[Octopus-devel] [Octopus-notify] svn commit: r3454 - trunk/testsuite/finite_systems_3d by xavier

[Xavier Andrade]

On Thu, 1 Nov 2007, Florian Lorenzen wrote:

> I cannot confirm this. Also, it does not make much sense: if we set
> ghost_update=.true. we have zero_bc and then ghost_update,
> i. e. Xvec_ghost_update works. If zero_bc is faulty, it will
> compromise some values when called again, not?

Yes, but perhaps ghost_update is faulty and requires zero_bc. It could also
be a compiler bug (very unlikely to affect 2 or 3 compilers at the same
time) or an aliasing problem.

>
> Also, it must be related to the kinetic operator acting on a complex
> wavefunction because for the ground state it works. I also started
> looking for the error but have not found something up to know.
>

It depends, if it is a small error it might be that the iterative solvers
absorb it, while the propagation doesn't. Using Born-Oppenheimer propagation
works correctly (and it uses complex wfs).

Cheers,

Xavier