Line data Source code
1 : /*
2 : - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
3 : SLEPc - Scalable Library for Eigenvalue Problem Computations
4 : Copyright (c) 2002-, Universitat Politecnica de Valencia, Spain
5 :
6 : This file is part of SLEPc.
7 : SLEPc is distributed under a 2-clause BSD license (see LICENSE).
8 : - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
9 : */
10 : /*
11 : SVD routines related to the solution process
12 : */
13 :
14 : #include <slepc/private/svdimpl.h> /*I "slepcsvd.h" I*/
15 : #include <slepc/private/bvimpl.h>
16 :
17 : /*
18 : SVDComputeVectors_Left - Compute left singular vectors as U=A*V.
19 : Only done if the leftbasis flag is false. Assumes V is available.
20 : */
21 55 : PetscErrorCode SVDComputeVectors_Left(SVD svd)
22 : {
23 55 : Vec tl,omega2,u,v,w;
24 55 : PetscInt i,oldsize;
25 55 : VecType vtype;
26 55 : const PetscScalar* varray;
27 :
28 55 : PetscFunctionBegin;
29 55 : if (!svd->leftbasis) {
30 : /* generate left singular vectors on U */
31 40 : if (!svd->U) PetscCall(SVDGetBV(svd,NULL,&svd->U));
32 40 : PetscCall(BVGetSizes(svd->U,NULL,NULL,&oldsize));
33 40 : if (!oldsize) {
34 35 : if (!((PetscObject)svd->U)->type_name) PetscCall(BVSetType(svd->U,((PetscObject)svd->V)->type_name));
35 35 : PetscCall(MatCreateVecsEmpty(svd->A,NULL,&tl));
36 35 : PetscCall(BVSetSizesFromVec(svd->U,tl,svd->ncv));
37 35 : PetscCall(VecDestroy(&tl));
38 : }
39 40 : PetscCall(BVSetActiveColumns(svd->V,0,svd->nconv));
40 40 : PetscCall(BVSetActiveColumns(svd->U,0,svd->nconv));
41 40 : if (!svd->ishyperbolic) PetscCall(BVMatMult(svd->V,svd->A,svd->U));
42 12 : else if (svd->swapped) { /* compute right singular vectors as V=A'*Omega*U */
43 4 : PetscCall(MatCreateVecs(svd->A,&w,NULL));
44 120 : for (i=0;i<svd->nconv;i++) {
45 116 : PetscCall(BVGetColumn(svd->V,i,&v));
46 116 : PetscCall(BVGetColumn(svd->U,i,&u));
47 116 : PetscCall(VecPointwiseMult(w,v,svd->omega));
48 116 : PetscCall(MatMult(svd->A,w,u));
49 116 : PetscCall(BVRestoreColumn(svd->V,i,&v));
50 116 : PetscCall(BVRestoreColumn(svd->U,i,&u));
51 : }
52 4 : PetscCall(VecDestroy(&w));
53 : } else { /* compute left singular vectors as usual U=A*V, and set-up Omega-orthogonalization of U */
54 8 : PetscCall(BV_SetMatrixDiagonal(svd->U,svd->omega,svd->A));
55 8 : PetscCall(BVMatMult(svd->V,svd->A,svd->U));
56 : }
57 40 : PetscCall(BVOrthogonalize(svd->U,NULL));
58 40 : if (svd->ishyperbolic && !svd->swapped) { /* store signature after Omega-orthogonalization */
59 8 : PetscCall(MatGetVecType(svd->A,&vtype));
60 8 : PetscCall(VecCreate(PETSC_COMM_SELF,&omega2));
61 8 : PetscCall(VecSetSizes(omega2,svd->nconv,svd->nconv));
62 8 : PetscCall(VecSetType(omega2,vtype));
63 8 : PetscCall(BVGetSignature(svd->U,omega2));
64 8 : PetscCall(VecGetArrayRead(omega2,&varray));
65 354 : for (i=0;i<svd->nconv;i++) {
66 346 : svd->sign[i] = PetscRealPart(varray[i]);
67 346 : if (PetscRealPart(varray[i])<0.0) PetscCall(BVScaleColumn(svd->U,i,-1.0));
68 : }
69 8 : PetscCall(VecRestoreArrayRead(omega2,&varray));
70 8 : PetscCall(VecDestroy(&omega2));
71 : }
72 : }
73 55 : PetscFunctionReturn(PETSC_SUCCESS);
74 : }
75 :
76 2666 : PetscErrorCode SVDComputeVectors(SVD svd)
77 : {
78 2666 : PetscFunctionBegin;
79 2666 : SVDCheckSolved(svd,1);
80 2666 : if (svd->state==SVD_STATE_SOLVED) PetscTryTypeMethod(svd,computevectors);
81 2666 : svd->state = SVD_STATE_VECTORS;
82 2666 : PetscFunctionReturn(PETSC_SUCCESS);
83 : }
84 :
85 : /*@
86 : SVDSolve - Solves the singular value problem.
