An improved algorithm for implication testing involving arithmetic inequalities

Implication testing of arithmetic inequalities has been widely used in different areas in database systems and has received extensive research as well. Klug and Ullman (A. Klug, 1988; and J.D. Ullman, 1989) proposed an algorithm that determines whether S implies T, where T and S consist of inequalities of form (X op Y), X and Y are two variables, and op/spl epsiv/ {=<, /spl les/, /spl ne/, >, /spl ges/}. The complexity of the algorithm is O(n/sup 3/), where n is the number of inequalities in S. We reduce the problem to matrix multiplication, thus improving the time bound to O(n/sup 2.376/). We also demonstrate an O(n/sup 2/) algorithm if the number of inequalities in T is bounded by O(n). Since matrix multiplication has been well studied, our reduction allows the possibility of directly adopting many practical results for managing matrices and their operations, such as parallel computation and efficient representation of sparse matrices.<
>