Exploiting the locality of data structures in multithreaded architecture

Multithreaded architectures taking a hybrid approach of von Neumann computers and dataflow computers are recently in active research. Multithreaded architectures can improve the performance by the locality exploitation within threads and asynchronous parallel execution among threads. However, it has been overlooked how to exploit effectively the locality of large shared data structures among threads in multithreaded architectures. In this paper, we propose a new data structure, called IB-structure, which is an extension of I-structure to facilitate the distribution of data structure and exploit the locality of data structure, and implement it on multithreaded architecture DAVRID (DAtaflow Von Neumann RISC hybrID). The simulation results show that IB-structure is superior to I-structure over several benchmarks.