xref: /petsc/src/mat/interface/matrix.c (revision 4c49b1284a7fad5511c23c1bb39bf91d51e85708)
1 /*$Id: matrix.c,v 1.357 2000/01/11 21:00:31 bsmith Exp bsmith $*/
2 
3 /*
4    This is where the abstract matrix operations are defined
5 */
6 
7 #include "src/mat/matimpl.h"        /*I "mat.h" I*/
8 #include "src/vec/vecimpl.h"
9 
10 #undef __FUNC__
11 #define __FUNC__ "MatGetRow"
12 /*@C
13    MatGetRow - Gets a row of a matrix.  You MUST call MatRestoreRow()
14    for each row that you get to ensure that your application does
15    not bleed memory.
16 
17    Not Collective
18 
19    Input Parameters:
20 +  mat - the matrix
21 -  row - the row to get
22 
23    Output Parameters:
24 +  ncols -  the number of nonzeros in the row
25 .  cols - if nonzero, the column numbers
26 -  vals - if nonzero, the values
27 
28    Notes:
29    This routine is provided for people who need to have direct access
30    to the structure of a matrix.  We hope that we provide enough
31    high-level matrix routines that few users will need it.
32 
33    MatGetRow() always returns 0-based column indices, regardless of
34    whether the internal representation is 0-based (default) or 1-based.
35 
36    For better efficiency, set cols and/or vals to PETSC_NULL if you do
37    not wish to extract these quantities.
38 
39    The user can only examine the values extracted with MatGetRow();
40    the values cannot be altered.  To change the matrix entries, one
41    must use MatSetValues().
42 
43    You can only have one call to MatGetRow() outstanding for a particular
44    matrix at a time, per processor. MatGetRow() can only obtained rows
45    associated with the given processor, it cannot get rows from the
46    other processors; for that we suggest using MatGetSubMatrices(), then
47    MatGetRow() on the submatrix. The row indix passed to MatGetRows()
48    is in the global number of rows.
49 
50    Fortran Notes:
51    The calling sequence from Fortran is
52 .vb
53    MatGetRow(matrix,row,ncols,cols,values,ierr)
54          Mat     matrix (input)
55          integer row    (input)
56          integer ncols  (output)
57          integer cols(maxcols) (output)
58          double precision (or double complex) values(maxcols) output
59 .ve
60    where maxcols >= maximum nonzeros in any row of the matrix.
61 
62    Caution:
63    Do not try to change the contents of the output arrays (cols and vals).
64    In some cases, this may corrupt the matrix.
65 
66    Level: advanced
67 
68 .keywords: matrix, row, get, extract
69 
70 .seealso: MatRestoreRow(), MatSetValues(), MatGetValues(), MatGetSubmatrices(), MatGetDiagonal()
71 @*/
72 int MatGetRow(Mat mat,int row,int *ncols,int **cols,Scalar **vals)
73 {
74   int   ierr;
75 
76   PetscFunctionBegin;
77   PetscValidHeaderSpecific(mat,MAT_COOKIE);
78   PetscValidIntPointer(ncols);
79   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
80   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
81   if (!mat->ops->getrow) SETERRQ(PETSC_ERR_SUP,0,"");
82   PLogEventBegin(MAT_GetRow,mat,0,0,0);
83   ierr = (*mat->ops->getrow)(mat,row,ncols,cols,vals);CHKERRQ(ierr);
84   PLogEventEnd(MAT_GetRow,mat,0,0,0);
85   PetscFunctionReturn(0);
86 }
87 
88 #undef __FUNC__
89 #define __FUNC__ "MatRestoreRow"
90 /*@C
91    MatRestoreRow - Frees any temporary space allocated by MatGetRow().
92 
93    Not Collective
94 
95    Input Parameters:
96 +  mat - the matrix
97 .  row - the row to get
98 .  ncols, cols - the number of nonzeros and their columns
99 -  vals - if nonzero the column values
100 
101    Notes:
102    This routine should be called after you have finished examining the entries.
103 
104    Fortran Notes:
105    The calling sequence from Fortran is
106 .vb
107    MatRestoreRow(matrix,row,ncols,cols,values,ierr)
108       Mat     matrix (input)
109       integer row    (input)
110       integer ncols  (output)
111       integer cols(maxcols) (output)
112       double precision (or double complex) values(maxcols) output
113 .ve
114    Where maxcols >= maximum nonzeros in any row of the matrix.
115 
116    In Fortran MatRestoreRow() MUST be called after MatGetRow()
117    before another call to MatGetRow() can be made.
118 
119    Level: advanced
120 
121 .keywords: matrix, row, restore
122 
123 .seealso:  MatGetRow()
124 @*/
125 int MatRestoreRow(Mat mat,int row,int *ncols,int **cols,Scalar **vals)
126 {
127   int ierr;
128 
129   PetscFunctionBegin;
130   PetscValidHeaderSpecific(mat,MAT_COOKIE);
131   PetscValidIntPointer(ncols);
132   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
133   if (!mat->ops->restorerow) PetscFunctionReturn(0);
134   ierr = (*mat->ops->restorerow)(mat,row,ncols,cols,vals);CHKERRQ(ierr);
135   PetscFunctionReturn(0);
136 }
137 
138 #undef __FUNC__
139 #define __FUNC__ "MatView"
140 /*@C
141    MatView - Visualizes a matrix object.
142 
143    Collective on Mat
144 
145    Input Parameters:
146 +  mat - the matrix
147 -  ptr - visualization context
148 
149   Notes:
150   The available visualization contexts include
151 +    VIEWER_STDOUT_SELF - standard output (default)
152 .    VIEWER_STDOUT_WORLD - synchronized standard
153         output where only the first processor opens
154         the file.  All other processors send their
155         data to the first processor to print.
156 -     VIEWER_DRAW_WORLD - graphical display of nonzero structure
157 
158    The user can open alternative visualization contexts with
159 +    ViewerASCIIOpen() - Outputs matrix to a specified file
160 .    ViewerBinaryOpen() - Outputs matrix in binary to a
161          specified file; corresponding input uses MatLoad()
162 .    ViewerDrawOpen() - Outputs nonzero matrix structure to
163          an X window display
164 -    ViewerSocketOpen() - Outputs matrix to Socket viewer.
165          Currently only the sequential dense and AIJ
166          matrix types support the Socket viewer.
167 
168    The user can call ViewerSetFormat() to specify the output
169    format of ASCII printed objects (when using VIEWER_STDOUT_SELF,
170    VIEWER_STDOUT_WORLD and ViewerASCIIOpen).  Available formats include
171 +    VIEWER_FORMAT_ASCII_DEFAULT - default, prints matrix contents
172 .    VIEWER_FORMAT_ASCII_MATLAB - prints matrix contents in Matlab format
173 .    VIEWER_FORMAT_ASCII_DENSE - prints entire matrix including zeros
174 .    VIEWER_FORMAT_ASCII_COMMON - prints matrix contents, using a sparse
175          format common among all matrix types
176 .    VIEWER_FORMAT_ASCII_IMPL - prints matrix contents, using an implementation-specific
177          format (which is in many cases the same as the default)
178 .    VIEWER_FORMAT_ASCII_INFO - prints basic information about the matrix
179          size and structure (not the matrix entries)
180 -    VIEWER_FORMAT_ASCII_INFO_LONG - prints more detailed information about
181          the matrix structure
182 
183    Level: beginner
184 
185 .keywords: matrix, view, visualize, output, print, write, draw
186 
187 .seealso: ViewerSetFormat(), ViewerASCIIOpen(), ViewerDrawOpen(),
188           ViewerSocketOpen(), ViewerBinaryOpen(), MatLoad()
189 @*/
190 int MatView(Mat mat,Viewer viewer)
191 {
192   int        format,ierr,rows,cols;
193   PetscTruth isascii;
194   char       *cstr;
195 
196   PetscFunctionBegin;
197   PetscValidHeaderSpecific(mat,MAT_COOKIE);
198   if (!viewer) viewer = VIEWER_STDOUT_SELF;
199   PetscValidHeaderSpecific(viewer,VIEWER_COOKIE);
200   PetscCheckSameComm(mat,viewer);
201 
202   ierr = PetscTypeCompare((PetscObject)viewer,ASCII_VIEWER,&isascii);CHKERRQ(ierr);
203   if (isascii) {
204     ierr = ViewerGetFormat(viewer,&format);CHKERRQ(ierr);
205     if (format == VIEWER_FORMAT_ASCII_INFO || format == VIEWER_FORMAT_ASCII_INFO_LONG) {
206       ierr = ViewerASCIIPrintf(viewer,"Matrix Object:\n");CHKERRQ(ierr);
207       ierr = ViewerASCIIPushTab(viewer);CHKERRQ(ierr);
208       ierr = MatGetType(mat,PETSC_NULL,&cstr);CHKERRQ(ierr);
209       ierr = MatGetSize(mat,&rows,&cols);CHKERRQ(ierr);
210       ierr = ViewerASCIIPrintf(viewer,"type=%s, rows=%d, cols=%d\n",cstr,rows,cols);CHKERRQ(ierr);
211       if (mat->ops->getinfo) {
212         MatInfo info;
213         ierr = MatGetInfo(mat,MAT_GLOBAL_SUM,&info);CHKERRQ(ierr);
214         ierr = ViewerASCIIPrintf(viewer,"total: nonzeros=%d, allocated nonzeros=%d\n",
215                           (int)info.nz_used,(int)info.nz_allocated);CHKERRQ(ierr);
216       }
217     }
218   }
219   if (mat->ops->view) {
220     ierr = ViewerASCIIPushTab(viewer);CHKERRQ(ierr);
221     ierr = (*mat->ops->view)(mat,viewer);CHKERRQ(ierr);
222     ierr = ViewerASCIIPopTab(viewer);CHKERRQ(ierr);
223   } else if (!isascii) {
224     SETERRQ1(1,1,"Viewer type %s not supported",((PetscObject)viewer)->type_name);
225   }
226   if (isascii) {
227     ierr = ViewerGetFormat(viewer,&format);CHKERRQ(ierr);
228     if (format == VIEWER_FORMAT_ASCII_INFO || format == VIEWER_FORMAT_ASCII_INFO_LONG) {
229       ierr = ViewerASCIIPopTab(viewer);CHKERRQ(ierr);
230     }
231   }
232   PetscFunctionReturn(0);
233 }
234 
235 #undef __FUNC__
236 #define __FUNC__ "MatScaleSystem"
237 /*@C
238    MatScaleSystem - Scale a vector solution and right hand side to
239    match the scaling of a scaled matrix.
240 
241    Collective on Mat
242 
243    Input Parameter:
244 +  mat - the matrix
245 .  x - solution vector (or PETSC_NULL)
246 +  b - right hand side vector (or PETSC_NULL)
247 
248 
249    Notes:
250    For AIJ, BAIJ, and BDiag matrix formats, the matrices are not
251    internally scaled, so this does nothing. For MPIROWBS it
252    permutes and diagonally scales.
253 
254    The SLES methods automatically call this routine when required
255    (via PCPreSolve()) so it is rarely used directly.
256 
257    Level: Developer
258 
259 .keywords: matrix, scale
260 
261 .seealso: MatUseScaledForm(), MatUnScaleSystem()
262 @*/
263 int MatScaleSystem(Mat mat,Vec x,Vec b)
264 {
265   int ierr;
266 
267   PetscFunctionBegin;
268   PetscValidHeaderSpecific(mat,MAT_COOKIE);
269   if (x) {PetscValidHeaderSpecific(x,VEC_COOKIE);PetscCheckSameComm(mat,x);}
270   if (b) {PetscValidHeaderSpecific(b,VEC_COOKIE);PetscCheckSameComm(mat,b);}
271 
272   if (mat->ops->scalesystem) {
273     ierr = (*mat->ops->scalesystem)(mat,x,b);CHKERRQ(ierr);
274   }
275   PetscFunctionReturn(0);
276 }
277 
278 #undef __FUNC__
279 #define __FUNC__ "MatUnScaleSystem"
280 /*@C
281    MatUnScaleSystem - Unscales a vector solution and right hand side to
282    match the original scaling of a scaled matrix.
283 
284    Collective on Mat
285 
286    Input Parameter:
287 +  mat - the matrix
288 .  x - solution vector (or PETSC_NULL)
289 +  b - right hand side vector (or PETSC_NULL)
290 
291 
292    Notes:
293    For AIJ, BAIJ, and BDiag matrix formats, the matrices are not
294    internally scaled, so this does nothing. For MPIROWBS it
295    permutes and diagonally scales.
296 
297    The SLES methods automatically call this routine when required
298    (via PCPreSolve()) so it is rarely used directly.
299 
300    Level: Developer
301 
302 .keywords: matrix, scale
303 
304 .seealso: MatUseScaledForm(), MatScaleSystem()
305 @*/
306 int MatUnScaleSystem(Mat mat,Vec x,Vec b)
307 {
308   int ierr;
309 
310   PetscFunctionBegin;
311   PetscValidHeaderSpecific(mat,MAT_COOKIE);
312   if (x) {PetscValidHeaderSpecific(x,VEC_COOKIE);PetscCheckSameComm(mat,x);}
313   if (b) {PetscValidHeaderSpecific(b,VEC_COOKIE);PetscCheckSameComm(mat,b);}
314   if (mat->ops->unscalesystem) {
315     ierr = (*mat->ops->unscalesystem)(mat,x,b);CHKERRQ(ierr);
316   }
317   PetscFunctionReturn(0);
318 }
319 
320 #undef __FUNC__
321 #define __FUNC__ "MatUseScaledForm"
322 /*@C
323    MatUseScaledForm - For matrix storage formats that scale the
324    matrix (for example MPIRowBS matrices are diagonally scaled on
325    assembly) indicates matrix operations (MatMult() etc) are
326    applied using the scaled matrix.
327 
328    Collective on Mat
329 
330    Input Parameter:
331 +  mat - the matrix
332 -  scaled - PETSC_TRUE for applying the scaled, PETSC_FALSE for
333             applying the original matrix
334 
335    Notes:
336    For scaled matrix formats, applying the original, unscaled matrix
337    will be slightly more expensive
338 
339    Level: Developer
340 
341 .keywords: matrix, scale
342 
343 .seealso: MatScaleSystem(), MatUnScaleSystem()
344 @*/
345 int MatUseScaledForm(Mat mat,PetscTruth scaled)
346 {
347   int ierr;
348 
349   PetscFunctionBegin;
350   PetscValidHeaderSpecific(mat,MAT_COOKIE);
351   if (mat->ops->usescaledform) {
352     ierr = (*mat->ops->usescaledform)(mat,scaled);CHKERRQ(ierr);
353   }
354   PetscFunctionReturn(0);
355 }
356 
357 #undef __FUNC__
358 #define __FUNC__ "MatDestroy"
359 /*@C
360    MatDestroy - Frees space taken by a matrix.
361 
362    Collective on Mat
363 
364    Input Parameter:
365 .  mat - the matrix
366 
367    Level: beginner
368 
369 .keywords: matrix, destroy
370 @*/
371 int MatDestroy(Mat mat)
372 {
373   int ierr;
374 
375   PetscFunctionBegin;
376   PetscValidHeaderSpecific(mat,MAT_COOKIE);
377   if (--mat->refct > 0) PetscFunctionReturn(0);
378 
379   /* if memory was published with AMS then destroy it */
380   ierr = PetscObjectDepublish(mat);CHKERRQ(ierr);
381 
382   ierr = (*mat->ops->destroy)(mat);CHKERRQ(ierr);
383   PetscFunctionReturn(0);
384 }
385 
386 #undef __FUNC__
387 #define __FUNC__ "MatValid"
388 /*@
389    MatValid - Checks whether a matrix object is valid.
390 
391    Collective on Mat
392 
393    Input Parameter:
394 .  m - the matrix to check
395 
396    Output Parameter:
397    flg - flag indicating matrix status, either
398    PETSC_TRUE if matrix is valid, or PETSC_FALSE otherwise.
399 
400    Level: developer
401 
402 .keywords: matrix, valid
403 @*/
404 int MatValid(Mat m,PetscTruth *flg)
405 {
406   PetscFunctionBegin;
407   PetscValidIntPointer(flg);
408   if (!m)                           *flg = PETSC_FALSE;
409   else if (m->cookie != MAT_COOKIE) *flg = PETSC_FALSE;
410   else                              *flg = PETSC_TRUE;
411   PetscFunctionReturn(0);
412 }
413 
414 #undef __FUNC__
415 #define __FUNC__ "MatSetValues"
416 /*@
417    MatSetValues - Inserts or adds a block of values into a matrix.
418    These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
419    MUST be called after all calls to MatSetValues() have been completed.
420 
421    Not Collective
422 
423    Input Parameters:
424 +  mat - the matrix
425 .  v - a logically two-dimensional array of values
426 .  m, idxm - the number of rows and their global indices
427 .  n, idxn - the number of columns and their global indices
428 -  addv - either ADD_VALUES or INSERT_VALUES, where
429    ADD_VALUES adds values to any existing entries, and
430    INSERT_VALUES replaces existing entries with new values
431 
432    Notes:
433    By default the values, v, are row-oriented and unsorted.
434    See MatSetOption() for other options.
435 
436    Calls to MatSetValues() with the INSERT_VALUES and ADD_VALUES
437    options cannot be mixed without intervening calls to the assembly
438    routines.
439 
440    MatSetValues() uses 0-based row and column numbers in Fortran
441    as well as in C.
442 
443    Negative indices may be passed in idxm and idxn, these rows and columns are
444    simply ignored. This allows easily inserting element stiffness matrices
445    with homogeneous Dirchlet boundary conditions that you don't want represented
446    in the matrix.
447 
448    Efficiency Alert:
449    The routine MatSetValuesBlocked() may offer much better efficiency
450    for users of block sparse formats (MATSEQBAIJ and MATMPIBAIJ).
451 
452    Level: beginner
453 
454 .keywords: matrix, insert, add, set, values
455 
456 .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
457 @*/
458 int MatSetValues(Mat mat,int m,int *idxm,int n,int *idxn,Scalar *v,InsertMode addv)
459 {
460   int ierr;
461 
462   PetscFunctionBegin;
463   if (!m || !n) PetscFunctionReturn(0); /* no values to insert */
464   PetscValidHeaderSpecific(mat,MAT_COOKIE);
465   PetscValidIntPointer(idxm);
466   PetscValidIntPointer(idxn);
467   PetscValidScalarPointer(v);
468   if (mat->insertmode == NOT_SET_VALUES) {
469     mat->insertmode = addv;
470   }
471 #if defined(PETSC_USE_BOPT_g)
472   else if (mat->insertmode != addv) {
473     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,1,"Cannot mix add values and insert values");
474   }
475   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
476 #endif
477 
478   if (mat->assembled) {
479     mat->was_assembled = PETSC_TRUE;
480     mat->assembled     = PETSC_FALSE;
481   }
482   PLogEventBegin(MAT_SetValues,mat,0,0,0);
483   if (!mat->ops->setvalues) SETERRQ(PETSC_ERR_SUP,1,"Not supported for this matrix type");
484   ierr = (*mat->ops->setvalues)(mat,m,idxm,n,idxn,v,addv);CHKERRQ(ierr);
485   PLogEventEnd(MAT_SetValues,mat,0,0,0);
486   PetscFunctionReturn(0);
487 }
488 
489 #undef __FUNC__
490 #define __FUNC__ "MatSetValuesBlocked"
491 /*@
492    MatSetValuesBlocked - Inserts or adds a block of values into a matrix.
493 
494    Not Collective
495 
496    Input Parameters:
497 +  mat - the matrix
498 .  v - a logically two-dimensional array of values
499 .  m, idxm - the number of block rows and their global block indices
500 .  n, idxn - the number of block columns and their global block indices
501 -  addv - either ADD_VALUES or INSERT_VALUES, where
502    ADD_VALUES adds values to any existing entries, and
503    INSERT_VALUES replaces existing entries with new values
504 
505    Notes:
506    By default the values, v, are row-oriented and unsorted. So the layout of
507    v is the same as for MatSetValues(). See MatSetOption() for other options.
508 
509    Calls to MatSetValuesBlocked() with the INSERT_VALUES and ADD_VALUES
510    options cannot be mixed without intervening calls to the assembly
511    routines.
