Warp-Level Parallelism: Enabling Multiple Replications In Parallel on GPU