(mat, &vtype));
4893: PetscCall(MatSetVecType(B, vtype));
4895: B->stencil.dim = mat->stencil.dim;
4896: B->stencil.noc = mat->stencil.noc;
4897: for (i = 0; i <= mat->stencil.dim + (mat->stencil.noc ? 0 : -1); i++) {
4898: B->stencil.dims[i] = mat->stencil.dims[i];
4899: B->stencil.starts[i] = mat->stencil.starts[i];
4900: }
4902: B->nooffproczerorows = mat->nooffproczerorows;
4903: B->nooffprocentries = mat->nooffprocentries;
4905: PetscCall(PetscObjectQuery((PetscObject)mat, "__PETSc_dm", &dm));
4906: if (dm) PetscCall(PetscObjectCompose((PetscObject)B, "__PETSc_dm", dm));
4907: PetscCall(PetscObjectQuery((PetscObject)mat, "__PETSc_MatCOOStruct_Host", &container_h));
4908: if (container_h) PetscCall(PetscObjectCompose((PetscObject)B, "__PETSc_MatCOOStruct_Host", container_h));
4909: PetscCall(PetscObjectQuery((PetscObject)mat, "__PETSc_MatCOOStruct_Device", &container_d));
4910: if (container_d) PetscCall(PetscObjectCompose((PetscObject)B, "__PETSc_MatCOOStruct_Device", container_d));
4911: PetscCall(PetscObjectStateIncrease((PetscObject)B));
4912: PetscFunctionReturn(PETSC_SUCCESS);
4913: }
4915: /*@
4916: MatGetDiagonal - Gets the diagonal of a matrix as a `Vec`
4918: Logically Collective
4920: Input Parameter:
4921: . mat - the matrix
4923: Output Parameter:
4924: . v - the diagonal of the matrix
4926: Level: intermediate
4928: Note:
4929: If `mat` has local sizes `n` x `m`, this routine fills the first `ndiag = min(n, m)` entries
4930: of `v` with the diagonal values. Thus `v` must have local size of at least `ndiag`. If `v`
4931: is larger than `ndiag`, the values of the remaining entries are unspecified.
4933: Currently only correct in parallel for square matrices.
4935: .seealso: [](ch_matrices), `Mat`, `Vec`, `MatGetRow()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMaxAbs()`
4936: @*/
4937: PetscErrorCode MatGetDiagonal(Mat mat, Vec v)
4938: {
4939: PetscFunctionBegin;
4943: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4944: MatCheckPreallocated(mat, 1);
4945: if (PetscDefined(USE_DEBUG)) {
4946: PetscInt nv, row, col, ndiag;
4948: PetscCall(VecGetLocalSize(v, &nv));
4949: PetscCall(MatGetLocalSize(mat, &row, &col));
4950: ndiag = PetscMin(row, col);
4951: 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);
4952: }
4954: PetscUseTypeMethod(mat, getdiagonal, v);
4955: PetscCall(PetscObjectStateIncrease((PetscObject)v));
4956: PetscFunctionReturn(PETSC_SUCCESS);
4957: }
4959: /*@C
4960: MatGetRowMin - Gets the minimum value (of the real part) of each
4961: row of the matrix
4963: Logically Collective
4965: Input Parameter:
4966: . mat - the matrix
4968: Output Parameters:
4969: + v - the vector for storing the maximums
4970: - idx - the indices of the column found for each row (optional)
4972: Level: intermediate
4974: Note:
4975: The result of this call are the same as if one converted the matrix to dense format
4976: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
4978: This code is only implemented for a couple of matrix formats.
4980: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMaxAbs()`, `MatGetRowMinAbs()`,
4981: `MatGetRowMax()`
4982: @*/
4983: PetscErrorCode MatGetRowMin(Mat mat, Vec v, PetscInt idx[])
4984: {
4985: PetscFunctionBegin;
4989: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4991: if (!mat->cmap->N) {
4992: PetscCall(VecSet(v, PETSC_MAX_REAL));
4993: if (idx) {
4994: PetscInt i, m = mat->rmap->n;
4995: for (i = 0; i < m; i++) idx[i] = -1;
4996: }
4997: } else {
4998: MatCheckPreallocated(mat, 1);
4999: }
5000: PetscUseTypeMethod(mat, getrowmin, v, idx);
5001: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5002: PetscFunctionReturn(PETSC_SUCCESS);
5003: }
5005: /*@C
5006: MatGetRowMinAbs - Gets the minimum value (in absolute value) of each
5007: row of the matrix
5009: Logically Collective
5011: Input Parameter:
5012: . mat - the matrix
5014: Output Parameters:
5015: + v - the vector for storing the minimums
5016: - idx - the indices of the column found for each row (or `NULL` if not needed)
5018: Level: intermediate
5020: Notes:
5021: if a row is completely empty or has only 0.0 values then the `idx` value for that
5022: row is 0 (the first column).
5024: This code is only implemented for a couple of matrix formats.
5026: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMax()`, `MatGetRowMaxAbs()`, `MatGetRowMin()`
5027: @*/
5028: PetscErrorCode MatGetRowMinAbs(Mat mat, Vec v, PetscInt idx[])
5029: {
5030: PetscFunctionBegin;
5034: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5035: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5037: if (!mat->cmap->N) {
5038: PetscCall(VecSet(v, 0.0));
5039: if (idx) {
5040: PetscInt i, m = mat->rmap->n;
5041: for (i = 0; i < m; i++) idx[i] = -1;
5042: }
5043: } else {
5044: MatCheckPreallocated(mat, 1);
5045: if (idx) PetscCall(PetscArrayzero(idx, mat->rmap->n));
5046: PetscUseTypeMethod(mat, getrowminabs, v, idx);
5047: }
5048: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5049: PetscFunctionReturn(PETSC_SUCCESS);
5050: }
5052: /*@C
5053: MatGetRowMax - Gets the maximum value (of the real part) of each
5054: row of the matrix
5056: Logically Collective
5058: Input Parameter:
5059: . mat - the matrix
5061: Output Parameters:
5062: + v - the vector for storing the maximums
5063: - idx - the indices of the column found for each row (optional)
5065: Level: intermediate
5067: Notes:
5068: The result of this call are the same as if one converted the matrix to dense format
5069: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
5071: This code is only implemented for a couple of matrix formats.
5073: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMaxAbs()`, `MatGetRowMin()`, `MatGetRowMinAbs()`
5074: @*/
5075: PetscErrorCode MatGetRowMax(Mat mat, Vec v, PetscInt idx[])
5076: {
5077: PetscFunctionBegin;
5081: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5083: if (!mat->cmap->N) {
5084: PetscCall(VecSet(v, PETSC_MIN_REAL));
5085: if (idx) {
5086: PetscInt i, m = mat->rmap->n;
5087: for (i = 0; i < m; i++) idx[i] = -1;
5088: }
5089: } else {
5090: MatCheckPreallocated(mat, 1);
5091: PetscUseTypeMethod(mat, getrowmax, v, idx);
5092: }
5093: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5094: PetscFunctionReturn(PETSC_SUCCESS);
5095: }
5097: /*@C
5098: MatGetRowMaxAbs - Gets the maximum value (in absolute value) of each
5099: row of the matrix
5101: Logically Collective
5103: Input Parameter:
5104: . mat - the matrix
5106: Output Parameters:
5107: + v - the vector for storing the maximums
5108: - idx - the indices of the column found for each row (or `NULL` if not needed)
5110: Level: intermediate
5112: Notes:
5113: if a row is completely empty or has only 0.0 values then the `idx` value for that
5114: row is 0 (the first column).
5116: This code is only implemented for a couple of matrix formats.
5118: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMax()`, `MatGetRowMin()`, `MatGetRowMinAbs()`
5119: @*/
5120: PetscErrorCode MatGetRowMaxAbs(Mat mat, Vec v, PetscInt idx[])
5121: {
5122: PetscFunctionBegin;
5126: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5128: if (!mat->cmap->N) {
5129: PetscCall(VecSet(v, 0.0));
5130: if (idx) {
5131: PetscInt i, m = mat->rmap->n;
5132: for (i = 0; i < m; i++) idx[i] = -1;
5133: }
5134: } else {
5135: MatCheckPreallocated(mat, 1);
5136: if (idx) PetscCall(PetscArrayzero(idx, mat->rmap->n));
5137: PetscUseTypeMethod(mat, getrowmaxabs, v, idx);
5138: }
5139: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5140: PetscFunctionReturn(PETSC_SUCCESS);
5141: }
5143: /*@
5144: MatGetRowSum - Gets the sum of each row of the matrix
5146: Logically or Neighborhood Collective
5148: Input Parameter:
5149: . mat - the matrix
5151: Output Parameter:
5152: . v - the vector for storing the sum of rows
5154: Level: intermediate
5156: Note:
5157: This code is slow since it is not currently specialized for different formats
5159: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMax()`, `MatGetRowMin()`, `MatGetRowMaxAbs()`, `MatGetRowMinAbs()`
5160: @*/
5161: PetscErrorCode MatGetRowSum(Mat mat, Vec v)
5162: {
5163: Vec ones;
5165: PetscFunctionBegin;
5169: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5170: MatCheckPreallocated(mat, 1);
5171: PetscCall(MatCreateVecs(mat, &ones, NULL));
5172: PetscCall(VecSet(ones, 1.));
5173: PetscCall(MatMult(mat, ones, v));
5174: PetscCall(VecDestroy(&ones));
5175: PetscFunctionReturn(PETSC_SUCCESS);
5176: }
5178: /*@
5179: MatTransposeSetPrecursor - Set the matrix from which the second matrix will receive numerical transpose data with a call to `MatTranspose`(A,`MAT_REUSE_MATRIX`,&B)
5180: when B was not obtained with `MatTranspose`(A,`MAT_INITIAL_MATRIX`,&B)
5182: Collective
5184: Input Parameter:
5185: . mat - the matrix to provide the transpose
5187: Output Parameter:
5188: . 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
5190: Level: advanced
5192: Note:
5193: 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
5194: routine allows bypassing that call.
5196: .seealso: [](ch_matrices), `Mat`, `MatTransposeSymbolic()`, `MatTranspose()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`
5197: @*/
5198: PetscErrorCode MatTransposeSetPrecursor(Mat mat, Mat B)
5199: {
5200: PetscContainer rB = NULL;
5201: MatParentState *rb = NULL;
5203: PetscFunctionBegin;
5204: PetscCall(PetscNew(&rb));
5205: rb->id = ((PetscObject)mat)->id;
5206: rb->state = 0;
5207: PetscCall(MatGetNonzeroState(mat, &rb->nonzerostate));
5208: PetscCall(PetscContainerCreate(PetscObjectComm((PetscObject)B), &rB));
5209: PetscCall(PetscContainerSetPointer(rB, rb));
5210: PetscCall(PetscContainerSetUserDestroy(rB, PetscContainerUserDestroyDefault));
5211: PetscCall(PetscObjectCompose((PetscObject)B, "MatTransposeParent", (PetscObject)rB));
5212: PetscCall(PetscObjectDereference((PetscObject)rB));
5213: PetscFunctionReturn(PETSC_SUCCESS);
5214: }
5216: /*@
5217: MatTranspose - Computes an in-place or out-of-place transpose of a matrix.
5219: Collective
5221: Input Parameters:
5222: + mat - the matrix to transpose
5223: - reuse - either `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, or `MAT_INPLACE_MATRIX`
5225: Output Parameter:
5226: . B - the transpose
5228: Level: intermediate
5230: Notes:
5231: If you use `MAT_INPLACE_MATRIX` then you must pass in `&mat` for `B`
5233: `MAT_REUSE_MATRIX` uses the `B` matrix obtained from a previous call to this function with `MAT_INITIAL_MATRIX`. If you already have a matrix to contain the
5234: transpose, call `MatTransposeSetPrecursor(mat, B)` before calling this routine.
5236: 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.
5238: Consider using `MatCreateTranspose()` instead if you only need a matrix that behaves like the transpose, but don't need the storage to be changed.
5240: If mat is unchanged from the last call this function returns immediately without recomputing the result
5242: If you only need the symbolic transpose, and not the numerical values, use `MatTransposeSymbolic()`
5244: .seealso: [](ch_matrices), `Mat`, `MatTransposeSetPrecursor()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`,
5245: `MatTransposeSymbolic()`, `MatCreateTranspose()`
5246: @*/
5247: PetscErrorCode MatTranspose(Mat mat, MatReuse reuse, Mat *B)
5248: {
5249: PetscContainer rB = NULL;
5250: MatParentState *rb = NULL;
5252: PetscFunctionBegin;
5255: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5256: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5257: PetscCheck(reuse != MAT_INPLACE_MATRIX || mat == *B, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "MAT_INPLACE_MATRIX requires last matrix to match first");
5258: PetscCheck(reuse != MAT_REUSE_MATRIX || mat != *B, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Perhaps you mean MAT_INPLACE_MATRIX");
5259: MatCheckPreallocated(mat, 1);
5260: if (reuse == MAT_REUSE_MATRIX) {
5261: PetscCall(PetscObjectQuery((PetscObject)*B, "MatTransposeParent", (PetscObject *)&rB));
5262: PetscCheck(rB, PetscObjectComm((PetscObject)*B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from call to MatTranspose(). Suggest MatTransposeSetPrecursor().");
5263: PetscCall(PetscContainerGetPointer(rB, (void **)&rb));
5264: PetscCheck(rb->id == ((PetscObject)mat)->id, PetscObjectComm((PetscObject)*B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from input matrix");
5265: if (rb->state == ((PetscObject)mat)->state) PetscFunctionReturn(PETSC_SUCCESS);
5266: }
5268: PetscCall(PetscLogEventBegin(MAT_Transpose, mat, 0, 0, 0));
5269: if (reuse != MAT_INPLACE_MATRIX || mat->symmetric != PETSC_BOOL3_TRUE) {
5270: PetscUseTypeMethod(mat, transpose, reuse, B);
5271: PetscCall(PetscObjectStateIncrease((PetscObject)*B));
5272: }
5273: PetscCall(PetscLogEventEnd(MAT_Transpose, mat, 0, 0, 0));
5275: if (reuse == MAT_INITIAL_MATRIX) PetscCall(MatTransposeSetPrecursor(mat, *B));
5276: if (reuse != MAT_INPLACE_MATRIX) {
5277: PetscCall(PetscObjectQuery((PetscObject)*B, "MatTransposeParent", (PetscObject *)&rB));
5278: PetscCall(PetscContainerGetPointer(rB, (void **)&rb));
5279: rb->state = ((PetscObject)mat)->state;
5280: rb->nonzerostate = mat->nonzerostate;
5281: }
5282: PetscFunctionReturn(PETSC_SUCCESS);
5283: }
5285: /*@
5286: MatTransposeSymbolic - Computes the symbolic part of the transpose of a matrix.
5288: Collective
5290: Input Parameter:
5291: . A - the matrix to transpose
5293: Output Parameter:
5294: . 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
5295: numerical portion.
5297: Level: intermediate
5299: Note:
5300: This is not supported for many matrix types, use `MatTranspose()` in those cases
5302: .seealso: [](ch_matrices), `Mat`, `MatTransposeSetPrecursor()`, `MatTranspose()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`
5303: @*/
5304: PetscErrorCode MatTransposeSymbolic(Mat A, Mat *B)
5305: {
5306: PetscFunctionBegin;
5309: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5310: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5311: PetscCall(PetscLogEventBegin(MAT_Transpose, A, 0, 0, 0));
5312: PetscUseTypeMethod(A, transposesymbolic, B);
5313: PetscCall(PetscLogEventEnd(MAT_Transpose, A, 0, 0, 0));
5315: PetscCall(MatTransposeSetPrecursor(A, *B));
5316: PetscFunctionReturn(PETSC_SUCCESS);
5317: }
5319: PetscErrorCode MatTransposeCheckNonzeroState_Private(Mat A, Mat B)
5320: {
5321: PetscContainer rB;
5322: MatParentState *rb;
5324: PetscFunctionBegin;
5327: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5328: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5329: PetscCall(PetscObjectQuery((PetscObject)B, "MatTransposeParent", (PetscObject *)&rB));
5330: PetscCheck(rB, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from call to MatTranspose()");
5331: PetscCall(PetscContainerGetPointer(rB, (void **)&rb));
5332: PetscCheck(rb->id == ((PetscObject)A)->id, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from input matrix");
5333: PetscCheck(rb->nonzerostate == A->nonzerostate, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONGSTATE, "Reuse matrix has changed nonzero structure");
5334: PetscFunctionReturn(PETSC_SUCCESS);
5335: }
5337: /*@
5338: MatIsTranspose - Test whether a matrix is another one's transpose,
5339: or its own, in which case it tests symmetry.
5341: Collective
5343: Input Parameters:
5344: + A - the matrix to test
5345: . B - the matrix to test against, this can equal the first parameter
5346: - tol - tolerance, differences between entries smaller than this are counted as zero
5348: Output Parameter:
5349: . flg - the result
5351: Level: intermediate
5353: Notes:
5354: Only available for `MATAIJ` matrices.
5356: The sequential algorithm has a running time of the order of the number of nonzeros; the parallel
5357: test involves parallel copies of the block off-diagonal parts of the matrix.
5359: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatIsSymmetric()`, `MatIsHermitian()`
5360: @*/
5361: PetscErrorCode MatIsTranspose(Mat A, Mat B, PetscReal tol, PetscBool *flg)
5362: {
5363: PetscErrorCode (*f)(Mat, Mat, PetscReal, PetscBool *), (*g)(Mat, Mat, PetscReal, PetscBool *);
5365: PetscFunctionBegin;
5368: PetscAssertPointer(flg, 4);
5369: PetscCall(PetscObjectQueryFunction((PetscObject)A, "MatIsTranspose_C", &f));
5370: PetscCall(PetscObjectQueryFunction((PetscObject)B, "MatIsTranspose_C", &g));
5371: *flg = PETSC_FALSE;
5372: if (f && g) {
5373: PetscCheck(f == g, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_NOTSAMETYPE, "Matrices do not have the same comparator for symmetry test");
5374: PetscCall((*f)(A, B, tol, flg));
5375: } else {
5376: MatType mattype;
5378: PetscCall(MatGetType(f ? B : A, &mattype));
5379: SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "Matrix of type %s does not support checking for transpose", mattype);
5380: }
5381: PetscFunctionReturn(PETSC_SUCCESS);
5382: }
5384: /*@
5385: MatHermitianTranspose - Computes an in-place or out-of-place Hermitian transpose of a matrix in complex conjugate.
5387: Collective
5389: Input Parameters:
5390: + mat - the matrix to transpose and complex conjugate
5391: - reuse - either `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, or `MAT_INPLACE_MATRIX`
5393: Output Parameter:
5394: . B - the Hermitian transpose
5396: Level: intermediate
5398: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`
5399: @*/
5400: PetscErrorCode MatHermitianTranspose(Mat mat, MatReuse reuse, Mat *B)
5401: {
5402: PetscFunctionBegin;
5403: PetscCall(MatTranspose(mat, reuse, B));
5404: #if defined(PETSC_USE_COMPLEX)
5405: PetscCall(MatConjugate(*B));
5406: #endif
5407: PetscFunctionReturn(PETSC_SUCCESS);
5408: }
5410: /*@
5411: MatIsHermitianTranspose - Test whether a matrix is another one's Hermitian transpose,
5413: Collective
5415: Input Parameters:
5416: + A - the matrix to test
5417: . B - the matrix to test against, this can equal the first parameter
5418: - tol - tolerance, differences between entries smaller than this are counted as zero
5420: Output Parameter:
5421: . flg - the result
5423: Level: intermediate
5425: Notes:
5426: Only available for `MATAIJ` matrices.
5428: The sequential algorithm
5429: has a running time of the order of the number of nonzeros; the parallel
5430: test involves parallel copies of the block off-diagonal parts of the matrix.
5432: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatIsSymmetric()`, `MatIsHermitian()`, `MatIsTranspose()`
5433: @*/
5434: PetscErrorCode MatIsHermitianTranspose(Mat A, Mat B, PetscReal tol, PetscBool *flg)
5435: {
5436: PetscErrorCode (*f)(Mat, Mat, PetscReal, PetscBool *), (*g)(Mat, Mat, PetscReal, PetscBool *);
5438: PetscFunctionBegin;
5441: PetscAssertPointer(flg, 4);
5442: PetscCall(PetscObjectQueryFunction((PetscObject)A, "MatIsHermitianTranspose_C", &f));
5443: PetscCall(PetscObjectQueryFunction((PetscObject)B, "MatIsHermitianTranspose_C", &g));
5444: if (f && g) {
5445: PetscCheck(f != g, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_NOTSAMETYPE, "Matrices do not have the same comparator for Hermitian test");
5446: PetscCall((*f)(A, B, tol, flg));
5447: }
5448: PetscFunctionReturn(PETSC_SUCCESS);
5449: }
5451: /*@
5452: MatPermute - Creates a new matrix with rows and columns permuted from the
5453: original.
5455: Collective
5457: Input Parameters:
5458: + mat - the matrix to permute
5459: . row - row permutation, each processor supplies only the permutation for its rows
5460: - col - column permutation, each processor supplies only the permutation for its columns
5462: Output Parameter:
5463: . B - the permuted matrix
5465: Level: advanced
5467: Note:
5468: The index sets map from row/col of permuted matrix to row/col of original matrix.
5469: The index sets should be on the same communicator as mat and have the same local sizes.
5471: Developer Note:
5472: If you want to implement `MatPermute()` for a matrix type, and your approach doesn't
5473: exploit the fact that row and col are permutations, consider implementing the
5474: more general `MatCreateSubMatrix()` instead.
5476: .seealso: [](ch_matrices), `Mat`, `MatGetOrdering()`, `ISAllGather()`, `MatCreateSubMatrix()`
5477: @*/
5478: PetscErrorCode MatPermute(Mat mat, IS row, IS col, Mat *B)
5479: {
5480: PetscFunctionBegin;
5485: PetscAssertPointer(B, 4);
5486: PetscCheckSameComm(mat, 1, row, 2);
5487: if (row != col) PetscCheckSameComm(row, 2, col, 3);
5488: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5489: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5490: PetscCheck(mat->ops->permute || mat->ops->createsubmatrix, PETSC_COMM_SELF, PETSC_ERR_SUP, "MatPermute not available for Mat type %s", ((PetscObject)mat)->type_name);
5491: MatCheckPreallocated(mat, 1);
5493: if (mat->ops->permute) {
5494: PetscUseTypeMethod(mat, permute, row, col, B);
5495: PetscCall(PetscObjectStateIncrease((PetscObject)*B));
5496: } else {
5497: PetscCall(MatCreateSubMatrix(mat, row, col, MAT_INITIAL_MATRIX, B));
5498: }
5499: PetscFunctionReturn(PETSC_SUCCESS);
5500: }
5502: /*@
5503: MatEqual - Compares two matrices.
5505: Collective
5507: Input Parameters:
5508: + A - the first matrix
5509: - B - the second matrix
5511: Output Parameter:
5512: . flg - `PETSC_TRUE` if the matrices are equal; `PETSC_FALSE` otherwise.
5514: Level: intermediate
5516: .seealso: [](ch_matrices), `Mat`
5517: @*/
5518: PetscErrorCode MatEqual(Mat A, Mat B, PetscBool *flg)
5519: {
5520: PetscFunctionBegin;
5525: PetscAssertPointer(flg, 3);
5526: PetscCheckSameComm(A, 1, B, 2);
5527: MatCheckPreallocated(A, 1);
5528: MatCheckPreallocated(B, 2);
5529: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5530: PetscCheck(B->assembled, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5531: 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,
5532: B->cmap->N);
5533: if (A->ops->equal && A->ops->equal == B->ops->equal) {
5534: PetscUseTypeMethod(A, equal, B, flg);
5535: } else {
5536: PetscCall(MatMultEqual(A, B, 10, flg));
5537: }
5538: PetscFunctionReturn(PETSC_SUCCESS);
5539: }
5541: /*@
5542: MatDiagonalScale - Scales a matrix on the left and right by diagonal
5543: matrices that are stored as vectors. Either of the two scaling
5544: matrices can be `NULL`.
5546: Collective
5548: Input Parameters:
5549: + mat - the matrix to be scaled
5550: . l - the left scaling vector (or `NULL`)
5551: - r - the right scaling vector (or `NULL`)
5553: Level: intermediate
5555: Note:
5556: `MatDiagonalScale()` computes $A = LAR$, where
5557: L = a diagonal matrix (stored as a vector), R = a diagonal matrix (stored as a vector)
5558: The L scales the rows of the matrix, the R scales the columns of the matrix.
5560: .seealso: [](ch_matrices), `Mat`, `MatScale()`, `MatShift()`, `MatDiagonalSet()`
5561: @*/
5562: PetscErrorCode MatDiagonalScale(Mat mat, Vec l, Vec r)
5563: {
5564: PetscFunctionBegin;
5567: if (l) {
5569: PetscCheckSameComm(mat, 1, l, 2);
5570: }
5571: if (r) {
5573: PetscCheckSameComm(mat, 1, r, 3);
5574: }
5575: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5576: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5577: MatCheckPreallocated(mat, 1);
5578: if (!l && !r) PetscFunctionReturn(PETSC_SUCCESS);
5580: PetscCall(PetscLogEventBegin(MAT_Scale, mat, 0, 0, 0));
5581: PetscUseTypeMethod(mat, diagonalscale, l, r);
5582: PetscCall(PetscLogEventEnd(MAT_Scale, mat, 0, 0, 0));
5583: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
5584: if (l != r) mat->symmetric = PETSC_BOOL3_FALSE;
5585: PetscFunctionReturn(PETSC_SUCCESS);
5586: }
5588: /*@
5589: MatScale - Scales all elements of a matrix by a given number.
5591: Logically Collective
5593: Input Parameters:
5594: + mat - the matrix to be scaled
5595: - a - the scaling value
5597: Level: intermediate
5599: .seealso: [](ch_matrices), `Mat`, `MatDiagonalScale()`
5600: @*/
5601: PetscErrorCode MatScale(Mat mat, PetscScalar a)
5602: {
5603: PetscFunctionBegin;
5606: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5607: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5609: MatCheckPreallocated(mat, 1);
5611: PetscCall(PetscLogEventBegin(MAT_Scale, mat, 0, 0, 0));
5612: if (a != (PetscScalar)1.0) {
5613: PetscUseTypeMethod(mat, scale, a);
5614: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
5615: }
5616: PetscCall(PetscLogEventEnd(MAT_Scale, mat, 0, 0, 0));
5617: PetscFunctionReturn(PETSC_SUCCESS);
5618: }
5620: /*@
5621: MatNorm - Calculates various norms of a matrix.
5623: Collective
5625: Input Parameters:
5626: + mat - the matrix
5627: - type - the type of norm, `NORM_1`, `NORM_FROBENIUS`, `NORM_INFINITY`
5629: Output Parameter:
5630: . nrm - the resulting norm
5632: Level: intermediate
5634: .seealso: [](ch_matrices), `Mat`
5635: @*/
5636: PetscErrorCode MatNorm(Mat mat, NormType type, PetscReal *nrm)
5637: {
5638: PetscFunctionBegin;
5641: PetscAssertPointer(nrm, 3);
5643: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5644: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5645: MatCheckPreallocated(mat, 1);
5647: PetscUseTypeMethod(mat, norm, type, nrm);
5648: PetscFunctionReturn(PETSC_SUCCESS);
5649: }
5651: /*
5652: This variable is used to prevent counting of MatAssemblyBegin() that
5653: are called from within a MatAssemblyEnd().
5654: */
5655: static PetscInt MatAssemblyEnd_InUse = 0;
5656: /*@
5657: MatAssemblyBegin - Begins assembling the matrix. This routine should
5658: be called after completing all calls to `MatSetValues()`.
5660: Collective
5662: Input Parameters:
5663: + mat - the matrix
5664: - type - type of assembly, either `MAT_FLUSH_ASSEMBLY` or `MAT_FINAL_ASSEMBLY`
5666: Level: beginner
5668: Notes:
5669: `MatSetValues()` generally caches the values that belong to other MPI processes. The matrix is ready to
5670: use only after `MatAssemblyBegin()` and `MatAssemblyEnd()` have been called.
5672: Use `MAT_FLUSH_ASSEMBLY` when switching between `ADD_VALUES` and `INSERT_VALUES`
5673: in `MatSetValues()`; use `MAT_FINAL_ASSEMBLY` for the final assembly before
5674: using the matrix.
5676: ALL processes that share a matrix MUST call `MatAssemblyBegin()` and `MatAssemblyEnd()` the SAME NUMBER of times, and each time with the
5677: 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
5678: a global collective operation requiring all processes that share the matrix.
5680: Space for preallocated nonzeros that is not filled by a call to `MatSetValues()` or a related routine are compressed
5681: out by assembly. If you intend to use that extra space on a subsequent assembly, be sure to insert explicit zeros
5682: before `MAT_FINAL_ASSEMBLY` so the space is not compressed out.
5684: .seealso: [](ch_matrices), `Mat`, `MatAssemblyEnd()`, `MatSetValues()`, `MatAssembled()`
5685: @*/
5686: PetscErrorCode MatAssemblyBegin(Mat mat, MatAssemblyType type)
5687: {
5688: PetscFunctionBegin;
5691: MatCheckPreallocated(mat, 1);
5692: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix.\nDid you forget to call MatSetUnfactored()?");
5693: if (mat->assembled) {
5694: mat->was_assembled = PETSC_TRUE;
5695: mat->assembled = PETSC_FALSE;
5696: }
5698: if (!MatAssemblyEnd_InUse) {
5699: PetscCall(PetscLogEventBegin(MAT_AssemblyBegin, mat, 0, 0, 0));
5700: PetscTryTypeMethod(mat, assemblybegin, type);
5701: PetscCall(PetscLogEventEnd(MAT_AssemblyBegin, mat, 0, 0, 0));
5702: } else PetscTryTypeMethod(mat, assemblybegin, type);
5703: PetscFunctionReturn(PETSC_SUCCESS);
5704: }
5706: /*@
5707: MatAssembled - Indicates if a matrix has been assembled and is ready for
5708: use; for example, in matrix-vector product.
5710: Not Collective
5712: Input Parameter:
5713: . mat - the matrix
5715: Output Parameter:
5716: . assembled - `PETSC_TRUE` or `PETSC_FALSE`
5718: Level: advanced
5720: .seealso: [](ch_matrices), `Mat`, `MatAssemblyEnd()`, `MatSetValues()`, `MatAssemblyBegin()`
5721: @*/
5722: PetscErrorCode MatAssembled(Mat mat, PetscBool *assembled)
5723: {
5724: PetscFunctionBegin;
5726: PetscAssertPointer(assembled, 2);
5727: *assembled = mat->assembled;
5728: PetscFunctionReturn(PETSC_SUCCESS);
5729: }
5731: /*@
5732: MatAssemblyEnd - Completes assembling the matrix. This routine should
5733: be called after `MatAssemblyBegin()`.
5735: Collective
5737: Input Parameters:
5738: + mat - the matrix
5739: - type - type of assembly, either `MAT_FLUSH_ASSEMBLY` or `MAT_FINAL_ASSEMBLY`
5741: Options Database Keys:
5742: + -mat_view ::ascii_info - Prints info on matrix at conclusion of `MatAssemblyEnd()`
5743: . -mat_view ::ascii_info_detail - Prints more detailed info
5744: . -mat_view - Prints matrix in ASCII format
5745: . -mat_view ::ascii_matlab - Prints matrix in MATLAB format
5746: . -mat_view draw - draws nonzero structure of matrix, using `MatView()` and `PetscDrawOpenX()`.
