[pull] master from hrydgard:master

Common/GPU/Shader.cpp

@@ -50,6 +50,7 @@ void ShaderLanguageDesc::Init(ShaderLanguage lang) {
 		lastFragData = nullptr;
 		gles = false;
 		forceMatrix4x4 = true;
+		depthMinusOneToOne = true;
 		break;
 	case GLSL_3xx:
 		// Just used in the shader test.
@@ -67,6 +68,7 @@ void ShaderLanguageDesc::Init(ShaderLanguage lang) {
 		gles = true;
 		forceMatrix4x4 = true;
 		glslES30 = true;
+		depthMinusOneToOne = true;
 		break;
 	case GLSL_VULKAN:
 		fragColor0 = "fragColor0";
@@ -86,6 +88,7 @@ void ShaderLanguageDesc::Init(ShaderLanguage lang) {
 		forceMatrix4x4 = false;
 		coefsFromBuffers = true;
 		vertexIndex = true;
+		depthMinusOneToOne = false;
 		break;
 	case HLSL_D3D11:
 		fragColor0 = "outfragment.target";
@@ -107,6 +110,7 @@ void ShaderLanguageDesc::Init(ShaderLanguage lang) {
 		coefsFromBuffers = true;
 		vsOutPrefix = "Out.";
 		viewportYSign = "-";
+		depthMinusOneToOne = false;
 		break;
 	}
 }

Common/GPU/Shader.h

@@ -57,6 +57,7 @@ struct ShaderLanguageDesc {
 	const char *vsOutPrefix = "";
 	const char *viewportYSign = "";
 
+	bool depthMinusOneToOne = false;
 	bool vertexIndex = false;
 	bool glslES30 = false;  // really glslES30Features. TODO: Clean this up.
 	bool bitwiseOps = false;

Common/GPU/ShaderWriter.cpp

@@ -361,6 +361,9 @@ void ShaderWriter::EndVSMain(Slice<VaryingDef> varyings) {
 		if (strlen(lang_.viewportYSign)) {
 			F("  gl_Position.y *= %s1.0;\n", lang_.viewportYSign);
 		}
+		if (lang_.depthMinusOneToOne) {
+			F("  gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");  // homogenous math... looks confusing.
+		}
 		C("  vs_out.pos = gl_Position;\n");
 		for (auto &varying : varyings) {
 			F("  vs_out.%s = %s;\n", varying.name, varying.name);

Common/GPU/Vulkan/VulkanContext.h

@@ -441,6 +441,10 @@ class VulkanContext {
 		return winsys_;
 	}
 
+	bool SupportsPreRotation() const {
+		return surfCapabilities_.supportedTransforms != VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR;
+	}
+
 private:
 	bool ChooseQueue();
 

Common/Log.h

@@ -57,6 +57,7 @@ enum class Log {
 	GeDebugger,
 	UI,
 	IAP,
+	CwCheats,
 
 	sceAudio,
 	sceCtrl,

Common/Log/LogManager.cpp

@@ -96,6 +96,7 @@ static const char * const g_logTypeNames[] = {
 	"GEDEBUGGER",
 	"UI",
 	"IAP",
+	"CWCHEATS",
 	"SCEAUDIO",
 	"SCECTRL",
 	"SCEDISP",

Common/Math/lin/matrix4x4.cpp

@@ -43,7 +43,7 @@ void Matrix4x4::setViewFrame(const Vec3 &pos, const Vec3 &vRight, const Vec3 &vV
 	ww = 1.0f;
 }
 
-void Matrix4x4::setOrtho(float left, float right, float bottom, float top, float near, float far) {
+void Matrix4x4::setOrthoGL(float left, float right, float bottom, float top, float near, float far) {
 	empty();
 	xx = 2.0f / (right - left);
 	yy = 2.0f / (top - bottom);

Common/Math/lin/matrix4x4.h

@@ -87,7 +87,7 @@ class Matrix4x4 {
 		ww = 1.0f;
 	}
 
-	void setOrtho(float left, float right, float bottom, float top, float near, float far);
+	void setOrthoGL(float left, float right, float bottom, float top, float near, float far);
 	void setOrthoD3D(float left, float right, float bottom, float top, float near, float far);
 	void setOrthoVulkan(float left, float right, float top, float bottom, float near, float far);
 

Common/System/Display.cpp

@@ -138,7 +138,7 @@ Lin::Matrix4x4 ComputeOrthoMatrix(float xres, float yres, CoordConvention coordC
 		break;
 	case CoordConvention::OpenGL:
 	default:
-		ortho.setOrtho(0.0f, xres, yres, 0.0f, -1.0f, 1.0f);
+		ortho.setOrthoGL(0.0f, xres, yres, 0.0f, -1.0f, 1.0f);
 		break;
 	}
 	// Compensate for rotated display if needed.