87 :
88 : Collective
89 :
90 : Input Parameter:
91 : . svd - singular value solver context obtained from SVDCreate()
92 :
93 : Options Database Keys:
94 : + -svd_view - print information about the solver used
95 : . -svd_view_mat0 - view the first matrix (A)
96 : . -svd_view_mat1 - view the second matrix (B)
97 : . -svd_view_signature - view the signature matrix (omega)
98 : . -svd_view_vectors - view the computed singular vectors
99 : . -svd_view_values - view the computed singular values
100 : . -svd_converged_reason - print reason for convergence, and number of iterations
101 : . -svd_error_absolute - print absolute errors of each singular triplet
102 : . -svd_error_relative - print relative errors of each singular triplet
103 : - -svd_error_norm - print errors relative to the matrix norms of each singular triplet
104 :
105 : Notes:
106 : All the command-line options listed above admit an optional argument specifying
107 : the viewer type and options. For instance, use '-svd_view_mat0 binary:amatrix.bin'
108 : to save the A matrix to a binary file, '-svd_view_values draw' to draw the computed
109 : singular values graphically, or '-svd_error_relative :myerr.m:ascii_matlab' to save
110 : the errors in a file that can be executed in Matlab.
111 :
112 : Level: beginner
113 :
114 : .seealso: SVDCreate(), SVDSetUp(), SVDDestroy()
115 : @*/
116 274 : PetscErrorCode SVDSolve(SVD svd)
117 : {
118 274 : PetscInt i,*workperm;
119 :
120 274 : PetscFunctionBegin;
121 274 : PetscValidHeaderSpecific(svd,SVD_CLASSID,1);
122 274 : if (svd->state>=SVD_STATE_SOLVED) PetscFunctionReturn(PETSC_SUCCESS);
123 274 : PetscCall(PetscLogEventBegin(SVD_Solve,svd,0,0,0));
124 :
125 : /* call setup */
126 274 : PetscCall(SVDSetUp(svd));
127 274 : svd->its = 0;
128 274 : svd->nconv = 0;
129 6221 : for (i=0;i<svd->ncv;i++) {
130 5947 : svd->sigma[i] = 0.0;
131 5947 : svd->errest[i] = 0.0;
132 5947 : svd->perm[i] = i;
133 : }
134 274 : PetscCall(SVDViewFromOptions(svd,NULL,"-svd_view_pre"));
135 :
136 274 : switch (svd->problem_type) {
137 140 : case SVD_STANDARD:
138 140 : PetscUseTypeMethod(svd,solve);
139 : break;
140 98 : case SVD_GENERALIZED:
141 98 : PetscUseTypeMethod(svd,solveg);
142 : break;
143 36 : case SVD_HYPERBOLIC:
144 36 : PetscUseTypeMethod(svd,solveh);
145 : break;
146 : }
147 274 : svd->state = SVD_STATE_SOLVED;
148 :
149 : /* sort singular triplets */
150 274 : if (svd->which == SVD_SMALLEST) PetscCall(PetscSortRealWithPermutation(svd->nconv,svd->sigma,svd->perm));
151 : else {
152 235 : PetscCall(PetscMalloc1(svd->nconv,&workperm));
153 2946 : for (i=0;i<svd->nconv;i++) workperm[i] = i;
154 235 : PetscCall(PetscSortRealWithPermutation(svd->nconv,svd->sigma,workperm));
155 2946 : for (i=0;i<svd->nconv;i++) svd->perm[i] = workperm[svd->nconv-i-1];
156 235 : PetscCall(PetscFree(workperm));
157 : }
158 274 : PetscCall(PetscLogEventEnd(SVD_Solve,svd,0,0,0));
159 :
160 : /* various viewers */
161 274 : PetscCall(SVDViewFromOptions(svd,NULL,"-svd_view"));
162 274 : PetscCall(SVDConvergedReasonViewFromOptions(svd));
163 274 : PetscCall(SVDErrorViewFromOptions(svd));
164 274 : PetscCall(SVDValuesViewFromOptions(svd));
