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, MAT_HIPCopyToGPU;
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: 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, st;
203: PetscFunctionBegin;
204: PetscCall(MatCreateVecs(mat, &r, &l));
205: if (!cols) { /* nonzero rows */
206: PetscCall(MatGetOwnershipRange(mat, &st, NULL));
207: PetscCall(MatGetSize(mat, &N, NULL));
208: PetscCall(MatGetLocalSize(mat, &n, NULL));
209: PetscCall(VecSet(l, 0.0));
210: PetscCall(VecSetRandom(r, NULL));
211: PetscCall(MatMult(mat, r, l));
212: PetscCall(VecGetArrayRead(l, &al));
213: } else { /* nonzero columns */
214: PetscCall(MatGetOwnershipRangeColumn(mat, &st, NULL));
215: PetscCall(MatGetSize(mat, NULL, &N));
216: PetscCall(MatGetLocalSize(mat, NULL, &n));
217: PetscCall(VecSet(r, 0.0));
218: PetscCall(VecSetRandom(l, NULL));
219: PetscCall(MatMultTranspose(mat, l, r));
220: PetscCall(VecGetArrayRead(r, &al));
221: }
222: if (tol <= 0.0) {
223: for (i = 0, nz = 0; i < n; i++)
224: if (al[i] != 0.0) nz++;
225: } else {
226: for (i = 0, nz = 0; i < n; i++)
227: if (PetscAbsScalar(al[i]) > tol) nz++;
228: }
229: PetscCall(MPIU_Allreduce(&nz, &gnz, 1, MPIU_INT, MPI_SUM, PetscObjectComm((PetscObject)mat)));
230: if (gnz != N) {
231: PetscInt *nzr;
232: PetscCall(PetscMalloc1(nz, &nzr));
233: if (nz) {
234: if (tol < 0) {
235: for (i = 0, nz = 0; i < n; i++)
236: if (al[i] != 0.0) nzr[nz++] = i + st;
237: } else {
238: for (i = 0, nz = 0; i < n; i++)
239: if (PetscAbsScalar(al[i]) > tol) nzr[nz++] = i + st;
240: }
241: }
242: PetscCall(ISCreateGeneral(PetscObjectComm((PetscObject)mat), nz, nzr, PETSC_OWN_POINTER, nonzero));
243: } else *nonzero = NULL;
244: if (!cols) { /* nonzero rows */
245: PetscCall(VecRestoreArrayRead(l, &al));
246: } else {
247: PetscCall(VecRestoreArrayRead(r, &al));
248: }
249: PetscCall(VecDestroy(&l));
250: PetscCall(VecDestroy(&r));
251: PetscFunctionReturn(PETSC_SUCCESS);
252: }
254: /*@
255: MatFindNonzeroRows - Locate all rows that are not completely zero in the matrix
257: Input Parameter:
258: . mat - the matrix
260: Output Parameter:
261: . keptrows - the rows that are not completely zero
263: Level: intermediate
265: Note:
266: `keptrows` is set to `NULL` if all rows are nonzero.
268: .seealso: [](ch_matrices), `Mat`, `MatFindZeroRows()`
269: @*/
270: PetscErrorCode MatFindNonzeroRows(Mat mat, IS *keptrows)
271: {
272: PetscFunctionBegin;
275: PetscAssertPointer(keptrows, 2);
276: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
277: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
278: if (mat->ops->findnonzerorows) PetscUseTypeMethod(mat, findnonzerorows, keptrows);
279: else PetscCall(MatFindNonzeroRowsOrCols_Basic(mat, PETSC_FALSE, 0.0, keptrows));
280: PetscFunctionReturn(PETSC_SUCCESS);
281: }
283: /*@
284: MatFindZeroRows - Locate all rows that are completely zero in the matrix
286: Input Parameter:
287: . mat - the matrix
289: Output Parameter:
290: . zerorows - the rows that are completely zero
292: Level: intermediate
294: Note:
295: `zerorows` is set to `NULL` if no rows are zero.
297: .seealso: [](ch_matrices), `Mat`, `MatFindNonzeroRows()`
298: @*/
299: PetscErrorCode MatFindZeroRows(Mat mat, IS *zerorows)
300: {
301: IS keptrows;
302: PetscInt m, n;
304: PetscFunctionBegin;
307: PetscAssertPointer(zerorows, 2);
308: PetscCall(MatFindNonzeroRows(mat, &keptrows));
309: /* MatFindNonzeroRows sets keptrows to NULL if there are no zero rows.
310: In keeping with this convention, we set zerorows to NULL if there are no zero
311: rows. */
312: if (keptrows == NULL) {
313: *zerorows = NULL;
314: } else {
315: PetscCall(MatGetOwnershipRange(mat, &m, &n));
316: PetscCall(ISComplement(keptrows, m, n, zerorows));
317: PetscCall(ISDestroy(&keptrows));
318: }
319: PetscFunctionReturn(PETSC_SUCCESS);
320: }
322: /*@
323: MatGetDiagonalBlock - Returns the part of the matrix associated with the on-process coupling
325: Not Collective
327: Input Parameter:
328: . A - the matrix
330: Output Parameter:
331: . a - the diagonal part (which is a SEQUENTIAL matrix)
333: Level: advanced
335: Notes:
336: See `MatCreateAIJ()` for more information on the "diagonal part" of the matrix.
338: Use caution, as the reference count on the returned matrix is not incremented and it is used as part of `A`'s normal operation.
340: .seealso: [](ch_matrices), `Mat`, `MatCreateAIJ()`, `MATAIJ`, `MATBAIJ`, `MATSBAIJ`
341: @*/
342: PetscErrorCode MatGetDiagonalBlock(Mat A, Mat *a)
343: {
344: PetscFunctionBegin;
347: PetscAssertPointer(a, 2);
348: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
349: if (A->ops->getdiagonalblock) PetscUseTypeMethod(A, getdiagonalblock, a);
350: else {
351: PetscMPIInt size;
353: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)A), &size));
354: PetscCheck(size == 1, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Not for parallel matrix type %s", ((PetscObject)A)->type_name);
355: *a = A;
356: }
357: PetscFunctionReturn(PETSC_SUCCESS);
358: }
360: /*@
361: MatGetTrace - Gets the trace of a matrix. The sum of the diagonal entries.
363: Collective
365: Input Parameter:
366: . mat - the matrix
368: Output Parameter:
369: . trace - the sum of the diagonal entries
371: Level: advanced
373: .seealso: [](ch_matrices), `Mat`
374: @*/
375: PetscErrorCode MatGetTrace(Mat mat, PetscScalar *trace)
376: {
377: Vec diag;
379: PetscFunctionBegin;
381: PetscAssertPointer(trace, 2);
382: PetscCall(MatCreateVecs(mat, &diag, NULL));
383: PetscCall(MatGetDiagonal(mat, diag));
384: PetscCall(VecSum(diag, trace));
385: PetscCall(VecDestroy(&diag));
386: PetscFunctionReturn(PETSC_SUCCESS);
387: }
389: /*@
390: MatRealPart - Zeros out the imaginary part of the matrix
392: Logically Collective
394: Input Parameter:
395: . mat - the matrix
397: Level: advanced
399: .seealso: [](ch_matrices), `Mat`, `MatImaginaryPart()`
400: @*/
401: PetscErrorCode MatRealPart(Mat mat)
402: {
403: PetscFunctionBegin;
406: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
407: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
408: MatCheckPreallocated(mat, 1);
409: PetscUseTypeMethod(mat, realpart);
410: PetscFunctionReturn(PETSC_SUCCESS);
411: }
413: /*@C
414: MatGetGhosts - Get the global indices of all ghost nodes defined by the sparse matrix
416: Collective
418: Input Parameter:
419: . mat - the matrix
421: Output Parameters:
422: + nghosts - number of ghosts (for `MATBAIJ` and `MATSBAIJ` matrices there is one ghost for each matrix block)
423: - ghosts - the global indices of the ghost points
425: Level: advanced
427: Note:
428: `nghosts` and `ghosts` are suitable to pass into `VecCreateGhost()` or `VecCreateGhostBlock()`
430: .seealso: [](ch_matrices), `Mat`, `VecCreateGhost()`, `VecCreateGhostBlock()`
431: @*/
432: PetscErrorCode MatGetGhosts(Mat mat, PetscInt *nghosts, const PetscInt *ghosts[])
433: {
434: PetscFunctionBegin;
437: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
438: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
439: if (mat->ops->getghosts) PetscUseTypeMethod(mat, getghosts, nghosts, ghosts);
440: else {
441: if (nghosts) *nghosts = 0;
442: if (ghosts) *ghosts = NULL;
443: }
444: PetscFunctionReturn(PETSC_SUCCESS);
445: }
447: /*@
448: MatImaginaryPart - Moves the imaginary part of the matrix to the real part and zeros the imaginary part
450: Logically Collective
452: Input Parameter:
453: . mat - the matrix
455: Level: advanced
457: .seealso: [](ch_matrices), `Mat`, `MatRealPart()`
458: @*/
459: PetscErrorCode MatImaginaryPart(Mat mat)
460: {
461: PetscFunctionBegin;
464: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
465: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
466: MatCheckPreallocated(mat, 1);
467: PetscUseTypeMethod(mat, imaginarypart);
468: PetscFunctionReturn(PETSC_SUCCESS);
469: }
471: /*@
472: MatMissingDiagonal - Determine if sparse matrix is missing a diagonal entry (or block entry for `MATBAIJ` and `MATSBAIJ` matrices) in the nonzero structure
474: Not Collective
476: Input Parameter:
477: . mat - the matrix
479: Output Parameters:
480: + missing - is any diagonal entry missing
481: - dd - first diagonal entry that is missing (optional) on this process
483: Level: advanced
485: Note:
486: This does not return diagonal entries that are in the nonzero structure but happen to have a zero numerical value
488: .seealso: [](ch_matrices), `Mat`
489: @*/
490: PetscErrorCode MatMissingDiagonal(Mat mat, PetscBool *missing, PetscInt *dd)
491: {
492: PetscFunctionBegin;
495: PetscAssertPointer(missing, 2);
496: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix %s", ((PetscObject)mat)->type_name);
497: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
498: PetscUseTypeMethod(mat, missingdiagonal, missing, dd);
499: PetscFunctionReturn(PETSC_SUCCESS);
500: }
502: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
503: /*@C
504: MatGetRow - Gets a row of a matrix. You MUST call `MatRestoreRow()`
505: for each row that you get to ensure that your application does
506: not bleed memory.
508: Not Collective
510: Input Parameters:
511: + mat - the matrix
512: - row - the row to get
514: Output Parameters:
515: + ncols - if not `NULL`, the number of nonzeros in `row`
516: . cols - if not `NULL`, the column numbers
517: - vals - if not `NULL`, the numerical values
519: Level: advanced
521: Notes:
522: This routine is provided for people who need to have direct access
523: to the structure of a matrix. We hope that we provide enough
524: high-level matrix routines that few users will need it.
526: `MatGetRow()` always returns 0-based column indices, regardless of
527: whether the internal representation is 0-based (default) or 1-based.
529: For better efficiency, set `cols` and/or `vals` to `NULL` if you do
530: not wish to extract these quantities.
532: The user can only examine the values extracted with `MatGetRow()`;
533: the values CANNOT be altered. To change the matrix entries, one
534: must use `MatSetValues()`.
536: You can only have one call to `MatGetRow()` outstanding for a particular
537: matrix at a time, per processor. `MatGetRow()` can only obtain rows
538: associated with the given processor, it cannot get rows from the
539: other processors; for that we suggest using `MatCreateSubMatrices()`, then
540: `MatGetRow()` on the submatrix. The row index passed to `MatGetRow()`
541: is in the global number of rows.
543: Use `MatGetRowIJ()` and `MatRestoreRowIJ()` to access all the local indices of the sparse matrix.
545: Use `MatSeqAIJGetArray()` and similar functions to access the numerical values for certain matrix types directly.
547: Fortran Note:
548: The calling sequence is
549: .vb
550: MatGetRow(matrix,row,ncols,cols,values,ierr)
551: Mat matrix (input)
552: integer row (input)
553: integer ncols (output)
554: integer cols(maxcols) (output)
555: double precision (or double complex) values(maxcols) output
556: .ve
557: where maxcols >= maximum nonzeros in any row of the matrix.
559: .seealso: [](ch_matrices), `Mat`, `MatRestoreRow()`, `MatSetValues()`, `MatGetValues()`, `MatCreateSubMatrices()`, `MatGetDiagonal()`, `MatGetRowIJ()`, `MatRestoreRowIJ()`
560: @*/
561: PetscErrorCode MatGetRow(Mat mat, PetscInt row, PetscInt *ncols, const PetscInt *cols[], const PetscScalar *vals[])
562: {
563: PetscInt incols;
565: PetscFunctionBegin;
568: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
569: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
570: MatCheckPreallocated(mat, 1);
571: 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);
572: PetscCall(PetscLogEventBegin(MAT_GetRow, mat, 0, 0, 0));
573: PetscUseTypeMethod(mat, getrow, row, &incols, (PetscInt **)cols, (PetscScalar **)vals);
574: if (ncols) *ncols = incols;
575: PetscCall(PetscLogEventEnd(MAT_GetRow, mat, 0, 0, 0));
576: PetscFunctionReturn(PETSC_SUCCESS);
577: }
579: /*@
580: MatConjugate - replaces the matrix values with their complex conjugates
582: Logically Collective
584: Input Parameter:
585: . mat - the matrix
587: Level: advanced
589: .seealso: [](ch_matrices), `Mat`, `MatRealPart()`, `MatImaginaryPart()`, `VecConjugate()`, `MatTranspose()`
590: @*/
591: PetscErrorCode MatConjugate(Mat mat)
592: {
593: PetscFunctionBegin;
595: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
596: if (PetscDefined(USE_COMPLEX) && mat->hermitian != PETSC_BOOL3_TRUE) {
597: PetscUseTypeMethod(mat, conjugate);
598: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
599: }
600: PetscFunctionReturn(PETSC_SUCCESS);
601: }
603: /*@C
604: MatRestoreRow - Frees any temporary space allocated by `MatGetRow()`.
606: Not Collective
608: Input Parameters:
609: + mat - the matrix
610: . row - the row to get
611: . ncols - the number of nonzeros
612: . cols - the columns of the nonzeros
613: - vals - if nonzero the column values
615: Level: advanced
617: Notes:
618: This routine should be called after you have finished examining the entries.
620: This routine zeros out `ncols`, `cols`, and `vals`. This is to prevent accidental
621: us of the array after it has been restored. If you pass `NULL`, it will
622: not zero the pointers. Use of `cols` or `vals` after `MatRestoreRow()` is invalid.
624: Fortran Notes:
625: The calling sequence is
626: .vb
627: MatRestoreRow(matrix,row,ncols,cols,values,ierr)
628: Mat matrix (input)
629: integer row (input)
630: integer ncols (output)
631: integer cols(maxcols) (output)
632: double precision (or double complex) values(maxcols) output
633: .ve
634: Where maxcols >= maximum nonzeros in any row of the matrix.
636: In Fortran `MatRestoreRow()` MUST be called after `MatGetRow()`
637: before another call to `MatGetRow()` can be made.
639: .seealso: [](ch_matrices), `Mat`, `MatGetRow()`
640: @*/
641: PetscErrorCode MatRestoreRow(Mat mat, PetscInt row, PetscInt *ncols, const PetscInt *cols[], const PetscScalar *vals[])
642: {
643: PetscFunctionBegin;
645: if (ncols) PetscAssertPointer(ncols, 3);
646: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
647: if (!mat->ops->restorerow) PetscFunctionReturn(PETSC_SUCCESS);
648: PetscUseTypeMethod(mat, restorerow, row, ncols, (PetscInt **)cols, (PetscScalar **)vals);
649: if (ncols) *ncols = 0;
650: if (cols) *cols = NULL;
651: if (vals) *vals = NULL;
652: PetscFunctionReturn(PETSC_SUCCESS);
653: }
655: /*@
656: MatGetRowUpperTriangular - Sets a flag to enable calls to `MatGetRow()` for matrix in `MATSBAIJ` format.
657: You should call `MatRestoreRowUpperTriangular()` after calling` MatGetRow()` and `MatRestoreRow()` to disable the flag.
659: Not Collective
661: Input Parameter:
662: . mat - the matrix
664: Level: advanced
666: Note:
667: 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.
669: .seealso: [](ch_matrices), `Mat`, `MATSBAIJ`, `MatRestoreRowUpperTriangular()`
670: @*/
671: PetscErrorCode MatGetRowUpperTriangular(Mat mat)
672: {
673: PetscFunctionBegin;
676: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
677: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
678: MatCheckPreallocated(mat, 1);
679: if (!mat->ops->getrowuppertriangular) PetscFunctionReturn(PETSC_SUCCESS);
680: PetscUseTypeMethod(mat, getrowuppertriangular);
681: PetscFunctionReturn(PETSC_SUCCESS);
682: }
684: /*@
685: MatRestoreRowUpperTriangular - Disable calls to `MatGetRow()` for matrix in `MATSBAIJ` format.
687: Not Collective
689: Input Parameter:
690: . mat - the matrix
692: Level: advanced
694: Note:
695: This routine should be called after you have finished calls to `MatGetRow()` and `MatRestoreRow()`.
697: .seealso: [](ch_matrices), `Mat`, `MATSBAIJ`, `MatGetRowUpperTriangular()`
698: @*/
699: PetscErrorCode MatRestoreRowUpperTriangular(Mat mat)
700: {
701: PetscFunctionBegin;
704: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
705: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
706: MatCheckPreallocated(mat, 1);
707: if (!mat->ops->restorerowuppertriangular) PetscFunctionReturn(PETSC_SUCCESS);
708: PetscUseTypeMethod(mat, restorerowuppertriangular);
709: PetscFunctionReturn(PETSC_SUCCESS);
710: }
712: /*@
713: MatSetOptionsPrefix - Sets the prefix used for searching for all
714: `Mat` options in the database.
716: Logically Collective
718: Input Parameters:
719: + A - the matrix
720: - prefix - the prefix to prepend to all option names
722: Level: advanced
724: Notes:
725: A hyphen (-) must NOT be given at the beginning of the prefix name.
726: The first character of all runtime options is AUTOMATICALLY the hyphen.
728: This is NOT used for options for the factorization of the matrix. Normally the
729: prefix is automatically passed in from the PC calling the factorization. To set
730: it directly use `MatSetOptionsPrefixFactor()`
732: .seealso: [](ch_matrices), `Mat`, `MatSetFromOptions()`, `MatSetOptionsPrefixFactor()`
733: @*/
734: PetscErrorCode MatSetOptionsPrefix(Mat A, const char prefix[])
735: {
736: PetscFunctionBegin;
738: PetscCall(PetscObjectSetOptionsPrefix((PetscObject)A, prefix));
739: PetscFunctionReturn(PETSC_SUCCESS);
740: }
742: /*@
743: MatSetOptionsPrefixFactor - Sets the prefix used for searching for all matrix factor options in the database for
744: for matrices created with `MatGetFactor()`
746: Logically Collective
748: Input Parameters:
749: + A - the matrix
750: - prefix - the prefix to prepend to all option names for the factored matrix
752: Level: developer
754: Notes:
755: A hyphen (-) must NOT be given at the beginning of the prefix name.
756: The first character of all runtime options is AUTOMATICALLY the hyphen.
758: Normally the prefix is automatically passed in from the `PC` calling the factorization. To set
759: it directly when not using `KSP`/`PC` use `MatSetOptionsPrefixFactor()`
761: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSetFromOptions()`, `MatSetOptionsPrefix()`, `MatAppendOptionsPrefixFactor()`
762: @*/
763: PetscErrorCode MatSetOptionsPrefixFactor(Mat A, const char prefix[])
764: {
765: PetscFunctionBegin;
767: if (prefix) {
768: PetscAssertPointer(prefix, 2);
769: PetscCheck(prefix[0] != '-', PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONG, "Options prefix should not begin with a hyphen");
770: if (prefix != A->factorprefix) {
771: PetscCall(PetscFree(A->factorprefix));
772: PetscCall(PetscStrallocpy(prefix, &A->factorprefix));
773: }
774: } else PetscCall(PetscFree(A->factorprefix));
775: PetscFunctionReturn(PETSC_SUCCESS);
776: }
778: /*@
779: MatAppendOptionsPrefixFactor - Appends to the prefix used for searching for all matrix factor options in the database for
780: for matrices created with `MatGetFactor()`
782: Logically Collective
784: Input Parameters:
785: + A - the matrix
786: - prefix - the prefix to prepend to all option names for the factored matrix
788: Level: developer
790: Notes:
791: A hyphen (-) must NOT be given at the beginning of the prefix name.
792: The first character of all runtime options is AUTOMATICALLY the hyphen.
794: Normally the prefix is automatically passed in from the `PC` calling the factorization. To set
795: it directly when not using `KSP`/`PC` use `MatAppendOptionsPrefixFactor()`
797: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `PetscOptionsCreate()`, `PetscOptionsDestroy()`, `PetscObjectSetOptionsPrefix()`, `PetscObjectPrependOptionsPrefix()`,
798: `PetscObjectGetOptionsPrefix()`, `TSAppendOptionsPrefix()`, `SNESAppendOptionsPrefix()`, `KSPAppendOptionsPrefix()`, `MatSetOptionsPrefixFactor()`,
799: `MatSetOptionsPrefix()`
800: @*/
801: PetscErrorCode MatAppendOptionsPrefixFactor(Mat A, const char prefix[])
802: {
803: size_t len1, len2, new_len;
805: PetscFunctionBegin;
807: if (!prefix) PetscFunctionReturn(PETSC_SUCCESS);
808: if (!A->factorprefix) {
809: PetscCall(MatSetOptionsPrefixFactor(A, prefix));
810: PetscFunctionReturn(PETSC_SUCCESS);
811: }
812: PetscCheck(prefix[0] != '-', PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONG, "Options prefix should not begin with a hyphen");
814: PetscCall(PetscStrlen(A->factorprefix, &len1));
815: PetscCall(PetscStrlen(prefix, &len2));
816: new_len = len1 + len2 + 1;
817: PetscCall(PetscRealloc(new_len * sizeof(*A->factorprefix), &A->factorprefix));
818: PetscCall(PetscStrncpy(A->factorprefix + len1, prefix, len2 + 1));
819: PetscFunctionReturn(PETSC_SUCCESS);
820: }
822: /*@
823: MatAppendOptionsPrefix - Appends to the prefix used for searching for all
824: matrix options in the database.
826: Logically Collective
828: Input Parameters:
829: + A - the matrix
830: - prefix - the prefix to prepend to all option names
832: Level: advanced
834: Note:
835: A hyphen (-) must NOT be given at the beginning of the prefix name.
836: The first character of all runtime options is AUTOMATICALLY the hyphen.
838: .seealso: [](ch_matrices), `Mat`, `MatGetOptionsPrefix()`, `MatAppendOptionsPrefixFactor()`, `MatSetOptionsPrefix()`
839: @*/
840: PetscErrorCode MatAppendOptionsPrefix(Mat A, const char prefix[])
841: {
842: PetscFunctionBegin;
844: PetscCall(PetscObjectAppendOptionsPrefix((PetscObject)A, prefix));
845: PetscFunctionReturn(PETSC_SUCCESS);
846: }
848: /*@
849: MatGetOptionsPrefix - Gets the prefix used for searching for all
850: matrix options in the database.
852: Not Collective
854: Input Parameter:
855: . A - the matrix
857: Output Parameter:
858: . prefix - pointer to the prefix string used
860: Level: advanced
862: Fortran Note:
863: The user should pass in a string `prefix` of
864: sufficient length to hold the prefix.
866: .seealso: [](ch_matrices), `Mat`, `MatAppendOptionsPrefix()`, `MatSetOptionsPrefix()`, `MatAppendOptionsPrefixFactor()`, `MatSetOptionsPrefixFactor()`
867: @*/
868: PetscErrorCode MatGetOptionsPrefix(Mat A, const char *prefix[])
869: {
870: PetscFunctionBegin;
872: PetscAssertPointer(prefix, 2);
873: PetscCall(PetscObjectGetOptionsPrefix((PetscObject)A, prefix));
874: PetscFunctionReturn(PETSC_SUCCESS);
875: }
877: /*@C
878: MatGetState - Gets the state of a `Mat`.
880: Not Collective
882: Input Parameter:
883: . A - the matrix
885: Output Parameter:
886: . state - the object state
888: Level: advanced
890: Note:
891: Object state is an integer which gets increased every time
892: the object is changed. By saving and later querying the object state
893: one can determine whether information about the object is still current.
895: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `PetscObjectStateGet()`
896: @*/
897: PetscErrorCode MatGetState(Mat A, PetscObjectState *state)
898: {
899: PetscFunctionBegin;
901: PetscAssertPointer(state, 2);
902: PetscCall(PetscObjectStateGet((PetscObject)A, state));
903: PetscFunctionReturn(PETSC_SUCCESS);
904: }
906: /*@
907: MatResetPreallocation - Reset matrix to use the original nonzero pattern provided by the user.
909: Collective
911: Input Parameter:
912: . A - the matrix
914: Level: beginner
916: Notes:
917: The allocated memory will be shrunk after calling `MatAssemblyBegin()` and `MatAssemblyEnd()` with `MAT_FINAL_ASSEMBLY`.
919: Users can reset the preallocation to access the original memory.
921: Currently only supported for `MATAIJ` matrices.
923: .seealso: [](ch_matrices), `Mat`, `MatSeqAIJSetPreallocation()`, `MatMPIAIJSetPreallocation()`, `MatXAIJSetPreallocation()`
924: @*/
925: PetscErrorCode MatResetPreallocation(Mat A)
926: {
927: PetscFunctionBegin;
930: PetscCheck(A->insertmode == NOT_SET_VALUES, PETSC_COMM_SELF, PETSC_ERR_SUP, "Cannot reset preallocation after setting some values but not yet calling MatAssemblyBegin()/MatAssemblyEnd()");
931: if (A->num_ass == 0) PetscFunctionReturn(PETSC_SUCCESS);
932: PetscUseMethod(A, "MatResetPreallocation_C", (Mat), (A));
933: PetscFunctionReturn(PETSC_SUCCESS);
934: }
936: /*@
937: MatSetUp - Sets up the internal matrix data structures for later use.
939: Collective
941: Input Parameter:
942: . A - the matrix
944: Level: intermediate
946: Notes:
947: If the user has not set preallocation for this matrix then an efficient algorithm will be used for the first round of
948: setting values in the matrix.
950: This routine is called internally by other matrix functions when needed so rarely needs to be called by users
952: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatCreate()`, `MatDestroy()`, `MatXAIJSetPreallocation()`
953: @*/
954: PetscErrorCode MatSetUp(Mat A)
955: {
956: PetscFunctionBegin;
958: if (!((PetscObject)A)->type_name) {
959: PetscMPIInt size;
961: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)A), &size));
962: PetscCall(MatSetType(A, size == 1 ? MATSEQAIJ : MATMPIAIJ));
963: }
964: if (!A->preallocated) PetscTryTypeMethod(A, setup);
965: PetscCall(PetscLayoutSetUp(A->rmap));
966: PetscCall(PetscLayoutSetUp(A->cmap));
967: A->preallocated = PETSC_TRUE;
968: PetscFunctionReturn(PETSC_SUCCESS);
969: }
971: #if defined(PETSC_HAVE_SAWS)
972: #include <petscviewersaws.h>
973: #endif
975: /*
976: If threadsafety is on extraneous matrices may be printed
978: This flag cannot be stored in the matrix because the original matrix in MatView() may assemble a new matrix which is passed into MatViewFromOptions()
979: */
980: #if !defined(PETSC_HAVE_THREADSAFETY)
981: static PetscInt insidematview = 0;
982: #endif
984: /*@C
985: MatViewFromOptions - View properties of the matrix based on options set in the options database
987: Collective
989: Input Parameters:
990: + A - the matrix
991: . obj - optional additional object that provides the options prefix to use
992: - name - command line option
994: Options Database Key:
995: . -mat_view [viewertype]:... - the viewer and its options
997: Level: intermediate
999: Note:
1000: .vb
1001: If no value is provided ascii:stdout is used
1002: ascii[:[filename][:[format][:append]]] defaults to stdout - format can be one of ascii_info, ascii_info_detail, or ascii_matlab,
1003: for example ascii::ascii_info prints just the information about the object not all details
1004: unless :append is given filename opens in write mode, overwriting what was already there
1005: binary[:[filename][:[format][:append]]] defaults to the file binaryoutput
1006: draw[:drawtype[:filename]] for example, draw:tikz, draw:tikz:figure.tex or draw:x
1007: socket[:port] defaults to the standard output port
1008: saws[:communicatorname] publishes object to the Scientific Application Webserver (SAWs)
1009: .ve
1011: .seealso: [](ch_matrices), `Mat`, `MatView()`, `PetscObjectViewFromOptions()`, `MatCreate()`
1012: @*/
1013: PetscErrorCode MatViewFromOptions(Mat A, PetscObject obj, const char name[])
1014: {
1015: PetscFunctionBegin;
1017: #if !defined(PETSC_HAVE_THREADSAFETY)
1018: if (insidematview) PetscFunctionReturn(PETSC_SUCCESS);
1019: #endif
1020: PetscCall(PetscObjectViewFromOptions((PetscObject)A, obj, name));
1021: PetscFunctionReturn(PETSC_SUCCESS);
1022: }
1024: /*@C
1025: MatView - display information about a matrix in a variety ways
1027: Collective on viewer
1029: Input Parameters:
1030: + mat - the matrix
1031: - viewer - visualization context
1033: Options Database Keys:
1034: + -mat_view ::ascii_info - Prints info on matrix at conclusion of `MatAssemblyEnd()`
1035: . -mat_view ::ascii_info_detail - Prints more detailed info
1036: . -mat_view - Prints matrix in ASCII format
1037: . -mat_view ::ascii_matlab - Prints matrix in MATLAB format
1038: . -mat_view draw - PetscDraws nonzero structure of matrix, using `MatView()` and `PetscDrawOpenX()`.
1039: . -display <name> - Sets display name (default is host)
1040: . -draw_pause <sec> - Sets number of seconds to pause after display
1041: . -mat_view socket - Sends matrix to socket, can be accessed from MATLAB (see Users-Manual: ch_matlab for details)
1042: . -viewer_socket_machine <machine> - -
1043: . -viewer_socket_port <port> - -
1044: . -mat_view binary - save matrix to file in binary format
1045: - -viewer_binary_filename <name> - -
1047: Level: beginner
1049: Notes:
1050: The available visualization contexts include
1051: + `PETSC_VIEWER_STDOUT_SELF` - for sequential matrices
1052: . `PETSC_VIEWER_STDOUT_WORLD` - for parallel matrices created on `PETSC_COMM_WORLD`
1053: . `PETSC_VIEWER_STDOUT_`(comm) - for matrices created on MPI communicator comm
1054: - `PETSC_VIEWER_DRAW_WORLD` - graphical display of nonzero structure
1056: The user can open alternative visualization contexts with
1057: + `PetscViewerASCIIOpen()` - Outputs matrix to a specified file
1058: . `PetscViewerBinaryOpen()` - Outputs matrix in binary to a
1059: specified file; corresponding input uses `MatLoad()`
1060: . `PetscViewerDrawOpen()` - Outputs nonzero matrix structure to
1061: an X window display
1062: - `PetscViewerSocketOpen()` - Outputs matrix to Socket viewer.
1063: Currently only the `MATSEQDENSE` and `MATAIJ`
1064: matrix types support the Socket viewer.
1066: The user can call `PetscViewerPushFormat()` to specify the output
1067: format of ASCII printed objects (when using `PETSC_VIEWER_STDOUT_SELF`,
1068: `PETSC_VIEWER_STDOUT_WORLD` and `PetscViewerASCIIOpen()`). Available formats include
1069: + `PETSC_VIEWER_DEFAULT` - default, prints matrix contents
1070: . `PETSC_VIEWER_ASCII_MATLAB` - prints matrix contents in MATLAB format
1071: . `PETSC_VIEWER_ASCII_DENSE` - prints entire matrix including zeros
1072: . `PETSC_VIEWER_ASCII_COMMON` - prints matrix contents, using a sparse
1073: format common among all matrix types
1074: . `PETSC_VIEWER_ASCII_IMPL` - prints matrix contents, using an implementation-specific
1075: format (which is in many cases the same as the default)
1076: . `PETSC_VIEWER_ASCII_INFO` - prints basic information about the matrix
1077: size and structure (not the matrix entries)
1078: - `PETSC_VIEWER_ASCII_INFO_DETAIL` - prints more detailed information about
1079: the matrix structure
1081: The ASCII viewers are only recommended for small matrices on at most a moderate number of processes,
1082: the program will seemingly hang and take hours for larger matrices, for larger matrices one should use the binary format.
1084: In the debugger you can do "call MatView(mat,0)" to display the matrix. (The same holds for any PETSc object viewer).
1086: See the manual page for `MatLoad()` for the exact format of the binary file when the binary
1087: viewer is used.
1089: See share/petsc/matlab/PetscBinaryRead.m for a MATLAB code that can read in the binary file when the binary
1090: viewer is used and lib/petsc/bin/PetscBinaryIO.py for loading them into Python.
1092: One can use '-mat_view draw -draw_pause -1' to pause the graphical display of matrix nonzero structure,
1093: and then use the following mouse functions.
1094: .vb
1095: left mouse: zoom in
1096: middle mouse: zoom out
1097: right mouse: continue with the simulation
1098: .ve
1100: .seealso: [](ch_matrices), `Mat`, `PetscViewerPushFormat()`, `PetscViewerASCIIOpen()`, `PetscViewerDrawOpen()`, `PetscViewer`,
1101: `PetscViewerSocketOpen()`, `PetscViewerBinaryOpen()`, `MatLoad()`, `MatViewFromOptions()`
1102: @*/
1103: PetscErrorCode MatView(Mat mat, PetscViewer viewer)
1104: {
1105: PetscInt rows, cols, rbs, cbs;
1106: PetscBool isascii, isstring, issaws;
1107: PetscViewerFormat format;
1108: PetscMPIInt size;
1110: PetscFunctionBegin;
1113: if (!viewer) PetscCall(PetscViewerASCIIGetStdout(PetscObjectComm((PetscObject)mat), &viewer));
1116: PetscCall(PetscViewerGetFormat(viewer, &format));
1117: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)viewer), &size));
1118: if (size == 1 && format == PETSC_VIEWER_LOAD_BALANCE) PetscFunctionReturn(PETSC_SUCCESS);
1120: #if !defined(PETSC_HAVE_THREADSAFETY)
1121: insidematview++;
1122: #endif
1123: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERSTRING, &isstring));
1124: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &isascii));
1125: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERSAWS, &issaws));
1126: PetscCheck((isascii && (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL)) || !mat->factortype, PetscObjectComm((PetscObject)viewer), PETSC_ERR_ARG_WRONGSTATE, "No viewers for factored matrix except ASCII, info, or info_detail");
1128: PetscCall(PetscLogEventBegin(MAT_View, mat, viewer, 0, 0));
1129: if (isascii) {
1130: if (!mat->preallocated) {
1131: PetscCall(PetscViewerASCIIPrintf(viewer, "Matrix has not been preallocated yet\n"));
1132: #if !defined(PETSC_HAVE_THREADSAFETY)
1133: insidematview--;
1134: #endif
1135: PetscCall(PetscLogEventEnd(MAT_View, mat, viewer, 0, 0));
1136: PetscFunctionReturn(PETSC_SUCCESS);
1137: }
1138: if (!mat->assembled) {
1139: PetscCall(PetscViewerASCIIPrintf(viewer, "Matrix has not been assembled yet\n"));
1140: #if !defined(PETSC_HAVE_THREADSAFETY)
1141: insidematview--;
1142: #endif
1143: PetscCall(PetscLogEventEnd(MAT_View, mat, viewer, 0, 0));
1144: PetscFunctionReturn(PETSC_SUCCESS);
1145: }
1146: PetscCall(PetscObjectPrintClassNamePrefixType((PetscObject)mat, viewer));
1147: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
1148: MatNullSpace nullsp, transnullsp;
1150: PetscCall(PetscViewerASCIIPushTab(viewer));
1151: PetscCall(MatGetSize(mat, &rows, &cols));
1152: PetscCall(MatGetBlockSizes(mat, &rbs, &cbs));
1153: if (rbs != 1 || cbs != 1) {
1154: if (rbs != cbs) PetscCall(PetscViewerASCIIPrintf(viewer, "rows=%" PetscInt_FMT ", cols=%" PetscInt_FMT ", rbs=%" PetscInt_FMT ", cbs=%" PetscInt_FMT "%s\n", rows, cols, rbs, cbs, mat->bsizes ? " variable blocks set" : ""));
1155: else PetscCall(PetscViewerASCIIPrintf(viewer, "rows=%" PetscInt_FMT ", cols=%" PetscInt_FMT ", bs=%" PetscInt_FMT "%s\n", rows, cols, rbs, mat->bsizes ? " variable blocks set" : ""));
1156: } else PetscCall(PetscViewerASCIIPrintf(viewer, "rows=%" PetscInt_FMT ", cols=%" PetscInt_FMT "\n", rows, cols));
1157: if (mat->factortype) {
1158: MatSolverType solver;
1159: PetscCall(MatFactorGetSolverType(mat, &solver));
1160: PetscCall(PetscViewerASCIIPrintf(viewer, "package used to perform factorization: %s\n", solver));
1161: }
1162: if (mat->ops->getinfo) {
1163: MatInfo info;
1164: PetscCall(MatGetInfo(mat, MAT_GLOBAL_SUM, &info));
1165: PetscCall(PetscViewerASCIIPrintf(viewer, "total: nonzeros=%.f, allocated nonzeros=%.f\n", info.nz_used, info.nz_allocated));
1166: if (!mat->factortype) PetscCall(PetscViewerASCIIPrintf(viewer, "total number of mallocs used during MatSetValues calls=%" PetscInt_FMT "\n", (PetscInt)info.mallocs));
1167: }
1168: PetscCall(MatGetNullSpace(mat, &nullsp));
1169: PetscCall(MatGetTransposeNullSpace(mat, &transnullsp));
1170: if (nullsp) PetscCall(PetscViewerASCIIPrintf(viewer, " has attached null space\n"));
1171: if (transnullsp && transnullsp != nullsp) PetscCall(PetscViewerASCIIPrintf(viewer, " has attached transposed null space\n"));
1172: PetscCall(MatGetNearNullSpace(mat, &nullsp));
1173: if (nullsp) PetscCall(PetscViewerASCIIPrintf(viewer, " has attached near null space\n"));
1174: PetscCall(PetscViewerASCIIPushTab(viewer));
1175: PetscCall(MatProductView(mat, viewer));
1176: PetscCall(PetscViewerASCIIPopTab(viewer));
1177: if (mat->bsizes && format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
1178: IS tmp;
1180: PetscCall(ISCreateGeneral(PetscObjectComm((PetscObject)viewer), mat->nblocks, mat->bsizes, PETSC_USE_POINTER, &tmp));
1181: PetscCall(PetscObjectSetName((PetscObject)tmp, "Block Sizes"));
1182: PetscCall(PetscViewerASCIIPushTab(viewer));
1183: PetscCall(ISView(tmp, viewer));
1184: PetscCall(PetscViewerASCIIPopTab(viewer));
1185: PetscCall(ISDestroy(&tmp));
1186: }
1187: }
1188: } else if (issaws) {
1189: #if defined(PETSC_HAVE_SAWS)
1190: PetscMPIInt rank;
1192: PetscCall(PetscObjectName((PetscObject)mat));
1193: PetscCallMPI(MPI_Comm_rank(PETSC_COMM_WORLD, &rank));
1194: if (!((PetscObject)mat)->amsmem && rank == 0) PetscCall(PetscObjectViewSAWs((PetscObject)mat, viewer));
1195: #endif
1196: } else if (isstring) {
1197: const char *type;
1198: PetscCall(MatGetType(mat, &type));
1199: PetscCall(PetscViewerStringSPrintf(viewer, " MatType: %-7.7s", type));
1200: PetscTryTypeMethod(mat, view, viewer);
1201: }
1202: if ((format == PETSC_VIEWER_NATIVE || format == PETSC_VIEWER_LOAD_BALANCE) && mat->ops->viewnative) {
1203: PetscCall(PetscViewerASCIIPushTab(viewer));
1204: PetscUseTypeMethod(mat, viewnative, viewer);
1205: PetscCall(PetscViewerASCIIPopTab(viewer));
1206: } else if (mat->ops->view) {
1207: PetscCall(PetscViewerASCIIPushTab(viewer));
1208: PetscUseTypeMethod(mat, view, viewer);
1209: PetscCall(PetscViewerASCIIPopTab(viewer));
1210: }
1211: if (isascii) {
1212: PetscCall(PetscViewerGetFormat(viewer, &format));
1213: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) PetscCall(PetscViewerASCIIPopTab(viewer));
1214: }
1215: PetscCall(PetscLogEventEnd(MAT_View, mat, viewer, 0, 0));
1216: #if !defined(PETSC_HAVE_THREADSAFETY)
1217: insidematview--;
1218: #endif
1219: PetscFunctionReturn(PETSC_SUCCESS);
1220: }
1222: #if defined(PETSC_USE_DEBUG)
1223: #include <../src/sys/totalview/tv_data_display.h>
1224: PETSC_UNUSED static int TV_display_type(const struct _p_Mat *mat)
1225: {
1226: TV_add_row("Local rows", "int", &mat->rmap->n);
1227: TV_add_row("Local columns", "int", &mat->cmap->n);
1228: TV_add_row("Global rows", "int", &mat->rmap->N);
1229: TV_add_row("Global columns", "int", &mat->cmap->N);
1230: TV_add_row("Typename", TV_ascii_string_type, ((PetscObject)mat)->type_name);
1231: return TV_format_OK;
1232: }
1233: #endif
1235: /*@C
1236: MatLoad - Loads a matrix that has been stored in binary/HDF5 format
1237: with `MatView()`. The matrix format is determined from the options database.
1238: Generates a parallel MPI matrix if the communicator has more than one
1239: processor. The default matrix type is `MATAIJ`.
1241: Collective
1243: Input Parameters:
1244: + mat - the newly loaded matrix, this needs to have been created with `MatCreate()`
1245: or some related function before a call to `MatLoad()`
1246: - viewer - `PETSCVIEWERBINARY`/`PETSCVIEWERHDF5` file viewer
1248: Options Database Key:
1249: . -matload_block_size <bs> - set block size
1251: Level: beginner
1253: Notes:
1254: If the `Mat` type has not yet been given then `MATAIJ` is used, call `MatSetFromOptions()` on the
1255: `Mat` before calling this routine if you wish to set it from the options database.
1257: `MatLoad()` automatically loads into the options database any options
1258: given in the file filename.info where filename is the name of the file
1259: that was passed to the `PetscViewerBinaryOpen()`. The options in the info
1260: file will be ignored if you use the -viewer_binary_skip_info option.
1262: If the type or size of mat is not set before a call to `MatLoad()`, PETSc
1263: sets the default matrix type AIJ and sets the local and global sizes.
1264: If type and/or size is already set, then the same are used.
1266: In parallel, each processor can load a subset of rows (or the
1267: entire matrix). This routine is especially useful when a large
1268: matrix is stored on disk and only part of it is desired on each
1269: processor. For example, a parallel solver may access only some of
1270: the rows from each processor. The algorithm used here reads
1271: relatively small blocks of data rather than reading the entire
1272: matrix and then subsetting it.
1274: Viewer's `PetscViewerType` must be either `PETSCVIEWERBINARY` or `PETSCVIEWERHDF5`.
1275: Such viewer can be created using `PetscViewerBinaryOpen()` or `PetscViewerHDF5Open()`,
1276: or the sequence like
1277: .vb
1278: `PetscViewer` v;
1279: `PetscViewerCreate`(`PETSC_COMM_WORLD`,&v);
1280: `PetscViewerSetType`(v,`PETSCVIEWERBINARY`);
1281: `PetscViewerSetFromOptions`(v);
1282: `PetscViewerFileSetMode`(v,`FILE_MODE_READ`);
1283: `PetscViewerFileSetName`(v,"datafile");
1284: .ve
1285: The optional `PetscViewerSetFromOptions()` call allows overriding `PetscViewerSetType()` using the option
1286: $ -viewer_type {binary, hdf5}
1288: See the example src/ksp/ksp/tutorials/ex27.c with the first approach,
1289: and src/mat/tutorials/ex10.c with the second approach.
1291: In case of `PETSCVIEWERBINARY`, a native PETSc binary format is used. Each of the blocks
1292: is read onto MPI rank 0 and then shipped to its destination MPI rank, one after another.
1293: Multiple objects, both matrices and vectors, can be stored within the same file.
1294: Their `PetscObject` name is ignored; they are loaded in the order of their storage.
1296: Most users should not need to know the details of the binary storage
1297: format, since `MatLoad()` and `MatView()` completely hide these details.
1298: But for anyone who is interested, the standard binary matrix storage
1299: format is
1301: .vb
1302: PetscInt MAT_FILE_CLASSID
1303: PetscInt number of rows
1304: PetscInt number of columns
1305: PetscInt total number of nonzeros
1306: PetscInt *number nonzeros in each row
1307: PetscInt *column indices of all nonzeros (starting index is zero)
1308: PetscScalar *values of all nonzeros
1309: .ve
1310: If PETSc was not configured with `--with-64-bit-indices` then only `MATMPIAIJ` matrices with more than `PETSC_INT_MAX` non-zeros can be
1311: 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
1312: case will not fit in a (32-bit) `PetscInt` the value `PETSC_INT_MAX` is used for the header entry `total number of nonzeros`.
1314: PETSc automatically does the byte swapping for
1315: machines that store the bytes reversed. Thus if you write your own binary
1316: read/write routines you have to swap the bytes; see `PetscBinaryRead()`
1317: and `PetscBinaryWrite()` to see how this may be done.
1319: In case of `PETSCVIEWERHDF5`, a parallel HDF5 reader is used.
1320: Each processor's chunk is loaded independently by its owning MPI process.
1321: Multiple objects, both matrices and vectors, can be stored within the same file.
1322: They are looked up by their PetscObject name.
1324: As the MATLAB MAT-File Version 7.3 format is also a HDF5 flavor, we decided to use
1325: by default the same structure and naming of the AIJ arrays and column count
1326: within the HDF5 file. This means that a MAT file saved with -v7.3 flag, e.g.
1327: $ save example.mat A b -v7.3
1328: can be directly read by this routine (see Reference 1 for details).
1330: Depending on your MATLAB version, this format might be a default,
1331: otherwise you can set it as default in Preferences.
1333: Unless -nocompression flag is used to save the file in MATLAB,
1334: PETSc must be configured with ZLIB package.
1336: See also examples src/mat/tutorials/ex10.c and src/ksp/ksp/tutorials/ex27.c
1338: This reader currently supports only real `MATSEQAIJ`, `MATMPIAIJ`, `MATSEQDENSE` and `MATMPIDENSE` matrices for `PETSCVIEWERHDF5`
1340: Corresponding `MatView()` is not yet implemented.
1342: The loaded matrix is actually a transpose of the original one in MATLAB,
1343: unless you push `PETSC_VIEWER_HDF5_MAT` format (see examples above).
1344: With this format, matrix is automatically transposed by PETSc,
1345: unless the matrix is marked as SPD or symmetric
1346: (see `MatSetOption()`, `MAT_SPD`, `MAT_SYMMETRIC`).
1348: See MATLAB Documentation on `save()`, <https://www.mathworks.com/help/matlab/ref/save.html#btox10b-1-version>
1350: .seealso: [](ch_matrices), `Mat`, `PetscViewerBinaryOpen()`, `PetscViewerSetType()`, `MatView()`, `VecLoad()`
1351: @*/
1352: PetscErrorCode MatLoad(Mat mat, PetscViewer viewer)
1353: {
1354: PetscBool flg;
1356: PetscFunctionBegin;
1360: if (!((PetscObject)mat)->type_name) PetscCall(MatSetType(mat, MATAIJ));
1362: flg = PETSC_FALSE;
1363: PetscCall(PetscOptionsGetBool(((PetscObject)mat)->options, ((PetscObject)mat)->prefix, "-matload_symmetric", &flg, NULL));
1364: if (flg) {
1365: PetscCall(MatSetOption(mat, MAT_SYMMETRIC, PETSC_TRUE));
1366: PetscCall(MatSetOption(mat, MAT_SYMMETRY_ETERNAL, PETSC_TRUE));
1367: }
1368: flg = PETSC_FALSE;
1369: PetscCall(PetscOptionsGetBool(((PetscObject)mat)->options, ((PetscObject)mat)->prefix, "-matload_spd", &flg, NULL));
1370: if (flg) PetscCall(MatSetOption(mat, MAT_SPD, PETSC_TRUE));
1372: PetscCall(PetscLogEventBegin(MAT_Load, mat, viewer, 0, 0));
1373: PetscUseTypeMethod(mat, load, viewer);
1374: PetscCall(PetscLogEventEnd(MAT_Load, mat, viewer, 0, 0));
1375: PetscFunctionReturn(PETSC_SUCCESS);
1376: }
1378: static PetscErrorCode MatDestroy_Redundant(Mat_Redundant **redundant)
1379: {
1380: Mat_Redundant *redund = *redundant;
1382: PetscFunctionBegin;
1383: if (redund) {
1384: if (redund->matseq) { /* via MatCreateSubMatrices() */
1385: PetscCall(ISDestroy(&redund->isrow));
1386: PetscCall(ISDestroy(&redund->iscol));
1387: PetscCall(MatDestroySubMatrices(1, &redund->matseq));
1388: } else {
1389: PetscCall(PetscFree2(redund->send_rank, redund->recv_rank));
1390: PetscCall(PetscFree(redund->sbuf_j));
1391: PetscCall(PetscFree(redund->sbuf_a));
1392: for (PetscInt i = 0; i < redund->nrecvs; i++) {
1393: PetscCall(PetscFree(redund->rbuf_j[i]));
1394: PetscCall(PetscFree(redund->rbuf_a[i]));
1395: }
1396: PetscCall(PetscFree4(redund->sbuf_nz, redund->rbuf_nz, redund->rbuf_j, redund->rbuf_a));
1397: }
1399: if (redund->subcomm) PetscCall(PetscCommDestroy(&redund->subcomm));
1400: PetscCall(PetscFree(redund));
1401: }
1402: PetscFunctionReturn(PETSC_SUCCESS);
1403: }
1405: /*@C
1406: MatDestroy - Frees space taken by a matrix.
1408: Collective
1410: Input Parameter:
1411: . A - the matrix
1413: Level: beginner
1415: Developer Note:
1416: Some special arrays of matrices are not destroyed in this routine but instead by the routines called by
1417: `MatDestroySubMatrices()`. Thus one must be sure that any changes here must also be made in those routines.
1418: `MatHeaderMerge()` and `MatHeaderReplace()` also manipulate the data in the `Mat` object and likely need changes
1419: if changes are needed here.
1421: .seealso: [](ch_matrices), `Mat`, `MatCreate()`
1422: @*/
1423: PetscErrorCode MatDestroy(Mat *A)
1424: {
1425: PetscFunctionBegin;
1426: if (!*A) PetscFunctionReturn(PETSC_SUCCESS);
1428: if (--((PetscObject)*A)->refct > 0) {
1429: *A = NULL;
1430: PetscFunctionReturn(PETSC_SUCCESS);
1431: }
1433: /* if memory was published with SAWs then destroy it */
1434: PetscCall(PetscObjectSAWsViewOff((PetscObject)*A));
1435: PetscTryTypeMethod(*A, destroy);
1437: PetscCall(PetscFree((*A)->factorprefix));
1438: PetscCall(PetscFree((*A)->defaultvectype));
1439: PetscCall(PetscFree((*A)->defaultrandtype));
1440: PetscCall(PetscFree((*A)->bsizes));
1441: PetscCall(PetscFree((*A)->solvertype));
1442: for (PetscInt i = 0; i < MAT_FACTOR_NUM_TYPES; i++) PetscCall(PetscFree((*A)->preferredordering[i]));
1443: if ((*A)->redundant && (*A)->redundant->matseq[0] == *A) (*A)->redundant->matseq[0] = NULL;
1444: PetscCall(MatDestroy_Redundant(&(*A)->redundant));
1445: PetscCall(MatProductClear(*A));
1446: PetscCall(MatNullSpaceDestroy(&(*A)->nullsp));
1447: PetscCall(MatNullSpaceDestroy(&(*A)->transnullsp));
1448: PetscCall(MatNullSpaceDestroy(&(*A)->nearnullsp));
1449: PetscCall(MatDestroy(&(*A)->schur));
1450: PetscCall(PetscLayoutDestroy(&(*A)->rmap));
1451: PetscCall(PetscLayoutDestroy(&(*A)->cmap));
1452: PetscCall(PetscHeaderDestroy(A));
1453: PetscFunctionReturn(PETSC_SUCCESS);
1454: }
1456: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
1457: /*@C
1458: MatSetValues - Inserts or adds a block of values into a matrix.
1459: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
1460: MUST be called after all calls to `MatSetValues()` have been completed.
1462: Not Collective
1464: Input Parameters:
1465: + mat - the matrix
1466: . v - a logically two-dimensional array of values
1467: . m - the number of rows
1468: . idxm - the global indices of the rows
1469: . n - the number of columns
1470: . idxn - the global indices of the columns
1471: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
1473: Level: beginner
1475: Notes:
1476: By default the values, `v`, are stored row-oriented. See `MatSetOption()` for other options.
1478: Calls to `MatSetValues()` with the `INSERT_VALUES` and `ADD_VALUES`
1479: options cannot be mixed without intervening calls to the assembly
1480: routines.
1482: `MatSetValues()` uses 0-based row and column numbers in Fortran
1483: as well as in C.
1485: Negative indices may be passed in `idxm` and `idxn`, these rows and columns are
1486: simply ignored. This allows easily inserting element stiffness matrices
1487: with homogeneous Dirichlet boundary conditions that you don't want represented
1488: in the matrix.
1490: Efficiency Alert:
1491: The routine `MatSetValuesBlocked()` may offer much better efficiency
1492: for users of block sparse formats (`MATSEQBAIJ` and `MATMPIBAIJ`).
1494: Developer Note:
1495: This is labeled with C so does not automatically generate Fortran stubs and interfaces
1496: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
1498: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1499: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`
1500: @*/
1501: PetscErrorCode MatSetValues(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], const PetscScalar v[], InsertMode addv)
1502: {
1503: PetscFunctionBeginHot;
1506: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1507: PetscAssertPointer(idxm, 3);
1508: PetscAssertPointer(idxn, 5);
1509: MatCheckPreallocated(mat, 1);
1511: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
1512: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
1514: if (PetscDefined(USE_DEBUG)) {
1515: PetscInt i, j;
1517: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
1518: if (v) {
1519: for (i = 0; i < m; i++) {
1520: for (j = 0; j < n; j++) {
1521: if (mat->erroriffailure && PetscIsInfOrNanScalar(v[i * n + j]))
1522: #if defined(PETSC_USE_COMPLEX)
1523: 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]);
1524: #else
1525: 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]);
1526: #endif
1527: }
1528: }
1529: }
1530: 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);
1531: 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);
1532: }
1534: if (mat->assembled) {
1535: mat->was_assembled = PETSC_TRUE;
1536: mat->assembled = PETSC_FALSE;
1537: }
1538: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
1539: PetscUseTypeMethod(mat, setvalues, m, idxm, n, idxn, v, addv);
1540: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
1541: PetscFunctionReturn(PETSC_SUCCESS);
1542: }
1544: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
1545: /*@
1546: MatSetValuesIS - Inserts or adds a block of values into a matrix using an `IS` to indicate the rows and columns
1547: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
1548: MUST be called after all calls to `MatSetValues()` have been completed.
