Develop Reference

How to texture a rectangle with multiple textures in OpenGL

I'm trying to make a 2d map (for a game, think tiled world map) in OpenGL ES 2.0 for an android game. Basically, there are a few tile types that have different textures, and the map is randomly generated from these types, so from game-to-game the map changes but for the duration of a single game it stays the same.
My first thought was to generate a single large texture / image / bitmap (independent from OpenGL) beforehand basically stitching duplicate tile textures together to make the larger map, and then using this single texture for one large map rectangle. In theory I think this is simple and would work fine, but I'm worried that it won't scale well for larger maps and especially on mobile I'll run out of memory with such a large image map. Plus, there's a small set of tiles that are duplicated over and over so it seems like a tremendous waste to duplicate the pixel data in a big texture over and over.
My second thought was having many textures, one for each of the tile textures. But I'm not sure how this would work, texture-binding-wise, would I need the shaders to contain multiple texture references and within the shader have logic for using the right one?
Finally, I thought using a texture atlas could work, have one texture / image with all of the tile data in it, this would be relatively small. But I'm struggling to imagine how to get the maths to work out such that "tiles" or subsections of the map rectangle would use completely different texture coordinates.
Am I approaching this the wrong way? Should I be using a rectangle for each tile? At least this way I can pass the shaders both vertex and texture coordinates for each tile independently. This seems easier, but also seems wrong since the map really is just one rectangle that won't be changing.

My first thought was to generate a single large texture...
Actualy, something like this has already been used in id Software's id Tech since version 4. It's called MegaTexture. Basicaly, it's a big texture, which could also hold additional data.
My second thought was having many textures...
You don't need to hold all the textures in a shader. Do it like this:
Implement a loop with n iterations, where n is how much different types of textures are used.
Inside a loop, bind the current texture type.
Pass any data, like position/color/texture coords to shaders.
Draw all tiles that use the bounded texture. You could use GLES30.glDrawElementsInstanced or GLES30.glDrawArraysInstanced if you are targeting devices with GLES 3.x or an appropriate extension support. Otherwise, draw your tiles using GLES20.glDrawArrays or GLES20.glDrawElements.
Shaders won't be complicated with this approach.
Finally, I thought using a texture atlas could work...
You could use loop here too and compute the texture coordinates for each tile type on CPU, then just pass them to shaders.
Considering your map is not changing through a game session, MegaTexture approach looks good. However, it depends on how large your map is and how much memory is available. Also, note that max texture size is limited. Max size differs from device to device but should be (AFAIK) equal or greater than screen size and at least 64 texels(16 for cube-mapped textures). You can get the maximum texture size on any device using glGet(GL_MAX_TEXTURE_SIZE ).

Libgdx - Texture packer n^2

If I create a texture atlas 2048x2048, must I also make each and every texture imported into the atlas a square of 2 dimension.
Because I'm wasting a lot of space if I do that.
But my guess also is that I have to, because the MipMap filter scales it down diagonally... So if I make a rectangle, it wouldn't know how to scale it...
Any ideas?

From my experience you don't need to use n^2 images in libGDX. You can just put in any sized image in the atlas and libGDX will handle it correctly.
I never had any problems with scaling rectangular images, just using the libGDX code..
I hope this answers your question.

How to create mipmap manually?

I'm not fully satisfied with the quality obtained with the mipmap automatically generated with this line of code:
glTexParameterf(GL10.GL_TEXTURE_2D, GL11.GL_GENERATE_MIPMAP, GL10.GL_TRUE);
I thought to create (with Gimp) various scaled version of my texture for every texture used in my game. For example for a texture for a ball I will have:
ball256.png 256x256 px
ball128.png 128x128 px
ball64.png 64x64 px
ball32.png 32x32 px
ball16.png 16x16 px
1. Do you think is a good idea?
2. How can I create a single mipmapped texture from all these images?

This is not only a good idea, but it is a pretty standard practice (particularly in Direct3D)!
OpenGL implementations tend to use a standard box filter (uniformly weighted) when you generate mipmaps. You can use a nice tent fiter (bilinear) or even cubic spline (bicubic) when downsampling textures in image editing suites. Personally, I would prefer a lanczos filter since this is going to be done offline.
You may already be aware of this, but Direct3D has a standard texture format known as DDS (Direct Draw Surface) which allows you to pre-compute and pre-compress every mipmap level at content creation time instead of load-time. This decreases compressed texture load time and more importantly allows for much higher quality sample reconstruction during downsampling into each LOD. OpenGL also has a standardized format that does the same thing: KTX. I brought up Direct3D because although OpenGL has a standardized format very few people seem to know about it; DDS is much more familiar to most people.
If you do not want to use the standardized format I mentioned above, you can always load your levels of detail one-at-a-time manually by calling glTexImage2D (..., n, ...) where n is the LOD (0 is the highest resolution level-of-detail). You would do this in a loop for each LOD when you create your texture, this is actually how things like gluBuild2DMipmaps (...) work.

