Actual source code: matrix.c
1: #define PETSCMAT_DLL
3: /*
4: This is where the abstract matrix operations are defined
5: */
7: #include include/private/matimpl.h
8: #include private/vecimpl.h
10: /* Logging support */
11: PetscCookie MAT_COOKIE = 0;
12: PetscEvent MAT_Mult = 0, MAT_Mults = 0, MAT_MultConstrained = 0, MAT_MultAdd = 0, MAT_MultTranspose = 0;
13: PetscEvent MAT_MultTransposeConstrained = 0, MAT_MultTransposeAdd = 0, MAT_Solve = 0, MAT_Solves = 0, MAT_SolveAdd = 0, MAT_SolveTranspose = 0, MAT_MatSolve = 0;
14: PetscEvent MAT_SolveTransposeAdd = 0, MAT_Relax = 0, MAT_ForwardSolve = 0, MAT_BackwardSolve = 0, MAT_LUFactor = 0, MAT_LUFactorSymbolic = 0;
15: PetscEvent MAT_LUFactorNumeric = 0, MAT_CholeskyFactor = 0, MAT_CholeskyFactorSymbolic = 0, MAT_CholeskyFactorNumeric = 0, MAT_ILUFactor = 0;
16: PetscEvent MAT_ILUFactorSymbolic = 0, MAT_ICCFactorSymbolic = 0, MAT_Copy = 0, MAT_Convert = 0, MAT_Scale = 0, MAT_AssemblyBegin = 0;
17: PetscEvent MAT_AssemblyEnd = 0, MAT_SetValues = 0, MAT_GetValues = 0, MAT_GetRow = 0, MAT_GetRowIJ = 0, MAT_GetSubMatrices = 0, MAT_GetColoring = 0, MAT_GetOrdering = 0, MAT_GetRedundantMatrix = 0, MAT_GetSeqNonzeroStructure = 0;
18: PetscEvent MAT_IncreaseOverlap = 0, MAT_Partitioning = 0, MAT_ZeroEntries = 0, MAT_Load = 0, MAT_View = 0, MAT_AXPY = 0, MAT_FDColoringCreate = 0;
19: PetscEvent MAT_FDColoringApply = 0,MAT_Transpose = 0,MAT_FDColoringFunction = 0;
20: PetscEvent MAT_MatMult = 0, MAT_MatMultSymbolic = 0, MAT_MatMultNumeric = 0;
21: PetscEvent MAT_PtAP = 0, MAT_PtAPSymbolic = 0, MAT_PtAPNumeric = 0;
22: PetscEvent MAT_MatMultTranspose = 0, MAT_MatMultTransposeSymbolic = 0, MAT_MatMultTransposeNumeric = 0;
23: PetscEvent MAT_Getsymtranspose = 0, MAT_Getsymtransreduced = 0, MAT_Transpose_SeqAIJ = 0, MAT_GetBrowsOfAcols = 0;
24: PetscEvent MAT_GetBrowsOfAocols = 0, MAT_Getlocalmat = 0, MAT_Getlocalmatcondensed = 0, MAT_Seqstompi = 0, MAT_Seqstompinum = 0, MAT_Seqstompisym = 0;
25: PetscEvent MAT_Applypapt = 0, MAT_Applypapt_numeric = 0, MAT_Applypapt_symbolic = 0, MAT_GetSequentialNonzeroStructure = 0;
27: /* nasty global values for MatSetValue() */
28: PetscInt MatSetValue_Row = 0;
29: PetscInt MatSetValue_Column = 0;
30: PetscScalar MatSetValue_Value = 0.0;
34: /*@
35: MatRealPart - Zeros out the imaginary part of the matrix
37: Collective on Mat
39: Input Parameters:
40: . mat - the matrix
42: Level: advanced
45: .seealso: MatImaginaryPart()
46: @*/
48: PetscErrorCode MatRealPart(Mat mat)
49: {
55: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
56: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
57: if (!mat->ops->realpart) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
58: MatPreallocated(mat);
59: (*mat->ops->realpart)(mat);
60: return(0);
61: }
66: /*@
67: MatImaginaryPart - Moves the imaginary part of the matrix to the real part and zeros the imaginary part
69: Collective on Mat
71: Input Parameters:
72: . mat - the matrix
74: Level: advanced
77: .seealso: MatRealPart()
78: @*/
80: PetscErrorCode MatImaginaryPart(Mat mat)
81: {
87: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
88: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
89: if (!mat->ops->imaginarypart) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
90: MatPreallocated(mat);
91: (*mat->ops->imaginarypart)(mat);
92: return(0);
93: }
97: /*@
98: MatMissingDiagonal - Determine if sparse matrix is missing a diagonal entry (or block entry for BAIJ matrices)
100: Collective on Mat
102: Input Parameter:
103: . mat - the matrix
105: Output Parameters:
106: + missing - is any diagonal missing
107: - dd - first diagonal entry that is missing (optional)
109: Level: advanced
112: .seealso: MatRealPart()
113: @*/
115: PetscErrorCode MatMissingDiagonal(Mat mat,PetscTruth *missing,PetscInt *dd)
116: {
122: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
123: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
124: if (!mat->ops->missingdiagonal) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
125: (*mat->ops->missingdiagonal)(mat,missing,dd);
126: return(0);
127: }
131: /*@C
132: MatGetRow - Gets a row of a matrix. You MUST call MatRestoreRow()
133: for each row that you get to ensure that your application does
134: not bleed memory.
136: Not Collective
138: Input Parameters:
139: + mat - the matrix
140: - row - the row to get
142: Output Parameters:
143: + ncols - if not NULL, the number of nonzeros in the row
144: . cols - if not NULL, the column numbers
145: - vals - if not NULL, the values
147: Notes:
148: This routine is provided for people who need to have direct access
149: to the structure of a matrix. We hope that we provide enough
150: high-level matrix routines that few users will need it.
152: MatGetRow() always returns 0-based column indices, regardless of
153: whether the internal representation is 0-based (default) or 1-based.
155: For better efficiency, set cols and/or vals to PETSC_NULL if you do
156: not wish to extract these quantities.
158: The user can only examine the values extracted with MatGetRow();
159: the values cannot be altered. To change the matrix entries, one
160: must use MatSetValues().
162: You can only have one call to MatGetRow() outstanding for a particular
163: matrix at a time, per processor. MatGetRow() can only obtain rows
164: associated with the given processor, it cannot get rows from the
165: other processors; for that we suggest using MatGetSubMatrices(), then
166: MatGetRow() on the submatrix. The row indix passed to MatGetRows()
167: is in the global number of rows.
169: Fortran Notes:
170: The calling sequence from Fortran is
171: .vb
172: MatGetRow(matrix,row,ncols,cols,values,ierr)
173: Mat matrix (input)
174: integer row (input)
175: integer ncols (output)
176: integer cols(maxcols) (output)
177: double precision (or double complex) values(maxcols) output
178: .ve
179: where maxcols >= maximum nonzeros in any row of the matrix.
182: Caution:
183: Do not try to change the contents of the output arrays (cols and vals).
184: In some cases, this may corrupt the matrix.
186: Level: advanced
188: Concepts: matrices^row access
190: .seealso: MatRestoreRow(), MatSetValues(), MatGetValues(), MatGetSubMatrices(), MatGetDiagonal()
191: @*/
193: PetscErrorCode MatGetRow(Mat mat,PetscInt row,PetscInt *ncols,const PetscInt *cols[],const PetscScalar *vals[])
194: {
196: PetscInt incols;
201: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
202: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
203: if (!mat->ops->getrow) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
204: MatPreallocated(mat);
206: (*mat->ops->getrow)(mat,row,&incols,(PetscInt **)cols,(PetscScalar **)vals);
207: if (ncols) *ncols = incols;
209: return(0);
210: }
214: /*@
215: MatConjugate - replaces the matrix values with their complex conjugates
217: Collective on Mat
219: Input Parameters:
220: . mat - the matrix
222: Level: advanced
224: .seealso: VecConjugate()
225: @*/
226: PetscErrorCode MatConjugate(Mat mat)
227: {
232: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
233: if (!mat->ops->conjugate) SETERRQ(PETSC_ERR_SUP,"Not provided for this matrix format, send email to petsc-maint@mcs.anl.gov");
234: (*mat->ops->conjugate)(mat);
235: return(0);
236: }
240: /*@C
241: MatRestoreRow - Frees any temporary space allocated by MatGetRow().
243: Not Collective
245: Input Parameters:
246: + mat - the matrix
247: . row - the row to get
248: . ncols, cols - the number of nonzeros and their columns
249: - vals - if nonzero the column values
251: Notes:
252: This routine should be called after you have finished examining the entries.
254: Fortran Notes:
255: The calling sequence from Fortran is
256: .vb
257: MatRestoreRow(matrix,row,ncols,cols,values,ierr)
258: Mat matrix (input)
259: integer row (input)
260: integer ncols (output)
261: integer cols(maxcols) (output)
262: double precision (or double complex) values(maxcols) output
263: .ve
264: Where maxcols >= maximum nonzeros in any row of the matrix.
266: In Fortran MatRestoreRow() MUST be called after MatGetRow()
267: before another call to MatGetRow() can be made.
269: Level: advanced
271: .seealso: MatGetRow()
272: @*/
273: PetscErrorCode MatRestoreRow(Mat mat,PetscInt row,PetscInt *ncols,const PetscInt *cols[],const PetscScalar *vals[])
274: {
280: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
281: if (!mat->ops->restorerow) return(0);
282: (*mat->ops->restorerow)(mat,row,ncols,(PetscInt **)cols,(PetscScalar **)vals);
283: return(0);
284: }
288: /*@C
289: MatGetRowUpperTriangular - Sets a flag to enable calls to MatGetRow() for matrix in MATSBAIJ format.
290: You should call MatRestoreRowUpperTriangular() after calling MatGetRow/MatRestoreRow() to disable the flag.
292: Not Collective
294: Input Parameters:
295: + mat - the matrix
297: Notes:
298: The flag is to ensure that users are aware of MatGetRow() only provides the upper trianglular part of the row for the matrices in MATSBAIJ format.
300: Level: advanced
302: Concepts: matrices^row access
304: .seealso: MatRestoreRowRowUpperTriangular()
305: @*/
307: PetscErrorCode MatGetRowUpperTriangular(Mat mat)
308: {
314: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
315: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
316: if (!mat->ops->getrowuppertriangular) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
317: MatPreallocated(mat);
318: (*mat->ops->getrowuppertriangular)(mat);
319: return(0);
320: }
324: /*@C
325: MatRestoreRowUpperTriangular - Disable calls to MatGetRow() for matrix in MATSBAIJ format.
327: Not Collective
329: Input Parameters:
330: + mat - the matrix
332: Notes:
333: This routine should be called after you have finished MatGetRow/MatRestoreRow().
336: Level: advanced
338: .seealso: MatGetRowUpperTriangular()
339: @*/
340: PetscErrorCode MatRestoreRowUpperTriangular(Mat mat)
341: {
346: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
347: if (!mat->ops->restorerowuppertriangular) return(0);
348: (*mat->ops->restorerowuppertriangular)(mat);
349: return(0);
350: }
354: /*@C
355: MatSetOptionsPrefix - Sets the prefix used for searching for all
356: Mat options in the database.
358: Collective on Mat
360: Input Parameter:
361: + A - the Mat context
362: - prefix - the prefix to prepend to all option names
364: Notes:
365: A hyphen (-) must NOT be given at the beginning of the prefix name.
366: The first character of all runtime options is AUTOMATICALLY the hyphen.
368: Level: advanced
370: .keywords: Mat, set, options, prefix, database
372: .seealso: MatSetFromOptions()
373: @*/
374: PetscErrorCode MatSetOptionsPrefix(Mat A,const char prefix[])
375: {
380: PetscObjectSetOptionsPrefix((PetscObject)A,prefix);
381: return(0);
382: }
386: /*@C
387: MatAppendOptionsPrefix - Appends to the prefix used for searching for all
388: Mat options in the database.
390: Collective on Mat
392: Input Parameters:
393: + A - the Mat context
394: - prefix - the prefix to prepend to all option names
396: Notes:
397: A hyphen (-) must NOT be given at the beginning of the prefix name.
398: The first character of all runtime options is AUTOMATICALLY the hyphen.
400: Level: advanced
402: .keywords: Mat, append, options, prefix, database
404: .seealso: MatGetOptionsPrefix()
405: @*/
406: PetscErrorCode MatAppendOptionsPrefix(Mat A,const char prefix[])
407: {
409:
412: PetscObjectAppendOptionsPrefix((PetscObject)A,prefix);
413: return(0);
414: }
418: /*@C
419: MatGetOptionsPrefix - Sets the prefix used for searching for all
420: Mat options in the database.
422: Not Collective
424: Input Parameter:
425: . A - the Mat context
427: Output Parameter:
428: . prefix - pointer to the prefix string used
430: Notes: On the fortran side, the user should pass in a string 'prefix' of
431: sufficient length to hold the prefix.
433: Level: advanced
435: .keywords: Mat, get, options, prefix, database
437: .seealso: MatAppendOptionsPrefix()
438: @*/
439: PetscErrorCode MatGetOptionsPrefix(Mat A,const char *prefix[])
440: {
445: PetscObjectGetOptionsPrefix((PetscObject)A,prefix);
446: return(0);
447: }
451: /*@
452: MatSetUp - Sets up the internal matrix data structures for the later use.
454: Collective on Mat
456: Input Parameters:
457: . A - the Mat context
459: Notes:
460: For basic use of the Mat classes the user need not explicitly call
461: MatSetUp(), since these actions will happen automatically.
463: Level: advanced
465: .keywords: Mat, setup
467: .seealso: MatCreate(), MatDestroy()
468: @*/
469: PetscErrorCode MatSetUp(Mat A)
470: {
471: PetscMPIInt size;
476: if (!((PetscObject)A)->type_name) {
477: MPI_Comm_size(((PetscObject)A)->comm, &size);
478: if (size == 1) {
479: MatSetType(A, MATSEQAIJ);
480: } else {
481: MatSetType(A, MATMPIAIJ);
482: }
483: }
484: MatSetUpPreallocation(A);
485: return(0);
486: }
490: /*@C
491: MatView - Visualizes a matrix object.
493: Collective on Mat
495: Input Parameters:
496: + mat - the matrix
497: - viewer - visualization context
499: Notes:
500: The available visualization contexts include
501: + PETSC_VIEWER_STDOUT_SELF - standard output (default)
502: . PETSC_VIEWER_STDOUT_WORLD - synchronized standard
503: output where only the first processor opens
504: the file. All other processors send their
505: data to the first processor to print.
506: - PETSC_VIEWER_DRAW_WORLD - graphical display of nonzero structure
508: The user can open alternative visualization contexts with
509: + PetscViewerASCIIOpen() - Outputs matrix to a specified file
510: . PetscViewerBinaryOpen() - Outputs matrix in binary to a
511: specified file; corresponding input uses MatLoad()
512: . PetscViewerDrawOpen() - Outputs nonzero matrix structure to
513: an X window display
514: - PetscViewerSocketOpen() - Outputs matrix to Socket viewer.
515: Currently only the sequential dense and AIJ
516: matrix types support the Socket viewer.
518: The user can call PetscViewerSetFormat() to specify the output
519: format of ASCII printed objects (when using PETSC_VIEWER_STDOUT_SELF,
520: PETSC_VIEWER_STDOUT_WORLD and PetscViewerASCIIOpen). Available formats include
521: + PETSC_VIEWER_ASCII_DEFAULT - default, prints matrix contents
522: . PETSC_VIEWER_ASCII_MATLAB - prints matrix contents in Matlab format
523: . PETSC_VIEWER_ASCII_DENSE - prints entire matrix including zeros
524: . PETSC_VIEWER_ASCII_COMMON - prints matrix contents, using a sparse
525: format common among all matrix types
526: . PETSC_VIEWER_ASCII_IMPL - prints matrix contents, using an implementation-specific
527: format (which is in many cases the same as the default)
528: . PETSC_VIEWER_ASCII_INFO - prints basic information about the matrix
529: size and structure (not the matrix entries)
530: . PETSC_VIEWER_ASCII_INFO_DETAIL - prints more detailed information about
531: the matrix structure
533: Options Database Keys:
534: + -mat_view_info - Prints info on matrix at conclusion of MatEndAssembly()
535: . -mat_view_info_detailed - Prints more detailed info
536: . -mat_view - Prints matrix in ASCII format
537: . -mat_view_matlab - Prints matrix in Matlab format
538: . -mat_view_draw - PetscDraws nonzero structure of matrix, using MatView() and PetscDrawOpenX().
539: . -display <name> - Sets display name (default is host)
540: . -draw_pause <sec> - Sets number of seconds to pause after display
541: . -mat_view_socket - Sends matrix to socket, can be accessed from Matlab (see users manual)
542: . -viewer_socket_machine <machine>
543: . -viewer_socket_port <port>
544: . -mat_view_binary - save matrix to file in binary format
545: - -viewer_binary_filename <name>
546: Level: beginner
548: Concepts: matrices^viewing
549: Concepts: matrices^plotting
550: Concepts: matrices^printing
552: .seealso: PetscViewerSetFormat(), PetscViewerASCIIOpen(), PetscViewerDrawOpen(),
553: PetscViewerSocketOpen(), PetscViewerBinaryOpen(), MatLoad()
554: @*/
555: PetscErrorCode MatView(Mat mat,PetscViewer viewer)
556: {
557: PetscErrorCode ierr;
558: PetscInt rows,cols;
559: PetscTruth iascii;
560: const char *cstr;
561: PetscViewerFormat format;
566: if (!viewer) {
567: PetscViewerASCIIGetStdout(((PetscObject)mat)->comm,&viewer);
568: }
571: if (!mat->assembled) SETERRQ(PETSC_ERR_ORDER,"Must call MatAssemblyBegin/End() before viewing matrix");
572: MatPreallocated(mat);
574: PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&iascii);
575: if (iascii) {
576: PetscViewerGetFormat(viewer,&format);
577: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
578: if (((PetscObject)mat)->prefix) {
579: PetscViewerASCIIPrintf(viewer,"Matrix Object:(%s)\n",((PetscObject)mat)->prefix);
580: } else {
581: PetscViewerASCIIPrintf(viewer,"Matrix Object:\n");
582: }
583: PetscViewerASCIIPushTab(viewer);
584: MatGetType(mat,&cstr);
585: MatGetSize(mat,&rows,&cols);
586: PetscViewerASCIIPrintf(viewer,"type=%s, rows=%D, cols=%D\n",cstr,rows,cols);
587: if (mat->ops->getinfo) {
588: MatInfo info;
589: MatGetInfo(mat,MAT_GLOBAL_SUM,&info);
590: PetscViewerASCIIPrintf(viewer,"total: nonzeros=%D, allocated nonzeros=%D\n",
591: (PetscInt)info.nz_used,(PetscInt)info.nz_allocated);
592: }
593: }
594: }
595: if (mat->ops->view) {
596: PetscViewerASCIIPushTab(viewer);
597: (*mat->ops->view)(mat,viewer);
598: PetscViewerASCIIPopTab(viewer);
599: } else if (!iascii) {
600: SETERRQ1(PETSC_ERR_SUP,"Viewer type %s not supported",((PetscObject)viewer)->type_name);
601: }
602: if (iascii) {
603: PetscViewerGetFormat(viewer,&format);
604: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
605: PetscViewerASCIIPopTab(viewer);
606: }
607: }
608: return(0);
609: }
613: /*@
614: MatScaleSystem - Scale a vector solution and right hand side to
615: match the scaling of a scaled matrix.
616:
617: Collective on Mat
619: Input Parameter:
620: + mat - the matrix
621: . b - right hand side vector (or PETSC_NULL)
622: - x - solution vector (or PETSC_NULL)
625: Notes:
626: For AIJ, BAIJ, and BDiag matrix formats, the matrices are not
627: internally scaled, so this does nothing. For MPIROWBS it
628: permutes and diagonally scales.
630: The KSP methods automatically call this routine when required
631: (via PCPreSolve()) so it is rarely used directly.
633: Level: Developer
635: Concepts: matrices^scaling
637: .seealso: MatUseScaledForm(), MatUnScaleSystem()
638: @*/
639: PetscErrorCode MatScaleSystem(Mat mat,Vec b,Vec x)
640: {
646: MatPreallocated(mat);
650: if (mat->ops->scalesystem) {
651: (*mat->ops->scalesystem)(mat,b,x);
652: }
653: return(0);
654: }
658: /*@
659: MatUnScaleSystem - Unscales a vector solution and right hand side to
660: match the original scaling of a scaled matrix.
661:
662: Collective on Mat
664: Input Parameter:
665: + mat - the matrix
666: . b - right hand side vector (or PETSC_NULL)
667: - x - solution vector (or PETSC_NULL)
670: Notes:
671: For AIJ, BAIJ, and BDiag matrix formats, the matrices are not
672: internally scaled, so this does nothing. For MPIROWBS it
673: permutes and diagonally scales.
675: The KSP methods automatically call this routine when required
676: (via PCPreSolve()) so it is rarely used directly.
678: Level: Developer
680: .seealso: MatUseScaledForm(), MatScaleSystem()
681: @*/
682: PetscErrorCode MatUnScaleSystem(Mat mat,Vec b,Vec x)
683: {
689: MatPreallocated(mat);
692: if (mat->ops->unscalesystem) {
693: (*mat->ops->unscalesystem)(mat,b,x);
694: }
695: return(0);
696: }
700: /*@
701: MatUseScaledForm - For matrix storage formats that scale the
702: matrix (for example MPIRowBS matrices are diagonally scaled on
703: assembly) indicates matrix operations (MatMult() etc) are
704: applied using the scaled matrix.
705:
706: Collective on Mat
708: Input Parameter:
709: + mat - the matrix
710: - scaled - PETSC_TRUE for applying the scaled, PETSC_FALSE for
711: applying the original matrix
713: Notes:
714: For scaled matrix formats, applying the original, unscaled matrix
715: will be slightly more expensive
717: Level: Developer
719: .seealso: MatScaleSystem(), MatUnScaleSystem()
720: @*/
721: PetscErrorCode MatUseScaledForm(Mat mat,PetscTruth scaled)
722: {
728: MatPreallocated(mat);
729: if (mat->ops->usescaledform) {
730: (*mat->ops->usescaledform)(mat,scaled);
731: }
732: return(0);
733: }
737: /*@
738: MatDestroy - Frees space taken by a matrix.
739:
740: Collective on Mat
742: Input Parameter:
743: . A - the matrix
745: Level: beginner
747: @*/
748: PetscErrorCode MatDestroy(Mat A)
749: {
753: if (--((PetscObject)A)->refct > 0) return(0);
754: MatPreallocated(A);
755: /* if memory was published with AMS then destroy it */
756: PetscObjectDepublish(A);
757: if (A->ops->destroy) {
758: (*A->ops->destroy)(A);
759: }
760: if (A->mapping) {
761: ISLocalToGlobalMappingDestroy(A->mapping);
762: }
763: if (A->bmapping) {
764: ISLocalToGlobalMappingDestroy(A->bmapping);
765: }
766: PetscFree(A->rmap.range);
767: PetscFree(A->cmap.range);
768: if (A->spptr){PetscFree(A->spptr);}
769: PetscHeaderDestroy(A);
770: return(0);
771: }
775: /*@
776: MatValid - Checks whether a matrix object is valid.
778: Collective on Mat
780: Input Parameter:
781: . m - the matrix to check
783: Output Parameter:
784: flg - flag indicating matrix status, either
785: PETSC_TRUE if matrix is valid, or PETSC_FALSE otherwise.
787: Level: developer
789: Concepts: matrices^validity
790: @*/
791: PetscErrorCode MatValid(Mat m,PetscTruth *flg)
792: {
795: if (!m) *flg = PETSC_FALSE;
796: else if (((PetscObject)m)->cookie != MAT_COOKIE) *flg = PETSC_FALSE;
797: else *flg = PETSC_TRUE;
798: return(0);
799: }
803: /*@
804: MatSetValues - Inserts or adds a block of values into a matrix.
805: These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
806: MUST be called after all calls to MatSetValues() have been completed.
808: Not Collective
810: Input Parameters:
811: + mat - the matrix
812: . v - a logically two-dimensional array of values
813: . m, idxm - the number of rows and their global indices
814: . n, idxn - the number of columns and their global indices
815: - addv - either ADD_VALUES or INSERT_VALUES, where
816: ADD_VALUES adds values to any existing entries, and
817: INSERT_VALUES replaces existing entries with new values
819: Notes:
820: By default the values, v, are row-oriented and unsorted.
821: See MatSetOption() for other options.
823: Calls to MatSetValues() with the INSERT_VALUES and ADD_VALUES
824: options cannot be mixed without intervening calls to the assembly
825: routines.
827: MatSetValues() uses 0-based row and column numbers in Fortran
828: as well as in C.
830: Negative indices may be passed in idxm and idxn, these rows and columns are
831: simply ignored. This allows easily inserting element stiffness matrices
832: with homogeneous Dirchlet boundary conditions that you don't want represented
833: in the matrix.
835: Efficiency Alert:
836: The routine MatSetValuesBlocked() may offer much better efficiency
837: for users of block sparse formats (MATSEQBAIJ and MATMPIBAIJ).
839: Level: beginner
841: Concepts: matrices^putting entries in
843: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
844: InsertMode, INSERT_VALUES, ADD_VALUES
845: @*/
846: PetscErrorCode MatSetValues(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],const PetscScalar v[],InsertMode addv)
847: {
851: if (!m || !n) return(0); /* no values to insert */
857: MatPreallocated(mat);
858: if (mat->insertmode == NOT_SET_VALUES) {
859: mat->insertmode = addv;
860: }
861: #if defined(PETSC_USE_DEBUG)
862: else if (mat->insertmode != addv) {
863: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
864: }
865: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
866: #endif
868: if (mat->assembled) {
869: mat->was_assembled = PETSC_TRUE;
870: mat->assembled = PETSC_FALSE;
871: }
873: if (!mat->ops->setvalues) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
874: (*mat->ops->setvalues)(mat,m,idxm,n,idxn,v,addv);
876: return(0);
877: }
882: /*@
883: MatSetValuesRowLocal - Inserts a row (block row for BAIJ matrices) of nonzero
884: values into a matrix
886: Not Collective
888: Input Parameters:
889: + mat - the matrix
890: . row - the (block) row to set
891: - v - a logically two-dimensional array of values
893: Notes:
894: By the values, v, are column-oriented (for the block version) and sorted
896: All the nonzeros in the row must be provided
898: The matrix must have previously had its column indices set
900: The row must belong to this process
902: Level: intermediate
904: Concepts: matrices^putting entries in
906: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
907: InsertMode, INSERT_VALUES, ADD_VALUES, MatSetValues(), MatSetValuesRow(), MatSetLocalToGlobalMapping()
908: @*/
909: PetscErrorCode MatSetValuesRowLocal(Mat mat,PetscInt row,const PetscScalar v[])
910: {
917: MatSetValuesRow(mat, mat->mapping->indices[row],v);
918: return(0);
919: }
923: /*@
924: MatSetValuesRow - Inserts a row (block row for BAIJ matrices) of nonzero
925: values into a matrix
927: Not Collective
929: Input Parameters:
930: + mat - the matrix
931: . row - the (block) row to set
932: - v - a logically two-dimensional array of values
934: Notes:
935: By the values, v, are column-oriented (for the block version) and sorted
937: All the nonzeros in the row must be provided
939: The matrix must have previously had its column indices set
941: The row must belong to this process
943: Level: intermediate
945: Concepts: matrices^putting entries in
947: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
948: InsertMode, INSERT_VALUES, ADD_VALUES, MatSetValues()
949: @*/
950: PetscErrorCode MatSetValuesRow(Mat mat,PetscInt row,const PetscScalar v[])
951: {
958: #if defined(PETSC_USE_DEBUG)
959: if (mat->insertmode == ADD_VALUES) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add and insert values");
960: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
961: #endif
962: mat->insertmode = INSERT_VALUES;
964: if (mat->assembled) {
965: mat->was_assembled = PETSC_TRUE;
966: mat->assembled = PETSC_FALSE;
967: }
969: if (!mat->ops->setvaluesrow) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
970: (*mat->ops->setvaluesrow)(mat,row,v);
972: return(0);
973: }
977: /*@
978: MatSetValuesStencil - Inserts or adds a block of values into a matrix.
