xref: /petsc/src/mat/interface/matrix.c (revision c8ecec4179e20ade7121e54035569ae47bffa58a)
1 #define PETSCMAT_DLL
2 
3 /*
4    This is where the abstract matrix operations are defined
5 */
6 
7 #include "private/matimpl.h"        /*I "petscmat.h" I*/
8 #include "private/vecimpl.h"
9 
10 /* Logging support */
11 PetscCookie PETSCMAT_DLLEXPORT MAT_COOKIE;
12 PetscCookie PETSCMAT_DLLEXPORT MAT_FDCOLORING_COOKIE;
13 
14 PetscLogEvent  MAT_Mult, MAT_Mults, MAT_MultConstrained, MAT_MultAdd, MAT_MultTranspose;
15 PetscLogEvent  MAT_MultTransposeConstrained, MAT_MultTransposeAdd, MAT_Solve, MAT_Solves, MAT_SolveAdd, MAT_SolveTranspose, MAT_MatSolve;
16 PetscLogEvent  MAT_SolveTransposeAdd, MAT_Relax, MAT_ForwardSolve, MAT_BackwardSolve, MAT_LUFactor, MAT_LUFactorSymbolic;
17 PetscLogEvent  MAT_LUFactorNumeric, MAT_CholeskyFactor, MAT_CholeskyFactorSymbolic, MAT_CholeskyFactorNumeric, MAT_ILUFactor;
18 PetscLogEvent  MAT_ILUFactorSymbolic, MAT_ICCFactorSymbolic, MAT_Copy, MAT_Convert, MAT_Scale, MAT_AssemblyBegin;
19 PetscLogEvent  MAT_AssemblyEnd, MAT_SetValues, MAT_GetValues, MAT_GetRow, MAT_GetRowIJ, MAT_GetSubMatrices, MAT_GetColoring, MAT_GetOrdering, MAT_GetRedundantMatrix, MAT_GetSeqNonzeroStructure;
20 PetscLogEvent  MAT_IncreaseOverlap, MAT_Partitioning, MAT_ZeroEntries, MAT_Load, MAT_View, MAT_AXPY, MAT_FDColoringCreate;
21 PetscLogEvent  MAT_FDColoringApply,MAT_Transpose,MAT_FDColoringFunction;
22 PetscLogEvent  MAT_MatMult, MAT_MatMultSymbolic, MAT_MatMultNumeric;
23 PetscLogEvent  MAT_PtAP, MAT_PtAPSymbolic, MAT_PtAPNumeric;
24 PetscLogEvent  MAT_MatMultTranspose, MAT_MatMultTransposeSymbolic, MAT_MatMultTransposeNumeric;
25 PetscLogEvent  MAT_Getsymtranspose, MAT_Getsymtransreduced, MAT_Transpose_SeqAIJ, MAT_GetBrowsOfAcols;
26 PetscLogEvent  MAT_GetBrowsOfAocols, MAT_Getlocalmat, MAT_Getlocalmatcondensed, MAT_Seqstompi, MAT_Seqstompinum, MAT_Seqstompisym;
27 PetscLogEvent  MAT_Applypapt, MAT_Applypapt_numeric, MAT_Applypapt_symbolic, MAT_GetSequentialNonzeroStructure;
28 PetscLogEvent  MAT_ILUDTFactorSymbolic, MAT_ILUDTFactorNumeric;
29 
30 /* nasty global values for MatSetValue() */
31 PetscInt    PETSCMAT_DLLEXPORT MatSetValue_Row = 0;
32 PetscInt    PETSCMAT_DLLEXPORT MatSetValue_Column = 0;
33 PetscScalar PETSCMAT_DLLEXPORT MatSetValue_Value = 0.0;
34 
35 #undef __FUNCT__
36 #define __FUNCT__ "MatGetDiagonalBlock"
37 /*@
38    MatGetDiagonalBlock - Returns the part of the matrix associated with the on-process coupling
39 
40    Not Collective
41 
42    Input Parameters:
43 +  mat - the matrix
44 -  reuse - indicates you are passing in the a matrix and want it reused
45 
46    Output Parameters:
47 +   iscopy - indicates a copy of the diagonal matrix was created and you should use MatDestroy() on it
48 -   a - the diagonal part (which is a SEQUENTIAL matrix)
49 
50    Notes: see the manual page for MatCreateMPIAIJ() for more information on the "diagonal part" of the matrix
51 
52    Level: advanced
53 
54 @*/
55 PetscErrorCode PETSCMAT_DLLEXPORT MatGetDiagonalBlock(Mat A,PetscTruth *iscopy,MatReuse reuse,Mat *a)
56 {
57   PetscErrorCode ierr,(*f)(Mat,PetscTruth*,MatReuse,Mat*);
58   PetscMPIInt    size;
59 
60   PetscFunctionBegin;
61   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
62   PetscValidType(A,1);
63   PetscValidPointer(iscopy,2);
64   PetscValidPointer(a,3);
65   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
66   if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
67   ierr = MPI_Comm_size(((PetscObject)A)->comm,&size);CHKERRQ(ierr);
68   ierr = PetscObjectQueryFunction((PetscObject)A,"MatGetDiagonalBlock_C",(void (**)(void))&f);CHKERRQ(ierr);
69   if (f) {
70     ierr = (*f)(A,iscopy,reuse,a);CHKERRQ(ierr);
71   } else if (size == 1) {
72     *a = A;
73     *iscopy = PETSC_FALSE;
74   } else {
75     SETERRQ(PETSC_ERR_SUP,"Cannot get diagonal part for this matrix");
76   }
77   PetscFunctionReturn(0);
78 }
79 
80 #undef __FUNCT__
81 #define __FUNCT__ "MatRealPart"
82 /*@
83    MatRealPart - Zeros out the imaginary part of the matrix
84 
85    Collective on Mat
86 
87    Input Parameters:
88 .  mat - the matrix
89 
90    Level: advanced
91 
92 
93 .seealso: MatImaginaryPart()
94 @*/
95 PetscErrorCode PETSCMAT_DLLEXPORT MatRealPart(Mat mat)
96 {
97   PetscErrorCode ierr;
98 
99   PetscFunctionBegin;
100   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
101   PetscValidType(mat,1);
102   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
103   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
104   if (!mat->ops->realpart) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
105   ierr = MatPreallocated(mat);CHKERRQ(ierr);
106   ierr = (*mat->ops->realpart)(mat);CHKERRQ(ierr);
107   PetscFunctionReturn(0);
108 }
109 
110 #undef __FUNCT__
111 #define __FUNCT__ "MatGetGhosts"
112 /*@C
113    MatGetGhosts - Get the global index of all ghost nodes defined by the sparse matrix
114 
115    Collective on Mat
116 
117    Input Parameter:
118 .  mat - the matrix
119 
120    Output Parameters:
121 +   nghosts - number of ghosts (note for BAIJ matrices there is one ghost for each block)
122 -   ghosts - the global indices of the ghost points
123 
124    Notes: the nghosts and ghosts are suitable to pass into VecCreateGhost()
125 
126    Level: advanced
127 
128 @*/
129 PetscErrorCode PETSCMAT_DLLEXPORT MatGetGhosts(Mat mat,PetscInt *nghosts,const PetscInt *ghosts[])
130 {
131   PetscErrorCode ierr;
132 
133   PetscFunctionBegin;
134   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
135   PetscValidType(mat,1);
136   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
137   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
138   if (!mat->ops->getghosts) {
139     if (nghosts) *nghosts = 0;
140     if (ghosts) *ghosts = 0;
141   } else {
142     ierr = (*mat->ops->getghosts)(mat,nghosts,ghosts);CHKERRQ(ierr);
143   }
144   PetscFunctionReturn(0);
145 }
146 
147 
148 #undef __FUNCT__
149 #define __FUNCT__ "MatImaginaryPart"
150 /*@
151    MatImaginaryPart - Moves the imaginary part of the matrix to the real part and zeros the imaginary part
152 
153    Collective on Mat
154 
155    Input Parameters:
156 .  mat - the matrix
157 
158    Level: advanced
159 
160 
161 .seealso: MatRealPart()
162 @*/
163 PetscErrorCode PETSCMAT_DLLEXPORT MatImaginaryPart(Mat mat)
164 {
165   PetscErrorCode ierr;
166 
167   PetscFunctionBegin;
168   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
169   PetscValidType(mat,1);
170   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
171   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
172   if (!mat->ops->imaginarypart) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
173   ierr = MatPreallocated(mat);CHKERRQ(ierr);
174   ierr = (*mat->ops->imaginarypart)(mat);CHKERRQ(ierr);
175   PetscFunctionReturn(0);
176 }
177 
178 #undef __FUNCT__
179 #define __FUNCT__ "MatMissingDiagonal"
180 /*@
181    MatMissingDiagonal - Determine if sparse matrix is missing a diagonal entry (or block entry for BAIJ matrices)
182 
183    Collective on Mat
184 
185    Input Parameter:
186 .  mat - the matrix
187 
188    Output Parameters:
189 +  missing - is any diagonal missing
190 -  dd - first diagonal entry that is missing (optional)
191 
192    Level: advanced
193 
194 
195 .seealso: MatRealPart()
196 @*/
197 PetscErrorCode PETSCMAT_DLLEXPORT MatMissingDiagonal(Mat mat,PetscTruth *missing,PetscInt *dd)
198 {
199   PetscErrorCode ierr;
200 
201   PetscFunctionBegin;
202   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
203   PetscValidType(mat,1);
204   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
205   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
206   if (!mat->ops->missingdiagonal) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
207   ierr = (*mat->ops->missingdiagonal)(mat,missing,dd);CHKERRQ(ierr);
208   PetscFunctionReturn(0);
209 }
210 
211 #undef __FUNCT__
212 #define __FUNCT__ "MatGetRow"
213 /*@C
214    MatGetRow - Gets a row of a matrix.  You MUST call MatRestoreRow()
215    for each row that you get to ensure that your application does
216    not bleed memory.
217 
218    Not Collective
219 
220    Input Parameters:
221 +  mat - the matrix
222 -  row - the row to get
223 
224    Output Parameters:
225 +  ncols -  if not NULL, the number of nonzeros in the row
226 .  cols - if not NULL, the column numbers
227 -  vals - if not NULL, the values
228 
229    Notes:
230    This routine is provided for people who need to have direct access
231    to the structure of a matrix.  We hope that we provide enough
232    high-level matrix routines that few users will need it.
233 
234    MatGetRow() always returns 0-based column indices, regardless of
235    whether the internal representation is 0-based (default) or 1-based.
236 
237    For better efficiency, set cols and/or vals to PETSC_NULL if you do
238    not wish to extract these quantities.
239 
240    The user can only examine the values extracted with MatGetRow();
241    the values cannot be altered.  To change the matrix entries, one
242    must use MatSetValues().
243 
244    You can only have one call to MatGetRow() outstanding for a particular
245    matrix at a time, per processor. MatGetRow() can only obtain rows
246    associated with the given processor, it cannot get rows from the
247    other processors; for that we suggest using MatGetSubMatrices(), then
248    MatGetRow() on the submatrix. The row indix passed to MatGetRows()
249    is in the global number of rows.
250 
251    Fortran Notes:
252    The calling sequence from Fortran is
253 .vb
254    MatGetRow(matrix,row,ncols,cols,values,ierr)
255          Mat     matrix (input)
256          integer row    (input)
257          integer ncols  (output)
258          integer cols(maxcols) (output)
259          double precision (or double complex) values(maxcols) output
260 .ve
261    where maxcols >= maximum nonzeros in any row of the matrix.
262 
263 
264    Caution:
265    Do not try to change the contents of the output arrays (cols and vals).
266    In some cases, this may corrupt the matrix.
267 
268    Level: advanced
269 
270    Concepts: matrices^row access
271 
272 .seealso: MatRestoreRow(), MatSetValues(), MatGetValues(), MatGetSubMatrices(), MatGetDiagonal()
273 @*/
274 PetscErrorCode PETSCMAT_DLLEXPORT MatGetRow(Mat mat,PetscInt row,PetscInt *ncols,const PetscInt *cols[],const PetscScalar *vals[])
275 {
276   PetscErrorCode ierr;
277   PetscInt       incols;
278 
279   PetscFunctionBegin;
280   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
281   PetscValidType(mat,1);
282   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
283   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
284   if (!mat->ops->getrow) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
285   ierr = MatPreallocated(mat);CHKERRQ(ierr);
286   ierr = PetscLogEventBegin(MAT_GetRow,mat,0,0,0);CHKERRQ(ierr);
287   ierr = (*mat->ops->getrow)(mat,row,&incols,(PetscInt **)cols,(PetscScalar **)vals);CHKERRQ(ierr);
288   if (ncols) *ncols = incols;
289   ierr = PetscLogEventEnd(MAT_GetRow,mat,0,0,0);CHKERRQ(ierr);
290   PetscFunctionReturn(0);
291 }
292 
293 #undef __FUNCT__
294 #define __FUNCT__ "MatConjugate"
295 /*@
296    MatConjugate - replaces the matrix values with their complex conjugates
297 
298    Collective on Mat
299 
300    Input Parameters:
301 .  mat - the matrix
302 
303    Level: advanced
304 
305 .seealso:  VecConjugate()
306 @*/
307 PetscErrorCode PETSCMAT_DLLEXPORT MatConjugate(Mat mat)
308 {
309   PetscErrorCode ierr;
310 
311   PetscFunctionBegin;
312   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
313   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
314   if (!mat->ops->conjugate) SETERRQ(PETSC_ERR_SUP,"Not provided for this matrix format, send email to petsc-maint@mcs.anl.gov");
315   ierr = (*mat->ops->conjugate)(mat);CHKERRQ(ierr);
316   PetscFunctionReturn(0);
317 }
318 
319 #undef __FUNCT__
320 #define __FUNCT__ "MatRestoreRow"
321 /*@C
322    MatRestoreRow - Frees any temporary space allocated by MatGetRow().
323 
324    Not Collective
325 
326    Input Parameters:
327 +  mat - the matrix
328 .  row - the row to get
329 .  ncols, cols - the number of nonzeros and their columns
330 -  vals - if nonzero the column values
331 
332    Notes:
333    This routine should be called after you have finished examining the entries.
334 
335    Fortran Notes:
336    The calling sequence from Fortran is
337 .vb
338    MatRestoreRow(matrix,row,ncols,cols,values,ierr)
339       Mat     matrix (input)
340       integer row    (input)
341       integer ncols  (output)
342       integer cols(maxcols) (output)
343       double precision (or double complex) values(maxcols) output
344 .ve
345    Where maxcols >= maximum nonzeros in any row of the matrix.
346 
347    In Fortran MatRestoreRow() MUST be called after MatGetRow()
348    before another call to MatGetRow() can be made.
349 
350    Level: advanced
351 
352 .seealso:  MatGetRow()
353 @*/
354 PetscErrorCode PETSCMAT_DLLEXPORT MatRestoreRow(Mat mat,PetscInt row,PetscInt *ncols,const PetscInt *cols[],const PetscScalar *vals[])
355 {
356   PetscErrorCode ierr;
357 
358   PetscFunctionBegin;
359   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
360   PetscValidIntPointer(ncols,3);
361   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
362   if (!mat->ops->restorerow) PetscFunctionReturn(0);
363   ierr = (*mat->ops->restorerow)(mat,row,ncols,(PetscInt **)cols,(PetscScalar **)vals);CHKERRQ(ierr);
364   PetscFunctionReturn(0);
365 }
366 
367 #undef __FUNCT__
368 #define __FUNCT__ "MatGetRowUpperTriangular"
369 /*@
370    MatGetRowUpperTriangular - Sets a flag to enable calls to MatGetRow() for matrix in MATSBAIJ format.
371    You should call MatRestoreRowUpperTriangular() after calling MatGetRow/MatRestoreRow() to disable the flag.
372 
373    Not Collective
374 
375    Input Parameters:
376 +  mat - the matrix
377 
378    Notes:
379    The flag is to ensure that users are aware of MatGetRow() only provides the upper trianglular part of the row for the matrices in MATSBAIJ format.
380 
381    Level: advanced
382 
383    Concepts: matrices^row access
384 
385 .seealso: MatRestoreRowRowUpperTriangular()
386 @*/
387 PetscErrorCode PETSCMAT_DLLEXPORT MatGetRowUpperTriangular(Mat mat)
388 {
389   PetscErrorCode ierr;
390 
391   PetscFunctionBegin;
392   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
393   PetscValidType(mat,1);
394   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
395   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
396   if (!mat->ops->getrowuppertriangular) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
397   ierr = MatPreallocated(mat);CHKERRQ(ierr);
398   ierr = (*mat->ops->getrowuppertriangular)(mat);CHKERRQ(ierr);
399   PetscFunctionReturn(0);
400 }
401 
402 #undef __FUNCT__
403 #define __FUNCT__ "MatRestoreRowUpperTriangular"
404 /*@
405    MatRestoreRowUpperTriangular - Disable calls to MatGetRow() for matrix in MATSBAIJ format.
406 
407    Not Collective
408 
409    Input Parameters:
410 +  mat - the matrix
411 
412    Notes:
413    This routine should be called after you have finished MatGetRow/MatRestoreRow().
414 
415 
416    Level: advanced
417 
418 .seealso:  MatGetRowUpperTriangular()
419 @*/
420 PetscErrorCode PETSCMAT_DLLEXPORT MatRestoreRowUpperTriangular(Mat mat)
421 {
422   PetscErrorCode ierr;
423 
424   PetscFunctionBegin;
425   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
426   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
427   if (!mat->ops->restorerowuppertriangular) PetscFunctionReturn(0);
428   ierr = (*mat->ops->restorerowuppertriangular)(mat);CHKERRQ(ierr);
429   PetscFunctionReturn(0);
430 }
431 
432 #undef __FUNCT__
433 #define __FUNCT__ "MatSetOptionsPrefix"
434 /*@C
435    MatSetOptionsPrefix - Sets the prefix used for searching for all
436    Mat options in the database.
437 
438    Collective on Mat
439 
440    Input Parameter:
441 +  A - the Mat context
442 -  prefix - the prefix to prepend to all option names
443 
444    Notes:
445    A hyphen (-) must NOT be given at the beginning of the prefix name.
446    The first character of all runtime options is AUTOMATICALLY the hyphen.
447 
448    Level: advanced
449 
450 .keywords: Mat, set, options, prefix, database
451 
452 .seealso: MatSetFromOptions()
453 @*/
454 PetscErrorCode PETSCMAT_DLLEXPORT MatSetOptionsPrefix(Mat A,const char prefix[])
455 {
456   PetscErrorCode ierr;
457 
458   PetscFunctionBegin;
459   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
460   ierr = PetscObjectSetOptionsPrefix((PetscObject)A,prefix);CHKERRQ(ierr);
461   PetscFunctionReturn(0);
462 }
463 
464 #undef __FUNCT__
465 #define __FUNCT__ "MatAppendOptionsPrefix"
466 /*@C
467    MatAppendOptionsPrefix - Appends to the prefix used for searching for all
468    Mat options in the database.
469 
470    Collective on Mat
471 
472    Input Parameters:
473 +  A - the Mat context
474 -  prefix - the prefix to prepend to all option names
475 
476    Notes:
477    A hyphen (-) must NOT be given at the beginning of the prefix name.
478    The first character of all runtime options is AUTOMATICALLY the hyphen.
479 
480    Level: advanced
481 
482 .keywords: Mat, append, options, prefix, database
483 
484 .seealso: MatGetOptionsPrefix()
485 @*/
486 PetscErrorCode PETSCMAT_DLLEXPORT MatAppendOptionsPrefix(Mat A,const char prefix[])
487 {
488   PetscErrorCode ierr;
489 
490   PetscFunctionBegin;
491   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
492   ierr = PetscObjectAppendOptionsPrefix((PetscObject)A,prefix);CHKERRQ(ierr);
493   PetscFunctionReturn(0);
494 }
495 
496 #undef __FUNCT__
497 #define __FUNCT__ "MatGetOptionsPrefix"
498 /*@C
499    MatGetOptionsPrefix - Sets the prefix used for searching for all
500    Mat options in the database.
501 
502    Not Collective
503 
504    Input Parameter:
505 .  A - the Mat context
506 
507    Output Parameter:
508 .  prefix - pointer to the prefix string used
509 
510    Notes: On the fortran side, the user should pass in a string 'prefix' of
511    sufficient length to hold the prefix.
512 
513    Level: advanced
514 
515 .keywords: Mat, get, options, prefix, database
516 
517 .seealso: MatAppendOptionsPrefix()
518 @*/
519 PetscErrorCode PETSCMAT_DLLEXPORT MatGetOptionsPrefix(Mat A,const char *prefix[])
520 {
521   PetscErrorCode ierr;
522 
523   PetscFunctionBegin;
524   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
525   ierr = PetscObjectGetOptionsPrefix((PetscObject)A,prefix);CHKERRQ(ierr);
526   PetscFunctionReturn(0);
527 }
528 
529 #undef __FUNCT__
530 #define __FUNCT__ "MatSetUp"
531 /*@
532    MatSetUp - Sets up the internal matrix data structures for the later use.
533 
534    Collective on Mat
535 
536    Input Parameters:
537 .  A - the Mat context
538 
539    Notes:
540    For basic use of the Mat classes the user need not explicitly call
541    MatSetUp(), since these actions will happen automatically.
542 
543    Level: advanced
544 
545 .keywords: Mat, setup
546 
547 .seealso: MatCreate(), MatDestroy()
548 @*/
549 PetscErrorCode PETSCMAT_DLLEXPORT MatSetUp(Mat A)
550 {
551   PetscMPIInt    size;
552   PetscErrorCode ierr;
553 
554   PetscFunctionBegin;
555   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
556   if (!((PetscObject)A)->type_name) {
557     ierr = MPI_Comm_size(((PetscObject)A)->comm, &size);CHKERRQ(ierr);
558     if (size == 1) {
559       ierr = MatSetType(A, MATSEQAIJ);CHKERRQ(ierr);
560     } else {
561       ierr = MatSetType(A, MATMPIAIJ);CHKERRQ(ierr);
562     }
563   }
564   ierr = MatSetUpPreallocation(A);CHKERRQ(ierr);
565   PetscFunctionReturn(0);
566 }
567 
568 #undef __FUNCT__
569 #define __FUNCT__ "MatView"
570 /*@C
571    MatView - Visualizes a matrix object.
572 
573    Collective on Mat
574 
575    Input Parameters:
576 +  mat - the matrix
577 -  viewer - visualization context
578 
579   Notes:
580   The available visualization contexts include
581 +    PETSC_VIEWER_STDOUT_SELF - standard output (default)
582 .    PETSC_VIEWER_STDOUT_WORLD - synchronized standard
583         output where only the first processor opens
584         the file.  All other processors send their
585         data to the first processor to print.
586 -     PETSC_VIEWER_DRAW_WORLD - graphical display of nonzero structure
587 
588    The user can open alternative visualization contexts with
589 +    PetscViewerASCIIOpen() - Outputs matrix to a specified file
590 .    PetscViewerBinaryOpen() - Outputs matrix in binary to a
591          specified file; corresponding input uses MatLoad()
592 .    PetscViewerDrawOpen() - Outputs nonzero matrix structure to
593          an X window display
594 -    PetscViewerSocketOpen() - Outputs matrix to Socket viewer.
595          Currently only the sequential dense and AIJ
596          matrix types support the Socket viewer.
597 
598    The user can call PetscViewerSetFormat() to specify the output
599    format of ASCII printed objects (when using PETSC_VIEWER_STDOUT_SELF,
600    PETSC_VIEWER_STDOUT_WORLD and PetscViewerASCIIOpen).  Available formats include
601 +    PETSC_VIEWER_DEFAULT - default, prints matrix contents
602 .    PETSC_VIEWER_ASCII_MATLAB - prints matrix contents in Matlab format
603 .    PETSC_VIEWER_ASCII_DENSE - prints entire matrix including zeros
604 .    PETSC_VIEWER_ASCII_COMMON - prints matrix contents, using a sparse
605          format common among all matrix types
606 .    PETSC_VIEWER_ASCII_IMPL - prints matrix contents, using an implementation-specific
607          format (which is in many cases the same as the default)
608 .    PETSC_VIEWER_ASCII_INFO - prints basic information about the matrix
609          size and structure (not the matrix entries)
610 .    PETSC_VIEWER_ASCII_INFO_DETAIL - prints more detailed information about
611          the matrix structure
612 
613    Options Database Keys:
614 +  -mat_view_info - Prints info on matrix at conclusion of MatEndAssembly()
615 .  -mat_view_info_detailed - Prints more detailed info
616 .  -mat_view - Prints matrix in ASCII format
617 .  -mat_view_matlab - Prints matrix in Matlab format
618 .  -mat_view_draw - PetscDraws nonzero structure of matrix, using MatView() and PetscDrawOpenX().
619 .  -display <name> - Sets display name (default is host)
620 .  -draw_pause <sec> - Sets number of seconds to pause after display
621 .  -mat_view_socket - Sends matrix to socket, can be accessed from Matlab (see users manual)
622 .  -viewer_socket_machine <machine>
623 .  -viewer_socket_port <port>
624 .  -mat_view_binary - save matrix to file in binary format
625 -  -viewer_binary_filename <name>
626    Level: beginner
627 
628    Notes: see the manual page for MatLoad() for the exact format of the binary file when the binary
629       viewer is used.
630 
631       See bin/matlab/PetscBinaryRead.m for a Matlab code that can read in the binary file when the binary
632       viewer is used.
633 
634    Concepts: matrices^viewing
635    Concepts: matrices^plotting
636    Concepts: matrices^printing
637 
638 .seealso: PetscViewerSetFormat(), PetscViewerASCIIOpen(), PetscViewerDrawOpen(),
639           PetscViewerSocketOpen(), PetscViewerBinaryOpen(), MatLoad()
640 @*/
641 PetscErrorCode PETSCMAT_DLLEXPORT MatView(Mat mat,PetscViewer viewer)
642 {
643   PetscErrorCode    ierr;
644   PetscInt          rows,cols;
645   PetscTruth        iascii;
646   const MatType     cstr;
647   PetscViewerFormat format;
648 
649   PetscFunctionBegin;
650   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
651   PetscValidType(mat,1);
652   if (!viewer) {
653     ierr = PetscViewerASCIIGetStdout(((PetscObject)mat)->comm,&viewer);CHKERRQ(ierr);
654   }
655   PetscValidHeaderSpecific(viewer,PETSC_VIEWER_COOKIE,2);
656   PetscCheckSameComm(mat,1,viewer,2);
657   if (!mat->assembled) SETERRQ(PETSC_ERR_ORDER,"Must call MatAssemblyBegin/End() before viewing matrix");
658   ierr = MatPreallocated(mat);CHKERRQ(ierr);
659 
660   ierr = PetscLogEventBegin(MAT_View,mat,viewer,0,0);CHKERRQ(ierr);
661   ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&iascii);CHKERRQ(ierr);
662   if (iascii) {
663     ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr);
664     if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
665       if (((PetscObject)mat)->prefix) {
666         ierr = PetscViewerASCIIPrintf(viewer,"Matrix Object:(%s)\n",((PetscObject)mat)->prefix);CHKERRQ(ierr);
667       } else {
668         ierr = PetscViewerASCIIPrintf(viewer,"Matrix Object:\n");CHKERRQ(ierr);
669       }
670       ierr = PetscViewerASCIIPushTab(viewer);CHKERRQ(ierr);
671       ierr = MatGetType(mat,&cstr);CHKERRQ(ierr);
672       ierr = MatGetSize(mat,&rows,&cols);CHKERRQ(ierr);
673       ierr = PetscViewerASCIIPrintf(viewer,"type=%s, rows=%D, cols=%D\n",cstr,rows,cols);CHKERRQ(ierr);
674       if (mat->factor) {
675         const MatSolverPackage solver;
676         ierr = MatFactorGetSolverPackage(mat,&solver);CHKERRQ(ierr);
677         ierr = PetscViewerASCIIPrintf(viewer,"package used to perform factorization: %s\n",solver);CHKERRQ(ierr);
678       }
679       if (mat->ops->getinfo) {
680         MatInfo info;
681         ierr = MatGetInfo(mat,MAT_GLOBAL_SUM,&info);CHKERRQ(ierr);
682         ierr = PetscViewerASCIIPrintf(viewer,"total: nonzeros=%D, allocated nonzeros=%D\n",(PetscInt)info.nz_used,(PetscInt)info.nz_allocated);CHKERRQ(ierr);
683       }
684     }
685   }
686   if (mat->ops->view) {
687     ierr = PetscViewerASCIIPushTab(viewer);CHKERRQ(ierr);
688     ierr = (*mat->ops->view)(mat,viewer);CHKERRQ(ierr);
689     ierr = PetscViewerASCIIPopTab(viewer);CHKERRQ(ierr);
690   } else if (!iascii) {
691     SETERRQ1(PETSC_ERR_SUP,"Viewer type %s not supported",((PetscObject)viewer)->type_name);
692   }
693   if (iascii) {
694     ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr);
695     if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
696       ierr = PetscViewerASCIIPopTab(viewer);CHKERRQ(ierr);
697     }
698   }
699   ierr = PetscLogEventEnd(MAT_View,mat,viewer,0,0);CHKERRQ(ierr);
700   PetscFunctionReturn(0);
701 }
702 
703 #undef __FUNCT__
704 #define __FUNCT__ "MatScaleSystem"
705 /*@
706    MatScaleSystem - Scale a vector solution and right hand side to
707    match the scaling of a scaled matrix.
708 
709    Collective on Mat
710 
711    Input Parameter:
712 +  mat - the matrix
713 .  b - right hand side vector (or PETSC_NULL)
714 -  x - solution vector (or PETSC_NULL)
715 
716 
717    Notes:
718    For AIJ, and BAIJ matrix formats, the matrices are not
719    internally scaled, so this does nothing. For MPIROWBS it
720    permutes and diagonally scales.
721 
722    The KSP methods automatically call this routine when required
723    (via PCPreSolve()) so it is rarely used directly.
724 
725    Level: Developer
726 
727    Concepts: matrices^scaling
728 
729 .seealso: MatUseScaledForm(), MatUnScaleSystem()
730 @*/
731 PetscErrorCode PETSCMAT_DLLEXPORT MatScaleSystem(Mat mat,Vec b,Vec x)
732 {
733   PetscErrorCode ierr;
734 
735   PetscFunctionBegin;
736   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
737   PetscValidType(mat,1);
738   ierr = MatPreallocated(mat);CHKERRQ(ierr);
739   if (x) {PetscValidHeaderSpecific(x,VEC_COOKIE,2);PetscCheckSameComm(mat,1,x,2);}
740   if (b) {PetscValidHeaderSpecific(b,VEC_COOKIE,3);PetscCheckSameComm(mat,1,b,3);}
741 
742   if (mat->ops->scalesystem) {
743     ierr = (*mat->ops->scalesystem)(mat,b,x);CHKERRQ(ierr);
744   }
745   PetscFunctionReturn(0);
746 }
747 
748 #undef __FUNCT__
749 #define __FUNCT__ "MatUnScaleSystem"
750 /*@
751    MatUnScaleSystem - Unscales a vector solution and right hand side to
752    match the original scaling of a scaled matrix.
753 
754    Collective on Mat
755 
756    Input Parameter:
757 +  mat - the matrix
758 .  b - right hand side vector (or PETSC_NULL)
759 -  x - solution vector (or PETSC_NULL)
760 
761 
762    Notes:
763    For AIJ and BAIJ matrix formats, the matrices are not
764    internally scaled, so this does nothing. For MPIROWBS it
765    permutes and diagonally scales.
766 
767    The KSP methods automatically call this routine when required
768    (via PCPreSolve()) so it is rarely used directly.
769 
770    Level: Developer
771 
772 .seealso: MatUseScaledForm(), MatScaleSystem()
773 @*/
774 PetscErrorCode PETSCMAT_DLLEXPORT MatUnScaleSystem(Mat mat,Vec b,Vec x)
775 {
776   PetscErrorCode ierr;
777 
778   PetscFunctionBegin;
779   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
780   PetscValidType(mat,1);
781   ierr = MatPreallocated(mat);CHKERRQ(ierr);
782   if (x) {PetscValidHeaderSpecific(x,VEC_COOKIE,2);PetscCheckSameComm(mat,1,x,2);}
783   if (b) {PetscValidHeaderSpecific(b,VEC_COOKIE,3);PetscCheckSameComm(mat,1,b,3);}
784   if (mat->ops->unscalesystem) {
785     ierr = (*mat->ops->unscalesystem)(mat,b,x);CHKERRQ(ierr);
786   }
787   PetscFunctionReturn(0);
788 }
789 
790 #undef __FUNCT__
791 #define __FUNCT__ "MatUseScaledForm"
792 /*@
793    MatUseScaledForm - For matrix storage formats that scale the
794    matrix (for example MPIRowBS matrices are diagonally scaled on
795    assembly) indicates matrix operations (MatMult() etc) are
796    applied using the scaled matrix.
797 
798    Collective on Mat
799 
800    Input Parameter:
801 +  mat - the matrix
802 -  scaled - PETSC_TRUE for applying the scaled, PETSC_FALSE for
803             applying the original matrix
804 
805    Notes:
806    For scaled matrix formats, applying the original, unscaled matrix
807    will be slightly more expensive
808 
809    Level: Developer
810 
811 .seealso: MatScaleSystem(), MatUnScaleSystem()
812 @*/
813 PetscErrorCode PETSCMAT_DLLEXPORT MatUseScaledForm(Mat mat,PetscTruth scaled)
814 {
815   PetscErrorCode ierr;
816 
817   PetscFunctionBegin;
818   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
819   PetscValidType(mat,1);
820   ierr = MatPreallocated(mat);CHKERRQ(ierr);
821   if (mat->ops->usescaledform) {
822     ierr = (*mat->ops->usescaledform)(mat,scaled);CHKERRQ(ierr);
823   }
824   PetscFunctionReturn(0);
825 }
826 
827 #undef __FUNCT__
828 #define __FUNCT__ "MatDestroy"
829 /*@
830    MatDestroy - Frees space taken by a matrix.
831 
832    Collective on Mat
833 
834    Input Parameter:
835 .  A - the matrix
836 
837    Level: beginner
838 
839 @*/
840 PetscErrorCode PETSCMAT_DLLEXPORT MatDestroy(Mat A)
841 {
842   PetscErrorCode ierr;
843   PetscFunctionBegin;
844   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
845   if (--((PetscObject)A)->refct > 0) PetscFunctionReturn(0);
846   ierr = MatPreallocated(A);CHKERRQ(ierr);
847   /* if memory was published with AMS then destroy it */
848   ierr = PetscObjectDepublish(A);CHKERRQ(ierr);
849   if (A->ops->destroy) {
850     ierr = (*A->ops->destroy)(A);CHKERRQ(ierr);
851   }
852   if (A->mapping) {
853     ierr = ISLocalToGlobalMappingDestroy(A->mapping);CHKERRQ(ierr);
854   }
855   if (A->bmapping) {
856     ierr = ISLocalToGlobalMappingDestroy(A->bmapping);CHKERRQ(ierr);
857   }
858 
859   if (A->spptr){ierr = PetscFree(A->spptr);CHKERRQ(ierr);}
860   ierr = PetscMapDestroy(A->rmap);CHKERRQ(ierr);
861   ierr = PetscMapDestroy(A->cmap);CHKERRQ(ierr);
862   ierr = PetscHeaderDestroy(A);CHKERRQ(ierr);
863   PetscFunctionReturn(0);
864 }
865 
866 #undef __FUNCT__
867 #define __FUNCT__ "MatValid"
868 /*@
869    MatValid - Checks whether a matrix object is valid.
870 
871    Collective on Mat
872 
873    Input Parameter:
874 .  m - the matrix to check
875 
876    Output Parameter:
877    flg - flag indicating matrix status, either
878    PETSC_TRUE if matrix is valid, or PETSC_FALSE otherwise.
879 
880    Level: developer
881 
882    Concepts: matrices^validity
883 @*/
884 PetscErrorCode PETSCMAT_DLLEXPORT MatValid(Mat m,PetscTruth *flg)
885 {
886   PetscFunctionBegin;
887   PetscValidIntPointer(flg,1);
888   if (!m)                                          *flg = PETSC_FALSE;
889   else if (((PetscObject)m)->cookie != MAT_COOKIE) *flg = PETSC_FALSE;
890   else                                             *flg = PETSC_TRUE;
891   PetscFunctionReturn(0);
892 }
893 
894 #undef __FUNCT__
895 #define __FUNCT__ "MatSetValues"
896 /*@
897    MatSetValues - Inserts or adds a block of values into a matrix.
898    These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
899    MUST be called after all calls to MatSetValues() have been completed.
900 
901    Not Collective
902 
903    Input Parameters:
904 +  mat - the matrix
905 .  v - a logically two-dimensional array of values
906 .  m, idxm - the number of rows and their global indices
907 .  n, idxn - the number of columns and their global indices
908 -  addv - either ADD_VALUES or INSERT_VALUES, where
909    ADD_VALUES adds values to any existing entries, and
910    INSERT_VALUES replaces existing entries with new values
911 
912    Notes:
913    By default the values, v, are row-oriented and unsorted.
914    See MatSetOption() for other options.
915 
916    Calls to MatSetValues() with the INSERT_VALUES and ADD_VALUES
917    options cannot be mixed without intervening calls to the assembly
918    routines.
919 
920    MatSetValues() uses 0-based row and column numbers in Fortran
921    as well as in C.
922 
923    Negative indices may be passed in idxm and idxn, these rows and columns are
924    simply ignored. This allows easily inserting element stiffness matrices
925    with homogeneous Dirchlet boundary conditions that you don't want represented
926    in the matrix.
927 
928    Efficiency Alert:
929    The routine MatSetValuesBlocked() may offer much better efficiency
930    for users of block sparse formats (MATSEQBAIJ and MATMPIBAIJ).
931 
932    Level: beginner
933 
934    Concepts: matrices^putting entries in
935 
936 .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
937           InsertMode, INSERT_VALUES, ADD_VALUES
938 @*/
939 PetscErrorCode PETSCMAT_DLLEXPORT MatSetValues(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],const PetscScalar v[],InsertMode addv)
940 {
941   PetscErrorCode ierr;
942 
943   PetscFunctionBegin;
944   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
945   PetscValidType(mat,1);
946   if (!m || !n) PetscFunctionReturn(0); /* no values to insert */
947   PetscValidIntPointer(idxm,3);
948   PetscValidIntPointer(idxn,5);
949   ierr = MatPreallocated(mat);CHKERRQ(ierr);
950   if (mat->insertmode == NOT_SET_VALUES) {
951     mat->insertmode = addv;
952   }
953 #if defined(PETSC_USE_DEBUG)
954   else if (mat->insertmode != addv) {
955     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
956   }
957   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
958 #endif
959 
960   if (mat->assembled) {
961     mat->was_assembled = PETSC_TRUE;
962     mat->assembled     = PETSC_FALSE;
963   }
964   ierr = PetscLogEventBegin(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
965   if (!mat->ops->setvalues) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
966   ierr = (*mat->ops->setvalues)(mat,m,idxm,n,idxn,v,addv);CHKERRQ(ierr);
967   ierr = PetscLogEventEnd(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
968   PetscFunctionReturn(0);
969 }
970 
971 
972 #undef __FUNCT__
973 #define __FUNCT__ "MatSetValuesRowLocal"
974 /*@
975    MatSetValuesRowLocal - Inserts a row (block row for BAIJ matrices) of nonzero
976         values into a matrix
977 
978    Not Collective
979 
980    Input Parameters:
981 +  mat - the matrix
982 .  row - the (block) row to set
983 -  v - a logically two-dimensional array of values
984 
985    Notes:
986    By the values, v, are column-oriented (for the block version) and sorted
987 
988    All the nonzeros in the row must be provided
989 
990    The matrix must have previously had its column indices set
991 
992    The row must belong to this process
993 
994    Level: intermediate
995 
996    Concepts: matrices^putting entries in
997 
998 .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
999           InsertMode, INSERT_VALUES, ADD_VALUES, MatSetValues(), MatSetValuesRow(), MatSetLocalToGlobalMapping()
1000 @*/
1001 PetscErrorCode PETSCMAT_DLLEXPORT MatSetValuesRowLocal(Mat mat,PetscInt row,const PetscScalar v[])
1002 {
1003   PetscErrorCode ierr;
1004 
1005   PetscFunctionBegin;
1006   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1007   PetscValidType(mat,1);
1008   PetscValidScalarPointer(v,2);
1009   ierr = MatSetValuesRow(mat, mat->mapping->indices[row],v);CHKERRQ(ierr);
1010   PetscFunctionReturn(0);
1011 }
1012 
1013 #undef __FUNCT__
1014 #define __FUNCT__ "MatSetValuesRow"
1015 /*@
1016    MatSetValuesRow - Inserts a row (block row for BAIJ matrices) of nonzero
1017         values into a matrix
1018 
1019    Not Collective
1020 
1021    Input Parameters:
1022 +  mat - the matrix
1023 .  row - the (block) row to set
1024 -  v - a logically two-dimensional array of values
1025 
1026    Notes:
1027    By the values, v, are column-oriented (for the block version) and sorted
1028 
1029    All the nonzeros in the row must be provided
1030 
1031    The matrix must have previously had its column indices set
1032 
1033    The row must belong to this process
1034 
1035    Level: intermediate
1036 
1037    Concepts: matrices^putting entries in
1038 
1039 .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
1040           InsertMode, INSERT_VALUES, ADD_VALUES, MatSetValues()
1041 @*/
1042 PetscErrorCode PETSCMAT_DLLEXPORT MatSetValuesRow(Mat mat,PetscInt row,const PetscScalar v[])
1043 {
1044   PetscErrorCode ierr;
1045 
1046   PetscFunctionBegin;
1047   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1048   PetscValidType(mat,1);
1049   PetscValidScalarPointer(v,2);
1050 #if defined(PETSC_USE_DEBUG)
1051   if (mat->insertmode == ADD_VALUES) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add and insert values");
1052   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1053 #endif
1054   mat->insertmode = INSERT_VALUES;
1055 
1056   if (mat->assembled) {
1057     mat->was_assembled = PETSC_TRUE;
1058     mat->assembled     = PETSC_FALSE;
1059   }
1060   ierr = PetscLogEventBegin(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
1061   if (!mat->ops->setvaluesrow) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1062   ierr = (*mat->ops->setvaluesrow)(mat,row,v);CHKERRQ(ierr);
1063   ierr = PetscLogEventEnd(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
1064   PetscFunctionReturn(0);
1065 }
1066 
1067 #undef __FUNCT__
1068 #define __FUNCT__ "MatSetValuesStencil"
1069 /*@
1070    MatSetValuesStencil - Inserts or adds a block of values into a matrix.
1071      Using structured grid indexing
1072 
1073    Not Collective
1074 
1075    Input Parameters:
1076 +  mat - the matrix
1077 .  v - a logically two-dimensional array of values
1078 .  m - number of rows being entered
1079 .  idxm - grid coordinates (and component number when dof > 1) for matrix rows being entered
1080 .  n - number of columns being entered
1081 .  idxn - grid coordinates (and component number when dof > 1) for matrix columns being entered
1082 -  addv - either ADD_VALUES or INSERT_VALUES, where
1083    ADD_VALUES adds values to any existing entries, and
1084    INSERT_VALUES replaces existing entries with new values
1085 
1086    Notes:
1087    By default the values, v, are row-oriented.  See MatSetOption() for other options.
1088 
1089    Calls to MatSetValuesStencil() with the INSERT_VALUES and ADD_VALUES
1090    options cannot be mixed without intervening calls to the assembly
1091    routines.
