Actual source code: pepopts.c
1: /*
2: - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
3: SLEPc - Scalable Library for Eigenvalue Problem Computations
4: Copyright (c) 2002-, Universitat Politecnica de Valencia, Spain
6: This file is part of SLEPc.
7: SLEPc is distributed under a 2-clause BSD license (see LICENSE).
8: - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
9: */
10: /*
11: PEP routines related to options that can be set via the command-line
12: or procedurally
13: */
15: #include <slepc/private/pepimpl.h>
16: #include <petscdraw.h>
18: /*@C
19: PEPMonitorSetFromOptions - Sets a monitor function and viewer appropriate for the type
20: indicated by the user.
22: Collective
24: Input Parameters:
25: + pep - the polynomial eigensolver context
26: . opt - the command line option for this monitor
27: . name - the monitor type one is seeking
28: . ctx - an optional user context for the monitor, or `NULL`
29: - trackall - whether this monitor tracks all eigenvalues or not
31: Level: developer
33: .seealso: [](ch:pep), `PEPMonitorSet()`, `PEPSetTrackAll()`
34: @*/
35: PetscErrorCode PEPMonitorSetFromOptions(PEP pep,const char opt[],const char name[],void *ctx,PetscBool trackall)
36: {
37: PetscErrorCode (*mfunc)(PEP,PetscInt,PetscInt,PetscScalar*,PetscScalar*,PetscReal*,PetscInt,void*);
38: PetscErrorCode (*cfunc)(PetscViewer,PetscViewerFormat,void*,PetscViewerAndFormat**);
39: PetscErrorCode (*dfunc)(PetscViewerAndFormat**);
40: PetscViewerAndFormat *vf;
41: PetscViewer viewer;
42: PetscViewerFormat format;
43: PetscViewerType vtype;
44: char key[PETSC_MAX_PATH_LEN];
45: PetscBool flg;
47: PetscFunctionBegin;
48: PetscCall(PetscOptionsCreateViewer(PetscObjectComm((PetscObject)pep),((PetscObject)pep)->options,((PetscObject)pep)->prefix,opt,&viewer,&format,&flg));
49: if (!flg) PetscFunctionReturn(PETSC_SUCCESS);
51: PetscCall(PetscViewerGetType(viewer,&vtype));
52: PetscCall(SlepcMonitorMakeKey_Internal(name,vtype,format,key));
53: PetscCall(PetscFunctionListFind(PEPMonitorList,key,&mfunc));
54: PetscCheck(mfunc,PetscObjectComm((PetscObject)pep),PETSC_ERR_SUP,"Specified viewer and format not supported");
55: PetscCall(PetscFunctionListFind(PEPMonitorCreateList,key,&cfunc));
56: PetscCall(PetscFunctionListFind(PEPMonitorDestroyList,key,&dfunc));
57: if (!cfunc) cfunc = PetscViewerAndFormatCreate_Internal;
58: if (!dfunc) dfunc = PetscViewerAndFormatDestroy;
60: PetscCall((*cfunc)(viewer,format,ctx,&vf));
61: PetscCall(PetscViewerDestroy(&viewer));
62: PetscCall(PEPMonitorSet(pep,mfunc,vf,(PetscCtxDestroyFn*)dfunc));
63: if (trackall) PetscCall(PEPSetTrackAll(pep,PETSC_TRUE));
64: PetscFunctionReturn(PETSC_SUCCESS);
65: }
67: /*@
68: PEPSetFromOptions - Sets `PEP` options from the options database.
69: This routine must be called before `PEPSetUp()` if the user is to be
70: allowed to configure the solver.
72: Collective
74: Input Parameter:
75: . pep - the polynomial eigensolver context
77: Note:
78: To see all options, run your program with the `-help` option.
80: Level: beginner
82: .seealso: [](ch:pep), `PEPSetOptionsPrefix()`
83: @*/
84: PetscErrorCode PEPSetFromOptions(PEP pep)
85: {
86: char type[256];
87: PetscBool set,flg,flg1,flg2,flg3,flg4,flg5;
88: PetscReal r,t,array[2]={0,0};
89: PetscScalar s;
90: PetscInt i,j,k;
91: PEPScale scale;
92: PEPRefine refine;
93: PEPRefineScheme scheme;
95: PetscFunctionBegin;
97: PetscCall(PEPRegisterAll());
98: PetscObjectOptionsBegin((PetscObject)pep);
99: PetscCall(PetscOptionsFList("-pep_type","Polynomial eigensolver method","PEPSetType",PEPList,(char*)(((PetscObject)pep)->type_name?((PetscObject)pep)->type_name:PEPTOAR),type,sizeof(type),&flg));
100: if (flg) PetscCall(PEPSetType(pep,type));
101: else if (!((PetscObject)pep)->type_name) PetscCall(PEPSetType(pep,PEPTOAR));
103: PetscCall(PetscOptionsBoolGroupBegin("-pep_general","General polynomial eigenvalue problem","PEPSetProblemType",&flg));
104: if (flg) PetscCall(PEPSetProblemType(pep,PEP_GENERAL));
105: PetscCall(PetscOptionsBoolGroup("-pep_hermitian","Hermitian polynomial eigenvalue problem","PEPSetProblemType",&flg));
106: if (flg) PetscCall(PEPSetProblemType(pep,PEP_HERMITIAN));
107: PetscCall(PetscOptionsBoolGroup("-pep_hyperbolic","Hyperbolic polynomial eigenvalue problem","PEPSetProblemType",&flg));
108: if (flg) PetscCall(PEPSetProblemType(pep,PEP_HYPERBOLIC));
109: PetscCall(PetscOptionsBoolGroupEnd("-pep_gyroscopic","Gyroscopic polynomial eigenvalue problem","PEPSetProblemType",&flg));
110: if (flg) PetscCall(PEPSetProblemType(pep,PEP_GYROSCOPIC));
112: scale = pep->scale;
113: PetscCall(PetscOptionsEnum("-pep_scale","Scaling strategy","PEPSetScale",PEPScaleTypes,(PetscEnum)scale,(PetscEnum*)&scale,&flg1));
114: r = pep->sfactor;
115: PetscCall(PetscOptionsReal("-pep_scale_factor","Scale factor","PEPSetScale",pep->sfactor,&r,&flg2));
116: if (!flg2 && r==1.0) r = PETSC_DETERMINE;
117: j = pep->sits;
118: PetscCall(PetscOptionsInt("-pep_scale_its","Number of iterations in diagonal scaling","PEPSetScale",pep->sits,&j,&flg3));
119: t = pep->slambda;
120: PetscCall(PetscOptionsReal("-pep_scale_lambda","Estimate of eigenvalue (modulus) for diagonal scaling","PEPSetScale",pep->slambda,&t,&flg4));
121: if (flg1 || flg2 || flg3 || flg4) PetscCall(PEPSetScale(pep,scale,r,NULL,NULL,j,t));
123: PetscCall(PetscOptionsEnum("-pep_extract","Extraction method","PEPSetExtract",PEPExtractTypes,(PetscEnum)pep->extract,(PetscEnum*)&pep->extract,NULL));
125: refine = pep->refine;
126: PetscCall(PetscOptionsEnum("-pep_refine","Iterative refinement method","PEPSetRefine",PEPRefineTypes,(PetscEnum)refine,(PetscEnum*)&refine,&flg1));
127: i = pep->npart;
128: PetscCall(PetscOptionsInt("-pep_refine_partitions","Number of partitions of the communicator for iterative refinement","PEPSetRefine",pep->npart,&i,&flg2));
129: r = pep->rtol;
130: PetscCall(PetscOptionsReal("-pep_refine_tol","Tolerance for iterative refinement","PEPSetRefine",pep->rtol==(PetscReal)PETSC_DETERMINE?SLEPC_DEFAULT_TOL/1000:pep->rtol,&r,&flg3));
131: j = pep->rits;