Scrolling/zooming a scene in OpenGL and subdivision

We are to develop a scrolling/zooming scene in OpenGL ES on Android, very much like a level in Angry Birds but more like a level in World Of Goo. More like the latter as the world will not consist of repeated layers as featured in Angry Birds but of a large image. As the scene needs to scroll/zoom and therefore a lot of it will not be visible, I was wondering about the most efficient way to implement the rendering, focusing on the environment only (ie not the objects within the world but background layers).
We will be using an orthographic projection.
The first that comes to mind is creating a large 4 vertices rectangle at world size, which has the background texture mapped to it, and translate/scale this using glTranslatef / glScalef. However, I was wondering if the non visible area outside of the screens boundaries is still being rendered by OpenGL as it is not being culled (you would lose the visible area as well as there are only 4 vertices). Therefore, would it be more efficient to subdivide this rectangle, so non visible smaller rectangles can be culled?
Another option would be creating a 4 vertice rectangle that would fill the screen, then move the background by adjusting its texture coordinates. However, I guess we would run into problems when building bigger worlds, considering the texture size limit. It seems like a nice implementation for repeated backgrounds like AngryBirds has.
Maybe there is another way..?
If someone has an idea on how it might be done in AngryBirds / World of Goo, please share as I'd love to hear. They seem to have implemented a system that allows for the world to be moved and zoomed very (WorldOfGoo = VERY) smoothly.

This is probably your best bet for implementation.
In my experience, keeping a large texture in memory is very expensive on Android. I would get quite a few OutOfMemoryError exceptions for the background texture before I moved to tiling.
I think the biggest rendering bottleneck would be with memory transfer speeds and fill rate instead of any graphics computation.
Edit: Check out 53:28 of this presentation from Google I/O 2009.

You could split the background rectangle into smaller rectangles, so that OpenGL only renders the visible rectangles. You won't have a big ass rectangle with a big ass texture loaded but smallers rectangles with smaller textures that you could load/unload, depending on what is visible on screen...

Afaik there would be no performance drop due to large areas being rendered off-screen, subdividing and culling is normally done just to reduce vertex count, but you would actually be adding to it here.
Putting that aside for now; from the way you phrased the question I am unsure whether you have a large background texture or a small repeating one. If it is large, then you will need to subdivide because of texture size limitations anyway, so the question is moot! If it is small, then I would suggest the second method, fit a quad to the screen and move the background by changing the texture coordinates.
I feel like I may have missed something, though, as I am unsure why you mentioned the texture size limitation issue when talking about the the texture coordinate method and not the large quad method. Surely for both this is not a problem for repeating textures as you can use GL_REPEAT texture wrap mode...
But for both it is a problem for a single large texture unless you subdivide, which would make the texture coordinate tactic way more complicated than necessary. In this case subdividing the mesh along texture subdivisions would be best, and culling off-screen sections. Deciding which parts to cull should be trivial with this technique.
Cheers.

Can OpenGL ES render textures of non base 2 dimensions?

This is just a quick question before I dive deeper into converting my current rendering system to openGL. I heard that textures needed to be in base 2 sizes in order to be stored for rendering. Is this true?
My application is very tight on memory, but most of the bitmaps are not powers of two. Does storing non-base 2 textures consume more memory?

It's true depending on the OpenGL ES version, OpenGL ES 1.0/1.1 have the power of two restriction. OpenGL ES 2.0 doesn't have the limitation, but it restrict the wrap modes for non power of two textures.
Creating bigger textures to match POT dimensions does waste texture memory.

Suresh, the power of 2 limitation was built into OpenGL back in the (very) early days of computer graphics (before affordable hardware acceleration), and it was done for performance reasons. Low-level rendering code gets a decent performance boost when it can be hard-coded for power-of-two textures. Even in modern GPU's, POT textures are faster than NPOT textures, but the speed difference is much smaller than it used to be (though it may still be noticeable on many ES devices).
GuyNoir, what you should do is build a texture atlas. I just solved this problem myself this past weekend for my own Android game. I created a class called TextureAtlas, and its constructor calls glTexImage2D() to create a large texture of any size I choose (passing null for the pixel values). Then I can call add(id, bitmap), which calls glTexSubImage2D(), repeatedly to pack in the smaller images. The TextureAtlas class tracks the used and free space within the larger texture and the rectangles each bitmap is stored in. Then the rendering code can call get(id) to get the rectangle for an image within the atlas (which it can then convert to texture coordinates).
Side note #1: Choosing the best way to pack in various texture sizes is NOT a trivial task. I chose to start with simple logic in the TextureAtlas class (think typewriter + carriage return + line feed) and make sure I load the images in the best order to take advantage of that logic. In my case, that was to start with the smallest square-ish images and work my way up to the medium square-ish images. Then I load any short+wide images, force a CR+LF, and then load any tall+skinny images. I load the largest square-ish images last.
Side note #2: If you need multiple texture atlases, try to group images inside each that will be rendered together to minimize the number of times you need to switch textures (which can kill performance). For example, in my Android game I put all the static game board elements into one atlas and all the frames of various animation effects in a second atlas. That way I can bind atlas #1 and draw everything on the game board, then I can bind atlas #2 and draw all the special effects on top of it. Two texture selects per frame is very efficient.
Side note #3: If you need repeating/mirroring textures, they need to go into their own textures, and you need to scale them (not add black pixels to fill in the edges).

No, it must be a 2base. However, you can get around this by adding black bars to the top and/or bottom of your image, then using the texture coordinates array to restrict where the texture will be mapped from your image. For example, lets say you have a 13 x 16 pixel texture. You can add 3 pixels of black to the right side then do the following:
static const GLfloat texCoords[] = {
0.0, 0.0,
0.0, 13.0/16.0,
1.0, 0.0,
1.0, 13.0/16.0
};
Now, you have a 2base image file, but a non-2base texture. Just make sure you use linear scaling :)

This is a bit late but Non-power of 2 textures are supported under OpenGL ES 1/2 through extensions.
The main one is GL_OES_texture_npot. There is also GL_IMG_texture_npot and GL_APPLE_texture_2D_limited_npot for iOS devices
Check for these extensions by calling glGetString(GL_EXTENSIONS) and searching for the extension you need.
I would also advise keeping your textures to sizes that are multiples of 4 as some hardware stretches textures if not.