Glu perspective 3d rendering issue

Newbie & Debugging Questions
#1

I am fairly new to using these , I usually either do it hands on or work in 2d .Currently the issue appears to arrise with gl_ModelViewProjectionMatrix * position , without gl_Model… it renders a triangle on the screen however with it it doesnt render.
Here is the current initialisation.


try {

			Display.setDisplayMode(new DisplayMode(800, 600));
			Display.sync(60);
			Display.create();
		} catch (LWJGLException e) {
			// TODO Auto-generated catch block
			e.printStackTrace();
		}
		glMatrixMode(GL_PROJECTION);
		glLoadIdentity();
		gluPerspective(45.0f, 800.0f / 600.0f, 1.0f, 500.0f);
		glMatrixMode(GL_MODELVIEW);
		glLoadIdentity();

and my attempt at the shader


attribute vec3 vertex;
attribute vec2 texturecoordinate;
uniform vec3 rotation;
uniform vec3 rotation_centre;
void main(){
	gl_Position = gl_ModelViewProjectionMatrix * vec4(vertex,1);
}

I am currently using a VBO , I am not too sure if there is anything extra that I will need to add to that for it to work , if so or if the code for that is needed just ask.

#2

try …

vec4 mvv = gl_ModelViewMatrix * vec4(vertex, 1.0);
gl_Position = gl_ProjectionMatrix * mvv;

this should render the same as [icode]gl_Position = vec4(vertex,1)[/icode] does since both matrices are identity. … oh wait, you got a [icode]glViewport(…)[/icode] somewhere ?

edit

using the modelview and projection matrix separately is nice cos’ you get the linear depth at [icode]-mvv.z[/icode] if you need it.

#3

yer the gl viewport is what I added just then after doing that.
This works


vec4 mvv = gl_ModelViewMatrix * vec4(vertex, 1.0);
gl_Position =  mvv;

this doesnt.


vec4 mvv = gl_ModelViewMatrix * vec4(vertex, 1.0);
gl_Position = gl_ProjectionMatrix * mvv;
#4

meh, it should :clue:

#5

I think the projection matrix is just 0 meaning that nothing is occuring to it.
Could you give an example of how I would setup a manual projectionmatrix?

#6

sure … if you do not trust gluPerspective …

i use rowmajor matrices, unlike opengl default.
the matrix class i use is basically just a double[4][4], double[row][column].

this is how i setup orthogonal projections :

public static void setOrtho(double left, double right,
                            double top, double bottom,
                            double near, double far,
                            Matrix m)
{
  m.set(2.0 / (right - left), 0, 0, (left + right) / (right - left),  // row 0
        0, 2.0 / (top - bottom), 0, (top + bottom) / (top - bottom),  // row 1
        0, 0, -2.0 / (far - near), (far + near) / (far - near),       // row 2
        0, 0, 0, 1);                                                  // row 3
}

i feed the same values i use with glViewport but in this order : 0.0, width, 0.0, height, -1.0 (near), 1.0 (far).

for perspective projections i use :

// fovr = y-fov in radians, vertical.
// aspect = viewport (float)width/height
// znear should be > 0.0
public static void setPerspective(double fovr, double aspect, double near, double far, Matrix m)
{
  final double y = near*StrictMath.tan(fovr*0.5);
  final double x = aspect*y;

  setFrustum(-x,x,y,-y,near,far,m);
}

setting the frustum :

public static void setFrustum(double left, double right,
                              double top, double bottom,
                              double near,double far,
                              Matrix m)
{
  m.set(( 2.0*near )/( right - left ),0,( left + right )/( right - left ),0,    // row 0
          0,( 2.0*near )/( top - bottom ),( top + bottom )/( top - bottom ),0,  // row 1
          0,0,-( near + far )/( far - near ),( -2*near*far )/( far - near ),    // row 2
          0,0,-1,0);                                                            // row 3
}

now, to move this to opengl i transpose the matrices and copy them … in one go :

public static void copy(Matrix m, FloatBuffer buffer)
{
  buffer.put((float)m.matrix[0][0]);
  buffer.put((float)m.matrix[1][0]);
  buffer.put((float)m.matrix[2][0]);
  buffer.put((float)m.matrix[3][0]);
  buffer.put((float)m.matrix[0][1]);
  buffer.put((float)m.matrix[1][1]);
  buffer.put((float)m.matrix[2][1]);
  buffer.put((float)m.matrix[3][1]);
  buffer.put((float)m.matrix[0][2]);
  buffer.put((float)m.matrix[1][2]);
  buffer.put((float)m.matrix[2][2]);
  buffer.put((float)m.matrix[3][2]);
  buffer.put((float)m.matrix[0][3]);
  buffer.put((float)m.matrix[1][3]);
  buffer.put((float)m.matrix[2][3]);
  buffer.put((float)m.matrix[3][3]);
  buffer.rewind();
}

that buffer was a FloatBuffer of size 16 (one could use a DoubleBuffer too but i do not) and is going to opengl by :

GL11.glMatrixMode(GL11.GL_PROJECTION);
GL11.glLoadMatrix(projection);
//GL11.glMatrixMode(GL11.GL_MODELVIEW);
//GL11.glLoadMatrix(modelview)

or if you use direct state :

EXTDirectStateAccess.glMatrixLoadEXT(GL11.GL_PROJECTION,projection);
//EXTDirectStateAccess.glMatrixLoadEXT(GL11.GL_MODELVIEW,modelview);

edit the modern way is write all values into a UBO-mapped-buffer, not use loadMatrix() and have all shaders source matrices from that UBO’s, this on the other hand should work with fixed-pipeline opengl 1.1.

next would be … look-at matrix into modelview :slight_smile:

o/

#7

Absolutely excellent , have a medal . However I managed to fix it , it was simply an ordering problem , I put the viewport after all the identity stuff so it wasn’t doing anything.

#8

cool