979: Using structured grid indexing
981: Not Collective
983: Input Parameters:
984: + mat - the matrix
985: . v - a logically two-dimensional array of values
986: . m - number of rows being entered
987: . idxm - grid coordinates (and component number when dof > 1) for matrix rows being entered
988: . n - number of columns being entered
989: . idxn - grid coordinates (and component number when dof > 1) for matrix columns being entered
990: - addv - either ADD_VALUES or INSERT_VALUES, where
991: ADD_VALUES adds values to any existing entries, and
992: INSERT_VALUES replaces existing entries with new values
994: Notes:
995: By default the values, v, are row-oriented and unsorted.
996: See MatSetOption() for other options.
998: Calls to MatSetValuesStencil() with the INSERT_VALUES and ADD_VALUES
999: options cannot be mixed without intervening calls to the assembly
1000: routines.
1002: The grid coordinates are across the entire grid, not just the local portion
1004: MatSetValuesStencil() uses 0-based row and column numbers in Fortran
1005: as well as in C.
1007: For setting/accessing vector values via array coordinates you can use the DAVecGetArray() routine
1009: In order to use this routine you must either obtain the matrix with DAGetMatrix()
1010: or call MatSetLocalToGlobalMapping() and MatSetStencil() first.
1012: The columns and rows in the stencil passed in MUST be contained within the
1013: ghost region of the given process as set with DACreateXXX() or MatSetStencil(). For example,
1014: if you create a DA with an overlap of one grid level and on a particular process its first
1015: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1016: first i index you can use in your column and row indices in MatSetStencil() is 5.
1018: In Fortran idxm and idxn should be declared as
1019: $ MatStencil idxm(4,m),idxn(4,n)
1020: and the values inserted using
1021: $ idxm(MatStencil_i,1) = i
1022: $ idxm(MatStencil_j,1) = j
1023: $ idxm(MatStencil_k,1) = k
1024: $ idxm(MatStencil_c,1) = c
1025: etc
1027: For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
1028: obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
1029: etc to obtain values that obtained by wrapping the values from the left edge.
1031: 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
1032: a single value per point) you can skip filling those indices.
1034: Inspired by the structured grid interface to the HYPRE package
1035: (http://www.llnl.gov/CASC/hypre)
1037: Efficiency Alert:
1038: The routine MatSetValuesBlockedStencil() may offer much better efficiency
1039: for users of block sparse formats (MATSEQBAIJ and MATMPIBAIJ).
1041: Level: beginner
1043: Concepts: matrices^putting entries in
1045: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
1046: MatSetValues(), MatSetValuesBlockedStencil(), MatSetStencil(), DAGetMatrix(), DAVecGetArray(), MatStencil
1047: @*/
1048: PetscErrorCode MatSetValuesStencil(Mat mat,PetscInt m,const MatStencil idxm[],PetscInt n,const MatStencil idxn[],const PetscScalar v[],InsertMode addv)
1049: {
1051: PetscInt j,i,jdxm[128],jdxn[256],dim = mat->stencil.dim,*dims = mat->stencil.dims+1,tmp;
1052: PetscInt *starts = mat->stencil.starts,*dxm = (PetscInt*)idxm,*dxn = (PetscInt*)idxn,sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1055: if (!m || !n) return(0); /* no values to insert */
1062: if (m > 128) SETERRQ1(PETSC_ERR_SUP,"Can only set 128 rows at a time; trying to set %D",m);
1063: if (n > 256) SETERRQ1(PETSC_ERR_SUP,"Can only set 256 columns at a time; trying to set %D",n);
1065: for (i=0; i<m; i++) {
1066: for (j=0; j<3-sdim; j++) dxm++;
1067: tmp = *dxm++ - starts[0];
1068: for (j=0; j<dim-1; j++) {
1069: if ((*dxm++ - starts[j+1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
1070: else tmp = tmp*dims[j] + *(dxm-1) - starts[j+1];
1071: }
1072: if (mat->stencil.noc) dxm++;
1073: jdxm[i] = tmp;
1074: }
1075: for (i=0; i<n; i++) {
1076: for (j=0; j<3-sdim; j++) dxn++;
1077: tmp = *dxn++ - starts[0];
1078: for (j=0; j<dim-1; j++) {
1079: if ((*dxn++ - starts[j+1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
1080: else tmp = tmp*dims[j] + *(dxn-1) - starts[j+1];
1081: }
1082: if (mat->stencil.noc) dxn++;
1083: jdxn[i] = tmp;
1084: }
1085: MatSetValuesLocal(mat,m,jdxm,n,jdxn,v,addv);
1086: return(0);
1087: }
1091: /*@C
1092: MatSetValuesBlockedStencil - Inserts or adds a block of values into a matrix.
1093: Using structured grid indexing
1095: Not Collective
1097: Input Parameters:
1098: + mat - the matrix
1099: . v - a logically two-dimensional array of values
1100: . m - number of rows being entered
1101: . idxm - grid coordinates for matrix rows being entered
1102: . n - number of columns being entered
1103: . idxn - grid coordinates for matrix columns being entered
1104: - addv - either ADD_VALUES or INSERT_VALUES, where
1105: ADD_VALUES adds values to any existing entries, and
1106: INSERT_VALUES replaces existing entries with new values
1108: Notes:
1109: By default the values, v, are row-oriented and unsorted.
1110: See MatSetOption() for other options.
1112: Calls to MatSetValuesBlockedStencil() with the INSERT_VALUES and ADD_VALUES
1113: options cannot be mixed without intervening calls to the assembly
1114: routines.
1116: The grid coordinates are across the entire grid, not just the local portion
1118: MatSetValuesBlockedStencil() uses 0-based row and column numbers in Fortran
1119: as well as in C.
1121: For setting/accessing vector values via array coordinates you can use the DAVecGetArray() routine
1123: In order to use this routine you must either obtain the matrix with DAGetMatrix()
1124: or call MatSetLocalToGlobalMapping() and MatSetStencil() first.
1126: The columns and rows in the stencil passed in MUST be contained within the
1127: ghost region of the given process as set with DACreateXXX() or MatSetStencil(). For example,
1128: if you create a DA with an overlap of one grid level and on a particular process its first
1129: local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1130: first i index you can use in your column and row indices in MatSetStencil() is 5.
1132: In Fortran idxm and idxn should be declared as
1133: $ MatStencil idxm(4,m),idxn(4,n)
1134: and the values inserted using
1135: $ idxm(MatStencil_i,1) = i
1136: $ idxm(MatStencil_j,1) = j
1137: $ idxm(MatStencil_k,1) = k
1138: etc
1140: Negative indices may be passed in idxm and idxn, these rows and columns are
1141: simply ignored. This allows easily inserting element stiffness matrices
1142: with homogeneous Dirchlet boundary conditions that you don't want represented
1143: in the matrix.
1145: Inspired by the structured grid interface to the HYPRE package
1146: (http://www.llnl.gov/CASC/hypre)
1148: Level: beginner
1150: Concepts: matrices^putting entries in
1152: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
1153: MatSetValues(), MatSetValuesStencil(), MatSetStencil(), DAGetMatrix(), DAVecGetArray(), MatStencil
1154: @*/
1155: PetscErrorCode MatSetValuesBlockedStencil(Mat mat,PetscInt m,const MatStencil idxm[],PetscInt n,const MatStencil idxn[],const PetscScalar v[],InsertMode addv)
1156: {
1158: PetscInt j,i,jdxm[128],jdxn[256],dim = mat->stencil.dim,*dims = mat->stencil.dims+1,tmp;
1159: PetscInt *starts = mat->stencil.starts,*dxm = (PetscInt*)idxm,*dxn = (PetscInt*)idxn,sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1162: if (!m || !n) return(0); /* no values to insert */
1169: if (m > 128) SETERRQ1(PETSC_ERR_SUP,"Can only set 128 rows at a time; trying to set %D",m);
1170: if (n > 128) SETERRQ1(PETSC_ERR_SUP,"Can only set 256 columns at a time; trying to set %D",n);
1172: for (i=0; i<m; i++) {
1173: for (j=0; j<3-sdim; j++) dxm++;
1174: tmp = *dxm++ - starts[0];
1175: for (j=0; j<sdim-1; j++) {
1176: if ((*dxm++ - starts[j+1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
1177: else tmp = tmp*dims[j] + *(dxm-1) - starts[j+1];
1178: }
1179: dxm++;
1180: jdxm[i] = tmp;
1181: }
1182: for (i=0; i<n; i++) {
1183: for (j=0; j<3-sdim; j++) dxn++;
1184: tmp = *dxn++ - starts[0];
1185: for (j=0; j<sdim-1; j++) {
1186: if ((*dxn++ - starts[j+1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
1187: else tmp = tmp*dims[j] + *(dxn-1) - starts[j+1];
1188: }
1189: dxn++;
1190: jdxn[i] = tmp;
1191: }
1192: MatSetValuesBlockedLocal(mat,m,jdxm,n,jdxn,v,addv);
1193: return(0);
1194: }
1198: /*@
1199: MatSetStencil - Sets the grid information for setting values into a matrix via
1200: MatSetValuesStencil()
1202: Not Collective
1204: Input Parameters:
1205: + mat - the matrix
1206: . dim - dimension of the grid 1, 2, or 3
1207: . dims - number of grid points in x, y, and z direction, including ghost points on your processor
1208: . starts - starting point of ghost nodes on your processor in x, y, and z direction
1209: - dof - number of degrees of freedom per node
1212: Inspired by the structured grid interface to the HYPRE package
1213: (www.llnl.gov/CASC/hyper)
1215: For matrices generated with DAGetMatrix() this routine is automatically called and so not needed by the
1216: user.
1217:
1218: Level: beginner
1220: Concepts: matrices^putting entries in
1222: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
1223: MatSetValues(), MatSetValuesBlockedStencil(), MatSetValuesStencil()
1224: @*/
1225: PetscErrorCode MatSetStencil(Mat mat,PetscInt dim,const PetscInt dims[],const PetscInt starts[],PetscInt dof)
1226: {
1227: PetscInt i;
1234: mat->stencil.dim = dim + (dof > 1);
1235: for (i=0; i<dim; i++) {
1236: mat->stencil.dims[i] = dims[dim-i-1]; /* copy the values in backwards */
1237: mat->stencil.starts[i] = starts[dim-i-1];
1238: }
1239: mat->stencil.dims[dim] = dof;
1240: mat->stencil.starts[dim] = 0;
1241: mat->stencil.noc = (PetscTruth)(dof == 1);
1242: return(0);
1243: }
1247: /*@
1248: MatSetValuesBlocked - Inserts or adds a block of values into a matrix.
1250: Not Collective
1252: Input Parameters:
1253: + mat - the matrix
1254: . v - a logically two-dimensional array of values
1255: . m, idxm - the number of block rows and their global block indices
1256: . n, idxn - the number of block columns and their global block indices
1257: - addv - either ADD_VALUES or INSERT_VALUES, where
1258: ADD_VALUES adds values to any existing entries, and
1259: INSERT_VALUES replaces existing entries with new values
1261: Notes:
1262: The m and n count the NUMBER of blocks in the row direction and column direction,
1263: NOT the total number of rows/columns; for example, if the block size is 2 and
1264: you are passing in values for rows 2,3,4,5 then m would be 2 (not 4).
1266: By default the values, v, are row-oriented and unsorted. So the layout of
1267: v is the same as for MatSetValues(). See MatSetOption() for other options.
1269: Calls to MatSetValuesBlocked() with the INSERT_VALUES and ADD_VALUES
1270: options cannot be mixed without intervening calls to the assembly
1271: routines.
1273: MatSetValuesBlocked() uses 0-based row and column numbers in Fortran
1274: as well as in C.
1276: Negative indices may be passed in idxm and idxn, these rows and columns are
1277: simply ignored. This allows easily inserting element stiffness matrices
1278: with homogeneous Dirchlet boundary conditions that you don't want represented
1279: in the matrix.
1281: Each time an entry is set within a sparse matrix via MatSetValues(),
1282: internal searching must be done to determine where to place the the
1283: data in the matrix storage space. By instead inserting blocks of
1284: entries via MatSetValuesBlocked(), the overhead of matrix assembly is
1285: reduced.
1287: Example:
1288: $ Suppose m=n=2 and block size(bs) = 2 The matrix is
1289: $
1290: $ 1 2 | 3 4
1291: $ 5 6 | 7 8
1292: $ - - - | - - -
1293: $ 9 10 | 11 12
1294: $ 13 14 | 15 16
1295: $
1296: $ v[] should be passed in like
1297: $ v[] = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]
1299: Restrictions:
1300: MatSetValuesBlocked() is currently supported only for the BAIJ and SBAIJ formats
1302: Level: intermediate
1304: Concepts: matrices^putting entries in blocked
1306: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesBlockedLocal()
1307: @*/
1308: PetscErrorCode MatSetValuesBlocked(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],const PetscScalar v[],InsertMode addv)
1309: {
1313: if (!m || !n) return(0); /* no values to insert */
1319: MatPreallocated(mat);
1320: if (mat->insertmode == NOT_SET_VALUES) {
1321: mat->insertmode = addv;
1322: }
1323: #if defined(PETSC_USE_DEBUG)
1324: else if (mat->insertmode != addv) {
1325: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
1326: }
1327: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1328: #endif
1330: if (mat->assembled) {
1331: mat->was_assembled = PETSC_TRUE;
1332: mat->assembled = PETSC_FALSE;
1333: }
1335: if (!mat->ops->setvaluesblocked) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1336: (*mat->ops->setvaluesblocked)(mat,m,idxm,n,idxn,v,addv);
1338: return(0);
1339: }
1343: /*@
1344: MatGetValues - Gets a block of values from a matrix.
1346: Not Collective; currently only returns a local block
1348: Input Parameters:
1349: + mat - the matrix
1350: . v - a logically two-dimensional array for storing the values
1351: . m, idxm - the number of rows and their global indices
1352: - n, idxn - the number of columns and their global indices
1354: Notes:
1355: The user must allocate space (m*n PetscScalars) for the values, v.
1356: The values, v, are then returned in a row-oriented format,
1357: analogous to that used by default in MatSetValues().
1359: MatGetValues() uses 0-based row and column numbers in
1360: Fortran as well as in C.
1362: MatGetValues() requires that the matrix has been assembled
1363: with MatAssemblyBegin()/MatAssemblyEnd(). Thus, calls to
1364: MatSetValues() and MatGetValues() CANNOT be made in succession
1365: without intermediate matrix assembly.
1367: Negative row or column indices will be ignored and those locations in v[] will be
1368: left unchanged.
1370: Level: advanced
1372: Concepts: matrices^accessing values
1374: .seealso: MatGetRow(), MatGetSubMatrices(), MatSetValues()
1375: @*/
1376: PetscErrorCode MatGetValues(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],PetscScalar v[])
1377: {
1386: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1387: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1388: if (!mat->ops->getvalues) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1389: MatPreallocated(mat);
1392: (*mat->ops->getvalues)(mat,m,idxm,n,idxn,v);
1394: return(0);
1395: }
1399: /*@
1400: MatSetLocalToGlobalMapping - Sets a local-to-global numbering for use by
1401: the routine MatSetValuesLocal() to allow users to insert matrix entries
1402: using a local (per-processor) numbering.
1404: Not Collective
1406: Input Parameters:
1407: + x - the matrix
1408: - mapping - mapping created with ISLocalToGlobalMappingCreate()
1409: or ISLocalToGlobalMappingCreateIS()
1411: Level: intermediate
1413: Concepts: matrices^local to global mapping
1414: Concepts: local to global mapping^for matrices
1416: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesLocal()
1417: @*/
1418: PetscErrorCode MatSetLocalToGlobalMapping(Mat x,ISLocalToGlobalMapping mapping)
1419: {
1425: if (x->mapping) {
1426: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Mapping already set for matrix");
1427: }
1428: MatPreallocated(x);
1430: if (x->ops->setlocaltoglobalmapping) {
1431: (*x->ops->setlocaltoglobalmapping)(x,mapping);
1432: } else {
1433: PetscObjectReference((PetscObject)mapping);
1434: if (x->mapping) { ISLocalToGlobalMappingDestroy(x->mapping); }
1435: x->mapping = mapping;
1436: }
1437: return(0);
1438: }
1442: /*@
1443: MatSetLocalToGlobalMappingBlock - Sets a local-to-global numbering for use
1444: by the routine MatSetValuesBlockedLocal() to allow users to insert matrix
1445: entries using a local (per-processor) numbering.
1447: Not Collective
1449: Input Parameters:
1450: + x - the matrix
1451: - mapping - mapping created with ISLocalToGlobalMappingCreate() or
1452: ISLocalToGlobalMappingCreateIS()
1454: Level: intermediate
1456: Concepts: matrices^local to global mapping blocked
1457: Concepts: local to global mapping^for matrices, blocked
1459: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesBlockedLocal(),
1460: MatSetValuesBlocked(), MatSetValuesLocal()
1461: @*/
1462: PetscErrorCode MatSetLocalToGlobalMappingBlock(Mat x,ISLocalToGlobalMapping mapping)
1463: {
1469: if (x->bmapping) {
1470: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Mapping already set for matrix");
1471: }
1472: PetscObjectReference((PetscObject)mapping);
1473: if (x->bmapping) { ISLocalToGlobalMappingDestroy(x->mapping); }
1474: x->bmapping = mapping;
1475: return(0);
1476: }
1480: /*@
1481: MatSetValuesLocal - Inserts or adds values into certain locations of a matrix,
1482: using a local ordering of the nodes.
1484: Not Collective
1486: Input Parameters:
1487: + x - the matrix
1488: . nrow, irow - number of rows and their local indices
1489: . ncol, icol - number of columns and their local indices
1490: . y - a logically two-dimensional array of values
1491: - addv - either INSERT_VALUES or ADD_VALUES, where
1492: ADD_VALUES adds values to any existing entries, and
1493: INSERT_VALUES replaces existing entries with new values
1495: Notes:
1496: Before calling MatSetValuesLocal(), the user must first set the
1497: local-to-global mapping by calling MatSetLocalToGlobalMapping().
1499: Calls to MatSetValuesLocal() with the INSERT_VALUES and ADD_VALUES
1500: options cannot be mixed without intervening calls to the assembly
1501: routines.
1503: These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
1504: MUST be called after all calls to MatSetValuesLocal() have been completed.
1506: Level: intermediate
1508: Concepts: matrices^putting entries in with local numbering
1510: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetLocalToGlobalMapping(),
1511: MatSetValueLocal()
1512: @*/
1513: PetscErrorCode MatSetValuesLocal(Mat mat,PetscInt nrow,const PetscInt irow[],PetscInt ncol,const PetscInt icol[],const PetscScalar y[],InsertMode addv)
1514: {
1516: PetscInt irowm[2048],icolm[2048];
1524: MatPreallocated(mat);
1525: if (mat->insertmode == NOT_SET_VALUES) {
1526: mat->insertmode = addv;
1527: }
1528: #if defined(PETSC_USE_DEBUG)
1529: else if (mat->insertmode != addv) {
1530: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
1531: }
1532: if (!mat->ops->setvalueslocal && (nrow > 2048 || ncol > 2048)) {
1533: SETERRQ2(PETSC_ERR_SUP,"Number column/row indices must be <= 2048: are %D %D",nrow,ncol);
1534: }
1535: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1536: #endif
1538: if (mat->assembled) {
1539: mat->was_assembled = PETSC_TRUE;
1540: mat->assembled = PETSC_FALSE;
1541: }
1543: if (!mat->ops->setvalueslocal) {
1544: ISLocalToGlobalMappingApply(mat->mapping,nrow,irow,irowm);
1545: ISLocalToGlobalMappingApply(mat->mapping,ncol,icol,icolm);
1546: (*mat->ops->setvalues)(mat,nrow,irowm,ncol,icolm,y,addv);
1547: } else {
1548: (*mat->ops->setvalueslocal)(mat,nrow,irow,ncol,icol,y,addv);
1549: }
1550: mat->same_nonzero = PETSC_FALSE;
1552: return(0);
1553: }
1557: /*@
1558: MatSetValuesBlockedLocal - Inserts or adds values into certain locations of a matrix,
1559: using a local ordering of the nodes a block at a time.
1561: Not Collective
1563: Input Parameters:
1564: + x - the matrix
1565: . nrow, irow - number of rows and their local indices
1566: . ncol, icol - number of columns and their local indices
1567: . y - a logically two-dimensional array of values
1568: - addv - either INSERT_VALUES or ADD_VALUES, where
1569: ADD_VALUES adds values to any existing entries, and
1570: INSERT_VALUES replaces existing entries with new values
1572: Notes:
1573: Before calling MatSetValuesBlockedLocal(), the user must first set the
1574: local-to-global mapping by calling MatSetLocalToGlobalMappingBlock(),
1575: where the mapping MUST be set for matrix blocks, not for matrix elements.
1577: Calls to MatSetValuesBlockedLocal() with the INSERT_VALUES and ADD_VALUES
1578: options cannot be mixed without intervening calls to the assembly
1579: routines.
1581: These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
1582: MUST be called after all calls to MatSetValuesBlockedLocal() have been completed.
1584: Level: intermediate
1586: Concepts: matrices^putting blocked values in with local numbering
1588: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesLocal(), MatSetLocalToGlobalMappingBlock(), MatSetValuesBlocked()
1589: @*/
1590: PetscErrorCode MatSetValuesBlockedLocal(Mat mat,PetscInt nrow,const PetscInt irow[],PetscInt ncol,const PetscInt icol[],const PetscScalar y[],InsertMode addv)
1591: {
1593: PetscInt irowm[2048],icolm[2048];
1601: MatPreallocated(mat);
1602: if (mat->insertmode == NOT_SET_VALUES) {
1603: mat->insertmode = addv;
1604: }
1605: #if defined(PETSC_USE_DEBUG)
1606: else if (mat->insertmode != addv) {
1607: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
1608: }
1609: if (!mat->bmapping) {
1610: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Local to global never set with MatSetLocalToGlobalMappingBlock()");
1611: }
1612: if (nrow > 2048 || ncol > 2048) {
1613: SETERRQ2(PETSC_ERR_SUP,"Number column/row indices must be <= 2048: are %D %D",nrow,ncol);
1614: }
1615: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1616: #endif
1618: if (mat->assembled) {
1619: mat->was_assembled = PETSC_TRUE;
1620: mat->assembled = PETSC_FALSE;
1621: }
1623: ISLocalToGlobalMappingApply(mat->bmapping,nrow,irow,irowm);
1624: ISLocalToGlobalMappingApply(mat->bmapping,ncol,icol,icolm);
1625: (*mat->ops->setvaluesblocked)(mat,nrow,irowm,ncol,icolm,y,addv);
1627: return(0);
1628: }
1630: /* --------------------------------------------------------*/
1633: /*@
1634: MatMult - Computes the matrix-vector product, y = Ax.
1636: Collective on Mat and Vec
1638: Input Parameters:
1639: + mat - the matrix
1640: - x - the vector to be multiplied
1642: Output Parameters:
1643: . y - the result
1645: Notes:
1646: The vectors x and y cannot be the same. I.e., one cannot
1647: call MatMult(A,y,y).
1649: Level: beginner
1651: Concepts: matrix-vector product
1653: .seealso: MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
1654: @*/
1655: PetscErrorCode MatMult(Mat mat,Vec x,Vec y)
1656: {
1665: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1666: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1667: if (x == y) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
1668: #ifndef PETSC_HAVE_CONSTRAINTS
1669: if (mat->cmap.N != x->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap.N,x->map.N);
1670: if (mat->rmap.N != y->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->rmap.N,y->map.N);
1671: if (mat->rmap.n != y->map.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: local dim %D %D",mat->rmap.n,y->map.n);
1672: #endif
1673: MatPreallocated(mat);
1675: if (mat->nullsp) {
1676: MatNullSpaceRemove(mat->nullsp,x,&x);
1677: }
1679: if (!mat->ops->mult) SETERRQ(PETSC_ERR_SUP,"This matrix type does not have a multiply defined");
1681: (*mat->ops->mult)(mat,x,y);
1684: if (mat->nullsp) {
1685: MatNullSpaceRemove(mat->nullsp,y,PETSC_NULL);
1686: }
1687: PetscObjectStateIncrease((PetscObject)y);
1688: return(0);
1689: }
1693: /*@
1694: MatMultTranspose - Computes matrix transpose times a vector.
1696: Collective on Mat and Vec
1698: Input Parameters:
1699: + mat - the matrix
1700: - x - the vector to be multilplied
1702: Output Parameters:
1703: . y - the result
1705: Notes:
1706: The vectors x and y cannot be the same. I.e., one cannot
1707: call MatMultTranspose(A,y,y).
1709: Level: beginner
1711: Concepts: matrix vector product^transpose
1713: .seealso: MatMult(), MatMultAdd(), MatMultTransposeAdd()
1714: @*/
1715: PetscErrorCode MatMultTranspose(Mat mat,Vec x,Vec y)
1716: {
1725: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1726: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1727: if (x == y) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
1728: #ifndef PETSC_HAVE_CONSTRAINTS
1729: if (mat->rmap.N != x->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->rmap.N,x->map.N);
1730: if (mat->cmap.N != y->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->cmap.N,y->map.N);
1731: #endif
1732: MatPreallocated(mat);
1734: if (!mat->ops->multtranspose) SETERRQ(PETSC_ERR_SUP,"This matrix type does not have a multiply tranpose defined");
1736: (*mat->ops->multtranspose)(mat,x,y);
1738: PetscObjectStateIncrease((PetscObject)y);
1739: return(0);
1740: }
1744: /*@
1745: MatMultAdd - Computes v3 = v2 + A * v1.
