The normals in the GBuffer appear wrong, but the debugger shows they are correct. Why is that?

[left coral]

• custom pass (C#)

rayTracingShader.SetTexture("_CameraNormalTexture", ctx.cameraNormalBuffer);

• shader part

uint2 normalPixel = uint2(pixel.x, resolution.y - 1 - pixel.y);
float4 encodedNormal = _CameraNormalTexture.Load(int4(normalPixel, 0, 0));
RenderTarget[pixel] = encodedNormal;

[lavish mural]

This seems like a very weird way to sample scene normals... 🤔

Mind to explain the reasoning behind that calculation?

[left coral]

reading data at the pixel coordinates, inverting y. The result is sent to rendering as rgb.

If you see what I'm doing wrong, please point it out. It's worth noting that I am doing all this in a .raytrace shader.

[lavish mural]

Oh, I see. I was concerned about the way you sample the texture, but I can see that Load takes pixel coordinates, rather than uvs.

How do you calculate the pixel value?

[left coral]

uint2 pixel = DispatchRaysIndex().xy;

[lavish mural]

And the resolution?

[left coral]

uint2 resolution = DispatchRaysDimensions().xy;

[lavish mural]

I see. In this case I'm not sure. I assume you've been referencing this page?
https://docs.unity3d.com/6000.2/Documentation/ScriptReference/Rendering.RayTracingShader.Dispatch.html

[left coral]

yes

I've been sitting with chat gpt and grok for over 5 hours now trying to figure out how to properly sample the gbuffer from a tracing shader. I've completely lost all understanding at this point.

[lavish mural]

I can only assume that using pixel as an indexer I to the render target might be incorrect. They do it differently in the docs I think.

[left coral]

I hope you're right. I'll try it a bit later, but thank you.

[wind river]

The normals are encoded in the normal buffer using octahedral coordinates.
Displaying the raw output of the buffer gives this "checker" look, but the rendering debugger decodes them for better visualization.

There is an Unpacking function available in the shader library : https://github.com/Unity-Technologies/Graphics/blob/master/Packages/com.unity.render-pipelines.core/ShaderLibrary/Packing.hlsl#L40

[left coral]

AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

after 8 hours

i want to cry....

@wind river you are my hero

[left coral]

@wind river

I'm trying to get the world position of a pixel using a depth map. I'm comparing the result with ray tracing output and getting position discrepancies when rotating the camera. is the depth map guaranteed to be rendered in the same frame?

[wind river]

Hum... Yes, the depth map is rendered at the current frame.
Could it be that it is the RT output that is delayed?

[left coral]

possibly, but I'm not sure how to verify this definitively.

ctx.cmd.DispatchRays(rayTracingShader, "RayGen", (uint)width, (uint)height, 1, ctx.hdCamera.camera);

HDUtils.DrawFullScreen(ctx.cmd, overlayMat, ctx.cameraColorBuffer,
            shaderPassId: overlayMat.FindPass("ShadowMask"));