Interleaving and Lock-Step Semantics for Analysis and Verification of GPU Kernels