1550: Not Collective
1552: Input Parameters:
1553: + mat - the matrix
1554: . v - a logically two-dimensional array of values
1555: . ism - the rows to provide
1556: . isn - the columns to provide
1557: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
1559: Level: beginner
1561: Notes:
1562: By default the values, `v`, are stored row-oriented. See `MatSetOption()` for other options.
1564: Calls to `MatSetValues()` with the `INSERT_VALUES` and `ADD_VALUES`
1565: options cannot be mixed without intervening calls to the assembly
1566: routines.
1568: `MatSetValues()` uses 0-based row and column numbers in Fortran
1569: as well as in C.
1571: Negative indices may be passed in `ism` and `isn`, these rows and columns are
1572: simply ignored. This allows easily inserting element stiffness matrices
1573: with homogeneous Dirichlet boundary conditions that you don't want represented
1574: in the matrix.
1576: Efficiency Alert:
1577: The routine `MatSetValuesBlocked()` may offer much better efficiency
1578: for users of block sparse formats (`MATSEQBAIJ` and `MATMPIBAIJ`).
1580: This is currently not optimized for any particular `ISType`
1582: Developer Note:
1583: This is labeled with C so does not automatically generate Fortran stubs and interfaces
1584: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
1586: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatSetValues()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1587: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`
1588: @*/
1589: PetscErrorCode MatSetValuesIS(Mat mat, IS ism, IS isn, const PetscScalar v[], InsertMode addv)
1590: {
1591: PetscInt m, n;
1592: const PetscInt *rows, *cols;
1594: PetscFunctionBeginHot;
1596: PetscCall(ISGetIndices(ism, &rows));
1597: PetscCall(ISGetIndices(isn, &cols));
1598: PetscCall(ISGetLocalSize(ism, &m));
1599: PetscCall(ISGetLocalSize(isn, &n));
1600: PetscCall(MatSetValues(mat, m, rows, n, cols, v, addv));
1601: PetscCall(ISRestoreIndices(ism, &rows));
1602: PetscCall(ISRestoreIndices(isn, &cols));
1603: PetscFunctionReturn(PETSC_SUCCESS);
1604: }
1606: /*@
1607: MatSetValuesRowLocal - Inserts a row (block row for `MATBAIJ` matrices) of nonzero
1608: values into a matrix
1610: Not Collective
1612: Input Parameters:
1613: + mat - the matrix
1614: . row - the (block) row to set
1615: - v - a logically two-dimensional array of values
1617: Level: intermediate
1619: Notes:
1620: The values, `v`, are column-oriented (for the block version) and sorted
1622: All the nonzero values in `row` must be provided
1624: The matrix must have previously had its column indices set, likely by having been assembled.
1626: `row` must belong to this MPI process
1628: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1629: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`, `MatSetValues()`, `MatSetValuesRow()`, `MatSetLocalToGlobalMapping()`
1630: @*/
1631: PetscErrorCode MatSetValuesRowLocal(Mat mat, PetscInt row, const PetscScalar v[])
1632: {
1633: PetscInt globalrow;
1635: PetscFunctionBegin;
1638: PetscAssertPointer(v, 3);
1639: PetscCall(ISLocalToGlobalMappingApply(mat->rmap->mapping, 1, &row, &globalrow));
1640: PetscCall(MatSetValuesRow(mat, globalrow, v));
1641: PetscFunctionReturn(PETSC_SUCCESS);
1642: }
1644: /*@
1645: MatSetValuesRow - Inserts a row (block row for `MATBAIJ` matrices) of nonzero
1646: values into a matrix
1648: Not Collective
1650: Input Parameters:
1651: + mat - the matrix
1652: . row - the (block) row to set
1653: - 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
1655: Level: advanced
1657: Notes:
1658: The values, `v`, are column-oriented for the block version.
1660: All the nonzeros in `row` must be provided
1662: THE MATRIX MUST HAVE PREVIOUSLY HAD ITS COLUMN INDICES SET. IT IS RARE THAT THIS ROUTINE IS USED, usually `MatSetValues()` is used.
1664: `row` must belong to this process
1666: .seealso: [](ch_matrices), `Mat`, `MatSetValues()`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1667: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`
1668: @*/
1669: PetscErrorCode MatSetValuesRow(Mat mat, PetscInt row, const PetscScalar v[])
1670: {
1671: PetscFunctionBeginHot;
1674: MatCheckPreallocated(mat, 1);
1675: PetscAssertPointer(v, 3);
1676: PetscCheck(mat->insertmode != ADD_VALUES, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add and insert values");
1677: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
1678: mat->insertmode = INSERT_VALUES;
1680: if (mat->assembled) {
1681: mat->was_assembled = PETSC_TRUE;
1682: mat->assembled = PETSC_FALSE;
1683: }
1684: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
1685: PetscUseTypeMethod(mat, setvaluesrow, row, v);
1686: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
1687: PetscFunctionReturn(PETSC_SUCCESS);
1688: }
1690: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
1691: /*@
1692: MatSetValuesStencil - Inserts or adds a block of values into a matrix.
1693: Using structured grid indexing
1695: Not Collective
1697: Input Parameters:
1698: + mat - the matrix
1699: . m - number of rows being entered
1700: . idxm - grid coordinates (and component number when dof > 1) for matrix rows being entered
1701: . n - number of columns being entered
1702: . idxn - grid coordinates (and component number when dof > 1) for matrix columns being entered
1703: . v - a logically two-dimensional array of values
1704: - addv - either `ADD_VALUES` to add to existing entries at that location or `INSERT_VALUES` to replace existing entries with new values
1706: Level: beginner
1708: Notes:
1709: By default the values, `v`, are row-oriented. See `MatSetOption()` for other options.
1711: Calls to `MatSetValuesStencil()` with the `INSERT_VALUES` and `ADD_VALUES`
1712: options cannot be mixed without intervening calls to the assembly
1713: routines.
1715: The grid coordinates are across the entire grid, not just the local portion
1717: `MatSetValuesStencil()` uses 0-based row and column numbers in Fortran
1718: as well as in C.
1720: For setting/accessing vector values via array coordinates you can use the `DMDAVecGetArray()` routine
1722: In order to use this routine you must either obtain the matrix with `DMCreateMatrix()`
1723: or call `MatSetLocalToGlobalMapping()` and `MatSetStencil()` first.
1725: The columns and rows in the stencil passed in MUST be contained within the
1726: ghost region of the given process as set with DMDACreateXXX() or `MatSetStencil()`. For example,
1727: if you create a `DMDA` with an overlap of one grid level and on a particular process its first
1728: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1729: first i index you can use in your column and row indices in `MatSetStencil()` is 5.
1731: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
1732: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
1733: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
1734: `DM_BOUNDARY_PERIODIC` boundary type.
1736: 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
1737: a single value per point) you can skip filling those indices.
1739: Inspired by the structured grid interface to the HYPRE package
1740: (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)
1742: Efficiency Alert:
1743: The routine `MatSetValuesBlockedStencil()` may offer much better efficiency
1744: for users of block sparse formats (`MATSEQBAIJ` and `MATMPIBAIJ`).
1746: Fortran Note:
1747: `idxm` and `idxn` should be declared as
1748: $ MatStencil idxm(4,m),idxn(4,n)
1749: and the values inserted using
1750: .vb
1751: idxm(MatStencil_i,1) = i
1752: idxm(MatStencil_j,1) = j
1753: idxm(MatStencil_k,1) = k
1754: idxm(MatStencil_c,1) = c
1755: etc
1756: .ve
1758: .seealso: [](ch_matrices), `Mat`, `DMDA`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`
1759: `MatSetValues()`, `MatSetValuesBlockedStencil()`, `MatSetStencil()`, `DMCreateMatrix()`, `DMDAVecGetArray()`, `MatStencil`
1760: @*/
1761: PetscErrorCode MatSetValuesStencil(Mat mat, PetscInt m, const MatStencil idxm[], PetscInt n, const MatStencil idxn[], const PetscScalar v[], InsertMode addv)
1762: {
1763: PetscInt buf[8192], *bufm = NULL, *bufn = NULL, *jdxm, *jdxn;
1764: PetscInt j, i, dim = mat->stencil.dim, *dims = mat->stencil.dims + 1, tmp;
1765: PetscInt *starts = mat->stencil.starts, *dxm = (PetscInt *)idxm, *dxn = (PetscInt *)idxn, sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1767: PetscFunctionBegin;
1768: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1771: PetscAssertPointer(idxm, 3);
1772: PetscAssertPointer(idxn, 5);
1774: if ((m + n) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
1775: jdxm = buf;
1776: jdxn = buf + m;
1777: } else {
1778: PetscCall(PetscMalloc2(m, &bufm, n, &bufn));
1779: jdxm = bufm;
1780: jdxn = bufn;
1781: }
1782: for (i = 0; i < m; i++) {
1783: for (j = 0; j < 3 - sdim; j++) dxm++;
1784: tmp = *dxm++ - starts[0];
1785: for (j = 0; j < dim - 1; j++) {
1786: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1787: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
1788: }
1789: if (mat->stencil.noc) dxm++;
1790: jdxm[i] = tmp;
1791: }
1792: for (i = 0; i < n; i++) {
1793: for (j = 0; j < 3 - sdim; j++) dxn++;
1794: tmp = *dxn++ - starts[0];
1795: for (j = 0; j < dim - 1; j++) {
1796: if ((*dxn++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1797: else tmp = tmp * dims[j] + *(dxn - 1) - starts[j + 1];
1798: }
1799: if (mat->stencil.noc) dxn++;
1800: jdxn[i] = tmp;
1801: }
1802: PetscCall(MatSetValuesLocal(mat, m, jdxm, n, jdxn, v, addv));
1803: PetscCall(PetscFree2(bufm, bufn));
1804: PetscFunctionReturn(PETSC_SUCCESS);
1805: }
1807: /*@
1808: MatSetValuesBlockedStencil - Inserts or adds a block of values into a matrix.
1809: Using structured grid indexing
1811: Not Collective
1813: Input Parameters:
1814: + mat - the matrix
1815: . m - number of rows being entered
1816: . idxm - grid coordinates for matrix rows being entered
1817: . n - number of columns being entered
1818: . idxn - grid coordinates for matrix columns being entered
1819: . v - a logically two-dimensional array of values
1820: - addv - either `ADD_VALUES` to add to existing entries or `INSERT_VALUES` to replace existing entries with new values
1822: Level: beginner
1824: Notes:
1825: By default the values, `v`, are row-oriented and unsorted.
1826: See `MatSetOption()` for other options.
1828: Calls to `MatSetValuesBlockedStencil()` with the `INSERT_VALUES` and `ADD_VALUES`
1829: options cannot be mixed without intervening calls to the assembly
1830: routines.
1832: The grid coordinates are across the entire grid, not just the local portion
1834: `MatSetValuesBlockedStencil()` uses 0-based row and column numbers in Fortran
1835: as well as in C.
1837: For setting/accessing vector values via array coordinates you can use the `DMDAVecGetArray()` routine
1839: In order to use this routine you must either obtain the matrix with `DMCreateMatrix()`
1840: or call `MatSetBlockSize()`, `MatSetLocalToGlobalMapping()` and `MatSetStencil()` first.
1842: The columns and rows in the stencil passed in MUST be contained within the
1843: ghost region of the given process as set with DMDACreateXXX() or `MatSetStencil()`. For example,
1844: if you create a `DMDA` with an overlap of one grid level and on a particular process its first
1845: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1846: first i index you can use in your column and row indices in `MatSetStencil()` is 5.
1848: Negative indices may be passed in idxm and idxn, these rows and columns are
1849: simply ignored. This allows easily inserting element stiffness matrices
1850: with homogeneous Dirichlet boundary conditions that you don't want represented
1851: in the matrix.
1853: Inspired by the structured grid interface to the HYPRE package
1854: (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)
1856: Fortran Note:
1857: `idxm` and `idxn` should be declared as
1858: $ MatStencil idxm(4,m),idxn(4,n)
1859: and the values inserted using
1860: .vb
1861: idxm(MatStencil_i,1) = i
1862: idxm(MatStencil_j,1) = j
1863: idxm(MatStencil_k,1) = k
1864: etc
1865: .ve
1867: .seealso: [](ch_matrices), `Mat`, `DMDA`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`
1868: `MatSetValues()`, `MatSetValuesStencil()`, `MatSetStencil()`, `DMCreateMatrix()`, `DMDAVecGetArray()`, `MatStencil`,
1869: `MatSetBlockSize()`, `MatSetLocalToGlobalMapping()`
1870: @*/
1871: PetscErrorCode MatSetValuesBlockedStencil(Mat mat, PetscInt m, const MatStencil idxm[], PetscInt n, const MatStencil idxn[], const PetscScalar v[], InsertMode addv)
1872: {
1873: PetscInt buf[8192], *bufm = NULL, *bufn = NULL, *jdxm, *jdxn;
1874: PetscInt j, i, dim = mat->stencil.dim, *dims = mat->stencil.dims + 1, tmp;
1875: PetscInt *starts = mat->stencil.starts, *dxm = (PetscInt *)idxm, *dxn = (PetscInt *)idxn, sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1877: PetscFunctionBegin;
1878: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1881: PetscAssertPointer(idxm, 3);
1882: PetscAssertPointer(idxn, 5);
1883: PetscAssertPointer(v, 6);
1885: if ((m + n) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
1886: jdxm = buf;
1887: jdxn = buf + m;
1888: } else {
1889: PetscCall(PetscMalloc2(m, &bufm, n, &bufn));
1890: jdxm = bufm;
1891: jdxn = bufn;
1892: }
1893: for (i = 0; i < m; i++) {
1894: for (j = 0; j < 3 - sdim; j++) dxm++;
1895: tmp = *dxm++ - starts[0];
1896: for (j = 0; j < sdim - 1; j++) {
1897: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1898: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
1899: }
1900: dxm++;
1901: jdxm[i] = tmp;
1902: }
1903: for (i = 0; i < n; i++) {
1904: for (j = 0; j < 3 - sdim; j++) dxn++;
1905: tmp = *dxn++ - starts[0];
1906: for (j = 0; j < sdim - 1; j++) {
1907: if ((*dxn++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1908: else tmp = tmp * dims[j] + *(dxn - 1) - starts[j + 1];
1909: }
1910: dxn++;
1911: jdxn[i] = tmp;
1912: }
1913: PetscCall(MatSetValuesBlockedLocal(mat, m, jdxm, n, jdxn, v, addv));
1914: PetscCall(PetscFree2(bufm, bufn));
1915: PetscFunctionReturn(PETSC_SUCCESS);
1916: }
1918: /*@
1919: MatSetStencil - Sets the grid information for setting values into a matrix via
1920: `MatSetValuesStencil()`
1922: Not Collective
1924: Input Parameters:
1925: + mat - the matrix
1926: . dim - dimension of the grid 1, 2, or 3
1927: . dims - number of grid points in x, y, and z direction, including ghost points on your processor
1928: . starts - starting point of ghost nodes on your processor in x, y, and z direction
1929: - dof - number of degrees of freedom per node
1931: Level: beginner
1933: Notes:
1934: Inspired by the structured grid interface to the HYPRE package
1935: (www.llnl.gov/CASC/hyper)
1937: For matrices generated with `DMCreateMatrix()` this routine is automatically called and so not needed by the
1938: user.
1940: .seealso: [](ch_matrices), `Mat`, `MatStencil`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`
1941: `MatSetValues()`, `MatSetValuesBlockedStencil()`, `MatSetValuesStencil()`
1942: @*/
1943: PetscErrorCode MatSetStencil(Mat mat, PetscInt dim, const PetscInt dims[], const PetscInt starts[], PetscInt dof)
1944: {
1945: PetscFunctionBegin;
1947: PetscAssertPointer(dims, 3);
1948: PetscAssertPointer(starts, 4);
1950: mat->stencil.dim = dim + (dof > 1);
1951: for (PetscInt i = 0; i < dim; i++) {
1952: mat->stencil.dims[i] = dims[dim - i - 1]; /* copy the values in backwards */
1953: mat->stencil.starts[i] = starts[dim - i - 1];
1954: }
1955: mat->stencil.dims[dim] = dof;
1956: mat->stencil.starts[dim] = 0;
1957: mat->stencil.noc = (PetscBool)(dof == 1);
1958: PetscFunctionReturn(PETSC_SUCCESS);
1959: }
1961: /*@C
1962: MatSetValuesBlocked - Inserts or adds a block of values into a matrix.
1964: Not Collective
1966: Input Parameters:
1967: + mat - the matrix
1968: . v - a logically two-dimensional array of values
1969: . m - the number of block rows
1970: . idxm - the global block indices
1971: . n - the number of block columns
1972: . idxn - the global block indices
1973: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` replaces existing entries with new values
1975: Level: intermediate
1977: Notes:
1978: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call
1979: MatXXXXSetPreallocation() or `MatSetUp()` before using this routine.
1981: The `m` and `n` count the NUMBER of blocks in the row direction and column direction,
1982: NOT the total number of rows/columns; for example, if the block size is 2 and
1983: you are passing in values for rows 2,3,4,5 then `m` would be 2 (not 4).
1984: The values in `idxm` would be 1 2; that is the first index for each block divided by
1985: the block size.
1987: You must call `MatSetBlockSize()` when constructing this matrix (before
1988: preallocating it).
1990: By default the values, `v`, are row-oriented, so the layout of
1991: `v` is the same as for `MatSetValues()`. See `MatSetOption()` for other options.
1993: Calls to `MatSetValuesBlocked()` with the `INSERT_VALUES` and `ADD_VALUES`
1994: options cannot be mixed without intervening calls to the assembly
1995: routines.
1997: `MatSetValuesBlocked()` uses 0-based row and column numbers in Fortran
1998: as well as in C.
2000: Negative indices may be passed in `idxm` and `idxn`, these rows and columns are
2001: simply ignored. This allows easily inserting element stiffness matrices
2002: with homogeneous Dirichlet boundary conditions that you don't want represented
2003: in the matrix.
2005: Each time an entry is set within a sparse matrix via `MatSetValues()`,
2006: internal searching must be done to determine where to place the
2007: data in the matrix storage space. By instead inserting blocks of
2008: entries via `MatSetValuesBlocked()`, the overhead of matrix assembly is
2009: reduced.
2011: Example:
2012: .vb
2013: Suppose m=n=2 and block size(bs) = 2 The array is
2015: 1 2 | 3 4
2016: 5 6 | 7 8
2017: - - - | - - -
2018: 9 10 | 11 12
2019: 13 14 | 15 16
2021: v[] should be passed in like
2022: v[] = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]
2024: If you are not using row-oriented storage of v (that is you called MatSetOption(mat,MAT_ROW_ORIENTED,PETSC_FALSE)) then
2025: v[] = [1,5,9,13,2,6,10,14,3,7,11,15,4,8,12,16]
2026: .ve
2028: .seealso: [](ch_matrices), `Mat`, `MatSetBlockSize()`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetValuesBlockedLocal()`
2029: @*/
2030: PetscErrorCode MatSetValuesBlocked(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], const PetscScalar v[], InsertMode addv)
2031: {
2032: PetscFunctionBeginHot;
2035: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
2036: PetscAssertPointer(idxm, 3);
2037: PetscAssertPointer(idxn, 5);
2038: MatCheckPreallocated(mat, 1);
2039: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
2040: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
2041: if (PetscDefined(USE_DEBUG)) {
2042: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2043: PetscCheck(mat->ops->setvaluesblocked || mat->ops->setvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2044: }
2045: if (PetscDefined(USE_DEBUG)) {
2046: PetscInt rbs, cbs, M, N, i;
2047: PetscCall(MatGetBlockSizes(mat, &rbs, &cbs));
2048: PetscCall(MatGetSize(mat, &M, &N));
2049: for (i = 0; i < m; i++) PetscCheck(idxm[i] * rbs < M, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Row block %" PetscInt_FMT " contains an index %" PetscInt_FMT "*%" PetscInt_FMT " greater than row length %" PetscInt_FMT, i, idxm[i], rbs, M);
2050: for (i = 0; i < n; i++)
2051: PetscCheck(idxn[i] * cbs < N, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column block %" PetscInt_FMT " contains an index %" PetscInt_FMT "*%" PetscInt_FMT " greater than column length %" PetscInt_FMT, i, idxn[i], cbs, N);
2052: }
2053: if (mat->assembled) {
2054: mat->was_assembled = PETSC_TRUE;
2055: mat->assembled = PETSC_FALSE;
2056: }
2057: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
2058: if (mat->ops->setvaluesblocked) {
2059: PetscUseTypeMethod(mat, setvaluesblocked, m, idxm, n, idxn, v, addv);
2060: } else {
2061: PetscInt buf[8192], *bufr = NULL, *bufc = NULL, *iidxm, *iidxn;
2062: PetscInt i, j, bs, cbs;
2064: PetscCall(MatGetBlockSizes(mat, &bs, &cbs));
2065: if ((m * bs + n * cbs) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
2066: iidxm = buf;
2067: iidxn = buf + m * bs;
2068: } else {
2069: PetscCall(PetscMalloc2(m * bs, &bufr, n * cbs, &bufc));
2070: iidxm = bufr;
2071: iidxn = bufc;
2072: }
2073: for (i = 0; i < m; i++) {
2074: for (j = 0; j < bs; j++) iidxm[i * bs + j] = bs * idxm[i] + j;
2075: }
2076: if (m != n || bs != cbs || idxm != idxn) {
2077: for (i = 0; i < n; i++) {
2078: for (j = 0; j < cbs; j++) iidxn[i * cbs + j] = cbs * idxn[i] + j;
2079: }
2080: } else iidxn = iidxm;
2081: PetscCall(MatSetValues(mat, m * bs, iidxm, n * cbs, iidxn, v, addv));
2082: PetscCall(PetscFree2(bufr, bufc));
2083: }
2084: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
2085: PetscFunctionReturn(PETSC_SUCCESS);
2086: }
2088: /*@C
2089: MatGetValues - Gets a block of local values from a matrix.
2091: Not Collective; can only return values that are owned by the give process
2093: Input Parameters:
2094: + mat - the matrix
2095: . v - a logically two-dimensional array for storing the values
2096: . m - the number of rows
2097: . idxm - the global indices of the rows
2098: . n - the number of columns
2099: - idxn - the global indices of the columns
2101: Level: advanced
2103: Notes:
2104: The user must allocate space (m*n `PetscScalar`s) for the values, `v`.
2105: The values, `v`, are then returned in a row-oriented format,
2106: analogous to that used by default in `MatSetValues()`.
2108: `MatGetValues()` uses 0-based row and column numbers in
2109: Fortran as well as in C.
2111: `MatGetValues()` requires that the matrix has been assembled
2112: with `MatAssemblyBegin()`/`MatAssemblyEnd()`. Thus, calls to
2113: `MatSetValues()` and `MatGetValues()` CANNOT be made in succession
2114: without intermediate matrix assembly.
2116: Negative row or column indices will be ignored and those locations in `v` will be
2117: left unchanged.
2119: For the standard row-based matrix formats, `idxm` can only contain rows owned by the requesting MPI process.
2120: That is, rows with global index greater than or equal to rstart and less than rend where rstart and rend are obtainable
2121: from `MatGetOwnershipRange`(mat,&rstart,&rend).
2123: .seealso: [](ch_matrices), `Mat`, `MatGetRow()`, `MatCreateSubMatrices()`, `MatSetValues()`, `MatGetOwnershipRange()`, `MatGetValuesLocal()`, `MatGetValue()`
2124: @*/
2125: PetscErrorCode MatGetValues(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], PetscScalar v[])
2126: {
2127: PetscFunctionBegin;
2130: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS);
2131: PetscAssertPointer(idxm, 3);
2132: PetscAssertPointer(idxn, 5);
2133: PetscAssertPointer(v, 6);
2134: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2135: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2136: MatCheckPreallocated(mat, 1);
2138: PetscCall(PetscLogEventBegin(MAT_GetValues, mat, 0, 0, 0));
2139: PetscUseTypeMethod(mat, getvalues, m, idxm, n, idxn, v);
2140: PetscCall(PetscLogEventEnd(MAT_GetValues, mat, 0, 0, 0));
2141: PetscFunctionReturn(PETSC_SUCCESS);
2142: }
2144: /*@C
2145: MatGetValuesLocal - retrieves values from certain locations in a matrix using the local numbering of the indices
2146: defined previously by `MatSetLocalToGlobalMapping()`
2148: Not Collective
2150: Input Parameters:
2151: + mat - the matrix
2152: . nrow - number of rows
2153: . irow - the row local indices
2154: . ncol - number of columns
2155: - icol - the column local indices
2157: Output Parameter:
2158: . y - a logically two-dimensional array of values
2160: Level: advanced
2162: Notes:
2163: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call `MatSetLocalToGlobalMapping()` before using this routine.
2165: 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,
2166: are greater than or equal to rstart and less than rend where rstart and rend are obtainable from `MatGetOwnershipRange`(mat,&rstart,&rend). One can
2167: determine if the resulting global row associated with the local row r is owned by the requesting MPI process by applying the `ISLocalToGlobalMapping` set
2168: with `MatSetLocalToGlobalMapping()`.
2170: Developer Note:
2171: This is labelled with C so does not automatically generate Fortran stubs and interfaces
2172: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2174: .seealso: [](ch_matrices), `Mat`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetLocalToGlobalMapping()`,
2175: `MatSetValuesLocal()`, `MatGetValues()`
2176: @*/
2177: PetscErrorCode MatGetValuesLocal(Mat mat, PetscInt nrow, const PetscInt irow[], PetscInt ncol, const PetscInt icol[], PetscScalar y[])
2178: {
2179: PetscFunctionBeginHot;
2182: MatCheckPreallocated(mat, 1);
2183: if (!nrow || !ncol) PetscFunctionReturn(PETSC_SUCCESS); /* no values to retrieve */
2184: PetscAssertPointer(irow, 3);
2185: PetscAssertPointer(icol, 5);
2186: if (PetscDefined(USE_DEBUG)) {
2187: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2188: PetscCheck(mat->ops->getvalueslocal || mat->ops->getvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2189: }
2190: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2191: PetscCall(PetscLogEventBegin(MAT_GetValues, mat, 0, 0, 0));
2192: if (mat->ops->getvalueslocal) PetscUseTypeMethod(mat, getvalueslocal, nrow, irow, ncol, icol, y);
2193: else {
2194: PetscInt buf[8192], *bufr = NULL, *bufc = NULL, *irowm, *icolm;
2195: if ((nrow + ncol) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
2196: irowm = buf;
2197: icolm = buf + nrow;
2198: } else {
2199: PetscCall(PetscMalloc2(nrow, &bufr, ncol, &bufc));
2200: irowm = bufr;
2201: icolm = bufc;
2202: }
2203: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "MatGetValuesLocal() cannot proceed without local-to-global row mapping (See MatSetLocalToGlobalMapping()).");
2204: PetscCheck(mat->cmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "MatGetValuesLocal() cannot proceed without local-to-global column mapping (See MatSetLocalToGlobalMapping()).");
2205: PetscCall(ISLocalToGlobalMappingApply(mat->rmap->mapping, nrow, irow, irowm));
2206: PetscCall(ISLocalToGlobalMappingApply(mat->cmap->mapping, ncol, icol, icolm));
2207: PetscCall(MatGetValues(mat, nrow, irowm, ncol, icolm, y));
2208: PetscCall(PetscFree2(bufr, bufc));
2209: }
2210: PetscCall(PetscLogEventEnd(MAT_GetValues, mat, 0, 0, 0));
2211: PetscFunctionReturn(PETSC_SUCCESS);
2212: }
2214: /*@
2215: MatSetValuesBatch - Adds (`ADD_VALUES`) many blocks of values into a matrix at once. The blocks must all be square and
2216: the same size. Currently, this can only be called once and creates the given matrix.
2218: Not Collective
2220: Input Parameters:
2221: + mat - the matrix
2222: . nb - the number of blocks
2223: . bs - the number of rows (and columns) in each block
2224: . rows - a concatenation of the rows for each block
2225: - v - a concatenation of logically two-dimensional arrays of values
2227: Level: advanced
2229: Notes:
2230: `MatSetPreallocationCOO()` and `MatSetValuesCOO()` may be a better way to provide the values
2232: In the future, we will extend this routine to handle rectangular blocks, and to allow multiple calls for a given matrix.
2234: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
2235: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`, `MatSetValues()`, `MatSetPreallocationCOO()`, `MatSetValuesCOO()`
2236: @*/
2237: PetscErrorCode MatSetValuesBatch(Mat mat, PetscInt nb, PetscInt bs, PetscInt rows[], const PetscScalar v[])
2238: {
2239: PetscFunctionBegin;
2242: PetscAssertPointer(rows, 4);
2243: PetscAssertPointer(v, 5);
2244: PetscAssert(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2246: PetscCall(PetscLogEventBegin(MAT_SetValuesBatch, mat, 0, 0, 0));
2247: if (mat->ops->setvaluesbatch) PetscUseTypeMethod(mat, setvaluesbatch, nb, bs, rows, v);
2248: else {
2249: for (PetscInt b = 0; b < nb; ++b) PetscCall(MatSetValues(mat, bs, &rows[b * bs], bs, &rows[b * bs], &v[b * bs * bs], ADD_VALUES));
2250: }
2251: PetscCall(PetscLogEventEnd(MAT_SetValuesBatch, mat, 0, 0, 0));
2252: PetscFunctionReturn(PETSC_SUCCESS);
2253: }
2255: /*@
2256: MatSetLocalToGlobalMapping - Sets a local-to-global numbering for use by
2257: the routine `MatSetValuesLocal()` to allow users to insert matrix entries
2258: using a local (per-processor) numbering.
2260: Not Collective
2262: Input Parameters:
2263: + x - the matrix
2264: . rmapping - row mapping created with `ISLocalToGlobalMappingCreate()` or `ISLocalToGlobalMappingCreateIS()`
2265: - cmapping - column mapping
2267: Level: intermediate
2269: Note:
2270: If the matrix is obtained with `DMCreateMatrix()` then this may already have been called on the matrix
2272: .seealso: [](ch_matrices), `Mat`, `DM`, `DMCreateMatrix()`, `MatGetLocalToGlobalMapping()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetValuesLocal()`, `MatGetValuesLocal()`
2273: @*/
2274: PetscErrorCode MatSetLocalToGlobalMapping(Mat x, ISLocalToGlobalMapping rmapping, ISLocalToGlobalMapping cmapping)
2275: {
2276: PetscFunctionBegin;
2281: if (x->ops->setlocaltoglobalmapping) PetscUseTypeMethod(x, setlocaltoglobalmapping, rmapping, cmapping);
2282: else {
2283: PetscCall(PetscLayoutSetISLocalToGlobalMapping(x->rmap, rmapping));
2284: PetscCall(PetscLayoutSetISLocalToGlobalMapping(x->cmap, cmapping));
2285: }
2286: PetscFunctionReturn(PETSC_SUCCESS);
2287: }
2289: /*@
2290: MatGetLocalToGlobalMapping - Gets the local-to-global numbering set by `MatSetLocalToGlobalMapping()`
2292: Not Collective
2294: Input Parameter:
2295: . A - the matrix
2297: Output Parameters:
2298: + rmapping - row mapping
2299: - cmapping - column mapping
2301: Level: advanced
2303: .seealso: [](ch_matrices), `Mat`, `MatSetLocalToGlobalMapping()`, `MatSetValuesLocal()`
2304: @*/
2305: PetscErrorCode MatGetLocalToGlobalMapping(Mat A, ISLocalToGlobalMapping *rmapping, ISLocalToGlobalMapping *cmapping)
2306: {
2307: PetscFunctionBegin;
2310: if (rmapping) {
2311: PetscAssertPointer(rmapping, 2);
2312: *rmapping = A->rmap->mapping;
2313: }
2314: if (cmapping) {
2315: PetscAssertPointer(cmapping, 3);
2316: *cmapping = A->cmap->mapping;
2317: }
2318: PetscFunctionReturn(PETSC_SUCCESS);
2319: }
2321: /*@
2322: MatSetLayouts - Sets the `PetscLayout` objects for rows and columns of a matrix
2324: Logically Collective
2326: Input Parameters:
2327: + A - the matrix
2328: . rmap - row layout
2329: - cmap - column layout
2331: Level: advanced
2333: Note:
2334: The `PetscLayout` objects are usually created automatically for the matrix so this routine rarely needs to be called.
2336: .seealso: [](ch_matrices), `Mat`, `PetscLayout`, `MatCreateVecs()`, `MatGetLocalToGlobalMapping()`, `MatGetLayouts()`
2337: @*/
2338: PetscErrorCode MatSetLayouts(Mat A, PetscLayout rmap, PetscLayout cmap)
2339: {
2340: PetscFunctionBegin;
2342: PetscCall(PetscLayoutReference(rmap, &A->rmap));
2343: PetscCall(PetscLayoutReference(cmap, &A->cmap));
2344: PetscFunctionReturn(PETSC_SUCCESS);
2345: }
2347: /*@
2348: MatGetLayouts - Gets the `PetscLayout` objects for rows and columns
2350: Not Collective
2352: Input Parameter:
2353: . A - the matrix
2355: Output Parameters:
2356: + rmap - row layout
2357: - cmap - column layout
2359: Level: advanced
2361: .seealso: [](ch_matrices), `Mat`, [Matrix Layouts](sec_matlayout), `PetscLayout`, `MatCreateVecs()`, `MatGetLocalToGlobalMapping()`, `MatSetLayouts()`
2362: @*/
2363: PetscErrorCode MatGetLayouts(Mat A, PetscLayout *rmap, PetscLayout *cmap)
2364: {
2365: PetscFunctionBegin;
2368: if (rmap) {
2369: PetscAssertPointer(rmap, 2);
2370: *rmap = A->rmap;
2371: }
2372: if (cmap) {
2373: PetscAssertPointer(cmap, 3);
2374: *cmap = A->cmap;
2375: }
2376: PetscFunctionReturn(PETSC_SUCCESS);
2377: }
2379: /*@C
2380: MatSetValuesLocal - Inserts or adds values into certain locations of a matrix,
2381: using a local numbering of the rows and columns.
2383: Not Collective
2385: Input Parameters:
2386: + mat - the matrix
2387: . nrow - number of rows
2388: . irow - the row local indices
2389: . ncol - number of columns
2390: . icol - the column local indices
2391: . y - a logically two-dimensional array of values
2392: - addv - either `INSERT_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
2394: Level: intermediate
2396: Notes:
2397: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call `MatSetLocalToGlobalMapping()` before using this routine
2399: Calls to `MatSetValuesLocal()` with the `INSERT_VALUES` and `ADD_VALUES`
2400: options cannot be mixed without intervening calls to the assembly
2401: routines.
2403: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
2404: MUST be called after all calls to `MatSetValuesLocal()` have been completed.
2406: Developer Note:
2407: This is labeled with C so does not automatically generate Fortran stubs and interfaces
2408: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2410: .seealso: [](ch_matrices), `Mat`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetLocalToGlobalMapping()`,
2411: `MatGetValuesLocal()`
2412: @*/
2413: PetscErrorCode MatSetValuesLocal(Mat mat, PetscInt nrow, const PetscInt irow[], PetscInt ncol, const PetscInt icol[], const PetscScalar y[], InsertMode addv)
2414: {
2415: PetscFunctionBeginHot;
2418: MatCheckPreallocated(mat, 1);
2419: if (!nrow || !ncol) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
2420: PetscAssertPointer(irow, 3);
2421: PetscAssertPointer(icol, 5);
2422: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
2423: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
2424: if (PetscDefined(USE_DEBUG)) {
2425: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2426: PetscCheck(mat->ops->setvalueslocal || mat->ops->setvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2427: }
2429: if (mat->assembled) {
2430: mat->was_assembled = PETSC_TRUE;
2431: mat->assembled = PETSC_FALSE;
2432: }
2433: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
2434: if (mat->ops->setvalueslocal) PetscUseTypeMethod(mat, setvalueslocal, nrow, irow, ncol, icol, y, addv);
2435: else {
2436: PetscInt buf[8192], *bufr = NULL, *bufc = NULL;
2437: const PetscInt *irowm, *icolm;
2439: if ((!mat->rmap->mapping && !mat->cmap->mapping) || (nrow + ncol) <= (PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf)) {
2440: bufr = buf;
2441: bufc = buf + nrow;
2442: irowm = bufr;
2443: icolm = bufc;
2444: } else {
2445: PetscCall(PetscMalloc2(nrow, &bufr, ncol, &bufc));
2446: irowm = bufr;
2447: icolm = bufc;
2448: }
2449: if (mat->rmap->mapping) PetscCall(ISLocalToGlobalMappingApply(mat->rmap->mapping, nrow, irow, bufr));
2450: else irowm = irow;
2451: if (mat->cmap->mapping) {
2452: if (mat->cmap->mapping != mat->rmap->mapping || ncol != nrow || icol != irow) {
2453: PetscCall(ISLocalToGlobalMappingApply(mat->cmap->mapping, ncol, icol, bufc));
2454: } else icolm = irowm;
2455: } else icolm = icol;
2456: PetscCall(MatSetValues(mat, nrow, irowm, ncol, icolm, y, addv));
2457: if (bufr != buf) PetscCall(PetscFree2(bufr, bufc));
2458: }
2459: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
2460: PetscFunctionReturn(PETSC_SUCCESS);
2461: }
2463: /*@C
2464: MatSetValuesBlockedLocal - Inserts or adds values into certain locations of a matrix,
2465: using a local ordering of the nodes a block at a time.
2467: Not Collective
2469: Input Parameters:
2470: + mat - the matrix
2471: . nrow - number of rows
2472: . irow - the row local indices
2473: . ncol - number of columns
2474: . icol - the column local indices
2475: . y - a logically two-dimensional array of values
2476: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
2478: Level: intermediate
2480: Notes:
2481: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call `MatSetBlockSize()` and `MatSetLocalToGlobalMapping()`
2482: before using this routineBefore calling `MatSetValuesLocal()`, the user must first set the
2484: Calls to `MatSetValuesBlockedLocal()` with the `INSERT_VALUES` and `ADD_VALUES`
2485: options cannot be mixed without intervening calls to the assembly
2486: routines.
2488: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
2489: MUST be called after all calls to `MatSetValuesBlockedLocal()` have been completed.
2491: Developer Note:
2492: This is labeled with C so does not automatically generate Fortran stubs and interfaces
2493: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2495: .seealso: [](ch_matrices), `Mat`, `MatSetBlockSize()`, `MatSetLocalToGlobalMapping()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`,
2496: `MatSetValuesLocal()`, `MatSetValuesBlocked()`
2497: @*/
2498: PetscErrorCode MatSetValuesBlockedLocal(Mat mat, PetscInt nrow, const PetscInt irow[], PetscInt ncol, const PetscInt icol[], const PetscScalar y[], InsertMode addv)
2499: {
2500: PetscFunctionBeginHot;
2503: MatCheckPreallocated(mat, 1);
2504: if (!nrow || !ncol) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
2505: PetscAssertPointer(irow, 3);
2506: PetscAssertPointer(icol, 5);
2507: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
2508: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
2509: if (PetscDefined(USE_DEBUG)) {
2510: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2511: 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);
2512: }
2514: if (mat->assembled) {
2515: mat->was_assembled = PETSC_TRUE;
2516: mat->assembled = PETSC_FALSE;
2517: }
2518: if (PetscUnlikelyDebug(mat->rmap->mapping)) { /* Condition on the mapping existing, because MatSetValuesBlockedLocal_IS does not require it to be set. */
2519: PetscInt irbs, rbs;
2520: PetscCall(MatGetBlockSizes(mat, &rbs, NULL));
2521: PetscCall(ISLocalToGlobalMappingGetBlockSize(mat->rmap->mapping, &irbs));
2522: PetscCheck(rbs == irbs, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Different row block sizes! mat %" PetscInt_FMT ", row l2g map %" PetscInt_FMT, rbs, irbs);
2523: }
2524: if (PetscUnlikelyDebug(mat->cmap->mapping)) {
2525: PetscInt icbs, cbs;
2526: PetscCall(MatGetBlockSizes(mat, NULL, &cbs));
2527: PetscCall(ISLocalToGlobalMappingGetBlockSize(mat->cmap->mapping, &icbs));
2528: PetscCheck(cbs == icbs, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Different col block sizes! mat %" PetscInt_FMT ", col l2g map %" PetscInt_FMT, cbs, icbs);
2529: }
2530: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
2531: if (mat->ops->setvaluesblockedlocal) PetscUseTypeMethod(mat, setvaluesblockedlocal, nrow, irow, ncol, icol, y, addv);
2532: else {
2533: PetscInt buf[8192], *bufr = NULL, *bufc = NULL;
2534: const PetscInt *irowm, *icolm;
2536: if ((!mat->rmap->mapping && !mat->cmap->mapping) || (nrow + ncol) <= ((PetscInt)PETSC_STATIC_ARRAY_LENGTH(buf))) {
2537: bufr = buf;
2538: bufc = buf + nrow;
2539: irowm = bufr;
2540: icolm = bufc;
2541: } else {
2542: PetscCall(PetscMalloc2(nrow, &bufr, ncol, &bufc));
2543: irowm = bufr;
2544: icolm = bufc;
2545: }
2546: if (mat->rmap->mapping) PetscCall(ISLocalToGlobalMappingApplyBlock(mat->rmap->mapping, nrow, irow, bufr));
2547: else irowm = irow;
2548: if (mat->cmap->mapping) {
2549: if (mat->cmap->mapping != mat->rmap->mapping || ncol != nrow || icol != irow) {
2550: PetscCall(ISLocalToGlobalMappingApplyBlock(mat->cmap->mapping, ncol, icol, bufc));
2551: } else icolm = irowm;
2552: } else icolm = icol;
2553: PetscCall(MatSetValuesBlocked(mat, nrow, irowm, ncol, icolm, y, addv));
2554: if (bufr != buf) PetscCall(PetscFree2(bufr, bufc));
2555: }
2556: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
2557: PetscFunctionReturn(PETSC_SUCCESS);
2558: }
2560: /*@
2561: MatMultDiagonalBlock - Computes the matrix-vector product, $y = Dx$. Where `D` is defined by the inode or block structure of the diagonal
2563: Collective
2565: Input Parameters:
2566: + mat - the matrix
2567: - x - the vector to be multiplied
2569: Output Parameter:
2570: . y - the result
2572: Level: developer
2574: Note:
2575: The vectors `x` and `y` cannot be the same. I.e., one cannot
2576: call `MatMultDiagonalBlock`(A,y,y).
2578: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
2579: @*/
2580: PetscErrorCode MatMultDiagonalBlock(Mat mat, Vec x, Vec y)
2581: {
2582: PetscFunctionBegin;
2588: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2589: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2590: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2591: MatCheckPreallocated(mat, 1);
2593: PetscUseTypeMethod(mat, multdiagonalblock, x, y);
2594: PetscCall(PetscObjectStateIncrease((PetscObject)y));
2595: PetscFunctionReturn(PETSC_SUCCESS);
2596: }
2598: /*@
2599: MatMult - Computes the matrix-vector product, $y = Ax$.
2601: Neighbor-wise Collective
2603: Input Parameters:
2604: + mat - the matrix
2605: - x - the vector to be multiplied
2607: Output Parameter:
2608: . y - the result
2610: Level: beginner
2612: Note:
2613: The vectors `x` and `y` cannot be the same. I.e., one cannot
2614: call `MatMult`(A,y,y).
2616: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
2617: @*/
2618: PetscErrorCode MatMult(Mat mat, Vec x, Vec y)
2619: {
2620: PetscFunctionBegin;
2624: VecCheckAssembled(x);
2626: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2627: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2628: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2629: 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);
2630: 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);
2631: 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);
2632: 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);
2633: PetscCall(VecSetErrorIfLocked(y, 3));
2634: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(x, 2, PETSC_TRUE));
2635: MatCheckPreallocated(mat, 1);
2637: PetscCall(VecLockReadPush(x));
2638: PetscCall(PetscLogEventBegin(MAT_Mult, mat, x, y, 0));
2639: PetscUseTypeMethod(mat, mult, x, y);
2640: PetscCall(PetscLogEventEnd(MAT_Mult, mat, x, y, 0));
2641: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(y, 3, PETSC_FALSE));
2642: PetscCall(VecLockReadPop(x));
2643: PetscFunctionReturn(PETSC_SUCCESS);
2644: }
2646: /*@
2647: MatMultTranspose - Computes matrix transpose times a vector $y = A^T * x$.
2649: Neighbor-wise Collective
2651: Input Parameters:
2652: + mat - the matrix
2653: - x - the vector to be multiplied
2655: Output Parameter:
2656: . y - the result
2658: Level: beginner
2660: Notes:
2661: The vectors `x` and `y` cannot be the same. I.e., one cannot
2662: call `MatMultTranspose`(A,y,y).
2664: For complex numbers this does NOT compute the Hermitian (complex conjugate) transpose multiple,
2665: use `MatMultHermitianTranspose()`
2667: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatMultHermitianTranspose()`, `MatTranspose()`
2668: @*/
2669: PetscErrorCode MatMultTranspose(Mat mat, Vec x, Vec y)
2670: {
2671: PetscErrorCode (*op)(Mat, Vec, Vec) = NULL;
2673: PetscFunctionBegin;
2677: VecCheckAssembled(x);
2680: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2681: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2682: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2683: 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);
2684: 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);
2685: 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);
2686: 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);
2687: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(x, 2, PETSC_TRUE));
2688: MatCheckPreallocated(mat, 1);
2690: if (!mat->ops->multtranspose) {
2691: if (mat->symmetric == PETSC_BOOL3_TRUE && mat->ops->mult) op = mat->ops->mult;
2692: 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);
2693: } else op = mat->ops->multtranspose;
2694: PetscCall(PetscLogEventBegin(MAT_MultTranspose, mat, x, y, 0));
2695: PetscCall(VecLockReadPush(x));
2696: PetscCall((*op)(mat, x, y));
2697: PetscCall(VecLockReadPop(x));
2698: PetscCall(PetscLogEventEnd(MAT_MultTranspose, mat, x, y, 0));
2699: PetscCall(PetscObjectStateIncrease((PetscObject)y));
2700: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(y, 3, PETSC_FALSE));
2701: PetscFunctionReturn(PETSC_SUCCESS);
2702: }
2704: /*@
2705: MatMultHermitianTranspose - Computes matrix Hermitian-transpose times a vector $y = A^H * x$.
2707: Neighbor-wise Collective
2709: Input Parameters:
2710: + mat - the matrix
2711: - x - the vector to be multiplied
2713: Output Parameter:
2714: . y - the result
2716: Level: beginner
2718: Notes:
2719: The vectors `x` and `y` cannot be the same. I.e., one cannot
2720: call `MatMultHermitianTranspose`(A,y,y).
2722: Also called the conjugate transpose, complex conjugate transpose, or adjoint.
2724: For real numbers `MatMultTranspose()` and `MatMultHermitianTranspose()` are identical.
2726: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatMultAdd()`, `MatMultHermitianTransposeAdd()`, `MatMultTranspose()`
2727: @*/
2728: PetscErrorCode MatMultHermitianTranspose(Mat mat, Vec x, Vec y)
2729: {
2730: PetscFunctionBegin;
2736: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2737: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2738: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2739: 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);
2740: 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);
2741: 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);
2742: 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);
2743: MatCheckPreallocated(mat, 1);
2745: PetscCall(PetscLogEventBegin(MAT_MultHermitianTranspose, mat, x, y, 0));
2746: #if defined(PETSC_USE_COMPLEX)
2747: if (mat->ops->multhermitiantranspose || (mat->hermitian == PETSC_BOOL3_TRUE && mat->ops->mult)) {
2748: PetscCall(VecLockReadPush(x));
2749: if (mat->ops->multhermitiantranspose) PetscUseTypeMethod(mat, multhermitiantranspose, x, y);
2750: else PetscUseTypeMethod(mat, mult, x, y);
2751: PetscCall(VecLockReadPop(x));
2752: } else {
2753: Vec w;
2754: PetscCall(VecDuplicate(x, &w));
2755: PetscCall(VecCopy(x, w));
2756: PetscCall(VecConjugate(w));
2757: PetscCall(MatMultTranspose(mat, w, y));
2758: PetscCall(VecDestroy(&w));
2759: PetscCall(VecConjugate(y));
2760: }
2761: PetscCall(PetscObjectStateIncrease((PetscObject)y));
2762: #else
2763: PetscCall(MatMultTranspose(mat, x, y));
2764: #endif
2765: PetscCall(PetscLogEventEnd(MAT_MultHermitianTranspose, mat, x, y, 0));
2766: PetscFunctionReturn(PETSC_SUCCESS);
2767: }
2769: /*@
2770: MatMultAdd - Computes $v3 = v2 + A * v1$.
2772: Neighbor-wise Collective
2774: Input Parameters:
2775: + mat - the matrix
2776: . v1 - the vector to be multiplied by `mat`
2777: - v2 - the vector to be added to the result
2779: Output Parameter:
2780: . v3 - the result
2782: Level: beginner
2784: Note:
2785: The vectors `v1` and `v3` cannot be the same. I.e., one cannot
2786: call `MatMultAdd`(A,v1,v2,v1).
2788: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMult()`, `MatMultTransposeAdd()`
2789: @*/
2790: PetscErrorCode MatMultAdd(Mat mat, Vec v1, Vec v2, Vec v3)
2791: {
2792: PetscFunctionBegin;
2799: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2800: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2801: 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);
2802: /* 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);
2803: 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); */
2804: 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);
2805: 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);
2806: PetscCheck(v1 != v3, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "v1 and v3 must be different vectors");
2807: MatCheckPreallocated(mat, 1);
2809: PetscCall(PetscLogEventBegin(MAT_MultAdd, mat, v1, v2, v3));
2810: PetscCall(VecLockReadPush(v1));
2811: PetscUseTypeMethod(mat, multadd, v1, v2, v3);
2812: PetscCall(VecLockReadPop(v1));
2813: PetscCall(PetscLogEventEnd(MAT_MultAdd, mat, v1, v2, v3));
2814: PetscCall(PetscObjectStateIncrease((PetscObject)v3));
2815: PetscFunctionReturn(PETSC_SUCCESS);
2816: }
2818: /*@
2819: MatMultTransposeAdd - Computes $v3 = v2 + A^T * v1$.
2821: Neighbor-wise Collective
2823: Input Parameters:
2824: + mat - the matrix
2825: . v1 - the vector to be multiplied by the transpose of the matrix
2826: - v2 - the vector to be added to the result
2828: Output Parameter:
2829: . v3 - the result
2831: Level: beginner
2833: Note:
2834: The vectors `v1` and `v3` cannot be the same. I.e., one cannot
2835: call `MatMultTransposeAdd`(A,v1,v2,v1).
2837: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMult()`
2838: @*/
2839: PetscErrorCode MatMultTransposeAdd(Mat mat, Vec v1, Vec v2, Vec v3)
2840: {
2841: PetscErrorCode (*op)(Mat, Vec, Vec, Vec) = (!mat->ops->multtransposeadd && mat->symmetric) ? mat->ops->multadd : mat->ops->multtransposeadd;
2843: PetscFunctionBegin;
2850: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2851: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2852: 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);
2853: 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);
2854: 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);
2855: PetscCheck(v1 != v3, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "v1 and v3 must be different vectors");
2856: PetscCheck(op, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2857: MatCheckPreallocated(mat, 1);
2859: PetscCall(PetscLogEventBegin(MAT_MultTransposeAdd, mat, v1, v2, v3));
2860: PetscCall(VecLockReadPush(v1));
2861: PetscCall((*op)(mat, v1, v2, v3));
2862: PetscCall(VecLockReadPop(v1));
2863: PetscCall(PetscLogEventEnd(MAT_MultTransposeAdd, mat, v1, v2, v3));
2864: PetscCall(PetscObjectStateIncrease((PetscObject)v3));
2865: PetscFunctionReturn(PETSC_SUCCESS);
2866: }
2868: /*@
2869: MatMultHermitianTransposeAdd - Computes $v3 = v2 + A^H * v1$.
2871: Neighbor-wise Collective
2873: Input Parameters:
2874: + mat - the matrix
2875: . v1 - the vector to be multiplied by the Hermitian transpose
2876: - v2 - the vector to be added to the result
2878: Output Parameter:
2879: . v3 - the result
2881: Level: beginner
2883: Note:
2884: The vectors `v1` and `v3` cannot be the same. I.e., one cannot
2885: call `MatMultHermitianTransposeAdd`(A,v1,v2,v1).
2887: .seealso: [](ch_matrices), `Mat`, `MatMultHermitianTranspose()`, `MatMultTranspose()`, `MatMultAdd()`, `MatMult()`
2888: @*/
2889: PetscErrorCode MatMultHermitianTransposeAdd(Mat mat, Vec v1, Vec v2, Vec v3)
2890: {
2891: PetscFunctionBegin;
2898: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2899: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2900: PetscCheck(v1 != v3, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "v1 and v3 must be different vectors");
2901: 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);
2902: 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);
2903: 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);
2904: MatCheckPreallocated(mat, 1);
2906: PetscCall(PetscLogEventBegin(MAT_MultHermitianTransposeAdd, mat, v1, v2, v3));
2907: PetscCall(VecLockReadPush(v1));
2908: if (mat->ops->multhermitiantransposeadd) PetscUseTypeMethod(mat, multhermitiantransposeadd, v1, v2, v3);
2909: else {
2910: Vec w, z;
2911: PetscCall(VecDuplicate(v1, &w));
2912: PetscCall(VecCopy(v1, w));
2913: PetscCall(VecConjugate(w));
2914: PetscCall(VecDuplicate(v3, &z));
2915: PetscCall(MatMultTranspose(mat, w, z));
2916: PetscCall(VecDestroy(&w));
2917: PetscCall(VecConjugate(z));
2918: if (v2 != v3) {
2919: PetscCall(VecWAXPY(v3, 1.0, v2, z));
2920: } else {
2921: PetscCall(VecAXPY(v3, 1.0, z));
2922: }
2923: PetscCall(VecDestroy(&z));
2924: }
2925: PetscCall(VecLockReadPop(v1));
2926: PetscCall(PetscLogEventEnd(MAT_MultHermitianTransposeAdd, mat, v1, v2, v3));
2927: PetscCall(PetscObjectStateIncrease((PetscObject)v3));
2928: PetscFunctionReturn(PETSC_SUCCESS);
2929: }
2931: /*@
2932: MatGetFactorType - gets the type of factorization a matrix is
2934: Not Collective
2936: Input Parameter:
2937: . mat - the matrix
2939: Output Parameter:
2940: . 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`
2942: Level: intermediate
2944: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorType`, `MatGetFactor()`, `MatSetFactorType()`, `MAT_FACTOR_NONE`, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ILU`,
2945: `MAT_FACTOR_ICC`,`MAT_FACTOR_ILUDT`, `MAT_FACTOR_QR`
2946: @*/
2947: PetscErrorCode MatGetFactorType(Mat mat, MatFactorType *t)
2948: {
2949: PetscFunctionBegin;
2952: PetscAssertPointer(t, 2);
2953: *t = mat->factortype;
2954: PetscFunctionReturn(PETSC_SUCCESS);
2955: }
2957: /*@
2958: MatSetFactorType - sets the type of factorization a matrix is
2960: Logically Collective
2962: Input Parameters:
2963: + mat - the matrix
2964: - 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`
2966: Level: intermediate
2968: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorType`, `MatGetFactor()`, `MatGetFactorType()`, `MAT_FACTOR_NONE`, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ILU`,
2969: `MAT_FACTOR_ICC`,`MAT_FACTOR_ILUDT`, `MAT_FACTOR_QR`
2970: @*/
2971: PetscErrorCode MatSetFactorType(Mat mat, MatFactorType t)
2972: {
2973: PetscFunctionBegin;
2976: mat->factortype = t;
2977: PetscFunctionReturn(PETSC_SUCCESS);
2978: }
2980: /*@C
2981: MatGetInfo - Returns information about matrix storage (number of
2982: nonzeros, memory, etc.).
