High-performance __device__ sorting function in CUDA

Question

Hi, All

I am looking for the __device__ sorting function for an array with variable length since I want a warp or block of threads to collab