自组织芯片设计和映射问题的研究

Research on self-organizing IC chip design and mapping

信息处理算法到芯片的映射是大规模集成电路设计的焦点问题。借助硬件进化(evolvable hardware,EHW)的工作机理,本文研究了另一种全新的信息处理算法的自组织映射方法。考虑到信息处理算法的多样性,以及算法映射对学习机制在数据构造问题上的不同针对性,本文从自适应无损图像压缩硬件实现这一典型的问题出发来考虑和研究算法映射的关键问题。分析指出了现有EHW工作模型存在的不足,提出并探讨了适应度评估机制的计算复杂性、模型的可扩展性以及模型的可移植性3个关键问题,并给出了针对性的解决方案。

The mapping of information processing algorithm to chip is the focus of large scale integrated circuit design. In virtue of the working mechanism of EHW （evolvable hardware）, a completely different self-organizing mapping method for information processing algorithm was studied. Considering the diversity of information processing algorithms and pertinence of data generation of learning mechanism, the paper studies the key problem - that is adaptive lossless image compression hardware realization, and points out the shortages of current EHW working model. Three key problems including computation complexity of fitness evaluation mechanism, model scalability and model portability are put forward and discussed, and a sound scheme is given.