2984: Collective if `MAT_GLOBAL_MAX` or `MAT_GLOBAL_SUM` is used as the flag
2986: Input Parameters:
2987: + mat - the matrix
2988: - 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)
2990: Output Parameter:
2991: . info - matrix information context
2993: Options Database Key:
2994: . -mat_view ::ascii_info - print matrix info to `PETSC_STDOUT`
2996: Notes:
2997: The `MatInfo` context contains a variety of matrix data, including
2998: number of nonzeros allocated and used, number of mallocs during
2999: matrix assembly, etc. Additional information for factored matrices
3000: is provided (such as the fill ratio, number of mallocs during
3001: factorization, etc.).
3003: Example:
3004: See the file ${PETSC_DIR}/include/petscmat.h for a complete list of
3005: data within the MatInfo context. For example,
3006: .vb
3007: MatInfo info;
3008: Mat A;
3009: double mal, nz_a, nz_u;
3011: MatGetInfo(A, MAT_LOCAL, &info);
3012: mal = info.mallocs;
3013: nz_a = info.nz_allocated;
3014: .ve
3016: Fortran users should declare info as a double precision
3017: array of dimension `MAT_INFO_SIZE`, and then extract the parameters
3018: of interest. See the file ${PETSC_DIR}/include/petsc/finclude/petscmat.h
3019: a complete list of parameter names.
3020: .vb
3021: double precision info(MAT_INFO_SIZE)
3022: double precision mal, nz_a
3023: Mat A
3024: integer ierr
3026: call MatGetInfo(A, MAT_LOCAL, info, ierr)
3027: mal = info(MAT_INFO_MALLOCS)
3028: nz_a = info(MAT_INFO_NZ_ALLOCATED)
3029: .ve
3031: Level: intermediate
3033: Developer Note:
3034: The Fortran interface is not autogenerated as the
3035: interface definition cannot be generated correctly [due to `MatInfo` argument]
3037: .seealso: [](ch_matrices), `Mat`, `MatInfo`, `MatStashGetInfo()`
3038: @*/
3039: PetscErrorCode MatGetInfo(Mat mat, MatInfoType flag, MatInfo *info)
3040: {
3041: PetscFunctionBegin;
3044: PetscAssertPointer(info, 3);
3045: MatCheckPreallocated(mat, 1);
3046: PetscUseTypeMethod(mat, getinfo, flag, info);
3047: PetscFunctionReturn(PETSC_SUCCESS);
3048: }
3050: /*
3051: This is used by external packages where it is not easy to get the info from the actual
3052: matrix factorization.
3053: */
3054: PetscErrorCode MatGetInfo_External(Mat A, MatInfoType flag, MatInfo *info)
3055: {
3056: PetscFunctionBegin;
3057: PetscCall(PetscMemzero(info, sizeof(MatInfo)));
3058: PetscFunctionReturn(PETSC_SUCCESS);
3059: }
3061: /*@C
3062: MatLUFactor - Performs in-place LU factorization of matrix.
3064: Collective
3066: Input Parameters:
3067: + mat - the matrix
3068: . row - row permutation
3069: . col - column permutation
3070: - info - options for factorization, includes
3071: .vb
3072: fill - expected fill as ratio of original fill.
3073: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3074: Run with the option -info to determine an optimal value to use
3075: .ve
3077: Level: developer
3079: Notes:
3080: Most users should employ the `KSP` interface for linear solvers
3081: instead of working directly with matrix algebra routines such as this.
3082: See, e.g., `KSPCreate()`.
3084: This changes the state of the matrix to a factored matrix; it cannot be used
3085: for example with `MatSetValues()` unless one first calls `MatSetUnfactored()`.
3087: This is really in-place only for dense matrices, the preferred approach is to use `MatGetFactor()`, `MatLUFactorSymbolic()`, and `MatLUFactorNumeric()`
3088: when not using `KSP`.
3090: Developer Note:
3091: The Fortran interface is not autogenerated as the
3092: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3094: .seealso: [](ch_matrices), [Matrix Factorization](sec_matfactor), `Mat`, `MatFactorType`, `MatLUFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`,
3095: `MatGetOrdering()`, `MatSetUnfactored()`, `MatFactorInfo`, `MatGetFactor()`
3096: @*/
3097: PetscErrorCode MatLUFactor(Mat mat, IS row, IS col, const MatFactorInfo *info)
3098: {
3099: MatFactorInfo tinfo;
3101: PetscFunctionBegin;
3105: if (info) PetscAssertPointer(info, 4);
3107: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3108: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3109: MatCheckPreallocated(mat, 1);
3110: if (!info) {
3111: PetscCall(MatFactorInfoInitialize(&tinfo));
3112: info = &tinfo;
3113: }
3115: PetscCall(PetscLogEventBegin(MAT_LUFactor, mat, row, col, 0));
3116: PetscUseTypeMethod(mat, lufactor, row, col, info);
3117: PetscCall(PetscLogEventEnd(MAT_LUFactor, mat, row, col, 0));
3118: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3119: PetscFunctionReturn(PETSC_SUCCESS);
3120: }
3122: /*@C
3123: MatILUFactor - Performs in-place ILU factorization of matrix.
3125: Collective
3127: Input Parameters:
3128: + mat - the matrix
3129: . row - row permutation
3130: . col - column permutation
3131: - info - structure containing
3132: .vb
3133: levels - number of levels of fill.
3134: expected fill - as ratio of original fill.
3135: 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
3136: missing diagonal entries)
3137: .ve
3139: Level: developer
3141: Notes:
3142: Most users should employ the `KSP` interface for linear solvers
3143: instead of working directly with matrix algebra routines such as this.
3144: See, e.g., `KSPCreate()`.
3146: Probably really in-place only when level of fill is zero, otherwise allocates
3147: new space to store factored matrix and deletes previous memory. The preferred approach is to use `MatGetFactor()`, `MatILUFactorSymbolic()`, and `MatILUFactorNumeric()`
3148: when not using `KSP`.
3150: Developer Note:
3151: The Fortran interface is not autogenerated as the
3152: interface definition cannot be generated correctly [due to MatFactorInfo]
3154: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatILUFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`, `MatFactorInfo`
3155: @*/
3156: PetscErrorCode MatILUFactor(Mat mat, IS row, IS col, const MatFactorInfo *info)
3157: {
3158: PetscFunctionBegin;
3162: PetscAssertPointer(info, 4);
3164: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "matrix must be square");
3165: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3166: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3167: MatCheckPreallocated(mat, 1);
3169: PetscCall(PetscLogEventBegin(MAT_ILUFactor, mat, row, col, 0));
3170: PetscUseTypeMethod(mat, ilufactor, row, col, info);
3171: PetscCall(PetscLogEventEnd(MAT_ILUFactor, mat, row, col, 0));
3172: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3173: PetscFunctionReturn(PETSC_SUCCESS);
3174: }
3176: /*@C
3177: MatLUFactorSymbolic - Performs symbolic LU factorization of matrix.
3178: Call this routine before calling `MatLUFactorNumeric()` and after `MatGetFactor()`.
3180: Collective
3182: Input Parameters:
3183: + fact - the factor matrix obtained with `MatGetFactor()`
3184: . mat - the matrix
3185: . row - the row permutation
3186: . col - the column permutation
3187: - info - options for factorization, includes
3188: .vb
3189: fill - expected fill as ratio of original fill. Run with the option -info to determine an optimal value to use
3190: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3191: .ve
3193: Level: developer
3195: Notes:
3196: See [Matrix Factorization](sec_matfactor) for additional information about factorizations
3198: Most users should employ the simplified `KSP` interface for linear solvers
3199: instead of working directly with matrix algebra routines such as this.
3200: See, e.g., `KSPCreate()`.
3202: Developer Note:
3203: The Fortran interface is not autogenerated as the
3204: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3206: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatLUFactor()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`, `MatFactorInfo`, `MatFactorInfoInitialize()`
3207: @*/
3208: PetscErrorCode MatLUFactorSymbolic(Mat fact, Mat mat, IS row, IS col, const MatFactorInfo *info)
3209: {
3210: MatFactorInfo tinfo;
3212: PetscFunctionBegin;
3217: if (info) PetscAssertPointer(info, 5);
3220: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3221: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3222: MatCheckPreallocated(mat, 2);
3223: if (!info) {
3224: PetscCall(MatFactorInfoInitialize(&tinfo));
3225: info = &tinfo;
3226: }
3228: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_LUFactorSymbolic, mat, row, col, 0));
3229: PetscUseTypeMethod(fact, lufactorsymbolic, mat, row, col, info);
3230: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_LUFactorSymbolic, mat, row, col, 0));
3231: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3232: PetscFunctionReturn(PETSC_SUCCESS);
3233: }
3235: /*@C
3236: MatLUFactorNumeric - Performs numeric LU factorization of a matrix.
3237: Call this routine after first calling `MatLUFactorSymbolic()` and `MatGetFactor()`.
3239: Collective
3241: Input Parameters:
3242: + fact - the factor matrix obtained with `MatGetFactor()`
3243: . mat - the matrix
3244: - info - options for factorization
3246: Level: developer
3248: Notes:
3249: See `MatLUFactor()` for in-place factorization. See
3250: `MatCholeskyFactorNumeric()` for the symmetric, positive definite case.
3252: Most users should employ the `KSP` interface for linear solvers
3253: instead of working directly with matrix algebra routines such as this.
3254: See, e.g., `KSPCreate()`.
3256: Developer Note:
3257: The Fortran interface is not autogenerated as the
3258: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3260: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatFactorInfo`, `MatLUFactorSymbolic()`, `MatLUFactor()`, `MatCholeskyFactor()`
3261: @*/
3262: PetscErrorCode MatLUFactorNumeric(Mat fact, Mat mat, const MatFactorInfo *info)
3263: {
3264: MatFactorInfo tinfo;
3266: PetscFunctionBegin;
3271: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3272: 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,
3273: mat->rmap->N, (fact)->rmap->N, mat->cmap->N, (fact)->cmap->N);
3275: MatCheckPreallocated(mat, 2);
3276: if (!info) {
3277: PetscCall(MatFactorInfoInitialize(&tinfo));
3278: info = &tinfo;
3279: }
3281: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_LUFactorNumeric, mat, fact, 0, 0));
3282: else PetscCall(PetscLogEventBegin(MAT_LUFactor, mat, fact, 0, 0));
3283: PetscUseTypeMethod(fact, lufactornumeric, mat, info);
3284: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_LUFactorNumeric, mat, fact, 0, 0));
3285: else PetscCall(PetscLogEventEnd(MAT_LUFactor, mat, fact, 0, 0));
3286: PetscCall(MatViewFromOptions(fact, NULL, "-mat_factor_view"));
3287: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3288: PetscFunctionReturn(PETSC_SUCCESS);
3289: }
3291: /*@C
3292: MatCholeskyFactor - Performs in-place Cholesky factorization of a
3293: symmetric matrix.
3295: Collective
3297: Input Parameters:
3298: + mat - the matrix
3299: . perm - row and column permutations
3300: - info - expected fill as ratio of original fill
3302: Level: developer
3304: Notes:
3305: See `MatLUFactor()` for the nonsymmetric case. See also `MatGetFactor()`,
3306: `MatCholeskyFactorSymbolic()`, and `MatCholeskyFactorNumeric()`.
3308: Most users should employ the `KSP` interface for linear solvers
3309: instead of working directly with matrix algebra routines such as this.
3310: See, e.g., `KSPCreate()`.
3312: Developer Note:
3313: The Fortran interface is not autogenerated as the
3314: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3316: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatFactorInfo`, `MatLUFactor()`, `MatCholeskyFactorSymbolic()`, `MatCholeskyFactorNumeric()`
3317: `MatGetOrdering()`
3318: @*/
3319: PetscErrorCode MatCholeskyFactor(Mat mat, IS perm, const MatFactorInfo *info)
3320: {
3321: MatFactorInfo tinfo;
3323: PetscFunctionBegin;
3326: if (info) PetscAssertPointer(info, 3);
3328: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "Matrix must be square");
3329: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3330: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3331: MatCheckPreallocated(mat, 1);
3332: if (!info) {
3333: PetscCall(MatFactorInfoInitialize(&tinfo));
3334: info = &tinfo;
3335: }
3337: PetscCall(PetscLogEventBegin(MAT_CholeskyFactor, mat, perm, 0, 0));
3338: PetscUseTypeMethod(mat, choleskyfactor, perm, info);
3339: PetscCall(PetscLogEventEnd(MAT_CholeskyFactor, mat, perm, 0, 0));
3340: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3341: PetscFunctionReturn(PETSC_SUCCESS);
3342: }
3344: /*@C
3345: MatCholeskyFactorSymbolic - Performs symbolic Cholesky factorization
3346: of a symmetric matrix.
3348: Collective
3350: Input Parameters:
3351: + fact - the factor matrix obtained with `MatGetFactor()`
3352: . mat - the matrix
3353: . perm - row and column permutations
3354: - info - options for factorization, includes
3355: .vb
3356: fill - expected fill as ratio of original fill.
3357: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3358: Run with the option -info to determine an optimal value to use
3359: .ve
3361: Level: developer
3363: Notes:
3364: See `MatLUFactorSymbolic()` for the nonsymmetric case. See also
3365: `MatCholeskyFactor()` and `MatCholeskyFactorNumeric()`.
3367: Most users should employ the `KSP` interface for linear solvers
3368: instead of working directly with matrix algebra routines such as this.
3369: See, e.g., `KSPCreate()`.
3371: Developer Note:
3372: The Fortran interface is not autogenerated as the
3373: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3375: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactor()`, `MatCholeskyFactorNumeric()`
3376: `MatGetOrdering()`
3377: @*/
3378: PetscErrorCode MatCholeskyFactorSymbolic(Mat fact, Mat mat, IS perm, const MatFactorInfo *info)
3379: {
3380: MatFactorInfo tinfo;
3382: PetscFunctionBegin;
3386: if (info) PetscAssertPointer(info, 4);
3389: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "Matrix must be square");
3390: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3391: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3392: MatCheckPreallocated(mat, 2);
3393: if (!info) {
3394: PetscCall(MatFactorInfoInitialize(&tinfo));
3395: info = &tinfo;
3396: }
3398: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_CholeskyFactorSymbolic, mat, perm, 0, 0));
3399: PetscUseTypeMethod(fact, choleskyfactorsymbolic, mat, perm, info);
3400: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_CholeskyFactorSymbolic, mat, perm, 0, 0));
3401: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3402: PetscFunctionReturn(PETSC_SUCCESS);
3403: }
3405: /*@C
3406: MatCholeskyFactorNumeric - Performs numeric Cholesky factorization
3407: of a symmetric matrix. Call this routine after first calling `MatGetFactor()` and
3408: `MatCholeskyFactorSymbolic()`.
3410: Collective
3412: Input Parameters:
3413: + fact - the factor matrix obtained with `MatGetFactor()`, where the factored values are stored
3414: . mat - the initial matrix that is to be factored
3415: - info - options for factorization
3417: Level: developer
3419: Note:
3420: Most users should employ the `KSP` interface for linear solvers
3421: instead of working directly with matrix algebra routines such as this.
3422: See, e.g., `KSPCreate()`.
3424: Developer Note:
3425: The Fortran interface is not autogenerated as the
3426: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3428: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatCholeskyFactorSymbolic()`, `MatCholeskyFactor()`, `MatLUFactorNumeric()`
3429: @*/
3430: PetscErrorCode MatCholeskyFactorNumeric(Mat fact, Mat mat, const MatFactorInfo *info)
3431: {
3432: MatFactorInfo tinfo;
3434: PetscFunctionBegin;
3439: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3440: 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,
3441: mat->rmap->N, (fact)->rmap->N, mat->cmap->N, (fact)->cmap->N);
3442: MatCheckPreallocated(mat, 2);
3443: if (!info) {
3444: PetscCall(MatFactorInfoInitialize(&tinfo));
3445: info = &tinfo;
3446: }
3448: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_CholeskyFactorNumeric, mat, fact, 0, 0));
3449: else PetscCall(PetscLogEventBegin(MAT_CholeskyFactor, mat, fact, 0, 0));
3450: PetscUseTypeMethod(fact, choleskyfactornumeric, mat, info);
3451: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_CholeskyFactorNumeric, mat, fact, 0, 0));
3452: else PetscCall(PetscLogEventEnd(MAT_CholeskyFactor, mat, fact, 0, 0));
3453: PetscCall(MatViewFromOptions(fact, NULL, "-mat_factor_view"));
3454: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3455: PetscFunctionReturn(PETSC_SUCCESS);
3456: }
3458: /*@
3459: MatQRFactor - Performs in-place QR factorization of matrix.
3461: Collective
3463: Input Parameters:
3464: + mat - the matrix
3465: . col - column permutation
3466: - info - options for factorization, includes
3467: .vb
3468: fill - expected fill as ratio of original fill.
3469: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3470: Run with the option -info to determine an optimal value to use
3471: .ve
3473: Level: developer
3475: Notes:
3476: Most users should employ the `KSP` interface for linear solvers
3477: instead of working directly with matrix algebra routines such as this.
3478: See, e.g., `KSPCreate()`.
3480: This changes the state of the matrix to a factored matrix; it cannot be used
3481: for example with `MatSetValues()` unless one first calls `MatSetUnfactored()`.
3483: Developer Note:
3484: The Fortran interface is not autogenerated as the
3485: interface definition cannot be generated correctly [due to MatFactorInfo]
3487: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatQRFactorSymbolic()`, `MatQRFactorNumeric()`, `MatLUFactor()`,
3488: `MatSetUnfactored()`
3489: @*/
3490: PetscErrorCode MatQRFactor(Mat mat, IS col, const MatFactorInfo *info)
3491: {
3492: PetscFunctionBegin;
3495: if (info) PetscAssertPointer(info, 3);
3497: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3498: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3499: MatCheckPreallocated(mat, 1);
3500: PetscCall(PetscLogEventBegin(MAT_QRFactor, mat, col, 0, 0));
3501: PetscUseMethod(mat, "MatQRFactor_C", (Mat, IS, const MatFactorInfo *), (mat, col, info));
3502: PetscCall(PetscLogEventEnd(MAT_QRFactor, mat, col, 0, 0));
3503: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3504: PetscFunctionReturn(PETSC_SUCCESS);
3505: }
3507: /*@
3508: MatQRFactorSymbolic - Performs symbolic QR factorization of matrix.
3509: Call this routine after `MatGetFactor()` but before calling `MatQRFactorNumeric()`.
3511: Collective
3513: Input Parameters:
3514: + fact - the factor matrix obtained with `MatGetFactor()`
3515: . mat - the matrix
3516: . col - column permutation
3517: - info - options for factorization, includes
3518: .vb
3519: fill - expected fill as ratio of original fill.
3520: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3521: Run with the option -info to determine an optimal value to use
3522: .ve
3524: Level: developer
3526: Note:
3527: Most users should employ the `KSP` interface for linear solvers
3528: instead of working directly with matrix algebra routines such as this.
3529: See, e.g., `KSPCreate()`.
3531: Developer Note:
3532: The Fortran interface is not autogenerated as the
3533: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3535: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatFactorInfo`, `MatQRFactor()`, `MatQRFactorNumeric()`, `MatLUFactor()`, `MatFactorInfoInitialize()`
3536: @*/
3537: PetscErrorCode MatQRFactorSymbolic(Mat fact, Mat mat, IS col, const MatFactorInfo *info)
3538: {
3539: MatFactorInfo tinfo;
3541: PetscFunctionBegin;
3545: if (info) PetscAssertPointer(info, 4);
3548: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3549: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3550: MatCheckPreallocated(mat, 2);
3551: if (!info) {
3552: PetscCall(MatFactorInfoInitialize(&tinfo));
3553: info = &tinfo;
3554: }
3556: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_QRFactorSymbolic, fact, mat, col, 0));
3557: PetscUseMethod(fact, "MatQRFactorSymbolic_C", (Mat, Mat, IS, const MatFactorInfo *), (fact, mat, col, info));
3558: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_QRFactorSymbolic, fact, mat, col, 0));
3559: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3560: PetscFunctionReturn(PETSC_SUCCESS);
3561: }
3563: /*@
3564: MatQRFactorNumeric - Performs numeric QR factorization of a matrix.
3565: Call this routine after first calling `MatGetFactor()`, and `MatQRFactorSymbolic()`.
3567: Collective
3569: Input Parameters:
3570: + fact - the factor matrix obtained with `MatGetFactor()`
3571: . mat - the matrix
3572: - info - options for factorization
3574: Level: developer
3576: Notes:
3577: See `MatQRFactor()` for in-place factorization.
3579: Most users should employ the `KSP` interface for linear solvers
3580: instead of working directly with matrix algebra routines such as this.
3581: See, e.g., `KSPCreate()`.
3583: Developer Note:
3584: The Fortran interface is not autogenerated as the
3585: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3587: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatQRFactor()`, `MatQRFactorSymbolic()`, `MatLUFactor()`
3588: @*/
3589: PetscErrorCode MatQRFactorNumeric(Mat fact, Mat mat, const MatFactorInfo *info)
3590: {
3591: MatFactorInfo tinfo;
3593: PetscFunctionBegin;
3598: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3599: 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,
3600: mat->rmap->N, (fact)->rmap->N, mat->cmap->N, (fact)->cmap->N);
3602: MatCheckPreallocated(mat, 2);
3603: if (!info) {
3604: PetscCall(MatFactorInfoInitialize(&tinfo));
3605: info = &tinfo;
3606: }
3608: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_QRFactorNumeric, mat, fact, 0, 0));
3609: else PetscCall(PetscLogEventBegin(MAT_QRFactor, mat, fact, 0, 0));
3610: PetscUseMethod(fact, "MatQRFactorNumeric_C", (Mat, Mat, const MatFactorInfo *), (fact, mat, info));
3611: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_QRFactorNumeric, mat, fact, 0, 0));
3612: else PetscCall(PetscLogEventEnd(MAT_QRFactor, mat, fact, 0, 0));
3613: PetscCall(MatViewFromOptions(fact, NULL, "-mat_factor_view"));
3614: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3615: PetscFunctionReturn(PETSC_SUCCESS);
3616: }
3618: /*@
3619: MatSolve - Solves $A x = b$, given a factored matrix.
3621: Neighbor-wise Collective
3623: Input Parameters:
3624: + mat - the factored matrix
3625: - b - the right-hand-side vector
3627: Output Parameter:
3628: . x - the result vector
3630: Level: developer
3632: Notes:
3633: The vectors `b` and `x` cannot be the same. I.e., one cannot
3634: call `MatSolve`(A,x,x).
3636: Most users should employ the `KSP` interface for linear solvers
3637: instead of working directly with matrix algebra routines such as this.
3638: See, e.g., `KSPCreate()`.
3640: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatLUFactor()`, `MatSolveAdd()`, `MatSolveTranspose()`, `MatSolveTransposeAdd()`
3641: @*/
3642: PetscErrorCode MatSolve(Mat mat, Vec b, Vec x)
3643: {
3644: PetscFunctionBegin;
3649: PetscCheckSameComm(mat, 1, b, 2);
3650: PetscCheckSameComm(mat, 1, x, 3);
3651: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3652: 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);
3653: 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);
3654: 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);
3655: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3656: MatCheckPreallocated(mat, 1);
3658: PetscCall(PetscLogEventBegin(MAT_Solve, mat, b, x, 0));
3659: if (mat->factorerrortype) {
3660: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
3661: PetscCall(VecSetInf(x));
3662: } else PetscUseTypeMethod(mat, solve, b, x);
3663: PetscCall(PetscLogEventEnd(MAT_Solve, mat, b, x, 0));
3664: PetscCall(PetscObjectStateIncrease((PetscObject)x));
3665: PetscFunctionReturn(PETSC_SUCCESS);
3666: }
3668: static PetscErrorCode MatMatSolve_Basic(Mat A, Mat B, Mat X, PetscBool trans)
3669: {
3670: Vec b, x;
3671: PetscInt N, i;
3672: PetscErrorCode (*f)(Mat, Vec, Vec);
3673: PetscBool Abound, Bneedconv = PETSC_FALSE, Xneedconv = PETSC_FALSE;
3675: PetscFunctionBegin;
3676: if (A->factorerrortype) {
3677: PetscCall(PetscInfo(A, "MatFactorError %d\n", A->factorerrortype));
3678: PetscCall(MatSetInf(X));
3679: PetscFunctionReturn(PETSC_SUCCESS);
3680: }
3681: f = (!trans || (!A->ops->solvetranspose && A->symmetric)) ? A->ops->solve : A->ops->solvetranspose;
3682: PetscCheck(f, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Mat type %s", ((PetscObject)A)->type_name);
3683: PetscCall(MatBoundToCPU(A, &Abound));
3684: if (!Abound) {
3685: PetscCall(PetscObjectTypeCompareAny((PetscObject)B, &Bneedconv, MATSEQDENSE, MATMPIDENSE, ""));
3686: PetscCall(PetscObjectTypeCompareAny((PetscObject)X, &Xneedconv, MATSEQDENSE, MATMPIDENSE, ""));
3687: }
3688: #if PetscDefined(HAVE_CUDA)
3689: if (Bneedconv) PetscCall(MatConvert(B, MATDENSECUDA, MAT_INPLACE_MATRIX, &B));
3690: if (Xneedconv) PetscCall(MatConvert(X, MATDENSECUDA, MAT_INPLACE_MATRIX, &X));
3691: #elif PetscDefined(HAVE_HIP)
3692: if (Bneedconv) PetscCall(MatConvert(B, MATDENSEHIP, MAT_INPLACE_MATRIX, &B));
3693: if (Xneedconv) PetscCall(MatConvert(X, MATDENSEHIP, MAT_INPLACE_MATRIX, &X));
3694: #endif
3695: PetscCall(MatGetSize(B, NULL, &N));
3696: for (i = 0; i < N; i++) {
3697: PetscCall(MatDenseGetColumnVecRead(B, i, &b));
3698: PetscCall(MatDenseGetColumnVecWrite(X, i, &x));
3699: PetscCall((*f)(A, b, x));
3700: PetscCall(MatDenseRestoreColumnVecWrite(X, i, &x));
3701: PetscCall(MatDenseRestoreColumnVecRead(B, i, &b));
3702: }
3703: if (Bneedconv) PetscCall(MatConvert(B, MATDENSE, MAT_INPLACE_MATRIX, &B));
3704: if (Xneedconv) PetscCall(MatConvert(X, MATDENSE, MAT_INPLACE_MATRIX, &X));
3705: PetscFunctionReturn(PETSC_SUCCESS);
3706: }
3708: /*@
3709: MatMatSolve - Solves $A X = B$, given a factored matrix.
3711: Neighbor-wise Collective
3713: Input Parameters:
3714: + A - the factored matrix
3715: - B - the right-hand-side matrix `MATDENSE` (or sparse `MATAIJ`-- when using MUMPS)
3717: Output Parameter:
3718: . X - the result matrix (dense matrix)
3720: Level: developer
3722: Note:
3723: If `B` is a `MATDENSE` matrix then one can call `MatMatSolve`(A,B,B) except with `MATSOLVERMKL_CPARDISO`;
3724: otherwise, `B` and `X` cannot be the same.
3726: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSolve()`, `MatMatSolveTranspose()`, `MatLUFactor()`, `MatCholeskyFactor()`
3727: @*/
3728: PetscErrorCode MatMatSolve(Mat A, Mat B, Mat X)
3729: {
3730: PetscFunctionBegin;
3735: PetscCheckSameComm(A, 1, B, 2);
3736: PetscCheckSameComm(A, 1, X, 3);
3737: 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);
3738: 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);
3739: 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");
3740: if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3741: MatCheckPreallocated(A, 1);
3743: PetscCall(PetscLogEventBegin(MAT_MatSolve, A, B, X, 0));
3744: if (!A->ops->matsolve) {
3745: PetscCall(PetscInfo(A, "Mat type %s using basic MatMatSolve\n", ((PetscObject)A)->type_name));
3746: PetscCall(MatMatSolve_Basic(A, B, X, PETSC_FALSE));
3747: } else PetscUseTypeMethod(A, matsolve, B, X);
3748: PetscCall(PetscLogEventEnd(MAT_MatSolve, A, B, X, 0));
3749: PetscCall(PetscObjectStateIncrease((PetscObject)X));
3750: PetscFunctionReturn(PETSC_SUCCESS);
3751: }
3753: /*@
3754: MatMatSolveTranspose - Solves $A^T X = B $, given a factored matrix.
3756: Neighbor-wise Collective
3758: Input Parameters:
3759: + A - the factored matrix
3760: - B - the right-hand-side matrix (`MATDENSE` matrix)
3762: Output Parameter:
3763: . X - the result matrix (dense matrix)
3765: Level: developer
3767: Note:
3768: The matrices `B` and `X` cannot be the same. I.e., one cannot
3769: call `MatMatSolveTranspose`(A,X,X).
3771: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSolveTranspose()`, `MatMatSolve()`, `MatLUFactor()`, `MatCholeskyFactor()`
3772: @*/
3773: PetscErrorCode MatMatSolveTranspose(Mat A, Mat B, Mat X)
3774: {
3775: PetscFunctionBegin;
3780: PetscCheckSameComm(A, 1, B, 2);
3781: PetscCheckSameComm(A, 1, X, 3);
3782: PetscCheck(X != B, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_IDN, "X and B must be different matrices");
3783: 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);
3784: 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);
3785: 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);
3786: 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");
3787: if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3788: MatCheckPreallocated(A, 1);
3790: PetscCall(PetscLogEventBegin(MAT_MatSolve, A, B, X, 0));
3791: if (!A->ops->matsolvetranspose) {
3792: PetscCall(PetscInfo(A, "Mat type %s using basic MatMatSolveTranspose\n", ((PetscObject)A)->type_name));
3793: PetscCall(MatMatSolve_Basic(A, B, X, PETSC_TRUE));
3794: } else PetscUseTypeMethod(A, matsolvetranspose, B, X);
3795: PetscCall(PetscLogEventEnd(MAT_MatSolve, A, B, X, 0));
3796: PetscCall(PetscObjectStateIncrease((PetscObject)X));
3797: PetscFunctionReturn(PETSC_SUCCESS);
3798: }
3800: /*@
3801: MatMatTransposeSolve - Solves $A X = B^T$, given a factored matrix.
3803: Neighbor-wise Collective
3805: Input Parameters:
3806: + A - the factored matrix
3807: - Bt - the transpose of right-hand-side matrix as a `MATDENSE`
3809: Output Parameter:
3810: . X - the result matrix (dense matrix)
3812: Level: developer
3814: Note:
3815: 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
3816: format on the host processor and call `MatMatTransposeSolve()` to implement MUMPS' `MatMatSolve()`.
3818: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatMatSolve()`, `MatMatSolveTranspose()`, `MatLUFactor()`, `MatCholeskyFactor()`
3819: @*/
3820: PetscErrorCode MatMatTransposeSolve(Mat A, Mat Bt, Mat X)
3821: {
3822: PetscFunctionBegin;
3827: PetscCheckSameComm(A, 1, Bt, 2);
3828: PetscCheckSameComm(A, 1, X, 3);
3830: PetscCheck(X != Bt, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_IDN, "X and B must be different matrices");
3831: 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);
3832: 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);
3833: 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");
3834: if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3835: PetscCheck(A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
3836: MatCheckPreallocated(A, 1);
3838: PetscCall(PetscLogEventBegin(MAT_MatTrSolve, A, Bt, X, 0));
3839: PetscUseTypeMethod(A, mattransposesolve, Bt, X);
3840: PetscCall(PetscLogEventEnd(MAT_MatTrSolve, A, Bt, X, 0));
3841: PetscCall(PetscObjectStateIncrease((PetscObject)X));
3842: PetscFunctionReturn(PETSC_SUCCESS);
3843: }
3845: /*@
3846: MatForwardSolve - Solves $ L x = b $, given a factored matrix, $A = LU $, or
3847: $U^T*D^(1/2) x = b$, given a factored symmetric matrix, $A = U^T*D*U$,
3849: Neighbor-wise Collective
3851: Input Parameters:
3852: + mat - the factored matrix
3853: - b - the right-hand-side vector
3855: Output Parameter:
3856: . x - the result vector
3858: Level: developer
3860: Notes:
3861: `MatSolve()` should be used for most applications, as it performs
3862: a forward solve followed by a backward solve.
3864: The vectors `b` and `x` cannot be the same, i.e., one cannot
3865: call `MatForwardSolve`(A,x,x).
3867: For matrix in `MATSEQBAIJ` format with block size larger than 1,
3868: the diagonal blocks are not implemented as $D = D^(1/2) * D^(1/2)$ yet.
3869: `MatForwardSolve()` solves $U^T*D y = b$, and
3870: `MatBackwardSolve()` solves $U x = y$.
3871: Thus they do not provide a symmetric preconditioner.
3873: .seealso: [](ch_matrices), `Mat`, `MatBackwardSolve()`, `MatGetFactor()`, `MatSolve()`
3874: @*/
3875: PetscErrorCode MatForwardSolve(Mat mat, Vec b, Vec x)
3876: {
3877: PetscFunctionBegin;
3882: PetscCheckSameComm(mat, 1, b, 2);
3883: PetscCheckSameComm(mat, 1, x, 3);
3884: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3885: 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);
3886: 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);
3887: 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);
3888: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3889: MatCheckPreallocated(mat, 1);
3891: PetscCall(PetscLogEventBegin(MAT_ForwardSolve, mat, b, x, 0));
3892: PetscUseTypeMethod(mat, forwardsolve, b, x);
3893: PetscCall(PetscLogEventEnd(MAT_ForwardSolve, mat, b, x, 0));
3894: PetscCall(PetscObjectStateIncrease((PetscObject)x));
3895: PetscFunctionReturn(PETSC_SUCCESS);
3896: }
3898: /*@
3899: MatBackwardSolve - Solves $U x = b$, given a factored matrix, $A = LU$.
3900: $D^(1/2) U x = b$, given a factored symmetric matrix, $A = U^T*D*U$,
3902: Neighbor-wise Collective
3904: Input Parameters:
3905: + mat - the factored matrix
3906: - b - the right-hand-side vector
3908: Output Parameter:
3909: . x - the result vector
3911: Level: developer
3913: Notes:
3914: `MatSolve()` should be used for most applications, as it performs
3915: a forward solve followed by a backward solve.
3917: The vectors `b` and `x` cannot be the same. I.e., one cannot
3918: call `MatBackwardSolve`(A,x,x).
3920: For matrix in `MATSEQBAIJ` format with block size larger than 1,
3921: the diagonal blocks are not implemented as $D = D^(1/2) * D^(1/2)$ yet.
3922: `MatForwardSolve()` solves $U^T*D y = b$, and
3923: `MatBackwardSolve()` solves $U x = y$.
3924: Thus they do not provide a symmetric preconditioner.
3926: .seealso: [](ch_matrices), `Mat`, `MatForwardSolve()`, `MatGetFactor()`, `MatSolve()`
3927: @*/
3928: PetscErrorCode MatBackwardSolve(Mat mat, Vec b, Vec x)
3929: {
3930: PetscFunctionBegin;
3935: PetscCheckSameComm(mat, 1, b, 2);
3936: PetscCheckSameComm(mat, 1, x, 3);
3937: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3938: 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);
3939: 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);
3940: 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);
3941: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3942: MatCheckPreallocated(mat, 1);
3944: PetscCall(PetscLogEventBegin(MAT_BackwardSolve, mat, b, x, 0));
3945: PetscUseTypeMethod(mat, backwardsolve, b, x);
3946: PetscCall(PetscLogEventEnd(MAT_BackwardSolve, mat, b, x, 0));
3947: PetscCall(PetscObjectStateIncrease((PetscObject)x));
3948: PetscFunctionReturn(PETSC_SUCCESS);
3949: }
3951: /*@
3952: MatSolveAdd - Computes $x = y + A^{-1}*b$, given a factored matrix.
3954: Neighbor-wise Collective
3956: Input Parameters:
3957: + mat - the factored matrix
3958: . b - the right-hand-side vector
3959: - y - the vector to be added to
3961: Output Parameter:
3962: . x - the result vector
3964: Level: developer
3966: Note:
3967: The vectors `b` and `x` cannot be the same. I.e., one cannot
3968: call `MatSolveAdd`(A,x,y,x).
3970: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatSolve()`, `MatGetFactor()`, `MatSolveTranspose()`, `MatSolveTransposeAdd()`
3971: @*/
3972: PetscErrorCode MatSolveAdd(Mat mat, Vec b, Vec y, Vec x)
3973: {
3974: PetscScalar one = 1.0;
3975: Vec tmp;
3977: PetscFunctionBegin;
3983: PetscCheckSameComm(mat, 1, b, 2);
3984: PetscCheckSameComm(mat, 1, y, 3);
3985: PetscCheckSameComm(mat, 1, x, 4);
3986: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3987: 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);
3988: 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);
3989: 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);
3990: 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);
3991: 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);
3992: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3993: MatCheckPreallocated(mat, 1);
3995: PetscCall(PetscLogEventBegin(MAT_SolveAdd, mat, b, x, y));
3996: if (mat->factorerrortype) {
3997: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
3998: PetscCall(VecSetInf(x));
3999: } else if (mat->ops->solveadd) {
4000: PetscUseTypeMethod(mat, solveadd, b, y, x);
4001: } else {
4002: /* do the solve then the add manually */
4003: if (x != y) {
4004: PetscCall(MatSolve(mat, b, x));
4005: PetscCall(VecAXPY(x, one, y));
4006: } else {
4007: PetscCall(VecDuplicate(x, &tmp));
4008: PetscCall(VecCopy(x, tmp));
4009: PetscCall(MatSolve(mat, b, x));
4010: PetscCall(VecAXPY(x, one, tmp));
4011: PetscCall(VecDestroy(&tmp));
4012: }
4013: }
4014: PetscCall(PetscLogEventEnd(MAT_SolveAdd, mat, b, x, y));
4015: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4016: PetscFunctionReturn(PETSC_SUCCESS);
4017: }
4019: /*@
4020: MatSolveTranspose - Solves $A^T x = b$, given a factored matrix.
4022: Neighbor-wise Collective
4024: Input Parameters:
4025: + mat - the factored matrix
4026: - b - the right-hand-side vector
4028: Output Parameter:
4029: . x - the result vector
4031: Level: developer
4033: Notes:
4034: The vectors `b` and `x` cannot be the same. I.e., one cannot
4035: call `MatSolveTranspose`(A,x,x).
4037: Most users should employ the `KSP` interface for linear solvers
4038: instead of working directly with matrix algebra routines such as this.
4039: See, e.g., `KSPCreate()`.
4041: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `KSP`, `MatSolve()`, `MatSolveAdd()`, `MatSolveTransposeAdd()`
4042: @*/
4043: PetscErrorCode MatSolveTranspose(Mat mat, Vec b, Vec x)
4044: {
4045: PetscErrorCode (*f)(Mat, Vec, Vec) = (!mat->ops->solvetranspose && mat->symmetric) ? mat->ops->solve : mat->ops->solvetranspose;
4047: PetscFunctionBegin;
4052: PetscCheckSameComm(mat, 1, b, 2);
4053: PetscCheckSameComm(mat, 1, x, 3);
4054: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
4055: 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);
4056: 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);
4057: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
4058: MatCheckPreallocated(mat, 1);
4059: PetscCall(PetscLogEventBegin(MAT_SolveTranspose, mat, b, x, 0));
4060: if (mat->factorerrortype) {
4061: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
4062: PetscCall(VecSetInf(x));
4063: } else {
4064: PetscCheck(f, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Matrix type %s", ((PetscObject)mat)->type_name);
4065: PetscCall((*f)(mat, b, x));
4066: }
4067: PetscCall(PetscLogEventEnd(MAT_SolveTranspose, mat, b, x, 0));
4068: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4069: PetscFunctionReturn(PETSC_SUCCESS);
4070: }
4072: /*@
4073: MatSolveTransposeAdd - Computes $x = y + A^{-T} b$
4074: factored matrix.
4076: Neighbor-wise Collective
4078: Input Parameters:
4079: + mat - the factored matrix
4080: . b - the right-hand-side vector
4081: - y - the vector to be added to
4083: Output Parameter:
4084: . x - the result vector
4086: Level: developer
4088: Note:
4089: The vectors `b` and `x` cannot be the same. I.e., one cannot
4090: call `MatSolveTransposeAdd`(A,x,y,x).
4092: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatSolve()`, `MatSolveAdd()`, `MatSolveTranspose()`
4093: @*/
4094: PetscErrorCode MatSolveTransposeAdd(Mat mat, Vec b, Vec y, Vec x)
4095: {
4096: PetscScalar one = 1.0;
4097: Vec tmp;
4098: PetscErrorCode (*f)(Mat, Vec, Vec, Vec) = (!mat->ops->solvetransposeadd && mat->symmetric) ? mat->ops->solveadd : mat->ops->solvetransposeadd;
4100: PetscFunctionBegin;
4106: PetscCheckSameComm(mat, 1, b, 2);
4107: PetscCheckSameComm(mat, 1, y, 3);
4108: PetscCheckSameComm(mat, 1, x, 4);
4109: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
4110: 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);
4111: 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);
4112: 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);
4113: 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);
4114: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
4115: MatCheckPreallocated(mat, 1);
4117: PetscCall(PetscLogEventBegin(MAT_SolveTransposeAdd, mat, b, x, y));
4118: if (mat->factorerrortype) {
4119: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
4120: PetscCall(VecSetInf(x));
4121: } else if (f) {
4122: PetscCall((*f)(mat, b, y, x));
4123: } else {
4124: /* do the solve then the add manually */
4125: if (x != y) {
4126: PetscCall(MatSolveTranspose(mat, b, x));
4127: PetscCall(VecAXPY(x, one, y));
4128: } else {
4129: PetscCall(VecDuplicate(x, &tmp));
4130: PetscCall(VecCopy(x, tmp));
4131: PetscCall(MatSolveTranspose(mat, b, x));
4132: PetscCall(VecAXPY(x, one, tmp));
4133: PetscCall(VecDestroy(&tmp));
4134: }
4135: }
4136: PetscCall(PetscLogEventEnd(MAT_SolveTransposeAdd, mat, b, x, y));
4137: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4138: PetscFunctionReturn(PETSC_SUCCESS);
4139: }
4141: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
4142: /*@
4143: MatSOR - Computes relaxation (SOR, Gauss-Seidel) sweeps.
4145: Neighbor-wise Collective
4147: Input Parameters:
4148: + mat - the matrix
4149: . b - the right-hand side
4150: . omega - the relaxation factor
4151: . flag - flag indicating the type of SOR (see below)
4152: . shift - diagonal shift
4153: . its - the number of iterations
4154: - lits - the number of local iterations
4156: Output Parameter:
4157: . x - the solution (can contain an initial guess, use option `SOR_ZERO_INITIAL_GUESS` to indicate no guess)
4159: SOR Flags:
4160: + `SOR_FORWARD_SWEEP` - forward SOR
4161: . `SOR_BACKWARD_SWEEP` - backward SOR
4162: . `SOR_SYMMETRIC_SWEEP` - SSOR (symmetric SOR)
4163: . `SOR_LOCAL_FORWARD_SWEEP` - local forward SOR
4164: . `SOR_LOCAL_BACKWARD_SWEEP` - local forward SOR
4165: . `SOR_LOCAL_SYMMETRIC_SWEEP` - local SSOR
4166: . `SOR_EISENSTAT` - SOR with Eisenstat trick
4167: . `SOR_APPLY_UPPER`, `SOR_APPLY_LOWER` - applies
4168: upper/lower triangular part of matrix to
4169: vector (with omega)
4170: - `SOR_ZERO_INITIAL_GUESS` - zero initial guess
4172: Level: developer
4174: Notes:
4175: `SOR_LOCAL_FORWARD_SWEEP`, `SOR_LOCAL_BACKWARD_SWEEP`, and
4176: `SOR_LOCAL_SYMMETRIC_SWEEP` perform separate independent smoothings
4177: on each processor.
4179: Application programmers will not generally use `MatSOR()` directly,
4180: but instead will employ the `KSP`/`PC` interface.
4182: For `MATBAIJ`, `MATSBAIJ`, and `MATAIJ` matrices with Inodes this does a block SOR smoothing, otherwise it does a pointwise smoothing
4184: Most users should employ the `KSP` interface for linear solvers
4185: instead of working directly with matrix algebra routines such as this.
4186: See, e.g., `KSPCreate()`.
4188: Vectors `x` and `b` CANNOT be the same
4190: The flags are implemented as bitwise inclusive or operations.
4191: For example, use (`SOR_ZERO_INITIAL_GUESS` | `SOR_SYMMETRIC_SWEEP`)
4192: to specify a zero initial guess for SSOR.
4194: Developer Note:
4195: We should add block SOR support for `MATAIJ` matrices with block size set to great than one and no inodes
4197: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `KSP`, `PC`, `MatGetFactor()`
4198: @*/
4199: PetscErrorCode MatSOR(Mat mat, Vec b, PetscReal omega, MatSORType flag, PetscReal shift, PetscInt its, PetscInt lits, Vec x)
4200: {
4201: PetscFunctionBegin;
4206: PetscCheckSameComm(mat, 1, b, 2);
4207: PetscCheckSameComm(mat, 1, x, 8);
4208: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4209: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4210: 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);
4211: 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);
4212: 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);
4213: PetscCheck(its > 0, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Relaxation requires global its %" PetscInt_FMT " positive", its);
4214: PetscCheck(lits > 0, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Relaxation requires local its %" PetscInt_FMT " positive", lits);
4215: PetscCheck(b != x, PETSC_COMM_SELF, PETSC_ERR_ARG_IDN, "b and x vector cannot be the same");
4217: MatCheckPreallocated(mat, 1);
4218: PetscCall(PetscLogEventBegin(MAT_SOR, mat, b, x, 0));
4219: PetscUseTypeMethod(mat, sor, b, omega, flag, shift, its, lits, x);
4220: PetscCall(PetscLogEventEnd(MAT_SOR, mat, b, x, 0));
4221: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4222: PetscFunctionReturn(PETSC_SUCCESS);
4223: }
4225: /*
4226: Default matrix copy routine.
4227: */
4228: PetscErrorCode MatCopy_Basic(Mat A, Mat B, MatStructure str)
4229: {
4230: PetscInt i, rstart = 0, rend = 0, nz;
4231: const PetscInt *cwork;
4232: const PetscScalar *vwork;
4234: PetscFunctionBegin;
4235: if (B->assembled) PetscCall(MatZeroEntries(B));
4236: if (str == SAME_NONZERO_PATTERN) {
4237: PetscCall(MatGetOwnershipRange(A, &rstart, &rend));
4238: for (i = rstart; i < rend; i++) {
4239: PetscCall(MatGetRow(A, i, &nz, &cwork, &vwork));
4240: PetscCall(MatSetValues(B, 1, &i, nz, cwork, vwork, INSERT_VALUES));
4241: PetscCall(MatRestoreRow(A, i, &nz, &cwork, &vwork));
4242: }
4243: } else {
4244: PetscCall(MatAYPX(B, 0.0, A, str));
4245: }
4246: PetscCall(MatAssemblyBegin(B, MAT_FINAL_ASSEMBLY));
4247: PetscCall(MatAssemblyEnd(B, MAT_FINAL_ASSEMBLY));
4248: PetscFunctionReturn(PETSC_SUCCESS);
4249: }
4251: /*@
4252: MatCopy - Copies a matrix to another matrix.
4254: Collective
4256: Input Parameters:
4257: + A - the matrix
4258: - str - `SAME_NONZERO_PATTERN` or `DIFFERENT_NONZERO_PATTERN`
4260: Output Parameter:
4261: . B - where the copy is put
4263: Level: intermediate
4265: Notes:
4266: If you use `SAME_NONZERO_PATTERN` then the two matrices must have the same nonzero pattern or the routine will crash.
4268: `MatCopy()` copies the matrix entries of a matrix to another existing
4269: matrix (after first zeroing the second matrix). A related routine is
4270: `MatConvert()`, which first creates a new matrix and then copies the data.
4272: .seealso: [](ch_matrices), `Mat`, `MatConvert()`, `MatDuplicate()`
4273: @*/
4274: PetscErrorCode MatCopy(Mat A, Mat B, MatStructure str)
4275: {
4276: PetscInt i;
4278: PetscFunctionBegin;
4283: PetscCheckSameComm(A, 1, B, 2);
4284: MatCheckPreallocated(B, 2);
4285: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4286: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4287: 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,
4288: A->cmap->N, B->cmap->N);
4289: MatCheckPreallocated(A, 1);
4290: if (A == B) PetscFunctionReturn(PETSC_SUCCESS);
4292: PetscCall(PetscLogEventBegin(MAT_Copy, A, B, 0, 0));
4293: if (A->ops->copy) PetscUseTypeMethod(A, copy, B, str);
4294: else PetscCall(MatCopy_Basic(A, B, str));
4296: B->stencil.dim = A->stencil.dim;
4297: B->stencil.noc = A->stencil.noc;
4298: for (i = 0; i <= A->stencil.dim + (A->stencil.noc ? 0 : -1); i++) {
4299: B->stencil.dims[i] = A->stencil.dims[i];
4300: B->stencil.starts[i] = A->stencil.starts[i];
4301: }
4303: PetscCall(PetscLogEventEnd(MAT_Copy, A, B, 0, 0));
4304: PetscCall(PetscObjectStateIncrease((PetscObject)B));
4305: PetscFunctionReturn(PETSC_SUCCESS);
4306: }
4308: /*@
4309: MatConvert - Converts a matrix to another matrix, either of the same
4310: or different type.
4312: Collective
4314: Input Parameters:
4315: + mat - the matrix
4316: . newtype - new matrix type. Use `MATSAME` to create a new matrix of the
4317: same type as the original matrix.
4318: - reuse - denotes if the destination matrix is to be created or reused.
4319: 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
4320: `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).
4322: Output Parameter:
4323: . M - pointer to place new matrix
4325: Level: intermediate
4327: Notes:
4328: `MatConvert()` first creates a new matrix and then copies the data from
4329: the first matrix. A related routine is `MatCopy()`, which copies the matrix
4330: entries of one matrix to another already existing matrix context.
4332: Cannot be used to convert a sequential matrix to parallel or parallel to sequential,
4333: the MPI communicator of the generated matrix is always the same as the communicator
4334: of the input matrix.
4336: .seealso: [](ch_matrices), `Mat`, `MatCopy()`, `MatDuplicate()`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`
4337: @*/
4338: PetscErrorCode MatConvert(Mat mat, MatType newtype, MatReuse reuse, Mat *M)
4339: {
4340: PetscBool sametype, issame, flg;
4341: PetscBool3 issymmetric, ishermitian;
4342: char convname[256], mtype[256];
4343: Mat B;
4345: PetscFunctionBegin;
4348: PetscAssertPointer(M, 4);
4349: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4350: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4351: MatCheckPreallocated(mat, 1);
4353: PetscCall(PetscOptionsGetString(((PetscObject)mat)->options, ((PetscObject)mat)->prefix, "-matconvert_type", mtype, sizeof(mtype), &flg));
4354: if (flg) newtype = mtype;
4356: PetscCall(PetscObjectTypeCompare((PetscObject)mat, newtype, &sametype));
4357: PetscCall(PetscStrcmp(newtype, "same", &issame));
4358: PetscCheck(!(reuse == MAT_INPLACE_MATRIX) || !(mat != *M), PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "MAT_INPLACE_MATRIX requires same input and output matrix");
4359: if (reuse == MAT_REUSE_MATRIX) {
4361: PetscCheck(mat != *M, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "MAT_REUSE_MATRIX means reuse matrix in final argument, perhaps you mean MAT_INPLACE_MATRIX");
4362: }
4364: if ((reuse == MAT_INPLACE_MATRIX) && (issame || sametype)) {
4365: PetscCall(PetscInfo(mat, "Early return for inplace %s %d %d\n", ((PetscObject)mat)->type_name, sametype, issame));
4366: PetscFunctionReturn(PETSC_SUCCESS);
4367: }
4369: /* Cache Mat options because some converters use MatHeaderReplace */
4370: issymmetric = mat->symmetric;
4371: ishermitian = mat->hermitian;
4373: if ((sametype || issame) && (reuse == MAT_INITIAL_MATRIX) && mat->ops->duplicate) {
4374: PetscCall(PetscInfo(mat, "Calling duplicate for initial matrix %s %d %d\n", ((PetscObject)mat)->type_name, sametype, issame));
4375: PetscUseTypeMethod(mat, duplicate, MAT_COPY_VALUES, M);
4376: } else {
4377: PetscErrorCode (*conv)(Mat, MatType, MatReuse, Mat *) = NULL;
4378: const char *prefix[3] = {"seq", "mpi", ""};
4379: PetscInt i;
4380: /*
4381: Order of precedence:
4382: 0) See if newtype is a superclass of the current matrix.
4383: 1) See if a specialized converter is known to the current matrix.
4384: 2) See if a specialized converter is known to the desired matrix class.
4385: 3) See if a good general converter is registered for the desired class
4386: (as of 6/27/03 only MATMPIADJ falls into this category).
4387: 4) See if a good general converter is known for the current matrix.
4388: 5) Use a really basic converter.
4389: */
4391: /* 0) See if newtype is a superclass of the current matrix.
