Actual source code: pciss.c

slepc-3.18.2 2023-01-26
Report Typos and Errors
  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:    SLEPc eigensolver: "ciss"

 13:    Method: Contour Integral Spectral Slicing

 15:    Algorithm:

 17:        Contour integral based on Sakurai-Sugiura method to construct a
 18:        subspace, with various eigenpair extractions (Rayleigh-Ritz,
 19:        explicit moment).

 21:    Based on code contributed by Y. Maeda, T. Sakurai.

 23:    References:

 25:        [1] J. Asakura, T. Sakurai, H. Tadano, T. Ikegami, K. Kimura, "A
 26:            numerical method for polynomial eigenvalue problems using contour
 27:            integral", Japan J. Indust. Appl. Math. 27:73-90, 2010.
 28: */

 30: #include <slepc/private/pepimpl.h>
 31: #include <slepc/private/slepccontour.h>

 33: typedef struct {
 34:   /* parameters */
 35:   PetscInt          N;             /* number of integration points (32) */
 36:   PetscInt          L;             /* block size (16) */
 37:   PetscInt          M;             /* moment degree (N/4 = 4) */
 38:   PetscReal         delta;         /* threshold of singular value (1e-12) */
 39:   PetscInt          L_max;         /* maximum number of columns of the source matrix V */
 40:   PetscReal         spurious_threshold; /* discard spurious eigenpairs */
 41:   PetscBool         isreal;        /* T(z) is real for real z */
 42:   PetscInt          npart;         /* number of partitions */
 43:   PetscInt          refine_inner;
 44:   PetscInt          refine_blocksize;
 45:   PEPCISSExtraction extraction;
 46:   /* private data */
 47:   SlepcContourData  contour;
 48:   PetscReal         *sigma;        /* threshold for numerical rank */
 49:   PetscScalar       *weight;
 50:   PetscScalar       *omega;
 51:   PetscScalar       *pp;
 52:   BV                V;
 53:   BV                S;
 54:   BV                Y;
 55:   PetscBool         useconj;
 56:   Mat               J,*Psplit;     /* auxiliary matrices */
 57:   BV                pV;
 58:   PetscObjectId     rgid;
 59:   PetscObjectState  rgstate;
 60: } PEP_CISS;

 62: static PetscErrorCode PEPComputeFunction(PEP pep,PetscScalar lambda,Mat T,Mat P,PetscBool deriv)
 63: {
 64:   PetscInt         i;
 65:   PetscScalar      *coeff;
 66:   Mat              *A,*K;
 67:   MatStructure     str,strp;
 68:   PEP_CISS         *ctx = (PEP_CISS*)pep->data;
 69:   SlepcContourData contour = ctx->contour;

 71:   A = (contour->pA)?contour->pA:pep->A;
 72:   K = (contour->pP)?contour->pP:ctx->Psplit;
 73:   PetscMalloc1(pep->nmat,&coeff);
 74:   if (deriv) PEPEvaluateBasisDerivative(pep,lambda,0,coeff,NULL);
 75:   else PEPEvaluateBasis(pep,lambda,0,coeff,NULL);
 76:   STGetMatStructure(pep->st,&str);
 77:   MatZeroEntries(T);
 78:   if (!deriv && T != P) {
 79:     STGetSplitPreconditionerInfo(pep->st,NULL,&strp);
 80:     MatZeroEntries(P);
 81:   }
 82:   i = deriv?1:0;
 83:   for (;i<pep->nmat;i++) {
 84:     MatAXPY(T,coeff[i],A[i],str);
 85:     if (!deriv && T != P) MatAXPY(P,coeff[i],K[i],strp);
 86:   }
 87:   PetscFree(coeff);
 88:   return 0;
 89: }

 91: /*
 92:   Set up KSP solvers for every integration point
 93: */
 94: static PetscErrorCode PEPCISSSetUp(PEP pep,Mat T,Mat P)
 95: {
 96:   PEP_CISS         *ctx = (PEP_CISS*)pep->data;
 97:   SlepcContourData contour;
 98:   PetscInt         i,p_id;
 99:   Mat              Amat,Pmat;

101:   if (!ctx->contour || !ctx->contour->ksp) PEPCISSGetKSPs(pep,NULL,NULL);
102:   contour = ctx->contour;
103:   PetscAssert(ctx->contour && ctx->contour->ksp,PetscObjectComm((PetscObject)pep),PETSC_ERR_PLIB,"Something went wrong with PEPCISSGetKSPs()");
104:   for (i=0;i<contour->npoints;i++) {
105:     p_id = i*contour->subcomm->n + contour->subcomm->color;
106:     MatDuplicate(T,MAT_DO_NOT_COPY_VALUES,&Amat);
107:     if (T != P) MatDuplicate(P,MAT_DO_NOT_COPY_VALUES,&Pmat); else Pmat = Amat;
108:     PEPComputeFunction(pep,ctx->omega[p_id],Amat,Pmat,PETSC_FALSE);
109:     PEP_KSPSetOperators(contour->ksp[i],Amat,Pmat);
110:     MatDestroy(&Amat);
111:     if (T != P) MatDestroy(&Pmat);
112:   }
113:   return 0;
114: }

116: /*
117:   Y_i = F(z_i)^{-1}Fp(z_i)V for every integration point, Y=[Y_i] is in the context
118: */
119: static PetscErrorCode PEPCISSSolve(PEP pep,Mat dT,BV V,PetscInt L_start,PetscInt L_end)
120: {
121:   PEP_CISS         *ctx = (PEP_CISS*)pep->data;
122:   SlepcContourData contour;
123:   PetscInt         i,p_id;
124:   Mat              MV,BMV=NULL,MC;

126:   contour = ctx->contour;
127:   BVSetActiveColumns(V,L_start,L_end);
128:   BVGetMat(V,&MV);
129:   for (i=0;i<contour->npoints;i++) {
130:     p_id = i*contour->subcomm->n + contour->subcomm->color;
131:     PEPComputeFunction(pep,ctx->omega[p_id],dT,NULL,PETSC_TRUE);
132:     BVSetActiveColumns(ctx->Y,i*ctx->L+L_start,i*ctx->L+L_end);
133:     BVGetMat(ctx->Y,&MC);
134:     if (!i) {
135:       MatProductCreate(dT,MV,NULL,&BMV);
136:       MatProductSetType(BMV,MATPRODUCT_AB);
137:       MatProductSetFromOptions(BMV);
138:       MatProductSymbolic(BMV);
139:     }
140:     MatProductNumeric(BMV);
141:     KSPMatSolve(contour->ksp[i],BMV,MC);
142:     BVRestoreMat(ctx->Y,&MC);
143:   }
144:   MatDestroy(&BMV);
145:   BVRestoreMat(V,&MV);
146:   return 0;
147: }

