#pragma kernel BigTileLightListGen

#include "ShaderBase.h"
#include "LightDefinitions.cs.hlsl"

#include "LightingConvexHullUtils.hlsl"
#include "SortingComputeUtils.hlsl"

#define EXACT_EDGE_TESTS
#define PERFORM_SPHERICAL_INTERSECTION_TESTS

#define MAX_NR_BIGTILE_LIGHTS				(MAX_NR_BIGTILE_LIGHTS_PLUSONE-1)


uniform int g_isOrthographic;
uniform int g_iNrVisibLights;
uniform uint2 g_viDimensions;
uniform float4x4 g_mInvScrProjection;
uniform float4x4 g_mScrProjection;
uniform float g_fNearPlane;
uniform float g_fFarPlane;

StructuredBuffer<float3> g_vBoundsBuffer : register( t1 );
StructuredBuffer<SFiniteLightData> g_vLightData : register( t2 );
StructuredBuffer<SFiniteLightBound> g_data : register( t3 );


#define NR_THREADS			64

// output buffer
RWStructuredBuffer<uint> g_vLightList : register( u0 );		// don't support RWBuffer yet in unity


// 2kB (room for roughly 30 wavefronts)
groupshared unsigned int lightsListLDS[MAX_NR_BIGTILE_LIGHTS_PLUSONE];
groupshared uint lightOffs;


float GetLinearDepth(float zDptBufSpace)	// 0 is near 1 is far
{
	float3 vP = float3(0.0f,0.0f,zDptBufSpace);
	float4 v4Pres = mul(g_mInvScrProjection, float4(vP,1.0));
	return v4Pres.z / v4Pres.w;
}


float3 GetViewPosFromLinDepth(float2 v2ScrPos, float fLinDepth)
{
	bool isOrthographic = g_isOrthographic!=0;
	float fSx = g_mScrProjection[0].x;
	float fSy = g_mScrProjection[1].y;
	float fCx = isOrthographic ? g_mScrProjection[0].w : g_mScrProjection[0].z;
	float fCy = isOrthographic ? g_mScrProjection[1].w : g_mScrProjection[1].z;

#if USE_LEFTHAND_CAMERASPACE
	bool useLeftHandVersion = true;
#else
	bool useLeftHandVersion = isOrthographic;
#endif

	float s = useLeftHandVersion ? 1 : (-1);
	float2 p = float2( (s*v2ScrPos.x-fCx)/fSx, (s*v2ScrPos.y-fCy)/fSy);

	return float3(isOrthographic ? p.xy : (fLinDepth*p.xy), fLinDepth);
}

float GetOnePixDiagWorldDistAtDepthOne()
{
	float fSx = g_mScrProjection[0].x;
	float fSy = g_mScrProjection[1].y;

	return length( float2(1.0/fSx,1.0/fSy) );
}


#ifdef PERFORM_SPHERICAL_INTERSECTION_TESTS
void SphericalIntersectionTests(uint threadID, int iNrCoarseLights, float2 screenCoordinate);
#endif

#ifdef EXACT_EDGE_TESTS
void CullByExactEdgeTests(uint threadID, int iNrCoarseLights, uint2 viTilLL, uint2 viTilUR);
#endif




[numthreads(NR_THREADS, 1, 1)]
void BigTileLightListGen(uint threadID : SV_GroupIndex, uint3 u3GroupID : SV_GroupID)
{
	uint2 tileIDX = u3GroupID.xy;
	uint t=threadID;

	uint iWidth = g_viDimensions.x;
	uint iHeight = g_viDimensions.y;
	uint nrBigTilesX = (iWidth+63)/64;
	uint nrBigTilesY = (iHeight+63)/64;

	if(t==0) lightOffs = 0;

#if !defined(SHADER_API_XBOXONE) && !defined(SHADER_API_PSSL)
	GroupMemoryBarrierWithGroupSync();
#endif


	uint2 viTilLL = 64*tileIDX;
	uint2 viTilUR = min( viTilLL+uint2(64,64), uint2(iWidth, iHeight) );			// not width and height minus 1 since viTilUR represents the end of the tile corner.

	float2 vTileLL = float2(viTilLL.x/(float) iWidth, viTilLL.y/(float) iHeight);
	float2 vTileUR = float2(viTilUR.x/(float) iWidth, viTilUR.y/(float) iHeight);

