Implements the Scherzinger-Dohrmann fast and accurate eigendecomposition of 3x3 symmetric matrices. I am intentionally not implementing derivatives, and avoiding the non-differentiability of the eigenvectors at eigenvalue degeneracies. The envisioned use case is for computing isotropic tensor functions, which themselves are differentiable even when repeated eigenvalues occur (where the underlying scalar function is differentiable, of course). I will add some of the commonly used ones in a successive PR.
Implements the Scherzinger-Dohrmann fast and accurate eigendecomposition of 3x3 symmetric matrices. I am intentionally not implementing derivatives, and avoiding the non-differentiability of the eigenvectors at eigenvalue degeneracies. The envisioned use case is for computing isotropic tensor functions, which themselves are differentiable even when repeated eigenvalues occur (where the underlying scalar function is differentiable, of course). I will add some of the commonly used ones in a successive PR.