1747: Collective on Mat and Vec
1749: Input Parameters:
1750: + mat - the matrix
1751: - v1, v2 - the vectors
1753: Output Parameters:
1754: . v3 - the result
1756: Notes:
1757: The vectors v1 and v3 cannot be the same. I.e., one cannot
1758: call MatMultAdd(A,v1,v2,v1).
1760: Level: beginner
1762: Concepts: matrix vector product^addition
1764: .seealso: MatMultTranspose(), MatMult(), MatMultTransposeAdd()
1765: @*/
1766: PetscErrorCode MatMultAdd(Mat mat,Vec v1,Vec v2,Vec v3)
1767: {
1777: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1778: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1779: if (mat->cmap.N != v1->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v1: global dim %D %D",mat->cmap.N,v1->map.N);
1780: if (mat->rmap.N != v2->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v2: global dim %D %D",mat->rmap.N,v2->map.N);
1781: if (mat->rmap.N != v3->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v3: global dim %D %D",mat->rmap.N,v3->map.N);
1782: if (mat->rmap.n != v3->map.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v3: local dim %D %D",mat->rmap.n,v3->map.n);
1783: if (mat->rmap.n != v2->map.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v2: local dim %D %D",mat->rmap.n,v2->map.n);
1784: if (v1 == v3) SETERRQ(PETSC_ERR_ARG_IDN,"v1 and v3 must be different vectors");
1785: MatPreallocated(mat);
1788: (*mat->ops->multadd)(mat,v1,v2,v3);
1790: PetscObjectStateIncrease((PetscObject)v3);
1791: return(0);
1792: }
1796: /*@
1797: MatMultTransposeAdd - Computes v3 = v2 + A' * v1.
1799: Collective on Mat and Vec
1801: Input Parameters:
1802: + mat - the matrix
1803: - v1, v2 - the vectors
1805: Output Parameters:
1806: . v3 - the result
1808: Notes:
1809: The vectors v1 and v3 cannot be the same. I.e., one cannot
1810: call MatMultTransposeAdd(A,v1,v2,v1).
1812: Level: beginner
1814: Concepts: matrix vector product^transpose and addition
1816: .seealso: MatMultTranspose(), MatMultAdd(), MatMult()
1817: @*/
1818: PetscErrorCode MatMultTransposeAdd(Mat mat,Vec v1,Vec v2,Vec v3)
1819: {
1829: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1830: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1831: if (!mat->ops->multtransposeadd) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1832: if (v1 == v3) SETERRQ(PETSC_ERR_ARG_IDN,"v1 and v3 must be different vectors");
1833: if (mat->rmap.N != v1->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v1: global dim %D %D",mat->rmap.N,v1->map.N);
1834: if (mat->cmap.N != v2->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v2: global dim %D %D",mat->cmap.N,v2->map.N);
1835: if (mat->cmap.N != v3->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v3: global dim %D %D",mat->cmap.N,v3->map.N);
1836: MatPreallocated(mat);
1839: (*mat->ops->multtransposeadd)(mat,v1,v2,v3);
1841: PetscObjectStateIncrease((PetscObject)v3);
1842: return(0);
1843: }
1847: /*@
1848: MatMultConstrained - The inner multiplication routine for a
1849: constrained matrix P^T A P.
1851: Collective on Mat and Vec
1853: Input Parameters:
1854: + mat - the matrix
1855: - x - the vector to be multilplied
1857: Output Parameters:
1858: . y - the result
1860: Notes:
1861: The vectors x and y cannot be the same. I.e., one cannot
1862: call MatMult(A,y,y).
1864: Level: beginner
1866: .keywords: matrix, multiply, matrix-vector product, constraint
1867: .seealso: MatMult(), MatMultTrans(), MatMultAdd(), MatMultTransAdd()
1868: @*/
1869: PetscErrorCode MatMultConstrained(Mat mat,Vec x,Vec y)
1870: {
1877: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1878: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1879: if (x == y) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
1880: if (mat->cmap.N != x->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap.N,x->map.N);
1881: if (mat->rmap.N != y->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->rmap.N,y->map.N);
1882: if (mat->rmap.n != y->map.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: local dim %D %D",mat->rmap.n,y->map.n);
1885: (*mat->ops->multconstrained)(mat,x,y);
1887: PetscObjectStateIncrease((PetscObject)y);
1889: return(0);
1890: }
1894: /*@
1895: MatMultTransposeConstrained - The inner multiplication routine for a
1896: constrained matrix P^T A^T P.
1898: Collective on Mat and Vec
1900: Input Parameters:
1901: + mat - the matrix
1902: - x - the vector to be multilplied
1904: Output Parameters:
1905: . y - the result
1907: Notes:
1908: The vectors x and y cannot be the same. I.e., one cannot
1909: call MatMult(A,y,y).
1911: Level: beginner
1913: .keywords: matrix, multiply, matrix-vector product, constraint
1914: .seealso: MatMult(), MatMultTrans(), MatMultAdd(), MatMultTransAdd()
1915: @*/
1916: PetscErrorCode MatMultTransposeConstrained(Mat mat,Vec x,Vec y)
1917: {
1924: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1925: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1926: if (x == y) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
1927: if (mat->rmap.N != x->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap.N,x->map.N);
1928: if (mat->cmap.N != y->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->rmap.N,y->map.N);
1931: (*mat->ops->multtransposeconstrained)(mat,x,y);
1933: PetscObjectStateIncrease((PetscObject)y);
1935: return(0);
1936: }
1937: /* ------------------------------------------------------------*/
1940: /*@
1941: MatGetInfo - Returns information about matrix storage (number of
1942: nonzeros, memory, etc.).
1944: Collective on Mat if MAT_GLOBAL_MAX or MAT_GLOBAL_SUM is used
1945: as the flag
1947: Input Parameters:
1948: . mat - the matrix
1950: Output Parameters:
1951: + flag - flag indicating the type of parameters to be returned
1952: (MAT_LOCAL - local matrix, MAT_GLOBAL_MAX - maximum over all processors,
1953: MAT_GLOBAL_SUM - sum over all processors)
1954: - info - matrix information context
1956: Notes:
1957: The MatInfo context contains a variety of matrix data, including
1958: number of nonzeros allocated and used, number of mallocs during
1959: matrix assembly, etc. Additional information for factored matrices
1960: is provided (such as the fill ratio, number of mallocs during
1961: factorization, etc.). Much of this info is printed to PETSC_STDOUT
1962: when using the runtime options
1963: $ -info -mat_view_info
1965: Example for C/C++ Users:
1966: See the file ${PETSC_DIR}/include/petscmat.h for a complete list of
1967: data within the MatInfo context. For example,
1968: .vb
1969: MatInfo info;
1970: Mat A;
1971: double mal, nz_a, nz_u;
1973: MatGetInfo(A,MAT_LOCAL,&info);
1974: mal = info.mallocs;
1975: nz_a = info.nz_allocated;
1976: .ve
1978: Example for Fortran Users:
1979: Fortran users should declare info as a double precision
1980: array of dimension MAT_INFO_SIZE, and then extract the parameters
1981: of interest. See the file ${PETSC_DIR}/include/finclude/petscmat.h
1982: a complete list of parameter names.
1983: .vb
1984: double precision info(MAT_INFO_SIZE)
1985: double precision mal, nz_a
1986: Mat A
1987: integer ierr
1989: call MatGetInfo(A,MAT_LOCAL,info,ierr)
1990: mal = info(MAT_INFO_MALLOCS)
1991: nz_a = info(MAT_INFO_NZ_ALLOCATED)
1992: .ve
1994: Level: intermediate
1996: Concepts: matrices^getting information on
1997:
1998: @*/
1999: PetscErrorCode MatGetInfo(Mat mat,MatInfoType flag,MatInfo *info)
2000: {
2007: if (!mat->ops->getinfo) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2008: MatPreallocated(mat);
2009: (*mat->ops->getinfo)(mat,flag,info);
2010: return(0);
2011: }
2013: /* ----------------------------------------------------------*/
2016: /*@C
2017: MatILUDTFactor - Performs a drop tolerance ILU factorization.
2019: Collective on Mat
2021: Input Parameters:
2022: + mat - the matrix
2023: . row - row permutation
2024: . col - column permutation
2025: - info - information about the factorization to be done
2027: Output Parameters:
2028: . fact - the factored matrix
2030: Level: developer
2032: Notes:
2033: Most users should employ the simplified KSP interface for linear solvers
2034: instead of working directly with matrix algebra routines such as this.
2035: See, e.g., KSPCreate().
2037: This is currently only supported for the SeqAIJ matrix format using code
2038: from Yousef Saad's SPARSEKIT2 package (translated to C with f2c) and/or
2039: Matlab. SPARSEKIT2 is copyrighted by Yousef Saad with the GNU copyright
2040: and thus can be distributed with PETSc.
2042: Concepts: matrices^ILUDT factorization
2044: .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor(), MatFactorInfo
2045: @*/
2046: PetscErrorCode MatILUDTFactor(Mat mat,IS row,IS col,MatFactorInfo *info,Mat *fact)
2047: {
2057: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2058: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2059: if (!mat->ops->iludtfactor) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2060: MatPreallocated(mat);
2062: (*mat->ops->iludtfactor)(mat,row,col,info,fact);
2064: PetscObjectStateIncrease((PetscObject)*fact);
2066: return(0);
2067: }
2071: /*@
2072: MatLUFactor - Performs in-place LU factorization of matrix.
2074: Collective on Mat
2076: Input Parameters:
2077: + mat - the matrix
2078: . row - row permutation
2079: . col - column permutation
2080: - info - options for factorization, includes
2081: $ fill - expected fill as ratio of original fill.
2082: $ dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
2083: $ Run with the option -info to determine an optimal value to use
2085: Notes:
2086: Most users should employ the simplified KSP interface for linear solvers
2087: instead of working directly with matrix algebra routines such as this.
2088: See, e.g., KSPCreate().
2090: This changes the state of the matrix to a factored matrix; it cannot be used
2091: for example with MatSetValues() unless one first calls MatSetUnfactored().
2093: Level: developer
2095: Concepts: matrices^LU factorization
2097: .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor(),
2098: MatGetOrdering(), MatSetUnfactored(), MatFactorInfo
2100: @*/
2101: PetscErrorCode MatLUFactor(Mat mat,IS row,IS col,MatFactorInfo *info)
2102: {
2111: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2112: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2113: if (!mat->ops->lufactor) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2114: MatPreallocated(mat);
2117: (*mat->ops->lufactor)(mat,row,col,info);
2119: PetscObjectStateIncrease((PetscObject)mat);
2120: return(0);
2121: }
2125: /*@
2126: MatILUFactor - Performs in-place ILU factorization of matrix.
2128: Collective on Mat
2130: Input Parameters:
2131: + mat - the matrix
2132: . row - row permutation
2133: . col - column permutation
2134: - info - structure containing
2135: $ levels - number of levels of fill.
2136: $ expected fill - as ratio of original fill.
2137: $ 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
2138: missing diagonal entries)
2140: Notes:
2141: Probably really in-place only when level of fill is zero, otherwise allocates
2142: new space to store factored matrix and deletes previous memory.
2144: Most users should employ the simplified KSP interface for linear solvers
2145: instead of working directly with matrix algebra routines such as this.
2146: See, e.g., KSPCreate().
2148: Level: developer
2150: Concepts: matrices^ILU factorization
2152: .seealso: MatILUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor(), MatFactorInfo
2153: @*/
2154: PetscErrorCode MatILUFactor(Mat mat,IS row,IS col,MatFactorInfo *info)
2155: {
2164: if (mat->rmap.N != mat->cmap.N) SETERRQ(PETSC_ERR_ARG_WRONG,"matrix must be square");
2165: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2166: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2167: if (!mat->ops->ilufactor) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2168: MatPreallocated(mat);
2171: (*mat->ops->ilufactor)(mat,row,col,info);
2173: PetscObjectStateIncrease((PetscObject)mat);
2174: return(0);
2175: }
2179: /*@
2180: MatLUFactorSymbolic - Performs symbolic LU factorization of matrix.
2181: Call this routine before calling MatLUFactorNumeric().
2183: Collective on Mat
2185: Input Parameters:
2186: + mat - the matrix
2187: . row, col - row and column permutations
2188: - info - options for factorization, includes
2189: $ fill - expected fill as ratio of original fill.
2190: $ dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
2191: $ Run with the option -info to determine an optimal value to use
2193: Output Parameter:
2194: . fact - new matrix that has been symbolically factored
2196: Notes:
2197: See the users manual for additional information about
2198: choosing the fill factor for better efficiency.
2200: Most users should employ the simplified KSP interface for linear solvers
2201: instead of working directly with matrix algebra routines such as this.
2202: See, e.g., KSPCreate().
2204: Level: developer
2206: Concepts: matrices^LU symbolic factorization
2208: .seealso: MatLUFactor(), MatLUFactorNumeric(), MatCholeskyFactor(), MatFactorInfo
2209: @*/
2210: PetscErrorCode MatLUFactorSymbolic(Mat mat,IS row,IS col,MatFactorInfo *info,Mat *fact)
2211: {
2221: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2222: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2223: if (!mat->ops->lufactorsymbolic) SETERRQ1(PETSC_ERR_SUP,"Matrix type %s symbolic LU",((PetscObject)mat)->type_name);
2224: MatPreallocated(mat);
2227: (*mat->ops->lufactorsymbolic)(mat,row,col,info,fact);
2229: PetscObjectStateIncrease((PetscObject)*fact);
2230: return(0);
2231: }
2235: /*@
2236: MatLUFactorNumeric - Performs numeric LU factorization of a matrix.
2237: Call this routine after first calling MatLUFactorSymbolic().
2239: Collective on Mat
2241: Input Parameters:
2242: + mat - the matrix
2243: . info - options for factorization
2244: - fact - the matrix generated for the factor, from MatLUFactorSymbolic()
2246: Notes:
2247: See MatLUFactor() for in-place factorization. See
2248: MatCholeskyFactorNumeric() for the symmetric, positive definite case.
2250: Most users should employ the simplified KSP interface for linear solvers
2251: instead of working directly with matrix algebra routines such as this.
2252: See, e.g., KSPCreate().
2254: Level: developer
2256: Concepts: matrices^LU numeric factorization
2258: .seealso: MatLUFactorSymbolic(), MatLUFactor(), MatCholeskyFactor()
2259: @*/
2260: PetscErrorCode MatLUFactorNumeric(Mat mat,MatFactorInfo *info,Mat *fact)
2261: {
2269: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2270: if (mat->rmap.N != (*fact)->rmap.N || mat->cmap.N != (*fact)->cmap.N) {
2271: SETERRQ4(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);
2272: }
2273: if (!(*fact)->ops->lufactornumeric) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2274: MatPreallocated(mat);
2276: (*(*fact)->ops->lufactornumeric)(mat,info,fact);
2279: MatView_Private(*fact);
2280: PetscObjectStateIncrease((PetscObject)*fact);
2281: return(0);
2282: }
2286: /*@
2287: MatCholeskyFactor - Performs in-place Cholesky factorization of a
2288: symmetric matrix.
2290: Collective on Mat
2292: Input Parameters:
2293: + mat - the matrix
2294: . perm - row and column permutations
2295: - f - expected fill as ratio of original fill
2297: Notes:
2298: See MatLUFactor() for the nonsymmetric case. See also
2299: MatCholeskyFactorSymbolic(), and MatCholeskyFactorNumeric().
2301: Most users should employ the simplified KSP interface for linear solvers
2302: instead of working directly with matrix algebra routines such as this.
2303: See, e.g., KSPCreate().
2305: Level: developer
2307: Concepts: matrices^Cholesky factorization
2309: .seealso: MatLUFactor(), MatCholeskyFactorSymbolic(), MatCholeskyFactorNumeric()
2310: MatGetOrdering()
2312: @*/
2313: PetscErrorCode MatCholeskyFactor(Mat mat,IS perm,MatFactorInfo *info)
2314: {
2322: if (mat->rmap.N != mat->cmap.N) SETERRQ(PETSC_ERR_ARG_WRONG,"Matrix must be square");
2323: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2324: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2325: if (!mat->ops->choleskyfactor) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2326: MatPreallocated(mat);
2329: (*mat->ops->choleskyfactor)(mat,perm,info);
2331: PetscObjectStateIncrease((PetscObject)mat);
2332: return(0);
2333: }
2337: /*@
2338: MatCholeskyFactorSymbolic - Performs symbolic Cholesky factorization
2339: of a symmetric matrix.
2341: Collective on Mat
2343: Input Parameters:
2344: + mat - the matrix
2345: . perm - row and column permutations
2346: - info - options for factorization, includes
2347: $ fill - expected fill as ratio of original fill.
2348: $ dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
2349: $ Run with the option -info to determine an optimal value to use
2351: Output Parameter:
2352: . fact - the factored matrix
2354: Notes:
2355: See MatLUFactorSymbolic() for the nonsymmetric case. See also
2356: MatCholeskyFactor() and MatCholeskyFactorNumeric().
2358: Most users should employ the simplified KSP interface for linear solvers
2359: instead of working directly with matrix algebra routines such as this.
2360: See, e.g., KSPCreate().
2362: Level: developer
2364: Concepts: matrices^Cholesky symbolic factorization
2366: .seealso: MatLUFactorSymbolic(), MatCholeskyFactor(), MatCholeskyFactorNumeric()
2367: MatGetOrdering()
2369: @*/
2370: PetscErrorCode MatCholeskyFactorSymbolic(Mat mat,IS perm,MatFactorInfo *info,Mat *fact)
2371: {
2380: if (mat->rmap.N != mat->cmap.N) SETERRQ(PETSC_ERR_ARG_WRONG,"Matrix must be square");
2381: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2382: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2383: if (!mat->ops->choleskyfactorsymbolic) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2384: MatPreallocated(mat);
2387: (*mat->ops->choleskyfactorsymbolic)(mat,perm,info,fact);
2389: PetscObjectStateIncrease((PetscObject)*fact);
2390: return(0);
2391: }
2395: /*@
2396: MatCholeskyFactorNumeric - Performs numeric Cholesky factorization
2397: of a symmetric matrix. Call this routine after first calling
2398: MatCholeskyFactorSymbolic().
2400: Collective on Mat
2402: Input Parameter:
2403: . mat - the initial matrix
2404: . info - options for factorization
2405: - fact - the symbolic factor of mat
2407: Output Parameter:
2408: . fact - the factored matrix
2410: Notes:
2411: Most users should employ the simplified KSP interface for linear solvers
2412: instead of working directly with matrix algebra routines such as this.
2413: See, e.g., KSPCreate().
2415: Level: developer
2417: Concepts: matrices^Cholesky numeric factorization
2419: .seealso: MatCholeskyFactorSymbolic(), MatCholeskyFactor(), MatLUFactorNumeric()
2420: @*/
2421: PetscErrorCode MatCholeskyFactorNumeric(Mat mat,MatFactorInfo *info,Mat *fact)
2422: {
2430: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2431: if (!(*fact)->ops->choleskyfactornumeric) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2432: if (mat->rmap.N != (*fact)->rmap.N || mat->cmap.N != (*fact)->cmap.N) {
2433: SETERRQ4(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);
2434: }
2435: MatPreallocated(mat);
2438: (*(*fact)->ops->choleskyfactornumeric)(mat,info,fact);
2441: MatView_Private(*fact);
2442: PetscObjectStateIncrease((PetscObject)*fact);
2443: return(0);
2444: }
2446: /* ----------------------------------------------------------------*/
2449: /*@
2450: MatSolve - Solves A x = b, given a factored matrix.
2452: Collective on Mat and Vec
2454: Input Parameters:
2455: + mat - the factored matrix
2456: - b - the right-hand-side vector
2458: Output Parameter:
2459: . x - the result vector
2461: Notes:
2462: The vectors b and x cannot be the same. I.e., one cannot
2463: call MatSolve(A,x,x).
2465: Notes:
2466: Most users should employ the simplified KSP interface for linear solvers
2467: instead of working directly with matrix algebra routines such as this.
2468: See, e.g., KSPCreate().
2470: Level: developer
2472: Concepts: matrices^triangular solves
2474: .seealso: MatSolveAdd(), MatSolveTranspose(), MatSolveTransposeAdd()
2475: @*/
2476: PetscErrorCode MatSolve(Mat mat,Vec b,Vec x)
2477: {
2487: if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
2488: if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2489: if (mat->cmap.N != x->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap.N,x->map.N);
2490: if (mat->rmap.N != b->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap.N,b->map.N);
2491: if (mat->rmap.n != b->map.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap.n,b->map.n);
2492: if (!mat->rmap.N && !mat->cmap.N) return(0);
2493: if (!mat->ops->solve) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2494: MatPreallocated(mat);
2497: (*mat->ops->solve)(mat,b,x);
2499: PetscObjectStateIncrease((PetscObject)x);
2500: return(0);
2501: }
2505: /*@
2506: MatMatSolve - Solves A X = B, given a factored matrix.
2508: Collective on Mat
2510: Input Parameters:
2511: + mat - the factored matrix
2512: - b - the right-hand-side matrix (dense matrix)
2514: Output Parameter:
2515: . x - the result matrix (dense matrix)
2517: Notes:
2518: The matrices b and x cannot be the same. I.e., one cannot
2519: call MatMatSolve(A,x,x).
2521: Notes:
2522: Most users should usually employ the simplified KSP interface for linear solvers
2523: instead of working directly with matrix algebra routines such as this.
2524: See, e.g., KSPCreate(). However KSP can only solve for one vector (column of X)
2525: at a time.
2527: Level: developer
2529: Concepts: matrices^triangular solves
2531: .seealso: MatMatSolveAdd(), MatMatSolveTranspose(), MatMatSolveTransposeAdd()
2532: @*/
2533: PetscErrorCode MatMatSolve(Mat A,Mat B,Mat X)
2534: {
2544: if (X == B) SETERRQ(PETSC_ERR_ARG_IDN,"X and B must be different matrices");
2545: if (!A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2546: if (A->cmap.N != X->rmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat A,Mat X: global dim %D %D",A->cmap.N,X->rmap.N);
2547: if (A->rmap.N != B->rmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat A,Mat B: global dim %D %D",A->rmap.N,B->rmap.N);
2548: if (A->rmap.n != B->rmap.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat A,Mat B: local dim %D %D",A->rmap.n,B->rmap.n);
2549: if (!A->rmap.N && !A->cmap.N) return(0);
2550: if (!A->ops->matsolve) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)A)->type_name);
2551: MatPreallocated(A);
2554: (*A->ops->matsolve)(A,B,X);
2556: PetscObjectStateIncrease((PetscObject)X);
2557: return(0);
2558: }
2563: /* @
2564: MatForwardSolve - Solves L x = b, given a factored matrix, A = LU, or
2565: U^T*D^(1/2) x = b, given a factored symmetric matrix, A = U^T*D*U,
2567: Collective on Mat and Vec
2569: Input Parameters:
2570: + mat - the factored matrix
2571: - b - the right-hand-side vector
2573: Output Parameter:
2574: . x - the result vector
2576: Notes:
2577: MatSolve() should be used for most applications, as it performs
2578: a forward solve followed by a backward solve.
2580: The vectors b and x cannot be the same, i.e., one cannot
2581: call MatForwardSolve(A,x,x).
2583: For matrix in seqsbaij format with block size larger than 1,
2584: the diagonal blocks are not implemented as D = D^(1/2) * D^(1/2) yet.
2585: MatForwardSolve() solves U^T*D y = b, and
2586: MatBackwardSolve() solves U x = y.
2587: Thus they do not provide a symmetric preconditioner.
2589: Most users should employ the simplified KSP interface for linear solvers
2590: instead of working directly with matrix algebra routines such as this.
2591: See, e.g., KSPCreate().
2593: Level: developer
2595: Concepts: matrices^forward solves
2597: .seealso: MatSolve(), MatBackwardSolve()
2598: @ */
2599: PetscErrorCode MatForwardSolve(Mat mat,Vec b,Vec x)
2600: {
2610: if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
2611: if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2612: if (!mat->ops->forwardsolve) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2613: if (mat->cmap.N != x->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap.N,x->map.N);
2614: if (mat->rmap.N != b->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap.N,b->map.N);
2615: if (mat->rmap.n != b->map.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap.n,b->map.n);
2616: MatPreallocated(mat);
2618: (*mat->ops->forwardsolve)(mat,b,x);
2620: PetscObjectStateIncrease((PetscObject)x);
2621: return(0);
2622: }
2626: /* @
2627: MatBackwardSolve - Solves U x = b, given a factored matrix, A = LU.
2628: D^(1/2) U x = b, given a factored symmetric matrix, A = U^T*D*U,
2630: Collective on Mat and Vec
2632: Input Parameters:
2633: + mat - the factored matrix
2634: - b - the right-hand-side vector
2636: Output Parameter:
2637: . x - the result vector
2639: Notes:
2640: MatSolve() should be used for most applications, as it performs
2641: a forward solve followed by a backward solve.
2643: The vectors b and x cannot be the same. I.e., one cannot
2644: call MatBackwardSolve(A,x,x).
2646: For matrix in seqsbaij format with block size larger than 1,
2647: the diagonal blocks are not implemented as D = D^(1/2) * D^(1/2) yet.
2648: MatForwardSolve() solves U^T*D y = b, and
2649: MatBackwardSolve() solves U x = y.
2650: Thus they do not provide a symmetric preconditioner.
2652: Most users should employ the simplified KSP interface for linear solvers
2653: instead of working directly with matrix algebra routines such as this.
2654: See, e.g., KSPCreate().
2656: Level: developer
2658: Concepts: matrices^backward solves
2660: .seealso: MatSolve(), MatForwardSolve()
2661: @ */
2662: PetscErrorCode MatBackwardSolve(Mat mat,Vec b,Vec x)
2663: {
2673: if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
2674: if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2675: if (!mat->ops->backwardsolve) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2676: if (mat->cmap.N != x->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap.N,x->map.N);
2677: if (mat->rmap.N != b->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap.N,b->map.N);
2678: if (mat->rmap.n != b->map.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap.n,b->map.n);
2679: MatPreallocated(mat);
2682: (*mat->ops->backwardsolve)(mat,b,x);
2684: PetscObjectStateIncrease((PetscObject)x);
2685: return(0);
2686: }
2690: /*@
2691: MatSolveAdd - Computes x = y + inv(A)*b, given a factored matrix.
2693: Collective on Mat and Vec
2695: Input Parameters:
2696: + mat - the factored matrix
2697: . b - the right-hand-side vector
2698: - y - the vector to be added to
2700: Output Parameter:
2701: . x - the result vector
2703: Notes:
2704: The vectors b and x cannot be the same. I.e., one cannot
2705: call MatSolveAdd(A,x,y,x).
2707: Most users should employ the simplified KSP interface for linear solvers
2708: instead of working directly with matrix algebra routines such as this.
