Geometry pipeline v2

Co-Authored-By: Sky <skylersaleh@gmail.com>

include/PICA/gpu.hpp

@@ -4,6 +4,7 @@
 #include "logger.hpp"
 #include "memory.hpp"
 #include "PICA/float_types.hpp"
+#include "PICA/regs.hpp"
 #include "PICA/shader_unit.hpp"
 #include "renderer_gl/renderer_gl.hpp"
 
@@ -19,6 +20,7 @@ class GPU {
 	static constexpr u32 regNum = 0x300;
 	static constexpr u32 vramSize = 6_MB;
 	std::array<u32, regNum> regs; // GPU internal registers
+	std::array<vec4f, 16> currentAttributes; // Vertex attributes before being passed to the shader
 
 	template <bool indexed>
 	void drawArrays();
@@ -31,12 +33,17 @@ class GPU {
 		int size = 0; // Bytes per vertex
 		u32 config1 = 0;
 		u32 config2 = 0;
+		u32 componentCount = 0; // Number of components for the attribute
 
 		u64 getConfigFull() {
 			return u64(config1) | (u64(config2) << 32);
 		}
 	};
 
+	u64 getVertexShaderInputConfig() {
+		return u64(regs[PICAInternalRegs::VertexShaderInputCfgLow]) | (u64(regs[PICAInternalRegs::VertexShaderInputCfgHigh]) << 32);
+	}
+
 	std::array<AttribInfo, maxAttribCount> attributeInfo; // Info for each of the 12 attributes
 	u32 totalAttribCount = 0; // Number of vertex attributes to send to VS
 	u32 fixedAttribMask = 0; // Which attributes are fixed?

include/PICA/regs.hpp

@@ -10,6 +10,7 @@ namespace PICAInternalRegs {
 
 		DepthScale = 0x4D,
 		DepthOffset = 0x4E,
+		ShaderOutputCount = 0x4F,
 
 		// Framebuffer registers
 		AlphaTestConfig = 0x104,
@@ -85,6 +86,10 @@ namespace PICAInternalRegs {
 		VertexFloatUniformData6 = 0x2C7,
 		VertexFloatUniformData7 = 0x2C8,
 
+		VertexShaderInputBufferCfg = 0x2B9,
+		VertexShaderInputCfgLow = 0x2BB,
+		VertexShaderInputCfgHigh = 0x2BC,
+
 		VertexShaderTransferIndex = 0x2CB,
 		VertexShaderData0 = 0x2CC,
 		VertexShaderData1 = 0x2CD,

include/PICA/shader.hpp

@@ -163,7 +163,7 @@ public:
 	std::array<vec4f, 96> floatUniforms;
 
 	std::array<vec4f, 16> fixedAttributes; // Fixed vertex attributes
-	std::array<vec4f, 16> attributes; // Attributes passed to the shader
+	std::array<vec4f, 16> inputs; // Attributes passed to the shader
 	std::array<vec4f, 16> outputs;
 
 	PICAShader(ShaderType type) : type(type) {}

include/helpers.hpp

@@ -120,10 +120,6 @@ namespace Helpers {
         static_for_impl<T, Begin>( std::forward<Func>(f), std::make_integer_sequence<T, End - Begin>{ } );
     }
 
-    static constexpr inline u8 get8BitColor (u8 colorRGB555) {
-        return (colorRGB555 << 3) | (colorRGB555 >> 2);
-    }
-
     // For values < 0x99
     static constexpr inline u8 incBCDByte(u8 value) {
         return ((value & 0xf) == 0x9) ? value + 7 : value + 1;

include/renderer_gl/renderer_gl.hpp

@@ -12,6 +12,7 @@ class GPU;
 struct Vertex {
 	OpenGL::vec4 position;
 	OpenGL::vec4 colour;
+	OpenGL::vec2 UVs;
 };
 
 class Renderer {

src/core/PICA/gpu.cpp

@@ -72,6 +72,10 @@ void GPU::drawArrays() {
 		log("PICA::DrawElements(vertex count = %d, index buffer config = %08X)\n", vertexCount, indexBufferConfig);
 	}
 