132: PetscCall(PetscOptionsInt("-pep_refine_its","Maximum number of iterations for iterative refinement","PEPSetRefine",pep->rits,&j,&flg4));
133: scheme = pep->scheme;
134: PetscCall(PetscOptionsEnum("-pep_refine_scheme","Scheme used for linear systems within iterative refinement","PEPSetRefine",PEPRefineSchemes,(PetscEnum)scheme,(PetscEnum*)&scheme,&flg5));
135: if (flg1 || flg2 || flg3 || flg4 || flg5) PetscCall(PEPSetRefine(pep,refine,i,r,j,scheme));
137: i = pep->max_it;
138: PetscCall(PetscOptionsInt("-pep_max_it","Maximum number of iterations","PEPSetTolerances",pep->max_it,&i,&flg1));
139: r = pep->tol;
140: PetscCall(PetscOptionsReal("-pep_tol","Tolerance","PEPSetTolerances",SlepcDefaultTol(pep->tol),&r,&flg2));
141: if (flg1 || flg2) PetscCall(PEPSetTolerances(pep,r,i));
143: PetscCall(PetscOptionsBoolGroupBegin("-pep_conv_rel","Relative error convergence test","PEPSetConvergenceTest",&flg));
144: if (flg) PetscCall(PEPSetConvergenceTest(pep,PEP_CONV_REL));
145: PetscCall(PetscOptionsBoolGroup("-pep_conv_norm","Convergence test relative to the matrix norms","PEPSetConvergenceTest",&flg));
146: if (flg) PetscCall(PEPSetConvergenceTest(pep,PEP_CONV_NORM));
147: PetscCall(PetscOptionsBoolGroup("-pep_conv_abs","Absolute error convergence test","PEPSetConvergenceTest",&flg));
148: if (flg) PetscCall(PEPSetConvergenceTest(pep,PEP_CONV_ABS));
149: PetscCall(PetscOptionsBoolGroupEnd("-pep_conv_user","User-defined convergence test","PEPSetConvergenceTest",&flg));
150: if (flg) PetscCall(PEPSetConvergenceTest(pep,PEP_CONV_USER));
152: PetscCall(PetscOptionsBoolGroupBegin("-pep_stop_basic","Stop iteration if all eigenvalues converged or max_it reached","PEPSetStoppingTest",&flg));
153: if (flg) PetscCall(PEPSetStoppingTest(pep,PEP_STOP_BASIC));
154: PetscCall(PetscOptionsBoolGroupEnd("-pep_stop_user","User-defined stopping test","PEPSetStoppingTest",&flg));
155: if (flg) PetscCall(PEPSetStoppingTest(pep,PEP_STOP_USER));
157: i = pep->nev;
158: PetscCall(PetscOptionsInt("-pep_nev","Number of eigenvalues to compute","PEPSetDimensions",pep->nev,&i,&flg1));
159: j = pep->ncv;
160: PetscCall(PetscOptionsInt("-pep_ncv","Number of basis vectors","PEPSetDimensions",pep->ncv,&j,&flg2));
161: k = pep->mpd;
162: PetscCall(PetscOptionsInt("-pep_mpd","Maximum dimension of projected problem","PEPSetDimensions",pep->mpd,&k,&flg3));
163: if (flg1 || flg2 || flg3) PetscCall(PEPSetDimensions(pep,i,j,k));
165: PetscCall(PetscOptionsEnum("-pep_basis","Polynomial basis","PEPSetBasis",PEPBasisTypes,(PetscEnum)pep->basis,(PetscEnum*)&pep->basis,NULL));
167: PetscCall(PetscOptionsBoolGroupBegin("-pep_largest_magnitude","Compute largest eigenvalues in magnitude","PEPSetWhichEigenpairs",&flg));
168: if (flg) PetscCall(PEPSetWhichEigenpairs(pep,PEP_LARGEST_MAGNITUDE));
169: PetscCall(PetscOptionsBoolGroup("-pep_smallest_magnitude","Compute smallest eigenvalues in magnitude","PEPSetWhichEigenpairs",&flg));
170: if (flg) PetscCall(PEPSetWhichEigenpairs(pep,PEP_SMALLEST_MAGNITUDE));
171: PetscCall(PetscOptionsBoolGroup("-pep_largest_real","Compute eigenvalues with largest real parts","PEPSetWhichEigenpairs",&flg));
172: if (flg) PetscCall(PEPSetWhichEigenpairs(pep,PEP_LARGEST_REAL));
173: PetscCall(PetscOptionsBoolGroup("-pep_smallest_real","Compute eigenvalues with smallest real parts","PEPSetWhichEigenpairs",&flg));
174: if (flg) PetscCall(PEPSetWhichEigenpairs(pep,PEP_SMALLEST_REAL));
175: PetscCall(PetscOptionsBoolGroup("-pep_largest_imaginary","Compute eigenvalues with largest imaginary parts","PEPSetWhichEigenpairs",&flg));
176: if (flg) PetscCall(PEPSetWhichEigenpairs(pep,PEP_LARGEST_IMAGINARY));
177: PetscCall(PetscOptionsBoolGroup("-pep_smallest_imaginary","Compute eigenvalues with smallest imaginary parts","PEPSetWhichEigenpairs",&flg));
178: if (flg) PetscCall(PEPSetWhichEigenpairs(pep,PEP_SMALLEST_IMAGINARY));
179: PetscCall(PetscOptionsBoolGroup("-pep_target_magnitude","Compute eigenvalues closest to target","PEPSetWhichEigenpairs",&flg));
180: if (flg) PetscCall(PEPSetWhichEigenpairs(pep,PEP_TARGET_MAGNITUDE));
181: PetscCall(PetscOptionsBoolGroup("-pep_target_real","Compute eigenvalues with real parts closest to target","PEPSetWhichEigenpairs",&flg));
182: if (flg) PetscCall(PEPSetWhichEigenpairs(pep,PEP_TARGET_REAL));
183: PetscCall(PetscOptionsBoolGroup("-pep_target_imaginary","Compute eigenvalues with imaginary parts closest to target","PEPSetWhichEigenpairs",&flg));
184: if (flg) PetscCall(PEPSetWhichEigenpairs(pep,PEP_TARGET_IMAGINARY));
185: PetscCall(PetscOptionsBoolGroupEnd("-pep_all","Compute all eigenvalues in an interval or a region","PEPSetWhichEigenpairs",&flg));
186: if (flg) PetscCall(PEPSetWhichEigenpairs(pep,PEP_ALL));
188: PetscCall(PetscOptionsScalar("-pep_target","Value of the target","PEPSetTarget",pep->target,&s,&flg));
189: if (flg) {
190: if (pep->which!=PEP_TARGET_REAL && pep->which!=PEP_TARGET_IMAGINARY) PetscCall(PEPSetWhichEigenpairs(pep,PEP_TARGET_MAGNITUDE));
191: PetscCall(PEPSetTarget(pep,s));
192: }
194: k = 2;
195: PetscCall(PetscOptionsRealArray("-pep_interval","Computational interval (two real values separated with a comma without spaces)","PEPSetInterval",array,&k,&flg));
196: if (flg) {
197: PetscCheck(k>1,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_SIZ,"Must pass two values in -pep_interval (comma-separated without spaces)");
198: PetscCall(PEPSetWhichEigenpairs(pep,PEP_ALL));
199: PetscCall(PEPSetInterval(pep,array[0],array[1]));
200: }
202: /* -----------------------------------------------------------------------*/
203: /*
204: Cancels all monitors hardwired into code before call to PEPSetFromOptions()
205: */
206: PetscCall(PetscOptionsBool("-pep_monitor_cancel","Remove any hardwired monitor routines","PEPMonitorCancel",PETSC_FALSE,&flg,&set));
207: if (set && flg) PetscCall(PEPMonitorCancel(pep));
208: PetscCall(PEPMonitorSetFromOptions(pep,"-pep_monitor","first_approximation",NULL,PETSC_FALSE));
209: PetscCall(PEPMonitorSetFromOptions(pep,"-pep_monitor_all","all_approximations",NULL,PETSC_TRUE));
210: PetscCall(PEPMonitorSetFromOptions(pep,"-pep_monitor_conv","convergence_history",NULL,PETSC_FALSE));
