A multi-threads runtime for the pandore data-parallel compiler

Abstract

Thanks to the multi-threads implementation we are able to efficiently overlap the communication and the computation phases. The unnecessary waiting times that may occur when using a sequential process are avoided. The limited number of threads involved in our implementation prevents the system overhead that in many cases counterbalances the benefit of the parallelization. We think that these improvements will contribute to a better maturity for data parallel compilers and so automatic code generation for distributed memory parallel architectures will become a realistic means of programming these architectures for application users. More details can be found in [1].