1092 
1093    The grid coordinates are across the entire grid, not just the local portion
1094 
1095    MatSetValuesStencil() uses 0-based row and column numbers in Fortran
1096    as well as in C.
1097 
1098    For setting/accessing vector values via array coordinates you can use the DAVecGetArray() routine
1099 
1100    In order to use this routine you must either obtain the matrix with DAGetMatrix()
1101    or call MatSetLocalToGlobalMapping() and MatSetStencil() first.
1102 
1103    The columns and rows in the stencil passed in MUST be contained within the
1104    ghost region of the given process as set with DACreateXXX() or MatSetStencil(). For example,
1105    if you create a DA with an overlap of one grid level and on a particular process its first
1106    local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1107    first i index you can use in your column and row indices in MatSetStencil() is 5.
1108 
1109    In Fortran idxm and idxn should be declared as
1110 $     MatStencil idxm(4,m),idxn(4,n)
1111    and the values inserted using
1112 $    idxm(MatStencil_i,1) = i
1113 $    idxm(MatStencil_j,1) = j
1114 $    idxm(MatStencil_k,1) = k
1115 $    idxm(MatStencil_c,1) = c
1116    etc
1117 
1118    For periodic boundary conditions use negative indices for values to the left (below 0; that are to be
1119    obtained by wrapping values from right edge). For values to the right of the last entry using that index plus one
1120    etc to obtain values that obtained by wrapping the values from the left edge. This does not work for the DA_NONPERIODIC
1121    wrap.
1122 
1123    For indices that don't mean anything for your case (like the k index when working in 2d) or the c index when you have
1124    a single value per point) you can skip filling those indices.
1125 
1126    Inspired by the structured grid interface to the HYPRE package
1127    (http://www.llnl.gov/CASC/hypre)
1128 
1129    Efficiency Alert:
1130    The routine MatSetValuesBlockedStencil() may offer much better efficiency
1131    for users of block sparse formats (MATSEQBAIJ and MATMPIBAIJ).
1132 
1133    Level: beginner
1134 
1135    Concepts: matrices^putting entries in
1136 
1137 .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
1138           MatSetValues(), MatSetValuesBlockedStencil(), MatSetStencil(), DAGetMatrix(), DAVecGetArray(), MatStencil
1139 @*/
1140 PetscErrorCode PETSCMAT_DLLEXPORT MatSetValuesStencil(Mat mat,PetscInt m,const MatStencil idxm[],PetscInt n,const MatStencil idxn[],const PetscScalar v[],InsertMode addv)
1141 {
1142   PetscErrorCode ierr;
1143   PetscInt       j,i,jdxm[128],jdxn[256],dim = mat->stencil.dim,*dims = mat->stencil.dims+1,tmp;
1144   PetscInt       *starts = mat->stencil.starts,*dxm = (PetscInt*)idxm,*dxn = (PetscInt*)idxn,sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1145 
1146   PetscFunctionBegin;
1147   if (!m || !n) PetscFunctionReturn(0); /* no values to insert */
1148   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1149   PetscValidType(mat,1);
1150   PetscValidIntPointer(idxm,3);
1151   PetscValidIntPointer(idxn,5);
1152   PetscValidScalarPointer(v,6);
1153 
1154   if (m > 128) SETERRQ1(PETSC_ERR_SUP,"Can only set 128 rows at a time; trying to set %D",m);
1155   if (n > 256) SETERRQ1(PETSC_ERR_SUP,"Can only set 256 columns at a time; trying to set %D",n);
1156 
1157   for (i=0; i<m; i++) {
1158     for (j=0; j<3-sdim; j++) dxm++;
1159     tmp = *dxm++ - starts[0];
1160     for (j=0; j<dim-1; j++) {
1161       if ((*dxm++ - starts[j+1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
1162       else                                       tmp = tmp*dims[j] + *(dxm-1) - starts[j+1];
1163     }
1164     if (mat->stencil.noc) dxm++;
1165     jdxm[i] = tmp;
1166   }
1167   for (i=0; i<n; i++) {
1168     for (j=0; j<3-sdim; j++) dxn++;
1169     tmp = *dxn++ - starts[0];
1170     for (j=0; j<dim-1; j++) {
1171       if ((*dxn++ - starts[j+1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
1172       else                                       tmp = tmp*dims[j] + *(dxn-1) - starts[j+1];
1173     }
1174     if (mat->stencil.noc) dxn++;
1175     jdxn[i] = tmp;
1176   }
1177   ierr = MatSetValuesLocal(mat,m,jdxm,n,jdxn,v,addv);CHKERRQ(ierr);
1178   PetscFunctionReturn(0);
1179 }
1180 
1181 #undef __FUNCT__
1182 #define __FUNCT__ "MatSetValuesBlockedStencil"
1183 /*@C
1184    MatSetValuesBlockedStencil - Inserts or adds a block of values into a matrix.
1185      Using structured grid indexing
1186 
1187    Not Collective
1188 
1189    Input Parameters:
1190 +  mat - the matrix
1191 .  v - a logically two-dimensional array of values
1192 .  m - number of rows being entered
1193 .  idxm - grid coordinates for matrix rows being entered
1194 .  n - number of columns being entered
1195 .  idxn - grid coordinates for matrix columns being entered
1196 -  addv - either ADD_VALUES or INSERT_VALUES, where
1197    ADD_VALUES adds values to any existing entries, and
1198    INSERT_VALUES replaces existing entries with new values
1199 
1200    Notes:
1201    By default the values, v, are row-oriented and unsorted.
1202    See MatSetOption() for other options.
1203 
1204    Calls to MatSetValuesBlockedStencil() with the INSERT_VALUES and ADD_VALUES
1205    options cannot be mixed without intervening calls to the assembly
1206    routines.
1207 
1208    The grid coordinates are across the entire grid, not just the local portion
1209 
1210    MatSetValuesBlockedStencil() uses 0-based row and column numbers in Fortran
1211    as well as in C.
1212 
1213    For setting/accessing vector values via array coordinates you can use the DAVecGetArray() routine
1214 
1215    In order to use this routine you must either obtain the matrix with DAGetMatrix()
1216    or call MatSetLocalToGlobalMapping() and MatSetStencil() first.
1217 
1218    The columns and rows in the stencil passed in MUST be contained within the
1219    ghost region of the given process as set with DACreateXXX() or MatSetStencil(). For example,
1220    if you create a DA with an overlap of one grid level and on a particular process its first
1221    local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the
1222    first i index you can use in your column and row indices in MatSetStencil() is 5.
1223 
1224    In Fortran idxm and idxn should be declared as
1225 $     MatStencil idxm(4,m),idxn(4,n)
1226    and the values inserted using
1227 $    idxm(MatStencil_i,1) = i
1228 $    idxm(MatStencil_j,1) = j
1229 $    idxm(MatStencil_k,1) = k
1230    etc
1231 
1232    Negative indices may be passed in idxm and idxn, these rows and columns are
1233    simply ignored. This allows easily inserting element stiffness matrices
1234    with homogeneous Dirchlet boundary conditions that you don't want represented
1235    in the matrix.
1236 
1237    Inspired by the structured grid interface to the HYPRE package
1238    (http://www.llnl.gov/CASC/hypre)
1239 
1240    Level: beginner
1241 
1242    Concepts: matrices^putting entries in
1243 
1244 .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
1245           MatSetValues(), MatSetValuesStencil(), MatSetStencil(), DAGetMatrix(), DAVecGetArray(), MatStencil
1246 @*/
1247 PetscErrorCode PETSCMAT_DLLEXPORT MatSetValuesBlockedStencil(Mat mat,PetscInt m,const MatStencil idxm[],PetscInt n,const MatStencil idxn[],const PetscScalar v[],InsertMode addv)
1248 {
1249   PetscErrorCode ierr;
1250   PetscInt       j,i,jdxm[128],jdxn[256],dim = mat->stencil.dim,*dims = mat->stencil.dims+1,tmp;
1251   PetscInt       *starts = mat->stencil.starts,*dxm = (PetscInt*)idxm,*dxn = (PetscInt*)idxn,sdim = dim - (1 - (PetscInt)mat->stencil.noc);
1252 
1253   PetscFunctionBegin;
1254   if (!m || !n) PetscFunctionReturn(0); /* no values to insert */
1255   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1256   PetscValidType(mat,1);
1257   PetscValidIntPointer(idxm,3);
1258   PetscValidIntPointer(idxn,5);
1259   PetscValidScalarPointer(v,6);
1260 
1261   if (m > 128) SETERRQ1(PETSC_ERR_SUP,"Can only set 128 rows at a time; trying to set %D",m);
1262   if (n > 128) SETERRQ1(PETSC_ERR_SUP,"Can only set 256 columns at a time; trying to set %D",n);
1263 
1264   for (i=0; i<m; i++) {
1265     for (j=0; j<3-sdim; j++) dxm++;
1266     tmp = *dxm++ - starts[0];
1267     for (j=0; j<sdim-1; j++) {
1268       if ((*dxm++ - starts[j+1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
1269       else                                      tmp = tmp*dims[j] + *(dxm-1) - starts[j+1];
1270     }
1271     dxm++;
1272     jdxm[i] = tmp;
1273   }
1274   for (i=0; i<n; i++) {
1275     for (j=0; j<3-sdim; j++) dxn++;
1276     tmp = *dxn++ - starts[0];
1277     for (j=0; j<sdim-1; j++) {
1278       if ((*dxn++ - starts[j+1]) < 0 || tmp < 0) tmp = PETSC_MIN_INT;
1279       else                                       tmp = tmp*dims[j] + *(dxn-1) - starts[j+1];
1280     }
1281     dxn++;
1282     jdxn[i] = tmp;
1283   }
1284   ierr = MatSetValuesBlockedLocal(mat,m,jdxm,n,jdxn,v,addv);CHKERRQ(ierr);
1285   PetscFunctionReturn(0);
1286 }
1287 
1288 #undef __FUNCT__
1289 #define __FUNCT__ "MatSetStencil"
1290 /*@
1291    MatSetStencil - Sets the grid information for setting values into a matrix via
1292         MatSetValuesStencil()
1293 
1294    Not Collective
1295 
1296    Input Parameters:
1297 +  mat - the matrix
1298 .  dim - dimension of the grid 1, 2, or 3
1299 .  dims - number of grid points in x, y, and z direction, including ghost points on your processor
1300 .  starts - starting point of ghost nodes on your processor in x, y, and z direction
1301 -  dof - number of degrees of freedom per node
1302 
1303 
1304    Inspired by the structured grid interface to the HYPRE package
1305    (www.llnl.gov/CASC/hyper)
1306 
1307    For matrices generated with DAGetMatrix() this routine is automatically called and so not needed by the
1308    user.
1309 
1310    Level: beginner
1311 
1312    Concepts: matrices^putting entries in
1313 
1314 .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal()
1315           MatSetValues(), MatSetValuesBlockedStencil(), MatSetValuesStencil()
1316 @*/
1317 PetscErrorCode PETSCMAT_DLLEXPORT MatSetStencil(Mat mat,PetscInt dim,const PetscInt dims[],const PetscInt starts[],PetscInt dof)
1318 {
1319   PetscInt i;
1320 
1321   PetscFunctionBegin;
1322   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1323   PetscValidIntPointer(dims,3);
1324   PetscValidIntPointer(starts,4);
1325 
1326   mat->stencil.dim = dim + (dof > 1);
1327   for (i=0; i<dim; i++) {
1328     mat->stencil.dims[i]   = dims[dim-i-1];      /* copy the values in backwards */
1329     mat->stencil.starts[i] = starts[dim-i-1];
1330   }
1331   mat->stencil.dims[dim]   = dof;
1332   mat->stencil.starts[dim] = 0;
1333   mat->stencil.noc         = (PetscTruth)(dof == 1);
1334   PetscFunctionReturn(0);
1335 }
1336 
1337 #undef __FUNCT__
1338 #define __FUNCT__ "MatSetValuesBlocked"
1339 /*@
1340    MatSetValuesBlocked - Inserts or adds a block of values into a matrix.
1341 
1342    Not Collective
1343 
1344    Input Parameters:
1345 +  mat - the matrix
1346 .  v - a logically two-dimensional array of values
1347 .  m, idxm - the number of block rows and their global block indices
1348 .  n, idxn - the number of block columns and their global block indices
1349 -  addv - either ADD_VALUES or INSERT_VALUES, where
1350    ADD_VALUES adds values to any existing entries, and
1351    INSERT_VALUES replaces existing entries with new values
1352 
1353    Notes:
1354    The m and n count the NUMBER of blocks in the row direction and column direction,
1355    NOT the total number of rows/columns; for example, if the block size is 2 and
1356    you are passing in values for rows 2,3,4,5  then m would be 2 (not 4).
1357    The values in idxm would be 1 2; that is the first index for each block divided by
1358    the block size.
1359 
1360    By default the values, v, are row-oriented. So the layout of
1361    v is the same as for MatSetValues(). See MatSetOption() for other options.
1362 
1363    Calls to MatSetValuesBlocked() with the INSERT_VALUES and ADD_VALUES
1364    options cannot be mixed without intervening calls to the assembly
1365    routines.
1366 
1367    MatSetValuesBlocked() uses 0-based row and column numbers in Fortran
1368    as well as in C.
1369 
1370    Negative indices may be passed in idxm and idxn, these rows and columns are
1371    simply ignored. This allows easily inserting element stiffness matrices
1372    with homogeneous Dirchlet boundary conditions that you don't want represented
1373    in the matrix.
1374 
1375    Each time an entry is set within a sparse matrix via MatSetValues(),
1376    internal searching must be done to determine where to place the the
1377    data in the matrix storage space.  By instead inserting blocks of
1378    entries via MatSetValuesBlocked(), the overhead of matrix assembly is
1379    reduced.
1380 
1381    Example:
1382 $   Suppose m=n=2 and block size(bs) = 2 The array is
1383 $
1384 $   1  2  | 3  4
1385 $   5  6  | 7  8
1386 $   - - - | - - -
1387 $   9  10 | 11 12
1388 $   13 14 | 15 16
1389 $
1390 $   v[] should be passed in like
1391 $   v[] = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]
1392 $
1393 $  If you are not using row oriented storage of v (that is you called MatSetOption(mat,MAT_ROW_ORIENTED,PETSC_FALSE)) then
1394 $   v[] = [1,5,9,13,2,6,10,14,3,7,11,15,4,8,12,16]
1395 
1396    Level: intermediate
1397 
1398    Concepts: matrices^putting entries in blocked
1399 
1400 .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesBlockedLocal()
1401 @*/
1402 PetscErrorCode PETSCMAT_DLLEXPORT MatSetValuesBlocked(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],const PetscScalar v[],InsertMode addv)
1403 {
1404   PetscErrorCode ierr;
1405 
1406   PetscFunctionBegin;
1407   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1408   PetscValidType(mat,1);
1409   if (!m || !n) PetscFunctionReturn(0); /* no values to insert */
1410   PetscValidIntPointer(idxm,3);
1411   PetscValidIntPointer(idxn,5);
1412   PetscValidScalarPointer(v,6);
1413   ierr = MatPreallocated(mat);CHKERRQ(ierr);
1414   if (mat->insertmode == NOT_SET_VALUES) {
1415     mat->insertmode = addv;
1416   }
1417 #if defined(PETSC_USE_DEBUG)
1418   else if (mat->insertmode != addv) {
1419     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
1420   }
1421   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1422 #endif
1423 
1424   if (mat->assembled) {
1425     mat->was_assembled = PETSC_TRUE;
1426     mat->assembled     = PETSC_FALSE;
1427   }
1428   ierr = PetscLogEventBegin(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
1429   if (mat->ops->setvaluesblocked) {
1430     ierr = (*mat->ops->setvaluesblocked)(mat,m,idxm,n,idxn,v,addv);CHKERRQ(ierr);
1431   } else {
1432     PetscInt buf[4096],*ibufm=0,*ibufn=0;
1433     PetscInt i,j,*iidxm,*iidxn,bs=mat->rmap->bs;
1434     if ((m+n)*bs <= 4096) {
1435       iidxm = buf; iidxn = buf + m*bs;
1436     } else {
1437       ierr = PetscMalloc2(m*bs,PetscInt,&ibufm,n*bs,PetscInt,&ibufn);CHKERRQ(ierr);
1438       iidxm = ibufm; iidxn = ibufn;
1439     }
1440     for (i=0; i<m; i++) {
1441       for (j=0; j<bs; j++) {
1442 	iidxm[i*bs+j] = bs*idxm[i] + j;
1443       }
1444     }
1445     for (i=0; i<n; i++) {
1446       for (j=0; j<bs; j++) {
1447 	iidxn[i*bs+j] = bs*idxn[i] + j;
1448       }
1449     }
1450     ierr = MatSetValues(mat,bs*m,iidxm,bs*n,iidxn,v,addv);CHKERRQ(ierr);
1451     ierr = PetscFree2(ibufm,ibufn);CHKERRQ(ierr);
1452   }
1453   ierr = PetscLogEventEnd(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
1454   PetscFunctionReturn(0);
1455 }
1456 
1457 #undef __FUNCT__
1458 #define __FUNCT__ "MatGetValues"
1459 /*@
1460    MatGetValues - Gets a block of values from a matrix.
1461 
1462    Not Collective; currently only returns a local block
1463 
1464    Input Parameters:
1465 +  mat - the matrix
1466 .  v - a logically two-dimensional array for storing the values
1467 .  m, idxm - the number of rows and their global indices
1468 -  n, idxn - the number of columns and their global indices
1469 
1470    Notes:
1471    The user must allocate space (m*n PetscScalars) for the values, v.
1472    The values, v, are then returned in a row-oriented format,
1473    analogous to that used by default in MatSetValues().
1474 
1475    MatGetValues() uses 0-based row and column numbers in
1476    Fortran as well as in C.
1477 
1478    MatGetValues() requires that the matrix has been assembled
1479    with MatAssemblyBegin()/MatAssemblyEnd().  Thus, calls to
1480    MatSetValues() and MatGetValues() CANNOT be made in succession
1481    without intermediate matrix assembly.
1482 
1483    Negative row or column indices will be ignored and those locations in v[] will be
1484    left unchanged.
1485 
1486    Level: advanced
1487 
1488    Concepts: matrices^accessing values
1489 
1490 .seealso: MatGetRow(), MatGetSubMatrices(), MatSetValues()
1491 @*/
1492 PetscErrorCode PETSCMAT_DLLEXPORT MatGetValues(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],PetscScalar v[])
1493 {
1494   PetscErrorCode ierr;
1495 
1496   PetscFunctionBegin;
1497   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1498   PetscValidType(mat,1);
1499   PetscValidIntPointer(idxm,3);
1500   PetscValidIntPointer(idxn,5);
1501   PetscValidScalarPointer(v,6);
1502   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1503   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1504   if (!mat->ops->getvalues) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1505   ierr = MatPreallocated(mat);CHKERRQ(ierr);
1506 
1507   ierr = PetscLogEventBegin(MAT_GetValues,mat,0,0,0);CHKERRQ(ierr);
1508   ierr = (*mat->ops->getvalues)(mat,m,idxm,n,idxn,v);CHKERRQ(ierr);
1509   ierr = PetscLogEventEnd(MAT_GetValues,mat,0,0,0);CHKERRQ(ierr);
1510   PetscFunctionReturn(0);
1511 }
1512 
1513 #undef __FUNCT__
1514 #define __FUNCT__ "MatSetLocalToGlobalMapping"
1515 /*@
1516    MatSetLocalToGlobalMapping - Sets a local-to-global numbering for use by
1517    the routine MatSetValuesLocal() to allow users to insert matrix entries
1518    using a local (per-processor) numbering.
1519 
1520    Not Collective
1521 
1522    Input Parameters:
1523 +  x - the matrix
1524 -  mapping - mapping created with ISLocalToGlobalMappingCreate()
1525              or ISLocalToGlobalMappingCreateIS()
1526 
1527    Level: intermediate
1528 
1529    Concepts: matrices^local to global mapping
1530    Concepts: local to global mapping^for matrices
1531 
1532 .seealso:  MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesLocal()
1533 @*/
1534 PetscErrorCode PETSCMAT_DLLEXPORT MatSetLocalToGlobalMapping(Mat x,ISLocalToGlobalMapping mapping)
1535 {
1536   PetscErrorCode ierr;
1537   PetscFunctionBegin;
1538   PetscValidHeaderSpecific(x,MAT_COOKIE,1);
1539   PetscValidType(x,1);
1540   PetscValidHeaderSpecific(mapping,IS_LTOGM_COOKIE,2);
1541   if (x->mapping) {
1542     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Mapping already set for matrix");
1543   }
1544   ierr = MatPreallocated(x);CHKERRQ(ierr);
1545 
1546   if (x->ops->setlocaltoglobalmapping) {
1547     ierr = (*x->ops->setlocaltoglobalmapping)(x,mapping);CHKERRQ(ierr);
1548   } else {
1549     ierr = PetscObjectReference((PetscObject)mapping);CHKERRQ(ierr);
1550     if (x->mapping) { ierr = ISLocalToGlobalMappingDestroy(x->mapping);CHKERRQ(ierr); }
1551     x->mapping = mapping;
1552   }
1553   PetscFunctionReturn(0);
1554 }
1555 
1556 #undef __FUNCT__
1557 #define __FUNCT__ "MatSetLocalToGlobalMappingBlock"
1558 /*@
1559    MatSetLocalToGlobalMappingBlock - Sets a local-to-global numbering for use
1560    by the routine MatSetValuesBlockedLocal() to allow users to insert matrix
1561    entries using a local (per-processor) numbering.
1562 
1563    Not Collective
1564 
1565    Input Parameters:
1566 +  x - the matrix
1567 -  mapping - mapping created with ISLocalToGlobalMappingCreate() or
1568              ISLocalToGlobalMappingCreateIS()
1569 
1570    Level: intermediate
1571 
1572    Concepts: matrices^local to global mapping blocked
1573    Concepts: local to global mapping^for matrices, blocked
1574 
1575 .seealso:  MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetValuesBlockedLocal(),
1576            MatSetValuesBlocked(), MatSetValuesLocal()
1577 @*/
1578 PetscErrorCode PETSCMAT_DLLEXPORT MatSetLocalToGlobalMappingBlock(Mat x,ISLocalToGlobalMapping mapping)
1579 {
1580   PetscErrorCode ierr;
1581   PetscFunctionBegin;
1582   PetscValidHeaderSpecific(x,MAT_COOKIE,1);
1583   PetscValidType(x,1);
1584   PetscValidHeaderSpecific(mapping,IS_LTOGM_COOKIE,2);
1585   if (x->bmapping) {
1586     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Mapping already set for matrix");
1587   }
1588   ierr = PetscObjectReference((PetscObject)mapping);CHKERRQ(ierr);
1589   if (x->bmapping) { ierr = ISLocalToGlobalMappingDestroy(x->bmapping);CHKERRQ(ierr); }
1590   x->bmapping = mapping;
1591   PetscFunctionReturn(0);
1592 }
1593 
1594 #undef __FUNCT__
1595 #define __FUNCT__ "MatSetValuesLocal"
1596 /*@
1597    MatSetValuesLocal - Inserts or adds values into certain locations of a matrix,
1598    using a local ordering of the nodes.
1599 
1600    Not Collective
1601 
1602    Input Parameters:
1603 +  x - the matrix
1604 .  nrow, irow - number of rows and their local indices
1605 .  ncol, icol - number of columns and their local indices
1606 .  y -  a logically two-dimensional array of values
1607 -  addv - either INSERT_VALUES or ADD_VALUES, where
1608    ADD_VALUES adds values to any existing entries, and
1609    INSERT_VALUES replaces existing entries with new values
1610 
1611    Notes:
1612    Before calling MatSetValuesLocal(), the user must first set the
1613    local-to-global mapping by calling MatSetLocalToGlobalMapping().
1614 
1615    Calls to MatSetValuesLocal() with the INSERT_VALUES and ADD_VALUES
1616    options cannot be mixed without intervening calls to the assembly
1617    routines.
1618 
1619    These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
1620    MUST be called after all calls to MatSetValuesLocal() have been completed.
1621 
1622    Level: intermediate
1623 
1624    Concepts: matrices^putting entries in with local numbering
1625 
1626 .seealso:  MatAssemblyBegin(), MatAssemblyEnd(), MatSetValues(), MatSetLocalToGlobalMapping(),
1627            MatSetValueLocal()
1628 @*/
1629 PetscErrorCode PETSCMAT_DLLEXPORT MatSetValuesLocal(Mat mat,PetscInt nrow,const PetscInt irow[],PetscInt ncol,const PetscInt icol[],const PetscScalar y[],InsertMode addv)
1630 {
1631   PetscErrorCode ierr;
1632   PetscInt       irowm[2048],icolm[2048];
1633 
1634   PetscFunctionBegin;
1635   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1636   PetscValidType(mat,1);
1637   if (!nrow || !ncol) PetscFunctionReturn(0); /* no values to insert */
1638   PetscValidIntPointer(irow,3);
1639   PetscValidIntPointer(icol,5);
1640   PetscValidScalarPointer(y,6);
1641   ierr = MatPreallocated(mat);CHKERRQ(ierr);
1642   if (mat->insertmode == NOT_SET_VALUES) {
1643     mat->insertmode = addv;
1644   }
1645 #if defined(PETSC_USE_DEBUG)
1646   else if (mat->insertmode != addv) {
1647     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
1648   }
1649   if (!mat->ops->setvalueslocal && (nrow > 2048 || ncol > 2048)) {
1650     SETERRQ2(PETSC_ERR_SUP,"Number column/row indices must be <= 2048: are %D %D",nrow,ncol);
1651   }
1652   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1653 #endif
1654 
1655   if (mat->assembled) {
1656     mat->was_assembled = PETSC_TRUE;
1657     mat->assembled     = PETSC_FALSE;
1658   }
1659   ierr = PetscLogEventBegin(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
1660   if (!mat->ops->setvalueslocal) {
1661     ierr = ISLocalToGlobalMappingApply(mat->mapping,nrow,irow,irowm);CHKERRQ(ierr);
1662     ierr = ISLocalToGlobalMappingApply(mat->mapping,ncol,icol,icolm);CHKERRQ(ierr);
1663     ierr = (*mat->ops->setvalues)(mat,nrow,irowm,ncol,icolm,y,addv);CHKERRQ(ierr);
1664   } else {
1665     ierr = (*mat->ops->setvalueslocal)(mat,nrow,irow,ncol,icol,y,addv);CHKERRQ(ierr);
1666   }
1667   mat->same_nonzero = PETSC_FALSE;
1668   ierr = PetscLogEventEnd(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
1669   PetscFunctionReturn(0);
1670 }
1671 
1672 #undef __FUNCT__
1673 #define __FUNCT__ "MatSetValuesBlockedLocal"
1674 /*@
1675    MatSetValuesBlockedLocal - Inserts or adds values into certain locations of a matrix,
1676    using a local ordering of the nodes a block at a time.
1677 
1678    Not Collective
1679 
1680    Input Parameters:
1681 +  x - the matrix
1682 .  nrow, irow - number of rows and their local indices
1683 .  ncol, icol - number of columns and their local indices
1684 .  y -  a logically two-dimensional array of values
1685 -  addv - either INSERT_VALUES or ADD_VALUES, where
1686    ADD_VALUES adds values to any existing entries, and
1687    INSERT_VALUES replaces existing entries with new values
1688 
1689    Notes:
1690    Before calling MatSetValuesBlockedLocal(), the user must first set the
1691    local-to-global mapping by calling MatSetLocalToGlobalMappingBlock(),
1692    where the mapping MUST be set for matrix blocks, not for matrix elements.
1693 
1694    Calls to MatSetValuesBlockedLocal() with the INSERT_VALUES and ADD_VALUES
1695    options cannot be mixed without intervening calls to the assembly
1696    routines.
1697 
1698    These values may be cached, so MatAssemblyBegin() and MatAssemblyEnd()
1699    MUST be called after all calls to MatSetValuesBlockedLocal() have been completed.
1700 
1701    Level: intermediate
1702 
1703    Concepts: matrices^putting blocked values in with local numbering
1704 
1705 .seealso:  MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesLocal(), MatSetLocalToGlobalMappingBlock(), MatSetValuesBlocked()
1706 @*/
1707 PetscErrorCode PETSCMAT_DLLEXPORT MatSetValuesBlockedLocal(Mat mat,PetscInt nrow,const PetscInt irow[],PetscInt ncol,const PetscInt icol[],const PetscScalar y[],InsertMode addv)
1708 {
1709   PetscErrorCode ierr;
1710   PetscInt       irowm[2048],icolm[2048];
1711 
1712   PetscFunctionBegin;
1713   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1714   PetscValidType(mat,1);
1715   if (!nrow || !ncol) PetscFunctionReturn(0); /* no values to insert */
1716   PetscValidIntPointer(irow,3);
1717   PetscValidIntPointer(icol,5);
1718   PetscValidScalarPointer(y,6);
1719   ierr = MatPreallocated(mat);CHKERRQ(ierr);
1720   if (mat->insertmode == NOT_SET_VALUES) {
1721     mat->insertmode = addv;
1722   }
1723 #if defined(PETSC_USE_DEBUG)
1724   else if (mat->insertmode != addv) {
1725     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
1726   }
1727   if (!mat->bmapping) {
1728     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Local to global never set with MatSetLocalToGlobalMappingBlock()");
1729   }
1730   if (nrow > 2048 || ncol > 2048) {
1731     SETERRQ2(PETSC_ERR_SUP,"Number column/row indices must be <= 2048: are %D %D",nrow,ncol);
1732   }
1733   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1734 #endif
1735 
1736   if (mat->assembled) {
1737     mat->was_assembled = PETSC_TRUE;
1738     mat->assembled     = PETSC_FALSE;
1739   }
1740   ierr = PetscLogEventBegin(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
1741   ierr = ISLocalToGlobalMappingApply(mat->bmapping,nrow,irow,irowm);CHKERRQ(ierr);
1742   ierr = ISLocalToGlobalMappingApply(mat->bmapping,ncol,icol,icolm);CHKERRQ(ierr);
1743   if (mat->ops->setvaluesblocked) {
1744   ierr = (*mat->ops->setvaluesblocked)(mat,nrow,irowm,ncol,icolm,y,addv);CHKERRQ(ierr);
1745   } else {
1746     PetscInt buf[4096],*ibufm=0,*ibufn=0;
1747     PetscInt i,j,*iirowm,*iicolm,bs=mat->rmap->bs;
1748     if ((nrow+ncol)*bs <= 4096) {
1749       iirowm = buf; iicolm = buf + nrow*bs;
1750     } else {
1751       ierr = PetscMalloc2(nrow*bs,PetscInt,&ibufm,ncol*bs,PetscInt,&ibufn);CHKERRQ(ierr);
1752       iirowm = ibufm; iicolm = ibufn;
1753     }
1754     for (i=0; i<nrow; i++) {
1755       for (j=0; j<bs; j++) {
1756 	iirowm[i*bs+j] = bs*irowm[i] + j;
1757       }
1758     }
1759     for (i=0; i<ncol; i++) {
1760       for (j=0; j<bs; j++) {
1761 	iicolm[i*bs+j] = bs*icolm[i] + j;
1762       }
1763     }
1764     ierr = MatSetValues(mat,bs*nrow,iirowm,bs*ncol,iicolm,y,addv);CHKERRQ(ierr);
1765     ierr = PetscFree2(ibufm,ibufn);CHKERRQ(ierr);
1766   }
1767   ierr = PetscLogEventEnd(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
1768   PetscFunctionReturn(0);
1769 }
1770 
1771 /* --------------------------------------------------------*/
1772 #undef __FUNCT__
1773 #define __FUNCT__ "MatMult"
1774 /*@
1775    MatMult - Computes the matrix-vector product, y = Ax.
1776 
1777    Collective on Mat and Vec
1778 
1779    Input Parameters:
1780 +  mat - the matrix
1781 -  x   - the vector to be multiplied
1782 
1783    Output Parameters:
1784 .  y - the result
1785 
1786    Notes:
1787    The vectors x and y cannot be the same.  I.e., one cannot
1788    call MatMult(A,y,y).
1789 
1790    Level: beginner
1791 
1792    Concepts: matrix-vector product
1793 
1794 .seealso: MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
1795 @*/
1796 PetscErrorCode PETSCMAT_DLLEXPORT MatMult(Mat mat,Vec x,Vec y)
1797 {
1798   PetscErrorCode ierr;
1799 
1800   PetscFunctionBegin;
1801   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1802   PetscValidType(mat,1);
1803   PetscValidHeaderSpecific(x,VEC_COOKIE,2);
1804   PetscValidHeaderSpecific(y,VEC_COOKIE,3);
1805 
1806   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1807   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1808   if (x == y) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
1809 #ifndef PETSC_HAVE_CONSTRAINTS
1810   if (mat->cmap->N != x->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
1811   if (mat->rmap->N != y->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->rmap->N,y->map->N);
1812   if (mat->rmap->n != y->map->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: local dim %D %D",mat->rmap->n,y->map->n);
1813 #endif
1814   ierr = MatPreallocated(mat);CHKERRQ(ierr);
1815 
1816   if (mat->nullsp) {
1817     ierr = MatNullSpaceRemove(mat->nullsp,x,&x);CHKERRQ(ierr);
1818   }
1819 
1820   if (!mat->ops->mult) SETERRQ(PETSC_ERR_SUP,"This matrix type does not have a multiply defined");
1821   ierr = PetscLogEventBegin(MAT_Mult,mat,x,y,0);CHKERRQ(ierr);
1822   ierr = (*mat->ops->mult)(mat,x,y);CHKERRQ(ierr);
1823   ierr = PetscLogEventEnd(MAT_Mult,mat,x,y,0);CHKERRQ(ierr);
1824 
1825   if (mat->nullsp) {
1826     ierr = MatNullSpaceRemove(mat->nullsp,y,PETSC_NULL);CHKERRQ(ierr);
1827   }
1828   ierr = PetscObjectStateIncrease((PetscObject)y);CHKERRQ(ierr);
1829   PetscFunctionReturn(0);
1830 }
1831 
1832 #undef __FUNCT__
1833 #define __FUNCT__ "MatMultTranspose"
1834 /*@
1835    MatMultTranspose - Computes matrix transpose times a vector.
1836 
1837    Collective on Mat and Vec
1838 
1839    Input Parameters:
1840 +  mat - the matrix
1841 -  x   - the vector to be multilplied
1842 
1843    Output Parameters:
1844 .  y - the result
1845 
1846    Notes:
1847    The vectors x and y cannot be the same.  I.e., one cannot
1848    call MatMultTranspose(A,y,y).
1849 
1850    Level: beginner
1851 
1852    Concepts: matrix vector product^transpose
1853 
1854 .seealso: MatMult(), MatMultAdd(), MatMultTransposeAdd()
1855 @*/
1856 PetscErrorCode PETSCMAT_DLLEXPORT MatMultTranspose(Mat mat,Vec x,Vec y)
1857 {
1858   PetscErrorCode ierr;
1859 
1860   PetscFunctionBegin;
1861   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1862   PetscValidType(mat,1);
1863   PetscValidHeaderSpecific(x,VEC_COOKIE,2);
1864   PetscValidHeaderSpecific(y,VEC_COOKIE,3);
1865 
1866   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1867   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1868   if (x == y) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
1869 #ifndef PETSC_HAVE_CONSTRAINTS
1870   if (mat->rmap->N != x->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->rmap->N,x->map->N);
1871   if (mat->cmap->N != y->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->cmap->N,y->map->N);
1872 #endif
1873   ierr = MatPreallocated(mat);CHKERRQ(ierr);
1874 
1875   if (!mat->ops->multtranspose) SETERRQ(PETSC_ERR_SUP,"This matrix type does not have a multiply tranpose defined");
1876   ierr = PetscLogEventBegin(MAT_MultTranspose,mat,x,y,0);CHKERRQ(ierr);
1877   ierr = (*mat->ops->multtranspose)(mat,x,y);CHKERRQ(ierr);
1878   ierr = PetscLogEventEnd(MAT_MultTranspose,mat,x,y,0);CHKERRQ(ierr);
1879   ierr = PetscObjectStateIncrease((PetscObject)y);CHKERRQ(ierr);
1880   PetscFunctionReturn(0);
1881 }
1882 
1883 #undef __FUNCT__
1884 #define __FUNCT__ "MatMultAdd"
1885 /*@
1886     MatMultAdd -  Computes v3 = v2 + A * v1.
1887 
1888     Collective on Mat and Vec
1889 
1890     Input Parameters:
1891 +   mat - the matrix
1892 -   v1, v2 - the vectors
1893 
1894     Output Parameters:
1895 .   v3 - the result
1896 
1897     Notes:
1898     The vectors v1 and v3 cannot be the same.  I.e., one cannot
1899     call MatMultAdd(A,v1,v2,v1).
1900 
1901     Level: beginner
1902 
1903     Concepts: matrix vector product^addition
1904 
1905 .seealso: MatMultTranspose(), MatMult(), MatMultTransposeAdd()
1906 @*/
1907 PetscErrorCode PETSCMAT_DLLEXPORT MatMultAdd(Mat mat,Vec v1,Vec v2,Vec v3)
1908 {
1909   PetscErrorCode ierr;
1910 
1911   PetscFunctionBegin;
1912   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1913   PetscValidType(mat,1);
1914   PetscValidHeaderSpecific(v1,VEC_COOKIE,2);
1915   PetscValidHeaderSpecific(v2,VEC_COOKIE,3);
1916   PetscValidHeaderSpecific(v3,VEC_COOKIE,4);
1917 
1918   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1919   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1920   if (mat->cmap->N != v1->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v1: global dim %D %D",mat->cmap->N,v1->map->N);
1921   if (mat->rmap->N != v2->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v2: global dim %D %D",mat->rmap->N,v2->map->N);
1922   if (mat->rmap->N != v3->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v3: global dim %D %D",mat->rmap->N,v3->map->N);
1923   if (mat->rmap->n != v3->map->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v3: local dim %D %D",mat->rmap->n,v3->map->n);
1924   if (mat->rmap->n != v2->map->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v2: local dim %D %D",mat->rmap->n,v2->map->n);
1925   if (v1 == v3) SETERRQ(PETSC_ERR_ARG_IDN,"v1 and v3 must be different vectors");
1926   ierr = MatPreallocated(mat);CHKERRQ(ierr);
1927 
1928   ierr = PetscLogEventBegin(MAT_MultAdd,mat,v1,v2,v3);CHKERRQ(ierr);
1929   ierr = (*mat->ops->multadd)(mat,v1,v2,v3);CHKERRQ(ierr);
1930   ierr = PetscLogEventEnd(MAT_MultAdd,mat,v1,v2,v3);CHKERRQ(ierr);
1931   ierr = PetscObjectStateIncrease((PetscObject)v3);CHKERRQ(ierr);
1932   PetscFunctionReturn(0);
1933 }
1934 
1935 #undef __FUNCT__
1936 #define __FUNCT__ "MatMultTransposeAdd"
1937 /*@
1938    MatMultTransposeAdd - Computes v3 = v2 + A' * v1.
1939 
1940    Collective on Mat and Vec
1941 
1942    Input Parameters:
1943 +  mat - the matrix
1944 -  v1, v2 - the vectors
1945 
1946    Output Parameters:
1947 .  v3 - the result
1948 
1949    Notes:
1950    The vectors v1 and v3 cannot be the same.  I.e., one cannot
1951    call MatMultTransposeAdd(A,v1,v2,v1).
1952 
1953    Level: beginner
1954 
1955    Concepts: matrix vector product^transpose and addition
1956 
1957 .seealso: MatMultTranspose(), MatMultAdd(), MatMult()
1958 @*/
1959 PetscErrorCode PETSCMAT_DLLEXPORT MatMultTransposeAdd(Mat mat,Vec v1,Vec v2,Vec v3)
1960 {
1961   PetscErrorCode ierr;
1962 
1963   PetscFunctionBegin;
1964   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
1965   PetscValidType(mat,1);
1966   PetscValidHeaderSpecific(v1,VEC_COOKIE,2);
1967   PetscValidHeaderSpecific(v2,VEC_COOKIE,3);
1968   PetscValidHeaderSpecific(v3,VEC_COOKIE,4);
1969 
1970   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
1971   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
1972   if (!mat->ops->multtransposeadd) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
1973   if (v1 == v3) SETERRQ(PETSC_ERR_ARG_IDN,"v1 and v3 must be different vectors");
1974   if (mat->rmap->N != v1->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v1: global dim %D %D",mat->rmap->N,v1->map->N);
1975   if (mat->cmap->N != v2->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v2: global dim %D %D",mat->cmap->N,v2->map->N);
1976   if (mat->cmap->N != v3->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec v3: global dim %D %D",mat->cmap->N,v3->map->N);
1977   ierr = MatPreallocated(mat);CHKERRQ(ierr);
1978 
1979   ierr = PetscLogEventBegin(MAT_MultTransposeAdd,mat,v1,v2,v3);CHKERRQ(ierr);
1980   ierr = (*mat->ops->multtransposeadd)(mat,v1,v2,v3);CHKERRQ(ierr);
1981   ierr = PetscLogEventEnd(MAT_MultTransposeAdd,mat,v1,v2,v3);CHKERRQ(ierr);
1982   ierr = PetscObjectStateIncrease((PetscObject)v3);CHKERRQ(ierr);
1983   PetscFunctionReturn(0);
1984 }
1985 
1986 #undef __FUNCT__
1987 #define __FUNCT__ "MatMultConstrained"
1988 /*@
1989    MatMultConstrained - The inner multiplication routine for a
1990    constrained matrix P^T A P.
1991 
1992    Collective on Mat and Vec
1993 
1994    Input Parameters:
1995 +  mat - the matrix
1996 -  x   - the vector to be multilplied
1997 
1998    Output Parameters:
1999 .  y - the result
2000 
2001    Notes:
2002    The vectors x and y cannot be the same.  I.e., one cannot
2003    call MatMult(A,y,y).
