Replication of uniformly accessed shared data for large-scale data-parallel algorithms
Chung‐Ming Chen, Soo-Young Lee · 2002
In this paper we show how to minimize data sharing overhead required in most parallel algorithms, especially in Large-Scale Data-Parallel (LSDP) algorithms on a 2D mesh. Two specific issues are addressed in this study. One is what the optimal group size is, i.e., how many PEs should share a copy of shared data. The other is where the replicated data should be allocated.>