512 
513    MatSetValuesBlocked() uses 0-based row and column numbers in Fortran
514    as well as in C.
515 
516    Negative indices may be passed in idxm and idxn, these rows and columns are
517    simply ignored. This allows easily inserting element stiffness matrices
518    with homogeneous Dirchlet boundary conditions that you don't want represented
519    in the matrix.
520 
521    Each time an entry is set within a sparse matrix via MatSetValues(),
522    internal searching must be done to determine where to place the the
523    data in the matrix storage space.  By instead inserting blocks of
524    entries via MatSetValuesBlocked(), the overhead of matrix assembly is
525    reduced.
526 
527    Restrictions:
528    MatSetValuesBlocked() is currently supported only for the block AIJ
529    matrix format (MATSEQBAIJ and MATMPIBAIJ, which are created via
530    MatCreateSeqBAIJ() and MatCreateMPIBAIJ()).
531 
532    Level: intermediate
533 
534 .keywords: matrix, insert, add, set, values
535 
536 .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesBlockedLocal()
537 @*/
538 int MatSetValuesBlocked(Mat mat,int m,int *idxm,int n,int *idxn,Scalar *v,InsertMode addv)
539 {
540   int ierr;
541 
542   PetscFunctionBegin;
543   if (!m || !n) PetscFunctionReturn(0); /* no values to insert */
544   PetscValidHeaderSpecific(mat,MAT_COOKIE);
545   PetscValidIntPointer(idxm);
546   PetscValidIntPointer(idxn);
547   PetscValidScalarPointer(v);
548   if (mat->insertmode == NOT_SET_VALUES) {
549     mat->insertmode = addv;
550   }
551 #if defined(PETSC_USE_BOPT_g)
552   else if (mat->insertmode != addv) {
553     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,1,"Cannot mix add values and insert values");
554   }
555   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
556 #endif
557 
558   if (mat->assembled) {
559     mat->was_assembled = PETSC_TRUE;
560     mat->assembled     = PETSC_FALSE;
561   }
562   PLogEventBegin(MAT_SetValues,mat,0,0,0);
563   if (!mat->ops->setvaluesblocked) SETERRQ(PETSC_ERR_SUP,1,"Not supported for this matrix type");
564   ierr = (*mat->ops->setvaluesblocked)(mat,m,idxm,n,idxn,v,addv);CHKERRQ(ierr);
565   PLogEventEnd(MAT_SetValues,mat,0,0,0);
566   PetscFunctionReturn(0);
567 }
568 
569 /*MC
570    MatSetValue - Set a single entry into a matrix.
571 
572    Synopsis:
573    void MatSetValue(Mat m,int row,int col,Scalar value,InsertMode mode);
574 
575    Not collective
576 
577    Input Parameters:
578 +  m - the matrix
579 .  row - the row location of the entry
580 .  col - the column location of the entry
581 .  value - the value to insert
582 -  mode - either INSERT_VALUES or ADD_VALUES
583 
584    Notes:
585    For efficiency one should use MatSetValues() and set several or many
586    values simultaneously if possible.
587 
588    Note that VecSetValue() does NOT return an error code (since this
589    is checked internally).
590 
591    Level: beginner
592 
593 .seealso: MatSetValues()
594 M*/
595 
596 #undef __FUNC__
597 #define __FUNC__ "MatGetValues"
598 /*@
599    MatGetValues - Gets a block of values from a matrix.
600 
601    Not Collective; currently only returns a local block
602 
603    Input Parameters:
604 +  mat - the matrix
605 .  v - a logically two-dimensional array for storing the values
606 .  m, idxm - the number of rows and their global indices
607 -  n, idxn - the number of columns and their global indices
608 
609    Notes:
610    The user must allocate space (m*n Scalars) for the values, v.
611    The values, v, are then returned in a row-oriented format,
612    analogous to that used by default in MatSetValues().
613 
614    MatGetValues() uses 0-based row and column numbers in
615    Fortran as well as in C.
616 
617    MatGetValues() requires that the matrix has been assembled
618    with MatAssemblyBegin()/MatAssemblyEnd().  Thus, calls to
619    MatSetValues() and MatGetValues() CANNOT be made in succession
620    without intermediate matrix assembly.
621 
622    Level: advanced
623 
624 .keywords: matrix, get, values
625 
626 .seealso: MatGetRow(), MatGetSubmatrices(), MatSetValues()
627 @*/
628 int MatGetValues(Mat mat,int m,int *idxm,int n,int *idxn,Scalar *v)
629 {
630   int ierr;
631 
632   PetscFunctionBegin;
633   PetscValidHeaderSpecific(mat,MAT_COOKIE);
634   PetscValidIntPointer(idxm);
635   PetscValidIntPointer(idxn);
636   PetscValidScalarPointer(v);
637   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
638   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
639   if (!mat->ops->getvalues) SETERRQ(PETSC_ERR_SUP,0,"");
640 
641   PLogEventBegin(MAT_GetValues,mat,0,0,0);
642   ierr = (*mat->ops->getvalues)(mat,m,idxm,n,idxn,v);CHKERRQ(ierr);
643   PLogEventEnd(MAT_GetValues,mat,0,0,0);
644   PetscFunctionReturn(0);
645 }
646 
647 #undef __FUNC__
648 #define __FUNC__ "MatSetLocalToGlobalMapping"
649 /*@
650    MatSetLocalToGlobalMapping - Sets a local-to-global numbering for use by
651    the routine MatSetValuesLocal() to allow users to insert matrix entries
652    using a local (per-processor) numbering.
653 
654    Not Collective
655 
656    Input Parameters:
657 +  x - the matrix
658 -  mapping - mapping created with ISLocalToGlobalMappingCreate()
659              or ISLocalToGlobalMappingCreateIS()
660 
661    Level: intermediate
662 
663 .keywords: matrix, set, values, local, global, mapping
664 
665 .seealso:  MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesLocal()
666 @*/
667 int MatSetLocalToGlobalMapping(Mat x,ISLocalToGlobalMapping mapping)
668 {
669   int ierr;
670   PetscFunctionBegin;
671   PetscValidHeaderSpecific(x,MAT_COOKIE);
672   PetscValidHeaderSpecific(mapping,IS_LTOGM_COOKIE);
673   if (x->mapping) {
674     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Mapping already set for matrix");
675   }
676 
677   x->mapping = mapping;
678   ierr = PetscObjectReference((PetscObject)mapping);CHKERRQ(ierr);
679   PetscFunctionReturn(0);
680 }
681 
682 #undef __FUNC__
683 #define __FUNC__ "MatSetLocalToGlobalMappingBlock"
684 /*@
685    MatSetLocalToGlobalMappingBlock - Sets a local-to-global numbering for use
686    by the routine MatSetValuesBlockedLocal() to allow users to insert matrix
687    entries using a local (per-processor) numbering.
688 
689    Not Collective
690 
691    Input Parameters:
692 +  x - the matrix
693 -  mapping - mapping created with ISLocalToGlobalMappingCreate() or
694              ISLocalToGlobalMappingCreateIS()
695 
696    Level: intermediate
697 
698 .keywords: matrix, set, values, local ordering
699 
700 .seealso:  MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesBlockedLocal(),
701            MatSetValuesBlocked(), MatSetValuesLocal()
702 @*/
703 int MatSetLocalToGlobalMappingBlock(Mat x,ISLocalToGlobalMapping mapping)
704 {
705   int ierr;
706   PetscFunctionBegin;
707   PetscValidHeaderSpecific(x,MAT_COOKIE);
708   PetscValidHeaderSpecific(mapping,IS_LTOGM_COOKIE);
709   if (x->bmapping) {
710     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Mapping already set for matrix");
711   }
712 
713   x->bmapping = mapping;
714   ierr = PetscObjectReference((PetscObject)mapping);CHKERRQ(ierr);
715   PetscFunctionReturn(0);
716 }
717 
718 #undef __FUNC__
719 #define __FUNC__ "MatSetValuesLocal"
720 /*@
721    MatSetValuesLocal - Inserts or adds values into certain locations of a matrix,
722    using a local ordering of the nodes.
723 
724    Not Collective
725 
726    Input Parameters:
727 +  x - the matrix
728 .  nrow, irow - number of rows and their local indices
729 .  ncol, icol - number of columns and their local indices
730 .  y -  a logically two-dimensional array of values
731 -  addv - either INSERT_VALUES or ADD_VALUES, where
732    ADD_VALUES adds values to any existing entries, and
733    INSERT_VALUES replaces existing entries with new values
734 
735    Notes:
736    Before calling MatSetValuesLocal(), the user must first set the
737    local-to-global mapping by calling MatSetLocalToGlobalMapping().
738 
739    Calls to MatSetValuesLocal() with the INSERT_VALUES and ADD_VALUES
740    options cannot be mixed without intervening calls to the assembly
741    routines.
742 
743    These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
744    MUST be called after all calls to MatSetValuesLocal() have been completed.
745 
746    Level: intermediate
747 
748 .keywords: matrix, set, values, local ordering
749 
750 .seealso:  MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetLocalToGlobalMapping()
751 @*/
752 int MatSetValuesLocal(Mat mat,int nrow,int *irow,int ncol,int *icol,Scalar *y,InsertMode addv)
753 {
754   int ierr,irowm[2048],icolm[2048];
755 
756   PetscFunctionBegin;
757   PetscValidHeaderSpecific(mat,MAT_COOKIE);
758   PetscValidIntPointer(irow);
759   PetscValidIntPointer(icol);
760   PetscValidScalarPointer(y);
761 
762   if (mat->insertmode == NOT_SET_VALUES) {
763     mat->insertmode = addv;
764   }
765 #if defined(PETSC_USE_BOPT_g)
766   else if (mat->insertmode != addv) {
767     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,1,"Cannot mix add values and insert values");
768   }
769   if (!mat->mapping) {
770     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Local to global never set with MatSetLocalToGlobalMapping()");
771   }
772   if (nrow > 2048 || ncol > 2048) {
773     SETERRQ2(PETSC_ERR_SUP,0,"Number column/row indices must be <= 2048: are %d %d",nrow,ncol);
774   }
775   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
776 #endif
777 
778   if (mat->assembled) {
779     mat->was_assembled = PETSC_TRUE;
780     mat->assembled     = PETSC_FALSE;
781   }
782   PLogEventBegin(MAT_SetValues,mat,0,0,0);
783   ierr = ISLocalToGlobalMappingApply(mat->mapping,nrow,irow,irowm);CHKERRQ(ierr);
784   ierr = ISLocalToGlobalMappingApply(mat->mapping,ncol,icol,icolm);CHKERRQ(ierr);
785   ierr = (*mat->ops->setvalues)(mat,nrow,irowm,ncol,icolm,y,addv);CHKERRQ(ierr);
786   PLogEventEnd(MAT_SetValues,mat,0,0,0);
787   PetscFunctionReturn(0);
788 }
789 
790 #undef __FUNC__
791 #define __FUNC__ "MatSetValuesBlockedLocal"
792 /*@
793    MatSetValuesBlockedLocal - Inserts or adds values into certain locations of a matrix,
794    using a local ordering of the nodes a block at a time.
795 
796    Not Collective
797 
798    Input Parameters:
799 +  x - the matrix
800 .  nrow, irow - number of rows and their local indices
801 .  ncol, icol - number of columns and their local indices
802 .  y -  a logically two-dimensional array of values
803 -  addv - either INSERT_VALUES or ADD_VALUES, where
804    ADD_VALUES adds values to any existing entries, and
805    INSERT_VALUES replaces existing entries with new values
806 
807    Notes:
808    Before calling MatSetValuesBlockedLocal(), the user must first set the
809    local-to-global mapping by calling MatSetLocalToGlobalMappingBlock(),
810    where the mapping MUST be set for matrix blocks, not for matrix elements.
811 
812    Calls to MatSetValuesBlockedLocal() with the INSERT_VALUES and ADD_VALUES
813    options cannot be mixed without intervening calls to the assembly
814    routines.
815 
816    These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
817    MUST be called after all calls to MatSetValuesBlockedLocal() have been completed.
818 
819    Level: intermediate
820 
821 .keywords: matrix, set, values, blocked, local
822 
823 .seealso:  MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesLocal(), MatSetLocalToGlobalMappingBlock(), MatSetValuesBlocked()
824 @*/
825 int MatSetValuesBlockedLocal(Mat mat,int nrow,int *irow,int ncol,int *icol,Scalar *y,InsertMode addv)
826 {
827   int ierr,irowm[2048],icolm[2048];
828 
829   PetscFunctionBegin;
830   PetscValidHeaderSpecific(mat,MAT_COOKIE);
831   PetscValidIntPointer(irow);
832   PetscValidIntPointer(icol);
833   PetscValidScalarPointer(y);
834   if (mat->insertmode == NOT_SET_VALUES) {
835     mat->insertmode = addv;
836   }
837 #if defined(PETSC_USE_BOPT_g)
838   else if (mat->insertmode != addv) {
839     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,1,"Cannot mix add values and insert values");
840   }
841   if (!mat->bmapping) {
842     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Local to global never set with MatSetLocalToGlobalMappingBlock()");
843   }
844   if (nrow > 2048 || ncol > 2048) {
845     SETERRQ2(PETSC_ERR_SUP,0,"Number column/row indices must be <= 2048: are %d %d",nrow,ncol);
846   }
847   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
848 #endif
849 
850   if (mat->assembled) {
851     mat->was_assembled = PETSC_TRUE;
852     mat->assembled     = PETSC_FALSE;
853   }
854   PLogEventBegin(MAT_SetValues,mat,0,0,0);
855   ierr = ISLocalToGlobalMappingApply(mat->bmapping,nrow,irow,irowm);CHKERRQ(ierr);
856   ierr = ISLocalToGlobalMappingApply(mat->bmapping,ncol,icol,icolm);CHKERRQ(ierr);
857   ierr = (*mat->ops->setvaluesblocked)(mat,nrow,irowm,ncol,icolm,y,addv);CHKERRQ(ierr);
858   PLogEventEnd(MAT_SetValues,mat,0,0,0);
859   PetscFunctionReturn(0);
860 }
861 
862 /* --------------------------------------------------------*/
863 #undef __FUNC__
864 #define __FUNC__ "MatMult"
865 /*@
866    MatMult - Computes the matrix-vector product, y = Ax.
867 
868    Collective on Mat and Vec
869 
870    Input Parameters:
871 +  mat - the matrix
872 -  x   - the vector to be multilplied
873 
874    Output Parameters:
875 .  y - the result
876 
877    Notes:
878    The vectors x and y cannot be the same.  I.e., one cannot
879    call MatMult(A,y,y).
880 
881    Level: beginner
882 
883 .keywords: matrix, multiply, matrix-vector product
884 
885 .seealso: MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
886 @*/
887 int MatMult(Mat mat,Vec x,Vec y)
888 {
889   int ierr;
890 
891   PetscFunctionBegin;
892   PetscValidHeaderSpecific(mat,MAT_COOKIE);
893   PetscValidHeaderSpecific(x,VEC_COOKIE);
894   PetscValidHeaderSpecific(y,VEC_COOKIE);
895   PetscCheckSameComm(mat,x);
896   PetscCheckSameComm(mat,y);
897   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
898   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
899   if (x == y) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"x and y must be different vectors");
900   if (mat->N != x->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec x: global dim %d %d",mat->N,x->N);
901   if (mat->M != y->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec y: global dim %d %d",mat->M,y->N);
902   if (mat->m != y->n) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec y: local dim %d %d",mat->m,y->n);
903 
904   PLogEventBegin(MAT_Mult,mat,x,y,0);
905   ierr = (*mat->ops->mult)(mat,x,y);CHKERRQ(ierr);
906   PLogEventEnd(MAT_Mult,mat,x,y,0);
907 
908   PetscFunctionReturn(0);
909 }
910 
911 #undef __FUNC__
912 #define __FUNC__ "MatMultTranspose"
913 /*@
914    MatMultTranspose - Computes matrix transpose times a vector.
915 
916    Collective on Mat and Vec
917 
918    Input Parameters:
919 +  mat - the matrix
920 -  x   - the vector to be multilplied
921 
922    Output Parameters:
923 .  y - the result
924 
925    Notes:
926    The vectors x and y cannot be the same.  I.e., one cannot
927    call MatMultTranspose(A,y,y).
928 
929    Level: beginner
930 
931 .keywords: matrix, multiply, matrix-vector product, transpose
932 
933 .seealso: MatMult(), MatMultAdd(), MatMultTransposeAdd()
934 @*/
935 int MatMultTranspose(Mat mat,Vec x,Vec y)
936 {
937   int ierr;
938 
939   PetscFunctionBegin;
940   PetscValidHeaderSpecific(mat,MAT_COOKIE);
941   PetscValidHeaderSpecific(x,VEC_COOKIE);
942   PetscValidHeaderSpecific(y,VEC_COOKIE);
943   PetscCheckSameComm(mat,x);
944   PetscCheckSameComm(mat,y);
945   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
946   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
947   if (x == y) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"x and y must be different vectors");
948   if (mat->M != x->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec x: global dim %d %d",mat->M,x->N);
949   if (mat->N != y->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec y: global dim %d %d",mat->N,y->N);
950 
951   PLogEventBegin(MAT_MultTranspose,mat,x,y,0);
952   ierr = (*mat->ops->multtranspose)(mat,x,y);CHKERRQ(ierr);
953   PLogEventEnd(MAT_MultTranspose,mat,x,y,0);
954   PetscFunctionReturn(0);
955 }
956 
957 #undef __FUNC__
958 #define __FUNC__ "MatMultAdd"
959 /*@
960     MatMultAdd -  Computes v3 = v2 + A * v1.
961 
962     Collective on Mat and Vec
963 
964     Input Parameters:
965 +   mat - the matrix
966 -   v1, v2 - the vectors
967 
968     Output Parameters:
969 .   v3 - the result
970 
971     Notes:
972     The vectors v1 and v3 cannot be the same.  I.e., one cannot
973     call MatMultAdd(A,v1,v2,v1).
974 
975     Level: beginner
976 
977 .keywords: matrix, multiply, matrix-vector product, add
978 
979 .seealso: MatMultTranspose(), MatMult(), MatMultTransposeAdd()
980 @*/
981 int MatMultAdd(Mat mat,Vec v1,Vec v2,Vec v3)
982 {
983   int ierr;
984 
985   PetscFunctionBegin;
986   PetscValidHeaderSpecific(mat,MAT_COOKIE);
987   PetscValidHeaderSpecific(v1,VEC_COOKIE);
988   PetscValidHeaderSpecific(v2,VEC_COOKIE);
989   PetscValidHeaderSpecific(v3,VEC_COOKIE);
990   PetscCheckSameComm(mat,v1);
991   PetscCheckSameComm(mat,v2);
992   PetscCheckSameComm(mat,v3);
993   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
994   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
995   if (mat->N != v1->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec v1: global dim %d %d",mat->N,v1->N);
996   if (mat->M != v2->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec v2: global dim %d %d",mat->M,v2->N);
997   if (mat->M != v3->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec v3: global dim %d %d",mat->M,v3->N);
998   if (mat->m != v3->n) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec v3: local dim %d %d",mat->m,v3->n);
999   if (mat->m != v2->n) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec v2: local dim %d %d",mat->m,v2->n);
1000   if (v1 == v3) SETERRQ(PETSC_ERR_ARG_IDN,0,"v1 and v3 must be different vectors");
1001 
1002   PLogEventBegin(MAT_MultAdd,mat,v1,v2,v3);
1003   ierr = (*mat->ops->multadd)(mat,v1,v2,v3);CHKERRQ(ierr);
1004   PLogEventEnd(MAT_MultAdd,mat,v1,v2,v3);
1005   PetscFunctionReturn(0);
1006 }
1007 
1008 #undef __FUNC__
1009 #define __FUNC__ "MatMultTransposeAdd"
1010 /*@
1011    MatMultTransposeAdd - Computes v3 = v2 + A' * v1.
1012 
1013    Collective on Mat and Vec
1014 
1015    Input Parameters:
1016 +  mat - the matrix
1017 -  v1, v2 - the vectors
1018 
1019    Output Parameters:
1020 .  v3 - the result
1021 
1022    Notes:
1023    The vectors v1 and v3 cannot be the same.  I.e., one cannot
1024    call MatMultTransposeAdd(A,v1,v2,v1).
