Merge pull request #4937 from unknownbrackets/gpu-minor

Ignore material update when no vertex color, softgpu vertex preview

GPU/Directx9/ShaderManagerDX9.cpp

@@ -26,6 +26,7 @@
 
 #include "Common/Common.h"
 #include "Core/Reporting.h"
+#include "GPU/Math3D.h"
 #include "GPU/GPUState.h"
 #include "GPU/ge_constants.h"
 #include "GPU/Directx9/ShaderManagerDX9.h"
@@ -257,29 +258,10 @@ void LinkedShaderDX9::SetColorUniform3ExtraFloat(D3DXHANDLE uniform, u32 color,
 	SetFloatArray(uniform, col, 4);
 }
 
-static void ConvertMatrix4x3To4x4(const float *m4x3, float *m4x4) {
-	m4x4[0] = m4x3[0];
-	m4x4[1] = m4x3[1];
-	m4x4[2] = m4x3[2];
-	m4x4[3] = 0.0f;
-	m4x4[4] = m4x3[3];
-	m4x4[5] = m4x3[4];
-	m4x4[6] = m4x3[5];
-	m4x4[7] = 0.0f;
-	m4x4[8] = m4x3[6];
-	m4x4[9] = m4x3[7];
-	m4x4[10] = m4x3[8];
-	m4x4[11] = 0.0f;
-	m4x4[12] = m4x3[9];
-	m4x4[13] = m4x3[10];
-	m4x4[14] = m4x3[11];
-	m4x4[15] = 1.0f;
-}
-
 // Utility
 void LinkedShaderDX9::SetMatrix4x3(D3DXHANDLE uniform, const float *m4x3) {
 	float m4x4[16];
-	ConvertMatrix4x3To4x4(m4x3, m4x4);
+	ConvertMatrix4x3To4x4(m4x4, m4x3);
 
 	if (m_vs->constant->SetMatrix(pD3Ddevice, uniform, (D3DXMATRIX*)m4x4) == D3D_OK); 
 	else
@@ -423,7 +405,7 @@ void LinkedShaderDX9::updateUniforms() {
 		bool allDirty = true;
 		for (int i = 0; i < numBones; i++) {
 			if (dirtyUniforms & (DIRTY_BONEMATRIX0 << i)) {
-				ConvertMatrix4x3To4x4(gstate.boneMatrix + 12 * i, allBones + 16 * i);
+				ConvertMatrix4x3To4x4(allBones + 16 * i, gstate.boneMatrix + 12 * i);
 			} else {
 				allDirty = false;
 			}
@@ -444,7 +426,7 @@ void LinkedShaderDX9::updateUniforms() {
 	float bonetemp[16];
 	for (int i = 0; i < numBones; i++) {
 		if (dirtyUniforms & (DIRTY_BONEMATRIX0 << i)) {
-			ConvertMatrix4x3To4x4(gstate.boneMatrix + 12 * i, bonetemp);
+			ConvertMatrix4x3To4x4(bonetemp, gstate.boneMatrix + 12 * i);
 
 			if (u_bone[i] != 0)
 				SetMatrix(u_bone[i], bonetemp);

GPU/Directx9/TransformPipelineDX9.cpp

@@ -137,7 +137,7 @@ using namespace DX9;
 // that's common between the many vertices of a draw call.
 class Lighter {
 public:
-	Lighter();
+	Lighter(int vertType);
 	void Light(float colorOut0[4], float colorOut1[4], const float colorIn[4], Vec3f pos, Vec3f normal);
 
 private:
@@ -152,7 +152,7 @@ private:
 	int materialUpdate_;
 };
 
-Lighter::Lighter() {
+Lighter::Lighter(int vertType) {
 	doShadeMapping_ = gstate.getUVGenMode() == GE_TEXMAP_ENVIRONMENT_MAP;
 	materialEmissive.GetFromRGB(gstate.materialemissive);
 	materialEmissive.a = 0.0f;
@@ -166,7 +166,8 @@ Lighter::Lighter() {
 	materialSpecular.a = 1.0f;
 	specCoef_ = getFloat24(gstate.materialspecularcoef);
 	// viewer_ = Vec3f(-gstate.viewMatrix[9], -gstate.viewMatrix[10], -gstate.viewMatrix[11]);
-	materialUpdate_ = gstate.materialupdate & 7;
+	bool hasColor = (vertType & GE_VTYPE_COL_MASK) != 0;
+	materialUpdate_ = hasColor ? gstate.materialupdate & 7 : 0;
 }
 
 void Lighter::Light(float colorOut0[4], float colorOut1[4], const float colorIn[4], Vec3f pos, Vec3f norm)
@@ -560,7 +561,7 @@ void TransformDrawEngineDX9::SoftwareTransformAndDraw(
 		float widthFactor = (float) w / (float) gstate_c.curTextureWidth;
 		float heightFactor = (float) h / (float) gstate_c.curTextureHeight;
 
