Actual source code: ciss.c
slepc-3.14.1 2020-12-08
1: /*
2: - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
3: SLEPc - Scalable Library for Eigenvalue Problem Computations
4: Copyright (c) 2002-2020, 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] T. Sakurai and H. Sugiura, "A projection method for generalized
26: eigenvalue problems", J. Comput. Appl. Math. 159:119-128, 2003.
28: [2] T. Sakurai and H. Tadano, "CIRR: a Rayleigh-Ritz type method with
29: contour integral for generalized eigenvalue problems", Hokkaido
30: Math. J. 36:745-757, 2007.
31: */
33: #include <slepc/private/epsimpl.h>
34: #include <slepcblaslapack.h>
36: PetscLogEvent EPS_CISS_SVD;
38: typedef struct {
39: /* parameters */
40: PetscInt N; /* number of integration points (32) */
41: PetscInt L; /* block size (16) */
42: PetscInt M; /* moment degree (N/4 = 4) */
43: PetscReal delta; /* threshold of singular value (1e-12) */
44: PetscInt L_max; /* maximum number of columns of the source matrix V */
45: PetscReal spurious_threshold; /* discard spurious eigenpairs */
46: PetscBool isreal; /* A and B are real */
47: PetscInt npart; /* number of partitions */
48: PetscInt refine_inner;
49: PetscInt refine_blocksize;
50: /* private data */
51: PetscReal *sigma; /* threshold for numerical rank */
52: PetscInt subcomm_id;
53: PetscInt num_solve_point;
54: PetscScalar *weight;
55: PetscScalar *omega;
56: PetscScalar *pp;
57: BV V;
58: BV S;
59: BV pV;
60: BV Y;
61: Vec xsub;
62: Vec xdup;
63: KSP *ksp; /* ksp array for storing factorizations at integration points */
64: PetscBool useconj;
65: PetscReal est_eig;
66: VecScatter scatterin;
67: Mat pA,pB;
68: PetscSubcomm subcomm;
69: PetscBool usest;
70: PetscBool usest_set; /* whether the user set the usest flag or not */
71: EPSCISSQuadRule quad;
72: EPSCISSExtraction extraction;
73: } EPS_CISS;
75: /* destroy KSP objects when the number of solve points changes */
76: PETSC_STATIC_INLINE PetscErrorCode EPSCISSResetSolvers(EPS eps)
77: {
79: PetscInt i;
80: EPS_CISS *ctx = (EPS_CISS*)eps->data;
83: if (ctx->ksp) {
84: for (i=0;i<ctx->num_solve_point;i++) {
85: KSPDestroy(&ctx->ksp[i]);
86: }
87: PetscFree(ctx->ksp);
88: }
89: return(0);
90: }
92: /* clean PetscSubcomm object when the number of partitions changes */
93: PETSC_STATIC_INLINE PetscErrorCode EPSCISSResetSubcomm(EPS eps)
94: {
96: EPS_CISS *ctx = (EPS_CISS*)eps->data;
99: EPSCISSResetSolvers(eps);
100: PetscSubcommDestroy(&ctx->subcomm);
101: return(0);
102: }
104: /* determine whether half of integration points can be avoided (use its conjugates);
105: depends on isreal and the center of the region */
106: PETSC_STATIC_INLINE PetscErrorCode EPSCISSSetUseConj(EPS eps,PetscBool *useconj)
107: {
109: PetscScalar center;
110: PetscReal c,d;
111: PetscBool isellipse,isinterval;
112: #if defined(PETSC_USE_COMPLEX)
113: EPS_CISS *ctx = (EPS_CISS*)eps->data;
114: #endif
117: *useconj = PETSC_FALSE;
118: PetscObjectTypeCompare((PetscObject)eps->rg,RGELLIPSE,&isellipse);
119: if (isellipse) {
120: RGEllipseGetParameters(eps->rg,¢er,NULL,NULL);
121: #if defined(PETSC_USE_COMPLEX)
122: *useconj = (ctx->isreal && PetscImaginaryPart(center) == 0.0)? PETSC_TRUE: PETSC_FALSE;
123: #endif
124: } else {
125: PetscObjectTypeCompare((PetscObject)eps->rg,RGINTERVAL,&isinterval);
126: if (isinterval) {
127: RGIntervalGetEndpoints(eps->rg,NULL,NULL,&c,&d);
128: #if defined(PETSC_USE_COMPLEX)
129: *useconj = (ctx->isreal && c==d)? PETSC_TRUE: PETSC_FALSE;
130: #endif
131: }
132: }
133: return(0);
134: }
136: /* create PetscSubcomm object and determine num_solve_point (depends on npart, N, useconj) */
137: PETSC_STATIC_INLINE PetscErrorCode EPSCISSSetUpSubComm(EPS eps,PetscInt *num_solve_point)
138: {
140: EPS_CISS *ctx = (EPS_CISS*)eps->data;
141: PetscInt N = ctx->N;
144: PetscSubcommCreate(PetscObjectComm((PetscObject)eps),&ctx->subcomm);
145: PetscSubcommSetNumber(ctx->subcomm,ctx->npart);
146: PetscSubcommSetType(ctx->subcomm,PETSC_SUBCOMM_INTERLACED);
147: PetscLogObjectMemory((PetscObject)eps,sizeof(PetscSubcomm));
148: ctx->subcomm_id = ctx->subcomm->color;
149: EPSCISSSetUseConj(eps,&ctx->useconj);
150: if (ctx->useconj) N = N/2;
151: *num_solve_point = N / ctx->npart;
152: if (N%ctx->npart > ctx->subcomm_id) (*num_solve_point)++;
153: return(0);
154: }
156: static PetscErrorCode CISSRedundantMat(EPS eps)
157: {
159: EPS_CISS *ctx = (EPS_CISS*)eps->data;
160: Mat A,B;
161: PetscInt nmat;
164: STGetNumMatrices(eps->st,&nmat);
165: if (ctx->subcomm->n != 1) {
166: STGetMatrix(eps->st,0,&A);
167: MatDestroy(&ctx->pA);
168: MatCreateRedundantMatrix(A,ctx->subcomm->n,PetscSubcommChild(ctx->subcomm),MAT_INITIAL_MATRIX,&ctx->pA);
169: if (nmat>1) {
170: STGetMatrix(eps->st,1,&B);
171: MatDestroy(&ctx->pB);
172: MatCreateRedundantMatrix(B,ctx->subcomm->n,PetscSubcommChild(ctx->subcomm),MAT_INITIAL_MATRIX,&ctx->pB);
173: } else ctx->pB = NULL;
174: } else {
175: ctx->pA = NULL;
176: ctx->pB = NULL;
177: }
178: return(0);
179: }
181: static PetscErrorCode CISSScatterVec(EPS eps)
182: {
184: EPS_CISS *ctx = (EPS_CISS*)eps->data;
185: IS is1,is2;
186: Vec v0;
187: PetscInt i,j,k,mstart,mend,mlocal;
188: PetscInt *idx1,*idx2,mloc_sub;
191: VecDestroy(&ctx->xsub);
192: MatCreateVecs(ctx->pA,&ctx->xsub,NULL);
194: VecDestroy(&ctx->xdup);
195: MatGetLocalSize(ctx->pA,&mloc_sub,NULL);
196: VecCreateMPI(PetscSubcommContiguousParent(ctx->subcomm),mloc_sub,PETSC_DECIDE,&ctx->xdup);
198: VecScatterDestroy(&ctx->scatterin);
199: BVGetColumn(ctx->V,0,&v0);
200: VecGetOwnershipRange(v0,&mstart,&mend);
201: mlocal = mend - mstart;
202: PetscMalloc2(ctx->subcomm->n*mlocal,&idx1,ctx->subcomm->n*mlocal,&idx2);
203: j = 0;
204: for (k=0;k<ctx->subcomm->n;k++) {
205: for (i=mstart;i<mend;i++) {
206: idx1[j] = i;
207: idx2[j++] = i + eps->n*k;
208: }
209: }
210: ISCreateGeneral(PetscObjectComm((PetscObject)eps),ctx->subcomm->n*mlocal,idx1,PETSC_COPY_VALUES,&is1);
211: ISCreateGeneral(PetscObjectComm((PetscObject)eps),ctx->subcomm->n*mlocal,idx2,PETSC_COPY_VALUES,&is2);
212: VecScatterCreate(v0,is1,ctx->xdup,is2,&ctx->scatterin);
213: ISDestroy(&is1);
214: ISDestroy(&is2);
215: PetscFree2(idx1,idx2);
216: BVRestoreColumn(ctx->V,0,&v0);
217: return(0);
218: }
220: static PetscErrorCode SetPathParameter(EPS eps)
221: {
223: EPS_CISS *ctx = (EPS_CISS*)eps->data;
224: PetscInt i,j;
225: PetscScalar center=0.0,tmp,tmp2,*omegai;
226: PetscReal theta,radius=1.0,vscale,a,b,c,d,max_w=0.0,rgscale;
227: #if defined(PETSC_USE_COMPLEX)
228: PetscReal start_ang,end_ang;
229: #endif
230: PetscBool isring=PETSC_FALSE,isellipse=PETSC_FALSE,isinterval=PETSC_FALSE;
233: PetscObjectTypeCompare((PetscObject)eps->rg,RGELLIPSE,&isellipse);
234: PetscObjectTypeCompare((PetscObject)eps->rg,RGRING,&isring);
235: PetscObjectTypeCompare((PetscObject)eps->rg,RGINTERVAL,&isinterval);
236: RGGetScale(eps->rg,&rgscale);
237: PetscMalloc1(ctx->N+1l,&omegai);
238: RGComputeContour(eps->rg,ctx->N,ctx->omega,omegai);
239: if (isellipse) {
240: RGEllipseGetParameters(eps->rg,¢er,&radius,&vscale);
241: for (i=0;i<ctx->N;i++) {
242: #if defined(PETSC_USE_COMPLEX)
243: theta = 2.0*PETSC_PI*(i+0.5)/ctx->N;
244: ctx->pp[i] = PetscCMPLX(PetscCosReal(theta),vscale*PetscSinReal(theta));
245: ctx->weight[i] = rgscale*radius*(PetscCMPLX(vscale*PetscCosReal(theta),PetscSinReal(theta)))/(PetscReal)ctx->N;
246: #else
247: theta = (PETSC_PI/ctx->N)*(i+0.5);
248: ctx->pp[i] = PetscCosReal(theta);
249: ctx->weight[i] = PetscCosReal((ctx->N-1)*theta)/ctx->N;
250: ctx->omega[i] = rgscale*(center + radius*ctx->pp[i]);
251: #endif
252: }
253: } else if (ctx->quad == EPS_CISS_QUADRULE_CHEBYSHEV) {
254: for (i=0;i<ctx->N;i++) {
255: theta = (PETSC_PI/ctx->N)*(i+0.5);
256: ctx->pp[i] = PetscCosReal(theta);
257: ctx->weight[i] = PetscCosReal((ctx->N-1)*theta)/ctx->N;
258: }
259: if (isinterval) {
260: RGIntervalGetEndpoints(eps->rg,&a,&b,&c,&d);
