Can anyone shed some light on ==>

    --- The algorithm (modelling, filling, clipping, etc etc)
        techniques used by the GDI & DirectX and how
        using different
        HW accelerators makes no difference to the
        look and feel of graphics(windows, buttons, etc etc)
        on a Windows desktop?

     --- Are these algorithms made open for the benefit
         of device drivers who would seek to directly
         emulate them on their HW(one less task for
         the OS)?

I am basically interested in dividing the work done between
Windows and my own software but am worried that the
look of the graphics would be noticeably different.

I want to have the same look but divide the work between
Windows and my SW to the extent(and this is vital) that
the part of the work done by the SW includes everything apart
from displaying the resultant BitMap which would of course
be done by Windows.

Has anyone done any research on this?