Enhancing Resource Utilization with Design Alternatives in Runtime Reconfigurable Systems

Average resource utilization in reconfigurable systems is dependent on how effectively modules are placed. Resources may be left unused because modules usually have different resource requirements, and consequently different module bounding boxes. When fitting such different modules on an FPGA, the placement can be imperfect. In order to investigate the effect of module design alternatives on resource utilization, we model the placement problem as a constraint satisfaction problem. In particular we are interested in real world FPGAs. The constraint satisfaction problem is solved using a constraint solver. The constraint problem is modelled as a minimization problem in order to attain the optimal placement. We have investigated module design alternatives which include representing the modules with different layout. We have found an increase in average resource utilization of 11% by considering module design alternatives. The computation time is short, so the method could be applied as a part of an interactive tool.