+	// Total number of input attributes to shader. Differs between GS and VS. Currently stubbed to the VS one, as we don't have geometry shaders.
+	const u32 inputAttrCount = (regs[PICAInternalRegs::VertexShaderInputBufferCfg] & 0xf) + 1;
+	const u64 inputAttrCfg = getVertexShaderInputConfig();
+		
 	for (u32 i = 0; i < vertexCount; i++) {
 		u32 vertexIndex; // Index of the vertex in the VBO
 
@@ -89,45 +93,71 @@ void GPU::drawArrays() {
 			}
 		}
 
-		int attrCount = 0; // Number of attributes we've passed to the shader
+		int attrCount = 0;
-		for (int attrCount = 0; attrCount < totalAttribCount; attrCount++) {
+		int buffer = 0; // Vertex buffer index for non-fixed attributes
+
+		while (attrCount < totalAttribCount) {
 			// Check if attribute is fixed or not
 			if (fixedAttribMask & (1 << attrCount)) { // Fixed attribute
 				vec4f& fixedAttr = shaderUnit.vs.fixedAttributes[attrCount]; // TODO: Is this how it works?
-				vec4f& inputAttr = shaderUnit.vs.attributes[attrCount];
+				vec4f& inputAttr = currentAttributes[attrCount];
 				std::memcpy(&inputAttr, &fixedAttr, sizeof(vec4f)); // Copy fixed attr to input attr
+				attrCount++;
 			} else { // Non-fixed attribute
-				auto& attr = attributeInfo[attrCount]; // Get information for this attribute
+				auto& attr = attributeInfo[buffer]; // Get information for this attribute
 				u64 attrCfg = attr.getConfigFull(); // Get config1 | (config2 << 32)
-				uint index = (attrCfg >> (attrCount * 4)) & 0xf; // Get index of attribute in vertexCfg
+				u32 attrAddress = vertexBase + attr.offset + (vertexIndex * attr.size);
 
+				for (int j = 0; j < attr.componentCount; j++) {
+					uint index = (attrCfg >> (j * 4)) & 0xf; // Get index of attribute in vertexCfg
 					if (index >= 12) Helpers::panic("[PICA] Vertex attribute used as padding");
 
 					u32 attribInfo = (vertexCfg >> (index * 4)) & 0xf;
 					u32 attribType = attribInfo & 0x3; //  Type of attribute(sbyte/ubyte/short/float)
-				u32 componentCount = (attribInfo >> 2) + 1; // Total number of components
+					u32 size = (attribInfo >> 2) + 1; // Total number of components
 
-				// Address to fetch the attribute from
+					//printf("vertex_attribute_strides[%d] = %d\n", attrCount, attr.size);
-				u32 attrAddress = vertexBase + attr.offset + (vertexIndex * attr.size);
+					vec4f& attribute = currentAttributes[attrCount];
-				vec4f& attribute = shaderUnit.vs.attributes[attrCount];
 					uint component; // Current component
 
 					switch (attribType) {
-					case 2: { // Short
+						case 0: { // Signed byte
-						s16* ptr = getPointerPhys<s16>(attrAddress);
+							s8* ptr = getPointerPhys<s8>(attrAddress);
-						for (component = 0; component < componentCount; component++) {
+							for (component = 0; component < size; component++) {
 								float val = static_cast<float>(*ptr++);
 								attribute[component] = f24::fromFloat32(val);
 							}
+							attrAddress += size * sizeof(s8);
+							break;
+						}
+
+						case 1: { // Unsigned byte
+							u8* ptr = getPointerPhys<u8>(attrAddress);
+							for (component = 0; component < size; component++) {
+								float val = static_cast<float>(*ptr++);
+								attribute[component] = f24::fromFloat32(val);
+							}
+							attrAddress += size * sizeof(u8);
+							break;
+						}
+
+						case 2: { // Short
+							s16* ptr = getPointerPhys<s16>(attrAddress);
+							for (component = 0; component < size; component++) {
+								float val = static_cast<float>(*ptr++);
+								attribute[component] = f24::fromFloat32(val);
+							}
+							attrAddress += size * sizeof(s16);
 							break;
 						}
 
