Actual source code: matrix.c
petsc-3.12.5 2020-03-29
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_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_",0};
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.
116: This routine should only be called if MatGetFactorError() returns a value of MAT_FACTOR_NUMERIC_ZEROPIVOT
118: This can be called on non-factored matrices that come from, for example, matrices used in SOR.
120: .seealso: MatZeroEntries(), MatFactor(), MatGetFactor(), MatFactorSymbolic(), MatFactorClearError(), MatFactorGetErrorZeroPivot()
121: @*/
122: PetscErrorCode MatFactorGetErrorZeroPivot(Mat mat,PetscReal *pivot,PetscInt *row)
123: {
126: *pivot = mat->factorerror_zeropivot_value;
127: *row = mat->factorerror_zeropivot_row;
128: return(0);
129: }
131: /*@
132: MatFactorGetError - gets the error code from a factorization
134: Logically Collective on Mat
136: Input Parameters:
137: . mat - the factored matrix
139: Output Parameter:
140: . err - the error code
142: Level: advanced
144: Notes:
145: This can be called on non-factored matrices that come from, for example, matrices used in SOR.
147: .seealso: MatZeroEntries(), MatFactor(), MatGetFactor(), MatFactorSymbolic(), MatFactorClearError(), MatFactorGetErrorZeroPivot()
148: @*/
149: PetscErrorCode MatFactorGetError(Mat mat,MatFactorError *err)
150: {
153: *err = mat->factorerrortype;
154: return(0);
155: }
157: /*@
158: MatFactorClearError - clears the error code in a factorization
160: Logically Collective on Mat
162: Input Parameter:
163: . mat - the factored matrix
165: Level: developer
167: Notes:
168: This can be called on non-factored matrices that come from, for example, matrices used in SOR.
170: .seealso: MatZeroEntries(), MatFactor(), MatGetFactor(), MatFactorSymbolic(), MatFactorGetError(), MatFactorGetErrorZeroPivot()
171: @*/
172: PetscErrorCode MatFactorClearError(Mat mat)
173: {
176: mat->factorerrortype = MAT_FACTOR_NOERROR;
177: mat->factorerror_zeropivot_value = 0.0;
178: mat->factorerror_zeropivot_row = 0;
179: return(0);
180: }
182: PETSC_INTERN PetscErrorCode MatFindNonzeroRowsOrCols_Basic(Mat mat,PetscBool cols,PetscReal tol,IS *nonzero)
183: {
184: PetscErrorCode ierr;
185: Vec r,l;
186: const PetscScalar *al;
187: PetscInt i,nz,gnz,N,n;
190: MatCreateVecs(mat,&r,&l);
191: if (!cols) { /* nonzero rows */
192: MatGetSize(mat,&N,NULL);
193: MatGetLocalSize(mat,&n,NULL);
194: VecSet(l,0.0);
195: VecSetRandom(r,NULL);
196: MatMult(mat,r,l);
197: VecGetArrayRead(l,&al);
198: } else { /* nonzero columns */
199: MatGetSize(mat,NULL,&N);
200: MatGetLocalSize(mat,NULL,&n);
201: VecSet(r,0.0);
202: VecSetRandom(l,NULL);
203: MatMultTranspose(mat,l,r);
204: VecGetArrayRead(r,&al);
205: }
206: if (tol <= 0.0) { for (i=0,nz=0;i<n;i++) if (al[i] != 0.0) nz++; }
207: else { for (i=0,nz=0;i<n;i++) if (PetscAbsScalar(al[i]) > tol) nz++; }
208: MPIU_Allreduce(&nz,&gnz,1,MPIU_INT,MPI_SUM,PetscObjectComm((PetscObject)mat));
209: if (gnz != N) {
210: PetscInt *nzr;
211: PetscMalloc1(nz,&nzr);
212: if (nz) {
213: if (tol < 0) { for (i=0,nz=0;i<n;i++) if (al[i] != 0.0) nzr[nz++] = i; }
214: else { for (i=0,nz=0;i<n;i++) if (PetscAbsScalar(al[i]) > tol) nzr[nz++] = i; }
215: }
216: ISCreateGeneral(PetscObjectComm((PetscObject)mat),nz,nzr,PETSC_OWN_POINTER,nonzero);
217: } else *nonzero = NULL;
218: if (!cols) { /* nonzero rows */
219: VecRestoreArrayRead(l,&al);
220: } else {
221: VecRestoreArrayRead(r,&al);
222: }
223: VecDestroy(&l);
224: VecDestroy(&r);
225: return(0);
226: }
228: /*@
229: MatFindNonzeroRows - Locate all rows that are not completely zero in the matrix
231: Input Parameter:
232: . A - the matrix
234: Output Parameter:
235: . keptrows - the rows that are not completely zero
237: Notes:
238: keptrows is set to NULL if all rows are nonzero.
240: Level: intermediate
242: @*/
243: PetscErrorCode MatFindNonzeroRows(Mat mat,IS *keptrows)
244: {
251: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
252: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
253: if (!mat->ops->findnonzerorows) {
254: MatFindNonzeroRowsOrCols_Basic(mat,PETSC_FALSE,0.0,keptrows);
255: } else {
256: (*mat->ops->findnonzerorows)(mat,keptrows);
257: }
258: return(0);
259: }
261: /*@
262: MatFindZeroRows - Locate all rows that are completely zero in the matrix
264: Input Parameter:
265: . A - the matrix
267: Output Parameter:
268: . zerorows - the rows that are completely zero
270: Notes:
271: zerorows is set to NULL if no rows are zero.
273: Level: intermediate
275: @*/
276: PetscErrorCode MatFindZeroRows(Mat mat,IS *zerorows)
277: {
279: IS keptrows;
280: PetscInt m, n;
285: MatFindNonzeroRows(mat, &keptrows);
286: /* MatFindNonzeroRows sets keptrows to NULL if there are no zero rows.
287: In keeping with this convention, we set zerorows to NULL if there are no zero
288: rows. */
289: if (keptrows == NULL) {
290: *zerorows = NULL;
291: } else {
292: MatGetOwnershipRange(mat,&m,&n);
293: ISComplement(keptrows,m,n,zerorows);
294: ISDestroy(&keptrows);
295: }
296: return(0);
297: }
299: /*@
300: MatGetDiagonalBlock - Returns the part of the matrix associated with the on-process coupling
302: Not Collective
304: Input Parameters:
305: . A - the matrix
307: Output Parameters:
308: . a - the diagonal part (which is a SEQUENTIAL matrix)
310: Notes:
311: see the manual page for MatCreateAIJ() for more information on the "diagonal part" of the matrix.
312: Use caution, as the reference count on the returned matrix is not incremented and it is used as
313: part of the containing MPI Mat's normal operation.
315: Level: advanced
317: @*/
318: PetscErrorCode MatGetDiagonalBlock(Mat A,Mat *a)
319: {
326: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
327: if (!A->ops->getdiagonalblock) {
328: PetscMPIInt size;
329: MPI_Comm_size(PetscObjectComm((PetscObject)A),&size);
330: if (size == 1) {
331: *a = A;
332: return(0);
333: } else SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Not coded for matrix type %s",((PetscObject)A)->type_name);
334: }
335: (*A->ops->getdiagonalblock)(A,a);
336: return(0);
337: }
339: /*@
340: MatGetTrace - Gets the trace of a matrix. The sum of the diagonal entries.
342: Collective on Mat
344: Input Parameters:
345: . mat - the matrix
347: Output Parameter:
348: . trace - the sum of the diagonal entries
350: Level: advanced
352: @*/
353: PetscErrorCode MatGetTrace(Mat mat,PetscScalar *trace)
354: {
356: Vec diag;
359: MatCreateVecs(mat,&diag,NULL);
360: MatGetDiagonal(mat,diag);
361: VecSum(diag,trace);
362: VecDestroy(&diag);
363: return(0);
364: }
366: /*@
367: MatRealPart - Zeros out the imaginary part of the matrix
369: Logically Collective on Mat
371: Input Parameters:
372: . mat - the matrix
374: Level: advanced
377: .seealso: MatImaginaryPart()
378: @*/
379: PetscErrorCode MatRealPart(Mat mat)
380: {
386: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
387: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
388: if (!mat->ops->realpart) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
389: MatCheckPreallocated(mat,1);
390: (*mat->ops->realpart)(mat);
391: return(0);
392: }
394: /*@C
395: MatGetGhosts - Get the global index of all ghost nodes defined by the sparse matrix
397: Collective on Mat
399: Input Parameter:
400: . mat - the matrix
402: Output Parameters:
403: + nghosts - number of ghosts (note for BAIJ matrices there is one ghost for each block)
404: - ghosts - the global indices of the ghost points
406: Notes:
407: the nghosts and ghosts are suitable to pass into VecCreateGhost()
409: Level: advanced
411: @*/
412: PetscErrorCode MatGetGhosts(Mat mat,PetscInt *nghosts,const PetscInt *ghosts[])
413: {
419: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
420: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
421: if (!mat->ops->getghosts) {
422: if (nghosts) *nghosts = 0;
423: if (ghosts) *ghosts = 0;
424: } else {
425: (*mat->ops->getghosts)(mat,nghosts,ghosts);
426: }
427: return(0);
428: }
431: /*@
432: MatImaginaryPart - Moves the imaginary part of the matrix to the real part and zeros the imaginary part
434: Logically Collective on Mat
436: Input Parameters:
437: . mat - the matrix
439: Level: advanced
442: .seealso: MatRealPart()
443: @*/
444: PetscErrorCode MatImaginaryPart(Mat mat)
445: {
451: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
452: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
453: if (!mat->ops->imaginarypart) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
454: MatCheckPreallocated(mat,1);
455: (*mat->ops->imaginarypart)(mat);
456: return(0);
457: }
459: /*@
460: MatMissingDiagonal - Determine if sparse matrix is missing a diagonal entry (or block entry for BAIJ matrices)
462: Not Collective
464: Input Parameter:
465: . mat - the matrix
467: Output Parameters:
468: + missing - is any diagonal missing
469: - dd - first diagonal entry that is missing (optional) on this process
471: Level: advanced
474: .seealso: MatRealPart()
475: @*/
476: PetscErrorCode MatMissingDiagonal(Mat mat,PetscBool *missing,PetscInt *dd)
477: {
484: if (!mat->assembled) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix %s",((PetscObject)mat)->type_name);
485: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
486: if (!mat->ops->missingdiagonal) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
487: (*mat->ops->missingdiagonal)(mat,missing,dd);
488: return(0);
489: }
491: /*@C
492: MatGetRow - Gets a row of a matrix. You MUST call MatRestoreRow()
493: for each row that you get to ensure that your application does
494: not bleed memory.
496: Not Collective
498: Input Parameters:
499: + mat - the matrix
500: - row - the row to get
502: Output Parameters:
503: + ncols - if not NULL, the number of nonzeros in the row
504: . cols - if not NULL, the column numbers
505: - vals - if not NULL, the values
507: Notes:
508: This routine is provided for people who need to have direct access
509: to the structure of a matrix. We hope that we provide enough
510: high-level matrix routines that few users will need it.
512: MatGetRow() always returns 0-based column indices, regardless of
513: whether the internal representation is 0-based (default) or 1-based.
515: For better efficiency, set cols and/or vals to NULL if you do
516: not wish to extract these quantities.
518: The user can only examine the values extracted with MatGetRow();
519: the values cannot be altered. To change the matrix entries, one
520: must use MatSetValues().
522: You can only have one call to MatGetRow() outstanding for a particular
523: matrix at a time, per processor. MatGetRow() can only obtain rows
524: associated with the given processor, it cannot get rows from the
525: other processors; for that we suggest using MatCreateSubMatrices(), then
526: MatGetRow() on the submatrix. The row index passed to MatGetRow()
527: is in the global number of rows.
529: Fortran Notes:
530: The calling sequence from Fortran is
531: .vb
532: MatGetRow(matrix,row,ncols,cols,values,ierr)
533: Mat matrix (input)
534: integer row (input)
535: integer ncols (output)
536: integer cols(maxcols) (output)
537: double precision (or double complex) values(maxcols) output
538: .ve
539: where maxcols >= maximum nonzeros in any row of the matrix.
542: Caution:
543: Do not try to change the contents of the output arrays (cols and vals).
544: In some cases, this may corrupt the matrix.
546: Level: advanced
548: .seealso: MatRestoreRow(), MatSetValues(), MatGetValues(), MatCreateSubMatrices(), MatGetDiagonal()
549: @*/
550: PetscErrorCode MatGetRow(Mat mat,PetscInt row,PetscInt *ncols,const PetscInt *cols[],const PetscScalar *vals[])
551: {
553: PetscInt incols;
558: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
559: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
560: if (!mat->ops->getrow) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
561: MatCheckPreallocated(mat,1);
562: PetscLogEventBegin(MAT_GetRow,mat,0,0,0);
563: (*mat->ops->getrow)(mat,row,&incols,(PetscInt**)cols,(PetscScalar**)vals);
564: if (ncols) *ncols = incols;
565: PetscLogEventEnd(MAT_GetRow,mat,0,0,0);
566: return(0);
567: }
569: /*@
570: MatConjugate - replaces the matrix values with their complex conjugates
572: Logically Collective on Mat
574: Input Parameters:
575: . mat - the matrix
577: Level: advanced
579: .seealso: VecConjugate()
580: @*/
581: PetscErrorCode MatConjugate(Mat mat)
582: {
583: #if defined(PETSC_USE_COMPLEX)
588: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
589: 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);
590: (*mat->ops->conjugate)(mat);
591: #else
593: #endif
594: return(0);
595: }
597: /*@C
598: MatRestoreRow - Frees any temporary space allocated by MatGetRow().
600: Not Collective
602: Input Parameters:
603: + mat - the matrix
604: . row - the row to get
605: . ncols, cols - the number of nonzeros and their columns
606: - vals - if nonzero the column values
608: Notes:
609: This routine should be called after you have finished examining the entries.
611: This routine zeros out ncols, cols, and vals. This is to prevent accidental
612: us of the array after it has been restored. If you pass NULL, it will
613: not zero the pointers. Use of cols or vals after MatRestoreRow is invalid.
615: Fortran Notes:
616: The calling sequence from Fortran is
617: .vb
618: MatRestoreRow(matrix,row,ncols,cols,values,ierr)
619: Mat matrix (input)
620: integer row (input)
621: integer ncols (output)
622: integer cols(maxcols) (output)
623: double precision (or double complex) values(maxcols) output
624: .ve
625: Where maxcols >= maximum nonzeros in any row of the matrix.
627: In Fortran MatRestoreRow() MUST be called after MatGetRow()
628: before another call to MatGetRow() can be made.
630: Level: advanced
632: .seealso: MatGetRow()
633: @*/
634: PetscErrorCode MatRestoreRow(Mat mat,PetscInt row,PetscInt *ncols,const PetscInt *cols[],const PetscScalar *vals[])
635: {
641: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
642: if (!mat->ops->restorerow) return(0);
643: (*mat->ops->restorerow)(mat,row,ncols,(PetscInt **)cols,(PetscScalar **)vals);
644: if (ncols) *ncols = 0;
645: if (cols) *cols = NULL;
646: if (vals) *vals = NULL;
647: return(0);
648: }
650: /*@
651: MatGetRowUpperTriangular - Sets a flag to enable calls to MatGetRow() for matrix in MATSBAIJ format.
652: You should call MatRestoreRowUpperTriangular() after calling MatGetRow/MatRestoreRow() to disable the flag.
654: Not Collective
656: Input Parameters:
657: . mat - the matrix
659: Notes:
660: 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.
662: Level: advanced
664: .seealso: MatRestoreRowUpperTriangular()
665: @*/
666: PetscErrorCode MatGetRowUpperTriangular(Mat mat)
667: {
673: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
674: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
675: MatCheckPreallocated(mat,1);
676: if (!mat->ops->getrowuppertriangular) return(0);
677: (*mat->ops->getrowuppertriangular)(mat);
678: return(0);
679: }
681: /*@
682: MatRestoreRowUpperTriangular - Disable calls to MatGetRow() for matrix in MATSBAIJ format.
684: Not Collective
686: Input Parameters:
687: . mat - the matrix
689: Notes:
690: This routine should be called after you have finished MatGetRow/MatRestoreRow().
693: Level: advanced
695: .seealso: MatGetRowUpperTriangular()
696: @*/
697: PetscErrorCode MatRestoreRowUpperTriangular(Mat mat)
698: {
704: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
705: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
706: MatCheckPreallocated(mat,1);
707: if (!mat->ops->restorerowuppertriangular) return(0);
708: (*mat->ops->restorerowuppertriangular)(mat);
709: return(0);
710: }
712: /*@C
713: MatSetOptionsPrefix - Sets the prefix used for searching for all
714: Mat options in the database.
716: Logically Collective on Mat
718: Input Parameter:
719: + A - the Mat context
720: - prefix - the prefix to prepend to all option names
722: Notes:
723: A hyphen (-) must NOT be given at the beginning of the prefix name.
724: The first character of all runtime options is AUTOMATICALLY the hyphen.
726: Level: advanced
728: .seealso: MatSetFromOptions()
729: @*/
730: PetscErrorCode MatSetOptionsPrefix(Mat A,const char prefix[])
731: {
736: PetscObjectSetOptionsPrefix((PetscObject)A,prefix);
737: return(0);
738: }
740: /*@C
741: MatAppendOptionsPrefix - Appends to the prefix used for searching for all
742: Mat options in the database.
744: Logically Collective on Mat
746: Input Parameters:
747: + A - the Mat context
748: - prefix - the prefix to prepend to all option names
750: Notes:
751: A hyphen (-) must NOT be given at the beginning of the prefix name.
752: The first character of all runtime options is AUTOMATICALLY the hyphen.
754: Level: advanced
756: .seealso: MatGetOptionsPrefix()
757: @*/
758: PetscErrorCode MatAppendOptionsPrefix(Mat A,const char prefix[])
759: {
764: PetscObjectAppendOptionsPrefix((PetscObject)A,prefix);
765: return(0);
766: }
768: /*@C
769: MatGetOptionsPrefix - Sets the prefix used for searching for all
770: Mat options in the database.
772: Not Collective
774: Input Parameter:
775: . A - the Mat context
777: Output Parameter:
778: . prefix - pointer to the prefix string used
780: Notes:
781: On the fortran side, the user should pass in a string 'prefix' of
782: sufficient length to hold the prefix.
784: Level: advanced
786: .seealso: MatAppendOptionsPrefix()
787: @*/
788: PetscErrorCode MatGetOptionsPrefix(Mat A,const char *prefix[])
789: {
794: PetscObjectGetOptionsPrefix((PetscObject)A,prefix);
795: return(0);
796: }
798: /*@
799: MatResetPreallocation - Reset mat to use the original nonzero pattern provided by users.
801: Collective on Mat
803: Input Parameters:
804: . A - the Mat context
806: Notes:
807: The allocated memory will be shrunk after calling MatAssembly with MAT_FINAL_ASSEMBLY. Users can reset the preallocation to access the original memory.
808: Currently support MPIAIJ and SEQAIJ.
810: Level: beginner
812: .seealso: MatSeqAIJSetPreallocation(), MatMPIAIJSetPreallocation(), MatXAIJSetPreallocation()
813: @*/
814: PetscErrorCode MatResetPreallocation(Mat A)
815: {
821: PetscUseMethod(A,"MatResetPreallocation_C",(Mat),(A));
822: return(0);
823: }
826: /*@
827: MatSetUp - Sets up the internal matrix data structures for later use.
829: Collective on Mat
831: Input Parameters:
832: . A - the Mat context
834: Notes:
835: If the user has not set preallocation for this matrix then a default preallocation that is likely to be inefficient is used.
837: If a suitable preallocation routine is used, this function does not need to be called.
839: See the Performance chapter of the PETSc users manual for how to preallocate matrices
841: Level: beginner
843: .seealso: MatCreate(), MatDestroy()
844: @*/
845: PetscErrorCode MatSetUp(Mat A)
846: {
847: PetscMPIInt size;
852: if (!((PetscObject)A)->type_name) {
853: MPI_Comm_size(PetscObjectComm((PetscObject)A), &size);
854: if (size == 1) {
855: MatSetType(A, MATSEQAIJ);
856: } else {
857: MatSetType(A, MATMPIAIJ);
858: }
859: }
860: if (!A->preallocated && A->ops->setup) {
861: PetscInfo(A,"Warning not preallocating matrix storage\n");
862: (*A->ops->setup)(A);
863: }
864: PetscLayoutSetUp(A->rmap);
865: PetscLayoutSetUp(A->cmap);
866: A->preallocated = PETSC_TRUE;
867: return(0);
868: }
870: #if defined(PETSC_HAVE_SAWS)
871: #include <petscviewersaws.h>
872: #endif
873: /*@C
874: MatView - Visualizes a matrix object.
876: Collective on Mat
878: Input Parameters:
879: + mat - the matrix
880: - viewer - visualization context
882: Notes:
883: The available visualization contexts include
884: + PETSC_VIEWER_STDOUT_SELF - for sequential matrices
885: . PETSC_VIEWER_STDOUT_WORLD - for parallel matrices created on PETSC_COMM_WORLD
886: . PETSC_VIEWER_STDOUT_(comm) - for matrices created on MPI communicator comm
887: - PETSC_VIEWER_DRAW_WORLD - graphical display of nonzero structure
889: The user can open alternative visualization contexts with
890: + PetscViewerASCIIOpen() - Outputs matrix to a specified file
891: . PetscViewerBinaryOpen() - Outputs matrix in binary to a
892: specified file; corresponding input uses MatLoad()
893: . PetscViewerDrawOpen() - Outputs nonzero matrix structure to
894: an X window display
895: - PetscViewerSocketOpen() - Outputs matrix to Socket viewer.
896: Currently only the sequential dense and AIJ
897: matrix types support the Socket viewer.
899: The user can call PetscViewerPushFormat() to specify the output
900: format of ASCII printed objects (when using PETSC_VIEWER_STDOUT_SELF,
901: PETSC_VIEWER_STDOUT_WORLD and PetscViewerASCIIOpen). Available formats include
902: + PETSC_VIEWER_DEFAULT - default, prints matrix contents
903: . PETSC_VIEWER_ASCII_MATLAB - prints matrix contents in Matlab format
904: . PETSC_VIEWER_ASCII_DENSE - prints entire matrix including zeros
905: . PETSC_VIEWER_ASCII_COMMON - prints matrix contents, using a sparse
906: format common among all matrix types
907: . PETSC_VIEWER_ASCII_IMPL - prints matrix contents, using an implementation-specific
908: format (which is in many cases the same as the default)
909: . PETSC_VIEWER_ASCII_INFO - prints basic information about the matrix
910: size and structure (not the matrix entries)
911: - PETSC_VIEWER_ASCII_INFO_DETAIL - prints more detailed information about
912: the matrix structure
914: Options Database Keys:
915: + -mat_view ::ascii_info - Prints info on matrix at conclusion of MatAssemblyEnd()
916: . -mat_view ::ascii_info_detail - Prints more detailed info
917: . -mat_view - Prints matrix in ASCII format
918: . -mat_view ::ascii_matlab - Prints matrix in Matlab format
919: . -mat_view draw - PetscDraws nonzero structure of matrix, using MatView() and PetscDrawOpenX().
920: . -display <name> - Sets display name (default is host)
921: . -draw_pause <sec> - Sets number of seconds to pause after display
922: . -mat_view socket - Sends matrix to socket, can be accessed from Matlab (see Users-Manual: Chapter 12 Using MATLAB with PETSc for details)
923: . -viewer_socket_machine <machine> -
924: . -viewer_socket_port <port> -
925: . -mat_view binary - save matrix to file in binary format
926: - -viewer_binary_filename <name> -
927: Level: beginner
929: Notes:
930: The ASCII viewers are only recommended for small matrices on at most a moderate number of processes,
931: the program will seemingly hang and take hours for larger matrices, for larger matrices one should use the binary format.
933: See the manual page for MatLoad() for the exact format of the binary file when the binary
934: viewer is used.
936: See share/petsc/matlab/PetscBinaryRead.m for a Matlab code that can read in the binary file when the binary
937: viewer is used.
939: One can use '-mat_view draw -draw_pause -1' to pause the graphical display of matrix nonzero structure,
940: and then use the following mouse functions.
941: + left mouse: zoom in
942: . middle mouse: zoom out
943: - right mouse: continue with the simulation
945: .seealso: PetscViewerPushFormat(), PetscViewerASCIIOpen(), PetscViewerDrawOpen(),
946: PetscViewerSocketOpen(), PetscViewerBinaryOpen(), MatLoad()
947: @*/
948: PetscErrorCode MatView(Mat mat,PetscViewer viewer)
949: {
950: PetscErrorCode ierr;
951: PetscInt rows,cols,rbs,cbs;
952: PetscBool iascii,ibinary,isstring;
953: PetscViewerFormat format;
954: PetscMPIInt size;
955: #if defined(PETSC_HAVE_SAWS)
956: PetscBool issaws;
957: #endif
962: if (!viewer) {
963: PetscViewerASCIIGetStdout(PetscObjectComm((PetscObject)mat),&viewer);
964: }
967: MatCheckPreallocated(mat,1);
968: PetscViewerGetFormat(viewer,&format);
969: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
970: if (size == 1 && format == PETSC_VIEWER_LOAD_BALANCE) return(0);
971: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&ibinary);
972: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERSTRING,&isstring);
973: if (ibinary) {
974: PetscBool mpiio;
975: PetscViewerBinaryGetUseMPIIO(viewer,&mpiio);
976: if (mpiio) SETERRQ(PetscObjectComm((PetscObject)viewer),PETSC_ERR_SUP,"PETSc matrix viewers do not support using MPI-IO, turn off that flag");
977: }
979: PetscLogEventBegin(MAT_View,mat,viewer,0,0);
980: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
981: if ((!iascii || (format != PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL)) && mat->factortype) {
982: SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"No viewers for factored matrix except ASCII info or info_detailed");
983: }
985: #if defined(PETSC_HAVE_SAWS)
986: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERSAWS,&issaws);
987: #endif
988: if (iascii) {
989: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ORDER,"Must call MatAssemblyBegin/End() before viewing matrix");
990: PetscObjectPrintClassNamePrefixType((PetscObject)mat,viewer);
991: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
992: MatNullSpace nullsp,transnullsp;
994: PetscViewerASCIIPushTab(viewer);
995: MatGetSize(mat,&rows,&cols);
996: MatGetBlockSizes(mat,&rbs,&cbs);
997: if (rbs != 1 || cbs != 1) {
998: if (rbs != cbs) {PetscViewerASCIIPrintf(viewer,"rows=%D, cols=%D, rbs=%D, cbs=%D\n",rows,cols,rbs,cbs);}
999: else {PetscViewerASCIIPrintf(viewer,"rows=%D, cols=%D, bs=%D\n",rows,cols,rbs);}
1000: } else {
1001: PetscViewerASCIIPrintf(viewer,"rows=%D, cols=%D\n",rows,cols);
1002: }
1003: if (mat->factortype) {
1004: MatSolverType solver;
1005: MatFactorGetSolverType(mat,&solver);
1006: PetscViewerASCIIPrintf(viewer,"package used to perform factorization: %s\n",solver);
1007: }
1008: if (mat->ops->getinfo) {
1009: MatInfo info;
1010: MatGetInfo(mat,MAT_GLOBAL_SUM,&info);
1011: PetscViewerASCIIPrintf(viewer,"total: nonzeros=%.f, allocated nonzeros=%.f\n",info.nz_used,info.nz_allocated);
1012: PetscViewerASCIIPrintf(viewer,"total number of mallocs used during MatSetValues calls=%D\n",(PetscInt)info.mallocs);
1013: }
1014: MatGetNullSpace(mat,&nullsp);
1015: MatGetTransposeNullSpace(mat,&transnullsp);
1016: if (nullsp) {PetscViewerASCIIPrintf(viewer," has attached null space\n");}
1017: if (transnullsp && transnullsp != nullsp) {PetscViewerASCIIPrintf(viewer," has attached transposed null space\n");}
1018: MatGetNearNullSpace(mat,&nullsp);
1019: if (nullsp) {PetscViewerASCIIPrintf(viewer," has attached near null space\n");}
1020: }
1021: #if defined(PETSC_HAVE_SAWS)
1022: } else if (issaws) {
1023: PetscMPIInt rank;
1025: PetscObjectName((PetscObject)mat);
1026: MPI_Comm_rank(PETSC_COMM_WORLD,&rank);
1027: if (!((PetscObject)mat)->amsmem && !rank) {
1028: PetscObjectViewSAWs((PetscObject)mat,viewer);
1029: }
1030: #endif
1031: } else if (isstring) {
1032: const char *type;
1033: MatGetType(mat,&type);
1034: PetscViewerStringSPrintf(viewer," MatType: %-7.7s",type);
1035: if (mat->ops->view) {(*mat->ops->view)(mat,viewer);}
1036: }
1037: if ((format == PETSC_VIEWER_NATIVE || format == PETSC_VIEWER_LOAD_BALANCE) && mat->ops->viewnative) {
1038: PetscViewerASCIIPushTab(viewer);
1039: (*mat->ops->viewnative)(mat,viewer);
1040: PetscViewerASCIIPopTab(viewer);
1041: } else if (mat->ops->view) {
1042: PetscViewerASCIIPushTab(viewer);
1043: (*mat->ops->view)(mat,viewer);
1044: PetscViewerASCIIPopTab(viewer);
1045: }
1046: if (iascii) {
1047: PetscViewerGetFormat(viewer,&format);
1048: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
1049: PetscViewerASCIIPopTab(viewer);
1050: }
1051: }
1052: PetscLogEventEnd(MAT_View,mat,viewer,0,0);
1053: return(0);
1054: }
1056: #if defined(PETSC_USE_DEBUG)
1057: #include <../src/sys/totalview/tv_data_display.h>
1058: PETSC_UNUSED static int TV_display_type(const struct _p_Mat *mat)
1059: {
1060: TV_add_row("Local rows", "int", &mat->rmap->n);
1061: TV_add_row("Local columns", "int", &mat->cmap->n);
1062: TV_add_row("Global rows", "int", &mat->rmap->N);
1063: TV_add_row("Global columns", "int", &mat->cmap->N);
1064: TV_add_row("Typename", TV_ascii_string_type, ((PetscObject)mat)->type_name);
1065: return TV_format_OK;
1066: }
1067: #endif
1069: /*@C
1070: MatLoad - Loads a matrix that has been stored in binary/HDF5 format
1071: with MatView(). The matrix format is determined from the options database.
1072: Generates a parallel MPI matrix if the communicator has more than one
1073: processor. The default matrix type is AIJ.
1075: Collective on PetscViewer
1077: Input Parameters:
1078: + newmat - the newly loaded matrix, this needs to have been created with MatCreate()
1079: or some related function before a call to MatLoad()
1080: - viewer - binary/HDF5 file viewer
1082: Options Database Keys:
1083: Used with block matrix formats (MATSEQBAIJ, ...) to specify
1084: block size
1085: . -matload_block_size <bs>
1087: Level: beginner
1089: Notes:
1090: If the Mat type has not yet been given then MATAIJ is used, call MatSetFromOptions() on the
1091: Mat before calling this routine if you wish to set it from the options database.
1093: MatLoad() automatically loads into the options database any options
1094: given in the file filename.info where filename is the name of the file
1095: that was passed to the PetscViewerBinaryOpen(). The options in the info
1096: file will be ignored if you use the -viewer_binary_skip_info option.
1098: If the type or size of newmat is not set before a call to MatLoad, PETSc
1099: sets the default matrix type AIJ and sets the local and global sizes.
1100: If type and/or size is already set, then the same are used.
1102: In parallel, each processor can load a subset of rows (or the
1103: entire matrix). This routine is especially useful when a large
1104: matrix is stored on disk and only part of it is desired on each
1105: processor. For example, a parallel solver may access only some of
1106: the rows from each processor. The algorithm used here reads
1107: relatively small blocks of data rather than reading the entire
1108: matrix and then subsetting it.
1110: Viewer's PetscViewerType must be either PETSCVIEWERBINARY or PETSCVIEWERHDF5.
1111: Such viewer can be created using PetscViewerBinaryOpen()/PetscViewerHDF5Open(),
1112: or the sequence like
1113: $ PetscViewer v;
1114: $ PetscViewerCreate(PETSC_COMM_WORLD,&v);
1115: $ PetscViewerSetType(v,PETSCVIEWERBINARY);
1116: $ PetscViewerSetFromOptions(v);
1117: $ PetscViewerFileSetMode(v,FILE_MODE_READ);
1118: $ PetscViewerFileSetName(v,"datafile");
1119: The optional PetscViewerSetFromOptions() call allows to override PetscViewerSetType() using option
1120: $ -viewer_type {binary,hdf5}
1122: See the example src/ksp/ksp/examples/tutorials/ex27.c with the first approach,
1123: and src/mat/examples/tutorials/ex10.c with the second approach.
1125: Notes about the PETSc binary format:
1126: In case of PETSCVIEWERBINARY, a native PETSc binary format is used. Each of the blocks
1127: is read onto rank 0 and then shipped to its destination rank, one after another.
1128: Multiple objects, both matrices and vectors, can be stored within the same file.
1129: Their PetscObject name is ignored; they are loaded in the order of their storage.
1131: Most users should not need to know the details of the binary storage
1132: format, since MatLoad() and MatView() completely hide these details.
1133: But for anyone who's interested, the standard binary matrix storage
1134: format is
1136: $ PetscInt MAT_FILE_CLASSID
1137: $ PetscInt number of rows
1138: $ PetscInt number of columns
1139: $ PetscInt total number of nonzeros
1140: $ PetscInt *number nonzeros in each row
1141: $ PetscInt *column indices of all nonzeros (starting index is zero)
1142: $ PetscScalar *values of all nonzeros
1144: PETSc automatically does the byte swapping for
1145: machines that store the bytes reversed, e.g. DEC alpha, freebsd,
1146: linux, Windows and the paragon; thus if you write your own binary
1147: read/write routines you have to swap the bytes; see PetscBinaryRead()
1148: and PetscBinaryWrite() to see how this may be done.
1150: Notes about the HDF5 (MATLAB MAT-File Version 7.3) format:
1151: In case of PETSCVIEWERHDF5, a parallel HDF5 reader is used.
1152: Each processor's chunk is loaded independently by its owning rank.
1153: Multiple objects, both matrices and vectors, can be stored within the same file.
1154: They are looked up by their PetscObject name.
1156: As the MATLAB MAT-File Version 7.3 format is also a HDF5 flavor, we decided to use
1157: by default the same structure and naming of the AIJ arrays and column count
1158: within the HDF5 file. This means that a MAT file saved with -v7.3 flag, e.g.
1159: $ save example.mat A b -v7.3
1160: can be directly read by this routine (see Reference 1 for details).
1161: Note that depending on your MATLAB version, this format might be a default,
1162: otherwise you can set it as default in Preferences.
1164: Unless -nocompression flag is used to save the file in MATLAB,
1165: PETSc must be configured with ZLIB package.
1167: See also examples src/mat/examples/tutorials/ex10.c and src/ksp/ksp/examples/tutorials/ex27.c
1169: Current HDF5 (MAT-File) limitations:
1170: This reader currently supports only real MATSEQAIJ, MATMPIAIJ, MATSEQDENSE and MATMPIDENSE matrices.
1172: Corresponding MatView() is not yet implemented.
1174: The loaded matrix is actually a transpose of the original one in MATLAB,
1175: unless you push PETSC_VIEWER_HDF5_MAT format (see examples above).
1176: With this format, matrix is automatically transposed by PETSc,
1177: unless the matrix is marked as SPD or symmetric
1178: (see MatSetOption(), MAT_SPD, MAT_SYMMETRIC).
1180: References:
1181: 1. MATLAB(R) Documentation, manual page of save(), https://www.mathworks.com/help/matlab/ref/save.html#btox10b-1-version
1183: .seealso: PetscViewerBinaryOpen(), PetscViewerSetType(), MatView(), VecLoad()
1185: @*/
1186: PetscErrorCode MatLoad(Mat newmat,PetscViewer viewer)
1187: {
1189: PetscBool flg;
1195: if (!((PetscObject)newmat)->type_name) {
1196: MatSetType(newmat,MATAIJ);
1197: }
1199: flg = PETSC_FALSE;
1200: PetscOptionsGetBool(((PetscObject)newmat)->options,((PetscObject)newmat)->prefix,"-matload_symmetric",&flg,NULL);
1201: if (flg) {
1202: MatSetOption(newmat,MAT_SYMMETRIC,PETSC_TRUE);
1203: MatSetOption(newmat,MAT_SYMMETRY_ETERNAL,PETSC_TRUE);
1204: }
1205: flg = PETSC_FALSE;
1206: PetscOptionsGetBool(((PetscObject)newmat)->options,((PetscObject)newmat)->prefix,"-matload_spd",&flg,NULL);
1207: if (flg) {
1208: MatSetOption(newmat,MAT_SPD,PETSC_TRUE);
1209: }
1211: if (!newmat->ops->load) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"MatLoad is not supported for type %s",((PetscObject)newmat)->type_name);
1212: PetscLogEventBegin(MAT_Load,viewer,0,0,0);
1213: (*newmat->ops->load)(newmat,viewer);
1214: PetscLogEventEnd(MAT_Load,viewer,0,0,0);
1215: return(0);
1216: }
1218: PetscErrorCode MatDestroy_Redundant(Mat_Redundant **redundant)
1219: {
1221: Mat_Redundant *redund = *redundant;
1222: PetscInt i;
1225: if (redund){
1226: if (redund->matseq) { /* via MatCreateSubMatrices() */
1227: ISDestroy(&redund->isrow);
1228: ISDestroy(&redund->iscol);
1229: MatDestroySubMatrices(1,&redund->matseq);
1230: } else {
1231: PetscFree2(redund->send_rank,redund->recv_rank);
1232: PetscFree(redund->sbuf_j);
1233: PetscFree(redund->sbuf_a);
1234: for (i=0; i<redund->nrecvs; i++) {
1235: PetscFree(redund->rbuf_j[i]);
1236: PetscFree(redund->rbuf_a[i]);
1237: }
1238: PetscFree4(redund->sbuf_nz,redund->rbuf_nz,redund->rbuf_j,redund->rbuf_a);
1239: }
1241: if (redund->subcomm) {
1242: PetscCommDestroy(&redund->subcomm);
1243: }
1244: PetscFree(redund);
1245: }
1246: return(0);
1247: }
1249: /*@
1250: MatDestroy - Frees space taken by a matrix.
1252: Collective on Mat
1254: Input Parameter:
1255: . A - the matrix
1257: Level: beginner
1259: @*/
1260: PetscErrorCode MatDestroy(Mat *A)
1261: {
1265: if (!*A) return(0);
1267: if (--((PetscObject)(*A))->refct > 0) {*A = NULL; return(0);}
1269: /* if memory was published with SAWs then destroy it */
1270: PetscObjectSAWsViewOff((PetscObject)*A);
1271: if ((*A)->ops->destroy) {
1272: (*(*A)->ops->destroy)(*A);
1273: }
1275: PetscFree((*A)->defaultvectype);
1276: PetscFree((*A)->bsizes);
1277: PetscFree((*A)->solvertype);
1278: MatDestroy_Redundant(&(*A)->redundant);
1279: MatNullSpaceDestroy(&(*A)->nullsp);
1280: MatNullSpaceDestroy(&(*A)->transnullsp);
1281: MatNullSpaceDestroy(&(*A)->nearnullsp);
1282: MatDestroy(&(*A)->schur);
1283: PetscLayoutDestroy(&(*A)->rmap);
1284: PetscLayoutDestroy(&(*A)->cmap);
1285: PetscHeaderDestroy(A);
1286: return(0);
1287: }
1289: /*@C
1290: MatSetValues - Inserts or adds a block of values into a matrix.
1291: These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
1292: MUST be called after all calls to MatSetValues() have been completed.
1294: Not Collective
1296: Input Parameters:
1297: + mat - the matrix
1298: . v - a logically two-dimensional array of values
1299: . m, idxm - the number of rows and their global indices
1300: . n, idxn - the number of columns and their global indices
1301: - addv - either ADD_VALUES or INSERT_VALUES, where
1302: ADD_VALUES adds values to any existing entries, and
1303: INSERT_VALUES replaces existing entries with new values
1305: Notes:
1306: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call MatXXXXSetPreallocation() or
1307: MatSetUp() before using this routine
1309: By default the values, v, are row-oriented. See MatSetOption() for other options.
1311: Calls to MatSetValues() with the INSERT_VALUES and ADD_VALUES
1312: options cannot be mixed without intervening calls to the assembly
1313: routines.
1315: MatSetValues() uses 0-based row and column numbers in Fortran
1316: as well as in C.
1318: Negative indices may be passed in idxm and idxn, these rows and columns are
1319: simply ignored. This allows easily inserting element stiffness matrices
1320: with homogeneous Dirchlet boundary conditions that you don't want represented
1321: in the matrix.
1323: Efficiency Alert:
1324: The routine MatSetValuesBlocked() may offer much better efficiency
1325: for users of block sparse formats (MATSEQBAIJ and MATMPIBAIJ).
1327: Level: beginner
1329: Developer Notes:
1330: This is labeled with C so does not automatically generate Fortran stubs and interfaces
1331: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
1333: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
1334: InsertMode, INSERT_VALUES, ADD_VALUES
1335: @*/
1336: PetscErrorCode MatSetValues(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],const PetscScalar v[],InsertMode addv)
1337: {
1339: #if defined(PETSC_USE_DEBUG)
1340: PetscInt i,j;
1341: #endif
1346: if (!m || !n) return(0); /* no values to insert */
1349: MatCheckPreallocated(mat,1);
1351: if (mat->insertmode == NOT_SET_VALUES) {
1352: mat->insertmode = addv;
1353: }
1354: #if defined(PETSC_USE_DEBUG)
1355: else if (mat->insertmode != addv) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
1356: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1357: if (!mat->ops->setvalues) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1359: for (i=0; i<m; i++) {
1360: for (j=0; j<n; j++) {
1361: if (mat->erroriffailure && PetscIsInfOrNanScalar(v[i*n+j]))
1362: #if defined(PETSC_USE_COMPLEX)
1363: 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]);
1364: #else
1365: SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_FP,"Inserting %g at matrix entry (%D,%D)",(double)v[i*n+j],idxm[i],idxn[j]);
1366: #endif
1367: }
1368: }
1369: #endif
1371: if (mat->assembled) {
1372: mat->was_assembled = PETSC_TRUE;
1373: mat->assembled = PETSC_FALSE;
1374: }
1375: PetscLogEventBegin(MAT_SetValues,mat,0,0,0);
1376: (*mat->ops->setvalues)(mat,m,idxm,n,idxn,v,addv);
1377: PetscLogEventEnd(MAT_SetValues,mat,0,0,0);
1378: return(0);
1379: }
1382: /*@
1383: MatSetValuesRowLocal - Inserts a row (block row for BAIJ matrices) of nonzero
1384: values into a matrix
1386: Not Collective
1388: Input Parameters:
1389: + mat - the matrix
1390: . row - the (block) row to set
1391: - v - a logically two-dimensional array of values
1393: Notes:
1394: By the values, v, are column-oriented (for the block version) and sorted
1396: All the nonzeros in the row must be provided
1398: The matrix must have previously had its column indices set
1400: The row must belong to this process
1402: Level: intermediate
1404: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
1405: InsertMode, INSERT_VALUES, ADD_VALUES, MatSetValues(), MatSetValuesRow(), MatSetLocalToGlobalMapping()
1406: @*/
1407: PetscErrorCode MatSetValuesRowLocal(Mat mat,PetscInt row,const PetscScalar v[])
1408: {
1410: PetscInt globalrow;
1416: ISLocalToGlobalMappingApply(mat->rmap->mapping,1,&row,&globalrow);
1417: MatSetValuesRow(mat,globalrow,v);
1418: return(0);
1419: }
1421: /*@
1422: MatSetValuesRow - Inserts a row (block row for BAIJ matrices) of nonzero
1423: values into a matrix
1425: Not Collective
1427: Input Parameters:
1428: + mat - the matrix
1429: . row - the (block) row to set
1430: - 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
1432: Notes:
1433: The values, v, are column-oriented for the block version.
1435: All the nonzeros in the row must be provided
1437: THE MATRIX MUST HAVE PREVIOUSLY HAD ITS COLUMN INDICES SET. IT IS RARE THAT THIS ROUTINE IS USED, usually MatSetValues() is used.
1439: The row must belong to this process
1441: Level: advanced
1443: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
1444: InsertMode, INSERT_VALUES, ADD_VALUES, MatSetValues()
1445: @*/
1446: PetscErrorCode MatSetValuesRow(Mat mat,PetscInt row,const PetscScalar v[])
1447: {
1453: MatCheckPreallocated(mat,1);
1455: #if defined(PETSC_USE_DEBUG)
1456: if (mat->insertmode == ADD_VALUES) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add and insert values");
1457: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1458: #endif
1459: mat->insertmode = INSERT_VALUES;
1461: if (mat->assembled) {
1462: mat->was_assembled = PETSC_TRUE;
1463: mat->assembled = PETSC_FALSE;
1464: }
1465: PetscLogEventBegin(MAT_SetValues,mat,0,0,0);
1466: if (!mat->ops->setvaluesrow) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1467: (*mat->ops->setvaluesrow)(mat,row,v);
1468: PetscLogEventEnd(MAT_SetValues,mat,0,0,0);
1469: return(0);
1470: }
1472: /*@
1473: MatSetValuesStencil - Inserts or adds a block of values into a matrix.
