high performance computing on graphics processing units: hgpu.org

Dataflow models are valuable tools for representing, analyzing, and synthesizing embedded systems. Heterogeneous computing platforms with multi-core CPU and Graphics Processing Units (GPUs) provide a low cost platform for high performance computations. In this report, we present a dataflow based automated design framework that incorporates analysis, optimization and synthesis tools for embedded systems. Our framework is capable of generating high-performance software from dataflow applications targeted on heterogeneous CPU-GPU platforms. This framework exploits task and data-level parallelism in the dataflow specification and automatically utilizes the heterogeneous platform for performance gain without the need for manual, platform and application specific optimization. We demonstrate the novel and useful capabilities of this framework through experiments on an adapted MP-Sched benchmark that is representative for a wide range of DSP applications.