Actual source code: matrix.c
petsc-3.14.6 2021-03-30
1: /*
2: This is where the abstract matrix operations are defined
3: */
5: #include <petsc/private/matimpl.h>
6: #include <petsc/private/isimpl.h>
7: #include <petsc/private/vecimpl.h>
9: /* Logging support */
10: PetscClassId MAT_CLASSID;
11: PetscClassId MAT_COLORING_CLASSID;
12: PetscClassId MAT_FDCOLORING_CLASSID;
13: PetscClassId MAT_TRANSPOSECOLORING_CLASSID;
15: PetscLogEvent MAT_Mult, MAT_Mults, MAT_MultConstrained, MAT_MultAdd, MAT_MultTranspose;
16: PetscLogEvent MAT_MultTransposeConstrained, MAT_MultTransposeAdd, MAT_Solve, MAT_Solves, MAT_SolveAdd, MAT_SolveTranspose, MAT_MatSolve,MAT_MatTrSolve;
17: PetscLogEvent MAT_SolveTransposeAdd, MAT_SOR, MAT_ForwardSolve, MAT_BackwardSolve, MAT_LUFactor, MAT_LUFactorSymbolic;
18: PetscLogEvent MAT_LUFactorNumeric, MAT_CholeskyFactor, MAT_CholeskyFactorSymbolic, MAT_CholeskyFactorNumeric, MAT_ILUFactor;
19: PetscLogEvent MAT_ILUFactorSymbolic, MAT_ICCFactorSymbolic, MAT_Copy, MAT_Convert, MAT_Scale, MAT_AssemblyBegin;
20: PetscLogEvent MAT_AssemblyEnd, MAT_SetValues, MAT_GetValues, MAT_GetRow, MAT_GetRowIJ, MAT_CreateSubMats, MAT_GetOrdering, MAT_RedundantMat, MAT_GetSeqNonzeroStructure;
21: PetscLogEvent MAT_IncreaseOverlap, MAT_Partitioning, MAT_PartitioningND, MAT_Coarsen, MAT_ZeroEntries, MAT_Load, MAT_View, MAT_AXPY, MAT_FDColoringCreate;
22: PetscLogEvent MAT_FDColoringSetUp, MAT_FDColoringApply,MAT_Transpose,MAT_FDColoringFunction, MAT_CreateSubMat;
23: PetscLogEvent MAT_TransposeColoringCreate;
24: PetscLogEvent MAT_MatMult, MAT_MatMultSymbolic, MAT_MatMultNumeric;
25: PetscLogEvent MAT_PtAP, MAT_PtAPSymbolic, MAT_PtAPNumeric,MAT_RARt, MAT_RARtSymbolic, MAT_RARtNumeric;
26: PetscLogEvent MAT_MatTransposeMult, MAT_MatTransposeMultSymbolic, MAT_MatTransposeMultNumeric;
27: PetscLogEvent MAT_TransposeMatMult, MAT_TransposeMatMultSymbolic, MAT_TransposeMatMultNumeric;
28: PetscLogEvent MAT_MatMatMult, MAT_MatMatMultSymbolic, MAT_MatMatMultNumeric;
29: PetscLogEvent MAT_MultHermitianTranspose,MAT_MultHermitianTransposeAdd;
30: PetscLogEvent MAT_Getsymtranspose, MAT_Getsymtransreduced, MAT_GetBrowsOfAcols;
31: PetscLogEvent MAT_GetBrowsOfAocols, MAT_Getlocalmat, MAT_Getlocalmatcondensed, MAT_Seqstompi, MAT_Seqstompinum, MAT_Seqstompisym;
32: PetscLogEvent MAT_Applypapt, MAT_Applypapt_numeric, MAT_Applypapt_symbolic, MAT_GetSequentialNonzeroStructure;
33: PetscLogEvent MAT_GetMultiProcBlock;
34: PetscLogEvent MAT_CUSPARSECopyToGPU, MAT_CUSPARSEGenerateTranspose, MAT_SetValuesBatch;
35: PetscLogEvent MAT_ViennaCLCopyToGPU;
36: PetscLogEvent MAT_DenseCopyToGPU, MAT_DenseCopyFromGPU;
37: PetscLogEvent MAT_Merge,MAT_Residual,MAT_SetRandom;
38: PetscLogEvent MAT_FactorFactS,MAT_FactorInvS;
39: PetscLogEvent MATCOLORING_Apply,MATCOLORING_Comm,MATCOLORING_Local,MATCOLORING_ISCreate,MATCOLORING_SetUp,MATCOLORING_Weights;
41: const char *const MatFactorTypes[] = {"NONE","LU","CHOLESKY","ILU","ICC","ILUDT","MatFactorType","MAT_FACTOR_",NULL};
43: /*@
44: MatSetRandom - Sets all components of a matrix to random numbers. For sparse matrices that have been preallocated but not been assembled it randomly selects appropriate locations,
45: for sparse matrices that already have locations it fills the locations with random numbers
47: Logically Collective on Mat
49: Input Parameters:
50: + x - the matrix
51: - rctx - the random number context, formed by PetscRandomCreate(), or NULL and
52: it will create one internally.
54: Output Parameter:
55: . x - the matrix
57: Example of Usage:
58: .vb
59: PetscRandomCreate(PETSC_COMM_WORLD,&rctx);
60: MatSetRandom(x,rctx);
61: PetscRandomDestroy(rctx);
62: .ve
64: Level: intermediate
67: .seealso: MatZeroEntries(), MatSetValues(), PetscRandomCreate(), PetscRandomDestroy()
68: @*/
69: PetscErrorCode MatSetRandom(Mat x,PetscRandom rctx)
70: {
72: PetscRandom randObj = NULL;
79: if (!x->ops->setrandom) SETERRQ1(PetscObjectComm((PetscObject)x),PETSC_ERR_SUP,"Mat type %s",((PetscObject)x)->type_name);
81: if (!rctx) {
82: MPI_Comm comm;
83: PetscObjectGetComm((PetscObject)x,&comm);
84: PetscRandomCreate(comm,&randObj);
85: PetscRandomSetFromOptions(randObj);
86: rctx = randObj;
87: }
89: PetscLogEventBegin(MAT_SetRandom,x,rctx,0,0);
90: (*x->ops->setrandom)(x,rctx);
91: PetscLogEventEnd(MAT_SetRandom,x,rctx,0,0);
93: MatAssemblyBegin(x, MAT_FINAL_ASSEMBLY);
94: MatAssemblyEnd(x, MAT_FINAL_ASSEMBLY);
95: PetscRandomDestroy(&randObj);
96: return(0);
97: }
99: /*@
100: MatFactorGetErrorZeroPivot - returns the pivot value that was determined to be zero and the row it occurred in
102: Logically Collective on Mat
104: Input Parameters:
105: . mat - the factored matrix
107: Output Parameter:
108: + pivot - the pivot value computed
109: - row - the row that the zero pivot occurred. Note that this row must be interpreted carefully due to row reorderings and which processes
110: the share the matrix
112: Level: advanced
114: Notes:
115: This routine does not work for factorizations done with external packages.
117: This routine should only be called if MatGetFactorError() returns a value of MAT_FACTOR_NUMERIC_ZEROPIVOT
119: This can be called on non-factored matrices that come from, for example, matrices used in SOR.
121: .seealso: MatZeroEntries(), MatFactor(), MatGetFactor(), MatLUFactorSymbolic(), MatCholeskyFactorSymbolic(), MatFactorClearError(), MatFactorGetErrorZeroPivot()
122: @*/
123: PetscErrorCode MatFactorGetErrorZeroPivot(Mat mat,PetscReal *pivot,PetscInt *row)
124: {
127: *pivot = mat->factorerror_zeropivot_value;
128: *row = mat->factorerror_zeropivot_row;
129: return(0);
130: }
132: /*@
133: MatFactorGetError - gets the error code from a factorization
135: Logically Collective on Mat
137: Input Parameters:
138: . mat - the factored matrix
140: Output Parameter:
141: . err - the error code
143: Level: advanced
145: Notes:
146: This can be called on non-factored matrices that come from, for example, matrices used in SOR.
148: .seealso: MatZeroEntries(), MatFactor(), MatGetFactor(), MatLUFactorSymbolic(), MatCholeskyFactorSymbolic(), MatFactorClearError(), MatFactorGetErrorZeroPivot()
149: @*/
150: PetscErrorCode MatFactorGetError(Mat mat,MatFactorError *err)
151: {
154: *err = mat->factorerrortype;
155: return(0);
156: }
158: /*@
159: MatFactorClearError - clears the error code in a factorization
161: Logically Collective on Mat
163: Input Parameter:
164: . mat - the factored matrix
166: Level: developer
168: Notes:
169: This can be called on non-factored matrices that come from, for example, matrices used in SOR.
171: .seealso: MatZeroEntries(), MatFactor(), MatGetFactor(), MatLUFactorSymbolic(), MatCholeskyFactorSymbolic(), MatFactorGetError(), MatFactorGetErrorZeroPivot()
172: @*/
173: PetscErrorCode MatFactorClearError(Mat mat)
174: {
177: mat->factorerrortype = MAT_FACTOR_NOERROR;
178: mat->factorerror_zeropivot_value = 0.0;
179: mat->factorerror_zeropivot_row = 0;
180: return(0);
181: }
183: PETSC_INTERN PetscErrorCode MatFindNonzeroRowsOrCols_Basic(Mat mat,PetscBool cols,PetscReal tol,IS *nonzero)
184: {
185: PetscErrorCode ierr;
186: Vec r,l;
187: const PetscScalar *al;
188: PetscInt i,nz,gnz,N,n;
191: MatCreateVecs(mat,&r,&l);
192: if (!cols) { /* nonzero rows */
193: MatGetSize(mat,&N,NULL);
194: MatGetLocalSize(mat,&n,NULL);
195: VecSet(l,0.0);
196: VecSetRandom(r,NULL);
197: MatMult(mat,r,l);
198: VecGetArrayRead(l,&al);
199: } else { /* nonzero columns */
200: MatGetSize(mat,NULL,&N);
201: MatGetLocalSize(mat,NULL,&n);
202: VecSet(r,0.0);
203: VecSetRandom(l,NULL);
204: MatMultTranspose(mat,l,r);
205: VecGetArrayRead(r,&al);
206: }
207: if (tol <= 0.0) { for (i=0,nz=0;i<n;i++) if (al[i] != 0.0) nz++; }
208: else { for (i=0,nz=0;i<n;i++) if (PetscAbsScalar(al[i]) > tol) nz++; }
209: MPIU_Allreduce(&nz,&gnz,1,MPIU_INT,MPI_SUM,PetscObjectComm((PetscObject)mat));
210: if (gnz != N) {
211: PetscInt *nzr;
212: PetscMalloc1(nz,&nzr);
213: if (nz) {
214: if (tol < 0) { for (i=0,nz=0;i<n;i++) if (al[i] != 0.0) nzr[nz++] = i; }
215: else { for (i=0,nz=0;i<n;i++) if (PetscAbsScalar(al[i]) > tol) nzr[nz++] = i; }
216: }
217: ISCreateGeneral(PetscObjectComm((PetscObject)mat),nz,nzr,PETSC_OWN_POINTER,nonzero);
218: } else *nonzero = NULL;
219: if (!cols) { /* nonzero rows */
220: VecRestoreArrayRead(l,&al);
221: } else {
222: VecRestoreArrayRead(r,&al);
223: }
224: VecDestroy(&l);
225: VecDestroy(&r);
226: return(0);
227: }
229: /*@
230: MatFindNonzeroRows - Locate all rows that are not completely zero in the matrix
232: Input Parameter:
233: . A - the matrix
235: Output Parameter:
236: . keptrows - the rows that are not completely zero
238: Notes:
239: keptrows is set to NULL if all rows are nonzero.
241: Level: intermediate
243: @*/
244: PetscErrorCode MatFindNonzeroRows(Mat mat,IS *keptrows)
245: {
252: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
253: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
254: if (!mat->ops->findnonzerorows) {
255: MatFindNonzeroRowsOrCols_Basic(mat,PETSC_FALSE,0.0,keptrows);
256: } else {
257: (*mat->ops->findnonzerorows)(mat,keptrows);
258: }
259: return(0);
260: }
262: /*@
263: MatFindZeroRows - Locate all rows that are completely zero in the matrix
265: Input Parameter:
266: . A - the matrix
268: Output Parameter:
269: . zerorows - the rows that are completely zero
271: Notes:
272: zerorows is set to NULL if no rows are zero.
274: Level: intermediate
276: @*/
277: PetscErrorCode MatFindZeroRows(Mat mat,IS *zerorows)
278: {
280: IS keptrows;
281: PetscInt m, n;
286: MatFindNonzeroRows(mat, &keptrows);
287: /* MatFindNonzeroRows sets keptrows to NULL if there are no zero rows.
288: In keeping with this convention, we set zerorows to NULL if there are no zero
289: rows. */
290: if (keptrows == NULL) {
291: *zerorows = NULL;
292: } else {
293: MatGetOwnershipRange(mat,&m,&n);
294: ISComplement(keptrows,m,n,zerorows);
295: ISDestroy(&keptrows);
296: }
297: return(0);
298: }
300: /*@
301: MatGetDiagonalBlock - Returns the part of the matrix associated with the on-process coupling
303: Not Collective
305: Input Parameters:
306: . A - the matrix
308: Output Parameters:
309: . a - the diagonal part (which is a SEQUENTIAL matrix)
311: Notes:
312: see the manual page for MatCreateAIJ() for more information on the "diagonal part" of the matrix.
313: Use caution, as the reference count on the returned matrix is not incremented and it is used as
314: part of the containing MPI Mat's normal operation.
316: Level: advanced
318: @*/
319: PetscErrorCode MatGetDiagonalBlock(Mat A,Mat *a)
320: {
327: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
328: if (!A->ops->getdiagonalblock) {
329: PetscMPIInt size;
330: MPI_Comm_size(PetscObjectComm((PetscObject)A),&size);
331: if (size == 1) {
332: *a = A;
333: return(0);
334: } else SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Not coded for matrix type %s",((PetscObject)A)->type_name);
335: }
336: (*A->ops->getdiagonalblock)(A,a);
337: return(0);
338: }
340: /*@
341: MatGetTrace - Gets the trace of a matrix. The sum of the diagonal entries.
343: Collective on Mat
345: Input Parameters:
346: . mat - the matrix
348: Output Parameter:
349: . trace - the sum of the diagonal entries
351: Level: advanced
353: @*/
354: PetscErrorCode MatGetTrace(Mat mat,PetscScalar *trace)
355: {
357: Vec diag;
360: MatCreateVecs(mat,&diag,NULL);
361: MatGetDiagonal(mat,diag);
362: VecSum(diag,trace);
363: VecDestroy(&diag);
364: return(0);
365: }
367: /*@
368: MatRealPart - Zeros out the imaginary part of the matrix
370: Logically Collective on Mat
372: Input Parameters:
373: . mat - the matrix
375: Level: advanced
378: .seealso: MatImaginaryPart()
379: @*/
380: PetscErrorCode MatRealPart(Mat mat)
381: {
387: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
388: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
389: if (!mat->ops->realpart) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
390: MatCheckPreallocated(mat,1);
391: (*mat->ops->realpart)(mat);
392: return(0);
393: }
395: /*@C
396: MatGetGhosts - Get the global index of all ghost nodes defined by the sparse matrix
398: Collective on Mat
400: Input Parameter:
401: . mat - the matrix
403: Output Parameters:
404: + nghosts - number of ghosts (note for BAIJ matrices there is one ghost for each block)
405: - ghosts - the global indices of the ghost points
407: Notes:
408: the nghosts and ghosts are suitable to pass into VecCreateGhost()
410: Level: advanced
412: @*/
413: PetscErrorCode MatGetGhosts(Mat mat,PetscInt *nghosts,const PetscInt *ghosts[])
414: {
420: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
421: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
422: if (!mat->ops->getghosts) {
423: if (nghosts) *nghosts = 0;
424: if (ghosts) *ghosts = NULL;
425: } else {
426: (*mat->ops->getghosts)(mat,nghosts,ghosts);
427: }
428: return(0);
429: }
432: /*@
433: MatImaginaryPart - Moves the imaginary part of the matrix to the real part and zeros the imaginary part
435: Logically Collective on Mat
437: Input Parameters:
438: . mat - the matrix
440: Level: advanced
443: .seealso: MatRealPart()
444: @*/
445: PetscErrorCode MatImaginaryPart(Mat mat)
446: {
452: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
453: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
454: if (!mat->ops->imaginarypart) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
455: MatCheckPreallocated(mat,1);
456: (*mat->ops->imaginarypart)(mat);
457: return(0);
458: }
460: /*@
461: MatMissingDiagonal - Determine if sparse matrix is missing a diagonal entry (or block entry for BAIJ matrices)
463: Not Collective
465: Input Parameter:
466: . mat - the matrix
468: Output Parameters:
469: + missing - is any diagonal missing
470: - dd - first diagonal entry that is missing (optional) on this process
472: Level: advanced
475: .seealso: MatRealPart()
476: @*/
477: PetscErrorCode MatMissingDiagonal(Mat mat,PetscBool *missing,PetscInt *dd)
478: {
485: if (!mat->assembled) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix %s",((PetscObject)mat)->type_name);
486: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
487: if (!mat->ops->missingdiagonal) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
488: (*mat->ops->missingdiagonal)(mat,missing,dd);
489: return(0);
490: }
492: /*@C
493: MatGetRow - Gets a row of a matrix. You MUST call MatRestoreRow()
494: for each row that you get to ensure that your application does
495: not bleed memory.
497: Not Collective
499: Input Parameters:
500: + mat - the matrix
501: - row - the row to get
503: Output Parameters:
504: + ncols - if not NULL, the number of nonzeros in the row
505: . cols - if not NULL, the column numbers
506: - vals - if not NULL, the values
508: Notes:
509: This routine is provided for people who need to have direct access
510: to the structure of a matrix. We hope that we provide enough
511: high-level matrix routines that few users will need it.
513: MatGetRow() always returns 0-based column indices, regardless of
514: whether the internal representation is 0-based (default) or 1-based.
516: For better efficiency, set cols and/or vals to NULL if you do
517: not wish to extract these quantities.
519: The user can only examine the values extracted with MatGetRow();
520: the values cannot be altered. To change the matrix entries, one
521: must use MatSetValues().
523: You can only have one call to MatGetRow() outstanding for a particular
524: matrix at a time, per processor. MatGetRow() can only obtain rows
525: associated with the given processor, it cannot get rows from the
526: other processors; for that we suggest using MatCreateSubMatrices(), then
527: MatGetRow() on the submatrix. The row index passed to MatGetRow()
528: is in the global number of rows.
530: Fortran Notes:
531: The calling sequence from Fortran is
532: .vb
533: MatGetRow(matrix,row,ncols,cols,values,ierr)
534: Mat matrix (input)
535: integer row (input)
536: integer ncols (output)
537: integer cols(maxcols) (output)
538: double precision (or double complex) values(maxcols) output
539: .ve
540: where maxcols >= maximum nonzeros in any row of the matrix.
543: Caution:
544: Do not try to change the contents of the output arrays (cols and vals).
545: In some cases, this may corrupt the matrix.
547: Level: advanced
549: .seealso: MatRestoreRow(), MatSetValues(), MatGetValues(), MatCreateSubMatrices(), MatGetDiagonal()
550: @*/
551: PetscErrorCode MatGetRow(Mat mat,PetscInt row,PetscInt *ncols,const PetscInt *cols[],const PetscScalar *vals[])
552: {
554: PetscInt incols;
559: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
560: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
561: if (!mat->ops->getrow) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
562: MatCheckPreallocated(mat,1);
563: PetscLogEventBegin(MAT_GetRow,mat,0,0,0);
564: (*mat->ops->getrow)(mat,row,&incols,(PetscInt**)cols,(PetscScalar**)vals);
565: if (ncols) *ncols = incols;
566: PetscLogEventEnd(MAT_GetRow,mat,0,0,0);
567: return(0);
568: }
570: /*@
571: MatConjugate - replaces the matrix values with their complex conjugates
573: Logically Collective on Mat
575: Input Parameters:
576: . mat - the matrix
578: Level: advanced
580: .seealso: VecConjugate()
581: @*/
582: PetscErrorCode MatConjugate(Mat mat)
583: {
584: #if defined(PETSC_USE_COMPLEX)
589: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
590: if (!mat->ops->conjugate) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Not provided for matrix type %s, send email to petsc-maint@mcs.anl.gov",((PetscObject)mat)->type_name);
591: (*mat->ops->conjugate)(mat);
592: #else
594: #endif
595: return(0);
596: }
598: /*@C
599: MatRestoreRow - Frees any temporary space allocated by MatGetRow().
601: Not Collective
603: Input Parameters:
604: + mat - the matrix
605: . row - the row to get
606: . ncols, cols - the number of nonzeros and their columns
607: - vals - if nonzero the column values
609: Notes:
610: This routine should be called after you have finished examining the entries.
612: This routine zeros out ncols, cols, and vals. This is to prevent accidental
613: us of the array after it has been restored. If you pass NULL, it will
614: not zero the pointers. Use of cols or vals after MatRestoreRow is invalid.
616: Fortran Notes:
617: The calling sequence from Fortran is
618: .vb
619: MatRestoreRow(matrix,row,ncols,cols,values,ierr)
620: Mat matrix (input)
621: integer row (input)
622: integer ncols (output)
623: integer cols(maxcols) (output)
624: double precision (or double complex) values(maxcols) output
625: .ve
626: Where maxcols >= maximum nonzeros in any row of the matrix.
628: In Fortran MatRestoreRow() MUST be called after MatGetRow()
629: before another call to MatGetRow() can be made.
631: Level: advanced
633: .seealso: MatGetRow()
634: @*/
635: PetscErrorCode MatRestoreRow(Mat mat,PetscInt row,PetscInt *ncols,const PetscInt *cols[],const PetscScalar *vals[])
636: {
642: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
643: if (!mat->ops->restorerow) return(0);
644: (*mat->ops->restorerow)(mat,row,ncols,(PetscInt **)cols,(PetscScalar **)vals);
645: if (ncols) *ncols = 0;
646: if (cols) *cols = NULL;
647: if (vals) *vals = NULL;
648: return(0);
649: }
651: /*@
652: MatGetRowUpperTriangular - Sets a flag to enable calls to MatGetRow() for matrix in MATSBAIJ format.
653: You should call MatRestoreRowUpperTriangular() after calling MatGetRow/MatRestoreRow() to disable the flag.
655: Not Collective
657: Input Parameters:
658: . mat - the matrix
660: Notes:
661: The flag is to ensure that users are aware of MatGetRow() only provides the upper triangular part of the row for the matrices in MATSBAIJ format.
663: Level: advanced
665: .seealso: MatRestoreRowUpperTriangular()
666: @*/
667: PetscErrorCode MatGetRowUpperTriangular(Mat mat)
668: {
674: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
675: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
676: MatCheckPreallocated(mat,1);
677: if (!mat->ops->getrowuppertriangular) return(0);
678: (*mat->ops->getrowuppertriangular)(mat);
679: return(0);
680: }
682: /*@
683: MatRestoreRowUpperTriangular - Disable calls to MatGetRow() for matrix in MATSBAIJ format.
685: Not Collective
687: Input Parameters:
688: . mat - the matrix
690: Notes:
691: This routine should be called after you have finished MatGetRow/MatRestoreRow().
694: Level: advanced
696: .seealso: MatGetRowUpperTriangular()
697: @*/
698: PetscErrorCode MatRestoreRowUpperTriangular(Mat mat)
699: {
705: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
706: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
707: MatCheckPreallocated(mat,1);
708: if (!mat->ops->restorerowuppertriangular) return(0);
709: (*mat->ops->restorerowuppertriangular)(mat);
710: return(0);
711: }
713: /*@C
714: MatSetOptionsPrefix - Sets the prefix used for searching for all
715: Mat options in the database.
717: Logically Collective on Mat
719: Input Parameter:
720: + A - the Mat context
721: - prefix - the prefix to prepend to all option names
723: Notes:
724: A hyphen (-) must NOT be given at the beginning of the prefix name.
725: The first character of all runtime options is AUTOMATICALLY the hyphen.
727: Level: advanced
729: .seealso: MatSetFromOptions()
730: @*/
731: PetscErrorCode MatSetOptionsPrefix(Mat A,const char prefix[])
732: {
737: PetscObjectSetOptionsPrefix((PetscObject)A,prefix);
738: return(0);
739: }
741: /*@C
742: MatAppendOptionsPrefix - Appends to the prefix used for searching for all
743: Mat options in the database.
745: Logically Collective on Mat
747: Input Parameters:
748: + A - the Mat context
749: - prefix - the prefix to prepend to all option names
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: Level: advanced
757: .seealso: MatGetOptionsPrefix()
758: @*/
759: PetscErrorCode MatAppendOptionsPrefix(Mat A,const char prefix[])
760: {
765: PetscObjectAppendOptionsPrefix((PetscObject)A,prefix);
766: return(0);
767: }
769: /*@C
770: MatGetOptionsPrefix - Gets the prefix used for searching for all
771: Mat options in the database.
773: Not Collective
775: Input Parameter:
776: . A - the Mat context
778: Output Parameter:
779: . prefix - pointer to the prefix string used
781: Notes:
782: On the fortran side, the user should pass in a string 'prefix' of
783: sufficient length to hold the prefix.
785: Level: advanced
787: .seealso: MatAppendOptionsPrefix()
788: @*/
789: PetscErrorCode MatGetOptionsPrefix(Mat A,const char *prefix[])
790: {
795: PetscObjectGetOptionsPrefix((PetscObject)A,prefix);
796: return(0);
797: }
799: /*@
800: MatResetPreallocation - Reset mat to use the original nonzero pattern provided by users.
802: Collective on Mat
804: Input Parameters:
805: . A - the Mat context
807: Notes:
808: The allocated memory will be shrunk after calling MatAssembly with MAT_FINAL_ASSEMBLY. Users can reset the preallocation to access the original memory.
809: Currently support MPIAIJ and SEQAIJ.
811: Level: beginner
813: .seealso: MatSeqAIJSetPreallocation(), MatMPIAIJSetPreallocation(), MatXAIJSetPreallocation()
814: @*/
815: PetscErrorCode MatResetPreallocation(Mat A)
816: {
822: PetscUseMethod(A,"MatResetPreallocation_C",(Mat),(A));
823: return(0);
824: }
827: /*@
828: MatSetUp - Sets up the internal matrix data structures for later use.
830: Collective on Mat
832: Input Parameters:
833: . A - the Mat context
835: Notes:
836: If the user has not set preallocation for this matrix then a default preallocation that is likely to be inefficient is used.
838: If a suitable preallocation routine is used, this function does not need to be called.
840: See the Performance chapter of the PETSc users manual for how to preallocate matrices
842: Level: beginner
844: .seealso: MatCreate(), MatDestroy()
845: @*/
846: PetscErrorCode MatSetUp(Mat A)
847: {
848: PetscMPIInt size;
853: if (!((PetscObject)A)->type_name) {
854: MPI_Comm_size(PetscObjectComm((PetscObject)A), &size);
855: if (size == 1) {
856: MatSetType(A, MATSEQAIJ);
857: } else {
858: MatSetType(A, MATMPIAIJ);
859: }
860: }
861: if (!A->preallocated && A->ops->setup) {
862: PetscInfo(A,"Warning not preallocating matrix storage\n");
863: (*A->ops->setup)(A);
864: }
865: PetscLayoutSetUp(A->rmap);
866: PetscLayoutSetUp(A->cmap);
867: A->preallocated = PETSC_TRUE;
868: return(0);
869: }
871: #if defined(PETSC_HAVE_SAWS)
872: #include <petscviewersaws.h>
873: #endif
875: /*@C
876: MatViewFromOptions - View from Options
878: Collective on Mat
880: Input Parameters:
881: + A - the Mat context
882: . obj - Optional object
883: - name - command line option
885: Level: intermediate
886: .seealso: Mat, MatView, PetscObjectViewFromOptions(), MatCreate()
887: @*/
888: PetscErrorCode MatViewFromOptions(Mat A,PetscObject obj,const char name[])
889: {
894: PetscObjectViewFromOptions((PetscObject)A,obj,name);
895: return(0);
896: }
898: /*@C
899: MatView - Visualizes a matrix object.
901: Collective on Mat
903: Input Parameters:
904: + mat - the matrix
905: - viewer - visualization context
907: Notes:
908: The available visualization contexts include
909: + PETSC_VIEWER_STDOUT_SELF - for sequential matrices
910: . PETSC_VIEWER_STDOUT_WORLD - for parallel matrices created on PETSC_COMM_WORLD
911: . PETSC_VIEWER_STDOUT_(comm) - for matrices created on MPI communicator comm
912: - PETSC_VIEWER_DRAW_WORLD - graphical display of nonzero structure
914: The user can open alternative visualization contexts with
915: + PetscViewerASCIIOpen() - Outputs matrix to a specified file
916: . PetscViewerBinaryOpen() - Outputs matrix in binary to a
917: specified file; corresponding input uses MatLoad()
918: . PetscViewerDrawOpen() - Outputs nonzero matrix structure to
919: an X window display
920: - PetscViewerSocketOpen() - Outputs matrix to Socket viewer.
921: Currently only the sequential dense and AIJ
922: matrix types support the Socket viewer.
924: The user can call PetscViewerPushFormat() to specify the output
925: format of ASCII printed objects (when using PETSC_VIEWER_STDOUT_SELF,
926: PETSC_VIEWER_STDOUT_WORLD and PetscViewerASCIIOpen). Available formats include
927: + PETSC_VIEWER_DEFAULT - default, prints matrix contents
928: . PETSC_VIEWER_ASCII_MATLAB - prints matrix contents in Matlab format
929: . PETSC_VIEWER_ASCII_DENSE - prints entire matrix including zeros
930: . PETSC_VIEWER_ASCII_COMMON - prints matrix contents, using a sparse
931: format common among all matrix types
932: . PETSC_VIEWER_ASCII_IMPL - prints matrix contents, using an implementation-specific
933: format (which is in many cases the same as the default)
934: . PETSC_VIEWER_ASCII_INFO - prints basic information about the matrix
935: size and structure (not the matrix entries)
936: - PETSC_VIEWER_ASCII_INFO_DETAIL - prints more detailed information about
937: the matrix structure
939: Options Database Keys:
940: + -mat_view ::ascii_info - Prints info on matrix at conclusion of MatAssemblyEnd()
941: . -mat_view ::ascii_info_detail - Prints more detailed info
942: . -mat_view - Prints matrix in ASCII format
943: . -mat_view ::ascii_matlab - Prints matrix in Matlab format
944: . -mat_view draw - PetscDraws nonzero structure of matrix, using MatView() and PetscDrawOpenX().
945: . -display <name> - Sets display name (default is host)
946: . -draw_pause <sec> - Sets number of seconds to pause after display
947: . -mat_view socket - Sends matrix to socket, can be accessed from Matlab (see Users-Manual: ch_matlab for details)
948: . -viewer_socket_machine <machine> -
949: . -viewer_socket_port <port> -
950: . -mat_view binary - save matrix to file in binary format
951: - -viewer_binary_filename <name> -
952: Level: beginner
954: Notes:
955: The ASCII viewers are only recommended for small matrices on at most a moderate number of processes,
956: the program will seemingly hang and take hours for larger matrices, for larger matrices one should use the binary format.
958: In the debugger you can do "call MatView(mat,0)" to display the matrix. (The same holds for any PETSc object viewer).
960: See the manual page for MatLoad() for the exact format of the binary file when the binary
961: viewer is used.
963: See share/petsc/matlab/PetscBinaryRead.m for a Matlab code that can read in the binary file when the binary
964: viewer is used and lib/petsc/bin/PetscBinaryIO.py for loading them into Python.
966: One can use '-mat_view draw -draw_pause -1' to pause the graphical display of matrix nonzero structure,
967: and then use the following mouse functions.
968: + left mouse: zoom in
969: . middle mouse: zoom out
970: - right mouse: continue with the simulation
972: .seealso: PetscViewerPushFormat(), PetscViewerASCIIOpen(), PetscViewerDrawOpen(),
973: PetscViewerSocketOpen(), PetscViewerBinaryOpen(), MatLoad()
974: @*/
975: PetscErrorCode MatView(Mat mat,PetscViewer viewer)
976: {
977: PetscErrorCode ierr;
978: PetscInt rows,cols,rbs,cbs;
979: PetscBool isascii,isstring,issaws;
980: PetscViewerFormat format;
981: PetscMPIInt size;
986: if (!viewer) {PetscViewerASCIIGetStdout(PetscObjectComm((PetscObject)mat),&viewer);}
989: MatCheckPreallocated(mat,1);
991: PetscViewerGetFormat(viewer,&format);
992: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
993: if (size == 1 && format == PETSC_VIEWER_LOAD_BALANCE) return(0);
995: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERSTRING,&isstring);
996: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&isascii);
997: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERSAWS,&issaws);
998: if ((!isascii || (format != PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL)) && mat->factortype) {
999: SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"No viewers for factored matrix except ASCII info or info_detail");
1000: }
1002: PetscLogEventBegin(MAT_View,mat,viewer,0,0);
1003: if (isascii) {
1004: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ORDER,"Must call MatAssemblyBegin/End() before viewing matrix");
1005: PetscObjectPrintClassNamePrefixType((PetscObject)mat,viewer);
1006: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
1007: MatNullSpace nullsp,transnullsp;
1009: PetscViewerASCIIPushTab(viewer);
1010: MatGetSize(mat,&rows,&cols);
1011: MatGetBlockSizes(mat,&rbs,&cbs);
1012: if (rbs != 1 || cbs != 1) {
1013: if (rbs != cbs) {PetscViewerASCIIPrintf(viewer,"rows=%D, cols=%D, rbs=%D, cbs=%D\n",rows,cols,rbs,cbs);}
1014: else {PetscViewerASCIIPrintf(viewer,"rows=%D, cols=%D, bs=%D\n",rows,cols,rbs);}
1015: } else {
1016: PetscViewerASCIIPrintf(viewer,"rows=%D, cols=%D\n",rows,cols);
1017: }
1018: if (mat->factortype) {
1019: MatSolverType solver;
1020: MatFactorGetSolverType(mat,&solver);
1021: PetscViewerASCIIPrintf(viewer,"package used to perform factorization: %s\n",solver);
1022: }
1023: if (mat->ops->getinfo) {
1024: MatInfo info;
1025: MatGetInfo(mat,MAT_GLOBAL_SUM,&info);
1026: PetscViewerASCIIPrintf(viewer,"total: nonzeros=%.f, allocated nonzeros=%.f\n",info.nz_used,info.nz_allocated);
1027: if (!mat->factortype) {
1028: PetscViewerASCIIPrintf(viewer,"total number of mallocs used during MatSetValues calls=%D\n",(PetscInt)info.mallocs);
1029: }
1030: }
1031: MatGetNullSpace(mat,&nullsp);
1032: MatGetTransposeNullSpace(mat,&transnullsp);
1033: if (nullsp) {PetscViewerASCIIPrintf(viewer," has attached null space\n");}
1034: if (transnullsp && transnullsp != nullsp) {PetscViewerASCIIPrintf(viewer," has attached transposed null space\n");}
1035: MatGetNearNullSpace(mat,&nullsp);
1036: if (nullsp) {PetscViewerASCIIPrintf(viewer," has attached near null space\n");}
1037: PetscViewerASCIIPushTab(viewer);
1038: MatProductView(mat,viewer);
1039: PetscViewerASCIIPopTab(viewer);
1040: }
1041: } else if (issaws) {
1042: #if defined(PETSC_HAVE_SAWS)
1043: PetscMPIInt rank;
1045: PetscObjectName((PetscObject)mat);
1046: MPI_Comm_rank(PETSC_COMM_WORLD,&rank);
1047: if (!((PetscObject)mat)->amsmem && !rank) {
1048: PetscObjectViewSAWs((PetscObject)mat,viewer);
1049: }
1050: #endif
1051: } else if (isstring) {
1052: const char *type;
1053: MatGetType(mat,&type);
1054: PetscViewerStringSPrintf(viewer," MatType: %-7.7s",type);
1055: if (mat->ops->view) {(*mat->ops->view)(mat,viewer);}
1056: }
1057: if ((format == PETSC_VIEWER_NATIVE || format == PETSC_VIEWER_LOAD_BALANCE) && mat->ops->viewnative) {
1058: PetscViewerASCIIPushTab(viewer);
1059: (*mat->ops->viewnative)(mat,viewer);
1060: PetscViewerASCIIPopTab(viewer);
1061: } else if (mat->ops->view) {
1062: PetscViewerASCIIPushTab(viewer);
1063: (*mat->ops->view)(mat,viewer);
1064: PetscViewerASCIIPopTab(viewer);
1065: }
1066: if (isascii) {
1067: PetscViewerGetFormat(viewer,&format);
1068: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
1069: PetscViewerASCIIPopTab(viewer);
1070: }
1071: }
1072: PetscLogEventEnd(MAT_View,mat,viewer,0,0);
1073: return(0);
1074: }
1076: #if defined(PETSC_USE_DEBUG)
1077: #include <../src/sys/totalview/tv_data_display.h>
1078: PETSC_UNUSED static int TV_display_type(const struct _p_Mat *mat)
1079: {
1080: TV_add_row("Local rows", "int", &mat->rmap->n);
1081: TV_add_row("Local columns", "int", &mat->cmap->n);
1082: TV_add_row("Global rows", "int", &mat->rmap->N);
1083: TV_add_row("Global columns", "int", &mat->cmap->N);
1084: TV_add_row("Typename", TV_ascii_string_type, ((PetscObject)mat)->type_name);
1085: return TV_format_OK;
1086: }
1087: #endif
1089: /*@C
1090: MatLoad - Loads a matrix that has been stored in binary/HDF5 format
1091: with MatView(). The matrix format is determined from the options database.
1092: Generates a parallel MPI matrix if the communicator has more than one
1093: processor. The default matrix type is AIJ.
1095: Collective on PetscViewer
1097: Input Parameters:
1098: + mat - the newly loaded matrix, this needs to have been created with MatCreate()
1099: or some related function before a call to MatLoad()
1100: - viewer - binary/HDF5 file viewer
1102: Options Database Keys:
1103: Used with block matrix formats (MATSEQBAIJ, ...) to specify
1104: block size
1105: . -matload_block_size <bs>
1107: Level: beginner
1109: Notes:
1110: If the Mat type has not yet been given then MATAIJ is used, call MatSetFromOptions() on the
1111: Mat before calling this routine if you wish to set it from the options database.
1113: MatLoad() automatically loads into the options database any options
1114: given in the file filename.info where filename is the name of the file
1115: that was passed to the PetscViewerBinaryOpen(). The options in the info
1116: file will be ignored if you use the -viewer_binary_skip_info option.
1118: If the type or size of mat is not set before a call to MatLoad, PETSc
1119: sets the default matrix type AIJ and sets the local and global sizes.
1120: If type and/or size is already set, then the same are used.
1122: In parallel, each processor can load a subset of rows (or the
1123: entire matrix). This routine is especially useful when a large
1124: matrix is stored on disk and only part of it is desired on each
1125: processor. For example, a parallel solver may access only some of
1126: the rows from each processor. The algorithm used here reads
1127: relatively small blocks of data rather than reading the entire
1128: matrix and then subsetting it.
1130: Viewer's PetscViewerType must be either PETSCVIEWERBINARY or PETSCVIEWERHDF5.
1131: Such viewer can be created using PetscViewerBinaryOpen()/PetscViewerHDF5Open(),
1132: or the sequence like
1133: $ PetscViewer v;
1134: $ PetscViewerCreate(PETSC_COMM_WORLD,&v);
1135: $ PetscViewerSetType(v,PETSCVIEWERBINARY);
1136: $ PetscViewerSetFromOptions(v);
1137: $ PetscViewerFileSetMode(v,FILE_MODE_READ);
1138: $ PetscViewerFileSetName(v,"datafile");
1139: The optional PetscViewerSetFromOptions() call allows to override PetscViewerSetType() using option
1140: $ -viewer_type {binary,hdf5}
1142: See the example src/ksp/ksp/tutorials/ex27.c with the first approach,
1143: and src/mat/tutorials/ex10.c with the second approach.
1145: Notes about the PETSc binary format:
1146: In case of PETSCVIEWERBINARY, a native PETSc binary format is used. Each of the blocks
1147: is read onto rank 0 and then shipped to its destination rank, one after another.
1148: Multiple objects, both matrices and vectors, can be stored within the same file.
1149: Their PetscObject name is ignored; they are loaded in the order of their storage.
1151: Most users should not need to know the details of the binary storage
1152: format, since MatLoad() and MatView() completely hide these details.
1153: But for anyone who's interested, the standard binary matrix storage
1154: format is
1156: $ PetscInt MAT_FILE_CLASSID
1157: $ PetscInt number of rows
1158: $ PetscInt number of columns
1159: $ PetscInt total number of nonzeros
1160: $ PetscInt *number nonzeros in each row
1161: $ PetscInt *column indices of all nonzeros (starting index is zero)
1162: $ PetscScalar *values of all nonzeros
1164: PETSc automatically does the byte swapping for
1165: machines that store the bytes reversed, e.g. DEC alpha, freebsd,
1166: linux, Windows and the paragon; thus if you write your own binary
1167: read/write routines you have to swap the bytes; see PetscBinaryRead()
1168: and PetscBinaryWrite() to see how this may be done.
1170: Notes about the HDF5 (MATLAB MAT-File Version 7.3) format:
1171: In case of PETSCVIEWERHDF5, a parallel HDF5 reader is used.
1172: Each processor's chunk is loaded independently by its owning rank.
1173: Multiple objects, both matrices and vectors, can be stored within the same file.
1174: They are looked up by their PetscObject name.
1176: As the MATLAB MAT-File Version 7.3 format is also a HDF5 flavor, we decided to use
1177: by default the same structure and naming of the AIJ arrays and column count
1178: within the HDF5 file. This means that a MAT file saved with -v7.3 flag, e.g.
1179: $ save example.mat A b -v7.3
1180: can be directly read by this routine (see Reference 1 for details).
1181: Note that depending on your MATLAB version, this format might be a default,
1182: otherwise you can set it as default in Preferences.
1184: Unless -nocompression flag is used to save the file in MATLAB,
1185: PETSc must be configured with ZLIB package.
1187: See also examples src/mat/tutorials/ex10.c and src/ksp/ksp/tutorials/ex27.c
1189: Current HDF5 (MAT-File) limitations:
1190: This reader currently supports only real MATSEQAIJ, MATMPIAIJ, MATSEQDENSE and MATMPIDENSE matrices.
1192: Corresponding MatView() is not yet implemented.
1194: The loaded matrix is actually a transpose of the original one in MATLAB,
1195: unless you push PETSC_VIEWER_HDF5_MAT format (see examples above).
1196: With this format, matrix is automatically transposed by PETSc,
1197: unless the matrix is marked as SPD or symmetric
1198: (see MatSetOption(), MAT_SPD, MAT_SYMMETRIC).
1200: References:
1201: 1. MATLAB(R) Documentation, manual page of save(), https://www.mathworks.com/help/matlab/ref/save.html#btox10b-1-version
1203: .seealso: PetscViewerBinaryOpen(), PetscViewerSetType(), MatView(), VecLoad()
1205: @*/
1206: PetscErrorCode MatLoad(Mat mat,PetscViewer viewer)
1207: {
1209: PetscBool flg;
1215: if (!((PetscObject)mat)->type_name) {
1216: MatSetType(mat,MATAIJ);
1217: }
1219: flg = PETSC_FALSE;
1220: PetscOptionsGetBool(((PetscObject)mat)->options,((PetscObject)mat)->prefix,"-matload_symmetric",&flg,NULL);
1221: if (flg) {
1222: MatSetOption(mat,MAT_SYMMETRIC,PETSC_TRUE);
1223: MatSetOption(mat,MAT_SYMMETRY_ETERNAL,PETSC_TRUE);
1224: }
1225: flg = PETSC_FALSE;
1226: PetscOptionsGetBool(((PetscObject)mat)->options,((PetscObject)mat)->prefix,"-matload_spd",&flg,NULL);
1227: if (flg) {
1228: MatSetOption(mat,MAT_SPD,PETSC_TRUE);
1229: }
1231: if (!mat->ops->load) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"MatLoad is not supported for type %s",((PetscObject)mat)->type_name);
1232: PetscLogEventBegin(MAT_Load,mat,viewer,0,0);
1233: (*mat->ops->load)(mat,viewer);
1234: PetscLogEventEnd(MAT_Load,mat,viewer,0,0);
1235: return(0);
1236: }
1238: static PetscErrorCode MatDestroy_Redundant(Mat_Redundant **redundant)
1239: {
1241: Mat_Redundant *redund = *redundant;
1242: PetscInt i;
1245: if (redund){
1246: if (redund->matseq) { /* via MatCreateSubMatrices() */
1247: ISDestroy(&redund->isrow);
1248: ISDestroy(&redund->iscol);
1249: MatDestroySubMatrices(1,&redund->matseq);
1250: } else {
1251: PetscFree2(redund->send_rank,redund->recv_rank);
1252: PetscFree(redund->sbuf_j);
1253: PetscFree(redund->sbuf_a);
1254: for (i=0; i<redund->nrecvs; i++) {
1255: PetscFree(redund->rbuf_j[i]);
1256: PetscFree(redund->rbuf_a[i]);
1257: }
1258: PetscFree4(redund->sbuf_nz,redund->rbuf_nz,redund->rbuf_j,redund->rbuf_a);
1259: }
1261: if (redund->subcomm) {
1262: PetscCommDestroy(&redund->subcomm);
1263: }
1264: PetscFree(redund);
1265: }
1266: return(0);
1267: }
1269: /*@
1270: MatDestroy - Frees space taken by a matrix.
