Actual source code: dsnep.c
slepc-3.14.2 2021-02-01
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: */
11: #include <slepc/private/dsimpl.h>
12: #include <slepcblaslapack.h>
14: typedef struct {
15: PetscInt nf; /* number of functions in f[] */
16: FN f[DS_NUM_EXTRA]; /* functions defining the nonlinear operator */
17: PetscInt neig; /* number of available eigenpairs */
18: void *computematrixctx;
19: PetscErrorCode (*computematrix)(DS,PetscScalar,PetscBool,DSMatType,void*);
20: } DS_NEP;
22: /*
23: DSNEPComputeMatrix - Build the matrix associated with a nonlinear operator
24: T(lambda) or its derivative T'(lambda), given the parameter lambda, where
25: T(lambda) = sum_i E_i*f_i(lambda). The result is written in mat.
26: */
27: static PetscErrorCode DSNEPComputeMatrix(DS ds,PetscScalar lambda,PetscBool deriv,DSMatType mat)
28: {
30: DS_NEP *ctx = (DS_NEP*)ds->data;
31: PetscScalar *T,*E,alpha;
32: PetscInt i,ld,n;
33: PetscBLASInt k,inc=1;
36: PetscLogEventBegin(DS_Other,ds,0,0,0);
37: if (ctx->computematrix) {
38: (*ctx->computematrix)(ds,lambda,deriv,mat,ctx->computematrixctx);
39: } else {
40: DSGetDimensions(ds,&n,NULL,NULL,NULL,NULL);
41: DSGetLeadingDimension(ds,&ld);
42: PetscBLASIntCast(ld*n,&k);
43: DSGetArray(ds,mat,&T);
44: PetscArrayzero(T,k);
45: for (i=0;i<ctx->nf;i++) {
46: if (deriv) {
47: FNEvaluateDerivative(ctx->f[i],lambda,&alpha);
48: } else {
49: FNEvaluateFunction(ctx->f[i],lambda,&alpha);
50: }
51: E = ds->mat[DSMatExtra[i]];
52: PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&k,&alpha,E,&inc,T,&inc));
53: }
54: DSRestoreArray(ds,mat,&T);
55: }
56: PetscLogEventEnd(DS_Other,ds,0,0,0);
57: return(0);
58: }
60: PetscErrorCode DSAllocate_NEP(DS ds,PetscInt ld)
61: {
63: DS_NEP *ctx = (DS_NEP*)ds->data;
64: PetscInt i;
67: if (!ctx->nf) SETERRQ(PetscObjectComm((PetscObject)ds),PETSC_ERR_ARG_WRONGSTATE,"DSNEP requires passing some functions via DSSetFN()");
68: DSAllocateMat_Private(ds,DS_MAT_X);
69: for (i=0;i<ctx->nf;i++) {
70: DSAllocateMat_Private(ds,DSMatExtra[i]);
71: }
72: PetscFree(ds->perm);
73: PetscMalloc1(ld,&ds->perm);
74: PetscLogObjectMemory((PetscObject)ds,ld*sizeof(PetscInt));
75: return(0);
76: }
78: PetscErrorCode DSView_NEP(DS ds,PetscViewer viewer)
79: {
80: PetscErrorCode ierr;
81: DS_NEP *ctx = (DS_NEP*)ds->data;
82: PetscViewerFormat format;
83: PetscInt i;
86: PetscViewerGetFormat(viewer,&format);
87: PetscViewerASCIIPrintf(viewer,"number of functions: %D\n",ctx->nf);
88: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) return(0);
89: for (i=0;i<ctx->nf;i++) {
90: FNView(ctx->f[i],viewer);
91: DSViewMat(ds,viewer,DSMatExtra[i]);
92: }
93: if (ds->state>DS_STATE_INTERMEDIATE) { DSViewMat(ds,viewer,DS_MAT_X); }
94: return(0);
95: }
97: PetscErrorCode DSVectors_NEP(DS ds,DSMatType mat,PetscInt *j,PetscReal *rnorm)
98: {
100: if (rnorm) SETERRQ(PetscObjectComm((PetscObject)ds),PETSC_ERR_SUP,"Not implemented yet");
101: switch (mat) {
102: case DS_MAT_X:
103: break;
104: case DS_MAT_Y:
105: SETERRQ(PetscObjectComm((PetscObject)ds),PETSC_ERR_SUP,"Not implemented yet");
106: default:
107: SETERRQ(PetscObjectComm((PetscObject)ds),PETSC_ERR_ARG_OUTOFRANGE,"Invalid mat parameter");
108: }
109: return(0);
110: }
112: PetscErrorCode DSSort_NEP(DS ds,PetscScalar *wr,PetscScalar *wi,PetscScalar *rr,PetscScalar *ri,PetscInt *dummy)
113: {
115: DS_NEP *ctx = (DS_NEP*)ds->data;
116: PetscInt n,l,i,j,k,p,*perm,told,ld=ds->ld;
117: PetscScalar *A,*X,rtmp;
120: if (!ds->sc) return(0);
121: n = ds->n;
122: l = ds->l;
123: A = ds->mat[DS_MAT_A];
124: perm = ds->perm;
125: for (i=0;i<ctx->neig;i++) perm[i] = i;
126: told = ds->t;
127: ds->t = ctx->neig; /* force the sorting routines to consider ctx->neig eigenvalues */
128: if (rr) {
129: DSSortEigenvalues_Private(ds,rr,ri,perm,PETSC_FALSE);
130: } else {
131: DSSortEigenvalues_Private(ds,wr,NULL,perm,PETSC_FALSE);
132: }
133: ds->t = told; /* restore value of t */
134: for (i=l;i<n;i++) A[i+i*ld] = wr[perm[i]];
135: for (i=l;i<n;i++) wr[i] = A[i+i*ld];
136: /* cannot use DSPermuteColumns_Private() since not all columns are filled */
137: X = ds->mat[DS_MAT_X];
138: for (i=0;i<ctx->neig;i++) {
139: p = perm[i];
140: if (p != i) {
141: j = i + 1;
142: while (perm[j] != i) j++;
143: perm[j] = p; perm[i] = i;
144: /* swap columns i and j */
145: for (k=0;k<n;k++) {
146: rtmp = X[k+p*ld]; X[k+p*ld] = X[k+i*ld]; X[k+i*ld] = rtmp;
147: }
148: }
149: }
150: return(0);
151: }
153: PetscErrorCode DSSolve_NEP_SLP(DS ds,PetscScalar *wr,PetscScalar *wi)
154: {
156: DS_NEP *ctx = (DS_NEP*)ds->data;
157: PetscScalar *A,*B,*W,*X,*work,*alpha,*beta;
158: PetscScalar norm,sigma,lambda,mu,re,re2,sone=1.0,zero=0.0;
159: PetscBLASInt info,n,ld,lrwork=0,lwork,one=1;
160: PetscInt it,pos,j,maxit=100,result;
161: PetscReal tol;
162: #if defined(PETSC_USE_COMPLEX)
163: PetscReal *rwork;
164: #else
165: PetscReal *alphai,im,im2;
166: #endif
169: if (!ds->mat[DS_MAT_A]) {
170: DSAllocateMat_Private(ds,DS_MAT_A);
171: }
172: if (!ds->mat[DS_MAT_B]) {
173: DSAllocateMat_Private(ds,DS_MAT_B);
174: }
175: if (!ds->mat[DS_MAT_W]) {
176: DSAllocateMat_Private(ds,DS_MAT_W);
177: }
178: PetscBLASIntCast(ds->n,&n);
179: PetscBLASIntCast(ds->ld,&ld);
180: #if defined(PETSC_USE_COMPLEX)
181: PetscBLASIntCast(2*ds->n+2*ds->n,&lwork);
182: PetscBLASIntCast(8*ds->n,&lrwork);
183: #else
184: PetscBLASIntCast(3*ds->n+8*ds->n,&lwork);
185: #endif
186: DSAllocateWork_Private(ds,lwork,lrwork,0);
187: alpha = ds->work;
188: beta = ds->work + ds->n;
189: #if defined(PETSC_USE_COMPLEX)
190: work = ds->work + 2*ds->n;
191: lwork -= 2*ds->n;
192: #else
193: alphai = ds->work + 2*ds->n;
194: work = ds->work + 3*ds->n;
195: lwork -= 3*ds->n;
196: #endif
197: A = ds->mat[DS_MAT_A];
198: B = ds->mat[DS_MAT_B];
199: W = ds->mat[DS_MAT_W];
200: X = ds->mat[DS_MAT_X];
202: sigma = 0.0;
203: if (ds->sc->comparison==SlepcCompareTargetMagnitude || ds->sc->comparison==SlepcCompareTargetReal) sigma = *(PetscScalar*)ds->sc->comparisonctx;
204: lambda = sigma;
205: tol = 1000*n*PETSC_MACHINE_EPSILON;
207: for (it=0;it<maxit;it++) {
209: /* evaluate T and T' */
210: DSNEPComputeMatrix(ds,lambda,PETSC_FALSE,DS_MAT_A);
211: if (it) {
