From: Gerard Rowe (GerardR_at_usca.edu)
Date: Tue Jul 24 2018 - 10:55:53 CDT
I'm continuing to play with the QM/MM feature trying to get it to work correctly, and I found that I get the same memory allocation error when invoking mopac instead of orca. NAMD also exits with the same error message if launched with charmrun (I also built an MPI-enabled version that gave the same error, but I'm not certain I launched it correctly). The error persists with QMSimsPerNode=2 or if both QM regions are combined into one.
I'm working with a single, dual-socket node with 48 GB of memory. Has anyone experienced this issue before? The only reference to this particular error message I can find is from years ago when million atom MD simulations were literally running out of memory.
From: Gerard Rowe
Sent: Wednesday, July 18, 2018 6:18 PM
Subject: QM/MM with Orca: Memory allocation failed
I'm having trouble getting hybrid QM/MM to run properly on my Linux machine. Every time I try to run a QM job using orca, namd dies with the message:
Info: Startup phase 15 took 0.000211954 s, 194.461 MB of memory in use
Info: Finished startup at 8.24108 s, 194.461 MB of memory in use
TCL: Minimizing for 100 steps
FATAL ERROR: Memory allocation failed on processor 0.
The relevant part of the configuration file:
qmConfigLine "! ZINDO/1 ENGRAD"
qmConfigLine "%output PrintLevel Mini Print\[P_Mulliken \] 1 Print\[P_AtCharges_M\] end"
qmMult 1 6
qmCharge 1 2.00
qmMult 2 6
qmCharge 2 2.00
I have defined two QM regions with 83 atoms each. I have no problem running namd2 or orca by themselves. I find it hard to believe that this semiempirical job is running out of memory on a system with 48 GB of RAM (569 MB in use at the start of the the job). This error occurs regardless of the number of processors
Is there some way to find out how much memory the program is attempting to allocate? Or is there a common reason this error may arise?
University of South Carolina Aiken
This archive was generated by hypermail 2.1.6 : Wed Sep 18 2019 - 23:20:10 CDT