Multi-core processors with up to 8 cores and more as well as GPUs with thousands of cores are omnipresent. In order to fully exploit their parallel computing capability, programmers have to deal with low-level concepts of parallel programming. These low-level concepts constitute a high barrier to efficient development of parallel applications. A higher level of abstraction would be desirable.

In order to assist programmers in developing parallel applications Algorithmic Skeletons have been proposed. They encapsulate well-defined, frequently recurring parallel programming patterns, thereby shielding programmers from low-level aspects of parallel programming.

The main contribution of this paper is the design and implementation of data parallel skeletons with GPU support in Java. Additionally, on the basis of three benchmark applications, including Matrix multiplication, N-Body calculations, and Shortest paths, we evaluate the performance of the presented implementation on a GPU cluster.