QM/MM live solvent selection - segmentation fault (with ORCA)

From: Oleksii Zdorevskyi (zdorevskyi_at_bitp.kiev.ua)
Date: Tue Sep 28 2021 - 10:39:18 CDT

Dear NAMD developers,

I am facing an issue regarding the QM/MM setup of NAMD together with ORCA
using Live solvent selection. Whenever I include a substantial number of
water molecules to the QM region, my system firstly runs ok, but when it
reaches the step for LSS exchange (listed in LSSfreq), the run immediately
crashes with a Segmentation fault. I have tested this issue on different
workstations and clusters with different amounts of memory allocation. On
all the workstations, OpenMPI is configured correctly, so that NAMD+ORCA
runs smoothly without LSS.

To illustrate this problem with an example, please take a look at a very
simple system of an H2O2 molecule in a water solution (71 water molecules
in the QM region):

https://urldefense.com/v3/__https://drive.google.com/drive/folders/1O1vSpSUC5VKC3y45BfeLJMxw4DWDy4vV?usp=sharing__;!!DZ3fjg!t-Iw6EdhA4PMwE_R3OWUQ2q9R32FPD7w52RwWQhjHc6Wk2VRDT6Koc3HdWe1jS5irQ$

Neither NAMD nor ORCA does not show any error, just a segmentation fault
in the Slurm output file (qm_40c/slurm-719879.out):

/var/spool/slurmd/job719879/slurm_script: line 19: 196326 Segmentation
fault (core dumped) $NAMDDIR/namd2 +CmiSleepOnIdle qm1.conf > qm1.log

Does NAMD implement some sophisticated calculations during the LSS
exchange that clog up the memory immediately?
Are there any limitations to the number of water molecules that can be
chosen for LSS?

Thanks in advance.

Best regards,
Oleksii

This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:11 CST