Back to General and Gameplay Programming

Cascaded Shadow Map Issue

Sean Forbes · 2013-02-23T09:01:01

We have been trying to get cascaded shadow maps working from MJP's sample but I can't seem to get it working right. All I get is a diagonal area of shadow across the ground regardless of which direction I look or where the light is. From my debugging of the shader, it seems like the problem is that the shader that builds the occlusion map is sampling the cascade map from the wrong point. I've gone over the code several times and can't see what would be causing the issue. This is the part of the shader that calculates the texture coordinate which seems to be wrong: // Reconstruct view-space position from the depth buffer float pixelDepth = DepthMap.Sample(DepthMapSampler, input.TexCoord).r; float4 position = float4(pixelDepth * input.FrustumCornerVS, 1.0f); // Determine the depth of the pixel with respect to the light float4x4 inverseLVP = mul(InverseView, lightViewProjection); float4 positionLight = mul(position, inverseLVP); float lightDepth = (positionLight.z / positionLight.w); // Transform from light space to shadow map texture space. float2 shadowTexCoord = 0.5 * positionLight.xy / positionLight.w + float2(0.5f, 0.5f); shadowTexCoord.x = shadowTexCoord.x / SPLITS + offset; shadowTexCoord.y = 1.0f - shadowTexCoord.y; // Offset the coordinate by half a texel so we sample it correctly shadowTexCoord += (0.5f / ShadowMapSize); Any help with this would be greatly appreciated, thanks in advance.

General and Gameplay Programming Programming

Started by riuthamus February 13, 2013 11:17 PM

46 comments, last by riuthamus 11 years, 1 month ago

Telanor

1,486

February 20, 2013 11:10 AM

position_vs.xy = (screencoord.xy * screensize_ws.wx) * (far_ws*depth/near_ws);

screensize_ws.wx? Did you mean xy?

What range is screencoord supposed to be in? DX gives it as the actual on screen pixel coordinate. In my case screencoord.xy = x = 536.500000000, y = 308.500000000. position_vs is coming out to a huge number after the multiplication.

Ashaman73

13,718

February 20, 2013 01:11 PM

I've corrected my previous post. Screensize is the screensize in world space divided by the screensize in pixels. Best to look up my previous post.

Ashaman

Gnoblins: Website - Facebook - Twitter - Youtube - Steam Greenlit - IndieDB - Gamedev Log

Telanor

1,486

February 21, 2013 04:22 AM

I can't get the reconstructed position to even come close to the position_ws_VS. I found something else out, however, while reading MJPs blog about reconstructing linear depth. The GetCorners function in SharpDX returns the corners in a different order than XNA. So I switched the indexes around and now the first shader output seems a little closer to what it should be (at least I think so...). It no longer rotates with the camera, but it's still moving with it:

Ashaman73

13,718

February 21, 2013 06:32 AM

It no longer rotates with the camera, but it's still moving with it:

That's the desired behavior So, the world space transformation should work (better) now. Have you tested the complete shadowmap pipeline with this fix ?

Ashaman

Gnoblins: Website - Facebook - Twitter - Youtube - Steam Greenlit - IndieDB - Gamedev Log

Telanor

1,486

February 21, 2013 06:59 AM

Yea I did test them. Assuming I didn't leave in any weird test stuff (I didn't see any), the shadows don't make any sense at all now:

[attachment=13771:RuinValor 2013-02-21 01-58-54-96.png]

With 4 cascades and a recognizable object:

[attachment=13772:RuinValor 2013-02-21 02-04-45-41.png]

Edit - This is what it looks like with the precision increased from 8 to 32 bits:

[attachment=13773:RuinValor 2013-02-21 02-14-06-47.png]

[attachment=13774:RuinValor 2013-02-21 02-14-15-86.png]

Ashaman73

13,718

February 21, 2013 07:40 AM

the shadows don't make any sense at all now

It is hard to say, but you might have entered the next stage: fighting shadowmapping artifacts

Currently your scene could suffer from self-shadowing artifacts (or an other bug), but basically that could mean, that the transformation works. Best to use one recognizable shadow caster (e.g. sphere) and leave the rest only as shadow receiver (no caster).

precision increased from 8

8 bit depth buffer ?? You should use at least 24 bits (e.g. 2 half floats or 32f rendering target).

Ashaman

Gnoblins: Website - Facebook - Twitter - Youtube - Steam Greenlit - IndieDB - Gamedev Log

Telanor

1,486

February 21, 2013 08:21 AM

Increasing the shadow bias fixed the remaining issue:

[attachment=13775:RuinValor 2013-02-21 03-13-26-14.png]

So it's pretty much working now. I'm getting just about every other problem possible though. Peter panning, jittering when moving, shadows disappearing when the object is behind the camera, and some weird issue where a triangle of shadow pops in on the right hand side (you can see it in the screenshot). The last two issues are particularly problematic though.

Edit: Switching to our old ESM shadow filtering (instead of PCF from the sample) mostly fixes issues 1, 2, and possibly 4. 3rd one is still something I'd really like to fix.