165 274 : PetscCall(SVDVectorsViewFromOptions(svd));
166 274 : PetscCall(MatViewFromOptions(svd->OP,(PetscObject)svd,"-svd_view_mat0"));
167 274 : if (svd->isgeneralized) PetscCall(MatViewFromOptions(svd->OPb,(PetscObject)svd,"-svd_view_mat1"));
168 274 : if (svd->ishyperbolic) PetscCall(VecViewFromOptions(svd->omega,(PetscObject)svd,"-svd_view_signature"));
169 :
170 : /* Remove the initial subspaces */
171 274 : svd->nini = 0;
172 274 : svd->ninil = 0;
173 274 : PetscFunctionReturn(PETSC_SUCCESS);
174 : }
175 :
176 : /*@
177 : SVDGetIterationNumber - Gets the current iteration number. If the
178 : call to SVDSolve() is complete, then it returns the number of iterations
179 : carried out by the solution method.
180 :
181 : Not Collective
182 :
183 : Input Parameter:
184 : . svd - the singular value solver context
185 :
186 : Output Parameter:
187 : . its - number of iterations
188 :
189 : Note:
190 : During the i-th iteration this call returns i-1. If SVDSolve() is
191 : complete, then parameter "its" contains either the iteration number at
192 : which convergence was successfully reached, or failure was detected.
193 : Call SVDGetConvergedReason() to determine if the solver converged or
194 : failed and why.
195 :
196 : Level: intermediate
197 :
198 : .seealso: SVDGetConvergedReason(), SVDSetTolerances()
199 : @*/
200 62 : PetscErrorCode SVDGetIterationNumber(SVD svd,PetscInt *its)
201 : {
202 62 : PetscFunctionBegin;
203 62 : PetscValidHeaderSpecific(svd,SVD_CLASSID,1);
204 62 : PetscAssertPointer(its,2);
205 62 : *its = svd->its;
206 62 : PetscFunctionReturn(PETSC_SUCCESS);
207 : }
208 :
209 : /*@
210 : SVDGetConvergedReason - Gets the reason why the SVDSolve() iteration was
211 : stopped.
212 :
213 : Not Collective
214 :
215 : Input Parameter:
216 : . svd - the singular value solver context
217 :
218 : Output Parameter:
219 : . reason - negative value indicates diverged, positive value converged
220 : (see SVDConvergedReason)
221 :
222 : Options Database Key:
223 : . -svd_converged_reason - print the reason to a viewer
224 :
225 : Notes:
226 : Possible values for reason are
227 : + SVD_CONVERGED_TOL - converged up to tolerance
228 : . SVD_CONVERGED_USER - converged due to a user-defined condition
229 : . SVD_CONVERGED_MAXIT - reached the maximum number of iterations with SVD_CONV_MAXIT criterion
230 : . SVD_DIVERGED_ITS - required more than max_it iterations to reach convergence
231 : . SVD_DIVERGED_BREAKDOWN - generic breakdown in method
232 : - SVD_DIVERGED_SYMMETRY_LOST - underlying indefinite eigensolver was not able to keep symmetry
233 :
234 : Can only be called after the call to SVDSolve() is complete.
235 :
236 : Level: intermediate
237 :
238 : .seealso: SVDSetTolerances(), SVDSolve(), SVDConvergedReason
239 : @*/
240 13 : PetscErrorCode SVDGetConvergedReason(SVD svd,SVDConvergedReason *reason)
241 : {
242 13 : PetscFunctionBegin;
243 13 : PetscValidHeaderSpecific(svd,SVD_CLASSID,1);
244 13 : PetscAssertPointer(reason,2);
245 13 : SVDCheckSolved(svd,1);
246 13 : *reason = svd->reason;
247 13 : PetscFunctionReturn(PETSC_SUCCESS);
248 : }
249 :
250 : /*@
251 : SVDGetConverged - Gets the number of converged singular values.
