VideoCommon: use the Light struct in XF memory

Source/Core/VideoBackends/Software/TransformUnit.cpp

@@ -210,7 +210,7 @@ inline float SafeDivide(float n, float d)
 
 void LightColor(const Vec3 &pos, const Vec3 &normal, u8 lightNum, const LitChannel &chan, Vec3 &lightCol)
 {
-	const LightPointer *light = (const LightPointer*)&xfmem.lights[0x10*lightNum];
+	const LightPointer *light = (const LightPointer*)&xfmem.lights[lightNum];
 
 	if (!(chan.attnfunc & 1))
 	{
@@ -295,7 +295,7 @@ void LightColor(const Vec3 &pos, const Vec3 &normal, u8 lightNum, const LitChann
 
 void LightAlpha(const Vec3 &pos, const Vec3 &normal, u8 lightNum, const LitChannel &chan, float &lightCol)
 {
-	const LightPointer *light = (const LightPointer*)&xfmem.lights[0x10*lightNum];
+	const LightPointer *light = (const LightPointer*)&xfmem.lights[lightNum];
 
 	if (!(chan.attnfunc & 1))
 	{
@@ -476,7 +476,7 @@ void TransformTexCoord(const InputVertexData *src, OutputVertexData *dst, bool s
 			break;
 		case XF_TEXGEN_EMBOSS_MAP:
 			{
-				const LightPointer *light = (const LightPointer*)&xfmem.lights[0x10*texinfo.embosslightshift];
+				const LightPointer *light = (const LightPointer*)&xfmem.lights[texinfo.embosslightshift];
 
 				Vec3 ldir = (light->pos - dst->mvPosition).normalized();
 				float d1 = ldir * dst->normal[1];

Source/Core/VideoBackends/Software/XFMemLoader.cpp

@@ -23,7 +23,7 @@ void XFWritten(u32 transferSize, u32 baseAddress)
 	// fix lights so invalid values don't trash the lighting computations
 	if (baseAddress <= 0x067f && topAddress >= 0x0604)
 	{
-		u32* x = xfmem.lights;
+		u32* x = (u32*)xfmem.lights;
 
 		// go through all lights
 		for (int light = 0; light < 8; light++)

Source/Core/VideoCommon/PixelShaderManager.cpp

@@ -108,16 +108,15 @@ void PixelShaderManager::SetConstants()
 			// lights don't have a 1 to 1 mapping, the color component needs to be converted to 4 floats
 			int istart = nLightsChanged[0] / 0x10;
 			int iend = (nLightsChanged[1] + 15) / 0x10;
-			const float* xfmemptr = (const float*)&xfmem.lights[0x10 * istart];
 
 			for (int i = istart; i < iend; ++i)
 			{
-				u32 color = *(const u32*)(xfmemptr + 3);
-				constants.plight_colors[i][0] = (color >> 24) & 0xFF;
-				constants.plight_colors[i][1] = (color >> 16) & 0xFF;
-				constants.plight_colors[i][2] = (color >> 8)  & 0xFF;
-				constants.plight_colors[i][3] = (color)       & 0xFF;
-				xfmemptr += 4;
+				const Light& light = xfmem.lights[i];
+				constants.plight_colors[i][0] = light.color[3];
+				constants.plight_colors[i][1] = light.color[2];
+				constants.plight_colors[i][2] = light.color[1];
+				constants.plight_colors[i][3] = light.color[0];
+				const float* xfmemptr = light.cosatt;
 
 				for (int j = 0; j < 4; ++j, xfmemptr += 3)
 				{

Source/Core/VideoCommon/VertexShaderManager.cpp

@@ -247,16 +247,17 @@ void VertexShaderManager::SetConstants()
 		// lights don't have a 1 to 1 mapping, the color component needs to be converted to 4 floats
 		int istart = nLightsChanged[0] / 0x10;
 		int iend = (nLightsChanged[1] + 15) / 0x10;
-		const float* xfmemptr = (const float*)&xfmem.lights[0x10 * istart];
 
 		for (int i = istart; i < iend; ++i)
 		{
-			u32 color = *(const u32*)(xfmemptr + 3);
-			constants.light_colors[i][0] = (color >> 24) & 0xFF;
-			constants.light_colors[i][1] = (color >> 16) & 0xFF;
-			constants.light_colors[i][2] = (color >> 8)  & 0xFF;
-			constants.light_colors[i][3] = (color)       & 0xFF;
-			xfmemptr += 4;
+			const Light& light = xfmem.lights[i];
+
+			// xfmem.light.color is packed as abgr in u8[4], so we have to swap the order
+			constants.light_colors[i][0] = light.color[3];
+			constants.light_colors[i][1] = light.color[2];
+			constants.light_colors[i][2] = light.color[1];
+			constants.light_colors[i][3] = light.color[0];
+			const float* xfmemptr = light.cosatt;
 
 			for (int j = 0; j < 4; ++j, xfmemptr += 3)
 			{

Source/Core/VideoCommon/XFMemory.h

@@ -194,13 +194,9 @@ union DualTexInfo
 struct Light
 {
 	u32 useless[3];
-	u32 color; // rgba
-	float a0;  // attenuation
-	float a1;
-	float a2;
-	float k0;  // k stuff
-	float k1;
-	float k2;
+	u8 color[4];
+	float cosatt[3]; // cos attenuation
+	float distatt[3]; // dist attenuation
 
 	union
 	{
@@ -241,7 +237,7 @@ struct XFMemory
 	u32 normalMatrices[96];         // 0x0400 - 0x045f
 	u32 unk1[160];                  // 0x0460 - 0x04ff
 	u32 postMatrices[256];          // 0x0500 - 0x05ff
-	u32 lights[128];                // 0x0600 - 0x067f
+	Light lights[8];                // 0x0600 - 0x067f
 	u32 unk2[2432];                 // 0x0680 - 0x0fff
 	u32 error;                      // 0x1000
 	u32 diag;                       // 0x1001