Any ideas why for the following matrix: http://projects.scipy.org/numpy/raw-att ... 54/mat.txt

DGEEV returns vectors for which the relative error |A v - lambda v|/||A||_F is of the order of 1e-2? Test program here: http://projects.scipy.org/numpy/raw-att ... 4/test.f90

And so, with LAPACK 3.2.1,

- Code: Select all
`$ gfortran -o test test.f90 -llapack -lblas`

$ ./test < mat.txt

4.09736095556244961E-026

4.43089563288366399E-023

3.4612399980026800 LARGE ERROR

8.9078382202313087 LARGE ERROR

8.8674244567312162 LARGE ERROR

8.9066982533988046 LARGE ERROR

3.83826475736065412E-032

...