Merge pull request #1264 from bunnei/fragment-lighting-hw

Fragment lighting support in the HW renderer

src/video_core/CMakeLists.txt

@@ -33,6 +33,7 @@ set(HEADERS
             command_processor.h
             gpu_debugger.h
             pica.h
+            pica_types.h
             primitive_assembly.h
             rasterizer.h
             rasterizer_interface.h

src/video_core/clipper.cpp

@@ -59,15 +59,17 @@ static void InitScreenCoordinates(OutputVertex& vtx)
     } viewport;
 
     const auto& regs = g_state.regs;
-    viewport.halfsize_x = float24::FromRawFloat24(regs.viewport_size_x);
+    viewport.halfsize_x = float24::FromRaw(regs.viewport_size_x);
-    viewport.halfsize_y = float24::FromRawFloat24(regs.viewport_size_y);
+    viewport.halfsize_y = float24::FromRaw(regs.viewport_size_y);
     viewport.offset_x   = float24::FromFloat32(static_cast<float>(regs.viewport_corner.x));
     viewport.offset_y   = float24::FromFloat32(static_cast<float>(regs.viewport_corner.y));
-    viewport.zscale     = float24::FromRawFloat24(regs.viewport_depth_range);
+    viewport.zscale     = float24::FromRaw(regs.viewport_depth_range);
-    viewport.offset_z   = float24::FromRawFloat24(regs.viewport_depth_far_plane);
+    viewport.offset_z   = float24::FromRaw(regs.viewport_depth_far_plane);
 
     float24 inv_w = float24::FromFloat32(1.f) / vtx.pos.w;
     vtx.color *= inv_w;
+    vtx.view *= inv_w;
+    vtx.quat *= inv_w;
     vtx.tc0 *= inv_w;
     vtx.tc1 *= inv_w;
     vtx.tc2 *= inv_w;

src/video_core/command_processor.cpp

@@ -98,10 +98,10 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
                 Math::Vec4<float24>& attribute = g_state.vs.default_attributes[setup.index];
 
                 // NOTE: The destination component order indeed is "backwards"
-                attribute.w = float24::FromRawFloat24(default_attr_write_buffer[0] >> 8);
+                attribute.w = float24::FromRaw(default_attr_write_buffer[0] >> 8);
-                attribute.z = float24::FromRawFloat24(((default_attr_write_buffer[0] & 0xFF) << 16) | ((default_attr_write_buffer[1] >> 16) & 0xFFFF));
+                attribute.z = float24::FromRaw(((default_attr_write_buffer[0] & 0xFF) << 16) | ((default_attr_write_buffer[1] >> 16) & 0xFFFF));
-                attribute.y = float24::FromRawFloat24(((default_attr_write_buffer[1] & 0xFFFF) << 8) | ((default_attr_write_buffer[2] >> 24) & 0xFF));
+                attribute.y = float24::FromRaw(((default_attr_write_buffer[1] & 0xFFFF) << 8) | ((default_attr_write_buffer[2] >> 24) & 0xFF));
-                attribute.x = float24::FromRawFloat24(default_attr_write_buffer[2] & 0xFFFFFF);
+                attribute.x = float24::FromRaw(default_attr_write_buffer[2] & 0xFFFFFF);
 
