native depth interpolation option to help AMD gpus

When enabled, set 1/z in pos.w and use native interpolation for colors,
UV and depth.
Get rid of manual modvol clipping in dx11.

core/cfg/option.cpp

@@ -101,6 +101,7 @@ Option<int> TextureFiltering("rend.TextureFiltering", 0); // Default
 Option<bool> ThreadedRendering("rend.ThreadedRendering", true);
 Option<bool> DupeFrames("rend.DupeFrames", false);
 Option<int> PerPixelLayers("rend.PerPixelLayers", 32);
+Option<bool> NativeDepthInterpolation("rend.NativeDepthInterpolation", false);
 
 // Misc
 

core/cfg/option.h

@@ -459,6 +459,7 @@ extern Option<int> AnisotropicFiltering;
 extern Option<int> TextureFiltering; // 0: default, 1: force nearest, 2: force linear
 extern Option<bool> ThreadedRendering;
 extern Option<bool> DupeFrames;
+extern Option<bool> NativeDepthInterpolation;
 
 // Misc
 

core/rend/dx11/dx11_renderer.cpp

@@ -318,154 +318,6 @@ bool DX11Renderer::Process(TA_context* ctx)
 	}
 }
 
-//
-// Efficient Triangle and Quadrilateral Clipping within Shaders. M. McGuire
-// Journal of Graphics GPU and Game Tools - November 2011
-//
-static glm::vec3 intersect(const glm::vec3& A, float Adist , const glm::vec3& B, float Bdist)
-{
-	return (A * std::abs(Bdist) + B * std::abs(Adist)) / (std::abs(Adist) + std::abs(Bdist));
-}
-
-// Clip the triangle 'trig' with respect to the plane defined by the given point and normal vector.
-static int sutherlandHodgmanClip(const glm::vec2& point, const glm::vec2& normal, ModTriangle& trig, ModTriangle& newTrig)
-{
-	constexpr float clipEpsilon = 0.f; //0.00001;
-	constexpr float clipEpsilon2 = 0.f; //0.01;
-
-	glm::vec3 v0(trig.x0, trig.y0, trig.z0);
-	glm::vec3 v1(trig.x1, trig.y1, trig.z1);
-	glm::vec3 v2(trig.x2, trig.y2, trig.z2);
-
-	glm::vec3 dist = glm::vec3(
-			glm::dot(glm::vec2(v0) - point, normal),
-			glm::dot(glm::vec2(v1) - point, normal),
-			glm::dot(glm::vec2(v2) - point, normal));
-	if (!glm::any(glm::greaterThanEqual(dist , glm::vec3(clipEpsilon2))))
-		// all clipped
-		return 0;
-	if (glm::all(glm::greaterThanEqual(dist , glm::vec3(-clipEpsilon))))
-		// none clipped
-		return 3;
-
-	// There are either 1 or 2 vertices above the clipping plane.
-	glm::bvec3 above = glm::greaterThanEqual(dist, glm::vec3(0.f));
-	bool nextIsAbove;
-	glm::vec3 v3;
-	// Find the CCW-most vertex above the plane.
-	if (above[1] && !above[0])
-	{
-		// Cycle once CCW. Use v3 as a temp
-		nextIsAbove = above[2];
-		v3 = v0;
-		v0 = v1;
-		v1 = v2;
-		v2 = v3;
-		dist = glm::vec3(dist.y, dist.z, dist.x);
-	}
-	else if (above[2] && !above[1])
-	{
-		// Cycle once CW. Use v3 as a temp.
-		nextIsAbove = above[0];
-		v3 = v2;
-		v2 = v1;
-		v1 = v0;
-		v0 = v3;
-		dist = glm::vec3(dist.z, dist.x, dist.y);
-	}
-	else
-		nextIsAbove = above[1];
-	trig.x0 = v0.x;
-	trig.y0 = v0.y;
-	trig.z0 = v0.z;
-	// We always need to clip v2-v0.
-	v3 = intersect(v0, dist[0], v2, dist[2]);
-	if (nextIsAbove)
-	{
-		v2 = intersect(v1, dist[1], v2, dist[2]);
-		trig.x1 = v1.x;
-		trig.y1 = v1.y;
-		trig.z1 = v1.z;
-		trig.x2 = v2.x;
-		trig.y2 = v2.y;
-		trig.z2 = v2.z;
-		newTrig.x0 = v0.x;
-		newTrig.y0 = v0.y;
-		newTrig.z0 = v0.z;
-		newTrig.x1 = v2.x;
-		newTrig.y1 = v2.y;
-		newTrig.z1 = v2.z;
-		newTrig.x2 = v3.x;
-		newTrig.y2 = v3.y;
-		newTrig.z2 = v3.z;
-
-		return 4;
-	}
-	else
-	{
-		v1 = intersect(v0, dist[0], v1, dist[1]);
-		trig.x1 = v1.x;
-		trig.y1 = v1.y;
-		trig.z1 = v1.z;
-		trig.x2 = v3.x;
-		trig.y2 = v3.y;
-		trig.z2 = v3.z;
-
-		return 3;
-	}
-}
-
-static void clipModVols(List<ModifierVolumeParam>& params, std::vector<ModTriangle>& triangles)
-{
-	for (ModifierVolumeParam& param : params)
-	{
-		std::vector<ModTriangle> trigs(&pvrrc.modtrig.head()[param.first], &pvrrc.modtrig.head()[param.first + param.count]);
-		std::vector<ModTriangle> nextTrigs;
-		nextTrigs.reserve(trigs.size());
-		for (int axis = 0; axis < 4; axis++)
-		{
-			glm::vec2 point;
-			glm::vec2 normal;
-			switch (axis)
-			{
-			case 0: // left
-				point = glm::vec2(-6400.f, 0.f);
-				normal = glm::vec2(1.f, 0.f);
-				break;
-			case 1: // top
-				point = glm::vec2(0.f, -4800.f);
-				normal = glm::vec2(0.f, 1.f);
-				break;
-			case 2: // right
-				point = glm::vec2(7040.f, 0.f);
-				normal = glm::vec2(-1.f, 0.f);
-				break;
-			case 3: // bottom
-				point = glm::vec2(-0.f, 5280.f);
-				normal = glm::vec2(0.f, -1.f);
-				break;
-			}
-
-			for (ModTriangle& trig : trigs)
-			{
-				ModTriangle newTrig;
-				int size = sutherlandHodgmanClip(point, normal, trig, newTrig);
-				if (size > 0)
-				{
-					nextTrigs.push_back(trig);
-					if (size == 4)
-						nextTrigs.push_back(newTrig);
-				}
-			}
-			std::swap(trigs, nextTrigs);
-			nextTrigs.clear();
-		}
-		param.first = (u32)triangles.size();
-		param.count = (u32)trigs.size();
-		triangles.insert(triangles.end(), trigs.begin(), trigs.end());
-	}
-}
-
 void DX11Renderer::configVertexShader()
 {
 	matrices.CalcMatrices(&pvrrc, width, height);
@@ -522,33 +374,12 @@ void DX11Renderer::uploadGeometryBuffers()
 
 	if (config::ModifierVolumes && pvrrc.modtrig.used())
 	{
-		const ModTriangle *data = nullptr;
-		u32 size = 0;
-		std::vector<ModTriangle> modVolTriangles;
-		if (!settings.platform.isNaomi2()) // TODO for naomi2 as well?
-		{
-			// clip triangles
-			modVolTriangles.reserve(pvrrc.modtrig.used());
-			clipModVols(pvrrc.global_param_mvo, modVolTriangles);
-			clipModVols(pvrrc.global_param_mvo_tr, modVolTriangles);
-			if (!modVolTriangles.empty())
-			{
-				size = (u32)(modVolTriangles.size() * sizeof(ModTriangle));
-				data = modVolTriangles.data();
-			}
-		}
-		else
-		{
-			size = pvrrc.modtrig.bytes();
-			data = pvrrc.modtrig.head();
-		}
-		if (size > 0)
-		{
-			verify(ensureBufferSize(modvolBuffer, D3D11_BIND_VERTEX_BUFFER, modvolBufferSize, size));
-			deviceContext->Map(modvolBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedSubres);
-			memcpy(mappedSubres.pData, data, size);
-			deviceContext->Unmap(modvolBuffer, 0);
-		}
+		const ModTriangle *data = pvrrc.modtrig.head();
+		u32 size = pvrrc.modtrig.bytes();
+		verify(ensureBufferSize(modvolBuffer, D3D11_BIND_VERTEX_BUFFER, modvolBufferSize, size));
+		deviceContext->Map(modvolBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedSubres);
+		memcpy(mappedSubres.pData, data, size);
+		deviceContext->Unmap(modvolBuffer, 0);
 	}
     unsigned int stride = sizeof(Vertex);
     unsigned int offset = 0;
@@ -812,7 +643,6 @@ void DX11Renderer::setRenderState(const PolyParam *gp)
 			linearFiltering = false;
 		else
 			linearFiltering = true;
-
         auto sampler = samplers->getSampler(linearFiltering, gp->tsp.ClampU, gp->tsp.ClampV, gp->tsp.FlipU, gp->tsp.FlipV);
         deviceContext->PSSetSamplers(0, 1, &sampler.get());
 	}

core/rend/dx11/dx11_shaders.cpp

@@ -59,18 +59,26 @@ VertexOut main(in VertexIn vin)
 {
 	VertexOut vo;
 	vo.pos = mul(transMatrix, float4(vin.pos.xyz, 1.f));
-#if pp_Gouraud == 1
-	vo.col = vin.col * vo.pos.z;
-	vo.spec = vin.spec * vo.pos.z;
-#else
-	// flat shading: no interpolation
+#if DIV_POS_Z == 1
+	vo.pos /= vo.pos.z;
+	vo.pos.z = vo.pos.w;
+#endif
 	vo.col = vin.col;
 	vo.spec = vin.spec;
+#if pp_Gouraud == 1 && DIV_POS_Z != 1
+	vo.col *= vo.pos.z;
+	vo.spec *= vo.pos.z;
 #endif
-	vo.uv = float4(vin.uv * vo.pos.z, 0.f, vo.pos.z);
+	vo.uv.xyz = float3(vin.uv, 0.f);
 
+#if DIV_POS_Z == 1
+	vo.uv.w = vo.pos.w;
+#else
+	vo.uv.xy *= vo.pos.z;
+	vo.uv.w = vo.pos.z;
 	vo.pos.w = 1.f;
 	vo.pos.z = 0.f;
+#endif
 
 	return vo;
 }
@@ -102,11 +110,15 @@ VertexOut main(in VertexIn vin)
 {
 	VertexOut vo;
 	vo.pos = mul(transMatrix, float4(vin.pos.xyz, 1.f));
+#if DIV_POS_Z == 1
+	vo.pos /= vo.pos.z;
+	vo.pos.z = vo.pos.w;
+	vo.uv = float4(0.f, 0.f, 0.f, vo.pos.w);
+#else
 	vo.uv = float4(0.f, 0.f, 0.f, vo.pos.z);
-
 	vo.pos.w = 1.f;
 	vo.pos.z = 0.f;
-
+#endif
 	return vo;
 }
 
@@ -159,7 +171,13 @@ cbuffer polyConstantBuffer : register(b1)
 
 float fog_mode2(float w)
 {
-	float z = clamp(w * fogDensity, 1.0f, 255.9999f);
+	float z = clamp(
+#if DIV_POS_Z == 1
+					fogDensity / w
+#else
+					fogDensity * w
+#endif
+									, 1.0f, 255.9999f);
 	float exp = floor(log2(z));
 	float m = z * 16.0f / pow(2.0, exp) - 16.0f;
 	float idx = floor(m) + exp * 16.0f + 0.5f;
@@ -201,17 +219,16 @@ PSO main(in Pixel inpix)
 			&& inpix.pos.y >= clipTest.y && inpix.pos.y <= clipTest.w)
 		discard;
 #endif
-#if pp_Gouraud == 1
-	float4 color = inpix.col / inpix.uv.w;
-	#if pp_BumpMap == 1 || pp_Offset == 1
-		float4 specular = inpix.spec / inpix.uv.w;
-	#endif
-#else
 	float4 color = inpix.col;
 	#if pp_BumpMap == 1 || pp_Offset == 1
 		float4 specular = inpix.spec;
 	#endif
-#endif
+	#if pp_Gouraud == 1 && DIV_POS_Z != 1
+		color /= inpix.uv.w;
+		#if pp_BumpMap == 1 || pp_Offset == 1
+			specular /= inpix.uv.w;
+		#endif
+	#endif
 	#if pp_UseAlpha == 0
 		color.a = 1.0f;
 	#endif
@@ -220,7 +237,10 @@ PSO main(in Pixel inpix)
 	#endif
 	#if pp_Texture == 1
 	{
-		float2 uv = inpix.uv.xy / inpix.uv.w;
+		float2 uv = inpix.uv.xy;
+		#if DIV_POS_Z != 1
+			uv /= inpix.uv.w;
+		#endif
 		#if NearestWrapFix == 1
 			uv = min(fmod(uv, 1.f), 0.9997f);
 		#endif
@@ -279,7 +299,11 @@ PSO main(in Pixel inpix)
 	#endif
 
 	PSO pso;
-	float w = inpix.uv.w * 100000.0f;
+#if DIV_POS_Z == 1
+	float w = 100000.0f / inpix.uv.w;
+#else
+	float w = 100000.0f * inpix.uv.w;
+#endif
 	pso.z = log2(1.0f + w) / 34.0f;
 	pso.col = color;
 
@@ -295,7 +319,11 @@ struct MVPixel
 PSO modifierVolume(in MVPixel inpix)
 {
 	PSO pso;
-	float w = inpix.uv.w * 100000.0f;
+#if DIV_POS_Z == 1
+	float w = 100000.0f / inpix.uv.w;
+#else
+	float w = 100000.0f * inpix.uv.w;
+#endif
 	pso.z = log2(1.0f + w) / 34.0f;
 	pso.col = float4(0, 0, 0, 1.f - shadowScale);
 
@@ -354,9 +382,18 @@ float4 main(in VertexIn vin) : SV_Target
 
 const char * const MacroValues[] { "0", "1", "2", "3" };
 