5747: . -display <name> - Sets display name (default is host)
5748: . -draw_pause <sec> - Sets number of seconds to pause after display
5749: . -mat_view socket - Sends matrix to socket, can be accessed from MATLAB (See [Using MATLAB with PETSc](ch_matlab))
5750: . -viewer_socket_machine <machine> - Machine to use for socket
5751: . -viewer_socket_port <port> - Port number to use for socket
5752: - -mat_view binary:filename[:append] - Save matrix to file in binary format
5754: Level: beginner
5756: .seealso: [](ch_matrices), `Mat`, `MatAssemblyBegin()`, `MatSetValues()`, `PetscDrawOpenX()`, `PetscDrawCreate()`, `MatView()`, `MatAssembled()`, `PetscViewerSocketOpen()`
5757: @*/
5758: PetscErrorCode MatAssemblyEnd(Mat mat, MatAssemblyType type)
5759: {
5760: static PetscInt inassm = 0;
5761: PetscBool flg = PETSC_FALSE;
5763: PetscFunctionBegin;
5767: inassm++;
5768: MatAssemblyEnd_InUse++;
5769: if (MatAssemblyEnd_InUse == 1) { /* Do the logging only the first time through */
5770: PetscCall(PetscLogEventBegin(MAT_AssemblyEnd, mat, 0, 0, 0));
5771: PetscTryTypeMethod(mat, assemblyend, type);
5772: PetscCall(PetscLogEventEnd(MAT_AssemblyEnd, mat, 0, 0, 0));
5773: } else PetscTryTypeMethod(mat, assemblyend, type);
5775: /* Flush assembly is not a true assembly */
5776: if (type != MAT_FLUSH_ASSEMBLY) {
5777: if (mat->num_ass) {
5778: if (!mat->symmetry_eternal) {
5779: mat->symmetric = PETSC_BOOL3_UNKNOWN;
5780: mat->hermitian = PETSC_BOOL3_UNKNOWN;
5781: }
5782: if (!mat->structural_symmetry_eternal && mat->ass_nonzerostate != mat->nonzerostate) mat->structurally_symmetric = PETSC_BOOL3_UNKNOWN;
5783: if (!mat->spd_eternal) mat->spd = PETSC_BOOL3_UNKNOWN;
5784: }
5785: mat->num_ass++;
5786: mat->assembled = PETSC_TRUE;
5787: mat->ass_nonzerostate = mat->nonzerostate;
5788: }
5790: mat->insertmode = NOT_SET_VALUES;
5791: MatAssemblyEnd_InUse--;
5792: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
5793: if (inassm == 1 && type != MAT_FLUSH_ASSEMBLY) {
5794: PetscCall(MatViewFromOptions(mat, NULL, "-mat_view"));
5796: if (mat->checksymmetryonassembly) {
5797: PetscCall(MatIsSymmetric(mat, mat->checksymmetrytol, &flg));
5798: if (flg) {
5799: PetscCall(PetscPrintf(PetscObjectComm((PetscObject)mat), "Matrix is symmetric (tolerance %g)\n", (double)mat->checksymmetrytol));
5800: } else {
5801: PetscCall(PetscPrintf(PetscObjectComm((PetscObject)mat), "Matrix is not symmetric (tolerance %g)\n", (double)mat->checksymmetrytol));
5802: }
5803: }
5804: if (mat->nullsp && mat->checknullspaceonassembly) PetscCall(MatNullSpaceTest(mat->nullsp, mat, NULL));
5805: }
5806: inassm--;
5807: PetscFunctionReturn(PETSC_SUCCESS);
5808: }
5810: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
5811: /*@
5812: MatSetOption - Sets a parameter option for a matrix. Some options
5813: may be specific to certain storage formats. Some options
5814: determine how values will be inserted (or added). Sorted,
5815: row-oriented input will generally assemble the fastest. The default
5816: is row-oriented.
5818: Logically Collective for certain operations, such as `MAT_SPD`, not collective for `MAT_ROW_ORIENTED`, see `MatOption`
5820: Input Parameters:
5821: + mat - the matrix
5822: . op - the option, one of those listed below (and possibly others),
5823: - flg - turn the option on (`PETSC_TRUE`) or off (`PETSC_FALSE`)
5825: Options Describing Matrix Structure:
5826: + `MAT_SPD` - symmetric positive definite
5827: . `MAT_SYMMETRIC` - symmetric in terms of both structure and value
5828: . `MAT_HERMITIAN` - transpose is the complex conjugation
5829: . `MAT_STRUCTURALLY_SYMMETRIC` - symmetric nonzero structure
5830: . `MAT_SYMMETRY_ETERNAL` - indicates the symmetry (or Hermitian structure) or its absence will persist through any changes to the matrix
5831: . `MAT_STRUCTURAL_SYMMETRY_ETERNAL` - indicates the structural symmetry or its absence will persist through any changes to the matrix
5832: . `MAT_SPD_ETERNAL` - indicates the value of `MAT_SPD` (true or false) will persist through any changes to the matrix
5834: These are not really options of the matrix, they are knowledge about the structure of the matrix that users may provide so that they
5835: do not need to be computed (usually at a high cost)
5837: Options For Use with `MatSetValues()`:
5838: Insert a logically dense subblock, which can be
5839: . `MAT_ROW_ORIENTED` - row-oriented (default)
5841: These options reflect the data you pass in with `MatSetValues()`; it has
5842: nothing to do with how the data is stored internally in the matrix
5843: data structure.
5845: When (re)assembling a matrix, we can restrict the input for
5846: efficiency/debugging purposes. These options include
5847: . `MAT_NEW_NONZERO_LOCATIONS` - additional insertions will be allowed if they generate a new nonzero (slow)
5848: . `MAT_FORCE_DIAGONAL_ENTRIES` - forces diagonal entries to be allocated
5849: . `MAT_IGNORE_OFF_PROC_ENTRIES` - drops off-processor entries
5850: . `MAT_NEW_NONZERO_LOCATION_ERR` - generates an error for new matrix entry
5851: . `MAT_USE_HASH_TABLE` - uses a hash table to speed up matrix assembly
5852: . `MAT_NO_OFF_PROC_ENTRIES` - you know each process will only set values for its own rows, will generate an error if
5853: any process sets values for another process. This avoids all reductions in the MatAssembly routines and thus improves
5854: performance for very large process counts.
5855: - `MAT_SUBSET_OFF_PROC_ENTRIES` - you know that the first assembly after setting this flag will set a superset
5856: of the off-process entries required for all subsequent assemblies. This avoids a rendezvous step in the MatAssembly
5857: functions, instead sending only neighbor messages.
5859: Level: intermediate
5861: Notes:
5862: Except for `MAT_UNUSED_NONZERO_LOCATION_ERR` and `MAT_ROW_ORIENTED` all processes that share the matrix must pass the same value in flg!
5864: Some options are relevant only for particular matrix types and
5865: are thus ignored by others. Other options are not supported by
5866: certain matrix types and will generate an error message if set.
5868: If using Fortran to compute a matrix, one may need to
5869: use the column-oriented option (or convert to the row-oriented
5870: format).
5872: `MAT_NEW_NONZERO_LOCATIONS` set to `PETSC_FALSE` indicates that any add or insertion
5873: that would generate a new entry in the nonzero structure is instead
5874: ignored. Thus, if memory has not already been allocated for this particular
5875: data, then the insertion is ignored. For dense matrices, in which
5876: the entire array is allocated, no entries are ever ignored.
5877: Set after the first `MatAssemblyEnd()`. If this option is set then the `MatAssemblyBegin()`/`MatAssemblyEnd()` processes has one less global reduction
5879: `MAT_NEW_NONZERO_LOCATION_ERR` set to PETSC_TRUE indicates that any add or insertion
5880: that would generate a new entry in the nonzero structure instead produces
5881: 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
5883: `MAT_NEW_NONZERO_ALLOCATION_ERR` set to `PETSC_TRUE` indicates that any add or insertion
5884: that would generate a new entry that has not been preallocated will
5885: instead produce an error. (Currently supported for `MATAIJ` and `MATBAIJ` formats
5886: only.) This is a useful flag when debugging matrix memory preallocation.
5887: If this option is set then the `MatAssemblyBegin()`/`MatAssemblyEnd()` processes has one less global reduction
5889: `MAT_IGNORE_OFF_PROC_ENTRIES` set to `PETSC_TRUE` indicates entries destined for
5890: other processors should be dropped, rather than stashed.
5891: This is useful if you know that the "owning" processor is also
5892: always generating the correct matrix entries, so that PETSc need
5893: not transfer duplicate entries generated on another processor.
5895: `MAT_USE_HASH_TABLE` indicates that a hash table be used to improve the
5896: searches during matrix assembly. When this flag is set, the hash table
5897: is created during the first matrix assembly. This hash table is
5898: used the next time through, during `MatSetValues()`/`MatSetValuesBlocked()`
5899: to improve the searching of indices. `MAT_NEW_NONZERO_LOCATIONS` flag
5900: should be used with `MAT_USE_HASH_TABLE` flag. This option is currently
5901: supported by `MATMPIBAIJ` format only.
5903: `MAT_KEEP_NONZERO_PATTERN` indicates when `MatZeroRows()` is called the zeroed entries
5904: are kept in the nonzero structure
5906: `MAT_IGNORE_ZERO_ENTRIES` - for `MATAIJ` and `MATIS` matrices this will stop zero values from creating
5907: a zero location in the matrix
5909: `MAT_USE_INODES` - indicates using inode version of the code - works with `MATAIJ` matrix types
5911: `MAT_NO_OFF_PROC_ZERO_ROWS` - you know each process will only zero its own rows. This avoids all reductions in the
5912: zero row routines and thus improves performance for very large process counts.
5914: `MAT_IGNORE_LOWER_TRIANGULAR` - For `MATSBAIJ` matrices will ignore any insertions you make in the lower triangular
5915: part of the matrix (since they should match the upper triangular part).
5917: `MAT_SORTED_FULL` - each process provides exactly its local rows; all column indices for a given row are passed in a
5918: single call to `MatSetValues()`, preallocation is perfect, row oriented, `INSERT_VALUES` is used. Common
5919: with finite difference schemes with non-periodic boundary conditions.
5921: Developer Note:
5922: `MAT_SYMMETRY_ETERNAL`, `MAT_STRUCTURAL_SYMMETRY_ETERNAL`, and `MAT_SPD_ETERNAL` are used by `MatAssemblyEnd()` and in other
5923: places where otherwise the value of `MAT_SYMMETRIC`, `MAT_STRUCTURALLY_SYMMETRIC` or `MAT_SPD` would need to be changed back
5924: to `PETSC_BOOL3_UNKNOWN` because the matrix values had changed so the code cannot be certain that the related property had
5925: not changed.
5927: .seealso: [](ch_matrices), `MatOption`, `Mat`, `MatGetOption()`
5928: @*/
5929: PetscErrorCode MatSetOption(Mat mat, MatOption op, PetscBool flg)
5930: {
5931: PetscFunctionBegin;
5933: if (op > 0) {
5936: }
5938: 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);
5940: switch (op) {
5941: case MAT_FORCE_DIAGONAL_ENTRIES:
5942: mat->force_diagonals = flg;
5943: PetscFunctionReturn(PETSC_SUCCESS);
5944: case MAT_NO_OFF_PROC_ENTRIES:
5945: mat->nooffprocentries = flg;
5946: PetscFunctionReturn(PETSC_SUCCESS);
5947: case MAT_SUBSET_OFF_PROC_ENTRIES:
5948: mat->assembly_subset = flg;
5949: if (!mat->assembly_subset) { /* See the same logic in VecAssembly wrt VEC_SUBSET_OFF_PROC_ENTRIES */
5950: #if !defined(PETSC_HAVE_MPIUNI)
5951: PetscCall(MatStashScatterDestroy_BTS(&mat->stash));
5952: #endif
5953: mat->stash.first_assembly_done = PETSC_FALSE;
5954: }
5955: PetscFunctionReturn(PETSC_SUCCESS);
5956: case MAT_NO_OFF_PROC_ZERO_ROWS:
5957: mat->nooffproczerorows = flg;
5958: PetscFunctionReturn(PETSC_SUCCESS);
5959: case MAT_SPD:
5960: if (flg) {
5961: mat->spd = PETSC_BOOL3_TRUE;
5962: mat->symmetric = PETSC_BOOL3_TRUE;
5963: mat->structurally_symmetric = PETSC_BOOL3_TRUE;
5964: } else {
5965: mat->spd = PETSC_BOOL3_FALSE;
5966: }
5967: break;
5968: case MAT_SYMMETRIC:
5969: mat->symmetric = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
5970: if (flg) mat->structurally_symmetric = PETSC_BOOL3_TRUE;
5971: #if !defined(PETSC_USE_COMPLEX)
5972: mat->hermitian = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
5973: #endif
5974: break;
5975: case MAT_HERMITIAN:
5976: mat->hermitian = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
5977: if (flg) mat->structurally_symmetric = PETSC_BOOL3_TRUE;
5978: #if !defined(PETSC_USE_COMPLEX)
5979: mat->symmetric = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
5980: #endif
5981: break;
5982: case MAT_STRUCTURALLY_SYMMETRIC:
5983: mat->structurally_symmetric = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
5984: break;
5985: case MAT_SYMMETRY_ETERNAL:
5986: 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");
5987: mat->symmetry_eternal = flg;
5988: if (flg) mat->structural_symmetry_eternal = PETSC_TRUE;
5989: break;
5990: case MAT_STRUCTURAL_SYMMETRY_ETERNAL:
5991: 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");
5992: mat->structural_symmetry_eternal = flg;
5993: break;
5994: case MAT_SPD_ETERNAL:
5995: 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");
5996: mat->spd_eternal = flg;
5997: if (flg) {
5998: mat->structural_symmetry_eternal = PETSC_TRUE;
5999: mat->symmetry_eternal = PETSC_TRUE;
6000: }
6001: break;
6002: case MAT_STRUCTURE_ONLY:
6003: mat->structure_only = flg;
6004: break;
6005: case MAT_SORTED_FULL:
6006: mat->sortedfull = flg;
6007: break;
6008: default:
6009: break;
6010: }
6011: PetscTryTypeMethod(mat, setoption, op, flg);
6012: PetscFunctionReturn(PETSC_SUCCESS);
6013: }
6015: /*@
6016: MatGetOption - Gets a parameter option that has been set for a matrix.
6018: Logically Collective
6020: Input Parameters:
6021: + mat - the matrix
6022: - op - the option, this only responds to certain options, check the code for which ones
6024: Output Parameter:
6025: . flg - turn the option on (`PETSC_TRUE`) or off (`PETSC_FALSE`)
6027: Level: intermediate
6029: Notes:
6030: Can only be called after `MatSetSizes()` and `MatSetType()` have been set.
6032: Certain option values may be unknown, for those use the routines `MatIsSymmetric()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, or
6033: `MatIsSymmetricKnown()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetricKnown()`
6035: .seealso: [](ch_matrices), `Mat`, `MatOption`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`,
6036: `MatIsSymmetricKnown()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetricKnown()`
6037: @*/
6038: PetscErrorCode MatGetOption(Mat mat, MatOption op, PetscBool *flg)
6039: {
6040: PetscFunctionBegin;
6044: 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);
6045: 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()");
6047: switch (op) {
6048: case MAT_NO_OFF_PROC_ENTRIES:
6049: *flg = mat->nooffprocentries;
6050: break;
6051: case MAT_NO_OFF_PROC_ZERO_ROWS:
6052: *flg = mat->nooffproczerorows;
6053: break;
6054: case MAT_SYMMETRIC:
6055: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsSymmetric() or MatIsSymmetricKnown()");
6056: break;
6057: case MAT_HERMITIAN:
6058: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsHermitian() or MatIsHermitianKnown()");
6059: break;
6060: case MAT_STRUCTURALLY_SYMMETRIC:
6061: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsStructurallySymmetric() or MatIsStructurallySymmetricKnown()");
6062: break;
6063: case MAT_SPD:
6064: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsSPDKnown()");
6065: break;
6066: case MAT_SYMMETRY_ETERNAL:
6067: *flg = mat->symmetry_eternal;
6068: break;
6069: case MAT_STRUCTURAL_SYMMETRY_ETERNAL:
6070: *flg = mat->symmetry_eternal;
6071: break;
6072: default:
6073: break;
6074: }
6075: PetscFunctionReturn(PETSC_SUCCESS);
6076: }
6078: /*@
6079: MatZeroEntries - Zeros all entries of a matrix. For sparse matrices
6080: this routine retains the old nonzero structure.
6082: Logically Collective
6084: Input Parameter:
6085: . mat - the matrix
6087: Level: intermediate
6089: Note:
6090: 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.
6091: See the Performance chapter of the users manual for information on preallocating matrices.
6093: .seealso: [](ch_matrices), `Mat`, `MatZeroRows()`, `MatZeroRowsColumns()`
6094: @*/
6095: PetscErrorCode MatZeroEntries(Mat mat)
6096: {
6097: PetscFunctionBegin;
6100: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6101: 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");
6102: MatCheckPreallocated(mat, 1);
6104: PetscCall(PetscLogEventBegin(MAT_ZeroEntries, mat, 0, 0, 0));
6105: PetscUseTypeMethod(mat, zeroentries);
6106: PetscCall(PetscLogEventEnd(MAT_ZeroEntries, mat, 0, 0, 0));
6107: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6108: PetscFunctionReturn(PETSC_SUCCESS);
6109: }
6111: /*@
6112: MatZeroRowsColumns - Zeros all entries (except possibly the main diagonal)
6113: of a set of rows and columns of a matrix.
6115: Collective
6117: Input Parameters:
6118: + mat - the matrix
6119: . numRows - the number of rows/columns to zero
6120: . rows - the global row indices
6121: . diag - value put in the diagonal of the eliminated rows
6122: . x - optional vector of the solution for zeroed rows (other entries in vector are not used), these must be set before this call
6123: - b - optional vector of the right hand side, that will be adjusted by provided solution entries
6125: Level: intermediate
6127: Notes:
6128: This routine, along with `MatZeroRows()`, is typically used to eliminate known Dirichlet boundary conditions from a linear system.
6130: For each zeroed row, the value of the corresponding `b` is set to diag times the value of the corresponding `x`.
6131: 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
6133: If the resulting linear system is to be solved with `KSP` then one can (but does not have to) call `KSPSetInitialGuessNonzero()` to allow the
6134: Krylov method to take advantage of the known solution on the zeroed rows.
6136: For the parallel case, all processes that share the matrix (i.e.,
6137: those in the communicator used for matrix creation) MUST call this
6138: routine, regardless of whether any rows being zeroed are owned by
6139: them.
6141: Unlike `MatZeroRows()` this does not change the nonzero structure of the matrix, it merely zeros those entries in the matrix.
6143: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
6144: list only rows local to itself).
6146: The option `MAT_NO_OFF_PROC_ZERO_ROWS` does not apply to this routine.
6148: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRows()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6149: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6150: @*/
6151: PetscErrorCode MatZeroRowsColumns(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6152: {
6153: PetscFunctionBegin;
6156: if (numRows) PetscAssertPointer(rows, 3);
6157: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6158: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6159: MatCheckPreallocated(mat, 1);
6161: PetscUseTypeMethod(mat, zerorowscolumns, numRows, rows, diag, x, b);
6162: PetscCall(MatViewFromOptions(mat, NULL, "-mat_view"));
6163: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6164: PetscFunctionReturn(PETSC_SUCCESS);
6165: }
6167: /*@
6168: MatZeroRowsColumnsIS - Zeros all entries (except possibly the main diagonal)
6169: of a set of rows and columns of a matrix.
6171: Collective
6173: Input Parameters:
6174: + mat - the matrix
6175: . is - the rows to zero
6176: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6177: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6178: - b - optional vector of right hand side, that will be adjusted by provided solution
6180: Level: intermediate
6182: Note:
6183: See `MatZeroRowsColumns()` for details on how this routine operates.
6185: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6186: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRows()`, `MatZeroRowsColumnsStencil()`
6187: @*/
6188: PetscErrorCode MatZeroRowsColumnsIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6189: {
6190: PetscInt numRows;
6191: const PetscInt *rows;
6193: PetscFunctionBegin;
6198: PetscCall(ISGetLocalSize(is, &numRows));
6199: PetscCall(ISGetIndices(is, &rows));
6200: PetscCall(MatZeroRowsColumns(mat, numRows, rows, diag, x, b));
6201: PetscCall(ISRestoreIndices(is, &rows));
6202: PetscFunctionReturn(PETSC_SUCCESS);
6203: }
6205: /*@
6206: MatZeroRows - Zeros all entries (except possibly the main diagonal)
6207: of a set of rows of a matrix.
6209: Collective
6211: Input Parameters:
6212: + mat - the matrix
6213: . numRows - the number of rows to zero
6214: . rows - the global row indices
6215: . diag - value put in the diagonal of the zeroed rows
6216: . x - optional vector of solutions for zeroed rows (other entries in vector are not used), these must be set before this call
6217: - b - optional vector of right hand side, that will be adjusted by provided solution entries
6219: Level: intermediate
6221: Notes:
6222: This routine, along with `MatZeroRowsColumns()`, is typically used to eliminate known Dirichlet boundary conditions from a linear system.
6224: For each zeroed row, the value of the corresponding `b` is set to `diag` times the value of the corresponding `x`.
6226: If the resulting linear system is to be solved with `KSP` then one can (but does not have to) call `KSPSetInitialGuessNonzero()` to allow the
6227: Krylov method to take advantage of the known solution on the zeroed rows.
6229: 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)
6230: from the matrix.
6232: Unlike `MatZeroRowsColumns()` for the `MATAIJ` and `MATBAIJ` matrix formats this removes the old nonzero structure, from the eliminated rows of the matrix
6233: but does not release memory. Because of this removal matrix-vector products with the adjusted matrix will be a bit faster. For the dense and block diagonal
6234: formats this does not alter the nonzero structure.
6236: If the option `MatSetOption`(mat,`MAT_KEEP_NONZERO_PATTERN`,`PETSC_TRUE`) the nonzero structure
6237: of the matrix is not changed the values are
6238: merely zeroed.
6240: The user can set a value in the diagonal entry (or for the `MATAIJ` format
6241: formats can optionally remove the main diagonal entry from the
6242: nonzero structure as well, by passing 0.0 as the final argument).
6244: For the parallel case, all processes that share the matrix (i.e.,
6245: those in the communicator used for matrix creation) MUST call this
6246: routine, regardless of whether any rows being zeroed are owned by
6247: them.
6249: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
6250: list only rows local to itself).
6252: You can call `MatSetOption`(mat,`MAT_NO_OFF_PROC_ZERO_ROWS`,`PETSC_TRUE`) if each process indicates only rows it
6253: owns that are to be zeroed. This saves a global synchronization in the implementation.
6255: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6256: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`, `PCREDISTRIBUTE`
6257: @*/
6258: PetscErrorCode MatZeroRows(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6259: {
6260: PetscFunctionBegin;
6263: if (numRows) PetscAssertPointer(rows, 3);
6264: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6265: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6266: MatCheckPreallocated(mat, 1);
6268: PetscUseTypeMethod(mat, zerorows, numRows, rows, diag, x, b);
6269: PetscCall(MatViewFromOptions(mat, NULL, "-mat_view"));
6270: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6271: PetscFunctionReturn(PETSC_SUCCESS);
6272: }
6274: /*@
6275: MatZeroRowsIS - Zeros all entries (except possibly the main diagonal)
6276: of a set of rows of a matrix.
6278: Collective
6280: Input Parameters:
6281: + mat - the matrix
6282: . is - index set of rows to remove (if `NULL` then no row is removed)
6283: . diag - value put in all diagonals of eliminated rows
6284: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6285: - b - optional vector of right hand side, that will be adjusted by provided solution
6287: Level: intermediate
6289: Note:
6290: See `MatZeroRows()` for details on how this routine operates.
6292: .seealso: [](ch_matrices), `Mat`, `MatZeroRows()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6293: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6294: @*/
6295: PetscErrorCode MatZeroRowsIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6296: {
6297: PetscInt numRows = 0;
6298: const PetscInt *rows = NULL;
6300: PetscFunctionBegin;
6303: if (is) {
6305: PetscCall(ISGetLocalSize(is, &numRows));
6306: PetscCall(ISGetIndices(is, &rows));
6307: }
6308: PetscCall(MatZeroRows(mat, numRows, rows, diag, x, b));
6309: if (is) PetscCall(ISRestoreIndices(is, &rows));
6310: PetscFunctionReturn(PETSC_SUCCESS);
6311: }
6313: /*@
6314: MatZeroRowsStencil - Zeros all entries (except possibly the main diagonal)
6315: of a set of rows of a matrix. These rows must be local to the process.
6317: Collective
6319: Input Parameters:
6320: + mat - the matrix
6321: . numRows - the number of rows to remove
6322: . rows - the grid coordinates (and component number when dof > 1) for matrix rows
6323: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6324: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6325: - b - optional vector of right hand side, that will be adjusted by provided solution
6327: Level: intermediate
6329: Notes:
6330: See `MatZeroRows()` for details on how this routine operates.
6332: The grid coordinates are across the entire grid, not just the local portion
6334: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
6335: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
6336: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
6337: `DM_BOUNDARY_PERIODIC` boundary type.
6339: 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
6340: a single value per point) you can skip filling those indices.
6342: Fortran Note:
6343: `idxm` and `idxn` should be declared as
6344: $ MatStencil idxm(4, m)
6345: and the values inserted using
6346: .vb
6347: idxm(MatStencil_i, 1) = i
6348: idxm(MatStencil_j, 1) = j
6349: idxm(MatStencil_k, 1) = k
6350: idxm(MatStencil_c, 1) = c
6351: etc
6352: .ve
6354: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsl()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6355: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6356: @*/
6357: PetscErrorCode MatZeroRowsStencil(Mat mat, PetscInt numRows, const MatStencil rows[], PetscScalar diag, Vec x, Vec b)
6358: {
6359: PetscInt dim = mat->stencil.dim;
6360: PetscInt sdim = dim - (1 - (PetscInt)mat->stencil.noc);
6361: PetscInt *dims = mat->stencil.dims + 1;
6362: PetscInt *starts = mat->stencil.starts;
6363: PetscInt *dxm = (PetscInt *)rows;
6364: PetscInt *jdxm, i, j, tmp, numNewRows = 0;
6366: PetscFunctionBegin;
6369: if (numRows) PetscAssertPointer(rows, 3);
6371: PetscCall(PetscMalloc1(numRows, &jdxm));
6372: for (i = 0; i < numRows; ++i) {
6373: /* Skip unused dimensions (they are ordered k, j, i, c) */
6374: for (j = 0; j < 3 - sdim; ++j) dxm++;
6375: /* Local index in X dir */
6376: tmp = *dxm++ - starts[0];
6377: /* Loop over remaining dimensions */
6378: for (j = 0; j < dim - 1; ++j) {
6379: /* If nonlocal, set index to be negative */
6380: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
6381: /* Update local index */
6382: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
6383: }
6384: /* Skip component slot if necessary */
6385: if (mat->stencil.noc) dxm++;
6386: /* Local row number */
6387: if (tmp >= 0) jdxm[numNewRows++] = tmp;
6388: }
6389: PetscCall(MatZeroRowsLocal(mat, numNewRows, jdxm, diag, x, b));
6390: PetscCall(PetscFree(jdxm));
6391: PetscFunctionReturn(PETSC_SUCCESS);
6392: }
6394: /*@
6395: MatZeroRowsColumnsStencil - Zeros all row and column entries (except possibly the main diagonal)
6396: of a set of rows and columns of a matrix.
6398: Collective
6400: Input Parameters:
6401: + mat - the matrix
6402: . numRows - the number of rows/columns to remove
6403: . rows - the grid coordinates (and component number when dof > 1) for matrix rows
6404: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6405: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6406: - b - optional vector of right hand side, that will be adjusted by provided solution
6408: Level: intermediate
6410: Notes:
6411: See `MatZeroRowsColumns()` for details on how this routine operates.
6413: The grid coordinates are across the entire grid, not just the local portion
6415: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
6416: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
6417: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
6418: `DM_BOUNDARY_PERIODIC` boundary type.
6420: 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
6421: a single value per point) you can skip filling those indices.
6423: Fortran Note:
6424: `idxm` and `idxn` should be declared as
6425: $ MatStencil idxm(4, m)
6426: and the values inserted using
6427: .vb
6428: idxm(MatStencil_i, 1) = i
6429: idxm(MatStencil_j, 1) = j
6430: idxm(MatStencil_k, 1) = k
6431: idxm(MatStencil_c, 1) = c
6432: etc
6433: .ve
6435: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6436: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRows()`
6437: @*/
6438: PetscErrorCode MatZeroRowsColumnsStencil(Mat mat, PetscInt numRows, const MatStencil rows[], PetscScalar diag, Vec x, Vec b)
6439: {
6440: PetscInt dim = mat->stencil.dim;
6441: PetscInt sdim = dim - (1 - (PetscInt)mat->stencil.noc);
6442: PetscInt *dims = mat->stencil.dims + 1;
6443: PetscInt *starts = mat->stencil.starts;
6444: PetscInt *dxm = (PetscInt *)rows;
6445: PetscInt *jdxm, i, j, tmp, numNewRows = 0;
6447: PetscFunctionBegin;
6450: if (numRows) PetscAssertPointer(rows, 3);
6452: PetscCall(PetscMalloc1(numRows, &jdxm));
6453: for (i = 0; i < numRows; ++i) {
6454: /* Skip unused dimensions (they are ordered k, j, i, c) */
6455: for (j = 0; j < 3 - sdim; ++j) dxm++;
6456: /* Local index in X dir */
6457: tmp = *dxm++ - starts[0];
6458: /* Loop over remaining dimensions */
6459: for (j = 0; j < dim - 1; ++j) {
6460: /* If nonlocal, set index to be negative */
6461: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
6462: /* Update local index */
6463: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
6464: }
6465: /* Skip component slot if necessary */
6466: if (mat->stencil.noc) dxm++;
6467: /* Local row number */
6468: if (tmp >= 0) jdxm[numNewRows++] = tmp;
6469: }
6470: PetscCall(MatZeroRowsColumnsLocal(mat, numNewRows, jdxm, diag, x, b));
6471: PetscCall(PetscFree(jdxm));
6472: PetscFunctionReturn(PETSC_SUCCESS);
6473: }
6475: /*@C
6476: MatZeroRowsLocal - Zeros all entries (except possibly the main diagonal)
6477: of a set of rows of a matrix; using local numbering of rows.
6479: Collective
6481: Input Parameters:
6482: + mat - the matrix
6483: . numRows - the number of rows to remove
6484: . rows - the local row indices
6485: . diag - value put in all diagonals of eliminated rows
6486: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6487: - b - optional vector of right hand side, that will be adjusted by provided solution
6489: Level: intermediate
6491: Notes:
6492: Before calling `MatZeroRowsLocal()`, the user must first set the
6493: local-to-global mapping by calling MatSetLocalToGlobalMapping(), this is often already set for matrices obtained with `DMCreateMatrix()`.
6495: See `MatZeroRows()` for details on how this routine operates.
