Previous: Search Engine
Up: Constructing a Metagame-player
Next: Meta-Level Evaluation Function
Previous Page: Search Engine
Next Page: Meta-Level Evaluation Function

Automated Efficiency Optimisation

This powerful search engine should allow a playing program to search deeply. However, search speed is linked to the speed with which the primitive search operations of move-generation and goal detection can be performed. For a game-specific program, these issues can be easily hand-engineered, but for a program which is to play an entire class of games, the excess overhead of such generality initially caused the search to be unbearably slow. This problem was overcome by using a knowledge-compilation approach. We represent the semantics of the class of games in an extremely general but inefficient manner, and after receiving the rules of a given game the program automatically partially evaluates the general theory with respect to it, thus compiling away both unnecessary conditionality and the overhead of interpretation. This is discussed in detail in [\protect\citenamePell, 1993a].