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_Mults, 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_Getsymtranspose, MAT_Getsymtransreduced, MAT_GetBrowsOfAcols;
34: PetscLogEvent MAT_GetBrowsOfAocols, MAT_Getlocalmat, MAT_Getlocalmatcondensed, MAT_Seqstompi, MAT_Seqstompinum, MAT_Seqstompisym;
35: PetscLogEvent MAT_Applypapt, MAT_Applypapt_numeric, MAT_Applypapt_symbolic, MAT_GetSequentialNonzeroStructure;
36: PetscLogEvent MAT_GetMultiProcBlock;
37: PetscLogEvent MAT_CUSPARSECopyToGPU, MAT_CUSPARSECopyFromGPU, MAT_CUSPARSEGenerateTranspose, MAT_CUSPARSESolveAnalysis;
38: PetscLogEvent MAT_HIPSPARSECopyToGPU, MAT_HIPSPARSECopyFromGPU, MAT_HIPSPARSEGenerateTranspose, MAT_HIPSPARSESolveAnalysis;
39: PetscLogEvent MAT_PreallCOO, MAT_SetVCOO;
40: PetscLogEvent MAT_SetValuesBatch;
41: PetscLogEvent MAT_ViennaCLCopyToGPU;
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 locations it fills the locations with random numbers.
74: It generates an error if used on sparse matrices that have not been preallocated.
76: .seealso: [](chapter_matrices), `Mat`, `PetscRandom`, `PetscRandomCreate()`, `MatZeroEntries()`, `MatSetValues()`, `PetscRandomCreate()`, `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: [](chapter_matrices), `Mat`, `MatZeroEntries()`, `MatFactor()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`, `MatFactorClearError()`, `MatFactorGetErrorZeroPivot()`,
129: `MAT_FACTOR_NUMERIC_ZEROPIVOT`
130: @*/
131: PetscErrorCode MatFactorGetErrorZeroPivot(Mat mat, PetscReal *pivot, PetscInt *row)
132: {
133: PetscFunctionBegin;
137: *pivot = mat->factorerror_zeropivot_value;
138: *row = mat->factorerror_zeropivot_row;
139: PetscFunctionReturn(PETSC_SUCCESS);
140: }
142: /*@
143: MatFactorGetError - gets the error code from a factorization
145: Logically Collective
147: Input Parameter:
148: . mat - the factored matrix
150: Output Parameter:
151: . err - the error code
153: Level: advanced
155: Note:
156: This can also be called on non-factored matrices that come from, for example, matrices used in SOR.
158: .seealso: [](chapter_matrices), `Mat`, `MatZeroEntries()`, `MatFactor()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`,
159: `MatFactorClearError()`, `MatFactorGetErrorZeroPivot()`, `MatFactorError`
160: @*/
161: PetscErrorCode MatFactorGetError(Mat mat, MatFactorError *err)
162: {
163: PetscFunctionBegin;
166: *err = mat->factorerrortype;
167: PetscFunctionReturn(PETSC_SUCCESS);
168: }
170: /*@
171: MatFactorClearError - clears the error code in a factorization
173: Logically Collective
175: Input Parameter:
176: . mat - the factored matrix
178: Level: developer
180: Note:
181: This can also be called on non-factored matrices that come from, for example, matrices used in SOR.
183: .seealso: [](chapter_matrices), `Mat`, `MatZeroEntries()`, `MatFactor()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`, `MatFactorGetError()`, `MatFactorGetErrorZeroPivot()`,
184: `MatGetErrorCode()`, `MatFactorError`
185: @*/
186: PetscErrorCode MatFactorClearError(Mat mat)
187: {
188: PetscFunctionBegin;
190: mat->factorerrortype = MAT_FACTOR_NOERROR;
191: mat->factorerror_zeropivot_value = 0.0;
192: mat->factorerror_zeropivot_row = 0;
193: PetscFunctionReturn(PETSC_SUCCESS);
194: }
196: PETSC_INTERN PetscErrorCode MatFindNonzeroRowsOrCols_Basic(Mat mat, PetscBool cols, PetscReal tol, IS *nonzero)
197: {
198: Vec r, l;
199: const PetscScalar *al;
200: PetscInt i, nz, gnz, N, n;
202: PetscFunctionBegin;
203: PetscCall(MatCreateVecs(mat, &r, &l));
204: if (!cols) { /* nonzero rows */
205: PetscCall(MatGetSize(mat, &N, NULL));
206: PetscCall(MatGetLocalSize(mat, &n, NULL));
207: PetscCall(VecSet(l, 0.0));
208: PetscCall(VecSetRandom(r, NULL));
209: PetscCall(MatMult(mat, r, l));
210: PetscCall(VecGetArrayRead(l, &al));
211: } else { /* nonzero columns */
212: PetscCall(MatGetSize(mat, NULL, &N));
213: PetscCall(MatGetLocalSize(mat, NULL, &n));
214: PetscCall(VecSet(r, 0.0));
215: PetscCall(VecSetRandom(l, NULL));
216: PetscCall(MatMultTranspose(mat, l, r));
217: PetscCall(VecGetArrayRead(r, &al));
218: }
219: if (tol <= 0.0) {
220: for (i = 0, nz = 0; i < n; i++)
221: if (al[i] != 0.0) nz++;
222: } else {
223: for (i = 0, nz = 0; i < n; i++)
224: if (PetscAbsScalar(al[i]) > tol) nz++;
225: }
226: PetscCall(MPIU_Allreduce(&nz, &gnz, 1, MPIU_INT, MPI_SUM, PetscObjectComm((PetscObject)mat)));
227: if (gnz != N) {
228: PetscInt *nzr;
229: PetscCall(PetscMalloc1(nz, &nzr));
230: if (nz) {
231: if (tol < 0) {
232: for (i = 0, nz = 0; i < n; i++)
233: if (al[i] != 0.0) nzr[nz++] = i;
234: } else {
235: for (i = 0, nz = 0; i < n; i++)
236: if (PetscAbsScalar(al[i]) > tol) nzr[nz++] = i;
237: }
238: }
239: PetscCall(ISCreateGeneral(PetscObjectComm((PetscObject)mat), nz, nzr, PETSC_OWN_POINTER, nonzero));
240: } else *nonzero = NULL;
241: if (!cols) { /* nonzero rows */
242: PetscCall(VecRestoreArrayRead(l, &al));
243: } else {
244: PetscCall(VecRestoreArrayRead(r, &al));
245: }
246: PetscCall(VecDestroy(&l));
247: PetscCall(VecDestroy(&r));
248: PetscFunctionReturn(PETSC_SUCCESS);
249: }
251: /*@
252: MatFindNonzeroRows - Locate all rows that are not completely zero in the matrix
254: Input Parameter:
255: . A - the matrix
257: Output Parameter:
258: . keptrows - the rows that are not completely zero
260: Level: intermediate
262: Note:
263: `keptrows` is set to `NULL` if all rows are nonzero.
265: .seealso: [](chapter_matrices), `Mat`, `MatFindZeroRows()`
266: @*/
267: PetscErrorCode MatFindNonzeroRows(Mat mat, IS *keptrows)
268: {
269: PetscFunctionBegin;
273: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
274: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
275: if (mat->ops->findnonzerorows) PetscUseTypeMethod(mat, findnonzerorows, keptrows);
276: else PetscCall(MatFindNonzeroRowsOrCols_Basic(mat, PETSC_FALSE, 0.0, keptrows));
277: PetscFunctionReturn(PETSC_SUCCESS);
278: }
280: /*@
281: MatFindZeroRows - Locate all rows that are completely zero in the matrix
283: Input Parameter:
284: . A - the matrix
286: Output Parameter:
287: . zerorows - the rows that are completely zero
289: Level: intermediate
291: Note:
292: `zerorows` is set to `NULL` if no rows are zero.
294: .seealso: [](chapter_matrices), `Mat`, `MatFindNonzeroRows()`
295: @*/
296: PetscErrorCode MatFindZeroRows(Mat mat, IS *zerorows)
297: {
298: IS keptrows;
299: PetscInt m, n;
301: PetscFunctionBegin;
305: PetscCall(MatFindNonzeroRows(mat, &keptrows));
306: /* MatFindNonzeroRows sets keptrows to NULL if there are no zero rows.
307: In keeping with this convention, we set zerorows to NULL if there are no zero
308: rows. */
309: if (keptrows == NULL) {
310: *zerorows = NULL;
311: } else {
312: PetscCall(MatGetOwnershipRange(mat, &m, &n));
313: PetscCall(ISComplement(keptrows, m, n, zerorows));
314: PetscCall(ISDestroy(&keptrows));
315: }
316: PetscFunctionReturn(PETSC_SUCCESS);
317: }
319: /*@
320: MatGetDiagonalBlock - Returns the part of the matrix associated with the on-process coupling
322: Not Collective
324: Input Parameter:
325: . A - the matrix
327: Output Parameter:
328: . a - the diagonal part (which is a SEQUENTIAL matrix)
330: Level: advanced
332: Notes:
333: See `MatCreateAIJ()` for more information on the "diagonal part" of the matrix.
335: Use caution, as the reference count on the returned matrix is not incremented and it is used as part of `A`'s normal operation.
337: .seealso: [](chapter_matrices), `Mat`, `MatCreateAIJ()`, `MATAIJ`, `MATBAIJ`, `MATSBAIJ`
338: @*/
339: PetscErrorCode MatGetDiagonalBlock(Mat A, Mat *a)
340: {
341: PetscFunctionBegin;
345: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
346: if (A->ops->getdiagonalblock) PetscUseTypeMethod(A, getdiagonalblock, a);
347: else {
348: PetscMPIInt size;
350: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)A), &size));
351: PetscCheck(size == 1, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Not for parallel matrix type %s", ((PetscObject)A)->type_name);
352: *a = A;
353: }
354: PetscFunctionReturn(PETSC_SUCCESS);
355: }
357: /*@
358: MatGetTrace - Gets the trace of a matrix. The sum of the diagonal entries.
360: Collective
362: Input Parameter:
363: . mat - the matrix
365: Output Parameter:
366: . trace - the sum of the diagonal entries
368: Level: advanced
370: .seealso: [](chapter_matrices), `Mat`
371: @*/
372: PetscErrorCode MatGetTrace(Mat mat, PetscScalar *trace)
373: {
374: Vec diag;
376: PetscFunctionBegin;
379: PetscCall(MatCreateVecs(mat, &diag, NULL));
380: PetscCall(MatGetDiagonal(mat, diag));
381: PetscCall(VecSum(diag, trace));
382: PetscCall(VecDestroy(&diag));
383: PetscFunctionReturn(PETSC_SUCCESS);
384: }
386: /*@
387: MatRealPart - Zeros out the imaginary part of the matrix
389: Logically Collective
391: Input Parameter:
392: . mat - the matrix
394: Level: advanced
396: .seealso: [](chapter_matrices), `Mat`, `MatImaginaryPart()`
397: @*/
398: PetscErrorCode MatRealPart(Mat mat)
399: {
400: PetscFunctionBegin;
403: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
404: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
405: MatCheckPreallocated(mat, 1);
406: PetscUseTypeMethod(mat, realpart);
407: PetscFunctionReturn(PETSC_SUCCESS);
408: }
410: /*@C
411: MatGetGhosts - Get the global indices of all ghost nodes defined by the sparse matrix
413: Collective
415: Input Parameter:
416: . mat - the matrix
418: Output Parameters:
419: + nghosts - number of ghosts (for `MATBAIJ` and `MATSBAIJ` matrices there is one ghost for each block)
420: - ghosts - the global indices of the ghost points
422: Level: advanced
424: Note:
425: `nghosts` and `ghosts` are suitable to pass into `VecCreateGhost()`
427: .seealso: [](chapter_matrices), `Mat`, `VecCreateGhost()`
428: @*/
429: PetscErrorCode MatGetGhosts(Mat mat, PetscInt *nghosts, const PetscInt *ghosts[])
430: {
431: PetscFunctionBegin;
434: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
435: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
436: if (mat->ops->getghosts) PetscUseTypeMethod(mat, getghosts, nghosts, ghosts);
437: else {
438: if (nghosts) *nghosts = 0;
439: if (ghosts) *ghosts = NULL;
440: }
441: PetscFunctionReturn(PETSC_SUCCESS);
442: }
444: /*@
445: MatImaginaryPart - Moves the imaginary part of the matrix to the real part and zeros the imaginary part
447: Logically Collective
449: Input Parameter:
450: . mat - the matrix
452: Level: advanced
454: .seealso: [](chapter_matrices), `Mat`, `MatRealPart()`
455: @*/
456: PetscErrorCode MatImaginaryPart(Mat mat)
457: {
458: PetscFunctionBegin;
461: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
462: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
463: MatCheckPreallocated(mat, 1);
464: PetscUseTypeMethod(mat, imaginarypart);
465: PetscFunctionReturn(PETSC_SUCCESS);
466: }
468: /*@
469: MatMissingDiagonal - Determine if sparse matrix is missing a diagonal entry (or block entry for `MATBAIJ` and `MATSBAIJ` matrices)
471: Not Collective
473: Input Parameter:
474: . mat - the matrix
476: Output Parameters:
477: + missing - is any diagonal missing
478: - dd - first diagonal entry that is missing (optional) on this process
480: Level: advanced
482: .seealso: [](chapter_matrices), `Mat`
483: @*/
484: PetscErrorCode MatMissingDiagonal(Mat mat, PetscBool *missing, PetscInt *dd)
485: {
486: PetscFunctionBegin;
490: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix %s", ((PetscObject)mat)->type_name);
491: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
492: PetscUseTypeMethod(mat, missingdiagonal, missing, dd);
493: PetscFunctionReturn(PETSC_SUCCESS);
494: }
496: /*@C
497: MatGetRow - Gets a row of a matrix. You MUST call `MatRestoreRow()`
498: for each row that you get to ensure that your application does
499: not bleed memory.
501: Not Collective
503: Input Parameters:
504: + mat - the matrix
505: - row - the row to get
507: Output Parameters:
508: + ncols - if not `NULL`, the number of nonzeros in the row
509: . cols - if not `NULL`, the column numbers
510: - vals - if not `NULL`, the values
512: Level: advanced
514: Notes:
515: This routine is provided for people who need to have direct access
516: to the structure of a matrix. We hope that we provide enough
517: high-level matrix routines that few users will need it.
519: `MatGetRow()` always returns 0-based column indices, regardless of
520: whether the internal representation is 0-based (default) or 1-based.
522: For better efficiency, set cols and/or vals to `NULL` if you do
523: not wish to extract these quantities.
525: The user can only examine the values extracted with `MatGetRow()`;
526: the values cannot be altered. To change the matrix entries, one
527: must use `MatSetValues()`.
529: You can only have one call to `MatGetRow()` outstanding for a particular
530: matrix at a time, per processor. `MatGetRow()` can only obtain rows
531: associated with the given processor, it cannot get rows from the
532: other processors; for that we suggest using `MatCreateSubMatrices()`, then
533: MatGetRow() on the submatrix. The row index passed to `MatGetRow()`
534: is in the global number of rows.
536: Use `MatGetRowIJ()` and `MatRestoreRowIJ()` to access all the local indices of the sparse matrix.
538: Use `MatSeqAIJGetArray()` and similar functions to access the numerical values for certain matrix types directly.
540: Fortran Note:
541: The calling sequence is
542: .vb
543: MatGetRow(matrix,row,ncols,cols,values,ierr)
544: Mat matrix (input)
545: integer row (input)
546: integer ncols (output)
547: integer cols(maxcols) (output)
548: double precision (or double complex) values(maxcols) output
549: .ve
550: where maxcols >= maximum nonzeros in any row of the matrix.
552: Caution:
553: Do not try to change the contents of the output arrays (`cols` and `vals`).
554: In some cases, this may corrupt the matrix.
556: .seealso: [](chapter_matrices), `Mat`, `MatRestoreRow()`, `MatSetValues()`, `MatGetValues()`, `MatCreateSubMatrices()`, `MatGetDiagonal()`, `MatGetRowIJ()`, `MatRestoreRowIJ()`
557: @*/
558: PetscErrorCode MatGetRow(Mat mat, PetscInt row, PetscInt *ncols, const PetscInt *cols[], const PetscScalar *vals[])
559: {
560: PetscInt incols;
562: PetscFunctionBegin;
565: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
566: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
567: MatCheckPreallocated(mat, 1);
568: 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);
569: PetscCall(PetscLogEventBegin(MAT_GetRow, mat, 0, 0, 0));
570: PetscUseTypeMethod(mat, getrow, row, &incols, (PetscInt **)cols, (PetscScalar **)vals);
571: if (ncols) *ncols = incols;
572: PetscCall(PetscLogEventEnd(MAT_GetRow, mat, 0, 0, 0));
573: PetscFunctionReturn(PETSC_SUCCESS);
574: }
576: /*@
577: MatConjugate - replaces the matrix values with their complex conjugates
579: Logically Collective
581: Input Parameter:
582: . mat - the matrix
584: Level: advanced
586: .seealso: [](chapter_matrices), `Mat`, `MatRealPart()`, `MatImaginaryPart()`, `VecConjugate()`, `MatTranspose()`
587: @*/
588: PetscErrorCode MatConjugate(Mat mat)
589: {
590: PetscFunctionBegin;
592: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
593: if (PetscDefined(USE_COMPLEX) && mat->hermitian != PETSC_BOOL3_TRUE) {
594: PetscUseTypeMethod(mat, conjugate);
595: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
596: }
597: PetscFunctionReturn(PETSC_SUCCESS);
598: }
600: /*@C
601: MatRestoreRow - Frees any temporary space allocated by `MatGetRow()`.
603: Not Collective
605: Input Parameters:
606: + mat - the matrix
607: . row - the row to get
608: . ncols - the number of nonzeros
609: . cols - the columns of the nonzeros
610: - vals - if nonzero the column values
612: Level: advanced
614: Notes:
615: This routine should be called after you have finished examining the entries.
617: This routine zeros out `ncols`, `cols`, and `vals`. This is to prevent accidental
618: us of the array after it has been restored. If you pass `NULL`, it will
619: not zero the pointers. Use of `cols` or `vals` after `MatRestoreRow()` is invalid.
621: Fortran Notes:
622: The calling sequence is
623: .vb
624: MatRestoreRow(matrix,row,ncols,cols,values,ierr)
625: Mat matrix (input)
626: integer row (input)
627: integer ncols (output)
628: integer cols(maxcols) (output)
629: double precision (or double complex) values(maxcols) output
630: .ve
631: Where maxcols >= maximum nonzeros in any row of the matrix.
633: In Fortran `MatRestoreRow()` MUST be called after `MatGetRow()`
634: before another call to `MatGetRow()` can be made.
636: .seealso: [](chapter_matrices), `Mat`, `MatGetRow()`
637: @*/
638: PetscErrorCode MatRestoreRow(Mat mat, PetscInt row, PetscInt *ncols, const PetscInt *cols[], const PetscScalar *vals[])
639: {
640: PetscFunctionBegin;
643: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
644: if (!mat->ops->restorerow) PetscFunctionReturn(PETSC_SUCCESS);
645: PetscUseTypeMethod(mat, restorerow, row, ncols, (PetscInt **)cols, (PetscScalar **)vals);
646: if (ncols) *ncols = 0;
647: if (cols) *cols = NULL;
648: if (vals) *vals = NULL;
649: PetscFunctionReturn(PETSC_SUCCESS);
650: }
652: /*@
653: MatGetRowUpperTriangular - Sets a flag to enable calls to `MatGetRow()` for matrix in `MATSBAIJ` format.
654: You should call `MatRestoreRowUpperTriangular()` after calling` MatGetRow()` and `MatRestoreRow()` to disable the flag.
656: Not Collective
658: Input Parameter:
659: . mat - the matrix
661: Level: advanced
663: Note:
664: 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.
666: .seealso: [](chapter_matrices), `Mat`, `MATSBAIJ`, `MatRestoreRowUpperTriangular()`
667: @*/
668: PetscErrorCode MatGetRowUpperTriangular(Mat mat)
669: {
670: PetscFunctionBegin;
673: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
674: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
675: MatCheckPreallocated(mat, 1);
676: if (!mat->ops->getrowuppertriangular) PetscFunctionReturn(PETSC_SUCCESS);
677: PetscUseTypeMethod(mat, getrowuppertriangular);
678: PetscFunctionReturn(PETSC_SUCCESS);
679: }
681: /*@
682: MatRestoreRowUpperTriangular - Disable calls to `MatGetRow()` for matrix in `MATSBAIJ` format.
684: Not Collective
686: Input Parameter:
687: . mat - the matrix
689: Level: advanced
691: Note:
692: This routine should be called after you have finished calls to `MatGetRow()` and `MatRestoreRow()`.
694: .seealso: [](chapter_matrices), `Mat`, `MATSBAIJ`, `MatGetRowUpperTriangular()`
695: @*/
696: PetscErrorCode MatRestoreRowUpperTriangular(Mat mat)
697: {
698: PetscFunctionBegin;
701: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
702: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
703: MatCheckPreallocated(mat, 1);
704: if (!mat->ops->restorerowuppertriangular) PetscFunctionReturn(PETSC_SUCCESS);
705: PetscUseTypeMethod(mat, restorerowuppertriangular);
706: PetscFunctionReturn(PETSC_SUCCESS);
707: }
709: /*@C
710: MatSetOptionsPrefix - Sets the prefix used for searching for all
711: `Mat` options in the database.
713: Logically Collective
715: Input Parameters:
716: + A - the matrix
717: - prefix - the prefix to prepend to all option names
719: Level: advanced
721: Notes:
722: A hyphen (-) must NOT be given at the beginning of the prefix name.
723: The first character of all runtime options is AUTOMATICALLY the hyphen.
725: This is NOT used for options for the factorization of the matrix. Normally the
726: prefix is automatically passed in from the PC calling the factorization. To set
727: it directly use `MatSetOptionsPrefixFactor()`
729: .seealso: [](chapter_matrices), `Mat`, `MatSetFromOptions()`, `MatSetOptionsPrefixFactor()`
730: @*/
731: PetscErrorCode MatSetOptionsPrefix(Mat A, const char prefix[])
732: {
733: PetscFunctionBegin;
735: PetscCall(PetscObjectSetOptionsPrefix((PetscObject)A, prefix));
736: PetscFunctionReturn(PETSC_SUCCESS);
737: }
739: /*@C
740: MatSetOptionsPrefixFactor - Sets the prefix used for searching for all matrix factor options in the database for
741: for matrices created with `MatGetFactor()`
743: Logically Collective
745: Input Parameters:
746: + A - the matrix
747: - prefix - the prefix to prepend to all option names for the factored matrix
749: Level: developer
751: Notes:
752: A hyphen (-) must NOT be given at the beginning of the prefix name.
753: The first character of all runtime options is AUTOMATICALLY the hyphen.
755: Normally the prefix is automatically passed in from the `PC` calling the factorization. To set
756: it directly when not using `KSP`/`PC` use `MatSetOptionsPrefixFactor()`
758: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSetFromOptions()`, `MatSetOptionsPrefix()`, `MatAppendOptionsPrefixFactor()`
759: @*/
760: PetscErrorCode MatSetOptionsPrefixFactor(Mat A, const char prefix[])
761: {
762: PetscFunctionBegin;
764: if (prefix) {
766: PetscCheck(prefix[0] != '-', PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONG, "Options prefix should not begin with a hyphen");
767: if (prefix != A->factorprefix) {
768: PetscCall(PetscFree(A->factorprefix));
769: PetscCall(PetscStrallocpy(prefix, &A->factorprefix));
770: }
771: } else PetscCall(PetscFree(A->factorprefix));
772: PetscFunctionReturn(PETSC_SUCCESS);
773: }
775: /*@C
776: MatAppendOptionsPrefixFactor - Appends to the prefix used for searching for all matrix factor options in the database for
777: for matrices created with `MatGetFactor()`
779: Logically Collective
781: Input Parameters:
782: + A - the matrix
783: - prefix - the prefix to prepend to all option names for the factored matrix
785: Level: developer
787: Notes:
788: A hyphen (-) must NOT be given at the beginning of the prefix name.
789: The first character of all runtime options is AUTOMATICALLY the hyphen.
791: Normally the prefix is automatically passed in from the `PC` calling the factorization. To set
792: it directly when not using `KSP`/`PC` use `MatAppendOptionsPrefixFactor()`
794: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, PetscOptionsCreate()`, `PetscOptionsDestroy()`, `PetscObjectSetOptionsPrefix()`, `PetscObjectPrependOptionsPrefix()`,
795: `PetscObjectGetOptionsPrefix()`, `TSAppendOptionsPrefix()`, `SNESAppendOptionsPrefix()`, `KSPAppendOptionsPrefix()`, `MatSetOptionsPrefixFactor()`,
796: `MatSetOptionsPrefix()`
797: @*/
798: PetscErrorCode MatAppendOptionsPrefixFactor(Mat A, const char prefix[])
799: {
800: size_t len1, len2, new_len;
802: PetscFunctionBegin;
804: if (!prefix) PetscFunctionReturn(PETSC_SUCCESS);
805: if (!A->factorprefix) {
806: PetscCall(MatSetOptionsPrefixFactor(A, prefix));
807: PetscFunctionReturn(PETSC_SUCCESS);
808: }
809: PetscCheck(prefix[0] != '-', PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONG, "Options prefix should not begin with a hyphen");
811: PetscCall(PetscStrlen(A->factorprefix, &len1));
812: PetscCall(PetscStrlen(prefix, &len2));
813: new_len = len1 + len2 + 1;
814: PetscCall(PetscRealloc(new_len * sizeof(*(A->factorprefix)), &A->factorprefix));
815: PetscCall(PetscStrncpy(A->factorprefix + len1, prefix, len2 + 1));
816: PetscFunctionReturn(PETSC_SUCCESS);
817: }
819: /*@C
820: MatAppendOptionsPrefix - Appends to the prefix used for searching for all
821: matrix options in the database.
823: Logically Collective
825: Input Parameters:
826: + A - the matrix
827: - prefix - the prefix to prepend to all option names
829: Level: advanced
831: Note:
832: A hyphen (-) must NOT be given at the beginning of the prefix name.
833: The first character of all runtime options is AUTOMATICALLY the hyphen.
835: .seealso: [](chapter_matrices), `Mat`, `MatGetOptionsPrefix()`, `MatAppendOptionsPrefixFactor()`, `MatSetOptionsPrefix()`
836: @*/
837: PetscErrorCode MatAppendOptionsPrefix(Mat A, const char prefix[])
838: {
839: PetscFunctionBegin;
841: PetscCall(PetscObjectAppendOptionsPrefix((PetscObject)A, prefix));
842: PetscFunctionReturn(PETSC_SUCCESS);
843: }
845: /*@C
846: MatGetOptionsPrefix - Gets the prefix used for searching for all
847: matrix options in the database.
849: Not Collective
851: Input Parameter:
852: . A - the matrix
854: Output Parameter:
855: . prefix - pointer to the prefix string used
857: Level: advanced
859: Fortran Note:
860: The user should pass in a string `prefix` of
861: sufficient length to hold the prefix.
863: .seealso: [](chapter_matrices), `Mat`, `MatAppendOptionsPrefix()`, `MatSetOptionsPrefix()`, `MatAppendOptionsPrefixFactor()`, `MatSetOptionsPrefixFactor()`
864: @*/
865: PetscErrorCode MatGetOptionsPrefix(Mat A, const char *prefix[])
866: {
867: PetscFunctionBegin;
870: PetscCall(PetscObjectGetOptionsPrefix((PetscObject)A, prefix));
871: PetscFunctionReturn(PETSC_SUCCESS);
872: }
874: /*@
875: MatResetPreallocation - Reset matrix to use the original nonzero pattern provided by users.
877: Collective
879: Input Parameter:
880: . A - the matrix
882: Level: beginner
884: Notes:
885: The allocated memory will be shrunk after calling `MatAssemblyBegin()` and `MatAssemblyEnd()` with `MAT_FINAL_ASSEMBLY`.
887: Users can reset the preallocation to access the original memory.
889: Currently only supported for `MATAIJ` matrices.
891: .seealso: [](chapter_matrices), `Mat`, `MatSeqAIJSetPreallocation()`, `MatMPIAIJSetPreallocation()`, `MatXAIJSetPreallocation()`
892: @*/
893: PetscErrorCode MatResetPreallocation(Mat A)
894: {
895: PetscFunctionBegin;
898: PetscUseMethod(A, "MatResetPreallocation_C", (Mat), (A));
899: PetscFunctionReturn(PETSC_SUCCESS);
900: }
902: /*@
903: MatSetUp - Sets up the internal matrix data structures for later use.
905: Collective
907: Input Parameter:
908: . A - the matrix
910: Level: intermediate
912: Notes:
913: If the user has not set preallocation for this matrix then an efficient algorithm will be used for the first round of
914: setting values in the matrix.
916: If a suitable preallocation routine is used, this function does not need to be called.
918: This routine is called internally by other matrix functions when needed so rarely needs to be called by users
920: .seealso: [](chapter_matrices), `Mat`, `MatMult()`, `MatCreate()`, `MatDestroy()`, `MatXAIJSetPreallocation()`
921: @*/
922: PetscErrorCode MatSetUp(Mat A)
923: {
924: PetscFunctionBegin;
926: if (!((PetscObject)A)->type_name) {
927: PetscMPIInt size;
929: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)A), &size));
930: PetscCall(MatSetType(A, size == 1 ? MATSEQAIJ : MATMPIAIJ));
931: }
932: if (!A->preallocated) PetscTryTypeMethod(A, setup);
933: PetscCall(PetscLayoutSetUp(A->rmap));
934: PetscCall(PetscLayoutSetUp(A->cmap));
935: A->preallocated = PETSC_TRUE;
936: PetscFunctionReturn(PETSC_SUCCESS);
937: }
939: #if defined(PETSC_HAVE_SAWS)
940: #include <petscviewersaws.h>
941: #endif
943: /*@C
944: MatViewFromOptions - View properties of the matrix based on options set in the options database
946: Collective
948: Input Parameters:
949: + A - the matrix
950: . obj - optional additional object that provides the options prefix to use
951: - name - command line option
953: Options Database Key:
954: . -mat_view [viewertype]:... - the viewer and its options
956: Level: intermediate
958: Notes:
959: .vb
960: If no value is provided ascii:stdout is used
961: ascii[:[filename][:[format][:append]]] defaults to stdout - format can be one of ascii_info, ascii_info_detail, or ascii_matlab,
962: for example ascii::ascii_info prints just the information about the object not all details
963: unless :append is given filename opens in write mode, overwriting what was already there
964: binary[:[filename][:[format][:append]]] defaults to the file binaryoutput
965: draw[:drawtype[:filename]] for example, draw:tikz, draw:tikz:figure.tex or draw:x
966: socket[:port] defaults to the standard output port
967: saws[:communicatorname] publishes object to the Scientific Application Webserver (SAWs)
968: .ve
970: .seealso: [](chapter_matrices), `Mat`, `MatView()`, `PetscObjectViewFromOptions()`, `MatCreate()`
971: @*/
972: PetscErrorCode MatViewFromOptions(Mat A, PetscObject obj, const char name[])
973: {
974: PetscFunctionBegin;
976: PetscCall(PetscObjectViewFromOptions((PetscObject)A, obj, name));
977: PetscFunctionReturn(PETSC_SUCCESS);
978: }
980: /*@C
981: MatView - display information about a matrix in a variety ways
983: Collective
985: Input Parameters:
986: + mat - the matrix
987: - viewer - visualization context
989: Options Database Keys:
990: + -mat_view ::ascii_info - Prints info on matrix at conclusion of `MatAssemblyEnd()`
991: . -mat_view ::ascii_info_detail - Prints more detailed info
992: . -mat_view - Prints matrix in ASCII format
993: . -mat_view ::ascii_matlab - Prints matrix in Matlab format
994: . -mat_view draw - PetscDraws nonzero structure of matrix, using `MatView()` and `PetscDrawOpenX()`.
995: . -display <name> - Sets display name (default is host)
996: . -draw_pause <sec> - Sets number of seconds to pause after display
997: . -mat_view socket - Sends matrix to socket, can be accessed from Matlab (see Users-Manual: ch_matlab for details)
998: . -viewer_socket_machine <machine> -
999: . -viewer_socket_port <port> -
1000: . -mat_view binary - save matrix to file in binary format
1001: - -viewer_binary_filename <name> -
1003: Level: beginner
1005: Notes:
1006: The available visualization contexts include
1007: + `PETSC_VIEWER_STDOUT_SELF` - for sequential matrices
1008: . `PETSC_VIEWER_STDOUT_WORLD` - for parallel matrices created on `PETSC_COMM_WORLD`
1009: . `PETSC_VIEWER_STDOUT_`(comm) - for matrices created on MPI communicator comm
1010: - `PETSC_VIEWER_DRAW_WORLD` - graphical display of nonzero structure
1012: The user can open alternative visualization contexts with
1013: + `PetscViewerASCIIOpen()` - Outputs matrix to a specified file
1014: . `PetscViewerBinaryOpen()` - Outputs matrix in binary to a
1015: specified file; corresponding input uses MatLoad()
1016: . `PetscViewerDrawOpen()` - Outputs nonzero matrix structure to
1017: an X window display
1018: - `PetscViewerSocketOpen()` - Outputs matrix to Socket viewer.
1019: Currently only the sequential dense and AIJ
1020: matrix types support the Socket viewer.
1022: The user can call `PetscViewerPushFormat()` to specify the output
1023: format of ASCII printed objects (when using `PETSC_VIEWER_STDOUT_SELF`,
1024: `PETSC_VIEWER_STDOUT_WORLD` and `PetscViewerASCIIOpen()`). Available formats include
1025: + `PETSC_VIEWER_DEFAULT` - default, prints matrix contents
1026: . `PETSC_VIEWER_ASCII_MATLAB` - prints matrix contents in Matlab format
1027: . `PETSC_VIEWER_ASCII_DENSE` - prints entire matrix including zeros
1028: . `PETSC_VIEWER_ASCII_COMMON` - prints matrix contents, using a sparse
1029: format common among all matrix types
1030: . `PETSC_VIEWER_ASCII_IMPL` - prints matrix contents, using an implementation-specific
1031: format (which is in many cases the same as the default)
1032: . `PETSC_VIEWER_ASCII_INFO` - prints basic information about the matrix
1033: size and structure (not the matrix entries)
1034: - `PETSC_VIEWER_ASCII_INFO_DETAIL` - prints more detailed information about
1035: the matrix structure
1037: The ASCII viewers are only recommended for small matrices on at most a moderate number of processes,
1038: the program will seemingly hang and take hours for larger matrices, for larger matrices one should use the binary format.
1040: In the debugger you can do "call MatView(mat,0)" to display the matrix. (The same holds for any PETSc object viewer).
1042: See the manual page for `MatLoad()` for the exact format of the binary file when the binary
1043: viewer is used.
1045: See share/petsc/matlab/PetscBinaryRead.m for a Matlab code that can read in the binary file when the binary
1046: viewer is used and lib/petsc/bin/PetscBinaryIO.py for loading them into Python.
1048: One can use '-mat_view draw -draw_pause -1' to pause the graphical display of matrix nonzero structure,
1049: and then use the following mouse functions.
1050: .vb
1051: left mouse: zoom in
1052: middle mouse: zoom out
1053: right mouse: continue with the simulation
1054: .ve
1056: .seealso: [](chapter_matrices), `Mat`, `PetscViewerPushFormat()`, `PetscViewerASCIIOpen()`, `PetscViewerDrawOpen()`, `PetscViewer`,
1057: `PetscViewerSocketOpen()`, `PetscViewerBinaryOpen()`, `MatLoad()`, `MatViewFromOptions()`
1058: @*/
1059: PetscErrorCode MatView(Mat mat, PetscViewer viewer)
1060: {
1061: PetscInt rows, cols, rbs, cbs;
1062: PetscBool isascii, isstring, issaws;
1063: PetscViewerFormat format;
1064: PetscMPIInt size;
1066: PetscFunctionBegin;
1069: if (!viewer) PetscCall(PetscViewerASCIIGetStdout(PetscObjectComm((PetscObject)mat), &viewer));
1071: PetscCheckSameComm(mat, 1, viewer, 2);
1073: PetscCall(PetscViewerGetFormat(viewer, &format));
1074: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
1075: if (size == 1 && format == PETSC_VIEWER_LOAD_BALANCE) PetscFunctionReturn(PETSC_SUCCESS);
1077: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERSTRING, &isstring));
1078: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &isascii));
1079: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERSAWS, &issaws));
1080: PetscCheck((isascii && (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL)) || !mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "No viewers for factored matrix except ASCII, info, or info_detail");
1082: PetscCall(PetscLogEventBegin(MAT_View, mat, viewer, 0, 0));
1083: if (isascii) {
1084: if (!mat->preallocated) {
1085: PetscCall(PetscViewerASCIIPrintf(viewer, "Matrix has not been preallocated yet\n"));
1086: PetscFunctionReturn(PETSC_SUCCESS);
1087: }
1088: if (!mat->assembled) {
1089: PetscCall(PetscViewerASCIIPrintf(viewer, "Matrix has not been assembled yet\n"));
1090: PetscFunctionReturn(PETSC_SUCCESS);
1091: }
1092: PetscCall(PetscObjectPrintClassNamePrefixType((PetscObject)mat, viewer));
1093: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
1094: MatNullSpace nullsp, transnullsp;
1096: PetscCall(PetscViewerASCIIPushTab(viewer));
1097: PetscCall(MatGetSize(mat, &rows, &cols));
1098: PetscCall(MatGetBlockSizes(mat, &rbs, &cbs));
1099: if (rbs != 1 || cbs != 1) {
1100: if (rbs != cbs) PetscCall(PetscViewerASCIIPrintf(viewer, "rows=%" PetscInt_FMT ", cols=%" PetscInt_FMT ", rbs=%" PetscInt_FMT ", cbs=%" PetscInt_FMT "\n", rows, cols, rbs, cbs));
1101: else PetscCall(PetscViewerASCIIPrintf(viewer, "rows=%" PetscInt_FMT ", cols=%" PetscInt_FMT ", bs=%" PetscInt_FMT "\n", rows, cols, rbs));
1102: } else PetscCall(PetscViewerASCIIPrintf(viewer, "rows=%" PetscInt_FMT ", cols=%" PetscInt_FMT "\n", rows, cols));
1103: if (mat->factortype) {
1104: MatSolverType solver;
1105: PetscCall(MatFactorGetSolverType(mat, &solver));
1106: PetscCall(PetscViewerASCIIPrintf(viewer, "package used to perform factorization: %s\n", solver));
1107: }
1108: if (mat->ops->getinfo) {
1109: MatInfo info;
1110: PetscCall(MatGetInfo(mat, MAT_GLOBAL_SUM, &info));
1111: PetscCall(PetscViewerASCIIPrintf(viewer, "total: nonzeros=%.f, allocated nonzeros=%.f\n", info.nz_used, info.nz_allocated));
1112: if (!mat->factortype) PetscCall(PetscViewerASCIIPrintf(viewer, "total number of mallocs used during MatSetValues calls=%" PetscInt_FMT "\n", (PetscInt)info.mallocs));
1113: }
1114: PetscCall(MatGetNullSpace(mat, &nullsp));
1115: PetscCall(MatGetTransposeNullSpace(mat, &transnullsp));
1116: if (nullsp) PetscCall(PetscViewerASCIIPrintf(viewer, " has attached null space\n"));
1117: if (transnullsp && transnullsp != nullsp) PetscCall(PetscViewerASCIIPrintf(viewer, " has attached transposed null space\n"));
1118: PetscCall(MatGetNearNullSpace(mat, &nullsp));
1119: if (nullsp) PetscCall(PetscViewerASCIIPrintf(viewer, " has attached near null space\n"));
1120: PetscCall(PetscViewerASCIIPushTab(viewer));
1121: PetscCall(MatProductView(mat, viewer));
1122: PetscCall(PetscViewerASCIIPopTab(viewer));
1123: }
1124: } else if (issaws) {
1125: #if defined(PETSC_HAVE_SAWS)
1126: PetscMPIInt rank;
1128: PetscCall(PetscObjectName((PetscObject)mat));
1129: PetscCallMPI(MPI_Comm_rank(PETSC_COMM_WORLD, &rank));
1130: if (!((PetscObject)mat)->amsmem && rank == 0) PetscCall(PetscObjectViewSAWs((PetscObject)mat, viewer));
1131: #endif
1132: } else if (isstring) {
1133: const char *type;
1134: PetscCall(MatGetType(mat, &type));
1135: PetscCall(PetscViewerStringSPrintf(viewer, " MatType: %-7.7s", type));
1136: PetscTryTypeMethod(mat, view, viewer);
1137: }
1138: if ((format == PETSC_VIEWER_NATIVE || format == PETSC_VIEWER_LOAD_BALANCE) && mat->ops->viewnative) {
1139: PetscCall(PetscViewerASCIIPushTab(viewer));
1140: PetscUseTypeMethod(mat, viewnative, viewer);
1141: PetscCall(PetscViewerASCIIPopTab(viewer));
1142: } else if (mat->ops->view) {
1143: PetscCall(PetscViewerASCIIPushTab(viewer));
1144: PetscUseTypeMethod(mat, view, viewer);
1145: PetscCall(PetscViewerASCIIPopTab(viewer));
1146: }
1147: if (isascii) {
1148: PetscCall(PetscViewerGetFormat(viewer, &format));
1149: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) PetscCall(PetscViewerASCIIPopTab(viewer));
1150: }
1151: PetscCall(PetscLogEventEnd(MAT_View, mat, viewer, 0, 0));
1152: PetscFunctionReturn(PETSC_SUCCESS);
1153: }
1155: #if defined(PETSC_USE_DEBUG)
1156: #include <../src/sys/totalview/tv_data_display.h>
1157: PETSC_UNUSED static int TV_display_type(const struct _p_Mat *mat)
1158: {
1159: TV_add_row("Local rows", "int", &mat->rmap->n);
1160: TV_add_row("Local columns", "int", &mat->cmap->n);
1161: TV_add_row("Global rows", "int", &mat->rmap->N);
1162: TV_add_row("Global columns", "int", &mat->cmap->N);
1163: TV_add_row("Typename", TV_ascii_string_type, ((PetscObject)mat)->type_name);
1164: return TV_format_OK;
1165: }
1166: #endif
1168: /*@C
1169: MatLoad - Loads a matrix that has been stored in binary/HDF5 format
1170: with `MatView()`. The matrix format is determined from the options database.
1171: Generates a parallel MPI matrix if the communicator has more than one
1172: processor. The default matrix type is `MATAIJ`.
1174: Collective
1176: Input Parameters:
1177: + mat - the newly loaded matrix, this needs to have been created with `MatCreate()`
1178: or some related function before a call to `MatLoad()`
1179: - viewer - `PETSCVIEWERBINARY`/`PETSCVIEWERHDF5` file viewer
1181: Options Database Keys:
1182: Used with block matrix formats (`MATSEQBAIJ`, ...) to specify
1183: block size
1184: . -matload_block_size <bs> - set block size
1186: Level: beginner
1188: Notes:
1189: If the `Mat` type has not yet been given then `MATAIJ` is used, call `MatSetFromOptions()` on the
1190: `Mat` before calling this routine if you wish to set it from the options database.
1192: `MatLoad()` automatically loads into the options database any options
1193: given in the file filename.info where filename is the name of the file
1194: that was passed to the `PetscViewerBinaryOpen()`. The options in the info
1195: file will be ignored if you use the -viewer_binary_skip_info option.
1197: If the type or size of mat is not set before a call to `MatLoad()`, PETSc
1198: sets the default matrix type AIJ and sets the local and global sizes.
1199: If type and/or size is already set, then the same are used.
1201: In parallel, each processor can load a subset of rows (or the
1202: entire matrix). This routine is especially useful when a large
1203: matrix is stored on disk and only part of it is desired on each
1204: processor. For example, a parallel solver may access only some of
1205: the rows from each processor. The algorithm used here reads
1206: relatively small blocks of data rather than reading the entire
1207: matrix and then subsetting it.
1209: Viewer's `PetscViewerType` must be either `PETSCVIEWERBINARY` or `PETSCVIEWERHDF5`.
1210: Such viewer can be created using `PetscViewerBinaryOpen()` or `PetscViewerHDF5Open()`,
1211: or the sequence like
1212: .vb
1213: `PetscViewer` v;
1214: `PetscViewerCreate`(`PETSC_COMM_WORLD`,&v);
1215: `PetscViewerSetType`(v,`PETSCVIEWERBINARY`);
1216: `PetscViewerSetFromOptions`(v);
1217: `PetscViewerFileSetMode`(v,`FILE_MODE_READ`);
1218: `PetscViewerFileSetName`(v,"datafile");
1219: .ve
1220: The optional `PetscViewerSetFromOptions()` call allows overriding `PetscViewerSetType()` using the option
1221: $ -viewer_type {binary,hdf5}
1223: See the example src/ksp/ksp/tutorials/ex27.c with the first approach,
1224: and src/mat/tutorials/ex10.c with the second approach.
1226: In case of `PETSCVIEWERBINARY`, a native PETSc binary format is used. Each of the blocks
1227: is read onto rank 0 and then shipped to its destination rank, one after another.
1228: Multiple objects, both matrices and vectors, can be stored within the same file.
1229: Their PetscObject name is ignored; they are loaded in the order of their storage.
1231: Most users should not need to know the details of the binary storage
1232: format, since `MatLoad()` and `MatView()` completely hide these details.
1233: But for anyone who's interested, the standard binary matrix storage
1234: format is
1236: .vb
1237: PetscInt MAT_FILE_CLASSID
1238: PetscInt number of rows
1239: PetscInt number of columns
1240: PetscInt total number of nonzeros
1241: PetscInt *number nonzeros in each row
1242: PetscInt *column indices of all nonzeros (starting index is zero)
1243: PetscScalar *values of all nonzeros
1244: .ve
1246: PETSc automatically does the byte swapping for
1247: machines that store the bytes reversed. Thus if you write your own binary
1248: read/write routines you have to swap the bytes; see `PetscBinaryRead()`
1249: and `PetscBinaryWrite()` to see how this may be done.
1251: In case of `PETSCVIEWERHDF5`, a parallel HDF5 reader is used.
1252: Each processor's chunk is loaded independently by its owning rank.
1253: Multiple objects, both matrices and vectors, can be stored within the same file.
1254: They are looked up by their PetscObject name.
1256: As the MATLAB MAT-File Version 7.3 format is also a HDF5 flavor, we decided to use
1257: by default the same structure and naming of the AIJ arrays and column count
1258: within the HDF5 file. This means that a MAT file saved with -v7.3 flag, e.g.
1259: $ save example.mat A b -v7.3
1260: can be directly read by this routine (see Reference 1 for details).
1262: Depending on your MATLAB version, this format might be a default,
1263: otherwise you can set it as default in Preferences.
1265: Unless -nocompression flag is used to save the file in MATLAB,
1266: PETSc must be configured with ZLIB package.
1268: See also examples src/mat/tutorials/ex10.c and src/ksp/ksp/tutorials/ex27.c
1270: This reader currently supports only real `MATSEQAIJ`, `MATMPIAIJ`, `MATSEQDENSE` and `MATMPIDENSE` matrices for `PETSCVIEWERHDF5`
1272: Corresponding `MatView()` is not yet implemented.
1274: The loaded matrix is actually a transpose of the original one in MATLAB,
1275: unless you push `PETSC_VIEWER_HDF5_MAT` format (see examples above).
1276: With this format, matrix is automatically transposed by PETSc,
1277: unless the matrix is marked as SPD or symmetric
1278: (see `MatSetOption()`, `MAT_SPD`, `MAT_SYMMETRIC`).
1280: References:
1281: . * - MATLAB(R) Documentation, manual page of save(), https://www.mathworks.com/help/matlab/ref/save.html#btox10b-1-version
1283: .seealso: [](chapter_matrices), `Mat`, `PetscViewerBinaryOpen()`, `PetscViewerSetType()`, `MatView()`, `VecLoad()`
1284: @*/
1285: PetscErrorCode MatLoad(Mat mat, PetscViewer viewer)
1286: {
1287: PetscBool flg;
1289: PetscFunctionBegin;
1293: if (!((PetscObject)mat)->type_name) PetscCall(MatSetType(mat, MATAIJ));
1295: flg = PETSC_FALSE;
1296: PetscCall(PetscOptionsGetBool(((PetscObject)mat)->options, ((PetscObject)mat)->prefix, "-matload_symmetric", &flg, NULL));
1297: if (flg) {
1298: PetscCall(MatSetOption(mat, MAT_SYMMETRIC, PETSC_TRUE));
1299: PetscCall(MatSetOption(mat, MAT_SYMMETRY_ETERNAL, PETSC_TRUE));
1300: }
1301: flg = PETSC_FALSE;
1302: PetscCall(PetscOptionsGetBool(((PetscObject)mat)->options, ((PetscObject)mat)->prefix, "-matload_spd", &flg, NULL));
1303: if (flg) PetscCall(MatSetOption(mat, MAT_SPD, PETSC_TRUE));
1305: PetscCall(PetscLogEventBegin(MAT_Load, mat, viewer, 0, 0));
1306: PetscUseTypeMethod(mat, load, viewer);
1307: PetscCall(PetscLogEventEnd(MAT_Load, mat, viewer, 0, 0));
1308: PetscFunctionReturn(PETSC_SUCCESS);
1309: }
1311: static PetscErrorCode MatDestroy_Redundant(Mat_Redundant **redundant)
1312: {
1313: Mat_Redundant *redund = *redundant;
1315: PetscFunctionBegin;
1316: if (redund) {
1317: if (redund->matseq) { /* via MatCreateSubMatrices() */
1318: PetscCall(ISDestroy(&redund->isrow));
1319: PetscCall(ISDestroy(&redund->iscol));
1320: PetscCall(MatDestroySubMatrices(1, &redund->matseq));
1321: } else {
1322: PetscCall(PetscFree2(redund->send_rank, redund->recv_rank));
1323: PetscCall(PetscFree(redund->sbuf_j));
1324: PetscCall(PetscFree(redund->sbuf_a));
1325: for (PetscInt i = 0; i < redund->nrecvs; i++) {
1326: PetscCall(PetscFree(redund->rbuf_j[i]));
1327: PetscCall(PetscFree(redund->rbuf_a[i]));
1328: }
1329: PetscCall(PetscFree4(redund->sbuf_nz, redund->rbuf_nz, redund->rbuf_j, redund->rbuf_a));
1330: }
1332: if (redund->subcomm) PetscCall(PetscCommDestroy(&redund->subcomm));
1333: PetscCall(PetscFree(redund));
1334: }
1335: PetscFunctionReturn(PETSC_SUCCESS);
1336: }
1338: /*@C
1339: MatDestroy - Frees space taken by a matrix.
1341: Collective
1343: Input Parameter:
1344: . A - the matrix
1346: Level: beginner
1348: Developer Note:
1349: Some special arrays of matrices are not destroyed in this routine but instead by the routines called by
1350: `MatDestroySubMatrices()`. Thus one must be sure that any changes here must also be made in those routines.
1351: `MatHeaderMerge()` and `MatHeaderReplace()` also manipulate the data in the `Mat` object and likely need changes
1352: if changes are needed here.
1354: .seealso: [](chapter_matrices), `Mat`, `MatCreate()`
1355: @*/
1356: PetscErrorCode MatDestroy(Mat *A)
1357: {
1358: PetscFunctionBegin;
1359: if (!*A) PetscFunctionReturn(PETSC_SUCCESS);
1361: if (--((PetscObject)(*A))->refct > 0) {
1362: *A = NULL;
1363: PetscFunctionReturn(PETSC_SUCCESS);
1364: }
1366: /* if memory was published with SAWs then destroy it */
1367: PetscCall(PetscObjectSAWsViewOff((PetscObject)*A));
1368: PetscTryTypeMethod((*A), destroy);
1370: PetscCall(PetscFree((*A)->factorprefix));
1371: PetscCall(PetscFree((*A)->defaultvectype));
1372: PetscCall(PetscFree((*A)->defaultrandtype));
1373: PetscCall(PetscFree((*A)->bsizes));
1374: PetscCall(PetscFree((*A)->solvertype));
1375: for (PetscInt i = 0; i < MAT_FACTOR_NUM_TYPES; i++) PetscCall(PetscFree((*A)->preferredordering[i]));
1376: if ((*A)->redundant && (*A)->redundant->matseq[0] == *A) (*A)->redundant->matseq[0] = NULL;
1377: PetscCall(MatDestroy_Redundant(&(*A)->redundant));
1378: PetscCall(MatProductClear(*A));
1379: PetscCall(MatNullSpaceDestroy(&(*A)->nullsp));
1380: PetscCall(MatNullSpaceDestroy(&(*A)->transnullsp));
1381: PetscCall(MatNullSpaceDestroy(&(*A)->nearnullsp));
1382: PetscCall(MatDestroy(&(*A)->schur));
1383: PetscCall(PetscLayoutDestroy(&(*A)->rmap));
1384: PetscCall(PetscLayoutDestroy(&(*A)->cmap));
1385: PetscCall(PetscHeaderDestroy(A));
1386: PetscFunctionReturn(PETSC_SUCCESS);
1387: }
1389: /*@C
1390: MatSetValues - Inserts or adds a block of values into a matrix.
1391: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
1392: MUST be called after all calls to `MatSetValues()` have been completed.
1394: Not Collective
1396: Input Parameters:
1397: + mat - the matrix
1398: . v - a logically two-dimensional array of values
1399: . m - the number of rows
1400: . idxm - the global indices of the rows
1401: . n - the number of columns
1402: . idxn - the global indices of the columns
1403: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
1405: Level: beginner
1407: Notes:
1408: By default the values, `v`, are stored row-oriented. See `MatSetOption()` for other options.
1410: Calls to `MatSetValues()` with the `INSERT_VALUES` and `ADD_VALUES`
1411: options cannot be mixed without intervening calls to the assembly
1412: routines.
1414: `MatSetValues()` uses 0-based row and column numbers in Fortran
1415: as well as in C.
1417: Negative indices may be passed in `idxm` and `idxn`, these rows and columns are
1418: simply ignored. This allows easily inserting element stiffness matrices
1419: with homogeneous Dirchlet boundary conditions that you don't want represented
1420: in the matrix.
1422: Efficiency Alert:
1423: The routine `MatSetValuesBlocked()` may offer much better efficiency
1424: for users of block sparse formats (`MATSEQBAIJ` and `MATMPIBAIJ`).
