Memory-aware and user-aware mapping of applications to MPSoCs

Application mapping is one of the key issues in MPSoC design. For MPSoC systems running multiple applications, a memory-aware and user-aware application mapping strategy is proposed. This strategy classifies the applications into hot applications which access large amounts of data and non-hot applications with moderate memory access by offline analysis. Then, the applications are further identified as critical or non-critical application with respect to the user behavior at runtime. For each application, different mapping algorithm is applied according to the above classification. Hot applications are distributed as close as possible to the shared memory. The internal communication cost and contents are minimized for critical applications, and the external communication cost and contents are minimized for non-critical applications. Experimental results show that the proposed strategy saves the overall system communication energy, compared with the single memory-aware or user-aware strategy.