Generally speaking, is it better to compute rotations of vectors into tangent space with a vertex shader or with the cpu? ( Assuming that lighting positions can move and specular & other components will be computed in the fragment shader )
I know the questions really depends, but what would you rather do (or which do you use now? )
In most situations the GPU calculation is better. If you do it on GPU you will have following benefits:
[ul][]The power of GPU vertex shaders is mostly unused so offloading some work from CPU which can then be used for physics or better AI is usefull.[]All geometry can be static and stored on the card which reduces bus bandwith necessary for rendering.[]It allows you to do skinning of characters on the GPU.[]Single geometry can be shared by several instances of mesh (think army of soldiers or bar room with many chairs) which can significantly reduce memory requirements.[*]On some HW you can easily use various packed formats to store normal values so this also reduces memory requirements.[/ul]
This topic was automatically closed 183 days after the last reply. New replies are no longer allowed.