1426: Developer Note:
1427: This is labeled with C so does not automatically generate Fortran stubs and interfaces
1428: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
1430: .seealso: [](chapter_matrices), `Mat`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1431: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`
1432: @*/
1433: PetscErrorCode MatSetValues(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], const PetscScalar v[], InsertMode addv)
1434: {
1435: PetscFunctionBeginHot;
1438: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1441: MatCheckPreallocated(mat, 1);
1443: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
1444: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
1446: if (PetscDefined(USE_DEBUG)) {
1447: PetscInt i, j;
1449: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
1450: for (i = 0; i < m; i++) {
1451: for (j = 0; j < n; j++) {
1452: if (mat->erroriffailure && PetscIsInfOrNanScalar(v[i * n + j]))
1453: #if defined(PETSC_USE_COMPLEX)
1454: 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]);
1455: #else
1456: 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]);
1457: #endif
1458: }
1459: }
1460: 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);
1461: 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);
1462: }
1464: if (mat->assembled) {
1465: mat->was_assembled = PETSC_TRUE;
1466: mat->assembled = PETSC_FALSE;
1467: }
1468: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
1469: PetscUseTypeMethod(mat, setvalues, m, idxm, n, idxn, v, addv);
1470: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
1471: PetscFunctionReturn(PETSC_SUCCESS);
1472: }
1474: /*@C
1475: MatSetValuesIS - Inserts or adds a block of values into a matrix using an `IS` to indicate the rows and columns
1476: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
1477: MUST be called after all calls to `MatSetValues()` have been completed.
1479: Not Collective
1481: Input Parameters:
1482: + mat - the matrix
1483: . v - a logically two-dimensional array of values
1484: . ism - the rows to provide
1485: . isn - the columns to provide
1486: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
1488: Level: beginner
1490: Notes:
1491: By default the values, `v`, are stored row-oriented. See `MatSetOption()` for other options.
1493: Calls to `MatSetValues()` with the `INSERT_VALUES` and `ADD_VALUES`
1494: options cannot be mixed without intervening calls to the assembly
1495: routines.
1497: `MatSetValues()` uses 0-based row and column numbers in Fortran
1498: as well as in C.
1500: Negative indices may be passed in `ism` and `isn`, these rows and columns are
1501: simply ignored. This allows easily inserting element stiffness matrices
1502: with homogeneous Dirchlet boundary conditions that you don't want represented
1503: in the matrix.
1505: Efficiency Alert:
1506: The routine `MatSetValuesBlocked()` may offer much better efficiency
1507: for users of block sparse formats (`MATSEQBAIJ` and `MATMPIBAIJ`).
1509: This is currently not optimized for any particular `ISType`
1511: Developer Notes:
1512: This is labeled with C so does not automatically generate Fortran stubs and interfaces
1513: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
1515: .seealso: [](chapter_matrices), `Mat`, `MatSetOption()`, `MatSetValues()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1516: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`, `MatSetValues()`
1517: @*/
1518: PetscErrorCode MatSetValuesIS(Mat mat, IS ism, IS isn, const PetscScalar v[], InsertMode addv)
1519: {
1520: PetscInt m, n;
1521: const PetscInt *rows, *cols;
1523: PetscFunctionBeginHot;
1525: PetscCall(ISGetIndices(ism, &rows));
1526: PetscCall(ISGetIndices(isn, &cols));
1527: PetscCall(ISGetLocalSize(ism, &m));
1528: PetscCall(ISGetLocalSize(isn, &n));
1529: PetscCall(MatSetValues(mat, m, rows, n, cols, v, addv));
1530: PetscCall(ISRestoreIndices(ism, &rows));
1531: PetscCall(ISRestoreIndices(isn, &cols));
1532: PetscFunctionReturn(PETSC_SUCCESS);
1533: }
1535: /*@
1536: MatSetValuesRowLocal - Inserts a row (block row for `MATBAIJ` matrices) of nonzero
1537: values into a matrix
1539: Not Collective
1541: Input Parameters:
1542: + mat - the matrix
1543: . row - the (block) row to set
1544: - v - a logically two-dimensional array of values
1546: Level: intermediate
1548: Notes:
1549: The values, `v`, are column-oriented (for the block version) and sorted
1551: All the nonzeros in the row must be provided
1553: The matrix must have previously had its column indices set, likely by having been assembled.
1555: The row must belong to this process
1557: .seealso: [](chapter_matrices), `Mat`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1558: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`, `MatSetValues()`, `MatSetValuesRow()`, `MatSetLocalToGlobalMapping()`
1559: @*/
1560: PetscErrorCode MatSetValuesRowLocal(Mat mat, PetscInt row, const PetscScalar v[])
1561: {
1562: PetscInt globalrow;
1564: PetscFunctionBegin;
1568: PetscCall(ISLocalToGlobalMappingApply(mat->rmap->mapping, 1, &row, &globalrow));
1569: PetscCall(MatSetValuesRow(mat, globalrow, v));
1570: PetscFunctionReturn(PETSC_SUCCESS);
1571: }
1573: /*@
1574: MatSetValuesRow - Inserts a row (block row for `MATBAIJ` matrices) of nonzero
1575: values into a matrix
1577: Not Collective
1579: Input Parameters:
1580: + mat - the matrix
1581: . row - the (block) row to set
1582: - 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
1584: Level: advanced
1586: Notes:
1587: The values, `v`, are column-oriented for the block version.
1589: All the nonzeros in the row must be provided
1591: THE MATRIX MUST HAVE PREVIOUSLY HAD ITS COLUMN INDICES SET. IT IS RARE THAT THIS ROUTINE IS USED, usually `MatSetValues()` is used.
1593: The row must belong to this process
1595: .seealso: [](chapter_matrices), `Mat`, `MatSetValues()`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
1596: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`, `MatSetValues()`
1597: @*/
1598: PetscErrorCode MatSetValuesRow(Mat mat, PetscInt row, const PetscScalar v[])
1599: {
1600: PetscFunctionBeginHot;
1603: MatCheckPreallocated(mat, 1);
1605: PetscCheck(mat->insertmode != ADD_VALUES, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add and insert values");
1606: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
1607: mat->insertmode = INSERT_VALUES;
1609: if (mat->assembled) {
1610: mat->was_assembled = PETSC_TRUE;
1611: mat->assembled = PETSC_FALSE;
1612: }
1613: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
1614: PetscUseTypeMethod(mat, setvaluesrow, row, v);
1615: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
1616: PetscFunctionReturn(PETSC_SUCCESS);
1617: }
1619: /*@
1620: MatSetValuesStencil - Inserts or adds a block of values into a matrix.
1621: Using structured grid indexing
1623: Not Collective
1625: Input Parameters:
1626: + mat - the matrix
1627: . m - number of rows being entered
1628: . idxm - grid coordinates (and component number when dof > 1) for matrix rows being entered
1629: . n - number of columns being entered
1630: . idxn - grid coordinates (and component number when dof > 1) for matrix columns being entered
1631: . v - a logically two-dimensional array of values
1632: - addv - either `ADD_VALUES` to add to existing entries at that location or `INSERT_VALUES` to replace existing entries with new values
1634: Level: beginner
1636: Notes:
1637: By default the values, `v`, are row-oriented. See `MatSetOption()` for other options.
1639: Calls to `MatSetValuesStencil()` with the `INSERT_VALUES` and `ADD_VALUES`
1640: options cannot be mixed without intervening calls to the assembly
1641: routines.
1643: The grid coordinates are across the entire grid, not just the local portion
1645: `MatSetValuesStencil()` uses 0-based row and column numbers in Fortran
1646: as well as in C.
1648: For setting/accessing vector values via array coordinates you can use the `DMDAVecGetArray()` routine
1650: In order to use this routine you must either obtain the matrix with `DMCreateMatrix()`
1651: or call `MatSetLocalToGlobalMapping()` and `MatSetStencil()` first.
1653: The columns and rows in the stencil passed in MUST be contained within the
1654: ghost region of the given process as set with DMDACreateXXX() or `MatSetStencil()`. For example,
1655: if you create a `DMDA` with an overlap of one grid level and on a particular process its first
1656: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1657: first i index you can use in your column and row indices in `MatSetStencil()` is 5.
1659: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
1660: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
1661: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
1662: `DM_BOUNDARY_PERIODIC` boundary type.
1664: 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
1665: a single value per point) you can skip filling those indices.
1667: Inspired by the structured grid interface to the HYPRE package
1668: (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)
1670: Efficiency Alert:
1671: The routine `MatSetValuesBlockedStencil()` may offer much better efficiency
1672: for users of block sparse formats (`MATSEQBAIJ` and `MATMPIBAIJ`).
1674: Fortran Note:
1675: `idxm` and `idxn` should be declared as
1676: $ MatStencil idxm(4,m),idxn(4,n)
1677: and the values inserted using
1678: .vb
1679: idxm(MatStencil_i,1) = i
1680: idxm(MatStencil_j,1) = j
1681: idxm(MatStencil_k,1) = k
1682: idxm(MatStencil_c,1) = c
1683: etc
1684: .ve
1686: .seealso: [](chapter_matrices), `Mat`, `DMDA`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`
1687: `MatSetValues()`, `MatSetValuesBlockedStencil()`, `MatSetStencil()`, `DMCreateMatrix()`, `DMDAVecGetArray()`, `MatStencil`
1688: @*/
1689: PetscErrorCode MatSetValuesStencil(Mat mat, PetscInt m, const MatStencil idxm[], PetscInt n, const MatStencil idxn[], const PetscScalar v[], InsertMode addv)
1690: {
1691: PetscInt buf[8192], *bufm = NULL, *bufn = NULL, *jdxm, *jdxn;
1692: PetscInt j, i, dim = mat->stencil.dim, *dims = mat->stencil.dims + 1, tmp;
1693: PetscInt *starts = mat->stencil.starts, *dxm = (PetscInt *)idxm, *dxn = (PetscInt *)idxn, sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1695: PetscFunctionBegin;
1696: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1702: if ((m + n) <= (PetscInt)(sizeof(buf) / sizeof(PetscInt))) {
1703: jdxm = buf;
1704: jdxn = buf + m;
1705: } else {
1706: PetscCall(PetscMalloc2(m, &bufm, n, &bufn));
1707: jdxm = bufm;
1708: jdxn = bufn;
1709: }
1710: for (i = 0; i < m; i++) {
1711: for (j = 0; j < 3 - sdim; j++) dxm++;
1712: tmp = *dxm++ - starts[0];
1713: for (j = 0; j < dim - 1; j++) {
1714: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1715: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
1716: }
1717: if (mat->stencil.noc) dxm++;
1718: jdxm[i] = tmp;
1719: }
1720: for (i = 0; i < n; i++) {
1721: for (j = 0; j < 3 - sdim; j++) dxn++;
1722: tmp = *dxn++ - starts[0];
1723: for (j = 0; j < dim - 1; j++) {
1724: if ((*dxn++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1725: else tmp = tmp * dims[j] + *(dxn - 1) - starts[j + 1];
1726: }
1727: if (mat->stencil.noc) dxn++;
1728: jdxn[i] = tmp;
1729: }
1730: PetscCall(MatSetValuesLocal(mat, m, jdxm, n, jdxn, v, addv));
1731: PetscCall(PetscFree2(bufm, bufn));
1732: PetscFunctionReturn(PETSC_SUCCESS);
1733: }
1735: /*@
1736: MatSetValuesBlockedStencil - Inserts or adds a block of values into a matrix.
1737: Using structured grid indexing
1739: Not Collective
1741: Input Parameters:
1742: + mat - the matrix
1743: . m - number of rows being entered
1744: . idxm - grid coordinates for matrix rows being entered
1745: . n - number of columns being entered
1746: . idxn - grid coordinates for matrix columns being entered
1747: . v - a logically two-dimensional array of values
1748: - addv - either `ADD_VALUES` to add to existing entries or `INSERT_VALUES` to replace existing entries with new values
1750: Level: beginner
1752: Notes:
1753: By default the values, `v`, are row-oriented and unsorted.
1754: See `MatSetOption()` for other options.
1756: Calls to `MatSetValuesBlockedStencil()` with the `INSERT_VALUES` and `ADD_VALUES`
1757: options cannot be mixed without intervening calls to the assembly
1758: routines.
1760: The grid coordinates are across the entire grid, not just the local portion
1762: `MatSetValuesBlockedStencil()` uses 0-based row and column numbers in Fortran
1763: as well as in C.
1765: For setting/accessing vector values via array coordinates you can use the `DMDAVecGetArray()` routine
1767: In order to use this routine you must either obtain the matrix with `DMCreateMatrix()`
1768: or call `MatSetBlockSize()`, `MatSetLocalToGlobalMapping()` and `MatSetStencil()` first.
1770: The columns and rows in the stencil passed in MUST be contained within the
1771: ghost region of the given process as set with DMDACreateXXX() or `MatSetStencil()`. For example,
1772: if you create a `DMDA` with an overlap of one grid level and on a particular process its first
1773: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1774: first i index you can use in your column and row indices in `MatSetStencil()` is 5.
1776: Negative indices may be passed in idxm and idxn, these rows and columns are
1777: simply ignored. This allows easily inserting element stiffness matrices
1778: with homogeneous Dirchlet boundary conditions that you don't want represented
1779: in the matrix.
1781: Inspired by the structured grid interface to the HYPRE package
1782: (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)
1784: Fortran Note:
1785: `idxm` and `idxn` should be declared as
1786: $ MatStencil idxm(4,m),idxn(4,n)
1787: and the values inserted using
1788: .vb
1789: idxm(MatStencil_i,1) = i
1790: idxm(MatStencil_j,1) = j
1791: idxm(MatStencil_k,1) = k
1792: etc
1793: .ve
1795: .seealso: [](chapter_matrices), `Mat`, `DMDA`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`
1796: `MatSetValues()`, `MatSetValuesStencil()`, `MatSetStencil()`, `DMCreateMatrix()`, `DMDAVecGetArray()`, `MatStencil`,
1797: `MatSetBlockSize()`, `MatSetLocalToGlobalMapping()`
1798: @*/
1799: PetscErrorCode MatSetValuesBlockedStencil(Mat mat, PetscInt m, const MatStencil idxm[], PetscInt n, const MatStencil idxn[], const PetscScalar v[], InsertMode addv)
1800: {
1801: PetscInt buf[8192], *bufm = NULL, *bufn = NULL, *jdxm, *jdxn;
1802: PetscInt j, i, dim = mat->stencil.dim, *dims = mat->stencil.dims + 1, tmp;
1803: PetscInt *starts = mat->stencil.starts, *dxm = (PetscInt *)idxm, *dxn = (PetscInt *)idxn, sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1805: PetscFunctionBegin;
1806: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1813: if ((m + n) <= (PetscInt)(sizeof(buf) / sizeof(PetscInt))) {
1814: jdxm = buf;
1815: jdxn = buf + m;
1816: } else {
1817: PetscCall(PetscMalloc2(m, &bufm, n, &bufn));
1818: jdxm = bufm;
1819: jdxn = bufn;
1820: }
1821: for (i = 0; i < m; i++) {
1822: for (j = 0; j < 3 - sdim; j++) dxm++;
1823: tmp = *dxm++ - starts[0];
1824: for (j = 0; j < sdim - 1; j++) {
1825: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1826: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
1827: }
1828: dxm++;
1829: jdxm[i] = tmp;
1830: }
1831: for (i = 0; i < n; i++) {
1832: for (j = 0; j < 3 - sdim; j++) dxn++;
1833: tmp = *dxn++ - starts[0];
1834: for (j = 0; j < sdim - 1; j++) {
1835: if ((*dxn++ - starts[j + 1]) < 0 || tmp < 0) tmp = -1;
1836: else tmp = tmp * dims[j] + *(dxn - 1) - starts[j + 1];
1837: }
1838: dxn++;
1839: jdxn[i] = tmp;
1840: }
1841: PetscCall(MatSetValuesBlockedLocal(mat, m, jdxm, n, jdxn, v, addv));
1842: PetscCall(PetscFree2(bufm, bufn));
1843: PetscFunctionReturn(PETSC_SUCCESS);
1844: }
1846: /*@
1847: MatSetStencil - Sets the grid information for setting values into a matrix via
1848: `MatSetValuesStencil()`
1850: Not Collective
1852: Input Parameters:
1853: + mat - the matrix
1854: . dim - dimension of the grid 1, 2, or 3
1855: . dims - number of grid points in x, y, and z direction, including ghost points on your processor
1856: . starts - starting point of ghost nodes on your processor in x, y, and z direction
1857: - dof - number of degrees of freedom per node
1859: Level: beginner
1861: Notes:
1862: Inspired by the structured grid interface to the HYPRE package
1863: (www.llnl.gov/CASC/hyper)
1865: For matrices generated with `DMCreateMatrix()` this routine is automatically called and so not needed by the
1866: user.
1868: .seealso: [](chapter_matrices), `Mat`, `MatStencil`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`
1869: `MatSetValues()`, `MatSetValuesBlockedStencil()`, `MatSetValuesStencil()`
1870: @*/
1871: PetscErrorCode MatSetStencil(Mat mat, PetscInt dim, const PetscInt dims[], const PetscInt starts[], PetscInt dof)
1872: {
1873: PetscFunctionBegin;
1878: mat->stencil.dim = dim + (dof > 1);
1879: for (PetscInt i = 0; i < dim; i++) {
1880: mat->stencil.dims[i] = dims[dim - i - 1]; /* copy the values in backwards */
1881: mat->stencil.starts[i] = starts[dim - i - 1];
1882: }
1883: mat->stencil.dims[dim] = dof;
1884: mat->stencil.starts[dim] = 0;
1885: mat->stencil.noc = (PetscBool)(dof == 1);
1886: PetscFunctionReturn(PETSC_SUCCESS);
1887: }
1889: /*@C
1890: MatSetValuesBlocked - Inserts or adds a block of values into a matrix.
1892: Not Collective
1894: Input Parameters:
1895: + mat - the matrix
1896: . v - a logically two-dimensional array of values
1897: . m - the number of block rows
1898: . idxm - the global block indices
1899: . n - the number of block columns
1900: . idxn - the global block indices
1901: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` replaces existing entries with new values
1903: Level: intermediate
1905: Notes:
1906: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call
1907: MatXXXXSetPreallocation() or `MatSetUp()` before using this routine.
1909: The `m` and `n` count the NUMBER of blocks in the row direction and column direction,
1910: NOT the total number of rows/columns; for example, if the block size is 2 and
1911: you are passing in values for rows 2,3,4,5 then m would be 2 (not 4).
1912: The values in idxm would be 1 2; that is the first index for each block divided by
1913: the block size.
1915: You must call `MatSetBlockSize()` when constructing this matrix (before
1916: preallocating it).
1918: By default the values, `v`, are row-oriented, so the layout of
1919: `v` is the same as for `MatSetValues()`. See `MatSetOption()` for other options.
1921: Calls to `MatSetValuesBlocked()` with the `INSERT_VALUES` and `ADD_VALUES`
1922: options cannot be mixed without intervening calls to the assembly
1923: routines.
1925: `MatSetValuesBlocked()` uses 0-based row and column numbers in Fortran
1926: as well as in C.
1928: Negative indices may be passed in `idxm` and `idxn`, these rows and columns are
1929: simply ignored. This allows easily inserting element stiffness matrices
1930: with homogeneous Dirchlet boundary conditions that you don't want represented
1931: in the matrix.
1933: Each time an entry is set within a sparse matrix via `MatSetValues()`,
1934: internal searching must be done to determine where to place the
1935: data in the matrix storage space. By instead inserting blocks of
1936: entries via `MatSetValuesBlocked()`, the overhead of matrix assembly is
1937: reduced.
1939: Example:
1940: .vb
1941: Suppose m=n=2 and block size(bs) = 2 The array is
1943: 1 2 | 3 4
1944: 5 6 | 7 8
1945: - - - | - - -
1946: 9 10 | 11 12
1947: 13 14 | 15 16
1949: v[] should be passed in like
1950: v[] = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]
1952: If you are not using row oriented storage of v (that is you called MatSetOption(mat,MAT_ROW_ORIENTED,PETSC_FALSE)) then
1953: v[] = [1,5,9,13,2,6,10,14,3,7,11,15,4,8,12,16]
1954: .ve
1956: .seealso: [](chapter_matrices), `Mat`, `MatSetBlockSize()`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetValuesBlockedLocal()`
1957: @*/
1958: PetscErrorCode MatSetValuesBlocked(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], const PetscScalar v[], InsertMode addv)
1959: {
1960: PetscFunctionBeginHot;
1963: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
1966: MatCheckPreallocated(mat, 1);
1967: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
1968: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
1969: if (PetscDefined(USE_DEBUG)) {
1970: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
1971: PetscCheck(mat->ops->setvaluesblocked || mat->ops->setvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
1972: }
1973: if (PetscDefined(USE_DEBUG)) {
1974: PetscInt rbs, cbs, M, N, i;
1975: PetscCall(MatGetBlockSizes(mat, &rbs, &cbs));
1976: PetscCall(MatGetSize(mat, &M, &N));
1977: for (i = 0; i < m; i++) PetscCheck(idxm[i] * rbs < M, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Row block index %" PetscInt_FMT " (index %" PetscInt_FMT ") greater than row length %" PetscInt_FMT, i, idxm[i], M);
1978: for (i = 0; i < n; i++) PetscCheck(idxn[i] * cbs < N, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column block index %" PetscInt_FMT " (index %" PetscInt_FMT ") great than column length %" PetscInt_FMT, i, idxn[i], N);
1979: }
1980: if (mat->assembled) {
1981: mat->was_assembled = PETSC_TRUE;
1982: mat->assembled = PETSC_FALSE;
1983: }
1984: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
1985: if (mat->ops->setvaluesblocked) {
1986: PetscUseTypeMethod(mat, setvaluesblocked, m, idxm, n, idxn, v, addv);
1987: } else {
1988: PetscInt buf[8192], *bufr = NULL, *bufc = NULL, *iidxm, *iidxn;
1989: PetscInt i, j, bs, cbs;
1991: PetscCall(MatGetBlockSizes(mat, &bs, &cbs));
1992: if (m * bs + n * cbs <= (PetscInt)(sizeof(buf) / sizeof(PetscInt))) {
1993: iidxm = buf;
1994: iidxn = buf + m * bs;
1995: } else {
1996: PetscCall(PetscMalloc2(m * bs, &bufr, n * cbs, &bufc));
1997: iidxm = bufr;
1998: iidxn = bufc;
1999: }
2000: for (i = 0; i < m; i++) {
2001: for (j = 0; j < bs; j++) iidxm[i * bs + j] = bs * idxm[i] + j;
2002: }
2003: if (m != n || bs != cbs || idxm != idxn) {
2004: for (i = 0; i < n; i++) {
2005: for (j = 0; j < cbs; j++) iidxn[i * cbs + j] = cbs * idxn[i] + j;
2006: }
2007: } else iidxn = iidxm;
2008: PetscCall(MatSetValues(mat, m * bs, iidxm, n * cbs, iidxn, v, addv));
2009: PetscCall(PetscFree2(bufr, bufc));
2010: }
2011: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
2012: PetscFunctionReturn(PETSC_SUCCESS);
2013: }
2015: /*@C
2016: MatGetValues - Gets a block of local values from a matrix.
2018: Not Collective; can only return values that are owned by the give process
2020: Input Parameters:
2021: + mat - the matrix
2022: . v - a logically two-dimensional array for storing the values
2023: . m - the number of rows
2024: . idxm - the global indices of the rows
2025: . n - the number of columns
2026: - idxn - the global indices of the columns
2028: Level: advanced
2030: Notes:
2031: The user must allocate space (m*n `PetscScalar`s) for the values, `v`.
2032: The values, `v`, are then returned in a row-oriented format,
2033: analogous to that used by default in `MatSetValues()`.
2035: `MatGetValues()` uses 0-based row and column numbers in
2036: Fortran as well as in C.
2038: `MatGetValues()` requires that the matrix has been assembled
2039: with `MatAssemblyBegin()`/`MatAssemblyEnd()`. Thus, calls to
2040: `MatSetValues()` and `MatGetValues()` CANNOT be made in succession
2041: without intermediate matrix assembly.
2043: Negative row or column indices will be ignored and those locations in `v` will be
2044: left unchanged.
2046: For the standard row-based matrix formats, `idxm` can only contain rows owned by the requesting MPI rank.
2047: That is, rows with global index greater than or equal to rstart and less than rend where rstart and rend are obtainable
2048: from `MatGetOwnershipRange`(mat,&rstart,&rend).
2050: .seealso: [](chapter_matrices), `Mat`, `MatGetRow()`, `MatCreateSubMatrices()`, `MatSetValues()`, `MatGetOwnershipRange()`, `MatGetValuesLocal()`, `MatGetValue()`
2051: @*/
2052: PetscErrorCode MatGetValues(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], PetscScalar v[])
2053: {
2054: PetscFunctionBegin;
2057: if (!m || !n) PetscFunctionReturn(PETSC_SUCCESS);
2061: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2062: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2063: MatCheckPreallocated(mat, 1);
2065: PetscCall(PetscLogEventBegin(MAT_GetValues, mat, 0, 0, 0));
2066: PetscUseTypeMethod(mat, getvalues, m, idxm, n, idxn, v);
2067: PetscCall(PetscLogEventEnd(MAT_GetValues, mat, 0, 0, 0));
2068: PetscFunctionReturn(PETSC_SUCCESS);
2069: }
2071: /*@C
2072: MatGetValuesLocal - retrieves values from certain locations in a matrix using the local numbering of the indices
2073: defined previously by `MatSetLocalToGlobalMapping()`
2075: Not Collective
2077: Input Parameters:
2078: + mat - the matrix
2079: . nrow - number of rows
2080: . irow - the row local indices
2081: . ncol - number of columns
2082: - icol - the column local indices
2084: Output Parameter:
2085: . y - a logically two-dimensional array of values
2087: Level: advanced
2089: Notes:
2090: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call `MatSetLocalToGlobalMapping()` before using this routine.
2092: This routine can only return values that are owned by the requesting MPI rank. That is, for standard matrix formats, rows that, in the global numbering,
2093: are greater than or equal to rstart and less than rend where rstart and rend are obtainable from `MatGetOwnershipRange`(mat,&rstart,&rend). One can
2094: determine if the resulting global row associated with the local row r is owned by the requesting MPI rank by applying the `ISLocalToGlobalMapping` set
2095: with `MatSetLocalToGlobalMapping()`.
2097: Developer Note:
2098: This is labelled with C so does not automatically generate Fortran stubs and interfaces
2099: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2101: .seealso: [](chapter_matrices), `Mat`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetLocalToGlobalMapping()`,
2102: `MatSetValuesLocal()`, `MatGetValues()`
2103: @*/
2104: PetscErrorCode MatGetValuesLocal(Mat mat, PetscInt nrow, const PetscInt irow[], PetscInt ncol, const PetscInt icol[], PetscScalar y[])
2105: {
2106: PetscFunctionBeginHot;
2109: MatCheckPreallocated(mat, 1);
2110: if (!nrow || !ncol) PetscFunctionReturn(PETSC_SUCCESS); /* no values to retrieve */
2113: if (PetscDefined(USE_DEBUG)) {
2114: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2115: PetscCheck(mat->ops->getvalueslocal || mat->ops->getvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2116: }
2117: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2118: PetscCall(PetscLogEventBegin(MAT_GetValues, mat, 0, 0, 0));
2119: if (mat->ops->getvalueslocal) PetscUseTypeMethod(mat, getvalueslocal, nrow, irow, ncol, icol, y);
2120: else {
2121: PetscInt buf[8192], *bufr = NULL, *bufc = NULL, *irowm, *icolm;
2122: if ((nrow + ncol) <= (PetscInt)(sizeof(buf) / sizeof(PetscInt))) {
2123: irowm = buf;
2124: icolm = buf + nrow;
2125: } else {
2126: PetscCall(PetscMalloc2(nrow, &bufr, ncol, &bufc));
2127: irowm = bufr;
2128: icolm = bufc;
2129: }
2130: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "MatGetValuesLocal() cannot proceed without local-to-global row mapping (See MatSetLocalToGlobalMapping()).");
2131: PetscCheck(mat->cmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "MatGetValuesLocal() cannot proceed without local-to-global column mapping (See MatSetLocalToGlobalMapping()).");
2132: PetscCall(ISLocalToGlobalMappingApply(mat->rmap->mapping, nrow, irow, irowm));
2133: PetscCall(ISLocalToGlobalMappingApply(mat->cmap->mapping, ncol, icol, icolm));
2134: PetscCall(MatGetValues(mat, nrow, irowm, ncol, icolm, y));
2135: PetscCall(PetscFree2(bufr, bufc));
2136: }
2137: PetscCall(PetscLogEventEnd(MAT_GetValues, mat, 0, 0, 0));
2138: PetscFunctionReturn(PETSC_SUCCESS);
2139: }
2141: /*@
2142: MatSetValuesBatch - Adds (`ADD_VALUES`) many blocks of values into a matrix at once. The blocks must all be square and
2143: the same size. Currently, this can only be called once and creates the given matrix.
2145: Not Collective
2147: Input Parameters:
2148: + mat - the matrix
2149: . nb - the number of blocks
2150: . bs - the number of rows (and columns) in each block
2151: . rows - a concatenation of the rows for each block
2152: - v - a concatenation of logically two-dimensional arrays of values
2154: Level: advanced
2156: Note:
2157: `MatSetPreallocationCOO()` and `MatSetValuesCOO()` may be a better way to provide the values
2159: In the future, we will extend this routine to handle rectangular blocks, and to allow multiple calls for a given matrix.
2161: .seealso: [](chapter_matrices), `Mat`, `MatSetOption()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValuesBlocked()`, `MatSetValuesLocal()`,
2162: `InsertMode`, `INSERT_VALUES`, `ADD_VALUES`, `MatSetValues()`, `MatSetPreallocationCOO()`, `MatSetValuesCOO()`
2163: @*/
2164: PetscErrorCode MatSetValuesBatch(Mat mat, PetscInt nb, PetscInt bs, PetscInt rows[], const PetscScalar v[])
2165: {
2166: PetscFunctionBegin;
2171: PetscAssert(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2173: PetscCall(PetscLogEventBegin(MAT_SetValuesBatch, mat, 0, 0, 0));
2174: if (mat->ops->setvaluesbatch) PetscUseTypeMethod(mat, setvaluesbatch, nb, bs, rows, v);
2175: else {
2176: for (PetscInt b = 0; b < nb; ++b) PetscCall(MatSetValues(mat, bs, &rows[b * bs], bs, &rows[b * bs], &v[b * bs * bs], ADD_VALUES));
2177: }
2178: PetscCall(PetscLogEventEnd(MAT_SetValuesBatch, mat, 0, 0, 0));
2179: PetscFunctionReturn(PETSC_SUCCESS);
2180: }
2182: /*@
2183: MatSetLocalToGlobalMapping - Sets a local-to-global numbering for use by
2184: the routine `MatSetValuesLocal()` to allow users to insert matrix entries
2185: using a local (per-processor) numbering.
2187: Not Collective
2189: Input Parameters:
2190: + x - the matrix
2191: . rmapping - row mapping created with `ISLocalToGlobalMappingCreate()` or `ISLocalToGlobalMappingCreateIS()`
2192: - cmapping - column mapping
2194: Level: intermediate
2196: Note:
2197: If the matrix is obtained with `DMCreateMatrix()` then this may already have been called on the matrix
2199: .seealso: [](chapter_matrices), `Mat`, `DM`, `DMCreateMatrix()`, `MatGetLocalToGlobalMapping()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetValuesLocal()`, `MatGetValuesLocal()`
2200: @*/
2201: PetscErrorCode MatSetLocalToGlobalMapping(Mat x, ISLocalToGlobalMapping rmapping, ISLocalToGlobalMapping cmapping)
2202: {
2203: PetscFunctionBegin;
2208: if (x->ops->setlocaltoglobalmapping) PetscUseTypeMethod(x, setlocaltoglobalmapping, rmapping, cmapping);
2209: else {
2210: PetscCall(PetscLayoutSetISLocalToGlobalMapping(x->rmap, rmapping));
2211: PetscCall(PetscLayoutSetISLocalToGlobalMapping(x->cmap, cmapping));
2212: }
2213: PetscFunctionReturn(PETSC_SUCCESS);
2214: }
2216: /*@
2217: MatGetLocalToGlobalMapping - Gets the local-to-global numbering set by `MatSetLocalToGlobalMapping()`
2219: Not Collective
2221: Input Parameter:
2222: . A - the matrix
2224: Output Parameters:
2225: + rmapping - row mapping
2226: - cmapping - column mapping
2228: Level: advanced
2230: .seealso: [](chapter_matrices), `Mat`, `MatSetLocalToGlobalMapping()`, `MatSetValuesLocal()`
2231: @*/
2232: PetscErrorCode MatGetLocalToGlobalMapping(Mat A, ISLocalToGlobalMapping *rmapping, ISLocalToGlobalMapping *cmapping)
2233: {
2234: PetscFunctionBegin;
2237: if (rmapping) {
2239: *rmapping = A->rmap->mapping;
2240: }
2241: if (cmapping) {
2243: *cmapping = A->cmap->mapping;
2244: }
2245: PetscFunctionReturn(PETSC_SUCCESS);
2246: }
2248: /*@
2249: MatSetLayouts - Sets the `PetscLayout` objects for rows and columns of a matrix
2251: Logically Collective
2253: Input Parameters:
2254: + A - the matrix
2255: . rmap - row layout
2256: - cmap - column layout
2258: Level: advanced
2260: Note:
2261: The `PetscLayout` objects are usually created automatically for the matrix so this routine rarely needs to be called.
2263: .seealso: [](chapter_matrices), `Mat`, `PetscLayout`, `MatCreateVecs()`, `MatGetLocalToGlobalMapping()`, `MatGetLayouts()`
2264: @*/
2265: PetscErrorCode MatSetLayouts(Mat A, PetscLayout rmap, PetscLayout cmap)
2266: {
2267: PetscFunctionBegin;
2269: PetscCall(PetscLayoutReference(rmap, &A->rmap));
2270: PetscCall(PetscLayoutReference(cmap, &A->cmap));
2271: PetscFunctionReturn(PETSC_SUCCESS);
2272: }
2274: /*@
2275: MatGetLayouts - Gets the `PetscLayout` objects for rows and columns
2277: Not Collective
2279: Input Parameter:
2280: . A - the matrix
2282: Output Parameters:
2283: + rmap - row layout
2284: - cmap - column layout
2286: Level: advanced
2288: .seealso: [](chapter_matrices), `Mat`, [Matrix Layouts](sec_matlayout), `PetscLayout`, `MatCreateVecs()`, `MatGetLocalToGlobalMapping()`, `MatSetLayouts()`
2289: @*/
2290: PetscErrorCode MatGetLayouts(Mat A, PetscLayout *rmap, PetscLayout *cmap)
2291: {
2292: PetscFunctionBegin;
2295: if (rmap) {
2297: *rmap = A->rmap;
2298: }
2299: if (cmap) {
2301: *cmap = A->cmap;
2302: }
2303: PetscFunctionReturn(PETSC_SUCCESS);
2304: }
2306: /*@C
2307: MatSetValuesLocal - Inserts or adds values into certain locations of a matrix,
2308: using a local numbering of the nodes.
2310: Not Collective
2312: Input Parameters:
2313: + mat - the matrix
2314: . nrow - number of rows
2315: . irow - the row local indices
2316: . ncol - number of columns
2317: . icol - the column local indices
2318: . y - a logically two-dimensional array of values
2319: - addv - either `INSERT_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
2321: Level: intermediate
2323: Notes:
2324: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call MatXXXXSetPreallocation() or
2325: `MatSetUp()` before using this routine
2327: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call `MatSetLocalToGlobalMapping()` before using this routine
2329: Calls to `MatSetValuesLocal()` with the `INSERT_VALUES` and `ADD_VALUES`
2330: options cannot be mixed without intervening calls to the assembly
2331: routines.
2333: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
2334: MUST be called after all calls to `MatSetValuesLocal()` have been completed.
2336: Developer Note:
2337: This is labeled with C so does not automatically generate Fortran stubs and interfaces
2338: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2340: .seealso: [](chapter_matrices), `Mat`, `MatAssemblyBegin()`, `MatAssemblyEnd()`, `MatSetValues()`, `MatSetLocalToGlobalMapping()`,
2341: `MatGetValuesLocal()`
2342: @*/
2343: PetscErrorCode MatSetValuesLocal(Mat mat, PetscInt nrow, const PetscInt irow[], PetscInt ncol, const PetscInt icol[], const PetscScalar y[], InsertMode addv)
2344: {
2345: PetscFunctionBeginHot;
2348: MatCheckPreallocated(mat, 1);
2349: if (!nrow || !ncol) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
2352: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
2353: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
2354: if (PetscDefined(USE_DEBUG)) {
2355: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2356: PetscCheck(mat->ops->setvalueslocal || mat->ops->setvalues, PETSC_COMM_SELF, PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2357: }
2359: if (mat->assembled) {
2360: mat->was_assembled = PETSC_TRUE;
2361: mat->assembled = PETSC_FALSE;
2362: }
2363: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
2364: if (mat->ops->setvalueslocal) PetscUseTypeMethod(mat, setvalueslocal, nrow, irow, ncol, icol, y, addv);
2365: else {
2366: PetscInt buf[8192], *bufr = NULL, *bufc = NULL;
2367: const PetscInt *irowm, *icolm;
2369: if ((!mat->rmap->mapping && !mat->cmap->mapping) || (nrow + ncol) <= (PetscInt)(sizeof(buf) / sizeof(PetscInt))) {
2370: bufr = buf;
2371: bufc = buf + nrow;
2372: irowm = bufr;
2373: icolm = bufc;
2374: } else {
2375: PetscCall(PetscMalloc2(nrow, &bufr, ncol, &bufc));
2376: irowm = bufr;
2377: icolm = bufc;
2378: }
2379: if (mat->rmap->mapping) PetscCall(ISLocalToGlobalMappingApply(mat->rmap->mapping, nrow, irow, bufr));
2380: else irowm = irow;
2381: if (mat->cmap->mapping) {
2382: if (mat->cmap->mapping != mat->rmap->mapping || ncol != nrow || icol != irow) {
2383: PetscCall(ISLocalToGlobalMappingApply(mat->cmap->mapping, ncol, icol, bufc));
2384: } else icolm = irowm;
2385: } else icolm = icol;
2386: PetscCall(MatSetValues(mat, nrow, irowm, ncol, icolm, y, addv));
2387: if (bufr != buf) PetscCall(PetscFree2(bufr, bufc));
2388: }
2389: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
2390: PetscFunctionReturn(PETSC_SUCCESS);
2391: }
2393: /*@C
2394: MatSetValuesBlockedLocal - Inserts or adds values into certain locations of a matrix,
2395: using a local ordering of the nodes a block at a time.
2397: Not Collective
2399: Input Parameters:
2400: + x - the matrix
2401: . nrow - number of rows
2402: . irow - the row local indices
2403: . ncol - number of columns
2404: . icol - the column local indices
2405: . y - a logically two-dimensional array of values
2406: - addv - either `ADD_VALUES` to add values to any existing entries, or `INSERT_VALUES` to replace existing entries with new values
2408: Level: intermediate
2410: Notes:
2411: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call MatXXXXSetPreallocation() or
2412: `MatSetUp()` before using this routine
2414: If you create the matrix yourself (that is not with a call to `DMCreateMatrix()`) then you MUST call `MatSetBlockSize()` and `MatSetLocalToGlobalMapping()`
2415: before using this routineBefore calling `MatSetValuesLocal()`, the user must first set the
2417: Calls to `MatSetValuesBlockedLocal()` with the `INSERT_VALUES` and `ADD_VALUES`
2418: options cannot be mixed without intervening calls to the assembly
2419: routines.
2421: These values may be cached, so `MatAssemblyBegin()` and `MatAssemblyEnd()`
2422: MUST be called after all calls to `MatSetValuesBlockedLocal()` have been completed.
2424: Developer Note:
2425: This is labeled with C so does not automatically generate Fortran stubs and interfaces
2426: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2428: .seealso: [](chapter_matrices), `Mat`, `MatSetBlockSize()`, `MatSetLocalToGlobalMapping()`, `MatAssemblyBegin()`, `MatAssemblyEnd()`,
2429: `MatSetValuesLocal()`, `MatSetValuesBlocked()`
2430: @*/
2431: PetscErrorCode MatSetValuesBlockedLocal(Mat mat, PetscInt nrow, const PetscInt irow[], PetscInt ncol, const PetscInt icol[], const PetscScalar y[], InsertMode addv)
2432: {
2433: PetscFunctionBeginHot;
2436: MatCheckPreallocated(mat, 1);
2437: if (!nrow || !ncol) PetscFunctionReturn(PETSC_SUCCESS); /* no values to insert */
2440: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
2441: else PetscCheck(mat->insertmode == addv, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot mix add values and insert values");
2442: if (PetscDefined(USE_DEBUG)) {
2443: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2444: 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);
2445: }
2447: if (mat->assembled) {
2448: mat->was_assembled = PETSC_TRUE;
2449: mat->assembled = PETSC_FALSE;
2450: }
2451: if (PetscUnlikelyDebug(mat->rmap->mapping)) { /* Condition on the mapping existing, because MatSetValuesBlockedLocal_IS does not require it to be set. */
2452: PetscInt irbs, rbs;
2453: PetscCall(MatGetBlockSizes(mat, &rbs, NULL));
2454: PetscCall(ISLocalToGlobalMappingGetBlockSize(mat->rmap->mapping, &irbs));
2455: PetscCheck(rbs == irbs, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Different row block sizes! mat %" PetscInt_FMT ", row l2g map %" PetscInt_FMT, rbs, irbs);
2456: }
2457: if (PetscUnlikelyDebug(mat->cmap->mapping)) {
2458: PetscInt icbs, cbs;
2459: PetscCall(MatGetBlockSizes(mat, NULL, &cbs));
2460: PetscCall(ISLocalToGlobalMappingGetBlockSize(mat->cmap->mapping, &icbs));
2461: PetscCheck(cbs == icbs, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Different col block sizes! mat %" PetscInt_FMT ", col l2g map %" PetscInt_FMT, cbs, icbs);
2462: }
2463: PetscCall(PetscLogEventBegin(MAT_SetValues, mat, 0, 0, 0));
2464: if (mat->ops->setvaluesblockedlocal) PetscUseTypeMethod(mat, setvaluesblockedlocal, nrow, irow, ncol, icol, y, addv);
2465: else {
2466: PetscInt buf[8192], *bufr = NULL, *bufc = NULL;
2467: const PetscInt *irowm, *icolm;
2469: if ((!mat->rmap->mapping && !mat->cmap->mapping) || (nrow + ncol) <= (PetscInt)(sizeof(buf) / sizeof(PetscInt))) {
2470: bufr = buf;
2471: bufc = buf + nrow;
2472: irowm = bufr;
2473: icolm = bufc;
2474: } else {
2475: PetscCall(PetscMalloc2(nrow, &bufr, ncol, &bufc));
2476: irowm = bufr;
2477: icolm = bufc;
2478: }
2479: if (mat->rmap->mapping) PetscCall(ISLocalToGlobalMappingApplyBlock(mat->rmap->mapping, nrow, irow, bufr));
2480: else irowm = irow;
2481: if (mat->cmap->mapping) {
2482: if (mat->cmap->mapping != mat->rmap->mapping || ncol != nrow || icol != irow) {
2483: PetscCall(ISLocalToGlobalMappingApplyBlock(mat->cmap->mapping, ncol, icol, bufc));
2484: } else icolm = irowm;
2485: } else icolm = icol;
2486: PetscCall(MatSetValuesBlocked(mat, nrow, irowm, ncol, icolm, y, addv));
2487: if (bufr != buf) PetscCall(PetscFree2(bufr, bufc));
2488: }
2489: PetscCall(PetscLogEventEnd(MAT_SetValues, mat, 0, 0, 0));
2490: PetscFunctionReturn(PETSC_SUCCESS);
2491: }
2493: /*@
2494: MatMultDiagonalBlock - Computes the matrix-vector product, y = Dx. Where D is defined by the inode or block structure of the diagonal
2496: Collective
2498: Input Parameters:
2499: + mat - the matrix
2500: - x - the vector to be multiplied
2502: Output Parameter:
2503: . y - the result
2505: Level: developer
2507: Note:
2508: The vectors `x` and `y` cannot be the same. I.e., one cannot
2509: call `MatMultDiagonalBlock`(A,y,y).
2511: .seealso: [](chapter_matrices), `Mat`, `MatMult()`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
2512: @*/
2513: PetscErrorCode MatMultDiagonalBlock(Mat mat, Vec x, Vec y)
2514: {
2515: PetscFunctionBegin;
2521: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2522: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2523: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2524: MatCheckPreallocated(mat, 1);
2526: PetscUseTypeMethod(mat, multdiagonalblock, x, y);
2527: PetscCall(PetscObjectStateIncrease((PetscObject)y));
2528: PetscFunctionReturn(PETSC_SUCCESS);
2529: }
2531: /*@
2532: MatMult - Computes the matrix-vector product, y = Ax.
2534: Neighbor-wise Collective
2536: Input Parameters:
2537: + mat - the matrix
2538: - x - the vector to be multiplied
2540: Output Parameter:
2541: . y - the result
2543: Level: beginner
2545: Note:
2546: The vectors `x` and `y` cannot be the same. I.e., one cannot
2547: call `MatMult`(A,y,y).
2549: .seealso: [](chapter_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
2550: @*/
2551: PetscErrorCode MatMult(Mat mat, Vec x, Vec y)
2552: {
2553: PetscFunctionBegin;
2557: VecCheckAssembled(x);
2559: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2560: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2561: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2562: 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);
2563: 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);
2564: 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);
2565: 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);
2566: PetscCall(VecSetErrorIfLocked(y, 3));
2567: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(x, 2, PETSC_TRUE));
2568: MatCheckPreallocated(mat, 1);
2570: PetscCall(VecLockReadPush(x));
2571: PetscCall(PetscLogEventBegin(MAT_Mult, mat, x, y, 0));
2572: PetscUseTypeMethod(mat, mult, x, y);
2573: PetscCall(PetscLogEventEnd(MAT_Mult, mat, x, y, 0));
2574: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(y, 3, PETSC_FALSE));
2575: PetscCall(VecLockReadPop(x));
2576: PetscFunctionReturn(PETSC_SUCCESS);
2577: }
2579: /*@
2580: MatMultTranspose - Computes matrix transpose times a vector y = A^T * x.
2582: Neighbor-wise Collective
2584: Input Parameters:
2585: + mat - the matrix
2586: - x - the vector to be multiplied
2588: Output Parameter:
2589: . y - the result
2591: Level: beginner
2593: Notes:
2594: The vectors `x` and `y` cannot be the same. I.e., one cannot
2595: call `MatMultTranspose`(A,y,y).
2597: For complex numbers this does NOT compute the Hermitian (complex conjugate) transpose multiple,
2598: use `MatMultHermitianTranspose()`
2600: .seealso: [](chapter_matrices), `Mat`, `MatMult()`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatMultHermitianTranspose()`, `MatTranspose()`
2601: @*/
2602: PetscErrorCode MatMultTranspose(Mat mat, Vec x, Vec y)
2603: {
2604: PetscErrorCode (*op)(Mat, Vec, Vec) = NULL;
2606: PetscFunctionBegin;
2610: VecCheckAssembled(x);
2613: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2614: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2615: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2616: 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);
2617: 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);
2618: 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);
2619: 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);
2620: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(x, 2, PETSC_TRUE));
2621: MatCheckPreallocated(mat, 1);
2623: if (!mat->ops->multtranspose) {
2624: if (mat->symmetric == PETSC_BOOL3_TRUE && mat->ops->mult) op = mat->ops->mult;
2625: 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);
2626: } else op = mat->ops->multtranspose;
2627: PetscCall(PetscLogEventBegin(MAT_MultTranspose, mat, x, y, 0));
2628: PetscCall(VecLockReadPush(x));
2629: PetscCall((*op)(mat, x, y));
2630: PetscCall(VecLockReadPop(x));
2631: PetscCall(PetscLogEventEnd(MAT_MultTranspose, mat, x, y, 0));
2632: PetscCall(PetscObjectStateIncrease((PetscObject)y));
2633: if (mat->erroriffailure) PetscCall(VecValidValues_Internal(y, 3, PETSC_FALSE));
2634: PetscFunctionReturn(PETSC_SUCCESS);
2635: }
2637: /*@
2638: MatMultHermitianTranspose - Computes matrix Hermitian transpose times a vector.
2640: Neighbor-wise Collective
2642: Input Parameters:
2643: + mat - the matrix
2644: - x - the vector to be multilplied
2646: Output Parameter:
2647: . y - the result
2649: Level: beginner
2651: Notes:
2652: The vectors `x` and `y` cannot be the same. I.e., one cannot
2653: call `MatMultHermitianTranspose`(A,y,y).
2655: Also called the conjugate transpose, complex conjugate transpose, or adjoint.
2657: For real numbers `MatMultTranspose()` and `MatMultHermitianTranspose()` are identical.
2659: .seealso: [](chapter_matrices), `Mat`, `MatMult()`, `MatMultAdd()`, `MatMultHermitianTransposeAdd()`, `MatMultTranspose()`
2660: @*/
2661: PetscErrorCode MatMultHermitianTranspose(Mat mat, Vec x, Vec y)
2662: {
2663: PetscFunctionBegin;
2669: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2670: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2671: PetscCheck(x != y, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "x and y must be different vectors");
2672: 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);
2673: 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);
2674: 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);
2675: 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);
2676: MatCheckPreallocated(mat, 1);
2678: PetscCall(PetscLogEventBegin(MAT_MultHermitianTranspose, mat, x, y, 0));
2679: #if defined(PETSC_USE_COMPLEX)
2680: if (mat->ops->multhermitiantranspose || (mat->hermitian == PETSC_BOOL3_TRUE && mat->ops->mult)) {
2681: PetscCall(VecLockReadPush(x));
2682: if (mat->ops->multhermitiantranspose) PetscUseTypeMethod(mat, multhermitiantranspose, x, y);
2683: else PetscUseTypeMethod(mat, mult, x, y);
2684: PetscCall(VecLockReadPop(x));
2685: } else {
2686: Vec w;
2687: PetscCall(VecDuplicate(x, &w));
2688: PetscCall(VecCopy(x, w));
2689: PetscCall(VecConjugate(w));
2690: PetscCall(MatMultTranspose(mat, w, y));
2691: PetscCall(VecDestroy(&w));
2692: PetscCall(VecConjugate(y));
2693: }
2694: PetscCall(PetscObjectStateIncrease((PetscObject)y));
2695: #else
2696: PetscCall(MatMultTranspose(mat, x, y));
2697: #endif
2698: PetscCall(PetscLogEventEnd(MAT_MultHermitianTranspose, mat, x, y, 0));
2699: PetscFunctionReturn(PETSC_SUCCESS);
2700: }
2702: /*@
2703: MatMultAdd - Computes v3 = v2 + A * v1.
2705: Neighbor-wise Collective
2707: Input Parameters:
2708: + mat - the matrix
2709: . v1 - the vector to be multiplied by `mat`
2710: - v2 - the vector to be added to the result
2712: Output Parameter:
2713: . v3 - the result
2715: Level: beginner
2717: Note:
2718: The vectors `v1` and `v3` cannot be the same. I.e., one cannot
2719: call `MatMultAdd`(A,v1,v2,v1).
2721: .seealso: [](chapter_matrices), `Mat`, `MatMultTranspose()`, `MatMult()`, `MatMultTransposeAdd()`
2722: @*/
2723: PetscErrorCode MatMultAdd(Mat mat, Vec v1, Vec v2, Vec v3)
2724: {
2725: PetscFunctionBegin;
2732: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2733: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2734: 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);
2735: /* 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);
2736: 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); */
2737: 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);
2738: 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);
2739: PetscCheck(v1 != v3, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "v1 and v3 must be different vectors");
2740: MatCheckPreallocated(mat, 1);
2742: PetscCall(PetscLogEventBegin(MAT_MultAdd, mat, v1, v2, v3));
2743: PetscCall(VecLockReadPush(v1));
2744: PetscUseTypeMethod(mat, multadd, v1, v2, v3);
2745: PetscCall(VecLockReadPop(v1));
2746: PetscCall(PetscLogEventEnd(MAT_MultAdd, mat, v1, v2, v3));
2747: PetscCall(PetscObjectStateIncrease((PetscObject)v3));
2748: PetscFunctionReturn(PETSC_SUCCESS);
2749: }
2751: /*@
2752: MatMultTransposeAdd - Computes v3 = v2 + A' * v1.
2754: Neighbor-wise Collective
2756: Input Parameters:
2757: + mat - the matrix
2758: . v1 - the vector to be multiplied by the transpose of the matrix
2759: - v2 - the vector to be added to the result
2761: Output Parameter:
2762: . v3 - the result
2764: Level: beginner
2766: Note:
2767: The vectors `v1` and `v3` cannot be the same. I.e., one cannot
2768: call `MatMultTransposeAdd`(A,v1,v2,v1).
2770: .seealso: [](chapter_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMult()`
2771: @*/
2772: PetscErrorCode MatMultTransposeAdd(Mat mat, Vec v1, Vec v2, Vec v3)
2773: {
2774: PetscErrorCode (*op)(Mat, Vec, Vec, Vec) = (!mat->ops->multtransposeadd && mat->symmetric) ? mat->ops->multadd : mat->ops->multtransposeadd;
2776: PetscFunctionBegin;
2783: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2784: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2785: 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);
2786: 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);
2787: 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);
2788: PetscCheck(v1 != v3, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "v1 and v3 must be different vectors");
2789: PetscCheck(op, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Mat type %s", ((PetscObject)mat)->type_name);
2790: MatCheckPreallocated(mat, 1);
2792: PetscCall(PetscLogEventBegin(MAT_MultTransposeAdd, mat, v1, v2, v3));
2793: PetscCall(VecLockReadPush(v1));
2794: PetscCall((*op)(mat, v1, v2, v3));
2795: PetscCall(VecLockReadPop(v1));
2796: PetscCall(PetscLogEventEnd(MAT_MultTransposeAdd, mat, v1, v2, v3));
2797: PetscCall(PetscObjectStateIncrease((PetscObject)v3));
2798: PetscFunctionReturn(PETSC_SUCCESS);
2799: }
2801: /*@
2802: MatMultHermitianTransposeAdd - Computes v3 = v2 + A^H * v1.
2804: Neighbor-wise Collective
2806: Input Parameters:
2807: + mat - the matrix
2808: . v1 - the vector to be multiplied by the Hermitian transpose
2809: - v2 - the vector to be added to the result
2811: Output Parameter:
2812: . v3 - the result
2814: Level: beginner
2816: Note:
2817: The vectors `v1` and `v3` cannot be the same. I.e., one cannot
2818: call `MatMultHermitianTransposeAdd`(A,v1,v2,v1).