4392: i.e mat is mpiaij and newtype is aij */
4393: for (i = 0; i < 2; i++) {
4394: PetscCall(PetscStrncpy(convname, prefix[i], sizeof(convname)));
4395: PetscCall(PetscStrlcat(convname, newtype, sizeof(convname)));
4396: PetscCall(PetscStrcmp(convname, ((PetscObject)mat)->type_name, &flg));
4397: PetscCall(PetscInfo(mat, "Check superclass %s %s -> %d\n", convname, ((PetscObject)mat)->type_name, flg));
4398: if (flg) {
4399: if (reuse == MAT_INPLACE_MATRIX) {
4400: PetscCall(PetscInfo(mat, "Early return\n"));
4401: PetscFunctionReturn(PETSC_SUCCESS);
4402: } else if (reuse == MAT_INITIAL_MATRIX && mat->ops->duplicate) {
4403: PetscCall(PetscInfo(mat, "Calling MatDuplicate\n"));
4404: PetscUseTypeMethod(mat, duplicate, MAT_COPY_VALUES, M);
4405: PetscFunctionReturn(PETSC_SUCCESS);
4406: } else if (reuse == MAT_REUSE_MATRIX && mat->ops->copy) {
4407: PetscCall(PetscInfo(mat, "Calling MatCopy\n"));
4408: PetscCall(MatCopy(mat, *M, SAME_NONZERO_PATTERN));
4409: PetscFunctionReturn(PETSC_SUCCESS);
4410: }
4411: }
4412: }
4413: /* 1) See if a specialized converter is known to the current matrix and the desired class */
4414: for (i = 0; i < 3; i++) {
4415: PetscCall(PetscStrncpy(convname, "MatConvert_", sizeof(convname)));
4416: PetscCall(PetscStrlcat(convname, ((PetscObject)mat)->type_name, sizeof(convname)));
4417: PetscCall(PetscStrlcat(convname, "_", sizeof(convname)));
4418: PetscCall(PetscStrlcat(convname, prefix[i], sizeof(convname)));
4419: PetscCall(PetscStrlcat(convname, issame ? ((PetscObject)mat)->type_name : newtype, sizeof(convname)));
4420: PetscCall(PetscStrlcat(convname, "_C", sizeof(convname)));
4421: PetscCall(PetscObjectQueryFunction((PetscObject)mat, convname, &conv));
4422: PetscCall(PetscInfo(mat, "Check specialized (1) %s (%s) -> %d\n", convname, ((PetscObject)mat)->type_name, !!conv));
4423: if (conv) goto foundconv;
4424: }
4426: /* 2) See if a specialized converter is known to the desired matrix class. */
4427: PetscCall(MatCreate(PetscObjectComm((PetscObject)mat), &B));
4428: PetscCall(MatSetSizes(B, mat->rmap->n, mat->cmap->n, mat->rmap->N, mat->cmap->N));
4429: PetscCall(MatSetType(B, newtype));
4430: for (i = 0; i < 3; i++) {
4431: PetscCall(PetscStrncpy(convname, "MatConvert_", sizeof(convname)));
4432: PetscCall(PetscStrlcat(convname, ((PetscObject)mat)->type_name, sizeof(convname)));
4433: PetscCall(PetscStrlcat(convname, "_", sizeof(convname)));
4434: PetscCall(PetscStrlcat(convname, prefix[i], sizeof(convname)));
4435: PetscCall(PetscStrlcat(convname, newtype, sizeof(convname)));
4436: PetscCall(PetscStrlcat(convname, "_C", sizeof(convname)));
4437: PetscCall(PetscObjectQueryFunction((PetscObject)B, convname, &conv));
4438: PetscCall(PetscInfo(mat, "Check specialized (2) %s (%s) -> %d\n", convname, ((PetscObject)B)->type_name, !!conv));
4439: if (conv) {
4440: PetscCall(MatDestroy(&B));
4441: goto foundconv;
4442: }
4443: }
4445: /* 3) See if a good general converter is registered for the desired class */
4446: conv = B->ops->convertfrom;
4447: PetscCall(PetscInfo(mat, "Check convertfrom (%s) -> %d\n", ((PetscObject)B)->type_name, !!conv));
4448: PetscCall(MatDestroy(&B));
4449: if (conv) goto foundconv;
4451: /* 4) See if a good general converter is known for the current matrix */
4452: if (mat->ops->convert) conv = mat->ops->convert;
4453: PetscCall(PetscInfo(mat, "Check general convert (%s) -> %d\n", ((PetscObject)mat)->type_name, !!conv));
4454: if (conv) goto foundconv;
4456: /* 5) Use a really basic converter. */
4457: PetscCall(PetscInfo(mat, "Using MatConvert_Basic\n"));
4458: conv = MatConvert_Basic;
4460: foundconv:
4461: PetscCall(PetscLogEventBegin(MAT_Convert, mat, 0, 0, 0));
4462: PetscCall((*conv)(mat, newtype, reuse, M));
4463: if (mat->rmap->mapping && mat->cmap->mapping && !(*M)->rmap->mapping && !(*M)->cmap->mapping) {
4464: /* the block sizes must be same if the mappings are copied over */
4465: (*M)->rmap->bs = mat->rmap->bs;
4466: (*M)->cmap->bs = mat->cmap->bs;
4467: PetscCall(PetscObjectReference((PetscObject)mat->rmap->mapping));
4468: PetscCall(PetscObjectReference((PetscObject)mat->cmap->mapping));
4469: (*M)->rmap->mapping = mat->rmap->mapping;
4470: (*M)->cmap->mapping = mat->cmap->mapping;
4471: }
4472: (*M)->stencil.dim = mat->stencil.dim;
4473: (*M)->stencil.noc = mat->stencil.noc;
4474: for (i = 0; i <= mat->stencil.dim + (mat->stencil.noc ? 0 : -1); i++) {
4475: (*M)->stencil.dims[i] = mat->stencil.dims[i];
4476: (*M)->stencil.starts[i] = mat->stencil.starts[i];
4477: }
4478: PetscCall(PetscLogEventEnd(MAT_Convert, mat, 0, 0, 0));
4479: }
4480: PetscCall(PetscObjectStateIncrease((PetscObject)*M));
4482: /* Copy Mat options */
4483: if (issymmetric == PETSC_BOOL3_TRUE) PetscCall(MatSetOption(*M, MAT_SYMMETRIC, PETSC_TRUE));
4484: else if (issymmetric == PETSC_BOOL3_FALSE) PetscCall(MatSetOption(*M, MAT_SYMMETRIC, PETSC_FALSE));
4485: if (ishermitian == PETSC_BOOL3_TRUE) PetscCall(MatSetOption(*M, MAT_HERMITIAN, PETSC_TRUE));
4486: else if (ishermitian == PETSC_BOOL3_FALSE) PetscCall(MatSetOption(*M, MAT_HERMITIAN, PETSC_FALSE));
4487: PetscFunctionReturn(PETSC_SUCCESS);
4488: }
4490: /*@C
4491: MatFactorGetSolverType - Returns name of the package providing the factorization routines
4493: Not Collective
4495: Input Parameter:
4496: . mat - the matrix, must be a factored matrix
4498: Output Parameter:
4499: . type - the string name of the package (do not free this string)
4501: Level: intermediate
4503: Fortran Note:
4504: Pass in an empty string that is long enough and the package name will be copied into it.
4506: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSolverType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`
4507: @*/
4508: PetscErrorCode MatFactorGetSolverType(Mat mat, MatSolverType *type)
4509: {
4510: PetscErrorCode (*conv)(Mat, MatSolverType *);
4512: PetscFunctionBegin;
4515: PetscAssertPointer(type, 2);
4516: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Only for factored matrix");
4517: PetscCall(PetscObjectQueryFunction((PetscObject)mat, "MatFactorGetSolverType_C", &conv));
4518: if (conv) PetscCall((*conv)(mat, type));
4519: else *type = MATSOLVERPETSC;
4520: PetscFunctionReturn(PETSC_SUCCESS);
4521: }
4523: typedef struct _MatSolverTypeForSpecifcType *MatSolverTypeForSpecifcType;
4524: struct _MatSolverTypeForSpecifcType {
4525: MatType mtype;
4526: /* no entry for MAT_FACTOR_NONE */
4527: PetscErrorCode (*createfactor[MAT_FACTOR_NUM_TYPES - 1])(Mat, MatFactorType, Mat *);
4528: MatSolverTypeForSpecifcType next;
4529: };
4531: typedef struct _MatSolverTypeHolder *MatSolverTypeHolder;
4532: struct _MatSolverTypeHolder {
4533: char *name;
4534: MatSolverTypeForSpecifcType handlers;
4535: MatSolverTypeHolder next;
4536: };
4538: static MatSolverTypeHolder MatSolverTypeHolders = NULL;
4540: /*@C
4541: MatSolverTypeRegister - Registers a `MatSolverType` that works for a particular matrix type
4543: Logically Collective, No Fortran Support
4545: Input Parameters:
4546: + package - name of the package, for example petsc or superlu
4547: . mtype - the matrix type that works with this package
4548: . ftype - the type of factorization supported by the package
4549: - createfactor - routine that will create the factored matrix ready to be used
4551: Level: developer
4553: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorGetSolverType()`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`,
4554: `MatGetFactor()`
4555: @*/
4556: PetscErrorCode MatSolverTypeRegister(MatSolverType package, MatType mtype, MatFactorType ftype, PetscErrorCode (*createfactor)(Mat, MatFactorType, Mat *))
4557: {
4558: MatSolverTypeHolder next = MatSolverTypeHolders, prev = NULL;
4559: PetscBool flg;
4560: MatSolverTypeForSpecifcType inext, iprev = NULL;
4562: PetscFunctionBegin;
4563: PetscCall(MatInitializePackage());
4564: if (!next) {
4565: PetscCall(PetscNew(&MatSolverTypeHolders));
4566: PetscCall(PetscStrallocpy(package, &MatSolverTypeHolders->name));
4567: PetscCall(PetscNew(&MatSolverTypeHolders->handlers));
4568: PetscCall(PetscStrallocpy(mtype, (char **)&MatSolverTypeHolders->handlers->mtype));
4569: MatSolverTypeHolders->handlers->createfactor[(int)ftype - 1] = createfactor;
4570: PetscFunctionReturn(PETSC_SUCCESS);
4571: }
4572: while (next) {
4573: PetscCall(PetscStrcasecmp(package, next->name, &flg));
4574: if (flg) {
4575: PetscCheck(next->handlers, PETSC_COMM_SELF, PETSC_ERR_PLIB, "MatSolverTypeHolder is missing handlers");
4576: inext = next->handlers;
4577: while (inext) {
4578: PetscCall(PetscStrcasecmp(mtype, inext->mtype, &flg));
4579: if (flg) {
4580: inext->createfactor[(int)ftype - 1] = createfactor;
4581: PetscFunctionReturn(PETSC_SUCCESS);
4582: }
4583: iprev = inext;
4584: inext = inext->next;
4585: }
4586: PetscCall(PetscNew(&iprev->next));
4587: PetscCall(PetscStrallocpy(mtype, (char **)&iprev->next->mtype));
4588: iprev->next->createfactor[(int)ftype - 1] = createfactor;
4589: PetscFunctionReturn(PETSC_SUCCESS);
4590: }
4591: prev = next;
4592: next = next->next;
4593: }
4594: PetscCall(PetscNew(&prev->next));
4595: PetscCall(PetscStrallocpy(package, &prev->next->name));
4596: PetscCall(PetscNew(&prev->next->handlers));
4597: PetscCall(PetscStrallocpy(mtype, (char **)&prev->next->handlers->mtype));
4598: prev->next->handlers->createfactor[(int)ftype - 1] = createfactor;
4599: PetscFunctionReturn(PETSC_SUCCESS);
4600: }
4602: /*@C
4603: MatSolverTypeGet - Gets the function that creates the factor matrix if it exist
4605: Input Parameters:
4606: + 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
4607: . ftype - the type of factorization supported by the type
4608: - mtype - the matrix type that works with this type
4610: Output Parameters:
4611: + foundtype - `PETSC_TRUE` if the type was registered
4612: . foundmtype - `PETSC_TRUE` if the type supports the requested mtype
4613: - createfactor - routine that will create the factored matrix ready to be used or `NULL` if not found
4615: Calling sequence of `createfactor`:
4616: + A - the matrix providing the factor matrix
4617: . mtype - the `MatType` of the factor requested
4618: - B - the new factor matrix that responds to MatXXFactorSymbolic,Numeric() functions, such as `MatLUFactorSymbolic()`
4620: Level: developer
4622: Note:
4623: When `type` is `NULL` the available functions are searched for based on the order of the calls to `MatSolverTypeRegister()` in `MatInitializePackage()`.
4624: Since different PETSc configurations may have different external solvers, seemingly identical runs with different PETSc configurations may use a different solver.
4625: For example if one configuration had --download-mumps while a different one had --download-superlu_dist.
4627: .seealso: [](ch_matrices), `Mat`, `MatFactorType`, `MatType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatSolverTypeRegister()`, `MatGetFactor()`,
4628: `MatInitializePackage()`
4629: @*/
4630: PetscErrorCode MatSolverTypeGet(MatSolverType type, MatType mtype, MatFactorType ftype, PetscBool *foundtype, PetscBool *foundmtype, PetscErrorCode (**createfactor)(Mat A, MatFactorType mtype, Mat *B))
4631: {
4632: MatSolverTypeHolder next = MatSolverTypeHolders;
4633: PetscBool flg;
4634: MatSolverTypeForSpecifcType inext;
4636: PetscFunctionBegin;
4637: if (foundtype) *foundtype = PETSC_FALSE;
4638: if (foundmtype) *foundmtype = PETSC_FALSE;
4639: if (createfactor) *createfactor = NULL;
4641: if (type) {
4642: while (next) {
4643: PetscCall(PetscStrcasecmp(type, next->name, &flg));
4644: if (flg) {
4645: if (foundtype) *foundtype = PETSC_TRUE;
4646: inext = next->handlers;
4647: while (inext) {
4648: PetscCall(PetscStrbeginswith(mtype, inext->mtype, &flg));
4649: if (flg) {
4650: if (foundmtype) *foundmtype = PETSC_TRUE;
4651: if (createfactor) *createfactor = inext->createfactor[(int)ftype - 1];
4652: PetscFunctionReturn(PETSC_SUCCESS);
4653: }
4654: inext = inext->next;
4655: }
4656: }
4657: next = next->next;
4658: }
4659: } else {
4660: while (next) {
4661: inext = next->handlers;
4662: while (inext) {
4663: PetscCall(PetscStrcmp(mtype, inext->mtype, &flg));
4664: if (flg && inext->createfactor[(int)ftype - 1]) {
4665: if (foundtype) *foundtype = PETSC_TRUE;
4666: if (foundmtype) *foundmtype = PETSC_TRUE;
4667: if (createfactor) *createfactor = inext->createfactor[(int)ftype - 1];
4668: PetscFunctionReturn(PETSC_SUCCESS);
4669: }
4670: inext = inext->next;
4671: }
4672: next = next->next;
4673: }
4674: /* try with base classes inext->mtype */
4675: next = MatSolverTypeHolders;
4676: while (next) {
4677: inext = next->handlers;
4678: while (inext) {
4679: PetscCall(PetscStrbeginswith(mtype, inext->mtype, &flg));
4680: if (flg && inext->createfactor[(int)ftype - 1]) {
4681: if (foundtype) *foundtype = PETSC_TRUE;
4682: if (foundmtype) *foundmtype = PETSC_TRUE;
4683: if (createfactor) *createfactor = inext->createfactor[(int)ftype - 1];
4684: PetscFunctionReturn(PETSC_SUCCESS);
4685: }
4686: inext = inext->next;
4687: }
4688: next = next->next;
4689: }
4690: }
4691: PetscFunctionReturn(PETSC_SUCCESS);
4692: }
4694: PetscErrorCode MatSolverTypeDestroy(void)
4695: {
4696: MatSolverTypeHolder next = MatSolverTypeHolders, prev;
4697: MatSolverTypeForSpecifcType inext, iprev;
4699: PetscFunctionBegin;
4700: while (next) {
4701: PetscCall(PetscFree(next->name));
4702: inext = next->handlers;
4703: while (inext) {
4704: PetscCall(PetscFree(inext->mtype));
4705: iprev = inext;
4706: inext = inext->next;
4707: PetscCall(PetscFree(iprev));
4708: }
4709: prev = next;
4710: next = next->next;
4711: PetscCall(PetscFree(prev));
4712: }
4713: MatSolverTypeHolders = NULL;
4714: PetscFunctionReturn(PETSC_SUCCESS);
4715: }
4717: /*@
4718: MatFactorGetCanUseOrdering - Indicates if the factorization can use the ordering provided in `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`
4720: Logically Collective
4722: Input Parameter:
4723: . mat - the matrix
4725: Output Parameter:
4726: . flg - `PETSC_TRUE` if uses the ordering
4728: Level: developer
4730: Note:
4731: Most internal PETSc factorizations use the ordering passed to the factorization routine but external
4732: packages do not, thus we want to skip generating the ordering when it is not needed or used.
4734: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`
4735: @*/
4736: PetscErrorCode MatFactorGetCanUseOrdering(Mat mat, PetscBool *flg)
4737: {
4738: PetscFunctionBegin;
4739: *flg = mat->canuseordering;
4740: PetscFunctionReturn(PETSC_SUCCESS);
4741: }
4743: /*@
4744: MatFactorGetPreferredOrdering - The preferred ordering for a particular matrix factor object
4746: Logically Collective
4748: Input Parameters:
4749: + mat - the matrix obtained with `MatGetFactor()`
4750: - ftype - the factorization type to be used
4752: Output Parameter:
4753: . otype - the preferred ordering type
4755: Level: developer
4757: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorType`, `MatOrderingType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`
4758: @*/
4759: PetscErrorCode MatFactorGetPreferredOrdering(Mat mat, MatFactorType ftype, MatOrderingType *otype)
4760: {
4761: PetscFunctionBegin;
4762: *otype = mat->preferredordering[ftype];
4763: PetscCheck(*otype, PETSC_COMM_SELF, PETSC_ERR_PLIB, "MatFactor did not have a preferred ordering");
4764: PetscFunctionReturn(PETSC_SUCCESS);
4765: }
4767: /*@
4768: MatGetFactor - Returns a matrix suitable to calls to MatXXFactorSymbolic,Numeric()
4770: Collective
4772: Input Parameters:
4773: + mat - the matrix
4774: . 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
4775: the other criteria is returned
4776: - ftype - factor type, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`
4778: Output Parameter:
4779: . f - the factor matrix used with MatXXFactorSymbolic,Numeric() calls. Can be `NULL` in some cases, see notes below.
4781: Options Database Keys:
4782: + -pc_factor_mat_solver_type <type> - choose the type at run time. When using `KSP` solvers
4783: - -mat_factor_bind_factorization <host, device> - Where to do matrix factorization? Default is device (might consume more device memory.
4784: One can choose host to save device memory). Currently only supported with `MATSEQAIJCUSPARSE` matrices.
4786: Level: intermediate
4788: Notes:
4789: The return matrix can be `NULL` if the requested factorization is not available, since some combinations of matrix types and factorization
4790: types registered with `MatSolverTypeRegister()` cannot be fully tested if not at runtime.
4792: Users usually access the factorization solvers via `KSP`
4794: Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
4795: 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
4797: When `type` is `NULL` the available results are searched for based on the order of the calls to `MatSolverTypeRegister()` in `MatInitializePackage()`.
4798: Since different PETSc configurations may have different external solvers, seemingly identical runs with different PETSc configurations may use a different solver.
4799: For example if one configuration had --download-mumps while a different one had --download-superlu_dist.
4801: Some of the packages have options for controlling the factorization, these are in the form -prefix_mat_packagename_packageoption
4802: where prefix is normally obtained from the calling `KSP`/`PC`. If `MatGetFactor()` is called directly one can set
4803: call `MatSetOptionsPrefixFactor()` on the originating matrix or `MatSetOptionsPrefix()` on the resulting factor matrix.
4805: Developer Note:
4806: This should actually be called `MatCreateFactor()` since it creates a new factor object
4808: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `KSP`, `MatSolverType`, `MatFactorType`, `MatCopy()`, `MatDuplicate()`,
4809: `MatGetFactorAvailable()`, `MatFactorGetCanUseOrdering()`, `MatSolverTypeRegister()`, `MatSolverTypeGet()`
4810: `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`, `MatInitializePackage()`
4811: @*/
4812: PetscErrorCode MatGetFactor(Mat mat, MatSolverType type, MatFactorType ftype, Mat *f)
4813: {
4814: PetscBool foundtype, foundmtype;
4815: PetscErrorCode (*conv)(Mat, MatFactorType, Mat *);
4817: PetscFunctionBegin;
4821: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4822: MatCheckPreallocated(mat, 1);
4824: PetscCall(MatSolverTypeGet(type, ((PetscObject)mat)->type_name, ftype, &foundtype, &foundmtype, &conv));
4825: if (!foundtype) {
4826: if (type) {
4827: 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],
4828: ((PetscObject)mat)->type_name, type);
4829: } else {
4830: 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);
4831: }
4832: }
4833: PetscCheck(foundmtype, PetscObjectComm((PetscObject)mat), PETSC_ERR_MISSING_FACTOR, "MatSolverType %s does not support matrix type %s", type, ((PetscObject)mat)->type_name);
4834: 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);
4836: PetscCall((*conv)(mat, ftype, f));
4837: if (mat->factorprefix) PetscCall(MatSetOptionsPrefix(*f, mat->factorprefix));
4838: PetscFunctionReturn(PETSC_SUCCESS);
4839: }
4841: /*@
4842: MatGetFactorAvailable - Returns a flag if matrix supports particular type and factor type
4844: Not Collective
4846: Input Parameters:
4847: + mat - the matrix
4848: . type - name of solver type, for example, superlu, petsc (to use PETSc's default)
4849: - ftype - factor type, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`
4851: Output Parameter:
4852: . flg - PETSC_TRUE if the factorization is available
4854: Level: intermediate
4856: Notes:
4857: Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
4858: such as pastix, superlu, mumps etc.
4860: PETSc must have been ./configure to use the external solver, using the option --download-package
4862: Developer Note:
4863: This should actually be called `MatCreateFactorAvailable()` since `MatGetFactor()` creates a new factor object
4865: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatSolverType`, `MatFactorType`, `MatGetFactor()`, `MatCopy()`, `MatDuplicate()`, `MatSolverTypeRegister()`,
4866: `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`, `MatSolverTypeGet()`
4867: @*/
4868: PetscErrorCode MatGetFactorAvailable(Mat mat, MatSolverType type, MatFactorType ftype, PetscBool *flg)
4869: {
4870: PetscErrorCode (*gconv)(Mat, MatFactorType, Mat *);
4872: PetscFunctionBegin;
4874: PetscAssertPointer(flg, 4);
4876: *flg = PETSC_FALSE;
4877: if (!((PetscObject)mat)->type_name) PetscFunctionReturn(PETSC_SUCCESS);
4879: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4880: MatCheckPreallocated(mat, 1);
4882: PetscCall(MatSolverTypeGet(type, ((PetscObject)mat)->type_name, ftype, NULL, NULL, &gconv));
4883: *flg = gconv ? PETSC_TRUE : PETSC_FALSE;
4884: PetscFunctionReturn(PETSC_SUCCESS);
4885: }
4887: /*@
4888: MatDuplicate - Duplicates a matrix including the non-zero structure.
4890: Collective
4892: Input Parameters:
4893: + mat - the matrix
4894: - op - One of `MAT_DO_NOT_COPY_VALUES`, `MAT_COPY_VALUES`, or `MAT_SHARE_NONZERO_PATTERN`.
4895: See the manual page for `MatDuplicateOption()` for an explanation of these options.
4897: Output Parameter:
4898: . M - pointer to place new matrix
4900: Level: intermediate
4902: Notes:
4903: You cannot change the nonzero pattern for the parent or child matrix later if you use `MAT_SHARE_NONZERO_PATTERN`.
4905: If `op` is not `MAT_COPY_VALUES` the numerical values in the new matrix are zeroed.
4907: 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.
4909: When original mat is a product of matrix operation, e.g., an output of `MatMatMult()` or `MatCreateSubMatrix()`, only the matrix data structure of `mat`
4910: is duplicated and the internal data structures created for the reuse of previous matrix operations are not duplicated.
4911: User should not use `MatDuplicate()` to create new matrix `M` if `M` is intended to be reused as the product of matrix operation.
4913: .seealso: [](ch_matrices), `Mat`, `MatCopy()`, `MatConvert()`, `MatDuplicateOption`
4914: @*/
4915: PetscErrorCode MatDuplicate(Mat mat, MatDuplicateOption op, Mat *M)
4916: {
4917: Mat B;
4918: VecType vtype;
4919: PetscInt i;
4920: PetscObject dm, container_h, container_d;
4921: void (*viewf)(void);
4923: PetscFunctionBegin;
4926: PetscAssertPointer(M, 3);
4927: PetscCheck(op != MAT_COPY_VALUES || mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "MAT_COPY_VALUES not allowed for unassembled matrix");
4928: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4929: MatCheckPreallocated(mat, 1);
4931: *M = NULL;
4932: PetscCall(PetscLogEventBegin(MAT_Convert, mat, 0, 0, 0));
4933: PetscUseTypeMethod(mat, duplicate, op, M);
4934: PetscCall(PetscLogEventEnd(MAT_Convert, mat, 0, 0, 0));
4935: B = *M;
4937: PetscCall(MatGetOperation(mat, MATOP_VIEW, &viewf));
4938: if (viewf) PetscCall(MatSetOperation(B, MATOP_VIEW, viewf));
4939: PetscCall(MatGetVecType(mat, &vtype));
4940: PetscCall(MatSetVecType(B, vtype));
4942: B->stencil.dim = mat->stencil.dim;
4943: B->stencil.noc = mat->stencil.noc;
4944: for (i = 0; i <= mat->stencil.dim + (mat->stencil.noc ? 0 : -1); i++) {
4945: B->stencil.dims[i] = mat->stencil.dims[i];
4946: B->stencil.starts[i] = mat->stencil.starts[i];
4947: }
4949: B->nooffproczerorows = mat->nooffproczerorows;
4950: B->nooffprocentries = mat->nooffprocentries;
4952: PetscCall(PetscObjectQuery((PetscObject)mat, "__PETSc_dm", &dm));
4953: if (dm) PetscCall(PetscObjectCompose((PetscObject)B, "__PETSc_dm", dm));
4954: PetscCall(PetscObjectQuery((PetscObject)mat, "__PETSc_MatCOOStruct_Host", &container_h));
4955: if (container_h) PetscCall(PetscObjectCompose((PetscObject)B, "__PETSc_MatCOOStruct_Host", container_h));
4956: PetscCall(PetscObjectQuery((PetscObject)mat, "__PETSc_MatCOOStruct_Device", &container_d));
4957: if (container_d) PetscCall(PetscObjectCompose((PetscObject)B, "__PETSc_MatCOOStruct_Device", container_d));
4958: PetscCall(PetscObjectStateIncrease((PetscObject)B));
4959: PetscFunctionReturn(PETSC_SUCCESS);
4960: }
4962: /*@
4963: MatGetDiagonal - Gets the diagonal of a matrix as a `Vec`
4965: Logically Collective
4967: Input Parameter:
4968: . mat - the matrix
4970: Output Parameter:
4971: . v - the diagonal of the matrix
4973: Level: intermediate
4975: Note:
4976: If `mat` has local sizes `n` x `m`, this routine fills the first `ndiag = min(n, m)` entries
4977: of `v` with the diagonal values. Thus `v` must have local size of at least `ndiag`. If `v`
4978: is larger than `ndiag`, the values of the remaining entries are unspecified.
4980: Currently only correct in parallel for square matrices.
4982: .seealso: [](ch_matrices), `Mat`, `Vec`, `MatGetRow()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMaxAbs()`
4983: @*/
4984: PetscErrorCode MatGetDiagonal(Mat mat, Vec v)
4985: {
4986: PetscFunctionBegin;
4990: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4991: MatCheckPreallocated(mat, 1);
4992: if (PetscDefined(USE_DEBUG)) {
4993: PetscInt nv, row, col, ndiag;
4995: PetscCall(VecGetLocalSize(v, &nv));
4996: PetscCall(MatGetLocalSize(mat, &row, &col));
4997: ndiag = PetscMin(row, col);
4998: 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);
4999: }
5001: PetscUseTypeMethod(mat, getdiagonal, v);
5002: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5003: PetscFunctionReturn(PETSC_SUCCESS);
5004: }
5006: /*@
5007: MatGetRowMin - Gets the minimum value (of the real part) of each
5008: row of the matrix
5010: Logically Collective
5012: Input Parameter:
5013: . mat - the matrix
5015: Output Parameters:
5016: + v - the vector for storing the maximums
5017: - idx - the indices of the column found for each row (optional, pass `NULL` if not needed)
5019: Level: intermediate
5021: Note:
5022: The result of this call are the same as if one converted the matrix to dense format
5023: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
5025: This code is only implemented for a couple of matrix formats.
5027: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMaxAbs()`, `MatGetRowMinAbs()`,
5028: `MatGetRowMax()`
5029: @*/
5030: PetscErrorCode MatGetRowMin(Mat mat, Vec v, PetscInt idx[])
5031: {
5032: PetscFunctionBegin;
5036: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5038: if (!mat->cmap->N) {
5039: PetscCall(VecSet(v, PETSC_MAX_REAL));
5040: if (idx) {
5041: PetscInt i, m = mat->rmap->n;
5042: for (i = 0; i < m; i++) idx[i] = -1;
5043: }
5044: } else {
5045: MatCheckPreallocated(mat, 1);
5046: }
5047: PetscUseTypeMethod(mat, getrowmin, v, idx);
5048: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5049: PetscFunctionReturn(PETSC_SUCCESS);
5050: }
5052: /*@
5053: MatGetRowMinAbs - Gets the minimum value (in absolute value) 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 minimums
5063: - idx - the indices of the column found for each row (or `NULL` if not needed)
5065: Level: intermediate
5067: Notes:
5068: if a row is completely empty or has only 0.0 values then the `idx` value for that
5069: row is 0 (the first column).
5071: This code is only implemented for a couple of matrix formats.
5073: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMax()`, `MatGetRowMaxAbs()`, `MatGetRowMin()`
5074: @*/
5075: PetscErrorCode MatGetRowMinAbs(Mat mat, Vec v, PetscInt idx[])
5076: {
5077: PetscFunctionBegin;
5081: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5082: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5084: if (!mat->cmap->N) {
5085: PetscCall(VecSet(v, 0.0));
5086: if (idx) {
5087: PetscInt i, m = mat->rmap->n;
5088: for (i = 0; i < m; i++) idx[i] = -1;
5089: }
5090: } else {
5091: MatCheckPreallocated(mat, 1);
5092: if (idx) PetscCall(PetscArrayzero(idx, mat->rmap->n));
5093: PetscUseTypeMethod(mat, getrowminabs, v, idx);
5094: }
5095: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5096: PetscFunctionReturn(PETSC_SUCCESS);
5097: }
5099: /*@
5100: MatGetRowMax - Gets the maximum value (of the real part) of each
5101: row of the matrix
5103: Logically Collective
5105: Input Parameter:
5106: . mat - the matrix
5108: Output Parameters:
5109: + v - the vector for storing the maximums
5110: - idx - the indices of the column found for each row (optional, otherwise pass `NULL`)
5112: Level: intermediate
5114: Notes:
5115: The result of this call are the same as if one converted the matrix to dense format
5116: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
5118: This code is only implemented for a couple of matrix formats.
5120: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMaxAbs()`, `MatGetRowMin()`, `MatGetRowMinAbs()`
5121: @*/
5122: PetscErrorCode MatGetRowMax(Mat mat, Vec v, PetscInt idx[])
5123: {
5124: PetscFunctionBegin;
5128: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5130: if (!mat->cmap->N) {
5131: PetscCall(VecSet(v, PETSC_MIN_REAL));
5132: if (idx) {
5133: PetscInt i, m = mat->rmap->n;
5134: for (i = 0; i < m; i++) idx[i] = -1;
5135: }
5136: } else {
5137: MatCheckPreallocated(mat, 1);
5138: PetscUseTypeMethod(mat, getrowmax, v, idx);
5139: }
5140: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5141: PetscFunctionReturn(PETSC_SUCCESS);
5142: }
5144: /*@
5145: MatGetRowMaxAbs - Gets the maximum value (in absolute value) of each
5146: row of the matrix
5148: Logically Collective
5150: Input Parameter:
5151: . mat - the matrix
5153: Output Parameters:
5154: + v - the vector for storing the maximums
5155: - idx - the indices of the column found for each row (or `NULL` if not needed)
5157: Level: intermediate
5159: Notes:
5160: if a row is completely empty or has only 0.0 values then the `idx` value for that
5161: row is 0 (the first column).
5163: This code is only implemented for a couple of matrix formats.
5165: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowSum()`, `MatGetRowMin()`, `MatGetRowMinAbs()`
5166: @*/
5167: PetscErrorCode MatGetRowMaxAbs(Mat mat, Vec v, PetscInt idx[])
5168: {
5169: PetscFunctionBegin;
5173: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5175: if (!mat->cmap->N) {
5176: PetscCall(VecSet(v, 0.0));
5177: if (idx) {
5178: PetscInt i, m = mat->rmap->n;
5179: for (i = 0; i < m; i++) idx[i] = -1;
5180: }
5181: } else {
5182: MatCheckPreallocated(mat, 1);
5183: if (idx) PetscCall(PetscArrayzero(idx, mat->rmap->n));
5184: PetscUseTypeMethod(mat, getrowmaxabs, v, idx);
5185: }
5186: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5187: PetscFunctionReturn(PETSC_SUCCESS);
5188: }
5190: /*@
5191: MatGetRowSumAbs - Gets the sum value (in absolute value) of each row of the matrix
5193: Logically Collective
5195: Input Parameter:
5196: . mat - the matrix
5198: Output Parameter:
5199: . v - the vector for storing the sum
5201: Level: intermediate
5203: This code is only implemented for a couple of matrix formats.
5205: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMax()`, `MatGetRowMin()`, `MatGetRowMinAbs()`
5206: @*/
5207: PetscErrorCode MatGetRowSumAbs(Mat mat, Vec v)
5208: {
5209: PetscFunctionBegin;
5213: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5215: if (!mat->cmap->N) {
5216: PetscCall(VecSet(v, 0.0));
5217: } else {
5218: MatCheckPreallocated(mat, 1);
5219: PetscUseTypeMethod(mat, getrowsumabs, v);
5220: }
5221: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5222: PetscFunctionReturn(PETSC_SUCCESS);
5223: }
5225: /*@
5226: MatGetRowSum - Gets the sum of each row of the matrix
5228: Logically or Neighborhood Collective
5230: Input Parameter:
5231: . mat - the matrix
5233: Output Parameter:
5234: . v - the vector for storing the sum of rows
5236: Level: intermediate
5238: Note:
5239: This code is slow since it is not currently specialized for different formats
5241: .seealso: [](ch_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMax()`, `MatGetRowMin()`, `MatGetRowMaxAbs()`, `MatGetRowMinAbs()`, `MatGetRowSumAbs()`
5242: @*/
5243: PetscErrorCode MatGetRowSum(Mat mat, Vec v)
5244: {
5245: Vec ones;
5247: PetscFunctionBegin;
5251: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5252: MatCheckPreallocated(mat, 1);
5253: PetscCall(MatCreateVecs(mat, &ones, NULL));
5254: PetscCall(VecSet(ones, 1.));
5255: PetscCall(MatMult(mat, ones, v));
5256: PetscCall(VecDestroy(&ones));
5257: PetscFunctionReturn(PETSC_SUCCESS);
5258: }
5260: /*@
5261: MatTransposeSetPrecursor - Set the matrix from which the second matrix will receive numerical transpose data with a call to `MatTranspose`(A,`MAT_REUSE_MATRIX`,&B)
5262: when B was not obtained with `MatTranspose`(A,`MAT_INITIAL_MATRIX`,&B)
5264: Collective
5266: Input Parameter:
5267: . mat - the matrix to provide the transpose
5269: Output Parameter:
5270: . 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
5272: Level: advanced
5274: Note:
5275: 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
5276: routine allows bypassing that call.
5278: .seealso: [](ch_matrices), `Mat`, `MatTransposeSymbolic()`, `MatTranspose()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`
5279: @*/
5280: PetscErrorCode MatTransposeSetPrecursor(Mat mat, Mat B)
5281: {
5282: PetscContainer rB = NULL;
5283: MatParentState *rb = NULL;
5285: PetscFunctionBegin;
5286: PetscCall(PetscNew(&rb));
5287: rb->id = ((PetscObject)mat)->id;
5288: rb->state = 0;
5289: PetscCall(MatGetNonzeroState(mat, &rb->nonzerostate));
5290: PetscCall(PetscContainerCreate(PetscObjectComm((PetscObject)B), &rB));
5291: PetscCall(PetscContainerSetPointer(rB, rb));
5292: PetscCall(PetscContainerSetUserDestroy(rB, PetscContainerUserDestroyDefault));
5293: PetscCall(PetscObjectCompose((PetscObject)B, "MatTransposeParent", (PetscObject)rB));
5294: PetscCall(PetscObjectDereference((PetscObject)rB));
5295: PetscFunctionReturn(PETSC_SUCCESS);
5296: }
5298: /*@
5299: MatTranspose - Computes an in-place or out-of-place transpose of a matrix.
5301: Collective
5303: Input Parameters:
5304: + mat - the matrix to transpose
5305: - reuse - either `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, or `MAT_INPLACE_MATRIX`
5307: Output Parameter:
5308: . B - the transpose
5310: Level: intermediate
5312: Notes:
5313: If you use `MAT_INPLACE_MATRIX` then you must pass in `&mat` for `B`
5315: `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
5316: transpose, call `MatTransposeSetPrecursor(mat, B)` before calling this routine.
5318: 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.
5320: Consider using `MatCreateTranspose()` instead if you only need a matrix that behaves like the transpose, but don't need the storage to be changed.
5322: If mat is unchanged from the last call this function returns immediately without recomputing the result
5324: If you only need the symbolic transpose, and not the numerical values, use `MatTransposeSymbolic()`
5326: .seealso: [](ch_matrices), `Mat`, `MatTransposeSetPrecursor()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`,
5327: `MatTransposeSymbolic()`, `MatCreateTranspose()`
5328: @*/
5329: PetscErrorCode MatTranspose(Mat mat, MatReuse reuse, Mat *B)
5330: {
5331: PetscContainer rB = NULL;
5332: MatParentState *rb = NULL;
5334: PetscFunctionBegin;
5337: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5338: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5339: PetscCheck(reuse != MAT_INPLACE_MATRIX || mat == *B, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "MAT_INPLACE_MATRIX requires last matrix to match first");
5340: PetscCheck(reuse != MAT_REUSE_MATRIX || mat != *B, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Perhaps you mean MAT_INPLACE_MATRIX");
5341: MatCheckPreallocated(mat, 1);
5342: if (reuse == MAT_REUSE_MATRIX) {
5343: PetscCall(PetscObjectQuery((PetscObject)*B, "MatTransposeParent", (PetscObject *)&rB));
5344: PetscCheck(rB, PetscObjectComm((PetscObject)*B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from call to MatTranspose(). Suggest MatTransposeSetPrecursor().");
5345: PetscCall(PetscContainerGetPointer(rB, (void **)&rb));
5346: PetscCheck(rb->id == ((PetscObject)mat)->id, PetscObjectComm((PetscObject)*B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from input matrix");
5347: if (rb->state == ((PetscObject)mat)->state) PetscFunctionReturn(PETSC_SUCCESS);
5348: }
5350: PetscCall(PetscLogEventBegin(MAT_Transpose, mat, 0, 0, 0));
5351: if (reuse != MAT_INPLACE_MATRIX || mat->symmetric != PETSC_BOOL3_TRUE) {
5352: PetscUseTypeMethod(mat, transpose, reuse, B);
5353: PetscCall(PetscObjectStateIncrease((PetscObject)*B));
5354: }
5355: PetscCall(PetscLogEventEnd(MAT_Transpose, mat, 0, 0, 0));
5357: if (reuse == MAT_INITIAL_MATRIX) PetscCall(MatTransposeSetPrecursor(mat, *B));
5358: if (reuse != MAT_INPLACE_MATRIX) {
5359: PetscCall(PetscObjectQuery((PetscObject)*B, "MatTransposeParent", (PetscObject *)&rB));
5360: PetscCall(PetscContainerGetPointer(rB, (void **)&rb));
5361: rb->state = ((PetscObject)mat)->state;
5362: rb->nonzerostate = mat->nonzerostate;
5363: }
5364: PetscFunctionReturn(PETSC_SUCCESS);
5365: }
5367: /*@
5368: MatTransposeSymbolic - Computes the symbolic part of the transpose of a matrix.
5370: Collective
5372: Input Parameter:
5373: . A - the matrix to transpose
5375: Output Parameter:
5376: . 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
5377: numerical portion.
5379: Level: intermediate
5381: Note:
5382: This is not supported for many matrix types, use `MatTranspose()` in those cases
5384: .seealso: [](ch_matrices), `Mat`, `MatTransposeSetPrecursor()`, `MatTranspose()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`
5385: @*/
5386: PetscErrorCode MatTransposeSymbolic(Mat A, Mat *B)
5387: {
5388: PetscFunctionBegin;
5391: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5392: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5393: PetscCall(PetscLogEventBegin(MAT_Transpose, A, 0, 0, 0));
5394: PetscUseTypeMethod(A, transposesymbolic, B);
5395: PetscCall(PetscLogEventEnd(MAT_Transpose, A, 0, 0, 0));
5397: PetscCall(MatTransposeSetPrecursor(A, *B));
5398: PetscFunctionReturn(PETSC_SUCCESS);
5399: }
5401: PetscErrorCode MatTransposeCheckNonzeroState_Private(Mat A, Mat B)
5402: {
5403: PetscContainer rB;
5404: MatParentState *rb;
5406: PetscFunctionBegin;
5409: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5410: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5411: PetscCall(PetscObjectQuery((PetscObject)B, "MatTransposeParent", (PetscObject *)&rB));
5412: PetscCheck(rB, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from call to MatTranspose()");
5413: PetscCall(PetscContainerGetPointer(rB, (void **)&rb));
5414: PetscCheck(rb->id == ((PetscObject)A)->id, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from input matrix");
5415: PetscCheck(rb->nonzerostate == A->nonzerostate, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONGSTATE, "Reuse matrix has changed nonzero structure");
5416: PetscFunctionReturn(PETSC_SUCCESS);
5417: }
5419: /*@
5420: MatIsTranspose - Test whether a matrix is another one's transpose,
5421: or its own, in which case it tests symmetry.
5423: Collective
5425: Input Parameters:
5426: + A - the matrix to test
5427: . B - the matrix to test against, this can equal the first parameter
5428: - tol - tolerance, differences between entries smaller than this are counted as zero
5430: Output Parameter:
5431: . flg - the result
5433: Level: intermediate
5435: Notes:
5436: The sequential algorithm has a running time of the order of the number of nonzeros; the parallel
5437: test involves parallel copies of the block off-diagonal parts of the matrix.
5439: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatIsSymmetric()`, `MatIsHermitian()`
5440: @*/
5441: PetscErrorCode MatIsTranspose(Mat A, Mat B, PetscReal tol, PetscBool *flg)
5442: {
5443: PetscErrorCode (*f)(Mat, Mat, PetscReal, PetscBool *), (*g)(Mat, Mat, PetscReal, PetscBool *);
5445: PetscFunctionBegin;
5448: PetscAssertPointer(flg, 4);
5449: PetscCall(PetscObjectQueryFunction((PetscObject)A, "MatIsTranspose_C", &f));
5450: PetscCall(PetscObjectQueryFunction((PetscObject)B, "MatIsTranspose_C", &g));
5451: *flg = PETSC_FALSE;
5452: if (f && g) {
5453: PetscCheck(f == g, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_NOTSAMETYPE, "Matrices do not have the same comparator for symmetry test");
5454: PetscCall((*f)(A, B, tol, flg));
5455: } else {
5456: MatType mattype;
5458: PetscCall(MatGetType(f ? B : A, &mattype));
5459: SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "Matrix of type %s does not support checking for transpose", mattype);
5460: }
5461: PetscFunctionReturn(PETSC_SUCCESS);
5462: }
5464: /*@
5465: MatHermitianTranspose - Computes an in-place or out-of-place Hermitian transpose of a matrix in complex conjugate.
5467: Collective
5469: Input Parameters:
5470: + mat - the matrix to transpose and complex conjugate
5471: - reuse - either `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, or `MAT_INPLACE_MATRIX`
5473: Output Parameter:
5474: . B - the Hermitian transpose
5476: Level: intermediate
5478: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`
5479: @*/
5480: PetscErrorCode MatHermitianTranspose(Mat mat, MatReuse reuse, Mat *B)
5481: {
5482: PetscFunctionBegin;
5483: PetscCall(MatTranspose(mat, reuse, B));
5484: #if defined(PETSC_USE_COMPLEX)
5485: PetscCall(MatConjugate(*B));
5486: #endif
5487: PetscFunctionReturn(PETSC_SUCCESS);
5488: }
5490: /*@
5491: MatIsHermitianTranspose - Test whether a matrix is another one's Hermitian transpose,
5493: Collective
5495: Input Parameters:
5496: + A - the matrix to test
5497: . B - the matrix to test against, this can equal the first parameter
5498: - tol - tolerance, differences between entries smaller than this are counted as zero
5500: Output Parameter:
5501: . flg - the result
5503: Level: intermediate
5505: Notes:
5506: Only available for `MATAIJ` matrices.
5508: The sequential algorithm
5509: has a running time of the order of the number of nonzeros; the parallel
5510: test involves parallel copies of the block off-diagonal parts of the matrix.
5512: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatIsSymmetric()`, `MatIsHermitian()`, `MatIsTranspose()`
5513: @*/
5514: PetscErrorCode MatIsHermitianTranspose(Mat A, Mat B, PetscReal tol, PetscBool *flg)
5515: {
5516: PetscErrorCode (*f)(Mat, Mat, PetscReal, PetscBool *), (*g)(Mat, Mat, PetscReal, PetscBool *);
5518: PetscFunctionBegin;
5521: PetscAssertPointer(flg, 4);
5522: PetscCall(PetscObjectQueryFunction((PetscObject)A, "MatIsHermitianTranspose_C", &f));
5523: PetscCall(PetscObjectQueryFunction((PetscObject)B, "MatIsHermitianTranspose_C", &g));
5524: if (f && g) {
5525: PetscCheck(f != g, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_NOTSAMETYPE, "Matrices do not have the same comparator for Hermitian test");
5526: PetscCall((*f)(A, B, tol, flg));
5527: }
5528: PetscFunctionReturn(PETSC_SUCCESS);
5529: }
5531: /*@
5532: MatPermute - Creates a new matrix with rows and columns permuted from the
5533: original.
5535: Collective
5537: Input Parameters:
5538: + mat - the matrix to permute
5539: . row - row permutation, each processor supplies only the permutation for its rows
5540: - col - column permutation, each processor supplies only the permutation for its columns
5542: Output Parameter:
5543: . B - the permuted matrix
5545: Level: advanced
5547: Note:
5548: The index sets map from row/col of permuted matrix to row/col of original matrix.
5549: The index sets should be on the same communicator as mat and have the same local sizes.
5551: Developer Note:
5552: If you want to implement `MatPermute()` for a matrix type, and your approach doesn't
5553: exploit the fact that row and col are permutations, consider implementing the
5554: more general `MatCreateSubMatrix()` instead.
5556: .seealso: [](ch_matrices), `Mat`, `MatGetOrdering()`, `ISAllGather()`, `MatCreateSubMatrix()`
5557: @*/
5558: PetscErrorCode MatPermute(Mat mat, IS row, IS col, Mat *B)
5559: {
5560: PetscFunctionBegin;
5565: PetscAssertPointer(B, 4);
5566: PetscCheckSameComm(mat, 1, row, 2);
5567: if (row != col) PetscCheckSameComm(row, 2, col, 3);
5568: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5569: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5570: PetscCheck(mat->ops->permute || mat->ops->createsubmatrix, PETSC_COMM_SELF, PETSC_ERR_SUP, "MatPermute not available for Mat type %s", ((PetscObject)mat)->type_name);
5571: MatCheckPreallocated(mat, 1);
5573: if (mat->ops->permute) {
5574: PetscUseTypeMethod(mat, permute, row, col, B);
5575: PetscCall(PetscObjectStateIncrease((PetscObject)*B));
5576: } else {
5577: PetscCall(MatCreateSubMatrix(mat, row, col, MAT_INITIAL_MATRIX, B));
5578: }
5579: PetscFunctionReturn(PETSC_SUCCESS);
5580: }
5582: /*@
5583: MatEqual - Compares two matrices.
5585: Collective
5587: Input Parameters:
5588: + A - the first matrix
5589: - B - the second matrix
5591: Output Parameter:
5592: . flg - `PETSC_TRUE` if the matrices are equal; `PETSC_FALSE` otherwise.
5594: Level: intermediate
5596: .seealso: [](ch_matrices), `Mat`
5597: @*/
5598: PetscErrorCode MatEqual(Mat A, Mat B, PetscBool *flg)
5599: {
5600: PetscFunctionBegin;
5605: PetscAssertPointer(flg, 3);
5606: PetscCheckSameComm(A, 1, B, 2);
5607: MatCheckPreallocated(A, 1);
5608: MatCheckPreallocated(B, 2);
5609: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5610: PetscCheck(B->assembled, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5611: 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,
5612: B->cmap->N);
5613: if (A->ops->equal && A->ops->equal == B->ops->equal) {
5614: PetscUseTypeMethod(A, equal, B, flg);
5615: } else {
5616: PetscCall(MatMultEqual(A, B, 10, flg));
5617: }
5618: PetscFunctionReturn(PETSC_SUCCESS);
5619: }
5621: /*@
5622: MatDiagonalScale - Scales a matrix on the left and right by diagonal
5623: matrices that are stored as vectors. Either of the two scaling
5624: matrices can be `NULL`.
5626: Collective
5628: Input Parameters:
5629: + mat - the matrix to be scaled
5630: . l - the left scaling vector (or `NULL`)
5631: - r - the right scaling vector (or `NULL`)
5633: Level: intermediate
5635: Note:
5636: `MatDiagonalScale()` computes $A = LAR$, where
5637: L = a diagonal matrix (stored as a vector), R = a diagonal matrix (stored as a vector)
5638: The L scales the rows of the matrix, the R scales the columns of the matrix.
5640: .seealso: [](ch_matrices), `Mat`, `MatScale()`, `MatShift()`, `MatDiagonalSet()`
5641: @*/
5642: PetscErrorCode MatDiagonalScale(Mat mat, Vec l, Vec r)
5643: {
5644: PetscFunctionBegin;
5647: if (l) {
5649: PetscCheckSameComm(mat, 1, l, 2);
5650: }
5651: if (r) {
5653: PetscCheckSameComm(mat, 1, r, 3);
5654: }
5655: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5656: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5657: MatCheckPreallocated(mat, 1);
5658: if (!l && !r) PetscFunctionReturn(PETSC_SUCCESS);
5660: PetscCall(PetscLogEventBegin(MAT_Scale, mat, 0, 0, 0));
5661: PetscUseTypeMethod(mat, diagonalscale, l, r);
5662: PetscCall(PetscLogEventEnd(MAT_Scale, mat, 0, 0, 0));
5663: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
5664: if (l != r) mat->symmetric = PETSC_BOOL3_FALSE;
5665: PetscFunctionReturn(PETSC_SUCCESS);
5666: }
5668: /*@
5669: MatScale - Scales all elements of a matrix by a given number.
5671: Logically Collective
5673: Input Parameters:
5674: + mat - the matrix to be scaled
5675: - a - the scaling value
5677: Level: intermediate
5679: .seealso: [](ch_matrices), `Mat`, `MatDiagonalScale()`
5680: @*/
5681: PetscErrorCode MatScale(Mat mat, PetscScalar a)
5682: {
5683: PetscFunctionBegin;
5686: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5687: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5689: MatCheckPreallocated(mat, 1);
5691: PetscCall(PetscLogEventBegin(MAT_Scale, mat, 0, 0, 0));
5692: if (a != (PetscScalar)1.0) {
5693: PetscUseTypeMethod(mat, scale, a);
5694: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
5695: }
5696: PetscCall(PetscLogEventEnd(MAT_Scale, mat, 0, 0, 0));
5697: PetscFunctionReturn(PETSC_SUCCESS);
5698: }
5700: /*@
5701: MatNorm - Calculates various norms of a matrix.
5703: Collective
5705: Input Parameters:
5706: + mat - the matrix
5707: - type - the type of norm, `NORM_1`, `NORM_FROBENIUS`, `NORM_INFINITY`
5709: Output Parameter:
5710: . nrm - the resulting norm
5712: Level: intermediate
5714: .seealso: [](ch_matrices), `Mat`
5715: @*/
5716: PetscErrorCode MatNorm(Mat mat, NormType type, PetscReal *nrm)
5717: {
5718: PetscFunctionBegin;
5721: PetscAssertPointer(nrm, 3);
5723: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5724: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5725: MatCheckPreallocated(mat, 1);
5727: PetscUseTypeMethod(mat, norm, type, nrm);
5728: PetscFunctionReturn(PETSC_SUCCESS);
5729: }
5731: /*
5732: This variable is used to prevent counting of MatAssemblyBegin() that
5733: are called from within a MatAssemblyEnd().
5734: */
5735: static PetscInt MatAssemblyEnd_InUse = 0;
5736: /*@
5737: MatAssemblyBegin - Begins assembling the matrix. This routine should
5738: be called after completing all calls to `MatSetValues()`.
5740: Collective
5742: Input Parameters:
5743: + mat - the matrix
5744: - type - type of assembly, either `MAT_FLUSH_ASSEMBLY` or `MAT_FINAL_ASSEMBLY`
5746: Level: beginner
5748: Notes:
5749: `MatSetValues()` generally caches the values that belong to other MPI processes. The matrix is ready to
5750: use only after `MatAssemblyBegin()` and `MatAssemblyEnd()` have been called.
5752: Use `MAT_FLUSH_ASSEMBLY` when switching between `ADD_VALUES` and `INSERT_VALUES`
5753: in `MatSetValues()`; use `MAT_FINAL_ASSEMBLY` for the final assembly before
5754: using the matrix.
5756: ALL processes that share a matrix MUST call `MatAssemblyBegin()` and `MatAssemblyEnd()` the SAME NUMBER of times, and each time with the
5757: 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
5758: a global collective operation requiring all processes that share the matrix.
5760: Space for preallocated nonzeros that is not filled by a call to `MatSetValues()` or a related routine are compressed
5761: out by assembly. If you intend to use that extra space on a subsequent assembly, be sure to insert explicit zeros
5762: before `MAT_FINAL_ASSEMBLY` so the space is not compressed out.
5764: .seealso: [](ch_matrices), `Mat`, `MatAssemblyEnd()`, `MatSetValues()`, `MatAssembled()`
5765: @*/
5766: PetscErrorCode MatAssemblyBegin(Mat mat, MatAssemblyType type)
5767: {
5768: PetscFunctionBegin;
5771: MatCheckPreallocated(mat, 1);
5772: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix. Did you forget to call MatSetUnfactored()?");
5773: if (mat->assembled) {
5774: mat->was_assembled = PETSC_TRUE;
5775: mat->assembled = PETSC_FALSE;
5776: }
5778: if (!MatAssemblyEnd_InUse) {
5779: PetscCall(PetscLogEventBegin(MAT_AssemblyBegin, mat, 0, 0, 0));
5780: PetscTryTypeMethod(mat, assemblybegin, type);
5781: PetscCall(PetscLogEventEnd(MAT_AssemblyBegin, mat, 0, 0, 0));
5782: } else PetscTryTypeMethod(mat, assemblybegin, type);
5783: PetscFunctionReturn(PETSC_SUCCESS);
5784: }
5786: /*@
5787: MatAssembled - Indicates if a matrix has been assembled and is ready for
5788: use; for example, in matrix-vector product.
5790: Not Collective
5792: Input Parameter:
5793: . mat - the matrix
5795: Output Parameter:
5796: . assembled - `PETSC_TRUE` or `PETSC_FALSE`
5798: Level: advanced
5800: .seealso: [](ch_matrices), `Mat`, `MatAssemblyEnd()`, `MatSetValues()`, `MatAssemblyBegin()`
5801: @*/
5802: PetscErrorCode MatAssembled(Mat mat, PetscBool *assembled)
5803: {
5804: PetscFunctionBegin;
5806: PetscAssertPointer(assembled, 2);
5807: *assembled = mat->assembled;
5808: PetscFunctionReturn(PETSC_SUCCESS);
5809: }
5811: /*@
5812: MatAssemblyEnd - Completes assembling the matrix. This routine should
5813: be called after `MatAssemblyBegin()`.
5815: Collective
5817: Input Parameters:
5818: + mat - the matrix
5819: - type - type of assembly, either `MAT_FLUSH_ASSEMBLY` or `MAT_FINAL_ASSEMBLY`
5821: Options Database Keys:
5822: + -mat_view ::ascii_info - Prints info on matrix at conclusion of `MatAssemblyEnd()`
5823: . -mat_view ::ascii_info_detail - Prints more detailed info
5824: . -mat_view - Prints matrix in ASCII format
5825: . -mat_view ::ascii_matlab - Prints matrix in MATLAB format
5826: . -mat_view draw - draws nonzero structure of matrix, using `MatView()` and `PetscDrawOpenX()`.
