“…However, using more complex basis functions or support for non‐diffuse transport increases the computational overhead. Various optimizations have been developed that improve the computational complexity and memory requirements, such as adaptive meshing [LTG92], hierarchical [HS92] and adaptive subdivision schemes [DB95], progressive refinement [NFKP95], efficient support for dynamic scenes [BBS95,FYT95,ÁDE96, BMA03, CzSG08], and accelerated view factor computation on the GPU [TYSK06,KGP*15]. While the latter two build upon rasterization‐based visibility testing on the GPU to compute direct view (or form) factors between two elements, we compute complete multi‐bounce transport between any two elements via photon tracing , inspired by the Gebhart method [Geb61].…”