2820: .seealso: [](chapter_matrices), `Mat`, `MatMultHermitianTranspose()`, `MatMultTranspose()`, `MatMultAdd()`, `MatMult()`
2821: @*/
2822: PetscErrorCode MatMultHermitianTransposeAdd(Mat mat, Vec v1, Vec v2, Vec v3)
2823: {
2824: PetscFunctionBegin;
2831: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
2832: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
2833: PetscCheck(v1 != v3, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "v1 and v3 must be different vectors");
2834: 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);
2835: 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);
2836: 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);
2837: MatCheckPreallocated(mat, 1);
2839: PetscCall(PetscLogEventBegin(MAT_MultHermitianTransposeAdd, mat, v1, v2, v3));
2840: PetscCall(VecLockReadPush(v1));
2841: if (mat->ops->multhermitiantransposeadd) PetscUseTypeMethod(mat, multhermitiantransposeadd, v1, v2, v3);
2842: else {
2843: Vec w, z;
2844: PetscCall(VecDuplicate(v1, &w));
2845: PetscCall(VecCopy(v1, w));
2846: PetscCall(VecConjugate(w));
2847: PetscCall(VecDuplicate(v3, &z));
2848: PetscCall(MatMultTranspose(mat, w, z));
2849: PetscCall(VecDestroy(&w));
2850: PetscCall(VecConjugate(z));
2851: if (v2 != v3) {
2852: PetscCall(VecWAXPY(v3, 1.0, v2, z));
2853: } else {
2854: PetscCall(VecAXPY(v3, 1.0, z));
2855: }
2856: PetscCall(VecDestroy(&z));
2857: }
2858: PetscCall(VecLockReadPop(v1));
2859: PetscCall(PetscLogEventEnd(MAT_MultHermitianTransposeAdd, mat, v1, v2, v3));
2860: PetscCall(PetscObjectStateIncrease((PetscObject)v3));
2861: PetscFunctionReturn(PETSC_SUCCESS);
2862: }
2864: /*@C
2865: MatGetFactorType - gets the type of factorization it is
2867: Not Collective
2869: Input Parameter:
2870: . mat - the matrix
2872: Output Parameter:
2873: . 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`
2875: Level: intermediate
2877: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorType`, `MatGetFactor()`, `MatSetFactorType()`, `MAT_FACTOR_NONE`, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ILU`,
2878: `MAT_FACTOR_ICC,MAT_FACTOR_ILUDT`, `MAT_FACTOR_QR`
2879: @*/
2880: PetscErrorCode MatGetFactorType(Mat mat, MatFactorType *t)
2881: {
2882: PetscFunctionBegin;
2886: *t = mat->factortype;
2887: PetscFunctionReturn(PETSC_SUCCESS);
2888: }
2890: /*@C
2891: MatSetFactorType - sets the type of factorization it is
2893: Logically Collective
2895: Input Parameters:
2896: + mat - the matrix
2897: - t - the type, one of `MAT_FACTOR_NONE`, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ILU`, `MAT_FACTOR_ICC,MAT_FACTOR_ILUDT`, `MAT_FACTOR_QR`
2899: Level: intermediate
2901: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorType`, `MatGetFactor()`, `MatGetFactorType()`, `MAT_FACTOR_NONE`, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ILU`,
2902: `MAT_FACTOR_ICC,MAT_FACTOR_ILUDT`, `MAT_FACTOR_QR`
2903: @*/
2904: PetscErrorCode MatSetFactorType(Mat mat, MatFactorType t)
2905: {
2906: PetscFunctionBegin;
2909: mat->factortype = t;
2910: PetscFunctionReturn(PETSC_SUCCESS);
2911: }
2913: /*@C
2914: MatGetInfo - Returns information about matrix storage (number of
2915: nonzeros, memory, etc.).
2917: Collective if `MAT_GLOBAL_MAX` or `MAT_GLOBAL_SUM` is used as the flag
2919: Input Parameters:
2920: + mat - the matrix
2921: - 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)
2923: Output Parameter:
2924: . info - matrix information context
2926: Notes:
2927: The `MatInfo` context contains a variety of matrix data, including
2928: number of nonzeros allocated and used, number of mallocs during
2929: matrix assembly, etc. Additional information for factored matrices
2930: is provided (such as the fill ratio, number of mallocs during
2931: factorization, etc.). Much of this info is printed to `PETSC_STDOUT`
2932: when using the runtime options
2933: $ -info -mat_view ::ascii_info
2935: Example:
2936: See the file ${PETSC_DIR}/include/petscmat.h for a complete list of
2937: data within the MatInfo context. For example,
2938: .vb
2939: MatInfo info;
2940: Mat A;
2941: double mal, nz_a, nz_u;
2943: MatGetInfo(A,MAT_LOCAL,&info);
2944: mal = info.mallocs;
2945: nz_a = info.nz_allocated;
2946: .ve
2948: Fortran users should declare info as a double precision
2949: array of dimension `MAT_INFO_SIZE`, and then extract the parameters
2950: of interest. See the file ${PETSC_DIR}/include/petsc/finclude/petscmat.h
2951: a complete list of parameter names.
2952: .vb
2953: double precision info(MAT_INFO_SIZE)
2954: double precision mal, nz_a
2955: Mat A
2956: integer ierr
2958: call MatGetInfo(A,MAT_LOCAL,info,ierr)
2959: mal = info(MAT_INFO_MALLOCS)
2960: nz_a = info(MAT_INFO_NZ_ALLOCATED)
2961: .ve
2963: Level: intermediate
2965: Developer Note:
2966: The Fortran interface is not autogenerated as the
2967: interface definition cannot be generated correctly [due to `MatInfo` argument]
2969: .seealso: [](chapter_matrices), `Mat`, `MatInfo`, `MatStashGetInfo()`
2970: @*/
2971: PetscErrorCode MatGetInfo(Mat mat, MatInfoType flag, MatInfo *info)
2972: {
2973: PetscFunctionBegin;
2977: MatCheckPreallocated(mat, 1);
2978: PetscUseTypeMethod(mat, getinfo, flag, info);
2979: PetscFunctionReturn(PETSC_SUCCESS);
2980: }
2982: /*
2983: This is used by external packages where it is not easy to get the info from the actual
2984: matrix factorization.
2985: */
2986: PetscErrorCode MatGetInfo_External(Mat A, MatInfoType flag, MatInfo *info)
2987: {
2988: PetscFunctionBegin;
2989: PetscCall(PetscMemzero(info, sizeof(MatInfo)));
2990: PetscFunctionReturn(PETSC_SUCCESS);
2991: }
2993: /*@C
2994: MatLUFactor - Performs in-place LU factorization of matrix.
2996: Collective
2998: Input Parameters:
2999: + mat - the matrix
3000: . row - row permutation
3001: . col - column permutation
3002: - info - options for factorization, includes
3003: .vb
3004: fill - expected fill as ratio of original fill.
3005: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3006: Run with the option -info to determine an optimal value to use
3007: .ve
3008: Level: developer
3010: Notes:
3011: Most users should employ the `KSP` interface for linear solvers
3012: instead of working directly with matrix algebra routines such as this.
3013: See, e.g., `KSPCreate()`.
3015: This changes the state of the matrix to a factored matrix; it cannot be used
3016: for example with `MatSetValues()` unless one first calls `MatSetUnfactored()`.
3018: This is really in-place only for dense matrices, the preferred approach is to use `MatGetFactor()`, `MatLUFactorSymbolic()`, and `MatLUFactorNumeric()`
3019: when not using `KSP`.
3021: Developer Note:
3022: The Fortran interface is not autogenerated as the
3023: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3025: .seealso: [](chapter_matrices), [Matrix Factorization](sec_matfactor), `Mat`, `MatFactorType`, `MatLUFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`,
3026: `MatGetOrdering()`, `MatSetUnfactored()`, `MatFactorInfo`, `MatGetFactor()`
3027: @*/
3028: PetscErrorCode MatLUFactor(Mat mat, IS row, IS col, const MatFactorInfo *info)
3029: {
3030: MatFactorInfo tinfo;
3032: PetscFunctionBegin;
3038: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3039: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3040: MatCheckPreallocated(mat, 1);
3041: if (!info) {
3042: PetscCall(MatFactorInfoInitialize(&tinfo));
3043: info = &tinfo;
3044: }
3046: PetscCall(PetscLogEventBegin(MAT_LUFactor, mat, row, col, 0));
3047: PetscUseTypeMethod(mat, lufactor, row, col, info);
3048: PetscCall(PetscLogEventEnd(MAT_LUFactor, mat, row, col, 0));
3049: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3050: PetscFunctionReturn(PETSC_SUCCESS);
3051: }
3053: /*@C
3054: MatILUFactor - Performs in-place ILU factorization of matrix.
3056: Collective
3058: Input Parameters:
3059: + mat - the matrix
3060: . row - row permutation
3061: . col - column permutation
3062: - info - structure containing
3063: .vb
3064: levels - number of levels of fill.
3065: expected fill - as ratio of original fill.
3066: 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
3067: missing diagonal entries)
3068: .ve
3070: Level: developer
3072: Notes:
3073: Most users should employ the `KSP` interface for linear solvers
3074: instead of working directly with matrix algebra routines such as this.
3075: See, e.g., `KSPCreate()`.
3077: Probably really in-place only when level of fill is zero, otherwise allocates
3078: new space to store factored matrix and deletes previous memory. The preferred approach is to use `MatGetFactor()`, `MatILUFactorSymbolic()`, and `MatILUFactorNumeric()`
3079: when not using `KSP`.
3081: Developer Note:
3082: The Fortran interface is not autogenerated as the
3083: interface definition cannot be generated correctly [due to MatFactorInfo]
3085: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatILUFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`, `MatFactorInfo`
3086: @*/
3087: PetscErrorCode MatILUFactor(Mat mat, IS row, IS col, const MatFactorInfo *info)
3088: {
3089: PetscFunctionBegin;
3095: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "matrix must be square");
3096: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3097: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3098: MatCheckPreallocated(mat, 1);
3100: PetscCall(PetscLogEventBegin(MAT_ILUFactor, mat, row, col, 0));
3101: PetscUseTypeMethod(mat, ilufactor, row, col, info);
3102: PetscCall(PetscLogEventEnd(MAT_ILUFactor, mat, row, col, 0));
3103: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3104: PetscFunctionReturn(PETSC_SUCCESS);
3105: }
3107: /*@C
3108: MatLUFactorSymbolic - Performs symbolic LU factorization of matrix.
3109: Call this routine before calling `MatLUFactorNumeric()` and after `MatGetFactor()`.
3111: Collective
3113: Input Parameters:
3114: + fact - the factor matrix obtained with `MatGetFactor()`
3115: . mat - the matrix
3116: . row - the row permutation
3117: . col - the column permutation
3118: - info - options for factorization, includes
3119: .vb
3120: fill - expected fill as ratio of original fill. Run with the option -info to determine an optimal value to use
3121: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3122: .ve
3124: Level: developer
3126: Notes:
3127: See [Matrix Factorization](sec_matfactor) for additional information about factorizations
3129: Most users should employ the simplified `KSP` interface for linear solvers
3130: instead of working directly with matrix algebra routines such as this.
3131: See, e.g., `KSPCreate()`.
3133: Developer Note:
3134: The Fortran interface is not autogenerated as the
3135: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3137: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatLUFactor()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`, `MatFactorInfo`, `MatFactorInfoInitialize()`
3138: @*/
3139: PetscErrorCode MatLUFactorSymbolic(Mat fact, Mat mat, IS row, IS col, const MatFactorInfo *info)
3140: {
3141: MatFactorInfo tinfo;
3143: PetscFunctionBegin;
3151: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3152: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3153: MatCheckPreallocated(mat, 2);
3154: if (!info) {
3155: PetscCall(MatFactorInfoInitialize(&tinfo));
3156: info = &tinfo;
3157: }
3159: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_LUFactorSymbolic, mat, row, col, 0));
3160: PetscUseTypeMethod(fact, lufactorsymbolic, mat, row, col, info);
3161: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_LUFactorSymbolic, mat, row, col, 0));
3162: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3163: PetscFunctionReturn(PETSC_SUCCESS);
3164: }
3166: /*@C
3167: MatLUFactorNumeric - Performs numeric LU factorization of a matrix.
3168: Call this routine after first calling `MatLUFactorSymbolic()` and `MatGetFactor()`.
3170: Collective
3172: Input Parameters:
3173: + fact - the factor matrix obtained with `MatGetFactor()`
3174: . mat - the matrix
3175: - info - options for factorization
3177: Level: developer
3179: Notes:
3180: See `MatLUFactor()` for in-place factorization. See
3181: `MatCholeskyFactorNumeric()` for the symmetric, positive definite case.
3183: Most users should employ the `KSP` interface for linear solvers
3184: instead of working directly with matrix algebra routines such as this.
3185: See, e.g., `KSPCreate()`.
3187: Developer Note:
3188: The Fortran interface is not autogenerated as the
3189: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3191: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatFactorInfo`, `MatLUFactorSymbolic()`, `MatLUFactor()`, `MatCholeskyFactor()`
3192: @*/
3193: PetscErrorCode MatLUFactorNumeric(Mat fact, Mat mat, const MatFactorInfo *info)
3194: {
3195: MatFactorInfo tinfo;
3197: PetscFunctionBegin;
3203: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3204: 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,
3205: mat->rmap->N, (fact)->rmap->N, mat->cmap->N, (fact)->cmap->N);
3207: MatCheckPreallocated(mat, 2);
3208: if (!info) {
3209: PetscCall(MatFactorInfoInitialize(&tinfo));
3210: info = &tinfo;
3211: }
3213: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_LUFactorNumeric, mat, fact, 0, 0));
3214: else PetscCall(PetscLogEventBegin(MAT_LUFactor, mat, fact, 0, 0));
3215: PetscUseTypeMethod(fact, lufactornumeric, mat, info);
3216: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_LUFactorNumeric, mat, fact, 0, 0));
3217: else PetscCall(PetscLogEventEnd(MAT_LUFactor, mat, fact, 0, 0));
3218: PetscCall(MatViewFromOptions(fact, NULL, "-mat_factor_view"));
3219: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3220: PetscFunctionReturn(PETSC_SUCCESS);
3221: }
3223: /*@C
3224: MatCholeskyFactor - Performs in-place Cholesky factorization of a
3225: symmetric matrix.
3227: Collective
3229: Input Parameters:
3230: + mat - the matrix
3231: . perm - row and column permutations
3232: - f - expected fill as ratio of original fill
3234: Level: developer
3236: Notes:
3237: See `MatLUFactor()` for the nonsymmetric case. See also `MatGetFactor()`,
3238: `MatCholeskyFactorSymbolic()`, and `MatCholeskyFactorNumeric()`.
3240: Most users should employ the `KSP` interface for linear solvers
3241: instead of working directly with matrix algebra routines such as this.
3242: See, e.g., `KSPCreate()`.
3244: Developer Note:
3245: The Fortran interface is not autogenerated as the
3246: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3248: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatFactorInfo`, `MatLUFactor()`, `MatCholeskyFactorSymbolic()`, `MatCholeskyFactorNumeric()`
3249: `MatGetOrdering()`
3250: @*/
3251: PetscErrorCode MatCholeskyFactor(Mat mat, IS perm, const MatFactorInfo *info)
3252: {
3253: MatFactorInfo tinfo;
3255: PetscFunctionBegin;
3260: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "Matrix must be square");
3261: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3262: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3263: MatCheckPreallocated(mat, 1);
3264: if (!info) {
3265: PetscCall(MatFactorInfoInitialize(&tinfo));
3266: info = &tinfo;
3267: }
3269: PetscCall(PetscLogEventBegin(MAT_CholeskyFactor, mat, perm, 0, 0));
3270: PetscUseTypeMethod(mat, choleskyfactor, perm, info);
3271: PetscCall(PetscLogEventEnd(MAT_CholeskyFactor, mat, perm, 0, 0));
3272: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3273: PetscFunctionReturn(PETSC_SUCCESS);
3274: }
3276: /*@C
3277: MatCholeskyFactorSymbolic - Performs symbolic Cholesky factorization
3278: of a symmetric matrix.
3280: Collective
3282: Input Parameters:
3283: + fact - the factor matrix obtained with `MatGetFactor()`
3284: . mat - the matrix
3285: . perm - row and column permutations
3286: - info - options for factorization, includes
3287: .vb
3288: fill - expected fill as ratio of original fill.
3289: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3290: Run with the option -info to determine an optimal value to use
3291: .ve
3293: Level: developer
3295: Notes:
3296: See `MatLUFactorSymbolic()` for the nonsymmetric case. See also
3297: `MatCholeskyFactor()` and `MatCholeskyFactorNumeric()`.
3299: Most users should employ the `KSP` interface for linear solvers
3300: instead of working directly with matrix algebra routines such as this.
3301: See, e.g., `KSPCreate()`.
3303: Developer Note:
3304: The Fortran interface is not autogenerated as the
3305: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3307: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactor()`, `MatCholeskyFactorNumeric()`
3308: `MatGetOrdering()`
3309: @*/
3310: PetscErrorCode MatCholeskyFactorSymbolic(Mat fact, Mat mat, IS perm, const MatFactorInfo *info)
3311: {
3312: MatFactorInfo tinfo;
3314: PetscFunctionBegin;
3321: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "Matrix must be square");
3322: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3323: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3324: MatCheckPreallocated(mat, 2);
3325: if (!info) {
3326: PetscCall(MatFactorInfoInitialize(&tinfo));
3327: info = &tinfo;
3328: }
3330: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_CholeskyFactorSymbolic, mat, perm, 0, 0));
3331: PetscUseTypeMethod(fact, choleskyfactorsymbolic, mat, perm, info);
3332: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_CholeskyFactorSymbolic, mat, perm, 0, 0));
3333: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3334: PetscFunctionReturn(PETSC_SUCCESS);
3335: }
3337: /*@C
3338: MatCholeskyFactorNumeric - Performs numeric Cholesky factorization
3339: of a symmetric matrix. Call this routine after first calling `MatGetFactor()` and
3340: `MatCholeskyFactorSymbolic()`.
3342: Collective
3344: Input Parameters:
3345: + fact - the factor matrix obtained with `MatGetFactor()`, where the factored values are stored
3346: . mat - the initial matrix that is to be factored
3347: - info - options for factorization
3349: Level: developer
3351: Note:
3352: Most users should employ the `KSP` interface for linear solvers
3353: instead of working directly with matrix algebra routines such as this.
3354: See, e.g., `KSPCreate()`.
3356: Developer Note:
3357: The Fortran interface is not autogenerated as the
3358: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3360: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatCholeskyFactorSymbolic()`, `MatCholeskyFactor()`, `MatLUFactorNumeric()`
3361: @*/
3362: PetscErrorCode MatCholeskyFactorNumeric(Mat fact, Mat mat, const MatFactorInfo *info)
3363: {
3364: MatFactorInfo tinfo;
3366: PetscFunctionBegin;
3372: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3373: 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,
3374: mat->rmap->N, (fact)->rmap->N, mat->cmap->N, (fact)->cmap->N);
3375: MatCheckPreallocated(mat, 2);
3376: if (!info) {
3377: PetscCall(MatFactorInfoInitialize(&tinfo));
3378: info = &tinfo;
3379: }
3381: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_CholeskyFactorNumeric, mat, fact, 0, 0));
3382: else PetscCall(PetscLogEventBegin(MAT_CholeskyFactor, mat, fact, 0, 0));
3383: PetscUseTypeMethod(fact, choleskyfactornumeric, mat, info);
3384: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_CholeskyFactorNumeric, mat, fact, 0, 0));
3385: else PetscCall(PetscLogEventEnd(MAT_CholeskyFactor, mat, fact, 0, 0));
3386: PetscCall(MatViewFromOptions(fact, NULL, "-mat_factor_view"));
3387: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3388: PetscFunctionReturn(PETSC_SUCCESS);
3389: }
3391: /*@
3392: MatQRFactor - Performs in-place QR factorization of matrix.
3394: Collective
3396: Input Parameters:
3397: + mat - the matrix
3398: . col - column permutation
3399: - info - options for factorization, includes
3400: .vb
3401: fill - expected fill as ratio of original fill.
3402: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3403: Run with the option -info to determine an optimal value to use
3404: .ve
3406: Level: developer
3408: Notes:
3409: Most users should employ the `KSP` interface for linear solvers
3410: instead of working directly with matrix algebra routines such as this.
3411: See, e.g., `KSPCreate()`.
3413: This changes the state of the matrix to a factored matrix; it cannot be used
3414: for example with `MatSetValues()` unless one first calls `MatSetUnfactored()`.
3416: Developer Note:
3417: The Fortran interface is not autogenerated as the
3418: interface definition cannot be generated correctly [due to MatFactorInfo]
3420: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatQRFactorSymbolic()`, `MatQRFactorNumeric()`, `MatLUFactor()`,
3421: `MatSetUnfactored()`, `MatFactorInfo`, `MatGetFactor()`
3422: @*/
3423: PetscErrorCode MatQRFactor(Mat mat, IS col, const MatFactorInfo *info)
3424: {
3425: PetscFunctionBegin;
3430: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3431: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3432: MatCheckPreallocated(mat, 1);
3433: PetscCall(PetscLogEventBegin(MAT_QRFactor, mat, col, 0, 0));
3434: PetscUseMethod(mat, "MatQRFactor_C", (Mat, IS, const MatFactorInfo *), (mat, col, info));
3435: PetscCall(PetscLogEventEnd(MAT_QRFactor, mat, col, 0, 0));
3436: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
3437: PetscFunctionReturn(PETSC_SUCCESS);
3438: }
3440: /*@
3441: MatQRFactorSymbolic - Performs symbolic QR factorization of matrix.
3442: Call this routine after `MatGetFactor()` but before calling `MatQRFactorNumeric()`.
3444: Collective
3446: Input Parameters:
3447: + fact - the factor matrix obtained with `MatGetFactor()`
3448: . mat - the matrix
3449: . col - column permutation
3450: - info - options for factorization, includes
3451: .vb
3452: fill - expected fill as ratio of original fill.
3453: dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3454: Run with the option -info to determine an optimal value to use
3455: .ve
3457: Level: developer
3459: Note:
3460: Most users should employ the `KSP` interface for linear solvers
3461: instead of working directly with matrix algebra routines such as this.
3462: See, e.g., `KSPCreate()`.
3464: Developer Note:
3465: The Fortran interface is not autogenerated as the
3466: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3468: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatFactorInfo`, `MatQRFactor()`, `MatQRFactorNumeric()`, `MatLUFactor()`, `MatFactorInfo`, `MatFactorInfoInitialize()`
3469: @*/
3470: PetscErrorCode MatQRFactorSymbolic(Mat fact, Mat mat, IS col, const MatFactorInfo *info)
3471: {
3472: MatFactorInfo tinfo;
3474: PetscFunctionBegin;
3481: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3482: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3483: MatCheckPreallocated(mat, 2);
3484: if (!info) {
3485: PetscCall(MatFactorInfoInitialize(&tinfo));
3486: info = &tinfo;
3487: }
3489: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_QRFactorSymbolic, fact, mat, col, 0));
3490: PetscUseMethod(fact, "MatQRFactorSymbolic_C", (Mat, Mat, IS, const MatFactorInfo *), (fact, mat, col, info));
3491: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_QRFactorSymbolic, fact, mat, col, 0));
3492: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3493: PetscFunctionReturn(PETSC_SUCCESS);
3494: }
3496: /*@
3497: MatQRFactorNumeric - Performs numeric QR factorization of a matrix.
3498: Call this routine after first calling `MatGetFactor()`, and `MatQRFactorSymbolic()`.
3500: Collective
3502: Input Parameters:
3503: + fact - the factor matrix obtained with `MatGetFactor()`
3504: . mat - the matrix
3505: - info - options for factorization
3507: Level: developer
3509: Notes:
3510: See `MatQRFactor()` for in-place factorization.
3512: Most users should employ the `KSP` interface for linear solvers
3513: instead of working directly with matrix algebra routines such as this.
3514: See, e.g., `KSPCreate()`.
3516: Developer Note:
3517: The Fortran interface is not autogenerated as the
3518: interface definition cannot be generated correctly [due to `MatFactorInfo`]
3520: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorInfo`, `MatGetFactor()`, `MatQRFactor()`, `MatQRFactorSymbolic()`, `MatLUFactor()`
3521: @*/
3522: PetscErrorCode MatQRFactorNumeric(Mat fact, Mat mat, const MatFactorInfo *info)
3523: {
3524: MatFactorInfo tinfo;
3526: PetscFunctionBegin;
3531: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3532: 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,
3533: mat->rmap->N, (fact)->rmap->N, mat->cmap->N, (fact)->cmap->N);
3535: MatCheckPreallocated(mat, 2);
3536: if (!info) {
3537: PetscCall(MatFactorInfoInitialize(&tinfo));
3538: info = &tinfo;
3539: }
3541: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_QRFactorNumeric, mat, fact, 0, 0));
3542: else PetscCall(PetscLogEventBegin(MAT_QRFactor, mat, fact, 0, 0));
3543: PetscUseMethod(fact, "MatQRFactorNumeric_C", (Mat, Mat, const MatFactorInfo *), (fact, mat, info));
3544: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_QRFactorNumeric, mat, fact, 0, 0));
3545: else PetscCall(PetscLogEventEnd(MAT_QRFactor, mat, fact, 0, 0));
3546: PetscCall(MatViewFromOptions(fact, NULL, "-mat_factor_view"));
3547: PetscCall(PetscObjectStateIncrease((PetscObject)fact));
3548: PetscFunctionReturn(PETSC_SUCCESS);
3549: }
3551: /*@
3552: MatSolve - Solves A x = b, given a factored matrix.
3554: Neighbor-wise Collective
3556: Input Parameters:
3557: + mat - the factored matrix
3558: - b - the right-hand-side vector
3560: Output Parameter:
3561: . x - the result vector
3563: Level: developer
3565: Notes:
3566: The vectors `b` and `x` cannot be the same. I.e., one cannot
3567: call `MatSolve`(A,x,x).
3569: Most users should employ the `KSP` interface for linear solvers
3570: instead of working directly with matrix algebra routines such as this.
3571: See, e.g., `KSPCreate()`.
3573: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatLUFactor()`, `MatSolveAdd()`, `MatSolveTranspose()`, `MatSolveTransposeAdd()`
3574: @*/
3575: PetscErrorCode MatSolve(Mat mat, Vec b, Vec x)
3576: {
3577: PetscFunctionBegin;
3582: PetscCheckSameComm(mat, 1, b, 2);
3583: PetscCheckSameComm(mat, 1, x, 3);
3584: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3585: 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);
3586: 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);
3587: 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);
3588: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3589: MatCheckPreallocated(mat, 1);
3591: PetscCall(PetscLogEventBegin(MAT_Solve, mat, b, x, 0));
3592: if (mat->factorerrortype) {
3593: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
3594: PetscCall(VecSetInf(x));
3595: } else PetscUseTypeMethod(mat, solve, b, x);
3596: PetscCall(PetscLogEventEnd(MAT_Solve, mat, b, x, 0));
3597: PetscCall(PetscObjectStateIncrease((PetscObject)x));
3598: PetscFunctionReturn(PETSC_SUCCESS);
3599: }
3601: static PetscErrorCode MatMatSolve_Basic(Mat A, Mat B, Mat X, PetscBool trans)
3602: {
3603: Vec b, x;
3604: PetscInt N, i;
3605: PetscErrorCode (*f)(Mat, Vec, Vec);
3606: PetscBool Abound, Bneedconv = PETSC_FALSE, Xneedconv = PETSC_FALSE;
3608: PetscFunctionBegin;
3609: if (A->factorerrortype) {
3610: PetscCall(PetscInfo(A, "MatFactorError %d\n", A->factorerrortype));
3611: PetscCall(MatSetInf(X));
3612: PetscFunctionReturn(PETSC_SUCCESS);
3613: }
3614: f = (!trans || (!A->ops->solvetranspose && A->symmetric)) ? A->ops->solve : A->ops->solvetranspose;
3615: PetscCheck(f, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Mat type %s", ((PetscObject)A)->type_name);
3616: PetscCall(MatBoundToCPU(A, &Abound));
3617: if (!Abound) {
3618: PetscCall(PetscObjectTypeCompareAny((PetscObject)B, &Bneedconv, MATSEQDENSE, MATMPIDENSE, ""));
3619: PetscCall(PetscObjectTypeCompareAny((PetscObject)X, &Xneedconv, MATSEQDENSE, MATMPIDENSE, ""));
3620: }
3621: #if defined(PETSC_HAVE_CUDA)
3622: if (Bneedconv) PetscCall(MatConvert(B, MATDENSECUDA, MAT_INPLACE_MATRIX, &B));
3623: if (Xneedconv) PetscCall(MatConvert(X, MATDENSECUDA, MAT_INPLACE_MATRIX, &X));
3624: #elif (PETSC_HAVE_HIP)
3625: if (Bneedconv) PetscCall(MatConvert(B, MATDENSEHIP, MAT_INPLACE_MATRIX, &B));
3626: if (Xneedconv) PetscCall(MatConvert(X, MATDENSEHIP, MAT_INPLACE_MATRIX, &X));
3627: #endif
3628: PetscCall(MatGetSize(B, NULL, &N));
3629: for (i = 0; i < N; i++) {
3630: PetscCall(MatDenseGetColumnVecRead(B, i, &b));
3631: PetscCall(MatDenseGetColumnVecWrite(X, i, &x));
3632: PetscCall((*f)(A, b, x));
3633: PetscCall(MatDenseRestoreColumnVecWrite(X, i, &x));
3634: PetscCall(MatDenseRestoreColumnVecRead(B, i, &b));
3635: }
3636: if (Bneedconv) PetscCall(MatConvert(B, MATDENSE, MAT_INPLACE_MATRIX, &B));
3637: if (Xneedconv) PetscCall(MatConvert(X, MATDENSE, MAT_INPLACE_MATRIX, &X));
3638: PetscFunctionReturn(PETSC_SUCCESS);
3639: }
3641: /*@
3642: MatMatSolve - Solves A X = B, given a factored matrix.
3644: Neighbor-wise Collective
3646: Input Parameters:
3647: + A - the factored matrix
3648: - B - the right-hand-side matrix `MATDENSE` (or sparse `MATAIJ`-- when using MUMPS)
3650: Output Parameter:
3651: . X - the result matrix (dense matrix)
3653: Level: developer
3655: Note:
3656: If `B` is a `MATDENSE` matrix then one can call `MatMatSolve`(A,B,B) except with `MATSOLVERMKL_CPARDISO`;
3657: otherwise, `B` and `X` cannot be the same.
3659: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSolve()`, `MatMatSolveTranspose()`, `MatLUFactor()`, `MatCholeskyFactor()`
3660: @*/
3661: PetscErrorCode MatMatSolve(Mat A, Mat B, Mat X)
3662: {
3663: PetscFunctionBegin;
3668: PetscCheckSameComm(A, 1, B, 2);
3669: PetscCheckSameComm(A, 1, X, 3);
3670: 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);
3671: 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);
3672: 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");
3673: if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3674: PetscCheck(A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
3675: MatCheckPreallocated(A, 1);
3677: PetscCall(PetscLogEventBegin(MAT_MatSolve, A, B, X, 0));
3678: if (!A->ops->matsolve) {
3679: PetscCall(PetscInfo(A, "Mat type %s using basic MatMatSolve\n", ((PetscObject)A)->type_name));
3680: PetscCall(MatMatSolve_Basic(A, B, X, PETSC_FALSE));
3681: } else PetscUseTypeMethod(A, matsolve, B, X);
3682: PetscCall(PetscLogEventEnd(MAT_MatSolve, A, B, X, 0));
3683: PetscCall(PetscObjectStateIncrease((PetscObject)X));
3684: PetscFunctionReturn(PETSC_SUCCESS);
3685: }
3687: /*@
3688: MatMatSolveTranspose - Solves A^T X = B, given a factored matrix.
3690: Neighbor-wise Collective
3692: Input Parameters:
3693: + A - the factored matrix
3694: - B - the right-hand-side matrix (`MATDENSE` matrix)
3696: Output Parameter:
3697: . X - the result matrix (dense matrix)
3699: Level: developer
3701: Note:
3702: The matrices `B` and `X` cannot be the same. I.e., one cannot
3703: call `MatMatSolveTranspose`(A,X,X).
3705: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSolveTranspose()`, `MatMatSolve()`, `MatLUFactor()`, `MatCholeskyFactor()`
3706: @*/
3707: PetscErrorCode MatMatSolveTranspose(Mat A, Mat B, Mat X)
3708: {
3709: PetscFunctionBegin;
3714: PetscCheckSameComm(A, 1, B, 2);
3715: PetscCheckSameComm(A, 1, X, 3);
3716: PetscCheck(X != B, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_IDN, "X and B must be different matrices");
3717: 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);
3718: 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);
3719: 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);
3720: 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");
3721: if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3722: PetscCheck(A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
3723: MatCheckPreallocated(A, 1);
3725: PetscCall(PetscLogEventBegin(MAT_MatSolve, A, B, X, 0));
3726: if (!A->ops->matsolvetranspose) {
3727: PetscCall(PetscInfo(A, "Mat type %s using basic MatMatSolveTranspose\n", ((PetscObject)A)->type_name));
3728: PetscCall(MatMatSolve_Basic(A, B, X, PETSC_TRUE));
3729: } else PetscUseTypeMethod(A, matsolvetranspose, B, X);
3730: PetscCall(PetscLogEventEnd(MAT_MatSolve, A, B, X, 0));
3731: PetscCall(PetscObjectStateIncrease((PetscObject)X));
3732: PetscFunctionReturn(PETSC_SUCCESS);
3733: }
3735: /*@
3736: MatMatTransposeSolve - Solves A X = B^T, given a factored matrix.
3738: Neighbor-wise Collective
3740: Input Parameters:
3741: + A - the factored matrix
3742: - Bt - the transpose of right-hand-side matrix as a `MATDENSE`
3744: Output Parameter:
3745: . X - the result matrix (dense matrix)
3747: Level: developer
3749: Note:
3750: For MUMPS, it only supports centralized sparse compressed column format on the host processor for right hand side matrix. User must create B^T in sparse compressed row
3751: format on the host processor and call `MatMatTransposeSolve()` to implement MUMPS' `MatMatSolve()`.
3753: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatMatSolve()`, `MatMatSolveTranspose()`, `MatLUFactor()`, `MatCholeskyFactor()`
3754: @*/
3755: PetscErrorCode MatMatTransposeSolve(Mat A, Mat Bt, Mat X)
3756: {
3757: PetscFunctionBegin;
3762: PetscCheckSameComm(A, 1, Bt, 2);
3763: PetscCheckSameComm(A, 1, X, 3);
3765: PetscCheck(X != Bt, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_IDN, "X and B must be different matrices");
3766: 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);
3767: 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);
3768: 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");
3769: if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3770: PetscCheck(A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
3771: MatCheckPreallocated(A, 1);
3773: PetscCall(PetscLogEventBegin(MAT_MatTrSolve, A, Bt, X, 0));
3774: PetscUseTypeMethod(A, mattransposesolve, Bt, X);
3775: PetscCall(PetscLogEventEnd(MAT_MatTrSolve, A, Bt, X, 0));
3776: PetscCall(PetscObjectStateIncrease((PetscObject)X));
3777: PetscFunctionReturn(PETSC_SUCCESS);
3778: }
3780: /*@
3781: MatForwardSolve - Solves L x = b, given a factored matrix, A = LU, or
3782: U^T*D^(1/2) x = b, given a factored symmetric matrix, A = U^T*D*U,
3784: Neighbor-wise Collective
3786: Input Parameters:
3787: + mat - the factored matrix
3788: - b - the right-hand-side vector
3790: Output Parameter:
3791: . x - the result vector
3793: Level: developer
3795: Notes:
3796: `MatSolve()` should be used for most applications, as it performs
3797: a forward solve followed by a backward solve.
3799: The vectors `b` and `x` cannot be the same, i.e., one cannot
3800: call `MatForwardSolve`(A,x,x).
3802: For matrix in `MATSEQBAIJ` format with block size larger than 1,
3803: the diagonal blocks are not implemented as D = D^(1/2) * D^(1/2) yet.
3804: `MatForwardSolve()` solves U^T*D y = b, and
3805: `MatBackwardSolve()` solves U x = y.
3806: Thus they do not provide a symmetric preconditioner.
3808: .seealso: [](chapter_matrices), `Mat`, `MatBackwardSolve()`, `MatGetFactor()`, `MatSolve()`, `MatBackwardSolve()`
3809: @*/
3810: PetscErrorCode MatForwardSolve(Mat mat, Vec b, Vec x)
3811: {
3812: PetscFunctionBegin;
3817: PetscCheckSameComm(mat, 1, b, 2);
3818: PetscCheckSameComm(mat, 1, x, 3);
3819: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3820: 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);
3821: 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);
3822: 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);
3823: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3824: MatCheckPreallocated(mat, 1);
3826: PetscCall(PetscLogEventBegin(MAT_ForwardSolve, mat, b, x, 0));
3827: PetscUseTypeMethod(mat, forwardsolve, b, x);
3828: PetscCall(PetscLogEventEnd(MAT_ForwardSolve, mat, b, x, 0));
3829: PetscCall(PetscObjectStateIncrease((PetscObject)x));
3830: PetscFunctionReturn(PETSC_SUCCESS);
3831: }
3833: /*@
3834: MatBackwardSolve - Solves U x = b, given a factored matrix, A = LU.
3835: D^(1/2) U x = b, given a factored symmetric matrix, A = U^T*D*U,
3837: Neighbor-wise Collective
3839: Input Parameters:
3840: + mat - the factored matrix
3841: - b - the right-hand-side vector
3843: Output Parameter:
3844: . x - the result vector
3846: Level: developer
3848: Notes:
3849: `MatSolve()` should be used for most applications, as it performs
3850: a forward solve followed by a backward solve.
3852: The vectors `b` and `x` cannot be the same. I.e., one cannot
3853: call `MatBackwardSolve`(A,x,x).
3855: For matrix in `MATSEQBAIJ` format with block size larger than 1,
3856: the diagonal blocks are not implemented as D = D^(1/2) * D^(1/2) yet.
3857: `MatForwardSolve()` solves U^T*D y = b, and
3858: `MatBackwardSolve()` solves U x = y.
3859: Thus they do not provide a symmetric preconditioner.
3861: .seealso: [](chapter_matrices), `Mat`, `MatForwardSolve()`, `MatGetFactor()`, `MatSolve()`, `MatForwardSolve()`
3862: @*/
3863: PetscErrorCode MatBackwardSolve(Mat mat, Vec b, Vec x)
3864: {
3865: PetscFunctionBegin;
3870: PetscCheckSameComm(mat, 1, b, 2);
3871: PetscCheckSameComm(mat, 1, x, 3);
3872: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3873: 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);
3874: 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);
3875: 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);
3876: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3877: MatCheckPreallocated(mat, 1);
3879: PetscCall(PetscLogEventBegin(MAT_BackwardSolve, mat, b, x, 0));
3880: PetscUseTypeMethod(mat, backwardsolve, b, x);
3881: PetscCall(PetscLogEventEnd(MAT_BackwardSolve, mat, b, x, 0));
3882: PetscCall(PetscObjectStateIncrease((PetscObject)x));
3883: PetscFunctionReturn(PETSC_SUCCESS);
3884: }
3886: /*@
3887: MatSolveAdd - Computes x = y + inv(A)*b, given a factored matrix.
3889: Neighbor-wise Collective
3891: Input Parameters:
3892: + mat - the factored matrix
3893: . b - the right-hand-side vector
3894: - y - the vector to be added to
3896: Output Parameter:
3897: . x - the result vector
3899: Level: developer
3901: Note:
3902: The vectors `b` and `x` cannot be the same. I.e., one cannot
3903: call `MatSolveAdd`(A,x,y,x).
3905: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatSolve()`, `MatGetFactor()`, `MatSolveTranspose()`, `MatSolveTransposeAdd()`
3906: @*/
3907: PetscErrorCode MatSolveAdd(Mat mat, Vec b, Vec y, Vec x)
3908: {
3909: PetscScalar one = 1.0;
3910: Vec tmp;
3912: PetscFunctionBegin;
3918: PetscCheckSameComm(mat, 1, b, 2);
3919: PetscCheckSameComm(mat, 1, y, 3);
3920: PetscCheckSameComm(mat, 1, x, 4);
3921: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3922: 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);
3923: 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);
3924: 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);
3925: 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);
3926: 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);
3927: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3928: MatCheckPreallocated(mat, 1);
3930: PetscCall(PetscLogEventBegin(MAT_SolveAdd, mat, b, x, y));
3931: if (mat->factorerrortype) {
3932: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
3933: PetscCall(VecSetInf(x));
3934: } else if (mat->ops->solveadd) {
3935: PetscUseTypeMethod(mat, solveadd, b, y, x);
3936: } else {
3937: /* do the solve then the add manually */
3938: if (x != y) {
3939: PetscCall(MatSolve(mat, b, x));
3940: PetscCall(VecAXPY(x, one, y));
3941: } else {
3942: PetscCall(VecDuplicate(x, &tmp));
3943: PetscCall(VecCopy(x, tmp));
3944: PetscCall(MatSolve(mat, b, x));
3945: PetscCall(VecAXPY(x, one, tmp));
3946: PetscCall(VecDestroy(&tmp));
3947: }
3948: }
3949: PetscCall(PetscLogEventEnd(MAT_SolveAdd, mat, b, x, y));
3950: PetscCall(PetscObjectStateIncrease((PetscObject)x));
3951: PetscFunctionReturn(PETSC_SUCCESS);
3952: }
3954: /*@
3955: MatSolveTranspose - Solves A' x = b, given a factored matrix.
3957: Neighbor-wise Collective
3959: Input Parameters:
3960: + mat - the factored matrix
3961: - b - the right-hand-side vector
3963: Output Parameter:
3964: . x - the result vector
3966: Level: developer
3968: Notes:
3969: The vectors `b` and `x` cannot be the same. I.e., one cannot
3970: call `MatSolveTranspose`(A,x,x).
3972: Most users should employ the `KSP` interface for linear solvers
3973: instead of working directly with matrix algebra routines such as this.
3974: See, e.g., `KSPCreate()`.
3976: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `KSP`, `MatSolve()`, `MatSolveAdd()`, `MatSolveTransposeAdd()`
3977: @*/
3978: PetscErrorCode MatSolveTranspose(Mat mat, Vec b, Vec x)
3979: {
3980: PetscErrorCode (*f)(Mat, Vec, Vec) = (!mat->ops->solvetranspose && mat->symmetric) ? mat->ops->solve : mat->ops->solvetranspose;
3982: PetscFunctionBegin;
3987: PetscCheckSameComm(mat, 1, b, 2);
3988: PetscCheckSameComm(mat, 1, x, 3);
3989: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
3990: 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);
3991: 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);
3992: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
3993: MatCheckPreallocated(mat, 1);
3994: PetscCall(PetscLogEventBegin(MAT_SolveTranspose, mat, b, x, 0));
3995: if (mat->factorerrortype) {
3996: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
3997: PetscCall(VecSetInf(x));
3998: } else {
3999: PetscCheck(f, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Matrix type %s", ((PetscObject)mat)->type_name);
4000: PetscCall((*f)(mat, b, x));
4001: }
4002: PetscCall(PetscLogEventEnd(MAT_SolveTranspose, mat, b, x, 0));
4003: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4004: PetscFunctionReturn(PETSC_SUCCESS);
4005: }
4007: /*@
4008: MatSolveTransposeAdd - Computes x = y + inv(Transpose(A)) b, given a
4009: factored matrix.
4011: Neighbor-wise Collective
4013: Input Parameters:
4014: + mat - the factored matrix
4015: . b - the right-hand-side vector
4016: - y - the vector to be added to
4018: Output Parameter:
4019: . x - the result vector
4021: Level: developer
4023: Note:
4024: The vectors `b` and `x` cannot be the same. I.e., one cannot
4025: call `MatSolveTransposeAdd`(A,x,y,x).
4027: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `MatSolve()`, `MatSolveAdd()`, `MatSolveTranspose()`
4028: @*/
4029: PetscErrorCode MatSolveTransposeAdd(Mat mat, Vec b, Vec y, Vec x)
4030: {
4031: PetscScalar one = 1.0;
4032: Vec tmp;
4033: PetscErrorCode (*f)(Mat, Vec, Vec, Vec) = (!mat->ops->solvetransposeadd && mat->symmetric) ? mat->ops->solveadd : mat->ops->solvetransposeadd;
4035: PetscFunctionBegin;
4041: PetscCheckSameComm(mat, 1, b, 2);
4042: PetscCheckSameComm(mat, 1, y, 3);
4043: PetscCheckSameComm(mat, 1, x, 4);
4044: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
4045: 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);
4046: 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);
4047: 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);
4048: 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);
4049: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
4050: MatCheckPreallocated(mat, 1);
4052: PetscCall(PetscLogEventBegin(MAT_SolveTransposeAdd, mat, b, x, y));
4053: if (mat->factorerrortype) {
4054: PetscCall(PetscInfo(mat, "MatFactorError %d\n", mat->factorerrortype));
4055: PetscCall(VecSetInf(x));
4056: } else if (f) {
4057: PetscCall((*f)(mat, b, y, x));
4058: } else {
4059: /* do the solve then the add manually */
4060: if (x != y) {
4061: PetscCall(MatSolveTranspose(mat, b, x));
4062: PetscCall(VecAXPY(x, one, y));
4063: } else {
4064: PetscCall(VecDuplicate(x, &tmp));
4065: PetscCall(VecCopy(x, tmp));
4066: PetscCall(MatSolveTranspose(mat, b, x));
4067: PetscCall(VecAXPY(x, one, tmp));
4068: PetscCall(VecDestroy(&tmp));
4069: }
4070: }
4071: PetscCall(PetscLogEventEnd(MAT_SolveTransposeAdd, mat, b, x, y));
4072: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4073: PetscFunctionReturn(PETSC_SUCCESS);
4074: }
4076: /*@
4077: MatSOR - Computes relaxation (SOR, Gauss-Seidel) sweeps.
4079: Neighbor-wise Collective
4081: Input Parameters:
4082: + mat - the matrix
4083: . b - the right hand side
4084: . omega - the relaxation factor
4085: . flag - flag indicating the type of SOR (see below)
4086: . shift - diagonal shift
4087: . its - the number of iterations
4088: - lits - the number of local iterations
4090: Output Parameter:
4091: . x - the solution (can contain an initial guess, use option `SOR_ZERO_INITIAL_GUESS` to indicate no guess)
4093: SOR Flags:
4094: + `SOR_FORWARD_SWEEP` - forward SOR
4095: . `SOR_BACKWARD_SWEEP` - backward SOR
4096: . `SOR_SYMMETRIC_SWEEP` - SSOR (symmetric SOR)
4097: . `SOR_LOCAL_FORWARD_SWEEP` - local forward SOR
4098: . `SOR_LOCAL_BACKWARD_SWEEP` - local forward SOR
4099: . `SOR_LOCAL_SYMMETRIC_SWEEP` - local SSOR
4100: . `SOR_EISENSTAT` - SOR with Eisenstat trick
4101: . `SOR_APPLY_UPPER`, `SOR_APPLY_LOWER` - applies
4102: upper/lower triangular part of matrix to
4103: vector (with omega)
4104: - `SOR_ZERO_INITIAL_GUESS` - zero initial guess
4106: Level: developer
4108: Notes:
4109: `SOR_LOCAL_FORWARD_SWEEP`, `SOR_LOCAL_BACKWARD_SWEEP`, and
4110: `SOR_LOCAL_SYMMETRIC_SWEEP` perform separate independent smoothings
4111: on each processor.
4113: Application programmers will not generally use `MatSOR()` directly,
4114: but instead will employ the `KSP`/`PC` interface.
4116: For `MATBAIJ`, `MATSBAIJ`, and `MATAIJ` matrices with Inodes this does a block SOR smoothing, otherwise it does a pointwise smoothing
4118: Most users should employ the `KSP` interface for linear solvers
4119: instead of working directly with matrix algebra routines such as this.
4120: See, e.g., `KSPCreate()`.
4122: Vectors `x` and `b` CANNOT be the same
4124: The flags are implemented as bitwise inclusive or operations.
4125: For example, use (`SOR_ZERO_INITIAL_GUESS` | `SOR_SYMMETRIC_SWEEP`)
4126: to specify a zero initial guess for SSOR.
4128: Developer Note:
4129: We should add block SOR support for `MATAIJ` matrices with block size set to great than one and no inodes
4131: .seealso: [](chapter_matrices), `Mat`, `MatMult()`, `KSP`, `PC`, `MatGetFactor()`
4132: @*/
4133: PetscErrorCode MatSOR(Mat mat, Vec b, PetscReal omega, MatSORType flag, PetscReal shift, PetscInt its, PetscInt lits, Vec x)
4134: {
4135: PetscFunctionBegin;
4140: PetscCheckSameComm(mat, 1, b, 2);
4141: PetscCheckSameComm(mat, 1, x, 8);
4142: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4143: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4144: 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);
4145: 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);
4146: 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);
4147: PetscCheck(its > 0, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Relaxation requires global its %" PetscInt_FMT " positive", its);
4148: PetscCheck(lits > 0, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Relaxation requires local its %" PetscInt_FMT " positive", lits);
4149: PetscCheck(b != x, PETSC_COMM_SELF, PETSC_ERR_ARG_IDN, "b and x vector cannot be the same");
4151: MatCheckPreallocated(mat, 1);
4152: PetscCall(PetscLogEventBegin(MAT_SOR, mat, b, x, 0));
4153: PetscUseTypeMethod(mat, sor, b, omega, flag, shift, its, lits, x);
4154: PetscCall(PetscLogEventEnd(MAT_SOR, mat, b, x, 0));
4155: PetscCall(PetscObjectStateIncrease((PetscObject)x));
4156: PetscFunctionReturn(PETSC_SUCCESS);
4157: }
4159: /*
4160: Default matrix copy routine.
4161: */
4162: PetscErrorCode MatCopy_Basic(Mat A, Mat B, MatStructure str)
4163: {
4164: PetscInt i, rstart = 0, rend = 0, nz;
4165: const PetscInt *cwork;
4166: const PetscScalar *vwork;
4168: PetscFunctionBegin;
4169: if (B->assembled) PetscCall(MatZeroEntries(B));
4170: if (str == SAME_NONZERO_PATTERN) {
4171: PetscCall(MatGetOwnershipRange(A, &rstart, &rend));
4172: for (i = rstart; i < rend; i++) {
4173: PetscCall(MatGetRow(A, i, &nz, &cwork, &vwork));
4174: PetscCall(MatSetValues(B, 1, &i, nz, cwork, vwork, INSERT_VALUES));
4175: PetscCall(MatRestoreRow(A, i, &nz, &cwork, &vwork));
4176: }
4177: } else {
4178: PetscCall(MatAYPX(B, 0.0, A, str));
4179: }
4180: PetscCall(MatAssemblyBegin(B, MAT_FINAL_ASSEMBLY));
4181: PetscCall(MatAssemblyEnd(B, MAT_FINAL_ASSEMBLY));
4182: PetscFunctionReturn(PETSC_SUCCESS);
4183: }
4185: /*@
4186: MatCopy - Copies a matrix to another matrix.
4188: Collective
4190: Input Parameters:
4191: + A - the matrix
4192: - str - `SAME_NONZERO_PATTERN` or `DIFFERENT_NONZERO_PATTERN`
4194: Output Parameter:
4195: . B - where the copy is put
4197: Level: intermediate
4199: Notes:
4200: If you use `SAME_NONZERO_PATTERN` then the two matrices must have the same nonzero pattern or the routine will crash.
4202: `MatCopy()` copies the matrix entries of a matrix to another existing
4203: matrix (after first zeroing the second matrix). A related routine is
4204: `MatConvert()`, which first creates a new matrix and then copies the data.
4206: .seealso: [](chapter_matrices), `Mat`, `MatConvert()`, `MatDuplicate()`
4207: @*/
4208: PetscErrorCode MatCopy(Mat A, Mat B, MatStructure str)
4209: {
4210: PetscInt i;
4212: PetscFunctionBegin;
4217: PetscCheckSameComm(A, 1, B, 2);
4218: MatCheckPreallocated(B, 2);
4219: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4220: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4221: 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,
4222: A->cmap->N, B->cmap->N);
4223: MatCheckPreallocated(A, 1);
4224: if (A == B) PetscFunctionReturn(PETSC_SUCCESS);
4226: PetscCall(PetscLogEventBegin(MAT_Copy, A, B, 0, 0));
4227: if (A->ops->copy) PetscUseTypeMethod(A, copy, B, str);
4228: else PetscCall(MatCopy_Basic(A, B, str));
4230: B->stencil.dim = A->stencil.dim;
4231: B->stencil.noc = A->stencil.noc;
4232: for (i = 0; i <= A->stencil.dim; i++) {
4233: B->stencil.dims[i] = A->stencil.dims[i];
4234: B->stencil.starts[i] = A->stencil.starts[i];
4235: }
4237: PetscCall(PetscLogEventEnd(MAT_Copy, A, B, 0, 0));
4238: PetscCall(PetscObjectStateIncrease((PetscObject)B));
4239: PetscFunctionReturn(PETSC_SUCCESS);
4240: }
4242: /*@C
4243: MatConvert - Converts a matrix to another matrix, either of the same
4244: or different type.
4246: Collective
4248: Input Parameters:
4249: + mat - the matrix
4250: . newtype - new matrix type. Use `MATSAME` to create a new matrix of the
4251: same type as the original matrix.
4252: - reuse - denotes if the destination matrix is to be created or reused.
4253: 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
4254: `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).
4256: Output Parameter:
4257: . M - pointer to place new matrix
4259: Level: intermediate
4261: Notes:
4262: `MatConvert()` first creates a new matrix and then copies the data from
4263: the first matrix. A related routine is `MatCopy()`, which copies the matrix
4264: entries of one matrix to another already existing matrix context.
4266: Cannot be used to convert a sequential matrix to parallel or parallel to sequential,
4267: the MPI communicator of the generated matrix is always the same as the communicator
4268: of the input matrix.
4270: .seealso: [](chapter_matrices), `Mat`, `MatCopy()`, `MatDuplicate()`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`
4271: @*/
4272: PetscErrorCode MatConvert(Mat mat, MatType newtype, MatReuse reuse, Mat *M)
4273: {
4274: PetscBool sametype, issame, flg;
4275: PetscBool3 issymmetric, ishermitian;
4276: char convname[256], mtype[256];
4277: Mat B;
4279: PetscFunctionBegin;
4283: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4284: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4285: MatCheckPreallocated(mat, 1);
4287: PetscCall(PetscOptionsGetString(((PetscObject)mat)->options, ((PetscObject)mat)->prefix, "-matconvert_type", mtype, sizeof(mtype), &flg));
4288: if (flg) newtype = mtype;
4290: PetscCall(PetscObjectTypeCompare((PetscObject)mat, newtype, &sametype));
4291: PetscCall(PetscStrcmp(newtype, "same", &issame));
4292: PetscCheck(!(reuse == MAT_INPLACE_MATRIX) || !(mat != *M), PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "MAT_INPLACE_MATRIX requires same input and output matrix");
4293: PetscCheck(!(reuse == MAT_REUSE_MATRIX) || !(mat == *M), PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "MAT_REUSE_MATRIX means reuse matrix in final argument, perhaps you mean MAT_INPLACE_MATRIX");
4295: if ((reuse == MAT_INPLACE_MATRIX) && (issame || sametype)) {
4296: PetscCall(PetscInfo(mat, "Early return for inplace %s %d %d\n", ((PetscObject)mat)->type_name, sametype, issame));
4297: PetscFunctionReturn(PETSC_SUCCESS);
4298: }
4300: /* Cache Mat options because some converters use MatHeaderReplace */
4301: issymmetric = mat->symmetric;
4302: ishermitian = mat->hermitian;
4304: if ((sametype || issame) && (reuse == MAT_INITIAL_MATRIX) && mat->ops->duplicate) {
4305: PetscCall(PetscInfo(mat, "Calling duplicate for initial matrix %s %d %d\n", ((PetscObject)mat)->type_name, sametype, issame));
4306: PetscUseTypeMethod(mat, duplicate, MAT_COPY_VALUES, M);
4307: } else {
4308: PetscErrorCode (*conv)(Mat, MatType, MatReuse, Mat *) = NULL;
4309: const char *prefix[3] = {"seq", "mpi", ""};
4310: PetscInt i;
4311: /*
4312: Order of precedence:
4313: 0) See if newtype is a superclass of the current matrix.