252 :
253 : Not Collective
254 :
255 : Input Parameter:
256 : . svd - the singular value solver context
257 :
258 : Output Parameter:
259 : . nconv - number of converged singular values
260 :
261 : Note:
262 : This function should be called after SVDSolve() has finished.
263 :
264 : Level: beginner
265 :
266 : .seealso: SVDSetDimensions(), SVDSolve(), SVDGetSingularTriplet()
267 : @*/
268 127 : PetscErrorCode SVDGetConverged(SVD svd,PetscInt *nconv)
269 : {
270 127 : PetscFunctionBegin;
271 127 : PetscValidHeaderSpecific(svd,SVD_CLASSID,1);
272 127 : PetscAssertPointer(nconv,2);
273 127 : SVDCheckSolved(svd,1);
274 127 : *nconv = svd->nconv;
275 127 : PetscFunctionReturn(PETSC_SUCCESS);
276 : }
277 :
278 : /*@
279 : SVDGetSingularTriplet - Gets the i-th triplet of the singular value decomposition
280 : as computed by SVDSolve(). The solution consists in the singular value and its left
281 : and right singular vectors.
282 :
283 : Collective
284 :
285 : Input Parameters:
286 : + svd - singular value solver context
287 : - i - index of the solution
288 :
289 : Output Parameters:
290 : + sigma - singular value
291 : . u - left singular vector
292 : - v - right singular vector
293 :
294 : Note:
295 : Both u or v can be NULL if singular vectors are not required.
296 : Otherwise, the caller must provide valid Vec objects, i.e.,
297 : they must be created by the calling program with e.g. MatCreateVecs().
298 :
299 : The index i should be a value between 0 and nconv-1 (see SVDGetConverged()).
300 : Singular triplets are indexed according to the ordering criterion established
301 : with SVDSetWhichSingularTriplets().
302 :
303 : In the case of GSVD, the solution consists in three vectors u,v,x that are
304 : returned as follows. Vector x is returned in the right singular vector
305 : (argument v) and has length equal to the number of columns of A and B.
306 : The other two vectors are returned stacked on top of each other [u;v] in
307 : the left singular vector argument, with length equal to m+n (number of rows
308 : of A plus number of rows of B).
309 :
310 : Level: beginner
311 :
312 : .seealso: SVDSolve(), SVDGetConverged(), SVDSetWhichSingularTriplets()
313 : @*/
314 3972 : PetscErrorCode SVDGetSingularTriplet(SVD svd,PetscInt i,PetscReal *sigma,Vec u,Vec v)
315 : {
316 3972 : PetscInt M,N;
317 3972 : Vec w;
318 :
319 3972 : PetscFunctionBegin;
320 3972 : PetscValidHeaderSpecific(svd,SVD_CLASSID,1);
321 11916 : PetscValidLogicalCollectiveInt(svd,i,2);
322 3972 : SVDCheckSolved(svd,1);
323 3972 : if (u) { PetscValidHeaderSpecific(u,VEC_CLASSID,4); PetscCheckSameComm(svd,1,u,4); }
324 3972 : if (v) { PetscValidHeaderSpecific(v,VEC_CLASSID,5); PetscCheckSameComm(svd,1,v,5); }
325 3972 : PetscCheck(i>=0,PetscObjectComm((PetscObject)svd),PETSC_ERR_ARG_OUTOFRANGE,"The index cannot be negative");
326 3972 : PetscCheck(i<svd->nconv,PetscObjectComm((PetscObject)svd),PETSC_ERR_ARG_OUTOFRANGE,"The index can be nconv-1 at most, see SVDGetConverged()");
327 3972 : if (sigma) *sigma = svd->sigma[svd->perm[i]];
328 3972 : if (u || v) {
329 2665 : if (!svd->isgeneralized) {
330 2076 : PetscCall(MatGetSize(svd->OP,&M,&N));
331 2076 : if (M<N) { w = u; u = v; v = w; }
332 : }
333 2665 : PetscCall(SVDComputeVectors(svd));
334 2665 : if (u) PetscCall(BVCopyVec(svd->U,svd->perm[i],u));
335 2665 : if (v) PetscCall(BVCopyVec(svd->V,svd->perm[i],v));
336 : }
337 3972 : PetscFunctionReturn(PETSC_SUCCESS);
338 : }
339 :
340 : /*
341 : SVDComputeResidualNorms_Standard - Computes the norms of the left and
342 : right residuals associated with the i-th computed singular triplet.
