xref: /petsc/src/ksp/pc/impls/jacobi/jacobi.c (revision 7b6bb2c608b6fc6714ef38fda02c2dbb91c82665)
1 
2 /*  --------------------------------------------------------------------
3 
4      This file implements a Jacobi preconditioner in PETSc as part of PC.
5      You can use this as a starting point for implementing your own
6      preconditioner that is not provided with PETSc. (You might also consider
7      just using PCSHELL)
8 
9      The following basic routines are required for each preconditioner.
10           PCCreate_XXX()          - Creates a preconditioner context
11           PCSetFromOptions_XXX()  - Sets runtime options
12           PCApply_XXX()           - Applies the preconditioner
13           PCDestroy_XXX()         - Destroys the preconditioner context
14      where the suffix "_XXX" denotes a particular implementation, in
15      this case we use _Jacobi (e.g., PCCreate_Jacobi, PCApply_Jacobi).
16      These routines are actually called via the common user interface
17      routines PCCreate(), PCSetFromOptions(), PCApply(), and PCDestroy(),
18      so the application code interface remains identical for all
19      preconditioners.
20 
21      Another key routine is:
22           PCSetUp_XXX()           - Prepares for the use of a preconditioner
23      by setting data structures and options.   The interface routine PCSetUp()
24      is not usually called directly by the user, but instead is called by
25      PCApply() if necessary.
26 
27      Additional basic routines are:
28           PCView_XXX()            - Prints details of runtime options that
29                                     have actually been used.
30      These are called by application codes via the interface routines
31      PCView().
32 
33      The various types of solvers (preconditioners, Krylov subspace methods,
34      nonlinear solvers, timesteppers) are all organized similarly, so the
35      above description applies to these categories also.  One exception is
36      that the analogues of PCApply() for these components are KSPSolve(),
37      SNESSolve(), and TSSolve().
38 
39      Additional optional functionality unique to preconditioners is left and
40      right symmetric preconditioner application via PCApplySymmetricLeft()
41      and PCApplySymmetricRight().  The Jacobi implementation is
42      PCApplySymmetricLeftOrRight_Jacobi().
43 
44     -------------------------------------------------------------------- */
45 
46 /*
47    Include files needed for the Jacobi preconditioner:
48      pcimpl.h - private include file intended for use by all preconditioners
49 */
50 
51 #include <private/pcimpl.h>   /*I "petscpc.h" I*/
52 
53 /*
54    Private context (data structure) for the Jacobi preconditioner.
55 */
56 typedef struct {
57   Vec        diag;               /* vector containing the reciprocals of the diagonal elements
58                                     of the preconditioner matrix */
59   Vec        diagsqrt;           /* vector containing the reciprocals of the square roots of
60                                     the diagonal elements of the preconditioner matrix (used
61                                     only for symmetric preconditioner application) */
62   PetscBool  userowmax;
63   PetscBool  userowsum;
64   PetscBool  useabs;             /* use the absolute values of the diagonal entries */
65 } PC_Jacobi;
66 
67 EXTERN_C_BEGIN
68 #undef __FUNCT__
69 #define __FUNCT__ "PCJacobiSetUseRowMax_Jacobi"
70 PetscErrorCode  PCJacobiSetUseRowMax_Jacobi(PC pc)
71 {
72   PC_Jacobi *j;
73 
74   PetscFunctionBegin;
75   j            = (PC_Jacobi*)pc->data;
76   j->userowmax = PETSC_TRUE;
77   PetscFunctionReturn(0);
78 }
79 EXTERN_C_END
80 
81 EXTERN_C_BEGIN
82 #undef __FUNCT__
83 #define __FUNCT__ "PCJacobiSetUseRowSum_Jacobi"
84 PetscErrorCode  PCJacobiSetUseRowSum_Jacobi(PC pc)
85 {
86   PC_Jacobi *j;
87 
88   PetscFunctionBegin;
89   j            = (PC_Jacobi*)pc->data;
90   j->userowsum = PETSC_TRUE;
91   PetscFunctionReturn(0);
92 }
93 EXTERN_C_END
94 
95 EXTERN_C_BEGIN
96 #undef __FUNCT__
97 #define __FUNCT__ "PCJacobiSetUseAbs_Jacobi"
98 PetscErrorCode  PCJacobiSetUseAbs_Jacobi(PC pc)
99 {
100   PC_Jacobi *j;
101 
102   PetscFunctionBegin;
103   j         = (PC_Jacobi*)pc->data;
104   j->useabs = PETSC_TRUE;
105   PetscFunctionReturn(0);
106 }
107 EXTERN_C_END
108 
109 /* -------------------------------------------------------------------------- */
110 /*
111    PCSetUp_Jacobi - Prepares for the use of the Jacobi preconditioner
112                     by setting data structures and options.
