hgtransform is interpreted in software for the painters renderer or z-buffer renderer.
For the OpenGL renderer, command and data node structure is created enclosing all of the data, the transform commands being marked as applied to the result of the other operations, and the complete structure is sent to the OpenGL renderer to process. The actual transformations are not done at the MATLAB level in this case: instead what to do is sent to OpenGL. In turn OpenGL might send the structure (or parts of it) to hardware rendering routines.
If that is not what you meant, then perhaps read https://en.wikipedia.org/wiki/Transformation_matrix and understand that transform matrices are applied to coordinates to find new coordinates. After that, it is a matter of rendering the transformed coordinates just as if the the coordinates had always been that value.