Common/UI/Notice.h

@@ -34,9 +34,6 @@ class NoticeView : public UI::InertView {
 		level_ = level;
 		text_ = text;
 	}
-	void SetSquishy(bool squishy) {
-		squishy_ = squishy;
-	}
 	void SetWrapText(bool wrapText) {
 		wrapText_ = wrapText;
 	}
@@ -53,8 +50,7 @@ class NoticeView : public UI::InertView {
 	std::string iconName_;
 	NoticeLevel level_;
 	mutable float height1_ = 0.0f;
-	bool squishy_ = false;
-	bool wrapText_ = false;
+	bool wrapText_ = true;
 };
 
 ImageID GetOSDIcon(NoticeLevel level);

Common/UI/PopupScreens.cpp

@@ -816,6 +816,7 @@ void AskForInput(ScreenManager *screenManager, RequesterToken token, UI::View *s
 
 PopupTextInputChoice::PopupTextInputChoice(RequesterToken token, std::string *value, std::string_view title, std::string_view placeholder, int maxLen, ScreenManager *screenManager, LayoutParams *layoutParams)
 	: AbstractChoiceWithValueDisplay(title, layoutParams), screenManager_(screenManager), value_(value), placeHolder_(placeholder), maxLen_(maxLen), token_(token), restriction_(StringRestriction::None) {
+	_dbg_assert_(value);
 	OnClick.Handle(this, &PopupTextInputChoice::HandleClick);
 }
 

GPU/Common/FramebufferManagerCommon.cpp

@@ -3185,9 +3185,9 @@ bool GetOutputFramebuffer(Draw::DrawContext *draw, GPUDebugBuffer &buffer) {
 	if (fmt != Draw::DataFormat::B8G8R8A8_UNORM)
 		fmt = Draw::DataFormat::R8G8B8A8_UNORM;
 
-	bool flipped = g_Config.iGPUBackend == (int)GPUBackend::OPENGL;
+	bool flipY = g_Config.iGPUBackend == (int)GPUBackend::OPENGL;
 
-	buffer.Allocate(w, h, fmt == Draw::DataFormat::R8G8B8A8_UNORM ? GPU_DBG_FORMAT_8888 : GPU_DBG_FORMAT_8888_BGRA, flipped);
+	buffer.Allocate(w, h, fmt == Draw::DataFormat::R8G8B8A8_UNORM ? GPU_DBG_FORMAT_8888 : GPU_DBG_FORMAT_8888_BGRA, flipY);
 	buffer.SetIsBackbuffer(true);
 	return draw->CopyFramebufferToMemory(nullptr, Draw::Aspect::COLOR_BIT, 0, 0, w, h, fmt, buffer.GetData(), w, Draw::ReadbackMode::BLOCK, "GetOutputFramebuffer");
 }

GPU/Common/ShaderUniforms.cpp

@@ -14,18 +14,12 @@
 
 using namespace Lin;
 
-static void ConvertProjMatrixToVulkan(Matrix4x4 &in) {
+static void ConvertProjMatrixToZeroToOneDepth(Matrix4x4 &in) {
 	const Vec3 trans(gstate_c.vpXOffset, gstate_c.vpYOffset, gstate_c.vpZOffset * 0.5f + 0.5f);
 	const Vec3 scale(gstate_c.vpWidthScale, gstate_c.vpHeightScale, gstate_c.vpDepthScale * 0.5f);
 	in.translateAndScale(trans, scale);
 }
 
-static void ConvertProjMatrixToD3D11(Matrix4x4 &in) {
-	const Vec3 trans(gstate_c.vpXOffset, -gstate_c.vpYOffset, gstate_c.vpZOffset * 0.5f + 0.5f);
-	const Vec3 scale(gstate_c.vpWidthScale, -gstate_c.vpHeightScale, gstate_c.vpDepthScale * 0.5f);
-	in.translateAndScale(trans, scale);
-}
-
 void CalcCullRange(float minValues[4], float maxValues[4], bool flipViewport, bool hasNegZ) {
 	// Account for the projection viewport adjustment when viewport is too large.
 	auto reverseViewportX = [](float x) {
@@ -72,6 +66,40 @@ void CalcCullRange(float minValues[4], float maxValues[4], bool flipViewport, bo
 	maxValues[3] = NAN;
 }
 