2004 
2005    Level: beginner
2006 
2007 .keywords: matrix, multiply, matrix-vector product, constraint
2008 .seealso: MatMult(), MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
2009 @*/
2010 PetscErrorCode PETSCMAT_DLLEXPORT MatMultConstrained(Mat mat,Vec x,Vec y)
2011 {
2012   PetscErrorCode ierr;
2013 
2014   PetscFunctionBegin;
2015   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2016   PetscValidHeaderSpecific(x,VEC_COOKIE,2);
2017   PetscValidHeaderSpecific(y,VEC_COOKIE,3);
2018   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2019   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2020   if (x == y) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2021   if (mat->cmap->N != x->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
2022   if (mat->rmap->N != y->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->rmap->N,y->map->N);
2023   if (mat->rmap->n != y->map->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: local dim %D %D",mat->rmap->n,y->map->n);
2024 
2025   ierr = PetscLogEventBegin(MAT_MultConstrained,mat,x,y,0);CHKERRQ(ierr);
2026   ierr = (*mat->ops->multconstrained)(mat,x,y);CHKERRQ(ierr);
2027   ierr = PetscLogEventEnd(MAT_MultConstrained,mat,x,y,0);CHKERRQ(ierr);
2028   ierr = PetscObjectStateIncrease((PetscObject)y);CHKERRQ(ierr);
2029 
2030   PetscFunctionReturn(0);
2031 }
2032 
2033 #undef __FUNCT__
2034 #define __FUNCT__ "MatMultTransposeConstrained"
2035 /*@
2036    MatMultTransposeConstrained - The inner multiplication routine for a
2037    constrained matrix P^T A^T P.
2038 
2039    Collective on Mat and Vec
2040 
2041    Input Parameters:
2042 +  mat - the matrix
2043 -  x   - the vector to be multilplied
2044 
2045    Output Parameters:
2046 .  y - the result
2047 
2048    Notes:
2049    The vectors x and y cannot be the same.  I.e., one cannot
2050    call MatMult(A,y,y).
2051 
2052    Level: beginner
2053 
2054 .keywords: matrix, multiply, matrix-vector product, constraint
2055 .seealso: MatMult(), MatMultTranspose(), MatMultAdd(), MatMultTransposeAdd()
2056 @*/
2057 PetscErrorCode PETSCMAT_DLLEXPORT MatMultTransposeConstrained(Mat mat,Vec x,Vec y)
2058 {
2059   PetscErrorCode ierr;
2060 
2061   PetscFunctionBegin;
2062   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2063   PetscValidHeaderSpecific(x,VEC_COOKIE,2);
2064   PetscValidHeaderSpecific(y,VEC_COOKIE,3);
2065   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2066   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2067   if (x == y) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"x and y must be different vectors");
2068   if (mat->rmap->N != x->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
2069   if (mat->cmap->N != y->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->rmap->N,y->map->N);
2070 
2071   ierr = PetscLogEventBegin(MAT_MultConstrained,mat,x,y,0);CHKERRQ(ierr);
2072   ierr = (*mat->ops->multtransposeconstrained)(mat,x,y);CHKERRQ(ierr);
2073   ierr = PetscLogEventEnd(MAT_MultConstrained,mat,x,y,0);CHKERRQ(ierr);
2074   ierr = PetscObjectStateIncrease((PetscObject)y);CHKERRQ(ierr);
2075 
2076   PetscFunctionReturn(0);
2077 }
2078 /* ------------------------------------------------------------*/
2079 #undef __FUNCT__
2080 #define __FUNCT__ "MatGetInfo"
2081 /*@
2082    MatGetInfo - Returns information about matrix storage (number of
2083    nonzeros, memory, etc.).
2084 
2085    Collective on Mat if MAT_GLOBAL_MAX or MAT_GLOBAL_SUM is used
2086    as the flag
2087 
2088    Input Parameters:
2089 .  mat - the matrix
2090 
2091    Output Parameters:
2092 +  flag - flag indicating the type of parameters to be returned
2093    (MAT_LOCAL - local matrix, MAT_GLOBAL_MAX - maximum over all processors,
2094    MAT_GLOBAL_SUM - sum over all processors)
2095 -  info - matrix information context
2096 
2097    Notes:
2098    The MatInfo context contains a variety of matrix data, including
2099    number of nonzeros allocated and used, number of mallocs during
2100    matrix assembly, etc.  Additional information for factored matrices
2101    is provided (such as the fill ratio, number of mallocs during
2102    factorization, etc.).  Much of this info is printed to PETSC_STDOUT
2103    when using the runtime options
2104 $       -info -mat_view_info
2105 
2106    Example for C/C++ Users:
2107    See the file ${PETSC_DIR}/include/petscmat.h for a complete list of
2108    data within the MatInfo context.  For example,
2109 .vb
2110       MatInfo info;
2111       Mat     A;
2112       double  mal, nz_a, nz_u;
2113 
2114       MatGetInfo(A,MAT_LOCAL,&info);
2115       mal  = info.mallocs;
2116       nz_a = info.nz_allocated;
2117 .ve
2118 
2119    Example for Fortran Users:
2120    Fortran users should declare info as a double precision
2121    array of dimension MAT_INFO_SIZE, and then extract the parameters
2122    of interest.  See the file ${PETSC_DIR}/include/finclude/petscmat.h
2123    a complete list of parameter names.
2124 .vb
2125       double  precision info(MAT_INFO_SIZE)
2126       double  precision mal, nz_a
2127       Mat     A
2128       integer ierr
2129 
2130       call MatGetInfo(A,MAT_LOCAL,info,ierr)
2131       mal = info(MAT_INFO_MALLOCS)
2132       nz_a = info(MAT_INFO_NZ_ALLOCATED)
2133 .ve
2134 
2135     Level: intermediate
2136 
2137     Concepts: matrices^getting information on
2138 
2139 @*/
2140 PetscErrorCode PETSCMAT_DLLEXPORT MatGetInfo(Mat mat,MatInfoType flag,MatInfo *info)
2141 {
2142   PetscErrorCode ierr;
2143 
2144   PetscFunctionBegin;
2145   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2146   PetscValidType(mat,1);
2147   PetscValidPointer(info,3);
2148   if (!mat->ops->getinfo) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2149   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2150   ierr = (*mat->ops->getinfo)(mat,flag,info);CHKERRQ(ierr);
2151   PetscFunctionReturn(0);
2152 }
2153 
2154 /* ----------------------------------------------------------*/
2155 #undef __FUNCT__
2156 #define __FUNCT__ "MatILUDTFactor"
2157 /*@C
2158    MatILUDTFactor - Performs a drop tolerance ILU factorization.
2159 
2160    Collective on Mat
2161 
2162    Input Parameters:
2163 +  mat - the matrix
2164 .  row - row permutation
2165 .  col - column permutation
2166 -  info - information about the factorization to be done
2167 
2168    Output Parameters:
2169 .  fact - the factored matrix
2170 
2171    Level: developer
2172 
2173    Notes:
2174    Most users should employ the simplified KSP interface for linear solvers
2175    instead of working directly with matrix algebra routines such as this.
2176    See, e.g., KSPCreate().
2177 
2178    This is currently only supported for the SeqAIJ matrix format using code
2179    from Yousef Saad's SPARSEKIT2  package (translated to C with f2c) and/or
2180    Matlab. SPARSEKIT2 is copyrighted by Yousef Saad with the GNU copyright
2181    and thus can be distributed with PETSc.
2182 
2183     Concepts: matrices^ILUDT factorization
2184 
2185 .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor(), MatFactorInfo
2186 @*/
2187 PetscErrorCode PETSCMAT_DLLEXPORT MatILUDTFactor(Mat mat,IS row,IS col,const MatFactorInfo *info,Mat *fact)
2188 {
2189   PetscErrorCode ierr;
2190 
2191   PetscFunctionBegin;
2192   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2193   PetscValidType(mat,1);
2194   if (row) PetscValidHeaderSpecific(row,IS_COOKIE,2);
2195   if (col) PetscValidHeaderSpecific(col,IS_COOKIE,3);
2196   PetscValidPointer(info,4);
2197   PetscValidPointer(fact,5);
2198   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2199   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2200   if (!mat->ops->iludtfactor) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2201   if (mat->rmap->N != mat->cmap->N) SETERRQ(PETSC_ERR_ARG_WRONG,"matrix must be square");
2202   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2203   ierr = PetscLogEventBegin(MAT_ILUFactor,mat,row,col,0);CHKERRQ(ierr);
2204   ierr = (*mat->ops->iludtfactor)(mat,row,col,info,fact);CHKERRQ(ierr);
2205   ierr = PetscLogEventEnd(MAT_ILUFactor,mat,row,col,0);CHKERRQ(ierr);
2206   ierr = PetscObjectStateIncrease((PetscObject)*fact);CHKERRQ(ierr);
2207 
2208   PetscFunctionReturn(0);
2209 }
2210 
2211 #undef __FUNCT__
2212 #define __FUNCT__ "MatLUFactor"
2213 /*@
2214    MatLUFactor - Performs in-place LU factorization of matrix.
2215 
2216    Collective on Mat
2217 
2218    Input Parameters:
2219 +  mat - the matrix
2220 .  row - row permutation
2221 .  col - column permutation
2222 -  info - options for factorization, includes
2223 $          fill - expected fill as ratio of original fill.
2224 $          dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
2225 $                   Run with the option -info to determine an optimal value to use
2226 
2227    Notes:
2228    Most users should employ the simplified KSP interface for linear solvers
2229    instead of working directly with matrix algebra routines such as this.
2230    See, e.g., KSPCreate().
2231 
2232    This changes the state of the matrix to a factored matrix; it cannot be used
2233    for example with MatSetValues() unless one first calls MatSetUnfactored().
2234 
2235    Level: developer
2236 
2237    Concepts: matrices^LU factorization
2238 
2239 .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor(),
2240           MatGetOrdering(), MatSetUnfactored(), MatFactorInfo
2241 
2242 @*/
2243 PetscErrorCode PETSCMAT_DLLEXPORT MatLUFactor(Mat mat,IS row,IS col,const MatFactorInfo *info)
2244 {
2245   PetscErrorCode ierr;
2246 
2247   PetscFunctionBegin;
2248   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2249   if (row) PetscValidHeaderSpecific(row,IS_COOKIE,2);
2250   if (col) PetscValidHeaderSpecific(col,IS_COOKIE,3);
2251   PetscValidPointer(info,4);
2252   PetscValidType(mat,1);
2253   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2254   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2255   if (!mat->ops->lufactor) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2256   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2257 
2258   ierr = PetscLogEventBegin(MAT_LUFactor,mat,row,col,0);CHKERRQ(ierr);
2259   ierr = (*mat->ops->lufactor)(mat,row,col,info);CHKERRQ(ierr);
2260   ierr = PetscLogEventEnd(MAT_LUFactor,mat,row,col,0);CHKERRQ(ierr);
2261   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
2262   PetscFunctionReturn(0);
2263 }
2264 
2265 #undef __FUNCT__
2266 #define __FUNCT__ "MatILUFactor"
2267 /*@
2268    MatILUFactor - Performs in-place ILU factorization of matrix.
2269 
2270    Collective on Mat
2271 
2272    Input Parameters:
2273 +  mat - the matrix
2274 .  row - row permutation
2275 .  col - column permutation
2276 -  info - structure containing
2277 $      levels - number of levels of fill.
2278 $      expected fill - as ratio of original fill.
2279 $      1 or 0 - indicating force fill on diagonal (improves robustness for matrices
2280                 missing diagonal entries)
2281 
2282    Notes:
2283    Probably really in-place only when level of fill is zero, otherwise allocates
2284    new space to store factored matrix and deletes previous memory.
2285 
2286    Most users should employ the simplified KSP interface for linear solvers
2287    instead of working directly with matrix algebra routines such as this.
2288    See, e.g., KSPCreate().
2289 
2290    Level: developer
2291 
2292    Concepts: matrices^ILU factorization
2293 
2294 .seealso: MatILUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor(), MatFactorInfo
2295 @*/
2296 PetscErrorCode PETSCMAT_DLLEXPORT MatILUFactor(Mat mat,IS row,IS col,const MatFactorInfo *info)
2297 {
2298   PetscErrorCode ierr;
2299 
2300   PetscFunctionBegin;
2301   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2302   if (row) PetscValidHeaderSpecific(row,IS_COOKIE,2);
2303   if (col) PetscValidHeaderSpecific(col,IS_COOKIE,3);
2304   PetscValidPointer(info,4);
2305   PetscValidType(mat,1);
2306   if (mat->rmap->N != mat->cmap->N) SETERRQ(PETSC_ERR_ARG_WRONG,"matrix must be square");
2307   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2308   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2309   if (!mat->ops->ilufactor) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2310   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2311 
2312   ierr = PetscLogEventBegin(MAT_ILUFactor,mat,row,col,0);CHKERRQ(ierr);
2313   ierr = (*mat->ops->ilufactor)(mat,row,col,info);CHKERRQ(ierr);
2314   ierr = PetscLogEventEnd(MAT_ILUFactor,mat,row,col,0);CHKERRQ(ierr);
2315   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
2316   PetscFunctionReturn(0);
2317 }
2318 
2319 #undef __FUNCT__
2320 #define __FUNCT__ "MatLUFactorSymbolic"
2321 /*@
2322    MatLUFactorSymbolic - Performs symbolic LU factorization of matrix.
2323    Call this routine before calling MatLUFactorNumeric().
2324 
2325    Collective on Mat
2326 
2327    Input Parameters:
2328 +  fact - the factor matrix obtained with MatGetFactor()
2329 .  mat - the matrix
2330 .  row, col - row and column permutations
2331 -  info - options for factorization, includes
2332 $          fill - expected fill as ratio of original fill.
2333 $          dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
2334 $                   Run with the option -info to determine an optimal value to use
2335 
2336 
2337    Notes:
2338    See the users manual for additional information about
2339    choosing the fill factor for better efficiency.
2340 
2341    Most users should employ the simplified KSP interface for linear solvers
2342    instead of working directly with matrix algebra routines such as this.
2343    See, e.g., KSPCreate().
2344 
2345    Level: developer
2346 
2347    Concepts: matrices^LU symbolic factorization
2348 
2349 .seealso: MatLUFactor(), MatLUFactorNumeric(), MatCholeskyFactor(), MatFactorInfo
2350 @*/
2351 PetscErrorCode PETSCMAT_DLLEXPORT MatLUFactorSymbolic(Mat fact,Mat mat,IS row,IS col,const MatFactorInfo *info)
2352 {
2353   PetscErrorCode ierr;
2354 
2355   PetscFunctionBegin;
2356   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2357   if (row) PetscValidHeaderSpecific(row,IS_COOKIE,2);
2358   if (col) PetscValidHeaderSpecific(col,IS_COOKIE,3);
2359   PetscValidPointer(info,4);
2360   PetscValidType(mat,1);
2361   PetscValidPointer(fact,5);
2362   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2363   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2364   if (!(fact)->ops->lufactorsymbolic) SETERRQ1(PETSC_ERR_SUP,"Matrix type %s  symbolic LU",((PetscObject)mat)->type_name);
2365   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2366 
2367   ierr = PetscLogEventBegin(MAT_LUFactorSymbolic,mat,row,col,0);CHKERRQ(ierr);
2368   ierr = (fact->ops->lufactorsymbolic)(fact,mat,row,col,info);CHKERRQ(ierr);
2369   ierr = PetscLogEventEnd(MAT_LUFactorSymbolic,mat,row,col,0);CHKERRQ(ierr);
2370   ierr = PetscObjectStateIncrease((PetscObject)fact);CHKERRQ(ierr);
2371   PetscFunctionReturn(0);
2372 }
2373 
2374 #undef __FUNCT__
2375 #define __FUNCT__ "MatLUFactorNumeric"
2376 /*@
2377    MatLUFactorNumeric - Performs numeric LU factorization of a matrix.
2378    Call this routine after first calling MatLUFactorSymbolic().
2379 
2380    Collective on Mat
2381 
2382    Input Parameters:
2383 +  fact - the factor matrix obtained with MatGetFactor()
2384 .  mat - the matrix
2385 -  info - options for factorization
2386 
2387    Notes:
2388    See MatLUFactor() for in-place factorization.  See
2389    MatCholeskyFactorNumeric() for the symmetric, positive definite case.
2390 
2391    Most users should employ the simplified KSP interface for linear solvers
2392    instead of working directly with matrix algebra routines such as this.
2393    See, e.g., KSPCreate().
2394 
2395    Level: developer
2396 
2397    Concepts: matrices^LU numeric factorization
2398 
2399 .seealso: MatLUFactorSymbolic(), MatLUFactor(), MatCholeskyFactor()
2400 @*/
2401 PetscErrorCode PETSCMAT_DLLEXPORT MatLUFactorNumeric(Mat fact,Mat mat,const MatFactorInfo *info)
2402 {
2403   PetscErrorCode ierr;
2404 
2405   PetscFunctionBegin;
2406   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2407   PetscValidType(mat,1);
2408   PetscValidPointer(fact,2);
2409   PetscValidHeaderSpecific(fact,MAT_COOKIE,2);
2410   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2411   if (mat->rmap->N != (fact)->rmap->N || mat->cmap->N != (fact)->cmap->N) {
2412     SETERRQ4(PETSC_ERR_ARG_SIZ,"Mat mat,Mat fact: global dimensions are different %D should = %D %D should = %D",mat->rmap->N,(fact)->rmap->N,mat->cmap->N,(fact)->cmap->N);
2413   }
2414   if (!(fact)->ops->lufactornumeric) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2415   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2416   ierr = PetscLogEventBegin(MAT_LUFactorNumeric,mat,fact,0,0);CHKERRQ(ierr);
2417   ierr = (fact->ops->lufactornumeric)(fact,mat,info);CHKERRQ(ierr);
2418   ierr = PetscLogEventEnd(MAT_LUFactorNumeric,mat,fact,0,0);CHKERRQ(ierr);
2419 
2420   ierr = MatView_Private(fact);CHKERRQ(ierr);
2421   ierr = PetscObjectStateIncrease((PetscObject)fact);CHKERRQ(ierr);
2422   PetscFunctionReturn(0);
2423 }
2424 
2425 #undef __FUNCT__
2426 #define __FUNCT__ "MatCholeskyFactor"
2427 /*@
2428    MatCholeskyFactor - Performs in-place Cholesky factorization of a
2429    symmetric matrix.
2430 
2431    Collective on Mat
2432 
2433    Input Parameters:
2434 +  mat - the matrix
2435 .  perm - row and column permutations
2436 -  f - expected fill as ratio of original fill
2437 
2438    Notes:
2439    See MatLUFactor() for the nonsymmetric case.  See also
2440    MatCholeskyFactorSymbolic(), and MatCholeskyFactorNumeric().
2441 
2442    Most users should employ the simplified KSP interface for linear solvers
2443    instead of working directly with matrix algebra routines such as this.
2444    See, e.g., KSPCreate().
2445 
2446    Level: developer
2447 
2448    Concepts: matrices^Cholesky factorization
2449 
2450 .seealso: MatLUFactor(), MatCholeskyFactorSymbolic(), MatCholeskyFactorNumeric()
2451           MatGetOrdering()
2452 
2453 @*/
2454 PetscErrorCode PETSCMAT_DLLEXPORT MatCholeskyFactor(Mat mat,IS perm,const MatFactorInfo *info)
2455 {
2456   PetscErrorCode ierr;
2457 
2458   PetscFunctionBegin;
2459   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2460   PetscValidType(mat,1);
2461   PetscValidHeaderSpecific(perm,IS_COOKIE,2);
2462   PetscValidPointer(info,3);
2463   if (mat->rmap->N != mat->cmap->N) SETERRQ(PETSC_ERR_ARG_WRONG,"Matrix must be square");
2464   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2465   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2466   if (!mat->ops->choleskyfactor) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2467   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2468 
2469   ierr = PetscLogEventBegin(MAT_CholeskyFactor,mat,perm,0,0);CHKERRQ(ierr);
2470   ierr = (*mat->ops->choleskyfactor)(mat,perm,info);CHKERRQ(ierr);
2471   ierr = PetscLogEventEnd(MAT_CholeskyFactor,mat,perm,0,0);CHKERRQ(ierr);
2472   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
2473   PetscFunctionReturn(0);
2474 }
2475 
2476 #undef __FUNCT__
2477 #define __FUNCT__ "MatCholeskyFactorSymbolic"
2478 /*@
2479    MatCholeskyFactorSymbolic - Performs symbolic Cholesky factorization
2480    of a symmetric matrix.
2481 
2482    Collective on Mat
2483 
2484    Input Parameters:
2485 +  fact - the factor matrix obtained with MatGetFactor()
2486 .  mat - the matrix
2487 .  perm - row and column permutations
2488 -  info - options for factorization, includes
2489 $          fill - expected fill as ratio of original fill.
2490 $          dtcol - pivot tolerance (0 no pivot, 1 full column pivoting)
2491 $                   Run with the option -info to determine an optimal value to use
2492 
2493    Notes:
2494    See MatLUFactorSymbolic() for the nonsymmetric case.  See also
2495    MatCholeskyFactor() and MatCholeskyFactorNumeric().
2496 
2497    Most users should employ the simplified KSP interface for linear solvers
2498    instead of working directly with matrix algebra routines such as this.
2499    See, e.g., KSPCreate().
2500 
2501    Level: developer
2502 
2503    Concepts: matrices^Cholesky symbolic factorization
2504 
2505 .seealso: MatLUFactorSymbolic(), MatCholeskyFactor(), MatCholeskyFactorNumeric()
2506           MatGetOrdering()
2507 
2508 @*/
2509 PetscErrorCode PETSCMAT_DLLEXPORT MatCholeskyFactorSymbolic(Mat fact,Mat mat,IS perm,const MatFactorInfo *info)
2510 {
2511   PetscErrorCode ierr;
2512 
2513   PetscFunctionBegin;
2514   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2515   PetscValidType(mat,1);
2516   if (perm) PetscValidHeaderSpecific(perm,IS_COOKIE,2);
2517   PetscValidPointer(info,3);
2518   PetscValidPointer(fact,4);
2519   if (mat->rmap->N != mat->cmap->N) SETERRQ(PETSC_ERR_ARG_WRONG,"Matrix must be square");
2520   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2521   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2522   if (!(fact)->ops->choleskyfactorsymbolic) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2523   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2524 
2525   ierr = PetscLogEventBegin(MAT_CholeskyFactorSymbolic,mat,perm,0,0);CHKERRQ(ierr);
2526   ierr = (fact->ops->choleskyfactorsymbolic)(fact,mat,perm,info);CHKERRQ(ierr);
2527   ierr = PetscLogEventEnd(MAT_CholeskyFactorSymbolic,mat,perm,0,0);CHKERRQ(ierr);
2528   ierr = PetscObjectStateIncrease((PetscObject)fact);CHKERRQ(ierr);
2529   PetscFunctionReturn(0);
2530 }
2531 
2532 #undef __FUNCT__
2533 #define __FUNCT__ "MatCholeskyFactorNumeric"
2534 /*@
2535    MatCholeskyFactorNumeric - Performs numeric Cholesky factorization
2536    of a symmetric matrix. Call this routine after first calling
2537    MatCholeskyFactorSymbolic().
2538 
2539    Collective on Mat
2540 
2541    Input Parameters:
2542 +  fact - the factor matrix obtained with MatGetFactor()
2543 .  mat - the initial matrix
2544 .  info - options for factorization
2545 -  fact - the symbolic factor of mat
2546 
2547 
2548    Notes:
2549    Most users should employ the simplified KSP interface for linear solvers
2550    instead of working directly with matrix algebra routines such as this.
2551    See, e.g., KSPCreate().
2552 
2553    Level: developer
2554 
2555    Concepts: matrices^Cholesky numeric factorization
2556 
2557 .seealso: MatCholeskyFactorSymbolic(), MatCholeskyFactor(), MatLUFactorNumeric()
2558 @*/
2559 PetscErrorCode PETSCMAT_DLLEXPORT MatCholeskyFactorNumeric(Mat fact,Mat mat,const MatFactorInfo *info)
2560 {
2561   PetscErrorCode ierr;
2562 
2563   PetscFunctionBegin;
2564   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2565   PetscValidType(mat,1);
2566   PetscValidPointer(fact,2);
2567   PetscValidHeaderSpecific(fact,MAT_COOKIE,2);
2568   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2569   if (!(fact)->ops->choleskyfactornumeric) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2570   if (mat->rmap->N != (fact)->rmap->N || mat->cmap->N != (fact)->cmap->N) {
2571     SETERRQ4(PETSC_ERR_ARG_SIZ,"Mat mat,Mat fact: global dim %D should = %D %D should = %D",mat->rmap->N,(fact)->rmap->N,mat->cmap->N,(fact)->cmap->N);
2572   }
2573   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2574 
2575   ierr = PetscLogEventBegin(MAT_CholeskyFactorNumeric,mat,fact,0,0);CHKERRQ(ierr);
2576   ierr = (fact->ops->choleskyfactornumeric)(fact,mat,info);CHKERRQ(ierr);
2577   ierr = PetscLogEventEnd(MAT_CholeskyFactorNumeric,mat,fact,0,0);CHKERRQ(ierr);
2578 
2579   ierr = MatView_Private(fact);CHKERRQ(ierr);
2580   ierr = PetscObjectStateIncrease((PetscObject)fact);CHKERRQ(ierr);
2581   PetscFunctionReturn(0);
2582 }
2583 
2584 /* ----------------------------------------------------------------*/
2585 #undef __FUNCT__
2586 #define __FUNCT__ "MatSolve"
2587 /*@
2588    MatSolve - Solves A x = b, given a factored matrix.
2589 
2590    Collective on Mat and Vec
2591 
2592    Input Parameters:
2593 +  mat - the factored matrix
2594 -  b - the right-hand-side vector
2595 
2596    Output Parameter:
2597 .  x - the result vector
2598 
2599    Notes:
2600    The vectors b and x cannot be the same.  I.e., one cannot
2601    call MatSolve(A,x,x).
2602 
2603    Notes:
2604    Most users should employ the simplified KSP interface for linear solvers
2605    instead of working directly with matrix algebra routines such as this.
2606    See, e.g., KSPCreate().
2607 
2608    Level: developer
2609 
2610    Concepts: matrices^triangular solves
2611 
2612 .seealso: MatSolveAdd(), MatSolveTranspose(), MatSolveTransposeAdd()
2613 @*/
2614 PetscErrorCode PETSCMAT_DLLEXPORT MatSolve(Mat mat,Vec b,Vec x)
2615 {
2616   PetscErrorCode ierr;
2617 
2618   PetscFunctionBegin;
2619   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2620   PetscValidType(mat,1);
2621   PetscValidHeaderSpecific(b,VEC_COOKIE,2);
2622   PetscValidHeaderSpecific(x,VEC_COOKIE,3);
2623   PetscCheckSameComm(mat,1,b,2);
2624   PetscCheckSameComm(mat,1,x,3);
2625   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
2626   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2627   if (mat->cmap->N != x->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
2628   if (mat->rmap->N != b->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap->N,b->map->N);
2629   if (mat->rmap->n != b->map->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap->n,b->map->n);
2630   if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(0);
2631   if (!mat->ops->solve) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2632   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2633 
2634   ierr = PetscLogEventBegin(MAT_Solve,mat,b,x,0);CHKERRQ(ierr);
2635   ierr = (*mat->ops->solve)(mat,b,x);CHKERRQ(ierr);
2636   ierr = PetscLogEventEnd(MAT_Solve,mat,b,x,0);CHKERRQ(ierr);
2637   ierr = PetscObjectStateIncrease((PetscObject)x);CHKERRQ(ierr);
2638   PetscFunctionReturn(0);
2639 }
2640 
2641 #undef __FUNCT__
2642 #define __FUNCT__ "MatMatSolve_Basic"
2643 PetscErrorCode PETSCMAT_DLLEXPORT MatMatSolve_Basic(Mat A,Mat B,Mat X)
2644 {
2645   PetscErrorCode ierr;
2646   Vec            b,x;
2647   PetscInt       m,N,i;
2648   PetscScalar    *bb,*xx;
2649 
2650   PetscFunctionBegin;
2651   ierr = MatGetArray(B,&bb);CHKERRQ(ierr);
2652   ierr = MatGetArray(X,&xx);CHKERRQ(ierr);
2653   ierr = MatGetLocalSize(B,&m,PETSC_NULL);CHKERRQ(ierr);  /* number local rows */
2654   ierr = MatGetSize(B,PETSC_NULL,&N);CHKERRQ(ierr);       /* total columns in dense matrix */
2655   ierr = VecCreateMPIWithArray(((PetscObject)A)->comm,m,PETSC_DETERMINE,PETSC_NULL,&b);CHKERRQ(ierr);
2656   ierr = VecCreateMPIWithArray(((PetscObject)A)->comm,m,PETSC_DETERMINE,PETSC_NULL,&x);CHKERRQ(ierr);
2657   for (i=0; i<N; i++) {
2658     ierr = VecPlaceArray(b,bb + i*m);CHKERRQ(ierr);
2659     ierr = VecPlaceArray(x,xx + i*m);CHKERRQ(ierr);
2660     ierr = MatSolve(A,b,x);CHKERRQ(ierr);
2661     ierr = VecResetArray(x);CHKERRQ(ierr);
2662     ierr = VecResetArray(b);CHKERRQ(ierr);
2663   }
2664   ierr = VecDestroy(b);CHKERRQ(ierr);
2665   ierr = VecDestroy(x);CHKERRQ(ierr);
2666   ierr = MatRestoreArray(B,&bb);CHKERRQ(ierr);
2667   ierr = MatRestoreArray(X,&xx);CHKERRQ(ierr);
2668   PetscFunctionReturn(0);
2669 }
2670 
2671 #undef __FUNCT__
2672 #define __FUNCT__ "MatMatSolve"
2673 /*@
2674    MatMatSolve - Solves A X = B, given a factored matrix.
2675 
2676    Collective on Mat
2677 
2678    Input Parameters:
2679 +  mat - the factored matrix
2680 -  B - the right-hand-side matrix  (dense matrix)
2681 
2682    Output Parameter:
2683 .  X - the result matrix (dense matrix)
2684 
2685    Notes:
2686    The matrices b and x cannot be the same.  I.e., one cannot
2687    call MatMatSolve(A,x,x).
2688 
2689    Notes:
2690    Most users should usually employ the simplified KSP interface for linear solvers
2691    instead of working directly with matrix algebra routines such as this.
2692    See, e.g., KSPCreate(). However KSP can only solve for one vector (column of X)
2693    at a time.
2694 
2695    Level: developer
2696 
2697    Concepts: matrices^triangular solves
2698 
2699 .seealso: MatMatSolveAdd(), MatMatSolveTranspose(), MatMatSolveTransposeAdd(), MatLUFactor(), MatCholeskyFactor()
2700 @*/
2701 PetscErrorCode PETSCMAT_DLLEXPORT MatMatSolve(Mat A,Mat B,Mat X)
2702 {
2703   PetscErrorCode ierr;
2704 
2705   PetscFunctionBegin;
2706   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
2707   PetscValidType(A,1);
2708   PetscValidHeaderSpecific(B,MAT_COOKIE,2);
2709   PetscValidHeaderSpecific(X,MAT_COOKIE,3);
2710   PetscCheckSameComm(A,1,B,2);
2711   PetscCheckSameComm(A,1,X,3);
2712   if (X == B) SETERRQ(PETSC_ERR_ARG_IDN,"X and B must be different matrices");
2713   if (!A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2714   if (A->cmap->N != X->rmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat A,Mat X: global dim %D %D",A->cmap->N,X->rmap->N);
2715   if (A->rmap->N != B->rmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat A,Mat B: global dim %D %D",A->rmap->N,B->rmap->N);
2716   if (A->rmap->n != B->rmap->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat A,Mat B: local dim %D %D",A->rmap->n,B->rmap->n);
2717   if (!A->rmap->N && !A->cmap->N) PetscFunctionReturn(0);
2718   ierr = MatPreallocated(A);CHKERRQ(ierr);
2719 
2720   ierr = PetscLogEventBegin(MAT_MatSolve,A,B,X,0);CHKERRQ(ierr);
2721   if (!A->ops->matsolve) {
2722     ierr = PetscInfo1(A,"Mat type %s using basic MatMatSolve",((PetscObject)A)->type_name);CHKERRQ(ierr);
2723     ierr = MatMatSolve_Basic(A,B,X);CHKERRQ(ierr);
2724   } else {
2725     ierr = (*A->ops->matsolve)(A,B,X);CHKERRQ(ierr);
2726   }
2727   ierr = PetscLogEventEnd(MAT_MatSolve,A,B,X,0);CHKERRQ(ierr);
2728   ierr = PetscObjectStateIncrease((PetscObject)X);CHKERRQ(ierr);
2729   PetscFunctionReturn(0);
2730 }
2731 
2732 
2733 #undef __FUNCT__
2734 #define __FUNCT__ "MatForwardSolve"
2735 /* @
2736    MatForwardSolve - Solves L x = b, given a factored matrix, A = LU, or
2737                             U^T*D^(1/2) x = b, given a factored symmetric matrix, A = U^T*D*U,
2738 
2739    Collective on Mat and Vec
2740 
2741    Input Parameters:
2742 +  mat - the factored matrix
2743 -  b - the right-hand-side vector
2744 
2745    Output Parameter:
2746 .  x - the result vector
2747 
2748    Notes:
2749    MatSolve() should be used for most applications, as it performs
2750    a forward solve followed by a backward solve.
2751 
2752    The vectors b and x cannot be the same,  i.e., one cannot
2753    call MatForwardSolve(A,x,x).
2754 
2755    For matrix in seqsbaij format with block size larger than 1,
2756    the diagonal blocks are not implemented as D = D^(1/2) * D^(1/2) yet.
2757    MatForwardSolve() solves U^T*D y = b, and
2758    MatBackwardSolve() solves U x = y.
2759    Thus they do not provide a symmetric preconditioner.
2760 
2761    Most users should employ the simplified KSP interface for linear solvers
2762    instead of working directly with matrix algebra routines such as this.
2763    See, e.g., KSPCreate().
2764 
2765    Level: developer
2766 
2767    Concepts: matrices^forward solves
2768 
2769 .seealso: MatSolve(), MatBackwardSolve()
2770 @ */
2771 PetscErrorCode PETSCMAT_DLLEXPORT MatForwardSolve(Mat mat,Vec b,Vec x)
2772 {
2773   PetscErrorCode ierr;
2774 
2775   PetscFunctionBegin;
2776   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2777   PetscValidType(mat,1);
2778   PetscValidHeaderSpecific(b,VEC_COOKIE,2);
2779   PetscValidHeaderSpecific(x,VEC_COOKIE,3);
2780   PetscCheckSameComm(mat,1,b,2);
2781   PetscCheckSameComm(mat,1,x,3);
2782   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
2783   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2784   if (!mat->ops->forwardsolve) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2785   if (mat->cmap->N != x->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
2786   if (mat->rmap->N != b->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap->N,b->map->N);
2787   if (mat->rmap->n != b->map->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap->n,b->map->n);
2788   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2789   ierr = PetscLogEventBegin(MAT_ForwardSolve,mat,b,x,0);CHKERRQ(ierr);
2790   ierr = (*mat->ops->forwardsolve)(mat,b,x);CHKERRQ(ierr);
2791   ierr = PetscLogEventEnd(MAT_ForwardSolve,mat,b,x,0);CHKERRQ(ierr);
2792   ierr = PetscObjectStateIncrease((PetscObject)x);CHKERRQ(ierr);
2793   PetscFunctionReturn(0);
2794 }
2795 
2796 #undef __FUNCT__
2797 #define __FUNCT__ "MatBackwardSolve"
2798 /* @
2799    MatBackwardSolve - Solves U x = b, given a factored matrix, A = LU.
2800                              D^(1/2) U x = b, given a factored symmetric matrix, A = U^T*D*U,
2801 
2802    Collective on Mat and Vec
2803 
2804    Input Parameters:
2805 +  mat - the factored matrix
2806 -  b - the right-hand-side vector
2807 
2808    Output Parameter:
2809 .  x - the result vector
2810 
2811    Notes:
2812    MatSolve() should be used for most applications, as it performs
2813    a forward solve followed by a backward solve.
2814 
2815    The vectors b and x cannot be the same.  I.e., one cannot
2816    call MatBackwardSolve(A,x,x).
2817 
2818    For matrix in seqsbaij format with block size larger than 1,
2819    the diagonal blocks are not implemented as D = D^(1/2) * D^(1/2) yet.
2820    MatForwardSolve() solves U^T*D y = b, and
2821    MatBackwardSolve() solves U x = y.
2822    Thus they do not provide a symmetric preconditioner.
2823 
2824    Most users should employ the simplified KSP interface for linear solvers
2825    instead of working directly with matrix algebra routines such as this.
2826    See, e.g., KSPCreate().
2827 
2828    Level: developer
2829 
2830    Concepts: matrices^backward solves
2831 
2832 .seealso: MatSolve(), MatForwardSolve()
2833 @ */
2834 PetscErrorCode PETSCMAT_DLLEXPORT MatBackwardSolve(Mat mat,Vec b,Vec x)
2835 {
2836   PetscErrorCode ierr;
2837 
2838   PetscFunctionBegin;
2839   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2840   PetscValidType(mat,1);
2841   PetscValidHeaderSpecific(b,VEC_COOKIE,2);
2842   PetscValidHeaderSpecific(x,VEC_COOKIE,3);
2843   PetscCheckSameComm(mat,1,b,2);
2844   PetscCheckSameComm(mat,1,x,3);
2845   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
2846   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2847   if (!mat->ops->backwardsolve) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
2848   if (mat->cmap->N != x->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
2849   if (mat->rmap->N != b->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap->N,b->map->N);
2850   if (mat->rmap->n != b->map->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap->n,b->map->n);
2851   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2852 
2853   ierr = PetscLogEventBegin(MAT_BackwardSolve,mat,b,x,0);CHKERRQ(ierr);
2854   ierr = (*mat->ops->backwardsolve)(mat,b,x);CHKERRQ(ierr);
2855   ierr = PetscLogEventEnd(MAT_BackwardSolve,mat,b,x,0);CHKERRQ(ierr);
2856   ierr = PetscObjectStateIncrease((PetscObject)x);CHKERRQ(ierr);
2857   PetscFunctionReturn(0);
2858 }
2859 
2860 #undef __FUNCT__
2861 #define __FUNCT__ "MatSolveAdd"
2862 /*@
2863    MatSolveAdd - Computes x = y + inv(A)*b, given a factored matrix.
2864 
2865    Collective on Mat and Vec
2866 
2867    Input Parameters:
2868 +  mat - the factored matrix
2869 .  b - the right-hand-side vector
2870 -  y - the vector to be added to
2871 
2872    Output Parameter:
2873 .  x - the result vector
2874 
2875    Notes:
2876    The vectors b and x cannot be the same.  I.e., one cannot
2877    call MatSolveAdd(A,x,y,x).
2878 
2879    Most users should employ the simplified KSP interface for linear solvers
2880    instead of working directly with matrix algebra routines such as this.
2881    See, e.g., KSPCreate().
2882 
2883    Level: developer
2884 
2885    Concepts: matrices^triangular solves
2886 
2887 .seealso: MatSolve(), MatSolveTranspose(), MatSolveTransposeAdd()
2888 @*/
2889 PetscErrorCode PETSCMAT_DLLEXPORT MatSolveAdd(Mat mat,Vec b,Vec y,Vec x)
2890 {
2891   PetscScalar    one = 1.0;
2892   Vec            tmp;
2893   PetscErrorCode ierr;
2894 
2895   PetscFunctionBegin;
2896   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2897   PetscValidType(mat,1);
2898   PetscValidHeaderSpecific(y,VEC_COOKIE,2);
2899   PetscValidHeaderSpecific(b,VEC_COOKIE,3);
2900   PetscValidHeaderSpecific(x,VEC_COOKIE,4);
2901   PetscCheckSameComm(mat,1,b,2);
2902   PetscCheckSameComm(mat,1,y,2);
2903   PetscCheckSameComm(mat,1,x,3);
2904   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
2905   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2906   if (mat->cmap->N != x->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
2907   if (mat->rmap->N != b->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap->N,b->map->N);
2908   if (mat->rmap->N != y->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->rmap->N,y->map->N);
2909   if (mat->rmap->n != b->map->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap->n,b->map->n);
2910   if (x->map->n != y->map->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Vec x,Vec y: local dim %D %D",x->map->n,y->map->n);
2911   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2912 
2913   ierr = PetscLogEventBegin(MAT_SolveAdd,mat,b,x,y);CHKERRQ(ierr);
2914   if (mat->ops->solveadd)  {
2915     ierr = (*mat->ops->solveadd)(mat,b,y,x);CHKERRQ(ierr);
2916   } else {
2917     /* do the solve then the add manually */
2918     if (x != y) {
2919       ierr = MatSolve(mat,b,x);CHKERRQ(ierr);
2920       ierr = VecAXPY(x,one,y);CHKERRQ(ierr);
2921     } else {
2922       ierr = VecDuplicate(x,&tmp);CHKERRQ(ierr);
2923       ierr = PetscLogObjectParent(mat,tmp);CHKERRQ(ierr);
2924       ierr = VecCopy(x,tmp);CHKERRQ(ierr);
2925       ierr = MatSolve(mat,b,x);CHKERRQ(ierr);
2926       ierr = VecAXPY(x,one,tmp);CHKERRQ(ierr);
2927       ierr = VecDestroy(tmp);CHKERRQ(ierr);
2928     }
2929   }
2930   ierr = PetscLogEventEnd(MAT_SolveAdd,mat,b,x,y);CHKERRQ(ierr);
2931   ierr = PetscObjectStateIncrease((PetscObject)x);CHKERRQ(ierr);
2932   PetscFunctionReturn(0);
2933 }
2934 
2935 #undef __FUNCT__
2936 #define __FUNCT__ "MatSolveTranspose"
2937 /*@
2938    MatSolveTranspose - Solves A' x = b, given a factored matrix.
2939 
2940    Collective on Mat and Vec
2941 
2942    Input Parameters:
2943 +  mat - the factored matrix
2944 -  b - the right-hand-side vector
2945 
2946    Output Parameter:
2947 .  x - the result vector
2948 
2949    Notes:
2950    The vectors b and x cannot be the same.  I.e., one cannot
2951    call MatSolveTranspose(A,x,x).
2952 
2953    Most users should employ the simplified KSP interface for linear solvers
2954    instead of working directly with matrix algebra routines such as this.
