Simon Brown@sjb3dGot a good 20% speedup moving more non-coalesced memory reads to texture reads though (yay texobj-by-value on SM3+), becoming rule of thumb.9:35 AM · Jun 10, 2013·Twitter Web Client1 Like
