In the context of each multiprocessor (or SM, using NVIDIA terminology), my understanding is that the GPU executes warps in SIMD/SIMT fasion. What I don\'t understand is why