1474: Using structured grid indexing
1476: Not Collective
1478: Input Parameters:
1479: + mat - the matrix
1480: . m - number of rows being entered
1481: . idxm - grid coordinates (and component number when dof > 1) for matrix rows being entered
1482: . n - number of columns being entered
1483: . idxn - grid coordinates (and component number when dof > 1) for matrix columns being entered
1484: . v - a logically two-dimensional array of values
1485: - addv - either ADD_VALUES or INSERT_VALUES, where
1486: ADD_VALUES adds values to any existing entries, and
1487: INSERT_VALUES replaces existing entries with new values
1489: Notes:
1490: By default the values, v, are row-oriented. See MatSetOption() for other options.
1492: Calls to MatSetValuesStencil() with the INSERT_VALUES and ADD_VALUES
1493: options cannot be mixed without intervening calls to the assembly
1494: routines.
1496: The grid coordinates are across the entire grid, not just the local portion
1498: MatSetValuesStencil() uses 0-based row and column numbers in Fortran
1499: as well as in C.
1501: For setting/accessing vector values via array coordinates you can use the DMDAVecGetArray() routine
1503: In order to use this routine you must either obtain the matrix with DMCreateMatrix()
1504: or call MatSetLocalToGlobalMapping() and MatSetStencil() first.
1506: The columns and rows in the stencil passed in MUST be contained within the
1507: ghost region of the given process as set with DMDACreateXXX() or MatSetStencil(). For example,
1508: if you create a DMDA with an overlap of one grid level and on a particular process its first
1509: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1510: first i index you can use in your column and row indices in MatSetStencil() is 5.
1512: In Fortran idxm and idxn should be declared as
1513: $ MatStencil idxm(4,m),idxn(4,n)
1514: and the values inserted using
1515: $ idxm(MatStencil_i,1) = i
1516: $ idxm(MatStencil_j,1) = j
1517: $ idxm(MatStencil_k,1) = k
1518: $ idxm(MatStencil_c,1) = c
1519: etc
1521: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
1522: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
1523: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
1524: DM_BOUNDARY_PERIODIC boundary type.
1526: 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
1527: a single value per point) you can skip filling those indices.
1529: Inspired by the structured grid interface to the HYPRE package
1530: (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)
1532: Efficiency Alert:
1533: The routine MatSetValuesBlockedStencil() may offer much better efficiency
1534: for users of block sparse formats (MATSEQBAIJ and MATMPIBAIJ).
1536: Level: beginner
1538: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
1539: MatSetValues(), MatSetValuesBlockedStencil(), MatSetStencil(), DMCreateMatrix(), DMDAVecGetArray(), MatStencil
1540: @*/
1541: PetscErrorCode MatSetValuesStencil(Mat mat,PetscInt m,const MatStencil idxm[],PetscInt n,const MatStencil idxn[],const PetscScalar v[],InsertMode addv)
1542: {
1544: PetscInt buf[8192],*bufm=0,*bufn=0,*jdxm,*jdxn;
1545: PetscInt j,i,dim = mat->stencil.dim,*dims = mat->stencil.dims+1,tmp;
1546: PetscInt *starts = mat->stencil.starts,*dxm = (PetscInt*)idxm,*dxn = (PetscInt*)idxn,sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1549: if (!m || !n) return(0); /* no values to insert */
1555: if ((m+n) <= (PetscInt)(sizeof(buf)/sizeof(PetscInt))) {
1556: jdxm = buf; jdxn = buf+m;
1557: } else {
1558: PetscMalloc2(m,&bufm,n,&bufn);
1559: jdxm = bufm; jdxn = bufn;
1560: }
1561: for (i=0; i<m; i++) {
1562: for (j=0; j<3-sdim; j++) dxm++;
1563: tmp = *dxm++ - starts[0];
1564: for (j=0; j<dim-1; j++) {
1565: if ((*dxm++ - starts[j+1]) < 0 || tmp < 0) tmp = -1;
1566: else tmp = tmp*dims[j] + *(dxm-1) - starts[j+1];
1567: }
1568: if (mat->stencil.noc) dxm++;
1569: jdxm[i] = tmp;
1570: }
1571: for (i=0; i<n; i++) {
1572: for (j=0; j<3-sdim; j++) dxn++;
1573: tmp = *dxn++ - starts[0];
1574: for (j=0; j<dim-1; j++) {
1575: if ((*dxn++ - starts[j+1]) < 0 || tmp < 0) tmp = -1;
1576: else tmp = tmp*dims[j] + *(dxn-1) - starts[j+1];
1577: }
1578: if (mat->stencil.noc) dxn++;
1579: jdxn[i] = tmp;
1580: }
1581: MatSetValuesLocal(mat,m,jdxm,n,jdxn,v,addv);
1582: PetscFree2(bufm,bufn);
1583: return(0);
1584: }
1586: /*@
1587: MatSetValuesBlockedStencil - Inserts or adds a block of values into a matrix.
1588: Using structured grid indexing
1590: Not Collective
1592: Input Parameters:
1593: + mat - the matrix
1594: . m - number of rows being entered
1595: . idxm - grid coordinates for matrix rows being entered
1596: . n - number of columns being entered
1597: . idxn - grid coordinates for matrix columns being entered
1598: . v - a logically two-dimensional array of values
1599: - addv - either ADD_VALUES or INSERT_VALUES, where
1600: ADD_VALUES adds values to any existing entries, and
1601: INSERT_VALUES replaces existing entries with new values
1603: Notes:
1604: By default the values, v, are row-oriented and unsorted.
1605: See MatSetOption() for other options.
1607: Calls to MatSetValuesBlockedStencil() with the INSERT_VALUES and ADD_VALUES
1608: options cannot be mixed without intervening calls to the assembly
1609: routines.
1611: The grid coordinates are across the entire grid, not just the local portion
1613: MatSetValuesBlockedStencil() uses 0-based row and column numbers in Fortran
1614: as well as in C.
1616: For setting/accessing vector values via array coordinates you can use the DMDAVecGetArray() routine
1618: In order to use this routine you must either obtain the matrix with DMCreateMatrix()
1619: or call MatSetBlockSize(), MatSetLocalToGlobalMapping() and MatSetStencil() first.
1621: The columns and rows in the stencil passed in MUST be contained within the
1622: ghost region of the given process as set with DMDACreateXXX() or MatSetStencil(). For example,
1623: if you create a DMDA with an overlap of one grid level and on a particular process its first
1624: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1625: first i index you can use in your column and row indices in MatSetStencil() is 5.
1627: In Fortran idxm and idxn should be declared as
1628: $ MatStencil idxm(4,m),idxn(4,n)
1629: and the values inserted using
1630: $ idxm(MatStencil_i,1) = i
1631: $ idxm(MatStencil_j,1) = j
1632: $ idxm(MatStencil_k,1) = k
1633: etc
1635: Negative indices may be passed in idxm and idxn, these rows and columns are
1636: simply ignored. This allows easily inserting element stiffness matrices
1637: with homogeneous Dirchlet boundary conditions that you don't want represented
1638: in the matrix.
1640: Inspired by the structured grid interface to the HYPRE package
1641: (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)
1643: Level: beginner
1645: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
1646: MatSetValues(), MatSetValuesStencil(), MatSetStencil(), DMCreateMatrix(), DMDAVecGetArray(), MatStencil,
1647: MatSetBlockSize(), MatSetLocalToGlobalMapping()
1648: @*/
1649: PetscErrorCode MatSetValuesBlockedStencil(Mat mat,PetscInt m,const MatStencil idxm[],PetscInt n,const MatStencil idxn[],const PetscScalar v[],InsertMode addv)
1650: {
1652: PetscInt buf[8192],*bufm=0,*bufn=0,*jdxm,*jdxn;
1653: PetscInt j,i,dim = mat->stencil.dim,*dims = mat->stencil.dims+1,tmp;
1654: PetscInt *starts = mat->stencil.starts,*dxm = (PetscInt*)idxm,*dxn = (PetscInt*)idxn,sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1657: if (!m || !n) return(0); /* no values to insert */
1664: if ((m+n) <= (PetscInt)(sizeof(buf)/sizeof(PetscInt))) {
1665: jdxm = buf; jdxn = buf+m;
1666: } else {
1667: PetscMalloc2(m,&bufm,n,&bufn);
1668: jdxm = bufm; jdxn = bufn;
1669: }
1670: for (i=0; i<m; i++) {
1671: for (j=0; j<3-sdim; j++) dxm++;
1672: tmp = *dxm++ - starts[0];
1673: for (j=0; j<sdim-1; j++) {
1674: if ((*dxm++ - starts[j+1]) < 0 || tmp < 0) tmp = -1;
1675: else tmp = tmp*dims[j] + *(dxm-1) - starts[j+1];
1676: }
1677: dxm++;
1678: jdxm[i] = tmp;
1679: }
1680: for (i=0; i<n; i++) {
1681: for (j=0; j<3-sdim; j++) dxn++;
1682: tmp = *dxn++ - starts[0];
1683: for (j=0; j<sdim-1; j++) {
1684: if ((*dxn++ - starts[j+1]) < 0 || tmp < 0) tmp = -1;
1685: else tmp = tmp*dims[j] + *(dxn-1) - starts[j+1];
1686: }
1687: dxn++;
1688: jdxn[i] = tmp;
1689: }
1690: MatSetValuesBlockedLocal(mat,m,jdxm,n,jdxn,v,addv);
1691: PetscFree2(bufm,bufn);
1692: return(0);
1693: }
1695: /*@
1696: MatSetStencil - Sets the grid information for setting values into a matrix via
1697: MatSetValuesStencil()
1699: Not Collective
1701: Input Parameters:
1702: + mat - the matrix
1703: . dim - dimension of the grid 1, 2, or 3
1704: . dims - number of grid points in x, y, and z direction, including ghost points on your processor
1705: . starts - starting point of ghost nodes on your processor in x, y, and z direction
1706: - dof - number of degrees of freedom per node
1709: Inspired by the structured grid interface to the HYPRE package
1710: (www.llnl.gov/CASC/hyper)
1712: For matrices generated with DMCreateMatrix() this routine is automatically called and so not needed by the
1713: user.
1715: Level: beginner
1717: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
1718: MatSetValues(), MatSetValuesBlockedStencil(), MatSetValuesStencil()
1719: @*/
1720: PetscErrorCode MatSetStencil(Mat mat,PetscInt dim,const PetscInt dims[],const PetscInt starts[],PetscInt dof)
1721: {
1722: PetscInt i;
1729: mat->stencil.dim = dim + (dof > 1);
1730: for (i=0; i<dim; i++) {
1731: mat->stencil.dims[i] = dims[dim-i-1]; /* copy the values in backwards */
1732: mat->stencil.starts[i] = starts[dim-i-1];
1733: }
1734: mat->stencil.dims[dim] = dof;
1735: mat->stencil.starts[dim] = 0;
1736: mat->stencil.noc = (PetscBool)(dof == 1);
1737: return(0);
1738: }
1740: /*@C
1741: MatSetValuesBlocked - Inserts or adds a block of values into a matrix.
1743: Not Collective
1745: Input Parameters:
1746: + mat - the matrix
1747: . v - a logically two-dimensional array of values
1748: . m, idxm - the number of block rows and their global block indices
1749: . n, idxn - the number of block columns and their global block indices
1750: - addv - either ADD_VALUES or INSERT_VALUES, where
1751: ADD_VALUES adds values to any existing entries, and
1752: INSERT_VALUES replaces existing entries with new values
1754: Notes:
1755: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call
1756: MatXXXXSetPreallocation() or MatSetUp() before using this routine.
1758: The m and n count the NUMBER of blocks in the row direction and column direction,
1759: NOT the total number of rows/columns; for example, if the block size is 2 and
1760: you are passing in values for rows 2,3,4,5 then m would be 2 (not 4).
1761: The values in idxm would be 1 2; that is the first index for each block divided by
1762: the block size.
1764: Note that you must call MatSetBlockSize() when constructing this matrix (before
1765: preallocating it).
1767: By default the values, v, are row-oriented, so the layout of
1768: v is the same as for MatSetValues(). See MatSetOption() for other options.
1770: Calls to MatSetValuesBlocked() with the INSERT_VALUES and ADD_VALUES
1771: options cannot be mixed without intervening calls to the assembly
1772: routines.
1774: MatSetValuesBlocked() uses 0-based row and column numbers in Fortran
1775: as well as in C.
1777: Negative indices may be passed in idxm and idxn, these rows and columns are
1778: simply ignored. This allows easily inserting element stiffness matrices
1779: with homogeneous Dirchlet boundary conditions that you don't want represented
1780: in the matrix.
1782: Each time an entry is set within a sparse matrix via MatSetValues(),
1783: internal searching must be done to determine where to place the
1784: data in the matrix storage space. By instead inserting blocks of
1785: entries via MatSetValuesBlocked(), the overhead of matrix assembly is
1786: reduced.
1788: Example:
1789: $ Suppose m=n=2 and block size(bs) = 2 The array is
1790: $
1791: $ 1 2 | 3 4
1792: $ 5 6 | 7 8
1793: $ - - - | - - -
1794: $ 9 10 | 11 12
1795: $ 13 14 | 15 16
1796: $
1797: $ v[] should be passed in like
1798: $ v[] = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]
1799: $
1800: $ If you are not using row oriented storage of v (that is you called MatSetOption(mat,MAT_ROW_ORIENTED,PETSC_FALSE)) then
1801: $ v[] = [1,5,9,13,2,6,10,14,3,7,11,15,4,8,12,16]
1803: Level: intermediate
1805: .seealso: MatSetBlockSize(), MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesBlockedLocal()
1806: @*/
1807: PetscErrorCode MatSetValuesBlocked(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],const PetscScalar v[],InsertMode addv)
1808: {
1814: if (!m || !n) return(0); /* no values to insert */
1818: MatCheckPreallocated(mat,1);
1819: if (mat->insertmode == NOT_SET_VALUES) {
1820: mat->insertmode = addv;
1821: }
1822: #if defined(PETSC_USE_DEBUG)
1823: else if (mat->insertmode != addv) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
1824: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1825: if (!mat->ops->setvaluesblocked && !mat->ops->setvalues) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1826: #endif
1828: if (mat->assembled) {
1829: mat->was_assembled = PETSC_TRUE;
1830: mat->assembled = PETSC_FALSE;
1831: }
1832: PetscLogEventBegin(MAT_SetValues,mat,0,0,0);
1833: if (mat->ops->setvaluesblocked) {
1834: (*mat->ops->setvaluesblocked)(mat,m,idxm,n,idxn,v,addv);
1835: } else {
1836: PetscInt buf[8192],*bufr=0,*bufc=0,*iidxm,*iidxn;
1837: PetscInt i,j,bs,cbs;
1838: MatGetBlockSizes(mat,&bs,&cbs);
1839: if (m*bs+n*cbs <= (PetscInt)(sizeof(buf)/sizeof(PetscInt))) {
1840: iidxm = buf; iidxn = buf + m*bs;
1841: } else {
1842: PetscMalloc2(m*bs,&bufr,n*cbs,&bufc);
1843: iidxm = bufr; iidxn = bufc;
1844: }
1845: for (i=0; i<m; i++) {
1846: for (j=0; j<bs; j++) {
1847: iidxm[i*bs+j] = bs*idxm[i] + j;
1848: }
1849: }
1850: for (i=0; i<n; i++) {
1851: for (j=0; j<cbs; j++) {
1852: iidxn[i*cbs+j] = cbs*idxn[i] + j;
1853: }
1854: }
1855: MatSetValues(mat,m*bs,iidxm,n*cbs,iidxn,v,addv);
1856: PetscFree2(bufr,bufc);
1857: }
1858: PetscLogEventEnd(MAT_SetValues,mat,0,0,0);
1859: return(0);
1860: }
1862: /*@
1863: MatGetValues - Gets a block of values from a matrix.
1865: Not Collective; currently only returns a local block
1867: Input Parameters:
1868: + mat - the matrix
1869: . v - a logically two-dimensional array for storing the values
1870: . m, idxm - the number of rows and their global indices
1871: - n, idxn - the number of columns and their global indices
1873: Notes:
1874: The user must allocate space (m*n PetscScalars) for the values, v.
1875: The values, v, are then returned in a row-oriented format,
1876: analogous to that used by default in MatSetValues().
1878: MatGetValues() uses 0-based row and column numbers in
1879: Fortran as well as in C.
1881: MatGetValues() requires that the matrix has been assembled
1882: with MatAssemblyBegin()/MatAssemblyEnd(). Thus, calls to
1883: MatSetValues() and MatGetValues() CANNOT be made in succession
1884: without intermediate matrix assembly.
1886: Negative row or column indices will be ignored and those locations in v[] will be
1887: left unchanged.
1889: Level: advanced
1891: .seealso: MatGetRow(), MatCreateSubMatrices(), MatSetValues()
1892: @*/
1893: PetscErrorCode MatGetValues(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],PetscScalar v[])
1894: {
1900: if (!m || !n) return(0);
1904: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1905: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1906: if (!mat->ops->getvalues) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1907: MatCheckPreallocated(mat,1);
1909: PetscLogEventBegin(MAT_GetValues,mat,0,0,0);
1910: (*mat->ops->getvalues)(mat,m,idxm,n,idxn,v);
1911: PetscLogEventEnd(MAT_GetValues,mat,0,0,0);
1912: return(0);
1913: }
1915: /*@
1916: MatSetValuesBatch - Adds (ADD_VALUES) many blocks of values into a matrix at once. The blocks must all be square and
1917: the same size. Currently, this can only be called once and creates the given matrix.
1919: Not Collective
1921: Input Parameters:
1922: + mat - the matrix
1923: . nb - the number of blocks
1924: . bs - the number of rows (and columns) in each block
1925: . rows - a concatenation of the rows for each block
1926: - v - a concatenation of logically two-dimensional arrays of values
1928: Notes:
1929: In the future, we will extend this routine to handle rectangular blocks, and to allow multiple calls for a given matrix.
1931: Level: advanced
1933: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
1934: InsertMode, INSERT_VALUES, ADD_VALUES, MatSetValues()
1935: @*/
1936: PetscErrorCode MatSetValuesBatch(Mat mat, PetscInt nb, PetscInt bs, PetscInt rows[], const PetscScalar v[])
1937: {
1945: #if defined(PETSC_USE_DEBUG)
1946: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1947: #endif
1949: PetscLogEventBegin(MAT_SetValuesBatch,mat,0,0,0);
1950: if (mat->ops->setvaluesbatch) {
1951: (*mat->ops->setvaluesbatch)(mat,nb,bs,rows,v);
1952: } else {
1953: PetscInt b;
1954: for (b = 0; b < nb; ++b) {
1955: MatSetValues(mat, bs, &rows[b*bs], bs, &rows[b*bs], &v[b*bs*bs], ADD_VALUES);
1956: }
1957: }
1958: PetscLogEventEnd(MAT_SetValuesBatch,mat,0,0,0);
1959: return(0);
1960: }
1962: /*@
1963: MatSetLocalToGlobalMapping - Sets a local-to-global numbering for use by
1964: the routine MatSetValuesLocal() to allow users to insert matrix entries
1965: using a local (per-processor) numbering.
1967: Not Collective
1969: Input Parameters:
1970: + x - the matrix
1971: . rmapping - row mapping created with ISLocalToGlobalMappingCreate() or ISLocalToGlobalMappingCreateIS()
1972: - cmapping - column mapping
1974: Level: intermediate
1977: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesLocal()
1978: @*/
1979: PetscErrorCode MatSetLocalToGlobalMapping(Mat x,ISLocalToGlobalMapping rmapping,ISLocalToGlobalMapping cmapping)
1980: {
1989: if (x->ops->setlocaltoglobalmapping) {
1990: (*x->ops->setlocaltoglobalmapping)(x,rmapping,cmapping);
1991: } else {
1992: PetscLayoutSetISLocalToGlobalMapping(x->rmap,rmapping);
1993: PetscLayoutSetISLocalToGlobalMapping(x->cmap,cmapping);
1994: }
1995: return(0);
1996: }
1999: /*@
2000: MatGetLocalToGlobalMapping - Gets the local-to-global numbering set by MatSetLocalToGlobalMapping()
2002: Not Collective
2004: Input Parameters:
2005: . A - the matrix
2007: Output Parameters:
2008: + rmapping - row mapping
2009: - cmapping - column mapping
2011: Level: advanced
2014: .seealso: MatSetValuesLocal()
2015: @*/
2016: PetscErrorCode MatGetLocalToGlobalMapping(Mat A,ISLocalToGlobalMapping *rmapping,ISLocalToGlobalMapping *cmapping)
2017: {
2023: if (rmapping) *rmapping = A->rmap->mapping;
2024: if (cmapping) *cmapping = A->cmap->mapping;
2025: return(0);
2026: }
2028: /*@
2029: MatGetLayouts - Gets the PetscLayout objects for rows and columns
2031: Not Collective
2033: Input Parameters:
2034: . A - the matrix
2036: Output Parameters:
2037: + rmap - row layout
2038: - cmap - column layout
2040: Level: advanced
2042: .seealso: MatCreateVecs(), MatGetLocalToGlobalMapping()
2043: @*/
2044: PetscErrorCode MatGetLayouts(Mat A,PetscLayout *rmap,PetscLayout *cmap)
2045: {
2051: if (rmap) *rmap = A->rmap;
2052: if (cmap) *cmap = A->cmap;
2053: return(0);
2054: }
2056: /*@C
2057: MatSetValuesLocal - Inserts or adds values into certain locations of a matrix,
2058: using a local ordering of the nodes.
2060: Not Collective
2062: Input Parameters:
2063: + mat - the matrix
2064: . nrow, irow - number of rows and their local indices
2065: . ncol, icol - number of columns and their local indices
2066: . y - a logically two-dimensional array of values
2067: - addv - either INSERT_VALUES or ADD_VALUES, where
2068: ADD_VALUES adds values to any existing entries, and
2069: INSERT_VALUES replaces existing entries with new values
2071: Notes:
2072: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call MatXXXXSetPreallocation() or
2073: MatSetUp() before using this routine
2075: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call MatSetLocalToGlobalMapping() before using this routine
2077: Calls to MatSetValuesLocal() with the INSERT_VALUES and ADD_VALUES
2078: options cannot be mixed without intervening calls to the assembly
2079: routines.
2081: These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
2082: MUST be called after all calls to MatSetValuesLocal() have been completed.
2084: Level: intermediate
2086: Developer Notes:
2087: This is labeled with C so does not automatically generate Fortran stubs and interfaces
2088: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2090: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetLocalToGlobalMapping(),
2091: MatSetValueLocal()
2092: @*/
2093: PetscErrorCode MatSetValuesLocal(Mat mat,PetscInt nrow,const PetscInt irow[],PetscInt ncol,const PetscInt icol[],const PetscScalar y[],InsertMode addv)
2094: {
2100: MatCheckPreallocated(mat,1);
2101: if (!nrow || !ncol) return(0); /* no values to insert */
2104: if (mat->insertmode == NOT_SET_VALUES) {
2105: mat->insertmode = addv;
2106: }
2107: #if defined(PETSC_USE_DEBUG)
2108: else if (mat->insertmode != addv) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
2109: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2110: if (!mat->ops->setvalueslocal && !mat->ops->setvalues) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2111: #endif
2113: if (mat->assembled) {
2114: mat->was_assembled = PETSC_TRUE;
2115: mat->assembled = PETSC_FALSE;
2116: }
2117: PetscLogEventBegin(MAT_SetValues,mat,0,0,0);
2118: if (mat->ops->setvalueslocal) {
2119: (*mat->ops->setvalueslocal)(mat,nrow,irow,ncol,icol,y,addv);
2120: } else {
2121: PetscInt buf[8192],*bufr=0,*bufc=0,*irowm,*icolm;
2122: if ((nrow+ncol) <= (PetscInt)(sizeof(buf)/sizeof(PetscInt))) {
2123: irowm = buf; icolm = buf+nrow;
2124: } else {
2125: PetscMalloc2(nrow,&bufr,ncol,&bufc);
2126: irowm = bufr; icolm = bufc;
2127: }
2128: if (!mat->rmap->mapping) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"MatSetValuesLocal() cannot proceed without local-to-global row mapping (See MatSetLocalToGlobalMapping()).");
2129: if (!mat->cmap->mapping) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"MatSetValuesLocal() cannot proceed without local-to-global column mapping (See MatSetLocalToGlobalMapping()).");
2130: ISLocalToGlobalMappingApply(mat->rmap->mapping,nrow,irow,irowm);
2131: ISLocalToGlobalMappingApply(mat->cmap->mapping,ncol,icol,icolm);
2132: MatSetValues(mat,nrow,irowm,ncol,icolm,y,addv);
2133: PetscFree2(bufr,bufc);
2134: }
2135: PetscLogEventEnd(MAT_SetValues,mat,0,0,0);
2136: return(0);
2137: }
2139: /*@C
2140: MatSetValuesBlockedLocal - Inserts or adds values into certain locations of a matrix,
2141: using a local ordering of the nodes a block at a time.
2143: Not Collective
2145: Input Parameters:
2146: + x - the matrix
2147: . nrow, irow - number of rows and their local indices
2148: . ncol, icol - number of columns and their local indices
2149: . y - a logically two-dimensional array of values
2150: - addv - either INSERT_VALUES or ADD_VALUES, where
2151: ADD_VALUES adds values to any existing entries, and
2152: INSERT_VALUES replaces existing entries with new values
2154: Notes:
2155: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call MatXXXXSetPreallocation() or
2156: MatSetUp() before using this routine
2158: If you create the matrix yourself (that is not with a call to DMCreateMatrix()) then you MUST call MatSetBlockSize() and MatSetLocalToGlobalMapping()
2159: before using this routineBefore calling MatSetValuesLocal(), the user must first set the
2161: Calls to MatSetValuesBlockedLocal() with the INSERT_VALUES and ADD_VALUES
2162: options cannot be mixed without intervening calls to the assembly
2163: routines.
2165: These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
2166: MUST be called after all calls to MatSetValuesBlockedLocal() have been completed.
2168: Level: intermediate
2170: Developer Notes:
2171: This is labeled with C so does not automatically generate Fortran stubs and interfaces
2172: because it requires multiple Fortran interfaces depending on which arguments are scalar or arrays.
2174: .seealso: MatSetBlockSize(), MatSetLocalToGlobalMapping(), MatAssemblyBegin(), MatAssemblyEnd(),
2175: MatSetValuesLocal(), MatSetValuesBlocked()
2176: @*/
2177: PetscErrorCode MatSetValuesBlockedLocal(Mat mat,PetscInt nrow,const PetscInt irow[],PetscInt ncol,const PetscInt icol[],const PetscScalar y[],InsertMode addv)
2178: {
2184: MatCheckPreallocated(mat,1);
2185: if (!nrow || !ncol) return(0); /* no values to insert */
2189: if (mat->insertmode == NOT_SET_VALUES) {
2190: mat->insertmode = addv;
2191: }
2192: #if defined(PETSC_USE_DEBUG)
2193: else if (mat->insertmode != addv) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
2194: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2195: 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);
2196: #endif
2198: if (mat->assembled) {
2199: mat->was_assembled = PETSC_TRUE;
2200: mat->assembled = PETSC_FALSE;
2201: }
2202: #if defined(PETSC_USE_DEBUG)
2203: /* Condition on the mapping existing, because MatSetValuesBlockedLocal_IS does not require it to be set. */
2204: if (mat->rmap->mapping) {
2205: PetscInt irbs, rbs;
2206: MatGetBlockSizes(mat, &rbs, NULL);
2207: ISLocalToGlobalMappingGetBlockSize(mat->rmap->mapping,&irbs);
2208: if (rbs != irbs) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Different row block sizes! mat %D, row l2g map %D",rbs,irbs);
2209: }
2210: if (mat->cmap->mapping) {
2211: PetscInt icbs, cbs;
2212: MatGetBlockSizes(mat,NULL,&cbs);
2213: ISLocalToGlobalMappingGetBlockSize(mat->cmap->mapping,&icbs);
2214: if (cbs != icbs) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Different col block sizes! mat %D, col l2g map %D",cbs,icbs);
2215: }
2216: #endif
2217: PetscLogEventBegin(MAT_SetValues,mat,0,0,0);
2218: if (mat->ops->setvaluesblockedlocal) {
2219: (*mat->ops->setvaluesblockedlocal)(mat,nrow,irow,ncol,icol,y,addv);
2220: } else {
2221: PetscInt buf[8192],*bufr=0,*bufc=0,*irowm,*icolm;
2222: if ((nrow+ncol) <= (PetscInt)(sizeof(buf)/sizeof(PetscInt))) {
2223: irowm = buf; icolm = buf + nrow;
2224: } else {
2225: PetscMalloc2(nrow,&bufr,ncol,&bufc);
2226: irowm = bufr; icolm = bufc;
2227: }
2228: ISLocalToGlobalMappingApplyBlock(mat->rmap->mapping,nrow,irow,irowm);
2229: ISLocalToGlobalMappingApplyBlock(mat->cmap->mapping,ncol,icol,icolm);
2230: MatSetValuesBlocked(mat,nrow,irowm,ncol,icolm,y,addv);
2231: PetscFree2(bufr,bufc);
2232: }
2233: PetscLogEventEnd(MAT_SetValues,mat,0,0,0);
2234: return(0);
2235: }
2237: /*@
2238: MatMultDiagonalBlock - Computes the matrix-vector product, y = Dx. Where D is defined by the inode or block structure of the diagonal
2240: Collective on Mat
2242: Input Parameters:
2243: + mat - the matrix
2244: - x - the vector to be multiplied
2246: Output Parameters:
2247: . y - the result
2249: Notes:
2250: The vectors x and y cannot be the same. I.e., one cannot
2251: call MatMult(A,y,y).
2253: Level: developer
2255: .seealso: MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
2256: @*/
2257: PetscErrorCode MatMultDiagonalBlock(Mat mat,Vec x,Vec y)
2258: {
2267: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2268: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2269: if (x == y) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2270: MatCheckPreallocated(mat,1);
2272: if (!mat->ops->multdiagonalblock) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s does not have a multiply defined",((PetscObject)mat)->type_name);
2273: (*mat->ops->multdiagonalblock)(mat,x,y);
2274: PetscObjectStateIncrease((PetscObject)y);
2275: return(0);
2276: }
2278: /* --------------------------------------------------------*/
2279: /*@
2280: MatMult - Computes the matrix-vector product, y = Ax.
2282: Neighbor-wise Collective on Mat
2284: Input Parameters:
2285: + mat - the matrix
2286: - x - the vector to be multiplied
2288: Output Parameters:
2289: . y - the result
2291: Notes:
2292: The vectors x and y cannot be the same. I.e., one cannot
2293: call MatMult(A,y,y).
2295: Level: beginner
2297: .seealso: MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
2298: @*/
2299: PetscErrorCode MatMult(Mat mat,Vec x,Vec y)
2300: {
2308: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2309: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2310: if (x == y) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2311: #if !defined(PETSC_HAVE_CONSTRAINTS)
2312: 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);
2313: 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);
2314: 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);
2315: #endif
2316: VecSetErrorIfLocked(y,3);
2317: if (mat->erroriffailure) {VecValidValues(x,2,PETSC_TRUE);}
2318: MatCheckPreallocated(mat,1);
2320: VecLockReadPush(x);
2321: if (!mat->ops->mult) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s does not have a multiply defined",((PetscObject)mat)->type_name);
2322: PetscLogEventBegin(MAT_Mult,mat,x,y,0);
2323: (*mat->ops->mult)(mat,x,y);
2324: PetscLogEventEnd(MAT_Mult,mat,x,y,0);
2325: if (mat->erroriffailure) {VecValidValues(y,3,PETSC_FALSE);}
2326: VecLockReadPop(x);
2327: return(0);
2328: }
2330: /*@
2331: MatMultTranspose - Computes matrix transpose times a vector y = A^T * x.
2333: Neighbor-wise Collective on Mat
2335: Input Parameters:
2336: + mat - the matrix
2337: - x - the vector to be multiplied
2339: Output Parameters:
2340: . y - the result
2342: Notes:
2343: The vectors x and y cannot be the same. I.e., one cannot
2344: call MatMultTranspose(A,y,y).
2346: For complex numbers this does NOT compute the Hermitian (complex conjugate) transpose multiple,
2347: use MatMultHermitianTranspose()
2349: Level: beginner
2351: .seealso: MatMult(), MatMultAdd(), MatMultTransposeAdd(), MatMultHermitianTranspose(), MatTranspose()
2352: @*/
2353: PetscErrorCode MatMultTranspose(Mat mat,Vec x,Vec y)
2354: {
2363: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2364: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2365: if (x == y) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2366: #if !defined(PETSC_HAVE_CONSTRAINTS)
2367: 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);
2368: 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);
2369: #endif
2370: if (mat->erroriffailure) {VecValidValues(x,2,PETSC_TRUE);}
2371: MatCheckPreallocated(mat,1);
2373: if (!mat->ops->multtranspose) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s does not have a multiply transpose defined",((PetscObject)mat)->type_name);
2374: PetscLogEventBegin(MAT_MultTranspose,mat,x,y,0);
2375: VecLockReadPush(x);
2376: (*mat->ops->multtranspose)(mat,x,y);
2377: VecLockReadPop(x);
2378: PetscLogEventEnd(MAT_MultTranspose,mat,x,y,0);
2379: PetscObjectStateIncrease((PetscObject)y);
2380: if (mat->erroriffailure) {VecValidValues(y,3,PETSC_FALSE);}
2381: return(0);
2382: }
2384: /*@
2385: MatMultHermitianTranspose - Computes matrix Hermitian transpose times a vector.
2387: Neighbor-wise Collective on Mat
2389: Input Parameters:
2390: + mat - the matrix
2391: - x - the vector to be multilplied
2393: Output Parameters:
2394: . y - the result
2396: Notes:
2397: The vectors x and y cannot be the same. I.e., one cannot
2398: call MatMultHermitianTranspose(A,y,y).
2400: Also called the conjugate transpose, complex conjugate transpose, or adjoint.
2402: For real numbers MatMultTranspose() and MatMultHermitianTranspose() are identical.
2404: Level: beginner
2406: .seealso: MatMult(), MatMultAdd(), MatMultHermitianTransposeAdd(), MatMultTranspose()
2407: @*/
2408: PetscErrorCode MatMultHermitianTranspose(Mat mat,Vec x,Vec y)
2409: {
2411: Vec w;
2419: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2420: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2421: if (x == y) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2422: #if !defined(PETSC_HAVE_CONSTRAINTS)
2423: 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);
2424: 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);
2425: #endif
2426: MatCheckPreallocated(mat,1);
2428: PetscLogEventBegin(MAT_MultHermitianTranspose,mat,x,y,0);
2429: if (mat->ops->multhermitiantranspose) {
2430: VecLockReadPush(x);
2431: (*mat->ops->multhermitiantranspose)(mat,x,y);
2432: VecLockReadPop(x);
2433: } else {
2434: VecDuplicate(x,&w);
2435: VecCopy(x,w);
2436: VecConjugate(w);
2437: MatMultTranspose(mat,w,y);
2438: VecDestroy(&w);
2439: VecConjugate(y);
2440: }
2441: PetscLogEventEnd(MAT_MultHermitianTranspose,mat,x,y,0);
2442: PetscObjectStateIncrease((PetscObject)y);
2443: return(0);
2444: }
2446: /*@
2447: MatMultAdd - Computes v3 = v2 + A * v1.
2449: Neighbor-wise Collective on Mat
2451: Input Parameters:
2452: + mat - the matrix
2453: - v1, v2 - the vectors
2455: Output Parameters:
2456: . v3 - the result
2458: Notes:
2459: The vectors v1 and v3 cannot be the same. I.e., one cannot
2460: call MatMultAdd(A,v1,v2,v1).
2462: Level: beginner
2464: .seealso: MatMultTranspose(), MatMult(), MatMultTransposeAdd()
2465: @*/
2466: PetscErrorCode MatMultAdd(Mat mat,Vec v1,Vec v2,Vec v3)
2467: {
2477: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2478: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2479: 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);
2480: /* 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);
2481: 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); */
2482: 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);
2483: 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);
2484: if (v1 == v3) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"v1 and v3 must be different vectors");
2485: MatCheckPreallocated(mat,1);
2487: if (!mat->ops->multadd) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"No MatMultAdd() for matrix type %s",((PetscObject)mat)->type_name);
2488: PetscLogEventBegin(MAT_MultAdd,mat,v1,v2,v3);
2489: VecLockReadPush(v1);
2490: (*mat->ops->multadd)(mat,v1,v2,v3);
2491: VecLockReadPop(v1);
2492: PetscLogEventEnd(MAT_MultAdd,mat,v1,v2,v3);
2493: PetscObjectStateIncrease((PetscObject)v3);
2494: return(0);
2495: }
2497: /*@
2498: MatMultTransposeAdd - Computes v3 = v2 + A' * v1.
2500: Neighbor-wise Collective on Mat
2502: Input Parameters:
2503: + mat - the matrix
2504: - v1, v2 - the vectors
2506: Output Parameters:
2507: . v3 - the result
2509: Notes:
2510: The vectors v1 and v3 cannot be the same. I.e., one cannot
2511: call MatMultTransposeAdd(A,v1,v2,v1).
2513: Level: beginner
2515: .seealso: MatMultTranspose(), MatMultAdd(), MatMult()
2516: @*/
2517: PetscErrorCode MatMultTransposeAdd(Mat mat,Vec v1,Vec v2,Vec v3)
2518: {
2528: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2529: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2530: if (!mat->ops->multtransposeadd) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2531: if (v1 == v3) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"v1 and v3 must be different vectors");
2532: 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);
2533: 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);
2534: 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);
2535: MatCheckPreallocated(mat,1);
2537: PetscLogEventBegin(MAT_MultTransposeAdd,mat,v1,v2,v3);
2538: VecLockReadPush(v1);
2539: (*mat->ops->multtransposeadd)(mat,v1,v2,v3);
2540: VecLockReadPop(v1);
2541: PetscLogEventEnd(MAT_MultTransposeAdd,mat,v1,v2,v3);
2542: PetscObjectStateIncrease((PetscObject)v3);
2543: return(0);
2544: }
2546: /*@
2547: MatMultHermitianTransposeAdd - Computes v3 = v2 + A^H * v1.
2549: Neighbor-wise Collective on Mat
2551: Input Parameters:
2552: + mat - the matrix
2553: - v1, v2 - the vectors
2555: Output Parameters:
2556: . v3 - the result
2558: Notes:
2559: The vectors v1 and v3 cannot be the same. I.e., one cannot
2560: call MatMultHermitianTransposeAdd(A,v1,v2,v1).
2562: Level: beginner
2564: .seealso: MatMultHermitianTranspose(), MatMultTranspose(), MatMultAdd(), MatMult()
2565: @*/
2566: PetscErrorCode MatMultHermitianTransposeAdd(Mat mat,Vec v1,Vec v2,Vec v3)
2567: {
2577: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2578: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2579: if (v1 == v3) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"v1 and v3 must be different vectors");
2580: 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);
2581: 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);
2582: 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);
2583: MatCheckPreallocated(mat,1);
2585: PetscLogEventBegin(MAT_MultHermitianTransposeAdd,mat,v1,v2,v3);
2586: VecLockReadPush(v1);
2587: if (mat->ops->multhermitiantransposeadd) {
2588: (*mat->ops->multhermitiantransposeadd)(mat,v1,v2,v3);
2589: } else {
2590: Vec w,z;
2591: VecDuplicate(v1,&w);
2592: VecCopy(v1,w);
2593: VecConjugate(w);
2594: VecDuplicate(v3,&z);
2595: MatMultTranspose(mat,w,z);
2596: VecDestroy(&w);
2597: VecConjugate(z);
2598: if (v2 != v3) {
2599: VecWAXPY(v3,1.0,v2,z);
2600: } else {
2601: VecAXPY(v3,1.0,z);
2602: }
2603: VecDestroy(&z);
2604: }
2605: VecLockReadPop(v1);
2606: PetscLogEventEnd(MAT_MultHermitianTransposeAdd,mat,v1,v2,v3);
2607: PetscObjectStateIncrease((PetscObject)v3);
2608: return(0);
2609: }
2611: /*@
2612: MatMultConstrained - The inner multiplication routine for a
2613: constrained matrix P^T A P.
2615: Neighbor-wise Collective on Mat
2617: Input Parameters:
2618: + mat - the matrix
2619: - x - the vector to be multilplied
2621: Output Parameters:
2622: . y - the result
2624: Notes:
2625: The vectors x and y cannot be the same. I.e., one cannot
2626: call MatMult(A,y,y).
2628: Level: beginner
2630: .seealso: MatMult(), MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
2631: @*/
2632: PetscErrorCode MatMultConstrained(Mat mat,Vec x,Vec y)
2633: {
2640: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2641: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2642: if (x == y) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2643: 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);
2644: 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);
2645: 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);
2647: PetscLogEventBegin(MAT_MultConstrained,mat,x,y,0);
2648: VecLockReadPush(x);
2649: (*mat->ops->multconstrained)(mat,x,y);
2650: VecLockReadPop(x);
2651: PetscLogEventEnd(MAT_MultConstrained,mat,x,y,0);
2652: PetscObjectStateIncrease((PetscObject)y);
2653: return(0);
2654: }
2656: /*@
2657: MatMultTransposeConstrained - The inner multiplication routine for a
2658: constrained matrix P^T A^T P.
2660: Neighbor-wise Collective on Mat
2662: Input Parameters:
2663: + mat - the matrix
2664: - x - the vector to be multilplied
2666: Output Parameters:
2667: . y - the result
2669: Notes:
2670: The vectors x and y cannot be the same. I.e., one cannot
2671: call MatMult(A,y,y).
2673: Level: beginner
2675: .seealso: MatMult(), MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
2676: @*/
2677: PetscErrorCode MatMultTransposeConstrained(Mat mat,Vec x,Vec y)
2678: {
2685: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2686: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2687: if (x == y) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2688: 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);
2689: 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);
2691: PetscLogEventBegin(MAT_MultConstrained,mat,x,y,0);
2692: (*mat->ops->multtransposeconstrained)(mat,x,y);
2693: PetscLogEventEnd(MAT_MultConstrained,mat,x,y,0);
2694: PetscObjectStateIncrease((PetscObject)y);
2695: return(0);
2696: }
2698: /*@C
2699: MatGetFactorType - gets the type of factorization it is
2701: Not Collective
2703: Input Parameters:
2704: . mat - the matrix
2706: Output Parameters:
2707: . t - the type, one of MAT_FACTOR_NONE, MAT_FACTOR_LU, MAT_FACTOR_CHOLESKY, MAT_FACTOR_ILU, MAT_FACTOR_ICC,MAT_FACTOR_ILUDT
2709: Level: intermediate
2711: .seealso: MatFactorType, MatGetFactor(), MatSetFactorType()
2712: @*/
2713: PetscErrorCode MatGetFactorType(Mat mat,MatFactorType *t)
2714: {
2719: *t = mat->factortype;
2720: return(0);
2721: }
2723: /*@C
2724: MatSetFactorType - sets the type of factorization it is
2726: Logically Collective on Mat
2728: Input Parameters:
2729: + mat - the matrix
2730: - t - the type, one of MAT_FACTOR_NONE, MAT_FACTOR_LU, MAT_FACTOR_CHOLESKY, MAT_FACTOR_ILU, MAT_FACTOR_ICC,MAT_FACTOR_ILUDT
2732: Level: intermediate
2734: .seealso: MatFactorType, MatGetFactor(), MatGetFactorType()
2735: @*/
2736: PetscErrorCode MatSetFactorType(Mat mat, MatFactorType t)
2737: {
2741: mat->factortype = t;
2742: return(0);
2743: }
2745: /* ------------------------------------------------------------*/
2746: /*@C
2747: MatGetInfo - Returns information about matrix storage (number of
2748: nonzeros, memory, etc.).
2750: Collective on Mat if MAT_GLOBAL_MAX or MAT_GLOBAL_SUM is used as the flag
2752: Input Parameters:
2753: . mat - the matrix
2755: Output Parameters:
2756: + flag - flag indicating the type of parameters to be returned
2757: (MAT_LOCAL - local matrix, MAT_GLOBAL_MAX - maximum over all processors,
2758: MAT_GLOBAL_SUM - sum over all processors)
2759: - info - matrix information context
2761: Notes:
2762: The MatInfo context contains a variety of matrix data, including
2763: number of nonzeros allocated and used, number of mallocs during
2764: matrix assembly, etc. Additional information for factored matrices
2765: is provided (such as the fill ratio, number of mallocs during
2766: factorization, etc.). Much of this info is printed to PETSC_STDOUT
2767: when using the runtime options
2768: $ -info -mat_view ::ascii_info
2770: Example for C/C++ Users:
2771: See the file ${PETSC_DIR}/include/petscmat.h for a complete list of
2772: data within the MatInfo context. For example,
2773: .vb
2774: MatInfo info;
2775: Mat A;
2776: double mal, nz_a, nz_u;
2778: MatGetInfo(A,MAT_LOCAL,&info);
2779: mal = info.mallocs;
2780: nz_a = info.nz_allocated;
2781: .ve
2783: Example for Fortran Users:
2784: Fortran users should declare info as a double precision
2785: array of dimension MAT_INFO_SIZE, and then extract the parameters
2786: of interest. See the file ${PETSC_DIR}/include/petsc/finclude/petscmat.h
2787: a complete list of parameter names.