4314: 1) See if a specialized converter is known to the current matrix.
4315: 2) See if a specialized converter is known to the desired matrix class.
4316: 3) See if a good general converter is registered for the desired class
4317: (as of 6/27/03 only MATMPIADJ falls into this category).
4318: 4) See if a good general converter is known for the current matrix.
4319: 5) Use a really basic converter.
4320: */
4322: /* 0) See if newtype is a superclass of the current matrix.
4323: i.e mat is mpiaij and newtype is aij */
4324: for (i = 0; i < 2; i++) {
4325: PetscCall(PetscStrncpy(convname, prefix[i], sizeof(convname)));
4326: PetscCall(PetscStrlcat(convname, newtype, sizeof(convname)));
4327: PetscCall(PetscStrcmp(convname, ((PetscObject)mat)->type_name, &flg));
4328: PetscCall(PetscInfo(mat, "Check superclass %s %s -> %d\n", convname, ((PetscObject)mat)->type_name, flg));
4329: if (flg) {
4330: if (reuse == MAT_INPLACE_MATRIX) {
4331: PetscCall(PetscInfo(mat, "Early return\n"));
4332: PetscFunctionReturn(PETSC_SUCCESS);
4333: } else if (reuse == MAT_INITIAL_MATRIX && mat->ops->duplicate) {
4334: PetscCall(PetscInfo(mat, "Calling MatDuplicate\n"));
4335: PetscUseTypeMethod(mat, duplicate, MAT_COPY_VALUES, M);
4336: PetscFunctionReturn(PETSC_SUCCESS);
4337: } else if (reuse == MAT_REUSE_MATRIX && mat->ops->copy) {
4338: PetscCall(PetscInfo(mat, "Calling MatCopy\n"));
4339: PetscCall(MatCopy(mat, *M, SAME_NONZERO_PATTERN));
4340: PetscFunctionReturn(PETSC_SUCCESS);
4341: }
4342: }
4343: }
4344: /* 1) See if a specialized converter is known to the current matrix and the desired class */
4345: for (i = 0; i < 3; i++) {
4346: PetscCall(PetscStrncpy(convname, "MatConvert_", sizeof(convname)));
4347: PetscCall(PetscStrlcat(convname, ((PetscObject)mat)->type_name, sizeof(convname)));
4348: PetscCall(PetscStrlcat(convname, "_", sizeof(convname)));
4349: PetscCall(PetscStrlcat(convname, prefix[i], sizeof(convname)));
4350: PetscCall(PetscStrlcat(convname, issame ? ((PetscObject)mat)->type_name : newtype, sizeof(convname)));
4351: PetscCall(PetscStrlcat(convname, "_C", sizeof(convname)));
4352: PetscCall(PetscObjectQueryFunction((PetscObject)mat, convname, &conv));
4353: PetscCall(PetscInfo(mat, "Check specialized (1) %s (%s) -> %d\n", convname, ((PetscObject)mat)->type_name, !!conv));
4354: if (conv) goto foundconv;
4355: }
4357: /* 2) See if a specialized converter is known to the desired matrix class. */
4358: PetscCall(MatCreate(PetscObjectComm((PetscObject)mat), &B));
4359: PetscCall(MatSetSizes(B, mat->rmap->n, mat->cmap->n, mat->rmap->N, mat->cmap->N));
4360: PetscCall(MatSetType(B, newtype));
4361: for (i = 0; i < 3; i++) {
4362: PetscCall(PetscStrncpy(convname, "MatConvert_", sizeof(convname)));
4363: PetscCall(PetscStrlcat(convname, ((PetscObject)mat)->type_name, sizeof(convname)));
4364: PetscCall(PetscStrlcat(convname, "_", sizeof(convname)));
4365: PetscCall(PetscStrlcat(convname, prefix[i], sizeof(convname)));
4366: PetscCall(PetscStrlcat(convname, newtype, sizeof(convname)));
4367: PetscCall(PetscStrlcat(convname, "_C", sizeof(convname)));
4368: PetscCall(PetscObjectQueryFunction((PetscObject)B, convname, &conv));
4369: PetscCall(PetscInfo(mat, "Check specialized (2) %s (%s) -> %d\n", convname, ((PetscObject)B)->type_name, !!conv));
4370: if (conv) {
4371: PetscCall(MatDestroy(&B));
4372: goto foundconv;
4373: }
4374: }
4376: /* 3) See if a good general converter is registered for the desired class */
4377: conv = B->ops->convertfrom;
4378: PetscCall(PetscInfo(mat, "Check convertfrom (%s) -> %d\n", ((PetscObject)B)->type_name, !!conv));
4379: PetscCall(MatDestroy(&B));
4380: if (conv) goto foundconv;
4382: /* 4) See if a good general converter is known for the current matrix */
4383: if (mat->ops->convert) conv = mat->ops->convert;
4384: PetscCall(PetscInfo(mat, "Check general convert (%s) -> %d\n", ((PetscObject)mat)->type_name, !!conv));
4385: if (conv) goto foundconv;
4387: /* 5) Use a really basic converter. */
4388: PetscCall(PetscInfo(mat, "Using MatConvert_Basic\n"));
4389: conv = MatConvert_Basic;
4391: foundconv:
4392: PetscCall(PetscLogEventBegin(MAT_Convert, mat, 0, 0, 0));
4393: PetscCall((*conv)(mat, newtype, reuse, M));
4394: if (mat->rmap->mapping && mat->cmap->mapping && !(*M)->rmap->mapping && !(*M)->cmap->mapping) {
4395: /* the block sizes must be same if the mappings are copied over */
4396: (*M)->rmap->bs = mat->rmap->bs;
4397: (*M)->cmap->bs = mat->cmap->bs;
4398: PetscCall(PetscObjectReference((PetscObject)mat->rmap->mapping));
4399: PetscCall(PetscObjectReference((PetscObject)mat->cmap->mapping));
4400: (*M)->rmap->mapping = mat->rmap->mapping;
4401: (*M)->cmap->mapping = mat->cmap->mapping;
4402: }
4403: (*M)->stencil.dim = mat->stencil.dim;
4404: (*M)->stencil.noc = mat->stencil.noc;
4405: for (i = 0; i <= mat->stencil.dim; i++) {
4406: (*M)->stencil.dims[i] = mat->stencil.dims[i];
4407: (*M)->stencil.starts[i] = mat->stencil.starts[i];
4408: }
4409: PetscCall(PetscLogEventEnd(MAT_Convert, mat, 0, 0, 0));
4410: }
4411: PetscCall(PetscObjectStateIncrease((PetscObject)*M));
4413: /* Copy Mat options */
4414: if (issymmetric == PETSC_BOOL3_TRUE) PetscCall(MatSetOption(*M, MAT_SYMMETRIC, PETSC_TRUE));
4415: else if (issymmetric == PETSC_BOOL3_FALSE) PetscCall(MatSetOption(*M, MAT_SYMMETRIC, PETSC_FALSE));
4416: if (ishermitian == PETSC_BOOL3_TRUE) PetscCall(MatSetOption(*M, MAT_HERMITIAN, PETSC_TRUE));
4417: else if (ishermitian == PETSC_BOOL3_FALSE) PetscCall(MatSetOption(*M, MAT_HERMITIAN, PETSC_FALSE));
4418: PetscFunctionReturn(PETSC_SUCCESS);
4419: }
4421: /*@C
4422: MatFactorGetSolverType - Returns name of the package providing the factorization routines
4424: Not Collective
4426: Input Parameter:
4427: . mat - the matrix, must be a factored matrix
4429: Output Parameter:
4430: . type - the string name of the package (do not free this string)
4432: Level: intermediate
4434: Fortran Note:
4435: Pass in an empty string and the package name will be copied into it. Make sure the string is long enough.
4437: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatSolverType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatGetFactor()`
4438: @*/
4439: PetscErrorCode MatFactorGetSolverType(Mat mat, MatSolverType *type)
4440: {
4441: PetscErrorCode (*conv)(Mat, MatSolverType *);
4443: PetscFunctionBegin;
4447: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Only for factored matrix");
4448: PetscCall(PetscObjectQueryFunction((PetscObject)mat, "MatFactorGetSolverType_C", &conv));
4449: if (conv) PetscCall((*conv)(mat, type));
4450: else *type = MATSOLVERPETSC;
4451: PetscFunctionReturn(PETSC_SUCCESS);
4452: }
4454: typedef struct _MatSolverTypeForSpecifcType *MatSolverTypeForSpecifcType;
4455: struct _MatSolverTypeForSpecifcType {
4456: MatType mtype;
4457: /* no entry for MAT_FACTOR_NONE */
4458: PetscErrorCode (*createfactor[MAT_FACTOR_NUM_TYPES - 1])(Mat, MatFactorType, Mat *);
4459: MatSolverTypeForSpecifcType next;
4460: };
4462: typedef struct _MatSolverTypeHolder *MatSolverTypeHolder;
4463: struct _MatSolverTypeHolder {
4464: char *name;
4465: MatSolverTypeForSpecifcType handlers;
4466: MatSolverTypeHolder next;
4467: };
4469: static MatSolverTypeHolder MatSolverTypeHolders = NULL;
4471: /*@C
4472: MatSolverTypeRegister - Registers a `MatSolverType` that works for a particular matrix type
4474: Input Parameters:
4475: + package - name of the package, for example petsc or superlu
4476: . mtype - the matrix type that works with this package
4477: . ftype - the type of factorization supported by the package
4478: - createfactor - routine that will create the factored matrix ready to be used
4480: Level: developer
4482: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorGetSolverType()`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatGetFactor()`
4483: @*/
4484: PetscErrorCode MatSolverTypeRegister(MatSolverType package, MatType mtype, MatFactorType ftype, PetscErrorCode (*createfactor)(Mat, MatFactorType, Mat *))
4485: {
4486: MatSolverTypeHolder next = MatSolverTypeHolders, prev = NULL;
4487: PetscBool flg;
4488: MatSolverTypeForSpecifcType inext, iprev = NULL;
4490: PetscFunctionBegin;
4491: PetscCall(MatInitializePackage());
4492: if (!next) {
4493: PetscCall(PetscNew(&MatSolverTypeHolders));
4494: PetscCall(PetscStrallocpy(package, &MatSolverTypeHolders->name));
4495: PetscCall(PetscNew(&MatSolverTypeHolders->handlers));
4496: PetscCall(PetscStrallocpy(mtype, (char **)&MatSolverTypeHolders->handlers->mtype));
4497: MatSolverTypeHolders->handlers->createfactor[(int)ftype - 1] = createfactor;
4498: PetscFunctionReturn(PETSC_SUCCESS);
4499: }
4500: while (next) {
4501: PetscCall(PetscStrcasecmp(package, next->name, &flg));
4502: if (flg) {
4503: PetscCheck(next->handlers, PETSC_COMM_SELF, PETSC_ERR_PLIB, "MatSolverTypeHolder is missing handlers");
4504: inext = next->handlers;
4505: while (inext) {
4506: PetscCall(PetscStrcasecmp(mtype, inext->mtype, &flg));
4507: if (flg) {
4508: inext->createfactor[(int)ftype - 1] = createfactor;
4509: PetscFunctionReturn(PETSC_SUCCESS);
4510: }
4511: iprev = inext;
4512: inext = inext->next;
4513: }
4514: PetscCall(PetscNew(&iprev->next));
4515: PetscCall(PetscStrallocpy(mtype, (char **)&iprev->next->mtype));
4516: iprev->next->createfactor[(int)ftype - 1] = createfactor;
4517: PetscFunctionReturn(PETSC_SUCCESS);
4518: }
4519: prev = next;
4520: next = next->next;
4521: }
4522: PetscCall(PetscNew(&prev->next));
4523: PetscCall(PetscStrallocpy(package, &prev->next->name));
4524: PetscCall(PetscNew(&prev->next->handlers));
4525: PetscCall(PetscStrallocpy(mtype, (char **)&prev->next->handlers->mtype));
4526: prev->next->handlers->createfactor[(int)ftype - 1] = createfactor;
4527: PetscFunctionReturn(PETSC_SUCCESS);
4528: }
4530: /*@C
4531: MatSolverTypeGet - Gets the function that creates the factor matrix if it exist
4533: Input Parameters:
4534: + type - name of the package, for example petsc or superlu
4535: . ftype - the type of factorization supported by the type
4536: - mtype - the matrix type that works with this type
4538: Output Parameters:
4539: + foundtype - `PETSC_TRUE` if the type was registered
4540: . foundmtype - `PETSC_TRUE` if the type supports the requested mtype
4541: - createfactor - routine that will create the factored matrix ready to be used or `NULL` if not found
4543: Level: developer
4545: .seealso: [](chapter_matrices), `Mat`, `MatFactorType`, `MatType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatSolverTypeRegister()`, `MatGetFactor()`
4546: @*/
4547: PetscErrorCode MatSolverTypeGet(MatSolverType type, MatType mtype, MatFactorType ftype, PetscBool *foundtype, PetscBool *foundmtype, PetscErrorCode (**createfactor)(Mat, MatFactorType, Mat *))
4548: {
4549: MatSolverTypeHolder next = MatSolverTypeHolders;
4550: PetscBool flg;
4551: MatSolverTypeForSpecifcType inext;
4553: PetscFunctionBegin;
4554: if (foundtype) *foundtype = PETSC_FALSE;
4555: if (foundmtype) *foundmtype = PETSC_FALSE;
4556: if (createfactor) *createfactor = NULL;
4558: if (type) {
4559: while (next) {
4560: PetscCall(PetscStrcasecmp(type, next->name, &flg));
4561: if (flg) {
4562: if (foundtype) *foundtype = PETSC_TRUE;
4563: inext = next->handlers;
4564: while (inext) {
4565: PetscCall(PetscStrbeginswith(mtype, inext->mtype, &flg));
4566: if (flg) {
4567: if (foundmtype) *foundmtype = PETSC_TRUE;
4568: if (createfactor) *createfactor = inext->createfactor[(int)ftype - 1];
4569: PetscFunctionReturn(PETSC_SUCCESS);
4570: }
4571: inext = inext->next;
4572: }
4573: }
4574: next = next->next;
4575: }
4576: } else {
4577: while (next) {
4578: inext = next->handlers;
4579: while (inext) {
4580: PetscCall(PetscStrcmp(mtype, inext->mtype, &flg));
4581: if (flg && inext->createfactor[(int)ftype - 1]) {
4582: if (foundtype) *foundtype = PETSC_TRUE;
4583: if (foundmtype) *foundmtype = PETSC_TRUE;
4584: if (createfactor) *createfactor = inext->createfactor[(int)ftype - 1];
4585: PetscFunctionReturn(PETSC_SUCCESS);
4586: }
4587: inext = inext->next;
4588: }
4589: next = next->next;
4590: }
4591: /* try with base classes inext->mtype */
4592: next = MatSolverTypeHolders;
4593: while (next) {
4594: inext = next->handlers;
4595: while (inext) {
4596: PetscCall(PetscStrbeginswith(mtype, inext->mtype, &flg));
4597: if (flg && inext->createfactor[(int)ftype - 1]) {
4598: if (foundtype) *foundtype = PETSC_TRUE;
4599: if (foundmtype) *foundmtype = PETSC_TRUE;
4600: if (createfactor) *createfactor = inext->createfactor[(int)ftype - 1];
4601: PetscFunctionReturn(PETSC_SUCCESS);
4602: }
4603: inext = inext->next;
4604: }
4605: next = next->next;
4606: }
4607: }
4608: PetscFunctionReturn(PETSC_SUCCESS);
4609: }
4611: PetscErrorCode MatSolverTypeDestroy(void)
4612: {
4613: MatSolverTypeHolder next = MatSolverTypeHolders, prev;
4614: MatSolverTypeForSpecifcType inext, iprev;
4616: PetscFunctionBegin;
4617: while (next) {
4618: PetscCall(PetscFree(next->name));
4619: inext = next->handlers;
4620: while (inext) {
4621: PetscCall(PetscFree(inext->mtype));
4622: iprev = inext;
4623: inext = inext->next;
4624: PetscCall(PetscFree(iprev));
4625: }
4626: prev = next;
4627: next = next->next;
4628: PetscCall(PetscFree(prev));
4629: }
4630: MatSolverTypeHolders = NULL;
4631: PetscFunctionReturn(PETSC_SUCCESS);
4632: }
4634: /*@C
4635: MatFactorGetCanUseOrdering - Indicates if the factorization can use the ordering provided in `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`
4637: Logically Collective
4639: Input Parameter:
4640: . mat - the matrix
4642: Output Parameter:
4643: . flg - `PETSC_TRUE` if uses the ordering
4645: Level: developer
4647: Note:
4648: Most internal PETSc factorizations use the ordering passed to the factorization routine but external
4649: packages do not, thus we want to skip generating the ordering when it is not needed or used.
4651: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`
4652: @*/
4653: PetscErrorCode MatFactorGetCanUseOrdering(Mat mat, PetscBool *flg)
4654: {
4655: PetscFunctionBegin;
4656: *flg = mat->canuseordering;
4657: PetscFunctionReturn(PETSC_SUCCESS);
4658: }
4660: /*@C
4661: MatFactorGetPreferredOrdering - The preferred ordering for a particular matrix factor object
4663: Logically Collective
4665: Input Parameters:
4666: + mat - the matrix obtained with `MatGetFactor()`
4667: - ftype - the factorization type to be used
4669: Output Parameter:
4670: . otype - the preferred ordering type
4672: Level: developer
4674: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatFactorType`, `MatOrderingType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatCholeskyFactorSymbolic()`
4675: @*/
4676: PetscErrorCode MatFactorGetPreferredOrdering(Mat mat, MatFactorType ftype, MatOrderingType *otype)
4677: {
4678: PetscFunctionBegin;
4679: *otype = mat->preferredordering[ftype];
4680: PetscCheck(*otype, PETSC_COMM_SELF, PETSC_ERR_PLIB, "MatFactor did not have a preferred ordering");
4681: PetscFunctionReturn(PETSC_SUCCESS);
4682: }
4684: /*@C
4685: MatGetFactor - Returns a matrix suitable to calls to MatXXFactorSymbolic()
4687: Collective
4689: Input Parameters:
4690: + mat - the matrix
4691: . type - name of solver type, for example, superlu, petsc (to use PETSc's default)
4692: - ftype - factor type, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`
4694: Output Parameter:
4695: . f - the factor matrix used with MatXXFactorSymbolic() calls
4697: Options Database Key:
4698: . -mat_factor_bind_factorization <host, device> - Where to do matrix factorization? Default is device (might consume more device memory.
4699: One can choose host to save device memory). Currently only supported with `MATSEQAIJCUSPARSE` matrices.
4701: Level: intermediate
4703: Notes:
4704: Users usually access the factorization solvers via `KSP`
4706: Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
4707: such as pastix, superlu, mumps etc.
4709: PETSc must have been ./configure to use the external solver, using the option --download-package
4711: Some of the packages have options for controlling the factorization, these are in the form -prefix_mat_packagename_packageoption
4712: where prefix is normally obtained from the calling `KSP`/`PC`. If `MatGetFactor()` is called directly one can set
4713: call `MatSetOptionsPrefixFactor()` on the originating matrix or `MatSetOptionsPrefix()` on the resulting factor matrix.
4715: Developer Note:
4716: This should actually be called `MatCreateFactor()` since it creates a new factor object
4718: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `KSP`, `MatSolverType`, `MatFactorType`, `MatCopy()`, `MatDuplicate()`, `MatGetFactorAvailable()`, `MatFactorGetCanUseOrdering()`, `MatSolverTypeRegister()`,
4719: `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`
4720: @*/
4721: PetscErrorCode MatGetFactor(Mat mat, MatSolverType type, MatFactorType ftype, Mat *f)
4722: {
4723: PetscBool foundtype, foundmtype;
4724: PetscErrorCode (*conv)(Mat, MatFactorType, Mat *);
4726: PetscFunctionBegin;
4730: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4731: MatCheckPreallocated(mat, 1);
4733: PetscCall(MatSolverTypeGet(type, ((PetscObject)mat)->type_name, ftype, &foundtype, &foundmtype, &conv));
4734: if (!foundtype) {
4735: if (type) {
4736: 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],
4737: ((PetscObject)mat)->type_name, type);
4738: } else {
4739: 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);
4740: }
4741: }
4742: PetscCheck(foundmtype, PetscObjectComm((PetscObject)mat), PETSC_ERR_MISSING_FACTOR, "MatSolverType %s does not support matrix type %s", type, ((PetscObject)mat)->type_name);
4743: 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);
4745: PetscCall((*conv)(mat, ftype, f));
4746: if (mat->factorprefix) PetscCall(MatSetOptionsPrefix(*f, mat->factorprefix));
4747: PetscFunctionReturn(PETSC_SUCCESS);
4748: }
4750: /*@C
4751: MatGetFactorAvailable - Returns a a flag if matrix supports particular type and factor type
4753: Not Collective
4755: Input Parameters:
4756: + mat - the matrix
4757: . type - name of solver type, for example, superlu, petsc (to use PETSc's default)
4758: - ftype - factor type, `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`
4760: Output Parameter:
4761: . flg - PETSC_TRUE if the factorization is available
4763: Level: intermediate
4765: Notes:
4766: Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
4767: such as pastix, superlu, mumps etc.
4769: PETSc must have been ./configure to use the external solver, using the option --download-package
4771: Developer Note:
4772: This should actually be called MatCreateFactorAvailable() since MatGetFactor() creates a new factor object
4774: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatSolverType`, `MatFactorType`, `MatGetFactor()`, `MatCopy()`, `MatDuplicate()`, `MatGetFactor()`, `MatSolverTypeRegister()`,
4775: `MAT_FACTOR_LU`, `MAT_FACTOR_CHOLESKY`, `MAT_FACTOR_ICC`, `MAT_FACTOR_ILU`, `MAT_FACTOR_QR`
4776: @*/
4777: PetscErrorCode MatGetFactorAvailable(Mat mat, MatSolverType type, MatFactorType ftype, PetscBool *flg)
4778: {
4779: PetscErrorCode (*gconv)(Mat, MatFactorType, Mat *);
4781: PetscFunctionBegin;
4786: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4787: MatCheckPreallocated(mat, 1);
4789: PetscCall(MatSolverTypeGet(type, ((PetscObject)mat)->type_name, ftype, NULL, NULL, &gconv));
4790: *flg = gconv ? PETSC_TRUE : PETSC_FALSE;
4791: PetscFunctionReturn(PETSC_SUCCESS);
4792: }
4794: /*@
4795: MatDuplicate - Duplicates a matrix including the non-zero structure.
4797: Collective
4799: Input Parameters:
4800: + mat - the matrix
4801: - op - One of `MAT_DO_NOT_COPY_VALUES`, `MAT_COPY_VALUES`, or `MAT_SHARE_NONZERO_PATTERN`.
4802: See the manual page for `MatDuplicateOption()` for an explanation of these options.
4804: Output Parameter:
4805: . M - pointer to place new matrix
4807: Level: intermediate
4809: Notes:
4810: You cannot change the nonzero pattern for the parent or child matrix if you use `MAT_SHARE_NONZERO_PATTERN`.
4812: 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.
4814: When original mat is a product of matrix operation, e.g., an output of `MatMatMult()` or `MatCreateSubMatrix()`, only the simple matrix data structure of mat
4815: is duplicated and the internal data structures created for the reuse of previous matrix operations are not duplicated.
4816: User should not use `MatDuplicate()` to create new matrix M if M is intended to be reused as the product of matrix operation.
4818: .seealso: [](chapter_matrices), `Mat`, `MatCopy()`, `MatConvert()`, `MatDuplicateOption`
4819: @*/
4820: PetscErrorCode MatDuplicate(Mat mat, MatDuplicateOption op, Mat *M)
4821: {
4822: Mat B;
4823: VecType vtype;
4824: PetscInt i;
4825: PetscObject dm;
4826: void (*viewf)(void);
4828: PetscFunctionBegin;
4832: PetscCheck(op != MAT_COPY_VALUES || mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "MAT_COPY_VALUES not allowed for unassembled matrix");
4833: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4834: MatCheckPreallocated(mat, 1);
4836: *M = NULL;
4837: PetscCall(PetscLogEventBegin(MAT_Convert, mat, 0, 0, 0));
4838: PetscUseTypeMethod(mat, duplicate, op, M);
4839: PetscCall(PetscLogEventEnd(MAT_Convert, mat, 0, 0, 0));
4840: B = *M;
4842: PetscCall(MatGetOperation(mat, MATOP_VIEW, &viewf));
4843: if (viewf) PetscCall(MatSetOperation(B, MATOP_VIEW, viewf));
4844: PetscCall(MatGetVecType(mat, &vtype));
4845: PetscCall(MatSetVecType(B, vtype));
4847: B->stencil.dim = mat->stencil.dim;
4848: B->stencil.noc = mat->stencil.noc;
4849: for (i = 0; i <= mat->stencil.dim; i++) {
4850: B->stencil.dims[i] = mat->stencil.dims[i];
4851: B->stencil.starts[i] = mat->stencil.starts[i];
4852: }
4854: B->nooffproczerorows = mat->nooffproczerorows;
4855: B->nooffprocentries = mat->nooffprocentries;
4857: PetscCall(PetscObjectQuery((PetscObject)mat, "__PETSc_dm", &dm));
4858: if (dm) PetscCall(PetscObjectCompose((PetscObject)B, "__PETSc_dm", dm));
4859: PetscCall(PetscObjectStateIncrease((PetscObject)B));
4860: PetscFunctionReturn(PETSC_SUCCESS);
4861: }
4863: /*@
4864: MatGetDiagonal - Gets the diagonal of a matrix as a `Vec`
4866: Logically Collective
4868: Input Parameter:
4869: . mat - the matrix
4871: Output Parameter:
4872: . v - the diagonal of the matrix
4874: Level: intermediate
4876: Note:
4877: Currently only correct in parallel for square matrices.
4879: .seealso: [](chapter_matrices), `Mat`, `Vec`, `MatGetRow()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMaxAbs()`
4880: @*/
4881: PetscErrorCode MatGetDiagonal(Mat mat, Vec v)
4882: {
4883: PetscFunctionBegin;
4887: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4888: MatCheckPreallocated(mat, 1);
4890: PetscUseTypeMethod(mat, getdiagonal, v);
4891: PetscCall(PetscObjectStateIncrease((PetscObject)v));
4892: PetscFunctionReturn(PETSC_SUCCESS);
4893: }
4895: /*@C
4896: MatGetRowMin - Gets the minimum value (of the real part) of each
4897: row of the matrix
4899: Logically Collective
4901: Input Parameter:
4902: . mat - the matrix
4904: Output Parameters:
4905: + v - the vector for storing the maximums
4906: - idx - the indices of the column found for each row (optional)
4908: Level: intermediate
4910: Note:
4911: The result of this call are the same as if one converted the matrix to dense format
4912: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
4914: This code is only implemented for a couple of matrix formats.
4916: .seealso: [](chapter_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMaxAbs()`, `MatGetRowMinAbs()`,
4917: `MatGetRowMax()`
4918: @*/
4919: PetscErrorCode MatGetRowMin(Mat mat, Vec v, PetscInt idx[])
4920: {
4921: PetscFunctionBegin;
4925: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4927: if (!mat->cmap->N) {
4928: PetscCall(VecSet(v, PETSC_MAX_REAL));
4929: if (idx) {
4930: PetscInt i, m = mat->rmap->n;
4931: for (i = 0; i < m; i++) idx[i] = -1;
4932: }
4933: } else {
4934: MatCheckPreallocated(mat, 1);
4935: }
4936: PetscUseTypeMethod(mat, getrowmin, v, idx);
4937: PetscCall(PetscObjectStateIncrease((PetscObject)v));
4938: PetscFunctionReturn(PETSC_SUCCESS);
4939: }
4941: /*@C
4942: MatGetRowMinAbs - Gets the minimum value (in absolute value) of each
4943: row of the matrix
4945: Logically Collective
4947: Input Parameter:
4948: . mat - the matrix
4950: Output Parameters:
4951: + v - the vector for storing the minimums
4952: - idx - the indices of the column found for each row (or `NULL` if not needed)
4954: Level: intermediate
4956: Notes:
4957: if a row is completely empty or has only 0.0 values then the idx[] value for that
4958: row is 0 (the first column).
4960: This code is only implemented for a couple of matrix formats.
4962: .seealso: [](chapter_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMax()`, `MatGetRowMaxAbs()`, `MatGetRowMin()`
4963: @*/
4964: PetscErrorCode MatGetRowMinAbs(Mat mat, Vec v, PetscInt idx[])
4965: {
4966: PetscFunctionBegin;
4970: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4971: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4973: if (!mat->cmap->N) {
4974: PetscCall(VecSet(v, 0.0));
4975: if (idx) {
4976: PetscInt i, m = mat->rmap->n;
4977: for (i = 0; i < m; i++) idx[i] = -1;
4978: }
4979: } else {
4980: MatCheckPreallocated(mat, 1);
4981: if (idx) PetscCall(PetscArrayzero(idx, mat->rmap->n));
4982: PetscUseTypeMethod(mat, getrowminabs, v, idx);
4983: }
4984: PetscCall(PetscObjectStateIncrease((PetscObject)v));
4985: PetscFunctionReturn(PETSC_SUCCESS);
4986: }
4988: /*@C
4989: MatGetRowMax - Gets the maximum value (of the real part) of each
4990: row of the matrix
4992: Logically Collective
4994: Input Parameter:
4995: . mat - the matrix
4997: Output Parameters:
4998: + v - the vector for storing the maximums
4999: - idx - the indices of the column found for each row (optional)
5001: Level: intermediate
5003: Notes:
5004: The result of this call are the same as if one converted the matrix to dense format
5005: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
5007: This code is only implemented for a couple of matrix formats.
5009: .seealso: [](chapter_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMaxAbs()`, `MatGetRowMin()`, `MatGetRowMinAbs()`
5010: @*/
5011: PetscErrorCode MatGetRowMax(Mat mat, Vec v, PetscInt idx[])
5012: {
5013: PetscFunctionBegin;
5017: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5019: if (!mat->cmap->N) {
5020: PetscCall(VecSet(v, PETSC_MIN_REAL));
5021: if (idx) {
5022: PetscInt i, m = mat->rmap->n;
5023: for (i = 0; i < m; i++) idx[i] = -1;
5024: }
5025: } else {
5026: MatCheckPreallocated(mat, 1);
5027: PetscUseTypeMethod(mat, getrowmax, v, idx);
5028: }
5029: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5030: PetscFunctionReturn(PETSC_SUCCESS);
5031: }
5033: /*@C
5034: MatGetRowMaxAbs - Gets the maximum value (in absolute value) of each
5035: row of the matrix
5037: Logically Collective
5039: Input Parameter:
5040: . mat - the matrix
5042: Output Parameters:
5043: + v - the vector for storing the maximums
5044: - idx - the indices of the column found for each row (or `NULL` if not needed)
5046: Level: intermediate
5048: Notes:
5049: if a row is completely empty or has only 0.0 values then the idx[] value for that
5050: row is 0 (the first column).
5052: This code is only implemented for a couple of matrix formats.
5054: .seealso: [](chapter_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMax()`, `MatGetRowMin()`, `MatGetRowMinAbs()`
5055: @*/
5056: PetscErrorCode MatGetRowMaxAbs(Mat mat, Vec v, PetscInt idx[])
5057: {
5058: PetscFunctionBegin;
5062: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5064: if (!mat->cmap->N) {
5065: PetscCall(VecSet(v, 0.0));
5066: if (idx) {
5067: PetscInt i, m = mat->rmap->n;
5068: for (i = 0; i < m; i++) idx[i] = -1;
5069: }
5070: } else {
5071: MatCheckPreallocated(mat, 1);
5072: if (idx) PetscCall(PetscArrayzero(idx, mat->rmap->n));
5073: PetscUseTypeMethod(mat, getrowmaxabs, v, idx);
5074: }
5075: PetscCall(PetscObjectStateIncrease((PetscObject)v));
5076: PetscFunctionReturn(PETSC_SUCCESS);
5077: }
5079: /*@
5080: MatGetRowSum - Gets the sum of each row of the matrix
5082: Logically or Neighborhood Collective
5084: Input Parameter:
5085: . mat - the matrix
5087: Output Parameter:
5088: . v - the vector for storing the sum of rows
5090: Level: intermediate
5092: Notes:
5093: This code is slow since it is not currently specialized for different formats
5095: .seealso: [](chapter_matrices), `Mat`, `MatGetDiagonal()`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRowMax()`, `MatGetRowMin()`, `MatGetRowMaxAbs()`, `MatGetRowMinAbs()`
5096: @*/
5097: PetscErrorCode MatGetRowSum(Mat mat, Vec v)
5098: {
5099: Vec ones;
5101: PetscFunctionBegin;
5105: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5106: MatCheckPreallocated(mat, 1);
5107: PetscCall(MatCreateVecs(mat, &ones, NULL));
5108: PetscCall(VecSet(ones, 1.));
5109: PetscCall(MatMult(mat, ones, v));
5110: PetscCall(VecDestroy(&ones));
5111: PetscFunctionReturn(PETSC_SUCCESS);
5112: }
5114: /*@
5115: MatTransposeSetPrecursor - Set the matrix from which the second matrix will receive numerical transpose data with a call to `MatTranspose`(A,`MAT_REUSE_MATRIX`,&B)
5116: when B was not obtained with `MatTranspose`(A,`MAT_INITIAL_MATRIX`,&B)
5118: Collective
5120: Input Parameter:
5121: . mat - the matrix to provide the transpose
5123: Output Parameter:
5124: . mat - 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
5126: Level: advanced
5128: Note:
5129: 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
5130: routine allows bypassing that call.
5132: .seealso: [](chapter_matrices), `Mat`, `MatTransposeSymbolic()`, `MatTranspose()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`
5133: @*/
5134: PetscErrorCode MatTransposeSetPrecursor(Mat mat, Mat B)
5135: {
5136: PetscContainer rB = NULL;
5137: MatParentState *rb = NULL;
5139: PetscFunctionBegin;
5140: PetscCall(PetscNew(&rb));
5141: rb->id = ((PetscObject)mat)->id;
5142: rb->state = 0;
5143: PetscCall(MatGetNonzeroState(mat, &rb->nonzerostate));
5144: PetscCall(PetscContainerCreate(PetscObjectComm((PetscObject)B), &rB));
5145: PetscCall(PetscContainerSetPointer(rB, rb));
5146: PetscCall(PetscContainerSetUserDestroy(rB, PetscContainerUserDestroyDefault));
5147: PetscCall(PetscObjectCompose((PetscObject)B, "MatTransposeParent", (PetscObject)rB));
5148: PetscCall(PetscObjectDereference((PetscObject)rB));
5149: PetscFunctionReturn(PETSC_SUCCESS);
5150: }
5152: /*@
5153: MatTranspose - Computes an in-place or out-of-place transpose of a matrix.
5155: Collective
5157: Input Parameters:
5158: + mat - the matrix to transpose
5159: - reuse - either `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, or `MAT_INPLACE_MATRIX`
5161: Output Parameter:
5162: . B - the transpose
5164: Level: intermediate
5166: Notes:
5167: If you use `MAT_INPLACE_MATRIX` then you must pass in &mat for B
5169: `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
5170: transpose, call `MatTransposeSetPrecursor`(mat,B) before calling this routine.
5172: 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.
5174: Consider using `MatCreateTranspose()` instead if you only need a matrix that behaves like the transpose, but don't need the storage to be changed.
5176: If mat is unchanged from the last call this function returns immediately without recomputing the result
5178: If you only need the symbolic transpose, and not the numerical values, use `MatTransposeSymbolic()`
5180: .seealso: [](chapter_matrices), `Mat`, `MatTransposeSetPrecursor()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`,
5181: `MatTransposeSymbolic()`, `MatCreateTranspose()`
5182: @*/
5183: PetscErrorCode MatTranspose(Mat mat, MatReuse reuse, Mat *B)
5184: {
5185: PetscContainer rB = NULL;
5186: MatParentState *rb = NULL;
5188: PetscFunctionBegin;
5191: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5192: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5193: PetscCheck(reuse != MAT_INPLACE_MATRIX || mat == *B, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "MAT_INPLACE_MATRIX requires last matrix to match first");
5194: PetscCheck(reuse != MAT_REUSE_MATRIX || mat != *B, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Perhaps you mean MAT_INPLACE_MATRIX");
5195: MatCheckPreallocated(mat, 1);
5196: if (reuse == MAT_REUSE_MATRIX) {
5197: PetscCall(PetscObjectQuery((PetscObject)*B, "MatTransposeParent", (PetscObject *)&rB));
5198: PetscCheck(rB, PetscObjectComm((PetscObject)*B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from call to MatTranspose(). Suggest MatTransposeSetPrecursor().");
5199: PetscCall(PetscContainerGetPointer(rB, (void **)&rb));
5200: PetscCheck(rb->id == ((PetscObject)mat)->id, PetscObjectComm((PetscObject)*B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from input matrix");
5201: if (rb->state == ((PetscObject)mat)->state) PetscFunctionReturn(PETSC_SUCCESS);
5202: }
5204: PetscCall(PetscLogEventBegin(MAT_Transpose, mat, 0, 0, 0));
5205: if (reuse != MAT_INPLACE_MATRIX || mat->symmetric != PETSC_BOOL3_TRUE) {
5206: PetscUseTypeMethod(mat, transpose, reuse, B);
5207: PetscCall(PetscObjectStateIncrease((PetscObject)*B));
5208: }
5209: PetscCall(PetscLogEventEnd(MAT_Transpose, mat, 0, 0, 0));
5211: if (reuse == MAT_INITIAL_MATRIX) PetscCall(MatTransposeSetPrecursor(mat, *B));
5212: if (reuse != MAT_INPLACE_MATRIX) {
5213: PetscCall(PetscObjectQuery((PetscObject)*B, "MatTransposeParent", (PetscObject *)&rB));
5214: PetscCall(PetscContainerGetPointer(rB, (void **)&rb));
5215: rb->state = ((PetscObject)mat)->state;
5216: rb->nonzerostate = mat->nonzerostate;
5217: }
5218: PetscFunctionReturn(PETSC_SUCCESS);
5219: }
5221: /*@
5222: MatTransposeSymbolic - Computes the symbolic part of the transpose of a matrix.
5224: Collective
5226: Input Parameter:
5227: . A - the matrix to transpose
5229: Output Parameter:
5230: . 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
5231: numerical portion.
5233: Level: intermediate
5235: Note:
5236: This is not supported for many matrix types, use `MatTranspose()` in those cases
5238: .seealso: [](chapter_matrices), `Mat`, `MatTransposeSetPrecursor()`, `MatTranspose()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`, `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, `MAT_INPLACE_MATRIX`
5239: @*/
5240: PetscErrorCode MatTransposeSymbolic(Mat A, Mat *B)
5241: {
5242: PetscFunctionBegin;
5245: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5246: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5247: PetscCall(PetscLogEventBegin(MAT_Transpose, A, 0, 0, 0));
5248: PetscUseTypeMethod(A, transposesymbolic, B);
5249: PetscCall(PetscLogEventEnd(MAT_Transpose, A, 0, 0, 0));
5251: PetscCall(MatTransposeSetPrecursor(A, *B));
5252: PetscFunctionReturn(PETSC_SUCCESS);
5253: }
5255: PetscErrorCode MatTransposeCheckNonzeroState_Private(Mat A, Mat B)
5256: {
5257: PetscContainer rB;
5258: MatParentState *rb;
5260: PetscFunctionBegin;
5263: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5264: PetscCheck(!A->factortype, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5265: PetscCall(PetscObjectQuery((PetscObject)B, "MatTransposeParent", (PetscObject *)&rB));
5266: PetscCheck(rB, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from call to MatTranspose()");
5267: PetscCall(PetscContainerGetPointer(rB, (void **)&rb));
5268: PetscCheck(rb->id == ((PetscObject)A)->id, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONG, "Reuse matrix used was not generated from input matrix");
5269: PetscCheck(rb->nonzerostate == A->nonzerostate, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONGSTATE, "Reuse matrix has changed nonzero structure");
5270: PetscFunctionReturn(PETSC_SUCCESS);
5271: }
5273: /*@
5274: MatIsTranspose - Test whether a matrix is another one's transpose,
5275: or its own, in which case it tests symmetry.
5277: Collective
5279: Input Parameters:
5280: + A - the matrix to test
5281: . B - the matrix to test against, this can equal the first parameter
5282: - tol - tolerance, differences between entries smaller than this are counted as zero
5284: Output Parameter:
5285: . flg - the result
5287: Level: intermediate
5289: Notes:
5290: Only available for `MATAIJ` matrices.
5292: The sequential algorithm has a running time of the order of the number of nonzeros; the parallel
5293: test involves parallel copies of the block-offdiagonal parts of the matrix.
5295: .seealso: [](chapter_matrices), `Mat`, `MatTranspose()`, `MatIsSymmetric()`, `MatIsHermitian()`
5296: @*/
5297: PetscErrorCode MatIsTranspose(Mat A, Mat B, PetscReal tol, PetscBool *flg)
5298: {
5299: PetscErrorCode (*f)(Mat, Mat, PetscReal, PetscBool *), (*g)(Mat, Mat, PetscReal, PetscBool *);
5301: PetscFunctionBegin;
5305: PetscCall(PetscObjectQueryFunction((PetscObject)A, "MatIsTranspose_C", &f));
5306: PetscCall(PetscObjectQueryFunction((PetscObject)B, "MatIsTranspose_C", &g));
5307: *flg = PETSC_FALSE;
5308: if (f && g) {
5309: PetscCheck(f == g, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_NOTSAMETYPE, "Matrices do not have the same comparator for symmetry test");
5310: PetscCall((*f)(A, B, tol, flg));
5311: } else {
5312: MatType mattype;
5314: PetscCall(MatGetType(f ? B : A, &mattype));
5315: SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "Matrix of type %s does not support checking for transpose", mattype);
5316: }
5317: PetscFunctionReturn(PETSC_SUCCESS);
5318: }
5320: /*@
5321: MatHermitianTranspose - Computes an in-place or out-of-place Hermitian transpose of a matrix in complex conjugate.
5323: Collective
5325: Input Parameters:
5326: + mat - the matrix to transpose and complex conjugate
5327: - reuse - either `MAT_INITIAL_MATRIX`, `MAT_REUSE_MATRIX`, or `MAT_INPLACE_MATRIX`
5329: Output Parameter:
5330: . B - the Hermitian transpose
5332: Level: intermediate
5334: .seealso: [](chapter_matrices), `Mat`, `MatTranspose()`, `MatMultTranspose()`, `MatMultTransposeAdd()`, `MatIsTranspose()`, `MatReuse`
5335: @*/
5336: PetscErrorCode MatHermitianTranspose(Mat mat, MatReuse reuse, Mat *B)
5337: {
5338: PetscFunctionBegin;
5339: PetscCall(MatTranspose(mat, reuse, B));
5340: #if defined(PETSC_USE_COMPLEX)
5341: PetscCall(MatConjugate(*B));
5342: #endif
5343: PetscFunctionReturn(PETSC_SUCCESS);
5344: }
5346: /*@
5347: MatIsHermitianTranspose - Test whether a matrix is another one's Hermitian transpose,
5349: Collective
5351: Input Parameters:
5352: + A - the matrix to test
5353: . B - the matrix to test against, this can equal the first parameter
5354: - tol - tolerance, differences between entries smaller than this are counted as zero
5356: Output Parameter:
5357: . flg - the result
5359: Level: intermediate
5361: Notes:
5362: Only available for `MATAIJ` matrices.
5364: The sequential algorithm
5365: has a running time of the order of the number of nonzeros; the parallel
5366: test involves parallel copies of the block-offdiagonal parts of the matrix.
5368: .seealso: [](chapter_matrices), `Mat`, `MatTranspose()`, `MatIsSymmetric()`, `MatIsHermitian()`, `MatIsTranspose()`
5369: @*/
5370: PetscErrorCode MatIsHermitianTranspose(Mat A, Mat B, PetscReal tol, PetscBool *flg)
5371: {
5372: PetscErrorCode (*f)(Mat, Mat, PetscReal, PetscBool *), (*g)(Mat, Mat, PetscReal, PetscBool *);
5374: PetscFunctionBegin;
5378: PetscCall(PetscObjectQueryFunction((PetscObject)A, "MatIsHermitianTranspose_C", &f));
5379: PetscCall(PetscObjectQueryFunction((PetscObject)B, "MatIsHermitianTranspose_C", &g));
5380: if (f && g) {
5381: PetscCheck(f != g, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_NOTSAMETYPE, "Matrices do not have the same comparator for Hermitian test");
5382: PetscCall((*f)(A, B, tol, flg));
5383: }
5384: PetscFunctionReturn(PETSC_SUCCESS);
5385: }
5387: /*@
5388: MatPermute - Creates a new matrix with rows and columns permuted from the
5389: original.
5391: Collective
5393: Input Parameters:
5394: + mat - the matrix to permute
5395: . row - row permutation, each processor supplies only the permutation for its rows
5396: - col - column permutation, each processor supplies only the permutation for its columns
5398: Output Parameter:
5399: . B - the permuted matrix
5401: Level: advanced
5403: Note:
5404: The index sets map from row/col of permuted matrix to row/col of original matrix.
5405: The index sets should be on the same communicator as mat and have the same local sizes.
5407: Developer Note:
5408: If you want to implement `MatPermute()` for a matrix type, and your approach doesn't
5409: exploit the fact that row and col are permutations, consider implementing the
5410: more general `MatCreateSubMatrix()` instead.
5412: .seealso: [](chapter_matrices), `Mat`, `MatGetOrdering()`, `ISAllGather()`, `MatCreateSubMatrix()`
5413: @*/
5414: PetscErrorCode MatPermute(Mat mat, IS row, IS col, Mat *B)
5415: {
5416: PetscFunctionBegin;
5422: PetscCheckSameComm(mat, 1, row, 2);
5423: if (row != col) PetscCheckSameComm(row, 2, col, 3);
5424: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5425: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5426: PetscCheck(mat->ops->permute || mat->ops->createsubmatrix, PETSC_COMM_SELF, PETSC_ERR_SUP, "MatPermute not available for Mat type %s", ((PetscObject)mat)->type_name);
5427: MatCheckPreallocated(mat, 1);
5429: if (mat->ops->permute) {
5430: PetscUseTypeMethod(mat, permute, row, col, B);
5431: PetscCall(PetscObjectStateIncrease((PetscObject)*B));
5432: } else {
5433: PetscCall(MatCreateSubMatrix(mat, row, col, MAT_INITIAL_MATRIX, B));
5434: }
5435: PetscFunctionReturn(PETSC_SUCCESS);
5436: }
5438: /*@
5439: MatEqual - Compares two matrices.
5441: Collective
5443: Input Parameters:
5444: + A - the first matrix
5445: - B - the second matrix
5447: Output Parameter:
5448: . flg - `PETSC_TRUE` if the matrices are equal; `PETSC_FALSE` otherwise.
5450: Level: intermediate
5452: .seealso: [](chapter_matrices), `Mat`
5453: @*/
5454: PetscErrorCode MatEqual(Mat A, Mat B, PetscBool *flg)
5455: {
5456: PetscFunctionBegin;
5462: PetscCheckSameComm(A, 1, B, 2);
5463: MatCheckPreallocated(A, 1);
5464: MatCheckPreallocated(B, 2);
5465: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5466: PetscCheck(B->assembled, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5467: 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,
5468: B->cmap->N);
5469: if (A->ops->equal && A->ops->equal == B->ops->equal) {
5470: PetscUseTypeMethod(A, equal, B, flg);
5471: } else {
5472: PetscCall(MatMultEqual(A, B, 10, flg));
5473: }
5474: PetscFunctionReturn(PETSC_SUCCESS);
5475: }
5477: /*@
5478: MatDiagonalScale - Scales a matrix on the left and right by diagonal
5479: matrices that are stored as vectors. Either of the two scaling
5480: matrices can be `NULL`.
5482: Collective
5484: Input Parameters:
5485: + mat - the matrix to be scaled
5486: . l - the left scaling vector (or `NULL`)
5487: - r - the right scaling vector (or `NULL`)
5489: Level: intermediate
5491: Note:
5492: `MatDiagonalScale()` computes A = LAR, where
5493: L = a diagonal matrix (stored as a vector), R = a diagonal matrix (stored as a vector)
5494: The L scales the rows of the matrix, the R scales the columns of the matrix.
5496: .seealso: [](chapter_matrices), `Mat`, `MatScale()`, `MatShift()`, `MatDiagonalSet()`
5497: @*/
5498: PetscErrorCode MatDiagonalScale(Mat mat, Vec l, Vec r)
5499: {
5500: PetscFunctionBegin;
5503: if (l) {
5505: PetscCheckSameComm(mat, 1, l, 2);
5506: }
5507: if (r) {
5509: PetscCheckSameComm(mat, 1, r, 3);
5510: }
5511: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5512: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5513: MatCheckPreallocated(mat, 1);
5514: if (!l && !r) PetscFunctionReturn(PETSC_SUCCESS);
5516: PetscCall(PetscLogEventBegin(MAT_Scale, mat, 0, 0, 0));
5517: PetscUseTypeMethod(mat, diagonalscale, l, r);
5518: PetscCall(PetscLogEventEnd(MAT_Scale, mat, 0, 0, 0));
5519: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
5520: if (l != r) mat->symmetric = PETSC_BOOL3_FALSE;
5521: PetscFunctionReturn(PETSC_SUCCESS);
5522: }
5524: /*@
5525: MatScale - Scales all elements of a matrix by a given number.
5527: Logically Collective
5529: Input Parameters:
5530: + mat - the matrix to be scaled
5531: - a - the scaling value
5533: Level: intermediate
5535: .seealso: [](chapter_matrices), `Mat`, `MatDiagonalScale()`
5536: @*/
5537: PetscErrorCode MatScale(Mat mat, PetscScalar a)
5538: {
5539: PetscFunctionBegin;
5542: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5543: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5545: MatCheckPreallocated(mat, 1);
5547: PetscCall(PetscLogEventBegin(MAT_Scale, mat, 0, 0, 0));
5548: if (a != (PetscScalar)1.0) {
5549: PetscUseTypeMethod(mat, scale, a);
5550: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
5551: }
5552: PetscCall(PetscLogEventEnd(MAT_Scale, mat, 0, 0, 0));
5553: PetscFunctionReturn(PETSC_SUCCESS);
5554: }
5556: /*@
5557: MatNorm - Calculates various norms of a matrix.
5559: Collective
5561: Input Parameters:
5562: + mat - the matrix
5563: - type - the type of norm, `NORM_1`, `NORM_FROBENIUS`, `NORM_INFINITY`
5565: Output Parameter:
5566: . nrm - the resulting norm
5568: Level: intermediate
5570: .seealso: [](chapter_matrices), `Mat`
5571: @*/
5572: PetscErrorCode MatNorm(Mat mat, NormType type, PetscReal *nrm)
5573: {
5574: PetscFunctionBegin;
5579: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
5580: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
5581: MatCheckPreallocated(mat, 1);
5583: PetscUseTypeMethod(mat, norm, type, nrm);
5584: PetscFunctionReturn(PETSC_SUCCESS);
5585: }
5587: /*
5588: This variable is used to prevent counting of MatAssemblyBegin() that
5589: are called from within a MatAssemblyEnd().
5590: */
5591: static PetscInt MatAssemblyEnd_InUse = 0;
5592: /*@
5593: MatAssemblyBegin - Begins assembling the matrix. This routine should
5594: be called after completing all calls to `MatSetValues()`.
5596: Collective
5598: Input Parameters:
5599: + mat - the matrix
5600: - type - type of assembly, either `MAT_FLUSH_ASSEMBLY` or `MAT_FINAL_ASSEMBLY`
5602: Level: beginner
5604: Notes:
5605: `MatSetValues()` generally caches the values that belong to other MPI ranks. The matrix is ready to
5606: use only after `MatAssemblyBegin()` and `MatAssemblyEnd()` have been called.
5608: Use `MAT_FLUSH_ASSEMBLY` when switching between `ADD_VALUES` and `INSERT_VALUES`
5609: in `MatSetValues()`; use `MAT_FINAL_ASSEMBLY` for the final assembly before
5610: using the matrix.
5612: ALL processes that share a matrix MUST call `MatAssemblyBegin()` and `MatAssemblyEnd()` the SAME NUMBER of times, and each time with the
5613: 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
5614: a global collective operation requiring all processes that share the matrix.
5616: Space for preallocated nonzeros that is not filled by a call to `MatSetValues()` or a related routine are compressed
5617: out by assembly. If you intend to use that extra space on a subsequent assembly, be sure to insert explicit zeros
5618: before `MAT_FINAL_ASSEMBLY` so the space is not compressed out.
5620: .seealso: [](chapter_matrices), `Mat`, `MatAssemblyEnd()`, `MatSetValues()`, `MatAssembled()`
5621: @*/
5622: PetscErrorCode MatAssemblyBegin(Mat mat, MatAssemblyType type)
5623: {
5624: PetscFunctionBegin;
5627: MatCheckPreallocated(mat, 1);
5628: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix.\nDid you forget to call MatSetUnfactored()?");
5629: if (mat->assembled) {
5630: mat->was_assembled = PETSC_TRUE;
5631: mat->assembled = PETSC_FALSE;
5632: }
5634: if (!MatAssemblyEnd_InUse) {
5635: PetscCall(PetscLogEventBegin(MAT_AssemblyBegin, mat, 0, 0, 0));
5636: PetscTryTypeMethod(mat, assemblybegin, type);
5637: PetscCall(PetscLogEventEnd(MAT_AssemblyBegin, mat, 0, 0, 0));
5638: } else PetscTryTypeMethod(mat, assemblybegin, type);
5639: PetscFunctionReturn(PETSC_SUCCESS);
5640: }
5642: /*@
5643: MatAssembled - Indicates if a matrix has been assembled and is ready for
5644: use; for example, in matrix-vector product.
5646: Not Collective
5648: Input Parameter:
5649: . mat - the matrix
5651: Output Parameter:
5652: . assembled - `PETSC_TRUE` or `PETSC_FALSE`
5654: Level: advanced
5656: .seealso: [](chapter_matrices), `Mat`, `MatAssemblyEnd()`, `MatSetValues()`, `MatAssemblyBegin()`
5657: @*/
5658: PetscErrorCode MatAssembled(Mat mat, PetscBool *assembled)
5659: {
5660: PetscFunctionBegin;
5663: *assembled = mat->assembled;
5664: PetscFunctionReturn(PETSC_SUCCESS);
5665: }
5667: /*@
5668: MatAssemblyEnd - Completes assembling the matrix. This routine should
5669: be called after `MatAssemblyBegin()`.