212: /* -----------------------------------------------------------------------*/
213: PetscCall(PetscOptionsName("-pep_view","Print detailed information on solver used","PEPView",&set));
214: PetscCall(PetscOptionsName("-pep_view_vectors","View computed eigenvectors","PEPVectorsView",&set));
215: PetscCall(PetscOptionsName("-pep_view_values","View computed eigenvalues","PEPValuesView",&set));
216: PetscCall(PetscOptionsName("-pep_converged_reason","Print reason for convergence, and number of iterations","PEPConvergedReasonView",&set));
217: PetscCall(PetscOptionsName("-pep_error_absolute","Print absolute errors of each eigenpair","PEPErrorView",&set));
218: PetscCall(PetscOptionsName("-pep_error_relative","Print relative errors of each eigenpair","PEPErrorView",&set));
219: PetscCall(PetscOptionsName("-pep_error_backward","Print backward errors of each eigenpair","PEPErrorView",&set));
221: PetscTryTypeMethod(pep,setfromoptions,PetscOptionsObject);
222: PetscCall(PetscObjectProcessOptionsHandlers((PetscObject)pep,PetscOptionsObject));
223: PetscOptionsEnd();
225: if (!pep->V) PetscCall(PEPGetBV(pep,&pep->V));
226: PetscCall(BVSetFromOptions(pep->V));
227: if (!pep->rg) PetscCall(PEPGetRG(pep,&pep->rg));
228: PetscCall(RGSetFromOptions(pep->rg));
229: if (!pep->ds) PetscCall(PEPGetDS(pep,&pep->ds));
230: PetscCall(PEPSetDSType(pep));
231: PetscCall(DSSetFromOptions(pep->ds));
232: if (!pep->st) PetscCall(PEPGetST(pep,&pep->st));
233: PetscCall(PEPSetDefaultST(pep));
234: PetscCall(STSetFromOptions(pep->st));
235: if (!pep->refineksp) PetscCall(PEPRefineGetKSP(pep,&pep->refineksp));
236: PetscCall(KSPSetFromOptions(pep->refineksp));
237: PetscFunctionReturn(PETSC_SUCCESS);
238: }
240: /*@
241: PEPGetTolerances - Gets the tolerance and maximum iteration count used
242: by the `PEP` convergence tests.
244: Not Collective
246: Input Parameter:
247: . pep - the polynomial eigensolver context
249: Output Parameters:
250: + tol - the convergence tolerance
251: - maxits - maximum number of iterations
253: Note:
254: The user can specify `NULL` for any parameter that is not needed.
256: Level: intermediate
258: .seealso: [](ch:pep), `PEPSetTolerances()`
259: @*/
260: PetscErrorCode PEPGetTolerances(PEP pep,PetscReal *tol,PetscInt *maxits)
261: {
262: PetscFunctionBegin;
264: if (tol) *tol = pep->tol;
265: if (maxits) *maxits = pep->max_it;
266: PetscFunctionReturn(PETSC_SUCCESS);
267: }
269: /*@
270: PEPSetTolerances - Sets the tolerance and maximum iteration count used
271: by the `PEP` convergence tests.
273: Logically Collective
275: Input Parameters:
276: + pep - the polynomial eigensolver context
277: . tol - the convergence tolerance
278: - maxits - maximum number of iterations to use
280: Options Database Keys:
281: + -pep_tol \<tol\> - sets the convergence tolerance
282: - -pep_max_it \<maxits\> - sets the maximum number of iterations allowed
284: Note:
285: Use `PETSC_CURRENT` to retain the current value of any of the parameters.
286: Use `PETSC_DETERMINE` for either argument to assign a default value computed
287: internally (may be different in each solver).
288: For `maxits` use `PETSC_UNLIMITED` to indicate there is no upper bound on this value.
290: Level: intermediate
292: .seealso: [](ch:pep), `PEPGetTolerances()`
293: @*/
294: PetscErrorCode PEPSetTolerances(PEP pep,PetscReal tol,PetscInt maxits)
295: {
296: PetscFunctionBegin;
300: if (tol == (PetscReal)PETSC_DETERMINE) {
301: pep->tol = PETSC_DETERMINE;
302: pep->state = PEP_STATE_INITIAL;
303: } else if (tol != (PetscReal)PETSC_CURRENT) {
304: PetscCheck(tol>0.0,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Illegal value of tol. Must be > 0");
305: pep->tol = tol;
306: }
307: if (maxits == PETSC_DETERMINE) {
308: pep->max_it = PETSC_DETERMINE;
309: pep->state = PEP_STATE_INITIAL;
310: } else if (maxits == PETSC_UNLIMITED) {
311: pep->max_it = PETSC_INT_MAX;
312: } else if (maxits != PETSC_CURRENT) {
313: PetscCheck(maxits>0,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Illegal value of maxits. Must be > 0");
314: pep->max_it = maxits;
315: }
316: PetscFunctionReturn(PETSC_SUCCESS);
317: }
319: /*@
320: PEPGetDimensions - Gets the number of eigenvalues to compute
321: and the dimension of the subspace.
323: Not Collective
325: Input Parameter:
326: . pep - the polynomial eigensolver context
328: Output Parameters:
329: + nev - number of eigenvalues to compute
330: . ncv - the maximum dimension of the subspace to be used by the solver
331: - mpd - the maximum dimension allowed for the projected problem
333: Notes:
334: The user can specify `NULL` for any parameter that is not needed.
336: Level: intermediate
338: .seealso: [](ch:pep), `PEPSetDimensions()`
339: @*/
340: PetscErrorCode PEPGetDimensions(PEP pep,PetscInt *nev,PetscInt *ncv,PetscInt *mpd)
341: {
342: PetscFunctionBegin;
344: if (nev) *nev = pep->nev;
345: if (ncv) *ncv = pep->ncv;
346: if (mpd) *mpd = pep->mpd;
347: PetscFunctionReturn(PETSC_SUCCESS);
348: }
350: /*@
351: PEPSetDimensions - Sets the number of eigenvalues to compute
352: and the dimension of the subspace.
354: Logically Collective
356: Input Parameters:
357: + pep - the polynomial eigensolver context
358: . nev - number of eigenvalues to compute
359: . ncv - the maximum dimension of the subspace to be used by the solver
360: - mpd - the maximum dimension allowed for the projected problem
362: Options Database Keys:
363: + -pep_nev \<nev\> - sets the number of eigenvalues
364: . -pep_ncv \<ncv\> - sets the dimension of the subspace
365: - -pep_mpd \<mpd\> - sets the maximum projected dimension
367: Notes:
368: Use `PETSC_DETERMINE` for `ncv` and `mpd` to assign a reasonably good value, which is
369: dependent on the solution method. For any of the arguments, use `PETSC_CURRENT`
370: to preserve the current value.
372: The parameters `ncv` and `mpd` are intimately related, so that the user is advised
373: to set one of them at most. Normal usage is\:
375: 1. in cases where `nev` is small, the user sets `ncv` (a reasonable default is `2*nev`).
