Bohrium: Unmodified NumPy Code on CPU, GPU, and Cluster

In this paper we introduce Bohrium, a runtimesystem for mapping array-operations onto a number of different hardware platforms, from multi-core systems to clusters and GPU enabled systems. As a result, the Bohrium runtime system enables NumPy code to utilize CPU, GPU, and Clusters. Bohrium integrates seamlessly into NumPy through the implicit data parallelization of array operations, which are called Universal Functions in NumPy. Bohrium requires no annotations or other code modifications besides changing the original NumPy import statement to: “import bohrium as numpy”. We evaluate the presented design through a setup that targets a multi-core CPU, an eight-node Cluster, and a GPU, all implemented as preliminary prototypes. The evaluation includes three well-known benchmark applications, Black Sholes, Shallow Water, and N-body, implemented in Python/NumPy.