5671: Collective
5673: Input Parameters:
5674: + mat - the matrix
5675: - type - type of assembly, either `MAT_FLUSH_ASSEMBLY` or `MAT_FINAL_ASSEMBLY`
5677: Options Database Keys:
5678: + -mat_view ::ascii_info - Prints info on matrix at conclusion of `MatEndAssembly()`
5679: . -mat_view ::ascii_info_detail - Prints more detailed info
5680: . -mat_view - Prints matrix in ASCII format
5681: . -mat_view ::ascii_matlab - Prints matrix in Matlab format
5682: . -mat_view draw - draws nonzero structure of matrix, using `MatView()` and `PetscDrawOpenX()`.
5683: . -display <name> - Sets display name (default is host)
5684: . -draw_pause <sec> - Sets number of seconds to pause after display
5685: . -mat_view socket - Sends matrix to socket, can be accessed from Matlab (See [Using MATLAB with PETSc](ch_matlab))
5686: . -viewer_socket_machine <machine> - Machine to use for socket
5687: . -viewer_socket_port <port> - Port number to use for socket
5688: - -mat_view binary:filename[:append] - Save matrix to file in binary format
5690: Level: beginner
5692: .seealso: [](chapter_matrices), `Mat`, `MatAssemblyBegin()`, `MatSetValues()`, `PetscDrawOpenX()`, `PetscDrawCreate()`, `MatView()`, `MatAssembled()`, `PetscViewerSocketOpen()`
5693: @*/
5694: PetscErrorCode MatAssemblyEnd(Mat mat, MatAssemblyType type)
5695: {
5696: static PetscInt inassm = 0;
5697: PetscBool flg = PETSC_FALSE;
5699: PetscFunctionBegin;
5703: inassm++;
5704: MatAssemblyEnd_InUse++;
5705: if (MatAssemblyEnd_InUse == 1) { /* Do the logging only the first time through */
5706: PetscCall(PetscLogEventBegin(MAT_AssemblyEnd, mat, 0, 0, 0));
5707: PetscTryTypeMethod(mat, assemblyend, type);
5708: PetscCall(PetscLogEventEnd(MAT_AssemblyEnd, mat, 0, 0, 0));
5709: } else PetscTryTypeMethod(mat, assemblyend, type);
5711: /* Flush assembly is not a true assembly */
5712: if (type != MAT_FLUSH_ASSEMBLY) {
5713: if (mat->num_ass) {
5714: if (!mat->symmetry_eternal) {
5715: mat->symmetric = PETSC_BOOL3_UNKNOWN;
5716: mat->hermitian = PETSC_BOOL3_UNKNOWN;
5717: }
5718: if (!mat->structural_symmetry_eternal && mat->ass_nonzerostate != mat->nonzerostate) mat->structurally_symmetric = PETSC_BOOL3_UNKNOWN;
5719: if (!mat->spd_eternal) mat->spd = PETSC_BOOL3_UNKNOWN;
5720: }
5721: mat->num_ass++;
5722: mat->assembled = PETSC_TRUE;
5723: mat->ass_nonzerostate = mat->nonzerostate;
5724: }
5726: mat->insertmode = NOT_SET_VALUES;
5727: MatAssemblyEnd_InUse--;
5728: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
5729: if (inassm == 1 && type != MAT_FLUSH_ASSEMBLY) {
5730: PetscCall(MatViewFromOptions(mat, NULL, "-mat_view"));
5732: if (mat->checksymmetryonassembly) {
5733: PetscCall(MatIsSymmetric(mat, mat->checksymmetrytol, &flg));
5734: if (flg) {
5735: PetscCall(PetscPrintf(PetscObjectComm((PetscObject)mat), "Matrix is symmetric (tolerance %g)\n", (double)mat->checksymmetrytol));
5736: } else {
5737: PetscCall(PetscPrintf(PetscObjectComm((PetscObject)mat), "Matrix is not symmetric (tolerance %g)\n", (double)mat->checksymmetrytol));
5738: }
5739: }
5740: if (mat->nullsp && mat->checknullspaceonassembly) PetscCall(MatNullSpaceTest(mat->nullsp, mat, NULL));
5741: }
5742: inassm--;
5743: PetscFunctionReturn(PETSC_SUCCESS);
5744: }
5746: /*@
5747: MatSetOption - Sets a parameter option for a matrix. Some options
5748: may be specific to certain storage formats. Some options
5749: determine how values will be inserted (or added). Sorted,
5750: row-oriented input will generally assemble the fastest. The default
5751: is row-oriented.
5753: Logically Collective for certain operations, such as `MAT_SPD`, not collective for `MAT_ROW_ORIENTED`, see `MatOption`
5755: Input Parameters:
5756: + mat - the matrix
5757: . option - the option, one of those listed below (and possibly others),
5758: - flg - turn the option on (`PETSC_TRUE`) or off (`PETSC_FALSE`)
5760: Options Describing Matrix Structure:
5761: + `MAT_SPD` - symmetric positive definite
5762: . `MAT_SYMMETRIC` - symmetric in terms of both structure and value
5763: . `MAT_HERMITIAN` - transpose is the complex conjugation
5764: . `MAT_STRUCTURALLY_SYMMETRIC` - symmetric nonzero structure
5765: . `MAT_SYMMETRY_ETERNAL` - indicates the symmetry (or Hermitian structure) or its absence will persist through any changes to the matrix
5766: . `MAT_STRUCTURAL_SYMMETRY_ETERNAL` - indicates the structural symmetry or its absence will persist through any changes to the matrix
5767: - `MAT_SPD_ETERNAL` - indicates the value of `MAT_SPD` (true or false) will persist through any changes to the matrix
5769: These are not really options of the matrix, they are knowledge about the structure of the matrix that users may provide so that they
5770: do not need to be computed (usually at a high cost)
5772: Options For Use with `MatSetValues()`:
5773: Insert a logically dense subblock, which can be
5774: . `MAT_ROW_ORIENTED` - row-oriented (default)
5776: These options reflect the data you pass in with `MatSetValues()`; it has
5777: nothing to do with how the data is stored internally in the matrix
5778: data structure.
5780: When (re)assembling a matrix, we can restrict the input for
5781: efficiency/debugging purposes. These options include
5782: + `MAT_NEW_NONZERO_LOCATIONS` - additional insertions will be allowed if they generate a new nonzero (slow)
5783: . `MAT_FORCE_DIAGONAL_ENTRIES` - forces diagonal entries to be allocated
5784: . `MAT_IGNORE_OFF_PROC_ENTRIES` - drops off-processor entries
5785: . `MAT_NEW_NONZERO_LOCATION_ERR` - generates an error for new matrix entry
5786: . `MAT_USE_HASH_TABLE` - uses a hash table to speed up matrix assembly
5787: . `MAT_NO_OFF_PROC_ENTRIES` - you know each process will only set values for its own rows, will generate an error if
5788: any process sets values for another process. This avoids all reductions in the MatAssembly routines and thus improves
5789: performance for very large process counts.
5790: - `MAT_SUBSET_OFF_PROC_ENTRIES` - you know that the first assembly after setting this flag will set a superset
5791: of the off-process entries required for all subsequent assemblies. This avoids a rendezvous step in the MatAssembly
5792: functions, instead sending only neighbor messages.
5794: Level: intermediate
5796: Notes:
5797: Except for `MAT_UNUSED_NONZERO_LOCATION_ERR` and `MAT_ROW_ORIENTED` all processes that share the matrix must pass the same value in flg!
5799: Some options are relevant only for particular matrix types and
5800: are thus ignored by others. Other options are not supported by
5801: certain matrix types and will generate an error message if set.
5803: If using Fortran to compute a matrix, one may need to
5804: use the column-oriented option (or convert to the row-oriented
5805: format).
5807: `MAT_NEW_NONZERO_LOCATIONS` set to `PETSC_FALSE` indicates that any add or insertion
5808: that would generate a new entry in the nonzero structure is instead
5809: ignored. Thus, if memory has not already been allocated for this particular
5810: data, then the insertion is ignored. For dense matrices, in which
5811: the entire array is allocated, no entries are ever ignored.
5812: Set after the first `MatAssemblyEnd()`. If this option is set then the MatAssemblyBegin/End() processes has one less global reduction
5814: `MAT_NEW_NONZERO_LOCATION_ERR` set to PETSC_TRUE indicates that any add or insertion
5815: that would generate a new entry in the nonzero structure instead produces
5816: 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
5818: `MAT_NEW_NONZERO_ALLOCATION_ERR` set to `PETSC_TRUE` indicates that any add or insertion
5819: that would generate a new entry that has not been preallocated will
5820: instead produce an error. (Currently supported for `MATAIJ` and `MATBAIJ` formats
5821: only.) This is a useful flag when debugging matrix memory preallocation.
5822: If this option is set then the `MatAssemblyBegin()`/`MatAssemblyEnd()` processes has one less global reduction
5824: `MAT_IGNORE_OFF_PROC_ENTRIES` set to `PETSC_TRUE` indicates entries destined for
5825: other processors should be dropped, rather than stashed.
5826: This is useful if you know that the "owning" processor is also
5827: always generating the correct matrix entries, so that PETSc need
5828: not transfer duplicate entries generated on another processor.
5830: `MAT_USE_HASH_TABLE` indicates that a hash table be used to improve the
5831: searches during matrix assembly. When this flag is set, the hash table
5832: is created during the first matrix assembly. This hash table is
5833: used the next time through, during `MatSetValues()`/`MatSetValuesBlocked()`
5834: to improve the searching of indices. `MAT_NEW_NONZERO_LOCATIONS` flag
5835: should be used with `MAT_USE_HASH_TABLE` flag. This option is currently
5836: supported by `MATMPIBAIJ` format only.
5838: `MAT_KEEP_NONZERO_PATTERN` indicates when `MatZeroRows()` is called the zeroed entries
5839: are kept in the nonzero structure
5841: `MAT_IGNORE_ZERO_ENTRIES` - for `MATAIJ` and `MATIS` matrices this will stop zero values from creating
5842: a zero location in the matrix
5844: `MAT_USE_INODES` - indicates using inode version of the code - works with `MATAIJ` matrix types
5846: `MAT_NO_OFF_PROC_ZERO_ROWS` - you know each process will only zero its own rows. This avoids all reductions in the
5847: zero row routines and thus improves performance for very large process counts.
5849: `MAT_IGNORE_LOWER_TRIANGULAR` - For `MATSBAIJ` matrices will ignore any insertions you make in the lower triangular
5850: part of the matrix (since they should match the upper triangular part).
5852: `MAT_SORTED_FULL` - each process provides exactly its local rows; all column indices for a given row are passed in a
5853: single call to `MatSetValues()`, preallocation is perfect, row oriented, `INSERT_VALUES` is used. Common
5854: with finite difference schemes with non-periodic boundary conditions.
5856: Developer Note:
5857: `MAT_SYMMETRY_ETERNAL`, `MAT_STRUCTURAL_SYMMETRY_ETERNAL`, and `MAT_SPD_ETERNAL` are used by `MatAssemblyEnd()` and in other
5858: places where otherwise the value of `MAT_SYMMETRIC`, `MAT_STRUCTURAL_SYMMETRIC` or `MAT_SPD` would need to be changed back
5859: to `PETSC_BOOL3_UNKNOWN` because the matrix values had changed so the code cannot be certain that the related property had
5860: not changed.
5862: .seealso: [](chapter_matrices), `MatOption`, `Mat`, `MatGetOption()`
5863: @*/
5864: PetscErrorCode MatSetOption(Mat mat, MatOption op, PetscBool flg)
5865: {
5866: PetscFunctionBegin;
5868: if (op > 0) {
5871: }
5873: 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);
5875: switch (op) {
5876: case MAT_FORCE_DIAGONAL_ENTRIES:
5877: mat->force_diagonals = flg;
5878: PetscFunctionReturn(PETSC_SUCCESS);
5879: case MAT_NO_OFF_PROC_ENTRIES:
5880: mat->nooffprocentries = flg;
5881: PetscFunctionReturn(PETSC_SUCCESS);
5882: case MAT_SUBSET_OFF_PROC_ENTRIES:
5883: mat->assembly_subset = flg;
5884: if (!mat->assembly_subset) { /* See the same logic in VecAssembly wrt VEC_SUBSET_OFF_PROC_ENTRIES */
5885: #if !defined(PETSC_HAVE_MPIUNI)
5886: PetscCall(MatStashScatterDestroy_BTS(&mat->stash));
5887: #endif
5888: mat->stash.first_assembly_done = PETSC_FALSE;
5889: }
5890: PetscFunctionReturn(PETSC_SUCCESS);
5891: case MAT_NO_OFF_PROC_ZERO_ROWS:
5892: mat->nooffproczerorows = flg;
5893: PetscFunctionReturn(PETSC_SUCCESS);
5894: case MAT_SPD:
5895: if (flg) {
5896: mat->spd = PETSC_BOOL3_TRUE;
5897: mat->symmetric = PETSC_BOOL3_TRUE;
5898: mat->structurally_symmetric = PETSC_BOOL3_TRUE;
5899: } else {
5900: mat->spd = PETSC_BOOL3_FALSE;
5901: }
5902: break;
5903: case MAT_SYMMETRIC:
5904: mat->symmetric = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
5905: if (flg) mat->structurally_symmetric = PETSC_BOOL3_TRUE;
5906: #if !defined(PETSC_USE_COMPLEX)
5907: mat->hermitian = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
5908: #endif
5909: break;
5910: case MAT_HERMITIAN:
5911: mat->hermitian = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
5912: if (flg) mat->structurally_symmetric = PETSC_BOOL3_TRUE;
5913: #if !defined(PETSC_USE_COMPLEX)
5914: mat->symmetric = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
5915: #endif
5916: break;
5917: case MAT_STRUCTURALLY_SYMMETRIC:
5918: mat->structurally_symmetric = flg ? PETSC_BOOL3_TRUE : PETSC_BOOL3_FALSE;
5919: break;
5920: case MAT_SYMMETRY_ETERNAL:
5921: 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");
5922: mat->symmetry_eternal = flg;
5923: if (flg) mat->structural_symmetry_eternal = PETSC_TRUE;
5924: break;
5925: case MAT_STRUCTURAL_SYMMETRY_ETERNAL:
5926: PetscCheck(mat->structurally_symmetric != PETSC_BOOL3_UNKNOWN, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Cannot set MAT_STRUCTURAL_SYMMETRY_ETERNAL without first setting MAT_STRUCTURAL_SYMMETRIC to true or false");
5927: mat->structural_symmetry_eternal = flg;
5928: break;
5929: case MAT_SPD_ETERNAL:
5930: 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");
5931: mat->spd_eternal = flg;
5932: if (flg) {
5933: mat->structural_symmetry_eternal = PETSC_TRUE;
5934: mat->symmetry_eternal = PETSC_TRUE;
5935: }
5936: break;
5937: case MAT_STRUCTURE_ONLY:
5938: mat->structure_only = flg;
5939: break;
5940: case MAT_SORTED_FULL:
5941: mat->sortedfull = flg;
5942: break;
5943: default:
5944: break;
5945: }
5946: PetscTryTypeMethod(mat, setoption, op, flg);
5947: PetscFunctionReturn(PETSC_SUCCESS);
5948: }
5950: /*@
5951: MatGetOption - Gets a parameter option that has been set for a matrix.
5953: Logically Collective
5955: Input Parameters:
5956: + mat - the matrix
5957: - option - the option, this only responds to certain options, check the code for which ones
5959: Output Parameter:
5960: . flg - turn the option on (`PETSC_TRUE`) or off (`PETSC_FALSE`)
5962: Level: intermediate
5964: Notes:
5965: Can only be called after `MatSetSizes()` and `MatSetType()` have been set.
5967: Certain option values may be unknown, for those use the routines `MatIsSymmetric()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, or
5968: `MatIsSymmetricKnown()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetricKnown()`
5970: .seealso: [](chapter_matrices), `Mat`, `MatOption`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`,
5971: `MatIsSymmetricKnown()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetricKnown()`
5972: @*/
5973: PetscErrorCode MatGetOption(Mat mat, MatOption op, PetscBool *flg)
5974: {
5975: PetscFunctionBegin;
5979: 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);
5980: 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()");
5982: switch (op) {
5983: case MAT_NO_OFF_PROC_ENTRIES:
5984: *flg = mat->nooffprocentries;
5985: break;
5986: case MAT_NO_OFF_PROC_ZERO_ROWS:
5987: *flg = mat->nooffproczerorows;
5988: break;
5989: case MAT_SYMMETRIC:
5990: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsSymmetric() or MatIsSymmetricKnown()");
5991: break;
5992: case MAT_HERMITIAN:
5993: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsHermitian() or MatIsHermitianKnown()");
5994: break;
5995: case MAT_STRUCTURALLY_SYMMETRIC:
5996: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsStructurallySymmetric() or MatIsStructurallySymmetricKnown()");
5997: break;
5998: case MAT_SPD:
5999: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "Use MatIsSPDKnown()");
6000: break;
6001: case MAT_SYMMETRY_ETERNAL:
6002: *flg = mat->symmetry_eternal;
6003: break;
6004: case MAT_STRUCTURAL_SYMMETRY_ETERNAL:
6005: *flg = mat->symmetry_eternal;
6006: break;
6007: default:
6008: break;
6009: }
6010: PetscFunctionReturn(PETSC_SUCCESS);
6011: }
6013: /*@
6014: MatZeroEntries - Zeros all entries of a matrix. For sparse matrices
6015: this routine retains the old nonzero structure.
6017: Logically Collective
6019: Input Parameter:
6020: . mat - the matrix
6022: Level: intermediate
6024: Note:
6025: 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.
6026: See the Performance chapter of the users manual for information on preallocating matrices.
6028: .seealso: [](chapter_matrices), `Mat`, `MatZeroRows()`, `MatZeroRowsColumns()`
6029: @*/
6030: PetscErrorCode MatZeroEntries(Mat mat)
6031: {
6032: PetscFunctionBegin;
6035: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6036: 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");
6037: MatCheckPreallocated(mat, 1);
6039: PetscCall(PetscLogEventBegin(MAT_ZeroEntries, mat, 0, 0, 0));
6040: PetscUseTypeMethod(mat, zeroentries);
6041: PetscCall(PetscLogEventEnd(MAT_ZeroEntries, mat, 0, 0, 0));
6042: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6043: PetscFunctionReturn(PETSC_SUCCESS);
6044: }
6046: /*@
6047: MatZeroRowsColumns - Zeros all entries (except possibly the main diagonal)
6048: of a set of rows and columns of a matrix.
6050: Collective
6052: Input Parameters:
6053: + mat - the matrix
6054: . numRows - the number of rows/columns to zero
6055: . rows - the global row indices
6056: . diag - value put in the diagonal of the eliminated rows
6057: . x - optional vector of the solution for zeroed rows (other entries in vector are not used), these must be set before this call
6058: - b - optional vector of the right hand side, that will be adjusted by provided solution entries
6060: Level: intermediate
6062: Notes:
6063: This routine, along with `MatZeroRows()`, is typically used to eliminate known Dirichlet boundary conditions from a linear system.
6065: For each zeroed row, the value of the corresponding `b` is set to diag times the value of the corresponding `x`.
6066: 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
6068: If the resulting linear system is to be solved with `KSP` then one can (but does not have to) call `KSPSetInitialGuessNonzero()` to allow the
6069: Krylov method to take advantage of the known solution on the zeroed rows.
6071: For the parallel case, all processes that share the matrix (i.e.,
6072: those in the communicator used for matrix creation) MUST call this
6073: routine, regardless of whether any rows being zeroed are owned by
6074: them.
6076: Unlike `MatZeroRows()` this does not change the nonzero structure of the matrix, it merely zeros those entries in the matrix.
6078: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
6079: list only rows local to itself).
6081: The option `MAT_NO_OFF_PROC_ZERO_ROWS` does not apply to this routine.
6083: .seealso: [](chapter_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRows()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6084: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6085: @*/
6086: PetscErrorCode MatZeroRowsColumns(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6087: {
6088: PetscFunctionBegin;
6092: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6093: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6094: MatCheckPreallocated(mat, 1);
6096: PetscUseTypeMethod(mat, zerorowscolumns, numRows, rows, diag, x, b);
6097: PetscCall(MatViewFromOptions(mat, NULL, "-mat_view"));
6098: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6099: PetscFunctionReturn(PETSC_SUCCESS);
6100: }
6102: /*@
6103: MatZeroRowsColumnsIS - Zeros all entries (except possibly the main diagonal)
6104: of a set of rows and columns of a matrix.
6106: Collective
6108: Input Parameters:
6109: + mat - the matrix
6110: . is - the rows to zero
6111: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6112: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6113: - b - optional vector of right hand side, that will be adjusted by provided solution
6115: Level: intermediate
6117: Note:
6118: See `MatZeroRowsColumns()` for details on how this routine operates.
6120: .seealso: [](chapter_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6121: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRows()`, `MatZeroRowsColumnsStencil()`
6122: @*/
6123: PetscErrorCode MatZeroRowsColumnsIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6124: {
6125: PetscInt numRows;
6126: const PetscInt *rows;
6128: PetscFunctionBegin;
6133: PetscCall(ISGetLocalSize(is, &numRows));
6134: PetscCall(ISGetIndices(is, &rows));
6135: PetscCall(MatZeroRowsColumns(mat, numRows, rows, diag, x, b));
6136: PetscCall(ISRestoreIndices(is, &rows));
6137: PetscFunctionReturn(PETSC_SUCCESS);
6138: }
6140: /*@
6141: MatZeroRows - Zeros all entries (except possibly the main diagonal)
6142: of a set of rows of a matrix.
6144: Collective
6146: Input Parameters:
6147: + mat - the matrix
6148: . numRows - the number of rows to zero
6149: . rows - the global row indices
6150: . diag - value put in the diagonal of the zeroed rows
6151: . x - optional vector of solutions for zeroed rows (other entries in vector are not used), these must be set before this call
6152: - b - optional vector of right hand side, that will be adjusted by provided solution entries
6154: Level: intermediate
6156: Notes:
6157: This routine, along with `MatZeroRowsColumns()`, is typically used to eliminate known Dirichlet boundary conditions from a linear system.
6159: For each zeroed row, the value of the corresponding `b` is set to `diag` times the value of the corresponding `x`.
6161: If the resulting linear system is to be solved with `KSP` then one can (but does not have to) call `KSPSetInitialGuessNonzero()` to allow the
6162: Krylov method to take advantage of the known solution on the zeroed rows.
6164: 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)
6165: from the matrix.
6167: Unlike `MatZeroRowsColumns()` for the `MATAIJ` and `MATBAIJ` matrix formats this removes the old nonzero structure, from the eliminated rows of the matrix
6168: 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
6169: formats this does not alter the nonzero structure.
6171: If the option `MatSetOption`(mat,`MAT_KEEP_NONZERO_PATTERN`,`PETSC_TRUE`) the nonzero structure
6172: of the matrix is not changed the values are
6173: merely zeroed.
6175: The user can set a value in the diagonal entry (or for the `MATAIJ` format
6176: formats can optionally remove the main diagonal entry from the
6177: nonzero structure as well, by passing 0.0 as the final argument).
6179: For the parallel case, all processes that share the matrix (i.e.,
6180: those in the communicator used for matrix creation) MUST call this
6181: routine, regardless of whether any rows being zeroed are owned by
6182: them.
6184: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
6185: list only rows local to itself).
6187: You can call `MatSetOption`(mat,`MAT_NO_OFF_PROC_ZERO_ROWS`,`PETSC_TRUE`) if each process indicates only rows it
6188: owns that are to be zeroed. This saves a global synchronization in the implementation.
6190: .seealso: [](chapter_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6191: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`, `PCREDISTRIBUTE`
6192: @*/
6193: PetscErrorCode MatZeroRows(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6194: {
6195: PetscFunctionBegin;
6199: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6200: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6201: MatCheckPreallocated(mat, 1);
6203: PetscUseTypeMethod(mat, zerorows, numRows, rows, diag, x, b);
6204: PetscCall(MatViewFromOptions(mat, NULL, "-mat_view"));
6205: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6206: PetscFunctionReturn(PETSC_SUCCESS);
6207: }
6209: /*@
6210: MatZeroRowsIS - Zeros all entries (except possibly the main diagonal)
6211: of a set of rows of a matrix.
6213: Collective
6215: Input Parameters:
6216: + mat - the matrix
6217: . is - index set of rows to remove (if `NULL` then no row is removed)
6218: . diag - value put in all diagonals of eliminated rows
6219: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6220: - b - optional vector of right hand side, that will be adjusted by provided solution
6222: Level: intermediate
6224: Note:
6225: See `MatZeroRows()` for details on how this routine operates.
6227: .seealso: [](chapter_matrices), `Mat`, `MatZeroRows()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6228: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6229: @*/
6230: PetscErrorCode MatZeroRowsIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6231: {
6232: PetscInt numRows = 0;
6233: const PetscInt *rows = NULL;
6235: PetscFunctionBegin;
6238: if (is) {
6240: PetscCall(ISGetLocalSize(is, &numRows));
6241: PetscCall(ISGetIndices(is, &rows));
6242: }
6243: PetscCall(MatZeroRows(mat, numRows, rows, diag, x, b));
6244: if (is) PetscCall(ISRestoreIndices(is, &rows));
6245: PetscFunctionReturn(PETSC_SUCCESS);
6246: }
6248: /*@
6249: MatZeroRowsStencil - Zeros all entries (except possibly the main diagonal)
6250: of a set of rows of a matrix. These rows must be local to the process.
6252: Collective
6254: Input Parameters:
6255: + mat - the matrix
6256: . numRows - the number of rows to remove
6257: . rows - the grid coordinates (and component number when dof > 1) for matrix rows
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: Notes:
6265: See `MatZeroRows()` for details on how this routine operates.
6267: The grid coordinates are across the entire grid, not just the local portion
6269: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
6270: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
6271: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
6272: `DM_BOUNDARY_PERIODIC` boundary type.
6274: 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
6275: a single value per point) you can skip filling those indices.
6277: Fortran Note:
6278: `idxm` and `idxn` should be declared as
6279: $ MatStencil idxm(4,m)
6280: and the values inserted using
6281: .vb
6282: idxm(MatStencil_i,1) = i
6283: idxm(MatStencil_j,1) = j
6284: idxm(MatStencil_k,1) = k
6285: idxm(MatStencil_c,1) = c
6286: etc
6287: .ve
6289: .seealso: [](chapter_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsl()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6290: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6291: @*/
6292: PetscErrorCode MatZeroRowsStencil(Mat mat, PetscInt numRows, const MatStencil rows[], PetscScalar diag, Vec x, Vec b)
6293: {
6294: PetscInt dim = mat->stencil.dim;
6295: PetscInt sdim = dim - (1 - (PetscInt)mat->stencil.noc);
6296: PetscInt *dims = mat->stencil.dims + 1;
6297: PetscInt *starts = mat->stencil.starts;
6298: PetscInt *dxm = (PetscInt *)rows;
6299: PetscInt *jdxm, i, j, tmp, numNewRows = 0;
6301: PetscFunctionBegin;
6306: PetscCall(PetscMalloc1(numRows, &jdxm));
6307: for (i = 0; i < numRows; ++i) {
6308: /* Skip unused dimensions (they are ordered k, j, i, c) */
6309: for (j = 0; j < 3 - sdim; ++j) dxm++;
6310: /* Local index in X dir */
6311: tmp = *dxm++ - starts[0];
6312: /* Loop over remaining dimensions */
6313: for (j = 0; j < dim - 1; ++j) {
6314: /* If nonlocal, set index to be negative */
6315: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
6316: /* Update local index */
6317: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
6318: }
6319: /* Skip component slot if necessary */
6320: if (mat->stencil.noc) dxm++;
6321: /* Local row number */
6322: if (tmp >= 0) jdxm[numNewRows++] = tmp;
6323: }
6324: PetscCall(MatZeroRowsLocal(mat, numNewRows, jdxm, diag, x, b));
6325: PetscCall(PetscFree(jdxm));
6326: PetscFunctionReturn(PETSC_SUCCESS);
6327: }
6329: /*@
6330: MatZeroRowsColumnsStencil - Zeros all row and column entries (except possibly the main diagonal)
6331: of a set of rows and columns of a matrix.
6333: Collective
6335: Input Parameters:
6336: + mat - the matrix
6337: . numRows - the number of rows/columns to remove
6338: . rows - the grid coordinates (and component number when dof > 1) for matrix rows
6339: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6340: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6341: - b - optional vector of right hand side, that will be adjusted by provided solution
6343: Level: intermediate
6345: Notes:
6346: See `MatZeroRowsColumns()` for details on how this routine operates.
6348: The grid coordinates are across the entire grid, not just the local portion
6350: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
6351: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
6352: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
6353: `DM_BOUNDARY_PERIODIC` boundary type.
6355: 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
6356: a single value per point) you can skip filling those indices.
6358: Fortran Note:
6359: `idxm` and `idxn` should be declared as
6360: $ MatStencil idxm(4,m)
6361: and the values inserted using
6362: .vb
6363: idxm(MatStencil_i,1) = i
6364: idxm(MatStencil_j,1) = j
6365: idxm(MatStencil_k,1) = k
6366: idxm(MatStencil_c,1) = c
6367: etc
6368: .ve
6370: .seealso: [](chapter_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6371: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRows()`
6372: @*/
6373: PetscErrorCode MatZeroRowsColumnsStencil(Mat mat, PetscInt numRows, const MatStencil rows[], PetscScalar diag, Vec x, Vec b)
6374: {
6375: PetscInt dim = mat->stencil.dim;
6376: PetscInt sdim = dim - (1 - (PetscInt)mat->stencil.noc);
6377: PetscInt *dims = mat->stencil.dims + 1;
6378: PetscInt *starts = mat->stencil.starts;
6379: PetscInt *dxm = (PetscInt *)rows;
6380: PetscInt *jdxm, i, j, tmp, numNewRows = 0;
6382: PetscFunctionBegin;
6387: PetscCall(PetscMalloc1(numRows, &jdxm));
6388: for (i = 0; i < numRows; ++i) {
6389: /* Skip unused dimensions (they are ordered k, j, i, c) */
6390: for (j = 0; j < 3 - sdim; ++j) dxm++;
6391: /* Local index in X dir */
6392: tmp = *dxm++ - starts[0];
6393: /* Loop over remaining dimensions */
6394: for (j = 0; j < dim - 1; ++j) {
6395: /* If nonlocal, set index to be negative */
6396: if ((*dxm++ - starts[j + 1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
6397: /* Update local index */
6398: else tmp = tmp * dims[j] + *(dxm - 1) - starts[j + 1];
6399: }
6400: /* Skip component slot if necessary */
6401: if (mat->stencil.noc) dxm++;
6402: /* Local row number */
6403: if (tmp >= 0) jdxm[numNewRows++] = tmp;
6404: }
6405: PetscCall(MatZeroRowsColumnsLocal(mat, numNewRows, jdxm, diag, x, b));
6406: PetscCall(PetscFree(jdxm));
6407: PetscFunctionReturn(PETSC_SUCCESS);
6408: }
6410: /*@C
6411: MatZeroRowsLocal - Zeros all entries (except possibly the main diagonal)
6412: of a set of rows of a matrix; using local numbering of rows.
6414: Collective
6416: Input Parameters:
6417: + mat - the matrix
6418: . numRows - the number of rows to remove
6419: . rows - the local row indices
6420: . diag - value put in all diagonals of eliminated rows
6421: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6422: - b - optional vector of right hand side, that will be adjusted by provided solution
6424: Level: intermediate
6426: Notes:
6427: Before calling `MatZeroRowsLocal()`, the user must first set the
6428: local-to-global mapping by calling MatSetLocalToGlobalMapping(), this is often already set for matrices obtained with `DMCreateMatrix()`.
6430: See `MatZeroRows()` for details on how this routine operates.
6432: .seealso: [](chapter_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRows()`, `MatSetOption()`,
6433: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6434: @*/
6435: PetscErrorCode MatZeroRowsLocal(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6436: {
6437: PetscFunctionBegin;
6441: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6442: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6443: MatCheckPreallocated(mat, 1);
6445: if (mat->ops->zerorowslocal) {
6446: PetscUseTypeMethod(mat, zerorowslocal, numRows, rows, diag, x, b);
6447: } else {
6448: IS is, newis;
6449: const PetscInt *newRows;
6451: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Need to provide local to global mapping to matrix first");
6452: PetscCall(ISCreateGeneral(PETSC_COMM_SELF, numRows, rows, PETSC_COPY_VALUES, &is));
6453: PetscCall(ISLocalToGlobalMappingApplyIS(mat->rmap->mapping, is, &newis));
6454: PetscCall(ISGetIndices(newis, &newRows));
6455: PetscUseTypeMethod(mat, zerorows, numRows, newRows, diag, x, b);
6456: PetscCall(ISRestoreIndices(newis, &newRows));
6457: PetscCall(ISDestroy(&newis));
6458: PetscCall(ISDestroy(&is));
6459: }
6460: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6461: PetscFunctionReturn(PETSC_SUCCESS);
6462: }
6464: /*@
6465: MatZeroRowsLocalIS - Zeros all entries (except possibly the main diagonal)
6466: of a set of rows of a matrix; using local numbering of rows.
6468: Collective
6470: Input Parameters:
6471: + mat - the matrix
6472: . is - index set of rows to remove
6473: . diag - value put in all diagonals of eliminated rows
6474: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6475: - b - optional vector of right hand side, that will be adjusted by provided solution
6477: Level: intermediate
6479: Notes:
6480: Before calling `MatZeroRowsLocalIS()`, the user must first set the
6481: local-to-global mapping by calling `MatSetLocalToGlobalMapping()`, this is often already set for matrices obtained with `DMCreateMatrix()`.
6483: See `MatZeroRows()` for details on how this routine operates.
6485: .seealso: [](chapter_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRows()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6486: `MatZeroRowsColumnsLocal()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6487: @*/
6488: PetscErrorCode MatZeroRowsLocalIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6489: {
6490: PetscInt numRows;
6491: const PetscInt *rows;
6493: PetscFunctionBegin;
6497: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6498: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6499: MatCheckPreallocated(mat, 1);
6501: PetscCall(ISGetLocalSize(is, &numRows));
6502: PetscCall(ISGetIndices(is, &rows));
6503: PetscCall(MatZeroRowsLocal(mat, numRows, rows, diag, x, b));
6504: PetscCall(ISRestoreIndices(is, &rows));
6505: PetscFunctionReturn(PETSC_SUCCESS);
6506: }
6508: /*@
6509: MatZeroRowsColumnsLocal - Zeros all entries (except possibly the main diagonal)
6510: of a set of rows and columns of a matrix; using local numbering of rows.
6512: Collective
6514: Input Parameters:
6515: + mat - the matrix
6516: . numRows - the number of rows to remove
6517: . rows - the global row indices
6518: . diag - value put in all diagonals of eliminated rows
6519: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6520: - b - optional vector of right hand side, that will be adjusted by provided solution
6522: Level: intermediate
6524: Notes:
6525: Before calling `MatZeroRowsColumnsLocal()`, the user must first set the
6526: local-to-global mapping by calling `MatSetLocalToGlobalMapping()`, this is often already set for matrices obtained with `DMCreateMatrix()`.
6528: See `MatZeroRowsColumns()` for details on how this routine operates.
6530: .seealso: [](chapter_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6531: `MatZeroRows()`, `MatZeroRowsColumnsLocalIS()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6532: @*/
6533: PetscErrorCode MatZeroRowsColumnsLocal(Mat mat, PetscInt numRows, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
6534: {
6535: IS is, newis;
6536: const PetscInt *newRows;
6538: PetscFunctionBegin;
6542: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6543: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6544: MatCheckPreallocated(mat, 1);
6546: PetscCheck(mat->cmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Need to provide local to global mapping to matrix first");
6547: PetscCall(ISCreateGeneral(PETSC_COMM_SELF, numRows, rows, PETSC_COPY_VALUES, &is));
6548: PetscCall(ISLocalToGlobalMappingApplyIS(mat->cmap->mapping, is, &newis));
6549: PetscCall(ISGetIndices(newis, &newRows));
6550: PetscUseTypeMethod(mat, zerorowscolumns, numRows, newRows, diag, x, b);
6551: PetscCall(ISRestoreIndices(newis, &newRows));
6552: PetscCall(ISDestroy(&newis));
6553: PetscCall(ISDestroy(&is));
6554: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
6555: PetscFunctionReturn(PETSC_SUCCESS);
6556: }
6558: /*@
6559: MatZeroRowsColumnsLocalIS - Zeros all entries (except possibly the main diagonal)
6560: of a set of rows and columns of a matrix; using local numbering of rows.
6562: Collective
6564: Input Parameters:
6565: + mat - the matrix
6566: . is - index set of rows to remove
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 `MatZeroRowsColumnsLocalIS()`, 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 `MatZeroRowsColumns()` for details on how this routine operates.
6579: .seealso: [](chapter_matrices), `Mat`, `MatZeroRowsIS()`, `MatZeroRowsColumns()`, `MatZeroRowsLocalIS()`, `MatZeroRowsStencil()`, `MatZeroEntries()`, `MatZeroRowsLocal()`, `MatSetOption()`,
6580: `MatZeroRowsColumnsLocal()`, `MatZeroRows()`, `MatZeroRowsColumnsIS()`, `MatZeroRowsColumnsStencil()`
6581: @*/
6582: PetscErrorCode MatZeroRowsColumnsLocalIS(Mat mat, IS is, PetscScalar diag, Vec x, Vec b)
6583: {
6584: PetscInt numRows;
6585: const PetscInt *rows;
6587: PetscFunctionBegin;
6591: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6592: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6593: MatCheckPreallocated(mat, 1);
6595: PetscCall(ISGetLocalSize(is, &numRows));
6596: PetscCall(ISGetIndices(is, &rows));
6597: PetscCall(MatZeroRowsColumnsLocal(mat, numRows, rows, diag, x, b));
6598: PetscCall(ISRestoreIndices(is, &rows));
6599: PetscFunctionReturn(PETSC_SUCCESS);
6600: }
6602: /*@C
6603: MatGetSize - Returns the numbers of rows and columns in a matrix.
6605: Not Collective
6607: Input Parameter:
6608: . mat - the matrix
6610: Output Parameters:
6611: + m - the number of global rows
6612: - n - the number of global columns
6614: Level: beginner
6616: Note:
6617: Both output parameters can be `NULL` on input.
6619: .seealso: [](chapter_matrices), `Mat`, `MatSetSizes()`, `MatGetLocalSize()`
6620: @*/
6621: PetscErrorCode MatGetSize(Mat mat, PetscInt *m, PetscInt *n)
6622: {
6623: PetscFunctionBegin;
6625: if (m) *m = mat->rmap->N;
6626: if (n) *n = mat->cmap->N;
6627: PetscFunctionReturn(PETSC_SUCCESS);
6628: }
6630: /*@C
6631: MatGetLocalSize - For most matrix formats, excluding `MATELEMENTAL` and `MATSCALAPACK`, Returns the number of local rows and local columns
6632: of a matrix. For all matrices this is the local size of the left and right vectors as returned by `MatCreateVecs()`.
6634: Not Collective
6636: Input Parameter:
6637: . mat - the matrix
6639: Output Parameters:
6640: + m - the number of local rows, use `NULL` to not obtain this value
6641: - n - the number of local columns, use `NULL` to not obtain this value
6643: Level: beginner
6645: .seealso: [](chapter_matrices), `Mat`, `MatSetSizes()`, `MatGetSize()`
6646: @*/
6647: PetscErrorCode MatGetLocalSize(Mat mat, PetscInt *m, PetscInt *n)
6648: {
6649: PetscFunctionBegin;
6653: if (m) *m = mat->rmap->n;
6654: if (n) *n = mat->cmap->n;
6655: PetscFunctionReturn(PETSC_SUCCESS);
6656: }
6658: /*@C
6659: MatGetOwnershipRangeColumn - Returns the range of matrix columns associated with rows of a vector one multiplies this matrix by that are owned by
6660: this processor. (The columns of the "diagonal block" for most sparse matrix formats). See :any:`<sec_matlayout>` for details on matrix layouts.
6662: Not Collective, unless matrix has not been allocated, then collective
6664: Input Parameter:
6665: . mat - the matrix
6667: Output Parameters:
6668: + m - the global index of the first local column, use `NULL` to not obtain this value
6669: - n - one more than the global index of the last local column, use `NULL` to not obtain this value
6671: Level: developer
6673: .seealso: [](chapter_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRanges()`, `MatGetOwnershipRangesColumn()`, `PetscLayout`
6674: @*/
6675: PetscErrorCode MatGetOwnershipRangeColumn(Mat mat, PetscInt *m, PetscInt *n)
6676: {
6677: PetscFunctionBegin;
6682: MatCheckPreallocated(mat, 1);
6683: if (m) *m = mat->cmap->rstart;
6684: if (n) *n = mat->cmap->rend;
6685: PetscFunctionReturn(PETSC_SUCCESS);
6686: }
6688: /*@C
6689: MatGetOwnershipRange - For matrices that own values by row, excludes `MATELEMENTAL` and `MATSCALAPACK`, returns the range of matrix rows owned by
6690: this MPI rank. For all matrices it returns the range of matrix rows associated with rows of a vector that would contain the result of a matrix
6691: vector product with this matrix. See :any:`<sec_matlayout>` for details on matrix layouts
6693: Not Collective
6695: Input Parameter:
6696: . mat - the matrix
6698: Output Parameters:
6699: + m - the global index of the first local row, use `NULL` to not obtain this value
6700: - n - one more than the global index of the last local row, use `NULL` to not obtain this value
6702: Level: beginner
6704: Note:
6705: This function requires that the matrix be preallocated. If you have not preallocated, consider using
6706: `PetscSplitOwnership`(`MPI_Comm` comm, `PetscInt` *n, `PetscInt` *N)
6707: and then `MPI_Scan()` to calculate prefix sums of the local sizes.
6709: .seealso: [](chapter_matrices), `Mat`, `MatGetOwnershipRanges()`, `MatGetOwnershipRangeColumn()`, `MatGetOwnershipRangesColumn()`, `PetscSplitOwnership()`, `PetscSplitOwnershipBlock()`,
6710: `PetscLayout`
6711: @*/
6712: PetscErrorCode MatGetOwnershipRange(Mat mat, PetscInt *m, PetscInt *n)
6713: {
6714: PetscFunctionBegin;
6719: MatCheckPreallocated(mat, 1);
6720: if (m) *m = mat->rmap->rstart;
6721: if (n) *n = mat->rmap->rend;
6722: PetscFunctionReturn(PETSC_SUCCESS);
6723: }
6725: /*@C
6726: MatGetOwnershipRanges - For matrices that own values by row, excludes `MATELEMENTAL` and `MATSCALAPACK`, returns the range of matrix rows owned by
6727: each process. For all matrices it returns the ranges of matrix rows associated with rows of a vector that would contain the result of a matrix
6728: vector product with this matrix. See :any:`<sec_matlayout>` for details on matrix layouts
6730: Not Collective, unless matrix has not been allocated
6732: Input Parameter:
6733: . mat - the matrix
6735: Output Parameter:
6736: . ranges - start of each processors portion plus one more than the total length at the end
6738: Level: beginner
6740: .seealso: [](chapter_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`, `MatGetOwnershipRangesColumn()`, `PetscLayout`
6741: @*/
6742: PetscErrorCode MatGetOwnershipRanges(Mat mat, const PetscInt **ranges)
6743: {
6744: PetscFunctionBegin;
6747: MatCheckPreallocated(mat, 1);
6748: PetscCall(PetscLayoutGetRanges(mat->rmap, ranges));
6749: PetscFunctionReturn(PETSC_SUCCESS);
6750: }
6752: /*@C
6753: MatGetOwnershipRangesColumn - Returns the ranges of matrix columns associated with rows of a vector one multiplies this vector by that are owned by
6754: each processor. (The columns of the "diagonal blocks", for most sparse matrix formats). See :any:`<sec_matlayout>` for details on matrix layouts.
6756: Not Collective, unless matrix has not been allocated
6758: Input Parameter:
6759: . mat - the matrix
6761: Output Parameter:
6762: . ranges - start of each processors portion plus one more then the total length at the end
6764: Level: beginner
6766: .seealso: [](chapter_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`, `MatGetOwnershipRanges()`
6767: @*/
6768: PetscErrorCode MatGetOwnershipRangesColumn(Mat mat, const PetscInt **ranges)
6769: {
6770: PetscFunctionBegin;
6773: MatCheckPreallocated(mat, 1);
6774: PetscCall(PetscLayoutGetRanges(mat->cmap, ranges));
6775: PetscFunctionReturn(PETSC_SUCCESS);
6776: }
6778: /*@C
6779: MatGetOwnershipIS - Get row and column ownership of a matrices' values as index sets. For most matrices, excluding `MATELEMENTAL` and `MATSCALAPACK`, this
6780: corresponds to values returned by `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`. For `MATELEMENTAL` and `MATSCALAPACK` the ownership
6781: is more complicated. See :any:`<sec_matlayout>` for details on matrix layouts.
6783: Not Collective
6785: Input Parameter:
6786: . A - matrix
6788: Output Parameters:
6789: + rows - rows in which this process owns elements, , use `NULL` to not obtain this value
6790: - cols - columns in which this process owns elements, use `NULL` to not obtain this value
6792: Level: intermediate
6794: .seealso: [](chapter_matrices), `Mat`, `MatGetOwnershipRange()`, `MatGetOwnershipRangeColumn()`, `MatSetValues()`, ``MATELEMENTAL``, ``MATSCALAPACK``
6795: @*/
6796: PetscErrorCode MatGetOwnershipIS(Mat A, IS *rows, IS *cols)
6797: {
6798: PetscErrorCode (*f)(Mat, IS *, IS *);
6800: PetscFunctionBegin;
6801: MatCheckPreallocated(A, 1);
6802: PetscCall(PetscObjectQueryFunction((PetscObject)A, "MatGetOwnershipIS_C", &f));
6803: if (f) {
6804: PetscCall((*f)(A, rows, cols));
6805: } else { /* Create a standard row-based partition, each process is responsible for ALL columns in their row block */
6806: if (rows) PetscCall(ISCreateStride(PETSC_COMM_SELF, A->rmap->n, A->rmap->rstart, 1, rows));
6807: if (cols) PetscCall(ISCreateStride(PETSC_COMM_SELF, A->cmap->N, 0, 1, cols));
6808: }
6809: PetscFunctionReturn(PETSC_SUCCESS);
6810: }
6812: /*@C
6813: MatILUFactorSymbolic - Performs symbolic ILU factorization of a matrix obtained with `MatGetFactor()`
6814: Uses levels of fill only, not drop tolerance. Use `MatLUFactorNumeric()`
6815: to complete the factorization.
6817: Collective
6819: Input Parameters:
6820: + fact - the factorized matrix obtained with `MatGetFactor()`
6821: . mat - the matrix
6822: . row - row permutation
6823: . col - column permutation
6824: - info - structure containing
6825: .vb
6826: levels - number of levels of fill.
6827: expected fill - as ratio of original fill.
6828: 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
6829: missing diagonal entries)
6830: .ve
6832: Level: developer
6834: Notes:
6835: See [Matrix Factorization](sec_matfactor) for additional information.
6837: Most users should employ the `KSP` interface for linear solvers
6838: instead of working directly with matrix algebra routines such as this.
6839: See, e.g., `KSPCreate()`.
6841: Uses the definition of level of fill as in Y. Saad, 2003
6843: Developer Note:
6844: The Fortran interface is not autogenerated as the
6845: interface definition cannot be generated correctly [due to `MatFactorInfo`]
6847: References:
6848: . * - Y. Saad, Iterative methods for sparse linear systems Philadelphia: Society for Industrial and Applied Mathematics, 2003
6850: .seealso: [](chapter_matrices), `Mat`, [Matrix Factorization](sec_matfactor), `MatGetFactor()`, `MatLUFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`
6851: `MatGetOrdering()`, `MatFactorInfo`
6852: @*/
6853: PetscErrorCode MatILUFactorSymbolic(Mat fact, Mat mat, IS row, IS col, const MatFactorInfo *info)
6854: {
6855: PetscFunctionBegin;
6862: PetscCheck(info->levels >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Levels of fill negative %" PetscInt_FMT, (PetscInt)info->levels);
6863: PetscCheck(info->fill >= 1.0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Expected fill less than 1.0 %g", (double)info->fill);
6864: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6865: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6866: MatCheckPreallocated(mat, 2);
6868: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_ILUFactorSymbolic, mat, row, col, 0));
6869: PetscUseTypeMethod(fact, ilufactorsymbolic, mat, row, col, info);
6870: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_ILUFactorSymbolic, mat, row, col, 0));
6871: PetscFunctionReturn(PETSC_SUCCESS);
6872: }
6874: /*@C
6875: MatICCFactorSymbolic - Performs symbolic incomplete
6876: Cholesky factorization for a symmetric matrix. Use
6877: `MatCholeskyFactorNumeric()` to complete the factorization.
6879: Collective
6881: Input Parameters:
6882: + fact - the factorized matrix obtained with `MatGetFactor()`
6883: . mat - the matrix to be factored
6884: . perm - row and column permutation
6885: - info - structure containing
6886: .vb
6887: levels - number of levels of fill.
6888: expected fill - as ratio of original fill.
6889: .ve
6891: Level: developer
6893: Notes:
6894: Most users should employ the `KSP` interface for linear solvers
6895: instead of working directly with matrix algebra routines such as this.
6896: See, e.g., `KSPCreate()`.
6898: This uses the definition of level of fill as in Y. Saad, 2003
6900: Developer Note:
6901: The Fortran interface is not autogenerated as the
6902: interface definition cannot be generated correctly [due to `MatFactorInfo`]
6904: References:
6905: . * - Y. Saad, Iterative methods for sparse linear systems Philadelphia: Society for Industrial and Applied Mathematics, 2003
6907: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `MatCholeskyFactorNumeric()`, `MatCholeskyFactor()`, `MatFactorInfo`
6908: @*/
6909: PetscErrorCode MatICCFactorSymbolic(Mat fact, Mat mat, IS perm, const MatFactorInfo *info)
6910: {
6911: PetscFunctionBegin;
6917: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
6918: PetscCheck(info->levels >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Levels negative %" PetscInt_FMT, (PetscInt)info->levels);
6919: PetscCheck(info->fill >= 1.0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Expected fill less than 1.0 %g", (double)info->fill);
6920: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
6921: MatCheckPreallocated(mat, 2);
6923: if (!fact->trivialsymbolic) PetscCall(PetscLogEventBegin(MAT_ICCFactorSymbolic, mat, perm, 0, 0));
6924: PetscUseTypeMethod(fact, iccfactorsymbolic, mat, perm, info);
6925: if (!fact->trivialsymbolic) PetscCall(PetscLogEventEnd(MAT_ICCFactorSymbolic, mat, perm, 0, 0));
6926: PetscFunctionReturn(PETSC_SUCCESS);
6927: }
6929: /*@C
6930: MatCreateSubMatrices - Extracts several submatrices from a matrix. If submat
6931: points to an array of valid matrices, they may be reused to store the new
6932: submatrices.
6934: Collective
6936: Input Parameters:
6937: + mat - the matrix
6938: . n - the number of submatrixes to be extracted (on this processor, may be zero)
6939: . irow - index set of rows to extract
6940: . icol - index set of columns to extract
6941: - scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
6943: Output Parameter:
6944: . submat - the array of submatrices
6946: Level: advanced
6948: Notes:
6949: `MatCreateSubMatrices()` can extract ONLY sequential submatrices
6950: (from both sequential and parallel matrices). Use `MatCreateSubMatrix()`
6951: to extract a parallel submatrix.
6953: Some matrix types place restrictions on the row and column
6954: indices, such as that they be sorted or that they be equal to each other.
6956: The index sets may not have duplicate entries.
6958: When extracting submatrices from a parallel matrix, each processor can
6959: form a different submatrix by setting the rows and columns of its
6960: individual index sets according to the local submatrix desired.
6962: When finished using the submatrices, the user should destroy
6963: them with `MatDestroySubMatrices()`.
6965: `MAT_REUSE_MATRIX` can only be used when the nonzero structure of the
6966: original matrix has not changed from that last call to `MatCreateSubMatrices()`.
6968: This routine creates the matrices in submat; you should NOT create them before
6969: calling it. It also allocates the array of matrix pointers submat.
6971: For `MATBAIJ` matrices the index sets must respect the block structure, that is if they
6972: request one row/column in a block, they must request all rows/columns that are in
6973: that block. For example, if the block size is 2 you cannot request just row 0 and
6974: column 0.
6976: Fortran Note:
6977: The Fortran interface is slightly different from that given below; it
6978: requires one to pass in as `submat` a `Mat` (integer) array of size at least n+1.
6980: .seealso: [](chapter_matrices), `Mat`, `MatDestroySubMatrices()`, `MatCreateSubMatrix()`, `MatGetRow()`, `MatGetDiagonal()`, `MatReuse`
6981: @*/
6982: PetscErrorCode MatCreateSubMatrices(Mat mat, PetscInt n, const IS irow[], const IS icol[], MatReuse scall, Mat *submat[])
6983: {
6984: PetscInt i;
6985: PetscBool eq;
6987: PetscFunctionBegin;
6990: if (n) {
6995: }
6997: if (n && scall == MAT_REUSE_MATRIX) {
7000: }
7001: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7002: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7003: MatCheckPreallocated(mat, 1);
7004: PetscCall(PetscLogEventBegin(MAT_CreateSubMats, mat, 0, 0, 0));
7005: PetscUseTypeMethod(mat, createsubmatrices, n, irow, icol, scall, submat);
7006: PetscCall(PetscLogEventEnd(MAT_CreateSubMats, mat, 0, 0, 0));
7007: for (i = 0; i < n; i++) {
7008: (*submat)[i]->factortype = MAT_FACTOR_NONE; /* in case in place factorization was previously done on submatrix */
7009: PetscCall(ISEqualUnsorted(irow[i], icol[i], &eq));
7010: if (eq) PetscCall(MatPropagateSymmetryOptions(mat, (*submat)[i]));
7011: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
7012: if (mat->boundtocpu && mat->bindingpropagates) {
7013: PetscCall(MatBindToCPU((*submat)[i], PETSC_TRUE));
7014: PetscCall(MatSetBindingPropagates((*submat)[i], PETSC_TRUE));
7015: }
7016: #endif
7017: }
7018: PetscFunctionReturn(PETSC_SUCCESS);
7019: }
7021: /*@C
7022: MatCreateSubMatricesMPI - Extracts MPI submatrices across a sub communicator of mat (by pairs of `IS` that may live on subcomms).