376: 2. in cases where `nev` is large, the user sets `mpd`.
378: The value of `ncv` should always be between `nev` and `(nev+mpd)`, typically
379: `ncv=nev+mpd`. If `nev` is not too large, `mpd=nev` is a reasonable choice, otherwise
380: a smaller value should be used.
382: When computing all eigenvalues in an interval, see `PEPSetInterval()`, these
383: parameters lose relevance, and tuning must be done with
384: `PEPSTOARSetDimensions()`.
386: Level: intermediate
388: .seealso: [](ch:pep), `PEPGetDimensions()`, `PEPSetInterval()`, `PEPSTOARSetDimensions()`
389: @*/
390: PetscErrorCode PEPSetDimensions(PEP pep,PetscInt nev,PetscInt ncv,PetscInt mpd)
391: {
392: PetscFunctionBegin;
397: if (nev != PETSC_CURRENT) {
398: PetscCheck(nev>0,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Illegal value of nev. Must be > 0");
399: pep->nev = nev;
400: }
401: if (ncv == PETSC_DETERMINE) {
402: pep->ncv = PETSC_DETERMINE;
403: } else if (ncv != PETSC_CURRENT) {
404: PetscCheck(ncv>0,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Illegal value of ncv. Must be > 0");
405: pep->ncv = ncv;
406: }
407: if (mpd == PETSC_DETERMINE) {
408: pep->mpd = PETSC_DETERMINE;
409: } else if (mpd != PETSC_CURRENT) {
410: PetscCheck(mpd>0,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Illegal value of mpd. Must be > 0");
411: pep->mpd = mpd;
412: }
413: pep->state = PEP_STATE_INITIAL;
414: PetscFunctionReturn(PETSC_SUCCESS);
415: }
417: /*@
418: PEPSetWhichEigenpairs - Specifies which portion of the spectrum is
419: to be sought.
421: Logically Collective
423: Input Parameters:
424: + pep - the polynomial eigensolver context
425: - which - the portion of the spectrum to be sought, see `PEPWhich` for possible values
427: Options Database Keys:
428: + -pep_largest_magnitude - sets largest eigenvalues in magnitude
429: . -pep_smallest_magnitude - sets smallest eigenvalues in magnitude
430: . -pep_largest_real - sets largest real parts
431: . -pep_smallest_real - sets smallest real parts
432: . -pep_largest_imaginary - sets largest imaginary parts
433: . -pep_smallest_imaginary - sets smallest imaginary parts
434: . -pep_target_magnitude - sets eigenvalues closest to target
435: . -pep_target_real - sets real parts closest to target
436: . -pep_target_imaginary - sets imaginary parts closest to target
437: - -pep_all - sets all eigenvalues in an interval or region
439: Notes:
440: Not all eigensolvers implemented in `PEP` account for all the possible values
441: of `which`. Also, some values make sense only for certain types of
442: problems. If SLEPc is compiled for real numbers `PEP_LARGEST_IMAGINARY`
443: and `PEP_SMALLEST_IMAGINARY` use the absolute value of the imaginary part
444: for eigenvalue selection.
446: The target is a scalar value provided with `PEPSetTarget()`.
448: The criterion `PEP_TARGET_IMAGINARY` is available only in case PETSc and
449: SLEPc have been built with complex scalars.
451: `PEP_ALL` is intended for use in combination with an interval (see
452: `PEPSetInterval()`), when all eigenvalues within the interval are requested,
453: and also for computing all eigenvalues in a region with the `PEPCISS` solver.
455: Level: intermediate
457: .seealso: [](ch:pep), `PEPGetWhichEigenpairs()`, `PEPSetTarget()`, `PEPSetInterval()`, `PEPSetDimensions()`, `PEPSetEigenvalueComparison()`, `PEPWhich`
458: @*/
459: PetscErrorCode PEPSetWhichEigenpairs(PEP pep,PEPWhich which)
460: {
461: PetscFunctionBegin;
464: switch (which) {
465: case PEP_LARGEST_MAGNITUDE:
466: case PEP_SMALLEST_MAGNITUDE:
467: case PEP_LARGEST_REAL:
468: case PEP_SMALLEST_REAL:
469: case PEP_LARGEST_IMAGINARY:
470: case PEP_SMALLEST_IMAGINARY:
471: case PEP_TARGET_MAGNITUDE:
472: case PEP_TARGET_REAL:
473: #if defined(PETSC_USE_COMPLEX)
474: case PEP_TARGET_IMAGINARY:
475: #endif
476: case PEP_ALL:
477: case PEP_WHICH_USER:
478: if (pep->which != which) {
479: pep->state = PEP_STATE_INITIAL;
480: pep->which = which;
481: }
482: break;
483: #if !defined(PETSC_USE_COMPLEX)
484: case PEP_TARGET_IMAGINARY:
485: SETERRQ(PetscObjectComm((PetscObject)pep),PETSC_ERR_SUP,"PEP_TARGET_IMAGINARY can be used only with complex scalars");
486: #endif
487: default:
488: SETERRQ(PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Invalid 'which' value");
489: }
490: PetscFunctionReturn(PETSC_SUCCESS);
491: }
493: /*@
494: PEPGetWhichEigenpairs - Returns which portion of the spectrum is to be
495: sought.
497: Not Collective
499: Input Parameter:
500: . pep - the polynomial eigensolver context
502: Output Parameter:
503: . which - the portion of the spectrum to be sought
505: Level: intermediate
507: .seealso: [](ch:pep), `PEPSetWhichEigenpairs()`, `PEPWhich`
508: @*/
509: PetscErrorCode PEPGetWhichEigenpairs(PEP pep,PEPWhich *which)
510: {
511: PetscFunctionBegin;
513: PetscAssertPointer(which,2);
514: *which = pep->which;
515: PetscFunctionReturn(PETSC_SUCCESS);
516: }
518: /*@C
519: PEPSetEigenvalueComparison - Specifies the eigenvalue comparison function
520: when `PEPSetWhichEigenpairs()` is set to `PEP_WHICH_USER`.
522: Logically Collective
524: Input Parameters:
525: + pep - the polynomial eigensolver context
526: . comp - the comparison function, see `SlepcEigenvalueComparisonFn` for the calling sequence
527: - ctx - a context pointer (the last parameter to the comparison function)
529: Level: advanced
531: .seealso: [](ch:pep), `PEPSetWhichEigenpairs()`, `PEPWhich`
532: @*/
533: PetscErrorCode PEPSetEigenvalueComparison(PEP pep,SlepcEigenvalueComparisonFn *comp,void *ctx)
534: {
535: PetscFunctionBegin;
537: pep->sc->comparison = comp;
538: pep->sc->comparisonctx = ctx;
539: pep->which = PEP_WHICH_USER;
540: PetscFunctionReturn(PETSC_SUCCESS);
541: }
543: /*@
544: PEPSetProblemType - Specifies the type of the polynomial eigenvalue problem.
546: Logically Collective
548: Input Parameters:
549: + pep - the polynomial eigensolver context
550: - type - a known type of polynomial eigenvalue problem
552: Options Database Keys:
553: + -pep_general - general problem with no particular structure
554: . -pep_hermitian - problem whose coefficient matrices are Hermitian
555: . -pep_hyperbolic - Hermitian problem that satisfies the definition of hyperbolic
556: - -pep_gyroscopic - problem with gyroscopic structure
558: Notes:
559: See `PEPProblemType` for possible problem types.
561: This function is used to instruct SLEPc to exploit certain structure in
562: the polynomial eigenproblem. By default, no particular structure is assumed.
