Actual source code: itfunc.c
1: /*
2: Interface KSP routines that the user calls.
3: */
5: #include <petsc/private/kspimpl.h>
6: #include <petsc/private/matimpl.h>
7: #include <petscdm.h>
9: PETSC_STATIC_INLINE PetscErrorCode ObjectView(PetscObject obj, PetscViewer viewer, PetscViewerFormat format)
10: {
13: PetscViewerPushFormat(viewer, format);
14: PetscObjectView(obj, viewer);
15: PetscViewerPopFormat(viewer);
16: return(0);
17: }
19: /*@
20: KSPComputeExtremeSingularValues - Computes the extreme singular values
21: for the preconditioned operator. Called after or during KSPSolve().
23: Not Collective
25: Input Parameter:
26: . ksp - iterative context obtained from KSPCreate()
28: Output Parameters:
29: . emin, emax - extreme singular values
31: Options Database Keys:
32: . -ksp_view_singularvalues - compute extreme singular values and print when KSPSolve completes.
34: Notes:
35: One must call KSPSetComputeSingularValues() before calling KSPSetUp()
36: (or use the option -ksp_view_eigenvalues) in order for this routine to work correctly.
38: Many users may just want to use the monitoring routine
39: KSPMonitorSingularValue() (which can be set with option -ksp_monitor_singular_value)
40: to print the extreme singular values at each iteration of the linear solve.
42: Estimates of the smallest singular value may be very inaccurate, especially if the Krylov method has not converged.
43: The largest singular value is usually accurate to within a few percent if the method has converged, but is still not
44: intended for eigenanalysis.
46: Disable restarts if using KSPGMRES, otherwise this estimate will only be using those iterations after the last
47: restart. See KSPGMRESSetRestart() for more details.
49: Level: advanced
51: .seealso: KSPSetComputeSingularValues(), KSPMonitorSingularValue(), KSPComputeEigenvalues(), KSP
52: @*/
53: PetscErrorCode KSPComputeExtremeSingularValues(KSP ksp,PetscReal *emax,PetscReal *emin)
54: {
61: if (!ksp->calc_sings) SETERRQ(PetscObjectComm((PetscObject)ksp),4,"Singular values not requested before KSPSetUp()");
63: if (ksp->ops->computeextremesingularvalues) {
64: (*ksp->ops->computeextremesingularvalues)(ksp,emax,emin);
65: } else {
66: *emin = -1.0;
67: *emax = -1.0;
68: }
69: return(0);
70: }
72: /*@
73: KSPComputeEigenvalues - Computes the extreme eigenvalues for the
74: preconditioned operator. Called after or during KSPSolve().
76: Not Collective
78: Input Parameters:
79: + ksp - iterative context obtained from KSPCreate()
80: - n - size of arrays r and c. The number of eigenvalues computed (neig) will, in
81: general, be less than this.
83: Output Parameters:
84: + r - real part of computed eigenvalues, provided by user with a dimension of at least n
85: . c - complex part of computed eigenvalues, provided by user with a dimension of at least n
86: - neig - actual number of eigenvalues computed (will be less than or equal to n)
88: Options Database Keys:
89: . -ksp_view_eigenvalues - Prints eigenvalues to stdout
91: Notes:
92: The number of eigenvalues estimated depends on the size of the Krylov space
93: generated during the KSPSolve() ; for example, with
94: CG it corresponds to the number of CG iterations, for GMRES it is the number
95: of GMRES iterations SINCE the last restart. Any extra space in r[] and c[]
96: will be ignored.
98: KSPComputeEigenvalues() does not usually provide accurate estimates; it is
99: intended only for assistance in understanding the convergence of iterative
100: methods, not for eigenanalysis. For accurate computation of eigenvalues we recommend using
101: the excellent package SLEPc.
103: One must call KSPSetComputeEigenvalues() before calling KSPSetUp()
104: in order for this routine to work correctly.
106: Many users may just want to use the monitoring routine
107: KSPMonitorSingularValue() (which can be set with option -ksp_monitor_singular_value)
108: to print the singular values at each iteration of the linear solve.
110: Level: advanced
112: .seealso: KSPSetComputeSingularValues(), KSPMonitorSingularValue(), KSPComputeExtremeSingularValues(), KSP
113: @*/
114: PetscErrorCode KSPComputeEigenvalues(KSP ksp,PetscInt n,PetscReal r[],PetscReal c[],PetscInt *neig)
115: {
122: if (n<0) SETERRQ(PetscObjectComm((PetscObject)ksp),PETSC_ERR_ARG_OUTOFRANGE,"Requested < 0 Eigenvalues");
124: if (!ksp->calc_sings) SETERRQ(PetscObjectComm((PetscObject)ksp),4,"Eigenvalues not requested before KSPSetUp()");
126: if (n && ksp->ops->computeeigenvalues) {
127: (*ksp->ops->computeeigenvalues)(ksp,n,r,c,neig);
128: } else {
129: *neig = 0;
130: }
131: return(0);
132: }
134: /*@
135: KSPComputeRitz - Computes the Ritz or harmonic Ritz pairs associated to the
136: smallest or largest in modulus, for the preconditioned operator.
137: Called after KSPSolve().
139: Not Collective
141: Input Parameters:
142: + ksp - iterative context obtained from KSPCreate()
143: . ritz - PETSC_TRUE or PETSC_FALSE for ritz pairs or harmonic Ritz pairs, respectively
144: . small - PETSC_TRUE or PETSC_FALSE for smallest or largest (harmonic) Ritz values, respectively
145: - nrit - number of (harmonic) Ritz pairs to compute
147: Output Parameters:
148: + nrit - actual number of computed (harmonic) Ritz pairs
149: . S - multidimensional vector with Ritz vectors
150: . tetar - real part of the Ritz values
151: - tetai - imaginary part of the Ritz values
153: Notes:
154: -For GMRES, the (harmonic) Ritz pairs are computed from the Hessenberg matrix obtained during
155: the last complete cycle, or obtained at the end of the solution if the method is stopped before
156: a restart. Then, the number of actual (harmonic) Ritz pairs computed is less or equal to the restart
157: parameter for GMRES if a complete cycle has been performed or less or equal to the number of GMRES
158: iterations.
159: -Moreover, for real matrices, the (harmonic) Ritz pairs are possibly complex-valued. In such a case,
160: the routine selects the complex (harmonic) Ritz value and its conjugate, and two successive columns of S
161: are equal to the real and the imaginary parts of the associated vectors.
162: -the (harmonic) Ritz pairs are given in order of increasing (harmonic) Ritz values in modulus
163: -this is currently not implemented when PETSc is built with complex numbers
165: One must call KSPSetComputeRitz() before calling KSPSetUp()
166: in order for this routine to work correctly.
168: Level: advanced
170: .seealso: KSPSetComputeRitz(), KSP
171: @*/
172: PetscErrorCode KSPComputeRitz(KSP ksp,PetscBool ritz,PetscBool small,PetscInt *nrit,Vec S[],PetscReal tetar[],PetscReal tetai[])
173: {
178: if (!ksp->calc_ritz) SETERRQ(PetscObjectComm((PetscObject)ksp),4,"Ritz pairs not requested before KSPSetUp()");
179: if (ksp->ops->computeritz) {(*ksp->ops->computeritz)(ksp,ritz,small,nrit,S,tetar,tetai);}
180: return(0);
181: }
182: /*@
183: KSPSetUpOnBlocks - Sets up the preconditioner for each block in
184: the block Jacobi, block Gauss-Seidel, and overlapping Schwarz
185: methods.
187: Collective on ksp
189: Input Parameter:
190: . ksp - the KSP context
192: Notes:
193: KSPSetUpOnBlocks() is a routine that the user can optinally call for
194: more precise profiling (via -log_view) of the setup phase for these
195: block preconditioners. If the user does not call KSPSetUpOnBlocks(),
196: it will automatically be called from within KSPSolve().
198: Calling KSPSetUpOnBlocks() is the same as calling PCSetUpOnBlocks()
199: on the PC context within the KSP context.
201: Level: advanced
203: .seealso: PCSetUpOnBlocks(), KSPSetUp(), PCSetUp(), KSP
204: @*/
205: PetscErrorCode KSPSetUpOnBlocks(KSP ksp)
206: {
207: PC pc;
209: PCFailedReason pcreason;
213: KSPGetPC(ksp,&pc);
214: PCSetUpOnBlocks(pc);
215: PCGetFailedReasonRank(pc,&pcreason);
216: /* TODO: this code was wrong and is still wrong, there is no way to propogate the failure to all processes; their is no code to handle a ksp->reason on only some ranks */
217: if (pcreason) {
218: ksp->reason = KSP_DIVERGED_PC_FAILED;
219: }
220: return(0);
221: }
223: /*@
224: KSPSetReusePreconditioner - reuse the current preconditioner, do not construct a new one even if the operator changes
226: Collective on ksp
228: Input Parameters:
229: + ksp - iterative context obtained from KSPCreate()
230: - flag - PETSC_TRUE to reuse the current preconditioner
232: Level: intermediate
234: .seealso: KSPCreate(), KSPSolve(), KSPDestroy(), PCSetReusePreconditioner(), KSP
235: @*/
236: PetscErrorCode KSPSetReusePreconditioner(KSP ksp,PetscBool flag)
237: {
238: PC pc;
243: KSPGetPC(ksp,&pc);
244: PCSetReusePreconditioner(pc,flag);
245: return(0);
246: }
248: /*@
249: KSPGetReusePreconditioner - Determines if the KSP reuses the current preconditioner even if the operator in the preconditioner has changed.
251: Collective on ksp
253: Input Parameters:
254: . ksp - iterative context obtained from KSPCreate()
256: Output Parameters:
257: . flag - the boolean flag
259: Level: intermediate
261: .seealso: KSPCreate(), KSPSolve(), KSPDestroy(), KSPSetReusePreconditioner(), KSP
262: @*/
263: PetscErrorCode KSPGetReusePreconditioner(KSP ksp,PetscBool *flag)
264: {
270: *flag = PETSC_FALSE;
271: if (ksp->pc) {
272: PCGetReusePreconditioner(ksp->pc,flag);
273: }
274: return(0);
275: }
277: /*@
278: KSPSetSkipPCSetFromOptions - prevents KSPSetFromOptions() from call PCSetFromOptions(). This is used if the same PC is shared by more than one KSP so its options are not resetable for each KSP
280: Collective on ksp
282: Input Parameters:
283: + ksp - iterative context obtained from KSPCreate()
284: - flag - PETSC_TRUE to skip calling the PCSetFromOptions()
286: Level: intermediate
288: .seealso: KSPCreate(), KSPSolve(), KSPDestroy(), PCSetReusePreconditioner(), KSP
289: @*/
290: PetscErrorCode KSPSetSkipPCSetFromOptions(KSP ksp,PetscBool flag)
291: {
294: ksp->skippcsetfromoptions = flag;
295: return(0);
296: }
298: /*@
299: KSPSetUp - Sets up the internal data structures for the
300: later use of an iterative solver.
302: Collective on ksp
304: Input Parameter:
305: . ksp - iterative context obtained from KSPCreate()
307: Level: developer
309: .seealso: KSPCreate(), KSPSolve(), KSPDestroy(), KSP
310: @*/
311: PetscErrorCode KSPSetUp(KSP ksp)
312: {
314: Mat A,B;
315: Mat mat,pmat;
316: MatNullSpace nullsp;
317: PCFailedReason pcreason;
322: /* reset the convergence flag from the previous solves */
323: ksp->reason = KSP_CONVERGED_ITERATING;
325: if (!((PetscObject)ksp)->type_name) {
326: KSPSetType(ksp,KSPGMRES);
327: }
328: KSPSetUpNorms_Private(ksp,PETSC_TRUE,&ksp->normtype,&ksp->pc_side);
330: if (ksp->dmActive && !ksp->setupstage) {
331: /* first time in so build matrix and vector data structures using DM */
332: if (!ksp->vec_rhs) {DMCreateGlobalVector(ksp->dm,&ksp->vec_rhs);}
333: if (!ksp->vec_sol) {DMCreateGlobalVector(ksp->dm,&ksp->vec_sol);}
334: DMCreateMatrix(ksp->dm,&A);
335: KSPSetOperators(ksp,A,A);
336: PetscObjectDereference((PetscObject)A);
337: }
339: if (ksp->dmActive) {
340: DMKSP kdm;
341: DMGetDMKSP(ksp->dm,&kdm);
343: if (kdm->ops->computeinitialguess && ksp->setupstage != KSP_SETUP_NEWRHS) {
344: /* only computes initial guess the first time through */
345: (*kdm->ops->computeinitialguess)(ksp,ksp->vec_sol,kdm->initialguessctx);
346: KSPSetInitialGuessNonzero(ksp,PETSC_TRUE);
347: }
348: if (kdm->ops->computerhs) {
349: (*kdm->ops->computerhs)(ksp,ksp->vec_rhs,kdm->rhsctx);
350: }
352: if (ksp->setupstage != KSP_SETUP_NEWRHS) {
353: if (kdm->ops->computeoperators) {
354: KSPGetOperators(ksp,&A,&B);
355: (*kdm->ops->computeoperators)(ksp,A,B,kdm->operatorsctx);
356: } else SETERRQ(PetscObjectComm((PetscObject)ksp),PETSC_ERR_ARG_WRONGSTATE,"You called KSPSetDM() but did not use DMKSPSetComputeOperators() or KSPSetDMActive(ksp,PETSC_FALSE);");
357: }
358: }
360: if (ksp->setupstage == KSP_SETUP_NEWRHS) return(0);
361: PetscLogEventBegin(KSP_SetUp,ksp,ksp->vec_rhs,ksp->vec_sol,0);
363: switch (ksp->setupstage) {
364: case KSP_SETUP_NEW:
365: (*ksp->ops->setup)(ksp);
366: break;
367: case KSP_SETUP_NEWMATRIX: { /* This should be replaced with a more general mechanism */
368: if (ksp->setupnewmatrix) {
369: (*ksp->ops->setup)(ksp);
370: }
371: } break;
372: default: break;
373: }
375: if (!ksp->pc) {KSPGetPC(ksp,&ksp->pc);}
376: PCGetOperators(ksp->pc,&mat,&pmat);
377: /* scale the matrix if requested */
378: if (ksp->dscale) {
379: PetscScalar *xx;
380: PetscInt i,n;
381: PetscBool zeroflag = PETSC_FALSE;
382: if (!ksp->pc) {KSPGetPC(ksp,&ksp->pc);}
383: if (!ksp->diagonal) { /* allocate vector to hold diagonal */
384: MatCreateVecs(pmat,&ksp->diagonal,NULL);
385: }
386: MatGetDiagonal(pmat,ksp->diagonal);
387: VecGetLocalSize(ksp->diagonal,&n);
388: VecGetArray(ksp->diagonal,&xx);
389: for (i=0; i<n; i++) {
390: if (xx[i] != 0.0) xx[i] = 1.0/PetscSqrtReal(PetscAbsScalar(xx[i]));
391: else {
392: xx[i] = 1.0;
393: zeroflag = PETSC_TRUE;
394: }
395: }
396: VecRestoreArray(ksp->diagonal,&xx);
397: if (zeroflag) {
398: PetscInfo(ksp,"Zero detected in diagonal of matrix, using 1 at those locations\n");
399: }
400: MatDiagonalScale(pmat,ksp->diagonal,ksp->diagonal);
401: if (mat != pmat) {MatDiagonalScale(mat,ksp->diagonal,ksp->diagonal);}
402: ksp->dscalefix2 = PETSC_FALSE;
403: }
404: PetscLogEventEnd(KSP_SetUp,ksp,ksp->vec_rhs,ksp->vec_sol,0);
405: PCSetErrorIfFailure(ksp->pc,ksp->errorifnotconverged);
406: PCSetUp(ksp->pc);
407: PCGetFailedReasonRank(ksp->pc,&pcreason);
408: /* TODO: this code was wrong and is still wrong, there is no way to propogate the failure to all processes; their is no code to handle a ksp->reason on only some ranks */
409: if (pcreason) {
410: ksp->reason = KSP_DIVERGED_PC_FAILED;
411: }
413: MatGetNullSpace(mat,&nullsp);
414: if (nullsp) {
415: PetscBool test = PETSC_FALSE;
416: PetscOptionsGetBool(((PetscObject)ksp)->options,((PetscObject)ksp)->prefix,"-ksp_test_null_space",&test,NULL);
417: if (test) {
418: MatNullSpaceTest(nullsp,mat,NULL);
419: }
420: }
421: ksp->setupstage = KSP_SETUP_NEWRHS;
422: return(0);
423: }
425: /*@C
426: KSPConvergedReasonView - Displays the reason a KSP solve converged or diverged to a viewer
428: Collective on ksp
430: Parameter:
431: + ksp - iterative context obtained from KSPCreate()
432: - viewer - the viewer to display the reason
434: Options Database Keys:
435: + -ksp_converged_reason - print reason for converged or diverged, also prints number of iterations
436: - -ksp_converged_reason ::failed - only print reason and number of iterations when diverged
438: Notes:
439: To change the format of the output call PetscViewerPushFormat(viewer,format) before this call. Use PETSC_VIEWER_DEFAULT for the default,
440: use PETSC_VIEWER_FAILED to only display a reason if it fails.