+// TODO: This will be removed later.
+static inline void FlipProjMatrix(Matrix4x4 & in) {
+	const bool invertedY = gstate_c.vpHeight < 0;
+	if (invertedY) {
+		in[1] = -in[1];
+		in[5] = -in[5];
+		in[9] = -in[9];
+		in[13] = -in[13];
+	}
+	const bool invertedX = gstate_c.vpWidth < 0;
+	if (invertedX) {
+		in[0] = -in[0];
+		in[4] = -in[4];
+		in[8] = -in[8];
+		in[12] = -in[12];
+	}
+}
+
+void UpdateRotation(float rotMatrix[4], bool useBufferedRendering) {
+	if (!gstate_c.Use(GPU_USE_PRE_ROTATION) || useBufferedRendering) {
+		rotMatrix[0] = 1.0f;
+		rotMatrix[1] = 0.0f;
+		rotMatrix[2] = 0.0f;
+		rotMatrix[3] = 1.0f;
+	} else {
+		const DisplayRotation rotation = g_display.rotation;
+		// The others are ROTATE_90 and so on.
+		rotMatrix[0] = rotation == DisplayRotation::ROTATE_0 ? 1.0 : (rotation == DisplayRotation::ROTATE_180 ? -1.0 : 0.0);
+		rotMatrix[1] = rotation == DisplayRotation::ROTATE_90 ? 1.0 : (rotation == DisplayRotation::ROTATE_270 ? -1.0 : 0.0);
+		rotMatrix[2] = rotation == DisplayRotation::ROTATE_270 ? 1.0 : (rotation == DisplayRotation::ROTATE_90 ? -1.0 : 0.0);
+		rotMatrix[3] = rotation == DisplayRotation::ROTATE_0 ? 1.0 : (rotation == DisplayRotation::ROTATE_180 ? -1.0 : 0.0);
+	}
+}
+
 void BaseUpdateUniforms(UB_VS_FS_Base *ub, uint64_t dirtyUniforms, bool flipViewport, bool useBufferedRendering) {
 	if (dirtyUniforms & DIRTY_TEXENV) {
 		Uint8x3ToFloat3(ub->texEnvColor, gstate.texenvcolor);
@@ -115,44 +143,16 @@ void BaseUpdateUniforms(UB_VS_FS_Base *ub, uint64_t dirtyUniforms, bool flipView
 		Matrix4x4 flippedMatrix;
 		memcpy(&flippedMatrix, gstate.projMatrix, 16 * sizeof(float));
 
-		const bool invertedY = gstate_c.vpHeight < 0;
-		if (invertedY) {
-			flippedMatrix[1] = -flippedMatrix[1];
-			flippedMatrix[5] = -flippedMatrix[5];
-			flippedMatrix[9] = -flippedMatrix[9];
-			flippedMatrix[13] = -flippedMatrix[13];
-		}
-		const bool invertedX = gstate_c.vpWidth < 0;
-		if (invertedX) {
-			flippedMatrix[0] = -flippedMatrix[0];
-			flippedMatrix[4] = -flippedMatrix[4];
-			flippedMatrix[8] = -flippedMatrix[8];
-			flippedMatrix[12] = -flippedMatrix[12];
-		}
-		if (flipViewport) {
-			ConvertProjMatrixToD3D11(flippedMatrix);
-		} else {
-			ConvertProjMatrixToVulkan(flippedMatrix);
-		}
-
-		if (!useBufferedRendering && g_display.rotation != DisplayRotation::ROTATE_0) {
-			flippedMatrix = flippedMatrix * g_display.rot_matrix;
-		}
+		FlipProjMatrix(flippedMatrix);
+		ConvertProjMatrixToZeroToOneDepth(flippedMatrix);
 		CopyMatrix4x4(ub->proj, flippedMatrix.getReadPtr());
 
 		ub->rotation = useBufferedRendering ? 0 : (float)g_display.rotation;
 	}
 
 	if (dirtyUniforms & DIRTY_PROJTHROUGHMATRIX) {
 		Matrix4x4 proj_through;
-		if (flipViewport) {
-			proj_through.setOrthoD3D(0.0f, gstate_c.curRTWidth, gstate_c.curRTHeight, 0, 0, 1);
-		} else {
-			proj_through.setOrthoVulkan(0.0f, gstate_c.curRTWidth, 0, gstate_c.curRTHeight, 0, 1);
-		}
-		if (!useBufferedRendering && g_display.rotation != DisplayRotation::ROTATE_0) {
-			proj_through = proj_through * g_display.rot_matrix;
-		}
+		proj_through.setOrthoVulkan(0.0f, gstate_c.curRTWidth, 0, gstate_c.curRTHeight, 0, 1);
 
