more random fixes to new clustered implementation (still disabled)

more random fixes to new clustered implementation (still disabled)

Assets/ScriptableRenderLoop/common/ShaderBase.h

 	return 1 - depthTexture.Load(uint3(pixCoord.xy, 0)).x;
 }
 
+float FetchDepthMSAA(Texture2DMS<float> depthTexture, uint2 pixCoord, uint sampleIdx)
+{
+	return 1 - depthTexture.Load(uint3(pixCoord.xy, 0), sampleIdx).x;
+}
+
 #endif

Assets/ScriptableRenderLoop/fptl/FptlLighting.cs

 		static private int kGenAABBKernel;
 		static private int kGenListPerTileKernel;
         static private int kGenListPerVoxelKernel;
+        static private int kClearVoxelAtomicKernel;
 		static private ComputeBuffer m_lightDataBuffer;
 		static private ComputeBuffer m_convexBoundsBuffer;
 		static private ComputeBuffer m_aabbBoundsBuffer;
         static private ComputeBuffer m_perVoxelLightLists;
         static private ComputeBuffer m_perVoxelOffset;
         static private ComputeBuffer m_globalLightListAtomic;
-        static private ComputeBuffer m_singleZeroUInt;      // lazy way to clear m_globalLightListAtomic once per frame
         // clustered light list specific buffers and data end
 
         static private int m_WidthOnRecord;
             {
                 if(m_globalLightListAtomic!=null)
 				    m_globalLightListAtomic.Release();
-
-                if(m_singleZeroUInt!=null)
-				    m_singleZeroUInt.Release();
             }
 
 		}
 
             if(gEnableClustered)
             {
-                kGenListPerVoxelKernel = m_BuildPerVoxelLightListShader.FindKernel("TileLightListGen");
+                kGenListPerVoxelKernel = m_BuildPerVoxelLightListShader.FindKernel("TileLightListGen_NoDepthRT");
+                kClearVoxelAtomicKernel = m_BuildPerVoxelLightListShader.FindKernel("ClearAtomic");
-                m_singleZeroUInt  = new ComputeBuffer(1, sizeof(uint));
             }
 
 			m_cookieTexArray = new TextureCache2D();
             if(gEnableClustered)
             {
                 m_globalLightListAtomic.Release();
-                m_singleZeroUInt.Release();
             }
 		}
 
 
         void VoxelLightListGeneration(CommandBuffer cmd, Camera camera, int numLights, Matrix4x4 projscr, Matrix4x4 invProjscr)
         {
-            // copy resource m_globalLightListAtomic <-- m_singleZeroUInt
+            // clear atomic offset index
+            cmd.SetComputeBufferParam(m_BuildPerVoxelLightListShader, kClearVoxelAtomicKernel, "g_LayeredSingleIdxBuffer", m_globalLightListAtomic);
+            cmd.ComputeDispatch(m_BuildPerVoxelLightListShader, kClearVoxelAtomicKernel, 1, 1, 1);
+
             cmd.SetComputeIntParam(m_BuildPerVoxelLightListShader, "g_iNrVisibLights", numLights);
 			SetMatrixCS(cmd, m_BuildPerVoxelLightListShader, "g_mScrProjection", projscr);
 			SetMatrixCS(cmd, m_BuildPerVoxelLightListShader, "g_mInvScrProjection", invProjscr);
 			cmd.SetComputeBufferParam(m_BuildPerVoxelLightListShader, kGenListPerVoxelKernel, "g_vLayeredLightList", m_perVoxelLightLists);
             cmd.SetComputeBufferParam(m_BuildPerVoxelLightListShader, kGenListPerVoxelKernel, "g_LayeredOffset", m_perVoxelOffset);
             cmd.SetComputeBufferParam(m_BuildPerVoxelLightListShader, kGenListPerVoxelKernel, "g_LayeredSingleIdxBuffer", m_globalLightListAtomic);
-			//cmd.ComputeDispatch(m_BuildPerVoxelLightListShader, kGenListPerVoxelKernel, nrTilesX, nrTilesY, 1);
-
+            int nrTilesX = (camera.pixelWidth + 15) / 16;
+			int nrTilesY = (camera.pixelHeight + 15) / 16;
+			cmd.ComputeDispatch(m_BuildPerVoxelLightListShader, kGenListPerVoxelKernel, nrTilesX, nrTilesY, 1);
         }
 	}
 }

Assets/ScriptableRenderLoop/fptl/lightlistbuild-clustered.compute

-#pragma kernel TileLightListGen
+#pragma kernel TileLightListGen_NoDepthRT
+#pragma kernel TileLightListGen_DepthRT			ENABLE_DEPTH_TEXTURE_BACKPLANE
+#pragma kernel TileLightListGen_DepthRT_MSAA	ENABLE_DEPTH_TEXTURE_BACKPLANE		MSAA_ENABLED
+#pragma kernel ClearAtomic
 