442: Level: beginner
444: .seealso: KSPCreate(), KSPSetUp(), KSPDestroy(), KSPSetTolerances(), KSPConvergedDefault(),
445: KSPSolveTranspose(), KSPGetIterationNumber(), KSP, KSPGetConvergedReason(), PetscViewerPushFormat(), PetscViewerPopFormat()
446: @*/
447: PetscErrorCode KSPConvergedReasonView(KSP ksp, PetscViewer viewer)
448: {
449: PetscErrorCode ierr;
450: PetscBool isAscii;
451: PetscViewerFormat format;
454: if (!viewer) viewer = PETSC_VIEWER_STDOUT_(PetscObjectComm((PetscObject)ksp));
455: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&isAscii);
456: if (isAscii) {
457: PetscViewerGetFormat(viewer, &format);
458: PetscViewerASCIIAddTab(viewer,((PetscObject)ksp)->tablevel);
459: if (ksp->reason > 0 && format != PETSC_VIEWER_FAILED) {
460: if (((PetscObject) ksp)->prefix) {
461: PetscViewerASCIIPrintf(viewer,"Linear %s solve converged due to %s iterations %D\n",((PetscObject) ksp)->prefix,KSPConvergedReasons[ksp->reason],ksp->its);
462: } else {
463: PetscViewerASCIIPrintf(viewer,"Linear solve converged due to %s iterations %D\n",KSPConvergedReasons[ksp->reason],ksp->its);
464: }
465: } else if (ksp->reason <= 0) {
466: if (((PetscObject) ksp)->prefix) {
467: PetscViewerASCIIPrintf(viewer,"Linear %s solve did not converge due to %s iterations %D\n",((PetscObject) ksp)->prefix,KSPConvergedReasons[ksp->reason],ksp->its);
468: } else {
469: PetscViewerASCIIPrintf(viewer,"Linear solve did not converge due to %s iterations %D\n",KSPConvergedReasons[ksp->reason],ksp->its);
470: }
471: if (ksp->reason == KSP_DIVERGED_PC_FAILED) {
472: PCFailedReason reason;
473: PCGetFailedReason(ksp->pc,&reason);
474: PetscViewerASCIIPrintf(viewer," PC failed due to %s \n",PCFailedReasons[reason]);
475: }
476: }
477: PetscViewerASCIISubtractTab(viewer,((PetscObject)ksp)->tablevel);
478: }
479: return(0);
480: }
482: /*@C
483: KSPConvergedReasonViewSet - Sets an ADDITIONAL function that is to be used at the
484: end of the linear solver to display the convergence reason of the linear solver.
486: Logically Collective on KSP
488: Input Parameters:
489: + ksp - the KSP context
490: . f - the ksp converged reason view function
491: . vctx - [optional] user-defined context for private data for the
492: ksp converged reason view routine (use NULL if no context is desired)
493: - reasonviewdestroy - [optional] routine that frees reasonview context
494: (may be NULL)
496: Options Database Keys:
497: + -ksp_converged_reason - sets a default KSPConvergedReasonView()
498: - -ksp_converged_reason_view_cancel - cancels all converged reason viewers that have
499: been hardwired into a code by
500: calls to KSPConvergedReasonViewSet(), but
501: does not cancel those set via
502: the options database.
504: Notes:
505: Several different converged reason view routines may be set by calling
506: KSPConvergedReasonViewSet() multiple times; all will be called in the
507: order in which they were set.
509: Level: intermediate
511: .seealso: KSPConvergedReasonView(), KSPConvergedReasonViewCancel()
512: @*/
513: PetscErrorCode KSPConvergedReasonViewSet(KSP ksp,PetscErrorCode (*f)(KSP,void*),void *vctx,PetscErrorCode (*reasonviewdestroy)(void**))
514: {
515: PetscInt i;
517: PetscBool identical;
521: for (i=0; i<ksp->numberreasonviews;i++) {
522: PetscMonitorCompare((PetscErrorCode (*)(void))f,vctx,reasonviewdestroy,(PetscErrorCode (*)(void))ksp->reasonview[i],ksp->reasonviewcontext[i],ksp->reasonviewdestroy[i],&identical);
523: if (identical) return(0);
524: }
525: if (ksp->numberreasonviews >= MAXKSPREASONVIEWS) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Too many KSP reasonview set");
526: ksp->reasonview[ksp->numberreasonviews] = f;
527: ksp->reasonviewdestroy[ksp->numberreasonviews] = reasonviewdestroy;
528: ksp->reasonviewcontext[ksp->numberreasonviews++] = (void*)vctx;
529: return(0);
530: }
532: /*@
533: KSPConvergedReasonViewCancel - Clears all the reasonview functions for a KSP object.
535: Collective on KSP
537: Input Parameter:
538: . ksp - iterative context obtained from KSPCreate()
540: Level: intermediate
542: .seealso: KSPCreate(), KSPDestroy(), KSPReset()
543: @*/
544: PetscErrorCode KSPConvergedReasonViewCancel(KSP ksp)
545: {
547: PetscInt i;
551: for (i=0; i<ksp->numberreasonviews; i++) {
552: if (ksp->reasonviewdestroy[i]) {
553: (*ksp->reasonviewdestroy[i])(&ksp->reasonviewcontext[i]);
554: }
555: }
556: ksp->numberreasonviews = 0;
557: return(0);
558: }
560: /*@
561: KSPConvergedReasonViewFromOptions - Processes command line options to determine if/how a KSPReason is to be viewed.
563: Collective on ksp
565: Input Parameters:
566: . ksp - the KSP object
568: Level: intermediate
570: .seealso: KSPConvergedReasonView()
571: @*/
572: PetscErrorCode KSPConvergedReasonViewFromOptions(KSP ksp)
573: {
574: PetscViewer viewer;
575: PetscBool flg;
576: PetscViewerFormat format;
577: PetscErrorCode ierr;
578: PetscInt i;
582: /* Call all user-provided reason review routines */
583: for (i=0; i<ksp->numberreasonviews; i++) {
584: (*ksp->reasonview[i])(ksp,ksp->reasonviewcontext[i]);
585: }
587: /* Call the default PETSc routine */
588: PetscOptionsGetViewer(PetscObjectComm((PetscObject)ksp),((PetscObject)ksp)->options,((PetscObject)ksp)->prefix,"-ksp_converged_reason",&viewer,&format,&flg);
589: if (flg) {
590: PetscViewerPushFormat(viewer,format);
591: KSPConvergedReasonView(ksp, viewer);
592: PetscViewerPopFormat(viewer);
593: PetscViewerDestroy(&viewer);
594: }
595: return(0);
596: }
598: /*@C
599: KSPConvergedRateView - Displays the reason a KSP solve converged or diverged to a viewer
601: Collective on ksp
603: Input Parameters:
604: + ksp - iterative context obtained from KSPCreate()
605: - viewer - the viewer to display the reason
607: Options Database Keys:
608: . -ksp_converged_rate - print reason for convergence or divergence and the convergence rate (or 0.0 for divergence)
610: Notes:
611: To change the format of the output, call PetscViewerPushFormat(viewer,format) before this call.
613: Suppose that the residual is reduced linearly, $r_k = c^k r_0$, which means $log r_k = log r_0 + k log c$. After linear regression,
614: the slope is $\log c$. The coefficient of determination is given by $1 - \frac{\sum_i (y_i - f(x_i))^2}{\sum_i (y_i - \bar y)}$,
615: see also https://en.wikipedia.org/wiki/Coefficient_of_determination
617: Level: intermediate
619: .seealso: KSPConvergedReasonView(), KSPGetConvergedRate(), KSPSetTolerances(), KSPConvergedDefault()
620: @*/
621: PetscErrorCode KSPConvergedRateView(KSP ksp, PetscViewer viewer)
622: {
623: PetscViewerFormat format;
624: PetscBool isAscii;
625: PetscReal rrate, rRsq, erate = 0.0, eRsq = 0.0;
626: PetscInt its;
627: const char *prefix, *reason = KSPConvergedReasons[ksp->reason];
628: PetscErrorCode ierr;
631: KSPGetOptionsPrefix(ksp, &prefix);
632: KSPGetIterationNumber(ksp, &its);
633: KSPComputeConvergenceRate(ksp, &rrate, &rRsq, &erate, &eRsq);
634: if (!viewer) viewer = PETSC_VIEWER_STDOUT_(PetscObjectComm((PetscObject)ksp));
635: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&isAscii);
636: if (isAscii) {
637: PetscViewerGetFormat(viewer, &format);
638: PetscViewerASCIIAddTab(viewer,((PetscObject)ksp)->tablevel);
639: if (ksp->reason > 0) {
640: if (prefix) {PetscViewerASCIIPrintf(viewer, "Linear %s solve converged due to %s iterations %D", prefix, reason, its);}
641: else {PetscViewerASCIIPrintf(viewer, "Linear solve converged due to %s iterations %D", reason, its);}
642: PetscViewerASCIIUseTabs(viewer, PETSC_FALSE);
643: if (rRsq >= 0.0) {PetscViewerASCIIPrintf(viewer, " res rate %g R^2 %g", rrate, rRsq);}
644: if (eRsq >= 0.0) {PetscViewerASCIIPrintf(viewer, " error rate %g R^2 %g", erate, eRsq);}
645: PetscViewerASCIIPrintf(viewer, "\n");
646: PetscViewerASCIIUseTabs(viewer, PETSC_TRUE);
647: } else if (ksp->reason <= 0) {
648: if (prefix) {PetscViewerASCIIPrintf(viewer, "Linear %s solve did not converge due to %s iterations %D", prefix, reason, its);}
649: else {PetscViewerASCIIPrintf(viewer, "Linear solve did not converge due to %s iterations %D", reason, its);}
650: PetscViewerASCIIUseTabs(viewer, PETSC_FALSE);
651: if (rRsq >= 0.0) {PetscViewerASCIIPrintf(viewer, " res rate %g R^2 %g", rrate, rRsq);}
652: if (eRsq >= 0.0) {PetscViewerASCIIPrintf(viewer, " error rate %g R^2 %g", erate, eRsq);}
653: PetscViewerASCIIPrintf(viewer, "\n");
654: PetscViewerASCIIUseTabs(viewer, PETSC_TRUE);
655: if (ksp->reason == KSP_DIVERGED_PC_FAILED) {
656: PCFailedReason reason;
657: PCGetFailedReason(ksp->pc,&reason);
658: PetscViewerASCIIPrintf(viewer," PC failed due to %s \n",PCFailedReasons[reason]);
659: }
660: }
661: PetscViewerASCIISubtractTab(viewer,((PetscObject)ksp)->tablevel);
662: }
663: return(0);
664: }
666: #include <petscdraw.h>
668: static PetscErrorCode KSPViewEigenvalues_Internal(KSP ksp, PetscBool isExplicit, PetscViewer viewer, PetscViewerFormat format)
669: {
670: PetscReal *r, *c;
671: PetscInt n, i, neig;
672: PetscBool isascii, isdraw;
673: PetscMPIInt rank;
677: MPI_Comm_rank(PetscObjectComm((PetscObject) ksp), &rank);
678: PetscObjectTypeCompare((PetscObject) viewer, PETSCVIEWERASCII, &isascii);
679: PetscObjectTypeCompare((PetscObject) viewer, PETSCVIEWERDRAW, &isdraw);
680: if (isExplicit) {
681: VecGetSize(ksp->vec_sol,&n);
682: PetscMalloc2(n, &r, n, &c);
683: KSPComputeEigenvaluesExplicitly(ksp, n, r, c);
684: neig = n;
685: } else {
686: PetscInt nits;
688: KSPGetIterationNumber(ksp, &nits);
689: n = nits+2;
690: if (!nits) {PetscViewerASCIIPrintf(viewer, "Zero iterations in solver, cannot approximate any eigenvalues\n");return(0);}
691: PetscMalloc2(n, &r, n, &c);
692: KSPComputeEigenvalues(ksp, n, r, c, &neig);
693: }
694: if (isascii) {
695: PetscViewerASCIIPrintf(viewer, "%s computed eigenvalues\n", isExplicit ? "Explicitly" : "Iteratively");
696: for (i = 0; i < neig; ++i) {
697: if (c[i] >= 0.0) {PetscViewerASCIIPrintf(viewer, "%g + %gi\n", (double) r[i], (double) c[i]);}
698: else {PetscViewerASCIIPrintf(viewer, "%g - %gi\n", (double) r[i], -(double) c[i]);}
699: }
700: } else if (isdraw && !rank) {
701: PetscDraw draw;
702: PetscDrawSP drawsp;
704: if (format == PETSC_VIEWER_DRAW_CONTOUR) {
705: KSPPlotEigenContours_Private(ksp,neig,r,c);
706: } else {
707: if (!ksp->eigviewer) {PetscViewerDrawOpen(PETSC_COMM_SELF,NULL,isExplicit ? "Explicitly Computed Eigenvalues" : "Iteratively Computed Eigenvalues",PETSC_DECIDE,PETSC_DECIDE,400,400,&ksp->eigviewer);}