7024: Collective
7026: Input Parameters:
7027: + mat - the matrix
7028: . n - the number of submatrixes to be extracted
7029: . irow - index set of rows to extract
7030: . icol - index set of columns to extract
7031: - scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
7033: Output Parameter:
7034: . submat - the array of submatrices
7036: Level: advanced
7038: Note:
7039: This is used by `PCGASM`
7041: .seealso: [](chapter_matrices), `Mat`, `PCGASM`, `MatCreateSubMatrices()`, `MatCreateSubMatrix()`, `MatGetRow()`, `MatGetDiagonal()`, `MatReuse`
7042: @*/
7043: PetscErrorCode MatCreateSubMatricesMPI(Mat mat, PetscInt n, const IS irow[], const IS icol[], MatReuse scall, Mat *submat[])
7044: {
7045: PetscInt i;
7046: PetscBool eq;
7048: PetscFunctionBegin;
7051: if (n) {
7056: }
7058: if (n && scall == MAT_REUSE_MATRIX) {
7061: }
7062: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7063: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7064: MatCheckPreallocated(mat, 1);
7066: PetscCall(PetscLogEventBegin(MAT_CreateSubMats, mat, 0, 0, 0));
7067: PetscUseTypeMethod(mat, createsubmatricesmpi, n, irow, icol, scall, submat);
7068: PetscCall(PetscLogEventEnd(MAT_CreateSubMats, mat, 0, 0, 0));
7069: for (i = 0; i < n; i++) {
7070: PetscCall(ISEqualUnsorted(irow[i], icol[i], &eq));
7071: if (eq) PetscCall(MatPropagateSymmetryOptions(mat, (*submat)[i]));
7072: }
7073: PetscFunctionReturn(PETSC_SUCCESS);
7074: }
7076: /*@C
7077: MatDestroyMatrices - Destroys an array of matrices.
7079: Collective
7081: Input Parameters:
7082: + n - the number of local matrices
7083: - mat - the matrices (this is a pointer to the array of matrices)
7085: Level: advanced
7087: Note:
7088: Frees not only the matrices, but also the array that contains the matrices
7090: Fortran Note:
7091: This does not free the array.
7093: .seealso: [](chapter_matrices), `Mat`, `MatCreateSubMatrices()` `MatDestroySubMatrices()`
7094: @*/
7095: PetscErrorCode MatDestroyMatrices(PetscInt n, Mat *mat[])
7096: {
7097: PetscInt i;
7099: PetscFunctionBegin;
7100: if (!*mat) PetscFunctionReturn(PETSC_SUCCESS);
7101: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Trying to destroy negative number of matrices %" PetscInt_FMT, n);
7104: for (i = 0; i < n; i++) PetscCall(MatDestroy(&(*mat)[i]));
7106: /* memory is allocated even if n = 0 */
7107: PetscCall(PetscFree(*mat));
7108: PetscFunctionReturn(PETSC_SUCCESS);
7109: }
7111: /*@C
7112: MatDestroySubMatrices - Destroys a set of matrices obtained with `MatCreateSubMatrices()`.
7114: Collective
7116: Input Parameters:
7117: + n - the number of local matrices
7118: - mat - the matrices (this is a pointer to the array of matrices, just to match the calling
7119: sequence of `MatCreateSubMatrices()`)
7121: Level: advanced
7123: Note:
7124: Frees not only the matrices, but also the array that contains the matrices
7126: Fortran Note:
7127: This does not free the array.
7129: .seealso: [](chapter_matrices), `Mat`, `MatCreateSubMatrices()`, `MatDestroyMatrices()`
7130: @*/
7131: PetscErrorCode MatDestroySubMatrices(PetscInt n, Mat *mat[])
7132: {
7133: Mat mat0;
7135: PetscFunctionBegin;
7136: if (!*mat) PetscFunctionReturn(PETSC_SUCCESS);
7137: /* mat[] is an array of length n+1, see MatCreateSubMatrices_xxx() */
7138: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Trying to destroy negative number of matrices %" PetscInt_FMT, n);
7141: mat0 = (*mat)[0];
7142: if (mat0 && mat0->ops->destroysubmatrices) {
7143: PetscCall((*mat0->ops->destroysubmatrices)(n, mat));
7144: } else {
7145: PetscCall(MatDestroyMatrices(n, mat));
7146: }
7147: PetscFunctionReturn(PETSC_SUCCESS);
7148: }
7150: /*@C
7151: MatGetSeqNonzeroStructure - Extracts the nonzero structure from a matrix and stores it, in its entirety, on each process
7153: Collective
7155: Input Parameter:
7156: . mat - the matrix
7158: Output Parameter:
7159: . matstruct - the sequential matrix with the nonzero structure of mat
7161: Level: developer
7163: .seealso: [](chapter_matrices), `Mat`, `MatDestroySeqNonzeroStructure()`, `MatCreateSubMatrices()`, `MatDestroyMatrices()`
7164: @*/
7165: PetscErrorCode MatGetSeqNonzeroStructure(Mat mat, Mat *matstruct)
7166: {
7167: PetscFunctionBegin;
7172: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7173: MatCheckPreallocated(mat, 1);
7175: PetscCall(PetscLogEventBegin(MAT_GetSeqNonzeroStructure, mat, 0, 0, 0));
7176: PetscUseTypeMethod(mat, getseqnonzerostructure, matstruct);
7177: PetscCall(PetscLogEventEnd(MAT_GetSeqNonzeroStructure, mat, 0, 0, 0));
7178: PetscFunctionReturn(PETSC_SUCCESS);
7179: }
7181: /*@C
7182: MatDestroySeqNonzeroStructure - Destroys matrix obtained with `MatGetSeqNonzeroStructure()`.
7184: Collective
7186: Input Parameter:
7187: . mat - the matrix (this is a pointer to the array of matrices, just to match the calling
7188: sequence of `MatGetSequentialNonzeroStructure()`)
7190: Level: advanced
7192: Note:
7193: Frees not only the matrices, but also the array that contains the matrices
7195: .seealso: [](chapter_matrices), `Mat`, `MatGetSeqNonzeroStructure()`
7196: @*/
7197: PetscErrorCode MatDestroySeqNonzeroStructure(Mat *mat)
7198: {
7199: PetscFunctionBegin;
7201: PetscCall(MatDestroy(mat));
7202: PetscFunctionReturn(PETSC_SUCCESS);
7203: }
7205: /*@
7206: MatIncreaseOverlap - Given a set of submatrices indicated by index sets,
7207: replaces the index sets by larger ones that represent submatrices with
7208: additional overlap.
7210: Collective
7212: Input Parameters:
7213: + mat - the matrix
7214: . n - the number of index sets
7215: . is - the array of index sets (these index sets will changed during the call)
7216: - ov - the additional overlap requested
7218: Options Database Key:
7219: . -mat_increase_overlap_scalable - use a scalable algorithm to compute the overlap (supported by MPIAIJ matrix)
7221: Level: developer
7223: Note:
7224: The computed overlap preserves the matrix block sizes when the blocks are square.
7225: That is: if a matrix nonzero for a given block would increase the overlap all columns associated with
7226: that block are included in the overlap regardless of whether each specific column would increase the overlap.
7228: .seealso: [](chapter_matrices), `Mat`, `PCASM`, `MatSetBlockSize()`, `MatIncreaseOverlapSplit()`, `MatCreateSubMatrices()`
7229: @*/
7230: PetscErrorCode MatIncreaseOverlap(Mat mat, PetscInt n, IS is[], PetscInt ov)
7231: {
7232: PetscInt i, bs, cbs;
7234: PetscFunctionBegin;
7238: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Must have one or more domains, you have %" PetscInt_FMT, n);
7239: if (n) {
7242: }
7243: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7244: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7245: MatCheckPreallocated(mat, 1);
7247: if (!ov || !n) PetscFunctionReturn(PETSC_SUCCESS);
7248: PetscCall(PetscLogEventBegin(MAT_IncreaseOverlap, mat, 0, 0, 0));
7249: PetscUseTypeMethod(mat, increaseoverlap, n, is, ov);
7250: PetscCall(PetscLogEventEnd(MAT_IncreaseOverlap, mat, 0, 0, 0));
7251: PetscCall(MatGetBlockSizes(mat, &bs, &cbs));
7252: if (bs == cbs) {
7253: for (i = 0; i < n; i++) PetscCall(ISSetBlockSize(is[i], bs));
7254: }
7255: PetscFunctionReturn(PETSC_SUCCESS);
7256: }
7258: PetscErrorCode MatIncreaseOverlapSplit_Single(Mat, IS *, PetscInt);
7260: /*@
7261: MatIncreaseOverlapSplit - Given a set of submatrices indicated by index sets across
7262: a sub communicator, replaces the index sets by larger ones that represent submatrices with
7263: additional overlap.
7265: Collective
7267: Input Parameters:
7268: + mat - the matrix
7269: . n - the number of index sets
7270: . is - the array of index sets (these index sets will changed during the call)
7271: - ov - the additional overlap requested
7273: ` Options Database Key:
7274: . -mat_increase_overlap_scalable - use a scalable algorithm to compute the overlap (supported by MPIAIJ matrix)
7276: Level: developer
7278: .seealso: [](chapter_matrices), `Mat`, `MatCreateSubMatrices()`, `MatIncreaseOverlap()`
7279: @*/
7280: PetscErrorCode MatIncreaseOverlapSplit(Mat mat, PetscInt n, IS is[], PetscInt ov)
7281: {
7282: PetscInt i;
7284: PetscFunctionBegin;
7287: PetscCheck(n >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Must have one or more domains, you have %" PetscInt_FMT, n);
7288: if (n) {
7291: }
7292: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
7293: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
7294: MatCheckPreallocated(mat, 1);
7295: if (!ov) PetscFunctionReturn(PETSC_SUCCESS);
7296: PetscCall(PetscLogEventBegin(MAT_IncreaseOverlap, mat, 0, 0, 0));
7297: for (i = 0; i < n; i++) PetscCall(MatIncreaseOverlapSplit_Single(mat, &is[i], ov));
7298: PetscCall(PetscLogEventEnd(MAT_IncreaseOverlap, mat, 0, 0, 0));
7299: PetscFunctionReturn(PETSC_SUCCESS);
7300: }
7302: /*@
7303: MatGetBlockSize - Returns the matrix block size.
7305: Not Collective
7307: Input Parameter:
7308: . mat - the matrix
7310: Output Parameter:
7311: . bs - block size
7313: Level: intermediate
7315: Notes:
7316: Block row formats are `MATBAIJ` and `MATSBAIJ` ALWAYS have square block storage in the matrix.
7318: If the block size has not been set yet this routine returns 1.
7320: .seealso: [](chapter_matrices), `Mat`, `MATBAIJ`, `MATSBAIJ`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSizes()`
7321: @*/
7322: PetscErrorCode MatGetBlockSize(Mat mat, PetscInt *bs)
7323: {
7324: PetscFunctionBegin;
7327: *bs = PetscAbs(mat->rmap->bs);
7328: PetscFunctionReturn(PETSC_SUCCESS);
7329: }
7331: /*@
7332: MatGetBlockSizes - Returns the matrix block row and column sizes.
7334: Not Collective
7336: Input Parameter:
7337: . mat - the matrix
7339: Output Parameters:
7340: + rbs - row block size
7341: - cbs - column block size
7343: Level: intermediate
7345: Notes:
7346: Block row formats are `MATBAIJ` and `MATSBAIJ` ALWAYS have square block storage in the matrix.
7347: If you pass a different block size for the columns than the rows, the row block size determines the square block storage.
7349: If a block size has not been set yet this routine returns 1.
7351: .seealso: [](chapter_matrices), `Mat`, `MATBAIJ`, `MATSBAIJ`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSize()`, `MatSetBlockSizes()`
7352: @*/
7353: PetscErrorCode MatGetBlockSizes(Mat mat, PetscInt *rbs, PetscInt *cbs)
7354: {
7355: PetscFunctionBegin;
7359: if (rbs) *rbs = PetscAbs(mat->rmap->bs);
7360: if (cbs) *cbs = PetscAbs(mat->cmap->bs);
7361: PetscFunctionReturn(PETSC_SUCCESS);
7362: }
7364: /*@
7365: MatSetBlockSize - Sets the matrix block size.
7367: Logically Collective
7369: Input Parameters:
7370: + mat - the matrix
7371: - bs - block size
7373: Level: intermediate
7375: Notes:
7376: Block row formats are `MATBAIJ` and `MATSBAIJ` formats ALWAYS have square block storage in the matrix.
7377: This must be called before `MatSetUp()` or MatXXXSetPreallocation() (or will default to 1) and the block size cannot be changed later.
7379: For `MATAIJ` matrix format, this function can be called at a later stage, provided that the specified block size
7380: is compatible with the matrix local sizes.
7382: .seealso: [](chapter_matrices), `Mat`, `MATBAIJ`, `MATSBAIJ`, `MATAIJ`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`, `MatGetBlockSizes()`
7383: @*/
7384: PetscErrorCode MatSetBlockSize(Mat mat, PetscInt bs)
7385: {
7386: PetscFunctionBegin;
7389: PetscCall(MatSetBlockSizes(mat, bs, bs));
7390: PetscFunctionReturn(PETSC_SUCCESS);
7391: }
7393: typedef struct {
7394: PetscInt n;
7395: IS *is;
7396: Mat *mat;
7397: PetscObjectState nonzerostate;
7398: Mat C;
7399: } EnvelopeData;
7401: static PetscErrorCode EnvelopeDataDestroy(EnvelopeData *edata)
7402: {
7403: for (PetscInt i = 0; i < edata->n; i++) PetscCall(ISDestroy(&edata->is[i]));
7404: PetscCall(PetscFree(edata->is));
7405: PetscCall(PetscFree(edata));
7406: return PETSC_SUCCESS;
7407: }
7409: /*
7410: MatComputeVariableBlockEnvelope - Given a matrix whose nonzeros are in blocks along the diagonal this computes and stores
7411: the sizes of these blocks in the matrix. An individual block may lie over several processes.
7413: Collective
7415: Input Parameter:
7416: . mat - the matrix
7418: Notes:
7419: There can be zeros within the blocks
7421: The blocks can overlap between processes, including laying on more than two processes
7423: .seealso: [](chapter_matrices), `Mat`, `MatInvertVariableBlockEnvelope()`, `MatSetVariableBlockSizes()`
7424: */
7425: static PetscErrorCode MatComputeVariableBlockEnvelope(Mat mat)
7426: {
7427: PetscInt n, *sizes, *starts, i = 0, env = 0, tbs = 0, lblocks = 0, rstart, II, ln = 0, cnt = 0, cstart, cend;
7428: PetscInt *diag, *odiag, sc;
7429: VecScatter scatter;
7430: PetscScalar *seqv;
7431: const PetscScalar *parv;
7432: const PetscInt *ia, *ja;
7433: PetscBool set, flag, done;
7434: Mat AA = mat, A;
7435: MPI_Comm comm;
7436: PetscMPIInt rank, size, tag;
7437: MPI_Status status;
7438: PetscContainer container;
7439: EnvelopeData *edata;
7440: Vec seq, par;
7441: IS isglobal;
7443: PetscFunctionBegin;
7445: PetscCall(MatIsSymmetricKnown(mat, &set, &flag));
7446: if (!set || !flag) {
7447: /* TOO: only needs nonzero structure of transpose */
7448: PetscCall(MatTranspose(mat, MAT_INITIAL_MATRIX, &AA));
7449: PetscCall(MatAXPY(AA, 1.0, mat, DIFFERENT_NONZERO_PATTERN));
7450: }
7451: PetscCall(MatAIJGetLocalMat(AA, &A));
7452: PetscCall(MatGetRowIJ(A, 0, PETSC_FALSE, PETSC_FALSE, &n, &ia, &ja, &done));
7453: PetscCheck(done, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Unable to get IJ structure from matrix");
7455: PetscCall(MatGetLocalSize(mat, &n, NULL));
7456: PetscCall(PetscObjectGetNewTag((PetscObject)mat, &tag));
7457: PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
7458: PetscCallMPI(MPI_Comm_size(comm, &size));
7459: PetscCallMPI(MPI_Comm_rank(comm, &rank));
7461: PetscCall(PetscMalloc2(n, &sizes, n, &starts));
7463: if (rank > 0) {
7464: PetscCallMPI(MPI_Recv(&env, 1, MPIU_INT, rank - 1, tag, comm, &status));
7465: PetscCallMPI(MPI_Recv(&tbs, 1, MPIU_INT, rank - 1, tag, comm, &status));
7466: }
7467: PetscCall(MatGetOwnershipRange(mat, &rstart, NULL));
7468: for (i = 0; i < n; i++) {
7469: env = PetscMax(env, ja[ia[i + 1] - 1]);
7470: II = rstart + i;
7471: if (env == II) {
7472: starts[lblocks] = tbs;
7473: sizes[lblocks++] = 1 + II - tbs;
7474: tbs = 1 + II;
7475: }
7476: }
7477: if (rank < size - 1) {
7478: PetscCallMPI(MPI_Send(&env, 1, MPIU_INT, rank + 1, tag, comm));
7479: PetscCallMPI(MPI_Send(&tbs, 1, MPIU_INT, rank + 1, tag, comm));
7480: }
7482: PetscCall(MatRestoreRowIJ(A, 0, PETSC_FALSE, PETSC_FALSE, &n, &ia, &ja, &done));
7483: if (!set || !flag) PetscCall(MatDestroy(&AA));
7484: PetscCall(MatDestroy(&A));
7486: PetscCall(PetscNew(&edata));
7487: PetscCall(MatGetNonzeroState(mat, &edata->nonzerostate));
7488: edata->n = lblocks;
7489: /* create IS needed for extracting blocks from the original matrix */
7490: PetscCall(PetscMalloc1(lblocks, &edata->is));
7491: for (PetscInt i = 0; i < lblocks; i++) PetscCall(ISCreateStride(PETSC_COMM_SELF, sizes[i], starts[i], 1, &edata->is[i]));
7493: /* Create the resulting inverse matrix structure with preallocation information */
7494: PetscCall(MatCreate(PetscObjectComm((PetscObject)mat), &edata->C));
7495: PetscCall(MatSetSizes(edata->C, mat->rmap->n, mat->cmap->n, mat->rmap->N, mat->cmap->N));
7496: PetscCall(MatSetBlockSizesFromMats(edata->C, mat, mat));
7497: PetscCall(MatSetType(edata->C, MATAIJ));
7499: /* Communicate the start and end of each row, from each block to the correct rank */
7500: /* TODO: Use PetscSF instead of VecScatter */
7501: for (PetscInt i = 0; i < lblocks; i++) ln += sizes[i];
7502: PetscCall(VecCreateSeq(PETSC_COMM_SELF, 2 * ln, &seq));
7503: PetscCall(VecGetArrayWrite(seq, &seqv));
7504: for (PetscInt i = 0; i < lblocks; i++) {
7505: for (PetscInt j = 0; j < sizes[i]; j++) {
7506: seqv[cnt] = starts[i];
7507: seqv[cnt + 1] = starts[i] + sizes[i];
7508: cnt += 2;
7509: }
7510: }
7511: PetscCall(VecRestoreArrayWrite(seq, &seqv));
7512: PetscCallMPI(MPI_Scan(&cnt, &sc, 1, MPIU_INT, MPI_SUM, PetscObjectComm((PetscObject)mat)));
7513: sc -= cnt;
7514: PetscCall(VecCreateMPI(PetscObjectComm((PetscObject)mat), 2 * mat->rmap->n, 2 * mat->rmap->N, &par));
7515: PetscCall(ISCreateStride(PETSC_COMM_SELF, cnt, sc, 1, &isglobal));
7516: PetscCall(VecScatterCreate(seq, NULL, par, isglobal, &scatter));
7517: PetscCall(ISDestroy(&isglobal));
7518: PetscCall(VecScatterBegin(scatter, seq, par, INSERT_VALUES, SCATTER_FORWARD));
7519: PetscCall(VecScatterEnd(scatter, seq, par, INSERT_VALUES, SCATTER_FORWARD));
7520: PetscCall(VecScatterDestroy(&scatter));
7521: PetscCall(VecDestroy(&seq));
7522: PetscCall(MatGetOwnershipRangeColumn(mat, &cstart, &cend));
7523: PetscCall(PetscMalloc2(mat->rmap->n, &diag, mat->rmap->n, &odiag));
7524: PetscCall(VecGetArrayRead(par, &parv));
7525: cnt = 0;
7526: PetscCall(MatGetSize(mat, NULL, &n));
7527: for (PetscInt i = 0; i < mat->rmap->n; i++) {
7528: PetscInt start, end, d = 0, od = 0;
7530: start = (PetscInt)PetscRealPart(parv[cnt]);
7531: end = (PetscInt)PetscRealPart(parv[cnt + 1]);
7532: cnt += 2;
7534: if (start < cstart) {
7535: od += cstart - start + n - cend;
7536: d += cend - cstart;
7537: } else if (start < cend) {
7538: od += n - cend;
7539: d += cend - start;
7540: } else od += n - start;
7541: if (end <= cstart) {
7542: od -= cstart - end + n - cend;
7543: d -= cend - cstart;
7544: } else if (end < cend) {
7545: od -= n - cend;
7546: d -= cend - end;
7547: } else od -= n - end;
7549: odiag[i] = od;
7550: diag[i] = d;
7551: }
7552: PetscCall(VecRestoreArrayRead(par, &parv));
7553: PetscCall(VecDestroy(&par));
7554: PetscCall(MatXAIJSetPreallocation(edata->C, mat->rmap->bs, diag, odiag, NULL, NULL));
7555: PetscCall(PetscFree2(diag, odiag));
7556: PetscCall(PetscFree2(sizes, starts));
7558: PetscCall(PetscContainerCreate(PETSC_COMM_SELF, &container));
7559: PetscCall(PetscContainerSetPointer(container, edata));
7560: PetscCall(PetscContainerSetUserDestroy(container, (PetscErrorCode(*)(void *))EnvelopeDataDestroy));
7561: PetscCall(PetscObjectCompose((PetscObject)mat, "EnvelopeData", (PetscObject)container));
7562: PetscCall(PetscObjectDereference((PetscObject)container));
7563: PetscFunctionReturn(PETSC_SUCCESS);
7564: }
7566: /*@
7567: MatInvertVariableBlockEnvelope - set matrix C to be the inverted block diagonal of matrix A
7569: Collective
7571: Input Parameters:
7572: + A - the matrix
7573: - reuse - indicates if the `C` matrix was obtained from a previous call to this routine
7575: Output Parameter:
7576: . C - matrix with inverted block diagonal of `A`
7578: Level: advanced
7580: Note:
7581: For efficiency the matrix `A` should have all the nonzero entries clustered in smallish blocks along the diagonal.
7583: .seealso: [](chapter_matrices), `Mat`, `MatInvertBlockDiagonal()`, `MatComputeBlockDiagonal()`
7584: @*/
7585: PetscErrorCode MatInvertVariableBlockEnvelope(Mat A, MatReuse reuse, Mat *C)
7586: {
7587: PetscContainer container;
7588: EnvelopeData *edata;
7589: PetscObjectState nonzerostate;
7591: PetscFunctionBegin;
7592: PetscCall(PetscObjectQuery((PetscObject)A, "EnvelopeData", (PetscObject *)&container));
7593: if (!container) {
7594: PetscCall(MatComputeVariableBlockEnvelope(A));
7595: PetscCall(PetscObjectQuery((PetscObject)A, "EnvelopeData", (PetscObject *)&container));
7596: }
7597: PetscCall(PetscContainerGetPointer(container, (void **)&edata));
7598: PetscCall(MatGetNonzeroState(A, &nonzerostate));
7599: PetscCheck(nonzerostate <= edata->nonzerostate, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Cannot handle changes to matrix nonzero structure");
7600: PetscCheck(reuse != MAT_REUSE_MATRIX || *C == edata->C, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "C matrix must be the same as previously output");
7602: PetscCall(MatCreateSubMatrices(A, edata->n, edata->is, edata->is, MAT_INITIAL_MATRIX, &edata->mat));
7603: *C = edata->C;
7605: for (PetscInt i = 0; i < edata->n; i++) {
7606: Mat D;
7607: PetscScalar *dvalues;
7609: PetscCall(MatConvert(edata->mat[i], MATSEQDENSE, MAT_INITIAL_MATRIX, &D));
7610: PetscCall(MatSetOption(*C, MAT_ROW_ORIENTED, PETSC_FALSE));
7611: PetscCall(MatSeqDenseInvert(D));
7612: PetscCall(MatDenseGetArray(D, &dvalues));
7613: PetscCall(MatSetValuesIS(*C, edata->is[i], edata->is[i], dvalues, INSERT_VALUES));
7614: PetscCall(MatDestroy(&D));
7615: }
7616: PetscCall(MatDestroySubMatrices(edata->n, &edata->mat));
7617: PetscCall(MatAssemblyBegin(*C, MAT_FINAL_ASSEMBLY));
7618: PetscCall(MatAssemblyEnd(*C, MAT_FINAL_ASSEMBLY));
7619: PetscFunctionReturn(PETSC_SUCCESS);
7620: }
7622: /*@
7623: MatSetVariableBlockSizes - Sets diagonal point-blocks of the matrix that need not be of the same size
7625: Logically Collective
7627: Input Parameters:
7628: + mat - the matrix
7629: . nblocks - the number of blocks on this process, each block can only exist on a single process
7630: - bsizes - the block sizes
7632: Level: intermediate
7634: Notes:
7635: Currently used by `PCVPBJACOBI` for `MATAIJ` matrices
7637: Each variable point-block set of degrees of freedom must live on a single MPI rank. That is a point block cannot straddle two MPI ranks.
7639: .seealso: [](chapter_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`, `MatGetBlockSizes()`, `MatGetVariableBlockSizes()`,
7640: `MatComputeVariableBlockEnvelope()`, `PCVPBJACOBI`
7641: @*/
7642: PetscErrorCode MatSetVariableBlockSizes(Mat mat, PetscInt nblocks, PetscInt *bsizes)
7643: {
7644: PetscInt i, ncnt = 0, nlocal;
7646: PetscFunctionBegin;
7648: PetscCheck(nblocks >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Number of local blocks must be great than or equal to zero");
7649: PetscCall(MatGetLocalSize(mat, &nlocal, NULL));
7650: for (i = 0; i < nblocks; i++) ncnt += bsizes[i];
7651: 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);
7652: PetscCall(PetscFree(mat->bsizes));
7653: mat->nblocks = nblocks;
7654: PetscCall(PetscMalloc1(nblocks, &mat->bsizes));
7655: PetscCall(PetscArraycpy(mat->bsizes, bsizes, nblocks));
7656: PetscFunctionReturn(PETSC_SUCCESS);
7657: }
7659: /*@C
7660: MatGetVariableBlockSizes - Gets a diagonal blocks of the matrix that need not be of the same size
7662: Logically Collective; No Fortran Support
7664: Input Parameter:
7665: . mat - the matrix
7667: Output Parameters:
7668: + nblocks - the number of blocks on this process
7669: - bsizes - the block sizes
7671: Level: intermediate
7673: .seealso: [](chapter_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`, `MatGetBlockSizes()`, `MatSetVariableBlockSizes()`, `MatComputeVariableBlockEnvelope()`
7674: @*/
7675: PetscErrorCode MatGetVariableBlockSizes(Mat mat, PetscInt *nblocks, const PetscInt **bsizes)
7676: {
7677: PetscFunctionBegin;
7679: *nblocks = mat->nblocks;
7680: *bsizes = mat->bsizes;
7681: PetscFunctionReturn(PETSC_SUCCESS);
7682: }
7684: /*@
7685: MatSetBlockSizes - Sets the matrix block row and column sizes.
7687: Logically Collective
7689: Input Parameters:
7690: + mat - the matrix
7691: . rbs - row block size
7692: - cbs - column block size
7694: Level: intermediate
7696: Notes:
7697: Block row formats are `MATBAIJ` and `MATSBAIJ`. These formats ALWAYS have square block storage in the matrix.
7698: If you pass a different block size for the columns than the rows, the row block size determines the square block storage.
7699: This must be called before `MatSetUp()` or MatXXXSetPreallocation() (or will default to 1) and the block size cannot be changed later.
7701: For `MATAIJ` matrix this function can be called at a later stage, provided that the specified block sizes
7702: are compatible with the matrix local sizes.
7704: The row and column block size determine the blocksize of the "row" and "column" vectors returned by `MatCreateVecs()`.
7706: .seealso: [](chapter_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSize()`, `MatGetBlockSizes()`
7707: @*/
7708: PetscErrorCode MatSetBlockSizes(Mat mat, PetscInt rbs, PetscInt cbs)
7709: {
7710: PetscFunctionBegin;
7714: PetscTryTypeMethod(mat, setblocksizes, rbs, cbs);
7715: if (mat->rmap->refcnt) {
7716: ISLocalToGlobalMapping l2g = NULL;
7717: PetscLayout nmap = NULL;
7719: PetscCall(PetscLayoutDuplicate(mat->rmap, &nmap));
7720: if (mat->rmap->mapping) PetscCall(ISLocalToGlobalMappingDuplicate(mat->rmap->mapping, &l2g));
7721: PetscCall(PetscLayoutDestroy(&mat->rmap));
7722: mat->rmap = nmap;
7723: mat->rmap->mapping = l2g;
7724: }
7725: if (mat->cmap->refcnt) {
7726: ISLocalToGlobalMapping l2g = NULL;
7727: PetscLayout nmap = NULL;
7729: PetscCall(PetscLayoutDuplicate(mat->cmap, &nmap));
7730: if (mat->cmap->mapping) PetscCall(ISLocalToGlobalMappingDuplicate(mat->cmap->mapping, &l2g));
7731: PetscCall(PetscLayoutDestroy(&mat->cmap));
7732: mat->cmap = nmap;
7733: mat->cmap->mapping = l2g;
7734: }
7735: PetscCall(PetscLayoutSetBlockSize(mat->rmap, rbs));
7736: PetscCall(PetscLayoutSetBlockSize(mat->cmap, cbs));
7737: PetscFunctionReturn(PETSC_SUCCESS);
7738: }
7740: /*@
7741: MatSetBlockSizesFromMats - Sets the matrix block row and column sizes to match a pair of matrices
7743: Logically Collective
7745: Input Parameters:
7746: + mat - the matrix
7747: . fromRow - matrix from which to copy row block size
7748: - fromCol - matrix from which to copy column block size (can be same as fromRow)
7750: Level: developer
7752: .seealso: [](chapter_matrices), `Mat`, `MatCreateSeqBAIJ()`, `MatCreateBAIJ()`, `MatGetBlockSize()`, `MatSetBlockSizes()`
7753: @*/
7754: PetscErrorCode MatSetBlockSizesFromMats(Mat mat, Mat fromRow, Mat fromCol)
7755: {
7756: PetscFunctionBegin;
7760: if (fromRow->rmap->bs > 0) PetscCall(PetscLayoutSetBlockSize(mat->rmap, fromRow->rmap->bs));
7761: if (fromCol->cmap->bs > 0) PetscCall(PetscLayoutSetBlockSize(mat->cmap, fromCol->cmap->bs));
7762: PetscFunctionReturn(PETSC_SUCCESS);
7763: }
7765: /*@
7766: MatResidual - Default routine to calculate the residual r = b - Ax
7768: Collective
7770: Input Parameters:
7771: + mat - the matrix
7772: . b - the right-hand-side
7773: - x - the approximate solution
7775: Output Parameter:
7776: . r - location to store the residual
7778: Level: developer
7780: .seealso: [](chapter_matrices), `Mat`, `MatMult()`, `MatMultAdd()`, `PCMGSetResidual()`
7781: @*/
7782: PetscErrorCode MatResidual(Mat mat, Vec b, Vec x, Vec r)
7783: {
7784: PetscFunctionBegin;
7790: MatCheckPreallocated(mat, 1);
7791: PetscCall(PetscLogEventBegin(MAT_Residual, mat, 0, 0, 0));
7792: if (!mat->ops->residual) {
7793: PetscCall(MatMult(mat, x, r));
7794: PetscCall(VecAYPX(r, -1.0, b));
7795: } else {
7796: PetscUseTypeMethod(mat, residual, b, x, r);
7797: }
7798: PetscCall(PetscLogEventEnd(MAT_Residual, mat, 0, 0, 0));
7799: PetscFunctionReturn(PETSC_SUCCESS);
7800: }
7802: /*MC
7803: MatGetRowIJF90 - Obtains the compressed row storage i and j indices for the local rows of a sparse matrix
7805: Synopsis:
7806: MatGetRowIJF90(Mat A, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt n, {PetscInt, pointer :: ia(:)}, {PetscInt, pointer :: ja(:)}, PetscBool done,integer ierr)
7808: Not Collective
7810: Input Parameters:
7811: + A - the matrix
7812: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
7813: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
7814: - inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
7815: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
7816: always used.
7818: Output Parameters:
7819: + n - number of local rows in the (possibly compressed) matrix
7820: . ia - the row pointers; that is ia[0] = 0, ia[row] = ia[row-1] + number of elements in that row of the matrix
7821: . ja - the column indices
7822: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
7823: are responsible for handling the case when done == `PETSC_FALSE` and ia and ja are not set
7825: Level: developer
7827: Note:
7828: Use `MatRestoreRowIJF90()` when you no longer need access to the data
7830: .seealso: [](chapter_matrices), [](sec_fortranarrays), `Mat`, `MATMPIAIJ`, `MatGetRowIJ()`, `MatRestoreRowIJ()`, `MatRestoreRowIJF90()`
7831: M*/
7833: /*MC
7834: MatRestoreRowIJF90 - restores the compressed row storage i and j indices for the local rows of a sparse matrix obtained with `MatGetRowIJF90()`
7836: Synopsis:
7837: MatRestoreRowIJF90(Mat A, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt n, {PetscInt, pointer :: ia(:)}, {PetscInt, pointer :: ja(:)}, PetscBool done,integer ierr)
7839: Not Collective
7841: Input Parameters:
7842: + A - the matrix
7843: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
7844: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
7845: inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
7846: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
7847: always used.
7848: . n - number of local rows in the (possibly compressed) matrix
7849: . ia - the row pointers; that is ia[0] = 0, ia[row] = ia[row-1] + number of elements in that row of the matrix
7850: . ja - the column indices
7851: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
7852: are responsible for handling the case when done == `PETSC_FALSE` and ia and ja are not set
7854: Level: developer
7856: .seealso: [](chapter_matrices), [](sec_fortranarrays), `Mat`, `MATMPIAIJ`, `MatGetRowIJ()`, `MatRestoreRowIJ()`, `MatGetRowIJF90()`
7857: M*/
7859: /*@C
7860: MatGetRowIJ - Returns the compressed row storage i and j indices for the local rows of a sparse matrix
7862: Collective
7864: Input Parameters:
7865: + mat - the matrix
7866: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
7867: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
7868: - inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
7869: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
7870: always used.
7872: Output Parameters:
7873: + n - number of local rows in the (possibly compressed) matrix, use `NULL` if not needed
7874: . 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
7875: . ja - the column indices, use `NULL` if not needed
7876: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
7877: are responsible for handling the case when done == `PETSC_FALSE` and ia and ja are not set
7879: Level: developer
7881: Notes:
7882: You CANNOT change any of the ia[] or ja[] values.
7884: Use `MatRestoreRowIJ()` when you are finished accessing the ia[] and ja[] values.
7886: Fortran Notes:
7887: Use
7888: .vb
7889: PetscInt, pointer :: ia(:),ja(:)
7890: call MatGetRowIJF90(mat,shift,symmetric,inodecompressed,n,ia,ja,done,ierr)
7891: ! Access the ith and jth entries via ia(i) and ja(j)
7892: .ve
7893: `MatGetRowIJ()` Fortran binding is deprecated (since PETSc 3.19), use `MatGetRowIJF90()`
7895: .seealso: [](chapter_matrices), `Mat`, `MATAIJ`, `MatGetRowIJF90()`, `MatGetColumnIJ()`, `MatRestoreRowIJ()`, `MatSeqAIJGetArray()`
7896: @*/
7897: PetscErrorCode MatGetRowIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
7898: {
7899: PetscFunctionBegin;
7906: MatCheckPreallocated(mat, 1);
7907: if (!mat->ops->getrowij && done) *done = PETSC_FALSE;
7908: else {
7909: if (done) *done = PETSC_TRUE;
7910: PetscCall(PetscLogEventBegin(MAT_GetRowIJ, mat, 0, 0, 0));
7911: PetscUseTypeMethod(mat, getrowij, shift, symmetric, inodecompressed, n, ia, ja, done);
7912: PetscCall(PetscLogEventEnd(MAT_GetRowIJ, mat, 0, 0, 0));
7913: }
7914: PetscFunctionReturn(PETSC_SUCCESS);
7915: }
7917: /*@C
7918: MatGetColumnIJ - Returns the compressed column storage i and j indices for sequential matrices.
7920: Collective
7922: Input Parameters:
7923: + mat - the matrix
7924: . shift - 1 or zero indicating we want the indices starting at 0 or 1
7925: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be
7926: symmetrized
7927: . inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
7928: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
7929: always used.
7930: . n - number of columns in the (possibly compressed) matrix
7931: . ia - the column pointers; that is ia[0] = 0, ia[col] = i[col-1] + number of elements in that col of the matrix
7932: - ja - the row indices
7934: Output Parameter:
7935: . done - `PETSC_TRUE` or `PETSC_FALSE`, indicating whether the values have been returned
7937: Level: developer
7939: .seealso: [](chapter_matrices), `Mat`, `MatGetRowIJ()`, `MatRestoreColumnIJ()`
7940: @*/
7941: PetscErrorCode MatGetColumnIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
7942: {
7943: PetscFunctionBegin;
7950: MatCheckPreallocated(mat, 1);
7951: if (!mat->ops->getcolumnij) *done = PETSC_FALSE;
7952: else {
7953: *done = PETSC_TRUE;
7954: PetscUseTypeMethod(mat, getcolumnij, shift, symmetric, inodecompressed, n, ia, ja, done);
7955: }
7956: PetscFunctionReturn(PETSC_SUCCESS);
7957: }
7959: /*@C
7960: MatRestoreRowIJ - Call after you are completed with the ia,ja indices obtained with `MatGetRowIJ()`.
7962: Collective
7964: Input Parameters:
7965: + mat - the matrix
7966: . shift - 1 or zero indicating we want the indices starting at 0 or 1
7967: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
7968: . inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
7969: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
7970: always used.
7971: . n - size of (possibly compressed) matrix
7972: . ia - the row pointers
7973: - ja - the column indices
7975: Output Parameter:
7976: . done - `PETSC_TRUE` or `PETSC_FALSE` indicated that the values have been returned
7978: Level: developer
7980: Note:
7981: This routine zeros out `n`, `ia`, and `ja`. This is to prevent accidental
7982: us of the array after it has been restored. If you pass `NULL`, it will
7983: not zero the pointers. Use of ia or ja after `MatRestoreRowIJ()` is invalid.
7985: Fortran Note:
7986: `MatRestoreRowIJ()` Fortran binding is deprecated (since PETSc 3.19), use `MatRestoreRowIJF90()`
7988: .seealso: [](chapter_matrices), `Mat`, `MatGetRowIJ()`, `MatRestoreRowIJF90()`, `MatRestoreColumnIJ()`
7989: @*/
7990: PetscErrorCode MatRestoreRowIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
7991: {
7992: PetscFunctionBegin;
7998: MatCheckPreallocated(mat, 1);
8000: if (!mat->ops->restorerowij && done) *done = PETSC_FALSE;
8001: else {
8002: if (done) *done = PETSC_TRUE;
8003: PetscUseTypeMethod(mat, restorerowij, shift, symmetric, inodecompressed, n, ia, ja, done);
8004: if (n) *n = 0;
8005: if (ia) *ia = NULL;
8006: if (ja) *ja = NULL;
8007: }
8008: PetscFunctionReturn(PETSC_SUCCESS);
8009: }
8011: /*@C
8012: MatRestoreColumnIJ - Call after you are completed with the ia,ja indices obtained with `MatGetColumnIJ()`.
8014: Collective
8016: Input Parameters:
8017: + mat - the matrix
8018: . shift - 1 or zero indicating we want the indices starting at 0 or 1
8019: . symmetric - `PETSC_TRUE` or `PETSC_FALSE` indicating the matrix data structure should be symmetrized
8020: - inodecompressed - `PETSC_TRUE` or `PETSC_FALSE` indicating if the nonzero structure of the
8021: inodes or the nonzero elements is wanted. For `MATBAIJ` matrices the compressed version is
8022: always used.
8024: Output Parameters:
8025: + n - size of (possibly compressed) matrix
8026: . ia - the column pointers
8027: . ja - the row indices
8028: - done - `PETSC_TRUE` or `PETSC_FALSE` indicated that the values have been returned
8030: Level: developer
8032: .seealso: [](chapter_matrices), `Mat`, `MatGetColumnIJ()`, `MatRestoreRowIJ()`
8033: @*/
8034: PetscErrorCode MatRestoreColumnIJ(Mat mat, PetscInt shift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
8035: {
8036: PetscFunctionBegin;
8042: MatCheckPreallocated(mat, 1);
8044: if (!mat->ops->restorecolumnij) *done = PETSC_FALSE;
8045: else {
8046: *done = PETSC_TRUE;
8047: PetscUseTypeMethod(mat, restorecolumnij, shift, symmetric, inodecompressed, n, ia, ja, done);
8048: if (n) *n = 0;
8049: if (ia) *ia = NULL;
8050: if (ja) *ja = NULL;
8051: }
8052: PetscFunctionReturn(PETSC_SUCCESS);
8053: }
8055: /*@C
8056: MatColoringPatch -Used inside matrix coloring routines that use `MatGetRowIJ()` and/or `MatGetColumnIJ()`.
8058: Collective
8060: Input Parameters:
8061: + mat - the matrix
8062: . ncolors - maximum color value
8063: . n - number of entries in colorarray
8064: - colorarray - array indicating color for each column
8066: Output Parameter:
8067: . iscoloring - coloring generated using colorarray information
8069: Level: developer
8071: .seealso: [](chapter_matrices), `Mat`, `MatGetRowIJ()`, `MatGetColumnIJ()`
8072: @*/
8073: PetscErrorCode MatColoringPatch(Mat mat, PetscInt ncolors, PetscInt n, ISColoringValue colorarray[], ISColoring *iscoloring)
8074: {
8075: PetscFunctionBegin;
8080: MatCheckPreallocated(mat, 1);
8082: if (!mat->ops->coloringpatch) {
8083: PetscCall(ISColoringCreate(PetscObjectComm((PetscObject)mat), ncolors, n, colorarray, PETSC_OWN_POINTER, iscoloring));
8084: } else {
8085: PetscUseTypeMethod(mat, coloringpatch, ncolors, n, colorarray, iscoloring);
8086: }
8087: PetscFunctionReturn(PETSC_SUCCESS);
8088: }
8090: /*@
8091: MatSetUnfactored - Resets a factored matrix to be treated as unfactored.
8093: Logically Collective
8095: Input Parameter:
8096: . mat - the factored matrix to be reset
8098: Level: developer
8100: Notes:
8101: This routine should be used only with factored matrices formed by in-place
8102: factorization via ILU(0) (or by in-place LU factorization for the `MATSEQDENSE`
8103: format). This option can save memory, for example, when solving nonlinear
8104: systems with a matrix-free Newton-Krylov method and a matrix-based, in-place
8105: ILU(0) preconditioner.
8107: One can specify in-place ILU(0) factorization by calling
8108: .vb
8109: PCType(pc,PCILU);
8110: PCFactorSeUseInPlace(pc);
8111: .ve
8112: or by using the options -pc_type ilu -pc_factor_in_place
8114: In-place factorization ILU(0) can also be used as a local
8115: solver for the blocks within the block Jacobi or additive Schwarz
8116: methods (runtime option: -sub_pc_factor_in_place). See Users-Manual: ch_pc
8117: for details on setting local solver options.
8119: Most users should employ the `KSP` interface for linear solvers
8120: instead of working directly with matrix algebra routines such as this.
8121: See, e.g., `KSPCreate()`.
8123: .seealso: [](chapter_matrices), `Mat`, `PCFactorSetUseInPlace()`, `PCFactorGetUseInPlace()`
8124: @*/
8125: PetscErrorCode MatSetUnfactored(Mat mat)
8126: {
8127: PetscFunctionBegin;
8130: MatCheckPreallocated(mat, 1);
8131: mat->factortype = MAT_FACTOR_NONE;
8132: if (!mat->ops->setunfactored) PetscFunctionReturn(PETSC_SUCCESS);
8133: PetscUseTypeMethod(mat, setunfactored);
8134: PetscFunctionReturn(PETSC_SUCCESS);
8135: }
8137: /*MC
8138: MatDenseGetArrayF90 - Accesses a matrix array from Fortran
8140: Synopsis:
8141: MatDenseGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:,:)},integer ierr)
8143: Not Collective
8145: Input Parameter:
8146: . x - matrix
8148: Output Parameters:
8149: + xx_v - the Fortran pointer to the array
8150: - ierr - error code
8152: Example of Usage:
8153: .vb
8154: PetscScalar, pointer xx_v(:,:)
8155: ....
8156: call MatDenseGetArrayF90(x,xx_v,ierr)
8157: a = xx_v(3)
8158: call MatDenseRestoreArrayF90(x,xx_v,ierr)
8159: .ve
8161: Level: advanced
8163: .seealso: [](chapter_matrices), `Mat`, `MatDenseRestoreArrayF90()`, `MatDenseGetArray()`, `MatDenseRestoreArray()`, `MatSeqAIJGetArrayF90()`
8164: M*/
8166: /*MC
8167: MatDenseRestoreArrayF90 - Restores a matrix array that has been
8168: accessed with `MatDenseGetArrayF90()`.
8170: Synopsis:
8171: MatDenseRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:,:)},integer ierr)
8173: Not Collective
8175: Input Parameters:
8176: + x - matrix
8177: - xx_v - the Fortran90 pointer to the array
8179: Output Parameter:
8180: . ierr - error code
8182: Example of Usage:
8183: .vb
8184: PetscScalar, pointer xx_v(:,:)
8185: ....
8186: call MatDenseGetArrayF90(x,xx_v,ierr)
8187: a = xx_v(3)
8188: call MatDenseRestoreArrayF90(x,xx_v,ierr)
8189: .ve
8191: Level: advanced
8193: .seealso: [](chapter_matrices), `Mat`, `MatDenseGetArrayF90()`, `MatDenseGetArray()`, `MatDenseRestoreArray()`, `MatSeqAIJRestoreArrayF90()`
8194: M*/
8196: /*MC
8197: MatSeqAIJGetArrayF90 - Accesses a matrix array from Fortran.
8199: Synopsis:
8200: MatSeqAIJGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
8202: Not Collective
8204: Input Parameter:
8205: . x - matrix
8207: Output Parameters:
8208: + xx_v - the Fortran pointer to the array
8209: - ierr - error code
8211: Example of Usage:
8212: .vb
8213: PetscScalar, pointer xx_v(:)
8214: ....
8215: call MatSeqAIJGetArrayF90(x,xx_v,ierr)
8216: a = xx_v(3)
8217: call MatSeqAIJRestoreArrayF90(x,xx_v,ierr)
8218: .ve
8220: Level: advanced
8222: .seealso: [](chapter_matrices), `Mat`, `MatSeqAIJRestoreArrayF90()`, `MatSeqAIJGetArray()`, `MatSeqAIJRestoreArray()`, `MatDenseGetArrayF90()`
8223: M*/
8225: /*MC
8226: MatSeqAIJRestoreArrayF90 - Restores a matrix array that has been
8227: accessed with `MatSeqAIJGetArrayF90()`.
8229: Synopsis:
8230: MatSeqAIJRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
8232: Not Collective
8234: Input Parameters:
8235: + x - matrix
8236: - xx_v - the Fortran90 pointer to the array
8238: Output Parameter:
8239: . ierr - error code
8241: Example of Usage:
8242: .vb
8243: PetscScalar, pointer xx_v(:)
8244: ....
8245: call MatSeqAIJGetArrayF90(x,xx_v,ierr)
8246: a = xx_v(3)
8247: call MatSeqAIJRestoreArrayF90(x,xx_v,ierr)
8248: .ve
8250: Level: advanced
8252: .seealso: [](chapter_matrices), `Mat`, `MatSeqAIJGetArrayF90()`, `MatSeqAIJGetArray()`, `MatSeqAIJRestoreArray()`, `MatDenseRestoreArrayF90()`
8253: M*/
8255: /*@
8256: MatCreateSubMatrix - Gets a single submatrix on the same number of processors
8257: as the original matrix.
8259: Collective
8261: Input Parameters:
8262: + mat - the original matrix
8263: . isrow - parallel `IS` containing the rows this processor should obtain
8264: . 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.
8265: - cll - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
8267: Output Parameter:
8268: . newmat - the new submatrix, of the same type as the original matrix
8270: Level: advanced
8272: Notes:
8273: The submatrix will be able to be multiplied with vectors using the same layout as `iscol`.
8275: Some matrix types place restrictions on the row and column indices, such
8276: 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;
8277: for example, if the block size is 3 one cannot select the 0 and 2 rows without selecting the 1 row.
8279: The index sets may not have duplicate entries.
8281: The first time this is called you should use a cll of `MAT_INITIAL_MATRIX`,
8282: the `MatCreateSubMatrix()` routine will create the newmat for you. Any additional calls
8283: to this routine with a mat of the same nonzero structure and with a call of `MAT_REUSE_MATRIX`
8284: will reuse the matrix generated the first time. You should call `MatDestroy()` on `newmat` when
8285: you are finished using it.
8287: The communicator of the newly obtained matrix is ALWAYS the same as the communicator of
8288: the input matrix.
8290: If `iscol` is `NULL` then all columns are obtained (not supported in Fortran).
8292: Example usage:
8293: Consider the following 8x8 matrix with 34 non-zero values, that is
8294: assembled across 3 processors. Let's assume that proc0 owns 3 rows,
8295: proc1 owns 3 rows, proc2 owns 2 rows. This division can be shown
8296: as follows
8297: .vb
8298: 1 2 0 | 0 3 0 | 0 4
8299: Proc0 0 5 6 | 7 0 0 | 8 0
8300: 9 0 10 | 11 0 0 | 12 0
8301: -------------------------------------
8302: 13 0 14 | 15 16 17 | 0 0
8303: Proc1 0 18 0 | 19 20 21 | 0 0
8304: 0 0 0 | 22 23 0 | 24 0
8305: -------------------------------------
8306: Proc2 25 26 27 | 0 0 28 | 29 0
8307: 30 0 0 | 31 32 33 | 0 34
8308: .ve
8310: Suppose `isrow` = [0 1 | 4 | 6 7] and `iscol` = [1 2 | 3 4 5 | 6]. The resulting submatrix is
8312: .vb
8313: 2 0 | 0 3 0 | 0
8314: Proc0 5 6 | 7 0 0 | 8
8315: -------------------------------
8316: Proc1 18 0 | 19 20 21 | 0
8317: -------------------------------
8318: Proc2 26 27 | 0 0 28 | 29
8319: 0 0 | 31 32 33 | 0
8320: .ve
8322: .seealso: [](chapter_matrices), `Mat`, `MatCreateSubMatrices()`, `MatCreateSubMatricesMPI()`, `MatCreateSubMatrixVirtual()`, `MatSubMatrixVirtualUpdate()`
8323: @*/
8324: PetscErrorCode MatCreateSubMatrix(Mat mat, IS isrow, IS iscol, MatReuse cll, Mat *newmat)
8325: {
8326: PetscMPIInt size;
8327: Mat *local;
8328: IS iscoltmp;
8329: PetscBool flg;
8331: PetscFunctionBegin;
8338: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
8339: PetscCheck(cll != MAT_IGNORE_MATRIX, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Cannot use MAT_IGNORE_MATRIX");
8341: MatCheckPreallocated(mat, 1);
8342: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
8344: if (!iscol || isrow == iscol) {
8345: PetscBool stride;
8346: PetscMPIInt grabentirematrix = 0, grab;
8347: PetscCall(PetscObjectTypeCompare((PetscObject)isrow, ISSTRIDE, &stride));
8348: if (stride) {
8349: PetscInt first, step, n, rstart, rend;
8350: PetscCall(ISStrideGetInfo(isrow, &first, &step));
8351: if (step == 1) {
8352: PetscCall(MatGetOwnershipRange(mat, &rstart, &rend));
8353: if (rstart == first) {
8354: PetscCall(ISGetLocalSize(isrow, &n));
8355: if (n == rend - rstart) grabentirematrix = 1;
8356: }
8357: }
8358: }
8359: PetscCall(MPIU_Allreduce(&grabentirematrix, &grab, 1, MPI_INT, MPI_MIN, PetscObjectComm((PetscObject)mat)));
8360: if (grab) {
8361: PetscCall(PetscInfo(mat, "Getting entire matrix as submatrix\n"));
8362: if (cll == MAT_INITIAL_MATRIX) {
8363: *newmat = mat;
8364: PetscCall(PetscObjectReference((PetscObject)mat));
8365: }
8366: PetscFunctionReturn(PETSC_SUCCESS);
8367: }
8368: }
8370: if (!iscol) {
8371: PetscCall(ISCreateStride(PetscObjectComm((PetscObject)mat), mat->cmap->n, mat->cmap->rstart, 1, &iscoltmp));
8372: } else {
8373: iscoltmp = iscol;
8374: }
8376: /* if original matrix is on just one processor then use submatrix generated */
8377: if (mat->ops->createsubmatrices && !mat->ops->createsubmatrix && size == 1 && cll == MAT_REUSE_MATRIX) {
8378: PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscoltmp, MAT_REUSE_MATRIX, &newmat));
8379: goto setproperties;
8380: } else if (mat->ops->createsubmatrices && !mat->ops->createsubmatrix && size == 1) {
8381: PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscoltmp, MAT_INITIAL_MATRIX, &local));
8382: *newmat = *local;
8383: PetscCall(PetscFree(local));
8384: goto setproperties;
8385: } else if (!mat->ops->createsubmatrix) {
8386: /* Create a new matrix type that implements the operation using the full matrix */
8387: PetscCall(PetscLogEventBegin(MAT_CreateSubMat, mat, 0, 0, 0));
8388: switch (cll) {
8389: case MAT_INITIAL_MATRIX:
8390: PetscCall(MatCreateSubMatrixVirtual(mat, isrow, iscoltmp, newmat));
8391: break;
8392: case MAT_REUSE_MATRIX:
8393: PetscCall(MatSubMatrixVirtualUpdate(*newmat, mat, isrow, iscoltmp));
8394: break;
8395: default:
8396: SETERRQ(PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "Invalid MatReuse, must be either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX");
8397: }
8398: PetscCall(PetscLogEventEnd(MAT_CreateSubMat, mat, 0, 0, 0));
8399: goto setproperties;
8400: }
8402: PetscCall(PetscLogEventBegin(MAT_CreateSubMat, mat, 0, 0, 0));
8403: PetscUseTypeMethod(mat, createsubmatrix, isrow, iscoltmp, cll, newmat);
8404: PetscCall(PetscLogEventEnd(MAT_CreateSubMat, mat, 0, 0, 0));
8406: setproperties:
8407: PetscCall(ISEqualUnsorted(isrow, iscoltmp, &flg));
8408: if (flg) PetscCall(MatPropagateSymmetryOptions(mat, *newmat));
8409: if (!iscol) PetscCall(ISDestroy(&iscoltmp));
8410: if (*newmat && cll == MAT_INITIAL_MATRIX) PetscCall(PetscObjectStateIncrease((PetscObject)*newmat));
8411: PetscFunctionReturn(PETSC_SUCCESS);
8412: }
8414: /*@
8415: MatPropagateSymmetryOptions - Propagates symmetry options set on a matrix to another matrix
8417: Not Collective
8419: Input Parameters:
8420: + A - the matrix we wish to propagate options from
8421: - B - the matrix we wish to propagate options to
8423: Level: beginner
8425: Note:
8426: Propagates the options associated to `MAT_SYMMETRY_ETERNAL`, `MAT_STRUCTURALLY_SYMMETRIC`, `MAT_HERMITIAN`, `MAT_SPD`, `MAT_SYMMETRIC`, and `MAT_STRUCTURAL_SYMMETRY_ETERNAL`
8428: .seealso: [](chapter_matrices), `Mat`, `MatSetOption()`, `MatIsSymmetricKnown()`, `MatIsSPDKnown()`, `MatIsHermitianKnown()`, MatIsStructurallySymmetricKnown()`
8429: @*/
8430: PetscErrorCode MatPropagateSymmetryOptions(Mat A, Mat B)
8431: {
8432: PetscFunctionBegin;
8435: B->symmetry_eternal = A->symmetry_eternal;
8436: B->structural_symmetry_eternal = A->structural_symmetry_eternal;
8437: B->symmetric = A->symmetric;
8438: B->structurally_symmetric = A->structurally_symmetric;
8439: B->spd = A->spd;
8440: B->hermitian = A->hermitian;
8441: PetscFunctionReturn(PETSC_SUCCESS);
8442: }
8444: /*@
8445: MatStashSetInitialSize - sets the sizes of the matrix stash, that is
8446: used during the assembly process to store values that belong to
8447: other processors.