6497: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRows()`, `MatSetOption()`,
6498: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6499: @*/
6500: PetscErrorCode MatZeroRowsLocal(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6501: {
6502: PetscFunctionBegin;
6505: if (numRows) PetscAssertPointer(rows, 3);
6506: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6507: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6508: MatCheckPreallocated(mat, 1);
6510: if (mat->ops->zerorowslocal) {
6511: PetscUseTypeMethod(mat, zerorowslocal, numRows, rows, diag, x, b);
6512: } else {
6513: IS is, newis;
6514: const PetscInt *newRows;
6516: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Need to provide local to global mapping to matrix first");
6517: PetscCall(ISCreateGeneral(PETSC_COMM_SELF, numRows, rows, PETSC_COPY_VALUES, &is));
6518: PetscCall(ISLocalToGlobalMappingApplyIS(mat->rmap->mapping, is, &newis));
6519: PetscCall(ISGetIndices(newis, &newRows));
6520: PetscUseTypeMethod(mat, zerorows, numRows, newRows, diag, x, b);
6521: PetscCall(ISRestoreIndices(newis, &newRows));
6522: PetscCall(ISDestroy(&newis));
6523: PetscCall(ISDestroy(&is));
6524: }
6525: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6526: PetscFunctionReturn(PETSC_SUCCESS);
6527: }
6529: /*@
6530: MatZeroRowsLocalIS - Zeros all entries (except possibly the main diagonal)
6531: of a set of rows of a matrix; using local numbering of rows.
6533: Collective
6535: Input Parameters:
6536: + mat - the matrix
6537: . is - index set of rows to remove
6538: . diag - value put in all diagonals of eliminated rows
6539: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6540: - b - optional vector of right hand side, that will be adjusted by provided solution
6542: Level: intermediate
6544: Notes:
6545: Before calling `MatZeroRowsLocalIS()`, the user must first set the
6546: local-to-global mapping by calling `MatSetLocalToGlobalMapping()`, this is often already set for matrices obtained with `DMCreateMatrix()`.
6548: See `MatZeroRows()` for details on how this routine operates.
6550: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRows()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6551: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6552: @*/
6553: PetscErrorCode MatZeroRowsLocalIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6554: {
6555: PetscInt numRows;
6556: const PetscInt *rows;
6558: PetscFunctionBegin;
6562: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6563: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6564: MatCheckPreallocated(mat, 1);
6566: PetscCall(ISGetLocalSize(is, &numRows));
6567: PetscCall(ISGetIndices(is, &rows));
6568: PetscCall(MatZeroRowsLocal(mat, numRows, rows, diag, x, b));
6569: PetscCall(ISRestoreIndices(is, &rows));
6570: PetscFunctionReturn(PETSC_SUCCESS);
6571: }
6573: /*@
6574: MatZeroRowsColumnsLocal - Zeros all entries (except possibly the main diagonal)
6575: of a set of rows and columns of a matrix; using local numbering of rows.
6577: Collective
6579: Input Parameters:
6580: + mat - the matrix
6581: . numRows - the number of rows to remove
6582: . rows - the global row indices
6583: . diag - value put in all diagonals of eliminated rows
6584: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6585: - b - optional vector of right hand side, that will be adjusted by provided solution
6587: Level: intermediate
6589: Notes:
6590: Before calling `MatZeroRowsColumnsLocal()`, the user must first set the
6591: local-to-global mapping by calling `MatSetLocalToGlobalMapping()`, this is often already set for matrices obtained with `DMCreateMatrix()`.
6593: See `MatZeroRowsColumns()` for details on how this routine operates.
6595: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6596: `MatZeroRows()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6597: @*/
6598: PetscErrorCode MatZeroRowsColumnsLocal(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6599: {
6600: IS is, newis;
6601: const PetscInt *newRows;
6603: PetscFunctionBegin;
6606: if (numRows) PetscAssertPointer(rows, 3);
6607: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6608: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6609: MatCheckPreallocated(mat, 1);
6611: PetscCheck(mat->cmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Need to provide local to global mapping to matrix first");
6612: PetscCall(ISCreateGeneral(PETSC_COMM_SELF, numRows, rows, PETSC_COPY_VALUES, &is));
6613: PetscCall(ISLocalToGlobalMappingApplyIS(mat->cmap->mapping, is, &newis));
6614: PetscCall(ISGetIndices(newis, &newRows));
6615: PetscUseTypeMethod(mat, zerorowscolumns, numRows, newRows, diag, x, b);
6616: PetscCall(ISRestoreIndices(newis, &newRows));
6617: PetscCall(ISDestroy(&newis));
6618: PetscCall(ISDestroy(&is));
6619: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6620: PetscFunctionReturn(PETSC_SUCCESS);
6621: }
6623: /*@
6624: MatZeroRowsColumnsLocalIS - Zeros all entries (except possibly the main diagonal)
6625: of a set of rows and columns of a matrix; using local numbering of rows.
6627: Collective
6629: Input Parameters:
6630: + mat - the matrix
6631: . is - index set of rows to remove
6632: . diag - value put in all diagonals of eliminated rows
6633: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6634: - b - optional vector of right hand side, that will be adjusted by provided solution
6636: Level: intermediate
6638: Notes:
6639: Before calling `MatZeroRowsColumnsLocalIS()`, the user must first set the
6640: local-to-global mapping by calling `MatSetLocalToGlobalMapping()`, this is often already set for matrices obtained with `DMCreateMatrix()`.
6642: See `MatZeroRowsColumns()` for details on how this routine operates.
6644: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6645: `MatZeroRowsColumnsLocal()`, `MatZeroRows()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6646: @*/
6647: PetscErrorCode MatZeroRowsColumnsLocalIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6648: {
6649: PetscInt numRows;
6650: const PetscInt *rows;
6652: PetscFunctionBegin;
6656: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6657: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6658: MatCheckPreallocated(mat, 1);
6660: PetscCall(ISGetLocalSize(is, &numRows));
6661: PetscCall(ISGetIndices(is, &rows));
6662: PetscCall(MatZeroRowsColumnsLocal(mat, numRows, rows, diag, x, b));
6663: PetscCall(ISRestoreIndices(is, &rows));
6664: PetscFunctionReturn(PETSC_SUCCESS);
6665: }
6667: /*@C
6668: MatGetSize - Returns the numbers of rows and columns in a matrix.
6670: Not Collective
6672: Input Parameter:
6673: . mat - the matrix
6675: Output Parameters:
6676: + m - the number of global rows
6677: - n - the number of global columns
6679: Level: beginner
6681: Note:
6682: Both output parameters can be `NULL` on input.
6684: .seealso: [](ch_matrices), `Mat`, `MatSetSizes()`, `MatGetLocalSize()`
6685: @*/
6686: PetscErrorCode MatGetSize(Mat mat, PetscInt *m, PetscInt *n)
6687: {
6688: PetscFunctionBegin;
6690: if (m) *m = mat->rmap->N;
6691: if (n) *n = mat->cmap->N;
6692: PetscFunctionReturn(PETSC_SUCCESS);
6693: }
6695: /*@C
6696: MatGetLocalSize - For most matrix formats, excluding `MATELEMENTAL` and `MATSCALAPACK`, Returns the number of local rows and local columns
6697: of a matrix. For all matrices this is the local size of the left and right vectors as returned by `MatCreateVecs()`.
6699: Not Collective
6701: Input Parameter:
6702: . mat - the matrix
6704: Output Parameters:
6705: + m - the number of local rows, use `NULL` to not obtain this value
6706: - n - the number of local columns, use `NULL` to not obtain this value
6708: Level: beginner
6710: .seealso: [](ch_matrices), `Mat`, `MatSetSizes()`, `MatGetSize()`
6711: @*/
6712: PetscErrorCode MatGetLocalSize(Mat mat, PetscInt *m, PetscInt *n)
6713: {
6714: PetscFunctionBegin;
6716: if (m) PetscAssertPointer(m, 2);
6717: if (n) PetscAssertPointer(n, 3);
6718: if (m) *m = mat->rmap->n;
6719: if (n) *n = mat->cmap->n;
6720: PetscFunctionReturn(PETSC_SUCCESS);
6721: }
6723: /*@C
6724: MatGetOwnershipRangeColumn - Returns the range of matrix columns associated with rows of a
6725: vector one multiplies this matrix by that are owned by this processor.
6727: Not Collective, unless matrix has not been allocated, then collective
6729: Input Parameter:
6730: . mat - the matrix
6732: Output Parameters:
6733: + m - the global index of the first local column, use `NULL` to not obtain this value
6734: - n - one more than the global index of the last local column, use `NULL` to not obtain this value
6736: Level: developer
6738: Note:
6739: Returns the columns of the "diagonal block" for most sparse matrix formats. See [Matrix
6740: Layouts](sec_matlayout) for details on matrix layouts.
6742: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRanges()`, `MatGetOwnershipRangesColumn()`, `PetscLayout`
6743: @*/
6744: PetscErrorCode MatGetOwnershipRangeColumn(Mat mat, PetscInt *m, PetscInt *n)
6745: {
6746: PetscFunctionBegin;
6749: if (m) PetscAssertPointer(m, 2);
6750: if (n) PetscAssertPointer(n, 3);
6751: MatCheckPreallocated(mat, 1);
6752: if (m) *m = mat->cmap->rstart;
6753: if (n) *n = mat->cmap->rend;
6754: PetscFunctionReturn(PETSC_SUCCESS);
6755: }
6757: /*@C
6758: MatGetOwnershipRange - For matrices that own values by row, excludes `MATELEMENTAL` and `MATSCALAPACK`, returns the range of matrix rows owned by
6759: this MPI process.
6761: Not Collective
6763: Input Parameter:
6764: . mat - the matrix
6766: Output Parameters:
6767: + m - the global index of the first local row, use `NULL` to not obtain this value
6768: - n - one more than the global index of the last local row, use `NULL` to not obtain this value
6770: Level: beginner
6772: Note:
6773: For all matrices it returns the range of matrix rows associated with rows of a vector that
6774: would contain the result of a matrix vector product with this matrix. See [Matrix
6775: Layouts](sec_matlayout) for details on matrix layouts.
6777: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRanges()`, `MatGetOwnershipRangeColumn()`, `MatGetOwnershipRangesColumn()`, `PetscSplitOwnership()`, `PetscSplitOwnershipBlock()`,
6778: `PetscLayout`
6779: @*/
6780: PetscErrorCode MatGetOwnershipRange(Mat mat, PetscInt *m, PetscInt *n)
6781: {
6782: PetscFunctionBegin;
6785: if (m) PetscAssertPointer(m, 2);
6786: if (n) PetscAssertPointer(n, 3);
6787: MatCheckPreallocated(mat, 1);
6788: if (m) *m = mat->rmap->rstart;
6789: if (n) *n = mat->rmap->rend;
6790: PetscFunctionReturn(PETSC_SUCCESS);
6791: }
6793: /*@C
6794: MatGetOwnershipRanges - For matrices that own values by row, excludes `MATELEMENTAL` and
6795: `MATSCALAPACK`, returns the range of matrix rows owned by each process.
6797: Not Collective, unless matrix has not been allocated
6799: Input Parameter:
6800: . mat - the matrix
6802: Output Parameter:
6803: . ranges - start of each processors portion plus one more than the total length at the end
6805: Level: beginner
6807: Note:
6808: For all matrices it returns the ranges of matrix rows associated with rows of a vector that
6809: would contain the result of a matrix vector product with this matrix. See [Matrix
6810: Layouts](sec_matlayout) for details on matrix layouts.
6812: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`, `MatGetOwnershipRangesColumn()`, `PetscLayout`
6813: @*/
6814: PetscErrorCode MatGetOwnershipRanges(Mat mat, const PetscInt **ranges)
6815: {
6816: PetscFunctionBegin;
6819: MatCheckPreallocated(mat, 1);
6820: PetscCall(PetscLayoutGetRanges(mat->rmap, ranges));
6821: PetscFunctionReturn(PETSC_SUCCESS);
6822: }
6824: /*@C
6825: MatGetOwnershipRangesColumn - Returns the ranges of matrix columns associated with rows of a
6826: vector one multiplies this vector by that are owned by each processor.
6828: Not Collective, unless matrix has not been allocated
6830: Input Parameter:
6831: . mat - the matrix
6833: Output Parameter:
6834: . ranges - start of each processors portion plus one more than the total length at the end
6836: Level: beginner
6838: Note:
6839: Returns the columns of the "diagonal blocks", for most sparse matrix formats. See [Matrix
6840: Layouts](sec_matlayout) for details on matrix layouts.
6842: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`, `MatGetOwnershipRanges()`
6843: @*/
6844: PetscErrorCode MatGetOwnershipRangesColumn(Mat mat, const PetscInt **ranges)
6845: {
6846: PetscFunctionBegin;
6849: MatCheckPreallocated(mat, 1);
6850: PetscCall(PetscLayoutGetRanges(mat->cmap, ranges));
6851: PetscFunctionReturn(PETSC_SUCCESS);
6852: }
6854: /*@C
6855: MatGetOwnershipIS - Get row and column ownership of a matrices' values as index sets.
6857: Not Collective
6859: Input Parameter:
6860: . A - matrix
6862: Output Parameters:
6863: + rows - rows in which this process owns elements, , use `NULL` to not obtain this value
6864: - cols - columns in which this process owns elements, use `NULL` to not obtain this value
6866: Level: intermediate
6868: Note:
6869: For most matrices, excluding `MATELEMENTAL` and `MATSCALAPACK`, this corresponds to values
6870: returned by `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`. For `MATELEMENTAL` and
6871: `MATSCALAPACK` the ownership is more complicated. See [Matrix Layouts](sec_matlayout) for
6872: details on matrix layouts.
6874: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`, `MatSetValues()`, ``MATELEMENTAL``, ``MATSCALAPACK``
6875: @*/
6876: PetscErrorCode MatGetOwnershipIS(Mat A, IS *rows, IS *cols)
6877: {
6878: PetscErrorCode (*f)(Mat, IS *, IS *);
6880: PetscFunctionBegin;
6881: MatCheckPreallocated(A, 1);
6882: PetscCall(PetscObjectQueryFunction((PetscObject)A, "MatGetOwnershipIS_C", &f));
6883: if (f) {
6884: PetscCall((*f)(A, rows, cols));
6885: } else { /* Create a standard row-based partition, each process is responsible for ALL columns in their row block */
6886: if (rows) PetscCall(ISCreateStride(PETSC_COMM_SELF, A->rmap->n, A->rmap->rstart, 1, rows));
6887: if (cols) PetscCall(ISCreateStride(PETSC_COMM_SELF, A->cmap->N, 0, 1, cols));
6888: }
6889: PetscFunctionReturn(PETSC_SUCCESS);
6890: }
6892: /*@C
6893: MatILUFactorSymbolic - Performs symbolic ILU factorization of a matrix obtained with `MatGetFactor()`
6894: Uses levels of fill only, not drop tolerance. Use `MatLUFactorNumeric()`
6895: to complete the factorization.
6897: Collective
6899: Input Parameters:
6900: + fact - the factorized matrix obtained with `MatGetFactor()`
6901: . mat - the matrix
6902: . row - row permutation
6903: . col - column permutation
6904: - info - structure containing
6905: .vb
6906: levels - number of levels of fill.
6907: expected fill - as ratio of original fill.
6908: 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
6909: missing diagonal entries)
6910: .ve
6912: Level: developer
6914: Notes:
6915: See [Matrix Factorization](sec_matfactor) for additional information.
6917: Most users should employ the `KSP` interface for linear solvers
6918: instead of working directly with matrix algebra routines such as this.
6919: See, e.g., `KSPCreate()`.
6921: Uses the definition of level of fill as in Y. Saad, {cite}`saad2003`
6923: Developer Note:
6924: The Fortran interface is not autogenerated as the
6925: interface definition cannot be generated correctly [due to `MatFactorInfo`]
6927: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`
6928: `MatGetOrdering()`, `MatFactorInfo`
6929: @*/
6930: PetscErrorCode MatILUFactorSymbolic(Mat fact, Mat mat, IS row, IS col, const MatFactorInfo *info)
6931: {
6932: PetscFunctionBegin;
6937: PetscAssertPointer(info, 5);
6938: PetscAssertPointer(fact, 1);
6939: PetscCheck(info->levels >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Levels of fill negative %" PetscInt_FMT, (PetscInt)info->levels);
6940: PetscCheck(info->fill >= 1.0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Expected fill less than 1.0 %g", (double)info->fill);
6941: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6942: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6943: MatCheckPreallocated(mat, 2);
6945: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_ILUFactorSymbolic, mat, row, col, 0));
6946: PetscUseTypeMethod(fact, ilufactorsymbolic, mat, row, col, info);
6947: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_ILUFactorSymbolic, mat, row, col, 0));
6948: PetscFunctionReturn(PETSC_SUCCESS);
6949: }
6951: /*@C
6952: MatICCFactorSymbolic - Performs symbolic incomplete
6953: Cholesky factorization for a symmetric matrix. Use
6954: `MatCholeskyFactorNumeric()` to complete the factorization.
6956: Collective
6958: Input Parameters:
6959: + fact - the factorized matrix obtained with `MatGetFactor()`
6960: . mat - the matrix to be factored
6961: . perm - row and column permutation
6962: - info - structure containing
6963: .vb
6964: levels - number of levels of fill.
6965: expected fill - as ratio of original fill.
6966: .ve
6968: Level: developer
6970: Notes:
6971: Most users should employ the `KSP` interface for linear solvers
6972: instead of working directly with matrix algebra routines such as this.
6973: See, e.g., `KSPCreate()`.
6975: This uses the definition of level of fill as in Y. Saad {cite}`saad2003`
6977: Developer Note:
6978: The Fortran interface is not autogenerated as the
6979: interface definition cannot be generated correctly [due to `MatFactorInfo`]
6981: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatCholeskyFactorNumeric()`, `MatCholeskyFactor()`, `MatFactorInfo`
6982: @*/
6983: PetscErrorCode MatICCFactorSymbolic(Mat fact, Mat mat, IS perm, const MatFactorInfo *info)
6984: {
6985: PetscFunctionBegin;
6989: PetscAssertPointer(info, 4);
6990: PetscAssertPointer(fact, 1);
6991: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6992: PetscCheck(info->levels >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Levels negative %" PetscInt_FMT, (PetscInt)info->levels);
6993: PetscCheck(info->fill >= 1.0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Expected fill less than 1.0 %g", (double)info->fill);
6994: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6995: MatCheckPreallocated(mat, 2);
6997: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_ICCFactorSymbolic, mat, perm, 0, 0));
6998: PetscUseTypeMethod(fact, iccfactorsymbolic, mat, perm, info);
6999: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_ICCFactorSymbolic, mat, perm, 0, 0));
7000: PetscFunctionReturn(PETSC_SUCCESS);
7001: }
7003: /*@C
7004: MatCreateSubMatrices - Extracts several submatrices from a matrix. If submat
7005: points to an array of valid matrices, they may be reused to store the new
7006: submatrices.
7008: Collective
7010: Input Parameters:
7011: + mat - the matrix
7012: . n - the number of submatrixes to be extracted (on this processor, may be zero)
7013: . irow - index set of rows to extract
7014: . icol - index set of columns to extract
7015: - scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
7017: Output Parameter:
7018: . submat - the array of submatrices
7020: Level: advanced
7022: Notes:
7023: `MatCreateSubMatrices()` can extract ONLY sequential submatrices
7024: (from both sequential and parallel matrices). Use `MatCreateSubMatrix()`
7025: to extract a parallel submatrix.
7027: Some matrix types place restrictions on the row and column
7028: indices, such as that they be sorted or that they be equal to each other.
7030: The index sets may not have duplicate entries.
7032: When extracting submatrices from a parallel matrix, each processor can
7033: form a different submatrix by setting the rows and columns of its
7034: individual index sets according to the local submatrix desired.
7036: When finished using the submatrices, the user should destroy
7037: them with `MatDestroySubMatrices()`.
7039: `MAT_REUSE_MATRIX` can only be used when the nonzero structure of the
7040: original matrix has not changed from that last call to `MatCreateSubMatrices()`.
7042: This routine creates the matrices in submat; you should NOT create them before
7043: calling it. It also allocates the array of matrix pointers submat.
7045: For `MATBAIJ` matrices the index sets must respect the block structure, that is if they
7046: request one row/column in a block, they must request all rows/columns that are in
7047: that block. For example, if the block size is 2 you cannot request just row 0 and
7048: column 0.
7050: Fortran Note:
7051: The Fortran interface is slightly different from that given below; it
7052: requires one to pass in as `submat` a `Mat` (integer) array of size at least n+1.
7054: .seealso: [](ch_matrices), `Mat`, `MatDestroySubMatrices()`, `MatCreateSubMatrix()`, `MatGetRow()`, `MatGetDiagonal()`, `MatReuse`
7055: @*/
7056: PetscErrorCode MatCreateSubMatrices(Mat mat, PetscInt n, const IS irow[], const IS icol[], MatReuse scall, Mat *submat[])
7057: {
7058: PetscInt i;
7059: PetscBool eq;
7061: PetscFunctionBegin;
7064: if (n) {
7065: PetscAssertPointer(irow, 3);
7067: PetscAssertPointer(icol, 4);
7069: }
7070: PetscAssertPointer(submat, 6);
7071: if (n && scall == MAT_REUSE_MATRIX) {
7072: PetscAssertPointer(*submat, 6);
7074: }
7075: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7076: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7077: MatCheckPreallocated(mat, 1);
7078: PetscCall(PetscLogEventBegin(MAT_CreateSubMats, mat, 0, 0, 0));
7079: PetscUseTypeMethod(mat, createsubmatrices, n, irow, icol, scall, submat);
7080: PetscCall(PetscLogEventEnd(MAT_CreateSubMats, mat, 0, 0, 0));
7081: for (i = 0; i < n; i++) {
7082: (*submat)[i]->factortype = MAT_FACTOR_NONE; /* in case in place factorization was previously done on submatrix */
7083: PetscCall(ISEqualUnsorted(irow[i], icol[i], &eq));
7084: if (eq) PetscCall(MatPropagateSymmetryOptions(mat, (*submat)[i]));
7085: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
7086: if (mat->boundtocpu && mat->bindingpropagates) {
7087: PetscCall(MatBindToCPU((*submat)[i], PETSC_TRUE));
7088: PetscCall(MatSetBindingPropagates((*submat)[i], PETSC_TRUE));
7089: }
7090: #endif
7091: }
7092: PetscFunctionReturn(PETSC_SUCCESS);
7093: }
7095: /*@C
7096: MatCreateSubMatricesMPI - Extracts MPI submatrices across a sub communicator of mat (by pairs of `IS` that may live on subcomms).
7098: Collective
7100: Input Parameters:
7101: + mat - the matrix
7102: . n - the number of submatrixes to be extracted
7103: . irow - index set of rows to extract
7104: . icol - index set of columns to extract
7105: - scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
7107: Output Parameter:
7108: . submat - the array of submatrices
7110: Level: advanced
7112: Note:
7113: This is used by `PCGASM`
7115: .seealso: [](ch_matrices), `Mat`, `PCGASM`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRow()`, `MatGetDiagonal()`, `MatReuse`
7116: @*/
7117: PetscErrorCode MatCreateSubMatricesMPI(Mat mat, PetscInt n, const IS irow[], const IS icol[], MatReuse scall, Mat *submat[])
7118: {
7119: PetscInt i;
7120: PetscBool eq;
7122: PetscFunctionBegin;
7125: if (n) {
7126: PetscAssertPointer(irow, 3);
7128: PetscAssertPointer(icol, 4);
7130: }
7131: PetscAssertPointer(submat, 6);
7132: if (n && scall == MAT_REUSE_MATRIX) {
7133: PetscAssertPointer(*submat, 6);
7135: }
7136: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7137: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7138: MatCheckPreallocated(mat, 1);
7140: PetscCall(PetscLogEventBegin(MAT_CreateSubMats, mat, 0, 0, 0));
7141: PetscUseTypeMethod(mat, createsubmatricesmpi, n, irow, icol, scall, submat);
7142: PetscCall(PetscLogEventEnd(MAT_CreateSubMats, mat, 0, 0, 0));
7143: for (i = 0; i < n; i++) {
7144: PetscCall(ISEqualUnsorted(irow[i], icol[i], &eq));
7145: if (eq) PetscCall(MatPropagateSymmetryOptions(mat, (*submat)[i]));
7146: }
7147: PetscFunctionReturn(PETSC_SUCCESS);
7148: }
7150: /*@C
7151: MatDestroyMatrices - Destroys an array of matrices.
7153: Collective
7155: Input Parameters:
7156: + n - the number of local matrices
7157: - mat - the matrices (this is a pointer to the array of matrices)
7159: Level: advanced
7161: Note:
7162: Frees not only the matrices, but also the array that contains the matrices
7164: Fortran Note:
7165: This does not free the array.
7167: .seealso: [](ch_matrices), `Mat`, `MatCreateSubMatrices()` `MatDestroySubMatrices()`
7168: @*/
7169: PetscErrorCode MatDestroyMatrices(PetscInt n, Mat *mat[])
7170: {
7171: PetscInt i;
7173: PetscFunctionBegin;
7174: if (!*mat) PetscFunctionReturn(PETSC_SUCCESS);
7175: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Trying to destroy negative number of matrices %" PetscInt_FMT, n);
7176: PetscAssertPointer(mat, 2);
7178: for (i = 0; i < n; i++) PetscCall(MatDestroy(&(*mat)[i]));
7180: /* memory is allocated even if n = 0 */
7181: PetscCall(PetscFree(*mat));
7182: PetscFunctionReturn(PETSC_SUCCESS);
7183: }
7185: /*@C
7186: MatDestroySubMatrices - Destroys a set of matrices obtained with `MatCreateSubMatrices()`.
7188: Collective
7190: Input Parameters:
7191: + n - the number of local matrices
7192: - mat - the matrices (this is a pointer to the array of matrices, just to match the calling
7193: sequence of `MatCreateSubMatrices()`)
7195: Level: advanced
7197: Note:
7198: Frees not only the matrices, but also the array that contains the matrices
7200: Fortran Note:
7201: This does not free the array.
7203: .seealso: [](ch_matrices), `Mat`, `MatCreateSubMatrices()`, `MatDestroyMatrices()`
7204: @*/
7205: PetscErrorCode MatDestroySubMatrices(PetscInt n, Mat *mat[])
7206: {
7207: Mat mat0;
7209: PetscFunctionBegin;
7210: if (!*mat) PetscFunctionReturn(PETSC_SUCCESS);
7211: /* mat[] is an array of length n+1, see MatCreateSubMatrices_xxx() */
7212: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Trying to destroy negative number of matrices %" PetscInt_FMT, n);
7213: PetscAssertPointer(mat, 2);
7215: mat0 = (*mat)[0];
7216: if (mat0 && mat0->ops->destroysubmatrices) {
7217: PetscCall((*mat0->ops->destroysubmatrices)(n, mat));
7218: } else {
7219: PetscCall(MatDestroyMatrices(n, mat));
7220: }
7221: PetscFunctionReturn(PETSC_SUCCESS);
7222: }
7224: /*@C
7225: MatGetSeqNonzeroStructure - Extracts the nonzero structure from a matrix and stores it, in its entirety, on each process
7227: Collective
7229: Input Parameter:
7230: . mat - the matrix
7232: Output Parameter:
7233: . matstruct - the sequential matrix with the nonzero structure of mat
7235: Level: developer
7237: .seealso: [](ch_matrices), `Mat`, `MatDestroySeqNonzeroStructure()`, `MatCreateSubMatrices()`, `MatDestroyMatrices()`
7238: @*/
7239: PetscErrorCode MatGetSeqNonzeroStructure(Mat mat, Mat *matstruct)
7240: {
7241: PetscFunctionBegin;
7243: PetscAssertPointer(matstruct, 2);
7246: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7247: MatCheckPreallocated(mat, 1);
7249: PetscCall(PetscLogEventBegin(MAT_GetSeqNonzeroStructure, mat, 0, 0, 0));
7250: PetscUseTypeMethod(mat, getseqnonzerostructure, matstruct);
7251: PetscCall(PetscLogEventEnd(MAT_GetSeqNonzeroStructure, mat, 0, 0, 0));
7252: PetscFunctionReturn(PETSC_SUCCESS);
7253: }
7255: /*@C
7256: MatDestroySeqNonzeroStructure - Destroys matrix obtained with `MatGetSeqNonzeroStructure()`.
7258: Collective
7260: Input Parameter:
7261: . mat - the matrix (this is a pointer to the array of matrices, just to match the calling
7262: sequence of `MatGetSeqNonzeroStructure()`)
7264: Level: advanced
7266: Note:
7267: Frees not only the matrices, but also the array that contains the matrices
7269: .seealso: [](ch_matrices), `Mat`, `MatGetSeqNonzeroStructure()`
7270: @*/
7271: PetscErrorCode MatDestroySeqNonzeroStructure(Mat *mat)
7272: {
7273: PetscFunctionBegin;
7274: PetscAssertPointer(mat, 1);
7275: PetscCall(MatDestroy(mat));
7276: PetscFunctionReturn(PETSC_SUCCESS);
7277: }
7279: /*@
7280: MatIncreaseOverlap - Given a set of submatrices indicated by index sets,
7281: replaces the index sets by larger ones that represent submatrices with
7282: additional overlap.
7284: Collective
7286: Input Parameters:
7287: + mat - the matrix
7288: . n - the number of index sets
7289: . is - the array of index sets (these index sets will changed during the call)
7290: - ov - the additional overlap requested
7292: Options Database Key:
7293: . -mat_increase_overlap_scalable - use a scalable algorithm to compute the overlap (supported by MPIAIJ matrix)
7295: Level: developer
7297: Note:
7298: The computed overlap preserves the matrix block sizes when the blocks are square.
7299: That is: if a matrix nonzero for a given block would increase the overlap all columns associated with
7300: that block are included in the overlap regardless of whether each specific column would increase the overlap.
7302: .seealso: [](ch_matrices), `Mat`, `PCASM`, `MatSetBlockSize()`, `MatIncreaseOverlapSplit()`, `MatCreateSubMatrices()`
7303: @*/
7304: PetscErrorCode MatIncreaseOverlap(Mat mat, PetscInt n, IS is[], PetscInt ov)
7305: {
7306: PetscInt i, bs, cbs;
7308: PetscFunctionBegin;
7312: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Must have one or more domains, you have %" PetscInt_FMT, n);
7313: if (n) {
7314: PetscAssertPointer(is, 3);
7316: }
7317: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7318: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7319: MatCheckPreallocated(mat, 1);
7321: if (!ov || !n) PetscFunctionReturn(PETSC_SUCCESS);
7322: PetscCall(PetscLogEventBegin(MAT_IncreaseOverlap, mat, 0, 0, 0));
7323: PetscUseTypeMethod(mat, increaseoverlap, n, is, ov);
7324: PetscCall(PetscLogEventEnd(MAT_IncreaseOverlap, mat, 0, 0, 0));
7325: PetscCall(MatGetBlockSizes(mat, &bs, &cbs));
7326: if (bs == cbs) {
7327: for (i = 0; i < n; i++) PetscCall(ISSetBlockSize(is[i], bs));
7328: }
7329: PetscFunctionReturn(PETSC_SUCCESS);
7330: }
7332: PetscErrorCode MatIncreaseOverlapSplit_Single(Mat, IS *, PetscInt);
7334: /*@
7335: MatIncreaseOverlapSplit - Given a set of submatrices indicated by index sets across
7336: a sub communicator, replaces the index sets by larger ones that represent submatrices with
7337: additional overlap.