261: if ((c!=d || c!=0.0) && (a!=b || a!=0.0)) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_SUP,"Endpoints of the imaginary axis or the real axis must be both zero");
262: for (i=0;i<ctx->N;i++) {
263: if (c==d) ctx->omega[i] = ((b-a)*(ctx->pp[i]+1.0)/2.0+a)*rgscale;
264: if (a==b) {
265: #if defined(PETSC_USE_COMPLEX)
266: ctx->omega[i] = ((d-c)*(ctx->pp[i]+1.0)/2.0+c)*rgscale*PETSC_i;
267: #else
268: SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Integration points on a vertical line require complex arithmetic");
269: #endif
270: }
271: }
272: }
273: if (isring) { /* only supported in complex scalars */
274: #if defined(PETSC_USE_COMPLEX)
275: RGRingGetParameters(eps->rg,¢er,&radius,&vscale,&start_ang,&end_ang,NULL);
276: for (i=0;i<ctx->N;i++) {
277: theta = (start_ang*2.0+(end_ang-start_ang)*(PetscRealPart(ctx->pp[i])+1.0))*PETSC_PI;
278: ctx->omega[i] = rgscale*(center + radius*PetscCMPLX(PetscCosReal(theta),vscale*PetscSinReal(theta)));
279: }
280: #endif
281: }
282: } else {
283: if (isinterval) {
284: RGIntervalGetEndpoints(eps->rg,&a,&b,&c,&d);
285: center = rgscale*((b+a)/2.0+(d+c)/2.0*PETSC_PI);
286: radius = PetscSqrtReal(PetscPowRealInt(rgscale*(b-a)/2.0,2)+PetscPowRealInt(rgscale*(d-c)/2.0,2));
287: } else if (isring) {
288: RGRingGetParameters(eps->rg,¢er,&radius,NULL,NULL,NULL,NULL);
289: center *= rgscale;
290: radius *= rgscale;
291: }
292: for (i=0;i<ctx->N;i++) {
293: ctx->pp[i] = (ctx->omega[i]-center)/radius;
294: tmp = 1; tmp2 = 1;
295: for (j=0;j<ctx->N;j++) {
296: tmp *= ctx->omega[j];
297: if (i != j) tmp2 *= ctx->omega[j]-ctx->omega[i];
298: }
299: ctx->weight[i] = tmp/tmp2;
300: max_w = PetscMax(PetscAbsScalar(ctx->weight[i]),max_w);
301: }
302: for (i=0;i<ctx->N;i++) ctx->weight[i] /= (PetscScalar)max_w;
303: }
304: PetscFree(omegai);
305: return(0);
306: }
308: static PetscErrorCode CISSVecSetRandom(BV V,PetscInt i0,PetscInt i1)
309: {
311: PetscInt i,j,nlocal;
312: PetscScalar *vdata;
313: Vec x;
316: BVGetSizes(V,&nlocal,NULL,NULL);
317: for (i=i0;i<i1;i++) {
318: BVSetRandomColumn(V,i);
319: BVGetColumn(V,i,&x);
320: VecGetArray(x,&vdata);
321: for (j=0;j<nlocal;j++) {
322: vdata[j] = PetscRealPart(vdata[j]);
323: if (PetscRealPart(vdata[j]) < 0.5) vdata[j] = -1.0;
324: else vdata[j] = 1.0;
325: }
326: VecRestoreArray(x,&vdata);
327: BVRestoreColumn(V,i,&x);
328: }
329: return(0);
330: }
332: static PetscErrorCode VecScatterVecs(EPS eps,BV Vin,PetscInt n)
333: {
334: PetscErrorCode ierr;
335: EPS_CISS *ctx = (EPS_CISS*)eps->data;
336: PetscInt i;
337: Vec vi,pvi;
338: const PetscScalar *array;
341: for (i=0;i<n;i++) {
342: BVGetColumn(Vin,i,&vi);
343: VecScatterBegin(ctx->scatterin,vi,ctx->xdup,INSERT_VALUES,SCATTER_FORWARD);
344: VecScatterEnd(ctx->scatterin,vi,ctx->xdup,INSERT_VALUES,SCATTER_FORWARD);
345: BVRestoreColumn(Vin,i,&vi);
346: VecGetArrayRead(ctx->xdup,&array);
347: VecPlaceArray(ctx->xsub,array);
348: BVGetColumn(ctx->pV,i,&pvi);
349: VecCopy(ctx->xsub,pvi);
350: BVRestoreColumn(ctx->pV,i,&pvi);
351: VecResetArray(ctx->xsub);
352: VecRestoreArrayRead(ctx->xdup,&array);
353: }
354: return(0);
355: }
357: static PetscErrorCode SolveLinearSystem(EPS eps,Mat A,Mat B,BV V,PetscInt L_start,PetscInt L_end,PetscBool initksp)
358: {
360: EPS_CISS *ctx = (EPS_CISS*)eps->data;
361: PetscInt i,p_id;
362: Mat Fz,kspMat,MV,BMV=NULL,MC;
363: KSP ksp;
366: if (!ctx->ksp) { EPSCISSGetKSPs(eps,&ctx->num_solve_point,&ctx->ksp); }
367: if (ctx->usest) {
368: MatDuplicate(A,MAT_DO_NOT_COPY_VALUES,&Fz);
369: }
370: BVSetActiveColumns(V,L_start,L_end);
371: BVGetMat(V,&MV);
372: if (B) {
373: MatProductCreate(B,MV,NULL,&BMV);
374: MatProductSetType(BMV,MATPRODUCT_AB);
375: MatProductSetFromOptions(BMV);
376: MatProductSymbolic(BMV);
377: }
378: for (i=0;i<ctx->num_solve_point;i++) {
379: p_id = i*ctx->subcomm->n + ctx->subcomm_id;
380: if (!ctx->usest && initksp) {
381: MatDuplicate(A,MAT_COPY_VALUES,&kspMat);
382: if (B) {
383: MatAXPY(kspMat,-ctx->omega[p_id],B,DIFFERENT_NONZERO_PATTERN);
384: } else {
385: MatShift(kspMat,-ctx->omega[p_id]);
386: }
387: KSPSetOperators(ctx->ksp[i],kspMat,kspMat);
388: MatDestroy(&kspMat);
389: } else if (ctx->usest) {
390: STSetShift(eps->st,ctx->omega[p_id]);
391: STGetKSP(eps->st,&ksp);
392: }
393: BVSetActiveColumns(ctx->Y,i*ctx->L_max+L_start,i*ctx->L_max+L_end);
394: BVGetMat(ctx->Y,&MC);
395: if (B) {
396: MatProductNumeric(BMV);
397: if (ctx->usest) {
398: KSPMatSolve(ksp,BMV,MC);
399: } else {
400: KSPMatSolve(ctx->ksp[i],BMV,MC);
401: }
402: } else {
403: if (ctx->usest) {
404: KSPMatSolve(ksp,MV,MC);
405: } else {
406: KSPMatSolve(ctx->ksp[i],MV,MC);
407: }
408: }
409: if (ctx->usest && i<ctx->num_solve_point-1) { KSPReset(ksp); }
410: BVRestoreMat(ctx->Y,&MC);
411: }
412: MatDestroy(&BMV);
413: BVRestoreMat(V,&MV);
414: if (ctx->usest) { MatDestroy(&Fz); }
415: return(0);
416: }
418: #if defined(PETSC_USE_COMPLEX)
419: static PetscErrorCode EstimateNumberEigs(EPS eps,PetscInt *L_add)
420: {
422: EPS_CISS *ctx = (EPS_CISS*)eps->data;
423: PetscInt i,j,p_id;
424: PetscScalar tmp,m = 1,sum = 0.0;
425: PetscReal eta;
426: Vec v,vtemp,vj;
429: BVCreateVec(ctx->Y,&v);
430: BVCreateVec(ctx->V,&vtemp);
431: for (j=0;j<ctx->L;j++) {
432: VecSet(v,0);
433: for (i=0;i<ctx->num_solve_point; i++) {
434: p_id = i*ctx->subcomm->n + ctx->subcomm_id;
435: BVSetActiveColumns(ctx->Y,i*ctx->L_max+j,i*ctx->L_max+j+1);
436: BVMultVec(ctx->Y,ctx->weight[p_id],1,v,&m);
437: }
438: BVGetColumn(ctx->V,j,&vj);
439: if (ctx->pA) {
440: VecSet(vtemp,0);
441: VecScatterBegin(ctx->scatterin,v,vtemp,ADD_VALUES,SCATTER_REVERSE);
442: VecScatterEnd(ctx->scatterin,v,vtemp,ADD_VALUES,SCATTER_REVERSE);
443: VecDot(vj,vtemp,&tmp);
444: } else {
445: VecDot(vj,v,&tmp);
446: }
447: BVRestoreColumn(ctx->V,j,&vj);
448: if (ctx->useconj) sum += PetscRealPart(tmp)*2;
449: else sum += tmp;
450: }
451: ctx->est_eig = PetscAbsScalar(sum/(PetscReal)ctx->L);
452: eta = PetscPowReal(10.0,-PetscLog10Real(eps->tol)/ctx->N);
453: PetscInfo1(eps,"Estimation_#Eig %f\n",(double)ctx->est_eig);
454: *L_add = (PetscInt)PetscCeilReal((ctx->est_eig*eta)/ctx->M) - ctx->L;
455: if (*L_add < 0) *L_add = 0;
456: if (*L_add>ctx->L_max-ctx->L) {
457: PetscInfo(eps,"Number of eigenvalues around the contour path may be too large\n");
458: *L_add = ctx->L_max-ctx->L;
459: }
460: VecDestroy(&v);
461: VecDestroy(&vtemp);
462: return(0);
463: }
464: #endif
466: static PetscErrorCode CalcMu(EPS eps,PetscScalar *Mu)
467: {
469: PetscMPIInt sub_size,len;
470: PetscInt i,j,k,s;
471: PetscScalar *m,*temp,*temp2,*ppk,alp;
472: EPS_CISS *ctx = (EPS_CISS*)eps->data;
473: Mat M;
476: MPI_Comm_size(PetscSubcommChild(ctx->subcomm),&sub_size);
477: PetscMalloc3(ctx->num_solve_point*ctx->L*(ctx->L+1),&temp,2*ctx->M*ctx->L*ctx->L,&temp2,ctx->num_solve_point,&ppk);
478: MatCreateSeqDense(PETSC_COMM_SELF,ctx->L,ctx->L_max*ctx->num_solve_point,NULL,&M);
479: for (i=0;i<2*ctx->M*ctx->L*ctx->L;i++) temp2[i] = 0;
480: BVSetActiveColumns(ctx->Y,0,ctx->L_max*ctx->num_solve_point);
481: if (ctx->pA) {
482: BVSetActiveColumns(ctx->pV,0,ctx->L);
483: BVDot(ctx->Y,ctx->pV,M);
484: } else {
485: BVSetActiveColumns(ctx->V,0,ctx->L);
486: BVDot(ctx->Y,ctx->V,M);
487: }
488: MatDenseGetArray(M,&m);
489: for (i=0;i<ctx->num_solve_point;i++) {
490: for (j=0;j<ctx->L;j++) {
491: for (k=0;k<ctx->L;k++) {
492: temp[k+j*ctx->L+i*ctx->L*ctx->L]=m[k+j*ctx->L+i*ctx->L*ctx->L_max];
493: }
494: }
495: }
496: MatDenseRestoreArray(M,&m);
497: for (i=0;i<ctx->num_solve_point;i++) ppk[i] = 1;
498: for (k=0;k<2*ctx->M;k++) {
499: for (j=0;j<ctx->L;j++) {
500: for (i=0;i<ctx->num_solve_point;i++) {
501: alp = ppk[i]*ctx->weight[i*ctx->subcomm->n + ctx->subcomm_id];
502: for (s=0;s<ctx->L;s++) {
503: if (ctx->useconj) temp2[s+(j+k*ctx->L)*ctx->L] += PetscRealPart(alp*temp[s+(j+i*ctx->L)*ctx->L])*2;