2955    See, e.g., KSPCreate().
2956 
2957    Level: developer
2958 
2959    Concepts: matrices^triangular solves
2960 
2961 .seealso: MatSolve(), MatSolveAdd(), MatSolveTransposeAdd()
2962 @*/
2963 PetscErrorCode PETSCMAT_DLLEXPORT MatSolveTranspose(Mat mat,Vec b,Vec x)
2964 {
2965   PetscErrorCode ierr;
2966 
2967   PetscFunctionBegin;
2968   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
2969   PetscValidType(mat,1);
2970   PetscValidHeaderSpecific(b,VEC_COOKIE,2);
2971   PetscValidHeaderSpecific(x,VEC_COOKIE,3);
2972   PetscCheckSameComm(mat,1,b,2);
2973   PetscCheckSameComm(mat,1,x,3);
2974   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
2975   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
2976   if (!mat->ops->solvetranspose) SETERRQ1(PETSC_ERR_SUP,"Matrix type %s",((PetscObject)mat)->type_name);
2977   if (mat->rmap->N != x->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->rmap->N,x->map->N);
2978   if (mat->cmap->N != b->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->cmap->N,b->map->N);
2979   ierr = MatPreallocated(mat);CHKERRQ(ierr);
2980   ierr = PetscLogEventBegin(MAT_SolveTranspose,mat,b,x,0);CHKERRQ(ierr);
2981   ierr = (*mat->ops->solvetranspose)(mat,b,x);CHKERRQ(ierr);
2982   ierr = PetscLogEventEnd(MAT_SolveTranspose,mat,b,x,0);CHKERRQ(ierr);
2983   ierr = PetscObjectStateIncrease((PetscObject)x);CHKERRQ(ierr);
2984   PetscFunctionReturn(0);
2985 }
2986 
2987 #undef __FUNCT__
2988 #define __FUNCT__ "MatSolveTransposeAdd"
2989 /*@
2990    MatSolveTransposeAdd - Computes x = y + inv(Transpose(A)) b, given a
2991                       factored matrix.
2992 
2993    Collective on Mat and Vec
2994 
2995    Input Parameters:
2996 +  mat - the factored matrix
2997 .  b - the right-hand-side vector
2998 -  y - the vector to be added to
2999 
3000    Output Parameter:
3001 .  x - the result vector
3002 
3003    Notes:
3004    The vectors b and x cannot be the same.  I.e., one cannot
3005    call MatSolveTransposeAdd(A,x,y,x).
3006 
3007    Most users should employ the simplified KSP interface for linear solvers
3008    instead of working directly with matrix algebra routines such as this.
3009    See, e.g., KSPCreate().
3010 
3011    Level: developer
3012 
3013    Concepts: matrices^triangular solves
3014 
3015 .seealso: MatSolve(), MatSolveAdd(), MatSolveTranspose()
3016 @*/
3017 PetscErrorCode PETSCMAT_DLLEXPORT MatSolveTransposeAdd(Mat mat,Vec b,Vec y,Vec x)
3018 {
3019   PetscScalar    one = 1.0;
3020   PetscErrorCode ierr;
3021   Vec            tmp;
3022 
3023   PetscFunctionBegin;
3024   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3025   PetscValidType(mat,1);
3026   PetscValidHeaderSpecific(y,VEC_COOKIE,2);
3027   PetscValidHeaderSpecific(b,VEC_COOKIE,3);
3028   PetscValidHeaderSpecific(x,VEC_COOKIE,4);
3029   PetscCheckSameComm(mat,1,b,2);
3030   PetscCheckSameComm(mat,1,y,3);
3031   PetscCheckSameComm(mat,1,x,4);
3032   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
3033   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
3034   if (mat->rmap->N != x->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->rmap->N,x->map->N);
3035   if (mat->cmap->N != b->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->cmap->N,b->map->N);
3036   if (mat->cmap->N != y->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec y: global dim %D %D",mat->cmap->N,y->map->N);
3037   if (x->map->n != y->map->n)   SETERRQ2(PETSC_ERR_ARG_SIZ,"Vec x,Vec y: local dim %D %D",x->map->n,y->map->n);
3038   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3039 
3040   ierr = PetscLogEventBegin(MAT_SolveTransposeAdd,mat,b,x,y);CHKERRQ(ierr);
3041   if (mat->ops->solvetransposeadd) {
3042     ierr = (*mat->ops->solvetransposeadd)(mat,b,y,x);CHKERRQ(ierr);
3043   } else {
3044     /* do the solve then the add manually */
3045     if (x != y) {
3046       ierr = MatSolveTranspose(mat,b,x);CHKERRQ(ierr);
3047       ierr = VecAXPY(x,one,y);CHKERRQ(ierr);
3048     } else {
3049       ierr = VecDuplicate(x,&tmp);CHKERRQ(ierr);
3050       ierr = PetscLogObjectParent(mat,tmp);CHKERRQ(ierr);
3051       ierr = VecCopy(x,tmp);CHKERRQ(ierr);
3052       ierr = MatSolveTranspose(mat,b,x);CHKERRQ(ierr);
3053       ierr = VecAXPY(x,one,tmp);CHKERRQ(ierr);
3054       ierr = VecDestroy(tmp);CHKERRQ(ierr);
3055     }
3056   }
3057   ierr = PetscLogEventEnd(MAT_SolveTransposeAdd,mat,b,x,y);CHKERRQ(ierr);
3058   ierr = PetscObjectStateIncrease((PetscObject)x);CHKERRQ(ierr);
3059   PetscFunctionReturn(0);
3060 }
3061 /* ----------------------------------------------------------------*/
3062 
3063 #undef __FUNCT__
3064 #define __FUNCT__ "MatRelax"
3065 /*@
3066    MatRelax - Computes relaxation (SOR, Gauss-Seidel) sweeps.
3067 
3068    Collective on Mat and Vec
3069 
3070    Input Parameters:
3071 +  mat - the matrix
3072 .  b - the right hand side
3073 .  omega - the relaxation factor
3074 .  flag - flag indicating the type of SOR (see below)
3075 .  shift -  diagonal shift
3076 .  its - the number of iterations
3077 -  lits - the number of local iterations
3078 
3079    Output Parameters:
3080 .  x - the solution (can contain an initial guess)
3081 
3082    SOR Flags:
3083 .     SOR_FORWARD_SWEEP - forward SOR
3084 .     SOR_BACKWARD_SWEEP - backward SOR
3085 .     SOR_SYMMETRIC_SWEEP - SSOR (symmetric SOR)
3086 .     SOR_LOCAL_FORWARD_SWEEP - local forward SOR
3087 .     SOR_LOCAL_BACKWARD_SWEEP - local forward SOR
3088 .     SOR_LOCAL_SYMMETRIC_SWEEP - local SSOR
3089 .     SOR_APPLY_UPPER, SOR_APPLY_LOWER - applies
3090          upper/lower triangular part of matrix to
3091          vector (with omega)
3092 .     SOR_ZERO_INITIAL_GUESS - zero initial guess
3093 
3094    Notes:
3095    SOR_LOCAL_FORWARD_SWEEP, SOR_LOCAL_BACKWARD_SWEEP, and
3096    SOR_LOCAL_SYMMETRIC_SWEEP perform separate independent smoothings
3097    on each processor.
3098 
3099    Application programmers will not generally use MatRelax() directly,
3100    but instead will employ the KSP/PC interface.
3101 
3102    Notes for Advanced Users:
3103    The flags are implemented as bitwise inclusive or operations.
3104    For example, use (SOR_ZERO_INITIAL_GUESS | SOR_SYMMETRIC_SWEEP)
3105    to specify a zero initial guess for SSOR.
3106 
3107    Most users should employ the simplified KSP interface for linear solvers
3108    instead of working directly with matrix algebra routines such as this.
3109    See, e.g., KSPCreate().
3110 
3111    See also, MatPBRelax(). This routine will automatically call the point block
3112    version if the point version is not available.
3113 
3114    Level: developer
3115 
3116    Concepts: matrices^relaxation
3117    Concepts: matrices^SOR
3118    Concepts: matrices^Gauss-Seidel
3119 
3120 @*/
3121 PetscErrorCode PETSCMAT_DLLEXPORT MatRelax(Mat mat,Vec b,PetscReal omega,MatSORType flag,PetscReal shift,PetscInt its,PetscInt lits,Vec x)
3122 {
3123   PetscErrorCode ierr;
3124 
3125   PetscFunctionBegin;
3126   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3127   PetscValidType(mat,1);
3128   PetscValidHeaderSpecific(b,VEC_COOKIE,2);
3129   PetscValidHeaderSpecific(x,VEC_COOKIE,8);
3130   PetscCheckSameComm(mat,1,b,2);
3131   PetscCheckSameComm(mat,1,x,8);
3132   if (!mat->ops->relax && !mat->ops->pbrelax) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3133   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3134   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3135   if (mat->cmap->N != x->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
3136   if (mat->rmap->N != b->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap->N,b->map->N);
3137   if (mat->rmap->n != b->map->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap->n,b->map->n);
3138   if (its <= 0) SETERRQ1(PETSC_ERR_ARG_WRONG,"Relaxation requires global its %D positive",its);
3139   if (lits <= 0) SETERRQ1(PETSC_ERR_ARG_WRONG,"Relaxation requires local its %D positive",lits);
3140 
3141   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3142   ierr = PetscLogEventBegin(MAT_Relax,mat,b,x,0);CHKERRQ(ierr);
3143   if (mat->ops->relax) {
3144     ierr =(*mat->ops->relax)(mat,b,omega,flag,shift,its,lits,x);CHKERRQ(ierr);
3145   } else {
3146     ierr =(*mat->ops->pbrelax)(mat,b,omega,flag,shift,its,lits,x);CHKERRQ(ierr);
3147   }
3148   ierr = PetscLogEventEnd(MAT_Relax,mat,b,x,0);CHKERRQ(ierr);
3149   ierr = PetscObjectStateIncrease((PetscObject)x);CHKERRQ(ierr);
3150   PetscFunctionReturn(0);
3151 }
3152 
3153 #undef __FUNCT__
3154 #define __FUNCT__ "MatPBRelax"
3155 /*@
3156    MatPBRelax - Computes relaxation (SOR, Gauss-Seidel) sweeps.
3157 
3158    Collective on Mat and Vec
3159 
3160    See MatRelax() for usage
3161 
3162    For multi-component PDEs where the Jacobian is stored in a point block format
3163    (with the PETSc BAIJ matrix formats) the relaxation is done one point block at
3164    a time. That is, the small (for example, 4 by 4) blocks along the diagonal are solved
3165    simultaneously (that is a 4 by 4 linear solve is done) to update all the values at a point.
3166 
3167    Level: developer
3168 
3169 @*/
3170 PetscErrorCode PETSCMAT_DLLEXPORT MatPBRelax(Mat mat,Vec b,PetscReal omega,MatSORType flag,PetscReal shift,PetscInt its,PetscInt lits,Vec x)
3171 {
3172   PetscErrorCode ierr;
3173 
3174   PetscFunctionBegin;
3175   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3176   PetscValidType(mat,1);
3177   PetscValidHeaderSpecific(b,VEC_COOKIE,2);
3178   PetscValidHeaderSpecific(x,VEC_COOKIE,8);
3179   PetscCheckSameComm(mat,1,b,2);
3180   PetscCheckSameComm(mat,1,x,8);
3181   if (!mat->ops->pbrelax) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3182   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3183   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3184   if (mat->cmap->N != x->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec x: global dim %D %D",mat->cmap->N,x->map->N);
3185   if (mat->rmap->N != b->map->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: global dim %D %D",mat->rmap->N,b->map->N);
3186   if (mat->rmap->n != b->map->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"Mat mat,Vec b: local dim %D %D",mat->rmap->n,b->map->n);
3187   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3188 
3189   ierr = PetscLogEventBegin(MAT_Relax,mat,b,x,0);CHKERRQ(ierr);
3190   ierr =(*mat->ops->pbrelax)(mat,b,omega,flag,shift,its,lits,x);CHKERRQ(ierr);
3191   ierr = PetscLogEventEnd(MAT_Relax,mat,b,x,0);CHKERRQ(ierr);
3192   ierr = PetscObjectStateIncrease((PetscObject)x);CHKERRQ(ierr);
3193   PetscFunctionReturn(0);
3194 }
3195 
3196 #undef __FUNCT__
3197 #define __FUNCT__ "MatCopy_Basic"
3198 /*
3199       Default matrix copy routine.
3200 */
3201 PetscErrorCode MatCopy_Basic(Mat A,Mat B,MatStructure str)
3202 {
3203   PetscErrorCode    ierr;
3204   PetscInt          i,rstart,rend,nz;
3205   const PetscInt    *cwork;
3206   const PetscScalar *vwork;
3207 
3208   PetscFunctionBegin;
3209   if (B->assembled) {
3210     ierr = MatZeroEntries(B);CHKERRQ(ierr);
3211   }
3212   ierr = MatGetOwnershipRange(A,&rstart,&rend);CHKERRQ(ierr);
3213   for (i=rstart; i<rend; i++) {
3214     ierr = MatGetRow(A,i,&nz,&cwork,&vwork);CHKERRQ(ierr);
3215     ierr = MatSetValues(B,1,&i,nz,cwork,vwork,INSERT_VALUES);CHKERRQ(ierr);
3216     ierr = MatRestoreRow(A,i,&nz,&cwork,&vwork);CHKERRQ(ierr);
3217   }
3218   ierr = MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
3219   ierr = MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
3220   ierr = PetscObjectStateIncrease((PetscObject)B);CHKERRQ(ierr);
3221   PetscFunctionReturn(0);
3222 }
3223 
3224 #undef __FUNCT__
3225 #define __FUNCT__ "MatCopy"
3226 /*@
3227    MatCopy - Copys a matrix to another matrix.
3228 
3229    Collective on Mat
3230 
3231    Input Parameters:
3232 +  A - the matrix
3233 -  str - SAME_NONZERO_PATTERN or DIFFERENT_NONZERO_PATTERN
3234 
3235    Output Parameter:
3236 .  B - where the copy is put
3237 
3238    Notes:
3239    If you use SAME_NONZERO_PATTERN then the two matrices had better have the
3240    same nonzero pattern or the routine will crash.
3241 
3242    MatCopy() copies the matrix entries of a matrix to another existing
3243    matrix (after first zeroing the second matrix).  A related routine is
3244    MatConvert(), which first creates a new matrix and then copies the data.
3245 
3246    Level: intermediate
3247 
3248    Concepts: matrices^copying
3249 
3250 .seealso: MatConvert(), MatDuplicate()
3251 
3252 @*/
3253 PetscErrorCode PETSCMAT_DLLEXPORT MatCopy(Mat A,Mat B,MatStructure str)
3254 {
3255   PetscErrorCode ierr;
3256   PetscInt       i;
3257 
3258   PetscFunctionBegin;
3259   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
3260   PetscValidHeaderSpecific(B,MAT_COOKIE,2);
3261   PetscValidType(A,1);
3262   PetscValidType(B,2);
3263   PetscCheckSameComm(A,1,B,2);
3264   ierr = MatPreallocated(B);CHKERRQ(ierr);
3265   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3266   if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3267   if (A->rmap->N != B->rmap->N || A->cmap->N != B->cmap->N) SETERRQ4(PETSC_ERR_ARG_SIZ,"Mat A,Mat B: global dim (%D,%D) (%D,%D)",A->rmap->N,B->rmap->N,A->cmap->N,B->cmap->N);
3268   ierr = MatPreallocated(A);CHKERRQ(ierr);
3269 
3270   ierr = PetscLogEventBegin(MAT_Copy,A,B,0,0);CHKERRQ(ierr);
3271   if (A->ops->copy) {
3272     ierr = (*A->ops->copy)(A,B,str);CHKERRQ(ierr);
3273   } else { /* generic conversion */
3274     ierr = MatCopy_Basic(A,B,str);CHKERRQ(ierr);
3275   }
3276   if (A->mapping) {
3277     if (B->mapping) {ierr = ISLocalToGlobalMappingDestroy(B->mapping);CHKERRQ(ierr);B->mapping = 0;}
3278     ierr = MatSetLocalToGlobalMapping(B,A->mapping);CHKERRQ(ierr);
3279   }
3280   if (A->bmapping) {
3281     if (B->bmapping) {ierr = ISLocalToGlobalMappingDestroy(B->bmapping);CHKERRQ(ierr);B->bmapping = 0;}
3282     ierr = MatSetLocalToGlobalMappingBlock(B,A->mapping);CHKERRQ(ierr);
3283   }
3284 
3285   B->stencil.dim = A->stencil.dim;
3286   B->stencil.noc = A->stencil.noc;
3287   for (i=0; i<=A->stencil.dim; i++) {
3288     B->stencil.dims[i]   = A->stencil.dims[i];
3289     B->stencil.starts[i] = A->stencil.starts[i];
3290   }
3291 
3292   ierr = PetscLogEventEnd(MAT_Copy,A,B,0,0);CHKERRQ(ierr);
3293   ierr = PetscObjectStateIncrease((PetscObject)B);CHKERRQ(ierr);
3294   PetscFunctionReturn(0);
3295 }
3296 
3297 #undef __FUNCT__
3298 #define __FUNCT__ "MatConvert"
3299 /*@C
3300    MatConvert - Converts a matrix to another matrix, either of the same
3301    or different type.
3302 
3303    Collective on Mat
3304 
3305    Input Parameters:
3306 +  mat - the matrix
3307 .  newtype - new matrix type.  Use MATSAME to create a new matrix of the
3308    same type as the original matrix.
3309 -  reuse - denotes if the destination matrix is to be created or reused.  Currently
3310    MAT_REUSE_MATRIX is only supported for inplace conversion, otherwise use
3311    MAT_INITIAL_MATRIX.
3312 
3313    Output Parameter:
3314 .  M - pointer to place new matrix
3315 
3316    Notes:
3317    MatConvert() first creates a new matrix and then copies the data from
3318    the first matrix.  A related routine is MatCopy(), which copies the matrix
3319    entries of one matrix to another already existing matrix context.
3320 
3321    Cannot be used to convert a sequential matrix to parallel or parallel to sequential,
3322    the MPI communicator of the generated matrix is always the same as the communicator
3323    of the input matrix.
3324 
3325    Level: intermediate
3326 
3327    Concepts: matrices^converting between storage formats
3328 
3329 .seealso: MatCopy(), MatDuplicate()
3330 @*/
3331 PetscErrorCode PETSCMAT_DLLEXPORT MatConvert(Mat mat, const MatType newtype,MatReuse reuse,Mat *M)
3332 {
3333   PetscErrorCode         ierr;
3334   PetscTruth             sametype,issame,flg;
3335   char                   convname[256],mtype[256];
3336   Mat                    B;
3337 
3338   PetscFunctionBegin;
3339   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3340   PetscValidType(mat,1);
3341   PetscValidPointer(M,3);
3342   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3343   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3344   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3345 
3346   ierr = PetscOptionsGetString(((PetscObject)mat)->prefix,"-matconvert_type",mtype,256,&flg);CHKERRQ(ierr);
3347   if (flg) {
3348     newtype = mtype;
3349   }
3350   ierr = PetscTypeCompare((PetscObject)mat,newtype,&sametype);CHKERRQ(ierr);
3351   ierr = PetscStrcmp(newtype,"same",&issame);CHKERRQ(ierr);
3352   if ((reuse == MAT_REUSE_MATRIX) && (mat != *M)) {
3353     SETERRQ(PETSC_ERR_SUP,"MAT_REUSE_MATRIX only supported for in-place conversion currently");
3354   }
3355 
3356   if ((reuse == MAT_REUSE_MATRIX) && (issame || sametype)) PetscFunctionReturn(0);
3357 
3358   if ((sametype || issame) && (reuse==MAT_INITIAL_MATRIX) && mat->ops->duplicate) {
3359     ierr = (*mat->ops->duplicate)(mat,MAT_COPY_VALUES,M);CHKERRQ(ierr);
3360   } else {
3361     PetscErrorCode (*conv)(Mat, const MatType,MatReuse,Mat*)=PETSC_NULL;
3362     const char     *prefix[3] = {"seq","mpi",""};
3363     PetscInt       i;
3364     /*
3365        Order of precedence:
3366        1) See if a specialized converter is known to the current matrix.
3367        2) See if a specialized converter is known to the desired matrix class.
3368        3) See if a good general converter is registered for the desired class
3369           (as of 6/27/03 only MATMPIADJ falls into this category).
3370        4) See if a good general converter is known for the current matrix.
3371        5) Use a really basic converter.
3372     */
3373 
3374     /* 1) See if a specialized converter is known to the current matrix and the desired class */
3375     for (i=0; i<3; i++) {
3376       ierr = PetscStrcpy(convname,"MatConvert_");CHKERRQ(ierr);
3377       ierr = PetscStrcat(convname,((PetscObject)mat)->type_name);CHKERRQ(ierr);
3378       ierr = PetscStrcat(convname,"_");CHKERRQ(ierr);
3379       ierr = PetscStrcat(convname,prefix[i]);CHKERRQ(ierr);
3380       ierr = PetscStrcat(convname,newtype);CHKERRQ(ierr);
3381       ierr = PetscStrcat(convname,"_C");CHKERRQ(ierr);
3382       ierr = PetscObjectQueryFunction((PetscObject)mat,convname,(void (**)(void))&conv);CHKERRQ(ierr);
3383       if (conv) goto foundconv;
3384     }
3385 
3386     /* 2)  See if a specialized converter is known to the desired matrix class. */
3387     ierr = MatCreate(((PetscObject)mat)->comm,&B);CHKERRQ(ierr);
3388     ierr = MatSetSizes(B,mat->rmap->n,mat->cmap->n,mat->rmap->N,mat->cmap->N);CHKERRQ(ierr);
3389     ierr = MatSetType(B,newtype);CHKERRQ(ierr);
3390     for (i=0; i<3; i++) {
3391       ierr = PetscStrcpy(convname,"MatConvert_");CHKERRQ(ierr);
3392       ierr = PetscStrcat(convname,((PetscObject)mat)->type_name);CHKERRQ(ierr);
3393       ierr = PetscStrcat(convname,"_");CHKERRQ(ierr);
3394       ierr = PetscStrcat(convname,prefix[i]);CHKERRQ(ierr);
3395       ierr = PetscStrcat(convname,newtype);CHKERRQ(ierr);
3396       ierr = PetscStrcat(convname,"_C");CHKERRQ(ierr);
3397       ierr = PetscObjectQueryFunction((PetscObject)B,convname,(void (**)(void))&conv);CHKERRQ(ierr);
3398       if (conv) {
3399         ierr = MatDestroy(B);CHKERRQ(ierr);
3400         goto foundconv;
3401       }
3402     }
3403 
3404     /* 3) See if a good general converter is registered for the desired class */
3405     conv = B->ops->convertfrom;
3406     ierr = MatDestroy(B);CHKERRQ(ierr);
3407     if (conv) goto foundconv;
3408 
3409     /* 4) See if a good general converter is known for the current matrix */
3410     if (mat->ops->convert) {
3411       conv = mat->ops->convert;
3412     }
3413     if (conv) goto foundconv;
3414 
3415     /* 5) Use a really basic converter. */
3416     conv = MatConvert_Basic;
3417 
3418     foundconv:
3419     ierr = PetscLogEventBegin(MAT_Convert,mat,0,0,0);CHKERRQ(ierr);
3420     ierr = (*conv)(mat,newtype,reuse,M);CHKERRQ(ierr);
3421     ierr = PetscLogEventEnd(MAT_Convert,mat,0,0,0);CHKERRQ(ierr);
3422   }
3423   ierr = PetscObjectStateIncrease((PetscObject)*M);CHKERRQ(ierr);
3424   PetscFunctionReturn(0);
3425 }
3426 
3427 #undef __FUNCT__
3428 #define __FUNCT__ "MatFactorGetSolverPackage"
3429 /*@C
3430    MatFactorGetSolverPackage - Returns name of the package providing the factorization routines
3431 
3432    Not Collective
3433 
3434    Input Parameter:
3435 .  mat - the matrix, must be a factored matrix
3436 
3437    Output Parameter:
3438 .   type - the string name of the package (do not free this string)
3439 
3440    Notes:
3441       In Fortran you pass in a empty string and the package name will be copied into it.
3442     (Make sure the string is long enough)
3443 
3444    Level: intermediate
3445 
3446 .seealso: MatCopy(), MatDuplicate(), MatGetFactorAvailable(), MatGetFactor()
3447 @*/
3448 PetscErrorCode PETSCMAT_DLLEXPORT MatFactorGetSolverPackage(Mat mat, const MatSolverPackage *type)
3449 {
3450   PetscErrorCode         ierr;
3451   PetscErrorCode         (*conv)(Mat,const MatSolverPackage*);
3452 
3453   PetscFunctionBegin;
3454   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3455   PetscValidType(mat,1);
3456   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Only for factored matrix");
3457   ierr = PetscObjectQueryFunction((PetscObject)mat,"MatFactorGetSolverPackage_C",(void (**)(void))&conv);CHKERRQ(ierr);
3458   if (!conv) {
3459     *type = MAT_SOLVER_PETSC;
3460   } else {
3461     ierr = (*conv)(mat,type);CHKERRQ(ierr);
3462   }
3463   PetscFunctionReturn(0);
3464 }
3465 
3466 #undef __FUNCT__
3467 #define __FUNCT__ "MatGetFactor"
3468 /*@C
3469    MatGetFactor - Returns a matrix suitable to calls to MatXXFactorSymbolic()
3470 
3471    Collective on Mat
3472 
3473    Input Parameters:
3474 +  mat - the matrix
3475 .  type - name of solver type, for example, spooles, superlu, plapack, petsc (to use PETSc's default)
3476 -  ftype - factor type, MAT_FACTOR_LU, MAT_FACTOR_CHOLESKY, MAT_FACTOR_ICC, MAT_FACTOR_ILU,
3477 
3478    Output Parameters:
3479 .  f - the factor matrix used with MatXXFactorSymbolic() calls
3480 
3481    Notes:
3482       Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
3483      such as pastix, superlu, mumps, spooles etc.
3484 
3485       PETSc must have been config/configure.py to use the external solver, using the option --download-package
3486 
3487    Level: intermediate
3488 
3489 .seealso: MatCopy(), MatDuplicate(), MatGetFactorAvailable()
3490 @*/
3491 PetscErrorCode PETSCMAT_DLLEXPORT MatGetFactor(Mat mat, const MatSolverPackage type,MatFactorType ftype,Mat *f)
3492 {
3493   PetscErrorCode         ierr;
3494   char                   convname[256];
3495   PetscErrorCode         (*conv)(Mat,MatFactorType,Mat*);
3496 
3497   PetscFunctionBegin;
3498   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3499   PetscValidType(mat,1);
3500 
3501   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3502   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3503 
3504   ierr = PetscStrcpy(convname,"MatGetFactor_");CHKERRQ(ierr);
3505   ierr = PetscStrcat(convname,((PetscObject)mat)->type_name);CHKERRQ(ierr);
3506   ierr = PetscStrcat(convname,"_");CHKERRQ(ierr);
3507   ierr = PetscStrcat(convname,type);CHKERRQ(ierr);
3508   ierr = PetscStrcat(convname,"_C");CHKERRQ(ierr);
3509   ierr = PetscObjectQueryFunction((PetscObject)mat,convname,(void (**)(void))&conv);CHKERRQ(ierr);
3510   if (!conv) {
3511     PetscTruth flag;
3512     ierr = PetscStrcasecmp(MAT_SOLVER_PETSC,type,&flag);CHKERRQ(ierr);
3513     if (flag) {
3514       SETERRQ1(PETSC_ERR_SUP,"Matrix format %s does not have a built-in PETSc direct solver",((PetscObject)mat)->type_name);
3515     } else {
3516       SETERRQ3(PETSC_ERR_SUP,"Matrix format %s does not have a solver %s. Perhaps you must config/configure.py with --download-%s",((PetscObject)mat)->type_name,type,type);
3517     }
3518   }
3519   ierr = (*conv)(mat,ftype,f);CHKERRQ(ierr);
3520   PetscFunctionReturn(0);
3521 }
3522 
3523 #undef __FUNCT__
3524 #define __FUNCT__ "MatGetFactorAvailable"
3525 /*@C
3526    MatGetFactorAvailable - Returns a a flag if matrix supports particular package and factor type
3527 
3528    Collective on Mat
3529 
3530    Input Parameters:
3531 +  mat - the matrix
3532 .  type - name of solver type, for example, spooles, superlu, plapack, petsc (to use PETSc's default)
3533 -  ftype - factor type, MAT_FACTOR_LU, MAT_FACTOR_CHOLESKY, MAT_FACTOR_ICC, MAT_FACTOR_ILU,
3534 
3535    Output Parameter:
3536 .    flg - PETSC_TRUE if the factorization is available
3537 
3538    Notes:
3539       Some PETSc matrix formats have alternative solvers available that are contained in alternative packages
3540      such as pastix, superlu, mumps, spooles etc.
3541 
3542       PETSc must have been config/configure.py to use the external solver, using the option --download-package
3543 
3544    Level: intermediate
3545 
3546 .seealso: MatCopy(), MatDuplicate(), MatGetFactor()
3547 @*/
3548 PetscErrorCode PETSCMAT_DLLEXPORT MatGetFactorAvailable(Mat mat, const MatSolverPackage type,MatFactorType ftype,PetscTruth *flg)
3549 {
3550   PetscErrorCode         ierr;
3551   char                   convname[256];
3552   PetscErrorCode         (*conv)(Mat,MatFactorType,PetscTruth*);
3553 
3554   PetscFunctionBegin;
3555   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3556   PetscValidType(mat,1);
3557 
3558   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3559   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3560 
3561   ierr = PetscStrcpy(convname,"MatGetFactorAvailable_");CHKERRQ(ierr);
3562   ierr = PetscStrcat(convname,((PetscObject)mat)->type_name);CHKERRQ(ierr);
3563   ierr = PetscStrcat(convname,"_");CHKERRQ(ierr);
3564   ierr = PetscStrcat(convname,type);CHKERRQ(ierr);
3565   ierr = PetscStrcat(convname,"_C");CHKERRQ(ierr);
3566   ierr = PetscObjectQueryFunction((PetscObject)mat,convname,(void (**)(void))&conv);CHKERRQ(ierr);
3567   if (!conv) {
3568     *flg = PETSC_FALSE;
3569   } else {
3570     ierr = (*conv)(mat,ftype,flg);CHKERRQ(ierr);
3571   }
3572   PetscFunctionReturn(0);
3573 }
3574 
3575 
3576 #undef __FUNCT__
3577 #define __FUNCT__ "MatDuplicate"
3578 /*@
3579    MatDuplicate - Duplicates a matrix including the non-zero structure.
3580 
3581    Collective on Mat
3582 
3583    Input Parameters:
3584 +  mat - the matrix
3585 -  op - either MAT_DO_NOT_COPY_VALUES or MAT_COPY_VALUES, cause it to copy nonzero
3586         values as well or not
3587 
3588    Output Parameter:
3589 .  M - pointer to place new matrix
3590 
3591    Level: intermediate
3592 
3593    Concepts: matrices^duplicating
3594 
3595 .seealso: MatCopy(), MatConvert()
3596 @*/
3597 PetscErrorCode PETSCMAT_DLLEXPORT MatDuplicate(Mat mat,MatDuplicateOption op,Mat *M)
3598 {
3599   PetscErrorCode ierr;
3600   Mat            B;
3601   PetscInt       i;
3602 
3603   PetscFunctionBegin;
3604   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3605   PetscValidType(mat,1);
3606   PetscValidPointer(M,3);
3607   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3608   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3609   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3610 
3611   *M  = 0;
3612   if (!mat->ops->duplicate) {
3613     SETERRQ(PETSC_ERR_SUP,"Not written for this matrix type");
3614   }
3615   ierr = PetscLogEventBegin(MAT_Convert,mat,0,0,0);CHKERRQ(ierr);
3616   ierr = (*mat->ops->duplicate)(mat,op,M);CHKERRQ(ierr);
3617   B = *M;
3618   if (mat->mapping) {
3619     ierr = MatSetLocalToGlobalMapping(B,mat->mapping);CHKERRQ(ierr);
3620   }
3621   if (mat->bmapping) {
3622     ierr = MatSetLocalToGlobalMappingBlock(B,mat->bmapping);CHKERRQ(ierr);
3623   }
3624   ierr = PetscMapCopy(((PetscObject)mat)->comm,mat->rmap,B->rmap);CHKERRQ(ierr);
3625   ierr = PetscMapCopy(((PetscObject)mat)->comm,mat->cmap,B->cmap);CHKERRQ(ierr);
3626 
3627   B->stencil.dim = mat->stencil.dim;
3628   B->stencil.noc = mat->stencil.noc;
3629   for (i=0; i<=mat->stencil.dim; i++) {
3630     B->stencil.dims[i]   = mat->stencil.dims[i];
3631     B->stencil.starts[i] = mat->stencil.starts[i];
3632   }
3633 
3634   ierr = PetscLogEventEnd(MAT_Convert,mat,0,0,0);CHKERRQ(ierr);
3635   ierr = PetscObjectStateIncrease((PetscObject)B);CHKERRQ(ierr);
3636   PetscFunctionReturn(0);
3637 }
3638 
3639 #undef __FUNCT__
3640 #define __FUNCT__ "MatGetDiagonal"
3641 /*@
3642    MatGetDiagonal - Gets the diagonal of a matrix.
3643 
3644    Collective on Mat and Vec
3645 
3646    Input Parameters:
3647 +  mat - the matrix
3648 -  v - the vector for storing the diagonal
3649 
3650    Output Parameter:
3651 .  v - the diagonal of the matrix
3652 
3653    Level: intermediate
3654 
3655    Concepts: matrices^accessing diagonals
3656 
3657 .seealso: MatGetRow(), MatGetSubMatrices(), MatGetSubmatrix(), MatGetRowMaxAbs()
3658 @*/
3659 PetscErrorCode PETSCMAT_DLLEXPORT MatGetDiagonal(Mat mat,Vec v)
3660 {
3661   PetscErrorCode ierr;
3662 
3663   PetscFunctionBegin;
3664   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3665   PetscValidType(mat,1);
3666   PetscValidHeaderSpecific(v,VEC_COOKIE,2);
3667   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3668   if (!mat->ops->getdiagonal) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3669   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3670 
3671   ierr = (*mat->ops->getdiagonal)(mat,v);CHKERRQ(ierr);
3672   ierr = PetscObjectStateIncrease((PetscObject)v);CHKERRQ(ierr);
3673   PetscFunctionReturn(0);
3674 }
3675 
3676 #undef __FUNCT__
3677 #define __FUNCT__ "MatGetRowMin"
3678 /*@
3679    MatGetRowMin - Gets the minimum value (of the real part) of each
3680         row of the matrix
3681 
3682    Collective on Mat and Vec
3683 
3684    Input Parameters:
3685 .  mat - the matrix
3686 
3687    Output Parameter:
3688 +  v - the vector for storing the maximums
3689 -  idx - the indices of the column found for each row (optional)
3690 
3691    Level: intermediate
3692 
3693    Notes: The result of this call are the same as if one converted the matrix to dense format
3694       and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
3695 
3696     This code is only implemented for a couple of matrix formats.
3697 
3698    Concepts: matrices^getting row maximums
3699 
3700 .seealso: MatGetDiagonal(), MatGetSubMatrices(), MatGetSubmatrix(), MatGetRowMaxAbs(),
3701           MatGetRowMax()
3702 @*/
3703 PetscErrorCode PETSCMAT_DLLEXPORT MatGetRowMin(Mat mat,Vec v,PetscInt idx[])
3704 {
3705   PetscErrorCode ierr;
3706 
3707   PetscFunctionBegin;
3708   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3709   PetscValidType(mat,1);
3710   PetscValidHeaderSpecific(v,VEC_COOKIE,2);
3711   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3712   if (!mat->ops->getrowmax) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3713   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3714 
3715   ierr = (*mat->ops->getrowmin)(mat,v,idx);CHKERRQ(ierr);
3716   ierr = PetscObjectStateIncrease((PetscObject)v);CHKERRQ(ierr);
3717   PetscFunctionReturn(0);
3718 }
3719 
3720 #undef __FUNCT__
3721 #define __FUNCT__ "MatGetRowMinAbs"
3722 /*@
3723    MatGetRowMinAbs - Gets the minimum value (in absolute value) of each
3724         row of the matrix
3725 
3726    Collective on Mat and Vec
3727 
3728    Input Parameters:
3729 .  mat - the matrix
3730 
3731    Output Parameter:
3732 +  v - the vector for storing the minimums
3733 -  idx - the indices of the column found for each row (optional)
3734 
3735    Level: intermediate
3736 
3737    Notes: if a row is completely empty or has only 0.0 values then the idx[] value for that
3738     row is 0 (the first column).
3739 
3740     This code is only implemented for a couple of matrix formats.
3741 
3742    Concepts: matrices^getting row maximums
3743 
3744 .seealso: MatGetDiagonal(), MatGetSubMatrices(), MatGetSubmatrix(), MatGetRowMax(), MatGetRowMaxAbs(), MatGetRowMin()
3745 @*/
3746 PetscErrorCode PETSCMAT_DLLEXPORT MatGetRowMinAbs(Mat mat,Vec v,PetscInt idx[])
3747 {
3748   PetscErrorCode ierr;
3749 
3750   PetscFunctionBegin;
3751   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3752   PetscValidType(mat,1);
3753   PetscValidHeaderSpecific(v,VEC_COOKIE,2);
3754   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3755   if (!mat->ops->getrowminabs) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3756   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3757   if (idx) {ierr = PetscMemzero(idx,mat->rmap->n*sizeof(PetscInt));CHKERRQ(ierr);}
3758 
3759   ierr = (*mat->ops->getrowminabs)(mat,v,idx);CHKERRQ(ierr);
3760   ierr = PetscObjectStateIncrease((PetscObject)v);CHKERRQ(ierr);
3761   PetscFunctionReturn(0);
3762 }
3763 
3764 #undef __FUNCT__
3765 #define __FUNCT__ "MatGetRowMax"
3766 /*@
3767    MatGetRowMax - Gets the maximum value (of the real part) of each
3768         row of the matrix
3769 
3770    Collective on Mat and Vec
3771 
3772    Input Parameters:
3773 .  mat - the matrix
3774 
3775    Output Parameter:
3776 +  v - the vector for storing the maximums
3777 -  idx - the indices of the column found for each row (optional)
3778 
3779    Level: intermediate
3780 
3781    Notes: The result of this call are the same as if one converted the matrix to dense format
3782       and found the minimum value in each row (i.e. the implicit zeros are counted as zeros).
3783 
3784     This code is only implemented for a couple of matrix formats.
3785 
3786    Concepts: matrices^getting row maximums
3787 
3788 .seealso: MatGetDiagonal(), MatGetSubMatrices(), MatGetSubmatrix(), MatGetRowMaxAbs(), MatGetRowMin()
3789 @*/
3790 PetscErrorCode PETSCMAT_DLLEXPORT MatGetRowMax(Mat mat,Vec v,PetscInt idx[])
3791 {
3792   PetscErrorCode ierr;
3793 
3794   PetscFunctionBegin;
3795   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3796   PetscValidType(mat,1);
3797   PetscValidHeaderSpecific(v,VEC_COOKIE,2);
3798   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3799   if (!mat->ops->getrowmax) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3800   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3801 
3802   ierr = (*mat->ops->getrowmax)(mat,v,idx);CHKERRQ(ierr);
3803   ierr = PetscObjectStateIncrease((PetscObject)v);CHKERRQ(ierr);
3804   PetscFunctionReturn(0);
3805 }
3806 
3807 #undef __FUNCT__
3808 #define __FUNCT__ "MatGetRowMaxAbs"
3809 /*@
3810    MatGetRowMaxAbs - Gets the maximum value (in absolute value) of each
3811         row of the matrix
3812 
3813    Collective on Mat and Vec
3814 
3815    Input Parameters:
3816 .  mat - the matrix
3817 
3818    Output Parameter:
3819 +  v - the vector for storing the maximums
3820 -  idx - the indices of the column found for each row (optional)
3821 
3822    Level: intermediate
3823 
3824    Notes: if a row is completely empty or has only 0.0 values then the idx[] value for that
3825     row is 0 (the first column).
3826 
3827     This code is only implemented for a couple of matrix formats.
3828 
3829    Concepts: matrices^getting row maximums
3830 
3831 .seealso: MatGetDiagonal(), MatGetSubMatrices(), MatGetSubmatrix(), MatGetRowMax(), MatGetRowMin()
3832 @*/
3833 PetscErrorCode PETSCMAT_DLLEXPORT MatGetRowMaxAbs(Mat mat,Vec v,PetscInt idx[])
3834 {
3835   PetscErrorCode ierr;
3836 
3837   PetscFunctionBegin;
3838   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3839   PetscValidType(mat,1);
3840   PetscValidHeaderSpecific(v,VEC_COOKIE,2);
3841   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3842   if (!mat->ops->getrowmaxabs) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3843   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3844   if (idx) {ierr = PetscMemzero(idx,mat->rmap->n*sizeof(PetscInt));CHKERRQ(ierr);}
3845 
3846   ierr = (*mat->ops->getrowmaxabs)(mat,v,idx);CHKERRQ(ierr);
3847   ierr = PetscObjectStateIncrease((PetscObject)v);CHKERRQ(ierr);
3848   PetscFunctionReturn(0);
3849 }
3850 
3851 #undef __FUNCT__
3852 #define __FUNCT__ "MatGetRowSum"
3853 /*@
3854    MatGetRowSum - Gets the sum of each row of the matrix
3855 
3856    Collective on Mat and Vec
3857 
3858    Input Parameters:
3859 .  mat - the matrix
3860 
3861    Output Parameter:
3862 .  v - the vector for storing the maximums
3863 
3864    Level: intermediate
3865 
3866    Notes: This code is slow since it is not currently specialized for different formats
3867 
3868    Concepts: matrices^getting row sums
3869 
3870 .seealso: MatGetDiagonal(), MatGetSubMatrices(), MatGetSubmatrix(), MatGetRowMax(), MatGetRowMin()
3871 @*/
3872 PetscErrorCode PETSCMAT_DLLEXPORT MatGetRowSum(Mat mat, Vec v)
3873 {
3874   PetscInt       start, end, row;
3875   PetscScalar   *array;
3876   PetscErrorCode ierr;
3877 
3878   PetscFunctionBegin;
3879   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3880   PetscValidType(mat,1);
3881   PetscValidHeaderSpecific(v,VEC_COOKIE,2);
3882   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3883   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3884   ierr = MatGetOwnershipRange(mat, &start, &end);CHKERRQ(ierr);
3885   ierr = VecGetArray(v, &array);CHKERRQ(ierr);
3886   for(row = start; row < end; ++row) {
3887     PetscInt           ncols, col;
3888     const PetscInt    *cols;
3889     const PetscScalar *vals;
3890 
3891     array[row - start] = 0.0;
3892     ierr = MatGetRow(mat, row, &ncols, &cols, &vals);CHKERRQ(ierr);
3893     for(col = 0; col < ncols; col++) {
3894       array[row - start] += vals[col];
3895     }
3896     ierr = MatRestoreRow(mat, row, &ncols, &cols, &vals);CHKERRQ(ierr);
3897   }
3898   ierr = VecRestoreArray(v, &array);CHKERRQ(ierr);
3899   ierr = PetscObjectStateIncrease((PetscObject) v);CHKERRQ(ierr);
3900   PetscFunctionReturn(0);
3901 }
3902 
3903 #undef __FUNCT__
3904 #define __FUNCT__ "MatTranspose"
3905 /*@
3906    MatTranspose - Computes an in-place or out-of-place transpose of a matrix.
3907 
3908    Collective on Mat
3909 
3910    Input Parameter:
3911 +  mat - the matrix to transpose
3912 -  reuse - store the transpose matrix in the provided B
3913 
3914    Output Parameters:
3915 .  B - the transpose
3916 
3917    Notes:
3918      If you  pass in &mat for B the transpose will be done in place
3919 
3920    Level: intermediate
3921 
3922    Concepts: matrices^transposing
3923 
3924 .seealso: MatMultTranspose(), MatMultTransposeAdd(), MatIsTranspose()
3925 @*/
3926 PetscErrorCode PETSCMAT_DLLEXPORT MatTranspose(Mat mat,MatReuse reuse,Mat *B)
3927 {
3928   PetscErrorCode ierr;
3929 
3930   PetscFunctionBegin;
3931   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
3932   PetscValidType(mat,1);
3933   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3934   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3935   if (!mat->ops->transpose) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
3936   ierr = MatPreallocated(mat);CHKERRQ(ierr);
3937 
3938   ierr = PetscLogEventBegin(MAT_Transpose,mat,0,0,0);CHKERRQ(ierr);
3939   ierr = (*mat->ops->transpose)(mat,reuse,B);CHKERRQ(ierr);
3940   ierr = PetscLogEventEnd(MAT_Transpose,mat,0,0,0);CHKERRQ(ierr);
3941   if (B) {ierr = PetscObjectStateIncrease((PetscObject)*B);CHKERRQ(ierr);}
3942   PetscFunctionReturn(0);
3943 }
3944 
3945 #undef __FUNCT__
3946 #define __FUNCT__ "MatIsTranspose"
3947 /*@
3948    MatIsTranspose - Test whether a matrix is another one's transpose,
3949         or its own, in which case it tests symmetry.