+enum VertexMacroEnum {
+	MacroGouraud,
+	MacroDivPosZ,
+	MacroPositionOnly,
+	MacroTwoVolumes,
+	MacroLightOn,
+};
+
 static D3D_SHADER_MACRO VertexMacros[]
 {
 	{ "pp_Gouraud", "1" },
+	{ "DIV_POS_Z", "0" },
 	{ "POSITION_ONLY", "0" },
 	{ "pp_TwoVolumes", "0" },
 	{ "LIGHT_ON", "1" },
@@ -364,8 +401,7 @@ static D3D_SHADER_MACRO VertexMacros[]
 };
 
 enum PixelMacroEnum {
-	MacroGouraud,
-	MacroTexture,
+	MacroTexture = 2,
 	MacroUseAlpha,
 	MacroIgnoreTexA,
 	MacroShadInstr,
@@ -383,6 +419,7 @@ enum PixelMacroEnum {
 static D3D_SHADER_MACRO PixelMacros[]
 {
 	{ "pp_Gouraud", "1" },
+	{ "DIV_POS_Z", "0" },
 	{ "pp_Texture", "0" },
 	{ "pp_UseAlpha", "0" },
 	{ "pp_IgnoreTexA", "0" },
@@ -403,6 +440,7 @@ const ComPtr<ID3D11PixelShader>& DX11Shaders::getShader(bool pp_Texture, bool pp
 		bool pp_Offset, u32 pp_FogCtrl, bool pp_BumpMap, bool fog_clamping,
 		bool trilinear, bool palette, bool gouraud, bool alphaTest, bool clipInside, bool nearestWrapFix)
 {
+	bool divPosZ = !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
 	const u32 hash = (int)pp_Texture
 			| (pp_UseAlpha << 1)
 			| (pp_IgnoreTexA << 2)
@@ -416,7 +454,8 @@ const ComPtr<ID3D11PixelShader>& DX11Shaders::getShader(bool pp_Texture, bool pp
 			| (gouraud << 12)
 			| (alphaTest << 13)
 			| (clipInside << 14)
-			| (nearestWrapFix << 15);
+			| (nearestWrapFix << 15)
+			| (divPosZ << 16);
 	auto& shader = shaders[hash];
 	if (shader == nullptr)
 	{
@@ -436,6 +475,7 @@ const ComPtr<ID3D11PixelShader>& DX11Shaders::getShader(bool pp_Texture, bool pp
 		PixelMacros[MacroAlphaTest].Definition = MacroValues[alphaTest];
 		PixelMacros[MacroClipInside].Definition = MacroValues[clipInside];
 		PixelMacros[MacroNearestWrapFix].Definition = MacroValues[nearestWrapFix];
+		PixelMacros[MacroDivPosZ].Definition = MacroValues[divPosZ];
 
 		shader = compilePS(PixelShader, "main", PixelMacros);
 		verify(shader != nullptr);
@@ -445,20 +485,22 @@ const ComPtr<ID3D11PixelShader>& DX11Shaders::getShader(bool pp_Texture, bool pp
 
 const ComPtr<ID3D11VertexShader>& DX11Shaders::getVertexShader(bool gouraud, bool naomi2)
 {
-	int index = (int)gouraud | ((int)naomi2 << 1);
+	bool divPosZ = !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
+	int index = (int)gouraud | ((int)naomi2 << 1) | ((int)divPosZ << 2);
 	ComPtr<ID3D11VertexShader>& vertexShader = vertexShaders[index];
 	if (!vertexShader)
 	{
-		VertexMacros[0].Definition = MacroValues[gouraud];
+		VertexMacros[MacroGouraud].Definition = MacroValues[gouraud];
 		if (!naomi2)
 		{
+			VertexMacros[MacroDivPosZ].Definition = MacroValues[divPosZ];
 			vertexShader = compileVS(VertexShader, "main", VertexMacros);
 		}
 		else
 		{
-			VertexMacros[1].Definition = MacroValues[false];
-			VertexMacros[2].Definition = MacroValues[false];
-			VertexMacros[3].Definition = MacroValues[true];
+			VertexMacros[MacroPositionOnly].Definition = MacroValues[false];
+			VertexMacros[MacroTwoVolumes].Definition = MacroValues[false];
+			VertexMacros[MacroLightOn].Definition = MacroValues[true];
 			std::string source(DX11N2VertexShader);
 			source += std::string("\n") + DX11N2ColorShader;
 			vertexShader = compileVS(source.c_str(), "main", VertexMacros);
@@ -470,21 +512,26 @@ const ComPtr<ID3D11VertexShader>& DX11Shaders::getVertexShader(bool gouraud, boo
 
 const ComPtr<ID3D11VertexShader>& DX11Shaders::getMVVertexShader(bool naomi2)
 {
-	if (!modVolVertexShaders[naomi2])
+	bool divPosZ = !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
+	int index = (int)naomi2 | ((int)divPosZ << 1);
+	if (!modVolVertexShaders[index])
 	{
 		if (!naomi2)
-			modVolVertexShaders[0] = compileVS(ModVolVertexShader, "main", nullptr);
+		{
+			VertexMacros[MacroDivPosZ].Definition = MacroValues[divPosZ];
+			modVolVertexShaders[index] = compileVS(ModVolVertexShader, "main", VertexMacros);
+		}
 		else
 		{
-			VertexMacros[0].Definition = MacroValues[false];
-			VertexMacros[1].Definition = MacroValues[true];
-			VertexMacros[2].Definition = MacroValues[false];
-			VertexMacros[3].Definition = MacroValues[false];
-			modVolVertexShaders[1] = compileVS(DX11N2VertexShader, "main", VertexMacros);
+			VertexMacros[MacroGouraud].Definition = MacroValues[false];
+			VertexMacros[MacroPositionOnly].Definition = MacroValues[true];
+			VertexMacros[MacroTwoVolumes].Definition = MacroValues[false];
+			VertexMacros[MacroLightOn].Definition = MacroValues[false];
+			modVolVertexShaders[index] = compileVS(DX11N2VertexShader, "main", VertexMacros);
 		}
 	}
 
-	return modVolVertexShaders[naomi2];
+	return modVolVertexShaders[index];
 }
 
 const ComPtr<ID3D11PixelShader>& DX11Shaders::getModVolShader()
@@ -564,10 +611,10 @@ ComPtr<ID3D11PixelShader> DX11Shaders::compilePS(const char* source, const char*
 
 ComPtr<ID3DBlob> DX11Shaders::getVertexShaderBlob()
 {
-	VertexMacros[0].Definition = MacroValues[true];
+	VertexMacros[MacroGouraud].Definition = MacroValues[true];
 	// FIXME code dup
-	VertexMacros[1].Definition = MacroValues[false];
-	VertexMacros[2].Definition = MacroValues[false];
+	VertexMacros[MacroPositionOnly].Definition = MacroValues[false];
+	VertexMacros[MacroTwoVolumes].Definition = MacroValues[false];
 	std::string source(DX11N2VertexShader);
 	source += std::string("\n") + DX11N2ColorShader;
 	return compileShader(source.c_str(), "main", "vs_4_0", VertexMacros);
@@ -576,9 +623,9 @@ ComPtr<ID3DBlob> DX11Shaders::getVertexShaderBlob()
 ComPtr<ID3DBlob> DX11Shaders::getMVVertexShaderBlob()
 {
 	// FIXME code dup
-	VertexMacros[0].Definition = MacroValues[false];
-	VertexMacros[1].Definition = MacroValues[true];
-	VertexMacros[2].Definition = MacroValues[false];
+	VertexMacros[MacroGouraud].Definition = MacroValues[false];
+	VertexMacros[MacroPositionOnly].Definition = MacroValues[true];
+	VertexMacros[MacroTwoVolumes].Definition = MacroValues[false];
 	return compileShader(DX11N2VertexShader, "main", "vs_4_0", VertexMacros);
 }
 

core/rend/dx11/dx11_shaders.h

@@ -70,9 +70,9 @@ private:
 
 	ComPtr<ID3D11Device> device;
 	std::unordered_map<u32, ComPtr<ID3D11PixelShader>> shaders;
-	ComPtr<ID3D11VertexShader> vertexShaders[4];
+	ComPtr<ID3D11VertexShader> vertexShaders[8];
 	ComPtr<ID3D11PixelShader> modVolShader;
-	ComPtr<ID3D11VertexShader> modVolVertexShaders[2];
+	ComPtr<ID3D11VertexShader> modVolVertexShaders[4];
 	ComPtr<ID3D11PixelShader> quadPixelShader;
 	ComPtr<ID3D11VertexShader> quadVertexShader;
 	ComPtr<ID3D11VertexShader> quadRotateVertexShader;

core/rend/dx11/oit/dx11_oitshaders.cpp

@@ -71,24 +71,33 @@ VertexOut main(in VertexIn vin)
 {
 	VertexOut vo;
 	vo.pos = mul(transMatrix, float4(vin.pos.xyz, 1.f));
-#if pp_Gouraud == 1
-	vo.col = vin.col * vo.pos.z;
-	vo.spec = vin.spec * vo.pos.z;
-	vo.col1 = vin.col1 * vo.pos.z;
-	vo.spec1 = vin.spec1 * vo.pos.z;
-#else
-	// flat shading: no interpolation
+#if DIV_POS_Z == 1
+	vo.pos /= vo.pos.z;
+	vo.pos.z = vo.pos.w;
+#endif
 	vo.col = vin.col;
 	vo.spec = vin.spec;
 	vo.col1 = vin.col1;
 	vo.spec1 = vin.spec1;
+#if pp_Gouraud == 1 && DIV_POS_Z != 1
+	vo.col *= vo.pos.z;
+	vo.spec *= vo.pos.z;
+	vo.col1 *= vo.pos.z;
+	vo.spec1 *= vo.pos.z;
 #endif
-	vo.uv = float4(vin.uv * vo.pos.z, 0.f, vo.pos.z);
-	vo.uv1 = vin.uv1 * vo.pos.z;
+	vo.uv.xyz = float3(vin.uv, 0.f);
+	vo.uv1 = vin.uv1;
 	vo.index = uint(polyNumber) + vin.vertexId;
 
+#if DIV_POS_Z == 1
+	vo.uv.w = vo.pos.w;
+#else
+	vo.uv.xy *= vo.pos.z;
+	vo.uv.w = vo.pos.z;
+	vo.uv1 *= vo.pos.z;
 	vo.pos.w = 1.f;
 	vo.pos.z = 0.f;
+#endif
 
 	return vo;
 }
@@ -131,7 +140,11 @@ struct Pixel {
 
 float getFragDepth(float z)
 {
+#if DIV_POS_Z == 1
+	float w = 100000.0 / z;
+#else
 	float w = 100000.0 * z;
+#endif
 	return log2(1.0 + w) / 34.0;
 }
 
@@ -329,7 +342,13 @@ cbuffer polyConstantBuffer : register(b1)
 
 float fog_mode2(float w)
 {
-	float z = clamp(w * fogDensity, 1.0f, 255.9999f);
+	float z = clamp(
+#if DIV_POS_Z == 1
+					fogDensity / w
+#else
+					fogDensity * w
+#endif
+									, 1.0f, 255.9999f);
 	float exp = floor(log2(z));
 	float m = z * 16.0f / pow(2.0, exp) - 16.0f;
 	float idx = floor(m) + exp * 16.0f + 0.5f;
@@ -404,7 +423,7 @@ PSO main(in VertexIn inpix)
 			}
 		#endif
 	#endif
-	#if pp_Gouraud == 1
+	#if pp_Gouraud == 1 && DIV_POS_Z != 1
 		color /= inpix.uv.w;
 		specular /= inpix.uv.w;
 	#endif
@@ -421,10 +440,14 @@ PSO main(in VertexIn inpix)
 		float2 uv;
 		#if pp_TwoVolumes == 1
 			if (area1)
-				uv = inpix.uv1 / inpix.uv.w;
+				uv = inpix.uv1;
 			else
 		#endif
-				uv = inpix.uv.xy / inpix.uv.w;
+				uv = inpix.uv.xy;
+		#if DIV_POS_Z != 1
+			uv /= inpix.uv.w;
+		#endif
+
 		#if NearestWrapFix == 1
 			uv = min(fmod(uv, 1.f), 0.9997f);
 		#endif
@@ -564,7 +587,7 @@ PSO main(in VertexIn inpix)
 		
 		Pixel pixel;
 		pixel.color = packColors(clamp(color, 0.f, 1.f));
-		pixel.depth = inpix.uv.w;
+		pixel.depth = pso.z;
 		pixel.seq_num = inpix.index;
 		InterlockedExchange(abufferPointers[coords], idx, pixel.next);
 		Pixels[idx] = pixel;
@@ -829,18 +852,26 @@ struct IncludeManager : public ID3DInclude
 
 const char * const MacroValues[] { "0", "1", "2", "3" };
 
+enum VertexMacroEnum {
+	MacroGouraud,
+	MacroTwoVolumes,
+	MacroDivPosZ,
+	MacroPositionOnly,
+	MacroLightOn,
+};
+
 static D3D_SHADER_MACRO VertexMacros[]
 {
 	{ "pp_Gouraud", "1" },
-	{ "POSITION_ONLY", "0" },
 	{ "pp_TwoVolumes", "0" },
+	{ "DIV_POS_Z", "0" },
+	{ "POSITION_ONLY", "0" },
 	{ "LIGHT_ON", "1" },
 	{ nullptr, nullptr }
 };
 
 enum PixelMacroEnum {
-	MacroGouraud,
-	MacroTexture,
+	MacroTexture = 3,
 	MacroUseAlpha,
 	MacroIgnoreTexA,
 	MacroShadInstr,
@@ -852,13 +883,14 @@ enum PixelMacroEnum {
 	MacroAlphaTest,
 	MacroClipInside,
 	MacroNearestWrapFix,
-	MacroTwoVolumes,
 	MacroPass
 };
 
 static D3D_SHADER_MACRO PixelMacros[]
 {
 	{ "pp_Gouraud", "1" },
+	{ "pp_TwoVolumes", "0" },
+	{ "DIV_POS_Z", "0" },
 	{ "pp_Texture", "0" },
 	{ "pp_UseAlpha", "0" },
 	{ "pp_IgnoreTexA", "0" },
@@ -871,7 +903,6 @@ static D3D_SHADER_MACRO PixelMacros[]
 	{ "cp_AlphaTest", "0" },
 	{ "pp_ClipInside", "0" },
 	{ "NearestWrapFix", "0" },
-	{ "pp_TwoVolumes", "0" },
 	{ "PASS", "0" },
 	{ nullptr, nullptr }
 };
@@ -880,6 +911,7 @@ const ComPtr<ID3D11PixelShader>& DX11OITShaders::getShader(bool pp_Texture, bool
 		bool pp_Offset, u32 pp_FogCtrl, bool pp_BumpMap, bool fog_clamping,
 		bool palette, bool gouraud, bool alphaTest, bool clipInside, bool nearestWrapFix, bool twoVolumes, Pass pass)
 {
+	bool divPosZ = !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
 	const u32 hash = (int)pp_Texture
 			| (pp_UseAlpha << 1)
 			| (pp_IgnoreTexA << 2)
@@ -894,7 +926,8 @@ const ComPtr<ID3D11PixelShader>& DX11OITShaders::getShader(bool pp_Texture, bool
 			| (clipInside << 13)
 			| (nearestWrapFix << 14)
 			| (twoVolumes << 15)
-			| (pass << 16);
+			| (pass << 16)
+			| (divPosZ << 18);
 	auto& shader = shaders[hash];
 	if (shader == nullptr)
 	{
@@ -915,6 +948,7 @@ const ComPtr<ID3D11PixelShader>& DX11OITShaders::getShader(bool pp_Texture, bool
 		PixelMacros[MacroClipInside].Definition = MacroValues[clipInside];
 		PixelMacros[MacroNearestWrapFix].Definition = MacroValues[nearestWrapFix];
 		PixelMacros[MacroTwoVolumes].Definition = MacroValues[twoVolumes];
+		PixelMacros[MacroDivPosZ].Definition = MacroValues[divPosZ];
 		PixelMacros[MacroPass].Definition = MacroValues[pass];
 
 		shader = compilePS(PixelShader, "main", PixelMacros);
@@ -925,24 +959,28 @@ const ComPtr<ID3D11PixelShader>& DX11OITShaders::getShader(bool pp_Texture, bool
 
