During the last years, complex reactive systems used under real time constraints became more and more important. In this paper we discuss how a time-critical algorithm in form of a task-graph is mapped on a heterogeneous multi-processor system. For every single task of this graph we fix the spatial and temporal location as well as the inter-processor communication structure. Our methods are based on mixed integer linear programming. Though dealing with a NP-hard problem, we even find for larger applications optimal solutions because our methods exploit the restriction of the solution space as far as possible using static analysis and stochastic algorithms.