Dune Core Modules (2.9.1)

fmatrixev.hh
Go to the documentation of this file.
1// -*- tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 2 -*-
2// vi: set et ts=4 sw=2 sts=2:
3// SPDX-FileCopyrightInfo: Copyright (C) DUNE Project contributors, see file LICENSE.md in module root
4// SPDX-License-Identifier: LicenseRef-GPL-2.0-only-with-DUNE-exception
5#ifndef DUNE_FMATRIXEIGENVALUES_HH
6#define DUNE_FMATRIXEIGENVALUES_HH
7
12#include <algorithm>
13#include <iostream>
14#include <cmath>
15#include <cassert>
16
20#include <dune/common/math.hh>
21
22namespace Dune {
23
29 namespace FMatrixHelp {
30
31#if HAVE_LAPACK
32 // defined in fmatrixev.cc
33 extern void eigenValuesLapackCall(
34 const char* jobz, const char* uplo, const long
35 int* n, double* a, const long int* lda, double* w,
36 double* work, const long int* lwork, long int* info);
37
38 extern void eigenValuesNonsymLapackCall(
39 const char* jobvl, const char* jobvr, const long
40 int* n, double* a, const long int* lda, double* wr, double* wi, double* vl,
41 const long int* ldvl, double* vr, const long int* ldvr, double* work,
42 const long int* lwork, long int* info);
43
44 extern void eigenValuesLapackCall(
45 const char* jobz, const char* uplo, const long
46 int* n, float* a, const long int* lda, float* w,
47 float* work, const long int* lwork, long int* info);
48
49 extern void eigenValuesNonsymLapackCall(
50 const char* jobvl, const char* jobvr, const long
51 int* n, float* a, const long int* lda, float* wr, float* wi, float* vl,
52 const long int* ldvl, float* vr, const long int* ldvr, float* work,
53 const long int* lwork, long int* info);
54
55#endif
56
57 namespace Impl {
58 //internal tag to activate/disable code for eigenvector calculation at compile time
59 enum Jobs { OnlyEigenvalues=0, EigenvaluesEigenvectors=1 };
60
61 //internal dummy used if only eigenvalues are to be calculated
62 template<typename K, int dim>
63 using EVDummy = FieldMatrix<K, dim, dim>;
64
65 //compute the cross-product of two vectors
66 template<typename K>
67 inline FieldVector<K,3> crossProduct(const FieldVector<K,3>& vec0, const FieldVector<K,3>& vec1) {
68 return {vec0[1]*vec1[2] - vec0[2]*vec1[1], vec0[2]*vec1[0] - vec0[0]*vec1[2], vec0[0]*vec1[1] - vec0[1]*vec1[0]};
69 }
70
71 template <typename K>
72 static void eigenValues2dImpl(const FieldMatrix<K, 2, 2>& matrix,
73 FieldVector<K, 2>& eigenvalues)
74 {
75 using std::sqrt;
76 const K p = 0.5 * (matrix[0][0] + matrix [1][1]);
77 const K p2 = p - matrix[1][1];
78 K q = p2 * p2 + matrix[1][0] * matrix[0][1];
79 if( q < 0 && q > -1e-14 ) q = 0;
80 if (q < 0)
81 {
82 std::cout << matrix << std::endl;
83 // Complex eigenvalues are either caused by non-symmetric matrices or by round-off errors
84 DUNE_THROW(MathError, "Complex eigenvalue detected (which this implementation cannot handle).");
85 }
86
87 // get square root
88 q = sqrt(q);
89
90 // store eigenvalues in ascending order
91 eigenvalues[0] = p - q;
92 eigenvalues[1] = p + q;
93 }
94
95 /*
96 This implementation was adapted from the pseudo-code (Python?) implementation found on
97 http://en.wikipedia.org/wiki/Eigenvalue_algorithm (retrieved late August 2014).
