diff --git a/GPU/GLES/FragmentShaderGenerator.cpp b/GPU/GLES/FragmentShaderGenerator.cpp
index b225c299bf..d240894809 100644
--- a/GPU/GLES/FragmentShaderGenerator.cpp
+++ b/GPU/GLES/FragmentShaderGenerator.cpp
@@ -21,7 +21,6 @@
 #define GLSL_1_3
 #endif
 
-
 #include "FragmentShaderGenerator.h"
 #include "../ge_constants.h"
 #include "../GPUState.h"
@@ -77,7 +76,9 @@ char *GenerateFragmentShader()
 
 	int lmode = gstate.lmode & 1;
 
-	if (gstate.textureMapEnable & 1)
+	int doTexture = (gstate.textureMapEnable & 1) && !(gstate.clearmode & 1);
+
+	if (doTexture)
 		WRITE(p, "uniform sampler2D tex;\n");
 	if (gstate.alphaTestEnable & 1)
 		WRITE(p, "uniform vec4 u_alpharef;\n");
@@ -89,7 +90,8 @@ char *GenerateFragmentShader()
 	WRITE(p, "varying vec4 v_color0;\n");
 	if (lmode)
 		WRITE(p, "varying vec4 v_color1;\n");
-	WRITE(p, "varying vec2 v_texcoord;\n");
+	if (doTexture)
+		WRITE(p, "varying vec2 v_texcoord;\n");
 	if (gstate.isFogEnabled())
 		WRITE(p, "varying float v_depth;\n");
 
@@ -98,6 +100,7 @@ char *GenerateFragmentShader()
 
 	if (gstate.clearmode & 1)
 	{
+		// Clear mode does not allow any fancy shading.
 		WRITE(p, "  v = v_color0;\n");
 	}
 	else
@@ -156,7 +159,7 @@ char *GenerateFragmentShader()
 		if (gstate.texfunc & 0x10000) {
 			WRITE(p, "  v = v * vec4(2.0, 2.0, 2.0, 1.0);");
 		}
-		
+
 		if (gstate.alphaTestEnable & 1) {
 			int alphaTestFunc = gstate.alphatest & 7;
 			const char *alphaTestFuncs[] = { "#", "#", " == ", " != ", " < ", " <= ", " > ", " >= " };	// never/always don't make sense
@@ -169,7 +172,6 @@ char *GenerateFragmentShader()
 			// WRITE(p, "  v = mix(v, u_fogcolor, u_fogcoef.x + u_fogcoef.y * v_depth;\n");
 			// WRITE(p, "  v.x = v_depth;\n");
 		}
-		// Fogging should be added here - and disabled during clear mode
 	}
 
 	WRITE(p, "  gl_FragColor = v;\n");
diff --git a/GPU/GLES/ShaderManager.cpp b/GPU/GLES/ShaderManager.cpp
index 0f004653c2..93096d67d6 100644
--- a/GPU/GLES/ShaderManager.cpp
+++ b/GPU/GLES/ShaderManager.cpp
@@ -46,7 +46,7 @@ Shader::Shader(const char *code, uint32_t shaderType) {
 		ERROR_LOG(G3D, "Info log: %s\n", infoLog);
 		ERROR_LOG(G3D, "Shader source:\n%s\n", (const char *)code);
 	} else {
-		//NOTICE_LOG(G3D, "Compiled shader:\n%s\n", (const char *)code);
+		DEBUG_LOG(G3D, "Compiled shader:\n%s\n", (const char *)code);
 	}
 }
 
@@ -67,32 +67,30 @@ LinkedShader::LinkedShader(Shader *vs, Shader *fs)
 			glGetProgramInfoLog(program, bufLength, NULL, buf);
 			ERROR_LOG(G3D, "Could not link program:\n %s", buf);
 			delete [] buf;	// we're dead!
-
-			//_dbg_assert_msg_(HLE,0,"Could not link program");
 		}
 		return;
 	}
 