212: PetscStackCallBLAS("BLASgemv",BLASgemv_("N",&n,&n,&sone,A,&ld,X,&one,&zero,X+ld,&one));
213: norm = BLASnrm2_(&n,X+ld,&one);
214: if (PetscRealPart(norm)/PetscAbsScalar(lambda)<=tol) break;
215: }
216: DSNEPComputeMatrix(ds,lambda,PETSC_TRUE,DS_MAT_B);
218: /* compute eigenvalue correction mu and eigenvector u */
219: #if defined(PETSC_USE_COMPLEX)
220: rwork = ds->rwork;
221: PetscStackCallBLAS("LAPACKggev",LAPACKggev_("N","V",&n,A,&ld,B,&ld,alpha,beta,NULL,&ld,W,&ld,work,&lwork,rwork,&info));
222: #else
223: PetscStackCallBLAS("LAPACKggev",LAPACKggev_("N","V",&n,A,&ld,B,&ld,alpha,alphai,beta,NULL,&ld,W,&ld,work,&lwork,&info));
224: #endif
225: SlepcCheckLapackInfo("ggev",info);
227: /* find smallest eigenvalue */
228: j = 0;
229: if (beta[j]==0.0) re = (PetscRealPart(alpha[j])>0.0)? PETSC_MAX_REAL: PETSC_MIN_REAL;
230: else re = alpha[j]/beta[j];
231: #if !defined(PETSC_USE_COMPLEX)
232: if (beta[j]==0.0) im = (alphai[j]>0.0)? PETSC_MAX_REAL: PETSC_MIN_REAL;
233: else im = alphai[j]/beta[j];
234: #endif
235: pos = 0;
236: for (j=1;j<n;j++) {
237: if (beta[j]==0.0) re2 = (PetscRealPart(alpha[j])>0.0)? PETSC_MAX_REAL: PETSC_MIN_REAL;
238: else re2 = alpha[j]/beta[j];
239: #if !defined(PETSC_USE_COMPLEX)
240: if (beta[j]==0.0) im2 = (alphai[j]>0.0)? PETSC_MAX_REAL: PETSC_MIN_REAL;
241: else im2 = alphai[j]/beta[j];
242: SlepcCompareSmallestMagnitude(re,im,re2,im2,&result,NULL);
243: #else
244: SlepcCompareSmallestMagnitude(re,0.0,re2,0.0,&result,NULL);
245: #endif
246: if (result > 0) {
247: re = re2;
248: #if !defined(PETSC_USE_COMPLEX)
249: im = im2;
250: #endif
251: pos = j;
252: }
253: }
255: #if !defined(PETSC_USE_COMPLEX)
256: if (im!=0.0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"DSNEP found a complex eigenvalue; try rerunning with complex scalars");
257: #endif
258: mu = alpha[pos]/beta[pos];
259: PetscArraycpy(X,W+pos*ld,n);
260: norm = BLASnrm2_(&n,X,&one);
261: norm = 1.0/norm;
262: PetscStackCallBLAS("BLASscal",BLASscal_(&n,&norm,X,&one));
264: /* correct eigenvalue approximation */
265: lambda = lambda - mu;
266: }
268: if (it==maxit) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_CONV_FAILED,"DSNEP did not converge");
269: ctx->neig = 1;
270: wr[0] = lambda;
271: if (wi) wi[0] = 0.0;
272: return(0);
273: }
275: PetscErrorCode DSSynchronize_NEP(DS ds,PetscScalar eigr[],PetscScalar eigi[])
276: {
278: PetscInt k=0;
279: PetscMPIInt n,rank,size,off=0;
282: if (ds->state>=DS_STATE_CONDENSED) k += ds->n;
283: if (eigr) k += 1;
284: if (eigi) k += 1;
285: DSAllocateWork_Private(ds,k,0,0);
286: PetscMPIIntCast(k*sizeof(PetscScalar),&size);
287: PetscMPIIntCast(ds->n,&n);
288: MPI_Comm_rank(PetscObjectComm((PetscObject)ds),&rank);
289: if (!rank) {
290: if (ds->state>=DS_STATE_CONDENSED) {
291: MPI_Pack(ds->mat[DS_MAT_X],n,MPIU_SCALAR,ds->work,size,&off,PetscObjectComm((PetscObject)ds));
292: }
293: if (eigr) {
294: MPI_Pack(eigr,1,MPIU_SCALAR,ds->work,size,&off,PetscObjectComm((PetscObject)ds));
295: }
296: if (eigi) {
297: MPI_Pack(eigi,1,MPIU_SCALAR,ds->work,size,&off,PetscObjectComm((PetscObject)ds));
298: }
299: }
300: MPI_Bcast(ds->work,size,MPI_BYTE,0,PetscObjectComm((PetscObject)ds));