3950 
3951    Collective on Mat
3952 
3953    Input Parameter:
3954 +  A - the matrix to test
3955 -  B - the matrix to test against, this can equal the first parameter
3956 
3957    Output Parameters:
3958 .  flg - the result
3959 
3960    Notes:
3961    Only available for SeqAIJ/MPIAIJ matrices. The sequential algorithm
3962    has a running time of the order of the number of nonzeros; the parallel
3963    test involves parallel copies of the block-offdiagonal parts of the matrix.
3964 
3965    Level: intermediate
3966 
3967    Concepts: matrices^transposing, matrix^symmetry
3968 
3969 .seealso: MatTranspose(), MatIsSymmetric(), MatIsHermitian()
3970 @*/
3971 PetscErrorCode PETSCMAT_DLLEXPORT MatIsTranspose(Mat A,Mat B,PetscReal tol,PetscTruth *flg)
3972 {
3973   PetscErrorCode ierr,(*f)(Mat,Mat,PetscReal,PetscTruth*),(*g)(Mat,Mat,PetscReal,PetscTruth*);
3974 
3975   PetscFunctionBegin;
3976   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
3977   PetscValidHeaderSpecific(B,MAT_COOKIE,2);
3978   PetscValidPointer(flg,3);
3979   ierr = PetscObjectQueryFunction((PetscObject)A,"MatIsTranspose_C",(void (**)(void))&f);CHKERRQ(ierr);
3980   ierr = PetscObjectQueryFunction((PetscObject)B,"MatIsTranspose_C",(void (**)(void))&g);CHKERRQ(ierr);
3981   if (f && g) {
3982     if (f==g) {
3983       ierr = (*f)(A,B,tol,flg);CHKERRQ(ierr);
3984     } else {
3985       SETERRQ(PETSC_ERR_ARG_NOTSAMETYPE,"Matrices do not have the same comparator for symmetry test");
3986     }
3987   }
3988   PetscFunctionReturn(0);
3989 }
3990 
3991 #undef __FUNCT__
3992 #define __FUNCT__ "MatIsHermitianTranspose"
3993 /*@
3994    MatIsHermitianTranspose - Test whether a matrix is another one's Hermitian transpose,
3995 
3996    Collective on Mat
3997 
3998    Input Parameter:
3999 +  A - the matrix to test
4000 -  B - the matrix to test against, this can equal the first parameter
4001 
4002    Output Parameters:
4003 .  flg - the result
4004 
4005    Notes:
4006    Only available for SeqAIJ/MPIAIJ matrices. The sequential algorithm
4007    has a running time of the order of the number of nonzeros; the parallel
4008    test involves parallel copies of the block-offdiagonal parts of the matrix.
4009 
4010    Level: intermediate
4011 
4012    Concepts: matrices^transposing, matrix^symmetry
4013 
4014 .seealso: MatTranspose(), MatIsSymmetric(), MatIsHermitian(), MatIsTranspose()
4015 @*/
4016 PetscErrorCode PETSCMAT_DLLEXPORT MatIsHermitianTranspose(Mat A,Mat B,PetscReal tol,PetscTruth *flg)
4017 {
4018   PetscErrorCode ierr,(*f)(Mat,Mat,PetscReal,PetscTruth*),(*g)(Mat,Mat,PetscReal,PetscTruth*);
4019 
4020   PetscFunctionBegin;
4021   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
4022   PetscValidHeaderSpecific(B,MAT_COOKIE,2);
4023   PetscValidPointer(flg,3);
4024   ierr = PetscObjectQueryFunction((PetscObject)A,"MatIsHermitianTranspose_C",(void (**)(void))&f);CHKERRQ(ierr);
4025   ierr = PetscObjectQueryFunction((PetscObject)B,"MatIsHermitianTranspose_C",(void (**)(void))&g);CHKERRQ(ierr);
4026   if (f && g) {
4027     if (f==g) {
4028       ierr = (*f)(A,B,tol,flg);CHKERRQ(ierr);
4029     } else {
4030       SETERRQ(PETSC_ERR_ARG_NOTSAMETYPE,"Matrices do not have the same comparator for Hermitian test");
4031     }
4032   }
4033   PetscFunctionReturn(0);
4034 }
4035 
4036 #undef __FUNCT__
4037 #define __FUNCT__ "MatPermute"
4038 /*@
4039    MatPermute - Creates a new matrix with rows and columns permuted from the
4040    original.
4041 
4042    Collective on Mat
4043 
4044    Input Parameters:
4045 +  mat - the matrix to permute
4046 .  row - row permutation, each processor supplies only the permutation for its rows
4047 -  col - column permutation, each processor needs the entire column permutation, that is
4048          this is the same size as the total number of columns in the matrix. It can often
4049          be obtained with ISAllGather() on the row permutation
4050 
4051    Output Parameters:
4052 .  B - the permuted matrix
4053 
4054    Level: advanced
4055 
4056    Concepts: matrices^permuting
4057 
4058 .seealso: MatGetOrdering(), ISAllGather()
4059 
4060 @*/
4061 PetscErrorCode PETSCMAT_DLLEXPORT MatPermute(Mat mat,IS row,IS col,Mat *B)
4062 {
4063   PetscErrorCode ierr;
4064 
4065   PetscFunctionBegin;
4066   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
4067   PetscValidType(mat,1);
4068   PetscValidHeaderSpecific(row,IS_COOKIE,2);
4069   PetscValidHeaderSpecific(col,IS_COOKIE,3);
4070   PetscValidPointer(B,4);
4071   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4072   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4073   if (!mat->ops->permute) SETERRQ1(PETSC_ERR_SUP,"MatPermute not available for Mat type %s",((PetscObject)mat)->type_name);
4074   ierr = MatPreallocated(mat);CHKERRQ(ierr);
4075 
4076   ierr = (*mat->ops->permute)(mat,row,col,B);CHKERRQ(ierr);
4077   ierr = PetscObjectStateIncrease((PetscObject)*B);CHKERRQ(ierr);
4078   PetscFunctionReturn(0);
4079 }
4080 
4081 #undef __FUNCT__
4082 #define __FUNCT__ "MatPermuteSparsify"
4083 /*@
4084   MatPermuteSparsify - Creates a new matrix with rows and columns permuted from the
4085   original and sparsified to the prescribed tolerance.
4086 
4087   Collective on Mat
4088 
4089   Input Parameters:
4090 + A    - The matrix to permute
4091 . band - The half-bandwidth of the sparsified matrix, or PETSC_DECIDE
4092 . frac - The half-bandwidth as a fraction of the total size, or 0.0
4093 . tol  - The drop tolerance
4094 . rowp - The row permutation
4095 - colp - The column permutation
4096 
4097   Output Parameter:
4098 . B    - The permuted, sparsified matrix
4099 
4100   Level: advanced
4101 
4102   Note:
4103   The default behavior (band = PETSC_DECIDE and frac = 0.0) is to
4104   restrict the half-bandwidth of the resulting matrix to 5% of the
4105   total matrix size.
4106 
4107 .keywords: matrix, permute, sparsify
4108 
4109 .seealso: MatGetOrdering(), MatPermute()
4110 @*/
4111 PetscErrorCode PETSCMAT_DLLEXPORT MatPermuteSparsify(Mat A, PetscInt band, PetscReal frac, PetscReal tol, IS rowp, IS colp, Mat *B)
4112 {
4113   IS                irowp, icolp;
4114   const PetscInt    *rows, *cols;
4115   PetscInt          M, N, locRowStart, locRowEnd;
4116   PetscInt          nz, newNz;
4117   const PetscInt    *cwork;
4118   PetscInt          *cnew;
4119   const PetscScalar *vwork;
4120   PetscScalar       *vnew;
4121   PetscInt          bw, issize;
4122   PetscInt          row, locRow, newRow, col, newCol;
4123   PetscErrorCode    ierr;
4124 
4125   PetscFunctionBegin;
4126   PetscValidHeaderSpecific(A,    MAT_COOKIE,1);
4127   PetscValidHeaderSpecific(rowp, IS_COOKIE,5);
4128   PetscValidHeaderSpecific(colp, IS_COOKIE,6);
4129   PetscValidPointer(B,7);
4130   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4131   if (A->factor)     SETERRQ(PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4132   if (!A->ops->permutesparsify) {
4133     ierr = MatGetSize(A, &M, &N);CHKERRQ(ierr);
4134     ierr = MatGetOwnershipRange(A, &locRowStart, &locRowEnd);CHKERRQ(ierr);
4135     ierr = ISGetSize(rowp, &issize);CHKERRQ(ierr);
4136     if (issize != M) SETERRQ2(PETSC_ERR_ARG_WRONG, "Wrong size %D for row permutation, should be %D", issize, M);
4137     ierr = ISGetSize(colp, &issize);CHKERRQ(ierr);
4138     if (issize != N) SETERRQ2(PETSC_ERR_ARG_WRONG, "Wrong size %D for column permutation, should be %D", issize, N);
4139     ierr = ISInvertPermutation(rowp, 0, &irowp);CHKERRQ(ierr);
4140     ierr = ISGetIndices(irowp, &rows);CHKERRQ(ierr);
4141     ierr = ISInvertPermutation(colp, 0, &icolp);CHKERRQ(ierr);
4142     ierr = ISGetIndices(icolp, &cols);CHKERRQ(ierr);
4143     ierr = PetscMalloc(N * sizeof(PetscInt),         &cnew);CHKERRQ(ierr);
4144     ierr = PetscMalloc(N * sizeof(PetscScalar), &vnew);CHKERRQ(ierr);
4145 
4146     /* Setup bandwidth to include */
4147     if (band == PETSC_DECIDE) {
4148       if (frac <= 0.0)
4149         bw = (PetscInt) (M * 0.05);
4150       else
4151         bw = (PetscInt) (M * frac);
4152     } else {
4153       if (band <= 0) SETERRQ(PETSC_ERR_ARG_WRONG, "Bandwidth must be a positive integer");
4154       bw = band;
4155     }
4156 
4157     /* Put values into new matrix */
4158     ierr = MatDuplicate(A, MAT_DO_NOT_COPY_VALUES, B);CHKERRQ(ierr);
4159     for(row = locRowStart, locRow = 0; row < locRowEnd; row++, locRow++) {
4160       ierr = MatGetRow(A, row, &nz, &cwork, &vwork);CHKERRQ(ierr);
4161       newRow   = rows[locRow]+locRowStart;
4162       for(col = 0, newNz = 0; col < nz; col++) {
4163         newCol = cols[cwork[col]];
4164         if ((newCol >= newRow - bw) && (newCol < newRow + bw) && (PetscAbsScalar(vwork[col]) >= tol)) {
4165           cnew[newNz] = newCol;
4166           vnew[newNz] = vwork[col];
4167           newNz++;
4168         }
4169       }
4170       ierr = MatSetValues(*B, 1, &newRow, newNz, cnew, vnew, INSERT_VALUES);CHKERRQ(ierr);
4171       ierr = MatRestoreRow(A, row, &nz, &cwork, &vwork);CHKERRQ(ierr);
4172     }
4173     ierr = PetscFree(cnew);CHKERRQ(ierr);
4174     ierr = PetscFree(vnew);CHKERRQ(ierr);
4175     ierr = MatAssemblyBegin(*B, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
4176     ierr = MatAssemblyEnd(*B, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
4177     ierr = ISRestoreIndices(irowp, &rows);CHKERRQ(ierr);
4178     ierr = ISRestoreIndices(icolp, &cols);CHKERRQ(ierr);
4179     ierr = ISDestroy(irowp);CHKERRQ(ierr);
4180     ierr = ISDestroy(icolp);CHKERRQ(ierr);
4181   } else {
4182     ierr = (*A->ops->permutesparsify)(A, band, frac, tol, rowp, colp, B);CHKERRQ(ierr);
4183   }
4184   ierr = PetscObjectStateIncrease((PetscObject)*B);CHKERRQ(ierr);
4185   PetscFunctionReturn(0);
4186 }
4187 
4188 #undef __FUNCT__
4189 #define __FUNCT__ "MatEqual"
4190 /*@
4191    MatEqual - Compares two matrices.
4192 
4193    Collective on Mat
4194 
4195    Input Parameters:
4196 +  A - the first matrix
4197 -  B - the second matrix
4198 
4199    Output Parameter:
4200 .  flg - PETSC_TRUE if the matrices are equal; PETSC_FALSE otherwise.
4201 
4202    Level: intermediate
4203 
4204    Concepts: matrices^equality between
4205 @*/
4206 PetscErrorCode PETSCMAT_DLLEXPORT MatEqual(Mat A,Mat B,PetscTruth *flg)
4207 {
4208   PetscErrorCode ierr;
4209 
4210   PetscFunctionBegin;
4211   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
4212   PetscValidHeaderSpecific(B,MAT_COOKIE,2);
4213   PetscValidType(A,1);
4214   PetscValidType(B,2);
4215   PetscValidIntPointer(flg,3);
4216   PetscCheckSameComm(A,1,B,2);
4217   ierr = MatPreallocated(B);CHKERRQ(ierr);
4218   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4219   if (!B->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4220   if (A->rmap->N != B->rmap->N || A->cmap->N != B->cmap->N) SETERRQ4(PETSC_ERR_ARG_SIZ,"Mat A,Mat B: global dim %D %D %D %D",A->rmap->N,B->rmap->N,A->cmap->N,B->cmap->N);
4221   if (!A->ops->equal) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)A)->type_name);
4222   if (!B->ops->equal) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)B)->type_name);
4223   if (A->ops->equal != B->ops->equal) SETERRQ2(PETSC_ERR_ARG_INCOMP,"A is type: %s\nB is type: %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
4224   ierr = MatPreallocated(A);CHKERRQ(ierr);
4225 
4226   ierr = (*A->ops->equal)(A,B,flg);CHKERRQ(ierr);
4227   PetscFunctionReturn(0);
4228 }
4229 
4230 #undef __FUNCT__
4231 #define __FUNCT__ "MatDiagonalScale"
4232 /*@
4233    MatDiagonalScale - Scales a matrix on the left and right by diagonal
4234    matrices that are stored as vectors.  Either of the two scaling
4235    matrices can be PETSC_NULL.
4236 
4237    Collective on Mat
4238 
4239    Input Parameters:
4240 +  mat - the matrix to be scaled
4241 .  l - the left scaling vector (or PETSC_NULL)
4242 -  r - the right scaling vector (or PETSC_NULL)
4243 
4244    Notes:
4245    MatDiagonalScale() computes A = LAR, where
4246    L = a diagonal matrix (stored as a vector), R = a diagonal matrix (stored as a vector)
4247 
4248    Level: intermediate
4249 
4250    Concepts: matrices^diagonal scaling
4251    Concepts: diagonal scaling of matrices
4252 
4253 .seealso: MatScale()
4254 @*/
4255 PetscErrorCode PETSCMAT_DLLEXPORT MatDiagonalScale(Mat mat,Vec l,Vec r)
4256 {
4257   PetscErrorCode ierr;
4258 
4259   PetscFunctionBegin;
4260   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
4261   PetscValidType(mat,1);
4262   if (!mat->ops->diagonalscale) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4263   if (l) {PetscValidHeaderSpecific(l,VEC_COOKIE,2);PetscCheckSameComm(mat,1,l,2);}
4264   if (r) {PetscValidHeaderSpecific(r,VEC_COOKIE,3);PetscCheckSameComm(mat,1,r,3);}
4265   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4266   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4267   ierr = MatPreallocated(mat);CHKERRQ(ierr);
4268 
4269   ierr = PetscLogEventBegin(MAT_Scale,mat,0,0,0);CHKERRQ(ierr);
4270   ierr = (*mat->ops->diagonalscale)(mat,l,r);CHKERRQ(ierr);
4271   ierr = PetscLogEventEnd(MAT_Scale,mat,0,0,0);CHKERRQ(ierr);
4272   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
4273   PetscFunctionReturn(0);
4274 }
4275 
4276 #undef __FUNCT__
4277 #define __FUNCT__ "MatScale"
4278 /*@
4279     MatScale - Scales all elements of a matrix by a given number.
4280 
4281     Collective on Mat
4282 
4283     Input Parameters:
4284 +   mat - the matrix to be scaled
4285 -   a  - the scaling value
4286 
4287     Output Parameter:
4288 .   mat - the scaled matrix
4289 
4290     Level: intermediate
4291 
4292     Concepts: matrices^scaling all entries
4293 
4294 .seealso: MatDiagonalScale()
4295 @*/
4296 PetscErrorCode PETSCMAT_DLLEXPORT MatScale(Mat mat,PetscScalar a)
4297 {
4298   PetscErrorCode ierr;
4299 
4300   PetscFunctionBegin;
4301   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
4302   PetscValidType(mat,1);
4303   if (a != 1.0 && !mat->ops->scale) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4304   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4305   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4306   ierr = MatPreallocated(mat);CHKERRQ(ierr);
4307 
4308   ierr = PetscLogEventBegin(MAT_Scale,mat,0,0,0);CHKERRQ(ierr);
4309   if (a != 1.0) {
4310     ierr = (*mat->ops->scale)(mat,a);CHKERRQ(ierr);
4311     ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
4312   }
4313   ierr = PetscLogEventEnd(MAT_Scale,mat,0,0,0);CHKERRQ(ierr);
4314   PetscFunctionReturn(0);
4315 }
4316 
4317 #undef __FUNCT__
4318 #define __FUNCT__ "MatNorm"
4319 /*@
4320    MatNorm - Calculates various norms of a matrix.
4321 
4322    Collective on Mat
4323 
4324    Input Parameters:
4325 +  mat - the matrix
4326 -  type - the type of norm, NORM_1, NORM_FROBENIUS, NORM_INFINITY
4327 
4328    Output Parameters:
4329 .  nrm - the resulting norm
4330 
4331    Level: intermediate
4332 
4333    Concepts: matrices^norm
4334    Concepts: norm^of matrix
4335 @*/
4336 PetscErrorCode PETSCMAT_DLLEXPORT MatNorm(Mat mat,NormType type,PetscReal *nrm)
4337 {
4338   PetscErrorCode ierr;
4339 
4340   PetscFunctionBegin;
4341   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
4342   PetscValidType(mat,1);
4343   PetscValidScalarPointer(nrm,3);
4344 
4345   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4346   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4347   if (!mat->ops->norm) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4348   ierr = MatPreallocated(mat);CHKERRQ(ierr);
4349 
4350   ierr = (*mat->ops->norm)(mat,type,nrm);CHKERRQ(ierr);
4351   PetscFunctionReturn(0);
4352 }
4353 
4354 /*
4355      This variable is used to prevent counting of MatAssemblyBegin() that
4356    are called from within a MatAssemblyEnd().
4357 */
4358 static PetscInt MatAssemblyEnd_InUse = 0;
4359 #undef __FUNCT__
4360 #define __FUNCT__ "MatAssemblyBegin"
4361 /*@
4362    MatAssemblyBegin - Begins assembling the matrix.  This routine should
4363    be called after completing all calls to MatSetValues().
4364 
4365    Collective on Mat
4366 
4367    Input Parameters:
4368 +  mat - the matrix
4369 -  type - type of assembly, either MAT_FLUSH_ASSEMBLY or MAT_FINAL_ASSEMBLY
4370 
4371    Notes:
4372    MatSetValues() generally caches the values.  The matrix is ready to
4373    use only after MatAssemblyBegin() and MatAssemblyEnd() have been called.
4374    Use MAT_FLUSH_ASSEMBLY when switching between ADD_VALUES and INSERT_VALUES
4375    in MatSetValues(); use MAT_FINAL_ASSEMBLY for the final assembly before
4376    using the matrix.
4377 
4378    Level: beginner
4379 
4380    Concepts: matrices^assembling
4381 
4382 .seealso: MatAssemblyEnd(), MatSetValues(), MatAssembled()
4383 @*/
4384 PetscErrorCode PETSCMAT_DLLEXPORT MatAssemblyBegin(Mat mat,MatAssemblyType type)
4385 {
4386   PetscErrorCode ierr;
4387 
4388   PetscFunctionBegin;
4389   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
4390   PetscValidType(mat,1);
4391   ierr = MatPreallocated(mat);CHKERRQ(ierr);
4392   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix.\nDid you forget to call MatSetUnfactored()?");
4393   if (mat->assembled) {
4394     mat->was_assembled = PETSC_TRUE;
4395     mat->assembled     = PETSC_FALSE;
4396   }
4397   if (!MatAssemblyEnd_InUse) {
4398     ierr = PetscLogEventBegin(MAT_AssemblyBegin,mat,0,0,0);CHKERRQ(ierr);
4399     if (mat->ops->assemblybegin){ierr = (*mat->ops->assemblybegin)(mat,type);CHKERRQ(ierr);}
4400     ierr = PetscLogEventEnd(MAT_AssemblyBegin,mat,0,0,0);CHKERRQ(ierr);
4401   } else {
4402     if (mat->ops->assemblybegin){ierr = (*mat->ops->assemblybegin)(mat,type);CHKERRQ(ierr);}
4403   }
4404   PetscFunctionReturn(0);
4405 }
4406 
4407 #undef __FUNCT__
4408 #define __FUNCT__ "MatAssembed"
4409 /*@
4410    MatAssembled - Indicates if a matrix has been assembled and is ready for
4411      use; for example, in matrix-vector product.
4412 
4413    Collective on Mat
4414 
4415    Input Parameter:
4416 .  mat - the matrix
4417 
4418    Output Parameter:
4419 .  assembled - PETSC_TRUE or PETSC_FALSE
4420 
4421    Level: advanced
4422 
4423    Concepts: matrices^assembled?
4424 
4425 .seealso: MatAssemblyEnd(), MatSetValues(), MatAssemblyBegin()
4426 @*/
4427 PetscErrorCode PETSCMAT_DLLEXPORT MatAssembled(Mat mat,PetscTruth *assembled)
4428 {
4429   PetscFunctionBegin;
4430   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
4431   PetscValidType(mat,1);
4432   PetscValidPointer(assembled,2);
4433   *assembled = mat->assembled;
4434   PetscFunctionReturn(0);
4435 }
4436 
4437 #undef __FUNCT__
4438 #define __FUNCT__ "MatView_Private"
4439 /*
4440     Processes command line options to determine if/how a matrix
4441   is to be viewed. Called by MatAssemblyEnd() and MatLoad().
4442 */
4443 PetscErrorCode MatView_Private(Mat mat)
4444 {
4445   PetscErrorCode    ierr;
4446   PetscTruth        flg1 = PETSC_FALSE,flg2 = PETSC_FALSE,flg3 = PETSC_FALSE,flg4 = PETSC_FALSE,flg6 = PETSC_FALSE,flg7 = PETSC_FALSE,flg8 = PETSC_FALSE;
4447   static PetscTruth incall = PETSC_FALSE;
4448 #if defined(PETSC_USE_SOCKET_VIEWER)
4449   PetscTruth        flg5 = PETSC_FALSE;
4450 #endif
4451 
4452   PetscFunctionBegin;
4453   if (incall) PetscFunctionReturn(0);
4454   incall = PETSC_TRUE;
4455   ierr = PetscOptionsBegin(((PetscObject)mat)->comm,((PetscObject)mat)->prefix,"Matrix Options","Mat");CHKERRQ(ierr);
4456     ierr = PetscOptionsTruth("-mat_view_info","Information on matrix size","MatView",flg1,&flg1,PETSC_NULL);CHKERRQ(ierr);
4457     ierr = PetscOptionsTruth("-mat_view_info_detailed","Nonzeros in the matrix","MatView",flg2,&flg2,PETSC_NULL);CHKERRQ(ierr);
4458     ierr = PetscOptionsTruth("-mat_view","Print matrix to stdout","MatView",flg3,&flg3,PETSC_NULL);CHKERRQ(ierr);
4459     ierr = PetscOptionsTruth("-mat_view_matlab","Print matrix to stdout in a format Matlab can read","MatView",flg4,&flg4,PETSC_NULL);CHKERRQ(ierr);
4460 #if defined(PETSC_USE_SOCKET_VIEWER)
4461     ierr = PetscOptionsTruth("-mat_view_socket","Send matrix to socket (can be read from matlab)","MatView",flg5,&flg5,PETSC_NULL);CHKERRQ(ierr);
4462 #endif
4463     ierr = PetscOptionsTruth("-mat_view_binary","Save matrix to file in binary format","MatView",flg6,&flg6,PETSC_NULL);CHKERRQ(ierr);
4464     ierr = PetscOptionsTruth("-mat_view_draw","Draw the matrix nonzero structure","MatView",flg7,&flg7,PETSC_NULL);CHKERRQ(ierr);
4465   ierr = PetscOptionsEnd();CHKERRQ(ierr);
4466 
4467   if (flg1) {
4468     PetscViewer viewer;
4469 
4470     ierr = PetscViewerASCIIGetStdout(((PetscObject)mat)->comm,&viewer);CHKERRQ(ierr);
4471     ierr = PetscViewerPushFormat(viewer,PETSC_VIEWER_ASCII_INFO);CHKERRQ(ierr);
4472     ierr = MatView(mat,viewer);CHKERRQ(ierr);
4473     ierr = PetscViewerPopFormat(viewer);CHKERRQ(ierr);
4474   }
4475   if (flg2) {
4476     PetscViewer viewer;
4477 
4478     ierr = PetscViewerASCIIGetStdout(((PetscObject)mat)->comm,&viewer);CHKERRQ(ierr);
4479     ierr = PetscViewerPushFormat(viewer,PETSC_VIEWER_ASCII_INFO_DETAIL);CHKERRQ(ierr);
4480     ierr = MatView(mat,viewer);CHKERRQ(ierr);
4481     ierr = PetscViewerPopFormat(viewer);CHKERRQ(ierr);
4482   }
4483   if (flg3) {
4484     PetscViewer viewer;
4485 
4486     ierr = PetscViewerASCIIGetStdout(((PetscObject)mat)->comm,&viewer);CHKERRQ(ierr);
4487     ierr = MatView(mat,viewer);CHKERRQ(ierr);
4488   }
4489   if (flg4) {
4490     PetscViewer viewer;
4491 
4492     ierr = PetscViewerASCIIGetStdout(((PetscObject)mat)->comm,&viewer);CHKERRQ(ierr);
4493     ierr = PetscViewerPushFormat(viewer,PETSC_VIEWER_ASCII_MATLAB);CHKERRQ(ierr);
4494     ierr = MatView(mat,viewer);CHKERRQ(ierr);
4495     ierr = PetscViewerPopFormat(viewer);CHKERRQ(ierr);
4496   }
4497 #if defined(PETSC_USE_SOCKET_VIEWER)
4498   if (flg5) {
4499     ierr = MatView(mat,PETSC_VIEWER_SOCKET_(((PetscObject)mat)->comm));CHKERRQ(ierr);
4500     ierr = PetscViewerFlush(PETSC_VIEWER_SOCKET_(((PetscObject)mat)->comm));CHKERRQ(ierr);
4501   }
4502 #endif
4503   if (flg6) {
4504     ierr = MatView(mat,PETSC_VIEWER_BINARY_(((PetscObject)mat)->comm));CHKERRQ(ierr);
4505     ierr = PetscViewerFlush(PETSC_VIEWER_BINARY_(((PetscObject)mat)->comm));CHKERRQ(ierr);
4506   }
4507   if (flg7) {
4508     ierr = PetscOptionsGetTruth(((PetscObject)mat)->prefix,"-mat_view_contour",&flg8,PETSC_NULL);CHKERRQ(ierr);
4509     if (flg8) {
4510       PetscViewerPushFormat(PETSC_VIEWER_DRAW_(((PetscObject)mat)->comm),PETSC_VIEWER_DRAW_CONTOUR);CHKERRQ(ierr);
4511     }
4512     ierr = MatView(mat,PETSC_VIEWER_DRAW_(((PetscObject)mat)->comm));CHKERRQ(ierr);
4513     ierr = PetscViewerFlush(PETSC_VIEWER_DRAW_(((PetscObject)mat)->comm));CHKERRQ(ierr);
4514     if (flg8) {
4515       PetscViewerPopFormat(PETSC_VIEWER_DRAW_(((PetscObject)mat)->comm));CHKERRQ(ierr);
4516     }
4517   }
4518   incall = PETSC_FALSE;
4519   PetscFunctionReturn(0);
4520 }
4521 
4522 #undef __FUNCT__
4523 #define __FUNCT__ "MatAssemblyEnd"
4524 /*@
4525    MatAssemblyEnd - Completes assembling the matrix.  This routine should
4526    be called after MatAssemblyBegin().
4527 
4528    Collective on Mat
4529 
4530    Input Parameters:
4531 +  mat - the matrix
4532 -  type - type of assembly, either MAT_FLUSH_ASSEMBLY or MAT_FINAL_ASSEMBLY
4533 
4534    Options Database Keys:
4535 +  -mat_view_info - Prints info on matrix at conclusion of MatEndAssembly()
4536 .  -mat_view_info_detailed - Prints more detailed info
4537 .  -mat_view - Prints matrix in ASCII format
4538 .  -mat_view_matlab - Prints matrix in Matlab format
4539 .  -mat_view_draw - PetscDraws nonzero structure of matrix, using MatView() and PetscDrawOpenX().
4540 .  -display <name> - Sets display name (default is host)
4541 .  -draw_pause <sec> - Sets number of seconds to pause after display
4542 .  -mat_view_socket - Sends matrix to socket, can be accessed from Matlab (see users manual)
4543 .  -viewer_socket_machine <machine>
4544 .  -viewer_socket_port <port>
4545 .  -mat_view_binary - save matrix to file in binary format
4546 -  -viewer_binary_filename <name>
4547 
4548    Notes:
4549    MatSetValues() generally caches the values.  The matrix is ready to
4550    use only after MatAssemblyBegin() and MatAssemblyEnd() have been called.
4551    Use MAT_FLUSH_ASSEMBLY when switching between ADD_VALUES and INSERT_VALUES
4552    in MatSetValues(); use MAT_FINAL_ASSEMBLY for the final assembly before
4553    using the matrix.
4554 
4555    Level: beginner
4556 
4557 .seealso: MatAssemblyBegin(), MatSetValues(), PetscDrawOpenX(), MatView(), MatAssembled(), PetscViewerSocketOpen()
4558 @*/
4559 PetscErrorCode PETSCMAT_DLLEXPORT MatAssemblyEnd(Mat mat,MatAssemblyType type)
4560 {
4561   PetscErrorCode  ierr;
4562   static PetscInt inassm = 0;
4563   PetscTruth      flg = PETSC_FALSE;
4564 
4565   PetscFunctionBegin;
4566   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
4567   PetscValidType(mat,1);
4568 
4569   inassm++;
4570   MatAssemblyEnd_InUse++;
4571   if (MatAssemblyEnd_InUse == 1) { /* Do the logging only the first time through */
4572     ierr = PetscLogEventBegin(MAT_AssemblyEnd,mat,0,0,0);CHKERRQ(ierr);
4573     if (mat->ops->assemblyend) {
4574       ierr = (*mat->ops->assemblyend)(mat,type);CHKERRQ(ierr);
4575     }
4576     ierr = PetscLogEventEnd(MAT_AssemblyEnd,mat,0,0,0);CHKERRQ(ierr);
4577   } else {
4578     if (mat->ops->assemblyend) {
4579       ierr = (*mat->ops->assemblyend)(mat,type);CHKERRQ(ierr);
4580     }
4581   }
4582 
4583   /* Flush assembly is not a true assembly */
4584   if (type != MAT_FLUSH_ASSEMBLY) {
4585     mat->assembled  = PETSC_TRUE; mat->num_ass++;
4586   }
4587   mat->insertmode = NOT_SET_VALUES;
4588   MatAssemblyEnd_InUse--;
4589   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
4590   if (!mat->symmetric_eternal) {
4591     mat->symmetric_set              = PETSC_FALSE;
4592     mat->hermitian_set              = PETSC_FALSE;
4593     mat->structurally_symmetric_set = PETSC_FALSE;
4594   }
4595   if (inassm == 1 && type != MAT_FLUSH_ASSEMBLY) {
4596     ierr = MatView_Private(mat);CHKERRQ(ierr);
4597     ierr = PetscOptionsGetTruth(((PetscObject)mat)->prefix,"-mat_is_symmetric",&flg,PETSC_NULL);CHKERRQ(ierr);
4598     if (flg) {
4599       PetscReal tol = 0.0;
4600       ierr = PetscOptionsGetReal(((PetscObject)mat)->prefix,"-mat_is_symmetric",&tol,PETSC_NULL);CHKERRQ(ierr);
4601       ierr = MatIsSymmetric(mat,tol,&flg);CHKERRQ(ierr);
4602       if (flg) {
4603         ierr = PetscPrintf(((PetscObject)mat)->comm,"Matrix is symmetric (tolerance %G)\n",tol);CHKERRQ(ierr);
4604       } else {
4605         ierr = PetscPrintf(((PetscObject)mat)->comm,"Matrix is not symmetric (tolerance %G)\n",tol);CHKERRQ(ierr);
4606       }
4607     }
4608   }
4609   inassm--;
4610   PetscFunctionReturn(0);
4611 }
4612 
4613 #undef __FUNCT__
4614 #define __FUNCT__ "MatSetOption"
4615 /*@
4616    MatSetOption - Sets a parameter option for a matrix. Some options
4617    may be specific to certain storage formats.  Some options
4618    determine how values will be inserted (or added). Sorted,
4619    row-oriented input will generally assemble the fastest. The default
4620    is row-oriented, nonsorted input.
4621 
4622    Collective on Mat
4623 
4624    Input Parameters:
4625 +  mat - the matrix
4626 .  option - the option, one of those listed below (and possibly others),
4627 -  flg - turn the option on (PETSC_TRUE) or off (PETSC_FALSE)
4628 
4629   Options Describing Matrix Structure:
4630 +    MAT_SYMMETRIC - symmetric in terms of both structure and value
4631 .    MAT_HERMITIAN - transpose is the complex conjugation
4632 .    MAT_STRUCTURALLY_SYMMETRIC - symmetric nonzero structure
4633 -    MAT_SYMMETRY_ETERNAL - if you would like the symmetry/Hermitian flag
4634                             you set to be kept with all future use of the matrix
4635                             including after MatAssemblyBegin/End() which could
4636                             potentially change the symmetry structure, i.e. you
4637                             KNOW the matrix will ALWAYS have the property you set.
4638 
4639 
4640    Options For Use with MatSetValues():
4641    Insert a logically dense subblock, which can be
4642 .    MAT_ROW_ORIENTED - row-oriented (default)
4643 
4644    Note these options reflect the data you pass in with MatSetValues(); it has
4645    nothing to do with how the data is stored internally in the matrix
4646    data structure.
4647 
4648    When (re)assembling a matrix, we can restrict the input for
4649    efficiency/debugging purposes.  These options include
4650 +    MAT_NEW_NONZERO_LOCATIONS - additional insertions will be
4651         allowed if they generate a new nonzero
4652 .    MAT_NEW_DIAGONALS - new diagonals will be allowed (for block diagonal format only)
4653 .    MAT_IGNORE_OFF_PROC_ENTRIES - drops off-processor entries
4654 .    MAT_NEW_NONZERO_LOCATION_ERR - generates an error for new matrix entry
4655 -    MAT_USE_HASH_TABLE - uses a hash table to speed up matrix assembly
4656 
4657    Notes:
4658    Some options are relevant only for particular matrix types and
4659    are thus ignored by others.  Other options are not supported by
4660    certain matrix types and will generate an error message if set.
4661 
4662    If using a Fortran 77 module to compute a matrix, one may need to
4663    use the column-oriented option (or convert to the row-oriented
4664    format).
4665 
4666    MAT_NEW_NONZERO_LOCATIONS set to PETSC_FALSE indicates that any add or insertion
4667    that would generate a new entry in the nonzero structure is instead
4668    ignored.  Thus, if memory has not alredy been allocated for this particular
4669    data, then the insertion is ignored. For dense matrices, in which
4670    the entire array is allocated, no entries are ever ignored.
4671    Set after the first MatAssemblyEnd()
4672 
4673    MAT_NEW_NONZERO_LOCATION_ERR indicates that any add or insertion
4674    that would generate a new entry in the nonzero structure instead produces
4675    an error. (Currently supported for AIJ and BAIJ formats only.)
4676    This is a useful flag when using SAME_NONZERO_PATTERN in calling
4677    KSPSetOperators() to ensure that the nonzero pattern truely does
4678    remain unchanged. Set after the first MatAssemblyEnd()
4679 
4680    MAT_NEW_NONZERO_ALLOCATION_ERR indicates that any add or insertion
4681    that would generate a new entry that has not been preallocated will
4682    instead produce an error. (Currently supported for AIJ and BAIJ formats
4683    only.) This is a useful flag when debugging matrix memory preallocation.
4684 
4685    MAT_IGNORE_OFF_PROC_ENTRIES indicates entries destined for
4686    other processors should be dropped, rather than stashed.
4687    This is useful if you know that the "owning" processor is also
4688    always generating the correct matrix entries, so that PETSc need
4689    not transfer duplicate entries generated on another processor.
4690 
4691    MAT_USE_HASH_TABLE indicates that a hash table be used to improve the
4692    searches during matrix assembly. When this flag is set, the hash table
4693    is created during the first Matrix Assembly. This hash table is
4694    used the next time through, during MatSetVaules()/MatSetVaulesBlocked()
4695    to improve the searching of indices. MAT_NEW_NONZERO_LOCATIONS flag
4696    should be used with MAT_USE_HASH_TABLE flag. This option is currently
4697    supported by MATMPIBAIJ format only.
4698 
4699    MAT_KEEP_ZEROED_ROWS indicates when MatZeroRows() is called the zeroed entries
4700    are kept in the nonzero structure
4701 
4702    MAT_IGNORE_ZERO_ENTRIES - for AIJ/IS matrices this will stop zero values from creating
4703    a zero location in the matrix
4704 
4705    MAT_USE_INODES - indicates using inode version of the code - works with AIJ and
4706    ROWBS matrix types
4707 
4708    Level: intermediate
4709 
4710    Concepts: matrices^setting options
4711 
4712 @*/
4713 PetscErrorCode PETSCMAT_DLLEXPORT MatSetOption(Mat mat,MatOption op,PetscTruth flg)
4714 {
4715   PetscErrorCode ierr;
4716 
4717   PetscFunctionBegin;
4718   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
4719   PetscValidType(mat,1);
4720   if (((int) op) < 0 || ((int) op) >= NUM_MAT_OPTIONS) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Options %d is out of range",(int)op);
4721   ierr = MatPreallocated(mat);CHKERRQ(ierr);
4722   switch (op) {
4723   case MAT_SYMMETRIC:
4724     mat->symmetric                  = flg;
4725     if (flg) mat->structurally_symmetric     = PETSC_TRUE;
4726     mat->symmetric_set              = PETSC_TRUE;
4727     mat->structurally_symmetric_set = flg;
4728     break;
4729   case MAT_HERMITIAN:
4730     mat->hermitian                  = flg;
4731     if (flg) mat->structurally_symmetric     = PETSC_TRUE;
4732     mat->hermitian_set              = PETSC_TRUE;
4733     mat->structurally_symmetric_set = flg;
4734     break;
4735   case MAT_STRUCTURALLY_SYMMETRIC:
4736     mat->structurally_symmetric     = flg;
4737     mat->structurally_symmetric_set = PETSC_TRUE;
4738     break;
4739   case MAT_SYMMETRY_ETERNAL:
4740     mat->symmetric_eternal          = flg;
4741     break;
4742   default:
4743     break;
4744   }
4745   if (mat->ops->setoption) {
4746     ierr = (*mat->ops->setoption)(mat,op,flg);CHKERRQ(ierr);
4747   }
4748   PetscFunctionReturn(0);
4749 }
4750 
4751 #undef __FUNCT__
4752 #define __FUNCT__ "MatZeroEntries"
4753 /*@
4754    MatZeroEntries - Zeros all entries of a matrix.  For sparse matrices
4755    this routine retains the old nonzero structure.