-		Lighter lighter;
+		Lighter lighter(vertType);
 		float fog_end = getFloat24(gstate.fog1);
 		float fog_slope = getFloat24(gstate.fog2);
 

GPU/Directx9/VertexShaderGeneratorDX9.cpp

@@ -210,7 +210,7 @@ void GenerateVertexShaderDX9(int prim, char *buffer, bool useHWTransform) {
 		}
 		if (gstate.isLightingEnabled()) {
 			WRITE(p, "float4 u_ambient;\n");
-			if ((gstate.materialupdate & 2) == 0)
+			if ((gstate.materialupdate & 2) == 0 || !hasColor)
 				WRITE(p, "float3 u_matdiffuse;\n");
 			// if ((gstate.materialupdate & 4) == 0)
 			WRITE(p, "float4 u_matspecular;\n");  // Specular coef is contained in alpha
@@ -371,9 +371,9 @@ void GenerateVertexShaderDX9(int prim, char *buffer, bool useHWTransform) {
 
 		// TODO: Declare variables for dots for shade mapping if needed.
 
-		const char *ambientStr = (gstate.materialupdate & 1) ? (hasColor ? "In.C1" : "u_matambientalpha") : "u_matambientalpha";
-		const char *diffuseStr = (gstate.materialupdate & 2) ? (hasColor ? "In.C1.rgb" : "u_matambientalpha.rgb") : "u_matdiffuse";
-		const char *specularStr = (gstate.materialupdate & 4) ? (hasColor ? "In.C1.rgb" : "u_matambientalpha.rgb") : "u_matspecular.rgb";
+		const char *ambientStr = (gstate.materialupdate & 1) && hasColor ? "In.C1" : "u_matambientalpha";
+		const char *diffuseStr = (gstate.materialupdate & 2) && hasColor ? "In.C1.rgb" : "u_matdiffuse";
+		const char *specularStr = (gstate.materialupdate & 4) && hasColor ? "In.C1.rgb" : "u_matspecular.rgb";
 
 		bool diffuseIsZero = true;
 		bool specularIsZero = true;

GPU/GLES/ShaderManager.cpp

@@ -32,6 +32,7 @@
 
 #include "Core/Config.h"
 #include "Core/Reporting.h"
+#include "GPU/Math3D.h"
 #include "GPU/GPUState.h"
 #include "GPU/ge_constants.h"
 #include "GPU/GLES/ShaderManager.h"
@@ -307,28 +308,9 @@ static void SetColorUniform3ExtraFloat(int uniform, u32 color, float extra) {
 	glUniform4fv(uniform, 1, col);
 }
 
-static void ConvertMatrix4x3To4x4(const float *m4x3, float *m4x4) {
-	m4x4[0] = m4x3[0];
-	m4x4[1] = m4x3[1];
-	m4x4[2] = m4x3[2];
-	m4x4[3] = 0.0f;
-	m4x4[4] = m4x3[3];
-	m4x4[5] = m4x3[4];
-	m4x4[6] = m4x3[5];
-	m4x4[7] = 0.0f;
-	m4x4[8] = m4x3[6];
-	m4x4[9] = m4x3[7];
-	m4x4[10] = m4x3[8];
-	m4x4[11] = 0.0f;
-	m4x4[12] = m4x3[9];
-	m4x4[13] = m4x3[10];
-	m4x4[14] = m4x3[11];
-	m4x4[15] = 1.0f;
-}
-
 static void SetMatrix4x3(int uniform, const float *m4x3) {
 	float m4x4[16];
-	ConvertMatrix4x3To4x4(m4x3, m4x4);
+	ConvertMatrix4x3To4x4(m4x4, m4x3);
 	glUniformMatrix4fv(uniform, 1, GL_FALSE, m4x4);
 }
 
