The bug relates to MD simulations on GPU. In the documentation of oxDNA (as of 09.09.2024), the setting "cells_auto_optimisation = false" is proposed to improve performance.

I am working on a cluster and I used interactive SLURM jobs to run simulations within the notebook. In this context, any MD simulation that I tried to run on GPU with "cells_auto_optimisation = false" killed the notebook kernel directly. It then looks like this:

In this repo, you can find a minimal working example from what I used on the cluster to crash the kernel. I also tried to use SLURM jobs to run the simulations and there, a error message is shown - however I do not know to counteract it: It reads "CUDA error at /path/to/oxDNA/src/CUDA/Lists/CUDASimpleVerletList.cu 147: out of memory". However, in the documentation it is not specified that other flags need to be set if "cells_auto_optimisation = false".