149: PetscErrorCode PEPSetUp_CISS(PEP pep)
150: {
151:   PEP_CISS         *ctx = (PEP_CISS*)pep->data;
152:   SlepcContourData contour;
153:   PetscInt         i,nwork,nsplit;
154:   PetscBool        istrivial,isellipse,flg;
155:   PetscObjectId    id;
156:   PetscObjectState state;
157:   Vec              v0;

159:   if (pep->ncv==PETSC_DEFAULT) pep->ncv = ctx->L_max*ctx->M;
160:   else {
161:     ctx->L_max = pep->ncv/ctx->M;
162:     if (!ctx->L_max) {
163:       ctx->L_max = 1;
164:       pep->ncv = ctx->L_max*ctx->M;
165:     }
166:   }
167:   ctx->L = PetscMin(ctx->L,ctx->L_max);
168:   if (pep->max_it==PETSC_DEFAULT) pep->max_it = 5;
169:   if (pep->mpd==PETSC_DEFAULT) pep->mpd = pep->ncv;
170:   if (!pep->which) pep->which = PEP_ALL;
172:   PEPCheckUnsupported(pep,PEP_FEATURE_STOPPING);
173:   PEPCheckIgnored(pep,PEP_FEATURE_SCALE);

175:   /* check region */
176:   RGIsTrivial(pep->rg,&istrivial);
178:   RGGetComplement(pep->rg,&flg);
180:   PetscObjectTypeCompare((PetscObject)pep->rg,RGELLIPSE,&isellipse);

183:   /* if the region has changed, then reset contour data */
184:   PetscObjectGetId((PetscObject)pep->rg,&id);
185:   PetscObjectStateGet((PetscObject)pep->rg,&state);
186:   if (ctx->rgid && (id != ctx->rgid || state != ctx->rgstate)) {
187:     SlepcContourDataDestroy(&ctx->contour);
188:     PetscInfo(pep,"Resetting the contour data structure due to a change of region\n");
189:     ctx->rgid = id; ctx->rgstate = state;
190:   }

192:   /* create contour data structure */
193:   if (!ctx->contour) {
194:     RGCanUseConjugates(pep->rg,ctx->isreal,&ctx->useconj);
195:     SlepcContourDataCreate(ctx->useconj?ctx->N/2:ctx->N,ctx->npart,(PetscObject)pep,&ctx->contour);
196:   }

198:   PEPAllocateSolution(pep,0);
199:   if (ctx->weight) PetscFree4(ctx->weight,ctx->omega,ctx->pp,ctx->sigma);
200:   PetscMalloc4(ctx->N,&ctx->weight,ctx->N,&ctx->omega,ctx->N,&ctx->pp,ctx->L_max*ctx->M,&ctx->sigma);

202:   /* allocate basis vectors */
203:   BVDestroy(&ctx->S);
204:   BVDuplicateResize(pep->V,ctx->L*ctx->M,&ctx->S);
205:   BVDestroy(&ctx->V);
206:   BVDuplicateResize(pep->V,ctx->L,&ctx->V);

208:   /* check if a user-defined split preconditioner has been set */
209:   STGetSplitPreconditionerInfo(pep->st,&nsplit,NULL);
210:   if (nsplit) {
211:     PetscFree(ctx->Psplit);
212:     PetscMalloc1(nsplit,&ctx->Psplit);
213:     for (i=0;i<nsplit;i++) STGetSplitPreconditionerTerm(pep->st,i,&ctx->Psplit[i]);
214:   }

216:   contour = ctx->contour;
217:   SlepcContourRedundantMat(contour,pep->nmat,pep->A,ctx->Psplit);
218:   if (!ctx->J) MatDuplicate(contour->pA?contour->pA[0]:pep->A[0],MAT_DO_NOT_COPY_VALUES,&ctx->J);
219:   if (contour->pA) {
220:     BVGetColumn(ctx->V,0,&v0);
221:     SlepcContourScatterCreate(contour,v0);
222:     BVRestoreColumn(ctx->V,0,&v0);
223:     BVDestroy(&ctx->pV);
224:     BVCreate(PetscObjectComm((PetscObject)contour->xsub),&ctx->pV);
225:     BVSetSizesFromVec(ctx->pV,contour->xsub,pep->n);
226:     BVSetFromOptions(ctx->pV);
227:     BVResize(ctx->pV,ctx->L,PETSC_FALSE);
228:   }

230:   BVDestroy(&ctx->Y);
231:   if (contour->pA) {
232:     BVCreate(PetscObjectComm((PetscObject)contour->xsub),&ctx->Y);
233:     BVSetSizesFromVec(ctx->Y,contour->xsub,pep->n);
234:     BVSetFromOptions(ctx->Y);
235:     BVResize(ctx->Y,contour->npoints*ctx->L,PETSC_FALSE);
236:   } else BVDuplicateResize(pep->V,contour->npoints*ctx->L,&ctx->Y);

238:   if (ctx->extraction == PEP_CISS_EXTRACTION_HANKEL) DSSetType(pep->ds,DSGNHEP);
239:   else if (ctx->extraction == PEP_CISS_EXTRACTION_CAA) DSSetType(pep->ds,DSNHEP);
240:   else {
241:     DSSetType(pep->ds,DSPEP);
242:     DSPEPSetDegree(pep->ds,pep->nmat-1);
243:     DSPEPSetCoefficients(pep->ds,pep->pbc);
244:   }
245:   DSAllocate(pep->ds,pep->ncv);
246:   nwork = 2;
247:   PEPSetWorkVecs(pep,nwork);
248:   return 0;
249: }

251: PetscErrorCode PEPSolve_CISS(PEP pep)
252: {
253:   PEP_CISS         *ctx = (PEP_CISS*)pep->data;
254:   SlepcContourData contour = ctx->contour;
255:   Mat              X,M,E,T,P;
256:   PetscInt         i,j,ld,L_add=0,nv=0,L_base=ctx->L,inner,*inside,nsplit;
257:   PetscScalar      *Mu,*H0,*H1,*rr,*temp,center;
258:   PetscReal        error,max_error,radius,rgscale,est_eig,eta;
259:   PetscBool        isellipse,*fl1;
260:   Vec              si;
261:   SlepcSC          sc;
262:   PetscRandom      rand;

