SBEL uses Euler, a heterogenous CPU/GPU cluster built by the Wisconsin Applied Computing Center. Multi-core processing is vital to parallel computing, which allows researchers to break complex problems into many simple components. SBEL uses this approach to model and simulate complex mechanical systems.
Euler is accessible via SSH at euler.wacc.wisc.edu. Mac/Linux users should use rsync and Windows users should use WinSCP for file transfers. Below are some helpful links for tools available on Euler:
General Euler Courtesy
- Interactive jobs should be kept to a 20 minute maximum.
- Don’t build large projects on the head node.
- The maximum size for a home directory is no greater than
du -sh ~shows the size of your home directory.
- Please set a time limit on jobs – it allows the scheduler to run more efficiently.
- Please log out when you are not using Euler.
Creating and Accessing Temporary Files
- As of Wednesday, September 6th, files and folders created under /tmp will be deleted without further warning.
- Users may obtain a safe folder to create temporary files using the command
module load tmpdir.
- The path to the new folder will be stored in the environment variable
- Temporary files and folders contained within this path will not be deleted until the corresponding job completes.
Priority Hours and Accounts
- Hardware contributors can debit their share of hours to have access to a higher-priority queue such as slurm_priority or slurm_groupname.
- In order to use the priority queue, you need to pass the following flags along with your script
- -p slurm_priority or -p slurm_groupname
- --qos=priority or --qos=groupname_owner
- Replace groupname with the name of your lab or PI (e.g. sbel or cui)
- To ensure that hours are not debited accidentally, the default behavior is not to use any elevated priority levels when submitting a job.
Chrono is an open-source, C/C++ physics-based dynamics engine that models complex mechanical systems, like cars, trucks, tracked vehicles, or the flow of granular material. Chrono relies on parallel computing, specifically multi-core and GPU parallelism, to simulate rigid bodies interacting through friction and contact. Learn more about Project Chrono here.
Building Chrono on Euler
We recommend using GCC5.3 to build Chrono. To build Chrono::FSI, you will need to load the cuda and gcc/5.3.x modules and cmake like so:
ccmake <chrono_source_path> -DCUDA_HOST_COMPILER=$CCBIN -DCMAKE_C_COMPILER=$CCBIN -DCMAKE_CXX_COMPILER=$CXXBIN