-	NOTICE_LOG(G3D, "Linked shader!");
+	INFO_LOG(G3D, "Linked shader: vs %i fs %i", (int)vs->shader, (int)fs->shader);
 
-	u_tex		= glGetUniformLocation(program, "tex");
-	u_proj	 = glGetUniformLocation(program, "u_proj");
+	u_tex = glGetUniformLocation(program, "tex");
+	u_proj = glGetUniformLocation(program, "u_proj");
 	u_proj_through = glGetUniformLocation(program, "u_proj_through");
 	u_texenv = glGetUniformLocation(program, "u_texenv");
 	u_fogcolor = glGetUniformLocation(program, "u_fogcolor");
 	u_fogcoef = glGetUniformLocation(program, "u_fogcoef");
 	u_alpharef = glGetUniformLocation(program, "u_alpharef");
 
-	a_position	= glGetAttribLocation(program, "a_position");
-	a_color0		= glGetAttribLocation(program, "a_color0");
-	a_color1		= glGetAttribLocation(program, "a_color1");
-	a_texcoord	= glGetAttribLocation(program, "a_texcoord");
+	a_position = glGetAttribLocation(program, "a_position");
+	a_color0 = glGetAttribLocation(program, "a_color0");
+	a_color1 = glGetAttribLocation(program, "a_color1");
+	a_texcoord = glGetAttribLocation(program, "a_texcoord");
 
 	glUseProgram(program);
 	// Default uniform values
 	glUniform1i(u_tex, 0);
 	// The rest, use the "dirty" mechanism.
-	dirtyUniforms = DIRTY_PROJMATRIX | DIRTY_TEXENV | DIRTY_ALPHAREF;
+	dirtyUniforms = DIRTY_PROJMATRIX | DIRTY_PROJTHROUGHMATRIX | DIRTY_TEXENV | DIRTY_ALPHAREF;
 }
 
 LinkedShader::~LinkedShader() {
@@ -100,7 +98,7 @@ LinkedShader::~LinkedShader() {
 }
 
 void LinkedShader::use() {
-	glUseProgram(program);	
+	glUseProgram(program);
 	glUniform1i(u_tex, 0);
 	// Update any dirty uniforms before we draw
 	if (u_proj != -1 && (dirtyUniforms & DIRTY_PROJMATRIX)) {
@@ -175,7 +173,7 @@ void ShaderManager::DirtyShader()
 }
 
 
-LinkedShader *ShaderManager::ApplyShader()
+LinkedShader *ShaderManager::ApplyShader(int prim)
 {
 	if (globalDirty) {
 		// Deferred dirtying! Let's see if we can make this even more clever later.
@@ -187,7 +185,7 @@ LinkedShader *ShaderManager::ApplyShader()
 
 	VertexShaderID VSID;
 	FragmentShaderID FSID;
-	ComputeVertexShaderID(&VSID);
+	ComputeVertexShaderID(&VSID, prim);
 	ComputeFragmentShaderID(&FSID);
 
 	// Bail quickly in the no-op case. TODO: why does it cause trouble?
@@ -206,7 +204,7 @@ LinkedShader *ShaderManager::ApplyShader()
 	} else {
 		vs = vsIter->second;
 	}
-	
+
 	FSCache::iterator fsIter = fsCache.find(FSID);
 	Shader *fs;
 	if (fsIter == fsCache.end())	{
diff --git a/GPU/GLES/ShaderManager.h b/GPU/GLES/ShaderManager.h
index 080b5331e3..9fbe842ca9 100644
--- a/GPU/GLES/ShaderManager.h
+++ b/GPU/GLES/ShaderManager.h
@@ -53,8 +53,13 @@ struct LinkedShader
 	int u_alpharef;
 	int u_fogcolor;
 	int u_fogcoef;
+
+	// Lighting
+	int u_ambientcolor;
+	int u_light[4];  // each light consist of vec4[3]
 };
 
