Research on Evaluation Strategy of Heterogeneous Computing Chip Based on Improved Common Origin Grey Clustering

Heterogeneous computing chips are widely used in AI, cloud computing and other fields based on their high performance computing capability, scalability, and high utilization of computing resources. However, there are relatively few researches based on the performance evaluation of heterogeneous computing chips at present, and many companies and individuals have many uncertainties in the convenience of selecting heterogeneous chips. In response to the above problems, this paper constructs a heterogeneous computing evaluation index system based on chip network performance, business support capability, security, virtualization and other dimensions, and then proposes a heterogeneous chip evaluation method based on hybrid cross-weighting and improved co-origin grey clustering, and proposed a hybrid cross-weighting method based on the variation coefficient order relationship method. At the same time, in view of the inaccuracy of the gray class crossing and the rate change of membership of the common origin clustering function, the function interval and function form of the common origin gray clustering are modified to make the evaluation result more reasonable and authentic.