 #include "..\common\ShaderBase.h"
 #include "LightDefinitions.cs.hlsl"
 bool CheckIntersection(int l, int k, uint2 viTilLL, uint2 viTilUR, float fModulUserScale)
 {
 	unsigned int val = (clusterIdxs[l>>1]>>(16*(l&1)))&0xffff;
-	bool bIsHit = ((val>>0)&0xff)<=k && k<=((val>>8)&0xff);
+	bool bIsHit = ((val>>0)&0xff)<=((uint) k) && ((uint) k)<=((val>>8)&0xff);
 	if(bIsHit)
 	{
 #ifdef CONV_HULL_TEST_ENABLED
 bool CheckIntersectionBasic(int l, int k)
 {
 	unsigned int val = (clusterIdxs[l>>1]>>(16*(l&1)))&0xffff;
-	return ((val>>0)&0xff)<=k && k<=((val>>8)&0xff);
+	return ((val>>0)&0xff)<=((uint) k) && ((uint) k)<=((val>>8)&0xff);
 }
 
 
 #endif
 
 	//////////// cell specific code
-	for(int l=(int) t; l<((iNrCoarseLights+1)>>1); l += NR_THREADS)
-		const int l0 = coarseList[2*l+0], l1 = coarseList[min(2*l+1,iNrCoarseLights)];
-		const unsigned int clustIdxMi0 = (const unsigned int) min(255,SnapToClusterIdx(GetLinearDepth(g_vBoundsBuffer[l0].z), fModulUserScale));
-		const unsigned int clustIdxMa0 = (const unsigned int) min(255,SnapToClusterIdx(GetLinearDepth(g_vBoundsBuffer[l0+g_iNrVisibLights].z), fModulUserScale));
-		const unsigned int clustIdxMi1 = (const unsigned int) min(255,SnapToClusterIdx(GetLinearDepth(g_vBoundsBuffer[l1].z), fModulUserScale));
-		const unsigned int clustIdxMa1 = (const unsigned int) min(255,SnapToClusterIdx(GetLinearDepth(g_vBoundsBuffer[l1+g_iNrVisibLights].z), fModulUserScale));
+		for(int l=(int) t; l<((iNrCoarseLights+1)>>1); l += NR_THREADS)
+		{
+			const int l0 = coarseList[2*l+0], l1 = coarseList[min(2*l+1,iNrCoarseLights)];
+			const unsigned int clustIdxMi0 = (const unsigned int) min(255,SnapToClusterIdx(GetLinearDepth(g_vBoundsBuffer[l0].z), fModulUserScale));
+			const unsigned int clustIdxMa0 = (const unsigned int) min(255,SnapToClusterIdx(GetLinearDepth(g_vBoundsBuffer[l0+g_iNrVisibLights].z), fModulUserScale));
+			const unsigned int clustIdxMi1 = (const unsigned int) min(255,SnapToClusterIdx(GetLinearDepth(g_vBoundsBuffer[l1].z), fModulUserScale));
+			const unsigned int clustIdxMa1 = (const unsigned int) min(255,SnapToClusterIdx(GetLinearDepth(g_vBoundsBuffer[l1+g_iNrVisibLights].z), fModulUserScale));
-		clusterIdxs[l] = (clustIdxMa1<<24) | (clustIdxMi1<<16) | (clustIdxMa0<<8) | (clustIdxMi0<<0);
+			clusterIdxs[l] = (clustIdxMa1<<24) | (clustIdxMi1<<16) | (clustIdxMa0<<8) | (clustIdxMi0<<0);
+		}
 	}
 
 #if !defined(XBONE) && !defined(PLAYSTATION4)
 	}
 
 	uint localOffs=0;
-	uint idx = i*nrTilesX*nrTilesY + tileIDX.y*nrTilesX + tileIDX.x;
+	offs = i*nrTilesX*nrTilesY + tileIDX.y*nrTilesX + tileIDX.x;
-			g_LayeredOffset[idx] = (start+localOffs) | (((uint) numLights)<<27);
-			idx += (nrClusters*nrTilesX*nrTilesY);
+			g_LayeredOffset[offs] = (start+localOffs) | (((uint) numLights)<<27);
+			offs += (nrClusters*nrTilesX*nrTilesY);
 			localOffs += modelListCount[m];		// use unclamped count for localOffs
 		}
 	}
 #endif
 	return lightOffs2;
 }
-#endif
+#endif
+
+
+
+[numthreads(1, 1, 1)]
+void ClearAtomic(uint threadID : SV_GroupIndex, uint3 u3GroupID : SV_GroupID)
+{
+	g_LayeredSingleIdxBuffer[0]=0;
+}