Contacts

Probabilistic software analysis aims at quantifying the probability of a target event to occur during a program execution, given a probabilistic usage profile. The probability distributions over the input values specified by a profile can be used to quantify the probability of satisfying a branch condition (and its complement) of sequential programs. On the other hand, the behavior of multithreaded programs is affected not only by users inputs, but also by the choices of the scheduler assigning the CPU to each of the ready threads. A problem for multithreaded programs is to identify a scheduler that minimizes or maximizes the probability of the target event to occur. In [4] an approach reminiscent of dynamic programming is used for this purpose. The goal of this seminar is to replace the dynamic programming approach of [4] with an ant-optimization strategy, able to find a (near-) optimal scheduler via a partial exploration of the execution tree