+// Will reach 32 bits soon :P
 enum
 {
 	DIRTY_PROJMATRIX = (1 << 0),
@@ -63,8 +68,29 @@ enum
 	DIRTY_FOGCOEF    = (1 << 3),
 	DIRTY_TEXENV		 = (1 << 4),
 	DIRTY_ALPHAREF	 = (1 << 5),
+	DIRTY_COLORREF	 = (1 << 6),
 
-	DIRTY_ALL = (1 << 6) - 1
+	DIRTY_LIGHT0 = (1 << 12),
+	DIRTY_LIGHT1 = (1 << 13),
+	DIRTY_LIGHT2 = (1 << 14),
+	DIRTY_LIGHT3 = (1 << 15),
+
+	DIRTY_GLOBALAMBIENT = (1 << 16),
+	DIRTY_MATERIAL = (1 << 17),  // let's set all 4 together (emissive ambient diffuse specular). We hide specular coef in specular.a
+	DIRTY_UVSCALEOFFSET = (1 << 18),  // this will be dirtied ALL THE TIME... maybe we'll need to do "last value with this shader compares"
+
+	DIRTY_VIEWMATRIX = (1 << 22),  // Maybe we'll fold this into projmatrix eventually
+	DIRTY_TEXMATRIX = (1 << 23),
+	DIRTY_BONEMATRIX0 = (1 << 24),
+	DIRTY_BONEMATRIX1 = (1 << 25),
+	DIRTY_BONEMATRIX2 = (1 << 26),
+	DIRTY_BONEMATRIX3 = (1 << 27),
+	DIRTY_BONEMATRIX4 = (1 << 28),
+	DIRTY_BONEMATRIX5 = (1 << 29),
+	DIRTY_BONEMATRIX6 = (1 << 30),
+	DIRTY_BONEMATRIX7 = (1 << 31),
+
+	DIRTY_ALL = 0xFFFFFFFF
 };
 
 // Real public interface
@@ -84,7 +110,7 @@ public:
 	ShaderManager() : globalDirty(0xFFFFFFFF) {}
 
 	void ClearCache(bool deleteThem);  // TODO: deleteThem currently not respected
-	LinkedShader *ApplyShader();
+	LinkedShader *ApplyShader(int prim);
 	void DirtyShader();
 	void DirtyUniform(u32 what);
 
diff --git a/GPU/GLES/TransformPipeline.cpp b/GPU/GLES/TransformPipeline.cpp
index a1c9ad31fa..31071a23b6 100644
--- a/GPU/GLES/TransformPipeline.cpp
+++ b/GPU/GLES/TransformPipeline.cpp
@@ -119,9 +119,9 @@ void Lighter::Light(float colorOut0[4], float colorOut1[4], const float colorIn[
 		specular = &in;
 	else
 		specular = &materialSpecular;
-		
+
 	Color4 lightSum0 = globalAmbient * *ambient + materialEmissive;
-	Color4 lightSum1(0,0,0,0);
+	Color4 lightSum1(0, 0, 0, 0);
 
 	// Try lights.elf - there's something wrong with the lighting
 
@@ -131,8 +131,8 @@ void Lighter::Light(float colorOut0[4], float colorOut1[4], const float colorIn[
 		if ((gstate.lightEnable[l] & 1) == 0 && !doShadeMapping_)
 			continue;
 
-		GELightComputation comp = (GELightComputation)(gstate.ltype[l]&3);
-		GELightType type = (GELightType)((gstate.ltype[l]>>8)&3);
+		GELightComputation comp = (GELightComputation)(gstate.ltype[l] & 3);
+		GELightType type = (GELightType)((gstate.ltype[l] >> 8) & 3);
 		Vec3 toLight;
 
 		if (type == GE_LIGHTTYPE_DIRECTIONAL)
@@ -146,7 +146,7 @@ void Lighter::Light(float colorOut0[4], float colorOut1[4], const float colorIn[
 		float lightScale = 1.0f;
 		if (type != GE_LIGHTTYPE_DIRECTIONAL)
 		{
-			float distance = toLight.Normalize(); 
+			float distance = toLight.Normalize();
 			lightScale = 1.0f / (gstate_c.lightatt[l][0] + gstate_c.lightatt[l][1]*distance + gstate_c.lightatt[l][2]*distance*distance);
 			if (lightScale > 1.0f) lightScale = 1.0f;
 		}
@@ -159,12 +159,12 @@ void Lighter::Light(float colorOut0[4], float colorOut1[4], const float colorIn[
 		if (poweredDiffuse)
 			dot = powf(dot, specCoef_);
 