5827: . -display <name> - Sets display name (default is host)
5828: . -draw_pause <sec> - Sets number of seconds to pause after display
5829: . -mat_view socket - Sends matrix to socket, can be accessed from MATLAB (See [Using MATLAB with PETSc](ch_matlab))
5830: . -viewer_socket_machine <machine> - Machine to use for socket
5831: . -viewer_socket_port <port> - Port number to use for socket
5832: - -mat_view binary:filename[:append] - Save matrix to file in binary format
5834: Level: beginner
5836: .seealso: [](ch_matrices), `Mat`, `MatAssemblyBegin()`, `MatSetValues()`, `PetscDrawOpenX()`, `PetscDrawCreate()`, `MatView()`, `MatAssembled()`, `PetscViewerSocketOpen()`
5837: @*/
5838: PetscErrorCode MatAssemblyEnd(Mat mat, MatAssemblyType type)
5839: {
5840: static PetscInt inassm = 0;
5841: PetscBool flg = PETSC_FALSE;
5843: PetscFunctionBegin;
5847: inassm++;
5848: MatAssemblyEnd_InUse++;
5849: if (MatAssemblyEnd_InUse == 1) { /* Do the logging only the first time through */
5850: PetscCall(PetscLogEventBegin(MAT_AssemblyEnd, mat, 0, 0, 0));
5851: PetscTryTypeMethod(mat, assemblyend, type);
5852: PetscCall(PetscLogEventEnd(MAT_AssemblyEnd, mat, 0, 0, 0));
5853: } else PetscTryTypeMethod(mat, assemblyend, type);
5855: /* Flush assembly is not a true assembly */
5856: if (type != MAT_FLUSH_ASSEMBLY) {
5857: if (mat->num_ass) {
5858: if (!mat->symmetry_eternal) {
5859: mat->symmetric = PETSC_BOOL3_UNKNOWN;
5860: mat->hermitian = PETSC_BOOL3_UNKNOWN;
5861: }
5862: if (!mat->structural_symmetry_eternal && mat->ass_nonzerostate != mat->nonzerostate) mat->structurally_symmetric = PETSC_BOOL3_UNKNOWN;
5863: if (!mat->spd_eternal) mat->spd = PETSC_BOOL3_UNKNOWN;
5864: }
5865: mat->num_ass++;
5866: mat->assembled = PETSC_TRUE;
5867: mat->ass_nonzerostate = mat->nonzerostate;
5868: }
5870: mat->insertmode = NOT_SET_VALUES;
5871: MatAssemblyEnd_InUse--;
5872: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
5873: if (inassm == 1 && type != MAT_FLUSH_ASSEMBLY) {
5874: PetscCall(MatViewFromOptions(mat, NULL, "-mat_view"));
5876: if (mat->checksymmetryonassembly) {
5877: PetscCall(MatIsSymmetric(mat, mat->checksymmetrytol, &flg));
5878: if (flg) {
5879: PetscCall(PetscPrintf(PetscObjectComm((PetscObject)mat), "Matrix is symmetric (tolerance %g)\n", (double)mat->checksymmetrytol));
5880: } else {
5881: PetscCall(PetscPrintf(PetscObjectComm((PetscObject)mat), "Matrix is not symmetric (tolerance %g)\n", (double)mat->checksymmetrytol));
5882: }
5883: }
5884: if (mat->nullsp && mat->checknullspaceonassembly) PetscCall(MatNullSpaceTest(mat->nullsp, mat, NULL));
5885: }
5886: inassm--;
5887: PetscFunctionReturn(PETSC_SUCCESS);
5888: }
5890: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
5891: /*@
5892: MatSetOption - Sets a parameter option for a matrix. Some options
5893: may be specific to certain storage formats. Some options
5894: determine how values will be inserted (or added). Sorted,
5895: row-oriented input will generally assemble the fastest. The default
5896: is row-oriented.
5898: Logically Collective for certain operations, such as `MAT_SPD`, not collective for `MAT_ROW_ORIENTED`, see `MatOption`
5900: Input Parameters:
5901: + mat - the matrix
5902: . op - the option, one of those listed below (and possibly others),
5903: - flg - turn the option on (`PETSC_TRUE`) or off (`PETSC_FALSE`)
5905: Options Describing Matrix Structure:
5906: + `MAT_SPD` - symmetric positive definite
5907: . `MAT_SYMMETRIC` - symmetric in terms of both structure and value
5908: . `MAT_HERMITIAN` - transpose is the complex conjugation
5909: . `MAT_STRUCTURALLY_SYMMETRIC` - symmetric nonzero structure
5910: . `MAT_SYMMETRY_ETERNAL` - indicates the symmetry (or Hermitian structure) or its absence will persist through any changes to the matrix
5911: . `MAT_STRUCTURAL_SYMMETRY_ETERNAL` - indicates the structural symmetry or its absence will persist through any changes to the matrix
5912: . `MAT_SPD_ETERNAL` - indicates the value of `MAT_SPD` (true or false) will persist through any changes to the matrix
5914: These are not really options of the matrix, they are knowledge about the structure of the matrix that users may provide so that they
5915: do not need to be computed (usually at a high cost)
5917: Options For Use with `MatSetValues()`:
5918: Insert a logically dense subblock, which can be
5919: . `MAT_ROW_ORIENTED` - row-oriented (default)
5921: These options reflect the data you pass in with `MatSetValues()`; it has
5922: nothing to do with how the data is stored internally in the matrix
5923: data structure.
5925: When (re)assembling a matrix, we can restrict the input for
5926: efficiency/debugging purposes. These options include
5927: . `MAT_NEW_NONZERO_LOCATIONS` - additional insertions will be allowed if they generate a new nonzero (slow)
5928: . `MAT_FORCE_DIAGONAL_ENTRIES` - forces diagonal entries to be allocated
5929: . `MAT_IGNORE_OFF_PROC_ENTRIES` - drops off-processor entries
5930: . `MAT_NEW_NONZERO_LOCATION_ERR` - generates an error for new matrix entry
5931: . `MAT_USE_HASH_TABLE` - uses a hash table to speed up matrix assembly
5932: . `MAT_NO_OFF_PROC_ENTRIES` - you know each process will only set values for its own rows, will generate an error if
5933: any process sets values for another process. This avoids all reductions in the MatAssembly routines and thus improves
5934: performance for very large process counts.
5935: - `MAT_SUBSET_OFF_PROC_ENTRIES` - you know that the first assembly after setting this flag will set a superset
5936: of the off-process entries required for all subsequent assemblies. This avoids a rendezvous step in the MatAssembly
5937: functions, instead sending only neighbor messages.
5939: Level: intermediate
5941: Notes:
5942: Except for `MAT_UNUSED_NONZERO_LOCATION_ERR` and `MAT_ROW_ORIENTED` all processes that share the matrix must pass the same value in flg!
5944: Some options are relevant only for particular matrix types and
5945: are thus ignored by others. Other options are not supported by
5946: certain matrix types and will generate an error message if set.
5948: If using Fortran to compute a matrix, one may need to
5949: use the column-oriented option (or convert to the row-oriented
5950: format).
5952: `MAT_NEW_NONZERO_LOCATIONS` set to `PETSC_FALSE` indicates that any add or insertion
5953: that would generate a new entry in the nonzero structure is instead
5954: ignored. Thus, if memory has not already been allocated for this particular
5955: data, then the insertion is ignored. For dense matrices, in which
5956: the entire array is allocated, no entries are ever ignored.
5957: Set after the first `MatAssemblyEnd()`. If this option is set then the `MatAssemblyBegin()`/`MatAssemblyEnd()` processes has one less global reduction
5959: `MAT_NEW_NONZERO_LOCATION_ERR` set to PETSC_TRUE indicates that any add or insertion
5960: that would generate a new entry in the nonzero structure instead produces
5961: 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
5963: `MAT_NEW_NONZERO_ALLOCATION_ERR` set to `PETSC_TRUE` indicates that any add or insertion
5964: that would generate a new entry that has not been preallocated will
5965: instead produce an error. (Currently supported for `MATAIJ` and `MATBAIJ` formats
5966: only.) This is a useful flag when debugging matrix memory preallocation.
5967: If this option is set then the `MatAssemblyBegin()`/`MatAssemblyEnd()` processes has one less global reduction
5969: `MAT_IGNORE_OFF_PROC_ENTRIES` set to `PETSC_TRUE` indicates entries destined for
5970: other processors should be dropped, rather than stashed.
5971: This is useful if you know that the "owning" processor is also
5972: always generating the correct matrix entries, so that PETSc need
5973: not transfer duplicate entries generated on another processor.
5975: `MAT_USE_HASH_TABLE` indicates that a hash table be used to improve the
5976: searches during matrix assembly. When this flag is set, the hash table
5977: is created during the first matrix assembly. This hash table is
5978: used the next time through, during `MatSetValues()`/`MatSetValuesBlocked()`
5979: to improve the searching of indices. `MAT_NEW_NONZERO_LOCATIONS` flag
5980: should be used with `MAT_USE_HASH_TABLE` flag. This option is currently
5981: supported by `MATMPIBAIJ` format only.
5983: `MAT_KEEP_NONZERO_PATTERN` indicates when `MatZeroRows()` is called the zeroed entries
5984: are kept in the nonzero structure. This flag is not used for `MatZeroRowsColumns()`
5986: `MAT_IGNORE_ZERO_ENTRIES` - for `MATAIJ` and `MATIS` matrices this will stop zero values from creating
5987: a zero location in the matrix
5989: `MAT_USE_INODES` - indicates using inode version of the code - works with `MATAIJ` matrix types
5991: `MAT_NO_OFF_PROC_ZERO_ROWS` - you know each process will only zero its own rows. This avoids all reductions in the
5992: zero row routines and thus improves performance for very large process counts.
5994: `MAT_IGNORE_LOWER_TRIANGULAR` - For `MATSBAIJ` matrices will ignore any insertions you make in the lower triangular
5995: part of the matrix (since they should match the upper triangular part).
5997: `MAT_SORTED_FULL` - each process provides exactly its local rows; all column indices for a given row are passed in a
5998: single call to `MatSetValues()`, preallocation is perfect, row-oriented, `INSERT_VALUES` is used. Common
5999: with finite difference schemes with non-periodic boundary conditions.
6001: Developer Note:
6002: `MAT_SYMMETRY_ETERNAL`, `MAT_STRUCTURAL_SYMMETRY_ETERNAL`, and `MAT_SPD_ETERNAL` are used by `MatAssemblyEnd()` and in other
6003: places where otherwise the value of `MAT_SYMMETRIC`, `MAT_STRUCTURALLY_SYMMETRIC` or `MAT_SPD` would need to be changed back
6004: to `PETSC_BOOL3_UNKNOWN` because the matrix values had changed so the code cannot be certain that the related property had
6005: not changed.
6007: .seealso: [](ch_matrices), `MatOption`, `Mat`, `MatGetOption()`
6008: @*/
6009: PetscErrorCode MatSetOption(Mat mat, MatOption op, PetscBool flg)
6010: {
6011: PetscFunctionBegin;
6013: if (op > 0) {
6016: }
6018: 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);
6020: switch (op) {
6021: case MAT_FORCE_DIAGONAL_ENTRIES:
6022: mat->force_diagonals = flg;
6023: PetscFunctionReturn(PETSC_SUCCESS);
6024: case MAT_NO_OFF_PROC_ENTRIES:
6025: mat->nooffprocentries = flg;
6026: PetscFunctionReturn(PETSC_SUCCESS);
6027: case MAT_SUBSET_OFF_PROC_ENTRIES:
6028: mat->assembly_subset = flg;
6029: if (!mat->assembly_subset) { /* See the same logic in VecAssembly wrt VEC_SUBSET_OFF_PROC_ENTRIES */
6030: #if !defined(PETSC_HAVE_MPIUNI)
6031: PetscCall(MatStashScatterDestroy_BTS(&mat->stash));
6032: #endif
6033: mat->stash.first_assembly_done = PETSC_FALSE;
6034: }
6035: PetscFunctionReturn(PETSC_SUCCESS);
6036: case MAT_NO_OFF_PROC_ZERO_ROWS:
6037: mat->nooffproczerorows = flg;
6038: PetscFunctionReturn(PETSC_SUCCESS);
6039: case MAT_SPD:
6040: if (flg) {
6041: mat->spd = PETSC_BOOL3_TRUE;
6042: mat->symmetric = PETSC_BOOL3_TRUE;
6043: mat->structurally_symmetric = PETSC_BOOL3_TRUE;
6044: } else {
6045: mat->spd = PETSC_BOOL3_FALSE;
6046: }
6047: break;
6048: case MAT_SYMMETRIC:
6049: mat->symmetric = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
6050: if (flg) mat->structurally_symmetric = PETSC_BOOL3_TRUE;
6051: #if !defined(PETSC_USE_COMPLEX)
6052: mat->hermitian = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
6053: #endif
6054: break;
6055: case MAT_HERMITIAN:
6056: mat->hermitian = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
6057: if (flg) mat->structurally_symmetric = PETSC_BOOL3_TRUE;
6058: #if !defined(PETSC_USE_COMPLEX)
6059: mat->symmetric = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
6060: #endif
6061: break;
6062: case MAT_STRUCTURALLY_SYMMETRIC:
6063: mat->structurally_symmetric = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
6064: break;
6065: case MAT_SYMMETRY_ETERNAL:
6066: 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");
6067: mat->symmetry_eternal = flg;
6068: if (flg) mat->structural_symmetry_eternal = PETSC_TRUE;
6069: break;
6070: case MAT_STRUCTURAL_SYMMETRY_ETERNAL:
6071: 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");
6072: mat->structural_symmetry_eternal = flg;
6073: break;
6074: case MAT_SPD_ETERNAL:
6075: 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");
6076: mat->spd_eternal = flg;
6077: if (flg) {
6078: mat->structural_symmetry_eternal = PETSC_TRUE;
6079: mat->symmetry_eternal = PETSC_TRUE;
6080: }
6081: break;
6082: case MAT_STRUCTURE_ONLY:
6083: mat->structure_only = flg;
6084: break;
6085: case MAT_SORTED_FULL:
6086: mat->sortedfull = flg;
6087: break;
6088: default:
6089: break;
6090: }
6091: PetscTryTypeMethod(mat, setoption, op, flg);
6092: PetscFunctionReturn(PETSC_SUCCESS);
6093: }
6095: /*@
6096: MatGetOption - Gets a parameter option that has been set for a matrix.
6098: Logically Collective
6100: Input Parameters:
6101: + mat - the matrix
6102: - op - the option, this only responds to certain options, check the code for which ones
6104: Output Parameter:
6105: . flg - turn the option on (`PETSC_TRUE`) or off (`PETSC_FALSE`)
6107: Level: intermediate
6109: Notes:
6110: Can only be called after `MatSetSizes()` and `MatSetType()` have been set.
6112: Certain option values may be unknown, for those use the routines `MatIsSymmetric()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, or
6113: `MatIsSymmetricKnown()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetricKnown()`
6115: .seealso: [](ch_matrices), `Mat`, `MatOption`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`,
6116: `MatIsSymmetricKnown()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetricKnown()`
6117: @*/
6118: PetscErrorCode MatGetOption(Mat mat, MatOption op, PetscBool *flg)
6119: {
6120: PetscFunctionBegin;
6124: 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);
6125: 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()");
6127: switch (op) {
6128: case MAT_NO_OFF_PROC_ENTRIES:
6129: *flg = mat->nooffprocentries;
6130: break;
6131: case MAT_NO_OFF_PROC_ZERO_ROWS:
6132: *flg = mat->nooffproczerorows;
6133: break;
6134: case MAT_SYMMETRIC:
6135: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsSymmetric() or MatIsSymmetricKnown()");
6136: break;
6137: case MAT_HERMITIAN:
6138: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsHermitian() or MatIsHermitianKnown()");
6139: break;
6140: case MAT_STRUCTURALLY_SYMMETRIC:
6141: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsStructurallySymmetric() or MatIsStructurallySymmetricKnown()");
6142: break;
6143: case MAT_SPD:
6144: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsSPDKnown()");
6145: break;
6146: case MAT_SYMMETRY_ETERNAL:
6147: *flg = mat->symmetry_eternal;
6148: break;
6149: case MAT_STRUCTURAL_SYMMETRY_ETERNAL:
6150: *flg = mat->symmetry_eternal;
6151: break;
6152: default:
6153: break;
6154: }
6155: PetscFunctionReturn(PETSC_SUCCESS);
6156: }
6158: /*@
6159: MatZeroEntries - Zeros all entries of a matrix. For sparse matrices
6160: this routine retains the old nonzero structure.
6162: Logically Collective
6164: Input Parameter:
6165: . mat - the matrix
6167: Level: intermediate
6169: Note:
6170: 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.
6171: See the Performance chapter of the users manual for information on preallocating matrices.
6173: .seealso: [](ch_matrices), `Mat`, `MatZeroRows()`, `MatZeroRowsColumns()`
6174: @*/
6175: PetscErrorCode MatZeroEntries(Mat mat)
6176: {
6177: PetscFunctionBegin;
6180: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6181: 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");
6182: MatCheckPreallocated(mat, 1);
6184: PetscCall(PetscLogEventBegin(MAT_ZeroEntries, mat, 0, 0, 0));
6185: PetscUseTypeMethod(mat, zeroentries);
6186: PetscCall(PetscLogEventEnd(MAT_ZeroEntries, mat, 0, 0, 0));
6187: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6188: PetscFunctionReturn(PETSC_SUCCESS);
6189: }
6191: /*@
6192: MatZeroRowsColumns - Zeros all entries (except possibly the main diagonal)
6193: of a set of rows and columns of a matrix.
6195: Collective
6197: Input Parameters:
6198: + mat - the matrix
6199: . numRows - the number of rows/columns to zero
6200: . rows - the global row indices
6201: . diag - value put in the diagonal of the eliminated rows
6202: . x - optional vector of the solution for zeroed rows (other entries in vector are not used), these must be set before this call
6203: - b - optional vector of the right-hand side, that will be adjusted by provided solution entries
6205: Level: intermediate
6207: Notes:
6208: This routine, along with `MatZeroRows()`, is typically used to eliminate known Dirichlet boundary conditions from a linear system.
6210: For each zeroed row, the value of the corresponding `b` is set to diag times the value of the corresponding `x`.
6211: 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
6213: If the resulting linear system is to be solved with `KSP` then one can (but does not have to) call `KSPSetInitialGuessNonzero()` to allow the
6214: Krylov method to take advantage of the known solution on the zeroed rows.
6216: For the parallel case, all processes that share the matrix (i.e.,
6217: those in the communicator used for matrix creation) MUST call this
6218: routine, regardless of whether any rows being zeroed are owned by
6219: them.
6221: Unlike `MatZeroRows()`, this ignores the `MAT_KEEP_NONZERO_PATTERN` option value set with `MatSetOption()`, it merely zeros those entries in the matrix, but never
6222: removes them from the nonzero pattern. The nonzero pattern of the matrix can still change if a nonzero needs to be inserted on a diagonal entry that was previously
6223: missing.
6225: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
6226: list only rows local to itself).
6228: The option `MAT_NO_OFF_PROC_ZERO_ROWS` does not apply to this routine.
6230: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRows()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6231: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6232: @*/
6233: PetscErrorCode MatZeroRowsColumns(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6234: {
6235: PetscFunctionBegin;
6238: if (numRows) PetscAssertPointer(rows, 3);
6239: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6240: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6241: MatCheckPreallocated(mat, 1);
6243: PetscUseTypeMethod(mat, zerorowscolumns, numRows, rows, diag, x, b);
6244: PetscCall(MatViewFromOptions(mat, NULL, "-mat_view"));
6245: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6246: PetscFunctionReturn(PETSC_SUCCESS);
6247: }
6249: /*@
6250: MatZeroRowsColumnsIS - Zeros all entries (except possibly the main diagonal)
6251: of a set of rows and columns of a matrix.
6253: Collective
6255: Input Parameters:
6256: + mat - the matrix
6257: . is - the rows to zero
6258: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6259: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6260: - b - optional vector of right-hand side, that will be adjusted by provided solution
6262: Level: intermediate
6264: Note:
6265: See `MatZeroRowsColumns()` for details on how this routine operates.
6267: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6268: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRows()`, `MatZeroRowsColumnsStencil()`
6269: @*/
6270: PetscErrorCode MatZeroRowsColumnsIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6271: {
6272: PetscInt numRows;
6273: const PetscInt *rows;
6275: PetscFunctionBegin;
6280: PetscCall(ISGetLocalSize(is, &numRows));
6281: PetscCall(ISGetIndices(is, &rows));
6282: PetscCall(MatZeroRowsColumns(mat, numRows, rows, diag, x, b));
6283: PetscCall(ISRestoreIndices(is, &rows));
6284: PetscFunctionReturn(PETSC_SUCCESS);
6285: }
6287: /*@
6288: MatZeroRows - Zeros all entries (except possibly the main diagonal)
6289: of a set of rows of a matrix.
6291: Collective
6293: Input Parameters:
6294: + mat - the matrix
6295: . numRows - the number of rows to zero
6296: . rows - the global row indices
6297: . diag - value put in the diagonal of the zeroed rows
6298: . x - optional vector of solutions for zeroed rows (other entries in vector are not used), these must be set before this call
6299: - b - optional vector of right-hand side, that will be adjusted by provided solution entries
6301: Level: intermediate
6303: Notes:
6304: This routine, along with `MatZeroRowsColumns()`, is typically used to eliminate known Dirichlet boundary conditions from a linear system.
6306: For each zeroed row, the value of the corresponding `b` is set to `diag` times the value of the corresponding `x`.
6308: If the resulting linear system is to be solved with `KSP` then one can (but does not have to) call `KSPSetInitialGuessNonzero()` to allow the
6309: Krylov method to take advantage of the known solution on the zeroed rows.
6311: 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)
6312: from the matrix.
6314: Unlike `MatZeroRowsColumns()` for the `MATAIJ` and `MATBAIJ` matrix formats this removes the old nonzero structure, from the eliminated rows of the matrix
6315: 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
6316: formats this does not alter the nonzero structure.
6318: If the option `MatSetOption`(mat,`MAT_KEEP_NONZERO_PATTERN`,`PETSC_TRUE`) the nonzero structure
6319: of the matrix is not changed the values are
6320: merely zeroed.
6322: The user can set a value in the diagonal entry (or for the `MATAIJ` format
6323: formats can optionally remove the main diagonal entry from the
6324: nonzero structure as well, by passing 0.0 as the final argument).
6326: For the parallel case, all processes that share the matrix (i.e.,
6327: those in the communicator used for matrix creation) MUST call this
6328: routine, regardless of whether any rows being zeroed are owned by
6329: them.
6331: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
6332: list only rows local to itself).
6334: You can call `MatSetOption`(mat,`MAT_NO_OFF_PROC_ZERO_ROWS`,`PETSC_TRUE`) if each process indicates only rows it
6335: owns that are to be zeroed. This saves a global synchronization in the implementation.
6337: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6338: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`, `PCREDISTRIBUTE`, `MAT_KEEP_NONZERO_PATTERN`
6339: @*/
6340: PetscErrorCode MatZeroRows(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6341: {
6342: PetscFunctionBegin;
6345: if (numRows) PetscAssertPointer(rows, 3);
6346: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6347: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6348: MatCheckPreallocated(mat, 1);
6350: PetscUseTypeMethod(mat, zerorows, numRows, rows, diag, x, b);
6351: PetscCall(MatViewFromOptions(mat, NULL, "-mat_view"));
6352: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6353: PetscFunctionReturn(PETSC_SUCCESS);
6354: }
6356: /*@
6357: MatZeroRowsIS - Zeros all entries (except possibly the main diagonal)
6358: of a set of rows of a matrix.
6360: Collective
6362: Input Parameters:
6363: + mat - the matrix
6364: . is - index set of rows to remove (if `NULL` then no row is removed)
6365: . diag - value put in all diagonals of eliminated rows
6366: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6367: - b - optional vector of right-hand side, that will be adjusted by provided solution
6369: Level: intermediate
6371: Note:
6372: See `MatZeroRows()` for details on how this routine operates.
6374: .seealso: [](ch_matrices), `Mat`, `MatZeroRows()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6375: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6376: @*/
6377: PetscErrorCode MatZeroRowsIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6378: {
6379: PetscInt numRows = 0;
6380: const PetscInt *rows = NULL;
6382: PetscFunctionBegin;
6385: if (is) {
6387: PetscCall(ISGetLocalSize(is, &numRows));
6388: PetscCall(ISGetIndices(is, &rows));
6389: }
6390: PetscCall(MatZeroRows(mat, numRows, rows, diag, x, b));
6391: if (is) PetscCall(ISRestoreIndices(is, &rows));
6392: PetscFunctionReturn(PETSC_SUCCESS);
6393: }
6395: /*@
6396: MatZeroRowsStencil - Zeros all entries (except possibly the main diagonal)
6397: of a set of rows of a matrix. These rows must be local to the process.
6399: Collective
6401: Input Parameters:
6402: + mat - the matrix
6403: . numRows - the number of rows to remove
6404: . rows - the grid coordinates (and component number when dof > 1) for matrix rows
6405: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6406: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6407: - b - optional vector of right-hand side, that will be adjusted by provided solution
6409: Level: intermediate
6411: Notes:
6412: See `MatZeroRows()` for details on how this routine operates.
6414: The grid coordinates are across the entire grid, not just the local portion
6416: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
6417: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
6418: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
6419: `DM_BOUNDARY_PERIODIC` boundary type.
6421: 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
6422: a single value per point) you can skip filling those indices.
6424: Fortran Note:
6425: `idxm` and `idxn` should be declared as
6426: $ MatStencil idxm(4, m)
6427: and the values inserted using
6428: .vb
6429: idxm(MatStencil_i, 1) = i
6430: idxm(MatStencil_j, 1) = j
6431: idxm(MatStencil_k, 1) = k
6432: idxm(MatStencil_c, 1) = c
6433: etc
6434: .ve
6436: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsl()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6437: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6438: @*/
6439: PetscErrorCode MatZeroRowsStencil(Mat mat, PetscInt numRows, const MatStencil rows[], PetscScalar diag, Vec x, Vec b)
6440: {
6441: PetscInt dim = mat->stencil.dim;
6442: PetscInt sdim = dim - (1 - (PetscInt)mat->stencil.noc);
6443: PetscInt *dims = mat->stencil.dims + 1;
6444: PetscInt *starts = mat->stencil.starts;
6445: PetscInt *dxm = (PetscInt *)rows;
6446: PetscInt *jdxm, i, j, tmp, numNewRows = 0;
6448: PetscFunctionBegin;
6451: if (numRows) PetscAssertPointer(rows, 3);
6453: PetscCall(PetscMalloc1(numRows, &jdxm));
6454: for (i = 0; i < numRows; ++i) {
6455: /* Skip unused dimensions (they are ordered k, j, i, c) */
6456: for (j = 0; j < 3 - sdim; ++j) dxm++;
6457: /* Local index in X dir */
6458: tmp = *dxm++ - starts[0];
6459: /* Loop over remaining dimensions */
6460: for (j = 0; j < dim - 1; ++j) {
6461: /* If nonlocal, set index to be negative */
6462: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
6463: /* Update local index */
6464: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
6465: }
6466: /* Skip component slot if necessary */
6467: if (mat->stencil.noc) dxm++;
6468: /* Local row number */
6469: if (tmp >= 0) jdxm[numNewRows++] = tmp;
6470: }
6471: PetscCall(MatZeroRowsLocal(mat, numNewRows, jdxm, diag, x, b));
6472: PetscCall(PetscFree(jdxm));
6473: PetscFunctionReturn(PETSC_SUCCESS);
6474: }
6476: /*@
6477: MatZeroRowsColumnsStencil - Zeros all row and column entries (except possibly the main diagonal)
6478: of a set of rows and columns of a matrix.
6480: Collective
6482: Input Parameters:
6483: + mat - the matrix
6484: . numRows - the number of rows/columns to remove
6485: . rows - the grid coordinates (and component number when dof > 1) for matrix rows
6486: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6487: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6488: - b - optional vector of right-hand side, that will be adjusted by provided solution
6490: Level: intermediate
6492: Notes:
6493: See `MatZeroRowsColumns()` for details on how this routine operates.
6495: The grid coordinates are across the entire grid, not just the local portion
6497: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
6498: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
6499: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
6500: `DM_BOUNDARY_PERIODIC` boundary type.
6502: 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
6503: a single value per point) you can skip filling those indices.
6505: Fortran Note:
6506: `idxm` and `idxn` should be declared as
6507: $ MatStencil idxm(4, m)
6508: and the values inserted using
6509: .vb
6510: idxm(MatStencil_i, 1) = i
6511: idxm(MatStencil_j, 1) = j
6512: idxm(MatStencil_k, 1) = k
6513: idxm(MatStencil_c, 1) = c
6514: etc
6515: .ve
6517: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6518: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRows()`
6519: @*/
6520: PetscErrorCode MatZeroRowsColumnsStencil(Mat mat, PetscInt numRows, const MatStencil rows[], PetscScalar diag, Vec x, Vec b)
6521: {
6522: PetscInt dim = mat->stencil.dim;
6523: PetscInt sdim = dim - (1 - (PetscInt)mat->stencil.noc);
6524: PetscInt *dims = mat->stencil.dims + 1;
6525: PetscInt *starts = mat->stencil.starts;
6526: PetscInt *dxm = (PetscInt *)rows;
6527: PetscInt *jdxm, i, j, tmp, numNewRows = 0;
6529: PetscFunctionBegin;
6532: if (numRows) PetscAssertPointer(rows, 3);
6534: PetscCall(PetscMalloc1(numRows, &jdxm));
6535: for (i = 0; i < numRows; ++i) {
6536: /* Skip unused dimensions (they are ordered k, j, i, c) */
6537: for (j = 0; j < 3 - sdim; ++j) dxm++;
6538: /* Local index in X dir */
6539: tmp = *dxm++ - starts[0];
6540: /* Loop over remaining dimensions */
6541: for (j = 0; j < dim - 1; ++j) {
6542: /* If nonlocal, set index to be negative */
6543: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
6544: /* Update local index */
6545: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
6546: }
6547: /* Skip component slot if necessary */
6548: if (mat->stencil.noc) dxm++;
6549: /* Local row number */
6550: if (tmp >= 0) jdxm[numNewRows++] = tmp;
6551: }
6552: PetscCall(MatZeroRowsColumnsLocal(mat, numNewRows, jdxm, diag, x, b));
6553: PetscCall(PetscFree(jdxm));
6554: PetscFunctionReturn(PETSC_SUCCESS);
6555: }
6557: /*@C
6558: MatZeroRowsLocal - Zeros all entries (except possibly the main diagonal)
6559: of a set of rows of a matrix; using local numbering of rows.
6561: Collective
6563: Input Parameters:
6564: + mat - the matrix
6565: . numRows - the number of rows to remove
6566: . rows - the local row indices
6567: . diag - value put in all diagonals of eliminated rows
6568: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6569: - b - optional vector of right-hand side, that will be adjusted by provided solution
6571: Level: intermediate
6573: Notes:
6574: Before calling `MatZeroRowsLocal()`, the user must first set the
6575: local-to-global mapping by calling MatSetLocalToGlobalMapping(), this is often already set for matrices obtained with `DMCreateMatrix()`.
6577: See `MatZeroRows()` for details on how this routine operates.
6579: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRows()`, `MatSetOption()`,
6580: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6581: @*/
6582: PetscErrorCode MatZeroRowsLocal(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6583: {
6584: PetscFunctionBegin;
6587: if (numRows) PetscAssertPointer(rows, 3);
6588: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6589: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6590: MatCheckPreallocated(mat, 1);
6592: if (mat->ops->zerorowslocal) {
6593: PetscUseTypeMethod(mat, zerorowslocal, numRows, rows, diag, x, b);
6594: } else {
6595: IS is, newis;
6596: const PetscInt *newRows;
6598: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Need to provide local to global mapping to matrix first");
6599: PetscCall(ISCreateGeneral(PETSC_COMM_SELF, numRows, rows, PETSC_COPY_VALUES, &is));
6600: PetscCall(ISLocalToGlobalMappingApplyIS(mat->rmap->mapping, is, &newis));
6601: PetscCall(ISGetIndices(newis, &newRows));
6602: PetscUseTypeMethod(mat, zerorows, numRows, newRows, diag, x, b);
6603: PetscCall(ISRestoreIndices(newis, &newRows));
6604: PetscCall(ISDestroy(&newis));
6605: PetscCall(ISDestroy(&is));
6606: }
6607: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6608: PetscFunctionReturn(PETSC_SUCCESS);
6609: }
6611: /*@
6612: MatZeroRowsLocalIS - Zeros all entries (except possibly the main diagonal)
6613: of a set of rows of a matrix; using local numbering of rows.
6615: Collective
6617: Input Parameters:
6618: + mat - the matrix
6619: . is - index set of rows to remove
6620: . diag - value put in all diagonals of eliminated rows
6621: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6622: - b - optional vector of right-hand side, that will be adjusted by provided solution
6624: Level: intermediate
6626: Notes:
6627: Before calling `MatZeroRowsLocalIS()`, the user must first set the
6628: local-to-global mapping by calling `MatSetLocalToGlobalMapping()`, this is often already set for matrices obtained with `DMCreateMatrix()`.
6630: See `MatZeroRows()` for details on how this routine operates.
6632: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRows()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6633: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6634: @*/
6635: PetscErrorCode MatZeroRowsLocalIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6636: {
6637: PetscInt numRows;
6638: const PetscInt *rows;
6640: PetscFunctionBegin;
6644: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6645: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6646: MatCheckPreallocated(mat, 1);
6648: PetscCall(ISGetLocalSize(is, &numRows));
6649: PetscCall(ISGetIndices(is, &rows));
6650: PetscCall(MatZeroRowsLocal(mat, numRows, rows, diag, x, b));
6651: PetscCall(ISRestoreIndices(is, &rows));
6652: PetscFunctionReturn(PETSC_SUCCESS);
6653: }
6655: /*@
6656: MatZeroRowsColumnsLocal - Zeros all entries (except possibly the main diagonal)
6657: of a set of rows and columns of a matrix; using local numbering of rows.
6659: Collective
6661: Input Parameters:
6662: + mat - the matrix
6663: . numRows - the number of rows to remove
6664: . rows - the global row indices
6665: . diag - value put in all diagonals of eliminated rows
6666: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6667: - b - optional vector of right-hand side, that will be adjusted by provided solution
6669: Level: intermediate
6671: Notes:
6672: Before calling `MatZeroRowsColumnsLocal()`, the user must first set the
6673: local-to-global mapping by calling `MatSetLocalToGlobalMapping()`, this is often already set for matrices obtained with `DMCreateMatrix()`.
6675: See `MatZeroRowsColumns()` for details on how this routine operates.
6677: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6678: `MatZeroRows()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6679: @*/
6680: PetscErrorCode MatZeroRowsColumnsLocal(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6681: {
6682: IS is, newis;
6683: const PetscInt *newRows;
6685: PetscFunctionBegin;
6688: if (numRows) PetscAssertPointer(rows, 3);
6689: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6690: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6691: MatCheckPreallocated(mat, 1);
6693: PetscCheck(mat->cmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Need to provide local to global mapping to matrix first");
6694: PetscCall(ISCreateGeneral(PETSC_COMM_SELF, numRows, rows, PETSC_COPY_VALUES, &is));
6695: PetscCall(ISLocalToGlobalMappingApplyIS(mat->cmap->mapping, is, &newis));
6696: PetscCall(ISGetIndices(newis, &newRows));
6697: PetscUseTypeMethod(mat, zerorowscolumns, numRows, newRows, diag, x, b);
6698: PetscCall(ISRestoreIndices(newis, &newRows));
6699: PetscCall(ISDestroy(&newis));
6700: PetscCall(ISDestroy(&is));
6701: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6702: PetscFunctionReturn(PETSC_SUCCESS);
6703: }
6705: /*@
6706: MatZeroRowsColumnsLocalIS - Zeros all entries (except possibly the main diagonal)
6707: of a set of rows and columns of a matrix; using local numbering of rows.
6709: Collective
6711: Input Parameters:
6712: + mat - the matrix
6713: . is - index set of rows to remove
6714: . diag - value put in all diagonals of eliminated rows
6715: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6716: - b - optional vector of right-hand side, that will be adjusted by provided solution
6718: Level: intermediate
6720: Notes:
6721: Before calling `MatZeroRowsColumnsLocalIS()`, the user must first set the
6722: local-to-global mapping by calling `MatSetLocalToGlobalMapping()`, this is often already set for matrices obtained with `DMCreateMatrix()`.
6724: See `MatZeroRowsColumns()` for details on how this routine operates.
6726: .seealso: [](ch_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6727: `MatZeroRowsColumnsLocal()`, `MatZeroRows()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6728: @*/
6729: PetscErrorCode MatZeroRowsColumnsLocalIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6730: {
6731: PetscInt numRows;
6732: const PetscInt *rows;
6734: PetscFunctionBegin;
6738: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6739: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6740: MatCheckPreallocated(mat, 1);
6742: PetscCall(ISGetLocalSize(is, &numRows));
6743: PetscCall(ISGetIndices(is, &rows));
6744: PetscCall(MatZeroRowsColumnsLocal(mat, numRows, rows, diag, x, b));
6745: PetscCall(ISRestoreIndices(is, &rows));
6746: PetscFunctionReturn(PETSC_SUCCESS);
6747: }
6749: /*@C
6750: MatGetSize - Returns the numbers of rows and columns in a matrix.
6752: Not Collective
6754: Input Parameter:
6755: . mat - the matrix
6757: Output Parameters:
6758: + m - the number of global rows
6759: - n - the number of global columns
6761: Level: beginner
6763: Note:
6764: Both output parameters can be `NULL` on input.
6766: .seealso: [](ch_matrices), `Mat`, `MatSetSizes()`, `MatGetLocalSize()`
6767: @*/
6768: PetscErrorCode MatGetSize(Mat mat, PetscInt *m, PetscInt *n)
6769: {
6770: PetscFunctionBegin;
6772: if (m) *m = mat->rmap->N;
6773: if (n) *n = mat->cmap->N;
6774: PetscFunctionReturn(PETSC_SUCCESS);
6775: }
6777: /*@C
6778: MatGetLocalSize - For most matrix formats, excluding `MATELEMENTAL` and `MATSCALAPACK`, Returns the number of local rows and local columns
6779: of a matrix. For all matrices this is the local size of the left and right vectors as returned by `MatCreateVecs()`.
6781: Not Collective
6783: Input Parameter:
6784: . mat - the matrix
6786: Output Parameters:
6787: + m - the number of local rows, use `NULL` to not obtain this value
6788: - n - the number of local columns, use `NULL` to not obtain this value
6790: Level: beginner
6792: .seealso: [](ch_matrices), `Mat`, `MatSetSizes()`, `MatGetSize()`
6793: @*/
6794: PetscErrorCode MatGetLocalSize(Mat mat, PetscInt *m, PetscInt *n)
6795: {
6796: PetscFunctionBegin;
6798: if (m) PetscAssertPointer(m, 2);
6799: if (n) PetscAssertPointer(n, 3);
6800: if (m) *m = mat->rmap->n;
6801: if (n) *n = mat->cmap->n;
6802: PetscFunctionReturn(PETSC_SUCCESS);
6803: }
6805: /*@
6806: MatGetOwnershipRangeColumn - Returns the range of matrix columns associated with rows of a
6807: vector one multiplies this matrix by that are owned by this processor.
6809: Not Collective, unless matrix has not been allocated, then collective
6811: Input Parameter:
6812: . mat - the matrix
6814: Output Parameters:
6815: + m - the global index of the first local column, use `NULL` to not obtain this value
6816: - n - one more than the global index of the last local column, use `NULL` to not obtain this value
6818: Level: developer
6820: Note:
6821: Returns the columns of the "diagonal block" for most sparse matrix formats. See [Matrix
6822: Layouts](sec_matlayout) for details on matrix layouts.
6824: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRanges()`, `MatGetOwnershipRangesColumn()`, `PetscLayout`
6825: @*/
6826: PetscErrorCode MatGetOwnershipRangeColumn(Mat mat, PetscInt *m, PetscInt *n)
6827: {
6828: PetscFunctionBegin;
6831: if (m) PetscAssertPointer(m, 2);
6832: if (n) PetscAssertPointer(n, 3);
6833: MatCheckPreallocated(mat, 1);
6834: if (m) *m = mat->cmap->rstart;
6835: if (n) *n = mat->cmap->rend;
6836: PetscFunctionReturn(PETSC_SUCCESS);
6837: }
6839: /*@C
6840: MatGetOwnershipRange - For matrices that own values by row, excludes `MATELEMENTAL` and `MATSCALAPACK`, returns the range of matrix rows owned by
6841: this MPI process.
6843: Not Collective
6845: Input Parameter:
6846: . mat - the matrix
6848: Output Parameters:
6849: + m - the global index of the first local row, use `NULL` to not obtain this value
6850: - n - one more than the global index of the last local row, use `NULL` to not obtain this value
6852: Level: beginner
6854: Note:
6855: For all matrices it returns the range of matrix rows associated with rows of a vector that
6856: would contain the result of a matrix vector product with this matrix. See [Matrix
6857: Layouts](sec_matlayout) for details on matrix layouts.
6859: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRanges()`, `MatGetOwnershipRangeColumn()`, `MatGetOwnershipRangesColumn()`, `PetscSplitOwnership()`, `PetscSplitOwnershipBlock()`,
6860: `PetscLayout`
6861: @*/
6862: PetscErrorCode MatGetOwnershipRange(Mat mat, PetscInt *m, PetscInt *n)
6863: {
6864: PetscFunctionBegin;
6867: if (m) PetscAssertPointer(m, 2);
6868: if (n) PetscAssertPointer(n, 3);
6869: MatCheckPreallocated(mat, 1);
6870: if (m) *m = mat->rmap->rstart;
6871: if (n) *n = mat->rmap->rend;
6872: PetscFunctionReturn(PETSC_SUCCESS);
6873: }
6875: /*@C
6876: MatGetOwnershipRanges - For matrices that own values by row, excludes `MATELEMENTAL` and
6877: `MATSCALAPACK`, returns the range of matrix rows owned by each process.
6879: Not Collective, unless matrix has not been allocated
6881: Input Parameter:
6882: . mat - the matrix
6884: Output Parameter:
6885: . ranges - start of each processors portion plus one more than the total length at the end
6887: Level: beginner
6889: Note:
6890: For all matrices it returns the ranges of matrix rows associated with rows of a vector that
6891: would contain the result of a matrix vector product with this matrix. See [Matrix
6892: Layouts](sec_matlayout) for details on matrix layouts.
6894: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`, `MatGetOwnershipRangesColumn()`, `PetscLayout`
6895: @*/
6896: PetscErrorCode MatGetOwnershipRanges(Mat mat, const PetscInt *ranges[])
6897: {
6898: PetscFunctionBegin;
6901: MatCheckPreallocated(mat, 1);
6902: PetscCall(PetscLayoutGetRanges(mat->rmap, ranges));
6903: PetscFunctionReturn(PETSC_SUCCESS);
6904: }
6906: /*@C
6907: MatGetOwnershipRangesColumn - Returns the ranges of matrix columns associated with rows of a
6908: vector one multiplies this vector by that are owned by each processor.
6910: Not Collective, unless matrix has not been allocated
6912: Input Parameter:
6913: . mat - the matrix
6915: Output Parameter:
6916: . ranges - start of each processors portion plus one more than the total length at the end
6918: Level: beginner
6920: Note:
6921: Returns the columns of the "diagonal blocks", for most sparse matrix formats. See [Matrix
6922: Layouts](sec_matlayout) for details on matrix layouts.
6924: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`, `MatGetOwnershipRanges()`
6925: @*/
6926: PetscErrorCode MatGetOwnershipRangesColumn(Mat mat, const PetscInt *ranges[])
6927: {
6928: PetscFunctionBegin;
6931: MatCheckPreallocated(mat, 1);
6932: PetscCall(PetscLayoutGetRanges(mat->cmap, ranges));
6933: PetscFunctionReturn(PETSC_SUCCESS);
6934: }
6936: /*@C
6937: MatGetOwnershipIS - Get row and column ownership of a matrices' values as index sets.
6939: Not Collective
6941: Input Parameter:
6942: . A - matrix
6944: Output Parameters:
6945: + rows - rows in which this process owns elements, , use `NULL` to not obtain this value
6946: - cols - columns in which this process owns elements, use `NULL` to not obtain this value
6948: Level: intermediate
6950: Note:
6951: For most matrices, excluding `MATELEMENTAL` and `MATSCALAPACK`, this corresponds to values
6952: returned by `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`. For `MATELEMENTAL` and
6953: `MATSCALAPACK` the ownership is more complicated. See [Matrix Layouts](sec_matlayout) for
6954: details on matrix layouts.
6956: .seealso: [](ch_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`, `MatSetValues()`, ``MATELEMENTAL``, ``MATSCALAPACK``
6957: @*/
6958: PetscErrorCode MatGetOwnershipIS(Mat A, IS *rows, IS *cols)
6959: {
6960: PetscErrorCode (*f)(Mat, IS *, IS *);
6962: PetscFunctionBegin;
6963: MatCheckPreallocated(A, 1);
6964: PetscCall(PetscObjectQueryFunction((PetscObject)A, "MatGetOwnershipIS_C", &f));
6965: if (f) {
6966: PetscCall((*f)(A, rows, cols));
6967: } else { /* Create a standard row-based partition, each process is responsible for ALL columns in their row block */
6968: if (rows) PetscCall(ISCreateStride(PETSC_COMM_SELF, A->rmap->n, A->rmap->rstart, 1, rows));
6969: if (cols) PetscCall(ISCreateStride(PETSC_COMM_SELF, A->cmap->N, 0, 1, cols));
6970: }
6971: PetscFunctionReturn(PETSC_SUCCESS);
6972: }
6974: /*@C
6975: MatILUFactorSymbolic - Performs symbolic ILU factorization of a matrix obtained with `MatGetFactor()`
6976: Uses levels of fill only, not drop tolerance. Use `MatLUFactorNumeric()`
6977: to complete the factorization.
6979: Collective
6981: Input Parameters:
6982: + fact - the factorized matrix obtained with `MatGetFactor()`
6983: . mat - the matrix
6984: . row - row permutation
6985: . col - column permutation
6986: - info - structure containing
6987: .vb
6988: levels - number of levels of fill.
6989: expected fill - as ratio of original fill.
6990: 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
6991: missing diagonal entries)
6992: .ve
6994: Level: developer
6996: Notes:
6997: See [Matrix Factorization](sec_matfactor) for additional information.
6999: Most users should employ the `KSP` interface for linear solvers
7000: instead of working directly with matrix algebra routines such as this.
7001: See, e.g., `KSPCreate()`.
7003: Uses the definition of level of fill as in Y. Saad, {cite}`saad2003`
7005: Developer Note:
7006: The Fortran interface is not autogenerated as the
7007: interface definition cannot be generated correctly [due to `MatFactorInfo`]
7009: .seealso: [](ch_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`
7010: `MatGetOrdering()`, `MatFactorInfo`
7011: @*/
7012: PetscErrorCode MatILUFactorSymbolic(Mat fact, Mat mat, IS row, IS col, const MatFactorInfo *info)
7013: {
7014: PetscFunctionBegin;
7019: PetscAssertPointer(info, 5);
7020: PetscAssertPointer(fact, 1);
7021: PetscCheck(info->levels >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Levels of fill negative %" PetscInt_FMT, (PetscInt)info->levels);
7022: PetscCheck(info->fill >= 1.0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Expected fill less than 1.0 %g", (double)info->fill);
7023: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7024: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7025: MatCheckPreallocated(mat, 2);
7027: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_ILUFactorSymbolic, mat, row, col, 0));
7028: PetscUseTypeMethod(fact, ilufactorsymbolic, mat, row, col, info);
7029: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_ILUFactorSymbolic, mat, row, col, 0));
7030: PetscFunctionReturn(PETSC_SUCCESS);
7031: }
7033: /*@C
7034: MatICCFactorSymbolic - Performs symbolic incomplete
7035: Cholesky factorization for a symmetric matrix. Use
7036: `MatCholeskyFactorNumeric()` to complete the factorization.
7038: Collective
7040: Input Parameters:
7041: + fact - the factorized matrix obtained with `MatGetFactor()`
7042: . mat - the matrix to be factored
7043: . perm - row and column permutation
7044: - info - structure containing
7045: .vb
7046: levels - number of levels of fill.
7047: expected fill - as ratio of original fill.
7048: .ve
7050: Level: developer
7052: Notes:
7053: Most users should employ the `KSP` interface for linear solvers
7054: instead of working directly with matrix algebra routines such as this.
7055: See, e.g., `KSPCreate()`.
7057: This uses the definition of level of fill as in Y. Saad {cite}`saad2003`
7059: Developer Note:
7060: The Fortran interface is not autogenerated as the
7061: interface definition cannot be generated correctly [due to `MatFactorInfo`]
7063: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatCholeskyFactorNumeric()`, `MatCholeskyFactor()`, `MatFactorInfo`
7064: @*/
7065: PetscErrorCode MatICCFactorSymbolic(Mat fact, Mat mat, IS perm, const MatFactorInfo *info)
7066: {
7067: PetscFunctionBegin;
7071: PetscAssertPointer(info, 4);
7072: PetscAssertPointer(fact, 1);
7073: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7074: PetscCheck(info->levels >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Levels negative %" PetscInt_FMT, (PetscInt)info->levels);
7075: PetscCheck(info->fill >= 1.0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Expected fill less than 1.0 %g", (double)info->fill);
7076: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7077: MatCheckPreallocated(mat, 2);
7079: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_ICCFactorSymbolic, mat, perm, 0, 0));
7080: PetscUseTypeMethod(fact, iccfactorsymbolic, mat, perm, info);
7081: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_ICCFactorSymbolic, mat, perm, 0, 0));
7082: PetscFunctionReturn(PETSC_SUCCESS);
7083: }
7085: /*@C
7086: MatCreateSubMatrices - Extracts several submatrices from a matrix. If submat
7087: points to an array of valid matrices, they may be reused to store the new
7088: submatrices.
7090: Collective
7092: Input Parameters:
7093: + mat - the matrix
7094: . n - the number of submatrixes to be extracted (on this processor, may be zero)
7095: . irow - index set of rows to extract
7096: . icol - index set of columns to extract
7097: - scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
7099: Output Parameter:
7100: . submat - the array of submatrices
7102: Level: advanced
7104: Notes:
7105: `MatCreateSubMatrices()` can extract ONLY sequential submatrices
7106: (from both sequential and parallel matrices). Use `MatCreateSubMatrix()`
7107: to extract a parallel submatrix.
7109: Some matrix types place restrictions on the row and column
7110: indices, such as that they be sorted or that they be equal to each other.
7112: The index sets may not have duplicate entries.
7114: When extracting submatrices from a parallel matrix, each processor can
7115: form a different submatrix by setting the rows and columns of its
7116: individual index sets according to the local submatrix desired.
7118: When finished using the submatrices, the user should destroy
7119: them with `MatDestroySubMatrices()`.
7121: `MAT_REUSE_MATRIX` can only be used when the nonzero structure of the
7122: original matrix has not changed from that last call to `MatCreateSubMatrices()`.
7124: This routine creates the matrices in submat; you should NOT create them before
7125: calling it. It also allocates the array of matrix pointers submat.
7127: For `MATBAIJ` matrices the index sets must respect the block structure, that is if they
7128: request one row/column in a block, they must request all rows/columns that are in
7129: that block. For example, if the block size is 2 you cannot request just row 0 and
7130: column 0.
7132: Fortran Note:
7133: The Fortran interface is slightly different from that given below; it
7134: requires one to pass in as `submat` a `Mat` (integer) array of size at least n+1.
7136: .seealso: [](ch_matrices), `Mat`, `MatDestroySubMatrices()`, `MatCreateSubMatrix()`, `MatGetRow()`, `MatGetDiagonal()`, `MatReuse`
7137: @*/
7138: PetscErrorCode MatCreateSubMatrices(Mat mat, PetscInt n, const IS irow[], const IS icol[], MatReuse scall, Mat *submat[])
7139: {
7140: PetscInt i;
7141: PetscBool eq;
7143: PetscFunctionBegin;
7146: if (n) {
7147: PetscAssertPointer(irow, 3);
7149: PetscAssertPointer(icol, 4);
7151: }
7152: PetscAssertPointer(submat, 6);
7153: if (n && scall == MAT_REUSE_MATRIX) {
7154: PetscAssertPointer(*submat, 6);
7156: }
7157: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7158: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7159: MatCheckPreallocated(mat, 1);
7160: PetscCall(PetscLogEventBegin(MAT_CreateSubMats, mat, 0, 0, 0));
7161: PetscUseTypeMethod(mat, createsubmatrices, n, irow, icol, scall, submat);
7162: PetscCall(PetscLogEventEnd(MAT_CreateSubMats, mat, 0, 0, 0));
7163: for (i = 0; i < n; i++) {
7164: (*submat)[i]->factortype = MAT_FACTOR_NONE; /* in case in place factorization was previously done on submatrix */
7165: PetscCall(ISEqualUnsorted(irow[i], icol[i], &eq));
7166: if (eq) PetscCall(MatPropagateSymmetryOptions(mat, (*submat)[i]));
7167: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
7168: if (mat->boundtocpu && mat->bindingpropagates) {
7169: PetscCall(MatBindToCPU((*submat)[i], PETSC_TRUE));
7170: PetscCall(MatSetBindingPropagates((*submat)[i], PETSC_TRUE));
7171: }
7172: #endif
7173: }
7174: PetscFunctionReturn(PETSC_SUCCESS);
7175: }
7177: /*@C
7178: MatCreateSubMatricesMPI - Extracts MPI submatrices across a sub communicator of mat (by pairs of `IS` that may live on subcomms).
7180: Collective
7182: Input Parameters:
7183: + mat - the matrix
7184: . n - the number of submatrixes to be extracted
7185: . irow - index set of rows to extract
7186: . icol - index set of columns to extract
7187: - scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
7189: Output Parameter:
7190: . submat - the array of submatrices
7192: Level: advanced
7194: Note:
7195: This is used by `PCGASM`
7197: .seealso: [](ch_matrices), `Mat`, `PCGASM`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRow()`, `MatGetDiagonal()`, `MatReuse`
7198: @*/
7199: PetscErrorCode MatCreateSubMatricesMPI(Mat mat, PetscInt n, const IS irow[], const IS icol[], MatReuse scall, Mat *submat[])
7200: {
7201: PetscInt i;
7202: PetscBool eq;
7204: PetscFunctionBegin;
7207: if (n) {
7208: PetscAssertPointer(irow, 3);
7210: PetscAssertPointer(icol, 4);
7212: }
7213: PetscAssertPointer(submat, 6);
7214: if (n && scall == MAT_REUSE_MATRIX) {
7215: PetscAssertPointer(*submat, 6);
7217: }
7218: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7219: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7220: MatCheckPreallocated(mat, 1);
7222: PetscCall(PetscLogEventBegin(MAT_CreateSubMats, mat, 0, 0, 0));
7223: PetscUseTypeMethod(mat, createsubmatricesmpi, n, irow, icol, scall, submat);
7224: PetscCall(PetscLogEventEnd(MAT_CreateSubMats, mat, 0, 0, 0));
7225: for (i = 0; i < n; i++) {
7226: PetscCall(ISEqualUnsorted(irow[i], icol[i], &eq));
7227: if (eq) PetscCall(MatPropagateSymmetryOptions(mat, (*submat)[i]));
7228: }
7229: PetscFunctionReturn(PETSC_SUCCESS);
7230: }
7232: /*@C
7233: MatDestroyMatrices - Destroys an array of matrices.
