So I tried the first Idea and I liked the result, it had a few bugs that I need to work out so I'm back to my original, (this time I'll make a backup before I go on these excursions ).

So what I tried was having 16x16x16 blocks, in each block only the top-most cubes were set to "visible". If a block on top is destroyed, the corresponding block underneath is set to visible, if there is a block beside it which isn't visible, that is set to visible. However by creating objects to do this instead of a boolean array, it turned out to be just as intensive memory wise.

[EDIT]
I was unclear above, the FPS was dramatically increased from my current model which I was very pleased. I also rendered all cubes currently visible. I was unable to make a working algorithm to detet hidden-faces. Therefore I was still unable to place as many cubes as I would like.
[/EDIT]

I'll try again with the boolean idea. Perhaps I can find some documentation on an algorithm to detect hidden-faces quickly

I like this triangle reduction idea. How would you go about it?

My Theory:

Have two worlds, the main world has the singular cubes but doesn't render them (for collisions), the second world is a copy, where I merge all of the objects and then draw the singularity(will this reduce faces automatically?)

I think that sounds good

The only problem with this is the memory consumption, on previous games I've had to expand the allowed heap. This hasn't hindered anybody from playing my games (that I'm aware, only around 50 unique devices have tried them)

Perhaps implementing both methods would be beneficial?

I'm thinking of a 3 dimensional array of booleans and integers, and an enumeration for the type (for setting textures on the fly)


Let me know what you gentlemen think,
Donny

My game is coming along quite nicely. Right now I have a plane of cubes rendered (20x20) and performance is great. However, when I add a lot of cubes to the scene, the frame rate slows down and eventually it becomes unplayable.

I had tried to use picking to get objects from the scene that are currently "on top" but this is even slower (<1 fps).

Does anybody have any experience with culling interior objects like this?

I have generated a new test (10x10x10) and when you look at the center the framerate drops very quickly, but if you look at a glancing angle the framerate is fine.

All the objects share compiled-data with a single cube, the only thing that changes is the texture.

Thanks for your help again,

Donny

for (int i = 0; i < texList.length; i++)
      {
         Bitmap temp = BitmapFactory.decodeStream(c.getResources().openRawResource(texList));
         
      preview = Bitmap.createBitmap(temp, 16, 32, 16, 16);
      Texture texture = new Texture(temp,true);
      texture.defaultToMipmapping(true);
      
      if (!TextureManager.getInstance().containsTexture(c.getResources().getResourceEntryName(texList)))
      {
         TextureManager.getInstance().addTexture(c.getResources().getResourceEntryName(texList), texture);
         TextureManager.getInstance().addTexture(c.getResources().getResourceEntryName(texList) + "preview", new Texture(preview,true));
      }
      }

I have 64x64 textures but it always re-sizes to 256x256 and then looks blurry when applied to my object.

Here you go:

Code: [Select]

        SimpleVector dir = this.getDirection(point.x, point.y);
        Object[] res = world.calcMinDistanceAndObject3D(camera.getPosition(), dir, 10000);
        if (res[1] != null) { // something hit
            Object3D obj3D = ((Object3D)res[1]);
            Voxel hitVoxel = data.getVoxel(world.getVoxelId(obj3D.getID()));
            if (hitVoxel != null) {
                voxelPos = hitVoxel.getPosAsInt();
                // find collision point
                SimpleVector colPoint = camera.getPosition();
                dir.scalarMul((Float)res[0]);
                colPoint.add(dir);
                colPoint.sub(obj3D.getOrigin());
                // find side that it hits
                int lastActiveSide = 0; // assume it's zero (no need to check)
                float dist = colPoint.distance(directionVectors[0]);
                for (int i = 1; i < directionVectors.length; i++) {
                    float tempDist = colPoint.distance(directionVectors[i]);
                    if (dist > tempDist) {
                        dist = tempDist;
                        lastActiveSide = i;
                    }
                }
            }
        }
For the direction I use Interact2D.reproject2D3DWS

Please let me know if you have any questions.

Just curious, what are you up to?

Thank you - I figured this out by generating 6 cubes around it and detecting if those are collided.

I'm making a game called CubeCraft, yes, it's exactly what you think it is


Thanks,
Donny

I can't believe I let myself get this frustrated over something so incredibly simple, my cube's were spaced too close together and they were fighting to show through each other

I'm rendering 100 cubes with a simple texture applied to them. The way the scene works is you can only see the top and you can rotate the camera, however. Whenever I rotate there are lines and the texture seems to expand and shrink rapidly on each cube. Why does this happen? How can I stop it? I'm stumped and frustrated.

Hello forum,

This is my first post so hello and I hope to be a contributing member.

Unfortunately my first post has to be in the Support forum because, well I need support.
I'm using the software renderer. I'm only interested in including jpct.jar. It all appears to be working fine,
however; when I load a model, you can see through some of the faces. The same behavior is apparent when
using the plane primitive.  When rotated you can only see one face.

Just wondering if this is a problem with the way I have it configured, or if it's just simply the way it is. If it does have to be
one way, what are some tips for blender to make sure everything looks solid?

Thanks,
Donny.