 const ComPtr<ID3D11VertexShader>& DX11OITShaders::getVertexShader(bool gouraud, bool naomi2, bool positionOnly, bool lightOn, bool twoVolumes)
 {
+	bool divPosZ = !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
 	const u32 hash = (int)gouraud
 			| ((int)naomi2 << 1)
 			| ((int)positionOnly << 2)
 			| ((int)lightOn << 3)
-			| ((int)twoVolumes << 4);
+			| ((int)twoVolumes << 4)
+			| ((int)divPosZ << 5);
 	auto& shader = vertexShaders[hash];
 	if (shader == nullptr)
 	{
-		VertexMacros[0].Definition = MacroValues[gouraud];
+		VertexMacros[MacroGouraud].Definition = MacroValues[gouraud];
 		if (!naomi2)
 		{
+			VertexMacros[MacroDivPosZ].Definition = MacroValues[divPosZ];
 			shader = compileVS(VertexShader, "main", VertexMacros);
 		}
 		else
 		{
-			VertexMacros[1].Definition = MacroValues[positionOnly];
-			VertexMacros[2].Definition = MacroValues[twoVolumes];
-			VertexMacros[3].Definition = MacroValues[lightOn];
+			VertexMacros[MacroDivPosZ].Definition = MacroValues[false];
+			VertexMacros[MacroPositionOnly].Definition = MacroValues[positionOnly];
+			VertexMacros[MacroTwoVolumes].Definition = MacroValues[twoVolumes];
+			VertexMacros[MacroLightOn].Definition = MacroValues[lightOn];
 			std::string source(DX11N2VertexShader);
 			if (!positionOnly && lightOn)
 				source += std::string("\n") + DX11N2ColorShader;
@@ -955,27 +993,37 @@ const ComPtr<ID3D11VertexShader>& DX11OITShaders::getVertexShader(bool gouraud,
 
 const ComPtr<ID3D11VertexShader>& DX11OITShaders::getMVVertexShader(bool naomi2)
 {
-	if (!modVolVertexShaders[naomi2])
+	bool divPosZ = !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
+	auto& mvVertexShader = modVolVertexShaders[(int)naomi2 | ((int)divPosZ << 1)];
+	if (!mvVertexShader)
 	{
 		if (!naomi2)
-			modVolVertexShaders[0] = compileVS(ModVolVertexShader, "main", nullptr);
+		{
+			VertexMacros[MacroDivPosZ].Definition = MacroValues[divPosZ];
+			mvVertexShader = compileVS(ModVolVertexShader, "main", VertexMacros);
+		}
 		else
 		{
-			VertexMacros[0].Definition = MacroValues[false];
-			VertexMacros[1].Definition = MacroValues[true];
-			VertexMacros[2].Definition = MacroValues[false];
-			VertexMacros[3].Definition = MacroValues[false];
-			modVolVertexShaders[1] = compileVS(DX11N2VertexShader, "main", VertexMacros);
+			VertexMacros[MacroGouraud].Definition = MacroValues[false];
+			VertexMacros[MacroPositionOnly].Definition = MacroValues[true];
+			VertexMacros[MacroTwoVolumes].Definition = MacroValues[false];
+			VertexMacros[MacroLightOn].Definition = MacroValues[false];
+			mvVertexShader = compileVS(DX11N2VertexShader, "main", VertexMacros);
 		}
 	}
 
-	return modVolVertexShaders[naomi2];
+	return mvVertexShader;
 }
 
 const ComPtr<ID3D11PixelShader>& DX11OITShaders::getModVolShader()
 {
+	bool divPosZ = !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
+	auto& modVolShader = modVolShaders[divPosZ];
 	if (!modVolShader)
+	{
+		PixelMacros[MacroDivPosZ].Definition = MacroValues[divPosZ];
 		modVolShader = compilePS(PixelShader, "modifierVolume", PixelMacros);
+	}
 
 	return modVolShader;
 }
@@ -1082,10 +1130,10 @@ ComPtr<ID3D11PixelShader> DX11OITShaders::compilePS(const char* source, const ch
 
 ComPtr<ID3DBlob> DX11OITShaders::getVertexShaderBlob()
 {
-	VertexMacros[0].Definition = MacroValues[true];
+	VertexMacros[MacroGouraud].Definition = MacroValues[true];
 	// FIXME code dup
-	VertexMacros[1].Definition = MacroValues[false];
-	VertexMacros[2].Definition = MacroValues[true];
+	VertexMacros[MacroPositionOnly].Definition = MacroValues[false];
+	VertexMacros[MacroTwoVolumes].Definition = MacroValues[true];
 	std::string source(DX11N2VertexShader);
 	source += std::string("\n") + DX11N2ColorShader;
 	return compileShader(source.c_str(), "main", "vs_5_0", VertexMacros);
@@ -1094,9 +1142,9 @@ ComPtr<ID3DBlob> DX11OITShaders::getVertexShaderBlob()
 ComPtr<ID3DBlob> DX11OITShaders::getMVVertexShaderBlob()
 {
 	// FIXME code dup
-	VertexMacros[0].Definition = MacroValues[false];
-	VertexMacros[1].Definition = MacroValues[true];
-	VertexMacros[2].Definition = MacroValues[false];
+	VertexMacros[MacroGouraud].Definition = MacroValues[false];
+	VertexMacros[MacroPositionOnly].Definition = MacroValues[true];
+	VertexMacros[MacroTwoVolumes].Definition = MacroValues[false];
 	return compileShader(DX11N2VertexShader, "main", "vs_5_0", VertexMacros);
 }
 

core/rend/dx11/oit/dx11_oitshaders.h

@@ -35,7 +35,7 @@ public:
 	const ComPtr<ID3D11PixelShader>& getShader(bool pp_Texture, bool pp_UseAlpha, bool pp_IgnoreTexA, u32 pp_ShadInstr,
 			bool pp_Offset, u32 pp_FogCtrl, bool pp_BumpMap, bool fog_clamping,
 			bool palette, bool gouraud, bool alphaTest, bool clipInside, bool nearestWrapFix, bool twoVolumes, Pass pass);
-	const ComPtr<ID3D11VertexShader>& getVertexShader(bool gouraud, bool naomi2, bool positionOnly, bool lightOn, bool twoVolumes = true);
+	const ComPtr<ID3D11VertexShader>& getVertexShader(bool gouraud, bool naomi2, bool positionOnly, bool lightOn, bool twoVolumes = false);
 	const ComPtr<ID3D11PixelShader>& getModVolShader();
 	const ComPtr<ID3D11VertexShader>& getMVVertexShader(bool naomi2);
 	const ComPtr<ID3D11PixelShader>& getFinalShader();
@@ -47,9 +47,10 @@ public:
 		saveCache(CacheFile);
 		shaders.clear();
 		vertexShaders.clear();
-		modVolShader.reset();
 		for (auto& shader : modVolVertexShaders)
 			shader.reset();
+		for (auto& shader : modVolShaders)
+			shader.reset();
 		for (auto& shader : trModVolShaders)
 			shader.reset();
 		finalShader.reset();
@@ -69,8 +70,8 @@ private:
 	ComPtr<ID3D11Device> device;
 	std::unordered_map<u32, ComPtr<ID3D11PixelShader>> shaders;
 	std::unordered_map<u32, ComPtr<ID3D11VertexShader>> vertexShaders;
-	ComPtr<ID3D11PixelShader> modVolShader;
-	ComPtr<ID3D11VertexShader> modVolVertexShaders[2];
+	ComPtr<ID3D11PixelShader> modVolShaders[2];
+	ComPtr<ID3D11VertexShader> modVolVertexShaders[4];
 
 	ComPtr<ID3D11PixelShader> trModVolShaders[4];
 	ComPtr<ID3D11PixelShader> finalShader;

core/rend/dx9/d3d_shaders.cpp

@@ -17,6 +17,7 @@
     along with Flycast.  If not, see <https://www.gnu.org/licenses/>.
 */
 #include "d3d_shaders.h"
+#include "cfg/option.h"
 
 #define SHADER_DEBUG 0 // D3DXSHADER_DEBUG|D3DXSHADER_SKIPOPTIMIZATION
 
@@ -43,18 +44,23 @@ VertexOut main(in VertexIn vin)
 {
 	VertexOut vo;
 	vo.pos = mul(transMatrix, float4(vin.pos.xyz, 1.f));
-#if pp_Gouraud == 1
-	vo.col = vin.col * vo.pos.z;
-	vo.spec = vin.spec * vo.pos.z;
-#else
-	// flat shading: no interpolation
+#if DIV_POS_Z == 1
+	vo.pos /= vo.pos.z;
+	vo.pos.z = vo.pos.w;
+#endif
 	vo.col = vin.col;
 	vo.spec = vin.spec;
+#if pp_Gouraud == 1 && DIV_POS_Z != 1
+	vo.col *= vo.pos.z;
+	vo.spec *= vo.pos.z;
 #endif
-	vo.uv = float4(vin.uv * vo.pos.z, 0.f, vo.pos.z);
+	vo.uv = float4(vin.uv, 0.f, vo.pos.z);
 
+#if DIV_POS_Z != 1
+	vo.uv.xy *= vo.pos.z;
 	vo.pos.w = 1.f;
 	vo.pos.z = 0.f;
+#endif
 
 	return vo;
 }
@@ -75,7 +81,7 @@ struct pixel
 #endif
 	
 };
- 
+
 sampler2D samplr : register(s0);
 sampler2D tex_pal : register(s1);
 sampler2D fog_table : register(s2);
@@ -91,7 +97,13 @@ float4 colorClampMax : register(c7);
 
 float fog_mode2(float w)
 {
-	float z = clamp(w * FOG_DENSITY_SCALE.x, 1.0f, 255.9999f);
+	float z = clamp(
+#if DIV_POS_Z == 1
+					FOG_DENSITY_SCALE.x / w
+#else
+					FOG_DENSITY_SCALE.x * w
+#endif
+											, 1.0f, 255.9999f);
 	float exp = floor(log2(z));
 	float m = z * 16.0f / pow(2.0, exp) - 16.0f;
 	float idx = floor(m) + exp * 16.0f + 0.5f;
@@ -112,7 +124,12 @@ float4 clampColor(float4 color)
 
 float4 palettePixel(float4 coords)
 {
-	int colorIdx = int(floor(tex2Dproj(samplr, coords).a * 255.0f + 0.5f) + paletteIndex.x);
+#if DIV_POS_Z == 1
+	float texColIdx = tex2D(samplr, coords.xy).a;
+#else
+	float texColIdx = tex2Dproj(samplr, coords).a;
+#endif
+	int colorIdx = int(floor(texColIdx * 255.0f + 0.5f) + paletteIndex.x);
     float2 c = float2((fmod(float(colorIdx), 32.0f) * 2.0f + 1.0f) / 64.0f, (float(colorIdx / 32) * 2.0f + 1.0f) / 64.0f);
 	return tex2D(tex_pal, c);
 }
@@ -134,17 +151,16 @@ PSO main(in pixel inpix)
 		discard;
 #endif
 
-#if pp_Gouraud == 1
-	float4 color = inpix.col / inpix.uv.w;
-	#if pp_BumpMap == 1 || pp_Offset == 1
-		float4 specular = inpix.spec / inpix.uv.w;
-	#endif
-#else
 	float4 color = inpix.col;
 	#if pp_BumpMap == 1 || pp_Offset == 1
 		float4 specular = inpix.spec;
 	#endif
-#endif
+	#if pp_Gouraud == 1 && DIV_POS_Z != 1
+		color /= inpix.uv.w;
+		#if pp_BumpMap == 1 || pp_Offset == 1
+			specular /= inpix.uv.w;
+		#endif
+	#endif
 	#if pp_UseAlpha == 0
 		color.a = 1.0f;
 	#endif
@@ -154,7 +170,11 @@ PSO main(in pixel inpix)
 	#if pp_Texture == 1
 	{
 		#if pp_Palette == 0
-			float4 texcol = tex2Dproj(samplr, inpix.uv);
+			#if DIV_POS_Z == 1
+				float4 texcol = tex2D(samplr, inpix.uv.xy);
+			#else
+				float4 texcol = tex2Dproj(samplr, inpix.uv);
+			#endif
 		#else
 			float4 texcol = palettePixel(inpix.uv);
 		#endif
@@ -204,7 +224,11 @@ PSO main(in pixel inpix)
 
 	//color.rgb = float3(inpix.uv.w * FOG_DENSITY_SCALE.x / 128.0f);
 	PSO pso;
-	float w = inpix.uv.w * 100000.0f;
+#if DIV_POS_Z == 1
+	float w = 100000.0f / inpix.uv.w;
+#else
+	float w = 100000.0f * inpix.uv.w;
+#endif
 	pso.z = log2(1.0f + w) / 34.0f;
 	pso.col = color;
 
@@ -214,7 +238,11 @@ PSO main(in pixel inpix)
 PSO modifierVolume(float4 uv : TEXCOORD0)
 {
 	PSO pso;
-	float w = uv.w * 100000.0f;
+#if DIV_POS_Z == 1
+	float w = 100000.0f / uv.w;
+#else
+	float w = 100000.0f * uv.w;
+#endif
 	pso.z = log2(1.0f + w) / 34.0f;
 	pso.col = float4(0, 0, 0, FOG_DENSITY_SCALE.y);
 
@@ -227,24 +255,30 @@ const char * const MacroValues[] { "0", "1", "2", "3" };
 static D3DXMACRO VertexMacros[]
 {
 	{ "pp_Gouraud", "1" },
+	{ "DIV_POS_Z", "0" },
 	{ 0, 0 }
 };
 
-constexpr u32 MacroTexture = 0;
-constexpr u32 MacroOffset = 1;
-constexpr u32 MacroShadInstr = 2;
-constexpr u32 MacroIgnoreTexA = 3;
-constexpr u32 MacroUseAlpha = 4;
-constexpr u32 MacroFogCtrl = 5;
-constexpr u32 MacroFogClamping = 6;
-constexpr u32 MacroPalette = 7;
-constexpr u32 MacroBumpMap = 8;
-constexpr u32 MacroTriLinear = 9;
-constexpr u32 MacroGouraud = 10;
-constexpr u32 MacroClipInside = 11;
+enum ShaderMacros {
+	MacroGouraud,
+	MacroDivPosZ,
+	MacroTexture,
+	MacroOffset,
+	MacroShadInstr,
+	MacroIgnoreTexA,
+	MacroUseAlpha,
+	MacroFogCtrl,
+	MacroFogClamping,
+	MacroPalette,
+	MacroBumpMap,
+	MacroTriLinear,
+	MacroClipInside,
+};
 
 static D3DXMACRO PixelMacros[]
 {
+	{ "pp_Gouraud", "1" },
+	{ "DIV_POS_Z", "0" },
 	{ "pp_Texture", "0" },
 	{ "pp_Offset", "0" },
 	{ "pp_ShadInstr", "0" },
@@ -255,7 +289,6 @@ static D3DXMACRO PixelMacros[]
 	{ "pp_Palette", "0" },
 	{ "pp_BumpMap", "0" },
 	{ "pp_TriLinear", "0" },
-	{ "pp_Gouraud", "1" },
 	{ "pp_ClipInside", "0" },
 	{0, 0}
 };
@@ -275,7 +308,8 @@ const ComPtr<IDirect3DPixelShader9>& D3DShaders::getShader(bool pp_Texture, bool
 			| (trilinear << 10)
 			| (palette << 11)
 			| (gouraud << 12)
-			| (clipInside << 13);
+			| (clipInside << 13)
+			| ((int)config::NativeDepthInterpolation << 14);
 	auto it = shaders.find(hash);
 	if (it == shaders.end())
 	{
@@ -293,6 +327,7 @@ const ComPtr<IDirect3DPixelShader9>& D3DShaders::getShader(bool pp_Texture, bool
 		PixelMacros[MacroPalette].Definition = MacroValues[palette];
 		PixelMacros[MacroGouraud].Definition = MacroValues[gouraud];
 		PixelMacros[MacroClipInside].Definition = MacroValues[clipInside];
+		PixelMacros[MacroDivPosZ].Definition = MacroValues[config::NativeDepthInterpolation];
 		ComPtr<IDirect3DPixelShader9> shader = compilePS(PixelShader, "main", PixelMacros);
 		verify((bool )shader);
 		it = shaders.insert(std::make_pair(hash, shader)).first;
@@ -302,10 +337,11 @@ const ComPtr<IDirect3DPixelShader9>& D3DShaders::getShader(bool pp_Texture, bool
 
 const ComPtr<IDirect3DVertexShader9>& D3DShaders::getVertexShader(bool gouraud)
 {
-	ComPtr<IDirect3DVertexShader9>& vertexShader = gouraud ? gouraudVertexShader : flatVertexShader;
+	ComPtr<IDirect3DVertexShader9>& vertexShader = vertexShaders[(int)gouraud | ((int)config::NativeDepthInterpolation << 1)];
 	if (!vertexShader)
 	{
-		VertexMacros[0].Definition = MacroValues[gouraud];
+		VertexMacros[MacroGouraud].Definition = MacroValues[gouraud];
+		VertexMacros[MacroDivPosZ].Definition = MacroValues[config::NativeDepthInterpolation];
 		vertexShader = compileVS(VertexShader, "main", VertexMacros);
 	}
 
