Open jdakka opened 6 years ago
@jdakka : Do we have to user docker encapsulation for EnsembleTk? If so, it's possible, but a bit complicated to set up because I believe we need to match the NVIDIA drivers and CUDA version inside the docker image to the same one running on Titan.
If we don't need to use docker, we can simply build a special conda package for use on Titan that builds OpenMM with the appropriate CUDA version. What version of CUDA (and what driver release) does Titan have?
@jchodera EnTK does not need a docker encapsulation, I think the second approach is fine.
Unfortunately Titan has CUDA 7.5 I /opt/nvidia/cudatoolkit7.5/7.5.18-1.0502.10743.2.1/bin/nvcc
This is how I generated the conda environment with OpenMM running CUDA on Titan. That environment still exists.
Great! We should be able to set up an environment to create builds of OpenMM using our cuda75 build image. Hopefully @lnaden can help us set up a new repo to build this and push these conda packages to the omnia
channel under a separate cuda75
label.
@jchodera @Lnaden let me know if you need assistance in getting this going (just getting things in order before our meeting tomorrow).
@jchodera could assign someone to help or point me to a reference in order to create a docker image of OpenMM to be used on Titan. I checked here though not sure if this works.