I\'m hoping to draw a 2D Grid within a finite space on the X axis using OpengGL 4.0.
I wish to use GLSL using vert/frag shaders etc for rending the light (making the
GLSL is not meant for doing things like that. GLSL is meant for writing shaders which:
GLSL isn't intended to be used for actually making draw calls. Yes, you can use shaders for procedural geometry generation, but you always start off with geometry and draw calls made on the client side.
Yes, you can use just shaders to generate your geometry...
glDrawArrays()
gl_VertexID
in the vertex shader or gl_PrimitiveID
in the geometry shader to procedurally generate your stuff.It can be faster exactly because there are no vertex attributes or input data. Not to mention the space saved and initialization time is next to nothing.
Here's an example of a vertex shader that draws a grid with GL_TRIANGLES
:
#version 150
uniform mat4 modelviewMat;
uniform mat3 normalMat;
uniform mat4 projectionMat;
uniform ivec2 dim;
out vec3 esNorm;
const vec2 triOffset[] = vec2[](
vec2(0,0),
vec2(0,1),
vec2(1,1),
vec2(0,0),
vec2(1,1),
vec2(1,0));
void main()
{
int triVert = gl_VertexID % 6;
int gridIndex = gl_VertexID / 6;
vec2 coord = vec2(gridIndex / dim.x, gridIndex % dim.x);
coord = (coord + triOffset[triVert]) / vec2(dim);
vec4 esVert = modelviewMat * vec4(coord * 2.0 - 1.0, 0.0, 1.0);
gl_Position = projectionMat * esVert;
esNorm = normalMat * vec3(0.0, 0.0, 1.0);
}
I did have some trouble drawing without VBOs bound on my ancient ATI card. This approach works fine on my Nvidia cards with new drivers. Discussed further here: Opengl, DrawArrays without binding VBO, where glDrawArraysInstanced
/gl_InstanceID
is suggested as an alternative.
A further note. I've noticed modulo %
arithmetic can be a little slow in some cases. Using simpler bitwise operations or other tricks may speed things up.