708: PetscViewerDrawGetDraw(ksp->eigviewer,0,&draw);
709: PetscDrawSPCreate(draw,1,&drawsp);
710: PetscDrawSPReset(drawsp);
711: for (i = 0; i < neig; ++i) {PetscDrawSPAddPoint(drawsp,r+i,c+i);}
712: PetscDrawSPDraw(drawsp,PETSC_TRUE);
713: PetscDrawSPSave(drawsp);
714: PetscDrawSPDestroy(&drawsp);
715: }
716: }
717: PetscFree2(r, c);
718: return(0);
719: }
721: static PetscErrorCode KSPViewSingularvalues_Internal(KSP ksp, PetscViewer viewer, PetscViewerFormat format)
722: {
723: PetscReal smax, smin;
724: PetscInt nits;
725: PetscBool isascii;
729: PetscObjectTypeCompare((PetscObject) viewer, PETSCVIEWERASCII, &isascii);
730: KSPGetIterationNumber(ksp, &nits);
731: if (!nits) {PetscViewerASCIIPrintf(viewer, "Zero iterations in solver, cannot approximate any singular values\n");return(0);}
732: KSPComputeExtremeSingularValues(ksp, &smax, &smin);
733: if (isascii) {PetscViewerASCIIPrintf(viewer, "Iteratively computed extreme singular values: max %g min %g max/min %g\n",(double)smax,(double)smin,(double)(smax/smin));}
734: return(0);
735: }
737: static PetscErrorCode KSPViewFinalResidual_Internal(KSP ksp, PetscViewer viewer, PetscViewerFormat format)
738: {
739: PetscBool isascii;
743: PetscObjectTypeCompare((PetscObject) viewer, PETSCVIEWERASCII, &isascii);
744: if (ksp->dscale && !ksp->dscalefix) SETERRQ(PetscObjectComm((PetscObject) ksp), PETSC_ERR_ARG_WRONGSTATE, "Cannot compute final scale with -ksp_diagonal_scale except also with -ksp_diagonal_scale_fix");
745: if (isascii) {
746: Mat A;
747: Vec t;
748: PetscReal norm;
750: PCGetOperators(ksp->pc, &A, NULL);
751: VecDuplicate(ksp->vec_rhs, &t);
752: KSP_MatMult(ksp, A, ksp->vec_sol, t);
753: VecAYPX(t, -1.0, ksp->vec_rhs);
754: VecNorm(t, NORM_2, &norm);
755: VecDestroy(&t);
756: PetscViewerASCIIPrintf(viewer, "KSP final norm of residual %g\n", (double) norm);
757: }
758: return(0);
759: }
761: static PetscErrorCode KSPMonitorPauseFinal_Internal(KSP ksp)
762: {
763: PetscInt i;
767: if (!ksp->pauseFinal) return(0);
768: for (i = 0; i < ksp->numbermonitors; ++i) {
769: PetscViewerAndFormat *vf = (PetscViewerAndFormat *) ksp->monitorcontext[i];
770: PetscDraw draw;
771: PetscReal lpause;
773: if (!vf) continue;
774: if (vf->lg) {
776: if (((PetscObject) vf->lg)->classid != PETSC_DRAWLG_CLASSID) continue;
777: PetscDrawLGGetDraw(vf->lg, &draw);
778: PetscDrawGetPause(draw, &lpause);
779: PetscDrawSetPause(draw, -1.0);
780: PetscDrawPause(draw);
781: PetscDrawSetPause(draw, lpause);
782: } else {
783: PetscBool isdraw;
786: if (((PetscObject) vf->viewer)->classid != PETSC_VIEWER_CLASSID) continue;
787: PetscObjectTypeCompare((PetscObject) vf->viewer, PETSCVIEWERDRAW, &isdraw);
788: if (!isdraw) continue;
789: PetscViewerDrawGetDraw(vf->viewer, 0, &draw);
790: PetscDrawGetPause(draw, &lpause);
791: PetscDrawSetPause(draw, -1.0);
792: PetscDrawPause(draw);
793: PetscDrawSetPause(draw, lpause);
794: }
795: }
796: return(0);
797: }
799: static PetscErrorCode KSPSolve_Private(KSP ksp,Vec b,Vec x)
800: {
802: PetscBool flg = PETSC_FALSE,inXisinB=PETSC_FALSE,guess_zero;
803: Mat mat,pmat;
804: MPI_Comm comm;
805: MatNullSpace nullsp;
806: Vec btmp,vec_rhs=NULL;
809: comm = PetscObjectComm((PetscObject)ksp);
810: if (x && x == b) {
811: if (!ksp->guess_zero) SETERRQ(comm,PETSC_ERR_ARG_INCOMP,"Cannot use x == b with nonzero initial guess");
812: VecDuplicate(b,&x);
813: inXisinB = PETSC_TRUE;
814: }
815: if (b) {
816: PetscObjectReference((PetscObject)b);
817: VecDestroy(&ksp->vec_rhs);
818: ksp->vec_rhs = b;
819: }
820: if (x) {
821: PetscObjectReference((PetscObject)x);
822: VecDestroy(&ksp->vec_sol);
823: ksp->vec_sol = x;
824: }
826: if (ksp->viewPre) {ObjectView((PetscObject) ksp, ksp->viewerPre, ksp->formatPre);}
828: if (ksp->presolve) {(*ksp->presolve)(ksp,ksp->vec_rhs,ksp->vec_sol,ksp->prectx);}
830: /* reset the residual history list if requested */
831: if (ksp->res_hist_reset) ksp->res_hist_len = 0;
832: if (ksp->err_hist_reset) ksp->err_hist_len = 0;
834: if (ksp->guess) {
835: PetscObjectState ostate,state;
837: KSPGuessSetUp(ksp->guess);
838: PetscObjectStateGet((PetscObject)ksp->vec_sol,&ostate);
839: KSPGuessFormGuess(ksp->guess,ksp->vec_rhs,ksp->vec_sol);
840: PetscObjectStateGet((PetscObject)ksp->vec_sol,&state);
841: if (state != ostate) {
842: ksp->guess_zero = PETSC_FALSE;
843: } else {
844: PetscInfo(ksp,"Using zero initial guess since the KSPGuess object did not change the vector\n");
845: ksp->guess_zero = PETSC_TRUE;
846: }
847: }
849: /* KSPSetUp() scales the matrix if needed */
850: KSPSetUp(ksp);
851: KSPSetUpOnBlocks(ksp);
853: VecSetErrorIfLocked(ksp->vec_sol,3);
855: PetscLogEventBegin(KSP_Solve,ksp,ksp->vec_rhs,ksp->vec_sol,0);
856: PCGetOperators(ksp->pc,&mat,&pmat);
857: /* diagonal scale RHS if called for */
858: if (ksp->dscale) {
859: VecPointwiseMult(ksp->vec_rhs,ksp->vec_rhs,ksp->diagonal);
860: /* second time in, but matrix was scaled back to original */
861: if (ksp->dscalefix && ksp->dscalefix2) {
862: Mat mat,pmat;
864: PCGetOperators(ksp->pc,&mat,&pmat);
865: MatDiagonalScale(pmat,ksp->diagonal,ksp->diagonal);
866: if (mat != pmat) {MatDiagonalScale(mat,ksp->diagonal,ksp->diagonal);}
867: }
869: /* scale initial guess */
870: if (!ksp->guess_zero) {
871: if (!ksp->truediagonal) {
872: VecDuplicate(ksp->diagonal,&ksp->truediagonal);
873: VecCopy(ksp->diagonal,ksp->truediagonal);
874: VecReciprocal(ksp->truediagonal);
875: }
876: VecPointwiseMult(ksp->vec_sol,ksp->vec_sol,ksp->truediagonal);
877: }
878: }
879: PCPreSolve(ksp->pc,ksp);
881: if (ksp->guess_zero) { VecSet(ksp->vec_sol,0.0);}
882: if (ksp->guess_knoll) { /* The Knoll trick is independent on the KSPGuess specified */
883: PCApply(ksp->pc,ksp->vec_rhs,ksp->vec_sol);
884: KSP_RemoveNullSpace(ksp,ksp->vec_sol);
885: ksp->guess_zero = PETSC_FALSE;
886: }
888: /* can we mark the initial guess as zero for this solve? */
889: guess_zero = ksp->guess_zero;
890: if (!ksp->guess_zero) {
891: PetscReal norm;
893: VecNormAvailable(ksp->vec_sol,NORM_2,&flg,&norm);
894: if (flg && !norm) ksp->guess_zero = PETSC_TRUE;
895: }
896: if (ksp->transpose_solve) {
897: MatGetNullSpace(pmat,&nullsp);
898: } else {
899: MatGetTransposeNullSpace(pmat,&nullsp);
900: }
901: if (nullsp) {
902: VecDuplicate(ksp->vec_rhs,&btmp);
903: VecCopy(ksp->vec_rhs,btmp);
904: MatNullSpaceRemove(nullsp,btmp);
905: vec_rhs = ksp->vec_rhs;
906: ksp->vec_rhs = btmp;
907: }
908: VecLockReadPush(ksp->vec_rhs);
909: if (ksp->reason == KSP_DIVERGED_PC_FAILED) {
910: VecSetInf(ksp->vec_sol);
911: }
912: (*ksp->ops->solve)(ksp);
913: KSPMonitorPauseFinal_Internal(ksp);
915: VecLockReadPop(ksp->vec_rhs);
916: if (nullsp) {
917: ksp->vec_rhs = vec_rhs;
918: VecDestroy(&btmp);
919: }
921: ksp->guess_zero = guess_zero;
923: if (!ksp->reason) SETERRQ(comm,PETSC_ERR_PLIB,"Internal error, solver returned without setting converged reason");
924: ksp->totalits += ksp->its;
926: KSPConvergedReasonViewFromOptions(ksp);
928: if (ksp->viewRate) {
929: PetscViewerPushFormat(ksp->viewerRate,ksp->formatRate);
930: KSPConvergedRateView(ksp, ksp->viewerRate);
931: PetscViewerPopFormat(ksp->viewerRate);
932: }
933: PCPostSolve(ksp->pc,ksp);
935: /* diagonal scale solution if called for */
936: if (ksp->dscale) {
937: VecPointwiseMult(ksp->vec_sol,ksp->vec_sol,ksp->diagonal);
938: /* unscale right hand side and matrix */
939: if (ksp->dscalefix) {
940: Mat mat,pmat;
942: VecReciprocal(ksp->diagonal);
943: VecPointwiseMult(ksp->vec_rhs,ksp->vec_rhs,ksp->diagonal);
944: PCGetOperators(ksp->pc,&mat,&pmat);
945: MatDiagonalScale(pmat,ksp->diagonal,ksp->diagonal);
946: if (mat != pmat) {MatDiagonalScale(mat,ksp->diagonal,ksp->diagonal);}
947: VecReciprocal(ksp->diagonal);
948: ksp->dscalefix2 = PETSC_TRUE;
949: }
950: }
951: PetscLogEventEnd(KSP_Solve,ksp,ksp->vec_rhs,ksp->vec_sol,0);
952: if (ksp->guess) {
953: KSPGuessUpdate(ksp->guess,ksp->vec_rhs,ksp->vec_sol);
954: }
955: if (ksp->postsolve) {
956: (*ksp->postsolve)(ksp,ksp->vec_rhs,ksp->vec_sol,ksp->postctx);
957: }
959: PCGetOperators(ksp->pc,&mat,&pmat);
960: if (ksp->viewEV) {KSPViewEigenvalues_Internal(ksp, PETSC_FALSE, ksp->viewerEV, ksp->formatEV);}
961: if (ksp->viewEVExp) {KSPViewEigenvalues_Internal(ksp, PETSC_TRUE, ksp->viewerEVExp, ksp->formatEVExp);}
962: if (ksp->viewSV) {KSPViewSingularvalues_Internal(ksp, ksp->viewerSV, ksp->formatSV);}
963: if (ksp->viewFinalRes) {KSPViewFinalResidual_Internal(ksp, ksp->viewerFinalRes, ksp->formatFinalRes);}
964: if (ksp->viewMat) {ObjectView((PetscObject) mat, ksp->viewerMat, ksp->formatMat);}
965: if (ksp->viewPMat) {ObjectView((PetscObject) pmat, ksp->viewerPMat, ksp->formatPMat);}
966: if (ksp->viewRhs) {ObjectView((PetscObject) ksp->vec_rhs, ksp->viewerRhs, ksp->formatRhs);}
967: if (ksp->viewSol) {ObjectView((PetscObject) ksp->vec_sol, ksp->viewerSol, ksp->formatSol);}
968: if (ksp->view) {ObjectView((PetscObject) ksp, ksp->viewer, ksp->format);}
969: if (ksp->viewDScale) {ObjectView((PetscObject) ksp->diagonal, ksp->viewerDScale, ksp->formatDScale);}
970: if (ksp->viewMatExp) {
971: Mat A, B;
973: PCGetOperators(ksp->pc, &A, NULL);
974: if (ksp->transpose_solve) {
975: Mat AT;
977: MatCreateTranspose(A, &AT);
978: MatComputeOperator(AT, MATAIJ, &B);
979: MatDestroy(&AT);
980: } else {
981: MatComputeOperator(A, MATAIJ, &B);
982: }
983: ObjectView((PetscObject) B, ksp->viewerMatExp, ksp->formatMatExp);
984: MatDestroy(&B);
985: }
986: if (ksp->viewPOpExp) {
987: Mat B;
989: KSPComputeOperator(ksp, MATAIJ, &B);
990: ObjectView((PetscObject) B, ksp->viewerPOpExp, ksp->formatPOpExp);
991: MatDestroy(&B);
992: }
994: if (inXisinB) {
995: VecCopy(x,b);
996: VecDestroy(&x);
997: }
998: PetscObjectSAWsBlock((PetscObject)ksp);
999: if (ksp->errorifnotconverged && ksp->reason < 0 && ksp->reason != KSP_DIVERGED_ITS) {
1000: if (ksp->reason == KSP_DIVERGED_PC_FAILED) {
1001: PCFailedReason reason;
1002: PCGetFailedReason(ksp->pc,&reason);
1003: SETERRQ2(comm,PETSC_ERR_NOT_CONVERGED,"KSPSolve has not converged, reason %s PC failed due to %s",KSPConvergedReasons[ksp->reason],PCFailedReasons[reason]);
1004: } else SETERRQ1(comm,PETSC_ERR_NOT_CONVERGED,"KSPSolve has not converged, reason %s",KSPConvergedReasons[ksp->reason]);
1005: }
1006: return(0);
1007: }
1009: /*@
1010: KSPSolve - Solves linear system.
