In discrete event systems, a given task can start executing when all the required input data are available. The required input data for a given task may change along the evolution of the system. A way of modeling this changing requirement is through multi-guarded tasks. This paper studies the performance evaluation of the class of marked graphs extended with multi-guarded transitions (or tasks). Although the throughput of such systems can be computed through Markov chain analysis, two alternative methods are proposed to avoid the state explosion problem. The first one obtains throughput bounds in polynomial time through linear programming. The second one yields a small subsystem that estimates the throughput of the whole system.