7339: Collective
7341: Input Parameters:
7342: + mat - the matrix
7343: . n - the number of index sets
7344: . is - the array of index sets (these index sets will changed during the call)
7345: - ov - the additional overlap requested
7347: ` Options Database Key:
7348: . -mat_increase_overlap_scalable - use a scalable algorithm to compute the overlap (supported by MPIAIJ matrix)
7350: Level: developer
7352: .seealso: [](ch_matrices), `Mat`, `MatCreateSubMatrices()`, `MatIncreaseOverlap()`
7353: @*/
7354: PetscErrorCode MatIncreaseOverlapSplit(Mat mat, PetscInt n, IS is[], PetscInt ov)
7355: {
7356: PetscInt i;
7358: PetscFunctionBegin;
7361: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Must have one or more domains, you have %" PetscInt_FMT, n);
7362: if (n) {
7363: PetscAssertPointer(is, 3);
7365: }
7366: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7367: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7368: MatCheckPreallocated(mat, 1);
7369: if (!ov) PetscFunctionReturn(PETSC_SUCCESS);
7370: PetscCall(PetscLogEventBegin(MAT_IncreaseOverlap, mat, 0, 0, 0));
7371: for (i = 0; i < n; i++) PetscCall(MatIncreaseOverlapSplit_Single(mat, &is[i], ov));
7372: PetscCall(PetscLogEventEnd(MAT_IncreaseOverlap, mat, 0, 0, 0));
7373: PetscFunctionReturn(PETSC_SUCCESS);
7374: }
7376: /*@
7377: MatGetBlockSize - Returns the matrix block size.
7379: Not Collective
7381: Input Parameter:
7382: . mat - the matrix
7384: Output Parameter:
7385: . bs - block size
7387: Level: intermediate
7389: Notes:
7390: Block row formats are `MATBAIJ` and `MATSBAIJ` ALWAYS have square block storage in the matrix.
7392: If the block size has not been set yet this routine returns 1.
7394: .seealso: [](ch_matrices), `Mat`, `MATBAIJ`, `MATSBAIJ`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSizes()`
7395: @*/
7396: PetscErrorCode MatGetBlockSize(Mat mat, PetscInt *bs)
7397: {
7398: PetscFunctionBegin;
7400: PetscAssertPointer(bs, 2);
7401: *bs = PetscAbs(mat->rmap->bs);
7402: PetscFunctionReturn(PETSC_SUCCESS);
7403: }
7405: /*@
7406: MatGetBlockSizes - Returns the matrix block row and column sizes.
7408: Not Collective
7410: Input Parameter:
7411: . mat - the matrix
7413: Output Parameters:
7414: + rbs - row block size
7415: - cbs - column block size
7417: Level: intermediate
7419: Notes:
7420: Block row formats are `MATBAIJ` and `MATSBAIJ` ALWAYS have square block storage in the matrix.
7421: If you pass a different block size for the columns than the rows, the row block size determines the square block storage.
7423: If a block size has not been set yet this routine returns 1.
7425: .seealso: [](ch_matrices), `Mat`, `MATBAIJ`, `MATSBAIJ`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSize()`, `MatSetBlockSizes()`
7426: @*/
7427: PetscErrorCode MatGetBlockSizes(Mat mat, PetscInt *rbs, PetscInt *cbs)
7428: {
7429: PetscFunctionBegin;
7431: if (rbs) PetscAssertPointer(rbs, 2);
7432: if (cbs) PetscAssertPointer(cbs, 3);
7433: if (rbs) *rbs = PetscAbs(mat->rmap->bs);
7434: if (cbs) *cbs = PetscAbs(mat->cmap->bs);
7435: PetscFunctionReturn(PETSC_SUCCESS);
7436: }
7438: /*@
7439: MatSetBlockSize - Sets the matrix block size.
7441: Logically Collective
7443: Input Parameters:
7444: + mat - the matrix
7445: - bs - block size
7447: Level: intermediate
7449: Notes:
7450: Block row formats are `MATBAIJ` and `MATSBAIJ` formats ALWAYS have square block storage in the matrix.
7451: This must be called before `MatSetUp()` or MatXXXSetPreallocation() (or will default to 1) and the block size cannot be changed later.
7453: For `MATAIJ` matrix format, this function can be called at a later stage, provided that the specified block size
7454: is compatible with the matrix local sizes.
7456: .seealso: [](ch_matrices), `Mat`, `MATBAIJ`, `MATSBAIJ`, `MATAIJ`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`, `MatGetBlockSizes()`
7457: @*/
7458: PetscErrorCode MatSetBlockSize(Mat mat, PetscInt bs)
7459: {
7460: PetscFunctionBegin;
7463: PetscCall(MatSetBlockSizes(mat, bs, bs));
7464: PetscFunctionReturn(PETSC_SUCCESS);
7465: }
7467: typedef struct {
7468: PetscInt n;
7469: IS *is;
7470: Mat *mat;
7471: PetscObjectState nonzerostate;
7472: Mat C;
7473: } EnvelopeData;
7475: static PetscErrorCode EnvelopeDataDestroy(EnvelopeData *edata)
7476: {
7477: for (PetscInt i = 0; i < edata->n; i++) PetscCall(ISDestroy(&edata->is[i]));
7478: PetscCall(PetscFree(edata->is));
7479: PetscCall(PetscFree(edata));
7480: return PETSC_SUCCESS;
7481: }
7483: /*@
7484: MatComputeVariableBlockEnvelope - Given a matrix whose nonzeros are in blocks along the diagonal this computes and stores
7485: the sizes of these blocks in the matrix. An individual block may lie over several processes.
7487: Collective
7489: Input Parameter:
7490: . mat - the matrix
7492: Level: intermediate
7494: Notes:
7495: There can be zeros within the blocks
7497: The blocks can overlap between processes, including laying on more than two processes
7499: .seealso: [](ch_matrices), `Mat`, `MatInvertVariableBlockEnvelope()`, `MatSetVariableBlockSizes()`
7500: @*/
7501: PetscErrorCode MatComputeVariableBlockEnvelope(Mat mat)
7502: {
7503: PetscInt n, *sizes, *starts, i = 0, env = 0, tbs = 0, lblocks = 0, rstart, II, ln = 0, cnt = 0, cstart, cend;
7504: PetscInt *diag, *odiag, sc;
7505: VecScatter scatter;
7506: PetscScalar *seqv;
7507: const PetscScalar *parv;
7508: const PetscInt *ia, *ja;
7509: PetscBool set, flag, done;
7510: Mat AA = mat, A;
7511: MPI_Comm comm;
7512: PetscMPIInt rank, size, tag;
7513: MPI_Status status;
7514: PetscContainer container;
7515: EnvelopeData *edata;
7516: Vec seq, par;
7517: IS isglobal;
7519: PetscFunctionBegin;
7521: PetscCall(MatIsSymmetricKnown(mat, &set, &flag));
7522: if (!set || !flag) {
7523: /* TODO: only needs nonzero structure of transpose */
7524: PetscCall(MatTranspose(mat, MAT_INITIAL_MATRIX, &AA));
7525: PetscCall(MatAXPY(AA, 1.0, mat, DIFFERENT_NONZERO_PATTERN));
7526: }
7527: PetscCall(MatAIJGetLocalMat(AA, &A));
7528: PetscCall(MatGetRowIJ(A, 0, PETSC_FALSE, PETSC_FALSE, &n, &ia, &ja, &done));
7529: PetscCheck(done, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Unable to get IJ structure from matrix");
7531: PetscCall(MatGetLocalSize(mat, &n, NULL));
7532: PetscCall(PetscObjectGetNewTag((PetscObject)mat, &tag));
7533: PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
7534: PetscCallMPI(MPI_Comm_size(comm, &size));
7535: PetscCallMPI(MPI_Comm_rank(comm, &rank));
7537: PetscCall(PetscMalloc2(n, &sizes, n, &starts));
7539: if (rank > 0) {
7540: PetscCallMPI(MPI_Recv(&env, 1, MPIU_INT, rank - 1, tag, comm, &status));
7541: PetscCallMPI(MPI_Recv(&tbs, 1, MPIU_INT, rank - 1, tag, comm, &status));
7542: }
7543: PetscCall(MatGetOwnershipRange(mat, &rstart, NULL));
7544: for (i = 0; i < n; i++) {
7545: env = PetscMax(env, ja[ia[i + 1] - 1]);
7546: II = rstart + i;
7547: if (env == II) {
7548: starts[lblocks] = tbs;
7549: sizes[lblocks++] = 1 + II - tbs;
7550: tbs = 1 + II;
7551: }
7552: }
7553: if (rank < size - 1) {
7554: PetscCallMPI(MPI_Send(&env, 1, MPIU_INT, rank + 1, tag, comm));
7555: PetscCallMPI(MPI_Send(&tbs, 1, MPIU_INT, rank + 1, tag, comm));
7556: }
7558: PetscCall(MatRestoreRowIJ(A, 0, PETSC_FALSE, PETSC_FALSE, &n, &ia, &ja, &done));
7559: if (!set || !flag) PetscCall(MatDestroy(&AA));
7560: PetscCall(MatDestroy(&A));
7562: PetscCall(PetscNew(&edata));
7563: PetscCall(MatGetNonzeroState(mat, &edata->nonzerostate));
7564: edata->n = lblocks;
7565: /* create IS needed for extracting blocks from the original matrix */
7566: PetscCall(PetscMalloc1(lblocks, &edata->is));
7567: for (PetscInt i = 0; i < lblocks; i++) PetscCall(ISCreateStride(PETSC_COMM_SELF, sizes[i], starts[i], 1, &edata->is[i]));
7569: /* Create the resulting inverse matrix structure with preallocation information */
7570: PetscCall(MatCreate(PetscObjectComm((PetscObject)mat), &edata->C));
7571: PetscCall(MatSetSizes(edata->C, mat->rmap->n, mat->cmap->n, mat->rmap->N, mat->cmap->N));
7572: PetscCall(MatSetBlockSizesFromMats(edata->C, mat, mat));
7573: PetscCall(MatSetType(edata->C, MATAIJ));
7575: /* Communicate the start and end of each row, from each block to the correct rank */
7576: /* TODO: Use PetscSF instead of VecScatter */
7577: for (PetscInt i = 0; i < lblocks; i++) ln += sizes[i];
7578: PetscCall(VecCreateSeq(PETSC_COMM_SELF, 2 * ln, &seq));
7579: PetscCall(VecGetArrayWrite(seq, &seqv));
7580: for (PetscInt i = 0; i < lblocks; i++) {
7581: for (PetscInt j = 0; j < sizes[i]; j++) {
7582: seqv[cnt] = starts[i];
7583: seqv[cnt + 1] = starts[i] + sizes[i];
7584: cnt += 2;
7585: }
7586: }
7587: PetscCall(VecRestoreArrayWrite(seq, &seqv));
7588: PetscCallMPI(MPI_Scan(&cnt, &sc, 1, MPIU_INT, MPI_SUM, PetscObjectComm((PetscObject)mat)));
7589: sc -= cnt;
7590: PetscCall(VecCreateMPI(PetscObjectComm((PetscObject)mat), 2 * mat->rmap->n, 2 * mat->rmap->N, &par));
7591: PetscCall(ISCreateStride(PETSC_COMM_SELF, cnt, sc, 1, &isglobal));
7592: PetscCall(VecScatterCreate(seq, NULL, par, isglobal, &scatter));
7593: PetscCall(ISDestroy(&isglobal));
7594: PetscCall(VecScatterBegin(scatter, seq, par, INSERT_VALUES, SCATTER_FORWARD));
7595: PetscCall(VecScatterEnd(scatter, seq, par, INSERT_VALUES, SCATTER_FORWARD));
7596: PetscCall(VecScatterDestroy(&scatter));
7597: PetscCall(VecDestroy(&seq));
7598: PetscCall(MatGetOwnershipRangeColumn(mat, &cstart, &cend));
7599: PetscCall(PetscMalloc2(mat->rmap->n, &diag, mat->rmap->n, &odiag));
7600: PetscCall(VecGetArrayRead(par, &parv));
7601: cnt = 0;
7602: PetscCall(MatGetSize(mat, NULL, &n));
7603: for (PetscInt i = 0; i < mat->rmap->n; i++) {
7604: PetscInt start, end, d = 0, od = 0;
7606: start = (PetscInt)PetscRealPart(parv[cnt]);
7607: end = (PetscInt)PetscRealPart(parv[cnt + 1]);
7608: cnt += 2;
7610: if (start < cstart) {
7611: od += cstart - start + n - cend;
7612: d += cend - cstart;
7613: } else if (start < cend) {
7614: od += n - cend;
7615: d += cend - start;
7616: } else od += n - start;
7617: if (end <= cstart) {
7618: od -= cstart - end + n - cend;
7619: d -= cend - cstart;
7620: } else if (end < cend) {
7621: od -= n - cend;
7622: d -= cend - end;
7623: } else od -= n - end;
7625: odiag[i] = od;
7626: diag[i] = d;
7627: }
7628: PetscCall(VecRestoreArrayRead(par, &parv));
7629: PetscCall(VecDestroy(&par));
7630: PetscCall(MatXAIJSetPreallocation(edata->C, mat->rmap->bs, diag, odiag, NULL, NULL));
7631: PetscCall(PetscFree2(diag, odiag));
7632: PetscCall(PetscFree2(sizes, starts));
7634: PetscCall(PetscContainerCreate(PETSC_COMM_SELF, &container));
7635: PetscCall(PetscContainerSetPointer(container, edata));
7636: PetscCall(PetscContainerSetUserDestroy(container, (PetscErrorCode(*)(void *))EnvelopeDataDestroy));
7637: PetscCall(PetscObjectCompose((PetscObject)mat, "EnvelopeData", (PetscObject)container));
7638: PetscCall(PetscObjectDereference((PetscObject)container));
7639: PetscFunctionReturn(PETSC_SUCCESS);
7640: }
7642: /*@
7643: MatInvertVariableBlockEnvelope - set matrix C to be the inverted block diagonal of matrix A
7645: Collective
7647: Input Parameters:
7648: + A - the matrix
7649: - reuse - indicates if the `C` matrix was obtained from a previous call to this routine
7651: Output Parameter:
7652: . C - matrix with inverted block diagonal of `A`
7654: Level: advanced
7656: Note:
7657: For efficiency the matrix `A` should have all the nonzero entries clustered in smallish blocks along the diagonal.
7659: .seealso: [](ch_matrices), `Mat`, `MatInvertBlockDiagonal()`, `MatComputeBlockDiagonal()`
7660: @*/
7661: PetscErrorCode MatInvertVariableBlockEnvelope(Mat A, MatReuse reuse, Mat *C)
7662: {
7663: PetscContainer container;
7664: EnvelopeData *edata;
7665: PetscObjectState nonzerostate;
7667: PetscFunctionBegin;
7668: PetscCall(PetscObjectQuery((PetscObject)A, "EnvelopeData", (PetscObject *)&container));
7669: if (!container) {
7670: PetscCall(MatComputeVariableBlockEnvelope(A));
7671: PetscCall(PetscObjectQuery((PetscObject)A, "EnvelopeData", (PetscObject *)&container));
7672: }
7673: PetscCall(PetscContainerGetPointer(container, (void **)&edata));
7674: PetscCall(MatGetNonzeroState(A, &nonzerostate));
7675: PetscCheck(nonzerostate <= edata->nonzerostate, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Cannot handle changes to matrix nonzero structure");
7676: PetscCheck(reuse != MAT_REUSE_MATRIX || *C == edata->C, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "C matrix must be the same as previously output");
7678: PetscCall(MatCreateSubMatrices(A, edata->n, edata->is, edata->is, MAT_INITIAL_MATRIX, &edata->mat));
7679: *C = edata->C;
7681: for (PetscInt i = 0; i < edata->n; i++) {
7682: Mat D;
7683: PetscScalar *dvalues;
7685: PetscCall(MatConvert(edata->mat[i], MATSEQDENSE, MAT_INITIAL_MATRIX, &D));
7686: PetscCall(MatSetOption(*C, MAT_ROW_ORIENTED, PETSC_FALSE));
7687: PetscCall(MatSeqDenseInvert(D));
7688: PetscCall(MatDenseGetArray(D, &dvalues));
7689: PetscCall(MatSetValuesIS(*C, edata->is[i], edata->is[i], dvalues, INSERT_VALUES));
7690: PetscCall(MatDestroy(&D));
7691: }
7692: PetscCall(MatDestroySubMatrices(edata->n, &edata->mat));
7693: PetscCall(MatAssemblyBegin(*C, MAT_FINAL_ASSEMBLY));
7694: PetscCall(MatAssemblyEnd(*C, MAT_FINAL_ASSEMBLY));
7695: PetscFunctionReturn(PETSC_SUCCESS);
7696: }
7698: /*@
7699: MatSetVariableBlockSizes - Sets diagonal point-blocks of the matrix that need not be of the same size
7701: Logically Collective
7703: Input Parameters:
7704: + mat - the matrix
7705: . nblocks - the number of blocks on this process, each block can only exist on a single process
7706: - bsizes - the block sizes
7708: Level: intermediate
7710: Notes:
7711: Currently used by `PCVPBJACOBI` for `MATAIJ` matrices
7713: 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.
7715: .seealso: [](ch_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`, `MatGetBlockSizes()`, `MatGetVariableBlockSizes()`,
7716: `MatComputeVariableBlockEnvelope()`, `PCVPBJACOBI`
7717: @*/
7718: PetscErrorCode MatSetVariableBlockSizes(Mat mat, PetscInt nblocks, PetscInt *bsizes)
7719: {
7720: PetscInt i, ncnt = 0, nlocal;
7722: PetscFunctionBegin;
7724: PetscCheck(nblocks >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Number of local blocks must be great than or equal to zero");
7725: PetscCall(MatGetLocalSize(mat, &nlocal, NULL));
7726: for (i = 0; i < nblocks; i++) ncnt += bsizes[i];
7727: 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);
7728: PetscCall(PetscFree(mat->bsizes));
7729: mat->nblocks = nblocks;
7730: PetscCall(PetscMalloc1(nblocks, &mat->bsizes));
7731: PetscCall(PetscArraycpy(mat->bsizes, bsizes, nblocks));
7732: PetscFunctionReturn(PETSC_SUCCESS);
7733: }
7735: /*@C
7736: MatGetVariableBlockSizes - Gets a diagonal blocks of the matrix that need not be of the same size
7738: Logically Collective; No Fortran Support
7740: Input Parameter:
7741: . mat - the matrix
7743: Output Parameters:
7744: + nblocks - the number of blocks on this process
7745: - bsizes - the block sizes
7747: Level: intermediate
7749: .seealso: [](ch_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`, `MatGetBlockSizes()`, `MatSetVariableBlockSizes()`, `MatComputeVariableBlockEnvelope()`
7750: @*/
7751: PetscErrorCode MatGetVariableBlockSizes(Mat mat, PetscInt *nblocks, const PetscInt **bsizes)
7752: {
7753: PetscFunctionBegin;
7755: *nblocks = mat->nblocks;
7756: *bsizes = mat->bsizes;
7757: PetscFunctionReturn(PETSC_SUCCESS);
7758: }
7760: /*@
7761: MatSetBlockSizes - Sets the matrix block row and column sizes.
7763: Logically Collective
7765: Input Parameters:
7766: + mat - the matrix
7767: . rbs - row block size
7768: - cbs - column block size
7770: Level: intermediate
7772: Notes:
7773: Block row formats are `MATBAIJ` and `MATSBAIJ`. These formats ALWAYS have square block storage in the matrix.
7774: If you pass a different block size for the columns than the rows, the row block size determines the square block storage.
7775: This must be called before `MatSetUp()` or MatXXXSetPreallocation() (or will default to 1) and the block size cannot be changed later.
7777: For `MATAIJ` matrix this function can be called at a later stage, provided that the specified block sizes
7778: are compatible with the matrix local sizes.
7780: The row and column block size determine the blocksize of the "row" and "column" vectors returned by `MatCreateVecs()`.
7782: .seealso: [](ch_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSize()`, `MatGetBlockSizes()`
7783: @*/
7784: PetscErrorCode MatSetBlockSizes(Mat mat, PetscInt rbs, PetscInt cbs)
7785: {
7786: PetscFunctionBegin;
7790: PetscTryTypeMethod(mat, setblocksizes, rbs, cbs);
7791: if (mat->rmap->refcnt) {
7792: ISLocalToGlobalMapping l2g = NULL;
7793: PetscLayout nmap = NULL;
7795: PetscCall(PetscLayoutDuplicate(mat->rmap, &nmap));
7796: if (mat->rmap->mapping) PetscCall(ISLocalToGlobalMappingDuplicate(mat->rmap->mapping, &l2g));
7797: PetscCall(PetscLayoutDestroy(&mat->rmap));
7798: mat->rmap = nmap;
7799: mat->rmap->mapping = l2g;
7800: }
7801: if (mat->cmap->refcnt) {
7802: ISLocalToGlobalMapping l2g = NULL;
7803: PetscLayout nmap = NULL;
7805: PetscCall(PetscLayoutDuplicate(mat->cmap, &nmap));
7806: if (mat->cmap->mapping) PetscCall(ISLocalToGlobalMappingDuplicate(mat->cmap->mapping, &l2g));
7807: PetscCall(PetscLayoutDestroy(&mat->cmap));
7808: mat->cmap = nmap;
7809: mat->cmap->mapping = l2g;
7810: }
7811: PetscCall(PetscLayoutSetBlockSize(mat->rmap, rbs));
7812: PetscCall(PetscLayoutSetBlockSize(mat->cmap, cbs));
7813: PetscFunctionReturn(PETSC_SUCCESS);
7814: }
7816: /*@
7817: MatSetBlockSizesFromMats - Sets the matrix block row and column sizes to match a pair of matrices
7819: Logically Collective
7821: Input Parameters:
7822: + mat - the matrix
7823: . fromRow - matrix from which to copy row block size
7824: - fromCol - matrix from which to copy column block size (can be same as fromRow)
7826: Level: developer
7828: .seealso: [](ch_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`
7829: @*/
7830: PetscErrorCode MatSetBlockSizesFromMats(Mat mat, Mat fromRow, Mat fromCol)
7831: {
7832: PetscFunctionBegin;
7836: if (fromRow->rmap->bs > 0) PetscCall(PetscLayoutSetBlockSize(mat->rmap, fromRow->rmap->bs));
7837: if (fromCol->cmap->bs > 0) PetscCall(PetscLayoutSetBlockSize(mat->cmap, fromCol->cmap->bs));
7838: PetscFunctionReturn(PETSC_SUCCESS);
7839: }
7841: /*@
7842: MatResidual - Default routine to calculate the residual r = b - Ax
7844: Collective
7846: Input Parameters:
7847: + mat - the matrix
7848: . b - the right-hand-side
7849: - x - the approximate solution
7851: Output Parameter:
7852: . r - location to store the residual
7854: Level: developer
7856: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatMultAdd()`, `PCMGSetResidual()`
7857: @*/
7858: PetscErrorCode MatResidual(Mat mat, Vec b, Vec x, Vec r)
7859: {
7860: PetscFunctionBegin;
7866: MatCheckPreallocated(mat, 1);
7867: PetscCall(PetscLogEventBegin(MAT_Residual, mat, 0, 0, 0));
7868: if (!mat->ops->residual) {
7869: PetscCall(MatMult(mat, x, r));
7870: PetscCall(VecAYPX(r, -1.0, b));
7871: } else {
7872: PetscUseTypeMethod(mat, residual, b, x, r);
7873: }
7874: PetscCall(PetscLogEventEnd(MAT_Residual, mat, 0, 0, 0));
7875: PetscFunctionReturn(PETSC_SUCCESS);
7876: }
7878: /*MC
7879: MatGetRowIJF90 - Obtains the compressed row storage i and j indices for the local rows of a sparse matrix
7881: Synopsis:
7882: MatGetRowIJF90(Mat A, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt n, {PetscInt, pointer :: ia(:)}, {PetscInt, pointer :: ja(:)}, PetscBool done,integer ierr)
7884: Not Collective
7886: Input Parameters:
7887: + A - the matrix
7888: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
7889: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
7890: - inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
7891: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
7892: always used.
7894: Output Parameters:
7895: + n - number of local rows in the (possibly compressed) matrix
7896: . ia - the row pointers; that is ia[0] = 0, ia[row] = ia[row-1] + number of elements in that row of the matrix
7897: . ja - the column indices
7898: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
7899: are responsible for handling the case when done == `PETSC_FALSE` and ia and ja are not set
7901: Level: developer
7903: Note:
7904: Use `MatRestoreRowIJF90()` when you no longer need access to the data
7906: .seealso: [](ch_matrices), [](sec_fortranarrays), `Mat`, `MATMPIAIJ`, `MatGetRowIJ()`, `MatRestoreRowIJ()`, `MatRestoreRowIJF90()`
7907: M*/
7909: /*MC
7910: MatRestoreRowIJF90 - restores the compressed row storage i and j indices for the local rows of a sparse matrix obtained with `MatGetRowIJF90()`
7912: Synopsis:
7913: MatRestoreRowIJF90(Mat A, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt n, {PetscInt, pointer :: ia(:)}, {PetscInt, pointer :: ja(:)}, PetscBool done,integer ierr)
7915: Not Collective
7917: Input Parameters:
7918: + A - the matrix
7919: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
7920: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
7921: inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
7922: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
7923: always used.
7924: . n - number of local rows in the (possibly compressed) matrix
7925: . ia - the row pointers; that is ia[0] = 0, ia[row] = ia[row-1] + number of elements in that row of the matrix
7926: . ja - the column indices
7927: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
7928: are responsible for handling the case when done == `PETSC_FALSE` and ia and ja are not set
7930: Level: developer
7932: .seealso: [](ch_matrices), [](sec_fortranarrays), `Mat`, `MATMPIAIJ`, `MatGetRowIJ()`, `MatRestoreRowIJ()`, `MatGetRowIJF90()`
7933: M*/
7935: /*@C
7936: MatGetRowIJ - Returns the compressed row storage i and j indices for the local rows of a sparse matrix
7938: Collective
7940: Input Parameters:
7941: + mat - the matrix
7942: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
7943: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
7944: - inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
7945: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
7946: always used.
7948: Output Parameters:
7949: + n - number of local rows in the (possibly compressed) matrix, use `NULL` if not needed
7950: . 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
7951: . ja - the column indices, use `NULL` if not needed
7952: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
7953: are responsible for handling the case when done == `PETSC_FALSE` and ia and ja are not set
7955: Level: developer
7957: Notes:
7958: You CANNOT change any of the ia[] or ja[] values.
7960: Use `MatRestoreRowIJ()` when you are finished accessing the ia[] and ja[] values.
7962: Fortran Notes:
7963: Use
7964: .vb
7965: PetscInt, pointer :: ia(:),ja(:)
7966: call MatGetRowIJF90(mat,shift,symmetric,inodecompressed,n,ia,ja,done,ierr)
7967: ! Access the ith and jth entries via ia(i) and ja(j)
7968: .ve
7970: `MatGetRowIJ()` Fortran binding is deprecated (since PETSc 3.19), use `MatGetRowIJF90()`
7972: .seealso: [](ch_matrices), `Mat`, `MATAIJ`, `MatGetRowIJF90()`, `MatGetColumnIJ()`, `MatRestoreRowIJ()`, `MatSeqAIJGetArray()`
7973: @*/
7974: PetscErrorCode MatGetRowIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
7975: {
7976: PetscFunctionBegin;
7979: if (n) PetscAssertPointer(n, 5);
7980: if (ia) PetscAssertPointer(ia, 6);
7981: if (ja) PetscAssertPointer(ja, 7);
7982: if (done) PetscAssertPointer(done, 8);
7983: MatCheckPreallocated(mat, 1);
7984: if (!mat->ops->getrowij && done) *done = PETSC_FALSE;
7985: else {
7986: if (done) *done = PETSC_TRUE;
7987: PetscCall(PetscLogEventBegin(MAT_GetRowIJ, mat, 0, 0, 0));
7988: PetscUseTypeMethod(mat, getrowij, shift, symmetric, inodecompressed, n, ia, ja, done);
7989: PetscCall(PetscLogEventEnd(MAT_GetRowIJ, mat, 0, 0, 0));
7990: }
7991: PetscFunctionReturn(PETSC_SUCCESS);
7992: }
7994: /*@C
7995: MatGetColumnIJ - Returns the compressed column storage i and j indices for sequential matrices.
7997: Collective
7999: Input Parameters:
8000: + mat - the matrix
8001: . shift - 1 or zero indicating we want the indices starting at 0 or 1
8002: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be
8003: symmetrized
8004: . inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8005: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8006: always used.
8007: . n - number of columns in the (possibly compressed) matrix
8008: . ia - the column pointers; that is ia[0] = 0, ia[col] = i[col-1] + number of elements in that col of the matrix
8009: - ja - the row indices
8011: Output Parameter:
8012: . done - `PETSC_TRUE` or `PETSC_FALSE`, indicating whether the values have been returned
8014: Level: developer
8016: .seealso: [](ch_matrices), `Mat`, `MatGetRowIJ()`, `MatRestoreColumnIJ()`
8017: @*/
8018: PetscErrorCode MatGetColumnIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
8019: {
8020: PetscFunctionBegin;
8023: PetscAssertPointer(n, 5);
8024: if (ia) PetscAssertPointer(ia, 6);
8025: if (ja) PetscAssertPointer(ja, 7);
8026: PetscAssertPointer(done, 8);
8027: MatCheckPreallocated(mat, 1);
8028: if (!mat->ops->getcolumnij) *done = PETSC_FALSE;
8029: else {
8030: *done = PETSC_TRUE;
8031: PetscUseTypeMethod(mat, getcolumnij, shift, symmetric, inodecompressed, n, ia, ja, done);
8032: }
8033: PetscFunctionReturn(PETSC_SUCCESS);
8034: }
8036: /*@C
8037: MatRestoreRowIJ - Call after you are completed with the ia,ja indices obtained with `MatGetRowIJ()`.
8039: Collective
8041: Input Parameters:
8042: + mat - the matrix
8043: . shift - 1 or zero indicating we want the indices starting at 0 or 1
8044: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
8045: . inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8046: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8047: always used.
8048: . n - size of (possibly compressed) matrix
8049: . ia - the row pointers
8050: - ja - the column indices
8052: Output Parameter:
8053: . done - `PETSC_TRUE` or `PETSC_FALSE` indicated that the values have been returned
8055: Level: developer
8057: Note:
8058: This routine zeros out `n`, `ia`, and `ja`. This is to prevent accidental
8059: us of the array after it has been restored. If you pass `NULL`, it will
8060: not zero the pointers. Use of ia or ja after `MatRestoreRowIJ()` is invalid.
8062: Fortran Note:
8063: `MatRestoreRowIJ()` Fortran binding is deprecated (since PETSc 3.19), use `MatRestoreRowIJF90()`
8065: .seealso: [](ch_matrices), `Mat`, `MatGetRowIJ()`, `MatRestoreRowIJF90()`, `MatRestoreColumnIJ()`
8066: @*/
8067: PetscErrorCode MatRestoreRowIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
8068: {
8069: PetscFunctionBegin;
8072: if (ia) PetscAssertPointer(ia, 6);
8073: if (ja) PetscAssertPointer(ja, 7);
8074: if (done) PetscAssertPointer(done, 8);
8075: MatCheckPreallocated(mat, 1);
8077: if (!mat->ops->restorerowij && done) *done = PETSC_FALSE;
8078: else {
8079: if (done) *done = PETSC_TRUE;
8080: PetscUseTypeMethod(mat, restorerowij, shift, symmetric, inodecompressed, n, ia, ja, done);
8081: if (n) *n = 0;
8082: if (ia) *ia = NULL;
8083: if (ja) *ja = NULL;
8084: }
8085: PetscFunctionReturn(PETSC_SUCCESS);
8086: }
8088: /*@C
8089: MatRestoreColumnIJ - Call after you are completed with the ia,ja indices obtained with `MatGetColumnIJ()`.