2709: See, e.g., KSPCreate().
2711: Level: developer
2713: Concepts: matrices^triangular solves
2715: .seealso: MatSolve(), MatSolveTranspose(), MatSolveTransposeAdd()
2716: @*/
2717: PetscErrorCode MatSolveAdd(Mat mat,Vec b,Vec y,Vec x)
2718: {
2719: PetscScalar one = 1.0;
2720: Vec tmp;
2732: if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
2733: if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2734: if (mat->cmap.N != x->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap.N,x->map.N);
2735: if (mat->rmap.N != b->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap.N,b->map.N);
2736: if (mat->rmap.N != y->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->rmap.N,y->map.N);
2737: if (mat->rmap.n != b->map.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap.n,b->map.n);
2738: if (x->map.n != y->map.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Vec x,Vec y: local dim %D %D",x->map.n,y->map.n);
2739: MatPreallocated(mat);
2742: if (mat->ops->solveadd) {
2743: (*mat->ops->solveadd)(mat,b,y,x);
2744: } else {
2745: /* do the solve then the add manually */
2746: if (x != y) {
2747: MatSolve(mat,b,x);
2748: VecAXPY(x,one,y);
2749: } else {
2750: VecDuplicate(x,&tmp);
2751: PetscLogObjectParent(mat,tmp);
2752: VecCopy(x,tmp);
2753: MatSolve(mat,b,x);
2754: VecAXPY(x,one,tmp);
2755: VecDestroy(tmp);
2756: }
2757: }
2759: PetscObjectStateIncrease((PetscObject)x);
2760: return(0);
2761: }
2765: /*@
2766: MatSolveTranspose - Solves A' x = b, given a factored matrix.
2768: Collective on Mat and Vec
2770: Input Parameters:
2771: + mat - the factored matrix
2772: - b - the right-hand-side vector
2774: Output Parameter:
2775: . x - the result vector
2777: Notes:
2778: The vectors b and x cannot be the same. I.e., one cannot
2779: call MatSolveTranspose(A,x,x).
2781: Most users should employ the simplified KSP interface for linear solvers
2782: instead of working directly with matrix algebra routines such as this.
2783: See, e.g., KSPCreate().
2785: Level: developer
2787: Concepts: matrices^triangular solves
2789: .seealso: MatSolve(), MatSolveAdd(), MatSolveTransposeAdd()
2790: @*/
2791: PetscErrorCode MatSolveTranspose(Mat mat,Vec b,Vec x)
2792: {
2802: if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2803: if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
2804: if (!mat->ops->solvetranspose) SETERRQ1(PETSC_ERR_SUP,"Matrix type %s",((PetscObject)mat)->type_name);
2805: if (mat->rmap.N != x->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->rmap.N,x->map.N);
2806: if (mat->cmap.N != b->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->cmap.N,b->map.N);
2807: MatPreallocated(mat);
2809: (*mat->ops->solvetranspose)(mat,b,x);
2811: PetscObjectStateIncrease((PetscObject)x);
2812: return(0);
2813: }
2817: /*@
2818: MatSolveTransposeAdd - Computes x = y + inv(Transpose(A)) b, given a
2819: factored matrix.
2821: Collective on Mat and Vec
2823: Input Parameters:
2824: + mat - the factored matrix
2825: . b - the right-hand-side vector
2826: - y - the vector to be added to
2828: Output Parameter:
2829: . x - the result vector
2831: Notes:
2832: The vectors b and x cannot be the same. I.e., one cannot
2833: call MatSolveTransposeAdd(A,x,y,x).
2835: Most users should employ the simplified KSP interface for linear solvers
2836: instead of working directly with matrix algebra routines such as this.
2837: See, e.g., KSPCreate().
2839: Level: developer
2841: Concepts: matrices^triangular solves
2843: .seealso: MatSolve(), MatSolveAdd(), MatSolveTranspose()
2844: @*/
2845: PetscErrorCode MatSolveTransposeAdd(Mat mat,Vec b,Vec y,Vec x)
2846: {
2847: PetscScalar one = 1.0;
2849: Vec tmp;
2860: if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
2861: if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2862: if (mat->rmap.N != x->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->rmap.N,x->map.N);
2863: if (mat->cmap.N != b->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->cmap.N,b->map.N);
2864: if (mat->cmap.N != y->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->cmap.N,y->map.N);
2865: if (x->map.n != y->map.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Vec x,Vec y: local dim %D %D",x->map.n,y->map.n);
2866: MatPreallocated(mat);
2869: if (mat->ops->solvetransposeadd) {
2870: (*mat->ops->solvetransposeadd)(mat,b,y,x);
2871: } else {
2872: /* do the solve then the add manually */
2873: if (x != y) {
2874: MatSolveTranspose(mat,b,x);
2875: VecAXPY(x,one,y);
2876: } else {
2877: VecDuplicate(x,&tmp);
2878: PetscLogObjectParent(mat,tmp);
2879: VecCopy(x,tmp);
2880: MatSolveTranspose(mat,b,x);
2881: VecAXPY(x,one,tmp);
2882: VecDestroy(tmp);
2883: }
2884: }
2886: PetscObjectStateIncrease((PetscObject)x);
2887: return(0);
2888: }
2889: /* ----------------------------------------------------------------*/
2893: /*@
2894: MatRelax - Computes relaxation (SOR, Gauss-Seidel) sweeps.
2896: Collective on Mat and Vec
2898: Input Parameters:
2899: + mat - the matrix
2900: . b - the right hand side
2901: . omega - the relaxation factor
2902: . flag - flag indicating the type of SOR (see below)
2903: . shift - diagonal shift
2904: - its - the number of iterations
2905: - lits - the number of local iterations
2907: Output Parameters:
2908: . x - the solution (can contain an initial guess)
2910: SOR Flags:
2911: . SOR_FORWARD_SWEEP - forward SOR
2912: . SOR_BACKWARD_SWEEP - backward SOR
2913: . SOR_SYMMETRIC_SWEEP - SSOR (symmetric SOR)
2914: . SOR_LOCAL_FORWARD_SWEEP - local forward SOR
2915: . SOR_LOCAL_BACKWARD_SWEEP - local forward SOR
2916: . SOR_LOCAL_SYMMETRIC_SWEEP - local SSOR
2917: . SOR_APPLY_UPPER, SOR_APPLY_LOWER - applies
2918: upper/lower triangular part of matrix to
2919: vector (with omega)
2920: . SOR_ZERO_INITIAL_GUESS - zero initial guess
2922: Notes:
2923: SOR_LOCAL_FORWARD_SWEEP, SOR_LOCAL_BACKWARD_SWEEP, and
2924: SOR_LOCAL_SYMMETRIC_SWEEP perform separate independent smoothings
2925: on each processor.
2927: Application programmers will not generally use MatRelax() directly,
2928: but instead will employ the KSP/PC interface.
2930: Notes for Advanced Users:
2931: The flags are implemented as bitwise inclusive or operations.
2932: For example, use (SOR_ZERO_INITIAL_GUESS | SOR_SYMMETRIC_SWEEP)
2933: to specify a zero initial guess for SSOR.
2935: Most users should employ the simplified KSP interface for linear solvers
2936: instead of working directly with matrix algebra routines such as this.
2937: See, e.g., KSPCreate().
2939: See also, MatPBRelax(). This routine will automatically call the point block
2940: version if the point version is not available.
2942: Level: developer
2944: Concepts: matrices^relaxation
2945: Concepts: matrices^SOR
2946: Concepts: matrices^Gauss-Seidel
2948: @*/
2949: PetscErrorCode MatRelax(Mat mat,Vec b,PetscReal omega,MatSORType flag,PetscReal shift,PetscInt its,PetscInt lits,Vec x)
2950: {
2960: if (!mat->ops->relax && !mat->ops->pbrelax) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2961: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2962: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2963: if (mat->cmap.N != x->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap.N,x->map.N);
2964: if (mat->rmap.N != b->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap.N,b->map.N);
2965: if (mat->rmap.n != b->map.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap.n,b->map.n);
2966: if (its <= 0) SETERRQ1(PETSC_ERR_ARG_WRONG,"Relaxation requires global its %D positive",its);
2967: if (lits <= 0) SETERRQ1(PETSC_ERR_ARG_WRONG,"Relaxation requires local its %D positive",lits);
2969: MatPreallocated(mat);
2971: if (mat->ops->relax) {
2972: ierr =(*mat->ops->relax)(mat,b,omega,flag,shift,its,lits,x);
2973: } else {
2974: ierr =(*mat->ops->pbrelax)(mat,b,omega,flag,shift,its,lits,x);
2975: }
2977: PetscObjectStateIncrease((PetscObject)x);
2978: return(0);
2979: }
2983: /*@
2984: MatPBRelax - Computes relaxation (SOR, Gauss-Seidel) sweeps.
2986: Collective on Mat and Vec
2988: See MatRelax() for usage
2990: For multi-component PDEs where the Jacobian is stored in a point block format
2991: (with the PETSc BAIJ matrix formats) the relaxation is done one point block at
2992: a time. That is, the small (for example, 4 by 4) blocks along the diagonal are solved
2993: simultaneously (that is a 4 by 4 linear solve is done) to update all the values at a point.
2995: Level: developer
2997: @*/
2998: PetscErrorCode MatPBRelax(Mat mat,Vec b,PetscReal omega,MatSORType flag,PetscReal shift,PetscInt its,PetscInt lits,Vec x)
2999: {
3009: if (!mat->ops->pbrelax) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3010: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3011: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3012: if (mat->cmap.N != x->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap.N,x->map.N);
3013: if (mat->rmap.N != b->map.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap.N,b->map.N);
3014: if (mat->rmap.n != b->map.n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap.n,b->map.n);
3015: MatPreallocated(mat);
3018: ierr =(*mat->ops->pbrelax)(mat,b,omega,flag,shift,its,lits,x);
3020: PetscObjectStateIncrease((PetscObject)x);
3021: return(0);
3022: }
3026: /*
3027: Default matrix copy routine.
3028: */
3029: PetscErrorCode MatCopy_Basic(Mat A,Mat B,MatStructure str)
3030: {
3031: PetscErrorCode ierr;
3032: PetscInt i,rstart,rend,nz;
3033: const PetscInt *cwork;
3034: const PetscScalar *vwork;
3037: if (B->assembled) {
3038: MatZeroEntries(B);
3039: }
3040: MatGetOwnershipRange(A,&rstart,&rend);
3041: for (i=rstart; i<rend; i++) {
3042: MatGetRow(A,i,&nz,&cwork,&vwork);
3043: MatSetValues(B,1,&i,nz,cwork,vwork,INSERT_VALUES);
3044: MatRestoreRow(A,i,&nz,&cwork,&vwork);
3045: }
3046: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
3047: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
3048: PetscObjectStateIncrease((PetscObject)B);
3049: return(0);
3050: }
3054: /*@
3055: MatCopy - Copys a matrix to another matrix.
3057: Collective on Mat
3059: Input Parameters:
3060: + A - the matrix
3061: - str - SAME_NONZERO_PATTERN or DIFFERENT_NONZERO_PATTERN
3063: Output Parameter:
3064: . B - where the copy is put
3066: Notes:
3067: If you use SAME_NONZERO_PATTERN then the two matrices had better have the
3068: same nonzero pattern or the routine will crash.
3070: MatCopy() copies the matrix entries of a matrix to another existing
3071: matrix (after first zeroing the second matrix). A related routine is
3072: MatConvert(), which first creates a new matrix and then copies the data.
3074: Level: intermediate
3075:
3076: Concepts: matrices^copying
3078: .seealso: MatConvert(), MatDuplicate()
3080: @*/
3081: PetscErrorCode MatCopy(Mat A,Mat B,MatStructure str)
3082: {
3090: MatPreallocated(B);
3092: if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3093: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3094: if (A->rmap.N != B->rmap.N || A->cmap.N != B->cmap.N) SETERRQ4(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);
3095: MatPreallocated(A);
3098: if (A->ops->copy) {
3099: (*A->ops->copy)(A,B,str);
3100: } else { /* generic conversion */
3101: MatCopy_Basic(A,B,str);
3102: }
3103: if (A->mapping) {
3104: if (B->mapping) {ISLocalToGlobalMappingDestroy(B->mapping);B->mapping = 0;}
3105: MatSetLocalToGlobalMapping(B,A->mapping);
3106: }
3107: if (A->bmapping) {
3108: if (B->bmapping) {ISLocalToGlobalMappingDestroy(B->bmapping);B->bmapping = 0;}
3109: MatSetLocalToGlobalMappingBlock(B,A->mapping);
3110: }
3112: PetscObjectStateIncrease((PetscObject)B);
3113: return(0);
3114: }
3118: /*@C
3119: MatConvert - Converts a matrix to another matrix, either of the same
3120: or different type.
3122: Collective on Mat
3124: Input Parameters:
3125: + mat - the matrix
3126: . newtype - new matrix type. Use MATSAME to create a new matrix of the
3127: same type as the original matrix.
3128: - reuse - denotes if the destination matrix is to be created or reused. Currently
3129: MAT_REUSE_MATRIX is only supported for inplace conversion, otherwise use
3130: MAT_INITIAL_MATRIX.
3132: Output Parameter:
3133: . M - pointer to place new matrix
3135: Notes:
3136: MatConvert() first creates a new matrix and then copies the data from
3137: the first matrix. A related routine is MatCopy(), which copies the matrix
3138: entries of one matrix to another already existing matrix context.
3140: Level: intermediate
3142: Concepts: matrices^converting between storage formats
3144: .seealso: MatCopy(), MatDuplicate()
3145: @*/
3146: PetscErrorCode MatConvert(Mat mat, MatType newtype,MatReuse reuse,Mat *M)
3147: {
3148: PetscErrorCode ierr;
3149: PetscTruth sametype,issame,flg;
3150: char convname[256],mtype[256];
3151: Mat B;
3157: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3158: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3159: MatPreallocated(mat);
3161: PetscOptionsGetString(((PetscObject)mat)->prefix,"-matconvert_type",mtype,256,&flg);
3162: if (flg) {
3163: newtype = mtype;
3164: }
3165: PetscTypeCompare((PetscObject)mat,newtype,&sametype);
3166: PetscStrcmp(newtype,"same",&issame);
3167: if ((reuse==MAT_REUSE_MATRIX) && (mat != *M)) {
3168: SETERRQ(PETSC_ERR_SUP,"MAT_REUSE_MATRIX only supported for in-place conversion currently");
3169: }
3170: if ((sametype || issame) && (reuse==MAT_INITIAL_MATRIX) && mat->ops->duplicate) {
3171: (*mat->ops->duplicate)(mat,MAT_COPY_VALUES,M);
3172: } else {
3173: PetscErrorCode (*conv)(Mat, MatType,MatReuse,Mat*)=PETSC_NULL;
3174: const char *prefix[3] = {"seq","mpi",""};
3175: PetscInt i;
3176: /*
3177: Order of precedence:
3178: 1) See if a specialized converter is known to the current matrix.
3179: 2) See if a specialized converter is known to the desired matrix class.
3180: 3) See if a good general converter is registered for the desired class
3181: (as of 6/27/03 only MATMPIADJ falls into this category).
3182: 4) See if a good general converter is known for the current matrix.
3183: 5) Use a really basic converter.
3184: */
3185:
3186: /* 1) See if a specialized converter is known to the current matrix and the desired class */
3187: for (i=0; i<3; i++) {
3188: PetscStrcpy(convname,"MatConvert_");
3189: PetscStrcat(convname,((PetscObject)mat)->type_name);
3190: PetscStrcat(convname,"_");
3191: PetscStrcat(convname,prefix[i]);
3192: PetscStrcat(convname,newtype);
3193: PetscStrcat(convname,"_C");
3194: PetscObjectQueryFunction((PetscObject)mat,convname,(void (**)(void))&conv);
3195: if (conv) goto foundconv;
3196: }
3198: /* 2) See if a specialized converter is known to the desired matrix class. */
3199: MatCreate(((PetscObject)mat)->comm,&B);
3200: MatSetSizes(B,mat->rmap.n,mat->cmap.n,mat->rmap.N,mat->cmap.N);
3201: MatSetType(B,newtype);
3202: for (i=0; i<3; i++) {
3203: PetscStrcpy(convname,"MatConvert_");
3204: PetscStrcat(convname,((PetscObject)mat)->type_name);
3205: PetscStrcat(convname,"_");
3206: PetscStrcat(convname,prefix[i]);
3207: PetscStrcat(convname,newtype);
3208: PetscStrcat(convname,"_C");
3209: PetscObjectQueryFunction((PetscObject)B,convname,(void (**)(void))&conv);
3210: if (conv) {
3211: MatDestroy(B);
3212: goto foundconv;
3213: }
3214: }
3216: /* 3) See if a good general converter is registered for the desired class */
3217: conv = B->ops->convertfrom;
3218: MatDestroy(B);
3219: if (conv) goto foundconv;
3221: /* 4) See if a good general converter is known for the current matrix */
3222: if (mat->ops->convert) {
3223: conv = mat->ops->convert;
3224: }
3225: if (conv) goto foundconv;
3227: /* 5) Use a really basic converter. */
3228: conv = MatConvert_Basic;
3230: foundconv:
3232: (*conv)(mat,newtype,reuse,M);
3234: }
3235: B = *M;
3236: PetscObjectStateIncrease((PetscObject)B);
3237: return(0);
3238: }
3243: /*@
3244: MatDuplicate - Duplicates a matrix including the non-zero structure.
3246: Collective on Mat
3248: Input Parameters:
3249: + mat - the matrix
3250: - op - either MAT_DO_NOT_COPY_VALUES or MAT_COPY_VALUES, cause it to copy nonzero
3251: values as well or not
3253: Output Parameter:
3254: . M - pointer to place new matrix
3256: Level: intermediate
3258: Concepts: matrices^duplicating
3260: .seealso: MatCopy(), MatConvert()
3261: @*/
3262: PetscErrorCode MatDuplicate(Mat mat,MatDuplicateOption op,Mat *M)
3263: {
3265: Mat B;
3271: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3272: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3273: MatPreallocated(mat);
3275: *M = 0;
3276: if (!mat->ops->duplicate) {
3277: SETERRQ(PETSC_ERR_SUP,"Not written for this matrix type");
3278: }
3280: (*mat->ops->duplicate)(mat,op,M);
3281: B = *M;
3282: if (mat->mapping) {
3283: MatSetLocalToGlobalMapping(B,mat->mapping);
3284: }
3285: if (mat->bmapping) {
3286: MatSetLocalToGlobalMappingBlock(B,mat->bmapping);
3287: }
3288: PetscMapCopy(((PetscObject)mat)->comm,&mat->rmap,&B->rmap);
3289: PetscMapCopy(((PetscObject)mat)->comm,&mat->cmap,&B->cmap);
3290:
3292: PetscObjectStateIncrease((PetscObject)B);
3293: return(0);
3294: }
3298: /*@
3299: MatGetDiagonal - Gets the diagonal of a matrix.
3301: Collective on Mat and Vec
3303: Input Parameters:
3304: + mat - the matrix
3305: - v - the vector for storing the diagonal
3307: Output Parameter:
3308: . v - the diagonal of the matrix
3310: Notes: The result of this call are the same as if one converted the matrix to dense format
3311: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
3313: Level: intermediate
3315: Concepts: matrices^accessing diagonals
3317: .seealso: MatGetRow(), MatGetSubMatrices(), MatGetSubmatrix(), MatGetRowMaxAbs()
3318: @*/
3319: PetscErrorCode MatGetDiagonal(Mat mat,Vec v)
3320: {
3327: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3328: if (!mat->ops->getdiagonal) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3329: MatPreallocated(mat);
3331: (*mat->ops->getdiagonal)(mat,v);
3332: PetscObjectStateIncrease((PetscObject)v);
3333: return(0);
3334: }
3338: /*@
3339: MatGetRowMin - Gets the minimum value (of the real part) of each
3340: row of the matrix
3342: Collective on Mat and Vec
3344: Input Parameters:
3345: . mat - the matrix
3347: Output Parameter:
3348: + v - the vector for storing the maximums
3349: - idx - the indices of the column found for each row (optional)
3351: Level: intermediate
3353: Notes: The result of this call are the same as if one converted the matrix to dense format
3354: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
3356: This code is only implemented for a couple of matrix formats.
3358: Concepts: matrices^getting row maximums
3360: .seealso: MatGetDiagonal(), MatGetSubMatrices(), MatGetSubmatrix(), MatGetRowMaxAbs(),
3361: MatGetRowMax()
3362: @*/
3363: PetscErrorCode MatGetRowMin(Mat mat,Vec v,PetscInt idx[])
3364: {
3371: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3372: if (!mat->ops->getrowmax) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3373: MatPreallocated(mat);
3375: (*mat->ops->getrowmin)(mat,v,idx);
3376: PetscObjectStateIncrease((PetscObject)v);
3377: return(0);
3378: }
3382: /*@
3383: MatGetRowMax - Gets the maximum value (of the real part) of each
3384: row of the matrix
3386: Collective on Mat and Vec
3388: Input Parameters:
3389: . mat - the matrix
3391: Output Parameter:
3392: + v - the vector for storing the maximums
3393: - idx - the indices of the column found for each row (optional)
3395: Level: intermediate
3397: Notes: The result of this call are the same as if one converted the matrix to dense format
3398: and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
3400: This code is only implemented for a couple of matrix formats.
3402: Concepts: matrices^getting row maximums
3404: .seealso: MatGetDiagonal(), MatGetSubMatrices(), MatGetSubmatrix(), MatGetRowMaxAbs(), MatGetRowMin()
3405: @*/
3406: PetscErrorCode MatGetRowMax(Mat mat,Vec v,PetscInt idx[])
3407: {
3414: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3415: if (!mat->ops->getrowmax) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3416: MatPreallocated(mat);
3418: (*mat->ops->getrowmax)(mat,v,idx);
3419: PetscObjectStateIncrease((PetscObject)v);
3420: return(0);
3421: }
3425: /*@
3426: MatGetRowMaxAbs - Gets the maximum value (in absolute value) of each
3427: row of the matrix
3429: Collective on Mat and Vec
3431: Input Parameters:
3432: . mat - the matrix
3434: Output Parameter:
3435: + v - the vector for storing the maximums
3436: - idx - the indices of the column found for each row (optional)
3438: Level: intermediate
3440: Notes: if a row is completely empty or has only 0.0 values then the idx[] value for that
3441: row is 0 (the first column).
3443: This code is only implemented for a couple of matrix formats.
3445: Concepts: matrices^getting row maximums
3447: .seealso: MatGetDiagonal(), MatGetSubMatrices(), MatGetSubmatrix(), MatGetRowMax(), MatGetRowMin()
3448: @*/
3449: PetscErrorCode MatGetRowMaxAbs(Mat mat,Vec v,PetscInt idx[])
3450: {
3457: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3458: if (!mat->ops->getrowmaxabs) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3459: MatPreallocated(mat);
3461: (*mat->ops->getrowmaxabs)(mat,v,idx);
3462: PetscObjectStateIncrease((PetscObject)v);
3463: return(0);
3464: }
3468: /*@
3469: MatGetRowSum - Gets the sum of each row of the matrix
3471: Collective on Mat and Vec
3473: Input Parameters:
3474: . mat - the matrix
3476: Output Parameter:
3477: . v - the vector for storing the maximums
3479: Level: intermediate
3481: Notes: This code is slow since it is not currently specialized for different formats
3483: Concepts: matrices^getting row sums
3485: .seealso: MatGetDiagonal(), MatGetSubMatrices(), MatGetSubmatrix(), MatGetRowMax(), MatGetRowMin()
3486: @*/
3487: PetscErrorCode MatGetRowSum(Mat mat, Vec v)
3488: {
3489: PetscInt start, end, row;
3490: PetscScalar *array;
3497: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3498: MatPreallocated(mat);
3499: MatGetOwnershipRange(mat, &start, &end);
3500: VecGetArray(v, &array);
3501: for(row = start; row < end; ++row) {
3502: PetscInt ncols, col;
3503: const PetscInt *cols;
3504: const PetscScalar *vals;
3506: array[row - start] = 0.0;
3507: MatGetRow(mat, row, &ncols, &cols, &vals);
3508: for(col = 0; col < ncols; col++) {
3509: array[row - start] += vals[col];
3510: }
3511: MatRestoreRow(mat, row, &ncols, &cols, &vals);
3512: }
3513: VecRestoreArray(v, &array);
3514: PetscObjectStateIncrease((PetscObject) v);
3515: return(0);
3516: }
3520: /*@
3521: MatTranspose - Computes an in-place or out-of-place transpose of a matrix.
3523: Collective on Mat
3525: Input Parameter:
3526: . mat - the matrix to transpose
3528: Output Parameters:
3529: . B - the transpose
3531: Notes:
3532: If you pass in PETSC_NULL for B an in-place transpose in mat will be done
3534: Level: intermediate
3536: Concepts: matrices^transposing
3538: .seealso: MatMultTranspose(), MatMultTransposeAdd(), MatIsTranspose()
3539: @*/
3540: PetscErrorCode MatTranspose(Mat mat,Mat *B)
3541: {
3547: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3548: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3549: if (!mat->ops->transpose) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3550: MatPreallocated(mat);
3553: (*mat->ops->transpose)(mat,B);
3555: if (B) {PetscObjectStateIncrease((PetscObject)*B);}
3556: return(0);
3557: }
3561: /*@
3562: MatIsTranspose - Test whether a matrix is another one's transpose,
3563: or its own, in which case it tests symmetry.
3565: Collective on Mat
3567: Input Parameter:
3568: + A - the matrix to test
3569: - B - the matrix to test against, this can equal the first parameter
3571: Output Parameters:
3572: . flg - the result
3574: Notes:
3575: Only available for SeqAIJ/MPIAIJ matrices. The sequential algorithm
3576: has a running time of the order of the number of nonzeros; the parallel
3577: test involves parallel copies of the block-offdiagonal parts of the matrix.
3579: Level: intermediate
3581: Concepts: matrices^transposing, matrix^symmetry
3583: .seealso: MatTranspose(), MatIsSymmetric(), MatIsHermitian()
3584: @*/
3585: PetscErrorCode MatIsTranspose(Mat A,Mat B,PetscReal tol,PetscTruth *flg)
3586: {
3587: PetscErrorCode ierr,(*f)(Mat,Mat,PetscReal,PetscTruth*),(*g)(Mat,Mat,PetscReal,PetscTruth*);
3593: PetscObjectQueryFunction((PetscObject)A,"MatIsTranspose_C",(void (**)(void))&f);
3594: PetscObjectQueryFunction((PetscObject)B,"MatIsTranspose_C",(void (**)(void))&g);
3595: if (f && g) {
3596: if (f==g) {
3597: (*f)(A,B,tol,flg);
3598: } else {
3599: SETERRQ(PETSC_ERR_ARG_NOTSAMETYPE,"Matrices do not have the same comparator for symmetry test");
3600: }
3601: }
3602: return(0);
3603: }
3607: /*@
3608: MatIsHermitianTranspose - Test whether a matrix is another one's Hermitian transpose,
3610: Collective on Mat
3612: Input Parameter:
3613: + A - the matrix to test
3614: - B - the matrix to test against, this can equal the first parameter
3616: Output Parameters:
3617: . flg - the result
3619: Notes:
3620: Only available for SeqAIJ/MPIAIJ matrices. The sequential algorithm
3621: has a running time of the order of the number of nonzeros; the parallel
3622: test involves parallel copies of the block-offdiagonal parts of the matrix.