264:   DSSetFromOptions(pep->ds);
265:   DSGetSlepcSC(pep->ds,&sc);
266:   sc->comparison    = SlepcCompareLargestMagnitude;
267:   sc->comparisonctx = NULL;
268:   sc->map           = NULL;
269:   sc->mapobj        = NULL;
270:   DSGetLeadingDimension(pep->ds,&ld);
271:   RGComputeQuadrature(pep->rg,RG_QUADRULE_TRAPEZOIDAL,ctx->N,ctx->omega,ctx->pp,ctx->weight);
272:   STGetSplitPreconditionerInfo(pep->st,&nsplit,NULL);
273:   if (contour->pA) {
274:     T = contour->pA[0];
275:     P = nsplit? contour->pP[0]: T;
276:   } else {
277:     T = pep->A[0];
278:     P = nsplit? ctx->Psplit[0]: T;
279:   }
280:   PEPCISSSetUp(pep,T,P);
281:   BVSetActiveColumns(ctx->V,0,ctx->L);
282:   BVSetRandomSign(ctx->V);
283:   BVGetRandomContext(ctx->V,&rand);
284:   if (contour->pA) BVScatter(ctx->V,ctx->pV,contour->scatterin,contour->xdup);
285:   PEPCISSSolve(pep,ctx->J,(contour->pA)?ctx->pV:ctx->V,0,ctx->L);
286:   PetscObjectTypeCompare((PetscObject)pep->rg,RGELLIPSE,&isellipse);
287:   if (isellipse) {
288:     BVTraceQuadrature(ctx->Y,ctx->V,ctx->L,ctx->L,ctx->weight,contour->scatterin,contour->subcomm,contour->npoints,ctx->useconj,&est_eig);
289:     PetscInfo(pep,"Estimated eigenvalue count: %f\n",(double)est_eig);
290:     eta = PetscPowReal(10.0,-PetscLog10Real(pep->tol)/ctx->N);
291:     L_add = PetscMax(0,(PetscInt)PetscCeilReal((est_eig*eta)/ctx->M)-ctx->L);
292:     if (L_add>ctx->L_max-ctx->L) {
293:       PetscInfo(pep,"Number of eigenvalues inside the contour path may be too large\n");
294:       L_add = ctx->L_max-ctx->L;
295:     }
296:   }
297:   /* Updates L after estimate the number of eigenvalue */
298:   if (L_add>0) {
299:     PetscInfo(pep,"Changing L %" PetscInt_FMT " -> %" PetscInt_FMT " by Estimate #Eig\n",ctx->L,ctx->L+L_add);
300:     BVCISSResizeBases(ctx->S,contour->pA?ctx->pV:ctx->V,ctx->Y,ctx->L,ctx->L+L_add,ctx->M,contour->npoints);
301:     BVSetActiveColumns(ctx->V,ctx->L,ctx->L+L_add);
302:     BVSetRandomSign(ctx->V);
303:     if (contour->pA) BVScatter(ctx->V,ctx->pV,contour->scatterin,contour->xdup);
304:     ctx->L += L_add;
305:     PEPCISSSolve(pep,ctx->J,(contour->pA)?ctx->pV:ctx->V,ctx->L-L_add,ctx->L);
306:   }

308:   PetscMalloc2(ctx->L*ctx->L*ctx->M*2,&Mu,ctx->L*ctx->M*ctx->L*ctx->M,&H0);
309:   for (i=0;i<ctx->refine_blocksize;i++) {
310:     BVDotQuadrature(ctx->Y,(contour->pA)?ctx->pV:ctx->V,Mu,ctx->M,ctx->L,ctx->L,ctx->weight,ctx->pp,contour->subcomm,contour->npoints,ctx->useconj);
311:     CISS_BlockHankel(Mu,0,ctx->L,ctx->M,H0);
312:     PetscLogEventBegin(PEP_CISS_SVD,pep,0,0,0);
313:     SlepcCISS_BH_SVD(H0,ctx->L*ctx->M,ctx->delta,ctx->sigma,&nv);
314:     PetscLogEventEnd(PEP_CISS_SVD,pep,0,0,0);
315:     if (ctx->sigma[0]<=ctx->delta || nv < ctx->L*ctx->M || ctx->L == ctx->L_max) break;
316:     L_add = L_base;
317:     if (ctx->L+L_add>ctx->L_max) L_add = ctx->L_max-ctx->L;
318:     PetscInfo(pep,"Changing L %" PetscInt_FMT " -> %" PetscInt_FMT " by SVD(H0)\n",ctx->L,ctx->L+L_add);
319:     BVCISSResizeBases(ctx->S,contour->pA?ctx->pV:ctx->V,ctx->Y,ctx->L,ctx->L+L_add,ctx->M,contour->npoints);
320:     BVSetActiveColumns(ctx->V,ctx->L,ctx->L+L_add);
321:     BVSetRandomSign(ctx->V);
322:     if (contour->pA) BVScatter(ctx->V,ctx->pV,contour->scatterin,contour->xdup);
323:     ctx->L += L_add;
324:     PEPCISSSolve(pep,ctx->J,(contour->pA)?ctx->pV:ctx->V,ctx->L-L_add,ctx->L);
325:     if (L_add) {
326:       PetscFree2(Mu,H0);
327:       PetscMalloc2(ctx->L*ctx->L*ctx->M*2,&Mu,ctx->L*ctx->M*ctx->L*ctx->M,&H0);
328:     }
329:   }

331:   RGGetScale(pep->rg,&rgscale);
332:   RGEllipseGetParameters(pep->rg,&center,&radius,NULL);

334:   if (ctx->extraction == PEP_CISS_EXTRACTION_HANKEL) PetscMalloc1(ctx->L*ctx->M*ctx->L*ctx->M,&H1);

