LAPACK/ScaLAPACK Development

Hello,

I would like to use lapack to find the eigendecomposition of a relatively small (usually no more than 100x100) non-symmetric matrix whose eigenvalues are guaranteed to be real. Is there any way to speed up DGEEV by using real arithmetic instead of (what I assume it to be using) complex arithmetic?

Thanks!
__________________________________
-Danny greenavis

DGEEV already uses only real arithmetic. Internally it performs a real Schur decomposition with 1x1 or 2x2 blocks on the diagonal.