-		Color4 diff = (gstate_c.lightColor[1][l] * *diffuse) * (dot * lightScale);	
+		Color4 diff = (gstate_c.lightColor[1][l] * *diffuse) * (dot * lightScale);
 
 		// Real PSP specular
 		Vec3 toViewer(0,0,1);
 		// Better specular
-		//Vec3 toViewer = (viewer - pos).Normalized();
+		// Vec3 toViewer = (viewer - pos).Normalized();
 
 		if (doSpecular)
 		{
@@ -176,7 +176,7 @@ void Lighter::Light(float colorOut0[4], float colorOut1[4], const float colorIn[
 			if (dot >= 0)
 			{
 				lightSum1 += (gstate_c.lightColor[2][l] * *specular * (powf(dot, specCoef_)*lightScale));
-			}	
+			}
 		}
 		dots[l] = dot;
 		if (gstate.lightEnable[l] & 1)
@@ -203,7 +203,11 @@ void GLES_GPU::TransformAndDrawPrim(void *verts, void *inds, int prim, int verte
 	VertexDecoder dec;
 	dec.SetVertexType(gstate.vertType);
 	dec.DecodeVerts(decoded, verts, inds, prim, vertexCount, &indexLowerBound, &indexUpperBound);
-
+#if 0
+	for (int i = indexLowerBound; i <= indexUpperBound; i++) {
+		PrintDecodedVertex(decoded[i], gstate.vertType);
+	}
+#endif
 	bool useTexCoord = false;
 
 	// Check if anything needs updating
@@ -217,14 +221,14 @@ void GLES_GPU::TransformAndDrawPrim(void *verts, void *inds, int prim, int verte
 	}
 	gpuStats.numDrawCalls++;
 	gpuStats.numVertsTransformed += vertexCount;
-	
+
 	if (bytesRead)
 		*bytesRead = vertexCount * dec.VertexSize();
 
 	bool throughmode = (gstate.vertType & GE_VTYPE_THROUGH_MASK) != 0;
 	// Then, transform and draw in one big swoop (urgh!)
 	// need to move this to the shader.
-	
+
 	// We're gonna have to keep software transforming RECTANGLES, unless we use a geom shader which we can't on OpenGL ES 2.0.
 	// Usually, though, these primitives don't use lighting etc so it's no biggie performance wise, but it would be nice to get rid of
 	// this code.
@@ -322,7 +326,7 @@ void GLES_GPU::TransformAndDrawPrim(void *verts, void *inds, int prim, int verte
 			float litColor0[4];
 			float litColor1[4];
 			lighter.Light(litColor0, litColor1, unlitColor, out, norm, dots);
-				
+
 			if (gstate.lightingEnable & 1)
 			{
 				// TODO: don't ignore gstate.lmode - we should send two colors in that case
@@ -459,7 +463,7 @@ void GLES_GPU::TransformAndDrawPrim(void *verts, void *inds, int prim, int verte
 			if (indexType == GE_VTYPE_IDX_8BIT)
 			{
 				index = ((u8*)inds)[i];
-			} 
+			}
 			else if (indexType == GE_VTYPE_IDX_16BIT)
 			{
 				index = ((u16*)inds)[i];
@@ -480,7 +484,7 @@ void GLES_GPU::TransformAndDrawPrim(void *verts, void *inds, int prim, int verte
 				// We have to turn the rectangle into two triangles, so 6 points. Sigh.
 