8091: Collective
8093: Input Parameters:
8094: + mat - the matrix
8095: . shift - 1 or zero indicating we want the indices starting at 0 or 1
8096: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
8097: - inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8098: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8099: always used.
8101: Output Parameters:
8102: + n - size of (possibly compressed) matrix
8103: . ia - the column pointers
8104: . ja - the row indices
8105: - done - `PETSC_TRUE` or `PETSC_FALSE` indicated that the values have been returned
8107: Level: developer
8109: .seealso: [](ch_matrices), `Mat`, `MatGetColumnIJ()`, `MatRestoreRowIJ()`
8110: @*/
8111: PetscErrorCode MatRestoreColumnIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
8112: {
8113: PetscFunctionBegin;
8116: if (ia) PetscAssertPointer(ia, 6);
8117: if (ja) PetscAssertPointer(ja, 7);
8118: PetscAssertPointer(done, 8);
8119: MatCheckPreallocated(mat, 1);
8121: if (!mat->ops->restorecolumnij) *done = PETSC_FALSE;
8122: else {
8123: *done = PETSC_TRUE;
8124: PetscUseTypeMethod(mat, restorecolumnij, shift, symmetric, inodecompressed, n, ia, ja, done);
8125: if (n) *n = 0;
8126: if (ia) *ia = NULL;
8127: if (ja) *ja = NULL;
8128: }
8129: PetscFunctionReturn(PETSC_SUCCESS);
8130: }
8132: /*@C
8133: MatColoringPatch - Used inside matrix coloring routines that use `MatGetRowIJ()` and/or
8134: `MatGetColumnIJ()`.
8136: Collective
8138: Input Parameters:
8139: + mat - the matrix
8140: . ncolors - maximum color value
8141: . n - number of entries in colorarray
8142: - colorarray - array indicating color for each column
8144: Output Parameter:
8145: . iscoloring - coloring generated using colorarray information
8147: Level: developer
8149: .seealso: [](ch_matrices), `Mat`, `MatGetRowIJ()`, `MatGetColumnIJ()`
8150: @*/
8151: PetscErrorCode MatColoringPatch(Mat mat, PetscInt ncolors, PetscInt n, ISColoringValue colorarray[], ISColoring *iscoloring)
8152: {
8153: PetscFunctionBegin;
8156: PetscAssertPointer(colorarray, 4);
8157: PetscAssertPointer(iscoloring, 5);
8158: MatCheckPreallocated(mat, 1);
8160: if (!mat->ops->coloringpatch) {
8161: PetscCall(ISColoringCreate(PetscObjectComm((PetscObject)mat), ncolors, n, colorarray, PETSC_OWN_POINTER, iscoloring));
8162: } else {
8163: PetscUseTypeMethod(mat, coloringpatch, ncolors, n, colorarray, iscoloring);
8164: }
8165: PetscFunctionReturn(PETSC_SUCCESS);
8166: }
8168: /*@
8169: MatSetUnfactored - Resets a factored matrix to be treated as unfactored.
8171: Logically Collective
8173: Input Parameter:
8174: . mat - the factored matrix to be reset
8176: Level: developer
8178: Notes:
8179: This routine should be used only with factored matrices formed by in-place
8180: factorization via ILU(0) (or by in-place LU factorization for the `MATSEQDENSE`
8181: format). This option can save memory, for example, when solving nonlinear
8182: systems with a matrix-free Newton-Krylov method and a matrix-based, in-place
8183: ILU(0) preconditioner.
8185: One can specify in-place ILU(0) factorization by calling
8186: .vb
8187: PCType(pc,PCILU);
8188: PCFactorSeUseInPlace(pc);
8189: .ve
8190: or by using the options -pc_type ilu -pc_factor_in_place
8192: In-place factorization ILU(0) can also be used as a local
8193: solver for the blocks within the block Jacobi or additive Schwarz
8194: methods (runtime option: -sub_pc_factor_in_place). See Users-Manual: ch_pc
8195: for details on setting local solver options.
8197: Most users should employ the `KSP` interface for linear solvers
8198: instead of working directly with matrix algebra routines such as this.
8199: See, e.g., `KSPCreate()`.
8201: .seealso: [](ch_matrices), `Mat`, `PCFactorSetUseInPlace()`, `PCFactorGetUseInPlace()`
8202: @*/
8203: PetscErrorCode MatSetUnfactored(Mat mat)
8204: {
8205: PetscFunctionBegin;
8208: MatCheckPreallocated(mat, 1);
8209: mat->factortype = MAT_FACTOR_NONE;
8210: if (!mat->ops->setunfactored) PetscFunctionReturn(PETSC_SUCCESS);
8211: PetscUseTypeMethod(mat, setunfactored);
8212: PetscFunctionReturn(PETSC_SUCCESS);
8213: }
8215: /*MC
8216: MatDenseGetArrayF90 - Accesses a matrix array from Fortran
8218: Synopsis:
8219: MatDenseGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:,:)},integer ierr)
8221: Not Collective
8223: Input Parameter:
8224: . x - matrix
8226: Output Parameters:
8227: + xx_v - the Fortran pointer to the array
8228: - ierr - error code
8230: Example of Usage:
8231: .vb
8232: PetscScalar, pointer xx_v(:,:)
8233: ....
8234: call MatDenseGetArrayF90(x,xx_v,ierr)
8235: a = xx_v(3)
8236: call MatDenseRestoreArrayF90(x,xx_v,ierr)
8237: .ve
8239: Level: advanced
8241: .seealso: [](ch_matrices), `Mat`, `MatDenseRestoreArrayF90()`, `MatDenseGetArray()`, `MatDenseRestoreArray()`, `MatSeqAIJGetArrayF90()`
8242: M*/
8244: /*MC
8245: MatDenseRestoreArrayF90 - Restores a matrix array that has been
8246: accessed with `MatDenseGetArrayF90()`.
8248: Synopsis:
8249: MatDenseRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:,:)},integer ierr)
8251: Not Collective
8253: Input Parameters:
8254: + x - matrix
8255: - xx_v - the Fortran90 pointer to the array
8257: Output Parameter:
8258: . ierr - error code
8260: Example of Usage:
8261: .vb
8262: PetscScalar, pointer xx_v(:,:)
8263: ....
8264: call MatDenseGetArrayF90(x,xx_v,ierr)
8265: a = xx_v(3)
8266: call MatDenseRestoreArrayF90(x,xx_v,ierr)
8267: .ve
8269: Level: advanced
8271: .seealso: [](ch_matrices), `Mat`, `MatDenseGetArrayF90()`, `MatDenseGetArray()`, `MatDenseRestoreArray()`, `MatSeqAIJRestoreArrayF90()`
8272: M*/
8274: /*MC
8275: MatSeqAIJGetArrayF90 - Accesses a matrix array from Fortran.
8277: Synopsis:
8278: MatSeqAIJGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
8280: Not Collective
8282: Input Parameter:
8283: . x - matrix
8285: Output Parameters:
8286: + xx_v - the Fortran pointer to the array
8287: - ierr - error code
8289: Example of Usage:
8290: .vb
8291: PetscScalar, pointer xx_v(:)
8292: ....
8293: call MatSeqAIJGetArrayF90(x,xx_v,ierr)
8294: a = xx_v(3)
8295: call MatSeqAIJRestoreArrayF90(x,xx_v,ierr)
8296: .ve
8298: Level: advanced
8300: .seealso: [](ch_matrices), `Mat`, `MatSeqAIJRestoreArrayF90()`, `MatSeqAIJGetArray()`, `MatSeqAIJRestoreArray()`, `MatDenseGetArrayF90()`
8301: M*/
8303: /*MC
8304: MatSeqAIJRestoreArrayF90 - Restores a matrix array that has been
8305: accessed with `MatSeqAIJGetArrayF90()`.
8307: Synopsis:
8308: MatSeqAIJRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
8310: Not Collective
8312: Input Parameters:
8313: + x - matrix
8314: - xx_v - the Fortran90 pointer to the array
8316: Output Parameter:
8317: . ierr - error code
8319: Example of Usage:
8320: .vb
8321: PetscScalar, pointer xx_v(:)
8322: ....
8323: call MatSeqAIJGetArrayF90(x,xx_v,ierr)
8324: a = xx_v(3)
8325: call MatSeqAIJRestoreArrayF90(x,xx_v,ierr)
8326: .ve
8328: Level: advanced
8330: .seealso: [](ch_matrices), `Mat`, `MatSeqAIJGetArrayF90()`, `MatSeqAIJGetArray()`, `MatSeqAIJRestoreArray()`, `MatDenseRestoreArrayF90()`
8331: M*/
8333: /*@
8334: MatCreateSubMatrix - Gets a single submatrix on the same number of processors
8335: as the original matrix.
8337: Collective
8339: Input Parameters:
8340: + mat - the original matrix
8341: . isrow - parallel `IS` containing the rows this processor should obtain
8342: . 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.
8343: - cll - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
8345: Output Parameter:
8346: . newmat - the new submatrix, of the same type as the original matrix
8348: Level: advanced
8350: Notes:
8351: The submatrix will be able to be multiplied with vectors using the same layout as `iscol`.
8353: Some matrix types place restrictions on the row and column indices, such
8354: 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;
8355: for example, if the block size is 3 one cannot select the 0 and 2 rows without selecting the 1 row.
8357: The index sets may not have duplicate entries.
8359: The first time this is called you should use a cll of `MAT_INITIAL_MATRIX`,
8360: the `MatCreateSubMatrix()` routine will create the newmat for you. Any additional calls
8361: to this routine with a mat of the same nonzero structure and with a call of `MAT_REUSE_MATRIX`
8362: will reuse the matrix generated the first time. You should call `MatDestroy()` on `newmat` when
8363: you are finished using it.
8365: The communicator of the newly obtained matrix is ALWAYS the same as the communicator of
8366: the input matrix.
8368: If `iscol` is `NULL` then all columns are obtained (not supported in Fortran).
8370: If `isrow` and `iscol` have a nontrivial block-size then the resulting matrix has this block-size as well. This feature
8371: is used by `PCFIELDSPLIT` to allow easy nesting of its use.
8373: Example usage:
8374: Consider the following 8x8 matrix with 34 non-zero values, that is
8375: assembled across 3 processors. Let's assume that proc0 owns 3 rows,
8376: proc1 owns 3 rows, proc2 owns 2 rows. This division can be shown
8377: as follows
8378: .vb
8379: 1 2 0 | 0 3 0 | 0 4
8380: Proc0 0 5 6 | 7 0 0 | 8 0
8381: 9 0 10 | 11 0 0 | 12 0
8382: -------------------------------------
8383: 13 0 14 | 15 16 17 | 0 0
8384: Proc1 0 18 0 | 19 20 21 | 0 0
8385: 0 0 0 | 22 23 0 | 24 0
8386: -------------------------------------
8387: Proc2 25 26 27 | 0 0 28 | 29 0
8388: 30 0 0 | 31 32 33 | 0 34
8389: .ve
8391: Suppose `isrow` = [0 1 | 4 | 6 7] and `iscol` = [1 2 | 3 4 5 | 6]. The resulting submatrix is
8393: .vb
8394: 2 0 | 0 3 0 | 0
8395: Proc0 5 6 | 7 0 0 | 8
8396: -------------------------------
8397: Proc1 18 0 | 19 20 21 | 0
8398: -------------------------------
8399: Proc2 26 27 | 0 0 28 | 29
8400: 0 0 | 31 32 33 | 0
8401: .ve
8403: .seealso: [](ch_matrices), `Mat`, `MatCreateSubMatrices()`, `MatCreateSubMatricesMPI()`, `MatCreateSubMatrixVirtual()`, `MatSubMatrixVirtualUpdate()`
8404: @*/
8405: PetscErrorCode MatCreateSubMatrix(Mat mat, IS isrow, IS iscol, MatReuse cll, Mat *newmat)
8406: {
8407: PetscMPIInt size;
8408: Mat *local;
8409: IS iscoltmp;
8410: PetscBool flg;
8412: PetscFunctionBegin;
8416: PetscAssertPointer(newmat, 5);
8419: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
8420: PetscCheck(cll != MAT_IGNORE_MATRIX, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Cannot use MAT_IGNORE_MATRIX");
8422: MatCheckPreallocated(mat, 1);
8423: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
8425: if (!iscol || isrow == iscol) {
8426: PetscBool stride;
8427: PetscMPIInt grabentirematrix = 0, grab;
8428: PetscCall(PetscObjectTypeCompare((PetscObject)isrow, ISSTRIDE, &stride));
8429: if (stride) {
8430: PetscInt first, step, n, rstart, rend;
8431: PetscCall(ISStrideGetInfo(isrow, &first, &step));
8432: if (step == 1) {
8433: PetscCall(MatGetOwnershipRange(mat, &rstart, &rend));
8434: if (rstart == first) {
8435: PetscCall(ISGetLocalSize(isrow, &n));
8436: if (n == rend - rstart) grabentirematrix = 1;
8437: }
8438: }
8439: }
8440: PetscCall(MPIU_Allreduce(&grabentirematrix, &grab, 1, MPI_INT, MPI_MIN, PetscObjectComm((PetscObject)mat)));
8441: if (grab) {
8442: PetscCall(PetscInfo(mat, "Getting entire matrix as submatrix\n"));
8443: if (cll == MAT_INITIAL_MATRIX) {
8444: *newmat = mat;
8445: PetscCall(PetscObjectReference((PetscObject)mat));
8446: }
8447: PetscFunctionReturn(PETSC_SUCCESS);
8448: }
8449: }
8451: if (!iscol) {
8452: PetscCall(ISCreateStride(PetscObjectComm((PetscObject)mat), mat->cmap->n, mat->cmap->rstart, 1, &iscoltmp));
8453: } else {
8454: iscoltmp = iscol;
8455: }
8457: /* if original matrix is on just one processor then use submatrix generated */
8458: if (mat->ops->createsubmatrices && !mat->ops->createsubmatrix && size == 1 && cll == MAT_REUSE_MATRIX) {
8459: PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscoltmp, MAT_REUSE_MATRIX, &newmat));
8460: goto setproperties;
8461: } else if (mat->ops->createsubmatrices && !mat->ops->createsubmatrix && size == 1) {
8462: PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscoltmp, MAT_INITIAL_MATRIX, &local));
8463: *newmat = *local;
8464: PetscCall(PetscFree(local));
8465: goto setproperties;
8466: } else if (!mat->ops->createsubmatrix) {
8467: /* Create a new matrix type that implements the operation using the full matrix */
8468: PetscCall(PetscLogEventBegin(MAT_CreateSubMat, mat, 0, 0, 0));
8469: switch (cll) {
8470: case MAT_INITIAL_MATRIX:
8471: PetscCall(MatCreateSubMatrixVirtual(mat, isrow, iscoltmp, newmat));
8472: break;
8473: case MAT_REUSE_MATRIX:
8474: PetscCall(MatSubMatrixVirtualUpdate(*newmat, mat, isrow, iscoltmp));
8475: break;
8476: default:
8477: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Invalid MatReuse, must be either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX");
8478: }
8479: PetscCall(PetscLogEventEnd(MAT_CreateSubMat, mat, 0, 0, 0));
8480: goto setproperties;
8481: }
8483: PetscCall(PetscLogEventBegin(MAT_CreateSubMat, mat, 0, 0, 0));
8484: PetscUseTypeMethod(mat, createsubmatrix, isrow, iscoltmp, cll, newmat);
8485: PetscCall(PetscLogEventEnd(MAT_CreateSubMat, mat, 0, 0, 0));
8487: setproperties:
8488: PetscCall(ISEqualUnsorted(isrow, iscoltmp, &flg));
8489: if (flg) PetscCall(MatPropagateSymmetryOptions(mat, *newmat));
8490: if (!iscol) PetscCall(ISDestroy(&iscoltmp));
8491: if (*newmat && cll == MAT_INITIAL_MATRIX) PetscCall(PetscObjectStateIncrease((PetscObject)*newmat));
8492: PetscFunctionReturn(PETSC_SUCCESS);
8493: }
8495: /*@
8496: MatPropagateSymmetryOptions - Propagates symmetry options set on a matrix to another matrix
8498: Not Collective
8500: Input Parameters:
8501: + A - the matrix we wish to propagate options from
8502: - B - the matrix we wish to propagate options to
8504: Level: beginner
8506: Note:
8507: Propagates the options associated to `MAT_SYMMETRY_ETERNAL`, `MAT_STRUCTURALLY_SYMMETRIC`, `MAT_HERMITIAN`, `MAT_SPD`, `MAT_SYMMETRIC`, and `MAT_STRUCTURAL_SYMMETRY_ETERNAL`
8509: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatIsSymmetricKnown()`, `MatIsSPDKnown()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetricKnown()`
8510: @*/
8511: PetscErrorCode MatPropagateSymmetryOptions(Mat A, Mat B)
8512: {
8513: PetscFunctionBegin;
8516: B->symmetry_eternal = A->symmetry_eternal;
8517: B->structural_symmetry_eternal = A->structural_symmetry_eternal;
8518: B->symmetric = A->symmetric;
8519: B->structurally_symmetric = A->structurally_symmetric;
8520: B->spd = A->spd;
8521: B->hermitian = A->hermitian;
8522: PetscFunctionReturn(PETSC_SUCCESS);
8523: }
8525: /*@
8526: MatStashSetInitialSize - sets the sizes of the matrix stash, that is
8527: used during the assembly process to store values that belong to
8528: other processors.
8530: Not Collective
8532: Input Parameters:
8533: + mat - the matrix
8534: . size - the initial size of the stash.
8535: - bsize - the initial size of the block-stash(if used).
8537: Options Database Keys:
8538: + -matstash_initial_size <size> or <size0,size1,...sizep-1> - set initial size
8539: - -matstash_block_initial_size <bsize> or <bsize0,bsize1,...bsizep-1> - set initial block size
8541: Level: intermediate
8543: Notes:
8544: The block-stash is used for values set with `MatSetValuesBlocked()` while
8545: the stash is used for values set with `MatSetValues()`
8547: Run with the option -info and look for output of the form
8548: MatAssemblyBegin_MPIXXX:Stash has MM entries, uses nn mallocs.
8549: to determine the appropriate value, MM, to use for size and
8550: MatAssemblyBegin_MPIXXX:Block-Stash has BMM entries, uses nn mallocs.
8551: to determine the value, BMM to use for bsize
8553: .seealso: [](ch_matrices), `MatAssemblyBegin()`, `MatAssemblyEnd()`, `Mat`, `MatStashGetInfo()`
8554: @*/
8555: PetscErrorCode MatStashSetInitialSize(Mat mat, PetscInt size, PetscInt bsize)
8556: {
8557: PetscFunctionBegin;
8560: PetscCall(MatStashSetInitialSize_Private(&mat->stash, size));
8561: PetscCall(MatStashSetInitialSize_Private(&mat->bstash, bsize));
8562: PetscFunctionReturn(PETSC_SUCCESS);
8563: }
8565: /*@
8566: MatInterpolateAdd - $w = y + A*x$ or $A^T*x$ depending on the shape of
8567: the matrix
8569: Neighbor-wise Collective
8571: Input Parameters:
8572: + A - the matrix
8573: . x - the vector to be multiplied by the interpolation operator
8574: - y - the vector to be added to the result
8576: Output Parameter:
8577: . w - the resulting vector
8579: Level: intermediate
8581: Notes:
8582: `w` may be the same vector as `y`.
8584: This allows one to use either the restriction or interpolation (its transpose)
8585: matrix to do the interpolation
8587: .seealso: [](ch_matrices), `Mat`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatRestrict()`, `PCMG`
8588: @*/
8589: PetscErrorCode MatInterpolateAdd(Mat A, Vec x, Vec y, Vec w)
8590: {
8591: PetscInt M, N, Ny;
8593: PetscFunctionBegin;
8598: PetscCall(MatGetSize(A, &M, &N));
8599: PetscCall(VecGetSize(y, &Ny));
8600: if (M == Ny) {
8601: PetscCall(MatMultAdd(A, x, y, w));
8602: } else {
8603: PetscCall(MatMultTransposeAdd(A, x, y, w));
8604: }
8605: PetscFunctionReturn(PETSC_SUCCESS);
8606: }
8608: /*@
8609: MatInterpolate - $y = A*x$ or $A^T*x$ depending on the shape of
8610: the matrix
8612: Neighbor-wise Collective
8614: Input Parameters:
8615: + A - the matrix
8616: - x - the vector to be interpolated
8618: Output Parameter:
8619: . y - the resulting vector
8621: Level: intermediate
8623: Note:
8624: This allows one to use either the restriction or interpolation (its transpose)
8625: matrix to do the interpolation
8627: .seealso: [](ch_matrices), `Mat`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatRestrict()`, `PCMG`
8628: @*/
8629: PetscErrorCode MatInterpolate(Mat A, Vec x, Vec y)
8630: {
8631: PetscInt M, N, Ny;
8633: PetscFunctionBegin;
8637: PetscCall(MatGetSize(A, &M, &N));
8638: PetscCall(VecGetSize(y, &Ny));
8639: if (M == Ny) {
8640: PetscCall(MatMult(A, x, y));
8641: } else {
8642: PetscCall(MatMultTranspose(A, x, y));
8643: }
8644: PetscFunctionReturn(PETSC_SUCCESS);
8645: }
8647: /*@
8648: MatRestrict - $y = A*x$ or $A^T*x$
8650: Neighbor-wise Collective
8652: Input Parameters:
8653: + A - the matrix
8654: - x - the vector to be restricted
8656: Output Parameter:
8657: . y - the resulting vector
8659: Level: intermediate
8661: Note:
8662: This allows one to use either the restriction or interpolation (its transpose)
8663: matrix to do the restriction
8665: .seealso: [](ch_matrices), `Mat`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatInterpolate()`, `PCMG`
8666: @*/
8667: PetscErrorCode MatRestrict(Mat A, Vec x, Vec y)
8668: {
8669: PetscInt M, N, Ny;
8671: PetscFunctionBegin;
8675: PetscCall(MatGetSize(A, &M, &N));
8676: PetscCall(VecGetSize(y, &Ny));
8677: if (M == Ny) {
8678: PetscCall(MatMult(A, x, y));
8679: } else {
8680: PetscCall(MatMultTranspose(A, x, y));
8681: }
8682: PetscFunctionReturn(PETSC_SUCCESS);
8683: }
8685: /*@
8686: MatMatInterpolateAdd - $Y = W + A*X$ or $W + A^T*X$ depending on the shape of `A`
8688: Neighbor-wise Collective
8690: Input Parameters:
8691: + A - the matrix
8692: . x - the input dense matrix to be multiplied
8693: - w - the input dense matrix to be added to the result
8695: Output Parameter:
8696: . y - the output dense matrix
8698: Level: intermediate
8700: Note:
8701: This allows one to use either the restriction or interpolation (its transpose)
8702: matrix to do the interpolation. `y` matrix can be reused if already created with the proper sizes,
8703: otherwise it will be recreated. `y` must be initialized to `NULL` if not supplied.
8705: .seealso: [](ch_matrices), `Mat`, `MatInterpolateAdd()`, `MatMatInterpolate()`, `MatMatRestrict()`, `PCMG`
8706: @*/
8707: PetscErrorCode MatMatInterpolateAdd(Mat A, Mat x, Mat w, Mat *y)
8708: {
8709: PetscInt M, N, Mx, Nx, Mo, My = 0, Ny = 0;
8710: PetscBool trans = PETSC_TRUE;
8711: MatReuse reuse = MAT_INITIAL_MATRIX;
8713: PetscFunctionBegin;
8719: PetscCall(MatGetSize(A, &M, &N));
8720: PetscCall(MatGetSize(x, &Mx, &Nx));
8721: if (N == Mx) trans = PETSC_FALSE;
8722: 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);
8723: Mo = trans ? N : M;
8724: if (*y) {
8725: PetscCall(MatGetSize(*y, &My, &Ny));
8726: if (Mo == My && Nx == Ny) {
8727: reuse = MAT_REUSE_MATRIX;
8728: } else {
8729: 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);
8730: PetscCall(MatDestroy(y));
8731: }
8732: }
8734: if (w && *y == w) { /* this is to minimize changes in PCMG */
8735: PetscBool flg;
8737: PetscCall(PetscObjectQuery((PetscObject)*y, "__MatMatIntAdd_w", (PetscObject *)&w));
8738: if (w) {
8739: PetscInt My, Ny, Mw, Nw;
8741: PetscCall(PetscObjectTypeCompare((PetscObject)*y, ((PetscObject)w)->type_name, &flg));
8742: PetscCall(MatGetSize(*y, &My, &Ny));
8743: PetscCall(MatGetSize(w, &Mw, &Nw));
8744: if (!flg || My != Mw || Ny != Nw) w = NULL;
8745: }
8746: if (!w) {
8747: PetscCall(MatDuplicate(*y, MAT_COPY_VALUES, &w));
8748: PetscCall(PetscObjectCompose((PetscObject)*y, "__MatMatIntAdd_w", (PetscObject)w));
8749: PetscCall(PetscObjectDereference((PetscObject)w));
8750: } else {
8751: PetscCall(MatCopy(*y, w, UNKNOWN_NONZERO_PATTERN));
8752: }
8753: }
8754: if (!trans) {
8755: PetscCall(MatMatMult(A, x, reuse, PETSC_DEFAULT, y));
8756: } else {
8757: PetscCall(MatTransposeMatMult(A, x, reuse, PETSC_DEFAULT, y));
8758: }
8759: if (w) PetscCall(MatAXPY(*y, 1.0, w, UNKNOWN_NONZERO_PATTERN));
8760: PetscFunctionReturn(PETSC_SUCCESS);
8761: }
8763: /*@
8764: MatMatInterpolate - $Y = A*X$ or $A^T*X$ depending on the shape of `A`
8766: Neighbor-wise Collective
8768: Input Parameters:
8769: + A - the matrix
8770: - x - the input dense matrix
8772: Output Parameter:
8773: . y - the output dense matrix
8775: Level: intermediate
8777: Note:
8778: This allows one to use either the restriction or interpolation (its transpose)
8779: matrix to do the interpolation. `y` matrix can be reused if already created with the proper sizes,
8780: otherwise it will be recreated. `y` must be initialized to `NULL` if not supplied.
8782: .seealso: [](ch_matrices), `Mat`, `MatInterpolate()`, `MatRestrict()`, `MatMatRestrict()`, `PCMG`
8783: @*/
8784: PetscErrorCode MatMatInterpolate(Mat A, Mat x, Mat *y)
8785: {
8786: PetscFunctionBegin;
8787: PetscCall(MatMatInterpolateAdd(A, x, NULL, y));
8788: PetscFunctionReturn(PETSC_SUCCESS);
8789: }
8791: /*@
8792: MatMatRestrict - $Y = A*X$ or $A^T*X$ depending on the shape of `A`
8794: Neighbor-wise Collective
8796: Input Parameters:
8797: + A - the matrix
8798: - x - the input dense matrix
8800: Output Parameter:
8801: . y - the output dense matrix
8803: Level: intermediate
8805: Note:
8806: This allows one to use either the restriction or interpolation (its transpose)
8807: matrix to do the restriction. `y` matrix can be reused if already created with the proper sizes,
8808: otherwise it will be recreated. `y` must be initialized to `NULL` if not supplied.
8810: .seealso: [](ch_matrices), `Mat`, `MatRestrict()`, `MatInterpolate()`, `MatMatInterpolate()`, `PCMG`
8811: @*/
8812: PetscErrorCode MatMatRestrict(Mat A, Mat x, Mat *y)
8813: {
8814: PetscFunctionBegin;
8815: PetscCall(MatMatInterpolateAdd(A, x, NULL, y));
8816: PetscFunctionReturn(PETSC_SUCCESS);
8817: }
8819: /*@
8820: MatGetNullSpace - retrieves the null space of a matrix.
8822: Logically Collective
8824: Input Parameters:
8825: + mat - the matrix
8826: - nullsp - the null space object
8828: Level: developer
8830: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatSetNullSpace()`, `MatNullSpace`
8831: @*/
8832: PetscErrorCode MatGetNullSpace(Mat mat, MatNullSpace *nullsp)
8833: {
8834: PetscFunctionBegin;
8836: PetscAssertPointer(nullsp, 2);
8837: *nullsp = (mat->symmetric == PETSC_BOOL3_TRUE && !mat->nullsp) ? mat->transnullsp : mat->nullsp;
8838: PetscFunctionReturn(PETSC_SUCCESS);
8839: }
8841: /*@
8842: MatSetNullSpace - attaches a null space to a matrix.
8844: Logically Collective
8846: Input Parameters:
8847: + mat - the matrix
8848: - nullsp - the null space object
8850: Level: advanced
8852: Notes:
8853: This null space is used by the `KSP` linear solvers to solve singular systems.
8855: 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`
8857: 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 to
8858: to zero but the linear system will still be solved in a least squares sense.
8860: The fundamental theorem of linear algebra (Gilbert Strang, Introduction to Applied Mathematics, page 72) states that
8861: 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), + the range of $A^T$, $R(A^T)$.
8862: 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
8863: 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
8864: 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($A^T$).
8865: This \hat{b} can be obtained by calling `MatNullSpaceRemove()` with the null space of the transpose of the matrix.
8867: If the matrix is known to be symmetric because it is an `MATSBAIJ` matrix or one as called
8868: `MatSetOption`(mat,`MAT_SYMMETRIC` or possibly `MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`); this
8869: routine also automatically calls `MatSetTransposeNullSpace()`.
8871: The user should call `MatNullSpaceDestroy()`.
8873: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatSetTransposeNullSpace()`, `MatGetTransposeNullSpace()`, `MatNullSpaceRemove()`,
8874: `KSPSetPCSide()`
8875: @*/
8876: PetscErrorCode MatSetNullSpace(Mat mat, MatNullSpace nullsp)
8877: {
8878: PetscFunctionBegin;
8881: if (nullsp) PetscCall(PetscObjectReference((PetscObject)nullsp));
8882: PetscCall(MatNullSpaceDestroy(&mat->nullsp));
8883: mat->nullsp = nullsp;
8884: if (mat->symmetric == PETSC_BOOL3_TRUE) PetscCall(MatSetTransposeNullSpace(mat, nullsp));
8885: PetscFunctionReturn(PETSC_SUCCESS);
8886: }
8888: /*@
8889: MatGetTransposeNullSpace - retrieves the null space of the transpose of a matrix.
8891: Logically Collective
8893: Input Parameters:
8894: + mat - the matrix
8895: - nullsp - the null space object
8897: Level: developer
8899: .seealso: [](ch_matrices), `Mat`, `MatNullSpace`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatSetTransposeNullSpace()`, `MatSetNullSpace()`, `MatGetNullSpace()`
8900: @*/
8901: PetscErrorCode MatGetTransposeNullSpace(Mat mat, MatNullSpace *nullsp)
8902: {
8903: PetscFunctionBegin;
8906: PetscAssertPointer(nullsp, 2);
8907: *nullsp = (mat->symmetric == PETSC_BOOL3_TRUE && !mat->transnullsp) ? mat->nullsp : mat->transnullsp;
8908: PetscFunctionReturn(PETSC_SUCCESS);
8909: }
8911: /*@
8912: MatSetTransposeNullSpace - attaches the null space of a transpose of a matrix to the matrix
8914: Logically Collective
8916: Input Parameters:
8917: + mat - the matrix
8918: - nullsp - the null space object
8920: Level: advanced
8922: Notes:
8923: This allows solving singular linear systems defined by the transpose of the matrix using `KSP` solvers with left preconditioning.
