I am developing a simple 2D spriting engine. I was wondering if it would be a better idea to include matrix modelviewproj transformation data within the vertices themselves (will add an additional parameter to the vertex layout from CPU to GPU vertex shader) or simply use a constant buffer and change that constant buffer for each sprite.
From learning about Direct3D11 I find that you should try to minimze the calls to the GPU. If this is the case wouldn't packing the matrix data for each frame within the vertices themselves be a better idea? or would the extra bandwidth (matrix's cost alot of memory) for the extra parameter be too expensive?