1272: Collective on Mat
1274: Input Parameter:
1275: . A - the matrix
1277: Level: beginner
1279: @*/
1280: PetscErrorCode MatDestroy(Mat *A)
1281: {
1285: if (!*A) return(0);
1287: if (--((PetscObject)(*A))->refct > 0) {*A = NULL; return(0);}
1289: /* if memory was published with SAWs then destroy it */
1290: PetscObjectSAWsViewOff((PetscObject)*A);
1291: if ((*A)->ops->destroy) {
1292: (*(*A)->ops->destroy)(*A);
1293: }
1295: PetscFree((*A)->defaultvectype);
1296: PetscFree((*A)->bsizes);
1297: PetscFree((*A)->solvertype);
1298: MatDestroy_Redundant(&(*A)->redundant);
1299: MatProductClear(*A);
1300: MatNullSpaceDestroy(&(*A)->nullsp);
1301: MatNullSpaceDestroy(&(*A)->transnullsp);
1302: MatNullSpaceDestroy(&(*A)->nearnullsp);
1303: MatDestroy(&(*A)->schur);
1304: PetscLayoutDestroy(&(*A)->rmap);
1305: PetscLayoutDestroy(&(*A)->cmap);
1306: PetscHeaderDestroy(A);
1307: return(0);
1308: }
1310: /*@C
1311: MatSetValues - Inserts or adds a block of values into a matrix.
1312: These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
1313: MUST be called after all calls to MatSetValues() have been completed.
1315: Not Collective
1317: Input Parameters:
1318: + mat - the matrix
1319: . v - a logically two-dimensional array of values
1320: . m, idxm - the number of rows and their global indices
1321: . n, idxn - the number of columns and their global indices
1322: - addv - either ADD_VALUES or INSERT_VALUES, where
1323: ADD_VALUES adds values to any existing entries, and
1324: INSERT_VALUES replaces existing entries with new values
1326: Notes:
1327: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call MatXXXXSetPreallocation() or
1328: MatSetUp() before using this routine
1330: By default the values, v, are row-oriented. See MatSetOption() for other options.
1332: Calls to MatSetValues() with the INSERT_VALUES and ADD_VALUES
1333: options cannot be mixed without intervening calls to the assembly
1334: routines.
1336: MatSetValues() uses 0-based row and column numbers in Fortran
1337: as well as in C.
1339: Negative indices may be passed in idxm and idxn, these rows and columns are
1340: simply ignored. This allows easily inserting element stiffness matrices
1341: with homogeneous Dirchlet boundary conditions that you don't want represented
1342: in the matrix.
1344: Efficiency Alert:
1345: The routine MatSetValuesBlocked() may offer much better efficiency
1346: for users of block sparse formats (MATSEQBAIJ and MATMPIBAIJ).
1348: Level: beginner
1350: Developer Notes:
1351: This is labeled with C so does not automatically generate Fortran stubs and interfaces
1352: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
1354: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
1355: InsertMode, INSERT_VALUES, ADD_VALUES
1356: @*/
1357: PetscErrorCode MatSetValues(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],const PetscScalar v[],InsertMode addv)
1358: {
1364: if (!m || !n) return(0); /* no values to insert */
1367: MatCheckPreallocated(mat,1);
1369: if (mat->insertmode == NOT_SET_VALUES) {
1370: mat->insertmode = addv;
1371: } else if (PetscUnlikely(mat->insertmode != addv)) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
1372: if (PetscDefined(USE_DEBUG)) {
1373: PetscInt i,j;
1375: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1376: if (!mat->ops->setvalues) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1378: for (i=0; i<m; i++) {
1379: for (j=0; j<n; j++) {
1380: if (mat->erroriffailure && PetscIsInfOrNanScalar(v[i*n+j]))
1381: #if defined(PETSC_USE_COMPLEX)
1382: SETERRQ4(PETSC_COMM_SELF,PETSC_ERR_FP,"Inserting %g+ig at matrix entry (%D,%D)",(double)PetscRealPart(v[i*n+j]),(double)PetscImaginaryPart(v[i*n+j]),idxm[i],idxn[j]);
1383: #else
1384: SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_FP,"Inserting %g at matrix entry (%D,%D)",(double)v[i*n+j],idxm[i],idxn[j]);
1385: #endif
1386: }
1387: }
1388: }
1390: if (mat->assembled) {
1391: mat->was_assembled = PETSC_TRUE;
1392: mat->assembled = PETSC_FALSE;
1393: }
1394: PetscLogEventBegin(MAT_SetValues,mat,0,0,0);
1395: (*mat->ops->setvalues)(mat,m,idxm,n,idxn,v,addv);
1396: PetscLogEventEnd(MAT_SetValues,mat,0,0,0);
1397: return(0);
1398: }
1401: /*@
1402: MatSetValuesRowLocal - Inserts a row (block row for BAIJ matrices) of nonzero
1403: values into a matrix
1405: Not Collective
1407: Input Parameters:
1408: + mat - the matrix
1409: . row - the (block) row to set
1410: - v - a logically two-dimensional array of values
1412: Notes:
1413: By the values, v, are column-oriented (for the block version) and sorted
1415: All the nonzeros in the row must be provided
1417: The matrix must have previously had its column indices set
1419: The row must belong to this process
1421: Level: intermediate
1423: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
1424: InsertMode, INSERT_VALUES, ADD_VALUES, MatSetValues(), MatSetValuesRow(), MatSetLocalToGlobalMapping()
1425: @*/
1426: PetscErrorCode MatSetValuesRowLocal(Mat mat,PetscInt row,const PetscScalar v[])
1427: {
1429: PetscInt globalrow;
1435: ISLocalToGlobalMappingApply(mat->rmap->mapping,1,&row,&globalrow);
1436: MatSetValuesRow(mat,globalrow,v);
1437: return(0);
1438: }
1440: /*@
1441: MatSetValuesRow - Inserts a row (block row for BAIJ matrices) of nonzero
1442: values into a matrix
1444: Not Collective
1446: Input Parameters:
1447: + mat - the matrix
1448: . row - the (block) row to set
1449: - 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
1451: Notes:
1452: The values, v, are column-oriented for the block version.
1454: All the nonzeros in the row must be provided
1456: THE MATRIX MUST HAVE PREVIOUSLY HAD ITS COLUMN INDICES SET. IT IS RARE THAT THIS ROUTINE IS USED, usually MatSetValues() is used.
1458: The row must belong to this process
1460: Level: advanced
1462: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
1463: InsertMode, INSERT_VALUES, ADD_VALUES, MatSetValues()
1464: @*/
1465: PetscErrorCode MatSetValuesRow(Mat mat,PetscInt row,const PetscScalar v[])
1466: {
1472: MatCheckPreallocated(mat,1);
1474: if (PetscUnlikely(mat->insertmode == ADD_VALUES)) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add and insert values");
1475: if (PetscUnlikely(mat->factortype)) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1476: mat->insertmode = INSERT_VALUES;
1478: if (mat->assembled) {
1479: mat->was_assembled = PETSC_TRUE;
1480: mat->assembled = PETSC_FALSE;
1481: }
1482: PetscLogEventBegin(MAT_SetValues,mat,0,0,0);
1483: if (!mat->ops->setvaluesrow) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1484: (*mat->ops->setvaluesrow)(mat,row,v);
1485: PetscLogEventEnd(MAT_SetValues,mat,0,0,0);
1486: return(0);
1487: }
1489: /*@
1490: MatSetValuesStencil - Inserts or adds a block of values into a matrix.
1491: Using structured grid indexing
1493: Not Collective
1495: Input Parameters:
1496: + mat - the matrix
1497: . m - number of rows being entered
1498: . idxm - grid coordinates (and component number when dof > 1) for matrix rows being entered
1499: . n - number of columns being entered
1500: . idxn - grid coordinates (and component number when dof > 1) for matrix columns being entered
1501: . v - a logically two-dimensional array of values
1502: - addv - either ADD_VALUES or INSERT_VALUES, where
1503: ADD_VALUES adds values to any existing entries, and
1504: INSERT_VALUES replaces existing entries with new values
1506: Notes:
1507: By default the values, v, are row-oriented. See MatSetOption() for other options.
1509: Calls to MatSetValuesStencil() with the INSERT_VALUES and ADD_VALUES
1510: options cannot be mixed without intervening calls to the assembly
1511: routines.
1513: The grid coordinates are across the entire grid, not just the local portion
1515: MatSetValuesStencil() uses 0-based row and column numbers in Fortran
1516: as well as in C.
1518: For setting/accessing vector values via array coordinates you can use the DMDAVecGetArray() routine
1520: In order to use this routine you must either obtain the matrix with DMCreateMatrix()
1521: or call MatSetLocalToGlobalMapping() and MatSetStencil() first.
1523: The columns and rows in the stencil passed in MUST be contained within the
1524: ghost region of the given process as set with DMDACreateXXX() or MatSetStencil(). For example,
1525: if you create a DMDA with an overlap of one grid level and on a particular process its first
1526: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1527: first i index you can use in your column and row indices in MatSetStencil() is 5.
1529: In Fortran idxm and idxn should be declared as
1530: $ MatStencil idxm(4,m),idxn(4,n)
1531: and the values inserted using
1532: $ idxm(MatStencil_i,1) = i
1533: $ idxm(MatStencil_j,1) = j
1534: $ idxm(MatStencil_k,1) = k
1535: $ idxm(MatStencil_c,1) = c
1536: etc
1538: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
1539: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
1540: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
1541: DM_BOUNDARY_PERIODIC boundary type.
1543: 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
1544: a single value per point) you can skip filling those indices.
1546: Inspired by the structured grid interface to the HYPRE package
1547: (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)
1549: Efficiency Alert:
1550: The routine MatSetValuesBlockedStencil() may offer much better efficiency
1551: for users of block sparse formats (MATSEQBAIJ and MATMPIBAIJ).
1553: Level: beginner
1555: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
1556: MatSetValues(), MatSetValuesBlockedStencil(), MatSetStencil(), DMCreateMatrix(), DMDAVecGetArray(), MatStencil
1557: @*/
1558: PetscErrorCode MatSetValuesStencil(Mat mat,PetscInt m,const MatStencil idxm[],PetscInt n,const MatStencil idxn[],const PetscScalar v[],InsertMode addv)
1559: {
1561: PetscInt buf[8192],*bufm=NULL,*bufn=NULL,*jdxm,*jdxn;
1562: PetscInt j,i,dim = mat->stencil.dim,*dims = mat->stencil.dims+1,tmp;
1563: PetscInt *starts = mat->stencil.starts,*dxm = (PetscInt*)idxm,*dxn = (PetscInt*)idxn,sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1566: if (!m || !n) return(0); /* no values to insert */
1572: if ((m+n) <= (PetscInt)(sizeof(buf)/sizeof(PetscInt))) {
1573: jdxm = buf; jdxn = buf+m;
1574: } else {
1575: PetscMalloc2(m,&bufm,n,&bufn);
1576: jdxm = bufm; jdxn = bufn;
1577: }
1578: for (i=0; i<m; i++) {
1579: for (j=0; j<3-sdim; j++) dxm++;
1580: tmp = *dxm++ - starts[0];
1581: for (j=0; j<dim-1; j++) {
1582: if ((*dxm++ - starts[j+1]) < 0 || tmp < 0) tmp = -1;
1583: else tmp = tmp*dims[j] + *(dxm-1) - starts[j+1];
1584: }
1585: if (mat->stencil.noc) dxm++;
1586: jdxm[i] = tmp;
1587: }
1588: for (i=0; i<n; i++) {
1589: for (j=0; j<3-sdim; j++) dxn++;
1590: tmp = *dxn++ - starts[0];
1591: for (j=0; j<dim-1; j++) {
1592: if ((*dxn++ - starts[j+1]) < 0 || tmp < 0) tmp = -1;
1593: else tmp = tmp*dims[j] + *(dxn-1) - starts[j+1];
1594: }
1595: if (mat->stencil.noc) dxn++;
1596: jdxn[i] = tmp;
1597: }
1598: MatSetValuesLocal(mat,m,jdxm,n,jdxn,v,addv);
1599: PetscFree2(bufm,bufn);
1600: return(0);
1601: }
1603: /*@
1604: MatSetValuesBlockedStencil - Inserts or adds a block of values into a matrix.
1605: Using structured grid indexing
1607: Not Collective
1609: Input Parameters:
1610: + mat - the matrix
1611: . m - number of rows being entered
1612: . idxm - grid coordinates for matrix rows being entered
1613: . n - number of columns being entered
1614: . idxn - grid coordinates for matrix columns being entered
1615: . v - a logically two-dimensional array of values
1616: - addv - either ADD_VALUES or INSERT_VALUES, where
1617: ADD_VALUES adds values to any existing entries, and
1618: INSERT_VALUES replaces existing entries with new values
1620: Notes:
1621: By default the values, v, are row-oriented and unsorted.
1622: See MatSetOption() for other options.
1624: Calls to MatSetValuesBlockedStencil() with the INSERT_VALUES and ADD_VALUES
1625: options cannot be mixed without intervening calls to the assembly
1626: routines.
1628: The grid coordinates are across the entire grid, not just the local portion
1630: MatSetValuesBlockedStencil() uses 0-based row and column numbers in Fortran
1631: as well as in C.
1633: For setting/accessing vector values via array coordinates you can use the DMDAVecGetArray() routine
1635: In order to use this routine you must either obtain the matrix with DMCreateMatrix()
1636: or call MatSetBlockSize(), MatSetLocalToGlobalMapping() and MatSetStencil() first.
1638: The columns and rows in the stencil passed in MUST be contained within the
1639: ghost region of the given process as set with DMDACreateXXX() or MatSetStencil(). For example,
1640: if you create a DMDA with an overlap of one grid level and on a particular process its first
1641: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1642: first i index you can use in your column and row indices in MatSetStencil() is 5.
1644: In Fortran idxm and idxn should be declared as
1645: $ MatStencil idxm(4,m),idxn(4,n)
1646: and the values inserted using
1647: $ idxm(MatStencil_i,1) = i
1648: $ idxm(MatStencil_j,1) = j
1649: $ idxm(MatStencil_k,1) = k
1650: etc
1652: Negative indices may be passed in idxm and idxn, these rows and columns are
1653: simply ignored. This allows easily inserting element stiffness matrices
1654: with homogeneous Dirchlet boundary conditions that you don't want represented
1655: in the matrix.
1657: Inspired by the structured grid interface to the HYPRE package
1658: (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)
1660: Level: beginner
1662: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
1663: MatSetValues(), MatSetValuesStencil(), MatSetStencil(), DMCreateMatrix(), DMDAVecGetArray(), MatStencil,
1664: MatSetBlockSize(), MatSetLocalToGlobalMapping()
1665: @*/
1666: PetscErrorCode MatSetValuesBlockedStencil(Mat mat,PetscInt m,const MatStencil idxm[],PetscInt n,const MatStencil idxn[],const PetscScalar v[],InsertMode addv)
1667: {
1669: PetscInt buf[8192],*bufm=NULL,*bufn=NULL,*jdxm,*jdxn;
1670: PetscInt j,i,dim = mat->stencil.dim,*dims = mat->stencil.dims+1,tmp;
1671: PetscInt *starts = mat->stencil.starts,*dxm = (PetscInt*)idxm,*dxn = (PetscInt*)idxn,sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1674: if (!m || !n) return(0); /* no values to insert */
1681: if ((m+n) <= (PetscInt)(sizeof(buf)/sizeof(PetscInt))) {
1682: jdxm = buf; jdxn = buf+m;
1683: } else {
1684: PetscMalloc2(m,&bufm,n,&bufn);
1685: jdxm = bufm; jdxn = bufn;
1686: }
1687: for (i=0; i<m; i++) {
1688: for (j=0; j<3-sdim; j++) dxm++;
1689: tmp = *dxm++ - starts[0];
1690: for (j=0; j<sdim-1; j++) {
1691: if ((*dxm++ - starts[j+1]) < 0 || tmp < 0) tmp = -1;
1692: else tmp = tmp*dims[j] + *(dxm-1) - starts[j+1];
1693: }
1694: dxm++;
1695: jdxm[i] = tmp;
1696: }
1697: for (i=0; i<n; i++) {
1698: for (j=0; j<3-sdim; j++) dxn++;
1699: tmp = *dxn++ - starts[0];
1700: for (j=0; j<sdim-1; j++) {
1701: if ((*dxn++ - starts[j+1]) < 0 || tmp < 0) tmp = -1;
1702: else tmp = tmp*dims[j] + *(dxn-1) - starts[j+1];
1703: }
1704: dxn++;
1705: jdxn[i] = tmp;
1706: }
1707: MatSetValuesBlockedLocal(mat,m,jdxm,n,jdxn,v,addv);
1708: PetscFree2(bufm,bufn);
1709: return(0);
1710: }
1712: /*@
1713: MatSetStencil - Sets the grid information for setting values into a matrix via
1714: MatSetValuesStencil()
1716: Not Collective
1718: Input Parameters:
1719: + mat - the matrix
1720: . dim - dimension of the grid 1, 2, or 3
1721: . dims - number of grid points in x, y, and z direction, including ghost points on your processor
1722: . starts - starting point of ghost nodes on your processor in x, y, and z direction
1723: - dof - number of degrees of freedom per node
1726: Inspired by the structured grid interface to the HYPRE package
1727: (www.llnl.gov/CASC/hyper)
1729: For matrices generated with DMCreateMatrix() this routine is automatically called and so not needed by the
1730: user.
1732: Level: beginner
1734: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
1735: MatSetValues(), MatSetValuesBlockedStencil(), MatSetValuesStencil()
1736: @*/
1737: PetscErrorCode MatSetStencil(Mat mat,PetscInt dim,const PetscInt dims[],const PetscInt starts[],PetscInt dof)
1738: {
1739: PetscInt i;
1746: mat->stencil.dim = dim + (dof > 1);
1747: for (i=0; i<dim; i++) {
1748: mat->stencil.dims[i] = dims[dim-i-1]; /* copy the values in backwards */
1749: mat->stencil.starts[i] = starts[dim-i-1];
1750: }
1751: mat->stencil.dims[dim] = dof;
1752: mat->stencil.starts[dim] = 0;
1753: mat->stencil.noc = (PetscBool)(dof == 1);
1754: return(0);
1755: }
1757: /*@C
1758: MatSetValuesBlocked - Inserts or adds a block of values into a matrix.
1760: Not Collective
1762: Input Parameters:
1763: + mat - the matrix
1764: . v - a logically two-dimensional array of values
1765: . m, idxm - the number of block rows and their global block indices
1766: . n, idxn - the number of block columns and their global block indices
1767: - addv - either ADD_VALUES or INSERT_VALUES, where
1768: ADD_VALUES adds values to any existing entries, and
1769: INSERT_VALUES replaces existing entries with new values
1771: Notes:
1772: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call
1773: MatXXXXSetPreallocation() or MatSetUp() before using this routine.
1775: The m and n count the NUMBER of blocks in the row direction and column direction,
1776: NOT the total number of rows/columns; for example, if the block size is 2 and
1777: you are passing in values for rows 2,3,4,5 then m would be 2 (not 4).
1778: The values in idxm would be 1 2; that is the first index for each block divided by
1779: the block size.
1781: Note that you must call MatSetBlockSize() when constructing this matrix (before
1782: preallocating it).
1784: By default the values, v, are row-oriented, so the layout of
1785: v is the same as for MatSetValues(). See MatSetOption() for other options.
1787: Calls to MatSetValuesBlocked() with the INSERT_VALUES and ADD_VALUES
1788: options cannot be mixed without intervening calls to the assembly
1789: routines.
1791: MatSetValuesBlocked() uses 0-based row and column numbers in Fortran
1792: as well as in C.
1794: Negative indices may be passed in idxm and idxn, these rows and columns are
1795: simply ignored. This allows easily inserting element stiffness matrices
1796: with homogeneous Dirchlet boundary conditions that you don't want represented
1797: in the matrix.
1799: Each time an entry is set within a sparse matrix via MatSetValues(),
1800: internal searching must be done to determine where to place the
1801: data in the matrix storage space. By instead inserting blocks of
1802: entries via MatSetValuesBlocked(), the overhead of matrix assembly is
1803: reduced.
1805: Example:
1806: $ Suppose m=n=2 and block size(bs) = 2 The array is
1807: $
1808: $ 1 2 | 3 4
1809: $ 5 6 | 7 8
1810: $ - - - | - - -
1811: $ 9 10 | 11 12
1812: $ 13 14 | 15 16
1813: $
1814: $ v[] should be passed in like
1815: $ v[] = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]
1816: $
1817: $ If you are not using row oriented storage of v (that is you called MatSetOption(mat,MAT_ROW_ORIENTED,PETSC_FALSE)) then
1818: $ v[] = [1,5,9,13,2,6,10,14,3,7,11,15,4,8,12,16]
1820: Level: intermediate
1822: .seealso: MatSetBlockSize(), MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesBlockedLocal()
1823: @*/
1824: PetscErrorCode MatSetValuesBlocked(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],const PetscScalar v[],InsertMode addv)
1825: {
1831: if (!m || !n) return(0); /* no values to insert */
1835: MatCheckPreallocated(mat,1);
1836: if (mat->insertmode == NOT_SET_VALUES) {
1837: mat->insertmode = addv;
1838: } else if (PetscUnlikely(mat->insertmode != addv)) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
1839: if (PetscDefined(USE_DEBUG)) {
1840: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1841: if (!mat->ops->setvaluesblocked && !mat->ops->setvalues) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1842: }
1844: if (mat->assembled) {
1845: mat->was_assembled = PETSC_TRUE;
1846: mat->assembled = PETSC_FALSE;
1847: }
1848: PetscLogEventBegin(MAT_SetValues,mat,0,0,0);
1849: if (mat->ops->setvaluesblocked) {
1850: (*mat->ops->setvaluesblocked)(mat,m,idxm,n,idxn,v,addv);
1851: } else {
1852: PetscInt buf[8192],*bufr=NULL,*bufc=NULL,*iidxm,*iidxn;
1853: PetscInt i,j,bs,cbs;
1854: MatGetBlockSizes(mat,&bs,&cbs);
1855: if (m*bs+n*cbs <= (PetscInt)(sizeof(buf)/sizeof(PetscInt))) {
1856: iidxm = buf; iidxn = buf + m*bs;
1857: } else {
1858: PetscMalloc2(m*bs,&bufr,n*cbs,&bufc);
1859: iidxm = bufr; iidxn = bufc;
1860: }
1861: for (i=0; i<m; i++) {
1862: for (j=0; j<bs; j++) {
1863: iidxm[i*bs+j] = bs*idxm[i] + j;
1864: }
1865: }
1866: for (i=0; i<n; i++) {
1867: for (j=0; j<cbs; j++) {
1868: iidxn[i*cbs+j] = cbs*idxn[i] + j;
1869: }
1870: }
1871: MatSetValues(mat,m*bs,iidxm,n*cbs,iidxn,v,addv);
1872: PetscFree2(bufr,bufc);
1873: }
1874: PetscLogEventEnd(MAT_SetValues,mat,0,0,0);
1875: return(0);
1876: }
1878: /*@C
1879: MatGetValues - Gets a block of values from a matrix.
1881: Not Collective; can only return values that are owned by the give process
1883: Input Parameters:
1884: + mat - the matrix
1885: . v - a logically two-dimensional array for storing the values
1886: . m, idxm - the number of rows and their global indices
1887: - n, idxn - the number of columns and their global indices
1889: Notes:
1890: The user must allocate space (m*n PetscScalars) for the values, v.
1891: The values, v, are then returned in a row-oriented format,
1892: analogous to that used by default in MatSetValues().
1894: MatGetValues() uses 0-based row and column numbers in
1895: Fortran as well as in C.
1897: MatGetValues() requires that the matrix has been assembled
1898: with MatAssemblyBegin()/MatAssemblyEnd(). Thus, calls to
1899: MatSetValues() and MatGetValues() CANNOT be made in succession
1900: without intermediate matrix assembly.
1902: Negative row or column indices will be ignored and those locations in v[] will be
1903: left unchanged.
1905: For the standard row-based matrix formats, idxm[] can only contain rows owned by the requesting MPI rank.
1906: That is, rows with global index greater than or equal to restart and less than rend where restart and rend are obtainable
1907: from MatGetOwnershipRange(mat,&rstart,&rend).
1909: Level: advanced
1911: .seealso: MatGetRow(), MatCreateSubMatrices(), MatSetValues(), MatGetOwnershipRange(), MatGetValuesLocal()
1912: @*/
1913: PetscErrorCode MatGetValues(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],PetscScalar v[])
1914: {
1920: if (!m || !n) return(0);
1924: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1925: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1926: if (!mat->ops->getvalues) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1927: MatCheckPreallocated(mat,1);
1929: PetscLogEventBegin(MAT_GetValues,mat,0,0,0);
1930: (*mat->ops->getvalues)(mat,m,idxm,n,idxn,v);
1931: PetscLogEventEnd(MAT_GetValues,mat,0,0,0);
1932: return(0);
1933: }
1935: /*@C
1936: MatGetValuesLocal - retrieves values from certain locations in a matrix using the local numbering of the indices
1937: defined previously by MatSetLocalToGlobalMapping()
1939: Not Collective
1941: Input Parameters:
1942: + mat - the matrix
1943: . nrow, irow - number of rows and their local indices
1944: - ncol, icol - number of columns and their local indices
1946: Output Parameter:
1947: . y - a logically two-dimensional array of values
1949: Notes:
1950: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call MatSetLocalToGlobalMapping() before using this routine.
1952: 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,
1953: are greater than or equal to restart and less than rend where restart and rend are obtainable from MatGetOwnershipRange(mat,&rstart,&rend). One can
1954: determine if the resulting global row associated with the local row r is owned by the requesting MPI rank by applying the ISLocalToGlobalMapping set
1955: with MatSetLocalToGlobalMapping().
1957: Developer Notes:
1958: This is labelled with C so does not automatically generate Fortran stubs and interfaces
1959: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
1961: Level: advanced
1963: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetLocalToGlobalMapping(),
1964: MatSetValuesLocal(), MatGetValues()
1965: @*/
1966: PetscErrorCode MatGetValuesLocal(Mat mat,PetscInt nrow,const PetscInt irow[],PetscInt ncol,const PetscInt icol[],PetscScalar y[])
1967: {
1973: MatCheckPreallocated(mat,1);
1974: if (!nrow || !ncol) return(0); /* no values to retrieve */
1977: if (PetscDefined(USE_DEBUG)) {
1978: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1979: if (!mat->ops->getvalueslocal && !mat->ops->getvalues) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1980: }
1981: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1982: PetscLogEventBegin(MAT_GetValues,mat,0,0,0);
1983: if (mat->ops->getvalueslocal) {
1984: (*mat->ops->getvalueslocal)(mat,nrow,irow,ncol,icol,y);
1985: } else {
1986: PetscInt buf[8192],*bufr=NULL,*bufc=NULL,*irowm,*icolm;
1987: if ((nrow+ncol) <= (PetscInt)(sizeof(buf)/sizeof(PetscInt))) {
1988: irowm = buf; icolm = buf+nrow;
1989: } else {
1990: PetscMalloc2(nrow,&bufr,ncol,&bufc);
1991: irowm = bufr; icolm = bufc;
1992: }
1993: if (!mat->rmap->mapping) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"MatGetValuesLocal() cannot proceed without local-to-global row mapping (See MatSetLocalToGlobalMapping()).");
1994: if (!mat->cmap->mapping) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"MatGetValuesLocal() cannot proceed without local-to-global column mapping (See MatSetLocalToGlobalMapping()).");
1995: ISLocalToGlobalMappingApply(mat->rmap->mapping,nrow,irow,irowm);
1996: ISLocalToGlobalMappingApply(mat->cmap->mapping,ncol,icol,icolm);
1997: MatGetValues(mat,nrow,irowm,ncol,icolm,y);
1998: PetscFree2(bufr,bufc);
1999: }
2000: PetscLogEventEnd(MAT_GetValues,mat,0,0,0);
2001: return(0);
2002: }
2004: /*@
2005: MatSetValuesBatch - Adds (ADD_VALUES) many blocks of values into a matrix at once. The blocks must all be square and
2006: the same size. Currently, this can only be called once and creates the given matrix.
2008: Not Collective
2010: Input Parameters:
2011: + mat - the matrix
2012: . nb - the number of blocks
2013: . bs - the number of rows (and columns) in each block
2014: . rows - a concatenation of the rows for each block
2015: - v - a concatenation of logically two-dimensional arrays of values
2017: Notes:
2018: In the future, we will extend this routine to handle rectangular blocks, and to allow multiple calls for a given matrix.
2020: Level: advanced
2022: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
2023: InsertMode, INSERT_VALUES, ADD_VALUES, MatSetValues()
2024: @*/
2025: PetscErrorCode MatSetValuesBatch(Mat mat, PetscInt nb, PetscInt bs, PetscInt rows[], const PetscScalar v[])
2026: {
2034: if (PetscUnlikelyDebug(mat->factortype)) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2036: PetscLogEventBegin(MAT_SetValuesBatch,mat,0,0,0);
2037: if (mat->ops->setvaluesbatch) {
2038: (*mat->ops->setvaluesbatch)(mat,nb,bs,rows,v);
2039: } else {
2040: PetscInt b;
2041: for (b = 0; b < nb; ++b) {
2042: MatSetValues(mat, bs, &rows[b*bs], bs, &rows[b*bs], &v[b*bs*bs], ADD_VALUES);
2043: }
2044: }
2045: PetscLogEventEnd(MAT_SetValuesBatch,mat,0,0,0);
2046: return(0);
2047: }
2049: /*@
2050: MatSetLocalToGlobalMapping - Sets a local-to-global numbering for use by
2051: the routine MatSetValuesLocal() to allow users to insert matrix entries
2052: using a local (per-processor) numbering.
2054: Not Collective
2056: Input Parameters:
2057: + x - the matrix
2058: . rmapping - row mapping created with ISLocalToGlobalMappingCreate() or ISLocalToGlobalMappingCreateIS()
2059: - cmapping - column mapping
2061: Level: intermediate
2064: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesLocal(), MatGetValuesLocal()
2065: @*/
2066: PetscErrorCode MatSetLocalToGlobalMapping(Mat x,ISLocalToGlobalMapping rmapping,ISLocalToGlobalMapping cmapping)
2067: {
2076: if (x->ops->setlocaltoglobalmapping) {
2077: (*x->ops->setlocaltoglobalmapping)(x,rmapping,cmapping);
2078: } else {
2079: PetscLayoutSetISLocalToGlobalMapping(x->rmap,rmapping);
2080: PetscLayoutSetISLocalToGlobalMapping(x->cmap,cmapping);
2081: }
2082: return(0);
2083: }
2086: /*@
2087: MatGetLocalToGlobalMapping - Gets the local-to-global numbering set by MatSetLocalToGlobalMapping()
2089: Not Collective
2091: Input Parameters:
2092: . A - the matrix
2094: Output Parameters:
2095: + rmapping - row mapping
2096: - cmapping - column mapping
2098: Level: advanced
2101: .seealso: MatSetValuesLocal()
2102: @*/
2103: PetscErrorCode MatGetLocalToGlobalMapping(Mat A,ISLocalToGlobalMapping *rmapping,ISLocalToGlobalMapping *cmapping)
2104: {
2110: if (rmapping) *rmapping = A->rmap->mapping;
2111: if (cmapping) *cmapping = A->cmap->mapping;
2112: return(0);
2113: }
2115: /*@
2116: MatSetLayouts - Sets the PetscLayout objects for rows and columns of a matrix
2118: Logically Collective on A
2120: Input Parameters:
2121: + A - the matrix
2122: . rmap - row layout
2123: - cmap - column layout
2125: Level: advanced
2127: .seealso: MatCreateVecs(), MatGetLocalToGlobalMapping(), MatGetLayouts()
2128: @*/
2129: PetscErrorCode MatSetLayouts(Mat A,PetscLayout rmap,PetscLayout cmap)
2130: {
2136: PetscLayoutReference(rmap,&A->rmap);
2137: PetscLayoutReference(cmap,&A->cmap);
2138: return(0);
2139: }
2141: /*@
2142: MatGetLayouts - Gets the PetscLayout objects for rows and columns
2144: Not Collective
2146: Input Parameters:
2147: . A - the matrix
2149: Output Parameters:
2150: + rmap - row layout
2151: - cmap - column layout
2153: Level: advanced
2155: .seealso: MatCreateVecs(), MatGetLocalToGlobalMapping(), MatSetLayouts()
2156: @*/
2157: PetscErrorCode MatGetLayouts(Mat A,PetscLayout *rmap,PetscLayout *cmap)
2158: {
2164: if (rmap) *rmap = A->rmap;
2165: if (cmap) *cmap = A->cmap;
2166: return(0);
2167: }
2169: /*@C
2170: MatSetValuesLocal - Inserts or adds values into certain locations of a matrix,
2171: using a local numbering of the nodes.
2173: Not Collective
2175: Input Parameters:
2176: + mat - the matrix
2177: . nrow, irow - number of rows and their local indices
2178: . ncol, icol - number of columns and their local indices
2179: . y - a logically two-dimensional array of values
2180: - addv - either INSERT_VALUES or ADD_VALUES, where
2181: ADD_VALUES adds values to any existing entries, and
2182: INSERT_VALUES replaces existing entries with new values
2184: Notes:
2185: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call MatXXXXSetPreallocation() or
2186: MatSetUp() before using this routine
2188: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call MatSetLocalToGlobalMapping() before using this routine
2190: Calls to MatSetValuesLocal() with the INSERT_VALUES and ADD_VALUES
2191: options cannot be mixed without intervening calls to the assembly
2192: routines.
2194: These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
2195: MUST be called after all calls to MatSetValuesLocal() have been completed.
2197: Level: intermediate
2199: Developer Notes:
2200: This is labeled with C so does not automatically generate Fortran stubs and interfaces
2201: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2203: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetLocalToGlobalMapping(),
2204: MatSetValueLocal(), MatGetValuesLocal()
2205: @*/
2206: PetscErrorCode MatSetValuesLocal(Mat mat,PetscInt nrow,const PetscInt irow[],PetscInt ncol,const PetscInt icol[],const PetscScalar y[],InsertMode addv)
2207: {
2213: MatCheckPreallocated(mat,1);
2214: if (!nrow || !ncol) return(0); /* no values to insert */
2217: if (mat->insertmode == NOT_SET_VALUES) {
2218: mat->insertmode = addv;
2219: }
2220: else if (PetscUnlikely(mat->insertmode != addv)) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
2221: if (PetscDefined(USE_DEBUG)) {
2222: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2223: if (!mat->ops->setvalueslocal && !mat->ops->setvalues) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2224: }
2226: if (mat->assembled) {
2227: mat->was_assembled = PETSC_TRUE;
2228: mat->assembled = PETSC_FALSE;
2229: }
2230: PetscLogEventBegin(MAT_SetValues,mat,0,0,0);
2231: if (mat->ops->setvalueslocal) {
2232: (*mat->ops->setvalueslocal)(mat,nrow,irow,ncol,icol,y,addv);
2233: } else {
2234: PetscInt buf[8192],*bufr=NULL,*bufc=NULL,*irowm,*icolm;
2235: if ((nrow+ncol) <= (PetscInt)(sizeof(buf)/sizeof(PetscInt))) {
2236: irowm = buf; icolm = buf+nrow;
2237: } else {
2238: PetscMalloc2(nrow,&bufr,ncol,&bufc);
2239: irowm = bufr; icolm = bufc;
2240: }
2241: if (!mat->rmap->mapping) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"MatSetValuesLocal() cannot proceed without local-to-global row mapping (See MatSetLocalToGlobalMapping()).");
2242: if (!mat->cmap->mapping) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"MatSetValuesLocal() cannot proceed without local-to-global column mapping (See MatSetLocalToGlobalMapping()).");
2243: ISLocalToGlobalMappingApply(mat->rmap->mapping,nrow,irow,irowm);
2244: ISLocalToGlobalMappingApply(mat->cmap->mapping,ncol,icol,icolm);
2245: MatSetValues(mat,nrow,irowm,ncol,icolm,y,addv);
2246: PetscFree2(bufr,bufc);
2247: }
2248: PetscLogEventEnd(MAT_SetValues,mat,0,0,0);
2249: return(0);
2250: }
2252: /*@C
2253: MatSetValuesBlockedLocal - Inserts or adds values into certain locations of a matrix,
2254: using a local ordering of the nodes a block at a time.
2256: Not Collective
2258: Input Parameters:
2259: + x - the matrix
2260: . nrow, irow - number of rows and their local indices
2261: . ncol, icol - number of columns and their local indices
2262: . y - a logically two-dimensional array of values
2263: - addv - either INSERT_VALUES or ADD_VALUES, where
2264: ADD_VALUES adds values to any existing entries, and
2265: INSERT_VALUES replaces existing entries with new values
2267: Notes:
2268: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call MatXXXXSetPreallocation() or
2269: MatSetUp() before using this routine
2271: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call MatSetBlockSize() and MatSetLocalToGlobalMapping()
2272: before using this routineBefore calling MatSetValuesLocal(), the user must first set the
2274: Calls to MatSetValuesBlockedLocal() with the INSERT_VALUES and ADD_VALUES
2275: options cannot be mixed without intervening calls to the assembly
2276: routines.
2278: These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
2279: MUST be called after all calls to MatSetValuesBlockedLocal() have been completed.
2281: Level: intermediate
2283: Developer Notes:
2284: This is labeled with C so does not automatically generate Fortran stubs and interfaces
2285: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2287: .seealso: MatSetBlockSize(), MatSetLocalToGlobalMapping(), MatAssemblyBegin(), MatAssemblyEnd(),
2288: MatSetValuesLocal(), MatSetValuesBlocked()
2289: @*/
2290: PetscErrorCode MatSetValuesBlockedLocal(Mat mat,PetscInt nrow,const PetscInt irow[],PetscInt ncol,const PetscInt icol[],const PetscScalar y[],InsertMode addv)
2291: {
2297: MatCheckPreallocated(mat,1);
2298: if (!nrow || !ncol) return(0); /* no values to insert */
2302: if (mat->insertmode == NOT_SET_VALUES) {
2303: mat->insertmode = addv;
2304: } else if (PetscUnlikely(mat->insertmode != addv)) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
2305: if (PetscDefined(USE_DEBUG)) {
2306: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2307: if (!mat->ops->setvaluesblockedlocal && !mat->ops->setvaluesblocked && !mat->ops->setvalueslocal && !mat->ops->setvalues) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2308: }
2310: if (mat->assembled) {
2311: mat->was_assembled = PETSC_TRUE;
2312: mat->assembled = PETSC_FALSE;
2313: }
2314: if (PetscUnlikelyDebug(mat->rmap->mapping)) { /* Condition on the mapping existing, because MatSetValuesBlockedLocal_IS does not require it to be set. */
2315: PetscInt irbs, rbs;
2316: MatGetBlockSizes(mat, &rbs, NULL);
2317: ISLocalToGlobalMappingGetBlockSize(mat->rmap->mapping,&irbs);
2318: if (rbs != irbs) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Different row block sizes! mat %D, row l2g map %D",rbs,irbs);
2319: }
2320: if (PetscUnlikelyDebug(mat->cmap->mapping)) {
2321: PetscInt icbs, cbs;
2322: MatGetBlockSizes(mat,NULL,&cbs);
2323: ISLocalToGlobalMappingGetBlockSize(mat->cmap->mapping,&icbs);
2324: if (cbs != icbs) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Different col block sizes! mat %D, col l2g map %D",cbs,icbs);
2325: }
2326: PetscLogEventBegin(MAT_SetValues,mat,0,0,0);
2327: if (mat->ops->setvaluesblockedlocal) {
2328: (*mat->ops->setvaluesblockedlocal)(mat,nrow,irow,ncol,icol,y,addv);
2329: } else {
2330: PetscInt buf[8192],*bufr=NULL,*bufc=NULL,*irowm,*icolm;
2331: if ((nrow+ncol) <= (PetscInt)(sizeof(buf)/sizeof(PetscInt))) {
2332: irowm = buf; icolm = buf + nrow;
2333: } else {
2334: PetscMalloc2(nrow,&bufr,ncol,&bufc);
2335: irowm = bufr; icolm = bufc;
2336: }
2337: ISLocalToGlobalMappingApplyBlock(mat->rmap->mapping,nrow,irow,irowm);
2338: ISLocalToGlobalMappingApplyBlock(mat->cmap->mapping,ncol,icol,icolm);
2339: MatSetValuesBlocked(mat,nrow,irowm,ncol,icolm,y,addv);
2340: PetscFree2(bufr,bufc);
2341: }
2342: PetscLogEventEnd(MAT_SetValues,mat,0,0,0);
2343: return(0);
2344: }
2346: /*@
2347: MatMultDiagonalBlock - Computes the matrix-vector product, y = Dx. Where D is defined by the inode or block structure of the diagonal
2349: Collective on Mat
2351: Input Parameters:
2352: + mat - the matrix
2353: - x - the vector to be multiplied
2355: Output Parameters:
2356: . y - the result
2358: Notes:
2359: The vectors x and y cannot be the same. I.e., one cannot
2360: call MatMult(A,y,y).
2362: Level: developer
2364: .seealso: MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
2365: @*/
2366: PetscErrorCode MatMultDiagonalBlock(Mat mat,Vec x,Vec y)
2367: {
2376: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2377: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2378: if (x == y) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2379: MatCheckPreallocated(mat,1);
2381: if (!mat->ops->multdiagonalblock) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s does not have a multiply defined",((PetscObject)mat)->type_name);
2382: (*mat->ops->multdiagonalblock)(mat,x,y);
2383: PetscObjectStateIncrease((PetscObject)y);
2384: return(0);
2385: }
2387: /* --------------------------------------------------------*/
2388: /*@
2389: MatMult - Computes the matrix-vector product, y = Ax.
2391: Neighbor-wise Collective on Mat
2393: Input Parameters:
2394: + mat - the matrix
2395: - x - the vector to be multiplied
2397: Output Parameters:
2398: . y - the result
2400: Notes:
2401: The vectors x and y cannot be the same. I.e., one cannot
2402: call MatMult(A,y,y).
2404: Level: beginner
2406: .seealso: MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
2407: @*/
2408: PetscErrorCode MatMult(Mat mat,Vec x,Vec y)
2409: {
2417: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2418: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2419: if (x == y) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2420: #if !defined(PETSC_HAVE_CONSTRAINTS)
2421: if (mat->cmap->N != x->map->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
2422: if (mat->rmap->N != y->map->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->rmap->N,y->map->N);
2423: if (mat->rmap->n != y->map->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: local dim %D %D",mat->rmap->n,y->map->n);
2424: #endif
2425: VecSetErrorIfLocked(y,3);
2426: if (mat->erroriffailure) {VecValidValues(x,2,PETSC_TRUE);}
2427: MatCheckPreallocated(mat,1);
2429: VecLockReadPush(x);
2430: if (!mat->ops->mult) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s does not have a multiply defined",((PetscObject)mat)->type_name);
2431: PetscLogEventBegin(MAT_Mult,mat,x,y,0);
2432: (*mat->ops->mult)(mat,x,y);
2433: PetscLogEventEnd(MAT_Mult,mat,x,y,0);
2434: if (mat->erroriffailure) {VecValidValues(y,3,PETSC_FALSE);}
2435: VecLockReadPop(x);
2436: return(0);
2437: }
2439: /*@
2440: MatMultTranspose - Computes matrix transpose times a vector y = A^T * x.
2442: Neighbor-wise Collective on Mat
2444: Input Parameters:
2445: + mat - the matrix
2446: - x - the vector to be multiplied
2448: Output Parameters:
2449: . y - the result
2451: Notes:
2452: The vectors x and y cannot be the same. I.e., one cannot
2453: call MatMultTranspose(A,y,y).
2455: For complex numbers this does NOT compute the Hermitian (complex conjugate) transpose multiple,
2456: use MatMultHermitianTranspose()
2458: Level: beginner
2460: .seealso: MatMult(), MatMultAdd(), MatMultTransposeAdd(), MatMultHermitianTranspose(), MatTranspose()
2461: @*/
2462: PetscErrorCode MatMultTranspose(Mat mat,Vec x,Vec y)
2463: {
2464: PetscErrorCode (*op)(Mat,Vec,Vec)=NULL,ierr;
2472: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2473: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2474: if (x == y) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2475: #if !defined(PETSC_HAVE_CONSTRAINTS)
2476: if (mat->rmap->N != x->map->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->rmap->N,x->map->N);
2477: if (mat->cmap->N != y->map->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->cmap->N,y->map->N);
2478: #endif
2479: if (mat->erroriffailure) {VecValidValues(x,2,PETSC_TRUE);}
2480: MatCheckPreallocated(mat,1);
2482: if (!mat->ops->multtranspose) {
2483: if (mat->symmetric && mat->ops->mult) op = mat->ops->mult;
2484: if (!op) SETERRQ1(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);
2485: } else op = mat->ops->multtranspose;
2486: PetscLogEventBegin(MAT_MultTranspose,mat,x,y,0);
2487: VecLockReadPush(x);
2488: (*op)(mat,x,y);
2489: VecLockReadPop(x);
2490: PetscLogEventEnd(MAT_MultTranspose,mat,x,y,0);
2491: PetscObjectStateIncrease((PetscObject)y);
2492: if (mat->erroriffailure) {VecValidValues(y,3,PETSC_FALSE);}
2493: return(0);
2494: }
2496: /*@
2497: MatMultHermitianTranspose - Computes matrix Hermitian transpose times a vector.
