Jump to content

  • Log In with Google      Sign In   
  • Create Account

Problem with my sky dome - big hole in the top!


Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
5 replies to this topic

#1 Lil_Lloyd   Members   -  Reputation: 287

Like
0Likes
Like

Posted 16 January 2013 - 07:31 PM

I have an issue with my sky dome in a terrain project. I'm generating the vertices and indices for a hemisphere in a function, but for some reason when testing using texture triangles and open gl's line mode, there is a huge gaping hole in the top of the sky dome. I have tried changing the draw distance to see if it's a clipping problem but no luck!

 

Here is my code, sorry about the formatting as pasting it into the code reader squashed some lines together

bool	SkyDome::Init(const std::string& textureName,const std::string& shader){

	//TODO: Texture stuff
	//TODO shader stuff

	//init the objectbuffer and vertex info!

	int numVertices = (m_latSlices+1) * (m_longSlices+1); 

	vec3* vertexData = new vec3[numVertices];

	for(int latSlice = 0; latSlice <= m_latSlices; latSlice++){

          for(int longSlice = 0; longSlice <= m_longSlices; longSlice++){
			
            vertexData[latSlice * m_longSlices + longSlice] = 
                                vec3(m_radius * cos( (float)(latSlice)/(m_latSlices-1) * PI/2.0 ) * cos( 2.0 * (float)longSlice/(m_longSlices-1) * PI ),
				m_radius * sin( (float)(latSlice)/(m_latSlices-1) * PI/2.0 ),
			        m_radius * cos( (float)(latSlice)/(m_latSlices-1) * PI/2.0 ) * sin( 2.0 * (float)longSlice/(m_longSlices-1) * PI ) );
		 }
     
   }

   //generate indices
	int index = 0;
	unsigned int* indices = new unsigned int[numVertices * 6];
	//generate indices
	for(int row = 0; row <= m_latSlices; row++){
		for(int column = 0; column <= m_longSlices; column++){
			int start   = row * m_longSlices + column;
			indices[index++] = start;
			indices[index++] = start + 1;
			indices[index++] = start + m_longSlices;

			indices[index++] = start + 1;
			indices[index++] = start + 1 + m_longSlices ;
			indices[index++] = start + m_longSlices ;
		}
	}


	m_objectData = new ObjectData(1,numVertices);
	m_objectData->Init(indices,numVertices,vertexData,3);

    //delete [] vertexData;
	delete [] indices;

	return true;

 

 



Sponsor:

#2 /Jeff/   Members   -  Reputation: 648

Like
1Likes
Like

Posted 17 January 2013 - 06:54 AM

It would probably be easier to identify the issue if you could post a screenshot of the problem.



#3 Milcho   Crossbones+   -  Reputation: 1175

Like
1Likes
Like

Posted 17 January 2013 - 09:18 AM

Your spherical equation works fine - with those parameters it should generate a half-shere oriented along the +y axis - so I'm assuming your 'up' direction is +y.

 

I don't think this is exactly the problem, but when calculating the vertex positions you have:

latSlice in the range of [0, m_latSlices] and longSlice in the range of [0, m_longSlices] (I'm starting to bold these since there's no inline <code> tag)

 

- meaning you include both boundary values, but in your calculations you have:

(latSlice)/(m_latSlices-1) and longSlice/(m_longSlices-1)

- from that it looks like you're overshooting the values - you want both of them to range from 0..1 - and this will make them go to 0.. slightly above 1 

The calculation of angle should probably be (latSlice)/(m_latSlices) and longSlice/(m_longSlices) - without the -1

 

 

Also, in your loop to assign indices.. there's something odd (meaning, I can't put my finger on it yet) going on if you take a look at your very last run through that loop:

for(int row = 0; row <= m_latSlices; row++)
{
  for(int column = 0; column <= m_longSlices; column++)
  {

Here row and column also range to m_latSlices and m_longSlices inclusive. 

Then in your calcuation:

int start = row * m_longSlices + column;

the maximum value of start (which occurs at the last run through that loop) will therefore be

m_latSlices*m_longSlices + m_longSlices or (m_latSlices + 1)*m_longSlices

 

Then one of your indices becomes

indices[index++] = start + 1 + m_longSlices - which then, for your maximum value of start, actually becomes:

(m_latSlices + 1)*m_longSlices + 1 + m_longSlices or m_longSlices*(m_latSlices + 2) + 1

 

which seems like it's incorrect... From what I see you're making triangles out of 1 'column' ahead - but that usually means that you should limit your loop so that it doesn't do that for the last column, since the last column has no 'one column ahead'. 