4756 
4757    Collective on Mat
4758 
4759    Input Parameters:
4760 .  mat - the matrix
4761 
4762    Level: intermediate
4763 
4764    Concepts: matrices^zeroing
4765 
4766 .seealso: MatZeroRows()
4767 @*/
4768 PetscErrorCode PETSCMAT_DLLEXPORT MatZeroEntries(Mat mat)
4769 {
4770   PetscErrorCode ierr;
4771 
4772   PetscFunctionBegin;
4773   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
4774   PetscValidType(mat,1);
4775   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4776   if (mat->insertmode != NOT_SET_VALUES) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for matrices where you have set values but not yet assembled");
4777   if (!mat->ops->zeroentries) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4778   ierr = MatPreallocated(mat);CHKERRQ(ierr);
4779 
4780   ierr = PetscLogEventBegin(MAT_ZeroEntries,mat,0,0,0);CHKERRQ(ierr);
4781   ierr = (*mat->ops->zeroentries)(mat);CHKERRQ(ierr);
4782   ierr = PetscLogEventEnd(MAT_ZeroEntries,mat,0,0,0);CHKERRQ(ierr);
4783   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
4784   PetscFunctionReturn(0);
4785 }
4786 
4787 #undef __FUNCT__
4788 #define __FUNCT__ "MatZeroRows"
4789 /*@C
4790    MatZeroRows - Zeros all entries (except possibly the main diagonal)
4791    of a set of rows of a matrix.
4792 
4793    Collective on Mat
4794 
4795    Input Parameters:
4796 +  mat - the matrix
4797 .  numRows - the number of rows to remove
4798 .  rows - the global row indices
4799 -  diag - value put in all diagonals of eliminated rows (0.0 will even eliminate diagonal entry)
4800 
4801    Notes:
4802    For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
4803    but does not release memory.  For the dense and block diagonal
4804    formats this does not alter the nonzero structure.
4805 
4806    If the option MatSetOption(mat,MAT_KEEP_ZEROED_ROWS,PETSC_TRUE) the nonzero structure
4807    of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
4808    merely zeroed.
4809 
4810    The user can set a value in the diagonal entry (or for the AIJ and
4811    row formats can optionally remove the main diagonal entry from the
4812    nonzero structure as well, by passing 0.0 as the final argument).
4813 
4814    For the parallel case, all processes that share the matrix (i.e.,
4815    those in the communicator used for matrix creation) MUST call this
4816    routine, regardless of whether any rows being zeroed are owned by
4817    them.
4818 
4819    Each processor can indicate any rows in the entire matrix to be zeroed (i.e. each process does NOT have to
4820    list only rows local to itself).
4821 
4822    Level: intermediate
4823 
4824    Concepts: matrices^zeroing rows
4825 
4826 .seealso: MatZeroRowsIS(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption()
4827 @*/
4828 PetscErrorCode PETSCMAT_DLLEXPORT MatZeroRows(Mat mat,PetscInt numRows,const PetscInt rows[],PetscScalar diag)
4829 {
4830   PetscErrorCode ierr;
4831 
4832   PetscFunctionBegin;
4833   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
4834   PetscValidType(mat,1);
4835   if (numRows) PetscValidIntPointer(rows,3);
4836   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4837   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4838   if (!mat->ops->zerorows) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
4839   ierr = MatPreallocated(mat);CHKERRQ(ierr);
4840 
4841   ierr = (*mat->ops->zerorows)(mat,numRows,rows,diag);CHKERRQ(ierr);
4842   ierr = MatView_Private(mat);CHKERRQ(ierr);
4843   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
4844   PetscFunctionReturn(0);
4845 }
4846 
4847 #undef __FUNCT__
4848 #define __FUNCT__ "MatZeroRowsIS"
4849 /*@C
4850    MatZeroRowsIS - Zeros all entries (except possibly the main diagonal)
4851    of a set of rows of a matrix.
4852 
4853    Collective on Mat
4854 
4855    Input Parameters:
4856 +  mat - the matrix
4857 .  is - index set of rows to remove
4858 -  diag - value put in all diagonals of eliminated rows
4859 
4860    Notes:
4861    For the AIJ and BAIJ matrix formats this removes the old nonzero structure,
4862    but does not release memory.  For the dense and block diagonal
4863    formats this does not alter the nonzero structure.
4864 
4865    If the option MatSetOption(mat,MAT_KEEP_ZEROED_ROWS,PETSC_TRUE) the nonzero structure
4866    of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
4867    merely zeroed.
4868 
4869    The user can set a value in the diagonal entry (or for the AIJ and
4870    row formats can optionally remove the main diagonal entry from the
4871    nonzero structure as well, by passing 0.0 as the final argument).
4872 
4873    For the parallel case, all processes that share the matrix (i.e.,
4874    those in the communicator used for matrix creation) MUST call this
4875    routine, regardless of whether any rows being zeroed are owned by
4876    them.
4877 
4878    Each processor should list the rows that IT wants zeroed
4879 
4880    Level: intermediate
4881 
4882    Concepts: matrices^zeroing rows
4883 
4884 .seealso: MatZeroRows(), MatZeroEntries(), MatZeroRowsLocal(), MatSetOption()
4885 @*/
4886 PetscErrorCode PETSCMAT_DLLEXPORT MatZeroRowsIS(Mat mat,IS is,PetscScalar diag)
4887 {
4888   PetscInt       numRows;
4889   const PetscInt *rows;
4890   PetscErrorCode ierr;
4891 
4892   PetscFunctionBegin;
4893   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
4894   PetscValidType(mat,1);
4895   PetscValidHeaderSpecific(is,IS_COOKIE,2);
4896   ierr = ISGetLocalSize(is,&numRows);CHKERRQ(ierr);
4897   ierr = ISGetIndices(is,&rows);CHKERRQ(ierr);
4898   ierr = MatZeroRows(mat,numRows,rows,diag);CHKERRQ(ierr);
4899   ierr = ISRestoreIndices(is,&rows);CHKERRQ(ierr);
4900   PetscFunctionReturn(0);
4901 }
4902 
4903 #undef __FUNCT__
4904 #define __FUNCT__ "MatZeroRowsLocal"
4905 /*@C
4906    MatZeroRowsLocal - Zeros all entries (except possibly the main diagonal)
4907    of a set of rows of a matrix; using local numbering of rows.
4908 
4909    Collective on Mat
4910 
4911    Input Parameters:
4912 +  mat - the matrix
4913 .  numRows - the number of rows to remove
4914 .  rows - the global row indices
4915 -  diag - value put in all diagonals of eliminated rows
4916 
4917    Notes:
4918    Before calling MatZeroRowsLocal(), the user must first set the
4919    local-to-global mapping by calling MatSetLocalToGlobalMapping().
4920 
4921    For the AIJ matrix formats this removes the old nonzero structure,
4922    but does not release memory.  For the dense and block diagonal
4923    formats this does not alter the nonzero structure.
4924 
4925    If the option MatSetOption(mat,MAT_KEEP_ZEROED_ROWS,PETSC_TRUE) the nonzero structure
4926    of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
4927    merely zeroed.
4928 
4929    The user can set a value in the diagonal entry (or for the AIJ and
4930    row formats can optionally remove the main diagonal entry from the
4931    nonzero structure as well, by passing 0.0 as the final argument).
4932 
4933    Level: intermediate
4934 
4935    Concepts: matrices^zeroing
4936 
4937 .seealso: MatZeroRows(), MatZeroRowsLocalIS(), MatZeroEntries(), MatZeroRows(), MatSetLocalToGlobalMapping
4938 @*/
4939 PetscErrorCode PETSCMAT_DLLEXPORT MatZeroRowsLocal(Mat mat,PetscInt numRows,const PetscInt rows[],PetscScalar diag)
4940 {
4941   PetscErrorCode ierr;
4942 
4943   PetscFunctionBegin;
4944   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
4945   PetscValidType(mat,1);
4946   if (numRows) PetscValidIntPointer(rows,3);
4947   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
4948   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
4949   ierr = MatPreallocated(mat);CHKERRQ(ierr);
4950 
4951   if (mat->ops->zerorowslocal) {
4952     ierr = (*mat->ops->zerorowslocal)(mat,numRows,rows,diag);CHKERRQ(ierr);
4953   } else {
4954     IS             is, newis;
4955     const PetscInt *newRows;
4956 
4957     if (!mat->mapping) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Need to provide local to global mapping to matrix first");
4958     ierr = ISCreateGeneral(PETSC_COMM_SELF,numRows,rows,&is);CHKERRQ(ierr);
4959     ierr = ISLocalToGlobalMappingApplyIS(mat->mapping,is,&newis);CHKERRQ(ierr);
4960     ierr = ISGetIndices(newis,&newRows);CHKERRQ(ierr);
4961     ierr = (*mat->ops->zerorows)(mat,numRows,newRows,diag);CHKERRQ(ierr);
4962     ierr = ISRestoreIndices(newis,&newRows);CHKERRQ(ierr);
4963     ierr = ISDestroy(newis);CHKERRQ(ierr);
4964     ierr = ISDestroy(is);CHKERRQ(ierr);
4965   }
4966   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
4967   PetscFunctionReturn(0);
4968 }
4969 
4970 #undef __FUNCT__
4971 #define __FUNCT__ "MatZeroRowsLocalIS"
4972 /*@C
4973    MatZeroRowsLocalIS - Zeros all entries (except possibly the main diagonal)
4974    of a set of rows of a matrix; using local numbering of rows.
4975 
4976    Collective on Mat
4977 
4978    Input Parameters:
4979 +  mat - the matrix
4980 .  is - index set of rows to remove
4981 -  diag - value put in all diagonals of eliminated rows
4982 
4983    Notes:
4984    Before calling MatZeroRowsLocalIS(), the user must first set the
4985    local-to-global mapping by calling MatSetLocalToGlobalMapping().
4986 
4987    For the AIJ matrix formats this removes the old nonzero structure,
4988    but does not release memory.  For the dense and block diagonal
4989    formats this does not alter the nonzero structure.
4990 
4991    If the option MatSetOption(mat,MAT_KEEP_ZEROED_ROWS,PETSC_TRUE) the nonzero structure
4992    of the matrix is not changed (even for AIJ and BAIJ matrices) the values are
4993    merely zeroed.
4994 
4995    The user can set a value in the diagonal entry (or for the AIJ and
4996    row formats can optionally remove the main diagonal entry from the
4997    nonzero structure as well, by passing 0.0 as the final argument).
4998 
4999    Level: intermediate
5000 
5001    Concepts: matrices^zeroing
5002 
5003 .seealso: MatZeroRows(), MatZeroRowsLocal(), MatZeroEntries(), MatZeroRows(), MatSetLocalToGlobalMapping
5004 @*/
5005 PetscErrorCode PETSCMAT_DLLEXPORT MatZeroRowsLocalIS(Mat mat,IS is,PetscScalar diag)
5006 {
5007   PetscErrorCode ierr;
5008   PetscInt       numRows;
5009   const PetscInt *rows;
5010 
5011   PetscFunctionBegin;
5012   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5013   PetscValidType(mat,1);
5014   PetscValidHeaderSpecific(is,IS_COOKIE,2);
5015   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5016   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5017   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5018 
5019   ierr = ISGetLocalSize(is,&numRows);CHKERRQ(ierr);
5020   ierr = ISGetIndices(is,&rows);CHKERRQ(ierr);
5021   ierr = MatZeroRowsLocal(mat,numRows,rows,diag);CHKERRQ(ierr);
5022   ierr = ISRestoreIndices(is,&rows);CHKERRQ(ierr);
5023   PetscFunctionReturn(0);
5024 }
5025 
5026 #undef __FUNCT__
5027 #define __FUNCT__ "MatGetSize"
5028 /*@
5029    MatGetSize - Returns the numbers of rows and columns in a matrix.
5030 
5031    Not Collective
5032 
5033    Input Parameter:
5034 .  mat - the matrix
5035 
5036    Output Parameters:
5037 +  m - the number of global rows
5038 -  n - the number of global columns
5039 
5040    Note: both output parameters can be PETSC_NULL on input.
5041 
5042    Level: beginner
5043 
5044    Concepts: matrices^size
5045 
5046 .seealso: MatGetLocalSize()
5047 @*/
5048 PetscErrorCode PETSCMAT_DLLEXPORT MatGetSize(Mat mat,PetscInt *m,PetscInt* n)
5049 {
5050   PetscFunctionBegin;
5051   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5052   if (m) *m = mat->rmap->N;
5053   if (n) *n = mat->cmap->N;
5054   PetscFunctionReturn(0);
5055 }
5056 
5057 #undef __FUNCT__
5058 #define __FUNCT__ "MatGetLocalSize"
5059 /*@
5060    MatGetLocalSize - Returns the number of rows and columns in a matrix
5061    stored locally.  This information may be implementation dependent, so
5062    use with care.
5063 
5064    Not Collective
5065 
5066    Input Parameters:
5067 .  mat - the matrix
5068 
5069    Output Parameters:
5070 +  m - the number of local rows
5071 -  n - the number of local columns
5072 
5073    Note: both output parameters can be PETSC_NULL on input.
5074 
5075    Level: beginner
5076 
5077    Concepts: matrices^local size
5078 
5079 .seealso: MatGetSize()
5080 @*/
5081 PetscErrorCode PETSCMAT_DLLEXPORT MatGetLocalSize(Mat mat,PetscInt *m,PetscInt* n)
5082 {
5083   PetscFunctionBegin;
5084   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5085   if (m) PetscValidIntPointer(m,2);
5086   if (n) PetscValidIntPointer(n,3);
5087   if (m) *m = mat->rmap->n;
5088   if (n) *n = mat->cmap->n;
5089   PetscFunctionReturn(0);
5090 }
5091 
5092 #undef __FUNCT__
5093 #define __FUNCT__ "MatGetOwnershipRangeColumn"
5094 /*@
5095    MatGetOwnershipRangeColumn - Returns the range of matrix columns owned by
5096    this processor.
5097 
5098    Not Collective, unless matrix has not been allocated, then collective on Mat
5099 
5100    Input Parameters:
5101 .  mat - the matrix
5102 
5103    Output Parameters:
5104 +  m - the global index of the first local column
5105 -  n - one more than the global index of the last local column
5106 
5107    Notes: both output parameters can be PETSC_NULL on input.
5108 
5109    Level: developer
5110 
5111    Concepts: matrices^column ownership
5112 
5113 .seealso:  MatGetOwnershipRange(), MatGetOwnershipRanges(), MatGetOwnershipRangesColumn()
5114 
5115 @*/
5116 PetscErrorCode PETSCMAT_DLLEXPORT MatGetOwnershipRangeColumn(Mat mat,PetscInt *m,PetscInt* n)
5117 {
5118   PetscErrorCode ierr;
5119 
5120   PetscFunctionBegin;
5121   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5122   PetscValidType(mat,1);
5123   if (m) PetscValidIntPointer(m,2);
5124   if (n) PetscValidIntPointer(n,3);
5125   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5126   if (m) *m = mat->cmap->rstart;
5127   if (n) *n = mat->cmap->rend;
5128   PetscFunctionReturn(0);
5129 }
5130 
5131 #undef __FUNCT__
5132 #define __FUNCT__ "MatGetOwnershipRange"
5133 /*@
5134    MatGetOwnershipRange - Returns the range of matrix rows owned by
5135    this processor, assuming that the matrix is laid out with the first
5136    n1 rows on the first processor, the next n2 rows on the second, etc.
5137    For certain parallel layouts this range may not be well defined.
5138 
5139    Not Collective, unless matrix has not been allocated, then collective on Mat
5140 
5141    Input Parameters:
5142 .  mat - the matrix
5143 
5144    Output Parameters:
5145 +  m - the global index of the first local row
5146 -  n - one more than the global index of the last local row
5147 
5148    Note: both output parameters can be PETSC_NULL on input.
5149 
5150    Level: beginner
5151 
5152    Concepts: matrices^row ownership
5153 
5154 .seealso:   MatGetOwnershipRanges(), MatGetOwnershipRangeColumn(), MatGetOwnershipRangesColumn()
5155 
5156 @*/
5157 PetscErrorCode PETSCMAT_DLLEXPORT MatGetOwnershipRange(Mat mat,PetscInt *m,PetscInt* n)
5158 {
5159   PetscErrorCode ierr;
5160 
5161   PetscFunctionBegin;
5162   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5163   PetscValidType(mat,1);
5164   if (m) PetscValidIntPointer(m,2);
5165   if (n) PetscValidIntPointer(n,3);
5166   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5167   if (m) *m = mat->rmap->rstart;
5168   if (n) *n = mat->rmap->rend;
5169   PetscFunctionReturn(0);
5170 }
5171 
5172 #undef __FUNCT__
5173 #define __FUNCT__ "MatGetOwnershipRanges"
5174 /*@C
5175    MatGetOwnershipRanges - Returns the range of matrix rows owned by
5176    each process
5177 
5178    Not Collective, unless matrix has not been allocated, then collective on Mat
5179 
5180    Input Parameters:
5181 .  mat - the matrix
5182 
5183    Output Parameters:
5184 .  ranges - start of each processors portion plus one more then the total length at the end
5185 
5186    Level: beginner
5187 
5188    Concepts: matrices^row ownership
5189 
5190 .seealso:   MatGetOwnershipRange(), MatGetOwnershipRangeColumn(), MatGetOwnershipRangesColumn()
5191 
5192 @*/
5193 PetscErrorCode PETSCMAT_DLLEXPORT MatGetOwnershipRanges(Mat mat,const PetscInt **ranges)
5194 {
5195   PetscErrorCode ierr;
5196 
5197   PetscFunctionBegin;
5198   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5199   PetscValidType(mat,1);
5200   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5201   ierr = PetscMapGetRanges(mat->rmap,ranges);CHKERRQ(ierr);
5202   PetscFunctionReturn(0);
5203 }
5204 
5205 #undef __FUNCT__
5206 #define __FUNCT__ "MatGetOwnershipRangesColumn"
5207 /*@C
5208    MatGetOwnershipRangesColumn - Returns the range of local columns for each process
5209 
5210    Not Collective, unless matrix has not been allocated, then collective on Mat
5211 
5212    Input Parameters:
5213 .  mat - the matrix
5214 
5215    Output Parameters:
5216 .  ranges - start of each processors portion plus one more then the total length at the end
5217 
5218    Level: beginner
5219 
5220    Concepts: matrices^column ownership
5221 
5222 .seealso:   MatGetOwnershipRange(), MatGetOwnershipRangeColumn(), MatGetOwnershipRanges()
5223 
5224 @*/
5225 PetscErrorCode PETSCMAT_DLLEXPORT MatGetOwnershipRangesColumn(Mat mat,const PetscInt **ranges)
5226 {
5227   PetscErrorCode ierr;
5228 
5229   PetscFunctionBegin;
5230   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5231   PetscValidType(mat,1);
5232   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5233   ierr = PetscMapGetRanges(mat->cmap,ranges);CHKERRQ(ierr);
5234   PetscFunctionReturn(0);
5235 }
5236 
5237 #undef __FUNCT__
5238 #define __FUNCT__ "MatILUFactorSymbolic"
5239 /*@
5240    MatILUFactorSymbolic - Performs symbolic ILU factorization of a matrix.
5241    Uses levels of fill only, not drop tolerance. Use MatLUFactorNumeric()
5242    to complete the factorization.
5243 
5244    Collective on Mat
5245 
5246    Input Parameters:
5247 +  mat - the matrix
5248 .  row - row permutation
5249 .  column - column permutation
5250 -  info - structure containing
5251 $      levels - number of levels of fill.
5252 $      expected fill - as ratio of original fill.
5253 $      1 or 0 - indicating force fill on diagonal (improves robustness for matrices
5254                 missing diagonal entries)
5255 
5256    Output Parameters:
5257 .  fact - new matrix that has been symbolically factored
5258 
5259    Notes:
5260    See the users manual for additional information about
5261    choosing the fill factor for better efficiency.
5262 
5263    Most users should employ the simplified KSP interface for linear solvers
5264    instead of working directly with matrix algebra routines such as this.
5265    See, e.g., KSPCreate().
5266 
5267    Level: developer
5268 
5269   Concepts: matrices^symbolic LU factorization
5270   Concepts: matrices^factorization
5271   Concepts: LU^symbolic factorization
5272 
5273 .seealso: MatLUFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor()
5274           MatGetOrdering(), MatFactorInfo
5275 
5276 @*/
5277 PetscErrorCode PETSCMAT_DLLEXPORT MatILUFactorSymbolic(Mat fact,Mat mat,IS row,IS col,const MatFactorInfo *info)
5278 {
5279   PetscErrorCode ierr;
5280 
5281   PetscFunctionBegin;
5282   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5283   PetscValidType(mat,1);
5284   PetscValidHeaderSpecific(row,IS_COOKIE,2);
5285   PetscValidHeaderSpecific(col,IS_COOKIE,3);
5286   PetscValidPointer(info,4);
5287   PetscValidPointer(fact,5);
5288   if (info->levels < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Levels of fill negative %D",(PetscInt)info->levels);
5289   if (info->fill < 1.0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Expected fill less than 1.0 %G",info->fill);
5290   if (!(fact)->ops->ilufactorsymbolic) SETERRQ1(PETSC_ERR_SUP,"Matrix type %s  symbolic ILU",((PetscObject)mat)->type_name);
5291   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5292   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5293   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5294 
5295   ierr = PetscLogEventBegin(MAT_ILUFactorSymbolic,mat,row,col,0);CHKERRQ(ierr);
5296   ierr = (fact->ops->ilufactorsymbolic)(fact,mat,row,col,info);CHKERRQ(ierr);
5297   ierr = PetscLogEventEnd(MAT_ILUFactorSymbolic,mat,row,col,0);CHKERRQ(ierr);
5298   PetscFunctionReturn(0);
5299 }
5300 
5301 #undef __FUNCT__
5302 #define __FUNCT__ "MatICCFactorSymbolic"
5303 /*@
5304    MatICCFactorSymbolic - Performs symbolic incomplete
5305    Cholesky factorization for a symmetric matrix.  Use
5306    MatCholeskyFactorNumeric() to complete the factorization.
5307 
5308    Collective on Mat
5309 
5310    Input Parameters:
5311 +  mat - the matrix
5312 .  perm - row and column permutation
5313 -  info - structure containing
5314 $      levels - number of levels of fill.
5315 $      expected fill - as ratio of original fill.
5316 
5317    Output Parameter:
5318 .  fact - the factored matrix
5319 
5320    Notes:
5321    Most users should employ the KSP interface for linear solvers
5322    instead of working directly with matrix algebra routines such as this.
5323    See, e.g., KSPCreate().
5324 
5325    Level: developer
5326 
5327   Concepts: matrices^symbolic incomplete Cholesky factorization
5328   Concepts: matrices^factorization
5329   Concepts: Cholsky^symbolic factorization
5330 
5331 .seealso: MatCholeskyFactorNumeric(), MatCholeskyFactor(), MatFactorInfo
5332 @*/
5333 PetscErrorCode PETSCMAT_DLLEXPORT MatICCFactorSymbolic(Mat fact,Mat mat,IS perm,const MatFactorInfo *info)
5334 {
5335   PetscErrorCode ierr;
5336 
5337   PetscFunctionBegin;
5338   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5339   PetscValidType(mat,1);
5340   PetscValidHeaderSpecific(perm,IS_COOKIE,2);
5341   PetscValidPointer(info,3);
5342   PetscValidPointer(fact,4);
5343   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5344   if (info->levels < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Levels negative %D",(PetscInt) info->levels);
5345   if (info->fill < 1.0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Expected fill less than 1.0 %G",info->fill);
5346   if (!(fact)->ops->iccfactorsymbolic) SETERRQ1(PETSC_ERR_SUP,"Matrix type %s  symbolic ICC",((PetscObject)mat)->type_name);
5347   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5348   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5349 
5350   ierr = PetscLogEventBegin(MAT_ICCFactorSymbolic,mat,perm,0,0);CHKERRQ(ierr);
5351   ierr = (fact->ops->iccfactorsymbolic)(fact,mat,perm,info);CHKERRQ(ierr);
5352   ierr = PetscLogEventEnd(MAT_ICCFactorSymbolic,mat,perm,0,0);CHKERRQ(ierr);
5353   PetscFunctionReturn(0);
5354 }
5355 
5356 #undef __FUNCT__
5357 #define __FUNCT__ "MatILUDTFactorSymbolic"
5358 /*@
5359    MatILUDTFactorSymbolic - Performs symbolic pivoting drop-tolerance ILU factorization of a matrix.
5360    User provides the drop tolerance(dt) and the maximum nonzeros to be allowed per row(dtcount).
5361    Use MatILUDTFactorNumeric() to complete the factorization.
5362 
5363    Collective on Mat
5364 
5365    Input Parameters:
5366 +  mat - the matrix
5367 .  row - row permutation
5368 .  column - column permutation
5369 -  info - structure containing
5370 $      dt - drop tolerance.
5371 $      dtcount - maximum nonzeros to be allowed per row.
5372 
5373    Output Parameters:
5374 .  fact - factor matrix with memory preallocated
5375 
5376    Notes:
5377    See the ILUT algorithm written by Yousef Saad.
5378 
5379    Most users should employ the simplified KSP interface for linear solvers
5380    instead of working directly with matrix algebra routines such as this.
5381    See, e.g., KSPCreate().
5382 
5383    Level: developer
5384 
5385   Concepts: matrices^symbolic ILU factorization
5386   Concepts: matrices^factorization
5387 
5388 .seealso: MatILUDTFactorNumeric()
5389           MatGetOrdering(), MatGetFactor(), MatFactorInfo
5390 
5391 @*/
5392 PetscErrorCode PETSCMAT_DLLEXPORT MatILUDTFactorSymbolic(Mat fact,Mat mat,IS row,IS col,const MatFactorInfo *info)
5393 {
5394   PetscErrorCode ierr;
5395 
5396   PetscFunctionBegin;
5397   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5398   PetscValidType(mat,1);
5399   PetscValidHeaderSpecific(row,IS_COOKIE,2);
5400   PetscValidHeaderSpecific(col,IS_COOKIE,3);
5401   PetscValidPointer(info,4);
5402   PetscValidPointer(fact,5);
5403   if (info->dt < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"drop tolerance negative %G",(PetscInt)info->dt);
5404   if (info->dtcount < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"nonzeros per row %D <0",info->dtcount);
5405   if (!(fact)->ops->iludtfactorsymbolic) SETERRQ1(PETSC_ERR_SUP,"Matrix type %s  symbolic ILUDT",((PetscObject)mat)->type_name);
5406   if (mat->rmap->N != mat->cmap->N) SETERRQ(PETSC_ERR_ARG_WRONG,"matrix must be square");
5407   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5408   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5409   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5410 
5411   ierr = PetscLogEventBegin(MAT_ILUDTFactorSymbolic,mat,row,col,0);CHKERRQ(ierr);
5412   ierr = (fact->ops->iludtfactorsymbolic)(fact,mat,row,col,info);CHKERRQ(ierr);
5413   ierr = PetscLogEventEnd(MAT_ILUDTFactorSymbolic,mat,row,col,0);CHKERRQ(ierr);
5414   PetscFunctionReturn(0);
5415 }
5416 
5417 #undef __FUNCT__
5418 #define __FUNCT__ "MatILUDTFactorNumeric"
5419 /*@
5420    MatILUDTFactorNumeric - Performs numeric pivoting drop-tolerance ILU factorization of a matrix.
5421    Call this routine after first calling MatILUDTFactorSymbolic().
5422 
5423    Collective on Mat
5424 
5425    Input Parameters:
5426 +  fact - the factor matrix obtained with MatGetFactor()
5427 .  mat - the matrix
5428 -  info - options for factorization
5429 
5430    Output Parameters:
5431 .  fact - assembled factor matrix
5432 
5433    Notes:
5434    Most users should employ the simplified KSP interface for linear solvers
5435    instead of working directly with matrix algebra routines such as this.
5436    See, e.g., KSPCreate().
5437 
5438    Level: developer
5439 
5440 .seealso: MatILUDTFactorSymbolic()
5441 @*/
5442 PetscErrorCode PETSCMAT_DLLEXPORT MatILUDTFactorNumeric(Mat fact,Mat mat,const MatFactorInfo *info)
5443 {
5444   PetscErrorCode ierr;
5445 
5446   PetscFunctionBegin;
5447   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5448   PetscValidType(mat,1);
5449   PetscValidPointer(fact,2);
5450   PetscValidHeaderSpecific(fact,MAT_COOKIE,2);
5451   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5452   if (mat->rmap->N != (fact)->rmap->N || mat->cmap->N != (fact)->cmap->N) {
5453     SETERRQ4(PETSC_ERR_ARG_SIZ,"Mat mat,Mat fact: global dimensions are different %D should = %D %D should = %D",mat->rmap->N,(fact)->rmap->N,mat->cmap->N,(fact)->cmap->N);
5454   }
5455   if (!(fact)->ops->iludtfactornumeric) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5456   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5457   ierr = PetscLogEventBegin(MAT_ILUDTFactorNumeric,mat,fact,0,0);CHKERRQ(ierr);
5458   ierr = (fact->ops->iludtfactornumeric)(fact,mat,info);CHKERRQ(ierr);
5459   ierr = PetscLogEventEnd(MAT_ILUDTFactorNumeric,mat,fact,0,0);CHKERRQ(ierr);
5460 
5461   ierr = MatView_Private(fact);CHKERRQ(ierr);
5462   ierr = PetscObjectStateIncrease((PetscObject)fact);CHKERRQ(ierr);
5463   PetscFunctionReturn(0);
5464 }
5465 
5466 #undef __FUNCT__
5467 #define __FUNCT__ "MatGetArray"
5468 /*@C
5469    MatGetArray - Returns a pointer to the element values in the matrix.
5470    The result of this routine is dependent on the underlying matrix data
5471    structure, and may not even work for certain matrix types.  You MUST
5472    call MatRestoreArray() when you no longer need to access the array.
5473 
5474    Not Collective
5475 
5476    Input Parameter:
5477 .  mat - the matrix
5478 
5479    Output Parameter:
5480 .  v - the location of the values
5481 
5482 
5483    Fortran Note:
5484    This routine is used differently from Fortran, e.g.,
5485 .vb
5486         Mat         mat
5487         PetscScalar mat_array(1)
5488         PetscOffset i_mat
5489         PetscErrorCode ierr
5490         call MatGetArray(mat,mat_array,i_mat,ierr)
5491 
5492   C  Access first local entry in matrix; note that array is
5493   C  treated as one dimensional
5494         value = mat_array(i_mat + 1)
5495 
5496         [... other code ...]
5497         call MatRestoreArray(mat,mat_array,i_mat,ierr)
5498 .ve
5499 
5500    See the Fortran chapter of the users manual and
5501    petsc/src/mat/examples/tests for details.
5502 
5503    Level: advanced
5504 
5505    Concepts: matrices^access array
5506 
5507 .seealso: MatRestoreArray(), MatGetArrayF90(), MatGetRowIJ()
5508 @*/
5509 PetscErrorCode PETSCMAT_DLLEXPORT MatGetArray(Mat mat,PetscScalar *v[])
5510 {
5511   PetscErrorCode ierr;
5512 
5513   PetscFunctionBegin;
5514   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5515   PetscValidType(mat,1);
5516   PetscValidPointer(v,2);
5517   if (!mat->ops->getarray) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5518   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5519   ierr = (*mat->ops->getarray)(mat,v);CHKERRQ(ierr);
5520   CHKMEMQ;
5521   PetscFunctionReturn(0);
5522 }
5523 
5524 #undef __FUNCT__
5525 #define __FUNCT__ "MatRestoreArray"
5526 /*@C
5527    MatRestoreArray - Restores the matrix after MatGetArray() has been called.
5528 
5529    Not Collective
5530 
5531    Input Parameter:
5532 +  mat - the matrix
5533 -  v - the location of the values
5534 
5535    Fortran Note:
5536    This routine is used differently from Fortran, e.g.,
5537 .vb
5538         Mat         mat
5539         PetscScalar mat_array(1)
5540         PetscOffset i_mat
5541         PetscErrorCode ierr
5542         call MatGetArray(mat,mat_array,i_mat,ierr)
5543 
5544   C  Access first local entry in matrix; note that array is
5545   C  treated as one dimensional
5546         value = mat_array(i_mat + 1)
5547 
5548         [... other code ...]
5549         call MatRestoreArray(mat,mat_array,i_mat,ierr)
5550 .ve
5551 
5552    See the Fortran chapter of the users manual and
5553    petsc/src/mat/examples/tests for details
5554 
5555    Level: advanced
5556 
5557 .seealso: MatGetArray(), MatRestoreArrayF90()
5558 @*/
5559 PetscErrorCode PETSCMAT_DLLEXPORT MatRestoreArray(Mat mat,PetscScalar *v[])
5560 {
5561   PetscErrorCode ierr;
5562 
5563   PetscFunctionBegin;
5564   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5565   PetscValidType(mat,1);
5566   PetscValidPointer(v,2);
5567 #if defined(PETSC_USE_DEBUG)
5568   CHKMEMQ;
5569 #endif
5570   if (!mat->ops->restorearray) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5571   ierr = (*mat->ops->restorearray)(mat,v);CHKERRQ(ierr);
5572   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
5573   PetscFunctionReturn(0);
5574 }
5575 
5576 #undef __FUNCT__
5577 #define __FUNCT__ "MatGetSubMatrices"
5578 /*@C
5579    MatGetSubMatrices - Extracts several submatrices from a matrix. If submat
5580    points to an array of valid matrices, they may be reused to store the new
5581    submatrices.
5582 
5583    Collective on Mat
5584 
5585    Input Parameters:
5586 +  mat - the matrix
5587 .  n   - the number of submatrixes to be extracted (on this processor, may be zero)
5588 .  irow, icol - index sets of rows and columns to extract
5589 -  scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
5590 
5591    Output Parameter:
5592 .  submat - the array of submatrices
5593 
5594    Notes:
5595    MatGetSubMatrices() can extract ONLY sequential submatrices
5596    (from both sequential and parallel matrices). Use MatGetSubMatrix()
5597    to extract a parallel submatrix.
5598 
5599    When extracting submatrices from a parallel matrix, each processor can
5600    form a different submatrix by setting the rows and columns of its
5601    individual index sets according to the local submatrix desired.
5602 
5603    When finished using the submatrices, the user should destroy
5604    them with MatDestroyMatrices().
5605 
5606    MAT_REUSE_MATRIX can only be used when the nonzero structure of the
5607    original matrix has not changed from that last call to MatGetSubMatrices().
5608 
5609    This routine creates the matrices in submat; you should NOT create them before
5610    calling it. It also allocates the array of matrix pointers submat.
5611 
5612    For BAIJ matrices the index sets must respect the block structure, that is if they
5613    request one row/column in a block, they must request all rows/columns that are in
5614    that block. For example, if the block size is 2 you cannot request just row 0 and
5615    column 0.
5616 
5617    Fortran Note:
5618    The Fortran interface is slightly different from that given below; it
5619    requires one to pass in  as submat a Mat (integer) array of size at least m.
5620 
5621    Level: advanced
5622 
5623    Concepts: matrices^accessing submatrices
5624    Concepts: submatrices
5625 
5626 .seealso: MatDestroyMatrices(), MatGetSubMatrix(), MatGetRow(), MatGetDiagonal(), MatReuse
5627 @*/
5628 PetscErrorCode PETSCMAT_DLLEXPORT MatGetSubMatrices(Mat mat,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *submat[])
5629 {
5630   PetscErrorCode ierr;
5631   PetscInt        i;
5632   PetscTruth      eq;
5633 
5634   PetscFunctionBegin;
5635   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5636   PetscValidType(mat,1);
5637   if (n) {
5638     PetscValidPointer(irow,3);
5639     PetscValidHeaderSpecific(*irow,IS_COOKIE,3);
5640     PetscValidPointer(icol,4);
5641     PetscValidHeaderSpecific(*icol,IS_COOKIE,4);
5642   }
5643   PetscValidPointer(submat,6);
5644   if (n && scall == MAT_REUSE_MATRIX) {
5645     PetscValidPointer(*submat,6);
5646     PetscValidHeaderSpecific(**submat,MAT_COOKIE,6);
5647   }
5648   if (!mat->ops->getsubmatrices) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5649   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5650   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5651   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5652 
5653   ierr = PetscLogEventBegin(MAT_GetSubMatrices,mat,0,0,0);CHKERRQ(ierr);
5654   ierr = (*mat->ops->getsubmatrices)(mat,n,irow,icol,scall,submat);CHKERRQ(ierr);
5655   ierr = PetscLogEventEnd(MAT_GetSubMatrices,mat,0,0,0);CHKERRQ(ierr);
5656   for (i=0; i<n; i++) {
5657     if (mat->symmetric || mat->structurally_symmetric || mat->hermitian) {
5658       ierr = ISEqual(irow[i],icol[i],&eq);CHKERRQ(ierr);
5659       if (eq) {
5660 	if (mat->symmetric){
5661 	  ierr = MatSetOption((*submat)[i],MAT_SYMMETRIC,PETSC_TRUE);CHKERRQ(ierr);
5662 	} else if (mat->hermitian) {
5663 	  ierr = MatSetOption((*submat)[i],MAT_HERMITIAN,PETSC_TRUE);CHKERRQ(ierr);
5664 	} else if (mat->structurally_symmetric) {
5665 	  ierr = MatSetOption((*submat)[i],MAT_STRUCTURALLY_SYMMETRIC,PETSC_TRUE);CHKERRQ(ierr);
5666 	}
5667       }
5668     }
5669   }
5670   PetscFunctionReturn(0);
5671 }
5672 
5673 #undef __FUNCT__
5674 #define __FUNCT__ "MatDestroyMatrices"
5675 /*@C
5676    MatDestroyMatrices - Destroys a set of matrices obtained with MatGetSubMatrices().
5677 
5678    Collective on Mat
5679 
5680    Input Parameters:
5681 +  n - the number of local matrices
5682 -  mat - the matrices (note that this is a pointer to the array of matrices, just to match the calling
5683                        sequence of MatGetSubMatrices())
5684 
5685    Level: advanced
5686 
5687     Notes: Frees not only the matrices, but also the array that contains the matrices
5688            In Fortran will not free the array.
5689 
5690 .seealso: MatGetSubMatrices()
5691 @*/
5692 PetscErrorCode PETSCMAT_DLLEXPORT MatDestroyMatrices(PetscInt n,Mat *mat[])
5693 {
5694   PetscErrorCode ierr;
5695   PetscInt       i;
5696 
5697   PetscFunctionBegin;
5698   if (n < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Trying to destroy negative number of matrices %D",n);
5699   PetscValidPointer(mat,2);
5700   for (i=0; i<n; i++) {
5701     ierr = MatDestroy((*mat)[i]);CHKERRQ(ierr);
5702   }
5703   /* memory is allocated even if n = 0 */
5704   ierr = PetscFree(*mat);CHKERRQ(ierr);
5705   PetscFunctionReturn(0);
5706 }
5707 
5708 #undef __FUNCT__
5709 #define __FUNCT__ "MatGetSeqNonzeroStructure"
5710 /*@C
5711    MatGetSeqNonzeroStructure - Extracts the sequential nonzero structure from a matrix.
5712 
5713    Collective on Mat
5714 
5715    Input Parameters:
5716 .  mat - the matrix
5717 
5718    Output Parameter:
5719 .  matstruct - the sequential matrix with the nonzero structure of mat
5720 
5721   Level: intermediate
5722 
5723 .seealso: MatDestroySeqNonzeroStructure(), MatGetSubMatrices(), MatDestroyMatrices()
5724 @*/
5725 PetscErrorCode PETSCMAT_DLLEXPORT MatGetSeqNonzeroStructure(Mat mat,Mat *matstruct[])
5726 {
5727   PetscErrorCode ierr;
5728 
5729   PetscFunctionBegin;
5730   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5731   PetscValidPointer(matstruct,2);
5732 
5733   PetscValidType(mat,1);
5734   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5735   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5736 
5737   ierr = PetscLogEventBegin(MAT_GetSeqNonzeroStructure,mat,0,0,0);CHKERRQ(ierr);
5738   ierr = (*mat->ops->getseqnonzerostructure)(mat,matstruct);CHKERRQ(ierr);
5739   ierr = PetscLogEventEnd(MAT_GetSeqNonzeroStructure,mat,0,0,0);CHKERRQ(ierr);
5740   PetscFunctionReturn(0);
5741 }
5742 
5743 #undef __FUNCT__
5744 #define __FUNCT__ "MatDestroySeqNonzeroStructure"
5745 /*@C
5746    MatDestroySeqNonzeroStructure - Destroys matrix obtained with MatGetSeqNonzeroStructure().
5747 
5748    Collective on Mat
5749 
5750    Input Parameters:
5751 .  mat - the matrix (note that this is a pointer to the array of matrices, just to match the calling
5752                        sequence of MatGetSequentialNonzeroStructure())
5753 
5754    Level: advanced
5755 
5756     Notes: Frees not only the matrices, but also the array that contains the matrices
5757 
5758 .seealso: MatGetSeqNonzeroStructure()
5759 @*/
5760 PetscErrorCode PETSCMAT_DLLEXPORT MatDestroySeqNonzeroStructure(Mat *mat[])
5761 {
5762   PetscErrorCode ierr;
5763 
5764   PetscFunctionBegin;
5765   PetscValidPointer(mat,1);
5766   ierr = MatDestroyMatrices(1,mat);CHKERRQ(ierr);
5767   PetscFunctionReturn(0);
5768 }
5769 
5770 #undef __FUNCT__
5771 #define __FUNCT__ "MatIncreaseOverlap"
5772 /*@
5773    MatIncreaseOverlap - Given a set of submatrices indicated by index sets,
5774    replaces the index sets by larger ones that represent submatrices with
5775    additional overlap.
5776 
5777    Collective on Mat
5778 
5779    Input Parameters:
5780 +  mat - the matrix
5781 .  n   - the number of index sets
5782 .  is  - the array of index sets (these index sets will changed during the call)
5783 -  ov  - the additional overlap requested
5784 
5785    Level: developer
5786 
5787    Concepts: overlap
5788    Concepts: ASM^computing overlap
5789 
5790 .seealso: MatGetSubMatrices()
5791 @*/
5792 PetscErrorCode PETSCMAT_DLLEXPORT MatIncreaseOverlap(Mat mat,PetscInt n,IS is[],PetscInt ov)
5793 {
5794   PetscErrorCode ierr;
5795 
5796   PetscFunctionBegin;
5797   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5798   PetscValidType(mat,1);
5799   if (n < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Must have one or more domains, you have %D",n);
5800   if (n) {
5801     PetscValidPointer(is,3);
5802     PetscValidHeaderSpecific(*is,IS_COOKIE,3);
5803   }
5804   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
5805   if (mat->factor)     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
5806   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5807 
5808   if (!ov) PetscFunctionReturn(0);
5809   if (!mat->ops->increaseoverlap) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
5810   ierr = PetscLogEventBegin(MAT_IncreaseOverlap,mat,0,0,0);CHKERRQ(ierr);
5811   ierr = (*mat->ops->increaseoverlap)(mat,n,is,ov);CHKERRQ(ierr);
5812   ierr = PetscLogEventEnd(MAT_IncreaseOverlap,mat,0,0,0);CHKERRQ(ierr);
5813   PetscFunctionReturn(0);
5814 }
5815 
5816 #undef __FUNCT__
5817 #define __FUNCT__ "MatGetBlockSize"
5818 /*@
5819    MatGetBlockSize - Returns the matrix block size; useful especially for the
5820    block row and block diagonal formats.