1025 
1026    Level: beginner
1027 
1028 .keywords: matrix, multiply, matrix-vector product, transpose, add
1029 
1030 .seealso: MatMultTranspose(), MatMultAdd(), MatMult()
1031 @*/
1032 int MatMultTransposeAdd(Mat mat,Vec v1,Vec v2,Vec v3)
1033 {
1034   int ierr;
1035 
1036   PetscFunctionBegin;
1037   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1038   PetscValidHeaderSpecific(v1,VEC_COOKIE);
1039   PetscValidHeaderSpecific(v2,VEC_COOKIE);
1040   PetscValidHeaderSpecific(v3,VEC_COOKIE);
1041   PetscCheckSameComm(mat,v1);
1042   PetscCheckSameComm(mat,v2);
1043   PetscCheckSameComm(mat,v3);
1044   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
1045   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
1046   if (!mat->ops->multtransposeadd) SETERRQ(PETSC_ERR_SUP,0,"");
1047   if (v1 == v3) SETERRQ(PETSC_ERR_ARG_IDN,0,"v1 and v3 must be different vectors");
1048   if (mat->M != v1->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec v1: global dim %d %d",mat->M,v1->N);
1049   if (mat->N != v2->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec v2: global dim %d %d",mat->N,v2->N);
1050   if (mat->N != v3->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec v3: global dim %d %d",mat->N,v3->N);
1051 
1052   PLogEventBegin(MAT_MultTransposeAdd,mat,v1,v2,v3);
1053   ierr = (*mat->ops->multtransposeadd)(mat,v1,v2,v3);CHKERRQ(ierr);
1054   PLogEventEnd(MAT_MultTransposeAdd,mat,v1,v2,v3);
1055   PetscFunctionReturn(0);
1056 }
1057 /* ------------------------------------------------------------*/
1058 #undef __FUNC__
1059 #define __FUNC__ "MatGetInfo"
1060 /*@C
1061    MatGetInfo - Returns information about matrix storage (number of
1062    nonzeros, memory, etc.).
1063 
1064    Collective on Mat if MAT_GLOBAL_MAX or MAT_GLOBAL_SUM is used
1065    as the flag
1066 
1067    Input Parameters:
1068 .  mat - the matrix
1069 
1070    Output Parameters:
1071 +  flag - flag indicating the type of parameters to be returned
1072    (MAT_LOCAL - local matrix, MAT_GLOBAL_MAX - maximum over all processors,
1073    MAT_GLOBAL_SUM - sum over all processors)
1074 -  info - matrix information context
1075 
1076    Notes:
1077    The MatInfo context contains a variety of matrix data, including
1078    number of nonzeros allocated and used, number of mallocs during
1079    matrix assembly, etc.  Additional information for factored matrices
1080    is provided (such as the fill ratio, number of mallocs during
1081    factorization, etc.).  Much of this info is printed to STDOUT
1082    when using the runtime options
1083 $       -log_info -mat_view_info
1084 
1085    Example for C/C++ Users:
1086    See the file ${PETSC_DIR}/include/mat.h for a complete list of
1087    data within the MatInfo context.  For example,
1088 .vb
1089       MatInfo info;
1090       Mat     A;
1091       double  mal, nz_a, nz_u;
1092 
1093       MatGetInfo(A,MAT_LOCAL,&info);
1094       mal  = info.mallocs;
1095       nz_a = info.nz_allocated;
1096 .ve
1097 
1098    Example for Fortran Users:
1099    Fortran users should declare info as a double precision
1100    array of dimension MAT_INFO_SIZE, and then extract the parameters
1101    of interest.  See the file ${PETSC_DIR}/include/finclude/mat.h
1102    a complete list of parameter names.
1103 .vb
1104       double  precision info(MAT_INFO_SIZE)
1105       double  precision mal, nz_a
1106       Mat     A
1107       integer ierr
1108 
1109       call MatGetInfo(A,MAT_LOCAL,info,ierr)
1110       mal = info(MAT_INFO_MALLOCS)
1111       nz_a = info(MAT_INFO_NZ_ALLOCATED)
1112 .ve
1113 
1114     Level: intermediate
1115 
1116 .keywords: matrix, get, info, storage, nonzeros, memory, fill
1117 @*/
1118 int MatGetInfo(Mat mat,MatInfoType flag,MatInfo *info)
1119 {
1120   int ierr;
1121 
1122   PetscFunctionBegin;
1123   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1124   PetscValidPointer(info);
1125   if (!mat->ops->getinfo) SETERRQ(PETSC_ERR_SUP,0,"");
1126   ierr = (*mat->ops->getinfo)(mat,flag,info);CHKERRQ(ierr);
1127   PetscFunctionReturn(0);
1128 }
1129 
1130 /* ----------------------------------------------------------*/
1131 #undef __FUNC__
1132 #define __FUNC__ "MatILUDTFactor"
1133 /*@C
1134    MatILUDTFactor - Performs a drop tolerance ILU factorization.
1135 
1136    Collective on Mat
1137 
1138    Input Parameters:
1139 +  mat - the matrix
1140 .  info - information about the factorization to be done
1141 .  row - row permutation
1142 -  col - column permutation
1143 
1144    Output Parameters:
1145 .  fact - the factored matrix
1146 
1147    Level: developer
1148 
1149    Notes:
1150    Most users should employ the simplified SLES interface for linear solvers
1151    instead of working directly with matrix algebra routines such as this.
1152    See, e.g., SLESCreate().
1153 
1154    This is currently only supported for the SeqAIJ matrix format using code
1155    from Yousef Saad's SPARSEKIT2 package. That code is copyright by Yousef
1156    Saad with the GNU copyright.
1157 
1158 .keywords: matrix, factor, LU, in-place
1159 
1160 .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor()
1161 @*/
1162 int MatILUDTFactor(Mat mat,MatILUInfo *info,IS row,IS col,Mat *fact)
1163 {
1164   int ierr;
1165 
1166   PetscFunctionBegin;
1167   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1168   PetscValidPointer(fact);
1169   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
1170   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
1171   if (!mat->ops->iludtfactor) SETERRQ(PETSC_ERR_SUP,0,"");
1172 
1173   PLogEventBegin(MAT_ILUFactor,mat,row,col,0);
1174   ierr = (*mat->ops->iludtfactor)(mat,info,row,col,fact);CHKERRQ(ierr);
1175   PLogEventEnd(MAT_ILUFactor,mat,row,col,0);
1176 
1177   PetscFunctionReturn(0);
1178 }
1179 
1180 #undef __FUNC__
1181 #define __FUNC__ "MatLUFactor"
1182 /*@
1183    MatLUFactor - Performs in-place LU factorization of matrix.
1184 
1185    Collective on Mat
1186 
1187    Input Parameters:
1188 +  mat - the matrix
1189 .  row - row permutation
1190 .  col - column permutation
1191 -  f - expected fill as ratio of original fill.
1192 
1193    Notes:
1194    Most users should employ the simplified SLES interface for linear solvers
1195    instead of working directly with matrix algebra routines such as this.
1196    See, e.g., SLESCreate().
1197 
1198    This changes the state of the matrix to a factored matrix; it cannot be used
1199    for example with MatSetValues() unless one first calls MatSetUnfactored().
1200 
1201    Level: developer
1202 
1203 .keywords: matrix, factor, LU, in-place
1204 
1205 .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor(),
1206           MatGetOrdering(), MatSetUnfactored()
1207 
1208 @*/
1209 int MatLUFactor(Mat mat,IS row,IS col,PetscReal f)
1210 {
1211   int ierr;
1212 
1213   PetscFunctionBegin;
1214   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1215   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
1216   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
1217   if (!mat->ops->lufactor) SETERRQ(PETSC_ERR_SUP,0,"");
1218 
1219   PLogEventBegin(MAT_LUFactor,mat,row,col,0);
1220   ierr = (*mat->ops->lufactor)(mat,row,col,f);CHKERRQ(ierr);
1221   PLogEventEnd(MAT_LUFactor,mat,row,col,0);
1222   PetscFunctionReturn(0);
1223 }
1224 
1225 #undef __FUNC__
1226 #define __FUNC__ "MatILUFactor"
1227 /*@
1228    MatILUFactor - Performs in-place ILU factorization of matrix.
1229 
1230    Collective on Mat
1231 
1232    Input Parameters:
1233 +  mat - the matrix
1234 .  row - row permutation
1235 .  col - column permutation
1236 -  info - structure containing
1237 $      levels - number of levels of fill.
1238 $      expected fill - as ratio of original fill.
1239 $      1 or 0 - indicating force fill on diagonal (improves robustness for matrices
1240                 missing diagonal entries)
1241 
1242    Notes:
1243    Probably really in-place only when level of fill is zero, otherwise allocates
1244    new space to store factored matrix and deletes previous memory.
1245 
1246    Most users should employ the simplified SLES interface for linear solvers
1247    instead of working directly with matrix algebra routines such as this.
1248    See, e.g., SLESCreate().
1249 
1250    Level: developer
1251 
1252 .keywords: matrix, factor, ILU, in-place
1253 
1254 .seealso: MatILUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor()
1255 @*/
1256 int MatILUFactor(Mat mat,IS row,IS col,MatILUInfo *info)
1257 {
1258   int ierr;
1259 
1260   PetscFunctionBegin;
1261   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1262   if (mat->M != mat->N) SETERRQ(PETSC_ERR_ARG_WRONG,0,"matrix must be square");
1263   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
1264   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
1265   if (!mat->ops->ilufactor) SETERRQ(PETSC_ERR_SUP,0,"");
1266 
1267   PLogEventBegin(MAT_ILUFactor,mat,row,col,0);
1268   ierr = (*mat->ops->ilufactor)(mat,row,col,info);CHKERRQ(ierr);
1269   PLogEventEnd(MAT_ILUFactor,mat,row,col,0);
1270   PetscFunctionReturn(0);
1271 }
1272 
1273 #undef __FUNC__
1274 #define __FUNC__ "MatLUFactorSymbolic"
1275 /*@
1276    MatLUFactorSymbolic - Performs symbolic LU factorization of matrix.
1277    Call this routine before calling MatLUFactorNumeric().
1278 
1279    Collective on Mat
1280 
1281    Input Parameters:
1282 +  mat - the matrix
1283 .  row, col - row and column permutations
1284 -  f - expected fill as ratio of the original number of nonzeros,
1285        for example 3.0; choosing this parameter well can result in
1286        more efficient use of time and space. Run with the option -log_info
1287        to determine an optimal value to use
1288 
1289    Output Parameter:
1290 .  fact - new matrix that has been symbolically factored
1291 
1292    Notes:
1293    See the users manual for additional information about
1294    choosing the fill factor for better efficiency.
1295 
1296    Most users should employ the simplified SLES interface for linear solvers
1297    instead of working directly with matrix algebra routines such as this.
1298    See, e.g., SLESCreate().
1299 
1300    Level: developer
1301 
1302 .keywords: matrix, factor, LU, symbolic, fill
1303 
1304 .seealso: MatLUFactor(), MatLUFactorNumeric(), MatCholeskyFactor()
1305 @*/
1306 int MatLUFactorSymbolic(Mat mat,IS row,IS col,PetscReal f,Mat *fact)
1307 {
1308   int ierr;
1309 
1310   PetscFunctionBegin;
1311   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1312   PetscValidPointer(fact);
1313   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
1314   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
1315   if (!mat->ops->lufactorsymbolic) SETERRQ(PETSC_ERR_SUP,0,"");
1316 
1317   PLogEventBegin(MAT_LUFactorSymbolic,mat,row,col,0);
1318   ierr = (*mat->ops->lufactorsymbolic)(mat,row,col,f,fact);CHKERRQ(ierr);
1319   PLogEventEnd(MAT_LUFactorSymbolic,mat,row,col,0);
1320   PetscFunctionReturn(0);
1321 }
1322 
1323 #undef __FUNC__
1324 #define __FUNC__ "MatLUFactorNumeric"
1325 /*@
1326    MatLUFactorNumeric - Performs numeric LU factorization of a matrix.
1327    Call this routine after first calling MatLUFactorSymbolic().
1328 
1329    Collective on Mat
1330 
1331    Input Parameters:
1332 +  mat - the matrix
1333 -  row, col - row and  column permutations
1334 
1335    Output Parameters:
1336 .  fact - symbolically factored matrix that must have been generated
1337           by MatLUFactorSymbolic()
1338 
1339    Notes:
1340    See MatLUFactor() for in-place factorization.  See
1341    MatCholeskyFactorNumeric() for the symmetric, positive definite case.
1342 
1343    Most users should employ the simplified SLES interface for linear solvers
1344    instead of working directly with matrix algebra routines such as this.
1345    See, e.g., SLESCreate().
1346 
1347    Level: developer
1348 
1349 .keywords: matrix, factor, LU, numeric
1350 
1351 .seealso: MatLUFactorSymbolic(), MatLUFactor(), MatCholeskyFactor()
1352 @*/
1353 int MatLUFactorNumeric(Mat mat,Mat *fact)
1354 {
1355   int        ierr;
1356   PetscTruth flg;
1357 
1358   PetscFunctionBegin;
1359   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1360   PetscValidPointer(fact);
1361   PetscValidHeaderSpecific(*fact,MAT_COOKIE);
1362   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
1363   if (mat->M != (*fact)->M || mat->N != (*fact)->N) {
1364     SETERRQ4(PETSC_ERR_ARG_SIZ,0,"Mat mat,Mat *fact: global dimensions are different %d should = %d %d should = %d",
1365             mat->M,(*fact)->M,mat->N,(*fact)->N);
1366   }
1367   if (!(*fact)->ops->lufactornumeric) SETERRQ(PETSC_ERR_SUP,0,"");
1368 
1369   PLogEventBegin(MAT_LUFactorNumeric,mat,*fact,0,0);
1370   ierr = (*(*fact)->ops->lufactornumeric)(mat,fact);CHKERRQ(ierr);
1371   PLogEventEnd(MAT_LUFactorNumeric,mat,*fact,0,0);
1372   ierr = OptionsHasName(PETSC_NULL,"-mat_view_draw",&flg);CHKERRQ(ierr);
1373   if (flg) {
1374     ierr = OptionsHasName(PETSC_NULL,"-mat_view_contour",&flg);CHKERRQ(ierr);
1375     if (flg) {
1376       ViewerPushFormat(VIEWER_DRAW_(mat->comm),VIEWER_FORMAT_DRAW_CONTOUR,0);CHKERRQ(ierr);
1377     }
1378     ierr = MatView(*fact,VIEWER_DRAW_(mat->comm));CHKERRQ(ierr);
1379     ierr = ViewerFlush(VIEWER_DRAW_(mat->comm));CHKERRQ(ierr);
1380     if (flg) {
1381       ViewerPopFormat(VIEWER_DRAW_(mat->comm));CHKERRQ(ierr);
1382     }
1383   }
1384   PetscFunctionReturn(0);
1385 }
1386 
1387 #undef __FUNC__
1388 #define __FUNC__ "MatCholeskyFactor"
1389 /*@
1390    MatCholeskyFactor - Performs in-place Cholesky factorization of a
1391    symmetric matrix.
1392 
1393    Collective on Mat
1394 
1395    Input Parameters:
1396 +  mat - the matrix
1397 .  perm - row and column permutations
1398 -  f - expected fill as ratio of original fill
1399 
1400    Notes:
1401    See MatLUFactor() for the nonsymmetric case.  See also
1402    MatCholeskyFactorSymbolic(), and MatCholeskyFactorNumeric().
1403 
1404    Most users should employ the simplified SLES interface for linear solvers
1405    instead of working directly with matrix algebra routines such as this.
1406    See, e.g., SLESCreate().
1407 
1408    Level: developer
1409 
1410 .keywords: matrix, factor, in-place, Cholesky
1411 
1412 .seealso: MatLUFactor(), MatCholeskyFactorSymbolic(), MatCholeskyFactorNumeric()
1413           MatGetOrdering()
1414 
1415 @*/
1416 int MatCholeskyFactor(Mat mat,IS perm,PetscReal f)
1417 {
1418   int ierr;
1419 
1420   PetscFunctionBegin;
1421   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1422   if (mat->M != mat->N) SETERRQ(PETSC_ERR_ARG_WRONG,0,"Matrix must be square");
1423   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
1424   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
1425   if (!mat->ops->choleskyfactor) SETERRQ(PETSC_ERR_SUP,0,"");
1426 
1427   PLogEventBegin(MAT_CholeskyFactor,mat,perm,0,0);
1428   ierr = (*mat->ops->choleskyfactor)(mat,perm,f);CHKERRQ(ierr);
1429   PLogEventEnd(MAT_CholeskyFactor,mat,perm,0,0);
1430   PetscFunctionReturn(0);
1431 }
1432 
1433 #undef __FUNC__
1434 #define __FUNC__ "MatCholeskyFactorSymbolic"
1435 /*@
1436    MatCholeskyFactorSymbolic - Performs symbolic Cholesky factorization
1437    of a symmetric matrix.
1438 
1439    Collective on Mat
1440 
1441    Input Parameters:
1442 +  mat - the matrix
1443 .  perm - row and column permutations
1444 -  f - expected fill as ratio of original
1445 
1446    Output Parameter:
1447 .  fact - the factored matrix
1448 
1449    Notes:
1450    See MatLUFactorSymbolic() for the nonsymmetric case.  See also
1451    MatCholeskyFactor() and MatCholeskyFactorNumeric().
1452 
1453    Most users should employ the simplified SLES interface for linear solvers
1454    instead of working directly with matrix algebra routines such as this.
1455    See, e.g., SLESCreate().
1456 
1457    Level: developer
1458 
1459 .keywords: matrix, factor, factorization, symbolic, Cholesky
1460 
1461 .seealso: MatLUFactorSymbolic(), MatCholeskyFactor(), MatCholeskyFactorNumeric()
1462           MatGetOrdering()
1463 
1464 @*/
1465 int MatCholeskyFactorSymbolic(Mat mat,IS perm,PetscReal f,Mat *fact)
1466 {
1467   int ierr;
1468 
1469   PetscFunctionBegin;
1470   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1471   PetscValidPointer(fact);
1472   if (mat->M != mat->N) SETERRQ(PETSC_ERR_ARG_WRONG,0,"Matrix must be square");
1473   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
1474   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
1475   if (!mat->ops->choleskyfactorsymbolic) SETERRQ(PETSC_ERR_SUP,0,"");
1476 
1477   PLogEventBegin(MAT_CholeskyFactorSymbolic,mat,perm,0,0);
1478   ierr = (*mat->ops->choleskyfactorsymbolic)(mat,perm,f,fact);CHKERRQ(ierr);
1479   PLogEventEnd(MAT_CholeskyFactorSymbolic,mat,perm,0,0);
1480   PetscFunctionReturn(0);
1481 }
1482 
1483 #undef __FUNC__
1484 #define __FUNC__ "MatCholeskyFactorNumeric"
1485 /*@
1486    MatCholeskyFactorNumeric - Performs numeric Cholesky factorization
1487    of a symmetric matrix. Call this routine after first calling
1488    MatCholeskyFactorSymbolic().
1489 
1490    Collective on Mat
1491 
1492    Input Parameter:
1493 .  mat - the initial matrix
1494 
1495    Output Parameter:
1496 .  fact - the factored matrix
1497 
1498    Notes:
1499    Most users should employ the simplified SLES interface for linear solvers
1500    instead of working directly with matrix algebra routines such as this.