2499: Neighbor-wise Collective on Mat
2501: Input Parameters:
2502: + mat - the matrix
2503: - x - the vector to be multilplied
2505: Output Parameters:
2506: . y - the result
2508: Notes:
2509: The vectors x and y cannot be the same. I.e., one cannot
2510: call MatMultHermitianTranspose(A,y,y).
2512: Also called the conjugate transpose, complex conjugate transpose, or adjoint.
2514: For real numbers MatMultTranspose() and MatMultHermitianTranspose() are identical.
2516: Level: beginner
2518: .seealso: MatMult(), MatMultAdd(), MatMultHermitianTransposeAdd(), MatMultTranspose()
2519: @*/
2520: PetscErrorCode MatMultHermitianTranspose(Mat mat,Vec x,Vec y)
2521: {
2530: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2531: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2532: if (x == y) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2533: #if !defined(PETSC_HAVE_CONSTRAINTS)
2534: if (mat->rmap->N != x->map->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->rmap->N,x->map->N);
2535: if (mat->cmap->N != y->map->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->cmap->N,y->map->N);
2536: #endif
2537: MatCheckPreallocated(mat,1);
2539: PetscLogEventBegin(MAT_MultHermitianTranspose,mat,x,y,0);
2540: #if defined(PETSC_USE_COMPLEX)
2541: if (mat->ops->multhermitiantranspose || (mat->hermitian && mat->ops->mult)) {
2542: VecLockReadPush(x);
2543: if (mat->ops->multhermitiantranspose) {
2544: (*mat->ops->multhermitiantranspose)(mat,x,y);
2545: } else {
2546: (*mat->ops->mult)(mat,x,y);
2547: }
2548: VecLockReadPop(x);
2549: } else {
2550: Vec w;
2551: VecDuplicate(x,&w);
2552: VecCopy(x,w);
2553: VecConjugate(w);
2554: MatMultTranspose(mat,w,y);
2555: VecDestroy(&w);
2556: VecConjugate(y);
2557: }
2558: PetscObjectStateIncrease((PetscObject)y);
2559: #else
2560: MatMultTranspose(mat,x,y);
2561: #endif
2562: PetscLogEventEnd(MAT_MultHermitianTranspose,mat,x,y,0);
2563: return(0);
2564: }
2566: /*@
2567: MatMultAdd - Computes v3 = v2 + A * v1.
2569: Neighbor-wise Collective on Mat
2571: Input Parameters:
2572: + mat - the matrix
2573: - v1, v2 - the vectors
2575: Output Parameters:
2576: . v3 - the result
2578: Notes:
2579: The vectors v1 and v3 cannot be the same. I.e., one cannot
2580: call MatMultAdd(A,v1,v2,v1).
2582: Level: beginner
2584: .seealso: MatMultTranspose(), MatMult(), MatMultTransposeAdd()
2585: @*/
2586: PetscErrorCode MatMultAdd(Mat mat,Vec v1,Vec v2,Vec v3)
2587: {
2597: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2598: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2599: if (mat->cmap->N != v1->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec v1: global dim %D %D",mat->cmap->N,v1->map->N);
2600: /* if (mat->rmap->N != v2->map->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec v2: global dim %D %D",mat->rmap->N,v2->map->N);
2601: if (mat->rmap->N != v3->map->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec v3: global dim %D %D",mat->rmap->N,v3->map->N); */
2602: if (mat->rmap->n != v3->map->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec v3: local dim %D %D",mat->rmap->n,v3->map->n);
2603: if (mat->rmap->n != v2->map->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec v2: local dim %D %D",mat->rmap->n,v2->map->n);
2604: if (v1 == v3) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"v1 and v3 must be different vectors");
2605: MatCheckPreallocated(mat,1);
2607: if (!mat->ops->multadd) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"No MatMultAdd() for matrix type %s",((PetscObject)mat)->type_name);
2608: PetscLogEventBegin(MAT_MultAdd,mat,v1,v2,v3);
2609: VecLockReadPush(v1);
2610: (*mat->ops->multadd)(mat,v1,v2,v3);
2611: VecLockReadPop(v1);
2612: PetscLogEventEnd(MAT_MultAdd,mat,v1,v2,v3);
2613: PetscObjectStateIncrease((PetscObject)v3);
2614: return(0);
2615: }
2617: /*@
2618: MatMultTransposeAdd - Computes v3 = v2 + A' * v1.
2620: Neighbor-wise Collective on Mat
2622: Input Parameters:
2623: + mat - the matrix
2624: - v1, v2 - the vectors
2626: Output Parameters:
2627: . v3 - the result
2629: Notes:
2630: The vectors v1 and v3 cannot be the same. I.e., one cannot
2631: call MatMultTransposeAdd(A,v1,v2,v1).
2633: Level: beginner
2635: .seealso: MatMultTranspose(), MatMultAdd(), MatMult()
2636: @*/
2637: PetscErrorCode MatMultTransposeAdd(Mat mat,Vec v1,Vec v2,Vec v3)
2638: {
2648: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2649: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2650: if (!mat->ops->multtransposeadd) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2651: if (v1 == v3) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"v1 and v3 must be different vectors");
2652: if (mat->rmap->N != v1->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec v1: global dim %D %D",mat->rmap->N,v1->map->N);
2653: if (mat->cmap->N != v2->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec v2: global dim %D %D",mat->cmap->N,v2->map->N);
2654: if (mat->cmap->N != v3->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec v3: global dim %D %D",mat->cmap->N,v3->map->N);
2655: MatCheckPreallocated(mat,1);
2657: PetscLogEventBegin(MAT_MultTransposeAdd,mat,v1,v2,v3);
2658: VecLockReadPush(v1);
2659: (*mat->ops->multtransposeadd)(mat,v1,v2,v3);
2660: VecLockReadPop(v1);
2661: PetscLogEventEnd(MAT_MultTransposeAdd,mat,v1,v2,v3);
2662: PetscObjectStateIncrease((PetscObject)v3);
2663: return(0);
2664: }
2666: /*@
2667: MatMultHermitianTransposeAdd - Computes v3 = v2 + A^H * v1.
2669: Neighbor-wise Collective on Mat
2671: Input Parameters:
2672: + mat - the matrix
2673: - v1, v2 - the vectors
2675: Output Parameters:
2676: . v3 - the result
2678: Notes:
2679: The vectors v1 and v3 cannot be the same. I.e., one cannot
2680: call MatMultHermitianTransposeAdd(A,v1,v2,v1).
2682: Level: beginner
2684: .seealso: MatMultHermitianTranspose(), MatMultTranspose(), MatMultAdd(), MatMult()
2685: @*/
2686: PetscErrorCode MatMultHermitianTransposeAdd(Mat mat,Vec v1,Vec v2,Vec v3)
2687: {
2697: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2698: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2699: if (v1 == v3) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"v1 and v3 must be different vectors");
2700: if (mat->rmap->N != v1->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec v1: global dim %D %D",mat->rmap->N,v1->map->N);
2701: if (mat->cmap->N != v2->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec v2: global dim %D %D",mat->cmap->N,v2->map->N);
2702: if (mat->cmap->N != v3->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec v3: global dim %D %D",mat->cmap->N,v3->map->N);
2703: MatCheckPreallocated(mat,1);
2705: PetscLogEventBegin(MAT_MultHermitianTransposeAdd,mat,v1,v2,v3);
2706: VecLockReadPush(v1);
2707: if (mat->ops->multhermitiantransposeadd) {
2708: (*mat->ops->multhermitiantransposeadd)(mat,v1,v2,v3);
2709: } else {
2710: Vec w,z;
2711: VecDuplicate(v1,&w);
2712: VecCopy(v1,w);
2713: VecConjugate(w);
2714: VecDuplicate(v3,&z);
2715: MatMultTranspose(mat,w,z);
2716: VecDestroy(&w);
2717: VecConjugate(z);
2718: if (v2 != v3) {
2719: VecWAXPY(v3,1.0,v2,z);
2720: } else {
2721: VecAXPY(v3,1.0,z);
2722: }
2723: VecDestroy(&z);
2724: }
2725: VecLockReadPop(v1);
2726: PetscLogEventEnd(MAT_MultHermitianTransposeAdd,mat,v1,v2,v3);
2727: PetscObjectStateIncrease((PetscObject)v3);
2728: return(0);
2729: }
2731: /*@
2732: MatMultConstrained - The inner multiplication routine for a
2733: constrained matrix P^T A P.
2735: Neighbor-wise Collective on Mat
2737: Input Parameters:
2738: + mat - the matrix
2739: - x - the vector to be multilplied
2741: Output Parameters:
2742: . y - the result
2744: Notes:
2745: The vectors x and y cannot be the same. I.e., one cannot
2746: call MatMult(A,y,y).
2748: Level: beginner
2750: .seealso: MatMult(), MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
2751: @*/
2752: PetscErrorCode MatMultConstrained(Mat mat,Vec x,Vec y)
2753: {
2760: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2761: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2762: if (x == y) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2763: if (mat->cmap->N != x->map->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
2764: if (mat->rmap->N != y->map->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->rmap->N,y->map->N);
2765: if (mat->rmap->n != y->map->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: local dim %D %D",mat->rmap->n,y->map->n);
2767: PetscLogEventBegin(MAT_MultConstrained,mat,x,y,0);
2768: VecLockReadPush(x);
2769: (*mat->ops->multconstrained)(mat,x,y);
2770: VecLockReadPop(x);
2771: PetscLogEventEnd(MAT_MultConstrained,mat,x,y,0);
2772: PetscObjectStateIncrease((PetscObject)y);
2773: return(0);
2774: }
2776: /*@
2777: MatMultTransposeConstrained - The inner multiplication routine for a
2778: constrained matrix P^T A^T P.
2780: Neighbor-wise Collective on Mat
2782: Input Parameters:
2783: + mat - the matrix
2784: - x - the vector to be multilplied
2786: Output Parameters:
2787: . y - the result
2789: Notes:
2790: The vectors x and y cannot be the same. I.e., one cannot
2791: call MatMult(A,y,y).
2793: Level: beginner
2795: .seealso: MatMult(), MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
2796: @*/
2797: PetscErrorCode MatMultTransposeConstrained(Mat mat,Vec x,Vec y)
2798: {
2805: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2806: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2807: if (x == y) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2808: if (mat->rmap->N != x->map->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
2809: if (mat->cmap->N != y->map->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->rmap->N,y->map->N);
2811: PetscLogEventBegin(MAT_MultConstrained,mat,x,y,0);
2812: (*mat->ops->multtransposeconstrained)(mat,x,y);
2813: PetscLogEventEnd(MAT_MultConstrained,mat,x,y,0);
2814: PetscObjectStateIncrease((PetscObject)y);
2815: return(0);
2816: }
2818: /*@C
2819: MatGetFactorType - gets the type of factorization it is
2821: Not Collective
2823: Input Parameters:
2824: . mat - the matrix
2826: Output Parameters:
2827: . t - the type, one of MAT_FACTOR_NONE, MAT_FACTOR_LU, MAT_FACTOR_CHOLESKY, MAT_FACTOR_ILU, MAT_FACTOR_ICC,MAT_FACTOR_ILUDT
2829: Level: intermediate
2831: .seealso: MatFactorType, MatGetFactor(), MatSetFactorType()
2832: @*/
2833: PetscErrorCode MatGetFactorType(Mat mat,MatFactorType *t)
2834: {
2839: *t = mat->factortype;
2840: return(0);
2841: }
2843: /*@C
2844: MatSetFactorType - sets the type of factorization it is
2846: Logically Collective on Mat
2848: Input Parameters:
2849: + mat - the matrix
2850: - t - the type, one of MAT_FACTOR_NONE, MAT_FACTOR_LU, MAT_FACTOR_CHOLESKY, MAT_FACTOR_ILU, MAT_FACTOR_ICC,MAT_FACTOR_ILUDT
2852: Level: intermediate
2854: .seealso: MatFactorType, MatGetFactor(), MatGetFactorType()
2855: @*/
2856: PetscErrorCode MatSetFactorType(Mat mat, MatFactorType t)
2857: {
2861: mat->factortype = t;
2862: return(0);
2863: }
2865: /* ------------------------------------------------------------*/
2866: /*@C
2867: MatGetInfo - Returns information about matrix storage (number of
2868: nonzeros, memory, etc.).
2870: Collective on Mat if MAT_GLOBAL_MAX or MAT_GLOBAL_SUM is used as the flag
2872: Input Parameters:
2873: . mat - the matrix
2875: Output Parameters:
2876: + flag - flag indicating the type of parameters to be returned
2877: (MAT_LOCAL - local matrix, MAT_GLOBAL_MAX - maximum over all processors,
2878: MAT_GLOBAL_SUM - sum over all processors)
2879: - info - matrix information context
2881: Notes:
2882: The MatInfo context contains a variety of matrix data, including
2883: number of nonzeros allocated and used, number of mallocs during
2884: matrix assembly, etc. Additional information for factored matrices
2885: is provided (such as the fill ratio, number of mallocs during
2886: factorization, etc.). Much of this info is printed to PETSC_STDOUT
2887: when using the runtime options
2888: $ -info -mat_view ::ascii_info
2890: Example for C/C++ Users:
2891: See the file ${PETSC_DIR}/include/petscmat.h for a complete list of
2892: data within the MatInfo context. For example,
2893: .vb
2894: MatInfo info;
2895: Mat A;
2896: double mal, nz_a, nz_u;
2898: MatGetInfo(A,MAT_LOCAL,&info);
2899: mal = info.mallocs;
2900: nz_a = info.nz_allocated;
2901: .ve
2903: Example for Fortran Users:
2904: Fortran users should declare info as a double precision
2905: array of dimension MAT_INFO_SIZE, and then extract the parameters
2906: of interest. See the file ${PETSC_DIR}/include/petsc/finclude/petscmat.h
2907: a complete list of parameter names.
2908: .vb
2909: double precision info(MAT_INFO_SIZE)
2910: double precision mal, nz_a
2911: Mat A
2912: integer ierr
2914: call MatGetInfo(A,MAT_LOCAL,info,ierr)
2915: mal = info(MAT_INFO_MALLOCS)
2916: nz_a = info(MAT_INFO_NZ_ALLOCATED)
2917: .ve
2919: Level: intermediate
2921: Developer Note: fortran interface is not autogenerated as the f90
2922: interface defintion cannot be generated correctly [due to MatInfo]
2924: .seealso: MatStashGetInfo()
2926: @*/
2927: PetscErrorCode MatGetInfo(Mat mat,MatInfoType flag,MatInfo *info)
2928: {
2935: if (!mat->ops->getinfo) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2936: MatCheckPreallocated(mat,1);
2937: (*mat->ops->getinfo)(mat,flag,info);
2938: return(0);
2939: }
2941: /*
2942: This is used by external packages where it is not easy to get the info from the actual
2943: matrix factorization.
2944: */
2945: PetscErrorCode MatGetInfo_External(Mat A,MatInfoType flag,MatInfo *info)
2946: {
2950: PetscMemzero(info,sizeof(MatInfo));
2951: return(0);
2952: }
2954: /* ----------------------------------------------------------*/
2956: /*@C
2957: MatLUFactor - Performs in-place LU factorization of matrix.
2959: Collective on Mat
2961: Input Parameters:
2962: + mat - the matrix
2963: . row - row permutation
2964: . col - column permutation
2965: - info - options for factorization, includes
2966: $ fill - expected fill as ratio of original fill.
2967: $ dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
2968: $ Run with the option -info to determine an optimal value to use
2970: Notes:
2971: Most users should employ the simplified KSP interface for linear solvers
2972: instead of working directly with matrix algebra routines such as this.
2973: See, e.g., KSPCreate().
2975: This changes the state of the matrix to a factored matrix; it cannot be used
2976: for example with MatSetValues() unless one first calls MatSetUnfactored().
2978: Level: developer
2980: .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor(),
2981: MatGetOrdering(), MatSetUnfactored(), MatFactorInfo, MatGetFactor()
2983: Developer Note: fortran interface is not autogenerated as the f90
2984: interface defintion cannot be generated correctly [due to MatFactorInfo]
2986: @*/
2987: PetscErrorCode MatLUFactor(Mat mat,IS row,IS col,const MatFactorInfo *info)
2988: {
2990: MatFactorInfo tinfo;
2998: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2999: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3000: if (!mat->ops->lufactor) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3001: MatCheckPreallocated(mat,1);
3002: if (!info) {
3003: MatFactorInfoInitialize(&tinfo);
3004: info = &tinfo;
3005: }
3007: PetscLogEventBegin(MAT_LUFactor,mat,row,col,0);
3008: (*mat->ops->lufactor)(mat,row,col,info);
3009: PetscLogEventEnd(MAT_LUFactor,mat,row,col,0);
3010: PetscObjectStateIncrease((PetscObject)mat);
3011: return(0);
3012: }
3014: /*@C
3015: MatILUFactor - Performs in-place ILU factorization of matrix.
3017: Collective on Mat
3019: Input Parameters:
3020: + mat - the matrix
3021: . row - row permutation
3022: . col - column permutation
3023: - info - structure containing
3024: $ levels - number of levels of fill.
3025: $ expected fill - as ratio of original fill.
3026: $ 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
3027: missing diagonal entries)
3029: Notes:
3030: Probably really in-place only when level of fill is zero, otherwise allocates
3031: new space to store factored matrix and deletes previous memory.
3033: Most users should employ the simplified KSP interface for linear solvers
3034: instead of working directly with matrix algebra routines such as this.
3035: See, e.g., KSPCreate().
3037: Level: developer
3039: .seealso: MatILUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor(), MatFactorInfo
3041: Developer Note: fortran interface is not autogenerated as the f90
3042: interface defintion cannot be generated correctly [due to MatFactorInfo]
3044: @*/
3045: PetscErrorCode MatILUFactor(Mat mat,IS row,IS col,const MatFactorInfo *info)
3046: {
3055: if (mat->rmap->N != mat->cmap->N) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONG,"matrix must be square");
3056: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3057: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3058: if (!mat->ops->ilufactor) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3059: MatCheckPreallocated(mat,1);
3061: PetscLogEventBegin(MAT_ILUFactor,mat,row,col,0);
3062: (*mat->ops->ilufactor)(mat,row,col,info);
3063: PetscLogEventEnd(MAT_ILUFactor,mat,row,col,0);
3064: PetscObjectStateIncrease((PetscObject)mat);
3065: return(0);
3066: }
3068: /*@C
3069: MatLUFactorSymbolic - Performs symbolic LU factorization of matrix.
3070: Call this routine before calling MatLUFactorNumeric().
3072: Collective on Mat
3074: Input Parameters:
3075: + fact - the factor matrix obtained with MatGetFactor()
3076: . mat - the matrix
3077: . row, col - row and column permutations
3078: - info - options for factorization, includes
3079: $ fill - expected fill as ratio of original fill.
3080: $ dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3081: $ Run with the option -info to determine an optimal value to use
3084: Notes:
3085: See Users-Manual: ch_mat for additional information about choosing the fill factor for better efficiency.
3087: Most users should employ the simplified KSP interface for linear solvers
3088: instead of working directly with matrix algebra routines such as this.
3089: See, e.g., KSPCreate().
3091: Level: developer
3093: .seealso: MatLUFactor(), MatLUFactorNumeric(), MatCholeskyFactor(), MatFactorInfo, MatFactorInfoInitialize()
3095: Developer Note: fortran interface is not autogenerated as the f90
3096: interface defintion cannot be generated correctly [due to MatFactorInfo]
3098: @*/
3099: PetscErrorCode MatLUFactorSymbolic(Mat fact,Mat mat,IS row,IS col,const MatFactorInfo *info)
3100: {
3102: MatFactorInfo tinfo;
3111: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3112: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3113: if (!(fact)->ops->lufactorsymbolic) {
3114: MatSolverType stype;
3115: MatFactorGetSolverType(fact,&stype);
3116: SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s symbolic LU using solver package %s",((PetscObject)mat)->type_name,stype);
3117: }
3118: MatCheckPreallocated(mat,2);
3119: if (!info) {
3120: MatFactorInfoInitialize(&tinfo);
3121: info = &tinfo;
3122: }
3124: PetscLogEventBegin(MAT_LUFactorSymbolic,mat,row,col,0);
3125: (fact->ops->lufactorsymbolic)(fact,mat,row,col,info);
3126: PetscLogEventEnd(MAT_LUFactorSymbolic,mat,row,col,0);
3127: PetscObjectStateIncrease((PetscObject)fact);
3128: return(0);
3129: }
3131: /*@C
3132: MatLUFactorNumeric - Performs numeric LU factorization of a matrix.
3133: Call this routine after first calling MatLUFactorSymbolic().
3135: Collective on Mat
3137: Input Parameters:
3138: + fact - the factor matrix obtained with MatGetFactor()
3139: . mat - the matrix
3140: - info - options for factorization
3142: Notes:
3143: See MatLUFactor() for in-place factorization. See
3144: MatCholeskyFactorNumeric() for the symmetric, positive definite case.
3146: Most users should employ the simplified KSP interface for linear solvers
3147: instead of working directly with matrix algebra routines such as this.
3148: See, e.g., KSPCreate().
3150: Level: developer
3152: .seealso: MatLUFactorSymbolic(), MatLUFactor(), MatCholeskyFactor()
3154: Developer Note: fortran interface is not autogenerated as the f90
3155: interface defintion cannot be generated correctly [due to MatFactorInfo]
3157: @*/
3158: PetscErrorCode MatLUFactorNumeric(Mat fact,Mat mat,const MatFactorInfo *info)
3159: {
3160: MatFactorInfo tinfo;
3168: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3169: if (mat->rmap->N != (fact)->rmap->N || mat->cmap->N != (fact)->cmap->N) SETERRQ4(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Mat fact: global dimensions are different %D should = %D %D should = %D",mat->rmap->N,(fact)->rmap->N,mat->cmap->N,(fact)->cmap->N);
3171: if (!(fact)->ops->lufactornumeric) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s numeric LU",((PetscObject)mat)->type_name);
3172: MatCheckPreallocated(mat,2);
3173: if (!info) {
3174: MatFactorInfoInitialize(&tinfo);
3175: info = &tinfo;
3176: }
3178: PetscLogEventBegin(MAT_LUFactorNumeric,mat,fact,0,0);
3179: (fact->ops->lufactornumeric)(fact,mat,info);
3180: PetscLogEventEnd(MAT_LUFactorNumeric,mat,fact,0,0);
3181: MatViewFromOptions(fact,NULL,"-mat_factor_view");
3182: PetscObjectStateIncrease((PetscObject)fact);
3183: return(0);
3184: }
3186: /*@C
3187: MatCholeskyFactor - Performs in-place Cholesky factorization of a
3188: symmetric matrix.
3190: Collective on Mat
3192: Input Parameters:
3193: + mat - the matrix
3194: . perm - row and column permutations
3195: - f - expected fill as ratio of original fill
3197: Notes:
3198: See MatLUFactor() for the nonsymmetric case. See also
3199: MatCholeskyFactorSymbolic(), and MatCholeskyFactorNumeric().
3201: Most users should employ the simplified KSP interface for linear solvers
3202: instead of working directly with matrix algebra routines such as this.
3203: See, e.g., KSPCreate().
3205: Level: developer
3207: .seealso: MatLUFactor(), MatCholeskyFactorSymbolic(), MatCholeskyFactorNumeric()
3208: MatGetOrdering()
3210: Developer Note: fortran interface is not autogenerated as the f90
3211: interface defintion cannot be generated correctly [due to MatFactorInfo]
3213: @*/
3214: PetscErrorCode MatCholeskyFactor(Mat mat,IS perm,const MatFactorInfo *info)
3215: {
3217: MatFactorInfo tinfo;
3224: if (mat->rmap->N != mat->cmap->N) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONG,"Matrix must be square");
3225: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3226: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3227: if (!mat->ops->choleskyfactor) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"In-place factorization for Mat type %s is not supported, try out-of-place factorization. See MatCholeskyFactorSymbolic/Numeric",((PetscObject)mat)->type_name);
3228: MatCheckPreallocated(mat,1);
3229: if (!info) {
3230: MatFactorInfoInitialize(&tinfo);
3231: info = &tinfo;
3232: }
3234: PetscLogEventBegin(MAT_CholeskyFactor,mat,perm,0,0);
3235: (*mat->ops->choleskyfactor)(mat,perm,info);
3236: PetscLogEventEnd(MAT_CholeskyFactor,mat,perm,0,0);
3237: PetscObjectStateIncrease((PetscObject)mat);
3238: return(0);
3239: }
3241: /*@C
3242: MatCholeskyFactorSymbolic - Performs symbolic Cholesky factorization
3243: of a symmetric matrix.
3245: Collective on Mat
3247: Input Parameters:
3248: + fact - the factor matrix obtained with MatGetFactor()
3249: . mat - the matrix
3250: . perm - row and column permutations
3251: - info - options for factorization, includes
3252: $ fill - expected fill as ratio of original fill.
3253: $ dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3254: $ Run with the option -info to determine an optimal value to use
3256: Notes:
3257: See MatLUFactorSymbolic() for the nonsymmetric case. See also
3258: MatCholeskyFactor() and MatCholeskyFactorNumeric().
3260: Most users should employ the simplified KSP interface for linear solvers
3261: instead of working directly with matrix algebra routines such as this.
3262: See, e.g., KSPCreate().
3264: Level: developer
3266: .seealso: MatLUFactorSymbolic(), MatCholeskyFactor(), MatCholeskyFactorNumeric()
3267: MatGetOrdering()
3269: Developer Note: fortran interface is not autogenerated as the f90
3270: interface defintion cannot be generated correctly [due to MatFactorInfo]
3272: @*/
3273: PetscErrorCode MatCholeskyFactorSymbolic(Mat fact,Mat mat,IS perm,const MatFactorInfo *info)
3274: {
3276: MatFactorInfo tinfo;
3284: if (mat->rmap->N != mat->cmap->N) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONG,"Matrix must be square");
3285: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3286: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3287: if (!(fact)->ops->choleskyfactorsymbolic) {
3288: MatSolverType stype;
3289: MatFactorGetSolverType(fact,&stype);
3290: SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s symbolic factor Cholesky using solver package %s",((PetscObject)mat)->type_name,stype);
3291: }
3292: MatCheckPreallocated(mat,2);
3293: if (!info) {
3294: MatFactorInfoInitialize(&tinfo);
3295: info = &tinfo;
3296: }
3298: PetscLogEventBegin(MAT_CholeskyFactorSymbolic,mat,perm,0,0);
3299: (fact->ops->choleskyfactorsymbolic)(fact,mat,perm,info);
3300: PetscLogEventEnd(MAT_CholeskyFactorSymbolic,mat,perm,0,0);
3301: PetscObjectStateIncrease((PetscObject)fact);
3302: return(0);
3303: }
3305: /*@C
3306: MatCholeskyFactorNumeric - Performs numeric Cholesky factorization
3307: of a symmetric matrix. Call this routine after first calling
3308: MatCholeskyFactorSymbolic().
3310: Collective on Mat
3312: Input Parameters:
3313: + fact - the factor matrix obtained with MatGetFactor()
3314: . mat - the initial matrix
3315: . info - options for factorization
3316: - fact - the symbolic factor of mat
3319: Notes:
3320: Most users should employ the simplified KSP interface for linear solvers
3321: instead of working directly with matrix algebra routines such as this.
3322: See, e.g., KSPCreate().
3324: Level: developer
3326: .seealso: MatCholeskyFactorSymbolic(), MatCholeskyFactor(), MatLUFactorNumeric()
3328: Developer Note: fortran interface is not autogenerated as the f90
3329: interface defintion cannot be generated correctly [due to MatFactorInfo]
3331: @*/
3332: PetscErrorCode MatCholeskyFactorNumeric(Mat fact,Mat mat,const MatFactorInfo *info)
3333: {
3334: MatFactorInfo tinfo;
3342: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3343: if (!(fact)->ops->choleskyfactornumeric) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s numeric factor Cholesky",((PetscObject)mat)->type_name);
3344: if (mat->rmap->N != (fact)->rmap->N || mat->cmap->N != (fact)->cmap->N) SETERRQ4(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Mat fact: global dim %D should = %D %D should = %D",mat->rmap->N,(fact)->rmap->N,mat->cmap->N,(fact)->cmap->N);
3345: MatCheckPreallocated(mat,2);
3346: if (!info) {
3347: MatFactorInfoInitialize(&tinfo);
3348: info = &tinfo;
3349: }
3351: PetscLogEventBegin(MAT_CholeskyFactorNumeric,mat,fact,0,0);
3352: (fact->ops->choleskyfactornumeric)(fact,mat,info);
3353: PetscLogEventEnd(MAT_CholeskyFactorNumeric,mat,fact,0,0);
3354: MatViewFromOptions(fact,NULL,"-mat_factor_view");
3355: PetscObjectStateIncrease((PetscObject)fact);
3356: return(0);
3357: }
3359: /* ----------------------------------------------------------------*/
3360: /*@
3361: MatSolve - Solves A x = b, given a factored matrix.
3363: Neighbor-wise Collective on Mat
3365: Input Parameters:
3366: + mat - the factored matrix
3367: - b - the right-hand-side vector
3369: Output Parameter:
3370: . x - the result vector
3372: Notes:
3373: The vectors b and x cannot be the same. I.e., one cannot
3374: call MatSolve(A,x,x).
3376: Notes:
3377: Most users should employ the simplified KSP interface for linear solvers
3378: instead of working directly with matrix algebra routines such as this.
3379: See, e.g., KSPCreate().
3381: Level: developer
3383: .seealso: MatSolveAdd(), MatSolveTranspose(), MatSolveTransposeAdd()
3384: @*/
3385: PetscErrorCode MatSolve(Mat mat,Vec b,Vec x)
3386: {
3396: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3397: if (mat->cmap->N != x->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
3398: if (mat->rmap->N != b->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap->N,b->map->N);
3399: if (mat->rmap->n != b->map->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap->n,b->map->n);
3400: if (!mat->rmap->N && !mat->cmap->N) return(0);
3401: MatCheckPreallocated(mat,1);
3403: PetscLogEventBegin(MAT_Solve,mat,b,x,0);
3404: if (mat->factorerrortype) {
3405: PetscInfo1(mat,"MatFactorError %D\n",mat->factorerrortype);
3406: VecSetInf(x);
3407: } else {
3408: if (!mat->ops->solve) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3409: (*mat->ops->solve)(mat,b,x);
3410: }
3411: PetscLogEventEnd(MAT_Solve,mat,b,x,0);
3412: PetscObjectStateIncrease((PetscObject)x);
3413: return(0);
3414: }
3416: static PetscErrorCode MatMatSolve_Basic(Mat A,Mat B,Mat X,PetscBool trans)
3417: {
3419: Vec b,x;
3420: PetscInt m,N,i;
3421: PetscScalar *bb,*xx;
3424: MatDenseGetArrayRead(B,(const PetscScalar**)&bb);
3425: MatDenseGetArray(X,&xx);
3426: MatGetLocalSize(B,&m,NULL); /* number local rows */
3427: MatGetSize(B,NULL,&N); /* total columns in dense matrix */
3428: MatCreateVecs(A,&x,&b);
3429: for (i=0; i<N; i++) {
3430: VecPlaceArray(b,bb + i*m);
3431: VecPlaceArray(x,xx + i*m);
3432: if (trans) {
3433: MatSolveTranspose(A,b,x);
3434: } else {
3435: MatSolve(A,b,x);
3436: }
3437: VecResetArray(x);
3438: VecResetArray(b);
3439: }
3440: VecDestroy(&b);
3441: VecDestroy(&x);
3442: MatDenseRestoreArrayRead(B,(const PetscScalar**)&bb);
3443: MatDenseRestoreArray(X,&xx);
3444: return(0);
3445: }
3447: /*@
3448: MatMatSolve - Solves A X = B, given a factored matrix.
3450: Neighbor-wise Collective on Mat
3452: Input Parameters:
3453: + A - the factored matrix
3454: - B - the right-hand-side matrix MATDENSE (or sparse -- when using MUMPS)
3456: Output Parameter:
3457: . X - the result matrix (dense matrix)
3459: Notes:
3460: If B is a MATDENSE matrix then one can call MatMatSolve(A,B,B) except with MKL_CPARDISO;
3461: otherwise, B and X cannot be the same.
3463: Notes:
3464: Most users should usually employ the simplified KSP interface for linear solvers
3465: instead of working directly with matrix algebra routines such as this.
3466: See, e.g., KSPCreate(). However KSP can only solve for one vector (column of X)
3467: at a time.
3469: Level: developer
3471: .seealso: MatMatSolveTranspose(), MatLUFactor(), MatCholeskyFactor()
3472: @*/
3473: PetscErrorCode MatMatSolve(Mat A,Mat B,Mat X)
3474: {
3484: if (A->cmap->N != X->rmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Mat A,Mat X: global dim %D %D",A->cmap->N,X->rmap->N);
3485: if (A->rmap->N != B->rmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Mat A,Mat B: global dim %D %D",A->rmap->N,B->rmap->N);
3486: if (X->cmap->N != B->cmap->N) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Solution matrix must have same number of columns as rhs matrix");
3487: if (!A->rmap->N && !A->cmap->N) return(0);
3488: if (!A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
3489: MatCheckPreallocated(A,1);
3491: PetscLogEventBegin(MAT_MatSolve,A,B,X,0);
3492: if (!A->ops->matsolve) {
3493: PetscInfo1(A,"Mat type %s using basic MatMatSolve\n",((PetscObject)A)->type_name);
3494: MatMatSolve_Basic(A,B,X,PETSC_FALSE);
3495: } else {
3496: (*A->ops->matsolve)(A,B,X);
3497: }
3498: PetscLogEventEnd(MAT_MatSolve,A,B,X,0);
3499: PetscObjectStateIncrease((PetscObject)X);
3500: return(0);
3501: }
3503: /*@
3504: MatMatSolveTranspose - Solves A^T X = B, given a factored matrix.
3506: Neighbor-wise Collective on Mat
3508: Input Parameters:
3509: + A - the factored matrix
3510: - B - the right-hand-side matrix (dense matrix)
3512: Output Parameter:
3513: . X - the result matrix (dense matrix)
3515: Notes:
3516: The matrices B and X cannot be the same. I.e., one cannot
3517: call MatMatSolveTranspose(A,X,X).
3519: Notes:
3520: Most users should usually employ the simplified KSP interface for linear solvers
3521: instead of working directly with matrix algebra routines such as this.
3522: See, e.g., KSPCreate(). However KSP can only solve for one vector (column of X)
3523: at a time.
3525: When using SuperLU_Dist or MUMPS as a parallel solver, PETSc will use their functionality to solve multiple right hand sides simultaneously.
3527: Level: developer
3529: .seealso: MatMatSolve(), MatLUFactor(), MatCholeskyFactor()
3530: @*/
3531: PetscErrorCode MatMatSolveTranspose(Mat A,Mat B,Mat X)
3532: {
3542: if (X == B) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_IDN,"X and B must be different matrices");
3543: if (A->cmap->N != X->rmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Mat A,Mat X: global dim %D %D",A->cmap->N,X->rmap->N);
3544: if (A->rmap->N != B->rmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Mat A,Mat B: global dim %D %D",A->rmap->N,B->rmap->N);
3545: if (A->rmap->n != B->rmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat A,Mat B: local dim %D %D",A->rmap->n,B->rmap->n);
3546: if (X->cmap->N < B->cmap->N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Solution matrix must have same number of columns as rhs matrix");
3547: if (!A->rmap->N && !A->cmap->N) return(0);
3548: if (!A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
3549: MatCheckPreallocated(A,1);
3551: PetscLogEventBegin(MAT_MatSolve,A,B,X,0);
3552: if (!A->ops->matsolvetranspose) {
3553: PetscInfo1(A,"Mat type %s using basic MatMatSolveTranspose\n",((PetscObject)A)->type_name);
3554: MatMatSolve_Basic(A,B,X,PETSC_TRUE);
3555: } else {
3556: (*A->ops->matsolvetranspose)(A,B,X);
3557: }
3558: PetscLogEventEnd(MAT_MatSolve,A,B,X,0);
3559: PetscObjectStateIncrease((PetscObject)X);
3560: return(0);
3561: }
3563: /*@
3564: MatMatTransposeSolve - Solves A X = B^T, given a factored matrix.
3566: Neighbor-wise Collective on Mat
3568: Input Parameters:
3569: + A - the factored matrix
3570: - Bt - the transpose of right-hand-side matrix
3572: Output Parameter:
3573: . X - the result matrix (dense matrix)
3575: Notes:
3576: Most users should usually employ the simplified KSP interface for linear solvers
3577: instead of working directly with matrix algebra routines such as this.
3578: See, e.g., KSPCreate(). However KSP can only solve for one vector (column of X)
3579: at a time.
3581: 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 format on the host processor and call MatMatTransposeSolve() to implement MUMPS' MatMatSolve().
3583: Level: developer
3585: .seealso: MatMatSolve(), MatMatSolveTranspose(), MatLUFactor(), MatCholeskyFactor()
3586: @*/
3587: PetscErrorCode MatMatTransposeSolve(Mat A,Mat Bt,Mat X)
3588: {
3599: if (X == Bt) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_IDN,"X and B must be different matrices");
3600: if (A->cmap->N != X->rmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Mat A,Mat X: global dim %D %D",A->cmap->N,X->rmap->N);
3601: if (A->rmap->N != Bt->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Mat A,Mat Bt: global dim %D %D",A->rmap->N,Bt->cmap->N);
3602: if (X->cmap->N < Bt->rmap->N) SETERRQ(PetscObjectComm((PetscObject)X),PETSC_ERR_ARG_SIZ,"Solution matrix must have same number of columns as row number of the rhs matrix");
3603: if (!A->rmap->N && !A->cmap->N) return(0);
3604: if (!A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
3605: MatCheckPreallocated(A,1);
3607: if (!A->ops->mattransposesolve) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Mat type %s",((PetscObject)A)->type_name);
3608: PetscLogEventBegin(MAT_MatTrSolve,A,Bt,X,0);
3609: (*A->ops->mattransposesolve)(A,Bt,X);
3610: PetscLogEventEnd(MAT_MatTrSolve,A,Bt,X,0);
3611: PetscObjectStateIncrease((PetscObject)X);
3612: return(0);
3613: }
3615: /*@
3616: MatForwardSolve - Solves L x = b, given a factored matrix, A = LU, or
3617: U^T*D^(1/2) x = b, given a factored symmetric matrix, A = U^T*D*U,
3619: Neighbor-wise Collective on Mat
3621: Input Parameters:
3622: + mat - the factored matrix
3623: - b - the right-hand-side vector
3625: Output Parameter:
3626: . x - the result vector
3628: Notes:
3629: MatSolve() should be used for most applications, as it performs
3630: a forward solve followed by a backward solve.
3632: The vectors b and x cannot be the same, i.e., one cannot
3633: call MatForwardSolve(A,x,x).
3635: For matrix in seqsbaij format with block size larger than 1,
3636: the diagonal blocks are not implemented as D = D^(1/2) * D^(1/2) yet.
3637: MatForwardSolve() solves U^T*D y = b, and
3638: MatBackwardSolve() solves U x = y.
3639: Thus they do not provide a symmetric preconditioner.
3641: Most users should employ the simplified KSP interface for linear solvers
3642: instead of working directly with matrix algebra routines such as this.
3643: See, e.g., KSPCreate().
3645: Level: developer
3647: .seealso: MatSolve(), MatBackwardSolve()
3648: @*/
3649: PetscErrorCode MatForwardSolve(Mat mat,Vec b,Vec x)
3650: {
3660: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3661: if (mat->cmap->N != x->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
3662: if (mat->rmap->N != b->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap->N,b->map->N);
3663: if (mat->rmap->n != b->map->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap->n,b->map->n);
3664: if (!mat->rmap->N && !mat->cmap->N) return(0);
3665: MatCheckPreallocated(mat,1);
3667: if (!mat->ops->forwardsolve) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3668: PetscLogEventBegin(MAT_ForwardSolve,mat,b,x,0);
3669: (*mat->ops->forwardsolve)(mat,b,x);
3670: PetscLogEventEnd(MAT_ForwardSolve,mat,b,x,0);
3671: PetscObjectStateIncrease((PetscObject)x);
3672: return(0);
3673: }
3675: /*@
3676: MatBackwardSolve - Solves U x = b, given a factored matrix, A = LU.
3677: D^(1/2) U x = b, given a factored symmetric matrix, A = U^T*D*U,
3679: Neighbor-wise Collective on Mat
3681: Input Parameters:
3682: + mat - the factored matrix
3683: - b - the right-hand-side vector
3685: Output Parameter:
3686: . x - the result vector
3688: Notes:
3689: MatSolve() should be used for most applications, as it performs
3690: a forward solve followed by a backward solve.
3692: The vectors b and x cannot be the same. I.e., one cannot
3693: call MatBackwardSolve(A,x,x).
3695: For matrix in seqsbaij format with block size larger than 1,
3696: the diagonal blocks are not implemented as D = D^(1/2) * D^(1/2) yet.
3697: MatForwardSolve() solves U^T*D y = b, and
3698: MatBackwardSolve() solves U x = y.
3699: Thus they do not provide a symmetric preconditioner.
3701: Most users should employ the simplified KSP interface for linear solvers
3702: instead of working directly with matrix algebra routines such as this.
3703: See, e.g., KSPCreate().
3705: Level: developer
3707: .seealso: MatSolve(), MatForwardSolve()
3708: @*/
3709: PetscErrorCode MatBackwardSolve(Mat mat,Vec b,Vec x)
3710: {
3720: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3721: if (mat->cmap->N != x->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
3722: if (mat->rmap->N != b->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap->N,b->map->N);
3723: if (mat->rmap->n != b->map->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap->n,b->map->n);
3724: if (!mat->rmap->N && !mat->cmap->N) return(0);
3725: MatCheckPreallocated(mat,1);
3727: if (!mat->ops->backwardsolve) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3728: PetscLogEventBegin(MAT_BackwardSolve,mat,b,x,0);
3729: (*mat->ops->backwardsolve)(mat,b,x);
3730: PetscLogEventEnd(MAT_BackwardSolve,mat,b,x,0);
3731: PetscObjectStateIncrease((PetscObject)x);
3732: return(0);
3733: }
3735: /*@
3736: MatSolveAdd - Computes x = y + inv(A)*b, given a factored matrix.
3738: Neighbor-wise Collective on Mat
3740: Input Parameters:
3741: + mat - the factored matrix
3742: . b - the right-hand-side vector
3743: - y - the vector to be added to
3745: Output Parameter:
3746: . x - the result vector
3748: Notes:
3749: The vectors b and x cannot be the same. I.e., one cannot
3750: call MatSolveAdd(A,x,y,x).
3752: Most users should employ the simplified KSP interface for linear solvers
3753: instead of working directly with matrix algebra routines such as this.
3754: See, e.g., KSPCreate().
3756: Level: developer
3758: .seealso: MatSolve(), MatSolveTranspose(), MatSolveTransposeAdd()
3759: @*/
3760: PetscErrorCode MatSolveAdd(Mat mat,Vec b,Vec y,Vec x)
3761: {
3762: PetscScalar one = 1.0;
3763: Vec tmp;
3775: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3776: if (mat->cmap->N != x->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
3777: if (mat->rmap->N != b->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap->N,b->map->N);
3778: if (mat->rmap->N != y->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->rmap->N,y->map->N);
3779: if (mat->rmap->n != b->map->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap->n,b->map->n);
3780: if (x->map->n != y->map->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Vec x,Vec y: local dim %D %D",x->map->n,y->map->n);
3781: if (!mat->rmap->N && !mat->cmap->N) return(0);
3782: MatCheckPreallocated(mat,1);
3784: PetscLogEventBegin(MAT_SolveAdd,mat,b,x,y);
3785: if (mat->factorerrortype) {
3786: PetscInfo1(mat,"MatFactorError %D\n",mat->factorerrortype);
3787: VecSetInf(x);
3788: } else if (mat->ops->solveadd) {
3789: (*mat->ops->solveadd)(mat,b,y,x);
3790: } else {
3791: /* do the solve then the add manually */
3792: if (x != y) {
3793: MatSolve(mat,b,x);
3794: VecAXPY(x,one,y);
3795: } else {
3796: VecDuplicate(x,&tmp);
3797: PetscLogObjectParent((PetscObject)mat,(PetscObject)tmp);
3798: VecCopy(x,tmp);
3799: MatSolve(mat,b,x);
3800: VecAXPY(x,one,tmp);
3801: VecDestroy(&tmp);
3802: }
3803: }
3804: PetscLogEventEnd(MAT_SolveAdd,mat,b,x,y);
3805: PetscObjectStateIncrease((PetscObject)x);
3806: return(0);
3807: }
3809: /*@
3810: MatSolveTranspose - Solves A' x = b, given a factored matrix.
