Using more than one node with GPUs
The following steps illustrate the process of building NAMD 3.0 from the source code. This version is not yet built as a module on Kebnekaise. We will compile NAMD so that has supports for multi-node multi-GPU simulations. The stmv benchmark with 1M atoms performed ~1.3x faster when running it on 2 nodes w.r.t. 1 node case.