> Across all tested generations, divergent paths serialize linearly with the number of paths k, following T(k)≈sk with no super-linear reconvergence penalty. Warp execution efficiency falls as 32/k, the penalty is independent of occupancy, and predication removes the serialization cost
> Across all tested generations, divergent paths serialize linearly with the number of paths k, following T(k)≈sk with no super-linear reconvergence penalty. Warp execution efficiency falls as 32/k, the penalty is independent of occupancy, and predication removes the serialization cost
So... Nvidia did a good job?