301: if (rank) {
302: if (ds->state>=DS_STATE_CONDENSED) {
303: MPI_Unpack(ds->work,size,&off,ds->mat[DS_MAT_X],n,MPIU_SCALAR,PetscObjectComm((PetscObject)ds));
304: }
305: if (eigr) {
306: MPI_Unpack(ds->work,size,&off,eigr,1,MPIU_SCALAR,PetscObjectComm((PetscObject)ds));
307: }
308: if (eigi) {
309: MPI_Unpack(ds->work,size,&off,eigi,1,MPIU_SCALAR,PetscObjectComm((PetscObject)ds));
310: }
311: }
312: return(0);
313: }
315: static PetscErrorCode DSNEPSetFN_NEP(DS ds,PetscInt n,FN fn[])
316: {
318: DS_NEP *ctx = (DS_NEP*)ds->data;
319: PetscInt i;
322: if (n<=0) SETERRQ1(PetscObjectComm((PetscObject)ds),PETSC_ERR_ARG_OUTOFRANGE,"Must have one or more functions, you have %D",n);
323: if (n>DS_NUM_EXTRA) SETERRQ2(PetscObjectComm((PetscObject)ds),PETSC_ERR_ARG_OUTOFRANGE,"Too many functions, you specified %D but the limit is %D",n,DS_NUM_EXTRA);
324: if (ds->ld) { PetscInfo(ds,"DSNEPSetFN() called after DSAllocate()\n"); }
325: for (i=0;i<n;i++) {
326: PetscObjectReference((PetscObject)fn[i]);
327: }
328: for (i=0;i<ctx->nf;i++) {
329: FNDestroy(&ctx->f[i]);
330: }
331: for (i=0;i<n;i++) ctx->f[i] = fn[i];
332: ctx->nf = n;
333: return(0);
334: }
336: /*@
337: DSNEPSetFN - Sets a number of functions that define the nonlinear
338: eigenproblem.
340: Collective on ds
342: Input Parameters:
343: + ds - the direct solver context
344: . n - number of functions
345: - fn - array of functions
347: Notes:
348: The nonlinear eigenproblem is defined in terms of the split nonlinear
349: operator T(lambda) = sum_i A_i*f_i(lambda).
351: This function must be called before DSAllocate(). Then DSAllocate()
352: will allocate an extra matrix A_i per each function, that can be
353: filled in the usual way.
355: Level: advanced
357: .seealso: DSNEPGetFN(), DSAllocate()
358: @*/
359: PetscErrorCode DSNEPSetFN(DS ds,PetscInt n,FN fn[])
360: {
361: PetscInt i;
368: for (i=0;i<n;i++) {
371: }
372: PetscTryMethod(ds,"DSNEPSetFN_C",(DS,PetscInt,FN[]),(ds,n,fn));
373: return(0);
374: }
376: static PetscErrorCode DSNEPGetFN_NEP(DS ds,PetscInt k,FN *fn)
377: {
378: DS_NEP *ctx = (DS_NEP*)ds->data;
381: if (k<0 || k>=ctx->nf) SETERRQ1(PetscObjectComm((PetscObject)ds),PETSC_ERR_ARG_OUTOFRANGE,"k must be between 0 and %D",ctx->nf-1);
382: *fn = ctx->f[k];
383: return(0);
384: }
386: /*@
387: DSNEPGetFN - Gets the functions associated with the nonlinear DS.
389: Not collective, though parallel FNs are returned if the DS is parallel
391: Input Parameter:
392: + ds - the direct solver context
393: - k - the index of the requested function (starting in 0)
395: Output Parameter:
396: . fn - the function
398: Level: advanced
400: .seealso: DSNEPSetFN()
401: @*/
402: PetscErrorCode DSNEPGetFN(DS ds,PetscInt k,FN *fn)
403: {
409: PetscUseMethod(ds,"DSNEPGetFN_C",(DS,PetscInt,FN*),(ds,k,fn));
410: return(0);
411: }
413: static PetscErrorCode DSNEPGetNumFN_NEP(DS ds,PetscInt *n)
414: {
415: DS_NEP *ctx = (DS_NEP*)ds->data;
418: *n = ctx->nf;
419: return(0);
420: }
422: /*@
423: DSNEPGetNumFN - Returns the number of functions stored internally by
424: the DS.
