Visualize how GPU warps of 32 threads execute instructions in lockstep (SIMT), handle branch divergence with active masks, and get scheduled by the warp scheduler.
All 32 lanes execute the same path. Maximum SIMT efficiency with full lane utilization.