It is also strange that you say you think it is the ray origins and diretions that are not calculated correctly, (although you could very possibly be right, I wouldn't doubt it) But it is strange, because if I do a basic ray box intersection using the same ray directions and origins there is no warping at all. The warping exists entirely within the DDA algorithm I have. Here it is:
fn Sphere_DDA_March(
start_pos: vec3<f32>, // Starting position of the ray in world space
end_pos: vec3<f32>, // Ending position of the ray in world space
voxel_size: vec3<f32>, // Size of each voxel in the voxel grid
bounds: vec3<i32>, // Dimensions of the voxel grid (width, height, depth)
) -> vec4<f32> {
var direction: vec3<f32> = normalize(vec3<f32>(end_pos - start_pos));
var step: vec3<i32> = vec3<i32>(
select(-1, 1, direction.x > 0.0),
select(-1, 1, direction.y > 0.0),
select(-1, 1, direction.z > 0.0)
);
var tMax: vec3<f32> = vec3<f32>(
((select(floor(start_pos.x / voxel_size.x) + 1.0, floor(start_pos.x / voxel_size.x), direction.x < 0.0) * voxel_size.x) - start_pos.x) / direction.x,
((select(floor(start_pos.y / voxel_size.y) + 1.0, floor(start_pos.y / voxel_size.y), direction.y < 0.0) * voxel_size.y) - start_pos.y) / direction.y,
((select(floor(start_pos.z / voxel_size.z) + 1.0, floor(start_pos.z / voxel_size.z), direction.z < 0.0) * voxel_size.z) - start_pos.z) / direction.z
);
var tDelta: vec3<f32> = vec3<f32>(
abs(voxel_size.x / direction.x),
abs(voxel_size.y / direction.y),
abs(voxel_size.z / direction.z)
);
let center: vec3<f32> = vec3<f32>(15.0);
var ray_world_coord = vec3<f32>(vec3<f32>(current_voxel) * voxel_size);
var ray_dist: f32 = f32(sqrt(((ray_world_coord.x - camera.position.x) * (ray_world_coord.x - camera.position.x)) + ((ray_world_coord.y - camera.position.y) * (ray_world_coord.y - camera.position.y)) + ((ray_world_coord.z - camera.position.z) * (ray_world_coord.z - camera.position.z))));
while(ray_dist < 256.0){
ray_world_coord = vec3<f32>(vec3<f32>(current_voxel) * voxel_size);
var world_cord = vec3<f32>(vec3<f32>(current_voxel) * voxel_size);
ray_dist = f32(sqrt(((ray_world_coord.x - camera.position.x) * (ray_world_coord.x - camera.position.x)) + ((ray_world_coord.y - camera.position.y) * (ray_world_coord.y - camera.position.y)) + ((ray_world_coord.z - camera.position.z) * (ray_world_coord.z - camera.position.z))));
var dist: f32 = f32(sqrt(((world_cord.x - center.x) * (world_cord.x - center.x)) + ((world_cord.y - center.y) * (world_cord.y - center.y)) + ((world_cord.z - center.z) * (world_cord.z - center.z))));
if (dist < 60.0){
let start_light: vec3<f32> = vec3<f32>(world_cord.x, world_cord.y + -0.5, world_cord.z);
let end_light: vec3<f32> = vec3<f32>(world_cord.x, world_cord.y + 10.0, world_cord.z);
let output = DDA_Light(start_light, end_light, voxel_size, bounds);
if (output == 0) {
return vec4<f32>(1.0, ray_dist / 100.0, 0.0, 1.0);
} else {
return vec4<f32>(0.75, ray_dist / 100.0, 0.0, 1.0);
}
}
if (tMax.x < tMax.y && tMax.x < tMax.z) {
current_voxel.x += step.x;
tMax.x += tDelta.x;
} else if (tMax.y < tMax.z) {
current_voxel.y += step.y;
tMax.y += tDelta.y;
} else {
current_voxel.z += step.z;
tMax.z += tDelta.z;
}
}
return vec4<f32>(0.0);
}
Firstly, positive y should be up, anything else is heresy, and what is a projection matrix? I see everybody using one, but I don't use one myself. How do they work?
Hey this is awesome! I've spend the last 3 or so weeks changing how my engine uses buffers, and hilariously, I changed from fragment rendering to compute rendering this last week actually :) I'm glad to see your still making progress!
All of the above need to be added to the rules. At least the engine rules and copyright. I am not a participator I'm just making a suggestions :) Also I like how blasphemy is against the rules, its a nice touch.
Yeah I'm currently planning on open sourcing my engine. what can I say, I like free stuff, I assume everybody else does as well. royalties is a pretty fair way of doing it, personally I prefer paying for a onetime license (if just for the current version).
PS: Thats how I plan to get GI, I'll just have all the rays check if they hit a voxel and if so add it to a buffer of hit voxels, then just have yet another compute shader shade all hit voxels, the result of which will be per-voxel GI. (not complete GI just area wide not full world GI)
Indeed, If I run into any trouble I will ask :D I think I'll probably do the voxel raytracing first, and just have the meshing later, I'll have the voxel raytracer write to the Depth-Buffer and then just have the rasterizer compare depthbuffer distance and check if it should render that particular pixel or not. But I'll be doing all raytracing in a compute shader which will fill a texture and have that texture be read by the fragment shader later.
