I'm using assimp to load a model, and DirectX11 and DirectX ToolKit to render it.
In the document of DirectXTK's IEffectSkinning, it says:
As an optimization, SkinnedEffect and DGSLEffect assume that all the matrices are affine transformations, and that the final column is (0 0 0 1). This means that the value of the last column is effectively ignored when set into the constant buffer containing the bone transformations (i.e. the shaders use float4x3)
Is this the convention when implementing skeletal animation ?
Does it have negative repercussions ? If it does, what are they ?
Do some animated models require the translation components to be considered?