1012: Collective on ksp
1014: Parameters:
1015: + ksp - iterative context obtained from KSPCreate()
1016: . b - the right hand side vector
1017: - x - the solution (this may be the same vector as b, then b will be overwritten with answer)
1019: Options Database Keys:
1020: + -ksp_view_eigenvalues - compute preconditioned operators eigenvalues
1021: . -ksp_view_eigenvalues_explicit - compute the eigenvalues by forming the dense operator and using LAPACK
1022: . -ksp_view_mat binary - save matrix to the default binary viewer
1023: . -ksp_view_pmat binary - save matrix used to build preconditioner to the default binary viewer
1024: . -ksp_view_rhs binary - save right hand side vector to the default binary viewer
1025: . -ksp_view_solution binary - save computed solution vector to the default binary viewer
1026: (can be read later with src/ksp/tutorials/ex10.c for testing solvers)
1027: . -ksp_view_mat_explicit - for matrix-free operators, computes the matrix entries and views them
1028: . -ksp_view_preconditioned_operator_explicit - computes the product of the preconditioner and matrix as an explicit matrix and views it
1029: . -ksp_converged_reason - print reason for converged or diverged, also prints number of iterations
1030: . -ksp_view_final_residual - print 2-norm of true linear system residual at the end of the solution process
1031: - -ksp_view - print the ksp data structure at the end of the system solution
1033: Notes:
1035: If one uses KSPSetDM() then x or b need not be passed. Use KSPGetSolution() to access the solution in this case.
1037: The operator is specified with KSPSetOperators().
1039: Call KSPGetConvergedReason() to determine if the solver converged or failed and
1040: why. The number of iterations can be obtained from KSPGetIterationNumber().
1042: If you provide a matrix that has a MatSetNullSpace() and MatSetTransposeNullSpace() this will use that information to solve singular systems
1043: in the least squares sense with a norm minimizing solution.
1044: $
1045: $ A x = b where b = b_p + b_t where b_t is not in the range of A (and hence by the fundamental theorem of linear algebra is in the nullspace(A') see MatSetNullSpace()
1046: $
1047: $ KSP first removes b_t producing the linear system A x = b_p (which has multiple solutions) and solves this to find the ||x|| minimizing solution (and hence
1048: $ it finds the solution x orthogonal to the nullspace(A). The algorithm is simply in each iteration of the Krylov method we remove the nullspace(A) from the search
1049: $ direction thus the solution which is a linear combination of the search directions has no component in the nullspace(A).
1050: $
1051: $ We recommend always using GMRES for such singular systems.
1052: $ If nullspace(A) = nullspace(A') (note symmetric matrices always satisfy this property) then both left and right preconditioning will work
1053: $ If nullspace(A) != nullspace(A') then left preconditioning will work but right preconditioning may not work (or it may).
1055: Developer Note: The reason we cannot always solve nullspace(A) != nullspace(A') systems with right preconditioning is because we need to remove at each iteration
1056: the nullspace(AB) from the search direction. While we know the nullspace(A) the nullspace(AB) equals B^-1 times the nullspace(A) but except for trivial preconditioners
1057: such as diagonal scaling we cannot apply the inverse of the preconditioner to a vector and thus cannot compute the nullspace(AB).
1060: If using a direct method (e.g., via the KSP solver
1061: KSPPREONLY and a preconditioner such as PCLU/PCILU),
1062: then its=1. See KSPSetTolerances() and KSPConvergedDefault()
1063: for more details.
1065: Understanding Convergence:
1066: The routines KSPMonitorSet(), KSPComputeEigenvalues(), and
1067: KSPComputeEigenvaluesExplicitly() provide information on additional
1068: options to monitor convergence and print eigenvalue information.
1070: Level: beginner
1072: .seealso: KSPCreate(), KSPSetUp(), KSPDestroy(), KSPSetTolerances(), KSPConvergedDefault(),
1073: KSPSolveTranspose(), KSPGetIterationNumber(), MatNullSpaceCreate(), MatSetNullSpace(), MatSetTransposeNullSpace(), KSP,
1074: KSPConvergedReasonView()
1075: @*/
1076: PetscErrorCode KSPSolve(KSP ksp,Vec b,Vec x)
1077: {
1084: ksp->transpose_solve = PETSC_FALSE;
1085: KSPSolve_Private(ksp,b,x);
1086: return(0);
1087: }
1089: /*@
1090: KSPSolveTranspose - Solves the transpose of a linear system.
1092: Collective on ksp
1094: Input Parameters:
1095: + ksp - iterative context obtained from KSPCreate()
1096: . b - right hand side vector
1097: - x - solution vector
1099: Notes:
1100: For complex numbers this solve the non-Hermitian transpose system.
1102: Developer Notes:
1103: We need to implement a KSPSolveHermitianTranspose()
1105: Level: developer
1107: .seealso: KSPCreate(), KSPSetUp(), KSPDestroy(), KSPSetTolerances(), KSPConvergedDefault(),
1108: KSPSolve(), KSP
1109: @*/
1110: PetscErrorCode KSPSolveTranspose(KSP ksp,Vec b,Vec x)
1111: {
1118: if (ksp->transpose.use_explicittranspose) {
1119: Mat J,Jpre;
1120: KSPGetOperators(ksp,&J,&Jpre);
1121: if (!ksp->transpose.reuse_transpose) {
1122: MatTranspose(J,MAT_INITIAL_MATRIX,&ksp->transpose.AT);
1123: if (J != Jpre) {
1124: MatTranspose(Jpre,MAT_INITIAL_MATRIX,&ksp->transpose.BT);
1125: }
1126: ksp->transpose.reuse_transpose = PETSC_TRUE;
1127: } else {
1128: MatTranspose(J,MAT_REUSE_MATRIX,&ksp->transpose.AT);
1129: if (J != Jpre) {
1130: MatTranspose(Jpre,MAT_REUSE_MATRIX,&ksp->transpose.BT);
1131: }
1132: }
1133: if (J == Jpre && ksp->transpose.BT != ksp->transpose.AT) {
1134: PetscObjectReference((PetscObject)ksp->transpose.AT);
1135: ksp->transpose.BT = ksp->transpose.AT;
1136: }
1137: KSPSetOperators(ksp,ksp->transpose.AT,ksp->transpose.BT);
1138: } else {
1139: ksp->transpose_solve = PETSC_TRUE;
1140: }
1141: KSPSolve_Private(ksp,b,x);
1142: return(0);
1143: }
1145: static PetscErrorCode KSPViewFinalMatResidual_Internal(KSP ksp, Mat B, Mat X, PetscViewer viewer, PetscViewerFormat format, PetscInt shift)
1146: {
1147: Mat A, R;
1148: PetscReal *norms;
1149: PetscInt i, N;
1150: PetscBool flg;
1154: PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &flg);
1155: if (flg) {
1156: PCGetOperators(ksp->pc, &A, NULL);
1157: MatMatMult(A, X, MAT_INITIAL_MATRIX, PETSC_DEFAULT, &R);
1158: MatAYPX(R, -1.0, B, SAME_NONZERO_PATTERN);
1159: MatGetSize(R, NULL, &N);
1160: PetscMalloc1(N, &norms);
1161: MatGetColumnNorms(R, NORM_2, norms);
1162: MatDestroy(&R);
1163: for (i = 0; i < N; ++i) {
1164: PetscViewerASCIIPrintf(viewer, "%s #%D %g\n", i == 0 ? "KSP final norm of residual" : " ", shift + i, (double)norms[i]);
1165: }
1166: PetscFree(norms);
1167: }
1168: return(0);
1169: }
1171: /*@
1172: KSPMatSolve - Solves a linear system with multiple right-hand sides stored as a MATDENSE. Unlike KSPSolve(), B and X must be different matrices.
1174: Input Parameters:
1175: + ksp - iterative context
1176: - B - block of right-hand sides
1178: Output Parameter:
1179: . X - block of solutions
1181: Notes:
1182: This is a stripped-down version of KSPSolve(), which only handles -ksp_view, -ksp_converged_reason, and -ksp_view_final_residual.
1184: Level: intermediate
1186: .seealso: KSPSolve(), MatMatSolve(), MATDENSE, KSPHPDDM, PCBJACOBI, PCASM
1187: @*/
1188: PetscErrorCode KSPMatSolve(KSP ksp, Mat B, Mat X)
1189: {
1190: Mat A, vB, vX;
1191: Vec cb, cx;
1192: PetscInt n1, N1, n2, N2, Bbn = PETSC_DECIDE;
1193: PetscBool match;
1202: if (!B->assembled) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
1203: MatCheckPreallocated(X, 3);
1204: if (!X->assembled) {
1205: MatSetOption(X, MAT_NO_OFF_PROC_ENTRIES, PETSC_TRUE);
1206: MatAssemblyBegin(X, MAT_FINAL_ASSEMBLY);
1207: MatAssemblyEnd(X, MAT_FINAL_ASSEMBLY);
1208: }
1209: if (B == X) SETERRQ(PetscObjectComm((PetscObject)ksp), PETSC_ERR_ARG_IDN, "B and X must be different matrices");
1210: KSPGetOperators(ksp, &A, NULL);
1211: MatGetLocalSize(B, NULL, &n2);
1212: MatGetLocalSize(X, NULL, &n1);
1213: MatGetSize(B, NULL, &N2);
1214: MatGetSize(X, NULL, &N1);
1215: if (n1 != n2 || N1 != N2) SETERRQ4(PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Incompatible number of columns between block of right-hand sides (n,N) = (%D,%D) and block of solutions (n,N) = (%D,%D)", n2, N2, n1, N1);
1216: PetscObjectBaseTypeCompareAny((PetscObject)B, &match, MATSEQDENSE, MATMPIDENSE, "");
1217: if (!match) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Provided block of right-hand sides not stored in a dense Mat");
1218: PetscObjectBaseTypeCompareAny((PetscObject)X, &match, MATSEQDENSE, MATMPIDENSE, "");
1219: if (!match) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Provided block of solutions not stored in a dense Mat");
1220: KSPSetUp(ksp);
1221: KSPSetUpOnBlocks(ksp);
1222: if (ksp->ops->matsolve) {
1223: if (ksp->guess_zero) {
1224: MatZeroEntries(X);
1225: }
1226: PetscLogEventBegin(KSP_MatSolve, ksp, B, X, 0);
1227: KSPGetMatSolveBatchSize(ksp, &Bbn);
1228: /* by default, do a single solve with all columns */
1229: if (Bbn == PETSC_DECIDE) Bbn = N2;
1230: else if (Bbn < 1) SETERRQ1(PetscObjectComm((PetscObject)ksp), PETSC_ERR_ARG_OUTOFRANGE, "KSPMatSolve() block size %D must be positive", Bbn);
1231: PetscInfo2(ksp, "KSP type %s solving using blocks of width at most %D\n", ((PetscObject)ksp)->type_name, Bbn);
1232: /* if -ksp_matsolve_batch_size is greater than the actual number of columns, do a single solve with all columns */
1233: if (Bbn >= N2) {
1234: (*ksp->ops->matsolve)(ksp, B, X);
1235: if (ksp->viewFinalRes) {
1236: KSPViewFinalMatResidual_Internal(ksp, B, X, ksp->viewerFinalRes, ksp->formatFinalRes, 0);
1237: }
1239: KSPConvergedReasonViewFromOptions(ksp);
1241: if (ksp->viewRate) {
1242: PetscViewerPushFormat(ksp->viewerRate,PETSC_VIEWER_DEFAULT);
1243: KSPConvergedRateView(ksp, ksp->viewerRate);
1244: PetscViewerPopFormat(ksp->viewerRate);
1245: }
1246: } else {
1247: for (n2 = 0; n2 < N2; n2 += Bbn) {
1248: MatDenseGetSubMatrix(B, n2, PetscMin(n2+Bbn, N2), &vB);
1249: MatDenseGetSubMatrix(X, n2, PetscMin(n2+Bbn, N2), &vX);
1250: (*ksp->ops->matsolve)(ksp, vB, vX);
1251: if (ksp->viewFinalRes) {
1252: KSPViewFinalMatResidual_Internal(ksp, vB, vX, ksp->viewerFinalRes, ksp->formatFinalRes, n2);
1253: }
1255: KSPConvergedReasonViewFromOptions(ksp);
1257: if (ksp->viewRate) {
1258: PetscViewerPushFormat(ksp->viewerRate,PETSC_VIEWER_DEFAULT);
1259: KSPConvergedRateView(ksp, ksp->viewerRate);
1260: PetscViewerPopFormat(ksp->viewerRate);
1261: }
1262: MatDenseRestoreSubMatrix(B, &vB);
1263: MatDenseRestoreSubMatrix(X, &vX);
1264: }
1265: }
1266: if (ksp->view) {
1267: KSPView(ksp, ksp->viewer);
1268: }
1269: PetscLogEventEnd(KSP_MatSolve, ksp, B, X, 0);
1270: } else {
1271: PetscInfo1(ksp, "KSP type %s solving column by column\n", ((PetscObject)ksp)->type_name);
1272: for (n2 = 0; n2 < N2; ++n2) {
1273: MatDenseGetColumnVecRead(B, n2, &cb);
1274: MatDenseGetColumnVecWrite(X, n2, &cx);
1275: KSPSolve(ksp, cb, cx);
1276: MatDenseRestoreColumnVecWrite(X, n2, &cx);
1277: MatDenseRestoreColumnVecRead(B, n2, &cb);
1278: }
1279: }
1280: return(0);
1281: }
1283: /*@
1284: KSPSetMatSolveBatchSize - Sets the maximum number of columns treated simultaneously in KSPMatSolve().
1286: Logically collective
1288: Input Parameters:
1289: + ksp - iterative context
1290: - bs - block size
1292: Level: advanced
1294: .seealso: KSPMatSolve(), KSPGetMatSolveBatchSize(), -mat_mumps_icntl_27, -matmatmult_Bbn
1295: @*/
1296: PetscErrorCode KSPSetMatSolveBatchSize(KSP ksp, PetscInt bs)
1297: {
1301: ksp->nmax = bs;
1302: return(0);
1303: }
1305: /*@
1306: KSPGetMatSolveBatchSize - Gets the maximum number of columns treated simultaneously in KSPMatSolve().
