High performance computing (HPC) today offers three levels of parallelism: inter-node communication, intra-node multi-threading, and the usage of hardware accelerators. At the same time, parallel programming has left the HPC niche and is now also prevalent in standard workstations or laptops and has, thus, quietly entered every scientist’s domain. The workstation attached to a research instrument (such as a microscope, digital camera, or musical instrument) is the same as a compute node in an HPC system (minus the expensive inter-node network connectivity). It is the goal of this class to educate researchers in the efficient usage of compute resources which will serve multiple overarching goals:
- Establishing a common vocabulary with the HPC experts
- Enabling the researcher to “understand” and edit existing applications
- Better utilization of the already available workplace resources
- Easier transition to an HPC resource such as BigRed II
In order to achieve the goal the class introduces all levels of parallelism as well as common approaches for parallelization over the course of five days. The class contains lectures and labs so that the participants immediately put the newly acquired knowledge to work.
Go to RT events for a complete list of RT related events.