@@ -314,8 +350,12 @@ const ComPtr<IDirect3DVertexShader9>& D3DShaders::getVertexShader(bool gouraud)
 
 const ComPtr<IDirect3DPixelShader9>& D3DShaders::getModVolShader()
 {
+	ComPtr<IDirect3DPixelShader9>& modVolShader = modVolShaders[config::NativeDepthInterpolation];
 	if (!modVolShader)
+	{
+		PixelMacros[MacroDivPosZ].Definition = MacroValues[config::NativeDepthInterpolation];
 		modVolShader = compilePS(PixelShader, "modifierVolume", PixelMacros);
+	}
 
 	return modVolShader;
 }

core/rend/dx9/d3d_shaders.h

@@ -36,9 +36,10 @@ public:
 	const ComPtr<IDirect3DPixelShader9>& getModVolShader();
 	void term() {
 		shaders.clear();
-		gouraudVertexShader.reset();
-		flatVertexShader.reset();
-		modVolShader.reset();
+		for (auto& shader : vertexShaders)
+			shader.reset();
+		for (auto& shader : modVolShaders)
+			shader.reset();
 		device.reset();
 	}
 
@@ -49,7 +50,6 @@ private:
 
 	ComPtr<IDirect3DDevice9> device;
 	std::unordered_map<u32, ComPtr<IDirect3DPixelShader9>> shaders;
-	ComPtr<IDirect3DVertexShader9> gouraudVertexShader;
-	ComPtr<IDirect3DVertexShader9> flatVertexShader;
-	ComPtr<IDirect3DPixelShader9> modVolShader;
+	ComPtr<IDirect3DVertexShader9> vertexShaders[4];
+	ComPtr<IDirect3DPixelShader9> modVolShaders[2];
 };

core/rend/gl4/abuffer.cpp

@@ -198,7 +198,7 @@ void main(void)
 )";
 
 static const char *tr_modvol_shader_source = R"(
-noperspective in vec3 vtx_uv;
+in vec3 vtx_uv;
 
 // Must match ModifierVolumeMode enum values
 #define MV_XOR		 0

core/rend/gl4/gl4.h

@@ -100,6 +100,7 @@ struct gl4PipelineShader
 	bool fog_clamping;
 	bool palette;
 	bool naomi2;
+	bool divPosZ;
 };
 
 

core/rend/gl4/gl4naomi2.cpp

@@ -25,14 +25,14 @@ static const char *gouraudSource = R"(
 #if pp_Gouraud == 0
 #define INTERPOLATION flat
 #else
-#define INTERPOLATION noperspective
+#define INTERPOLATION
 #endif
-#define NOPERSPECTIVE noperspective
 )";
 
 N2Vertex4Source::N2Vertex4Source(const gl4PipelineShader* shader) : OpenGl4Source()
 {
 	addConstant("OIT_RENDER");
+	addConstant("DIV_POS_Z", false);
 	if (shader == nullptr)
 	{
 		addConstant("POSITION_ONLY", 1);

core/rend/gl4/gldraw.cpp

@@ -54,6 +54,7 @@ static gl4PipelineShader *gl4GetProgram(bool cp_AlphaTest, bool pp_InsideClippin
 	rv <<= 1; rv |= (int)palette;
 	rv <<= 1; rv |= (int)naomi2;
 	rv <<= 2; rv |= (int)pass;
+	rv <<= 1; rv |= (int)(!settings.platform.isNaomi2() && config::NativeDepthInterpolation);
 
 	gl4PipelineShader *shader = &gl4.shaders[rv];
 	if (shader->program == 0)
@@ -73,6 +74,7 @@ static gl4PipelineShader *gl4GetProgram(bool cp_AlphaTest, bool pp_InsideClippin
 		shader->palette = palette;
 		shader->naomi2 = naomi2;
 		shader->pass = pass;
+		shader->divPosZ = !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
 		gl4CompilePipelineShader(shader);
 	}
 

core/rend/gl4/gles.cpp

@@ -55,7 +55,7 @@ static const char* VertexShaderSource = R"(
 #if pp_Gouraud == 0
 #define INTERPOLATION flat
 #else
-#define INTERPOLATION noperspective
+#define INTERPOLATION
 #endif
 
 // Uniforms 
@@ -74,31 +74,39 @@ in vec2 in_uv1;
 // Output
 INTERPOLATION out vec4 vtx_base;
 INTERPOLATION out vec4 vtx_offs;
-noperspective out vec3 vtx_uv;
+out vec3 vtx_uv;
 INTERPOLATION out vec4 vtx_base1;
 INTERPOLATION out vec4 vtx_offs1;
-noperspective out vec2 vtx_uv1;
+out vec2 vtx_uv1;
 flat out uint vtx_index;
 
 void main()
 {
 	vec4 vpos = ndcMat * in_pos;
+	#if DIV_POS_Z == 1
+		vpos /= vpos.z;
+		vpos.z = vpos.w;
+	#endif
 	vtx_base = in_base;
 	vtx_offs = in_offs;
-	vtx_uv = vec3(in_uv * vpos.z, vpos.z);
+	vtx_uv = vec3(in_uv, vpos.z);
 	vtx_base1 = in_base1;
 	vtx_offs1 = in_offs1;
-	vtx_uv1 = in_uv1 * vpos.z;
+	vtx_uv1 = in_uv1;
 	vtx_index = uint(pp_Number) + uint(gl_VertexID);
-#if pp_Gouraud == 1
-	vtx_base *= vpos.z;
-	vtx_offs *= vpos.z;
-	vtx_base1 *= vpos.z;
-	vtx_offs1 *= vpos.z;
-#endif
+	#if pp_Gouraud == 1 && DIV_POS_Z != 1
+		vtx_base *= vpos.z;
+		vtx_offs *= vpos.z;
+		vtx_base1 *= vpos.z;
+		vtx_offs1 *= vpos.z;
+	#endif
 	
-	vpos.w = 1.0;
-	vpos.z = 0.0;
+	#if DIV_POS_Z != 1
+		vtx_uv.xy *= vpos.z;
+		vtx_uv1 *= vpos.z;
+		vpos.w = 1.0;
+		vpos.z = 0.0;
+	#endif
 	gl_Position = vpos;
 }
 )";
@@ -123,7 +131,7 @@ out vec4 FragColor;
 #if pp_Gouraud == 0
 #define INTERPOLATION flat
 #else
-#define INTERPOLATION noperspective
+#define INTERPOLATION
 #endif
 
 // Uniforms
@@ -155,15 +163,21 @@ uniform int fog_control[2];
 // Input
 INTERPOLATION in vec4 vtx_base;
 INTERPOLATION in vec4 vtx_offs;
-noperspective in vec3 vtx_uv;
+in vec3 vtx_uv;
 INTERPOLATION in vec4 vtx_base1;
 INTERPOLATION in vec4 vtx_offs1;
-noperspective in vec2 vtx_uv1;
+in vec2 vtx_uv1;
 flat in uint vtx_index;
 
 float fog_mode2(float w)
 {
-	float z = clamp(w * sp_FOG_DENSITY, 1.0, 255.9999);
+	float z = clamp(
+#if DIV_POS_Z == 1
+					sp_FOG_DENSITY / w
+#else
+					sp_FOG_DENSITY * w
+#endif
+									  , 1.0, 255.9999);
 	float exp = floor(log2(z));
 	float m = z * 16.0 / pow(2.0, exp) - 16.0;
 	float idx = floor(m) + exp * 16.0 + 0.5;
@@ -184,7 +198,12 @@ vec4 fog_clamp(vec4 col)
 
 vec4 palettePixel(sampler2D tex, vec3 coords)
 {
-	int color_idx = int(floor(textureProj(tex, coords).r * 255.0 + 0.5)) + palette_index;
+#if DIV_POS_Z == 1
+	float colIdx = texture(tex, coords.xy).r;
+#else
+	float colIdx = textureProj(tex, coords).r;
+#endif
+	int color_idx = int(floor(colIdx * 255.0 + 0.5)) + palette_index;
 	ivec2 c = ivec2(color_idx % 32, color_idx / 32);
 	return texelFetch(palette, c, 0);
 }
@@ -233,7 +252,7 @@ void main()
 			}
 		#endif
 	#endif
-	#if pp_Gouraud == 1
+	#if pp_Gouraud == 1 && DIV_POS_Z != 1
 		color /= vtx_uv.z;
 		offset /= vtx_uv.z;
 	#endif
@@ -250,10 +269,17 @@ void main()
 	{
 		vec4 texcol;
 		#if pp_Palette == 0
-			if (area1)
-				texcol = textureProj(tex1, vec3(vtx_uv1.xy, vtx_uv.z));
-			else
-				texcol = textureProj(tex0, vtx_uv);
+			#if DIV_POS_Z == 1
+				if (area1)
+					texcol = texture(tex1, vtx_uv1);
+				else
+					texcol = texture(tex0, vtx_uv.xy);
+			#else
+				if (area1)
+					texcol = textureProj(tex1, vec3(vtx_uv1.xy, vtx_uv.z));
+				else
+					texcol = textureProj(tex0, vtx_uv);
+			#endif
 		#else
 			if (area1)
 				texcol = palettePixel(tex1, vec3(vtx_uv1.xy, vtx_uv.z));
@@ -415,7 +441,7 @@ void main()
 )";
 
 static const char* ModifierVolumeShader = R"(
-noperspective in vec3 vtx_uv;
+in vec3 vtx_uv;
 
 void main()
 {
@@ -426,8 +452,9 @@ void main()
 class Vertex4Source : public OpenGl4Source
 {
 public:
-	Vertex4Source(bool gouraud) : OpenGl4Source() {
+	Vertex4Source(bool gouraud, bool divPosZ) : OpenGl4Source() {
 		addConstant("pp_Gouraud", gouraud);
+		addConstant("DIV_POS_Z", divPosZ);
 
 		addSource(VertexShaderSource);
 	}
@@ -453,6 +480,7 @@ public:
 		addConstant("pp_Palette", s->palette);
 		addConstant("NOUVEAU", gl.mesa_nouveau);
 		addConstant("PASS", (int)s->pass);
+		addConstant("DIV_POS_Z", s->divPosZ);
 
 		addSource(ShaderHeader);
 		addSource(gl4PixelPipelineShader);
@@ -472,7 +500,7 @@ bool gl4CompilePipelineShader(gl4PipelineShader* s, const char *fragment_source
 	if (s->naomi2)
 		vertexSource = N2Vertex4Source(s).generate();
 	else
-		vertexSource = Vertex4Source(s->pp_Gouraud).generate();
+		vertexSource = Vertex4Source(s->pp_Gouraud, s->divPosZ).generate();
 	Fragment4ShaderSource fragmentSource(s);
 
 	s->program = gl_CompileAndLink(vertex_source != nullptr ? vertex_source : vertexSource.c_str(),
@@ -588,15 +616,17 @@ static void create_modvol_shader()
 {
 	if (gl4.modvol_shader.program != 0)
 		return;
-	Vertex4Source vertexShader(false);
+	Vertex4Source vertexShader(false, config::NativeDepthInterpolation);
 	OpenGl4Source fragmentShader;
-	fragmentShader.addSource(ShaderHeader)
+	fragmentShader.addConstant("DIV_POS_Z", config::NativeDepthInterpolation)
+		.addSource(ShaderHeader)
 		.addSource(ModifierVolumeShader);
 
 	gl4.modvol_shader.program = gl_CompileAndLink(vertexShader.generate().c_str(), fragmentShader.generate().c_str());
 	gl4.modvol_shader.ndcMat = glGetUniformLocation(gl4.modvol_shader.program, "ndcMat");
 
 	N2Vertex4Source n2VertexShader;
+	fragmentShader.setConstant("DIV_POS_Z", false);
 	gl4.n2ModVolShader.program = gl_CompileAndLink(n2VertexShader.generate().c_str(), fragmentShader.generate().c_str());
 	gl4.n2ModVolShader.ndcMat = glGetUniformLocation(gl4.n2ModVolShader.program, "ndcMat");
 	gl4.n2ModVolShader.mvMat = glGetUniformLocation(gl4.n2ModVolShader.program, "mvMat");
@@ -626,8 +656,6 @@ static bool gl_create_resources()
 		gl4SetupModvolVBO();
 	}
 
-	create_modvol_shader();
-
 	initQuad();
 	glCheck();
 
@@ -757,8 +785,9 @@ static bool RenderFrame(int width, int height)
 	pvrrc.fog_clamp_min.getRGBAColor(gl4ShaderUniforms.fog_clamp_min);
 	pvrrc.fog_clamp_max.getRGBAColor(gl4ShaderUniforms.fog_clamp_max);
 	