1308: Input Parameter:
1309: . ksp - iterative context
1311: Output Parameter:
1312: . bs - block size
1314: Level: advanced
1316: .seealso: KSPMatSolve(), KSPSetMatSolveBatchSize(), -mat_mumps_icntl_27, -matmatmult_Bbn
1317: @*/
1318: PetscErrorCode KSPGetMatSolveBatchSize(KSP ksp, PetscInt *bs)
1319: {
1323: *bs = ksp->nmax;
1324: return(0);
1325: }
1327: /*@
1328: KSPResetViewers - Resets all the viewers set from the options database during KSPSetFromOptions()
1330: Collective on ksp
1332: Input Parameter:
1333: . ksp - iterative context obtained from KSPCreate()
1335: Level: beginner
1337: .seealso: KSPCreate(), KSPSetUp(), KSPSolve(), KSPSetFromOptions(), KSP
1338: @*/
1339: PetscErrorCode KSPResetViewers(KSP ksp)
1340: {
1345: if (!ksp) return(0);
1346: PetscViewerDestroy(&ksp->viewer);
1347: PetscViewerDestroy(&ksp->viewerPre);
1348: PetscViewerDestroy(&ksp->viewerRate);
1349: PetscViewerDestroy(&ksp->viewerMat);
1350: PetscViewerDestroy(&ksp->viewerPMat);
1351: PetscViewerDestroy(&ksp->viewerRhs);
1352: PetscViewerDestroy(&ksp->viewerSol);
1353: PetscViewerDestroy(&ksp->viewerMatExp);
1354: PetscViewerDestroy(&ksp->viewerEV);
1355: PetscViewerDestroy(&ksp->viewerSV);
1356: PetscViewerDestroy(&ksp->viewerEVExp);
1357: PetscViewerDestroy(&ksp->viewerFinalRes);
1358: PetscViewerDestroy(&ksp->viewerPOpExp);
1359: PetscViewerDestroy(&ksp->viewerDScale);
1360: ksp->view = PETSC_FALSE;
1361: ksp->viewPre = PETSC_FALSE;
1362: ksp->viewMat = PETSC_FALSE;
1363: ksp->viewPMat = PETSC_FALSE;
1364: ksp->viewRhs = PETSC_FALSE;
1365: ksp->viewSol = PETSC_FALSE;
1366: ksp->viewMatExp = PETSC_FALSE;
1367: ksp->viewEV = PETSC_FALSE;
1368: ksp->viewSV = PETSC_FALSE;
1369: ksp->viewEVExp = PETSC_FALSE;
1370: ksp->viewFinalRes = PETSC_FALSE;
1371: ksp->viewPOpExp = PETSC_FALSE;
1372: ksp->viewDScale = PETSC_FALSE;
1373: return(0);
1374: }
1376: /*@
1377: KSPReset - Resets a KSP context to the kspsetupcalled = 0 state and removes any allocated Vecs and Mats
1379: Collective on ksp
1381: Input Parameter:
1382: . ksp - iterative context obtained from KSPCreate()
1384: Level: beginner
1386: .seealso: KSPCreate(), KSPSetUp(), KSPSolve(), KSP
1387: @*/
1388: PetscErrorCode KSPReset(KSP ksp)
1389: {
1394: if (!ksp) return(0);
1395: if (ksp->ops->reset) {
1396: (*ksp->ops->reset)(ksp);
1397: }
1398: if (ksp->pc) {PCReset(ksp->pc);}
1399: if (ksp->guess) {
1400: KSPGuess guess = ksp->guess;
1401: if (guess->ops->reset) { (*guess->ops->reset)(guess); }
1402: }
1403: VecDestroyVecs(ksp->nwork,&ksp->work);
1404: VecDestroy(&ksp->vec_rhs);
1405: VecDestroy(&ksp->vec_sol);
1406: VecDestroy(&ksp->diagonal);
1407: VecDestroy(&ksp->truediagonal);
1409: KSPResetViewers(ksp);
1411: ksp->setupstage = KSP_SETUP_NEW;
1412: ksp->nmax = PETSC_DECIDE;
1413: return(0);
1414: }
1416: /*@C
1417: KSPDestroy - Destroys KSP context.
1419: Collective on ksp
1421: Input Parameter:
1422: . ksp - iterative context obtained from KSPCreate()
1424: Level: beginner
1426: .seealso: KSPCreate(), KSPSetUp(), KSPSolve(), KSP
1427: @*/
1428: PetscErrorCode KSPDestroy(KSP *ksp)
1429: {
1431: PC pc;
1434: if (!*ksp) return(0);
1436: if (--((PetscObject)(*ksp))->refct > 0) {*ksp = NULL; return(0);}
1438: PetscObjectSAWsViewOff((PetscObject)*ksp);
1440: /*
1441: Avoid a cascading call to PCReset(ksp->pc) from the following call:
1442: PCReset() shouldn't be called from KSPDestroy() as it is unprotected by pc's
1443: refcount (and may be shared, e.g., by other ksps).
1444: */
1445: pc = (*ksp)->pc;
1446: (*ksp)->pc = NULL;
1447: KSPReset((*ksp));
1448: (*ksp)->pc = pc;
1449: if ((*ksp)->ops->destroy) {(*(*ksp)->ops->destroy)(*ksp);}
1451: if ((*ksp)->transpose.use_explicittranspose) {
1452: MatDestroy(&(*ksp)->transpose.AT);
1453: MatDestroy(&(*ksp)->transpose.BT);
1454: (*ksp)->transpose.reuse_transpose = PETSC_FALSE;
1455: }
1457: KSPGuessDestroy(&(*ksp)->guess);
1458: DMDestroy(&(*ksp)->dm);
1459: PCDestroy(&(*ksp)->pc);
1460: PetscFree((*ksp)->res_hist_alloc);
1461: PetscFree((*ksp)->err_hist_alloc);
1462: if ((*ksp)->convergeddestroy) {
1463: (*(*ksp)->convergeddestroy)((*ksp)->cnvP);
1464: }
1465: KSPMonitorCancel((*ksp));
1466: KSPConvergedReasonViewCancel((*ksp));
1467: PetscViewerDestroy(&(*ksp)->eigviewer);
1468: PetscHeaderDestroy(ksp);
1469: return(0);
1470: }
1472: /*@
1473: KSPSetPCSide - Sets the preconditioning side.
1475: Logically Collective on ksp
1477: Input Parameter:
1478: . ksp - iterative context obtained from KSPCreate()
1480: Output Parameter:
1481: . side - the preconditioning side, where side is one of
1482: .vb
1483: PC_LEFT - left preconditioning (default)
1484: PC_RIGHT - right preconditioning
1485: PC_SYMMETRIC - symmetric preconditioning
1486: .ve
1488: Options Database Keys:
1489: . -ksp_pc_side <right,left,symmetric>
1491: Notes:
1492: Left preconditioning is used by default for most Krylov methods except KSPFGMRES which only supports right preconditioning.
1494: For methods changing the side of the preconditioner changes the norm type that is used, see KSPSetNormType().
1496: Symmetric preconditioning is currently available only for the KSPQCG method. Note, however, that
1497: symmetric preconditioning can be emulated by using either right or left
1498: preconditioning and a pre or post processing step.
1500: Setting the PC side often affects the default norm type. See KSPSetNormType() for details.
1502: Level: intermediate
1504: .seealso: KSPGetPCSide(), KSPSetNormType(), KSPGetNormType(), KSP
1505: @*/
1506: PetscErrorCode KSPSetPCSide(KSP ksp,PCSide side)
1507: {
1511: ksp->pc_side = ksp->pc_side_set = side;
1512: return(0);
1513: }
1515: /*@
1516: KSPGetPCSide - Gets the preconditioning side.
1518: Not Collective
1520: Input Parameter:
1521: . ksp - iterative context obtained from KSPCreate()
1523: Output Parameter:
1524: . side - the preconditioning side, where side is one of
1525: .vb
1526: PC_LEFT - left preconditioning (default)
1527: PC_RIGHT - right preconditioning
1528: PC_SYMMETRIC - symmetric preconditioning
1529: .ve
1531: Level: intermediate
1533: .seealso: KSPSetPCSide(), KSP
1534: @*/
1535: PetscErrorCode KSPGetPCSide(KSP ksp,PCSide *side)
1536: {
1542: KSPSetUpNorms_Private(ksp,PETSC_TRUE,&ksp->normtype,&ksp->pc_side);
1543: *side = ksp->pc_side;
1544: return(0);
1545: }
1547: /*@
1548: KSPGetTolerances - Gets the relative, absolute, divergence, and maximum
1549: iteration tolerances used by the default KSP convergence tests.
1551: Not Collective
1553: Input Parameter:
1554: . ksp - the Krylov subspace context
1556: Output Parameters:
1557: + rtol - the relative convergence tolerance
1558: . abstol - the absolute convergence tolerance
1559: . dtol - the divergence tolerance
1560: - maxits - maximum number of iterations
1562: Notes:
1563: The user can specify NULL for any parameter that is not needed.
1565: Level: intermediate
1567: maximum, iterations
1569: .seealso: KSPSetTolerances(), KSP
1570: @*/
1571: PetscErrorCode KSPGetTolerances(KSP ksp,PetscReal *rtol,PetscReal *abstol,PetscReal *dtol,PetscInt *maxits)
1572: {
1575: if (abstol) *abstol = ksp->abstol;
1576: if (rtol) *rtol = ksp->rtol;
1577: if (dtol) *dtol = ksp->divtol;
1578: if (maxits) *maxits = ksp->max_it;
1579: return(0);
1580: }
1582: /*@
1583: KSPSetTolerances - Sets the relative, absolute, divergence, and maximum
1584: iteration tolerances used by the default KSP convergence testers.
1586: Logically Collective on ksp
1588: Input Parameters:
1589: + ksp - the Krylov subspace context
1590: . rtol - the relative convergence tolerance, relative decrease in the (possibly preconditioned) residual norm
1591: . abstol - the absolute convergence tolerance absolute size of the (possibly preconditioned) residual norm
1592: . dtol - the divergence tolerance, amount (possibly preconditioned) residual norm can increase before KSPConvergedDefault() concludes that the method is diverging
1593: - maxits - maximum number of iterations to use
1595: Options Database Keys:
1596: + -ksp_atol <abstol> - Sets abstol
1597: . -ksp_rtol <rtol> - Sets rtol
1598: . -ksp_divtol <dtol> - Sets dtol
1599: - -ksp_max_it <maxits> - Sets maxits
1601: Notes:
1602: Use PETSC_DEFAULT to retain the default value of any of the tolerances.
1604: See KSPConvergedDefault() for details how these parameters are used in the default convergence test. See also KSPSetConvergenceTest()
1605: for setting user-defined stopping criteria.
1607: Level: intermediate
1609: convergence, maximum, iterations
1611: .seealso: KSPGetTolerances(), KSPConvergedDefault(), KSPSetConvergenceTest(), KSP
1612: @*/
1613: PetscErrorCode KSPSetTolerances(KSP ksp,PetscReal rtol,PetscReal abstol,PetscReal dtol,PetscInt maxits)
1614: {
1622: if (rtol != PETSC_DEFAULT) {
1623: if (rtol < 0.0 || 1.0 <= rtol) SETERRQ1(PetscObjectComm((PetscObject)ksp),PETSC_ERR_ARG_OUTOFRANGE,"Relative tolerance %g must be non-negative and less than 1.0",(double)rtol);
1624: ksp->rtol = rtol;
1625: }
1626: if (abstol != PETSC_DEFAULT) {
1627: if (abstol < 0.0) SETERRQ1(PetscObjectComm((PetscObject)ksp),PETSC_ERR_ARG_OUTOFRANGE,"Absolute tolerance %g must be non-negative",(double)abstol);
1628: ksp->abstol = abstol;
1629: }
1630: if (dtol != PETSC_DEFAULT) {
1631: if (dtol < 0.0) SETERRQ1(PetscObjectComm((PetscObject)ksp),PETSC_ERR_ARG_OUTOFRANGE,"Divergence tolerance %g must be larger than 1.0",(double)dtol);
1632: ksp->divtol = dtol;
1633: }
1634: if (maxits != PETSC_DEFAULT) {
1635: if (maxits < 0) SETERRQ1(PetscObjectComm((PetscObject)ksp),PETSC_ERR_ARG_OUTOFRANGE,"Maximum number of iterations %D must be non-negative",maxits);
1636: ksp->max_it = maxits;
1637: }
1638: return(0);
1639: }
1641: /*@
1642: KSPSetInitialGuessNonzero - Tells the iterative solver that the
1643: initial guess is nonzero; otherwise KSP assumes the initial guess
1644: is to be zero (and thus zeros it out before solving).
1646: Logically Collective on ksp
1648: Input Parameters:
1649: + ksp - iterative context obtained from KSPCreate()
1650: - flg - PETSC_TRUE indicates the guess is non-zero, PETSC_FALSE indicates the guess is zero
1652: Options database keys:
1653: . -ksp_initial_guess_nonzero : use nonzero initial guess; this takes an optional truth value (0/1/no/yes/true/false)
1655: Level: beginner
1657: Notes:
1658: If this is not called the X vector is zeroed in the call to KSPSolve().
1660: .seealso: KSPGetInitialGuessNonzero(), KSPSetGuessType(), KSPGuessType, KSP
1661: @*/
1662: PetscErrorCode KSPSetInitialGuessNonzero(KSP ksp,PetscBool flg)
1663: {
1667: ksp->guess_zero = (PetscBool) !(int)flg;
1668: return(0);
1669: }
1671: /*@
1672: KSPGetInitialGuessNonzero - Determines whether the KSP solver is using
1673: a zero initial guess.
1675: Not Collective
1677: Input Parameter:
1678: . ksp - iterative context obtained from KSPCreate()
1680: Output Parameter:
1681: . flag - PETSC_TRUE if guess is nonzero, else PETSC_FALSE
1683: Level: intermediate
1685: .seealso: KSPSetInitialGuessNonzero(), KSP
1686: @*/
1687: PetscErrorCode KSPGetInitialGuessNonzero(KSP ksp,PetscBool *flag)
1688: {
1692: if (ksp->guess_zero) *flag = PETSC_FALSE;
1693: else *flag = PETSC_TRUE;
1694: return(0);
1695: }
1697: /*@
1698: KSPSetErrorIfNotConverged - Causes KSPSolve() to generate an error if the solver has not converged.