2788: .vb
2789: double precision info(MAT_INFO_SIZE)
2790: double precision mal, nz_a
2791: Mat A
2792: integer ierr
2794: call MatGetInfo(A,MAT_LOCAL,info,ierr)
2795: mal = info(MAT_INFO_MALLOCS)
2796: nz_a = info(MAT_INFO_NZ_ALLOCATED)
2797: .ve
2799: Level: intermediate
2801: Developer Note: fortran interface is not autogenerated as the f90
2802: interface defintion cannot be generated correctly [due to MatInfo]
2804: .seealso: MatStashGetInfo()
2806: @*/
2807: PetscErrorCode MatGetInfo(Mat mat,MatInfoType flag,MatInfo *info)
2808: {
2815: if (!mat->ops->getinfo) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2816: MatCheckPreallocated(mat,1);
2817: (*mat->ops->getinfo)(mat,flag,info);
2818: return(0);
2819: }
2821: /*
2822: This is used by external packages where it is not easy to get the info from the actual
2823: matrix factorization.
2824: */
2825: PetscErrorCode MatGetInfo_External(Mat A,MatInfoType flag,MatInfo *info)
2826: {
2830: PetscMemzero(info,sizeof(MatInfo));
2831: return(0);
2832: }
2834: /* ----------------------------------------------------------*/
2836: /*@C
2837: MatLUFactor - Performs in-place LU factorization of matrix.
2839: Collective on Mat
2841: Input Parameters:
2842: + mat - the matrix
2843: . row - row permutation
2844: . col - column permutation
2845: - info - options for factorization, includes
2846: $ fill - expected fill as ratio of original fill.
2847: $ dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
2848: $ Run with the option -info to determine an optimal value to use
2850: Notes:
2851: Most users should employ the simplified KSP interface for linear solvers
2852: instead of working directly with matrix algebra routines such as this.
2853: See, e.g., KSPCreate().
2855: This changes the state of the matrix to a factored matrix; it cannot be used
2856: for example with MatSetValues() unless one first calls MatSetUnfactored().
2858: Level: developer
2860: .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor(),
2861: MatGetOrdering(), MatSetUnfactored(), MatFactorInfo, MatGetFactor()
2863: Developer Note: fortran interface is not autogenerated as the f90
2864: interface defintion cannot be generated correctly [due to MatFactorInfo]
2866: @*/
2867: PetscErrorCode MatLUFactor(Mat mat,IS row,IS col,const MatFactorInfo *info)
2868: {
2870: MatFactorInfo tinfo;
2878: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2879: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2880: if (!mat->ops->lufactor) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2881: MatCheckPreallocated(mat,1);
2882: if (!info) {
2883: MatFactorInfoInitialize(&tinfo);
2884: info = &tinfo;
2885: }
2887: PetscLogEventBegin(MAT_LUFactor,mat,row,col,0);
2888: (*mat->ops->lufactor)(mat,row,col,info);
2889: PetscLogEventEnd(MAT_LUFactor,mat,row,col,0);
2890: PetscObjectStateIncrease((PetscObject)mat);
2891: return(0);
2892: }
2894: /*@C
2895: MatILUFactor - Performs in-place ILU factorization of matrix.
2897: Collective on Mat
2899: Input Parameters:
2900: + mat - the matrix
2901: . row - row permutation
2902: . col - column permutation
2903: - info - structure containing
2904: $ levels - number of levels of fill.
2905: $ expected fill - as ratio of original fill.
2906: $ 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
2907: missing diagonal entries)
2909: Notes:
2910: Probably really in-place only when level of fill is zero, otherwise allocates
2911: new space to store factored matrix and deletes previous memory.
2913: Most users should employ the simplified KSP interface for linear solvers
2914: instead of working directly with matrix algebra routines such as this.
2915: See, e.g., KSPCreate().
2917: Level: developer
2919: .seealso: MatILUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor(), MatFactorInfo
2921: Developer Note: fortran interface is not autogenerated as the f90
2922: interface defintion cannot be generated correctly [due to MatFactorInfo]
2924: @*/
2925: PetscErrorCode MatILUFactor(Mat mat,IS row,IS col,const MatFactorInfo *info)
2926: {
2935: if (mat->rmap->N != mat->cmap->N) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONG,"matrix must be square");
2936: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2937: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2938: if (!mat->ops->ilufactor) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2939: MatCheckPreallocated(mat,1);
2941: PetscLogEventBegin(MAT_ILUFactor,mat,row,col,0);
2942: (*mat->ops->ilufactor)(mat,row,col,info);
2943: PetscLogEventEnd(MAT_ILUFactor,mat,row,col,0);
2944: PetscObjectStateIncrease((PetscObject)mat);
2945: return(0);
2946: }
2948: /*@C
2949: MatLUFactorSymbolic - Performs symbolic LU factorization of matrix.
2950: Call this routine before calling MatLUFactorNumeric().
2952: Collective on Mat
2954: Input Parameters:
2955: + fact - the factor matrix obtained with MatGetFactor()
2956: . mat - the matrix
2957: . row, col - row and column permutations
2958: - info - options for factorization, includes
2959: $ fill - expected fill as ratio of original fill.
2960: $ dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
2961: $ Run with the option -info to determine an optimal value to use
2964: Notes:
2965: See Users-Manual: ch_mat for additional information about choosing the fill factor for better efficiency.
2967: Most users should employ the simplified KSP interface for linear solvers
2968: instead of working directly with matrix algebra routines such as this.
2969: See, e.g., KSPCreate().
2971: Level: developer
2973: .seealso: MatLUFactor(), MatLUFactorNumeric(), MatCholeskyFactor(), MatFactorInfo, MatFactorInfoInitialize()
2975: Developer Note: fortran interface is not autogenerated as the f90
2976: interface defintion cannot be generated correctly [due to MatFactorInfo]
2978: @*/
2979: PetscErrorCode MatLUFactorSymbolic(Mat fact,Mat mat,IS row,IS col,const MatFactorInfo *info)
2980: {
2982: MatFactorInfo tinfo;
2991: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2992: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2993: if (!(fact)->ops->lufactorsymbolic) {
2994: MatSolverType spackage;
2995: MatFactorGetSolverType(fact,&spackage);
2996: SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s symbolic LU using solver package %s",((PetscObject)mat)->type_name,spackage);
2997: }
2998: MatCheckPreallocated(mat,2);
2999: if (!info) {
3000: MatFactorInfoInitialize(&tinfo);
3001: info = &tinfo;
3002: }
3004: PetscLogEventBegin(MAT_LUFactorSymbolic,mat,row,col,0);
3005: (fact->ops->lufactorsymbolic)(fact,mat,row,col,info);
3006: PetscLogEventEnd(MAT_LUFactorSymbolic,mat,row,col,0);
3007: PetscObjectStateIncrease((PetscObject)fact);
3008: return(0);
3009: }
3011: /*@C
3012: MatLUFactorNumeric - Performs numeric LU factorization of a matrix.
3013: Call this routine after first calling MatLUFactorSymbolic().
3015: Collective on Mat
3017: Input Parameters:
3018: + fact - the factor matrix obtained with MatGetFactor()
3019: . mat - the matrix
3020: - info - options for factorization
3022: Notes:
3023: See MatLUFactor() for in-place factorization. See
3024: MatCholeskyFactorNumeric() for the symmetric, positive definite case.
3026: Most users should employ the simplified KSP interface for linear solvers
3027: instead of working directly with matrix algebra routines such as this.
3028: See, e.g., KSPCreate().
3030: Level: developer
3032: .seealso: MatLUFactorSymbolic(), MatLUFactor(), MatCholeskyFactor()
3034: Developer Note: fortran interface is not autogenerated as the f90
3035: interface defintion cannot be generated correctly [due to MatFactorInfo]
3037: @*/
3038: PetscErrorCode MatLUFactorNumeric(Mat fact,Mat mat,const MatFactorInfo *info)
3039: {
3040: MatFactorInfo tinfo;
3048: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3049: 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);
3051: if (!(fact)->ops->lufactornumeric) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s numeric LU",((PetscObject)mat)->type_name);
3052: MatCheckPreallocated(mat,2);
3053: if (!info) {
3054: MatFactorInfoInitialize(&tinfo);
3055: info = &tinfo;
3056: }
3058: PetscLogEventBegin(MAT_LUFactorNumeric,mat,fact,0,0);
3059: (fact->ops->lufactornumeric)(fact,mat,info);
3060: PetscLogEventEnd(MAT_LUFactorNumeric,mat,fact,0,0);
3061: MatViewFromOptions(fact,NULL,"-mat_factor_view");
3062: PetscObjectStateIncrease((PetscObject)fact);
3063: return(0);
3064: }
3066: /*@C
3067: MatCholeskyFactor - Performs in-place Cholesky factorization of a
3068: symmetric matrix.
3070: Collective on Mat
3072: Input Parameters:
3073: + mat - the matrix
3074: . perm - row and column permutations
3075: - f - expected fill as ratio of original fill
3077: Notes:
3078: See MatLUFactor() for the nonsymmetric case. See also
3079: MatCholeskyFactorSymbolic(), and MatCholeskyFactorNumeric().
3081: Most users should employ the simplified KSP interface for linear solvers
3082: instead of working directly with matrix algebra routines such as this.
3083: See, e.g., KSPCreate().
3085: Level: developer
3087: .seealso: MatLUFactor(), MatCholeskyFactorSymbolic(), MatCholeskyFactorNumeric()
3088: MatGetOrdering()
3090: Developer Note: fortran interface is not autogenerated as the f90
3091: interface defintion cannot be generated correctly [due to MatFactorInfo]
3093: @*/
3094: PetscErrorCode MatCholeskyFactor(Mat mat,IS perm,const MatFactorInfo *info)
3095: {
3097: MatFactorInfo tinfo;
3104: if (mat->rmap->N != mat->cmap->N) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONG,"Matrix must be square");
3105: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3106: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3107: 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);
3108: MatCheckPreallocated(mat,1);
3109: if (!info) {
3110: MatFactorInfoInitialize(&tinfo);
3111: info = &tinfo;
3112: }
3114: PetscLogEventBegin(MAT_CholeskyFactor,mat,perm,0,0);
3115: (*mat->ops->choleskyfactor)(mat,perm,info);
3116: PetscLogEventEnd(MAT_CholeskyFactor,mat,perm,0,0);
3117: PetscObjectStateIncrease((PetscObject)mat);
3118: return(0);
3119: }
3121: /*@C
3122: MatCholeskyFactorSymbolic - Performs symbolic Cholesky factorization
3123: of a symmetric matrix.
3125: Collective on Mat
3127: Input Parameters:
3128: + fact - the factor matrix obtained with MatGetFactor()
3129: . mat - the matrix
3130: . perm - row and column permutations
3131: - info - options for factorization, includes
3132: $ fill - expected fill as ratio of original fill.
3133: $ dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
3134: $ Run with the option -info to determine an optimal value to use
3136: Notes:
3137: See MatLUFactorSymbolic() for the nonsymmetric case. See also
3138: MatCholeskyFactor() and MatCholeskyFactorNumeric().
3140: Most users should employ the simplified KSP interface for linear solvers
3141: instead of working directly with matrix algebra routines such as this.
3142: See, e.g., KSPCreate().
3144: Level: developer
3146: .seealso: MatLUFactorSymbolic(), MatCholeskyFactor(), MatCholeskyFactorNumeric()
3147: MatGetOrdering()
3149: Developer Note: fortran interface is not autogenerated as the f90
3150: interface defintion cannot be generated correctly [due to MatFactorInfo]
3152: @*/
3153: PetscErrorCode MatCholeskyFactorSymbolic(Mat fact,Mat mat,IS perm,const MatFactorInfo *info)
3154: {
3156: MatFactorInfo tinfo;
3164: if (mat->rmap->N != mat->cmap->N) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONG,"Matrix must be square");
3165: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3166: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3167: if (!(fact)->ops->choleskyfactorsymbolic) {
3168: MatSolverType spackage;
3169: MatFactorGetSolverType(fact,&spackage);
3170: SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s symbolic factor Cholesky using solver package %s",((PetscObject)mat)->type_name,spackage);
3171: }
3172: MatCheckPreallocated(mat,2);
3173: if (!info) {
3174: MatFactorInfoInitialize(&tinfo);
3175: info = &tinfo;
3176: }
3178: PetscLogEventBegin(MAT_CholeskyFactorSymbolic,mat,perm,0,0);
3179: (fact->ops->choleskyfactorsymbolic)(fact,mat,perm,info);
3180: PetscLogEventEnd(MAT_CholeskyFactorSymbolic,mat,perm,0,0);
3181: PetscObjectStateIncrease((PetscObject)fact);
3182: return(0);
3183: }
3185: /*@C
3186: MatCholeskyFactorNumeric - Performs numeric Cholesky factorization
3187: of a symmetric matrix. Call this routine after first calling
3188: MatCholeskyFactorSymbolic().
3190: Collective on Mat
3192: Input Parameters:
3193: + fact - the factor matrix obtained with MatGetFactor()
3194: . mat - the initial matrix
3195: . info - options for factorization
3196: - fact - the symbolic factor of mat
3199: Notes:
3200: Most users should employ the simplified KSP interface for linear solvers
3201: instead of working directly with matrix algebra routines such as this.
3202: See, e.g., KSPCreate().
3204: Level: developer
3206: .seealso: MatCholeskyFactorSymbolic(), MatCholeskyFactor(), MatLUFactorNumeric()
3208: Developer Note: fortran interface is not autogenerated as the f90
3209: interface defintion cannot be generated correctly [due to MatFactorInfo]
3211: @*/
3212: PetscErrorCode MatCholeskyFactorNumeric(Mat fact,Mat mat,const MatFactorInfo *info)
3213: {
3214: MatFactorInfo tinfo;
3222: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3223: if (!(fact)->ops->choleskyfactornumeric) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s numeric factor Cholesky",((PetscObject)mat)->type_name);
3224: 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);
3225: MatCheckPreallocated(mat,2);
3226: if (!info) {
3227: MatFactorInfoInitialize(&tinfo);
3228: info = &tinfo;
3229: }
3231: PetscLogEventBegin(MAT_CholeskyFactorNumeric,mat,fact,0,0);
3232: (fact->ops->choleskyfactornumeric)(fact,mat,info);
3233: PetscLogEventEnd(MAT_CholeskyFactorNumeric,mat,fact,0,0);
3234: MatViewFromOptions(fact,NULL,"-mat_factor_view");
3235: PetscObjectStateIncrease((PetscObject)fact);
3236: return(0);
3237: }
3239: /* ----------------------------------------------------------------*/
3240: /*@
3241: MatSolve - Solves A x = b, given a factored matrix.
3243: Neighbor-wise Collective on Mat
3245: Input Parameters:
3246: + mat - the factored matrix
3247: - b - the right-hand-side vector
3249: Output Parameter:
3250: . x - the result vector
3252: Notes:
3253: The vectors b and x cannot be the same. I.e., one cannot
3254: call MatSolve(A,x,x).
3256: Notes:
3257: Most users should employ the simplified KSP interface for linear solvers
3258: instead of working directly with matrix algebra routines such as this.
3259: See, e.g., KSPCreate().
3261: Level: developer
3263: .seealso: MatSolveAdd(), MatSolveTranspose(), MatSolveTransposeAdd()
3264: @*/
3265: PetscErrorCode MatSolve(Mat mat,Vec b,Vec x)
3266: {
3276: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3277: 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);
3278: 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);
3279: 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);
3280: if (!mat->rmap->N && !mat->cmap->N) return(0);
3281: if (!mat->ops->solve) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3282: MatCheckPreallocated(mat,1);
3284: PetscLogEventBegin(MAT_Solve,mat,b,x,0);
3285: if (mat->factorerrortype) {
3286: PetscInfo1(mat,"MatFactorError %D\n",mat->factorerrortype);
3287: VecSetInf(x);
3288: } else {
3289: if (!mat->ops->solve) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3290: (*mat->ops->solve)(mat,b,x);
3291: }
3292: PetscLogEventEnd(MAT_Solve,mat,b,x,0);
3293: PetscObjectStateIncrease((PetscObject)x);
3294: return(0);
3295: }
3297: static PetscErrorCode MatMatSolve_Basic(Mat A,Mat B,Mat X,PetscBool trans)
3298: {
3300: Vec b,x;
3301: PetscInt m,N,i;
3302: PetscScalar *bb,*xx;
3305: MatDenseGetArrayRead(B,(const PetscScalar**)&bb);
3306: MatDenseGetArray(X,&xx);
3307: MatGetLocalSize(B,&m,NULL); /* number local rows */
3308: MatGetSize(B,NULL,&N); /* total columns in dense matrix */
3309: MatCreateVecs(A,&x,&b);
3310: for (i=0; i<N; i++) {
3311: VecPlaceArray(b,bb + i*m);
3312: VecPlaceArray(x,xx + i*m);
3313: if (trans) {
3314: MatSolveTranspose(A,b,x);
3315: } else {
3316: MatSolve(A,b,x);
3317: }
3318: VecResetArray(x);
3319: VecResetArray(b);
3320: }
3321: VecDestroy(&b);
3322: VecDestroy(&x);
3323: MatDenseRestoreArrayRead(B,(const PetscScalar**)&bb);
3324: MatDenseRestoreArray(X,&xx);
3325: return(0);
3326: }
3328: /*@
3329: MatMatSolve - Solves A X = B, given a factored matrix.
3331: Neighbor-wise Collective on Mat
3333: Input Parameters:
3334: + A - the factored matrix
3335: - B - the right-hand-side matrix (dense matrix)
3337: Output Parameter:
3338: . X - the result matrix (dense matrix)
3340: Notes:
3341: The matrices b and x cannot be the same. I.e., one cannot
3342: call MatMatSolve(A,x,x).
3344: Notes:
3345: Most users should usually employ the simplified KSP interface for linear solvers
3346: instead of working directly with matrix algebra routines such as this.
3347: See, e.g., KSPCreate(). However KSP can only solve for one vector (column of X)
3348: at a time.
3350: When using SuperLU_Dist as a parallel solver PETSc will use the SuperLU_Dist functionality to solve multiple right hand sides simultaneously. For MUMPS
3351: it calls a separate solve for each right hand side since MUMPS does not yet support distributed right hand sides.
3353: Since the resulting matrix X must always be dense we do not support sparse representation of the matrix B.
3355: Level: developer
3357: .seealso: MatMatSolveTranspose(), MatLUFactor(), MatCholeskyFactor()
3358: @*/
3359: PetscErrorCode MatMatSolve(Mat A,Mat B,Mat X)
3360: {
3370: if (X == B) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_IDN,"X and B must be different matrices");
3371: 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);
3372: 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);
3373: 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");
3374: if (!A->rmap->N && !A->cmap->N) return(0);
3375: if (!A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
3376: MatCheckPreallocated(A,1);
3378: PetscLogEventBegin(MAT_MatSolve,A,B,X,0);
3379: if (!A->ops->matsolve) {
3380: PetscInfo1(A,"Mat type %s using basic MatMatSolve\n",((PetscObject)A)->type_name);
3381: MatMatSolve_Basic(A,B,X,PETSC_FALSE);
3382: } else {
3383: (*A->ops->matsolve)(A,B,X);
3384: }
3385: PetscLogEventEnd(MAT_MatSolve,A,B,X,0);
3386: PetscObjectStateIncrease((PetscObject)X);
3387: return(0);
3388: }
3390: /*@
3391: MatMatSolveTranspose - Solves A^T X = B, given a factored matrix.
3393: Neighbor-wise Collective on Mat
3395: Input Parameters:
3396: + A - the factored matrix
3397: - B - the right-hand-side matrix (dense matrix)
3399: Output Parameter:
3400: . X - the result matrix (dense matrix)
3402: Notes:
3403: The matrices B and X cannot be the same. I.e., one cannot
3404: call MatMatSolveTranspose(A,X,X).
3406: Notes:
3407: Most users should usually employ the simplified KSP interface for linear solvers
3408: instead of working directly with matrix algebra routines such as this.
3409: See, e.g., KSPCreate(). However KSP can only solve for one vector (column of X)
3410: at a time.
3412: When using SuperLU_Dist or MUMPS as a parallel solver, PETSc will use their functionality to solve multiple right hand sides simultaneously.
3414: Level: developer
3416: .seealso: MatMatSolve(), MatLUFactor(), MatCholeskyFactor()
3417: @*/
3418: PetscErrorCode MatMatSolveTranspose(Mat A,Mat B,Mat X)
3419: {
3429: if (X == B) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_IDN,"X and B must be different matrices");
3430: 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);
3431: 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);
3432: 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);
3433: 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");
3434: if (!A->rmap->N && !A->cmap->N) return(0);
3435: if (!A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
3436: MatCheckPreallocated(A,1);
3438: PetscLogEventBegin(MAT_MatSolve,A,B,X,0);
3439: if (!A->ops->matsolvetranspose) {
3440: PetscInfo1(A,"Mat type %s using basic MatMatSolveTranspose\n",((PetscObject)A)->type_name);
3441: MatMatSolve_Basic(A,B,X,PETSC_TRUE);
3442: } else {
3443: (*A->ops->matsolvetranspose)(A,B,X);
3444: }
3445: PetscLogEventEnd(MAT_MatSolve,A,B,X,0);
3446: PetscObjectStateIncrease((PetscObject)X);
3447: return(0);
3448: }
3450: /*@
3451: MatMatTransposeSolve - Solves A X = B^T, given a factored matrix.
3453: Neighbor-wise Collective on Mat
3455: Input Parameters:
3456: + A - the factored matrix
3457: - Bt - the transpose of right-hand-side matrix
3459: Output Parameter:
3460: . X - the result matrix (dense matrix)
3462: Notes:
3463: Most users should usually employ the simplified KSP interface for linear solvers
3464: instead of working directly with matrix algebra routines such as this.
3465: See, e.g., KSPCreate(). However KSP can only solve for one vector (column of X)
3466: at a time.
3468: 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().
3470: Level: developer
3472: .seealso: MatMatSolve(), MatMatSolveTranspose(), MatLUFactor(), MatCholeskyFactor()
3473: @*/
3474: PetscErrorCode MatMatTransposeSolve(Mat A,Mat Bt,Mat X)
3475: {
3486: if (X == Bt) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_IDN,"X and B must be different matrices");
3487: 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);
3488: 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);
3489: 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");
3490: if (!A->rmap->N && !A->cmap->N) return(0);
3491: if (!A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
3492: MatCheckPreallocated(A,1);
3494: if (!A->ops->mattransposesolve) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Mat type %s",((PetscObject)A)->type_name);
3495: PetscLogEventBegin(MAT_MatTrSolve,A,Bt,X,0);
3496: (*A->ops->mattransposesolve)(A,Bt,X);
3497: PetscLogEventEnd(MAT_MatTrSolve,A,Bt,X,0);
3498: PetscObjectStateIncrease((PetscObject)X);
3499: return(0);
3500: }
3502: /*@
3503: MatForwardSolve - Solves L x = b, given a factored matrix, A = LU, or
3504: U^T*D^(1/2) x = b, given a factored symmetric matrix, A = U^T*D*U,
3506: Neighbor-wise Collective on Mat
3508: Input Parameters:
3509: + mat - the factored matrix
3510: - b - the right-hand-side vector
3512: Output Parameter:
3513: . x - the result vector
3515: Notes:
3516: MatSolve() should be used for most applications, as it performs
3517: a forward solve followed by a backward solve.
3519: The vectors b and x cannot be the same, i.e., one cannot
3520: call MatForwardSolve(A,x,x).
3522: For matrix in seqsbaij format with block size larger than 1,
3523: the diagonal blocks are not implemented as D = D^(1/2) * D^(1/2) yet.
3524: MatForwardSolve() solves U^T*D y = b, and
3525: MatBackwardSolve() solves U x = y.
3526: Thus they do not provide a symmetric preconditioner.
3528: Most users should employ the simplified KSP interface for linear solvers
3529: instead of working directly with matrix algebra routines such as this.
3530: See, e.g., KSPCreate().
3532: Level: developer
3534: .seealso: MatSolve(), MatBackwardSolve()
3535: @*/
3536: PetscErrorCode MatForwardSolve(Mat mat,Vec b,Vec x)
3537: {
3547: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3548: 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);
3549: 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);
3550: 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);
3551: if (!mat->rmap->N && !mat->cmap->N) return(0);
3552: if (!mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
3553: MatCheckPreallocated(mat,1);
3555: if (!mat->ops->forwardsolve) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3556: PetscLogEventBegin(MAT_ForwardSolve,mat,b,x,0);
3557: (*mat->ops->forwardsolve)(mat,b,x);
3558: PetscLogEventEnd(MAT_ForwardSolve,mat,b,x,0);
3559: PetscObjectStateIncrease((PetscObject)x);
3560: return(0);
3561: }
3563: /*@
3564: MatBackwardSolve - Solves U x = b, given a factored matrix, A = LU.
3565: D^(1/2) U x = b, given a factored symmetric matrix, A = U^T*D*U,
3567: Neighbor-wise Collective on Mat
3569: Input Parameters:
3570: + mat - the factored matrix
3571: - b - the right-hand-side vector
3573: Output Parameter:
3574: . x - the result vector
3576: Notes:
3577: MatSolve() should be used for most applications, as it performs
3578: a forward solve followed by a backward solve.
3580: The vectors b and x cannot be the same. I.e., one cannot
3581: call MatBackwardSolve(A,x,x).
3583: For matrix in seqsbaij format with block size larger than 1,
3584: the diagonal blocks are not implemented as D = D^(1/2) * D^(1/2) yet.
3585: MatForwardSolve() solves U^T*D y = b, and
3586: MatBackwardSolve() solves U x = y.
3587: Thus they do not provide a symmetric preconditioner.
3589: Most users should employ the simplified KSP interface for linear solvers
3590: instead of working directly with matrix algebra routines such as this.
3591: See, e.g., KSPCreate().
3593: Level: developer
3595: .seealso: MatSolve(), MatForwardSolve()
3596: @*/
3597: PetscErrorCode MatBackwardSolve(Mat mat,Vec b,Vec x)
3598: {
3608: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3609: 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);
3610: 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);
3611: 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);
3612: if (!mat->rmap->N && !mat->cmap->N) return(0);
3613: if (!mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
3614: MatCheckPreallocated(mat,1);
3616: if (!mat->ops->backwardsolve) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3617: PetscLogEventBegin(MAT_BackwardSolve,mat,b,x,0);
3618: (*mat->ops->backwardsolve)(mat,b,x);
3619: PetscLogEventEnd(MAT_BackwardSolve,mat,b,x,0);
3620: PetscObjectStateIncrease((PetscObject)x);
3621: return(0);
3622: }
3624: /*@
3625: MatSolveAdd - Computes x = y + inv(A)*b, given a factored matrix.
3627: Neighbor-wise Collective on Mat
3629: Input Parameters:
3630: + mat - the factored matrix
3631: . b - the right-hand-side vector
3632: - y - the vector to be added to
3634: Output Parameter:
3635: . x - the result vector
3637: Notes:
3638: The vectors b and x cannot be the same. I.e., one cannot
3639: call MatSolveAdd(A,x,y,x).
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(), MatSolveTranspose(), MatSolveTransposeAdd()
3648: @*/
3649: PetscErrorCode MatSolveAdd(Mat mat,Vec b,Vec y,Vec x)
3650: {
3651: PetscScalar one = 1.0;
3652: Vec tmp;
3664: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3665: 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);
3666: 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);
3667: 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);
3668: 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);
3669: 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);
3670: if (!mat->rmap->N && !mat->cmap->N) return(0);
3671: if (!mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
3672: MatCheckPreallocated(mat,1);
3674: PetscLogEventBegin(MAT_SolveAdd,mat,b,x,y);
3675: if (mat->ops->solveadd) {
3676: (*mat->ops->solveadd)(mat,b,y,x);
3677: } else {
3678: /* do the solve then the add manually */
3679: if (x != y) {
3680: MatSolve(mat,b,x);
3681: VecAXPY(x,one,y);
3682: } else {
3683: VecDuplicate(x,&tmp);
3684: PetscLogObjectParent((PetscObject)mat,(PetscObject)tmp);
3685: VecCopy(x,tmp);
3686: MatSolve(mat,b,x);
3687: VecAXPY(x,one,tmp);
3688: VecDestroy(&tmp);
3689: }
3690: }
3691: PetscLogEventEnd(MAT_SolveAdd,mat,b,x,y);
3692: PetscObjectStateIncrease((PetscObject)x);
3693: return(0);
3694: }
3696: /*@
3697: MatSolveTranspose - Solves A' x = b, given a factored matrix.
3699: Neighbor-wise Collective on Mat
3701: Input Parameters:
3702: + mat - the factored matrix
3703: - b - the right-hand-side vector
3705: Output Parameter:
3706: . x - the result vector
3708: Notes:
3709: The vectors b and x cannot be the same. I.e., one cannot
3710: call MatSolveTranspose(A,x,x).
3712: Most users should employ the simplified KSP interface for linear solvers
3713: instead of working directly with matrix algebra routines such as this.
3714: See, e.g., KSPCreate().
3716: Level: developer
3718: .seealso: MatSolve(), MatSolveAdd(), MatSolveTransposeAdd()
3719: @*/
3720: PetscErrorCode MatSolveTranspose(Mat mat,Vec b,Vec x)
3721: {
3731: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3732: 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);
3733: 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);
3734: if (!mat->rmap->N && !mat->cmap->N) return(0);
3735: MatCheckPreallocated(mat,1);
3736: PetscLogEventBegin(MAT_SolveTranspose,mat,b,x,0);
3737: if (mat->factorerrortype) {
3738: PetscInfo1(mat,"MatFactorError %D\n",mat->factorerrortype);
3739: VecSetInf(x);
3740: } else {
3741: if (!mat->ops->solvetranspose) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s",((PetscObject)mat)->type_name);
3742: (*mat->ops->solvetranspose)(mat,b,x);
3743: }
3744: PetscLogEventEnd(MAT_SolveTranspose,mat,b,x,0);
3745: PetscObjectStateIncrease((PetscObject)x);
3746: return(0);
3747: }
3749: /*@
3750: MatSolveTransposeAdd - Computes x = y + inv(Transpose(A)) b, given a
3751: factored matrix.
3753: Neighbor-wise Collective on Mat
3755: Input Parameters:
3756: + mat - the factored matrix
3757: . b - the right-hand-side vector
3758: - y - the vector to be added to
3760: Output Parameter:
3761: . x - the result vector
3763: Notes:
3764: The vectors b and x cannot be the same. I.e., one cannot
3765: call MatSolveTransposeAdd(A,x,y,x).
3767: Most users should employ the simplified KSP interface for linear solvers
3768: instead of working directly with matrix algebra routines such as this.
3769: See, e.g., KSPCreate().
3771: Level: developer
3773: .seealso: MatSolve(), MatSolveAdd(), MatSolveTranspose()
3774: @*/
3775: PetscErrorCode MatSolveTransposeAdd(Mat mat,Vec b,Vec y,Vec x)
3776: {
3777: PetscScalar one = 1.0;
3779: Vec tmp;
3790: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3791: 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);
3792: 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);
3793: 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);
3794: 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);
3795: if (!mat->rmap->N && !mat->cmap->N) return(0);
3796: if (!mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
3797: MatCheckPreallocated(mat,1);
3799: PetscLogEventBegin(MAT_SolveTransposeAdd,mat,b,x,y);
3800: if (mat->ops->solvetransposeadd) {
3801: if (mat->factorerrortype) {
3802: PetscInfo1(mat,"MatFactorError %D\n",mat->factorerrortype);
3803: VecSetInf(x);
3804: } else {
3805: (*mat->ops->solvetransposeadd)(mat,b,y,x);
3806: }
3807: } else {
3808: /* do the solve then the add manually */
3809: if (x != y) {
3810: MatSolveTranspose(mat,b,x);
3811: VecAXPY(x,one,y);
3812: } else {
3813: VecDuplicate(x,&tmp);
3814: PetscLogObjectParent((PetscObject)mat,(PetscObject)tmp);
3815: VecCopy(x,tmp);
3816: MatSolveTranspose(mat,b,x);
3817: VecAXPY(x,one,tmp);
3818: VecDestroy(&tmp);
3819: }
3820: }
3821: PetscLogEventEnd(MAT_SolveTransposeAdd,mat,b,x,y);
3822: PetscObjectStateIncrease((PetscObject)x);
3823: return(0);
3824: }
3825: /* ----------------------------------------------------------------*/
3827: /*@
3828: MatSOR - Computes relaxation (SOR, Gauss-Seidel) sweeps.
3830: Neighbor-wise Collective on Mat
3832: Input Parameters:
3833: + mat - the matrix
3834: . b - the right hand side
3835: . omega - the relaxation factor
3836: . flag - flag indicating the type of SOR (see below)
3837: . shift - diagonal shift
3838: . its - the number of iterations
3839: - lits - the number of local iterations
3841: Output Parameters:
3842: . x - the solution (can contain an initial guess, use option SOR_ZERO_INITIAL_GUESS to indicate no guess)
3844: SOR Flags:
3845: + SOR_FORWARD_SWEEP - forward SOR
3846: . SOR_BACKWARD_SWEEP - backward SOR
3847: . SOR_SYMMETRIC_SWEEP - SSOR (symmetric SOR)
3848: . SOR_LOCAL_FORWARD_SWEEP - local forward SOR
3849: . SOR_LOCAL_BACKWARD_SWEEP - local forward SOR
3850: . SOR_LOCAL_SYMMETRIC_SWEEP - local SSOR
3851: . SOR_APPLY_UPPER, SOR_APPLY_LOWER - applies
3852: upper/lower triangular part of matrix to
3853: vector (with omega)
3854: - SOR_ZERO_INITIAL_GUESS - zero initial guess
3856: Notes:
3857: SOR_LOCAL_FORWARD_SWEEP, SOR_LOCAL_BACKWARD_SWEEP, and
3858: SOR_LOCAL_SYMMETRIC_SWEEP perform separate independent smoothings
3859: on each processor.
3861: Application programmers will not generally use MatSOR() directly,
3862: but instead will employ the KSP/PC interface.
3864: Notes:
3865: for BAIJ, SBAIJ, and AIJ matrices with Inodes this does a block SOR smoothing, otherwise it does a pointwise smoothing
3867: Notes for Advanced Users:
3868: The flags are implemented as bitwise inclusive or operations.
3869: For example, use (SOR_ZERO_INITIAL_GUESS | SOR_SYMMETRIC_SWEEP)
3870: to specify a zero initial guess for SSOR.
3872: Most users should employ the simplified KSP interface for linear solvers
3873: instead of working directly with matrix algebra routines such as this.
3874: See, e.g., KSPCreate().
3876: Vectors x and b CANNOT be the same
3878: Developer Note: We should add block SOR support for AIJ matrices with block size set to great than one and no inodes
3880: Level: developer
3882: @*/
3883: PetscErrorCode MatSOR(Mat mat,Vec b,PetscReal omega,MatSORType flag,PetscReal shift,PetscInt its,PetscInt lits,Vec x)
3884: {
3894: if (!mat->ops->sor) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3895: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3896: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3897: 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);
3898: 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);
3899: 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);
3900: if (its <= 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Relaxation requires global its %D positive",its);
3901: if (lits <= 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Relaxation requires local its %D positive",lits);
3902: if (b == x) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_IDN,"b and x vector cannot be the same");
3904: MatCheckPreallocated(mat,1);
3905: PetscLogEventBegin(MAT_SOR,mat,b,x,0);
3906: ierr =(*mat->ops->sor)(mat,b,omega,flag,shift,its,lits,x);
3907: PetscLogEventEnd(MAT_SOR,mat,b,x,0);
3908: PetscObjectStateIncrease((PetscObject)x);
3909: return(0);
3910: }
3912: /*
3913: Default matrix copy routine.
3914: */
3915: PetscErrorCode MatCopy_Basic(Mat A,Mat B,MatStructure str)
3916: {
3917: PetscErrorCode ierr;
3918: PetscInt i,rstart = 0,rend = 0,nz;
3919: const PetscInt *cwork;
3920: const PetscScalar *vwork;
3923: if (B->assembled) {
3924: MatZeroEntries(B);
3925: }
3926: if (str == SAME_NONZERO_PATTERN) {
3927: MatGetOwnershipRange(A,&rstart,&rend);
3928: for (i=rstart; i<rend; i++) {
3929: MatGetRow(A,i,&nz,&cwork,&vwork);
3930: MatSetValues(B,1,&i,nz,cwork,vwork,INSERT_VALUES);
3931: MatRestoreRow(A,i,&nz,&cwork,&vwork);
3932: }
3933: } else {
3934: MatAYPX(B,0.0,A,str);
3935: }
3936: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
3937: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
3938: return(0);
3939: }
3941: /*@
3942: MatCopy - Copies a matrix to another matrix.
3944: Collective on Mat
3946: Input Parameters:
3947: + A - the matrix
3948: - str - SAME_NONZERO_PATTERN or DIFFERENT_NONZERO_PATTERN
3950: Output Parameter:
3951: . B - where the copy is put
3953: Notes:
3954: If you use SAME_NONZERO_PATTERN then the two matrices had better have the
3955: same nonzero pattern or the routine will crash.
3957: MatCopy() copies the matrix entries of a matrix to another existing
3958: matrix (after first zeroing the second matrix). A related routine is
3959: MatConvert(), which first creates a new matrix and then copies the data.
3961: Level: intermediate
3963: .seealso: MatConvert(), MatDuplicate()
3965: @*/
3966: PetscErrorCode MatCopy(Mat A,Mat B,MatStructure str)
3967: {
3969: PetscInt i;
3977: MatCheckPreallocated(B,2);
3978: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3979: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3980: 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);
3981: MatCheckPreallocated(A,1);
3982: if (A == B) return(0);
3984: PetscLogEventBegin(MAT_Copy,A,B,0,0);
3985: if (A->ops->copy) {
3986: (*A->ops->copy)(A,B,str);
3987: } else { /* generic conversion */
3988: MatCopy_Basic(A,B,str);
3989: }
3991: B->stencil.dim = A->stencil.dim;
3992: B->stencil.noc = A->stencil.noc;
3993: for (i=0; i<=A->stencil.dim; i++) {
3994: B->stencil.dims[i] = A->stencil.dims[i];
3995: B->stencil.starts[i] = A->stencil.starts[i];
3996: }
3998: PetscLogEventEnd(MAT_Copy,A,B,0,0);
3999: PetscObjectStateIncrease((PetscObject)B);
4000: return(0);
4001: }
4003: /*@C
4004: MatConvert - Converts a matrix to another matrix, either of the same
4005: or different type.
4007: Collective on Mat
4009: Input Parameters:
4010: + mat - the matrix
4011: . newtype - new matrix type. Use MATSAME to create a new matrix of the
4012: same type as the original matrix.
4013: - reuse - denotes if the destination matrix is to be created or reused.
4014: 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
4015: 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).
4017: Output Parameter:
4018: . M - pointer to place new matrix
4020: Notes:
4021: MatConvert() first creates a new matrix and then copies the data from
4022: the first matrix. A related routine is MatCopy(), which copies the matrix
4023: entries of one matrix to another already existing matrix context.
4025: Cannot be used to convert a sequential matrix to parallel or parallel to sequential,
4026: the MPI communicator of the generated matrix is always the same as the communicator
4027: of the input matrix.
4029: Level: intermediate
4031: .seealso: MatCopy(), MatDuplicate()
4032: @*/
4033: PetscErrorCode MatConvert(Mat mat, MatType newtype,MatReuse reuse,Mat *M)
4034: {
4036: PetscBool sametype,issame,flg;
4037: char convname[256],mtype[256];
4038: Mat B;
4044: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4045: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4046: MatCheckPreallocated(mat,1);
4048: PetscOptionsGetString(((PetscObject)mat)->options,((PetscObject)mat)->prefix,"-matconvert_type",mtype,256,&flg);
4049: if (flg) newtype = mtype;
4051: PetscObjectTypeCompare((PetscObject)mat,newtype,&sametype);
4052: PetscStrcmp(newtype,"same",&issame);
4053: if ((reuse == MAT_INPLACE_MATRIX) && (mat != *M)) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"MAT_INPLACE_MATRIX requires same input and output matrix");
4054: 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");
4056: if ((reuse == MAT_INPLACE_MATRIX) && (issame || sametype)) {
4057: PetscInfo3(mat,"Early return for inplace %s %d %d\n",((PetscObject)mat)->type_name,sametype,issame);
4058: return(0);
4059: }
4061: if ((sametype || issame) && (reuse==MAT_INITIAL_MATRIX) && mat->ops->duplicate) {
4062: PetscInfo3(mat,"Calling duplicate for initial matrix %s %d %d\n",((PetscObject)mat)->type_name,sametype,issame);
4063: (*mat->ops->duplicate)(mat,MAT_COPY_VALUES,M);
4064: } else {
4065: PetscErrorCode (*conv)(Mat, MatType,MatReuse,Mat*)=NULL;
4066: const char *prefix[3] = {"seq","mpi",""};
4067: PetscInt i;
4068: /*
4069: Order of precedence:
4070: 0) See if newtype is a superclass of the current matrix.
4071: 1) See if a specialized converter is known to the current matrix.
4072: 2) See if a specialized converter is known to the desired matrix class.
4073: 3) See if a good general converter is registered for the desired class
4074: (as of 6/27/03 only MATMPIADJ falls into this category).
4075: 4) See if a good general converter is known for the current matrix.
4076: 5) Use a really basic converter.
4077: */
4079: /* 0) See if newtype is a superclass of the current matrix.
4080: i.e mat is mpiaij and newtype is aij */
4081: for (i=0; i<2; i++) {
4082: PetscStrncpy(convname,prefix[i],sizeof(convname));
4083: PetscStrlcat(convname,newtype,sizeof(convname));
4084: PetscStrcmp(convname,((PetscObject)mat)->type_name,&flg);
4085: PetscInfo3(mat,"Check superclass %s %s -> %d\n",convname,((PetscObject)mat)->type_name,flg);
4086: if (flg) {
4087: if (reuse == MAT_INPLACE_MATRIX) {
4088: return(0);
4089: } else if (reuse == MAT_INITIAL_MATRIX && mat->ops->duplicate) {
4090: (*mat->ops->duplicate)(mat,MAT_COPY_VALUES,M);
4091: return(0);
4092: } else if (reuse == MAT_REUSE_MATRIX && mat->ops->copy) {
4093: MatCopy(mat,*M,SAME_NONZERO_PATTERN);
4094: return(0);
4095: }
4096: }
4097: }
4098: /* 1) See if a specialized converter is known to the current matrix and the desired class */
4099: for (i=0; i<3; i++) {
4100: PetscStrncpy(convname,"MatConvert_",sizeof(convname));
4101: PetscStrlcat(convname,((PetscObject)mat)->type_name,sizeof(convname));
4102: PetscStrlcat(convname,"_",sizeof(convname));
4103: PetscStrlcat(convname,prefix[i],sizeof(convname));
4104: PetscStrlcat(convname,issame ? ((PetscObject)mat)->type_name : newtype,sizeof(convname));
4105: PetscStrlcat(convname,"_C",sizeof(convname));
4106: PetscObjectQueryFunction((PetscObject)mat,convname,&conv);
4107: PetscInfo3(mat,"Check specialized (1) %s (%s) -> %d\n",convname,((PetscObject)mat)->type_name,!!conv);
4108: if (conv) goto foundconv;
4109: }
4111: /* 2) See if a specialized converter is known to the desired matrix class. */
4112: MatCreate(PetscObjectComm((PetscObject)mat),&B);
4113: MatSetSizes(B,mat->rmap->n,mat->cmap->n,mat->rmap->N,mat->cmap->N);
4114: MatSetType(B,newtype);
4115: for (i=0; i<3; i++) {
4116: PetscStrncpy(convname,"MatConvert_",sizeof(convname));
4117: PetscStrlcat(convname,((PetscObject)mat)->type_name,sizeof(convname));
4118: PetscStrlcat(convname,"_",sizeof(convname));
4119: PetscStrlcat(convname,prefix[i],sizeof(convname));
4120: PetscStrlcat(convname,newtype,sizeof(convname));
4121: PetscStrlcat(convname,"_C",sizeof(convname));
4122: PetscObjectQueryFunction((PetscObject)B,convname,&conv);
4123: PetscInfo3(mat,"Check specialized (2) %s (%s) -> %d\n",convname,((PetscObject)B)->type_name,!!conv);
4124: if (conv) {
4125: MatDestroy(&B);
4126: goto foundconv;
4127: }
4128: }
4130: /* 3) See if a good general converter is registered for the desired class */
4131: conv = B->ops->convertfrom;
4132: PetscInfo2(mat,"Check convertfrom (%s) -> %d\n",((PetscObject)B)->type_name,!!conv);
4133: MatDestroy(&B);
4134: if (conv) goto foundconv;
4136: /* 4) See if a good general converter is known for the current matrix */
4137: if (mat->ops->convert) conv = mat->ops->convert;
4139: PetscInfo2(mat,"Check general convert (%s) -> %d\n",((PetscObject)mat)->type_name,!!conv);
4140: if (conv) goto foundconv;
4142: /* 5) Use a really basic converter. */
4143: PetscInfo(mat,"Using MatConvert_Basic\n");
4144: conv = MatConvert_Basic;
4146: foundconv:
4147: PetscLogEventBegin(MAT_Convert,mat,0,0,0);
4148: (*conv)(mat,newtype,reuse,M);
4149: if (mat->rmap->mapping && mat->cmap->mapping && !(*M)->rmap->mapping && !(*M)->cmap->mapping) {
4150: /* the block sizes must be same if the mappings are copied over */
4151: (*M)->rmap->bs = mat->rmap->bs;
4152: (*M)->cmap->bs = mat->cmap->bs;
4153: PetscObjectReference((PetscObject)mat->rmap->mapping);
4154: PetscObjectReference((PetscObject)mat->cmap->mapping);
4155: (*M)->rmap->mapping = mat->rmap->mapping;
4156: (*M)->cmap->mapping = mat->cmap->mapping;
4157: }
4158: (*M)->stencil.dim = mat->stencil.dim;
4159: (*M)->stencil.noc = mat->stencil.noc;
4160: for (i=0; i<=mat->stencil.dim; i++) {
4161: (*M)->stencil.dims[i] = mat->stencil.dims[i];
4162: (*M)->stencil.starts[i] = mat->stencil.starts[i];
4163: }
4164: PetscLogEventEnd(MAT_Convert,mat,0,0,0);
4165: }
4166: PetscObjectStateIncrease((PetscObject)*M);
4168: /* Copy Mat options */
4169: if (mat->symmetric) {MatSetOption(*M,MAT_SYMMETRIC,PETSC_TRUE);}
4170: if (mat->hermitian) {MatSetOption(*M,MAT_HERMITIAN,PETSC_TRUE);}
4171: return(0);
4172: }
4174: /*@C
4175: MatFactorGetSolverType - Returns name of the package providing the factorization routines
4177: Not Collective
4179: Input Parameter:
4180: . mat - the matrix, must be a factored matrix
4182: Output Parameter:
4183: . type - the string name of the package (do not free this string)
4185: Notes:
4186: In Fortran you pass in a empty string and the package name will be copied into it.