3624: Level: intermediate
3626: Concepts: matrices^transposing, matrix^symmetry
3628: .seealso: MatTranspose(), MatIsSymmetric(), MatIsHermitian(), MatIsTranspose()
3629: @*/
3630: PetscErrorCode MatIsHermitianTranspose(Mat A,Mat B,PetscReal tol,PetscTruth *flg)
3631: {
3632: PetscErrorCode ierr,(*f)(Mat,Mat,PetscReal,PetscTruth*),(*g)(Mat,Mat,PetscReal,PetscTruth*);
3638: PetscObjectQueryFunction((PetscObject)A,"MatIsHermitianTranspose_C",(void (**)(void))&f);
3639: PetscObjectQueryFunction((PetscObject)B,"MatIsHermitianTranspose_C",(void (**)(void))&g);
3640: if (f && g) {
3641: if (f==g) {
3642: (*f)(A,B,tol,flg);
3643: } else {
3644: SETERRQ(PETSC_ERR_ARG_NOTSAMETYPE,"Matrices do not have the same comparator for Hermitian test");
3645: }
3646: }
3647: return(0);
3648: }
3652: /*@
3653: MatPermute - Creates a new matrix with rows and columns permuted from the
3654: original.
3656: Collective on Mat
3658: Input Parameters:
3659: + mat - the matrix to permute
3660: . row - row permutation, each processor supplies only the permutation for its rows
3661: - col - column permutation, each processor needs the entire column permutation, that is
3662: this is the same size as the total number of columns in the matrix
3664: Output Parameters:
3665: . B - the permuted matrix
3667: Level: advanced
3669: Concepts: matrices^permuting
3671: .seealso: MatGetOrdering()
3672: @*/
3673: PetscErrorCode MatPermute(Mat mat,IS row,IS col,Mat *B)
3674: {
3683: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3684: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3685: if (!mat->ops->permute) SETERRQ1(PETSC_ERR_SUP,"MatPermute not available for Mat type %s",((PetscObject)mat)->type_name);
3686: MatPreallocated(mat);
3688: (*mat->ops->permute)(mat,row,col,B);
3689: PetscObjectStateIncrease((PetscObject)*B);
3690: return(0);
3691: }
3695: /*@
3696: MatPermuteSparsify - Creates a new matrix with rows and columns permuted from the
3697: original and sparsified to the prescribed tolerance.
3699: Collective on Mat
3701: Input Parameters:
3702: + A - The matrix to permute
3703: . band - The half-bandwidth of the sparsified matrix, or PETSC_DECIDE
3704: . frac - The half-bandwidth as a fraction of the total size, or 0.0
3705: . tol - The drop tolerance
3706: . rowp - The row permutation
3707: - colp - The column permutation
3709: Output Parameter:
3710: . B - The permuted, sparsified matrix
3712: Level: advanced
3714: Note:
3715: The default behavior (band = PETSC_DECIDE and frac = 0.0) is to
3716: restrict the half-bandwidth of the resulting matrix to 5% of the
3717: total matrix size.
3719: .keywords: matrix, permute, sparsify
3721: .seealso: MatGetOrdering(), MatPermute()
3722: @*/
3723: PetscErrorCode MatPermuteSparsify(Mat A, PetscInt band, PetscReal frac, PetscReal tol, IS rowp, IS colp, Mat *B)
3724: {
3725: IS irowp, icolp;
3726: PetscInt *rows, *cols;
3727: PetscInt M, N, locRowStart, locRowEnd;
3728: PetscInt nz, newNz;
3729: const PetscInt *cwork;
3730: PetscInt *cnew;
3731: const PetscScalar *vwork;
3732: PetscScalar *vnew;
3733: PetscInt bw, issize;
3734: PetscInt row, locRow, newRow, col, newCol;
3735: PetscErrorCode ierr;
3742: if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3743: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3744: if (!A->ops->permutesparsify) {
3745: MatGetSize(A, &M, &N);
3746: MatGetOwnershipRange(A, &locRowStart, &locRowEnd);
3747: ISGetSize(rowp, &issize);
3748: if (issize != M) SETERRQ2(PETSC_ERR_ARG_WRONG, "Wrong size %D for row permutation, should be %D", issize, M);
3749: ISGetSize(colp, &issize);
3750: if (issize != N) SETERRQ2(PETSC_ERR_ARG_WRONG, "Wrong size %D for column permutation, should be %D", issize, N);
3751: ISInvertPermutation(rowp, 0, &irowp);
3752: ISGetIndices(irowp, &rows);
3753: ISInvertPermutation(colp, 0, &icolp);
3754: ISGetIndices(icolp, &cols);
3755: PetscMalloc(N * sizeof(PetscInt), &cnew);
3756: PetscMalloc(N * sizeof(PetscScalar), &vnew);
3758: /* Setup bandwidth to include */
3759: if (band == PETSC_DECIDE) {
3760: if (frac <= 0.0)
3761: bw = (PetscInt) (M * 0.05);
3762: else
3763: bw = (PetscInt) (M * frac);
3764: } else {
3765: if (band <= 0) SETERRQ(PETSC_ERR_ARG_WRONG, "Bandwidth must be a positive integer");
3766: bw = band;
3767: }
3769: /* Put values into new matrix */
3770: MatDuplicate(A, MAT_DO_NOT_COPY_VALUES, B);
3771: for(row = locRowStart, locRow = 0; row < locRowEnd; row++, locRow++) {
3772: MatGetRow(A, row, &nz, &cwork, &vwork);
3773: newRow = rows[locRow]+locRowStart;
3774: for(col = 0, newNz = 0; col < nz; col++) {
3775: newCol = cols[cwork[col]];
3776: if ((newCol >= newRow - bw) && (newCol < newRow + bw) && (PetscAbsScalar(vwork[col]) >= tol)) {
3777: cnew[newNz] = newCol;
3778: vnew[newNz] = vwork[col];
3779: newNz++;
3780: }
3781: }
3782: MatSetValues(*B, 1, &newRow, newNz, cnew, vnew, INSERT_VALUES);
3783: MatRestoreRow(A, row, &nz, &cwork, &vwork);
3784: }
3785: PetscFree(cnew);
3786: PetscFree(vnew);
3787: MatAssemblyBegin(*B, MAT_FINAL_ASSEMBLY);
3788: MatAssemblyEnd(*B, MAT_FINAL_ASSEMBLY);
3789: ISRestoreIndices(irowp, &rows);
3790: ISRestoreIndices(icolp, &cols);
3791: ISDestroy(irowp);
3792: ISDestroy(icolp);
3793: } else {
3794: (*A->ops->permutesparsify)(A, band, frac, tol, rowp, colp, B);
3795: }
3796: PetscObjectStateIncrease((PetscObject)*B);
3797: return(0);
3798: }
3802: /*@
3803: MatEqual - Compares two matrices.
3805: Collective on Mat
3807: Input Parameters:
3808: + A - the first matrix
3809: - B - the second matrix
3811: Output Parameter:
3812: . flg - PETSC_TRUE if the matrices are equal; PETSC_FALSE otherwise.
3814: Level: intermediate
3816: Concepts: matrices^equality between
3817: @*/
3818: PetscErrorCode MatEqual(Mat A,Mat B,PetscTruth *flg)
3819: {
3827: MatPreallocated(B);
3830: if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3831: if (!B->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3832: if (A->rmap.N != B->rmap.N || A->cmap.N != B->cmap.N) SETERRQ4(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);
3833: if (!A->ops->equal) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)A)->type_name);
3834: if (!B->ops->equal) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)B)->type_name);
3835: if (A->ops->equal != B->ops->equal) SETERRQ2(PETSC_ERR_ARG_INCOMP,"A is type: %s\nB is type: %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
3836: MatPreallocated(A);
3838: (*A->ops->equal)(A,B,flg);
3839: return(0);
3840: }
3844: /*@
3845: MatDiagonalScale - Scales a matrix on the left and right by diagonal
3846: matrices that are stored as vectors. Either of the two scaling
3847: matrices can be PETSC_NULL.
3849: Collective on Mat
3851: Input Parameters:
3852: + mat - the matrix to be scaled
3853: . l - the left scaling vector (or PETSC_NULL)
3854: - r - the right scaling vector (or PETSC_NULL)
3856: Notes:
3857: MatDiagonalScale() computes A = LAR, where
3858: L = a diagonal matrix (stored as a vector), R = a diagonal matrix (stored as a vector)
3860: Level: intermediate
3862: Concepts: matrices^diagonal scaling
3863: Concepts: diagonal scaling of matrices
3865: .seealso: MatScale()
3866: @*/
3867: PetscErrorCode MatDiagonalScale(Mat mat,Vec l,Vec r)
3868: {
3874: if (!mat->ops->diagonalscale) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3877: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3878: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3879: MatPreallocated(mat);
3882: (*mat->ops->diagonalscale)(mat,l,r);
3884: PetscObjectStateIncrease((PetscObject)mat);
3885: return(0);
3886: }
3890: /*@
3891: MatScale - Scales all elements of a matrix by a given number.
3893: Collective on Mat
3895: Input Parameters:
3896: + mat - the matrix to be scaled
3897: - a - the scaling value
3899: Output Parameter:
3900: . mat - the scaled matrix
3902: Level: intermediate
3904: Concepts: matrices^scaling all entries
3906: .seealso: MatDiagonalScale()
3907: @*/
3908: PetscErrorCode MatScale(Mat mat,PetscScalar a)
3909: {
3915: if (!mat->ops->scale) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3916: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3917: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3918: MatPreallocated(mat);
3921: (*mat->ops->scale)(mat,a);
3923: PetscObjectStateIncrease((PetscObject)mat);
3924: return(0);
3925: }
3929: /*@
3930: MatNorm - Calculates various norms of a matrix.
3932: Collective on Mat
3934: Input Parameters:
3935: + mat - the matrix
3936: - type - the type of norm, NORM_1, NORM_FROBENIUS, NORM_INFINITY
3938: Output Parameters:
3939: . nrm - the resulting norm
3941: Level: intermediate
3943: Concepts: matrices^norm
3944: Concepts: norm^of matrix
3945: @*/
3946: PetscErrorCode MatNorm(Mat mat,NormType type,PetscReal *nrm)
3947: {
3955: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3956: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3957: if (!mat->ops->norm) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3958: MatPreallocated(mat);
3960: (*mat->ops->norm)(mat,type,nrm);
3961: return(0);
3962: }
3964: /*
3965: This variable is used to prevent counting of MatAssemblyBegin() that
3966: are called from within a MatAssemblyEnd().
3967: */
3968: static PetscInt MatAssemblyEnd_InUse = 0;
3971: /*@
3972: MatAssemblyBegin - Begins assembling the matrix. This routine should
3973: be called after completing all calls to MatSetValues().
3975: Collective on Mat
3977: Input Parameters:
3978: + mat - the matrix
3979: - type - type of assembly, either MAT_FLUSH_ASSEMBLY or MAT_FINAL_ASSEMBLY
3980:
3981: Notes:
3982: MatSetValues() generally caches the values. The matrix is ready to
3983: use only after MatAssemblyBegin() and MatAssemblyEnd() have been called.
3984: Use MAT_FLUSH_ASSEMBLY when switching between ADD_VALUES and INSERT_VALUES
3985: in MatSetValues(); use MAT_FINAL_ASSEMBLY for the final assembly before
3986: using the matrix.
3988: Level: beginner
3990: Concepts: matrices^assembling
3992: .seealso: MatAssemblyEnd(), MatSetValues(), MatAssembled()
3993: @*/
3994: PetscErrorCode MatAssemblyBegin(Mat mat,MatAssemblyType type)
3995: {
4001: MatPreallocated(mat);
4002: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix.\nDid you forget to call MatSetUnfactored()?");
4003: if (mat->assembled) {
4004: mat->was_assembled = PETSC_TRUE;
4005: mat->assembled = PETSC_FALSE;
4006: }
4007: if (!MatAssemblyEnd_InUse) {
4009: if (mat->ops->assemblybegin){(*mat->ops->assemblybegin)(mat,type);}
4011: } else {
4012: if (mat->ops->assemblybegin){(*mat->ops->assemblybegin)(mat,type);}
4013: }
4014: return(0);
4015: }
4019: /*@
4020: MatAssembled - Indicates if a matrix has been assembled and is ready for
4021: use; for example, in matrix-vector product.
4023: Collective on Mat
4025: Input Parameter:
4026: . mat - the matrix
4028: Output Parameter:
4029: . assembled - PETSC_TRUE or PETSC_FALSE
4031: Level: advanced
4033: Concepts: matrices^assembled?
4035: .seealso: MatAssemblyEnd(), MatSetValues(), MatAssemblyBegin()
4036: @*/
4037: PetscErrorCode MatAssembled(Mat mat,PetscTruth *assembled)
4038: {
4043: *assembled = mat->assembled;
4044: return(0);
4045: }
4049: /*
4050: Processes command line options to determine if/how a matrix
4051: is to be viewed. Called by MatAssemblyEnd() and MatLoad().
4052: */
4053: PetscErrorCode MatView_Private(Mat mat)
4054: {
4055: PetscErrorCode ierr;
4056: PetscTruth flg1,flg2,flg3,flg4,flg6,flg7,flg8;
4057: static PetscTruth incall = PETSC_FALSE;
4058: #if defined(PETSC_USE_SOCKET_VIEWER)
4059: PetscTruth flg5;
4060: #endif
4063: if (incall) return(0);
4064: incall = PETSC_TRUE;
4065: PetscOptionsBegin(((PetscObject)mat)->comm,((PetscObject)mat)->prefix,"Matrix Options","Mat");
4066: PetscOptionsName("-mat_view_info","Information on matrix size","MatView",&flg1);
4067: PetscOptionsName("-mat_view_info_detailed","Nonzeros in the matrix","MatView",&flg2);
4068: PetscOptionsName("-mat_view","Print matrix to stdout","MatView",&flg3);
4069: PetscOptionsName("-mat_view_matlab","Print matrix to stdout in a format Matlab can read","MatView",&flg4);
4070: #if defined(PETSC_USE_SOCKET_VIEWER)
4071: PetscOptionsName("-mat_view_socket","Send matrix to socket (can be read from matlab)","MatView",&flg5);
4072: #endif
4073: PetscOptionsName("-mat_view_binary","Save matrix to file in binary format","MatView",&flg6);
4074: PetscOptionsName("-mat_view_draw","Draw the matrix nonzero structure","MatView",&flg7);
4075: PetscOptionsEnd();
4077: if (flg1) {
4078: PetscViewer viewer;
4080: PetscViewerASCIIGetStdout(((PetscObject)mat)->comm,&viewer);
4081: PetscViewerPushFormat(viewer,PETSC_VIEWER_ASCII_INFO);
4082: MatView(mat,viewer);
4083: PetscViewerPopFormat(viewer);
4084: }
4085: if (flg2) {
4086: PetscViewer viewer;
4088: PetscViewerASCIIGetStdout(((PetscObject)mat)->comm,&viewer);
4089: PetscViewerPushFormat(viewer,PETSC_VIEWER_ASCII_INFO_DETAIL);
4090: MatView(mat,viewer);
4091: PetscViewerPopFormat(viewer);
4092: }
4093: if (flg3) {
4094: PetscViewer viewer;
4096: PetscViewerASCIIGetStdout(((PetscObject)mat)->comm,&viewer);
4097: MatView(mat,viewer);
4098: }
4099: if (flg4) {
4100: PetscViewer viewer;
4102: PetscViewerASCIIGetStdout(((PetscObject)mat)->comm,&viewer);
4103: PetscViewerPushFormat(viewer,PETSC_VIEWER_ASCII_MATLAB);
4104: MatView(mat,viewer);
4105: PetscViewerPopFormat(viewer);
4106: }
4107: #if defined(PETSC_USE_SOCKET_VIEWER)
4108: if (flg5) {
4109: MatView(mat,PETSC_VIEWER_SOCKET_(((PetscObject)mat)->comm));
4110: PetscViewerFlush(PETSC_VIEWER_SOCKET_(((PetscObject)mat)->comm));
4111: }
4112: #endif
4113: if (flg6) {
4114: MatView(mat,PETSC_VIEWER_BINARY_(((PetscObject)mat)->comm));
4115: PetscViewerFlush(PETSC_VIEWER_BINARY_(((PetscObject)mat)->comm));
4116: }
4117: if (flg7) {
4118: PetscOptionsHasName(((PetscObject)mat)->prefix,"-mat_view_contour",&flg8);
4119: if (flg8) {
4120: PetscViewerPushFormat(PETSC_VIEWER_DRAW_(((PetscObject)mat)->comm),PETSC_VIEWER_DRAW_CONTOUR);
4121: }
4122: MatView(mat,PETSC_VIEWER_DRAW_(((PetscObject)mat)->comm));
4123: PetscViewerFlush(PETSC_VIEWER_DRAW_(((PetscObject)mat)->comm));
4124: if (flg8) {
4125: PetscViewerPopFormat(PETSC_VIEWER_DRAW_(((PetscObject)mat)->comm));
4126: }
4127: }
4128: incall = PETSC_FALSE;
4129: return(0);
4130: }
4134: /*@
4135: MatAssemblyEnd - Completes assembling the matrix. This routine should
4136: be called after MatAssemblyBegin().
4138: Collective on Mat
4140: Input Parameters:
4141: + mat - the matrix
4142: - type - type of assembly, either MAT_FLUSH_ASSEMBLY or MAT_FINAL_ASSEMBLY
4144: Options Database Keys:
4145: + -mat_view_info - Prints info on matrix at conclusion of MatEndAssembly()
4146: . -mat_view_info_detailed - Prints more detailed info
4147: . -mat_view - Prints matrix in ASCII format
4148: . -mat_view_matlab - Prints matrix in Matlab format
4149: . -mat_view_draw - PetscDraws nonzero structure of matrix, using MatView() and PetscDrawOpenX().
4150: . -display <name> - Sets display name (default is host)
4151: . -draw_pause <sec> - Sets number of seconds to pause after display
4152: . -mat_view_socket - Sends matrix to socket, can be accessed from Matlab (see users manual)
4153: . -viewer_socket_machine <machine>
4154: . -viewer_socket_port <port>
4155: . -mat_view_binary - save matrix to file in binary format
4156: - -viewer_binary_filename <name>
4158: Notes:
4159: MatSetValues() generally caches the values. The matrix is ready to
4160: use only after MatAssemblyBegin() and MatAssemblyEnd() have been called.
4161: Use MAT_FLUSH_ASSEMBLY when switching between ADD_VALUES and INSERT_VALUES
4162: in MatSetValues(); use MAT_FINAL_ASSEMBLY for the final assembly before
4163: using the matrix.
4165: Level: beginner
4167: .seealso: MatAssemblyBegin(), MatSetValues(), PetscDrawOpenX(), MatView(), MatAssembled(), PetscViewerSocketOpen()
4168: @*/
4169: PetscErrorCode MatAssemblyEnd(Mat mat,MatAssemblyType type)
4170: {
4171: PetscErrorCode ierr;
4172: static PetscInt inassm = 0;
4173: PetscTruth flg;
4179: inassm++;
4180: MatAssemblyEnd_InUse++;
4181: if (MatAssemblyEnd_InUse == 1) { /* Do the logging only the first time through */
4183: if (mat->ops->assemblyend) {
4184: (*mat->ops->assemblyend)(mat,type);
4185: }
4187: } else {
4188: if (mat->ops->assemblyend) {
4189: (*mat->ops->assemblyend)(mat,type);
4190: }
4191: }
4193: /* Flush assembly is not a true assembly */
4194: if (type != MAT_FLUSH_ASSEMBLY) {
4195: mat->assembled = PETSC_TRUE; mat->num_ass++;
4196: }
4197: mat->insertmode = NOT_SET_VALUES;
4198: MatAssemblyEnd_InUse--;
4199: PetscObjectStateIncrease((PetscObject)mat);
4200: if (!mat->symmetric_eternal) {
4201: mat->symmetric_set = PETSC_FALSE;
4202: mat->hermitian_set = PETSC_FALSE;
4203: mat->structurally_symmetric_set = PETSC_FALSE;
4204: }
4205: if (inassm == 1 && type != MAT_FLUSH_ASSEMBLY) {
4206: MatView_Private(mat);
4207: PetscOptionsHasName(((PetscObject)mat)->prefix,"-mat_is_symmetric",&flg);
4208: if (flg) {
4209: PetscReal tol = 0.0;
4210: PetscOptionsGetReal(((PetscObject)mat)->prefix,"-mat_is_symmetric",&tol,PETSC_NULL);
4211: MatIsSymmetric(mat,tol,&flg);
4212: if (flg) {
4213: PetscPrintf(((PetscObject)mat)->comm,"Matrix is symmetric (tolerance %G)\n",tol);
4214: } else {
4215: PetscPrintf(((PetscObject)mat)->comm,"Matrix is not symmetric (tolerance %G)\n",tol);
4216: }
4217: }
4218: }
4219: inassm--;
4220: return(0);
4221: }
4226: /*@
4227: MatCompress - Tries to store the matrix in as little space as
4228: possible. May fail if memory is already fully used, since it
4229: tries to allocate new space.
4231: Collective on Mat
4233: Input Parameters:
4234: . mat - the matrix
4236: Level: advanced
4238: @*/
4239: PetscErrorCode MatCompress(Mat mat)
4240: {
4246: MatPreallocated(mat);
4247: if (mat->ops->compress) {(*mat->ops->compress)(mat);}
4248: return(0);
4249: }
4253: /*@
4254: MatSetOption - Sets a parameter option for a matrix. Some options
4255: may be specific to certain storage formats. Some options
4256: determine how values will be inserted (or added). Sorted,
4257: row-oriented input will generally assemble the fastest. The default
4258: is row-oriented, nonsorted input.
4260: Collective on Mat
4262: Input Parameters:
4263: + mat - the matrix
4264: - option - the option, one of those listed below (and possibly others),
4266: Options Describing Matrix Structure:
4267: + MAT_SYMMETRIC - symmetric in terms of both structure and value
4268: . MAT_HERMITIAN - transpose is the complex conjugation
4269: . MAT_STRUCTURALLY_SYMMETRIC - symmetric nonzero structure
4270: - MAT_SYMMETRY_ETERNAL - if you would like the symmetry/Hermitian flag
4271: you set to be kept with all future use of the matrix
4272: including after MatAssemblyBegin/End() which could
4273: potentially change the symmetry structure, i.e. you
4274: KNOW the matrix will ALWAYS have the property you set.
4277: Options For Use with MatSetValues():
4278: Insert a logically dense subblock, which can be
4279: . MAT_ROW_ORIENTED - row-oriented (default)
4281: Not these options reflect the data you pass in with MatSetValues(); it has
4282: nothing to do with how the data is stored internally in the matrix
4283: data structure.
4285: When (re)assembling a matrix, we can restrict the input for
4286: efficiency/debugging purposes. These options include
4287: + MAT_NEW_NONZERO_LOCATIONS - additional insertions will be
4288: allowed if they generate a new nonzero
4289: . MAT_NEW_DIAGONALS - new diagonals will be allowed (for block diagonal format only)
4290: . MAT_IGNORE_OFF_PROC_ENTRIES - drops off-processor entries
4291: . MAT_NEW_NONZERO_LOCATION_ERR - generates an error for new matrix entry
4292: - MAT_USE_HASH_TABLE - uses a hash table to speed up matrix assembly
4294: Notes:
4295: Some options are relevant only for particular matrix types and
4296: are thus ignored by others. Other options are not supported by
4297: certain matrix types and will generate an error message if set.
4299: If using a Fortran 77 module to compute a matrix, one may need to
4300: use the column-oriented option (or convert to the row-oriented
4301: format).
4303: MAT_NEW_NONZERO_LOCATIONS set to PETSC_FALSE indicates that any add or insertion
4304: that would generate a new entry in the nonzero structure is instead
4305: ignored. Thus, if memory has not alredy been allocated for this particular
4306: data, then the insertion is ignored. For dense matrices, in which
4307: the entire array is allocated, no entries are ever ignored.
4308: Set after the first MatAssemblyEnd()
4310: MAT_NEW_NONZERO_LOCATION_ERR indicates that any add or insertion
4311: that would generate a new entry in the nonzero structure instead produces
4312: an error. (Currently supported for AIJ and BAIJ formats only.)
4313: This is a useful flag when using SAME_NONZERO_PATTERN in calling
4314: KSPSetOperators() to ensure that the nonzero pattern truely does
4315: remain unchanged. Set after the first MatAssemblyEnd()
4317: MAT_NEW_NONZERO_ALLOCATION_ERR indicates that any add or insertion
4318: that would generate a new entry that has not been preallocated will
4319: instead produce an error. (Currently supported for AIJ and BAIJ formats
4320: only.) This is a useful flag when debugging matrix memory preallocation.
4322: MAT_IGNORE_OFF_PROC_ENTRIES indicates entries destined for
4323: other processors should be dropped, rather than stashed.
4324: This is useful if you know that the "owning" processor is also
4325: always generating the correct matrix entries, so that PETSc need
4326: not transfer duplicate entries generated on another processor.
4327:
4328: MAT_USE_HASH_TABLE indicates that a hash table be used to improve the
4329: searches during matrix assembly. When this flag is set, the hash table
4330: is created during the first Matrix Assembly. This hash table is
4331: used the next time through, during MatSetVaules()/MatSetVaulesBlocked()
4332: to improve the searching of indices. MAT_NEW_NONZERO_LOCATIONS flag
4333: should be used with MAT_USE_HASH_TABLE flag. This option is currently
4334: supported by MATMPIBAIJ format only.
