A data-driven approach for supporting extension processing

Extension processing (Cai, 1999) has been developed and studied for solving contradictory problems in the real world. Due to the huge amount of computing power needed in the extension, it is important to develop a higher performance scheme for supporting extension processing. In this paper, a data-driven (Gandiot et al., 1996; Nikhl et al., 1992) approach to support the parallel processing of matter elements in an extension system is presented. According to the nature of matter elements, divergent, conjugate, correlative, and implicative, the related attributes can be derived and found efficiently through the proposed data-driven mechanism. In the data-driven execution model, any operation with needed operands available is ready to fire. Those attributes are deemed as the operands needed by the operation, execution is ready to fire as long as the operands are available. Therefore, a higher performance and throughput can be achieved. The evaluation of the scheme is developed and tested according to the requirement of the extension process. From the simulation results, the proposed scheme and associated architecture can provide an effective environment for Extenics