1501    See, e.g., SLESCreate().
1502 
1503    Level: developer
1504 
1505 .keywords: matrix, factor, numeric, Cholesky
1506 
1507 .seealso: MatCholeskyFactorSymbolic(), MatCholeskyFactor(), MatLUFactorNumeric()
1508 @*/
1509 int MatCholeskyFactorNumeric(Mat mat,Mat *fact)
1510 {
1511   int ierr;
1512 
1513   PetscFunctionBegin;
1514   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1515   PetscValidPointer(fact);
1516   if (!mat->ops->choleskyfactornumeric) SETERRQ(PETSC_ERR_SUP,0,"");
1517   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
1518   if (mat->M != (*fact)->M || mat->N != (*fact)->N) {
1519     SETERRQ4(PETSC_ERR_ARG_SIZ,0,"Mat mat,Mat *fact: global dim %d should = %d %d should = %d",
1520             mat->M,(*fact)->M,mat->N,(*fact)->N);
1521   }
1522 
1523   PLogEventBegin(MAT_CholeskyFactorNumeric,mat,*fact,0,0);
1524   ierr = (*mat->ops->choleskyfactornumeric)(mat,fact);CHKERRQ(ierr);
1525   PLogEventEnd(MAT_CholeskyFactorNumeric,mat,*fact,0,0);
1526   PetscFunctionReturn(0);
1527 }
1528 
1529 /* ----------------------------------------------------------------*/
1530 #undef __FUNC__
1531 #define __FUNC__ "MatSolve"
1532 /*@
1533    MatSolve - Solves A x = b, given a factored matrix.
1534 
1535    Collective on Mat and Vec
1536 
1537    Input Parameters:
1538 +  mat - the factored matrix
1539 -  b - the right-hand-side vector
1540 
1541    Output Parameter:
1542 .  x - the result vector
1543 
1544    Notes:
1545    The vectors b and x cannot be the same.  I.e., one cannot
1546    call MatSolve(A,x,x).
1547 
1548    Notes:
1549    Most users should employ the simplified SLES interface for linear solvers
1550    instead of working directly with matrix algebra routines such as this.
1551    See, e.g., SLESCreate().
1552 
1553    Level: developer
1554 
1555 .keywords: matrix, linear system, solve, LU, Cholesky, triangular solve
1556 
1557 .seealso: MatSolveAdd(), MatSolveTranspose(), MatSolveTransposeAdd()
1558 @*/
1559 int MatSolve(Mat mat,Vec b,Vec x)
1560 {
1561   int ierr;
1562 
1563   PetscFunctionBegin;
1564   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1565   PetscValidHeaderSpecific(b,VEC_COOKIE);
1566   PetscValidHeaderSpecific(x,VEC_COOKIE);
1567   PetscCheckSameComm(mat,b);
1568   PetscCheckSameComm(mat,x);
1569   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,0,"x and b must be different vectors");
1570   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Unfactored matrix");
1571   if (mat->N != x->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec x: global dim %d %d",mat->N,x->N);
1572   if (mat->M != b->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec b: global dim %d %d",mat->M,b->N);
1573   if (mat->m != b->n) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec b: local dim %d %d",mat->m,b->n);
1574   if (mat->M == 0 && mat->N == 0) PetscFunctionReturn(0);
1575 
1576   if (!mat->ops->solve) SETERRQ(PETSC_ERR_SUP,0,"");
1577   PLogEventBegin(MAT_Solve,mat,b,x,0);
1578   ierr = (*mat->ops->solve)(mat,b,x);CHKERRQ(ierr);
1579   PLogEventEnd(MAT_Solve,mat,b,x,0);
1580   PetscFunctionReturn(0);
1581 }
1582 
1583 #undef __FUNC__
1584 #define __FUNC__ "MatForwardSolve"
1585 /* @
1586    MatForwardSolve - Solves L x = b, given a factored matrix, A = LU.
1587 
1588    Collective on Mat and Vec
1589 
1590    Input Parameters:
1591 +  mat - the factored matrix
1592 -  b - the right-hand-side vector
1593 
1594    Output Parameter:
1595 .  x - the result vector
1596 
1597    Notes:
1598    MatSolve() should be used for most applications, as it performs
1599    a forward solve followed by a backward solve.
1600 
1601    The vectors b and x cannot be the same.  I.e., one cannot
1602    call MatForwardSolve(A,x,x).
1603 
1604    Most users should employ the simplified SLES interface for linear solvers
1605    instead of working directly with matrix algebra routines such as this.
1606    See, e.g., SLESCreate().
1607 
1608    Level: developer
1609 
1610 .keywords: matrix, forward, LU, Cholesky, triangular solve
1611 
1612 .seealso: MatSolve(), MatBackwardSolve()
1613 @ */
1614 int MatForwardSolve(Mat mat,Vec b,Vec x)
1615 {
1616   int ierr;
1617 
1618   PetscFunctionBegin;
1619   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1620   PetscValidHeaderSpecific(b,VEC_COOKIE);
1621   PetscValidHeaderSpecific(x,VEC_COOKIE);
1622   PetscCheckSameComm(mat,b);
1623   PetscCheckSameComm(mat,x);
1624   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,0,"x and b must be different vectors");
1625   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Unfactored matrix");
1626   if (!mat->ops->forwardsolve) SETERRQ(PETSC_ERR_SUP,0,"");
1627   if (mat->N != x->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec x: global dim %d %d",mat->N,x->N);
1628   if (mat->M != b->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec b: global dim %d %d",mat->M,b->N);
1629   if (mat->m != b->n) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec b: local dim %d %d",mat->m,b->n);
1630 
1631   PLogEventBegin(MAT_ForwardSolve,mat,b,x,0);
1632   ierr = (*mat->ops->forwardsolve)(mat,b,x);CHKERRQ(ierr);
1633   PLogEventEnd(MAT_ForwardSolve,mat,b,x,0);
1634   PetscFunctionReturn(0);
1635 }
1636 
1637 #undef __FUNC__
1638 #define __FUNC__ "MatBackwardSolve"
1639 /* @
1640    MatBackwardSolve - Solves U x = b, given a factored matrix, A = LU.
1641 
1642    Collective on Mat and Vec
1643 
1644    Input Parameters:
1645 +  mat - the factored matrix
1646 -  b - the right-hand-side vector
1647 
1648    Output Parameter:
1649 .  x - the result vector
1650 
1651    Notes:
1652    MatSolve() should be used for most applications, as it performs
1653    a forward solve followed by a backward solve.
1654 
1655    The vectors b and x cannot be the same.  I.e., one cannot
1656    call MatBackwardSolve(A,x,x).
1657 
1658    Most users should employ the simplified SLES interface for linear solvers
1659    instead of working directly with matrix algebra routines such as this.
1660    See, e.g., SLESCreate().
1661 
1662    Level: developer
1663 
1664 .keywords: matrix, backward, LU, Cholesky, triangular solve
1665 
1666 .seealso: MatSolve(), MatForwardSolve()
1667 @ */
1668 int MatBackwardSolve(Mat mat,Vec b,Vec x)
1669 {
1670   int ierr;
1671 
1672   PetscFunctionBegin;
1673   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1674   PetscValidHeaderSpecific(b,VEC_COOKIE);
1675   PetscValidHeaderSpecific(x,VEC_COOKIE);
1676   PetscCheckSameComm(mat,b);
1677   PetscCheckSameComm(mat,x);
1678   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,0,"x and b must be different vectors");
1679   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Unfactored matrix");
1680   if (!mat->ops->backwardsolve) SETERRQ(PETSC_ERR_SUP,0,"");
1681   if (mat->N != x->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec x: global dim %d %d",mat->N,x->N);
1682   if (mat->M != b->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec b: global dim %d %d",mat->M,b->N);
1683   if (mat->m != b->n) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec b: local dim %d %d",mat->m,b->n);
1684 
1685   PLogEventBegin(MAT_BackwardSolve,mat,b,x,0);
1686   ierr = (*mat->ops->backwardsolve)(mat,b,x);CHKERRQ(ierr);
1687   PLogEventEnd(MAT_BackwardSolve,mat,b,x,0);
1688   PetscFunctionReturn(0);
1689 }
1690 
1691 #undef __FUNC__
1692 #define __FUNC__ "MatSolveAdd"
1693 /*@
1694    MatSolveAdd - Computes x = y + inv(A)*b, given a factored matrix.
1695 
1696    Collective on Mat and Vec
1697 
1698    Input Parameters:
1699 +  mat - the factored matrix
1700 .  b - the right-hand-side vector
1701 -  y - the vector to be added to
1702 
1703    Output Parameter:
1704 .  x - the result vector
1705 
1706    Notes:
1707    The vectors b and x cannot be the same.  I.e., one cannot
1708    call MatSolveAdd(A,x,y,x).
1709 
1710    Most users should employ the simplified SLES interface for linear solvers
1711    instead of working directly with matrix algebra routines such as this.
1712    See, e.g., SLESCreate().
1713 
1714    Level: developer
1715 
1716 .keywords: matrix, linear system, solve, LU, Cholesky, add
1717 
1718 .seealso: MatSolve(), MatSolveTranspose(), MatSolveTransposeAdd()
1719 @*/
1720 int MatSolveAdd(Mat mat,Vec b,Vec y,Vec x)
1721 {
1722   Scalar one = 1.0;
1723   Vec    tmp;
1724   int    ierr;
1725 
1726   PetscFunctionBegin;
1727   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1728   PetscValidHeaderSpecific(y,VEC_COOKIE);
1729   PetscValidHeaderSpecific(b,VEC_COOKIE);
1730   PetscValidHeaderSpecific(x,VEC_COOKIE);
1731   PetscCheckSameComm(mat,b);
1732   PetscCheckSameComm(mat,y);
1733   PetscCheckSameComm(mat,x);
1734   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,0,"x and b must be different vectors");
1735   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Unfactored matrix");
1736   if (mat->N != x->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec x: global dim %d %d",mat->N,x->N);
1737   if (mat->M != b->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec b: global dim %d %d",mat->M,b->N);
1738   if (mat->M != y->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec y: global dim %d %d",mat->M,y->N);
1739   if (mat->m != b->n) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec b: local dim %d %d",mat->m,b->n);
1740   if (x->n != y->n) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Vec x,Vec y: local dim %d %d",x->n,y->n);
1741 
1742   PLogEventBegin(MAT_SolveAdd,mat,b,x,y);
1743   if (mat->ops->solveadd)  {
1744     ierr = (*mat->ops->solveadd)(mat,b,y,x);CHKERRQ(ierr);
1745   } else {
1746     /* do the solve then the add manually */
1747     if (x != y) {
1748       ierr = MatSolve(mat,b,x);CHKERRQ(ierr);
1749       ierr = VecAXPY(&one,y,x);CHKERRQ(ierr);
1750     } else {
1751       ierr = VecDuplicate(x,&tmp);CHKERRQ(ierr);
1752       PLogObjectParent(mat,tmp);
1753       ierr = VecCopy(x,tmp);CHKERRQ(ierr);
1754       ierr = MatSolve(mat,b,x);CHKERRQ(ierr);
1755       ierr = VecAXPY(&one,tmp,x);CHKERRQ(ierr);
1756       ierr = VecDestroy(tmp);CHKERRQ(ierr);
1757     }
1758   }
1759   PLogEventEnd(MAT_SolveAdd,mat,b,x,y);
1760   PetscFunctionReturn(0);
1761 }
1762 
1763 #undef __FUNC__
1764 #define __FUNC__ "MatSolveTranspose"
1765 /*@
1766    MatSolveTranspose - Solves A' x = b, given a factored matrix.
1767 
1768    Collective on Mat and Vec
1769 
1770    Input Parameters:
1771 +  mat - the factored matrix
1772 -  b - the right-hand-side vector
1773 
1774    Output Parameter:
1775 .  x - the result vector
1776 
1777    Notes:
1778    The vectors b and x cannot be the same.  I.e., one cannot
1779    call MatSolveTranspose(A,x,x).
1780 
1781    Most users should employ the simplified SLES interface for linear solvers
1782    instead of working directly with matrix algebra routines such as this.
1783    See, e.g., SLESCreate().
1784 
1785    Level: developer
1786 
1787 .keywords: matrix, linear system, solve, LU, Cholesky, transpose
1788 
1789 .seealso: MatSolve(), MatSolveAdd(), MatSolveTransposeAdd()
1790 @*/
1791 int MatSolveTranspose(Mat mat,Vec b,Vec x)
1792 {
1793   int ierr;
1794 
1795   PetscFunctionBegin;
1796   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1797   PetscValidHeaderSpecific(b,VEC_COOKIE);
1798   PetscValidHeaderSpecific(x,VEC_COOKIE);
1799   PetscCheckSameComm(mat,b);
1800   PetscCheckSameComm(mat,x);
1801   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Unfactored matrix");
1802   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,0,"x and b must be different vectors");
1803   if (!mat->ops->solvetranspose) SETERRQ(PETSC_ERR_SUP,0,"");
1804   if (mat->M != x->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec x: global dim %d %d",mat->M,x->N);
1805   if (mat->N != b->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec b: global dim %d %d",mat->N,b->N);
1806 
1807   PLogEventBegin(MAT_SolveTranspose,mat,b,x,0);
1808   ierr = (*mat->ops->solvetranspose)(mat,b,x);CHKERRQ(ierr);
1809   PLogEventEnd(MAT_SolveTranspose,mat,b,x,0);
1810   PetscFunctionReturn(0);
1811 }
1812 
1813 #undef __FUNC__
1814 #define __FUNC__ "MatSolveTransposeAdd"
1815 /*@
1816    MatSolveTransposeAdd - Computes x = y + inv(Transpose(A)) b, given a
1817                       factored matrix.
1818 
1819    Collective on Mat and Vec
1820 
1821    Input Parameters:
1822 +  mat - the factored matrix
1823 .  b - the right-hand-side vector
1824 -  y - the vector to be added to
1825 
1826    Output Parameter:
1827 .  x - the result vector
1828 
1829    Notes:
1830    The vectors b and x cannot be the same.  I.e., one cannot
1831    call MatSolveTransposeAdd(A,x,y,x).
1832 
1833    Most users should employ the simplified SLES interface for linear solvers
1834    instead of working directly with matrix algebra routines such as this.
1835    See, e.g., SLESCreate().
1836 
1837    Level: developer
1838 
1839 .keywords: matrix, linear system, solve, LU, Cholesky, transpose, add
1840 
1841 .seealso: MatSolve(), MatSolveAdd(), MatSolveTranspose()
1842 @*/
1843 int MatSolveTransposeAdd(Mat mat,Vec b,Vec y,Vec x)
1844 {
1845   Scalar one = 1.0;
1846   int    ierr;
1847   Vec    tmp;
1848 
1849   PetscFunctionBegin;
1850   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1851   PetscValidHeaderSpecific(y,VEC_COOKIE);
1852   PetscValidHeaderSpecific(b,VEC_COOKIE);
1853   PetscValidHeaderSpecific(x,VEC_COOKIE);
1854   PetscCheckSameComm(mat,b);
1855   PetscCheckSameComm(mat,y);
1856   PetscCheckSameComm(mat,x);
1857   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,0,"x and b must be different vectors");
1858   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Unfactored matrix");
1859   if (mat->M != x->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec x: global dim %d %d",mat->M,x->N);
1860   if (mat->N != b->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec b: global dim %d %d",mat->N,b->N);
1861   if (mat->N != y->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec y: global dim %d %d",mat->N,y->N);
1862   if (x->n != y->n)   SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Vec x,Vec y: local dim %d %d",x->n,y->n);
1863 
1864   PLogEventBegin(MAT_SolveTransposeAdd,mat,b,x,y);
1865   if (mat->ops->solvetransposeadd) {
1866     ierr = (*mat->ops->solvetransposeadd)(mat,b,y,x);CHKERRQ(ierr);
1867   } else {
1868     /* do the solve then the add manually */
1869     if (x != y) {
1870       ierr = MatSolveTranspose(mat,b,x);CHKERRQ(ierr);
1871       ierr = VecAXPY(&one,y,x);CHKERRQ(ierr);
1872     } else {
1873       ierr = VecDuplicate(x,&tmp);CHKERRQ(ierr);
1874       PLogObjectParent(mat,tmp);
1875       ierr = VecCopy(x,tmp);CHKERRQ(ierr);
1876       ierr = MatSolveTranspose(mat,b,x);CHKERRQ(ierr);
1877       ierr = VecAXPY(&one,tmp,x);CHKERRQ(ierr);
1878       ierr = VecDestroy(tmp);CHKERRQ(ierr);
1879     }
1880   }
1881   PLogEventEnd(MAT_SolveTransposeAdd,mat,b,x,y);
1882   PetscFunctionReturn(0);
1883 }
1884 /* ----------------------------------------------------------------*/
1885 
1886 #undef __FUNC__
1887 #define __FUNC__ "MatRelax"
1888 /*@
1889    MatRelax - Computes one relaxation sweep.
1890 
1891    Collective on Mat and Vec
1892 
1893    Input Parameters:
1894 +  mat - the matrix
1895 .  b - the right hand side
1896 .  omega - the relaxation factor
1897 .  flag - flag indicating the type of SOR (see below)
1898 .  shift -  diagonal shift
1899 -  its - the number of iterations
1900 
1901    Output Parameters:
1902 .  x - the solution (can contain an initial guess)
1903 
1904    SOR Flags:
1905 .     SOR_FORWARD_SWEEP - forward SOR
1906 .     SOR_BACKWARD_SWEEP - backward SOR
1907 .     SOR_SYMMETRIC_SWEEP - SSOR (symmetric SOR)
1908 .     SOR_LOCAL_FORWARD_SWEEP - local forward SOR
1909 .     SOR_LOCAL_BACKWARD_SWEEP - local forward SOR
1910 .     SOR_LOCAL_SYMMETRIC_SWEEP - local SSOR
1911 .     SOR_APPLY_UPPER, SOR_APPLY_LOWER - applies
1912          upper/lower triangular part of matrix to
1913          vector (with omega)
1914 .     SOR_ZERO_INITIAL_GUESS - zero initial guess
1915 
1916    Notes:
1917    SOR_LOCAL_FORWARD_SWEEP, SOR_LOCAL_BACKWARD_SWEEP, and
1918    SOR_LOCAL_SYMMETRIC_SWEEP perform seperate independent smoothings
1919    on each processor.
1920 
1921    Application programmers will not generally use MatRelax() directly,
1922    but instead will employ the SLES/PC interface.
1923 
1924    Notes for Advanced Users:
1925    The flags are implemented as bitwise inclusive or operations.
1926    For example, use (SOR_ZERO_INITIAL_GUESS | SOR_SYMMETRIC_SWEEP)
1927    to specify a zero initial guess for SSOR.
1928 
1929    Most users should employ the simplified SLES interface for linear solvers
1930    instead of working directly with matrix algebra routines such as this.
1931    See, e.g., SLESCreate().
1932 
1933    Level: developer
1934 
1935 .keywords: matrix, relax, relaxation, sweep
1936 @*/
1937 int MatRelax(Mat mat,Vec b,PetscReal omega,MatSORType flag,PetscReal shift,int its,Vec x)
1938 {
1939   int ierr;
1940 
1941   PetscFunctionBegin;
1942   PetscValidHeaderSpecific(mat,MAT_COOKIE);
1943   PetscValidHeaderSpecific(b,VEC_COOKIE);
1944   PetscValidHeaderSpecific(x,VEC_COOKIE);
1945   PetscCheckSameComm(mat,b);
1946   PetscCheckSameComm(mat,x);
1947   if (!mat->ops->relax) SETERRQ(PETSC_ERR_SUP,0,"");
1948   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
1949   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
1950   if (mat->N != x->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec x: global dim %d %d",mat->N,x->N);
1951   if (mat->M != b->N) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec b: global dim %d %d",mat->M,b->N);
1952   if (mat->m != b->n) SETERRQ2(PETSC_ERR_ARG_SIZ,0,"Mat mat,Vec b: local dim %d %d",mat->m,b->n);
1953 
1954   PLogEventBegin(MAT_Relax,mat,b,x,0);
1955   ierr =(*mat->ops->relax)(mat,b,omega,flag,shift,its,x);CHKERRQ(ierr);
1956   PLogEventEnd(MAT_Relax,mat,b,x,0);
1957   PetscFunctionReturn(0);
1958 }
1959 
1960 #undef __FUNC__
1961 #define __FUNC__ "MatCopy_Basic"
1962 /*
1963       Default matrix copy routine.
1964 */
1965 int MatCopy_Basic(Mat A,Mat B,MatStructure str)
1966 {
1967   int    ierr,i,rstart,rend,nz,*cwork;
1968   Scalar *vwork;
1969 
1970   PetscFunctionBegin;
1971   ierr = MatZeroEntries(B);CHKERRQ(ierr);
1972   ierr = MatGetOwnershipRange(A,&rstart,&rend);CHKERRQ(ierr);
1973   for (i=rstart; i<rend; i++) {
1974     ierr = MatGetRow(A,i,&nz,&cwork,&vwork);CHKERRQ(ierr);
1975     ierr = MatSetValues(B,1,&i,nz,cwork,vwork,INSERT_VALUES);CHKERRQ(ierr);
1976     ierr = MatRestoreRow(A,i,&nz,&cwork,&vwork);CHKERRQ(ierr);
1977   }
1978   ierr = MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
1979   ierr = MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
1980   PetscFunctionReturn(0);
1981 }
1982 
1983 #undef __FUNC__
1984 #define __FUNC__ "MatCopy"
1985 /*@C
1986    MatCopy - Copys a matrix to another matrix.
1987 
1988    Collective on Mat
1989 
1990    Input Parameters:
1991 +  A - the matrix
1992 -  str - SAME_NONZERO_PATTERN or DIFFERENT_NONZERO_PATTERN
1993 
1994    Output Parameter:
1995 .  B - where the copy is put
1996 
1997    Notes:
1998    If you use SAME_NONZERO_PATTERN then the zero matrices had better have the
1999    same nonzero pattern or the routine will crash.