564: If the problem matrices are Hermitian (symmetric in the real case) or
565: Hermitian/skew-Hermitian then the solver can exploit this fact to perform
566: less operations or provide better stability. Hyperbolic problems are a
567: particular case of Hermitian problems, some solvers may treat them simply as
568: Hermitian.
570: Level: intermediate
572: .seealso: [](ch:pep), `PEPSetOperators()`, `PEPSetType()`, `PEPGetProblemType()`, `PEPProblemType`
573: @*/
574: PetscErrorCode PEPSetProblemType(PEP pep,PEPProblemType type)
575: {
576: PetscFunctionBegin;
579: PetscCheck(type==PEP_GENERAL || type==PEP_HERMITIAN || type==PEP_HYPERBOLIC || type==PEP_GYROSCOPIC,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_WRONG,"Unknown eigenvalue problem type");
580: if (type != pep->problem_type) {
581: pep->problem_type = type;
582: pep->state = PEP_STATE_INITIAL;
583: }
584: PetscFunctionReturn(PETSC_SUCCESS);
585: }
587: /*@
588: PEPGetProblemType - Gets the problem type from the `PEP` object.
590: Not Collective
592: Input Parameter:
593: . pep - the polynomial eigensolver context
595: Output Parameter:
596: . type - the problem type
598: Level: intermediate
600: .seealso: [](ch:pep), `PEPSetProblemType()`, `PEPProblemType`
601: @*/
602: PetscErrorCode PEPGetProblemType(PEP pep,PEPProblemType *type)
603: {
604: PetscFunctionBegin;
606: PetscAssertPointer(type,2);
607: *type = pep->problem_type;
608: PetscFunctionReturn(PETSC_SUCCESS);
609: }
611: /*@
612: PEPSetBasis - Specifies the type of polynomial basis used to describe the
613: polynomial eigenvalue problem.
615: Logically Collective
617: Input Parameters:
618: + pep - the polynomial eigensolver context
619: - basis - the type of polynomial basis, see `PEPBasis` for possible values
621: Options Database Key:
622: . -pep_basis \<basis\> - Select the basis type
624: Note:
625: By default, the coefficient matrices passed via `PEPSetOperators()` are
626: expressed in the monomial basis, i.e.
627: $P(\lambda) = A_0 + \lambda A_1 + \lambda^2 A_2 + \dots + \lambda^d A_d$.
628: Other polynomial bases may have better numerical behavior, but the user
629: must then pass the coefficient matrices accordingly.
631: Level: intermediate
633: .seealso: [](ch:pep), `PEPSetOperators()`, `PEPGetBasis()`, `PEPBasis`
634: @*/
635: PetscErrorCode PEPSetBasis(PEP pep,PEPBasis basis)
636: {
637: PetscFunctionBegin;
640: pep->basis = basis;
641: PetscFunctionReturn(PETSC_SUCCESS);
642: }
644: /*@
645: PEPGetBasis - Gets the type of polynomial basis from the `PEP` object.
647: Not Collective
649: Input Parameter:
650: . pep - the polynomial eigensolver context
652: Output Parameter:
653: . basis - the polynomial basis
655: Level: intermediate
657: .seealso: [](ch:pep), `PEPSetBasis()`, `PEPBasis`
658: @*/
659: PetscErrorCode PEPGetBasis(PEP pep,PEPBasis *basis)
660: {
661: PetscFunctionBegin;
663: PetscAssertPointer(basis,2);
664: *basis = pep->basis;
665: PetscFunctionReturn(PETSC_SUCCESS);
666: }
668: /*@
669: PEPSetTrackAll - Specifies if the solver must compute the residual of all
670: approximate eigenpairs or not.
672: Logically Collective
674: Input Parameters:
675: + pep - the polynomial eigensolver context
676: - trackall - whether compute all residuals or not
678: Notes:
679: If the user sets `trackall`=`PETSC_TRUE` then the solver explicitly computes
680: the residual for each eigenpair approximation. Computing the residual is
681: usually an expensive operation and solvers commonly compute the associated
682: residual to the first unconverged eigenpair.
684: The option `-pep_monitor_all` automatically activates this option.
686: Level: developer
688: .seealso: [](ch:pep), `PEPGetTrackAll()`
689: @*/
690: PetscErrorCode PEPSetTrackAll(PEP pep,PetscBool trackall)
691: {
692: PetscFunctionBegin;
695: pep->trackall = trackall;
696: PetscFunctionReturn(PETSC_SUCCESS);
697: }
699: /*@
700: PEPGetTrackAll - Returns the flag indicating whether all residual norms must
701: be computed or not.
703: Not Collective
705: Input Parameter:
706: . pep - the polynomial eigensolver context
708: Output Parameter:
709: . trackall - the returned flag
711: Level: developer
713: .seealso: [](ch:pep), `PEPSetTrackAll()`
714: @*/
715: PetscErrorCode PEPGetTrackAll(PEP pep,PetscBool *trackall)
716: {
717: PetscFunctionBegin;
719: PetscAssertPointer(trackall,2);
720: *trackall = pep->trackall;
721: PetscFunctionReturn(PETSC_SUCCESS);
722: }
724: /*@C
725: PEPSetConvergenceTestFunction - Sets a function to compute the error estimate
726: used in the convergence test.
728: Logically Collective
730: Input Parameters:
731: + pep - the polynomial eigensolver context
732: . conv - convergence test function, see `PEPConvergenceTestFn` for the calling sequence
733: . ctx - context for private data for the convergence routine (may be `NULL`)
734: - destroy - a routine for destroying the context (may be `NULL`), see `PetscCtxDestroyFn`
735: for the calling sequence
737: Notes:
738: When this is called with a user-defined function, then the convergence
739: criterion is set to `PEP_CONV_USER`, see `PEPSetConvergenceTest()`.
741: If the error estimate returned by the convergence test function is less than
742: the tolerance, then the eigenvalue is accepted as converged.
744: Level: advanced
746: .seealso: [](ch:pep), `PEPSetConvergenceTest()`, `PEPSetTolerances()`
747: @*/
748: PetscErrorCode PEPSetConvergenceTestFunction(PEP pep,PEPConvergenceTestFn *conv,void *ctx,PetscCtxDestroyFn *destroy)
749: {
750: PetscFunctionBegin;
752: if (pep->convergeddestroy) PetscCall((*pep->convergeddestroy)(&pep->convergedctx));
753: pep->convergeduser = conv;
754: pep->convergeddestroy = destroy;
755: pep->convergedctx = ctx;
756: if (conv == PEPConvergedRelative) pep->conv = PEP_CONV_REL;
757: else if (conv == PEPConvergedNorm) pep->conv = PEP_CONV_NORM;
758: else if (conv == PEPConvergedAbsolute) pep->conv = PEP_CONV_ABS;
759: else {
760: pep->conv = PEP_CONV_USER;
761: pep->converged = pep->convergeduser;
762: }
763: PetscFunctionReturn(PETSC_SUCCESS);
764: }
766: /*@
767: PEPSetConvergenceTest - Specifies how to compute the error estimate
768: used in the convergence test.