5821 
5822    Not Collective
5823 
5824    Input Parameter:
5825 .  mat - the matrix
5826 
5827    Output Parameter:
5828 .  bs - block size
5829 
5830    Notes:
5831    Block row formats are MATSEQBAIJ, MATMPIBAIJ, MATSEQSBAIJ, MATMPISBAIJ
5832 
5833    Level: intermediate
5834 
5835    Concepts: matrices^block size
5836 
5837 .seealso: MatCreateSeqBAIJ(), MatCreateMPIBAIJ()
5838 @*/
5839 PetscErrorCode PETSCMAT_DLLEXPORT MatGetBlockSize(Mat mat,PetscInt *bs)
5840 {
5841   PetscErrorCode ierr;
5842 
5843   PetscFunctionBegin;
5844   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5845   PetscValidType(mat,1);
5846   PetscValidIntPointer(bs,2);
5847   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5848   *bs = mat->rmap->bs;
5849   PetscFunctionReturn(0);
5850 }
5851 
5852 #undef __FUNCT__
5853 #define __FUNCT__ "MatSetBlockSize"
5854 /*@
5855    MatSetBlockSize - Sets the matrix block size; for many matrix types you
5856      cannot use this and MUST set the blocksize when you preallocate the matrix
5857 
5858    Collective on Mat
5859 
5860    Input Parameters:
5861 +  mat - the matrix
5862 -  bs - block size
5863 
5864    Notes:
5865      Only works for shell and AIJ matrices
5866 
5867    Level: intermediate
5868 
5869    Concepts: matrices^block size
5870 
5871 .seealso: MatCreateSeqBAIJ(), MatCreateMPIBAIJ(), MatGetBlockSize()
5872 @*/
5873 PetscErrorCode PETSCMAT_DLLEXPORT MatSetBlockSize(Mat mat,PetscInt bs)
5874 {
5875   PetscErrorCode ierr;
5876 
5877   PetscFunctionBegin;
5878   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5879   PetscValidType(mat,1);
5880   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5881   if (mat->ops->setblocksize) {
5882     /* XXX should check if (bs < 1) ??? */
5883     ierr = PetscMapSetBlockSize(mat->rmap,bs);CHKERRQ(ierr);
5884     ierr = PetscMapSetBlockSize(mat->cmap,bs);CHKERRQ(ierr);
5885     ierr = (*mat->ops->setblocksize)(mat,bs);CHKERRQ(ierr);
5886   } else {
5887     SETERRQ1(PETSC_ERR_ARG_INCOMP,"Cannot set the blocksize for matrix type %s",((PetscObject)mat)->type_name);
5888   }
5889   PetscFunctionReturn(0);
5890 }
5891 
5892 #undef __FUNCT__
5893 #define __FUNCT__ "MatGetRowIJ"
5894 /*@C
5895     MatGetRowIJ - Returns the compressed row storage i and j indices for sequential matrices.
5896 
5897    Collective on Mat
5898 
5899     Input Parameters:
5900 +   mat - the matrix
5901 .   shift -  0 or 1 indicating we want the indices starting at 0 or 1
5902 .   symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
5903                 symmetrized
5904 -   blockcompressed - PETSC_TRUE or PETSC_FALSE  indicating if the nonzero structure of the
5905                  blockcompressed matrix is desired or not [inode, baij have blockcompressed
5906                  nonzero structure which is different than the full nonzero structure]
5907 
5908     Output Parameters:
5909 +   n - number of rows in the (possibly compressed) matrix
5910 .   ia - the row pointers [of length n+1]
5911 .   ja - the column indices
5912 -   done - indicates if the routine actually worked and returned appropriate ia[] and ja[] arrays; callers
5913            are responsible for handling the case when done == PETSC_FALSE and ia and ja are not set
5914 
5915     Level: developer
5916 
5917     Notes: You CANNOT change any of the ia[] or ja[] values.
5918 
5919            Use MatRestoreRowIJ() when you are finished accessing the ia[] and ja[] values
5920 
5921     Fortran Node
5922 
5923            In Fortran use
5924 $           PetscInt ia(1), ja(1)
5925 $           PetscOffset iia, jja
5926 $      call MatGetRowIJ(mat,shift,symmetric,blockcompressed,n,ia,iia,ja,jja,done,ierr)
5927 
5928        Acess the ith and jth entries via ia(iia + i) and ja(jja + j)
5929 
5930 .seealso: MatGetColumnIJ(), MatRestoreRowIJ(), MatGetArray()
5931 @*/
5932 PetscErrorCode PETSCMAT_DLLEXPORT MatGetRowIJ(Mat mat,PetscInt shift,PetscTruth symmetric,PetscTruth blockcompressed,PetscInt *n,PetscInt *ia[],PetscInt* ja[],PetscTruth *done)
5933 {
5934   PetscErrorCode ierr;
5935 
5936   PetscFunctionBegin;
5937   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5938   PetscValidType(mat,1);
5939   PetscValidIntPointer(n,4);
5940   if (ia) PetscValidIntPointer(ia,5);
5941   if (ja) PetscValidIntPointer(ja,6);
5942   PetscValidIntPointer(done,7);
5943   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5944   if (!mat->ops->getrowij) *done = PETSC_FALSE;
5945   else {
5946     *done = PETSC_TRUE;
5947     ierr = PetscLogEventBegin(MAT_GetRowIJ,mat,0,0,0);CHKERRQ(ierr);
5948     ierr  = (*mat->ops->getrowij)(mat,shift,symmetric,blockcompressed,n,ia,ja,done);CHKERRQ(ierr);
5949     ierr = PetscLogEventEnd(MAT_GetRowIJ,mat,0,0,0);CHKERRQ(ierr);
5950   }
5951   PetscFunctionReturn(0);
5952 }
5953 
5954 #undef __FUNCT__
5955 #define __FUNCT__ "MatGetColumnIJ"
5956 /*@C
5957     MatGetColumnIJ - Returns the compressed column storage i and j indices for sequential matrices.
5958 
5959     Collective on Mat
5960 
5961     Input Parameters:
5962 +   mat - the matrix
5963 .   shift - 1 or zero indicating we want the indices starting at 0 or 1
5964 .   symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
5965                 symmetrized
5966 -   blockcompressed - PETSC_TRUE or PETSC_FALSE  indicating if the nonzero structure of the
5967                  blockcompressed matrix is desired or not [inode, baij have blockcompressed
5968                  nonzero structure which is different than the full nonzero structure]
5969 
5970     Output Parameters:
5971 +   n - number of columns in the (possibly compressed) matrix
5972 .   ia - the column pointers
5973 .   ja - the row indices
5974 -   done - PETSC_TRUE or PETSC_FALSE, indicating whether the values have been returned
5975 
5976     Level: developer
5977 
5978 .seealso: MatGetRowIJ(), MatRestoreColumnIJ()
5979 @*/
5980 PetscErrorCode PETSCMAT_DLLEXPORT MatGetColumnIJ(Mat mat,PetscInt shift,PetscTruth symmetric,PetscTruth blockcompressed,PetscInt *n,PetscInt *ia[],PetscInt* ja[],PetscTruth *done)
5981 {
5982   PetscErrorCode ierr;
5983 
5984   PetscFunctionBegin;
5985   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
5986   PetscValidType(mat,1);
5987   PetscValidIntPointer(n,4);
5988   if (ia) PetscValidIntPointer(ia,5);
5989   if (ja) PetscValidIntPointer(ja,6);
5990   PetscValidIntPointer(done,7);
5991   ierr = MatPreallocated(mat);CHKERRQ(ierr);
5992   if (!mat->ops->getcolumnij) *done = PETSC_FALSE;
5993   else {
5994     *done = PETSC_TRUE;
5995     ierr  = (*mat->ops->getcolumnij)(mat,shift,symmetric,blockcompressed,n,ia,ja,done);CHKERRQ(ierr);
5996   }
5997   PetscFunctionReturn(0);
5998 }
5999 
6000 #undef __FUNCT__
6001 #define __FUNCT__ "MatRestoreRowIJ"
6002 /*@C
6003     MatRestoreRowIJ - Call after you are completed with the ia,ja indices obtained with
6004     MatGetRowIJ().
6005 
6006     Collective on Mat
6007 
6008     Input Parameters:
6009 +   mat - the matrix
6010 .   shift - 1 or zero indicating we want the indices starting at 0 or 1
6011 .   symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
6012                 symmetrized
6013 -   blockcompressed - PETSC_TRUE or PETSC_FALSE  indicating if the nonzero structure of the
6014                  blockcompressed matrix is desired or not [inode, baij have blockcompressed
6015                  nonzero structure which is different than the full nonzero structure]
6016 
6017     Output Parameters:
6018 +   n - size of (possibly compressed) matrix
6019 .   ia - the row pointers
6020 .   ja - the column indices
6021 -   done - PETSC_TRUE or PETSC_FALSE indicated that the values have been returned
6022 
6023     Level: developer
6024 
6025 .seealso: MatGetRowIJ(), MatRestoreColumnIJ()
6026 @*/
6027 PetscErrorCode PETSCMAT_DLLEXPORT MatRestoreRowIJ(Mat mat,PetscInt shift,PetscTruth symmetric,PetscTruth blockcompressed,PetscInt *n,PetscInt *ia[],PetscInt* ja[],PetscTruth *done)
6028 {
6029   PetscErrorCode ierr;
6030 
6031   PetscFunctionBegin;
6032   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6033   PetscValidType(mat,1);
6034   if (ia) PetscValidIntPointer(ia,5);
6035   if (ja) PetscValidIntPointer(ja,6);
6036   PetscValidIntPointer(done,7);
6037   ierr = MatPreallocated(mat);CHKERRQ(ierr);
6038 
6039   if (!mat->ops->restorerowij) *done = PETSC_FALSE;
6040   else {
6041     *done = PETSC_TRUE;
6042     ierr  = (*mat->ops->restorerowij)(mat,shift,symmetric,blockcompressed,n,ia,ja,done);CHKERRQ(ierr);
6043   }
6044   PetscFunctionReturn(0);
6045 }
6046 
6047 #undef __FUNCT__
6048 #define __FUNCT__ "MatRestoreColumnIJ"
6049 /*@C
6050     MatRestoreColumnIJ - Call after you are completed with the ia,ja indices obtained with
6051     MatGetColumnIJ().
6052 
6053     Collective on Mat
6054 
6055     Input Parameters:
6056 +   mat - the matrix
6057 .   shift - 1 or zero indicating we want the indices starting at 0 or 1
6058 -   symmetric - PETSC_TRUE or PETSC_FALSE indicating the matrix data structure should be
6059                 symmetrized
6060 -   blockcompressed - PETSC_TRUE or PETSC_FALSE  indicating if the nonzero structure of the
6061                  blockcompressed matrix is desired or not [inode, baij have blockcompressed
6062                  nonzero structure which is different than the full nonzero structure]
6063 
6064     Output Parameters:
6065 +   n - size of (possibly compressed) matrix
6066 .   ia - the column pointers
6067 .   ja - the row indices
6068 -   done - PETSC_TRUE or PETSC_FALSE indicated that the values have been returned
6069 
6070     Level: developer
6071 
6072 .seealso: MatGetColumnIJ(), MatRestoreRowIJ()
6073 @*/
6074 PetscErrorCode PETSCMAT_DLLEXPORT MatRestoreColumnIJ(Mat mat,PetscInt shift,PetscTruth symmetric,PetscTruth blockcompressed,PetscInt *n,PetscInt *ia[],PetscInt* ja[],PetscTruth *done)
6075 {
6076   PetscErrorCode ierr;
6077 
6078   PetscFunctionBegin;
6079   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6080   PetscValidType(mat,1);
6081   if (ia) PetscValidIntPointer(ia,5);
6082   if (ja) PetscValidIntPointer(ja,6);
6083   PetscValidIntPointer(done,7);
6084   ierr = MatPreallocated(mat);CHKERRQ(ierr);
6085 
6086   if (!mat->ops->restorecolumnij) *done = PETSC_FALSE;
6087   else {
6088     *done = PETSC_TRUE;
6089     ierr  = (*mat->ops->restorecolumnij)(mat,shift,symmetric,blockcompressed,n,ia,ja,done);CHKERRQ(ierr);
6090   }
6091   PetscFunctionReturn(0);
6092 }
6093 
6094 #undef __FUNCT__
6095 #define __FUNCT__ "MatColoringPatch"
6096 /*@C
6097     MatColoringPatch -Used inside matrix coloring routines that
6098     use MatGetRowIJ() and/or MatGetColumnIJ().
6099 
6100     Collective on Mat
6101 
6102     Input Parameters:
6103 +   mat - the matrix
6104 .   ncolors - max color value
6105 .   n   - number of entries in colorarray
6106 -   colorarray - array indicating color for each column
6107 
6108     Output Parameters:
6109 .   iscoloring - coloring generated using colorarray information
6110 
6111     Level: developer
6112 
6113 .seealso: MatGetRowIJ(), MatGetColumnIJ()
6114 
6115 @*/
6116 PetscErrorCode PETSCMAT_DLLEXPORT MatColoringPatch(Mat mat,PetscInt ncolors,PetscInt n,ISColoringValue colorarray[],ISColoring *iscoloring)
6117 {
6118   PetscErrorCode ierr;
6119 
6120   PetscFunctionBegin;
6121   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6122   PetscValidType(mat,1);
6123   PetscValidIntPointer(colorarray,4);
6124   PetscValidPointer(iscoloring,5);
6125   ierr = MatPreallocated(mat);CHKERRQ(ierr);
6126 
6127   if (!mat->ops->coloringpatch){
6128     ierr = ISColoringCreate(((PetscObject)mat)->comm,ncolors,n,colorarray,iscoloring);CHKERRQ(ierr);
6129   } else {
6130     ierr = (*mat->ops->coloringpatch)(mat,ncolors,n,colorarray,iscoloring);CHKERRQ(ierr);
6131   }
6132   PetscFunctionReturn(0);
6133 }
6134 
6135 
6136 #undef __FUNCT__
6137 #define __FUNCT__ "MatSetUnfactored"
6138 /*@
6139    MatSetUnfactored - Resets a factored matrix to be treated as unfactored.
6140 
6141    Collective on Mat
6142 
6143    Input Parameter:
6144 .  mat - the factored matrix to be reset
6145 
6146    Notes:
6147    This routine should be used only with factored matrices formed by in-place
6148    factorization via ILU(0) (or by in-place LU factorization for the MATSEQDENSE
6149    format).  This option can save memory, for example, when solving nonlinear
6150    systems with a matrix-free Newton-Krylov method and a matrix-based, in-place
6151    ILU(0) preconditioner.
6152 
6153    Note that one can specify in-place ILU(0) factorization by calling
6154 .vb
6155      PCType(pc,PCILU);
6156      PCFactorSeUseInPlace(pc);
6157 .ve
6158    or by using the options -pc_type ilu -pc_factor_in_place
6159 
6160    In-place factorization ILU(0) can also be used as a local
6161    solver for the blocks within the block Jacobi or additive Schwarz
6162    methods (runtime option: -sub_pc_factor_in_place).  See the discussion
6163    of these preconditioners in the users manual for details on setting
6164    local solver options.
6165 
6166    Most users should employ the simplified KSP interface for linear solvers
6167    instead of working directly with matrix algebra routines such as this.
6168    See, e.g., KSPCreate().
6169 
6170    Level: developer
6171 
6172 .seealso: PCFactorSetUseInPlace()
6173 
6174    Concepts: matrices^unfactored
6175 
6176 @*/
6177 PetscErrorCode PETSCMAT_DLLEXPORT MatSetUnfactored(Mat mat)
6178 {
6179   PetscErrorCode ierr;
6180 
6181   PetscFunctionBegin;
6182   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6183   PetscValidType(mat,1);
6184   ierr = MatPreallocated(mat);CHKERRQ(ierr);
6185   mat->factor = MAT_FACTOR_NONE;
6186   if (!mat->ops->setunfactored) PetscFunctionReturn(0);
6187   ierr = (*mat->ops->setunfactored)(mat);CHKERRQ(ierr);
6188   PetscFunctionReturn(0);
6189 }
6190 
6191 /*MC
6192     MatGetArrayF90 - Accesses a matrix array from Fortran90.
6193 
6194     Synopsis:
6195     MatGetArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
6196 
6197     Not collective
6198 
6199     Input Parameter:
6200 .   x - matrix
6201 
6202     Output Parameters:
6203 +   xx_v - the Fortran90 pointer to the array
6204 -   ierr - error code
6205 
6206     Example of Usage:
6207 .vb
6208       PetscScalar, pointer xx_v(:)
6209       ....
6210       call MatGetArrayF90(x,xx_v,ierr)
6211       a = xx_v(3)
6212       call MatRestoreArrayF90(x,xx_v,ierr)
6213 .ve
6214 
6215     Notes:
6216     Not yet supported for all F90 compilers
6217 
6218     Level: advanced
6219 
6220 .seealso:  MatRestoreArrayF90(), MatGetArray(), MatRestoreArray()
6221 
6222     Concepts: matrices^accessing array
6223 
6224 M*/
6225 
6226 /*MC
6227     MatRestoreArrayF90 - Restores a matrix array that has been
6228     accessed with MatGetArrayF90().
6229 
6230     Synopsis:
6231     MatRestoreArrayF90(Mat x,{Scalar, pointer :: xx_v(:)},integer ierr)
6232 
6233     Not collective
6234 
6235     Input Parameters:
6236 +   x - matrix
6237 -   xx_v - the Fortran90 pointer to the array
6238 
6239     Output Parameter:
6240 .   ierr - error code
6241 
6242     Example of Usage:
6243 .vb
6244        PetscScalar, pointer xx_v(:)
6245        ....
6246        call MatGetArrayF90(x,xx_v,ierr)
6247        a = xx_v(3)
6248        call MatRestoreArrayF90(x,xx_v,ierr)
6249 .ve
6250 
6251     Notes:
6252     Not yet supported for all F90 compilers
6253 
6254     Level: advanced
6255 
6256 .seealso:  MatGetArrayF90(), MatGetArray(), MatRestoreArray()
6257 
6258 M*/
6259 
6260 
6261 #undef __FUNCT__
6262 #define __FUNCT__ "MatGetSubMatrix"
6263 /*@
6264     MatGetSubMatrix - Gets a single submatrix on the same number of processors
6265                       as the original matrix.
6266 
6267     Collective on Mat
6268 
6269     Input Parameters:
6270 +   mat - the original matrix
6271 .   isrow - rows this processor should obtain
6272 .   iscol - columns for all processors you wish to keep
6273 .   csize - number of columns "local" to this processor (does nothing for sequential
6274             matrices). This should match the result from VecGetLocalSize(x,...) if you
6275             plan to use the matrix in a A*x; alternatively, you can use PETSC_DECIDE
6276 -   cll - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
6277 
6278     Output Parameter:
6279 .   newmat - the new submatrix, of the same type as the old
6280 
6281     Level: advanced
6282 
6283     Notes: the iscol argument MUST be the same on each processor. You might be
6284     able to create the iscol argument with ISAllGather(). The rows is isrow will be
6285     sorted into the same order as the original matrix.
6286 
6287       The first time this is called you should use a cll of MAT_INITIAL_MATRIX,
6288    the MatGetSubMatrix() routine will create the newmat for you. Any additional calls
6289    to this routine with a mat of the same nonzero structure and with a call of MAT_REUSE_MATRIX
6290    will reuse the matrix generated the first time.  You should call MatDestroy() on newmat when
6291    you are finished using it.
6292 
6293     The communicator of the newly obtained matrix is ALWAYS the same as the communicator of
6294     the input matrix.
6295 
6296     If iscol is PETSC_NULL then all columns are obtained (not supported in Fortran), you should
6297     use csize = PETSC_DECIDE also in this case.
6298 
6299     Concepts: matrices^submatrices
6300 
6301 .seealso: MatGetSubMatrices(), ISAllGather()
6302 @*/
6303 PetscErrorCode PETSCMAT_DLLEXPORT MatGetSubMatrix(Mat mat,IS isrow,IS iscol,PetscInt csize,MatReuse cll,Mat *newmat)
6304 {
6305   PetscErrorCode ierr;
6306   PetscMPIInt    size;
6307   Mat            *local;
6308   IS             iscoltmp;
6309 
6310   PetscFunctionBegin;
6311   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6312   PetscValidHeaderSpecific(isrow,IS_COOKIE,2);
6313   if (iscol) PetscValidHeaderSpecific(iscol,IS_COOKIE,3);
6314   PetscValidPointer(newmat,6);
6315   if (cll == MAT_REUSE_MATRIX) PetscValidHeaderSpecific(*newmat,MAT_COOKIE,6);
6316   PetscValidType(mat,1);
6317   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6318   ierr = MatPreallocated(mat);CHKERRQ(ierr);
6319   ierr = MPI_Comm_size(((PetscObject)mat)->comm,&size);CHKERRQ(ierr);
6320 
6321   if (!iscol) {
6322     if (csize == PETSC_DECIDE) csize = mat->cmap->n;
6323     ierr = ISCreateStride(((PetscObject)mat)->comm,mat->cmap->N,0,1,&iscoltmp);CHKERRQ(ierr);
6324   } else {
6325     iscoltmp = iscol;
6326   }
6327 
6328   /* if original matrix is on just one processor then use submatrix generated */
6329   if (!mat->ops->getsubmatrix && size == 1 && cll == MAT_REUSE_MATRIX) {
6330     ierr = MatGetSubMatrices(mat,1,&isrow,&iscoltmp,MAT_REUSE_MATRIX,&newmat);CHKERRQ(ierr);
6331     if (!iscol) {ierr = ISDestroy(iscoltmp);CHKERRQ(ierr);}
6332     PetscFunctionReturn(0);
6333   } else if (!mat->ops->getsubmatrix && size == 1) {
6334     ierr    = MatGetSubMatrices(mat,1,&isrow,&iscoltmp,MAT_INITIAL_MATRIX,&local);CHKERRQ(ierr);
6335     *newmat = *local;
6336     ierr    = PetscFree(local);CHKERRQ(ierr);
6337     if (!iscol) {ierr = ISDestroy(iscoltmp);CHKERRQ(ierr);}
6338     PetscFunctionReturn(0);
6339   }
6340 
6341   if (!mat->ops->getsubmatrix) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6342   ierr = (*mat->ops->getsubmatrix)(mat,isrow,iscoltmp,csize,cll,newmat);CHKERRQ(ierr);
6343   if (!iscol) {ierr = ISDestroy(iscoltmp);CHKERRQ(ierr);}
6344   ierr = PetscObjectStateIncrease((PetscObject)*newmat);CHKERRQ(ierr);
6345   PetscFunctionReturn(0);
6346 }
6347 
6348 #undef __FUNCT__
6349 #define __FUNCT__ "MatGetSubMatrixRaw"
6350 /*@
6351     MatGetSubMatrixRaw - Gets a single submatrix on the same number of processors
6352                          as the original matrix.
6353 
6354     Collective on Mat
6355 
6356     Input Parameters:
6357 +   mat - the original matrix
6358 .   nrows - the number of rows this processor should obtain
6359 .   rows - rows this processor should obtain
6360 .   ncols - the number of columns for all processors you wish to keep
6361 .   cols - columns for all processors you wish to keep
6362 .   csize - number of columns "local" to this processor (does nothing for sequential
6363             matrices). This should match the result from VecGetLocalSize(x,...) if you
6364             plan to use the matrix in a A*x; alternatively, you can use PETSC_DECIDE
6365 -   cll - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
6366 
6367     Output Parameter:
6368 .   newmat - the new submatrix, of the same type as the old
6369 
6370     Level: advanced
6371 
6372     Notes: the iscol argument MUST be the same on each processor. You might be
6373     able to create the iscol argument with ISAllGather().
6374 
6375       The first time this is called you should use a cll of MAT_INITIAL_MATRIX,
6376    the MatGetSubMatrix() routine will create the newmat for you. Any additional calls
6377    to this routine with a mat of the same nonzero structure and with a cll of MAT_REUSE_MATRIX
6378    will reuse the matrix generated the first time.
6379 
6380     Concepts: matrices^submatrices
6381 
6382 .seealso: MatGetSubMatrices(), ISAllGather()
6383 @*/
6384 PetscErrorCode PETSCMAT_DLLEXPORT MatGetSubMatrixRaw(Mat mat,PetscInt nrows,const PetscInt rows[],PetscInt ncols,const PetscInt cols[],PetscInt csize,MatReuse cll,Mat *newmat)
6385 {
6386   IS             isrow, iscol;
6387   PetscErrorCode ierr;
6388 
6389   PetscFunctionBegin;
6390   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6391   PetscValidIntPointer(rows,2);
6392   PetscValidIntPointer(cols,3);
6393   PetscValidPointer(newmat,6);
6394   if (cll == MAT_REUSE_MATRIX) PetscValidHeaderSpecific(*newmat,MAT_COOKIE,6);
6395   PetscValidType(mat,1);
6396   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6397   ierr = MatPreallocated(mat);CHKERRQ(ierr);
6398   ierr = ISCreateGeneralWithArray(PETSC_COMM_SELF, nrows, (PetscInt *) rows, &isrow);CHKERRQ(ierr);
6399   ierr = ISCreateGeneralWithArray(PETSC_COMM_SELF, ncols, (PetscInt *) cols, &iscol);CHKERRQ(ierr);
6400   ierr = MatGetSubMatrix(mat, isrow, iscol, csize, cll, newmat);CHKERRQ(ierr);
6401   ierr = ISDestroy(isrow);CHKERRQ(ierr);
6402   ierr = ISDestroy(iscol);CHKERRQ(ierr);
6403   PetscFunctionReturn(0);
6404 }
6405 
6406 #undef __FUNCT__
6407 #define __FUNCT__ "MatStashSetInitialSize"
6408 /*@
6409    MatStashSetInitialSize - sets the sizes of the matrix stash, that is
6410    used during the assembly process to store values that belong to
6411    other processors.
6412 
6413    Not Collective
6414 
6415    Input Parameters:
6416 +  mat   - the matrix
6417 .  size  - the initial size of the stash.
6418 -  bsize - the initial size of the block-stash(if used).
6419 
6420    Options Database Keys:
6421 +   -matstash_initial_size <size> or <size0,size1,...sizep-1>
6422 -   -matstash_block_initial_size <bsize>  or <bsize0,bsize1,...bsizep-1>
6423 
6424    Level: intermediate
6425 
6426    Notes:
6427      The block-stash is used for values set with MatSetValuesBlocked() while
6428      the stash is used for values set with MatSetValues()
6429 
6430      Run with the option -info and look for output of the form
6431      MatAssemblyBegin_MPIXXX:Stash has MM entries, uses nn mallocs.
6432      to determine the appropriate value, MM, to use for size and
6433      MatAssemblyBegin_MPIXXX:Block-Stash has BMM entries, uses nn mallocs.
6434      to determine the value, BMM to use for bsize
6435 
6436    Concepts: stash^setting matrix size
6437    Concepts: matrices^stash
6438 
6439 .seealso: MatAssemblyBegin(), MatAssemblyEnd(), Mat, MatStashGetInfo()
6440 
6441 @*/
6442 PetscErrorCode PETSCMAT_DLLEXPORT MatStashSetInitialSize(Mat mat,PetscInt size, PetscInt bsize)
6443 {
6444   PetscErrorCode ierr;
6445 
6446   PetscFunctionBegin;
6447   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6448   PetscValidType(mat,1);
6449   ierr = MatStashSetInitialSize_Private(&mat->stash,size);CHKERRQ(ierr);
6450   ierr = MatStashSetInitialSize_Private(&mat->bstash,bsize);CHKERRQ(ierr);
6451   PetscFunctionReturn(0);
6452 }
6453 
6454 #undef __FUNCT__
6455 #define __FUNCT__ "MatInterpolateAdd"
6456 /*@
6457    MatInterpolateAdd - w = y + A*x or A'*x depending on the shape of
6458      the matrix
6459 
6460    Collective on Mat
6461 
6462    Input Parameters:
6463 +  mat   - the matrix
6464 .  x,y - the vectors
6465 -  w - where the result is stored
6466 
6467    Level: intermediate
6468 
6469    Notes:
6470     w may be the same vector as y.
6471 
6472     This allows one to use either the restriction or interpolation (its transpose)
6473     matrix to do the interpolation
6474 
6475     Concepts: interpolation
6476 
6477 .seealso: MatMultAdd(), MatMultTransposeAdd(), MatRestrict()
6478 
6479 @*/
6480 PetscErrorCode PETSCMAT_DLLEXPORT MatInterpolateAdd(Mat A,Vec x,Vec y,Vec w)
6481 {
6482   PetscErrorCode ierr;
6483   PetscInt       M,N;
6484 
6485   PetscFunctionBegin;
6486   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
6487   PetscValidHeaderSpecific(x,VEC_COOKIE,2);
6488   PetscValidHeaderSpecific(y,VEC_COOKIE,3);
6489   PetscValidHeaderSpecific(w,VEC_COOKIE,4);
6490   PetscValidType(A,1);
6491   ierr = MatPreallocated(A);CHKERRQ(ierr);
6492   ierr = MatGetSize(A,&M,&N);CHKERRQ(ierr);
6493   if (N > M) {
6494     ierr = MatMultTransposeAdd(A,x,y,w);CHKERRQ(ierr);
6495   } else {
6496     ierr = MatMultAdd(A,x,y,w);CHKERRQ(ierr);
6497   }
6498   PetscFunctionReturn(0);
6499 }
6500 
6501 #undef __FUNCT__
6502 #define __FUNCT__ "MatInterpolate"
6503 /*@
6504    MatInterpolate - y = A*x or A'*x depending on the shape of
6505      the matrix
6506 
6507    Collective on Mat
6508 
6509    Input Parameters:
6510 +  mat   - the matrix
6511 -  x,y - the vectors
6512 
6513    Level: intermediate
6514 
6515    Notes:
6516     This allows one to use either the restriction or interpolation (its transpose)
6517     matrix to do the interpolation
6518 
6519    Concepts: matrices^interpolation
6520 
6521 .seealso: MatMultAdd(), MatMultTransposeAdd(), MatRestrict()
6522 
6523 @*/
6524 PetscErrorCode PETSCMAT_DLLEXPORT MatInterpolate(Mat A,Vec x,Vec y)
6525 {
6526   PetscErrorCode ierr;
6527   PetscInt       M,N;
6528 
6529   PetscFunctionBegin;
6530   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
6531   PetscValidHeaderSpecific(x,VEC_COOKIE,2);
6532   PetscValidHeaderSpecific(y,VEC_COOKIE,3);
6533   PetscValidType(A,1);
6534   ierr = MatPreallocated(A);CHKERRQ(ierr);
6535   ierr = MatGetSize(A,&M,&N);CHKERRQ(ierr);
6536   if (N > M) {
6537     ierr = MatMultTranspose(A,x,y);CHKERRQ(ierr);
6538   } else {
6539     ierr = MatMult(A,x,y);CHKERRQ(ierr);
6540   }
6541   PetscFunctionReturn(0);
6542 }
6543 
6544 #undef __FUNCT__
6545 #define __FUNCT__ "MatRestrict"
6546 /*@
6547    MatRestrict - y = A*x or A'*x
6548 
6549    Collective on Mat
6550 
6551    Input Parameters:
6552 +  mat   - the matrix
6553 -  x,y - the vectors
6554 
6555    Level: intermediate
6556 
6557    Notes:
6558     This allows one to use either the restriction or interpolation (its transpose)
6559     matrix to do the restriction
6560 
6561    Concepts: matrices^restriction
6562 
6563 .seealso: MatMultAdd(), MatMultTransposeAdd(), MatInterpolate()
6564 
6565 @*/
6566 PetscErrorCode PETSCMAT_DLLEXPORT MatRestrict(Mat A,Vec x,Vec y)
6567 {
6568   PetscErrorCode ierr;
6569   PetscInt       M,N;
6570 
6571   PetscFunctionBegin;
6572   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
6573   PetscValidHeaderSpecific(x,VEC_COOKIE,2);
6574   PetscValidHeaderSpecific(y,VEC_COOKIE,3);
6575   PetscValidType(A,1);
6576   ierr = MatPreallocated(A);CHKERRQ(ierr);
6577 
6578   ierr = MatGetSize(A,&M,&N);CHKERRQ(ierr);
6579   if (N > M) {
6580     ierr = MatMult(A,x,y);CHKERRQ(ierr);
6581   } else {
6582     ierr = MatMultTranspose(A,x,y);CHKERRQ(ierr);
6583   }
6584   PetscFunctionReturn(0);
6585 }
6586 
6587 #undef __FUNCT__
6588 #define __FUNCT__ "MatNullSpaceAttach"
6589 /*@
6590    MatNullSpaceAttach - attaches a null space to a matrix.
6591         This null space will be removed from the resulting vector whenever
6592         MatMult() is called
6593 
6594    Collective on Mat
6595 
6596    Input Parameters:
6597 +  mat - the matrix
6598 -  nullsp - the null space object
6599 
6600    Level: developer
6601 
6602    Notes:
6603       Overwrites any previous null space that may have been attached
6604 
6605    Concepts: null space^attaching to matrix
6606 
6607 .seealso: MatCreate(), MatNullSpaceCreate()
6608 @*/
6609 PetscErrorCode PETSCMAT_DLLEXPORT MatNullSpaceAttach(Mat mat,MatNullSpace nullsp)
6610 {
6611   PetscErrorCode ierr;
6612 
6613   PetscFunctionBegin;
6614   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6615   PetscValidType(mat,1);
6616   PetscValidHeaderSpecific(nullsp,MAT_NULLSPACE_COOKIE,2);
6617   ierr = MatPreallocated(mat);CHKERRQ(ierr);
6618   ierr = PetscObjectReference((PetscObject)nullsp);CHKERRQ(ierr);
6619   if (mat->nullsp) { ierr = MatNullSpaceDestroy(mat->nullsp);CHKERRQ(ierr); }
6620   mat->nullsp = nullsp;
6621   PetscFunctionReturn(0);
6622 }
6623 
6624 #undef __FUNCT__
6625 #define __FUNCT__ "MatICCFactor"
6626 /*@
6627    MatICCFactor - Performs in-place incomplete Cholesky factorization of matrix.
6628 
6629    Collective on Mat
6630 
6631    Input Parameters:
6632 +  mat - the matrix
6633 .  row - row/column permutation
6634 .  fill - expected fill factor >= 1.0
6635 -  level - level of fill, for ICC(k)
6636 
6637    Notes:
6638    Probably really in-place only when level of fill is zero, otherwise allocates
6639    new space to store factored matrix and deletes previous memory.
6640 
6641    Most users should employ the simplified KSP interface for linear solvers
6642    instead of working directly with matrix algebra routines such as this.
6643    See, e.g., KSPCreate().
6644 
6645    Level: developer
6646 
6647    Concepts: matrices^incomplete Cholesky factorization
6648    Concepts: Cholesky factorization
6649 
6650 .seealso: MatICCFactorSymbolic(), MatLUFactorNumeric(), MatCholeskyFactor()
6651 @*/
6652 PetscErrorCode PETSCMAT_DLLEXPORT MatICCFactor(Mat mat,IS row,const MatFactorInfo* info)
6653 {
6654   PetscErrorCode ierr;
6655 
6656   PetscFunctionBegin;
6657   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6658   PetscValidType(mat,1);
6659   if (row) PetscValidHeaderSpecific(row,IS_COOKIE,2);
6660   PetscValidPointer(info,3);
6661   if (mat->rmap->N != mat->cmap->N) SETERRQ(PETSC_ERR_ARG_WRONG,"matrix must be square");
6662   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
6663   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
6664   if (!mat->ops->iccfactor) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6665   ierr = MatPreallocated(mat);CHKERRQ(ierr);
6666   ierr = (*mat->ops->iccfactor)(mat,row,info);CHKERRQ(ierr);
6667   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
6668   PetscFunctionReturn(0);
6669 }
6670 
6671 #undef __FUNCT__
6672 #define __FUNCT__ "MatSetValuesAdic"
6673 /*@
6674    MatSetValuesAdic - Sets values computed with ADIC automatic differentiation into a matrix.
6675 
6676    Not Collective
6677 
6678    Input Parameters:
6679 +  mat - the matrix
6680 -  v - the values compute with ADIC
6681 
6682    Level: developer
6683 
6684    Notes:
6685      Must call MatSetColoring() before using this routine. Also this matrix must already
6686      have its nonzero pattern determined.
6687 
6688 .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
6689           MatSetValues(), MatSetColoring(), MatSetValuesAdifor()
6690 @*/
6691 PetscErrorCode PETSCMAT_DLLEXPORT MatSetValuesAdic(Mat mat,void *v)
6692 {
6693   PetscErrorCode ierr;
6694 
6695   PetscFunctionBegin;
6696   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6697   PetscValidType(mat,1);
6698   PetscValidPointer(mat,2);
6699 
6700   if (!mat->assembled) {
6701     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Matrix must be already assembled");
6702   }
6703   ierr = PetscLogEventBegin(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
6704   if (!mat->ops->setvaluesadic) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6705   ierr = (*mat->ops->setvaluesadic)(mat,v);CHKERRQ(ierr);
6706   ierr = PetscLogEventEnd(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
6707   ierr = MatView_Private(mat);CHKERRQ(ierr);
6708   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
6709   PetscFunctionReturn(0);
6710 }
6711 
6712 
6713 #undef __FUNCT__
6714 #define __FUNCT__ "MatSetColoring"
6715 /*@
6716    MatSetColoring - Sets a coloring used by calls to MatSetValuesAdic()
6717 
6718    Not Collective
6719 
6720    Input Parameters:
6721 +  mat - the matrix
6722 -  coloring - the coloring
6723 
6724    Level: developer
6725 
6726 .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
6727           MatSetValues(), MatSetValuesAdic()
6728 @*/
6729 PetscErrorCode PETSCMAT_DLLEXPORT MatSetColoring(Mat mat,ISColoring coloring)
6730 {
6731   PetscErrorCode ierr;
6732 
6733   PetscFunctionBegin;
6734   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6735   PetscValidType(mat,1);
6736   PetscValidPointer(coloring,2);
6737 
6738   if (!mat->assembled) {
6739     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Matrix must be already assembled");
6740   }
6741   if (!mat->ops->setcoloring) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6742   ierr = (*mat->ops->setcoloring)(mat,coloring);CHKERRQ(ierr);
6743   PetscFunctionReturn(0);
6744 }
6745 
6746 #undef __FUNCT__
6747 #define __FUNCT__ "MatSetValuesAdifor"
6748 /*@
6749    MatSetValuesAdifor - Sets values computed with automatic differentiation into a matrix.
6750 
6751    Not Collective
6752 
6753    Input Parameters:
6754 +  mat - the matrix
6755 .  nl - leading dimension of v
6756 -  v - the values compute with ADIFOR
6757 
6758    Level: developer
6759 
6760    Notes:
6761      Must call MatSetColoring() before using this routine. Also this matrix must already
6762      have its nonzero pattern determined.
6763 
6764 .seealso: MatSetOption(), MatAssemblyBegin(), MatAssemblyEnd(), MatSetValuesBlocked(), MatSetValuesLocal(),
6765           MatSetValues(), MatSetColoring()
6766 @*/
6767 PetscErrorCode PETSCMAT_DLLEXPORT MatSetValuesAdifor(Mat mat,PetscInt nl,void *v)
6768 {
6769   PetscErrorCode ierr;
6770 
6771   PetscFunctionBegin;
6772   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6773   PetscValidType(mat,1);
6774   PetscValidPointer(v,3);
6775 
6776   if (!mat->assembled) {
6777     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Matrix must be already assembled");
6778   }
6779   ierr = PetscLogEventBegin(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
6780   if (!mat->ops->setvaluesadifor) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6781   ierr = (*mat->ops->setvaluesadifor)(mat,nl,v);CHKERRQ(ierr);
6782   ierr = PetscLogEventEnd(MAT_SetValues,mat,0,0,0);CHKERRQ(ierr);
6783   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
6784   PetscFunctionReturn(0);
6785 }
6786 
6787 #undef __FUNCT__
6788 #define __FUNCT__ "MatDiagonalScaleLocal"
6789 /*@
6790    MatDiagonalScaleLocal - Scales columns of a matrix given the scaling values including the
6791          ghosted ones.
6792 
6793    Not Collective
6794 
6795    Input Parameters:
6796 +  mat - the matrix
6797 -  diag = the diagonal values, including ghost ones
6798 
6799    Level: developer
6800 
6801    Notes: Works only for MPIAIJ and MPIBAIJ matrices
6802 
6803 .seealso: MatDiagonalScale()
6804 @*/
6805 PetscErrorCode PETSCMAT_DLLEXPORT MatDiagonalScaleLocal(Mat mat,Vec diag)
6806 {
6807   PetscErrorCode ierr;
6808   PetscMPIInt    size;
6809 
6810   PetscFunctionBegin;
6811   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6812   PetscValidHeaderSpecific(diag,VEC_COOKIE,2);
6813   PetscValidType(mat,1);
6814 
6815   if (!mat->assembled) {
6816     SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Matrix must be already assembled");
6817   }
6818   ierr = PetscLogEventBegin(MAT_Scale,mat,0,0,0);CHKERRQ(ierr);
6819   ierr = MPI_Comm_size(((PetscObject)mat)->comm,&size);CHKERRQ(ierr);
6820   if (size == 1) {
6821     PetscInt n,m;
6822     ierr = VecGetSize(diag,&n);CHKERRQ(ierr);
6823     ierr = MatGetSize(mat,0,&m);CHKERRQ(ierr);
6824     if (m == n) {
6825       ierr = MatDiagonalScale(mat,0,diag);CHKERRQ(ierr);
6826     } else {
6827       SETERRQ(PETSC_ERR_SUP,"Only supported for sequential matrices when no ghost points/periodic conditions");
6828     }
6829   } else {
6830     PetscErrorCode (*f)(Mat,Vec);
6831     ierr = PetscObjectQueryFunction((PetscObject)mat,"MatDiagonalScaleLocal_C",(void (**)(void))&f);CHKERRQ(ierr);
6832     if (f) {
6833       ierr = (*f)(mat,diag);CHKERRQ(ierr);
6834     } else {
6835       SETERRQ(PETSC_ERR_SUP,"Only supported for MPIAIJ and MPIBAIJ parallel matrices");
6836     }
6837   }
6838   ierr = PetscLogEventEnd(MAT_Scale,mat,0,0,0);CHKERRQ(ierr);
6839   ierr = PetscObjectStateIncrease((PetscObject)mat);CHKERRQ(ierr);
6840   PetscFunctionReturn(0);
6841 }
6842 
6843 #undef __FUNCT__
6844 #define __FUNCT__ "MatGetInertia"
6845 /*@
6846    MatGetInertia - Gets the inertia from a factored matrix
6847 
6848    Collective on Mat
6849 
6850    Input Parameter:
6851 .  mat - the matrix
6852 
6853    Output Parameters:
6854 +   nneg - number of negative eigenvalues
6855 .   nzero - number of zero eigenvalues
6856 -   npos - number of positive eigenvalues
6857 
6858    Level: advanced
6859 
6860    Notes: Matrix must have been factored by MatCholeskyFactor()
6861 
6862 
6863 @*/
6864 PetscErrorCode PETSCMAT_DLLEXPORT MatGetInertia(Mat mat,PetscInt *nneg,PetscInt *nzero,PetscInt *npos)
6865 {
6866   PetscErrorCode ierr;
6867 
6868   PetscFunctionBegin;
6869   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6870   PetscValidType(mat,1);
6871   if (!mat->factor)    SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
6872   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Numeric factor mat is not assembled");
6873   if (!mat->ops->getinertia) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6874   ierr = (*mat->ops->getinertia)(mat,nneg,nzero,npos);CHKERRQ(ierr);
6875   PetscFunctionReturn(0);
6876 }
6877 
6878 /* ----------------------------------------------------------------*/
6879 #undef __FUNCT__
6880 #define __FUNCT__ "MatSolves"
6881 /*@C
6882    MatSolves - Solves A x = b, given a factored matrix, for a collection of vectors
6883 
6884    Collective on Mat and Vecs
6885 
6886    Input Parameters:
6887 +  mat - the factored matrix
6888 -  b - the right-hand-side vectors
6889 
6890    Output Parameter:
6891 .  x - the result vectors
6892 
6893    Notes:
6894    The vectors b and x cannot be the same.  I.e., one cannot
6895    call MatSolves(A,x,x).
