Tilemap: vertex array or prerendered texture?

[Tenry]

With SFML 2.0 there are new possibilities drawing repeated backgrounds and tilemaps. I've already read something about using vertex arrays, but would it be faster than using prerendered textures?

E.g., a static tilemap could be prerendered on a render target and then drawn on screen as it would be a normal sprite. Of course, it uses a little bit more of memory, but wouldn't this method be faster?

[Chris12]

Yes it would be faster. But you will still have to split the complete, huge, prerendered texture, into smaller textures.
Since the maximum "renderTexture" is dependant on your GPU,
i think 1024 would be ok for most gpus, but you will have to check it yourself.


Whatever, as I said you will still have to split it.
I will implement some kind of grouping into my tilemap so a 20x20 part of my levels tilemap will be rendered into one rendertexture.

[Tenry]

Yes it would be faster. But you will still have to split the complete, huge, prerendered texture, into smaller textures.
Since the maximum "renderTexture" is dependant on your GPU,
i think 1024 would be ok for most gpus, but you will have to check it yourself.


Whatever, as I said you will still have to split it.
I will implement some kind of grouping into my tilemap so a 20x20 part of my levels tilemap will be rendered into one rendertexture.

Yeah, thought of it, too.
But wouldn't it be waste as well when vertex arraying a such large map and drawing it?

[Laurent]

Use vertex arrays. You shouldn't get performance issues, and your code will be much cleaner.

[Tenry]

Use vertex arrays. You shouldn't get performance issues, and your code will be much cleaner.

Okay.

... drawing repeated backgrounds ...

I noticed a method in the texture class: SetRepeated.

That way, I don't need to define a lot of quads for each time the background is drawn, I could also use a single quad and the making tex coords like tex_width*2, tex_height*2, right?

[Laurent]

That way, I don't need to define a lot of quads for each time the background is drawn, I could also use a single quad and the making tex coords like tex_width*2, tex_height*2, right?

Yep.