The ABINIT Users Mailing List ( CLOSED )

[PGanesh <pganesh@ciw.edu>]

Hi,

I am doing response-function calculations using abinit (tried with 
v5.4.4p, v5.6.3, 5.6.4, 5.6.5) compiled with intel 10.1 compilers on a 
cluster of xeons.  I always find that the calculations die prematurely 
in the 'ddk' or later during the phonon perturbations with the last few 
lines in the 'log' file being:

rank 17 in job 1  abe0932_57147   caused collective abort of all ranks
 exit status of rank 17: killed by signal 9

(The numbers above change from one job to the other)

The standard error is:

forrtl: severe (41): insufficient virtual memory
Image              PC                Routine            Line        
Source            
abinip             0000000001083936  Unknown               Unknown  Unknown
abinip             0000000001082990  Unknown               Unknown  Unknown
..........                
..........................................................................................

On xeons the code is compiled with Intel 10.1 compilers and there is 1GB 
of memory per processor, way more than the job should require from 
estimates in the 'log' file.

scf, geometry optmizations and nsf calculations run perfectly fine.

I would like to know if anyone has come across such an error and if not 
any suggestion of what may cause such an error would be helpful.  I can 
attach the input file if needed.

Thanks,
Ganesh