336:   while (pep->reason == PEP_CONVERGED_ITERATING) {
337:     pep->its++;
338:     for (inner=0;inner<=ctx->refine_inner;inner++) {
339:       if (ctx->extraction == PEP_CISS_EXTRACTION_HANKEL) {
340:         BVDotQuadrature(ctx->Y,(contour->pA)?ctx->pV:ctx->V,Mu,ctx->M,ctx->L,ctx->L,ctx->weight,ctx->pp,contour->subcomm,contour->npoints,ctx->useconj);
341:         CISS_BlockHankel(Mu,0,ctx->L,ctx->M,H0);
342:         PetscLogEventBegin(PEP_CISS_SVD,pep,0,0,0);
343:         SlepcCISS_BH_SVD(H0,ctx->L*ctx->M,ctx->delta,ctx->sigma,&nv);
344:         PetscLogEventEnd(PEP_CISS_SVD,pep,0,0,0);
345:       } else {
346:         BVSumQuadrature(ctx->S,ctx->Y,ctx->M,ctx->L,ctx->L,ctx->weight,ctx->pp,contour->scatterin,contour->subcomm,contour->npoints,ctx->useconj);
347:         /* compute SVD of S */
348:         BVSVDAndRank(ctx->S,ctx->M,ctx->L,ctx->delta,(ctx->extraction==PEP_CISS_EXTRACTION_CAA)?BV_SVD_METHOD_QR_CAA:BV_SVD_METHOD_QR,H0,ctx->sigma,&nv);
349:       }
350:       PetscInfo(pep,"Estimated rank: nv = %" PetscInt_FMT "\n",nv);
351:       if (ctx->sigma[0]>ctx->delta && nv==ctx->L*ctx->M && inner!=ctx->refine_inner) {
352:         BVSumQuadrature(ctx->S,ctx->Y,ctx->M,ctx->L,ctx->L,ctx->weight,ctx->pp,contour->scatterin,contour->subcomm,contour->npoints,ctx->useconj);
353:         BVSetActiveColumns(ctx->S,0,ctx->L);
354:         BVSetActiveColumns(ctx->V,0,ctx->L);
355:         BVCopy(ctx->S,ctx->V);
356:         if (contour->pA) BVScatter(ctx->V,ctx->pV,contour->scatterin,contour->xdup);
357:         PEPCISSSolve(pep,ctx->J,(contour->pA)?ctx->pV:ctx->V,0,ctx->L);
358:       } else break;
359:     }
360:     pep->nconv = 0;
361:     if (nv == 0) { pep->reason = PEP_CONVERGED_TOL; break; }
362:     else {
363:       /* Extracting eigenpairs */
364:       DSSetDimensions(pep->ds,nv,0,0);
365:       DSSetState(pep->ds,DS_STATE_RAW);
366:       if (ctx->extraction == PEP_CISS_EXTRACTION_HANKEL) {
367:         CISS_BlockHankel(Mu,0,ctx->L,ctx->M,H0);
368:         CISS_BlockHankel(Mu,1,ctx->L,ctx->M,H1);
369:         DSGetArray(pep->ds,DS_MAT_A,&temp);
370:         for (j=0;j<nv;j++)
371:           for (i=0;i<nv;i++)
372:             temp[i+j*ld] = H1[i+j*ctx->L*ctx->M];
373:         DSRestoreArray(pep->ds,DS_MAT_A,&temp);
374:         DSGetArray(pep->ds,DS_MAT_B,&temp);
375:         for (j=0;j<nv;j++)
376:           for (i=0;i<nv;i++)
377:             temp[i+j*ld] = H0[i+j*ctx->L*ctx->M];
378:         DSRestoreArray(pep->ds,DS_MAT_B,&temp);
379:       } else if (ctx->extraction == PEP_CISS_EXTRACTION_CAA) {
380:         BVSetActiveColumns(ctx->S,0,nv);
381:         DSGetArray(pep->ds,DS_MAT_A,&temp);
382:         for (i=0;i<nv;i++) PetscArraycpy(temp+i*ld,H0+i*nv,nv);
383:         DSRestoreArray(pep->ds,DS_MAT_A,&temp);
384:       } else {
385:         BVSetActiveColumns(ctx->S,0,nv);
386:         for (i=0;i<pep->nmat;i++) {
387:           DSGetMat(pep->ds,DSMatExtra[i],&E);
388:           BVMatProject(ctx->S,pep->A[i],ctx->S,E);
389:           DSRestoreMat(pep->ds,DSMatExtra[i],&E);
390:         }
391:         nv = (pep->nmat-1)*nv;
392:       }
393:       DSSolve(pep->ds,pep->eigr,pep->eigi);
394:       DSSynchronize(pep->ds,pep->eigr,pep->eigi);
395:       if (ctx->extraction == PEP_CISS_EXTRACTION_CAA || ctx->extraction == PEP_CISS_EXTRACTION_HANKEL) {
396:         for (i=0;i<nv;i++) {
397:           pep->eigr[i] = (pep->eigr[i]*radius+center)*rgscale;
398:         }
399:       }
400:       PetscMalloc3(nv,&fl1,nv,&inside,nv,&rr);
401:       DSVectors(pep->ds,DS_MAT_X,NULL,NULL);
402:       DSGetMat(pep->ds,DS_MAT_X,&X);
403:       SlepcCISS_isGhost(X,nv,ctx->sigma,ctx->spurious_threshold,fl1);
404:       DSRestoreMat(pep->ds,DS_MAT_X,&X);
405:       RGCheckInside(pep->rg,nv,pep->eigr,pep->eigi,inside);
406:       for (i=0;i<nv;i++) {
407:         if (fl1[i] && inside[i]>=0) {
408:           rr[i] = 1.0;
409:           pep->nconv++;
410:         } else rr[i] = 0.0;
411:       }
412:       DSSort(pep->ds,pep->eigr,pep->eigi,rr,NULL,&pep->nconv);
413:       DSSynchronize(pep->ds,pep->eigr,pep->eigi);
414:       if (ctx->extraction == PEP_CISS_EXTRACTION_CAA || ctx->extraction == PEP_CISS_EXTRACTION_HANKEL) {
415:         for (i=0;i<nv;i++) pep->eigr[i] = (pep->eigr[i]*radius+center)*rgscale;
416:       }
417:       PetscFree3(fl1,inside,rr);
418:       BVSetActiveColumns(pep->V,0,nv);
419:       DSVectors(pep->ds,DS_MAT_X,NULL,NULL);
420:       if (ctx->extraction == PEP_CISS_EXTRACTION_HANKEL) {
421:         BVSumQuadrature(ctx->S,ctx->Y,ctx->M,ctx->L,ctx->L,ctx->weight,ctx->pp,contour->scatterin,contour->subcomm,contour->npoints,ctx->useconj);
422:         BVSetActiveColumns(ctx->S,0,nv);
423:         BVCopy(ctx->S,pep->V);
424:         DSGetMat(pep->ds,DS_MAT_X,&X);
425:         BVMultInPlace(ctx->S,X,0,pep->nconv);
426:         BVMultInPlace(pep->V,X,0,pep->nconv);
427:         DSRestoreMat(pep->ds,DS_MAT_X,&X);
428:       } else {
429:         DSGetMat(pep->ds,DS_MAT_X,&X);
430:         BVMultInPlace(ctx->S,X,0,pep->nconv);
431:         DSRestoreMat(pep->ds,DS_MAT_X,&X);
432:         BVSetActiveColumns(ctx->S,0,pep->nconv);
433:         BVCopy(ctx->S,pep->V);
434:       }
435:       max_error = 0.0;
436:       for (i=0;i<pep->nconv;i++) {
437:         BVGetColumn(pep->V,i,&si);
438:         VecNormalize(si,NULL);
439:         PEPComputeResidualNorm_Private(pep,pep->eigr[i],0,si,NULL,pep->work,&error);
440:         (*pep->converged)(pep,pep->eigr[i],0,error,&error,pep->convergedctx);
441:         BVRestoreColumn(pep->V,i,&si);
442:         max_error = PetscMax(max_error,error);
443:       }
444:       if (max_error <= pep->tol) pep->reason = PEP_CONVERGED_TOL;
445:       else if (pep->its > pep->max_it) pep->reason = PEP_DIVERGED_ITS;
446:       else {
447:         if (pep->nconv > ctx->L) nv = pep->nconv;
448:         else if (ctx->L > nv) nv = ctx->L;
449:         nv = PetscMin(nv,ctx->L*ctx->M);
450:         MatCreateSeqDense(PETSC_COMM_SELF,nv,ctx->L,NULL,&M);
451:         MatSetRandom(M,rand);
452:         BVSetActiveColumns(ctx->S,0,nv);
453:         BVMultInPlace(ctx->S,M,0,ctx->L);
454:         MatDestroy(&M);
455:         BVSetActiveColumns(ctx->S,0,ctx->L);
456:         BVSetActiveColumns(ctx->V,0,ctx->L);
457:         BVCopy(ctx->S,ctx->V);
458:         if (contour->pA) BVScatter(ctx->V,ctx->pV,contour->scatterin,contour->xdup);
459:         PEPCISSSolve(pep,ctx->J,(contour->pA)?ctx->pV:ctx->V,0,ctx->L);
460:       }
461:     }
462:   }
463:   PetscFree2(Mu,H0);
464:   if (ctx->extraction == PEP_CISS_EXTRACTION_HANKEL) PetscFree(H1);
465:   return 0;
466: }