@@ -496,7 +478,7 @@ void LinkedShader::UpdateUniforms(u32 vertType) {
 		bool allDirty = true;
 		for (int i = 0; i < numBones; i++) {
 			if (dirty & (DIRTY_BONEMATRIX0 << i)) {
-				ConvertMatrix4x3To4x4(gstate.boneMatrix + 12 * i, allBones + 16 * i);
+				ConvertMatrix4x3To4x4(allBones + 16 * i, gstate.boneMatrix + 12 * i);
 			} else {
 				allDirty = false;
 			}
@@ -517,7 +499,7 @@ void LinkedShader::UpdateUniforms(u32 vertType) {
 	float bonetemp[16];
 	for (int i = 0; i < numBones; i++) {
 		if (dirty & (DIRTY_BONEMATRIX0 << i)) {
-			ConvertMatrix4x3To4x4(gstate.boneMatrix + 12 * i, bonetemp);
+			ConvertMatrix4x3To4x4(bonetemp, gstate.boneMatrix + 12 * i);
 			glUniformMatrix4fv(u_bone[i], 1, GL_FALSE, bonetemp);
 		}
 	}

GPU/GLES/SoftwareTransform.cpp

@@ -51,7 +51,7 @@ inline float clamp(float in, float min, float max) {
 // that's common between the many vertices of a draw call.
 class Lighter {
 public:
-	Lighter();
+	Lighter(int vertType);
 	void Light(float colorOut0[4], float colorOut1[4], const float colorIn[4], Vec3f pos, Vec3f normal);
 
 private:
@@ -66,7 +66,7 @@ private:
 	int materialUpdate_;
 };
 
-Lighter::Lighter() {
+Lighter::Lighter(int vertType) {
 	doShadeMapping_ = gstate.getUVGenMode() == GE_TEXMAP_ENVIRONMENT_MAP;
 	materialEmissive.GetFromRGB(gstate.materialemissive);
 	materialEmissive.a = 0.0f;
@@ -80,7 +80,8 @@ Lighter::Lighter() {
 	materialSpecular.a = 1.0f;
 	specCoef_ = getFloat24(gstate.materialspecularcoef);
 	// viewer_ = Vec3f(-gstate.viewMatrix[9], -gstate.viewMatrix[10], -gstate.viewMatrix[11]);
-	materialUpdate_ = gstate.materialupdate & 7;
+	bool hasColor = (vertType & GE_VTYPE_COL_MASK) != 0;
+	materialUpdate_ = hasColor ? gstate.materialupdate & 7 : 0;
 }
 
 void Lighter::Light(float colorOut0[4], float colorOut1[4], const float colorIn[4], Vec3f pos, Vec3f norm)
@@ -307,7 +308,7 @@ void TransformDrawEngine::SoftwareTransformAndDraw(
 	float widthFactor = (float) w / (float) gstate_c.curTextureWidth;
 	float heightFactor = (float) h / (float) gstate_c.curTextureHeight;
 
-	Lighter lighter;
+	Lighter lighter(vertType);
 	float fog_end = getFloat24(gstate.fog1);
 	float fog_slope = getFloat24(gstate.fog2);
 

GPU/GLES/Spline.cpp

@@ -28,10 +28,9 @@
 // The rest of the transform pipeline like lighting will go as normal, either hardware or software.
 // The implementation is initially a bit inefficient but shouldn't be a big deal.
 // An intermediate buffer of not-easy-to-predict size is stored at bufPtr.
-u32 TransformDrawEngine::NormalizeVertices(u8 *outPtr, u8 *bufPtr, const u8 *inPtr, int lowerBound, int upperBound, u32 vertType) {
+u32 TransformDrawEngine::NormalizeVertices(u8 *outPtr, u8 *bufPtr, const u8 *inPtr, VertexDecoder *dec, int lowerBound, int upperBound, u32 vertType) {
 	// First, decode the vertices into a GPU compatible format. This step can be eliminated but will need a separate
 	// implementation of the vertex decoder.
-	VertexDecoder *dec = GetVertexDecoder(vertType);
 	dec->DecodeVerts(bufPtr, inPtr, lowerBound, upperBound);
 
