My new 3d graphics engine coputes vertex transformation matrices manually. I
want to try d3d vertex blending but I'm still not sure if I need to use it.
It is easy to use. Also I think it is fast. It will be realy fast If there
is a hardware vertex processing supported graphics accelerator IMHO, but I
don't have any definite opinion about computation speed using software
vertex processing. It can be faster than manual computation because
optimizations, but can't be as fast as hardware vertex processing IMHO.
Consider that I use d3d vertex blending. I hold weights and matrix indices
in d3d vertex buffers with other vertex data (xyz, normal, tex0, tex1) and
directx does the calculations of transformed vertices. In this case I still
need transformed vertex positions to use in shadow volumes and physics.
Actually, transformed vertices may be excluded while processing physics,
because there can be a different model (more low poly) with different
vertices to process physics. However, transformed vertices are still
necessary for shadow volumes. So I still need a manuel coputation to obtain
transformed vertices. If I won't use d3d vertex blending, transformed
vertices can be computed only one time by manual processing.
Is there any advice?