504: else temp2[s+(j+k*ctx->L)*ctx->L] += alp*temp[s+(j+i*ctx->L)*ctx->L];
505: }
506: }
507: }
508: for (i=0;i<ctx->num_solve_point;i++)
509: ppk[i] *= ctx->pp[i*ctx->subcomm->n + ctx->subcomm_id];
510: }
511: for (i=0;i<2*ctx->M*ctx->L*ctx->L;i++) temp2[i] /= sub_size;
512: PetscMPIIntCast(2*ctx->M*ctx->L*ctx->L,&len);
513: MPIU_Allreduce(temp2,Mu,len,MPIU_SCALAR,MPIU_SUM,PetscObjectComm((PetscObject)eps));
514: PetscFree3(temp,temp2,ppk);
515: MatDestroy(&M);
516: return(0);
517: }
519: static PetscErrorCode BlockHankel(EPS eps,PetscScalar *Mu,PetscInt s,PetscScalar *H)
520: {
521: EPS_CISS *ctx = (EPS_CISS*)eps->data;
522: PetscInt i,j,k,L=ctx->L,M=ctx->M;
525: for (k=0;k<L*M;k++)
526: for (j=0;j<M;j++)
527: for (i=0;i<L;i++)
528: H[j*L+i+k*L*M] = Mu[i+k*L+(j+s)*L*L];
529: return(0);
530: }
532: static PetscErrorCode SVD_H0(EPS eps,PetscScalar *S,PetscInt *K)
533: {
535: EPS_CISS *ctx = (EPS_CISS*)eps->data;
536: PetscInt i,ml=ctx->L*ctx->M;
537: PetscBLASInt m,n,lda,ldu,ldvt,lwork,info;
538: PetscScalar *work;
539: #if defined(PETSC_USE_COMPLEX)
540: PetscReal *rwork;
541: #endif
544: PetscLogEventBegin(EPS_CISS_SVD,eps,0,0,0);
545: PetscMalloc1(5*ml,&work);
546: #if defined(PETSC_USE_COMPLEX)
547: PetscMalloc1(5*ml,&rwork);
548: #endif
549: PetscBLASIntCast(ml,&m);
550: n = m; lda = m; ldu = m; ldvt = m; lwork = 5*m;
551: PetscFPTrapPush(PETSC_FP_TRAP_OFF);
552: #if defined(PETSC_USE_COMPLEX)
553: PetscStackCallBLAS("LAPACKgesvd",LAPACKgesvd_("N","N",&m,&n,S,&lda,ctx->sigma,NULL,&ldu,NULL,&ldvt,work,&lwork,rwork,&info));
554: #else
555: PetscStackCallBLAS("LAPACKgesvd",LAPACKgesvd_("N","N",&m,&n,S,&lda,ctx->sigma,NULL,&ldu,NULL,&ldvt,work,&lwork,&info));
556: #endif
557: SlepcCheckLapackInfo("gesvd",info);
558: PetscFPTrapPop();
559: (*K) = 0;
560: for (i=0;i<ml;i++) {
561: if (ctx->sigma[i]/PetscMax(ctx->sigma[0],1)>ctx->delta) (*K)++;
562: }
563: PetscFree(work);
564: #if defined(PETSC_USE_COMPLEX)
565: PetscFree(rwork);
566: #endif
567: PetscLogEventEnd(EPS_CISS_SVD,eps,0,0,0);
568: return(0);
569: }
571: static PetscErrorCode ConstructS(EPS eps)
572: {
574: EPS_CISS *ctx = (EPS_CISS*)eps->data;
575: PetscInt i,j,k,vec_local_size,p_id;
576: Vec v,sj;
577: PetscScalar *ppk, *v_data, m = 1;
580: BVGetSizes(ctx->Y,&vec_local_size,NULL,NULL);
581: PetscMalloc1(ctx->num_solve_point,&ppk);
582: for (i=0;i<ctx->num_solve_point;i++) ppk[i] = 1;
583: BVCreateVec(ctx->Y,&v);
584: for (k=0;k<ctx->M;k++) {
585: for (j=0;j<ctx->L;j++) {
586: VecSet(v,0);
587: for (i=0;i<ctx->num_solve_point;i++) {
588: p_id = i*ctx->subcomm->n + ctx->subcomm_id;
589: BVSetActiveColumns(ctx->Y,i*ctx->L_max+j,i*ctx->L_max+j+1);
590: BVMultVec(ctx->Y,ppk[i]*ctx->weight[p_id],1.0,v,&m);
591: }
592: if (ctx->useconj) {
593: VecGetArray(v,&v_data);
594: for (i=0;i<vec_local_size;i++) v_data[i] = PetscRealPart(v_data[i])*2;
595: VecRestoreArray(v,&v_data);
596: }
597: BVGetColumn(ctx->S,k*ctx->L+j,&sj);
598: if (ctx->pA) {
599: VecSet(sj,0);
600: VecScatterBegin(ctx->scatterin,v,sj,ADD_VALUES,SCATTER_REVERSE);
601: VecScatterEnd(ctx->scatterin,v,sj,ADD_VALUES,SCATTER_REVERSE);
602: } else {
603: VecCopy(v,sj);
604: }
605: BVRestoreColumn(ctx->S,k*ctx->L+j,&sj);
606: }
607: for (i=0;i<ctx->num_solve_point;i++) {
608: p_id = i*ctx->subcomm->n + ctx->subcomm_id;
609: ppk[i] *= ctx->pp[p_id];
610: }
611: }
612: PetscFree(ppk);
613: VecDestroy(&v);
614: return(0);
615: }
617: static PetscErrorCode SVD_S(BV S,PetscInt ml,PetscReal delta,PetscReal *sigma,PetscInt *K)
618: {
620: PetscInt i,j,k,local_size;
621: PetscMPIInt len;
622: PetscScalar *work,*temp,*B,*tempB,*s_data,*Q1,*Q2,*temp2,alpha=1,beta=0;
623: PetscBLASInt l,m,n,lda,ldu,ldvt,lwork,info,ldb,ldc;
624: #if defined(PETSC_USE_COMPLEX)
625: PetscReal *rwork;
626: #endif
629: BVGetSizes(S,&local_size,NULL,NULL);
630: BVGetArray(S,&s_data);
631: PetscMalloc7(ml*ml,&temp,ml*ml,&temp2,local_size*ml,&Q1,local_size*ml,&Q2,ml*ml,&B,ml*ml,&tempB,5*ml,&work);
632: PetscArrayzero(B,ml*ml);
633: #if defined(PETSC_USE_COMPLEX)
634: PetscMalloc1(5*ml,&rwork);
635: #endif
636: PetscFPTrapPush(PETSC_FP_TRAP_OFF);
638: for (i=0;i<ml;i++) B[i*ml+i]=1;
640: for (k=0;k<2;k++) {
641: PetscBLASIntCast(local_size,&m);
642: PetscBLASIntCast(ml,&l);
643: n = l; lda = m; ldb = m; ldc = l;
644: if (k == 0) {
645: PetscStackCallBLAS("BLASgemm",BLASgemm_("C","N",&l,&n,&m,&alpha,s_data,&lda,s_data,&ldb,&beta,temp,&ldc));
646: } else if ((k%2)==1) {
647: PetscStackCallBLAS("BLASgemm",BLASgemm_("C","N",&l,&n,&m,&alpha,Q1,&lda,Q1,&ldb,&beta,temp,&ldc));
648: } else {
649: PetscStackCallBLAS("BLASgemm",BLASgemm_("C","N",&l,&n,&m,&alpha,Q2,&lda,Q2,&ldb,&beta,temp,&ldc));
650: }
651: PetscArrayzero(temp2,ml*ml);
652: PetscMPIIntCast(ml*ml,&len);
653: MPIU_Allreduce(temp,temp2,len,MPIU_SCALAR,MPIU_SUM,PetscObjectComm((PetscObject)S));
655: PetscBLASIntCast(ml,&m);
656: n = m; lda = m; lwork = 5*m, ldu = 1; ldvt = 1;
657: #if defined(PETSC_USE_COMPLEX)
658: PetscStackCallBLAS("LAPACKgesvd",LAPACKgesvd_("O","N",&m,&n,temp2,&lda,sigma,NULL,&ldu,NULL,&ldvt,work,&lwork,rwork,&info));
659: #else
660: PetscStackCallBLAS("LAPACKgesvd",LAPACKgesvd_("O","N",&m,&n,temp2,&lda,sigma,NULL,&ldu,NULL,&ldvt,work,&lwork,&info));
661: #endif
662: SlepcCheckLapackInfo("gesvd",info);
664: PetscBLASIntCast(local_size,&l);
665: PetscBLASIntCast(ml,&n);
666: m = n; lda = l; ldb = m; ldc = l;
667: if (k==0) {
668: PetscStackCallBLAS("BLASgemm",BLASgemm_("N","N",&l,&n,&m,&alpha,s_data,&lda,temp2,&ldb,&beta,Q1,&ldc));
669: } else if ((k%2)==1) {
670: PetscStackCallBLAS("BLASgemm",BLASgemm_("N","N",&l,&n,&m,&alpha,Q1,&lda,temp2,&ldb,&beta,Q2,&ldc));
671: } else {
672: PetscStackCallBLAS("BLASgemm",BLASgemm_("N","N",&l,&n,&m,&alpha,Q2,&lda,temp2,&ldb,&beta,Q1,&ldc));
673: }
675: PetscBLASIntCast(ml,&l);
676: m = l; n = l; lda = l; ldb = m; ldc = l;
677: PetscStackCallBLAS("BLASgemm",BLASgemm_("N","N",&l,&n,&m,&alpha,B,&lda,temp2,&ldb,&beta,tempB,&ldc));
678: for (i=0;i<ml;i++) {
679: sigma[i] = sqrt(sigma[i]);
680: for (j=0;j<local_size;j++) {
681: if ((k%2)==1) Q2[j+i*local_size]/=sigma[i];
682: else Q1[j+i*local_size]/=sigma[i];
683: }
684: for (j=0;j<ml;j++) {
685: B[j+i*ml]=tempB[j+i*ml]*sigma[i];
686: }
687: }
688: }
690: PetscBLASIntCast(ml,&m);
691: n = m; lda = m; ldu=1; ldvt=1;
692: #if defined(PETSC_USE_COMPLEX)
693: PetscStackCallBLAS("LAPACKgesvd",LAPACKgesvd_("N","O",&m,&n,B,&lda,sigma,NULL,&ldu,NULL,&ldvt,work,&lwork,rwork,&info));
694: #else
695: PetscStackCallBLAS("LAPACKgesvd",LAPACKgesvd_("N","O",&m,&n,B,&lda,sigma,NULL,&ldu,NULL,&ldvt,work,&lwork,&info));
696: #endif
697: SlepcCheckLapackInfo("gesvd",info);
699: PetscBLASIntCast(local_size,&l);
700: PetscBLASIntCast(ml,&n);
701: m = n; lda = l; ldb = m; ldc = l;
702: if ((k%2)==1) {
703: PetscStackCallBLAS("BLASgemm",BLASgemm_("N","T",&l,&n,&m,&alpha,Q1,&lda,B,&ldb,&beta,s_data,&ldc));
704: } else {
705: PetscStackCallBLAS("BLASgemm",BLASgemm_("N","T",&l,&n,&m,&alpha,Q2,&lda,B,&ldb,&beta,s_data,&ldc));
706: }
708: PetscFPTrapPop();
709: BVRestoreArray(S,&s_data);
711: (*K) = 0;
712: for (i=0;i<ml;i++) {
713: if (sigma[i]/PetscMax(sigma[0],1)>delta) (*K)++;
714: }
715: PetscFree7(temp,temp2,Q1,Q2,B,tempB,work);
716: #if defined(PETSC_USE_COMPLEX)
717: PetscFree(rwork);
718: #endif
719: return(0);
720: }
722: static PetscErrorCode isGhost(EPS eps,PetscInt ld,PetscInt nv,PetscBool *fl)
723: {
725: EPS_CISS *ctx = (EPS_CISS*)eps->data;
726: PetscInt i,j;
727: PetscScalar *pX;
728: PetscReal *tau,s1,s2,tau_max=0.0;
731: PetscMalloc1(nv,&tau);
732: DSVectors(eps->ds,DS_MAT_X,NULL,NULL);
733: DSGetArray(eps->ds,DS_MAT_X,&pX);
735: for (i=0;i<nv;i++) {
736: s1 = 0;
737: s2 = 0;
738: for (j=0;j<nv;j++) {
739: s1 += PetscAbsScalar(PetscPowScalarInt(pX[i*ld+j],2));
740: s2 += PetscPowRealInt(PetscAbsScalar(pX[i*ld+j]),2)/ctx->sigma[j];
741: }
742: tau[i] = s1/s2;
743: tau_max = PetscMax(tau_max,tau[i]);
