"Performance Evaluation of Data Intensive Logic Programs," F. Bancilhon and R. Ramakrishnan, To appear in Foundations of Deductive Databases and Logic Programming, Ed. J. Minker, Morgan Kaufman, 1987.  Home/Search   Document Not in Database   Summary   Related Articles

....generated by the Counting strategies. As we remarked earlier, since computing the indices represents an additional overhead, we expect the Counting methods to be used only when the optimizations described in this section are applicable. The improvement can be considerable, as the analysis in [Bancilhon and Ramakrishnan 87] indicates. Consider the rewritten program produced by the Counting method for the same generation example (Example 6) If we examine the second rule defining cnt sg ind bf : cnt sg ind bf (I 1, k 2 2, h 5 4, Z3) cnt sg ind bf (I, k, h, X) up(X,Z1) sg ind bf (I 1, k 2 2, h 5 2, ....

....costs. This favors strategies such as Prolog since the cost of generating queries is not measured, at the expense of strategies like Generalized Magic Sets, which generate additional facts (the facts in magic predicates) in order to generate subqueries. This was, in fact, the approach taken in [Bancilhon and Ramakrishnan 87] and the results of that study indicate that the number of magic facts is, in general, a small fraction of the generated facts. We note that any method proceeding according to a given collection of sips must evaluate all queries in Q and all facts in F, and the Magic Set method does not generate ....

"Performance Evaluation of Data Intensive Logic Programs," F. Bancilhon and R. Ramakrishnan, To appear in Foundations of Deductive Databases and Logic Programming, Ed. J. Minker, Morgan Kaufman, 1987.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC