There are some smaller things that you can take from it:
- if you are going to do 4-plane graphics with fairly small polygons, then you could construct 8 pixels x 4 bitplanes worth of data in a longword and use MOVEP for reading/writing. However if you expect to be having larger polygons, or perhaps not doing 4 full bitplanes, then it is probably faster to do directly planar with word/longword accesses for individual bitplanes.
- If you are going to do 2-plane graphics with fairly small polygons and 2x1/2x2 chunky pixelsize, then you can have a linear packed framebuffer (a1a0b1b0c1c0d1d0 ...), render like that to the first bitplane, and construct the contents of the second bitplane by doing a 1-bit swap on the contents on the first bitplane (a0a1b0b1c0c1d0d1 ...) and use a predefined 01010101 mask in bitplane 3. The 1-bit swap can be done as a table lookup on the bitplane data (256 bytes or 128k table needed) or perhaps via a blit (I don't know the STEs blitter capabilities well enough).
But overall, c2p approaches become much more useful if you are not looking to draw contiguous segments of pixels with the same color.