Task scheduling algorithm based on priority transformation in heterogeneous platforms

In heterogeneous platforms, efficient task scheduling is a crucial condition for achieving high-performance computing. To address this requirement, this paper proposes an improved genetic algorithm and conducts task scheduling based on priority transformation. The algorithm combines a priority queue and a processor mapping queue to form a hybrid encoding of chromosomes in the genetic algorithm. It enhances the implementation methods of selection, crossover, and mutation to improve the efficiency of task scheduling. Adaptive mechanisms, elitism preservation and degradation-extinction mechanisms are introduced to prevent the genetic algorithm from falling into local optima, thereby increasing convergence speed and stabilizing scheduling results. Finally, simulation experiments are conducted, and a CPU-GPU-FPGA heterogeneous platform is built using OpenCL for validation. Compared to standard genetic algorithms and classic algorithms such as HEFT and CPOP, the proposed algorithm demonstrates more stable optimal solution acquisition capabilities and better scheduling performance.