343 :
344 : Input Parameters:
345 : sigma - singular value
346 : u,v - singular vectors
347 : x,y - two work vectors with the same dimensions as u,v
348 : */
349 425 : static PetscErrorCode SVDComputeResidualNorms_Standard(SVD svd,PetscReal sigma,Vec u,Vec v,Vec x,Vec y,PetscReal *norm1,PetscReal *norm2)
350 : {
351 425 : PetscInt M,N;
352 :
353 425 : PetscFunctionBegin;
354 : /* norm1 = ||A*v-sigma*u||_2 */
355 425 : if (norm1) {
356 425 : PetscCall(MatMult(svd->OP,v,x));
357 425 : PetscCall(VecAXPY(x,-sigma,u));
358 425 : PetscCall(VecNorm(x,NORM_2,norm1));
359 : }
360 : /* norm2 = ||A^T*u-sigma*v||_2 */
361 425 : if (norm2) {
362 425 : PetscCall(MatGetSize(svd->OP,&M,&N));
363 425 : if (M<N) PetscCall(MatMult(svd->A,u,y));
364 354 : else PetscCall(MatMult(svd->AT,u,y));
365 425 : PetscCall(VecAXPY(y,-sigma,v));
366 425 : PetscCall(VecNorm(y,NORM_2,norm2));
367 : }
368 425 : PetscFunctionReturn(PETSC_SUCCESS);
369 : }
370 :
371 : /*
372 : SVDComputeResidualNorms_Generalized - In GSVD, compute the residual norms
373 : norm1 = ||s^2*A'*u-c*B'*B*x||_2 and norm2 = ||c^2*B'*v-s*A'*A*x||_2.
374 :
375 : Input Parameters:
376 : sigma - singular value
377 : uv - left singular vectors [u;v]
378 : x - right singular vector
379 : y,z - two work vectors with the same dimension as x
380 : */
381 388 : static PetscErrorCode SVDComputeResidualNorms_Generalized(SVD svd,PetscReal sigma,Vec uv,Vec x,Vec y,Vec z,PetscReal *norm1,PetscReal *norm2)
382 : {
383 388 : Vec u,v,ax,bx,nest,aux[2];
384 388 : PetscReal c,s;
385 :
386 388 : PetscFunctionBegin;
387 388 : PetscCall(MatCreateVecs(svd->OP,NULL,&u));
388 388 : PetscCall(MatCreateVecs(svd->OPb,NULL,&v));
389 388 : aux[0] = u;
390 388 : aux[1] = v;
391 388 : PetscCall(VecCreateNest(PetscObjectComm((PetscObject)svd),2,NULL,aux,&nest));
392 388 : PetscCall(VecCopy(uv,nest));
393 :
394 388 : s = 1.0/PetscSqrtReal(1.0+sigma*sigma);
395 388 : c = sigma*s;
396 :
397 : /* norm1 = ||s^2*A'*u-c*B'*B*x||_2 */
398 388 : if (norm1) {
399 388 : PetscCall(VecDuplicate(v,&bx));
400 388 : PetscCall(MatMultHermitianTranspose(svd->OP,u,z));
401 388 : PetscCall(MatMult(svd->OPb,x,bx));
402 388 : PetscCall(MatMultHermitianTranspose(svd->OPb,bx,y));
403 388 : PetscCall(VecAXPBY(y,s*s,-c,z));
404 388 : PetscCall(VecNorm(y,NORM_2,norm1));
405 388 : PetscCall(VecDestroy(&bx));
406 : }
407 : /* norm2 = ||c^2*B'*v-s*A'*A*x||_2 */
408 388 : if (norm2) {
409 388 : PetscCall(VecDuplicate(u,&ax));
410 388 : PetscCall(MatMultHermitianTranspose(svd->OPb,v,z));
411 388 : PetscCall(MatMult(svd->OP,x,ax));
412 388 : PetscCall(MatMultHermitianTranspose(svd->OP,ax,y));
413 388 : PetscCall(VecAXPBY(y,c*c,-s,z));
414 388 : PetscCall(VecNorm(y,NORM_2,norm2));
415 388 : PetscCall(VecDestroy(&ax));
416 : }
417 :
418 388 : PetscCall(VecDestroy(&v));
419 388 : PetscCall(VecDestroy(&u));
420 388 : PetscCall(VecDestroy(&nest));
421 388 : PetscFunctionReturn(PETSC_SUCCESS);
422 : }
423 :
424 : /*
425 : SVDComputeResidualNorms_Hyperbolic - Computes the norms of the left and
426 : right residuals associated with the i-th computed singular triplet.