744: }
745: DSRestoreArray(eps->ds,DS_MAT_X,&pX);
746: for (i=0;i<nv;i++) {
747: tau[i] /= tau_max;
748: }
749: for (i=0;i<nv;i++) {
750: if (tau[i]>=ctx->spurious_threshold) fl[i] = PETSC_TRUE;
751: else fl[i] = PETSC_FALSE;
752: }
753: PetscFree(tau);
754: return(0);
755: }
757: static PetscErrorCode rescale_eig(EPS eps,PetscInt nv)
758: {
760: EPS_CISS *ctx = (EPS_CISS*)eps->data;
761: PetscInt i;
762: PetscScalar center;
763: PetscReal radius,a,b,c,d,rgscale;
764: #if defined(PETSC_USE_COMPLEX)
765: PetscReal start_ang,end_ang,vscale,theta;
766: #endif
767: PetscBool isring,isellipse,isinterval;
770: PetscObjectTypeCompare((PetscObject)eps->rg,RGELLIPSE,&isellipse);
771: PetscObjectTypeCompare((PetscObject)eps->rg,RGRING,&isring);
772: PetscObjectTypeCompare((PetscObject)eps->rg,RGINTERVAL,&isinterval);
773: RGGetScale(eps->rg,&rgscale);
774: if (isinterval) {
775: RGIntervalGetEndpoints(eps->rg,NULL,NULL,&c,&d);
776: if (c==d) {
777: for (i=0;i<nv;i++) {
778: #if defined(PETSC_USE_COMPLEX)
779: eps->eigr[i] = PetscRealPart(eps->eigr[i]);
780: #else
781: eps->eigi[i] = 0;
782: #endif
783: }
784: }
785: }
786: if (ctx->extraction == EPS_CISS_EXTRACTION_HANKEL) {
787: if (isellipse) {
788: RGEllipseGetParameters(eps->rg,¢er,&radius,NULL);
789: for (i=0;i<nv;i++) eps->eigr[i] = rgscale*(center + radius*eps->eigr[i]);
790: } else if (isinterval) {
791: RGIntervalGetEndpoints(eps->rg,&a,&b,&c,&d);
792: if (ctx->quad == EPS_CISS_QUADRULE_CHEBYSHEV) {
793: for (i=0;i<nv;i++) {
794: if (c==d) eps->eigr[i] = ((b-a)*(eps->eigr[i]+1.0)/2.0+a)*rgscale;
795: if (a==b) {
796: #if defined(PETSC_USE_COMPLEX)
797: eps->eigr[i] = ((d-c)*(eps->eigr[i]+1.0)/2.0+c)*rgscale*PETSC_i;
798: #else
799: SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Integration points on a vertical line require complex arithmetic");
800: #endif
801: }
802: }
803: } else {
804: center = (b+a)/2.0+(d+c)/2.0*PETSC_PI;
805: radius = PetscSqrtReal(PetscPowRealInt((b-a)/2.0,2)+PetscPowRealInt((d-c)/2.0,2));
806: for (i=0;i<nv;i++) eps->eigr[i] = center + radius*eps->eigr[i];
807: }
808: } else if (isring) { /* only supported in complex scalars */
809: #if defined(PETSC_USE_COMPLEX)
810: RGRingGetParameters(eps->rg,¢er,&radius,&vscale,&start_ang,&end_ang,NULL);
811: if (ctx->quad == EPS_CISS_QUADRULE_CHEBYSHEV) {
812: for (i=0;i<nv;i++) {
813: theta = (start_ang*2.0+(end_ang-start_ang)*(PetscRealPart(eps->eigr[i])+1.0))*PETSC_PI;
814: eps->eigr[i] = rgscale*center + (rgscale*radius+PetscImaginaryPart(eps->eigr[i]))*PetscCMPLX(PetscCosReal(theta),vscale*PetscSinReal(theta));
815: }
816: } else {
817: for (i=0;i<nv;i++) eps->eigr[i] = rgscale*(center + radius*eps->eigr[i]);
818: }
819: #endif
820: }
821: }
822: return(0);
823: }
825: PetscErrorCode EPSSetUp_CISS(EPS eps)
826: {
828: EPS_CISS *ctx = (EPS_CISS*)eps->data;
829: PetscBool istrivial,isring,isellipse,isinterval,flg,useconj;
830: PetscReal c,d;
831: Mat A;
834: if (eps->ncv==PETSC_DEFAULT) {
835: eps->ncv = ctx->L_max*ctx->M;
836: if (eps->ncv>eps->n) {
837: eps->ncv = eps->n;
838: ctx->L_max = eps->ncv/ctx->M;
839: if (!ctx->L_max) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_SUP,"Cannot adjust solver parameters, try setting a smaller value of M (moment size)");
840: }
841: } else {
842: EPSSetDimensions_Default(eps,eps->nev,&eps->ncv,&eps->mpd);
843: ctx->L_max = eps->ncv/ctx->M;
844: if (!ctx->L_max) {
845: ctx->L_max = 1;
846: eps->ncv = ctx->L_max*ctx->M;
847: }
848: }
849: ctx->L = PetscMin(ctx->L,ctx->L_max);
850: if (eps->max_it==PETSC_DEFAULT) eps->max_it = 1;
851: if (eps->mpd==PETSC_DEFAULT) eps->mpd = eps->ncv;
852: if (!eps->which) eps->which = EPS_ALL;
853: if (eps->which!=EPS_ALL) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_SUP,"This solver supports only computing all eigenvalues");
854: EPSCheckUnsupported(eps,EPS_FEATURE_BALANCE | EPS_FEATURE_ARBITRARY | EPS_FEATURE_EXTRACTION | EPS_FEATURE_STOPPING | EPS_FEATURE_TWOSIDED);
856: /* check region */
857: RGIsTrivial(eps->rg,&istrivial);
858: if (istrivial) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_SUP,"CISS requires a nontrivial region, e.g. -rg_type ellipse ...");
859: RGGetComplement(eps->rg,&flg);
860: if (flg) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_SUP,"A region with complement flag set is not allowed");
861: PetscObjectTypeCompare((PetscObject)eps->rg,RGELLIPSE,&isellipse);
862: PetscObjectTypeCompare((PetscObject)eps->rg,RGRING,&isring);
863: PetscObjectTypeCompare((PetscObject)eps->rg,RGINTERVAL,&isinterval);
864: if (!isellipse && !isring && !isinterval) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_SUP,"Currently only implemented for interval, elliptic or ring regions");
865: /* if useconj has changed, then reset subcomm data */
866: EPSCISSSetUseConj(eps,&useconj);
867: if (useconj!=ctx->useconj) { EPSCISSResetSubcomm(eps); }
869: #if !defined(PETSC_USE_COMPLEX)
870: if (isring) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_SUP,"Ring region only supported for complex scalars");
871: #endif
872: if (isinterval) {
873: RGIntervalGetEndpoints(eps->rg,NULL,NULL,&c,&d);
874: #if !defined(PETSC_USE_COMPLEX)
875: if (c!=d || c!=0.0) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_SUP,"In real scalars, endpoints of the imaginary axis must be both zero");
876: #endif
877: if (!ctx->quad && c==d) ctx->quad = EPS_CISS_QUADRULE_CHEBYSHEV;
878: }
879: if (!ctx->quad) ctx->quad = EPS_CISS_QUADRULE_TRAPEZOIDAL;
881: /* create split comm */
882: if (!ctx->subcomm) { EPSCISSSetUpSubComm(eps,&ctx->num_solve_point); }
884: EPSAllocateSolution(eps,0);
885: if (ctx->weight) { PetscFree4(ctx->weight,ctx->omega,ctx->pp,ctx->sigma); }
886: PetscMalloc4(ctx->N,&ctx->weight,ctx->N+1,&ctx->omega,ctx->N,&ctx->pp,ctx->L_max*ctx->M,&ctx->sigma);
887: PetscLogObjectMemory((PetscObject)eps,3*ctx->N*sizeof(PetscScalar)+ctx->L_max*ctx->N*sizeof(PetscReal));
889: /* allocate basis vectors */
890: BVDestroy(&ctx->S);
891: BVDuplicateResize(eps->V,ctx->L_max*ctx->M,&ctx->S);
892: PetscLogObjectParent((PetscObject)eps,(PetscObject)ctx->S);
893: BVDestroy(&ctx->V);
894: BVDuplicateResize(eps->V,ctx->L_max,&ctx->V);
895: PetscLogObjectParent((PetscObject)eps,(PetscObject)ctx->V);
897: STGetMatrix(eps->st,0,&A);
898: PetscObjectTypeCompare((PetscObject)A,MATSHELL,&flg);
899: if (flg) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_SUP,"Matrix type shell is not supported in this solver");
901: if (!ctx->usest_set) ctx->usest = (ctx->npart>1)? PETSC_FALSE: PETSC_TRUE;
902: if (ctx->usest && ctx->npart>1) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_SUP,"The usest flag is not supported when partitions > 1");
904: CISSRedundantMat(eps);
905: if (ctx->pA) {
906: CISSScatterVec(eps);
907: BVDestroy(&ctx->pV);
908: BVCreate(PetscObjectComm((PetscObject)ctx->xsub),&ctx->pV);
909: BVSetSizesFromVec(ctx->pV,ctx->xsub,eps->n);
910: BVSetFromOptions(ctx->pV);
911: BVResize(ctx->pV,ctx->L_max,PETSC_FALSE);
912: PetscLogObjectParent((PetscObject)eps,(PetscObject)ctx->pV);
913: }
915: EPSCheckSinvertCondition(eps,ctx->usest," (with the usest flag set)");
917: BVDestroy(&ctx->Y);
918: if (ctx->pA) {
919: BVCreate(PetscObjectComm((PetscObject)ctx->xsub),&ctx->Y);
920: BVSetSizesFromVec(ctx->Y,ctx->xsub,eps->n);
921: BVSetFromOptions(ctx->Y);
922: BVResize(ctx->Y,ctx->num_solve_point*ctx->L_max,PETSC_FALSE);
923: } else {
924: BVDuplicateResize(eps->V,ctx->num_solve_point*ctx->L_max,&ctx->Y);
925: }
926: PetscLogObjectParent((PetscObject)eps,(PetscObject)ctx->Y);
928: if (ctx->extraction == EPS_CISS_EXTRACTION_HANKEL) {
929: DSSetType(eps->ds,DSGNHEP);
930: } else if (eps->isgeneralized) {
931: if (eps->ishermitian && eps->ispositive) {
932: DSSetType(eps->ds,DSGHEP);
933: } else {
934: DSSetType(eps->ds,DSGNHEP);
935: }
936: } else {
937: if (eps->ishermitian) {
938: DSSetType(eps->ds,DSHEP);
