MICHELLE for high-level optimization, large scale problems and HPC environments

New capability has been implemented in the MICHELLE code and its environment over the past two years in support of rapid device design and optimization. Central to this is the adoption of High Performance Computing (HPC) tools and the porting of the code to HPC clusters. The solver now supports distributed- memory, MPI-based domain decomposition. The development also supports per-node accelerators such as GPUs and multicore CPUs. The code can handle large- scale models supporting the ability to simulate multiple beam devices, such as MBK guns and MB-IOTs, and at adequate resolution to properly resolve the dynamics. Significant to this development has been the integration of MICHELLE with the CAPSTONE mesher, the ParaView data visualizer, and the GSB/DAKOTA execution environment and optimization library, which enables the rapid device design and optimization process. With the current development program in its final stage, this presentation provides an overview of the new capability that has been achieved.