Skip to Main Content
In recent years, consumers' demand for high-performance IT equipment has grown rapidly. Thus, the Multiple-Processor System on a Chip (MPSoC) and the distributed memory system are widely researched for improvement in the performance of the embedded system. The Message Passing Interface (MPI) specification is the software platform for using the distributed memory system on MPSoC. In addition, MPI_Bcast function is one of the most frequently used functions. Thus, we proposed a novel MPI broadcasting algorithm and hardware architecture for performance improvement. Since the proposed algorithm checks the status of processing nodes and reschedules the order of transmission, processing time can be reduced. In simulation, the proposed algorithm reduced the processing time a maximum of 286605ns (general sequential tree algorithm: 2851160ns, proposed sequential algorithm: 2294355ns) and improved the performance up to 71.3% with an 8-processing-node system.