 		// Negative RT offsets come from split framebuffers (Killzone)
 		if (gstate_c.curRTOffsetX < 0 || gstate_c.curRTOffsetY < 0) {
@@ -161,6 +161,8 @@ void BaseUpdateUniforms(UB_VS_FS_Base *ub, uint64_t dirtyUniforms, bool flipView
 		}
 
 		CopyMatrix4x4(ub->proj_through, proj_through.getReadPtr());
+
+		ub->rotation = useBufferedRendering ? 0 : (float)g_display.rotation;
 	}
 
 	// Transform

GPU/Common/ShaderUniforms.h

@@ -17,7 +17,7 @@ enum : uint64_t {
 	DIRTY_MATDIFFUSE | DIRTY_MATSPECULAR | DIRTY_MATEMISSIVE | DIRTY_AMBIENT,
 };
 
-// Currently 496 bytes.
+// Currently 480 bytes.
 // Every line here is a 4-float.
 struct alignas(16) UB_VS_FS_Base {
 	float proj[16];
@@ -43,6 +43,7 @@ struct alignas(16) UB_VS_FS_Base {
 	// VR stuff is to go here, later. For normal drawing, we can then get away
 	// with just uploading the first 448 bytes of the struct (up to and including fogCoef).
 };
+static_assert(sizeof(UB_VS_FS_Base) == 480, "UB_VS_FS_Base should be 480 bytes");
 
 static const char * const ub_baseStr =
 R"(  mat4 u_proj;
@@ -84,6 +85,7 @@ struct alignas(16) UB_VS_Lights {
 	float lightDiffuse[4][4];
 	float lightSpecular[4][4];
 };
+static_assert(sizeof(UB_VS_Lights) == 512);  // it's ok to optimize this, it's just an assumption check.
 
 static const char * const ub_vs_lightsStr =
 R"(  vec4 u_ambient;
@@ -105,6 +107,7 @@ R"(  vec4 u_ambient;
 struct alignas(16) UB_VS_Bones {
 	float bones[8][12];
 };
+static_assert(sizeof(UB_VS_Bones) == 384);  // No way to optimize this further.
 
 static const char * const ub_vs_bonesStr =
 R"(	mat3x4 u_bone0; mat3x4 u_bone1; mat3x4 u_bone2; mat3x4 u_bone3; mat3x4 u_bone4; mat3x4 u_bone5; mat3x4 u_bone6; mat3x4 u_bone7; mat3x4 u_bone8;

GPU/Common/SoftwareTransformCommon.cpp

@@ -65,27 +65,15 @@ static void SwapUVs(TransformedVertex &a, TransformedVertex &b) {
 
 // Note: 0 is BR and 2 is TL.
 
-static void RotateUV(TransformedVertex v[4], bool flippedY) {
-	// We use the transformed tl/br coordinates to figure out whether they're flipped or not.
-	float ySign = flippedY ? -1.0 : 1.0;
-
+static void RotateUV(TransformedVertex v[4]) {
 	const float x1 = v[2].x;
 	const float x2 = v[0].x;
-	const float y1 = v[2].y * ySign;
-	const float y2 = v[0].y * ySign;
-
-	if ((x1 < x2 && y1 < y2) || (x1 > x2 && y1 > y2))
-		SwapUVs(v[1], v[3]);
-}
+	const float y1 = v[2].y;
+	const float y2 = v[0].y;
 
-static void RotateUVThrough(TransformedVertex v[4]) {
-	float x1 = v[2].x;
-	float x2 = v[0].x;
-	float y1 = v[2].y;
-	float y2 = v[0].y;
-
-	if ((x1 < x2 && y1 > y2) || (x1 > x2 && y1 < y2))
+	if ((x1 < x2 && y1 > y2) || (x1 > x2 && y1 < y2)) {
 		SwapUVs(v[1], v[3]);
+	}
 }
 
 // Clears on the PSP are best done by drawing a series of vertical strips
@@ -658,11 +646,7 @@ bool SoftwareTransform::ExpandRectangles(int vertexCount, int &numDecodedVerts,
 		trans[3].v = transVtxBR.v * vscale;
 
 		// That's the four corners. Now process UV rotation.
-		if (throughmode) {
-			RotateUVThrough(trans);
-		} else {
-			RotateUV(trans, params_.flippedY);
-		}
+		RotateUV(trans);
 
 		// Triangle: BR-TR-TL
 		indsOut[0] = i * 2 + 0;