4336: MAT_KEEP_ZEROED_ROWS indicates when MatZeroRows() is called the zeroed entries
4337: are kept in the nonzero structure
4339: MAT_IGNORE_ZERO_ENTRIES - for AIJ/IS matrices this will stop zero values from creating
4340: a zero location in the matrix
4342: MAT_USE_INODES - indicates using inode version of the code - works with AIJ and
4343: ROWBS matrix types
4345: Level: intermediate
4347: Concepts: matrices^setting options
4349: @*/
4350: PetscErrorCode MatSetOption(Mat mat,MatOption op,PetscTruth flg)
4351: {
4357: MatPreallocated(mat);
4358: switch (op) {
4359: case MAT_SYMMETRIC:
4360: mat->symmetric = flg;
4361: if (flg) mat->structurally_symmetric = PETSC_TRUE;
4362: mat->symmetric_set = PETSC_TRUE;
4363: mat->structurally_symmetric_set = flg;
4364: break;
4365: case MAT_HERMITIAN:
4366: mat->hermitian = flg;
4367: if (flg) mat->structurally_symmetric = PETSC_TRUE;
4368: mat->hermitian_set = PETSC_TRUE;
4369: mat->structurally_symmetric_set = flg;
4370: break;
4371: case MAT_STRUCTURALLY_SYMMETRIC:
4372: mat->structurally_symmetric = flg;
4373: mat->structurally_symmetric_set = PETSC_TRUE;
4374: break;
4375: case MAT_SYMMETRY_ETERNAL:
4376: mat->symmetric_eternal = flg;
4377: break;
4378: default:
4379: break;
4380: }
4381: if (mat->ops->setoption) {
4382: (*mat->ops->setoption)(mat,op,flg);
4383: }
4384: return(0);
4385: }
4389: /*@
4390: MatZeroEntries - Zeros all entries of a matrix. For sparse matrices
4391: this routine retains the old nonzero structure.
4393: Collective on Mat
4395: Input Parameters:
4396: . mat - the matrix
4398: Level: intermediate
4400: Concepts: matrices^zeroing
4402: .seealso: MatZeroRows()
4403: @*/
4404: PetscErrorCode MatZeroEntries(Mat mat)
4405: {
4411: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4412: if (mat->insertmode != NOT_SET_VALUES) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for matrices where you have set values but not yet assembled");
4413: if (!mat->ops->zeroentries) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4414: MatPreallocated(mat);
4417: (*mat->ops->zeroentries)(mat);
4419: PetscObjectStateIncrease((PetscObject)mat);
4420: return(0);
4421: }
4425: /*@C
4426: MatZeroRows - Zeros all entries (except possibly the main diagonal)
4427: of a set of rows of a matrix.
4429: Collective on Mat
4431: Input Parameters:
4432: + mat - the matrix
4433: . numRows - the number of rows to remove
4434: . rows - the global row indices
4435: - diag - value put in all diagonals of eliminated rows
4437: Notes:
4438: For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
4439: but does not release memory. For the dense and block diagonal
4440: formats this does not alter the nonzero structure.
4442: If the option MatSetOption(mat,MAT_KEEP_ZEROED_ROWS,PETSC_TRUE) the nonzero structure
4443: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
4444: merely zeroed.
4446: The user can set a value in the diagonal entry (or for the AIJ and
4447: row formats can optionally remove the main diagonal entry from the
4448: nonzero structure as well, by passing 0.0 as the final argument).
4450: For the parallel case, all processes that share the matrix (i.e.,
4451: those in the communicator used for matrix creation) MUST call this
4452: routine, regardless of whether any rows being zeroed are owned by
4453: them.
4455: Each processor should list the rows that IT wants zeroed
4457: Level: intermediate
4459: Concepts: matrices^zeroing rows
4461: .seealso: MatZeroRowsIS(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption()
4462: @*/
4463: PetscErrorCode MatZeroRows(Mat mat,PetscInt numRows,const PetscInt rows[],PetscScalar diag)
4464: {
4471: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4472: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4473: if (!mat->ops->zerorows) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4474: MatPreallocated(mat);
4476: (*mat->ops->zerorows)(mat,numRows,rows,diag);
4477: MatView_Private(mat);
4478: PetscObjectStateIncrease((PetscObject)mat);
4479: return(0);
4480: }
4484: /*@C
4485: MatZeroRowsIS - Zeros all entries (except possibly the main diagonal)
4486: of a set of rows of a matrix.
4488: Collective on Mat
4490: Input Parameters:
4491: + mat - the matrix
4492: . is - index set of rows to remove
4493: - diag - value put in all diagonals of eliminated rows
4495: Notes:
4496: For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
4497: but does not release memory. For the dense and block diagonal
4498: formats this does not alter the nonzero structure.
4500: If the option MatSetOption(mat,MAT_KEEP_ZEROED_ROWS,PETSC_TRUE) the nonzero structure
4501: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
4502: merely zeroed.
4504: The user can set a value in the diagonal entry (or for the AIJ and
4505: row formats can optionally remove the main diagonal entry from the
4506: nonzero structure as well, by passing 0.0 as the final argument).
4508: For the parallel case, all processes that share the matrix (i.e.,
4509: those in the communicator used for matrix creation) MUST call this
4510: routine, regardless of whether any rows being zeroed are owned by
4511: them.
4513: Each processor should list the rows that IT wants zeroed
4515: Level: intermediate
4517: Concepts: matrices^zeroing rows
4519: .seealso: MatZeroRows(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption()
4520: @*/
4521: PetscErrorCode MatZeroRowsIS(Mat mat,IS is,PetscScalar diag)
4522: {
4523: PetscInt numRows;
4524: PetscInt *rows;
4531: ISGetLocalSize(is,&numRows);
4532: ISGetIndices(is,&rows);
4533: MatZeroRows(mat,numRows,rows,diag);
4534: ISRestoreIndices(is,&rows);
4535: return(0);
4536: }
4540: /*@C
4541: MatZeroRowsLocal - Zeros all entries (except possibly the main diagonal)
4542: of a set of rows of a matrix; using local numbering of rows.
4544: Collective on Mat
4546: Input Parameters:
4547: + mat - the matrix
4548: . numRows - the number of rows to remove
4549: . rows - the global row indices
4550: - diag - value put in all diagonals of eliminated rows
4552: Notes:
4553: Before calling MatZeroRowsLocal(), the user must first set the
4554: local-to-global mapping by calling MatSetLocalToGlobalMapping().
4556: For the AIJ matrix formats this removes the old nonzero structure,
4557: but does not release memory. For the dense and block diagonal
4558: formats this does not alter the nonzero structure.
4560: If the option MatSetOption(mat,MAT_KEEP_ZEROED_ROWS,PETSC_TRUE) the nonzero structure
4561: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
4562: merely zeroed.
4564: The user can set a value in the diagonal entry (or for the AIJ and
4565: row formats can optionally remove the main diagonal entry from the
4566: nonzero structure as well, by passing 0.0 as the final argument).
4568: Level: intermediate
4570: Concepts: matrices^zeroing
4572: .seealso: MatZeroRows(), MatZeroRowsLocalIS(), MatZeroEntries(), MatZeroRows(), MatSetLocalToGlobalMapping
4573: @*/
4574: PetscErrorCode MatZeroRowsLocal(Mat mat,PetscInt numRows,const PetscInt rows[],PetscScalar diag)
4575: {
4582: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4583: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4584: MatPreallocated(mat);
4586: if (mat->ops->zerorowslocal) {
4587: (*mat->ops->zerorowslocal)(mat,numRows,rows,diag);
4588: } else {
4589: IS is, newis;
4590: PetscInt *newRows;
4592: if (!mat->mapping) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Need to provide local to global mapping to matrix first");
4593: ISCreateGeneral(PETSC_COMM_SELF,numRows,rows,&is);
4594: ISLocalToGlobalMappingApplyIS(mat->mapping,is,&newis);
4595: ISGetIndices(newis,&newRows);
4596: (*mat->ops->zerorows)(mat,numRows,newRows,diag);
4597: ISRestoreIndices(newis,&newRows);
4598: ISDestroy(newis);
4599: ISDestroy(is);
4600: }
4601: PetscObjectStateIncrease((PetscObject)mat);
4602: return(0);
4603: }
4607: /*@C
4608: MatZeroRowsLocalIS - Zeros all entries (except possibly the main diagonal)
4609: of a set of rows of a matrix; using local numbering of rows.
4611: Collective on Mat
4613: Input Parameters:
4614: + mat - the matrix
4615: . is - index set of rows to remove
4616: - diag - value put in all diagonals of eliminated rows
4618: Notes:
4619: Before calling MatZeroRowsLocalIS(), the user must first set the
4620: local-to-global mapping by calling MatSetLocalToGlobalMapping().
4622: For the AIJ matrix formats this removes the old nonzero structure,
4623: but does not release memory. For the dense and block diagonal
4624: formats this does not alter the nonzero structure.
4626: If the option MatSetOption(mat,MAT_KEEP_ZEROED_ROWS,PETSC_TRUE) the nonzero structure
4627: of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
4628: merely zeroed.
4630: The user can set a value in the diagonal entry (or for the AIJ and
4631: row formats can optionally remove the main diagonal entry from the
4632: nonzero structure as well, by passing 0.0 as the final argument).
4634: Level: intermediate
4636: Concepts: matrices^zeroing
4638: .seealso: MatZeroRows(), MatZeroRowsLocal(), MatZeroEntries(), MatZeroRows(), MatSetLocalToGlobalMapping
4639: @*/
4640: PetscErrorCode MatZeroRowsLocalIS(Mat mat,IS is,PetscScalar diag)
4641: {
4643: PetscInt numRows;
4644: PetscInt *rows;
4650: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4651: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4652: MatPreallocated(mat);
4654: ISGetLocalSize(is,&numRows);
4655: ISGetIndices(is,&rows);
4656: MatZeroRowsLocal(mat,numRows,rows,diag);
4657: ISRestoreIndices(is,&rows);
4658: return(0);
4659: }
4663: /*@
4664: MatGetSize - Returns the numbers of rows and columns in a matrix.
4666: Not Collective
4668: Input Parameter:
4669: . mat - the matrix
4671: Output Parameters:
4672: + m - the number of global rows
4673: - n - the number of global columns
4675: Note: both output parameters can be PETSC_NULL on input.
4677: Level: beginner
4679: Concepts: matrices^size
4681: .seealso: MatGetLocalSize()
4682: @*/
4683: PetscErrorCode MatGetSize(Mat mat,PetscInt *m,PetscInt* n)
4684: {
4687: if (m) *m = mat->rmap.N;
4688: if (n) *n = mat->cmap.N;
4689: return(0);
4690: }
4694: /*@
4695: MatGetLocalSize - Returns the number of rows and columns in a matrix
4696: stored locally. This information may be implementation dependent, so
4697: use with care.
4699: Not Collective
4701: Input Parameters:
4702: . mat - the matrix
4704: Output Parameters:
4705: + m - the number of local rows
4706: - n - the number of local columns
4708: Note: both output parameters can be PETSC_NULL on input.
4710: Level: beginner
4712: Concepts: matrices^local size
4714: .seealso: MatGetSize()
4715: @*/
4716: PetscErrorCode MatGetLocalSize(Mat mat,PetscInt *m,PetscInt* n)
4717: {
4722: if (m) *m = mat->rmap.n;
4723: if (n) *n = mat->cmap.n;
4724: return(0);
4725: }
4729: /*@
4730: MatGetOwnershipRangeColumn - Returns the range of matrix columns owned by
4731: this processor.
4733: Not Collective
4735: Input Parameters:
4736: . mat - the matrix
4738: Output Parameters:
4739: + m - the global index of the first local column
4740: - n - one more than the global index of the last local column
4742: Notes: both output parameters can be PETSC_NULL on input.
4744: Level: developer
4746: Concepts: matrices^column ownership
4747: @*/
4748: PetscErrorCode MatGetOwnershipRangeColumn(Mat mat,PetscInt *m,PetscInt* n)
4749: {
4757: MatPreallocated(mat);
4758: if (m) *m = mat->cmap.rstart;
4759: if (n) *n = mat->cmap.rend;
4760: return(0);
4761: }
4765: /*@
4766: MatGetOwnershipRange - Returns the range of matrix rows owned by
4767: this processor, assuming that the matrix is laid out with the first
4768: n1 rows on the first processor, the next n2 rows on the second, etc.
4769: For certain parallel layouts this range may not be well defined.
4771: Not Collective
4773: Input Parameters:
4774: . mat - the matrix
4776: Output Parameters:
4777: + m - the global index of the first local row
4778: - n - one more than the global index of the last local row
4780: Note: both output parameters can be PETSC_NULL on input.
4782: Level: beginner
4784: Concepts: matrices^row ownership
4786: .seealso: MatGetOwnershipRanges()
4788: @*/
4789: PetscErrorCode MatGetOwnershipRange(Mat mat,PetscInt *m,PetscInt* n)
4790: {
4798: MatPreallocated(mat);
4799: if (m) *m = mat->rmap.rstart;
4800: if (n) *n = mat->rmap.rend;
4801: return(0);
4802: }
4806: /*@C
4807: MatGetOwnershipRanges - Returns the range of matrix rows owned by
4808: each process
4810: Not Collective
4812: Input Parameters:
4813: . mat - the matrix
4815: Output Parameters:
4816: . ranges - start of each processors portion plus one more then the total length at the end
4818: Level: beginner
4820: Concepts: matrices^row ownership
4822: .seealso: MatGetOwnershipRange()
4824: @*/
4825: PetscErrorCode MatGetOwnershipRanges(Mat mat,const PetscInt **ranges)
4826: {
4832: PetscMapGetGlobalRange(&mat->rmap,ranges);
4833: return(0);
4834: }
4838: /*@
4839: MatILUFactorSymbolic - Performs symbolic ILU factorization of a matrix.
4840: Uses levels of fill only, not drop tolerance. Use MatLUFactorNumeric()
4841: to complete the factorization.
4843: Collective on Mat
4845: Input Parameters:
4846: + mat - the matrix
4847: . row - row permutation
4848: . column - column permutation
4849: - info - structure containing
4850: $ levels - number of levels of fill.
4851: $ expected fill - as ratio of original fill.
4852: $ 1 or 0 - indicating force fill on diagonal (improves robustness for matrices
4853: missing diagonal entries)
4855: Output Parameters:
4856: . fact - new matrix that has been symbolically factored
4858: Notes:
4859: See the users manual for additional information about
4860: choosing the fill factor for better efficiency.
4862: Most users should employ the simplified KSP interface for linear solvers
4863: instead of working directly with matrix algebra routines such as this.
4864: See, e.g., KSPCreate().
4866: Level: developer
4868: Concepts: matrices^symbolic LU factorization
4869: Concepts: matrices^factorization
4870: Concepts: LU^symbolic factorization
4872: .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor()
4873: MatGetOrdering(), MatFactorInfo
4875: @*/
4876: PetscErrorCode MatILUFactorSymbolic(Mat mat,IS row,IS col,MatFactorInfo *info,Mat *fact)
4877: {
4887: if (info->levels < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Levels of fill negative %D",(PetscInt)info->levels);
4888: if (info->fill < 1.0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Expected fill less than 1.0 %G",info->fill);
4889: if (!mat->ops->ilufactorsymbolic) SETERRQ1(PETSC_ERR_SUP,"Matrix type %s symbolic ILU",((PetscObject)mat)->type_name);
4890: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4891: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4892: MatPreallocated(mat);
4895: (*mat->ops->ilufactorsymbolic)(mat,row,col,info,fact);
4897: return(0);
4898: }
4902: /*@
4903: MatICCFactorSymbolic - Performs symbolic incomplete
4904: Cholesky factorization for a symmetric matrix. Use
4905: MatCholeskyFactorNumeric() to complete the factorization.
4907: Collective on Mat
4909: Input Parameters:
4910: + mat - the matrix
4911: . perm - row and column permutation
4912: - info - structure containing
4913: $ levels - number of levels of fill.
4914: $ expected fill - as ratio of original fill.
4916: Output Parameter:
4917: . fact - the factored matrix
4919: Notes:
4920: Most users should employ the KSP interface for linear solvers
4921: instead of working directly with matrix algebra routines such as this.
4922: See, e.g., KSPCreate().
4924: Level: developer
4926: Concepts: matrices^symbolic incomplete Cholesky factorization
4927: Concepts: matrices^factorization
4928: Concepts: Cholsky^symbolic factorization
4930: .seealso: MatCholeskyFactorNumeric(), MatCholeskyFactor(), MatFactorInfo
4931: @*/
4932: PetscErrorCode MatICCFactorSymbolic(Mat mat,IS perm,MatFactorInfo *info,Mat *fact)
4933: {
4942: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4943: if (info->levels < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Levels negative %D",(PetscInt) info->levels);
4944: if (info->fill < 1.0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Expected fill less than 1.0 %G",info->fill);
4945: if (!mat->ops->iccfactorsymbolic) SETERRQ1(PETSC_ERR_SUP,"Matrix type %s symbolic ICC",((PetscObject)mat)->type_name);
4946: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4947: MatPreallocated(mat);
4950: (*mat->ops->iccfactorsymbolic)(mat,perm,info,fact);
4952: return(0);
4953: }
4957: /*@C
4958: MatGetArray - Returns a pointer to the element values in the matrix.
4959: The result of this routine is dependent on the underlying matrix data
4960: structure, and may not even work for certain matrix types. You MUST
4961: call MatRestoreArray() when you no longer need to access the array.
4963: Not Collective
4965: Input Parameter:
4966: . mat - the matrix
4968: Output Parameter:
4969: . v - the location of the values
4972: Fortran Note:
4973: This routine is used differently from Fortran, e.g.,
4974: .vb
4975: Mat mat
4976: PetscScalar mat_array(1)
4977: PetscOffset i_mat
4978: PetscErrorCode ierr
4979: call MatGetArray(mat,mat_array,i_mat,ierr)
4981: C Access first local entry in matrix; note that array is
4982: C treated as one dimensional
4983: value = mat_array(i_mat + 1)
4985: [... other code ...]
4986: call MatRestoreArray(mat,mat_array,i_mat,ierr)
4987: .ve
4989: See the Fortran chapter of the users manual and
4990: petsc/src/mat/examples/tests for details.
4992: Level: advanced
4994: Concepts: matrices^access array
4996: .seealso: MatRestoreArray(), MatGetArrayF90(), MatGetRowIJ()
4997: @*/
4998: PetscErrorCode MatGetArray(Mat mat,PetscScalar *v[])
4999: {
5006: if (!mat->ops->getarray) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5007: MatPreallocated(mat);
5008: (*mat->ops->getarray)(mat,v);
5009: CHKMEMQ;
5010: return(0);
5011: }
5015: /*@C
5016: MatRestoreArray - Restores the matrix after MatGetArray() has been called.
5018: Not Collective
5020: Input Parameter:
5021: + mat - the matrix
5022: - v - the location of the values
5024: Fortran Note:
5025: This routine is used differently from Fortran, e.g.,
5026: .vb
5027: Mat mat
5028: PetscScalar mat_array(1)
5029: PetscOffset i_mat
5030: PetscErrorCode ierr
5031: call MatGetArray(mat,mat_array,i_mat,ierr)
5033: C Access first local entry in matrix; note that array is
5034: C treated as one dimensional
5035: value = mat_array(i_mat + 1)
5037: [... other code ...]
5038: call MatRestoreArray(mat,mat_array,i_mat,ierr)
5039: .ve
5041: See the Fortran chapter of the users manual and
5042: petsc/src/mat/examples/tests for details
5044: Level: advanced
5046: .seealso: MatGetArray(), MatRestoreArrayF90()
5047: @*/
5048: PetscErrorCode MatRestoreArray(Mat mat,PetscScalar *v[])
5049: {
5056: #if defined(PETSC_USE_DEBUG)
5057: CHKMEMQ;
5058: #endif
5059: if (!mat->ops->restorearray) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5060: (*mat->ops->restorearray)(mat,v);
5061: PetscObjectStateIncrease((PetscObject)mat);
5062: return(0);
5063: }
5067: /*@C
5068: MatGetSubMatrices - Extracts several submatrices from a matrix. If submat
5069: points to an array of valid matrices, they may be reused to store the new
5070: submatrices.
5072: Collective on Mat
5074: Input Parameters:
5075: + mat - the matrix
5076: . n - the number of submatrixes to be extracted (on this processor, may be zero)
5077: . irow, icol - index sets of rows and columns to extract
5078: - scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
5080: Output Parameter:
5081: . submat - the array of submatrices
5083: Notes:
5084: MatGetSubMatrices() can extract only sequential submatrices
5085: (from both sequential and parallel matrices). Use MatGetSubMatrix()
5086: to extract a parallel submatrix.
5088: When extracting submatrices from a parallel matrix, each processor can
5089: form a different submatrix by setting the rows and columns of its
5090: individual index sets according to the local submatrix desired.
5092: When finished using the submatrices, the user should destroy
5093: them with MatDestroyMatrices().
5095: MAT_REUSE_MATRIX can only be used when the nonzero structure of the
5096: original matrix has not changed from that last call to MatGetSubMatrices().
5098: This routine creates the matrices in submat; you should NOT create them before
5099: calling it. It also allocates the array of matrix pointers submat.
5101: For BAIJ matrices the index sets must respect the block structure, that is if they
5102: request one row/column in a block, they must request all rows/columns that are in
5103: that block. For example, if the block size is 2 you cannot request just row 0 and
5104: column 0.
5106: Fortran Note:
5107: The Fortran interface is slightly different from that given below; it
5108: requires one to pass in as submat a Mat (integer) array of size at least m.
5110: Level: advanced
5112: Concepts: matrices^accessing submatrices
5113: Concepts: submatrices
5115: .seealso: MatDestroyMatrices(), MatGetSubMatrix(), MatGetRow(), MatGetDiagonal()
5116: @*/
5117: PetscErrorCode MatGetSubMatrices(Mat mat,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *submat[])
5118: {
5120: PetscInt i;
5121: PetscTruth eq;
5126: if (n) {
5131: }
5133: if (n && scall == MAT_REUSE_MATRIX) {
5136: }
5137: if (!mat->ops->getsubmatrices) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5138: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5139: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5140: MatPreallocated(mat);
5143: (*mat->ops->getsubmatrices)(mat,n,irow,icol,scall,submat);
5145: for (i=0; i<n; i++) {
5146: if (mat->symmetric || mat->structurally_symmetric || mat->hermitian) {
5147: ISEqual(irow[i],icol[i],&eq);
5148: if (eq) {
5149: if (mat->symmetric){
5150: MatSetOption((*submat)[i],MAT_SYMMETRIC,PETSC_TRUE);
5151: } else if (mat->hermitian) {
5152: MatSetOption((*submat)[i],MAT_HERMITIAN,PETSC_TRUE);
5153: } else if (mat->structurally_symmetric) {
5154: MatSetOption((*submat)[i],MAT_STRUCTURALLY_SYMMETRIC,PETSC_TRUE);
5155: }
5156: }
5157: }
5158: }
5159: return(0);
5160: }
5164: /*@C
5165: MatDestroyMatrices - Destroys a set of matrices obtained with MatGetSubMatrices().
5167: Collective on Mat
5169: Input Parameters:
5170: + n - the number of local matrices
5171: - mat - the matrices (note that this is a pointer to the array of matrices, just to match the calling
5172: sequence of MatGetSubMatrices())
5174: Level: advanced
5176: Notes: Frees not only the matrices, but also the array that contains the matrices
5178: .seealso: MatGetSubMatrices()
5179: @*/
5180: PetscErrorCode MatDestroyMatrices(PetscInt n,Mat *mat[])
5181: {
5183: PetscInt i;
5186: if (n < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Trying to destroy negative number of matrices %D",n);
5188: for (i=0; i<n; i++) {
5189: MatDestroy((*mat)[i]);
5190: }
5191: /* memory is allocated even if n = 0 */
5192: PetscFree(*mat);
5193: return(0);
5194: }
5198: /*@C
5199: MatGetSeqNonzeroStructure - Extracts the sequential nonzero structure from a matrix.
5201: Collective on Mat
5203: Input Parameters:
5204: . mat - the matrix
5206: Output Parameter:
5207: . matstruct - the sequential matrix with the nonzero structure of mat
5208: .seealso: MatDestroySeqNonzeroStructure(), MatGetSubMatrices(), MatDestroyMatrices()
5209: @*/
5210: PetscErrorCode MatGetSeqNonzeroStructure(Mat mat,Mat *matstruct[])
5211: {
5217:
5219: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5220: MatPreallocated(mat);
5223: (*mat->ops->getseqnonzerostructure)(mat,matstruct);
5225: return(0);
5226: }
5230: /*@C
5231: MatDestroySeqNonzeroStructure - Destroys matrix obtained with MatGetSeqNonzeroStructure().
5233: Collective on Mat
5235: Input Parameters:
5236: . mat - the matrix (note that this is a pointer to the array of matrices, just to match the calling
5237: sequence of MatGetSequentialNonzeroStructure())
5239: Level: advanced
5241: Notes: Frees not only the matrices, but also the array that contains the matrices
5243: .seealso: MatGetSeqNonzeroStructure()
5244: @*/
5245: PetscErrorCode MatDestroySeqNonzeroStructure(Mat *mat[])
5246: {
5251: MatDestroyMatrices(1,mat);
5252: return(0);
5253: }
5257: /*@
5258: MatIncreaseOverlap - Given a set of submatrices indicated by index sets,
5259: replaces the index sets by larger ones that represent submatrices with
5260: additional overlap.
5262: Collective on Mat
5264: Input Parameters:
5265: + mat - the matrix
5266: . n - the number of index sets
5267: . is - the array of index sets (these index sets will changed during the call)
5268: - ov - the additional overlap requested
5270: Level: developer
5272: Concepts: overlap
5273: Concepts: ASM^computing overlap
5275: .seealso: MatGetSubMatrices()
5276: @*/
5277: PetscErrorCode MatIncreaseOverlap(Mat mat,PetscInt n,IS is[],PetscInt ov)
5278: {
5284: if (n < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Must have one or more domains, you have %D",n);
5285: if (n) {
5288: }
5289: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5290: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5291: MatPreallocated(mat);
5293: if (!ov) return(0);
5294: if (!mat->ops->increaseoverlap) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5296: (*mat->ops->increaseoverlap)(mat,n,is,ov);
5298: return(0);
5299: }
5303: /*@
5304: MatGetBlockSize - Returns the matrix block size; useful especially for the
5305: block row and block diagonal formats.
5306:
5307: Not Collective
5309: Input Parameter:
5310: . mat - the matrix
5312: Output Parameter:
5313: . bs - block size
5315: Notes:
5316: Block diagonal formats are MATSEQBDIAG, MATMPIBDIAG.
5317: Block row formats are MATSEQBAIJ, MATMPIBAIJ, MATSEQSBAIJ, MATMPISBAIJ
5319: Level: intermediate
5321: Concepts: matrices^block size
5323: .seealso: MatCreateSeqBAIJ(), MatCreateMPIBAIJ(), MatCreateSeqBDiag(), MatCreateMPIBDiag()
5324: @*/
5325: PetscErrorCode MatGetBlockSize(Mat mat,PetscInt *bs)
5326: {
5333: MatPreallocated(mat);
5334: *bs = mat->rmap.bs;
5335: return(0);
5336: }
5340: /*@
5341: MatSetBlockSize - Sets the matrix block size; for many matrix types you
5342: cannot use this and MUST set the blocksize when you preallocate the matrix
5343:
5344: Not Collective
5346: Input Parameters:
5347: + mat - the matrix
5348: - bs - block size
5350: Notes:
5351: Only works for shell and AIJ matrices
5353: Level: intermediate
5355: Concepts: matrices^block size
5357: .seealso: MatCreateSeqBAIJ(), MatCreateMPIBAIJ(), MatCreateSeqBDiag(), MatCreateMPIBDiag(), MatGetBlockSize()
5358: @*/
5359: PetscErrorCode MatSetBlockSize(Mat mat,PetscInt bs)
5360: {
5366: MatPreallocated(mat);
5367: if (mat->ops->setblocksize) {
5368: mat->rmap.bs = bs;
5369: (*mat->ops->setblocksize)(mat,bs);
5370: } else {
5371: SETERRQ1(PETSC_ERR_ARG_INCOMP,"Cannot set the blocksize for matrix type %s",((PetscObject)mat)->type_name);
5372: }
5373: return(0);
5374: }
5378: /*@C
5379: MatGetRowIJ - Returns the compressed row storage i and j indices for sequential matrices.