468: static PetscErrorCode PEPCISSSetSizes_CISS(PEP pep,PetscInt ip,PetscInt bs,PetscInt ms,PetscInt npart,PetscInt bsmax,PetscBool realmats)
469: {
470:   PEP_CISS       *ctx = (PEP_CISS*)pep->data;
471:   PetscInt       oN,oL,oM,oLmax,onpart;
472:   PetscMPIInt    size;

474:   oN = ctx->N;
475:   if (ip == PETSC_DECIDE || ip == PETSC_DEFAULT) {
476:     if (ctx->N!=32) { ctx->N =32; ctx->M = ctx->N/4; }
477:   } else {
480:     if (ctx->N!=ip) { ctx->N = ip; ctx->M = ctx->N/4; }
481:   }
482:   oL = ctx->L;
483:   if (bs == PETSC_DECIDE || bs == PETSC_DEFAULT) {
484:     ctx->L = 16;
485:   } else {
487:     ctx->L = bs;
488:   }
489:   oM = ctx->M;
490:   if (ms == PETSC_DECIDE || ms == PETSC_DEFAULT) {
491:     ctx->M = ctx->N/4;
492:   } else {
495:     ctx->M = PetscMax(ms,2);
496:   }
497:   onpart = ctx->npart;
498:   if (npart == PETSC_DECIDE || npart == PETSC_DEFAULT) {
499:     ctx->npart = 1;
500:   } else {
501:     MPI_Comm_size(PetscObjectComm((PetscObject)pep),&size);
503:     ctx->npart = npart;
504:   }
505:   oLmax = ctx->L_max;
506:   if (bsmax == PETSC_DECIDE || bsmax == PETSC_DEFAULT) {
507:     ctx->L_max = 64;
508:   } else {
510:     ctx->L_max = PetscMax(bsmax,ctx->L);
511:   }
512:   if (onpart != ctx->npart || oN != ctx->N || realmats != ctx->isreal) {
513:     SlepcContourDataDestroy(&ctx->contour);
514:     PetscInfo(pep,"Resetting the contour data structure due to a change of parameters\n");
515:     pep->state = PEP_STATE_INITIAL;
516:   }
517:   ctx->isreal = realmats;
518:   if (oL != ctx->L || oM != ctx->M || oLmax != ctx->L_max) pep->state = PEP_STATE_INITIAL;
519:   return 0;
520: }

522: /*@
523:    PEPCISSSetSizes - Sets the values of various size parameters in the CISS solver.

525:    Logically Collective on pep

527:    Input Parameters:
528: +  pep   - the polynomial eigensolver context
529: .  ip    - number of integration points
530: .  bs    - block size
531: .  ms    - moment size
532: .  npart - number of partitions when splitting the communicator
533: .  bsmax - max block size
534: -  realmats - all coefficient matrices of P(.) are real

536:    Options Database Keys:
537: +  -pep_ciss_integration_points - Sets the number of integration points
538: .  -pep_ciss_blocksize - Sets the block size
539: .  -pep_ciss_moments - Sets the moment size
540: .  -pep_ciss_partitions - Sets the number of partitions
541: .  -pep_ciss_maxblocksize - Sets the maximum block size
542: -  -pep_ciss_realmats - all coefficient matrices of P(.) are real

544:    Notes:
545:    The default number of partitions is 1. This means the internal KSP object is shared
546:    among all processes of the PEP communicator. Otherwise, the communicator is split
547:    into npart communicators, so that npart KSP solves proceed simultaneously.

549:    Level: advanced

551: .seealso: PEPCISSGetSizes()
552: @*/
553: PetscErrorCode PEPCISSSetSizes(PEP pep,PetscInt ip,PetscInt bs,PetscInt ms,PetscInt npart,PetscInt bsmax,PetscBool realmats)
554: {
562:   PetscTryMethod(pep,"PEPCISSSetSizes_C",(PEP,PetscInt,PetscInt,PetscInt,PetscInt,PetscInt,PetscBool),(pep,ip,bs,ms,npart,bsmax,realmats));
563:   return 0;
564: }

566: static PetscErrorCode PEPCISSGetSizes_CISS(PEP pep,PetscInt *ip,PetscInt *bs,PetscInt *ms,PetscInt *npart,PetscInt *bsmax,PetscBool *realmats)
567: {
568:   PEP_CISS *ctx = (PEP_CISS*)pep->data;

570:   if (ip) *ip = ctx->N;
571:   if (bs) *bs = ctx->L;
572:   if (ms) *ms = ctx->M;
573:   if (npart) *npart = ctx->npart;
574:   if (bsmax) *bsmax = ctx->L_max;
575:   if (realmats) *realmats = ctx->isreal;
576:   return 0;
577: }

579: /*@
580:    PEPCISSGetSizes - Gets the values of various size parameters in the CISS solver.

582:    Not Collective

584:    Input Parameter:
585: .  pep - the polynomial eigensolver context

587:    Output Parameters:
588: +  ip    - number of integration points
589: .  bs    - block size
590: .  ms    - moment size
591: .  npart - number of partitions when splitting the communicator
592: .  bsmax - max block size
593: -  realmats - all coefficient matrices of P(.) are real

595:    Level: advanced

597: .seealso: PEPCISSSetSizes()
598: @*/
599: PetscErrorCode PEPCISSGetSizes(PEP pep,PetscInt *ip,PetscInt *bs,PetscInt *ms,PetscInt *npart,PetscInt *bsmax,PetscBool *realmats)
600: {
602:   PetscUseMethod(pep,"PEPCISSGetSizes_C",(PEP,PetscInt*,PetscInt*,PetscInt*,PetscInt*,PetscInt*,PetscBool*),(pep,ip,bs,ms,npart,bsmax,realmats));
603:   return 0;
604: }

606: static PetscErrorCode PEPCISSSetThreshold_CISS(PEP pep,PetscReal delta,PetscReal spur)
607: {
608:   PEP_CISS *ctx = (PEP_CISS*)pep->data;

610:   if (delta == PETSC_DEFAULT) {
611:     ctx->delta = SLEPC_DEFAULT_TOL*1e-4;
612:   } else {
614:     ctx->delta = delta;
615:   }
616:   if (spur == PETSC_DEFAULT) {
617:     ctx->spurious_threshold = PetscSqrtReal(SLEPC_DEFAULT_TOL);
618:   } else {
620:     ctx->spurious_threshold = spur;
621:   }
622:   return 0;
623: }

625: /*@
626:    PEPCISSSetThreshold - Sets the values of various threshold parameters in
627:    the CISS solver.