3812: Neighbor-wise Collective on Mat
3814: Input Parameters:
3815: + mat - the factored matrix
3816: - b - the right-hand-side vector
3818: Output Parameter:
3819: . x - the result vector
3821: Notes:
3822: The vectors b and x cannot be the same. I.e., one cannot
3823: call MatSolveTranspose(A,x,x).
3825: Most users should employ the simplified KSP interface for linear solvers
3826: instead of working directly with matrix algebra routines such as this.
3827: See, e.g., KSPCreate().
3829: Level: developer
3831: .seealso: MatSolve(), MatSolveAdd(), MatSolveTransposeAdd()
3832: @*/
3833: PetscErrorCode MatSolveTranspose(Mat mat,Vec b,Vec x)
3834: {
3844: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3845: if (mat->rmap->N != x->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->rmap->N,x->map->N);
3846: if (mat->cmap->N != b->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->cmap->N,b->map->N);
3847: if (!mat->rmap->N && !mat->cmap->N) return(0);
3848: MatCheckPreallocated(mat,1);
3849: PetscLogEventBegin(MAT_SolveTranspose,mat,b,x,0);
3850: if (mat->factorerrortype) {
3851: PetscInfo1(mat,"MatFactorError %D\n",mat->factorerrortype);
3852: VecSetInf(x);
3853: } else {
3854: if (!mat->ops->solvetranspose) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s",((PetscObject)mat)->type_name);
3855: (*mat->ops->solvetranspose)(mat,b,x);
3856: }
3857: PetscLogEventEnd(MAT_SolveTranspose,mat,b,x,0);
3858: PetscObjectStateIncrease((PetscObject)x);
3859: return(0);
3860: }
3862: /*@
3863: MatSolveTransposeAdd - Computes x = y + inv(Transpose(A)) b, given a
3864: factored matrix.
3866: Neighbor-wise Collective on Mat
3868: Input Parameters:
3869: + mat - the factored matrix
3870: . b - the right-hand-side vector
3871: - y - the vector to be added to
3873: Output Parameter:
3874: . x - the result vector
3876: Notes:
3877: The vectors b and x cannot be the same. I.e., one cannot
3878: call MatSolveTransposeAdd(A,x,y,x).
3880: Most users should employ the simplified KSP interface for linear solvers
3881: instead of working directly with matrix algebra routines such as this.
3882: See, e.g., KSPCreate().
3884: Level: developer
3886: .seealso: MatSolve(), MatSolveAdd(), MatSolveTranspose()
3887: @*/
3888: PetscErrorCode MatSolveTransposeAdd(Mat mat,Vec b,Vec y,Vec x)
3889: {
3890: PetscScalar one = 1.0;
3892: Vec tmp;
3903: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3904: if (mat->rmap->N != x->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->rmap->N,x->map->N);
3905: if (mat->cmap->N != b->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->cmap->N,b->map->N);
3906: if (mat->cmap->N != y->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->cmap->N,y->map->N);
3907: if (x->map->n != y->map->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Vec x,Vec y: local dim %D %D",x->map->n,y->map->n);
3908: if (!mat->rmap->N && !mat->cmap->N) return(0);
3909: MatCheckPreallocated(mat,1);
3911: PetscLogEventBegin(MAT_SolveTransposeAdd,mat,b,x,y);
3912: if (mat->factorerrortype) {
3913: PetscInfo1(mat,"MatFactorError %D\n",mat->factorerrortype);
3914: VecSetInf(x);
3915: } else if (mat->ops->solvetransposeadd){
3916: (*mat->ops->solvetransposeadd)(mat,b,y,x);
3917: } else {
3918: /* do the solve then the add manually */
3919: if (x != y) {
3920: MatSolveTranspose(mat,b,x);
3921: VecAXPY(x,one,y);
3922: } else {
3923: VecDuplicate(x,&tmp);
3924: PetscLogObjectParent((PetscObject)mat,(PetscObject)tmp);
3925: VecCopy(x,tmp);
3926: MatSolveTranspose(mat,b,x);
3927: VecAXPY(x,one,tmp);
3928: VecDestroy(&tmp);
3929: }
3930: }
3931: PetscLogEventEnd(MAT_SolveTransposeAdd,mat,b,x,y);
3932: PetscObjectStateIncrease((PetscObject)x);
3933: return(0);
3934: }
3935: /* ----------------------------------------------------------------*/
3937: /*@
3938: MatSOR - Computes relaxation (SOR, Gauss-Seidel) sweeps.
3940: Neighbor-wise Collective on Mat
3942: Input Parameters:
3943: + mat - the matrix
3944: . b - the right hand side
3945: . omega - the relaxation factor
3946: . flag - flag indicating the type of SOR (see below)
3947: . shift - diagonal shift
3948: . its - the number of iterations
3949: - lits - the number of local iterations
3951: Output Parameters:
3952: . x - the solution (can contain an initial guess, use option SOR_ZERO_INITIAL_GUESS to indicate no guess)
3954: SOR Flags:
3955: + SOR_FORWARD_SWEEP - forward SOR
3956: . SOR_BACKWARD_SWEEP - backward SOR
3957: . SOR_SYMMETRIC_SWEEP - SSOR (symmetric SOR)
3958: . SOR_LOCAL_FORWARD_SWEEP - local forward SOR
3959: . SOR_LOCAL_BACKWARD_SWEEP - local forward SOR
3960: . SOR_LOCAL_SYMMETRIC_SWEEP - local SSOR
3961: . SOR_APPLY_UPPER, SOR_APPLY_LOWER - applies
3962: upper/lower triangular part of matrix to
3963: vector (with omega)
3964: - SOR_ZERO_INITIAL_GUESS - zero initial guess
3966: Notes:
3967: SOR_LOCAL_FORWARD_SWEEP, SOR_LOCAL_BACKWARD_SWEEP, and
3968: SOR_LOCAL_SYMMETRIC_SWEEP perform separate independent smoothings
3969: on each processor.
3971: Application programmers will not generally use MatSOR() directly,
3972: but instead will employ the KSP/PC interface.
3974: Notes:
3975: for BAIJ, SBAIJ, and AIJ matrices with Inodes this does a block SOR smoothing, otherwise it does a pointwise smoothing
3977: Notes for Advanced Users:
3978: The flags are implemented as bitwise inclusive or operations.
3979: For example, use (SOR_ZERO_INITIAL_GUESS | SOR_SYMMETRIC_SWEEP)
3980: to specify a zero initial guess for SSOR.
3982: Most users should employ the simplified KSP interface for linear solvers
3983: instead of working directly with matrix algebra routines such as this.
3984: See, e.g., KSPCreate().
3986: Vectors x and b CANNOT be the same
3988: Developer Note: We should add block SOR support for AIJ matrices with block size set to great than one and no inodes
3990: Level: developer
3992: @*/
3993: PetscErrorCode MatSOR(Mat mat,Vec b,PetscReal omega,MatSORType flag,PetscReal shift,PetscInt its,PetscInt lits,Vec x)
3994: {
4004: if (!mat->ops->sor) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4005: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4006: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4007: if (mat->cmap->N != x->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
4008: if (mat->rmap->N != b->map->N) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap->N,b->map->N);
4009: if (mat->rmap->n != b->map->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap->n,b->map->n);
4010: if (its <= 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Relaxation requires global its %D positive",its);
4011: if (lits <= 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Relaxation requires local its %D positive",lits);
4012: if (b == x) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_IDN,"b and x vector cannot be the same");
4014: MatCheckPreallocated(mat,1);
4015: PetscLogEventBegin(MAT_SOR,mat,b,x,0);
4016: ierr =(*mat->ops->sor)(mat,b,omega,flag,shift,its,lits,x);
4017: PetscLogEventEnd(MAT_SOR,mat,b,x,0);
4018: PetscObjectStateIncrease((PetscObject)x);
4019: return(0);
4020: }
4022: /*
4023: Default matrix copy routine.
4024: */
4025: PetscErrorCode MatCopy_Basic(Mat A,Mat B,MatStructure str)
4026: {
4027: PetscErrorCode ierr;
4028: PetscInt i,rstart = 0,rend = 0,nz;
4029: const PetscInt *cwork;
4030: const PetscScalar *vwork;
4033: if (B->assembled) {
4034: MatZeroEntries(B);
4035: }
4036: if (str == SAME_NONZERO_PATTERN) {
4037: MatGetOwnershipRange(A,&rstart,&rend);
4038: for (i=rstart; i<rend; i++) {
4039: MatGetRow(A,i,&nz,&cwork,&vwork);
4040: MatSetValues(B,1,&i,nz,cwork,vwork,INSERT_VALUES);
4041: MatRestoreRow(A,i,&nz,&cwork,&vwork);
4042: }
4043: } else {
4044: MatAYPX(B,0.0,A,str);
4045: }
4046: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
4047: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
4048: return(0);
4049: }
4051: /*@
4052: MatCopy - Copies a matrix to another matrix.
4054: Collective on Mat
4056: Input Parameters:
4057: + A - the matrix
4058: - str - SAME_NONZERO_PATTERN or DIFFERENT_NONZERO_PATTERN
4060: Output Parameter:
4061: . B - where the copy is put
4063: Notes:
4064: If you use SAME_NONZERO_PATTERN then the two matrices had better have the
4065: same nonzero pattern or the routine will crash.
4067: MatCopy() copies the matrix entries of a matrix to another existing
4068: matrix (after first zeroing the second matrix). A related routine is
4069: MatConvert(), which first creates a new matrix and then copies the data.
4071: Level: intermediate
4073: .seealso: MatConvert(), MatDuplicate()
4075: @*/
4076: PetscErrorCode MatCopy(Mat A,Mat B,MatStructure str)
4077: {
4079: PetscInt i;
4087: MatCheckPreallocated(B,2);
4088: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4089: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4090: if (A->rmap->N != B->rmap->N || A->cmap->N != B->cmap->N) SETERRQ4(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Mat A,Mat B: global dim (%D,%D) (%D,%D)",A->rmap->N,B->rmap->N,A->cmap->N,B->cmap->N);
4091: MatCheckPreallocated(A,1);
4092: if (A == B) return(0);
4094: PetscLogEventBegin(MAT_Copy,A,B,0,0);
4095: if (A->ops->copy) {
4096: (*A->ops->copy)(A,B,str);
4097: } else { /* generic conversion */
4098: MatCopy_Basic(A,B,str);
4099: }
4101: B->stencil.dim = A->stencil.dim;
4102: B->stencil.noc = A->stencil.noc;
4103: for (i=0; i<=A->stencil.dim; i++) {
4104: B->stencil.dims[i] = A->stencil.dims[i];
4105: B->stencil.starts[i] = A->stencil.starts[i];
4106: }
4108: PetscLogEventEnd(MAT_Copy,A,B,0,0);
4109: PetscObjectStateIncrease((PetscObject)B);
4110: return(0);
4111: }
4113: /*@C
4114: MatConvert - Converts a matrix to another matrix, either of the same
4115: or different type.
4117: Collective on Mat
4119: Input Parameters:
4120: + mat - the matrix
4121: . newtype - new matrix type. Use MATSAME to create a new matrix of the
4122: same type as the original matrix.
4123: - reuse - denotes if the destination matrix is to be created or reused.
4124: 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
4125: 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).
4127: Output Parameter:
4128: . M - pointer to place new matrix
4130: Notes:
4131: MatConvert() first creates a new matrix and then copies the data from
4132: the first matrix. A related routine is MatCopy(), which copies the matrix
4133: entries of one matrix to another already existing matrix context.
4135: Cannot be used to convert a sequential matrix to parallel or parallel to sequential,
4136: the MPI communicator of the generated matrix is always the same as the communicator
4137: of the input matrix.
4139: Level: intermediate
4141: .seealso: MatCopy(), MatDuplicate()
4142: @*/
4143: PetscErrorCode MatConvert(Mat mat, MatType newtype,MatReuse reuse,Mat *M)
4144: {
4146: PetscBool sametype,issame,flg,issymmetric,ishermitian;
4147: char convname[256],mtype[256];
4148: Mat B;
4154: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4155: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4156: MatCheckPreallocated(mat,1);
4158: PetscOptionsGetString(((PetscObject)mat)->options,((PetscObject)mat)->prefix,"-matconvert_type",mtype,sizeof(mtype),&flg);
4159: if (flg) newtype = mtype;
4161: PetscObjectTypeCompare((PetscObject)mat,newtype,&sametype);
4162: PetscStrcmp(newtype,"same",&issame);
4163: if ((reuse == MAT_INPLACE_MATRIX) && (mat != *M)) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"MAT_INPLACE_MATRIX requires same input and output matrix");
4164: if ((reuse == MAT_REUSE_MATRIX) && (mat == *M)) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"MAT_REUSE_MATRIX means reuse matrix in final argument, perhaps you mean MAT_INPLACE_MATRIX");
4166: if ((reuse == MAT_INPLACE_MATRIX) && (issame || sametype)) {
4167: PetscInfo3(mat,"Early return for inplace %s %d %d\n",((PetscObject)mat)->type_name,sametype,issame);
4168: return(0);
4169: }
4171: /* Cache Mat options because some converter use MatHeaderReplace */
4172: issymmetric = mat->symmetric;
4173: ishermitian = mat->hermitian;
4175: if ((sametype || issame) && (reuse==MAT_INITIAL_MATRIX) && mat->ops->duplicate) {
4176: PetscInfo3(mat,"Calling duplicate for initial matrix %s %d %d\n",((PetscObject)mat)->type_name,sametype,issame);
4177: (*mat->ops->duplicate)(mat,MAT_COPY_VALUES,M);
4178: } else {
4179: PetscErrorCode (*conv)(Mat, MatType,MatReuse,Mat*)=NULL;
4180: const char *prefix[3] = {"seq","mpi",""};
4181: PetscInt i;
4182: /*
4183: Order of precedence:
4184: 0) See if newtype is a superclass of the current matrix.
4185: 1) See if a specialized converter is known to the current matrix.
4186: 2) See if a specialized converter is known to the desired matrix class.
4187: 3) See if a good general converter is registered for the desired class
4188: (as of 6/27/03 only MATMPIADJ falls into this category).
4189: 4) See if a good general converter is known for the current matrix.
4190: 5) Use a really basic converter.
4191: */
4193: /* 0) See if newtype is a superclass of the current matrix.
4194: i.e mat is mpiaij and newtype is aij */
4195: for (i=0; i<2; i++) {
4196: PetscStrncpy(convname,prefix[i],sizeof(convname));
4197: PetscStrlcat(convname,newtype,sizeof(convname));
4198: PetscStrcmp(convname,((PetscObject)mat)->type_name,&flg);
4199: PetscInfo3(mat,"Check superclass %s %s -> %d\n",convname,((PetscObject)mat)->type_name,flg);
4200: if (flg) {
4201: if (reuse == MAT_INPLACE_MATRIX) {
4202: PetscInfo(mat,"Early return\n");
4203: return(0);
4204: } else if (reuse == MAT_INITIAL_MATRIX && mat->ops->duplicate) {
4205: PetscInfo(mat,"Calling MatDuplicate\n");
4206: (*mat->ops->duplicate)(mat,MAT_COPY_VALUES,M);
4207: return(0);
4208: } else if (reuse == MAT_REUSE_MATRIX && mat->ops->copy) {
4209: PetscInfo(mat,"Calling MatCopy\n");
4210: MatCopy(mat,*M,SAME_NONZERO_PATTERN);
4211: return(0);
4212: }
4213: }
4214: }
4215: /* 1) See if a specialized converter is known to the current matrix and the desired class */
4216: for (i=0; i<3; i++) {
4217: PetscStrncpy(convname,"MatConvert_",sizeof(convname));
4218: PetscStrlcat(convname,((PetscObject)mat)->type_name,sizeof(convname));
4219: PetscStrlcat(convname,"_",sizeof(convname));
4220: PetscStrlcat(convname,prefix[i],sizeof(convname));
4221: PetscStrlcat(convname,issame ? ((PetscObject)mat)->type_name : newtype,sizeof(convname));
4222: PetscStrlcat(convname,"_C",sizeof(convname));
4223: PetscObjectQueryFunction((PetscObject)mat,convname,&conv);
4224: PetscInfo3(mat,"Check specialized (1) %s (%s) -> %d\n",convname,((PetscObject)mat)->type_name,!!conv);
4225: if (conv) goto foundconv;
4226: }
4228: /* 2) See if a specialized converter is known to the desired matrix class. */
4229: MatCreate(PetscObjectComm((PetscObject)mat),&B);
4230: MatSetSizes(B,mat->rmap->n,mat->cmap->n,mat->rmap->N,mat->cmap->N);
4231: MatSetType(B,newtype);
4232: for (i=0; i<3; i++) {
4233: PetscStrncpy(convname,"MatConvert_",sizeof(convname));
4234: PetscStrlcat(convname,((PetscObject)mat)->type_name,sizeof(convname));
4235: PetscStrlcat(convname,"_",sizeof(convname));
4236: PetscStrlcat(convname,prefix[i],sizeof(convname));
4237: PetscStrlcat(convname,newtype,sizeof(convname));
4238: PetscStrlcat(convname,"_C",sizeof(convname));
4239: PetscObjectQueryFunction((PetscObject)B,convname,&conv);
4240: PetscInfo3(mat,"Check specialized (2) %s (%s) -> %d\n",convname,((PetscObject)B)->type_name,!!conv);
4241: if (conv) {
4242: MatDestroy(&B);
4243: goto foundconv;
4244: }
4245: }
4247: /* 3) See if a good general converter is registered for the desired class */
4248: conv = B->ops->convertfrom;
4249: PetscInfo2(mat,"Check convertfrom (%s) -> %d\n",((PetscObject)B)->type_name,!!conv);
4250: MatDestroy(&B);
4251: if (conv) goto foundconv;
4253: /* 4) See if a good general converter is known for the current matrix */
4254: if (mat->ops->convert) conv = mat->ops->convert;
4256: PetscInfo2(mat,"Check general convert (%s) -> %d\n",((PetscObject)mat)->type_name,!!conv);
4257: if (conv) goto foundconv;
4259: /* 5) Use a really basic converter. */
4260: PetscInfo(mat,"Using MatConvert_Basic\n");
4261: conv = MatConvert_Basic;
4263: foundconv:
4264: PetscLogEventBegin(MAT_Convert,mat,0,0,0);
4265: (*conv)(mat,newtype,reuse,M);
4266: if (mat->rmap->mapping && mat->cmap->mapping && !(*M)->rmap->mapping && !(*M)->cmap->mapping) {
4267: /* the block sizes must be same if the mappings are copied over */
4268: (*M)->rmap->bs = mat->rmap->bs;
4269: (*M)->cmap->bs = mat->cmap->bs;
4270: PetscObjectReference((PetscObject)mat->rmap->mapping);
4271: PetscObjectReference((PetscObject)mat->cmap->mapping);
4272: (*M)->rmap->mapping = mat->rmap->mapping;
4273: (*M)->cmap->mapping = mat->cmap->mapping;
4274: }
4275: (*M)->stencil.dim = mat->stencil.dim;
4276: (*M)->stencil.noc = mat->stencil.noc;
4277: for (i=0; i<=mat->stencil.dim; i++) {
4278: (*M)->stencil.dims[i] = mat->stencil.dims[i];
4279: (*M)->stencil.starts[i] = mat->stencil.starts[i];
4280: }
4281: PetscLogEventEnd(MAT_Convert,mat,0,0,0);
4282: }
4283: PetscObjectStateIncrease((PetscObject)*M);
4285: /* Copy Mat options */
4286: if (issymmetric) {
4287: MatSetOption(*M,MAT_SYMMETRIC,PETSC_TRUE);
4288: }
4289: if (ishermitian) {
4290: MatSetOption(*M,MAT_HERMITIAN,PETSC_TRUE);
4291: }
4292: return(0);
4293: }
4295: /*@C
4296: MatFactorGetSolverType - Returns name of the package providing the factorization routines
4298: Not Collective
4300: Input Parameter:
4301: . mat - the matrix, must be a factored matrix
4303: Output Parameter:
4304: . type - the string name of the package (do not free this string)
4306: Notes:
4307: In Fortran you pass in a empty string and the package name will be copied into it.
4308: (Make sure the string is long enough)
4310: Level: intermediate
4312: .seealso: MatCopy(), MatDuplicate(), MatGetFactorAvailable(), MatGetFactor()
4313: @*/
4314: PetscErrorCode MatFactorGetSolverType(Mat mat, MatSolverType *type)
4315: {
4316: PetscErrorCode ierr, (*conv)(Mat,MatSolverType*);
4321: if (!mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Only for factored matrix");
4322: PetscObjectQueryFunction((PetscObject)mat,"MatFactorGetSolverType_C",&conv);
4323: if (!conv) {
4324: *type = MATSOLVERPETSC;
4325: } else {
4326: (*conv)(mat,type);
4327: }
4328: return(0);
4329: }
4331: typedef struct _MatSolverTypeForSpecifcType* MatSolverTypeForSpecifcType;
4332: struct _MatSolverTypeForSpecifcType {
4333: MatType mtype;
4334: PetscErrorCode (*createfactor[4])(Mat,MatFactorType,Mat*);
4335: MatSolverTypeForSpecifcType next;
4336: };
4338: typedef struct _MatSolverTypeHolder* MatSolverTypeHolder;
4339: struct _MatSolverTypeHolder {
4340: char *name;
4341: MatSolverTypeForSpecifcType handlers;
4342: MatSolverTypeHolder next;
4343: };
4345: static MatSolverTypeHolder MatSolverTypeHolders = NULL;
4347: /*@C
4348: MatSolveTypeRegister - Registers a MatSolverType that works for a particular matrix type
4350: Input Parameters:
4351: + package - name of the package, for example petsc or superlu
4352: . mtype - the matrix type that works with this package
4353: . ftype - the type of factorization supported by the package
4354: - createfactor - routine that will create the factored matrix ready to be used
4356: Level: intermediate
4358: .seealso: MatCopy(), MatDuplicate(), MatGetFactorAvailable(), MatGetFactor()
4359: @*/
4360: PetscErrorCode MatSolverTypeRegister(MatSolverType package,MatType mtype,MatFactorType ftype,PetscErrorCode (*createfactor)(Mat,MatFactorType,Mat*))
4361: {
4362: PetscErrorCode ierr;
4363: MatSolverTypeHolder next = MatSolverTypeHolders,prev = NULL;
4364: PetscBool flg;
4365: MatSolverTypeForSpecifcType inext,iprev = NULL;
4368: MatInitializePackage();
4369: if (!next) {
4370: PetscNew(&MatSolverTypeHolders);
4371: PetscStrallocpy(package,&MatSolverTypeHolders->name);
4372: PetscNew(&MatSolverTypeHolders->handlers);
4373: PetscStrallocpy(mtype,(char **)&MatSolverTypeHolders->handlers->mtype);
4374: MatSolverTypeHolders->handlers->createfactor[(int)ftype-1] = createfactor;
4375: return(0);
4376: }
4377: while (next) {
4378: PetscStrcasecmp(package,next->name,&flg);
4379: if (flg) {
4380: if (!next->handlers) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"MatSolverTypeHolder is missing handlers");
4381: inext = next->handlers;
4382: while (inext) {
4383: PetscStrcasecmp(mtype,inext->mtype,&flg);
4384: if (flg) {
4385: inext->createfactor[(int)ftype-1] = createfactor;
4386: return(0);
4387: }
4388: iprev = inext;
4389: inext = inext->next;
4390: }
4391: PetscNew(&iprev->next);
4392: PetscStrallocpy(mtype,(char **)&iprev->next->mtype);
4393: iprev->next->createfactor[(int)ftype-1] = createfactor;
4394: return(0);
4395: }
4396: prev = next;
4397: next = next->next;
4398: }
4399: PetscNew(&prev->next);
4400: PetscStrallocpy(package,&prev->next->name);
4401: PetscNew(&prev->next->handlers);
4402: PetscStrallocpy(mtype,(char **)&prev->next->handlers->mtype);
4403: prev->next->handlers->createfactor[(int)ftype-1] = createfactor;
4404: return(0);
4405: }
4407: /*@C
4408: MatSolveTypeGet - Gets the function that creates the factor matrix if it exist
4410: Input Parameters:
4411: + type - name of the package, for example petsc or superlu
4412: . ftype - the type of factorization supported by the type
4413: - mtype - the matrix type that works with this type
4415: Output Parameters:
4416: + foundtype - PETSC_TRUE if the type was registered
4417: . foundmtype - PETSC_TRUE if the type supports the requested mtype
4418: - createfactor - routine that will create the factored matrix ready to be used or NULL if not found
4420: Level: intermediate
4422: .seealso: MatCopy(), MatDuplicate(), MatGetFactorAvailable(), MatSolvePackageRegister), MatGetFactor()
4423: @*/
4424: PetscErrorCode MatSolverTypeGet(MatSolverType type,MatType mtype,MatFactorType ftype,PetscBool *foundtype,PetscBool *foundmtype,PetscErrorCode (**createfactor)(Mat,MatFactorType,Mat*))
4425: {
4426: PetscErrorCode ierr;
4427: MatSolverTypeHolder next = MatSolverTypeHolders;
4428: PetscBool flg;
4429: MatSolverTypeForSpecifcType inext;
4432: if (foundtype) *foundtype = PETSC_FALSE;
4433: if (foundmtype) *foundmtype = PETSC_FALSE;
4434: if (createfactor) *createfactor = NULL;
4436: if (type) {
4437: while (next) {
4438: PetscStrcasecmp(type,next->name,&flg);
4439: if (flg) {
4440: if (foundtype) *foundtype = PETSC_TRUE;
4441: inext = next->handlers;
4442: while (inext) {
4443: PetscStrbeginswith(mtype,inext->mtype,&flg);
4444: if (flg) {
4445: if (foundmtype) *foundmtype = PETSC_TRUE;
4446: if (createfactor) *createfactor = inext->createfactor[(int)ftype-1];
4447: return(0);
4448: }
4449: inext = inext->next;
4450: }
4451: }
4452: next = next->next;
4453: }
4454: } else {
4455: while (next) {
4456: inext = next->handlers;
4457: while (inext) {
4458: PetscStrbeginswith(mtype,inext->mtype,&flg);
4459: if (flg && inext->createfactor[(int)ftype-1]) {
4460: if (foundtype) *foundtype = PETSC_TRUE;
4461: if (foundmtype) *foundmtype = PETSC_TRUE;
4462: if (createfactor) *createfactor = inext->createfactor[(int)ftype-1];
4463: return(0);
4464: }
4465: inext = inext->next;
4466: }
4467: next = next->next;
4468: }
4469: }
4470: return(0);
4471: }
4473: PetscErrorCode MatSolverTypeDestroy(void)
4474: {
4475: PetscErrorCode ierr;
4476: MatSolverTypeHolder next = MatSolverTypeHolders,prev;
4477: MatSolverTypeForSpecifcType inext,iprev;
4480: while (next) {
4481: PetscFree(next->name);
4482: inext = next->handlers;
4483: while (inext) {
4484: PetscFree(inext->mtype);
4485: iprev = inext;
4486: inext = inext->next;
4487: PetscFree(iprev);
4488: }
4489: prev = next;
4490: next = next->next;
4491: PetscFree(prev);
4492: }
4493: MatSolverTypeHolders = NULL;
4494: return(0);
4495: }
4497: /*@C
4498: MatFactorGetUseOrdering - Indicates if the factorization uses the ordering provided in MatLUFactorSymbolic(), MatCholeskyFactorSymbolic()
4500: Logically Collective on Mat
4502: Input Parameters:
4503: . mat - the matrix
4505: Output Parameters:
4506: . flg - PETSC_TRUE if uses the ordering
4508: Notes:
4509: Most internal PETSc factorizations use the ordering past to the factorization routine but external
4510: packages do no, thus we want to skip the ordering when it is not needed.
4512: Level: developer
4514: .seealso: MatCopy(), MatDuplicate(), MatGetFactorAvailable(), MatGetFactor(), MatLUFactorSymbolic(), MatCholeskyFactorSymbolic()
4515: @*/
4516: PetscErrorCode MatFactorGetUseOrdering(Mat mat, PetscBool *flg)
4517: {
4519: *flg = mat->useordering;
4520: return(0);
4521: }
4523: /*@C
4524: MatGetFactor - Returns a matrix suitable to calls to MatXXFactorSymbolic()
4526: Collective on Mat
4528: Input Parameters:
4529: + mat - the matrix
4530: . type - name of solver type, for example, superlu, petsc (to use PETSc's default)
4531: - ftype - factor type, MAT_FACTOR_LU, MAT_FACTOR_CHOLESKY, MAT_FACTOR_ICC, MAT_FACTOR_ILU,
4533: Output Parameters:
4534: . f - the factor matrix used with MatXXFactorSymbolic() calls
4536: Notes:
4537: Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
4538: such as pastix, superlu, mumps etc.
4540: PETSc must have been ./configure to use the external solver, using the option --download-package
4542: Developer Notes:
4543: This should actually be called MatCreateFactor() since it creates a new factor object
4545: Level: intermediate
4547: .seealso: MatCopy(), MatDuplicate(), MatGetFactorAvailable(), MatFactorGetUseOrdering(), MatSolverTypeRegister()
4548: @*/
4549: PetscErrorCode MatGetFactor(Mat mat, MatSolverType type,MatFactorType ftype,Mat *f)
4550: {
4551: PetscErrorCode ierr,(*conv)(Mat,MatFactorType,Mat*);
4552: PetscBool foundtype,foundmtype;
4558: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4559: MatCheckPreallocated(mat,1);
4561: MatSolverTypeGet(type,((PetscObject)mat)->type_name,ftype,&foundtype,&foundmtype,&conv);
4562: if (!foundtype) {
4563: if (type) {
4564: SETERRQ4(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],((PetscObject)mat)->type_name,type);
4565: } else {
4566: SETERRQ2(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);
4567: }
4568: }
4569: if (!foundmtype) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_MISSING_FACTOR,"MatSolverType %s does not support matrix type %s",type,((PetscObject)mat)->type_name);
4570: if (!conv) SETERRQ3(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);
4572: (*conv)(mat,ftype,f);
4573: return(0);
4574: }
4576: /*@C
4577: MatGetFactorAvailable - Returns a a flag if matrix supports particular type and factor type
4579: Not Collective
4581: Input Parameters:
4582: + mat - the matrix
4583: . type - name of solver type, for example, superlu, petsc (to use PETSc's default)
4584: - ftype - factor type, MAT_FACTOR_LU, MAT_FACTOR_CHOLESKY, MAT_FACTOR_ICC, MAT_FACTOR_ILU,
4586: Output Parameter:
4587: . flg - PETSC_TRUE if the factorization is available
4589: Notes:
4590: Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
4591: such as pastix, superlu, mumps etc.
4593: PETSc must have been ./configure to use the external solver, using the option --download-package
4595: Developer Notes:
4596: This should actually be called MatCreateFactorAvailable() since MatGetFactor() creates a new factor object
4598: Level: intermediate
4600: .seealso: MatCopy(), MatDuplicate(), MatGetFactor(), MatSolverTypeRegister()
4601: @*/
4602: PetscErrorCode MatGetFactorAvailable(Mat mat, MatSolverType type,MatFactorType ftype,PetscBool *flg)
4603: {
4604: PetscErrorCode ierr, (*gconv)(Mat,MatFactorType,Mat*);
4610: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4611: MatCheckPreallocated(mat,1);
4613: *flg = PETSC_FALSE;
4614: MatSolverTypeGet(type,((PetscObject)mat)->type_name,ftype,NULL,NULL,&gconv);
4615: if (gconv) {
4616: *flg = PETSC_TRUE;
4617: }
4618: return(0);
4619: }
4621: #include <petscdmtypes.h>
4623: /*@
4624: MatDuplicate - Duplicates a matrix including the non-zero structure.
4626: Collective on Mat
4628: Input Parameters:
4629: + mat - the matrix
4630: - op - One of MAT_DO_NOT_COPY_VALUES, MAT_COPY_VALUES, or MAT_SHARE_NONZERO_PATTERN.
4631: See the manual page for MatDuplicateOption for an explanation of these options.
4633: Output Parameter:
4634: . M - pointer to place new matrix
4636: Level: intermediate
4638: Notes:
4639: You cannot change the nonzero pattern for the parent or child matrix if you use MAT_SHARE_NONZERO_PATTERN.
4640: 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 is duplicated and the internal data structures created for the reuse of previous matrix operations are not duplicated. User should not use MatDuplicate() to create new matrix M if M is intended to be reused as the product of matrix operation.
4642: .seealso: MatCopy(), MatConvert(), MatDuplicateOption
4643: @*/
4644: PetscErrorCode MatDuplicate(Mat mat,MatDuplicateOption op,Mat *M)
4645: {
4647: Mat B;
4648: PetscInt i;
4649: DM dm;
4650: void (*viewf)(void);
4656: if (op == MAT_COPY_VALUES && !mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"MAT_COPY_VALUES not allowed for unassembled matrix");
4657: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4658: MatCheckPreallocated(mat,1);
4660: *M = NULL;
4661: if (!mat->ops->duplicate) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Not written for matrix type %s\n",((PetscObject)mat)->type_name);
4662: PetscLogEventBegin(MAT_Convert,mat,0,0,0);
4663: (*mat->ops->duplicate)(mat,op,M);
4664: B = *M;
4666: MatGetOperation(mat,MATOP_VIEW,&viewf);
4667: if (viewf) {
4668: MatSetOperation(B,MATOP_VIEW,viewf);
4669: }
4671: B->stencil.dim = mat->stencil.dim;
4672: B->stencil.noc = mat->stencil.noc;
4673: for (i=0; i<=mat->stencil.dim; i++) {
4674: B->stencil.dims[i] = mat->stencil.dims[i];
4675: B->stencil.starts[i] = mat->stencil.starts[i];
4676: }
4678: B->nooffproczerorows = mat->nooffproczerorows;
4679: B->nooffprocentries = mat->nooffprocentries;
4681: PetscObjectQuery((PetscObject) mat, "__PETSc_dm", (PetscObject*) &dm);
4682: if (dm) {
4683: PetscObjectCompose((PetscObject) B, "__PETSc_dm", (PetscObject) dm);
4684: }
4685: PetscLogEventEnd(MAT_Convert,mat,0,0,0);
4686: PetscObjectStateIncrease((PetscObject)B);
4687: return(0);
4688: }
4690: /*@
4691: MatGetDiagonal - Gets the diagonal of a matrix.
4693: Logically Collective on Mat
4695: Input Parameters:
4696: + mat - the matrix
4697: - v - the vector for storing the diagonal
4699: Output Parameter:
4700: . v - the diagonal of the matrix
4702: Level: intermediate
4704: Note:
4705: Currently only correct in parallel for square matrices.
4707: .seealso: MatGetRow(), MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRowMaxAbs()
4708: @*/
4709: PetscErrorCode MatGetDiagonal(Mat mat,Vec v)
4710: {
4717: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4718: if (!mat->ops->getdiagonal) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4719: MatCheckPreallocated(mat,1);
4721: (*mat->ops->getdiagonal)(mat,v);
4722: PetscObjectStateIncrease((PetscObject)v);
4723: return(0);
4724: }
4726: /*@C
4727: MatGetRowMin - Gets the minimum value (of the real part) of each
4728: row of the matrix
4730: Logically Collective on Mat
4732: Input Parameters:
4733: . mat - the matrix
4735: Output Parameter:
4736: + v - the vector for storing the maximums
4737: - idx - the indices of the column found for each row (optional)
4739: Level: intermediate
4741: Notes:
4742: The result of this call are the same as if one converted the matrix to dense format
4743: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
4745: This code is only implemented for a couple of matrix formats.
4747: .seealso: MatGetDiagonal(), MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRowMaxAbs(),
4748: MatGetRowMax()
4749: @*/
4750: PetscErrorCode MatGetRowMin(Mat mat,Vec v,PetscInt idx[])
4751: {
4758: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4760: if (!mat->cmap->N) {
4761: VecSet(v,PETSC_MAX_REAL);
4762: if (idx) {
4763: PetscInt i,m = mat->rmap->n;
4764: for (i=0; i<m; i++) idx[i] = -1;
4765: }
4766: } else {
4767: if (!mat->ops->getrowmin) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4768: MatCheckPreallocated(mat,1);
4769: }
4770: (*mat->ops->getrowmin)(mat,v,idx);
4771: PetscObjectStateIncrease((PetscObject)v);
4772: return(0);
4773: }
4775: /*@C
4776: MatGetRowMinAbs - Gets the minimum value (in absolute value) of each
4777: row of the matrix
4779: Logically Collective on Mat
4781: Input Parameters:
4782: . mat - the matrix
4784: Output Parameter:
4785: + v - the vector for storing the minimums
4786: - idx - the indices of the column found for each row (or NULL if not needed)
4788: Level: intermediate
4790: Notes:
4791: if a row is completely empty or has only 0.0 values then the idx[] value for that
4792: row is 0 (the first column).
4794: This code is only implemented for a couple of matrix formats.
4796: .seealso: MatGetDiagonal(), MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRowMax(), MatGetRowMaxAbs(), MatGetRowMin()
4797: @*/
4798: PetscErrorCode MatGetRowMinAbs(Mat mat,Vec v,PetscInt idx[])
4799: {
4806: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4807: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4809: if (!mat->cmap->N) {
4810: VecSet(v,0.0);
4811: if (idx) {
4812: PetscInt i,m = mat->rmap->n;
4813: for (i=0; i<m; i++) idx[i] = -1;
4814: }
4815: } else {
4816: if (!mat->ops->getrowminabs) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4817: MatCheckPreallocated(mat,1);
4818: if (idx) {PetscArrayzero(idx,mat->rmap->n);}
4819: (*mat->ops->getrowminabs)(mat,v,idx);
4820: }
4821: PetscObjectStateIncrease((PetscObject)v);
4822: return(0);
4823: }
4825: /*@C
4826: MatGetRowMax - Gets the maximum value (of the real part) of each
4827: row of the matrix
4829: Logically Collective on Mat
4831: Input Parameters:
4832: . mat - the matrix
4834: Output Parameter:
4835: + v - the vector for storing the maximums
4836: - idx - the indices of the column found for each row (optional)
4838: Level: intermediate
4840: Notes:
4841: The result of this call are the same as if one converted the matrix to dense format
4842: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
4844: This code is only implemented for a couple of matrix formats.
4846: .seealso: MatGetDiagonal(), MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRowMaxAbs(), MatGetRowMin()
4847: @*/
4848: PetscErrorCode MatGetRowMax(Mat mat,Vec v,PetscInt idx[])
4849: {
4856: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4858: if (!mat->cmap->N) {
4859: VecSet(v,PETSC_MIN_REAL);
4860: if (idx) {
4861: PetscInt i,m = mat->rmap->n;
4862: for (i=0; i<m; i++) idx[i] = -1;
4863: }
4864: } else {
4865: if (!mat->ops->getrowmax) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4866: MatCheckPreallocated(mat,1);
4867: (*mat->ops->getrowmax)(mat,v,idx);
4868: }
4869: PetscObjectStateIncrease((PetscObject)v);
4870: return(0);
4871: }
4873: /*@C
4874: MatGetRowMaxAbs - Gets the maximum value (in absolute value) of each
4875: row of the matrix
4877: Logically Collective on Mat
4879: Input Parameters:
4880: . mat - the matrix
4882: Output Parameter:
4883: + v - the vector for storing the maximums
4884: - idx - the indices of the column found for each row (or NULL if not needed)
4886: Level: intermediate
4888: Notes:
4889: if a row is completely empty or has only 0.0 values then the idx[] value for that
4890: row is 0 (the first column).
4892: This code is only implemented for a couple of matrix formats.
4894: .seealso: MatGetDiagonal(), MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRowMax(), MatGetRowMin()
4895: @*/
4896: PetscErrorCode MatGetRowMaxAbs(Mat mat,Vec v,PetscInt idx[])
4897: {
4904: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4906: if (!mat->cmap->N) {
4907: VecSet(v,0.0);
4908: if (idx) {
4909: PetscInt i,m = mat->rmap->n;
4910: for (i=0; i<m; i++) idx[i] = -1;
4911: }
4912: } else {
4913: if (!mat->ops->getrowmaxabs) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4914: MatCheckPreallocated(mat,1);
4915: if (idx) {PetscArrayzero(idx,mat->rmap->n);}
4916: (*mat->ops->getrowmaxabs)(mat,v,idx);
4917: }
4918: PetscObjectStateIncrease((PetscObject)v);
4919: return(0);
4920: }
4922: /*@
4923: MatGetRowSum - Gets the sum of each row of the matrix
4925: Logically or Neighborhood Collective on Mat
4927: Input Parameters:
4928: . mat - the matrix
4930: Output Parameter:
4931: . v - the vector for storing the sum of rows
4933: Level: intermediate
4935: Notes:
4936: This code is slow since it is not currently specialized for different formats
4938: .seealso: MatGetDiagonal(), MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRowMax(), MatGetRowMin()
4939: @*/
4940: PetscErrorCode MatGetRowSum(Mat mat, Vec v)
4941: {
4942: Vec ones;
4949: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4950: MatCheckPreallocated(mat,1);
4951: MatCreateVecs(mat,&ones,NULL);
4952: VecSet(ones,1.);
4953: MatMult(mat,ones,v);
4954: VecDestroy(&ones);
4955: return(0);
4956: }
4958: /*@
4959: MatTranspose - Computes an in-place or out-of-place transpose of a matrix.
4961: Collective on Mat
4963: Input Parameter:
4964: + mat - the matrix to transpose
4965: - reuse - either MAT_INITIAL_MATRIX, MAT_REUSE_MATRIX, or MAT_INPLACE_MATRIX
4967: Output Parameters:
4968: . B - the transpose
4970: Notes:
4971: If you use MAT_INPLACE_MATRIX then you must pass in &mat for B
4973: MAT_REUSE_MATRIX causes the B matrix from a previous call to this function with MAT_INITIAL_MATRIX to be used
4975: Consider using MatCreateTranspose() instead if you only need a matrix that behaves like the transpose, but don't need the storage to be changed.
4977: Level: intermediate
4979: .seealso: MatMultTranspose(), MatMultTransposeAdd(), MatIsTranspose(), MatReuse
4980: @*/
4981: PetscErrorCode MatTranspose(Mat mat,MatReuse reuse,Mat *B)
4982: {
4988: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4989: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4990: if (!mat->ops->transpose) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4991: if (reuse == MAT_INPLACE_MATRIX && mat != *B) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"MAT_INPLACE_MATRIX requires last matrix to match first");
4992: if (reuse == MAT_REUSE_MATRIX && mat == *B) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Perhaps you mean MAT_INPLACE_MATRIX");
4993: MatCheckPreallocated(mat,1);
4995: PetscLogEventBegin(MAT_Transpose,mat,0,0,0);
4996: (*mat->ops->transpose)(mat,reuse,B);
4997: PetscLogEventEnd(MAT_Transpose,mat,0,0,0);
4998: if (B) {PetscObjectStateIncrease((PetscObject)*B);}
4999: return(0);
5000: }
5002: /*@
5003: MatIsTranspose - Test whether a matrix is another one's transpose,
5004: or its own, in which case it tests symmetry.
5006: Collective on Mat
5008: Input Parameter:
5009: + A - the matrix to test
5010: - B - the matrix to test against, this can equal the first parameter
5012: Output Parameters:
5013: . flg - the result
5015: Notes:
5016: Only available for SeqAIJ/MPIAIJ matrices. The sequential algorithm
5017: has a running time of the order of the number of nonzeros; the parallel
5018: test involves parallel copies of the block-offdiagonal parts of the matrix.
5020: Level: intermediate
5022: .seealso: MatTranspose(), MatIsSymmetric(), MatIsHermitian()
5023: @*/
5024: PetscErrorCode MatIsTranspose(Mat A,Mat B,PetscReal tol,PetscBool *flg)
5025: {
5026: PetscErrorCode ierr,(*f)(Mat,Mat,PetscReal,PetscBool*),(*g)(Mat,Mat,PetscReal,PetscBool*);
5032: PetscObjectQueryFunction((PetscObject)A,"MatIsTranspose_C",&f);
5033: PetscObjectQueryFunction((PetscObject)B,"MatIsTranspose_C",&g);
5034: *flg = PETSC_FALSE;
5035: if (f && g) {
5036: if (f == g) {
5037: (*f)(A,B,tol,flg);
5038: } else SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_NOTSAMETYPE,"Matrices do not have the same comparator for symmetry test");
5039: } else {
5040: MatType mattype;
5041: if (!f) {
5042: MatGetType(A,&mattype);
5043: } else {
5044: MatGetType(B,&mattype);
5045: }
5046: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix of type %s does not support checking for transpose",mattype);
5047: }
5048: return(0);
5049: }
5051: /*@
5052: MatHermitianTranspose - Computes an in-place or out-of-place transpose of a matrix in complex conjugate.