8925: See `MatSetNullSpace()`
8927: .seealso: [](ch_matrices), `Mat`, `MatNullSpace`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatSetNullSpace()`, `MatGetTransposeNullSpace()`, `MatNullSpaceRemove()`, `KSPSetPCSide()`
8928: @*/
8929: PetscErrorCode MatSetTransposeNullSpace(Mat mat, MatNullSpace nullsp)
8930: {
8931: PetscFunctionBegin;
8934: if (nullsp) PetscCall(PetscObjectReference((PetscObject)nullsp));
8935: PetscCall(MatNullSpaceDestroy(&mat->transnullsp));
8936: mat->transnullsp = nullsp;
8937: PetscFunctionReturn(PETSC_SUCCESS);
8938: }
8940: /*@
8941: MatSetNearNullSpace - attaches a null space to a matrix, which is often the null space (rigid body modes) of the operator without boundary conditions
8942: This null space will be used to provide near null space vectors to a multigrid preconditioner built from this matrix.
8944: Logically Collective
8946: Input Parameters:
8947: + mat - the matrix
8948: - nullsp - the null space object
8950: Level: advanced
8952: Notes:
8953: Overwrites any previous near null space that may have been attached
8955: You can remove the null space by calling this routine with an `nullsp` of `NULL`
8957: .seealso: [](ch_matrices), `Mat`, `MatNullSpace`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNullSpace()`, `MatNullSpaceCreateRigidBody()`, `MatGetNearNullSpace()`
8958: @*/
8959: PetscErrorCode MatSetNearNullSpace(Mat mat, MatNullSpace nullsp)
8960: {
8961: PetscFunctionBegin;
8965: MatCheckPreallocated(mat, 1);
8966: if (nullsp) PetscCall(PetscObjectReference((PetscObject)nullsp));
8967: PetscCall(MatNullSpaceDestroy(&mat->nearnullsp));
8968: mat->nearnullsp = nullsp;
8969: PetscFunctionReturn(PETSC_SUCCESS);
8970: }
8972: /*@
8973: MatGetNearNullSpace - Get null space attached with `MatSetNearNullSpace()`
8975: Not Collective
8977: Input Parameter:
8978: . mat - the matrix
8980: Output Parameter:
8981: . nullsp - the null space object, `NULL` if not set
8983: Level: advanced
8985: .seealso: [](ch_matrices), `Mat`, `MatNullSpace`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatNullSpaceCreate()`
8986: @*/
8987: PetscErrorCode MatGetNearNullSpace(Mat mat, MatNullSpace *nullsp)
8988: {
8989: PetscFunctionBegin;
8992: PetscAssertPointer(nullsp, 2);
8993: MatCheckPreallocated(mat, 1);
8994: *nullsp = mat->nearnullsp;
8995: PetscFunctionReturn(PETSC_SUCCESS);
8996: }
8998: /*@C
8999: MatICCFactor - Performs in-place incomplete Cholesky factorization of matrix.
9001: Collective
9003: Input Parameters:
9004: + mat - the matrix
9005: . row - row/column permutation
9006: - info - information on desired factorization process
9008: Level: developer
9010: Notes:
9011: Probably really in-place only when level of fill is zero, otherwise allocates
9012: new space to store factored matrix and deletes previous memory.
9014: Most users should employ the `KSP` interface for linear solvers
9015: instead of working directly with matrix algebra routines such as this.
9016: See, e.g., `KSPCreate()`.
9018: Developer Note:
9019: The Fortran interface is not autogenerated as the
9020: interface definition cannot be generated correctly [due to `MatFactorInfo`]
9022: .seealso: [](ch_matrices), `Mat`, `MatFactorInfo`, `MatGetFactor()`, `MatICCFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`
9023: @*/
9024: PetscErrorCode MatICCFactor(Mat mat, IS row, const MatFactorInfo *info)
9025: {
9026: PetscFunctionBegin;
9030: PetscAssertPointer(info, 3);
9031: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "matrix must be square");
9032: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
9033: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
9034: MatCheckPreallocated(mat, 1);
9035: PetscUseTypeMethod(mat, iccfactor, row, info);
9036: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
9037: PetscFunctionReturn(PETSC_SUCCESS);
9038: }
9040: /*@
9041: MatDiagonalScaleLocal - Scales columns of a matrix given the scaling values including the
9042: ghosted ones.
9044: Not Collective
9046: Input Parameters:
9047: + mat - the matrix
9048: - diag - the diagonal values, including ghost ones
9050: Level: developer
9052: Notes:
9053: Works only for `MATMPIAIJ` and `MATMPIBAIJ` matrices
9055: This allows one to avoid during communication to perform the scaling that must be done with `MatDiagonalScale()`
9057: .seealso: [](ch_matrices), `Mat`, `MatDiagonalScale()`
9058: @*/
9059: PetscErrorCode MatDiagonalScaleLocal(Mat mat, Vec diag)
9060: {
9061: PetscMPIInt size;
9063: PetscFunctionBegin;
9068: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Matrix must be already assembled");
9069: PetscCall(PetscLogEventBegin(MAT_Scale, mat, 0, 0, 0));
9070: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
9071: if (size == 1) {
9072: PetscInt n, m;
9073: PetscCall(VecGetSize(diag, &n));
9074: PetscCall(MatGetSize(mat, NULL, &m));
9075: PetscCheck(m == n, PETSC_COMM_SELF, PETSC_ERR_SUP, "Only supported for sequential matrices when no ghost points/periodic conditions");
9076: PetscCall(MatDiagonalScale(mat, NULL, diag));
9077: } else {
9078: PetscUseMethod(mat, "MatDiagonalScaleLocal_C", (Mat, Vec), (mat, diag));
9079: }
9080: PetscCall(PetscLogEventEnd(MAT_Scale, mat, 0, 0, 0));
9081: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
9082: PetscFunctionReturn(PETSC_SUCCESS);
9083: }
9085: /*@
9086: MatGetInertia - Gets the inertia from a factored matrix
9088: Collective
9090: Input Parameter:
9091: . mat - the matrix
9093: Output Parameters:
9094: + nneg - number of negative eigenvalues
9095: . nzero - number of zero eigenvalues
9096: - npos - number of positive eigenvalues
9098: Level: advanced
9100: Note:
9101: Matrix must have been factored by `MatCholeskyFactor()`
9103: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatCholeskyFactor()`
9104: @*/
9105: PetscErrorCode MatGetInertia(Mat mat, PetscInt *nneg, PetscInt *nzero, PetscInt *npos)
9106: {
9107: PetscFunctionBegin;
9110: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
9111: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Numeric factor mat is not assembled");
9112: PetscUseTypeMethod(mat, getinertia, nneg, nzero, npos);
9113: PetscFunctionReturn(PETSC_SUCCESS);
9114: }
9116: /*@C
9117: MatSolves - Solves $A x = b$, given a factored matrix, for a collection of vectors
9119: Neighbor-wise Collective
9121: Input Parameters:
9122: + mat - the factored matrix obtained with `MatGetFactor()`
9123: - b - the right-hand-side vectors
9125: Output Parameter:
9126: . x - the result vectors
9128: Level: developer
9130: Note:
9131: The vectors `b` and `x` cannot be the same. I.e., one cannot
9132: call `MatSolves`(A,x,x).
9134: .seealso: [](ch_matrices), `Mat`, `Vecs`, `MatSolveAdd()`, `MatSolveTranspose()`, `MatSolveTransposeAdd()`, `MatSolve()`
9135: @*/
9136: PetscErrorCode MatSolves(Mat mat, Vecs b, Vecs x)
9137: {
9138: PetscFunctionBegin;
9141: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
9142: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
9143: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
9145: MatCheckPreallocated(mat, 1);
9146: PetscCall(PetscLogEventBegin(MAT_Solves, mat, 0, 0, 0));
9147: PetscUseTypeMethod(mat, solves, b, x);
9148: PetscCall(PetscLogEventEnd(MAT_Solves, mat, 0, 0, 0));
9149: PetscFunctionReturn(PETSC_SUCCESS);
9150: }
9152: /*@
9153: MatIsSymmetric - Test whether a matrix is symmetric
9155: Collective
9157: Input Parameters:
9158: + A - the matrix to test
9159: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact transpose)
9161: Output Parameter:
9162: . flg - the result
9164: Level: intermediate
9166: Notes:
9167: For real numbers `MatIsSymmetric()` and `MatIsHermitian()` return identical results
9169: If the matrix does not yet know if it is symmetric or not this can be an expensive operation, also available `MatIsSymmetricKnown()`
9171: One can declare that a matrix is symmetric with `MatSetOption`(mat,`MAT_SYMMETRIC`,`PETSC_TRUE`) and if it is known to remain symmetric
9172: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9174: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetricKnown()`,
9175: `MAT_SYMMETRIC`, `MAT_SYMMETRY_ETERNAL`
9176: @*/
9177: PetscErrorCode MatIsSymmetric(Mat A, PetscReal tol, PetscBool *flg)
9178: {
9179: PetscFunctionBegin;
9181: PetscAssertPointer(flg, 3);
9183: if (A->symmetric == PETSC_BOOL3_TRUE) *flg = PETSC_TRUE;
9184: else if (A->symmetric == PETSC_BOOL3_FALSE) *flg = PETSC_FALSE;
9185: else {
9186: PetscUseTypeMethod(A, issymmetric, tol, flg);
9187: if (!tol) PetscCall(MatSetOption(A, MAT_SYMMETRIC, *flg));
9188: }
9189: PetscFunctionReturn(PETSC_SUCCESS);
9190: }
9192: /*@
9193: MatIsHermitian - Test whether a matrix is Hermitian
9195: Collective
9197: Input Parameters:
9198: + A - the matrix to test
9199: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact Hermitian)
9201: Output Parameter:
9202: . flg - the result
9204: Level: intermediate
9206: Notes:
9207: For real numbers `MatIsSymmetric()` and `MatIsHermitian()` return identical results
9209: If the matrix does not yet know if it is Hermitian or not this can be an expensive operation, also available `MatIsHermitianKnown()`
9211: One can declare that a matrix is Hermitian with `MatSetOption`(mat,`MAT_HERMITIAN`,`PETSC_TRUE`) and if it is known to remain Hermitian
9212: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMEMTRY_ETERNAL`,`PETSC_TRUE`)
9214: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetric()`, `MatSetOption()`,
9215: `MatIsSymmetricKnown()`, `MatIsSymmetric()`, `MAT_HERMITIAN`, `MAT_SYMMETRY_ETERNAL`
9216: @*/
9217: PetscErrorCode MatIsHermitian(Mat A, PetscReal tol, PetscBool *flg)
9218: {
9219: PetscFunctionBegin;
9221: PetscAssertPointer(flg, 3);
9223: if (A->hermitian == PETSC_BOOL3_TRUE) *flg = PETSC_TRUE;
9224: else if (A->hermitian == PETSC_BOOL3_FALSE) *flg = PETSC_FALSE;
9225: else {
9226: PetscUseTypeMethod(A, ishermitian, tol, flg);
9227: if (!tol) PetscCall(MatSetOption(A, MAT_HERMITIAN, *flg));
9228: }
9229: PetscFunctionReturn(PETSC_SUCCESS);
9230: }
9232: /*@
9233: MatIsSymmetricKnown - Checks if a matrix knows if it is symmetric or not and its symmetric state
9235: Not Collective
9237: Input Parameter:
9238: . A - the matrix to check
9240: Output Parameters:
9241: + set - `PETSC_TRUE` if the matrix knows its symmetry state (this tells you if the next flag is valid)
9242: - flg - the result (only valid if set is `PETSC_TRUE`)
9244: Level: advanced
9246: Notes:
9247: Does not check the matrix values directly, so this may return unknown (set = `PETSC_FALSE`). Use `MatIsSymmetric()`
9248: if you want it explicitly checked
9250: One can declare that a matrix is symmetric with `MatSetOption`(mat,`MAT_SYMMETRIC`,`PETSC_TRUE`) and if it is known to remain symmetric
9251: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9253: .seealso: [](ch_matrices), `Mat`, `MAT_SYMMETRY_ETERNAL`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitianKnown()`
9254: @*/
9255: PetscErrorCode MatIsSymmetricKnown(Mat A, PetscBool *set, PetscBool *flg)
9256: {
9257: PetscFunctionBegin;
9259: PetscAssertPointer(set, 2);
9260: PetscAssertPointer(flg, 3);
9261: if (A->symmetric != PETSC_BOOL3_UNKNOWN) {
9262: *set = PETSC_TRUE;
9263: *flg = PetscBool3ToBool(A->symmetric);
9264: } else {
9265: *set = PETSC_FALSE;
9266: }
9267: PetscFunctionReturn(PETSC_SUCCESS);
9268: }
9270: /*@
9271: MatIsSPDKnown - Checks if a matrix knows if it is symmetric positive definite or not and its symmetric positive definite state
9273: Not Collective
9275: Input Parameter:
9276: . A - the matrix to check
9278: Output Parameters:
9279: + set - `PETSC_TRUE` if the matrix knows its symmetric positive definite state (this tells you if the next flag is valid)
9280: - flg - the result (only valid if set is `PETSC_TRUE`)
9282: Level: advanced
9284: Notes:
9285: Does not check the matrix values directly, so this may return unknown (set = `PETSC_FALSE`).
9287: One can declare that a matrix is SPD with `MatSetOption`(mat,`MAT_SPD`,`PETSC_TRUE`) and if it is known to remain SPD
9288: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SPD_ETERNAL`,`PETSC_TRUE`)
9290: .seealso: [](ch_matrices), `Mat`, `MAT_SPD_ETERNAL`, `MAT_SPD`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitianKnown()`
9291: @*/
9292: PetscErrorCode MatIsSPDKnown(Mat A, PetscBool *set, PetscBool *flg)
9293: {
9294: PetscFunctionBegin;
9296: PetscAssertPointer(set, 2);
9297: PetscAssertPointer(flg, 3);
9298: if (A->spd != PETSC_BOOL3_UNKNOWN) {
9299: *set = PETSC_TRUE;
9300: *flg = PetscBool3ToBool(A->spd);
9301: } else {
9302: *set = PETSC_FALSE;
9303: }
9304: PetscFunctionReturn(PETSC_SUCCESS);
9305: }
9307: /*@
9308: MatIsHermitianKnown - Checks if a matrix knows if it is Hermitian or not and its Hermitian state
9310: Not Collective
9312: Input Parameter:
9313: . A - the matrix to check
9315: Output Parameters:
9316: + set - `PETSC_TRUE` if the matrix knows its Hermitian state (this tells you if the next flag is valid)
9317: - flg - the result (only valid if set is `PETSC_TRUE`)
9319: Level: advanced
9321: Notes:
9322: Does not check the matrix values directly, so this may return unknown (set = `PETSC_FALSE`). Use `MatIsHermitian()`
9323: if you want it explicitly checked
9325: One can declare that a matrix is Hermitian with `MatSetOption`(mat,`MAT_HERMITIAN`,`PETSC_TRUE`) and if it is known to remain Hermitian
9326: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9328: .seealso: [](ch_matrices), `Mat`, `MAT_SYMMETRY_ETERNAL`, `MAT_HERMITIAN`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`
9329: @*/
9330: PetscErrorCode MatIsHermitianKnown(Mat A, PetscBool *set, PetscBool *flg)
9331: {
9332: PetscFunctionBegin;
9334: PetscAssertPointer(set, 2);
9335: PetscAssertPointer(flg, 3);
9336: if (A->hermitian != PETSC_BOOL3_UNKNOWN) {
9337: *set = PETSC_TRUE;
9338: *flg = PetscBool3ToBool(A->hermitian);
9339: } else {
9340: *set = PETSC_FALSE;
9341: }
9342: PetscFunctionReturn(PETSC_SUCCESS);
9343: }
9345: /*@
9346: MatIsStructurallySymmetric - Test whether a matrix is structurally symmetric
9348: Collective
9350: Input Parameter:
9351: . A - the matrix to test
9353: Output Parameter:
9354: . flg - the result
9356: Level: intermediate
9358: Notes:
9359: If the matrix does yet know it is structurally symmetric this can be an expensive operation, also available `MatIsStructurallySymmetricKnown()`
9361: 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
9362: symmetric after changes to the matrices values one can call `MatSetOption`(mat,`MAT_STRUCTURAL_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9364: .seealso: [](ch_matrices), `Mat`, `MAT_STRUCTURALLY_SYMMETRIC`, `MAT_STRUCTURAL_SYMMETRY_ETERNAL`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsSymmetric()`, `MatSetOption()`, `MatIsStructurallySymmetricKnown()`
9365: @*/
9366: PetscErrorCode MatIsStructurallySymmetric(Mat A, PetscBool *flg)
9367: {
9368: PetscFunctionBegin;
9370: PetscAssertPointer(flg, 2);
9371: if (A->structurally_symmetric != PETSC_BOOL3_UNKNOWN) {
9372: *flg = PetscBool3ToBool(A->structurally_symmetric);
9373: } else {
9374: PetscUseTypeMethod(A, isstructurallysymmetric, flg);
9375: PetscCall(MatSetOption(A, MAT_STRUCTURALLY_SYMMETRIC, *flg));
9376: }
9377: PetscFunctionReturn(PETSC_SUCCESS);
9378: }
9380: /*@
9381: MatIsStructurallySymmetricKnown - Checks if a matrix knows if it is structurally symmetric or not and its structurally symmetric state
9383: Not Collective
9385: Input Parameter:
9386: . A - the matrix to check
9388: Output Parameters:
9389: + set - PETSC_TRUE if the matrix knows its structurally symmetric state (this tells you if the next flag is valid)
9390: - flg - the result (only valid if set is PETSC_TRUE)
9392: Level: advanced
9394: Notes:
9395: 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
9396: symmetric after changes to the matrices values one can call `MatSetOption`(mat,`MAT_STRUCTURAL_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9398: Use `MatIsStructurallySymmetric()` to explicitly check if a matrix is structurally symmetric (this is an expensive operation)
9400: .seealso: [](ch_matrices), `Mat`, `MAT_STRUCTURALLY_SYMMETRIC`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitianKnown()`
9401: @*/
9402: PetscErrorCode MatIsStructurallySymmetricKnown(Mat A, PetscBool *set, PetscBool *flg)
9403: {
9404: PetscFunctionBegin;
9406: PetscAssertPointer(set, 2);
9407: PetscAssertPointer(flg, 3);
9408: if (A->structurally_symmetric != PETSC_BOOL3_UNKNOWN) {
9409: *set = PETSC_TRUE;
9410: *flg = PetscBool3ToBool(A->structurally_symmetric);
9411: } else {
9412: *set = PETSC_FALSE;
9413: }
9414: PetscFunctionReturn(PETSC_SUCCESS);
9415: }
9417: /*@
9418: MatStashGetInfo - Gets how many values are currently in the matrix stash, i.e. need
9419: to be communicated to other processors during the `MatAssemblyBegin()`/`MatAssemblyEnd()` process
9421: Not Collective
9423: Input Parameter:
9424: . mat - the matrix
9426: Output Parameters:
9427: + nstash - the size of the stash
9428: . reallocs - the number of additional mallocs incurred.
9429: . bnstash - the size of the block stash
9430: - breallocs - the number of additional mallocs incurred.in the block stash
9432: Level: advanced
9434: .seealso: [](ch_matrices), `MatAssemblyBegin()`, `MatAssemblyEnd()`, `Mat`, `MatStashSetInitialSize()`
9435: @*/
9436: PetscErrorCode MatStashGetInfo(Mat mat, PetscInt *nstash, PetscInt *reallocs, PetscInt *bnstash, PetscInt *breallocs)
9437: {
9438: PetscFunctionBegin;
9439: PetscCall(MatStashGetInfo_Private(&mat->stash, nstash, reallocs));
9440: PetscCall(MatStashGetInfo_Private(&mat->bstash, bnstash, breallocs));
9441: PetscFunctionReturn(PETSC_SUCCESS);
9442: }
9444: /*@C
9445: MatCreateVecs - Get vector(s) compatible with the matrix, i.e. with the same
9446: parallel layout, `PetscLayout` for rows and columns
9448: Collective
9450: Input Parameter:
9451: . mat - the matrix
9453: Output Parameters:
9454: + right - (optional) vector that the matrix can be multiplied against
9455: - left - (optional) vector that the matrix vector product can be stored in
9457: Level: advanced
9459: Notes:
9460: 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()`.
9462: These are new vectors which are not owned by the mat, they should be destroyed in `VecDestroy()` when no longer needed
9464: .seealso: [](ch_matrices), `Mat`, `Vec`, `VecCreate()`, `VecDestroy()`, `DMCreateGlobalVector()`
9465: @*/
9466: PetscErrorCode MatCreateVecs(Mat mat, Vec *right, Vec *left)
9467: {
9468: PetscFunctionBegin;
9471: if (mat->ops->getvecs) {
9472: PetscUseTypeMethod(mat, getvecs, right, left);
9473: } else {
9474: if (right) {
9475: PetscCheck(mat->cmap->n >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "PetscLayout for columns not yet setup");
9476: PetscCall(VecCreateWithLayout_Private(mat->cmap, right));
9477: PetscCall(VecSetType(*right, mat->defaultvectype));
9478: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
9479: if (mat->boundtocpu && mat->bindingpropagates) {
9480: PetscCall(VecSetBindingPropagates(*right, PETSC_TRUE));
9481: PetscCall(VecBindToCPU(*right, PETSC_TRUE));
9482: }
9483: #endif
9484: }
9485: if (left) {
9486: PetscCheck(mat->rmap->n >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "PetscLayout for rows not yet setup");
9487: PetscCall(VecCreateWithLayout_Private(mat->rmap, left));
9488: PetscCall(VecSetType(*left, mat->defaultvectype));
9489: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
9490: if (mat->boundtocpu && mat->bindingpropagates) {
9491: PetscCall(VecSetBindingPropagates(*left, PETSC_TRUE));
9492: PetscCall(VecBindToCPU(*left, PETSC_TRUE));
9493: }
9494: #endif
9495: }
9496: }
9497: PetscFunctionReturn(PETSC_SUCCESS);
9498: }
9500: /*@C
9501: MatFactorInfoInitialize - Initializes a `MatFactorInfo` data structure
9502: with default values.
9504: Not Collective
9506: Input Parameter:
9507: . info - the `MatFactorInfo` data structure
9509: Level: developer
9511: Notes:
9512: The solvers are generally used through the `KSP` and `PC` objects, for example
9513: `PCLU`, `PCILU`, `PCCHOLESKY`, `PCICC`
9515: Once the data structure is initialized one may change certain entries as desired for the particular factorization to be performed
9517: Developer Note:
9518: The Fortran interface is not autogenerated as the
9519: interface definition cannot be generated correctly [due to `MatFactorInfo`]
9521: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorInfo`
9522: @*/
9523: PetscErrorCode MatFactorInfoInitialize(MatFactorInfo *info)
9524: {
9525: PetscFunctionBegin;
9526: PetscCall(PetscMemzero(info, sizeof(MatFactorInfo)));
9527: PetscFunctionReturn(PETSC_SUCCESS);
9528: }
9530: /*@
9531: MatFactorSetSchurIS - Set indices corresponding to the Schur complement you wish to have computed
9533: Collective
9535: Input Parameters:
9536: + mat - the factored matrix
9537: - is - the index set defining the Schur indices (0-based)
9539: Level: advanced
9541: Notes:
9542: Call `MatFactorSolveSchurComplement()` or `MatFactorSolveSchurComplementTranspose()` after this call to solve a Schur complement system.
9544: You can call `MatFactorGetSchurComplement()` or `MatFactorCreateSchurComplement()` after this call.
9546: This functionality is only supported for `MATSOLVERMUMPS` and `MATSOLVERMKL_PARDISO`
9548: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorGetSchurComplement()`, `MatFactorRestoreSchurComplement()`, `MatFactorCreateSchurComplement()`, `MatFactorSolveSchurComplement()`,
9549: `MatFactorSolveSchurComplementTranspose()`, `MATSOLVERMUMPS`, `MATSOLVERMKL_PARDISO`
9550: @*/
9551: PetscErrorCode MatFactorSetSchurIS(Mat mat, IS is)
9552: {
9553: PetscErrorCode (*f)(Mat, IS);
9555: PetscFunctionBegin;
9560: PetscCheckSameComm(mat, 1, is, 2);
9561: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Only for factored matrix");
9562: PetscCall(PetscObjectQueryFunction((PetscObject)mat, "MatFactorSetSchurIS_C", &f));
9563: PetscCheck(f, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "The selected MatSolverType does not support Schur complement computation. You should use MATSOLVERMUMPS or MATSOLVERMKL_PARDISO");
9564: PetscCall(MatDestroy(&mat->schur));
9565: PetscCall((*f)(mat, is));
9566: PetscCheck(mat->schur, PetscObjectComm((PetscObject)mat), PETSC_ERR_PLIB, "Schur complement has not been created");
9567: PetscFunctionReturn(PETSC_SUCCESS);
9568: }
9570: /*@
9571: MatFactorCreateSchurComplement - Create a Schur complement matrix object using Schur data computed during the factorization step
9573: Logically Collective
9575: Input Parameters:
9576: + F - the factored matrix obtained by calling `MatGetFactor()`
9577: . S - location where to return the Schur complement, can be `NULL`
9578: - status - the status of the Schur complement matrix, can be `NULL`
9580: Level: advanced
9582: Notes:
9583: You must call `MatFactorSetSchurIS()` before calling this routine.
9585: This functionality is only supported for `MATSOLVERMUMPS` and `MATSOLVERMKL_PARDISO`
9587: The routine provides a copy of the Schur matrix stored within the solver data structures.
9588: The caller must destroy the object when it is no longer needed.
9589: If `MatFactorInvertSchurComplement()` has been called, the routine gets back the inverse.
9591: 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)
9593: See `MatCreateSchurComplement()` or `MatGetSchurComplement()` for ways to create virtual or approximate Schur complements.
9595: Developer Note:
9596: The reason this routine exists is because the representation of the Schur complement within the factor matrix may be different than a standard PETSc
9597: matrix representation and we normally do not want to use the time or memory to make a copy as a regular PETSc matrix.
9599: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorGetSchurComplement()`, `MatFactorSchurStatus`, `MATSOLVERMUMPS`, `MATSOLVERMKL_PARDISO`
9600: @*/
9601: PetscErrorCode MatFactorCreateSchurComplement(Mat F, Mat *S, MatFactorSchurStatus *status)
9602: {
9603: PetscFunctionBegin;
9605: if (S) PetscAssertPointer(S, 2);
9606: if (status) PetscAssertPointer(status, 3);
9607: if (S) {
9608: PetscErrorCode (*f)(Mat, Mat *);
9610: PetscCall(PetscObjectQueryFunction((PetscObject)F, "MatFactorCreateSchurComplement_C", &f));
9611: if (f) {
9612: PetscCall((*f)(F, S));
9613: } else {
9614: PetscCall(MatDuplicate(F->schur, MAT_COPY_VALUES, S));
9615: }
9616: }
9617: if (status) *status = F->schur_status;
9618: PetscFunctionReturn(PETSC_SUCCESS);
9619: }
9621: /*@
9622: MatFactorGetSchurComplement - Gets access to a Schur complement matrix using the current Schur data within a factored matrix
9624: Logically Collective
9626: Input Parameters:
9627: + F - the factored matrix obtained by calling `MatGetFactor()`
9628: . S - location where to return the Schur complement, can be `NULL`
9629: - status - the status of the Schur complement matrix, can be `NULL`
9631: Level: advanced
9633: Notes:
9634: You must call `MatFactorSetSchurIS()` before calling this routine.
9636: Schur complement mode is currently implemented for sequential matrices with factor type of `MATSOLVERMUMPS`
9638: The routine returns a the Schur Complement stored within the data structures of the solver.
9640: If `MatFactorInvertSchurComplement()` has previously been called, the returned matrix is actually the inverse of the Schur complement.
9642: The returned matrix should not be destroyed; the caller should call `MatFactorRestoreSchurComplement()` when the object is no longer needed.
9644: Use `MatFactorCreateSchurComplement()` to create a copy of the Schur complement matrix that is within a factored matrix
9646: See `MatCreateSchurComplement()` or `MatGetSchurComplement()` for ways to create virtual or approximate Schur complements.
