A Task Scheduling and Placement Strategy Based on Tasks' Aspect Ratio

Abstract

Reconfigurable computing (RC) is a compromise of General-propose processor (GPP) computing and Application Specific Integrated Circuit (ASIC) computing with both hardware efficiency and software flexibility. An efficient algorithm to tackle the scheduling and placement problem for the dynamically reconfigurable Field-Programmable Gate Arrays (FPGAs) with real time decisions is highly concerned for the performance of RC. By exploring 2-dimensonal (2-D) reconfigurable hardware resource model and hardware task scale information, we propose an efficient task scheduling and placement algorithm by cooperating Tasks' Aspect Ratio (TAR) with 2-D reconfigurable resource model. Simulations show that our method improves the success ratio up to 9.7% compared with Horizon algorithm and DATS algorithm. TAR algorithm also achieves better resource utilization.