7235: Collective
7237: Input Parameters:
7238: + n - the number of local matrices
7239: - mat - the matrices (this is a pointer to the array of matrices)
7241: Level: advanced
7243: Note:
7244: Frees not only the matrices, but also the array that contains the matrices
7246: Fortran Note:
7247: This does not free the array.
7249: .seealso: [](ch_matrices), `Mat`, `MatCreateSubMatrices()` `MatDestroySubMatrices()`
7250: @*/
7251: PetscErrorCode MatDestroyMatrices(PetscInt n, Mat *mat[])
7252: {
7253: PetscInt i;
7255: PetscFunctionBegin;
7256: if (!*mat) PetscFunctionReturn(PETSC_SUCCESS);
7257: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Trying to destroy negative number of matrices %" PetscInt_FMT, n);
7258: PetscAssertPointer(mat, 2);
7260: for (i = 0; i < n; i++) PetscCall(MatDestroy(&(*mat)[i]));
7262: /* memory is allocated even if n = 0 */
7263: PetscCall(PetscFree(*mat));
7264: PetscFunctionReturn(PETSC_SUCCESS);
7265: }
7267: /*@C
7268: MatDestroySubMatrices - Destroys a set of matrices obtained with `MatCreateSubMatrices()`.
7270: Collective
7272: Input Parameters:
7273: + n - the number of local matrices
7274: - mat - the matrices (this is a pointer to the array of matrices, just to match the calling
7275: sequence of `MatCreateSubMatrices()`)
7277: Level: advanced
7279: Note:
7280: Frees not only the matrices, but also the array that contains the matrices
7282: Fortran Note:
7283: This does not free the array.
7285: .seealso: [](ch_matrices), `Mat`, `MatCreateSubMatrices()`, `MatDestroyMatrices()`
7286: @*/
7287: PetscErrorCode MatDestroySubMatrices(PetscInt n, Mat *mat[])
7288: {
7289: Mat mat0;
7291: PetscFunctionBegin;
7292: if (!*mat) PetscFunctionReturn(PETSC_SUCCESS);
7293: /* mat[] is an array of length n+1, see MatCreateSubMatrices_xxx() */
7294: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Trying to destroy negative number of matrices %" PetscInt_FMT, n);
7295: PetscAssertPointer(mat, 2);
7297: mat0 = (*mat)[0];
7298: if (mat0 && mat0->ops->destroysubmatrices) {
7299: PetscCall((*mat0->ops->destroysubmatrices)(n, mat));
7300: } else {
7301: PetscCall(MatDestroyMatrices(n, mat));
7302: }
7303: PetscFunctionReturn(PETSC_SUCCESS);
7304: }
7306: /*@
7307: MatGetSeqNonzeroStructure - Extracts the nonzero structure from a matrix and stores it, in its entirety, on each process
7309: Collective
7311: Input Parameter:
7312: . mat - the matrix
7314: Output Parameter:
7315: . matstruct - the sequential matrix with the nonzero structure of `mat`
7317: Level: developer
7319: .seealso: [](ch_matrices), `Mat`, `MatDestroySeqNonzeroStructure()`, `MatCreateSubMatrices()`, `MatDestroyMatrices()`
7320: @*/
7321: PetscErrorCode MatGetSeqNonzeroStructure(Mat mat, Mat *matstruct)
7322: {
7323: PetscFunctionBegin;
7325: PetscAssertPointer(matstruct, 2);
7328: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7329: MatCheckPreallocated(mat, 1);
7331: PetscCall(PetscLogEventBegin(MAT_GetSeqNonzeroStructure, mat, 0, 0, 0));
7332: PetscUseTypeMethod(mat, getseqnonzerostructure, matstruct);
7333: PetscCall(PetscLogEventEnd(MAT_GetSeqNonzeroStructure, mat, 0, 0, 0));
7334: PetscFunctionReturn(PETSC_SUCCESS);
7335: }
7337: /*@C
7338: MatDestroySeqNonzeroStructure - Destroys matrix obtained with `MatGetSeqNonzeroStructure()`.
7340: Collective
7342: Input Parameter:
7343: . mat - the matrix
7345: Level: advanced
7347: Note:
7348: This is not needed, one can just call `MatDestroy()`
7350: .seealso: [](ch_matrices), `Mat`, `MatGetSeqNonzeroStructure()`
7351: @*/
7352: PetscErrorCode MatDestroySeqNonzeroStructure(Mat *mat)
7353: {
7354: PetscFunctionBegin;
7355: PetscAssertPointer(mat, 1);
7356: PetscCall(MatDestroy(mat));
7357: PetscFunctionReturn(PETSC_SUCCESS);
7358: }
7360: /*@
7361: MatIncreaseOverlap - Given a set of submatrices indicated by index sets,
7362: replaces the index sets by larger ones that represent submatrices with
7363: additional overlap.
7365: Collective
7367: Input Parameters:
7368: + mat - the matrix
7369: . n - the number of index sets
7370: . is - the array of index sets (these index sets will changed during the call)
7371: - ov - the additional overlap requested
7373: Options Database Key:
7374: . -mat_increase_overlap_scalable - use a scalable algorithm to compute the overlap (supported by MPIAIJ matrix)
7376: Level: developer
7378: Note:
7379: The computed overlap preserves the matrix block sizes when the blocks are square.
7380: That is: if a matrix nonzero for a given block would increase the overlap all columns associated with
7381: that block are included in the overlap regardless of whether each specific column would increase the overlap.
7383: .seealso: [](ch_matrices), `Mat`, `PCASM`, `MatSetBlockSize()`, `MatIncreaseOverlapSplit()`, `MatCreateSubMatrices()`
7384: @*/
7385: PetscErrorCode MatIncreaseOverlap(Mat mat, PetscInt n, IS is[], PetscInt ov)
7386: {
7387: PetscInt i, bs, cbs;
7389: PetscFunctionBegin;
7393: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Must have one or more domains, you have %" PetscInt_FMT, n);
7394: if (n) {
7395: PetscAssertPointer(is, 3);
7397: }
7398: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7399: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7400: MatCheckPreallocated(mat, 1);
7402: if (!ov || !n) PetscFunctionReturn(PETSC_SUCCESS);
7403: PetscCall(PetscLogEventBegin(MAT_IncreaseOverlap, mat, 0, 0, 0));
7404: PetscUseTypeMethod(mat, increaseoverlap, n, is, ov);
7405: PetscCall(PetscLogEventEnd(MAT_IncreaseOverlap, mat, 0, 0, 0));
7406: PetscCall(MatGetBlockSizes(mat, &bs, &cbs));
7407: if (bs == cbs) {
7408: for (i = 0; i < n; i++) PetscCall(ISSetBlockSize(is[i], bs));
7409: }
7410: PetscFunctionReturn(PETSC_SUCCESS);
7411: }
7413: PetscErrorCode MatIncreaseOverlapSplit_Single(Mat, IS *, PetscInt);
7415: /*@
7416: MatIncreaseOverlapSplit - Given a set of submatrices indicated by index sets across
7417: a sub communicator, replaces the index sets by larger ones that represent submatrices with
7418: additional overlap.
7420: Collective
7422: Input Parameters:
7423: + mat - the matrix
7424: . n - the number of index sets
7425: . is - the array of index sets (these index sets will changed during the call)
7426: - ov - the additional overlap requested
7428: ` Options Database Key:
7429: . -mat_increase_overlap_scalable - use a scalable algorithm to compute the overlap (supported by MPIAIJ matrix)
7431: Level: developer
7433: .seealso: [](ch_matrices), `Mat`, `MatCreateSubMatrices()`, `MatIncreaseOverlap()`
7434: @*/
7435: PetscErrorCode MatIncreaseOverlapSplit(Mat mat, PetscInt n, IS is[], PetscInt ov)
7436: {
7437: PetscInt i;
7439: PetscFunctionBegin;
7442: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Must have one or more domains, you have %" PetscInt_FMT, n);
7443: if (n) {
7444: PetscAssertPointer(is, 3);
7446: }
7447: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7448: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7449: MatCheckPreallocated(mat, 1);
7450: if (!ov) PetscFunctionReturn(PETSC_SUCCESS);
7451: PetscCall(PetscLogEventBegin(MAT_IncreaseOverlap, mat, 0, 0, 0));
7452: for (i = 0; i < n; i++) PetscCall(MatIncreaseOverlapSplit_Single(mat, &is[i], ov));
7453: PetscCall(PetscLogEventEnd(MAT_IncreaseOverlap, mat, 0, 0, 0));
7454: PetscFunctionReturn(PETSC_SUCCESS);
7455: }
7457: /*@
7458: MatGetBlockSize - Returns the matrix block size.
7460: Not Collective
7462: Input Parameter:
7463: . mat - the matrix
7465: Output Parameter:
7466: . bs - block size
7468: Level: intermediate
7470: Notes:
7471: Block row formats are `MATBAIJ` and `MATSBAIJ` ALWAYS have square block storage in the matrix.
7473: If the block size has not been set yet this routine returns 1.
7475: .seealso: [](ch_matrices), `Mat`, `MATBAIJ`, `MATSBAIJ`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSizes()`
7476: @*/
7477: PetscErrorCode MatGetBlockSize(Mat mat, PetscInt *bs)
7478: {
7479: PetscFunctionBegin;
7481: PetscAssertPointer(bs, 2);
7482: *bs = PetscAbs(mat->rmap->bs);
7483: PetscFunctionReturn(PETSC_SUCCESS);
7484: }
7486: /*@
7487: MatGetBlockSizes - Returns the matrix block row and column sizes.
7489: Not Collective
7491: Input Parameter:
7492: . mat - the matrix
7494: Output Parameters:
7495: + rbs - row block size
7496: - cbs - column block size
7498: Level: intermediate
7500: Notes:
7501: Block row formats are `MATBAIJ` and `MATSBAIJ` ALWAYS have square block storage in the matrix.
7502: If you pass a different block size for the columns than the rows, the row block size determines the square block storage.
7504: If a block size has not been set yet this routine returns 1.
7506: .seealso: [](ch_matrices), `Mat`, `MATBAIJ`, `MATSBAIJ`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSize()`, `MatSetBlockSizes()`
7507: @*/
7508: PetscErrorCode MatGetBlockSizes(Mat mat, PetscInt *rbs, PetscInt *cbs)
7509: {
7510: PetscFunctionBegin;
7512: if (rbs) PetscAssertPointer(rbs, 2);
7513: if (cbs) PetscAssertPointer(cbs, 3);
7514: if (rbs) *rbs = PetscAbs(mat->rmap->bs);
7515: if (cbs) *cbs = PetscAbs(mat->cmap->bs);
7516: PetscFunctionReturn(PETSC_SUCCESS);
7517: }
7519: /*@
7520: MatSetBlockSize - Sets the matrix block size.
7522: Logically Collective
7524: Input Parameters:
7525: + mat - the matrix
7526: - bs - block size
7528: Level: intermediate
7530: Notes:
7531: Block row formats are `MATBAIJ` and `MATSBAIJ` formats ALWAYS have square block storage in the matrix.
7532: This must be called before `MatSetUp()` or MatXXXSetPreallocation() (or will default to 1) and the block size cannot be changed later.
7534: For `MATAIJ` matrix format, this function can be called at a later stage, provided that the specified block size
7535: is compatible with the matrix local sizes.
7537: .seealso: [](ch_matrices), `Mat`, `MATBAIJ`, `MATSBAIJ`, `MATAIJ`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`, `MatGetBlockSizes()`
7538: @*/
7539: PetscErrorCode MatSetBlockSize(Mat mat, PetscInt bs)
7540: {
7541: PetscFunctionBegin;
7544: PetscCall(MatSetBlockSizes(mat, bs, bs));
7545: PetscFunctionReturn(PETSC_SUCCESS);
7546: }
7548: typedef struct {
7549: PetscInt n;
7550: IS *is;
7551: Mat *mat;
7552: PetscObjectState nonzerostate;
7553: Mat C;
7554: } EnvelopeData;
7556: static PetscErrorCode EnvelopeDataDestroy(void *ptr)
7557: {
7558: EnvelopeData *edata = (EnvelopeData *)ptr;
7560: PetscFunctionBegin;
7561: for (PetscInt i = 0; i < edata->n; i++) PetscCall(ISDestroy(&edata->is[i]));
7562: PetscCall(PetscFree(edata->is));
7563: PetscCall(PetscFree(edata));
7564: PetscFunctionReturn(PETSC_SUCCESS);
7565: }
7567: /*@
7568: MatComputeVariableBlockEnvelope - Given a matrix whose nonzeros are in blocks along the diagonal this computes and stores
7569: the sizes of these blocks in the matrix. An individual block may lie over several processes.
7571: Collective
7573: Input Parameter:
7574: . mat - the matrix
7576: Level: intermediate
7578: Notes:
7579: There can be zeros within the blocks
7581: The blocks can overlap between processes, including laying on more than two processes
7583: .seealso: [](ch_matrices), `Mat`, `MatInvertVariableBlockEnvelope()`, `MatSetVariableBlockSizes()`
7584: @*/
7585: PetscErrorCode MatComputeVariableBlockEnvelope(Mat mat)
7586: {
7587: PetscInt n, *sizes, *starts, i = 0, env = 0, tbs = 0, lblocks = 0, rstart, II, ln = 0, cnt = 0, cstart, cend;
7588: PetscInt *diag, *odiag, sc;
7589: VecScatter scatter;
7590: PetscScalar *seqv;
7591: const PetscScalar *parv;
7592: const PetscInt *ia, *ja;
7593: PetscBool set, flag, done;
7594: Mat AA = mat, A;
7595: MPI_Comm comm;
7596: PetscMPIInt rank, size, tag;
7597: MPI_Status status;
7598: PetscContainer container;
7599: EnvelopeData *edata;
7600: Vec seq, par;
7601: IS isglobal;
7603: PetscFunctionBegin;
7605: PetscCall(MatIsSymmetricKnown(mat, &set, &flag));
7606: if (!set || !flag) {
7607: /* TODO: only needs nonzero structure of transpose */
7608: PetscCall(MatTranspose(mat, MAT_INITIAL_MATRIX, &AA));
7609: PetscCall(MatAXPY(AA, 1.0, mat, DIFFERENT_NONZERO_PATTERN));
7610: }
7611: PetscCall(MatAIJGetLocalMat(AA, &A));
7612: PetscCall(MatGetRowIJ(A, 0, PETSC_FALSE, PETSC_FALSE, &n, &ia, &ja, &done));
7613: PetscCheck(done, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Unable to get IJ structure from matrix");
7615: PetscCall(MatGetLocalSize(mat, &n, NULL));
7616: PetscCall(PetscObjectGetNewTag((PetscObject)mat, &tag));
7617: PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
7618: PetscCallMPI(MPI_Comm_size(comm, &size));
7619: PetscCallMPI(MPI_Comm_rank(comm, &rank));
7621: PetscCall(PetscMalloc2(n, &sizes, n, &starts));
7623: if (rank > 0) {
7624: PetscCallMPI(MPI_Recv(&env, 1, MPIU_INT, rank - 1, tag, comm, &status));
7625: PetscCallMPI(MPI_Recv(&tbs, 1, MPIU_INT, rank - 1, tag, comm, &status));
7626: }
7627: PetscCall(MatGetOwnershipRange(mat, &rstart, NULL));
7628: for (i = 0; i < n; i++) {
7629: env = PetscMax(env, ja[ia[i + 1] - 1]);
7630: II = rstart + i;
7631: if (env == II) {
7632: starts[lblocks] = tbs;
7633: sizes[lblocks++] = 1 + II - tbs;
7634: tbs = 1 + II;
7635: }
7636: }
7637: if (rank < size - 1) {
7638: PetscCallMPI(MPI_Send(&env, 1, MPIU_INT, rank + 1, tag, comm));
7639: PetscCallMPI(MPI_Send(&tbs, 1, MPIU_INT, rank + 1, tag, comm));
7640: }
7642: PetscCall(MatRestoreRowIJ(A, 0, PETSC_FALSE, PETSC_FALSE, &n, &ia, &ja, &done));
7643: if (!set || !flag) PetscCall(MatDestroy(&AA));
7644: PetscCall(MatDestroy(&A));
7646: PetscCall(PetscNew(&edata));
7647: PetscCall(MatGetNonzeroState(mat, &edata->nonzerostate));
7648: edata->n = lblocks;
7649: /* create IS needed for extracting blocks from the original matrix */
7650: PetscCall(PetscMalloc1(lblocks, &edata->is));
7651: for (PetscInt i = 0; i < lblocks; i++) PetscCall(ISCreateStride(PETSC_COMM_SELF, sizes[i], starts[i], 1, &edata->is[i]));
7653: /* Create the resulting inverse matrix structure with preallocation information */
7654: PetscCall(MatCreate(PetscObjectComm((PetscObject)mat), &edata->C));
7655: PetscCall(MatSetSizes(edata->C, mat->rmap->n, mat->cmap->n, mat->rmap->N, mat->cmap->N));
7656: PetscCall(MatSetBlockSizesFromMats(edata->C, mat, mat));
7657: PetscCall(MatSetType(edata->C, MATAIJ));
7659: /* Communicate the start and end of each row, from each block to the correct rank */
7660: /* TODO: Use PetscSF instead of VecScatter */
7661: for (PetscInt i = 0; i < lblocks; i++) ln += sizes[i];
7662: PetscCall(VecCreateSeq(PETSC_COMM_SELF, 2 * ln, &seq));
7663: PetscCall(VecGetArrayWrite(seq, &seqv));
7664: for (PetscInt i = 0; i < lblocks; i++) {
7665: for (PetscInt j = 0; j < sizes[i]; j++) {
7666: seqv[cnt] = starts[i];
7667: seqv[cnt + 1] = starts[i] + sizes[i];
7668: cnt += 2;
7669: }
7670: }
7671: PetscCall(VecRestoreArrayWrite(seq, &seqv));
7672: PetscCallMPI(MPI_Scan(&cnt, &sc, 1, MPIU_INT, MPI_SUM, PetscObjectComm((PetscObject)mat)));
7673: sc -= cnt;
7674: PetscCall(VecCreateMPI(PetscObjectComm((PetscObject)mat), 2 * mat->rmap->n, 2 * mat->rmap->N, &par));
7675: PetscCall(ISCreateStride(PETSC_COMM_SELF, cnt, sc, 1, &isglobal));
7676: PetscCall(VecScatterCreate(seq, NULL, par, isglobal, &scatter));
7677: PetscCall(ISDestroy(&isglobal));
7678: PetscCall(VecScatterBegin(scatter, seq, par, INSERT_VALUES, SCATTER_FORWARD));
7679: PetscCall(VecScatterEnd(scatter, seq, par, INSERT_VALUES, SCATTER_FORWARD));
7680: PetscCall(VecScatterDestroy(&scatter));
7681: PetscCall(VecDestroy(&seq));
7682: PetscCall(MatGetOwnershipRangeColumn(mat, &cstart, &cend));
7683: PetscCall(PetscMalloc2(mat->rmap->n, &diag, mat->rmap->n, &odiag));
7684: PetscCall(VecGetArrayRead(par, &parv));
7685: cnt = 0;
7686: PetscCall(MatGetSize(mat, NULL, &n));
7687: for (PetscInt i = 0; i < mat->rmap->n; i++) {
7688: PetscInt start, end, d = 0, od = 0;
7690: start = (PetscInt)PetscRealPart(parv[cnt]);
7691: end = (PetscInt)PetscRealPart(parv[cnt + 1]);
7692: cnt += 2;
7694: if (start < cstart) {
7695: od += cstart - start + n - cend;
7696: d += cend - cstart;
7697: } else if (start < cend) {
7698: od += n - cend;
7699: d += cend - start;
7700: } else od += n - start;
7701: if (end <= cstart) {
7702: od -= cstart - end + n - cend;
7703: d -= cend - cstart;
7704: } else if (end < cend) {
7705: od -= n - cend;
7706: d -= cend - end;
7707: } else od -= n - end;
7709: odiag[i] = od;
7710: diag[i] = d;
7711: }
7712: PetscCall(VecRestoreArrayRead(par, &parv));
7713: PetscCall(VecDestroy(&par));
7714: PetscCall(MatXAIJSetPreallocation(edata->C, mat->rmap->bs, diag, odiag, NULL, NULL));
7715: PetscCall(PetscFree2(diag, odiag));
7716: PetscCall(PetscFree2(sizes, starts));
7718: PetscCall(PetscContainerCreate(PETSC_COMM_SELF, &container));
7719: PetscCall(PetscContainerSetPointer(container, edata));
7720: PetscCall(PetscContainerSetUserDestroy(container, (PetscErrorCode(*)(void *))EnvelopeDataDestroy));
7721: PetscCall(PetscObjectCompose((PetscObject)mat, "EnvelopeData", (PetscObject)container));
7722: PetscCall(PetscObjectDereference((PetscObject)container));
7723: PetscFunctionReturn(PETSC_SUCCESS);
7724: }
7726: /*@
7727: MatInvertVariableBlockEnvelope - set matrix C to be the inverted block diagonal of matrix A
7729: Collective
7731: Input Parameters:
7732: + A - the matrix
7733: - reuse - indicates if the `C` matrix was obtained from a previous call to this routine
7735: Output Parameter:
7736: . C - matrix with inverted block diagonal of `A`
7738: Level: advanced
7740: Note:
7741: For efficiency the matrix `A` should have all the nonzero entries clustered in smallish blocks along the diagonal.
7743: .seealso: [](ch_matrices), `Mat`, `MatInvertBlockDiagonal()`, `MatComputeBlockDiagonal()`
7744: @*/
7745: PetscErrorCode MatInvertVariableBlockEnvelope(Mat A, MatReuse reuse, Mat *C)
7746: {
7747: PetscContainer container;
7748: EnvelopeData *edata;
7749: PetscObjectState nonzerostate;
7751: PetscFunctionBegin;
7752: PetscCall(PetscObjectQuery((PetscObject)A, "EnvelopeData", (PetscObject *)&container));
7753: if (!container) {
7754: PetscCall(MatComputeVariableBlockEnvelope(A));
7755: PetscCall(PetscObjectQuery((PetscObject)A, "EnvelopeData", (PetscObject *)&container));
7756: }
7757: PetscCall(PetscContainerGetPointer(container, (void **)&edata));
7758: PetscCall(MatGetNonzeroState(A, &nonzerostate));
7759: PetscCheck(nonzerostate <= edata->nonzerostate, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Cannot handle changes to matrix nonzero structure");
7760: PetscCheck(reuse != MAT_REUSE_MATRIX || *C == edata->C, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "C matrix must be the same as previously output");
7762: PetscCall(MatCreateSubMatrices(A, edata->n, edata->is, edata->is, MAT_INITIAL_MATRIX, &edata->mat));
7763: *C = edata->C;
7765: for (PetscInt i = 0; i < edata->n; i++) {
7766: Mat D;
7767: PetscScalar *dvalues;
7769: PetscCall(MatConvert(edata->mat[i], MATSEQDENSE, MAT_INITIAL_MATRIX, &D));
7770: PetscCall(MatSetOption(*C, MAT_ROW_ORIENTED, PETSC_FALSE));
7771: PetscCall(MatSeqDenseInvert(D));
7772: PetscCall(MatDenseGetArray(D, &dvalues));
7773: PetscCall(MatSetValuesIS(*C, edata->is[i], edata->is[i], dvalues, INSERT_VALUES));
7774: PetscCall(MatDestroy(&D));
7775: }
7776: PetscCall(MatDestroySubMatrices(edata->n, &edata->mat));
7777: PetscCall(MatAssemblyBegin(*C, MAT_FINAL_ASSEMBLY));
7778: PetscCall(MatAssemblyEnd(*C, MAT_FINAL_ASSEMBLY));
7779: PetscFunctionReturn(PETSC_SUCCESS);
7780: }
7782: /*@
7783: MatSetVariableBlockSizes - Sets diagonal point-blocks of the matrix that need not be of the same size
7785: Not Collective
7787: Input Parameters:
7788: + mat - the matrix
7789: . nblocks - the number of blocks on this process, each block can only exist on a single process
7790: - bsizes - the block sizes
7792: Level: intermediate
7794: Notes:
7795: Currently used by `PCVPBJACOBI` for `MATAIJ` matrices
7797: 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.
7799: .seealso: [](ch_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`, `MatGetBlockSizes()`, `MatGetVariableBlockSizes()`,
7800: `MatComputeVariableBlockEnvelope()`, `PCVPBJACOBI`
7801: @*/
7802: PetscErrorCode MatSetVariableBlockSizes(Mat mat, PetscInt nblocks, const PetscInt bsizes[])
7803: {
7804: PetscInt ncnt = 0, nlocal;
7806: PetscFunctionBegin;
7808: PetscCall(MatGetLocalSize(mat, &nlocal, NULL));
7809: PetscCheck(nblocks >= 0 && nblocks <= nlocal, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Number of local blocks %" PetscInt_FMT " is not in [0, %" PetscInt_FMT "]", nblocks, nlocal);
7810: for (PetscInt i = 0; i < nblocks; i++) ncnt += bsizes[i];
7811: 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);
7812: PetscCall(PetscFree(mat->bsizes));
7813: mat->nblocks = nblocks;
7814: PetscCall(PetscMalloc1(nblocks, &mat->bsizes));
7815: PetscCall(PetscArraycpy(mat->bsizes, bsizes, nblocks));
7816: PetscFunctionReturn(PETSC_SUCCESS);
7817: }
7819: /*@C
7820: MatGetVariableBlockSizes - Gets a diagonal blocks of the matrix that need not be of the same size
7822: Not Collective; No Fortran Support
7824: Input Parameter:
7825: . mat - the matrix
7827: Output Parameters:
7828: + nblocks - the number of blocks on this process
7829: - bsizes - the block sizes
7831: Level: intermediate
7833: .seealso: [](ch_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`, `MatGetBlockSizes()`, `MatSetVariableBlockSizes()`, `MatComputeVariableBlockEnvelope()`
7834: @*/
7835: PetscErrorCode MatGetVariableBlockSizes(Mat mat, PetscInt *nblocks, const PetscInt *bsizes[])
7836: {
7837: PetscFunctionBegin;
7839: if (nblocks) *nblocks = mat->nblocks;
7840: if (bsizes) *bsizes = mat->bsizes;
7841: PetscFunctionReturn(PETSC_SUCCESS);
7842: }
7844: /*@
7845: MatSetBlockSizes - Sets the matrix block row and column sizes.
7847: Logically Collective
7849: Input Parameters:
7850: + mat - the matrix
7851: . rbs - row block size
7852: - cbs - column block size
7854: Level: intermediate
7856: Notes:
7857: Block row formats are `MATBAIJ` and `MATSBAIJ`. These formats ALWAYS have square block storage in the matrix.
7858: If you pass a different block size for the columns than the rows, the row block size determines the square block storage.
7859: This must be called before `MatSetUp()` or MatXXXSetPreallocation() (or will default to 1) and the block size cannot be changed later.
7861: For `MATAIJ` matrix this function can be called at a later stage, provided that the specified block sizes
7862: are compatible with the matrix local sizes.
7864: The row and column block size determine the blocksize of the "row" and "column" vectors returned by `MatCreateVecs()`.
7866: .seealso: [](ch_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSize()`, `MatGetBlockSizes()`
7867: @*/
7868: PetscErrorCode MatSetBlockSizes(Mat mat, PetscInt rbs, PetscInt cbs)
7869: {
7870: PetscFunctionBegin;
7874: PetscTryTypeMethod(mat, setblocksizes, rbs, cbs);
7875: if (mat->rmap->refcnt) {
7876: ISLocalToGlobalMapping l2g = NULL;
7877: PetscLayout nmap = NULL;
7879: PetscCall(PetscLayoutDuplicate(mat->rmap, &nmap));
7880: if (mat->rmap->mapping) PetscCall(ISLocalToGlobalMappingDuplicate(mat->rmap->mapping, &l2g));
7881: PetscCall(PetscLayoutDestroy(&mat->rmap));
7882: mat->rmap = nmap;
7883: mat->rmap->mapping = l2g;
7884: }
7885: if (mat->cmap->refcnt) {
7886: ISLocalToGlobalMapping l2g = NULL;
7887: PetscLayout nmap = NULL;
7889: PetscCall(PetscLayoutDuplicate(mat->cmap, &nmap));
7890: if (mat->cmap->mapping) PetscCall(ISLocalToGlobalMappingDuplicate(mat->cmap->mapping, &l2g));
7891: PetscCall(PetscLayoutDestroy(&mat->cmap));
7892: mat->cmap = nmap;
7893: mat->cmap->mapping = l2g;
7894: }
7895: PetscCall(PetscLayoutSetBlockSize(mat->rmap, rbs));
7896: PetscCall(PetscLayoutSetBlockSize(mat->cmap, cbs));
7897: PetscFunctionReturn(PETSC_SUCCESS);
7898: }
7900: /*@
7901: MatSetBlockSizesFromMats - Sets the matrix block row and column sizes to match a pair of matrices
7903: Logically Collective
7905: Input Parameters:
7906: + mat - the matrix
7907: . fromRow - matrix from which to copy row block size
7908: - fromCol - matrix from which to copy column block size (can be same as fromRow)
7910: Level: developer
7912: .seealso: [](ch_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`
7913: @*/
7914: PetscErrorCode MatSetBlockSizesFromMats(Mat mat, Mat fromRow, Mat fromCol)
7915: {
7916: PetscFunctionBegin;
7920: if (fromRow->rmap->bs > 0) PetscCall(PetscLayoutSetBlockSize(mat->rmap, fromRow->rmap->bs));
7921: if (fromCol->cmap->bs > 0) PetscCall(PetscLayoutSetBlockSize(mat->cmap, fromCol->cmap->bs));
7922: PetscFunctionReturn(PETSC_SUCCESS);
7923: }
7925: /*@
7926: MatResidual - Default routine to calculate the residual r = b - Ax
7928: Collective
7930: Input Parameters:
7931: + mat - the matrix
7932: . b - the right-hand-side
7933: - x - the approximate solution
7935: Output Parameter:
7936: . r - location to store the residual
7938: Level: developer
7940: .seealso: [](ch_matrices), `Mat`, `MatMult()`, `MatMultAdd()`, `PCMGSetResidual()`
7941: @*/
7942: PetscErrorCode MatResidual(Mat mat, Vec b, Vec x, Vec r)
7943: {
7944: PetscFunctionBegin;
7950: MatCheckPreallocated(mat, 1);
7951: PetscCall(PetscLogEventBegin(MAT_Residual, mat, 0, 0, 0));
7952: if (!mat->ops->residual) {
7953: PetscCall(MatMult(mat, x, r));
7954: PetscCall(VecAYPX(r, -1.0, b));
7955: } else {
7956: PetscUseTypeMethod(mat, residual, b, x, r);
7957: }
7958: PetscCall(PetscLogEventEnd(MAT_Residual, mat, 0, 0, 0));
7959: PetscFunctionReturn(PETSC_SUCCESS);
7960: }
7962: /*MC
7963: MatGetRowIJF90 - Obtains the compressed row storage i and j indices for the local rows of a sparse matrix
7965: Synopsis:
7966: MatGetRowIJF90(Mat A, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt n, {PetscInt, pointer :: ia(:)}, {PetscInt, pointer :: ja(:)}, PetscBool done,integer ierr)
7968: Not Collective
7970: Input Parameters:
7971: + A - the matrix
7972: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
7973: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
7974: - inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
7975: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
7976: always used.
7978: Output Parameters:
7979: + n - number of local rows in the (possibly compressed) matrix
7980: . ia - the row pointers; that is ia[0] = 0, ia[row] = ia[row-1] + number of elements in that row of the matrix
7981: . ja - the column indices
7982: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
7983: are responsible for handling the case when done == `PETSC_FALSE` and ia and ja are not set
7985: Level: developer
7987: Note:
7988: Use `MatRestoreRowIJF90()` when you no longer need access to the data
7990: .seealso: [](ch_matrices), [](sec_fortranarrays), `Mat`, `MATMPIAIJ`, `MatGetRowIJ()`, `MatRestoreRowIJ()`, `MatRestoreRowIJF90()`
7991: M*/
7993: /*MC
7994: MatRestoreRowIJF90 - restores the compressed row storage i and j indices for the local rows of a sparse matrix obtained with `MatGetRowIJF90()`
7996: Synopsis:
7997: MatRestoreRowIJF90(Mat A, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt n, {PetscInt, pointer :: ia(:)}, {PetscInt, pointer :: ja(:)}, PetscBool done,integer ierr)
7999: Not Collective
8001: Input Parameters:
8002: + A - the matrix
8003: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
8004: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
8005: inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8006: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8007: always used.
8008: . n - number of local rows in the (possibly compressed) matrix
8009: . ia - the row pointers; that is ia[0] = 0, ia[row] = ia[row-1] + number of elements in that row of the matrix
8010: . ja - the column indices
8011: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
8012: are responsible for handling the case when done == `PETSC_FALSE` and ia and ja are not set
8014: Level: developer
8016: .seealso: [](ch_matrices), [](sec_fortranarrays), `Mat`, `MATMPIAIJ`, `MatGetRowIJ()`, `MatRestoreRowIJ()`, `MatGetRowIJF90()`
8017: M*/
8019: /*@C
8020: MatGetRowIJ - Returns the compressed row storage i and j indices for the local rows of a sparse matrix
8022: Collective
8024: Input Parameters:
8025: + mat - the matrix
8026: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
8027: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
8028: - inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8029: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8030: always used.
8032: Output Parameters:
8033: + n - number of local rows in the (possibly compressed) matrix, use `NULL` if not needed
8034: . 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
8035: . ja - the column indices, use `NULL` if not needed
8036: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
8037: are responsible for handling the case when done == `PETSC_FALSE` and ia and ja are not set
8039: Level: developer
8041: Notes:
8042: You CANNOT change any of the ia[] or ja[] values.
8044: Use `MatRestoreRowIJ()` when you are finished accessing the ia[] and ja[] values.
8046: Fortran Notes:
8047: Use
8048: .vb
8049: PetscInt, pointer :: ia(:),ja(:)
8050: call MatGetRowIJF90(mat,shift,symmetric,inodecompressed,n,ia,ja,done,ierr)
8051: ! Access the ith and jth entries via ia(i) and ja(j)
8052: .ve
8054: `MatGetRowIJ()` Fortran binding is deprecated (since PETSc 3.19), use `MatGetRowIJF90()`
8056: .seealso: [](ch_matrices), `Mat`, `MATAIJ`, `MatGetRowIJF90()`, `MatGetColumnIJ()`, `MatRestoreRowIJ()`, `MatSeqAIJGetArray()`
8057: @*/
8058: PetscErrorCode MatGetRowIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
8059: {
8060: PetscFunctionBegin;
8063: if (n) PetscAssertPointer(n, 5);
8064: if (ia) PetscAssertPointer(ia, 6);
8065: if (ja) PetscAssertPointer(ja, 7);
8066: if (done) PetscAssertPointer(done, 8);
8067: MatCheckPreallocated(mat, 1);
8068: if (!mat->ops->getrowij && done) *done = PETSC_FALSE;
8069: else {
8070: if (done) *done = PETSC_TRUE;
8071: PetscCall(PetscLogEventBegin(MAT_GetRowIJ, mat, 0, 0, 0));
8072: PetscUseTypeMethod(mat, getrowij, shift, symmetric, inodecompressed, n, ia, ja, done);
8073: PetscCall(PetscLogEventEnd(MAT_GetRowIJ, mat, 0, 0, 0));
8074: }
8075: PetscFunctionReturn(PETSC_SUCCESS);
8076: }
8078: /*@C
8079: MatGetColumnIJ - Returns the compressed column storage i and j indices for sequential matrices.
8081: Collective
8083: Input Parameters:
8084: + mat - the matrix
8085: . shift - 1 or zero indicating we want the indices starting at 0 or 1
8086: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be
8087: symmetrized
8088: . inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8089: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8090: always used.
8091: . n - number of columns in the (possibly compressed) matrix
8092: . ia - the column pointers; that is ia[0] = 0, ia[col] = i[col-1] + number of elements in that col of the matrix
8093: - ja - the row indices
8095: Output Parameter:
8096: . done - `PETSC_TRUE` or `PETSC_FALSE`, indicating whether the values have been returned
8098: Level: developer
8100: .seealso: [](ch_matrices), `Mat`, `MatGetRowIJ()`, `MatRestoreColumnIJ()`
8101: @*/
8102: PetscErrorCode MatGetColumnIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
8103: {
8104: PetscFunctionBegin;
8107: PetscAssertPointer(n, 5);
8108: if (ia) PetscAssertPointer(ia, 6);
8109: if (ja) PetscAssertPointer(ja, 7);
8110: PetscAssertPointer(done, 8);
8111: MatCheckPreallocated(mat, 1);
8112: if (!mat->ops->getcolumnij) *done = PETSC_FALSE;
8113: else {
8114: *done = PETSC_TRUE;
8115: PetscUseTypeMethod(mat, getcolumnij, shift, symmetric, inodecompressed, n, ia, ja, done);
8116: }
8117: PetscFunctionReturn(PETSC_SUCCESS);
8118: }
8120: /*@C
8121: MatRestoreRowIJ - Call after you are completed with the ia,ja indices obtained with `MatGetRowIJ()`.
8123: Collective
8125: Input Parameters:
8126: + mat - the matrix
8127: . shift - 1 or zero indicating we want the indices starting at 0 or 1
8128: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
8129: . inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8130: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8131: always used.
8132: . n - size of (possibly compressed) matrix
8133: . ia - the row pointers
8134: - ja - the column indices
8136: Output Parameter:
8137: . done - `PETSC_TRUE` or `PETSC_FALSE` indicated that the values have been returned
8139: Level: developer
8141: Note:
8142: This routine zeros out `n`, `ia`, and `ja`. This is to prevent accidental
8143: us of the array after it has been restored. If you pass `NULL`, it will
8144: not zero the pointers. Use of ia or ja after `MatRestoreRowIJ()` is invalid.
8146: Fortran Note:
8147: `MatRestoreRowIJ()` Fortran binding is deprecated (since PETSc 3.19), use `MatRestoreRowIJF90()`
8149: .seealso: [](ch_matrices), `Mat`, `MatGetRowIJ()`, `MatRestoreRowIJF90()`, `MatRestoreColumnIJ()`
8150: @*/
8151: PetscErrorCode MatRestoreRowIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
8152: {
8153: PetscFunctionBegin;
8156: if (ia) PetscAssertPointer(ia, 6);
8157: if (ja) PetscAssertPointer(ja, 7);
8158: if (done) PetscAssertPointer(done, 8);
8159: MatCheckPreallocated(mat, 1);
8161: if (!mat->ops->restorerowij && done) *done = PETSC_FALSE;
8162: else {
8163: if (done) *done = PETSC_TRUE;
8164: PetscUseTypeMethod(mat, restorerowij, shift, symmetric, inodecompressed, n, ia, ja, done);
8165: if (n) *n = 0;
8166: if (ia) *ia = NULL;
8167: if (ja) *ja = NULL;
8168: }
8169: PetscFunctionReturn(PETSC_SUCCESS);
8170: }
8172: /*@C
8173: MatRestoreColumnIJ - Call after you are completed with the ia,ja indices obtained with `MatGetColumnIJ()`.
8175: Collective
8177: Input Parameters:
8178: + mat - the matrix
8179: . shift - 1 or zero indicating we want the indices starting at 0 or 1
8180: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
8181: - inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8182: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8183: always used.
8185: Output Parameters:
8186: + n - size of (possibly compressed) matrix
8187: . ia - the column pointers
8188: . ja - the row indices
8189: - done - `PETSC_TRUE` or `PETSC_FALSE` indicated that the values have been returned
8191: Level: developer
8193: .seealso: [](ch_matrices), `Mat`, `MatGetColumnIJ()`, `MatRestoreRowIJ()`
8194: @*/
8195: PetscErrorCode MatRestoreColumnIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
8196: {
8197: PetscFunctionBegin;
8200: if (ia) PetscAssertPointer(ia, 6);
8201: if (ja) PetscAssertPointer(ja, 7);
8202: PetscAssertPointer(done, 8);
8203: MatCheckPreallocated(mat, 1);
8205: if (!mat->ops->restorecolumnij) *done = PETSC_FALSE;
8206: else {
8207: *done = PETSC_TRUE;
8208: PetscUseTypeMethod(mat, restorecolumnij, shift, symmetric, inodecompressed, n, ia, ja, done);
8209: if (n) *n = 0;
8210: if (ia) *ia = NULL;
8211: if (ja) *ja = NULL;
8212: }
8213: PetscFunctionReturn(PETSC_SUCCESS);
8214: }
8216: /*@
8217: MatColoringPatch - Used inside matrix coloring routines that use `MatGetRowIJ()` and/or
8218: `MatGetColumnIJ()`.
8220: Collective
8222: Input Parameters:
8223: + mat - the matrix
8224: . ncolors - maximum color value
8225: . n - number of entries in colorarray
8226: - colorarray - array indicating color for each column
8228: Output Parameter:
8229: . iscoloring - coloring generated using colorarray information
8231: Level: developer
8233: .seealso: [](ch_matrices), `Mat`, `MatGetRowIJ()`, `MatGetColumnIJ()`
8234: @*/
8235: PetscErrorCode MatColoringPatch(Mat mat, PetscInt ncolors, PetscInt n, ISColoringValue colorarray[], ISColoring *iscoloring)
8236: {
8237: PetscFunctionBegin;
8240: PetscAssertPointer(colorarray, 4);
8241: PetscAssertPointer(iscoloring, 5);
8242: MatCheckPreallocated(mat, 1);
8244: if (!mat->ops->coloringpatch) {
8245: PetscCall(ISColoringCreate(PetscObjectComm((PetscObject)mat), ncolors, n, colorarray, PETSC_OWN_POINTER, iscoloring));
8246: } else {
8247: PetscUseTypeMethod(mat, coloringpatch, ncolors, n, colorarray, iscoloring);
8248: }
8249: PetscFunctionReturn(PETSC_SUCCESS);
8250: }
8252: /*@
8253: MatSetUnfactored - Resets a factored matrix to be treated as unfactored.
8255: Logically Collective
8257: Input Parameter:
8258: . mat - the factored matrix to be reset
8260: Level: developer
8262: Notes:
8263: This routine should be used only with factored matrices formed by in-place
8264: factorization via ILU(0) (or by in-place LU factorization for the `MATSEQDENSE`
8265: format). This option can save memory, for example, when solving nonlinear
8266: systems with a matrix-free Newton-Krylov method and a matrix-based, in-place
8267: ILU(0) preconditioner.
8269: One can specify in-place ILU(0) factorization by calling
8270: .vb
8271: PCType(pc,PCILU);
8272: PCFactorSeUseInPlace(pc);
8273: .ve
8274: or by using the options -pc_type ilu -pc_factor_in_place
8276: In-place factorization ILU(0) can also be used as a local
8277: solver for the blocks within the block Jacobi or additive Schwarz
8278: methods (runtime option: -sub_pc_factor_in_place). See Users-Manual: ch_pc
8279: for details on setting local solver options.
8281: Most users should employ the `KSP` interface for linear solvers
8282: instead of working directly with matrix algebra routines such as this.
8283: See, e.g., `KSPCreate()`.
8285: .seealso: [](ch_matrices), `Mat`, `PCFactorSetUseInPlace()`, `PCFactorGetUseInPlace()`
8286: @*/
8287: PetscErrorCode MatSetUnfactored(Mat mat)
8288: {
8289: PetscFunctionBegin;
8292: MatCheckPreallocated(mat, 1);
8293: mat->factortype = MAT_FACTOR_NONE;
8294: if (!mat->ops->setunfactored) PetscFunctionReturn(PETSC_SUCCESS);
8295: PetscUseTypeMethod(mat, setunfactored);
8296: PetscFunctionReturn(PETSC_SUCCESS);
8297: }
8299: /*MC
8300: MatDenseGetArrayF90 - Accesses a matrix array from Fortran
8302: Synopsis:
8303: MatDenseGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:,:)},integer ierr)
8305: Not Collective
8307: Input Parameter:
8308: . x - matrix
8310: Output Parameters:
8311: + xx_v - the Fortran pointer to the array
8312: - ierr - error code
8314: Example of Usage:
8315: .vb
8316: PetscScalar, pointer xx_v(:,:)
8317: ....
8318: call MatDenseGetArrayF90(x,xx_v,ierr)
8319: a = xx_v(3)
8320: call MatDenseRestoreArrayF90(x,xx_v,ierr)
8321: .ve
8323: Level: advanced
8325: .seealso: [](ch_matrices), `Mat`, `MatDenseRestoreArrayF90()`, `MatDenseGetArray()`, `MatDenseRestoreArray()`, `MatSeqAIJGetArrayF90()`
8326: M*/
8328: /*MC
8329: MatDenseRestoreArrayF90 - Restores a matrix array that has been
8330: accessed with `MatDenseGetArrayF90()`.
8332: Synopsis:
8333: MatDenseRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:,:)},integer ierr)
8335: Not Collective
8337: Input Parameters:
8338: + x - matrix
8339: - xx_v - the Fortran90 pointer to the array
8341: Output Parameter:
8342: . ierr - error code
8344: Example of Usage:
8345: .vb
8346: PetscScalar, pointer xx_v(:,:)
8347: ....
8348: call MatDenseGetArrayF90(x,xx_v,ierr)
8349: a = xx_v(3)
8350: call MatDenseRestoreArrayF90(x,xx_v,ierr)
8351: .ve
8353: Level: advanced
8355: .seealso: [](ch_matrices), `Mat`, `MatDenseGetArrayF90()`, `MatDenseGetArray()`, `MatDenseRestoreArray()`, `MatSeqAIJRestoreArrayF90()`
8356: M*/
8358: /*MC
8359: MatSeqAIJGetArrayF90 - Accesses a matrix array from Fortran.
8361: Synopsis:
8362: MatSeqAIJGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
8364: Not Collective
8366: Input Parameter:
8367: . x - matrix
8369: Output Parameters:
8370: + xx_v - the Fortran pointer to the array
8371: - ierr - error code
8373: Example of Usage:
8374: .vb
8375: PetscScalar, pointer xx_v(:)
8376: ....
8377: call MatSeqAIJGetArrayF90(x,xx_v,ierr)
8378: a = xx_v(3)
8379: call MatSeqAIJRestoreArrayF90(x,xx_v,ierr)
8380: .ve
8382: Level: advanced
8384: .seealso: [](ch_matrices), `Mat`, `MatSeqAIJRestoreArrayF90()`, `MatSeqAIJGetArray()`, `MatSeqAIJRestoreArray()`, `MatDenseGetArrayF90()`
8385: M*/
8387: /*MC
8388: MatSeqAIJRestoreArrayF90 - Restores a matrix array that has been
8389: accessed with `MatSeqAIJGetArrayF90()`.
8391: Synopsis:
8392: MatSeqAIJRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
8394: Not Collective
8396: Input Parameters:
8397: + x - matrix
8398: - xx_v - the Fortran90 pointer to the array
8400: Output Parameter:
8401: . ierr - error code
8403: Example of Usage:
8404: .vb
8405: PetscScalar, pointer xx_v(:)
8406: ....
8407: call MatSeqAIJGetArrayF90(x,xx_v,ierr)
8408: a = xx_v(3)
8409: call MatSeqAIJRestoreArrayF90(x,xx_v,ierr)
8410: .ve
8412: Level: advanced
8414: .seealso: [](ch_matrices), `Mat`, `MatSeqAIJGetArrayF90()`, `MatSeqAIJGetArray()`, `MatSeqAIJRestoreArray()`, `MatDenseRestoreArrayF90()`
8415: M*/
8417: /*@
8418: MatCreateSubMatrix - Gets a single submatrix on the same number of processors
8419: as the original matrix.
8421: Collective
8423: Input Parameters:
8424: + mat - the original matrix
8425: . isrow - parallel `IS` containing the rows this processor should obtain
8426: . 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.
8427: - cll - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
8429: Output Parameter:
8430: . newmat - the new submatrix, of the same type as the original matrix
8432: Level: advanced
8434: Notes:
8435: The submatrix will be able to be multiplied with vectors using the same layout as `iscol`.
8437: Some matrix types place restrictions on the row and column indices, such
8438: 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;
8439: for example, if the block size is 3 one cannot select the 0 and 2 rows without selecting the 1 row.
8441: The index sets may not have duplicate entries.
8443: The first time this is called you should use a cll of `MAT_INITIAL_MATRIX`,
8444: the `MatCreateSubMatrix()` routine will create the newmat for you. Any additional calls
8445: to this routine with a mat of the same nonzero structure and with a call of `MAT_REUSE_MATRIX`
8446: will reuse the matrix generated the first time. You should call `MatDestroy()` on `newmat` when
8447: you are finished using it.
8449: The communicator of the newly obtained matrix is ALWAYS the same as the communicator of
8450: the input matrix.
8452: If `iscol` is `NULL` then all columns are obtained (not supported in Fortran).
8454: If `isrow` and `iscol` have a nontrivial block-size then the resulting matrix has this block-size as well. This feature
8455: is used by `PCFIELDSPLIT` to allow easy nesting of its use.
8457: Example usage:
8458: Consider the following 8x8 matrix with 34 non-zero values, that is
8459: assembled across 3 processors. Let's assume that proc0 owns 3 rows,
8460: proc1 owns 3 rows, proc2 owns 2 rows. This division can be shown
8461: as follows
8462: .vb
8463: 1 2 0 | 0 3 0 | 0 4
8464: Proc0 0 5 6 | 7 0 0 | 8 0
8465: 9 0 10 | 11 0 0 | 12 0
8466: -------------------------------------
8467: 13 0 14 | 15 16 17 | 0 0
8468: Proc1 0 18 0 | 19 20 21 | 0 0
8469: 0 0 0 | 22 23 0 | 24 0
8470: -------------------------------------
8471: Proc2 25 26 27 | 0 0 28 | 29 0
8472: 30 0 0 | 31 32 33 | 0 34
8473: .ve
8475: Suppose `isrow` = [0 1 | 4 | 6 7] and `iscol` = [1 2 | 3 4 5 | 6]. The resulting submatrix is
8477: .vb
8478: 2 0 | 0 3 0 | 0
8479: Proc0 5 6 | 7 0 0 | 8
8480: -------------------------------
8481: Proc1 18 0 | 19 20 21 | 0
8482: -------------------------------
8483: Proc2 26 27 | 0 0 28 | 29
8484: 0 0 | 31 32 33 | 0
8485: .ve
8487: .seealso: [](ch_matrices), `Mat`, `MatCreateSubMatrices()`, `MatCreateSubMatricesMPI()`, `MatCreateSubMatrixVirtual()`, `MatSubMatrixVirtualUpdate()`
8488: @*/
8489: PetscErrorCode MatCreateSubMatrix(Mat mat, IS isrow, IS iscol, MatReuse cll, Mat *newmat)
8490: {
8491: PetscMPIInt size;
8492: Mat *local;
8493: IS iscoltmp;
8494: PetscBool flg;
8496: PetscFunctionBegin;
8500: PetscAssertPointer(newmat, 5);
8503: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
8504: PetscCheck(cll != MAT_IGNORE_MATRIX, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Cannot use MAT_IGNORE_MATRIX");
8506: MatCheckPreallocated(mat, 1);
8507: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
8509: if (!iscol || isrow == iscol) {
8510: PetscBool stride;
8511: PetscMPIInt grabentirematrix = 0, grab;
8512: PetscCall(PetscObjectTypeCompare((PetscObject)isrow, ISSTRIDE, &stride));
8513: if (stride) {
8514: PetscInt first, step, n, rstart, rend;
8515: PetscCall(ISStrideGetInfo(isrow, &first, &step));
8516: if (step == 1) {
8517: PetscCall(MatGetOwnershipRange(mat, &rstart, &rend));
8518: if (rstart == first) {
8519: PetscCall(ISGetLocalSize(isrow, &n));
8520: if (n == rend - rstart) grabentirematrix = 1;
8521: }
8522: }
8523: }
8524: PetscCall(MPIU_Allreduce(&grabentirematrix, &grab, 1, MPI_INT, MPI_MIN, PetscObjectComm((PetscObject)mat)));
8525: if (grab) {
8526: PetscCall(PetscInfo(mat, "Getting entire matrix as submatrix\n"));
8527: if (cll == MAT_INITIAL_MATRIX) {
8528: *newmat = mat;
8529: PetscCall(PetscObjectReference((PetscObject)mat));
8530: }
8531: PetscFunctionReturn(PETSC_SUCCESS);
8532: }
8533: }
8535: if (!iscol) {
8536: PetscCall(ISCreateStride(PetscObjectComm((PetscObject)mat), mat->cmap->n, mat->cmap->rstart, 1, &iscoltmp));
8537: } else {
8538: iscoltmp = iscol;
8539: }
8541: /* if original matrix is on just one processor then use submatrix generated */
8542: if (mat->ops->createsubmatrices && !mat->ops->createsubmatrix && size == 1 && cll == MAT_REUSE_MATRIX) {
8543: PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscoltmp, MAT_REUSE_MATRIX, &newmat));
8544: goto setproperties;
8545: } else if (mat->ops->createsubmatrices && !mat->ops->createsubmatrix && size == 1) {
8546: PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscoltmp, MAT_INITIAL_MATRIX, &local));
8547: *newmat = *local;
8548: PetscCall(PetscFree(local));
8549: goto setproperties;
8550: } else if (!mat->ops->createsubmatrix) {
8551: /* Create a new matrix type that implements the operation using the full matrix */
8552: PetscCall(PetscLogEventBegin(MAT_CreateSubMat, mat, 0, 0, 0));
8553: switch (cll) {
8554: case MAT_INITIAL_MATRIX:
8555: PetscCall(MatCreateSubMatrixVirtual(mat, isrow, iscoltmp, newmat));
8556: break;
8557: case MAT_REUSE_MATRIX:
8558: PetscCall(MatSubMatrixVirtualUpdate(*newmat, mat, isrow, iscoltmp));
8559: break;
8560: default:
8561: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Invalid MatReuse, must be either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX");
8562: }
8563: PetscCall(PetscLogEventEnd(MAT_CreateSubMat, mat, 0, 0, 0));
8564: goto setproperties;
8565: }
8567: PetscCall(PetscLogEventBegin(MAT_CreateSubMat, mat, 0, 0, 0));
8568: PetscUseTypeMethod(mat, createsubmatrix, isrow, iscoltmp, cll, newmat);
8569: PetscCall(PetscLogEventEnd(MAT_CreateSubMat, mat, 0, 0, 0));
8571: setproperties:
8572: PetscCall(ISEqualUnsorted(isrow, iscoltmp, &flg));
8573: if (flg) PetscCall(MatPropagateSymmetryOptions(mat, *newmat));
8574: if (!iscol) PetscCall(ISDestroy(&iscoltmp));
8575: if (*newmat && cll == MAT_INITIAL_MATRIX) PetscCall(PetscObjectStateIncrease((PetscObject)*newmat));
8576: PetscFunctionReturn(PETSC_SUCCESS);
8577: }
8579: /*@
8580: MatPropagateSymmetryOptions - Propagates symmetry options set on a matrix to another matrix
8582: Not Collective
8584: Input Parameters:
8585: + A - the matrix we wish to propagate options from
8586: - B - the matrix we wish to propagate options to
8588: Level: beginner
8590: Note:
8591: Propagates the options associated to `MAT_SYMMETRY_ETERNAL`, `MAT_STRUCTURALLY_SYMMETRIC`, `MAT_HERMITIAN`, `MAT_SPD`, `MAT_SYMMETRIC`, and `MAT_STRUCTURAL_SYMMETRY_ETERNAL`
8593: .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MatIsSymmetricKnown()`, `MatIsSPDKnown()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetricKnown()`
8594: @*/
8595: PetscErrorCode MatPropagateSymmetryOptions(Mat A, Mat B)
8596: {
8597: PetscFunctionBegin;
8600: B->symmetry_eternal = A->symmetry_eternal;
8601: B->structural_symmetry_eternal = A->structural_symmetry_eternal;
8602: B->symmetric = A->symmetric;
8603: B->structurally_symmetric = A->structurally_symmetric;
8604: B->spd = A->spd;
8605: B->hermitian = A->hermitian;
8606: PetscFunctionReturn(PETSC_SUCCESS);
8607: }
8609: /*@
8610: MatStashSetInitialSize - sets the sizes of the matrix stash, that is
8611: used during the assembly process to store values that belong to
8612: other processors.
