专用指令集处理器(ASIP)评估方法研究

Evaluation method of application specific instruction set processor

针对专用指令集处理器(ASIP)评估具有多属性维数、多目标类型、多数据类型的特点,提出一种基于比较的评估方法。在评估的不同阶段,选取不同的参照指标,对数据进行处理、集结,从而获取候选方案的排序向量。根据评估目标类型和数据信息类型选取特定的参照数据,进行数据信息与参照数据的比较;利用模糊矩阵对评估指标的重要程度进行两两比较,并对其进行一致性判断和修正,获取指标权值向量;最后,利用有序加权平均法获取各候选方案的综合属性值,并提出区间比较度的概念,为评估提供量化依据。实例计算表明了该方法的有效性。

Due to the features of application specific instruction set processor （AS/P） evaluation including multi-attribute, multi-data type and multi-attribute type, a comparison-based evaluation method, which processes data by selecting different references to generate vector for candidates is presented. Firstly, reference data is selected according to the attribute type and data type and comparison between evaluation data and reference data is done. Then each two attributes are compared to build and adjust fuzzy judgment matrix to obtain attribute weight vector. Finally, the method of ordered weighted average （OWA） is used to compute integrated value for each candidate and the concept ofinterval comparison degree is presented to provide quantifiable base for the evaluation of ASIP. An application instance demonstrates the validity of the presented method.