However, our Pull–Model approach shows significant performance gains over the more traditional push–model approach,nearlytwoorders ofmagnitude.Sincethepush–model and the Pull–Model both execute the same number of clock tick events for a given program, the gains clearly come from the limited scope of the transitive propagation of the querying for a known output value as discussed previously. The three different experiments differ only by the number of outer–loop iterations, and the performance numbers show linear growth of execution time,as the outer–loop count increases,as expected