A series of large-scaled simulations on cascaded double-jet film cooling are performed by using hybrid thermal lattice Boltzmann method and multiple Graphic Processing Units to study the effect of streamwise spacing l between coolant jets on the flow and heat transfer mechanism in film cooling. The streamwise spacing is set as l = 3D, 5D, 7D, 9D, 12D. Here, D is the diameter of film cooling hole. The coolant jet is injected at an inclined angle α = 30° into a turbulent flat plate boundary layer profile at a free-stream Reynolds number of Re = 4000 . The simulated results show that the flow and heat transfer behaviors of film cooling vary much with l. The mixing between the second (downstream) coolant jet and hot crossflow is strengthened with the increasing in l. Meanwhile, the flow and heat transfer characteristics vary considerably when the second jet hole respectively locates in the rotating domain and dissipation domain of the first (upstream) jet. On the other hand, the present simulation with about 2.22 × 10⁸ grids is fulfilled by CUDA-MPI on 12 K20M GPUs and a high computational performance of 935.28 MLUPS is achieved.