These performance instabilities make it impossible, now and in the future, to use a single number to evaluate a given machine, as has been done in the past. performance ranges have arisen on single machines. As vector and parallel machines have been introduced, wider and wider.
The effort was started in order to shed light on computer system performance for scientific and engineering applications.
GRAPHIC CARD BENCHMARK 2015 SERIES
This report is the second in a projected series of periodic reports intended to present the results to date of the Perfect benchmarking effort and to encourage wider participation in the computing community. Finally, we discuss several considerations for systematically constructing future benchmark suites. The experimental results show that we are able to achieve satisfactory performance predictions, although errors are higher for outlier applications. We demonstrate the prediction methodology and conduct predictions with benchmarks from different suites to achieve better workload coverage. We first identify a set of important GPU application characteristics and then use them to predict performance of an arbitrary application by determining its most similar proxy benchmarks. One tenet of benchmark design, also a motivation of this paper, is that users should be given the capability to leverage standard workloads to infer the performance of applications of their interest. In this work, we study the approach of predicting the performance of GPU applications by correlating them to existing workloads. Furthermore, methodologies and associated tools are needed to analyze and predict the performance of GPU applications and help guide users’ purchasing decisions. It is critical for researchers to understand the first-order metrics that most influence GPU performance and scalability. This makes GPU workloads demonstrate application characteristics different from those of CPU workloads. GPUs present significantly different architectures from CPUs and require specific mappings and optimizations to achieve high performance. Graphics processing units (GPUs) have become an important platform for general-purpose computing, thanks to their high parallel throughput and high memory bandwidth.