I found that in DirectX 9 if you use an already transformed vertice that no other transformations matter, ie. projection, world, view. At the moment I have all my vertices software processed into the correct position and size for any resolution. What I am wondering is how I can use those transformations to hardware process my vertices? And... Does DirectX 10 do rotations for point sprites?

I am working on a 2D game, but I am doing this for each vertice that represents each sprite:

(Just an example)
posX = (float)pos_x*posRatioX;
posY = (float)pos_y*posRatioY;

These position ratios determine the correct location for each vertice of the sprite, but the problem is that I am using the CPU and not the GPU for this. It would be much faster to implement matrices for the GPU to support the same functionality.

As a side note...

I am not looking for instancing as each sprite is unique in location, and I have already implemented texture similarity rendering. Using pretransformed vertices automatically suggests to the GPU to not do anything whatsoever to the vertice during the rendering process.

I was wondering if anyone knew of a way to show 2D images in a sprite like state while being able to use the Projection, View, and World transformations for these sprites.