8449: Not Collective
8451: Input Parameters:
8452: + mat - the matrix
8453: . size - the initial size of the stash.
8454: - bsize - the initial size of the block-stash(if used).
8456: Options Database Keys:
8457: + -matstash_initial_size <size> or <size0,size1,...sizep-1>
8458: - -matstash_block_initial_size <bsize> or <bsize0,bsize1,...bsizep-1>
8460: Level: intermediate
8462: Notes:
8463: The block-stash is used for values set with `MatSetValuesBlocked()` while
8464: the stash is used for values set with `MatSetValues()`
8466: Run with the option -info and look for output of the form
8467: MatAssemblyBegin_MPIXXX:Stash has MM entries, uses nn mallocs.
8468: to determine the appropriate value, MM, to use for size and
8469: MatAssemblyBegin_MPIXXX:Block-Stash has BMM entries, uses nn mallocs.
8470: to determine the value, BMM to use for bsize
8472: .seealso: [](chapter_matrices), `MatAssemblyBegin()`, `MatAssemblyEnd()`, `Mat`, `MatStashGetInfo()`
8473: @*/
8474: PetscErrorCode MatStashSetInitialSize(Mat mat, PetscInt size, PetscInt bsize)
8475: {
8476: PetscFunctionBegin;
8479: PetscCall(MatStashSetInitialSize_Private(&mat->stash, size));
8480: PetscCall(MatStashSetInitialSize_Private(&mat->bstash, bsize));
8481: PetscFunctionReturn(PETSC_SUCCESS);
8482: }
8484: /*@
8485: MatInterpolateAdd - w = y + A*x or A'*x depending on the shape of
8486: the matrix
8488: Neighbor-wise Collective
8490: Input Parameters:
8491: + mat - the matrix
8492: . x - the vector to be multiplied by the interpolation operator
8493: - y - the vector to be added to the result
8495: Output Parameter:
8496: . w - the resulting vector
8498: Level: intermediate
8500: Notes:
8501: `w` may be the same vector as `y`.
8503: This allows one to use either the restriction or interpolation (its transpose)
8504: matrix to do the interpolation
8506: .seealso: [](chapter_matrices), `Mat`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatRestrict()`, `PCMG`
8507: @*/
8508: PetscErrorCode MatInterpolateAdd(Mat A, Vec x, Vec y, Vec w)
8509: {
8510: PetscInt M, N, Ny;
8512: PetscFunctionBegin;
8517: PetscCall(MatGetSize(A, &M, &N));
8518: PetscCall(VecGetSize(y, &Ny));
8519: if (M == Ny) {
8520: PetscCall(MatMultAdd(A, x, y, w));
8521: } else {
8522: PetscCall(MatMultTransposeAdd(A, x, y, w));
8523: }
8524: PetscFunctionReturn(PETSC_SUCCESS);
8525: }
8527: /*@
8528: MatInterpolate - y = A*x or A'*x depending on the shape of
8529: the matrix
8531: Neighbor-wise Collective
8533: Input Parameters:
8534: + mat - the matrix
8535: - x - the vector to be interpolated
8537: Output Parameter:
8538: . y - the resulting vector
8540: Level: intermediate
8542: Note:
8543: This allows one to use either the restriction or interpolation (its transpose)
8544: matrix to do the interpolation
8546: .seealso: [](chapter_matrices), `Mat`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatRestrict()`, `PCMG`
8547: @*/
8548: PetscErrorCode MatInterpolate(Mat A, Vec x, Vec y)
8549: {
8550: PetscInt M, N, Ny;
8552: PetscFunctionBegin;
8556: PetscCall(MatGetSize(A, &M, &N));
8557: PetscCall(VecGetSize(y, &Ny));
8558: if (M == Ny) {
8559: PetscCall(MatMult(A, x, y));
8560: } else {
8561: PetscCall(MatMultTranspose(A, x, y));
8562: }
8563: PetscFunctionReturn(PETSC_SUCCESS);
8564: }
8566: /*@
8567: MatRestrict - y = A*x or A'*x
8569: Neighbor-wise Collective
8571: Input Parameters:
8572: + mat - the matrix
8573: - x - the vector to be restricted
8575: Output Parameter:
8576: . y - the resulting vector
8578: Level: intermediate
8580: Note:
8581: This allows one to use either the restriction or interpolation (its transpose)
8582: matrix to do the restriction
8584: .seealso: [](chapter_matrices), `Mat`, `MatMultAdd()`, `MatMultTransposeAdd()`, `MatInterpolate()`, `PCMG`
8585: @*/
8586: PetscErrorCode MatRestrict(Mat A, Vec x, Vec y)
8587: {
8588: PetscInt M, N, Ny;
8590: PetscFunctionBegin;
8594: PetscCall(MatGetSize(A, &M, &N));
8595: PetscCall(VecGetSize(y, &Ny));
8596: if (M == Ny) {
8597: PetscCall(MatMult(A, x, y));
8598: } else {
8599: PetscCall(MatMultTranspose(A, x, y));
8600: }
8601: PetscFunctionReturn(PETSC_SUCCESS);
8602: }
8604: /*@
8605: MatMatInterpolateAdd - Y = W + A*X or W + A'*X
8607: Neighbor-wise Collective
8609: Input Parameters:
8610: + mat - the matrix
8611: . x - the input dense matrix to be multiplied
8612: - w - the input dense matrix to be added to the result
8614: Output Parameter:
8615: . y - the output dense matrix
8617: Level: intermediate
8619: Note:
8620: This allows one to use either the restriction or interpolation (its transpose)
8621: matrix to do the interpolation. y matrix can be reused if already created with the proper sizes,
8622: otherwise it will be recreated. y must be initialized to `NULL` if not supplied.
8624: .seealso: [](chapter_matrices), `Mat`, `MatInterpolateAdd()`, `MatMatInterpolate()`, `MatMatRestrict()`, `PCMG`
8625: @*/
8626: PetscErrorCode MatMatInterpolateAdd(Mat A, Mat x, Mat w, Mat *y)
8627: {
8628: PetscInt M, N, Mx, Nx, Mo, My = 0, Ny = 0;
8629: PetscBool trans = PETSC_TRUE;
8630: MatReuse reuse = MAT_INITIAL_MATRIX;
8632: PetscFunctionBegin;
8638: PetscCall(MatGetSize(A, &M, &N));
8639: PetscCall(MatGetSize(x, &Mx, &Nx));
8640: if (N == Mx) trans = PETSC_FALSE;
8641: 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);
8642: Mo = trans ? N : M;
8643: if (*y) {
8644: PetscCall(MatGetSize(*y, &My, &Ny));
8645: if (Mo == My && Nx == Ny) {
8646: reuse = MAT_REUSE_MATRIX;
8647: } else {
8648: 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);
8649: PetscCall(MatDestroy(y));
8650: }
8651: }
8653: if (w && *y == w) { /* this is to minimize changes in PCMG */
8654: PetscBool flg;
8656: PetscCall(PetscObjectQuery((PetscObject)*y, "__MatMatIntAdd_w", (PetscObject *)&w));
8657: if (w) {
8658: PetscInt My, Ny, Mw, Nw;
8660: PetscCall(PetscObjectTypeCompare((PetscObject)*y, ((PetscObject)w)->type_name, &flg));
8661: PetscCall(MatGetSize(*y, &My, &Ny));
8662: PetscCall(MatGetSize(w, &Mw, &Nw));
8663: if (!flg || My != Mw || Ny != Nw) w = NULL;
8664: }
8665: if (!w) {
8666: PetscCall(MatDuplicate(*y, MAT_COPY_VALUES, &w));
8667: PetscCall(PetscObjectCompose((PetscObject)*y, "__MatMatIntAdd_w", (PetscObject)w));
8668: PetscCall(PetscObjectDereference((PetscObject)w));
8669: } else {
8670: PetscCall(MatCopy(*y, w, UNKNOWN_NONZERO_PATTERN));
8671: }
8672: }
8673: if (!trans) {
8674: PetscCall(MatMatMult(A, x, reuse, PETSC_DEFAULT, y));
8675: } else {
8676: PetscCall(MatTransposeMatMult(A, x, reuse, PETSC_DEFAULT, y));
8677: }
8678: if (w) PetscCall(MatAXPY(*y, 1.0, w, UNKNOWN_NONZERO_PATTERN));
8679: PetscFunctionReturn(PETSC_SUCCESS);
8680: }
8682: /*@
8683: MatMatInterpolate - Y = A*X or A'*X
8685: Neighbor-wise Collective
8687: Input Parameters:
8688: + mat - the matrix
8689: - x - the input dense matrix
8691: Output Parameter:
8692: . y - the output dense matrix
8694: Level: intermediate
8696: Note:
8697: This allows one to use either the restriction or interpolation (its transpose)
8698: matrix to do the interpolation. y matrix can be reused if already created with the proper sizes,
8699: otherwise it will be recreated. y must be initialized to `NULL` if not supplied.
8701: .seealso: [](chapter_matrices), `Mat`, `MatInterpolate()`, `MatRestrict()`, `MatMatRestrict()`, `PCMG`
8702: @*/
8703: PetscErrorCode MatMatInterpolate(Mat A, Mat x, Mat *y)
8704: {
8705: PetscFunctionBegin;
8706: PetscCall(MatMatInterpolateAdd(A, x, NULL, y));
8707: PetscFunctionReturn(PETSC_SUCCESS);
8708: }
8710: /*@
8711: MatMatRestrict - Y = A*X or A'*X
8713: Neighbor-wise Collective
8715: Input Parameters:
8716: + mat - the matrix
8717: - x - the input dense matrix
8719: Output Parameter:
8720: . y - the output dense matrix
8722: Level: intermediate
8724: Note:
8725: This allows one to use either the restriction or interpolation (its transpose)
8726: matrix to do the restriction. y matrix can be reused if already created with the proper sizes,
8727: otherwise it will be recreated. y must be initialized to `NULL` if not supplied.
8729: .seealso: [](chapter_matrices), `Mat`, `MatRestrict()`, `MatInterpolate()`, `MatMatInterpolate()`, `PCMG`
8730: @*/
8731: PetscErrorCode MatMatRestrict(Mat A, Mat x, Mat *y)
8732: {
8733: PetscFunctionBegin;
8734: PetscCall(MatMatInterpolateAdd(A, x, NULL, y));
8735: PetscFunctionReturn(PETSC_SUCCESS);
8736: }
8738: /*@
8739: MatGetNullSpace - retrieves the null space of a matrix.
8741: Logically Collective
8743: Input Parameters:
8744: + mat - the matrix
8745: - nullsp - the null space object
8747: Level: developer
8749: .seealso: [](chapter_matrices), `Mat`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatSetNullSpace()`, `MatNullSpace`
8750: @*/
8751: PetscErrorCode MatGetNullSpace(Mat mat, MatNullSpace *nullsp)
8752: {
8753: PetscFunctionBegin;
8756: *nullsp = (mat->symmetric == PETSC_BOOL3_TRUE && !mat->nullsp) ? mat->transnullsp : mat->nullsp;
8757: PetscFunctionReturn(PETSC_SUCCESS);
8758: }
8760: /*@
8761: MatSetNullSpace - attaches a null space to a matrix.
8763: Logically Collective
8765: Input Parameters:
8766: + mat - the matrix
8767: - nullsp - the null space object
8769: Level: advanced
8771: Notes:
8772: This null space is used by the `KSP` linear solvers to solve singular systems.
8774: 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`
8776: 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
8777: to zero but the linear system will still be solved in a least squares sense.
8779: The fundamental theorem of linear algebra (Gilbert Strang, Introduction to Applied Mathematics, page 72) states that
8780: 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).
8781: 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
8782: 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
8783: 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).
8784: This \hat{b} can be obtained by calling MatNullSpaceRemove() with the null space of the transpose of the matrix.
8786: If the matrix is known to be symmetric because it is an `MATSBAIJ` matrix or one as called
8787: `MatSetOption`(mat,`MAT_SYMMETRIC` or possibly `MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`); this
8788: routine also automatically calls `MatSetTransposeNullSpace()`.
8790: The user should call `MatNullSpaceDestroy()`.
8792: .seealso: [](chapter_matrices), `Mat`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatSetTransposeNullSpace()`, `MatGetTransposeNullSpace()`, `MatNullSpaceRemove()`,
8793: `KSPSetPCSide()`
8794: @*/
8795: PetscErrorCode MatSetNullSpace(Mat mat, MatNullSpace nullsp)
8796: {
8797: PetscFunctionBegin;
8800: if (nullsp) PetscCall(PetscObjectReference((PetscObject)nullsp));
8801: PetscCall(MatNullSpaceDestroy(&mat->nullsp));
8802: mat->nullsp = nullsp;
8803: if (mat->symmetric == PETSC_BOOL3_TRUE) PetscCall(MatSetTransposeNullSpace(mat, nullsp));
8804: PetscFunctionReturn(PETSC_SUCCESS);
8805: }
8807: /*@
8808: MatGetTransposeNullSpace - retrieves the null space of the transpose of a matrix.
8810: Logically Collective
8812: Input Parameters:
8813: + mat - the matrix
8814: - nullsp - the null space object
8816: Level: developer
8818: .seealso: [](chapter_matrices), `Mat`, `MatNullSpace`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatSetTransposeNullSpace()`, `MatSetNullSpace()`, `MatGetNullSpace()`
8819: @*/
8820: PetscErrorCode MatGetTransposeNullSpace(Mat mat, MatNullSpace *nullsp)
8821: {
8822: PetscFunctionBegin;
8826: *nullsp = (mat->symmetric == PETSC_BOOL3_TRUE && !mat->transnullsp) ? mat->nullsp : mat->transnullsp;
8827: PetscFunctionReturn(PETSC_SUCCESS);
8828: }
8830: /*@
8831: MatSetTransposeNullSpace - attaches the null space of a transpose of a matrix to the matrix
8833: Logically Collective
8835: Input Parameters:
8836: + mat - the matrix
8837: - nullsp - the null space object
8839: Level: advanced
8841: Notes:
8842: This allows solving singular linear systems defined by the transpose of the matrix using `KSP` solvers with left preconditioning.
8844: See `MatSetNullSpace()`
8846: .seealso: [](chapter_matrices), `Mat`, `MatNullSpace`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatSetNullSpace()`, `MatGetTransposeNullSpace()`, `MatNullSpaceRemove()`, `KSPSetPCSide()`
8847: @*/
8848: PetscErrorCode MatSetTransposeNullSpace(Mat mat, MatNullSpace nullsp)
8849: {
8850: PetscFunctionBegin;
8853: if (nullsp) PetscCall(PetscObjectReference((PetscObject)nullsp));
8854: PetscCall(MatNullSpaceDestroy(&mat->transnullsp));
8855: mat->transnullsp = nullsp;
8856: PetscFunctionReturn(PETSC_SUCCESS);
8857: }
8859: /*@
8860: MatSetNearNullSpace - attaches a null space to a matrix, which is often the null space (rigid body modes) of the operator without boundary conditions
8861: This null space will be used to provide near null space vectors to a multigrid preconditioner built from this matrix.
8863: Logically Collective
8865: Input Parameters:
8866: + mat - the matrix
8867: - nullsp - the null space object
8869: Level: advanced
8871: Notes:
8872: Overwrites any previous near null space that may have been attached
8874: You can remove the null space by calling this routine with an nullsp of `NULL`
8876: .seealso: [](chapter_matrices), `Mat`, `MatNullSpace`, `MatCreate()`, `MatNullSpaceCreate()`, `MatSetNullSpace()`, `MatNullSpaceCreateRigidBody()`, `MatGetNearNullSpace()`
8877: @*/
8878: PetscErrorCode MatSetNearNullSpace(Mat mat, MatNullSpace nullsp)
8879: {
8880: PetscFunctionBegin;
8884: MatCheckPreallocated(mat, 1);
8885: if (nullsp) PetscCall(PetscObjectReference((PetscObject)nullsp));
8886: PetscCall(MatNullSpaceDestroy(&mat->nearnullsp));
8887: mat->nearnullsp = nullsp;
8888: PetscFunctionReturn(PETSC_SUCCESS);
8889: }
8891: /*@
8892: MatGetNearNullSpace - Get null space attached with `MatSetNearNullSpace()`
8894: Not Collective
8896: Input Parameter:
8897: . mat - the matrix
8899: Output Parameter:
8900: . nullsp - the null space object, `NULL` if not set
8902: Level: advanced
8904: .seealso: [](chapter_matrices), `Mat`, `MatNullSpace`, `MatSetNearNullSpace()`, `MatGetNullSpace()`, `MatNullSpaceCreate()`
8905: @*/
8906: PetscErrorCode MatGetNearNullSpace(Mat mat, MatNullSpace *nullsp)
8907: {
8908: PetscFunctionBegin;
8912: MatCheckPreallocated(mat, 1);
8913: *nullsp = mat->nearnullsp;
8914: PetscFunctionReturn(PETSC_SUCCESS);
8915: }
8917: /*@C
8918: MatICCFactor - Performs in-place incomplete Cholesky factorization of matrix.
8920: Collective
8922: Input Parameters:
8923: + mat - the matrix
8924: . row - row/column permutation
8925: - info - information on desired factorization process
8927: Level: developer
8929: Notes:
8930: Probably really in-place only when level of fill is zero, otherwise allocates
8931: new space to store factored matrix and deletes previous memory.
8933: Most users should employ the `KSP` interface for linear solvers
8934: instead of working directly with matrix algebra routines such as this.
8935: See, e.g., `KSPCreate()`.
8937: Developer Note:
8938: The Fortran interface is not autogenerated as the
8939: interface definition cannot be generated correctly [due to `MatFactorInfo`]
8941: .seealso: [](chapter_matrices), `Mat`, `MatFactorInfo`, `MatGetFactor()`, `MatICCFactorSymbolic()`, `MatLUFactorNumeric()`, `MatCholeskyFactor()`
8942: @*/
8943: PetscErrorCode MatICCFactor(Mat mat, IS row, const MatFactorInfo *info)
8944: {
8945: PetscFunctionBegin;
8950: PetscCheck(mat->rmap->N == mat->cmap->N, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONG, "matrix must be square");
8951: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
8952: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
8953: MatCheckPreallocated(mat, 1);
8954: PetscUseTypeMethod(mat, iccfactor, row, info);
8955: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
8956: PetscFunctionReturn(PETSC_SUCCESS);
8957: }
8959: /*@
8960: MatDiagonalScaleLocal - Scales columns of a matrix given the scaling values including the
8961: ghosted ones.
8963: Not Collective
8965: Input Parameters:
8966: + mat - the matrix
8967: - diag - the diagonal values, including ghost ones
8969: Level: developer
8971: Notes:
8972: Works only for `MATMPIAIJ` and `MATMPIBAIJ` matrices
8974: This allows one to avoid during communication to perform the scaling that must be done with `MatDiagonalScale()`
8976: .seealso: [](chapter_matrices), `Mat`, `MatDiagonalScale()`
8977: @*/
8978: PetscErrorCode MatDiagonalScaleLocal(Mat mat, Vec diag)
8979: {
8980: PetscMPIInt size;
8982: PetscFunctionBegin;
8987: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Matrix must be already assembled");
8988: PetscCall(PetscLogEventBegin(MAT_Scale, mat, 0, 0, 0));
8989: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
8990: if (size == 1) {
8991: PetscInt n, m;
8992: PetscCall(VecGetSize(diag, &n));
8993: PetscCall(MatGetSize(mat, NULL, &m));
8994: PetscCheck(m == n, PETSC_COMM_SELF, PETSC_ERR_SUP, "Only supported for sequential matrices when no ghost points/periodic conditions");
8995: PetscCall(MatDiagonalScale(mat, NULL, diag));
8996: } else {
8997: PetscUseMethod(mat, "MatDiagonalScaleLocal_C", (Mat, Vec), (mat, diag));
8998: }
8999: PetscCall(PetscLogEventEnd(MAT_Scale, mat, 0, 0, 0));
9000: PetscCall(PetscObjectStateIncrease((PetscObject)mat));
9001: PetscFunctionReturn(PETSC_SUCCESS);
9002: }
9004: /*@
9005: MatGetInertia - Gets the inertia from a factored matrix
9007: Collective
9009: Input Parameter:
9010: . mat - the matrix
9012: Output Parameters:
9013: + nneg - number of negative eigenvalues
9014: . nzero - number of zero eigenvalues
9015: - npos - number of positive eigenvalues
9017: Level: advanced
9019: Note:
9020: Matrix must have been factored by `MatCholeskyFactor()`
9022: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `MatCholeskyFactor()`
9023: @*/
9024: PetscErrorCode MatGetInertia(Mat mat, PetscInt *nneg, PetscInt *nzero, PetscInt *npos)
9025: {
9026: PetscFunctionBegin;
9029: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
9030: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Numeric factor mat is not assembled");
9031: PetscUseTypeMethod(mat, getinertia, nneg, nzero, npos);
9032: PetscFunctionReturn(PETSC_SUCCESS);
9033: }
9035: /*@C
9036: MatSolves - Solves A x = b, given a factored matrix, for a collection of vectors
9038: Neighbor-wise Collective
9040: Input Parameters:
9041: + mat - the factored matrix obtained with `MatGetFactor()`
9042: - b - the right-hand-side vectors
9044: Output Parameter:
9045: . x - the result vectors
9047: Level: developer
9049: Note:
9050: The vectors `b` and `x` cannot be the same. I.e., one cannot
9051: call `MatSolves`(A,x,x).
9053: .seealso: [](chapter_matrices), `Mat`, `Vecs`, `MatSolveAdd()`, `MatSolveTranspose()`, `MatSolveTransposeAdd()`, `MatSolve()`
9054: @*/
9055: PetscErrorCode MatSolves(Mat mat, Vecs b, Vecs x)
9056: {
9057: PetscFunctionBegin;
9060: PetscCheck(x != b, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_IDN, "x and b must be different vectors");
9061: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Unfactored matrix");
9062: if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(PETSC_SUCCESS);
9064: MatCheckPreallocated(mat, 1);
9065: PetscCall(PetscLogEventBegin(MAT_Solves, mat, 0, 0, 0));
9066: PetscUseTypeMethod(mat, solves, b, x);
9067: PetscCall(PetscLogEventEnd(MAT_Solves, mat, 0, 0, 0));
9068: PetscFunctionReturn(PETSC_SUCCESS);
9069: }
9071: /*@
9072: MatIsSymmetric - Test whether a matrix is symmetric
9074: Collective
9076: Input Parameters:
9077: + A - the matrix to test
9078: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact transpose)
9080: Output Parameter:
9081: . flg - the result
9083: Level: intermediate
9085: Notes:
9086: For real numbers `MatIsSymmetric()` and `MatIsHermitian()` return identical results
9088: If the matrix does not yet know if it is symmetric or not this can be an expensive operation, also available `MatIsSymmetricKnown()`
9090: One can declare that a matrix is symmetric with `MatSetOption`(mat,`MAT_SYMMETRIC`,`PETSC_TRUE`) and if it is known to remain symmetric
9091: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9093: .seealso: [](chapter_matrices), `Mat`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetricKnown()`,
9094: `MAT_SYMMETRIC`, `MAT_SYMMETRY_ETERNAL`, `MatSetOption()`
9095: @*/
9096: PetscErrorCode MatIsSymmetric(Mat A, PetscReal tol, PetscBool *flg)
9097: {
9098: PetscFunctionBegin;
9102: if (A->symmetric == PETSC_BOOL3_TRUE) *flg = PETSC_TRUE;
9103: else if (A->symmetric == PETSC_BOOL3_FALSE) *flg = PETSC_FALSE;
9104: else {
9105: PetscUseTypeMethod(A, issymmetric, tol, flg);
9106: if (!tol) PetscCall(MatSetOption(A, MAT_SYMMETRIC, *flg));
9107: }
9108: PetscFunctionReturn(PETSC_SUCCESS);
9109: }
9111: /*@
9112: MatIsHermitian - Test whether a matrix is Hermitian
9114: Collective
9116: Input Parameters:
9117: + A - the matrix to test
9118: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact Hermitian)
9120: Output Parameter:
9121: . flg - the result
9123: Level: intermediate
9125: Notes:
9126: For real numbers `MatIsSymmetric()` and `MatIsHermitian()` return identical results
9128: If the matrix does not yet know if it is Hermitian or not this can be an expensive operation, also available `MatIsHermitianKnown()`
9130: One can declare that a matrix is Hermitian with `MatSetOption`(mat,`MAT_HERMITIAN`,`PETSC_TRUE`) and if it is known to remain Hermitian
9131: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMEMTRY_ETERNAL`,`PETSC_TRUE`)
9133: .seealso: [](chapter_matrices), `Mat`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitianKnown()`, `MatIsStructurallySymmetric()`, `MatSetOption()`,
9134: `MatIsSymmetricKnown()`, `MatIsSymmetric()`, `MAT_HERMITIAN`, `MAT_SYMMETRY_ETERNAL`, `MatSetOption()`
9135: @*/
9136: PetscErrorCode MatIsHermitian(Mat A, PetscReal tol, PetscBool *flg)
9137: {
9138: PetscFunctionBegin;
9142: if (A->hermitian == PETSC_BOOL3_TRUE) *flg = PETSC_TRUE;
9143: else if (A->hermitian == PETSC_BOOL3_FALSE) *flg = PETSC_FALSE;
9144: else {
9145: PetscUseTypeMethod(A, ishermitian, tol, flg);
9146: if (!tol) PetscCall(MatSetOption(A, MAT_HERMITIAN, *flg));
9147: }
9148: PetscFunctionReturn(PETSC_SUCCESS);
9149: }
9151: /*@
9152: MatIsSymmetricKnown - Checks if a matrix knows if it is symmetric or not and its symmetric state
9154: Not Collective
9156: Input Parameter:
9157: . A - the matrix to check
9159: Output Parameters:
9160: + set - `PETSC_TRUE` if the matrix knows its symmetry state (this tells you if the next flag is valid)
9161: - flg - the result (only valid if set is `PETSC_TRUE`)
9163: Level: advanced
9165: Notes:
9166: Does not check the matrix values directly, so this may return unknown (set = `PETSC_FALSE`). Use `MatIsSymmetric()`
9167: if you want it explicitly checked
9169: One can declare that a matrix is symmetric with `MatSetOption`(mat,`MAT_SYMMETRIC`,`PETSC_TRUE`) and if it is known to remain symmetric
9170: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9172: .seealso: [](chapter_matrices), `Mat`, `MAT_SYMMETRY_ETERNAL`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitianKnown()`
9173: @*/
9174: PetscErrorCode MatIsSymmetricKnown(Mat A, PetscBool *set, PetscBool *flg)
9175: {
9176: PetscFunctionBegin;
9180: if (A->symmetric != PETSC_BOOL3_UNKNOWN) {
9181: *set = PETSC_TRUE;
9182: *flg = PetscBool3ToBool(A->symmetric);
9183: } else {
9184: *set = PETSC_FALSE;
9185: }
9186: PetscFunctionReturn(PETSC_SUCCESS);
9187: }
9189: /*@
9190: MatIsSPDKnown - Checks if a matrix knows if it is symmetric positive definite or not and its symmetric positive definite state
9192: Not Collective
9194: Input Parameter:
9195: . A - the matrix to check
9197: Output Parameters:
9198: + set - `PETSC_TRUE` if the matrix knows its symmetric positive definite state (this tells you if the next flag is valid)
9199: - flg - the result (only valid if set is `PETSC_TRUE`)
9201: Level: advanced
9203: Notes:
9204: Does not check the matrix values directly, so this may return unknown (set = `PETSC_FALSE`).
9206: One can declare that a matrix is SPD with `MatSetOption`(mat,`MAT_SPD`,`PETSC_TRUE`) and if it is known to remain SPD
9207: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SPD_ETERNAL`,`PETSC_TRUE`)
9209: .seealso: [](chapter_matrices), `Mat`, `MAT_SPD_ETERNAL`, `MAT_SPD`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitianKnown()`
9210: @*/
9211: PetscErrorCode MatIsSPDKnown(Mat A, PetscBool *set, PetscBool *flg)
9212: {
9213: PetscFunctionBegin;
9217: if (A->spd != PETSC_BOOL3_UNKNOWN) {
9218: *set = PETSC_TRUE;
9219: *flg = PetscBool3ToBool(A->spd);
9220: } else {
9221: *set = PETSC_FALSE;
9222: }
9223: PetscFunctionReturn(PETSC_SUCCESS);
9224: }
9226: /*@
9227: MatIsHermitianKnown - Checks if a matrix knows if it is Hermitian or not and its Hermitian state
9229: Not Collective
9231: Input Parameter:
9232: . A - the matrix to check
9234: Output Parameters:
9235: + set - `PETSC_TRUE` if the matrix knows its Hermitian state (this tells you if the next flag is valid)
9236: - flg - the result (only valid if set is `PETSC_TRUE`)
9238: Level: advanced
9240: Notes:
9241: Does not check the matrix values directly, so this may return unknown (set = `PETSC_FALSE`). Use `MatIsHermitian()`
9242: if you want it explicitly checked
9244: One can declare that a matrix is Hermitian with `MatSetOption`(mat,`MAT_HERMITIAN`,`PETSC_TRUE`) and if it is known to remain Hermitian
9245: after changes to the matrices values one can call `MatSetOption`(mat,`MAT_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9247: .seealso: [](chapter_matrices), `Mat`, `MAT_SYMMETRY_ETERNAL`, `MAT_HERMITIAN`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`
9248: @*/
9249: PetscErrorCode MatIsHermitianKnown(Mat A, PetscBool *set, PetscBool *flg)
9250: {
9251: PetscFunctionBegin;
9255: if (A->hermitian != PETSC_BOOL3_UNKNOWN) {
9256: *set = PETSC_TRUE;
9257: *flg = PetscBool3ToBool(A->hermitian);
9258: } else {
9259: *set = PETSC_FALSE;
9260: }
9261: PetscFunctionReturn(PETSC_SUCCESS);
9262: }
9264: /*@
9265: MatIsStructurallySymmetric - Test whether a matrix is structurally symmetric
9267: Collective
9269: Input Parameter:
9270: . A - the matrix to test
9272: Output Parameter:
9273: . flg - the result
9275: Level: intermediate
9277: Notes:
9278: If the matrix does yet know it is structurally symmetric this can be an expensive operation, also available `MatIsStructurallySymmetricKnown()`
9280: 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
9281: symmetric after changes to the matrices values one can call `MatSetOption`(mat,`MAT_STRUCTURAL_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9283: .seealso: [](chapter_matrices), `Mat`, `MAT_STRUCTURALLY_SYMMETRIC`, `MAT_STRUCTURAL_SYMMETRY_ETERNAL`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsSymmetric()`, `MatSetOption()`, `MatIsStructurallySymmetricKnown()`
9284: @*/
9285: PetscErrorCode MatIsStructurallySymmetric(Mat A, PetscBool *flg)
9286: {
9287: PetscFunctionBegin;
9290: if (A->structurally_symmetric != PETSC_BOOL3_UNKNOWN) {
9291: *flg = PetscBool3ToBool(A->structurally_symmetric);
9292: } else {
9293: PetscUseTypeMethod(A, isstructurallysymmetric, flg);
9294: PetscCall(MatSetOption(A, MAT_STRUCTURALLY_SYMMETRIC, *flg));
9295: }
9296: PetscFunctionReturn(PETSC_SUCCESS);
9297: }
9299: /*@
9300: MatIsStructurallySymmetricKnown - Checks if a matrix knows if it is structurally symmetric or not and its structurally symmetric state
9302: Not Collective
9304: Input Parameter:
9305: . A - the matrix to check
9307: Output Parameters:
9308: + set - PETSC_TRUE if the matrix knows its structurally symmetric state (this tells you if the next flag is valid)
9309: - flg - the result (only valid if set is PETSC_TRUE)
9311: Level: advanced
9313: Notes:
9314: 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
9315: symmetric after changes to the matrices values one can call `MatSetOption`(mat,`MAT_STRUCTURAL_SYMMETRY_ETERNAL`,`PETSC_TRUE`)
9317: Use `MatIsStructurallySymmetric()` to explicitly check if a matrix is structurally symmetric (this is an expensive operation)
9319: .seealso: [](chapter_matrices), `Mat`, `MAT_STRUCTURALLY_SYMMETRIC`, `MatTranspose()`, `MatIsTranspose()`, `MatIsHermitian()`, `MatIsStructurallySymmetric()`, `MatSetOption()`, `MatIsSymmetric()`, `MatIsHermitianKnown()`
9320: @*/
9321: PetscErrorCode MatIsStructurallySymmetricKnown(Mat A, PetscBool *set, PetscBool *flg)
9322: {
9323: PetscFunctionBegin;
9327: if (A->structurally_symmetric != PETSC_BOOL3_UNKNOWN) {
9328: *set = PETSC_TRUE;
9329: *flg = PetscBool3ToBool(A->structurally_symmetric);
9330: } else {
9331: *set = PETSC_FALSE;
9332: }
9333: PetscFunctionReturn(PETSC_SUCCESS);
9334: }
9336: /*@
9337: MatStashGetInfo - Gets how many values are currently in the matrix stash, i.e. need
9338: to be communicated to other processors during the `MatAssemblyBegin()`/`MatAssemblyEnd()` process
9340: Not Collective
9342: Input Parameter:
9343: . mat - the matrix
9345: Output Parameters:
9346: + nstash - the size of the stash
9347: . reallocs - the number of additional mallocs incurred.
9348: . bnstash - the size of the block stash
9349: - breallocs - the number of additional mallocs incurred.in the block stash
9351: Level: advanced
9353: .seealso: [](chapter_matrices), `MatAssemblyBegin()`, `MatAssemblyEnd()`, `Mat`, `MatStashSetInitialSize()`
9354: @*/
9355: PetscErrorCode MatStashGetInfo(Mat mat, PetscInt *nstash, PetscInt *reallocs, PetscInt *bnstash, PetscInt *breallocs)
9356: {
9357: PetscFunctionBegin;
9358: PetscCall(MatStashGetInfo_Private(&mat->stash, nstash, reallocs));
9359: PetscCall(MatStashGetInfo_Private(&mat->bstash, bnstash, breallocs));
9360: PetscFunctionReturn(PETSC_SUCCESS);
9361: }
9363: /*@C
9364: MatCreateVecs - Get vector(s) compatible with the matrix, i.e. with the same
9365: parallel layout, `PetscLayout` for rows and columns
9367: Collective
9369: Input Parameter:
9370: . mat - the matrix
9372: Output Parameters:
9373: + right - (optional) vector that the matrix can be multiplied against
9374: - left - (optional) vector that the matrix vector product can be stored in
9376: Level: advanced
9378: Notes:
9379: 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()`.
9381: These are new vectors which are not owned by the mat, they should be destroyed in `VecDestroy()` when no longer needed
9383: .seealso: [](chapter_matrices), `Mat`, `Vec`, `VecCreate()`, `VecDestroy()`, `DMCreateGlobalVector()`
9384: @*/
9385: PetscErrorCode MatCreateVecs(Mat mat, Vec *right, Vec *left)
9386: {
9387: PetscFunctionBegin;
9390: if (mat->ops->getvecs) {
9391: PetscUseTypeMethod(mat, getvecs, right, left);
9392: } else {
9393: PetscInt rbs, cbs;
9394: PetscCall(MatGetBlockSizes(mat, &rbs, &cbs));
9395: if (right) {
9396: PetscCheck(mat->cmap->n >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "PetscLayout for columns not yet setup");
9397: PetscCall(VecCreate(PetscObjectComm((PetscObject)mat), right));
9398: PetscCall(VecSetSizes(*right, mat->cmap->n, PETSC_DETERMINE));
9399: PetscCall(VecSetBlockSize(*right, cbs));
9400: PetscCall(VecSetType(*right, mat->defaultvectype));
9401: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
9402: if (mat->boundtocpu && mat->bindingpropagates) {
9403: PetscCall(VecSetBindingPropagates(*right, PETSC_TRUE));
9404: PetscCall(VecBindToCPU(*right, PETSC_TRUE));
9405: }
9406: #endif
9407: PetscCall(PetscLayoutReference(mat->cmap, &(*right)->map));
9408: }
9409: if (left) {
9410: PetscCheck(mat->rmap->n >= 0, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "PetscLayout for rows not yet setup");
9411: PetscCall(VecCreate(PetscObjectComm((PetscObject)mat), left));
9412: PetscCall(VecSetSizes(*left, mat->rmap->n, PETSC_DETERMINE));
9413: PetscCall(VecSetBlockSize(*left, rbs));
9414: PetscCall(VecSetType(*left, mat->defaultvectype));
9415: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
9416: if (mat->boundtocpu && mat->bindingpropagates) {
9417: PetscCall(VecSetBindingPropagates(*left, PETSC_TRUE));
9418: PetscCall(VecBindToCPU(*left, PETSC_TRUE));
9419: }
9420: #endif
9421: PetscCall(PetscLayoutReference(mat->rmap, &(*left)->map));
9422: }
9423: }
9424: PetscFunctionReturn(PETSC_SUCCESS);
9425: }
9427: /*@C
9428: MatFactorInfoInitialize - Initializes a `MatFactorInfo` data structure
9429: with default values.
9431: Not Collective
9433: Input Parameter:
9434: . info - the `MatFactorInfo` data structure
9436: Level: developer
9438: Notes:
9439: The solvers are generally used through the `KSP` and `PC` objects, for example
9440: `PCLU`, `PCILU`, `PCCHOLESKY`, `PCICC`
9442: Once the data structure is initialized one may change certain entries as desired for the particular factorization to be performed
9444: Developer Note:
9445: The Fortran interface is not autogenerated as the
9446: interface definition cannot be generated correctly [due to `MatFactorInfo`]
9448: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `MatFactorInfo`
9449: @*/
9450: PetscErrorCode MatFactorInfoInitialize(MatFactorInfo *info)
9451: {
9452: PetscFunctionBegin;
9453: PetscCall(PetscMemzero(info, sizeof(MatFactorInfo)));
9454: PetscFunctionReturn(PETSC_SUCCESS);
9455: }
9457: /*@
9458: MatFactorSetSchurIS - Set indices corresponding to the Schur complement you wish to have computed
9460: Collective
9462: Input Parameters:
9463: + mat - the factored matrix
9464: - is - the index set defining the Schur indices (0-based)
9466: Level: advanced
9468: Notes:
9469: Call `MatFactorSolveSchurComplement()` or `MatFactorSolveSchurComplementTranspose()` after this call to solve a Schur complement system.
9471: You can call `MatFactorGetSchurComplement()` or `MatFactorCreateSchurComplement()` after this call.
9473: This functionality is only supported for `MATSOLVERMUMPS` and `MATSOLVERMKL_PARDISO`
9475: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `MatFactorGetSchurComplement()`, `MatFactorRestoreSchurComplement()`, `MatFactorCreateSchurComplement()`, `MatFactorSolveSchurComplement()`,
9476: `MatFactorSolveSchurComplementTranspose()`, `MatFactorSolveSchurComplement()`, `MATSOLVERMUMPS`, `MATSOLVERMKL_PARDISO`
9477: @*/
9478: PetscErrorCode MatFactorSetSchurIS(Mat mat, IS is)
9479: {
9480: PetscErrorCode (*f)(Mat, IS);
9482: PetscFunctionBegin;
9487: PetscCheckSameComm(mat, 1, is, 2);
9488: PetscCheck(mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Only for factored matrix");
9489: PetscCall(PetscObjectQueryFunction((PetscObject)mat, "MatFactorSetSchurIS_C", &f));
9490: PetscCheck(f, PetscObjectComm((PetscObject)mat), PETSC_ERR_SUP, "The selected MatSolverType does not support Schur complement computation. You should use MATSOLVERMUMPS or MATSOLVERMKL_PARDISO");
9491: PetscCall(MatDestroy(&mat->schur));
9492: PetscCall((*f)(mat, is));
9493: PetscCheck(mat->schur, PetscObjectComm((PetscObject)mat), PETSC_ERR_PLIB, "Schur complement has not been created");
9494: PetscFunctionReturn(PETSC_SUCCESS);
9495: }
9497: /*@
9498: MatFactorCreateSchurComplement - Create a Schur complement matrix object using Schur data computed during the factorization step
9500: Logically Collective
9502: Input Parameters:
9503: + F - the factored matrix obtained by calling `MatGetFactor()`
9504: . S - location where to return the Schur complement, can be `NULL`
9505: - status - the status of the Schur complement matrix, can be `NULL`
9507: Level: advanced
9509: Notes:
9510: You must call `MatFactorSetSchurIS()` before calling this routine.
9512: This functionality is only supported for `MATSOLVERMUMPS` and `MATSOLVERMKL_PARDISO`
9514: The routine provides a copy of the Schur matrix stored within the solver data structures.
9515: The caller must destroy the object when it is no longer needed.
9516: If `MatFactorInvertSchurComplement()` has been called, the routine gets back the inverse.
9518: 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)
9520: See `MatCreateSchurComplement()` or `MatGetSchurComplement()` for ways to create virtual or approximate Schur complements.
9522: Developer Note:
9523: The reason this routine exists is because the representation of the Schur complement within the factor matrix may be different than a standard PETSc
9524: matrix representation and we normally do not want to use the time or memory to make a copy as a regular PETSc matrix.
9526: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorGetSchurComplement()`, `MatFactorSchurStatus`, `MATSOLVERMUMPS`, `MATSOLVERMKL_PARDISO`
9527: @*/
9528: PetscErrorCode MatFactorCreateSchurComplement(Mat F, Mat *S, MatFactorSchurStatus *status)
9529: {
9530: PetscFunctionBegin;
9534: if (S) {
9535: PetscErrorCode (*f)(Mat, Mat *);
9537: PetscCall(PetscObjectQueryFunction((PetscObject)F, "MatFactorCreateSchurComplement_C", &f));
9538: if (f) {
9539: PetscCall((*f)(F, S));
9540: } else {
9541: PetscCall(MatDuplicate(F->schur, MAT_COPY_VALUES, S));
9542: }
9543: }
9544: if (status) *status = F->schur_status;
9545: PetscFunctionReturn(PETSC_SUCCESS);
9546: }
9548: /*@
9549: MatFactorGetSchurComplement - Gets access to a Schur complement matrix using the current Schur data within a factored matrix
9551: Logically Collective
9553: Input Parameters:
9554: + F - the factored matrix obtained by calling `MatGetFactor()`
9555: . *S - location where to return the Schur complement, can be `NULL`
9556: - status - the status of the Schur complement matrix, can be `NULL`
9558: Level: advanced
9560: Notes:
9561: You must call `MatFactorSetSchurIS()` before calling this routine.
9563: Schur complement mode is currently implemented for sequential matrices with factor type of `MATSOLVERMUMPS`
9565: The routine returns a the Schur Complement stored within the data structures of the solver.
9567: If `MatFactorInvertSchurComplement()` has previously been called, the returned matrix is actually the inverse of the Schur complement.
9569: The returned matrix should not be destroyed; the caller should call `MatFactorRestoreSchurComplement()` when the object is no longer needed.
9571: Use `MatFactorCreateSchurComplement()` to create a copy of the Schur complement matrix that is within a factored matrix
9573: See `MatCreateSchurComplement()` or `MatGetSchurComplement()` for ways to create virtual or approximate Schur complements.
9575: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorRestoreSchurComplement()`, `MatFactorCreateSchurComplement()`, `MatFactorSchurStatus`
9576: @*/
9577: PetscErrorCode MatFactorGetSchurComplement(Mat F, Mat *S, MatFactorSchurStatus *status)
9578: {
9579: PetscFunctionBegin;
9583: if (S) *S = F->schur;
9584: if (status) *status = F->schur_status;
9585: PetscFunctionReturn(PETSC_SUCCESS);
9586: }
9588: static PetscErrorCode MatFactorUpdateSchurStatus_Private(Mat F)
9589: {
9590: Mat S = F->schur;
9592: PetscFunctionBegin;
9593: switch (F->schur_status) {
9594: case MAT_FACTOR_SCHUR_UNFACTORED: // fall-through
9595: case MAT_FACTOR_SCHUR_INVERTED:
9596: if (S) {
9597: S->ops->solve = NULL;
9598: S->ops->matsolve = NULL;
9599: S->ops->solvetranspose = NULL;
9600: S->ops->matsolvetranspose = NULL;
9601: S->ops->solveadd = NULL;
9602: S->ops->solvetransposeadd = NULL;
9603: S->factortype = MAT_FACTOR_NONE;
9604: PetscCall(PetscFree(S->solvertype));
9605: }
9606: case MAT_FACTOR_SCHUR_FACTORED: // fall-through
9607: break;
9608: default:
9609: SETERRQ(PetscObjectComm((PetscObject)F), PETSC_ERR_SUP, "Unhandled MatFactorSchurStatus %d", F->schur_status);
9610: }
9611: PetscFunctionReturn(PETSC_SUCCESS);
9612: }
9614: /*@
9615: MatFactorRestoreSchurComplement - Restore the Schur complement matrix object obtained from a call to `MatFactorGetSchurComplement()`
9617: Logically Collective
9619: Input Parameters:
9620: + F - the factored matrix obtained by calling `MatGetFactor()`
9621: . *S - location where the Schur complement is stored
9622: - status - the status of the Schur complement matrix (see `MatFactorSchurStatus`)
9624: Level: advanced
9626: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorRestoreSchurComplement()`, `MatFactorCreateSchurComplement()`, `MatFactorSchurStatus`
9627: @*/
9628: PetscErrorCode MatFactorRestoreSchurComplement(Mat F, Mat *S, MatFactorSchurStatus status)
9629: {
9630: PetscFunctionBegin;
9632: if (S) {
9634: *S = NULL;
9635: }
9636: F->schur_status = status;
9637: PetscCall(MatFactorUpdateSchurStatus_Private(F));
9638: PetscFunctionReturn(PETSC_SUCCESS);
9639: }
9641: /*@
9642: MatFactorSolveSchurComplementTranspose - Solve the transpose of the Schur complement system computed during the factorization step
9644: Logically Collective
9646: Input Parameters:
9647: + F - the factored matrix obtained by calling `MatGetFactor()`
9648: . rhs - location where the right hand side of the Schur complement system is stored
9649: - sol - location where the solution of the Schur complement system has to be returned
9651: Level: advanced
9653: Notes:
9654: The sizes of the vectors should match the size of the Schur complement
9656: Must be called after `MatFactorSetSchurIS()`
9658: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorSolveSchurComplement()`
9659: @*/
9660: PetscErrorCode MatFactorSolveSchurComplementTranspose(Mat F, Vec rhs, Vec sol)
9661: {
9662: PetscFunctionBegin;
9669: PetscCheckSameComm(F, 1, rhs, 2);
9670: PetscCheckSameComm(F, 1, sol, 3);
9671: PetscCall(MatFactorFactorizeSchurComplement(F));
9672: switch (F->schur_status) {
9673: case MAT_FACTOR_SCHUR_FACTORED:
9674: PetscCall(MatSolveTranspose(F->schur, rhs, sol));
9675: break;
9676: case MAT_FACTOR_SCHUR_INVERTED:
9677: PetscCall(MatMultTranspose(F->schur, rhs, sol));
9678: break;
9679: default:
9680: SETERRQ(PetscObjectComm((PetscObject)F), PETSC_ERR_SUP, "Unhandled MatFactorSchurStatus %d", F->schur_status);
9681: }
9682: PetscFunctionReturn(PETSC_SUCCESS);
9683: }
9685: /*@
9686: MatFactorSolveSchurComplement - Solve the Schur complement system computed during the factorization step
9688: Logically Collective
9690: Input Parameters:
9691: + F - the factored matrix obtained by calling `MatGetFactor()`
9692: . rhs - location where the right hand side of the Schur complement system is stored
9693: - sol - location where the solution of the Schur complement system has to be returned
9695: Level: advanced
9697: Notes:
9698: The sizes of the vectors should match the size of the Schur complement
9700: Must be called after `MatFactorSetSchurIS()`
9702: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorSolveSchurComplementTranspose()`
9703: @*/
9704: PetscErrorCode MatFactorSolveSchurComplement(Mat F, Vec rhs, Vec sol)
9705: {
9706: PetscFunctionBegin;
9713: PetscCheckSameComm(F, 1, rhs, 2);
9714: PetscCheckSameComm(F, 1, sol, 3);
9715: PetscCall(MatFactorFactorizeSchurComplement(F));
9716: switch (F->schur_status) {
9717: case MAT_FACTOR_SCHUR_FACTORED:
9718: PetscCall(MatSolve(F->schur, rhs, sol));
9719: break;
9720: case MAT_FACTOR_SCHUR_INVERTED:
9721: PetscCall(MatMult(F->schur, rhs, sol));
9722: break;
9723: default:
9724: SETERRQ(PetscObjectComm((PetscObject)F), PETSC_ERR_SUP, "Unhandled MatFactorSchurStatus %d", F->schur_status);
9725: }
9726: PetscFunctionReturn(PETSC_SUCCESS);
9727: }
9729: PETSC_EXTERN PetscErrorCode MatSeqDenseInvertFactors_Private(Mat);
9730: #if PetscDefined(HAVE_CUDA)
9731: PETSC_SINGLE_LIBRARY_INTERN PetscErrorCode MatSeqDenseCUDAInvertFactors_Internal(Mat);
9732: #endif
9734: /* Schur status updated in the interface */
9735: static PetscErrorCode MatFactorInvertSchurComplement_Private(Mat F)
9736: {
9737: Mat S = F->schur;
9739: PetscFunctionBegin;
9740: if (S) {
9741: PetscMPIInt size;
9742: PetscBool isdense, isdensecuda;
9744: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)S), &size));
9745: PetscCheck(size <= 1, PetscObjectComm((PetscObject)S), PETSC_ERR_SUP, "Not yet implemented");
9746: PetscCall(PetscObjectTypeCompare((PetscObject)S, MATSEQDENSE, &isdense));
9747: PetscCall(PetscObjectTypeCompare((PetscObject)S, MATSEQDENSECUDA, &isdensecuda));
9748: PetscCheck(isdense || isdensecuda, PetscObjectComm((PetscObject)S), PETSC_ERR_SUP, "Not implemented for type %s", ((PetscObject)S)->type_name);
9749: PetscCall(PetscLogEventBegin(MAT_FactorInvS, F, 0, 0, 0));
9750: if (isdense) {
9751: PetscCall(MatSeqDenseInvertFactors_Private(S));
9752: } else if (isdensecuda) {
9753: #if defined(PETSC_HAVE_CUDA)
9754: PetscCall(MatSeqDenseCUDAInvertFactors_Internal(S));
9755: #endif
9756: }
9757: // HIP??????????????
9758: PetscCall(PetscLogEventEnd(MAT_FactorInvS, F, 0, 0, 0));
9759: }
9760: PetscFunctionReturn(PETSC_SUCCESS);
9761: }
9763: /*@
9764: MatFactorInvertSchurComplement - Invert the Schur complement matrix computed during the factorization step
9766: Logically Collective
9768: Input Parameter:
9769: . F - the factored matrix obtained by calling `MatGetFactor()`
9771: Level: advanced
9773: Notes:
9774: Must be called after `MatFactorSetSchurIS()`.
9776: Call `MatFactorGetSchurComplement()` or `MatFactorCreateSchurComplement()` AFTER this call to actually compute the inverse and get access to it.
9778: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorGetSchurComplement()`, `MatFactorCreateSchurComplement()`
9779: @*/
9780: PetscErrorCode MatFactorInvertSchurComplement(Mat F)
9781: {
9782: PetscFunctionBegin;
9785: if (F->schur_status == MAT_FACTOR_SCHUR_INVERTED) PetscFunctionReturn(PETSC_SUCCESS);
9786: PetscCall(MatFactorFactorizeSchurComplement(F));
9787: PetscCall(MatFactorInvertSchurComplement_Private(F));
9788: F->schur_status = MAT_FACTOR_SCHUR_INVERTED;
9789: PetscFunctionReturn(PETSC_SUCCESS);
9790: }
9792: /*@
9793: MatFactorFactorizeSchurComplement - Factorize the Schur complement matrix computed during the factorization step
9795: Logically Collective
9797: Input Parameter:
9798: . F - the factored matrix obtained by calling `MatGetFactor()`
9800: Level: advanced
9802: Note:
9803: Must be called after `MatFactorSetSchurIS()`
9805: .seealso: [](chapter_matrices), `Mat`, `MatGetFactor()`, `MatFactorSetSchurIS()`, `MatFactorInvertSchurComplement()`
9806: @*/
9807: PetscErrorCode MatFactorFactorizeSchurComplement(Mat F)
9808: {
9809: MatFactorInfo info;
9811: PetscFunctionBegin;
9814: if (F->schur_status == MAT_FACTOR_SCHUR_INVERTED || F->schur_status == MAT_FACTOR_SCHUR_FACTORED) PetscFunctionReturn(PETSC_SUCCESS);
9815: PetscCall(PetscLogEventBegin(MAT_FactorFactS, F, 0, 0, 0));
9816: PetscCall(PetscMemzero(&info, sizeof(MatFactorInfo)));
9817: if (F->factortype == MAT_FACTOR_CHOLESKY) { /* LDL^t regarded as Cholesky */
9818: PetscCall(MatCholeskyFactor(F->schur, NULL, &info));
9819: } else {
9820: PetscCall(MatLUFactor(F->schur, NULL, NULL, &info));
9821: }
9822: PetscCall(PetscLogEventEnd(MAT_FactorFactS, F, 0, 0, 0));
9823: F->schur_status = MAT_FACTOR_SCHUR_FACTORED;
9824: PetscFunctionReturn(PETSC_SUCCESS);
9825: }
9827: /*@
9828: MatPtAP - Creates the matrix product C = P^T * A * P
9830: Neighbor-wise Collective
9832: Input Parameters:
9833: + A - the matrix
9834: . P - the projection matrix
9835: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
9836: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(P)), use `PETSC_DEFAULT` if you do not have a good estimate
9837: if the result is a dense matrix this is irrelevant
9839: Output Parameter:
9840: . C - the product matrix
9842: Level: intermediate
9844: Notes:
9845: C will be created and must be destroyed by the user with `MatDestroy()`.