939: } else {
940: DSSetType(eps->ds,DSNHEP);
941: }
942: }
943: DSAllocate(eps->ds,eps->ncv);
944: EPSSetWorkVecs(eps,2);
946: #if !defined(PETSC_USE_COMPLEX)
947: if (!eps->ishermitian) { PetscInfo(eps,"Warning: complex eigenvalues are not calculated exactly without --with-scalar-type=complex in PETSc\n"); }
948: #endif
949: return(0);
950: }
952: PetscErrorCode EPSSetUpSort_CISS(EPS eps)
953: {
955: SlepcSC sc;
958: /* fill sorting criterion context */
959: eps->sc->comparison = SlepcCompareSmallestReal;
960: eps->sc->comparisonctx = NULL;
961: eps->sc->map = NULL;
962: eps->sc->mapobj = NULL;
964: /* fill sorting criterion for DS */
965: DSGetSlepcSC(eps->ds,&sc);
966: sc->comparison = SlepcCompareLargestMagnitude;
967: sc->comparisonctx = NULL;
968: sc->map = NULL;
969: sc->mapobj = NULL;
970: return(0);
971: }
973: PetscErrorCode EPSSolve_CISS(EPS eps)
974: {
976: EPS_CISS *ctx = (EPS_CISS*)eps->data;
977: Mat A,B,X,M,pA,pB;
978: PetscInt i,j,ld,nmat,L_add=0,nv=0,L_base=ctx->L,inner,nlocal,*inside;
979: PetscScalar *Mu,*H0,*H1=NULL,*rr,*temp;
980: PetscReal error,max_error,norm;
981: PetscBool *fl1;
982: Vec si,w[3];
983: PetscRandom rand;
984: #if defined(PETSC_USE_COMPLEX)
985: PetscBool isellipse;
986: #endif
989: w[0] = eps->work[0];
990: w[1] = NULL;
991: w[2] = eps->work[1];
992: VecGetLocalSize(w[0],&nlocal);
993: DSGetLeadingDimension(eps->ds,&ld);
994: STGetNumMatrices(eps->st,&nmat);
995: STGetMatrix(eps->st,0,&A);
996: if (nmat>1) { STGetMatrix(eps->st,1,&B); }
997: else B = NULL;
998: SetPathParameter(eps);
999: CISSVecSetRandom(ctx->V,0,ctx->L);
1000: BVGetRandomContext(ctx->V,&rand);
1002: if (ctx->pA) {
1003: VecScatterVecs(eps,ctx->V,ctx->L);
1004: SolveLinearSystem(eps,ctx->pA,ctx->pB,ctx->pV,0,ctx->L,PETSC_TRUE);
1005: } else {
1006: SolveLinearSystem(eps,A,B,ctx->V,0,ctx->L,PETSC_TRUE);
1007: }
1008: #if defined(PETSC_USE_COMPLEX)
1009: PetscObjectTypeCompare((PetscObject)eps->rg,RGELLIPSE,&isellipse);
1010: if (isellipse) {
1011: EstimateNumberEigs(eps,&L_add);
1012: } else {
1013: L_add = 0;
1014: }
1015: #else
1016: L_add = 0;
1017: #endif
1018: if (L_add>0) {
1019: PetscInfo2(eps,"Changing L %D -> %D by Estimate #Eig\n",ctx->L,ctx->L+L_add);
1020: CISSVecSetRandom(ctx->V,ctx->L,ctx->L+L_add);
1021: if (ctx->pA) {
1022: VecScatterVecs(eps,ctx->V,ctx->L+L_add);
1023: SolveLinearSystem(eps,ctx->pA,ctx->pB,ctx->pV,ctx->L,ctx->L+L_add,PETSC_FALSE);
1024: } else {
1025: SolveLinearSystem(eps,A,B,ctx->V,ctx->L,ctx->L+L_add,PETSC_FALSE);
1026: }
1027: ctx->L += L_add;
1028: }
1029: PetscMalloc2(ctx->L*ctx->L*ctx->M*2,&Mu,ctx->L*ctx->M*ctx->L*ctx->M,&H0);
1030: for (i=0;i<ctx->refine_blocksize;i++) {
1031: CalcMu(eps,Mu);
1032: BlockHankel(eps,Mu,0,H0);
1033: SVD_H0(eps,H0,&nv);
1034: if (ctx->sigma[0]<=ctx->delta || nv < ctx->L*ctx->M || ctx->L == ctx->L_max) break;
1035: L_add = L_base;
1036: if (ctx->L+L_add>ctx->L_max) L_add = ctx->L_max-ctx->L;
1037: PetscInfo2(eps,"Changing L %D -> %D by SVD(H0)\n",ctx->L,ctx->L+L_add);
1038: CISSVecSetRandom(ctx->V,ctx->L,ctx->L+L_add);
1039: if (ctx->pA) {
1040: VecScatterVecs(eps,ctx->V,ctx->L+L_add);
1041: SolveLinearSystem(eps,ctx->pA,ctx->pB,ctx->pV,ctx->L,ctx->L+L_add,PETSC_FALSE);
1042: } else {
1043: SolveLinearSystem(eps,A,B,ctx->V,ctx->L,ctx->L+L_add,PETSC_FALSE);
1044: }
1045: ctx->L += L_add;
1046: }
1047: if (ctx->extraction == EPS_CISS_EXTRACTION_HANKEL) {
1048: PetscMalloc1(ctx->L*ctx->M*ctx->L*ctx->M,&H1);
1049: }
1051: while (eps->reason == EPS_CONVERGED_ITERATING) {
1052: eps->its++;
1053: for (inner=0;inner<=ctx->refine_inner;inner++) {
1054: if (ctx->extraction == EPS_CISS_EXTRACTION_HANKEL) {
1055: CalcMu(eps,Mu);
1056: BlockHankel(eps,Mu,0,H0);
1057: SVD_H0(eps,H0,&nv);
1058: break;
1059: } else {
1060: ConstructS(eps);
1061: BVSetActiveColumns(ctx->S,0,ctx->L);
1062: BVCopy(ctx->S,ctx->V);
1063: PetscLogEventBegin(EPS_CISS_SVD,eps,0,0,0);
1064: SVD_S(ctx->S,ctx->L*ctx->M,ctx->delta,ctx->sigma,&nv);
1065: PetscLogEventEnd(EPS_CISS_SVD,eps,0,0,0);
1066: if (ctx->sigma[0]>ctx->delta && nv==ctx->L*ctx->M && inner!=ctx->refine_inner) {
1067: if (ctx->pA) {
1068: VecScatterVecs(eps,ctx->V,ctx->L);
1069: SolveLinearSystem(eps,ctx->pA,ctx->pB,ctx->pV,0,ctx->L,PETSC_FALSE);
1070: } else {
1071: SolveLinearSystem(eps,A,B,ctx->V,0,ctx->L,PETSC_FALSE);
1072: }
1073: } else break;
1074: }
1075: }
1076: eps->nconv = 0;
1077: if (nv == 0) eps->reason = EPS_CONVERGED_TOL;
1078: else {
1079: DSSetDimensions(eps->ds,nv,0,0,0);
1080: DSSetState(eps->ds,DS_STATE_RAW);
1082: if (ctx->extraction == EPS_CISS_EXTRACTION_HANKEL) {
1083: BlockHankel(eps,Mu,0,H0);
1084: BlockHankel(eps,Mu,1,H1);
1085: DSGetArray(eps->ds,DS_MAT_A,&temp);
1086: for (j=0;j<nv;j++) {
1087: for (i=0;i<nv;i++) {
1088: temp[i+j*ld] = H1[i+j*ctx->L*ctx->M];
1089: }
1090: }
1091: DSRestoreArray(eps->ds,DS_MAT_A,&temp);
1092: DSGetArray(eps->ds,DS_MAT_B,&temp);
1093: for (j=0;j<nv;j++) {
1094: for (i=0;i<nv;i++) {
1095: temp[i+j*ld] = H0[i+j*ctx->L*ctx->M];
1096: }
1097: }
1098: DSRestoreArray(eps->ds,DS_MAT_B,&temp);
1099: } else {
1100: BVSetActiveColumns(ctx->S,0,nv);
1101: DSGetMat(eps->ds,DS_MAT_A,&pA);
1102: MatZeroEntries(pA);
1103: BVMatProject(ctx->S,A,ctx->S,pA);
1104: DSRestoreMat(eps->ds,DS_MAT_A,&pA);
1105: if (B) {
1106: DSGetMat(eps->ds,DS_MAT_B,&pB);
1107: MatZeroEntries(pB);
1108: BVMatProject(ctx->S,B,ctx->S,pB);
1109: DSRestoreMat(eps->ds,DS_MAT_B,&pB);
1110: }
1111: }
1113: DSSolve(eps->ds,eps->eigr,eps->eigi);
1114: DSSynchronize(eps->ds,eps->eigr,eps->eigi);
1116: PetscMalloc3(nv,&fl1,nv,&inside,nv,&rr);
1117: rescale_eig(eps,nv);
1118: isGhost(eps,ld,nv,fl1);
1119: RGCheckInside(eps->rg,nv,eps->eigr,eps->eigi,inside);
1120: for (i=0;i<nv;i++) {
1121: if (fl1[i] && inside[i]>=0) {
1122: rr[i] = 1.0;
1123: eps->nconv++;
1124: } else rr[i] = 0.0;
1125: }
1126: DSSort(eps->ds,eps->eigr,eps->eigi,rr,NULL,&eps->nconv);
1127: DSSynchronize(eps->ds,eps->eigr,eps->eigi);
1128: rescale_eig(eps,nv);
1129: PetscFree3(fl1,inside,rr);
1130: BVSetActiveColumns(eps->V,0,nv);
1131: if (ctx->extraction == EPS_CISS_EXTRACTION_HANKEL) {
1132: ConstructS(eps);
1133: BVSetActiveColumns(ctx->S,0,ctx->L);
1134: BVCopy(ctx->S,ctx->V);
1135: BVSetActiveColumns(ctx->S,0,nv);
1136: }
1137: BVCopy(ctx->S,eps->V);
1139: DSVectors(eps->ds,DS_MAT_X,NULL,NULL);
1140: DSGetMat(eps->ds,DS_MAT_X,&X);
1141: BVMultInPlace(ctx->S,X,0,eps->nconv);
1142: if (eps->ishermitian) {
1143: BVMultInPlace(eps->V,X,0,eps->nconv);
1144: }
1145: MatDestroy(&X);
1146: max_error = 0.0;
1147: for (i=0;i<eps->nconv;i++) {
1148: BVGetColumn(ctx->S,i,&si);
1149: EPSComputeResidualNorm_Private(eps,PETSC_FALSE,eps->eigr[i],eps->eigi[i],si,NULL,w,&error);
1150: if (ctx->extraction == EPS_CISS_EXTRACTION_HANKEL) { /* vector is not normalized */
1151: VecNorm(si,NORM_2,&norm);
1152: error /= norm;
1153: }
1154: (*eps->converged)(eps,eps->eigr[i],eps->eigi[i],error,&error,eps->convergedctx);
1155: BVRestoreColumn(ctx->S,i,&si);
1156: max_error = PetscMax(max_error,error);
1157: }
1159: if (max_error <= eps->tol) eps->reason = EPS_CONVERGED_TOL;
1160: else if (eps->its >= eps->max_it) eps->reason = EPS_DIVERGED_ITS;
1161: else {
1162: if (eps->nconv > ctx->L) {
1163: MatCreateSeqDense(PETSC_COMM_SELF,eps->nconv,ctx->L,NULL,&M);
1164: MatDenseGetArray(M,&temp);
1165: for (i=0;i<ctx->L*eps->nconv;i++) {
1166: PetscRandomGetValue(rand,&temp[i]);
1167: temp[i] = PetscRealPart(temp[i]);
1168: }
1169: MatDenseRestoreArray(M,&temp);
1170: BVSetActiveColumns(ctx->S,0,eps->nconv);
1171: BVMultInPlace(ctx->S,M,0,ctx->L);
1172: MatDestroy(&M);
1173: BVSetActiveColumns(ctx->S,0,ctx->L);
1174: BVCopy(ctx->S,ctx->V);
1175: }
1176: if (ctx->pA) {
1177: VecScatterVecs(eps,ctx->V,ctx->L);
1178: SolveLinearSystem(eps,ctx->pA,ctx->pB,ctx->pV,0,ctx->L,PETSC_FALSE);
1179: } else {
1180: SolveLinearSystem(eps,A,B,ctx->V,0,ctx->L,PETSC_FALSE);