427 :
428 : Input Parameters:
429 : sigma - singular value
430 : sign - corresponding element of the signature Omega2
431 : u,v - singular vectors
432 : x,y,z - three work vectors with the same dimensions as u,v,u
433 : */
434 691 : static PetscErrorCode SVDComputeResidualNorms_Hyperbolic(SVD svd,PetscReal sigma,PetscReal sign,Vec u,Vec v,Vec x,Vec y,Vec z,PetscReal *norm1,PetscReal *norm2)
435 : {
436 691 : PetscInt M,N;
437 :
438 691 : PetscFunctionBegin;
439 : /* norm1 = ||A*v-sigma*u||_2 */
440 691 : if (norm1) {
441 691 : PetscCall(MatMult(svd->OP,v,x));
442 691 : PetscCall(VecAXPY(x,-sigma,u));
443 691 : PetscCall(VecNorm(x,NORM_2,norm1));
444 : }
445 : /* norm2 = ||A^T*Omega*u-sigma*sign*v||_2 */
446 691 : if (norm2) {
447 691 : PetscCall(MatGetSize(svd->OP,&M,&N));
448 691 : PetscCall(VecPointwiseMult(z,u,svd->omega));
449 691 : if (M<N) PetscCall(MatMult(svd->A,z,y));
450 631 : else PetscCall(MatMult(svd->AT,z,y));
451 691 : PetscCall(VecAXPY(y,-sigma*sign,v));
452 691 : PetscCall(VecNorm(y,NORM_2,norm2));
453 : }
454 691 : PetscFunctionReturn(PETSC_SUCCESS);
455 : }
456 :
457 : /*@
458 : SVDComputeError - Computes the error (based on the residual norm) associated
459 : with the i-th singular triplet.
460 :
461 : Collective
462 :
463 : Input Parameters:
464 : + svd - the singular value solver context
465 : . i - the solution index
466 : - type - the type of error to compute
467 :
468 : Output Parameter:
469 : . error - the error
470 :
471 : Notes:
472 : The error can be computed in various ways, all of them based on the residual
473 : norm obtained as sqrt(n1^2+n2^2) with n1 = ||A*v-sigma*u||_2 and
474 : n2 = ||A^T*u-sigma*v||_2, where sigma is the singular value, u is the left
475 : singular vector and v is the right singular vector.
476 :
477 : In the case of the GSVD, the two components of the residual norm are
478 : n1 = ||s^2*A'*u-c*B'*B*x||_2 and n2 = ||c^2*B'*v-s*A'*A*x||_2, where [u;v]
479 : are the left singular vectors and x is the right singular vector, with
480 : sigma=c/s.