98 Wikipedia claims to have taken it from
99 Smith, Oliver K. (April 1961), Eigenvalues of a symmetric 3 × 3 matrix.,
100 Communications of the ACM 4 (4): 168, doi:10.1145/355578.366316
101 */
102 template <typename K>
103 static K eigenValues3dImpl(const FieldMatrix<K, 3, 3>& matrix,
104 FieldVector<K, 3>& eigenvalues)
105 {
106 using std::sqrt;
107 using std::acos;
108 using real_type = typename FieldTraits<K>::real_type;
109 const K pi = MathematicalConstants<K>::pi();
110 K p1 = matrix[0][1]*matrix[0][1] + matrix[0][2]*matrix[0][2] + matrix[1][2]*matrix[1][2];
111
112 if (p1 <= std::numeric_limits<K>::epsilon()) {
113 // A is diagonal.
114 eigenvalues[0] = matrix[0][0];
115 eigenvalues[1] = matrix[1][1];
116 eigenvalues[2] = matrix[2][2];
117 std::sort(eigenvalues.begin(), eigenvalues.end());
118
119 return 0.0;
120 }
121 else
122 {
123 // q = trace(A)/3
124 K q = 0;
125 for (int i=0; i<3; i++)
126 q += matrix[i][i] / 3.0;
127
128 K p2 = (matrix[0][0] - q)*(matrix[0][0] - q) + (matrix[1][1] - q)*(matrix[1][1] - q) + (matrix[2][2] - q)*(matrix[2][2] - q) + 2.0 * p1;
129 K p = sqrt(p2 / 6);
130 // B = (1 / p) * (A - q * I); // I is the identity matrix
131 FieldMatrix<K,3,3> B;
132 for (int i=0; i<3; i++)
133 for (int j=0; j<3; j++)
134 B[i][j] = (real_type(1.0)/p) * (matrix[i][j] - q*(i==j));
135
136 K r = B.determinant() / 2.0;
137
138 /*In exact arithmetic for a symmetric matrix -1 <= r <= 1
139 but computation error can leave it slightly outside this range.
140 acos(z) function requires |z| <= 1, but will fail silently
141 and return NaN if the input is larger than 1 in magnitude.
142 Thus r is clamped to [-1,1].*/
143 using std::clamp;
144 r = clamp<K>(r, -1.0, 1.0);
145 K phi = acos(r) / 3.0;
146
147 // the eigenvalues satisfy eig[2] <= eig[1] <= eig[0]
148 eigenvalues[2] = q + 2 * p * cos(phi);
149 eigenvalues[0] = q + 2 * p * cos(phi + (2*pi/3));
150 eigenvalues[1] = 3 * q - eigenvalues[0] - eigenvalues[2]; // since trace(matrix) = eig1 + eig2 + eig3
151
152 return r;
153 }
154 }
155
156 //see https://www.geometrictools.com/Documentation/RobustEigenSymmetric3x3.pdf
157 //Robustly compute a right-handed orthonormal set {u, v, evec0}.
158 template<typename K>
159 void orthoComp(const FieldVector<K,3>& evec0, FieldVector<K,3>& u, FieldVector<K,3>& v) {
160 using std::abs;
161 if(abs(evec0[0]) > abs(evec0[1])) {
162 //The component of maximum absolute value is either evec0[0] or evec0[2].
163 FieldVector<K,2> temp = {evec0[0], evec0[2]};
164 auto L = 1.0 / temp.two_norm();
165 u = L * FieldVector<K,3>({-evec0[2], 0.0, evec0[0]});
166 }
167 else {
168 //The component of maximum absolute value is either evec0[1] or evec0[2].
169 FieldVector<K,2> temp = {evec0[1], evec0[2]};
170 auto L = 1.0 / temp.two_norm();
171 u = L * FieldVector<K,3>({0.0, evec0[2], -evec0[1]});
172 }
173 v = crossProduct(evec0, u);
174 }
175
176 //see https://www.geometrictools.com/Documentation/RobustEigenSymmetric3x3.pdf
177 template<typename K>
178 void eig0(const FieldMatrix<K,3,3>& matrix, K eval0, FieldVector<K,3>& evec0) {
179 /* Compute a unit-length eigenvector for eigenvalue[i0]. The
180 matrix is rank 2, so two of the rows are linearly independent.
