ABINIT Discussion Forums

Hello,

a little bug I haven't fixed: if you set dilatmx too high (1.5 here) then sphereboundary complains:

=============
sphereboundary: BUG -
iloop,igb,mgb= 2 21 20
about to overwrite gbound.
Action : contact ABINIT group.

.Delivered 15 WARNINGs and 1 COMMENTs to log file.

leave_new : decision taken to exit ...
=============

which is not clear at all! The dilatmx should be tested or caught beforehand, or the bug fixed if possible.

Matthieu


Input file on ifort 11 openmpi xeon RHE linux:

ecut 11
ngkpt 1 1 1
acell 4 4 4
dilatmx 1.5
nsppol 2
spinat 0 0 2.0 0 0 2.0
kptopt 1
toldfe 1.e-10
xred
0 0 0
1/3 2/3 1/2
angdeg 90 90 120
znucl 27
natom 2
ntypat 1
typat 1 1
nband 16
nbdbuf 2
tsmear 0.001
occopt 7
nshiftk 1
shiftk 0. 0. 0.
nstep 100
prtwf 0

Hi Matthieu,

I think that there's a problem somewhere in the calculation of one of the different mgfft
used in the code as the G-sphere should always be well within the FFT box.
We has similar problem with PAW and iprcell > 0
In our case, however, the problem is related to the setup of the ngfft array
used for the susceptibility.

Ok, I have noticed that this bug is probably compiler dependent: on gfortran the posted input file goes through fine, with 5.9.4 and 6.0 abinits. Even dilatmx 10 works...

Now to go back to ifort to check

Matthieu

Ok, it appears I can no longer reproduce this bug. Declared solved unless someone else digs it up again or we get clearer information.

Matthieu

Hi matthieu

i'm facing a sphereboundary bug when i calculate the energy barrier of diffusion. i read your posts, are you sure the bug can be fixed by changing compiler?