SPAGHETtI: Scheduling/Placement Approach for Task-Graphs on HETerogeneous archItecture

We propose a new algorithm, called SPAGHETtI, for static scheduling tasks on an unbounded heterogeneous resources where resources belongs to di↵erent architecture (e.g. CPU or GPU). We show that this algorithm is optimal in complexity O(|E||A| + |V ||A|), where |E| is the number of edges, |V | the number of vertices of the scheduled DAG and |A| the number of architectures – usually a small value – and that it is able to compute the optimal makespan. Moreover, the number of resources to be used for executing the schedule is given by a linear time algorithm. When the resources are bounded we provide a method to reduce the number of necessary resources up to the bound providing a set of compromises between the makespan and the size of the infrastructure.