-	if (config::Fog)
+	if (config::ModifierVolumes)
 	{
+		create_modvol_shader();
 		glcache.UseProgram(gl4.modvol_shader.program);
 		glUniformMatrix4fv(gl4.modvol_shader.ndcMat, 1, GL_FALSE, &gl4ShaderUniforms.ndcMat[0][0]);
 

core/rend/gl4/glsl.h

@@ -40,7 +40,11 @@ struct Pixel { \n\
  \n\
 void setFragDepth(float z) \n\
 { \n\
+#if DIV_POS_Z == 1 \n\
+	float w = 100000.0 / z; \n\
+#else \n\
 	float w = 100000.0 * z; \n\
+#endif \n\
 	gl_FragDepth = log2(1.0 + w) / 34.0; \n\
 } \n\
 \n\

core/rend/gles/gles.cpp

@@ -69,22 +69,9 @@ const char *PixelCompatShader = R"(
 )";
 
 const char* GouraudSource = R"(
-#if TARGET_GL == GL3
-	#define NOPERSPECTIVE noperspective
-	#if pp_Gouraud == 0
-		#define INTERPOLATION flat
-	#else
-		#define INTERPOLATION noperspective
-	#endif
-#elif TARGET_GL == GLES3
-	#define NOPERSPECTIVE
-	#if pp_Gouraud == 0
-		#define INTERPOLATION flat
-	#else
-		#define INTERPOLATION
-	#endif
+#if (TARGET_GL == GL3 || TARGET_GL == GLES3) && pp_Gouraud == 0
+	#define INTERPOLATION flat
 #else
-	#define NOPERSPECTIVE
 	#define INTERPOLATION
 #endif
 )";
@@ -103,7 +90,7 @@ in highp vec2 in_uv;
 /* output */
 INTERPOLATION out highp vec4 vtx_base;
 INTERPOLATION out highp vec4 vtx_offs;
-NOPERSPECTIVE out highp vec3 vtx_uv;
+out highp vec3 vtx_uv;
 
 void main()
 {
@@ -116,13 +103,20 @@ void main()
 	vpos.z = depth_scale.x + depth_scale.y * vpos.w;
 	vpos.xy *= vpos.w;
 #else
-#if pp_Gouraud == 1
-	vtx_base *= vpos.z;
-	vtx_offs *= vpos.z;
-#endif
-	vtx_uv = vec3(in_uv * vpos.z, vpos.z);
-	vpos.w = 1.0;
-	vpos.z = 0.0;
+	#if DIV_POS_Z == 1
+		vpos /= vpos.z;
+		vpos.z = vpos.w;
+	#endif
+	#if pp_Gouraud == 1 && DIV_POS_Z != 1
+		vtx_base *= vpos.z;
+		vtx_offs *= vpos.z;
+	#endif
+	vtx_uv = vec3(in_uv, vpos.z);
+	#if DIV_POS_Z != 1
+		vtx_uv.xy *= vpos.z;
+		vpos.w = 1.0;
+		vpos.z = 0.0;
+	#endif
 #endif
 	gl_Position = vpos;
 }
@@ -149,15 +143,19 @@ uniform mediump int palette_index;
 /* Vertex input*/
 INTERPOLATION in highp vec4 vtx_base;
 INTERPOLATION in highp vec4 vtx_offs;
-NOPERSPECTIVE in highp vec3 vtx_uv;
+in highp vec3 vtx_uv;
 
 lowp float fog_mode2(highp float w)
 {
+	highp float z = clamp(
 #if TARGET_GL == GLES2
-	highp float z = clamp(vtx_uv.z, 1.0, 255.9999);
+						  vtx_uv.z
+#elif DIV_POS_Z == 1
+						  sp_FOG_DENSITY / w
 #else
-	highp float z = clamp(w * sp_FOG_DENSITY, 1.0, 255.9999);
+						  sp_FOG_DENSITY * w
 #endif
+											, 1.0, 255.9999);
 	mediump float exp = floor(log2(z));
 	highp float m = z * 16.0 / pow(2.0, exp) - 16.0;
 	mediump float idx = floor(m) + exp * 16.0 + 0.5;
@@ -178,7 +176,7 @@ highp vec4 fog_clamp(lowp vec4 col)
 
 lowp vec4 palettePixel(highp vec3 coords)
 {
-#if TARGET_GL == GLES2 || TARGET_GL == GL2
+#if TARGET_GL == GLES2 || TARGET_GL == GL2 || DIV_POS_Z == 1
 	highp int color_idx = int(floor(texture(tex, coords.xy).FOG_CHANNEL * 255.0 + 0.5)) + palette_index;
     highp vec2 c = vec2((mod(float(color_idx), 32.0) * 2.0 + 1.0) / 64.0, (float(color_idx / 32) * 2.0 + 1.0) / 64.0);
 	return texture(palette, c);
@@ -208,7 +206,7 @@ void main()
 	
 	highp vec4 color = vtx_base;
 	highp vec4 offset = vtx_offs;
-	#if pp_Gouraud == 1 && TARGET_GL != GLES2
+	#if pp_Gouraud == 1 && TARGET_GL != GLES2 && DIV_POS_Z != 1
 		color /= vtx_uv.z;
 		offset /= vtx_uv.z;
 	#endif
@@ -221,7 +219,7 @@ void main()
 	#if pp_Texture==1
 	{
 		#if pp_Palette == 0
-		  #if TARGET_GL == GLES2 || TARGET_GL == GL2
+		  #if TARGET_GL == GLES2 || TARGET_GL == GL2 || DIV_POS_Z == 1
 			lowp vec4 texcol = texture(tex, vtx_uv.xy);
 		  #else
 			lowp vec4 texcol = textureProj(tex, vtx_uv);
@@ -289,7 +287,11 @@ void main()
 	
 	//color.rgb = vec3(vtx_uv.z * sp_FOG_DENSITY / 128.0);
 #if TARGET_GL != GLES2
-	highp float w = vtx_uv.z * 100000.0;
+#if DIV_POS_Z == 1
+	highp float w = 100000.0 / vtx_uv.z;
+#else
+	highp float w = 100000.0 * vtx_uv.z;
+#endif
 	gl_FragDepth = log2(1.0 + w) / 34.0;
 #endif
 	gl_FragColor = color;
@@ -300,12 +302,16 @@ static const char* ModifierVolumeShader = R"(
 uniform lowp float sp_ShaderColor;
 
 /* Vertex input*/
-NOPERSPECTIVE in highp vec3 vtx_uv;
+in highp vec3 vtx_uv;
 
 void main()
 {
 #if TARGET_GL != GLES2
-	highp float w = vtx_uv.z * 100000.0;
+#if DIV_POS_Z == 1
+	highp float w = 100000.0 / vtx_uv.z;
+#else
+	highp float w = 100000.0 * vtx_uv.z;
+#endif
 	gl_FragDepth = log2(1.0 + w) / 34.0;
 #endif
 	gl_FragColor=vec4(0.0, 0.0, 0.0, sp_ShaderColor);
@@ -635,19 +641,20 @@ PipelineShader *GetProgram(bool cp_AlphaTest, bool pp_InsideClipping,
 	u32 rv=0;
 
 	rv |= pp_InsideClipping;
-	rv<<=1; rv|=cp_AlphaTest;
-	rv<<=1; rv|=pp_Texture;
-	rv<<=1; rv|=pp_UseAlpha;
-	rv<<=1; rv|=pp_IgnoreTexA;
-	rv<<=2; rv|=pp_ShadInstr;
-	rv<<=1; rv|=pp_Offset;
-	rv<<=2; rv|=pp_FogCtrl;
-	rv<<=1; rv|=pp_Gouraud;
-	rv<<=1; rv|=pp_BumpMap;
-	rv<<=1; rv|=fog_clamping;
-	rv<<=1; rv|=trilinear;
-	rv<<=1; rv|=palette;
-	rv<<=1; rv|=naomi2;
+	rv <<= 1; rv |= cp_AlphaTest;
+	rv <<= 1; rv |= pp_Texture;
+	rv <<= 1; rv |= pp_UseAlpha;
+	rv <<= 1; rv |= pp_IgnoreTexA;
+	rv <<= 2; rv |= pp_ShadInstr;
+	rv <<= 1; rv |= pp_Offset;
+	rv <<= 2; rv |= pp_FogCtrl;
+	rv <<= 1; rv |= pp_Gouraud;
+	rv <<= 1; rv |= pp_BumpMap;
+	rv <<= 1; rv |= fog_clamping;
+	rv <<= 1; rv |= trilinear;
+	rv <<= 1; rv |= palette;
+	rv <<= 1; rv |= naomi2;
+	rv <<= 1, rv |= !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
 
 	PipelineShader *shader = &gl.shaders[rv];
 	if (shader->program == 0)
@@ -666,6 +673,7 @@ PipelineShader *GetProgram(bool cp_AlphaTest, bool pp_InsideClipping,
 		shader->trilinear = trilinear;
 		shader->palette = palette;
 		shader->naomi2 = naomi2;
+		shader->divPosZ = !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
 		CompilePipelineShader(shader);
 	}
 
@@ -675,8 +683,9 @@ PipelineShader *GetProgram(bool cp_AlphaTest, bool pp_InsideClipping,
 class VertexSource : public OpenGlSource
 {
 public:
-	VertexSource(bool gouraud) : OpenGlSource() {
+	VertexSource(bool gouraud, bool divPosZ) : OpenGlSource() {
 		addConstant("pp_Gouraud", gouraud);
+		addConstant("DIV_POS_Z", divPosZ);
 
 		addSource(VertexCompatShader);
 		addSource(GouraudSource);
@@ -702,6 +711,7 @@ public:
 		addConstant("FogClamping", s->fog_clamping);
 		addConstant("pp_TriLinear", s->trilinear);
 		addConstant("pp_Palette", s->palette);
+		addConstant("DIV_POS_Z", s->divPosZ);
 
 		addSource(PixelCompatShader);
 		addSource(GouraudSource);
@@ -715,7 +725,7 @@ bool CompilePipelineShader(PipelineShader* s)
 	if (s->naomi2)
 		vertexShader = N2VertexSource(s->pp_Gouraud, false, s->pp_Texture).generate();
 	else
-		vertexShader = VertexSource(s->pp_Gouraud).generate();
+		vertexShader = VertexSource(s->pp_Gouraud, s->divPosZ).generate();
 	FragmentShaderSource fragmentSource(s);
 
 	s->program = gl_CompileAndLink(vertexShader.c_str(), fragmentSource.generate().c_str());
@@ -860,10 +870,11 @@ static void create_modvol_shader()
 {
 	if (gl.modvol_shader.program != 0)
 		return;
-	VertexSource vertexShader(false);
+	VertexSource vertexShader(false, config::NativeDepthInterpolation);
 
 	OpenGlSource fragmentShader;
 	fragmentShader.addConstant("pp_Gouraud", 0)
+			.addConstant("DIV_POS_Z", config::NativeDepthInterpolation)
 			.addSource(PixelCompatShader)
 			.addSource(GouraudSource)
 			.addSource(ModifierVolumeShader);
@@ -876,6 +887,7 @@ static void create_modvol_shader()
 	if (gl.gl_major >= 3)
 	{
 		N2VertexSource n2vertexShader(false, true, false);
+		fragmentShader.setConstant("DIV_POS_Z", false);
 		gl.n2ModVolShader.program = gl_CompileAndLink(n2vertexShader.generate().c_str(), fragmentShader.generate().c_str());
 		gl.n2ModVolShader.ndcMat = glGetUniformLocation(gl.n2ModVolShader.program, "ndcMat");
 		gl.n2ModVolShader.sp_ShaderColor = glGetUniformLocation(gl.n2ModVolShader.program, "sp_ShaderColor");
@@ -885,7 +897,7 @@ static void create_modvol_shader()
 	}
 }
 
-bool gl_create_resources()
+static bool gl_create_resources()
 {
 	if (gl.vbo.geometry != nullptr)
 		// Assume the resources have already been created
@@ -903,7 +915,6 @@ bool gl_create_resources()
 	gl.vbo.idxs = std::unique_ptr<GlBuffer>(new GlBuffer(GL_ELEMENT_ARRAY_BUFFER));
 	gl.vbo.idxs2 = std::unique_ptr<GlBuffer>(new GlBuffer(GL_ELEMENT_ARRAY_BUFFER));
 
-	create_modvol_shader();
 	initQuad();
 
 	return true;
@@ -911,8 +922,6 @@ bool gl_create_resources()
 
 GLuint gl_CompileShader(const char* shader,GLuint type);
 
-bool gl_create_resources();
-
 //setup
 
 #ifndef __APPLE__
@@ -1200,17 +1209,21 @@ bool RenderFrame(int width, int height)
 	pvrrc.fog_clamp_min.getRGBAColor(ShaderUniforms.fog_clamp_min);
 	pvrrc.fog_clamp_max.getRGBAColor(ShaderUniforms.fog_clamp_max);
 	
-	glcache.UseProgram(gl.modvol_shader.program);
-	if (gl.modvol_shader.depth_scale != -1)
-		glUniform4fv(gl.modvol_shader.depth_scale, 1, ShaderUniforms.depth_coefs);
-	glUniformMatrix4fv(gl.modvol_shader.ndcMat, 1, GL_FALSE, &ShaderUniforms.ndcMat[0][0]);
-	glUniform1f(gl.modvol_shader.sp_ShaderColor, 1 - FPU_SHAD_SCALE.scale_factor / 256.f);
+	if (config::ModifierVolumes)
+	{
+		create_modvol_shader();
+		glcache.UseProgram(gl.modvol_shader.program);
+		if (gl.modvol_shader.depth_scale != -1)
+			glUniform4fv(gl.modvol_shader.depth_scale, 1, ShaderUniforms.depth_coefs);
+		glUniformMatrix4fv(gl.modvol_shader.ndcMat, 1, GL_FALSE, &ShaderUniforms.ndcMat[0][0]);
+		glUniform1f(gl.modvol_shader.sp_ShaderColor, 1 - FPU_SHAD_SCALE.scale_factor / 256.f);
 
-	glcache.UseProgram(gl.n2ModVolShader.program);
-	if (gl.n2ModVolShader.depth_scale != -1)
-		glUniform4fv(gl.n2ModVolShader.depth_scale, 1, ShaderUniforms.depth_coefs);
-	glUniformMatrix4fv(gl.n2ModVolShader.ndcMat, 1, GL_FALSE, &ShaderUniforms.ndcMat[0][0]);
-	glUniform1f(gl.n2ModVolShader.sp_ShaderColor, 1 - FPU_SHAD_SCALE.scale_factor / 256.f);
+		glcache.UseProgram(gl.n2ModVolShader.program);
+		if (gl.n2ModVolShader.depth_scale != -1)
+			glUniform4fv(gl.n2ModVolShader.depth_scale, 1, ShaderUniforms.depth_coefs);
+		glUniformMatrix4fv(gl.n2ModVolShader.ndcMat, 1, GL_FALSE, &ShaderUniforms.ndcMat[0][0]);
+		glUniform1f(gl.n2ModVolShader.sp_ShaderColor, 1 - FPU_SHAD_SCALE.scale_factor / 256.f);
+	}
 
 	ShaderUniforms.PT_ALPHA=(PT_ALPHA_REF&0xFF)/255.0f;
 

core/rend/gles/gles.h

@@ -110,6 +110,7 @@ struct PipelineShader
 	bool trilinear;
 	bool palette;
 	bool naomi2;
+	bool divPosZ;
 };
 
 class GlBuffer

core/rend/gles/naomi2.cpp

@@ -46,10 +46,10 @@ INTERPOLATION out highp vec4 vtx_offs;
 #if pp_TwoVolumes == 1
 INTERPOLATION out vec4 vtx_base1;
 INTERPOLATION out vec4 vtx_offs1;
-noperspective out vec2 vtx_uv1;
+out vec2 vtx_uv1;
 #endif
 #endif
-NOPERSPECTIVE out highp vec3 vtx_uv;
+out highp vec3 vtx_uv;
 #ifdef OIT_RENDER
 flat out uint vtx_index;
 #endif

core/rend/gui.cpp

@@ -1731,25 +1731,30 @@ static void gui_display_settings()
 