 	// OK, morphing eliminated but bones still remain to be taken care of.
@@ -128,6 +127,11 @@ u32 TransformDrawEngine::NormalizeVertices(u8 *outPtr, u8 *bufPtr, const u8 *inP
 	return GE_VTYPE_TC_FLOAT | GE_VTYPE_COL_8888 | GE_VTYPE_NRM_FLOAT | GE_VTYPE_POS_FLOAT | (vertType & GE_VTYPE_IDX_MASK);
 }
 
+u32 TransformDrawEngine::NormalizeVertices(u8 *outPtr, u8 *bufPtr, const u8 *inPtr, int lowerBound, int upperBound, u32 vertType) {
+	VertexDecoder *dec = GetVertexDecoder(vertType);
+	return NormalizeVertices(outPtr, bufPtr, inPtr, dec, lowerBound, upperBound, vertType);
+}
+
 #define START_OPEN 1
 #define END_OPEN 2
 

GPU/GLES/TransformPipeline.cpp

@@ -751,25 +751,6 @@ void PlanesFromMatrix(float mtx[16], Plane planes[6]) {
 	planes[5].Set(mtx[3]-mtx[2], mtx[7]-mtx[6], mtx[11]-mtx[10], mtx[15]-mtx[14]); // Far
 }
 
-static void ConvertMatrix4x3To4x4(float *m4x4, const float *m4x3) {
-	m4x4[0] = m4x3[0];
-	m4x4[1] = m4x3[1];
-	m4x4[2] = m4x3[2];
-	m4x4[3] = 0.0f;
-	m4x4[4] = m4x3[3];
-	m4x4[5] = m4x3[4];
-	m4x4[6] = m4x3[5];
-	m4x4[7] = 0.0f;
-	m4x4[8] = m4x3[6];
-	m4x4[9] = m4x3[7];
-	m4x4[10] = m4x3[8];
-	m4x4[11] = 0.0f;
-	m4x4[12] = m4x3[9];
-	m4x4[13] = m4x3[10];
-	m4x4[14] = m4x3[11];
-	m4x4[15] = 1.0f;
-}
-
 // This code is HIGHLY unoptimized!
 //
 // It does the simplest and safest test possible: If all points of a bbox is outside a single of

GPU/GLES/TransformPipeline.h

@@ -132,6 +132,9 @@ public:
 		return decJitCache_->IsInSpace(ptr);
 	}
 
+	// Really just for convenience to share with softgpu.
+	static u32 NormalizeVertices(u8 *outPtr, u8 *bufPtr, const u8 *inPtr, VertexDecoder *dec, int lowerBound, int upperBound, u32 vertType);
+
 private:
 	void DecodeVerts();
 	void DecodeVertsStep();

GPU/GLES/VertexShaderGenerator.cpp

@@ -303,7 +303,7 @@ void GenerateVertexShader(int prim, u32 vertType, char *buffer, bool useHWTransf
 		}
 		if (gstate.isLightingEnabled()) {
 			WRITE(p, "uniform lowp vec4 u_ambient;\n");
-			if ((gstate.materialupdate & 2) == 0)
+			if ((gstate.materialupdate & 2) == 0 || !hasColor)
 				WRITE(p, "uniform lowp vec3 u_matdiffuse;\n");
 			// if ((gstate.materialupdate & 4) == 0)
 			WRITE(p, "uniform lowp vec4 u_matspecular;\n");  // Specular coef is contained in alpha
@@ -457,9 +457,9 @@ void GenerateVertexShader(int prim, u32 vertType, char *buffer, bool useHWTransf
 
 		// TODO: Declare variables for dots for shade mapping if needed.
 
-		const char *ambientStr = (gstate.materialupdate & 1) ? (hasColor ? "color0" : "u_matambientalpha") : "u_matambientalpha";
-		const char *diffuseStr = (gstate.materialupdate & 2) ? (hasColor ? "color0.rgb" : "u_matambientalpha.rgb") : "u_matdiffuse";
-		const char *specularStr = (gstate.materialupdate & 4) ? (hasColor ? "color0.rgb" : "u_matambientalpha.rgb") : "u_matspecular.rgb";
+		const char *ambientStr = (gstate.materialupdate & 1) && hasColor ? "color0" : "u_matambientalpha";
+		const char *diffuseStr = (gstate.materialupdate & 2) && hasColor ? "color0.rgb" : "u_matdiffuse";
+		const char *specularStr = (gstate.materialupdate & 4) && hasColor ? "color0.rgb" : "u_matspecular.rgb";
 
 		bool diffuseIsZero = true;
 		bool specularIsZero = true;