                 LOG_TRACE(HW_GPU, "Set default VS attribute %x to (%f %f %f %f)", (int)setup.index,
                           attribute.x.ToFloat32(), attribute.y.ToFloat32(), attribute.z.ToFloat32(),
@@ -418,10 +418,10 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
                         uniform[3 - i] = float24::FromFloat32(*(float*)(&uniform_write_buffer[i]));
                 } else {
                     // TODO: Untested
-                    uniform.w = float24::FromRawFloat24(uniform_write_buffer[0] >> 8);
+                    uniform.w = float24::FromRaw(uniform_write_buffer[0] >> 8);
-                    uniform.z = float24::FromRawFloat24(((uniform_write_buffer[0] & 0xFF)<<16) | ((uniform_write_buffer[1] >> 16) & 0xFFFF));
+                    uniform.z = float24::FromRaw(((uniform_write_buffer[0] & 0xFF) << 16) | ((uniform_write_buffer[1] >> 16) & 0xFFFF));
-                    uniform.y = float24::FromRawFloat24(((uniform_write_buffer[1] & 0xFFFF)<<8) | ((uniform_write_buffer[2] >> 24) & 0xFF));
+                    uniform.y = float24::FromRaw(((uniform_write_buffer[1] & 0xFFFF) << 8) | ((uniform_write_buffer[2] >> 24) & 0xFF));
-                    uniform.x = float24::FromRawFloat24(uniform_write_buffer[2] & 0xFFFFFF);
+                    uniform.x = float24::FromRaw(uniform_write_buffer[2] & 0xFFFFFF);
                 }
 
                 LOG_TRACE(HW_GPU, "Set uniform %x to (%f %f %f %f)", (int)uniform_setup.index,
@@ -464,6 +464,24 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
             break;
         }
 
+        case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[0], 0x1c8):
+        case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[1], 0x1c9):
+        case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[2], 0x1ca):
+        case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[3], 0x1cb):
+        case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[4], 0x1cc):
+        case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[5], 0x1cd):
+        case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[6], 0x1ce):
+        case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[7], 0x1cf):
+        {
+            auto& lut_config = regs.lighting.lut_config;
+
+            ASSERT_MSG(lut_config.index < 256, "lut_config.index exceeded maximum value of 255!");
+
+            g_state.lighting.luts[lut_config.type][lut_config.index].raw = value;
+            lut_config.index = lut_config.index + 1;
+            break;
+        }
+
         default:
             break;
     }

src/video_core/pica.h

@@ -16,6 +16,8 @@
 #include "common/vector_math.h"
 #include "common/logging/log.h"
 
+#include "pica_types.h"
+
 namespace Pica {
 
 // Returns index corresponding to the Regs member labeled by field_name
@@ -239,7 +241,8 @@ struct Regs {
     TextureConfig texture0;
     INSERT_PADDING_WORDS(0x8);
     BitField<0, 4, TextureFormat> texture0_format;
-    INSERT_PADDING_WORDS(0x2);
+    BitField<0, 1, u32> fragment_lighting_enable;
+    INSERT_PADDING_WORDS(0x1);
     TextureConfig texture1;
     BitField<0, 4, TextureFormat> texture1_format;
     INSERT_PADDING_WORDS(0x2);
@@ -641,7 +644,268 @@ struct Regs {
         }
     }
 