1700: Logically Collective on ksp
1702: Input Parameters:
1703: + ksp - iterative context obtained from KSPCreate()
1704: - flg - PETSC_TRUE indicates you want the error generated
1706: Options database keys:
1707: . -ksp_error_if_not_converged : this takes an optional truth value (0/1/no/yes/true/false)
1709: Level: intermediate
1711: Notes:
1712: Normally PETSc continues if a linear solver fails to converge, you can call KSPGetConvergedReason() after a KSPSolve()
1713: to determine if it has converged.
1716: .seealso: KSPGetErrorIfNotConverged(), KSP
1717: @*/
1718: PetscErrorCode KSPSetErrorIfNotConverged(KSP ksp,PetscBool flg)
1719: {
1723: ksp->errorifnotconverged = flg;
1724: return(0);
1725: }
1727: /*@
1728: KSPGetErrorIfNotConverged - Will KSPSolve() generate an error if the solver does not converge?
1730: Not Collective
1732: Input Parameter:
1733: . ksp - iterative context obtained from KSPCreate()
1735: Output Parameter:
1736: . flag - PETSC_TRUE if it will generate an error, else PETSC_FALSE
1738: Level: intermediate
1740: .seealso: KSPSetErrorIfNotConverged(), KSP
1741: @*/
1742: PetscErrorCode KSPGetErrorIfNotConverged(KSP ksp,PetscBool *flag)
1743: {
1747: *flag = ksp->errorifnotconverged;
1748: return(0);
1749: }
1751: /*@
1752: KSPSetInitialGuessKnoll - Tells the iterative solver to use PCApply(pc,b,..) to compute the initial guess (The Knoll trick)
1754: Logically Collective on ksp
1756: Input Parameters:
1757: + ksp - iterative context obtained from KSPCreate()
1758: - flg - PETSC_TRUE or PETSC_FALSE
1760: Level: advanced
1762: Developer Note: the Knoll trick is not currently implemented using the KSPGuess class
1764: .seealso: KSPGetInitialGuessKnoll(), KSPSetInitialGuessNonzero(), KSPGetInitialGuessNonzero(), KSP
1765: @*/
1766: PetscErrorCode KSPSetInitialGuessKnoll(KSP ksp,PetscBool flg)
1767: {
1771: ksp->guess_knoll = flg;
1772: return(0);
1773: }
1775: /*@
1776: KSPGetInitialGuessKnoll - Determines whether the KSP solver is using the Knoll trick (using PCApply(pc,b,...) to compute
1777: the initial guess
1779: Not Collective
1781: Input Parameter:
1782: . ksp - iterative context obtained from KSPCreate()
1784: Output Parameter:
1785: . flag - PETSC_TRUE if using Knoll trick, else PETSC_FALSE
1787: Level: advanced
1789: .seealso: KSPSetInitialGuessKnoll(), KSPSetInitialGuessNonzero(), KSPGetInitialGuessNonzero(), KSP
1790: @*/
1791: PetscErrorCode KSPGetInitialGuessKnoll(KSP ksp,PetscBool *flag)
1792: {
1796: *flag = ksp->guess_knoll;
1797: return(0);
1798: }
1800: /*@
1801: KSPGetComputeSingularValues - Gets the flag indicating whether the extreme singular
1802: values will be calculated via a Lanczos or Arnoldi process as the linear
1803: system is solved.
1805: Not Collective
1807: Input Parameter:
1808: . ksp - iterative context obtained from KSPCreate()
1810: Output Parameter:
1811: . flg - PETSC_TRUE or PETSC_FALSE
1813: Options Database Key:
1814: . -ksp_monitor_singular_value - Activates KSPSetComputeSingularValues()
1816: Notes:
1817: Currently this option is not valid for all iterative methods.
1819: Many users may just want to use the monitoring routine
1820: KSPMonitorSingularValue() (which can be set with option -ksp_monitor_singular_value)
1821: to print the singular values at each iteration of the linear solve.
1823: Level: advanced
1825: .seealso: KSPComputeExtremeSingularValues(), KSPMonitorSingularValue(), KSP
1826: @*/
1827: PetscErrorCode KSPGetComputeSingularValues(KSP ksp,PetscBool *flg)
1828: {
1832: *flg = ksp->calc_sings;
1833: return(0);
1834: }
1836: /*@
1837: KSPSetComputeSingularValues - Sets a flag so that the extreme singular
1838: values will be calculated via a Lanczos or Arnoldi process as the linear
1839: system is solved.
1841: Logically Collective on ksp
1843: Input Parameters:
1844: + ksp - iterative context obtained from KSPCreate()
1845: - flg - PETSC_TRUE or PETSC_FALSE
1847: Options Database Key:
1848: . -ksp_monitor_singular_value - Activates KSPSetComputeSingularValues()
1850: Notes:
1851: Currently this option is not valid for all iterative methods.
1853: Many users may just want to use the monitoring routine
1854: KSPMonitorSingularValue() (which can be set with option -ksp_monitor_singular_value)
1855: to print the singular values at each iteration of the linear solve.
1857: Level: advanced
1859: .seealso: KSPComputeExtremeSingularValues(), KSPMonitorSingularValue(), KSP
1860: @*/
1861: PetscErrorCode KSPSetComputeSingularValues(KSP ksp,PetscBool flg)
1862: {
1866: ksp->calc_sings = flg;
1867: return(0);
1868: }
1870: /*@
1871: KSPGetComputeEigenvalues - Gets the flag indicating that the extreme eigenvalues
1872: values will be calculated via a Lanczos or Arnoldi process as the linear
1873: system is solved.
1875: Not Collective
1877: Input Parameter:
1878: . ksp - iterative context obtained from KSPCreate()
1880: Output Parameter:
1881: . flg - PETSC_TRUE or PETSC_FALSE
1883: Notes:
1884: Currently this option is not valid for all iterative methods.
1886: Level: advanced
1888: .seealso: KSPComputeEigenvalues(), KSPComputeEigenvaluesExplicitly(), KSP
1889: @*/
1890: PetscErrorCode KSPGetComputeEigenvalues(KSP ksp,PetscBool *flg)
1891: {
1895: *flg = ksp->calc_sings;
1896: return(0);
1897: }
1899: /*@
1900: KSPSetComputeEigenvalues - Sets a flag so that the extreme eigenvalues
1901: values will be calculated via a Lanczos or Arnoldi process as the linear
1902: system is solved.
1904: Logically Collective on ksp
1906: Input Parameters:
1907: + ksp - iterative context obtained from KSPCreate()
1908: - flg - PETSC_TRUE or PETSC_FALSE
1910: Notes:
1911: Currently this option is not valid for all iterative methods.
1913: Level: advanced
1915: .seealso: KSPComputeEigenvalues(), KSPComputeEigenvaluesExplicitly(), KSP
1916: @*/
1917: PetscErrorCode KSPSetComputeEigenvalues(KSP ksp,PetscBool flg)
1918: {
1922: ksp->calc_sings = flg;
1923: return(0);
1924: }
1926: /*@
1927: KSPSetComputeRitz - Sets a flag so that the Ritz or harmonic Ritz pairs
1928: will be calculated via a Lanczos or Arnoldi process as the linear
1929: system is solved.
1931: Logically Collective on ksp
1933: Input Parameters:
1934: + ksp - iterative context obtained from KSPCreate()
1935: - flg - PETSC_TRUE or PETSC_FALSE
1937: Notes:
1938: Currently this option is only valid for the GMRES method.
1940: Level: advanced
1942: .seealso: KSPComputeRitz(), KSP
1943: @*/
1944: PetscErrorCode KSPSetComputeRitz(KSP ksp, PetscBool flg)
1945: {
1949: ksp->calc_ritz = flg;
1950: return(0);
1951: }
1953: /*@
1954: KSPGetRhs - Gets the right-hand-side vector for the linear system to
1955: be solved.
1957: Not Collective
1959: Input Parameter:
1960: . ksp - iterative context obtained from KSPCreate()
1962: Output Parameter:
1963: . r - right-hand-side vector
1965: Level: developer
1967: .seealso: KSPGetSolution(), KSPSolve(), KSP
1968: @*/
1969: PetscErrorCode KSPGetRhs(KSP ksp,Vec *r)
1970: {
1974: *r = ksp->vec_rhs;
1975: return(0);
1976: }
1978: /*@
1979: KSPGetSolution - Gets the location of the solution for the
1980: linear system to be solved. Note that this may not be where the solution
1981: is stored during the iterative process; see KSPBuildSolution().
1983: Not Collective
1985: Input Parameters:
1986: . ksp - iterative context obtained from KSPCreate()
1988: Output Parameters:
1989: . v - solution vector
1991: Level: developer
1993: .seealso: KSPGetRhs(), KSPBuildSolution(), KSPSolve(), KSP
1994: @*/
1995: PetscErrorCode KSPGetSolution(KSP ksp,Vec *v)
1996: {
2000: *v = ksp->vec_sol;
2001: return(0);
2002: }
2004: /*@
2005: KSPSetPC - Sets the preconditioner to be used to calculate the
2006: application of the preconditioner on a vector.
2008: Collective on ksp
2010: Input Parameters:
2011: + ksp - iterative context obtained from KSPCreate()
2012: - pc - the preconditioner object (can be NULL)
2014: Notes:
2015: Use KSPGetPC() to retrieve the preconditioner context.
2017: Level: developer
2019: .seealso: KSPGetPC(), KSP
2020: @*/
2021: PetscErrorCode KSPSetPC(KSP ksp,PC pc)
2022: {
2027: if (pc) {
2030: }
2031: PetscObjectReference((PetscObject)pc);
2032: PCDestroy(&ksp->pc);
2033: ksp->pc = pc;
2034: PetscLogObjectParent((PetscObject)ksp,(PetscObject)ksp->pc);
2035: return(0);
2036: }
2038: /*@
2039: KSPGetPC - Returns a pointer to the preconditioner context
2040: set with KSPSetPC().
2042: Not Collective
2044: Input Parameters:
2045: . ksp - iterative context obtained from KSPCreate()
2047: Output Parameter:
2048: . pc - preconditioner context
2050: Level: developer
2052: .seealso: KSPSetPC(), KSP
2053: @*/
2054: PetscErrorCode KSPGetPC(KSP ksp,PC *pc)
2055: {
2061: if (!ksp->pc) {
2062: PCCreate(PetscObjectComm((PetscObject)ksp),&ksp->pc);
2063: PetscObjectIncrementTabLevel((PetscObject)ksp->pc,(PetscObject)ksp,0);
2064: PetscLogObjectParent((PetscObject)ksp,(PetscObject)ksp->pc);
2065: PetscObjectSetOptions((PetscObject)ksp->pc,((PetscObject)ksp)->options);
2066: }
2067: *pc = ksp->pc;
2068: return(0);
2069: }
2071: /*@
2072: KSPMonitor - runs the user provided monitor routines, if they exist
2074: Collective on ksp
2076: Input Parameters:
2077: + ksp - iterative context obtained from KSPCreate()
2078: . it - iteration number
2079: - rnorm - relative norm of the residual
2081: Notes:
2082: This routine is called by the KSP implementations.
2083: It does not typically need to be called by the user.
2085: Level: developer
2087: .seealso: KSPMonitorSet()
2088: @*/
2089: PetscErrorCode KSPMonitor(KSP ksp,PetscInt it,PetscReal rnorm)
2090: {
2091: PetscInt i, n = ksp->numbermonitors;
2095: for (i=0; i<n; i++) {
2096: (*ksp->monitor[i])(ksp,it,rnorm,ksp->monitorcontext[i]);
2097: }
2098: return(0);
2099: }
2101: /*@C
2102: KSPMonitorSet - Sets an ADDITIONAL function to be called at every iteration to monitor
2103: the residual/error etc.
2105: Logically Collective on ksp
2107: Input Parameters:
2108: + ksp - iterative context obtained from KSPCreate()
2109: . monitor - pointer to function (if this is NULL, it turns off monitoring
2110: . mctx - [optional] context for private data for the
2111: monitor routine (use NULL if no context is desired)
2112: - monitordestroy - [optional] routine that frees monitor context
2113: (may be NULL)
2115: Calling Sequence of monitor:
2116: $ monitor (KSP ksp, PetscInt it, PetscReal rnorm, void *mctx)
2118: + ksp - iterative context obtained from KSPCreate()
2119: . it - iteration number
2120: . rnorm - (estimated) 2-norm of (preconditioned) residual
2121: - mctx - optional monitoring context, as set by KSPMonitorSet()
2123: Options Database Keys:
2124: + -ksp_monitor - sets KSPMonitorResidual()
2125: . -ksp_monitor draw - sets KSPMonitorResidualDraw() and plots residual
2126: . -ksp_monitor draw::draw_lg - sets KSPMonitorResidualDrawLG() and plots residual
2127: . -ksp_monitor_pause_final - Pauses any graphics when the solve finishes (only works for internal monitors)
2128: . -ksp_monitor_true_residual - sets KSPMonitorTrueResidual()
2129: . -ksp_monitor_true_residual draw::draw_lg - sets KSPMonitorTrueResidualDrawLG() and plots residual
2130: . -ksp_monitor_max - sets KSPMonitorTrueResidualMax()
2131: . -ksp_monitor_singular_value - sets KSPMonitorSingularValue()
2132: - -ksp_monitor_cancel - cancels all monitors that have
2133: been hardwired into a code by
2134: calls to KSPMonitorSet(), but
2135: does not cancel those set via
2136: the options database.
2138: Notes:
2139: The default is to do nothing. To print the residual, or preconditioned
2140: residual if KSPSetNormType(ksp,KSP_NORM_PRECONDITIONED) was called, use
2141: KSPMonitorResidual() as the monitoring routine, with a ASCII viewer as the
2142: context.
2144: Several different monitoring routines may be set by calling
2145: KSPMonitorSet() multiple times; all will be called in the
2146: order in which they were set.
2148: Fortran Notes:
2149: Only a single monitor function can be set for each KSP object
2151: Level: beginner
2153: .seealso: KSPMonitorResidual(), KSPMonitorCancel(), KSP
2154: @*/
2155: PetscErrorCode KSPMonitorSet(KSP ksp,PetscErrorCode (*monitor)(KSP,PetscInt,PetscReal,void*),void *mctx,PetscErrorCode (*monitordestroy)(void**))
2156: {
2157: PetscInt i;
2159: PetscBool identical;
2163: for (i=0; i<ksp->numbermonitors;i++) {
2164: PetscMonitorCompare((PetscErrorCode (*)(void))monitor,mctx,monitordestroy,(PetscErrorCode (*)(void))ksp->monitor[i],ksp->monitorcontext[i],ksp->monitordestroy[i],&identical);
2165: if (identical) return(0);
2166: }
2167: if (ksp->numbermonitors >= MAXKSPMONITORS) SETERRQ(PetscObjectComm((PetscObject)ksp),PETSC_ERR_ARG_OUTOFRANGE,"Too many KSP monitors set");
2168: ksp->monitor[ksp->numbermonitors] = monitor;
2169: ksp->monitordestroy[ksp->numbermonitors] = monitordestroy;
2170: ksp->monitorcontext[ksp->numbermonitors++] = (void*)mctx;
2171: return(0);
2172: }
2174: /*@
2175: KSPMonitorCancel - Clears all monitors for a KSP object.