770: Logically Collective
772: Input Parameters:
773: + pep - the polynomial eigensolver context
774: - conv - the type of convergence test, see `PEPConv` for possible values
776: Options Database Keys:
777: + -pep_conv_abs - sets the absolute convergence test
778: . -pep_conv_rel - sets the convergence test relative to the eigenvalue
779: . -pep_conv_norm - sets the convergence test relative to the matrix norms
780: - -pep_conv_user - selects the user-defined convergence test
782: Level: intermediate
784: .seealso: [](ch:pep), `PEPGetConvergenceTest()`, `PEPSetConvergenceTestFunction()`, `PEPSetStoppingTest()`, `PEPConv`
785: @*/
786: PetscErrorCode PEPSetConvergenceTest(PEP pep,PEPConv conv)
787: {
788: PetscFunctionBegin;
791: switch (conv) {
792: case PEP_CONV_ABS: pep->converged = PEPConvergedAbsolute; break;
793: case PEP_CONV_REL: pep->converged = PEPConvergedRelative; break;
794: case PEP_CONV_NORM: pep->converged = PEPConvergedNorm; break;
795: case PEP_CONV_USER:
796: PetscCheck(pep->convergeduser,PetscObjectComm((PetscObject)pep),PETSC_ERR_ORDER,"Must call PEPSetConvergenceTestFunction() first");
797: pep->converged = pep->convergeduser;
798: break;
799: default:
800: SETERRQ(PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Invalid 'conv' value");
801: }
802: pep->conv = conv;
803: PetscFunctionReturn(PETSC_SUCCESS);
804: }
806: /*@
807: PEPGetConvergenceTest - Gets the method used to compute the error estimate
808: used in the convergence test.
810: Not Collective
812: Input Parameter:
813: . pep - the polynomial eigensolver context
815: Output Parameter:
816: . conv - the type of convergence test
818: Level: intermediate
820: .seealso: [](ch:pep), `PEPSetConvergenceTest()`, `PEPConv`
821: @*/
822: PetscErrorCode PEPGetConvergenceTest(PEP pep,PEPConv *conv)
823: {
824: PetscFunctionBegin;
826: PetscAssertPointer(conv,2);
827: *conv = pep->conv;
828: PetscFunctionReturn(PETSC_SUCCESS);
829: }
831: /*@C
832: PEPSetStoppingTestFunction - Sets a function to decide when to stop the outer
833: iteration of the eigensolver.
835: Logically Collective
837: Input Parameters:
838: + pep - the polynomial eigensolver context
839: . stop - stopping test function, see `PEPStoppingTestFn` for the calling sequence
840: . ctx - context for private data for the stopping routine (may be `NULL`)
841: - destroy - a routine for destroying the context (may be `NULL`), see `PetscCtxDestroyFn`
842: for the calling sequence
844: Note:
845: When implementing a function for this, normal usage is to first call the
846: default routine `PEPStoppingBasic()` and then set `reason` to `PEP_CONVERGED_USER`
847: if some user-defined conditions have been met. To let the eigensolver continue
848: iterating, the result must be left as `PEP_CONVERGED_ITERATING`.
850: Level: advanced
852: .seealso: [](ch:pep), `PEPSetStoppingTest()`, `PEPStoppingBasic()`
853: @*/
854: PetscErrorCode PEPSetStoppingTestFunction(PEP pep,PEPStoppingTestFn *stop,void *ctx,PetscCtxDestroyFn *destroy)
855: {
856: PetscFunctionBegin;
858: if (pep->stoppingdestroy) PetscCall((*pep->stoppingdestroy)(&pep->stoppingctx));
859: pep->stoppinguser = stop;
860: pep->stoppingdestroy = destroy;
861: pep->stoppingctx = ctx;
862: if (stop == PEPStoppingBasic) pep->stop = PEP_STOP_BASIC;
863: else {
864: pep->stop = PEP_STOP_USER;
865: pep->stopping = pep->stoppinguser;
866: }
867: PetscFunctionReturn(PETSC_SUCCESS);
868: }
870: /*@
871: PEPSetStoppingTest - Specifies how to decide the termination of the outer
872: loop of the eigensolver.
874: Logically Collective
876: Input Parameters:
877: + pep - the polynomial eigensolver context
878: - stop - the type of stopping test, see `PEPStop`
880: Options Database Keys:
881: + -pep_stop_basic - sets the default stopping test
882: - -pep_stop_user - selects the user-defined stopping test
884: Level: advanced
886: .seealso: [](ch:pep), `PEPGetStoppingTest()`, `PEPSetStoppingTestFunction()`, `PEPSetConvergenceTest()`, `PEPStop`
887: @*/
888: PetscErrorCode PEPSetStoppingTest(PEP pep,PEPStop stop)
889: {
890: PetscFunctionBegin;
893: switch (stop) {
894: case PEP_STOP_BASIC: pep->stopping = PEPStoppingBasic; break;
895: case PEP_STOP_USER:
896: PetscCheck(pep->stoppinguser,PetscObjectComm((PetscObject)pep),PETSC_ERR_ORDER,"Must call PEPSetStoppingTestFunction() first");
897: pep->stopping = pep->stoppinguser;
898: break;
899: default:
900: SETERRQ(PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Invalid 'stop' value");
901: }
902: pep->stop = stop;
903: PetscFunctionReturn(PETSC_SUCCESS);
904: }
906: /*@
907: PEPGetStoppingTest - Gets the method used to decide the termination of the outer
908: loop of the eigensolver.
910: Not Collective
912: Input Parameter:
913: . pep - the polynomial eigensolver context
915: Output Parameter:
916: . stop - the type of stopping test
918: Level: advanced
920: .seealso: [](ch:pep), `PEPSetStoppingTest()`, `PEPStop`
921: @*/
922: PetscErrorCode PEPGetStoppingTest(PEP pep,PEPStop *stop)
923: {
924: PetscFunctionBegin;
926: PetscAssertPointer(stop,2);
927: *stop = pep->stop;
928: PetscFunctionReturn(PETSC_SUCCESS);
929: }
931: /*@
932: PEPSetScale - Specifies the scaling strategy to be used.
934: Collective
936: Input Parameters:
937: + pep - the polynomial eigensolver context
938: . scale - scaling strategy, see `PEPScale` for possible values
939: . alpha - the scaling factor used in the scalar strategy
940: . Dl - the left diagonal matrix of the diagonal scaling algorithm
941: . Dr - the right diagonal matrix of the diagonal scaling algorithm
942: . its - number of iterations of the diagonal scaling algorithm
943: - lambda - approximation to wanted eigenvalues (modulus)
945: Options Database Keys:
946: + -pep_scale \<scale\> - set the scaling type, one of `none`,`scalar`,`diagonal`,`both`
947: . -pep_scale_factor \<alpha\> - set the scaling factor
948: . -pep_scale_its \<its\> - set the number of iterations
949: - -pep_scale_lambda \<lambda\> - set the approximation to eigenvalues
951: Notes:
952: There are two non-exclusive scaling strategies, scalar and diagonal.
953: See discussion in section [](#sec:scaling).
955: In the scalar strategy, scaling is applied to the eigenvalue, that is,
956: $\mu = \lambda/\alpha$ is the new eigenvalue and all matrices are scaled
957: accordingly. After solving the scaled problem, the original $\lambda$ is
958: recovered. Parameter `alpha` must be positive. Use `PETSC_DETERMINE` to let
959: the solver compute a reasonable scaling factor, and `PETSC_CURRENT` to
960: retain a previously set value.
962: In the diagonal strategy, the solver works implicitly with matrix $D_\ell P(\lambda)D_r$,
963: where $D_\ell$ and $D_r$ are appropriate diagonal matrices. This improves the accuracy
964: of the computed results in some cases. The user may provide the `Dl` and `Dr`
965: matrices represented as `Vec` objects storing diagonal elements. If not
966: provided, these matrices are computed internally. This option requires
967: that the polynomial coefficient matrices are of `MATAIJ` type.
968: The parameter `its` is the number of iterations performed by the method.
969: Parameter `lambda` must be positive. Use `PETSC_DETERMINE` or set `lambda` = 1.0
970: if no information about eigenvalues is available. `PETSC_CURRENT` can also
971: be used to leave `its` and `lambda` unchanged.
