An Interactive Environment for Data Partitioning and Distribution
Vasanth Balasundaram, Geoffrey Fox, Ken Kennedy, Ulrich Kremer · 2005
An approach to distributed memory parallel programming that has recently become popular is one where the programmer explicitly specifies the data decomposition using language extensions, and a compiler generates all the communication. While this frees the programmer from the tedium of thinking about message-passing, no assistance is provided in determining the data decomposition scheme that gives the best performance on the target machine. In this paper, we propose an interactive software tool that provides assistance for this very task. The proposed tool also computes performance estimates for any chosen data partitioning scheme, allowing the programmer to experiment with several different strategies without ever running the program on the machine.