5054: Collective on Mat
5056: Input Parameter:
5057: + mat - the matrix to transpose and complex conjugate
5058: - reuse - MAT_INITIAL_MATRIX to create a new matrix, MAT_INPLACE_MATRIX to reuse the first argument to store the transpose
5060: Output Parameters:
5061: . B - the Hermitian
5063: Level: intermediate
5065: .seealso: MatTranspose(), MatMultTranspose(), MatMultTransposeAdd(), MatIsTranspose(), MatReuse
5066: @*/
5067: PetscErrorCode MatHermitianTranspose(Mat mat,MatReuse reuse,Mat *B)
5068: {
5072: MatTranspose(mat,reuse,B);
5073: #if defined(PETSC_USE_COMPLEX)
5074: MatConjugate(*B);
5075: #endif
5076: return(0);
5077: }
5079: /*@
5080: MatIsHermitianTranspose - Test whether a matrix is another one's Hermitian transpose,
5082: Collective on Mat
5084: Input Parameter:
5085: + A - the matrix to test
5086: - B - the matrix to test against, this can equal the first parameter
5088: Output Parameters:
5089: . flg - the result
5091: Notes:
5092: Only available for SeqAIJ/MPIAIJ matrices. The sequential algorithm
5093: has a running time of the order of the number of nonzeros; the parallel
5094: test involves parallel copies of the block-offdiagonal parts of the matrix.
5096: Level: intermediate
5098: .seealso: MatTranspose(), MatIsSymmetric(), MatIsHermitian(), MatIsTranspose()
5099: @*/
5100: PetscErrorCode MatIsHermitianTranspose(Mat A,Mat B,PetscReal tol,PetscBool *flg)
5101: {
5102: PetscErrorCode ierr,(*f)(Mat,Mat,PetscReal,PetscBool*),(*g)(Mat,Mat,PetscReal,PetscBool*);
5108: PetscObjectQueryFunction((PetscObject)A,"MatIsHermitianTranspose_C",&f);
5109: PetscObjectQueryFunction((PetscObject)B,"MatIsHermitianTranspose_C",&g);
5110: if (f && g) {
5111: if (f==g) {
5112: (*f)(A,B,tol,flg);
5113: } else SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_NOTSAMETYPE,"Matrices do not have the same comparator for Hermitian test");
5114: }
5115: return(0);
5116: }
5118: /*@
5119: MatPermute - Creates a new matrix with rows and columns permuted from the
5120: original.
5122: Collective on Mat
5124: Input Parameters:
5125: + mat - the matrix to permute
5126: . row - row permutation, each processor supplies only the permutation for its rows
5127: - col - column permutation, each processor supplies only the permutation for its columns
5129: Output Parameters:
5130: . B - the permuted matrix
5132: Level: advanced
5134: Note:
5135: The index sets map from row/col of permuted matrix to row/col of original matrix.
5136: The index sets should be on the same communicator as Mat and have the same local sizes.
5138: .seealso: MatGetOrdering(), ISAllGather()
5140: @*/
5141: PetscErrorCode MatPermute(Mat mat,IS row,IS col,Mat *B)
5142: {
5151: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5152: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5153: if (!mat->ops->permute) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"MatPermute not available for Mat type %s",((PetscObject)mat)->type_name);
5154: MatCheckPreallocated(mat,1);
5156: (*mat->ops->permute)(mat,row,col,B);
5157: PetscObjectStateIncrease((PetscObject)*B);
5158: return(0);
5159: }
5161: /*@
5162: MatEqual - Compares two matrices.
5164: Collective on Mat
5166: Input Parameters:
5167: + A - the first matrix
5168: - B - the second matrix
5170: Output Parameter:
5171: . flg - PETSC_TRUE if the matrices are equal; PETSC_FALSE otherwise.
5173: Level: intermediate
5175: @*/
5176: PetscErrorCode MatEqual(Mat A,Mat B,PetscBool *flg)
5177: {
5187: MatCheckPreallocated(B,2);
5188: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5189: if (!B->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5190: if (A->rmap->N != B->rmap->N || A->cmap->N != B->cmap->N) SETERRQ4(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Mat A,Mat B: global dim %D %D %D %D",A->rmap->N,B->rmap->N,A->cmap->N,B->cmap->N);
5191: if (!A->ops->equal) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Mat type %s",((PetscObject)A)->type_name);
5192: if (!B->ops->equal) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Mat type %s",((PetscObject)B)->type_name);
5193: if (A->ops->equal != B->ops->equal) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_INCOMP,"A is type: %s\nB is type: %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
5194: MatCheckPreallocated(A,1);
5196: (*A->ops->equal)(A,B,flg);
5197: return(0);
5198: }
5200: /*@
5201: MatDiagonalScale - Scales a matrix on the left and right by diagonal
5202: matrices that are stored as vectors. Either of the two scaling
5203: matrices can be NULL.
5205: Collective on Mat
5207: Input Parameters:
5208: + mat - the matrix to be scaled
5209: . l - the left scaling vector (or NULL)
5210: - r - the right scaling vector (or NULL)
5212: Notes:
5213: MatDiagonalScale() computes A = LAR, where
5214: L = a diagonal matrix (stored as a vector), R = a diagonal matrix (stored as a vector)
5215: The L scales the rows of the matrix, the R scales the columns of the matrix.
5217: Level: intermediate
5220: .seealso: MatScale(), MatShift(), MatDiagonalSet()
5221: @*/
5222: PetscErrorCode MatDiagonalScale(Mat mat,Vec l,Vec r)
5223: {
5231: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5232: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5233: MatCheckPreallocated(mat,1);
5234: if (!l && !r) return(0);
5236: if (!mat->ops->diagonalscale) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5237: PetscLogEventBegin(MAT_Scale,mat,0,0,0);
5238: (*mat->ops->diagonalscale)(mat,l,r);
5239: PetscLogEventEnd(MAT_Scale,mat,0,0,0);
5240: PetscObjectStateIncrease((PetscObject)mat);
5241: return(0);
5242: }
5244: /*@
5245: MatScale - Scales all elements of a matrix by a given number.
5247: Logically Collective on Mat
5249: Input Parameters:
5250: + mat - the matrix to be scaled
5251: - a - the scaling value
5253: Output Parameter:
5254: . mat - the scaled matrix
5256: Level: intermediate
5258: .seealso: MatDiagonalScale()
5259: @*/
5260: PetscErrorCode MatScale(Mat mat,PetscScalar a)
5261: {
5267: if (a != (PetscScalar)1.0 && !mat->ops->scale) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5268: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5269: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5271: MatCheckPreallocated(mat,1);
5273: PetscLogEventBegin(MAT_Scale,mat,0,0,0);
5274: if (a != (PetscScalar)1.0) {
5275: (*mat->ops->scale)(mat,a);
5276: PetscObjectStateIncrease((PetscObject)mat);
5277: }
5278: PetscLogEventEnd(MAT_Scale,mat,0,0,0);
5279: return(0);
5280: }
5282: /*@
5283: MatNorm - Calculates various norms of a matrix.
5285: Collective on Mat
5287: Input Parameters:
5288: + mat - the matrix
5289: - type - the type of norm, NORM_1, NORM_FROBENIUS, NORM_INFINITY
5291: Output Parameters:
5292: . nrm - the resulting norm
5294: Level: intermediate
5296: @*/
5297: PetscErrorCode MatNorm(Mat mat,NormType type,PetscReal *nrm)
5298: {
5306: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5307: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5308: if (!mat->ops->norm) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5309: MatCheckPreallocated(mat,1);
5311: (*mat->ops->norm)(mat,type,nrm);
5312: return(0);
5313: }
5315: /*
5316: This variable is used to prevent counting of MatAssemblyBegin() that
5317: are called from within a MatAssemblyEnd().
5318: */
5319: static PetscInt MatAssemblyEnd_InUse = 0;
5320: /*@
5321: MatAssemblyBegin - Begins assembling the matrix. This routine should
5322: be called after completing all calls to MatSetValues().
5324: Collective on Mat
5326: Input Parameters:
5327: + mat - the matrix
5328: - type - type of assembly, either MAT_FLUSH_ASSEMBLY or MAT_FINAL_ASSEMBLY
5330: Notes:
5331: MatSetValues() generally caches the values. The matrix is ready to
5332: use only after MatAssemblyBegin() and MatAssemblyEnd() have been called.
5333: Use MAT_FLUSH_ASSEMBLY when switching between ADD_VALUES and INSERT_VALUES
5334: in MatSetValues(); use MAT_FINAL_ASSEMBLY for the final assembly before
5335: using the matrix.
5337: ALL processes that share a matrix MUST call MatAssemblyBegin() and MatAssemblyEnd() the SAME NUMBER of times, and each time with the
5338: 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
5339: a global collective operation requring all processes that share the matrix.
5341: Space for preallocated nonzeros that is not filled by a call to MatSetValues() or a related routine are compressed
5342: out by assembly. If you intend to use that extra space on a subsequent assembly, be sure to insert explicit zeros
5343: before MAT_FINAL_ASSEMBLY so the space is not compressed out.
5345: Level: beginner
5347: .seealso: MatAssemblyEnd(), MatSetValues(), MatAssembled()
5348: @*/
5349: PetscErrorCode MatAssemblyBegin(Mat mat,MatAssemblyType type)
5350: {
5356: MatCheckPreallocated(mat,1);
5357: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix.\nDid you forget to call MatSetUnfactored()?");
5358: if (mat->assembled) {
5359: mat->was_assembled = PETSC_TRUE;
5360: mat->assembled = PETSC_FALSE;
5361: }
5363: if (!MatAssemblyEnd_InUse) {
5364: PetscLogEventBegin(MAT_AssemblyBegin,mat,0,0,0);
5365: if (mat->ops->assemblybegin) {(*mat->ops->assemblybegin)(mat,type);}
5366: PetscLogEventEnd(MAT_AssemblyBegin,mat,0,0,0);
5367: } else if (mat->ops->assemblybegin) {
5368: (*mat->ops->assemblybegin)(mat,type);
5369: }
5370: return(0);
5371: }
5373: /*@
5374: MatAssembled - Indicates if a matrix has been assembled and is ready for
5375: use; for example, in matrix-vector product.
5377: Not Collective
5379: Input Parameter:
5380: . mat - the matrix
5382: Output Parameter:
5383: . assembled - PETSC_TRUE or PETSC_FALSE
5385: Level: advanced
5387: .seealso: MatAssemblyEnd(), MatSetValues(), MatAssemblyBegin()
5388: @*/
5389: PetscErrorCode MatAssembled(Mat mat,PetscBool *assembled)
5390: {
5394: *assembled = mat->assembled;
5395: return(0);
5396: }
5398: /*@
5399: MatAssemblyEnd - Completes assembling the matrix. This routine should
5400: be called after MatAssemblyBegin().
5402: Collective on Mat
5404: Input Parameters:
5405: + mat - the matrix
5406: - type - type of assembly, either MAT_FLUSH_ASSEMBLY or MAT_FINAL_ASSEMBLY
5408: Options Database Keys:
5409: + -mat_view ::ascii_info - Prints info on matrix at conclusion of MatEndAssembly()
5410: . -mat_view ::ascii_info_detail - Prints more detailed info
5411: . -mat_view - Prints matrix in ASCII format
5412: . -mat_view ::ascii_matlab - Prints matrix in Matlab format
5413: . -mat_view draw - PetscDraws nonzero structure of matrix, using MatView() and PetscDrawOpenX().
5414: . -display <name> - Sets display name (default is host)
5415: . -draw_pause <sec> - Sets number of seconds to pause after display
5416: . -mat_view socket - Sends matrix to socket, can be accessed from Matlab (See Users-Manual: ch_matlab)
5417: . -viewer_socket_machine <machine> - Machine to use for socket
5418: . -viewer_socket_port <port> - Port number to use for socket
5419: - -mat_view binary:filename[:append] - Save matrix to file in binary format
5421: Notes:
5422: MatSetValues() generally caches the values. The matrix is ready to
5423: use only after MatAssemblyBegin() and MatAssemblyEnd() have been called.
5424: Use MAT_FLUSH_ASSEMBLY when switching between ADD_VALUES and INSERT_VALUES
5425: in MatSetValues(); use MAT_FINAL_ASSEMBLY for the final assembly before
5426: using the matrix.
5428: Space for preallocated nonzeros that is not filled by a call to MatSetValues() or a related routine are compressed
5429: out by assembly. If you intend to use that extra space on a subsequent assembly, be sure to insert explicit zeros
5430: before MAT_FINAL_ASSEMBLY so the space is not compressed out.
5432: Level: beginner
5434: .seealso: MatAssemblyBegin(), MatSetValues(), PetscDrawOpenX(), PetscDrawCreate(), MatView(), MatAssembled(), PetscViewerSocketOpen()
5435: @*/
5436: PetscErrorCode MatAssemblyEnd(Mat mat,MatAssemblyType type)
5437: {
5438: PetscErrorCode ierr;
5439: static PetscInt inassm = 0;
5440: PetscBool flg = PETSC_FALSE;
5446: inassm++;
5447: MatAssemblyEnd_InUse++;
5448: if (MatAssemblyEnd_InUse == 1) { /* Do the logging only the first time through */
5449: PetscLogEventBegin(MAT_AssemblyEnd,mat,0,0,0);
5450: if (mat->ops->assemblyend) {
5451: (*mat->ops->assemblyend)(mat,type);
5452: }
5453: PetscLogEventEnd(MAT_AssemblyEnd,mat,0,0,0);
5454: } else if (mat->ops->assemblyend) {
5455: (*mat->ops->assemblyend)(mat,type);
5456: }
5458: /* Flush assembly is not a true assembly */
5459: if (type != MAT_FLUSH_ASSEMBLY) {
5460: mat->num_ass++;
5461: mat->assembled = PETSC_TRUE;
5462: mat->ass_nonzerostate = mat->nonzerostate;
5463: }
5465: mat->insertmode = NOT_SET_VALUES;
5466: MatAssemblyEnd_InUse--;
5467: PetscObjectStateIncrease((PetscObject)mat);
5468: if (!mat->symmetric_eternal) {
5469: mat->symmetric_set = PETSC_FALSE;
5470: mat->hermitian_set = PETSC_FALSE;
5471: mat->structurally_symmetric_set = PETSC_FALSE;
5472: }
5473: if (inassm == 1 && type != MAT_FLUSH_ASSEMBLY) {
5474: MatViewFromOptions(mat,NULL,"-mat_view");
5476: if (mat->checksymmetryonassembly) {
5477: MatIsSymmetric(mat,mat->checksymmetrytol,&flg);
5478: if (flg) {
5479: PetscPrintf(PetscObjectComm((PetscObject)mat),"Matrix is symmetric (tolerance %g)\n",(double)mat->checksymmetrytol);
5480: } else {
5481: PetscPrintf(PetscObjectComm((PetscObject)mat),"Matrix is not symmetric (tolerance %g)\n",(double)mat->checksymmetrytol);
5482: }
5483: }
5484: if (mat->nullsp && mat->checknullspaceonassembly) {
5485: MatNullSpaceTest(mat->nullsp,mat,NULL);
5486: }
5487: }
5488: inassm--;
5489: return(0);
5490: }
5492: /*@
5493: MatSetOption - Sets a parameter option for a matrix. Some options
5494: may be specific to certain storage formats. Some options
5495: determine how values will be inserted (or added). Sorted,
5496: row-oriented input will generally assemble the fastest. The default
5497: is row-oriented.
5499: Logically Collective on Mat for certain operations, such as MAT_SPD, not collective for MAT_ROW_ORIENTED, see MatOption
5501: Input Parameters:
5502: + mat - the matrix
5503: . option - the option, one of those listed below (and possibly others),
5504: - flg - turn the option on (PETSC_TRUE) or off (PETSC_FALSE)
5506: Options Describing Matrix Structure:
5507: + MAT_SPD - symmetric positive definite
5508: . MAT_SYMMETRIC - symmetric in terms of both structure and value
5509: . MAT_HERMITIAN - transpose is the complex conjugation
5510: . MAT_STRUCTURALLY_SYMMETRIC - symmetric nonzero structure
5511: - MAT_SYMMETRY_ETERNAL - if you would like the symmetry/Hermitian flag
5512: you set to be kept with all future use of the matrix
5513: including after MatAssemblyBegin/End() which could
5514: potentially change the symmetry structure, i.e. you
5515: KNOW the matrix will ALWAYS have the property you set.
5516: Note that setting this flag alone implies nothing about whether the matrix is symmetric/Hermitian;
5517: the relevant flags must be set independently.
5520: Options For Use with MatSetValues():
5521: Insert a logically dense subblock, which can be
5522: . MAT_ROW_ORIENTED - row-oriented (default)
5524: Note these options reflect the data you pass in with MatSetValues(); it has
5525: nothing to do with how the data is stored internally in the matrix
5526: data structure.
5528: When (re)assembling a matrix, we can restrict the input for
5529: efficiency/debugging purposes. These options include:
5530: + MAT_NEW_NONZERO_LOCATIONS - additional insertions will be allowed if they generate a new nonzero (slow)
5531: . MAT_NEW_DIAGONALS - new diagonals will be allowed (for block diagonal format only)
5532: . MAT_IGNORE_OFF_PROC_ENTRIES - drops off-processor entries
5533: . MAT_NEW_NONZERO_LOCATION_ERR - generates an error for new matrix entry
5534: . MAT_USE_HASH_TABLE - uses a hash table to speed up matrix assembly
5535: . MAT_NO_OFF_PROC_ENTRIES - you know each process will only set values for its own rows, will generate an error if
5536: any process sets values for another process. This avoids all reductions in the MatAssembly routines and thus improves
5537: performance for very large process counts.
5538: - MAT_SUBSET_OFF_PROC_ENTRIES - you know that the first assembly after setting this flag will set a superset
5539: of the off-process entries required for all subsequent assemblies. This avoids a rendezvous step in the MatAssembly
5540: functions, instead sending only neighbor messages.
5542: Notes:
5543: Except for MAT_UNUSED_NONZERO_LOCATION_ERR and MAT_ROW_ORIENTED all processes that share the matrix must pass the same value in flg!
5545: Some options are relevant only for particular matrix types and
5546: are thus ignored by others. Other options are not supported by
5547: certain matrix types and will generate an error message if set.
5549: If using a Fortran 77 module to compute a matrix, one may need to
5550: use the column-oriented option (or convert to the row-oriented
5551: format).
5553: MAT_NEW_NONZERO_LOCATIONS set to PETSC_FALSE indicates that any add or insertion
5554: that would generate a new entry in the nonzero structure is instead
5555: ignored. Thus, if memory has not alredy been allocated for this particular
5556: data, then the insertion is ignored. For dense matrices, in which
5557: the entire array is allocated, no entries are ever ignored.
5558: Set after the first MatAssemblyEnd(). If this option is set then the MatAssemblyBegin/End() processes has one less global reduction
5560: MAT_NEW_NONZERO_LOCATION_ERR set to PETSC_TRUE indicates that any add or insertion
5561: that would generate a new entry in the nonzero structure instead produces
5562: an error. (Currently supported for AIJ and BAIJ formats only.) If this option is set then the MatAssemblyBegin/End() processes has one less global reduction
5564: MAT_NEW_NONZERO_ALLOCATION_ERR set to PETSC_TRUE indicates that any add or insertion
5565: that would generate a new entry that has not been preallocated will
5566: instead produce an error. (Currently supported for AIJ and BAIJ formats
5567: only.) This is a useful flag when debugging matrix memory preallocation.
5568: If this option is set then the MatAssemblyBegin/End() processes has one less global reduction
5570: MAT_IGNORE_OFF_PROC_ENTRIES set to PETSC_TRUE indicates entries destined for
5571: other processors should be dropped, rather than stashed.
5572: This is useful if you know that the "owning" processor is also
5573: always generating the correct matrix entries, so that PETSc need
5574: not transfer duplicate entries generated on another processor.
5576: MAT_USE_HASH_TABLE indicates that a hash table be used to improve the
5577: searches during matrix assembly. When this flag is set, the hash table
5578: is created during the first Matrix Assembly. This hash table is
5579: used the next time through, during MatSetVaules()/MatSetVaulesBlocked()
5580: to improve the searching of indices. MAT_NEW_NONZERO_LOCATIONS flag
5581: should be used with MAT_USE_HASH_TABLE flag. This option is currently
5582: supported by MATMPIBAIJ format only.
5584: MAT_KEEP_NONZERO_PATTERN indicates when MatZeroRows() is called the zeroed entries
5585: are kept in the nonzero structure
5587: MAT_IGNORE_ZERO_ENTRIES - for AIJ/IS matrices this will stop zero values from creating
5588: a zero location in the matrix
5590: MAT_USE_INODES - indicates using inode version of the code - works with AIJ matrix types
5592: MAT_NO_OFF_PROC_ZERO_ROWS - you know each process will only zero its own rows. This avoids all reductions in the
5593: zero row routines and thus improves performance for very large process counts.
5595: MAT_IGNORE_LOWER_TRIANGULAR - For SBAIJ matrices will ignore any insertions you make in the lower triangular
5596: part of the matrix (since they should match the upper triangular part).
5598: MAT_SORTED_FULL - each process provides exactly its local rows; all column indices for a given row are passed in a
5599: single call to MatSetValues(), preallocation is perfect, row oriented, INSERT_VALUES is used. Common
5600: with finite difference schemes with non-periodic boundary conditions.
5601: Notes:
5602: Can only be called after MatSetSizes() and MatSetType() have been set.
5604: Level: intermediate
5606: .seealso: MatOption, Mat
5608: @*/
5609: PetscErrorCode MatSetOption(Mat mat,MatOption op,PetscBool flg)
5610: {
5616: if (op > 0) {
5619: }
5621: if (((int) op) <= MAT_OPTION_MIN || ((int) op) >= MAT_OPTION_MAX) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"Options %d is out of range",(int)op);
5622: if (!((PetscObject)mat)->type_name) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_TYPENOTSET,"Cannot set options until type and size have been set, see MatSetType() and MatSetSizes()");
5624: switch (op) {
5625: case MAT_NO_OFF_PROC_ENTRIES:
5626: mat->nooffprocentries = flg;
5627: return(0);
5628: case MAT_SUBSET_OFF_PROC_ENTRIES:
5629: mat->assembly_subset = flg;
5630: if (!mat->assembly_subset) { /* See the same logic in VecAssembly wrt VEC_SUBSET_OFF_PROC_ENTRIES */
5631: #if !defined(PETSC_HAVE_MPIUNI)
5632: MatStashScatterDestroy_BTS(&mat->stash);
5633: #endif
5634: mat->stash.first_assembly_done = PETSC_FALSE;
5635: }
5636: return(0);
5637: case MAT_NO_OFF_PROC_ZERO_ROWS:
5638: mat->nooffproczerorows = flg;
5639: return(0);
5640: case MAT_SPD:
5641: mat->spd_set = PETSC_TRUE;
5642: mat->spd = flg;
5643: if (flg) {
5644: mat->symmetric = PETSC_TRUE;
5645: mat->structurally_symmetric = PETSC_TRUE;
5646: mat->symmetric_set = PETSC_TRUE;
5647: mat->structurally_symmetric_set = PETSC_TRUE;
5648: }
5649: break;
5650: case MAT_SYMMETRIC:
5651: mat->symmetric = flg;
5652: if (flg) mat->structurally_symmetric = PETSC_TRUE;
5653: mat->symmetric_set = PETSC_TRUE;
5654: mat->structurally_symmetric_set = flg;
5655: #if !defined(PETSC_USE_COMPLEX)
5656: mat->hermitian = flg;
5657: mat->hermitian_set = PETSC_TRUE;
5658: #endif
5659: break;
5660: case MAT_HERMITIAN:
5661: mat->hermitian = flg;
5662: if (flg) mat->structurally_symmetric = PETSC_TRUE;
5663: mat->hermitian_set = PETSC_TRUE;
5664: mat->structurally_symmetric_set = flg;
5665: #if !defined(PETSC_USE_COMPLEX)
5666: mat->symmetric = flg;
5667: mat->symmetric_set = PETSC_TRUE;
5668: #endif
5669: break;
5670: case MAT_STRUCTURALLY_SYMMETRIC:
5671: mat->structurally_symmetric = flg;
5672: mat->structurally_symmetric_set = PETSC_TRUE;
5673: break;
5674: case MAT_SYMMETRY_ETERNAL:
5675: mat->symmetric_eternal = flg;
5676: break;
5677: case MAT_STRUCTURE_ONLY:
5678: mat->structure_only = flg;
5679: break;
5680: case MAT_SORTED_FULL:
5681: mat->sortedfull = flg;
5682: break;
5683: default:
5684: break;
5685: }
5686: if (mat->ops->setoption) {
5687: (*mat->ops->setoption)(mat,op,flg);
5688: }
5689: return(0);
5690: }
5692: /*@
5693: MatGetOption - Gets a parameter option that has been set for a matrix.
5695: Logically Collective on Mat for certain operations, such as MAT_SPD, not collective for MAT_ROW_ORIENTED, see MatOption
5697: Input Parameters:
5698: + mat - the matrix
5699: - option - the option, this only responds to certain options, check the code for which ones
5701: Output Parameter:
5702: . flg - turn the option on (PETSC_TRUE) or off (PETSC_FALSE)
5704: Notes:
5705: Can only be called after MatSetSizes() and MatSetType() have been set.
5707: Level: intermediate
5709: .seealso: MatOption, MatSetOption()
5711: @*/
5712: PetscErrorCode MatGetOption(Mat mat,MatOption op,PetscBool *flg)
5713: {
5718: if (((int) op) <= MAT_OPTION_MIN || ((int) op) >= MAT_OPTION_MAX) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"Options %d is out of range",(int)op);
5719: if (!((PetscObject)mat)->type_name) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_TYPENOTSET,"Cannot get options until type and size have been set, see MatSetType() and MatSetSizes()");
5721: switch (op) {
5722: case MAT_NO_OFF_PROC_ENTRIES:
5723: *flg = mat->nooffprocentries;
5724: break;
5725: case MAT_NO_OFF_PROC_ZERO_ROWS:
5726: *flg = mat->nooffproczerorows;
5727: break;
5728: case MAT_SYMMETRIC:
5729: *flg = mat->symmetric;
5730: break;
5731: case MAT_HERMITIAN:
5732: *flg = mat->hermitian;
5733: break;
5734: case MAT_STRUCTURALLY_SYMMETRIC:
5735: *flg = mat->structurally_symmetric;
5736: break;
5737: case MAT_SYMMETRY_ETERNAL:
5738: *flg = mat->symmetric_eternal;
5739: break;
5740: case MAT_SPD:
5741: *flg = mat->spd;
5742: break;
5743: default:
5744: break;
5745: }
5746: return(0);
5747: }
5749: /*@
5750: MatZeroEntries - Zeros all entries of a matrix. For sparse matrices
5751: this routine retains the old nonzero structure.
5753: Logically Collective on Mat
5755: Input Parameters:
5756: . mat - the matrix
5758: Level: intermediate
5760: Notes:
5761: 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.
5762: See the Performance chapter of the users manual for information on preallocating matrices.
5764: .seealso: MatZeroRows()
5765: @*/
5766: PetscErrorCode MatZeroEntries(Mat mat)
5767: {
5773: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5774: if (mat->insertmode != NOT_SET_VALUES) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for matrices where you have set values but not yet assembled");
5775: if (!mat->ops->zeroentries) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5776: MatCheckPreallocated(mat,1);
5778: PetscLogEventBegin(MAT_ZeroEntries,mat,0,0,0);
5779: (*mat->ops->zeroentries)(mat);
5780: PetscLogEventEnd(MAT_ZeroEntries,mat,0,0,0);
5781: PetscObjectStateIncrease((PetscObject)mat);
5782: return(0);
5783: }
5785: /*@
5786: MatZeroRowsColumns - Zeros all entries (except possibly the main diagonal)
5787: of a set of rows and columns of a matrix.
5789: Collective on Mat
5791: Input Parameters:
5792: + mat - the matrix
5793: . numRows - the number of rows to remove
5794: . rows - the global row indices
5795: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
5796: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
5797: - b - optional vector of right hand side, that will be adjusted by provided solution
5799: Notes:
5800: This does not change the nonzero structure of the matrix, it merely zeros those entries in the matrix.
5802: The user can set a value in the diagonal entry (or for the AIJ and
5803: row formats can optionally remove the main diagonal entry from the
5804: nonzero structure as well, by passing 0.0 as the final argument).
5806: For the parallel case, all processes that share the matrix (i.e.,
5807: those in the communicator used for matrix creation) MUST call this
5808: routine, regardless of whether any rows being zeroed are owned by
5809: them.
5811: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
5812: list only rows local to itself).
5814: The option MAT_NO_OFF_PROC_ZERO_ROWS does not apply to this routine.
5816: Level: intermediate
5818: .seealso: MatZeroRowsIS(), MatZeroRows(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
5819: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
5820: @*/
5821: PetscErrorCode MatZeroRowsColumns(Mat mat,PetscInt numRows,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
5822: {
5829: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5830: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5831: if (!mat->ops->zerorowscolumns) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5832: MatCheckPreallocated(mat,1);
5834: (*mat->ops->zerorowscolumns)(mat,numRows,rows,diag,x,b);
5835: MatViewFromOptions(mat,NULL,"-mat_view");
5836: PetscObjectStateIncrease((PetscObject)mat);
5837: return(0);
5838: }
5840: /*@
5841: MatZeroRowsColumnsIS - Zeros all entries (except possibly the main diagonal)
5842: of a set of rows and columns of a matrix.
5844: Collective on Mat
5846: Input Parameters:
5847: + mat - the matrix
5848: . is - the rows to zero
5849: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
5850: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
5851: - b - optional vector of right hand side, that will be adjusted by provided solution
5853: Notes:
5854: This does not change the nonzero structure of the matrix, it merely zeros those entries in the matrix.
5856: The user can set a value in the diagonal entry (or for the AIJ and
5857: row formats can optionally remove the main diagonal entry from the
5858: nonzero structure as well, by passing 0.0 as the final argument).
5860: For the parallel case, all processes that share the matrix (i.e.,
5861: those in the communicator used for matrix creation) MUST call this
5862: routine, regardless of whether any rows being zeroed are owned by
5863: them.
5865: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
5866: list only rows local to itself).
5868: The option MAT_NO_OFF_PROC_ZERO_ROWS does not apply to this routine.
5870: Level: intermediate
5872: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
5873: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRows(), MatZeroRowsColumnsStencil()
5874: @*/
5875: PetscErrorCode MatZeroRowsColumnsIS(Mat mat,IS is,PetscScalar diag,Vec x,Vec b)
5876: {
5878: PetscInt numRows;
5879: const PetscInt *rows;
5886: ISGetLocalSize(is,&numRows);
5887: ISGetIndices(is,&rows);
5888: MatZeroRowsColumns(mat,numRows,rows,diag,x,b);
5889: ISRestoreIndices(is,&rows);
5890: return(0);
5891: }
5893: /*@
5894: MatZeroRows - Zeros all entries (except possibly the main diagonal)
5895: of a set of rows of a matrix.
5897: Collective on Mat
5899: Input Parameters:
5900: + mat - the matrix
5901: . numRows - the number of rows to remove
5902: . rows - the global row indices
5903: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
5904: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
5905: - b - optional vector of right hand side, that will be adjusted by provided solution
5907: Notes:
5908: For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
5909: but does not release memory. For the dense and block diagonal
5910: formats this does not alter the nonzero structure.
5912: If the option MatSetOption(mat,MAT_KEEP_NONZERO_PATTERN,PETSC_TRUE) the nonzero structure
5913: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
5914: merely zeroed.
5916: The user can set a value in the diagonal entry (or for the AIJ and
5917: row formats can optionally remove the main diagonal entry from the
5918: nonzero structure as well, by passing 0.0 as the final argument).
5920: For the parallel case, all processes that share the matrix (i.e.,
5921: those in the communicator used for matrix creation) MUST call this
5922: routine, regardless of whether any rows being zeroed are owned by
5923: them.
5925: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
5926: list only rows local to itself).
5928: You can call MatSetOption(mat,MAT_NO_OFF_PROC_ZERO_ROWS,PETSC_TRUE) if each process indicates only rows it
5929: owns that are to be zeroed. This saves a global synchronization in the implementation.
5931: Level: intermediate
5933: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
5934: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
5935: @*/
5936: PetscErrorCode MatZeroRows(Mat mat,PetscInt numRows,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
5937: {
5944: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5945: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5946: if (!mat->ops->zerorows) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5947: MatCheckPreallocated(mat,1);
5949: (*mat->ops->zerorows)(mat,numRows,rows,diag,x,b);
5950: MatViewFromOptions(mat,NULL,"-mat_view");
5951: PetscObjectStateIncrease((PetscObject)mat);
5952: return(0);
5953: }
5955: /*@
5956: MatZeroRowsIS - Zeros all entries (except possibly the main diagonal)
5957: of a set of rows of a matrix.
5959: Collective on Mat
5961: Input Parameters:
5962: + mat - the matrix
5963: . is - index set of rows to remove
5964: . diag - value put in all diagonals of eliminated rows
5965: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
5966: - b - optional vector of right hand side, that will be adjusted by provided solution
5968: Notes:
5969: For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
5970: but does not release memory. For the dense and block diagonal
5971: formats this does not alter the nonzero structure.
5973: If the option MatSetOption(mat,MAT_KEEP_NONZERO_PATTERN,PETSC_TRUE) the nonzero structure
5974: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
5975: merely zeroed.
5977: The user can set a value in the diagonal entry (or for the AIJ and
5978: row formats can optionally remove the main diagonal entry from the
5979: nonzero structure as well, by passing 0.0 as the final argument).
5981: For the parallel case, all processes that share the matrix (i.e.,
5982: those in the communicator used for matrix creation) MUST call this
5983: routine, regardless of whether any rows being zeroed are owned by
5984: them.
5986: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
5987: list only rows local to itself).
5989: You can call MatSetOption(mat,MAT_NO_OFF_PROC_ZERO_ROWS,PETSC_TRUE) if each process indicates only rows it
5990: owns that are to be zeroed. This saves a global synchronization in the implementation.
5992: Level: intermediate
5994: .seealso: MatZeroRows(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
5995: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
5996: @*/
5997: PetscErrorCode MatZeroRowsIS(Mat mat,IS is,PetscScalar diag,Vec x,Vec b)
5998: {
5999: PetscInt numRows;
6000: const PetscInt *rows;
6007: ISGetLocalSize(is,&numRows);
6008: ISGetIndices(is,&rows);
6009: MatZeroRows(mat,numRows,rows,diag,x,b);
6010: ISRestoreIndices(is,&rows);
6011: return(0);
6012: }
6014: /*@
6015: MatZeroRowsStencil - Zeros all entries (except possibly the main diagonal)
6016: of a set of rows of a matrix. These rows must be local to the process.
6018: Collective on Mat
6020: Input Parameters:
6021: + mat - the matrix
6022: . numRows - the number of rows to remove
6023: . rows - the grid coordinates (and component number when dof > 1) for matrix rows
6024: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6025: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6026: - b - optional vector of right hand side, that will be adjusted by provided solution
6028: Notes:
6029: For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
6030: but does not release memory. For the dense and block diagonal
6031: formats this does not alter the nonzero structure.
6033: If the option MatSetOption(mat,MAT_KEEP_NONZERO_PATTERN,PETSC_TRUE) the nonzero structure
6034: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
6035: merely zeroed.
6037: The user can set a value in the diagonal entry (or for the AIJ and
6038: row formats can optionally remove the main diagonal entry from the
6039: nonzero structure as well, by passing 0.0 as the final argument).
6041: For the parallel case, all processes that share the matrix (i.e.,
6042: those in the communicator used for matrix creation) MUST call this
6043: routine, regardless of whether any rows being zeroed are owned by
6044: them.
6046: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
6047: list only rows local to itself).
6049: The grid coordinates are across the entire grid, not just the local portion
6051: In Fortran idxm and idxn should be declared as
6052: $ MatStencil idxm(4,m)
6053: and the values inserted using
6054: $ idxm(MatStencil_i,1) = i
6055: $ idxm(MatStencil_j,1) = j
6056: $ idxm(MatStencil_k,1) = k
6057: $ idxm(MatStencil_c,1) = c
6058: etc
6060: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
6061: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
6062: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
6063: DM_BOUNDARY_PERIODIC boundary type.
6065: 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
6066: a single value per point) you can skip filling those indices.
6068: Level: intermediate
6070: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsl(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
6071: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
6072: @*/
6073: PetscErrorCode MatZeroRowsStencil(Mat mat,PetscInt numRows,const MatStencil rows[],PetscScalar diag,Vec x,Vec b)
6074: {
6075: PetscInt dim = mat->stencil.dim;
6076: PetscInt sdim = dim - (1 - (PetscInt) mat->stencil.noc);
6077: PetscInt *dims = mat->stencil.dims+1;
6078: PetscInt *starts = mat->stencil.starts;
6079: PetscInt *dxm = (PetscInt*) rows;
6080: PetscInt *jdxm, i, j, tmp, numNewRows = 0;
6088: PetscMalloc1(numRows, &jdxm);
6089: for (i = 0; i < numRows; ++i) {
6090: /* Skip unused dimensions (they are ordered k, j, i, c) */
6091: for (j = 0; j < 3-sdim; ++j) dxm++;
6092: /* Local index in X dir */
6093: tmp = *dxm++ - starts[0];
6094: /* Loop over remaining dimensions */
6095: for (j = 0; j < dim-1; ++j) {
6096: /* If nonlocal, set index to be negative */
6097: if ((*dxm++ - starts[j+1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
6098: /* Update local index */
6099: else tmp = tmp*dims[j] + *(dxm-1) - starts[j+1];
6100: }
6101: /* Skip component slot if necessary */
6102: if (mat->stencil.noc) dxm++;
6103: /* Local row number */
6104: if (tmp >= 0) {
6105: jdxm[numNewRows++] = tmp;
6106: }
6107: }
6108: MatZeroRowsLocal(mat,numNewRows,jdxm,diag,x,b);
6109: PetscFree(jdxm);
6110: return(0);
6111: }
6113: /*@
6114: MatZeroRowsColumnsStencil - Zeros all row and column entries (except possibly the main diagonal)
6115: of a set of rows and columns of a matrix.
6117: Collective on Mat
6119: Input Parameters:
6120: + mat - the matrix
6121: . numRows - the number of rows/columns to remove
6122: . rows - the grid coordinates (and component number when dof > 1) for matrix rows
6123: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
6124: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6125: - b - optional vector of right hand side, that will be adjusted by provided solution
6127: Notes:
6128: For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
6129: but does not release memory. For the dense and block diagonal
6130: formats this does not alter the nonzero structure.
6132: If the option MatSetOption(mat,MAT_KEEP_NONZERO_PATTERN,PETSC_TRUE) the nonzero structure
6133: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
6134: merely zeroed.
6136: The user can set a value in the diagonal entry (or for the AIJ and
6137: row formats can optionally remove the main diagonal entry from the
6138: nonzero structure as well, by passing 0.0 as the final argument).
6140: For the parallel case, all processes that share the matrix (i.e.,
6141: those in the communicator used for matrix creation) MUST call this
6142: routine, regardless of whether any rows being zeroed are owned by
6143: them.
6145: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
6146: list only rows local to itself, but the row/column numbers are given in local numbering).
6148: The grid coordinates are across the entire grid, not just the local portion
6150: In Fortran idxm and idxn should be declared as
6151: $ MatStencil idxm(4,m)
6152: and the values inserted using
6153: $ idxm(MatStencil_i,1) = i
6154: $ idxm(MatStencil_j,1) = j
6155: $ idxm(MatStencil_k,1) = k
6156: $ idxm(MatStencil_c,1) = c
6157: etc
6159: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
6160: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
6161: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
6162: DM_BOUNDARY_PERIODIC boundary type.
6164: 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
6165: a single value per point) you can skip filling those indices.
6167: Level: intermediate
6169: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
6170: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRows()
6171: @*/
6172: PetscErrorCode MatZeroRowsColumnsStencil(Mat mat,PetscInt numRows,const MatStencil rows[],PetscScalar diag,Vec x,Vec b)
6173: {
6174: PetscInt dim = mat->stencil.dim;
6175: PetscInt sdim = dim - (1 - (PetscInt) mat->stencil.noc);
6176: PetscInt *dims = mat->stencil.dims+1;
6177: PetscInt *starts = mat->stencil.starts;
6178: PetscInt *dxm = (PetscInt*) rows;
6179: PetscInt *jdxm, i, j, tmp, numNewRows = 0;
6187: PetscMalloc1(numRows, &jdxm);
6188: for (i = 0; i < numRows; ++i) {
6189: /* Skip unused dimensions (they are ordered k, j, i, c) */
6190: for (j = 0; j < 3-sdim; ++j) dxm++;
6191: /* Local index in X dir */
6192: tmp = *dxm++ - starts[0];
6193: /* Loop over remaining dimensions */
6194: for (j = 0; j < dim-1; ++j) {
6195: /* If nonlocal, set index to be negative */
6196: if ((*dxm++ - starts[j+1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
6197: /* Update local index */
6198: else tmp = tmp*dims[j] + *(dxm-1) - starts[j+1];
6199: }
6200: /* Skip component slot if necessary */
6201: if (mat->stencil.noc) dxm++;
6202: /* Local row number */
6203: if (tmp >= 0) {
6204: jdxm[numNewRows++] = tmp;
6205: }
6206: }
6207: MatZeroRowsColumnsLocal(mat,numNewRows,jdxm,diag,x,b);
6208: PetscFree(jdxm);
6209: return(0);
6210: }
6212: /*@C
6213: MatZeroRowsLocal - Zeros all entries (except possibly the main diagonal)
6214: of a set of rows of a matrix; using local numbering of rows.
6216: Collective on Mat
6218: Input Parameters:
6219: + mat - the matrix
6220: . numRows - the number of rows to remove
6221: . rows - the global row indices
6222: . diag - value put in all diagonals of eliminated rows
6223: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6224: - b - optional vector of right hand side, that will be adjusted by provided solution
6226: Notes:
6227: Before calling MatZeroRowsLocal(), the user must first set the
6228: local-to-global mapping by calling MatSetLocalToGlobalMapping().
6230: For the AIJ matrix formats this removes the old nonzero structure,
6231: but does not release memory. For the dense and block diagonal
6232: formats this does not alter the nonzero structure.
6234: If the option MatSetOption(mat,MAT_KEEP_NONZERO_PATTERN,PETSC_TRUE) the nonzero structure
6235: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
6236: merely zeroed.
6238: The user can set a value in the diagonal entry (or for the AIJ and
6239: row formats can optionally remove the main diagonal entry from the
6240: nonzero structure as well, by passing 0.0 as the final argument).
6242: You can call MatSetOption(mat,MAT_NO_OFF_PROC_ZERO_ROWS,PETSC_TRUE) if each process indicates only rows it
6243: owns that are to be zeroed. This saves a global synchronization in the implementation.
6245: Level: intermediate
6247: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRows(), MatSetOption(),
6248: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
6249: @*/
6250: PetscErrorCode MatZeroRowsLocal(Mat mat,PetscInt numRows,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
6251: {
6258: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6259: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6260: MatCheckPreallocated(mat,1);
6262: if (mat->ops->zerorowslocal) {
6263: (*mat->ops->zerorowslocal)(mat,numRows,rows,diag,x,b);
6264: } else {
6265: IS is, newis;
6266: const PetscInt *newRows;
6268: if (!mat->rmap->mapping) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Need to provide local to global mapping to matrix first");
6269: ISCreateGeneral(PETSC_COMM_SELF,numRows,rows,PETSC_COPY_VALUES,&is);
6270: ISLocalToGlobalMappingApplyIS(mat->rmap->mapping,is,&newis);
6271: ISGetIndices(newis,&newRows);
6272: (*mat->ops->zerorows)(mat,numRows,newRows,diag,x,b);
6273: ISRestoreIndices(newis,&newRows);
6274: ISDestroy(&newis);
6275: ISDestroy(&is);
6276: }
6277: PetscObjectStateIncrease((PetscObject)mat);
6278: return(0);
6279: }
6281: /*@
6282: MatZeroRowsLocalIS - Zeros all entries (except possibly the main diagonal)
6283: of a set of rows of a matrix; using local numbering of rows.
6285: Collective on Mat
6287: Input Parameters:
6288: + mat - the matrix
6289: . is - index set of rows to remove
6290: . diag - value put in all diagonals of eliminated rows
6291: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6292: - b - optional vector of right hand side, that will be adjusted by provided solution
6294: Notes:
6295: Before calling MatZeroRowsLocalIS(), the user must first set the
6296: local-to-global mapping by calling MatSetLocalToGlobalMapping().
6298: For the AIJ matrix formats this removes the old nonzero structure,
6299: but does not release memory. For the dense and block diagonal
6300: formats this does not alter the nonzero structure.
6302: If the option MatSetOption(mat,MAT_KEEP_NONZERO_PATTERN,PETSC_TRUE) the nonzero structure
6303: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
6304: merely zeroed.
6306: The user can set a value in the diagonal entry (or for the AIJ and
6307: row formats can optionally remove the main diagonal entry from the
6308: nonzero structure as well, by passing 0.0 as the final argument).
6310: You can call MatSetOption(mat,MAT_NO_OFF_PROC_ZERO_ROWS,PETSC_TRUE) if each process indicates only rows it
6311: owns that are to be zeroed. This saves a global synchronization in the implementation.
