Optimum Power-Performance GPU Configuration Prediction Based on Code Attributes
Ali Jooya, N.J. Dimopoulos, Amirali Baniasadi · 2017
GPUs have been widely used in the past decade to speed up the execution of general purpose applications with high level of parallelism. The efficiency of running general purpose applications on GPUs depends on how well the processing and memory demands of the application is balanced with the hardware resources available on the target GPU and it can significantly affect the power and performance of the application execution. In this study, we are proposing a model that relates the architectural parameters of the GPU to the characteristics of the application running on it. The model is used to predict the GPU configuration that results in the best power-performance that the application can achieve running on the GPU. We compare the model produced Optimal configurations to actual optimal configurations obtained from simulations and show that the Optimal configurations obtained from the model is very close to the actual ones.