 #ifndef TARGET_IPHONE
 		    	OptionCheckbox("VSync", config::VSync, "Synchronizes the frame rate with the screen refresh rate. Recommended");
-		    	ImGui::Indent();
-		    	if (!config::VSync || !isVulkan(config::RendererType))
+		    	if (isVulkan(config::RendererType))
 		    	{
-			        ImGui::PushItemFlag(ImGuiItemFlags_Disabled, true);
-			        ImGui::PushStyleVar(ImGuiStyleVar_Alpha, ImGui::GetStyle().Alpha * 0.5f);
+			    	ImGui::Indent();
+			    	if (!config::VSync)
+			    	{
+				        ImGui::PushItemFlag(ImGuiItemFlags_Disabled, true);
+				        ImGui::PushStyleVar(ImGuiStyleVar_Alpha, ImGui::GetStyle().Alpha * 0.5f);
+			    	}
+		    		OptionCheckbox("Duplicate frames", config::DupeFrames, "Duplicate frames on high refresh rate monitors (120 Hz and higher)");
+			    	if (!config::VSync)
+			    	{
+				        ImGui::PopItemFlag();
+				        ImGui::PopStyleVar();
+			    	}
+			    	ImGui::Unindent();
 		    	}
-	    		OptionCheckbox("Duplicate frames", config::DupeFrames, "Duplicate frames on high refresh rate monitors (120 Hz and higher)");
-		    	if (!config::VSync || !isVulkan(config::RendererType))
-		    	{
-			        ImGui::PopItemFlag();
-			        ImGui::PopStyleVar();
-		    	}
-		    	ImGui::Unindent();
 #endif
 		    	OptionCheckbox("Show FPS Counter", config::ShowFPS, "Show on-screen frame/sec counter");
 		    	OptionCheckbox("Show VMU In-game", config::FloatVMUs, "Show the VMU LCD screens while in-game");
 		    	OptionCheckbox("Rotate Screen 90°", config::Rotate90, "Rotate the screen 90° counterclockwise");
 		    	OptionCheckbox("Delay Frame Swapping", config::DelayFrameSwapping,
 		    			"Useful to avoid flashing screen or glitchy videos. Not recommended on slow platforms");
+		    	OptionCheckbox("Native Depth Interpolation", config::NativeDepthInterpolation,
+		    			"Helps with texture corruption and depth issues on AMD GPUs. Can also help Intel GPUs in some cases.");
 		    	constexpr int apiCount = 0
 					#ifdef USE_VULKAN
 		    			+ 1

core/rend/vulkan/oit/oit_pipeline.cpp

@@ -140,8 +140,9 @@ void OITPipelineManager::CreatePipeline(u32 listType, bool autosort, const PolyP
 	vk::PipelineDynamicStateCreateInfo pipelineDynamicStateCreateInfo(vk::PipelineDynamicStateCreateFlags(), 2, dynamicStates);
 
 	bool twoVolume = pp.tsp1.full != (u32)-1 || pp.tcw1.full != (u32)-1;
+	bool divPosZ = !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
 	vk::ShaderModule vertex_module = shaderManager->GetVertexShader(
-			OITShaderManager::VertexShaderParams{ pp.pcw.Gouraud == 1, pp.isNaomi2(), pass != Pass::Depth, twoVolume, pp.pcw.Texture == 1 });
+			OITShaderManager::VertexShaderParams{ pp.pcw.Gouraud == 1, pp.isNaomi2(), pass != Pass::Depth, twoVolume, pp.pcw.Texture == 1, divPosZ });
 	OITShaderManager::FragmentShaderParams params = {};
 	params.alphaTest = listType == ListType_Punch_Through;
 	params.bumpmap = pp.tcw.PixelFmt == PixelBumpMap;
@@ -157,6 +158,7 @@ void OITPipelineManager::CreatePipeline(u32 listType, bool autosort, const PolyP
 	params.pass = pass;
 	params.twoVolume = twoVolume;
 	params.palette = gpuPalette;
+	params.divPosZ = divPosZ;
 	vk::ShaderModule fragment_module = shaderManager->GetFragmentShader(params);
 
 	vk::PipelineShaderStageCreateInfo stages[] = {
@@ -440,8 +442,8 @@ void OITPipelineManager::CreateModVolPipeline(ModVolMode mode, int cullMode, boo
 	vk::DynamicState dynamicStates[2] = { vk::DynamicState::eViewport, vk::DynamicState::eScissor };
 	vk::PipelineDynamicStateCreateInfo pipelineDynamicStateCreateInfo(vk::PipelineDynamicStateCreateFlags(), 2, dynamicStates);
 
-	vk::ShaderModule vertex_module = shaderManager->GetModVolVertexShader(naomi2);
-	vk::ShaderModule fragment_module = shaderManager->GetModVolShader();
+	vk::ShaderModule vertex_module = shaderManager->GetModVolVertexShader(OITShaderManager::ModVolShaderParams{ naomi2, !settings.platform.isNaomi2() && config::NativeDepthInterpolation });
+	vk::ShaderModule fragment_module = shaderManager->GetModVolShader(!settings.platform.isNaomi2() && config::NativeDepthInterpolation);
 
 	vk::PipelineShaderStageCreateInfo stages[] = {
 			{ vk::PipelineShaderStageCreateFlags(), vk::ShaderStageFlagBits::eVertex, vertex_module, "main" },
@@ -535,8 +537,9 @@ void OITPipelineManager::CreateTrModVolPipeline(ModVolMode mode, int cullMode, b
 	vk::DynamicState dynamicStates[2] = { vk::DynamicState::eViewport, vk::DynamicState::eScissor };
 	vk::PipelineDynamicStateCreateInfo pipelineDynamicStateCreateInfo(vk::PipelineDynamicStateCreateFlags(), 2, dynamicStates);
 
-	vk::ShaderModule vertex_module = shaderManager->GetModVolVertexShader(naomi2);
-	vk::ShaderModule fragment_module = shaderManager->GetTrModVolShader(mode);
+	bool divPosZ = !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
+	vk::ShaderModule vertex_module = shaderManager->GetModVolVertexShader(OITShaderManager::ModVolShaderParams{ naomi2, divPosZ });
+	vk::ShaderModule fragment_module = shaderManager->GetTrModVolShader(OITShaderManager::TrModVolShaderParams{ mode, divPosZ });
 
 	vk::PipelineShaderStageCreateInfo stages[] = {
 			{ vk::PipelineShaderStageCreateFlags(), vk::ShaderStageFlagBits::eVertex, vertex_module, "main" },

core/rend/vulkan/oit/oit_pipeline.h

@@ -405,12 +405,13 @@ private:
 		}
 		hash |= (pp->isp.ZWriteDis << 20) | (pp->isp.CullMode << 21) | ((autosort ? 6 : pp->isp.DepthMode) << 23);
 		hash |= ((u32)gpuPalette << 26) | ((u32)pass << 27) | ((u32)pp->isNaomi2() << 29);
+		hash |= (u32)(!settings.platform.isNaomi2() && config::NativeDepthInterpolation) << 30;
 
 		return hash;
 	}
 	u32 hash(ModVolMode mode, int cullMode, bool naomi2) const
 	{
-		return ((int)mode << 2) | cullMode | ((u32)naomi2 << 5);
+		return ((int)mode << 2) | cullMode | ((u32)naomi2 << 5) | ((u32)(!settings.platform.isNaomi2() && config::NativeDepthInterpolation) << 6);
 	}
 
 	vk::PipelineVertexInputStateCreateInfo GetMainVertexInputStateCreateInfo(bool full = true) const

core/rend/vulkan/oit/oit_shaders.cpp

@@ -44,30 +44,38 @@ layout (location = 6) in mediump vec2 in_uv1;
 
 layout (location = 0) INTERPOLATION out highp vec4 vtx_base;
 layout (location = 1) INTERPOLATION out highp vec4 vtx_offs;
-layout (location = 2) noperspective out highp vec3 vtx_uv;
+layout (location = 2) out highp vec3 vtx_uv;
 layout (location = 3) INTERPOLATION out highp vec4 vtx_base1;		// New for OIT, only for OP/PT with 2-volume
 layout (location = 4) INTERPOLATION out highp vec4 vtx_offs1;
-layout (location = 5) noperspective out highp vec2 vtx_uv1;
+layout (location = 5) out highp vec2 vtx_uv1;
 layout (location = 6) flat out uint vtx_index;
 
 void main()
 {
 	vec4 vpos = uniformBuffer.ndcMat * in_pos;
+#if DIV_POS_Z == 1
+	vpos /= vpos.z;
+	vpos.z = vpos.w;
+#endif
 	vtx_base = vec4(in_base) / 255.0;
 	vtx_offs = vec4(in_offs) / 255.0;
-	vtx_uv = vec3(in_uv * vpos.z, vpos.z);
+	vtx_uv = vec3(in_uv, vpos.z);
 	vtx_base1 = vec4(in_base1) / 255.0;
 	vtx_offs1 = vec4(in_offs1) / 255.0;
-	vtx_uv1 = in_uv1 * vpos.z;
-#if pp_Gouraud == 1
+	vtx_uv1 = in_uv1;
+#if pp_Gouraud == 1 && DIV_POS_Z != 1
 	vtx_base *= vpos.z;
 	vtx_offs *= vpos.z;
 	vtx_base1 *= vpos.z;
 	vtx_offs1 *= vpos.z;
 #endif
 	vtx_index = uint(pushConstants.polyNumber) + uint(gl_VertexIndex);
+#if DIV_POS_Z != 1
+	vtx_uv.xy *= vpos.z;
+	vtx_uv1 *= vpos.z;
 	vpos.w = 1.0;
 	vpos.z = 0.0;
+#endif
 	gl_Position = vpos;
 }
 )";
@@ -179,10 +187,10 @@ layout (input_attachment_index = 0, set = 0, binding = 5) uniform subpassInput D
 // Vertex input
 layout (location = 0) INTERPOLATION in highp vec4 vtx_base;
 layout (location = 1) INTERPOLATION in highp vec4 vtx_offs;
-layout (location = 2) noperspective in highp vec3 vtx_uv;
+layout (location = 2) in highp vec3 vtx_uv;
 layout (location = 3) INTERPOLATION in highp vec4 vtx_base1;			// new for OIT. Only if 2 vol
 layout (location = 4) INTERPOLATION in highp vec4 vtx_offs1;
-layout (location = 5) noperspective in highp vec2 vtx_uv1;
+layout (location = 5) in highp vec2 vtx_uv1;
 layout (location = 6) flat in uint vtx_index;
 
 #if pp_FogCtrl != 2 || pp_TwoVolumes == 1
@@ -190,7 +198,13 @@ layout (set = 0, binding = 2) uniform sampler2D fog_table;
 
 float fog_mode2(float w)
 {
-	float z = clamp(w * uniformBuffer.sp_FOG_DENSITY, 1.0, 255.9999);
+	float z = clamp(
+#if DIV_POS_Z == 1
+					uniformBuffer.sp_FOG_DENSITY / w
+#else
+					uniformBuffer.sp_FOG_DENSITY * w
+#endif
+													, 1.0, 255.9999);
 	float exp = floor(log2(z));
 	float m = z * 16.0 / pow(2.0, exp) - 16.0;
 	float idx = floor(m) + exp * 16.0 + 0.5;
@@ -213,7 +227,12 @@ vec4 colorClamp(vec4 col)
 
 vec4 palettePixel(sampler2D tex, vec3 coords)
 {
-	vec4 c = vec4(textureProj(tex, coords).r * 255.0 / 1023.0 + pushConstants.palette_index, 0.5, 0.0, 0.0);
+#if DIV_POS_Z == 1
+	float texIdx = texture(tex, coords.xy).r;
+#else
+	float texIdx = textureProj(tex, coords).r;
+#endif
+	vec4 c = vec4(texIdx * 255.0 / 1023.0 + pushConstants.palette_index, 0.5, 0.0, 0.0);
 	return texture(palette, c.xy);
 }
 
@@ -261,7 +280,7 @@ void main()
 			}
 		#endif
 	#endif
-	#if pp_Gouraud == 1
+	#if pp_Gouraud == 1 && DIV_POS_Z != 1
 		color /= vtx_uv.z;
 		offset /= vtx_uv.z;
 	#endif
@@ -280,14 +299,22 @@ void main()
 		#if pp_TwoVolumes == 1
 			if (area1)
 				#if pp_Palette == 0
-					texcol = textureProj(tex1, vec3(vtx_uv1, vtx_uv.z));
+					#if DIV_POS_Z == 1
+						texcol = texture(tex1, vtx_uv1);
+					#else
+						texcol = textureProj(tex1, vec3(vtx_uv1, vtx_uv.z));
+					#endif
 				#else
 					texcol = palettePixel(tex1, vec3(vtx_uv1, vtx_uv.z));
 				#endif
 			else
 		#endif
 		#if pp_Palette == 0
+			#if DIV_POS_Z == 1
+				texcol = texture(tex0, vtx_uv.xy);
+			#else
 				texcol = textureProj(tex0, vtx_uv);
+			#endif
 		#else
 				texcol = palettePixel(tex0, vtx_uv);
 		#endif
@@ -428,7 +455,7 @@ void main()
 		
 		Pixel pixel;
 		pixel.color = packColors(clamp(color, vec4(0.0), vec4(1.0)));
-		pixel.depth = vtx_uv.z;
+		pixel.depth = gl_FragDepth;
 		pixel.seq_num = vtx_index;
 		pixel.next = atomicExchange(abufferPointer.pointers[coords.x + coords.y * uniformBuffer.viewportWidth], idx);
 		PixelBuffer.pixels[idx] = pixel;
@@ -438,7 +465,7 @@ void main()
 )";
 
 static const char OITModifierVolumeShader[] = R"(
-layout (location = 0) noperspective in highp float depth;
+layout (location = 0) in highp float depth;
 
 void main()
 {
@@ -614,7 +641,7 @@ void main(void)
 )";
 
 static const char OITTranslucentModvolShaderSource[] = R"(
-layout (location = 0) noperspective in highp float depth;
+layout (location = 0) in highp float depth;
 
