Analysis and Characterization of GPU Benchmarks for Kernel Concurrency Efficiency

Graphical Processing Units (GPUs) became an important platform to general purpose computing, thanks to their high performance and low cost when compared to CPUs. However, programming GPUs requires a different mindset and optimization techniques that take advantage of the peculiarities of the GPU architecture. Moreover, GPUs are rapidly changing, in the sense of including capabilities that can improve performance of general purpose applications, such as support for concurrent execution. Thus, benchmark suites developed to evaluate GPU performance and scalability should take those aspects into account and could be quite different from traditional CPU benchmarks. Nowadays, Rodinia, Parboil and SHOC are the main benchmark suites for evaluating GPUs. This work analyzes these benchmark suites in detail and categorizes their behavior in terms of computation type (integer or float), usage of memory hierarchy, efficiency and hardware occupancy. We also intend to evaluate similarities between the kernels of those suites. This characterization will be useful to disclosure the resource requirements of the kernels of these benchmarks that may affect further concurrent execution.