4187: (Make sure the string is long enough)
4189: Level: intermediate
4191: .seealso: MatCopy(), MatDuplicate(), MatGetFactorAvailable(), MatGetFactor()
4192: @*/
4193: PetscErrorCode MatFactorGetSolverType(Mat mat, MatSolverType *type)
4194: {
4195: PetscErrorCode ierr, (*conv)(Mat,MatSolverType*);
4200: if (!mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Only for factored matrix");
4201: PetscObjectQueryFunction((PetscObject)mat,"MatFactorGetSolverType_C",&conv);
4202: if (!conv) {
4203: *type = MATSOLVERPETSC;
4204: } else {
4205: (*conv)(mat,type);
4206: }
4207: return(0);
4208: }
4210: typedef struct _MatSolverTypeForSpecifcType* MatSolverTypeForSpecifcType;
4211: struct _MatSolverTypeForSpecifcType {
4212: MatType mtype;
4213: PetscErrorCode (*getfactor[4])(Mat,MatFactorType,Mat*);
4214: MatSolverTypeForSpecifcType next;
4215: };
4217: typedef struct _MatSolverTypeHolder* MatSolverTypeHolder;
4218: struct _MatSolverTypeHolder {
4219: char *name;
4220: MatSolverTypeForSpecifcType handlers;
4221: MatSolverTypeHolder next;
4222: };
4224: static MatSolverTypeHolder MatSolverTypeHolders = NULL;
4226: /*@C
4227: MatSolvePackageRegister - Registers a MatSolverType that works for a particular matrix type
4229: Input Parameters:
4230: + package - name of the package, for example petsc or superlu
4231: . mtype - the matrix type that works with this package
4232: . ftype - the type of factorization supported by the package
4233: - getfactor - routine that will create the factored matrix ready to be used
4235: Level: intermediate
4237: .seealso: MatCopy(), MatDuplicate(), MatGetFactorAvailable()
4238: @*/
4239: PetscErrorCode MatSolverTypeRegister(MatSolverType package,MatType mtype,MatFactorType ftype,PetscErrorCode (*getfactor)(Mat,MatFactorType,Mat*))
4240: {
4241: PetscErrorCode ierr;
4242: MatSolverTypeHolder next = MatSolverTypeHolders,prev = NULL;
4243: PetscBool flg;
4244: MatSolverTypeForSpecifcType inext,iprev = NULL;
4247: MatInitializePackage();
4248: if (!next) {
4249: PetscNew(&MatSolverTypeHolders);
4250: PetscStrallocpy(package,&MatSolverTypeHolders->name);
4251: PetscNew(&MatSolverTypeHolders->handlers);
4252: PetscStrallocpy(mtype,(char **)&MatSolverTypeHolders->handlers->mtype);
4253: MatSolverTypeHolders->handlers->getfactor[(int)ftype-1] = getfactor;
4254: return(0);
4255: }
4256: while (next) {
4257: PetscStrcasecmp(package,next->name,&flg);
4258: if (flg) {
4259: if (!next->handlers) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"MatSolverTypeHolder is missing handlers");
4260: inext = next->handlers;
4261: while (inext) {
4262: PetscStrcasecmp(mtype,inext->mtype,&flg);
4263: if (flg) {
4264: inext->getfactor[(int)ftype-1] = getfactor;
4265: return(0);
4266: }
4267: iprev = inext;
4268: inext = inext->next;
4269: }
4270: PetscNew(&iprev->next);
4271: PetscStrallocpy(mtype,(char **)&iprev->next->mtype);
4272: iprev->next->getfactor[(int)ftype-1] = getfactor;
4273: return(0);
4274: }
4275: prev = next;
4276: next = next->next;
4277: }
4278: PetscNew(&prev->next);
4279: PetscStrallocpy(package,&prev->next->name);
4280: PetscNew(&prev->next->handlers);
4281: PetscStrallocpy(mtype,(char **)&prev->next->handlers->mtype);
4282: prev->next->handlers->getfactor[(int)ftype-1] = getfactor;
4283: return(0);
4284: }
4286: /*@C
4287: MatSolvePackageGet - Get's the function that creates the factor matrix if it exist
4289: Input Parameters:
4290: + package - name of the package, for example petsc or superlu
4291: . ftype - the type of factorization supported by the package
4292: - mtype - the matrix type that works with this package
4294: Output Parameters:
4295: + foundpackage - PETSC_TRUE if the package was registered
4296: . foundmtype - PETSC_TRUE if the package supports the requested mtype
4297: - getfactor - routine that will create the factored matrix ready to be used or NULL if not found
4299: Level: intermediate
4301: .seealso: MatCopy(), MatDuplicate(), MatGetFactorAvailable()
4302: @*/
4303: PetscErrorCode MatSolverTypeGet(MatSolverType package,MatType mtype,MatFactorType ftype,PetscBool *foundpackage,PetscBool *foundmtype,PetscErrorCode (**getfactor)(Mat,MatFactorType,Mat*))
4304: {
4305: PetscErrorCode ierr;
4306: MatSolverTypeHolder next = MatSolverTypeHolders;
4307: PetscBool flg;
4308: MatSolverTypeForSpecifcType inext;
4311: if (foundpackage) *foundpackage = PETSC_FALSE;
4312: if (foundmtype) *foundmtype = PETSC_FALSE;
4313: if (getfactor) *getfactor = NULL;
4315: if (package) {
4316: while (next) {
4317: PetscStrcasecmp(package,next->name,&flg);
4318: if (flg) {
4319: if (foundpackage) *foundpackage = PETSC_TRUE;
4320: inext = next->handlers;
4321: while (inext) {
4322: PetscStrbeginswith(mtype,inext->mtype,&flg);
4323: if (flg) {
4324: if (foundmtype) *foundmtype = PETSC_TRUE;
4325: if (getfactor) *getfactor = inext->getfactor[(int)ftype-1];
4326: return(0);
4327: }
4328: inext = inext->next;
4329: }
4330: }
4331: next = next->next;
4332: }
4333: } else {
4334: while (next) {
4335: inext = next->handlers;
4336: while (inext) {
4337: PetscStrbeginswith(mtype,inext->mtype,&flg);
4338: if (flg && inext->getfactor[(int)ftype-1]) {
4339: if (foundpackage) *foundpackage = PETSC_TRUE;
4340: if (foundmtype) *foundmtype = PETSC_TRUE;
4341: if (getfactor) *getfactor = inext->getfactor[(int)ftype-1];
4342: return(0);
4343: }
4344: inext = inext->next;
4345: }
4346: next = next->next;
4347: }
4348: }
4349: return(0);
4350: }
4352: PetscErrorCode MatSolverTypeDestroy(void)
4353: {
4354: PetscErrorCode ierr;
4355: MatSolverTypeHolder next = MatSolverTypeHolders,prev;
4356: MatSolverTypeForSpecifcType inext,iprev;
4359: while (next) {
4360: PetscFree(next->name);
4361: inext = next->handlers;
4362: while (inext) {
4363: PetscFree(inext->mtype);
4364: iprev = inext;
4365: inext = inext->next;
4366: PetscFree(iprev);
4367: }
4368: prev = next;
4369: next = next->next;
4370: PetscFree(prev);
4371: }
4372: MatSolverTypeHolders = NULL;
4373: return(0);
4374: }
4376: /*@C
4377: MatGetFactor - Returns a matrix suitable to calls to MatXXFactorSymbolic()
4379: Collective on Mat
4381: Input Parameters:
4382: + mat - the matrix
4383: . type - name of solver type, for example, superlu, petsc (to use PETSc's default)
4384: - ftype - factor type, MAT_FACTOR_LU, MAT_FACTOR_CHOLESKY, MAT_FACTOR_ICC, MAT_FACTOR_ILU,
4386: Output Parameters:
4387: . f - the factor matrix used with MatXXFactorSymbolic() calls
4389: Notes:
4390: Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
4391: such as pastix, superlu, mumps etc.
4393: PETSc must have been ./configure to use the external solver, using the option --download-package
4395: Level: intermediate
4397: .seealso: MatCopy(), MatDuplicate(), MatGetFactorAvailable()
4398: @*/
4399: PetscErrorCode MatGetFactor(Mat mat, MatSolverType type,MatFactorType ftype,Mat *f)
4400: {
4401: PetscErrorCode ierr,(*conv)(Mat,MatFactorType,Mat*);
4402: PetscBool foundpackage,foundmtype;
4408: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4409: MatCheckPreallocated(mat,1);
4411: MatSolverTypeGet(type,((PetscObject)mat)->type_name,ftype,&foundpackage,&foundmtype,&conv);
4412: if (!foundpackage) {
4413: if (type) {
4414: SETERRQ4(PetscObjectComm((PetscObject)mat),PETSC_ERR_MISSING_FACTOR,"Could not locate solver package %s for factorization type %s and matrix type %s. Perhaps you must ./configure with --download-%s",type,MatFactorTypes[ftype],((PetscObject)mat)->type_name,type);
4415: } else {
4416: SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_MISSING_FACTOR,"Could not locate a solver package for factorization type %s and matrix type %s.",MatFactorTypes[ftype],((PetscObject)mat)->type_name);
4417: }
4418: }
4419: if (!foundmtype) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_MISSING_FACTOR,"MatSolverType %s does not support matrix type %s",type,((PetscObject)mat)->type_name);
4420: 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);
4422: #if defined(PETSC_USE_COMPLEX)
4423: if (mat->hermitian && !mat->symmetric && (ftype == MAT_FACTOR_CHOLESKY||ftype == MAT_FACTOR_ICC)) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Hermitian CHOLESKY or ICC Factor is not supported");
4424: #endif
4426: (*conv)(mat,ftype,f);
4427: return(0);
4428: }
4430: /*@C
4431: MatGetFactorAvailable - Returns a a flag if matrix supports particular package and factor type
4433: Not Collective
4435: Input Parameters:
4436: + mat - the matrix
4437: . type - name of solver type, for example, superlu, petsc (to use PETSc's default)
4438: - ftype - factor type, MAT_FACTOR_LU, MAT_FACTOR_CHOLESKY, MAT_FACTOR_ICC, MAT_FACTOR_ILU,
4440: Output Parameter:
4441: . flg - PETSC_TRUE if the factorization is available
4443: Notes:
4444: Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
4445: such as pastix, superlu, mumps etc.
4447: PETSc must have been ./configure to use the external solver, using the option --download-package
4449: Level: intermediate
4451: .seealso: MatCopy(), MatDuplicate(), MatGetFactor()
4452: @*/
4453: PetscErrorCode MatGetFactorAvailable(Mat mat, MatSolverType type,MatFactorType ftype,PetscBool *flg)
4454: {
4455: PetscErrorCode ierr, (*gconv)(Mat,MatFactorType,Mat*);
4461: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4462: MatCheckPreallocated(mat,1);
4464: *flg = PETSC_FALSE;
4465: MatSolverTypeGet(type,((PetscObject)mat)->type_name,ftype,NULL,NULL,&gconv);
4466: if (gconv) {
4467: *flg = PETSC_TRUE;
4468: }
4469: return(0);
4470: }
4472: #include <petscdmtypes.h>
4474: /*@
4475: MatDuplicate - Duplicates a matrix including the non-zero structure.
4477: Collective on Mat
4479: Input Parameters:
4480: + mat - the matrix
4481: - op - One of MAT_DO_NOT_COPY_VALUES, MAT_COPY_VALUES, or MAT_SHARE_NONZERO_PATTERN.
4482: See the manual page for MatDuplicateOption for an explanation of these options.
4484: Output Parameter:
4485: . M - pointer to place new matrix
4487: Level: intermediate
4489: Notes:
4490: You cannot change the nonzero pattern for the parent or child matrix if you use MAT_SHARE_NONZERO_PATTERN.
4491: 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.
4493: .seealso: MatCopy(), MatConvert(), MatDuplicateOption
4494: @*/
4495: PetscErrorCode MatDuplicate(Mat mat,MatDuplicateOption op,Mat *M)
4496: {
4498: Mat B;
4499: PetscInt i;
4500: DM dm;
4501: void (*viewf)(void);
4507: if (op == MAT_COPY_VALUES && !mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"MAT_COPY_VALUES not allowed for unassembled matrix");
4508: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4509: MatCheckPreallocated(mat,1);
4511: *M = 0;
4512: if (!mat->ops->duplicate) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Not written for matrix type %s\n",((PetscObject)mat)->type_name);
4513: PetscLogEventBegin(MAT_Convert,mat,0,0,0);
4514: (*mat->ops->duplicate)(mat,op,M);
4515: B = *M;
4517: MatGetOperation(mat,MATOP_VIEW,&viewf);
4518: if (viewf) {
4519: MatSetOperation(B,MATOP_VIEW,viewf);
4520: }
4522: B->stencil.dim = mat->stencil.dim;
4523: B->stencil.noc = mat->stencil.noc;
4524: for (i=0; i<=mat->stencil.dim; i++) {
4525: B->stencil.dims[i] = mat->stencil.dims[i];
4526: B->stencil.starts[i] = mat->stencil.starts[i];
4527: }
4529: B->nooffproczerorows = mat->nooffproczerorows;
4530: B->nooffprocentries = mat->nooffprocentries;
4532: PetscObjectQuery((PetscObject) mat, "__PETSc_dm", (PetscObject*) &dm);
4533: if (dm) {
4534: PetscObjectCompose((PetscObject) B, "__PETSc_dm", (PetscObject) dm);
4535: }
4536: PetscLogEventEnd(MAT_Convert,mat,0,0,0);
4537: PetscObjectStateIncrease((PetscObject)B);
4538: return(0);
4539: }
4541: /*@
4542: MatGetDiagonal - Gets the diagonal of a matrix.
4544: Logically Collective on Mat
4546: Input Parameters:
4547: + mat - the matrix
4548: - v - the vector for storing the diagonal
4550: Output Parameter:
4551: . v - the diagonal of the matrix
4553: Level: intermediate
4555: Note:
4556: Currently only correct in parallel for square matrices.
4558: .seealso: MatGetRow(), MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRowMaxAbs()
4559: @*/
4560: PetscErrorCode MatGetDiagonal(Mat mat,Vec v)
4561: {
4568: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4569: if (!mat->ops->getdiagonal) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4570: MatCheckPreallocated(mat,1);
4572: (*mat->ops->getdiagonal)(mat,v);
4573: PetscObjectStateIncrease((PetscObject)v);
4574: return(0);
4575: }
4577: /*@C
4578: MatGetRowMin - Gets the minimum value (of the real part) of each
4579: row of the matrix
4581: Logically Collective on Mat
4583: Input Parameters:
4584: . mat - the matrix
4586: Output Parameter:
4587: + v - the vector for storing the maximums
4588: - idx - the indices of the column found for each row (optional)
4590: Level: intermediate
4592: Notes:
4593: The result of this call are the same as if one converted the matrix to dense format
4594: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
4596: This code is only implemented for a couple of matrix formats.
4598: .seealso: MatGetDiagonal(), MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRowMaxAbs(),
4599: MatGetRowMax()
4600: @*/
4601: PetscErrorCode MatGetRowMin(Mat mat,Vec v,PetscInt idx[])
4602: {
4609: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4610: if (!mat->ops->getrowmax) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4611: MatCheckPreallocated(mat,1);
4613: (*mat->ops->getrowmin)(mat,v,idx);
4614: PetscObjectStateIncrease((PetscObject)v);
4615: return(0);
4616: }
4618: /*@C
4619: MatGetRowMinAbs - Gets the minimum value (in absolute value) of each
4620: row of the matrix
4622: Logically Collective on Mat
4624: Input Parameters:
4625: . mat - the matrix
4627: Output Parameter:
4628: + v - the vector for storing the minimums
4629: - idx - the indices of the column found for each row (or NULL if not needed)
4631: Level: intermediate
4633: Notes:
4634: if a row is completely empty or has only 0.0 values then the idx[] value for that
4635: row is 0 (the first column).
4637: This code is only implemented for a couple of matrix formats.
4639: .seealso: MatGetDiagonal(), MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRowMax(), MatGetRowMaxAbs(), MatGetRowMin()
4640: @*/
4641: PetscErrorCode MatGetRowMinAbs(Mat mat,Vec v,PetscInt idx[])
4642: {
4649: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4650: if (!mat->ops->getrowminabs) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4651: MatCheckPreallocated(mat,1);
4652: if (idx) {PetscArrayzero(idx,mat->rmap->n);}
4654: (*mat->ops->getrowminabs)(mat,v,idx);
4655: PetscObjectStateIncrease((PetscObject)v);
4656: return(0);
4657: }
4659: /*@C
4660: MatGetRowMax - Gets the maximum value (of the real part) of each
4661: row of the matrix
4663: Logically Collective on Mat
4665: Input Parameters:
4666: . mat - the matrix
4668: Output Parameter:
4669: + v - the vector for storing the maximums
4670: - idx - the indices of the column found for each row (optional)
4672: Level: intermediate
4674: Notes:
4675: The result of this call are the same as if one converted the matrix to dense format
4676: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
4678: This code is only implemented for a couple of matrix formats.
4680: .seealso: MatGetDiagonal(), MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRowMaxAbs(), MatGetRowMin()
4681: @*/
4682: PetscErrorCode MatGetRowMax(Mat mat,Vec v,PetscInt idx[])
4683: {
4690: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4691: if (!mat->ops->getrowmax) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4692: MatCheckPreallocated(mat,1);
4694: (*mat->ops->getrowmax)(mat,v,idx);
4695: PetscObjectStateIncrease((PetscObject)v);
4696: return(0);
4697: }
4699: /*@C
4700: MatGetRowMaxAbs - Gets the maximum value (in absolute value) of each
4701: row of the matrix
4703: Logically Collective on Mat
4705: Input Parameters:
4706: . mat - the matrix
4708: Output Parameter:
4709: + v - the vector for storing the maximums
4710: - idx - the indices of the column found for each row (or NULL if not needed)
4712: Level: intermediate
4714: Notes:
4715: if a row is completely empty or has only 0.0 values then the idx[] value for that
4716: row is 0 (the first column).
4718: This code is only implemented for a couple of matrix formats.
4720: .seealso: MatGetDiagonal(), MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRowMax(), MatGetRowMin()
4721: @*/
4722: PetscErrorCode MatGetRowMaxAbs(Mat mat,Vec v,PetscInt idx[])
4723: {
4730: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4731: if (!mat->ops->getrowmaxabs) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4732: MatCheckPreallocated(mat,1);
4733: if (idx) {PetscArrayzero(idx,mat->rmap->n);}
4735: (*mat->ops->getrowmaxabs)(mat,v,idx);
4736: PetscObjectStateIncrease((PetscObject)v);
4737: return(0);
4738: }
4740: /*@
4741: MatGetRowSum - Gets the sum of each row of the matrix
4743: Logically or Neighborhood Collective on Mat
4745: Input Parameters:
4746: . mat - the matrix
4748: Output Parameter:
4749: . v - the vector for storing the sum of rows
4751: Level: intermediate
4753: Notes:
4754: This code is slow since it is not currently specialized for different formats
4756: .seealso: MatGetDiagonal(), MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRowMax(), MatGetRowMin()
4757: @*/
4758: PetscErrorCode MatGetRowSum(Mat mat, Vec v)
4759: {
4760: Vec ones;
4767: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4768: MatCheckPreallocated(mat,1);
4769: MatCreateVecs(mat,&ones,NULL);
4770: VecSet(ones,1.);
4771: MatMult(mat,ones,v);
4772: VecDestroy(&ones);
4773: return(0);
4774: }
4776: /*@
4777: MatTranspose - Computes an in-place or out-of-place transpose of a matrix.
4779: Collective on Mat
4781: Input Parameter:
4782: + mat - the matrix to transpose
4783: - reuse - either MAT_INITIAL_MATRIX, MAT_REUSE_MATRIX, or MAT_INPLACE_MATRIX
4785: Output Parameters:
4786: . B - the transpose
4788: Notes:
4789: If you use MAT_INPLACE_MATRIX then you must pass in &mat for B
4791: MAT_REUSE_MATRIX causes the B matrix from a previous call to this function with MAT_INITIAL_MATRIX to be used
4793: Consider using MatCreateTranspose() instead if you only need a matrix that behaves like the transpose, but don't need the storage to be changed.
4795: Level: intermediate
4797: .seealso: MatMultTranspose(), MatMultTransposeAdd(), MatIsTranspose(), MatReuse
4798: @*/
4799: PetscErrorCode MatTranspose(Mat mat,MatReuse reuse,Mat *B)
4800: {
4806: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4807: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4808: if (!mat->ops->transpose) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4809: if (reuse == MAT_INPLACE_MATRIX && mat != *B) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"MAT_INPLACE_MATRIX requires last matrix to match first");
4810: if (reuse == MAT_REUSE_MATRIX && mat == *B) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Perhaps you mean MAT_INPLACE_MATRIX");
4811: MatCheckPreallocated(mat,1);
4813: PetscLogEventBegin(MAT_Transpose,mat,0,0,0);
4814: (*mat->ops->transpose)(mat,reuse,B);
4815: PetscLogEventEnd(MAT_Transpose,mat,0,0,0);
4816: if (B) {PetscObjectStateIncrease((PetscObject)*B);}
4817: return(0);
4818: }
4820: /*@
4821: MatIsTranspose - Test whether a matrix is another one's transpose,
4822: or its own, in which case it tests symmetry.
4824: Collective on Mat
4826: Input Parameter:
4827: + A - the matrix to test
4828: - B - the matrix to test against, this can equal the first parameter
4830: Output Parameters:
4831: . flg - the result
4833: Notes:
4834: Only available for SeqAIJ/MPIAIJ matrices. The sequential algorithm
4835: has a running time of the order of the number of nonzeros; the parallel
4836: test involves parallel copies of the block-offdiagonal parts of the matrix.
4838: Level: intermediate
4840: .seealso: MatTranspose(), MatIsSymmetric(), MatIsHermitian()
4841: @*/
4842: PetscErrorCode MatIsTranspose(Mat A,Mat B,PetscReal tol,PetscBool *flg)
4843: {
4844: PetscErrorCode ierr,(*f)(Mat,Mat,PetscReal,PetscBool*),(*g)(Mat,Mat,PetscReal,PetscBool*);
4850: PetscObjectQueryFunction((PetscObject)A,"MatIsTranspose_C",&f);
4851: PetscObjectQueryFunction((PetscObject)B,"MatIsTranspose_C",&g);
4852: *flg = PETSC_FALSE;
4853: if (f && g) {
4854: if (f == g) {
4855: (*f)(A,B,tol,flg);
4856: } else SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_NOTSAMETYPE,"Matrices do not have the same comparator for symmetry test");
4857: } else {
4858: MatType mattype;
4859: if (!f) {
4860: MatGetType(A,&mattype);
4861: } else {
4862: MatGetType(B,&mattype);
4863: }
4864: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix of type %s does not support checking for transpose",mattype);
4865: }
4866: return(0);
4867: }
4869: /*@
4870: MatHermitianTranspose - Computes an in-place or out-of-place transpose of a matrix in complex conjugate.
4872: Collective on Mat
4874: Input Parameter:
4875: + mat - the matrix to transpose and complex conjugate
4876: - reuse - MAT_INITIAL_MATRIX to create a new matrix, MAT_INPLACE_MATRIX to reuse the first argument to store the transpose
4878: Output Parameters:
4879: . B - the Hermitian
4881: Level: intermediate
4883: .seealso: MatTranspose(), MatMultTranspose(), MatMultTransposeAdd(), MatIsTranspose(), MatReuse
4884: @*/
4885: PetscErrorCode MatHermitianTranspose(Mat mat,MatReuse reuse,Mat *B)
4886: {
4890: MatTranspose(mat,reuse,B);
4891: #if defined(PETSC_USE_COMPLEX)
4892: MatConjugate(*B);
4893: #endif
4894: return(0);
4895: }
4897: /*@
4898: MatIsHermitianTranspose - Test whether a matrix is another one's Hermitian transpose,
4900: Collective on Mat
4902: Input Parameter:
4903: + A - the matrix to test
4904: - B - the matrix to test against, this can equal the first parameter
4906: Output Parameters:
4907: . flg - the result
4909: Notes:
4910: Only available for SeqAIJ/MPIAIJ matrices. The sequential algorithm
4911: has a running time of the order of the number of nonzeros; the parallel
4912: test involves parallel copies of the block-offdiagonal parts of the matrix.
4914: Level: intermediate
4916: .seealso: MatTranspose(), MatIsSymmetric(), MatIsHermitian(), MatIsTranspose()
4917: @*/
4918: PetscErrorCode MatIsHermitianTranspose(Mat A,Mat B,PetscReal tol,PetscBool *flg)
4919: {
4920: PetscErrorCode ierr,(*f)(Mat,Mat,PetscReal,PetscBool*),(*g)(Mat,Mat,PetscReal,PetscBool*);
4926: PetscObjectQueryFunction((PetscObject)A,"MatIsHermitianTranspose_C",&f);
4927: PetscObjectQueryFunction((PetscObject)B,"MatIsHermitianTranspose_C",&g);
4928: if (f && g) {
4929: if (f==g) {
4930: (*f)(A,B,tol,flg);
4931: } else SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_NOTSAMETYPE,"Matrices do not have the same comparator for Hermitian test");
4932: }
4933: return(0);
4934: }
4936: /*@
4937: MatPermute - Creates a new matrix with rows and columns permuted from the
4938: original.
4940: Collective on Mat
4942: Input Parameters:
4943: + mat - the matrix to permute
4944: . row - row permutation, each processor supplies only the permutation for its rows
4945: - col - column permutation, each processor supplies only the permutation for its columns
4947: Output Parameters:
4948: . B - the permuted matrix
4950: Level: advanced
4952: Note:
4953: The index sets map from row/col of permuted matrix to row/col of original matrix.
4954: The index sets should be on the same communicator as Mat and have the same local sizes.
4956: .seealso: MatGetOrdering(), ISAllGather()
4958: @*/
4959: PetscErrorCode MatPermute(Mat mat,IS row,IS col,Mat *B)
4960: {
4969: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4970: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4971: if (!mat->ops->permute) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"MatPermute not available for Mat type %s",((PetscObject)mat)->type_name);
4972: MatCheckPreallocated(mat,1);
4974: (*mat->ops->permute)(mat,row,col,B);
4975: PetscObjectStateIncrease((PetscObject)*B);
4976: return(0);
4977: }
4979: /*@
4980: MatEqual - Compares two matrices.
4982: Collective on Mat
4984: Input Parameters:
4985: + A - the first matrix
4986: - B - the second matrix
4988: Output Parameter:
4989: . flg - PETSC_TRUE if the matrices are equal; PETSC_FALSE otherwise.
4991: Level: intermediate
4993: @*/
4994: PetscErrorCode MatEqual(Mat A,Mat B,PetscBool *flg)
4995: {
5005: MatCheckPreallocated(B,2);
5006: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5007: if (!B->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5008: 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);
5009: if (!A->ops->equal) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Mat type %s",((PetscObject)A)->type_name);
5010: if (!B->ops->equal) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Mat type %s",((PetscObject)B)->type_name);
5011: 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);
5012: MatCheckPreallocated(A,1);
5014: (*A->ops->equal)(A,B,flg);
5015: return(0);
5016: }
5018: /*@
5019: MatDiagonalScale - Scales a matrix on the left and right by diagonal
5020: matrices that are stored as vectors. Either of the two scaling
5021: matrices can be NULL.
5023: Collective on Mat
5025: Input Parameters:
5026: + mat - the matrix to be scaled
5027: . l - the left scaling vector (or NULL)
5028: - r - the right scaling vector (or NULL)
5030: Notes:
5031: MatDiagonalScale() computes A = LAR, where
5032: L = a diagonal matrix (stored as a vector), R = a diagonal matrix (stored as a vector)
5033: The L scales the rows of the matrix, the R scales the columns of the matrix.
5035: Level: intermediate
5038: .seealso: MatScale(), MatShift(), MatDiagonalSet()
5039: @*/
5040: PetscErrorCode MatDiagonalScale(Mat mat,Vec l,Vec r)
5041: {
5047: if (!mat->ops->diagonalscale) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5050: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5051: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5052: MatCheckPreallocated(mat,1);
5054: PetscLogEventBegin(MAT_Scale,mat,0,0,0);
5055: (*mat->ops->diagonalscale)(mat,l,r);
5056: PetscLogEventEnd(MAT_Scale,mat,0,0,0);
5057: PetscObjectStateIncrease((PetscObject)mat);
5058: return(0);
5059: }
5061: /*@
5062: MatScale - Scales all elements of a matrix by a given number.
5064: Logically Collective on Mat
5066: Input Parameters:
5067: + mat - the matrix to be scaled
5068: - a - the scaling value
5070: Output Parameter:
5071: . mat - the scaled matrix
5073: Level: intermediate
5075: .seealso: MatDiagonalScale()
5076: @*/
5077: PetscErrorCode MatScale(Mat mat,PetscScalar a)
5078: {
5084: if (a != (PetscScalar)1.0 && !mat->ops->scale) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5085: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5086: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5088: MatCheckPreallocated(mat,1);
5090: PetscLogEventBegin(MAT_Scale,mat,0,0,0);
5091: if (a != (PetscScalar)1.0) {
5092: (*mat->ops->scale)(mat,a);
5093: PetscObjectStateIncrease((PetscObject)mat);
5094: }
5095: PetscLogEventEnd(MAT_Scale,mat,0,0,0);
5096: return(0);
5097: }
5099: /*@
5100: MatNorm - Calculates various norms of a matrix.
5102: Collective on Mat
5104: Input Parameters:
5105: + mat - the matrix
5106: - type - the type of norm, NORM_1, NORM_FROBENIUS, NORM_INFINITY
5108: Output Parameters:
5109: . nrm - the resulting norm
5111: Level: intermediate
5113: @*/
5114: PetscErrorCode MatNorm(Mat mat,NormType type,PetscReal *nrm)
5115: {
5123: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5124: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5125: if (!mat->ops->norm) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5126: MatCheckPreallocated(mat,1);
5128: (*mat->ops->norm)(mat,type,nrm);
5129: return(0);
5130: }
5132: /*
5133: This variable is used to prevent counting of MatAssemblyBegin() that
5134: are called from within a MatAssemblyEnd().
5135: */
5136: static PetscInt MatAssemblyEnd_InUse = 0;
5137: /*@
5138: MatAssemblyBegin - Begins assembling the matrix. This routine should
5139: be called after completing all calls to MatSetValues().
5141: Collective on Mat
5143: Input Parameters:
5144: + mat - the matrix
5145: - type - type of assembly, either MAT_FLUSH_ASSEMBLY or MAT_FINAL_ASSEMBLY
5147: Notes:
5148: MatSetValues() generally caches the values. The matrix is ready to
5149: use only after MatAssemblyBegin() and MatAssemblyEnd() have been called.
5150: Use MAT_FLUSH_ASSEMBLY when switching between ADD_VALUES and INSERT_VALUES
5151: in MatSetValues(); use MAT_FINAL_ASSEMBLY for the final assembly before
5152: using the matrix.
5154: ALL processes that share a matrix MUST call MatAssemblyBegin() and MatAssemblyEnd() the SAME NUMBER of times, and each time with the
5155: 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
5156: a global collective operation requring all processes that share the matrix.
5158: Space for preallocated nonzeros that is not filled by a call to MatSetValues() or a related routine are compressed
5159: out by assembly. If you intend to use that extra space on a subsequent assembly, be sure to insert explicit zeros
5160: before MAT_FINAL_ASSEMBLY so the space is not compressed out.
5162: Level: beginner
5164: .seealso: MatAssemblyEnd(), MatSetValues(), MatAssembled()
5165: @*/
5166: PetscErrorCode MatAssemblyBegin(Mat mat,MatAssemblyType type)
5167: {
5173: MatCheckPreallocated(mat,1);
5174: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix.\nDid you forget to call MatSetUnfactored()?");
5175: if (mat->assembled) {
5176: mat->was_assembled = PETSC_TRUE;
5177: mat->assembled = PETSC_FALSE;
5178: }
5180: if (!MatAssemblyEnd_InUse) {
5181: PetscLogEventBegin(MAT_AssemblyBegin,mat,0,0,0);
5182: if (mat->ops->assemblybegin) {(*mat->ops->assemblybegin)(mat,type);}
5183: PetscLogEventEnd(MAT_AssemblyBegin,mat,0,0,0);
5184: } else if (mat->ops->assemblybegin) {
5185: (*mat->ops->assemblybegin)(mat,type);
5186: }
5187: return(0);
5188: }
5190: /*@
5191: MatAssembled - Indicates if a matrix has been assembled and is ready for
5192: use; for example, in matrix-vector product.
5194: Not Collective
5196: Input Parameter:
5197: . mat - the matrix
5199: Output Parameter:
5200: . assembled - PETSC_TRUE or PETSC_FALSE
5202: Level: advanced
5204: .seealso: MatAssemblyEnd(), MatSetValues(), MatAssemblyBegin()
5205: @*/
5206: PetscErrorCode MatAssembled(Mat mat,PetscBool *assembled)
5207: {
5211: *assembled = mat->assembled;
5212: return(0);
5213: }
5215: /*@
5216: MatAssemblyEnd - Completes assembling the matrix. This routine should
5217: be called after MatAssemblyBegin().
5219: Collective on Mat
5221: Input Parameters:
5222: + mat - the matrix
5223: - type - type of assembly, either MAT_FLUSH_ASSEMBLY or MAT_FINAL_ASSEMBLY
5225: Options Database Keys:
5226: + -mat_view ::ascii_info - Prints info on matrix at conclusion of MatEndAssembly()
5227: . -mat_view ::ascii_info_detail - Prints more detailed info
5228: . -mat_view - Prints matrix in ASCII format
5229: . -mat_view ::ascii_matlab - Prints matrix in Matlab format
5230: . -mat_view draw - PetscDraws nonzero structure of matrix, using MatView() and PetscDrawOpenX().
5231: . -display <name> - Sets display name (default is host)
5232: . -draw_pause <sec> - Sets number of seconds to pause after display
5233: . -mat_view socket - Sends matrix to socket, can be accessed from Matlab (See Users-Manual: Chapter 12 Using MATLAB with PETSc )
5234: . -viewer_socket_machine <machine> - Machine to use for socket
5235: . -viewer_socket_port <port> - Port number to use for socket
5236: - -mat_view binary:filename[:append] - Save matrix to file in binary format
5238: Notes:
5239: MatSetValues() generally caches the values. The matrix is ready to
5240: use only after MatAssemblyBegin() and MatAssemblyEnd() have been called.
5241: Use MAT_FLUSH_ASSEMBLY when switching between ADD_VALUES and INSERT_VALUES
5242: in MatSetValues(); use MAT_FINAL_ASSEMBLY for the final assembly before
5243: using the matrix.
5245: Space for preallocated nonzeros that is not filled by a call to MatSetValues() or a related routine are compressed
5246: out by assembly. If you intend to use that extra space on a subsequent assembly, be sure to insert explicit zeros
5247: before MAT_FINAL_ASSEMBLY so the space is not compressed out.
5249: Level: beginner
5251: .seealso: MatAssemblyBegin(), MatSetValues(), PetscDrawOpenX(), PetscDrawCreate(), MatView(), MatAssembled(), PetscViewerSocketOpen()
5252: @*/
5253: PetscErrorCode MatAssemblyEnd(Mat mat,MatAssemblyType type)
5254: {
5255: PetscErrorCode ierr;
5256: static PetscInt inassm = 0;
5257: PetscBool flg = PETSC_FALSE;
5263: inassm++;
5264: MatAssemblyEnd_InUse++;
5265: if (MatAssemblyEnd_InUse == 1) { /* Do the logging only the first time through */
5266: PetscLogEventBegin(MAT_AssemblyEnd,mat,0,0,0);
5267: if (mat->ops->assemblyend) {
5268: (*mat->ops->assemblyend)(mat,type);
5269: }
5270: PetscLogEventEnd(MAT_AssemblyEnd,mat,0,0,0);
5271: } else if (mat->ops->assemblyend) {
5272: (*mat->ops->assemblyend)(mat,type);
5273: }
5275: /* Flush assembly is not a true assembly */
5276: if (type != MAT_FLUSH_ASSEMBLY) {
5277: mat->num_ass++;
5278: mat->assembled = PETSC_TRUE;
5279: mat->ass_nonzerostate = mat->nonzerostate;
5280: }
5282: mat->insertmode = NOT_SET_VALUES;
5283: MatAssemblyEnd_InUse--;
5284: PetscObjectStateIncrease((PetscObject)mat);
5285: if (!mat->symmetric_eternal) {
5286: mat->symmetric_set = PETSC_FALSE;
5287: mat->hermitian_set = PETSC_FALSE;
5288: mat->structurally_symmetric_set = PETSC_FALSE;
5289: }
5290: if (inassm == 1 && type != MAT_FLUSH_ASSEMBLY) {
5291: MatViewFromOptions(mat,NULL,"-mat_view");
5293: if (mat->checksymmetryonassembly) {
5294: MatIsSymmetric(mat,mat->checksymmetrytol,&flg);
5295: if (flg) {
5296: PetscPrintf(PetscObjectComm((PetscObject)mat),"Matrix is symmetric (tolerance %g)\n",(double)mat->checksymmetrytol);
5297: } else {
5298: PetscPrintf(PetscObjectComm((PetscObject)mat),"Matrix is not symmetric (tolerance %g)\n",(double)mat->checksymmetrytol);
5299: }
5300: }
5301: if (mat->nullsp && mat->checknullspaceonassembly) {
5302: MatNullSpaceTest(mat->nullsp,mat,NULL);
5303: }
5304: }
5305: inassm--;
5306: return(0);
5307: }
5309: /*@
5310: MatSetOption - Sets a parameter option for a matrix. Some options
5311: may be specific to certain storage formats. Some options
5312: determine how values will be inserted (or added). Sorted,
5313: row-oriented input will generally assemble the fastest. The default
5314: is row-oriented.
5316: Logically Collective on Mat for certain operations, such as MAT_SPD, not collective for MAT_ROW_ORIENTED, see MatOption
5318: Input Parameters:
5319: + mat - the matrix
5320: . option - the option, one of those listed below (and possibly others),
5321: - flg - turn the option on (PETSC_TRUE) or off (PETSC_FALSE)
5323: Options Describing Matrix Structure:
5324: + MAT_SPD - symmetric positive definite
5325: . MAT_SYMMETRIC - symmetric in terms of both structure and value
5326: . MAT_HERMITIAN - transpose is the complex conjugation
5327: . MAT_STRUCTURALLY_SYMMETRIC - symmetric nonzero structure
5328: - MAT_SYMMETRY_ETERNAL - if you would like the symmetry/Hermitian flag
5329: you set to be kept with all future use of the matrix
5330: including after MatAssemblyBegin/End() which could
5331: potentially change the symmetry structure, i.e. you
5332: KNOW the matrix will ALWAYS have the property you set.
5335: Options For Use with MatSetValues():
5336: Insert a logically dense subblock, which can be
5337: . MAT_ROW_ORIENTED - row-oriented (default)
5339: Note these options reflect the data you pass in with MatSetValues(); it has
5340: nothing to do with how the data is stored internally in the matrix
5341: data structure.
5343: When (re)assembling a matrix, we can restrict the input for
5344: efficiency/debugging purposes. These options include:
5345: + MAT_NEW_NONZERO_LOCATIONS - additional insertions will be allowed if they generate a new nonzero (slow)
5346: . MAT_NEW_DIAGONALS - new diagonals will be allowed (for block diagonal format only)
5347: . MAT_IGNORE_OFF_PROC_ENTRIES - drops off-processor entries
5348: . MAT_NEW_NONZERO_LOCATION_ERR - generates an error for new matrix entry
5349: . MAT_USE_HASH_TABLE - uses a hash table to speed up matrix assembly
5350: . MAT_NO_OFF_PROC_ENTRIES - you know each process will only set values for its own rows, will generate an error if
5351: any process sets values for another process. This avoids all reductions in the MatAssembly routines and thus improves
5352: performance for very large process counts.
5353: - MAT_SUBSET_OFF_PROC_ENTRIES - you know that the first assembly after setting this flag will set a superset
5354: of the off-process entries required for all subsequent assemblies. This avoids a rendezvous step in the MatAssembly
5355: functions, instead sending only neighbor messages.
5357: Notes:
5358: Except for MAT_UNUSED_NONZERO_LOCATION_ERR and MAT_ROW_ORIENTED all processes that share the matrix must pass the same value in flg!
5360: Some options are relevant only for particular matrix types and
5361: are thus ignored by others. Other options are not supported by
5362: certain matrix types and will generate an error message if set.
5364: If using a Fortran 77 module to compute a matrix, one may need to
5365: use the column-oriented option (or convert to the row-oriented
5366: format).
5368: MAT_NEW_NONZERO_LOCATIONS set to PETSC_FALSE indicates that any add or insertion
5369: that would generate a new entry in the nonzero structure is instead
5370: ignored. Thus, if memory has not alredy been allocated for this particular
5371: data, then the insertion is ignored. For dense matrices, in which
5372: the entire array is allocated, no entries are ever ignored.
5373: Set after the first MatAssemblyEnd(). If this option is set then the MatAssemblyBegin/End() processes has one less global reduction
5375: MAT_NEW_NONZERO_LOCATION_ERR set to PETSC_TRUE indicates that any add or insertion
5376: that would generate a new entry in the nonzero structure instead produces
5377: 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
5379: MAT_NEW_NONZERO_ALLOCATION_ERR set to PETSC_TRUE indicates that any add or insertion
5380: that would generate a new entry that has not been preallocated will
5381: instead produce an error. (Currently supported for AIJ and BAIJ formats
5382: only.) This is a useful flag when debugging matrix memory preallocation.
5383: If this option is set then the MatAssemblyBegin/End() processes has one less global reduction
5385: MAT_IGNORE_OFF_PROC_ENTRIES set to PETSC_TRUE indicates entries destined for
5386: other processors should be dropped, rather than stashed.
5387: This is useful if you know that the "owning" processor is also
5388: always generating the correct matrix entries, so that PETSc need
5389: not transfer duplicate entries generated on another processor.
5391: MAT_USE_HASH_TABLE indicates that a hash table be used to improve the
5392: searches during matrix assembly. When this flag is set, the hash table
5393: is created during the first Matrix Assembly. This hash table is
5394: used the next time through, during MatSetVaules()/MatSetVaulesBlocked()
5395: to improve the searching of indices. MAT_NEW_NONZERO_LOCATIONS flag
5396: should be used with MAT_USE_HASH_TABLE flag. This option is currently
5397: supported by MATMPIBAIJ format only.
5399: MAT_KEEP_NONZERO_PATTERN indicates when MatZeroRows() is called the zeroed entries
5400: are kept in the nonzero structure
5402: MAT_IGNORE_ZERO_ENTRIES - for AIJ/IS matrices this will stop zero values from creating
5403: a zero location in the matrix
5405: MAT_USE_INODES - indicates using inode version of the code - works with AIJ matrix types
5407: MAT_NO_OFF_PROC_ZERO_ROWS - you know each process will only zero its own rows. This avoids all reductions in the
5408: zero row routines and thus improves performance for very large process counts.
5410: MAT_IGNORE_LOWER_TRIANGULAR - For SBAIJ matrices will ignore any insertions you make in the lower triangular
5411: part of the matrix (since they should match the upper triangular part).
5413: MAT_SORTED_FULL - each process provides exactly its local rows; all column indices for a given row are passed in a
5414: single call to MatSetValues(), preallocation is perfect, row oriented, INSERT_VALUES is used. Common
5415: with finite difference schemes with non-periodic boundary conditions.
5416: Notes:
5417: Can only be called after MatSetSizes() and MatSetType() have been set.
