ECE497NC Final Report: RC Kernel Selection and Scheduling

Runtime Reconfiguration (RTR) allows reconfigurable architectures to adjust to application phased behavior at runtime. In this paper, we examine the performance of one such RTR system and the impact of system parameters such as reconfigurable array size and reconfiguration time. To gauge the effectiveness of our RTR algorithm, we compare its performance against an equivalent system that relies on static kernel scheduling where the compiler selects a set of kernels that will be mapped onto reconfigurable logic for the duration of an application. We demonstrate a speedup of up to 50% for a range of logic sizes over a static schedule when using a simple RTR algorithm for reasonably-sized reconfigurable coprocessors and suggest improvements to the algorithm that should widen this gap even further.