181 For a robust computation of the eigenvector, select the two
182 rows whose cross product has largest length of all pairs of
183 rows. */
184 using Vector = FieldVector<K,3>;
185 Vector row0 = {matrix[0][0]-eval0, matrix[0][1], matrix[0][2]};
186 Vector row1 = {matrix[1][0], matrix[1][1]-eval0, matrix[1][2]};
187 Vector row2 = {matrix[2][0], matrix[2][1], matrix[2][2]-eval0};
188
189 Vector r0xr1 = crossProduct(row0, row1);
190 Vector r0xr2 = crossProduct(row0, row2);
191 Vector r1xr2 = crossProduct(row1, row2);
192 auto d0 = r0xr1.two_norm();
193 auto d1 = r0xr2.two_norm();
194 auto d2 = r1xr2.two_norm();
195
196 auto dmax = d0 ;
197 int imax = 0;
198 if(d1>dmax) {
199 dmax = d1;
200 imax = 1;
201 }
202 if(d2>dmax)
203 imax = 2;
204
205 if(imax == 0)
206 evec0 = r0xr1 / d0;
207 else if(imax == 1)
208 evec0 = r0xr2 / d1;
209 else
210 evec0 = r1xr2 / d2;
211 }
212
213 //see https://www.geometrictools.com/Documentation/RobustEigenSymmetric3x3.pdf
214 template<typename K>
215 void eig1(const FieldMatrix<K,3,3>& matrix, const FieldVector<K,3>& evec0, FieldVector<K,3>& evec1, K eval1) {
216 using Vector = FieldVector<K,3>;
217
218 //Robustly compute a right-handed orthonormal set {u, v, evec0}.
219 Vector u,v;
220 orthoComp(evec0, u, v);
221
222 /* Let e be eval1 and let E be a corresponding eigenvector which
223 is a solution to the linear system (A - e*I)*E = 0. The matrix
224 (A - e*I) is 3x3, not invertible (so infinitely many
225 solutions), and has rank 2 when eval1 and eval are different.
226 It has rank 1 when eval1 and eval2 are equal. Numerically, it
227 is difficult to compute robustly the rank of a matrix. Instead,
228 the 3x3 linear system is reduced to a 2x2 system as follows.
229 Define the 3x2 matrix J = [u,v] whose columns are the u and v
230 computed previously. Define the 2x1 vector X = J*E. The 2x2
231 system is 0 = M * X = (J^T * (A - e*I) * J) * X where J^T is
232 the transpose of J and M = J^T * (A - e*I) * J is a 2x2 matrix.
233 The system may be written as
234 +- -++- -+ +- -+
235 | U^T*A*U - e U^T*A*V || x0 | = e * | x0 |
236 | V^T*A*U V^T*A*V - e || x1 | | x1 |
237 +- -++ -+ +- -+
238 where X has row entries x0 and x1. */
239
240 Vector Au, Av;
241 matrix.mv(u, Au);
242 matrix.mv(v, Av);
243
244 auto m00 = u.dot(Au) - eval1;
245 auto m01 = u.dot(Av);
246 auto m11 = v.dot(Av) - eval1;
247
248 /* For robustness, choose the largest-length row of M to compute
249 the eigenvector. The 2-tuple of coefficients of U and V in the
250 assignments to eigenvector[1] lies on a circle, and U and V are
251 unit length and perpendicular, so eigenvector[1] is unit length
252 (within numerical tolerance). */
253 using std::abs, std::sqrt, std::max;
254 auto absM00 = abs(m00);
255 auto absM01 = abs(m01);