5419: Level: intermediate
5421: .seealso: MatOption, Mat
5423: @*/
5424: PetscErrorCode MatSetOption(Mat mat,MatOption op,PetscBool flg)
5425: {
5431: if (op > 0) {
5434: }
5436: 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);
5437: 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()");
5439: switch (op) {
5440: case MAT_NO_OFF_PROC_ENTRIES:
5441: mat->nooffprocentries = flg;
5442: return(0);
5443: break;
5444: case MAT_SUBSET_OFF_PROC_ENTRIES:
5445: mat->assembly_subset = flg;
5446: if (!mat->assembly_subset) { /* See the same logic in VecAssembly wrt VEC_SUBSET_OFF_PROC_ENTRIES */
5447: #if !defined(PETSC_HAVE_MPIUNI)
5448: MatStashScatterDestroy_BTS(&mat->stash);
5449: #endif
5450: mat->stash.first_assembly_done = PETSC_FALSE;
5451: }
5452: return(0);
5453: case MAT_NO_OFF_PROC_ZERO_ROWS:
5454: mat->nooffproczerorows = flg;
5455: return(0);
5456: break;
5457: case MAT_SPD:
5458: mat->spd_set = PETSC_TRUE;
5459: mat->spd = flg;
5460: if (flg) {
5461: mat->symmetric = PETSC_TRUE;
5462: mat->structurally_symmetric = PETSC_TRUE;
5463: mat->symmetric_set = PETSC_TRUE;
5464: mat->structurally_symmetric_set = PETSC_TRUE;
5465: }
5466: break;
5467: case MAT_SYMMETRIC:
5468: mat->symmetric = flg;
5469: if (flg) mat->structurally_symmetric = PETSC_TRUE;
5470: mat->symmetric_set = PETSC_TRUE;
5471: mat->structurally_symmetric_set = flg;
5472: #if !defined(PETSC_USE_COMPLEX)
5473: mat->hermitian = flg;
5474: mat->hermitian_set = PETSC_TRUE;
5475: #endif
5476: break;
5477: case MAT_HERMITIAN:
5478: mat->hermitian = flg;
5479: if (flg) mat->structurally_symmetric = PETSC_TRUE;
5480: mat->hermitian_set = PETSC_TRUE;
5481: mat->structurally_symmetric_set = flg;
5482: #if !defined(PETSC_USE_COMPLEX)
5483: mat->symmetric = flg;
5484: mat->symmetric_set = PETSC_TRUE;
5485: #endif
5486: break;
5487: case MAT_STRUCTURALLY_SYMMETRIC:
5488: mat->structurally_symmetric = flg;
5489: mat->structurally_symmetric_set = PETSC_TRUE;
5490: break;
5491: case MAT_SYMMETRY_ETERNAL:
5492: mat->symmetric_eternal = flg;
5493: break;
5494: case MAT_STRUCTURE_ONLY:
5495: mat->structure_only = flg;
5496: break;
5497: case MAT_SORTED_FULL:
5498: mat->sortedfull = flg;
5499: break;
5500: default:
5501: break;
5502: }
5503: if (mat->ops->setoption) {
5504: (*mat->ops->setoption)(mat,op,flg);
5505: }
5506: return(0);
5507: }
5509: /*@
5510: MatGetOption - Gets a parameter option that has been set for a matrix.
5512: Logically Collective on Mat for certain operations, such as MAT_SPD, not collective for MAT_ROW_ORIENTED, see MatOption
5514: Input Parameters:
5515: + mat - the matrix
5516: - option - the option, this only responds to certain options, check the code for which ones
5518: Output Parameter:
5519: . flg - turn the option on (PETSC_TRUE) or off (PETSC_FALSE)
5521: Notes:
5522: Can only be called after MatSetSizes() and MatSetType() have been set.
5524: Level: intermediate
5526: .seealso: MatOption, MatSetOption()
5528: @*/
5529: PetscErrorCode MatGetOption(Mat mat,MatOption op,PetscBool *flg)
5530: {
5535: 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);
5536: 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()");
5538: switch (op) {
5539: case MAT_NO_OFF_PROC_ENTRIES:
5540: *flg = mat->nooffprocentries;
5541: break;
5542: case MAT_NO_OFF_PROC_ZERO_ROWS:
5543: *flg = mat->nooffproczerorows;
5544: break;
5545: case MAT_SYMMETRIC:
5546: *flg = mat->symmetric;
5547: break;
5548: case MAT_HERMITIAN:
5549: *flg = mat->hermitian;
5550: break;
5551: case MAT_STRUCTURALLY_SYMMETRIC:
5552: *flg = mat->structurally_symmetric;
5553: break;
5554: case MAT_SYMMETRY_ETERNAL:
5555: *flg = mat->symmetric_eternal;
5556: break;
5557: case MAT_SPD:
5558: *flg = mat->spd;
5559: break;
5560: default:
5561: break;
5562: }
5563: return(0);
5564: }
5566: /*@
5567: MatZeroEntries - Zeros all entries of a matrix. For sparse matrices
5568: this routine retains the old nonzero structure.
5570: Logically Collective on Mat
5572: Input Parameters:
5573: . mat - the matrix
5575: Level: intermediate
5577: Notes:
5578: 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.
5579: See the Performance chapter of the users manual for information on preallocating matrices.
5581: .seealso: MatZeroRows()
5582: @*/
5583: PetscErrorCode MatZeroEntries(Mat mat)
5584: {
5590: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5591: 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");
5592: if (!mat->ops->zeroentries) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5593: MatCheckPreallocated(mat,1);
5595: PetscLogEventBegin(MAT_ZeroEntries,mat,0,0,0);
5596: (*mat->ops->zeroentries)(mat);
5597: PetscLogEventEnd(MAT_ZeroEntries,mat,0,0,0);
5598: PetscObjectStateIncrease((PetscObject)mat);
5599: return(0);
5600: }
5602: /*@
5603: MatZeroRowsColumns - Zeros all entries (except possibly the main diagonal)
5604: of a set of rows and columns of a matrix.
5606: Collective on Mat
5608: Input Parameters:
5609: + mat - the matrix
5610: . numRows - the number of rows to remove
5611: . rows - the global row indices
5612: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
5613: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
5614: - b - optional vector of right hand side, that will be adjusted by provided solution
5616: Notes:
5617: This does not change the nonzero structure of the matrix, it merely zeros those entries in the matrix.
5619: The user can set a value in the diagonal entry (or for the AIJ and
5620: row formats can optionally remove the main diagonal entry from the
5621: nonzero structure as well, by passing 0.0 as the final argument).
5623: For the parallel case, all processes that share the matrix (i.e.,
5624: those in the communicator used for matrix creation) MUST call this
5625: routine, regardless of whether any rows being zeroed are owned by
5626: them.
5628: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
5629: list only rows local to itself).
5631: The option MAT_NO_OFF_PROC_ZERO_ROWS does not apply to this routine.
5633: Level: intermediate
5635: .seealso: MatZeroRowsIS(), MatZeroRows(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
5636: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
5637: @*/
5638: PetscErrorCode MatZeroRowsColumns(Mat mat,PetscInt numRows,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
5639: {
5646: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5647: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5648: if (!mat->ops->zerorowscolumns) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5649: MatCheckPreallocated(mat,1);
5651: (*mat->ops->zerorowscolumns)(mat,numRows,rows,diag,x,b);
5652: MatViewFromOptions(mat,NULL,"-mat_view");
5653: PetscObjectStateIncrease((PetscObject)mat);
5654: return(0);
5655: }
5657: /*@
5658: MatZeroRowsColumnsIS - Zeros all entries (except possibly the main diagonal)
5659: of a set of rows and columns of a matrix.
5661: Collective on Mat
5663: Input Parameters:
5664: + mat - the matrix
5665: . is - the rows to zero
5666: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
5667: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
5668: - b - optional vector of right hand side, that will be adjusted by provided solution
5670: Notes:
5671: This does not change the nonzero structure of the matrix, it merely zeros those entries in the matrix.
5673: The user can set a value in the diagonal entry (or for the AIJ and
5674: row formats can optionally remove the main diagonal entry from the
5675: nonzero structure as well, by passing 0.0 as the final argument).
5677: For the parallel case, all processes that share the matrix (i.e.,
5678: those in the communicator used for matrix creation) MUST call this
5679: routine, regardless of whether any rows being zeroed are owned by
5680: them.
5682: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
5683: list only rows local to itself).
5685: The option MAT_NO_OFF_PROC_ZERO_ROWS does not apply to this routine.
5687: Level: intermediate
5689: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
5690: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRows(), MatZeroRowsColumnsStencil()
5691: @*/
5692: PetscErrorCode MatZeroRowsColumnsIS(Mat mat,IS is,PetscScalar diag,Vec x,Vec b)
5693: {
5695: PetscInt numRows;
5696: const PetscInt *rows;
5703: ISGetLocalSize(is,&numRows);
5704: ISGetIndices(is,&rows);
5705: MatZeroRowsColumns(mat,numRows,rows,diag,x,b);
5706: ISRestoreIndices(is,&rows);
5707: return(0);
5708: }
5710: /*@
5711: MatZeroRows - Zeros all entries (except possibly the main diagonal)
5712: of a set of rows of a matrix.
5714: Collective on Mat
5716: Input Parameters:
5717: + mat - the matrix
5718: . numRows - the number of rows to remove
5719: . rows - the global row indices
5720: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
5721: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
5722: - b - optional vector of right hand side, that will be adjusted by provided solution
5724: Notes:
5725: For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
5726: but does not release memory. For the dense and block diagonal
5727: formats this does not alter the nonzero structure.
5729: If the option MatSetOption(mat,MAT_KEEP_NONZERO_PATTERN,PETSC_TRUE) the nonzero structure
5730: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
5731: merely zeroed.
5733: The user can set a value in the diagonal entry (or for the AIJ and
5734: row formats can optionally remove the main diagonal entry from the
5735: nonzero structure as well, by passing 0.0 as the final argument).
5737: For the parallel case, all processes that share the matrix (i.e.,
5738: those in the communicator used for matrix creation) MUST call this
5739: routine, regardless of whether any rows being zeroed are owned by
5740: them.
5742: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
5743: list only rows local to itself).
5745: You can call MatSetOption(mat,MAT_NO_OFF_PROC_ZERO_ROWS,PETSC_TRUE) if each process indicates only rows it
5746: owns that are to be zeroed. This saves a global synchronization in the implementation.
5748: Level: intermediate
5750: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
5751: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
5752: @*/
5753: PetscErrorCode MatZeroRows(Mat mat,PetscInt numRows,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
5754: {
5761: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5762: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5763: if (!mat->ops->zerorows) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5764: MatCheckPreallocated(mat,1);
5766: (*mat->ops->zerorows)(mat,numRows,rows,diag,x,b);
5767: MatViewFromOptions(mat,NULL,"-mat_view");
5768: PetscObjectStateIncrease((PetscObject)mat);
5769: return(0);
5770: }
5772: /*@
5773: MatZeroRowsIS - Zeros all entries (except possibly the main diagonal)
5774: of a set of rows of a matrix.
5776: Collective on Mat
5778: Input Parameters:
5779: + mat - the matrix
5780: . is - index set of rows to remove
5781: . diag - value put in all diagonals of eliminated rows
5782: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
5783: - b - optional vector of right hand side, that will be adjusted by provided solution
5785: Notes:
5786: For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
5787: but does not release memory. For the dense and block diagonal
5788: formats this does not alter the nonzero structure.
5790: If the option MatSetOption(mat,MAT_KEEP_NONZERO_PATTERN,PETSC_TRUE) the nonzero structure
5791: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
5792: merely zeroed.
5794: The user can set a value in the diagonal entry (or for the AIJ and
5795: row formats can optionally remove the main diagonal entry from the
5796: nonzero structure as well, by passing 0.0 as the final argument).
5798: For the parallel case, all processes that share the matrix (i.e.,
5799: those in the communicator used for matrix creation) MUST call this
5800: routine, regardless of whether any rows being zeroed are owned by
5801: them.
5803: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
5804: list only rows local to itself).
5806: You can call MatSetOption(mat,MAT_NO_OFF_PROC_ZERO_ROWS,PETSC_TRUE) if each process indicates only rows it
5807: owns that are to be zeroed. This saves a global synchronization in the implementation.
5809: Level: intermediate
5811: .seealso: MatZeroRows(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
5812: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
5813: @*/
5814: PetscErrorCode MatZeroRowsIS(Mat mat,IS is,PetscScalar diag,Vec x,Vec b)
5815: {
5816: PetscInt numRows;
5817: const PetscInt *rows;
5824: ISGetLocalSize(is,&numRows);
5825: ISGetIndices(is,&rows);
5826: MatZeroRows(mat,numRows,rows,diag,x,b);
5827: ISRestoreIndices(is,&rows);
5828: return(0);
5829: }
5831: /*@
5832: MatZeroRowsStencil - Zeros all entries (except possibly the main diagonal)
5833: of a set of rows of a matrix. These rows must be local to the process.
5835: Collective on Mat
5837: Input Parameters:
5838: + mat - the matrix
5839: . numRows - the number of rows to remove
5840: . rows - the grid coordinates (and component number when dof > 1) for matrix rows
5841: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
5842: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
5843: - b - optional vector of right hand side, that will be adjusted by provided solution
5845: Notes:
5846: For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
5847: but does not release memory. For the dense and block diagonal
5848: formats this does not alter the nonzero structure.
5850: If the option MatSetOption(mat,MAT_KEEP_NONZERO_PATTERN,PETSC_TRUE) the nonzero structure
5851: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
5852: merely zeroed.
5854: The user can set a value in the diagonal entry (or for the AIJ and
5855: row formats can optionally remove the main diagonal entry from the
5856: nonzero structure as well, by passing 0.0 as the final argument).
5858: For the parallel case, all processes that share the matrix (i.e.,
5859: those in the communicator used for matrix creation) MUST call this
5860: routine, regardless of whether any rows being zeroed are owned by
5861: them.
5863: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
5864: list only rows local to itself).
5866: The grid coordinates are across the entire grid, not just the local portion
5868: In Fortran idxm and idxn should be declared as
5869: $ MatStencil idxm(4,m)
5870: and the values inserted using
5871: $ idxm(MatStencil_i,1) = i
5872: $ idxm(MatStencil_j,1) = j
5873: $ idxm(MatStencil_k,1) = k
5874: $ idxm(MatStencil_c,1) = c
5875: etc
5877: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
5878: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
5879: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
5880: DM_BOUNDARY_PERIODIC boundary type.
5882: 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
5883: a single value per point) you can skip filling those indices.
5885: Level: intermediate
5887: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsl(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
5888: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
5889: @*/
5890: PetscErrorCode MatZeroRowsStencil(Mat mat,PetscInt numRows,const MatStencil rows[],PetscScalar diag,Vec x,Vec b)
5891: {
5892: PetscInt dim = mat->stencil.dim;
5893: PetscInt sdim = dim - (1 - (PetscInt) mat->stencil.noc);
5894: PetscInt *dims = mat->stencil.dims+1;
5895: PetscInt *starts = mat->stencil.starts;
5896: PetscInt *dxm = (PetscInt*) rows;
5897: PetscInt *jdxm, i, j, tmp, numNewRows = 0;
5905: PetscMalloc1(numRows, &jdxm);
5906: for (i = 0; i < numRows; ++i) {
5907: /* Skip unused dimensions (they are ordered k, j, i, c) */
5908: for (j = 0; j < 3-sdim; ++j) dxm++;
5909: /* Local index in X dir */
5910: tmp = *dxm++ - starts[0];
5911: /* Loop over remaining dimensions */
5912: for (j = 0; j < dim-1; ++j) {
5913: /* If nonlocal, set index to be negative */
5914: if ((*dxm++ - starts[j+1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
5915: /* Update local index */
5916: else tmp = tmp*dims[j] + *(dxm-1) - starts[j+1];
5917: }
5918: /* Skip component slot if necessary */
5919: if (mat->stencil.noc) dxm++;
5920: /* Local row number */
5921: if (tmp >= 0) {
5922: jdxm[numNewRows++] = tmp;
5923: }
5924: }
5925: MatZeroRowsLocal(mat,numNewRows,jdxm,diag,x,b);
5926: PetscFree(jdxm);
5927: return(0);
5928: }
5930: /*@
5931: MatZeroRowsColumnsStencil - Zeros all row and column entries (except possibly the main diagonal)
5932: of a set of rows and columns of a matrix.
5934: Collective on Mat
5936: Input Parameters:
5937: + mat - the matrix
5938: . numRows - the number of rows/columns to remove
5939: . rows - the grid coordinates (and component number when dof > 1) for matrix rows
5940: . diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
5941: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
5942: - b - optional vector of right hand side, that will be adjusted by provided solution
5944: Notes:
5945: For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
5946: but does not release memory. For the dense and block diagonal
5947: formats this does not alter the nonzero structure.
5949: If the option MatSetOption(mat,MAT_KEEP_NONZERO_PATTERN,PETSC_TRUE) the nonzero structure
5950: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
5951: merely zeroed.
5953: The user can set a value in the diagonal entry (or for the AIJ and
5954: row formats can optionally remove the main diagonal entry from the
5955: nonzero structure as well, by passing 0.0 as the final argument).
5957: For the parallel case, all processes that share the matrix (i.e.,
5958: those in the communicator used for matrix creation) MUST call this
5959: routine, regardless of whether any rows being zeroed are owned by
5960: them.
5962: Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
5963: list only rows local to itself, but the row/column numbers are given in local numbering).
5965: The grid coordinates are across the entire grid, not just the local portion
5967: In Fortran idxm and idxn should be declared as
5968: $ MatStencil idxm(4,m)
5969: and the values inserted using
5970: $ idxm(MatStencil_i,1) = i
5971: $ idxm(MatStencil_j,1) = j
5972: $ idxm(MatStencil_k,1) = k
5973: $ idxm(MatStencil_c,1) = c
5974: etc
5976: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
5977: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
5978: etc to obtain values that obtained by wrapping the values from the left edge. This does not work for anything but the
5979: DM_BOUNDARY_PERIODIC boundary type.
5981: 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
5982: a single value per point) you can skip filling those indices.
5984: Level: intermediate
5986: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
5987: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRows()
5988: @*/
5989: PetscErrorCode MatZeroRowsColumnsStencil(Mat mat,PetscInt numRows,const MatStencil rows[],PetscScalar diag,Vec x,Vec b)
5990: {
5991: PetscInt dim = mat->stencil.dim;
5992: PetscInt sdim = dim - (1 - (PetscInt) mat->stencil.noc);
5993: PetscInt *dims = mat->stencil.dims+1;
5994: PetscInt *starts = mat->stencil.starts;
5995: PetscInt *dxm = (PetscInt*) rows;
5996: PetscInt *jdxm, i, j, tmp, numNewRows = 0;
6004: PetscMalloc1(numRows, &jdxm);
6005: for (i = 0; i < numRows; ++i) {
6006: /* Skip unused dimensions (they are ordered k, j, i, c) */
6007: for (j = 0; j < 3-sdim; ++j) dxm++;
6008: /* Local index in X dir */
6009: tmp = *dxm++ - starts[0];
6010: /* Loop over remaining dimensions */
6011: for (j = 0; j < dim-1; ++j) {
6012: /* If nonlocal, set index to be negative */
6013: if ((*dxm++ - starts[j+1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
6014: /* Update local index */
6015: else tmp = tmp*dims[j] + *(dxm-1) - starts[j+1];
6016: }
6017: /* Skip component slot if necessary */
6018: if (mat->stencil.noc) dxm++;
6019: /* Local row number */
6020: if (tmp >= 0) {
6021: jdxm[numNewRows++] = tmp;
6022: }
6023: }
6024: MatZeroRowsColumnsLocal(mat,numNewRows,jdxm,diag,x,b);
6025: PetscFree(jdxm);
6026: return(0);
6027: }
6029: /*@C
6030: MatZeroRowsLocal - Zeros all entries (except possibly the main diagonal)
6031: of a set of rows of a matrix; using local numbering of rows.
6033: Collective on Mat
6035: Input Parameters:
6036: + mat - the matrix
6037: . numRows - the number of rows to remove
6038: . rows - the global row indices
6039: . diag - value put in all diagonals of eliminated rows
6040: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6041: - b - optional vector of right hand side, that will be adjusted by provided solution
6043: Notes:
6044: Before calling MatZeroRowsLocal(), the user must first set the
6045: local-to-global mapping by calling MatSetLocalToGlobalMapping().
6047: For the AIJ matrix formats this removes the old nonzero structure,
6048: but does not release memory. For the dense and block diagonal
6049: formats this does not alter the nonzero structure.
6051: If the option MatSetOption(mat,MAT_KEEP_NONZERO_PATTERN,PETSC_TRUE) the nonzero structure
6052: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
6053: merely zeroed.
6055: The user can set a value in the diagonal entry (or for the AIJ and
6056: row formats can optionally remove the main diagonal entry from the
6057: nonzero structure as well, by passing 0.0 as the final argument).
6059: You can call MatSetOption(mat,MAT_NO_OFF_PROC_ZERO_ROWS,PETSC_TRUE) if each process indicates only rows it
6060: owns that are to be zeroed. This saves a global synchronization in the implementation.
6062: Level: intermediate
6064: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRows(), MatSetOption(),
6065: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
6066: @*/
6067: PetscErrorCode MatZeroRowsLocal(Mat mat,PetscInt numRows,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
6068: {
6075: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6076: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6077: MatCheckPreallocated(mat,1);
6079: if (mat->ops->zerorowslocal) {
6080: (*mat->ops->zerorowslocal)(mat,numRows,rows,diag,x,b);
6081: } else {
6082: IS is, newis;
6083: const PetscInt *newRows;
6085: if (!mat->rmap->mapping) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Need to provide local to global mapping to matrix first");
6086: ISCreateGeneral(PETSC_COMM_SELF,numRows,rows,PETSC_COPY_VALUES,&is);
6087: ISLocalToGlobalMappingApplyIS(mat->rmap->mapping,is,&newis);
6088: ISGetIndices(newis,&newRows);
6089: (*mat->ops->zerorows)(mat,numRows,newRows,diag,x,b);
6090: ISRestoreIndices(newis,&newRows);
6091: ISDestroy(&newis);
6092: ISDestroy(&is);
6093: }
6094: PetscObjectStateIncrease((PetscObject)mat);
6095: return(0);
6096: }
6098: /*@
6099: MatZeroRowsLocalIS - Zeros all entries (except possibly the main diagonal)
6100: of a set of rows of a matrix; using local numbering of rows.
6102: Collective on Mat
6104: Input Parameters:
6105: + mat - the matrix
6106: . is - index set of rows to remove
6107: . diag - value put in all diagonals of eliminated rows
6108: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6109: - b - optional vector of right hand side, that will be adjusted by provided solution
6111: Notes:
6112: Before calling MatZeroRowsLocalIS(), the user must first set the
6113: local-to-global mapping by calling MatSetLocalToGlobalMapping().
6115: For the AIJ matrix formats this removes the old nonzero structure,
6116: but does not release memory. For the dense and block diagonal
6117: formats this does not alter the nonzero structure.
6119: If the option MatSetOption(mat,MAT_KEEP_NONZERO_PATTERN,PETSC_TRUE) the nonzero structure
6120: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
6121: merely zeroed.
6123: The user can set a value in the diagonal entry (or for the AIJ and
6124: row formats can optionally remove the main diagonal entry from the
6125: nonzero structure as well, by passing 0.0 as the final argument).
6127: You can call MatSetOption(mat,MAT_NO_OFF_PROC_ZERO_ROWS,PETSC_TRUE) if each process indicates only rows it
6128: owns that are to be zeroed. This saves a global synchronization in the implementation.
6130: Level: intermediate
6132: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRows(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
6133: MatZeroRowsColumnsLocal(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
6134: @*/
6135: PetscErrorCode MatZeroRowsLocalIS(Mat mat,IS is,PetscScalar diag,Vec x,Vec b)
6136: {
6138: PetscInt numRows;
6139: const PetscInt *rows;
6145: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6146: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6147: MatCheckPreallocated(mat,1);
6149: ISGetLocalSize(is,&numRows);
6150: ISGetIndices(is,&rows);
6151: MatZeroRowsLocal(mat,numRows,rows,diag,x,b);
6152: ISRestoreIndices(is,&rows);
6153: return(0);
6154: }
6156: /*@
6157: MatZeroRowsColumnsLocal - Zeros all entries (except possibly the main diagonal)
6158: of a set of rows and columns of a matrix; using local numbering of rows.
6160: Collective on Mat
6162: Input Parameters:
6163: + mat - the matrix
6164: . numRows - the number of rows to remove
6165: . rows - the global row indices
6166: . diag - value put in all diagonals of eliminated rows
6167: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6168: - b - optional vector of right hand side, that will be adjusted by provided solution
6170: Notes:
6171: Before calling MatZeroRowsColumnsLocal(), the user must first set the
6172: local-to-global mapping by calling MatSetLocalToGlobalMapping().
6174: The user can set a value in the diagonal entry (or for the AIJ and
6175: row formats can optionally remove the main diagonal entry from the
6176: nonzero structure as well, by passing 0.0 as the final argument).
6178: Level: intermediate
6180: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
6181: MatZeroRows(), MatZeroRowsColumnsLocalIS(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
6182: @*/
6183: PetscErrorCode MatZeroRowsColumnsLocal(Mat mat,PetscInt numRows,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
6184: {
6186: IS is, newis;
6187: const PetscInt *newRows;
6193: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6194: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6195: MatCheckPreallocated(mat,1);
6197: if (!mat->cmap->mapping) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Need to provide local to global mapping to matrix first");
6198: ISCreateGeneral(PETSC_COMM_SELF,numRows,rows,PETSC_COPY_VALUES,&is);
6199: ISLocalToGlobalMappingApplyIS(mat->cmap->mapping,is,&newis);
6200: ISGetIndices(newis,&newRows);
6201: (*mat->ops->zerorowscolumns)(mat,numRows,newRows,diag,x,b);
6202: ISRestoreIndices(newis,&newRows);
6203: ISDestroy(&newis);
6204: ISDestroy(&is);
6205: PetscObjectStateIncrease((PetscObject)mat);
6206: return(0);
6207: }
6209: /*@
6210: MatZeroRowsColumnsLocalIS - Zeros all entries (except possibly the main diagonal)
6211: of a set of rows and columns of a matrix; using local numbering of rows.
6213: Collective on Mat
6215: Input Parameters:
6216: + mat - the matrix
6217: . is - index set of rows to remove
6218: . diag - value put in all diagonals of eliminated rows
6219: . x - optional vector of solutions for zeroed rows (other entries in vector are not used)
6220: - b - optional vector of right hand side, that will be adjusted by provided solution
6222: Notes:
6223: Before calling MatZeroRowsColumnsLocalIS(), the user must first set the
6224: local-to-global mapping by calling MatSetLocalToGlobalMapping().
6226: The user can set a value in the diagonal entry (or for the AIJ and
6227: row formats can optionally remove the main diagonal entry from the
6228: nonzero structure as well, by passing 0.0 as the final argument).
6230: Level: intermediate
6232: .seealso: MatZeroRowsIS(), MatZeroRowsColumns(), MatZeroRowsLocalIS(), MatZeroRowsStencil(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption(),
6233: MatZeroRowsColumnsLocal(), MatZeroRows(), MatZeroRowsColumnsIS(), MatZeroRowsColumnsStencil()
6234: @*/
6235: PetscErrorCode MatZeroRowsColumnsLocalIS(Mat mat,IS is,PetscScalar diag,Vec x,Vec b)
6236: {
6238: PetscInt numRows;
6239: const PetscInt *rows;
6245: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6246: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6247: MatCheckPreallocated(mat,1);
6249: ISGetLocalSize(is,&numRows);
6250: ISGetIndices(is,&rows);
6251: MatZeroRowsColumnsLocal(mat,numRows,rows,diag,x,b);
6252: ISRestoreIndices(is,&rows);
6253: return(0);
6254: }
6256: /*@C
6257: MatGetSize - Returns the numbers of rows and columns in a matrix.
6259: Not Collective
6261: Input Parameter:
6262: . mat - the matrix
6264: Output Parameters:
6265: + m - the number of global rows
6266: - n - the number of global columns
6268: Note: both output parameters can be NULL on input.
6270: Level: beginner
6272: .seealso: MatGetLocalSize()
6273: @*/
6274: PetscErrorCode MatGetSize(Mat mat,PetscInt *m,PetscInt *n)
6275: {
6278: if (m) *m = mat->rmap->N;
6279: if (n) *n = mat->cmap->N;
6280: return(0);
6281: }
6283: /*@C
6284: MatGetLocalSize - Returns the number of rows and columns in a matrix
6285: stored locally. This information may be implementation dependent, so
6286: use with care.
6288: Not Collective
6290: Input Parameters:
6291: . mat - the matrix
6293: Output Parameters:
6294: + m - the number of local rows
6295: - n - the number of local columns
6297: Note: both output parameters can be NULL on input.
6299: Level: beginner
6301: .seealso: MatGetSize()
6302: @*/
6303: PetscErrorCode MatGetLocalSize(Mat mat,PetscInt *m,PetscInt *n)
6304: {
6309: if (m) *m = mat->rmap->n;
6310: if (n) *n = mat->cmap->n;
6311: return(0);
6312: }
6314: /*@C
6315: MatGetOwnershipRangeColumn - Returns the range of matrix columns associated with rows of a vector one multiplies by that owned by
6316: this processor. (The columns of the "diagonal block")
6318: Not Collective, unless matrix has not been allocated, then collective on Mat
6320: Input Parameters:
6321: . mat - the matrix
6323: Output Parameters:
6324: + m - the global index of the first local column
6325: - n - one more than the global index of the last local column
6327: Notes:
6328: both output parameters can be NULL on input.
6330: Level: developer
6332: .seealso: MatGetOwnershipRange(), MatGetOwnershipRanges(), MatGetOwnershipRangesColumn()
6334: @*/
6335: PetscErrorCode MatGetOwnershipRangeColumn(Mat mat,PetscInt *m,PetscInt *n)
6336: {
6342: MatCheckPreallocated(mat,1);
6343: if (m) *m = mat->cmap->rstart;
6344: if (n) *n = mat->cmap->rend;
6345: return(0);
6346: }
6348: /*@C
6349: MatGetOwnershipRange - Returns the range of matrix rows owned by
6350: this processor, assuming that the matrix is laid out with the first
6351: n1 rows on the first processor, the next n2 rows on the second, etc.
6352: For certain parallel layouts this range may not be well defined.
6354: Not Collective
6356: Input Parameters:
6357: . mat - the matrix
6359: Output Parameters:
6360: + m - the global index of the first local row
6361: - n - one more than the global index of the last local row
6363: Note: Both output parameters can be NULL on input.
6364: $ This function requires that the matrix be preallocated. If you have not preallocated, consider using
6365: $ PetscSplitOwnership(MPI_Comm comm, PetscInt *n, PetscInt *N)
6366: $ and then MPI_Scan() to calculate prefix sums of the local sizes.
6368: Level: beginner
6370: .seealso: MatGetOwnershipRanges(), MatGetOwnershipRangeColumn(), MatGetOwnershipRangesColumn(), PetscSplitOwnership(), PetscSplitOwnershipBlock()
6372: @*/
6373: PetscErrorCode MatGetOwnershipRange(Mat mat,PetscInt *m,PetscInt *n)
6374: {
6380: MatCheckPreallocated(mat,1);
6381: if (m) *m = mat->rmap->rstart;
6382: if (n) *n = mat->rmap->rend;
6383: return(0);
6384: }
6386: /*@C
6387: MatGetOwnershipRanges - Returns the range of matrix rows owned by
6388: each process
6390: Not Collective, unless matrix has not been allocated, then collective on Mat
6392: Input Parameters:
6393: . mat - the matrix
6395: Output Parameters:
6396: . ranges - start of each processors portion plus one more than the total length at the end
6398: Level: beginner
6400: .seealso: MatGetOwnershipRange(), MatGetOwnershipRangeColumn(), MatGetOwnershipRangesColumn()
6402: @*/
6403: PetscErrorCode MatGetOwnershipRanges(Mat mat,const PetscInt **ranges)
6404: {
6410: MatCheckPreallocated(mat,1);
6411: PetscLayoutGetRanges(mat->rmap,ranges);
6412: return(0);
6413: }
6415: /*@C
6416: MatGetOwnershipRangesColumn - Returns the range of matrix columns associated with rows of a vector one multiplies by that owned by
6417: this processor. (The columns of the "diagonal blocks" for each process)
6419: Not Collective, unless matrix has not been allocated, then collective on Mat
6421: Input Parameters:
6422: . mat - the matrix
6424: Output Parameters:
6425: . ranges - start of each processors portion plus one more then the total length at the end
6427: Level: beginner
6429: .seealso: MatGetOwnershipRange(), MatGetOwnershipRangeColumn(), MatGetOwnershipRanges()
6431: @*/
6432: PetscErrorCode MatGetOwnershipRangesColumn(Mat mat,const PetscInt **ranges)
6433: {
6439: MatCheckPreallocated(mat,1);
6440: PetscLayoutGetRanges(mat->cmap,ranges);
6441: return(0);
6442: }
6444: /*@C
6445: MatGetOwnershipIS - Get row and column ownership as index sets
6447: Not Collective
6449: Input Arguments:
6450: . A - matrix of type Elemental
6452: Output Arguments:
6453: + rows - rows in which this process owns elements
6454: - cols - columns in which this process owns elements
6456: Level: intermediate
6458: .seealso: MatGetOwnershipRange(), MatGetOwnershipRangeColumn(), MatSetValues(), MATELEMENTAL
6459: @*/
6460: PetscErrorCode MatGetOwnershipIS(Mat A,IS *rows,IS *cols)
6461: {
6462: PetscErrorCode ierr,(*f)(Mat,IS*,IS*);
6465: MatCheckPreallocated(A,1);
6466: PetscObjectQueryFunction((PetscObject)A,"MatGetOwnershipIS_C",&f);
6467: if (f) {
6468: (*f)(A,rows,cols);
6469: } else { /* Create a standard row-based partition, each process is responsible for ALL columns in their row block */
6470: if (rows) {ISCreateStride(PETSC_COMM_SELF,A->rmap->n,A->rmap->rstart,1,rows);}
6471: if (cols) {ISCreateStride(PETSC_COMM_SELF,A->cmap->N,0,1,cols);}
6472: }
6473: return(0);
6474: }
6476: /*@C
6477: MatILUFactorSymbolic - Performs symbolic ILU factorization of a matrix.
6478: Uses levels of fill only, not drop tolerance. Use MatLUFactorNumeric()
6479: to complete the factorization.
6481: Collective on Mat
6483: Input Parameters:
6484: + mat - the matrix
6485: . row - row permutation
6486: . column - column permutation
6487: - info - structure containing
6488: $ levels - number of levels of fill.
6489: $ expected fill - as ratio of original fill.
6490: $ 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
6491: missing diagonal entries)
6493: Output Parameters:
6494: . fact - new matrix that has been symbolically factored
6496: Notes:
6497: See Users-Manual: ch_mat for additional information about choosing the fill factor for better efficiency.
6499: Most users should employ the simplified KSP interface for linear solvers
6500: instead of working directly with matrix algebra routines such as this.
6501: See, e.g., KSPCreate().
6503: Level: developer
6505: .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor()
6506: MatGetOrdering(), MatFactorInfo
6508: Note: this uses the definition of level of fill as in Y. Saad, 2003
6510: Developer Note: fortran interface is not autogenerated as the f90
6511: interface defintion cannot be generated correctly [due to MatFactorInfo]
6513: References:
6514: Y. Saad, Iterative methods for sparse linear systems Philadelphia: Society for Industrial and Applied Mathematics, 2003
6515: @*/
6516: PetscErrorCode MatILUFactorSymbolic(Mat fact,Mat mat,IS row,IS col,const MatFactorInfo *info)
6517: {
6527: if (info->levels < 0) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"Levels of fill negative %D",(PetscInt)info->levels);
6528: if (info->fill < 1.0) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"Expected fill less than 1.0 %g",(double)info->fill);
6529: if (!(fact)->ops->ilufactorsymbolic) {
6530: MatSolverType spackage;
6531: MatFactorGetSolverType(fact,&spackage);
6532: SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s symbolic ILU using solver package %s",((PetscObject)mat)->type_name,spackage);
6533: }
6534: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6535: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6536: MatCheckPreallocated(mat,2);
6538: PetscLogEventBegin(MAT_ILUFactorSymbolic,mat,row,col,0);
6539: (fact->ops->ilufactorsymbolic)(fact,mat,row,col,info);
6540: PetscLogEventEnd(MAT_ILUFactorSymbolic,mat,row,col,0);
6541: return(0);
6542: }
6544: /*@C
6545: MatICCFactorSymbolic - Performs symbolic incomplete
6546: Cholesky factorization for a symmetric matrix. Use
6547: MatCholeskyFactorNumeric() to complete the factorization.
6549: Collective on Mat
6551: Input Parameters:
6552: + mat - the matrix
6553: . perm - row and column permutation
6554: - info - structure containing
6555: $ levels - number of levels of fill.
6556: $ expected fill - as ratio of original fill.
6558: Output Parameter:
6559: . fact - the factored matrix
6561: Notes:
6562: Most users should employ the KSP interface for linear solvers
6563: instead of working directly with matrix algebra routines such as this.
6564: See, e.g., KSPCreate().
6566: Level: developer
6568: .seealso: MatCholeskyFactorNumeric(), MatCholeskyFactor(), MatFactorInfo
6570: Note: this uses the definition of level of fill as in Y. Saad, 2003
6572: Developer Note: fortran interface is not autogenerated as the f90
6573: interface defintion cannot be generated correctly [due to MatFactorInfo]
6575: References:
6576: Y. Saad, Iterative methods for sparse linear systems Philadelphia: Society for Industrial and Applied Mathematics, 2003
6577: @*/
6578: PetscErrorCode MatICCFactorSymbolic(Mat fact,Mat mat,IS perm,const MatFactorInfo *info)
6579: {
6588: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6589: if (info->levels < 0) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"Levels negative %D",(PetscInt) info->levels);
6590: if (info->fill < 1.0) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"Expected fill less than 1.0 %g",(double)info->fill);
6591: if (!(fact)->ops->iccfactorsymbolic) {
6592: MatSolverType spackage;
6593: MatFactorGetSolverType(fact,&spackage);
6594: SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Matrix type %s symbolic ICC using solver package %s",((PetscObject)mat)->type_name,spackage);
6595: }
6596: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6597: MatCheckPreallocated(mat,2);
6599: PetscLogEventBegin(MAT_ICCFactorSymbolic,mat,perm,0,0);
6600: (fact->ops->iccfactorsymbolic)(fact,mat,perm,info);
6601: PetscLogEventEnd(MAT_ICCFactorSymbolic,mat,perm,0,0);
6602: return(0);
6603: }
6605: /*@C
6606: MatCreateSubMatrices - Extracts several submatrices from a matrix. If submat
6607: points to an array of valid matrices, they may be reused to store the new
6608: submatrices.
6610: Collective on Mat
6612: Input Parameters:
6613: + mat - the matrix
6614: . n - the number of submatrixes to be extracted (on this processor, may be zero)
6615: . irow, icol - index sets of rows and columns to extract
6616: - scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
6618: Output Parameter:
6619: . submat - the array of submatrices
6621: Notes:
6622: MatCreateSubMatrices() can extract ONLY sequential submatrices
6623: (from both sequential and parallel matrices). Use MatCreateSubMatrix()
6624: to extract a parallel submatrix.
6626: Some matrix types place restrictions on the row and column
6627: indices, such as that they be sorted or that they be equal to each other.
6629: The index sets may not have duplicate entries.
6631: When extracting submatrices from a parallel matrix, each processor can
6632: form a different submatrix by setting the rows and columns of its
6633: individual index sets according to the local submatrix desired.
6635: When finished using the submatrices, the user should destroy
6636: them with MatDestroySubMatrices().
6638: MAT_REUSE_MATRIX can only be used when the nonzero structure of the
6639: original matrix has not changed from that last call to MatCreateSubMatrices().
6641: This routine creates the matrices in submat; you should NOT create them before
6642: calling it. It also allocates the array of matrix pointers submat.
6644: For BAIJ matrices the index sets must respect the block structure, that is if they
6645: request one row/column in a block, they must request all rows/columns that are in
6646: that block. For example, if the block size is 2 you cannot request just row 0 and
6647: column 0.
6649: Fortran Note:
6650: The Fortran interface is slightly different from that given below; it
6651: requires one to pass in as submat a Mat (integer) array of size at least n+1.
6653: Level: advanced
6656: .seealso: MatDestroySubMatrices(), MatCreateSubMatrix(), MatGetRow(), MatGetDiagonal(), MatReuse
6657: @*/
6658: PetscErrorCode MatCreateSubMatrices(Mat mat,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *submat[])
6659: {
6661: PetscInt i;
6662: PetscBool eq;
6667: if (n) {
6672: }
6674: if (n && scall == MAT_REUSE_MATRIX) {
6677: }
6678: if (!mat->ops->createsubmatrices) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6679: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6680: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6681: MatCheckPreallocated(mat,1);
6683: PetscLogEventBegin(MAT_CreateSubMats,mat,0,0,0);
6684: (*mat->ops->createsubmatrices)(mat,n,irow,icol,scall,submat);
6685: PetscLogEventEnd(MAT_CreateSubMats,mat,0,0,0);
6686: for (i=0; i<n; i++) {
6687: (*submat)[i]->factortype = MAT_FACTOR_NONE; /* in case in place factorization was previously done on submatrix */
6688: if (mat->symmetric || mat->structurally_symmetric || mat->hermitian) {
6689: ISEqual(irow[i],icol[i],&eq);
6690: if (eq) {
6691: if (mat->symmetric) {
6692: MatSetOption((*submat)[i],MAT_SYMMETRIC,PETSC_TRUE);
6693: } else if (mat->hermitian) {
6694: MatSetOption((*submat)[i],MAT_HERMITIAN,PETSC_TRUE);
6695: } else if (mat->structurally_symmetric) {
6696: MatSetOption((*submat)[i],MAT_STRUCTURALLY_SYMMETRIC,PETSC_TRUE);
6697: }
6698: }
6699: }
6700: }
6701: return(0);
6702: }
6704: /*@C
6705: MatCreateSubMatricesMPI - Extracts MPI submatrices across a sub communicator of mat (by pairs of IS that may live on subcomms).
6707: Collective on Mat
6709: Input Parameters:
6710: + mat - the matrix
6711: . n - the number of submatrixes to be extracted
6712: . irow, icol - index sets of rows and columns to extract
6713: - scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
6715: Output Parameter:
6716: . submat - the array of submatrices
6718: Level: advanced
6721: .seealso: MatCreateSubMatrices(), MatCreateSubMatrix(), MatGetRow(), MatGetDiagonal(), MatReuse
6722: @*/
6723: PetscErrorCode MatCreateSubMatricesMPI(Mat mat,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *submat[])
6724: {
6726: PetscInt i;
6727: PetscBool eq;
6732: if (n) {
6737: }
6739: if (n && scall == MAT_REUSE_MATRIX) {
6742: }
6743: if (!mat->ops->createsubmatricesmpi) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6744: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6745: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6746: MatCheckPreallocated(mat,1);
6748: PetscLogEventBegin(MAT_CreateSubMats,mat,0,0,0);
6749: (*mat->ops->createsubmatricesmpi)(mat,n,irow,icol,scall,submat);
6750: PetscLogEventEnd(MAT_CreateSubMats,mat,0,0,0);
6751: for (i=0; i<n; i++) {
6752: if (mat->symmetric || mat->structurally_symmetric || mat->hermitian) {
6753: ISEqual(irow[i],icol[i],&eq);
6754: if (eq) {
6755: if (mat->symmetric) {
6756: MatSetOption((*submat)[i],MAT_SYMMETRIC,PETSC_TRUE);
6757: } else if (mat->hermitian) {
6758: MatSetOption((*submat)[i],MAT_HERMITIAN,PETSC_TRUE);
6759: } else if (mat->structurally_symmetric) {
6760: MatSetOption((*submat)[i],MAT_STRUCTURALLY_SYMMETRIC,PETSC_TRUE);
6761: }
6762: }
6763: }
6764: }
6765: return(0);
6766: }
6768: /*@C
6769: MatDestroyMatrices - Destroys an array of matrices.
6771: Collective on Mat
6773: Input Parameters:
6774: + n - the number of local matrices
6775: - mat - the matrices (note that this is a pointer to the array of matrices)
6777: Level: advanced
6779: Notes:
6780: Frees not only the matrices, but also the array that contains the matrices
6781: In Fortran will not free the array.
6783: .seealso: MatCreateSubMatrices() MatDestroySubMatrices()
6784: @*/
6785: PetscErrorCode MatDestroyMatrices(PetscInt n,Mat *mat[])
6786: {
6788: PetscInt i;
6791: if (!*mat) return(0);
6792: if (n < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Trying to destroy negative number of matrices %D",n);
6795: for (i=0; i<n; i++) {
6796: MatDestroy(&(*mat)[i]);
6797: }
6799: /* memory is allocated even if n = 0 */
6800: PetscFree(*mat);
6801: return(0);
6802: }
6804: /*@C
6805: MatDestroySubMatrices - Destroys a set of matrices obtained with MatCreateSubMatrices().
6807: Collective on Mat
6809: Input Parameters:
6810: + n - the number of local matrices
6811: - mat - the matrices (note that this is a pointer to the array of matrices, just to match the calling
6812: sequence of MatCreateSubMatrices())
6814: Level: advanced
6816: Notes:
6817: Frees not only the matrices, but also the array that contains the matrices
6818: In Fortran will not free the array.
6820: .seealso: MatCreateSubMatrices()
6821: @*/
6822: PetscErrorCode MatDestroySubMatrices(PetscInt n,Mat *mat[])
6823: {
6825: Mat mat0;
6828: if (!*mat) return(0);
6829: /* mat[] is an array of length n+1, see MatCreateSubMatrices_xxx() */
6830: if (n < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Trying to destroy negative number of matrices %D",n);
6833: mat0 = (*mat)[0];
6834: if (mat0 && mat0->ops->destroysubmatrices) {
6835: (mat0->ops->destroysubmatrices)(n,mat);
6836: } else {
6837: MatDestroyMatrices(n,mat);
6838: }
6839: return(0);
6840: }
6842: /*@C
6843: MatGetSeqNonzeroStructure - Extracts the sequential nonzero structure from a matrix.