2177: Logically Collective on ksp
2179: Input Parameters:
2180: . ksp - iterative context obtained from KSPCreate()
2182: Options Database Key:
2183: . -ksp_monitor_cancel - Cancels all monitors that have
2184: been hardwired into a code by calls to KSPMonitorSet(),
2185: but does not cancel those set via the options database.
2187: Level: intermediate
2189: .seealso: KSPMonitorResidual(), KSPMonitorSet(), KSP
2190: @*/
2191: PetscErrorCode KSPMonitorCancel(KSP ksp)
2192: {
2194: PetscInt i;
2198: for (i=0; i<ksp->numbermonitors; i++) {
2199: if (ksp->monitordestroy[i]) {
2200: (*ksp->monitordestroy[i])(&ksp->monitorcontext[i]);
2201: }
2202: }
2203: ksp->numbermonitors = 0;
2204: return(0);
2205: }
2207: /*@C
2208: KSPGetMonitorContext - Gets the monitoring context, as set by
2209: KSPMonitorSet() for the FIRST monitor only.
2211: Not Collective
2213: Input Parameter:
2214: . ksp - iterative context obtained from KSPCreate()
2216: Output Parameter:
2217: . ctx - monitoring context
2219: Level: intermediate
2221: .seealso: KSPMonitorResidual(), KSP
2222: @*/
2223: PetscErrorCode KSPGetMonitorContext(KSP ksp,void **ctx)
2224: {
2227: *ctx = (ksp->monitorcontext[0]);
2228: return(0);
2229: }
2231: /*@
2232: KSPSetResidualHistory - Sets the array used to hold the residual history.
2233: If set, this array will contain the residual norms computed at each
2234: iteration of the solver.
2236: Not Collective
2238: Input Parameters:
2239: + ksp - iterative context obtained from KSPCreate()
2240: . a - array to hold history
2241: . na - size of a
2242: - reset - PETSC_TRUE indicates the history counter is reset to zero
2243: for each new linear solve
2245: Level: advanced
2247: Notes:
2248: If provided, he array is NOT freed by PETSc so the user needs to keep track of it and destroy once the KSP object is destroyed.
2249: If 'a' is NULL then space is allocated for the history. If 'na' PETSC_DECIDE or PETSC_DEFAULT then a
2250: default array of length 10000 is allocated.
2252: .seealso: KSPGetResidualHistory(), KSP
2254: @*/
2255: PetscErrorCode KSPSetResidualHistory(KSP ksp,PetscReal a[],PetscInt na,PetscBool reset)
2256: {
2262: PetscFree(ksp->res_hist_alloc);
2263: if (na != PETSC_DECIDE && na != PETSC_DEFAULT && a) {
2264: ksp->res_hist = a;
2265: ksp->res_hist_max = na;
2266: } else {
2267: if (na != PETSC_DECIDE && na != PETSC_DEFAULT) ksp->res_hist_max = na;
2268: else ksp->res_hist_max = 10000; /* like default ksp->max_it */
2269: PetscCalloc1(ksp->res_hist_max,&ksp->res_hist_alloc);
2271: ksp->res_hist = ksp->res_hist_alloc;
2272: }
2273: ksp->res_hist_len = 0;
2274: ksp->res_hist_reset = reset;
2275: return(0);
2276: }
2278: /*@C
2279: KSPGetResidualHistory - Gets the array used to hold the residual history
2280: and the number of residuals it contains.
2282: Not Collective
2284: Input Parameter:
2285: . ksp - iterative context obtained from KSPCreate()
2287: Output Parameters:
2288: + a - pointer to array to hold history (or NULL)
2289: - na - number of used entries in a (or NULL)
2291: Level: advanced
2293: Notes:
2294: This array is borrowed and should not be freed by the caller.
2295: Can only be called after a KSPSetResidualHistory() otherwise a and na are set to zero
2297: The Fortran version of this routine has a calling sequence
2298: $ call KSPGetResidualHistory(KSP ksp, integer na, integer ierr)
2299: note that you have passed a Fortran array into KSPSetResidualHistory() and you need
2300: to access the residual values from this Fortran array you provided. Only the na (number of
2301: residual norms currently held) is set.
2303: .seealso: KSPSetResidualHistory(), KSP
2305: @*/
2306: PetscErrorCode KSPGetResidualHistory(KSP ksp, const PetscReal *a[],PetscInt *na)
2307: {
2310: if (a) *a = ksp->res_hist;
2311: if (na) *na = ksp->res_hist_len;
2312: return(0);
2313: }
2315: /*@
2316: KSPSetErrorHistory - Sets the array used to hold the error history. If set, this array will contain the error norms computed at each iteration of the solver.
2318: Not Collective
2320: Input Parameters:
2321: + ksp - iterative context obtained from KSPCreate()
2322: . a - array to hold history
2323: . na - size of a
2324: - reset - PETSC_TRUE indicates the history counter is reset to zero for each new linear solve
2326: Level: advanced
2328: Notes:
2329: If provided, the array is NOT freed by PETSc so the user needs to keep track of it and destroy once the KSP object is destroyed.
2330: If 'a' is NULL then space is allocated for the history. If 'na' PETSC_DECIDE or PETSC_DEFAULT then a default array of length 10000 is allocated.
2332: .seealso: KSPGetErrorHistory(), KSPSetResidualHistory(), KSP
2333: @*/
2334: PetscErrorCode KSPSetErrorHistory(KSP ksp, PetscReal a[], PetscInt na, PetscBool reset)
2335: {
2341: PetscFree(ksp->err_hist_alloc);
2342: if (na != PETSC_DECIDE && na != PETSC_DEFAULT && a) {
2343: ksp->err_hist = a;
2344: ksp->err_hist_max = na;
2345: } else {
2346: if (na != PETSC_DECIDE && na != PETSC_DEFAULT) ksp->err_hist_max = na;
2347: else ksp->err_hist_max = 10000; /* like default ksp->max_it */
2348: PetscCalloc1(ksp->err_hist_max, &ksp->err_hist_alloc);
2350: ksp->err_hist = ksp->err_hist_alloc;
2351: }
2352: ksp->err_hist_len = 0;
2353: ksp->err_hist_reset = reset;
2354: return(0);
2355: }
2357: /*@C
2358: KSPGetErrorHistory - Gets the array used to hold the error history and the number of residuals it contains.
2360: Not Collective
2362: Input Parameter:
2363: . ksp - iterative context obtained from KSPCreate()
2365: Output Parameters:
2366: + a - pointer to array to hold history (or NULL)
2367: - na - number of used entries in a (or NULL)
2369: Level: advanced
2371: Notes:
2372: This array is borrowed and should not be freed by the caller.
2373: Can only be called after a KSPSetErrorHistory() otherwise a and na are set to zero
2374: The Fortran version of this routine has a calling sequence
2375: $ call KSPGetErrorHistory(KSP ksp, integer na, integer ierr)
2376: note that you have passed a Fortran array into KSPSetErrorHistory() and you need
2377: to access the residual values from this Fortran array you provided. Only the na (number of
2378: residual norms currently held) is set.
2380: .seealso: KSPSetErrorHistory(), KSPGetResidualHistory(), KSP
2381: @*/
2382: PetscErrorCode KSPGetErrorHistory(KSP ksp, const PetscReal *a[], PetscInt *na)
2383: {
2386: if (a) *a = ksp->err_hist;
2387: if (na) *na = ksp->err_hist_len;
2388: return(0);
2389: }
2391: /*
2392: KSPComputeConvergenceRate - Compute the convergence rate for the iteration
2394: Not collective
2396: Input Parameter:
2397: . ksp - The KSP
2399: Output Parameters:
2400: + cr - The residual contraction rate
2401: . rRsq - The coefficient of determination, R^2, indicating the linearity of the data
2402: . ce - The error contraction rate
2403: - eRsq - The coefficient of determination, R^2, indicating the linearity of the data
2405: Note:
2406: Suppose that the residual is reduced linearly, $r_k = c^k r_0$, which means $log r_k = log r_0 + k log c$. After linear regression,
2407: the slope is $\log c$. The coefficient of determination is given by $1 - \frac{\sum_i (y_i - f(x_i))^2}{\sum_i (y_i - \bar y)}$,
2408: see also https://en.wikipedia.org/wiki/Coefficient_of_determination
2410: Level: advanced
2412: .seealso: KSPConvergedRateView()
2413: */
2414: PetscErrorCode KSPComputeConvergenceRate(KSP ksp, PetscReal *cr, PetscReal *rRsq, PetscReal *ce, PetscReal *eRsq)
2415: {
2416: PetscReal const *hist;
2417: PetscReal *x, *y, slope, intercept, mean = 0.0, var = 0.0, res = 0.0;
2418: PetscInt n, k;
2422: if (cr || rRsq) {
2423: KSPGetResidualHistory(ksp, &hist, &n);
2424: if (!n) {
2425: if (cr) *cr = 0.0;
2426: if (rRsq) *rRsq = -1.0;
2427: } else {
2428: PetscMalloc2(n, &x, n, &y);
2429: for (k = 0; k < n; ++k) {
2430: x[k] = k;
2431: y[k] = PetscLogReal(hist[k]);
2432: mean += y[k];
2433: }
2434: mean /= n;
2435: PetscLinearRegression(n, x, y, &slope, &intercept);
2436: for (k = 0; k < n; ++k) {
2437: res += PetscSqr(y[k] - (slope*x[k] + intercept));
2438: var += PetscSqr(y[k] - mean);
2439: }
2440: PetscFree2(x, y);
2441: if (cr) *cr = PetscExpReal(slope);
2442: if (rRsq) *rRsq = var < PETSC_MACHINE_EPSILON ? 0.0 : 1.0 - (res / var);
2443: }
2444: }
2445: if (ce || eRsq) {
2446: KSPGetErrorHistory(ksp, &hist, &n);
2447: if (!n) {
2448: if (ce) *ce = 0.0;
2449: if (eRsq) *eRsq = -1.0;
2450: } else {
2451: PetscMalloc2(n, &x, n, &y);
2452: for (k = 0; k < n; ++k) {
2453: x[k] = k;
2454: y[k] = PetscLogReal(hist[k]);
2455: mean += y[k];
2456: }
2457: mean /= n;
2458: PetscLinearRegression(n, x, y, &slope, &intercept);
2459: for (k = 0; k < n; ++k) {
2460: res += PetscSqr(y[k] - (slope*x[k] + intercept));
2461: var += PetscSqr(y[k] - mean);
2462: }
2463: PetscFree2(x, y);
2464: if (ce) *ce = PetscExpReal(slope);
2465: if (eRsq) *eRsq = var < PETSC_MACHINE_EPSILON ? 0.0 : 1.0 - (res / var);
2466: }
2467: }
2468: return(0);
2469: }
2471: /*@C
2472: KSPSetConvergenceTest - Sets the function to be used to determine
2473: convergence.
2475: Logically Collective on ksp
2477: Input Parameters:
2478: + ksp - iterative context obtained from KSPCreate()
2479: . converge - pointer to the function
2480: . cctx - context for private data for the convergence routine (may be null)
2481: - destroy - a routine for destroying the context (may be null)
2483: Calling sequence of converge:
2484: $ converge (KSP ksp, PetscInt it, PetscReal rnorm, KSPConvergedReason *reason,void *mctx)
2486: + ksp - iterative context obtained from KSPCreate()
2487: . it - iteration number
2488: . rnorm - (estimated) 2-norm of (preconditioned) residual
2489: . reason - the reason why it has converged or diverged
2490: - cctx - optional convergence context, as set by KSPSetConvergenceTest()
2493: Notes:
2494: Must be called after the KSP type has been set so put this after
2495: a call to KSPSetType(), or KSPSetFromOptions().
2497: The default convergence test, KSPConvergedDefault(), aborts if the
2498: residual grows to more than 10000 times the initial residual.
2500: The default is a combination of relative and absolute tolerances.
2501: The residual value that is tested may be an approximation; routines
2502: that need exact values should compute them.
2504: In the default PETSc convergence test, the precise values of reason
2505: are macros such as KSP_CONVERGED_RTOL, which are defined in petscksp.h.
2507: Level: advanced
2509: .seealso: KSPConvergedDefault(), KSPGetConvergenceContext(), KSPSetTolerances(), KSP, KSPGetConvergenceTest(), KSPGetAndClearConvergenceTest()
2510: @*/
2511: PetscErrorCode KSPSetConvergenceTest(KSP ksp,PetscErrorCode (*converge)(KSP,PetscInt,PetscReal,KSPConvergedReason*,void*),void *cctx,PetscErrorCode (*destroy)(void*))
2512: {
2517: if (ksp->convergeddestroy) {
2518: (*ksp->convergeddestroy)(ksp->cnvP);
2519: }
2520: ksp->converged = converge;
2521: ksp->convergeddestroy = destroy;
2522: ksp->cnvP = (void*)cctx;
2523: return(0);
2524: }
2526: /*@C
2527: KSPGetConvergenceTest - Gets the function to be used to determine
2528: convergence.