256 auto absM11 = abs(m11);
257 if(absM00 >= absM11) {
258 auto maxAbsComp = max(absM00, absM01);
259 if(maxAbsComp > 0.0) {
260 if(absM00 >= absM01) {
261 m01 /= m00;
262 m00 = 1.0 / sqrt(1.0 + m01*m01);
263 m01 *= m00;
264 }
265 else {
266 m00 /= m01;
267 m01 = 1.0 / sqrt(1.0 + m00*m00);
268 m00 *= m01;
269 }
270 evec1 = m01*u - m00*v;
271 }
272 else
273 evec1 = u;
274 }
275 else {
276 auto maxAbsComp = max(absM11, absM01);
277 if(maxAbsComp > 0.0) {
278 if(absM11 >= absM01) {
279 m01 /= m11;
280 m11 = 1.0 / sqrt(1.0 + m01*m01);
281 m01 *= m11;
282 }
283 else {
284 m11 /= m01;
285 m01 = 1.0 / sqrt(1.0 + m11*m11);
286 m11 *= m01;
287 }
288 evec1 = m11*u - m01*v;
289 }
290 else
291 evec1 = u;
292 }
293 }
294
295 // 1d specialization
296 template<Jobs Tag, typename K>
297 static void eigenValuesVectorsImpl(const FieldMatrix<K, 1, 1>& matrix,
298 FieldVector<K, 1>& eigenValues,
299 FieldMatrix<K, 1, 1>& eigenVectors)
300 {
301 eigenValues[0] = matrix[0][0];
302 if constexpr(Tag==EigenvaluesEigenvectors)
303 eigenVectors[0] = {1.0};
304 }
305
306
307 // 2d specialization
308 template <Jobs Tag, typename K>
309 static void eigenValuesVectorsImpl(const FieldMatrix<K, 2, 2>& matrix,
310 FieldVector<K, 2>& eigenValues,
311 FieldMatrix<K, 2, 2>& eigenVectors)
312 {
313 // Compute eigen values
314 Impl::eigenValues2dImpl(matrix, eigenValues);
315
316 // Compute eigenvectors by exploiting the Cayley–Hamilton theorem.
317 // If λ_1, λ_2 are the eigenvalues, then (A - λ_1I )(A - λ_2I ) = (A - λ_2I )(A - λ_1I ) = 0,
318 // so the columns of (A - λ_2I ) are annihilated by (A - λ_1I ) and vice versa.
319 // Assuming neither matrix is zero, the columns of each must include eigenvectors
320 // for the other eigenvalue. (If either matrix is zero, then A is a multiple of the
321 // identity and any non-zero vector is an eigenvector.)
322 // From: https://en.wikipedia.org/wiki/Eigenvalue_algorithm#2x2_matrices
323 if constexpr(Tag==EigenvaluesEigenvectors) {
324
325 // Special casing for multiples of the identity
326 FieldMatrix<K,2,2> temp = matrix;
327 temp[0][0] -= eigenValues[0];
328 temp[1][1] -= eigenValues[0];
329 if(temp.infinity_norm() <= 1e-14) {
330 eigenVectors[0] = {1.0, 0.0};
331 eigenVectors[1] = {0.0, 1.0};
332 }
333 else {
334 // The columns of A - λ_2I are eigenvectors for λ_1, or zero.
335 // Take the column with the larger norm to avoid zero columns.
336 FieldVector<K,2> ev0 = {matrix[0][0]-eigenValues[1], matrix[1][0]};
337 FieldVector<K,2> ev1 = {matrix[0][1], matrix[1][1]-eigenValues[1]};
338 eigenVectors[0] = (ev0.two_norm2() >= ev1.two_norm2()) ? ev0/ev0.two_norm() : ev1/ev1.two_norm();
339
340 // The columns of A - λ_1I are eigenvectors for λ_2, or zero.
341 // Take the column with the larger norm to avoid zero columns.
