Even with introduction of new computational models and improvements in processor speeds, the memory system is still a stumbling block to achieve comparable performance of computer systems. Especially the existence of shared global data may cause a con...
Even with introduction of new computational models and improvements in processor speeds, the memory system is still a stumbling block to achieve comparable performance of computer systems. Especially the existence of shared global data may cause a considerable performance degradation in parallel computing. Therefore, it is very important to exploit data locality by distributing data structure and scheduling computations according to the distributed data.
In this paper, we present a data parallelism scheme in nonstrict data structures for a multithreaded architecture DAVRID. We suggest a data parallel programming language, DAPID(DAta Parallel ID), which is an extension of parallel programming language Id. In a bundle. For the efficient implementation of this mechanism, we also suggest a new structure and a new scheme of tag manipulation in nonstrict data structures.