The colour-boosting feature allows you to obtain 50-100 colours from a single 16-colour palette without raster splits. It involves heavy offline processing of the graphics with a tool and the display arrangement is more complex. It also costs more to maintain the display updates.

A few notes:

- it flickers more than the others because the input map is already dithered (mainly between grey and brown), and my convertor re-dithered it into 2 fields. so it basically needs more care over the conversion for this one.
- the slow scroll is deliberate - its still 50hz though. I was asked about subpixel scroll recently so decided to use it here. the scroll vector and entity vectors are 16:16 fixedpoint, although the demo uses a table of only 32 equally spaced directions.
- the test map is 1264x1440 in 4 planes. the original version was 5 planes.
- the code is nearly identical to the shmup demo, except for the entity list and AI code (and extra bit for BG collisions). the rest is almost line-for-line.
- there are 40+ monsters on the map. they are all randomly spawned on non-solid tiles.
- each monster performs object-object collision reaction and object-background collision reaction. object-object one of the more sophisticated parts of the engine (as the shmup demos should hopefully show - it allows a lot of dynamics at very low cost).
- the object-background collision is tile based, using a list of solid tile indices. I won't detail the process here but it didn't take very long even if it was done manually, using map-diagnostic images emitted from agtcut. it helps that the map is quite well made.
- the sprites don't use many colours but drawn 4-planes regardless.
- this doesn't use the fastest sprite path in the engine, because the animation+rotation frames at 32x32 using my code generator just made it too big - about 300k (!). while I still can cut this back to maybe 200k its still too expensive - so it uses the lower cost version (like Anima's first method minus codegen, no line reordering). It would be worth using his newer in-between method for this demo, for ram vs speed - it seems an ideal fit. I've not had time to add it yet though. it does seem the masks for these sprites do not optimize too well so its hard to tell what a final version would be like at this point, just need to experiment more.