 				// TODO: there's supposed to be extra magic here to rotate the UV coordinates depending on if upside down etc.
-				
+
 				// bottom right
 				*trans = transVtx;
 				trans++;
@@ -606,17 +610,19 @@ void GLES_GPU::TransformAndDrawPrim(void *verts, void *inds, int prim, int verte
 	glstate.depthTest.set(wantDepthTest);
 	if(wantDepthTest) {
 		// Force GL_ALWAYS if mode clear
-		u8 depthTestFunc = gstate.isModeClear() ? 1 : gstate.getDepthTestFunc();
+		int depthTestFunc = gstate.isModeClear() ? 1 : gstate.getDepthTestFunc();
 		glstate.depthFunc.set(ztests[depthTestFunc]);
 	}
 
 	bool wantDepthWrite = gstate.isModeClear() || gstate.isDepthWriteEnabled();
 	glstate.depthWrite.set(wantDepthWrite ? GL_TRUE : GL_FALSE);
 
-	glstate.depthRange.set(gstate_c.zOff - gstate_c.zScale, gstate_c.zOff + gstate_c.zScale);
+	float depthRangeMin = gstate_c.zOff - gstate_c.zScale;
+	float depthRangeMax = gstate_c.zOff + gstate_c.zScale;
+	glstate.depthRange.set(depthRangeMin, depthRangeMax);
 
 	UpdateViewportAndProjection();
-	LinkedShader *program = shaderManager_->ApplyShader();
+	LinkedShader *program = shaderManager_->ApplyShader(prim);
 
 	// TODO: Make a cache for glEnableVertexAttribArray and glVertexAttribPtr states, these spam the gDebugger log.
 	glEnableVertexAttribArray(program->a_position);
@@ -625,7 +631,7 @@ void GLES_GPU::TransformAndDrawPrim(void *verts, void *inds, int prim, int verte
 	if (program->a_color1 != -1) glEnableVertexAttribArray(program->a_color1);
 	const int vertexSize = sizeof(transformed[0]);
 	glVertexAttribPointer(program->a_position, 3, GL_FLOAT, GL_FALSE, vertexSize, drawBuffer);
-	if (useTexCoord && program->a_texcoord != -1) glVertexAttribPointer(program->a_texcoord, 2, GL_FLOAT, GL_FALSE, vertexSize, ((uint8_t*)drawBuffer) + 3 * 4);	
+	if (useTexCoord && program->a_texcoord != -1) glVertexAttribPointer(program->a_texcoord, 2, GL_FLOAT, GL_FALSE, vertexSize, ((uint8_t*)drawBuffer) + 3 * 4);
 	if (program->a_color0 != -1) glVertexAttribPointer(program->a_color0, 4, GL_FLOAT, GL_FALSE, vertexSize, ((uint8_t*)drawBuffer) + 5 * 4);
 	if (program->a_color1 != -1) glVertexAttribPointer(program->a_color1, 4, GL_FLOAT, GL_FALSE, vertexSize, ((uint8_t*)drawBuffer) + 9 * 4);
 	// NOTICE_LOG(G3D,"DrawPrimitive: %i", numTrans);
@@ -643,7 +649,7 @@ void GLES_GPU::TransformAndDrawPrim(void *verts, void *inds, int prim, int verte
 void GLES_GPU::UpdateViewportAndProjection()
 {
 	bool throughmode = (gstate.vertType & GE_VTYPE_THROUGH_MASK) != 0;
-	
+
 	// We can probably use these to simply set scissors? Maybe we need to offset by regionX1/Y1
 	int regionX1 = gstate.region1 & 0x3FF;
 	int regionY1 = (gstate.region1 >> 10) & 0x3FF;
@@ -654,8 +660,8 @@ void GLES_GPU::UpdateViewportAndProjection()
 	float offsetY = (float)(gstate.offsety & 0xFFFF) / 16.0f;
 
 	if (throughmode) {
-		return;
 		// No viewport transform here. Let's experiment with using region.
+		return;
 		glViewport((0 + regionX1) * renderWidthFactor_, (0 - regionY1) * renderHeightFactor_, (regionX2 - regionX1) * renderWidthFactor_, (regionY2 - regionY1) * renderHeightFactor_);
 	} else {
 		// These we can turn into a glViewport call, offset by offsetX and offsetY. Math after.
@@ -693,8 +699,8 @@ void GLES_GPU::UpdateViewportAndProjection()
 