GPU/Math3D.cpp

@@ -17,6 +17,8 @@
 
 #include "Math3D.h"
 
+namespace Math3D {
+
 template<>
 float Vec2<float>::Length() const
 {
@@ -188,3 +190,5 @@ float Vec4<float>::Normalize()
 	(*this) = (*this)/len;
 	return len;
 }
+
+}; // namespace Math3D

GPU/Math3D.h

@@ -19,6 +19,8 @@
 
 #include <cmath>
 
+namespace Math3D {
+
 // Helper for Vec classes to clamp values.
 template<typename T>
 inline static T VecClamp(const T &v, const T &low, const T &high)
@@ -318,8 +320,6 @@ public:
 #undef _DEFINE_SWIZZLER2
 };
 
-typedef Vec3<float> Vec3f;
-
 template<typename T>
 class Vec4
 {
@@ -488,8 +488,6 @@ public:
 #undef _DEFINE_SWIZZLER3
 };
 
-typedef Vec4<float> Vec4f;
-
 
 template<typename BaseType>
 class Mat3x3
@@ -590,6 +588,11 @@ private:
 	BaseType values[4*4];
 };
 
+}; // namespace Math3D
+
+typedef Math3D::Vec3<float> Vec3f;
+typedef Math3D::Vec4<float> Vec4f;
+
 
 inline void Vec3ByMatrix43(float vecOut[3], const float v[3], const float m[12])
 {
@@ -629,12 +632,33 @@ inline void Matrix4ByMatrix4(float out[16], const float a[16], const float b[16]
 	Vec4ByMatrix44(out + 12, a + 12, b);
 }
 
+inline void ConvertMatrix4x3To4x4(float *m4x4, const float *m4x3) {
+	m4x4[0] = m4x3[0];
+	m4x4[1] = m4x3[1];
+	m4x4[2] = m4x3[2];
+	m4x4[3] = 0.0f;
+	m4x4[4] = m4x3[3];
+	m4x4[5] = m4x3[4];
+	m4x4[6] = m4x3[5];
+	m4x4[7] = 0.0f;
+	m4x4[8] = m4x3[6];
+	m4x4[9] = m4x3[7];
+	m4x4[10] = m4x3[8];
+	m4x4[11] = 0.0f;
+	m4x4[12] = m4x3[9];
+	m4x4[13] = m4x3[10];
+	m4x4[14] = m4x3[11];
+	m4x4[15] = 1.0f;
+}
+
 
 inline float Vec3Dot(const float v1[3], const float v2[3])
 {
 	return v1[0]*v2[0] + v1[1]*v2[1] + v1[2]*v2[2];
 }
 
+namespace Math3D {
+
 template<typename T>
 inline T Dot(const Vec2<T>& a, const Vec2<T>& b)
 {
@@ -659,6 +683,8 @@ inline Vec3<T> Cross(const Vec3<T>& a, const Vec3<T>& b)
 	return Vec3<T>(a.y*b.z-a.z*b.y, a.z*b.x-a.x*b.z, a.x*b.y-a.y*b.x);
 }
 
+}; // namespace Math3D
+
 // linear interpolation via float: 0.0=begin, 1.0=end
 template<typename X>
 inline X Lerp(const X& begin, const X& end, const float t)

GPU/Software/Lighting.cpp

@@ -21,11 +21,11 @@
 
 namespace Lighting {
 
-void Process(VertexData& vertex)
+void Process(VertexData& vertex, bool hasColor)
 {
 	Vec3<int> mec = Vec3<int>(gstate.getMaterialEmissiveR(), gstate.getMaterialEmissiveG(), gstate.getMaterialEmissiveB());
 
-	Vec3<int> mac = (gstate.materialupdate&1)
+	Vec3<int> mac = hasColor && (gstate.materialupdate&1)
 						? vertex.color0.rgb()
 						: Vec3<int>(gstate.getMaterialAmbientR(), gstate.getMaterialAmbientG(), gstate.getMaterialAmbientB());
 	Vec3<int> final_color = mec + mac * Vec3<int>(gstate.getAmbientR(), gstate.getAmbientG(), gstate.getAmbientB()) / 255;
@@ -92,7 +92,7 @@ void Process(VertexData& vertex)
 