629:    Logically Collective on pep

631:    Input Parameters:
632: +  pep   - the polynomial eigensolver context
633: .  delta - threshold for numerical rank
634: -  spur  - spurious threshold (to discard spurious eigenpairs)

636:    Options Database Keys:
637: +  -pep_ciss_delta - Sets the delta
638: -  -pep_ciss_spurious_threshold - Sets the spurious threshold

640:    Level: advanced

642: .seealso: PEPCISSGetThreshold()
643: @*/
644: PetscErrorCode PEPCISSSetThreshold(PEP pep,PetscReal delta,PetscReal spur)
645: {
649:   PetscTryMethod(pep,"PEPCISSSetThreshold_C",(PEP,PetscReal,PetscReal),(pep,delta,spur));
650:   return 0;
651: }

653: static PetscErrorCode PEPCISSGetThreshold_CISS(PEP pep,PetscReal *delta,PetscReal *spur)
654: {
655:   PEP_CISS *ctx = (PEP_CISS*)pep->data;

657:   if (delta) *delta = ctx->delta;
658:   if (spur)  *spur = ctx->spurious_threshold;
659:   return 0;
660: }

662: /*@
663:    PEPCISSGetThreshold - Gets the values of various threshold parameters in
664:    the CISS solver.

666:    Not Collective

668:    Input Parameter:
669: .  pep - the polynomial eigensolver context

671:    Output Parameters:
672: +  delta - threshold for numerical rank
673: -  spur  - spurious threshold (to discard spurious eigenpairs)

675:    Level: advanced

677: .seealso: PEPCISSSetThreshold()
678: @*/
679: PetscErrorCode PEPCISSGetThreshold(PEP pep,PetscReal *delta,PetscReal *spur)
680: {
682:   PetscUseMethod(pep,"PEPCISSGetThreshold_C",(PEP,PetscReal*,PetscReal*),(pep,delta,spur));
683:   return 0;
684: }

686: static PetscErrorCode PEPCISSSetRefinement_CISS(PEP pep,PetscInt inner,PetscInt blsize)
687: {
688:   PEP_CISS *ctx = (PEP_CISS*)pep->data;

690:   if (inner == PETSC_DEFAULT) {
691:     ctx->refine_inner = 0;
692:   } else {
694:     ctx->refine_inner = inner;
695:   }
696:   if (blsize == PETSC_DEFAULT) {
697:     ctx->refine_blocksize = 0;
698:   } else {
700:     ctx->refine_blocksize = blsize;
701:   }
702:   return 0;
703: }

705: /*@
706:    PEPCISSSetRefinement - Sets the values of various refinement parameters
707:    in the CISS solver.

709:    Logically Collective on pep

711:    Input Parameters:
712: +  pep    - the polynomial eigensolver context
713: .  inner  - number of iterative refinement iterations (inner loop)
714: -  blsize - number of iterative refinement iterations (blocksize loop)

716:    Options Database Keys:
717: +  -pep_ciss_refine_inner - Sets number of inner iterations
718: -  -pep_ciss_refine_blocksize - Sets number of blocksize iterations

720:    Level: advanced

722: .seealso: PEPCISSGetRefinement()
723: @*/
724: PetscErrorCode PEPCISSSetRefinement(PEP pep,PetscInt inner,PetscInt blsize)
725: {
729:   PetscTryMethod(pep,"PEPCISSSetRefinement_C",(PEP,PetscInt,PetscInt),(pep,inner,blsize));
730:   return 0;
731: }

733: static PetscErrorCode PEPCISSGetRefinement_CISS(PEP pep,PetscInt *inner,PetscInt *blsize)
734: {
735:   PEP_CISS *ctx = (PEP_CISS*)pep->data;

737:   if (inner)  *inner = ctx->refine_inner;
738:   if (blsize) *blsize = ctx->refine_blocksize;
739:   return 0;
740: }

742: /*@
743:    PEPCISSGetRefinement - Gets the values of various refinement parameters
744:    in the CISS solver.

746:    Not Collective

748:    Input Parameter:
749: .  pep - the polynomial eigensolver context

751:    Output Parameters:
752: +  inner  - number of iterative refinement iterations (inner loop)
753: -  blsize - number of iterative refinement iterations (blocksize loop)

755:    Level: advanced

757: .seealso: PEPCISSSetRefinement()
758: @*/
759: PetscErrorCode PEPCISSGetRefinement(PEP pep, PetscInt *inner, PetscInt *blsize)
760: {
762:   PetscUseMethod(pep,"PEPCISSGetRefinement_C",(PEP,PetscInt*,PetscInt*),(pep,inner,blsize));
763:   return 0;
764: }

766: static PetscErrorCode PEPCISSSetExtraction_CISS(PEP pep,PEPCISSExtraction extraction)
767: {
768:   PEP_CISS *ctx = (PEP_CISS*)pep->data;

770:   if (ctx->extraction != extraction) {
771:     ctx->extraction = extraction;
772:     pep->state      = PEP_STATE_INITIAL;
773:   }
774:   return 0;
775: }

777: /*@
778:    PEPCISSSetExtraction - Sets the extraction technique used in the CISS solver.

780:    Logically Collective on pep

782:    Input Parameters:
783: +  pep        - the polynomial eigensolver context
784: -  extraction - the extraction technique

786:    Options Database Key:
787: .  -pep_ciss_extraction - Sets the extraction technique (either 'ritz', 'hankel' or 'caa')

789:    Notes:
790:    By default, the Rayleigh-Ritz extraction is used (PEP_CISS_EXTRACTION_RITZ).

792:    If the 'hankel' or the 'caa' option is specified (PEP_CISS_EXTRACTION_HANKEL or
793:    PEP_CISS_EXTRACTION_CAA), then the Block Hankel method, or the Communication-avoiding
794:    Arnoldi method, respectively, is used for extracting eigenpairs.

796:    Level: advanced

798: .seealso: PEPCISSGetExtraction(), PEPCISSExtraction
799: @*/
800: PetscErrorCode PEPCISSSetExtraction(PEP pep,PEPCISSExtraction extraction)
801: {
804:   PetscTryMethod(pep,"PEPCISSSetExtraction_C",(PEP,PEPCISSExtraction),(pep,extraction));
805:   return 0;
806: }

808: static PetscErrorCode PEPCISSGetExtraction_CISS(PEP pep,PEPCISSExtraction *extraction)
809: {
810:   PEP_CISS *ctx = (PEP_CISS*)pep->data;

812:   *extraction = ctx->extraction;
813:   return 0;
814: }

816: /*@
817:    PEPCISSGetExtraction - Gets the extraction technique used in the CISS solver.