1181: }
1182: }
1183: }
1184: }
1185: if (ctx->extraction == EPS_CISS_EXTRACTION_HANKEL) {
1186: PetscFree(H1);
1187: }
1188: PetscFree2(Mu,H0);
1189: return(0);
1190: }
1192: PetscErrorCode EPSComputeVectors_CISS(EPS eps)
1193: {
1195: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1196: PetscInt n;
1197: Mat Z,B=NULL;
1200: if (eps->ishermitian) {
1201: if (eps->isgeneralized && !eps->ispositive) {
1202: EPSComputeVectors_Indefinite(eps);
1203: } else {
1204: EPSComputeVectors_Hermitian(eps);
1205: }
1206: return(0);
1207: }
1208: DSGetDimensions(eps->ds,&n,NULL,NULL,NULL,NULL);
1209: BVSetActiveColumns(eps->V,0,n);
1211: /* right eigenvectors */
1212: DSVectors(eps->ds,DS_MAT_X,NULL,NULL);
1214: /* V = V * Z */
1215: DSGetMat(eps->ds,DS_MAT_X,&Z);
1216: BVMultInPlace(eps->V,Z,0,n);
1217: MatDestroy(&Z);
1218: BVSetActiveColumns(eps->V,0,eps->nconv);
1220: /* normalize */
1221: if (ctx->extraction == EPS_CISS_EXTRACTION_HANKEL) {
1222: if (eps->isgeneralized && eps->ishermitian && eps->ispositive) {
1223: STGetMatrix(eps->st,1,&B);
1224: BVSetMatrix(eps->V,B,PETSC_FALSE);
1225: }
1226: BVNormalize(eps->V,NULL);
1227: if (B) { BVSetMatrix(eps->V,NULL,PETSC_FALSE); }
1228: }
1229: return(0);
1230: }
1232: static PetscErrorCode EPSCISSSetSizes_CISS(EPS eps,PetscInt ip,PetscInt bs,PetscInt ms,PetscInt npart,PetscInt bsmax,PetscBool realmats)
1233: {
1235: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1236: PetscInt oN,onpart;
1239: oN = ctx->N;
1240: if (ip == PETSC_DECIDE || ip == PETSC_DEFAULT) {
1241: if (ctx->N!=32) { ctx->N =32; ctx->M = ctx->N/4; }
1242: } else {
1243: if (ip<1) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_ARG_OUTOFRANGE,"The ip argument must be > 0");
1244: if (ip%2) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_ARG_OUTOFRANGE,"The ip argument must be an even number");
1245: if (ctx->N!=ip) { ctx->N = ip; ctx->M = ctx->N/4; }
1246: }
1247: if (bs == PETSC_DECIDE || bs == PETSC_DEFAULT) {
1248: ctx->L = 16;
1249: } else {
1250: if (bs<1) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_ARG_OUTOFRANGE,"The bs argument must be > 0");
1251: ctx->L = bs;
1252: }
1253: if (ms == PETSC_DECIDE || ms == PETSC_DEFAULT) {
1254: ctx->M = ctx->N/4;
1255: } else {
1256: if (ms<1) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_ARG_OUTOFRANGE,"The ms argument must be > 0");
1257: if (ms>ctx->N) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_ARG_OUTOFRANGE,"The ms argument must be less than or equal to the number of integration points");
1258: ctx->M = ms;
1259: }
1260: onpart = ctx->npart;
1261: if (npart == PETSC_DECIDE || npart == PETSC_DEFAULT) {
1262: ctx->npart = 1;
1263: } else {
1264: if (npart<1) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_ARG_OUTOFRANGE,"The npart argument must be > 0");
1265: ctx->npart = npart;
1266: }
1267: if (bsmax == PETSC_DECIDE || bsmax == PETSC_DEFAULT) {
1268: ctx->L_max = 64;
1269: } else {
1270: if (bsmax<1) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_ARG_OUTOFRANGE,"The bsmax argument must be > 0");
1271: ctx->L_max = PetscMax(bsmax,ctx->L);
1272: }
1273: if (onpart != ctx->npart || oN != ctx->N || realmats != ctx->isreal) { EPSCISSResetSubcomm(eps); }
1274: ctx->isreal = realmats;
1275: eps->state = EPS_STATE_INITIAL;
1276: return(0);
1277: }
1279: /*@
1280: EPSCISSSetSizes - Sets the values of various size parameters in the CISS solver.
1282: Logically Collective on eps
1284: Input Parameters:
1285: + eps - the eigenproblem solver context
1286: . ip - number of integration points
1287: . bs - block size
1288: . ms - moment size
1289: . npart - number of partitions when splitting the communicator
1290: . bsmax - max block size
1291: - realmats - A and B are real
1293: Options Database Keys:
1294: + -eps_ciss_integration_points - Sets the number of integration points
1295: . -eps_ciss_blocksize - Sets the block size
1296: . -eps_ciss_moments - Sets the moment size
1297: . -eps_ciss_partitions - Sets the number of partitions
1298: . -eps_ciss_maxblocksize - Sets the maximum block size
1299: - -eps_ciss_realmats - A and B are real
1301: Note:
1302: The default number of partitions is 1. This means the internal KSP object is shared
1303: among all processes of the EPS communicator. Otherwise, the communicator is split
1304: into npart communicators, so that npart KSP solves proceed simultaneously.
1306: Level: advanced
1308: .seealso: EPSCISSGetSizes()
1309: @*/
1310: PetscErrorCode EPSCISSSetSizes(EPS eps,PetscInt ip,PetscInt bs,PetscInt ms,PetscInt npart,PetscInt bsmax,PetscBool realmats)
1311: {
1322: PetscTryMethod(eps,"EPSCISSSetSizes_C",(EPS,PetscInt,PetscInt,PetscInt,PetscInt,PetscInt,PetscBool),(eps,ip,bs,ms,npart,bsmax,realmats));
1323: return(0);
1324: }
1326: static PetscErrorCode EPSCISSGetSizes_CISS(EPS eps,PetscInt *ip,PetscInt *bs,PetscInt *ms,PetscInt *npart,PetscInt *bsmax,PetscBool *realmats)
1327: {
1328: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1331: if (ip) *ip = ctx->N;
1332: if (bs) *bs = ctx->L;
1333: if (ms) *ms = ctx->M;
1334: if (npart) *npart = ctx->npart;
1335: if (bsmax) *bsmax = ctx->L_max;
1336: if (realmats) *realmats = ctx->isreal;
1337: return(0);
1338: }
1340: /*@
1341: EPSCISSGetSizes - Gets the values of various size parameters in the CISS solver.
1343: Not Collective
1345: Input Parameter:
1346: . eps - the eigenproblem solver context
1348: Output Parameters:
1349: + ip - number of integration points
1350: . bs - block size
1351: . ms - moment size
1352: . npart - number of partitions when splitting the communicator
1353: . bsmax - max block size
1354: - realmats - A and B are real
1356: Level: advanced
1358: .seealso: EPSCISSSetSizes()
1359: @*/
1360: PetscErrorCode EPSCISSGetSizes(EPS eps,PetscInt *ip,PetscInt *bs,PetscInt *ms,PetscInt *npart,PetscInt *bsmax,PetscBool *realmats)
1361: {
1366: PetscUseMethod(eps,"EPSCISSGetSizes_C",(EPS,PetscInt*,PetscInt*,PetscInt*,PetscInt*,PetscInt*,PetscBool*),(eps,ip,bs,ms,npart,bsmax,realmats));
1367: return(0);
1368: }
1370: static PetscErrorCode EPSCISSSetThreshold_CISS(EPS eps,PetscReal delta,PetscReal spur)
1371: {
1372: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1375: if (delta == PETSC_DEFAULT) {
1376: ctx->delta = 1e-12;
1377: } else {
1378: if (delta<=0.0) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_ARG_OUTOFRANGE,"The delta argument must be > 0.0");
1379: ctx->delta = delta;
1380: }
1381: if (spur == PETSC_DEFAULT) {
1382: ctx->spurious_threshold = 1e-4;
1383: } else {
1384: if (spur<=0.0) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_ARG_OUTOFRANGE,"The spurious threshold argument must be > 0.0");
1385: ctx->spurious_threshold = spur;
1386: }
1387: return(0);
1388: }
1390: /*@
1391: EPSCISSSetThreshold - Sets the values of various threshold parameters in
1392: the CISS solver.
1394: Logically Collective on eps
1396: Input Parameters:
1397: + eps - the eigenproblem solver context
1398: . delta - threshold for numerical rank
1399: - spur - spurious threshold (to discard spurious eigenpairs)
1401: Options Database Keys:
1402: + -eps_ciss_delta - Sets the delta
1403: - -eps_ciss_spurious_threshold - Sets the spurious threshold
1405: Level: advanced
1407: .seealso: EPSCISSGetThreshold()
1408: @*/
1409: PetscErrorCode EPSCISSSetThreshold(EPS eps,PetscReal delta,PetscReal spur)
1410: {
1417: PetscTryMethod(eps,"EPSCISSSetThreshold_C",(EPS,PetscReal,PetscReal),(eps,delta,spur));
1418: return(0);
1419: }
1421: static PetscErrorCode EPSCISSGetThreshold_CISS(EPS eps,PetscReal *delta,PetscReal *spur)
1422: {
1423: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1426: if (delta) *delta = ctx->delta;
1427: if (spur) *spur = ctx->spurious_threshold;
1428: return(0);
1429: }
1431: /*@
1432: EPSCISSGetThreshold - Gets the values of various threshold parameters
1433: in the CISS solver.