 		// diffuse lighting
 		Vec3<int> ldc = Vec3<int>(gstate.getDiffuseColorR(light), gstate.getDiffuseColorG(light), gstate.getDiffuseColorB(light));
-		Vec3<int> mdc = (gstate.materialupdate&2)
+		Vec3<int> mdc = hasColor && (gstate.materialupdate&2)
 							? vertex.color0.rgb()
 							: Vec3<int>(gstate.getMaterialDiffuseR(), gstate.getMaterialDiffuseG(), gstate.getMaterialDiffuseB());
 
@@ -115,7 +115,7 @@ void Process(VertexData& vertex)
 			Vec3<float> H = worldE / worldE.Length() + L / L.Length();
 
 			Vec3<int> lsc = Vec3<int>(gstate.getSpecularColorR(light), gstate.getSpecularColorG(light), gstate.getSpecularColorB(light));
-			Vec3<int> msc = (gstate.materialupdate&4)
+			Vec3<int> msc = hasColor && (gstate.materialupdate&4)
 								? vertex.color0.rgb()
 								: Vec3<int>(gstate.getMaterialSpecularR(), gstate.getMaterialSpecularG(), gstate.getMaterialSpecularB());
 
@@ -144,7 +144,7 @@ void Process(VertexData& vertex)
 		vertex.color1 = Vec3<int>(0, 0, 0);
 	}
 
-	int maa = (gstate.materialupdate&1) ? vertex.color0.a() : gstate.getMaterialAmbientA();
+	int maa = hasColor && (gstate.materialupdate&1) ? vertex.color0.a() : gstate.getMaterialAmbientA();
 	vertex.color0.a() = gstate.getAmbientA() * maa / 255;
 
 	vertex.color0 = vertex.color0.Clamp(0, 255);

GPU/Software/Lighting.h

@@ -21,6 +21,6 @@
 
 namespace Lighting {
 
-void Process(VertexData& vertex);
+void Process(VertexData& vertex, bool hasColor);
 
 }

GPU/Software/Rasterizer.cpp

@@ -433,7 +433,7 @@ static inline bool DepthTestPassed(int x, int y, u16 z)
 	}
 }
 
-static inline bool IsRightSideOrFlatBottomLine(const Vec2<u10>& vertex, const Vec2<u10>& line1, const Vec2<u10>& line2)
+static inline bool IsRightSideOrFlatBottomLine(const Vec2<fixed16>& vertex, const Vec2<fixed16>& line1, const Vec2<fixed16>& line2)
 {
 	if (line1.y == line2.y) {
 		// just check if vertex is above us => bottom line parallel to x-axis
@@ -1260,14 +1260,16 @@ void DrawLine(const VertexData &v0, const VertexData &v1)
 
 bool GetCurrentStencilbuffer(GPUDebugBuffer &buffer)
 {
-	buffer.Allocate(gstate.DepthBufStride(), 512, GPU_DBG_FORMAT_8BIT);
+	int w = gstate.getRegionX2() - gstate.getRegionX1() + 1;
+	int h = gstate.getRegionY2() - gstate.getRegionY1() + 1;
+	buffer.Allocate(w, h, GPU_DBG_FORMAT_8BIT);
 
 	u8 *row = buffer.GetData();
-	for (int y = 0; y < 512; ++y) {
-		for (int x = 0; x < gstate.DepthBufStride(); ++x) {
-			row[x] = GetPixelStencil(x, y);
+	for (int y = gstate.getRegionY1(); y <= gstate.getRegionY2(); ++y) {
+		for (int x = gstate.getRegionX1(); x <= gstate.getRegionX2(); ++x) {
+			row[x - gstate.getRegionX1()] = GetPixelStencil(x, y);
 		}
-		row += gstate.DepthBufStride();
+		row += w;
 	}
 	return true;
 }