2530: Logically Collective on ksp
2532: Input Parameter:
2533: . ksp - iterative context obtained from KSPCreate()
2535: Output Parameter:
2536: + converge - pointer to convergence test function
2537: . cctx - context for private data for the convergence routine (may be null)
2538: - destroy - a routine for destroying the context (may be null)
2540: Calling sequence of converge:
2541: $ converge (KSP ksp, PetscInt it, PetscReal rnorm, KSPConvergedReason *reason,void *mctx)
2543: + ksp - iterative context obtained from KSPCreate()
2544: . it - iteration number
2545: . rnorm - (estimated) 2-norm of (preconditioned) residual
2546: . reason - the reason why it has converged or diverged
2547: - cctx - optional convergence context, as set by KSPSetConvergenceTest()
2549: Level: advanced
2551: .seealso: KSPConvergedDefault(), KSPGetConvergenceContext(), KSPSetTolerances(), KSP, KSPSetConvergenceTest(), KSPGetAndClearConvergenceTest()
2552: @*/
2553: PetscErrorCode KSPGetConvergenceTest(KSP ksp,PetscErrorCode (**converge)(KSP,PetscInt,PetscReal,KSPConvergedReason*,void*),void **cctx,PetscErrorCode (**destroy)(void*))
2554: {
2557: if (converge) *converge = ksp->converged;
2558: if (destroy) *destroy = ksp->convergeddestroy;
2559: if (cctx) *cctx = ksp->cnvP;
2560: return(0);
2561: }
2563: /*@C
2564: KSPGetAndClearConvergenceTest - Gets the function to be used to determine convergence. Removes the current test without calling destroy on the test context
2566: Logically Collective on ksp
2568: Input Parameter:
2569: . ksp - iterative context obtained from KSPCreate()
2571: Output Parameter:
2572: + converge - pointer to convergence test function
2573: . cctx - context for private data for the convergence routine
2574: - destroy - a routine for destroying the context
2576: Calling sequence of converge:
2577: $ converge (KSP ksp, PetscInt it, PetscReal rnorm, KSPConvergedReason *reason,void *mctx)
2579: + ksp - iterative context obtained from KSPCreate()
2580: . it - iteration number
2581: . rnorm - (estimated) 2-norm of (preconditioned) residual
2582: . reason - the reason why it has converged or diverged
2583: - cctx - optional convergence context, as set by KSPSetConvergenceTest()
2585: Level: advanced
2587: Notes: This is intended to be used to allow transferring the convergence test (and its context) to another testing object (for example another KSP) and then calling
2588: KSPSetConvergenceTest() on this original KSP. If you just called KSPGetConvergenceTest() followed by KSPSetConvergenceTest() the original context information
2589: would be destroyed and hence the transferred context would be invalid and trigger a crash on use
2591: .seealso: KSPConvergedDefault(), KSPGetConvergenceContext(), KSPSetTolerances(), KSP, KSPSetConvergenceTest(), KSPGetConvergenceTest()
2592: @*/
2593: PetscErrorCode KSPGetAndClearConvergenceTest(KSP ksp,PetscErrorCode (**converge)(KSP,PetscInt,PetscReal,KSPConvergedReason*,void*),void **cctx,PetscErrorCode (**destroy)(void*))
2594: {
2597: *converge = ksp->converged;
2598: *destroy = ksp->convergeddestroy;
2599: *cctx = ksp->cnvP;
2600: ksp->converged = NULL;
2601: ksp->cnvP = NULL;
2602: ksp->convergeddestroy = NULL;
2603: return(0);
2604: }
2606: /*@C
2607: KSPGetConvergenceContext - Gets the convergence context set with
2608: KSPSetConvergenceTest().
2610: Not Collective
2612: Input Parameter:
2613: . ksp - iterative context obtained from KSPCreate()
2615: Output Parameter:
2616: . ctx - monitoring context
2618: Level: advanced
2620: .seealso: KSPConvergedDefault(), KSPSetConvergenceTest(), KSP
2621: @*/
2622: PetscErrorCode KSPGetConvergenceContext(KSP ksp,void **ctx)
2623: {
2626: *ctx = ksp->cnvP;
2627: return(0);
2628: }
2630: /*@C
2631: KSPBuildSolution - Builds the approximate solution in a vector provided.
2632: This routine is NOT commonly needed (see KSPSolve()).
2634: Collective on ksp
2636: Input Parameter:
2637: . ctx - iterative context obtained from KSPCreate()
2639: Output Parameter:
2640: Provide exactly one of
2641: + v - location to stash solution.
2642: - V - the solution is returned in this location. This vector is created
2643: internally. This vector should NOT be destroyed by the user with
2644: VecDestroy().
2646: Notes:
2647: This routine can be used in one of two ways
2648: .vb
2649: KSPBuildSolution(ksp,NULL,&V);
2650: or
2651: KSPBuildSolution(ksp,v,NULL); or KSPBuildSolution(ksp,v,&v);
2652: .ve
2653: In the first case an internal vector is allocated to store the solution
2654: (the user cannot destroy this vector). In the second case the solution
2655: is generated in the vector that the user provides. Note that for certain
2656: methods, such as KSPCG, the second case requires a copy of the solution,
2657: while in the first case the call is essentially free since it simply
2658: returns the vector where the solution already is stored. For some methods
2659: like GMRES this is a reasonably expensive operation and should only be
2660: used in truly needed.
2662: Level: advanced
2664: .seealso: KSPGetSolution(), KSPBuildResidual(), KSP
2665: @*/
2666: PetscErrorCode KSPBuildSolution(KSP ksp,Vec v,Vec *V)
2667: {
2672: if (!V && !v) SETERRQ(PetscObjectComm((PetscObject)ksp),PETSC_ERR_ARG_WRONG,"Must provide either v or V");
2673: if (!V) V = &v;
2674: (*ksp->ops->buildsolution)(ksp,v,V);
2675: return(0);
2676: }
2678: /*@C
2679: KSPBuildResidual - Builds the residual in a vector provided.
2681: Collective on ksp
2683: Input Parameter:
2684: . ksp - iterative context obtained from KSPCreate()
2686: Output Parameters:
2687: + v - optional location to stash residual. If v is not provided,
2688: then a location is generated.
2689: . t - work vector. If not provided then one is generated.
2690: - V - the residual
2692: Notes:
2693: Regardless of whether or not v is provided, the residual is
2694: returned in V.
2696: Level: advanced
2698: .seealso: KSPBuildSolution()
2699: @*/
2700: PetscErrorCode KSPBuildResidual(KSP ksp,Vec t,Vec v,Vec *V)
2701: {
2703: PetscBool flag = PETSC_FALSE;
2704: Vec w = v,tt = t;
2708: if (!w) {
2709: VecDuplicate(ksp->vec_rhs,&w);
2710: PetscLogObjectParent((PetscObject)ksp,(PetscObject)w);
2711: }
2712: if (!tt) {
2713: VecDuplicate(ksp->vec_sol,&tt); flag = PETSC_TRUE;
2714: PetscLogObjectParent((PetscObject)ksp,(PetscObject)tt);
2715: }
2716: (*ksp->ops->buildresidual)(ksp,tt,w,V);
2717: if (flag) {VecDestroy(&tt);}
2718: return(0);
2719: }
2721: /*@
2722: KSPSetDiagonalScale - Tells KSP to symmetrically diagonally scale the system
2723: before solving. This actually CHANGES the matrix (and right hand side).
2725: Logically Collective on ksp
2727: Input Parameter:
2728: + ksp - the KSP context
2729: - scale - PETSC_TRUE or PETSC_FALSE
2731: Options Database Key:
2732: + -ksp_diagonal_scale -
2733: - -ksp_diagonal_scale_fix - scale the matrix back AFTER the solve
2736: Notes:
2737: Scales the matrix by D^(-1/2) A D^(-1/2) [D^(1/2) x ] = D^(-1/2) b
2738: where D_{ii} is 1/abs(A_{ii}) unless A_{ii} is zero and then it is 1.
2740: BE CAREFUL with this routine: it actually scales the matrix and right
2741: hand side that define the system. After the system is solved the matrix
2742: and right hand side remain scaled unless you use KSPSetDiagonalScaleFix()
2744: This should NOT be used within the SNES solves if you are using a line
2745: search.
2747: If you use this with the PCType Eisenstat preconditioner than you can
2748: use the PCEisenstatSetNoDiagonalScaling() option, or -pc_eisenstat_no_diagonal_scaling
2749: to save some unneeded, redundant flops.
2751: Level: intermediate
2753: .seealso: KSPGetDiagonalScale(), KSPSetDiagonalScaleFix(), KSP
2754: @*/
2755: PetscErrorCode KSPSetDiagonalScale(KSP ksp,PetscBool scale)
2756: {
2760: ksp->dscale = scale;
2761: return(0);
2762: }
2764: /*@
2765: KSPGetDiagonalScale - Checks if KSP solver scales the matrix and
2766: right hand side
2768: Not Collective
2770: Input Parameter:
2771: . ksp - the KSP context
2773: Output Parameter:
2774: . scale - PETSC_TRUE or PETSC_FALSE
2776: Notes:
2777: BE CAREFUL with this routine: it actually scales the matrix and right
2778: hand side that define the system. After the system is solved the matrix
2779: and right hand side remain scaled unless you use KSPSetDiagonalScaleFix()
2781: Level: intermediate
2783: .seealso: KSPSetDiagonalScale(), KSPSetDiagonalScaleFix(), KSP
2784: @*/
2785: PetscErrorCode KSPGetDiagonalScale(KSP ksp,PetscBool *scale)
2786: {
2790: *scale = ksp->dscale;
2791: return(0);
2792: }
2794: /*@
2795: KSPSetDiagonalScaleFix - Tells KSP to diagonally scale the system
2796: back after solving.
2798: Logically Collective on ksp
2800: Input Parameter:
2801: + ksp - the KSP context
2802: - fix - PETSC_TRUE to scale back after the system solve, PETSC_FALSE to not
2803: rescale (default)
2805: Notes:
2806: Must be called after KSPSetDiagonalScale()
2808: Using this will slow things down, because it rescales the matrix before and
2809: after each linear solve. This is intended mainly for testing to allow one
2810: to easily get back the original system to make sure the solution computed is
2811: accurate enough.
2813: Level: intermediate
2815: .seealso: KSPGetDiagonalScale(), KSPSetDiagonalScale(), KSPGetDiagonalScaleFix(), KSP
2816: @*/
2817: PetscErrorCode KSPSetDiagonalScaleFix(KSP ksp,PetscBool fix)
2818: {
2822: ksp->dscalefix = fix;
2823: return(0);
2824: }
2826: /*@
2827: KSPGetDiagonalScaleFix - Determines if KSP diagonally scales the system
2828: back after solving.
2830: Not Collective
2832: Input Parameter:
2833: . ksp - the KSP context
2835: Output Parameter:
2836: . fix - PETSC_TRUE to scale back after the system solve, PETSC_FALSE to not
2837: rescale (default)
2839: Notes:
2840: Must be called after KSPSetDiagonalScale()
2842: If PETSC_TRUE will slow things down, because it rescales the matrix before and
2843: after each linear solve. This is intended mainly for testing to allow one
2844: to easily get back the original system to make sure the solution computed is
2845: accurate enough.
2847: Level: intermediate
2849: .seealso: KSPGetDiagonalScale(), KSPSetDiagonalScale(), KSPSetDiagonalScaleFix(), KSP
2850: @*/
2851: PetscErrorCode KSPGetDiagonalScaleFix(KSP ksp,PetscBool *fix)
2852: {
2856: *fix = ksp->dscalefix;
2857: return(0);
2858: }
2860: /*@C
2861: KSPSetComputeOperators - set routine to compute the linear operators
2863: Logically Collective
2865: Input Arguments:
2866: + ksp - the KSP context
2867: . func - function to compute the operators
2868: - ctx - optional context
2870: Calling sequence of func:
2871: $ func(KSP ksp,Mat A,Mat B,void *ctx)
2873: + ksp - the KSP context
2874: . A - the linear operator
2875: . B - preconditioning matrix
2876: - ctx - optional user-provided context
2878: Notes:
2879: The user provided func() will be called automatically at the very next call to KSPSolve(). It will not be called at future KSPSolve() calls
2880: unless either KSPSetComputeOperators() or KSPSetOperators() is called before that KSPSolve() is called.
2882: To reuse the same preconditioner for the next KSPSolve() and not compute a new one based on the most recently computed matrix call KSPSetReusePreconditioner()
2884: Level: beginner
2886: .seealso: KSPSetOperators(), KSPSetComputeRHS(), DMKSPSetComputeOperators(), KSPSetComputeInitialGuess()
2887: @*/
2888: PetscErrorCode KSPSetComputeOperators(KSP ksp,PetscErrorCode (*func)(KSP,Mat,Mat,void*),void *ctx)
2889: {
2891: DM dm;
2895: KSPGetDM(ksp,&dm);
2896: DMKSPSetComputeOperators(dm,func,ctx);
2897: if (ksp->setupstage == KSP_SETUP_NEWRHS) ksp->setupstage = KSP_SETUP_NEWMATRIX;
2898: return(0);
2899: }
2901: /*@C
2902: KSPSetComputeRHS - set routine to compute the right hand side of the linear system
2904: Logically Collective
2906: Input Arguments:
2907: + ksp - the KSP context
2908: . func - function to compute the right hand side
2909: - ctx - optional context
2911: Calling sequence of func:
2912: $ func(KSP ksp,Vec b,void *ctx)
2914: + ksp - the KSP context
2915: . b - right hand side of linear system
2916: - ctx - optional user-provided context
2918: Notes:
2919: The routine you provide will be called EACH you call KSPSolve() to prepare the new right hand side for that solve
2921: Level: beginner
2923: .seealso: KSPSolve(), DMKSPSetComputeRHS(), KSPSetComputeOperators()
2924: @*/
2925: PetscErrorCode KSPSetComputeRHS(KSP ksp,PetscErrorCode (*func)(KSP,Vec,void*),void *ctx)
2926: {
2928: DM dm;
2932: KSPGetDM(ksp,&dm);
2933: DMKSPSetComputeRHS(dm,func,ctx);
2934: return(0);
2935: }
2937: /*@C
2938: KSPSetComputeInitialGuess - set routine to compute the initial guess of the linear system
2940: Logically Collective
2942: Input Arguments:
2943: + ksp - the KSP context
2944: . func - function to compute the initial guess
2945: - ctx - optional context
2947: Calling sequence of func:
2948: $ func(KSP ksp,Vec x,void *ctx)
2950: + ksp - the KSP context
2951: . x - solution vector
2952: - ctx - optional user-provided context
2954: Notes: This should only be used in conjunction with KSPSetComputeRHS(), KSPSetComputeOperators(), otherwise
2955: call KSPSetInitialGuessNonzero() and set the initial guess values in the solution vector passed to KSPSolve().
2957: Level: beginner
2959: .seealso: KSPSolve(), KSPSetComputeRHS(), KSPSetComputeOperators(), DMKSPSetComputeInitialGuess()
2960: @*/
2961: PetscErrorCode KSPSetComputeInitialGuess(KSP ksp,PetscErrorCode (*func)(KSP,Vec,void*),void *ctx)
2962: {
2964: DM dm;
2968: KSPGetDM(ksp,&dm);
2969: DMKSPSetComputeInitialGuess(dm,func,ctx);
2970: return(0);
2971: }
2973: /*@
2974: KSPSetUseExplicitTranspose - Determines if transpose the system explicitly
2975: in KSPSolveTranspose.
2977: Logically Collective on ksp
2979: Input Parameter:
2980: . ksp - the KSP context
2982: Output Parameter:
2983: . flg - PETSC_TRUE to transpose the system in KSPSolveTranspose, PETSC_FALSE to not
2984: transpose (default)
2986: Level: advanced
2988: .seealso: KSPSolveTranspose(), KSP
2989: @*/
2990: PetscErrorCode KSPSetUseExplicitTranspose(KSP ksp,PetscBool flg)
2991: {
2994: ksp->transpose.use_explicittranspose = flg;
2995: return(0);
2996: }