113 
114    Input Parameter:
115 .  pc - the preconditioner context
116 
117    Application Interface Routine: PCSetUp()
118 
119    Notes:
120    The interface routine PCSetUp() is not usually called directly by
121    the user, but instead is called by PCApply() if necessary.
122 */
123 #undef __FUNCT__
124 #define __FUNCT__ "PCSetUp_Jacobi"
125 static PetscErrorCode PCSetUp_Jacobi(PC pc)
126 {
127   PC_Jacobi      *jac = (PC_Jacobi*)pc->data;
128   Vec            diag,diagsqrt;
129   PetscErrorCode ierr;
130   PetscInt       n,i;
131   PetscScalar    *x;
132   PetscBool      zeroflag = PETSC_FALSE;
133 
134   PetscFunctionBegin;
135   /*
136        For most preconditioners the code would begin here something like
137 
138   if (pc->setupcalled == 0) { allocate space the first time this is ever called
139     ierr = MatGetVecs(pc->mat,&jac->diag);CHKERRQ(ierr);
140     PetscLogObjectParent(pc,jac->diag);
141   }
142 
143     But for this preconditioner we want to support use of both the matrix' diagonal
144     elements (for left or right preconditioning) and square root of diagonal elements
145     (for symmetric preconditioning).  Hence we do not allocate space here, since we
146     don't know at this point which will be needed (diag and/or diagsqrt) until the user
147     applies the preconditioner, and we don't want to allocate BOTH unless we need
148     them both.  Thus, the diag and diagsqrt are allocated in PCSetUp_Jacobi_NonSymmetric()
149     and PCSetUp_Jacobi_Symmetric(), respectively.
150   */
151 
152   /*
153     Here we set up the preconditioner; that is, we copy the diagonal values from
154     the matrix and put them into a format to make them quick to apply as a preconditioner.
155   */
156   diag     = jac->diag;
157   diagsqrt = jac->diagsqrt;
158 
159   if (diag) {
160     if (jac->userowmax) {
161       ierr = MatGetRowMaxAbs(pc->pmat,diag,PETSC_NULL);CHKERRQ(ierr);
162     } else if (jac->userowsum) {
163       ierr = MatGetRowSum(pc->pmat,diag);CHKERRQ(ierr);
164     } else {
165       ierr = MatGetDiagonal(pc->pmat,diag);CHKERRQ(ierr);
166     }
167     ierr = VecReciprocal(diag);CHKERRQ(ierr);
168     ierr = VecGetLocalSize(diag,&n);CHKERRQ(ierr);
169     ierr = VecGetArray(diag,&x);CHKERRQ(ierr);
170     if (jac->useabs) {
171       for (i=0; i<n; i++) {
172         x[i]     = PetscAbsScalar(x[i]);
173       }
174     }
175     for (i=0; i<n; i++) {
176       if (x[i] == 0.0) {
177         x[i]     = 1.0;
178         zeroflag = PETSC_TRUE;
179       }
180     }
181     ierr = VecRestoreArray(diag,&x);CHKERRQ(ierr);
182   }
183   if (diagsqrt) {
184     if (jac->userowmax) {
185       ierr = MatGetRowMaxAbs(pc->pmat,diagsqrt,PETSC_NULL);CHKERRQ(ierr);
186     } else if (jac->userowsum) {
187       ierr = MatGetRowSum(pc->pmat,diagsqrt);CHKERRQ(ierr);
188     } else {
189       ierr = MatGetDiagonal(pc->pmat,diagsqrt);CHKERRQ(ierr);
190     }
191     ierr = VecGetLocalSize(diagsqrt,&n);CHKERRQ(ierr);
192     ierr = VecGetArray(diagsqrt,&x);CHKERRQ(ierr);
193     for (i=0; i<n; i++) {
