Problems with complex routine

[Luke]

I have been using the routine DGGEV to solve a generalised eigenvalue problem for the values E satisfying A.x = E B.x
The routine DGGEV works fine and gives me the right results for the system, however when I try to make the problem complex and use ZGGEV I get results that seem to bear no relation to those generated by DGGEV. The input has only changed in that some arguments have been assigned as complex16, where previously they were double precision, (and that ZGGEV drops an argument) however the values of these arguments contain no complex part at present and I would have thought that the outcome should consequently be the same.

In any case, any help would be much appreciated!

[delmas]

Hello,

Actually, the result must be the same for the real and the complex problems, there must be a mistake in your code.

Just make sure that you are solving the same problem in both cases. To check for the correctness of the output, you can compute the residual ||A*x-E*B*x||. If it is not close to the machine precision, there must be an error somewhere in your code.

Hope that helps, let us know if you have any other problems.

--
Remi