6313: Level: intermediate
6315: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRows(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
6316: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
6317: @*/
6318: PetscErrorCode MatZeroRowsLocalIS(Mat mat,IS is,PetscScalar diag,Vec x,Vec b)
6319: {
6321: PetscInt numRows;
6322: const PetscInt *rows;
6328: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6329: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6330: MatCheckPreallocated(mat,1);
6332: ISGetLocalSize(is,&numRows);
6333: ISGetIndices(is,&rows);
6334: MatZeroRowsLocal(mat,numRows,rows,diag,x,b);
6335: ISRestoreIndices(is,&rows);
6336: return(0);
6337: }
6339: /*@
6340: MatZeroRowsColumnsLocal - Zeros all entries (except possibly the main diagonal)
6341: of a set of rows and columns of a matrix; using local numbering of rows.
6343: Collective on Mat
6345: Input Parameters:
6346: + mat - the matrix
6347: . numRows - the number of rows to remove
6348: . rows - the global row indices
6349: . diag - value put in all diagonals of eliminated rows
6350: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6351: - b - optional vector of right hand side, that will be adjusted by provided solution
6353: Notes:
6354: Before calling MatZeroRowsColumnsLocal(), the user must first set the
6355: local-to-global mapping by calling MatSetLocalToGlobalMapping().
6357: The user can set a value in the diagonal entry (or for the AIJ and
6358: row formats can optionally remove the main diagonal entry from the
6359: nonzero structure as well, by passing 0.0 as the final argument).
6361: Level: intermediate
6363: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
6364: MatZeroRows(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
6365: @*/
6366: PetscErrorCode MatZeroRowsColumnsLocal(Mat mat,PetscInt numRows,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
6367: {
6369: IS is, newis;
6370: const PetscInt *newRows;
6376: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6377: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6378: MatCheckPreallocated(mat,1);
6380: if (!mat->cmap->mapping) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Need to provide local to global mapping to matrix first");
6381: ISCreateGeneral(PETSC_COMM_SELF,numRows,rows,PETSC_COPY_VALUES,&is);
6382: ISLocalToGlobalMappingApplyIS(mat->cmap->mapping,is,&newis);
6383: ISGetIndices(newis,&newRows);
6384: (*mat->ops->zerorowscolumns)(mat,numRows,newRows,diag,x,b);
6385: ISRestoreIndices(newis,&newRows);
6386: ISDestroy(&newis);
6387: ISDestroy(&is);
6388: PetscObjectStateIncrease((PetscObject)mat);
6389: return(0);
6390: }
6392: /*@
6393: MatZeroRowsColumnsLocalIS - Zeros all entries (except possibly the main diagonal)
6394: of a set of rows and columns of a matrix; using local numbering of rows.
6396: Collective on Mat
6398: Input Parameters:
6399: + mat - the matrix
6400: . is - index set of rows to remove
6401: . diag - value put in all diagonals of eliminated rows
6402: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6403: - b - optional vector of right hand side, that will be adjusted by provided solution
6405: Notes:
6406: Before calling MatZeroRowsColumnsLocalIS(), the user must first set the
6407: local-to-global mapping by calling MatSetLocalToGlobalMapping().
6409: The user can set a value in the diagonal entry (or for the AIJ and
6410: row formats can optionally remove the main diagonal entry from the
6411: nonzero structure as well, by passing 0.0 as the final argument).
6413: Level: intermediate
6415: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
6416: MatZeroRowsColumnsLocal(), MatZeroRows(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
6417: @*/
6418: PetscErrorCode MatZeroRowsColumnsLocalIS(Mat mat,IS is,PetscScalar diag,Vec x,Vec b)
6419: {
6421: PetscInt numRows;
6422: const PetscInt *rows;
6428: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6429: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6430: MatCheckPreallocated(mat,1);
6432: ISGetLocalSize(is,&numRows);
6433: ISGetIndices(is,&rows);
6434: MatZeroRowsColumnsLocal(mat,numRows,rows,diag,x,b);
6435: ISRestoreIndices(is,&rows);
6436: return(0);
6437: }
6439: /*@C
6440: MatGetSize - Returns the numbers of rows and columns in a matrix.
6442: Not Collective
6444: Input Parameter:
6445: . mat - the matrix
6447: Output Parameters:
6448: + m - the number of global rows
6449: - n - the number of global columns
6451: Note: both output parameters can be NULL on input.
6453: Level: beginner
6455: .seealso: MatGetLocalSize()
6456: @*/
6457: PetscErrorCode MatGetSize(Mat mat,PetscInt *m,PetscInt *n)
6458: {
6461: if (m) *m = mat->rmap->N;
6462: if (n) *n = mat->cmap->N;
6463: return(0);
6464: }
6466: /*@C
6467: MatGetLocalSize - Returns the number of local rows and local columns
6468: of a matrix, that is the local size of the left and right vectors as returned by MatCreateVecs().
6470: Not Collective
6472: Input Parameters:
6473: . mat - the matrix
6475: Output Parameters:
6476: + m - the number of local rows
6477: - n - the number of local columns
6479: Note: both output parameters can be NULL on input.
6481: Level: beginner
6483: .seealso: MatGetSize()
6484: @*/
6485: PetscErrorCode MatGetLocalSize(Mat mat,PetscInt *m,PetscInt *n)
6486: {
6491: if (m) *m = mat->rmap->n;
6492: if (n) *n = mat->cmap->n;
6493: return(0);
6494: }
6496: /*@C
6497: MatGetOwnershipRangeColumn - Returns the range of matrix columns associated with rows of a vector one multiplies by that owned by
6498: this processor. (The columns of the "diagonal block")
6500: Not Collective, unless matrix has not been allocated, then collective on Mat
6502: Input Parameters:
6503: . mat - the matrix
6505: Output Parameters:
6506: + m - the global index of the first local column
6507: - n - one more than the global index of the last local column
6509: Notes:
6510: both output parameters can be NULL on input.
6512: Level: developer
6514: .seealso: MatGetOwnershipRange(), MatGetOwnershipRanges(), MatGetOwnershipRangesColumn()
6516: @*/
6517: PetscErrorCode MatGetOwnershipRangeColumn(Mat mat,PetscInt *m,PetscInt *n)
6518: {
6524: MatCheckPreallocated(mat,1);
6525: if (m) *m = mat->cmap->rstart;
6526: if (n) *n = mat->cmap->rend;
6527: return(0);
6528: }
6530: /*@C
6531: MatGetOwnershipRange - Returns the range of matrix rows owned by
6532: this processor, assuming that the matrix is laid out with the first
6533: n1 rows on the first processor, the next n2 rows on the second, etc.
6534: For certain parallel layouts this range may not be well defined.
6536: Not Collective
6538: Input Parameters:
6539: . mat - the matrix
6541: Output Parameters:
6542: + m - the global index of the first local row
6543: - n - one more than the global index of the last local row
6545: Note: Both output parameters can be NULL on input.
6546: $ This function requires that the matrix be preallocated. If you have not preallocated, consider using
6547: $ PetscSplitOwnership(MPI_Comm comm, PetscInt *n, PetscInt *N)
6548: $ and then MPI_Scan() to calculate prefix sums of the local sizes.
6550: Level: beginner
6552: .seealso: MatGetOwnershipRanges(), MatGetOwnershipRangeColumn(), MatGetOwnershipRangesColumn(), PetscSplitOwnership(), PetscSplitOwnershipBlock()
6554: @*/
6555: PetscErrorCode MatGetOwnershipRange(Mat mat,PetscInt *m,PetscInt *n)
6556: {
6562: MatCheckPreallocated(mat,1);
6563: if (m) *m = mat->rmap->rstart;
6564: if (n) *n = mat->rmap->rend;
6565: return(0);
6566: }
6568: /*@C
6569: MatGetOwnershipRanges - Returns the range of matrix rows owned by
6570: each process
6572: Not Collective, unless matrix has not been allocated, then collective on Mat
6574: Input Parameters:
6575: . mat - the matrix
6577: Output Parameters:
6578: . ranges - start of each processors portion plus one more than the total length at the end
6580: Level: beginner
6582: .seealso: MatGetOwnershipRange(), MatGetOwnershipRangeColumn(), MatGetOwnershipRangesColumn()
6584: @*/
6585: PetscErrorCode MatGetOwnershipRanges(Mat mat,const PetscInt **ranges)
6586: {
6592: MatCheckPreallocated(mat,1);
6593: PetscLayoutGetRanges(mat->rmap,ranges);
6594: return(0);
6595: }
6597: /*@C
6598: MatGetOwnershipRangesColumn - Returns the range of matrix columns associated with rows of a vector one multiplies by that owned by
6599: this processor. (The columns of the "diagonal blocks" for each process)
6601: Not Collective, unless matrix has not been allocated, then collective on Mat
6603: Input Parameters:
6604: . mat - the matrix
6606: Output Parameters:
6607: . ranges - start of each processors portion plus one more then the total length at the end
6609: Level: beginner
6611: .seealso: MatGetOwnershipRange(), MatGetOwnershipRangeColumn(), MatGetOwnershipRanges()
6613: @*/
6614: PetscErrorCode MatGetOwnershipRangesColumn(Mat mat,const PetscInt **ranges)
6615: {
6621: MatCheckPreallocated(mat,1);
6622: PetscLayoutGetRanges(mat->cmap,ranges);
6623: return(0);
6624: }
6626: /*@C
6627: MatGetOwnershipIS - Get row and column ownership as index sets
6629: Not Collective
6631: Input Arguments:
6632: . A - matrix of type Elemental or ScaLAPACK
6634: Output Arguments:
6635: + rows - rows in which this process owns elements
6636: - cols - columns in which this process owns elements
6638: Level: intermediate
6640: .seealso: MatGetOwnershipRange(), MatGetOwnershipRangeColumn(), MatSetValues(), MATELEMENTAL
6641: @*/
6642: PetscErrorCode MatGetOwnershipIS(Mat A,IS *rows,IS *cols)
6643: {
6644: PetscErrorCode ierr,(*f)(Mat,IS*,IS*);
6647: MatCheckPreallocated(A,1);
6648: PetscObjectQueryFunction((PetscObject)A,"MatGetOwnershipIS_C",&f);
6649: if (f) {
6650: (*f)(A,rows,cols);
6651: } else { /* Create a standard row-based partition, each process is responsible for ALL columns in their row block */
6652: if (rows) {ISCreateStride(PETSC_COMM_SELF,A->rmap->n,A->rmap->rstart,1,rows);}
6653: if (cols) {ISCreateStride(PETSC_COMM_SELF,A->cmap->N,0,1,cols);}
6654: }
6655: return(0);
6656: }
6658: /*@C
6659: MatILUFactorSymbolic - Performs symbolic ILU factorization of a matrix.
6660: Uses levels of fill only, not drop tolerance. Use MatLUFactorNumeric()
6661: to complete the factorization.
6663: Collective on Mat
6665: Input Parameters:
6666: + mat - the matrix
6667: . row - row permutation
6668: . column - column permutation
6669: - info - structure containing
6670: $ levels - number of levels of fill.
6671: $ expected fill - as ratio of original fill.
6672: $ 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
6673: missing diagonal entries)
6675: Output Parameters:
6676: . fact - new matrix that has been symbolically factored
6678: Notes:
6679: See Users-Manual: ch_mat for additional information about choosing the fill factor for better efficiency.
6681: Most users should employ the simplified KSP interface for linear solvers
6682: instead of working directly with matrix algebra routines such as this.
6683: See, e.g., KSPCreate().
6685: Level: developer
6687: .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor()
6688: MatGetOrdering(), MatFactorInfo
6690: Note: this uses the definition of level of fill as in Y. Saad, 2003
6692: Developer Note: fortran interface is not autogenerated as the f90
6693: interface defintion cannot be generated correctly [due to MatFactorInfo]
6695: References:
6696: Y. Saad, Iterative methods for sparse linear systems Philadelphia: Society for Industrial and Applied Mathematics, 2003
6697: @*/
6698: PetscErrorCode MatILUFactorSymbolic(Mat fact,Mat mat,IS row,IS col,const MatFactorInfo *info)
6699: {
6709: if (info->levels < 0) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"Levels of fill negative %D",(PetscInt)info->levels);
6710: if (info->fill < 1.0) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"Expected fill less than 1.0 %g",(double)info->fill);
6711: if (!fact->ops->ilufactorsymbolic) {
6712: MatSolverType stype;
6713: MatFactorGetSolverType(fact,&stype);
6714: SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s symbolic ILU using solver type %s",((PetscObject)mat)->type_name,stype);
6715: }
6716: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6717: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6718: MatCheckPreallocated(mat,2);
6720: PetscLogEventBegin(MAT_ILUFactorSymbolic,mat,row,col,0);
6721: (fact->ops->ilufactorsymbolic)(fact,mat,row,col,info);
6722: PetscLogEventEnd(MAT_ILUFactorSymbolic,mat,row,col,0);
6723: return(0);
6724: }
6726: /*@C
6727: MatICCFactorSymbolic - Performs symbolic incomplete
6728: Cholesky factorization for a symmetric matrix. Use
6729: MatCholeskyFactorNumeric() to complete the factorization.
6731: Collective on Mat
6733: Input Parameters:
6734: + mat - the matrix
6735: . perm - row and column permutation
6736: - info - structure containing
6737: $ levels - number of levels of fill.
6738: $ expected fill - as ratio of original fill.
6740: Output Parameter:
6741: . fact - the factored matrix
6743: Notes:
6744: Most users should employ the KSP interface for linear solvers
6745: instead of working directly with matrix algebra routines such as this.
6746: See, e.g., KSPCreate().
6748: Level: developer
6750: .seealso: MatCholeskyFactorNumeric(), MatCholeskyFactor(), MatFactorInfo
6752: Note: this uses the definition of level of fill as in Y. Saad, 2003
6754: Developer Note: fortran interface is not autogenerated as the f90
6755: interface defintion cannot be generated correctly [due to MatFactorInfo]
6757: References:
6758: Y. Saad, Iterative methods for sparse linear systems Philadelphia: Society for Industrial and Applied Mathematics, 2003
6759: @*/
6760: PetscErrorCode MatICCFactorSymbolic(Mat fact,Mat mat,IS perm,const MatFactorInfo *info)
6761: {
6770: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6771: if (info->levels < 0) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"Levels negative %D",(PetscInt) info->levels);
6772: if (info->fill < 1.0) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"Expected fill less than 1.0 %g",(double)info->fill);
6773: if (!(fact)->ops->iccfactorsymbolic) {
6774: MatSolverType stype;
6775: MatFactorGetSolverType(fact,&stype);
6776: SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s symbolic ICC using solver type %s",((PetscObject)mat)->type_name,stype);
6777: }
6778: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6779: MatCheckPreallocated(mat,2);
6781: PetscLogEventBegin(MAT_ICCFactorSymbolic,mat,perm,0,0);
6782: (fact->ops->iccfactorsymbolic)(fact,mat,perm,info);
6783: PetscLogEventEnd(MAT_ICCFactorSymbolic,mat,perm,0,0);
6784: return(0);
6785: }
6787: /*@C
6788: MatCreateSubMatrices - Extracts several submatrices from a matrix. If submat
6789: points to an array of valid matrices, they may be reused to store the new
6790: submatrices.
6792: Collective on Mat
6794: Input Parameters:
6795: + mat - the matrix
6796: . n - the number of submatrixes to be extracted (on this processor, may be zero)
6797: . irow, icol - index sets of rows and columns to extract
6798: - scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
6800: Output Parameter:
6801: . submat - the array of submatrices
6803: Notes:
6804: MatCreateSubMatrices() can extract ONLY sequential submatrices
6805: (from both sequential and parallel matrices). Use MatCreateSubMatrix()
6806: to extract a parallel submatrix.
6808: Some matrix types place restrictions on the row and column
6809: indices, such as that they be sorted or that they be equal to each other.
6811: The index sets may not have duplicate entries.
6813: When extracting submatrices from a parallel matrix, each processor can
6814: form a different submatrix by setting the rows and columns of its
6815: individual index sets according to the local submatrix desired.
6817: When finished using the submatrices, the user should destroy
6818: them with MatDestroySubMatrices().
6820: MAT_REUSE_MATRIX can only be used when the nonzero structure of the
6821: original matrix has not changed from that last call to MatCreateSubMatrices().
6823: This routine creates the matrices in submat; you should NOT create them before
6824: calling it. It also allocates the array of matrix pointers submat.
6826: For BAIJ matrices the index sets must respect the block structure, that is if they
6827: request one row/column in a block, they must request all rows/columns that are in
6828: that block. For example, if the block size is 2 you cannot request just row 0 and
6829: column 0.
6831: Fortran Note:
6832: The Fortran interface is slightly different from that given below; it
6833: requires one to pass in as submat a Mat (integer) array of size at least n+1.
6835: Level: advanced
6838: .seealso: MatDestroySubMatrices(), MatCreateSubMatrix(), MatGetRow(), MatGetDiagonal(), MatReuse
6839: @*/
6840: PetscErrorCode MatCreateSubMatrices(Mat mat,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *submat[])
6841: {
6843: PetscInt i;
6844: PetscBool eq;
6849: if (n) {
6854: }
6856: if (n && scall == MAT_REUSE_MATRIX) {
6859: }
6860: if (!mat->ops->createsubmatrices) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6861: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6862: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6863: MatCheckPreallocated(mat,1);
6865: PetscLogEventBegin(MAT_CreateSubMats,mat,0,0,0);
6866: (*mat->ops->createsubmatrices)(mat,n,irow,icol,scall,submat);
6867: PetscLogEventEnd(MAT_CreateSubMats,mat,0,0,0);
6868: for (i=0; i<n; i++) {
6869: (*submat)[i]->factortype = MAT_FACTOR_NONE; /* in case in place factorization was previously done on submatrix */
6870: ISEqualUnsorted(irow[i],icol[i],&eq);
6871: if (eq) {
6872: MatPropagateSymmetryOptions(mat,(*submat)[i]);
6873: }
6874: }
6875: return(0);
6876: }
6878: /*@C
6879: MatCreateSubMatricesMPI - Extracts MPI submatrices across a sub communicator of mat (by pairs of IS that may live on subcomms).
6881: Collective on Mat
6883: Input Parameters:
6884: + mat - the matrix
6885: . n - the number of submatrixes to be extracted
6886: . irow, icol - index sets of rows and columns to extract
6887: - scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
6889: Output Parameter:
6890: . submat - the array of submatrices
6892: Level: advanced
6895: .seealso: MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRow(), MatGetDiagonal(), MatReuse
6896: @*/
6897: PetscErrorCode MatCreateSubMatricesMPI(Mat mat,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *submat[])
6898: {
6900: PetscInt i;
6901: PetscBool eq;
6906: if (n) {
6911: }
6913: if (n && scall == MAT_REUSE_MATRIX) {
6916: }
6917: if (!mat->ops->createsubmatricesmpi) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6918: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6919: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6920: MatCheckPreallocated(mat,1);
6922: PetscLogEventBegin(MAT_CreateSubMats,mat,0,0,0);
6923: (*mat->ops->createsubmatricesmpi)(mat,n,irow,icol,scall,submat);
6924: PetscLogEventEnd(MAT_CreateSubMats,mat,0,0,0);
6925: for (i=0; i<n; i++) {
6926: ISEqualUnsorted(irow[i],icol[i],&eq);
6927: if (eq) {
6928: MatPropagateSymmetryOptions(mat,(*submat)[i]);
6929: }
6930: }
6931: return(0);
6932: }
6934: /*@C
6935: MatDestroyMatrices - Destroys an array of matrices.
6937: Collective on Mat
6939: Input Parameters:
6940: + n - the number of local matrices
6941: - mat - the matrices (note that this is a pointer to the array of matrices)
6943: Level: advanced
6945: Notes:
6946: Frees not only the matrices, but also the array that contains the matrices
6947: In Fortran will not free the array.
6949: .seealso: MatCreateSubMatrices() MatDestroySubMatrices()
6950: @*/
6951: PetscErrorCode MatDestroyMatrices(PetscInt n,Mat *mat[])
6952: {
6954: PetscInt i;
6957: if (!*mat) return(0);
6958: if (n < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Trying to destroy negative number of matrices %D",n);
6961: for (i=0; i<n; i++) {
6962: MatDestroy(&(*mat)[i]);
6963: }
6965: /* memory is allocated even if n = 0 */
6966: PetscFree(*mat);
6967: return(0);
6968: }
6970: /*@C
6971: MatDestroySubMatrices - Destroys a set of matrices obtained with MatCreateSubMatrices().
6973: Collective on Mat
6975: Input Parameters:
6976: + n - the number of local matrices
6977: - mat - the matrices (note that this is a pointer to the array of matrices, just to match the calling
6978: sequence of MatCreateSubMatrices())
6980: Level: advanced
6982: Notes:
6983: Frees not only the matrices, but also the array that contains the matrices
6984: In Fortran will not free the array.
6986: .seealso: MatCreateSubMatrices()
6987: @*/
6988: PetscErrorCode MatDestroySubMatrices(PetscInt n,Mat *mat[])
6989: {
6991: Mat mat0;
6994: if (!*mat) return(0);
6995: /* mat[] is an array of length n+1, see MatCreateSubMatrices_xxx() */
6996: if (n < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Trying to destroy negative number of matrices %D",n);
6999: mat0 = (*mat)[0];
7000: if (mat0 && mat0->ops->destroysubmatrices) {
7001: (mat0->ops->destroysubmatrices)(n,mat);
7002: } else {
7003: MatDestroyMatrices(n,mat);
7004: }
7005: return(0);
7006: }
7008: /*@C
7009: MatGetSeqNonzeroStructure - Extracts the sequential nonzero structure from a matrix.
7011: Collective on Mat
7013: Input Parameters:
7014: . mat - the matrix
7016: Output Parameter:
7017: . matstruct - the sequential matrix with the nonzero structure of mat
7019: Level: intermediate
7021: .seealso: MatDestroySeqNonzeroStructure(), MatCreateSubMatrices(), MatDestroyMatrices()
7022: @*/
7023: PetscErrorCode MatGetSeqNonzeroStructure(Mat mat,Mat *matstruct)
7024: {
7032: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7033: MatCheckPreallocated(mat,1);
7035: if (!mat->ops->getseqnonzerostructure) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Not for matrix type %s\n",((PetscObject)mat)->type_name);
7036: PetscLogEventBegin(MAT_GetSeqNonzeroStructure,mat,0,0,0);
7037: (*mat->ops->getseqnonzerostructure)(mat,matstruct);
7038: PetscLogEventEnd(MAT_GetSeqNonzeroStructure,mat,0,0,0);
7039: return(0);
7040: }
7042: /*@C
7043: MatDestroySeqNonzeroStructure - Destroys matrix obtained with MatGetSeqNonzeroStructure().
7045: Collective on Mat
7047: Input Parameters:
7048: . mat - the matrix (note that this is a pointer to the array of matrices, just to match the calling
7049: sequence of MatGetSequentialNonzeroStructure())
7051: Level: advanced
7053: Notes:
7054: Frees not only the matrices, but also the array that contains the matrices
7056: .seealso: MatGetSeqNonzeroStructure()
7057: @*/
7058: PetscErrorCode MatDestroySeqNonzeroStructure(Mat *mat)
7059: {
7064: MatDestroy(mat);
7065: return(0);
7066: }
7068: /*@
7069: MatIncreaseOverlap - Given a set of submatrices indicated by index sets,
7070: replaces the index sets by larger ones that represent submatrices with
7071: additional overlap.
7073: Collective on Mat
7075: Input Parameters:
7076: + mat - the matrix
7077: . n - the number of index sets
7078: . is - the array of index sets (these index sets will changed during the call)
7079: - ov - the additional overlap requested
7081: Options Database:
7082: . -mat_increase_overlap_scalable - use a scalable algorithm to compute the overlap (supported by MPIAIJ matrix)
7084: Level: developer
7087: .seealso: MatCreateSubMatrices()
7088: @*/
7089: PetscErrorCode MatIncreaseOverlap(Mat mat,PetscInt n,IS is[],PetscInt ov)
7090: {
7096: if (n < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Must have one or more domains, you have %D",n);
7097: if (n) {
7100: }
7101: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7102: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7103: MatCheckPreallocated(mat,1);
7105: if (!ov) return(0);
7106: if (!mat->ops->increaseoverlap) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
7107: PetscLogEventBegin(MAT_IncreaseOverlap,mat,0,0,0);
7108: (*mat->ops->increaseoverlap)(mat,n,is,ov);
7109: PetscLogEventEnd(MAT_IncreaseOverlap,mat,0,0,0);
7110: return(0);
7111: }
7114: PetscErrorCode MatIncreaseOverlapSplit_Single(Mat,IS*,PetscInt);
7116: /*@
7117: MatIncreaseOverlapSplit - Given a set of submatrices indicated by index sets across
7118: a sub communicator, replaces the index sets by larger ones that represent submatrices with
7119: additional overlap.
7121: Collective on Mat
7123: Input Parameters:
7124: + mat - the matrix
7125: . n - the number of index sets
7126: . is - the array of index sets (these index sets will changed during the call)
7127: - ov - the additional overlap requested
7129: Options Database:
7130: . -mat_increase_overlap_scalable - use a scalable algorithm to compute the overlap (supported by MPIAIJ matrix)
7132: Level: developer
7135: .seealso: MatCreateSubMatrices()
7136: @*/
7137: PetscErrorCode MatIncreaseOverlapSplit(Mat mat,PetscInt n,IS is[],PetscInt ov)
7138: {
7139: PetscInt i;
7145: if (n < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Must have one or more domains, you have %D",n);
7146: if (n) {
7149: }
7150: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7151: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7152: MatCheckPreallocated(mat,1);
7153: if (!ov) return(0);
7154: PetscLogEventBegin(MAT_IncreaseOverlap,mat,0,0,0);
7155: for (i=0; i<n; i++){
7156: MatIncreaseOverlapSplit_Single(mat,&is[i],ov);
7157: }
7158: PetscLogEventEnd(MAT_IncreaseOverlap,mat,0,0,0);
7159: return(0);
7160: }
7165: /*@
7166: MatGetBlockSize - Returns the matrix block size.
7168: Not Collective
7170: Input Parameter:
7171: . mat - the matrix
7173: Output Parameter:
7174: . bs - block size
7176: Notes:
7177: Block row formats are MATSEQBAIJ, MATMPIBAIJ, MATSEQSBAIJ, MATMPISBAIJ. These formats ALWAYS have square block storage in the matrix.
7179: If the block size has not been set yet this routine returns 1.
7181: Level: intermediate
7183: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSizes()
7184: @*/
7185: PetscErrorCode MatGetBlockSize(Mat mat,PetscInt *bs)
7186: {
7190: *bs = PetscAbs(mat->rmap->bs);
7191: return(0);
7192: }
7194: /*@
7195: MatGetBlockSizes - Returns the matrix block row and column sizes.
7197: Not Collective
7199: Input Parameter:
7200: . mat - the matrix
7202: Output Parameter:
7203: + rbs - row block size
7204: - cbs - column block size
7206: Notes:
7207: Block row formats are MATSEQBAIJ, MATMPIBAIJ, MATSEQSBAIJ, MATMPISBAIJ. These formats ALWAYS have square block storage in the matrix.
7208: If you pass a different block size for the columns than the rows, the row block size determines the square block storage.
7210: If a block size has not been set yet this routine returns 1.
7212: Level: intermediate
7214: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSize(), MatSetBlockSize(), MatSetBlockSizes()
7215: @*/
7216: PetscErrorCode MatGetBlockSizes(Mat mat,PetscInt *rbs, PetscInt *cbs)
7217: {
7222: if (rbs) *rbs = PetscAbs(mat->rmap->bs);
7223: if (cbs) *cbs = PetscAbs(mat->cmap->bs);
7224: return(0);
7225: }
7227: /*@
7228: MatSetBlockSize - Sets the matrix block size.
7230: Logically Collective on Mat
7232: Input Parameters:
7233: + mat - the matrix
7234: - bs - block size
7236: Notes:
7237: Block row formats are MATSEQBAIJ, MATMPIBAIJ, MATSEQSBAIJ, MATMPISBAIJ. These formats ALWAYS have square block storage in the matrix.
7238: This must be called before MatSetUp() or MatXXXSetPreallocation() (or will default to 1) and the block size cannot be changed later.
7240: For MATMPIAIJ and MATSEQAIJ matrix formats, this function can be called at a later stage, provided that the specified block size
7241: is compatible with the matrix local sizes.
7243: Level: intermediate
7245: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSize(), MatSetBlockSizes(), MatGetBlockSizes()
7246: @*/
7247: PetscErrorCode MatSetBlockSize(Mat mat,PetscInt bs)
7248: {
7254: MatSetBlockSizes(mat,bs,bs);
7255: return(0);
7256: }
7258: /*@
7259: MatSetVariableBlockSizes - Sets a diagonal blocks of the matrix that need not be of the same size
7261: Logically Collective on Mat
7263: Input Parameters:
7264: + mat - the matrix
7265: . nblocks - the number of blocks on this process
7266: - bsizes - the block sizes
7268: Notes:
7269: Currently used by PCVPBJACOBI for SeqAIJ matrices
7271: Level: intermediate
7273: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSize(), MatSetBlockSizes(), MatGetBlockSizes(), MatGetVariableBlockSizes()
7274: @*/
7275: PetscErrorCode MatSetVariableBlockSizes(Mat mat,PetscInt nblocks,PetscInt *bsizes)
7276: {
7278: PetscInt i,ncnt = 0, nlocal;
7282: if (nblocks < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Number of local blocks must be great than or equal to zero");
7283: MatGetLocalSize(mat,&nlocal,NULL);
7284: for (i=0; i<nblocks; i++) ncnt += bsizes[i];
7285: if (ncnt != nlocal) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Sum of local block sizes %D does not equal local size of matrix %D",ncnt,nlocal);
7286: PetscFree(mat->bsizes);
7287: mat->nblocks = nblocks;
7288: PetscMalloc1(nblocks,&mat->bsizes);
7289: PetscArraycpy(mat->bsizes,bsizes,nblocks);
7290: return(0);
7291: }
7293: /*@C
7294: MatGetVariableBlockSizes - Gets a diagonal blocks of the matrix that need not be of the same size
7296: Logically Collective on Mat
7298: Input Parameters:
7299: . mat - the matrix
7301: Output Parameters:
7302: + nblocks - the number of blocks on this process
7303: - bsizes - the block sizes
7305: Notes: Currently not supported from Fortran
7307: Level: intermediate
7309: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSize(), MatSetBlockSizes(), MatGetBlockSizes(), MatSetVariableBlockSizes()
7310: @*/
7311: PetscErrorCode MatGetVariableBlockSizes(Mat mat,PetscInt *nblocks,const PetscInt **bsizes)
7312: {
7315: *nblocks = mat->nblocks;
7316: *bsizes = mat->bsizes;
7317: return(0);
7318: }
7320: /*@
7321: MatSetBlockSizes - Sets the matrix block row and column sizes.
7323: Logically Collective on Mat
7325: Input Parameters:
7326: + mat - the matrix
7327: . rbs - row block size
7328: - cbs - column block size
7330: Notes:
7331: Block row formats are MATSEQBAIJ, MATMPIBAIJ, MATSEQSBAIJ, MATMPISBAIJ. These formats ALWAYS have square block storage in the matrix.
7332: If you pass a different block size for the columns than the rows, the row block size determines the square block storage.
7333: This must be called before MatSetUp() or MatXXXSetPreallocation() (or will default to 1) and the block size cannot be changed later.
7335: For MATMPIAIJ and MATSEQAIJ matrix formats, this function can be called at a later stage, provided that the specified block sizes
7336: are compatible with the matrix local sizes.
7338: The row and column block size determine the blocksize of the "row" and "column" vectors returned by MatCreateVecs().
7340: Level: intermediate
7342: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSize(), MatSetBlockSize(), MatGetBlockSizes()
7343: @*/
7344: PetscErrorCode MatSetBlockSizes(Mat mat,PetscInt rbs,PetscInt cbs)
7345: {
7352: if (mat->ops->setblocksizes) {
7353: (*mat->ops->setblocksizes)(mat,rbs,cbs);
7354: }
7355: if (mat->rmap->refcnt) {
7356: ISLocalToGlobalMapping l2g = NULL;
7357: PetscLayout nmap = NULL;
7359: PetscLayoutDuplicate(mat->rmap,&nmap);
7360: if (mat->rmap->mapping) {
7361: ISLocalToGlobalMappingDuplicate(mat->rmap->mapping,&l2g);
7362: }
7363: PetscLayoutDestroy(&mat->rmap);
7364: mat->rmap = nmap;
7365: mat->rmap->mapping = l2g;
7366: }
7367: if (mat->cmap->refcnt) {
7368: ISLocalToGlobalMapping l2g = NULL;
7369: PetscLayout nmap = NULL;
7371: PetscLayoutDuplicate(mat->cmap,&nmap);
7372: if (mat->cmap->mapping) {
7373: ISLocalToGlobalMappingDuplicate(mat->cmap->mapping,&l2g);
7374: }
7375: PetscLayoutDestroy(&mat->cmap);
7376: mat->cmap = nmap;
7377: mat->cmap->mapping = l2g;
7378: }
7379: PetscLayoutSetBlockSize(mat->rmap,rbs);
7380: PetscLayoutSetBlockSize(mat->cmap,cbs);
7381: return(0);
7382: }
7384: /*@
7385: MatSetBlockSizesFromMats - Sets the matrix block row and column sizes to match a pair of matrices
7387: Logically Collective on Mat
7389: Input Parameters:
7390: + mat - the matrix
7391: . fromRow - matrix from which to copy row block size
7392: - fromCol - matrix from which to copy column block size (can be same as fromRow)
7394: Level: developer
7396: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSize(), MatSetBlockSizes()
7397: @*/
7398: PetscErrorCode MatSetBlockSizesFromMats(Mat mat,Mat fromRow,Mat fromCol)
7399: {
7406: if (fromRow->rmap->bs > 0) {PetscLayoutSetBlockSize(mat->rmap,fromRow->rmap->bs);}
7407: if (fromCol->cmap->bs > 0) {PetscLayoutSetBlockSize(mat->cmap,fromCol->cmap->bs);}
7408: return(0);
7409: }
7411: /*@
7412: MatResidual - Default routine to calculate the residual.
7414: Collective on Mat
7416: Input Parameters:
7417: + mat - the matrix
7418: . b - the right-hand-side
7419: - x - the approximate solution
7421: Output Parameter:
7422: . r - location to store the residual
7424: Level: developer
7426: .seealso: PCMGSetResidual()
7427: @*/
7428: PetscErrorCode MatResidual(Mat mat,Vec b,Vec x,Vec r)
7429: {
7438: MatCheckPreallocated(mat,1);
7439: PetscLogEventBegin(MAT_Residual,mat,0,0,0);
7440: if (!mat->ops->residual) {
7441: MatMult(mat,x,r);
7442: VecAYPX(r,-1.0,b);
7443: } else {
7444: (*mat->ops->residual)(mat,b,x,r);
7445: }
7446: PetscLogEventEnd(MAT_Residual,mat,0,0,0);
7447: return(0);
7448: }
7450: /*@C
7451: MatGetRowIJ - Returns the compressed row storage i and j indices for sequential matrices.
7453: Collective on Mat
7455: Input Parameters:
7456: + mat - the matrix
7457: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
7458: . symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be symmetrized
7459: - inodecompressed - PETSC_TRUE or PETSC_FALSE indicating if the nonzero structure of the
7460: inodes or the nonzero elements is wanted. For BAIJ matrices the compressed version is
7461: always used.
7463: Output Parameters:
7464: + n - number of rows in the (possibly compressed) matrix
7465: . ia - the row pointers; that is ia[0] = 0, ia[row] = ia[row-1] + number of elements in that row of the matrix
7466: . ja - the column indices
7467: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
7468: are responsible for handling the case when done == PETSC_FALSE and ia and ja are not set
7470: Level: developer
7472: Notes:
7473: You CANNOT change any of the ia[] or ja[] values.
7475: Use MatRestoreRowIJ() when you are finished accessing the ia[] and ja[] values.
7477: Fortran Notes:
7478: In Fortran use
7479: $
7480: $ PetscInt ia(1), ja(1)
7481: $ PetscOffset iia, jja
7482: $ call MatGetRowIJ(mat,shift,symmetric,inodecompressed,n,ia,iia,ja,jja,done,ierr)
7483: $ ! Access the ith and jth entries via ia(iia + i) and ja(jja + j)
7485: or
7486: $
7487: $ PetscInt, pointer :: ia(:),ja(:)
7488: $ call MatGetRowIJF90(mat,shift,symmetric,inodecompressed,n,ia,ja,done,ierr)
7489: $ ! Access the ith and jth entries via ia(i) and ja(j)
7491: .seealso: MatGetColumnIJ(), MatRestoreRowIJ(), MatSeqAIJGetArray()
7492: @*/
7493: PetscErrorCode MatGetRowIJ(Mat mat,PetscInt shift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
7494: {
7504: MatCheckPreallocated(mat,1);
7505: if (!mat->ops->getrowij) *done = PETSC_FALSE;
7506: else {
7507: *done = PETSC_TRUE;
7508: PetscLogEventBegin(MAT_GetRowIJ,mat,0,0,0);
7509: (*mat->ops->getrowij)(mat,shift,symmetric,inodecompressed,n,ia,ja,done);
7510: PetscLogEventEnd(MAT_GetRowIJ,mat,0,0,0);
7511: }
7512: return(0);
7513: }
7515: /*@C
7516: MatGetColumnIJ - Returns the compressed column storage i and j indices for sequential matrices.
7518: Collective on Mat
7520: Input Parameters:
7521: + mat - the matrix
7522: . shift - 1 or zero indicating we want the indices starting at 0 or 1
7523: . symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
7524: symmetrized
7525: . inodecompressed - PETSC_TRUE or PETSC_FALSE indicating if the nonzero structure of the
7526: inodes or the nonzero elements is wanted. For BAIJ matrices the compressed version is
7527: always used.
7528: . n - number of columns in the (possibly compressed) matrix
7529: . ia - the column pointers; that is ia[0] = 0, ia[col] = i[col-1] + number of elements in that col of the matrix
7530: - ja - the row indices
7532: Output Parameters:
7533: . done - PETSC_TRUE or PETSC_FALSE, indicating whether the values have been returned
7535: Level: developer
7537: .seealso: MatGetRowIJ(), MatRestoreColumnIJ()
7538: @*/
7539: PetscErrorCode MatGetColumnIJ(Mat mat,PetscInt shift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
7540: {
7550: MatCheckPreallocated(mat,1);
7551: if (!mat->ops->getcolumnij) *done = PETSC_FALSE;
7552: else {
7553: *done = PETSC_TRUE;
7554: (*mat->ops->getcolumnij)(mat,shift,symmetric,inodecompressed,n,ia,ja,done);
7555: }
7556: return(0);
7557: }
7559: /*@C
7560: MatRestoreRowIJ - Call after you are completed with the ia,ja indices obtained with
7561: MatGetRowIJ().
7563: Collective on Mat
7565: Input Parameters:
7566: + mat - the matrix
7567: . shift - 1 or zero indicating we want the indices starting at 0 or 1
7568: . symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
7569: symmetrized
7570: . inodecompressed - PETSC_TRUE or PETSC_FALSE indicating if the nonzero structure of the
7571: inodes or the nonzero elements is wanted. For BAIJ matrices the compressed version is
7572: always used.
7573: . n - size of (possibly compressed) matrix
7574: . ia - the row pointers
7575: - ja - the column indices
7577: Output Parameters:
7578: . done - PETSC_TRUE or PETSC_FALSE indicated that the values have been returned
7580: Note:
7581: This routine zeros out n, ia, and ja. This is to prevent accidental
7582: us of the array after it has been restored. If you pass NULL, it will
7583: not zero the pointers. Use of ia or ja after MatRestoreRowIJ() is invalid.
7585: Level: developer
7587: .seealso: MatGetRowIJ(), MatRestoreColumnIJ()
7588: @*/
7589: PetscErrorCode MatRestoreRowIJ(Mat mat,PetscInt shift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
7590: {
7599: MatCheckPreallocated(mat,1);
7601: if (!mat->ops->restorerowij) *done = PETSC_FALSE;
7602: else {
7603: *done = PETSC_TRUE;
7604: (*mat->ops->restorerowij)(mat,shift,symmetric,inodecompressed,n,ia,ja,done);
7605: if (n) *n = 0;
7606: if (ia) *ia = NULL;
7607: if (ja) *ja = NULL;
7608: }
7609: return(0);
7610: }
7612: /*@C
7613: MatRestoreColumnIJ - Call after you are completed with the ia,ja indices obtained with
7614: MatGetColumnIJ().
7616: Collective on Mat
7618: Input Parameters:
7619: + mat - the matrix
7620: . shift - 1 or zero indicating we want the indices starting at 0 or 1
7621: - symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
7622: symmetrized
7623: - inodecompressed - PETSC_TRUE or PETSC_FALSE indicating if the nonzero structure of the
7624: inodes or the nonzero elements is wanted. For BAIJ matrices the compressed version is
7625: always used.
7627: Output Parameters:
7628: + n - size of (possibly compressed) matrix
7629: . ia - the column pointers
7630: . ja - the row indices
7631: - done - PETSC_TRUE or PETSC_FALSE indicated that the values have been returned
7633: Level: developer
7635: .seealso: MatGetColumnIJ(), MatRestoreRowIJ()
7636: @*/
7637: PetscErrorCode MatRestoreColumnIJ(Mat mat,PetscInt shift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
7638: {
7647: MatCheckPreallocated(mat,1);
7649: if (!mat->ops->restorecolumnij) *done = PETSC_FALSE;
7650: else {
7651: *done = PETSC_TRUE;
7652: (*mat->ops->restorecolumnij)(mat,shift,symmetric,inodecompressed,n,ia,ja,done);
7653: if (n) *n = 0;
7654: if (ia) *ia = NULL;
7655: if (ja) *ja = NULL;
7656: }
7657: return(0);
7658: }
7660: /*@C
7661: MatColoringPatch -Used inside matrix coloring routines that
7662: use MatGetRowIJ() and/or MatGetColumnIJ().
7664: Collective on Mat
7666: Input Parameters:
7667: + mat - the matrix
7668: . ncolors - max color value
7669: . n - number of entries in colorarray
7670: - colorarray - array indicating color for each column
7672: Output Parameters:
7673: . iscoloring - coloring generated using colorarray information
7675: Level: developer
7677: .seealso: MatGetRowIJ(), MatGetColumnIJ()
7679: @*/
7680: PetscErrorCode MatColoringPatch(Mat mat,PetscInt ncolors,PetscInt n,ISColoringValue colorarray[],ISColoring *iscoloring)
7681: {
7689: MatCheckPreallocated(mat,1);
7691: if (!mat->ops->coloringpatch) {
7692: ISColoringCreate(PetscObjectComm((PetscObject)mat),ncolors,n,colorarray,PETSC_OWN_POINTER,iscoloring);
7693: } else {
7694: (*mat->ops->coloringpatch)(mat,ncolors,n,colorarray,iscoloring);
7695: }
7696: return(0);
7697: }
7700: /*@
7701: MatSetUnfactored - Resets a factored matrix to be treated as unfactored.
7703: Logically Collective on Mat
7705: Input Parameter:
7706: . mat - the factored matrix to be reset
7708: Notes:
7709: This routine should be used only with factored matrices formed by in-place
7710: factorization via ILU(0) (or by in-place LU factorization for the MATSEQDENSE
7711: format). This option can save memory, for example, when solving nonlinear
7712: systems with a matrix-free Newton-Krylov method and a matrix-based, in-place
7713: ILU(0) preconditioner.
7715: Note that one can specify in-place ILU(0) factorization by calling
7716: .vb
7717: PCType(pc,PCILU);
7718: PCFactorSeUseInPlace(pc);
7719: .ve
7720: or by using the options -pc_type ilu -pc_factor_in_place
7722: In-place factorization ILU(0) can also be used as a local
7723: solver for the blocks within the block Jacobi or additive Schwarz
7724: methods (runtime option: -sub_pc_factor_in_place). See Users-Manual: ch_pc
7725: for details on setting local solver options.
7727: Most users should employ the simplified KSP interface for linear solvers
7728: instead of working directly with matrix algebra routines such as this.
7729: See, e.g., KSPCreate().
7731: Level: developer
7733: .seealso: PCFactorSetUseInPlace(), PCFactorGetUseInPlace()
7735: @*/
7736: PetscErrorCode MatSetUnfactored(Mat mat)
7737: {
7743: MatCheckPreallocated(mat,1);
7744: mat->factortype = MAT_FACTOR_NONE;
7745: if (!mat->ops->setunfactored) return(0);
7746: (*mat->ops->setunfactored)(mat);
7747: return(0);
7748: }
7750: /*MC
7751: MatDenseGetArrayF90 - Accesses a matrix array from Fortran90.
7753: Synopsis:
7754: MatDenseGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:,:)},integer ierr)
7756: Not collective
7758: Input Parameter:
7759: . x - matrix
7761: Output Parameters:
7762: + xx_v - the Fortran90 pointer to the array
7763: - ierr - error code
7765: Example of Usage:
7766: .vb
7767: PetscScalar, pointer xx_v(:,:)
7768: ....
7769: call MatDenseGetArrayF90(x,xx_v,ierr)
7770: a = xx_v(3)
7771: call MatDenseRestoreArrayF90(x,xx_v,ierr)
7772: .ve
7774: Level: advanced
7776: .seealso: MatDenseRestoreArrayF90(), MatDenseGetArray(), MatDenseRestoreArray(), MatSeqAIJGetArrayF90()
7778: M*/
7780: /*MC
7781: MatDenseRestoreArrayF90 - Restores a matrix array that has been
7782: accessed with MatDenseGetArrayF90().
