Hollo,

I wrote the C++ code with eigen to solve a Av = lBv problem,
here

A= -0.0291673 -0.448951 -0.48359 -0.445526
-0.448951 -0.289696 -0.445526 -0.319255
-0.48359 -0.445526 -0.0291673 -0.448951
-0.445526 -0.319255 -0.448951 -0.289696

B= 0.999977 0.658283 0.491743 0.523003
0.658283 1 0.523003 0.868399
0.491743 0.523003 0.999977 0.658283
0.523003 0.868399 0.658283 1

Then, I used GeneralizedSelfAdjointEigenSolver to get the eigenvalue "l" and eigenvector "v".
Obtained eigenvalues "l" is almost the same with the eigenvalues obtained by using lapack dsygv.
However, obtained eigenvector is different from these of lapack dsygv.

The results from GeneralizedSelfAdjointEigenSolver:

l = -0.509984
0.203801
0.687355
1.35542

v = -0.321235 0.156434 0.753708 -1.15322
-0.272789 1.7707 -0.679355 1.44747
-0.321235 -0.156434 0.753708 1.15322
-0.272789 -1.7707 -0.679355 -1.44747

The result from lapack dsygv:

l = -0.5099767970
0.2038101470
0.6873309840
1.3553507406

v= 0.3212362790 0.3212362790 0.3212362790 0.3212362790
0.1564368400 0.1564368400 0.1564368400 0.1564368400
-0.7536975044 -0.7536975044 -0.7536975044 -0.7536975044
1.1532122077 1.1532122077 1.1532122077 1.1532122077

Do anyone knows why I got the eigenvector different from the results of lapack
and how method to get the same eigenvector of lapack dsygv?

The ones returned by lapack do not looks right as they do not satisfy v' * B * v = 1, nor Av = lBv, and they are not orthogonal.