 		// Flip vpY0 to match the OpenGL coordinate system.
 		vpY0 = renderHeight_ - (vpY0 + vpHeight);
-		glViewport(vpX0, vpY0, vpWidth, vpHeight); 
+		glViewport(vpX0, vpY0, vpWidth, vpHeight);
 		// Sadly, as glViewport takes integers, we will not be able to support sub pixel offsets this way. But meh.
 		shaderManager_->DirtyUniform(DIRTY_PROJMATRIX);
 	}
-}
\ No newline at end of file
+}
diff --git a/GPU/GLES/VertexDecoder.cpp b/GPU/GLES/VertexDecoder.cpp
index 5f7acbb29c..33f3281865 100644
--- a/GPU/GLES/VertexDecoder.cpp
+++ b/GPU/GLES/VertexDecoder.cpp
@@ -22,6 +22,15 @@
 
 #include "VertexDecoder.h"
 
+void PrintDecodedVertex(const DecodedVertex &vtx, u32 vtype)
+{
+	if (vtype & GE_VTYPE_NRM_MASK) printf("N: %f %f %f\n", vtx.normal[0], vtx.normal[1], vtx.normal[2]);
+	if (vtype & GE_VTYPE_TC_MASK) printf("TC: %f %f\n", vtx.uv[0], vtx.uv[1]);
+	if (vtype & GE_VTYPE_COL_MASK) printf("C: %02x %02x %02x %02x\n", vtx.color[0], vtx.color[1], vtx.color[2], vtx.color[3]);
+	if (vtype & GE_VTYPE_WEIGHT_MASK) printf("W: TODO\n");
+	printf("P: %f %f %f\n", vtx.pos[0], vtx.pos[1], vtx.pos[2]);
+}
+
 const int tcsize[4] = {0,2,4,8}, tcalign[4] = {0,1,2,4};
 const int colsize[8] = {0,0,0,0,2,2,2,4}, colalign[8] = {0,0,0,0,2,2,2,4};
 const int nrmsize[4] = {0,3,6,12}, nrmalign[4] = {0,1,2,4};
@@ -30,7 +39,7 @@ const int wtsize[4] = {0,1,2,4}, wtalign[4] = {0,1,2,4};
 
 inline int align(int n, int align)
 {
-	return (n+(align-1)) & ~(align-1);
+	return (n + (align - 1)) & ~(align - 1);
 }
 
 void VertexDecoder::SetVertexType(u32 fmt)
@@ -40,7 +49,7 @@ void VertexDecoder::SetVertexType(u32 fmt)
 
 	int biggest = 0;
 	size = 0;
-	
+
 	tc				 = fmt & 0x3;
 	col				= (fmt >> 2) & 0x7;
 	nrm				= (fmt >> 5) & 0x3;
@@ -51,7 +60,7 @@ void VertexDecoder::SetVertexType(u32 fmt)
 	nweights	 = ((fmt >> 14) & 0x7)+1;
 
 	DEBUG_LOG(G3D,"VTYPE: THRU=%i TC=%i COL=%i POS=%i NRM=%i WT=%i NW=%i IDX=%i MC=%i", (int)throughmode, tc,col,pos,nrm,weighttype,nweights,idx,morphcount);
-	
+
 	if (weighttype)
 	{
 		//size = align(size, wtalign[weighttype]);	unnecessary
diff --git a/GPU/GLES/VertexDecoder.h b/GPU/GLES/VertexDecoder.h
index a16725338c..464ebf72a1 100644
--- a/GPU/GLES/VertexDecoder.h
+++ b/GPU/GLES/VertexDecoder.h
@@ -40,26 +40,27 @@ struct TransformedVertex
 
 
 
