3D scheduling based on code space exploration for dynamically reconfigurable systems

This paper treats scheduling tasks on a dynamically and partly reconfigurable device. We propose a coding scheme which represents both temporal schedule of module reconfigurations and task executions and spatial placement of reconfigured modules. Registers for transferring data between tasks are also considered as run-time configured modules, and the highlight of the proposed coding scheme is the mechanism to guarantee nonoverlapping between 3D (spatial×temporal) life-durations of different objects and temporal overlapping between some pairs of modules to complete data transfer. As a design method to optimize schedule on a dynamically and partly reconfigurable device, simulated annealing (SA) based exploration of the solution space constructed by our coding scheme is demonstrated.