973: Level: intermediate
975: .seealso: [](ch:pep), [](#sec:scaling), `PEPGetScale()`
976: @*/
977: PetscErrorCode PEPSetScale(PEP pep,PEPScale scale,PetscReal alpha,Vec Dl,Vec Dr,PetscInt its,PetscReal lambda)
978: {
979: PetscFunctionBegin;
982: pep->scale = scale;
983: if (scale==PEP_SCALE_SCALAR || scale==PEP_SCALE_BOTH) {
985: if (alpha == (PetscReal)PETSC_DETERMINE) {
986: pep->sfactor = 0.0;
987: pep->sfactor_set = PETSC_FALSE;
988: } else if (alpha != (PetscReal)PETSC_CURRENT) {
989: PetscCheck(alpha>0.0,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Illegal value of alpha. Must be > 0");
990: pep->sfactor = alpha;
991: pep->sfactor_set = PETSC_TRUE;
992: }
993: }
994: if (scale==PEP_SCALE_DIAGONAL || scale==PEP_SCALE_BOTH) {
995: if (Dl) {
997: PetscCheckSameComm(pep,1,Dl,4);
998: PetscCall(PetscObjectReference((PetscObject)Dl));
999: PetscCall(VecDestroy(&pep->Dl));
1000: pep->Dl = Dl;
1001: }
1002: if (Dr) {
1004: PetscCheckSameComm(pep,1,Dr,5);
1005: PetscCall(PetscObjectReference((PetscObject)Dr));
1006: PetscCall(VecDestroy(&pep->Dr));
1007: pep->Dr = Dr;
1008: }
1011: if (its==PETSC_DETERMINE) pep->sits = 5;
1012: else if (its!=PETSC_CURRENT) {
1013: PetscCheck(its>0,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Illegal value of its. Must be > 0");
1014: pep->sits = its;
1015: }
1016: if (lambda == (PetscReal)PETSC_DETERMINE) pep->slambda = 1.0;
1017: else if (lambda != (PetscReal)PETSC_CURRENT) {
1018: PetscCheck(lambda>0.0,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Illegal value of lambda. Must be > 0");
1019: pep->slambda = lambda;
1020: }
1021: }
1022: pep->state = PEP_STATE_INITIAL;
1023: PetscFunctionReturn(PETSC_SUCCESS);
1024: }
1026: /*@
1027: PEPGetScale - Gets the scaling strategy used by the `PEP` object, and the
1028: associated parameters.
1030: Not Collective
1032: Input Parameter:
1033: . pep - the polynomial eigensolver context
1035: Output Parameters:
1036: + scale - scaling strategy
1037: . alpha - the scaling factor used in the scalar strategy
1038: . Dl - the left diagonal matrix of the diagonal scaling algorithm
1039: . Dr - the right diagonal matrix of the diagonal scaling algorithm
1040: . its - number of iterations of the diagonal scaling algorithm
1041: - lambda - approximation to wanted eigenvalues (modulus)
1043: Notes:
1044: The user can specify `NULL` for any parameter that is not needed.
1046: If `Dl` or `Dr` were not set by the user, then the ones computed internally are
1047: returned (or a `NULL` pointer if called before `PEPSetUp()`).
1049: Level: intermediate
1051: .seealso: [](ch:pep), `PEPSetScale()`, `PEPSetUp()`
1052: @*/
1053: PetscErrorCode PEPGetScale(PEP pep,PEPScale *scale,PetscReal *alpha,Vec *Dl,Vec *Dr,PetscInt *its,PetscReal *lambda)
1054: {
1055: PetscFunctionBegin;
1057: if (scale) *scale = pep->scale;
1058: if (alpha) *alpha = pep->sfactor;
1059: if (Dl) *Dl = pep->Dl;
1060: if (Dr) *Dr = pep->Dr;
1061: if (its) *its = pep->sits;
1062: if (lambda) *lambda = pep->slambda;
1063: PetscFunctionReturn(PETSC_SUCCESS);
1064: }
1066: /*@
1067: PEPSetExtract - Specifies the extraction strategy to be used.
1069: Logically Collective
1071: Input Parameters:
1072: + pep - the polynomial eigensolver context
1073: - extract - extraction strategy, see `PEPExtract` for possible values
1075: Options Database Key:
1076: . -pep_extract \<extract\> - extraction type, one of `none`,`norm`,`residual`,`structured`
1078: Note:
1079: This is relevant for solvers based on linearization. Once the solver has
1080: converged, the polynomial eigenvectors can be extracted from the
1081: eigenvectors of the linearized problem in different ways. See the
1082: discussion in section [](#sec:pepextr).
1084: Level: intermediate
1086: .seealso: [](ch:pep), [](#sec:pepextr), `PEPExtract`, `PEPGetExtract()`
1087: @*/
1088: PetscErrorCode PEPSetExtract(PEP pep,PEPExtract extract)
1089: {
1090: PetscFunctionBegin;
1093: pep->extract = extract;
1094: PetscFunctionReturn(PETSC_SUCCESS);
1095: }
1097: /*@
1098: PEPGetExtract - Gets the extraction strategy used by the `PEP` object.
1100: Not Collective
1102: Input Parameter:
1103: . pep - the polynomial eigensolver context
1105: Output Parameter:
1106: . extract - extraction strategy
1108: Level: intermediate
1110: .seealso: [](ch:pep), `PEPSetExtract()`, `PEPExtract`
1111: @*/
1112: PetscErrorCode PEPGetExtract(PEP pep,PEPExtract *extract)
1113: {
1114: PetscFunctionBegin;
1116: PetscAssertPointer(extract,2);
1117: *extract = pep->extract;
1118: PetscFunctionReturn(PETSC_SUCCESS);
1119: }
1121: /*@
1122: PEPSetRefine - Specifies the refinement type (and options) to be used
1123: after the solve.
1125: Logically Collective
1127: Input Parameters:
1128: + pep - the polynomial eigensolver context
1129: . refine - refinement type, see `PEPRefine` for possible values
1130: . npart - number of partitions of the communicator
1131: . tol - the convergence tolerance
1132: . its - maximum number of refinement iterations
1133: - scheme - the scheme for solving the involved linear systems, see `PEPRefineScheme`
1134: for possible values
1136: Options Database Keys:
1137: + -pep_refine \<refine\> - set the refinement type, one of `none`,`simple`,`multiple`
1138: . -pep_refine_partitions \<npart\> - set the number of partitions
1139: . -pep_refine_tol \<tol\> - set the tolerance
1140: . -pep_refine_its \<its\> - set the number of iterations
1141: - -pep_refine_scheme \<scheme\> - set the scheme for the linear solves, `schur`,`mbe`, or `explicit`
1143: Notes:
1144: This function configures the parameters of Newton iterative refinement,
1145: see section [](#sec:refine) for a discussion of the different strategies.
1147: By default, iterative refinement is disabled, since it may be very
1148: costly. There are two possible refinement strategies, simple and multiple.
1149: The simple approach performs iterative refinement on each of the
1150: converged eigenpairs individually, whereas the multiple strategy works
1151: with the invariant pair as a whole, refining all eigenpairs simultaneously.
1152: The latter may be required for the case of multiple eigenvalues.
1154: In some cases, especially when using direct solvers within the
1155: iterative refinement method, it may be helpful for improved scalability
1156: to split the communicator in several partitions. The `npart` parameter
1157: indicates how many partitions to use (defaults to 1).
1159: The `tol` and `its` parameters specify the stopping criterion. In the simple
1160: method, refinement continues until the residual of each eigenpair is
1161: below the tolerance (`tol` defaults to the `PEP` tolerance, but may be set to a
1162: different value). In contrast, the multiple method simply performs its
1163: refinement iterations (just one by default).
1165: The `scheme` argument is used to change the way in which linear systems are
1166: solved. Possible choices are explicit, mixed block elimination (MBE),
1167: and Schur complement.
