Sprite mapping format thoughts

mdtiler can generate sprite mappings (manually) using the sprite instruction, which both reads tiles and generates a sprite mapping entry in one go. The format used by mdtiler looks like this:

• 2 bytes: X coordinate
• 2 bytes: Y coordinate
• 2 bytes: Tile ID and flags
• 2 bytes: Sprite size

And the list is ended with $8000.

But in hindsight, it's kinda wasteful: the upper 12 bits of the sprite size are never used (always 0). Even if you want to cram in flags in the unused space of the sprite table, the upper 8 bits are never used at all. Also, to check the end of the list, you need to check if X = $8000, which wastes cycles.

Probably a better idea would have been to do this:

• 2 bytes: Sprite size
• 2 bytes: X coordinate
• 2 bytes: Y coordinate
• 2 bytes: Tile ID and flags

And again the list ends with $8000 (or larger). Then the "end of list" condition comes for free when reading the sprite size since the sign flag will be set, avoiding wasting cycles on a cmp instruction.

Maybe one day I'll update mdtiler to support the new format, or even custom formats (where you specify the order of each entry).