System-level architectural exploration using allocation-on-demand technique

Architectural exploration is very important in embedded system design and SoC design. In this paper, a new heuristic algorithm using the allocation-on-demand technique is proposed to solve this problem. Unlike previous research efforts, this algorithm allocates new resources only when it fails to schedule tasks under the performance constraints. So the resource costs of the system increase monotonously in running, which is apt to determine the feasibility of current solution earlier. Experimental results show that this approach is helpful for an efficient architectural exploration process.