 						case 3: { // Float
 							float* ptr = getPointerPhys<float>(attrAddress);
-						for (component = 0; component < componentCount; component++) {
+							for (component = 0; component < size; component++) {
 								float val = *ptr++;
 								attribute[component] = f24::fromFloat32(val);
 							}
+							attrAddress += size * sizeof(float);
 							break;
 						}
 
@@ -141,20 +171,34 @@ void GPU::drawArrays() {
 						attribute[component] = (component == 3) ? f24::fromFloat32(1.0) : f24::fromFloat32(0.0);
 						component++;
 					}
+
+					attrCount++;
 				}
+				buffer++;
+			}
+		}
+
+		// Before running the shader, the PICA maps the fetched attributes from the attribute registers to the shader input registers
+		// Based on the SH_ATTRIBUTES_PERMUTATION registers.
+		// Ie it might attribute #0 to v2, #1 to v7, etc
+		for (int j = 0; j < totalAttribCount; j++) {
+			const u32 mapping = (inputAttrCfg >> (j * 4)) & 0xf;
+			std::memcpy(&shaderUnit.vs.inputs[mapping], &currentAttributes[j], sizeof(vec4f));
 		}
 		
 		shaderUnit.vs.run();
 		std::memcpy(&vertices[i].position, &shaderUnit.vs.outputs[0], sizeof(vec4f));
 		std::memcpy(&vertices[i].colour, &shaderUnit.vs.outputs[1], sizeof(vec4f));
+		std::memcpy(&vertices[i].UVs, &shaderUnit.vs.outputs[2], 2 * sizeof(f24));
 
 		//printf("(x, y, z, w) = (%f, %f, %f, %f)\n", (double)vertices[i].position.x(), (double)vertices[i].position.y(), (double)vertices[i].position.z(), (double)vertices[i].position.w());
 		//printf("(r, g, b, a) = (%f, %f, %f, %f)\n", (double)vertices[i].colour.r(), (double)vertices[i].colour.g(), (double)vertices[i].colour.b(), (double)vertices[i].colour.a());
+		//printf("U: %f, V: %f\n", vertices[i].UVs.u(), vertices[i].UVs.v());
 	}
 
 	// The fourth type is meant to be "Geometry primitive". TODO: Find out what that is
 	static constexpr std::array<OpenGL::Primitives, 4> primTypes = {
-		OpenGL::Triangle, OpenGL::TriangleStrip, OpenGL::TriangleFan, OpenGL::LineStrip
+		OpenGL::Triangle, OpenGL::TriangleStrip, OpenGL::TriangleFan, OpenGL::Triangle
 	};
 	const auto shape = primTypes[primType];
 	renderer.drawVertices(shape, vertices, vertexCount);

src/core/PICA/shader_interpreter.cpp

@@ -94,7 +94,7 @@ u8 PICAShader::getIndexedSource(u32 source, u32 index) {
 
 PICAShader::vec4f PICAShader::getSource(u32 source) {
 	if (source < 0x10)
-		return attributes[source];
+		return inputs[source];
 	else if (source < 0x20)
 		return tempRegisters[source - 0x10];
 	else if (source <= 0x7f)
@@ -237,7 +237,6 @@ void PICAShader::mova(u32 instruction) {
 	const u32 operandDescriptor = operandDescriptors[instruction & 0x7f];
 	const u32 src = (instruction >> 12) & 0x7f;
 	const u32 idx = (instruction >> 19) & 3;
-	const u32 dest = (instruction >> 21) & 0x1f;
 
 	if (idx) Helpers::panic("[PICA] MOVA: idx != 0");
 	vec4f srcVector = getSourceSwizzled<1>(src, operandDescriptor);