1435: Not Collective
1437: Input Parameter:
1438: . eps - the eigenproblem solver context
1440: Output Parameters:
1441: + delta - threshold for numerical rank
1442: - spur - spurious threshold (to discard spurious eigenpairs)
1444: Level: advanced
1446: .seealso: EPSCISSSetThreshold()
1447: @*/
1448: PetscErrorCode EPSCISSGetThreshold(EPS eps,PetscReal *delta,PetscReal *spur)
1449: {
1454: PetscUseMethod(eps,"EPSCISSGetThreshold_C",(EPS,PetscReal*,PetscReal*),(eps,delta,spur));
1455: return(0);
1456: }
1458: static PetscErrorCode EPSCISSSetRefinement_CISS(EPS eps,PetscInt inner,PetscInt blsize)
1459: {
1460: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1463: if (inner == PETSC_DEFAULT) {
1464: ctx->refine_inner = 0;
1465: } else {
1466: if (inner<0) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_ARG_OUTOFRANGE,"The refine inner argument must be >= 0");
1467: ctx->refine_inner = inner;
1468: }
1469: if (blsize == PETSC_DEFAULT) {
1470: ctx->refine_blocksize = 0;
1471: } else {
1472: if (blsize<0) SETERRQ(PetscObjectComm((PetscObject)eps),PETSC_ERR_ARG_OUTOFRANGE,"The refine blocksize argument must be >= 0");
1473: ctx->refine_blocksize = blsize;
1474: }
1475: return(0);
1476: }
1478: /*@
1479: EPSCISSSetRefinement - Sets the values of various refinement parameters
1480: in the CISS solver.
1482: Logically Collective on eps
1484: Input Parameters:
1485: + eps - the eigenproblem solver context
1486: . inner - number of iterative refinement iterations (inner loop)
1487: - blsize - number of iterative refinement iterations (blocksize loop)
1489: Options Database Keys:
1490: + -eps_ciss_refine_inner - Sets number of inner iterations
1491: - -eps_ciss_refine_blocksize - Sets number of blocksize iterations
1493: Level: advanced
1495: .seealso: EPSCISSGetRefinement()
1496: @*/
1497: PetscErrorCode EPSCISSSetRefinement(EPS eps,PetscInt inner,PetscInt blsize)
1498: {
1505: PetscTryMethod(eps,"EPSCISSSetRefinement_C",(EPS,PetscInt,PetscInt),(eps,inner,blsize));
1506: return(0);
1507: }
1509: static PetscErrorCode EPSCISSGetRefinement_CISS(EPS eps,PetscInt *inner,PetscInt *blsize)
1510: {
1511: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1514: if (inner) *inner = ctx->refine_inner;
1515: if (blsize) *blsize = ctx->refine_blocksize;
1516: return(0);
1517: }
1519: /*@
1520: EPSCISSGetRefinement - Gets the values of various refinement parameters
1521: in the CISS solver.
1523: Not Collective
1525: Input Parameter:
1526: . eps - the eigenproblem solver context
1528: Output Parameters:
1529: + inner - number of iterative refinement iterations (inner loop)
1530: - blsize - number of iterative refinement iterations (blocksize loop)
1532: Level: advanced
1534: .seealso: EPSCISSSetRefinement()
1535: @*/
1536: PetscErrorCode EPSCISSGetRefinement(EPS eps, PetscInt *inner, PetscInt *blsize)
1537: {
1542: PetscUseMethod(eps,"EPSCISSGetRefinement_C",(EPS,PetscInt*,PetscInt*),(eps,inner,blsize));
1543: return(0);
1544: }
1546: static PetscErrorCode EPSCISSSetUseST_CISS(EPS eps,PetscBool usest)
1547: {
1548: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1551: ctx->usest = usest;
1552: ctx->usest_set = PETSC_TRUE;
1553: eps->state = EPS_STATE_INITIAL;
1554: return(0);
1555: }
1557: /*@
1558: EPSCISSSetUseST - Sets a flag indicating that the CISS solver will
1559: use the ST object for the linear solves.
1561: Logically Collective on eps
1563: Input Parameters:
1564: + eps - the eigenproblem solver context
1565: - usest - boolean flag to use the ST object or not
1567: Options Database Keys:
1568: . -eps_ciss_usest <bool> - whether the ST object will be used or not
1570: Level: advanced
1572: .seealso: EPSCISSGetUseST()
1573: @*/
1574: PetscErrorCode EPSCISSSetUseST(EPS eps,PetscBool usest)
1575: {
1581: PetscTryMethod(eps,"EPSCISSSetUseST_C",(EPS,PetscBool),(eps,usest));
1582: return(0);
1583: }
1585: static PetscErrorCode EPSCISSGetUseST_CISS(EPS eps,PetscBool *usest)
1586: {
1587: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1590: *usest = ctx->usest;
1591: return(0);
1592: }
1594: /*@
1595: EPSCISSGetUseST - Gets the flag for using the ST object
1596: in the CISS solver.
1598: Not Collective
1600: Input Parameter:
1601: . eps - the eigenproblem solver context
1603: Output Parameters:
1604: . usest - boolean flag indicating if the ST object is being used
1606: Level: advanced
1608: .seealso: EPSCISSSetUseST()
1609: @*/
1610: PetscErrorCode EPSCISSGetUseST(EPS eps,PetscBool *usest)
1611: {
1617: PetscUseMethod(eps,"EPSCISSGetUseST_C",(EPS,PetscBool*),(eps,usest));
1618: return(0);
1619: }
1621: static PetscErrorCode EPSCISSSetQuadRule_CISS(EPS eps,EPSCISSQuadRule quad)
1622: {
1623: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1626: ctx->quad = quad;
1627: return(0);
1628: }
1630: /*@
1631: EPSCISSSetQuadRule - Sets the quadrature rule used in the CISS solver.
1633: Logically Collective on eps
1635: Input Parameters:
1636: + eps - the eigenproblem solver context
1637: - quad - the quadrature rule
1639: Options Database Key:
1640: . -eps_ciss_quadrule - Sets the quadrature rule (either 'trapezoidal' or
1641: 'chebyshev')
1643: Notes:
1644: By default, the trapezoidal rule is used (EPS_CISS_QUADRULE_TRAPEZOIDAL).
1646: If the 'chebyshev' option is specified (EPS_CISS_QUADRULE_CHEBYSHEV), then
1647: Chebyshev points are used as quadrature points.
1649: Level: advanced
1651: .seealso: EPSCISSGetQuadRule(), EPSCISSQuadRule
1652: @*/
1653: PetscErrorCode EPSCISSSetQuadRule(EPS eps,EPSCISSQuadRule quad)
1654: {
1660: PetscTryMethod(eps,"EPSCISSSetQuadRule_C",(EPS,EPSCISSQuadRule),(eps,quad));
1661: return(0);
1662: }
1664: static PetscErrorCode EPSCISSGetQuadRule_CISS(EPS eps,EPSCISSQuadRule *quad)
1665: {
1666: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1669: *quad = ctx->quad;
1670: return(0);
1671: }
1673: /*@
1674: EPSCISSGetQuadRule - Gets the quadrature rule used in the CISS solver.
1676: Not Collective
1678: Input Parameter:
1679: . eps - the eigenproblem solver context
1681: Output Parameters:
1682: . quad - quadrature rule
1684: Level: advanced
1686: .seealso: EPSCISSSetQuadRule() EPSCISSQuadRule
1687: @*/
1688: PetscErrorCode EPSCISSGetQuadRule(EPS eps,EPSCISSQuadRule *quad)
1689: {
1695: PetscUseMethod(eps,"EPSCISSGetQuadRule_C",(EPS,EPSCISSQuadRule*),(eps,quad));
1696: return(0);
1697: }
1699: static PetscErrorCode EPSCISSSetExtraction_CISS(EPS eps,EPSCISSExtraction extraction)
1700: {
1701: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1704: ctx->extraction = extraction;
1705: return(0);
1706: }
1708: /*@
1709: EPSCISSSetExtraction - Sets the extraction technique used in the CISS solver.
1711: Logically Collective on eps
1713: Input Parameters:
1714: + eps - the eigenproblem solver context
1715: - extraction - the extraction technique
1717: Options Database Key:
1718: . -eps_ciss_extraction - Sets the extraction technique (either 'ritz' or
1719: 'hankel')
1721: Notes:
1722: By default, the Rayleigh-Ritz extraction is used (EPS_CISS_EXTRACTION_RITZ).
1724: If the 'hankel' option is specified (EPS_CISS_EXTRACTION_HANKEL), then
1725: the Block Hankel method is used for extracting eigenpairs.
1727: Level: advanced
1729: .seealso: EPSCISSGetExtraction(), EPSCISSExtraction
1730: @*/
1731: PetscErrorCode EPSCISSSetExtraction(EPS eps,EPSCISSExtraction extraction)
1732: {
1738: PetscTryMethod(eps,"EPSCISSSetExtraction_C",(EPS,EPSCISSExtraction),(eps,extraction));
1739: return(0);
1740: }
1742: static PetscErrorCode EPSCISSGetExtraction_CISS(EPS eps,EPSCISSExtraction *extraction)
1743: {
1744: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1747: *extraction = ctx->extraction;
1748: return(0);
1749: }
1751: /*@
1752: EPSCISSGetExtraction - Gets the extraction technique used in the CISS solver.
1754: Not Collective
1756: Input Parameter:
1757: . eps - the eigenproblem solver context
1759: Output Parameters:
1760: . extraction - extraction technique
1762: Level: advanced
1764: .seealso: EPSCISSSetExtraction() EPSCISSExtraction
1765: @*/
1766: PetscErrorCode EPSCISSGetExtraction(EPS eps,EPSCISSExtraction *extraction)
1767: {
1773: PetscUseMethod(eps,"EPSCISSGetExtraction_C",(EPS,EPSCISSExtraction*),(eps,extraction));
1774: return(0);
1775: }
1777: static PetscErrorCode EPSCISSGetKSPs_CISS(EPS eps,PetscInt *nsolve,KSP **ksp)
1778: {
1780: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1781: PetscInt i;
1782: PC pc;
1785: if (!ctx->ksp) {
1786: if (!ctx->subcomm) { /* initialize subcomm first */
1787: EPSCISSSetUseConj(eps,&ctx->useconj);
1788: EPSCISSSetUpSubComm(eps,&ctx->num_solve_point);
1789: }
1790: PetscMalloc1(ctx->num_solve_point,&ctx->ksp);
1791: for (i=0;i<ctx->num_solve_point;i++) {
1792: KSPCreate(PetscSubcommChild(ctx->subcomm),&ctx->ksp[i]);
1793: PetscObjectIncrementTabLevel((PetscObject)ctx->ksp[i],(PetscObject)eps,1);
1794: KSPSetOptionsPrefix(ctx->ksp[i],((PetscObject)eps)->prefix);
1795: KSPAppendOptionsPrefix(ctx->ksp[i],"eps_ciss_");
1796: PetscLogObjectParent((PetscObject)eps,(PetscObject)ctx->ksp[i]);
1797: PetscObjectSetOptions((PetscObject)ctx->ksp[i],((PetscObject)eps)->options);
1798: KSPSetErrorIfNotConverged(ctx->ksp[i],PETSC_TRUE);
1799: KSPSetTolerances(ctx->ksp[i],SLEPC_DEFAULT_TOL,PETSC_DEFAULT,PETSC_DEFAULT,PETSC_DEFAULT);
1800: KSPGetPC(ctx->ksp[i],&pc);
1801: KSPSetType(ctx->ksp[i],KSPPREONLY);
1802: PCSetType(pc,PCLU);
1803: }
1804: }
1805: if (nsolve) *nsolve = ctx->num_solve_point;
1806: if (ksp) *ksp = ctx->ksp;
1807: return(0);
1808: }
1810: /*@C
1811: EPSCISSGetKSPs - Retrieve the array of linear solver objects associated with
1812: the CISS solver.
