Improving Accessibility of Heterogeneous Compute Resources for Application Developers Using Programming Model Abstractions Max Plauth, Felix Eberhardt and Andreas Polze The ever-growing demand for compute resources has reached a wide range of application domains, and with that has created a larger audience for compute-intensive tasks. At the same time, heterogeneous computer architectures have become vital in order to deliver the corresponding growth in compute capacity. However, in order to make use of the diverse capabilities of heterogeneous compute resources, application developers have to deal with the complexities introduced by each class of compute resources. With the goal of making heterogeneous computing accessible for a wider range of application developers, we have investigated programming model abstractions for various types of compute resources. Targetting large-scale NUMA systems, the C++-based PGASUS framework simplifies data placement based on polymorphic memory allocators. Adressing scale-out GPU deployments, CloudCL attempts to provide a unified programming model that hides complexities of both OpenCL and distributed computing frameworks. Finally, MetalFS introduces FPGA-based near-storage computing facilities by exposing near-storage operators on the file system level. In our proposed talk, we want to provide a brief overlook of all three approaches and discuss their capabilities.