194       if (x[i] != 0.0) x[i] = 1.0/sqrt(PetscAbsScalar(x[i]));
195       else {
196         x[i]     = 1.0;
197         zeroflag = PETSC_TRUE;
198       }
199     }
200     ierr = VecRestoreArray(diagsqrt,&x);CHKERRQ(ierr);
201   }
202   if (zeroflag) {
203     ierr = PetscInfo(pc,"Zero detected in diagonal of matrix, using 1 at those locations\n");CHKERRQ(ierr);
204   }
205   PetscFunctionReturn(0);
206 }
207 /* -------------------------------------------------------------------------- */
208 /*
209    PCSetUp_Jacobi_Symmetric - Allocates the vector needed to store the
210    inverse of the square root of the diagonal entries of the matrix.  This
211    is used for symmetric application of the Jacobi preconditioner.
212 
213    Input Parameter:
214 .  pc - the preconditioner context
215 */
216 #undef __FUNCT__
217 #define __FUNCT__ "PCSetUp_Jacobi_Symmetric"
218 static PetscErrorCode PCSetUp_Jacobi_Symmetric(PC pc)
219 {
220   PetscErrorCode ierr;
221   PC_Jacobi      *jac = (PC_Jacobi*)pc->data;
222 
223   PetscFunctionBegin;
224   ierr = MatGetVecs(pc->pmat,&jac->diagsqrt,0);CHKERRQ(ierr);
225   ierr = PetscLogObjectParent(pc,jac->diagsqrt);CHKERRQ(ierr);
226   ierr = PCSetUp_Jacobi(pc);CHKERRQ(ierr);
227   PetscFunctionReturn(0);
228 }
229 /* -------------------------------------------------------------------------- */
230 /*
231    PCSetUp_Jacobi_NonSymmetric - Allocates the vector needed to store the
232    inverse of the diagonal entries of the matrix.  This is used for left of
233    right application of the Jacobi preconditioner.
234 
235    Input Parameter:
236 .  pc - the preconditioner context
237 */
238 #undef __FUNCT__
239 #define __FUNCT__ "PCSetUp_Jacobi_NonSymmetric"
240 static PetscErrorCode PCSetUp_Jacobi_NonSymmetric(PC pc)
241 {
242   PetscErrorCode ierr;
243   PC_Jacobi      *jac = (PC_Jacobi*)pc->data;
244 
245   PetscFunctionBegin;
246   ierr = MatGetVecs(pc->pmat,&jac->diag,0);CHKERRQ(ierr);
247   ierr = PetscLogObjectParent(pc,jac->diag);CHKERRQ(ierr);
248   ierr = PCSetUp_Jacobi(pc);CHKERRQ(ierr);
249   PetscFunctionReturn(0);
250 }
251 /* -------------------------------------------------------------------------- */
252 /*
253    PCApply_Jacobi - Applies the Jacobi preconditioner to a vector.
254 
255    Input Parameters:
256 .  pc - the preconditioner context
257 .  x - input vector
258 
259    Output Parameter:
260 .  y - output vector
261 
262    Application Interface Routine: PCApply()
263  */
264 #undef __FUNCT__
265 #define __FUNCT__ "PCApply_Jacobi"
266 static PetscErrorCode PCApply_Jacobi(PC pc,Vec x,Vec y)
267 {
268   PC_Jacobi      *jac = (PC_Jacobi*)pc->data;
269   PetscErrorCode ierr;
270 
271   PetscFunctionBegin;
272   if (!jac->diag) {
273     ierr = PCSetUp_Jacobi_NonSymmetric(pc);CHKERRQ(ierr);
274   }
275   ierr = VecPointwiseMult(y,x,jac->diag);CHKERRQ(ierr);
276   PetscFunctionReturn(0);
277 }
278 /* -------------------------------------------------------------------------- */
279 /*
280    PCApplySymmetricLeftOrRight_Jacobi - Applies the left or right part of a
281    symmetric preconditioner to a vector.