6845: Collective on Mat
6847: Input Parameters:
6848: . mat - the matrix
6850: Output Parameter:
6851: . matstruct - the sequential matrix with the nonzero structure of mat
6853: Level: intermediate
6855: .seealso: MatDestroySeqNonzeroStructure(), MatCreateSubMatrices(), MatDestroyMatrices()
6856: @*/
6857: PetscErrorCode MatGetSeqNonzeroStructure(Mat mat,Mat *matstruct)
6858: {
6866: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6867: MatCheckPreallocated(mat,1);
6869: if (!mat->ops->getseqnonzerostructure) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Not for matrix type %s\n",((PetscObject)mat)->type_name);
6870: PetscLogEventBegin(MAT_GetSeqNonzeroStructure,mat,0,0,0);
6871: (*mat->ops->getseqnonzerostructure)(mat,matstruct);
6872: PetscLogEventEnd(MAT_GetSeqNonzeroStructure,mat,0,0,0);
6873: return(0);
6874: }
6876: /*@C
6877: MatDestroySeqNonzeroStructure - Destroys matrix obtained with MatGetSeqNonzeroStructure().
6879: Collective on Mat
6881: Input Parameters:
6882: . mat - the matrix (note that this is a pointer to the array of matrices, just to match the calling
6883: sequence of MatGetSequentialNonzeroStructure())
6885: Level: advanced
6887: Notes:
6888: Frees not only the matrices, but also the array that contains the matrices
6890: .seealso: MatGetSeqNonzeroStructure()
6891: @*/
6892: PetscErrorCode MatDestroySeqNonzeroStructure(Mat *mat)
6893: {
6898: MatDestroy(mat);
6899: return(0);
6900: }
6902: /*@
6903: MatIncreaseOverlap - Given a set of submatrices indicated by index sets,
6904: replaces the index sets by larger ones that represent submatrices with
6905: additional overlap.
6907: Collective on Mat
6909: Input Parameters:
6910: + mat - the matrix
6911: . n - the number of index sets
6912: . is - the array of index sets (these index sets will changed during the call)
6913: - ov - the additional overlap requested
6915: Options Database:
6916: . -mat_increase_overlap_scalable - use a scalable algorithm to compute the overlap (supported by MPIAIJ matrix)
6918: Level: developer
6921: .seealso: MatCreateSubMatrices()
6922: @*/
6923: PetscErrorCode MatIncreaseOverlap(Mat mat,PetscInt n,IS is[],PetscInt ov)
6924: {
6930: if (n < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Must have one or more domains, you have %D",n);
6931: if (n) {
6934: }
6935: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6936: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6937: MatCheckPreallocated(mat,1);
6939: if (!ov) return(0);
6940: if (!mat->ops->increaseoverlap) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6941: PetscLogEventBegin(MAT_IncreaseOverlap,mat,0,0,0);
6942: (*mat->ops->increaseoverlap)(mat,n,is,ov);
6943: PetscLogEventEnd(MAT_IncreaseOverlap,mat,0,0,0);
6944: return(0);
6945: }
6948: PetscErrorCode MatIncreaseOverlapSplit_Single(Mat,IS*,PetscInt);
6950: /*@
6951: MatIncreaseOverlapSplit - Given a set of submatrices indicated by index sets across
6952: a sub communicator, replaces the index sets by larger ones that represent submatrices with
6953: additional overlap.
6955: Collective on Mat
6957: Input Parameters:
6958: + mat - the matrix
6959: . n - the number of index sets
6960: . is - the array of index sets (these index sets will changed during the call)
6961: - ov - the additional overlap requested
6963: Options Database:
6964: . -mat_increase_overlap_scalable - use a scalable algorithm to compute the overlap (supported by MPIAIJ matrix)
6966: Level: developer
6969: .seealso: MatCreateSubMatrices()
6970: @*/
6971: PetscErrorCode MatIncreaseOverlapSplit(Mat mat,PetscInt n,IS is[],PetscInt ov)
6972: {
6973: PetscInt i;
6979: if (n < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Must have one or more domains, you have %D",n);
6980: if (n) {
6983: }
6984: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6985: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6986: MatCheckPreallocated(mat,1);
6987: if (!ov) return(0);
6988: PetscLogEventBegin(MAT_IncreaseOverlap,mat,0,0,0);
6989: for(i=0; i<n; i++){
6990: MatIncreaseOverlapSplit_Single(mat,&is[i],ov);
6991: }
6992: PetscLogEventEnd(MAT_IncreaseOverlap,mat,0,0,0);
6993: return(0);
6994: }
6999: /*@
7000: MatGetBlockSize - Returns the matrix block size.
7002: Not Collective
7004: Input Parameter:
7005: . mat - the matrix
7007: Output Parameter:
7008: . bs - block size
7010: Notes:
7011: Block row formats are MATSEQBAIJ, MATMPIBAIJ, MATSEQSBAIJ, MATMPISBAIJ. These formats ALWAYS have square block storage in the matrix.
7013: If the block size has not been set yet this routine returns 1.
7015: Level: intermediate
7017: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSizes()
7018: @*/
7019: PetscErrorCode MatGetBlockSize(Mat mat,PetscInt *bs)
7020: {
7024: *bs = PetscAbs(mat->rmap->bs);
7025: return(0);
7026: }
7028: /*@
7029: MatGetBlockSizes - Returns the matrix block row and column sizes.
7031: Not Collective
7033: Input Parameter:
7034: . mat - the matrix
7036: Output Parameter:
7037: + rbs - row block size
7038: - cbs - column block size
7040: Notes:
7041: Block row formats are MATSEQBAIJ, MATMPIBAIJ, MATSEQSBAIJ, MATMPISBAIJ. These formats ALWAYS have square block storage in the matrix.
7042: If you pass a different block size for the columns than the rows, the row block size determines the square block storage.
7044: If a block size has not been set yet this routine returns 1.
7046: Level: intermediate
7048: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSize(), MatSetBlockSize(), MatSetBlockSizes()
7049: @*/
7050: PetscErrorCode MatGetBlockSizes(Mat mat,PetscInt *rbs, PetscInt *cbs)
7051: {
7056: if (rbs) *rbs = PetscAbs(mat->rmap->bs);
7057: if (cbs) *cbs = PetscAbs(mat->cmap->bs);
7058: return(0);
7059: }
7061: /*@
7062: MatSetBlockSize - Sets the matrix block size.
7064: Logically Collective on Mat
7066: Input Parameters:
7067: + mat - the matrix
7068: - bs - block size
7070: Notes:
7071: Block row formats are MATSEQBAIJ, MATMPIBAIJ, MATSEQSBAIJ, MATMPISBAIJ. These formats ALWAYS have square block storage in the matrix.
7072: This must be called before MatSetUp() or MatXXXSetPreallocation() (or will default to 1) and the block size cannot be changed later.
7074: For MATMPIAIJ and MATSEQAIJ matrix formats, this function can be called at a later stage, provided that the specified block size
7075: is compatible with the matrix local sizes.
7077: Level: intermediate
7079: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSize(), MatSetBlockSizes(), MatGetBlockSizes()
7080: @*/
7081: PetscErrorCode MatSetBlockSize(Mat mat,PetscInt bs)
7082: {
7088: MatSetBlockSizes(mat,bs,bs);
7089: return(0);
7090: }
7092: /*@
7093: MatSetVariableBlockSizes - Sets a diagonal blocks of the matrix that need not be of the same size
7095: Logically Collective on Mat
7097: Input Parameters:
7098: + mat - the matrix
7099: . nblocks - the number of blocks on this process
7100: - bsizes - the block sizes
7102: Notes:
7103: Currently used by PCVPBJACOBI for SeqAIJ matrices
7105: Level: intermediate
7107: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSize(), MatSetBlockSizes(), MatGetBlockSizes(), MatGetVariableBlockSizes()
7108: @*/
7109: PetscErrorCode MatSetVariableBlockSizes(Mat mat,PetscInt nblocks,PetscInt *bsizes)
7110: {
7112: PetscInt i,ncnt = 0, nlocal;
7116: if (nblocks < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Number of local blocks must be great than or equal to zero");
7117: MatGetLocalSize(mat,&nlocal,NULL);
7118: for (i=0; i<nblocks; i++) ncnt += bsizes[i];
7119: 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);
7120: PetscFree(mat->bsizes);
7121: mat->nblocks = nblocks;
7122: PetscMalloc1(nblocks,&mat->bsizes);
7123: PetscArraycpy(mat->bsizes,bsizes,nblocks);
7124: return(0);
7125: }
7127: /*@C
7128: MatGetVariableBlockSizes - Gets a diagonal blocks of the matrix that need not be of the same size
7130: Logically Collective on Mat
7132: Input Parameters:
7133: . mat - the matrix
7135: Output Parameters:
7136: + nblocks - the number of blocks on this process
7137: - bsizes - the block sizes
7139: Notes: Currently not supported from Fortran
7141: Level: intermediate
7143: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSize(), MatSetBlockSizes(), MatGetBlockSizes(), MatSetVariableBlockSizes()
7144: @*/
7145: PetscErrorCode MatGetVariableBlockSizes(Mat mat,PetscInt *nblocks,const PetscInt **bsizes)
7146: {
7149: *nblocks = mat->nblocks;
7150: *bsizes = mat->bsizes;
7151: return(0);
7152: }
7154: /*@
7155: MatSetBlockSizes - Sets the matrix block row and column sizes.
7157: Logically Collective on Mat
7159: Input Parameters:
7160: + mat - the matrix
7161: - rbs - row block size
7162: - cbs - column block size
7164: Notes:
7165: Block row formats are MATSEQBAIJ, MATMPIBAIJ, MATSEQSBAIJ, MATMPISBAIJ. These formats ALWAYS have square block storage in the matrix.
7166: If you pass a different block size for the columns than the rows, the row block size determines the square block storage.
7167: This must be called before MatSetUp() or MatXXXSetPreallocation() (or will default to 1) and the block size cannot be changed later
7169: For MATMPIAIJ and MATSEQAIJ matrix formats, this function can be called at a later stage, provided that the specified block sizes
7170: are compatible with the matrix local sizes.
7172: The row and column block size determine the blocksize of the "row" and "column" vectors returned by MatCreateVecs().
7174: Level: intermediate
7176: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSize(), MatSetBlockSize(), MatGetBlockSizes()
7177: @*/
7178: PetscErrorCode MatSetBlockSizes(Mat mat,PetscInt rbs,PetscInt cbs)
7179: {
7186: if (mat->ops->setblocksizes) {
7187: (*mat->ops->setblocksizes)(mat,rbs,cbs);
7188: }
7189: if (mat->rmap->refcnt) {
7190: ISLocalToGlobalMapping l2g = NULL;
7191: PetscLayout nmap = NULL;
7193: PetscLayoutDuplicate(mat->rmap,&nmap);
7194: if (mat->rmap->mapping) {
7195: ISLocalToGlobalMappingDuplicate(mat->rmap->mapping,&l2g);
7196: }
7197: PetscLayoutDestroy(&mat->rmap);
7198: mat->rmap = nmap;
7199: mat->rmap->mapping = l2g;
7200: }
7201: if (mat->cmap->refcnt) {
7202: ISLocalToGlobalMapping l2g = NULL;
7203: PetscLayout nmap = NULL;
7205: PetscLayoutDuplicate(mat->cmap,&nmap);
7206: if (mat->cmap->mapping) {
7207: ISLocalToGlobalMappingDuplicate(mat->cmap->mapping,&l2g);
7208: }
7209: PetscLayoutDestroy(&mat->cmap);
7210: mat->cmap = nmap;
7211: mat->cmap->mapping = l2g;
7212: }
7213: PetscLayoutSetBlockSize(mat->rmap,rbs);
7214: PetscLayoutSetBlockSize(mat->cmap,cbs);
7215: return(0);
7216: }
7218: /*@
7219: MatSetBlockSizesFromMats - Sets the matrix block row and column sizes to match a pair of matrices
7221: Logically Collective on Mat
7223: Input Parameters:
7224: + mat - the matrix
7225: . fromRow - matrix from which to copy row block size
7226: - fromCol - matrix from which to copy column block size (can be same as fromRow)
7228: Level: developer
7230: .seealso: MatCreateSeqBAIJ(), MatCreateBAIJ(), MatGetBlockSize(), MatSetBlockSizes()
7231: @*/
7232: PetscErrorCode MatSetBlockSizesFromMats(Mat mat,Mat fromRow,Mat fromCol)
7233: {
7240: if (fromRow->rmap->bs > 0) {PetscLayoutSetBlockSize(mat->rmap,fromRow->rmap->bs);}
7241: if (fromCol->cmap->bs > 0) {PetscLayoutSetBlockSize(mat->cmap,fromCol->cmap->bs);}
7242: return(0);
7243: }
7245: /*@
7246: MatResidual - Default routine to calculate the residual.
7248: Collective on Mat
7250: Input Parameters:
7251: + mat - the matrix
7252: . b - the right-hand-side
7253: - x - the approximate solution
7255: Output Parameter:
7256: . r - location to store the residual
7258: Level: developer
7260: .seealso: PCMGSetResidual()
7261: @*/
7262: PetscErrorCode MatResidual(Mat mat,Vec b,Vec x,Vec r)
7263: {
7272: MatCheckPreallocated(mat,1);
7273: PetscLogEventBegin(MAT_Residual,mat,0,0,0);
7274: if (!mat->ops->residual) {
7275: MatMult(mat,x,r);
7276: VecAYPX(r,-1.0,b);
7277: } else {
7278: (*mat->ops->residual)(mat,b,x,r);
7279: }
7280: PetscLogEventEnd(MAT_Residual,mat,0,0,0);
7281: return(0);
7282: }
7284: /*@C
7285: MatGetRowIJ - Returns the compressed row storage i and j indices for sequential matrices.
7287: Collective on Mat
7289: Input Parameters:
7290: + mat - the matrix
7291: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
7292: . symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be symmetrized
7293: - inodecompressed - PETSC_TRUE or PETSC_FALSE indicating if the nonzero structure of the
7294: inodes or the nonzero elements is wanted. For BAIJ matrices the compressed version is
7295: always used.
7297: Output Parameters:
7298: + n - number of rows in the (possibly compressed) matrix
7299: . ia - the row pointers; that is ia[0] = 0, ia[row] = ia[row-1] + number of elements in that row of the matrix
7300: . ja - the column indices
7301: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
7302: are responsible for handling the case when done == PETSC_FALSE and ia and ja are not set
7304: Level: developer
7306: Notes:
7307: You CANNOT change any of the ia[] or ja[] values.
7309: Use MatRestoreRowIJ() when you are finished accessing the ia[] and ja[] values.
7311: Fortran Notes:
7312: In Fortran use
7313: $
7314: $ PetscInt ia(1), ja(1)
7315: $ PetscOffset iia, jja
7316: $ call MatGetRowIJ(mat,shift,symmetric,inodecompressed,n,ia,iia,ja,jja,done,ierr)
7317: $ ! Access the ith and jth entries via ia(iia + i) and ja(jja + j)
7319: or
7320: $
7321: $ PetscInt, pointer :: ia(:),ja(:)
7322: $ call MatGetRowIJF90(mat,shift,symmetric,inodecompressed,n,ia,ja,done,ierr)
7323: $ ! Access the ith and jth entries via ia(i) and ja(j)
7325: .seealso: MatGetColumnIJ(), MatRestoreRowIJ(), MatSeqAIJGetArray()
7326: @*/
7327: PetscErrorCode MatGetRowIJ(Mat mat,PetscInt shift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
7328: {
7338: MatCheckPreallocated(mat,1);
7339: if (!mat->ops->getrowij) *done = PETSC_FALSE;
7340: else {
7341: *done = PETSC_TRUE;
7342: PetscLogEventBegin(MAT_GetRowIJ,mat,0,0,0);
7343: (*mat->ops->getrowij)(mat,shift,symmetric,inodecompressed,n,ia,ja,done);
7344: PetscLogEventEnd(MAT_GetRowIJ,mat,0,0,0);
7345: }
7346: return(0);
7347: }
7349: /*@C
7350: MatGetColumnIJ - Returns the compressed column storage i and j indices for sequential matrices.
7352: Collective on Mat
7354: Input Parameters:
7355: + mat - the matrix
7356: . shift - 1 or zero indicating we want the indices starting at 0 or 1
7357: . symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
7358: symmetrized
7359: . inodecompressed - PETSC_TRUE or PETSC_FALSE indicating if the nonzero structure of the
7360: inodes or the nonzero elements is wanted. For BAIJ matrices the compressed version is
7361: always used.
7362: . n - number of columns in the (possibly compressed) matrix
7363: . ia - the column pointers; that is ia[0] = 0, ia[col] = i[col-1] + number of elements in that col of the matrix
7364: - ja - the row indices
7366: Output Parameters:
7367: . done - PETSC_TRUE or PETSC_FALSE, indicating whether the values have been returned
7369: Level: developer
7371: .seealso: MatGetRowIJ(), MatRestoreColumnIJ()
7372: @*/
7373: PetscErrorCode MatGetColumnIJ(Mat mat,PetscInt shift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
7374: {
7384: MatCheckPreallocated(mat,1);
7385: if (!mat->ops->getcolumnij) *done = PETSC_FALSE;
7386: else {
7387: *done = PETSC_TRUE;
7388: (*mat->ops->getcolumnij)(mat,shift,symmetric,inodecompressed,n,ia,ja,done);
7389: }
7390: return(0);
7391: }
7393: /*@C
7394: MatRestoreRowIJ - Call after you are completed with the ia,ja indices obtained with
7395: MatGetRowIJ().
7397: Collective on Mat
7399: Input Parameters:
7400: + mat - the matrix
7401: . shift - 1 or zero indicating we want the indices starting at 0 or 1
7402: . symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
7403: symmetrized
7404: . inodecompressed - PETSC_TRUE or PETSC_FALSE indicating if the nonzero structure of the
7405: inodes or the nonzero elements is wanted. For BAIJ matrices the compressed version is
7406: always used.
7407: . n - size of (possibly compressed) matrix
7408: . ia - the row pointers
7409: - ja - the column indices
7411: Output Parameters:
7412: . done - PETSC_TRUE or PETSC_FALSE indicated that the values have been returned
7414: Note:
7415: This routine zeros out n, ia, and ja. This is to prevent accidental
7416: us of the array after it has been restored. If you pass NULL, it will
7417: not zero the pointers. Use of ia or ja after MatRestoreRowIJ() is invalid.
7419: Level: developer
7421: .seealso: MatGetRowIJ(), MatRestoreColumnIJ()
7422: @*/
7423: PetscErrorCode MatRestoreRowIJ(Mat mat,PetscInt shift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
7424: {
7433: MatCheckPreallocated(mat,1);
7435: if (!mat->ops->restorerowij) *done = PETSC_FALSE;
7436: else {
7437: *done = PETSC_TRUE;
7438: (*mat->ops->restorerowij)(mat,shift,symmetric,inodecompressed,n,ia,ja,done);
7439: if (n) *n = 0;
7440: if (ia) *ia = NULL;
7441: if (ja) *ja = NULL;
7442: }
7443: return(0);
7444: }
7446: /*@C
7447: MatRestoreColumnIJ - Call after you are completed with the ia,ja indices obtained with
7448: MatGetColumnIJ().
7450: Collective on Mat
7452: Input Parameters:
7453: + mat - the matrix
7454: . shift - 1 or zero indicating we want the indices starting at 0 or 1
7455: - symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
7456: symmetrized
7457: - inodecompressed - PETSC_TRUE or PETSC_FALSE indicating if the nonzero structure of the
7458: inodes or the nonzero elements is wanted. For BAIJ matrices the compressed version is
7459: always used.
7461: Output Parameters:
7462: + n - size of (possibly compressed) matrix
7463: . ia - the column pointers
7464: . ja - the row indices
7465: - done - PETSC_TRUE or PETSC_FALSE indicated that the values have been returned
7467: Level: developer
7469: .seealso: MatGetColumnIJ(), MatRestoreRowIJ()
7470: @*/
7471: PetscErrorCode MatRestoreColumnIJ(Mat mat,PetscInt shift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
7472: {
7481: MatCheckPreallocated(mat,1);
7483: if (!mat->ops->restorecolumnij) *done = PETSC_FALSE;
7484: else {
7485: *done = PETSC_TRUE;
7486: (*mat->ops->restorecolumnij)(mat,shift,symmetric,inodecompressed,n,ia,ja,done);
7487: if (n) *n = 0;
7488: if (ia) *ia = NULL;
7489: if (ja) *ja = NULL;
7490: }
7491: return(0);
7492: }
7494: /*@C
7495: MatColoringPatch -Used inside matrix coloring routines that
7496: use MatGetRowIJ() and/or MatGetColumnIJ().
7498: Collective on Mat
7500: Input Parameters:
7501: + mat - the matrix
7502: . ncolors - max color value
7503: . n - number of entries in colorarray
7504: - colorarray - array indicating color for each column
7506: Output Parameters:
7507: . iscoloring - coloring generated using colorarray information
7509: Level: developer
7511: .seealso: MatGetRowIJ(), MatGetColumnIJ()
7513: @*/
7514: PetscErrorCode MatColoringPatch(Mat mat,PetscInt ncolors,PetscInt n,ISColoringValue colorarray[],ISColoring *iscoloring)
7515: {
7523: MatCheckPreallocated(mat,1);
7525: if (!mat->ops->coloringpatch) {
7526: ISColoringCreate(PetscObjectComm((PetscObject)mat),ncolors,n,colorarray,PETSC_OWN_POINTER,iscoloring);
7527: } else {
7528: (*mat->ops->coloringpatch)(mat,ncolors,n,colorarray,iscoloring);
7529: }
7530: return(0);
7531: }
7534: /*@
7535: MatSetUnfactored - Resets a factored matrix to be treated as unfactored.
7537: Logically Collective on Mat
7539: Input Parameter:
7540: . mat - the factored matrix to be reset
7542: Notes:
7543: This routine should be used only with factored matrices formed by in-place
7544: factorization via ILU(0) (or by in-place LU factorization for the MATSEQDENSE
7545: format). This option can save memory, for example, when solving nonlinear
7546: systems with a matrix-free Newton-Krylov method and a matrix-based, in-place
7547: ILU(0) preconditioner.
7549: Note that one can specify in-place ILU(0) factorization by calling
7550: .vb
7551: PCType(pc,PCILU);
7552: PCFactorSeUseInPlace(pc);
7553: .ve
7554: or by using the options -pc_type ilu -pc_factor_in_place
7556: In-place factorization ILU(0) can also be used as a local
7557: solver for the blocks within the block Jacobi or additive Schwarz
7558: methods (runtime option: -sub_pc_factor_in_place). See Users-Manual: ch_pc
7559: for details on setting local solver options.
7561: Most users should employ the simplified KSP interface for linear solvers
7562: instead of working directly with matrix algebra routines such as this.
7563: See, e.g., KSPCreate().
7565: Level: developer
7567: .seealso: PCFactorSetUseInPlace(), PCFactorGetUseInPlace()
7569: @*/
7570: PetscErrorCode MatSetUnfactored(Mat mat)
7571: {
7577: MatCheckPreallocated(mat,1);
7578: mat->factortype = MAT_FACTOR_NONE;
7579: if (!mat->ops->setunfactored) return(0);
7580: (*mat->ops->setunfactored)(mat);
7581: return(0);
7582: }
7584: /*MC
7585: MatDenseGetArrayF90 - Accesses a matrix array from Fortran90.
7587: Synopsis:
7588: MatDenseGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:,:)},integer ierr)
7590: Not collective
7592: Input Parameter:
7593: . x - matrix
7595: Output Parameters:
7596: + xx_v - the Fortran90 pointer to the array
7597: - ierr - error code
7599: Example of Usage:
7600: .vb
7601: PetscScalar, pointer xx_v(:,:)
7602: ....
7603: call MatDenseGetArrayF90(x,xx_v,ierr)
7604: a = xx_v(3)
7605: call MatDenseRestoreArrayF90(x,xx_v,ierr)
7606: .ve
7608: Level: advanced
7610: .seealso: MatDenseRestoreArrayF90(), MatDenseGetArray(), MatDenseRestoreArray(), MatSeqAIJGetArrayF90()
7612: M*/
7614: /*MC
7615: MatDenseRestoreArrayF90 - Restores a matrix array that has been
7616: accessed with MatDenseGetArrayF90().
7618: Synopsis:
7619: MatDenseRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:,:)},integer ierr)
7621: Not collective
7623: Input Parameters:
7624: + x - matrix
7625: - xx_v - the Fortran90 pointer to the array
7627: Output Parameter:
7628: . ierr - error code
7630: Example of Usage:
7631: .vb
7632: PetscScalar, pointer xx_v(:,:)
7633: ....
7634: call MatDenseGetArrayF90(x,xx_v,ierr)
7635: a = xx_v(3)
7636: call MatDenseRestoreArrayF90(x,xx_v,ierr)
7637: .ve
7639: Level: advanced
7641: .seealso: MatDenseGetArrayF90(), MatDenseGetArray(), MatDenseRestoreArray(), MatSeqAIJRestoreArrayF90()
7643: M*/
7646: /*MC
7647: MatSeqAIJGetArrayF90 - Accesses a matrix array from Fortran90.
7649: Synopsis:
7650: MatSeqAIJGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
7652: Not collective
7654: Input Parameter:
7655: . x - matrix
7657: Output Parameters:
7658: + xx_v - the Fortran90 pointer to the array
7659: - ierr - error code
7661: Example of Usage:
7662: .vb
7663: PetscScalar, pointer xx_v(:)
7664: ....
7665: call MatSeqAIJGetArrayF90(x,xx_v,ierr)
7666: a = xx_v(3)
7667: call MatSeqAIJRestoreArrayF90(x,xx_v,ierr)
7668: .ve
7670: Level: advanced
7672: .seealso: MatSeqAIJRestoreArrayF90(), MatSeqAIJGetArray(), MatSeqAIJRestoreArray(), MatDenseGetArrayF90()
7674: M*/
7676: /*MC
7677: MatSeqAIJRestoreArrayF90 - Restores a matrix array that has been
7678: accessed with MatSeqAIJGetArrayF90().
7680: Synopsis:
7681: MatSeqAIJRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
7683: Not collective
7685: Input Parameters:
7686: + x - matrix
7687: - xx_v - the Fortran90 pointer to the array
7689: Output Parameter:
7690: . ierr - error code
7692: Example of Usage:
7693: .vb
7694: PetscScalar, pointer xx_v(:)
7695: ....
7696: call MatSeqAIJGetArrayF90(x,xx_v,ierr)
7697: a = xx_v(3)
7698: call MatSeqAIJRestoreArrayF90(x,xx_v,ierr)
7699: .ve
7701: Level: advanced
7703: .seealso: MatSeqAIJGetArrayF90(), MatSeqAIJGetArray(), MatSeqAIJRestoreArray(), MatDenseRestoreArrayF90()
7705: M*/
7708: /*@
7709: MatCreateSubMatrix - Gets a single submatrix on the same number of processors
7710: as the original matrix.
7712: Collective on Mat
7714: Input Parameters:
7715: + mat - the original matrix
7716: . isrow - parallel IS containing the rows this processor should obtain
7717: . 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.
7718: - cll - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
7720: Output Parameter:
7721: . newmat - the new submatrix, of the same type as the old
7723: Level: advanced
7725: Notes:
7726: The submatrix will be able to be multiplied with vectors using the same layout as iscol.
7728: Some matrix types place restrictions on the row and column indices, such
7729: as that they be sorted or that they be equal to each other.
7731: The index sets may not have duplicate entries.
7733: The first time this is called you should use a cll of MAT_INITIAL_MATRIX,
7734: the MatCreateSubMatrix() routine will create the newmat for you. Any additional calls
7735: to this routine with a mat of the same nonzero structure and with a call of MAT_REUSE_MATRIX
7736: will reuse the matrix generated the first time. You should call MatDestroy() on newmat when
7737: you are finished using it.
7739: The communicator of the newly obtained matrix is ALWAYS the same as the communicator of
7740: the input matrix.
7742: If iscol is NULL then all columns are obtained (not supported in Fortran).
7744: Example usage:
7745: Consider the following 8x8 matrix with 34 non-zero values, that is
7746: assembled across 3 processors. Let's assume that proc0 owns 3 rows,
7747: proc1 owns 3 rows, proc2 owns 2 rows. This division can be shown
7748: as follows:
7750: .vb
7751: 1 2 0 | 0 3 0 | 0 4
7752: Proc0 0 5 6 | 7 0 0 | 8 0
7753: 9 0 10 | 11 0 0 | 12 0
7754: -------------------------------------
7755: 13 0 14 | 15 16 17 | 0 0
7756: Proc1 0 18 0 | 19 20 21 | 0 0
7757: 0 0 0 | 22 23 0 | 24 0
7758: -------------------------------------
7759: Proc2 25 26 27 | 0 0 28 | 29 0
7760: 30 0 0 | 31 32 33 | 0 34
7761: .ve
7763: Suppose isrow = [0 1 | 4 | 6 7] and iscol = [1 2 | 3 4 5 | 6]. The resulting submatrix is
7765: .vb
7766: 2 0 | 0 3 0 | 0
7767: Proc0 5 6 | 7 0 0 | 8
7768: -------------------------------
7769: Proc1 18 0 | 19 20 21 | 0
7770: -------------------------------
7771: Proc2 26 27 | 0 0 28 | 29
7772: 0 0 | 31 32 33 | 0
7773: .ve
7776: .seealso: MatCreateSubMatrices(), MatCreateSubMatricesMPI(), MatCreateSubMatrixVirtual(), MatSubMatrixVirtualUpdate()
7777: @*/
7778: PetscErrorCode MatCreateSubMatrix(Mat mat,IS isrow,IS iscol,MatReuse cll,Mat *newmat)
7779: {
7781: PetscMPIInt size;
7782: Mat *local;
7783: IS iscoltmp;
7792: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7793: if (cll == MAT_IGNORE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Cannot use MAT_IGNORE_MATRIX");
7795: MatCheckPreallocated(mat,1);
7796: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
7798: if (!iscol || isrow == iscol) {
7799: PetscBool stride;
7800: PetscMPIInt grabentirematrix = 0,grab;
7801: PetscObjectTypeCompare((PetscObject)isrow,ISSTRIDE,&stride);
7802: if (stride) {
7803: PetscInt first,step,n,rstart,rend;
7804: ISStrideGetInfo(isrow,&first,&step);
7805: if (step == 1) {
7806: MatGetOwnershipRange(mat,&rstart,&rend);
7807: if (rstart == first) {
7808: ISGetLocalSize(isrow,&n);
7809: if (n == rend-rstart) {
7810: grabentirematrix = 1;
7811: }
7812: }
7813: }
7814: }
7815: MPIU_Allreduce(&grabentirematrix,&grab,1,MPI_INT,MPI_MIN,PetscObjectComm((PetscObject)mat));
7816: if (grab) {
7817: PetscInfo(mat,"Getting entire matrix as submatrix\n");
7818: if (cll == MAT_INITIAL_MATRIX) {
7819: *newmat = mat;
7820: PetscObjectReference((PetscObject)mat);
7821: }
7822: return(0);
7823: }
7824: }
7826: if (!iscol) {
7827: ISCreateStride(PetscObjectComm((PetscObject)mat),mat->cmap->n,mat->cmap->rstart,1,&iscoltmp);
7828: } else {
7829: iscoltmp = iscol;
7830: }
7832: /* if original matrix is on just one processor then use submatrix generated */
7833: if (mat->ops->createsubmatrices && !mat->ops->createsubmatrix && size == 1 && cll == MAT_REUSE_MATRIX) {
7834: MatCreateSubMatrices(mat,1,&isrow,&iscoltmp,MAT_REUSE_MATRIX,&newmat);
7835: goto setproperties;
7836: } else if (mat->ops->createsubmatrices && !mat->ops->createsubmatrix && size == 1) {
7837: MatCreateSubMatrices(mat,1,&isrow,&iscoltmp,MAT_INITIAL_MATRIX,&local);
7838: *newmat = *local;
7839: PetscFree(local);
7840: goto setproperties;
7841: } else if (!mat->ops->createsubmatrix) {
7842: /* Create a new matrix type that implements the operation using the full matrix */
7843: PetscLogEventBegin(MAT_CreateSubMat,mat,0,0,0);
7844: switch (cll) {
7845: case MAT_INITIAL_MATRIX:
7846: MatCreateSubMatrixVirtual(mat,isrow,iscoltmp,newmat);
7847: break;
7848: case MAT_REUSE_MATRIX:
7849: MatSubMatrixVirtualUpdate(*newmat,mat,isrow,iscoltmp);
7850: break;
7851: default: SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"Invalid MatReuse, must be either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX");
7852: }
7853: PetscLogEventEnd(MAT_CreateSubMat,mat,0,0,0);
7854: goto setproperties;
7855: }
7857: if (!mat->ops->createsubmatrix) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
7858: PetscLogEventBegin(MAT_CreateSubMat,mat,0,0,0);
7859: (*mat->ops->createsubmatrix)(mat,isrow,iscoltmp,cll,newmat);
7860: PetscLogEventEnd(MAT_CreateSubMat,mat,0,0,0);
7862: /* Propagate symmetry information for diagonal blocks */
7863: setproperties:
7864: if (isrow == iscoltmp) {
7865: if (mat->symmetric_set && mat->symmetric) {
7866: MatSetOption(*newmat,MAT_SYMMETRIC,PETSC_TRUE);
7867: }
7868: if (mat->structurally_symmetric_set && mat->structurally_symmetric) {
7869: MatSetOption(*newmat,MAT_STRUCTURALLY_SYMMETRIC,PETSC_TRUE);
7870: }
7871: if (mat->hermitian_set && mat->hermitian) {
7872: MatSetOption(*newmat,MAT_HERMITIAN,PETSC_TRUE);
7873: }
7874: if (mat->spd_set && mat->spd) {
7875: MatSetOption(*newmat,MAT_SPD,PETSC_TRUE);
7876: }
7877: }
7879: if (!iscol) {ISDestroy(&iscoltmp);}
7880: if (*newmat && cll == MAT_INITIAL_MATRIX) {PetscObjectStateIncrease((PetscObject)*newmat);}
7881: return(0);
7882: }
7884: /*@
7885: MatStashSetInitialSize - sets the sizes of the matrix stash, that is
7886: used during the assembly process to store values that belong to
7887: other processors.
7889: Not Collective
7891: Input Parameters:
7892: + mat - the matrix
7893: . size - the initial size of the stash.
7894: - bsize - the initial size of the block-stash(if used).
7896: Options Database Keys:
7897: + -matstash_initial_size <size> or <size0,size1,...sizep-1>
7898: - -matstash_block_initial_size <bsize> or <bsize0,bsize1,...bsizep-1>
7900: Level: intermediate
7902: Notes:
7903: The block-stash is used for values set with MatSetValuesBlocked() while
7904: the stash is used for values set with MatSetValues()
7906: Run with the option -info and look for output of the form
7907: MatAssemblyBegin_MPIXXX:Stash has MM entries, uses nn mallocs.
7908: to determine the appropriate value, MM, to use for size and
7909: MatAssemblyBegin_MPIXXX:Block-Stash has BMM entries, uses nn mallocs.
7910: to determine the value, BMM to use for bsize
7913: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), Mat, MatStashGetInfo()
7915: @*/
7916: PetscErrorCode MatStashSetInitialSize(Mat mat,PetscInt size, PetscInt bsize)
7917: {
7923: MatStashSetInitialSize_Private(&mat->stash,size);
7924: MatStashSetInitialSize_Private(&mat->bstash,bsize);
7925: return(0);
7926: }
7928: /*@
7929: MatInterpolateAdd - w = y + A*x or A'*x depending on the shape of
7930: the matrix
7932: Neighbor-wise Collective on Mat
7934: Input Parameters:
7935: + mat - the matrix
7936: . x,y - the vectors
7937: - w - where the result is stored
7939: Level: intermediate
7941: Notes:
7942: w may be the same vector as y.
7944: This allows one to use either the restriction or interpolation (its transpose)
7945: matrix to do the interpolation
7947: .seealso: MatMultAdd(), MatMultTransposeAdd(), MatRestrict()
7949: @*/
7950: PetscErrorCode MatInterpolateAdd(Mat A,Vec x,Vec y,Vec w)
7951: {
7953: PetscInt M,N,Ny;
7961: MatCheckPreallocated(A,1);
7962: MatGetSize(A,&M,&N);
7963: VecGetSize(y,&Ny);
7964: if (M == Ny) {
7965: MatMultAdd(A,x,y,w);
7966: } else {
7967: MatMultTransposeAdd(A,x,y,w);
7968: }
7969: return(0);
7970: }
7972: /*@
7973: MatInterpolate - y = A*x or A'*x depending on the shape of
7974: the matrix
7976: Neighbor-wise Collective on Mat
7978: Input Parameters:
7979: + mat - the matrix
7980: - x,y - the vectors
7982: Level: intermediate
7984: Notes:
7985: This allows one to use either the restriction or interpolation (its transpose)
7986: matrix to do the interpolation
7988: .seealso: MatMultAdd(), MatMultTransposeAdd(), MatRestrict()
7990: @*/
7991: PetscErrorCode MatInterpolate(Mat A,Vec x,Vec y)
7992: {
7994: PetscInt M,N,Ny;
8001: MatCheckPreallocated(A,1);
8002: MatGetSize(A,&M,&N);
8003: VecGetSize(y,&Ny);
8004: if (M == Ny) {
8005: MatMult(A,x,y);
8006: } else {
8007: MatMultTranspose(A,x,y);
8008: }
8009: return(0);
8010: }
8012: /*@
8013: MatRestrict - y = A*x or A'*x
8015: Neighbor-wise Collective on Mat
8017: Input Parameters:
8018: + mat - the matrix
8019: - x,y - the vectors
8021: Level: intermediate
8023: Notes:
8024: This allows one to use either the restriction or interpolation (its transpose)
8025: matrix to do the restriction
8027: .seealso: MatMultAdd(), MatMultTransposeAdd(), MatInterpolate()
8029: @*/
8030: PetscErrorCode MatRestrict(Mat A,Vec x,Vec y)
8031: {
8033: PetscInt M,N,Ny;
8040: MatCheckPreallocated(A,1);
8042: MatGetSize(A,&M,&N);
8043: VecGetSize(y,&Ny);
8044: if (M == Ny) {
8045: MatMult(A,x,y);
8046: } else {
8047: MatMultTranspose(A,x,y);
8048: }
8049: return(0);
8050: }
8052: /*@
8053: MatGetNullSpace - retrieves the null space of a matrix.
8055: Logically Collective on Mat
8057: Input Parameters:
8058: + mat - the matrix
8059: - nullsp - the null space object
8061: Level: developer
8063: .seealso: MatCreate(), MatNullSpaceCreate(), MatSetNearNullSpace(), MatSetNullSpace()
8064: @*/
8065: PetscErrorCode MatGetNullSpace(Mat mat, MatNullSpace *nullsp)
8066: {
8070: *nullsp = (mat->symmetric_set && mat->symmetric && !mat->nullsp) ? mat->transnullsp : mat->nullsp;
8071: return(0);
8072: }
8074: /*@
8075: MatSetNullSpace - attaches a null space to a matrix.
8077: Logically Collective on Mat
8079: Input Parameters:
8080: + mat - the matrix
8081: - nullsp - the null space object
8083: Level: advanced
8085: Notes:
8086: This null space is used by the linear solvers. Overwrites any previous null space that may have been attached
8088: For inconsistent singular systems (linear systems where the right hand side is not in the range of the operator) you also likely should
8089: call MatSetTransposeNullSpace(). This allows the linear system to be solved in a least squares sense.
8091: You can remove the null space by calling this routine with an nullsp of NULL
8094: The fundamental theorem of linear algebra (Gilbert Strang, Introduction to Applied Mathematics, page 72) states that
8095: 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).
8096: 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
8097: 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
8098: 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).
8100: Krylov solvers can produce the minimal norm solution to the least squares problem by utilizing MatNullSpaceRemove().
8102: 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
8103: routine also automatically calls MatSetTransposeNullSpace().
8105: .seealso: MatCreate(), MatNullSpaceCreate(), MatSetNearNullSpace(), MatGetNullSpace(), MatSetTransposeNullSpace(), MatGetTransposeNullSpace(), MatNullSpaceRemove()
8106: @*/
8107: PetscErrorCode MatSetNullSpace(Mat mat,MatNullSpace nullsp)
8108: {
8114: if (nullsp) {PetscObjectReference((PetscObject)nullsp);}
8115: MatNullSpaceDestroy(&mat->nullsp);
8116: mat->nullsp = nullsp;
8117: if (mat->symmetric_set && mat->symmetric) {
8118: MatSetTransposeNullSpace(mat,nullsp);
8119: }
8120: return(0);
8121: }
8123: /*@
8124: MatGetTransposeNullSpace - retrieves the null space of the transpose of a matrix.
8126: Logically Collective on Mat
8128: Input Parameters:
8129: + mat - the matrix
8130: - nullsp - the null space object
8132: Level: developer
8134: .seealso: MatCreate(), MatNullSpaceCreate(), MatSetNearNullSpace(), MatSetTransposeNullSpace(), MatSetNullSpace(), MatGetNullSpace()
8135: @*/
8136: PetscErrorCode MatGetTransposeNullSpace(Mat mat, MatNullSpace *nullsp)
8137: {
8142: *nullsp = (mat->symmetric_set && mat->symmetric && !mat->transnullsp) ? mat->nullsp : mat->transnullsp;
8143: return(0);
8144: }
8146: /*@
8147: MatSetTransposeNullSpace - attaches a null space to a matrix.
8149: Logically Collective on Mat
8151: Input Parameters:
8152: + mat - the matrix
8153: - nullsp - the null space object
8155: Level: advanced
8157: Notes:
8158: 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.
8159: You must also call MatSetNullSpace()
8162: The fundamental theorem of linear algebra (Gilbert Strang, Introduction to Applied Mathematics, page 72) states that
8163: 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).
8164: 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
8165: 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
8166: 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).
8168: Krylov solvers can produce the minimal norm solution to the least squares problem by utilizing MatNullSpaceRemove().
8170: .seealso: MatCreate(), MatNullSpaceCreate(), MatSetNearNullSpace(), MatGetNullSpace(), MatSetNullSpace(), MatGetTransposeNullSpace(), MatNullSpaceRemove()
8171: @*/
8172: PetscErrorCode MatSetTransposeNullSpace(Mat mat,MatNullSpace nullsp)
8173: {
8179: if (nullsp) {PetscObjectReference((PetscObject)nullsp);}
8180: MatNullSpaceDestroy(&mat->transnullsp);
8181: mat->transnullsp = nullsp;
8182: return(0);
8183: }
8185: /*@
8186: MatSetNearNullSpace - attaches a null space to a matrix, which is often the null space (rigid body modes) of the operator without boundary conditions
8187: This null space will be used to provide near null space vectors to a multigrid preconditioner built from this matrix.
8189: Logically Collective on Mat
8191: Input Parameters:
8192: + mat - the matrix
8193: - nullsp - the null space object
8195: Level: advanced
8197: Notes:
8198: Overwrites any previous near null space that may have been attached
8200: You can remove the null space by calling this routine with an nullsp of NULL
8202: .seealso: MatCreate(), MatNullSpaceCreate(), MatSetNullSpace(), MatNullSpaceCreateRigidBody(), MatGetNearNullSpace()
8203: @*/
8204: PetscErrorCode MatSetNearNullSpace(Mat mat,MatNullSpace nullsp)
8205: {
8212: MatCheckPreallocated(mat,1);
8213: if (nullsp) {PetscObjectReference((PetscObject)nullsp);}
8214: MatNullSpaceDestroy(&mat->nearnullsp);
8215: mat->nearnullsp = nullsp;
8216: return(0);
8217: }
8219: /*@
8220: MatGetNearNullSpace - Get null space attached with MatSetNearNullSpace()
8222: Not Collective
8224: Input Parameter:
8225: . mat - the matrix
8227: Output Parameter:
8228: . nullsp - the null space object, NULL if not set
8230: Level: developer
8232: .seealso: MatSetNearNullSpace(), MatGetNullSpace(), MatNullSpaceCreate()
8233: @*/
8234: PetscErrorCode MatGetNearNullSpace(Mat mat,MatNullSpace *nullsp)
8235: {
8240: MatCheckPreallocated(mat,1);
8241: *nullsp = mat->nearnullsp;
8242: return(0);
8243: }
8245: /*@C
8246: MatICCFactor - Performs in-place incomplete Cholesky factorization of matrix.
8248: Collective on Mat
8250: Input Parameters:
8251: + mat - the matrix
8252: . row - row/column permutation
8253: . fill - expected fill factor >= 1.0
8254: - level - level of fill, for ICC(k)
8256: Notes:
8257: Probably really in-place only when level of fill is zero, otherwise allocates
8258: new space to store factored matrix and deletes previous memory.
8260: Most users should employ the simplified KSP interface for linear solvers
8261: instead of working directly with matrix algebra routines such as this.
8262: See, e.g., KSPCreate().