2000 
2001    MatCopy() copies the matrix entries of a matrix to another existing
2002    matrix (after first zeroing the second matrix).  A related routine is
2003    MatConvert(), which first creates a new matrix and then copies the data.
2004 
2005    Level: intermediate
2006 
2007 .keywords: matrix, copy, convert
2008 
2009 .seealso: MatConvert()
2010 @*/
2011 int MatCopy(Mat A,Mat B,MatStructure str)
2012 {
2013   int ierr;
2014 
2015   PetscFunctionBegin;
2016   PetscValidHeaderSpecific(A,MAT_COOKIE);
2017   PetscValidHeaderSpecific(B,MAT_COOKIE);
2018   PetscCheckSameComm(A,B);
2019   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2020   if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
2021   if (A->M != B->M || A->N != B->N) SETERRQ4(PETSC_ERR_ARG_SIZ,0,"Mat A,Mat B: global dim %d %d",A->M,B->M,
2022                                              A->N,B->N);
2023 
2024   PLogEventBegin(MAT_Copy,A,B,0,0);
2025   if (A->ops->copy) {
2026     ierr = (*A->ops->copy)(A,B,str);CHKERRQ(ierr);
2027   } else { /* generic conversion */
2028     ierr = MatCopy_Basic(A,B,str);CHKERRQ(ierr);
2029   }
2030   PLogEventEnd(MAT_Copy,A,B,0,0);
2031   PetscFunctionReturn(0);
2032 }
2033 
2034 static int MatConvertersSet = 0;
2035 static int (*MatConverters[MAX_MATRIX_TYPES][MAX_MATRIX_TYPES])(Mat,MatType,Mat*) =
2036            {{0,0,0,0,0,0,0,0,0,0,0,0},{0,0,0,0,0,0,0,0,0,0,0,0},
2037             {0,0,0,0,0,0,0,0,0,0,0,0},{0,0,0,0,0,0,0,0,0,0,0,0},
2038             {0,0,0,0,0,0,0,0,0,0,0,0},{0,0,0,0,0,0,0,0,0,0,0,0},
2039             {0,0,0,0,0,0,0,0,0,0,0,0},{0,0,0,0,0,0,0,0,0,0,0,0},
2040             {0,0,0,0,0,0,0,0,0,0,0,0},{0,0,0,0,0,0,0,0,0,0,0,0},
2041             {0,0,0,0,0,0,0,0,0,0,0,0},{0,0,0,0,0,0,0,0,0,0,0,0}};
2042 
2043 #undef __FUNC__
2044 #define __FUNC__ "MatConvertRegister"
2045 /*@C
2046     MatConvertRegister - Allows one to register a routine that converts between
2047     two matrix types.
2048 
2049     Not Collective
2050 
2051     Input Parameters:
2052 +   intype - the type of matrix (defined in include/mat.h), for example, MATSEQAIJ.
2053 -   outtype - new matrix type, or MATSAME
2054 
2055     Level: advanced
2056 
2057 .seealso: MatConvertRegisterAll()
2058 @*/
2059 int MatConvertRegister(MatType intype,MatType outtype,int (*converter)(Mat,MatType,Mat*))
2060 {
2061   PetscFunctionBegin;
2062   MatConverters[intype][outtype] = converter;
2063   MatConvertersSet               = 1;
2064   PetscFunctionReturn(0);
2065 }
2066 
2067 #undef __FUNC__
2068 #define __FUNC__ "MatConvert"
2069 /*@C
2070    MatConvert - Converts a matrix to another matrix, either of the same
2071    or different type.
2072 
2073    Collective on Mat
2074 
2075    Input Parameters:
2076 +  mat - the matrix
2077 -  newtype - new matrix type.  Use MATSAME to create a new matrix of the
2078    same type as the original matrix.
2079 
2080    Output Parameter:
2081 .  M - pointer to place new matrix
2082 
2083    Notes:
2084    MatConvert() first creates a new matrix and then copies the data from
2085    the first matrix.  A related routine is MatCopy(), which copies the matrix
2086    entries of one matrix to another already existing matrix context.
2087 
2088    Level: intermediate
2089 
2090 .keywords: matrix, copy, convert
2091 
2092 .seealso: MatCopy(), MatDuplicate()
2093 @*/
2094 int MatConvert(Mat mat,MatType newtype,Mat *M)
2095 {
2096   int ierr;
2097 
2098   PetscFunctionBegin;
2099   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2100   PetscValidPointer(M);
2101   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2102   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
2103 
2104   if (newtype > MAX_MATRIX_TYPES || newtype < -1) {
2105     SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,1,"Not a valid matrix type");
2106   }
2107   *M  = 0;
2108 
2109   if (!MatConvertersSet) {
2110     ierr = MatLoadRegisterAll();CHKERRQ(ierr);
2111   }
2112 
2113   PLogEventBegin(MAT_Convert,mat,0,0,0);
2114   if ((newtype == mat->type || newtype == MATSAME) && mat->ops->duplicate) {
2115     ierr = (*mat->ops->duplicate)(mat,MAT_COPY_VALUES,M);CHKERRQ(ierr);
2116   } else {
2117     if (!MatConvertersSet) {
2118       ierr = MatConvertRegisterAll();CHKERRQ(ierr);
2119     }
2120     if (!MatConverters[mat->type][newtype]) {
2121       SETERRQ(PETSC_ERR_ARG_WRONG,1,"Invalid matrix type, or matrix converter not registered");
2122     }
2123     ierr = (*MatConverters[mat->type][newtype])(mat,newtype,M);CHKERRQ(ierr);
2124   }
2125   PLogEventEnd(MAT_Convert,mat,0,0,0);
2126   PetscFunctionReturn(0);
2127 }
2128 
2129 #undef __FUNC__
2130 #define __FUNC__ "MatDuplicate"
2131 /*@C
2132    MatDuplicate - Duplicates a matrix including the non-zero structure.
2133 
2134    Collective on Mat
2135 
2136    Input Parameters:
2137 +  mat - the matrix
2138 -  op - either MAT_DO_NOT_COPY_VALUES or MAT_COPY_VALUES, cause it to copy nonzero
2139         values as well or not
2140 
2141    Output Parameter:
2142 .  M - pointer to place new matrix
2143 
2144    Level: intermediate
2145 
2146 .keywords: matrix, copy, convert, duplicate
2147 
2148 .seealso: MatCopy(), MatConvert()
2149 @*/
2150 int MatDuplicate(Mat mat,MatDuplicateOption op,Mat *M)
2151 {
2152   int ierr;
2153 
2154   PetscFunctionBegin;
2155   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2156   PetscValidPointer(M);
2157   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2158   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
2159 
2160   *M  = 0;
2161   PLogEventBegin(MAT_Convert,mat,0,0,0);
2162   if (!mat->ops->duplicate) {
2163     SETERRQ(PETSC_ERR_SUP,1,"Not written for this matrix type");
2164   }
2165   ierr = (*mat->ops->duplicate)(mat,op,M);CHKERRQ(ierr);
2166   PLogEventEnd(MAT_Convert,mat,0,0,0);
2167   PetscFunctionReturn(0);
2168 }
2169 
2170 #undef __FUNC__
2171 #define __FUNC__ "MatGetDiagonal"
2172 /*@
2173    MatGetDiagonal - Gets the diagonal of a matrix.
2174 
2175    Collective on Mat and Vec
2176 
2177    Input Parameters:
2178 +  mat - the matrix
2179 -  v - the vector for storing the diagonal
2180 
2181    Output Parameter:
2182 .  v - the diagonal of the matrix
2183 
2184    Notes:
2185    For the SeqAIJ matrix format, this routine may also be called
2186    on a LU factored matrix; in that case it routines the reciprocal of
2187    the diagonal entries in U. It returns the entries permuted by the
2188    row and column permutation used during the symbolic factorization.
2189 
2190    Level: intermediate
2191 
2192 .keywords: matrix, get, diagonal
2193 
2194 .seealso: MatGetRow(), MatGetSubmatrices(), MatGetSubmatrix()
2195 @*/
2196 int MatGetDiagonal(Mat mat,Vec v)
2197 {
2198   int ierr;
2199 
2200   PetscFunctionBegin;
2201   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2202   PetscValidHeaderSpecific(v,VEC_COOKIE);
2203   /* PetscCheckSameComm(mat,v); Could be MPI vector but Seq matrix cause of two submatrix storage */
2204   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2205   if (!mat->ops->getdiagonal) SETERRQ(PETSC_ERR_SUP,0,"");
2206 
2207   ierr = (*mat->ops->getdiagonal)(mat,v);CHKERRQ(ierr);
2208   PetscFunctionReturn(0);
2209 }
2210 
2211 #undef __FUNC__
2212 #define __FUNC__ "MatTranspose"
2213 /*@C
2214    MatTranspose - Computes an in-place or out-of-place transpose of a matrix.
2215 
2216    Collective on Mat
2217 
2218    Input Parameter:
2219 .  mat - the matrix to transpose
2220 
2221    Output Parameters:
2222 .  B - the transpose (or pass in PETSC_NULL for an in-place transpose)
2223 
2224    Level: intermediate
2225 
2226 .keywords: matrix, transpose
2227 
2228 .seealso: MatMultTranspose(), MatMultTransposeAdd()
2229 @*/
2230 int MatTranspose(Mat mat,Mat *B)
2231 {
2232   int ierr;
2233 
2234   PetscFunctionBegin;
2235   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2236   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2237   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
2238   if (!mat->ops->transpose) SETERRQ(PETSC_ERR_SUP,0,"");
2239   ierr = (*mat->ops->transpose)(mat,B);CHKERRQ(ierr);
2240   PetscFunctionReturn(0);
2241 }
2242 
2243 #undef __FUNC__
2244 #define __FUNC__ "MatPermute"
2245 /*@C
2246    MatPermute - Creates a new matrix with rows and columns permuted from the
2247    original.
2248 
2249    Collective on Mat
2250 
2251    Input Parameters:
2252 +  mat - the matrix to permute
2253 .  row - row permutation, each processor supplies only the permutation for its rows
2254 -  col - column permutation, each processor needs the entire column permutation, that is
2255          this is the same size as the total number of columns in the matrix
2256 
2257    Output Parameters:
2258 .  B - the permuted matrix
2259 
2260    Level: advanced
2261 
2262 .keywords: matrix, transpose
2263 
2264 .seealso: MatGetOrdering()
2265 @*/
2266 int MatPermute(Mat mat,IS row,IS col,Mat *B)
2267 {
2268   int ierr;
2269 
2270   PetscFunctionBegin;
2271   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2272   PetscValidHeaderSpecific(row,IS_COOKIE);
2273   PetscValidHeaderSpecific(col,IS_COOKIE);
2274   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2275   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
2276   if (!mat->ops->permute) SETERRQ(PETSC_ERR_SUP,0,"");
2277   ierr = (*mat->ops->permute)(mat,row,col,B);CHKERRQ(ierr);
2278   PetscFunctionReturn(0);
2279 }
2280 
2281 #undef __FUNC__
2282 #define __FUNC__ "MatEqual"
2283 /*@
2284    MatEqual - Compares two matrices.
2285 
2286    Collective on Mat
2287 
2288    Input Parameters:
2289 +  A - the first matrix
2290 -  B - the second matrix
2291 
2292    Output Parameter:
2293 .  flg - PETSC_TRUE if the matrices are equal; PETSC_FALSE otherwise.
2294 
2295    Level: intermediate
2296 
2297 .keywords: matrix, equal, equivalent
2298 @*/
2299 int MatEqual(Mat A,Mat B,PetscTruth *flg)
2300 {
2301   int ierr;
2302 
2303   PetscFunctionBegin;
2304   PetscValidHeaderSpecific(A,MAT_COOKIE);
2305   PetscValidHeaderSpecific(B,MAT_COOKIE);
2306   PetscValidIntPointer(flg);
2307   PetscCheckSameComm(A,B);
2308   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2309   if (!B->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2310   if (A->M != B->M || A->N != B->N) SETERRQ4(PETSC_ERR_ARG_SIZ,0,"Mat A,Mat B: global dim %d %d %d %d",
2311                                              A->M,B->M,A->N,B->N);
2312   if (!A->ops->equal) SETERRQ(PETSC_ERR_SUP,0,"");
2313   ierr = (*A->ops->equal)(A,B,flg);CHKERRQ(ierr);
2314   PetscFunctionReturn(0);
2315 }
2316 
2317 #undef __FUNC__
2318 #define __FUNC__ "MatDiagonalScale"
2319 /*@
2320    MatDiagonalScale - Scales a matrix on the left and right by diagonal
2321    matrices that are stored as vectors.  Either of the two scaling
2322    matrices can be PETSC_NULL.
2323 
2324    Collective on Mat
2325 
2326    Input Parameters:
2327 +  mat - the matrix to be scaled
2328 .  l - the left scaling vector (or PETSC_NULL)
2329 -  r - the right scaling vector (or PETSC_NULL)
2330 
2331    Notes:
2332    MatDiagonalScale() computes A = LAR, where
2333    L = a diagonal matrix, R = a diagonal matrix
2334 
2335    Level: intermediate
2336 
2337 .keywords: matrix, diagonal, scale
2338 
2339 .seealso: MatScale()
2340 @*/
2341 int MatDiagonalScale(Mat mat,Vec l,Vec r)
2342 {
2343   int ierr;
2344 
2345   PetscFunctionBegin;
2346   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2347   if (!mat->ops->diagonalscale) SETERRQ(PETSC_ERR_SUP,0,"");
2348   if (l) {PetscValidHeaderSpecific(l,VEC_COOKIE);PetscCheckSameComm(mat,l);}
2349   if (r) {PetscValidHeaderSpecific(r,VEC_COOKIE);PetscCheckSameComm(mat,r);}
2350   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2351   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
2352 
2353   PLogEventBegin(MAT_Scale,mat,0,0,0);
2354   ierr = (*mat->ops->diagonalscale)(mat,l,r);CHKERRQ(ierr);
2355   PLogEventEnd(MAT_Scale,mat,0,0,0);
2356   PetscFunctionReturn(0);
2357 }
2358 
2359 #undef __FUNC__
2360 #define __FUNC__ "MatScale"
2361 /*@
2362     MatScale - Scales all elements of a matrix by a given number.
2363 
2364     Collective on Mat
2365 
2366     Input Parameters:
2367 +   mat - the matrix to be scaled
2368 -   a  - the scaling value
2369 
2370     Output Parameter:
2371 .   mat - the scaled matrix
2372 
2373     Level: intermediate
2374 
2375 .keywords: matrix, scale
2376 
2377 .seealso: MatDiagonalScale()
2378 @*/
2379 int MatScale(Scalar *a,Mat mat)
2380 {
2381   int ierr;
2382 
2383   PetscFunctionBegin;
2384   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2385   PetscValidScalarPointer(a);
2386   if (!mat->ops->scale) SETERRQ(PETSC_ERR_SUP,0,"");
2387   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2388   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
2389 
2390   PLogEventBegin(MAT_Scale,mat,0,0,0);
2391   ierr = (*mat->ops->scale)(a,mat);CHKERRQ(ierr);
2392   PLogEventEnd(MAT_Scale,mat,0,0,0);
2393   PetscFunctionReturn(0);
2394 }
2395 
2396 #undef __FUNC__
2397 #define __FUNC__ "MatNorm"
2398 /*@
2399    MatNorm - Calculates various norms of a matrix.
2400 
2401    Collective on Mat
2402 
2403    Input Parameters:
2404 +  mat - the matrix
2405 -  type - the type of norm, NORM_1, NORM_2, NORM_FROBENIUS, NORM_INFINITY
2406 
2407    Output Parameters:
2408 .  norm - the resulting norm
2409 
2410    Level: intermediate
2411 
2412 .keywords: matrix, norm, Frobenius
2413 @*/
2414 int MatNorm(Mat mat,NormType type,PetscReal *norm)
2415 {
2416   int ierr;
2417 
2418   PetscFunctionBegin;
2419   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2420   PetscValidScalarPointer(norm);
2421 
2422   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2423   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
2424   if (!mat->ops->norm) SETERRQ(PETSC_ERR_SUP,0,"Not for this matrix type");
2425   ierr = (*mat->ops->norm)(mat,type,norm);CHKERRQ(ierr);
2426   PetscFunctionReturn(0);
2427 }
2428 
2429 /*
2430      This variable is used to prevent counting of MatAssemblyBegin() that
2431    are called from within a MatAssemblyEnd().
2432 */
2433 static int MatAssemblyEnd_InUse = 0;
2434 #undef __FUNC__
2435 #define __FUNC__ "MatAssemblyBegin"
2436 /*@
2437    MatAssemblyBegin - Begins assembling the matrix.  This routine should
2438    be called after completing all calls to MatSetValues().
2439 
2440    Collective on Mat
2441 
2442    Input Parameters:
2443 +  mat - the matrix
2444 -  type - type of assembly, either MAT_FLUSH_ASSEMBLY or MAT_FINAL_ASSEMBLY
2445 
2446    Notes:
2447    MatSetValues() generally caches the values.  The matrix is ready to
2448    use only after MatAssemblyBegin() and MatAssemblyEnd() have been called.
2449    Use MAT_FLUSH_ASSEMBLY when switching between ADD_VALUES and INSERT_VALUES
2450    in MatSetValues(); use MAT_FINAL_ASSEMBLY for the final assembly before
2451    using the matrix.
2452 
2453    Level: beginner
2454 
2455 .keywords: matrix, assembly, assemble, begin
2456 
2457 .seealso: MatAssemblyEnd(), MatSetValues(), MatAssembled()
2458 @*/
2459 int MatAssemblyBegin(Mat mat,MatAssemblyType type)
2460 {
2461   int ierr;
2462 
2463   PetscFunctionBegin;
2464   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2465   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix.\n did you forget to call MatSetUnfactored()?");
2466   if (mat->assembled) {
2467     mat->was_assembled = PETSC_TRUE;
2468     mat->assembled     = PETSC_FALSE;
2469   }
2470   if (!MatAssemblyEnd_InUse) {
2471     PLogEventBegin(MAT_AssemblyBegin,mat,0,0,0);
2472     if (mat->ops->assemblybegin){ierr = (*mat->ops->assemblybegin)(mat,type);CHKERRQ(ierr);}
2473     PLogEventEnd(MAT_AssemblyBegin,mat,0,0,0);
2474   } else {
2475     if (mat->ops->assemblybegin){ierr = (*mat->ops->assemblybegin)(mat,type);CHKERRQ(ierr);}
2476   }
2477   PetscFunctionReturn(0);
2478 }
2479 
2480 #undef __FUNC__
2481 #define __FUNC__ "MatAssembed"
2482 /*@
2483    MatAssembled - Indicates if a matrix has been assembled and is ready for
2484      use; for example, in matrix-vector product.