	// build coarse list using AABB
	for(int l=(int) t; l<(int) g_iNrVisibLights; l += NR_THREADS)
	{
		const float2 vMi = g_vBoundsBuffer[l].xy;
		const float2 vMa = g_vBoundsBuffer[l+g_iNrVisibLights].xy;

		if( all(vMa>vTileLL) && all(vMi<vTileUR))
		{
			unsigned int uInc = 1;
			unsigned int uIndex;
			InterlockedAdd(lightOffs, uInc, uIndex);
			if(uIndex<MAX_NR_BIGTILE_LIGHTS) lightsListLDS[uIndex] = l;		// add to light list
		}
	}

#if !defined(SHADER_API_XBOXONE) && !defined(SHADER_API_PSSL)
	GroupMemoryBarrierWithGroupSync();
#endif

	int iNrCoarseLights = min(lightOffs,MAX_NR_BIGTILE_LIGHTS);

#ifdef PERFORM_SPHERICAL_INTERSECTION_TESTS
	SphericalIntersectionTests( t, iNrCoarseLights, float2(min(viTilLL.xy+uint2(64/2,64/2), uint2(iWidth-1, iHeight-1))) );
#endif

#ifdef EXACT_EDGE_TESTS
	CullByExactEdgeTests(t, iNrCoarseLights, viTilLL.xy, viTilUR.xy);
#endif


	// sort lights
	SORTLIST(lightsListLDS, iNrCoarseLights, MAX_NR_BIGTILE_LIGHTS_PLUSONE, t, NR_THREADS);

	if(t==0) lightOffs = 0;
	GroupMemoryBarrierWithGroupSync();
	int i;
	for(i=t; i<iNrCoarseLights; i+=NR_THREADS) if(lightsListLDS[i]<(uint)g_iNrVisibLights) InterlockedAdd(lightOffs, 1);
	GroupMemoryBarrierWithGroupSync();
	iNrCoarseLights = lightOffs;

	int offs = tileIDX.y*nrBigTilesX + tileIDX.x;

	for(i=t; i<(iNrCoarseLights+1); i+=NR_THREADS)
		g_vLightList[MAX_NR_BIGTILE_LIGHTS_PLUSONE*offs + i] = i==0 ? iNrCoarseLights : lightsListLDS[max(i-1, 0)];
}


#ifdef PERFORM_SPHERICAL_INTERSECTION_TESTS
void SphericalIntersectionTests(uint threadID, int iNrCoarseLights, float2 screenCoordinate)
{
#if USE_LEFTHAND_CAMERASPACE
	float3 V = GetViewPosFromLinDepth( screenCoordinate, 1.0);
#else
	float3 V = GetViewPosFromLinDepth( screenCoordinate, -1.0);
#endif

	float onePixDiagDist = GetOnePixDiagWorldDistAtDepthOne();
	float halfTileSizeAtZDistOne = 32*onePixDiagDist;		// scale by half a tile

	for(int l=threadID; l<iNrCoarseLights; l+=NR_THREADS)
	{
		SFiniteLightBound lgtDat = g_data[lightsListLDS[l]];

		if( !DoesSphereOverlapTile(V, halfTileSizeAtZDistOne, lgtDat.center.xyz, lgtDat.radius, g_isOrthographic!=0) )
			lightsListLDS[l]=0xffffffff;
	}

#if !defined(SHADER_API_XBOXONE) && !defined(SHADER_API_PSSL)
		GroupMemoryBarrierWithGroupSync();
#endif
}
#endif







#ifdef EXACT_EDGE_TESTS
float3 GetTileVertex(uint2 viTilLL, uint2 viTilUR, int i, float fTileFarPlane)
{
	float x = (i&1)==0 ? viTilLL.x : viTilUR.x;
	float y = (i&2)==0 ? viTilLL.y : viTilUR.y;
	float z = (i&4)==0 ? g_fNearPlane : fTileFarPlane;
#if !USE_LEFTHAND_CAMERASPACE
	z = -z;
#endif
	return GetViewPosFromLinDepth( float2(x, y), z);
}

void GetFrustEdge(out float3 vP0, out float3 vE0, const int e0, uint2 viTilLL, uint2 viTilUR, float fTileFarPlane)
{
	int iSection = e0>>2;		// section 0 is side edges, section 1 is near edges and section 2 is far edges
	int iSwizzle = e0&0x3;

	int i=iSwizzle + (2*(iSection&0x2));	// offset by 4 at section 2
	vP0 = GetTileVertex(uint2(viTilLL.x, viTilUR.y), uint2(viTilUR.x, viTilLL.y), i, fTileFarPlane);

