[Pw_forum] IBM JS20 pw tests problems

[Paolo Giannozzi]

Lazaro Calderin wrote:

> Well .. the most suspicious difference seems to be in the fft grids:
> 
>>From .out
> <      G cutoff =  137.8834  (   3367 G-vectors)     FFT grid: ( 22, 22, 22)
> <      G cutoff =   68.9417  (   1205 G-vectors)  smooth grid: ( 16, 16, 16)
> ---
>>From .ref
>>      G cutoff =  137.8834  (   3367 G-vectors)     FFT grid: ( 24, 24, 24)
>>      G cutoff =   68.9417  (   1205 G-vectors)  smooth grid: ( 15, 15, 15)

try to set the charge grid to 24 (nr1=24,nr2=24,nr3=24 in
namelist system) and see if it makes a difference

Paolo
-- 
Paolo Giannozzi, Democritos and University of Udine, Italy