Parallel Programming Model for the Epiphany Many-Core Coprocessor Using Threaded MPI

摘要

The Adapteva Epiphany many-core architecture comprises a 2D tiled meshNetwork-on-Chip (NoC) of low-power RISC cores with minimal uncorefunctionality. It offers high computational energy efficiency for both integerand floating point calculations as well as parallel scalability. Yet despitethe interesting architectural features, a compelling programming model has notbeen presented to date. This paper demonstrates an efficient parallelprogramming model for the Epiphany architecture based on the Message PassingInterface (MPI) standard. Using MPI exploits the similarities between theEpiphany architecture and a conventional parallel distributed cluster of serialcores. Our approach enables MPI codes to execute on the RISC array processorwith little modification and achieve high performance. We report benchmarkresults for the threaded MPI implementation of four algorithms (densematrix-matrix multiplication, N-body particle interaction, a five-point 2Dstencil update, and 2D FFT) and highlight the importance of fast inter-corecommunication for the architecture.