342 ev0 = {matrix[0][0]-eigenValues[0], matrix[1][0]};
343 ev1 = {matrix[0][1], matrix[1][1]-eigenValues[0]};
344 eigenVectors[1] = (ev0.two_norm2() >= ev1.two_norm2()) ? ev0/ev0.two_norm() : ev1/ev1.two_norm();
345 }
346 }
347 }
348
349 // 3d specialization
350 template <Jobs Tag, typename K>
351 static void eigenValuesVectorsImpl(const FieldMatrix<K, 3, 3>& matrix,
352 FieldVector<K, 3>& eigenValues,
353 FieldMatrix<K, 3, 3>& eigenVectors)
354 {
355 using Vector = FieldVector<K,3>;
356 using Matrix = FieldMatrix<K,3,3>;
357
358 //compute eigenvalues
359 /* Precondition the matrix by factoring out the maximum absolute
360 value of the components. This guards against floating-point
361 overflow when computing the eigenvalues.*/
362 using std::isnormal;
363 K maxAbsElement = (isnormal(matrix.infinity_norm())) ? matrix.infinity_norm() : K(1.0);
364 Matrix scaledMatrix = matrix / maxAbsElement;
365 K r = Impl::eigenValues3dImpl(scaledMatrix, eigenValues);
366
367 if constexpr(Tag==EigenvaluesEigenvectors) {
368 K offDiagNorm = Vector{scaledMatrix[0][1],scaledMatrix[0][2],scaledMatrix[1][2]}.two_norm2();
369 if (offDiagNorm <= std::numeric_limits<K>::epsilon())
370 {
371 eigenValues = {scaledMatrix[0][0], scaledMatrix[1][1], scaledMatrix[2][2]};
372 eigenVectors = {{1.0, 0.0, 0.0}, {0.0, 1.0, 0.0}, {0.0, 0.0, 1.0}};
373
374 // Use bubble sort to jointly sort eigenvalues and eigenvectors
375 // such that eigenvalues are ascending
376 if (eigenValues[0] > eigenValues[1])
377 {
378 std::swap(eigenValues[0], eigenValues[1]);
379 std::swap(eigenVectors[0], eigenVectors[1]);
380 }
381 if (eigenValues[1] > eigenValues[2])
382 {
383 std::swap(eigenValues[1], eigenValues[2]);
384 std::swap(eigenVectors[1], eigenVectors[2]);
385 }
386 if (eigenValues[0] > eigenValues[1])
387 {
388 std::swap(eigenValues[0], eigenValues[1]);
389 std::swap(eigenVectors[0], eigenVectors[1]);
390 }
391 }
392 else {
393 /*Compute the eigenvectors so that the set
394 [evec[0], evec[1], evec[2]] is right handed and
395 orthonormal. */
396
397 Matrix evec(0.0);
398 Vector eval(eigenValues);
399 if(r >= 0) {
400 Impl::eig0(scaledMatrix, eval[2], evec[2]);
401 Impl::eig1(scaledMatrix, evec[2], evec[1], eval[1]);
402 evec[0] = Impl::crossProduct(evec[1], evec[2]);
403 }
404 else {
405 Impl::eig0(scaledMatrix, eval[0], evec[0]);
406 Impl::eig1(scaledMatrix, evec[0], evec[1], eval[1]);
407 evec[2] = Impl::crossProduct(evec[0], evec[1]);
408 }
409 //sort eval/evec-pairs in ascending order
410 using EVPair = std::pair<K, Vector>;
411 std::vector<EVPair> pairs;
412 for(std::size_t i=0; i<=2; ++i)
413 pairs.push_back(EVPair(eval[i], evec[i]));
414 auto comp = [](EVPair x, EVPair y){ return x.first < y.first; };
415 std::sort(pairs.begin(), pairs.end(), comp);
416 for(std::size_t i=0; i<=2; ++i){
417 eigenValues[i] = pairs[i].first;
418 eigenVectors[i] = pairs[i].second;
419 }
420 }
421 }
422 //The preconditioning scaled the matrix, which scales the eigenvalues. Revert the scaling.