8264: Level: developer
8267: .seealso: MatICCFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor()
8269: Developer Note: fortran interface is not autogenerated as the f90
8270: interface defintion cannot be generated correctly [due to MatFactorInfo]
8272: @*/
8273: PetscErrorCode MatICCFactor(Mat mat,IS row,const MatFactorInfo *info)
8274: {
8282: if (mat->rmap->N != mat->cmap->N) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONG,"matrix must be square");
8283: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
8284: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
8285: if (!mat->ops->iccfactor) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
8286: MatCheckPreallocated(mat,1);
8287: (*mat->ops->iccfactor)(mat,row,info);
8288: PetscObjectStateIncrease((PetscObject)mat);
8289: return(0);
8290: }
8292: /*@
8293: MatDiagonalScaleLocal - Scales columns of a matrix given the scaling values including the
8294: ghosted ones.
8296: Not Collective
8298: Input Parameters:
8299: + mat - the matrix
8300: - diag = the diagonal values, including ghost ones
8302: Level: developer
8304: Notes:
8305: Works only for MPIAIJ and MPIBAIJ matrices
8307: .seealso: MatDiagonalScale()
8308: @*/
8309: PetscErrorCode MatDiagonalScaleLocal(Mat mat,Vec diag)
8310: {
8312: PetscMPIInt size;
8319: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Matrix must be already assembled");
8320: PetscLogEventBegin(MAT_Scale,mat,0,0,0);
8321: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
8322: if (size == 1) {
8323: PetscInt n,m;
8324: VecGetSize(diag,&n);
8325: MatGetSize(mat,0,&m);
8326: if (m == n) {
8327: MatDiagonalScale(mat,0,diag);
8328: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only supported for sequential matrices when no ghost points/periodic conditions");
8329: } else {
8330: PetscUseMethod(mat,"MatDiagonalScaleLocal_C",(Mat,Vec),(mat,diag));
8331: }
8332: PetscLogEventEnd(MAT_Scale,mat,0,0,0);
8333: PetscObjectStateIncrease((PetscObject)mat);
8334: return(0);
8335: }
8337: /*@
8338: MatGetInertia - Gets the inertia from a factored matrix
8340: Collective on Mat
8342: Input Parameter:
8343: . mat - the matrix
8345: Output Parameters:
8346: + nneg - number of negative eigenvalues
8347: . nzero - number of zero eigenvalues
8348: - npos - number of positive eigenvalues
8350: Level: advanced
8352: Notes:
8353: Matrix must have been factored by MatCholeskyFactor()
8356: @*/
8357: PetscErrorCode MatGetInertia(Mat mat,PetscInt *nneg,PetscInt *nzero,PetscInt *npos)
8358: {
8364: if (!mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
8365: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Numeric factor mat is not assembled");
8366: if (!mat->ops->getinertia) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
8367: (*mat->ops->getinertia)(mat,nneg,nzero,npos);
8368: return(0);
8369: }
8371: /* ----------------------------------------------------------------*/
8372: /*@C
8373: MatSolves - Solves A x = b, given a factored matrix, for a collection of vectors
8375: Neighbor-wise Collective on Mats
8377: Input Parameters:
8378: + mat - the factored matrix
8379: - b - the right-hand-side vectors
8381: Output Parameter:
8382: . x - the result vectors
8384: Notes:
8385: The vectors b and x cannot be the same. I.e., one cannot
8386: call MatSolves(A,x,x).
8388: Notes:
8389: Most users should employ the simplified KSP interface for linear solvers
8390: instead of working directly with matrix algebra routines such as this.
8391: See, e.g., KSPCreate().
8393: Level: developer
8395: .seealso: MatSolveAdd(), MatSolveTranspose(), MatSolveTransposeAdd(), MatSolve()
8396: @*/
8397: PetscErrorCode MatSolves(Mat mat,Vecs b,Vecs x)
8398: {
8404: if (x == b) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_IDN,"x and b must be different vectors");
8405: if (!mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
8406: if (!mat->rmap->N && !mat->cmap->N) return(0);
8408: if (!mat->ops->solves) SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
8409: MatCheckPreallocated(mat,1);
8410: PetscLogEventBegin(MAT_Solves,mat,0,0,0);
8411: (*mat->ops->solves)(mat,b,x);
8412: PetscLogEventEnd(MAT_Solves,mat,0,0,0);
8413: return(0);
8414: }
8416: /*@
8417: MatIsSymmetric - Test whether a matrix is symmetric
8419: Collective on Mat
8421: Input Parameter:
8422: + A - the matrix to test
8423: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact transpose)
8425: Output Parameters:
8426: . flg - the result
8428: Notes:
8429: For real numbers MatIsSymmetric() and MatIsHermitian() return identical results
8431: Level: intermediate
8433: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetricKnown()
8434: @*/
8435: PetscErrorCode MatIsSymmetric(Mat A,PetscReal tol,PetscBool *flg)
8436: {
8443: if (!A->symmetric_set) {
8444: if (!A->ops->issymmetric) {
8445: MatType mattype;
8446: MatGetType(A,&mattype);
8447: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix of type %s does not support checking for symmetric",mattype);
8448: }
8449: (*A->ops->issymmetric)(A,tol,flg);
8450: if (!tol) {
8451: A->symmetric_set = PETSC_TRUE;
8452: A->symmetric = *flg;
8453: if (A->symmetric) {
8454: A->structurally_symmetric_set = PETSC_TRUE;
8455: A->structurally_symmetric = PETSC_TRUE;
8456: }
8457: }
8458: } else if (A->symmetric) {
8459: *flg = PETSC_TRUE;
8460: } else if (!tol) {
8461: *flg = PETSC_FALSE;
8462: } else {
8463: if (!A->ops->issymmetric) {
8464: MatType mattype;
8465: MatGetType(A,&mattype);
8466: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix of type %s does not support checking for symmetric",mattype);
8467: }
8468: (*A->ops->issymmetric)(A,tol,flg);
8469: }
8470: return(0);
8471: }
8473: /*@
8474: MatIsHermitian - Test whether a matrix is Hermitian
8476: Collective on Mat
8478: Input Parameter:
8479: + A - the matrix to test
8480: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact Hermitian)
8482: Output Parameters:
8483: . flg - the result
8485: Level: intermediate
8487: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(),
8488: MatIsSymmetricKnown(), MatIsSymmetric()
8489: @*/
8490: PetscErrorCode MatIsHermitian(Mat A,PetscReal tol,PetscBool *flg)
8491: {
8498: if (!A->hermitian_set) {
8499: if (!A->ops->ishermitian) {
8500: MatType mattype;
8501: MatGetType(A,&mattype);
8502: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix of type %s does not support checking for hermitian",mattype);
8503: }
8504: (*A->ops->ishermitian)(A,tol,flg);
8505: if (!tol) {
8506: A->hermitian_set = PETSC_TRUE;
8507: A->hermitian = *flg;
8508: if (A->hermitian) {
8509: A->structurally_symmetric_set = PETSC_TRUE;
8510: A->structurally_symmetric = PETSC_TRUE;
8511: }
8512: }
8513: } else if (A->hermitian) {
8514: *flg = PETSC_TRUE;
8515: } else if (!tol) {
8516: *flg = PETSC_FALSE;
8517: } else {
8518: if (!A->ops->ishermitian) {
8519: MatType mattype;
8520: MatGetType(A,&mattype);
8521: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix of type %s does not support checking for hermitian",mattype);
8522: }
8523: (*A->ops->ishermitian)(A,tol,flg);
8524: }
8525: return(0);
8526: }
8528: /*@
8529: MatIsSymmetricKnown - Checks the flag on the matrix to see if it is symmetric.
8531: Not Collective
8533: Input Parameter:
8534: . A - the matrix to check
8536: Output Parameters:
8537: + set - if the symmetric flag is set (this tells you if the next flag is valid)
8538: - flg - the result
8540: Level: advanced
8542: Note: Does not check the matrix values directly, so this may return unknown (set = PETSC_FALSE). Use MatIsSymmetric()
8543: if you want it explicitly checked
8545: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetric()
8546: @*/
8547: PetscErrorCode MatIsSymmetricKnown(Mat A,PetscBool *set,PetscBool *flg)
8548: {
8553: if (A->symmetric_set) {
8554: *set = PETSC_TRUE;
8555: *flg = A->symmetric;
8556: } else {
8557: *set = PETSC_FALSE;
8558: }
8559: return(0);
8560: }
8562: /*@
8563: MatIsHermitianKnown - Checks the flag on the matrix to see if it is hermitian.
8565: Not Collective
8567: Input Parameter:
8568: . A - the matrix to check
8570: Output Parameters:
8571: + set - if the hermitian flag is set (this tells you if the next flag is valid)
8572: - flg - the result
8574: Level: advanced
8576: Note: Does not check the matrix values directly, so this may return unknown (set = PETSC_FALSE). Use MatIsHermitian()
8577: if you want it explicitly checked
8579: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetric()
8580: @*/
8581: PetscErrorCode MatIsHermitianKnown(Mat A,PetscBool *set,PetscBool *flg)
8582: {
8587: if (A->hermitian_set) {
8588: *set = PETSC_TRUE;
8589: *flg = A->hermitian;
8590: } else {
8591: *set = PETSC_FALSE;
8592: }
8593: return(0);
8594: }
8596: /*@
8597: MatIsStructurallySymmetric - Test whether a matrix is structurally symmetric
8599: Collective on Mat
8601: Input Parameter:
8602: . A - the matrix to test
8604: Output Parameters:
8605: . flg - the result
8607: Level: intermediate
8609: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsSymmetric(), MatSetOption()
8610: @*/
8611: PetscErrorCode MatIsStructurallySymmetric(Mat A,PetscBool *flg)
8612: {
8618: if (!A->structurally_symmetric_set) {
8619: 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);
8620: (*A->ops->isstructurallysymmetric)(A,&A->structurally_symmetric);
8622: A->structurally_symmetric_set = PETSC_TRUE;
8623: }
8624: *flg = A->structurally_symmetric;
8625: return(0);
8626: }
8628: /*@
8629: MatStashGetInfo - Gets how many values are currently in the matrix stash, i.e. need
8630: to be communicated to other processors during the MatAssemblyBegin/End() process
8632: Not collective
8634: Input Parameter:
8635: . vec - the vector
8637: Output Parameters:
8638: + nstash - the size of the stash
8639: . reallocs - the number of additional mallocs incurred.
8640: . bnstash - the size of the block stash
8641: - breallocs - the number of additional mallocs incurred.in the block stash
8643: Level: advanced
8645: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), Mat, MatStashSetInitialSize()
8647: @*/
8648: PetscErrorCode MatStashGetInfo(Mat mat,PetscInt *nstash,PetscInt *reallocs,PetscInt *bnstash,PetscInt *breallocs)
8649: {
8653: MatStashGetInfo_Private(&mat->stash,nstash,reallocs);
8654: MatStashGetInfo_Private(&mat->bstash,bnstash,breallocs);
8655: return(0);
8656: }
8658: /*@C
8659: MatCreateVecs - Get vector(s) compatible with the matrix, i.e. with the same
8660: parallel layout
8662: Collective on Mat
8664: Input Parameter:
8665: . mat - the matrix
8667: Output Parameter:
8668: + right - (optional) vector that the matrix can be multiplied against
8669: - left - (optional) vector that the matrix vector product can be stored in
8671: Notes:
8672: 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().
8674: Notes:
8675: These are new vectors which are not owned by the Mat, they should be destroyed in VecDestroy() when no longer needed
8677: Level: advanced
8679: .seealso: MatCreate(), VecDestroy()
8680: @*/
8681: PetscErrorCode MatCreateVecs(Mat mat,Vec *right,Vec *left)
8682: {
8688: if (mat->ops->getvecs) {
8689: (*mat->ops->getvecs)(mat,right,left);
8690: } else {
8691: PetscInt rbs,cbs;
8692: MatGetBlockSizes(mat,&rbs,&cbs);
8693: if (right) {
8694: if (mat->cmap->n < 0) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"PetscLayout for columns not yet setup");
8695: VecCreate(PetscObjectComm((PetscObject)mat),right);
8696: VecSetSizes(*right,mat->cmap->n,PETSC_DETERMINE);
8697: VecSetBlockSize(*right,cbs);
8698: VecSetType(*right,mat->defaultvectype);
8699: PetscLayoutReference(mat->cmap,&(*right)->map);
8700: }
8701: if (left) {
8702: if (mat->rmap->n < 0) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"PetscLayout for rows not yet setup");
8703: VecCreate(PetscObjectComm((PetscObject)mat),left);
8704: VecSetSizes(*left,mat->rmap->n,PETSC_DETERMINE);
8705: VecSetBlockSize(*left,rbs);
8706: VecSetType(*left,mat->defaultvectype);
8707: PetscLayoutReference(mat->rmap,&(*left)->map);
8708: }
8709: }
8710: return(0);
8711: }
8713: /*@C
8714: MatFactorInfoInitialize - Initializes a MatFactorInfo data structure
8715: with default values.
8717: Not Collective
8719: Input Parameters:
8720: . info - the MatFactorInfo data structure
8723: Notes:
8724: The solvers are generally used through the KSP and PC objects, for example
8725: PCLU, PCILU, PCCHOLESKY, PCICC
8727: Level: developer
8729: .seealso: MatFactorInfo
8731: Developer Note: fortran interface is not autogenerated as the f90
8732: interface defintion cannot be generated correctly [due to MatFactorInfo]
8734: @*/
8736: PetscErrorCode MatFactorInfoInitialize(MatFactorInfo *info)
8737: {
8741: PetscMemzero(info,sizeof(MatFactorInfo));
8742: return(0);
8743: }
8745: /*@
8746: MatFactorSetSchurIS - Set indices corresponding to the Schur complement you wish to have computed
8748: Collective on Mat
8750: Input Parameters:
8751: + mat - the factored matrix
8752: - is - the index set defining the Schur indices (0-based)
8754: Notes:
8755: Call MatFactorSolveSchurComplement() or MatFactorSolveSchurComplementTranspose() after this call to solve a Schur complement system.
8757: You can call MatFactorGetSchurComplement() or MatFactorCreateSchurComplement() after this call.
8759: Level: developer
8761: .seealso: MatGetFactor(), MatFactorGetSchurComplement(), MatFactorRestoreSchurComplement(), MatFactorCreateSchurComplement(), MatFactorSolveSchurComplement(),
8762: MatFactorSolveSchurComplementTranspose(), MatFactorSolveSchurComplement()
8764: @*/
8765: PetscErrorCode MatFactorSetSchurIS(Mat mat,IS is)
8766: {
8767: PetscErrorCode ierr,(*f)(Mat,IS);
8775: if (!mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Only for factored matrix");
8776: PetscObjectQueryFunction((PetscObject)mat,"MatFactorSetSchurIS_C",&f);
8777: 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");
8778: MatDestroy(&mat->schur);
8779: (*f)(mat,is);
8780: if (!mat->schur) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_PLIB,"Schur complement has not been created");
8781: return(0);
8782: }
8784: /*@
8785: MatFactorCreateSchurComplement - Create a Schur complement matrix object using Schur data computed during the factorization step
8787: Logically Collective on Mat
8789: Input Parameters:
8790: + F - the factored matrix obtained by calling MatGetFactor() from PETSc-MUMPS interface
8791: . S - location where to return the Schur complement, can be NULL
8792: - status - the status of the Schur complement matrix, can be NULL
8794: Notes:
8795: You must call MatFactorSetSchurIS() before calling this routine.
8797: The routine provides a copy of the Schur matrix stored within the solver data structures.
8798: The caller must destroy the object when it is no longer needed.
8799: If MatFactorInvertSchurComplement() has been called, the routine gets back the inverse.
8801: 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)
8803: Developer Notes:
8804: The reason this routine exists is because the representation of the Schur complement within the factor matrix may be different than a standard PETSc
8805: matrix representation and we normally do not want to use the time or memory to make a copy as a regular PETSc matrix.
8807: See MatCreateSchurComplement() or MatGetSchurComplement() for ways to create virtual or approximate Schur complements.
8809: Level: advanced
8811: References:
8813: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorGetSchurComplement(), MatFactorSchurStatus
8814: @*/
8815: PetscErrorCode MatFactorCreateSchurComplement(Mat F,Mat* S,MatFactorSchurStatus* status)
8816: {
8823: if (S) {
8824: PetscErrorCode (*f)(Mat,Mat*);
8826: PetscObjectQueryFunction((PetscObject)F,"MatFactorCreateSchurComplement_C",&f);
8827: if (f) {
8828: (*f)(F,S);
8829: } else {
8830: MatDuplicate(F->schur,MAT_COPY_VALUES,S);
8831: }
8832: }
8833: if (status) *status = F->schur_status;
8834: return(0);
8835: }
8837: /*@
8838: MatFactorGetSchurComplement - Gets access to a Schur complement matrix using the current Schur data within a factored matrix
8840: Logically Collective on Mat
8842: Input Parameters:
8843: + F - the factored matrix obtained by calling MatGetFactor()
8844: . *S - location where to return the Schur complement, can be NULL
8845: - status - the status of the Schur complement matrix, can be NULL
8847: Notes:
8848: You must call MatFactorSetSchurIS() before calling this routine.
8850: Schur complement mode is currently implemented for sequential matrices.
8851: The routine returns a the Schur Complement stored within the data strutures of the solver.
8852: If MatFactorInvertSchurComplement() has previously been called, the returned matrix is actually the inverse of the Schur complement.
8853: The returned matrix should not be destroyed; the caller should call MatFactorRestoreSchurComplement() when the object is no longer needed.
8855: Use MatFactorCreateSchurComplement() to create a copy of the Schur complement matrix that is within a factored matrix
8857: See MatCreateSchurComplement() or MatGetSchurComplement() for ways to create virtual or approximate Schur complements.
8859: Level: advanced
8861: References:
8863: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorRestoreSchurComplement(), MatFactorCreateSchurComplement(), MatFactorSchurStatus
8864: @*/
8865: PetscErrorCode MatFactorGetSchurComplement(Mat F,Mat* S,MatFactorSchurStatus* status)
8866: {
8871: if (S) *S = F->schur;
8872: if (status) *status = F->schur_status;
8873: return(0);
8874: }
8876: /*@
8877: MatFactorRestoreSchurComplement - Restore the Schur complement matrix object obtained from a call to MatFactorGetSchurComplement
8879: Logically Collective on Mat
8881: Input Parameters:
8882: + F - the factored matrix obtained by calling MatGetFactor()
8883: . *S - location where the Schur complement is stored
8884: - status - the status of the Schur complement matrix (see MatFactorSchurStatus)
8886: Notes:
8888: Level: advanced
8890: References:
8892: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorRestoreSchurComplement(), MatFactorCreateSchurComplement(), MatFactorSchurStatus
8893: @*/
8894: PetscErrorCode MatFactorRestoreSchurComplement(Mat F,Mat* S,MatFactorSchurStatus status)
8895: {
8900: if (S) {
8902: *S = NULL;
8903: }
8904: F->schur_status = status;
8905: MatFactorUpdateSchurStatus_Private(F);
8906: return(0);
8907: }
8909: /*@
8910: MatFactorSolveSchurComplementTranspose - Solve the transpose of the Schur complement system computed during the factorization step
8912: Logically Collective on Mat
8914: Input Parameters:
8915: + F - the factored matrix obtained by calling MatGetFactor()
8916: . rhs - location where the right hand side of the Schur complement system is stored
8917: - sol - location where the solution of the Schur complement system has to be returned
8919: Notes:
8920: The sizes of the vectors should match the size of the Schur complement
8922: Must be called after MatFactorSetSchurIS()
8924: Level: advanced
8926: References:
8928: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorSolveSchurComplement()
8929: @*/
8930: PetscErrorCode MatFactorSolveSchurComplementTranspose(Mat F, Vec rhs, Vec sol)
8931: {
8943: MatFactorFactorizeSchurComplement(F);
8944: switch (F->schur_status) {
8945: case MAT_FACTOR_SCHUR_FACTORED:
8946: MatSolveTranspose(F->schur,rhs,sol);
8947: break;
8948: case MAT_FACTOR_SCHUR_INVERTED:
8949: MatMultTranspose(F->schur,rhs,sol);
8950: break;
8951: default:
8952: SETERRQ1(PetscObjectComm((PetscObject)F),PETSC_ERR_SUP,"Unhandled MatFactorSchurStatus %D",F->schur_status);
8953: break;
8954: }
8955: return(0);
8956: }
8958: /*@
8959: MatFactorSolveSchurComplement - Solve the Schur complement system computed during the factorization step
8961: Logically Collective on Mat
8963: Input Parameters:
8964: + F - the factored matrix obtained by calling MatGetFactor()
8965: . rhs - location where the right hand side of the Schur complement system is stored
8966: - sol - location where the solution of the Schur complement system has to be returned
8968: Notes:
8969: The sizes of the vectors should match the size of the Schur complement
8971: Must be called after MatFactorSetSchurIS()
8973: Level: advanced
8975: References:
8977: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorSolveSchurComplementTranspose()
8978: @*/
8979: PetscErrorCode MatFactorSolveSchurComplement(Mat F, Vec rhs, Vec sol)
8980: {
8992: MatFactorFactorizeSchurComplement(F);
8993: switch (F->schur_status) {
8994: case MAT_FACTOR_SCHUR_FACTORED:
8995: MatSolve(F->schur,rhs,sol);
8996: break;
8997: case MAT_FACTOR_SCHUR_INVERTED:
8998: MatMult(F->schur,rhs,sol);
8999: break;
9000: default:
9001: SETERRQ1(PetscObjectComm((PetscObject)F),PETSC_ERR_SUP,"Unhandled MatFactorSchurStatus %D",F->schur_status);
9002: break;
9003: }
9004: return(0);
9005: }
9007: /*@
9008: MatFactorInvertSchurComplement - Invert the Schur complement matrix computed during the factorization step
9010: Logically Collective on Mat
9012: Input Parameters:
9013: . F - the factored matrix obtained by calling MatGetFactor()
9015: Notes:
9016: Must be called after MatFactorSetSchurIS().
9018: Call MatFactorGetSchurComplement() or MatFactorCreateSchurComplement() AFTER this call to actually compute the inverse and get access to it.
9020: Level: advanced
9022: References:
9024: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorGetSchurComplement(), MatFactorCreateSchurComplement()
9025: @*/
9026: PetscErrorCode MatFactorInvertSchurComplement(Mat F)
9027: {
9033: if (F->schur_status == MAT_FACTOR_SCHUR_INVERTED) return(0);
9034: MatFactorFactorizeSchurComplement(F);
9035: MatFactorInvertSchurComplement_Private(F);
9036: F->schur_status = MAT_FACTOR_SCHUR_INVERTED;
9037: return(0);
9038: }
9040: /*@
9041: MatFactorFactorizeSchurComplement - Factorize the Schur complement matrix computed during the factorization step
9043: Logically Collective on Mat
9045: Input Parameters:
9046: . F - the factored matrix obtained by calling MatGetFactor()
9048: Notes:
9049: Must be called after MatFactorSetSchurIS().
9051: Level: advanced
9053: References:
9055: .seealso: MatGetFactor(), MatFactorSetSchurIS(), MatFactorInvertSchurComplement()
9056: @*/
9057: PetscErrorCode MatFactorFactorizeSchurComplement(Mat F)
9058: {
9064: if (F->schur_status == MAT_FACTOR_SCHUR_INVERTED || F->schur_status == MAT_FACTOR_SCHUR_FACTORED) return(0);
9065: MatFactorFactorizeSchurComplement_Private(F);
9066: F->schur_status = MAT_FACTOR_SCHUR_FACTORED;
9067: return(0);
9068: }
9070: PetscErrorCode MatPtAP_Basic(Mat A,Mat P,MatReuse scall,PetscReal fill,Mat *C)
9071: {
9072: Mat AP;
9076: PetscInfo2(A,"Mat types %s and %s using basic PtAP\n",((PetscObject)A)->type_name,((PetscObject)P)->type_name);
9077: MatMatMult(A,P,MAT_INITIAL_MATRIX,PETSC_DEFAULT,&AP);
9078: MatTransposeMatMult(P,AP,scall,fill,C);
9079: MatDestroy(&AP);
9080: return(0);
9081: }
9083: /*@
9084: MatPtAP - Creates the matrix product C = P^T * A * P
9086: Neighbor-wise Collective on Mat
9088: Input Parameters:
9089: + A - the matrix
9090: . P - the projection matrix
9091: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9092: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(P)), use PETSC_DEFAULT if you do not have a good estimate
9093: if the result is a dense matrix this is irrelevent
9095: Output Parameters:
9096: . C - the product matrix
9098: Notes:
9099: C will be created and must be destroyed by the user with MatDestroy().
9101: For matrix types without special implementation the function fallbacks to MatMatMult() followed by MatTransposeMatMult().
9103: Level: intermediate
9105: .seealso: MatPtAPSymbolic(), MatPtAPNumeric(), MatMatMult(), MatRARt()
9106: @*/
9107: PetscErrorCode MatPtAP(Mat A,Mat P,MatReuse scall,PetscReal fill,Mat *C)
9108: {
9110: PetscErrorCode (*fA)(Mat,Mat,MatReuse,PetscReal,Mat*);
9111: PetscErrorCode (*fP)(Mat,Mat,MatReuse,PetscReal,Mat*);
9112: PetscErrorCode (*ptap)(Mat,Mat,MatReuse,PetscReal,Mat*)=NULL;
9113: PetscBool sametype;
9118: MatCheckPreallocated(A,1);
9119: if (scall == MAT_INPLACE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Inplace product not supported");
9120: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9121: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9124: MatCheckPreallocated(P,2);
9125: if (!P->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9126: if (P->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9128: if (A->rmap->N != A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix A must be square, %D != %D",A->rmap->N,A->cmap->N);
9129: if (P->rmap->N != A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->rmap->N,A->cmap->N);
9130: if (fill == PETSC_DEFAULT || fill == PETSC_DECIDE) fill = 2.0;
9131: if (fill < 1.0) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Expected fill=%g must be >= 1.0",(double)fill);
9133: if (scall == MAT_REUSE_MATRIX) {
9137: PetscLogEventBegin(MAT_PtAP,A,P,0,0);
9138: PetscLogEventBegin(MAT_PtAPNumeric,A,P,0,0);
9139: if ((*C)->ops->ptapnumeric) {
9140: (*(*C)->ops->ptapnumeric)(A,P,*C);
9141: } else {
9142: MatPtAP_Basic(A,P,scall,fill,C);
9143: }
9144: PetscLogEventEnd(MAT_PtAPNumeric,A,P,0,0);
9145: PetscLogEventEnd(MAT_PtAP,A,P,0,0);
9146: return(0);
9147: }
9149: if (fill == PETSC_DEFAULT || fill == PETSC_DECIDE) fill = 2.0;
9150: if (fill < 1.0) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Expected fill=%g must be >= 1.0",(double)fill);
9152: fA = A->ops->ptap;
9153: fP = P->ops->ptap;
9154: PetscStrcmp(((PetscObject)A)->type_name,((PetscObject)P)->type_name,&sametype);
9155: if (fP == fA && sametype) {
9156: ptap = fA;
9157: } else {
9158: /* dispatch based on the type of A and P from their PetscObject's PetscFunctionLists. */
9159: char ptapname[256];
9160: PetscStrncpy(ptapname,"MatPtAP_",sizeof(ptapname));
9161: PetscStrlcat(ptapname,((PetscObject)A)->type_name,sizeof(ptapname));
9162: PetscStrlcat(ptapname,"_",sizeof(ptapname));
9163: PetscStrlcat(ptapname,((PetscObject)P)->type_name,sizeof(ptapname));
9164: PetscStrlcat(ptapname,"_C",sizeof(ptapname)); /* e.g., ptapname = "MatPtAP_seqdense_seqaij_C" */
9165: PetscObjectQueryFunction((PetscObject)P,ptapname,&ptap);
9166: }
9168: if (!ptap) ptap = MatPtAP_Basic;
9169: PetscLogEventBegin(MAT_PtAP,A,P,0,0);
9170: (*ptap)(A,P,scall,fill,C);
9171: PetscLogEventEnd(MAT_PtAP,A,P,0,0);
9172: if (A->symmetric_set && A->symmetric) {
9173: MatSetOption(*C,MAT_SYMMETRIC,PETSC_TRUE);
9174: }
9175: return(0);
9176: }
9178: /*@
9179: MatPtAPNumeric - Computes the matrix product C = P^T * A * P
9181: Neighbor-wise Collective on Mat
9183: Input Parameters:
9184: + A - the matrix
9185: - P - the projection matrix
9187: Output Parameters:
9188: . C - the product matrix
9190: Notes:
9191: C must have been created by calling MatPtAPSymbolic and must be destroyed by
9192: the user using MatDeatroy().
9194: This routine is currently only implemented for pairs of AIJ matrices and classes
9195: which inherit from AIJ. C will be of type MATAIJ.
9197: Level: intermediate
9199: .seealso: MatPtAP(), MatPtAPSymbolic(), MatMatMultNumeric()
9200: @*/
9201: PetscErrorCode MatPtAPNumeric(Mat A,Mat P,Mat C)
9202: {
9208: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9209: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9212: MatCheckPreallocated(P,2);
9213: if (!P->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9214: if (P->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9217: MatCheckPreallocated(C,3);
9218: if (C->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9219: if (P->cmap->N!=C->rmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->cmap->N,C->rmap->N);
9220: if (P->rmap->N!=A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->rmap->N,A->cmap->N);
9221: if (A->rmap->N!=A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix 'A' must be square, %D != %D",A->rmap->N,A->cmap->N);
9222: if (P->cmap->N!=C->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->cmap->N,C->cmap->N);
9223: MatCheckPreallocated(A,1);
9225: if (!C->ops->ptapnumeric) SETERRQ(PetscObjectComm((PetscObject)C),PETSC_ERR_ARG_WRONGSTATE,"MatPtAPNumeric implementation is missing. You should call MatPtAPSymbolic first");
9226: PetscLogEventBegin(MAT_PtAPNumeric,A,P,0,0);
9227: (*C->ops->ptapnumeric)(A,P,C);
9228: PetscLogEventEnd(MAT_PtAPNumeric,A,P,0,0);
9229: return(0);
9230: }
9232: /*@
9233: MatPtAPSymbolic - Creates the (i,j) structure of the matrix product C = P^T * A * P
9235: Neighbor-wise Collective on Mat
9237: Input Parameters:
9238: + A - the matrix
9239: - P - the projection matrix
9241: Output Parameters:
9242: . C - the (i,j) structure of the product matrix
9244: Notes:
9245: C will be created and must be destroyed by the user with MatDestroy().
9247: This routine is currently only implemented for pairs of SeqAIJ matrices and classes
9248: which inherit from SeqAIJ. C will be of type MATSEQAIJ. The product is computed using
9249: this (i,j) structure by calling MatPtAPNumeric().
9251: Level: intermediate
9253: .seealso: MatPtAP(), MatPtAPNumeric(), MatMatMultSymbolic()
9254: @*/
9255: PetscErrorCode MatPtAPSymbolic(Mat A,Mat P,PetscReal fill,Mat *C)
9256: {
9262: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9263: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9264: if (fill <1.0) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Expected fill=%g must be >= 1.0",(double)fill);
9267: MatCheckPreallocated(P,2);
9268: if (!P->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9269: if (P->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9272: if (P->rmap->N!=A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->rmap->N,A->cmap->N);
9273: if (A->rmap->N!=A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix 'A' must be square, %D != %D",A->rmap->N,A->cmap->N);
9274: MatCheckPreallocated(A,1);
9276: if (!A->ops->ptapsymbolic) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"MatType %s",((PetscObject)A)->type_name);
9277: PetscLogEventBegin(MAT_PtAPSymbolic,A,P,0,0);
9278: (*A->ops->ptapsymbolic)(A,P,fill,C);
9279: PetscLogEventEnd(MAT_PtAPSymbolic,A,P,0,0);
9281: /* MatSetBlockSize(*C,A->rmap->bs); NO! this is not always true -ma */
9282: return(0);
9283: }
9285: /*@
9286: MatRARt - Creates the matrix product C = R * A * R^T
9288: Neighbor-wise Collective on Mat
9290: Input Parameters:
9291: + A - the matrix
9292: . R - the projection matrix
9293: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9294: - fill - expected fill as ratio of nnz(C)/nnz(A), use PETSC_DEFAULT if you do not have a good estimate
9295: if the result is a dense matrix this is irrelevent
9297: Output Parameters:
9298: . C - the product matrix
9300: Notes:
9301: C will be created and must be destroyed by the user with MatDestroy().
9303: This routine is currently only implemented for pairs of AIJ matrices and classes
9304: which inherit from AIJ. Due to PETSc sparse matrix block row distribution among processes,
9305: parallel MatRARt is implemented via explicit transpose of R, which could be very expensive.
9306: We recommend using MatPtAP().
9308: Level: intermediate
9310: .seealso: MatRARtSymbolic(), MatRARtNumeric(), MatMatMult(), MatPtAP()
9311: @*/
9312: PetscErrorCode MatRARt(Mat A,Mat R,MatReuse scall,PetscReal fill,Mat *C)
9313: {
9319: if (scall == MAT_INPLACE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Inplace product not supported");
9320: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9321: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9324: MatCheckPreallocated(R,2);
9325: if (!R->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9326: if (R->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9328: if (R->cmap->N!=A->rmap->N) SETERRQ2(PetscObjectComm((PetscObject)R),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",R->cmap->N,A->rmap->N);
9330: if (fill == PETSC_DEFAULT || fill == PETSC_DECIDE) fill = 2.0;
9331: if (fill < 1.0) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Expected fill=%g must be >= 1.0",(double)fill);
9332: MatCheckPreallocated(A,1);
9334: if (!A->ops->rart) {
9335: Mat Rt;
9336: MatTranspose(R,MAT_INITIAL_MATRIX,&Rt);
9337: MatMatMatMult(R,A,Rt,scall,fill,C);
9338: MatDestroy(&Rt);
9339: return(0);
9340: }
9341: PetscLogEventBegin(MAT_RARt,A,R,0,0);
9342: (*A->ops->rart)(A,R,scall,fill,C);
9343: PetscLogEventEnd(MAT_RARt,A,R,0,0);
9344: return(0);
9345: }
9347: /*@
9348: MatRARtNumeric - Computes the matrix product C = R * A * R^T
9350: Neighbor-wise Collective on Mat
9352: Input Parameters:
9353: + A - the matrix
9354: - R - the projection matrix
9356: Output Parameters:
9357: . C - the product matrix
9359: Notes:
9360: C must have been created by calling MatRARtSymbolic and must be destroyed by
9361: the user using MatDestroy().
9363: This routine is currently only implemented for pairs of AIJ matrices and classes
9364: which inherit from AIJ. C will be of type MATAIJ.
9366: Level: intermediate
9368: .seealso: MatRARt(), MatRARtSymbolic(), MatMatMultNumeric()
9369: @*/
9370: PetscErrorCode MatRARtNumeric(Mat A,Mat R,Mat C)
9371: {
9377: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9378: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9381: MatCheckPreallocated(R,2);
9382: if (!R->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9383: if (R->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9386: MatCheckPreallocated(C,3);
9387: if (C->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9388: if (R->rmap->N!=C->rmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",R->rmap->N,C->rmap->N);
9389: if (R->cmap->N!=A->rmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",R->cmap->N,A->rmap->N);
9390: if (A->rmap->N!=A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix 'A' must be square, %D != %D",A->rmap->N,A->cmap->N);
9391: if (R->rmap->N!=C->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",R->rmap->N,C->cmap->N);
9392: MatCheckPreallocated(A,1);
9394: PetscLogEventBegin(MAT_RARtNumeric,A,R,0,0);
9395: (*A->ops->rartnumeric)(A,R,C);
9396: PetscLogEventEnd(MAT_RARtNumeric,A,R,0,0);
9397: return(0);
9398: }
9400: /*@
9401: MatRARtSymbolic - Creates the (i,j) structure of the matrix product C = R * A * R^T
9403: Neighbor-wise Collective on Mat
9405: Input Parameters:
9406: + A - the matrix
9407: - R - the projection matrix
9409: Output Parameters:
9410: . C - the (i,j) structure of the product matrix
9412: Notes:
9413: C will be created and must be destroyed by the user with MatDestroy().
9415: This routine is currently only implemented for pairs of SeqAIJ matrices and classes
9416: which inherit from SeqAIJ. C will be of type MATSEQAIJ. The product is computed using
9417: this (i,j) structure by calling MatRARtNumeric().
9419: Level: intermediate
9421: .seealso: MatRARt(), MatRARtNumeric(), MatMatMultSymbolic()
9422: @*/
9423: PetscErrorCode MatRARtSymbolic(Mat A,Mat R,PetscReal fill,Mat *C)
9424: {
9430: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9431: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9432: if (fill <1.0) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Expected fill=%g must be >= 1.0",(double)fill);
9435: MatCheckPreallocated(R,2);
9436: if (!R->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9437: if (R->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9440: if (R->cmap->N!=A->rmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",R->cmap->N,A->rmap->N);
9441: if (A->rmap->N!=A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix 'A' must be square, %D != %D",A->rmap->N,A->cmap->N);
9442: MatCheckPreallocated(A,1);
9443: PetscLogEventBegin(MAT_RARtSymbolic,A,R,0,0);
9444: (*A->ops->rartsymbolic)(A,R,fill,C);
9445: PetscLogEventEnd(MAT_RARtSymbolic,A,R,0,0);
9447: MatSetBlockSizes(*C,PetscAbs(R->rmap->bs),PetscAbs(R->rmap->bs));
9448: return(0);
9449: }
9451: /*@
9452: MatMatMult - Performs Matrix-Matrix Multiplication C=A*B.
9454: Neighbor-wise Collective on Mat
9456: Input Parameters:
9457: + A - the left matrix
9458: . B - the right matrix
9459: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9460: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use PETSC_DEFAULT if you do not have a good estimate
9461: if the result is a dense matrix this is irrelevent
9463: Output Parameters:
9464: . C - the product matrix
9466: Notes:
9467: Unless scall is MAT_REUSE_MATRIX C will be created.
9469: 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
9470: call to this function with either MAT_INITIAL_MATRIX or MatMatMultSymbolic()
9472: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
9473: actually needed.
9475: If you have many matrices with the same non-zero structure to multiply, you
9476: should either
9477: $ 1) use MAT_REUSE_MATRIX in all calls but the first or
9478: $ 2) call MatMatMultSymbolic() once and then MatMatMultNumeric() for each product needed
9479: 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
9480: with MAT_REUSE_MATRIX, rather than first having MatMatMult() create it for you. You can NEVER do this if the matrix C is sparse.
9482: Level: intermediate
9484: .seealso: MatMatMultSymbolic(), MatMatMultNumeric(), MatTransposeMatMult(), MatMatTransposeMult(), MatPtAP()
9485: @*/
9486: PetscErrorCode MatMatMult(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
9487: {
9489: PetscErrorCode (*fA)(Mat,Mat,MatReuse,PetscReal,Mat*);
9490: PetscErrorCode (*fB)(Mat,Mat,MatReuse,PetscReal,Mat*);
9491: PetscErrorCode (*mult)(Mat,Mat,MatReuse,PetscReal,Mat*)=NULL;
9492: Mat T;
9493: PetscBool istrans;
9498: MatCheckPreallocated(A,1);
9499: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9500: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9503: MatCheckPreallocated(B,2);
9504: if (!B->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9505: if (B->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9507: if (scall == MAT_INPLACE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Inplace product not supported");
9508: if (B->rmap->N!=A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->rmap->N,A->cmap->N);
9509: PetscObjectTypeCompare((PetscObject)A,MATTRANSPOSEMAT,&istrans);
9510: if (istrans) {
9511: MatTransposeGetMat(A,&T);
9512: MatTransposeMatMult(T,B,scall,fill,C);
9513: return(0);
9514: } else {
9515: PetscObjectTypeCompare((PetscObject)B,MATTRANSPOSEMAT,&istrans);
9516: if (istrans) {
9517: MatTransposeGetMat(B,&T);
9518: MatMatTransposeMult(A,T,scall,fill,C);
9519: return(0);
9520: }
9521: }
9522: if (scall == MAT_REUSE_MATRIX) {
9525: PetscLogEventBegin(MAT_MatMult,A,B,0,0);
9526: PetscLogEventBegin(MAT_MatMultNumeric,A,B,0,0);
9527: (*(*C)->ops->matmultnumeric)(A,B,*C);
9528: PetscLogEventEnd(MAT_MatMultNumeric,A,B,0,0);
9529: PetscLogEventEnd(MAT_MatMult,A,B,0,0);
9530: return(0);
9531: }
9532: if (fill == PETSC_DEFAULT || fill == PETSC_DECIDE) fill = 2.0;
9533: if (fill < 1.0) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Expected fill=%g must be >= 1.0",(double)fill);
9535: fA = A->ops->matmult;
9536: fB = B->ops->matmult;
9537: if (fB == fA && fB) mult = fB;
9538: else {
9539: /* dispatch based on the type of A and B from their PetscObject's PetscFunctionLists. */
9540: char multname[256];
9541: PetscStrncpy(multname,"MatMatMult_",sizeof(multname));
9542: PetscStrlcat(multname,((PetscObject)A)->type_name,sizeof(multname));
9543: PetscStrlcat(multname,"_",sizeof(multname));
9544: PetscStrlcat(multname,((PetscObject)B)->type_name,sizeof(multname));
9545: PetscStrlcat(multname,"_C",sizeof(multname)); /* e.g., multname = "MatMatMult_seqdense_seqaij_C" */
9546: PetscObjectQueryFunction((PetscObject)B,multname,&mult);
9547: if (!mult) {
9548: PetscObjectQueryFunction((PetscObject)A,multname,&mult);
9549: }
9550: if (!mult) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_INCOMP,"MatMatMult requires A, %s, to be compatible with B, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
9551: }
9552: PetscLogEventBegin(MAT_MatMult,A,B,0,0);
9553: (*mult)(A,B,scall,fill,C);
9554: PetscLogEventEnd(MAT_MatMult,A,B,0,0);
9555: return(0);
9556: }
9558: /*@
9559: MatMatMultSymbolic - Performs construction, preallocation, and computes the ij structure
9560: of the matrix-matrix product C=A*B. Call this routine before calling MatMatMultNumeric().
9562: Neighbor-wise Collective on Mat
9564: Input Parameters:
9565: + A - the left matrix
9566: . B - the right matrix
9567: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use PETSC_DEFAULT if you do not have a good estimate,
9568: if C is a dense matrix this is irrelevent
9570: Output Parameters:
9571: . C - the product matrix
9573: Notes:
9574: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
9575: actually needed.
9577: This routine is currently implemented for
9578: - pairs of AIJ matrices and classes which inherit from AIJ, C will be of type AIJ
9579: - pairs of AIJ (A) and Dense (B) matrix, C will be of type Dense.
9580: - pairs of Dense (A) and AIJ (B) matrix, C will be of type Dense.
9582: Level: intermediate
9584: Developers Note: There are ways to estimate the number of nonzeros in the resulting product, see for example, https://arxiv.org/abs/1006.4173
9585: We should incorporate them into PETSc.
9587: .seealso: MatMatMult(), MatMatMultNumeric()
9588: @*/
9589: PetscErrorCode MatMatMultSymbolic(Mat A,Mat B,PetscReal fill,Mat *C)
9590: {
9592: PetscErrorCode (*Asymbolic)(Mat,Mat,PetscReal,Mat*);
9593: PetscErrorCode (*Bsymbolic)(Mat,Mat,PetscReal,Mat*);
9594: PetscErrorCode (*symbolic)(Mat,Mat,PetscReal,Mat*)=NULL;
9599: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9600: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9604: MatCheckPreallocated(B,2);
9605: if (!B->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9606: if (B->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9609: if (B->rmap->N!=A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->rmap->N,A->cmap->N);
9610: if (fill == PETSC_DEFAULT) fill = 2.0;
9611: if (fill < 1.0) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Expected fill=%g must be > 1.0",(double)fill);
9612: MatCheckPreallocated(A,1);
9614: Asymbolic = A->ops->matmultsymbolic;
9615: Bsymbolic = B->ops->matmultsymbolic;
9616: if (Asymbolic == Bsymbolic) {
9617: if (!Bsymbolic) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"C=A*B not implemented for B of type %s",((PetscObject)B)->type_name);
9618: symbolic = Bsymbolic;
9619: } else { /* dispatch based on the type of A and B */
9620: char symbolicname[256];
9621: PetscStrncpy(symbolicname,"MatMatMultSymbolic_",sizeof(symbolicname));
9622: PetscStrlcat(symbolicname,((PetscObject)A)->type_name,sizeof(symbolicname));
9623: PetscStrlcat(symbolicname,"_",sizeof(symbolicname));
9624: PetscStrlcat(symbolicname,((PetscObject)B)->type_name,sizeof(symbolicname));
9625: PetscStrlcat(symbolicname,"_C",sizeof(symbolicname));
9626: PetscObjectQueryFunction((PetscObject)B,symbolicname,&symbolic);
9627: if (!symbolic) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_INCOMP,"MatMatMultSymbolic requires A, %s, to be compatible with B, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
9628: }
9629: PetscLogEventBegin(MAT_MatMultSymbolic,A,B,0,0);
9630: *C = NULL;
9631: (*symbolic)(A,B,fill,C);
9632: PetscLogEventEnd(MAT_MatMultSymbolic,A,B,0,0);
9633: return(0);
9634: }
9636: /*@
9637: MatMatMultNumeric - Performs the numeric matrix-matrix product.
9638: Call this routine after first calling MatMatMultSymbolic().