5381: Collective on Mat
5383: Input Parameters:
5384: + mat - the matrix
5385: . shift - 0 or 1 indicating we want the indices starting at 0 or 1
5386: . symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
5387: symmetrized
5388: - blockcompressed - PETSC_TRUE or PETSC_FALSE indicating if the nonzero structure of the
5389: blockcompressed matrix is desired or not [inode, baij have blockcompressed
5390: nonzero structure which is different than the full nonzero structure]
5392: Output Parameters:
5393: + n - number of rows in the (possibly compressed) matrix
5394: . ia - the row pointers [of length n+1]
5395: . ja - the column indices
5396: - done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
5397: are responsible for handling the case when done == PETSC_FALSE and ia and ja are not set
5399: Level: developer
5401: Notes: You CANNOT change any of the ia[] or ja[] values.
5403: Use MatRestoreRowIJ() when you are finished accessing the ia[] and ja[] values
5405: Fortran Node
5407: In Fortran use
5408: $ PetscInt ia(1), ja(1)
5409: $ PetscOffset iia, jja
5410: $ call MatGetRowIJ(mat,shift,symmetric,blockcompressed,n,ia,iia,ja,jja,done,ierr)
5411:
5412: Acess the ith and jth entries via ia(iia + i) and ja(jja + j)
5414: .seealso: MatGetColumnIJ(), MatRestoreRowIJ(), MatGetArray()
5415: @*/
5416: PetscErrorCode MatGetRowIJ(Mat mat,PetscInt shift,PetscTruth symmetric,PetscTruth blockcompressed,PetscInt *n,PetscInt *ia[],PetscInt* ja[],PetscTruth *done)
5417: {
5427: MatPreallocated(mat);
5428: if (!mat->ops->getrowij) *done = PETSC_FALSE;
5429: else {
5430: *done = PETSC_TRUE;
5432: (*mat->ops->getrowij)(mat,shift,symmetric,blockcompressed,n,ia,ja,done);
5434: }
5435: return(0);
5436: }
5440: /*@C
5441: MatGetColumnIJ - Returns the compressed column storage i and j indices for sequential matrices.
5443: Collective on Mat
5445: Input Parameters:
5446: + mat - the matrix
5447: . shift - 1 or zero indicating we want the indices starting at 0 or 1
5448: . symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
5449: symmetrized
5450: - blockcompressed - PETSC_TRUE or PETSC_FALSE indicating if the nonzero structure of the
5451: blockcompressed matrix is desired or not [inode, baij have blockcompressed
5452: nonzero structure which is different than the full nonzero structure]
5454: Output Parameters:
5455: + n - number of columns in the (possibly compressed) matrix
5456: . ia - the column pointers
5457: . ja - the row indices
5458: - done - PETSC_TRUE or PETSC_FALSE, indicating whether the values have been returned
5460: Level: developer
5462: .seealso: MatGetRowIJ(), MatRestoreColumnIJ()
5463: @*/
5464: PetscErrorCode MatGetColumnIJ(Mat mat,PetscInt shift,PetscTruth symmetric,PetscTruth blockcompressed,PetscInt *n,PetscInt *ia[],PetscInt* ja[],PetscTruth *done)
5465: {
5475: MatPreallocated(mat);
5476: if (!mat->ops->getcolumnij) *done = PETSC_FALSE;
5477: else {
5478: *done = PETSC_TRUE;
5479: (*mat->ops->getcolumnij)(mat,shift,symmetric,blockcompressed,n,ia,ja,done);
5480: }
5481: return(0);
5482: }
5486: /*@C
5487: MatRestoreRowIJ - Call after you are completed with the ia,ja indices obtained with
5488: MatGetRowIJ().
5490: Collective on Mat
5492: Input Parameters:
5493: + mat - the matrix
5494: . shift - 1 or zero indicating we want the indices starting at 0 or 1
5495: . symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
5496: symmetrized
5497: - blockcompressed - PETSC_TRUE or PETSC_FALSE indicating if the nonzero structure of the
5498: blockcompressed matrix is desired or not [inode, baij have blockcompressed
5499: nonzero structure which is different than the full nonzero structure]
5501: Output Parameters:
5502: + n - size of (possibly compressed) matrix
5503: . ia - the row pointers
5504: . ja - the column indices
5505: - done - PETSC_TRUE or PETSC_FALSE indicated that the values have been returned
5507: Level: developer
5509: .seealso: MatGetRowIJ(), MatRestoreColumnIJ()
5510: @*/
5511: PetscErrorCode MatRestoreRowIJ(Mat mat,PetscInt shift,PetscTruth symmetric,PetscTruth blockcompressed,PetscInt *n,PetscInt *ia[],PetscInt* ja[],PetscTruth *done)
5512: {
5521: MatPreallocated(mat);
5523: if (!mat->ops->restorerowij) *done = PETSC_FALSE;
5524: else {
5525: *done = PETSC_TRUE;
5526: (*mat->ops->restorerowij)(mat,shift,symmetric,blockcompressed,n,ia,ja,done);
5527: }
5528: return(0);
5529: }
5533: /*@C
5534: MatRestoreColumnIJ - Call after you are completed with the ia,ja indices obtained with
5535: MatGetColumnIJ().
5537: Collective on Mat
5539: Input Parameters:
5540: + mat - the matrix
5541: . shift - 1 or zero indicating we want the indices starting at 0 or 1
5542: - symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
5543: symmetrized
5544: - blockcompressed - PETSC_TRUE or PETSC_FALSE indicating if the nonzero structure of the
5545: blockcompressed matrix is desired or not [inode, baij have blockcompressed
5546: nonzero structure which is different than the full nonzero structure]
5548: Output Parameters:
5549: + n - size of (possibly compressed) matrix
5550: . ia - the column pointers
5551: . ja - the row indices
5552: - done - PETSC_TRUE or PETSC_FALSE indicated that the values have been returned
5554: Level: developer
5556: .seealso: MatGetColumnIJ(), MatRestoreRowIJ()
5557: @*/
5558: PetscErrorCode MatRestoreColumnIJ(Mat mat,PetscInt shift,PetscTruth symmetric,PetscTruth blockcompressed,PetscInt *n,PetscInt *ia[],PetscInt* ja[],PetscTruth *done)
5559: {
5568: MatPreallocated(mat);
5570: if (!mat->ops->restorecolumnij) *done = PETSC_FALSE;
5571: else {
5572: *done = PETSC_TRUE;
5573: (*mat->ops->restorecolumnij)(mat,shift,symmetric,blockcompressed,n,ia,ja,done);
5574: }
5575: return(0);
5576: }
5580: /*@C
5581: MatColoringPatch -Used inside matrix coloring routines that
5582: use MatGetRowIJ() and/or MatGetColumnIJ().
5584: Collective on Mat
5586: Input Parameters:
5587: + mat - the matrix
5588: . ncolors - max color value
5589: . n - number of entries in colorarray
5590: - colorarray - array indicating color for each column
5592: Output Parameters:
5593: . iscoloring - coloring generated using colorarray information
5595: Level: developer
5597: .seealso: MatGetRowIJ(), MatGetColumnIJ()
5599: @*/
5600: PetscErrorCode MatColoringPatch(Mat mat,PetscInt ncolors,PetscInt n,ISColoringValue colorarray[],ISColoring *iscoloring)
5601: {
5609: MatPreallocated(mat);
5611: if (!mat->ops->coloringpatch){
5612: ISColoringCreate(((PetscObject)mat)->comm,ncolors,n,colorarray,iscoloring);
5613: } else {
5614: (*mat->ops->coloringpatch)(mat,ncolors,n,colorarray,iscoloring);
5615: }
5616: return(0);
5617: }
5622: /*@
5623: MatSetUnfactored - Resets a factored matrix to be treated as unfactored.
5625: Collective on Mat
5627: Input Parameter:
5628: . mat - the factored matrix to be reset
5630: Notes:
5631: This routine should be used only with factored matrices formed by in-place
5632: factorization via ILU(0) (or by in-place LU factorization for the MATSEQDENSE
5633: format). This option can save memory, for example, when solving nonlinear
5634: systems with a matrix-free Newton-Krylov method and a matrix-based, in-place
5635: ILU(0) preconditioner.
5637: Note that one can specify in-place ILU(0) factorization by calling
5638: .vb
5639: PCType(pc,PCILU);
5640: PCFactorSeUseInPlace(pc);
5641: .ve
5642: or by using the options -pc_type ilu -pc_factor_in_place
5644: In-place factorization ILU(0) can also be used as a local
5645: solver for the blocks within the block Jacobi or additive Schwarz
5646: methods (runtime option: -sub_pc_factor_in_place). See the discussion
5647: of these preconditioners in the users manual for details on setting
5648: local solver options.
5650: Most users should employ the simplified KSP interface for linear solvers
5651: instead of working directly with matrix algebra routines such as this.
5652: See, e.g., KSPCreate().
5654: Level: developer
5656: .seealso: PCFactorSetUseInPlace()
5658: Concepts: matrices^unfactored
5660: @*/
5661: PetscErrorCode MatSetUnfactored(Mat mat)
5662: {
5668: MatPreallocated(mat);
5669: mat->factor = 0;
5670: if (!mat->ops->setunfactored) return(0);
5671: (*mat->ops->setunfactored)(mat);
5672: return(0);
5673: }
5675: /*MC
5676: MatGetArrayF90 - Accesses a matrix array from Fortran90.
5678: Synopsis:
5679: MatGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
5681: Not collective
5683: Input Parameter:
5684: . x - matrix
5686: Output Parameters:
5687: + xx_v - the Fortran90 pointer to the array
5688: - ierr - error code
5690: Example of Usage:
5691: .vb
5692: PetscScalar, pointer xx_v(:)
5693: ....
5694: call MatGetArrayF90(x,xx_v,ierr)
5695: a = xx_v(3)
5696: call MatRestoreArrayF90(x,xx_v,ierr)
5697: .ve
5699: Notes:
5700: Not yet supported for all F90 compilers
5702: Level: advanced
5704: .seealso: MatRestoreArrayF90(), MatGetArray(), MatRestoreArray()
5706: Concepts: matrices^accessing array
5708: M*/
5710: /*MC
5711: MatRestoreArrayF90 - Restores a matrix array that has been
5712: accessed with MatGetArrayF90().
5714: Synopsis:
5715: MatRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
5717: Not collective
5719: Input Parameters:
5720: + x - matrix
5721: - xx_v - the Fortran90 pointer to the array
5723: Output Parameter:
5724: . ierr - error code
5726: Example of Usage:
5727: .vb
5728: PetscScalar, pointer xx_v(:)
5729: ....
5730: call MatGetArrayF90(x,xx_v,ierr)
5731: a = xx_v(3)
5732: call MatRestoreArrayF90(x,xx_v,ierr)
5733: .ve
5734:
5735: Notes:
5736: Not yet supported for all F90 compilers
5738: Level: advanced
5740: .seealso: MatGetArrayF90(), MatGetArray(), MatRestoreArray()
5742: M*/
5747: /*@
5748: MatGetSubMatrix - Gets a single submatrix on the same number of processors
5749: as the original matrix.
5751: Collective on Mat
5753: Input Parameters:
5754: + mat - the original matrix
5755: . isrow - rows this processor should obtain
5756: . iscol - columns for all processors you wish to keep
5757: . csize - number of columns "local" to this processor (does nothing for sequential
5758: matrices). This should match the result from VecGetLocalSize(x,...) if you
5759: plan to use the matrix in a A*x; alternatively, you can use PETSC_DECIDE
5760: - cll - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
5762: Output Parameter:
5763: . newmat - the new submatrix, of the same type as the old
5765: Level: advanced
5767: Notes: the iscol argument MUST be the same on each processor. You might be
5768: able to create the iscol argument with ISAllGather(). The rows is isrow will be
5769: sorted into the same order as the original matrix.
5771: The first time this is called you should use a cll of MAT_INITIAL_MATRIX,
5772: the MatGetSubMatrix() routine will create the newmat for you. Any additional calls
5773: to this routine with a mat of the same nonzero structure and with a call of MAT_REUSE_MATRIX
5774: will reuse the matrix generated the first time. You should call MatDestroy() on newmat when
5775: you are finished using it.
5777: Concepts: matrices^submatrices
5779: .seealso: MatGetSubMatrices(), ISAllGather()
5780: @*/
5781: PetscErrorCode MatGetSubMatrix(Mat mat,IS isrow,IS iscol,PetscInt csize,MatReuse cll,Mat *newmat)
5782: {
5784: PetscMPIInt size;
5785: Mat *local;
5794: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5795: MatPreallocated(mat);
5796: MPI_Comm_size(((PetscObject)mat)->comm,&size);
5798: /* if original matrix is on just one processor then use submatrix generated */
5799: if (!mat->ops->getsubmatrix && size == 1 && cll == MAT_REUSE_MATRIX) {
5800: MatGetSubMatrices(mat,1,&isrow,&iscol,MAT_REUSE_MATRIX,&newmat);
5801: return(0);
5802: } else if (!mat->ops->getsubmatrix && size == 1) {
5803: MatGetSubMatrices(mat,1,&isrow,&iscol,MAT_INITIAL_MATRIX,&local);
5804: *newmat = *local;
5805: PetscFree(local);
5806: return(0);
5807: }
5809: if (!mat->ops->getsubmatrix) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5810: (*mat->ops->getsubmatrix)(mat,isrow,iscol,csize,cll,newmat);
5811: PetscObjectStateIncrease((PetscObject)*newmat);
5812: return(0);
5813: }
5817: /*@
5818: MatGetSubMatrixRaw - Gets a single submatrix on the same number of processors
5819: as the original matrix.
5821: Collective on Mat
5823: Input Parameters:
5824: + mat - the original matrix
5825: . nrows - the number of rows this processor should obtain
5826: . rows - rows this processor should obtain
5827: . ncols - the number of columns for all processors you wish to keep
5828: . cols - columns for all processors you wish to keep
5829: . csize - number of columns "local" to this processor (does nothing for sequential
5830: matrices). This should match the result from VecGetLocalSize(x,...) if you
5831: plan to use the matrix in a A*x; alternatively, you can use PETSC_DECIDE
5832: - cll - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
5834: Output Parameter:
5835: . newmat - the new submatrix, of the same type as the old
5837: Level: advanced
5839: Notes: the iscol argument MUST be the same on each processor. You might be
5840: able to create the iscol argument with ISAllGather().
5842: The first time this is called you should use a cll of MAT_INITIAL_MATRIX,
5843: the MatGetSubMatrix() routine will create the newmat for you. Any additional calls
5844: to this routine with a mat of the same nonzero structure and with a cll of MAT_REUSE_MATRIX
5845: will reuse the matrix generated the first time.
5847: Concepts: matrices^submatrices
5849: .seealso: MatGetSubMatrices(), ISAllGather()
5850: @*/
5851: PetscErrorCode MatGetSubMatrixRaw(Mat mat,PetscInt nrows,const PetscInt rows[],PetscInt ncols,const PetscInt cols[],PetscInt csize,MatReuse cll,Mat *newmat)
5852: {
5853: IS isrow, iscol;
5863: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5864: MatPreallocated(mat);
5865: ISCreateGeneralWithArray(PETSC_COMM_SELF, nrows, (PetscInt *) rows, &isrow);
5866: ISCreateGeneralWithArray(PETSC_COMM_SELF, ncols, (PetscInt *) cols, &iscol);
5867: MatGetSubMatrix(mat, isrow, iscol, csize, cll, newmat);
5868: ISDestroy(isrow);
5869: ISDestroy(iscol);
5870: return(0);
5871: }
5875: /*@
5876: MatStashSetInitialSize - sets the sizes of the matrix stash, that is
5877: used during the assembly process to store values that belong to
5878: other processors.
5880: Not Collective
5882: Input Parameters:
5883: + mat - the matrix
5884: . size - the initial size of the stash.
5885: - bsize - the initial size of the block-stash(if used).
5887: Options Database Keys:
5888: + -matstash_initial_size <size> or <size0,size1,...sizep-1>
5889: - -matstash_block_initial_size <bsize> or <bsize0,bsize1,...bsizep-1>
5891: Level: intermediate
5893: Notes:
5894: The block-stash is used for values set with MatSetValuesBlocked() while
5895: the stash is used for values set with MatSetValues()
5897: Run with the option -info and look for output of the form
5898: MatAssemblyBegin_MPIXXX:Stash has MM entries, uses nn mallocs.
5899: to determine the appropriate value, MM, to use for size and
5900: MatAssemblyBegin_MPIXXX:Block-Stash has BMM entries, uses nn mallocs.
5901: to determine the value, BMM to use for bsize
5903: Concepts: stash^setting matrix size
5904: Concepts: matrices^stash
5906: @*/
5907: PetscErrorCode MatStashSetInitialSize(Mat mat,PetscInt size, PetscInt bsize)
5908: {
5914: MatStashSetInitialSize_Private(&mat->stash,size);
5915: MatStashSetInitialSize_Private(&mat->bstash,bsize);
5916: return(0);
5917: }
5921: /*@
5922: MatInterpolateAdd - w = y + A*x or A'*x depending on the shape of
5923: the matrix
5925: Collective on Mat
5927: Input Parameters:
5928: + mat - the matrix
5929: . x,y - the vectors
5930: - w - where the result is stored
5932: Level: intermediate
5934: Notes:
5935: w may be the same vector as y.
5937: This allows one to use either the restriction or interpolation (its transpose)
5938: matrix to do the interpolation
5940: Concepts: interpolation
5942: .seealso: MatMultAdd(), MatMultTransposeAdd(), MatRestrict()
5944: @*/
5945: PetscErrorCode MatInterpolateAdd(Mat A,Vec x,Vec y,Vec w)
5946: {
5948: PetscInt M,N;
5956: MatPreallocated(A);
5957: MatGetSize(A,&M,&N);
5958: if (N > M) {
5959: MatMultTransposeAdd(A,x,y,w);
5960: } else {
5961: MatMultAdd(A,x,y,w);
5962: }
5963: return(0);
5964: }
5968: /*@
5969: MatInterpolate - y = A*x or A'*x depending on the shape of
5970: the matrix
5972: Collective on Mat
5974: Input Parameters:
5975: + mat - the matrix
5976: - x,y - the vectors
5978: Level: intermediate
5980: Notes:
5981: This allows one to use either the restriction or interpolation (its transpose)
5982: matrix to do the interpolation
5984: Concepts: matrices^interpolation
5986: .seealso: MatMultAdd(), MatMultTransposeAdd(), MatRestrict()
5988: @*/
5989: PetscErrorCode MatInterpolate(Mat A,Vec x,Vec y)
5990: {
5992: PetscInt M,N;
5999: MatPreallocated(A);
6000: MatGetSize(A,&M,&N);
6001: if (N > M) {
6002: MatMultTranspose(A,x,y);
6003: } else {
6004: MatMult(A,x,y);
6005: }
6006: return(0);
6007: }
6011: /*@
6012: MatRestrict - y = A*x or A'*x
6014: Collective on Mat
6016: Input Parameters:
6017: + mat - the matrix
6018: - x,y - the vectors
6020: Level: intermediate
6022: Notes:
6023: This allows one to use either the restriction or interpolation (its transpose)
6024: matrix to do the restriction
6026: Concepts: matrices^restriction
6028: .seealso: MatMultAdd(), MatMultTransposeAdd(), MatInterpolate()
6030: @*/
6031: PetscErrorCode MatRestrict(Mat A,Vec x,Vec y)
6032: {
6034: PetscInt M,N;
6041: MatPreallocated(A);
6043: MatGetSize(A,&M,&N);
6044: if (N > M) {
6045: MatMult(A,x,y);
6046: } else {
6047: MatMultTranspose(A,x,y);
6048: }
6049: return(0);
6050: }
6054: /*@
6055: MatNullSpaceAttach - attaches a null space to a matrix.
6056: This null space will be removed from the resulting vector whenever
6057: MatMult() is called
6059: Collective on Mat
6061: Input Parameters:
6062: + mat - the matrix
6063: - nullsp - the null space object
6065: Level: developer
6067: Notes:
6068: Overwrites any previous null space that may have been attached
6070: Concepts: null space^attaching to matrix
6072: .seealso: MatCreate(), MatNullSpaceCreate()
6073: @*/
6074: PetscErrorCode MatNullSpaceAttach(Mat mat,MatNullSpace nullsp)
6075: {
6082: MatPreallocated(mat);
6083: PetscObjectReference((PetscObject)nullsp);
6084: if (mat->nullsp) { MatNullSpaceDestroy(mat->nullsp); }
6085: mat->nullsp = nullsp;
6086: return(0);
6087: }
6091: /*@
6092: MatICCFactor - Performs in-place incomplete Cholesky factorization of matrix.
6094: Collective on Mat
6096: Input Parameters:
6097: + mat - the matrix
6098: . row - row/column permutation
6099: . fill - expected fill factor >= 1.0
6100: - level - level of fill, for ICC(k)
6102: Notes:
6103: Probably really in-place only when level of fill is zero, otherwise allocates
6104: new space to store factored matrix and deletes previous memory.
6106: Most users should employ the simplified KSP interface for linear solvers
6107: instead of working directly with matrix algebra routines such as this.
6108: See, e.g., KSPCreate().
6110: Level: developer
6112: Concepts: matrices^incomplete Cholesky factorization
6113: Concepts: Cholesky factorization
6115: .seealso: MatICCFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor()
6116: @*/
6117: PetscErrorCode MatICCFactor(Mat mat,IS row,MatFactorInfo* info)
6118: {
6126: if (mat->rmap.N != mat->cmap.N) SETERRQ(PETSC_ERR_ARG_WRONG,"matrix must be square");
6127: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6128: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6129: if (!mat->ops->iccfactor) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6130: MatPreallocated(mat);
6131: (*mat->ops->iccfactor)(mat,row,info);
6132: PetscObjectStateIncrease((PetscObject)mat);
6133: return(0);
6134: }
6138: /*@
6139: MatSetValuesAdic - Sets values computed with ADIC automatic differentiation into a matrix.
6141: Not Collective
6143: Input Parameters:
6144: + mat - the matrix
6145: - v - the values compute with ADIC
6147: Level: developer
6149: Notes:
6150: Must call MatSetColoring() before using this routine. Also this matrix must already
6151: have its nonzero pattern determined.
6153: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
6154: MatSetValues(), MatSetColoring(), MatSetValuesAdifor()
6155: @*/
6156: PetscErrorCode MatSetValuesAdic(Mat mat,void *v)
6157: {
6165: if (!mat->assembled) {
6166: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Matrix must be already assembled");
6167: }
6169: if (!mat->ops->setvaluesadic) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6170: (*mat->ops->setvaluesadic)(mat,v);
6172: MatView_Private(mat);
6173: PetscObjectStateIncrease((PetscObject)mat);
6174: return(0);
6175: }
6180: /*@
6181: MatSetColoring - Sets a coloring used by calls to MatSetValuesAdic()
6183: Not Collective
6185: Input Parameters:
6186: + mat - the matrix
6187: - coloring - the coloring
6189: Level: developer
6191: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
6192: MatSetValues(), MatSetValuesAdic()
6193: @*/
6194: PetscErrorCode MatSetColoring(Mat mat,ISColoring coloring)
6195: {
6203: if (!mat->assembled) {
6204: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Matrix must be already assembled");
6205: }
6206: if (!mat->ops->setcoloring) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6207: (*mat->ops->setcoloring)(mat,coloring);
6208: return(0);
6209: }
6213: /*@
6214: MatSetValuesAdifor - Sets values computed with automatic differentiation into a matrix.
6216: Not Collective
6218: Input Parameters:
6219: + mat - the matrix
6220: . nl - leading dimension of v
6221: - v - the values compute with ADIFOR
6223: Level: developer
6225: Notes:
6226: Must call MatSetColoring() before using this routine. Also this matrix must already
6227: have its nonzero pattern determined.
6229: .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
6230: MatSetValues(), MatSetColoring()
6231: @*/
6232: PetscErrorCode MatSetValuesAdifor(Mat mat,PetscInt nl,void *v)
6233: {
6241: if (!mat->assembled) {
6242: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Matrix must be already assembled");
6243: }
6245: if (!mat->ops->setvaluesadifor) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6246: (*mat->ops->setvaluesadifor)(mat,nl,v);
6248: PetscObjectStateIncrease((PetscObject)mat);
6249: return(0);
6250: }
6254: /*@
6255: MatDiagonalScaleLocal - Scales columns of a matrix given the scaling values including the
6256: ghosted ones.
6258: Not Collective
6260: Input Parameters:
6261: + mat - the matrix
6262: - diag = the diagonal values, including ghost ones
6264: Level: developer
6266: Notes: Works only for MPIAIJ and MPIBAIJ matrices
6267:
6268: .seealso: MatDiagonalScale()
6269: @*/
6270: PetscErrorCode MatDiagonalScaleLocal(Mat mat,Vec diag)
6271: {
6273: PetscMPIInt size;
6280: if (!mat->assembled) {
6281: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Matrix must be already assembled");
6282: }
6284: MPI_Comm_size(((PetscObject)mat)->comm,&size);
6285: if (size == 1) {
6286: PetscInt n,m;
6287: VecGetSize(diag,&n);
6288: MatGetSize(mat,0,&m);
6289: if (m == n) {
6290: MatDiagonalScale(mat,0,diag);
6291: } else {
6292: SETERRQ(PETSC_ERR_SUP,"Only supported for sequential matrices when no ghost points/periodic conditions");
6293: }
6294: } else {
6295: PetscErrorCode (*f)(Mat,Vec);
6296: PetscObjectQueryFunction((PetscObject)mat,"MatDiagonalScaleLocal_C",(void (**)(void))&f);
6297: if (f) {
6298: (*f)(mat,diag);
6299: } else {
6300: SETERRQ(PETSC_ERR_SUP,"Only supported for MPIAIJ and MPIBAIJ parallel matrices");
6301: }
6302: }
6304: PetscObjectStateIncrease((PetscObject)mat);
6305: return(0);
6306: }
6310: /*@
6311: MatGetInertia - Gets the inertia from a factored matrix
6313: Collective on Mat
6315: Input Parameter:
6316: . mat - the matrix
6318: Output Parameters:
6319: + nneg - number of negative eigenvalues
6320: . nzero - number of zero eigenvalues
6321: - npos - number of positive eigenvalues
6323: Level: advanced
6325: Notes: Matrix must have been factored by MatCholeskyFactor()
6328: @*/
6329: PetscErrorCode MatGetInertia(Mat mat,PetscInt *nneg,PetscInt *nzero,PetscInt *npos)
6330: {
6336: if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
6337: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Numeric factor mat is not assembled");
6338: if (!mat->ops->getinertia) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6339: (*mat->ops->getinertia)(mat,nneg,nzero,npos);
6340: return(0);
6341: }
6343: /* ----------------------------------------------------------------*/
6346: /*@
6347: MatSolves - Solves A x = b, given a factored matrix, for a collection of vectors
6349: Collective on Mat and Vecs
6351: Input Parameters:
6352: + mat - the factored matrix
6353: - b - the right-hand-side vectors
6355: Output Parameter:
6356: . x - the result vectors
6358: Notes:
6359: The vectors b and x cannot be the same. I.e., one cannot
6360: call MatSolves(A,x,x).
