Abstract. We present a domain specific language, embedded in Haskell, for general purpose parallel programming on GPUs. Our intention is to explore the use of connection patterns in parallel programming. We briefly present our earlier work on hardware generation, and outline the current state of GPU architectures and programming models. Finally, we present the current status of the Obsidian project, which aims to make GPU programming easier, without relinquishing detailed control of GPU resources. Both a programming example and some details of the implementation are presented. This is a report on work in progress. 1

We present a novel high-level parallel programming model for graphics processing units (GPUs). We embed GPU kernels as data-parallel array computations in the purely functional language Haskell. GPU and CPU computations can be freely interleaved with the type system tracking the two different modes of computation. The embedded language of array computations is sufficiently limited that our system can automatically isolate and extract these computations and compile them to efficient GPU code. In this paper, we outline our approach and present the results of a few preliminary benchmarks. 1.