I need to take a closer look at this later, when I can run your code (my brain can only hold so much of this info in my head at once), and try with really low values of m_longSlices and m_latSlices - which should probably give a good indication of what's messing up.



#4 Milcho   Crossbones+   -  Reputation: 1175

Like
1Likes
Like

Posted 17 January 2013 - 02:37 PM

I was running just a numeric version of this -
What I said above seems to be a problem - your division of (latSlice)/(m_latSlices-1) and longSlice/(m_longSlices-1) - should not have that -1, and your indexing does indeed set an index that is out of bounds - only one index it sets is out of bounds, but it's still not accurate - and I don't have a quick solution to that right now.

There's also another problem: if you check how you calculate which vertex's position you're calculating, it looks like this:
latSlice * m_longSlices + longSlice - however, this calcuation actually gives inaccurate results. For 4 latSlices and 4longSlices, you make an array of 25 vertices, but you end up doing this:

 

vertexData[0] = 1 0 0
vertexData[1] = 0 0 1
vertexData[2] = -1 0 0
vertexData[3] = 0 0 -1
vertexData[4] = 1 0 0
vertexData[4] = 0.92388 0.382683 0
vertexData[5] = 0 0.382683 0.92388
vertexData[6] = -0.92388 0.382683 0
vertexData[7] = 0 0.382683 -0.92388
vertexData[8] = 0.92388 0.382683 -0
vertexData[8] = 0.707107 0.707107 0
vertexData[9] = 0 0.707107 0.707107
vertexData[10] = -0.707107 0.707107 0
vertexData[11] = 0 0.707107 -0.707107
vertexData[12] = 0.707107 0.707107 0
vertexData[12] = 0.382683 0.92388 0
vertexData[13] = 0 0.92388 0.382683
vertexData[14] = -0.382683 0.92388 0
vertexData[15] = 0 0.92388 -0.382683
vertexData[16] = 0.382683 0.92388 0
vertexData[16] = 0 1 0
vertexData[17] = 0 1 0
vertexData[18] = 0 1 0
vertexData[19] = 0 1 0
vertexData[20] = 0 1 0

 

You can see that some of the indices get repeated (because latSlice * m_longSlices + longSlice has the same value for (latSlice = n, longSlice = m_longSlices) and (latSlice = n + 1, longSlice = 0) - and because of that, you don't set all your indices - in the above, there are 4 repetitions and you should have also set vertexData[21...24] as well, but the program did not. Making an incremental counter to which vertex you're setting fixes that, but I still don't know if this will fix your problems.

 

I'm pretty sure that the hole you're getting on the top of your hemi-sphere is due to the fact that your last m_longSlices number of vertices don't get filled at all.


By incremental counter, I mean setting vertexData like this:

int index = 0;
for(int latSlice = 0; latSlice <= m_latSlices; latSlice++){
  for(int longSlice = 0; longSlice <= m_longSlices; longSlice++){
    vertexData[index++] =...

I think you need to carefully re-examine the boundaries of your loops, and re-write them while making sure they don't overlap and the way you calculate your angles varies between the 0..pi/2 and 0..2pi values correctly (as I posted in my previous post).
Edit: fixing weird gap that occured

Edit2: added additional comment clarifying cause of your 'hole'



#5 Lil_Lloyd   Members   -  Reputation: 287

Like
0Likes
Like

Posted 17 January 2013 - 06:11 PM

Thanks gentlemen. However, now I have taken the sensible route - using a sphere .obj mesh and using a cube map to texture it. I don't know why I didn't think of doing it in the first place! 



#6 swiftcoder   Senior Moderators   -  Reputation: 9987

Like
0Likes
Like

Posted 17 January 2013 - 08:40 PM

Don't feel bad - the first time anyone sits down to write a sphere tessellator, they usually mess up the indices like this (I certainly did).

 

If you end up revisiting this code, it's worth sketching the indices up on a piece of paper, and then reasoning very carefully about where your indices should/shouldn't wrap around.


Tristam MacDonald - Software Engineer @Amazon - [swiftcoding]





Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.



PARTNERS