异构多核片上系统的帧任务节能分配算法

针对异构多核片上系统的高能耗问题,提出了一种对帧任务进行分配的高能效两级优化算法。算法将系统能耗最小化问题定义为非线性整数规划问题。第一级优化将问题进行松弛处理,求解得到任务分配矩阵;第二级优化采用带判决门限的取整函数更新分配矩阵,构造新的规划问题并求解;最后结合两级优化得出最终的任务分配矩阵。仿真结果表明,与部分启发式算法相比,该算法能耗降低了20%50%,接近理论最优能耗;与能耗接近的其他优化算法相比,求解时间减少了54%75%。

Algorithm for scheduling energy-saving frame-based tasks on the heterogeneous multi-core SoC

Aiming at the high energy consumption of the heterogeneous multi-core system on the chip, an efficient two-level optimization algorithm for frame task assignment is proposed. The algorithm defines the problem of minimizing system energy consumption as a nonlinear integer programming problem. The first-level optimization is used to relax and solve the problem. The second-level optimization uses the rounding function with the decision threshold to update the assignment matrix, and then constructs a new planning problem and solves it. Finally, combining the two-level optimization results to formulate the final assignment matrix. Simulation shows that compared with some heuristic algorithms, the energy consumption of the algorithm is reduced by 20%50%, which is close to the theoretical optimal energy consumption. Compared with other optimization algorithms with similar energy consumption, the complete time is reduced by 54%75%.