7784: Synopsis:
7785: MatDenseRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:,:)},integer ierr)
7787: Not collective
7789: Input Parameters:
7790: + x - matrix
7791: - xx_v - the Fortran90 pointer to the array
7793: Output Parameter:
7794: . ierr - error code
7796: Example of Usage:
7797: .vb
7798: PetscScalar, pointer xx_v(:,:)
7799: ....
7800: call MatDenseGetArrayF90(x,xx_v,ierr)
7801: a = xx_v(3)
7802: call MatDenseRestoreArrayF90(x,xx_v,ierr)
7803: .ve
7805: Level: advanced
7807: .seealso: MatDenseGetArrayF90(), MatDenseGetArray(), MatDenseRestoreArray(), MatSeqAIJRestoreArrayF90()
7809: M*/
7812: /*MC
7813: MatSeqAIJGetArrayF90 - Accesses a matrix array from Fortran90.
7815: Synopsis:
7816: MatSeqAIJGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
7818: Not collective
7820: Input Parameter:
7821: . x - matrix
7823: Output Parameters:
7824: + xx_v - the Fortran90 pointer to the array
7825: - ierr - error code
7827: Example of Usage:
7828: .vb
7829: PetscScalar, pointer xx_v(:)
7830: ....
7831: call MatSeqAIJGetArrayF90(x,xx_v,ierr)
7832: a = xx_v(3)
7833: call MatSeqAIJRestoreArrayF90(x,xx_v,ierr)
7834: .ve
7836: Level: advanced
7838: .seealso: MatSeqAIJRestoreArrayF90(), MatSeqAIJGetArray(), MatSeqAIJRestoreArray(), MatDenseGetArrayF90()
7840: M*/
7842: /*MC
7843: MatSeqAIJRestoreArrayF90 - Restores a matrix array that has been
7844: accessed with MatSeqAIJGetArrayF90().
7846: Synopsis:
7847: MatSeqAIJRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
7849: Not collective
7851: Input Parameters:
7852: + x - matrix
7853: - xx_v - the Fortran90 pointer to the array
7855: Output Parameter:
7856: . ierr - error code
7858: Example of Usage:
7859: .vb
7860: PetscScalar, pointer xx_v(:)
7861: ....
7862: call MatSeqAIJGetArrayF90(x,xx_v,ierr)
7863: a = xx_v(3)
7864: call MatSeqAIJRestoreArrayF90(x,xx_v,ierr)
7865: .ve
7867: Level: advanced
7869: .seealso: MatSeqAIJGetArrayF90(), MatSeqAIJGetArray(), MatSeqAIJRestoreArray(), MatDenseRestoreArrayF90()
7871: M*/
7874: /*@
7875: MatCreateSubMatrix - Gets a single submatrix on the same number of processors
7876: as the original matrix.
7878: Collective on Mat
7880: Input Parameters:
7881: + mat - the original matrix
7882: . isrow - parallel IS containing the rows this processor should obtain
7883: . 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.
7884: - cll - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
7886: Output Parameter:
7887: . newmat - the new submatrix, of the same type as the old
7889: Level: advanced
7891: Notes:
7892: The submatrix will be able to be multiplied with vectors using the same layout as iscol.
7894: Some matrix types place restrictions on the row and column indices, such
7895: as that they be sorted or that they be equal to each other.
7897: The index sets may not have duplicate entries.
7899: The first time this is called you should use a cll of MAT_INITIAL_MATRIX,
7900: the MatCreateSubMatrix() routine will create the newmat for you. Any additional calls
7901: to this routine with a mat of the same nonzero structure and with a call of MAT_REUSE_MATRIX
7902: will reuse the matrix generated the first time. You should call MatDestroy() on newmat when
7903: you are finished using it.
7905: The communicator of the newly obtained matrix is ALWAYS the same as the communicator of
7906: the input matrix.
7908: If iscol is NULL then all columns are obtained (not supported in Fortran).
7910: Example usage:
7911: Consider the following 8x8 matrix with 34 non-zero values, that is
7912: assembled across 3 processors. Let's assume that proc0 owns 3 rows,
7913: proc1 owns 3 rows, proc2 owns 2 rows. This division can be shown
7914: as follows:
7916: .vb
7917: 1 2 0 | 0 3 0 | 0 4
7918: Proc0 0 5 6 | 7 0 0 | 8 0
7919: 9 0 10 | 11 0 0 | 12 0
7920: -------------------------------------
7921: 13 0 14 | 15 16 17 | 0 0
7922: Proc1 0 18 0 | 19 20 21 | 0 0
7923: 0 0 0 | 22 23 0 | 24 0
7924: -------------------------------------
7925: Proc2 25 26 27 | 0 0 28 | 29 0
7926: 30 0 0 | 31 32 33 | 0 34
7927: .ve
7929: Suppose isrow = [0 1 | 4 | 6 7] and iscol = [1 2 | 3 4 5 | 6]. The resulting submatrix is
7931: .vb
7932: 2 0 | 0 3 0 | 0
7933: Proc0 5 6 | 7 0 0 | 8
7934: -------------------------------
7935: Proc1 18 0 | 19 20 21 | 0
7936: -------------------------------
7937: Proc2 26 27 | 0 0 28 | 29
7938: 0 0 | 31 32 33 | 0
7939: .ve
7942: .seealso: MatCreateSubMatrices(), MatCreateSubMatricesMPI(), MatCreateSubMatrixVirtual(), MatSubMatrixVirtualUpdate()
7943: @*/
7944: PetscErrorCode MatCreateSubMatrix(Mat mat,IS isrow,IS iscol,MatReuse cll,Mat *newmat)
7945: {
7947: PetscMPIInt size;
7948: Mat *local;
7949: IS iscoltmp;
7950: PetscBool flg;
7959: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7960: if (cll == MAT_IGNORE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Cannot use MAT_IGNORE_MATRIX");
7962: MatCheckPreallocated(mat,1);
7963: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
7965: if (!iscol || isrow == iscol) {
7966: PetscBool stride;
7967: PetscMPIInt grabentirematrix = 0,grab;
7968: PetscObjectTypeCompare((PetscObject)isrow,ISSTRIDE,&stride);
7969: if (stride) {
7970: PetscInt first,step,n,rstart,rend;
7971: ISStrideGetInfo(isrow,&first,&step);
7972: if (step == 1) {
7973: MatGetOwnershipRange(mat,&rstart,&rend);
7974: if (rstart == first) {
7975: ISGetLocalSize(isrow,&n);
7976: if (n == rend-rstart) {
7977: grabentirematrix = 1;
7978: }
7979: }
7980: }
7981: }
7982: MPIU_Allreduce(&grabentirematrix,&grab,1,MPI_INT,MPI_MIN,PetscObjectComm((PetscObject)mat));
7983: if (grab) {
7984: PetscInfo(mat,"Getting entire matrix as submatrix\n");
7985: if (cll == MAT_INITIAL_MATRIX) {
7986: *newmat = mat;
7987: PetscObjectReference((PetscObject)mat);
7988: }
7989: return(0);
7990: }
7991: }
7993: if (!iscol) {
7994: ISCreateStride(PetscObjectComm((PetscObject)mat),mat->cmap->n,mat->cmap->rstart,1,&iscoltmp);
7995: } else {
7996: iscoltmp = iscol;
7997: }
7999: /* if original matrix is on just one processor then use submatrix generated */
8000: if (mat->ops->createsubmatrices && !mat->ops->createsubmatrix && size == 1 && cll == MAT_REUSE_MATRIX) {
8001: MatCreateSubMatrices(mat,1,&isrow,&iscoltmp,MAT_REUSE_MATRIX,&newmat);
8002: goto setproperties;
8003: } else if (mat->ops->createsubmatrices && !mat->ops->createsubmatrix && size == 1) {
8004: MatCreateSubMatrices(mat,1,&isrow,&iscoltmp,MAT_INITIAL_MATRIX,&local);
8005: *newmat = *local;
8006: PetscFree(local);
8007: goto setproperties;
8008: } else if (!mat->ops->createsubmatrix) {
8009: /* Create a new matrix type that implements the operation using the full matrix */
8010: PetscLogEventBegin(MAT_CreateSubMat,mat,0,0,0);
8011: switch (cll) {
8012: case MAT_INITIAL_MATRIX:
8013: MatCreateSubMatrixVirtual(mat,isrow,iscoltmp,newmat);
8014: break;
8015: case MAT_REUSE_MATRIX:
8016: MatSubMatrixVirtualUpdate(*newmat,mat,isrow,iscoltmp);
8017: break;
8018: default: SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"Invalid MatReuse, must be either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX");
8019: }
8020: PetscLogEventEnd(MAT_CreateSubMat,mat,0,0,0);
8021: goto setproperties;
8022: }
8024: if (!mat->ops->createsubmatrix) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
8025: PetscLogEventBegin(MAT_CreateSubMat,mat,0,0,0);
8026: (*mat->ops->createsubmatrix)(mat,isrow,iscoltmp,cll,newmat);
8027: PetscLogEventEnd(MAT_CreateSubMat,mat,0,0,0);
8029: setproperties:
8030: ISEqualUnsorted(isrow,iscoltmp,&flg);
8031: if (flg) {
8032: MatPropagateSymmetryOptions(mat,*newmat);
8033: }
8034: if (!iscol) {ISDestroy(&iscoltmp);}
8035: if (*newmat && cll == MAT_INITIAL_MATRIX) {PetscObjectStateIncrease((PetscObject)*newmat);}
8036: return(0);
8037: }
8039: /*@
8040: MatPropagateSymmetryOptions - Propagates symmetry options set on a matrix to another matrix
8042: Not Collective
8044: Input Parameters:
8045: + A - the matrix we wish to propagate options from
8046: - B - the matrix we wish to propagate options to
8048: Level: beginner
8050: Notes: Propagates the options associated to MAT_SYMMETRY_ETERNAL, MAT_STRUCTURALLY_SYMMETRIC, MAT_HERMITIAN, MAT_SPD and MAT_SYMMETRIC
8052: .seealso: MatSetOption()
8053: @*/
8054: PetscErrorCode MatPropagateSymmetryOptions(Mat A, Mat B)
8055: {
8061: if (A->symmetric_eternal) { /* symmetric_eternal does not have a corresponding *set flag */
8062: MatSetOption(B,MAT_SYMMETRY_ETERNAL,A->symmetric_eternal);
8063: }
8064: if (A->structurally_symmetric_set) {
8065: MatSetOption(B,MAT_STRUCTURALLY_SYMMETRIC,A->structurally_symmetric);
8066: }
8067: if (A->hermitian_set) {
8068: MatSetOption(B,MAT_HERMITIAN,A->hermitian);
8069: }
8070: if (A->spd_set) {
8071: MatSetOption(B,MAT_SPD,A->spd);
8072: }
8073: if (A->symmetric_set) {
8074: MatSetOption(B,MAT_SYMMETRIC,A->symmetric);
8075: }
8076: return(0);
8077: }
8079: /*@
8080: MatStashSetInitialSize - sets the sizes of the matrix stash, that is
8081: used during the assembly process to store values that belong to
8082: other processors.
8084: Not Collective
8086: Input Parameters:
8087: + mat - the matrix
8088: . size - the initial size of the stash.
8089: - bsize - the initial size of the block-stash(if used).
8091: Options Database Keys:
8092: + -matstash_initial_size <size> or <size0,size1,...sizep-1>
8093: - -matstash_block_initial_size <bsize> or <bsize0,bsize1,...bsizep-1>
8095: Level: intermediate
8097: Notes:
8098: The block-stash is used for values set with MatSetValuesBlocked() while
8099: the stash is used for values set with MatSetValues()
8101: Run with the option -info and look for output of the form
8102: MatAssemblyBegin_MPIXXX:Stash has MM entries, uses nn mallocs.
8103: to determine the appropriate value, MM, to use for size and
8104: MatAssemblyBegin_MPIXXX:Block-Stash has BMM entries, uses nn mallocs.
8105: to determine the value, BMM to use for bsize
8108: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), Mat, MatStashGetInfo()
8110: @*/
8111: PetscErrorCode MatStashSetInitialSize(Mat mat,PetscInt size, PetscInt bsize)
8112: {
8118: MatStashSetInitialSize_Private(&mat->stash,size);
8119: MatStashSetInitialSize_Private(&mat->bstash,bsize);
8120: return(0);
8121: }
8123: /*@
8124: MatInterpolateAdd - w = y + A*x or A'*x depending on the shape of
8125: the matrix
8127: Neighbor-wise Collective on Mat
8129: Input Parameters:
8130: + mat - the matrix
8131: . x,y - the vectors
8132: - w - where the result is stored
8134: Level: intermediate
8136: Notes:
8137: w may be the same vector as y.
8139: This allows one to use either the restriction or interpolation (its transpose)
8140: matrix to do the interpolation
8142: .seealso: MatMultAdd(), MatMultTransposeAdd(), MatRestrict()
8144: @*/
8145: PetscErrorCode MatInterpolateAdd(Mat A,Vec x,Vec y,Vec w)
8146: {
8148: PetscInt M,N,Ny;
8156: MatCheckPreallocated(A,1);
8157: MatGetSize(A,&M,&N);
8158: VecGetSize(y,&Ny);
8159: if (M == Ny) {
8160: MatMultAdd(A,x,y,w);
8161: } else {
8162: MatMultTransposeAdd(A,x,y,w);
8163: }
8164: return(0);
8165: }
8167: /*@
8168: MatInterpolate - y = A*x or A'*x depending on the shape of
8169: the matrix
8171: Neighbor-wise Collective on Mat
8173: Input Parameters:
8174: + mat - the matrix
8175: - x,y - the vectors
8177: Level: intermediate
8179: Notes:
8180: This allows one to use either the restriction or interpolation (its transpose)
8181: matrix to do the interpolation
8183: .seealso: MatMultAdd(), MatMultTransposeAdd(), MatRestrict()
8185: @*/
8186: PetscErrorCode MatInterpolate(Mat A,Vec x,Vec y)
8187: {
8189: PetscInt M,N,Ny;
8196: MatCheckPreallocated(A,1);
8197: MatGetSize(A,&M,&N);
8198: VecGetSize(y,&Ny);
8199: if (M == Ny) {
8200: MatMult(A,x,y);
8201: } else {
8202: MatMultTranspose(A,x,y);
8203: }
8204: return(0);
8205: }
8207: /*@
8208: MatRestrict - y = A*x or A'*x
8210: Neighbor-wise Collective on Mat
8212: Input Parameters:
8213: + mat - the matrix
8214: - x,y - the vectors
8216: Level: intermediate
8218: Notes:
8219: This allows one to use either the restriction or interpolation (its transpose)
8220: matrix to do the restriction
8222: .seealso: MatMultAdd(), MatMultTransposeAdd(), MatInterpolate()
8224: @*/
8225: PetscErrorCode MatRestrict(Mat A,Vec x,Vec y)
8226: {
8228: PetscInt M,N,Ny;
8235: MatCheckPreallocated(A,1);
8237: MatGetSize(A,&M,&N);
8238: VecGetSize(y,&Ny);
8239: if (M == Ny) {
8240: MatMult(A,x,y);
8241: } else {
8242: MatMultTranspose(A,x,y);
8243: }
8244: return(0);
8245: }
8247: /*@
8248: MatGetNullSpace - retrieves the null space of a matrix.
8250: Logically Collective on Mat
8252: Input Parameters:
8253: + mat - the matrix
8254: - nullsp - the null space object
8256: Level: developer
8258: .seealso: MatCreate(), MatNullSpaceCreate(), MatSetNearNullSpace(), MatSetNullSpace()
8259: @*/
8260: PetscErrorCode MatGetNullSpace(Mat mat, MatNullSpace *nullsp)
8261: {
8265: *nullsp = (mat->symmetric_set && mat->symmetric && !mat->nullsp) ? mat->transnullsp : mat->nullsp;
8266: return(0);
8267: }
8269: /*@
8270: MatSetNullSpace - attaches a null space to a matrix.
8272: Logically Collective on Mat
8274: Input Parameters:
8275: + mat - the matrix
8276: - nullsp - the null space object
8278: Level: advanced
8280: Notes:
8281: This null space is used by the linear solvers. Overwrites any previous null space that may have been attached
8283: For inconsistent singular systems (linear systems where the right hand side is not in the range of the operator) you also likely should
8284: call MatSetTransposeNullSpace(). This allows the linear system to be solved in a least squares sense.
8286: You can remove the null space by calling this routine with an nullsp of NULL
8289: The fundamental theorem of linear algebra (Gilbert Strang, Introduction to Applied Mathematics, page 72) states that
8290: 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).
8291: 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
8292: 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
8293: 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).
8295: Krylov solvers can produce the minimal norm solution to the least squares problem by utilizing MatNullSpaceRemove().
8297: If the matrix is known to be symmetric because it is an SBAIJ matrix or one as called MatSetOption(mat,MAT_SYMMETRIC or MAT_SYMMETRIC_ETERNAL,PETSC_TRUE); this
8298: routine also automatically calls MatSetTransposeNullSpace().
8300: .seealso: MatCreate(), MatNullSpaceCreate(), MatSetNearNullSpace(), MatGetNullSpace(), MatSetTransposeNullSpace(), MatGetTransposeNullSpace(), MatNullSpaceRemove()
8301: @*/
8302: PetscErrorCode MatSetNullSpace(Mat mat,MatNullSpace nullsp)
8303: {
8309: if (nullsp) {PetscObjectReference((PetscObject)nullsp);}
8310: MatNullSpaceDestroy(&mat->nullsp);
8311: mat->nullsp = nullsp;
8312: if (mat->symmetric_set && mat->symmetric) {
8313: MatSetTransposeNullSpace(mat,nullsp);
8314: }
8315: return(0);
8316: }
8318: /*@
8319: MatGetTransposeNullSpace - retrieves the null space of the transpose of a matrix.
8321: Logically Collective on Mat
8323: Input Parameters:
8324: + mat - the matrix
8325: - nullsp - the null space object
8327: Level: developer
8329: .seealso: MatCreate(), MatNullSpaceCreate(), MatSetNearNullSpace(), MatSetTransposeNullSpace(), MatSetNullSpace(), MatGetNullSpace()
8330: @*/
8331: PetscErrorCode MatGetTransposeNullSpace(Mat mat, MatNullSpace *nullsp)
8332: {
8337: *nullsp = (mat->symmetric_set && mat->symmetric && !mat->transnullsp) ? mat->nullsp : mat->transnullsp;
8338: return(0);
8339: }
8341: /*@
8342: MatSetTransposeNullSpace - attaches a null space to a matrix.
8344: Logically Collective on Mat
8346: Input Parameters:
8347: + mat - the matrix
8348: - nullsp - the null space object
8350: Level: advanced
8352: Notes:
8353: For inconsistent singular systems (linear systems where the right hand side is not in the range of the operator) this allows the linear system to be solved in a least squares sense.
8354: You must also call MatSetNullSpace()
8357: The fundamental theorem of linear algebra (Gilbert Strang, Introduction to Applied Mathematics, page 72) states that
8358: 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).
8359: 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
8360: 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
8361: 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).
8363: Krylov solvers can produce the minimal norm solution to the least squares problem by utilizing MatNullSpaceRemove().
8365: .seealso: MatCreate(), MatNullSpaceCreate(), MatSetNearNullSpace(), MatGetNullSpace(), MatSetNullSpace(), MatGetTransposeNullSpace(), MatNullSpaceRemove()
8366: @*/
8367: PetscErrorCode MatSetTransposeNullSpace(Mat mat,MatNullSpace nullsp)
8368: {
8374: if (nullsp) {PetscObjectReference((PetscObject)nullsp);}
8375: MatNullSpaceDestroy(&mat->transnullsp);
8376: mat->transnullsp = nullsp;
8377: return(0);
8378: }
8380: /*@
8381: MatSetNearNullSpace - attaches a null space to a matrix, which is often the null space (rigid body modes) of the operator without boundary conditions
8382: This null space will be used to provide near null space vectors to a multigrid preconditioner built from this matrix.
8384: Logically Collective on Mat
8386: Input Parameters:
8387: + mat - the matrix
8388: - nullsp - the null space object
8390: Level: advanced
8392: Notes:
8393: Overwrites any previous near null space that may have been attached
8395: You can remove the null space by calling this routine with an nullsp of NULL
8397: .seealso: MatCreate(), MatNullSpaceCreate(), MatSetNullSpace(), MatNullSpaceCreateRigidBody(), MatGetNearNullSpace()
8398: @*/
8399: PetscErrorCode MatSetNearNullSpace(Mat mat,MatNullSpace nullsp)
8400: {
8407: MatCheckPreallocated(mat,1);
8408: if (nullsp) {PetscObjectReference((PetscObject)nullsp);}
8409: MatNullSpaceDestroy(&mat->nearnullsp);
8410: mat->nearnullsp = nullsp;
8411: return(0);
8412: }
8414: /*@
8415: MatGetNearNullSpace - Get null space attached with MatSetNearNullSpace()
8417: Not Collective
8419: Input Parameter:
8420: . mat - the matrix
8422: Output Parameter:
8423: . nullsp - the null space object, NULL if not set
8425: Level: developer
8427: .seealso: MatSetNearNullSpace(), MatGetNullSpace(), MatNullSpaceCreate()
8428: @*/
8429: PetscErrorCode MatGetNearNullSpace(Mat mat,MatNullSpace *nullsp)
8430: {
8435: MatCheckPreallocated(mat,1);
8436: *nullsp = mat->nearnullsp;
8437: return(0);
8438: }
8440: /*@C
8441: MatICCFactor - Performs in-place incomplete Cholesky factorization of matrix.
8443: Collective on Mat
8445: Input Parameters:
8446: + mat - the matrix
8447: . row - row/column permutation
8448: . fill - expected fill factor >= 1.0
8449: - level - level of fill, for ICC(k)
8451: Notes:
8452: Probably really in-place only when level of fill is zero, otherwise allocates
8453: new space to store factored matrix and deletes previous memory.
8455: Most users should employ the simplified KSP interface for linear solvers
8456: instead of working directly with matrix algebra routines such as this.
8457: See, e.g., KSPCreate().
8459: Level: developer
8462: .seealso: MatICCFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor()
8464: Developer Note: fortran interface is not autogenerated as the f90
8465: interface defintion cannot be generated correctly [due to MatFactorInfo]
8467: @*/
8468: PetscErrorCode MatICCFactor(Mat mat,IS row,const MatFactorInfo *info)
8469: {
8477: if (mat->rmap->N != mat->cmap->N) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONG,"matrix must be square");
8478: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
8479: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
8480: if (!mat->ops->iccfactor) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
8481: MatCheckPreallocated(mat,1);
8482: (*mat->ops->iccfactor)(mat,row,info);
8483: PetscObjectStateIncrease((PetscObject)mat);
8484: return(0);
8485: }
8487: /*@
8488: MatDiagonalScaleLocal - Scales columns of a matrix given the scaling values including the
8489: ghosted ones.
8491: Not Collective
8493: Input Parameters:
8494: + mat - the matrix
8495: - diag = the diagonal values, including ghost ones
8497: Level: developer
8499: Notes:
8500: Works only for MPIAIJ and MPIBAIJ matrices
8502: .seealso: MatDiagonalScale()
8503: @*/
8504: PetscErrorCode MatDiagonalScaleLocal(Mat mat,Vec diag)
8505: {
8507: PetscMPIInt size;
8514: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Matrix must be already assembled");
8515: PetscLogEventBegin(MAT_Scale,mat,0,0,0);
8516: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
8517: if (size == 1) {
8518: PetscInt n,m;
8519: VecGetSize(diag,&n);
8520: MatGetSize(mat,NULL,&m);
8521: if (m == n) {
8522: MatDiagonalScale(mat,NULL,diag);
8523: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only supported for sequential matrices when no ghost points/periodic conditions");
8524: } else {
8525: PetscUseMethod(mat,"MatDiagonalScaleLocal_C",(Mat,Vec),(mat,diag));
8526: }
8527: PetscLogEventEnd(MAT_Scale,mat,0,0,0);
8528: PetscObjectStateIncrease((PetscObject)mat);
8529: return(0);
8530: }
8532: /*@
8533: MatGetInertia - Gets the inertia from a factored matrix
8535: Collective on Mat
8537: Input Parameter:
8538: . mat - the matrix
8540: Output Parameters:
8541: + nneg - number of negative eigenvalues
8542: . nzero - number of zero eigenvalues
8543: - npos - number of positive eigenvalues
8545: Level: advanced
8547: Notes:
8548: Matrix must have been factored by MatCholeskyFactor()
8551: @*/
8552: PetscErrorCode MatGetInertia(Mat mat,PetscInt *nneg,PetscInt *nzero,PetscInt *npos)
8553: {
8559: if (!mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
8560: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Numeric factor mat is not assembled");
8561: if (!mat->ops->getinertia) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
8562: (*mat->ops->getinertia)(mat,nneg,nzero,npos);
8563: return(0);
8564: }
8566: /* ----------------------------------------------------------------*/
8567: /*@C
8568: MatSolves - Solves A x = b, given a factored matrix, for a collection of vectors
8570: Neighbor-wise Collective on Mats
8572: Input Parameters:
8573: + mat - the factored matrix
8574: - b - the right-hand-side vectors
8576: Output Parameter:
8577: . x - the result vectors
8579: Notes:
8580: The vectors b and x cannot be the same. I.e., one cannot
8581: call MatSolves(A,x,x).
8583: Notes:
8584: Most users should employ the simplified KSP interface for linear solvers
8585: instead of working directly with matrix algebra routines such as this.
8586: See, e.g., KSPCreate().
8588: Level: developer
8590: .seealso: MatSolveAdd(), MatSolveTranspose(), MatSolveTransposeAdd(), MatSolve()
8591: @*/
8592: PetscErrorCode MatSolves(Mat mat,Vecs b,Vecs x)
8593: {
8599: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
8600: if (!mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
8601: if (!mat->rmap->N && !mat->cmap->N) return(0);
8603: if (!mat->ops->solves) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
8604: MatCheckPreallocated(mat,1);
8605: PetscLogEventBegin(MAT_Solves,mat,0,0,0);
8606: (*mat->ops->solves)(mat,b,x);
8607: PetscLogEventEnd(MAT_Solves,mat,0,0,0);
8608: return(0);
8609: }
8611: /*@
8612: MatIsSymmetric - Test whether a matrix is symmetric
8614: Collective on Mat
8616: Input Parameter:
8617: + A - the matrix to test
8618: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact transpose)
8620: Output Parameters:
8621: . flg - the result
8623: Notes:
8624: For real numbers MatIsSymmetric() and MatIsHermitian() return identical results
8626: Level: intermediate
8628: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetricKnown()
8629: @*/
8630: PetscErrorCode MatIsSymmetric(Mat A,PetscReal tol,PetscBool *flg)
8631: {
8638: if (!A->symmetric_set) {
8639: if (!A->ops->issymmetric) {
8640: MatType mattype;
8641: MatGetType(A,&mattype);
8642: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix of type %s does not support checking for symmetric",mattype);
8643: }
8644: (*A->ops->issymmetric)(A,tol,flg);
8645: if (!tol) {
8646: MatSetOption(A,MAT_SYMMETRIC,*flg);
8647: }
8648: } else if (A->symmetric) {
8649: *flg = PETSC_TRUE;
8650: } else if (!tol) {
8651: *flg = PETSC_FALSE;
8652: } else {
8653: if (!A->ops->issymmetric) {
8654: MatType mattype;
8655: MatGetType(A,&mattype);
8656: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix of type %s does not support checking for symmetric",mattype);
8657: }
8658: (*A->ops->issymmetric)(A,tol,flg);
8659: }
8660: return(0);
8661: }
8663: /*@
8664: MatIsHermitian - Test whether a matrix is Hermitian
8666: Collective on Mat
8668: Input Parameter:
8669: + A - the matrix to test
8670: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact Hermitian)
8672: Output Parameters:
8673: . flg - the result
8675: Level: intermediate
8677: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(),
8678: MatIsSymmetricKnown(), MatIsSymmetric()
8679: @*/
8680: PetscErrorCode MatIsHermitian(Mat A,PetscReal tol,PetscBool *flg)
8681: {
8688: if (!A->hermitian_set) {
8689: if (!A->ops->ishermitian) {
8690: MatType mattype;
8691: MatGetType(A,&mattype);
8692: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix of type %s does not support checking for hermitian",mattype);
8693: }
8694: (*A->ops->ishermitian)(A,tol,flg);
8695: if (!tol) {
8696: MatSetOption(A,MAT_HERMITIAN,*flg);
8697: }
8698: } else if (A->hermitian) {
8699: *flg = PETSC_TRUE;
8700: } else if (!tol) {
8701: *flg = PETSC_FALSE;
8702: } else {
8703: if (!A->ops->ishermitian) {
8704: MatType mattype;
8705: MatGetType(A,&mattype);
8706: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix of type %s does not support checking for hermitian",mattype);
8707: }
8708: (*A->ops->ishermitian)(A,tol,flg);
8709: }
8710: return(0);
8711: }
8713: /*@
8714: MatIsSymmetricKnown - Checks the flag on the matrix to see if it is symmetric.
8716: Not Collective
8718: Input Parameter:
8719: . A - the matrix to check
8721: Output Parameters:
8722: + set - if the symmetric flag is set (this tells you if the next flag is valid)
8723: - flg - the result
8725: Level: advanced
8727: Note: Does not check the matrix values directly, so this may return unknown (set = PETSC_FALSE). Use MatIsSymmetric()
8728: if you want it explicitly checked
8730: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetric()
8731: @*/
8732: PetscErrorCode MatIsSymmetricKnown(Mat A,PetscBool *set,PetscBool *flg)
8733: {
8738: if (A->symmetric_set) {
8739: *set = PETSC_TRUE;
8740: *flg = A->symmetric;
8741: } else {
8742: *set = PETSC_FALSE;
8743: }
8744: return(0);
8745: }
8747: /*@
8748: MatIsHermitianKnown - Checks the flag on the matrix to see if it is hermitian.
8750: Not Collective
8752: Input Parameter:
8753: . A - the matrix to check
8755: Output Parameters:
8756: + set - if the hermitian flag is set (this tells you if the next flag is valid)
8757: - flg - the result
8759: Level: advanced
8761: Note: Does not check the matrix values directly, so this may return unknown (set = PETSC_FALSE). Use MatIsHermitian()
8762: if you want it explicitly checked
8764: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetric()
8765: @*/
8766: PetscErrorCode MatIsHermitianKnown(Mat A,PetscBool *set,PetscBool *flg)
8767: {
8772: if (A->hermitian_set) {
8773: *set = PETSC_TRUE;
8774: *flg = A->hermitian;
8775: } else {
8776: *set = PETSC_FALSE;
8777: }
8778: return(0);
8779: }
8781: /*@
8782: MatIsStructurallySymmetric - Test whether a matrix is structurally symmetric
8784: Collective on Mat
8786: Input Parameter:
8787: . A - the matrix to test
8789: Output Parameters:
8790: . flg - the result
8792: Level: intermediate
8794: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsSymmetric(), MatSetOption()
8795: @*/
8796: PetscErrorCode MatIsStructurallySymmetric(Mat A,PetscBool *flg)
8797: {
8803: if (!A->structurally_symmetric_set) {
8804: if (!A->ops->isstructurallysymmetric) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Matrix of type %s does not support checking for structural symmetric",((PetscObject)A)->type_name);
8805: (*A->ops->isstructurallysymmetric)(A,flg);
8806: MatSetOption(A,MAT_STRUCTURALLY_SYMMETRIC,*flg);
8807: } else *flg = A->structurally_symmetric;
8808: return(0);
8809: }
8811: /*@
8812: MatStashGetInfo - Gets how many values are currently in the matrix stash, i.e. need
8813: to be communicated to other processors during the MatAssemblyBegin/End() process
8815: Not collective
8817: Input Parameter:
8818: . vec - the vector
8820: Output Parameters:
8821: + nstash - the size of the stash
8822: . reallocs - the number of additional mallocs incurred.
8823: . bnstash - the size of the block stash
8824: - breallocs - the number of additional mallocs incurred.in the block stash
8826: Level: advanced
8828: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), Mat, MatStashSetInitialSize()
8830: @*/
8831: PetscErrorCode MatStashGetInfo(Mat mat,PetscInt *nstash,PetscInt *reallocs,PetscInt *bnstash,PetscInt *breallocs)
8832: {
8836: MatStashGetInfo_Private(&mat->stash,nstash,reallocs);
8837: MatStashGetInfo_Private(&mat->bstash,bnstash,breallocs);
8838: return(0);
8839: }
8841: /*@C
8842: MatCreateVecs - Get vector(s) compatible with the matrix, i.e. with the same
8843: parallel layout
8845: Collective on Mat
8847: Input Parameter:
8848: . mat - the matrix
8850: Output Parameter:
8851: + right - (optional) vector that the matrix can be multiplied against
8852: - left - (optional) vector that the matrix vector product can be stored in
8854: Notes:
8855: 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().
8857: Notes:
8858: These are new vectors which are not owned by the Mat, they should be destroyed in VecDestroy() when no longer needed
8860: Level: advanced
8862: .seealso: MatCreate(), VecDestroy()
8863: @*/
8864: PetscErrorCode MatCreateVecs(Mat mat,Vec *right,Vec *left)
8865: {
8871: if (mat->ops->getvecs) {
8872: (*mat->ops->getvecs)(mat,right,left);
8873: } else {
8874: PetscInt rbs,cbs;
8875: MatGetBlockSizes(mat,&rbs,&cbs);
8876: if (right) {
8877: if (mat->cmap->n < 0) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"PetscLayout for columns not yet setup");
8878: VecCreate(PetscObjectComm((PetscObject)mat),right);
8879: VecSetSizes(*right,mat->cmap->n,PETSC_DETERMINE);
8880: VecSetBlockSize(*right,cbs);
8881: VecSetType(*right,mat->defaultvectype);
8882: PetscLayoutReference(mat->cmap,&(*right)->map);
8883: }
8884: if (left) {
8885: if (mat->rmap->n < 0) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"PetscLayout for rows not yet setup");
8886: VecCreate(PetscObjectComm((PetscObject)mat),left);
8887: VecSetSizes(*left,mat->rmap->n,PETSC_DETERMINE);
8888: VecSetBlockSize(*left,rbs);
8889: VecSetType(*left,mat->defaultvectype);
8890: PetscLayoutReference(mat->rmap,&(*left)->map);
8891: }
8892: }
8893: return(0);
8894: }
8896: /*@C
8897: MatFactorInfoInitialize - Initializes a MatFactorInfo data structure
8898: with default values.
8900: Not Collective
8902: Input Parameters:
8903: . info - the MatFactorInfo data structure
8906: Notes:
8907: The solvers are generally used through the KSP and PC objects, for example
8908: PCLU, PCILU, PCCHOLESKY, PCICC
8910: Level: developer
8912: .seealso: MatFactorInfo
8914: Developer Note: fortran interface is not autogenerated as the f90
8915: interface defintion cannot be generated correctly [due to MatFactorInfo]
8917: @*/
8919: PetscErrorCode MatFactorInfoInitialize(MatFactorInfo *info)
8920: {
8924: PetscMemzero(info,sizeof(MatFactorInfo));
8925: return(0);
8926: }
8928: /*@
8929: MatFactorSetSchurIS - Set indices corresponding to the Schur complement you wish to have computed
8931: Collective on Mat
8933: Input Parameters:
8934: + mat - the factored matrix
8935: - is - the index set defining the Schur indices (0-based)
8937: Notes:
8938: Call MatFactorSolveSchurComplement() or MatFactorSolveSchurComplementTranspose() after this call to solve a Schur complement system.
8940: You can call MatFactorGetSchurComplement() or MatFactorCreateSchurComplement() after this call.
8942: Level: developer
8944: .seealso: MatGetFactor(), MatFactorGetSchurComplement(), MatFactorRestoreSchurComplement(), MatFactorCreateSchurComplement(), MatFactorSolveSchurComplement(),
8945: MatFactorSolveSchurComplementTranspose(), MatFactorSolveSchurComplement()
8947: @*/
8948: PetscErrorCode MatFactorSetSchurIS(Mat mat,IS is)
8949: {
8950: PetscErrorCode ierr,(*f)(Mat,IS);
8958: if (!mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Only for factored matrix");
8959: PetscObjectQueryFunction((PetscObject)mat,"MatFactorSetSchurIS_C",&f);
8960: if (!f) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"The selected MatSolverType does not support Schur complement computation. You should use MATSOLVERMUMPS or MATSOLVERMKL_PARDISO");
8961: MatDestroy(&mat->schur);
8962: (*f)(mat,is);
8963: if (!mat->schur) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_PLIB,"Schur complement has not been created");
8964: return(0);
8965: }
8967: /*@
8968: MatFactorCreateSchurComplement - Create a Schur complement matrix object using Schur data computed during the factorization step
8970: Logically Collective on Mat
8972: Input Parameters:
8973: + F - the factored matrix obtained by calling MatGetFactor() from PETSc-MUMPS interface
8974: . S - location where to return the Schur complement, can be NULL
8975: - status - the status of the Schur complement matrix, can be NULL
8977: Notes:
8978: You must call MatFactorSetSchurIS() before calling this routine.
8980: The routine provides a copy of the Schur matrix stored within the solver data structures.
8981: The caller must destroy the object when it is no longer needed.
8982: If MatFactorInvertSchurComplement() has been called, the routine gets back the inverse.
8984: 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)
8986: Developer Notes:
8987: The reason this routine exists is because the representation of the Schur complement within the factor matrix may be different than a standard PETSc
8988: matrix representation and we normally do not want to use the time or memory to make a copy as a regular PETSc matrix.
8990: See MatCreateSchurComplement() or MatGetSchurComplement() for ways to create virtual or approximate Schur complements.
8992: Level: advanced
8994: References:
8996: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorGetSchurComplement(), MatFactorSchurStatus
8997: @*/
8998: PetscErrorCode MatFactorCreateSchurComplement(Mat F,Mat* S,MatFactorSchurStatus* status)
8999: {
9006: if (S) {
9007: PetscErrorCode (*f)(Mat,Mat*);
9009: PetscObjectQueryFunction((PetscObject)F,"MatFactorCreateSchurComplement_C",&f);
9010: if (f) {
9011: (*f)(F,S);
9012: } else {
9013: MatDuplicate(F->schur,MAT_COPY_VALUES,S);
9014: }
9015: }
9016: if (status) *status = F->schur_status;
9017: return(0);
9018: }
9020: /*@
9021: MatFactorGetSchurComplement - Gets access to a Schur complement matrix using the current Schur data within a factored matrix
9023: Logically Collective on Mat
9025: Input Parameters:
9026: + F - the factored matrix obtained by calling MatGetFactor()
9027: . *S - location where to return the Schur complement, can be NULL
9028: - status - the status of the Schur complement matrix, can be NULL
9030: Notes:
9031: You must call MatFactorSetSchurIS() before calling this routine.
9033: Schur complement mode is currently implemented for sequential matrices.
9034: The routine returns a the Schur Complement stored within the data strutures of the solver.
9035: If MatFactorInvertSchurComplement() has previously been called, the returned matrix is actually the inverse of the Schur complement.
9036: The returned matrix should not be destroyed; the caller should call MatFactorRestoreSchurComplement() when the object is no longer needed.
9038: Use MatFactorCreateSchurComplement() to create a copy of the Schur complement matrix that is within a factored matrix
9040: See MatCreateSchurComplement() or MatGetSchurComplement() for ways to create virtual or approximate Schur complements.
9042: Level: advanced
9044: References:
9046: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorRestoreSchurComplement(), MatFactorCreateSchurComplement(), MatFactorSchurStatus
9047: @*/
9048: PetscErrorCode MatFactorGetSchurComplement(Mat F,Mat* S,MatFactorSchurStatus* status)
9049: {
9054: if (S) *S = F->schur;
9055: if (status) *status = F->schur_status;
9056: return(0);
9057: }
9059: /*@
9060: MatFactorRestoreSchurComplement - Restore the Schur complement matrix object obtained from a call to MatFactorGetSchurComplement
9062: Logically Collective on Mat
9064: Input Parameters:
9065: + F - the factored matrix obtained by calling MatGetFactor()
9066: . *S - location where the Schur complement is stored
9067: - status - the status of the Schur complement matrix (see MatFactorSchurStatus)
9069: Notes:
9071: Level: advanced
9073: References:
9075: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorRestoreSchurComplement(), MatFactorCreateSchurComplement(), MatFactorSchurStatus
9076: @*/
9077: PetscErrorCode MatFactorRestoreSchurComplement(Mat F,Mat* S,MatFactorSchurStatus status)
9078: {
9083: if (S) {
9085: *S = NULL;
9086: }
9087: F->schur_status = status;
9088: MatFactorUpdateSchurStatus_Private(F);
9089: return(0);
9090: }
9092: /*@
9093: MatFactorSolveSchurComplementTranspose - Solve the transpose of the Schur complement system computed during the factorization step
9095: Logically Collective on Mat
9097: Input Parameters:
9098: + F - the factored matrix obtained by calling MatGetFactor()
9099: . rhs - location where the right hand side of the Schur complement system is stored
9100: - sol - location where the solution of the Schur complement system has to be returned
9102: Notes:
9103: The sizes of the vectors should match the size of the Schur complement
9105: Must be called after MatFactorSetSchurIS()
9107: Level: advanced
9109: References:
9111: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorSolveSchurComplement()
9112: @*/
9113: PetscErrorCode MatFactorSolveSchurComplementTranspose(Mat F, Vec rhs, Vec sol)
9114: {
9126: MatFactorFactorizeSchurComplement(F);
9127: switch (F->schur_status) {
9128: case MAT_FACTOR_SCHUR_FACTORED:
9129: MatSolveTranspose(F->schur,rhs,sol);
9130: break;
9131: case MAT_FACTOR_SCHUR_INVERTED:
9132: MatMultTranspose(F->schur,rhs,sol);
9133: break;
9134: default:
9135: SETERRQ1(PetscObjectComm((PetscObject)F),PETSC_ERR_SUP,"Unhandled MatFactorSchurStatus %D",F->schur_status);
9136: }
9137: return(0);
9138: }
9140: /*@
9141: MatFactorSolveSchurComplement - Solve the Schur complement system computed during the factorization step
9143: Logically Collective on Mat
9145: Input Parameters:
9146: + F - the factored matrix obtained by calling MatGetFactor()
9147: . rhs - location where the right hand side of the Schur complement system is stored
9148: - sol - location where the solution of the Schur complement system has to be returned
9150: Notes:
9151: The sizes of the vectors should match the size of the Schur complement
9153: Must be called after MatFactorSetSchurIS()
9155: Level: advanced
9157: References:
9159: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorSolveSchurComplementTranspose()
9160: @*/
9161: PetscErrorCode MatFactorSolveSchurComplement(Mat F, Vec rhs, Vec sol)
9162: {
9174: MatFactorFactorizeSchurComplement(F);
9175: switch (F->schur_status) {
9176: case MAT_FACTOR_SCHUR_FACTORED:
9177: MatSolve(F->schur,rhs,sol);
9178: break;
9179: case MAT_FACTOR_SCHUR_INVERTED:
9180: MatMult(F->schur,rhs,sol);
9181: break;
9182: default:
9183: SETERRQ1(PetscObjectComm((PetscObject)F),PETSC_ERR_SUP,"Unhandled MatFactorSchurStatus %D",F->schur_status);
9184: }
9185: return(0);
9186: }
9188: /*@
9189: MatFactorInvertSchurComplement - Invert the Schur complement matrix computed during the factorization step
9191: Logically Collective on Mat
9193: Input Parameters:
9194: . F - the factored matrix obtained by calling MatGetFactor()
9196: Notes:
9197: Must be called after MatFactorSetSchurIS().
9199: Call MatFactorGetSchurComplement() or MatFactorCreateSchurComplement() AFTER this call to actually compute the inverse and get access to it.
9201: Level: advanced
9203: References:
9205: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorGetSchurComplement(), MatFactorCreateSchurComplement()
9206: @*/
9207: PetscErrorCode MatFactorInvertSchurComplement(Mat F)
9208: {
9214: if (F->schur_status == MAT_FACTOR_SCHUR_INVERTED) return(0);
9215: MatFactorFactorizeSchurComplement(F);
9216: MatFactorInvertSchurComplement_Private(F);
9217: F->schur_status = MAT_FACTOR_SCHUR_INVERTED;
9218: return(0);
9219: }
9221: /*@
9222: MatFactorFactorizeSchurComplement - Factorize the Schur complement matrix computed during the factorization step
9224: Logically Collective on Mat
9226: Input Parameters:
9227: . F - the factored matrix obtained by calling MatGetFactor()
9229: Notes:
9230: Must be called after MatFactorSetSchurIS().
9232: Level: advanced
9234: References:
9236: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorInvertSchurComplement()
9237: @*/
9238: PetscErrorCode MatFactorFactorizeSchurComplement(Mat F)
9239: {
9245: if (F->schur_status == MAT_FACTOR_SCHUR_INVERTED || F->schur_status == MAT_FACTOR_SCHUR_FACTORED) return(0);
9246: MatFactorFactorizeSchurComplement_Private(F);
9247: F->schur_status = MAT_FACTOR_SCHUR_FACTORED;
9248: return(0);
9249: }
9251: /*@
9252: MatPtAP - Creates the matrix product C = P^T * A * P
9254: Neighbor-wise Collective on Mat
9256: Input Parameters:
9257: + A - the matrix
9258: . P - the projection matrix
9259: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9260: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(P)), use PETSC_DEFAULT if you do not have a good estimate
9261: if the result is a dense matrix this is irrelevent
9263: Output Parameters:
9264: . C - the product matrix
9266: Notes:
9267: C will be created and must be destroyed by the user with MatDestroy().