2485 
2486    Collective on Mat
2487 
2488    Input Parameter:
2489 .  mat - the matrix
2490 
2491    Output Parameter:
2492 .  assembled - PETSC_TRUE or PETSC_FALSE
2493 
2494    Level: advanced
2495 
2496 .keywords: matrix, assembly, assemble, begin
2497 
2498 .seealso: MatAssemblyEnd(), MatSetValues(), MatAssemblyBegin()
2499 @*/
2500 int MatAssembled(Mat mat,PetscTruth *assembled)
2501 {
2502   PetscFunctionBegin;
2503   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2504   *assembled = mat->assembled;
2505   PetscFunctionReturn(0);
2506 }
2507 
2508 #undef __FUNC__
2509 #define __FUNC__ "MatView_Private"
2510 /*
2511     Processes command line options to determine if/how a matrix
2512   is to be viewed. Called by MatAssemblyEnd() and MatLoad().
2513 */
2514 int MatView_Private(Mat mat)
2515 {
2516   int        ierr;
2517   PetscTruth flg;
2518 
2519   PetscFunctionBegin;
2520   ierr = OptionsHasName(mat->prefix,"-mat_view_info",&flg);CHKERRQ(ierr);
2521   if (flg) {
2522     ierr = ViewerPushFormat(VIEWER_STDOUT_(mat->comm),VIEWER_FORMAT_ASCII_INFO,0);CHKERRQ(ierr);
2523     ierr = MatView(mat,VIEWER_STDOUT_(mat->comm));CHKERRQ(ierr);
2524     ierr = ViewerPopFormat(VIEWER_STDOUT_(mat->comm));CHKERRQ(ierr);
2525   }
2526   ierr = OptionsHasName(mat->prefix,"-mat_view_info_detailed",&flg);CHKERRQ(ierr);
2527   if (flg) {
2528     ierr = ViewerPushFormat(VIEWER_STDOUT_(mat->comm),VIEWER_FORMAT_ASCII_INFO_LONG,0);CHKERRQ(ierr);
2529     ierr = MatView(mat,VIEWER_STDOUT_(mat->comm));CHKERRQ(ierr);
2530     ierr = ViewerPopFormat(VIEWER_STDOUT_(mat->comm));CHKERRQ(ierr);
2531   }
2532   ierr = OptionsHasName(mat->prefix,"-mat_view",&flg);CHKERRQ(ierr);
2533   if (flg) {
2534     ierr = MatView(mat,VIEWER_STDOUT_(mat->comm));CHKERRQ(ierr);
2535   }
2536   ierr = OptionsHasName(mat->prefix,"-mat_view_matlab",&flg);CHKERRQ(ierr);
2537   if (flg) {
2538     ierr = ViewerPushFormat(VIEWER_STDOUT_(mat->comm),VIEWER_FORMAT_ASCII_MATLAB,"M");CHKERRQ(ierr);
2539     ierr = MatView(mat,VIEWER_STDOUT_(mat->comm));CHKERRQ(ierr);
2540     ierr = ViewerPopFormat(VIEWER_STDOUT_(mat->comm));CHKERRQ(ierr);
2541   }
2542   ierr = OptionsHasName(mat->prefix,"-mat_view_draw",&flg);CHKERRQ(ierr);
2543   if (flg) {
2544     ierr = OptionsHasName(mat->prefix,"-mat_view_contour",&flg);CHKERRQ(ierr);
2545     if (flg) {
2546       ViewerPushFormat(VIEWER_DRAW_(mat->comm),VIEWER_FORMAT_DRAW_CONTOUR,0);CHKERRQ(ierr);
2547     }
2548     ierr = MatView(mat,VIEWER_DRAW_(mat->comm));CHKERRQ(ierr);
2549     ierr = ViewerFlush(VIEWER_DRAW_(mat->comm));CHKERRQ(ierr);
2550     if (flg) {
2551       ViewerPopFormat(VIEWER_DRAW_(mat->comm));CHKERRQ(ierr);
2552     }
2553   }
2554   ierr = OptionsHasName(mat->prefix,"-mat_view_socket",&flg);CHKERRQ(ierr);
2555   if (flg) {
2556     ierr = MatView(mat,VIEWER_SOCKET_(mat->comm));CHKERRQ(ierr);
2557     ierr = ViewerFlush(VIEWER_SOCKET_(mat->comm));CHKERRQ(ierr);
2558   }
2559   PetscFunctionReturn(0);
2560 }
2561 
2562 #undef __FUNC__
2563 #define __FUNC__ "MatAssemblyEnd"
2564 /*@
2565    MatAssemblyEnd - Completes assembling the matrix.  This routine should
2566    be called after MatAssemblyBegin().
2567 
2568    Collective on Mat
2569 
2570    Input Parameters:
2571 +  mat - the matrix
2572 -  type - type of assembly, either MAT_FLUSH_ASSEMBLY or MAT_FINAL_ASSEMBLY
2573 
2574    Options Database Keys:
2575 +  -mat_view_info - Prints info on matrix at conclusion of MatEndAssembly()
2576 .  -mat_view_info_detailed - Prints more detailed info
2577 .  -mat_view - Prints matrix in ASCII format
2578 .  -mat_view_matlab - Prints matrix in Matlab format
2579 .  -mat_view_draw - Draws nonzero structure of matrix, using MatView() and DrawOpenX().
2580 .  -display <name> - Sets display name (default is host)
2581 -  -draw_pause <sec> - Sets number of seconds to pause after display
2582 
2583    Notes:
2584    MatSetValues() generally caches the values.  The matrix is ready to
2585    use only after MatAssemblyBegin() and MatAssemblyEnd() have been called.
2586    Use MAT_FLUSH_ASSEMBLY when switching between ADD_VALUES and INSERT_VALUES
2587    in MatSetValues(); use MAT_FINAL_ASSEMBLY for the final assembly before
2588    using the matrix.
2589 
2590    Level: beginner
2591 
2592 .keywords: matrix, assembly, assemble, end
2593 
2594 .seealso: MatAssemblyBegin(), MatSetValues(), DrawOpenX(), MatView(), MatAssembled()
2595 @*/
2596 int MatAssemblyEnd(Mat mat,MatAssemblyType type)
2597 {
2598   int        ierr;
2599   static int inassm = 0;
2600 
2601   PetscFunctionBegin;
2602   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2603 
2604   inassm++;
2605   MatAssemblyEnd_InUse++;
2606   if (MatAssemblyEnd_InUse == 1) { /* Do the logging only the first time through */
2607     PLogEventBegin(MAT_AssemblyEnd,mat,0,0,0);
2608     if (mat->ops->assemblyend) {
2609       ierr = (*mat->ops->assemblyend)(mat,type);CHKERRQ(ierr);
2610     }
2611     PLogEventEnd(MAT_AssemblyEnd,mat,0,0,0);
2612   } else {
2613     if (mat->ops->assemblyend) {
2614       ierr = (*mat->ops->assemblyend)(mat,type);CHKERRQ(ierr);
2615     }
2616   }
2617 
2618   /* Flush assembly is not a true assembly */
2619   if (type != MAT_FLUSH_ASSEMBLY) {
2620     mat->assembled  = PETSC_TRUE; mat->num_ass++;
2621   }
2622   mat->insertmode = NOT_SET_VALUES;
2623   MatAssemblyEnd_InUse--;
2624 
2625   if (inassm == 1 && type != MAT_FLUSH_ASSEMBLY) {
2626     ierr = MatView_Private(mat);CHKERRQ(ierr);
2627   }
2628   inassm--;
2629   PetscFunctionReturn(0);
2630 }
2631 
2632 
2633 #undef __FUNC__
2634 #define __FUNC__ "MatCompress"
2635 /*@
2636    MatCompress - Tries to store the matrix in as little space as
2637    possible.  May fail if memory is already fully used, since it
2638    tries to allocate new space.
2639 
2640    Collective on Mat
2641 
2642    Input Parameters:
2643 .  mat - the matrix
2644 
2645    Level: advanced
2646 
2647 .keywords: matrix, compress
2648 @*/
2649 int MatCompress(Mat mat)
2650 {
2651   int ierr;
2652 
2653   PetscFunctionBegin;
2654   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2655   if (mat->ops->compress) {ierr = (*mat->ops->compress)(mat);CHKERRQ(ierr);}
2656   PetscFunctionReturn(0);
2657 }
2658 
2659 #undef __FUNC__
2660 #define __FUNC__ "MatSetOption"
2661 /*@
2662    MatSetOption - Sets a parameter option for a matrix. Some options
2663    may be specific to certain storage formats.  Some options
2664    determine how values will be inserted (or added). Sorted,
2665    row-oriented input will generally assemble the fastest. The default
2666    is row-oriented, nonsorted input.
2667 
2668    Collective on Mat
2669 
2670    Input Parameters:
2671 +  mat - the matrix
2672 -  option - the option, one of those listed below (and possibly others),
2673              e.g., MAT_ROWS_SORTED, MAT_NEW_NONZERO_LOCATION_ERR
2674 
2675    Options Describing Matrix Structure:
2676 +    MAT_SYMMETRIC - symmetric in terms of both structure and value
2677 -    MAT_STRUCTURALLY_SYMMETRIC - symmetric nonzero structure
2678 
2679    Options For Use with MatSetValues():
2680    Insert a logically dense subblock, which can be
2681 +    MAT_ROW_ORIENTED - row-oriented
2682 .    MAT_COLUMN_ORIENTED - column-oriented
2683 .    MAT_ROWS_SORTED - sorted by row
2684 .    MAT_ROWS_UNSORTED - not sorted by row
2685 .    MAT_COLUMNS_SORTED - sorted by column
2686 -    MAT_COLUMNS_UNSORTED - not sorted by column
2687 
2688    Not these options reflect the data you pass in with MatSetValues(); it has
2689    nothing to do with how the data is stored internally in the matrix
2690    data structure.
2691 
2692    When (re)assembling a matrix, we can restrict the input for
2693    efficiency/debugging purposes.  These options include
2694 +    MAT_NO_NEW_NONZERO_LOCATIONS - additional insertions will not be
2695         allowed if they generate a new nonzero
2696 .    MAT_YES_NEW_NONZERO_LOCATIONS - additional insertions will be allowed
2697 .    MAT_NO_NEW_DIAGONALS - additional insertions will not be allowed if
2698          they generate a nonzero in a new diagonal (for block diagonal format only)
2699 .    MAT_YES_NEW_DIAGONALS - new diagonals will be allowed (for block diagonal format only)
2700 .    MAT_IGNORE_OFF_PROC_ENTRIES - drops off-processor entries
2701 .    MAT_NEW_NONZERO_LOCATION_ERR - generates an error for new matrix entry
2702 -    MAT_USE_HASH_TABLE - uses a hash table to speed up matrix assembly
2703 
2704    Notes:
2705    Some options are relevant only for particular matrix types and
2706    are thus ignored by others.  Other options are not supported by
2707    certain matrix types and will generate an error message if set.
2708 
2709    If using a Fortran 77 module to compute a matrix, one may need to
2710    use the column-oriented option (or convert to the row-oriented
2711    format).
2712 
2713    MAT_NO_NEW_NONZERO_LOCATIONS indicates that any add or insertion
2714    that would generate a new entry in the nonzero structure is instead
2715    ignored.  Thus, if memory has not alredy been allocated for this particular
2716    data, then the insertion is ignored. For dense matrices, in which
2717    the entire array is allocated, no entries are ever ignored.
2718 
2719    MAT_NEW_NONZERO_LOCATION_ERR indicates that any add or insertion
2720    that would generate a new entry in the nonzero structure instead produces
2721    an error. (Currently supported for AIJ and BAIJ formats only.)
2722    This is a useful flag when using SAME_NONZERO_PATTERN in calling
2723    SLESSetOperators() to ensure that the nonzero pattern truely does
2724    remain unchanged.
2725 
2726    MAT_NEW_NONZERO_ALLOCATION_ERR indicates that any add or insertion
2727    that would generate a new entry that has not been preallocated will
2728    instead produce an error. (Currently supported for AIJ and BAIJ formats
2729    only.) This is a useful flag when debugging matrix memory preallocation.
2730 
2731    MAT_IGNORE_OFF_PROC_ENTRIES indicates entries destined for
2732    other processors should be dropped, rather than stashed.
2733    This is useful if you know that the "owning" processor is also
2734    always generating the correct matrix entries, so that PETSc need
2735    not transfer duplicate entries generated on another processor.
2736 
2737    MAT_USE_HASH_TABLE indicates that a hash table be used to improve the
2738    searches during matrix assembly. When this flag is set, the hash table
2739    is created during the first Matrix Assembly. This hash table is
2740    used the next time through, during MatSetVaules()/MatSetVaulesBlocked()
2741    to improve the searching of indices. MAT_NO_NEW_NONZERO_LOCATIONS flag
2742    should be used with MAT_USE_HASH_TABLE flag. This option is currently
2743    supported by MATMPIBAIJ format only.
2744 
2745    MAT_KEEP_ZEROED_ROWS indicates when MatZeroRows() is called the zeroed entries
2746    are kept in the nonzero structure
2747 
2748    MAT_IGNORE_ZERO_ENTRIES - when using ADD_VALUES for AIJ matrices this will stop
2749    zero values from creating a zero location in the matrix
2750 
2751    Level: intermediate
2752 
2753 .keywords: matrix, option, row-oriented, column-oriented, sorted, nonzero
2754 @*/
2755 int MatSetOption(Mat mat,MatOption op)
2756 {
2757   int ierr;
2758 
2759   PetscFunctionBegin;
2760   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2761   if (mat->ops->setoption) {ierr = (*mat->ops->setoption)(mat,op);CHKERRQ(ierr);}
2762   PetscFunctionReturn(0);
2763 }
2764 
2765 #undef __FUNC__
2766 #define __FUNC__ "MatZeroEntries"
2767 /*@
2768    MatZeroEntries - Zeros all entries of a matrix.  For sparse matrices
2769    this routine retains the old nonzero structure.
2770 
2771    Collective on Mat
2772 
2773    Input Parameters:
2774 .  mat - the matrix
2775 
2776    Level: intermediate
2777 
2778 .keywords: matrix, zero, entries
2779 
2780 .seealso: MatZeroRows()
2781 @*/
2782 int MatZeroEntries(Mat mat)
2783 {
2784   int ierr;
2785 
2786   PetscFunctionBegin;
2787   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2788   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
2789   if (!mat->ops->zeroentries) SETERRQ(PETSC_ERR_SUP,0,"");
2790 
2791   PLogEventBegin(MAT_ZeroEntries,mat,0,0,0);
2792   ierr = (*mat->ops->zeroentries)(mat);CHKERRQ(ierr);
2793   PLogEventEnd(MAT_ZeroEntries,mat,0,0,0);
2794   PetscFunctionReturn(0);
2795 }
2796 
2797 #undef __FUNC__
2798 #define __FUNC__ "MatZeroRows"
2799 /*@C
2800    MatZeroRows - Zeros all entries (except possibly the main diagonal)
2801    of a set of rows of a matrix.
2802 
2803    Collective on Mat
2804 
2805    Input Parameters:
2806 +  mat - the matrix
2807 .  is - index set of rows to remove
2808 -  diag - pointer to value put in all diagonals of eliminated rows.
2809           Note that diag is not a pointer to an array, but merely a
2810           pointer to a single value.
2811 
2812    Notes:
2813    For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
2814    but does not release memory.  For the dense and block diagonal
2815    formats this does not alter the nonzero structure.
2816 
2817    If the option MatSetOption(mat,MAT_KEEP_ZEROED_ROWS) the nonzero structure
2818    of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
2819    merely zeroed.
2820 
2821    The user can set a value in the diagonal entry (or for the AIJ and
2822    row formats can optionally remove the main diagonal entry from the
2823    nonzero structure as well, by passing a null pointer (PETSC_NULL
2824    in C or PETSC_NULL_SCALAR in Fortran) as the final argument).
2825 
2826    For the parallel case, all processes that share the matrix (i.e.,
2827    those in the communicator used for matrix creation) MUST call this
2828    routine, regardless of whether any rows being zeroed are owned by
2829    them.
2830 
2831 
2832    Level: intermediate
2833 
2834 .keywords: matrix, zero, rows, boundary conditions
2835 
2836 .seealso: MatZeroEntries(), MatZeroRowsLocal(), MatSetOption()
2837 @*/
2838 int MatZeroRows(Mat mat,IS is,Scalar *diag)
2839 {
2840   int ierr;
2841 
2842   PetscFunctionBegin;
2843   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2844   PetscValidHeaderSpecific(is,IS_COOKIE);
2845   if (diag) PetscValidScalarPointer(diag);
2846   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2847   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
2848   if (!mat->ops->zerorows) SETERRQ(PETSC_ERR_SUP,0,"");
2849 
2850   ierr = (*mat->ops->zerorows)(mat,is,diag);CHKERRQ(ierr);
2851   ierr = MatView_Private(mat);CHKERRQ(ierr);
2852   PetscFunctionReturn(0);
2853 }
2854 
2855 #undef __FUNC__
2856 #define __FUNC__ "MatZeroRowsLocal"
2857 /*@C
2858    MatZeroRowsLocal - Zeros all entries (except possibly the main diagonal)
2859    of a set of rows of a matrix; using local numbering of rows.
2860 
2861    Collective on Mat
2862 
2863    Input Parameters:
2864 +  mat - the matrix
2865 .  is - index set of rows to remove
2866 -  diag - pointer to value put in all diagonals of eliminated rows.
2867           Note that diag is not a pointer to an array, but merely a
2868           pointer to a single value.
2869 
2870    Notes:
2871    For the AIJ matrix formats this removes the old nonzero structure,
2872    but does not release memory.  For the dense and block diagonal
2873    formats this does not alter the nonzero structure.
2874 
2875    The user can set a value in the diagonal entry (or for the AIJ and
2876    row formats can optionally remove the main diagonal entry from the
2877    nonzero structure as well, by passing a null pointer (PETSC_NULL
2878    in C or PETSC_NULL_SCALAR in Fortran) as the final argument).
2879 
2880    Level: intermediate
2881 
2882 .keywords: matrix, zero, rows, boundary conditions
2883 
2884 .seealso: MatZeroEntries(), MatZeroRows()
2885 @*/
2886 int MatZeroRowsLocal(Mat mat,IS is,Scalar *diag)
2887 {
2888   int ierr;
2889   IS  newis;
2890 
2891   PetscFunctionBegin;
2892   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2893   PetscValidHeaderSpecific(is,IS_COOKIE);
2894   if (diag) PetscValidScalarPointer(diag);
2895   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
2896   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
2897   if (!mat->ops->zerorows) SETERRQ(PETSC_ERR_SUP,0,"");
2898 
2899   ierr = ISLocalToGlobalMappingApplyIS(mat->mapping,is,&newis);CHKERRQ(ierr);
2900   ierr = (*mat->ops->zerorows)(mat,newis,diag);CHKERRQ(ierr);
2901   ierr = ISDestroy(newis);CHKERRQ(ierr);
2902   PetscFunctionReturn(0);
2903 }
2904 
2905 #undef __FUNC__
2906 #define __FUNC__ "MatGetSize"
2907 /*@
2908    MatGetSize - Returns the numbers of rows and columns in a matrix.
2909 
2910    Not Collective
2911 
2912    Input Parameter:
2913 .  mat - the matrix
2914 
2915    Output Parameters:
2916 +  m - the number of global rows
2917 -  n - the number of global columns
2918 
2919    Level: beginner
2920 
2921 .keywords: matrix, dimension, size, rows, columns, global, get
2922 
2923 .seealso: MatGetLocalSize()
2924 @*/
2925 int MatGetSize(Mat mat,int *m,int* n)
2926 {
2927   int ierr;
2928 
2929   PetscFunctionBegin;
2930   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2931   ierr = (*mat->ops->getsize)(mat,m,n);CHKERRQ(ierr);
2932   PetscFunctionReturn(0);
2933 }
2934 
2935 #undef __FUNC__
2936 #define __FUNC__ "MatGetLocalSize"
2937 /*@
2938    MatGetLocalSize - Returns the number of rows and columns in a matrix
2939    stored locally.  This information may be implementation dependent, so
2940    use with care.