6896 
6897    Notes:
6898    Most users should employ the simplified KSP interface for linear solvers
6899    instead of working directly with matrix algebra routines such as this.
6900    See, e.g., KSPCreate().
6901 
6902    Level: developer
6903 
6904    Concepts: matrices^triangular solves
6905 
6906 .seealso: MatSolveAdd(), MatSolveTranspose(), MatSolveTransposeAdd(), MatSolve()
6907 @*/
6908 PetscErrorCode PETSCMAT_DLLEXPORT MatSolves(Mat mat,Vecs b,Vecs x)
6909 {
6910   PetscErrorCode ierr;
6911 
6912   PetscFunctionBegin;
6913   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
6914   PetscValidType(mat,1);
6915   if (x == b) SETERRQ(PETSC_ERR_ARG_IDN,"x and b must be different vectors");
6916   if (!mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Unfactored matrix");
6917   if (!mat->rmap->N && !mat->cmap->N) PetscFunctionReturn(0);
6918 
6919   if (!mat->ops->solves) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
6920   ierr = MatPreallocated(mat);CHKERRQ(ierr);
6921   ierr = PetscLogEventBegin(MAT_Solves,mat,0,0,0);CHKERRQ(ierr);
6922   ierr = (*mat->ops->solves)(mat,b,x);CHKERRQ(ierr);
6923   ierr = PetscLogEventEnd(MAT_Solves,mat,0,0,0);CHKERRQ(ierr);
6924   PetscFunctionReturn(0);
6925 }
6926 
6927 #undef __FUNCT__
6928 #define __FUNCT__ "MatIsSymmetric"
6929 /*@
6930    MatIsSymmetric - Test whether a matrix is symmetric
6931 
6932    Collective on Mat
6933 
6934    Input Parameter:
6935 +  A - the matrix to test
6936 -  tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact transpose)
6937 
6938    Output Parameters:
6939 .  flg - the result
6940 
6941    Level: intermediate
6942 
6943    Concepts: matrix^symmetry
6944 
6945 .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetricKnown()
6946 @*/
6947 PetscErrorCode PETSCMAT_DLLEXPORT MatIsSymmetric(Mat A,PetscReal tol,PetscTruth *flg)
6948 {
6949   PetscErrorCode ierr;
6950 
6951   PetscFunctionBegin;
6952   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
6953   PetscValidPointer(flg,2);
6954   if (!A->symmetric_set) {
6955     if (!A->ops->issymmetric) {
6956       const MatType mattype;
6957       ierr = MatGetType(A,&mattype);CHKERRQ(ierr);
6958       SETERRQ1(PETSC_ERR_SUP,"Matrix of type <%s> does not support checking for symmetric",mattype);
6959     }
6960     ierr = (*A->ops->issymmetric)(A,tol,&A->symmetric);CHKERRQ(ierr);
6961     A->symmetric_set = PETSC_TRUE;
6962     if (A->symmetric) {
6963       A->structurally_symmetric_set = PETSC_TRUE;
6964       A->structurally_symmetric     = PETSC_TRUE;
6965     }
6966   }
6967   *flg = A->symmetric;
6968   PetscFunctionReturn(0);
6969 }
6970 
6971 #undef __FUNCT__
6972 #define __FUNCT__ "MatIsHermitian"
6973 /*@
6974    MatIsHermitian - Test whether a matrix is Hermitian
6975 
6976    Collective on Mat
6977 
6978    Input Parameter:
6979 +  A - the matrix to test
6980 -  tol - difference between value and its transpose less than this amount counts as equal (use 0.0 for exact Hermitian)
6981 
6982    Output Parameters:
6983 .  flg - the result
6984 
6985    Level: intermediate
6986 
6987    Concepts: matrix^symmetry
6988 
6989 .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetricKnown()
6990 @*/
6991 PetscErrorCode PETSCMAT_DLLEXPORT MatIsHermitian(Mat A,PetscReal tol,PetscTruth *flg)
6992 {
6993   PetscErrorCode ierr;
6994 
6995   PetscFunctionBegin;
6996   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
6997   PetscValidPointer(flg,2);
6998   if (!A->hermitian_set) {
6999     if (!A->ops->ishermitian) {
7000       const MatType mattype;
7001       ierr = MatGetType(A,&mattype);CHKERRQ(ierr);
7002       SETERRQ1(PETSC_ERR_SUP,"Matrix of type <%s> does not support checking for Hermitian",mattype);
7003     }
7004     ierr = (*A->ops->ishermitian)(A,tol,&A->hermitian);CHKERRQ(ierr);
7005     A->hermitian_set = PETSC_TRUE;
7006     if (A->hermitian) {
7007       A->structurally_symmetric_set = PETSC_TRUE;
7008       A->structurally_symmetric     = PETSC_TRUE;
7009     }
7010   }
7011   *flg = A->hermitian;
7012   PetscFunctionReturn(0);
7013 }
7014 
7015 #undef __FUNCT__
7016 #define __FUNCT__ "MatIsSymmetricKnown"
7017 /*@
7018    MatIsSymmetricKnown - Checks the flag on the matrix to see if it is symmetric.
7019 
7020    Collective on Mat
7021 
7022    Input Parameter:
7023 .  A - the matrix to check
7024 
7025    Output Parameters:
7026 +  set - if the symmetric flag is set (this tells you if the next flag is valid)
7027 -  flg - the result
7028 
7029    Level: advanced
7030 
7031    Concepts: matrix^symmetry
7032 
7033    Note: Does not check the matrix values directly, so this may return unknown (set = PETSC_FALSE). Use MatIsSymmetric()
7034          if you want it explicitly checked
7035 
7036 .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetric()
7037 @*/
7038 PetscErrorCode PETSCMAT_DLLEXPORT MatIsSymmetricKnown(Mat A,PetscTruth *set,PetscTruth *flg)
7039 {
7040   PetscFunctionBegin;
7041   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
7042   PetscValidPointer(set,2);
7043   PetscValidPointer(flg,3);
7044   if (A->symmetric_set) {
7045     *set = PETSC_TRUE;
7046     *flg = A->symmetric;
7047   } else {
7048     *set = PETSC_FALSE;
7049   }
7050   PetscFunctionReturn(0);
7051 }
7052 
7053 #undef __FUNCT__
7054 #define __FUNCT__ "MatIsHermitianKnown"
7055 /*@
7056    MatIsHermitianKnown - Checks the flag on the matrix to see if it is hermitian.
7057 
7058    Collective on Mat
7059 
7060    Input Parameter:
7061 .  A - the matrix to check
7062 
7063    Output Parameters:
7064 +  set - if the hermitian flag is set (this tells you if the next flag is valid)
7065 -  flg - the result
7066 
7067    Level: advanced
7068 
7069    Concepts: matrix^symmetry
7070 
7071    Note: Does not check the matrix values directly, so this may return unknown (set = PETSC_FALSE). Use MatIsHermitian()
7072          if you want it explicitly checked
7073 
7074 .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsStructurallySymmetric(), MatSetOption(), MatIsSymmetric()
7075 @*/
7076 PetscErrorCode PETSCMAT_DLLEXPORT MatIsHermitianKnown(Mat A,PetscTruth *set,PetscTruth *flg)
7077 {
7078   PetscFunctionBegin;
7079   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
7080   PetscValidPointer(set,2);
7081   PetscValidPointer(flg,3);
7082   if (A->hermitian_set) {
7083     *set = PETSC_TRUE;
7084     *flg = A->hermitian;
7085   } else {
7086     *set = PETSC_FALSE;
7087   }
7088   PetscFunctionReturn(0);
7089 }
7090 
7091 #undef __FUNCT__
7092 #define __FUNCT__ "MatIsStructurallySymmetric"
7093 /*@
7094    MatIsStructurallySymmetric - Test whether a matrix is structurally symmetric
7095 
7096    Collective on Mat
7097 
7098    Input Parameter:
7099 .  A - the matrix to test
7100 
7101    Output Parameters:
7102 .  flg - the result
7103 
7104    Level: intermediate
7105 
7106    Concepts: matrix^symmetry
7107 
7108 .seealso: MatTranspose(), MatIsTranspose(), MatIsHermitian(), MatIsSymmetric(), MatSetOption()
7109 @*/
7110 PetscErrorCode PETSCMAT_DLLEXPORT MatIsStructurallySymmetric(Mat A,PetscTruth *flg)
7111 {
7112   PetscErrorCode ierr;
7113 
7114   PetscFunctionBegin;
7115   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
7116   PetscValidPointer(flg,2);
7117   if (!A->structurally_symmetric_set) {
7118     if (!A->ops->isstructurallysymmetric) SETERRQ(PETSC_ERR_SUP,"Matrix does not support checking for structural symmetric");
7119     ierr = (*A->ops->isstructurallysymmetric)(A,&A->structurally_symmetric);CHKERRQ(ierr);
7120     A->structurally_symmetric_set = PETSC_TRUE;
7121   }
7122   *flg = A->structurally_symmetric;
7123   PetscFunctionReturn(0);
7124 }
7125 
7126 #undef __FUNCT__
7127 #define __FUNCT__ "MatStashGetInfo"
7128 extern PetscErrorCode MatStashGetInfo_Private(MatStash*,PetscInt*,PetscInt*);
7129 /*@
7130    MatStashGetInfo - Gets how many values are currently in the vector stash, i.e. need
7131        to be communicated to other processors during the MatAssemblyBegin/End() process
7132 
7133     Not collective
7134 
7135    Input Parameter:
7136 .   vec - the vector
7137 
7138    Output Parameters:
7139 +   nstash   - the size of the stash
7140 .   reallocs - the number of additional mallocs incurred.
7141 .   bnstash   - the size of the block stash
7142 -   breallocs - the number of additional mallocs incurred.in the block stash
7143 
7144    Level: advanced
7145 
7146 .seealso: MatAssemblyBegin(), MatAssemblyEnd(), Mat, MatStashSetInitialSize()
7147 
7148 @*/
7149 PetscErrorCode PETSCMAT_DLLEXPORT MatStashGetInfo(Mat mat,PetscInt *nstash,PetscInt *reallocs,PetscInt *bnstash,PetscInt *breallocs)
7150 {
7151   PetscErrorCode ierr;
7152   PetscFunctionBegin;
7153   ierr = MatStashGetInfo_Private(&mat->stash,nstash,reallocs);CHKERRQ(ierr);
7154   ierr = MatStashGetInfo_Private(&mat->bstash,bnstash,breallocs);CHKERRQ(ierr);
7155   PetscFunctionReturn(0);
7156 }
7157 
7158 #undef __FUNCT__
7159 #define __FUNCT__ "MatGetVecs"
7160 /*@C
7161    MatGetVecs - Get vector(s) compatible with the matrix, i.e. with the same
7162      parallel layout
7163 
7164    Collective on Mat
7165 
7166    Input Parameter:
7167 .  mat - the matrix
7168 
7169    Output Parameter:
7170 +   right - (optional) vector that the matrix can be multiplied against
7171 -   left - (optional) vector that the matrix vector product can be stored in
7172 
7173   Level: advanced
7174 
7175 .seealso: MatCreate()
7176 @*/
7177 PetscErrorCode PETSCMAT_DLLEXPORT MatGetVecs(Mat mat,Vec *right,Vec *left)
7178 {
7179   PetscErrorCode ierr;
7180 
7181   PetscFunctionBegin;
7182   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
7183   PetscValidType(mat,1);
7184   ierr = MatPreallocated(mat);CHKERRQ(ierr);
7185   if (mat->ops->getvecs) {
7186     ierr = (*mat->ops->getvecs)(mat,right,left);CHKERRQ(ierr);
7187   } else {
7188     PetscMPIInt size;
7189     ierr = MPI_Comm_size(((PetscObject)mat)->comm, &size);CHKERRQ(ierr);
7190     if (right) {
7191       ierr = VecCreate(((PetscObject)mat)->comm,right);CHKERRQ(ierr);
7192       ierr = VecSetSizes(*right,mat->cmap->n,PETSC_DETERMINE);CHKERRQ(ierr);
7193       ierr = VecSetBlockSize(*right,mat->rmap->bs);CHKERRQ(ierr);
7194       if (size > 1) {
7195         /* New vectors uses Mat cmap and does not create a new one */
7196 	ierr = PetscMapDestroy((*right)->map);CHKERRQ(ierr);
7197 	(*right)->map = mat->cmap;
7198 	mat->cmap->refcnt++;
7199 
7200         ierr = VecSetType(*right,VECMPI);CHKERRQ(ierr);
7201       } else {ierr = VecSetType(*right,VECSEQ);CHKERRQ(ierr);}
7202     }
7203     if (left) {
7204       ierr = VecCreate(((PetscObject)mat)->comm,left);CHKERRQ(ierr);
7205       ierr = VecSetSizes(*left,mat->rmap->n,PETSC_DETERMINE);CHKERRQ(ierr);
7206       ierr = VecSetBlockSize(*left,mat->rmap->bs);CHKERRQ(ierr);
7207       if (size > 1) {
7208         /* New vectors uses Mat rmap and does not create a new one */
7209 	ierr = PetscMapDestroy((*left)->map);CHKERRQ(ierr);
7210 	(*left)->map = mat->rmap;
7211 	mat->rmap->refcnt++;
7212 
7213         ierr = VecSetType(*left,VECMPI);CHKERRQ(ierr);
7214       } else {ierr = VecSetType(*left,VECSEQ);CHKERRQ(ierr);}
7215     }
7216   }
7217   if (mat->mapping) {
7218     if (right) {ierr = VecSetLocalToGlobalMapping(*right,mat->mapping);CHKERRQ(ierr);}
7219     if (left) {ierr = VecSetLocalToGlobalMapping(*left,mat->mapping);CHKERRQ(ierr);}
7220   }
7221   if (mat->bmapping) {
7222     if (right) {ierr = VecSetLocalToGlobalMappingBlock(*right,mat->bmapping);CHKERRQ(ierr);}
7223     if (left) {ierr = VecSetLocalToGlobalMappingBlock(*left,mat->bmapping);CHKERRQ(ierr);}
7224   }
7225   PetscFunctionReturn(0);
7226 }
7227 
7228 #undef __FUNCT__
7229 #define __FUNCT__ "MatFactorInfoInitialize"
7230 /*@
7231    MatFactorInfoInitialize - Initializes a MatFactorInfo data structure
7232      with default values.
7233 
7234    Not Collective
7235 
7236    Input Parameters:
7237 .    info - the MatFactorInfo data structure
7238 
7239 
7240    Notes: The solvers are generally used through the KSP and PC objects, for example
7241           PCLU, PCILU, PCCHOLESKY, PCICC
7242 
7243    Level: developer
7244 
7245 .seealso: MatFactorInfo
7246 @*/
7247 
7248 PetscErrorCode PETSCMAT_DLLEXPORT MatFactorInfoInitialize(MatFactorInfo *info)
7249 {
7250   PetscErrorCode ierr;
7251 
7252   PetscFunctionBegin;
7253   ierr = PetscMemzero(info,sizeof(MatFactorInfo));CHKERRQ(ierr);
7254   PetscFunctionReturn(0);
7255 }
7256 
7257 #undef __FUNCT__
7258 #define __FUNCT__ "MatPtAP"
7259 /*@
7260    MatPtAP - Creates the matrix projection C = P^T * A * P
7261 
7262    Collective on Mat
7263 
7264    Input Parameters:
7265 +  A - the matrix
7266 .  P - the projection matrix
7267 .  scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
7268 -  fill - expected fill as ratio of nnz(C)/nnz(A)
7269 
7270    Output Parameters:
7271 .  C - the product matrix
7272 
7273    Notes:
7274    C will be created and must be destroyed by the user with MatDestroy().
7275 
7276    This routine is currently only implemented for pairs of AIJ matrices and classes
7277    which inherit from AIJ.
7278 
7279    Level: intermediate
7280 
7281 .seealso: MatPtAPSymbolic(), MatPtAPNumeric(), MatMatMult()
7282 @*/
7283 PetscErrorCode PETSCMAT_DLLEXPORT MatPtAP(Mat A,Mat P,MatReuse scall,PetscReal fill,Mat *C)
7284 {
7285   PetscErrorCode ierr;
7286 
7287   PetscFunctionBegin;
7288   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
7289   PetscValidType(A,1);
7290   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7291   if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7292   PetscValidHeaderSpecific(P,MAT_COOKIE,2);
7293   PetscValidType(P,2);
7294   ierr = MatPreallocated(P);CHKERRQ(ierr);
7295   if (!P->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7296   if (P->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7297   PetscValidPointer(C,3);
7298   if (P->rmap->N!=A->cmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->rmap->N,A->cmap->N);
7299   if (fill < 1.0) SETERRQ1(PETSC_ERR_ARG_SIZ,"Expected fill=%G must be >= 1.0",fill);
7300   ierr = MatPreallocated(A);CHKERRQ(ierr);
7301 
7302   ierr = PetscLogEventBegin(MAT_PtAP,A,P,0,0);CHKERRQ(ierr);
7303   ierr = (*A->ops->ptap)(A,P,scall,fill,C);CHKERRQ(ierr);
7304   ierr = PetscLogEventEnd(MAT_PtAP,A,P,0,0);CHKERRQ(ierr);
7305 
7306   PetscFunctionReturn(0);
7307 }
7308 
7309 #undef __FUNCT__
7310 #define __FUNCT__ "MatPtAPNumeric"
7311 /*@
7312    MatPtAPNumeric - Computes the matrix projection C = P^T * A * P
7313 
7314    Collective on Mat
7315 
7316    Input Parameters:
7317 +  A - the matrix
7318 -  P - the projection matrix
7319 
7320    Output Parameters:
7321 .  C - the product matrix
7322 
7323    Notes:
7324    C must have been created by calling MatPtAPSymbolic and must be destroyed by
7325    the user using MatDeatroy().
7326 
7327    This routine is currently only implemented for pairs of AIJ matrices and classes
7328    which inherit from AIJ.  C will be of type MATAIJ.
7329 
7330    Level: intermediate
7331 
7332 .seealso: MatPtAP(), MatPtAPSymbolic(), MatMatMultNumeric()
7333 @*/
7334 PetscErrorCode PETSCMAT_DLLEXPORT MatPtAPNumeric(Mat A,Mat P,Mat C)
7335 {
7336   PetscErrorCode ierr;
7337 
7338   PetscFunctionBegin;
7339   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
7340   PetscValidType(A,1);
7341   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7342   if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7343   PetscValidHeaderSpecific(P,MAT_COOKIE,2);
7344   PetscValidType(P,2);
7345   ierr = MatPreallocated(P);CHKERRQ(ierr);
7346   if (!P->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7347   if (P->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7348   PetscValidHeaderSpecific(C,MAT_COOKIE,3);
7349   PetscValidType(C,3);
7350   ierr = MatPreallocated(C);CHKERRQ(ierr);
7351   if (C->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7352   if (P->cmap->N!=C->rmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->cmap->N,C->rmap->N);
7353   if (P->rmap->N!=A->cmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->rmap->N,A->cmap->N);
7354   if (A->rmap->N!=A->cmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix 'A' must be square, %D != %D",A->rmap->N,A->cmap->N);
7355   if (P->cmap->N!=C->cmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->cmap->N,C->cmap->N);
7356   ierr = MatPreallocated(A);CHKERRQ(ierr);
7357 
7358   ierr = PetscLogEventBegin(MAT_PtAPNumeric,A,P,0,0);CHKERRQ(ierr);
7359   ierr = (*A->ops->ptapnumeric)(A,P,C);CHKERRQ(ierr);
7360   ierr = PetscLogEventEnd(MAT_PtAPNumeric,A,P,0,0);CHKERRQ(ierr);
7361   PetscFunctionReturn(0);
7362 }
7363 
7364 #undef __FUNCT__
7365 #define __FUNCT__ "MatPtAPSymbolic"
7366 /*@
7367    MatPtAPSymbolic - Creates the (i,j) structure of the matrix projection C = P^T * A * P
7368 
7369    Collective on Mat
7370 
7371    Input Parameters:
7372 +  A - the matrix
7373 -  P - the projection matrix
7374 
7375    Output Parameters:
7376 .  C - the (i,j) structure of the product matrix
7377 
7378    Notes:
7379    C will be created and must be destroyed by the user with MatDestroy().
7380 
7381    This routine is currently only implemented for pairs of SeqAIJ matrices and classes
7382    which inherit from SeqAIJ.  C will be of type MATSEQAIJ.  The product is computed using
7383    this (i,j) structure by calling MatPtAPNumeric().
7384 
7385    Level: intermediate
7386 
7387 .seealso: MatPtAP(), MatPtAPNumeric(), MatMatMultSymbolic()
7388 @*/
7389 PetscErrorCode PETSCMAT_DLLEXPORT MatPtAPSymbolic(Mat A,Mat P,PetscReal fill,Mat *C)
7390 {
7391   PetscErrorCode ierr;
7392 
7393   PetscFunctionBegin;
7394   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
7395   PetscValidType(A,1);
7396   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7397   if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7398   if (fill <1.0) SETERRQ1(PETSC_ERR_ARG_SIZ,"Expected fill=%G must be >= 1.0",fill);
7399   PetscValidHeaderSpecific(P,MAT_COOKIE,2);
7400   PetscValidType(P,2);
7401   ierr = MatPreallocated(P);CHKERRQ(ierr);
7402   if (!P->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7403   if (P->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7404   PetscValidPointer(C,3);
7405 
7406   if (P->rmap->N!=A->cmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",P->rmap->N,A->cmap->N);
7407   if (A->rmap->N!=A->cmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix 'A' must be square, %D != %D",A->rmap->N,A->cmap->N);
7408   ierr = MatPreallocated(A);CHKERRQ(ierr);
7409   ierr = PetscLogEventBegin(MAT_PtAPSymbolic,A,P,0,0);CHKERRQ(ierr);
7410   ierr = (*A->ops->ptapsymbolic)(A,P,fill,C);CHKERRQ(ierr);
7411   ierr = PetscLogEventEnd(MAT_PtAPSymbolic,A,P,0,0);CHKERRQ(ierr);
7412 
7413   ierr = MatSetBlockSize(*C,A->rmap->bs);CHKERRQ(ierr);
7414 
7415   PetscFunctionReturn(0);
7416 }
7417 
7418 #undef __FUNCT__
7419 #define __FUNCT__ "MatMatMult"
7420 /*@
7421    MatMatMult - Performs Matrix-Matrix Multiplication C=A*B.
7422 
7423    Collective on Mat
7424 
7425    Input Parameters:
7426 +  A - the left matrix
7427 .  B - the right matrix
7428 .  scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
7429 -  fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use PETSC_DEFAULT if you do not have a good estimate
7430           if the result is a dense matrix this is irrelevent
7431 
7432    Output Parameters:
7433 .  C - the product matrix
7434 
7435    Notes:
7436    Unless scall is MAT_REUSE_MATRIX C will be created.
7437 
7438    MAT_REUSE_MATRIX can only be used if the matrices A and B have the same nonzero pattern as in the previous call
7439 
7440    To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
7441    actually needed.
7442 
7443    If you have many matrices with the same non-zero structure to multiply, you
7444    should either
7445 $   1) use MAT_REUSE_MATRIX in all calls but the first or
7446 $   2) call MatMatMultSymbolic() once and then MatMatMultNumeric() for each product needed
7447 
7448    Level: intermediate
7449 
7450 .seealso: MatMatMultSymbolic(), MatMatMultNumeric(), MatPtAP()
7451 @*/
7452 PetscErrorCode PETSCMAT_DLLEXPORT MatMatMult(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
7453 {
7454   PetscErrorCode ierr;
7455   PetscErrorCode (*fA)(Mat,Mat,MatReuse,PetscReal,Mat*);
7456   PetscErrorCode (*fB)(Mat,Mat,MatReuse,PetscReal,Mat*);
7457   PetscErrorCode (*mult)(Mat,Mat,MatReuse,PetscReal,Mat *)=PETSC_NULL;
7458 
7459   PetscFunctionBegin;
7460   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
7461   PetscValidType(A,1);
7462   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7463   if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7464   PetscValidHeaderSpecific(B,MAT_COOKIE,2);
7465   PetscValidType(B,2);
7466   ierr = MatPreallocated(B);CHKERRQ(ierr);
7467   if (!B->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7468   if (B->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7469   PetscValidPointer(C,3);
7470   if (B->rmap->N!=A->cmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->rmap->N,A->cmap->N);
7471   if (scall == MAT_REUSE_MATRIX){
7472     PetscValidPointer(*C,5);
7473     PetscValidHeaderSpecific(*C,MAT_COOKIE,5);
7474   }
7475   if (fill == PETSC_DEFAULT || fill == PETSC_DECIDE) fill = 2.0;
7476   if (fill < 1.0) SETERRQ1(PETSC_ERR_ARG_SIZ,"Expected fill=%G must be >= 1.0",fill);
7477   ierr = MatPreallocated(A);CHKERRQ(ierr);
7478 
7479   fA = A->ops->matmult;
7480   fB = B->ops->matmult;
7481   if (fB == fA) {
7482     if (!fB) SETERRQ1(PETSC_ERR_SUP,"MatMatMult not supported for B of type %s",((PetscObject)B)->type_name);
7483     mult = fB;
7484   } else {
7485     /* dispatch based on the type of A and B */
7486     char  multname[256];
7487     ierr = PetscStrcpy(multname,"MatMatMult_");CHKERRQ(ierr);
7488     ierr = PetscStrcat(multname,((PetscObject)A)->type_name);CHKERRQ(ierr);
7489     ierr = PetscStrcat(multname,"_");CHKERRQ(ierr);
7490     ierr = PetscStrcat(multname,((PetscObject)B)->type_name);CHKERRQ(ierr);
7491     ierr = PetscStrcat(multname,"_C");CHKERRQ(ierr); /* e.g., multname = "MatMatMult_seqdense_seqaij_C" */
7492     ierr = PetscObjectQueryFunction((PetscObject)B,multname,(void (**)(void))&mult);CHKERRQ(ierr);
7493     if (!mult) SETERRQ2(PETSC_ERR_ARG_INCOMP,"MatMatMult requires A, %s, to be compatible with B, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
7494   }
7495   ierr = PetscLogEventBegin(MAT_MatMult,A,B,0,0);CHKERRQ(ierr);
7496   ierr = (*mult)(A,B,scall,fill,C);CHKERRQ(ierr);
7497   ierr = PetscLogEventEnd(MAT_MatMult,A,B,0,0);CHKERRQ(ierr);
7498   PetscFunctionReturn(0);
7499 }
7500 
7501 #undef __FUNCT__
7502 #define __FUNCT__ "MatMatMultSymbolic"
7503 /*@
7504    MatMatMultSymbolic - Performs construction, preallocation, and computes the ij structure
7505    of the matrix-matrix product C=A*B.  Call this routine before calling MatMatMultNumeric().
7506 
7507    Collective on Mat
7508 
7509    Input Parameters:
7510 +  A - the left matrix
7511 .  B - the right matrix
7512 -  fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use PETSC_DEFAULT if you do not have a good estimate,
7513       if C is a dense matrix this is irrelevent
7514 
7515    Output Parameters:
7516 .  C - the product matrix
7517 
7518    Notes:
7519    Unless scall is MAT_REUSE_MATRIX C will be created.
7520 
7521    To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
7522    actually needed.
7523 
7524    This routine is currently implemented for
7525     - pairs of AIJ matrices and classes which inherit from AIJ, C will be of type AIJ
7526     - pairs of AIJ (A) and Dense (B) matrix, C will be of type Dense.
7527     - pairs of Dense (A) and AIJ (B) matrix, C will be of type Dense.
7528 
7529    Level: intermediate
7530 
7531 .seealso: MatMatMult(), MatMatMultNumeric()
7532 @*/
7533 PetscErrorCode PETSCMAT_DLLEXPORT MatMatMultSymbolic(Mat A,Mat B,PetscReal fill,Mat *C)
7534 {
7535   PetscErrorCode ierr;
7536   PetscErrorCode (*Asymbolic)(Mat,Mat,PetscReal,Mat *);
7537   PetscErrorCode (*Bsymbolic)(Mat,Mat,PetscReal,Mat *);
7538   PetscErrorCode (*symbolic)(Mat,Mat,PetscReal,Mat *)=PETSC_NULL;
7539 
7540   PetscFunctionBegin;
7541   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
7542   PetscValidType(A,1);
7543   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7544   if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7545 
7546   PetscValidHeaderSpecific(B,MAT_COOKIE,2);
7547   PetscValidType(B,2);
7548   ierr = MatPreallocated(B);CHKERRQ(ierr);
7549   if (!B->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7550   if (B->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7551   PetscValidPointer(C,3);
7552 
7553   if (B->rmap->N!=A->cmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->rmap->N,A->cmap->N);
7554   if (fill == PETSC_DEFAULT) fill = 2.0;
7555   if (fill < 1.0) SETERRQ1(PETSC_ERR_ARG_SIZ,"Expected fill=%G must be > 1.0",fill);
7556   ierr = MatPreallocated(A);CHKERRQ(ierr);
7557 
7558   Asymbolic = A->ops->matmultsymbolic;
7559   Bsymbolic = B->ops->matmultsymbolic;
7560   if (Asymbolic == Bsymbolic){
7561     if (!Bsymbolic) SETERRQ1(PETSC_ERR_SUP,"C=A*B not implemented for B of type %s",((PetscObject)B)->type_name);
7562     symbolic = Bsymbolic;
7563   } else { /* dispatch based on the type of A and B */
7564     char  symbolicname[256];
7565     ierr = PetscStrcpy(symbolicname,"MatMatMultSymbolic_");CHKERRQ(ierr);
7566     ierr = PetscStrcat(symbolicname,((PetscObject)A)->type_name);CHKERRQ(ierr);
7567     ierr = PetscStrcat(symbolicname,"_");CHKERRQ(ierr);
7568     ierr = PetscStrcat(symbolicname,((PetscObject)B)->type_name);CHKERRQ(ierr);
7569     ierr = PetscStrcat(symbolicname,"_C");CHKERRQ(ierr);
7570     ierr = PetscObjectQueryFunction((PetscObject)B,symbolicname,(void (**)(void))&symbolic);CHKERRQ(ierr);
7571     if (!symbolic) SETERRQ2(PETSC_ERR_ARG_INCOMP,"MatMatMultSymbolic requires A, %s, to be compatible with B, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
7572   }
7573   ierr = PetscLogEventBegin(MAT_MatMultSymbolic,A,B,0,0);CHKERRQ(ierr);
7574   ierr = (*symbolic)(A,B,fill,C);CHKERRQ(ierr);
7575   ierr = PetscLogEventEnd(MAT_MatMultSymbolic,A,B,0,0);CHKERRQ(ierr);
7576   PetscFunctionReturn(0);
7577 }
7578 
7579 #undef __FUNCT__
7580 #define __FUNCT__ "MatMatMultNumeric"
7581 /*@
7582    MatMatMultNumeric - Performs the numeric matrix-matrix product.
7583    Call this routine after first calling MatMatMultSymbolic().
7584 
7585    Collective on Mat
7586 
7587    Input Parameters:
7588 +  A - the left matrix
7589 -  B - the right matrix
7590 
7591    Output Parameters:
7592 .  C - the product matrix, which was created by from MatMatMultSymbolic() or a call to MatMatMult().
7593 
7594    Notes:
7595    C must have been created with MatMatMultSymbolic().
7596 
7597    This routine is currently implemented for
7598     - pairs of AIJ matrices and classes which inherit from AIJ, C will be of type MATAIJ.
7599     - pairs of AIJ (A) and Dense (B) matrix, C will be of type Dense.
7600     - pairs of Dense (A) and AIJ (B) matrix, C will be of type Dense.
7601 
7602    Level: intermediate
7603 
7604 .seealso: MatMatMult(), MatMatMultSymbolic()
7605 @*/
7606 PetscErrorCode PETSCMAT_DLLEXPORT MatMatMultNumeric(Mat A,Mat B,Mat C)
7607 {
7608   PetscErrorCode ierr;
7609   PetscErrorCode (*Anumeric)(Mat,Mat,Mat);
7610   PetscErrorCode (*Bnumeric)(Mat,Mat,Mat);
7611   PetscErrorCode (*numeric)(Mat,Mat,Mat)=PETSC_NULL;
7612 
7613   PetscFunctionBegin;
7614   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
7615   PetscValidType(A,1);
7616   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7617   if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7618 
7619   PetscValidHeaderSpecific(B,MAT_COOKIE,2);
7620   PetscValidType(B,2);
7621   ierr = MatPreallocated(B);CHKERRQ(ierr);
7622   if (!B->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7623   if (B->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7624 
7625   PetscValidHeaderSpecific(C,MAT_COOKIE,3);
7626   PetscValidType(C,3);
7627   ierr = MatPreallocated(C);CHKERRQ(ierr);
7628   if (!C->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7629   if (C->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7630 
7631   if (B->cmap->N!=C->cmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->cmap->N,C->cmap->N);
7632   if (B->rmap->N!=A->cmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->rmap->N,A->cmap->N);
7633   if (A->rmap->N!=C->rmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",A->rmap->N,C->rmap->N);
7634   ierr = MatPreallocated(A);CHKERRQ(ierr);
7635 
7636   Anumeric = A->ops->matmultnumeric;
7637   Bnumeric = B->ops->matmultnumeric;
7638   if (Anumeric == Bnumeric){
7639     if (!Bnumeric) SETERRQ1(PETSC_ERR_SUP,"MatMatMultNumeric not supported for B of type %s",((PetscObject)B)->type_name);
7640     numeric = Bnumeric;
7641   } else {
7642     char  numericname[256];
7643     ierr = PetscStrcpy(numericname,"MatMatMultNumeric_");CHKERRQ(ierr);
7644     ierr = PetscStrcat(numericname,((PetscObject)A)->type_name);CHKERRQ(ierr);
7645     ierr = PetscStrcat(numericname,"_");CHKERRQ(ierr);
7646     ierr = PetscStrcat(numericname,((PetscObject)B)->type_name);CHKERRQ(ierr);
7647     ierr = PetscStrcat(numericname,"_C");CHKERRQ(ierr);
7648     ierr = PetscObjectQueryFunction((PetscObject)B,numericname,(void (**)(void))&numeric);CHKERRQ(ierr);
7649     if (!numeric)
7650       SETERRQ2(PETSC_ERR_ARG_INCOMP,"MatMatMultNumeric requires A, %s, to be compatible with B, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
7651   }
7652   ierr = PetscLogEventBegin(MAT_MatMultNumeric,A,B,0,0);CHKERRQ(ierr);
7653   ierr = (*numeric)(A,B,C);CHKERRQ(ierr);
7654   ierr = PetscLogEventEnd(MAT_MatMultNumeric,A,B,0,0);CHKERRQ(ierr);
7655   PetscFunctionReturn(0);
7656 }
7657 
7658 #undef __FUNCT__
7659 #define __FUNCT__ "MatMatMultTranspose"
7660 /*@
7661    MatMatMultTranspose - Performs Matrix-Matrix Multiplication C=A^T*B.
7662 
7663    Collective on Mat
7664 
7665    Input Parameters:
7666 +  A - the left matrix
7667 .  B - the right matrix
7668 .  scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
7669 -  fill - expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)), use PETSC_DEFAULT if not known
7670 
7671    Output Parameters:
7672 .  C - the product matrix
7673 
7674    Notes:
7675    C will be created if MAT_INITIAL_MATRIX and must be destroyed by the user with MatDestroy().
7676 
7677    MAT_REUSE_MATRIX can only be used if the matrices A and B have the same nonzero pattern as in the previous call
7678 
7679   To determine the correct fill value, run with -info and search for the string "Fill ratio" to see the value
7680    actually needed.
7681 
7682    This routine is currently only implemented for pairs of SeqAIJ matrices and pairs of SeqDense matrices and classes
7683    which inherit from SeqAIJ.  C will be of type MATSEQAIJ.
7684 
7685    Level: intermediate
7686 
7687 .seealso: MatMatMultTransposeSymbolic(), MatMatMultTransposeNumeric(), MatPtAP()
7688 @*/
7689 PetscErrorCode PETSCMAT_DLLEXPORT MatMatMultTranspose(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
7690 {
7691   PetscErrorCode ierr;
7692   PetscErrorCode (*fA)(Mat,Mat,MatReuse,PetscReal,Mat*);
7693   PetscErrorCode (*fB)(Mat,Mat,MatReuse,PetscReal,Mat*);
7694 
7695   PetscFunctionBegin;
7696   PetscValidHeaderSpecific(A,MAT_COOKIE,1);
7697   PetscValidType(A,1);
7698   if (!A->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7699   if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7700   PetscValidHeaderSpecific(B,MAT_COOKIE,2);
7701   PetscValidType(B,2);
7702   ierr = MatPreallocated(B);CHKERRQ(ierr);
7703   if (!B->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7704   if (B->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7705   PetscValidPointer(C,3);
7706   if (B->rmap->N!=A->rmap->N) SETERRQ2(PETSC_ERR_ARG_SIZ,"Matrix dimensions are incompatible, %D != %D",B->rmap->N,A->rmap->N);
7707   if (fill < 1.0) SETERRQ1(PETSC_ERR_ARG_SIZ,"Expected fill=%G must be > 1.0",fill);
7708   ierr = MatPreallocated(A);CHKERRQ(ierr);
7709 
7710   fA = A->ops->matmulttranspose;
7711   if (!fA) SETERRQ1(PETSC_ERR_SUP,"MatMatMultTranspose not supported for A of type %s",((PetscObject)A)->type_name);
7712   fB = B->ops->matmulttranspose;
7713   if (!fB) SETERRQ1(PETSC_ERR_SUP,"MatMatMultTranspose not supported for B of type %s",((PetscObject)B)->type_name);
7714   if (fB!=fA) SETERRQ2(PETSC_ERR_ARG_INCOMP,"MatMatMultTranspose requires A, %s, to be compatible with B, %s",((PetscObject)A)->type_name,((PetscObject)B)->type_name);
7715 
7716   ierr = PetscLogEventBegin(MAT_MatMultTranspose,A,B,0,0);CHKERRQ(ierr);
7717   ierr = (*A->ops->matmulttranspose)(A,B,scall,fill,C);CHKERRQ(ierr);
7718   ierr = PetscLogEventEnd(MAT_MatMultTranspose,A,B,0,0);CHKERRQ(ierr);
7719 
7720   PetscFunctionReturn(0);
7721 }
7722 
7723 #undef __FUNCT__
7724 #define __FUNCT__ "MatGetRedundantMatrix"
7725 /*@C
7726    MatGetRedundantMatrix - Create redundant matrices and put them into processors of subcommunicators.
7727 
7728    Collective on Mat
7729 
7730    Input Parameters:
7731 +  mat - the matrix
7732 .  nsubcomm - the number of subcommunicators (= number of redundant pareallel or sequential matrices)
7733 .  subcomm - MPI communicator split from the communicator where mat resides in
7734 .  mlocal_red - number of local rows of the redundant matrix
7735 -  reuse - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
7736 
7737    Output Parameter:
7738 .  matredundant - redundant matrix
7739 
7740    Notes:
7741    MAT_REUSE_MATRIX can only be used when the nonzero structure of the
7742    original matrix has not changed from that last call to MatGetRedundantMatrix().
7743 
7744    This routine creates the duplicated matrices in subcommunicators; you should NOT create them before
7745    calling it.
7746 
7747    Only MPIAIJ matrix is supported.
7748 
7749    Level: advanced
7750 
7751    Concepts: subcommunicator
7752    Concepts: duplicate matrix
7753 
7754 .seealso: MatDestroy()
7755 @*/
7756 PetscErrorCode PETSCMAT_DLLEXPORT MatGetRedundantMatrix(Mat mat,PetscInt nsubcomm,MPI_Comm subcomm,PetscInt mlocal_red,MatReuse reuse,Mat *matredundant)
7757 {
7758   PetscErrorCode ierr;
7759 
7760   PetscFunctionBegin;
7761   PetscValidHeaderSpecific(mat,MAT_COOKIE,1);
7762   if (nsubcomm && reuse == MAT_REUSE_MATRIX) {
7763     PetscValidPointer(*matredundant,6);
7764     PetscValidHeaderSpecific(*matredundant,MAT_COOKIE,6);
7765   }
7766   if (!mat->ops->getredundantmatrix) SETERRQ1(PETSC_ERR_SUP,"Mat type %s",((PetscObject)mat)->type_name);
7767   if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
7768   if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
7769   ierr = MatPreallocated(mat);CHKERRQ(ierr);
7770 
7771   ierr = PetscLogEventBegin(MAT_GetRedundantMatrix,mat,0,0,0);CHKERRQ(ierr);
7772   ierr = (*mat->ops->getredundantmatrix)(mat,nsubcomm,subcomm,mlocal_red,reuse,matredundant);CHKERRQ(ierr);
7773   ierr = PetscLogEventEnd(MAT_GetRedundantMatrix,mat,0,0,0);CHKERRQ(ierr);
7774   PetscFunctionReturn(0);
7775 }
7776