 // Must match ModifierVolumeMode enum values
 #define MV_XOR		 0
@@ -624,6 +651,9 @@ layout (location = 0) noperspective in highp float depth;
 
 void main()
 {
+#if MV_MODE == MV_XOR || MV_MODE == MV_OR
+	setFragDepth(depth);
+#endif
 	ivec2 coords = ivec2(gl_FragCoord.xy);
 	
 	uint idx = abufferPointer.pointers[coords.x + coords.y * uniformBuffer.viewportWidth];
@@ -635,10 +665,10 @@ void main()
 		if (getShadowEnable(pp))
 		{
 #if MV_MODE == MV_XOR
-			if (depth >= pixel.depth)
+			if (gl_FragDepth >= pixel.depth)
 				atomicXor(PixelBuffer.pixels[idx].seq_num, SHADOW_STENCIL);
 #elif MV_MODE == MV_OR
-			if (depth >= pixel.depth)
+			if (gl_FragDepth >= pixel.depth)
 				atomicOr(PixelBuffer.pixels[idx].seq_num, SHADOW_STENCIL);
 #elif MV_MODE == MV_INCLUSION
 			uint prev_val = atomicAnd(PixelBuffer.pixels[idx].seq_num, ~(SHADOW_STENCIL));
@@ -692,10 +722,10 @@ layout (location = 7) in vec3         in_normal;
 
 layout (location = 0) INTERPOLATION out highp vec4 vtx_base;
 layout (location = 1) INTERPOLATION out highp vec4 vtx_offs;
-layout (location = 2) noperspective out highp vec3 vtx_uv;
+layout (location = 2) out highp vec3 vtx_uv;
 layout (location = 3) INTERPOLATION out highp vec4 vtx_base1;
 layout (location = 4) INTERPOLATION out highp vec4 vtx_offs1;
-layout (location = 5) noperspective out highp vec2 vtx_uv1;
+layout (location = 5) out highp vec2 vtx_uv1;
 layout (location = 6) flat out uint vtx_index;
 
 void wDivide(inout vec4 vpos)
@@ -779,7 +809,8 @@ vk::UniqueShaderModule OITShaderManager::compileShader(const VertexShaderParams&
 {
 	VulkanSource src;
 	src.addConstant("pp_Gouraud", (int)params.gouraud)
-			.addSource(GouraudSource);
+		.addConstant("DIV_POS_Z", (int)params.divPosZ)
+		.addSource(GouraudSource);
 	if (params.naomi2)
 		src.addConstant("pp_TwoVolumes", (int)params.twoVolume)
 			.addConstant("LIGHT_ON", (int)params.lightOn)
@@ -807,6 +838,7 @@ vk::UniqueShaderModule OITShaderManager::compileShader(const FragmentShaderParam
 		.addConstant("pp_BumpMap", (int)params.bumpmap)
 		.addConstant("ColorClamping", (int)params.clamping)
 		.addConstant("pp_Palette", (int)params.palette)
+		.addConstant("DIV_POS_Z", (int)params.divPosZ)
 		.addConstant("PASS", (int)params.pass)
 		.addSource(GouraudSource)
 		.addSource(OITShaderHeader)
@@ -836,30 +868,31 @@ vk::UniqueShaderModule OITShaderManager::compileClearShader()
 	return ShaderCompiler::Compile(vk::ShaderStageFlagBits::eFragment, src.generate());
 }
 
-vk::UniqueShaderModule OITShaderManager::compileModVolVertexShader(bool naomi2)
+vk::UniqueShaderModule OITShaderManager::compileShader(const ModVolShaderParams& params)
 {
 	VulkanSource src;
-	if (naomi2)
+	if (params.naomi2)
 		src.addSource(N2ModVolVertexShaderSource);
 	else
-		src.addSource(ModVolVertexShaderSource);
+		src.addConstant("DIV_POS_Z", (int)params.divPosZ)
+			.addSource(ModVolVertexShaderSource);
 	return ShaderCompiler::Compile(vk::ShaderStageFlagBits::eVertex, src.generate());
 }
-vk::UniqueShaderModule OITShaderManager::compileModVolFragmentShader()
+vk::UniqueShaderModule OITShaderManager::compileModVolFragmentShader(bool divPosZ)
 {
 	VulkanSource src;
-	src.addSource(OITShaderHeader)
+	src.addConstant("DIV_POS_Z", (int)divPosZ)
+		.addSource(OITShaderHeader)
 		.addSource(OITModifierVolumeShader);
 	return ShaderCompiler::Compile(vk::ShaderStageFlagBits::eFragment, src.generate());
 }
-void OITShaderManager::compileTrModVolFragmentShader(ModVolMode mode)
+vk::UniqueShaderModule OITShaderManager::compileShader(const TrModVolShaderParams& params)
 {
-	if (trModVolShaders.empty())
-		trModVolShaders.resize((size_t)ModVolMode::Final);
 	VulkanSource src;
 	src.addConstant("MAX_PIXELS_PER_FRAGMENT", config::PerPixelLayers)
-		.addConstant("MV_MODE", (int)mode)
+		.addConstant("MV_MODE", (int)params.mode)
+		.addConstant("DIV_POS_Z", (int)params.divPosZ)
 		.addSource(OITShaderHeader)
 		.addSource(OITTranslucentModvolShaderSource);
-	trModVolShaders[(size_t)mode] = ShaderCompiler::Compile(vk::ShaderStageFlagBits::eFragment, src.generate());
+	return ShaderCompiler::Compile(vk::ShaderStageFlagBits::eFragment, src.generate());
 }

core/rend/vulkan/oit/oit_shaders.h

@@ -35,9 +35,10 @@ public:
 		bool lightOn;
 		bool twoVolume;
 		bool texture;
+		bool divPosZ;
 
 		u32 hash() { return (u32)gouraud | ((u32)naomi2 << 1) | ((u32)lightOn << 2)
-				| ((u32)twoVolume << 3) | ((u32)texture << 4); }
+				| ((u32)twoVolume << 3) | ((u32)texture << 4) | ((u32)divPosZ << 5); }
 	};
 
 	// alpha test, clip test, use alpha, texture, ignore alpha, shader instr, offset, fog, gouraud, bump, clamp
@@ -56,6 +57,7 @@ public:
 		bool clamping;
 		bool twoVolume;
 		bool palette;
+		bool divPosZ;
 		Pass pass;
 
 		u32 hash()
@@ -64,39 +66,39 @@ public:
 				| ((u32)texture << 3) | ((u32)ignoreTexAlpha << 4) | (shaderInstr << 5)
 				| ((u32)offset << 7) | ((u32)fog << 8) | ((u32)gouraud << 10)
 				| ((u32)bumpmap << 11) | ((u32)clamping << 12) | ((u32)twoVolume << 13)
-				| ((u32)palette << 14) | ((int)pass << 15);
+				| ((u32)palette << 14) | ((int)pass << 15) | ((u32)divPosZ << 17);
 		}
 	};
 
+	struct ModVolShaderParams
+	{
+		bool naomi2;
+		bool divPosZ;
+
+		u32 hash() { return (u32)naomi2 | ((u32)divPosZ << 1); }
+	};
+
+	struct TrModVolShaderParams
+	{
+		ModVolMode mode;
+		bool divPosZ;
+
+		u32 hash() { return (u32)mode | ((u32)divPosZ << 3); }
+	};
+
 	vk::ShaderModule GetVertexShader(const VertexShaderParams& params) { return getShader(vertexShaders, params); }
 	vk::ShaderModule GetFragmentShader(const FragmentShaderParams& params) { return getShader(fragmentShaders, params); }
-	vk::ShaderModule GetModVolVertexShader(bool naomi2)
-	{
-		vk::UniqueShaderModule& shader = naomi2 ? n2ModVolVertexShader : modVolVertexShader;
-		if (!shader)
-			shader = compileModVolVertexShader(naomi2);
-		return *shader;
-	}
-	vk::ShaderModule GetModVolShader()
+	vk::ShaderModule GetModVolVertexShader(const ModVolShaderParams& params) { return getShader(modVolVertexShaders, params); }
+
+	vk::ShaderModule GetModVolShader(bool divPosZ)
 	{
+		auto& modVolShader = modVolShaders[divPosZ];
 		if (!modVolShader)
-			modVolShader = compileModVolFragmentShader();
+			modVolShader = compileModVolFragmentShader(divPosZ);
 		return *modVolShader;
 	}
-	vk::ShaderModule GetTrModVolShader(ModVolMode mode)
-	{
-		if (trModVolShaders.empty() || !trModVolShaders[(size_t)mode] || maxLayers != config::PerPixelLayers)
-		{
-			if (maxLayers != config::PerPixelLayers)
-			{
-				trModVolShaders.clear();
-				finalFragmentShader.reset();
-			}
-			compileTrModVolFragmentShader(mode);
-			maxLayers = config::PerPixelLayers;
-		}
-		return *trModVolShaders[(size_t)mode];
-	}
+
+	vk::ShaderModule GetTrModVolShader(const TrModVolShaderParams& params) { return getShader(trModVolShaders, params); }
 
 	vk::ShaderModule GetFinalShader()
 	{
@@ -126,27 +128,27 @@ private:
 	template<typename T>
 	vk::ShaderModule getShader(std::map<u32, vk::UniqueShaderModule>& map, T params)
 	{
-		auto it = map.find(params.hash());
+		u32 h = params.hash();
+		auto it = map.find(h);
 		if (it != map.end())
 			return it->second.get();
-		map[params.hash()] = compileShader(params);
-		return map[params.hash()].get();
+		map[h] = compileShader(params);
+		return map[h].get();
 	}
 	vk::UniqueShaderModule compileShader(const VertexShaderParams& params);
 	vk::UniqueShaderModule compileShader(const FragmentShaderParams& params);
-	vk::UniqueShaderModule compileModVolVertexShader(bool naomi2);
-	vk::UniqueShaderModule compileModVolFragmentShader();
-	void compileTrModVolFragmentShader(ModVolMode mode);
+	vk::UniqueShaderModule compileShader(const ModVolShaderParams& params);
+	vk::UniqueShaderModule compileModVolFragmentShader(bool divPosZ);
+	vk::UniqueShaderModule compileShader(const TrModVolShaderParams& params);
 	vk::UniqueShaderModule compileFinalShader();
 	vk::UniqueShaderModule compileFinalVertexShader();
 	vk::UniqueShaderModule compileClearShader();
 
 	std::map<u32, vk::UniqueShaderModule> vertexShaders;
 	std::map<u32, vk::UniqueShaderModule> fragmentShaders;
-	vk::UniqueShaderModule modVolVertexShader;
-	vk::UniqueShaderModule n2ModVolVertexShader;
-	vk::UniqueShaderModule modVolShader;
-	std::vector<vk::UniqueShaderModule> trModVolShaders;
+	std::map<u32, vk::UniqueShaderModule> modVolVertexShaders;
+	vk::UniqueShaderModule modVolShaders[2];
+	std::map<u32, vk::UniqueShaderModule> trModVolShaders;
 
 	vk::UniqueShaderModule finalVertexShader;
 	vk::UniqueShaderModule finalFragmentShader;

core/rend/vulkan/pipeline.cpp

@@ -138,8 +138,9 @@ void PipelineManager::CreateModVolPipeline(ModVolMode mode, int cullMode, bool n
 	vk::DynamicState dynamicStates[2] = { vk::DynamicState::eViewport, vk::DynamicState::eScissor };
 	vk::PipelineDynamicStateCreateInfo pipelineDynamicStateCreateInfo(vk::PipelineDynamicStateCreateFlags(), 2, dynamicStates);
 
-	vk::ShaderModule vertex_module = shaderManager->GetModVolVertexShader(naomi2);
-	vk::ShaderModule fragment_module = shaderManager->GetModVolShader();
+	ModVolShaderParams shaderParams { naomi2, !settings.platform.isNaomi2() && config::NativeDepthInterpolation };
+	vk::ShaderModule vertex_module = shaderManager->GetModVolVertexShader(shaderParams);
+	vk::ShaderModule fragment_module = shaderManager->GetModVolShader(!settings.platform.isNaomi2() && config::NativeDepthInterpolation);
 
 	vk::PipelineShaderStageCreateInfo stages[] = {
 			{ vk::PipelineShaderStageCreateFlags(), vk::ShaderStageFlagBits::eVertex, vertex_module, "main" },
@@ -241,8 +242,9 @@ void PipelineManager::CreateDepthPassPipeline(int cullMode, bool naomi2)
 	vk::DynamicState dynamicStates[2] = { vk::DynamicState::eViewport, vk::DynamicState::eScissor };
 	vk::PipelineDynamicStateCreateInfo pipelineDynamicStateCreateInfo(vk::PipelineDynamicStateCreateFlags(), 2, dynamicStates);
 
-	vk::ShaderModule vertex_module = shaderManager->GetModVolVertexShader(naomi2);
-	vk::ShaderModule fragment_module = shaderManager->GetModVolShader();
+	ModVolShaderParams shaderParams { naomi2, !settings.platform.isNaomi2() && config::NativeDepthInterpolation };
+	vk::ShaderModule vertex_module = shaderManager->GetModVolVertexShader(shaderParams);
+	vk::ShaderModule fragment_module = shaderManager->GetModVolShader(!settings.platform.isNaomi2() && config::NativeDepthInterpolation);
 
 	vk::PipelineShaderStageCreateInfo stages[] = {
 			{ vk::PipelineShaderStageCreateFlags(), vk::ShaderStageFlagBits::eVertex, vertex_module, "main" },
@@ -392,7 +394,8 @@ void PipelineManager::CreatePipeline(u32 listType, bool sortTriangles, const Pol
 	vk::DynamicState dynamicStates[2] = { vk::DynamicState::eViewport, vk::DynamicState::eScissor };
 	vk::PipelineDynamicStateCreateInfo pipelineDynamicStateCreateInfo(vk::PipelineDynamicStateCreateFlags(), 2, dynamicStates);
 
-	vk::ShaderModule vertex_module = shaderManager->GetVertexShader(VertexShaderParams{ pp.pcw.Gouraud == 1, pp.isNaomi2() });
+	bool divPosZ = !settings.platform.isNaomi2() && config::NativeDepthInterpolation;
+	vk::ShaderModule vertex_module = shaderManager->GetVertexShader(VertexShaderParams { pp.pcw.Gouraud == 1, pp.isNaomi2(), divPosZ });
 	FragmentShaderParams params = {};
 	params.alphaTest = listType == ListType_Punch_Through;
 	params.bumpmap = pp.tcw.PixelFmt == PixelBumpMap;
@@ -407,6 +410,7 @@ void PipelineManager::CreatePipeline(u32 listType, bool sortTriangles, const Pol
 	params.trilinear = pp.pcw.Texture && pp.tsp.FilterMode > 1 && listType != ListType_Punch_Through && pp.tcw.MipMapped == 1;
 	params.useAlpha = pp.tsp.UseAlpha;
 	params.palette = gpuPalette;
+	params.divPosZ = divPosZ;
 	vk::ShaderModule fragment_module = shaderManager->GetFragmentShader(params);
 
 	vk::PipelineShaderStageCreateInfo stages[] = {

core/rend/vulkan/pipeline.h

@@ -275,16 +275,17 @@ private:
 			| (pp->tsp.DstInstr << 17);
 		hash |= (pp->isp.ZWriteDis << 20) | (pp->isp.CullMode << 21) | (pp->isp.DepthMode << 23);
 		hash |= ((u32)sortTriangles << 26) | ((u32)gpuPalette << 27) | ((u32)pp->isNaomi2() << 28);
+		hash |= (u32)(!settings.platform.isNaomi2() && config::NativeDepthInterpolation) << 29;
 
 		return hash;
 	}
 	u32 hash(ModVolMode mode, int cullMode, bool naomi2) const
 	{
-		return ((int)mode << 2) | cullMode | ((int)naomi2 << 5);
+		return ((int)mode << 2) | cullMode | ((int)naomi2 << 5) | ((int)(!settings.platform.isNaomi2() && config::NativeDepthInterpolation) << 6);
 	}
 	u32 hash(int cullMode, bool naomi2) const
 	{
-		return cullMode | ((int)naomi2 << 2);
+		return cullMode | ((int)naomi2 << 2) | ((int)(!settings.platform.isNaomi2() && config::NativeDepthInterpolation) << 3);
 	}
 
 	vk::PipelineVertexInputStateCreateInfo GetMainVertexInputStateCreateInfo(bool full = true) const

core/rend/vulkan/shaders.cpp

@@ -36,20 +36,28 @@ layout (location = 3) in mediump vec2 in_uv;
 
 layout (location = 0) INTERPOLATION out highp vec4 vtx_base;
 layout (location = 1) INTERPOLATION out highp vec4 vtx_offs;
-layout (location = 2) noperspective out highp vec3 vtx_uv;
+layout (location = 2) out highp vec3 vtx_uv;
 
 void main()
 {
 	vec4 vpos = uniformBuffer.ndcMat * in_pos;
+#if DIV_POS_Z == 1
+	vpos /= vpos.z;
+	vpos.z = vpos.w;
+#endif
 	vtx_base = vec4(in_base) / 255.0;
 	vtx_offs = vec4(in_offs) / 255.0;
-	vtx_uv = vec3(in_uv * vpos.z, vpos.z);
-#if pp_Gouraud == 1
+	vtx_uv = vec3(in_uv, vpos.z);
+#if pp_Gouraud == 1 && DIV_POS_Z != 1
 	vtx_base *= vpos.z;
 	vtx_offs *= vpos.z;
 #endif
+
+#if DIV_POS_Z != 1
+	vtx_uv.xy *= vpos.z;
 	vpos.w = 1.0;
 	vpos.z = 0.0;
+#endif
 	gl_Position = vpos;
 }
 )";
@@ -87,14 +95,20 @@ layout (set = 0, binding = 3) uniform sampler2D palette;
 // Vertex input
 layout (location = 0) INTERPOLATION in highp vec4 vtx_base;
 layout (location = 1) INTERPOLATION in highp vec4 vtx_offs;
-layout (location = 2) noperspective in highp vec3 vtx_uv;
+layout (location = 2) in highp vec3 vtx_uv;
 
