Task mapping on supercomputers with cellular networks

Several models are developed to map reasonably arbitrary application problems to computing platforms for achieving high performance by minimizing communication and by balancing computation. In our models, we assume that the underlying applications be appropriately decomposed to subtasks with known computational loads and the inter-subtask communicational demands, and assume that the computing system's specifications such as individual processor speed and inter-processor communication cost be given or easily measurable. Therefore, the model abstracting the application as a demand matrix and the computer as a network supply matrix can be designed with the objective function as minimum time to complete the application on the given computer. An application, 2D wave equation, was introduced to test our models on the BG/L supercomputer. The mappings generated by our models, reduced communication by 51% for 3D-mesh and 31% for 3D-torus over the default MPI rank order mapping.