423 eigenValues *= maxAbsElement;
424 }
425
426 // forwarding to LAPACK with corresponding tag
427 template <Jobs Tag, int dim, typename K>
428 static void eigenValuesVectorsLapackImpl(const FieldMatrix<K, dim, dim>& matrix,
429 FieldVector<K, dim>& eigenValues,
430 FieldMatrix<K, dim, dim>& eigenVectors)
431 {
432 {
433#if HAVE_LAPACK
434 /*Lapack uses a proprietary tag to determine whether both eigenvalues and
435 -vectors ('v') or only eigenvalues ('n') should be calculated */
436 const char jobz = "nv"[Tag];
437
438 const long int N = dim ;
439 const char uplo = 'u'; // use upper triangular matrix
440
441 // length of matrix vector, LWORK >= max(1,3*N-1)
442 const long int lwork = 3*N -1 ;
443
444 constexpr bool isKLapackType = std::is_same_v<K,double> || std::is_same_v<K,float>;
445 using LapackNumType = std::conditional_t<isKLapackType, K, double>;
446
447 // matrix to put into dsyev
448 LapackNumType matrixVector[dim * dim];
449
450 // copy matrix
451 int row = 0;
452 for(int i=0; i<dim; ++i)
453 {
454 for(int j=0; j<dim; ++j, ++row)
455 {
456 matrixVector[ row ] = matrix[ i ][ j ];
457 }
458 }
459
460 // working memory
461 LapackNumType workSpace[lwork];
462
463 // return value information
464 long int info = 0;
465 LapackNumType* ev;
466 if constexpr (isKLapackType){
467 ev = &eigenValues[0];
468 }else{
469 ev = new LapackNumType[dim];
470 }
471
472 // call LAPACK routine (see fmatrixev.cc)
473 eigenValuesLapackCall(&jobz, &uplo, &N, &matrixVector[0], &N,
474 ev, &workSpace[0], &lwork, &info);
475
476 if constexpr (!isKLapackType){
477 for(size_t i=0;i<dim;++i)
478 eigenValues[i] = ev[i];
479 delete[] ev;
480 }
481
482 // restore eigenvectors matrix
483 if (Tag==Jobs::EigenvaluesEigenvectors){
484 row = 0;
485 for(int i=0; i<dim; ++i)
486 {
487 for(int j=0; j<dim; ++j, ++row)
488 {
489 eigenVectors[ i ][ j ] = matrixVector[ row ];
490 }
491 }
492 }
493
494 if( info != 0 )
495 {
496 std::cerr << "For matrix " << matrix << " eigenvalue calculation failed! " << std::endl;
497 DUNE_THROW(InvalidStateException,"eigenValues: Eigenvalue calculation failed!");
498 }
499#else
500 DUNE_THROW(NotImplemented,"LAPACK not found!");
501#endif
502 }
503 }
504
505 // generic specialization
506 template <Jobs Tag, int dim, typename K>
507 static void eigenValuesVectorsImpl(const FieldMatrix<K, dim, dim>& matrix,
508 FieldVector<K, dim>& eigenValues,
509 FieldMatrix<K, dim, dim>& eigenVectors)
510 {
511 eigenValuesVectorsLapackImpl<Tag>(matrix,eigenValues,eigenVectors);
512 }
513 } //namespace Impl
514
522 template <int dim, typename K>
523 static void eigenValues(const FieldMatrix<K, dim, dim>& matrix,
524 FieldVector<K ,dim>& eigenValues)
525 {
527 Impl::eigenValuesVectorsImpl<Impl::Jobs::OnlyEigenvalues>(matrix, eigenValues, dummy);
528 }
529
538 template <int dim, typename K>
540 FieldVector<K ,dim>& eigenValues,
541 FieldMatrix<K, dim, dim>& eigenVectors)
542 {
543 Impl::eigenValuesVectorsImpl<Impl::Jobs::EigenvaluesEigenvectors>(matrix, eigenValues, eigenVectors);
544 }
545
553 template <int dim, typename K>
555 FieldVector<K, dim>& eigenValues)
556 {
558 Impl::eigenValuesVectorsLapackImpl<Impl::Jobs::EigenvaluesEigenvectors>(matrix, eigenValues, dummy);
559 }
560
569 template <int dim, typename K>
571 FieldVector<K, dim>& eigenValues,
572 FieldMatrix<K, dim, dim>& eigenVectors)
573 {
574 Impl::eigenValuesVectorsLapackImpl<Impl::Jobs::EigenvaluesEigenvectors>(matrix, eigenValues, eigenVectors);
575 }
576
584 template <int dim, typename K, class C>
586 FieldVector<C, dim>& eigenValues)