9640: Neighbor-wise Collective on Mat
9642: Input Parameters:
9643: + A - the left matrix
9644: - B - the right matrix
9646: Output Parameters:
9647: . C - the product matrix, which was created by from MatMatMultSymbolic() or a call to MatMatMult().
9649: Notes:
9650: C must have been created with MatMatMultSymbolic().
9652: This routine is currently implemented for
9653: - pairs of AIJ matrices and classes which inherit from AIJ, C will be of type MATAIJ.
9654: - pairs of AIJ (A) and Dense (B) matrix, C will be of type Dense.
9655: - pairs of Dense (A) and AIJ (B) matrix, C will be of type Dense.
9657: Level: intermediate
9659: .seealso: MatMatMult(), MatMatMultSymbolic()
9660: @*/
9661: PetscErrorCode MatMatMultNumeric(Mat A,Mat B,Mat C)
9662: {
9666: MatMatMult(A,B,MAT_REUSE_MATRIX,0.0,&C);
9667: return(0);
9668: }
9670: /*@
9671: MatMatTransposeMult - Performs Matrix-Matrix Multiplication C=A*B^T.
9673: Neighbor-wise Collective on Mat
9675: Input Parameters:
9676: + A - the left matrix
9677: . B - the right matrix
9678: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9679: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use PETSC_DEFAULT if not known
9681: Output Parameters:
9682: . C - the product matrix
9684: Notes:
9685: C will be created if MAT_INITIAL_MATRIX and must be destroyed by the user with MatDestroy().
9687: MAT_REUSE_MATRIX can only be used if the matrices A and B have the same nonzero pattern as in the previous call
9689: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
9690: actually needed.
9692: This routine is currently only implemented for pairs of SeqAIJ matrices, for the SeqDense class,
9693: and for pairs of MPIDense matrices.
9695: Options Database Keys:
9696: . -matmattransmult_mpidense_mpidense_via {allgatherv,cyclic} - Choose between algorthims for MPIDense matrices: the
9697: first redundantly copies the transposed B matrix on each process and requiers O(log P) communication complexity;
9698: the second never stores more than one portion of the B matrix at a time by requires O(P) communication complexity.
9700: Level: intermediate
9702: .seealso: MatMatTransposeMultSymbolic(), MatMatTransposeMultNumeric(), MatMatMult(), MatTransposeMatMult() MatPtAP()
9703: @*/
9704: PetscErrorCode MatMatTransposeMult(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
9705: {
9707: PetscErrorCode (*fA)(Mat,Mat,MatReuse,PetscReal,Mat*);
9708: PetscErrorCode (*fB)(Mat,Mat,MatReuse,PetscReal,Mat*);
9709: Mat T;
9710: PetscBool istrans;
9715: if (scall == MAT_INPLACE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Inplace product not supported");
9716: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9717: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9720: MatCheckPreallocated(B,2);
9721: if (!B->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9722: if (B->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9724: if (B->cmap->N!=A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, AN %D != BN %D",A->cmap->N,B->cmap->N);
9725: if (fill == PETSC_DEFAULT || fill == PETSC_DECIDE) fill = 2.0;
9726: if (fill < 1.0) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Expected fill=%g must be > 1.0",(double)fill);
9727: MatCheckPreallocated(A,1);
9729: PetscObjectTypeCompare((PetscObject)B,MATTRANSPOSEMAT,&istrans);
9730: if (istrans) {
9731: MatTransposeGetMat(B,&T);
9732: MatMatMult(A,T,scall,fill,C);
9733: return(0);
9734: }
9735: fA = A->ops->mattransposemult;
9736: if (!fA) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"MatMatTransposeMult not supported for A of type %s",((PetscObject)A)->type_name);
9737: fB = B->ops->mattransposemult;
9738: if (!fB) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"MatMatTransposeMult not supported for B of type %s",((PetscObject)B)->type_name);
9739: if (fB!=fA) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_INCOMP,"MatMatTransposeMult requires A, %s, to be compatible with B, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
9741: PetscLogEventBegin(MAT_MatTransposeMult,A,B,0,0);
9742: if (scall == MAT_INITIAL_MATRIX) {
9743: PetscLogEventBegin(MAT_MatTransposeMultSymbolic,A,B,0,0);
9744: (*A->ops->mattransposemultsymbolic)(A,B,fill,C);
9745: PetscLogEventEnd(MAT_MatTransposeMultSymbolic,A,B,0,0);
9746: }
9747: PetscLogEventBegin(MAT_MatTransposeMultNumeric,A,B,0,0);
9748: (*A->ops->mattransposemultnumeric)(A,B,*C);
9749: PetscLogEventEnd(MAT_MatTransposeMultNumeric,A,B,0,0);
9750: PetscLogEventEnd(MAT_MatTransposeMult,A,B,0,0);
9751: return(0);
9752: }
9754: /*@
9755: MatTransposeMatMult - Performs Matrix-Matrix Multiplication C=A^T*B.
9757: Neighbor-wise Collective on Mat
9759: Input Parameters:
9760: + A - the left matrix
9761: . B - the right matrix
9762: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9763: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use PETSC_DEFAULT if not known
9765: Output Parameters:
9766: . C - the product matrix
9768: Notes:
9769: C will be created if MAT_INITIAL_MATRIX and must be destroyed by the user with MatDestroy().
9771: MAT_REUSE_MATRIX can only be used if the matrices A and B have the same nonzero pattern as in the previous call
9773: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
9774: actually needed.
9776: This routine is currently implemented for pairs of AIJ matrices and pairs of SeqDense matrices and classes
9777: which inherit from SeqAIJ. C will be of same type as the input matrices.
9779: Level: intermediate
9781: .seealso: MatTransposeMatMultSymbolic(), MatTransposeMatMultNumeric(), MatMatMult(), MatMatTransposeMult(), MatPtAP()
9782: @*/
9783: PetscErrorCode MatTransposeMatMult(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
9784: {
9786: PetscErrorCode (*fA)(Mat,Mat,MatReuse,PetscReal,Mat*);
9787: PetscErrorCode (*fB)(Mat,Mat,MatReuse,PetscReal,Mat*);
9788: PetscErrorCode (*transposematmult)(Mat,Mat,MatReuse,PetscReal,Mat*) = NULL;
9789: Mat T;
9790: PetscBool istrans;
9795: if (scall == MAT_INPLACE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Inplace product not supported");
9796: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9797: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9800: MatCheckPreallocated(B,2);
9801: if (!B->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9802: if (B->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9804: if (B->rmap->N!=A->rmap->N) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->rmap->N,A->rmap->N);
9805: if (fill == PETSC_DEFAULT || fill == PETSC_DECIDE) fill = 2.0;
9806: if (fill < 1.0) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Expected fill=%g must be > 1.0",(double)fill);
9807: MatCheckPreallocated(A,1);
9809: PetscObjectTypeCompare((PetscObject)A,MATTRANSPOSEMAT,&istrans);
9810: if (istrans) {
9811: MatTransposeGetMat(A,&T);
9812: MatMatMult(T,B,scall,fill,C);
9813: return(0);
9814: }
9815: fA = A->ops->transposematmult;
9816: fB = B->ops->transposematmult;
9817: if (fB == fA && fA) transposematmult = fA;
9818: else {
9819: /* dispatch based on the type of A and B from their PetscObject's PetscFunctionLists. */
9820: char multname[256];
9821: PetscStrncpy(multname,"MatTransposeMatMult_",sizeof(multname));
9822: PetscStrlcat(multname,((PetscObject)A)->type_name,sizeof(multname));
9823: PetscStrlcat(multname,"_",sizeof(multname));
9824: PetscStrlcat(multname,((PetscObject)B)->type_name,sizeof(multname));
9825: PetscStrlcat(multname,"_C",sizeof(multname)); /* e.g., multname = "MatMatMult_seqdense_seqaij_C" */
9826: PetscObjectQueryFunction((PetscObject)B,multname,&transposematmult);
9827: if (!transposematmult) {
9828: PetscObjectQueryFunction((PetscObject)A,multname,&transposematmult);
9829: }
9830: if (!transposematmult) SETERRQ2(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_INCOMP,"MatTransposeMatMult requires A, %s, to be compatible with B, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
9831: }
9832: PetscLogEventBegin(MAT_TransposeMatMult,A,B,0,0);
9833: (*transposematmult)(A,B,scall,fill,C);
9834: PetscLogEventEnd(MAT_TransposeMatMult,A,B,0,0);
9835: return(0);
9836: }
9838: /*@
9839: MatMatMatMult - Performs Matrix-Matrix-Matrix Multiplication D=A*B*C.
9841: Neighbor-wise Collective on Mat
9843: Input Parameters:
9844: + A - the left matrix
9845: . B - the middle matrix
9846: . C - the right matrix
9847: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9848: - 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
9849: if the result is a dense matrix this is irrelevent
9851: Output Parameters:
9852: . D - the product matrix
9854: Notes:
9855: Unless scall is MAT_REUSE_MATRIX D will be created.
9857: MAT_REUSE_MATRIX can only be used if the matrices A, B and C have the same nonzero pattern as in the previous call
9859: To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
9860: actually needed.
9862: If you have many matrices with the same non-zero structure to multiply, you
9863: should use MAT_REUSE_MATRIX in all calls but the first or
9865: Level: intermediate
9867: .seealso: MatMatMult, MatPtAP()
9868: @*/
9869: PetscErrorCode MatMatMatMult(Mat A,Mat B,Mat C,MatReuse scall,PetscReal fill,Mat *D)
9870: {
9872: PetscErrorCode (*fA)(Mat,Mat,Mat,MatReuse,PetscReal,Mat*);
9873: PetscErrorCode (*fB)(Mat,Mat,Mat,MatReuse,PetscReal,Mat*);
9874: PetscErrorCode (*fC)(Mat,Mat,Mat,MatReuse,PetscReal,Mat*);
9875: PetscErrorCode (*mult)(Mat,Mat,Mat,MatReuse,PetscReal,Mat*)=NULL;
9880: MatCheckPreallocated(A,1);
9881: if (scall == MAT_INPLACE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Inplace product not supported");
9882: if (!A->assembled) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9883: if (A->factortype) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9886: MatCheckPreallocated(B,2);
9887: if (!B->assembled) SETERRQ(PetscObjectComm((PetscObject)B),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9888: if (B->factortype) SETERRQ(PetscObjectComm((PetscObject)B),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9891: MatCheckPreallocated(C,3);
9892: if (!C->assembled) SETERRQ(PetscObjectComm((PetscObject)C),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9893: if (C->factortype) SETERRQ(PetscObjectComm((PetscObject)C),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9894: if (B->rmap->N!=A->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)B),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->rmap->N,A->cmap->N);
9895: if (C->rmap->N!=B->cmap->N) SETERRQ2(PetscObjectComm((PetscObject)C),PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",C->rmap->N,B->cmap->N);
9896: if (scall == MAT_REUSE_MATRIX) {
9899: PetscLogEventBegin(MAT_MatMatMult,A,B,0,0);
9900: (*(*D)->ops->matmatmult)(A,B,C,scall,fill,D);
9901: PetscLogEventEnd(MAT_MatMatMult,A,B,0,0);
9902: return(0);
9903: }
9904: if (fill == PETSC_DEFAULT || fill == PETSC_DECIDE) fill = 2.0;
9905: if (fill < 1.0) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_SIZ,"Expected fill=%g must be >= 1.0",(double)fill);
9907: fA = A->ops->matmatmult;
9908: fB = B->ops->matmatmult;
9909: fC = C->ops->matmatmult;
9910: if (fA == fB && fA == fC) {
9911: if (!fA) SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"MatMatMatMult not supported for A of type %s",((PetscObject)A)->type_name);
9912: mult = fA;
9913: } else {
9914: /* dispatch based on the type of A, B and C from their PetscObject's PetscFunctionLists. */
9915: char multname[256];
9916: PetscStrncpy(multname,"MatMatMatMult_",sizeof(multname));
9917: PetscStrlcat(multname,((PetscObject)A)->type_name,sizeof(multname));
9918: PetscStrlcat(multname,"_",sizeof(multname));
9919: PetscStrlcat(multname,((PetscObject)B)->type_name,sizeof(multname));
9920: PetscStrlcat(multname,"_",sizeof(multname));
9921: PetscStrlcat(multname,((PetscObject)C)->type_name,sizeof(multname));
9922: PetscStrlcat(multname,"_C",sizeof(multname));
9923: PetscObjectQueryFunction((PetscObject)B,multname,&mult);
9924: if (!mult) SETERRQ3(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_INCOMP,"MatMatMatMult requires A, %s, to be compatible with B, %s, C, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name,((PetscObject)C)->type_name);
9925: }
9926: PetscLogEventBegin(MAT_MatMatMult,A,B,0,0);
9927: (*mult)(A,B,C,scall,fill,D);
9928: PetscLogEventEnd(MAT_MatMatMult,A,B,0,0);
9929: return(0);
9930: }
9932: /*@
9933: MatCreateRedundantMatrix - Create redundant matrices and put them into processors of subcommunicators.
9935: Collective on Mat
9937: Input Parameters:
9938: + mat - the matrix
9939: . nsubcomm - the number of subcommunicators (= number of redundant parallel or sequential matrices)
9940: . subcomm - MPI communicator split from the communicator where mat resides in (or MPI_COMM_NULL if nsubcomm is used)
9941: - reuse - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
9943: Output Parameter:
9944: . matredundant - redundant matrix
9946: Notes:
9947: MAT_REUSE_MATRIX can only be used when the nonzero structure of the
9948: original matrix has not changed from that last call to MatCreateRedundantMatrix().
9950: This routine creates the duplicated matrices in subcommunicators; you should NOT create them before
9951: calling it.
9953: Level: advanced
9956: .seealso: MatDestroy()
9957: @*/
9958: PetscErrorCode MatCreateRedundantMatrix(Mat mat,PetscInt nsubcomm,MPI_Comm subcomm,MatReuse reuse,Mat *matredundant)
9959: {
9961: MPI_Comm comm;
9962: PetscMPIInt size;
9963: PetscInt mloc_sub,nloc_sub,rstart,rend,M=mat->rmap->N,N=mat->cmap->N,bs=mat->rmap->bs;
9964: Mat_Redundant *redund=NULL;
9965: PetscSubcomm psubcomm=NULL;
9966: MPI_Comm subcomm_in=subcomm;
9967: Mat *matseq;
9968: IS isrow,iscol;
9969: PetscBool newsubcomm=PETSC_FALSE;
9973: if (nsubcomm && reuse == MAT_REUSE_MATRIX) {
9976: }
9978: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
9979: if (size == 1 || nsubcomm == 1) {
9980: if (reuse == MAT_INITIAL_MATRIX) {
9981: MatDuplicate(mat,MAT_COPY_VALUES,matredundant);
9982: } else {
9983: 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");
9984: MatCopy(mat,*matredundant,SAME_NONZERO_PATTERN);
9985: }
9986: return(0);
9987: }
9989: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
9990: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
9991: MatCheckPreallocated(mat,1);
9993: PetscLogEventBegin(MAT_RedundantMat,mat,0,0,0);
9994: if (subcomm_in == MPI_COMM_NULL && reuse == MAT_INITIAL_MATRIX) { /* get subcomm if user does not provide subcomm */
9995: /* create psubcomm, then get subcomm */
9996: PetscObjectGetComm((PetscObject)mat,&comm);
9997: MPI_Comm_size(comm,&size);
9998: if (nsubcomm < 1 || nsubcomm > size) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"nsubcomm must between 1 and %D",size);
10000: PetscSubcommCreate(comm,&psubcomm);
10001: PetscSubcommSetNumber(psubcomm,nsubcomm);
10002: PetscSubcommSetType(psubcomm,PETSC_SUBCOMM_CONTIGUOUS);
10003: PetscSubcommSetFromOptions(psubcomm);
10004: PetscCommDuplicate(PetscSubcommChild(psubcomm),&subcomm,NULL);
10005: newsubcomm = PETSC_TRUE;
10006: PetscSubcommDestroy(&psubcomm);
10007: }
10009: /* get isrow, iscol and a local sequential matrix matseq[0] */
10010: if (reuse == MAT_INITIAL_MATRIX) {
10011: mloc_sub = PETSC_DECIDE;
10012: nloc_sub = PETSC_DECIDE;
10013: if (bs < 1) {
10014: PetscSplitOwnership(subcomm,&mloc_sub,&M);
10015: PetscSplitOwnership(subcomm,&nloc_sub,&N);
10016: } else {
10017: PetscSplitOwnershipBlock(subcomm,bs,&mloc_sub,&M);
10018: PetscSplitOwnershipBlock(subcomm,bs,&nloc_sub,&N);
10019: }
10020: MPI_Scan(&mloc_sub,&rend,1,MPIU_INT,MPI_SUM,subcomm);
10021: rstart = rend - mloc_sub;
10022: ISCreateStride(PETSC_COMM_SELF,mloc_sub,rstart,1,&isrow);
10023: ISCreateStride(PETSC_COMM_SELF,N,0,1,&iscol);
10024: } else { /* reuse == MAT_REUSE_MATRIX */
10025: 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");
10026: /* retrieve subcomm */
10027: PetscObjectGetComm((PetscObject)(*matredundant),&subcomm);
10028: redund = (*matredundant)->redundant;
10029: isrow = redund->isrow;
10030: iscol = redund->iscol;
10031: matseq = redund->matseq;
10032: }
10033: MatCreateSubMatrices(mat,1,&isrow,&iscol,reuse,&matseq);
10035: /* get matredundant over subcomm */
10036: if (reuse == MAT_INITIAL_MATRIX) {
10037: MatCreateMPIMatConcatenateSeqMat(subcomm,matseq[0],nloc_sub,reuse,matredundant);
10039: /* create a supporting struct and attach it to C for reuse */
10040: PetscNewLog(*matredundant,&redund);
10041: (*matredundant)->redundant = redund;
10042: redund->isrow = isrow;
10043: redund->iscol = iscol;
10044: redund->matseq = matseq;
10045: if (newsubcomm) {
10046: redund->subcomm = subcomm;
10047: } else {
10048: redund->subcomm = MPI_COMM_NULL;
10049: }
10050: } else {
10051: MatCreateMPIMatConcatenateSeqMat(subcomm,matseq[0],PETSC_DECIDE,reuse,matredundant);
10052: }
10053: PetscLogEventEnd(MAT_RedundantMat,mat,0,0,0);
10054: return(0);
10055: }
10057: /*@C
10058: MatGetMultiProcBlock - Create multiple [bjacobi] 'parallel submatrices' from
10059: a given 'mat' object. Each submatrix can span multiple procs.
10061: Collective on Mat
10063: Input Parameters:
10064: + mat - the matrix
10065: . subcomm - the subcommunicator obtained by com_split(comm)
10066: - scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
10068: Output Parameter:
10069: . subMat - 'parallel submatrices each spans a given subcomm
10071: Notes:
10072: The submatrix partition across processors is dictated by 'subComm' a
10073: communicator obtained by com_split(comm). The comm_split
10074: is not restriced to be grouped with consecutive original ranks.
10076: Due the comm_split() usage, the parallel layout of the submatrices
10077: map directly to the layout of the original matrix [wrt the local
10078: row,col partitioning]. So the original 'DiagonalMat' naturally maps
10079: into the 'DiagonalMat' of the subMat, hence it is used directly from
10080: the subMat. However the offDiagMat looses some columns - and this is
10081: reconstructed with MatSetValues()
10083: Level: advanced
10086: .seealso: MatCreateSubMatrices()
10087: @*/
10088: PetscErrorCode MatGetMultiProcBlock(Mat mat, MPI_Comm subComm, MatReuse scall,Mat *subMat)
10089: {
10091: PetscMPIInt commsize,subCommSize;
10094: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&commsize);
10095: MPI_Comm_size(subComm,&subCommSize);
10096: if (subCommSize > commsize) SETERRQ2(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_OUTOFRANGE,"CommSize %D < SubCommZize %D",commsize,subCommSize);
10098: 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");
10099: PetscLogEventBegin(MAT_GetMultiProcBlock,mat,0,0,0);
10100: (*mat->ops->getmultiprocblock)(mat,subComm,scall,subMat);
10101: PetscLogEventEnd(MAT_GetMultiProcBlock,mat,0,0,0);
10102: return(0);
10103: }
10105: /*@
10106: MatGetLocalSubMatrix - Gets a reference to a submatrix specified in local numbering
10108: Not Collective
10110: Input Arguments:
10111: + mat - matrix to extract local submatrix from
10112: . isrow - local row indices for submatrix
10113: - iscol - local column indices for submatrix
10115: Output Arguments:
10116: . submat - the submatrix
10118: Level: intermediate
10120: Notes:
10121: The submat should be returned with MatRestoreLocalSubMatrix().
10123: Depending on the format of mat, the returned submat may not implement MatMult(). Its communicator may be
10124: the same as mat, it may be PETSC_COMM_SELF, or some other subcomm of mat's.
10126: The submat always implements MatSetValuesLocal(). If isrow and iscol have the same block size, then
10127: MatSetValuesBlockedLocal() will also be implemented.
10129: The mat must have had a ISLocalToGlobalMapping provided to it with MatSetLocalToGlobalMapping(). Note that
10130: matrices obtained with DMCreateMatrix() generally already have the local to global mapping provided.
10132: .seealso: MatRestoreLocalSubMatrix(), MatCreateLocalRef(), MatSetLocalToGlobalMapping()
10133: @*/
10134: PetscErrorCode MatGetLocalSubMatrix(Mat mat,IS isrow,IS iscol,Mat *submat)
10135: {
10144: if (!mat->rmap->mapping) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Matrix must have local to global mapping provided before this call");
10146: if (mat->ops->getlocalsubmatrix) {
10147: (*mat->ops->getlocalsubmatrix)(mat,isrow,iscol,submat);
10148: } else {
10149: MatCreateLocalRef(mat,isrow,iscol,submat);
10150: }
10151: return(0);
10152: }
10154: /*@
10155: MatRestoreLocalSubMatrix - Restores a reference to a submatrix specified in local numbering
10157: Not Collective
10159: Input Arguments:
10160: mat - matrix to extract local submatrix from
10161: isrow - local row indices for submatrix
10162: iscol - local column indices for submatrix
10163: submat - the submatrix
10165: Level: intermediate
10167: .seealso: MatGetLocalSubMatrix()
10168: @*/
10169: PetscErrorCode MatRestoreLocalSubMatrix(Mat mat,IS isrow,IS iscol,Mat *submat)
10170: {
10179: if (*submat) {
10181: }
10183: if (mat->ops->restorelocalsubmatrix) {
10184: (*mat->ops->restorelocalsubmatrix)(mat,isrow,iscol,submat);
10185: } else {
10186: MatDestroy(submat);
10187: }
10188: *submat = NULL;
10189: return(0);
10190: }
10192: /* --------------------------------------------------------*/
10193: /*@
10194: MatFindZeroDiagonals - Finds all the rows of a matrix that have zero or no diagonal entry in the matrix
10196: Collective on Mat
10198: Input Parameter:
10199: . mat - the matrix
10201: Output Parameter:
10202: . is - if any rows have zero diagonals this contains the list of them
10204: Level: developer
10206: .seealso: MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
10207: @*/
10208: PetscErrorCode MatFindZeroDiagonals(Mat mat,IS *is)
10209: {
10215: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
10216: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
10218: if (!mat->ops->findzerodiagonals) {
10219: Vec diag;
10220: const PetscScalar *a;
10221: PetscInt *rows;
10222: PetscInt rStart, rEnd, r, nrow = 0;
10224: MatCreateVecs(mat, &diag, NULL);
10225: MatGetDiagonal(mat, diag);
10226: MatGetOwnershipRange(mat, &rStart, &rEnd);
10227: VecGetArrayRead(diag, &a);
10228: for (r = 0; r < rEnd-rStart; ++r) if (a[r] == 0.0) ++nrow;
10229: PetscMalloc1(nrow, &rows);
10230: nrow = 0;
10231: for (r = 0; r < rEnd-rStart; ++r) if (a[r] == 0.0) rows[nrow++] = r+rStart;
10232: VecRestoreArrayRead(diag, &a);
10233: VecDestroy(&diag);
10234: ISCreateGeneral(PetscObjectComm((PetscObject) mat), nrow, rows, PETSC_OWN_POINTER, is);
10235: } else {
10236: (*mat->ops->findzerodiagonals)(mat, is);
10237: }
10238: return(0);
10239: }
10241: /*@
10242: MatFindOffBlockDiagonalEntries - Finds all the rows of a matrix that have entries outside of the main diagonal block (defined by the matrix block size)
10244: Collective on Mat
10246: Input Parameter:
10247: . mat - the matrix
10249: Output Parameter:
10250: . is - contains the list of rows with off block diagonal entries
10252: Level: developer
10254: .seealso: MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
10255: @*/
10256: PetscErrorCode MatFindOffBlockDiagonalEntries(Mat mat,IS *is)
10257: {
10263: if (!mat->assembled) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
10264: if (mat->factortype) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
10266: 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);
10267: (*mat->ops->findoffblockdiagonalentries)(mat,is);
10268: return(0);
10269: }
10271: /*@C
10272: MatInvertBlockDiagonal - Inverts the block diagonal entries.
10274: Collective on Mat
10276: Input Parameters:
10277: . mat - the matrix
10279: Output Parameters:
10280: . values - the block inverses in column major order (FORTRAN-like)
10282: Note:
10283: This routine is not available from Fortran.
10285: Level: advanced
10287: .seealso: MatInvertBockDiagonalMat
10288: @*/
10289: PetscErrorCode MatInvertBlockDiagonal(Mat mat,const PetscScalar **values)
10290: {
10295: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
10296: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
10297: if (!mat->ops->invertblockdiagonal) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not supported for type %s",((PetscObject)mat)->type_name);
10298: (*mat->ops->invertblockdiagonal)(mat,values);
10299: return(0);
10300: }
10302: /*@C
10303: MatInvertVariableBlockDiagonal - Inverts the block diagonal entries.
10305: Collective on Mat
10307: Input Parameters:
10308: + mat - the matrix
10309: . nblocks - the number of blocks
10310: - bsizes - the size of each block
10312: Output Parameters:
10313: . values - the block inverses in column major order (FORTRAN-like)
10315: Note:
10316: This routine is not available from Fortran.
10318: Level: advanced
10320: .seealso: MatInvertBockDiagonal()
10321: @*/
10322: PetscErrorCode MatInvertVariableBlockDiagonal(Mat mat,PetscInt nblocks,const PetscInt *bsizes,PetscScalar *values)
10323: {
10328: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
10329: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
10330: if (!mat->ops->invertvariableblockdiagonal) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not supported for type",((PetscObject)mat)->type_name);
10331: (*mat->ops->invertvariableblockdiagonal)(mat,nblocks,bsizes,values);
10332: return(0);
10333: }
10335: /*@
10336: MatInvertBlockDiagonalMat - set matrix C to be the inverted block diagonal of matrix A
10338: Collective on Mat
10340: Input Parameters:
10341: . A - the matrix
10343: Output Parameters:
10344: . C - matrix with inverted block diagonal of A. This matrix should be created and may have its type set.
10346: Notes: the blocksize of the matrix is used to determine the blocks on the diagonal of C
10348: Level: advanced
10350: .seealso: MatInvertBockDiagonal()
10351: @*/
10352: PetscErrorCode MatInvertBlockDiagonalMat(Mat A,Mat C)
10353: {
10354: PetscErrorCode ierr;
10355: const PetscScalar *vals;
10356: PetscInt *dnnz;
10357: PetscInt M,N,m,n,rstart,rend,bs,i,j;
10360: MatInvertBlockDiagonal(A,&vals);
10361: MatGetBlockSize(A,&bs);
10362: MatGetSize(A,&M,&N);
10363: MatGetLocalSize(A,&m,&n);
10364: MatSetSizes(C,m,n,M,N);
10365: MatSetBlockSize(C,bs);
10366: PetscMalloc1(m/bs,&dnnz);
10367: for (j = 0; j < m/bs; j++) dnnz[j] = 1;
10368: MatXAIJSetPreallocation(C,bs,dnnz,NULL,NULL,NULL);
10369: PetscFree(dnnz);
10370: MatGetOwnershipRange(C,&rstart,&rend);
10371: MatSetOption(C,MAT_ROW_ORIENTED,PETSC_FALSE);
10372: for (i = rstart/bs; i < rend/bs; i++) {
10373: MatSetValuesBlocked(C,1,&i,1,&i,&vals[(i-rstart/bs)*bs*bs],INSERT_VALUES);
10374: }
10375: MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);
10376: MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);
10377: MatSetOption(C,MAT_ROW_ORIENTED,PETSC_TRUE);
10378: return(0);
10379: }
10381: /*@C
10382: MatTransposeColoringDestroy - Destroys a coloring context for matrix product C=A*B^T that was created
10383: via MatTransposeColoringCreate().
10385: Collective on MatTransposeColoring
10387: Input Parameter:
10388: . c - coloring context
10390: Level: intermediate
10392: .seealso: MatTransposeColoringCreate()
10393: @*/
10394: PetscErrorCode MatTransposeColoringDestroy(MatTransposeColoring *c)
10395: {
10396: PetscErrorCode ierr;
10397: MatTransposeColoring matcolor=*c;
10400: if (!matcolor) return(0);
10401: if (--((PetscObject)matcolor)->refct > 0) {matcolor = 0; return(0);}
10403: PetscFree3(matcolor->ncolumns,matcolor->nrows,matcolor->colorforrow);
10404: PetscFree(matcolor->rows);
10405: PetscFree(matcolor->den2sp);
10406: PetscFree(matcolor->colorforcol);
10407: PetscFree(matcolor->columns);
10408: if (matcolor->brows>0) {
10409: PetscFree(matcolor->lstart);
10410: }
10411: PetscHeaderDestroy(c);
10412: return(0);
10413: }
10415: /*@C
10416: MatTransColoringApplySpToDen - Given a symbolic matrix product C=A*B^T for which
10417: a MatTransposeColoring context has been created, computes a dense B^T by Apply
10418: MatTransposeColoring to sparse B.
10420: Collective on MatTransposeColoring
10422: Input Parameters:
10423: + B - sparse matrix B
10424: . Btdense - symbolic dense matrix B^T
10425: - coloring - coloring context created with MatTransposeColoringCreate()
10427: Output Parameter:
10428: . Btdense - dense matrix B^T
10430: Level: advanced
10432: Notes:
10433: These are used internally for some implementations of MatRARt()
10435: .seealso: MatTransposeColoringCreate(), MatTransposeColoringDestroy(), MatTransColoringApplyDenToSp()
10437: @*/
10438: PetscErrorCode MatTransColoringApplySpToDen(MatTransposeColoring coloring,Mat B,Mat Btdense)
10439: {
10447: if (!B->ops->transcoloringapplysptoden) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not supported for this matrix type %s",((PetscObject)B)->type_name);
10448: (B->ops->transcoloringapplysptoden)(coloring,B,Btdense);
10449: return(0);
10450: }
10452: /*@C
10453: MatTransColoringApplyDenToSp - Given a symbolic matrix product Csp=A*B^T for which
10454: a MatTransposeColoring context has been created and a dense matrix Cden=A*Btdense
10455: in which Btdens is obtained from MatTransColoringApplySpToDen(), recover sparse matrix
10456: Csp from Cden.
10458: Collective on MatTransposeColoring
10460: Input Parameters:
10461: + coloring - coloring context created with MatTransposeColoringCreate()
10462: - Cden - matrix product of a sparse matrix and a dense matrix Btdense
10464: Output Parameter:
10465: . Csp - sparse matrix
10467: Level: advanced
10469: Notes:
10470: These are used internally for some implementations of MatRARt()
10472: .seealso: MatTransposeColoringCreate(), MatTransposeColoringDestroy(), MatTransColoringApplySpToDen()
10474: @*/
10475: PetscErrorCode MatTransColoringApplyDenToSp(MatTransposeColoring matcoloring,Mat Cden,Mat Csp)
10476: {
10484: if (!Csp->ops->transcoloringapplydentosp) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not supported for this matrix type %s",((PetscObject)Csp)->type_name);
10485: (Csp->ops->transcoloringapplydentosp)(matcoloring,Cden,Csp);
10486: return(0);
10487: }
10489: /*@C
10490: MatTransposeColoringCreate - Creates a matrix coloring context for matrix product C=A*B^T.
10492: Collective on Mat
10494: Input Parameters:
10495: + mat - the matrix product C
10496: - iscoloring - the coloring of the matrix; usually obtained with MatColoringCreate() or DMCreateColoring()
10498: Output Parameter:
10499: . color - the new coloring context
10501: Level: intermediate
10503: .seealso: MatTransposeColoringDestroy(), MatTransColoringApplySpToDen(),
10504: MatTransColoringApplyDenToSp()
10505: @*/
10506: PetscErrorCode MatTransposeColoringCreate(Mat mat,ISColoring iscoloring,MatTransposeColoring *color)
10507: {
10508: MatTransposeColoring c;
10509: MPI_Comm comm;
10510: PetscErrorCode ierr;
10513: PetscLogEventBegin(MAT_TransposeColoringCreate,mat,0,0,0);
10514: PetscObjectGetComm((PetscObject)mat,&comm);
10515: PetscHeaderCreate(c,MAT_TRANSPOSECOLORING_CLASSID,"MatTransposeColoring","Matrix product C=A*B^T via coloring","Mat",comm,MatTransposeColoringDestroy,NULL);
10517: c->ctype = iscoloring->ctype;
10518: if (mat->ops->transposecoloringcreate) {
10519: (*mat->ops->transposecoloringcreate)(mat,iscoloring,c);
10520: } else SETERRQ1(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"Code not yet written for matrix type %s",((PetscObject)mat)->type_name);
10522: *color = c;
10523: PetscLogEventEnd(MAT_TransposeColoringCreate,mat,0,0,0);
10524: return(0);
10525: }
10527: /*@
10528: MatGetNonzeroState - Returns a 64 bit integer representing the current state of nonzeros in the matrix. If the
10529: matrix has had no new nonzero locations added to the matrix since the previous call then the value will be the
10530: same, otherwise it will be larger
10532: Not Collective
10534: Input Parameter:
10535: . A - the matrix
10537: Output Parameter:
10538: . state - the current state
10540: Notes:
10541: You can only compare states from two different calls to the SAME matrix, you cannot compare calls between
10542: different matrices
10544: Level: intermediate
10546: @*/
10547: PetscErrorCode MatGetNonzeroState(Mat mat,PetscObjectState *state)
10548: {
10551: *state = mat->nonzerostate;
10552: return(0);
10553: }
10555: /*@
10556: MatCreateMPIMatConcatenateSeqMat - Creates a single large PETSc matrix by concatenating sequential
10557: matrices from each processor
10559: Collective
10561: Input Parameters:
10562: + comm - the communicators the parallel matrix will live on
10563: . seqmat - the input sequential matrices
10564: . n - number of local columns (or PETSC_DECIDE)
10565: - reuse - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
10567: Output Parameter:
10568: . mpimat - the parallel matrix generated
10570: Level: advanced
10572: Notes:
10573: The number of columns of the matrix in EACH processor MUST be the same.
10575: @*/
10576: PetscErrorCode MatCreateMPIMatConcatenateSeqMat(MPI_Comm comm,Mat seqmat,PetscInt n,MatReuse reuse,Mat *mpimat)
10577: {
10581: if (!seqmat->ops->creatempimatconcatenateseqmat) SETERRQ1(PetscObjectComm((PetscObject)seqmat),PETSC_ERR_SUP,"Mat type %s",((PetscObject)seqmat)->type_name);
10582: 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");
10584: PetscLogEventBegin(MAT_Merge,seqmat,0,0,0);
10585: (*seqmat->ops->creatempimatconcatenateseqmat)(comm,seqmat,n,reuse,mpimat);
10586: PetscLogEventEnd(MAT_Merge,seqmat,0,0,0);
10587: return(0);
10588: }
10590: /*@
10591: MatSubdomainsCreateCoalesce - Creates index subdomains by coalescing adjacent
10592: ranks' ownership ranges.
10594: Collective on A
10596: Input Parameters:
10597: + A - the matrix to create subdomains from
10598: - N - requested number of subdomains
10601: Output Parameters:
10602: + n - number of subdomains resulting on this rank
10603: - iss - IS list with indices of subdomains on this rank
10605: Level: advanced
10607: Notes:
10608: number of subdomains must be smaller than the communicator size
10609: @*/
10610: PetscErrorCode MatSubdomainsCreateCoalesce(Mat A,PetscInt N,PetscInt *n,IS *iss[])
10611: {
10612: MPI_Comm comm,subcomm;
10613: PetscMPIInt size,rank,color;
10614: PetscInt rstart,rend,k;
10615: PetscErrorCode ierr;
10618: PetscObjectGetComm((PetscObject)A,&comm);
10619: MPI_Comm_size(comm,&size);
10620: MPI_Comm_rank(comm,&rank);
10621: 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);
10622: *n = 1;
10623: k = ((PetscInt)size)/N + ((PetscInt)size%N>0); /* There are up to k ranks to a color */
10624: color = rank/k;
10625: MPI_Comm_split(comm,color,rank,&subcomm);
10626: PetscMalloc1(1,iss);
10627: MatGetOwnershipRange(A,&rstart,&rend);
10628: ISCreateStride(subcomm,rend-rstart,rstart,1,iss[0]);
10629: MPI_Comm_free(&subcomm);
10630: return(0);
10631: }
10633: /*@
10634: MatGalerkin - Constructs the coarse grid problem via Galerkin projection.
10636: If the interpolation and restriction operators are the same, uses MatPtAP.
10637: If they are not the same, use MatMatMatMult.
10639: Once the coarse grid problem is constructed, correct for interpolation operators
10640: that are not of full rank, which can legitimately happen in the case of non-nested
10641: geometric multigrid.
10643: Input Parameters:
10644: + restrct - restriction operator
10645: . dA - fine grid matrix
10646: . interpolate - interpolation operator
10647: . reuse - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
10648: - fill - expected fill, use PETSC_DEFAULT if you do not have a good estimate
10650: Output Parameters:
10651: . A - the Galerkin coarse matrix
10653: Options Database Key:
10654: . -pc_mg_galerkin <both,pmat,mat,none>
10656: Level: developer
10658: .seealso: MatPtAP(), MatMatMatMult()
10659: @*/
10660: PetscErrorCode MatGalerkin(Mat restrct, Mat dA, Mat interpolate, MatReuse reuse, PetscReal fill, Mat *A)
10661: {
10663: IS zerorows;
10664: Vec diag;
10667: if (reuse == MAT_INPLACE_MATRIX) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Inplace product not supported");
10668: /* Construct the coarse grid matrix */
10669: if (interpolate == restrct) {
10670: MatPtAP(dA,interpolate,reuse,fill,A);
10671: } else {
10672: MatMatMatMult(restrct,dA,interpolate,reuse,fill,A);
10673: }
10675: /* If the interpolation matrix is not of full rank, A will have zero rows.
10676: This can legitimately happen in the case of non-nested geometric multigrid.
10677: In that event, we set the rows of the matrix to the rows of the identity,
10678: ignoring the equations (as the RHS will also be zero). */
10680: MatFindZeroRows(*A, &zerorows);
10682: if (zerorows != NULL) { /* if there are any zero rows */
10683: MatCreateVecs(*A, &diag, NULL);
10684: MatGetDiagonal(*A, diag);
10685: VecISSet(diag, zerorows, 1.0);
10686: MatDiagonalSet(*A, diag, INSERT_VALUES);
10687: VecDestroy(&diag);
10688: ISDestroy(&zerorows);
10689: }
10690: return(0);
10691: }
10693: /*@C
10694: MatSetOperation - Allows user to set a matrix operation for any matrix type
10696: Logically Collective on Mat
10698: Input Parameters:
10699: + mat - the matrix
10700: . op - the name of the operation
10701: - f - the function that provides the operation
10703: Level: developer
10705: Usage:
10706: $ extern PetscErrorCode usermult(Mat,Vec,Vec);
10707: $ MatCreateXXX(comm,...&A);
10708: $ MatSetOperation(A,MATOP_MULT,(void(*)(void))usermult);
10710: Notes:
10711: See the file include/petscmat.h for a complete list of matrix
10712: operations, which all have the form MATOP_<OPERATION>, where
10713: <OPERATION> is the name (in all capital letters) of the
10714: user interface routine (e.g., MatMult() -> MATOP_MULT).
10716: All user-provided functions (except for MATOP_DESTROY) should have the same calling
10717: sequence as the usual matrix interface routines, since they
10718: are intended to be accessed via the usual matrix interface
10719: routines, e.g.,
10720: $ MatMult(Mat,Vec,Vec) -> usermult(Mat,Vec,Vec)
10722: In particular each function MUST return an error code of 0 on success and
10723: nonzero on failure.
10725: This routine is distinct from MatShellSetOperation() in that it can be called on any matrix type.
10727: .seealso: MatGetOperation(), MatCreateShell(), MatShellSetContext(), MatShellSetOperation()
10728: @*/
10729: PetscErrorCode MatSetOperation(Mat mat,MatOperation op,void (*f)(void))
10730: {
10733: if (op == MATOP_VIEW && !mat->ops->viewnative && f != (void (*)(void))(mat->ops->view)) {
10734: mat->ops->viewnative = mat->ops->view;
10735: }
10736: (((void(**)(void))mat->ops)[op]) = f;
10737: return(0);
10738: }
10740: /*@C
10741: MatGetOperation - Gets a matrix operation for any matrix type.
10743: Not Collective
10745: Input Parameters:
10746: + mat - the matrix
10747: - op - the name of the operation
10749: Output Parameter:
10750: . f - the function that provides the operation
10752: Level: developer
10754: Usage:
10755: $ PetscErrorCode (*usermult)(Mat,Vec,Vec);
10756: $ MatGetOperation(A,MATOP_MULT,(void(**)(void))&usermult);
10758: Notes:
10759: See the file include/petscmat.h for a complete list of matrix
10760: operations, which all have the form MATOP_<OPERATION>, where
10761: <OPERATION> is the name (in all capital letters) of the
10762: user interface routine (e.g., MatMult() -> MATOP_MULT).
10764: This routine is distinct from MatShellGetOperation() in that it can be called on any matrix type.
10766: .seealso: MatSetOperation(), MatCreateShell(), MatShellGetContext(), MatShellGetOperation()
10767: @*/
10768: PetscErrorCode MatGetOperation(Mat mat,MatOperation op,void(**f)(void))
10769: {
10772: *f = (((void (**)(void))mat->ops)[op]);
10773: return(0);
10774: }
10776: /*@
10777: MatHasOperation - Determines whether the given matrix supports the particular
10778: operation.
10780: Not Collective
10782: Input Parameters:
10783: + mat - the matrix
10784: - op - the operation, for example, MATOP_GET_DIAGONAL
10786: Output Parameter:
10787: . has - either PETSC_TRUE or PETSC_FALSE
10789: Level: advanced
10791: Notes:
10792: See the file include/petscmat.h for a complete list of matrix
10793: operations, which all have the form MATOP_<OPERATION>, where
10794: <OPERATION> is the name (in all capital letters) of the
10795: user-level routine. E.g., MatNorm() -> MATOP_NORM.
10797: .seealso: MatCreateShell()
10798: @*/
10799: PetscErrorCode MatHasOperation(Mat mat,MatOperation op,PetscBool *has)
10800: {
10807: if (mat->ops->hasoperation) {
10808: (*mat->ops->hasoperation)(mat,op,has);
10809: } else {
10810: if (((void**)mat->ops)[op]) *has = PETSC_TRUE;
10811: else {
10812: *has = PETSC_FALSE;
10813: if (op == MATOP_CREATE_SUBMATRIX) {
10814: PetscMPIInt size;
10816: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
10817: if (size == 1) {
10818: MatHasOperation(mat,MATOP_CREATE_SUBMATRICES,has);
10819: }
10820: }
10821: }
10822: }
10823: return(0);
10824: }
10826: /*@
10827: MatHasCongruentLayouts - Determines whether the rows and columns layouts
10828: of the matrix are congruent
10830: Collective on mat
10832: Input Parameters:
10833: . mat - the matrix
10835: Output Parameter:
10836: . cong - either PETSC_TRUE or PETSC_FALSE
10838: Level: beginner
10840: Notes:
10842: .seealso: MatCreate(), MatSetSizes()
10843: @*/
10844: PetscErrorCode MatHasCongruentLayouts(Mat mat,PetscBool *cong)
10845: {
10852: if (!mat->rmap || !mat->cmap) {
10853: *cong = mat->rmap == mat->cmap ? PETSC_TRUE : PETSC_FALSE;
10854: return(0);
10855: }
10856: if (mat->congruentlayouts == PETSC_DECIDE) { /* first time we compare rows and cols layouts */
10857: PetscLayoutCompare(mat->rmap,mat->cmap,cong);
10858: if (*cong) mat->congruentlayouts = 1;
10859: else mat->congruentlayouts = 0;
10860: } else *cong = mat->congruentlayouts ? PETSC_TRUE : PETSC_FALSE;
10861: return(0);
10862: }
10864: /*@
10865: MatFreeIntermediateDataStructures - Free intermediate data structures created for reuse,
10866: e.g., matrx product of MatPtAP.
10868: Collective on mat
10870: Input Parameters:
10871: . mat - the matrix
10873: Output Parameter:
10874: . mat - the matrix with intermediate data structures released
10876: Level: advanced
10878: Notes:
10880: .seealso: MatPtAP(), MatMatMult()
10881: @*/
10882: PetscErrorCode MatFreeIntermediateDataStructures(Mat mat)
10883: {
10889: if (mat->ops->freeintermediatedatastructures) {
10890: (*mat->ops->freeintermediatedatastructures)(mat);
10891: }
10892: return(0);
10893: }