GPU/Software/SoftGpu.cpp

@@ -865,17 +865,35 @@ bool SoftGPU::FramebufferDirty() {
 
 bool SoftGPU::GetCurrentFramebuffer(GPUDebugBuffer &buffer)
 {
-	// We don't know the height, so just use 512, which should be the max (hopefully?)
-	// TODO: Could check clipping and such, though...?
-	buffer = GPUDebugBuffer(fb.data, gstate.FrameBufStride(), 512, gstate.FrameBufFormat());
+	const int w = gstate.getRegionX2() - gstate.getRegionX1() + 1;
+	const int h = gstate.getRegionY2() - gstate.getRegionY1() + 1;
+	buffer.Allocate(w, h, gstate.FrameBufFormat());
+
+	const int depth = gstate.FrameBufFormat() == GE_FORMAT_8888 ? 4 : 2;
+	const u8 *src = fb.data + gstate.FrameBufStride() * depth * gstate.getRegionY1();
+	u8 *dst = buffer.GetData();
+	for (int y = gstate.getRegionY1(); y <= gstate.getRegionY2(); ++y) {
+		memcpy(dst, src + gstate.getRegionX1(), (gstate.getRegionX2() + 1) * depth);
+		dst += w * depth;
+		src += gstate.FrameBufStride() * depth;
+	}
 	return true;
 }
 
 bool SoftGPU::GetCurrentDepthbuffer(GPUDebugBuffer &buffer)
 {
-	// We don't know the height, so just use 512, which should be the max (hopefully?)
-	// TODO: Could check clipping and such, though...?
-	buffer = GPUDebugBuffer(depthbuf.data, gstate.DepthBufStride(), 512, GPU_DBG_FORMAT_16BIT);
+	const int w = gstate.getRegionX2() - gstate.getRegionX1() + 1;
+	const int h = gstate.getRegionY2() - gstate.getRegionY1() + 1;
+	buffer.Allocate(w, h, GPU_DBG_FORMAT_16BIT);
+
+	const int depth = 2;
+	const u8 *src = depthbuf.data + gstate.DepthBufStride() * depth * gstate.getRegionY1();
+	u8 *dst = buffer.GetData();
+	for (int y = gstate.getRegionY1(); y <= gstate.getRegionY2(); ++y) {
+		memcpy(dst, src + gstate.getRegionX1(), (gstate.getRegionX2() + 1) * depth);
+		dst += w * depth;
+		src += gstate.DepthBufStride() * depth;
+	}
 	return true;
 }
 
@@ -888,3 +906,8 @@ bool SoftGPU::GetCurrentTexture(GPUDebugBuffer &buffer)
 {
 	return Rasterizer::GetCurrentTexture(buffer);
 }
+
+bool SoftGPU::GetCurrentSimpleVertices(int count, std::vector<GPUDebugVertex> &vertices, std::vector<u16> &indices)
+{
+	return TransformUnit::GetCurrentSimpleVertices(count, vertices, indices);
+}

GPU/Software/SoftGpu.h

@@ -86,6 +86,7 @@ public:
 	virtual bool GetCurrentDepthbuffer(GPUDebugBuffer &buffer);
 	virtual bool GetCurrentStencilbuffer(GPUDebugBuffer &buffer);
 	virtual bool GetCurrentTexture(GPUDebugBuffer &buffer);
+	bool GetCurrentSimpleVertices(int count, std::vector<GPUDebugVertex> &vertices, std::vector<u16> &indices);
 
 protected:
 	virtual void FastRunLoop(DisplayList &list);

GPU/Software/TransformUnit.cpp

@@ -16,8 +16,11 @@
 // https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
 
 #include "Core/Host.h"
-#include "../GPUState.h"
-#include "../GLES/VertexDecoder.h"
+#include "Core/Config.h"
+#include "GPU/GPUState.h"
+#include "GPU/GLES/VertexDecoder.h"
+#include "GPU/GLES/TransformPipeline.h"
+#include "GPU/Common/SplineCommon.h"
 
 #include "TransformUnit.h"
 #include "Clipper.h"
@@ -177,7 +180,7 @@ static VertexData ReadVertex(VertexReader& vreader)
 			vertex.worldnormal /= vertex.worldnormal.Length();
 		}
 
