[QE-users] array index out of bounds error in gww.x

[Michal Krompiec]

Hello,
I'm trying to run GWL on a ~200-atom system (an insulator), starting
from a pw.x calculation with PBESOL and norm-conserving
pseudopotentials from pseudo dojo. Unfortunately, gww.x crashed
immediately after printing this to stdout (432nd band is the highest
occupied):

 Products in imaginary time:
 Loop on KS:         432           1
 Fourier trasform:
 Products in imaginary time:

and the following to stderr:
forrtl: severe (154): array index out of bounds
Image              PC                Routine            Line        Source
gww.x              0000000004E5757B  Unknown               Unknown  Unknown

What could be the cause of this? Is it because my settings don't make
sense, or have I stumbled upon a bug?

Some context:
My system has 864 valence electrons, so in pw.x scf calculation I have
nbnds = (864/2)*1.5 = 648, and I'm using Gamma point only (because
this is a supercell). Then I ran pw4gww.x with the following settings:
&inputpw4gww
    prefix='s'
    num_nbndv(1)=432
    num_nbnds=648
    l_truncated_coulomb=.true.
    truncation_radius=20
    numw_prod=100
    pmat_cutoff=3d0
    s_self_lanczos=1d-8
    l_simple=.true.
/
So far so good. Then I ran gww.x with the following settings (and got
the error mentioned above):
&inputgww
ggwin%prefix='s'
ggwin%n=97,
ggwin%n_fit=120,
ggwin%max_i=32,
ggwin%i_min=1
ggwin%i_max=432
ggwin%l_truncated_coulomb=.true.
ggwin%grid_time=3
ggwin%grid_freq=5
ggwin%second_grid_i=1
ggwin%second_grid_n=10
ggwin%omega=20
ggwin%omega_fit=20
ggwin%n_grid_fit=240
/

Thanks in advance,
Michal Krompiec
Merck KGaA