[Molpro-user] Inconsistent memory 2006.1

dedey at alumni.bilkent.edu.tr dedey at alumni.bilkent.edu.tr
Thu Sep 18 13:29:16 BST 2008


Hi all,

I have problems in running Molpro in parallel. The serial job invoked by
the command below works fine. When I request a parallel job with the
second command below I see the 8 compute processes start on the compute
node but after it enters the MULTI code it dies with the error:

 USED MEMORY IN cislow:        22302786  22305334  22305334  22305334
22305334  22305334  22305334  22305334
 FREE MEMORY IN cislow:        77697114  77694566  77694566  77694566
77694566  77694566  77694566  77694566
 ? Error
 ?  Inconsistent memory
 ? The problem occurs in check_address

 GA ERROR fehler on processor   0
 CLOSEW FILE 31  NAME=eaf_T3100013063.TMP  IMPLEMENTATION=eaf   HANDLE=    2

>From the error printed above it is seen that one of the eight compute
processes tries to allocate a slightly smaller memory. The
"check_address" mentioned in the error report can be found in the utils.f
source file. I played with the -m and -G flags to increase and decrease
the memory, but no success out of tens of trials. Specification of the
memory from the input file also did not work. Jobs with 2 or 4 or more
than 8 CPUs also die giving the same error.

Am I missing something obvious? Any ideas?

Thanks in advance...

Yavuz


The successfull serial run is with this command in the PBS script:

molpro -I $PWD -W /N/dc/scratch/$USER -d /N/dc/scratch/$USER -o
YD_WO_M1.mout YD_WO_M1.mlp

The parallel jobs dying are initiated by this command:

molpro -I $PWD -W /N/dc/scratch/$USER -d /N/dc/scratch/$USER -o
YD_WO_M2.mout -n 8 YD_WO_M2.mlp

Other errors of the same type with different number of cores are:

 ITER. MIC  NCI  NEG     ENERGY(VAR)     ENERGY(PROJ)   ENERGY CHANGE    
GRAD(0)  GRAD(ORB)   GRAD(CI)     STEP       TIME

 USED MEMORY IN cislow:        11296035  11411293  11411293  11411293 
11411293  11411293  11411293  11411293  11411293  11411293
                               11411293  11411293  11411293  11411293 
11411293  11411293
 FREE MEMORY IN cislow:        88703865  88588607  88588607  88588607 
88588607  88588607  88588607  88588607  88588607  88588607
                               88588607  88588607  88588607  88588607 
88588607  88588607
 ? Error
 ?  Inconsistent memory
 ? The problem occurs in check_address


 ITER. MIC  NCI  NEG     ENERGY(VAR)     ENERGY(PROJ)   ENERGY CHANGE    
GRAD(0)  GRAD(ORB)   GRAD(CI)     STEP       TIME

 USED MEMORY IN cislow:        11296035  11411293  11411293  11411293 
11411293  11411293  11411293  11411293  11411293  11411293
                               11411293  11411293  11411293  11411293 
11411293  11411293
 FREE MEMORY IN cislow:        38703865  38588607  38588607  38588607 
38588607  38588607  38588607  38588607  38588607  38588607
                               38588607  38588607  38588607  38588607 
38588607  38588607
 ? Error
 ?  Inconsistent memory
 ? The problem occurs in check_address

 ITER. MIC  NCI  NEG     ENERGY(VAR)     ENERGY(PROJ)   ENERGY CHANGE    
GRAD(0)  GRAD(ORB)   GRAD(CI)     STEP       TIME

 USED MEMORY IN cislow:        11296035  11411293  11411293  11411293 
11411293  11411293  11411293  11411293  11411293  11411293
                               11411293  11411293  11411293  11411293 
11411293  11411293
 FREE MEMORY IN cislow:        25404025  25288767  25288767  25288767 
25288767  25288767  25288767  25288767  25288767  25288767
                               25288767  25288767  25288767  25288767 
25288767  25288767
 ? Error
 ?  Inconsistent memory
 ? The problem occurs in check_address

 GA ERROR fehler on processor   0



||||||||||||||||||||||||||||||||||||||||||||||

Yavuz Dede, Ph.D.
Theo./Comp. Chem.

IU Bloomington - USA

METU Ankara - TURKEY

||||||||||||||||||||||||||||||||||||||||||||||






More information about the Molpro-user mailing list