How to manipulate actual vertex position in fragment shader

Llunatix · Oct 28, 2023

No, I don't misunderstand how this works and how a vertex is interpolated from vertex to fragment shader but still a good explanation.
But I finally got what you actually trying to tell me and I think we are talking past each other so let me clearify some things:

I know that I can't alter the value of a varying sent (and interpolated) from vertex to fragment shader.
I also know that you can't modify the position of a fragment.

But I can interpret a specific fragment how I like, for example, manipulate it's color, lighting, shadow etc.
And that's exactly what I want to achieve, I want to tell the code being executed after my custom shader what position it should use for the shadow calculations.

I don't want to modify the vertex varying, I'm just searching a way to tell the shader what my raymarching algrithm calculated.

So what I basically want is a POSITION parameter which can be read and written to, like this:

// Code before custom shader function
vec3 vertex = vertex_interp;
vec3 view = -normalize(vertex);

#if POSITION_USED
   vec3 position = vertex_interp;
#endif

// custom shader function
vec3 displacedPos = raymarch(...);
POSITION = displacedPos;

// Code after custom shader:
#if POSITION_USED
   view = -normalize(position);
#endif

// shadow calculation
			float shadow = 1.0;

			if (directional_lights.data[i].shadow_opacity > 0.001) {
				float depth_z = -position.z; // <-- use position here instead of vertex
				vec3 light_dir = directional_lights.data[i].direction;
                                // this also looks fishy as the fragment shader allows to write to NORMAL but uses normal_interp here
				vec3 base_normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp))));

I hope I made it crystal clear now what I want to achieve and what my problem actually is.
I'm not good at explaining things in an academic way, I'm better at practially showing what I want to do via code and images, so that may be the reason why my intentions where not good to understand.

P.S.: VERTEX is still being directly replaced with vertex so what I'm actually writing to after transpiling is the vec3 vertex variable, not the vertex_interp varying coming from the vertex stage.

xyz · Oct 28, 2023

lunatix I want to tell the code being executed after my custom shader what position it should use for the shadow calculations.

You can't tell it that. Shadow mapping doesn't work like that. It doesn't use pixels rendered in your fragment shader. At all. Any shadows that your voxels cast on other voxels in your volume are responsibility of your raymarcher. And any shadows that they are casting onto other objects can only be properly calculated if a shadow map renderer runs the raymarcher again (from the light's viewpoint) when rendering the shadow depth texture, and write the raymarch results into it. This depth is what needs to be displaced, not pixel's world positions in your fragment shader. Those can't affect the shadow map. Engine could not use this information for really anything further down the pipeline. That's why it doesn't exist as an output from the fragment function.

Llunatix · Oct 28, 2023

I think we still not on the same boat but that's okay, I won't discuss this any further because I already solved it on my own with my custom fork.
This fork just makes it possible to manipulate the "vertex" and geometric normal in the custom shader code and makes sure that, if used, afterwards some calculations are repeated to align multiview stuff view vector.

And just to prove it, here is a screenshot and code snippet on how to use it:

void fragment() {
	RaymarchHit hit;
	
        mat4 LocalToWorld = VIEW_MATRIX * MODEL_MATRIX;
	mat4 WorldToLocal = inverse(MODEL_MATRIX);
	
	vec3 cameraPosLS = TransformWorldToObject(WorldToLocal, CAMERA_POSITION_WORLD);
	vec3 rayOrigin = cameraPosLS * VoxelScale + VoxelBoxSize * 0.5;
	vec3 vertexPosWS = (INV_VIEW_MATRIX * vec4(VERTEX, 1.0)).xyz;
	
        vec3 cameraDirWS = normalize(vertexPosWS - CAMERA_POSITION_WORLD);
	vec3 rayDir = TransformWorldToObjectNormal(WorldToLocal, cameraDirWS);

	if (!RaymarchLoop(rayOrigin, rayDir, VoxelBoxSize, DataTexture, hit)) {
		discard;
	}
	
	float invVoxelScale = 1.0 / VoxelScale;
	vec3 localPos = (hit.position - VoxelBoxSize * 0.5) * invVoxelScale;
	vec3 worldPos = (LocalToWorld * vec4(localPos, 1.0)).xyz;
	vec4 projPos = PROJECTION_MATRIX * vec4(worldPos, 1.0);
	
        ALBEDO = hit.color;
	NORMAL = ((mat3(LocalToWorld)) * hit.normal);
	DEPTH = (projPos.z / projPos.w);
	VERTEX_OUT = worldPos; // this one is new
	NORMAL_GEOM = normalize(hit.normal); // this one is new
}

So this shader is now able to fake 3D voxel geometry by just rendering an inverse box and raymarching against a 3D texture.
It outputs a custom depth to gl_FragDepth and manipulates parameters in the fragment stage to make shadow mapping work (because it uses z coordinate of the vertex parameter).

And if you still don't believe me, here is a PR in my fork which shows the modifications I had to do: https://github.com/Lunatix89/godot/pull/1/files

Haystack · Oct 28, 2023

Note that while you can't modify the (x,y) position of a fragment, you can modify its depth value.

out float DEPTH

Custom depth value (0..1). If DEPTH is being written to in any shader branch, then you are responsible for setting the DEPTH for all other branches. Otherwise, the graphics API will leave them uninitialized.

https://docs.godotengine.org/en/stable/tutorials/shaders/shader_reference/spatial_shader.html#doc-spatial-shader
(Search for 'depth')

xyz · Oct 28, 2023

lunatix So this shader is now able to fake 3D voxel geometry by just rendering an inverse box and raymarching against a 3D texture.
It outputs a custom depth to gl_FragDepth and manipulates parameters in the fragment stage to make shadow mapping work (because it uses z coordinate of the vertex parameter).