6362: Notes:
6363: Most users should employ the simplified KSP interface for linear solvers
6364: instead of working directly with matrix algebra routines such as this.
6365: See, e.g., KSPCreate().
6367: Level: developer
6369: Concepts: matrices^triangular solves
6371: .seealso: MatSolveAdd(), MatSolveTranspose(), MatSolveTransposeAdd(), MatSolve()
6372: @*/
6373: PetscErrorCode MatSolves(Mat mat,Vecs b,Vecs x)
6374: {
6380: if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
6381: if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
6382: if (!mat->rmap.N && !mat->cmap.N) return(0);
6384: if (!mat->ops->solves) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6385: MatPreallocated(mat);
6387: (*mat->ops->solves)(mat,b,x);
6389: return(0);
6390: }
6394: /*@
6395: MatIsSymmetric - Test whether a matrix is symmetric
6397: Collective on Mat
6399: Input Parameter:
6400: + A - the matrix to test
6401: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact transpose)
6403: Output Parameters:
6404: . flg - the result
6406: Level: intermediate
6408: Concepts: matrix^symmetry
6410: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetricKnown()
6411: @*/
6412: PetscErrorCode MatIsSymmetric(Mat A,PetscReal tol,PetscTruth *flg)
6413: {
6419: if (!A->symmetric_set) {
6420: if (!A->ops->issymmetric) {
6421: MatType mattype;
6422: MatGetType(A,&mattype);
6423: SETERRQ1(PETSC_ERR_SUP,"Matrix of type <%s> does not support checking for symmetric",mattype);
6424: }
6425: (*A->ops->issymmetric)(A,tol,&A->symmetric);
6426: A->symmetric_set = PETSC_TRUE;
6427: if (A->symmetric) {
6428: A->structurally_symmetric_set = PETSC_TRUE;
6429: A->structurally_symmetric = PETSC_TRUE;
6430: }
6431: }
6432: *flg = A->symmetric;
6433: return(0);
6434: }
6438: /*@
6439: MatIsHermitian - Test whether a matrix is Hermitian
6441: Collective on Mat
6443: Input Parameter:
6444: + A - the matrix to test
6445: - tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact Hermitian)
6447: Output Parameters:
6448: . flg - the result
6450: Level: intermediate
6452: Concepts: matrix^symmetry
6454: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetricKnown()
6455: @*/
6456: PetscErrorCode MatIsHermitian(Mat A,PetscReal tol,PetscTruth *flg)
6457: {
6463: if (!A->hermitian_set) {
6464: if (!A->ops->ishermitian) {
6465: MatType mattype;
6466: MatGetType(A,&mattype);
6467: SETERRQ1(PETSC_ERR_SUP,"Matrix of type <%s> does not support checking for Hermitian",mattype);
6468: }
6469: (*A->ops->ishermitian)(A,tol,&A->hermitian);
6470: A->hermitian_set = PETSC_TRUE;
6471: if (A->hermitian) {
6472: A->structurally_symmetric_set = PETSC_TRUE;
6473: A->structurally_symmetric = PETSC_TRUE;
6474: }
6475: }
6476: *flg = A->hermitian;
6477: return(0);
6478: }
6482: /*@
6483: MatIsSymmetricKnown - Checks the flag on the matrix to see if it is symmetric.
6485: Collective on Mat
6487: Input Parameter:
6488: . A - the matrix to check
6490: Output Parameters:
6491: + set - if the symmetric flag is set (this tells you if the next flag is valid)
6492: - flg - the result
6494: Level: advanced
6496: Concepts: matrix^symmetry
6498: Note: Does not check the matrix values directly, so this may return unknown (set = PETSC_FALSE). Use MatIsSymmetric()
6499: if you want it explicitly checked
6501: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetric()
6502: @*/
6503: PetscErrorCode MatIsSymmetricKnown(Mat A,PetscTruth *set,PetscTruth *flg)
6504: {
6509: if (A->symmetric_set) {
6510: *set = PETSC_TRUE;
6511: *flg = A->symmetric;
6512: } else {
6513: *set = PETSC_FALSE;
6514: }
6515: return(0);
6516: }
6520: /*@
6521: MatIsHermitianKnown - Checks the flag on the matrix to see if it is hermitian.
6523: Collective on Mat
6525: Input Parameter:
6526: . A - the matrix to check
6528: Output Parameters:
6529: + set - if the hermitian flag is set (this tells you if the next flag is valid)
6530: - flg - the result
6532: Level: advanced
6534: Concepts: matrix^symmetry
6536: Note: Does not check the matrix values directly, so this may return unknown (set = PETSC_FALSE). Use MatIsHermitian()
6537: if you want it explicitly checked
6539: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetric()
6540: @*/
6541: PetscErrorCode MatIsHermitianKnown(Mat A,PetscTruth *set,PetscTruth *flg)
6542: {
6547: if (A->hermitian_set) {
6548: *set = PETSC_TRUE;
6549: *flg = A->hermitian;
6550: } else {
6551: *set = PETSC_FALSE;
6552: }
6553: return(0);
6554: }
6558: /*@
6559: MatIsStructurallySymmetric - Test whether a matrix is structurally symmetric
6561: Collective on Mat
6563: Input Parameter:
6564: . A - the matrix to test
6566: Output Parameters:
6567: . flg - the result
6569: Level: intermediate
6571: Concepts: matrix^symmetry
6573: .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsSymmetric(), MatSetOption()
6574: @*/
6575: PetscErrorCode MatIsStructurallySymmetric(Mat A,PetscTruth *flg)
6576: {
6582: if (!A->structurally_symmetric_set) {
6583: if (!A->ops->isstructurallysymmetric) SETERRQ(PETSC_ERR_SUP,"Matrix does not support checking for structural symmetric");
6584: (*A->ops->isstructurallysymmetric)(A,&A->structurally_symmetric);
6585: A->structurally_symmetric_set = PETSC_TRUE;
6586: }
6587: *flg = A->structurally_symmetric;
6588: return(0);
6589: }
6594: /*@
6595: MatStashGetInfo - Gets how many values are currently in the vector stash, i.e. need
6596: to be communicated to other processors during the MatAssemblyBegin/End() process
6598: Not collective
6600: Input Parameter:
6601: . vec - the vector
6603: Output Parameters:
6604: + nstash - the size of the stash
6605: . reallocs - the number of additional mallocs incurred.
6606: . bnstash - the size of the block stash
6607: - breallocs - the number of additional mallocs incurred.in the block stash
6608:
6609: Level: advanced
6611: .seealso: MatAssemblyBegin(), MatAssemblyEnd(), Mat, MatStashSetInitialSize()
6612:
6613: @*/
6614: PetscErrorCode MatStashGetInfo(Mat mat,PetscInt *nstash,PetscInt *reallocs,PetscInt *bnstash,PetscInt *breallocs)
6615: {
6618: MatStashGetInfo_Private(&mat->stash,nstash,reallocs);
6619: MatStashGetInfo_Private(&mat->bstash,bnstash,breallocs);
6620: return(0);
6621: }
6625: /*@
6626: MatGetVecs - Get vector(s) compatible with the matrix, i.e. with the same
6627: parallel layout
6628:
6629: Collective on Mat
6631: Input Parameter:
6632: . mat - the matrix
6634: Output Parameter:
6635: + right - (optional) vector that the matrix can be multiplied against
6636: - left - (optional) vector that the matrix vector product can be stored in
6638: Level: advanced
6640: .seealso: MatCreate()
6641: @*/
6642: PetscErrorCode MatGetVecs(Mat mat,Vec *right,Vec *left)
6643: {
6649: MatPreallocated(mat);
6650: if (mat->ops->getvecs) {
6651: (*mat->ops->getvecs)(mat,right,left);
6652: } else {
6653: PetscMPIInt size;
6654: MPI_Comm_size(((PetscObject)mat)->comm, &size);
6655: if (right) {
6656: VecCreate(((PetscObject)mat)->comm,right);
6657: VecSetSizes(*right,mat->cmap.n,PETSC_DETERMINE);
6658: if (size > 1) {VecSetType(*right,VECMPI);}
6659: else {VecSetType(*right,VECSEQ);}
6660: }
6661: if (left) {
6662: VecCreate(((PetscObject)mat)->comm,left);
6663: VecSetSizes(*left,mat->rmap.n,PETSC_DETERMINE);
6664: if (size > 1) {VecSetType(*left,VECMPI);}
6665: else {VecSetType(*left,VECSEQ);}
6666: }
6667: }
6668: if (right) {VecSetBlockSize(*right,mat->rmap.bs);}
6669: if (left) {VecSetBlockSize(*left,mat->rmap.bs);}
6670: if (mat->mapping) {
6671: if (right) {VecSetLocalToGlobalMapping(*right,mat->mapping);}
6672: if (left) {VecSetLocalToGlobalMapping(*left,mat->mapping);}
6673: }
6674: if (mat->bmapping) {
6675: if (right) {VecSetLocalToGlobalMappingBlock(*right,mat->bmapping);}
6676: if (left) {VecSetLocalToGlobalMappingBlock(*left,mat->bmapping);}
6677: }
6678: return(0);
6679: }
6683: /*@
6684: MatFactorInfoInitialize - Initializes a MatFactorInfo data structure
6685: with default values.
6687: Not Collective
6689: Input Parameters:
6690: . info - the MatFactorInfo data structure
6693: Notes: The solvers are generally used through the KSP and PC objects, for example
6694: PCLU, PCILU, PCCHOLESKY, PCICC
6696: Level: developer
6698: .seealso: MatFactorInfo
6699: @*/
6701: PetscErrorCode MatFactorInfoInitialize(MatFactorInfo *info)
6702: {
6706: PetscMemzero(info,sizeof(MatFactorInfo));
6707: return(0);
6708: }
6712: /*@
6713: MatPtAP - Creates the matrix projection C = P^T * A * P
6715: Collective on Mat
6717: Input Parameters:
6718: + A - the matrix
6719: . P - the projection matrix
6720: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
6721: - fill - expected fill as ratio of nnz(C)/nnz(A)
6723: Output Parameters:
6724: . C - the product matrix
6726: Notes:
6727: C will be created and must be destroyed by the user with MatDestroy().
6729: This routine is currently only implemented for pairs of AIJ matrices and classes
6730: which inherit from AIJ.
6732: Level: intermediate
6734: .seealso: MatPtAPSymbolic(), MatPtAPNumeric(), MatMatMult()
6735: @*/
6736: PetscErrorCode MatPtAP(Mat A,Mat P,MatReuse scall,PetscReal fill,Mat *C)
6737: {
6743: if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6744: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6747: MatPreallocated(P);
6748: if (!P->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6749: if (P->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6751: if (P->rmap.N!=A->cmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->rmap.N,A->cmap.N);
6752: if (fill < 1.0) SETERRQ1(PETSC_ERR_ARG_SIZ,"Expected fill=%G must be >= 1.0",fill);
6753: MatPreallocated(A);
6756: (*A->ops->ptap)(A,P,scall,fill,C);
6759: return(0);
6760: }
6764: /*@
6765: MatPtAPNumeric - Computes the matrix projection C = P^T * A * P
6767: Collective on Mat
6769: Input Parameters:
6770: + A - the matrix
6771: - P - the projection matrix
6773: Output Parameters:
6774: . C - the product matrix
6776: Notes:
6777: C must have been created by calling MatPtAPSymbolic and must be destroyed by
6778: the user using MatDeatroy().
6780: This routine is currently only implemented for pairs of AIJ matrices and classes
6781: which inherit from AIJ. C will be of type MATAIJ.
6783: Level: intermediate
6785: .seealso: MatPtAP(), MatPtAPSymbolic(), MatMatMultNumeric()
6786: @*/
6787: PetscErrorCode MatPtAPNumeric(Mat A,Mat P,Mat C)
6788: {
6794: if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6795: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6798: MatPreallocated(P);
6799: if (!P->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6800: if (P->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6803: MatPreallocated(C);
6804: if (C->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6805: if (P->cmap.N!=C->rmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->cmap.N,C->rmap.N);
6806: if (P->rmap.N!=A->cmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->rmap.N,A->cmap.N);
6807: if (A->rmap.N!=A->cmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix 'A' must be square, %D != %D",A->rmap.N,A->cmap.N);
6808: if (P->cmap.N!=C->cmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->cmap.N,C->cmap.N);
6809: MatPreallocated(A);
6812: (*A->ops->ptapnumeric)(A,P,C);
6814: return(0);
6815: }
6819: /*@
6820: MatPtAPSymbolic - Creates the (i,j) structure of the matrix projection C = P^T * A * P
6822: Collective on Mat
6824: Input Parameters:
6825: + A - the matrix
6826: - P - the projection matrix
6828: Output Parameters:
6829: . C - the (i,j) structure of the product matrix
6831: Notes:
6832: C will be created and must be destroyed by the user with MatDestroy().
6834: This routine is currently only implemented for pairs of SeqAIJ matrices and classes
6835: which inherit from SeqAIJ. C will be of type MATSEQAIJ. The product is computed using
6836: this (i,j) structure by calling MatPtAPNumeric().
6838: Level: intermediate
6840: .seealso: MatPtAP(), MatPtAPNumeric(), MatMatMultSymbolic()
6841: @*/
6842: PetscErrorCode MatPtAPSymbolic(Mat A,Mat P,PetscReal fill,Mat *C)
6843: {
6849: if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6850: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6851: if (fill <1.0) SETERRQ1(PETSC_ERR_ARG_SIZ,"Expected fill=%G must be >= 1.0",fill);
6854: MatPreallocated(P);
6855: if (!P->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6856: if (P->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6859: if (P->rmap.N!=A->cmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->rmap.N,A->cmap.N);
6860: if (A->rmap.N!=A->cmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix 'A' must be square, %D != %D",A->rmap.N,A->cmap.N);
6861: MatPreallocated(A);
6863: (*A->ops->ptapsymbolic)(A,P,fill,C);
6866: MatSetBlockSize(*C,A->rmap.bs);
6868: return(0);
6869: }
6873: /*@
6874: MatMatMult - Performs Matrix-Matrix Multiplication C=A*B.
6876: Collective on Mat
6878: Input Parameters:
6879: + A - the left matrix
6880: . B - the right matrix
6881: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
6882: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B))
6884: Output Parameters:
6885: . C - the product matrix
6887: Notes:
6888: C will be created and must be destroyed by the user with MatDestroy().
6889: Unless scall is MAT_REUSE_MATRIX
6891: If you have many matrices with the same non-zero structure to multiply, you
6892: should either
6893: $ 1) use MAT_REUSE_MATRIX in all calls but the first or
6894: $ 2) call MatMatMultSymbolic() once and then MatMatMultNumeric() for each product needed
6896: Level: intermediate
6898: .seealso: MatMatMultSymbolic(), MatMatMultNumeric(), MatPtAP()
6899: @*/
6900: PetscErrorCode MatMatMult(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
6901: {
6903: PetscErrorCode (*fA)(Mat,Mat,MatReuse,PetscReal,Mat*);
6904: PetscErrorCode (*fB)(Mat,Mat,MatReuse,PetscReal,Mat*);
6905: PetscErrorCode (*mult)(Mat,Mat,MatReuse,PetscReal,Mat *)=PETSC_NULL;
6910: if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6911: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6914: MatPreallocated(B);
6915: if (!B->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6916: if (B->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6918: if (B->rmap.N!=A->cmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->rmap.N,A->cmap.N);
6919: if (scall == MAT_REUSE_MATRIX){
6922: }
6923: if (fill == PETSC_DEFAULT) fill = 2.0;
6924: if (fill < 1.0) SETERRQ1(PETSC_ERR_ARG_SIZ,"Expected fill=%G must be >= 1.0",fill);
6925: MatPreallocated(A);
6927: fA = A->ops->matmult;
6928: fB = B->ops->matmult;
6929: if (fB == fA) {
6930: if (!fB) SETERRQ1(PETSC_ERR_SUP,"MatMatMult not supported for B of type %s",((PetscObject)B)->type_name);
6931: mult = fB;
6932: } else {
6933: /* dispatch based on the type of A and B */
6934: char multname[256];
6935: PetscStrcpy(multname,"MatMatMult_");
6936: PetscStrcat(multname,((PetscObject)A)->type_name);
6937: PetscStrcat(multname,"_");
6938: PetscStrcat(multname,((PetscObject)B)->type_name);
6939: PetscStrcat(multname,"_C"); /* e.g., multname = "MatMatMult_aij_dense_C" */
6940: PetscObjectQueryFunction((PetscObject)B,multname,(void (**)(void))&mult);
6941: if (!mult) SETERRQ2(PETSC_ERR_ARG_INCOMP,"MatMatMult requires A, %s, to be compatible with B, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
6942: }
6944: (*mult)(A,B,scall,fill,C);
6946: return(0);
6947: }
6951: /*@
6952: MatMatMultSymbolic - Performs construction, preallocation, and computes the ij structure
6953: of the matrix-matrix product C=A*B. Call this routine before calling MatMatMultNumeric().
6955: Collective on Mat
6957: Input Parameters:
6958: + A - the left matrix
6959: . B - the right matrix
6960: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B))
6962: Output Parameters:
6963: . C - the matrix containing the ij structure of product matrix
6965: Notes:
6966: C will be created and must be destroyed by the user with MatDestroy().
6968: This routine is currently implemented for
6969: - pairs of AIJ matrices and classes which inherit from AIJ, C will be of type MATAIJ.
6970: - pairs of AIJ (A) and Dense (B) matrix, C will be of type MATDENSE.
6972: Level: intermediate
6974: .seealso: MatMatMult(), MatMatMultNumeric()
6975: @*/
6976: PetscErrorCode MatMatMultSymbolic(Mat A,Mat B,PetscReal fill,Mat *C)
6977: {
6979: PetscErrorCode (*Asymbolic)(Mat,Mat,PetscReal,Mat *);
6980: PetscErrorCode (*Bsymbolic)(Mat,Mat,PetscReal,Mat *);
6981: PetscErrorCode (*symbolic)(Mat,Mat,PetscReal,Mat *)=PETSC_NULL;
6986: if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6987: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6991: MatPreallocated(B);
6992: if (!B->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6993: if (B->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6996: if (B->rmap.N!=A->cmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->rmap.N,A->cmap.N);
6997: if (fill == PETSC_DEFAULT) fill = 2.0;
6998: if (fill < 1.0) SETERRQ1(PETSC_ERR_ARG_SIZ,"Expected fill=%G must be > 1.0",fill);
6999: MatPreallocated(A);
7000:
7001: Asymbolic = A->ops->matmultsymbolic;
7002: Bsymbolic = B->ops->matmultsymbolic;
7003: if (Asymbolic == Bsymbolic){
7004: if (!Bsymbolic) SETERRQ1(PETSC_ERR_SUP,"C=A*B not implemented for B of type %s",((PetscObject)B)->type_name);
7005: symbolic = Bsymbolic;
7006: } else { /* dispatch based on the type of A and B */
7007: char symbolicname[256];
7008: PetscStrcpy(symbolicname,"MatMatMultSymbolic_");
7009: PetscStrcat(symbolicname,((PetscObject)A)->type_name);
7010: PetscStrcat(symbolicname,"_");
7011: PetscStrcat(symbolicname,((PetscObject)B)->type_name);
7012: PetscStrcat(symbolicname,"_C");
7013: PetscObjectQueryFunction((PetscObject)B,symbolicname,(void (**)(void))&symbolic);
7014: if (!symbolic) SETERRQ2(PETSC_ERR_ARG_INCOMP,"MatMatMultSymbolic requires A, %s, to be compatible with B, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
7015: }
7017: (*symbolic)(A,B,fill,C);
7019: return(0);
7020: }
7024: /*@
7025: MatMatMultNumeric - Performs the numeric matrix-matrix product.
7026: Call this routine after first calling MatMatMultSymbolic().
7028: Collective on Mat
7030: Input Parameters:
7031: + A - the left matrix
7032: - B - the right matrix
7034: Output Parameters:
7035: . C - the product matrix, whose ij structure was defined from MatMatMultSymbolic().
7037: Notes:
7038: C must have been created with MatMatMultSymbolic.
7040: This routine is currently implemented for
7041: - pairs of AIJ matrices and classes which inherit from AIJ, C will be of type MATAIJ.
7042: - pairs of AIJ (A) and Dense (B) matrix, C will be of type MATDENSE.
7044: Level: intermediate
7046: .seealso: MatMatMult(), MatMatMultSymbolic()
7047: @*/
7048: PetscErrorCode MatMatMultNumeric(Mat A,Mat B,Mat C)
7049: {
7051: PetscErrorCode (*Anumeric)(Mat,Mat,Mat);
7052: PetscErrorCode (*Bnumeric)(Mat,Mat,Mat);
7053: PetscErrorCode (*numeric)(Mat,Mat,Mat)=PETSC_NULL;
7058: if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7059: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7063: MatPreallocated(B);
7064: if (!B->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7065: if (B->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7069: MatPreallocated(C);
7070: if (!C->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7071: if (C->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7073: if (B->cmap.N!=C->cmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->cmap.N,C->cmap.N);
7074: if (B->rmap.N!=A->cmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->rmap.N,A->cmap.N);
7075: if (A->rmap.N!=C->rmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",A->rmap.N,C->rmap.N);
7076: MatPreallocated(A);
7078: Anumeric = A->ops->matmultnumeric;
7079: Bnumeric = B->ops->matmultnumeric;
7080: if (Anumeric == Bnumeric){
7081: if (!Bnumeric) SETERRQ1(PETSC_ERR_SUP,"MatMatMultNumeric not supported for B of type %s",((PetscObject)B)->type_name);
7082: numeric = Bnumeric;
7083: } else {
7084: char numericname[256];
7085: PetscStrcpy(numericname,"MatMatMultNumeric_");
7086: PetscStrcat(numericname,((PetscObject)A)->type_name);
7087: PetscStrcat(numericname,"_");
7088: PetscStrcat(numericname,((PetscObject)B)->type_name);
7089: PetscStrcat(numericname,"_C");
7090: PetscObjectQueryFunction((PetscObject)B,numericname,(void (**)(void))&numeric);
7091: if (!numeric)
7092: SETERRQ2(PETSC_ERR_ARG_INCOMP,"MatMatMultNumeric requires A, %s, to be compatible with B, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
7093: }
7095: (*numeric)(A,B,C);
7097: return(0);
7098: }
7102: /*@
7103: MatMatMultTranspose - Performs Matrix-Matrix Multiplication C=A^T*B.
7105: Collective on Mat
7107: Input Parameters:
7108: + A - the left matrix
7109: . B - the right matrix
7110: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
7111: - fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B))
7113: Output Parameters:
7114: . C - the product matrix
7116: Notes:
7117: C will be created and must be destroyed by the user with MatDestroy().
7119: This routine is currently only implemented for pairs of SeqAIJ matrices and classes
7120: which inherit from SeqAIJ. C will be of type MATSEQAIJ.
7122: Level: intermediate
7124: .seealso: MatMatMultTransposeSymbolic(), MatMatMultTransposeNumeric(), MatPtAP()
7125: @*/
7126: PetscErrorCode MatMatMultTranspose(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
7127: {
7129: PetscErrorCode (*fA)(Mat,Mat,MatReuse,PetscReal,Mat*);
7130: PetscErrorCode (*fB)(Mat,Mat,MatReuse,PetscReal,Mat*);
7135: if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7136: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7139: MatPreallocated(B);
7140: if (!B->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7141: if (B->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7143: if (B->rmap.N!=A->rmap.N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->rmap.N,A->rmap.N);
7144: if (fill < 1.0) SETERRQ1(PETSC_ERR_ARG_SIZ,"Expected fill=%G must be > 1.0",fill);
7145: MatPreallocated(A);
7147: fA = A->ops->matmulttranspose;
7148: if (!fA) SETERRQ1(PETSC_ERR_SUP,"MatMatMultTranspose not supported for A of type %s",((PetscObject)A)->type_name);
7149: fB = B->ops->matmulttranspose;
7150: if (!fB) SETERRQ1(PETSC_ERR_SUP,"MatMatMultTranspose not supported for B of type %s",((PetscObject)B)->type_name);
7151: if (fB!=fA) SETERRQ2(PETSC_ERR_ARG_INCOMP,"MatMatMultTranspose requires A, %s, to be compatible with B, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
7154: (*A->ops->matmulttranspose)(A,B,scall,fill,C);
7156:
7157: return(0);
7158: }
7162: /*@C
7163: MatGetRedundantMatrix - Create redundant matrices and put them into processors of subcommunicators.
7165: Collective on Mat
7167: Input Parameters:
7168: + mat - the matrix
7169: . nsubcomm - the number of subcommunicators (= number of redundant pareallel or sequential matrices)
7170: . subcomm - MPI communicator split from the communicator where mat resides in
7171: . mlocal_red - number of local rows of the redundant matrix
7172: - reuse - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
7174: Output Parameter:
7175: . matredundant - redundant matrix
7177: Notes:
7178: MAT_REUSE_MATRIX can only be used when the nonzero structure of the
7179: original matrix has not changed from that last call to MatGetRedundantMatrix().
7181: This routine creates the duplicated matrices in subcommunicators; you should NOT create them before
7182: calling it.
7184: Only MPIAIJ matrix is supported.
7185:
7186: Level: advanced
7188: Concepts: subcommunicator
7189: Concepts: duplicate matrix
7191: .seealso: MatDestroy()
7192: @*/
7193: PetscErrorCode MatGetRedundantMatrix(Mat mat,PetscInt nsubcomm,MPI_Comm subcomm,PetscInt mlocal_red,MatReuse reuse,Mat *matredundant)
7194: {
7199: if (nsubcomm && reuse == MAT_REUSE_MATRIX) {
7202: }
7203: if (!mat->ops->getredundantmatrix) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
7204: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7205: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7206: MatPreallocated(mat);
7209: (*mat->ops->getredundantmatrix)(mat,nsubcomm,subcomm,mlocal_red,reuse,matredundant);
7211: return(0);
7212: }