PANDORE: a system to manage data distribution
暂无分享,去创建一个
We investigate here the implementation of “high level” languages on Distributed Memory Parallel Computers (DMPCs). The Pandore system provides a language, a compiler and a run-time support which allows to use DMPCs with a sequential language including data distribution statements. In Pandore, code parallelization is achieved according to data distribution. No explicit process definition nor interprocess communications are needed at the user level. One of the main features of our approach is that data distribution do not modify the result of the program in any case; on the contrary, efficiency is closely related to data distribution.
[1] Michael Gerndt,et al. SUPERB: A tool for semi-automatic MIMD/SIMD parallelization , 1988, Parallel Comput..