481 :
482 : Level: beginner
483 :
484 : .seealso: SVDErrorType, SVDSolve()
485 : @*/
486 1504 : PetscErrorCode SVDComputeError(SVD svd,PetscInt i,SVDErrorType type,PetscReal *error)
487 : {
488 1504 : PetscReal sigma,norm1,norm2,c,s;
489 1504 : Vec u=NULL,v=NULL,x=NULL,y=NULL,z=NULL;
490 1504 : PetscReal vecnorm=1.0;
491 :
492 1504 : PetscFunctionBegin;
493 1504 : PetscValidHeaderSpecific(svd,SVD_CLASSID,1);
494 4512 : PetscValidLogicalCollectiveInt(svd,i,2);
495 4512 : PetscValidLogicalCollectiveEnum(svd,type,3);
496 1504 : PetscAssertPointer(error,4);
497 1504 : SVDCheckSolved(svd,1);
498 :
499 : /* allocate work vectors */
500 1504 : switch (svd->problem_type) {
501 425 : case SVD_STANDARD:
502 425 : PetscCall(SVDSetWorkVecs(svd,2,2));
503 425 : u = svd->workl[0];
504 425 : v = svd->workr[0];
505 425 : x = svd->workl[1];
506 425 : y = svd->workr[1];
507 425 : break;
508 388 : case SVD_GENERALIZED:
509 388 : PetscCall(SVDSetWorkVecs(svd,1,3));
510 388 : u = svd->workl[0];
511 388 : v = svd->workr[0];
512 388 : x = svd->workr[1];
513 388 : y = svd->workr[2];
514 388 : break;
515 691 : case SVD_HYPERBOLIC:
516 691 : PetscCall(SVDSetWorkVecs(svd,3,2));
517 691 : u = svd->workl[0];
518 691 : v = svd->workr[0];
519 691 : x = svd->workl[1];
520 691 : y = svd->workr[1];
521 691 : z = svd->workl[2];
522 691 : break;
523 : }
524 :
525 : /* compute residual norm */
526 1504 : PetscCall(SVDGetSingularTriplet(svd,i,&sigma,u,v));
527 1504 : switch (svd->problem_type) {
528 425 : case SVD_STANDARD:
529 425 : PetscCall(SVDComputeResidualNorms_Standard(svd,sigma,u,v,x,y,&norm1,&norm2));
530 : break;
531 388 : case SVD_GENERALIZED:
532 388 : PetscCall(SVDComputeResidualNorms_Generalized(svd,sigma,u,v,x,y,&norm1,&norm2));
533 : break;
534 691 : case SVD_HYPERBOLIC:
535 691 : PetscCall(SVDComputeResidualNorms_Hyperbolic(svd,sigma,svd->sign[svd->perm[i]],u,v,x,y,z,&norm1,&norm2));
536 : break;
537 : }
538 1504 : *error = SlepcAbs(norm1,norm2);
539 :
540 : /* compute 2-norm of eigenvector of the cyclic form */
541 1504 : if (type!=SVD_ERROR_ABSOLUTE) {
542 1491 : switch (svd->problem_type) {
543 412 : case SVD_STANDARD:
544 412 : vecnorm = PETSC_SQRT2;
545 412 : break;
546 388 : case SVD_GENERALIZED:
547 388 : PetscCall(VecNorm(v,NORM_2,&vecnorm));
548 388 : vecnorm = PetscSqrtReal(1.0+vecnorm*vecnorm);
549 388 : break;
550 691 : case SVD_HYPERBOLIC:
551 691 : PetscCall(VecNorm(u,NORM_2,&vecnorm));
552 691 : vecnorm = PetscSqrtReal(1.0+vecnorm*vecnorm);
553 691 : break;
554 : }
555 13 : }
556 :
557 : /* compute error */
558 1504 : switch (type) {
559 : case SVD_ERROR_ABSOLUTE:
560 : break;
561 405 : case SVD_ERROR_RELATIVE:
562 405 : if (svd->isgeneralized) {
563 0 : s = 1.0/PetscSqrtReal(1.0+sigma*sigma);
564 0 : c = sigma*s;
565 0 : norm1 /= c*vecnorm;
566 0 : norm2 /= s*vecnorm;
567 0 : *error = PetscMax(norm1,norm2);
568 405 : } else *error /= sigma*vecnorm;
569 : break;
570 1086 : case SVD_ERROR_NORM:
571 1086 : if (!svd->nrma) PetscCall(MatNorm(svd->OP,NORM_INFINITY,&svd->nrma));
572 1086 : if (svd->isgeneralized && !svd->nrmb) PetscCall(MatNorm(svd->OPb,NORM_INFINITY,&svd->nrmb));
573 1086 : *error /= PetscMax(svd->nrma,svd->nrmb)*vecnorm;
574 1086 : break;
575 0 : default:
576 0 : SETERRQ(PetscObjectComm((PetscObject)svd),PETSC_ERR_ARG_OUTOFRANGE,"Invalid error type");
577 : }
578 1504 : PetscFunctionReturn(PETSC_SUCCESS);
579 : }
|