8614: Not Collective
8616: Input Parameters:
8617: + mat - the matrix
8618: . size - the initial size of the stash.
8619: - bsize - the initial size of the block-stash(if used).
8621: Options Database Keys:
8622: + -matstash_initial_size <size> or <size0,size1,...sizep-1> - set initial size
8623: - -matstash_block_initial_size <bsize> or <bsize0,bsize1,...bsizep-1> - set initial block size
8625: Level: intermediate
8627: Notes:
8628: The block-stash is used for values set with `MatSetValuesBlocked()` while
8629: the stash is used for values set with `MatSetValues()`
8631: Run with the option -info and look for output of the form
8632: MatAssemblyBegin_MPIXXX:Stash has MM entries, uses nn mallocs.
8633: to determine the appropriate value, MM, to use for size and
8634: MatAssemblyBegin_MPIXXX:Block-Stash has BMM entries, uses nn mallocs.
8635: to determine the value, BMM to use for bsize
8637: .seealso: [](ch_matrices), `MatAssemblyBegin()`, `MatAssemblyEnd()`, `Mat`, `MatStashGetInfo()`
8638: @*/
8639: PetscErrorCode MatStashSetInitialSize(Mat mat, PetscInt size, PetscInt bsize)
8640: {
8641: PetscFunctionBegin;
8644: PetscCall(MatStashSetInitialSize_Private(&mat->stash, size));
8645: PetscCall(MatStashSetInitialSize_Private(&mat->bstash, bsize));
8646: PetscFunctionReturn(PETSC_SUCCESS);
8647: }
8649: /*@
8650: MatInterpolateAdd - $w = y + A*x$ or $A^T*x$ depending on the shape of
8651: the matrix
8653: Neighbor-wise Collective
8655: Input Parameters:
8656: + A - the matrix
8657: . x - the vector to be multiplied by the interpolation operator
8658: - y - the vector to be added to the result
8660: Output Parameter:
8661: . w - the resulting vector
8663: Level: intermediate
8665: Notes:
8666: `w` may be the same vector as `y`.
8668: This allows one to use either the restriction or interpolation (its transpose)
8669: matrix to do the interpolation
8671: .seealso: [](ch_matrices), `Mat`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatRestrict()`, `PCMG`
8672: @*/
8673: PetscErrorCode MatInterpolateAdd(Mat A, Vec x, Vec y, Vec w)
8674: {
8675: PetscInt M, N, Ny;
8677: PetscFunctionBegin;
8682: PetscCall(MatGetSize(A, &M, &N));
8683: PetscCall(VecGetSize(y, &Ny));
8684: if (M == Ny) {
8685: PetscCall(MatMultAdd(A, x, y, w));
8686: } else {
8687: PetscCall(MatMultTransposeAdd(A, x, y, w));
8688: }
8689: PetscFunctionReturn(PETSC_SUCCESS);
8690: }
8692: /*@
8693: MatInterpolate - $y = A*x$ or $A^T*x$ depending on the shape of
8694: the matrix
8696: Neighbor-wise Collective
8698: Input Parameters:
8699: + A - the matrix
8700: - x - the vector to be interpolated
8702: Output Parameter:
8703: . y - the resulting vector
8705: Level: intermediate
8707: Note:
8708: This allows one to use either the restriction or interpolation (its transpose)
8709: matrix to do the interpolation
8711: .seealso: [](ch_matrices), `Mat`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatRestrict()`, `PCMG`
8712: @*/
8713: PetscErrorCode MatInterpolate(Mat A, Vec x, Vec y)
8714: {
8715: PetscInt M, N, Ny;
8717: PetscFunctionBegin;
8721: PetscCall(MatGetSize(A, &M, &N));
8722: PetscCall(VecGetSize(y, &Ny));
8723: if (M == Ny) {
8724: PetscCall(MatMult(A, x, y));
8725: } else {
8726: PetscCall(MatMultTranspose(A, x, y));
8727: }
8728: PetscFunctionReturn(PETSC_SUCCESS);
8729: }
8731: /*@
8732: MatRestrict - $y = A*x$ or $A^T*x$
8734: Neighbor-wise Collective
8736: Input Parameters:
8737: + A - the matrix
8738: - x - the vector to be restricted
8740: Output Parameter:
8741: . y - the resulting vector
8743: Level: intermediate
8745: Note:
8746: This allows one to use either the restriction or interpolation (its transpose)
8747: matrix to do the restriction
8749: .seealso: [](ch_matrices), `Mat`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatInterpolate()`, `PCMG`
8750: @*/
8751: PetscErrorCode MatRestrict(Mat A, Vec x, Vec y)
8752: {
8753: PetscInt M, N, Nx;
8755: PetscFunctionBegin;
8759: PetscCall(MatGetSize(A, &M, &N));
8760: PetscCall(VecGetSize(x, &Nx));
8761: if (M == Nx) {
8762: PetscCall(MatMultTranspose(A, x, y));
8763: } else {
8764: PetscCall(MatMult(A, x, y));
8765: }
8766: PetscFunctionReturn(PETSC_SUCCESS);
8767: }
8769: /*@
8770: MatMatInterpolateAdd - $Y = W + A*X$ or $W + A^T*X$ depending on the shape of `A`
8772: Neighbor-wise Collective
8774: Input Parameters:
8775: + A - the matrix
8776: . x - the input dense matrix to be multiplied
8777: - w - the input dense matrix to be added to the result
8779: Output Parameter:
8780: . y - the output dense matrix
8782: Level: intermediate
8784: Note:
8785: This allows one to use either the restriction or interpolation (its transpose)
8786: matrix to do the interpolation. `y` matrix can be reused if already created with the proper sizes,
8787: otherwise it will be recreated. `y` must be initialized to `NULL` if not supplied.
8789: .seealso: [](ch_matrices), `Mat`, `MatInterpolateAdd()`, `MatMatInterpolate()`, `MatMatRestrict()`, `PCMG`
8790: @*/
8791: PetscErrorCode MatMatInterpolateAdd(Mat A, Mat x, Mat w, Mat *y)
8792: {
8793: PetscInt M, N, Mx, Nx, Mo, My = 0, Ny = 0;
8794: PetscBool trans = PETSC_TRUE;
8795: MatReuse reuse = MAT_INITIAL_MATRIX;
8797: PetscFunctionBegin;
8803: PetscCall(MatGetSize(A, &M, &N));
8804: PetscCall(MatGetSize(x, &Mx, &Nx));
8805: if (N == Mx) trans = PETSC_FALSE;
8806: 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);
8807: Mo = trans ? N : M;
8808: if (*y) {
8809: PetscCall(MatGetSize(*y, &My, &Ny));
8810: if (Mo == My && Nx == Ny) {
8811: reuse = MAT_REUSE_MATRIX;
8812: } else {
8813: 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);
8814: PetscCall(MatDestroy(y));
8815: }
8816: }
8818: if (w && *y == w) { /* this is to minimize changes in PCMG */
8819: PetscBool flg;
8821: PetscCall(PetscObjectQuery((PetscObject)*y, "__MatMatIntAdd_w", (PetscObject *)&w));
8822: if (w) {
8823: PetscInt My, Ny, Mw, Nw;
8825: PetscCall(PetscObjectTypeCompare((PetscObject)*y, ((PetscObject)w)->type_name, &flg));
8826: PetscCall(MatGetSize(*y, &My, &Ny));
8827: PetscCall(MatGetSize(w, &Mw, &Nw));
8828: if (!flg || My != Mw || Ny != Nw) w = NULL;
8829: }
8830: if (!w) {
8831: PetscCall(MatDuplicate(*y, MAT_COPY_VALUES, &w));
8832: PetscCall(PetscObjectCompose((PetscObject)*y, "__MatMatIntAdd_w", (PetscObject)w));
8833: PetscCall(PetscObjectDereference((PetscObject)w));
8834: } else {
8835: PetscCall(MatCopy(*y, w, UNKNOWN_NONZERO_PATTERN));
8836: }
8837: }
8838: if (!trans) {
8839: PetscCall(MatMatMult(A, x, reuse, PETSC_DEFAULT, y));
8840: } else {
8841: PetscCall(MatTransposeMatMult(A, x, reuse, PETSC_DEFAULT, y));
8842: }
8843: if (w) PetscCall(MatAXPY(*y, 1.0, w, UNKNOWN_NONZERO_PATTERN));
8844: PetscFunctionReturn(PETSC_SUCCESS);
8845: }
8847: /*@
8848: MatMatInterpolate - $Y = A*X$ or $A^T*X$ depending on the shape of `A`
8850: Neighbor-wise Collective
8852: Input Parameters:
8853: + A - the matrix
8854: - x - the input dense matrix
8856: Output Parameter:
8857: . y - the output dense matrix
8859: Level: intermediate
8861: Note:
8862: This allows one to use either the restriction or interpolation (its transpose)
8863: matrix to do the interpolation. `y` matrix can be reused if already created with the proper sizes,
8864: otherwise it will be recreated. `y` must be initialized to `NULL` if not supplied.
8866: .seealso: [](ch_matrices), `Mat`, `MatInterpolate()`, `MatRestrict()`, `MatMatRestrict()`, `PCMG`
8867: @*/
8868: PetscErrorCode MatMatInterpolate(Mat A, Mat x, Mat *y)
8869: {
8870: PetscFunctionBegin;
8871: PetscCall(MatMatInterpolateAdd(A, x, NULL, y));
8872: PetscFunctionReturn(PETSC_SUCCESS);
8873: }
8875: /*@
8876: MatMatRestrict - $Y = A*X$ or $A^T*X$ depending on the shape of `A`
8878: Neighbor-wise Collective
8880: Input Parameters:
8881: + A - the matrix
8882: - x - the input dense matrix
8884: Output Parameter:
8885: . y - the output dense matrix
8887: Level: intermediate
8889: Note:
8890: This allows one to use either the restriction or interpolation (its transpose)
8891: matrix to do the restriction. `y` matrix can be reused if already created with the proper sizes,
8892: otherwise it will be recreated. `y` must be initialized to `NULL` if not supplied.
8894: .seealso: [](ch_matrices), `Mat`, `MatRestrict()`, `MatInterpolate()`, `MatMatInterpolate()`, `PCMG`
8895: @*/
8896: PetscErrorCode MatMatRestrict(Mat A, Mat x, Mat *y)
8897: {
8898: PetscFunctionBegin;
8899: PetscCall(MatMatInterpolateAdd(A, x, NULL, y));
8900: PetscFunctionReturn(PETSC_SUCCESS);
8901: }
8903: /*@
8904: MatGetNullSpace - retrieves the null space of a matrix.
8906: Logically Collective
8908: Input Parameters:
8909: + mat - the matrix
8910: - nullsp - the null space object
8912: Level: developer
8914: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatSetNullSpace()`, `MatNullSpace`
8915: @*/
8916: PetscErrorCode MatGetNullSpace(Mat mat, MatNullSpace *nullsp)
8917: {
8918: PetscFunctionBegin;
8920: PetscAssertPointer(nullsp, 2);
8921: *nullsp = (mat->symmetric == PETSC_BOOL3_TRUE && !mat->nullsp) ? mat->transnullsp : mat->nullsp;
8922: PetscFunctionReturn(PETSC_SUCCESS);
8923: }
8925: /*@C
8926: MatGetNullSpaces - gets the null spaces, transpose null spaces, and near null spaces from an array of matrices
8928: Logically Collective
8930: Input Parameters:
8931: + n - the number of matrices
8932: - mat - the array of matrices
8934: Output Parameters:
8935: . nullsp - an array of null spaces, `NULL` for each matrix that does not have a null space
8937: Level: developer
8939: Note:
8940: Call `MatRestoreNullspaces()` to provide these to another array of matrices
8942: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatSetTransposeNullSpace()`, `MatGetTransposeNullSpace()`,
8943: `MatNullSpaceRemove()`, `MatRestoreNullSpaces()`
8944: @*/
8945: PetscErrorCode MatGetNullSpaces(PetscInt n, Mat mat[], MatNullSpace *nullsp[])
8946: {
8947: PetscFunctionBegin;
8948: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Number of matrices %" PetscInt_FMT " must be non-negative", n);
8949: PetscAssertPointer(mat, 2);
8950: PetscAssertPointer(nullsp, 3);
8952: PetscCall(PetscCalloc1(3 * n, nullsp));
8953: for (PetscInt i = 0; i < n; i++) {
8955: (*nullsp)[i] = mat[i]->nullsp;
8956: PetscCall(PetscObjectReference((PetscObject)(*nullsp)[i]));
8957: (*nullsp)[n + i] = mat[i]->nearnullsp;
8958: PetscCall(PetscObjectReference((PetscObject)(*nullsp)[n + i]));
8959: (*nullsp)[2 * n + i] = mat[i]->transnullsp;
8960: PetscCall(PetscObjectReference((PetscObject)(*nullsp)[2 * n + i]));
8961: }
8962: PetscFunctionReturn(PETSC_SUCCESS);
8963: }
8965: /*@C
8966: MatRestoreNullSpaces - sets the null spaces, transpose null spaces, and near null spaces obtained with `MatGetNullSpaces()` for an array of matrices
8968: Logically Collective
8970: Input Parameters:
8971: + n - the number of matrices
8972: . mat - the array of matrices
8973: - nullsp - an array of null spaces, `NULL` if the null space does not exist
8975: Level: developer
8977: Note:
8978: Call `MatGetNullSpaces()` to create `nullsp`
8980: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatSetTransposeNullSpace()`, `MatGetTransposeNullSpace()`,
8981: `MatNullSpaceRemove()`, `MatGetNullSpaces()`
8982: @*/
8983: PetscErrorCode MatRestoreNullSpaces(PetscInt n, Mat mat[], MatNullSpace *nullsp[])
8984: {
8985: PetscFunctionBegin;
8986: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Number of matrices %" PetscInt_FMT " must be non-negative", n);
8987: PetscAssertPointer(mat, 2);
8988: PetscAssertPointer(nullsp, 3);
8989: PetscAssertPointer(*nullsp, 3);
8991: for (PetscInt i = 0; i < n; i++) {
8993: PetscCall(MatSetNullSpace(mat[i], (*nullsp)[i]));
8994: PetscCall(PetscObjectDereference((PetscObject)(*nullsp)[i]));
8995: PetscCall(MatSetNearNullSpace(mat[i], (*nullsp)[n + i]));
8996: PetscCall(PetscObjectDereference((PetscObject)(*nullsp)[n + i]));
8997: PetscCall(MatSetTransposeNullSpace(mat[i], (*nullsp)[2 * n + i]));
8998: PetscCall(PetscObjectDereference((PetscObject)(*nullsp)[2 * n + i]));
8999: }
9000: PetscCall(PetscFree(*nullsp));
9001: PetscFunctionReturn(PETSC_SUCCESS);
9002: }
9004: /*@
9005: MatSetNullSpace - attaches a null space to a matrix.
9007: Logically Collective
9009: Input Parameters:
9010: + mat - the matrix
9011: - nullsp - the null space object
9013: Level: advanced
9015: Notes:
9016: This null space is used by the `KSP` linear solvers to solve singular systems.
9018: 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`
9020: 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
9021: to zero but the linear system will still be solved in a least squares sense.
9023: The fundamental theorem of linear algebra (Gilbert Strang, Introduction to Applied Mathematics, page 72) states that
9024: 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 $A^T$, $R(A^T)$.
9025: 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
9026: 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
9027: 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($A^T$).
9028: This \hat{b} can be obtained by calling `MatNullSpaceRemove()` with the null space of the transpose of the matrix.
9030: If the matrix is known to be symmetric because it is an `MATSBAIJ` matrix or one as called
9031: `MatSetOption`(mat,`MAT_SYMMETRIC` or possibly `MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`); this
9032: routine also automatically calls `MatSetTransposeNullSpace()`.
9034: The user should call `MatNullSpaceDestroy()`.
9036: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatSetTransposeNullSpace()`, `MatGetTransposeNullSpace()`, `MatNullSpaceRemove()`,
9037: `KSPSetPCSide()`
9038: @*/
9039: PetscErrorCode MatSetNullSpace(Mat mat, MatNullSpace nullsp)
9040: {
9041: PetscFunctionBegin;
9044: if (nullsp) PetscCall(PetscObjectReference((PetscObject)nullsp));
9045: PetscCall(MatNullSpaceDestroy(&mat->nullsp));
9046: mat->nullsp = nullsp;
9047: if (mat->symmetric == PETSC_BOOL3_TRUE) PetscCall(MatSetTransposeNullSpace(mat, nullsp));
9048: PetscFunctionReturn(PETSC_SUCCESS);
9049: }
9051: /*@
9052: MatGetTransposeNullSpace - retrieves the null space of the transpose of a matrix.
9054: Logically Collective
9056: Input Parameters:
9057: + mat - the matrix
9058: - nullsp - the null space object
9060: Level: developer
9062: .seealso: [](ch_matrices), `Mat`, `MatNullSpace`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatSetTransposeNullSpace()`, `MatSetNullSpace()`, `MatGetNullSpace()`
9063: @*/
9064: PetscErrorCode MatGetTransposeNullSpace(Mat mat, MatNullSpace *nullsp)
9065: {
9066: PetscFunctionBegin;
9069: PetscAssertPointer(nullsp, 2);
9070: *nullsp = (mat->symmetric == PETSC_BOOL3_TRUE && !mat->transnullsp) ? mat->nullsp : mat->transnullsp;
9071: PetscFunctionReturn(PETSC_SUCCESS);
9072: }
9074: /*@
9075: MatSetTransposeNullSpace - attaches the null space of a transpose of a matrix to the matrix
9077: Logically Collective
9079: Input Parameters:
9080: + mat - the matrix
9081: - nullsp - the null space object
9083: Level: advanced
9085: Notes:
9086: This allows solving singular linear systems defined by the transpose of the matrix using `KSP` solvers with left preconditioning.
9088: See `MatSetNullSpace()`
9090: .seealso: [](ch_matrices), `Mat`, `MatNullSpace`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatSetNullSpace()`, `MatGetTransposeNullSpace()`, `MatNullSpaceRemove()`, `KSPSetPCSide()`
9091: @*/
9092: PetscErrorCode MatSetTransposeNullSpace(Mat mat, MatNullSpace nullsp)
9093: {
9094: PetscFunctionBegin;
9097: if (nullsp) PetscCall(PetscObjectReference((PetscObject)nullsp));
9098: PetscCall(MatNullSpaceDestroy(&mat->transnullsp));
9099: mat->transnullsp = nullsp;
9100: PetscFunctionReturn(PETSC_SUCCESS);
9101: }
9103: /*@
9104: MatSetNearNullSpace - attaches a null space to a matrix, which is often the null space (rigid body modes) of the operator without boundary conditions
9105: This null space will be used to provide near null space vectors to a multigrid preconditioner built from this matrix.
9107: Logically Collective
9109: Input Parameters:
9110: + mat - the matrix
9111: - nullsp - the null space object
9113: Level: advanced
9115: Notes:
9116: Overwrites any previous near null space that may have been attached
9118: You can remove the null space by calling this routine with an `nullsp` of `NULL`
9120: .seealso: [](ch_matrices), `Mat`, `MatNullSpace`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNullSpace()`, `MatNullSpaceCreateRigidBody()`, `MatGetNearNullSpace()`
9121: @*/
9122: PetscErrorCode MatSetNearNullSpace(Mat mat, MatNullSpace nullsp)
9123: {
9124: PetscFunctionBegin;
9128: MatCheckPreallocated(mat, 1);
9129: if (nullsp) PetscCall(PetscObjectReference((PetscObject)nullsp));
9130: PetscCall(MatNullSpaceDestroy(&mat->nearnullsp));
9131: mat->nearnullsp = nullsp;
9132: PetscFunctionReturn(PETSC_SUCCESS);
9133: }
9135: /*@
9136: MatGetNearNullSpace - Get null space attached with `MatSetNearNullSpace()`
9138: Not Collective
9140: Input Parameter:
9141: . mat - the matrix
9143: Output Parameter:
9144: . nullsp - the null space object, `NULL` if not set
9146: Level: advanced
9148: .seealso: [](ch_matrices), `Mat`, `MatNullSpace`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatNullSpaceCreate()`
9149: @*/
9150: PetscErrorCode MatGetNearNullSpace(Mat mat, MatNullSpace *nullsp)
9151: {
9152: PetscFunctionBegin;
9155: PetscAssertPointer(nullsp, 2);
9156: MatCheckPreallocated(mat, 1);
9157: *nullsp = mat->nearnullsp;
9158: PetscFunctionReturn(PETSC_SUCCESS);
9159: }
9161: /*@C
9162: MatICCFactor - Performs in-place incomplete Cholesky factorization of matrix.
9164: Collective
9166: Input Parameters:
9167: + mat - the matrix
9168: . row - row/column permutation
9169: - info - information on desired factorization process
9171: Level: developer
9173: Notes:
9174: Probably really in-place only when level of fill is zero, otherwise allocates
9175: new space to store factored matrix and deletes previous memory.
9177: Most users should employ the `KSP` interface for linear solvers
9178: instead of working directly with matrix algebra routines such as this.
9179: See, e.g., `KSPCreate()`.
9181: Developer Note:
9182: The Fortran interface is not autogenerated as the
9183: interface definition cannot be generated correctly [due to `MatFactorInfo`]
9185: .seealso: [](ch_matrices), `Mat`, `MatFactorInfo`, `MatGetFactor()`, `MatICCFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`
9186: @*/
9187: PetscErrorCode MatICCFactor(Mat mat, IS row, const MatFactorInfo *info)
9188: {
9189: PetscFunctionBegin;
9193: PetscAssertPointer(info, 3);
9194: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "matrix must be square");
9195: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
9196: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
9197: MatCheckPreallocated(mat, 1);
9198: PetscUseTypeMethod(mat, iccfactor, row, info);
9199: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
9200: PetscFunctionReturn(PETSC_SUCCESS);
9201: }
9203: /*@
9204: MatDiagonalScaleLocal - Scales columns of a matrix given the scaling values including the
9205: ghosted ones.
9207: Not Collective
9209: Input Parameters:
9210: + mat - the matrix
9211: - diag - the diagonal values, including ghost ones
9213: Level: developer
9215: Notes:
9216: Works only for `MATMPIAIJ` and `MATMPIBAIJ` matrices
9218: This allows one to avoid during communication to perform the scaling that must be done with `MatDiagonalScale()`
9220: .seealso: [](ch_matrices), `Mat`, `MatDiagonalScale()`
9221: @*/
9222: PetscErrorCode MatDiagonalScaleLocal(Mat mat, Vec diag)
9223: {
9224: PetscMPIInt size;
9226: PetscFunctionBegin;
9231: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Matrix must be already assembled");
9232: PetscCall(PetscLogEventBegin(MAT_Scale, mat, 0, 0, 0));
9233: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
9234: if (size == 1) {
9235: PetscInt n, m;
9236: PetscCall(VecGetSize(diag, &n));
9237: PetscCall(MatGetSize(mat, NULL, &m));
9238: PetscCheck(m == n, PETSC_COMM_SELF, PETSC_ERR_SUP, "Only supported for sequential matrices when no ghost points/periodic conditions");
9239: PetscCall(MatDiagonalScale(mat, NULL, diag));
9240: } else {
9241: PetscUseMethod(mat, "MatDiagonalScaleLocal_C", (Mat, Vec), (mat, diag));
9242: }
9243: PetscCall(PetscLogEventEnd(MAT_Scale, mat, 0, 0, 0));
9244: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
9245: PetscFunctionReturn(PETSC_SUCCESS);
9246: }
9248: /*@
9249: MatGetInertia - Gets the inertia from a factored matrix
9251: Collective
9253: Input Parameter:
9254: . mat - the matrix
9256: Output Parameters:
9257: + nneg - number of negative eigenvalues
9258: . nzero - number of zero eigenvalues
9259: - npos - number of positive eigenvalues
9261: Level: advanced
9263: Note:
9264: Matrix must have been factored by `MatCholeskyFactor()`
9266: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatCholeskyFactor()`
9267: @*/
9268: PetscErrorCode MatGetInertia(Mat mat, PetscInt *nneg, PetscInt *nzero, PetscInt *npos)
9269: {
9270: PetscFunctionBegin;
9273: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
9274: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Numeric factor mat is not assembled");
9275: PetscUseTypeMethod(mat, getinertia, nneg, nzero, npos);
9276: PetscFunctionReturn(PETSC_SUCCESS);
9277: }
9279: /*@C
9280: MatSolves - Solves $A x = b$, given a factored matrix, for a collection of vectors
9282: Neighbor-wise Collective
9284: Input Parameters:
9285: + mat - the factored matrix obtained with `MatGetFactor()`
9286: - b - the right-hand-side vectors
9288: Output Parameter:
9289: . x - the result vectors
9291: Level: developer
9293: Note:
9294: The vectors `b` and `x` cannot be the same. I.e., one cannot
9295: call `MatSolves`(A,x,x).
9297: .seealso: [](ch_matrices), `Mat`, `Vecs`, `MatSolveAdd()`, `MatSolveTranspose()`, `MatSolveTransposeAdd()`, `MatSolve()`
9298: @*/
9299: PetscErrorCode MatSolves(Mat mat, Vecs b, Vecs x)
9300: {
9301: PetscFunctionBegin;
9304: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
9305: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
9306: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
9308: MatCheckPreallocated(mat, 1);
9309: PetscCall(PetscLogEventBegin(MAT_Solves, mat, 0, 0, 0));
9310: PetscUseTypeMethod(mat, solves, b, x);
9311: PetscCall(PetscLogEventEnd(MAT_Solves, mat, 0, 0, 0));
9312: PetscFunctionReturn(PETSC_SUCCESS);
9313: }
9315: /*@
9316: MatIsSymmetric - Test whether a matrix is symmetric
9318: Collective
9320: Input Parameters:
9321: + A - the matrix to test
9322: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact transpose)
9324: Output Parameter:
9325: . flg - the result
9327: Level: intermediate
9329: Notes:
9330: For real numbers `MatIsSymmetric()` and `MatIsHermitian()` return identical results
9332: If the matrix does not yet know if it is symmetric or not this can be an expensive operation, also available `MatIsSymmetricKnown()`
9334: One can declare that a matrix is symmetric with `MatSetOption`(mat,`MAT_SYMMETRIC`,`PETSC_TRUE`) and if it is known to remain symmetric
9335: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9337: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetricKnown()`,
9338: `MAT_SYMMETRIC`, `MAT_SYMMETRY_ETERNAL`
9339: @*/
9340: PetscErrorCode MatIsSymmetric(Mat A, PetscReal tol, PetscBool *flg)
9341: {
9342: PetscFunctionBegin;
9344: PetscAssertPointer(flg, 3);
9345: if (A->symmetric != PETSC_BOOL3_UNKNOWN) *flg = PetscBool3ToBool(A->symmetric);
9346: else {
9347: if (A->ops->issymmetric) PetscUseTypeMethod(A, issymmetric, tol, flg);
9348: else PetscCall(MatIsTranspose(A, A, tol, flg));
9349: if (!tol) PetscCall(MatSetOption(A, MAT_SYMMETRIC, *flg));
9350: }
9351: PetscFunctionReturn(PETSC_SUCCESS);
9352: }
9354: /*@
9355: MatIsHermitian - Test whether a matrix is Hermitian
9357: Collective
9359: Input Parameters:
9360: + A - the matrix to test
9361: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact Hermitian)
9363: Output Parameter:
9364: . flg - the result
9366: Level: intermediate
9368: Notes:
9369: For real numbers `MatIsSymmetric()` and `MatIsHermitian()` return identical results
9371: If the matrix does not yet know if it is Hermitian or not this can be an expensive operation, also available `MatIsHermitianKnown()`
9373: One can declare that a matrix is Hermitian with `MatSetOption`(mat,`MAT_HERMITIAN`,`PETSC_TRUE`) and if it is known to remain Hermitian
9374: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMEMTRY_ETERNAL`,`PETSC_TRUE`)
9376: .seealso: [](ch_matrices), `Mat`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetric()`, `MatSetOption()`,
9377: `MatIsSymmetricKnown()`, `MatIsSymmetric()`, `MAT_HERMITIAN`, `MAT_SYMMETRY_ETERNAL`
9378: @*/
9379: PetscErrorCode MatIsHermitian(Mat A, PetscReal tol, PetscBool *flg)
9380: {
9381: PetscFunctionBegin;
9383: PetscAssertPointer(flg, 3);
9384: if (A->hermitian != PETSC_BOOL3_UNKNOWN) *flg = PetscBool3ToBool(A->hermitian);
9385: else {
9386: if (A->ops->ishermitian) PetscUseTypeMethod(A, ishermitian, tol, flg);
9387: else PetscCall(MatIsHermitianTranspose(A, A, tol, flg));
9388: if (!tol) PetscCall(MatSetOption(A, MAT_HERMITIAN, *flg));
9389: }
9390: PetscFunctionReturn(PETSC_SUCCESS);
9391: }
9393: /*@
9394: MatIsSymmetricKnown - Checks if a matrix knows if it is symmetric or not and its symmetric state
9396: Not Collective
9398: Input Parameter:
9399: . A - the matrix to check
9401: Output Parameters:
9402: + set - `PETSC_TRUE` if the matrix knows its symmetry state (this tells you if the next flag is valid)
9403: - flg - the result (only valid if set is `PETSC_TRUE`)
9405: Level: advanced
9407: Notes:
9408: Does not check the matrix values directly, so this may return unknown (set = `PETSC_FALSE`). Use `MatIsSymmetric()`
9409: if you want it explicitly checked
9411: One can declare that a matrix is symmetric with `MatSetOption`(mat,`MAT_SYMMETRIC`,`PETSC_TRUE`) and if it is known to remain symmetric
9412: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9414: .seealso: [](ch_matrices), `Mat`, `MAT_SYMMETRY_ETERNAL`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitianKnown()`
9415: @*/
9416: PetscErrorCode MatIsSymmetricKnown(Mat A, PetscBool *set, PetscBool *flg)
9417: {
9418: PetscFunctionBegin;
9420: PetscAssertPointer(set, 2);
9421: PetscAssertPointer(flg, 3);
9422: if (A->symmetric != PETSC_BOOL3_UNKNOWN) {
9423: *set = PETSC_TRUE;
9424: *flg = PetscBool3ToBool(A->symmetric);
9425: } else {
9426: *set = PETSC_FALSE;
9427: }
9428: PetscFunctionReturn(PETSC_SUCCESS);
9429: }
9431: /*@
9432: MatIsSPDKnown - Checks if a matrix knows if it is symmetric positive definite or not and its symmetric positive definite state
9434: Not Collective
9436: Input Parameter:
9437: . A - the matrix to check
9439: Output Parameters:
9440: + set - `PETSC_TRUE` if the matrix knows its symmetric positive definite state (this tells you if the next flag is valid)
9441: - flg - the result (only valid if set is `PETSC_TRUE`)
9443: Level: advanced
9445: Notes:
9446: Does not check the matrix values directly, so this may return unknown (set = `PETSC_FALSE`).
9448: One can declare that a matrix is SPD with `MatSetOption`(mat,`MAT_SPD`,`PETSC_TRUE`) and if it is known to remain SPD
9449: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SPD_ETERNAL`,`PETSC_TRUE`)
9451: .seealso: [](ch_matrices), `Mat`, `MAT_SPD_ETERNAL`, `MAT_SPD`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitianKnown()`
9452: @*/
9453: PetscErrorCode MatIsSPDKnown(Mat A, PetscBool *set, PetscBool *flg)
9454: {
9455: PetscFunctionBegin;
9457: PetscAssertPointer(set, 2);
9458: PetscAssertPointer(flg, 3);
9459: if (A->spd != PETSC_BOOL3_UNKNOWN) {
9460: *set = PETSC_TRUE;
9461: *flg = PetscBool3ToBool(A->spd);
9462: } else {
9463: *set = PETSC_FALSE;
9464: }
9465: PetscFunctionReturn(PETSC_SUCCESS);
9466: }
9468: /*@
9469: MatIsHermitianKnown - Checks if a matrix knows if it is Hermitian or not and its Hermitian state
9471: Not Collective
9473: Input Parameter:
9474: . A - the matrix to check
9476: Output Parameters:
9477: + set - `PETSC_TRUE` if the matrix knows its Hermitian state (this tells you if the next flag is valid)
9478: - flg - the result (only valid if set is `PETSC_TRUE`)
9480: Level: advanced
9482: Notes:
9483: Does not check the matrix values directly, so this may return unknown (set = `PETSC_FALSE`). Use `MatIsHermitian()`
9484: if you want it explicitly checked
9486: One can declare that a matrix is Hermitian with `MatSetOption`(mat,`MAT_HERMITIAN`,`PETSC_TRUE`) and if it is known to remain Hermitian
9487: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9489: .seealso: [](ch_matrices), `Mat`, `MAT_SYMMETRY_ETERNAL`, `MAT_HERMITIAN`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`
9490: @*/
9491: PetscErrorCode MatIsHermitianKnown(Mat A, PetscBool *set, PetscBool *flg)
9492: {
9493: PetscFunctionBegin;
9495: PetscAssertPointer(set, 2);
9496: PetscAssertPointer(flg, 3);
9497: if (A->hermitian != PETSC_BOOL3_UNKNOWN) {
9498: *set = PETSC_TRUE;
9499: *flg = PetscBool3ToBool(A->hermitian);
9500: } else {
9501: *set = PETSC_FALSE;
9502: }
9503: PetscFunctionReturn(PETSC_SUCCESS);
9504: }
9506: /*@
9507: MatIsStructurallySymmetric - Test whether a matrix is structurally symmetric
9509: Collective
9511: Input Parameter:
9512: . A - the matrix to test
9514: Output Parameter:
9515: . flg - the result
9517: Level: intermediate
9519: Notes:
9520: If the matrix does yet know it is structurally symmetric this can be an expensive operation, also available `MatIsStructurallySymmetricKnown()`
9522: 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
9523: symmetric after changes to the matrices values one can call `MatSetOption`(mat,`MAT_STRUCTURAL_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9525: .seealso: [](ch_matrices), `Mat`, `MAT_STRUCTURALLY_SYMMETRIC`, `MAT_STRUCTURAL_SYMMETRY_ETERNAL`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsSymmetric()`, `MatSetOption()`, `MatIsStructurallySymmetricKnown()`
9526: @*/
9527: PetscErrorCode MatIsStructurallySymmetric(Mat A, PetscBool *flg)
9528: {
9529: PetscFunctionBegin;
9531: PetscAssertPointer(flg, 2);
9532: if (A->structurally_symmetric != PETSC_BOOL3_UNKNOWN) {
9533: *flg = PetscBool3ToBool(A->structurally_symmetric);
9534: } else {
9535: PetscUseTypeMethod(A, isstructurallysymmetric, flg);
9536: PetscCall(MatSetOption(A, MAT_STRUCTURALLY_SYMMETRIC, *flg));
9537: }
9538: PetscFunctionReturn(PETSC_SUCCESS);
9539: }
9541: /*@
9542: MatIsStructurallySymmetricKnown - Checks if a matrix knows if it is structurally symmetric or not and its structurally symmetric state
9544: Not Collective
9546: Input Parameter:
9547: . A - the matrix to check
9549: Output Parameters:
9550: + set - PETSC_TRUE if the matrix knows its structurally symmetric state (this tells you if the next flag is valid)
9551: - flg - the result (only valid if set is PETSC_TRUE)
9553: Level: advanced
9555: Notes:
9556: 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
9557: symmetric after changes to the matrices values one can call `MatSetOption`(mat,`MAT_STRUCTURAL_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9559: Use `MatIsStructurallySymmetric()` to explicitly check if a matrix is structurally symmetric (this is an expensive operation)
9561: .seealso: [](ch_matrices), `Mat`, `MAT_STRUCTURALLY_SYMMETRIC`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitianKnown()`
9562: @*/
9563: PetscErrorCode MatIsStructurallySymmetricKnown(Mat A, PetscBool *set, PetscBool *flg)
9564: {
9565: PetscFunctionBegin;
9567: PetscAssertPointer(set, 2);
9568: PetscAssertPointer(flg, 3);
9569: if (A->structurally_symmetric != PETSC_BOOL3_UNKNOWN) {
9570: *set = PETSC_TRUE;
9571: *flg = PetscBool3ToBool(A->structurally_symmetric);
9572: } else {
9573: *set = PETSC_FALSE;
9574: }
9575: PetscFunctionReturn(PETSC_SUCCESS);
9576: }
9578: /*@
9579: MatStashGetInfo - Gets how many values are currently in the matrix stash, i.e. need
9580: to be communicated to other processors during the `MatAssemblyBegin()`/`MatAssemblyEnd()` process
9582: Not Collective
9584: Input Parameter:
9585: . mat - the matrix
9587: Output Parameters:
9588: + nstash - the size of the stash
9589: . reallocs - the number of additional mallocs incurred.
9590: . bnstash - the size of the block stash
9591: - breallocs - the number of additional mallocs incurred.in the block stash
9593: Level: advanced
9595: .seealso: [](ch_matrices), `MatAssemblyBegin()`, `MatAssemblyEnd()`, `Mat`, `MatStashSetInitialSize()`
9596: @*/
9597: PetscErrorCode MatStashGetInfo(Mat mat, PetscInt *nstash, PetscInt *reallocs, PetscInt *bnstash, PetscInt *breallocs)
9598: {
9599: PetscFunctionBegin;
9600: PetscCall(MatStashGetInfo_Private(&mat->stash, nstash, reallocs));
9601: PetscCall(MatStashGetInfo_Private(&mat->bstash, bnstash, breallocs));
9602: PetscFunctionReturn(PETSC_SUCCESS);
9603: }
9605: /*@C
9606: MatCreateVecs - Get vector(s) compatible with the matrix, i.e. with the same
9607: parallel layout, `PetscLayout` for rows and columns
9609: Collective
9611: Input Parameter:
9612: . mat - the matrix
9614: Output Parameters:
9615: + right - (optional) vector that the matrix can be multiplied against
9616: - left - (optional) vector that the matrix vector product can be stored in
9618: Level: advanced
9620: Notes:
9621: 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()`.
9623: These are new vectors which are not owned by the mat, they should be destroyed in `VecDestroy()` when no longer needed
9625: .seealso: [](ch_matrices), `Mat`, `Vec`, `VecCreate()`, `VecDestroy()`, `DMCreateGlobalVector()`
9626: @*/
9627: PetscErrorCode MatCreateVecs(Mat mat, Vec *right, Vec *left)
9628: {
9629: PetscFunctionBegin;
9632: if (mat->ops->getvecs) {
9633: PetscUseTypeMethod(mat, getvecs, right, left);
9634: } else {
9635: if (right) {
9636: PetscCheck(mat->cmap->n >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "PetscLayout for columns not yet setup");
9637: PetscCall(VecCreateWithLayout_Private(mat->cmap, right));
9638: PetscCall(VecSetType(*right, mat->defaultvectype));
9639: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
9640: if (mat->boundtocpu && mat->bindingpropagates) {
9641: PetscCall(VecSetBindingPropagates(*right, PETSC_TRUE));
9642: PetscCall(VecBindToCPU(*right, PETSC_TRUE));
9643: }
9644: #endif
9645: }
9646: if (left) {
9647: PetscCheck(mat->rmap->n >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "PetscLayout for rows not yet setup");
9648: PetscCall(VecCreateWithLayout_Private(mat->rmap, left));
9649: PetscCall(VecSetType(*left, mat->defaultvectype));
9650: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
9651: if (mat->boundtocpu && mat->bindingpropagates) {
9652: PetscCall(VecSetBindingPropagates(*left, PETSC_TRUE));
9653: PetscCall(VecBindToCPU(*left, PETSC_TRUE));
9654: }
9655: #endif
9656: }
9657: }
9658: PetscFunctionReturn(PETSC_SUCCESS);
9659: }
9661: /*@C
9662: MatFactorInfoInitialize - Initializes a `MatFactorInfo` data structure
9663: with default values.
9665: Not Collective
9667: Input Parameter:
9668: . info - the `MatFactorInfo` data structure
9670: Level: developer
9672: Notes:
9673: The solvers are generally used through the `KSP` and `PC` objects, for example
9674: `PCLU`, `PCILU`, `PCCHOLESKY`, `PCICC`
9676: Once the data structure is initialized one may change certain entries as desired for the particular factorization to be performed
9678: Developer Note:
9679: The Fortran interface is not autogenerated as the
9680: interface definition cannot be generated correctly [due to `MatFactorInfo`]
9682: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorInfo`
9683: @*/
9684: PetscErrorCode MatFactorInfoInitialize(MatFactorInfo *info)
9685: {
9686: PetscFunctionBegin;
9687: PetscCall(PetscMemzero(info, sizeof(MatFactorInfo)));
9688: PetscFunctionReturn(PETSC_SUCCESS);
9689: }
9691: /*@
9692: MatFactorSetSchurIS - Set indices corresponding to the Schur complement you wish to have computed
9694: Collective
9696: Input Parameters:
9697: + mat - the factored matrix
9698: - is - the index set defining the Schur indices (0-based)
9700: Level: advanced
9702: Notes:
9703: Call `MatFactorSolveSchurComplement()` or `MatFactorSolveSchurComplementTranspose()` after this call to solve a Schur complement system.
9705: You can call `MatFactorGetSchurComplement()` or `MatFactorCreateSchurComplement()` after this call.
9707: This functionality is only supported for `MATSOLVERMUMPS` and `MATSOLVERMKL_PARDISO`
9709: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorGetSchurComplement()`, `MatFactorRestoreSchurComplement()`, `MatFactorCreateSchurComplement()`, `MatFactorSolveSchurComplement()`,
9710: `MatFactorSolveSchurComplementTranspose()`, `MATSOLVERMUMPS`, `MATSOLVERMKL_PARDISO`
9711: @*/
9712: PetscErrorCode MatFactorSetSchurIS(Mat mat, IS is)
9713: {
9714: PetscErrorCode (*f)(Mat, IS);
9716: PetscFunctionBegin;
9721: PetscCheckSameComm(mat, 1, is, 2);
9722: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Only for factored matrix");
9723: PetscCall(PetscObjectQueryFunction((PetscObject)mat, "MatFactorSetSchurIS_C", &f));
9724: PetscCheck(f, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "The selected MatSolverType does not support Schur complement computation. You should use MATSOLVERMUMPS or MATSOLVERMKL_PARDISO");
9725: PetscCall(MatDestroy(&mat->schur));
9726: PetscCall((*f)(mat, is));
9727: PetscCheck(mat->schur, PetscObjectComm((PetscObject)mat), PETSC_ERR_PLIB, "Schur complement has not been created");
9728: PetscFunctionReturn(PETSC_SUCCESS);
9729: }
9731: /*@
9732: MatFactorCreateSchurComplement - Create a Schur complement matrix object using Schur data computed during the factorization step
9734: Logically Collective
9736: Input Parameters:
9737: + F - the factored matrix obtained by calling `MatGetFactor()`
9738: . S - location where to return the Schur complement, can be `NULL`
9739: - status - the status of the Schur complement matrix, can be `NULL`
9741: Level: advanced
9743: Notes:
9744: You must call `MatFactorSetSchurIS()` before calling this routine.
9746: This functionality is only supported for `MATSOLVERMUMPS` and `MATSOLVERMKL_PARDISO`
9748: The routine provides a copy of the Schur matrix stored within the solver data structures.
9749: The caller must destroy the object when it is no longer needed.
9750: If `MatFactorInvertSchurComplement()` has been called, the routine gets back the inverse.
9752: 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)
9754: See `MatCreateSchurComplement()` or `MatGetSchurComplement()` for ways to create virtual or approximate Schur complements.
9756: Developer Note:
9757: The reason this routine exists is because the representation of the Schur complement within the factor matrix may be different than a standard PETSc
9758: matrix representation and we normally do not want to use the time or memory to make a copy as a regular PETSc matrix.
9760: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorGetSchurComplement()`, `MatFactorSchurStatus`, `MATSOLVERMUMPS`, `MATSOLVERMKL_PARDISO`
9761: @*/
9762: PetscErrorCode MatFactorCreateSchurComplement(Mat F, Mat *S, MatFactorSchurStatus *status)
9763: {
9764: PetscFunctionBegin;
9766: if (S) PetscAssertPointer(S, 2);
9767: if (status) PetscAssertPointer(status, 3);
9768: if (S) {
9769: PetscErrorCode (*f)(Mat, Mat *);
9771: PetscCall(PetscObjectQueryFunction((PetscObject)F, "MatFactorCreateSchurComplement_C", &f));
9772: if (f) {
9773: PetscCall((*f)(F, S));
9774: } else {
9775: PetscCall(MatDuplicate(F->schur, MAT_COPY_VALUES, S));
9776: }
9777: }
9778: if (status) *status = F->schur_status;
9779: PetscFunctionReturn(PETSC_SUCCESS);
9780: }
9782: /*@
9783: MatFactorGetSchurComplement - Gets access to a Schur complement matrix using the current Schur data within a factored matrix
9785: Logically Collective
9787: Input Parameters:
9788: + F - the factored matrix obtained by calling `MatGetFactor()`
9789: . S - location where to return the Schur complement, can be `NULL`
9790: - status - the status of the Schur complement matrix, can be `NULL`
9792: Level: advanced
9794: Notes:
9795: You must call `MatFactorSetSchurIS()` before calling this routine.
9797: Schur complement mode is currently implemented for sequential matrices with factor type of `MATSOLVERMUMPS`
9799: The routine returns a the Schur Complement stored within the data structures of the solver.
9801: If `MatFactorInvertSchurComplement()` has previously been called, the returned matrix is actually the inverse of the Schur complement.
9803: The returned matrix should not be destroyed; the caller should call `MatFactorRestoreSchurComplement()` when the object is no longer needed.
9805: Use `MatFactorCreateSchurComplement()` to create a copy of the Schur complement matrix that is within a factored matrix
9807: See `MatCreateSchurComplement()` or `MatGetSchurComplement()` for ways to create virtual or approximate Schur complements.
9809: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorRestoreSchurComplement()`, `MatFactorCreateSchurComplement()`, `MatFactorSchurStatus`
9810: @*/
9811: PetscErrorCode MatFactorGetSchurComplement(Mat F, Mat *S, MatFactorSchurStatus *status)
9812: {
9813: PetscFunctionBegin;
9815: if (S) {
9816: PetscAssertPointer(S, 2);
9817: *S = F->schur;
9818: }
9819: if (status) {
9820: PetscAssertPointer(status, 3);
9821: *status = F->schur_status;
9822: }
9823: PetscFunctionReturn(PETSC_SUCCESS);
9824: }
9826: static PetscErrorCode MatFactorUpdateSchurStatus_Private(Mat F)
9827: {
9828: Mat S = F->schur;
9830: PetscFunctionBegin;
9831: switch (F->schur_status) {
9832: case MAT_FACTOR_SCHUR_UNFACTORED: // fall-through
9833: case MAT_FACTOR_SCHUR_INVERTED:
9834: if (S) {
9835: S->ops->solve = NULL;
9836: S->ops->matsolve = NULL;
9837: S->ops->solvetranspose = NULL;
9838: S->ops->matsolvetranspose = NULL;
9839: S->ops->solveadd = NULL;
9840: S->ops->solvetransposeadd = NULL;
9841: S->factortype = MAT_FACTOR_NONE;
9842: PetscCall(PetscFree(S->solvertype));
9843: }
9844: case MAT_FACTOR_SCHUR_FACTORED: // fall-through
9845: break;
9846: default:
9847: SETERRQ(PetscObjectComm((PetscObject)F), PETSC_ERR_SUP, "Unhandled MatFactorSchurStatus %d", F->schur_status);
9848: }
9849: PetscFunctionReturn(PETSC_SUCCESS);
9850: }
9852: /*@
9853: MatFactorRestoreSchurComplement - Restore the Schur complement matrix object obtained from a call to `MatFactorGetSchurComplement()`
9855: Logically Collective
9857: Input Parameters:
9858: + F - the factored matrix obtained by calling `MatGetFactor()`
9859: . S - location where the Schur complement is stored
9860: - status - the status of the Schur complement matrix (see `MatFactorSchurStatus`)
9862: Level: advanced
9864: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorCreateSchurComplement()`, `MatFactorSchurStatus`
9865: @*/
9866: PetscErrorCode MatFactorRestoreSchurComplement(Mat F, Mat *S, MatFactorSchurStatus status)
9867: {
9868: PetscFunctionBegin;
9870: if (S) {
9872: *S = NULL;
9873: }
9874: F->schur_status = status;
9875: PetscCall(MatFactorUpdateSchurStatus_Private(F));
9876: PetscFunctionReturn(PETSC_SUCCESS);
9877: }
9879: /*@
9880: MatFactorSolveSchurComplementTranspose - Solve the transpose of the Schur complement system computed during the factorization step
9882: Logically Collective
9884: Input Parameters:
9885: + F - the factored matrix obtained by calling `MatGetFactor()`
9886: . rhs - location where the right-hand side of the Schur complement system is stored
9887: - sol - location where the solution of the Schur complement system has to be returned
9889: Level: advanced
9891: Notes:
9892: The sizes of the vectors should match the size of the Schur complement
9894: Must be called after `MatFactorSetSchurIS()`
9896: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorSolveSchurComplement()`
9897: @*/
9898: PetscErrorCode MatFactorSolveSchurComplementTranspose(Mat F, Vec rhs, Vec sol)
9899: {
9900: PetscFunctionBegin;
9907: PetscCheckSameComm(F, 1, rhs, 2);
9908: PetscCheckSameComm(F, 1, sol, 3);
9909: PetscCall(MatFactorFactorizeSchurComplement(F));
9910: switch (F->schur_status) {
9911: case MAT_FACTOR_SCHUR_FACTORED:
9912: PetscCall(MatSolveTranspose(F->schur, rhs, sol));
9913: break;
9914: case MAT_FACTOR_SCHUR_INVERTED:
9915: PetscCall(MatMultTranspose(F->schur, rhs, sol));
9916: break;
9917: default:
9918: SETERRQ(PetscObjectComm((PetscObject)F), PETSC_ERR_SUP, "Unhandled MatFactorSchurStatus %d", F->schur_status);
9919: }
9920: PetscFunctionReturn(PETSC_SUCCESS);
9921: }
9923: /*@
9924: MatFactorSolveSchurComplement - Solve the Schur complement system computed during the factorization step
9926: Logically Collective
9928: Input Parameters:
9929: + F - the factored matrix obtained by calling `MatGetFactor()`
9930: . rhs - location where the right-hand side of the Schur complement system is stored
9931: - sol - location where the solution of the Schur complement system has to be returned
9933: Level: advanced
9935: Notes:
9936: The sizes of the vectors should match the size of the Schur complement
9938: Must be called after `MatFactorSetSchurIS()`
9940: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorSolveSchurComplementTranspose()`
9941: @*/
9942: PetscErrorCode MatFactorSolveSchurComplement(Mat F, Vec rhs, Vec sol)
9943: {
9944: PetscFunctionBegin;
9951: PetscCheckSameComm(F, 1, rhs, 2);
9952: PetscCheckSameComm(F, 1, sol, 3);
9953: PetscCall(MatFactorFactorizeSchurComplement(F));
9954: switch (F->schur_status) {
9955: case MAT_FACTOR_SCHUR_FACTORED:
9956: PetscCall(MatSolve(F->schur, rhs, sol));
9957: break;
9958: case MAT_FACTOR_SCHUR_INVERTED:
9959: PetscCall(MatMult(F->schur, rhs, sol));
9960: break;
9961: default:
9962: SETERRQ(PetscObjectComm((PetscObject)F), PETSC_ERR_SUP, "Unhandled MatFactorSchurStatus %d", F->schur_status);
9963: }
9964: PetscFunctionReturn(PETSC_SUCCESS);
9965: }
9967: PETSC_EXTERN PetscErrorCode MatSeqDenseInvertFactors_Private(Mat);
9968: #if PetscDefined(HAVE_CUDA)
9969: PETSC_SINGLE_LIBRARY_INTERN PetscErrorCode MatSeqDenseCUDAInvertFactors_Internal(Mat);
9970: #endif
9972: /* Schur status updated in the interface */
9973: static PetscErrorCode MatFactorInvertSchurComplement_Private(Mat F)
9974: {
9975: Mat S = F->schur;
9977: PetscFunctionBegin;
9978: if (S) {
9979: PetscMPIInt size;
9980: PetscBool isdense, isdensecuda;
9982: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)S), &size));
9983: PetscCheck(size <= 1, PetscObjectComm((PetscObject)S), PETSC_ERR_SUP, "Not yet implemented");
9984: PetscCall(PetscObjectTypeCompare((PetscObject)S, MATSEQDENSE, &isdense));
9985: PetscCall(PetscObjectTypeCompare((PetscObject)S, MATSEQDENSECUDA, &isdensecuda));
9986: PetscCheck(isdense || isdensecuda, PetscObjectComm((PetscObject)S), PETSC_ERR_SUP, "Not implemented for type %s", ((PetscObject)S)->type_name);
9987: PetscCall(PetscLogEventBegin(MAT_FactorInvS, F, 0, 0, 0));
9988: if (isdense) {
9989: PetscCall(MatSeqDenseInvertFactors_Private(S));
9990: } else if (isdensecuda) {
9991: #if defined(PETSC_HAVE_CUDA)
9992: PetscCall(MatSeqDenseCUDAInvertFactors_Internal(S));
9993: #endif
9994: }
9995: // HIP??????????????
9996: PetscCall(PetscLogEventEnd(MAT_FactorInvS, F, 0, 0, 0));
9997: }
9998: PetscFunctionReturn(PETSC_SUCCESS);
9999: }
10001: /*@
10002: MatFactorInvertSchurComplement - Invert the Schur complement matrix computed during the factorization step
10004: Logically Collective
10006: Input Parameter:
10007: . F - the factored matrix obtained by calling `MatGetFactor()`
10009: Level: advanced
10011: Notes:
10012: Must be called after `MatFactorSetSchurIS()`.
10014: Call `MatFactorGetSchurComplement()` or `MatFactorCreateSchurComplement()` AFTER this call to actually compute the inverse and get access to it.
