Efficiency Mechanisms for a Class of Blackboard Systems

Abstract

This paper presents efficient mechanisms for activation, execution and rating that are suitable for use in BB1-style blackboard architectures. We describe a knowledge source compiler that produces match networks and demons for efficient activation and rating while compiling the entire system for increased execution speed. Experiments using the enhancements in a general-purpose blackboard shell illustrate approximately a doubling of run time speed, including an increase in activation speed by a factor of 7.6 on the average. We have also resolved a subclass of blackboard systems that can be compiled down to the machine level by using a condensed representation where low-level blackboard accesses are replaced by vector references. Our analysis shows that the time complexity of the execution cycle of a condensed system is faster than the conventional approach by the ratio of the time required for blackboard retrievals to the time required for vector element retrievals. In practice, this ratio is approximately four orders of magnitude.