TOUGH2-MP : simulations halted after 1 day simulations

Hi every body,

I'm using TOUGH2-MP version 2.01 Under Linux (Debian). I've did some changes in the code and compile it without any problem (named later as testv2.1). Calculations are done for a sample problem of 55000 elements during 13 hours without any problem with 64 CPUs. However, when I'm trying with another sample problem of 90483 elements, simulations halted after 10 min without finishing the job and gave the message error shown below (even with 4 CPUs):

-------------------------------------------------------------------------------------------------------------

$ mpirun -n 4 ./testv2.1
  nex2adj:: rank=           0  MNCON=NCON=      265472  MNEL=NEL=        90483
 Calling METIS (Version5)METIS_Kway              
  rank=           0  done calling Metis istatus=           1  edgecut=        6896
   0sending to pid=   1   22438   90484   22438  151973  151974   22437
   1recving           1   22438   90484   22438  151973  151974   22437
   0sending to pid=   2   22656   90484   22656  153425  153426   22655
   2recving           2   22656   90484   22656  153425  153426   22655
   0sending to pid=   3   22667   90484   22667  157812  157813   22666
   3recving           3   22667   90484   22667  157812  157813   22666
 time for preprocessing ---- including data input, partition, distribution:   36.350855827331543    
*** Error in `./testv2.1': double free or corruption (out): 0x00000000017d7000 ***

Program received signal SIGABRT: Process abort signal.

Backtrace for this error:
#0  0x7F51AF0C4407
#1  0x7F51AF0C4A1E
#2  0x7F51AE3C41BF
#3  0x7F51AE3C4147
#4  0x7F51AE3C5527
#5  0x7F51AE402293
#6  0x7F51AE407A6D
#7  0x7F51AE408775
#8  0x4D5A0A in AZ_fact_ilut
#9  0x4A9743 in AZ_factor_subdomain
#10  0x4E278E in AZ_domain_decomp
#11  0x4D8EFC in AZ_precondition
#12  0x4C0E7D in AZ_pbicgstab
#13  0x4A40EE in AZ_oldsolve
#14  0x4A4BED in AZ_solve
#15  0x445FA5 in lineq_
#16  0x44C84B in cycit_
#17  0x427686 in MAIN__ at tough2.01-mp.f90:?

----------------------------------------

I think that the problem comes from AZTEC solver. May I have to make simulations with the file PARAL.prn by changing default options in Aztec solver ? is-it a problem of memory ? 

Thanks for your help