9648: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorRestoreSchurComplement()`, `MatFactorCreateSchurComplement()`, `MatFactorSchurStatus`
9649: @*/
9650: PetscErrorCode MatFactorGetSchurComplement(Mat F, Mat *S, MatFactorSchurStatus *status)
9651: {
9652: PetscFunctionBegin;
9654: if (S) {
9655: PetscAssertPointer(S, 2);
9656: *S = F->schur;
9657: }
9658: if (status) {
9659: PetscAssertPointer(status, 3);
9660: *status = F->schur_status;
9661: }
9662: PetscFunctionReturn(PETSC_SUCCESS);
9663: }
9665: static PetscErrorCode MatFactorUpdateSchurStatus_Private(Mat F)
9666: {
9667: Mat S = F->schur;
9669: PetscFunctionBegin;
9670: switch (F->schur_status) {
9671: case MAT_FACTOR_SCHUR_UNFACTORED: // fall-through
9672: case MAT_FACTOR_SCHUR_INVERTED:
9673: if (S) {
9674: S->ops->solve = NULL;
9675: S->ops->matsolve = NULL;
9676: S->ops->solvetranspose = NULL;
9677: S->ops->matsolvetranspose = NULL;
9678: S->ops->solveadd = NULL;
9679: S->ops->solvetransposeadd = NULL;
9680: S->factortype = MAT_FACTOR_NONE;
9681: PetscCall(PetscFree(S->solvertype));
9682: }
9683: case MAT_FACTOR_SCHUR_FACTORED: // fall-through
9684: break;
9685: default:
9686: SETERRQ(PetscObjectComm((PetscObject)F), PETSC_ERR_SUP, "Unhandled MatFactorSchurStatus %d", F->schur_status);
9687: }
9688: PetscFunctionReturn(PETSC_SUCCESS);
9689: }
9691: /*@
9692: MatFactorRestoreSchurComplement - Restore the Schur complement matrix object obtained from a call to `MatFactorGetSchurComplement()`
9694: Logically Collective
9696: Input Parameters:
9697: + F - the factored matrix obtained by calling `MatGetFactor()`
9698: . S - location where the Schur complement is stored
9699: - status - the status of the Schur complement matrix (see `MatFactorSchurStatus`)
9701: Level: advanced
9703: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorCreateSchurComplement()`, `MatFactorSchurStatus`
9704: @*/
9705: PetscErrorCode MatFactorRestoreSchurComplement(Mat F, Mat *S, MatFactorSchurStatus status)
9706: {
9707: PetscFunctionBegin;
9709: if (S) {
9711: *S = NULL;
9712: }
9713: F->schur_status = status;
9714: PetscCall(MatFactorUpdateSchurStatus_Private(F));
9715: PetscFunctionReturn(PETSC_SUCCESS);
9716: }
9718: /*@
9719: MatFactorSolveSchurComplementTranspose - Solve the transpose of the Schur complement system computed during the factorization step
9721: Logically Collective
9723: Input Parameters:
9724: + F - the factored matrix obtained by calling `MatGetFactor()`
9725: . rhs - location where the right hand side of the Schur complement system is stored
9726: - sol - location where the solution of the Schur complement system has to be returned
9728: Level: advanced
9730: Notes:
9731: The sizes of the vectors should match the size of the Schur complement
9733: Must be called after `MatFactorSetSchurIS()`
9735: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorSolveSchurComplement()`
9736: @*/
9737: PetscErrorCode MatFactorSolveSchurComplementTranspose(Mat F, Vec rhs, Vec sol)
9738: {
9739: PetscFunctionBegin;
9746: PetscCheckSameComm(F, 1, rhs, 2);
9747: PetscCheckSameComm(F, 1, sol, 3);
9748: PetscCall(MatFactorFactorizeSchurComplement(F));
9749: switch (F->schur_status) {
9750: case MAT_FACTOR_SCHUR_FACTORED:
9751: PetscCall(MatSolveTranspose(F->schur, rhs, sol));
9752: break;
9753: case MAT_FACTOR_SCHUR_INVERTED:
9754: PetscCall(MatMultTranspose(F->schur, rhs, sol));
9755: break;
9756: default:
9757: SETERRQ(PetscObjectComm((PetscObject)F), PETSC_ERR_SUP, "Unhandled MatFactorSchurStatus %d", F->schur_status);
9758: }
9759: PetscFunctionReturn(PETSC_SUCCESS);
9760: }
9762: /*@
9763: MatFactorSolveSchurComplement - Solve the Schur complement system computed during the factorization step
9765: Logically Collective
9767: Input Parameters:
9768: + F - the factored matrix obtained by calling `MatGetFactor()`
9769: . rhs - location where the right hand side of the Schur complement system is stored
9770: - sol - location where the solution of the Schur complement system has to be returned
9772: Level: advanced
9774: Notes:
9775: The sizes of the vectors should match the size of the Schur complement
9777: Must be called after `MatFactorSetSchurIS()`
9779: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorSolveSchurComplementTranspose()`
9780: @*/
9781: PetscErrorCode MatFactorSolveSchurComplement(Mat F, Vec rhs, Vec sol)
9782: {
9783: PetscFunctionBegin;
9790: PetscCheckSameComm(F, 1, rhs, 2);
9791: PetscCheckSameComm(F, 1, sol, 3);
9792: PetscCall(MatFactorFactorizeSchurComplement(F));
9793: switch (F->schur_status) {
9794: case MAT_FACTOR_SCHUR_FACTORED:
9795: PetscCall(MatSolve(F->schur, rhs, sol));
9796: break;
9797: case MAT_FACTOR_SCHUR_INVERTED:
9798: PetscCall(MatMult(F->schur, rhs, sol));
9799: break;
9800: default:
9801: SETERRQ(PetscObjectComm((PetscObject)F), PETSC_ERR_SUP, "Unhandled MatFactorSchurStatus %d", F->schur_status);
9802: }
9803: PetscFunctionReturn(PETSC_SUCCESS);
9804: }
9806: PETSC_EXTERN PetscErrorCode MatSeqDenseInvertFactors_Private(Mat);
9807: #if PetscDefined(HAVE_CUDA)
9808: PETSC_SINGLE_LIBRARY_INTERN PetscErrorCode MatSeqDenseCUDAInvertFactors_Internal(Mat);
9809: #endif
9811: /* Schur status updated in the interface */
9812: static PetscErrorCode MatFactorInvertSchurComplement_Private(Mat F)
9813: {
9814: Mat S = F->schur;
9816: PetscFunctionBegin;
9817: if (S) {
9818: PetscMPIInt size;
9819: PetscBool isdense, isdensecuda;
9821: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)S), &size));
9822: PetscCheck(size <= 1, PetscObjectComm((PetscObject)S), PETSC_ERR_SUP, "Not yet implemented");
9823: PetscCall(PetscObjectTypeCompare((PetscObject)S, MATSEQDENSE, &isdense));
9824: PetscCall(PetscObjectTypeCompare((PetscObject)S, MATSEQDENSECUDA, &isdensecuda));
9825: PetscCheck(isdense || isdensecuda, PetscObjectComm((PetscObject)S), PETSC_ERR_SUP, "Not implemented for type %s", ((PetscObject)S)->type_name);
9826: PetscCall(PetscLogEventBegin(MAT_FactorInvS, F, 0, 0, 0));
9827: if (isdense) {
9828: PetscCall(MatSeqDenseInvertFactors_Private(S));
9829: } else if (isdensecuda) {
9830: #if defined(PETSC_HAVE_CUDA)
9831: PetscCall(MatSeqDenseCUDAInvertFactors_Internal(S));
9832: #endif
9833: }
9834: // HIP??????????????
9835: PetscCall(PetscLogEventEnd(MAT_FactorInvS, F, 0, 0, 0));
9836: }
9837: PetscFunctionReturn(PETSC_SUCCESS);
9838: }
9840: /*@
9841: MatFactorInvertSchurComplement - Invert the Schur complement matrix computed during the factorization step
9843: Logically Collective
9845: Input Parameter:
9846: . F - the factored matrix obtained by calling `MatGetFactor()`
9848: Level: advanced
9850: Notes:
9851: Must be called after `MatFactorSetSchurIS()`.
9853: Call `MatFactorGetSchurComplement()` or `MatFactorCreateSchurComplement()` AFTER this call to actually compute the inverse and get access to it.
9855: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorGetSchurComplement()`, `MatFactorCreateSchurComplement()`
9856: @*/
9857: PetscErrorCode MatFactorInvertSchurComplement(Mat F)
9858: {
9859: PetscFunctionBegin;
9862: if (F->schur_status == MAT_FACTOR_SCHUR_INVERTED) PetscFunctionReturn(PETSC_SUCCESS);
9863: PetscCall(MatFactorFactorizeSchurComplement(F));
9864: PetscCall(MatFactorInvertSchurComplement_Private(F));
9865: F->schur_status = MAT_FACTOR_SCHUR_INVERTED;
9866: PetscFunctionReturn(PETSC_SUCCESS);
9867: }
9869: /*@
9870: MatFactorFactorizeSchurComplement - Factorize the Schur complement matrix computed during the factorization step
9872: Logically Collective
9874: Input Parameter:
9875: . F - the factored matrix obtained by calling `MatGetFactor()`
9877: Level: advanced
9879: Note:
9880: Must be called after `MatFactorSetSchurIS()`
9882: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorInvertSchurComplement()`
9883: @*/
9884: PetscErrorCode MatFactorFactorizeSchurComplement(Mat F)
9885: {
9886: MatFactorInfo info;
9888: PetscFunctionBegin;
9891: if (F->schur_status == MAT_FACTOR_SCHUR_INVERTED || F->schur_status == MAT_FACTOR_SCHUR_FACTORED) PetscFunctionReturn(PETSC_SUCCESS);
9892: PetscCall(PetscLogEventBegin(MAT_FactorFactS, F, 0, 0, 0));
9893: PetscCall(PetscMemzero(&info, sizeof(MatFactorInfo)));
9894: if (F->factortype == MAT_FACTOR_CHOLESKY) { /* LDL^t regarded as Cholesky */
9895: PetscCall(MatCholeskyFactor(F->schur, NULL, &info));
9896: } else {
9897: PetscCall(MatLUFactor(F->schur, NULL, NULL, &info));
9898: }
9899: PetscCall(PetscLogEventEnd(MAT_FactorFactS, F, 0, 0, 0));
9900: F->schur_status = MAT_FACTOR_SCHUR_FACTORED;
9901: PetscFunctionReturn(PETSC_SUCCESS);
9902: }
9904: /*@
9905: MatPtAP - Creates the matrix product $C = P^T * A * P$
9907: Neighbor-wise Collective
9909: Input Parameters:
9910: + A - the matrix
9911: . P - the projection matrix
9912: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
9913: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(P)), use `PETSC_DEFAULT` if you do not have a good estimate
9914: if the result is a dense matrix this is irrelevant
9916: Output Parameter:
9917: . C - the product matrix
9919: Level: intermediate
9921: Notes:
9922: C will be created and must be destroyed by the user with `MatDestroy()`.
9924: An alternative approach to this function is to use `MatProductCreate()` and set the desired options before the computation is done
9926: Developer Note:
9927: For matrix types without special implementation the function fallbacks to `MatMatMult()` followed by `MatTransposeMatMult()`.
9929: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MatMatMult()`, `MatRARt()`
9930: @*/
9931: PetscErrorCode MatPtAP(Mat A, Mat P, MatReuse scall, PetscReal fill, Mat *C)
9932: {
9933: PetscFunctionBegin;
9934: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*C, 5);
9935: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
9937: if (scall == MAT_INITIAL_MATRIX) {
9938: PetscCall(MatProductCreate(A, P, NULL, C));
9939: PetscCall(MatProductSetType(*C, MATPRODUCT_PtAP));
9940: PetscCall(MatProductSetAlgorithm(*C, "default"));
9941: PetscCall(MatProductSetFill(*C, fill));
9943: (*C)->product->api_user = PETSC_TRUE;
9944: PetscCall(MatProductSetFromOptions(*C));
9945: 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);
9946: PetscCall(MatProductSymbolic(*C));
9947: } else { /* scall == MAT_REUSE_MATRIX */
9948: PetscCall(MatProductReplaceMats(A, P, NULL, *C));
9949: }
9951: PetscCall(MatProductNumeric(*C));
9952: (*C)->symmetric = A->symmetric;
9953: (*C)->spd = A->spd;
9954: PetscFunctionReturn(PETSC_SUCCESS);
9955: }
9957: /*@
9958: MatRARt - Creates the matrix product $C = R * A * R^T$
9960: Neighbor-wise Collective
9962: Input Parameters:
9963: + A - the matrix
9964: . R - the projection matrix
9965: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
9966: - fill - expected fill as ratio of nnz(C)/nnz(A), use `PETSC_DEFAULT` if you do not have a good estimate
9967: if the result is a dense matrix this is irrelevant
9969: Output Parameter:
9970: . C - the product matrix
9972: Level: intermediate
9974: Notes:
9975: C will be created and must be destroyed by the user with `MatDestroy()`.
9977: An alternative approach to this function is to use `MatProductCreate()` and set the desired options before the computation is done
9979: This routine is currently only implemented for pairs of `MATAIJ` matrices and classes
9980: which inherit from `MATAIJ`. Due to PETSc sparse matrix block row distribution among processes,
9981: parallel MatRARt is implemented via explicit transpose of R, which could be very expensive.
9982: We recommend using MatPtAP().
9984: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MatMatMult()`, `MatPtAP()`
9985: @*/
9986: PetscErrorCode MatRARt(Mat A, Mat R, MatReuse scall, PetscReal fill, Mat *C)
9987: {
9988: PetscFunctionBegin;
9989: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*C, 5);
9990: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
9992: if (scall == MAT_INITIAL_MATRIX) {
9993: PetscCall(MatProductCreate(A, R, NULL, C));
9994: PetscCall(MatProductSetType(*C, MATPRODUCT_RARt));
9995: PetscCall(MatProductSetAlgorithm(*C, "default"));
9996: PetscCall(MatProductSetFill(*C, fill));
9998: (*C)->product->api_user = PETSC_TRUE;
9999: PetscCall(MatProductSetFromOptions(*C));
10000: 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);
10001: PetscCall(MatProductSymbolic(*C));
10002: } else { /* scall == MAT_REUSE_MATRIX */
10003: PetscCall(MatProductReplaceMats(A, R, NULL, *C));
10004: }
10006: PetscCall(MatProductNumeric(*C));
10007: if (A->symmetric == PETSC_BOOL3_TRUE) PetscCall(MatSetOption(*C, MAT_SYMMETRIC, PETSC_TRUE));
10008: PetscFunctionReturn(PETSC_SUCCESS);
10009: }
10011: static PetscErrorCode MatProduct_Private(Mat A, Mat B, MatReuse scall, PetscReal fill, MatProductType ptype, Mat *C)
10012: {
10013: PetscFunctionBegin;
10014: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
10016: if (scall == MAT_INITIAL_MATRIX) {
10017: PetscCall(PetscInfo(A, "Calling MatProduct API with MAT_INITIAL_MATRIX and product type %s\n", MatProductTypes[ptype]));
10018: PetscCall(MatProductCreate(A, B, NULL, C));
10019: PetscCall(MatProductSetType(*C, ptype));
10020: PetscCall(MatProductSetAlgorithm(*C, MATPRODUCTALGORITHMDEFAULT));
10021: PetscCall(MatProductSetFill(*C, fill));
10023: (*C)->product->api_user = PETSC_TRUE;
10024: PetscCall(MatProductSetFromOptions(*C));
10025: PetscCall(MatProductSymbolic(*C));
10026: } else { /* scall == MAT_REUSE_MATRIX */
10027: Mat_Product *product = (*C)->product;
10028: PetscBool isdense;
10030: PetscCall(PetscObjectBaseTypeCompareAny((PetscObject)(*C), &isdense, MATSEQDENSE, MATMPIDENSE, ""));
10031: if (isdense && product && product->type != ptype) {
10032: PetscCall(MatProductClear(*C));
10033: product = NULL;
10034: }
10035: PetscCall(PetscInfo(A, "Calling MatProduct API with MAT_REUSE_MATRIX %s product present and product type %s\n", product ? "with" : "without", MatProductTypes[ptype]));
10036: if (!product) { /* user provide the dense matrix *C without calling MatProductCreate() or reusing it from previous calls */
10037: PetscCheck(isdense, PetscObjectComm((PetscObject)(*C)), PETSC_ERR_SUP, "Call MatProductCreate() first");
10038: PetscCall(MatProductCreate_Private(A, B, NULL, *C));
10039: product = (*C)->product;
10040: product->fill = fill;
10041: product->api_user = PETSC_TRUE;
10042: product->clear = PETSC_TRUE;
10044: PetscCall(MatProductSetType(*C, ptype));
10045: PetscCall(MatProductSetFromOptions(*C));
10046: PetscCheck((*C)->ops->productsymbolic, PetscObjectComm((PetscObject)(*C)), PETSC_ERR_SUP, "MatProduct %s not supported for %s and %s", MatProductTypes[ptype], ((PetscObject)A)->type_name, ((PetscObject)B)->type_name);
10047: PetscCall(MatProductSymbolic(*C));
10048: } else { /* user may change input matrices A or B when REUSE */
10049: PetscCall(MatProductReplaceMats(A, B, NULL, *C));
10050: }
10051: }
10052: PetscCall(MatProductNumeric(*C));
10053: PetscFunctionReturn(PETSC_SUCCESS);
10054: }
10056: /*@
10057: MatMatMult - Performs matrix-matrix multiplication C=A*B.
10059: Neighbor-wise Collective
10061: Input Parameters:
10062: + A - the left matrix
10063: . B - the right matrix
10064: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10065: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use `PETSC_DEFAULT` if you do not have a good estimate
10066: if the result is a dense matrix this is irrelevant
10068: Output Parameter:
10069: . C - the product matrix
10071: Notes:
10072: Unless scall is `MAT_REUSE_MATRIX` C will be created.
10074: `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
10075: call to this function with `MAT_INITIAL_MATRIX`.
10077: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value actually needed.
10079: 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`,
10080: rather than first having `MatMatMult()` create it for you. You can NEVER do this if the matrix C is sparse.
10082: Example of Usage:
10083: .vb
10084: MatProductCreate(A,B,NULL,&C);
10085: MatProductSetType(C,MATPRODUCT_AB);
10086: MatProductSymbolic(C);
10087: MatProductNumeric(C); // compute C=A * B
10088: MatProductReplaceMats(A1,B1,NULL,C); // compute C=A1 * B1
10089: MatProductNumeric(C);
10090: MatProductReplaceMats(A2,NULL,NULL,C); // compute C=A2 * B1
10091: MatProductNumeric(C);
10092: .ve
10094: Level: intermediate
10096: .seealso: [](ch_matrices), `Mat`, `MatProductType`, `MATPRODUCT_AB`, `MatTransposeMatMult()`, `MatMatTransposeMult()`, `MatPtAP()`, `MatProductCreate()`, `MatProductSymbolic()`, `MatProductReplaceMats()`, `MatProductNumeric()`
10097: @*/
10098: PetscErrorCode MatMatMult(Mat A, Mat B, MatReuse scall, PetscReal fill, Mat *C)
10099: {
10100: PetscFunctionBegin;
10101: PetscCall(MatProduct_Private(A, B, scall, fill, MATPRODUCT_AB, C));
10102: PetscFunctionReturn(PETSC_SUCCESS);
10103: }
10105: /*@
10106: MatMatTransposeMult - Performs matrix-matrix multiplication $C = A*B^T$.
10108: Neighbor-wise Collective
10110: Input Parameters:
10111: + A - the left matrix
10112: . B - the right matrix
10113: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10114: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use `PETSC_DEFAULT` if not known
10116: Output Parameter:
10117: . C - the product matrix
10119: Options Database Key:
10120: . -matmattransmult_mpidense_mpidense_via {allgatherv,cyclic} - Choose between algorithms for `MATMPIDENSE` matrices: the
10121: first redundantly copies the transposed `B` matrix on each process and requires O(log P) communication complexity;
10122: the second never stores more than one portion of the `B` matrix at a time but requires O(P) communication complexity.
10124: Level: intermediate
10126: Notes:
10127: C will be created if `MAT_INITIAL_MATRIX` and must be destroyed by the user with `MatDestroy()`.
10129: `MAT_REUSE_MATRIX` can only be used if the matrices A and B have the same nonzero pattern as in the previous call
10131: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
10132: actually needed.
10134: This routine is currently only implemented for pairs of `MATSEQAIJ` matrices, for the `MATSEQDENSE` class,
10135: and for pairs of `MATMPIDENSE` matrices.
10137: This routine is shorthand for using `MatProductCreate()` with the `MatProductType` of `MATPRODUCT_ABt`
10139: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MATPRODUCT_ABt`, `MatMatMult()`, `MatTransposeMatMult()` `MatPtAP()`, `MatProductAlgorithm`, `MatProductType`
10140: @*/
10141: PetscErrorCode MatMatTransposeMult(Mat A, Mat B, MatReuse scall, PetscReal fill, Mat *C)
10142: {
10143: PetscFunctionBegin;
10144: PetscCall(MatProduct_Private(A, B, scall, fill, MATPRODUCT_ABt, C));
10145: if (A == B) PetscCall(MatSetOption(*C, MAT_SYMMETRIC, PETSC_TRUE));
10146: PetscFunctionReturn(PETSC_SUCCESS);
10147: }
10149: /*@
10150: MatTransposeMatMult - Performs matrix-matrix multiplication $C = A^T*B$.
10152: Neighbor-wise Collective
10154: Input Parameters:
10155: + A - the left matrix
10156: . B - the right matrix
10157: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10158: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use `PETSC_DEFAULT` if not known
10160: Output Parameter:
10161: . C - the product matrix
10163: Level: intermediate
10165: Notes:
10166: `C` will be created if `MAT_INITIAL_MATRIX` and must be destroyed by the user with `MatDestroy()`.
10168: `MAT_REUSE_MATRIX` can only be used if the matrices A and B have the same nonzero pattern as in the previous call.
10170: This routine is shorthand for using `MatProductCreate()` with the `MatProductType` of `MATPRODUCT_AtB`
10172: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
10173: actually needed.
10175: This routine is currently implemented for pairs of `MATAIJ` matrices and pairs of `MATSEQDENSE` matrices and classes
10176: which inherit from `MATSEQAIJ`. `C` will be of the same type as the input matrices.
10178: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MATPRODUCT_AtB`, `MatMatMult()`, `MatMatTransposeMult()`, `MatPtAP()`
10179: @*/
10180: PetscErrorCode MatTransposeMatMult(Mat A, Mat B, MatReuse scall, PetscReal fill, Mat *C)
10181: {
10182: PetscFunctionBegin;
10183: PetscCall(MatProduct_Private(A, B, scall, fill, MATPRODUCT_AtB, C));
10184: PetscFunctionReturn(PETSC_SUCCESS);
10185: }
10187: /*@
10188: MatMatMatMult - Performs matrix-matrix-matrix multiplication D=A*B*C.
10190: Neighbor-wise Collective
10192: Input Parameters:
10193: + A - the left matrix
10194: . B - the middle matrix
10195: . C - the right matrix
10196: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10197: - fill - expected fill as ratio of nnz(D)/(nnz(A) + nnz(B)+nnz(C)), use `PETSC_DEFAULT` if you do not have a good estimate
10198: if the result is a dense matrix this is irrelevant
10200: Output Parameter:
10201: . D - the product matrix
10203: Level: intermediate
10205: Notes:
10206: Unless `scall` is `MAT_REUSE_MATRIX` D will be created.
10208: `MAT_REUSE_MATRIX` can only be used if the matrices `A`, `B`, and `C` have the same nonzero pattern as in the previous call
10210: This routine is shorthand for using `MatProductCreate()` with the `MatProductType` of `MATPRODUCT_ABC`
10212: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
10213: actually needed.
10215: If you have many matrices with the same non-zero structure to multiply, you
10216: should use `MAT_REUSE_MATRIX` in all calls but the first
10218: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MATPRODUCT_ABC`, `MatMatMult`, `MatPtAP()`, `MatMatTransposeMult()`, `MatTransposeMatMult()`
10219: @*/
10220: PetscErrorCode MatMatMatMult(Mat A, Mat B, Mat C, MatReuse scall, PetscReal fill, Mat *D)
10221: {
10222: PetscFunctionBegin;
10223: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*D, 6);
10224: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
10226: if (scall == MAT_INITIAL_MATRIX) {
10227: PetscCall(MatProductCreate(A, B, C, D));
10228: PetscCall(MatProductSetType(*D, MATPRODUCT_ABC));
10229: PetscCall(MatProductSetAlgorithm(*D, "default"));
10230: PetscCall(MatProductSetFill(*D, fill));
10232: (*D)->product->api_user = PETSC_TRUE;
10233: PetscCall(MatProductSetFromOptions(*D));
10234: 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,
10235: ((PetscObject)C)->type_name);
10236: PetscCall(MatProductSymbolic(*D));
10237: } else { /* user may change input matrices when REUSE */
10238: PetscCall(MatProductReplaceMats(A, B, C, *D));
10239: }
10240: PetscCall(MatProductNumeric(*D));
10241: PetscFunctionReturn(PETSC_SUCCESS);
10242: }
10244: /*@
10245: MatCreateRedundantMatrix - Create redundant matrices and put them into processors of subcommunicators.
10247: Collective
10249: Input Parameters:
10250: + mat - the matrix
10251: . nsubcomm - the number of subcommunicators (= number of redundant parallel or sequential matrices)
10252: . subcomm - MPI communicator split from the communicator where mat resides in (or `MPI_COMM_NULL` if nsubcomm is used)
10253: - reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10255: Output Parameter:
10256: . matredundant - redundant matrix
10258: Level: advanced
10260: Notes:
10261: `MAT_REUSE_MATRIX` can only be used when the nonzero structure of the
10262: original matrix has not changed from that last call to `MatCreateRedundantMatrix()`.
10264: This routine creates the duplicated matrices in the subcommunicators; you should NOT create them before
10265: calling it.
10267: `PetscSubcommCreate()` can be used to manage the creation of the subcomm but need not be.
10269: .seealso: [](ch_matrices), `Mat`, `MatDestroy()`, `PetscSubcommCreate()`, `PetscSubcomm`
10270: @*/
10271: PetscErrorCode MatCreateRedundantMatrix(Mat mat, PetscInt nsubcomm, MPI_Comm subcomm, MatReuse reuse, Mat *matredundant)
10272: {
10273: MPI_Comm comm;
10274: PetscMPIInt size;
10275: PetscInt mloc_sub, nloc_sub, rstart, rend, M = mat->rmap->N, N = mat->cmap->N, bs = mat->rmap->bs;
10276: Mat_Redundant *redund = NULL;
10277: PetscSubcomm psubcomm = NULL;
10278: MPI_Comm subcomm_in = subcomm;
10279: Mat *matseq;
10280: IS isrow, iscol;
10281: PetscBool newsubcomm = PETSC_FALSE;
10283: PetscFunctionBegin;
10285: if (nsubcomm && reuse == MAT_REUSE_MATRIX) {
10286: PetscAssertPointer(*matredundant, 5);
10288: }
10290: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
10291: if (size == 1 || nsubcomm == 1) {
10292: if (reuse == MAT_INITIAL_MATRIX) {
10293: PetscCall(MatDuplicate(mat, MAT_COPY_VALUES, matredundant));
10294: } else {
10295: 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");
10296: PetscCall(MatCopy(mat, *matredundant, SAME_NONZERO_PATTERN));
10297: }
10298: PetscFunctionReturn(PETSC_SUCCESS);
10299: }
10301: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10302: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10303: MatCheckPreallocated(mat, 1);
10305: PetscCall(PetscLogEventBegin(MAT_RedundantMat, mat, 0, 0, 0));
10306: if (subcomm_in == MPI_COMM_NULL && reuse == MAT_INITIAL_MATRIX) { /* get subcomm if user does not provide subcomm */
10307: /* create psubcomm, then get subcomm */
10308: PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
10309: PetscCallMPI(MPI_Comm_size(comm, &size));
10310: PetscCheck(nsubcomm >= 1 && nsubcomm <= size, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "nsubcomm must between 1 and %d", size);
10312: PetscCall(PetscSubcommCreate(comm, &psubcomm));
10313: PetscCall(PetscSubcommSetNumber(psubcomm, nsubcomm));
10314: PetscCall(PetscSubcommSetType(psubcomm, PETSC_SUBCOMM_CONTIGUOUS));
10315: PetscCall(PetscSubcommSetFromOptions(psubcomm));
10316: PetscCall(PetscCommDuplicate(PetscSubcommChild(psubcomm), &subcomm, NULL));
10317: newsubcomm = PETSC_TRUE;
10318: PetscCall(PetscSubcommDestroy(&psubcomm));
10319: }
10321: /* get isrow, iscol and a local sequential matrix matseq[0] */
10322: if (reuse == MAT_INITIAL_MATRIX) {
10323: mloc_sub = PETSC_DECIDE;
10324: nloc_sub = PETSC_DECIDE;
10325: if (bs < 1) {
10326: PetscCall(PetscSplitOwnership(subcomm, &mloc_sub, &M));
10327: PetscCall(PetscSplitOwnership(subcomm, &nloc_sub, &N));
10328: } else {
10329: PetscCall(PetscSplitOwnershipBlock(subcomm, bs, &mloc_sub, &M));
10330: PetscCall(PetscSplitOwnershipBlock(subcomm, bs, &nloc_sub, &N));
10331: }
10332: PetscCallMPI(MPI_Scan(&mloc_sub, &rend, 1, MPIU_INT, MPI_SUM, subcomm));
10333: rstart = rend - mloc_sub;
10334: PetscCall(ISCreateStride(PETSC_COMM_SELF, mloc_sub, rstart, 1, &isrow));
10335: PetscCall(ISCreateStride(PETSC_COMM_SELF, N, 0, 1, &iscol));
10336: PetscCall(ISSetIdentity(iscol));
10337: } else { /* reuse == MAT_REUSE_MATRIX */
10338: 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");
10339: /* retrieve subcomm */
10340: PetscCall(PetscObjectGetComm((PetscObject)(*matredundant), &subcomm));
10341: redund = (*matredundant)->redundant;
10342: isrow = redund->isrow;
10343: iscol = redund->iscol;
10344: matseq = redund->matseq;
10345: }
10346: PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscol, reuse, &matseq));
10348: /* get matredundant over subcomm */
10349: if (reuse == MAT_INITIAL_MATRIX) {
10350: PetscCall(MatCreateMPIMatConcatenateSeqMat(subcomm, matseq[0], nloc_sub, reuse, matredundant));
10352: /* create a supporting struct and attach it to C for reuse */
10353: PetscCall(PetscNew(&redund));
10354: (*matredundant)->redundant = redund;
10355: redund->isrow = isrow;
10356: redund->iscol = iscol;
10357: redund->matseq = matseq;
10358: if (newsubcomm) {
10359: redund->subcomm = subcomm;
10360: } else {
10361: redund->subcomm = MPI_COMM_NULL;
10362: }
10363: } else {
10364: PetscCall(MatCreateMPIMatConcatenateSeqMat(subcomm, matseq[0], PETSC_DECIDE, reuse, matredundant));
10365: }
10366: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
10367: if (matseq[0]->boundtocpu && matseq[0]->bindingpropagates) {
10368: PetscCall(MatBindToCPU(*matredundant, PETSC_TRUE));
10369: PetscCall(MatSetBindingPropagates(*matredundant, PETSC_TRUE));
10370: }
10371: #endif
10372: PetscCall(PetscLogEventEnd(MAT_RedundantMat, mat, 0, 0, 0));
10373: PetscFunctionReturn(PETSC_SUCCESS);
10374: }
10376: /*@C
10377: MatGetMultiProcBlock - Create multiple 'parallel submatrices' from
10378: a given `Mat`. Each submatrix can span multiple procs.
10380: Collective
10382: Input Parameters:
10383: + mat - the matrix
10384: . subComm - the sub communicator obtained as if by `MPI_Comm_split(PetscObjectComm((PetscObject)mat))`
10385: - scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10387: Output Parameter:
10388: . subMat - parallel sub-matrices each spanning a given `subcomm`
10390: Level: advanced
10392: Notes:
10393: The submatrix partition across processors is dictated by `subComm` a
10394: communicator obtained by `MPI_comm_split()` or via `PetscSubcommCreate()`. The `subComm`
10395: is not restricted to be grouped with consecutive original MPI processes.
10397: Due the `MPI_Comm_split()` usage, the parallel layout of the submatrices
10398: map directly to the layout of the original matrix [wrt the local
10399: row,col partitioning]. So the original 'DiagonalMat' naturally maps
10400: into the 'DiagonalMat' of the `subMat`, hence it is used directly from
10401: the `subMat`. However the offDiagMat looses some columns - and this is
10402: reconstructed with `MatSetValues()`
10404: This is used by `PCBJACOBI` when a single block spans multiple MPI processes.
10406: .seealso: [](ch_matrices), `Mat`, `MatCreateRedundantMatrix()`, `MatCreateSubMatrices()`, `PCBJACOBI`
10407: @*/
10408: PetscErrorCode MatGetMultiProcBlock(Mat mat, MPI_Comm subComm, MatReuse scall, Mat *subMat)
10409: {
10410: PetscMPIInt commsize, subCommSize;
10412: PetscFunctionBegin;
10413: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &commsize));
10414: PetscCallMPI(MPI_Comm_size(subComm, &subCommSize));
10415: PetscCheck(subCommSize <= commsize, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "CommSize %d < SubCommZize %d", commsize, subCommSize);
10417: 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");
10418: PetscCall(PetscLogEventBegin(MAT_GetMultiProcBlock, mat, 0, 0, 0));
10419: PetscUseTypeMethod(mat, getmultiprocblock, subComm, scall, subMat);
10420: PetscCall(PetscLogEventEnd(MAT_GetMultiProcBlock, mat, 0, 0, 0));
10421: PetscFunctionReturn(PETSC_SUCCESS);
10422: }
10424: /*@
10425: MatGetLocalSubMatrix - Gets a reference to a submatrix specified in local numbering
10427: Not Collective
10429: Input Parameters:
10430: + mat - matrix to extract local submatrix from
10431: . isrow - local row indices for submatrix
10432: - iscol - local column indices for submatrix
10434: Output Parameter:
10435: . submat - the submatrix
10437: Level: intermediate
10439: Notes:
10440: `submat` should be disposed of with `MatRestoreLocalSubMatrix()`.