1814: Not Collective
1816: Input Parameter:
1817: . eps - the eigenproblem solver solver
1819: Output Parameters:
1820: + nsolve - number of solver objects
1821: - ksp - array of linear solver object
1823: Notes:
1824: The number of KSP solvers is equal to the number of integration points divided by
1825: the number of partitions. This value is halved in the case of real matrices with
1826: a region centered at the real axis.
1828: Level: advanced
1830: .seealso: EPSCISSSetSizes()
1831: @*/
1832: PetscErrorCode EPSCISSGetKSPs(EPS eps,PetscInt *nsolve,KSP **ksp)
1833: {
1838: PetscUseMethod(eps,"EPSCISSGetKSPs_C",(EPS,PetscInt*,KSP**),(eps,nsolve,ksp));
1839: return(0);
1840: }
1842: PetscErrorCode EPSReset_CISS(EPS eps)
1843: {
1845: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1846: PetscInt i;
1849: BVDestroy(&ctx->S);
1850: BVDestroy(&ctx->V);
1851: BVDestroy(&ctx->Y);
1852: if (!ctx->usest) {
1853: for (i=0;i<ctx->num_solve_point;i++) {
1854: KSPReset(ctx->ksp[i]);
1855: }
1856: }
1857: VecScatterDestroy(&ctx->scatterin);
1858: VecDestroy(&ctx->xsub);
1859: VecDestroy(&ctx->xdup);
1860: if (ctx->pA) {
1861: MatDestroy(&ctx->pA);
1862: MatDestroy(&ctx->pB);
1863: BVDestroy(&ctx->pV);
1864: }
1865: return(0);
1866: }
1868: PetscErrorCode EPSSetFromOptions_CISS(PetscOptionItems *PetscOptionsObject,EPS eps)
1869: {
1870: PetscErrorCode ierr;
1871: PetscReal r3,r4;
1872: PetscInt i,i1,i2,i3,i4,i5,i6,i7;
1873: PetscBool b1,b2,flg;
1874: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1875: EPSCISSQuadRule quad;
1876: EPSCISSExtraction extraction;
1879: PetscOptionsHead(PetscOptionsObject,"EPS CISS Options");
1881: EPSCISSGetSizes(eps,&i1,&i2,&i3,&i4,&i5,&b1);
1882: PetscOptionsInt("-eps_ciss_integration_points","Number of integration points","EPSCISSSetSizes",i1,&i1,NULL);
1883: PetscOptionsInt("-eps_ciss_blocksize","Block size","EPSCISSSetSizes",i2,&i2,NULL);
1884: PetscOptionsInt("-eps_ciss_moments","Moment size","EPSCISSSetSizes",i3,&i3,NULL);
1885: PetscOptionsInt("-eps_ciss_partitions","Number of partitions","EPSCISSSetSizes",i4,&i4,NULL);
1886: PetscOptionsInt("-eps_ciss_maxblocksize","Maximum block size","EPSCISSSetSizes",i5,&i5,NULL);
1887: PetscOptionsBool("-eps_ciss_realmats","True if A and B are real","EPSCISSSetSizes",b1,&b1,NULL);
1888: EPSCISSSetSizes(eps,i1,i2,i3,i4,i5,b1);
1890: EPSCISSGetThreshold(eps,&r3,&r4);
1891: PetscOptionsReal("-eps_ciss_delta","Threshold for numerical rank","EPSCISSSetThreshold",r3,&r3,NULL);
1892: PetscOptionsReal("-eps_ciss_spurious_threshold","Threshold for the spurious eigenpairs","EPSCISSSetThreshold",r4,&r4,NULL);
1893: EPSCISSSetThreshold(eps,r3,r4);
1895: EPSCISSGetRefinement(eps,&i6,&i7);
1896: PetscOptionsInt("-eps_ciss_refine_inner","Number of inner iterative refinement iterations","EPSCISSSetRefinement",i6,&i6,NULL);
1897: PetscOptionsInt("-eps_ciss_refine_blocksize","Number of blocksize iterative refinement iterations","EPSCISSSetRefinement",i7,&i7,NULL);
1898: EPSCISSSetRefinement(eps,i6,i7);
1900: EPSCISSGetUseST(eps,&b2);
1901: PetscOptionsBool("-eps_ciss_usest","Use ST for linear solves","EPSCISSSetUseST",b2,&b2,&flg);
1902: if (flg) { EPSCISSSetUseST(eps,b2); }
1904: PetscOptionsEnum("-eps_ciss_quadrule","Quadrature rule","EPSCISSSetQuadRule",EPSCISSQuadRules,(PetscEnum)ctx->quad,(PetscEnum*)&quad,&flg);
1905: if (flg) { EPSCISSSetQuadRule(eps,quad); }
1907: PetscOptionsEnum("-eps_ciss_extraction","Extraction technique","EPSCISSSetExtraction",EPSCISSExtractions,(PetscEnum)ctx->extraction,(PetscEnum*)&extraction,&flg);
1908: if (flg) { EPSCISSSetExtraction(eps,extraction); }
1910: PetscOptionsTail();
1912: if (!eps->rg) { EPSGetRG(eps,&eps->rg); }
1913: RGSetFromOptions(eps->rg); /* this is necessary here to set useconj */
1914: if (!ctx->ksp) { EPSCISSGetKSPs(eps,&ctx->num_solve_point,&ctx->ksp); }
1915: for (i=0;i<ctx->num_solve_point;i++) {
1916: KSPSetFromOptions(ctx->ksp[i]);
1917: }
1918: PetscSubcommSetFromOptions(ctx->subcomm);
1919: return(0);
1920: }
1922: PetscErrorCode EPSDestroy_CISS(EPS eps)
1923: {
1925: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1928: EPSCISSResetSubcomm(eps);
1929: PetscFree4(ctx->weight,ctx->omega,ctx->pp,ctx->sigma);
1930: PetscFree(eps->data);
1931: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSSetSizes_C",NULL);
1932: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetSizes_C",NULL);
1933: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSSetThreshold_C",NULL);
1934: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetThreshold_C",NULL);
1935: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSSetRefinement_C",NULL);
1936: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetRefinement_C",NULL);
1937: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSSetUseST_C",NULL);
1938: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetUseST_C",NULL);
1939: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSSetQuadRule_C",NULL);
1940: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetQuadRule_C",NULL);
1941: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSSetExtraction_C",NULL);
1942: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetExtraction_C",NULL);
1943: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetKSPs_C",NULL);
1944: return(0);
1945: }
1947: PetscErrorCode EPSView_CISS(EPS eps,PetscViewer viewer)
1948: {
1950: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1951: PetscBool isascii;
1954: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&isascii);
1955: if (isascii) {
1956: PetscViewerASCIIPrintf(viewer," sizes { integration points: %D, block size: %D, moment size: %D, partitions: %D, maximum block size: %D }\n",ctx->N,ctx->L,ctx->M,ctx->npart,ctx->L_max);
1957: if (ctx->isreal) {
1958: PetscViewerASCIIPrintf(viewer," exploiting symmetry of integration points\n");
1959: }
1960: PetscViewerASCIIPrintf(viewer," threshold { delta: %g, spurious threshold: %g }\n",(double)ctx->delta,(double)ctx->spurious_threshold);
1961: PetscViewerASCIIPrintf(viewer," iterative refinement { inner: %D, blocksize: %D }\n",ctx->refine_inner, ctx->refine_blocksize);
1962: PetscViewerASCIIPrintf(viewer," extraction: %s\n",EPSCISSExtractions[ctx->extraction]);
1963: PetscViewerASCIIPrintf(viewer," quadrature rule: %s\n",EPSCISSQuadRules[ctx->quad]);
1964: if (ctx->usest) {
1965: PetscViewerASCIIPrintf(viewer," using ST for linear solves\n");
1966: } else {
1967: if (!ctx->ksp) { EPSCISSGetKSPs(eps,&ctx->num_solve_point,&ctx->ksp); }
1968: PetscViewerASCIIPushTab(viewer);
1969: KSPView(ctx->ksp[0],viewer);
1970: PetscViewerASCIIPopTab(viewer);
1971: }
1972: }
1973: return(0);
1974: }
1976: PetscErrorCode EPSSetDefaultST_CISS(EPS eps)
1977: {
1979: EPS_CISS *ctx = (EPS_CISS*)eps->data;
1980: PetscBool usest = ctx->usest;
1983: if (!((PetscObject)eps->st)->type_name) {
1984: if (!ctx->usest_set) usest = (ctx->npart>1)? PETSC_FALSE: PETSC_TRUE;
1985: if (usest) {
1986: STSetType(eps->st,STSINVERT);
1987: } else {
1988: /* we are not going to use ST, so avoid factorizing the matrix */
1989: STSetType(eps->st,STSHIFT);
1990: }
1991: }
1992: return(0);
1993: }
1995: SLEPC_EXTERN PetscErrorCode EPSCreate_CISS(EPS eps)
1996: {
1998: EPS_CISS *ctx = (EPS_CISS*)eps->data;
2001: PetscNewLog(eps,&ctx);
2002: eps->data = ctx;
2004: eps->useds = PETSC_TRUE;
2005: eps->categ = EPS_CATEGORY_CONTOUR;
2007: eps->ops->solve = EPSSolve_CISS;
2008: eps->ops->setup = EPSSetUp_CISS;
2009: eps->ops->setupsort = EPSSetUpSort_CISS;
2010: eps->ops->setfromoptions = EPSSetFromOptions_CISS;
2011: eps->ops->destroy = EPSDestroy_CISS;
2012: eps->ops->reset = EPSReset_CISS;
2013: eps->ops->view = EPSView_CISS;
2014: eps->ops->computevectors = EPSComputeVectors_CISS;
2015: eps->ops->setdefaultst = EPSSetDefaultST_CISS;
2017: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSSetSizes_C",EPSCISSSetSizes_CISS);
2018: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetSizes_C",EPSCISSGetSizes_CISS);
2019: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSSetThreshold_C",EPSCISSSetThreshold_CISS);
2020: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetThreshold_C",EPSCISSGetThreshold_CISS);
2021: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSSetRefinement_C",EPSCISSSetRefinement_CISS);
2022: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetRefinement_C",EPSCISSGetRefinement_CISS);
2023: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSSetUseST_C",EPSCISSSetUseST_CISS);
2024: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetUseST_C",EPSCISSGetUseST_CISS);
2025: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSSetQuadRule_C",EPSCISSSetQuadRule_CISS);
2026: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetQuadRule_C",EPSCISSGetQuadRule_CISS);
2027: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSSetExtraction_C",EPSCISSSetExtraction_CISS);
2028: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetExtraction_C",EPSCISSGetExtraction_CISS);
2029: PetscObjectComposeFunction((PetscObject)eps,"EPSCISSGetKSPs_C",EPSCISSGetKSPs_CISS);
2031: /* log events */
2032: PetscLogEventRegister("EPSCISS_SVD",EPS_CLASSID,&EPS_CISS_SVD);
2034: /* set default values of parameters */
2035: ctx->N = 32;
2036: ctx->L = 16;
2037: ctx->M = ctx->N/4;
2038: ctx->delta = 1e-12;
2039: ctx->L_max = 64;
2040: ctx->spurious_threshold = 1e-4;
2041: ctx->usest = PETSC_TRUE;
2042: ctx->usest_set = PETSC_FALSE;
2043: ctx->isreal = PETSC_FALSE;
2044: ctx->refine_inner = 0;
2045: ctx->refine_blocksize = 0;
2046: ctx->npart = 1;
2047: ctx->quad = (EPSCISSQuadRule)0;
2048: ctx->extraction = EPS_CISS_EXTRACTION_RITZ;
2049: return(0);
2050: }