[Pw_forum] dE0s is positive which should never happen

[Guntram Schmidt]

>>      0:"recvec.f90", line 106: 1525-108 Error encountered while
>> attempting to allocate a data object.  The program will stop.
>
> not enough memory


There seems to be more than this.
When I start the same job (different prefix and outdir) on the same 
machine (without any other jobs running), but two times using half the 
processors, one of the latter jobs crashes and the other survives.

E.g.:
One job, consuming 7.000MB with 8 processors (56.000MB in total - the 
amount of the machine, where it is running) --> runs fine

Two jobs, consuming 7.000MB with 4 processors, each (= 56.000 in total, 
too) --> one job "survives", the other crashes:

   0:"xc_vdW_DF.f90", line 348: 1525-108 Error encountered while 
attempting to allocate a data object.  The program will stop.
    1:"xc_vdW_DF.f90", line 348: 1525-108 Error encountered while 
attempting to allocate a data object.  The program will stop.

Any idea?

The machine is a IBM 575 
(http://www-01.ibm.com/common/ssi/cgi-bin/ssialias?infotype=AN&subtype=CA&htmlfid=897/ENUS107-675&appname=USN) 
running Powerlinux/SLES 11 with loadleveler-job-balancing.

Thanks,
Guntram