426: Not collective
428: Input Parameter:
429: . ds - the direct solver context
431: Output Parameters:
432: . n - the number of functions passed in DSNEPSetFN()
434: Level: advanced
436: .seealso: DSNEPSetFN()
437: @*/
438: PetscErrorCode DSNEPGetNumFN(DS ds,PetscInt *n)
439: {
445: PetscUseMethod(ds,"DSNEPGetNumFN_C",(DS,PetscInt*),(ds,n));
446: return(0);
447: }
449: static PetscErrorCode DSNEPSetComputeMatrixFunction_NEP(DS ds,PetscErrorCode (*fun)(DS,PetscScalar,PetscBool,DSMatType,void*),void *ctx)
450: {
451: DS_NEP *dsctx = (DS_NEP*)ds->data;
454: dsctx->computematrix = fun;
455: dsctx->computematrixctx = ctx;
456: return(0);
457: }
459: /*@
460: DSNEPSetComputeMatrixFunction - Sets a user-provided subroutine to compute
461: the matrices T(lambda) or T'(lambda).
463: Logically Collective on ds
465: Input Parameters:
466: + ds - the direct solver context
467: . fun - a pointer to the user function
468: - ctx - a context pointer (the last parameter to the user function)
470: Calling Sequence of fun:
471: $ fun(DS ds,PetscScalar lambda,PetscBool deriv,DSMatType mat,void *ctx)
473: + ds - the direct solver object
474: . lambda - point where T(lambda) or T'(lambda) must be evaluated
475: . deriv - if true compute T'(lambda), otherwise compute T(lambda)
476: . mat - the DS matrix where the result must be stored
477: - ctx - optional context, as set by DSNEPSetComputeMatrixFunction()
479: Note:
480: The result is computed as T(lambda) = sum_i E_i*f_i(lambda), and similarly
481: for the derivative.
483: Level: developer
484: @*/
485: PetscErrorCode DSNEPSetComputeMatrixFunction(DS ds,PetscErrorCode (*fun)(DS,PetscScalar,PetscBool,DSMatType,void*),void *ctx)
486: {
491: PetscTryMethod(ds,"DSNEPSetComputeMatrixFunction_C",(DS,PetscErrorCode (*)(DS,PetscScalar,PetscBool,DSMatType,void*),void*),(ds,fun,ctx));
492: return(0);
493: }
495: PetscErrorCode DSDestroy_NEP(DS ds)
496: {
498: DS_NEP *ctx = (DS_NEP*)ds->data;
499: PetscInt i;
502: for (i=0;i<ctx->nf;i++) {
503: FNDestroy(&ctx->f[i]);
504: }
505: PetscFree(ds->data);
506: PetscObjectComposeFunction((PetscObject)ds,"DSNEPSetFN_C",NULL);
507: PetscObjectComposeFunction((PetscObject)ds,"DSNEPGetFN_C",NULL);
508: PetscObjectComposeFunction((PetscObject)ds,"DSNEPGetNumFN_C",NULL);
509: PetscObjectComposeFunction((PetscObject)ds,"DSNEPSetComputeMatrixFunction_C",NULL);
510: return(0);
511: }
513: SLEPC_EXTERN PetscErrorCode DSCreate_NEP(DS ds)
514: {
515: DS_NEP *ctx;
519: PetscNewLog(ds,&ctx);
520: ds->data = (void*)ctx;
522: ds->ops->allocate = DSAllocate_NEP;
523: ds->ops->view = DSView_NEP;
524: ds->ops->vectors = DSVectors_NEP;
525: ds->ops->solve[0] = DSSolve_NEP_SLP;
526: ds->ops->sort = DSSort_NEP;
527: ds->ops->synchronize = DSSynchronize_NEP;
528: ds->ops->destroy = DSDestroy_NEP;
529: PetscObjectComposeFunction((PetscObject)ds,"DSNEPSetFN_C",DSNEPSetFN_NEP);
530: PetscObjectComposeFunction((PetscObject)ds,"DSNEPGetFN_C",DSNEPGetFN_NEP);
531: PetscObjectComposeFunction((PetscObject)ds,"DSNEPGetNumFN_C",DSNEPGetNumFN_NEP);
532: PetscObjectComposeFunction((PetscObject)ds,"DSNEPSetComputeMatrixFunction_C",DSNEPSetComputeMatrixFunction_NEP);
533: return(0);
534: }