Hi! Welcome!
The main point of using tiles is so that they can be repeated.
If you are not repeating anything at all, there's no point to split things into smaller pieces.
With that said, the texture size limit then forces breaking down larger images into multiple textures but each textures does not need to be broken into multiple smaller tiles unless they are going to be repeated (or will be or can be used in a different order).
If I remember correctly, Thor (library for use with SFML) provides the ability to draw an image larger than you can with a single texture using multiple textures.
This might be moot depending on the size of the textures available to the more modern graphics cards but increasing resolutions basically tip the balance back again. If a card can handle 32k x 32k textures, you can work with 4k and 8k resolutions directly. If not, those resolutions would require either stitched textures (see Thor recommendation above) or using a lower resolution scaled up.
It's not fully clear what your map requirements actually are. Do you need full, single images that don't repeat or break down into smaller components?
It's worth noting that often it can be optimistic (and also have the side-effect of helping with parallax effects) to use multiple layers. For example, a background that is not so complicated and can be represented using tile and probably doesn't move as much so less image needed overall, and a foreground that comprises of specific components; these components can be repeated when needed but, even if not, they don't need to represent the entire view, just some parts. I personally would call these components "free tiles" or something similar (as I did in Cheese Map).
Providing multiple layers
may 'cost' more, of course, depending on the design.
So, really, what do you need? Full-view maps that can be represented by a single image? If so, apart from the consideration mentioned above, draw it with a single tile from a single texture. You would need to be able to draw the next (in any direction) map image so you'd need to be prepared for that by drawing the 8 (all direction) images around the one you can see. If a texture cannot store multiple of these images, you'd need a different texture for each.
Your considerations then would be when to load a texture that a player is approaching but no yet reached and also when to unload a texture that a player has left or, better still, will definitely no be returning to.
You will also need to consider your process for dealing with the different possibilities of each graphics card and work different for each situation (e.g. splitting a single image into multiple 'tiles' with different textures if it can't deal with a size of texture that doesn't fill the screen (this is quite rare but 1k x 1k or 2k x 2k are usually expected so 4 (for 2k textures, or 16 for 1k textures) textures should be enough to fill a 4k view.
Again, though, really consider if you might be able to break down an image into smaller, preferably repeated, components. Most images are made up of smaller components and if you can find a way that it works for your design, it should be considered.
Rendering ALL the world in a RenderTexture? But is RenderTexture limited in size as well as regular textures? And would it consume large GPU memory as well?
Render textures render to a texture so the requirements for textures also apply to render textures.
Show Posts
