A Task Scheduling Algorithm For Heterogeneous Multicore Systems Under The Constraint Of FPGA Computing Resources

Because the heterogeneous multicore system contains different types of computing units,it can flexibly execute different types of tasks and improve the computing per-formance of complex embedded systems.Now FPGA accelerators are being integrated into heterogeneous multicore systems.The heterogeneous multicore system integrated with FPGA is a multicore system with both efficiency and flexibility.Unlike general purpose processors that use instruction sets to execute tasks,FPGA use hardware com-puting resources to execute tasks,but hardware computing resources on FPGA are lim-ited,so this should be considered when scheduling tasks on heterogeneous multicore systems integrated with FPGA.This paper proposed a task scheduling algorithm for heterogeneous multicore systems under the constraint of FPGA computing resources.The purpose was to reduce the processing time of applications and consider the resource limit of FPGA.The algorithm optimized the scheme of allocating tasks to computing units,which not only considered the execution time of tasks on different computing units,but also made full use of the resources on the FPGA,and ensured that the com-puting unit allocation of tasks during the algorithm operation always met the resource limit on the FPGA.The scheduling sequence of tasks could also be optimized in the algorithm.The main contributions of this paper were as follows:1.This paper realized the modeling of heterogeneous multicore systems integrated with FPGA and the modeling of program tasks.This paper analyzed the communi-cation rate between computing units of a heterogeneous multicore system integrated with FPGA,the limit of FPGA computing resources,the dependencies between tasks in applications,the execution time of tasks on heterogeneous computing units,and the amount of FPGA computing resources consumed by a task.2.This paper proposed a task scheduling algorithm for heterogeneous multicore systems under the constraint of FPGA computing resources.The algorithm was based on a genetic algorithm,where chromosomes consisted of the number of computing units to which tasks were allocated.When generating the initial population,some tasks were randomly assigned to the FPGA under the FPGA resource constraint.The algorithm modified the crossover operation and mutation operation of the genetic algorithm to en-sure that the FPGA resource usages hid in the chromosomes did not exceed the resource limit of the FPGA.Task scheduling was done in the chromosome evaluation phase.Through the genetic algorithm,the improved task allocation scheme and scheduling sequence were obtained.3.This paper developed a task scheduling tool for heterogeneous multicore sys-tems integrated with FPGA.The tool could give the task scheduling results presented in Gantt charts through the data and files input by the user,and visually displayed the allocation of tasks on the computing units,the duration of each task,and the direction of each communication process.A practical application was analyzed based on this tool,illustrating the feasibility of the method in this paper.The experiments in this paper scheduled random applications and three real appli-cations used in multiple works.Through analysis and comparison,our method achieved better performance than three state-of-the-art methods in the same field.