Recently, an analysis has been developed that allows the static prediction of cache contents and thus of execution times. In this talk we will show how this analysis can be incorporated into the computation of a pre-runtime schedule, in order to obtain better processor utilization.