9269: For matrix types without special implementation the function fallbacks to MatMatMult() followed by MatTransposeMatMult().
9271: Level: intermediate
9273: .seealso: MatMatMult(), MatRARt()
9274: @*/
9275: PetscErrorCode MatPtAP(Mat A,Mat P,MatReuse scall,PetscReal fill,Mat *C)
9276: {
9280: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*C,5);
9281: if (scall == MAT_INPLACE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Inplace product not supported");
9283: if (scall == MAT_INITIAL_MATRIX) {
9284: MatProductCreate(A,P,NULL,C);
9285: MatProductSetType(*C,MATPRODUCT_PtAP);
9286: MatProductSetAlgorithm(*C,"default");
9287: MatProductSetFill(*C,fill);
9289: (*C)->product->api_user = PETSC_TRUE;
9290: MatProductSetFromOptions(*C);
9291: if (!(*C)->ops->productsymbolic) SETERRQ3(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);
9292: MatProductSymbolic(*C);
9293: } else { /* scall == MAT_REUSE_MATRIX */
9294: MatProductReplaceMats(A,P,NULL,*C);
9295: }
9297: MatProductNumeric(*C);
9298: if (A->symmetric_set && A->symmetric) {
9299: MatSetOption(*C,MAT_SYMMETRIC,PETSC_TRUE);
9300: }
9301: return(0);
9302: }
9304: /*@
9305: MatRARt - Creates the matrix product C = R * A * R^T
9307: Neighbor-wise Collective on Mat
9309: Input Parameters:
9310: + A - the matrix
9311: . R - the projection matrix
9312: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9313: - fill - expected fill as ratio of nnz(C)/nnz(A), use PETSC_DEFAULT if you do not have a good estimate
9314: if the result is a dense matrix this is irrelevent
9316: Output Parameters:
9317: . C - the product matrix
9319: Notes:
9320: C will be created and must be destroyed by the user with MatDestroy().
9322: This routine is currently only implemented for pairs of AIJ matrices and classes
9323: which inherit from AIJ. Due to PETSc sparse matrix block row distribution among processes,
9324: parallel MatRARt is implemented via explicit transpose of R, which could be very expensive.
9325: We recommend using MatPtAP().
9327: Level: intermediate
9329: .seealso: MatMatMult(), MatPtAP()
9330: @*/
9331: PetscErrorCode MatRARt(Mat A,Mat R,MatReuse scall,PetscReal fill,Mat *C)
9332: {
9336: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*C,5);
9337: if (scall == MAT_INPLACE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Inplace product not supported");
9339: if (scall == MAT_INITIAL_MATRIX) {
9340: MatProductCreate(A,R,NULL,C);
9341: MatProductSetType(*C,MATPRODUCT_RARt);
9342: MatProductSetAlgorithm(*C,"default");
9343: MatProductSetFill(*C,fill);
9345: (*C)->product->api_user = PETSC_TRUE;
9346: MatProductSetFromOptions(*C);
9347: if (!(*C)->ops->productsymbolic) SETERRQ3(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);
9348: MatProductSymbolic(*C);
9349: } else { /* scall == MAT_REUSE_MATRIX */
9350: MatProductReplaceMats(A,R,NULL,*C);
9351: }
9353: MatProductNumeric(*C);
9354: if (A->symmetric_set && A->symmetric) {
9355: MatSetOption(*C,MAT_SYMMETRIC,PETSC_TRUE);
9356: }
9357: return(0);
9358: }
9361: static PetscErrorCode MatProduct_Private(Mat A,Mat B,MatReuse scall,PetscReal fill,MatProductType ptype, Mat *C)
9362: {
9366: if (scall == MAT_INPLACE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Inplace product not supported");
9368: if (scall == MAT_INITIAL_MATRIX) {
9369: PetscInfo1(A,"Calling MatProduct API with MAT_INITIAL_MATRIX and product type %s\n",MatProductTypes[ptype]);
9370: MatProductCreate(A,B,NULL,C);
9371: MatProductSetType(*C,ptype);
9372: MatProductSetAlgorithm(*C,MATPRODUCTALGORITHM_DEFAULT);
9373: MatProductSetFill(*C,fill);
9375: (*C)->product->api_user = PETSC_TRUE;
9376: MatProductSetFromOptions(*C);
9377: MatProductSymbolic(*C);
9378: } else { /* scall == MAT_REUSE_MATRIX */
9379: Mat_Product *product = (*C)->product;
9381: PetscInfo2(A,"Calling MatProduct API with MAT_REUSE_MATRIX %s product present and product type %s\n",product ? "with" : "without",MatProductTypes[ptype]);
9382: if (!product) {
9383: /* user provide the dense matrix *C without calling MatProductCreate() */
9384: PetscBool isdense;
9386: PetscObjectBaseTypeCompareAny((PetscObject)(*C),&isdense,MATSEQDENSE,MATMPIDENSE,"");
9387: if (isdense) {
9388: /* user wants to reuse an assembled dense matrix */
9389: /* Create product -- see MatCreateProduct() */
9390: MatProductCreate_Private(A,B,NULL,*C);
9391: product = (*C)->product;
9392: product->fill = fill;
9393: product->api_user = PETSC_TRUE;
9394: product->clear = PETSC_TRUE;
9396: MatProductSetType(*C,ptype);
9397: MatProductSetFromOptions(*C);
9398: if (!(*C)->ops->productsymbolic) SETERRQ3(PetscObjectComm((PetscObject)(*C)),PETSC_ERR_SUP,"MatProduct %s not supported for %s and %s",MatProductTypes[ptype],((PetscObject)A)->type_name,((PetscObject)B)->type_name);
9399: MatProductSymbolic(*C);
9400: } else SETERRQ(PetscObjectComm((PetscObject)(*C)),PETSC_ERR_SUP,"Call MatProductCreate() first");
9401: } else { /* user may change input matrices A or B when REUSE */
9402: MatProductReplaceMats(A,B,NULL,*C);
9403: }
9404: }
9405: MatProductNumeric(*C);
9406: return(0);
9407: }
9409: /*@
9410: MatMatMult - Performs Matrix-Matrix Multiplication C=A*B.
9412: Neighbor-wise Collective on Mat
9414: Input Parameters:
9415: + A - the left matrix
9416: . B - the right matrix
9417: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9418: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use PETSC_DEFAULT if you do not have a good estimate
9419: if the result is a dense matrix this is irrelevent
9421: Output Parameters:
9422: . C - the product matrix
9424: Notes:
9425: Unless scall is MAT_REUSE_MATRIX C will be created.
9427: 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
9428: call to this function with MAT_INITIAL_MATRIX.
9430: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value actually needed.
9432: If you have many matrices with the same non-zero structure to multiply, you should use MatProductCreate()/MatProductSymbolic(C)/ReplaceMats(), and call MatProductNumeric() repeatedly.
9434: 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, rather than first having MatMatMult() create it for you. You can NEVER do this if the matrix C is sparse.
9436: Level: intermediate
9438: .seealso: MatTransposeMatMult(), MatMatTransposeMult(), MatPtAP()
9439: @*/
9440: PetscErrorCode MatMatMult(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
9441: {
9445: MatProduct_Private(A,B,scall,fill,MATPRODUCT_AB,C);
9446: return(0);
9447: }
9449: /*@
9450: MatMatTransposeMult - Performs Matrix-Matrix Multiplication C=A*B^T.
9452: Neighbor-wise Collective on Mat
9454: Input Parameters:
9455: + A - the left matrix
9456: . B - the right matrix
9457: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9458: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use PETSC_DEFAULT if not known
9460: Output Parameters:
9461: . C - the product matrix
9463: Notes:
9464: C will be created if MAT_INITIAL_MATRIX and must be destroyed by the user with MatDestroy().
9466: MAT_REUSE_MATRIX can only be used if the matrices A and B have the same nonzero pattern as in the previous call
9468: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
9469: actually needed.
9471: This routine is currently only implemented for pairs of SeqAIJ matrices, for the SeqDense class,
9472: and for pairs of MPIDense matrices.
9474: Options Database Keys:
9475: . -matmattransmult_mpidense_mpidense_via {allgatherv,cyclic} - Choose between algorthims for MPIDense matrices: the
9476: first redundantly copies the transposed B matrix on each process and requiers O(log P) communication complexity;
9477: the second never stores more than one portion of the B matrix at a time by requires O(P) communication complexity.
9479: Level: intermediate
9481: .seealso: MatMatMult(), MatTransposeMatMult() MatPtAP()
9482: @*/
9483: PetscErrorCode MatMatTransposeMult(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
9484: {
9488: MatProduct_Private(A,B,scall,fill,MATPRODUCT_ABt,C);
9489: return(0);
9490: }
9492: /*@
9493: MatTransposeMatMult - Performs Matrix-Matrix Multiplication C=A^T*B.
9495: Neighbor-wise Collective on Mat
9497: Input Parameters:
9498: + A - the left matrix
9499: . B - the right matrix
9500: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9501: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use PETSC_DEFAULT if not known
9503: Output Parameters:
9504: . C - the product matrix
9506: Notes:
9507: C will be created if MAT_INITIAL_MATRIX and must be destroyed by the user with MatDestroy().
9509: MAT_REUSE_MATRIX can only be used if the matrices A and B have the same nonzero pattern as in the previous call.
9511: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
9512: actually needed.
9514: This routine is currently implemented for pairs of AIJ matrices and pairs of SeqDense matrices and classes
9515: which inherit from SeqAIJ. C will be of same type as the input matrices.
9517: Level: intermediate
9519: .seealso: MatMatMult(), MatMatTransposeMult(), MatPtAP()
9520: @*/
9521: PetscErrorCode MatTransposeMatMult(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
9522: {
9526: MatProduct_Private(A,B,scall,fill,MATPRODUCT_AtB,C);
9527: return(0);
9528: }
9530: /*@
9531: MatMatMatMult - Performs Matrix-Matrix-Matrix Multiplication D=A*B*C.
9533: Neighbor-wise Collective on Mat
9535: Input Parameters:
9536: + A - the left matrix
9537: . B - the middle matrix
9538: . C - the right matrix
9539: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9540: - 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
9541: if the result is a dense matrix this is irrelevent
9543: Output Parameters:
9544: . D - the product matrix
9546: Notes:
9547: Unless scall is MAT_REUSE_MATRIX D will be created.
9549: MAT_REUSE_MATRIX can only be used if the matrices A, B and C have the same nonzero pattern as in the previous call
9551: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
9552: actually needed.
9554: If you have many matrices with the same non-zero structure to multiply, you
9555: should use MAT_REUSE_MATRIX in all calls but the first or
9557: Level: intermediate
9559: .seealso: MatMatMult, MatPtAP()
9560: @*/
9561: PetscErrorCode MatMatMatMult(Mat A,Mat B,Mat C,MatReuse scall,PetscReal fill,Mat *D)
9562: {
9566: if (scall == MAT_REUSE_MATRIX) MatCheckProduct(*D,6);
9567: if (scall == MAT_INPLACE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Inplace product not supported");
9569: if (scall == MAT_INITIAL_MATRIX) {
9570: MatProductCreate(A,B,C,D);
9571: MatProductSetType(*D,MATPRODUCT_ABC);
9572: MatProductSetAlgorithm(*D,"default");
9573: MatProductSetFill(*D,fill);
9575: (*D)->product->api_user = PETSC_TRUE;
9576: MatProductSetFromOptions(*D);
9577: if (!(*D)->ops->productsymbolic) SETERRQ4(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,((PetscObject)C)->type_name);
9578: MatProductSymbolic(*D);
9579: } else { /* user may change input matrices when REUSE */
9580: MatProductReplaceMats(A,B,C,*D);
9581: }
9582: MatProductNumeric(*D);
9583: return(0);
9584: }
9586: /*@
9587: MatCreateRedundantMatrix - Create redundant matrices and put them into processors of subcommunicators.
9589: Collective on Mat
9591: Input Parameters:
9592: + mat - the matrix
9593: . nsubcomm - the number of subcommunicators (= number of redundant parallel or sequential matrices)
9594: . subcomm - MPI communicator split from the communicator where mat resides in (or MPI_COMM_NULL if nsubcomm is used)
9595: - reuse - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9597: Output Parameter:
9598: . matredundant - redundant matrix
9600: Notes:
9601: MAT_REUSE_MATRIX can only be used when the nonzero structure of the
9602: original matrix has not changed from that last call to MatCreateRedundantMatrix().
9604: This routine creates the duplicated matrices in subcommunicators; you should NOT create them before
9605: calling it.
9607: Level: advanced
9610: .seealso: MatDestroy()
9611: @*/
9612: PetscErrorCode MatCreateRedundantMatrix(Mat mat,PetscInt nsubcomm,MPI_Comm subcomm,MatReuse reuse,Mat *matredundant)
9613: {
9615: MPI_Comm comm;
9616: PetscMPIInt size;
9617: PetscInt mloc_sub,nloc_sub,rstart,rend,M=mat->rmap->N,N=mat->cmap->N,bs=mat->rmap->bs;
9618: Mat_Redundant *redund=NULL;
9619: PetscSubcomm psubcomm=NULL;
9620: MPI_Comm subcomm_in=subcomm;
9621: Mat *matseq;
9622: IS isrow,iscol;
9623: PetscBool newsubcomm=PETSC_FALSE;
9627: if (nsubcomm && reuse == MAT_REUSE_MATRIX) {
9630: }
9632: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
9633: if (size == 1 || nsubcomm == 1) {
9634: if (reuse == MAT_INITIAL_MATRIX) {
9635: MatDuplicate(mat,MAT_COPY_VALUES,matredundant);
9636: } else {
9637: if (*matredundant == mat) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"MAT_REUSE_MATRIX means reuse the matrix passed in as the final argument, not the original matrix");
9638: MatCopy(mat,*matredundant,SAME_NONZERO_PATTERN);
9639: }
9640: return(0);
9641: }
9643: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9644: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9645: MatCheckPreallocated(mat,1);
9647: PetscLogEventBegin(MAT_RedundantMat,mat,0,0,0);
9648: if (subcomm_in == MPI_COMM_NULL && reuse == MAT_INITIAL_MATRIX) { /* get subcomm if user does not provide subcomm */
9649: /* create psubcomm, then get subcomm */
9650: PetscObjectGetComm((PetscObject)mat,&comm);
9651: MPI_Comm_size(comm,&size);
9652: if (nsubcomm < 1 || nsubcomm > size) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"nsubcomm must between 1 and %D",size);
9654: PetscSubcommCreate(comm,&psubcomm);
9655: PetscSubcommSetNumber(psubcomm,nsubcomm);
9656: PetscSubcommSetType(psubcomm,PETSC_SUBCOMM_CONTIGUOUS);
9657: PetscSubcommSetFromOptions(psubcomm);
9658: PetscCommDuplicate(PetscSubcommChild(psubcomm),&subcomm,NULL);
9659: newsubcomm = PETSC_TRUE;
9660: PetscSubcommDestroy(&psubcomm);
9661: }
9663: /* get isrow, iscol and a local sequential matrix matseq[0] */
9664: if (reuse == MAT_INITIAL_MATRIX) {
9665: mloc_sub = PETSC_DECIDE;
9666: nloc_sub = PETSC_DECIDE;
9667: if (bs < 1) {
9668: PetscSplitOwnership(subcomm,&mloc_sub,&M);
9669: PetscSplitOwnership(subcomm,&nloc_sub,&N);
9670: } else {
9671: PetscSplitOwnershipBlock(subcomm,bs,&mloc_sub,&M);
9672: PetscSplitOwnershipBlock(subcomm,bs,&nloc_sub,&N);
9673: }
9674: MPI_Scan(&mloc_sub,&rend,1,MPIU_INT,MPI_SUM,subcomm);
9675: rstart = rend - mloc_sub;
9676: ISCreateStride(PETSC_COMM_SELF,mloc_sub,rstart,1,&isrow);
9677: ISCreateStride(PETSC_COMM_SELF,N,0,1,&iscol);
9678: } else { /* reuse == MAT_REUSE_MATRIX */
9679: if (*matredundant == mat) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"MAT_REUSE_MATRIX means reuse the matrix passed in as the final argument, not the original matrix");
9680: /* retrieve subcomm */
9681: PetscObjectGetComm((PetscObject)(*matredundant),&subcomm);
9682: redund = (*matredundant)->redundant;
9683: isrow = redund->isrow;
9684: iscol = redund->iscol;
9685: matseq = redund->matseq;
9686: }
9687: MatCreateSubMatrices(mat,1,&isrow,&iscol,reuse,&matseq);
9689: /* get matredundant over subcomm */
9690: if (reuse == MAT_INITIAL_MATRIX) {
9691: MatCreateMPIMatConcatenateSeqMat(subcomm,matseq[0],nloc_sub,reuse,matredundant);
9693: /* create a supporting struct and attach it to C for reuse */
9694: PetscNewLog(*matredundant,&redund);
9695: (*matredundant)->redundant = redund;
9696: redund->isrow = isrow;
9697: redund->iscol = iscol;
9698: redund->matseq = matseq;
9699: if (newsubcomm) {
9700: redund->subcomm = subcomm;
9701: } else {
9702: redund->subcomm = MPI_COMM_NULL;
9703: }
9704: } else {
9705: MatCreateMPIMatConcatenateSeqMat(subcomm,matseq[0],PETSC_DECIDE,reuse,matredundant);
9706: }
9707: PetscLogEventEnd(MAT_RedundantMat,mat,0,0,0);
9708: return(0);
9709: }
9711: /*@C
9712: MatGetMultiProcBlock - Create multiple [bjacobi] 'parallel submatrices' from
9713: a given 'mat' object. Each submatrix can span multiple procs.
9715: Collective on Mat
9717: Input Parameters:
9718: + mat - the matrix
9719: . subcomm - the subcommunicator obtained by com_split(comm)
9720: - scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9722: Output Parameter:
9723: . subMat - 'parallel submatrices each spans a given subcomm
9725: Notes:
9726: The submatrix partition across processors is dictated by 'subComm' a
9727: communicator obtained by com_split(comm). The comm_split
9728: is not restriced to be grouped with consecutive original ranks.
9730: Due the comm_split() usage, the parallel layout of the submatrices
9731: map directly to the layout of the original matrix [wrt the local
9732: row,col partitioning]. So the original 'DiagonalMat' naturally maps
9733: into the 'DiagonalMat' of the subMat, hence it is used directly from
9734: the subMat. However the offDiagMat looses some columns - and this is
9735: reconstructed with MatSetValues()
9737: Level: advanced
9740: .seealso: MatCreateSubMatrices()
9741: @*/
9742: PetscErrorCode MatGetMultiProcBlock(Mat mat, MPI_Comm subComm, MatReuse scall,Mat *subMat)
9743: {
9745: PetscMPIInt commsize,subCommSize;
9748: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&commsize);
9749: MPI_Comm_size(subComm,&subCommSize);
9750: if (subCommSize > commsize) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"CommSize %D < SubCommZize %D",commsize,subCommSize);
9752: if (scall == MAT_REUSE_MATRIX && *subMat == mat) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"MAT_REUSE_MATRIX means reuse the matrix passed in as the final argument, not the original matrix");
9753: PetscLogEventBegin(MAT_GetMultiProcBlock,mat,0,0,0);
9754: (*mat->ops->getmultiprocblock)(mat,subComm,scall,subMat);
9755: PetscLogEventEnd(MAT_GetMultiProcBlock,mat,0,0,0);
9756: return(0);
9757: }
9759: /*@
9760: MatGetLocalSubMatrix - Gets a reference to a submatrix specified in local numbering
9762: Not Collective
9764: Input Arguments:
9765: + mat - matrix to extract local submatrix from
9766: . isrow - local row indices for submatrix
9767: - iscol - local column indices for submatrix
9769: Output Arguments:
9770: . submat - the submatrix
9772: Level: intermediate
9774: Notes:
9775: The submat should be returned with MatRestoreLocalSubMatrix().
9777: Depending on the format of mat, the returned submat may not implement MatMult(). Its communicator may be
9778: the same as mat, it may be PETSC_COMM_SELF, or some other subcomm of mat's.
9780: The submat always implements MatSetValuesLocal(). If isrow and iscol have the same block size, then
9781: MatSetValuesBlockedLocal() will also be implemented.
9783: The mat must have had a ISLocalToGlobalMapping provided to it with MatSetLocalToGlobalMapping(). Note that
9784: matrices obtained with DMCreateMatrix() generally already have the local to global mapping provided.
9786: .seealso: MatRestoreLocalSubMatrix(), MatCreateLocalRef(), MatSetLocalToGlobalMapping()
9787: @*/
9788: PetscErrorCode MatGetLocalSubMatrix(Mat mat,IS isrow,IS iscol,Mat *submat)
9789: {
9798: if (!mat->rmap->mapping) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Matrix must have local to global mapping provided before this call");
9800: if (mat->ops->getlocalsubmatrix) {
9801: (*mat->ops->getlocalsubmatrix)(mat,isrow,iscol,submat);
9802: } else {
9803: MatCreateLocalRef(mat,isrow,iscol,submat);
9804: }
9805: return(0);
9806: }
9808: /*@
9809: MatRestoreLocalSubMatrix - Restores a reference to a submatrix specified in local numbering
9811: Not Collective
9813: Input Arguments:
9814: mat - matrix to extract local submatrix from
9815: isrow - local row indices for submatrix
9816: iscol - local column indices for submatrix
9817: submat - the submatrix
9819: Level: intermediate
9821: .seealso: MatGetLocalSubMatrix()
9822: @*/
9823: PetscErrorCode MatRestoreLocalSubMatrix(Mat mat,IS isrow,IS iscol,Mat *submat)
9824: {
9833: if (*submat) {
9835: }
9837: if (mat->ops->restorelocalsubmatrix) {
9838: (*mat->ops->restorelocalsubmatrix)(mat,isrow,iscol,submat);
9839: } else {
9840: MatDestroy(submat);
9841: }
9842: *submat = NULL;
9843: return(0);
9844: }
9846: /* --------------------------------------------------------*/
9847: /*@
9848: MatFindZeroDiagonals - Finds all the rows of a matrix that have zero or no diagonal entry in the matrix
9850: Collective on Mat
9852: Input Parameter:
9853: . mat - the matrix
9855: Output Parameter:
9856: . is - if any rows have zero diagonals this contains the list of them
9858: Level: developer
9860: .seealso: MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
9861: @*/
9862: PetscErrorCode MatFindZeroDiagonals(Mat mat,IS *is)
9863: {
9869: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9870: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9872: if (!mat->ops->findzerodiagonals) {
9873: Vec diag;
9874: const PetscScalar *a;
9875: PetscInt *rows;
9876: PetscInt rStart, rEnd, r, nrow = 0;
9878: MatCreateVecs(mat, &diag, NULL);
9879: MatGetDiagonal(mat, diag);
9880: MatGetOwnershipRange(mat, &rStart, &rEnd);
9881: VecGetArrayRead(diag, &a);
9882: for (r = 0; r < rEnd-rStart; ++r) if (a[r] == 0.0) ++nrow;
9883: PetscMalloc1(nrow, &rows);
9884: nrow = 0;
9885: for (r = 0; r < rEnd-rStart; ++r) if (a[r] == 0.0) rows[nrow++] = r+rStart;
9886: VecRestoreArrayRead(diag, &a);
9887: VecDestroy(&diag);
9888: ISCreateGeneral(PetscObjectComm((PetscObject) mat), nrow, rows, PETSC_OWN_POINTER, is);
9889: } else {
9890: (*mat->ops->findzerodiagonals)(mat, is);
9891: }
9892: return(0);
9893: }
9895: /*@
9896: MatFindOffBlockDiagonalEntries - Finds all the rows of a matrix that have entries outside of the main diagonal block (defined by the matrix block size)
9898: Collective on Mat
9900: Input Parameter:
9901: . mat - the matrix
9903: Output Parameter:
9904: . is - contains the list of rows with off block diagonal entries
9906: Level: developer
9908: .seealso: MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
9909: @*/
9910: PetscErrorCode MatFindOffBlockDiagonalEntries(Mat mat,IS *is)
9911: {
9917: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9918: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9920: if (!mat->ops->findoffblockdiagonalentries) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s does not have a find off block diagonal entries defined",((PetscObject)mat)->type_name);
9921: (*mat->ops->findoffblockdiagonalentries)(mat,is);
9922: return(0);
9923: }
9925: /*@C
9926: MatInvertBlockDiagonal - Inverts the block diagonal entries.
9928: Collective on Mat
9930: Input Parameters:
9931: . mat - the matrix
9933: Output Parameters:
9934: . values - the block inverses in column major order (FORTRAN-like)
9936: Note:
9937: This routine is not available from Fortran.
9939: Level: advanced
9941: .seealso: MatInvertBockDiagonalMat
9942: @*/
9943: PetscErrorCode MatInvertBlockDiagonal(Mat mat,const PetscScalar **values)
9944: {
9949: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9950: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9951: if (!mat->ops->invertblockdiagonal) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not supported for type %s",((PetscObject)mat)->type_name);
9952: (*mat->ops->invertblockdiagonal)(mat,values);
9953: return(0);
9954: }
9956: /*@C
9957: MatInvertVariableBlockDiagonal - Inverts the block diagonal entries.
9959: Collective on Mat
9961: Input Parameters:
9962: + mat - the matrix
9963: . nblocks - the number of blocks
9964: - bsizes - the size of each block
9966: Output Parameters:
9967: . values - the block inverses in column major order (FORTRAN-like)
9969: Note:
9970: This routine is not available from Fortran.
9972: Level: advanced
9974: .seealso: MatInvertBockDiagonal()
9975: @*/
9976: PetscErrorCode MatInvertVariableBlockDiagonal(Mat mat,PetscInt nblocks,const PetscInt *bsizes,PetscScalar *values)
9977: {
9982: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9983: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9984: if (!mat->ops->invertvariableblockdiagonal) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not supported for type",((PetscObject)mat)->type_name);
9985: (*mat->ops->invertvariableblockdiagonal)(mat,nblocks,bsizes,values);
9986: return(0);
9987: }
9989: /*@
9990: MatInvertBlockDiagonalMat - set matrix C to be the inverted block diagonal of matrix A
9992: Collective on Mat
9994: Input Parameters:
9995: . A - the matrix
9997: Output Parameters:
9998: . C - matrix with inverted block diagonal of A. This matrix should be created and may have its type set.
10000: Notes: the blocksize of the matrix is used to determine the blocks on the diagonal of C
10002: Level: advanced
10004: .seealso: MatInvertBockDiagonal()
10005: @*/
10006: PetscErrorCode MatInvertBlockDiagonalMat(Mat A,Mat C)
10007: {
10008: PetscErrorCode ierr;
10009: const PetscScalar *vals;
10010: PetscInt *dnnz;
10011: PetscInt M,N,m,n,rstart,rend,bs,i,j;
10014: MatInvertBlockDiagonal(A,&vals);
10015: MatGetBlockSize(A,&bs);
10016: MatGetSize(A,&M,&N);
10017: MatGetLocalSize(A,&m,&n);
10018: MatSetSizes(C,m,n,M,N);
10019: MatSetBlockSize(C,bs);
10020: PetscMalloc1(m/bs,&dnnz);
10021: for (j = 0; j < m/bs; j++) dnnz[j] = 1;
10022: MatXAIJSetPreallocation(C,bs,dnnz,NULL,NULL,NULL);
10023: PetscFree(dnnz);
10024: MatGetOwnershipRange(C,&rstart,&rend);
10025: MatSetOption(C,MAT_ROW_ORIENTED,PETSC_FALSE);
10026: for (i = rstart/bs; i < rend/bs; i++) {
10027: MatSetValuesBlocked(C,1,&i,1,&i,&vals[(i-rstart/bs)*bs*bs],INSERT_VALUES);
10028: }
10029: MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);
10030: MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);
10031: MatSetOption(C,MAT_ROW_ORIENTED,PETSC_TRUE);
10032: return(0);
10033: }
10035: /*@C
10036: MatTransposeColoringDestroy - Destroys a coloring context for matrix product C=A*B^T that was created
10037: via MatTransposeColoringCreate().
10039: Collective on MatTransposeColoring
10041: Input Parameter:
10042: . c - coloring context
10044: Level: intermediate
10046: .seealso: MatTransposeColoringCreate()
10047: @*/
10048: PetscErrorCode MatTransposeColoringDestroy(MatTransposeColoring *c)
10049: {
10050: PetscErrorCode ierr;
10051: MatTransposeColoring matcolor=*c;
10054: if (!matcolor) return(0);
10055: if (--((PetscObject)matcolor)->refct > 0) {matcolor = NULL; return(0);}
10057: PetscFree3(matcolor->ncolumns,matcolor->nrows,matcolor->colorforrow);
10058: PetscFree(matcolor->rows);
10059: PetscFree(matcolor->den2sp);
10060: PetscFree(matcolor->colorforcol);
10061: PetscFree(matcolor->columns);
10062: if (matcolor->brows>0) {
10063: PetscFree(matcolor->lstart);
10064: }
10065: PetscHeaderDestroy(c);
10066: return(0);
10067: }
10069: /*@C
10070: MatTransColoringApplySpToDen - Given a symbolic matrix product C=A*B^T for which
10071: a MatTransposeColoring context has been created, computes a dense B^T by Apply
10072: MatTransposeColoring to sparse B.
10074: Collective on MatTransposeColoring
10076: Input Parameters:
10077: + B - sparse matrix B
10078: . Btdense - symbolic dense matrix B^T
10079: - coloring - coloring context created with MatTransposeColoringCreate()
10081: Output Parameter:
10082: . Btdense - dense matrix B^T
10084: Level: advanced
10086: Notes:
10087: These are used internally for some implementations of MatRARt()
10089: .seealso: MatTransposeColoringCreate(), MatTransposeColoringDestroy(), MatTransColoringApplyDenToSp()
10091: @*/
10092: PetscErrorCode MatTransColoringApplySpToDen(MatTransposeColoring coloring,Mat B,Mat Btdense)
10093: {
10101: if (!B->ops->transcoloringapplysptoden) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not supported for this matrix type %s",((PetscObject)B)->type_name);
10102: (B->ops->transcoloringapplysptoden)(coloring,B,Btdense);
10103: return(0);
10104: }
10106: /*@C
10107: MatTransColoringApplyDenToSp - Given a symbolic matrix product Csp=A*B^T for which
10108: a MatTransposeColoring context has been created and a dense matrix Cden=A*Btdense
10109: in which Btdens is obtained from MatTransColoringApplySpToDen(), recover sparse matrix
10110: Csp from Cden.
10112: Collective on MatTransposeColoring
10114: Input Parameters:
10115: + coloring - coloring context created with MatTransposeColoringCreate()
10116: - Cden - matrix product of a sparse matrix and a dense matrix Btdense
10118: Output Parameter:
10119: . Csp - sparse matrix
10121: Level: advanced
10123: Notes:
10124: These are used internally for some implementations of MatRARt()
10126: .seealso: MatTransposeColoringCreate(), MatTransposeColoringDestroy(), MatTransColoringApplySpToDen()
10128: @*/
10129: PetscErrorCode MatTransColoringApplyDenToSp(MatTransposeColoring matcoloring,Mat Cden,Mat Csp)
10130: {
10138: if (!Csp->ops->transcoloringapplydentosp) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not supported for this matrix type %s",((PetscObject)Csp)->type_name);
10139: (Csp->ops->transcoloringapplydentosp)(matcoloring,Cden,Csp);
10140: MatAssemblyBegin(Csp,MAT_FINAL_ASSEMBLY);
10141: MatAssemblyEnd(Csp,MAT_FINAL_ASSEMBLY);
10142: return(0);
10143: }
10145: /*@C
10146: MatTransposeColoringCreate - Creates a matrix coloring context for matrix product C=A*B^T.
10148: Collective on Mat
10150: Input Parameters:
10151: + mat - the matrix product C
10152: - iscoloring - the coloring of the matrix; usually obtained with MatColoringCreate() or DMCreateColoring()
10154: Output Parameter:
10155: . color - the new coloring context
10157: Level: intermediate
10159: .seealso: MatTransposeColoringDestroy(), MatTransColoringApplySpToDen(),
10160: MatTransColoringApplyDenToSp()
10161: @*/
10162: PetscErrorCode MatTransposeColoringCreate(Mat mat,ISColoring iscoloring,MatTransposeColoring *color)
10163: {
10164: MatTransposeColoring c;
10165: MPI_Comm comm;
10166: PetscErrorCode ierr;
10169: PetscLogEventBegin(MAT_TransposeColoringCreate,mat,0,0,0);
10170: PetscObjectGetComm((PetscObject)mat,&comm);
10171: PetscHeaderCreate(c,MAT_TRANSPOSECOLORING_CLASSID,"MatTransposeColoring","Matrix product C=A*B^T via coloring","Mat",comm,MatTransposeColoringDestroy,NULL);
10173: c->ctype = iscoloring->ctype;
10174: if (mat->ops->transposecoloringcreate) {
10175: (*mat->ops->transposecoloringcreate)(mat,iscoloring,c);
10176: } else SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Code not yet written for matrix type %s",((PetscObject)mat)->type_name);
10178: *color = c;
10179: PetscLogEventEnd(MAT_TransposeColoringCreate,mat,0,0,0);
10180: return(0);
10181: }
10183: /*@
10184: MatGetNonzeroState - Returns a 64 bit integer representing the current state of nonzeros in the matrix. If the
10185: matrix has had no new nonzero locations added to the matrix since the previous call then the value will be the
10186: same, otherwise it will be larger
10188: Not Collective
10190: Input Parameter:
10191: . A - the matrix
10193: Output Parameter:
10194: . state - the current state
10196: Notes:
10197: You can only compare states from two different calls to the SAME matrix, you cannot compare calls between
10198: different matrices
10200: Level: intermediate
10202: @*/
10203: PetscErrorCode MatGetNonzeroState(Mat mat,PetscObjectState *state)
10204: {
10207: *state = mat->nonzerostate;
10208: return(0);
10209: }
10211: /*@
10212: MatCreateMPIMatConcatenateSeqMat - Creates a single large PETSc matrix by concatenating sequential
10213: matrices from each processor
10215: Collective
10217: Input Parameters:
10218: + comm - the communicators the parallel matrix will live on
10219: . seqmat - the input sequential matrices
10220: . n - number of local columns (or PETSC_DECIDE)
10221: - reuse - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
10223: Output Parameter:
10224: . mpimat - the parallel matrix generated
10226: Level: advanced
10228: Notes:
10229: The number of columns of the matrix in EACH processor MUST be the same.
10231: @*/
10232: PetscErrorCode MatCreateMPIMatConcatenateSeqMat(MPI_Comm comm,Mat seqmat,PetscInt n,MatReuse reuse,Mat *mpimat)
10233: {
10237: if (!seqmat->ops->creatempimatconcatenateseqmat) SETERRQ1(PetscObjectComm((PetscObject)seqmat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)seqmat)->type_name);
10238: if (reuse == MAT_REUSE_MATRIX && seqmat == *mpimat) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"MAT_REUSE_MATRIX means reuse the matrix passed in as the final argument, not the original matrix");
10240: PetscLogEventBegin(MAT_Merge,seqmat,0,0,0);
10241: (*seqmat->ops->creatempimatconcatenateseqmat)(comm,seqmat,n,reuse,mpimat);
10242: PetscLogEventEnd(MAT_Merge,seqmat,0,0,0);
10243: return(0);
10244: }
10246: /*@
10247: MatSubdomainsCreateCoalesce - Creates index subdomains by coalescing adjacent
10248: ranks' ownership ranges.
10250: Collective on A
10252: Input Parameters:
10253: + A - the matrix to create subdomains from
10254: - N - requested number of subdomains
10257: Output Parameters:
10258: + n - number of subdomains resulting on this rank
10259: - iss - IS list with indices of subdomains on this rank
10261: Level: advanced
10263: Notes:
10264: number of subdomains must be smaller than the communicator size
10265: @*/
10266: PetscErrorCode MatSubdomainsCreateCoalesce(Mat A,PetscInt N,PetscInt *n,IS *iss[])
10267: {
10268: MPI_Comm comm,subcomm;
10269: PetscMPIInt size,rank,color;
10270: PetscInt rstart,rend,k;
10271: PetscErrorCode ierr;
10274: PetscObjectGetComm((PetscObject)A,&comm);
10275: MPI_Comm_size(comm,&size);
10276: MPI_Comm_rank(comm,&rank);
10277: if (N < 1 || N >= (PetscInt)size) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"number of subdomains must be > 0 and < %D, got N = %D",size,N);
10278: *n = 1;
10279: k = ((PetscInt)size)/N + ((PetscInt)size%N>0); /* There are up to k ranks to a color */
10280: color = rank/k;
10281: MPI_Comm_split(comm,color,rank,&subcomm);
10282: PetscMalloc1(1,iss);
10283: MatGetOwnershipRange(A,&rstart,&rend);
10284: ISCreateStride(subcomm,rend-rstart,rstart,1,iss[0]);
10285: MPI_Comm_free(&subcomm);
10286: return(0);
10287: }
10289: /*@
10290: MatGalerkin - Constructs the coarse grid problem via Galerkin projection.
10292: If the interpolation and restriction operators are the same, uses MatPtAP.
10293: If they are not the same, use MatMatMatMult.
10295: Once the coarse grid problem is constructed, correct for interpolation operators
10296: that are not of full rank, which can legitimately happen in the case of non-nested
10297: geometric multigrid.
10299: Input Parameters:
10300: + restrct - restriction operator
10301: . dA - fine grid matrix
10302: . interpolate - interpolation operator
10303: . reuse - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
10304: - fill - expected fill, use PETSC_DEFAULT if you do not have a good estimate
10306: Output Parameters:
10307: . A - the Galerkin coarse matrix
10309: Options Database Key:
10310: . -pc_mg_galerkin <both,pmat,mat,none>
10312: Level: developer
10314: .seealso: MatPtAP(), MatMatMatMult()
10315: @*/
10316: PetscErrorCode MatGalerkin(Mat restrct, Mat dA, Mat interpolate, MatReuse reuse, PetscReal fill, Mat *A)
10317: {
10319: IS zerorows;
10320: Vec diag;
10323: if (reuse == MAT_INPLACE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Inplace product not supported");
10324: /* Construct the coarse grid matrix */
10325: if (interpolate == restrct) {
10326: MatPtAP(dA,interpolate,reuse,fill,A);
10327: } else {
10328: MatMatMatMult(restrct,dA,interpolate,reuse,fill,A);
10329: }
10331: /* If the interpolation matrix is not of full rank, A will have zero rows.
10332: This can legitimately happen in the case of non-nested geometric multigrid.
10333: In that event, we set the rows of the matrix to the rows of the identity,
10334: ignoring the equations (as the RHS will also be zero). */
10336: MatFindZeroRows(*A, &zerorows);
10338: if (zerorows != NULL) { /* if there are any zero rows */
10339: MatCreateVecs(*A, &diag, NULL);
10340: MatGetDiagonal(*A, diag);
10341: VecISSet(diag, zerorows, 1.0);
10342: MatDiagonalSet(*A, diag, INSERT_VALUES);
10343: VecDestroy(&diag);
10344: ISDestroy(&zerorows);
10345: }
10346: return(0);
10347: }
10349: /*@C
10350: MatSetOperation - Allows user to set a matrix operation for any matrix type
10352: Logically Collective on Mat
10354: Input Parameters:
10355: + mat - the matrix
10356: . op - the name of the operation
10357: - f - the function that provides the operation
10359: Level: developer
10361: Usage:
10362: $ extern PetscErrorCode usermult(Mat,Vec,Vec);
10363: $ MatCreateXXX(comm,...&A);
10364: $ MatSetOperation(A,MATOP_MULT,(void(*)(void))usermult);
10366: Notes:
10367: See the file include/petscmat.h for a complete list of matrix
10368: operations, which all have the form MATOP_<OPERATION>, where
10369: <OPERATION> is the name (in all capital letters) of the
10370: user interface routine (e.g., MatMult() -> MATOP_MULT).
10372: All user-provided functions (except for MATOP_DESTROY) should have the same calling
10373: sequence as the usual matrix interface routines, since they
10374: are intended to be accessed via the usual matrix interface
10375: routines, e.g.,
10376: $ MatMult(Mat,Vec,Vec) -> usermult(Mat,Vec,Vec)
10378: In particular each function MUST return an error code of 0 on success and
10379: nonzero on failure.
10381: This routine is distinct from MatShellSetOperation() in that it can be called on any matrix type.
10383: .seealso: MatGetOperation(), MatCreateShell(), MatShellSetContext(), MatShellSetOperation()
10384: @*/
10385: PetscErrorCode MatSetOperation(Mat mat,MatOperation op,void (*f)(void))
10386: {
10389: if (op == MATOP_VIEW && !mat->ops->viewnative && f != (void (*)(void))(mat->ops->view)) {
10390: mat->ops->viewnative = mat->ops->view;
10391: }
10392: (((void(**)(void))mat->ops)[op]) = f;
10393: return(0);
10394: }
10396: /*@C
10397: MatGetOperation - Gets a matrix operation for any matrix type.
10399: Not Collective
10401: Input Parameters:
10402: + mat - the matrix
10403: - op - the name of the operation
10405: Output Parameter:
10406: . f - the function that provides the operation
10408: Level: developer
10410: Usage:
10411: $ PetscErrorCode (*usermult)(Mat,Vec,Vec);
10412: $ MatGetOperation(A,MATOP_MULT,(void(**)(void))&usermult);
10414: Notes:
10415: See the file include/petscmat.h for a complete list of matrix
10416: operations, which all have the form MATOP_<OPERATION>, where
10417: <OPERATION> is the name (in all capital letters) of the
10418: user interface routine (e.g., MatMult() -> MATOP_MULT).
10420: This routine is distinct from MatShellGetOperation() in that it can be called on any matrix type.
10422: .seealso: MatSetOperation(), MatCreateShell(), MatShellGetContext(), MatShellGetOperation()
10423: @*/
10424: PetscErrorCode MatGetOperation(Mat mat,MatOperation op,void(**f)(void))
10425: {
10428: *f = (((void (**)(void))mat->ops)[op]);
10429: return(0);
10430: }
10432: /*@
10433: MatHasOperation - Determines whether the given matrix supports the particular
10434: operation.
10436: Not Collective
10438: Input Parameters:
10439: + mat - the matrix
10440: - op - the operation, for example, MATOP_GET_DIAGONAL
10442: Output Parameter:
10443: . has - either PETSC_TRUE or PETSC_FALSE
10445: Level: advanced
10447: Notes:
10448: See the file include/petscmat.h for a complete list of matrix
10449: operations, which all have the form MATOP_<OPERATION>, where
10450: <OPERATION> is the name (in all capital letters) of the
10451: user-level routine. E.g., MatNorm() -> MATOP_NORM.
10453: .seealso: MatCreateShell()
10454: @*/
10455: PetscErrorCode MatHasOperation(Mat mat,MatOperation op,PetscBool *has)
10456: {
10461: /* symbolic product can be set before matrix type */
10464: if (mat->ops->hasoperation) {
10465: (*mat->ops->hasoperation)(mat,op,has);
10466: } else {
10467: if (((void**)mat->ops)[op]) *has = PETSC_TRUE;
10468: else {
10469: *has = PETSC_FALSE;
10470: if (op == MATOP_CREATE_SUBMATRIX) {
10471: PetscMPIInt size;
10473: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
10474: if (size == 1) {
10475: MatHasOperation(mat,MATOP_CREATE_SUBMATRICES,has);
10476: }
10477: }
10478: }
10479: }
10480: return(0);
10481: }
10483: /*@
10484: MatHasCongruentLayouts - Determines whether the rows and columns layouts
10485: of the matrix are congruent
10487: Collective on mat
10489: Input Parameters:
10490: . mat - the matrix
10492: Output Parameter:
10493: . cong - either PETSC_TRUE or PETSC_FALSE
10495: Level: beginner
10497: Notes:
10499: .seealso: MatCreate(), MatSetSizes()
10500: @*/
10501: PetscErrorCode MatHasCongruentLayouts(Mat mat,PetscBool *cong)
10502: {
10509: if (!mat->rmap || !mat->cmap) {
10510: *cong = mat->rmap == mat->cmap ? PETSC_TRUE : PETSC_FALSE;
10511: return(0);
10512: }
10513: if (mat->congruentlayouts == PETSC_DECIDE) { /* first time we compare rows and cols layouts */
10514: PetscLayoutCompare(mat->rmap,mat->cmap,cong);
10515: if (*cong) mat->congruentlayouts = 1;
10516: else mat->congruentlayouts = 0;
10517: } else *cong = mat->congruentlayouts ? PETSC_TRUE : PETSC_FALSE;
10518: return(0);
10519: }
10521: PetscErrorCode MatSetInf(Mat A)
10522: {
10526: if (!A->ops->setinf) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"No support for this operation for this matrix type");
10527: (*A->ops->setinf)(A);
10528: return(0);
10529: }