Abstract: In order to solve large-scale high-dimensional nonlinear function optimization problem, a novel cooperative differential evolution algorithm based on GPU is proposed. The coevolution principle is introduced to the heuristic differential evolution algorithm；and then the large-scale computational problems are decomposited randomly. Using GPU parallelism, simultaneous calculate decomposed sub-problems, which improves the accuracy and speeds up the algorithm convergence process. Experimental comparison results show that the proposed GPU-based cooperative differential evolution algorithm has better accuracy and efficiency for large-scale nonlinear function optimizations.