-		Lighting::Process(vertex);
+		Lighting::Process(vertex, vreader.hasColor0());
 	} else {
 		vertex.screenpos.x = (u32)pos[0] * 16 + gstate.getOffsetX16();
 		vertex.screenpos.y = (u32)pos[1] * 16 + gstate.getOffsetY16();
@@ -473,3 +476,82 @@ void TransformUnit::SubmitPrimitive(void* vertices, void* indices, u32 prim_type
 
 	host->GPUNotifyDraw();
 }
+
+// TODO: This probably is not the best interface.
+bool TransformUnit::GetCurrentSimpleVertices(int count, std::vector<GPUDebugVertex> &vertices, std::vector<u16> &indices) {
+	// This is always for the current vertices.
+	u16 indexLowerBound = 0;
+	u16 indexUpperBound = count - 1;
+
+	if ((gstate.vertType & GE_VTYPE_IDX_MASK) != GE_VTYPE_IDX_NONE) {
+		const u8 *inds = Memory::GetPointer(gstate_c.indexAddr);
+		const u16 *inds16 = (const u16 *)inds;
+
+		if (inds) {
+			GetIndexBounds(inds, count, gstate.vertType, &indexLowerBound, &indexUpperBound);
+			indices.resize(count);
+			switch (gstate.vertType & GE_VTYPE_IDX_MASK) {
+			case GE_VTYPE_IDX_16BIT:
+				for (int i = 0; i < count; ++i) {
+					indices[i] = inds16[i];
+				}
+				break;
+			case GE_VTYPE_IDX_8BIT:
+				for (int i = 0; i < count; ++i) {
+					indices[i] = inds[i];
+				}
+				break;
+			default:
+				return false;
+			}
+		} else {
+			indices.clear();
+		}
+	} else {
+		indices.clear();
+	}
+
+	static std::vector<u32> temp_buffer;
+	static std::vector<SimpleVertex> simpleVertices;
+	temp_buffer.resize(65536 * 24 / sizeof(u32));
+	simpleVertices.resize(indexUpperBound + 1);
+
+	VertexDecoder vdecoder;
+	vdecoder.SetVertexType(gstate.vertType);
+	TransformDrawEngine::NormalizeVertices((u8 *)(&simpleVertices[0]), (u8 *)(&temp_buffer[0]), Memory::GetPointer(gstate_c.vertexAddr), &vdecoder, indexLowerBound, indexUpperBound, gstate.vertType);
+
+	float world[16];
+	float view[16];
+	float worldview[16];
+	float worldviewproj[16];
+	ConvertMatrix4x3To4x4(world, gstate.worldMatrix);
+	ConvertMatrix4x3To4x4(view, gstate.viewMatrix);
+	Matrix4ByMatrix4(worldview, world, view);
+	Matrix4ByMatrix4(worldviewproj, worldview, gstate.projMatrix);
+
+	vertices.resize(indexUpperBound + 1);
+	for (int i = indexLowerBound; i <= indexUpperBound; ++i) {
+		const SimpleVertex &vert = simpleVertices[i];
+
+		if (gstate.isModeThrough()) {
+			vertices[i].u = vert.uv[0];
+			vertices[i].v = vert.uv[1];
+			vertices[i].x = vert.pos.x;
+			vertices[i].y = vert.pos.y;
+			vertices[i].z = vert.pos.z;
+		} else {
+			float clipPos[4];
+			Vec3ByMatrix44(clipPos, vert.pos.AsArray(), worldviewproj);
+			ScreenCoords screenPos = ClipToScreen(clipPos);
+			DrawingCoords drawPos = ScreenToDrawing(screenPos);
+
+			vertices[i].u = vert.uv[0];
+			vertices[i].v = vert.uv[1];
+			vertices[i].x = drawPos.x;
+			vertices[i].y = drawPos.y;
+			vertices[i].z = 1.0;
+		}
+	}
+
+	return true;
+}

GPU/Software/TransformUnit.h

@@ -18,8 +18,11 @@
 #pragma once
 
 #include "CommonTypes.h"
+#include "GPU/Common/GPUDebugInterface.h"
 #include "../Math3D.h"
 
+using namespace Math3D;
+
 typedef u16 fixed16;
 typedef u16 u10; // TODO: erm... :/
 
@@ -118,4 +121,6 @@ public:
 
 	static void SubmitSpline(void* control_points, void* indices, int count_u, int count_v, int type_u, int type_v, GEPatchPrimType prim_type, u32 vertex_type);
 	static void SubmitPrimitive(void* vertices, void* indices, u32 prim_type, int vertex_count, u32 vertex_type, int *bytesRead);
+
+	static bool GetCurrentSimpleVertices(int count, std::vector<GPUDebugVertex> &vertices, std::vector<u16> &indices);
 };