2941 
2942    Not Collective
2943 
2944    Input Parameters:
2945 .  mat - the matrix
2946 
2947    Output Parameters:
2948 +  m - the number of local rows
2949 -  n - the number of local columns
2950 
2951    Level: beginner
2952 
2953 .keywords: matrix, dimension, size, local, rows, columns, get
2954 
2955 .seealso: MatGetSize()
2956 @*/
2957 int MatGetLocalSize(Mat mat,int *m,int* n)
2958 {
2959   int ierr;
2960 
2961   PetscFunctionBegin;
2962   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2963   ierr = (*mat->ops->getlocalsize)(mat,m,n);CHKERRQ(ierr);
2964   PetscFunctionReturn(0);
2965 }
2966 
2967 #undef __FUNC__
2968 #define __FUNC__ "MatGetOwnershipRange"
2969 /*@
2970    MatGetOwnershipRange - Returns the range of matrix rows owned by
2971    this processor, assuming that the matrix is laid out with the first
2972    n1 rows on the first processor, the next n2 rows on the second, etc.
2973    For certain parallel layouts this range may not be well defined.
2974 
2975    Not Collective
2976 
2977    Input Parameters:
2978 .  mat - the matrix
2979 
2980    Output Parameters:
2981 +  m - the global index of the first local row
2982 -  n - one more than the global index of the last local row
2983 
2984    Level: beginner
2985 
2986 .keywords: matrix, get, range, ownership
2987 @*/
2988 int MatGetOwnershipRange(Mat mat,int *m,int* n)
2989 {
2990   int ierr;
2991 
2992   PetscFunctionBegin;
2993   PetscValidHeaderSpecific(mat,MAT_COOKIE);
2994   if (m) PetscValidIntPointer(m);
2995   if (n) PetscValidIntPointer(n);
2996   if (!mat->ops->getownershiprange) SETERRQ(PETSC_ERR_SUP,0,"");
2997   ierr = (*mat->ops->getownershiprange)(mat,m,n);CHKERRQ(ierr);
2998   PetscFunctionReturn(0);
2999 }
3000 
3001 #undef __FUNC__
3002 #define __FUNC__ "MatILUFactorSymbolic"
3003 /*@
3004    MatILUFactorSymbolic - Performs symbolic ILU factorization of a matrix.
3005    Uses levels of fill only, not drop tolerance. Use MatLUFactorNumeric()
3006    to complete the factorization.
3007 
3008    Collective on Mat
3009 
3010    Input Parameters:
3011 +  mat - the matrix
3012 .  row - row permutation
3013 .  column - column permutation
3014 -  info - structure containing
3015 $      levels - number of levels of fill.
3016 $      expected fill - as ratio of original fill.
3017 $      1 or 0 - indicating force fill on diagonal (improves robustness for matrices
3018                 missing diagonal entries)
3019 
3020    Output Parameters:
3021 .  fact - new matrix that has been symbolically factored
3022 
3023    Notes:
3024    See the users manual for additional information about
3025    choosing the fill factor for better efficiency.
3026 
3027    Most users should employ the simplified SLES interface for linear solvers
3028    instead of working directly with matrix algebra routines such as this.
3029    See, e.g., SLESCreate().
3030 
3031    Level: developer
3032 
3033 .keywords: matrix, factor, incomplete, ILU, symbolic, fill
3034 
3035 .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor()
3036           MatGetOrdering()
3037 
3038 @*/
3039 int MatILUFactorSymbolic(Mat mat,IS row,IS col,MatILUInfo *info,Mat *fact)
3040 {
3041   int ierr;
3042 
3043   PetscFunctionBegin;
3044   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3045   PetscValidPointer(fact);
3046   if (info && info->levels < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,0,"Levels of fill negative %d",info->levels);
3047   if (info && info->fill < 1.0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,0,"Expected fill less than 1.0 %g",info->fill);
3048   if (!mat->ops->ilufactorsymbolic) SETERRQ(PETSC_ERR_SUP,0,"Only MatCreateMPIRowbs() matrices support parallel ILU");
3049   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
3050   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
3051 
3052   PLogEventBegin(MAT_ILUFactorSymbolic,mat,row,col,0);
3053   ierr = (*mat->ops->ilufactorsymbolic)(mat,row,col,info,fact);CHKERRQ(ierr);
3054   PLogEventEnd(MAT_ILUFactorSymbolic,mat,row,col,0);
3055   PetscFunctionReturn(0);
3056 }
3057 
3058 #undef __FUNC__
3059 #define __FUNC__ "MatIncompleteCholeskyFactorSymbolic"
3060 /*@
3061    MatIncompleteCholeskyFactorSymbolic - Performs symbolic incomplete
3062    Cholesky factorization for a symmetric matrix.  Use
3063    MatCholeskyFactorNumeric() to complete the factorization.
3064 
3065    Collective on Mat
3066 
3067    Input Parameters:
3068 +  mat - the matrix
3069 .  perm - row and column permutation
3070 .  fill - levels of fill
3071 -  f - expected fill as ratio of original fill
3072 
3073    Output Parameter:
3074 .  fact - the factored matrix
3075 
3076    Notes:
3077    Currently only no-fill factorization is supported.
3078 
3079    Most users should employ the simplified SLES interface for linear solvers
3080    instead of working directly with matrix algebra routines such as this.
3081    See, e.g., SLESCreate().
3082 
3083    Level: developer
3084 
3085 .keywords: matrix, factor, incomplete, ICC, Cholesky, symbolic, fill
3086 
3087 .seealso: MatCholeskyFactorNumeric(), MatCholeskyFactor()
3088 @*/
3089 int MatIncompleteCholeskyFactorSymbolic(Mat mat,IS perm,PetscReal f,int fill,Mat *fact)
3090 {
3091   int ierr;
3092 
3093   PetscFunctionBegin;
3094   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3095   PetscValidPointer(fact);
3096   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
3097   if (fill < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,0,"Fill negative %d",fill);
3098   if (f < 1.0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,0,"Expected fill less than 1.0 %g",f);
3099   if (!mat->ops->incompletecholeskyfactorsymbolic) SETERRQ(PETSC_ERR_SUP,0,"Currently only MatCreateMPIRowbs() matrices support ICC in parallel");
3100   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
3101 
3102   PLogEventBegin(MAT_IncompleteCholeskyFactorSymbolic,mat,perm,0,0);
3103   ierr = (*mat->ops->incompletecholeskyfactorsymbolic)(mat,perm,f,fill,fact);CHKERRQ(ierr);
3104   PLogEventEnd(MAT_IncompleteCholeskyFactorSymbolic,mat,perm,0,0);
3105   PetscFunctionReturn(0);
3106 }
3107 
3108 #undef __FUNC__
3109 #define __FUNC__ "MatGetArray"
3110 /*@C
3111    MatGetArray - Returns a pointer to the element values in the matrix.
3112    The result of this routine is dependent on the underlying matrix data
3113    structure, and may not even work for certain matrix types.  You MUST
3114    call MatRestoreArray() when you no longer need to access the array.
3115 
3116    Not Collective
3117 
3118    Input Parameter:
3119 .  mat - the matrix
3120 
3121    Output Parameter:
3122 .  v - the location of the values
3123 
3124    Currently returns an array only for the dense formats, giving access to
3125    the local portion of the matrix in the usual Fortran column-oriented format.
3126 
3127    Fortran Note:
3128    This routine is used differently from Fortran, e.g.,
3129 .vb
3130         Mat         mat
3131         Scalar      mat_array(1)
3132         PetscOffset i_mat
3133         int         ierr
3134         call MatGetArray(mat,mat_array,i_mat,ierr)
3135 
3136   C  Access first local entry in matrix; note that array is
3137   C  treated as one dimensional
3138         value = mat_array(i_mat + 1)
3139 
3140         [... other code ...]
3141         call MatRestoreArray(mat,mat_array,i_mat,ierr)
3142 .ve
3143 
3144    See the Fortran chapter of the users manual and
3145    petsc/src/mat/examples/tests for details.
3146 
3147    Level: advanced
3148 
3149 .keywords: matrix, array, elements, values
3150 
3151 .seealso: MatRestoreArray(), MatGetArrayF90()
3152 @*/
3153 int MatGetArray(Mat mat,Scalar **v)
3154 {
3155   int ierr;
3156 
3157   PetscFunctionBegin;
3158   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3159   PetscValidPointer(v);
3160   if (!mat->ops->getarray) SETERRQ(PETSC_ERR_SUP,0,"");
3161   ierr = (*mat->ops->getarray)(mat,v);CHKERRQ(ierr);
3162   PetscFunctionReturn(0);
3163 }
3164 
3165 #undef __FUNC__
3166 #define __FUNC__ "MatRestoreArray"
3167 /*@C
3168    MatRestoreArray - Restores the matrix after MatGetArray() has been called.
3169 
3170    Not Collective
3171 
3172    Input Parameter:
3173 +  mat - the matrix
3174 -  v - the location of the values
3175 
3176    Fortran Note:
3177    This routine is used differently from Fortran, e.g.,
3178 .vb
3179         Mat         mat
3180         Scalar      mat_array(1)
3181         PetscOffset i_mat
3182         int         ierr
3183         call MatGetArray(mat,mat_array,i_mat,ierr)
3184 
3185   C  Access first local entry in matrix; note that array is
3186   C  treated as one dimensional
3187         value = mat_array(i_mat + 1)
3188 
3189         [... other code ...]
3190         call MatRestoreArray(mat,mat_array,i_mat,ierr)
3191 .ve
3192 
3193    See the Fortran chapter of the users manual and
3194    petsc/src/mat/examples/tests for details
3195 
3196    Level: advanced
3197 
3198 .keywords: matrix, array, elements, values, restore
3199 
3200 .seealso: MatGetArray(), MatRestoreArrayF90()
3201 @*/
3202 int MatRestoreArray(Mat mat,Scalar **v)
3203 {
3204   int ierr;
3205 
3206   PetscFunctionBegin;
3207   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3208   PetscValidPointer(v);
3209   if (!mat->ops->restorearray) SETERRQ(PETSC_ERR_SUP,0,"");
3210   ierr = (*mat->ops->restorearray)(mat,v);CHKERRQ(ierr);
3211   PetscFunctionReturn(0);
3212 }
3213 
3214 #undef __FUNC__
3215 #define __FUNC__ "MatGetSubMatrices"
3216 /*@C
3217    MatGetSubMatrices - Extracts several submatrices from a matrix. If submat
3218    points to an array of valid matrices, they may be reused to store the new
3219    submatrices.
3220 
3221    Collective on Mat
3222 
3223    Input Parameters:
3224 +  mat - the matrix
3225 .  n   - the number of submatrixes to be extracted (on this processor, may be zero)
3226 .  irow, icol - index sets of rows and columns to extract
3227 -  scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
3228 
3229    Output Parameter:
3230 .  submat - the array of submatrices
3231 
3232    Notes:
3233    MatGetSubMatrices() can extract only sequential submatrices
3234    (from both sequential and parallel matrices). Use MatGetSubMatrix()
3235    to extract a parallel submatrix.
3236 
3237    When extracting submatrices from a parallel matrix, each processor can
3238    form a different submatrix by setting the rows and columns of its
3239    individual index sets according to the local submatrix desired.
3240 
3241    When finished using the submatrices, the user should destroy
3242    them with MatDestroySubMatrices().
3243 
3244    MAT_REUSE_MATRIX can only be used when the nonzero structure of the
3245    original matrix has not changed from that last call to MatGetSubMatrices().
3246 
3247    Fortran Note:
3248    The Fortran interface is slightly different from that given below; it
3249    requires one to pass in  as submat a Mat (integer) array of size at least m.
3250 
3251    Level: advanced
3252 
3253 .keywords: matrix, get, submatrix, submatrices
3254 
3255 .seealso: MatDestroyMatrices(), MatGetSubMatrix(), MatGetRow(), MatGetDiagonal()
3256 @*/
3257 int MatGetSubMatrices(Mat mat,int n,IS *irow,IS *icol,MatReuse scall,Mat **submat)
3258 {
3259   int ierr;
3260 
3261   PetscFunctionBegin;
3262   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3263   if (!mat->ops->getsubmatrices) SETERRQ(PETSC_ERR_SUP,0,"");
3264   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
3265 
3266   PLogEventBegin(MAT_GetSubMatrices,mat,0,0,0);
3267   ierr = (*mat->ops->getsubmatrices)(mat,n,irow,icol,scall,submat);CHKERRQ(ierr);
3268   PLogEventEnd(MAT_GetSubMatrices,mat,0,0,0);
3269 
3270   PetscFunctionReturn(0);
3271 }
3272 
3273 #undef __FUNC__
3274 #define __FUNC__ "MatDestroyMatrices"
3275 /*@C
3276    MatDestroyMatrices - Destroys a set of matrices obtained with MatGetSubMatrices().
3277 
3278    Collective on Mat
3279 
3280    Input Parameters:
3281 +  n - the number of local matrices
3282 -  mat - the matrices
3283 
3284    Level: advanced
3285 
3286 .keywords: matrix, destroy, submatrix, submatrices
3287 
3288 .seealso: MatGetSubMatrices()
3289 @*/
3290 int MatDestroyMatrices(int n,Mat **mat)
3291 {
3292   int ierr,i;
3293 
3294   PetscFunctionBegin;
3295   if (n < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,1,"Trying to destroy negative number of matrices %d",n);
3296   PetscValidPointer(mat);
3297   for (i=0; i<n; i++) {
3298     ierr = MatDestroy((*mat)[i]);CHKERRQ(ierr);
3299   }
3300   if (n) {ierr = PetscFree(*mat);CHKERRQ(ierr);}
3301   PetscFunctionReturn(0);
3302 }
3303 
3304 #undef __FUNC__
3305 #define __FUNC__ "MatIncreaseOverlap"
3306 /*@
3307    MatIncreaseOverlap - Given a set of submatrices indicated by index sets,
3308    replaces the index sets by larger ones that represent submatrices with
3309    additional overlap.
3310 
3311    Collective on Mat
3312 
3313    Input Parameters:
3314 +  mat - the matrix
3315 .  n   - the number of index sets
3316 .  is  - the array of pointers to index sets
3317 -  ov  - the additional overlap requested
3318 
3319    Level: developer
3320 
3321 .keywords: matrix, overlap, Schwarz
3322 
3323 .seealso: MatGetSubMatrices()
3324 @*/
3325 int MatIncreaseOverlap(Mat mat,int n,IS *is,int ov)
3326 {
3327   int ierr;
3328 
3329   PetscFunctionBegin;
3330   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3331   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for unassembled matrix");
3332   if (mat->factor)     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Not for factored matrix");
3333 
3334   if (!ov) PetscFunctionReturn(0);
3335   if (!mat->ops->increaseoverlap) SETERRQ(PETSC_ERR_SUP,0,"");
3336   PLogEventBegin(MAT_IncreaseOverlap,mat,0,0,0);
3337   ierr = (*mat->ops->increaseoverlap)(mat,n,is,ov);CHKERRQ(ierr);
3338   PLogEventEnd(MAT_IncreaseOverlap,mat,0,0,0);
3339   PetscFunctionReturn(0);
3340 }
3341 
3342 #undef __FUNC__
3343 #define __FUNC__ "MatPrintHelp"
3344 /*@
3345    MatPrintHelp - Prints all the options for the matrix.
3346 
3347    Collective on Mat
3348 
3349    Input Parameter:
3350 .  mat - the matrix
3351 
3352    Options Database Keys:
3353 +  -help - Prints matrix options
3354 -  -h - Prints matrix options
3355 
3356    Level: developer
3357 
3358 .keywords: mat, help
3359 
3360 .seealso: MatCreate(), MatCreateXXX()
3361 @*/
3362 int MatPrintHelp(Mat mat)
3363 {
3364   static PetscTruth called = PETSC_FALSE;
3365   int               ierr;
3366   MPI_Comm          comm;
3367 
3368   PetscFunctionBegin;
3369   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3370 
3371   comm = mat->comm;
3372   if (!called) {
3373     ierr = (*PetscHelpPrintf)(comm,"General matrix options:\n");CHKERRQ(ierr);
3374     ierr = (*PetscHelpPrintf)(comm,"  -mat_view_info: view basic matrix info during MatAssemblyEnd()\n");CHKERRQ(ierr);
3375     ierr = (*PetscHelpPrintf)(comm,"  -mat_view_info_detailed: view detailed matrix info during MatAssemblyEnd()\n");CHKERRQ(ierr);
3376     ierr = (*PetscHelpPrintf)(comm,"  -mat_view_draw: draw nonzero matrix structure during MatAssemblyEnd()\n");CHKERRQ(ierr);
3377     ierr = (*PetscHelpPrintf)(comm,"      -draw_pause <sec>: set seconds of display pause\n");CHKERRQ(ierr);
3378     ierr = (*PetscHelpPrintf)(comm,"      -display <name>: set alternate display\n");CHKERRQ(ierr);
3379     called = PETSC_TRUE;
3380   }
3381   if (mat->ops->printhelp) {
3382     ierr = (*mat->ops->printhelp)(mat);CHKERRQ(ierr);
3383   }
3384   PetscFunctionReturn(0);
3385 }
3386 
3387 #undef __FUNC__
3388 #define __FUNC__ "MatGetBlockSize"
3389 /*@
3390    MatGetBlockSize - Returns the matrix block size; useful especially for the
3391    block row and block diagonal formats.
3392 
3393    Not Collective
3394 
3395    Input Parameter:
3396 .  mat - the matrix
3397 
3398    Output Parameter:
3399 .  bs - block size
3400 
3401    Notes:
3402    Block diagonal formats are MATSEQBDIAG, MATMPIBDIAG.
3403    Block row formats are MATSEQBAIJ, MATMPIBAIJ
3404 
3405    Level: intermediate
3406 
3407 .keywords: matrix, get, block, size
3408 
3409 .seealso: MatCreateSeqBAIJ(), MatCreateMPIBAIJ(), MatCreateSeqBDiag(), MatCreateMPIBDiag()
3410 @*/
3411 int MatGetBlockSize(Mat mat,int *bs)
3412 {
3413   int ierr;
3414 
3415   PetscFunctionBegin;
3416   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3417   PetscValidIntPointer(bs);
3418   if (!mat->ops->getblocksize) SETERRQ(PETSC_ERR_SUP,0,"");
3419   ierr = (*mat->ops->getblocksize)(mat,bs);CHKERRQ(ierr);
3420   PetscFunctionReturn(0);
3421 }
3422 
3423 #undef __FUNC__
3424 #define __FUNC__ "MatGetRowIJ"
3425 /*@C
3426     MatGetRowIJ - Returns the compressed row storage i and j indices for sequential matrices.
3427 
3428    Collective on Mat
3429 
3430     Input Parameters:
3431 +   mat - the matrix
3432 .   shift -  0 or 1 indicating we want the indices starting at 0 or 1
3433 -   symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
3434                 symmetrized
3435 
3436     Output Parameters:
3437 +   n - number of rows in the (possibly compressed) matrix
3438 .   ia - the row pointers
3439 .   ja - the column indices
3440 -   done - PETSC_TRUE or PETSC_FALSE, indicating whether the values have been returned
3441 
3442     Level: developer
3443 
3444 .seealso: MatGetColumnIJ(), MatRestoreRowIJ()
3445 @*/
3446 int MatGetRowIJ(Mat mat,int shift,PetscTruth symmetric,int *n,int **ia,int** ja,PetscTruth *done)
3447 {
3448   int ierr;
3449 
3450   PetscFunctionBegin;
3451   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3452   if (ia) PetscValidIntPointer(ia);
3453   if (ja) PetscValidIntPointer(ja);
3454   PetscValidIntPointer(done);
3455   if (!mat->ops->getrowij) *done = PETSC_FALSE;
3456   else {
3457     *done = PETSC_TRUE;
3458     ierr  = (*mat->ops->getrowij)(mat,shift,symmetric,n,ia,ja,done);CHKERRQ(ierr);
3459   }
3460   PetscFunctionReturn(0);
3461 }
3462 
3463 #undef __FUNC__
3464 #define __FUNC__ "MatGetColumnIJ"