282 
283    Input Parameters:
284 .  pc - the preconditioner context
285 .  x - input vector
286 
287    Output Parameter:
288 .  y - output vector
289 
290    Application Interface Routines: PCApplySymmetricLeft(), PCApplySymmetricRight()
291 */
292 #undef __FUNCT__
293 #define __FUNCT__ "PCApplySymmetricLeftOrRight_Jacobi"
294 static PetscErrorCode PCApplySymmetricLeftOrRight_Jacobi(PC pc,Vec x,Vec y)
295 {
296   PetscErrorCode ierr;
297   PC_Jacobi      *jac = (PC_Jacobi*)pc->data;
298 
299   PetscFunctionBegin;
300   if (!jac->diagsqrt) {
301     ierr = PCSetUp_Jacobi_Symmetric(pc);CHKERRQ(ierr);
302   }
303   VecPointwiseMult(y,x,jac->diagsqrt);
304   PetscFunctionReturn(0);
305 }
306 /* -------------------------------------------------------------------------- */
307 #undef __FUNCT__
308 #define __FUNCT__ "PCReset_Jacobi"
309 static PetscErrorCode PCReset_Jacobi(PC pc)
310 {
311   PC_Jacobi      *jac = (PC_Jacobi*)pc->data;
312   PetscErrorCode ierr;
313 
314   PetscFunctionBegin;
315   ierr = VecDestroy(&jac->diag);CHKERRQ(ierr);
316   ierr = VecDestroy(&jac->diagsqrt);CHKERRQ(ierr);
317   PetscFunctionReturn(0);
318 }
319 
320 /*
321    PCDestroy_Jacobi - Destroys the private context for the Jacobi preconditioner
322    that was created with PCCreate_Jacobi().
323 
324    Input Parameter:
325 .  pc - the preconditioner context
326 
327    Application Interface Routine: PCDestroy()
328 */
329 #undef __FUNCT__
330 #define __FUNCT__ "PCDestroy_Jacobi"
331 static PetscErrorCode PCDestroy_Jacobi(PC pc)
332 {
333   PetscErrorCode ierr;
334 
335   PetscFunctionBegin;
336   ierr = PCReset_Jacobi(pc);CHKERRQ(ierr);
337 
338   /*
339       Free the private data structure that was hanging off the PC
340   */
341   ierr = PetscFree(pc->data);CHKERRQ(ierr);
342   PetscFunctionReturn(0);
343 }
344 
345 #undef __FUNCT__
346 #define __FUNCT__ "PCSetFromOptions_Jacobi"
347 static PetscErrorCode PCSetFromOptions_Jacobi(PC pc)
348 {
349   PC_Jacobi      *jac = (PC_Jacobi*)pc->data;
350   PetscErrorCode ierr;
351 
352   PetscFunctionBegin;
353   ierr = PetscOptionsHead("Jacobi options");CHKERRQ(ierr);
354     ierr = PetscOptionsBool("-pc_jacobi_rowmax","Use row maximums for diagonal","PCJacobiSetUseRowMax",jac->userowmax,
355                           &jac->userowmax,PETSC_NULL);CHKERRQ(ierr);
356     ierr = PetscOptionsBool("-pc_jacobi_rowsum","Use row sums for diagonal","PCJacobiSetUseRowSum",jac->userowsum,
357                           &jac->userowsum,PETSC_NULL);CHKERRQ(ierr);
358     ierr = PetscOptionsBool("-pc_jacobi_abs","Use absolute values of diagaonal entries","PCJacobiSetUseAbs",jac->useabs,
359                           &jac->useabs,PETSC_NULL);CHKERRQ(ierr);
360   ierr = PetscOptionsTail();CHKERRQ(ierr);
361   PetscFunctionReturn(0);
362 }
363 
364 /* -------------------------------------------------------------------------- */
365 /*
366    PCCreate_Jacobi - Creates a Jacobi preconditioner context, PC_Jacobi,
367    and sets this as the private data within the generic preconditioning
368    context, PC, that was created within PCCreate().
