What is the best way to display background images when using OpenGL (GLSurfaceView)?

The way I did it was by creating a texture with the image... and then rendering it... but this makes the performance extremely slow depending on the image size.

What other options do I have?

Thanks.

You can use GL11 Ext function glDrawTexfOES(x, y, z, width, height).

But I think rendering backgroud quad shouldn't make a big fps drop. Just try to make your texture size near to screen resolution as much as possible.

hardcoras wrote:You can use GL11 Ext function glDrawTexfOES(x, y, z, width, height).

But I think rendering backgroud quad shouldn't make a big fps drop. Just try to make your texture size near to screen resolution as much as possible.

Thanks. I'm still having trouble. But there seem to be some good options for android outside opengl so I'll try that for now.

Having the same problem. I'm rendering a textured quad as background which is placed at 0.0f, 0.0f, -8.0f. The quad is covering exactly the whole screen and I get 15fps. Same quad with same texture rendered at 0.0f, 0.0f, -16.0f gives 35fps. I don't get it, what's the problem? One quad rendered in front of the camera is slower, than rendering 50 quads in the distance.

Here's the code executed at initialization:

Code: Select all

   float x = 5.5f;
   float y = 3.3f;
   
   float   vertex[] = new float[] {
         -x, y, 0,
          x, y, 0,
         -x,-y, 0,
          x,-y, 0
   };

   float   texel[] = new float[] {
         0, 0,
         1, 0,
         0, 1,
         1, 1
   };

      ByteBuffer vb = ByteBuffer.allocateDirect(vertex.length * 4);
      vb.order(ByteOrder.nativeOrder());
      vertexBuffer = vb.asFloatBuffer();
      vertexBuffer.put(vertex);
      vertexBuffer.position(0);

      ByteBuffer tb = ByteBuffer.allocateDirect(texel.length * 4);
      tb.order(ByteOrder.nativeOrder());
      texelBuffer = tb.asFloatBuffer();
      texelBuffer.put(texel);
      texelBuffer.position(0);

And this one renders it on screen:

Code: Select all

      gl.glMatrixMode(GL10.GL_MODELVIEW);
      gl.glLoadIdentity();
      gl.glTranslatef(0.0f, 0.0f, -18.0f);
      gl.glDisable(GL10.GL_BLEND);
      gl.glEnable(GL10.GL_TEXTURE_2D);
      gl.glBindTexture(GL10.GL_TEXTURE_2D, globals.scnTextures[1]);
      gl.glVertexPointer(3, GL10.GL_FLOAT, 0, vertexBuffer);
      gl.glEnableClientState(GL10.GL_VERTEX_ARRAY);
      gl.glTexCoordPointer(2, GL10.GL_FLOAT, 0, texelBuffer);
      gl.glEnableClientState(GL10.GL_TEXTURE_COORD_ARRAY);
      gl.glDrawArrays(GL10.GL_TRIANGLE_STRIP, 0, vertex.length / 3);

EDIT: 40fps same background quad without texture.

This is how i generate textures:

Code: Select all

      gl.glBindTexture(GL10.GL_TEXTURE_2D, tx);
      gl.glTexParameterf(GL10.GL_TEXTURE_2D, GL10.GL_TEXTURE_MIN_FILTER, GL10.GL_NEAREST);
      gl.glTexParameterf(GL10.GL_TEXTURE_2D, GL10.GL_TEXTURE_MAG_FILTER, GL10.GL_LINEAR);
      GLUtils.texImage2D(GL10.GL_TEXTURE_2D, 0, bmp, 0);


bmp is a 8bit image.

Patton wrote:Having the same problem. I'm rendering a textured quad as background which is placed at 0.0f, 0.0f, -8.0f. The quad is covering exactly the whole screen and I get 15fps. Same quad with same texture rendered at 0.0f, 0.0f, -16.0f gives 35fps. I don't get it, what's the problem? One quad rendered in front of the camera is slower, than rendering 50 quads in the distance.

Interesting news. And what is your projection matrix?
I'm a n00b to opengl, but i don't see how z value (depth) could reflect on a fps in orto2d projection matrix.

Maybe opengl experts could explain it?

I've tried first with GLU.gluPerspective - 15fps... then I've used gluOrtho2D and 2d coordinates for that background - still 15fps It's pretty hilarious. So any OpenGL pro's reading this thread/forum? Please help...

If I draw only one quarter of the screen with gluOrtho2D, I get about 25-30fps.

Ehm, thanks to my friend, I've tested the same program on a real device HTC Hero.

60fps

Damn emulator!

Interesting. I didn't think the emulator would be so off...

Hi,

I was just wondering what kind of hardware you were using. Looks like any

@ameetnsharma
An emulator does not have any hardware acceleration. Plus an emulator only runs on a single thread (at least on a linux host). For a complex scene rendering it is almost useless.
From my experience textures rendering is very slow on emulator. Mostly if you are using bi-linear mag filter.

Thanks Patton and hardcoras.

Is there any cheap device I can use to test my programs? Is there a non-phone android device (like ipod touch instead of iphone for apple)

You can grab an used G1 on ebay. They are quite cheap these days.

There is a gaming handheld device called ODroid (http://www.hardkernel.com/). But i think in the near future there will be a lot of different consumer electronic devices, not only phones, tablets or netbooks.

The future is ours

Another question now about blending backgrounds.

I have one BMP image which has black background. I use gl.glBlendFunc(GL10.GL_SRC_ALPHA, GL10.GL_ONE) to make that black background transparent. How to achieve that red portions of the image are transparent, and not the black ones?

This looks like a ultra noob question, sorry about that

Why don't you use textures with alpha channel? Many devs use PNGs (including me).

How to load png textures into opengl?

For bmp's I used something like this:

Code: Select all

   private static int initBmpTexture(GL10 gl, Bitmap bmp) {
      
      ByteBuffer bb = ByteBuffer.allocateDirect(bmp.getHeight() * bmp.getWidth() * 4);
      bb.order(ByteOrder.nativeOrder());
      IntBuffer ib = bb.asIntBuffer();
      
      for (int y=0; y<bmp.getHeight(); y++)
         for (int x=0; x<bmp.getWidth(); x++) {
            ib.put(bmp.getPixel(x, y));
         }
      ib.position(0);
      bb.position(0);
      ...
      ...

You can do smt like



Exceptions handling, texture initialization and similar stuff is skipped.

Bitmap.Config.RGB_565 - for smaller textures (16bits) and native color model.