1169: Use `PETSC_CURRENT` to retain the current value of `npart`, `tol` or `its`. Use
1170: `PETSC_DETERMINE` to assign a default value.
1172: Level: intermediate
1174: .seealso: [](ch:pep), [](#sec:refine), `PEPGetRefine()`
1175: @*/
1176: PetscErrorCode PEPSetRefine(PEP pep,PEPRefine refine,PetscInt npart,PetscReal tol,PetscInt its,PEPRefineScheme scheme)
1177: {
1178: PetscMPIInt size;
1180: PetscFunctionBegin;
1187: pep->refine = refine;
1188: if (refine) { /* process parameters only if not REFINE_NONE */
1189: if (npart!=pep->npart) {
1190: PetscCall(PetscSubcommDestroy(&pep->refinesubc));
1191: PetscCall(KSPDestroy(&pep->refineksp));
1192: }
1193: if (npart == PETSC_DETERMINE) {
1194: pep->npart = 1;
1195: } else if (npart != PETSC_CURRENT) {
1196: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)pep),&size));
1197: PetscCheck(npart>0 && npart<=size,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Illegal value of npart");
1198: pep->npart = npart;
1199: }
1200: if (tol == (PetscReal)PETSC_DETERMINE) {
1201: pep->rtol = PETSC_DETERMINE;
1202: } else if (tol != (PetscReal)PETSC_CURRENT) {
1203: PetscCheck(tol>0.0,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Illegal value of tol. Must be > 0");
1204: pep->rtol = tol;
1205: }
1206: if (its==PETSC_DETERMINE) {
1207: pep->rits = PETSC_DETERMINE;
1208: } else if (its != PETSC_CURRENT) {
1209: PetscCheck(its>=0,PetscObjectComm((PetscObject)pep),PETSC_ERR_ARG_OUTOFRANGE,"Illegal value of its. Must be >= 0");
1210: pep->rits = its;
1211: }
1212: pep->scheme = scheme;
1213: }
1214: pep->state = PEP_STATE_INITIAL;
1215: PetscFunctionReturn(PETSC_SUCCESS);
1216: }
1218: /*@
1219: PEPGetRefine - Gets the refinement strategy used by the `PEP` object, and the
1220: associated parameters.
1222: Not Collective
1224: Input Parameter:
1225: . pep - the polynomial eigensolver context
1227: Output Parameters:
1228: + refine - refinement type
1229: . npart - number of partitions of the communicator
1230: . tol - the convergence tolerance
1231: . its - maximum number of refinement iterations
1232: - scheme - the scheme used for solving linear systems
1234: Note:
1235: The user can specify `NULL` for any parameter that is not needed.
1237: Level: intermediate
1239: .seealso: [](ch:pep), `PEPSetRefine()`
1240: @*/
1241: PetscErrorCode PEPGetRefine(PEP pep,PEPRefine *refine,PetscInt *npart,PetscReal *tol,PetscInt *its,PEPRefineScheme *scheme)
1242: {
1243: PetscFunctionBegin;
1245: if (refine) *refine = pep->refine;
1246: if (npart) *npart = pep->npart;
1247: if (tol) *tol = pep->rtol;
1248: if (its) *its = pep->rits;
1249: if (scheme) *scheme = pep->scheme;
1250: PetscFunctionReturn(PETSC_SUCCESS);
1251: }
1253: /*@
1254: PEPSetOptionsPrefix - Sets the prefix used for searching for all
1255: `PEP` options in the database.
1257: Logically Collective
1259: Input Parameters:
1260: + pep - the polynomial eigensolver context
1261: - prefix - the prefix string to prepend to all `PEP` option requests
1263: Notes:
1264: A hyphen (-) must NOT be given at the beginning of the prefix name.
1265: The first character of all runtime options is AUTOMATICALLY the
1266: hyphen.
1268: For example, to distinguish between the runtime options for two
1269: different `PEP` contexts, one could call
1270: .vb
1271: PEPSetOptionsPrefix(pep1,"qeig1_")
1272: PEPSetOptionsPrefix(pep2,"qeig2_")
1273: .ve
1275: Level: advanced
1277: .seealso: [](ch:pep), `PEPAppendOptionsPrefix()`, `PEPGetOptionsPrefix()`
1278: @*/
1279: PetscErrorCode PEPSetOptionsPrefix(PEP pep,const char prefix[])
1280: {
1281: PetscFunctionBegin;
1283: if (!pep->st) PetscCall(PEPGetST(pep,&pep->st));
1284: PetscCall(STSetOptionsPrefix(pep->st,prefix));
1285: if (!pep->V) PetscCall(PEPGetBV(pep,&pep->V));
1286: PetscCall(BVSetOptionsPrefix(pep->V,prefix));
1287: if (!pep->ds) PetscCall(PEPGetDS(pep,&pep->ds));
1288: PetscCall(DSSetOptionsPrefix(pep->ds,prefix));
1289: if (!pep->rg) PetscCall(PEPGetRG(pep,&pep->rg));
1290: PetscCall(RGSetOptionsPrefix(pep->rg,prefix));
1291: PetscCall(PetscObjectSetOptionsPrefix((PetscObject)pep,prefix));
1292: PetscFunctionReturn(PETSC_SUCCESS);
1293: }
1295: /*@
1296: PEPAppendOptionsPrefix - Appends to the prefix used for searching for all
1297: `PEP` options in the database.
1299: Logically Collective
1301: Input Parameters:
1302: + pep - the polynomial eigensolver context
1303: - prefix - the prefix string to prepend to all `PEP` option requests
1305: Notes:
1306: A hyphen (-) must NOT be given at the beginning of the prefix name.
1307: The first character of all runtime options is AUTOMATICALLY the hyphen.
1309: Level: advanced
1311: .seealso: [](ch:pep), `PEPSetOptionsPrefix()`, `PEPGetOptionsPrefix()`
1312: @*/
1313: PetscErrorCode PEPAppendOptionsPrefix(PEP pep,const char prefix[])
1314: {
1315: PetscFunctionBegin;
1317: if (!pep->st) PetscCall(PEPGetST(pep,&pep->st));
1318: PetscCall(STAppendOptionsPrefix(pep->st,prefix));
1319: if (!pep->V) PetscCall(PEPGetBV(pep,&pep->V));
1320: PetscCall(BVAppendOptionsPrefix(pep->V,prefix));
1321: if (!pep->ds) PetscCall(PEPGetDS(pep,&pep->ds));
1322: PetscCall(DSAppendOptionsPrefix(pep->ds,prefix));
1323: if (!pep->rg) PetscCall(PEPGetRG(pep,&pep->rg));
1324: PetscCall(RGAppendOptionsPrefix(pep->rg,prefix));
1325: PetscCall(PetscObjectAppendOptionsPrefix((PetscObject)pep,prefix));
1326: PetscFunctionReturn(PETSC_SUCCESS);
1327: }
1329: /*@
1330: PEPGetOptionsPrefix - Gets the prefix used for searching for all
1331: `PEP` options in the database.
1333: Not Collective
1335: Input Parameter:
1336: . pep - the polynomial eigensolver context
1338: Output Parameter:
1339: . prefix - pointer to the prefix string used is returned
1341: Level: advanced
1343: .seealso: [](ch:pep), `PEPSetOptionsPrefix()`, `PEPAppendOptionsPrefix()`
1344: @*/
1345: PetscErrorCode PEPGetOptionsPrefix(PEP pep,const char *prefix[])
1346: {
1347: PetscFunctionBegin;
1349: PetscAssertPointer(prefix,2);
1350: PetscCall(PetscObjectGetOptionsPrefix((PetscObject)pep,prefix));
1351: PetscFunctionReturn(PETSC_SUCCESS);
1352: }