3465 /*@C
3466     MatGetColumnIJ - Returns the compressed column storage i and j indices for sequential matrices.
3467 
3468     Collective on Mat
3469 
3470     Input Parameters:
3471 +   mat - the matrix
3472 .   shift - 1 or zero indicating we want the indices starting at 0 or 1
3473 -   symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
3474                 symmetrized
3475 
3476     Output Parameters:
3477 +   n - number of columns in the (possibly compressed) matrix
3478 .   ia - the column pointers
3479 .   ja - the row indices
3480 -   done - PETSC_TRUE or PETSC_FALSE, indicating whether the values have been returned
3481 
3482     Level: developer
3483 
3484 .seealso: MatGetRowIJ(), MatRestoreColumnIJ()
3485 @*/
3486 int MatGetColumnIJ(Mat mat,int shift,PetscTruth symmetric,int *n,int **ia,int** ja,PetscTruth *done)
3487 {
3488   int ierr;
3489 
3490   PetscFunctionBegin;
3491   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3492   if (ia) PetscValidIntPointer(ia);
3493   if (ja) PetscValidIntPointer(ja);
3494   PetscValidIntPointer(done);
3495 
3496   if (!mat->ops->getcolumnij) *done = PETSC_FALSE;
3497   else {
3498     *done = PETSC_TRUE;
3499     ierr  = (*mat->ops->getcolumnij)(mat,shift,symmetric,n,ia,ja,done);CHKERRQ(ierr);
3500   }
3501   PetscFunctionReturn(0);
3502 }
3503 
3504 #undef __FUNC__
3505 #define __FUNC__ "MatRestoreRowIJ"
3506 /*@C
3507     MatRestoreRowIJ - Call after you are completed with the ia,ja indices obtained with
3508     MatGetRowIJ().
3509 
3510     Collective on Mat
3511 
3512     Input Parameters:
3513 +   mat - the matrix
3514 .   shift - 1 or zero indicating we want the indices starting at 0 or 1
3515 -   symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
3516                 symmetrized
3517 
3518     Output Parameters:
3519 +   n - size of (possibly compressed) matrix
3520 .   ia - the row pointers
3521 .   ja - the column indices
3522 -   done - PETSC_TRUE or PETSC_FALSE indicated that the values have been returned
3523 
3524     Level: developer
3525 
3526 .seealso: MatGetRowIJ(), MatRestoreColumnIJ()
3527 @*/
3528 int MatRestoreRowIJ(Mat mat,int shift,PetscTruth symmetric,int *n,int **ia,int** ja,PetscTruth *done)
3529 {
3530   int ierr;
3531 
3532   PetscFunctionBegin;
3533   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3534   if (ia) PetscValidIntPointer(ia);
3535   if (ja) PetscValidIntPointer(ja);
3536   PetscValidIntPointer(done);
3537 
3538   if (!mat->ops->restorerowij) *done = PETSC_FALSE;
3539   else {
3540     *done = PETSC_TRUE;
3541     ierr  = (*mat->ops->restorerowij)(mat,shift,symmetric,n,ia,ja,done);CHKERRQ(ierr);
3542   }
3543   PetscFunctionReturn(0);
3544 }
3545 
3546 #undef __FUNC__
3547 #define __FUNC__ "MatRestoreColumnIJ"
3548 /*@C
3549     MatRestoreColumnIJ - Call after you are completed with the ia,ja indices obtained with
3550     MatGetColumnIJ().
3551 
3552     Collective on Mat
3553 
3554     Input Parameters:
3555 +   mat - the matrix
3556 .   shift - 1 or zero indicating we want the indices starting at 0 or 1
3557 -   symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
3558                 symmetrized
3559 
3560     Output Parameters:
3561 +   n - size of (possibly compressed) matrix
3562 .   ia - the column pointers
3563 .   ja - the row indices
3564 -   done - PETSC_TRUE or PETSC_FALSE indicated that the values have been returned
3565 
3566     Level: developer
3567 
3568 .seealso: MatGetColumnIJ(), MatRestoreRowIJ()
3569 @*/
3570 int MatRestoreColumnIJ(Mat mat,int shift,PetscTruth symmetric,int *n,int **ia,int** ja,PetscTruth *done)
3571 {
3572   int ierr;
3573 
3574   PetscFunctionBegin;
3575   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3576   if (ia) PetscValidIntPointer(ia);
3577   if (ja) PetscValidIntPointer(ja);
3578   PetscValidIntPointer(done);
3579 
3580   if (!mat->ops->restorecolumnij) *done = PETSC_FALSE;
3581   else {
3582     *done = PETSC_TRUE;
3583     ierr  = (*mat->ops->restorecolumnij)(mat,shift,symmetric,n,ia,ja,done);CHKERRQ(ierr);
3584   }
3585   PetscFunctionReturn(0);
3586 }
3587 
3588 #undef __FUNC__
3589 #define __FUNC__ "MatColoringPatch"
3590 /*@C
3591     MatColoringPatch -Used inside matrix coloring routines that
3592     use MatGetRowIJ() and/or MatGetColumnIJ().
3593 
3594     Collective on Mat
3595 
3596     Input Parameters:
3597 +   mat - the matrix
3598 .   n   - number of colors
3599 -   colorarray - array indicating color for each column
3600 
3601     Output Parameters:
3602 .   iscoloring - coloring generated using colorarray information
3603 
3604     Level: developer
3605 
3606 .seealso: MatGetRowIJ(), MatGetColumnIJ()
3607 
3608 .keywords: mat, coloring, patch
3609 @*/
3610 int MatColoringPatch(Mat mat,int n,int *colorarray,ISColoring *iscoloring)
3611 {
3612   int ierr;
3613 
3614   PetscFunctionBegin;
3615   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3616   PetscValidIntPointer(colorarray);
3617 
3618   if (!mat->ops->coloringpatch) {SETERRQ(PETSC_ERR_SUP,0,"");}
3619   else {
3620     ierr  = (*mat->ops->coloringpatch)(mat,n,colorarray,iscoloring);CHKERRQ(ierr);
3621   }
3622   PetscFunctionReturn(0);
3623 }
3624 
3625 
3626 #undef __FUNC__
3627 #define __FUNC__ "MatSetUnfactored"
3628 /*@
3629    MatSetUnfactored - Resets a factored matrix to be treated as unfactored.
3630 
3631    Collective on Mat
3632 
3633    Input Parameter:
3634 .  mat - the factored matrix to be reset
3635 
3636    Notes:
3637    This routine should be used only with factored matrices formed by in-place
3638    factorization via ILU(0) (or by in-place LU factorization for the MATSEQDENSE
3639    format).  This option can save memory, for example, when solving nonlinear
3640    systems with a matrix-free Newton-Krylov method and a matrix-based, in-place
3641    ILU(0) preconditioner.
3642 
3643    Note that one can specify in-place ILU(0) factorization by calling
3644 .vb
3645      PCType(pc,PCILU);
3646      PCILUSeUseInPlace(pc);
3647 .ve
3648    or by using the options -pc_type ilu -pc_ilu_in_place
3649 
3650    In-place factorization ILU(0) can also be used as a local
3651    solver for the blocks within the block Jacobi or additive Schwarz
3652    methods (runtime option: -sub_pc_ilu_in_place).  See the discussion
3653    of these preconditioners in the users manual for details on setting
3654    local solver options.
3655 
3656    Most users should employ the simplified SLES interface for linear solvers
3657    instead of working directly with matrix algebra routines such as this.
3658    See, e.g., SLESCreate().
3659 
3660    Level: developer
3661 
3662 .seealso: PCILUSetUseInPlace(), PCLUSetUseInPlace()
3663 
3664 .keywords: matrix-free, in-place ILU, in-place LU
3665 @*/
3666 int MatSetUnfactored(Mat mat)
3667 {
3668   int ierr;
3669 
3670   PetscFunctionBegin;
3671   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3672   mat->factor = 0;
3673   if (!mat->ops->setunfactored) PetscFunctionReturn(0);
3674   ierr = (*mat->ops->setunfactored)(mat);CHKERRQ(ierr);
3675   PetscFunctionReturn(0);
3676 }
3677 
3678 #undef __FUNC__
3679 #define __FUNC__ "MatGetType"
3680 /*@C
3681    MatGetType - Gets the matrix type and name (as a string) from the matrix.
3682 
3683    Not Collective
3684 
3685    Input Parameter:
3686 .  mat - the matrix
3687 
3688    Output Parameter:
3689 +  type - the matrix type (or use PETSC_NULL)
3690 -  name - name of matrix type (or use PETSC_NULL)
3691 
3692    Level: intermediate
3693 
3694 .keywords: matrix, get, type, name
3695 @*/
3696 int MatGetType(Mat mat,MatType *type,char **name)
3697 {
3698   int  itype = (int)mat->type;
3699   char *matname[10];
3700 
3701   PetscFunctionBegin;
3702   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3703 
3704   if (type) *type = (MatType) mat->type;
3705   if (name) {
3706     /* Note:  Be sure that this list corresponds to the enum in mat.h */
3707     matname[0] = "MATSEQDENSE";
3708     matname[1] = "MATSEQAIJ";
3709     matname[2] = "MATMPIAIJ";
3710     matname[3] = "MATSHELL";
3711     matname[4] = "MATMPIROWBS";
3712     matname[5] = "MATSEQBDIAG";
3713     matname[6] = "MATMPIBDIAG";
3714     matname[7] = "MATMPIDENSE";
3715     matname[8] = "MATSEQBAIJ";
3716     matname[9] = "MATMPIBAIJ";
3717 
3718     if (itype < 0 || itype > 9) *name = "Unknown matrix type";
3719     else                        *name = matname[itype];
3720   }
3721   PetscFunctionReturn(0);
3722 }
3723 
3724 /*MC
3725     MatGetArrayF90 - Accesses a matrix array from Fortran90.
3726 
3727     Synopsis:
3728     MatGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
3729 
3730     Not collective
3731 
3732     Input Parameter:
3733 .   x - matrix
3734 
3735     Output Parameters:
3736 +   xx_v - the Fortran90 pointer to the array
3737 -   ierr - error code
3738 
3739     Example of Usage:
3740 .vb
3741       Scalar, pointer xx_v(:)
3742       ....
3743       call MatGetArrayF90(x,xx_v,ierr)
3744       a = xx_v(3)
3745       call MatRestoreArrayF90(x,xx_v,ierr)
3746 .ve
3747 
3748     Notes:
3749     Not yet supported for all F90 compilers
3750 
3751     Level: advanced
3752 
3753 .seealso:  MatRestoreArrayF90(), MatGetArray(), MatRestoreArray()
3754 
3755 .keywords:  matrix, array, f90
3756 M*/
3757 
3758 /*MC
3759     MatRestoreArrayF90 - Restores a matrix array that has been
3760     accessed with MatGetArrayF90().
3761 
3762     Synopsis:
3763     MatRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
3764 
3765     Not collective
3766 
3767     Input Parameters:
3768 +   x - matrix
3769 -   xx_v - the Fortran90 pointer to the array
3770 
3771     Output Parameter:
3772 .   ierr - error code
3773 
3774     Example of Usage:
3775 .vb
3776        Scalar, pointer xx_v(:)
3777        ....
3778        call MatGetArrayF90(x,xx_v,ierr)
3779        a = xx_v(3)
3780        call MatRestoreArrayF90(x,xx_v,ierr)
3781 .ve
3782 
3783     Notes:
3784     Not yet supported for all F90 compilers
3785 
3786     Level: advanced
3787 
3788 .seealso:  MatGetArrayF90(), MatGetArray(), MatRestoreArray()
3789 
3790 .keywords:  matrix, array, f90
3791 M*/
3792 
3793 
3794 #undef __FUNC__
3795 #define __FUNC__ "MatGetSubMatrix"
3796 /*@
3797     MatGetSubMatrix - Gets a single submatrix on the same number of processors
3798                       as the original matrix.
3799 
3800     Collective on Mat
3801 
3802     Input Parameters:
3803 +   mat - the original matrix
3804 .   isrow - rows this processor should obtain
3805 .   iscol - columns for all processors you wish to keep
3806 .   csize - number of columns "local" to this processor (does nothing for sequential
3807             matrices). This should match the result from VecGetLocalSize(x,...) if you
3808             plan to use the matrix in a A*x; alternatively, you can use PETSC_DECIDE
3809 -   cll - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
3810 
3811     Output Parameter:
3812 .   newmat - the new submatrix, of the same type as the old
3813 
3814     Level: advanced
3815 
3816 .keywords: matrix, get, submatrix, submatrices
3817 
3818 .seealso: MatGetSubMatrices()
3819 @*/
3820 int MatGetSubMatrix(Mat mat,IS isrow,IS iscol,int csize,MatReuse cll,Mat *newmat)
3821 {
3822   int     ierr, size;
3823   Mat     *local;
3824 
3825   PetscFunctionBegin;
3826   ierr = MPI_Comm_size(mat->comm,&size);CHKERRQ(ierr);
3827 
3828   /* if original matrix is on just one processor then use submatrix generated */
3829   if (!mat->ops->getsubmatrix && size == 1 && cll == MAT_REUSE_MATRIX) {
3830     ierr = MatGetSubMatrices(mat,1,&isrow,&iscol,MAT_REUSE_MATRIX,&newmat);CHKERRQ(ierr);
3831     PetscFunctionReturn(0);
3832   } else if (!mat->ops->getsubmatrix && size == 1) {
3833     ierr = MatGetSubMatrices(mat,1,&isrow,&iscol,MAT_INITIAL_MATRIX,&local);CHKERRQ(ierr);
3834     *newmat = *local;
3835     ierr = PetscFree(local);CHKERRQ(ierr);
3836     PetscFunctionReturn(0);
3837   }
3838 
3839   if (!mat->ops->getsubmatrix) SETERRQ(PETSC_ERR_SUP,0,"Not currently implemented");
3840   ierr = (*mat->ops->getsubmatrix)(mat,isrow,iscol,csize,cll,newmat);CHKERRQ(ierr);
3841   PetscFunctionReturn(0);
3842 }
3843 
3844 #undef __FUNC__
3845 #define __FUNC__ "MatGetMaps"
3846 /*@C
3847    MatGetMaps - Returns the maps associated with the matrix.
3848 
3849    Not Collective
3850 
3851    Input Parameter:
3852 .  mat - the matrix
3853 
3854    Output Parameters:
3855 +  rmap - the row (right) map
3856 -  cmap - the column (left) map
3857 
3858    Level: developer
3859 
3860 .keywords: matrix, get, map
3861 @*/
3862 int MatGetMaps(Mat mat,Map *rmap,Map *cmap)
3863 {
3864   int ierr;
3865 
3866   PetscFunctionBegin;
3867   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3868   ierr = (*mat->ops->getmaps)(mat,rmap,cmap);CHKERRQ(ierr);
3869   PetscFunctionReturn(0);
3870 }
3871 
3872 /*
3873       Version that works for all PETSc matrices
3874 */
3875 #undef __FUNC__
3876 #define __FUNC__ "MatGetMaps_Petsc"
3877 int MatGetMaps_Petsc(Mat mat,Map *rmap,Map *cmap)
3878 {
3879   PetscFunctionBegin;
3880   if (rmap) *rmap = mat->rmap;
3881   if (cmap) *cmap = mat->cmap;
3882   PetscFunctionReturn(0);
3883 }
3884 
3885 #undef __FUNC__
3886 #define __FUNC__ "MatSetStashInitialSize"
3887 /*@
3888    MatSetStashInitialSize - sets the sizes of the matrix stash, that is
3889    used during the assembly process to store values that belong to
3890    other processors.
3891 
3892    Not Collective
3893 
3894    Input Parameters:
3895 +  mat   - the matrix
3896 .  size  - the initial size of the stash.
3897 -  bsize - the initial size of the block-stash(if used).
3898 
3899    Options Database Keys:
3900 +   -matstash_initial_size <size> or <size0,size1,...sizep-1>
3901 -   -matstash_block_initial_size <bsize>  or <bsize0,bsize1,...bsizep-1>
3902 
3903    Level: intermediate
3904 
3905    Notes:
3906      The block-stash is used for values set with VecSetValuesBlocked() while
3907      the stash is used for values set with VecSetValues()
3908 
3909      Run with the option -log_info and look for output of the form
3910      MatAssemblyBegin_MPIXXX:Stash has MM entries, uses nn mallocs.
3911      to determine the appropriate value, MM, to use for size and
3912      MatAssemblyBegin_MPIXXX:Block-Stash has BMM entries, uses nn mallocs.
3913      to determine the value, BMM to use for bsize
3914 
3915 .keywords: matrix, stash, assembly
3916 @*/
3917 int MatSetStashInitialSize(Mat mat,int size, int bsize)
3918 {
3919   int ierr;
3920 
3921   PetscFunctionBegin;
3922   PetscValidHeaderSpecific(mat,MAT_COOKIE);
3923   ierr = MatStashSetInitialSize_Private(&mat->stash,size);CHKERRQ(ierr);
3924   ierr = MatStashSetInitialSize_Private(&mat->bstash,bsize);CHKERRQ(ierr);
3925   PetscFunctionReturn(0);
3926 }
3927 
3928 #undef __FUNC__
3929 #define __FUNC__ "MatInterpolateAdd"
3930 /*@
3931    MatInterpolateAdd - w = y + A*x or A'*x depending on the shape of
3932      the matrix
3933 
3934    Collective on Mat
3935 
3936    Input Parameters:
3937 +  mat   - the matrix
3938 .  x,y - the vectors
3939 -  w - where the result is stored
3940 
3941    Level: intermediate
3942 
3943    Notes:
3944     w may be the same vector as y.
3945 
3946     This allows one to use either the restriction or interpolation (its transpose)
3947     matrix to do the interpolation
3948 
3949 .keywords: interpolate,
3950 
3951 .seealso: MatMultAdd(), MatMultTransposeAdd(), MatRestrict()
3952 
3953 @*/
3954 int MatInterpolateAdd(Mat A,Vec x,Vec y,Vec w)
3955 {
3956   int M,N,ierr;
3957 
3958   PetscFunctionBegin;
3959   ierr = MatGetSize(A,&M,&N);CHKERRQ(ierr);
3960   if (N > M) {
3961     ierr = MatMultTransposeAdd(A,x,y,w);CHKERRQ(ierr);
3962   } else {
3963     ierr = MatMultAdd(A,x,y,w);CHKERRQ(ierr);
3964   }
3965   PetscFunctionReturn(0);
3966 }
3967 
3968 #undef __FUNC__
3969 #define __FUNC__ "MatInterpolate"
3970 /*@
3971    MatInterpolate - y = A*x or A'*x depending on the shape of
3972      the matrix
3973 
3974    Collective on Mat
3975 
3976    Input Parameters:
3977 +  mat   - the matrix
3978 -  x,y - the vectors
3979 
3980    Level: intermediate
3981 
3982    Notes:
3983     This allows one to use either the restriction or interpolation (its transpose)
3984     matrix to do the interpolation
3985 
3986 .keywords: interpolate,
3987 
3988 .seealso: MatMultAdd(), MatMultTransposeAdd(), MatRestrict()
3989 
3990 @*/
3991 int MatInterpolate(Mat A,Vec x,Vec y)
3992 {
3993   int M,N,ierr;
3994 
3995   PetscFunctionBegin;
3996   ierr = MatGetSize(A,&M,&N);CHKERRQ(ierr);
3997   if (N > M) {
3998     ierr = MatMultTranspose(A,x,y);CHKERRQ(ierr);
3999   } else {
4000     ierr = MatMult(A,x,y);CHKERRQ(ierr);
4001   }
4002   PetscFunctionReturn(0);
4003 }
4004 
4005 #undef __FUNC__
4006 #define __FUNC__ "MatRestrict"
4007 /*@
4008    MatRestrict - y = A*x or A'*x
4009 
4010    Collective on Mat
4011 
4012    Input Parameters:
4013 +  mat   - the matrix
4014 -  x,y - the vectors
4015 
4016    Level: intermediate
4017 
4018    Notes:
4019     This allows one to use either the restriction or interpolation (its transpose)
4020     matrix to do the restriction
4021 
4022 .keywords: interpolate,
4023 
4024 .seealso: MatMultAdd(), MatMultTransposeAdd(), MatInterpolate()
4025 
4026 @*/
4027 int MatRestrict(Mat A,Vec x,Vec y)
4028 {
4029   int M,N,ierr;
4030 
4031   PetscFunctionBegin;
4032   ierr = MatGetSize(A,&M,&N);CHKERRQ(ierr);
4033   if (N > M) {
4034     ierr = MatMult(A,x,y);CHKERRQ(ierr);
4035   } else {
4036     ierr = MatMultTranspose(A,x,y);CHKERRQ(ierr);
4037   }
4038   PetscFunctionReturn(0);
4039 }
4040