Light Culling still doesn't work properly

res/shaders/LightCulling.slang

@@ -23,19 +23,26 @@ layout(set = INDEX_VIEW_PARAMS, binding = 3)
 StructuredBuffer<Frustum> frustums;
 
 layout(set = INDEX_VIEW_PARAMS, binding = 4)
-RWStructuredBuffer<uint> oLightIndexCounter;
+globallycoherent RWStructuredBuffer<uint> oLightIndexCounter;
 layout(set = INDEX_VIEW_PARAMS, binding = 5)
-RWStructuredBuffer<uint> tLightIndexCounter;
+globallycoherent RWStructuredBuffer<uint> tLightIndexCounter;
 
 layout(set = INDEX_VIEW_PARAMS, binding = 6)
 RWStructuredBuffer<uint> oLightIndexList;
 layout(set = INDEX_VIEW_PARAMS, binding = 7)
 RWStructuredBuffer<uint> tLightIndexList;
+
 layout(set = INDEX_VIEW_PARAMS, binding = 8)
 RWTexture2D<uint2> oLightGrid;
 layout(set = INDEX_VIEW_PARAMS, binding = 9)
 RWTexture2D<uint2> tLightGrid;
 
+// Debug
+//layout(set = INDEX_VIEW_PARAMS, binding = 10)
+//Texture2D lightCountHeatMap;
+//layout(set = INDEX_VIEW_PARAMS, binding = 11)
+//RWTexture2D<float4> debugTexture;
+
 groupshared uint uMinDepth;
 groupshared uint uMaxDepth;
 
@@ -74,8 +81,8 @@ void tAppendLight(uint lightIndex)
 [shader("compute")]
 void cullLights(ComputeShaderInput in)
 {
-    int3 texCoord = int3(in.dispatchThreadID.xy, 0);
-    float fDepth = depthTextureVS.Load(texCoord).r;
+    int2 texCoord = in.dispatchThreadID.xy;
+    float fDepth = depthTextureVS.Load(int3(texCoord, 0)).r;
 
     uint uDepth = asuint(fDepth);
     if(in.groupIndex == 0)
@@ -88,6 +95,7 @@ void cullLights(ComputeShaderInput in)
     }
 
     GroupMemoryBarrierWithGroupSync();
+
     InterlockedMin(uMinDepth, uDepth);
     InterlockedMax(uMaxDepth, uDepth);
 
@@ -120,22 +128,24 @@ void cullLights(ComputeShaderInput in)
 
     if(in.groupIndex == 0)
     {
-        InterlockedAdd(oLightIndexCounter[0], (uint)oLightCount, oLightIndexStartOffset);
+        InterlockedAdd(oLightIndexCounter[0], oLightCount, oLightIndexStartOffset);
         oLightGrid[in.groupID.xy] = uint2(oLightIndexStartOffset, oLightCount);
 
-        InterlockedAdd(tLightIndexCounter[0], (uint)tLightCount, tLightIndexStartOffset);
+        InterlockedAdd(tLightIndexCounter[0], tLightCount, tLightIndexStartOffset);
         tLightGrid[in.groupID.xy] = uint2(tLightIndexStartOffset, tLightCount);
     }
     GroupMemoryBarrierWithGroupSync();
 
-    for (uint j = in.groupIndex; j < (uint)oLightCount; j += BLOCK_SIZE * BLOCK_SIZE)
+    for (uint j = in.groupIndex; j < oLightCount; j += BLOCK_SIZE * BLOCK_SIZE)
     {
         oLightIndexList[oLightIndexStartOffset + j] = oLightList[j];
     }
     
     // For transparent geometry.
-    for ( uint k = in.groupIndex; k < (uint)tLightCount; k += BLOCK_SIZE * BLOCK_SIZE )
+    for ( uint k = in.groupIndex; k < tLightCount; k += BLOCK_SIZE * BLOCK_SIZE )
     {
         tLightIndexList[tLightIndexStartOffset + k] = tLightList[k];
     }
+
+    
 }

res/shaders/lib/Common.slang

@@ -1,14 +1,14 @@
 const static float PI = 3.1415926535897932f;
 const static uint MAX_PARTICLES = 65536;
-const static uint BLOCK_SIZE = 8;
+const static uint BLOCK_SIZE = 32;
 
 struct ViewParameter
 {
     float4x4 viewMatrix;
     float4x4 projectionMatrix;
-	float4x4 inverseProjection;
+    float4x4 inverseProjection;
     float4 cameraPos_WS;
-	float2 screenDimensions;
+    float2 screenDimensions;
 }
 layout(set = INDEX_VIEW_PARAMS, binding = 0, std430)
 ConstantBuffer<ViewParameter> gViewParams;
@@ -31,37 +31,31 @@ float4 screenToView( float4 screen )
     float2 texCoord = screen.xy / gViewParams.screenDimensions;
  
     // Convert to clip space
-    float4 clip = float4( float2( texCoord.x, -texCoord.y ) * 2.0f - 1.0f, screen.z, screen.w );
+    float4 clip = float4( float2( texCoord.x, 1.0f-texCoord.y ) * 2.0f - 1.0f, screen.z, screen.w );
  
     return clipToView( clip );
 }
 
 struct Plane
 {
-	float3 n;
-	float d;
-    float3 p0;
-    float3 p1;
-    float3 p2;
+    float3 n;
+    float d;
 };
 
 struct Frustum
 {
-	Plane planes[4];
+    Plane planes[4];
 };
 Plane computePlane(float3 p0, float3 p1, float3 p2)
 {
-	Plane plane;
+    Plane plane;
 
-	float3 v0 = p2 - p0;
-	float3 v2 = p1 - p0;
+    float3 v0 = p1 - p0;
+    float3 v2 = p2 - p0;
 
-	plane.n = normalize(cross(v0, v2));
+    plane.n = normalize(cross(v0, v2));
 
-	plane.d = dot(plane.n, p0);
-    plane.p0 = p0;
-    plane.p1 = p1;
-    plane.p2 = p2;
+    plane.d = dot(plane.n, p0);
 
-	return plane;
+    return plane;
 }

res/shaders/lib/LightEnv.slang

@@ -22,7 +22,6 @@ struct DirectionalLight : ILightEnv
 struct PointLight : ILightEnv
 {
     float4 positionWS;
-    float4 positionVS;
     float4 colorRange;
     
     float3 illuminate<B:IBRDF>(MaterialFragmentParameter input, B brdf, float3 viewDir)
@@ -36,14 +35,14 @@ struct PointLight : ILightEnv
 
 	bool insidePlane(Plane plane)
 	{
-        return dot(plane.n, positionVS.xyz) - plane.d < -colorRange.w;
+        return dot(plane.n, getViewPos().xyz) - plane.d < -colorRange.w;
 	}
 
 	bool insideFrustum(Frustum frustum, float zNear, float zFar)
 	{
         bool result = true;
 
-		if(positionVS.z - colorRange.w > zNear || positionVS.z + colorRange.w < zFar)
+		if(getViewPos().z - colorRange.w > zNear || getViewPos().z + colorRange.w < zFar)
 		{
 			result = false;
 		}
@@ -56,6 +55,10 @@ struct PointLight : ILightEnv
 		}
 		return result;
 	}
+	float3 getViewPos()
+	{
+		return mul(gViewParams.viewMatrix, positionWS).xyz;
+	}
 };
 
 layout(set = INDEX_LIGHT_ENV, binding = 0, std430)