Ciao Giacomo,
Sorry, I've never seen that.
Obviously, something is going wrong with memory management, and
it is probably independent of the 2x1x1 k-point grid.
Some few questions I can think of:
Does the molecule in its own optimized structure run ok ?.
Do you start from a fresh density matrix ?.
Have you played with the Diag.Memory parameter (check manual) ?.
On a another hand, from the source I can see that array ftmp2 is
being used in a MPI context. Thus I doubt you're really running
a serial code. Namely, a single processor, MPI-unabled code
may not be the same as a true serial one.
HTH
Roberto
On Sun, 8 Jul 2012, Giacomo Giorgi wrote:
Hi Roberto,
differently from what I was thinking the serial version gives me
this error message
siesta: 3 -2779.6431 -2749.0589 -2749.2928 1.0045 -0.6149
alloc_err: allocate status error 41
alloc_err: array ftmp2 requested by redistribXY
alloc_err: dim, lbound, ubound: 1 1 2
alloc_err: dim, lbound, ubound: 2 1 144
alloc_err: dim, lbound, ubound: 3 1 250
alloc_err: dim, lbound, ubound: 4 1 486
alloc_err: allocate error
This both in the case of the ghost atoms and the single point calculation of
the molecule in the optimized geometry of the molecule+surface system.
Could you help me in figuring out the reason for this?
Once more I repeat: molecule, 2x1x1 k-point sampling, serial run.
Thanks a lot!
With best regards,
Giacomo