9847: An alternative approach to this function is to use `MatProductCreate()` and set the desired options before the computation is done
9849: Developer Note:
9850: For matrix types without special implementation the function fallbacks to `MatMatMult()` followed by `MatTransposeMatMult()`.
9852: .seealso: [](chapter_matrices), `Mat`, `MatProductCreate()`, `MatMatMult()`, `MatRARt()`
9853: @*/
9854: PetscErrorCode MatPtAP(Mat A, Mat P, MatReuse scall, PetscReal fill, Mat *C)
9855: {
9856: PetscFunctionBegin;
9857: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*C, 5);
9858: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
9860: if (scall == MAT_INITIAL_MATRIX) {
9861: PetscCall(MatProductCreate(A, P, NULL, C));
9862: PetscCall(MatProductSetType(*C, MATPRODUCT_PtAP));
9863: PetscCall(MatProductSetAlgorithm(*C, "default"));
9864: PetscCall(MatProductSetFill(*C, fill));
9866: (*C)->product->api_user = PETSC_TRUE;
9867: PetscCall(MatProductSetFromOptions(*C));
9868: 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);
9869: PetscCall(MatProductSymbolic(*C));
9870: } else { /* scall == MAT_REUSE_MATRIX */
9871: PetscCall(MatProductReplaceMats(A, P, NULL, *C));
9872: }
9874: PetscCall(MatProductNumeric(*C));
9875: (*C)->symmetric = A->symmetric;
9876: (*C)->spd = A->spd;
9877: PetscFunctionReturn(PETSC_SUCCESS);
9878: }
9880: /*@
9881: MatRARt - Creates the matrix product C = R * A * R^T
9883: Neighbor-wise Collective
9885: Input Parameters:
9886: + A - the matrix
9887: . R - the projection matrix
9888: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
9889: - fill - expected fill as ratio of nnz(C)/nnz(A), use `PETSC_DEFAULT` if you do not have a good estimate
9890: if the result is a dense matrix this is irrelevant
9892: Output Parameter:
9893: . C - the product matrix
9895: Level: intermediate
9897: Notes:
9898: C will be created and must be destroyed by the user with `MatDestroy()`.
9900: An alternative approach to this function is to use `MatProductCreate()` and set the desired options before the computation is done
9902: This routine is currently only implemented for pairs of `MATAIJ` matrices and classes
9903: which inherit from `MATAIJ`. Due to PETSc sparse matrix block row distribution among processes,
9904: parallel MatRARt is implemented via explicit transpose of R, which could be very expensive.
9905: We recommend using MatPtAP().
9907: .seealso: [](chapter_matrices), `Mat`, `MatProductCreate()`, `MatMatMult()`, `MatPtAP()`
9908: @*/
9909: PetscErrorCode MatRARt(Mat A, Mat R, MatReuse scall, PetscReal fill, Mat *C)
9910: {
9911: PetscFunctionBegin;
9912: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*C, 5);
9913: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
9915: if (scall == MAT_INITIAL_MATRIX) {
9916: PetscCall(MatProductCreate(A, R, NULL, C));
9917: PetscCall(MatProductSetType(*C, MATPRODUCT_RARt));
9918: PetscCall(MatProductSetAlgorithm(*C, "default"));
9919: PetscCall(MatProductSetFill(*C, fill));
9921: (*C)->product->api_user = PETSC_TRUE;
9922: PetscCall(MatProductSetFromOptions(*C));
9923: 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);
9924: PetscCall(MatProductSymbolic(*C));
9925: } else { /* scall == MAT_REUSE_MATRIX */
9926: PetscCall(MatProductReplaceMats(A, R, NULL, *C));
9927: }
9929: PetscCall(MatProductNumeric(*C));
9930: if (A->symmetric == PETSC_BOOL3_TRUE) PetscCall(MatSetOption(*C, MAT_SYMMETRIC, PETSC_TRUE));
9931: PetscFunctionReturn(PETSC_SUCCESS);
9932: }
9934: static PetscErrorCode MatProduct_Private(Mat A, Mat B, MatReuse scall, PetscReal fill, MatProductType ptype, Mat *C)
9935: {
9936: PetscFunctionBegin;
9937: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
9939: if (scall == MAT_INITIAL_MATRIX) {
9940: PetscCall(PetscInfo(A, "Calling MatProduct API with MAT_INITIAL_MATRIX and product type %s\n", MatProductTypes[ptype]));
9941: PetscCall(MatProductCreate(A, B, NULL, C));
9942: PetscCall(MatProductSetType(*C, ptype));
9943: PetscCall(MatProductSetAlgorithm(*C, MATPRODUCTALGORITHMDEFAULT));
9944: PetscCall(MatProductSetFill(*C, fill));
9946: (*C)->product->api_user = PETSC_TRUE;
9947: PetscCall(MatProductSetFromOptions(*C));
9948: PetscCall(MatProductSymbolic(*C));
9949: } else { /* scall == MAT_REUSE_MATRIX */
9950: Mat_Product *product = (*C)->product;
9951: PetscBool isdense;
9953: PetscCall(PetscObjectBaseTypeCompareAny((PetscObject)(*C), &isdense, MATSEQDENSE, MATMPIDENSE, ""));
9954: if (isdense && product && product->type != ptype) {
9955: PetscCall(MatProductClear(*C));
9956: product = NULL;
9957: }
9958: PetscCall(PetscInfo(A, "Calling MatProduct API with MAT_REUSE_MATRIX %s product present and product type %s\n", product ? "with" : "without", MatProductTypes[ptype]));
9959: if (!product) { /* user provide the dense matrix *C without calling MatProductCreate() or reusing it from previous calls */
9960: PetscCheck(isdense, PetscObjectComm((PetscObject)(*C)), PETSC_ERR_SUP, "Call MatProductCreate() first");
9961: PetscCall(MatProductCreate_Private(A, B, NULL, *C));
9962: product = (*C)->product;
9963: product->fill = fill;
9964: product->api_user = PETSC_TRUE;
9965: product->clear = PETSC_TRUE;
9967: PetscCall(MatProductSetType(*C, ptype));
9968: PetscCall(MatProductSetFromOptions(*C));
9969: PetscCheck((*C)->ops->productsymbolic, PetscObjectComm((PetscObject)(*C)), PETSC_ERR_SUP, "MatProduct %s not supported for %s and %s", MatProductTypes[ptype], ((PetscObject)A)->type_name, ((PetscObject)B)->type_name);
9970: PetscCall(MatProductSymbolic(*C));
9971: } else { /* user may change input matrices A or B when REUSE */
9972: PetscCall(MatProductReplaceMats(A, B, NULL, *C));
9973: }
9974: }
9975: PetscCall(MatProductNumeric(*C));
9976: PetscFunctionReturn(PETSC_SUCCESS);
9977: }
9979: /*@
9980: MatMatMult - Performs matrix-matrix multiplication C=A*B.
9982: Neighbor-wise Collective
9984: Input Parameters:
9985: + A - the left matrix
9986: . B - the right matrix
9987: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
9988: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use `PETSC_DEFAULT` if you do not have a good estimate
9989: if the result is a dense matrix this is irrelevant
9991: Output Parameter:
9992: . C - the product matrix
9994: Notes:
9995: Unless scall is `MAT_REUSE_MATRIX` C will be created.
9997: `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
9998: call to this function with `MAT_INITIAL_MATRIX`.
10000: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value actually needed.
10002: 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`,
10003: rather than first having `MatMatMult()` create it for you. You can NEVER do this if the matrix C is sparse.
10005: Example of Usage:
10006: .vb
10007: MatProductCreate(A,B,NULL,&C);
10008: MatProductSetType(C,MATPRODUCT_AB);
10009: MatProductSymbolic(C);
10010: MatProductNumeric(C); // compute C=A * B
10011: MatProductReplaceMats(A1,B1,NULL,C); // compute C=A1 * B1
10012: MatProductNumeric(C);
10013: MatProductReplaceMats(A2,NULL,NULL,C); // compute C=A2 * B1
10014: MatProductNumeric(C);
10015: .ve
10017: Level: intermediate
10019: .seealso: [](chapter_matrices), `Mat`, `MatProductType`, `MATPRODUCT_AB`, `MatTransposeMatMult()`, `MatMatTransposeMult()`, `MatPtAP()`, `MatProductCreate()`, `MatProductSymbolic()`, `MatProductReplaceMats()`, `MatProductNumeric()`
10020: @*/
10021: PetscErrorCode MatMatMult(Mat A, Mat B, MatReuse scall, PetscReal fill, Mat *C)
10022: {
10023: PetscFunctionBegin;
10024: PetscCall(MatProduct_Private(A, B, scall, fill, MATPRODUCT_AB, C));
10025: PetscFunctionReturn(PETSC_SUCCESS);
10026: }
10028: /*@
10029: MatMatTransposeMult - Performs matrix-matrix multiplication C=A*B^T.
10031: Neighbor-wise Collective
10033: Input Parameters:
10034: + A - the left matrix
10035: . B - the right matrix
10036: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10037: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use `PETSC_DEFAULT` if not known
10039: Output Parameter:
10040: . C - the product matrix
10042: Level: intermediate
10044: Notes:
10045: C will be created if `MAT_INITIAL_MATRIX` and must be destroyed by the user with `MatDestroy()`.
10047: `MAT_REUSE_MATRIX` can only be used if the matrices A and B have the same nonzero pattern as in the previous call
10049: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
10050: actually needed.
10052: This routine is currently only implemented for pairs of `MATSEQAIJ` matrices, for the `MATSEQDENSE` class,
10053: and for pairs of `MATMPIDENSE` matrices.
10055: This routine is shorthand for using `MatProductCreate()` with the `MatProductType` of `MATPRODUCT_ABt`
10057: Options Database Keys:
10058: . -matmattransmult_mpidense_mpidense_via {allgatherv,cyclic} - Choose between algorithms for `MATMPIDENSE` matrices: the
10059: first redundantly copies the transposed B matrix on each process and requiers O(log P) communication complexity;
10060: the second never stores more than one portion of the B matrix at a time by requires O(P) communication complexity.
10062: .seealso: [](chapter_matrices), `Mat`, `MatProductCreate()`, `MATPRODUCT_ABt`, `MatMatMult()`, `MatTransposeMatMult()` `MatPtAP()`, `MatProductCreate()`, `MatProductAlgorithm`, `MatProductType`, `MATPRODUCT_ABt`
10063: @*/
10064: PetscErrorCode MatMatTransposeMult(Mat A, Mat B, MatReuse scall, PetscReal fill, Mat *C)
10065: {
10066: PetscFunctionBegin;
10067: PetscCall(MatProduct_Private(A, B, scall, fill, MATPRODUCT_ABt, C));
10068: if (A == B) PetscCall(MatSetOption(*C, MAT_SYMMETRIC, PETSC_TRUE));
10069: PetscFunctionReturn(PETSC_SUCCESS);
10070: }
10072: /*@
10073: MatTransposeMatMult - Performs matrix-matrix multiplication C=A^T*B.
10075: Neighbor-wise Collective
10077: Input Parameters:
10078: + A - the left matrix
10079: . B - the right matrix
10080: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10081: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use `PETSC_DEFAULT` if not known
10083: Output Parameter:
10084: . C - the product matrix
10086: Level: intermediate
10088: Notes:
10089: C will be created if `MAT_INITIAL_MATRIX` and must be destroyed by the user with `MatDestroy()`.
10091: `MAT_REUSE_MATRIX` can only be used if the matrices A and B have the same nonzero pattern as in the previous call.
10093: This routine is shorthand for using `MatProductCreate()` with the `MatProductType` of `MATPRODUCT_AtB`
10095: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
10096: actually needed.
10098: This routine is currently implemented for pairs of `MATAIJ` matrices and pairs of `MATSEQDENSE` matrices and classes
10099: which inherit from `MATSEQAIJ`. C will be of the same type as the input matrices.
10101: .seealso: [](chapter_matrices), `Mat`, `MatProductCreate()`, `MATPRODUCT_AtB`, `MatMatMult()`, `MatMatTransposeMult()`, `MatPtAP()`
10102: @*/
10103: PetscErrorCode MatTransposeMatMult(Mat A, Mat B, MatReuse scall, PetscReal fill, Mat *C)
10104: {
10105: PetscFunctionBegin;
10106: PetscCall(MatProduct_Private(A, B, scall, fill, MATPRODUCT_AtB, C));
10107: PetscFunctionReturn(PETSC_SUCCESS);
10108: }
10110: /*@
10111: MatMatMatMult - Performs matrix-matrix-matrix multiplication D=A*B*C.
10113: Neighbor-wise Collective
10115: Input Parameters:
10116: + A - the left matrix
10117: . B - the middle matrix
10118: . C - the right matrix
10119: . scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10120: - 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
10121: if the result is a dense matrix this is irrelevant
10123: Output Parameter:
10124: . D - the product matrix
10126: Level: intermediate
10128: Notes:
10129: Unless scall is `MAT_REUSE_MATRIX` D will be created.
10131: `MAT_REUSE_MATRIX` can only be used if the matrices A, B and C have the same nonzero pattern as in the previous call
10133: This routine is shorthand for using `MatProductCreate()` with the `MatProductType` of `MATPRODUCT_ABC`
10135: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
10136: actually needed.
10138: If you have many matrices with the same non-zero structure to multiply, you
10139: should use `MAT_REUSE_MATRIX` in all calls but the first
10141: .seealso: [](chapter_matrices), `Mat`, `MatProductCreate()`, `MATPRODUCT_ABC`, `MatMatMult`, `MatPtAP()`, `MatMatTransposeMult()`, `MatTransposeMatMult()`
10142: @*/
10143: PetscErrorCode MatMatMatMult(Mat A, Mat B, Mat C, MatReuse scall, PetscReal fill, Mat *D)
10144: {
10145: PetscFunctionBegin;
10146: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*D, 6);
10147: PetscCheck(scall != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
10149: if (scall == MAT_INITIAL_MATRIX) {
10150: PetscCall(MatProductCreate(A, B, C, D));
10151: PetscCall(MatProductSetType(*D, MATPRODUCT_ABC));
10152: PetscCall(MatProductSetAlgorithm(*D, "default"));
10153: PetscCall(MatProductSetFill(*D, fill));
10155: (*D)->product->api_user = PETSC_TRUE;
10156: PetscCall(MatProductSetFromOptions(*D));
10157: 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,
10158: ((PetscObject)C)->type_name);
10159: PetscCall(MatProductSymbolic(*D));
10160: } else { /* user may change input matrices when REUSE */
10161: PetscCall(MatProductReplaceMats(A, B, C, *D));
10162: }
10163: PetscCall(MatProductNumeric(*D));
10164: PetscFunctionReturn(PETSC_SUCCESS);
10165: }
10167: /*@
10168: MatCreateRedundantMatrix - Create redundant matrices and put them into processors of subcommunicators.
10170: Collective
10172: Input Parameters:
10173: + mat - the matrix
10174: . nsubcomm - the number of subcommunicators (= number of redundant parallel or sequential matrices)
10175: . subcomm - MPI communicator split from the communicator where mat resides in (or `MPI_COMM_NULL` if nsubcomm is used)
10176: - reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10178: Output Parameter:
10179: . matredundant - redundant matrix
10181: Level: advanced
10183: Notes:
10184: `MAT_REUSE_MATRIX` can only be used when the nonzero structure of the
10185: original matrix has not changed from that last call to MatCreateRedundantMatrix().
10187: This routine creates the duplicated matrices in the subcommunicators; you should NOT create them before
10188: calling it.
10190: `PetscSubcommCreate()` can be used to manage the creation of the subcomm but need not be.
10192: .seealso: [](chapter_matrices), `Mat`, `MatDestroy()`, `PetscSubcommCreate()`, `PetscSubComm`
10193: @*/
10194: PetscErrorCode MatCreateRedundantMatrix(Mat mat, PetscInt nsubcomm, MPI_Comm subcomm, MatReuse reuse, Mat *matredundant)
10195: {
10196: MPI_Comm comm;
10197: PetscMPIInt size;
10198: PetscInt mloc_sub, nloc_sub, rstart, rend, M = mat->rmap->N, N = mat->cmap->N, bs = mat->rmap->bs;
10199: Mat_Redundant *redund = NULL;
10200: PetscSubcomm psubcomm = NULL;
10201: MPI_Comm subcomm_in = subcomm;
10202: Mat *matseq;
10203: IS isrow, iscol;
10204: PetscBool newsubcomm = PETSC_FALSE;
10206: PetscFunctionBegin;
10208: if (nsubcomm && reuse == MAT_REUSE_MATRIX) {
10211: }
10213: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
10214: if (size == 1 || nsubcomm == 1) {
10215: if (reuse == MAT_INITIAL_MATRIX) {
10216: PetscCall(MatDuplicate(mat, MAT_COPY_VALUES, matredundant));
10217: } else {
10218: 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");
10219: PetscCall(MatCopy(mat, *matredundant, SAME_NONZERO_PATTERN));
10220: }
10221: PetscFunctionReturn(PETSC_SUCCESS);
10222: }
10224: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10225: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10226: MatCheckPreallocated(mat, 1);
10228: PetscCall(PetscLogEventBegin(MAT_RedundantMat, mat, 0, 0, 0));
10229: if (subcomm_in == MPI_COMM_NULL && reuse == MAT_INITIAL_MATRIX) { /* get subcomm if user does not provide subcomm */
10230: /* create psubcomm, then get subcomm */
10231: PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
10232: PetscCallMPI(MPI_Comm_size(comm, &size));
10233: PetscCheck(nsubcomm >= 1 && nsubcomm <= size, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "nsubcomm must between 1 and %d", size);
10235: PetscCall(PetscSubcommCreate(comm, &psubcomm));
10236: PetscCall(PetscSubcommSetNumber(psubcomm, nsubcomm));
10237: PetscCall(PetscSubcommSetType(psubcomm, PETSC_SUBCOMM_CONTIGUOUS));
10238: PetscCall(PetscSubcommSetFromOptions(psubcomm));
10239: PetscCall(PetscCommDuplicate(PetscSubcommChild(psubcomm), &subcomm, NULL));
10240: newsubcomm = PETSC_TRUE;
10241: PetscCall(PetscSubcommDestroy(&psubcomm));
10242: }
10244: /* get isrow, iscol and a local sequential matrix matseq[0] */
10245: if (reuse == MAT_INITIAL_MATRIX) {
10246: mloc_sub = PETSC_DECIDE;
10247: nloc_sub = PETSC_DECIDE;
10248: if (bs < 1) {
10249: PetscCall(PetscSplitOwnership(subcomm, &mloc_sub, &M));
10250: PetscCall(PetscSplitOwnership(subcomm, &nloc_sub, &N));
10251: } else {
10252: PetscCall(PetscSplitOwnershipBlock(subcomm, bs, &mloc_sub, &M));
10253: PetscCall(PetscSplitOwnershipBlock(subcomm, bs, &nloc_sub, &N));
10254: }
10255: PetscCallMPI(MPI_Scan(&mloc_sub, &rend, 1, MPIU_INT, MPI_SUM, subcomm));
10256: rstart = rend - mloc_sub;
10257: PetscCall(ISCreateStride(PETSC_COMM_SELF, mloc_sub, rstart, 1, &isrow));
10258: PetscCall(ISCreateStride(PETSC_COMM_SELF, N, 0, 1, &iscol));
10259: } else { /* reuse == MAT_REUSE_MATRIX */
10260: 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");
10261: /* retrieve subcomm */
10262: PetscCall(PetscObjectGetComm((PetscObject)(*matredundant), &subcomm));
10263: redund = (*matredundant)->redundant;
10264: isrow = redund->isrow;
10265: iscol = redund->iscol;
10266: matseq = redund->matseq;
10267: }
10268: PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscol, reuse, &matseq));
10270: /* get matredundant over subcomm */
10271: if (reuse == MAT_INITIAL_MATRIX) {
10272: PetscCall(MatCreateMPIMatConcatenateSeqMat(subcomm, matseq[0], nloc_sub, reuse, matredundant));
10274: /* create a supporting struct and attach it to C for reuse */
10275: PetscCall(PetscNew(&redund));
10276: (*matredundant)->redundant = redund;
10277: redund->isrow = isrow;
10278: redund->iscol = iscol;
10279: redund->matseq = matseq;
10280: if (newsubcomm) {
10281: redund->subcomm = subcomm;
10282: } else {
10283: redund->subcomm = MPI_COMM_NULL;
10284: }
10285: } else {
10286: PetscCall(MatCreateMPIMatConcatenateSeqMat(subcomm, matseq[0], PETSC_DECIDE, reuse, matredundant));
10287: }
10288: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP)
10289: if (matseq[0]->boundtocpu && matseq[0]->bindingpropagates) {
10290: PetscCall(MatBindToCPU(*matredundant, PETSC_TRUE));
10291: PetscCall(MatSetBindingPropagates(*matredundant, PETSC_TRUE));
10292: }
10293: #endif
10294: PetscCall(PetscLogEventEnd(MAT_RedundantMat, mat, 0, 0, 0));
10295: PetscFunctionReturn(PETSC_SUCCESS);
10296: }
10298: /*@C
10299: MatGetMultiProcBlock - Create multiple 'parallel submatrices' from
10300: a given `Mat`. Each submatrix can span multiple procs.
10302: Collective
10304: Input Parameters:
10305: + mat - the matrix
10306: . subcomm - the sub communicator obtained as if by `MPI_Comm_split(PetscObjectComm((PetscObject)mat))`
10307: - scall - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10309: Output Parameter:
10310: . subMat - parallel sub-matrices each spanning a given `subcomm`
10312: Level: advanced
10314: Notes:
10315: The submatrix partition across processors is dictated by `subComm` a
10316: communicator obtained by `MPI_comm_split()` or via `PetscSubcommCreate()`. The `subComm`
10317: is not restricted to be grouped with consecutive original ranks.
10319: Due the `MPI_Comm_split()` usage, the parallel layout of the submatrices
10320: map directly to the layout of the original matrix [wrt the local
10321: row,col partitioning]. So the original 'DiagonalMat' naturally maps
10322: into the 'DiagonalMat' of the `subMat`, hence it is used directly from
10323: the `subMat`. However the offDiagMat looses some columns - and this is
10324: reconstructed with `MatSetValues()`
10326: This is used by `PCBJACOBI` when a single block spans multiple MPI ranks
10328: .seealso: [](chapter_matrices), `Mat`, `MatCreateRedundantMatrix()`, `MatCreateSubMatrices()`, `PCBJACOBI`
10329: @*/
10330: PetscErrorCode MatGetMultiProcBlock(Mat mat, MPI_Comm subComm, MatReuse scall, Mat *subMat)
10331: {
10332: PetscMPIInt commsize, subCommSize;
10334: PetscFunctionBegin;
10335: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &commsize));
10336: PetscCallMPI(MPI_Comm_size(subComm, &subCommSize));
10337: PetscCheck(subCommSize <= commsize, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_OUTOFRANGE, "CommSize %d < SubCommZize %d", commsize, subCommSize);
10339: 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");
10340: PetscCall(PetscLogEventBegin(MAT_GetMultiProcBlock, mat, 0, 0, 0));
10341: PetscUseTypeMethod(mat, getmultiprocblock, subComm, scall, subMat);
10342: PetscCall(PetscLogEventEnd(MAT_GetMultiProcBlock, mat, 0, 0, 0));
10343: PetscFunctionReturn(PETSC_SUCCESS);
10344: }
10346: /*@
10347: MatGetLocalSubMatrix - Gets a reference to a submatrix specified in local numbering
10349: Not Collective
10351: Input Parameters:
10352: + mat - matrix to extract local submatrix from
10353: . isrow - local row indices for submatrix
10354: - iscol - local column indices for submatrix
10356: Output Parameter:
10357: . submat - the submatrix
10359: Level: intermediate
10361: Notes:
10362: `submat` should be disposed of with `MatRestoreLocalSubMatrix()`.
10364: Depending on the format of `mat`, the returned submat may not implement `MatMult()`. Its communicator may be
10365: the same as mat, it may be `PETSC_COMM_SELF`, or some other subcomm of `mat`'s.
10367: `submat` always implements `MatSetValuesLocal()`. If `isrow` and `iscol` have the same block size, then
10368: `MatSetValuesBlockedLocal()` will also be implemented.
10370: `mat` must have had a `ISLocalToGlobalMapping` provided to it with `MatSetLocalToGlobalMapping()`.
10371: Matrices obtained with `DMCreateMatrix()` generally already have the local to global mapping provided.
10373: .seealso: [](chapter_matrices), `Mat`, `MatRestoreLocalSubMatrix()`, `MatCreateLocalRef()`, `MatSetLocalToGlobalMapping()`
10374: @*/
10375: PetscErrorCode MatGetLocalSubMatrix(Mat mat, IS isrow, IS iscol, Mat *submat)
10376: {
10377: PetscFunctionBegin;
10381: PetscCheckSameComm(isrow, 2, iscol, 3);
10383: PetscCheck(mat->rmap->mapping, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Matrix must have local to global mapping provided before this call");
10385: if (mat->ops->getlocalsubmatrix) {
10386: PetscUseTypeMethod(mat, getlocalsubmatrix, isrow, iscol, submat);
10387: } else {
10388: PetscCall(MatCreateLocalRef(mat, isrow, iscol, submat));
10389: }
10390: PetscFunctionReturn(PETSC_SUCCESS);
10391: }
10393: /*@
10394: MatRestoreLocalSubMatrix - Restores a reference to a submatrix specified in local numbering obtained with `MatGetLocalSubMatrix()`
10396: Not Collective
10398: Input Parameters:
10399: + mat - matrix to extract local submatrix from
10400: . isrow - local row indices for submatrix
10401: . iscol - local column indices for submatrix
10402: - submat - the submatrix
10404: Level: intermediate
10406: .seealso: [](chapter_matrices), `Mat`, `MatGetLocalSubMatrix()`
10407: @*/
10408: PetscErrorCode MatRestoreLocalSubMatrix(Mat mat, IS isrow, IS iscol, Mat *submat)
10409: {
10410: PetscFunctionBegin;
10414: PetscCheckSameComm(isrow, 2, iscol, 3);
10418: if (mat->ops->restorelocalsubmatrix) {
10419: PetscUseTypeMethod(mat, restorelocalsubmatrix, isrow, iscol, submat);
10420: } else {
10421: PetscCall(MatDestroy(submat));
10422: }
10423: *submat = NULL;
10424: PetscFunctionReturn(PETSC_SUCCESS);
10425: }
10427: /*@
10428: MatFindZeroDiagonals - Finds all the rows of a matrix that have zero or no diagonal entry in the matrix
10430: Collective
10432: Input Parameter:
10433: . mat - the matrix
10435: Output Parameter:
10436: . is - if any rows have zero diagonals this contains the list of them
10438: Level: developer
10440: .seealso: [](chapter_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
10441: @*/
10442: PetscErrorCode MatFindZeroDiagonals(Mat mat, IS *is)
10443: {
10444: PetscFunctionBegin;
10447: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10448: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10450: if (!mat->ops->findzerodiagonals) {
10451: Vec diag;
10452: const PetscScalar *a;
10453: PetscInt *rows;
10454: PetscInt rStart, rEnd, r, nrow = 0;
10456: PetscCall(MatCreateVecs(mat, &diag, NULL));
10457: PetscCall(MatGetDiagonal(mat, diag));
10458: PetscCall(MatGetOwnershipRange(mat, &rStart, &rEnd));
10459: PetscCall(VecGetArrayRead(diag, &a));
10460: for (r = 0; r < rEnd - rStart; ++r)
10461: if (a[r] == 0.0) ++nrow;
10462: PetscCall(PetscMalloc1(nrow, &rows));
10463: nrow = 0;
10464: for (r = 0; r < rEnd - rStart; ++r)
10465: if (a[r] == 0.0) rows[nrow++] = r + rStart;
10466: PetscCall(VecRestoreArrayRead(diag, &a));
10467: PetscCall(VecDestroy(&diag));
10468: PetscCall(ISCreateGeneral(PetscObjectComm((PetscObject)mat), nrow, rows, PETSC_OWN_POINTER, is));
10469: } else {
10470: PetscUseTypeMethod(mat, findzerodiagonals, is);
10471: }
10472: PetscFunctionReturn(PETSC_SUCCESS);
10473: }
10475: /*@
10476: MatFindOffBlockDiagonalEntries - Finds all the rows of a matrix that have entries outside of the main diagonal block (defined by the matrix block size)
10478: Collective
10480: Input Parameter:
10481: . mat - the matrix
10483: Output Parameter:
10484: . is - contains the list of rows with off block diagonal entries
10486: Level: developer
10488: .seealso: [](chapter_matrices), `Mat`, `MatMultTranspose()`, `MatMultAdd()`, `MatMultTransposeAdd()`
10489: @*/
10490: PetscErrorCode MatFindOffBlockDiagonalEntries(Mat mat, IS *is)
10491: {
10492: PetscFunctionBegin;
10495: PetscCheck(mat->assembled, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10496: PetscCheck(!mat->factortype, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10498: PetscUseTypeMethod(mat, findoffblockdiagonalentries, is);
10499: PetscFunctionReturn(PETSC_SUCCESS);
10500: }
10502: /*@C
10503: MatInvertBlockDiagonal - Inverts the block diagonal entries.
10505: Collective; No Fortran Support
10507: Input Parameter:
10508: . mat - the matrix
10510: Output Parameter:
10511: . values - the block inverses in column major order (FORTRAN-like)
10513: Level: advanced
10515: Notes:
10516: The size of the blocks is determined by the block size of the matrix.
10518: The blocks never overlap between two MPI ranks, use `MatInvertVariableBlockEnvelope()` for that case
10520: The blocks all have the same size, use `MatInvertVariableBlockDiagonal()` for variable block size
10522: .seealso: [](chapter_matrices), `Mat`, `MatInvertVariableBlockEnvelope()`, `MatInvertBlockDiagonalMat()`
10523: @*/
10524: PetscErrorCode MatInvertBlockDiagonal(Mat mat, const PetscScalar **values)
10525: {
10526: PetscFunctionBegin;
10528: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10529: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10530: PetscUseTypeMethod(mat, invertblockdiagonal, values);
10531: PetscFunctionReturn(PETSC_SUCCESS);
10532: }
10534: /*@C
10535: MatInvertVariableBlockDiagonal - Inverts the point block diagonal entries.
10537: Collective; No Fortran Support
10539: Input Parameters:
10540: + mat - the matrix
10541: . nblocks - the number of blocks on the process, set with `MatSetVariableBlockSizes()`
10542: - bsizes - the size of each block on the process, set with `MatSetVariableBlockSizes()`
10544: Output Parameter:
10545: . values - the block inverses in column major order (FORTRAN-like)
10547: Level: advanced
10549: Notes:
10550: Use `MatInvertBlockDiagonal()` if all blocks have the same size
10552: The blocks never overlap between two MPI ranks, use `MatInvertVariableBlockEnvelope()` for that case
10554: .seealso: [](chapter_matrices), `Mat`, `MatInvertBlockDiagonal()`, `MatSetVariableBlockSizes()`, `MatInvertVariableBlockEnvelope()`
10555: @*/
10556: PetscErrorCode MatInvertVariableBlockDiagonal(Mat mat, PetscInt nblocks, const PetscInt *bsizes, PetscScalar *values)
10557: {
10558: PetscFunctionBegin;
10560: PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
10561: PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
10562: PetscUseTypeMethod(mat, invertvariableblockdiagonal, nblocks, bsizes, values);
10563: PetscFunctionReturn(PETSC_SUCCESS);
10564: }
10566: /*@
10567: MatInvertBlockDiagonalMat - set the values of matrix C to be the inverted block diagonal of matrix A
10569: Collective
10571: Input Parameters:
10572: + A - the matrix
10573: - C - matrix with inverted block diagonal of `A`. This matrix should be created and may have its type set.
10575: Level: advanced
10577: Note:
10578: The blocksize of the matrix is used to determine the blocks on the diagonal of `C`
10580: .seealso: [](chapter_matrices), `Mat`, `MatInvertBlockDiagonal()`
10581: @*/
10582: PetscErrorCode MatInvertBlockDiagonalMat(Mat A, Mat C)
10583: {
10584: const PetscScalar *vals;
10585: PetscInt *dnnz;
10586: PetscInt m, rstart, rend, bs, i, j;
10588: PetscFunctionBegin;
10589: PetscCall(MatInvertBlockDiagonal(A, &vals));
10590: PetscCall(MatGetBlockSize(A, &bs));
10591: PetscCall(MatGetLocalSize(A, &m, NULL));
10592: PetscCall(MatSetLayouts(C, A->rmap, A->cmap));
10593: PetscCall(PetscMalloc1(m / bs, &dnnz));
10594: for (j = 0; j < m / bs; j++) dnnz[j] = 1;
10595: PetscCall(MatXAIJSetPreallocation(C, bs, dnnz, NULL, NULL, NULL));
10596: PetscCall(PetscFree(dnnz));
10597: PetscCall(MatGetOwnershipRange(C, &rstart, &rend));
10598: PetscCall(MatSetOption(C, MAT_ROW_ORIENTED, PETSC_FALSE));
10599: for (i = rstart / bs; i < rend / bs; i++) PetscCall(MatSetValuesBlocked(C, 1, &i, 1, &i, &vals[(i - rstart / bs) * bs * bs], INSERT_VALUES));
10600: PetscCall(MatAssemblyBegin(C, MAT_FINAL_ASSEMBLY));
10601: PetscCall(MatAssemblyEnd(C, MAT_FINAL_ASSEMBLY));
10602: PetscCall(MatSetOption(C, MAT_ROW_ORIENTED, PETSC_TRUE));
10603: PetscFunctionReturn(PETSC_SUCCESS);
10604: }
10606: /*@C
10607: MatTransposeColoringDestroy - Destroys a coloring context for matrix product C=A*B^T that was created
10608: via `MatTransposeColoringCreate()`.
10610: Collective
10612: Input Parameter:
10613: . c - coloring context
10615: Level: intermediate
10617: .seealso: [](chapter_matrices), `Mat`, `MatTransposeColoringCreate()`
10618: @*/
10619: PetscErrorCode MatTransposeColoringDestroy(MatTransposeColoring *c)
10620: {
10621: MatTransposeColoring matcolor = *c;
10623: PetscFunctionBegin;
10624: if (!matcolor) PetscFunctionReturn(PETSC_SUCCESS);
10625: if (--((PetscObject)matcolor)->refct > 0) {
10626: matcolor = NULL;
10627: PetscFunctionReturn(PETSC_SUCCESS);
10628: }
10630: PetscCall(PetscFree3(matcolor->ncolumns, matcolor->nrows, matcolor->colorforrow));
10631: PetscCall(PetscFree(matcolor->rows));
10632: PetscCall(PetscFree(matcolor->den2sp));
10633: PetscCall(PetscFree(matcolor->colorforcol));
10634: PetscCall(PetscFree(matcolor->columns));
10635: if (matcolor->brows > 0) PetscCall(PetscFree(matcolor->lstart));
10636: PetscCall(PetscHeaderDestroy(c));
10637: PetscFunctionReturn(PETSC_SUCCESS);
10638: }
10640: /*@C
10641: MatTransColoringApplySpToDen - Given a symbolic matrix product C=A*B^T for which
10642: a `MatTransposeColoring` context has been created, computes a dense B^T by applying
10643: `MatTransposeColoring` to sparse B.
10645: Collective
10647: Input Parameters:
10648: + coloring - coloring context created with `MatTransposeColoringCreate()`
10649: - B - sparse matrix
10651: Output Parameter:
10652: . Btdense - dense matrix B^T
10654: Level: developer
10656: Note:
10657: These are used internally for some implementations of `MatRARt()`
10659: .seealso: [](chapter_matrices), `Mat`, `MatTransposeColoringCreate()`, `MatTransposeColoringDestroy()`, `MatTransColoringApplyDenToSp()`
10660: @*/
10661: PetscErrorCode MatTransColoringApplySpToDen(MatTransposeColoring coloring, Mat B, Mat Btdense)
10662: {
10663: PetscFunctionBegin;
10668: PetscCall((*B->ops->transcoloringapplysptoden)(coloring, B, Btdense));
10669: PetscFunctionReturn(PETSC_SUCCESS);
10670: }
10672: /*@C
10673: MatTransColoringApplyDenToSp - Given a symbolic matrix product Csp=A*B^T for which
10674: a `MatTransposeColoring` context has been created and a dense matrix Cden=A*Btdense
10675: in which Btdens is obtained from `MatTransColoringApplySpToDen()`, recover sparse matrix
10676: `Csp` from `Cden`.
10678: Collective
10680: Input Parameters:
10681: + matcoloring - coloring context created with `MatTransposeColoringCreate()`
10682: - Cden - matrix product of a sparse matrix and a dense matrix Btdense
10684: Output Parameter:
10685: . Csp - sparse matrix
10687: Level: developer
10689: Note:
10690: These are used internally for some implementations of `MatRARt()`
10692: .seealso: [](chapter_matrices), `Mat`, `MatTransposeColoringCreate()`, `MatTransposeColoringDestroy()`, `MatTransColoringApplySpToDen()`
10693: @*/
10694: PetscErrorCode MatTransColoringApplyDenToSp(MatTransposeColoring matcoloring, Mat Cden, Mat Csp)
10695: {
10696: PetscFunctionBegin;
10701: PetscCall((*Csp->ops->transcoloringapplydentosp)(matcoloring, Cden, Csp));
10702: PetscCall(MatAssemblyBegin(Csp, MAT_FINAL_ASSEMBLY));
10703: PetscCall(MatAssemblyEnd(Csp, MAT_FINAL_ASSEMBLY));
10704: PetscFunctionReturn(PETSC_SUCCESS);
10705: }
10707: /*@C
10708: MatTransposeColoringCreate - Creates a matrix coloring context for the matrix product C=A*B^T.
10710: Collective
10712: Input Parameters:
10713: + mat - the matrix product C
10714: - iscoloring - the coloring of the matrix; usually obtained with `MatColoringCreate()` or `DMCreateColoring()`
10716: Output Parameter:
10717: . color - the new coloring context
10719: Level: intermediate
10721: .seealso: [](chapter_matrices), `Mat`, `MatTransposeColoringDestroy()`, `MatTransColoringApplySpToDen()`,
10722: `MatTransColoringApplyDenToSp()`
10723: @*/
10724: PetscErrorCode MatTransposeColoringCreate(Mat mat, ISColoring iscoloring, MatTransposeColoring *color)
10725: {
10726: MatTransposeColoring c;
10727: MPI_Comm comm;
10729: PetscFunctionBegin;
10730: PetscCall(PetscLogEventBegin(MAT_TransposeColoringCreate, mat, 0, 0, 0));
10731: PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
10732: PetscCall(PetscHeaderCreate(c, MAT_TRANSPOSECOLORING_CLASSID, "MatTransposeColoring", "Matrix product C=A*B^T via coloring", "Mat", comm, MatTransposeColoringDestroy, NULL));
10734: c->ctype = iscoloring->ctype;
10735: PetscUseTypeMethod(mat, transposecoloringcreate, iscoloring, c);
10737: *color = c;
10738: PetscCall(PetscLogEventEnd(MAT_TransposeColoringCreate, mat, 0, 0, 0));
10739: PetscFunctionReturn(PETSC_SUCCESS);
10740: }
10742: /*@
10743: MatGetNonzeroState - Returns a 64 bit integer representing the current state of nonzeros in the matrix. If the
10744: matrix has had no new nonzero locations added to (or removed from) the matrix since the previous call then the value will be the
10745: same, otherwise it will be larger
10747: Not Collective
10749: Input Parameter:
10750: . A - the matrix
10752: Output Parameter:
10753: . state - the current state
10755: Level: intermediate
10757: Notes:
10758: You can only compare states from two different calls to the SAME matrix, you cannot compare calls between
10759: different matrices
10761: Use `PetscObjectStateGet()` to check for changes to the numerical values in a matrix
10763: Use the result of `PetscObjectGetId()` to compare if a previously checked matrix is the same as the current matrix, do not compare object pointers.
10765: .seealso: [](chapter_matrices), `Mat`, `PetscObjectStateGet()`, `PetscObjectGetId()`
10766: @*/
10767: PetscErrorCode MatGetNonzeroState(Mat mat, PetscObjectState *state)
10768: {
10769: PetscFunctionBegin;
10771: *state = mat->nonzerostate;
10772: PetscFunctionReturn(PETSC_SUCCESS);
10773: }
10775: /*@
10776: MatCreateMPIMatConcatenateSeqMat - Creates a single large PETSc matrix by concatenating sequential
10777: matrices from each processor
10779: Collective
10781: Input Parameters:
10782: + comm - the communicators the parallel matrix will live on
10783: . seqmat - the input sequential matrices
10784: . n - number of local columns (or `PETSC_DECIDE`)
10785: - reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10787: Output Parameter:
10788: . mpimat - the parallel matrix generated
10790: Level: developer
10792: Note:
10793: The number of columns of the matrix in EACH processor MUST be the same.
10795: .seealso: [](chapter_matrices), `Mat`
10796: @*/
10797: PetscErrorCode MatCreateMPIMatConcatenateSeqMat(MPI_Comm comm, Mat seqmat, PetscInt n, MatReuse reuse, Mat *mpimat)
10798: {
10799: PetscMPIInt size;
10801: PetscFunctionBegin;
10802: PetscCallMPI(MPI_Comm_size(comm, &size));
10803: if (size == 1) {
10804: if (reuse == MAT_INITIAL_MATRIX) {
10805: PetscCall(MatDuplicate(seqmat, MAT_COPY_VALUES, mpimat));
10806: } else {
10807: PetscCall(MatCopy(seqmat, *mpimat, SAME_NONZERO_PATTERN));
10808: }
10809: PetscFunctionReturn(PETSC_SUCCESS);
10810: }
10812: 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");
10814: PetscCall(PetscLogEventBegin(MAT_Merge, seqmat, 0, 0, 0));
10815: PetscCall((*seqmat->ops->creatempimatconcatenateseqmat)(comm, seqmat, n, reuse, mpimat));
10816: PetscCall(PetscLogEventEnd(MAT_Merge, seqmat, 0, 0, 0));
10817: PetscFunctionReturn(PETSC_SUCCESS);
10818: }
10820: /*@
10821: MatSubdomainsCreateCoalesce - Creates index subdomains by coalescing adjacent ranks' ownership ranges.
10823: Collective
10825: Input Parameters:
10826: + A - the matrix to create subdomains from
10827: - N - requested number of subdomains
10829: Output Parameters:
10830: + n - number of subdomains resulting on this rank
10831: - iss - `IS` list with indices of subdomains on this rank
10833: Level: advanced
10835: Note:
10836: The number of subdomains must be smaller than the communicator size
10838: .seealso: [](chapter_matrices), `Mat`, `IS`
10839: @*/
10840: PetscErrorCode MatSubdomainsCreateCoalesce(Mat A, PetscInt N, PetscInt *n, IS *iss[])
10841: {
10842: MPI_Comm comm, subcomm;
10843: PetscMPIInt size, rank, color;
10844: PetscInt rstart, rend, k;
10846: PetscFunctionBegin;
10847: PetscCall(PetscObjectGetComm((PetscObject)A, &comm));
10848: PetscCallMPI(MPI_Comm_size(comm, &size));
10849: PetscCallMPI(MPI_Comm_rank(comm, &rank));
10850: 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);
10851: *n = 1;
10852: k = ((PetscInt)size) / N + ((PetscInt)size % N > 0); /* There are up to k ranks to a color */
10853: color = rank / k;
10854: PetscCallMPI(MPI_Comm_split(comm, color, rank, &subcomm));
10855: PetscCall(PetscMalloc1(1, iss));
10856: PetscCall(MatGetOwnershipRange(A, &rstart, &rend));
10857: PetscCall(ISCreateStride(subcomm, rend - rstart, rstart, 1, iss[0]));
10858: PetscCallMPI(MPI_Comm_free(&subcomm));
10859: PetscFunctionReturn(PETSC_SUCCESS);
10860: }
10862: /*@
10863: MatGalerkin - Constructs the coarse grid problem matrix via Galerkin projection.
10865: If the interpolation and restriction operators are the same, uses `MatPtAP()`.
10866: If they are not the same, uses `MatMatMatMult()`.
10868: Once the coarse grid problem is constructed, correct for interpolation operators
10869: that are not of full rank, which can legitimately happen in the case of non-nested
10870: geometric multigrid.
10872: Input Parameters:
10873: + restrct - restriction operator
10874: . dA - fine grid matrix
10875: . interpolate - interpolation operator
10876: . reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
10877: - fill - expected fill, use `PETSC_DEFAULT` if you do not have a good estimate
10879: Output Parameter:
10880: . A - the Galerkin coarse matrix
10882: Options Database Key:
10883: . -pc_mg_galerkin <both,pmat,mat,none> - for what matrices the Galerkin process should be used
10885: Level: developer
10887: .seealso: [](chapter_matrices), `Mat`, `MatPtAP()`, `MatMatMatMult()`
10888: @*/
10889: PetscErrorCode MatGalerkin(Mat restrct, Mat dA, Mat interpolate, MatReuse reuse, PetscReal fill, Mat *A)
10890: {
10891: IS zerorows;
10892: Vec diag;
10894: PetscFunctionBegin;
10895: PetscCheck(reuse != MAT_INPLACE_MATRIX, PetscObjectComm((PetscObject)A), PETSC_ERR_SUP, "Inplace product not supported");
10896: /* Construct the coarse grid matrix */
10897: if (interpolate == restrct) {
10898: PetscCall(MatPtAP(dA, interpolate, reuse, fill, A));
10899: } else {
10900: PetscCall(MatMatMatMult(restrct, dA, interpolate, reuse, fill, A));
10901: }
10903: /* If the interpolation matrix is not of full rank, A will have zero rows.
10904: This can legitimately happen in the case of non-nested geometric multigrid.
10905: In that event, we set the rows of the matrix to the rows of the identity,
10906: ignoring the equations (as the RHS will also be zero). */
10908: PetscCall(MatFindZeroRows(*A, &zerorows));
10910: if (zerorows != NULL) { /* if there are any zero rows */
10911: PetscCall(MatCreateVecs(*A, &diag, NULL));
10912: PetscCall(MatGetDiagonal(*A, diag));
10913: PetscCall(VecISSet(diag, zerorows, 1.0));
10914: PetscCall(MatDiagonalSet(*A, diag, INSERT_VALUES));
10915: PetscCall(VecDestroy(&diag));
10916: PetscCall(ISDestroy(&zerorows));
10917: }
10918: PetscFunctionReturn(PETSC_SUCCESS);
10919: }
10921: /*@C
10922: MatSetOperation - Allows user to set a matrix operation for any matrix type
10924: Logically Collective
10926: Input Parameters:
10927: + mat - the matrix
10928: . op - the name of the operation
10929: - f - the function that provides the operation
10931: Level: developer
10933: Usage:
10934: .vb
10935: extern PetscErrorCode usermult(Mat, Vec, Vec);
10937: PetscCall(MatCreateXXX(comm, ..., &A));
10938: PetscCall(MatSetOperation(A, MATOP_MULT, (PetscVoidFunction)usermult));
10939: .ve
10941: Notes:
10942: See the file `include/petscmat.h` for a complete list of matrix
10943: operations, which all have the form MATOP_<OPERATION>, where
10944: <OPERATION> is the name (in all capital letters) of the
10945: user interface routine (e.g., `MatMult()` -> `MATOP_MULT`).
10947: All user-provided functions (except for `MATOP_DESTROY`) should have the same calling
10948: sequence as the usual matrix interface routines, since they
10949: are intended to be accessed via the usual matrix interface
10950: routines, e.g.,
10951: .vb
10952: MatMult(Mat, Vec, Vec) -> usermult(Mat, Vec, Vec)
10953: .ve
10955: In particular each function MUST return `PETSC_SUCCESS` on success and
10956: nonzero on failure.
10958: This routine is distinct from `MatShellSetOperation()` in that it can be called on any matrix type.
10960: .seealso: [](chapter_matrices), `Mat`, `MatGetOperation()`, `MatCreateShell()`, `MatShellSetContext()`, `MatShellSetOperation()`
10961: @*/
10962: PetscErrorCode MatSetOperation(Mat mat, MatOperation op, void (*f)(void))
10963: {
10964: PetscFunctionBegin;
10966: if (op == MATOP_VIEW && !mat->ops->viewnative && f != (void (*)(void))(mat->ops->view)) mat->ops->viewnative = mat->ops->view;
10967: (((void (**)(void))mat->ops)[op]) = f;
10968: PetscFunctionReturn(PETSC_SUCCESS);
10969: }
10971: /*@C
10972: MatGetOperation - Gets a matrix operation for any matrix type.
10974: Not Collective
10976: Input Parameters:
10977: + mat - the matrix
10978: - op - the name of the operation
10980: Output Parameter:
10981: . f - the function that provides the operation
10983: Level: developer
10985: Usage:
10986: .vb
10987: PetscErrorCode (*usermult)(Mat, Vec, Vec);
10988: MatGetOperation(A, MATOP_MULT, (void (**)(void))&usermult);
10989: .ve
10991: Notes:
10992: See the file include/petscmat.h for a complete list of matrix
10993: operations, which all have the form MATOP_<OPERATION>, where
10994: <OPERATION> is the name (in all capital letters) of the
10995: user interface routine (e.g., `MatMult()` -> `MATOP_MULT`).
10997: This routine is distinct from `MatShellGetOperation()` in that it can be called on any matrix type.
10999: .seealso: [](chapter_matrices), `Mat`, `MatSetOperation()`, `MatCreateShell()`, `MatShellGetContext()`, `MatShellGetOperation()`
11000: @*/
11001: PetscErrorCode MatGetOperation(Mat mat, MatOperation op, void (**f)(void))
11002: {
11003: PetscFunctionBegin;
11005: *f = (((void (**)(void))mat->ops)[op]);
11006: PetscFunctionReturn(PETSC_SUCCESS);
11007: }
11009: /*@
11010: MatHasOperation - Determines whether the given matrix supports the particular operation.
11012: Not Collective
11014: Input Parameters:
11015: + mat - the matrix
11016: - op - the operation, for example, `MATOP_GET_DIAGONAL`
11018: Output Parameter:
11019: . has - either `PETSC_TRUE` or `PETSC_FALSE`
11021: Level: advanced
11023: Note:
11024: See `MatSetOperation()` for additional discussion on naming convention and usage of `op`.
11026: .seealso: [](chapter_matrices), `Mat`, `MatCreateShell()`, `MatGetOperation()`, `MatSetOperation()`
11027: @*/
11028: PetscErrorCode MatHasOperation(Mat mat, MatOperation op, PetscBool *has)
11029: {
11030: PetscFunctionBegin;
11033: if (mat->ops->hasoperation) {
11034: PetscUseTypeMethod(mat, hasoperation, op, has);
11035: } else {
11036: if (((void **)mat->ops)[op]) *has = PETSC_TRUE;
11037: else {
11038: *has = PETSC_FALSE;
11039: if (op == MATOP_CREATE_SUBMATRIX) {
11040: PetscMPIInt size;
11042: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
11043: if (size == 1) PetscCall(MatHasOperation(mat, MATOP_CREATE_SUBMATRICES, has));
11044: }
11045: }
11046: }
11047: PetscFunctionReturn(PETSC_SUCCESS);
11048: }
11050: /*@
11051: MatHasCongruentLayouts - Determines whether the rows and columns layouts of the matrix are congruent
11053: Collective
11055: Input Parameter:
11056: . mat - the matrix
11058: Output Parameter:
11059: . cong - either `PETSC_TRUE` or `PETSC_FALSE`
11061: Level: beginner
11063: .seealso: [](chapter_matrices), `Mat`, `MatCreate()`, `MatSetSizes()`, `PetscLayout`
11064: @*/
11065: PetscErrorCode MatHasCongruentLayouts(Mat mat, PetscBool *cong)
11066: {
11067: PetscFunctionBegin;
11071: if (!mat->rmap || !mat->cmap) {
11072: *cong = mat->rmap == mat->cmap ? PETSC_TRUE : PETSC_FALSE;
11073: PetscFunctionReturn(PETSC_SUCCESS);
11074: }
11075: if (mat->congruentlayouts == PETSC_DECIDE) { /* first time we compare rows and cols layouts */
11076: PetscCall(PetscLayoutSetUp(mat->rmap));
11077: PetscCall(PetscLayoutSetUp(mat->cmap));
11078: PetscCall(PetscLayoutCompare(mat->rmap, mat->cmap, cong));
11079: if (*cong) mat->congruentlayouts = 1;
11080: else mat->congruentlayouts = 0;
11081: } else *cong = mat->congruentlayouts ? PETSC_TRUE : PETSC_FALSE;
11082: PetscFunctionReturn(PETSC_SUCCESS);
11083: }
11085: PetscErrorCode MatSetInf(Mat A)
11086: {
11087: PetscFunctionBegin;
11088: PetscUseTypeMethod(A, setinf);
11089: PetscFunctionReturn(PETSC_SUCCESS);
11090: }
11092: /*@C
11093: 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
11094: and possibly removes small values from the graph structure.
11096: Collective
11098: Input Parameters:
11099: + A - the matrix
11100: . sym - `PETSC_TRUE` indicates that the graph should be symmetrized
11101: . scale - `PETSC_TRUE` indicates that the graph edge weights should be symmetrically scaled with the diagonal entry
11102: - filter - filter value - < 0: does nothing; == 0: removes only 0.0 entries; otherwise: removes entries with abs(entries) <= value
11104: Output Parameter:
11105: . graph - the resulting graph
11107: Level: advanced
11109: .seealso: [](chapter_matrices), `Mat`, `MatCreate()`, `PCGAMG`
11110: @*/
11111: PetscErrorCode MatCreateGraph(Mat A, PetscBool sym, PetscBool scale, PetscReal filter, Mat *graph)
11112: {
11113: PetscFunctionBegin;
11118: PetscUseTypeMethod(A, creategraph, sym, scale, filter, graph);
11119: PetscFunctionReturn(PETSC_SUCCESS);
11120: }
11122: /*@
11123: MatEliminateZeros - eliminate the nondiagonal zero entries in place from the nonzero structure of a sparse `Mat` in place,
11124: meaning the same memory is used for the matrix, and no new memory is allocated.
11126: Collective
11128: Input Parameter:
11129: . A - the matrix
11131: Level: intermediate
11133: Developer Note:
11134: The entries in the sparse matrix data structure are shifted to fill in the unneeded locations in the data. Thus the end
11135: of the arrays in the data structure are unneeded.
11137: .seealso: [](chapter_matrices), `Mat`, `MatCreate()`, `MatCreateGraph()`, `MatChop()`
11138: @*/
11139: PetscErrorCode MatEliminateZeros(Mat A)
11140: {
11141: PetscFunctionBegin;
11143: PetscUseTypeMethod(A, eliminatezeros);
11144: PetscFunctionReturn(PETSC_SUCCESS);
11145: }