819:    Not Collective

821:    Input Parameter:
822: .  pep - the polynomial eigensolver context

824:    Output Parameters:
825: .  extraction - extraction technique

827:    Level: advanced

829: .seealso: PEPCISSSetExtraction() PEPCISSExtraction
830: @*/
831: PetscErrorCode PEPCISSGetExtraction(PEP pep,PEPCISSExtraction *extraction)
832: {
835:   PetscUseMethod(pep,"PEPCISSGetExtraction_C",(PEP,PEPCISSExtraction*),(pep,extraction));
836:   return 0;
837: }

839: static PetscErrorCode PEPCISSGetKSPs_CISS(PEP pep,PetscInt *nsolve,KSP **ksp)
840: {
841:   PEP_CISS         *ctx = (PEP_CISS*)pep->data;
842:   SlepcContourData contour;
843:   PetscInt         i,nsplit;
844:   PC               pc;
845:   MPI_Comm         child;

847:   if (!ctx->contour) {  /* initialize contour data structure first */
848:     RGCanUseConjugates(pep->rg,ctx->isreal,&ctx->useconj);
849:     SlepcContourDataCreate(ctx->useconj?ctx->N/2:ctx->N,ctx->npart,(PetscObject)pep,&ctx->contour);
850:   }
851:   contour = ctx->contour;
852:   if (!contour->ksp) {
853:     PetscMalloc1(contour->npoints,&contour->ksp);
854:     PEPGetST(pep,&pep->st);
855:     STGetSplitPreconditionerInfo(pep->st,&nsplit,NULL);
856:     PetscSubcommGetChild(contour->subcomm,&child);
857:     for (i=0;i<contour->npoints;i++) {
858:       KSPCreate(child,&contour->ksp[i]);
859:       PetscObjectIncrementTabLevel((PetscObject)contour->ksp[i],(PetscObject)pep,1);
860:       KSPSetOptionsPrefix(contour->ksp[i],((PetscObject)pep)->prefix);
861:       KSPAppendOptionsPrefix(contour->ksp[i],"pep_ciss_");
862:       PetscObjectSetOptions((PetscObject)contour->ksp[i],((PetscObject)pep)->options);
863:       KSPSetErrorIfNotConverged(contour->ksp[i],PETSC_TRUE);
864:       KSPSetTolerances(contour->ksp[i],SlepcDefaultTol(pep->tol),PETSC_DEFAULT,PETSC_DEFAULT,PETSC_DEFAULT);
865:       KSPGetPC(contour->ksp[i],&pc);
866:       if (nsplit) {
867:         KSPSetType(contour->ksp[i],KSPBCGS);
868:         PCSetType(pc,PCBJACOBI);
869:       } else {
870:         KSPSetType(contour->ksp[i],KSPPREONLY);
871:         PCSetType(pc,PCLU);
872:       }
873:     }
874:   }
875:   if (nsolve) *nsolve = contour->npoints;
876:   if (ksp)    *ksp    = contour->ksp;
877:   return 0;
878: }

880: /*@C
881:    PEPCISSGetKSPs - Retrieve the array of linear solver objects associated with
882:    the CISS solver.

884:    Not Collective

886:    Input Parameter:
887: .  pep - polynomial eigenvalue solver

889:    Output Parameters:
890: +  nsolve - number of solver objects
891: -  ksp - array of linear solver object

893:    Notes:
894:    The number of KSP solvers is equal to the number of integration points divided by
895:    the number of partitions. This value is halved in the case of real matrices with
896:    a region centered at the real axis.

898:    Level: advanced

900: .seealso: PEPCISSSetSizes()
901: @*/
902: PetscErrorCode PEPCISSGetKSPs(PEP pep,PetscInt *nsolve,KSP **ksp)
903: {
905:   PetscUseMethod(pep,"PEPCISSGetKSPs_C",(PEP,PetscInt*,KSP**),(pep,nsolve,ksp));
906:   return 0;
907: }

909: PetscErrorCode PEPReset_CISS(PEP pep)
910: {
911:   PEP_CISS       *ctx = (PEP_CISS*)pep->data;

913:   BVDestroy(&ctx->S);
914:   BVDestroy(&ctx->V);
915:   BVDestroy(&ctx->Y);
916:   SlepcContourDataReset(ctx->contour);
917:   MatDestroy(&ctx->J);
918:   BVDestroy(&ctx->pV);
919:   PetscFree(ctx->Psplit);
920:   return 0;
921: }

923: PetscErrorCode PEPSetFromOptions_CISS(PEP pep,PetscOptionItems *PetscOptionsObject)
924: {
925:   PEP_CISS          *ctx = (PEP_CISS*)pep->data;
926:   PetscReal         r1,r2;
927:   PetscInt          i,i1,i2,i3,i4,i5,i6,i7;
928:   PetscBool         b1,flg,flg2,flg3,flg4,flg5,flg6;
929:   PEPCISSExtraction extraction;

931:   PetscOptionsHeadBegin(PetscOptionsObject,"PEP CISS Options");

933:     PEPCISSGetSizes(pep,&i1,&i2,&i3,&i4,&i5,&b1);
934:     PetscOptionsInt("-pep_ciss_integration_points","Number of integration points","PEPCISSSetSizes",i1,&i1,&flg);
935:     PetscOptionsInt("-pep_ciss_blocksize","Block size","PEPCISSSetSizes",i2,&i2,&flg2);
936:     PetscOptionsInt("-pep_ciss_moments","Moment size","PEPCISSSetSizes",i3,&i3,&flg3);
937:     PetscOptionsInt("-pep_ciss_partitions","Number of partitions","PEPCISSSetSizes",i4,&i4,&flg4);
938:     PetscOptionsInt("-pep_ciss_maxblocksize","Maximum block size","PEPCISSSetSizes",i5,&i5,&flg5);
939:     PetscOptionsBool("-pep_ciss_realmats","True if all coefficient matrices of P(.) are real","PEPCISSSetSizes",b1,&b1,&flg6);
940:     if (flg || flg2 || flg3 || flg4 || flg5 || flg6) PEPCISSSetSizes(pep,i1,i2,i3,i4,i5,b1);

942:     PEPCISSGetThreshold(pep,&r1,&r2);
943:     PetscOptionsReal("-pep_ciss_delta","Threshold for numerical rank","PEPCISSSetThreshold",r1,&r1,&flg);
944:     PetscOptionsReal("-pep_ciss_spurious_threshold","Threshold for the spurious eigenpairs","PEPCISSSetThreshold",r2,&r2,&flg2);
945:     if (flg || flg2) PEPCISSSetThreshold(pep,r1,r2);