10442: Depending on the format of `mat`, the returned `submat` may not implement `MatMult()`. Its communicator may be
10443: the same as `mat`, it may be `PETSC_COMM_SELF`, or some other sub-communictor of `mat`'s.
10445: `submat` always implements `MatSetValuesLocal()`. If `isrow` and `iscol` have the same block size, then
10446: `MatSetValuesBlockedLocal()` will also be implemented.
10448: `mat` must have had a `ISLocalToGlobalMapping` provided to it with `MatSetLocalToGlobalMapping()`.
10449: Matrices obtained with `DMCreateMatrix()` generally already have the local to global mapping provided.
10451: .seealso: [](ch_matrices), `Mat`, `MatRestoreLocalSubMatrix()`, `MatCreateLocalRef()`, `MatSetLocalToGlobalMapping()`
10452: @*/
10453: PetscErrorCode MatGetLocalSubMatrix(Mat mat, IS isrow, IS iscol, Mat *submat)
10454: {
10455: PetscFunctionBegin;
10459: PetscCheckSameComm(isrow, 2, iscol, 3);
10460: PetscAssertPointer(submat, 4);
10461: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Matrix must have local to global mapping provided before this call");
10463: if (mat->ops->getlocalsubmatrix) {
10464: PetscUseTypeMethod(mat, getlocalsubmatrix, isrow, iscol, submat);
10465: } else {
10466: PetscCall(MatCreateLocalRef(mat, isrow, iscol, submat));
10467: }
10468: PetscFunctionReturn(PETSC_SUCCESS);
10469: }
10471: /*@
10472: MatRestoreLocalSubMatrix - Restores a reference to a submatrix specified in local numbering obtained with `MatGetLocalSubMatrix()`
10474: Not Collective
10476: Input Parameters:
10477: + mat - matrix to extract local submatrix from
10478: . isrow - local row indices for submatrix
10479: . iscol - local column indices for submatrix
10480: - submat - the submatrix
10482: Level: intermediate
10484: .seealso: [](ch_matrices), `Mat`, `MatGetLocalSubMatrix()`
10485: @*/
10486: PetscErrorCode MatRestoreLocalSubMatrix(Mat mat, IS isrow, IS iscol, Mat *submat)
10487: {
10488: PetscFunctionBegin;
10492: PetscCheckSameComm(isrow, 2, iscol, 3);
10493: PetscAssertPointer(submat, 4);
10496: if (mat->ops->restorelocalsubmatrix) {
10497: PetscUseTypeMethod(mat, restorelocalsubmatrix, isrow, iscol, submat);
10498: } else {
10499: PetscCall(MatDestroy(submat));
10500: }
10501: *submat = NULL;
10502: PetscFunctionReturn(PETSC_SUCCESS);
10503: }
10505: /*@
10506: MatFindZeroDiagonals - Finds all the rows of a matrix that have zero or no diagonal entry in the matrix
10508: Collective
10510: Input Parameter:
10511: . mat - the matrix
10513: Output Parameter:
10514: . is - if any rows have zero diagonals this contains the list of them
10516: Level: developer
10518: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
10519: @*/
10520: PetscErrorCode MatFindZeroDiagonals(Mat mat, IS *is)
10521: {
10522: PetscFunctionBegin;
10525: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10526: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10528: if (!mat->ops->findzerodiagonals) {
10529: Vec diag;
10530: const PetscScalar *a;
10531: PetscInt *rows;
10532: PetscInt rStart, rEnd, r, nrow = 0;
10534: PetscCall(MatCreateVecs(mat, &diag, NULL));
10535: PetscCall(MatGetDiagonal(mat, diag));
10536: PetscCall(MatGetOwnershipRange(mat, &rStart, &rEnd));
10537: PetscCall(VecGetArrayRead(diag, &a));
10538: for (r = 0; r < rEnd - rStart; ++r)
10539: if (a[r] == 0.0) ++nrow;
10540: PetscCall(PetscMalloc1(nrow, &rows));
10541: nrow = 0;
10542: for (r = 0; r < rEnd - rStart; ++r)
10543: if (a[r] == 0.0) rows[nrow++] = r + rStart;
10544: PetscCall(VecRestoreArrayRead(diag, &a));
10545: PetscCall(VecDestroy(&diag));
10546: PetscCall(ISCreateGeneral(PetscObjectComm((PetscObject)mat), nrow, rows, PETSC_OWN_POINTER, is));
10547: } else {
10548: PetscUseTypeMethod(mat, findzerodiagonals, is);
10549: }
10550: PetscFunctionReturn(PETSC_SUCCESS);
10551: }
10553: /*@
10554: MatFindOffBlockDiagonalEntries - Finds all the rows of a matrix that have entries outside of the main diagonal block (defined by the matrix block size)
10556: Collective
10558: Input Parameter:
10559: . mat - the matrix
10561: Output Parameter:
10562: . is - contains the list of rows with off block diagonal entries
10564: Level: developer
10566: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
10567: @*/
10568: PetscErrorCode MatFindOffBlockDiagonalEntries(Mat mat, IS *is)
10569: {
10570: PetscFunctionBegin;
10573: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10574: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10576: PetscUseTypeMethod(mat, findoffblockdiagonalentries, is);
10577: PetscFunctionReturn(PETSC_SUCCESS);
10578: }
10580: /*@C
10581: MatInvertBlockDiagonal - Inverts the block diagonal entries.
10583: Collective; No Fortran Support
10585: Input Parameter:
10586: . mat - the matrix
10588: Output Parameter:
10589: . values - the block inverses in column major order (FORTRAN-like)
10591: Level: advanced
10593: Notes:
10594: The size of the blocks is determined by the block size of the matrix.
10596: The blocks never overlap between two MPI processes, use `MatInvertVariableBlockEnvelope()` for that case
10598: The blocks all have the same size, use `MatInvertVariableBlockDiagonal()` for variable block size
10600: .seealso: [](ch_matrices), `Mat`, `MatInvertVariableBlockEnvelope()`, `MatInvertBlockDiagonalMat()`
10601: @*/
10602: PetscErrorCode MatInvertBlockDiagonal(Mat mat, const PetscScalar **values)
10603: {
10604: PetscFunctionBegin;
10606: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10607: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10608: PetscUseTypeMethod(mat, invertblockdiagonal, values);
10609: PetscFunctionReturn(PETSC_SUCCESS);
10610: }
10612: /*@C
10613: MatInvertVariableBlockDiagonal - Inverts the point block diagonal entries.
10615: Collective; No Fortran Support
10617: Input Parameters:
10618: + mat - the matrix
10619: . nblocks - the number of blocks on the process, set with `MatSetVariableBlockSizes()`
10620: - bsizes - the size of each block on the process, set with `MatSetVariableBlockSizes()`
10622: Output Parameter:
10623: . values - the block inverses in column major order (FORTRAN-like)
10625: Level: advanced
10627: Notes:
10628: Use `MatInvertBlockDiagonal()` if all blocks have the same size
10630: The blocks never overlap between two MPI processes, use `MatInvertVariableBlockEnvelope()` for that case
10632: .seealso: [](ch_matrices), `Mat`, `MatInvertBlockDiagonal()`, `MatSetVariableBlockSizes()`, `MatInvertVariableBlockEnvelope()`
10633: @*/
10634: PetscErrorCode MatInvertVariableBlockDiagonal(Mat mat, PetscInt nblocks, const PetscInt *bsizes, PetscScalar *values)
10635: {
10636: PetscFunctionBegin;
10638: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10639: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10640: PetscUseTypeMethod(mat, invertvariableblockdiagonal, nblocks, bsizes, values);
10641: PetscFunctionReturn(PETSC_SUCCESS);
10642: }
10644: /*@
10645: MatInvertBlockDiagonalMat - set the values of matrix C to be the inverted block diagonal of matrix A
10647: Collective
10649: Input Parameters:
10650: + A - the matrix
10651: - C - matrix with inverted block diagonal of `A`. This matrix should be created and may have its type set.
10653: Level: advanced
10655: Note:
10656: The blocksize of the matrix is used to determine the blocks on the diagonal of `C`
10658: .seealso: [](ch_matrices), `Mat`, `MatInvertBlockDiagonal()`
10659: @*/
10660: PetscErrorCode MatInvertBlockDiagonalMat(Mat A, Mat C)
10661: {
10662: const PetscScalar *vals;
10663: PetscInt *dnnz;
10664: PetscInt m, rstart, rend, bs, i, j;
10666: PetscFunctionBegin;
10667: PetscCall(MatInvertBlockDiagonal(A, &vals));
10668: PetscCall(MatGetBlockSize(A, &bs));
10669: PetscCall(MatGetLocalSize(A, &m, NULL));
10670: PetscCall(MatSetLayouts(C, A->rmap, A->cmap));
10671: PetscCall(PetscMalloc1(m / bs, &dnnz));
10672: for (j = 0; j < m / bs; j++) dnnz[j] = 1;
10673: PetscCall(MatXAIJSetPreallocation(C, bs, dnnz, NULL, NULL, NULL));
10674: PetscCall(PetscFree(dnnz));
10675: PetscCall(MatGetOwnershipRange(C, &rstart, &rend));
10676: PetscCall(MatSetOption(C, MAT_ROW_ORIENTED, PETSC_FALSE));
10677: for (i = rstart / bs; i < rend / bs; i++) PetscCall(MatSetValuesBlocked(C, 1, &i, 1, &i, &vals[(i - rstart / bs) * bs * bs], INSERT_VALUES));
10678: PetscCall(MatAssemblyBegin(C, MAT_FINAL_ASSEMBLY));
10679: PetscCall(MatAssemblyEnd(C, MAT_FINAL_ASSEMBLY));
10680: PetscCall(MatSetOption(C, MAT_ROW_ORIENTED, PETSC_TRUE));
10681: PetscFunctionReturn(PETSC_SUCCESS);
10682: }
10684: /*@C
10685: MatTransposeColoringDestroy - Destroys a coloring context for matrix product $C = A*B^T$ that was created
10686: via `MatTransposeColoringCreate()`.
10688: Collective
10690: Input Parameter:
10691: . c - coloring context
10693: Level: intermediate
10695: .seealso: [](ch_matrices), `Mat`, `MatTransposeColoringCreate()`
10696: @*/
10697: PetscErrorCode MatTransposeColoringDestroy(MatTransposeColoring *c)
10698: {
10699: MatTransposeColoring matcolor = *c;
10701: PetscFunctionBegin;
10702: if (!matcolor) PetscFunctionReturn(PETSC_SUCCESS);
10703: if (--((PetscObject)matcolor)->refct > 0) {
10704: matcolor = NULL;
10705: PetscFunctionReturn(PETSC_SUCCESS);
10706: }
10708: PetscCall(PetscFree3(matcolor->ncolumns, matcolor->nrows, matcolor->colorforrow));
10709: PetscCall(PetscFree(matcolor->rows));
10710: PetscCall(PetscFree(matcolor->den2sp));
10711: PetscCall(PetscFree(matcolor->colorforcol));
10712: PetscCall(PetscFree(matcolor->columns));
10713: if (matcolor->brows > 0) PetscCall(PetscFree(matcolor->lstart));
10714: PetscCall(PetscHeaderDestroy(c));
10715: PetscFunctionReturn(PETSC_SUCCESS);
10716: }
10718: /*@C
10719: MatTransColoringApplySpToDen - Given a symbolic matrix product $C = A*B^T$ for which
10720: a `MatTransposeColoring` context has been created, computes a dense $B^T$ by applying
10721: `MatTransposeColoring` to sparse `B`.
10723: Collective
10725: Input Parameters:
10726: + coloring - coloring context created with `MatTransposeColoringCreate()`
10727: - B - sparse matrix
10729: Output Parameter:
10730: . Btdense - dense matrix $B^T$
10732: Level: developer
10734: Note:
10735: These are used internally for some implementations of `MatRARt()`
10737: .seealso: [](ch_matrices), `Mat`, `MatTransposeColoringCreate()`, `MatTransposeColoringDestroy()`, `MatTransColoringApplyDenToSp()`
10738: @*/
10739: PetscErrorCode MatTransColoringApplySpToDen(MatTransposeColoring coloring, Mat B, Mat Btdense)
10740: {
10741: PetscFunctionBegin;
10746: PetscCall((*B->ops->transcoloringapplysptoden)(coloring, B, Btdense));
10747: PetscFunctionReturn(PETSC_SUCCESS);
10748: }
10750: /*@C
10751: MatTransColoringApplyDenToSp - Given a symbolic matrix product $C_{sp} = A*B^T$ for which
10752: a `MatTransposeColoring` context has been created and a dense matrix $C_{den} = A*B^T_{dense}$
10753: in which `B^T_{dens}` is obtained from `MatTransColoringApplySpToDen()`, recover sparse matrix
10754: $C_{sp}$ from $C_{den}$.
10756: Collective
10758: Input Parameters:
10759: + matcoloring - coloring context created with `MatTransposeColoringCreate()`
10760: - Cden - matrix product of a sparse matrix and a dense matrix Btdense
10762: Output Parameter:
10763: . Csp - sparse matrix
10765: Level: developer
10767: Note:
10768: These are used internally for some implementations of `MatRARt()`
10770: .seealso: [](ch_matrices), `Mat`, `MatTransposeColoringCreate()`, `MatTransposeColoringDestroy()`, `MatTransColoringApplySpToDen()`
10771: @*/
10772: PetscErrorCode MatTransColoringApplyDenToSp(MatTransposeColoring matcoloring, Mat Cden, Mat Csp)
10773: {
10774: PetscFunctionBegin;
10779: PetscCall((*Csp->ops->transcoloringapplydentosp)(matcoloring, Cden, Csp));
10780: PetscCall(MatAssemblyBegin(Csp, MAT_FINAL_ASSEMBLY));
10781: PetscCall(MatAssemblyEnd(Csp, MAT_FINAL_ASSEMBLY));
10782: PetscFunctionReturn(PETSC_SUCCESS);
10783: }
10785: /*@C
10786: MatTransposeColoringCreate - Creates a matrix coloring context for the matrix product $C = A*B^T$.
10788: Collective
10790: Input Parameters:
10791: + mat - the matrix product C
10792: - iscoloring - the coloring of the matrix; usually obtained with `MatColoringCreate()` or `DMCreateColoring()`
10794: Output Parameter:
10795: . color - the new coloring context
10797: Level: intermediate
10799: .seealso: [](ch_matrices), `Mat`, `MatTransposeColoringDestroy()`, `MatTransColoringApplySpToDen()`,
10800: `MatTransColoringApplyDenToSp()`
10801: @*/
10802: PetscErrorCode MatTransposeColoringCreate(Mat mat, ISColoring iscoloring, MatTransposeColoring *color)
10803: {
10804: MatTransposeColoring c;
10805: MPI_Comm comm;
10807: PetscFunctionBegin;
10808: PetscCall(PetscLogEventBegin(MAT_TransposeColoringCreate, mat, 0, 0, 0));
10809: PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
10810: PetscCall(PetscHeaderCreate(c, MAT_TRANSPOSECOLORING_CLASSID, "MatTransposeColoring", "Matrix product C=A*B^T via coloring", "Mat", comm, MatTransposeColoringDestroy, NULL));
10812: c->ctype = iscoloring->ctype;
10813: PetscUseTypeMethod(mat, transposecoloringcreate, iscoloring, c);
10815: *color = c;
10816: PetscCall(PetscLogEventEnd(MAT_TransposeColoringCreate, mat, 0, 0, 0));
10817: PetscFunctionReturn(PETSC_SUCCESS);
10818: }
10820: /*@
10821: MatGetNonzeroState - Returns a 64-bit integer representing the current state of nonzeros in the matrix. If the
10822: matrix has had no new nonzero locations added to (or removed from) the matrix since the previous call then the value will be the
10823: same, otherwise it will be larger
10825: Not Collective
10827: Input Parameter:
10828: . mat - the matrix
10830: Output Parameter:
10831: . state - the current state
10833: Level: intermediate
10835: Notes:
10836: You can only compare states from two different calls to the SAME matrix, you cannot compare calls between
10837: different matrices
10839: Use `PetscObjectStateGet()` to check for changes to the numerical values in a matrix
10841: Use the result of `PetscObjectGetId()` to compare if a previously checked matrix is the same as the current matrix, do not compare object pointers.
10843: .seealso: [](ch_matrices), `Mat`, `PetscObjectStateGet()`, `PetscObjectGetId()`
10844: @*/
10845: PetscErrorCode MatGetNonzeroState(Mat mat, PetscObjectState *state)
10846: {
10847: PetscFunctionBegin;
10849: *state = mat->nonzerostate;
10850: PetscFunctionReturn(PETSC_SUCCESS);
10851: }
10853: /*@
10854: MatCreateMPIMatConcatenateSeqMat - Creates a single large PETSc matrix by concatenating sequential
10855: matrices from each processor
10857: Collective
10859: Input Parameters:
10860: + comm - the communicators the parallel matrix will live on
10861: . seqmat - the input sequential matrices
10862: . n - number of local columns (or `PETSC_DECIDE`)
10863: - reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10865: Output Parameter:
10866: . mpimat - the parallel matrix generated
10868: Level: developer
10870: Note:
10871: The number of columns of the matrix in EACH processor MUST be the same.
10873: .seealso: [](ch_matrices), `Mat`
10874: @*/
10875: PetscErrorCode MatCreateMPIMatConcatenateSeqMat(MPI_Comm comm, Mat seqmat, PetscInt n, MatReuse reuse, Mat *mpimat)
10876: {
10877: PetscMPIInt size;
10879: PetscFunctionBegin;
10880: PetscCallMPI(MPI_Comm_size(comm, &size));
10881: if (size == 1) {
10882: if (reuse == MAT_INITIAL_MATRIX) {
10883: PetscCall(MatDuplicate(seqmat, MAT_COPY_VALUES, mpimat));
10884: } else {
10885: PetscCall(MatCopy(seqmat, *mpimat, SAME_NONZERO_PATTERN));
10886: }
10887: PetscFunctionReturn(PETSC_SUCCESS);
10888: }
10890: 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");
10892: PetscCall(PetscLogEventBegin(MAT_Merge, seqmat, 0, 0, 0));
10893: PetscCall((*seqmat->ops->creatempimatconcatenateseqmat)(comm, seqmat, n, reuse, mpimat));
10894: PetscCall(PetscLogEventEnd(MAT_Merge, seqmat, 0, 0, 0));
10895: PetscFunctionReturn(PETSC_SUCCESS);
10896: }
10898: /*@
10899: MatSubdomainsCreateCoalesce - Creates index subdomains by coalescing adjacent MPI processes' ownership ranges.
10901: Collective
10903: Input Parameters:
10904: + A - the matrix to create subdomains from
10905: - N - requested number of subdomains
10907: Output Parameters:
10908: + n - number of subdomains resulting on this MPI process
10909: - iss - `IS` list with indices of subdomains on this MPI process
10911: Level: advanced
10913: Note:
10914: The number of subdomains must be smaller than the communicator size
10916: .seealso: [](ch_matrices), `Mat`, `IS`
10917: @*/
10918: PetscErrorCode MatSubdomainsCreateCoalesce(Mat A, PetscInt N, PetscInt *n, IS *iss[])
10919: {
10920: MPI_Comm comm, subcomm;
10921: PetscMPIInt size, rank, color;
10922: PetscInt rstart, rend, k;
10924: PetscFunctionBegin;
10925: PetscCall(PetscObjectGetComm((PetscObject)A, &comm));
10926: PetscCallMPI(MPI_Comm_size(comm, &size));
10927: PetscCallMPI(MPI_Comm_rank(comm, &rank));
10928: PetscCheck(N >= 1 && N < (PetscInt)size, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "number of subdomains must be > 0 and < %d, got N = %" PetscInt_FMT, size, N);
10929: *n = 1;
10930: k = ((PetscInt)size) / N + ((PetscInt)size % N > 0); /* There are up to k ranks to a color */
10931: color = rank / k;
10932: PetscCallMPI(MPI_Comm_split(comm, color, rank, &subcomm));
10933: PetscCall(PetscMalloc1(1, iss));
10934: PetscCall(MatGetOwnershipRange(A, &rstart, &rend));
10935: PetscCall(ISCreateStride(subcomm, rend - rstart, rstart, 1, iss[0]));
10936: PetscCallMPI(MPI_Comm_free(&subcomm));
10937: PetscFunctionReturn(PETSC_SUCCESS);
10938: }
10940: /*@
10941: MatGalerkin - Constructs the coarse grid problem matrix via Galerkin projection.
10943: If the interpolation and restriction operators are the same, uses `MatPtAP()`.
10944: If they are not the same, uses `MatMatMatMult()`.
10946: Once the coarse grid problem is constructed, correct for interpolation operators
10947: that are not of full rank, which can legitimately happen in the case of non-nested
10948: geometric multigrid.
10950: Input Parameters:
10951: + restrct - restriction operator
10952: . dA - fine grid matrix
10953: . interpolate - interpolation operator
10954: . reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10955: - fill - expected fill, use `PETSC_DEFAULT` if you do not have a good estimate
10957: Output Parameter:
10958: . A - the Galerkin coarse matrix
10960: Options Database Key:
10961: . -pc_mg_galerkin <both,pmat,mat,none> - for what matrices the Galerkin process should be used
10963: Level: developer
10965: .seealso: [](ch_matrices), `Mat`, `MatPtAP()`, `MatMatMatMult()`
10966: @*/
10967: PetscErrorCode MatGalerkin(Mat restrct, Mat dA, Mat interpolate, MatReuse reuse, PetscReal fill, Mat *A)
10968: {
10969: IS zerorows;
10970: Vec diag;
10972: PetscFunctionBegin;
10973: PetscCheck(reuse != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
10974: /* Construct the coarse grid matrix */
10975: if (interpolate == restrct) {
10976: PetscCall(MatPtAP(dA, interpolate, reuse, fill, A));
10977: } else {
10978: PetscCall(MatMatMatMult(restrct, dA, interpolate, reuse, fill, A));
10979: }
10981: /* If the interpolation matrix is not of full rank, A will have zero rows.
10982: This can legitimately happen in the case of non-nested geometric multigrid.
10983: In that event, we set the rows of the matrix to the rows of the identity,
10984: ignoring the equations (as the RHS will also be zero). */
10986: PetscCall(MatFindZeroRows(*A, &zerorows));
10988: if (zerorows != NULL) { /* if there are any zero rows */
10989: PetscCall(MatCreateVecs(*A, &diag, NULL));
10990: PetscCall(MatGetDiagonal(*A, diag));
10991: PetscCall(VecISSet(diag, zerorows, 1.0));
10992: PetscCall(MatDiagonalSet(*A, diag, INSERT_VALUES));
10993: PetscCall(VecDestroy(&diag));
10994: PetscCall(ISDestroy(&zerorows));
10995: }
10996: PetscFunctionReturn(PETSC_SUCCESS);
10997: }
10999: /*@C
11000: MatSetOperation - Allows user to set a matrix operation for any matrix type
11002: Logically Collective
11004: Input Parameters:
11005: + mat - the matrix
11006: . op - the name of the operation
11007: - f - the function that provides the operation
11009: Level: developer
11011: Example Usage:
11012: .vb
11013: extern PetscErrorCode usermult(Mat, Vec, Vec);
11015: PetscCall(MatCreateXXX(comm, ..., &A));
11016: PetscCall(MatSetOperation(A, MATOP_MULT, (PetscVoidFunction)usermult));
11017: .ve
11019: Notes:
11020: See the file `include/petscmat.h` for a complete list of matrix
11021: operations, which all have the form MATOP_<OPERATION>, where
11022: <OPERATION> is the name (in all capital letters) of the
11023: user interface routine (e.g., `MatMult()` -> `MATOP_MULT`).
11025: All user-provided functions (except for `MATOP_DESTROY`) should have the same calling
11026: sequence as the usual matrix interface routines, since they
11027: are intended to be accessed via the usual matrix interface
11028: routines, e.g.,
11029: .vb
11030: MatMult(Mat, Vec, Vec) -> usermult(Mat, Vec, Vec)
11031: .ve
11033: In particular each function MUST return `PETSC_SUCCESS` on success and
11034: nonzero on failure.
11036: This routine is distinct from `MatShellSetOperation()` in that it can be called on any matrix type.
11038: .seealso: [](ch_matrices), `Mat`, `MatGetOperation()`, `MatCreateShell()`, `MatShellSetContext()`, `MatShellSetOperation()`
11039: @*/
11040: PetscErrorCode MatSetOperation(Mat mat, MatOperation op, void (*f)(void))
11041: {
11042: PetscFunctionBegin;
11044: if (op == MATOP_VIEW && !mat->ops->viewnative && f != (void (*)(void))(mat->ops->view)) mat->ops->viewnative = mat->ops->view;
11045: (((void (**)(void))mat->ops)[op]) = f;
11046: PetscFunctionReturn(PETSC_SUCCESS);
11047: }
11049: /*@C
11050: MatGetOperation - Gets a matrix operation for any matrix type.
11052: Not Collective
11054: Input Parameters:
11055: + mat - the matrix
11056: - op - the name of the operation
11058: Output Parameter:
11059: . f - the function that provides the operation
11061: Level: developer
11063: Example Usage:
11064: .vb
11065: PetscErrorCode (*usermult)(Mat, Vec, Vec);
11067: MatGetOperation(A, MATOP_MULT, (void (**)(void))&usermult);
11068: .ve
11070: Notes:
11071: See the file include/petscmat.h for a complete list of matrix
11072: operations, which all have the form MATOP_<OPERATION>, where
11073: <OPERATION> is the name (in all capital letters) of the
11074: user interface routine (e.g., `MatMult()` -> `MATOP_MULT`).
11076: This routine is distinct from `MatShellGetOperation()` in that it can be called on any matrix type.
11078: .seealso: [](ch_matrices), `Mat`, `MatSetOperation()`, `MatCreateShell()`, `MatShellGetContext()`, `MatShellGetOperation()`
11079: @*/
11080: PetscErrorCode MatGetOperation(Mat mat, MatOperation op, void (**f)(void))
11081: {
11082: PetscFunctionBegin;
11084: *f = (((void (**)(void))mat->ops)[op]);
11085: PetscFunctionReturn(PETSC_SUCCESS);
11086: }
11088: /*@
11089: MatHasOperation - Determines whether the given matrix supports the particular operation.
11091: Not Collective
11093: Input Parameters:
11094: + mat - the matrix
11095: - op - the operation, for example, `MATOP_GET_DIAGONAL`
11097: Output Parameter:
11098: . has - either `PETSC_TRUE` or `PETSC_FALSE`
11100: Level: advanced
11102: Note:
11103: See `MatSetOperation()` for additional discussion on naming convention and usage of `op`.
11105: .seealso: [](ch_matrices), `Mat`, `MatCreateShell()`, `MatGetOperation()`, `MatSetOperation()`
11106: @*/
11107: PetscErrorCode MatHasOperation(Mat mat, MatOperation op, PetscBool *has)
11108: {
11109: PetscFunctionBegin;
11111: PetscAssertPointer(has, 3);
11112: if (mat->ops->hasoperation) {
11113: PetscUseTypeMethod(mat, hasoperation, op, has);
11114: } else {
11115: if (((void **)mat->ops)[op]) *has = PETSC_TRUE;
11116: else {
11117: *has = PETSC_FALSE;
11118: if (op == MATOP_CREATE_SUBMATRIX) {
11119: PetscMPIInt size;
11121: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
11122: if (size == 1) PetscCall(MatHasOperation(mat, MATOP_CREATE_SUBMATRICES, has));
11123: }
11124: }
11125: }
11126: PetscFunctionReturn(PETSC_SUCCESS);
11127: }
11129: /*@
11130: MatHasCongruentLayouts - Determines whether the rows and columns layouts of the matrix are congruent
11132: Collective
11134: Input Parameter:
11135: . mat - the matrix
11137: Output Parameter:
11138: . cong - either `PETSC_TRUE` or `PETSC_FALSE`
11140: Level: beginner
11142: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatSetSizes()`, `PetscLayout`
11143: @*/
11144: PetscErrorCode MatHasCongruentLayouts(Mat mat, PetscBool *cong)
11145: {
11146: PetscFunctionBegin;
11149: PetscAssertPointer(cong, 2);
11150: if (!mat->rmap || !mat->cmap) {
11151: *cong = mat->rmap == mat->cmap ? PETSC_TRUE : PETSC_FALSE;
11152: PetscFunctionReturn(PETSC_SUCCESS);
11153: }
11154: if (mat->congruentlayouts == PETSC_DECIDE) { /* first time we compare rows and cols layouts */
11155: PetscCall(PetscLayoutSetUp(mat->rmap));
11156: PetscCall(PetscLayoutSetUp(mat->cmap));
11157: PetscCall(PetscLayoutCompare(mat->rmap, mat->cmap, cong));
11158: if (*cong) mat->congruentlayouts = 1;
11159: else mat->congruentlayouts = 0;
11160: } else *cong = mat->congruentlayouts ? PETSC_TRUE : PETSC_FALSE;
11161: PetscFunctionReturn(PETSC_SUCCESS);
11162: }
11164: PetscErrorCode MatSetInf(Mat A)
11165: {
11166: PetscFunctionBegin;
11167: PetscUseTypeMethod(A, setinf);
11168: PetscFunctionReturn(PETSC_SUCCESS);
11169: }
11171: /*@C
11172: 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
11173: and possibly removes small values from the graph structure.
11175: Collective
11177: Input Parameters:
11178: + A - the matrix
11179: . sym - `PETSC_TRUE` indicates that the graph should be symmetrized
11180: . scale - `PETSC_TRUE` indicates that the graph edge weights should be symmetrically scaled with the diagonal entry
11181: - filter - filter value - < 0: does nothing; == 0: removes only 0.0 entries; otherwise: removes entries with abs(entries) <= value
11183: Output Parameter:
11184: . graph - the resulting graph
11186: Level: advanced
11188: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `PCGAMG`
11189: @*/
11190: PetscErrorCode MatCreateGraph(Mat A, PetscBool sym, PetscBool scale, PetscReal filter, Mat *graph)
11191: {
11192: PetscFunctionBegin;
11196: PetscAssertPointer(graph, 5);
11197: PetscUseTypeMethod(A, creategraph, sym, scale, filter, graph);
11198: PetscFunctionReturn(PETSC_SUCCESS);
11199: }
11201: /*@
11202: MatEliminateZeros - eliminate the nondiagonal zero entries in place from the nonzero structure of a sparse `Mat` in place,
11203: meaning the same memory is used for the matrix, and no new memory is allocated.
11205: Collective
11207: Input Parameters:
11208: + A - the matrix
11209: - 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
11211: Level: intermediate
11213: Developer Note:
11214: The entries in the sparse matrix data structure are shifted to fill in the unneeded locations in the data. Thus the end
11215: of the arrays in the data structure are unneeded.
11217: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatCreateGraph()`, `MatFilter()`
11218: @*/
11219: PetscErrorCode MatEliminateZeros(Mat A, PetscBool keep)
11220: {
11221: PetscFunctionBegin;
11223: PetscUseTypeMethod(A, eliminatezeros, keep);
11224: PetscFunctionReturn(PETSC_SUCCESS);
11225: }