10016: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorGetSchurComplement()`, `MatFactorCreateSchurComplement()`
10017: @*/
10018: PetscErrorCode MatFactorInvertSchurComplement(Mat F)
10019: {
10020: PetscFunctionBegin;
10023: if (F->schur_status == MAT_FACTOR_SCHUR_INVERTED) PetscFunctionReturn(PETSC_SUCCESS);
10024: PetscCall(MatFactorFactorizeSchurComplement(F));
10025: PetscCall(MatFactorInvertSchurComplement_Private(F));
10026: F->schur_status = MAT_FACTOR_SCHUR_INVERTED;
10027: PetscFunctionReturn(PETSC_SUCCESS);
10028: }
10030: /*@
10031: MatFactorFactorizeSchurComplement - Factorize the Schur complement matrix computed during the factorization step
10033: Logically Collective
10035: Input Parameter:
10036: . F - the factored matrix obtained by calling `MatGetFactor()`
10038: Level: advanced
10040: Note:
10041: Must be called after `MatFactorSetSchurIS()`
10043: .seealso: [](ch_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorInvertSchurComplement()`
10044: @*/
10045: PetscErrorCode MatFactorFactorizeSchurComplement(Mat F)
10046: {
10047: MatFactorInfo info;
10049: PetscFunctionBegin;
10052: if (F->schur_status == MAT_FACTOR_SCHUR_INVERTED || F->schur_status == MAT_FACTOR_SCHUR_FACTORED) PetscFunctionReturn(PETSC_SUCCESS);
10053: PetscCall(PetscLogEventBegin(MAT_FactorFactS, F, 0, 0, 0));
10054: PetscCall(PetscMemzero(&info, sizeof(MatFactorInfo)));
10055: if (F->factortype == MAT_FACTOR_CHOLESKY) { /* LDL^t regarded as Cholesky */
10056: PetscCall(MatCholeskyFactor(F->schur, NULL, &info));
10057: } else {
10058: PetscCall(MatLUFactor(F->schur, NULL, NULL, &info));
10059: }
10060: PetscCall(PetscLogEventEnd(MAT_FactorFactS, F, 0, 0, 0));
10061: F->schur_status = MAT_FACTOR_SCHUR_FACTORED;
10062: PetscFunctionReturn(PETSC_SUCCESS);
10063: }
10065: /*@
10066: MatPtAP - Creates the matrix product $C = P^T * A * P$
10068: Neighbor-wise Collective
10070: Input Parameters:
10071: + A - the matrix
10072: . P - the projection matrix
10073: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10074: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(P)), use `PETSC_DEFAULT` if you do not have a good estimate
10075: if the result is a dense matrix this is irrelevant
10077: Output Parameter:
10078: . C - the product matrix
10080: Level: intermediate
10082: Notes:
10083: C will be created and must be destroyed by the user with `MatDestroy()`.
10085: An alternative approach to this function is to use `MatProductCreate()` and set the desired options before the computation is done
10087: Developer Note:
10088: For matrix types without special implementation the function fallbacks to `MatMatMult()` followed by `MatTransposeMatMult()`.
10090: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MatMatMult()`, `MatRARt()`
10091: @*/
10092: PetscErrorCode MatPtAP(Mat A, Mat P, MatReuse scall, PetscReal fill, Mat *C)
10093: {
10094: PetscFunctionBegin;
10095: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*C, 5);
10096: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
10098: if (scall == MAT_INITIAL_MATRIX) {
10099: PetscCall(MatProductCreate(A, P, NULL, C));
10100: PetscCall(MatProductSetType(*C, MATPRODUCT_PtAP));
10101: PetscCall(MatProductSetAlgorithm(*C, "default"));
10102: PetscCall(MatProductSetFill(*C, fill));
10104: (*C)->product->api_user = PETSC_TRUE;
10105: PetscCall(MatProductSetFromOptions(*C));
10106: 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);
10107: PetscCall(MatProductSymbolic(*C));
10108: } else { /* scall == MAT_REUSE_MATRIX */
10109: PetscCall(MatProductReplaceMats(A, P, NULL, *C));
10110: }
10112: PetscCall(MatProductNumeric(*C));
10113: (*C)->symmetric = A->symmetric;
10114: (*C)->spd = A->spd;
10115: PetscFunctionReturn(PETSC_SUCCESS);
10116: }
10118: /*@
10119: MatRARt - Creates the matrix product $C = R * A * R^T$
10121: Neighbor-wise Collective
10123: Input Parameters:
10124: + A - the matrix
10125: . R - the projection matrix
10126: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10127: - fill - expected fill as ratio of nnz(C)/nnz(A), use `PETSC_DEFAULT` if you do not have a good estimate
10128: if the result is a dense matrix this is irrelevant
10130: Output Parameter:
10131: . C - the product matrix
10133: Level: intermediate
10135: Notes:
10136: C will be created and must be destroyed by the user with `MatDestroy()`.
10138: An alternative approach to this function is to use `MatProductCreate()` and set the desired options before the computation is done
10140: This routine is currently only implemented for pairs of `MATAIJ` matrices and classes
10141: which inherit from `MATAIJ`. Due to PETSc sparse matrix block row distribution among processes,
10142: parallel MatRARt is implemented via explicit transpose of R, which could be very expensive.
10143: We recommend using MatPtAP().
10145: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MatMatMult()`, `MatPtAP()`
10146: @*/
10147: PetscErrorCode MatRARt(Mat A, Mat R, MatReuse scall, PetscReal fill, Mat *C)
10148: {
10149: PetscFunctionBegin;
10150: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*C, 5);
10151: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
10153: if (scall == MAT_INITIAL_MATRIX) {
10154: PetscCall(MatProductCreate(A, R, NULL, C));
10155: PetscCall(MatProductSetType(*C, MATPRODUCT_RARt));
10156: PetscCall(MatProductSetAlgorithm(*C, "default"));
10157: PetscCall(MatProductSetFill(*C, fill));
10159: (*C)->product->api_user = PETSC_TRUE;
10160: PetscCall(MatProductSetFromOptions(*C));
10161: 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);
10162: PetscCall(MatProductSymbolic(*C));
10163: } else { /* scall == MAT_REUSE_MATRIX */
10164: PetscCall(MatProductReplaceMats(A, R, NULL, *C));
10165: }
10167: PetscCall(MatProductNumeric(*C));
10168: if (A->symmetric == PETSC_BOOL3_TRUE) PetscCall(MatSetOption(*C, MAT_SYMMETRIC, PETSC_TRUE));
10169: PetscFunctionReturn(PETSC_SUCCESS);
10170: }
10172: static PetscErrorCode MatProduct_Private(Mat A, Mat B, MatReuse scall, PetscReal fill, MatProductType ptype, Mat *C)
10173: {
10174: PetscBool flg = PETSC_TRUE;
10176: PetscFunctionBegin;
10177: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "MAT_INPLACE_MATRIX product not supported");
10178: if (scall == MAT_INITIAL_MATRIX) {
10179: PetscCall(PetscInfo(A, "Calling MatProduct API with MAT_INITIAL_MATRIX and product type %s\n", MatProductTypes[ptype]));
10180: PetscCall(MatProductCreate(A, B, NULL, C));
10181: PetscCall(MatProductSetAlgorithm(*C, MATPRODUCTALGORITHMDEFAULT));
10182: PetscCall(MatProductSetFill(*C, fill));
10183: } else { /* scall == MAT_REUSE_MATRIX */
10184: Mat_Product *product = (*C)->product;
10186: PetscCall(PetscObjectBaseTypeCompareAny((PetscObject)*C, &flg, MATSEQDENSE, MATMPIDENSE, ""));
10187: if (flg && product && product->type != ptype) {
10188: PetscCall(MatProductClear(*C));
10189: product = NULL;
10190: }
10191: PetscCall(PetscInfo(A, "Calling MatProduct API with MAT_REUSE_MATRIX %s product present and product type %s\n", product ? "with" : "without", MatProductTypes[ptype]));
10192: if (!product) { /* user provide the dense matrix *C without calling MatProductCreate() or reusing it from previous calls */
10193: PetscCheck(flg, PetscObjectComm((PetscObject)*C), PETSC_ERR_SUP, "Call MatProductCreate() first");
10194: PetscCall(MatProductCreate_Private(A, B, NULL, *C));
10195: product = (*C)->product;
10196: product->fill = fill;
10197: product->clear = PETSC_TRUE;
10198: } else { /* user may change input matrices A or B when MAT_REUSE_MATRIX */
10199: flg = PETSC_FALSE;
10200: PetscCall(MatProductReplaceMats(A, B, NULL, *C));
10201: }
10202: }
10203: if (flg) {
10204: (*C)->product->api_user = PETSC_TRUE;
10205: PetscCall(MatProductSetType(*C, ptype));
10206: PetscCall(MatProductSetFromOptions(*C));
10207: PetscCall(MatProductSymbolic(*C));
10208: }
10209: PetscCall(MatProductNumeric(*C));
10210: PetscFunctionReturn(PETSC_SUCCESS);
10211: }
10213: /*@
10214: MatMatMult - Performs matrix-matrix multiplication C=A*B.
10216: Neighbor-wise Collective
10218: Input Parameters:
10219: + A - the left matrix
10220: . B - the right matrix
10221: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10222: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use `PETSC_DEFAULT` if you do not have a good estimate
10223: if the result is a dense matrix this is irrelevant
10225: Output Parameter:
10226: . C - the product matrix
10228: Notes:
10229: Unless scall is `MAT_REUSE_MATRIX` C will be created.
10231: `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
10232: call to this function with `MAT_INITIAL_MATRIX`.
10234: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value actually needed.
10236: 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`,
10237: rather than first having `MatMatMult()` create it for you. You can NEVER do this if the matrix C is sparse.
10239: Example of Usage:
10240: .vb
10241: MatProductCreate(A,B,NULL,&C);
10242: MatProductSetType(C,MATPRODUCT_AB);
10243: MatProductSymbolic(C);
10244: MatProductNumeric(C); // compute C=A * B
10245: MatProductReplaceMats(A1,B1,NULL,C); // compute C=A1 * B1
10246: MatProductNumeric(C);
10247: MatProductReplaceMats(A2,NULL,NULL,C); // compute C=A2 * B1
10248: MatProductNumeric(C);
10249: .ve
10251: Level: intermediate
10253: .seealso: [](ch_matrices), `Mat`, `MatProductType`, `MATPRODUCT_AB`, `MatTransposeMatMult()`, `MatMatTransposeMult()`, `MatPtAP()`, `MatProductCreate()`, `MatProductSymbolic()`, `MatProductReplaceMats()`, `MatProductNumeric()`
10254: @*/
10255: PetscErrorCode MatMatMult(Mat A, Mat B, MatReuse scall, PetscReal fill, Mat *C)
10256: {
10257: PetscFunctionBegin;
10258: PetscCall(MatProduct_Private(A, B, scall, fill, MATPRODUCT_AB, C));
10259: PetscFunctionReturn(PETSC_SUCCESS);
10260: }
10262: /*@
10263: MatMatTransposeMult - Performs matrix-matrix multiplication $C = A*B^T$.
10265: Neighbor-wise Collective
10267: Input Parameters:
10268: + A - the left matrix
10269: . B - the right matrix
10270: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10271: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use `PETSC_DEFAULT` if not known
10273: Output Parameter:
10274: . C - the product matrix
10276: Options Database Key:
10277: . -matmattransmult_mpidense_mpidense_via {allgatherv,cyclic} - Choose between algorithms for `MATMPIDENSE` matrices: the
10278: first redundantly copies the transposed `B` matrix on each process and requires O(log P) communication complexity;
10279: the second never stores more than one portion of the `B` matrix at a time but requires O(P) communication complexity.
10281: Level: intermediate
10283: Notes:
10284: C will be created if `MAT_INITIAL_MATRIX` and must be destroyed by the user with `MatDestroy()`.
10286: `MAT_REUSE_MATRIX` can only be used if the matrices A and B have the same nonzero pattern as in the previous call
10288: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
10289: actually needed.
10291: This routine is currently only implemented for pairs of `MATSEQAIJ` matrices, for the `MATSEQDENSE` class,
10292: and for pairs of `MATMPIDENSE` matrices.
10294: This routine is shorthand for using `MatProductCreate()` with the `MatProductType` of `MATPRODUCT_ABt`
10296: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MATPRODUCT_ABt`, `MatMatMult()`, `MatTransposeMatMult()` `MatPtAP()`, `MatProductAlgorithm`, `MatProductType`
10297: @*/
10298: PetscErrorCode MatMatTransposeMult(Mat A, Mat B, MatReuse scall, PetscReal fill, Mat *C)
10299: {
10300: PetscFunctionBegin;
10301: PetscCall(MatProduct_Private(A, B, scall, fill, MATPRODUCT_ABt, C));
10302: if (A == B) PetscCall(MatSetOption(*C, MAT_SYMMETRIC, PETSC_TRUE));
10303: PetscFunctionReturn(PETSC_SUCCESS);
10304: }
10306: /*@
10307: MatTransposeMatMult - Performs matrix-matrix multiplication $C = A^T*B$.
10309: Neighbor-wise Collective
10311: Input Parameters:
10312: + A - the left matrix
10313: . B - the right matrix
10314: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10315: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use `PETSC_DEFAULT` if not known
10317: Output Parameter:
10318: . C - the product matrix
10320: Level: intermediate
10322: Notes:
10323: `C` will be created if `MAT_INITIAL_MATRIX` and must be destroyed by the user with `MatDestroy()`.
10325: `MAT_REUSE_MATRIX` can only be used if the matrices A and B have the same nonzero pattern as in the previous call.
10327: This routine is shorthand for using `MatProductCreate()` with the `MatProductType` of `MATPRODUCT_AtB`
10329: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
10330: actually needed.
10332: This routine is currently implemented for pairs of `MATAIJ` matrices and pairs of `MATSEQDENSE` matrices and classes
10333: which inherit from `MATSEQAIJ`. `C` will be of the same type as the input matrices.
10335: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MATPRODUCT_AtB`, `MatMatMult()`, `MatMatTransposeMult()`, `MatPtAP()`
10336: @*/
10337: PetscErrorCode MatTransposeMatMult(Mat A, Mat B, MatReuse scall, PetscReal fill, Mat *C)
10338: {
10339: PetscFunctionBegin;
10340: PetscCall(MatProduct_Private(A, B, scall, fill, MATPRODUCT_AtB, C));
10341: PetscFunctionReturn(PETSC_SUCCESS);
10342: }
10344: /*@
10345: MatMatMatMult - Performs matrix-matrix-matrix multiplication D=A*B*C.
10347: Neighbor-wise Collective
10349: Input Parameters:
10350: + A - the left matrix
10351: . B - the middle matrix
10352: . C - the right matrix
10353: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10354: - 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
10355: if the result is a dense matrix this is irrelevant
10357: Output Parameter:
10358: . D - the product matrix
10360: Level: intermediate
10362: Notes:
10363: Unless `scall` is `MAT_REUSE_MATRIX` D will be created.
10365: `MAT_REUSE_MATRIX` can only be used if the matrices `A`, `B`, and `C` have the same nonzero pattern as in the previous call
10367: This routine is shorthand for using `MatProductCreate()` with the `MatProductType` of `MATPRODUCT_ABC`
10369: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
10370: actually needed.
10372: If you have many matrices with the same non-zero structure to multiply, you
10373: should use `MAT_REUSE_MATRIX` in all calls but the first
10375: .seealso: [](ch_matrices), `Mat`, `MatProductCreate()`, `MATPRODUCT_ABC`, `MatMatMult`, `MatPtAP()`, `MatMatTransposeMult()`, `MatTransposeMatMult()`
10376: @*/
10377: PetscErrorCode MatMatMatMult(Mat A, Mat B, Mat C, MatReuse scall, PetscReal fill, Mat *D)
10378: {
10379: PetscFunctionBegin;
10380: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*D, 6);
10381: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
10383: if (scall == MAT_INITIAL_MATRIX) {
10384: PetscCall(MatProductCreate(A, B, C, D));
10385: PetscCall(MatProductSetType(*D, MATPRODUCT_ABC));
10386: PetscCall(MatProductSetAlgorithm(*D, "default"));
10387: PetscCall(MatProductSetFill(*D, fill));
10389: (*D)->product->api_user = PETSC_TRUE;
10390: PetscCall(MatProductSetFromOptions(*D));
10391: 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,
10392: ((PetscObject)C)->type_name);
10393: PetscCall(MatProductSymbolic(*D));
10394: } else { /* user may change input matrices when REUSE */
10395: PetscCall(MatProductReplaceMats(A, B, C, *D));
10396: }
10397: PetscCall(MatProductNumeric(*D));
10398: PetscFunctionReturn(PETSC_SUCCESS);
10399: }
10401: /*@
10402: MatCreateRedundantMatrix - Create redundant matrices and put them into processors of subcommunicators.
10404: Collective
10406: Input Parameters:
10407: + mat - the matrix
10408: . nsubcomm - the number of subcommunicators (= number of redundant parallel or sequential matrices)
10409: . subcomm - MPI communicator split from the communicator where mat resides in (or `MPI_COMM_NULL` if nsubcomm is used)
10410: - reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10412: Output Parameter:
10413: . matredundant - redundant matrix
10415: Level: advanced
10417: Notes:
10418: `MAT_REUSE_MATRIX` can only be used when the nonzero structure of the
10419: original matrix has not changed from that last call to `MatCreateRedundantMatrix()`.
10421: This routine creates the duplicated matrices in the subcommunicators; you should NOT create them before
10422: calling it.
10424: `PetscSubcommCreate()` can be used to manage the creation of the subcomm but need not be.
10426: .seealso: [](ch_matrices), `Mat`, `MatDestroy()`, `PetscSubcommCreate()`, `PetscSubcomm`
10427: @*/
10428: PetscErrorCode MatCreateRedundantMatrix(Mat mat, PetscInt nsubcomm, MPI_Comm subcomm, MatReuse reuse, Mat *matredundant)
10429: {
10430: MPI_Comm comm;
10431: PetscMPIInt size;
10432: PetscInt mloc_sub, nloc_sub, rstart, rend, M = mat->rmap->N, N = mat->cmap->N, bs = mat->rmap->bs;
10433: Mat_Redundant *redund = NULL;
10434: PetscSubcomm psubcomm = NULL;
10435: MPI_Comm subcomm_in = subcomm;
10436: Mat *matseq;
10437: IS isrow, iscol;
10438: PetscBool newsubcomm = PETSC_FALSE;
10440: PetscFunctionBegin;
10442: if (nsubcomm && reuse == MAT_REUSE_MATRIX) {
10443: PetscAssertPointer(*matredundant, 5);
10445: }
10447: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
10448: if (size == 1 || nsubcomm == 1) {
10449: if (reuse == MAT_INITIAL_MATRIX) {
10450: PetscCall(MatDuplicate(mat, MAT_COPY_VALUES, matredundant));
10451: } else {
10452: 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");
10453: PetscCall(MatCopy(mat, *matredundant, SAME_NONZERO_PATTERN));
10454: }
10455: PetscFunctionReturn(PETSC_SUCCESS);
10456: }
10458: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10459: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10460: MatCheckPreallocated(mat, 1);
10462: PetscCall(PetscLogEventBegin(MAT_RedundantMat, mat, 0, 0, 0));
10463: if (subcomm_in == MPI_COMM_NULL && reuse == MAT_INITIAL_MATRIX) { /* get subcomm if user does not provide subcomm */
10464: /* create psubcomm, then get subcomm */
10465: PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
10466: PetscCallMPI(MPI_Comm_size(comm, &size));
10467: PetscCheck(nsubcomm >= 1 && nsubcomm <= size, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "nsubcomm must between 1 and %d", size);
10469: PetscCall(PetscSubcommCreate(comm, &psubcomm));
10470: PetscCall(PetscSubcommSetNumber(psubcomm, nsubcomm));
10471: PetscCall(PetscSubcommSetType(psubcomm, PETSC_SUBCOMM_CONTIGUOUS));
10472: PetscCall(PetscSubcommSetFromOptions(psubcomm));
10473: PetscCall(PetscCommDuplicate(PetscSubcommChild(psubcomm), &subcomm, NULL));
10474: newsubcomm = PETSC_TRUE;
10475: PetscCall(PetscSubcommDestroy(&psubcomm));
10476: }
10478: /* get isrow, iscol and a local sequential matrix matseq[0] */
10479: if (reuse == MAT_INITIAL_MATRIX) {
10480: mloc_sub = PETSC_DECIDE;
10481: nloc_sub = PETSC_DECIDE;
10482: if (bs < 1) {
10483: PetscCall(PetscSplitOwnership(subcomm, &mloc_sub, &M));
10484: PetscCall(PetscSplitOwnership(subcomm, &nloc_sub, &N));
10485: } else {
10486: PetscCall(PetscSplitOwnershipBlock(subcomm, bs, &mloc_sub, &M));
10487: PetscCall(PetscSplitOwnershipBlock(subcomm, bs, &nloc_sub, &N));
10488: }
10489: PetscCallMPI(MPI_Scan(&mloc_sub, &rend, 1, MPIU_INT, MPI_SUM, subcomm));
10490: rstart = rend - mloc_sub;
10491: PetscCall(ISCreateStride(PETSC_COMM_SELF, mloc_sub, rstart, 1, &isrow));
10492: PetscCall(ISCreateStride(PETSC_COMM_SELF, N, 0, 1, &iscol));
10493: PetscCall(ISSetIdentity(iscol));
10494: } else { /* reuse == MAT_REUSE_MATRIX */
10495: 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");
10496: /* retrieve subcomm */
10497: PetscCall(PetscObjectGetComm((PetscObject)*matredundant, &subcomm));
10498: redund = (*matredundant)->redundant;
10499: isrow = redund->isrow;
10500: iscol = redund->iscol;
10501: matseq = redund->matseq;
10502: }
10503: PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscol, reuse, &matseq));
10505: /* get matredundant over subcomm */
10506: if (reuse == MAT_INITIAL_MATRIX) {
10507: PetscCall(MatCreateMPIMatConcatenateSeqMat(subcomm, matseq[0], nloc_sub, reuse, matredundant));
10509: /* create a supporting struct and attach it to C for reuse */
10510: PetscCall(PetscNew(&redund));
10511: (*matredundant)->redundant = redund;
10512: redund->isrow = isrow;
10513: redund->iscol = iscol;
10514: redund->matseq = matseq;
10515: if (newsubcomm) {
10516: redund->subcomm = subcomm;
10517: } else {
10518: redund->subcomm = MPI_COMM_NULL;
10519: }
10520: } else {
10521: PetscCall(MatCreateMPIMatConcatenateSeqMat(subcomm, matseq[0], PETSC_DECIDE, reuse, matredundant));
10522: }
10523: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
10524: if (matseq[0]->boundtocpu && matseq[0]->bindingpropagates) {
10525: PetscCall(MatBindToCPU(*matredundant, PETSC_TRUE));
10526: PetscCall(MatSetBindingPropagates(*matredundant, PETSC_TRUE));
10527: }
10528: #endif
10529: PetscCall(PetscLogEventEnd(MAT_RedundantMat, mat, 0, 0, 0));
10530: PetscFunctionReturn(PETSC_SUCCESS);
10531: }
10533: /*@C
10534: MatGetMultiProcBlock - Create multiple 'parallel submatrices' from
10535: a given `Mat`. Each submatrix can span multiple procs.
10537: Collective
10539: Input Parameters:
10540: + mat - the matrix
10541: . subComm - the sub communicator obtained as if by `MPI_Comm_split(PetscObjectComm((PetscObject)mat))`
10542: - scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10544: Output Parameter:
10545: . subMat - parallel sub-matrices each spanning a given `subcomm`
10547: Level: advanced
10549: Notes:
10550: The submatrix partition across processors is dictated by `subComm` a
10551: communicator obtained by `MPI_comm_split()` or via `PetscSubcommCreate()`. The `subComm`
10552: is not restricted to be grouped with consecutive original MPI processes.
10554: Due the `MPI_Comm_split()` usage, the parallel layout of the submatrices
10555: map directly to the layout of the original matrix [wrt the local
10556: row,col partitioning]. So the original 'DiagonalMat' naturally maps
10557: into the 'DiagonalMat' of the `subMat`, hence it is used directly from
10558: the `subMat`. However the offDiagMat looses some columns - and this is
10559: reconstructed with `MatSetValues()`
10561: This is used by `PCBJACOBI` when a single block spans multiple MPI processes.
10563: .seealso: [](ch_matrices), `Mat`, `MatCreateRedundantMatrix()`, `MatCreateSubMatrices()`, `PCBJACOBI`
10564: @*/
10565: PetscErrorCode MatGetMultiProcBlock(Mat mat, MPI_Comm subComm, MatReuse scall, Mat *subMat)
10566: {
10567: PetscMPIInt commsize, subCommSize;
10569: PetscFunctionBegin;
10570: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &commsize));
10571: PetscCallMPI(MPI_Comm_size(subComm, &subCommSize));
10572: PetscCheck(subCommSize <= commsize, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "CommSize %d < SubCommZize %d", commsize, subCommSize);
10574: 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");
10575: PetscCall(PetscLogEventBegin(MAT_GetMultiProcBlock, mat, 0, 0, 0));
10576: PetscUseTypeMethod(mat, getmultiprocblock, subComm, scall, subMat);
10577: PetscCall(PetscLogEventEnd(MAT_GetMultiProcBlock, mat, 0, 0, 0));
10578: PetscFunctionReturn(PETSC_SUCCESS);
10579: }
10581: /*@
10582: MatGetLocalSubMatrix - Gets a reference to a submatrix specified in local numbering
10584: Not Collective
10586: Input Parameters:
10587: + mat - matrix to extract local submatrix from
10588: . isrow - local row indices for submatrix
10589: - iscol - local column indices for submatrix
10591: Output Parameter:
10592: . submat - the submatrix
10594: Level: intermediate
10596: Notes:
10597: `submat` should be disposed of with `MatRestoreLocalSubMatrix()`.
10599: Depending on the format of `mat`, the returned `submat` may not implement `MatMult()`. Its communicator may be
10600: the same as `mat`, it may be `PETSC_COMM_SELF`, or some other sub-communictor of `mat`'s.
10602: `submat` always implements `MatSetValuesLocal()`. If `isrow` and `iscol` have the same block size, then
10603: `MatSetValuesBlockedLocal()` will also be implemented.
10605: `mat` must have had a `ISLocalToGlobalMapping` provided to it with `MatSetLocalToGlobalMapping()`.
10606: Matrices obtained with `DMCreateMatrix()` generally already have the local to global mapping provided.
10608: .seealso: [](ch_matrices), `Mat`, `MatRestoreLocalSubMatrix()`, `MatCreateLocalRef()`, `MatSetLocalToGlobalMapping()`
10609: @*/
10610: PetscErrorCode MatGetLocalSubMatrix(Mat mat, IS isrow, IS iscol, Mat *submat)
10611: {
10612: PetscFunctionBegin;
10616: PetscCheckSameComm(isrow, 2, iscol, 3);
10617: PetscAssertPointer(submat, 4);
10618: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Matrix must have local to global mapping provided before this call");
10620: if (mat->ops->getlocalsubmatrix) {
10621: PetscUseTypeMethod(mat, getlocalsubmatrix, isrow, iscol, submat);
10622: } else {
10623: PetscCall(MatCreateLocalRef(mat, isrow, iscol, submat));
10624: }
10625: PetscFunctionReturn(PETSC_SUCCESS);
10626: }
10628: /*@
10629: MatRestoreLocalSubMatrix - Restores a reference to a submatrix specified in local numbering obtained with `MatGetLocalSubMatrix()`
10631: Not Collective
10633: Input Parameters:
10634: + mat - matrix to extract local submatrix from
10635: . isrow - local row indices for submatrix
10636: . iscol - local column indices for submatrix
10637: - submat - the submatrix
10639: Level: intermediate
10641: .seealso: [](ch_matrices), `Mat`, `MatGetLocalSubMatrix()`
10642: @*/
10643: PetscErrorCode MatRestoreLocalSubMatrix(Mat mat, IS isrow, IS iscol, Mat *submat)
10644: {
10645: PetscFunctionBegin;
10649: PetscCheckSameComm(isrow, 2, iscol, 3);
10650: PetscAssertPointer(submat, 4);
10653: if (mat->ops->restorelocalsubmatrix) {
10654: PetscUseTypeMethod(mat, restorelocalsubmatrix, isrow, iscol, submat);
10655: } else {
10656: PetscCall(MatDestroy(submat));
10657: }
10658: *submat = NULL;
10659: PetscFunctionReturn(PETSC_SUCCESS);
10660: }
10662: /*@
10663: MatFindZeroDiagonals - Finds all the rows of a matrix that have zero or no diagonal entry in the matrix
10665: Collective
10667: Input Parameter:
10668: . mat - the matrix
10670: Output Parameter:
10671: . is - if any rows have zero diagonals this contains the list of them
10673: Level: developer
10675: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
10676: @*/
10677: PetscErrorCode MatFindZeroDiagonals(Mat mat, IS *is)
10678: {
10679: PetscFunctionBegin;
10682: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10683: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10685: if (!mat->ops->findzerodiagonals) {
10686: Vec diag;
10687: const PetscScalar *a;
10688: PetscInt *rows;
10689: PetscInt rStart, rEnd, r, nrow = 0;
10691: PetscCall(MatCreateVecs(mat, &diag, NULL));
10692: PetscCall(MatGetDiagonal(mat, diag));
10693: PetscCall(MatGetOwnershipRange(mat, &rStart, &rEnd));
10694: PetscCall(VecGetArrayRead(diag, &a));
10695: for (r = 0; r < rEnd - rStart; ++r)
10696: if (a[r] == 0.0) ++nrow;
10697: PetscCall(PetscMalloc1(nrow, &rows));
10698: nrow = 0;
10699: for (r = 0; r < rEnd - rStart; ++r)
10700: if (a[r] == 0.0) rows[nrow++] = r + rStart;
10701: PetscCall(VecRestoreArrayRead(diag, &a));
10702: PetscCall(VecDestroy(&diag));
10703: PetscCall(ISCreateGeneral(PetscObjectComm((PetscObject)mat), nrow, rows, PETSC_OWN_POINTER, is));
10704: } else {
10705: PetscUseTypeMethod(mat, findzerodiagonals, is);
10706: }
10707: PetscFunctionReturn(PETSC_SUCCESS);
10708: }
10710: /*@
10711: MatFindOffBlockDiagonalEntries - Finds all the rows of a matrix that have entries outside of the main diagonal block (defined by the matrix block size)
10713: Collective
10715: Input Parameter:
10716: . mat - the matrix
10718: Output Parameter:
10719: . is - contains the list of rows with off block diagonal entries
10721: Level: developer
10723: .seealso: [](ch_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
10724: @*/
10725: PetscErrorCode MatFindOffBlockDiagonalEntries(Mat mat, IS *is)
10726: {
10727: PetscFunctionBegin;
10730: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10731: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10733: PetscUseTypeMethod(mat, findoffblockdiagonalentries, is);
10734: PetscFunctionReturn(PETSC_SUCCESS);
10735: }
10737: /*@C
10738: MatInvertBlockDiagonal - Inverts the block diagonal entries.
10740: Collective; No Fortran Support
10742: Input Parameter:
10743: . mat - the matrix
10745: Output Parameter:
10746: . values - the block inverses in column major order (FORTRAN-like)
10748: Level: advanced
10750: Notes:
10751: The size of the blocks is determined by the block size of the matrix.
10753: The blocks never overlap between two MPI processes, use `MatInvertVariableBlockEnvelope()` for that case
10755: The blocks all have the same size, use `MatInvertVariableBlockDiagonal()` for variable block size
10757: .seealso: [](ch_matrices), `Mat`, `MatInvertVariableBlockEnvelope()`, `MatInvertBlockDiagonalMat()`
10758: @*/
10759: PetscErrorCode MatInvertBlockDiagonal(Mat mat, const PetscScalar *values[])
10760: {
10761: PetscFunctionBegin;
10763: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10764: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10765: PetscUseTypeMethod(mat, invertblockdiagonal, values);
10766: PetscFunctionReturn(PETSC_SUCCESS);
10767: }
10769: /*@
10770: MatInvertVariableBlockDiagonal - Inverts the point block diagonal entries.
10772: Collective; No Fortran Support
10774: Input Parameters:
10775: + mat - the matrix
10776: . nblocks - the number of blocks on the process, set with `MatSetVariableBlockSizes()`
10777: - bsizes - the size of each block on the process, set with `MatSetVariableBlockSizes()`
10779: Output Parameter:
10780: . values - the block inverses in column major order (FORTRAN-like)
10782: Level: advanced
10784: Notes:
10785: Use `MatInvertBlockDiagonal()` if all blocks have the same size
10787: The blocks never overlap between two MPI processes, use `MatInvertVariableBlockEnvelope()` for that case
10789: .seealso: [](ch_matrices), `Mat`, `MatInvertBlockDiagonal()`, `MatSetVariableBlockSizes()`, `MatInvertVariableBlockEnvelope()`
10790: @*/
10791: PetscErrorCode MatInvertVariableBlockDiagonal(Mat mat, PetscInt nblocks, const PetscInt bsizes[], PetscScalar values[])
10792: {
10793: PetscFunctionBegin;
10795: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10796: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10797: PetscUseTypeMethod(mat, invertvariableblockdiagonal, nblocks, bsizes, values);
10798: PetscFunctionReturn(PETSC_SUCCESS);
10799: }
10801: /*@
10802: MatInvertBlockDiagonalMat - set the values of matrix C to be the inverted block diagonal of matrix A
10804: Collective
10806: Input Parameters:
10807: + A - the matrix
10808: - C - matrix with inverted block diagonal of `A`. This matrix should be created and may have its type set.
10810: Level: advanced
10812: Note:
10813: The blocksize of the matrix is used to determine the blocks on the diagonal of `C`
10815: .seealso: [](ch_matrices), `Mat`, `MatInvertBlockDiagonal()`
10816: @*/
10817: PetscErrorCode MatInvertBlockDiagonalMat(Mat A, Mat C)
10818: {
10819: const PetscScalar *vals;
10820: PetscInt *dnnz;
10821: PetscInt m, rstart, rend, bs, i, j;
10823: PetscFunctionBegin;
10824: PetscCall(MatInvertBlockDiagonal(A, &vals));
10825: PetscCall(MatGetBlockSize(A, &bs));
10826: PetscCall(MatGetLocalSize(A, &m, NULL));
10827: PetscCall(MatSetLayouts(C, A->rmap, A->cmap));
10828: PetscCall(PetscMalloc1(m / bs, &dnnz));
10829: for (j = 0; j < m / bs; j++) dnnz[j] = 1;
10830: PetscCall(MatXAIJSetPreallocation(C, bs, dnnz, NULL, NULL, NULL));
10831: PetscCall(PetscFree(dnnz));
10832: PetscCall(MatGetOwnershipRange(C, &rstart, &rend));
10833: PetscCall(MatSetOption(C, MAT_ROW_ORIENTED, PETSC_FALSE));
10834: for (i = rstart / bs; i < rend / bs; i++) PetscCall(MatSetValuesBlocked(C, 1, &i, 1, &i, &vals[(i - rstart / bs) * bs * bs], INSERT_VALUES));
10835: PetscCall(MatAssemblyBegin(C, MAT_FINAL_ASSEMBLY));
10836: PetscCall(MatAssemblyEnd(C, MAT_FINAL_ASSEMBLY));
10837: PetscCall(MatSetOption(C, MAT_ROW_ORIENTED, PETSC_TRUE));
10838: PetscFunctionReturn(PETSC_SUCCESS);
10839: }
10841: /*@C
10842: MatTransposeColoringDestroy - Destroys a coloring context for matrix product $C = A*B^T$ that was created
10843: via `MatTransposeColoringCreate()`.
10845: Collective
10847: Input Parameter:
10848: . c - coloring context
10850: Level: intermediate
10852: .seealso: [](ch_matrices), `Mat`, `MatTransposeColoringCreate()`
10853: @*/
10854: PetscErrorCode MatTransposeColoringDestroy(MatTransposeColoring *c)
10855: {
10856: MatTransposeColoring matcolor = *c;
10858: PetscFunctionBegin;
10859: if (!matcolor) PetscFunctionReturn(PETSC_SUCCESS);
10860: if (--((PetscObject)matcolor)->refct > 0) {
10861: matcolor = NULL;
10862: PetscFunctionReturn(PETSC_SUCCESS);
10863: }
10865: PetscCall(PetscFree3(matcolor->ncolumns, matcolor->nrows, matcolor->colorforrow));
10866: PetscCall(PetscFree(matcolor->rows));
10867: PetscCall(PetscFree(matcolor->den2sp));
10868: PetscCall(PetscFree(matcolor->colorforcol));
10869: PetscCall(PetscFree(matcolor->columns));
10870: if (matcolor->brows > 0) PetscCall(PetscFree(matcolor->lstart));
10871: PetscCall(PetscHeaderDestroy(c));
10872: PetscFunctionReturn(PETSC_SUCCESS);
10873: }
10875: /*@
10876: MatTransColoringApplySpToDen - Given a symbolic matrix product $C = A*B^T$ for which
10877: a `MatTransposeColoring` context has been created, computes a dense $B^T$ by applying
10878: `MatTransposeColoring` to sparse `B`.
10880: Collective
10882: Input Parameters:
10883: + coloring - coloring context created with `MatTransposeColoringCreate()`
10884: - B - sparse matrix
10886: Output Parameter:
10887: . Btdense - dense matrix $B^T$
10889: Level: developer
10891: Note:
10892: These are used internally for some implementations of `MatRARt()`
10894: .seealso: [](ch_matrices), `Mat`, `MatTransposeColoringCreate()`, `MatTransposeColoringDestroy()`, `MatTransColoringApplyDenToSp()`
10895: @*/
10896: PetscErrorCode MatTransColoringApplySpToDen(MatTransposeColoring coloring, Mat B, Mat Btdense)
10897: {
10898: PetscFunctionBegin;
10903: PetscCall((*B->ops->transcoloringapplysptoden)(coloring, B, Btdense));
10904: PetscFunctionReturn(PETSC_SUCCESS);
10905: }
10907: /*@
10908: MatTransColoringApplyDenToSp - Given a symbolic matrix product $C_{sp} = A*B^T$ for which
10909: a `MatTransposeColoring` context has been created and a dense matrix $C_{den} = A*B^T_{dense}$
10910: in which `B^T_{dens}` is obtained from `MatTransColoringApplySpToDen()`, recover sparse matrix
10911: $C_{sp}$ from $C_{den}$.
10913: Collective
10915: Input Parameters:
10916: + matcoloring - coloring context created with `MatTransposeColoringCreate()`
10917: - Cden - matrix product of a sparse matrix and a dense matrix Btdense
10919: Output Parameter:
10920: . Csp - sparse matrix
10922: Level: developer
10924: Note:
10925: These are used internally for some implementations of `MatRARt()`
10927: .seealso: [](ch_matrices), `Mat`, `MatTransposeColoringCreate()`, `MatTransposeColoringDestroy()`, `MatTransColoringApplySpToDen()`
10928: @*/
10929: PetscErrorCode MatTransColoringApplyDenToSp(MatTransposeColoring matcoloring, Mat Cden, Mat Csp)
10930: {
10931: PetscFunctionBegin;
10936: PetscCall((*Csp->ops->transcoloringapplydentosp)(matcoloring, Cden, Csp));
10937: PetscCall(MatAssemblyBegin(Csp, MAT_FINAL_ASSEMBLY));
10938: PetscCall(MatAssemblyEnd(Csp, MAT_FINAL_ASSEMBLY));
10939: PetscFunctionReturn(PETSC_SUCCESS);
10940: }
10942: /*@
10943: MatTransposeColoringCreate - Creates a matrix coloring context for the matrix product $C = A*B^T$.
10945: Collective
10947: Input Parameters:
10948: + mat - the matrix product C
10949: - iscoloring - the coloring of the matrix; usually obtained with `MatColoringCreate()` or `DMCreateColoring()`
10951: Output Parameter:
10952: . color - the new coloring context
10954: Level: intermediate
10956: .seealso: [](ch_matrices), `Mat`, `MatTransposeColoringDestroy()`, `MatTransColoringApplySpToDen()`,
10957: `MatTransColoringApplyDenToSp()`
10958: @*/
10959: PetscErrorCode MatTransposeColoringCreate(Mat mat, ISColoring iscoloring, MatTransposeColoring *color)
10960: {
10961: MatTransposeColoring c;
10962: MPI_Comm comm;
10964: PetscFunctionBegin;
10965: PetscCall(PetscLogEventBegin(MAT_TransposeColoringCreate, mat, 0, 0, 0));
10966: PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
10967: PetscCall(PetscHeaderCreate(c, MAT_TRANSPOSECOLORING_CLASSID, "MatTransposeColoring", "Matrix product C=A*B^T via coloring", "Mat", comm, MatTransposeColoringDestroy, NULL));
10969: c->ctype = iscoloring->ctype;
10970: PetscUseTypeMethod(mat, transposecoloringcreate, iscoloring, c);
10972: *color = c;
10973: PetscCall(PetscLogEventEnd(MAT_TransposeColoringCreate, mat, 0, 0, 0));
10974: PetscFunctionReturn(PETSC_SUCCESS);
10975: }
10977: /*@
10978: MatGetNonzeroState - Returns a 64-bit integer representing the current state of nonzeros in the matrix. If the
10979: matrix has had no new nonzero locations added to (or removed from) the matrix since the previous call then the value will be the
10980: same, otherwise it will be larger
10982: Not Collective
10984: Input Parameter:
10985: . mat - the matrix
10987: Output Parameter:
10988: . state - the current state
10990: Level: intermediate
10992: Notes:
10993: You can only compare states from two different calls to the SAME matrix, you cannot compare calls between
10994: different matrices
10996: Use `PetscObjectStateGet()` to check for changes to the numerical values in a matrix
10998: Use the result of `PetscObjectGetId()` to compare if a previously checked matrix is the same as the current matrix, do not compare object pointers.
11000: .seealso: [](ch_matrices), `Mat`, `PetscObjectStateGet()`, `PetscObjectGetId()`
11001: @*/
11002: PetscErrorCode MatGetNonzeroState(Mat mat, PetscObjectState *state)
11003: {
11004: PetscFunctionBegin;
11006: *state = mat->nonzerostate;
11007: PetscFunctionReturn(PETSC_SUCCESS);
11008: }
11010: /*@
11011: MatCreateMPIMatConcatenateSeqMat - Creates a single large PETSc matrix by concatenating sequential
11012: matrices from each processor
11014: Collective
11016: Input Parameters:
11017: + comm - the communicators the parallel matrix will live on
11018: . seqmat - the input sequential matrices
11019: . n - number of local columns (or `PETSC_DECIDE`)
11020: - reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
11022: Output Parameter:
11023: . mpimat - the parallel matrix generated
11025: Level: developer
11027: Note:
11028: The number of columns of the matrix in EACH processor MUST be the same.
11030: .seealso: [](ch_matrices), `Mat`
11031: @*/
11032: PetscErrorCode MatCreateMPIMatConcatenateSeqMat(MPI_Comm comm, Mat seqmat, PetscInt n, MatReuse reuse, Mat *mpimat)
11033: {
11034: PetscMPIInt size;
11036: PetscFunctionBegin;
11037: PetscCallMPI(MPI_Comm_size(comm, &size));
11038: if (size == 1) {
11039: if (reuse == MAT_INITIAL_MATRIX) {
11040: PetscCall(MatDuplicate(seqmat, MAT_COPY_VALUES, mpimat));
11041: } else {
11042: PetscCall(MatCopy(seqmat, *mpimat, SAME_NONZERO_PATTERN));
11043: }
11044: PetscFunctionReturn(PETSC_SUCCESS);
11045: }
11047: 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");
11049: PetscCall(PetscLogEventBegin(MAT_Merge, seqmat, 0, 0, 0));
11050: PetscCall((*seqmat->ops->creatempimatconcatenateseqmat)(comm, seqmat, n, reuse, mpimat));
11051: PetscCall(PetscLogEventEnd(MAT_Merge, seqmat, 0, 0, 0));
11052: PetscFunctionReturn(PETSC_SUCCESS);
11053: }
11055: /*@
11056: MatSubdomainsCreateCoalesce - Creates index subdomains by coalescing adjacent MPI processes' ownership ranges.
11058: Collective
11060: Input Parameters:
11061: + A - the matrix to create subdomains from
11062: - N - requested number of subdomains
11064: Output Parameters:
11065: + n - number of subdomains resulting on this MPI process
11066: - iss - `IS` list with indices of subdomains on this MPI process
11068: Level: advanced
11070: Note:
11071: The number of subdomains must be smaller than the communicator size
11073: .seealso: [](ch_matrices), `Mat`, `IS`
11074: @*/
11075: PetscErrorCode MatSubdomainsCreateCoalesce(Mat A, PetscInt N, PetscInt *n, IS *iss[])
11076: {
11077: MPI_Comm comm, subcomm;
11078: PetscMPIInt size, rank, color;
11079: PetscInt rstart, rend, k;
11081: PetscFunctionBegin;
11082: PetscCall(PetscObjectGetComm((PetscObject)A, &comm));
11083: PetscCallMPI(MPI_Comm_size(comm, &size));
11084: PetscCallMPI(MPI_Comm_rank(comm, &rank));
11085: 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);
11086: *n = 1;
11087: k = ((PetscInt)size) / N + ((PetscInt)size % N > 0); /* There are up to k ranks to a color */
11088: color = rank / k;
11089: PetscCallMPI(MPI_Comm_split(comm, color, rank, &subcomm));
11090: PetscCall(PetscMalloc1(1, iss));
11091: PetscCall(MatGetOwnershipRange(A, &rstart, &rend));
11092: PetscCall(ISCreateStride(subcomm, rend - rstart, rstart, 1, iss[0]));
11093: PetscCallMPI(MPI_Comm_free(&subcomm));
11094: PetscFunctionReturn(PETSC_SUCCESS);
11095: }
11097: /*@
11098: MatGalerkin - Constructs the coarse grid problem matrix via Galerkin projection.
11100: If the interpolation and restriction operators are the same, uses `MatPtAP()`.
11101: If they are not the same, uses `MatMatMatMult()`.
11103: Once the coarse grid problem is constructed, correct for interpolation operators
11104: that are not of full rank, which can legitimately happen in the case of non-nested
11105: geometric multigrid.
11107: Input Parameters:
11108: + restrct - restriction operator
11109: . dA - fine grid matrix
11110: . interpolate - interpolation operator
11111: . reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
11112: - fill - expected fill, use `PETSC_DEFAULT` if you do not have a good estimate
11114: Output Parameter:
11115: . A - the Galerkin coarse matrix
11117: Options Database Key:
11118: . -pc_mg_galerkin <both,pmat,mat,none> - for what matrices the Galerkin process should be used
11120: Level: developer
11122: .seealso: [](ch_matrices), `Mat`, `MatPtAP()`, `MatMatMatMult()`
11123: @*/
11124: PetscErrorCode MatGalerkin(Mat restrct, Mat dA, Mat interpolate, MatReuse reuse, PetscReal fill, Mat *A)
11125: {
11126: IS zerorows;
11127: Vec diag;
11129: PetscFunctionBegin;
11130: PetscCheck(reuse != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
11131: /* Construct the coarse grid matrix */
11132: if (interpolate == restrct) {
11133: PetscCall(MatPtAP(dA, interpolate, reuse, fill, A));
11134: } else {
11135: PetscCall(MatMatMatMult(restrct, dA, interpolate, reuse, fill, A));
11136: }
11138: /* If the interpolation matrix is not of full rank, A will have zero rows.
11139: This can legitimately happen in the case of non-nested geometric multigrid.
11140: In that event, we set the rows of the matrix to the rows of the identity,
11141: ignoring the equations (as the RHS will also be zero). */
11143: PetscCall(MatFindZeroRows(*A, &zerorows));
11145: if (zerorows != NULL) { /* if there are any zero rows */
11146: PetscCall(MatCreateVecs(*A, &diag, NULL));
11147: PetscCall(MatGetDiagonal(*A, diag));
11148: PetscCall(VecISSet(diag, zerorows, 1.0));
11149: PetscCall(MatDiagonalSet(*A, diag, INSERT_VALUES));
11150: PetscCall(VecDestroy(&diag));
11151: PetscCall(ISDestroy(&zerorows));
11152: }
11153: PetscFunctionReturn(PETSC_SUCCESS);
11154: }
11156: /*@C
11157: MatSetOperation - Allows user to set a matrix operation for any matrix type
11159: Logically Collective
11161: Input Parameters:
11162: + mat - the matrix
11163: . op - the name of the operation
11164: - f - the function that provides the operation
11166: Level: developer
11168: Example Usage:
11169: .vb
11170: extern PetscErrorCode usermult(Mat, Vec, Vec);
11172: PetscCall(MatCreateXXX(comm, ..., &A));
11173: PetscCall(MatSetOperation(A, MATOP_MULT, (PetscVoidFn *)usermult));
11174: .ve
11176: Notes:
11177: See the file `include/petscmat.h` for a complete list of matrix
11178: operations, which all have the form MATOP_<OPERATION>, where
11179: <OPERATION> is the name (in all capital letters) of the
11180: user interface routine (e.g., `MatMult()` -> `MATOP_MULT`).
11182: All user-provided functions (except for `MATOP_DESTROY`) should have the same calling
11183: sequence as the usual matrix interface routines, since they
11184: are intended to be accessed via the usual matrix interface
11185: routines, e.g.,
11186: .vb
11187: MatMult(Mat, Vec, Vec) -> usermult(Mat, Vec, Vec)
11188: .ve
11190: In particular each function MUST return `PETSC_SUCCESS` on success and
11191: nonzero on failure.
11193: This routine is distinct from `MatShellSetOperation()` in that it can be called on any matrix type.
11195: .seealso: [](ch_matrices), `Mat`, `MatGetOperation()`, `MatCreateShell()`, `MatShellSetContext()`, `MatShellSetOperation()`
11196: @*/
11197: PetscErrorCode MatSetOperation(Mat mat, MatOperation op, void (*f)(void))
11198: {
11199: PetscFunctionBegin;
11201: if (op == MATOP_VIEW && !mat->ops->viewnative && f != (void (*)(void))mat->ops->view) mat->ops->viewnative = mat->ops->view;
11202: (((void (**)(void))mat->ops)[op]) = f;
11203: PetscFunctionReturn(PETSC_SUCCESS);
11204: }
11206: /*@C
11207: MatGetOperation - Gets a matrix operation for any matrix type.
11209: Not Collective
11211: Input Parameters:
11212: + mat - the matrix
11213: - op - the name of the operation
11215: Output Parameter:
11216: . f - the function that provides the operation
11218: Level: developer
11220: Example Usage:
11221: .vb
11222: PetscErrorCode (*usermult)(Mat, Vec, Vec);
11224: MatGetOperation(A, MATOP_MULT, (void (**)(void))&usermult);
11225: .ve
11227: Notes:
11228: See the file include/petscmat.h for a complete list of matrix
11229: operations, which all have the form MATOP_<OPERATION>, where
11230: <OPERATION> is the name (in all capital letters) of the
11231: user interface routine (e.g., `MatMult()` -> `MATOP_MULT`).
11233: This routine is distinct from `MatShellGetOperation()` in that it can be called on any matrix type.
11235: .seealso: [](ch_matrices), `Mat`, `MatSetOperation()`, `MatCreateShell()`, `MatShellGetContext()`, `MatShellGetOperation()`
11236: @*/
11237: PetscErrorCode MatGetOperation(Mat mat, MatOperation op, void (**f)(void))
11238: {
11239: PetscFunctionBegin;
11241: *f = (((void (**)(void))mat->ops)[op]);
11242: PetscFunctionReturn(PETSC_SUCCESS);
11243: }
11245: /*@
11246: MatHasOperation - Determines whether the given matrix supports the particular operation.
11248: Not Collective
11250: Input Parameters:
11251: + mat - the matrix
11252: - op - the operation, for example, `MATOP_GET_DIAGONAL`
11254: Output Parameter:
11255: . has - either `PETSC_TRUE` or `PETSC_FALSE`
11257: Level: advanced
11259: Note:
11260: See `MatSetOperation()` for additional discussion on naming convention and usage of `op`.
11262: .seealso: [](ch_matrices), `Mat`, `MatCreateShell()`, `MatGetOperation()`, `MatSetOperation()`
11263: @*/
11264: PetscErrorCode MatHasOperation(Mat mat, MatOperation op, PetscBool *has)
11265: {
11266: PetscFunctionBegin;
11268: PetscAssertPointer(has, 3);
11269: if (mat->ops->hasoperation) {
11270: PetscUseTypeMethod(mat, hasoperation, op, has);
11271: } else {
11272: if (((void **)mat->ops)[op]) *has = PETSC_TRUE;
11273: else {
11274: *has = PETSC_FALSE;
11275: if (op == MATOP_CREATE_SUBMATRIX) {
11276: PetscMPIInt size;
11278: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
11279: if (size == 1) PetscCall(MatHasOperation(mat, MATOP_CREATE_SUBMATRICES, has));
11280: }
11281: }
11282: }
11283: PetscFunctionReturn(PETSC_SUCCESS);
11284: }
11286: /*@
11287: MatHasCongruentLayouts - Determines whether the rows and columns layouts of the matrix are congruent
11289: Collective
11291: Input Parameter:
11292: . mat - the matrix
11294: Output Parameter:
11295: . cong - either `PETSC_TRUE` or `PETSC_FALSE`
11297: Level: beginner
11299: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatSetSizes()`, `PetscLayout`
11300: @*/
11301: PetscErrorCode MatHasCongruentLayouts(Mat mat, PetscBool *cong)
11302: {
11303: PetscFunctionBegin;
11306: PetscAssertPointer(cong, 2);
11307: if (!mat->rmap || !mat->cmap) {
11308: *cong = mat->rmap == mat->cmap ? PETSC_TRUE : PETSC_FALSE;
11309: PetscFunctionReturn(PETSC_SUCCESS);
11310: }
11311: if (mat->congruentlayouts == PETSC_DECIDE) { /* first time we compare rows and cols layouts */
11312: PetscCall(PetscLayoutSetUp(mat->rmap));
11313: PetscCall(PetscLayoutSetUp(mat->cmap));
11314: PetscCall(PetscLayoutCompare(mat->rmap, mat->cmap, cong));
11315: if (*cong) mat->congruentlayouts = 1;
11316: else mat->congruentlayouts = 0;
11317: } else *cong = mat->congruentlayouts ? PETSC_TRUE : PETSC_FALSE;
11318: PetscFunctionReturn(PETSC_SUCCESS);
11319: }
11321: PetscErrorCode MatSetInf(Mat A)
11322: {
11323: PetscFunctionBegin;
11324: PetscUseTypeMethod(A, setinf);
11325: PetscFunctionReturn(PETSC_SUCCESS);
11326: }
11328: /*@
11329: 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
11330: and possibly removes small values from the graph structure.
11332: Collective
11334: Input Parameters:
11335: + A - the matrix
11336: . sym - `PETSC_TRUE` indicates that the graph should be symmetrized
11337: . scale - `PETSC_TRUE` indicates that the graph edge weights should be symmetrically scaled with the diagonal entry
11338: . filter - filter value - < 0: does nothing; == 0: removes only 0.0 entries; otherwise: removes entries with abs(entries) <= value
11339: . num_idx - size of 'index' array
11340: - index - array of block indices to use for graph strength of connection weight
11342: Output Parameter:
11343: . graph - the resulting graph
11345: Level: advanced
11347: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `PCGAMG`
11348: @*/
11349: PetscErrorCode MatCreateGraph(Mat A, PetscBool sym, PetscBool scale, PetscReal filter, PetscInt num_idx, PetscInt index[], Mat *graph)
11350: {
11351: PetscFunctionBegin;
11355: PetscAssertPointer(graph, 7);
11356: PetscUseTypeMethod(A, creategraph, sym, scale, filter, num_idx, index, graph);
11357: PetscFunctionReturn(PETSC_SUCCESS);
11358: }
11360: /*@
11361: MatEliminateZeros - eliminate the nondiagonal zero entries in place from the nonzero structure of a sparse `Mat` in place,
11362: meaning the same memory is used for the matrix, and no new memory is allocated.
11364: Collective
11366: Input Parameters:
11367: + A - the matrix
11368: - 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
11370: Level: intermediate
11372: Developer Note:
11373: The entries in the sparse matrix data structure are shifted to fill in the unneeded locations in the data. Thus the end
11374: of the arrays in the data structure are unneeded.
11376: .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatCreateGraph()`, `MatFilter()`
11377: @*/
11378: PetscErrorCode MatEliminateZeros(Mat A, PetscBool keep)
11379: {
11380: PetscFunctionBegin;
11382: PetscUseTypeMethod(A, eliminatezeros, keep);
11383: PetscFunctionReturn(PETSC_SUCCESS);
11384: }