-// Right now 
+// Right now
 //   - only contains computed information
 //   - does decoding in nasty branchfilled loops
-// Future TODO 
-//   - will compile into lighting fast specialized x86 
+// Future TODO
+//   - will compile into lighting fast specialized x86 and ARM
 //   - will not bother translating components that can be read directly
 //     by OpenGL ES. Will still have to translate 565 colors, and things
 //     like that. DecodedVertex will not be a fixed struct. Will have to
 //     do morphing here.
 //
 // We want 100% perf on 1Ghz even in vertex complex games!
-class VertexDecoder 
+class VertexDecoder
 {
 public:
 	VertexDecoder() : coloff(0), nrmoff(0), posoff(0) {}
 	~VertexDecoder() {}
-	void SetVertexType(u32 fmt);
+	void SetVertexType(u32 vtype);
 	void DecodeVerts(DecodedVertex *decoded, const void *verts, const void *inds, int prim, int count, int *indexLowerBound, int *indexUpperBound) const;
 	bool hasColor() const { return col != 0; }
 	int VertexSize() const { return size; }
+
 private:
 	u32 fmt;
 	bool throughmode;
@@ -81,5 +82,8 @@ private:
 	int idx;
 	int morphcount;
 	int nweights;
-
 };
+
+// Debugging utilities
+void PrintDecodedVertex(const DecodedVertex &vtx, u32 vtype);
+
diff --git a/GPU/GLES/VertexShaderGenerator.cpp b/GPU/GLES/VertexShaderGenerator.cpp
index e974d9928d..f6387f5603 100644
--- a/GPU/GLES/VertexShaderGenerator.cpp
+++ b/GPU/GLES/VertexShaderGenerator.cpp
@@ -38,12 +38,21 @@ static char buffer[16384];
 
 #define WRITE(x, ...) p+=sprintf(p, x "\n" __VA_ARGS__)
 
-void ComputeVertexShaderID(VertexShaderID *id)
+// prim so we can special case for RECTANGLES :(
+void ComputeVertexShaderID(VertexShaderID *id, int prim)
 {
 	memset(id->d, 0, sizeof(id->d));
 	id->d[0] = gstate.lmode & 1;
 	id->d[0] |= ((int)gstate.isModeThrough()) << 1;
 	id->d[0] |= ((int)gstate.isFogEnabled()) << 2;
+
+	// Bits that we will need:
+	// lightenable * 4
+	// lighttype * 4
+	// lightcomp * 4
+	// uv gen:
+	// mapping type
+	// texshade light choices (ONLY IF uv mapping type is shade)
 }
 
 void WriteLight(char *p, int l) {
@@ -61,8 +70,11 @@ char *GenerateVertexShader()
 
 	int lmode = gstate.lmode & 1;
 
-	WRITE("attribute vec4 a_position;");
-	WRITE("attribute vec2 a_texcoord;");
+	int doTexture = (gstate.textureMapEnable & 1) && !(gstate.clearmode & 1);
+
+	WRITE("attribute vec3 a_position;");
+	if (doTexture)
+		WRITE("attribute vec2 a_texcoord;");
 	WRITE("attribute vec4 a_color0;");
 	if (lmode)
 		WRITE("attribute vec4 a_color1;");
@@ -77,18 +89,20 @@ char *GenerateVertexShader()
 	WRITE("varying vec4 v_color0;");
 	if (lmode)
 		WRITE("varying vec4 v_color1;");
-	WRITE("varying vec2 v_texcoord;");
+	if (doTexture)
+		WRITE("varying vec2 v_texcoord;");
 	if (gstate.isFogEnabled())
 		WRITE("varying float v_depth;");
 	WRITE("void main() {");
 	WRITE("  v_color0 = a_color0;");
 	if (lmode)
 		WRITE("  v_color1 = a_color1;");
-	WRITE("  v_texcoord = a_texcoord;");
+	if (doTexture)
+		WRITE("  v_texcoord = a_texcoord;");
 	if (gstate.isModeThrough())	{
-		WRITE("  gl_Position = u_proj_through * a_position;");
+		WRITE("  gl_Position = u_proj_through * vec4(a_position, 1.0);");
 	} else {
-		WRITE("  gl_Position = u_proj * a_position;");
+		WRITE("  gl_Position = u_proj * vec4(a_position, 1.0);");
 	}
 	if (gstate.isFogEnabled()) {
 		WRITE("  v_depth = gl_Position.z;");
diff --git a/GPU/GLES/VertexShaderGenerator.h b/GPU/GLES/VertexShaderGenerator.h
index 10a1fd1e58..4ce2bdb87b 100644
--- a/GPU/GLES/VertexShaderGenerator.h
+++ b/GPU/GLES/VertexShaderGenerator.h
@@ -46,7 +46,7 @@ struct VertexShaderID
 	}
 };
 