587 {
588#if HAVE_LAPACK
589 {
590 const long int N = dim ;
591 const char jobvl = 'n';
592 const char jobvr = 'n';
593
594 constexpr bool isKLapackType = std::is_same_v<K,double> || std::is_same_v<K,float>;
595 using LapackNumType = std::conditional_t<isKLapackType, K, double>;
596
597 // matrix to put into dgeev
598 LapackNumType matrixVector[dim * dim];
599
600 // copy matrix
601 int row = 0;
602 for(int i=0; i<dim; ++i)
603 {
604 for(int j=0; j<dim; ++j, ++row)
605 {
606 matrixVector[ row ] = matrix[ i ][ j ];
607 }
608 }
609
610 // working memory
611 LapackNumType eigenR[dim];
612 LapackNumType eigenI[dim];
613 LapackNumType work[3*dim];
614
615 // return value information
616 long int info = 0;
617 const long int lwork = 3*dim;
618
619 // call LAPACK routine (see fmatrixev_ext.cc)
620 eigenValuesNonsymLapackCall(&jobvl, &jobvr, &N, &matrixVector[0], &N,
621 &eigenR[0], &eigenI[0], nullptr, &N, nullptr, &N, &work[0],
622 &lwork, &info);
623
624 if( info != 0 )
625 {
626 std::cerr << "For matrix " << matrix << " eigenvalue calculation failed! " << std::endl;
627 DUNE_THROW(InvalidStateException,"eigenValues: Eigenvalue calculation failed!");
628 }
629 for (int i=0; i<N; ++i) {
630 eigenValues[i].real = eigenR[i];
631 eigenValues[i].imag = eigenI[i];
632 }
633 }
634#else
635 DUNE_THROW(NotImplemented,"LAPACK not found!");
636#endif
637 }
638 } // end namespace FMatrixHelp
639
642} // end namespace Dune
643#endif
A dense n x m matrix.
Definition: fmatrix.hh:117
vector space out of a tensor product of fields.
Definition: fvector.hh:95
Default exception if a function was called while the object is not in a valid state for that function...
Definition: exceptions.hh:281
Default exception for dummy implementations.
Definition: exceptions.hh:263
A few common exception classes.
Implements a matrix constructed from a given type representing a field and compile-time given number ...
static void eigenValuesNonSym(const FieldMatrix< K, dim, dim > &matrix, FieldVector< C, dim > &eigenValues)
calculates the eigenvalues of a non-symmetric field matrix
Definition: fmatrixev.hh:585
static void eigenValues(const FieldMatrix< K, dim, dim > &matrix, FieldVector< K, dim > &eigenValues)
calculates the eigenvalues of a symmetric field matrix
Definition: fmatrixev.hh:523
static void eigenValuesLapack(const FieldMatrix< K, dim, dim > &matrix, FieldVector< K, dim > &eigenValues)
calculates the eigenvalues of a symmetric field matrix
Definition: fmatrixev.hh:554
static void eigenValuesVectors(const FieldMatrix< K, dim, dim > &matrix, FieldVector< K, dim > &eigenValues, FieldMatrix< K, dim, dim > &eigenVectors)
calculates the eigenvalues and eigenvectors of a symmetric field matrix
Definition: fmatrixev.hh:539
static void eigenValuesVectorsLapack(const FieldMatrix< K, dim, dim > &matrix, FieldVector< K, dim > &eigenValues, FieldMatrix< K, dim, dim > &eigenVectors)
calculates the eigenvalues and -vectors of a symmetric field matrix
Definition: fmatrixev.hh:570
Implements a vector constructed from a given type representing a field and a compile-time given size.
#define DUNE_THROW(E, m)
Definition: exceptions.hh:218
auto max(ADLTag< 0 >, const V &v1, const V &v2)
implements binary Simd::max()
Definition: defaults.hh:81
Some useful basic math stuff.
Dune namespace.
Definition: alignedallocator.hh:13
static const Field pi()
Archimedes' constant.
Definition: math.hh:48
Creative Commons License   |  Legal Statements / Impressum  |  Hosted by TU Dresden  |  generated with Hugo v0.111.3 (Jul 15, 22:36, 2024)