A Uniform Platform to Support Multigenerational GPUs for High Performance Stream-based Computing

Abstract

GPU-based computing has become one of the popular high performance computing fields. The field is called GPGPU. This paper is focused on design and implementation of a uniform GPGPU application that is optimized for both the legacy and the recent GPU architectures. As a typical example of such the GPGPU application, this paper will discuss the uniform implementation of the Caravela platform. Especially the flow-model execution mechanism will be considered referring the recent GPU architectures. To verify the design and the implementation on CUDA and OpenCL platform, this paper will evaluate the compatibility among the architectures, and also test measurements of performance.