Add support for Verlet scheme with Buckingham
We are trying to simulate a simple alumina crystal. The forcefield only contains non-bonded interactions defined with our own Buckingham potentials. Everything works fine with the group cutoff scheme. But with the similar settings in Verlet scheme, the simulation would blow up right after step 0 with a segmentation fault error. GDB shows the error happens in pme.c: 415 idxptr[ZZ] = pme->nnz[tiz]. Taking the group scheme .tpr and run mdrun -testverlet fails in the same way.
The step 0 information shows absurdly high buckingham energy and kinetic energy. We tried modifying parameters related to the Verlet scheme, such as verlet-buffer-drift, vdw-modifier and rvdw-switch, but had no luck. We also compared the outputs from group and Verlet schemes but couldn't see anything weird.
I attached the .mdp; changing the cutoff-scheme to group and setting rlist=1.2 will make the simulation run. The .tpr and topology are also attached.
I would greatly appreciate any suggestions. Thanks in advance!
#8 Updated by Rossen Apostolov almost 6 years ago
- Tracker changed from Bug to Feature
- Subject changed from Segmentation fault switching from group to verlet cutoff shceme to Add support for Verlet scheme with Buckingham
- Status changed from Fix uploaded to Accepted
- Priority changed from High to Normal
Changed the report to a "feature"
#9 Updated by Mark Abraham almost 6 years ago
- Target version changed from 4.6.x to 5.x
There are plans to support tabulated potentials with the Verlet scheme and GPUs, perhaps in 5.1. Native Buckingham kernels would only be considered after we produce a Verlet kernel generator sometime this year; perhaps it might be straightforward at that time.