369 
370    Input Parameter:
371 .  pc - the preconditioner context
372 
373    Application Interface Routine: PCCreate()
374 */
375 
376 /*MC
377      PCJACOBI - Jacobi (i.e. diagonal scaling preconditioning)
378 
379    Options Database Key:
380 +    -pc_jacobi_rowmax - use the maximum absolute value in each row as the scaling factor,
381                         rather than the diagonal
382 .    -pc_jacobi_rowsum - use the maximum absolute value in each row as the scaling factor,
383                         rather than the diagonal
384 -    -pc_jacobi_abs - use the absolute value of the diagaonl entry
385 
386    Level: beginner
387 
388   Concepts: Jacobi, diagonal scaling, preconditioners
389 
390   Notes: By using KSPSetPCSide(ksp,PC_SYMMETRIC) or -ksp_pc_side symmetric
391          can scale each side of the matrix by the squareroot of the diagonal entries.
392 
393          Zero entries along the diagonal are replaced with the value 1.0
394 
395 .seealso:  PCCreate(), PCSetType(), PCType (for list of available types), PC,
396            PCJacobiSetUseRowMax(), PCJacobiSetUseRowSum(), PCJacobiSetUseAbs()
397 M*/
398 
399 EXTERN_C_BEGIN
400 #undef __FUNCT__
401 #define __FUNCT__ "PCCreate_Jacobi"
402 PetscErrorCode  PCCreate_Jacobi(PC pc)
403 {
404   PC_Jacobi      *jac;
405   PetscErrorCode ierr;
406 
407   PetscFunctionBegin;
408   /*
409      Creates the private data structure for this preconditioner and
410      attach it to the PC object.
411   */
412   ierr      = PetscNewLog(pc,PC_Jacobi,&jac);CHKERRQ(ierr);
413   pc->data  = (void*)jac;
414 
415   /*
416      Initialize the pointers to vectors to ZERO; these will be used to store
417      diagonal entries of the matrix for fast preconditioner application.
418   */
419   jac->diag          = 0;
420   jac->diagsqrt      = 0;
421   jac->userowmax     = PETSC_FALSE;
422   jac->userowsum     = PETSC_FALSE;
423   jac->useabs        = PETSC_FALSE;
424 
425   /*
426       Set the pointers for the functions that are provided above.
427       Now when the user-level routines (such as PCApply(), PCDestroy(), etc.)
428       are called, they will automatically call these functions.  Note we
429       choose not to provide a couple of these functions since they are
430       not needed.
