Applications of the Michelle Code in HPC Environments

The MICHELLE charged particle beam optics code [1], [2] has been extended to perform on HPC class clusters in support of rapid device design and optimization. Central to the work is that the code now supports new distributed-memory domain decomposition and per-node GPU accelerators as well as multicore processing. For use on the HPC, MICHELLE has been integrated with the CAPSTONE mesher [3], the Para View data visualizer, and the GSB/DAKOTA execution environment and optimization library. This talk presents the current status of this program.