#if USE_LEFTHAND_CAMERASPACE
	float3 edgeSectionZero = g_isOrthographic==0 ? vP0 : float3(0.0,0.0,1.0); 
#else
	float3 edgeSectionZero = g_isOrthographic==0 ? vP0 : float3(0.0,0.0,-1.0); 
#endif

	vE0 = iSection == 0 ? edgeSectionZero : (((iSwizzle & 0x2) == 0 ? 1.0f : (-1.0f)) * ((int)(iSwizzle & 0x1) == (iSwizzle >> 1) ? float3(1, 0, 0) : float3(0, 1, 0)));
}

void CullByExactEdgeTests(uint threadID, int iNrCoarseLights, uint2 viTilLL, uint2 viTilUR)
{
	const bool bOnlyNeedFrustumSideEdges = true;
	const int nrFrustEdges = bOnlyNeedFrustumSideEdges ? 4 : 8;	// max 8 since we never need to test 4 far edges of frustum since they are identical vectors to near edges and plane is placed at vP0 on light hull.

	const int totNrEdgePairs = 12*nrFrustEdges;
	for(int l=0; l<iNrCoarseLights; l++)
	{
		const uint idxCoarse = lightsListLDS[l];

		bool canEnter = idxCoarse<(uint) g_iNrVisibLights;
		if(canEnter) canEnter = g_vLightData[idxCoarse].lightType!=SPHERE_LIGHT;		// don't bother doing edge tests for sphere lights since these have camera aligned bboxes.
		[branch]if(canEnter)
		{
			SFiniteLightBound lgtDat = g_data[idxCoarse];

			const float3 boxX = lgtDat.boxAxisX.xyz;
			const float3 boxY = lgtDat.boxAxisY.xyz;
			const float3 boxZ = -lgtDat.boxAxisZ.xyz;	// flip axis (so it points away from the light direction for a spot-light)
			const float3 center = lgtDat.center.xyz;
			const float2 scaleXY = lgtDat.scaleXY;

			for(int i=threadID; i<totNrEdgePairs; i+=NR_THREADS)
			{
				int e0 = (int) (((uint)i)/((uint) nrFrustEdges)); // should become a shift right
				int e1 = i - e0*nrFrustEdges;

				int idx_cur=0, idx_twin=0;
				float3 vP0, vE0;
				GetHullEdge(idx_cur, idx_twin, vP0, vE0, e0, boxX, boxY, boxZ, center, scaleXY);


				float3 vP1, vE1;
				GetFrustEdge(vP1, vE1, e1, viTilLL, viTilUR, g_fFarPlane);

				// potential separation plane
				float3 vN = cross(vE0, vE1);

				int positive=0, negative=0;
				for(int k=1; k<8; k++)		// only need to test 7 verts (technically just 6).
				{
					int j = (idx_cur+k)&0x7;
					float3 vPh = GetHullVertex(boxX, boxY, boxZ, center, scaleXY, j);
					float fSignDist = idx_twin==j ? 0.0 : dot(vN, vPh-vP0);
					if(fSignDist>0) ++positive; else if(fSignDist<0) ++negative;
				}
				int resh = (positive>0 && negative>0) ? 0 : (positive>0 ? 1 : (negative>0 ? (-1) : 0));

				positive=0; negative=0;
				for(int j=0; j<8; j++)
				{
					float3 vPf = GetTileVertex(viTilLL, viTilUR, j, g_fFarPlane);
					float fSignDist = dot(vN, vPf-vP0);
					if(fSignDist>0) ++positive; else if(fSignDist<0) ++negative;
				}
				int resf = (positive>0 && negative>0) ? 0 : (positive>0 ? 1 : (negative>0 ? (-1) : 0));

				bool bFoundSepPlane = (resh*resf)<0;
				if(bFoundSepPlane) lightsListLDS[l]=0xffffffff;
			}
		}
	}
#if !defined(SHADER_API_XBOXONE) && !defined(SHADER_API_PSSL)
		GroupMemoryBarrierWithGroupSync();
#endif
}
#endif