431   */
432   pc->ops->apply               = PCApply_Jacobi;
433   pc->ops->applytranspose      = PCApply_Jacobi;
434   pc->ops->setup               = PCSetUp_Jacobi;
435   pc->ops->reset               = PCReset_Jacobi;
436   pc->ops->destroy             = PCDestroy_Jacobi;
437   pc->ops->setfromoptions      = PCSetFromOptions_Jacobi;
438   pc->ops->view                = 0;
439   pc->ops->applyrichardson     = 0;
440   pc->ops->applysymmetricleft  = PCApplySymmetricLeftOrRight_Jacobi;
441   pc->ops->applysymmetricright = PCApplySymmetricLeftOrRight_Jacobi;
442   ierr = PetscObjectComposeFunctionDynamic((PetscObject)pc,"PCJacobiSetUseRowMax_C","PCJacobiSetUseRowMax_Jacobi",PCJacobiSetUseRowMax_Jacobi);CHKERRQ(ierr);
443   ierr = PetscObjectComposeFunctionDynamic((PetscObject)pc,"PCJacobiSetUseRowSum_C","PCJacobiSetUseRowSum_Jacobi",PCJacobiSetUseRowSum_Jacobi);CHKERRQ(ierr);
444   ierr = PetscObjectComposeFunctionDynamic((PetscObject)pc,"PCJacobiSetUseAbs_C","PCJacobiSetUseAbs_Jacobi",PCJacobiSetUseAbs_Jacobi);CHKERRQ(ierr);
445   PetscFunctionReturn(0);
446 }
447 EXTERN_C_END
448 
449 
450 #undef __FUNCT__
451 #define __FUNCT__ "PCJacobiSetUseAbs"
452 /*@
453    PCJacobiSetUseAbs - Causes the Jacobi preconditioner to use the
454       absolute value of the diagonal to for the preconditioner
455 
456    Logically Collective on PC
457 
458    Input Parameters:
459 .  pc - the preconditioner context
460 
461 
462    Options Database Key:
463 .  -pc_jacobi_abs
464 
465    Level: intermediate
466 
467    Concepts: Jacobi preconditioner
468 
469 .seealso: PCJacobiaUseRowMax(), PCJacobiaUseRowSum()
470 
471 @*/
472 PetscErrorCode  PCJacobiSetUseAbs(PC pc)
473 {
474   PetscErrorCode ierr;
475 
476   PetscFunctionBegin;
477   PetscValidHeaderSpecific(pc,PC_CLASSID,1);
478   ierr = PetscTryMethod(pc,"PCJacobiSetUseAbs_C",(PC),(pc));CHKERRQ(ierr);
479   PetscFunctionReturn(0);
480 }
481 
482 #undef __FUNCT__
483 #define __FUNCT__ "PCJacobiSetUseRowMax"
484 /*@
485    PCJacobiSetUseRowMax - Causes the Jacobi preconditioner to use the
486       maximum entry in each row as the diagonal preconditioner, instead of
487       the diagonal entry
488 
489    Logically Collective on PC
490 
491    Input Parameters:
492 .  pc - the preconditioner context
493 
494 
495    Options Database Key:
496 .  -pc_jacobi_rowmax
497 
498    Level: intermediate
499 
500    Concepts: Jacobi preconditioner
501 
502 .seealso: PCJacobiaUseAbs()
503 @*/
504 PetscErrorCode  PCJacobiSetUseRowMax(PC pc)
505 {
506   PetscErrorCode ierr;
507 
508   PetscFunctionBegin;
509   PetscValidHeaderSpecific(pc,PC_CLASSID,1);
510   ierr = PetscTryMethod(pc,"PCJacobiSetUseRowMax_C",(PC),(pc));CHKERRQ(ierr);
511   PetscFunctionReturn(0);
512 }
513 
514 #undef __FUNCT__
515 #define __FUNCT__ "PCJacobiSetUseRowSum"
516 /*@
517    PCJacobiSetUseRowSum - Causes the Jacobi preconditioner to use the
518       sum of each row as the diagonal preconditioner, instead of
519       the diagonal entry
520 
521    Logical Collective on PC
522 
523    Input Parameters:
524 .  pc - the preconditioner context
525 
526 
527    Options Database Key:
528 .  -pc_jacobi_rowsum
529 
530    Level: intermediate
531 
532    Concepts: Jacobi preconditioner
533 
534 .seealso: PCJacobiaUseAbs(), PCJacobiaUseRowSum()
535 @*/
536 PetscErrorCode  PCJacobiSetUseRowSum(PC pc)
537 {
538   PetscErrorCode ierr;
539 
540   PetscFunctionBegin;
541   PetscValidHeaderSpecific(pc,PC_CLASSID,1);
542   ierr = PetscTryMethod(pc,"PCJacobiSetUseRowSum_C",(PC),(pc));CHKERRQ(ierr);
543   PetscFunctionReturn(0);
544 }
545 
546