-void ComputeVertexShaderID(VertexShaderID *id);
+void ComputeVertexShaderID(VertexShaderID *id, int prim);
 
 // The return value is only valid until the function is called again.
 char *GenerateVertexShader();
diff --git a/GPU/GPUState.h b/GPU/GPUState.h
index 39adabc391..f11eb33bff 100644
--- a/GPU/GPUState.h
+++ b/GPU/GPUState.h
@@ -79,8 +79,8 @@ struct GPUgstate
 		u32 cmdmem[256];
 		struct
 		{
-			int nop;
-			u32 vaddr,
+			u32 nop,
+				vaddr,
 				iaddr,
 				pad00,
 				prim,
@@ -226,11 +226,9 @@ struct GPUgstate
 				transferstart,
 				transfersrcpos,
 				transferdstpos,
-				transfersize				
-				;
+				transfersize;
 
 			u32 pad05[0x63-0x40];
-
 		};
 	};
 
diff --git a/android/ab.sh b/android/ab.sh
old mode 100644
new mode 100755
diff --git a/android/jni/EmuScreen.cpp b/android/jni/EmuScreen.cpp
index 7fb7bf35bc..b21ab0b36c 100644
--- a/android/jni/EmuScreen.cpp
+++ b/android/jni/EmuScreen.cpp
@@ -16,6 +16,7 @@
 // https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
 
 #include "gfx_es2/glsl_program.h"
+#include "gfx_es2/gl_state.h"
 #include "gfx_es2/fbo.h"
 
 #include "input/input_state.h"
diff --git a/android/jni/NativeApp.cpp b/android/jni/NativeApp.cpp
index 4e7053ed7c..c4052bf84f 100644
--- a/android/jni/NativeApp.cpp
+++ b/android/jni/NativeApp.cpp
@@ -27,7 +27,7 @@
 #include "base/NativeApp.h"
 #include "file/vfs.h"
 #include "file/zip_read.h"
-#include "gfx/gl_state.h"
+#include "gfx_es2/gl_state.h"
 #include "gfx/gl_lost_manager.h"
 #include "gfx/texture.h"
 #include "input/input_state.h"
@@ -214,7 +214,7 @@ void NativeInit(int argc, const char *argv[], const char *savegame_directory, co
 	{
 		LogTypes::LOG_TYPE type = (LogTypes::LOG_TYPE)i;
 		logman->SetEnable(type, true);
-		logman->SetLogLevel(type, logLevel);
+		logman->SetLogLevel(type, gfxLog && i == LogTypes::G3D ? LogTypes::LDEBUG : logLevel);
 #ifdef ANDROID
 		logman->AddListener(type, logger);
 #endif
@@ -268,6 +268,7 @@ void NativeInitGraphics()
 
 void NativeRender()
 {
+	glstate.Restore();
 	glViewport(0, 0, pixel_xres, pixel_yres);
 	Matrix4x4 ortho;
 	ortho.setOrtho(0.0f, dp_xres, dp_yres, 0.0f, -1.0f, 1.0f);
diff --git a/native b/native
index 92bd22ffe5..7ed5811db3 160000
--- a/native
+++ b/native
@@ -1 +1 @@
-Subproject commit 92bd22ffe5fd90b1441660b9136fc6dc230399ec
+Subproject commit 7ed5811db38f3547f8eec3d3b577dae38405b1e0