src/core/PICA/shader_unit.cpp

@@ -18,7 +18,7 @@ void PICAShader::reset() {
 	f32UniformTransfer = false;
 
 	const vec4f zero = vec4f({ f24::zero(), f24::zero(), f24::zero(), f24::zero() });
-	attributes.fill(zero);
+	inputs.fill(zero);
 	floatUniforms.fill(zero);
 	outputs.fill(zero);
 	tempRegisters.fill(zero);

src/core/renderer_gl/renderer_gl.cpp

@@ -12,12 +12,15 @@ const char* vertexShader = R"(
 	
 	layout (location = 0) in vec4 coords;
 	layout (location = 1) in vec4 vertexColour;
+	layout (location = 2) in vec2 inUVs;
 
 	out vec4 colour;
+	out vec2 UVs;
 
 	void main() {
 		gl_Position = coords * vec4(1.0, 1.0, -1.0, 1.0);
 		colour = vertexColour;
+		UVs = inUVs;
 	}
 )";
 
@@ -25,12 +28,16 @@ const char* fragmentShader = R"(
 	#version 420 core
 	
 	in vec4 colour;
+	in vec2 UVs;
+
 	out vec4 fragColour;
 
 	uniform uint u_alphaControl;
+	uniform sampler2D u_tex0;
 
 	void main() {
-		fragColour = colour;
+		//fragColour = colour;
+		fragColour = texture(u_tex0, UVs);
 
 		if ((u_alphaControl & 1u) != 0u) { // Check if alpha test is on
 			uint func = (u_alphaControl >> 4u) & 7u;
@@ -127,6 +134,7 @@ void Renderer::initGraphicsContext() {
 	
 	alphaControlLoc = OpenGL::uniformLocation(triangleProgram, "u_alphaControl");
 	glUniform1ui(alphaControlLoc, 0); // Default alpha control to 0
+	glUniform1i(OpenGL::uniformLocation(triangleProgram, "u_tex0"), 0);
 
 	OpenGL::Shader vertDisplay(displayVertexShader, OpenGL::Vertex);
 	OpenGL::Shader fragDisplay(displayFragmentShader, OpenGL::Fragment);
@@ -143,6 +151,9 @@ void Renderer::initGraphicsContext() {
 	// Colour attribute
 	vao.setAttributeFloat<float>(1, 4, sizeof(Vertex), offsetof(Vertex, colour));
 	vao.enableAttribute(1);
+	// UV attribute
+	vao.setAttributeFloat<float>(2, 2, sizeof(Vertex), offsetof(Vertex, UVs));
+	vao.enableAttribute(2);
 
 	dummyVBO.create();
 	dummyVAO.create();
@@ -182,6 +193,8 @@ void Renderer::drawVertices(OpenGL::Primitives primType, Vertex* vertices, u32 c
 	f24 depthScale = f24::fromRaw(regs[PICAInternalRegs::DepthScale] & 0xffffff);
 	f24 depthOffset = f24::fromRaw(regs[PICAInternalRegs::DepthOffset] & 0xffffff);
 	printf("Depth enable: %d, func: %d, writeEnable: %d\n", depthEnable, depthFunc, depthWriteEnable);
+	printf("Blending enabled: %d\n", (regs[0x100] >> 8) & 1);
+	printf("Blend func: %08X\n", regs[0x101]);
 
 	//if (depthScale.toFloat32() != -1.0 || depthOffset.toFloat32() != 0.0)
 	//	Helpers::panic("TODO: Implement depth scale/offset. Remove the depth *= -1.0 from vertex shader");
@@ -194,6 +207,7 @@ void Renderer::drawVertices(OpenGL::Primitives primType, Vertex* vertices, u32 c
 
 		Texture targetTex(addr, static_cast<Texture::Formats>(format), width, height);
 		OpenGL::Texture tex = getTexture(targetTex);
+		tex.bind();
 	}
 
 	// TODO: Actually use this