Heterogeneous Programming with Single Operation Multiple Data
01 Oct 2014
Heterogeneity is omnipresent in today’s commodity computational systems, which comprise at least one multi-core CPU and one Graphics Processing Unit. Nonetheless, all this computing power is not being exploited by main- stream computing, as the programming of these systems entails many details of the underlying architecture and its distinct execution models. Current research on high level parallel programming is addressing these issues, but, still, the systems’ heterogeneity is exposed at language level.
This paper proposes a uniform framework, grounded on Single Operation Multiple Data model, for the programming of such heterogeneous systems. The model is declarative, empowering the compiler to generate code for mul- tiple architectures from the same source. To this extent, we designed a sim- ple extension of the Java programming language that embodies the model, and developed a compiler that generates code for both multi-core CPUs and GPUs. A performance evaluation attests the validity of the approach that, despite based on a simple programming model, is able to deliver performance gains on par with hand-tuned data parallel multi-threaded Java applications.