947:     PEPCISSGetRefinement(pep,&i6,&i7);
948:     PetscOptionsInt("-pep_ciss_refine_inner","Number of inner iterative refinement iterations","PEPCISSSetRefinement",i6,&i6,&flg);
949:     PetscOptionsInt("-pep_ciss_refine_blocksize","Number of blocksize iterative refinement iterations","PEPCISSSetRefinement",i7,&i7,&flg2);
950:     if (flg || flg2) PEPCISSSetRefinement(pep,i6,i7);

952:     PetscOptionsEnum("-pep_ciss_extraction","Extraction technique","PEPCISSSetExtraction",PEPCISSExtractions,(PetscEnum)ctx->extraction,(PetscEnum*)&extraction,&flg);
953:     if (flg) PEPCISSSetExtraction(pep,extraction);

955:   PetscOptionsHeadEnd();

957:   if (!pep->rg) PEPGetRG(pep,&pep->rg);
958:   RGSetFromOptions(pep->rg); /* this is necessary here to set useconj */
959:   if (!ctx->contour || !ctx->contour->ksp) PEPCISSGetKSPs(pep,NULL,NULL);
960:   PetscAssert(ctx->contour && ctx->contour->ksp,PetscObjectComm((PetscObject)pep),PETSC_ERR_PLIB,"Something went wrong with PEPCISSGetKSPs()");
961:   for (i=0;i<ctx->contour->npoints;i++) KSPSetFromOptions(ctx->contour->ksp[i]);
962:   PetscSubcommSetFromOptions(ctx->contour->subcomm);
963:   return 0;
964: }

966: PetscErrorCode PEPDestroy_CISS(PEP pep)
967: {
968:   PEP_CISS       *ctx = (PEP_CISS*)pep->data;

970:   SlepcContourDataDestroy(&ctx->contour);
971:   PetscFree4(ctx->weight,ctx->omega,ctx->pp,ctx->sigma);
972:   PetscFree(pep->data);
973:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSSetSizes_C",NULL);
974:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSGetSizes_C",NULL);
975:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSSetThreshold_C",NULL);
976:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSGetThreshold_C",NULL);
977:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSSetRefinement_C",NULL);
978:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSGetRefinement_C",NULL);
979:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSSetExtraction_C",NULL);
980:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSGetExtraction_C",NULL);
981:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSGetKSPs_C",NULL);
982:   return 0;
983: }

985: PetscErrorCode PEPView_CISS(PEP pep,PetscViewer viewer)
986: {
987:   PEP_CISS       *ctx = (PEP_CISS*)pep->data;
988:   PetscBool      isascii;
989:   PetscViewer    sviewer;

991:   PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&isascii);
992:   if (isascii) {
993:     PetscViewerASCIIPrintf(viewer,"  sizes { integration points: %" PetscInt_FMT ", block size: %" PetscInt_FMT ", moment size: %" PetscInt_FMT ", partitions: %" PetscInt_FMT ", maximum block size: %" PetscInt_FMT " }\n",ctx->N,ctx->L,ctx->M,ctx->npart,ctx->L_max);
994:     if (ctx->isreal) PetscViewerASCIIPrintf(viewer,"  exploiting symmetry of integration points\n");
995:     PetscViewerASCIIPrintf(viewer,"  threshold { delta: %g, spurious threshold: %g }\n",(double)ctx->delta,(double)ctx->spurious_threshold);
996:     PetscViewerASCIIPrintf(viewer,"  iterative refinement  { inner: %" PetscInt_FMT ", blocksize: %" PetscInt_FMT " }\n",ctx->refine_inner, ctx->refine_blocksize);
997:     PetscViewerASCIIPrintf(viewer,"  extraction: %s\n",PEPCISSExtractions[ctx->extraction]);
998:     if (!ctx->contour || !ctx->contour->ksp) PEPCISSGetKSPs(pep,NULL,NULL);
999:     PetscAssert(ctx->contour && ctx->contour->ksp,PetscObjectComm((PetscObject)pep),PETSC_ERR_PLIB,"Something went wrong with PEPCISSGetKSPs()");
1000:     PetscViewerASCIIPushTab(viewer);
1001:     if (ctx->npart>1 && ctx->contour->subcomm) {
1002:       PetscViewerGetSubViewer(viewer,ctx->contour->subcomm->child,&sviewer);
1003:       if (!ctx->contour->subcomm->color) KSPView(ctx->contour->ksp[0],sviewer);
1004:       PetscViewerFlush(sviewer);
1005:       PetscViewerRestoreSubViewer(viewer,ctx->contour->subcomm->child,&sviewer);
1006:       PetscViewerFlush(viewer);
1007:       /* extra call needed because of the two calls to PetscViewerASCIIPushSynchronized() in PetscViewerGetSubViewer() */
1008:       PetscViewerASCIIPopSynchronized(viewer);
1009:     } else KSPView(ctx->contour->ksp[0],viewer);
1010:     PetscViewerASCIIPopTab(viewer);
1011:   }
1012:   return 0;
1013: }

1015: SLEPC_EXTERN PetscErrorCode PEPCreate_CISS(PEP pep)
1016: {
1017:   PEP_CISS       *ctx = (PEP_CISS*)pep->data;

1019:   PetscNew(&ctx);
1020:   pep->data = ctx;
1021:   /* set default values of parameters */
1022:   ctx->N                  = 32;
1023:   ctx->L                  = 16;
1024:   ctx->M                  = ctx->N/4;
1025:   ctx->delta              = SLEPC_DEFAULT_TOL*1e-4;
1026:   ctx->L_max              = 64;
1027:   ctx->spurious_threshold = PetscSqrtReal(SLEPC_DEFAULT_TOL);
1028:   ctx->isreal             = PETSC_FALSE;
1029:   ctx->npart              = 1;

1031:   pep->ops->solve          = PEPSolve_CISS;
1032:   pep->ops->setup          = PEPSetUp_CISS;
1033:   pep->ops->setfromoptions = PEPSetFromOptions_CISS;
1034:   pep->ops->reset          = PEPReset_CISS;
1035:   pep->ops->destroy        = PEPDestroy_CISS;
1036:   pep->ops->view           = PEPView_CISS;

1038:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSSetSizes_C",PEPCISSSetSizes_CISS);
1039:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSGetSizes_C",PEPCISSGetSizes_CISS);
1040:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSSetThreshold_C",PEPCISSSetThreshold_CISS);
1041:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSGetThreshold_C",PEPCISSGetThreshold_CISS);
1042:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSSetRefinement_C",PEPCISSSetRefinement_CISS);
1043:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSGetRefinement_C",PEPCISSGetRefinement_CISS);
1044:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSSetExtraction_C",PEPCISSSetExtraction_CISS);
1045:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSGetExtraction_C",PEPCISSGetExtraction_CISS);
1046:   PetscObjectComposeFunction((PetscObject)pep,"PEPCISSGetKSPs_C",PEPCISSGetKSPs_CISS);
1047:   return 0;
1048: }