high performance computing on graphics processing units: hgpu.org

The current trend towards heterogeneous architectures requires a global rethinking of software and hardware design. The focus is centered around new parallel programming models, design space exploration and run-time resource management techniques to exploit the features of many-core processor architectures. Graphics Processing Units (GPU) have become the platform of choice in this area for accelerating a large range of data and task parallel applications. The rapid adoption of GPU computing has been greatly aided by the introduction of high-level programming environments such as NVIDIA’s CUDA C and Khronos’ OpenCL. In this work, we are interested in analyzing the design of a run-time system for heterogeneous architectures. Specifically, it is our goal to propose a robust intermediate representation that compilers can use to represent applications in this field and to evaluate optimizations that can be implemented by such compilers. We have shown in prior work that heterogeneous processors like GPUs have special characteristics that should be taken into consideration by compiler writers in order to achieve better performance and higher throughput. We have also shown that optimizations are highly sensitive to the characteristics of the applications and should be tuned to each architecture. Our current work focuses on extending a compiler framework that we have implemented to translate between different run-time systems and evaluating compiler optimizations for heterogeneous architectures.