I doubt you can get correct looking shadows without rendering the volume into shadow map. So that might be the intervention worth a fork.

I'd like to see your solution in motion with rotating volume, changing directional light direction or a couple of moving omni lights, and a 3D texture that has some see-through holes.

Llunatix · Oct 29, 2023

I'll take a closer look at this in the next days and will post an update.
I had the same shader in Unity and it worked pretty well, however, I always used the deferred rendering path and forward only for transparents.

Megalomaniak · Oct 29, 2023

have you looked into parallax occlusion mapping?

xyz · Oct 29, 2023

lunatix I had the same shader in Unity and it worked pretty well, however, I always used the deferred rendering path and forward only for transparents.

Rendering path wouldn't make much difference. For volume's shadow to be cast correctly, the volume needs to be rendered into shadow map by a raymarching shader.

If you can tell Unity's pipeline to use your custom raymarching shader when rendering the object into shadow map, then it may be possible to have correct shadows. Afaik, in Godot you can't use custom shadow map shaders.

So no amount of post-rasterization manipulation of pixel world positions could do this. There's simply not enough information about the volume there. You might get almost ok looking shadow shapes for voxel distributions that are close to completely filled box. This is precisely what you've shown in your last screenshot. But even that shadow doesn't look correct. Compare it to just regular boxes in same positions.

Here's a thought experiment. You can even try it with your Unity shader to see if it indeed does things correctly, which I'm doubtful of.

Take a 3D texture that is full block with a hole going all the way through from one side to another. Like this:

Position the camera so it doesn't see the hole. It will render something like this:

Now shine a directional light sideways in the direction of the hole. The light "sees" something like this:

The proper shadow would then look something like this:

As you can see, the pixels rendered from the camera view can't store any information about the hole. That information can be only obtained by looking from the light's point of view. And that's precisely what the shadow map is - the look at objects from light's perspective. The renderer needs that information to properly render shadows, and you cannot supply it only via pixels seen by the camera.

Llunatix · Oct 29, 2023

I've reproduced your experiment and it looks like everything is working fine because the shadow map actually uses my custom shader for rendering.

And a more complex geometry.

The only thing which is tricky is the bias because self shadowing doesn't look so nice with voxel geometry.

xyz · Oct 29, 2023

lunatix Great! Brownie points for Godot. And you as well

I'm just curious why did you need that fork really. It's enough for raymarch shader to write volumetric pixel depth into z buffer by simply assigning to built-in DEPTH. If such shader is invoked for both; camera and shadow map, then no other intervention is needed.

Llunatix · Oct 30, 2023

Thanks a lot

The fork was needed because I could not tell the code executed after my raymarching shader to use the raymarched position instead of the one passed from vertex to fragment shader.

There are several places which use the fragment position to calculate things like fog, lighting, shadows and so on and all those functions would calculate with the interpolated vertex position of the cube I use as a volume to raymarch.

However, that would end up in wrong lighting and shadowing because the z coordinate from the vertex is used to find out in which shadow cascade the fragment lies within and it's also used to get the PSSM coords.

So that's why it looks wrong in this screenshot where I could not provide the raymarched position:

You can see clearly that the actual box was used to calculate which fragment is in shadow (wrong angle but I think you'll get the point).

In Unity this was easy to do because it would let custom shaders use a function to calculate the fragments color, lighting, shadow, metallic and so on because there was a function provided for it which one can pass position, albedo, normal etc. but Godot is not that flexible here.

However, I don't know how much value this would bring to the general purpose user so I'm not sure whether it would be usefull to open a PR to the original Godot engine.
But I could imagine that it might also be useful for rendering SDF volumes or raytracing in the fragment shader, like screen space shadows or screen space reflections.

xyz · Oct 30, 2023

lunatix Oh, I finally grok what you were talking about. Your terminology was confusing. You want to be able to alter fragment's world position used by Godot further down the pipeline.

This is indeed useful for any kind of volumetric rendering. In fact it may be essential to make volumetrics play nice with the rest of the scene. You should definitely open a PR or at least a feature request to see how the devs would react and get some discussion going.

VERTEX_OUT is not a good name though. It should be something like FRAGMENT_WORLD

Llunatix · Oct 30, 2023

Yes, fragment world position is a good name for it.. in unity they always just use the term "position" in fragment shaders, like "positionWS" (world space), positionSS (screen space) or positionOS (object space) etc.

And you're right, VERTEX_OUT is bad and I knew that but I did not had a better term for it yet

Llunatix · Oct 31, 2023

Just found out after opening the PR that there is already another user who wanted to achieve the same: https://github.com/godotengine/godot/pull/65307

This user also wrote a really nice addon which uses the DDA algorithm but the plugin won't work until the PR has been merged into the main repo.
I think I like this users solution more because there is actually no need to supply the fragments world position since it can be reconstructed from linear depth and I really like the conclusion that you can't modify the position of a fragment in 3D space to anything you want to, you can only modify it along the ray i.e. farther or nearer to the camera.

However, it looks like the Godot core devs did not yet came to a solution and the team is currently focusing on the 4.2 release so I will maybe just continue to use a fork and stick to this users solution as it is way more optimized than my version was.