University of Houston

The High Performance Computing (HPC) community is heading toward the era of exascale computing. Although the final form of an exascale machine is yet unknown, these platforms are expected to exhibit a hitherto unprecedented level of complexity and size. A major challenge with respect to their design is the need to provide higher levels of computational power at dramatically lower rates of power consumption. The process of porting a scientific application to a supercomputer such as Titan at Oak Ridge National Laboratory begins with the formation of a multidisciplinary team composed of domain experts, computer scientists, hardware vendor representatives and professional programmers. This team will initiate the effort to restructure the code in order to exploit the new system’s features, a very time consuming, labor intensive and error-prone process; the quality of the results will depend critically on the experience of the experts involved. Worse, most of the work will be performed by application scientists at the expense of their scientific endeavors and without any guidance or tool support. In this dissertation, we created a tool called “Klonos” which is able to effectively guide the user in the process of poritng their code to a new system and can greatly reduce the human effort involved. We verified our tool by applying it to a real scientific application.