I was just going to ask this, too. From the tilemapping end, 8x8 is much easier and it is faster. I don't know if you saw the screenies in some of the other topics, but I included some tiles that were grouped together to form 16x16 tiles and some even larger. Using 8x8 tiles will give us the ability to still create and use 'fake' 16x16 tiles, along with 8x8 tiles for better detail in that way. From the perspective of data size and compression, I have been working on a tilemap compression routine that will compress be able to store large tiles (like houses or very large trees or anything else that is comprised of multiple 8x8 tiles) as a single byte, then upon decompression to RAM, it will simply take that byte and expand it to however large it needs to be. I would get into more detail on the compression methods I have in mind, but I will save that for the Coding topic.

YOu have a struct like this:

Code: [Select]

typedef struct{
  unsigned short x;
  unsigned short y;
  short xvel;
  short yvel;
}Bullet;
And you can init it like this:

Code: [Select]

Bullet bul = {.x = 20, .y = 35, .xvel = 2, .yvel = 3};

YOu can later add things like type id, lifetime, or whatever you want!
And adding xvel and yvel to the bullet:

Code: [Select]

bul.x += bul.xvel;
bul.y += bul.yvel;