 #if pp_FogCtrl != 2
 layout (set = 0, binding = 2) uniform sampler2D fog_table;
 
 float fog_mode2(float w)
 {
-	float z = clamp(w * uniformBuffer.sp_FOG_DENSITY, 1.0, 255.9999);
+	float z = clamp(
+#if DIV_POS_Z == 1
+					uniformBuffer.sp_FOG_DENSITY / w
+#else
+					uniformBuffer.sp_FOG_DENSITY * w
+#endif
+													, 1.0, 255.9999);
 	float exp = floor(log2(z));
 	float m = z * 16.0 / pow(2.0, exp) - 16.0;
 	float idx = floor(m) + exp * 16.0 + 0.5;
@@ -116,7 +130,12 @@ vec4 colorClamp(vec4 col)
 
 vec4 palettePixel(sampler2D tex, vec3 coords)
 {
-	vec4 c = vec4(textureProj(tex, coords).r * 255.0 / 1023.0 + pushConstants.palette_index, 0.5, 0.0, 0.0);
+#if DIV_POS_Z == 1
+	float texIdx = texture(tex, coords.xy).r;
+#else
+	float texIdx = textureProj(tex, coords).r;
+#endif
+	vec4 c = vec4(texIdx * 255.0 / 1023.0 + pushConstants.palette_index, 0.5, 0.0, 0.0);
 	return texture(palette, c.xy);
 }
 
@@ -133,7 +152,7 @@ void main()
 	
 	highp vec4 color = vtx_base;
 	highp vec4 offset = vtx_offs;
-	#if pp_Gouraud == 1
+	#if pp_Gouraud == 1 && DIV_POS_Z != 1
 		color /= vtx_uv.z;
 		offset /= vtx_uv.z;
 	#endif
@@ -141,12 +160,16 @@ void main()
 		color.a = 1.0;
 	#endif
 	#if pp_FogCtrl == 3
-		color = vec4(uniformBuffer.sp_FOG_COL_RAM.rgb, fog_mode2(gl_FragCoord.w));
+		color = vec4(uniformBuffer.sp_FOG_COL_RAM.rgb, fog_mode2(vtx_uv.z));
 	#endif
 	#if pp_Texture == 1
 	{
 		#if pp_Palette == 0
-			vec4 texcol = textureProj(tex, vtx_uv);
+			#if DIV_POS_Z == 1
+				vec4 texcol = texture(tex, vtx_uv.xy);
+			#else
+				vec4 texcol = textureProj(tex, vtx_uv);
+			#endif
 		#else
 			vec4 texcol = palettePixel(tex, vtx_uv);
 		#endif
@@ -216,7 +239,11 @@ void main()
 	
 	//color.rgb = vec3(gl_FragCoord.w * uniformBuffer.sp_FOG_DENSITY / 128.0);
 
-	highp float w = vtx_uv.z * 100000.0;
+#if DIV_POS_Z == 1
+	highp float w = 100000.0 / vtx_uv.z;
+#else
+	highp float w = 100000.0 * vtx_uv.z;
+#endif
 	gl_FragDepth = log2(1.0 + w) / 34.0;
 
 	gl_FragColor = color;
@@ -230,20 +257,26 @@ layout (std140, set = 0, binding = 0) uniform VertexShaderUniforms
 } uniformBuffer;
 
 layout (location = 0) in vec4 in_pos;
-layout (location = 0) noperspective out highp float depth;
+layout (location = 0) out highp float depth;
 
 void main()
 {
 	vec4 vpos = uniformBuffer.ndcMat * in_pos;
+#if DIV_POS_Z == 1
+	vpos /= vpos.z;
+	vpos.z = vpos.w;
+	depth = vpos.w;
+#else
 	depth = vpos.z;
 	vpos.w = 1.0;
 	vpos.z = 0.0;
+#endif
 	gl_Position = vpos;
 }
 )";
 
 static const char ModVolFragmentShaderSource[] = R"(
-layout (location = 0) noperspective in highp float depth;
+layout (location = 0) in highp float depth;
 layout (location = 0) out vec4 FragColor;
 
 layout (push_constant) uniform pushBlock
@@ -253,7 +286,11 @@ layout (push_constant) uniform pushBlock
 
 void main()
 {
-	highp float w = depth * 100000.0;
+#if DIV_POS_Z == 1
+	highp float w = 100000.0 / depth;
+#else
+	highp float w = 100000.0 * depth;
+#endif
 	gl_FragDepth = log2(1.0 + w) / 34.0;
 	FragColor = vec4(0.0, 0.0, 0.0, pushConstants.sp_ShaderColor);
 }
@@ -570,7 +607,7 @@ layout (location = 4) in vec3         in_normal;
 
 layout (location = 0) INTERPOLATION out highp vec4 vtx_base;
 layout (location = 1) INTERPOLATION out highp vec4 vtx_offs;
-layout (location = 2) noperspective out highp vec3 vtx_uv;
+layout (location = 2) out highp vec3 vtx_uv;
 
 void wDivide(inout vec4 vpos)
 {
@@ -633,7 +670,7 @@ layout (std140, set = 1, binding = 2) uniform N2VertexShaderUniforms
 } n2Uniform;
 
 layout (location = 0) in vec4 in_pos;
-layout (location = 0) noperspective out highp float depth;
+layout (location = 0) out highp float depth;
 
 void wDivide(inout vec4 vpos)
 {
@@ -660,6 +697,7 @@ vk::UniqueShaderModule ShaderManager::compileShader(const VertexShaderParams& pa
 	if (!params.naomi2)
 	{
 		src.addConstant("pp_Gouraud", (int)params.gouraud)
+				.addConstant("DIV_POS_Z", (int)params.divPosZ)
 				.addSource(GouraudSource)
 				.addSource(VertexShaderSource);
 	}
@@ -689,20 +727,24 @@ vk::UniqueShaderModule ShaderManager::compileShader(const FragmentShaderParams&
 		.addConstant("ColorClamping", (int)params.clamping)
 		.addConstant("pp_TriLinear", (int)params.trilinear)
 		.addConstant("pp_Palette", (int)params.palette)
+		.addConstant("DIV_POS_Z", (int)params.divPosZ)
 		.addSource(GouraudSource)
 		.addSource(FragmentShaderSource);
 	return ShaderCompiler::Compile(vk::ShaderStageFlagBits::eFragment, src.generate());
 }
 
-vk::UniqueShaderModule ShaderManager::compileModVolVertexShader(bool naomi2)
+vk::UniqueShaderModule ShaderManager::compileShader(const ModVolShaderParams& params)
 {
 	return ShaderCompiler::Compile(vk::ShaderStageFlagBits::eVertex,
-			VulkanSource().addSource(naomi2 ? N2ModVolVertexShaderSource : ModVolVertexShaderSource).generate());
+			VulkanSource().addConstant("DIV_POS_Z", (int)params.divPosZ)
+				.addSource(params.naomi2 ? N2ModVolVertexShaderSource : ModVolVertexShaderSource).generate());
 }
 
-vk::UniqueShaderModule ShaderManager::compileModVolFragmentShader()
+vk::UniqueShaderModule ShaderManager::compileModVolFragmentShader(bool divPosZ)
 {
-	return ShaderCompiler::Compile(vk::ShaderStageFlagBits::eFragment, VulkanSource().addSource(ModVolFragmentShaderSource).generate());
+	return ShaderCompiler::Compile(vk::ShaderStageFlagBits::eFragment,
+			VulkanSource().addConstant("DIV_POS_Z", (int)divPosZ)
+				.addSource(ModVolFragmentShaderSource).generate());
 }
 
 vk::UniqueShaderModule ShaderManager::compileQuadVertexShader(bool rotate)

core/rend/vulkan/shaders.h

@@ -29,8 +29,9 @@ struct VertexShaderParams
 {
 	bool gouraud;
 	bool naomi2;
+	bool divPosZ;
 
-	u32 hash() { return (u32)gouraud | ((u32)naomi2 << 1); }
+	u32 hash() { return (u32)gouraud | ((u32)naomi2 << 1) | ((u32)divPosZ << 2); }
 };
 
 // alpha test, clip test, use alpha, texture, ignore alpha, shader instr, offset, fog, gouraud, bump, clamp, trilinear
@@ -49,6 +50,7 @@ struct FragmentShaderParams
 	bool clamping;
 	bool trilinear;
 	bool palette;
+	bool divPosZ;
 
 	u32 hash()
 	{
@@ -56,10 +58,18 @@ struct FragmentShaderParams
 			| ((u32)texture << 3) | ((u32)ignoreTexAlpha << 4) | (shaderInstr << 5)
 			| ((u32)offset << 7) | ((u32)fog << 8) | ((u32)gouraud << 10)
 			| ((u32)bumpmap << 11) | ((u32)clamping << 12) | ((u32)trilinear << 13)
-			| ((u32)palette << 14);
+			| ((u32)palette << 14) | ((u32)divPosZ << 15);
 	}
 };
 
+struct ModVolShaderParams
+{
+	bool naomi2;
+	bool divPosZ;
+
+	u32 hash() { return (u32)naomi2 | ((u32)divPosZ << 1); }
+};
+
 // std140 alignment required
 struct VertexShaderUniforms
 {
@@ -96,17 +106,13 @@ class ShaderManager
 public:
 	vk::ShaderModule GetVertexShader(const VertexShaderParams& params) { return getShader(vertexShaders, params); }
 	vk::ShaderModule GetFragmentShader(const FragmentShaderParams& params) { return getShader(fragmentShaders, params); }
-	vk::ShaderModule GetModVolVertexShader(bool naomi2)
-	{
-		vk::UniqueShaderModule& shader = naomi2 ? n2ModVolVertexShader : modVolVertexShader;
-		if (!shader)
-			shader = compileModVolVertexShader(naomi2);
-		return *shader;
-	}
-	vk::ShaderModule GetModVolShader()
+	vk::ShaderModule GetModVolVertexShader(const ModVolShaderParams& params) { return getShader(modVolVertexShaders, params); }
+
+	vk::ShaderModule GetModVolShader(bool divPosZ)
 	{
+		auto& modVolShader = modVolShaders[divPosZ];
 		if (!modVolShader)
-			modVolShader = compileModVolFragmentShader();
+			modVolShader = compileModVolFragmentShader(divPosZ);
 		return *modVolShader;
 	}
 	vk::ShaderModule GetQuadVertexShader(bool rotate = false)
@@ -156,16 +162,17 @@ private:
 	template<typename T>
 	vk::ShaderModule getShader(std::map<u32, vk::UniqueShaderModule>& map, T params)
 	{
-		auto it = map.find(params.hash());
+		u32 h = params.hash();
+		auto it = map.find(h);
 		if (it != map.end())
 			return it->second.get();
-		map[params.hash()] = compileShader(params);
-		return map[params.hash()].get();
+		map[h] = compileShader(params);
+		return map[h].get();
 	}
 	vk::UniqueShaderModule compileShader(const VertexShaderParams& params);
 	vk::UniqueShaderModule compileShader(const FragmentShaderParams& params);
-	vk::UniqueShaderModule compileModVolVertexShader(bool naomi2);
-	vk::UniqueShaderModule compileModVolFragmentShader();
+	vk::UniqueShaderModule compileShader(const ModVolShaderParams& params);
+	vk::UniqueShaderModule compileModVolFragmentShader(bool divPosZ);
 	vk::UniqueShaderModule compileQuadVertexShader(bool rotate);
 	vk::UniqueShaderModule compileQuadFragmentShader(bool ignoreTexAlpha);
 	vk::UniqueShaderModule compileOSDVertexShader();
@@ -173,9 +180,8 @@ private:
 
 	std::map<u32, vk::UniqueShaderModule> vertexShaders;
 	std::map<u32, vk::UniqueShaderModule> fragmentShaders;
-	vk::UniqueShaderModule modVolVertexShader;
-	vk::UniqueShaderModule n2ModVolVertexShader;
-	vk::UniqueShaderModule modVolShader;
+	std::map<u32, vk::UniqueShaderModule> modVolVertexShaders;
+	vk::UniqueShaderModule modVolShaders[2];
 	vk::UniqueShaderModule quadVertexShader;
 	vk::UniqueShaderModule quadRotateVertexShader;
 	vk::UniqueShaderModule quadFragmentShader;

core/rend/vulkan/utils.h

@@ -82,7 +82,7 @@ static const char GouraudSource[] = R"(
 #if pp_Gouraud == 0
 #define INTERPOLATION flat
 #else
-#define INTERPOLATION noperspective
+#define INTERPOLATION
 #endif
 )";
 

shell/libretro/libretro_core_options.h

@@ -534,6 +534,20 @@ struct retro_core_option_v2_definition option_defs_us[] = {
       "256",
    },
 #endif
+   {
+      CORE_OPTION_NAME "_native_depth_interpolation",
+	  "Native Depth Interpolation",
+	  NULL,
+	  "Helps with texture corruption and depth issues on AMD GPUs. Can also help Intel GPUs in some cases.",
+	  NULL,
+	  "video",
+      {
+         { "disabled", NULL },
+         { "enabled",  NULL },
+         { NULL, NULL },
+      },
+      "disabled",
+   },
    {
       CORE_OPTION_NAME "_threaded_rendering",
       "Threaded Rendering",

shell/libretro/option.cpp

@@ -90,6 +90,7 @@ Option<int> TextureFiltering(CORE_OPTION_NAME "_texture_filtering");
 Option<bool> PowerVR2Filter(CORE_OPTION_NAME "_pvr2_filtering");
 Option<int64_t> PixelBufferSize("", 512 * 1024 * 1024);
 IntOption PerPixelLayers(CORE_OPTION_NAME "_oit_layers");
+Option<bool> NativeDepthInterpolation(CORE_OPTION_NAME "_native_depth_interpolation");
 
 // Misc