Hello both of you,
The eigenvalue decomposition of a real symmetric matrix is: "given A, find A = V * D * V^T" and, since A is real symmetric, we additional impose V^TV=I.
If you perform column permutation in V (permutation of the eigenvectors) and symmetric permutation in D (permutation of the eigenvalues) you obtain "another" eigenvalue decomposition.
There is not one better than another.
The concept of "original order" in the sentence:
the original order of the eigenvalues/eigenvectors is not preserved [ in LAPACK ]
makes no sense. There is nothing as an "original order".
Julien.