ABSTRACT
In this chapter we present a new tiled lighting variant with a primary focus on optimization for the AMD Graphics Core Next (GCN) architecture. Our approach was used for the game Rise of the Tomb Raider. In particular, we leverage asynchronous compute by interleaving light list generation with rendering of shadow maps. Light list generation is done per tile in two steps within the same compute kernel. An initial coarse pass that generates a light list in local storage using simple screen-space AABB bounding volume intersection testing regardless of light type. The second step is fine pruning, which performs further testing on the coarse list by testing each pixel in the tile if the point in 3D space is inside the true shape of the light source.
