diff --git a/include/renderer_mtl/mtl_draw_pipeline_cache.hpp b/include/renderer_mtl/mtl_draw_pipeline_cache.hpp index c4bf6e22..fec36b8f 100644 --- a/include/renderer_mtl/mtl_draw_pipeline_cache.hpp +++ b/include/renderer_mtl/mtl_draw_pipeline_cache.hpp @@ -6,19 +6,21 @@ using namespace PICA; namespace Metal { -struct DrawPipelineHash { +struct DrawPipelineHash { // 62 bits // Formats - ColorFmt colorFmt; - DepthFmt depthFmt; + ColorFmt colorFmt; // 3 bits + DepthFmt depthFmt; // 3 bits // Blending - bool blendEnabled; - u32 blendControl; + bool blendEnabled; // 1 bit + u32 blendControl; // 32 bits - // Specialization constants - bool lightingEnabled; - u8 lightingNumLights; - u8 lightingConfig1; + // Specialization constants (23 bits) + bool lightingEnabled; // 1 bit + u8 lightingNumLights; // 3 bits + u8 lightingConfig1; // 7 bits + // | ref | func | on | + u16 alphaControl; // 12 bits (mask: 11111111 0111 0001) }; // Bind the vertex buffer to binding 30 so that it doesn't occupy the lower indices @@ -43,8 +45,8 @@ public: } MTL::RenderPipelineState* get(DrawPipelineHash hash) { - u16 fragmentFunctionHash = ((u8)hash.lightingEnabled << 12) | (hash.lightingNumLights << 8) | hash.lightingConfig1; - u64 pipelineHash = ((u64)hash.colorFmt << 52) | ((u64)hash.depthFmt << 49) | ((u64)hash.blendEnabled << 48) | ((u64)hash.blendControl << 16) | fragmentFunctionHash; + u32 fragmentFunctionHash = ((u32)hash.lightingEnabled << 22) | ((u32)hash.lightingNumLights << 19) | ((u32)hash.lightingConfig1 << 12) | ((((u32)hash.alphaControl & 0b1111111100000000) >> 8) << 4) | ((((u32)hash.alphaControl & 0b01110000) >> 4) << 1) | ((u32)hash.alphaControl & 0b0001); + u64 pipelineHash = ((u64)hash.colorFmt << 59) | ((u64)hash.depthFmt << 56) | ((u64)hash.blendEnabled << 55) | ((u64)hash.blendControl << 23) | fragmentFunctionHash; auto& pipeline = pipelineCache[pipelineHash]; if (!pipeline) { auto& fragmentFunction = fragmentFunctionCache[fragmentFunctionHash]; @@ -53,6 +55,7 @@ public: constants->setConstantValue(&hash.lightingEnabled, MTL::DataTypeBool, NS::UInteger(0)); constants->setConstantValue(&hash.lightingNumLights, MTL::DataTypeUChar, NS::UInteger(1)); constants->setConstantValue(&hash.lightingConfig1, MTL::DataTypeUChar, NS::UInteger(2)); + constants->setConstantValue(&hash.alphaControl, MTL::DataTypeUShort, NS::UInteger(3)); NS::Error* error = nullptr; fragmentFunction = library->newFunction(NS::String::string("fragmentDraw", NS::ASCIIStringEncoding), constants, &error); @@ -116,7 +119,7 @@ public: private: std::unordered_map pipelineCache; - std::unordered_map fragmentFunctionCache; + std::unordered_map fragmentFunctionCache; MTL::Device* device; MTL::Library* library; diff --git a/src/core/renderer_mtl/renderer_mtl.cpp b/src/core/renderer_mtl/renderer_mtl.cpp index 30a1fc72..e2384a79 100644 --- a/src/core/renderer_mtl/renderer_mtl.cpp +++ b/src/core/renderer_mtl/renderer_mtl.cpp @@ -385,8 +385,9 @@ void RendererMTL::drawVertices(PICA::PrimType primType, std::spanformat; } pipelineHash.lightingEnabled = regs[0x008F] & 1; - pipelineHash.lightingNumLights = (regs[0x01C2] & 0x7) + 1; + pipelineHash.lightingNumLights = regs[0x01C2] & 0x7; pipelineHash.lightingConfig1 = regs[0x01C4u] >> 16; // Last 16 bits are unused, so skip them + pipelineHash.alphaControl = regs[0x104]; // Blending and logic op pipelineHash.blendEnabled = (regs[PICA::InternalRegs::ColourOperation] & (1 << 8)) != 0; diff --git a/src/host_shaders/metal_shaders.metal b/src/host_shaders/metal_shaders.metal index 85401a74..1d4151c6 100644 --- a/src/host_shaders/metal_shaders.metal +++ b/src/host_shaders/metal_shaders.metal @@ -386,6 +386,7 @@ float3 regToColor(uint reg) { constant bool lightingEnabled [[function_constant(0)]]; constant uint8_t lightingNumLights [[function_constant(1)]]; constant uint8_t lightingConfig1 [[function_constant(2)]]; +constant uint16_t alphaControl [[function_constant(3)]]; // Implements the following algorthm: https://mathb.in/26766 void calcLighting(thread DrawVertexOut& in, constant PicaRegs& picaRegs, texture1d_array texLightingLut, sampler linearSampler, thread float4& primaryColor, thread float4& secondaryColor) { @@ -413,7 +414,7 @@ void calcLighting(thread DrawVertexOut& in, constant PicaRegs& picaRegs, texture bool errorUnimpl = false; - for (uint i = 0u; i < lightingNumLights; i++) { + for (uint i = 0u; i < lightingNumLights + 1; i++) { uint lightID = extract_bits(GPUREG_LIGHTING_LIGHT_PERMUTATION, int(i * 3u), 3); uint GPUREG_LIGHTi_SPECULAR0 = picaRegs.read(0x0140u + 0x10u * lightID); @@ -580,7 +581,6 @@ fragment float4 fragmentDraw(DrawVertexOut in [[stage_in]], float4 prevColor [[c float4 color = performLogicOp(logicOp, globals.tevSources[15], prevColor); // Perform alpha test - uint alphaControl = picaRegs.read(0x104u); if ((alphaControl & 1u) != 0u) { // Check if alpha test is on uint func = (alphaControl >> 4u) & 7u; float reference = float((alphaControl >> 8u) & 0xffu) / 255.0;