-    INSERT_PADDING_WORDS(0xe0);
+    INSERT_PADDING_WORDS(0x20);
+
+    enum class LightingSampler {
+        Distribution0 = 0,
+        Distribution1 = 1,
+        Fresnel = 3,
+        ReflectBlue = 4,
+        ReflectGreen = 5,
+        ReflectRed = 6,
+        SpotlightAttenuation = 8,
+        DistanceAttenuation = 16,
+    };
+
+    /**
+     * Pica fragment lighting supports using different LUTs for each lighting component:
+     * Reflectance R, G, and B channels, distribution function for specular components 0 and 1,
+     * fresnel factor, and spotlight attenuation. Furthermore, which LUTs are used for each channel
+     * (or whether a channel is enabled at all) is specified by various pre-defined lighting
+     * configurations. With configurations that require more LUTs, more cycles are required on HW to
+     * perform lighting computations.
+     */
+    enum class LightingConfig {
+        Config0 = 0, ///< Reflect Red, Distribution 0, Spotlight
+        Config1 = 1, ///< Reflect Red, Fresnel, Spotlight
+        Config2 = 2, ///< Reflect Red, Distribution 0/1
+        Config3 = 3, ///< Distribution 0/1, Fresnel
+        Config4 = 4, ///< Reflect Red/Green/Blue, Distribution 0/1, Spotlight
+        Config5 = 5, ///< Reflect Red/Green/Blue, Distribution 0, Fresnel, Spotlight
+        Config6 = 6, ///< Reflect Red, Distribution 0/1, Fresnel, Spotlight
+        Config7 = 8, ///< Reflect Red/Green/Blue, Distribution 0/1, Fresnel, Spotlight
+                     ///< NOTE: '8' is intentional, '7' does not appear to be a valid configuration
+    };
+
+    /// Selects which lighting components are affected by fresnel
+    enum class LightingFresnelSelector {
+        None = 0,                             ///< Fresnel is disabled
+        PrimaryAlpha = 1,                     ///< Primary (diffuse) lighting alpha is affected by fresnel
+        SecondaryAlpha = 2,                   ///< Secondary (specular) lighting alpha is affected by fresnel
+        Both = PrimaryAlpha | SecondaryAlpha, ///< Both primary and secondary lighting alphas are affected by fresnel
+    };
+
+    /// Factor used to scale the output of a lighting LUT
+    enum class LightingScale {
+        Scale1 = 0,   ///< Scale is 1x
+        Scale2 = 1,   ///< Scale is 2x
+        Scale4 = 2,   ///< Scale is 4x
+        Scale8 = 3,   ///< Scale is 8x
+        Scale1_4 = 6, ///< Scale is 0.25x
+        Scale1_2 = 7, ///< Scale is 0.5x
+    };
+
+    enum class LightingLutInput {
+        NH = 0, // Cosine of the angle between the normal and half-angle vectors
+        VH = 1, // Cosine of the angle between the view and half-angle vectors
+        NV = 2, // Cosine of the angle between the normal and the view vector
+        LN = 3, // Cosine of the angle between the light and the normal vectors
+    };
+
+    enum class LightingBumpMode : u32 {
+        None = 0,
+        NormalMap = 1,
+        TangentMap = 2,
+    };
+
+    union LightColor {
+        BitField< 0, 10, u32> b;
+        BitField<10, 10, u32> g;
+        BitField<20, 10, u32> r;
+
+        Math::Vec3f ToVec3f() const {
+            // These fields are 10 bits wide, however 255 corresponds to 1.0f for each color component
+            return Math::MakeVec((f32)r / 255.f, (f32)g / 255.f, (f32)b / 255.f);
+        }
+    };
+
+    /// Returns true if the specified lighting sampler is supported by the current Pica lighting configuration
+    static bool IsLightingSamplerSupported(LightingConfig config, LightingSampler sampler) {
+        switch (sampler) {
+        case LightingSampler::Distribution0:
+            return (config != LightingConfig::Config1);
+
+        case LightingSampler::Distribution1:
+            return (config != LightingConfig::Config0) && (config != LightingConfig::Config1) && (config != LightingConfig::Config5);
+
+        case LightingSampler::Fresnel:
+            return (config != LightingConfig::Config0) && (config != LightingConfig::Config2) && (config != LightingConfig::Config4);
+
+        case LightingSampler::ReflectRed:
+            return (config != LightingConfig::Config3);
+
+        case LightingSampler::ReflectGreen:
+        case LightingSampler::ReflectBlue:
+            return (config == LightingConfig::Config4) || (config == LightingConfig::Config5) || (config == LightingConfig::Config7);
+        }
+        return false;
+    }
+
+    struct {
+        struct LightSrc {
+            LightColor specular_0;  // material.specular_0 * light.specular_0
+            LightColor specular_1;  // material.specular_1 * light.specular_1
+            LightColor diffuse;     // material.diffuse * light.diffuse
+            LightColor ambient;     // material.ambient * light.ambient
+
+            struct {
+                // Encoded as 16-bit floating point
+                union {
+                    BitField< 0, 16, u32> x;
+                    BitField<16, 16, u32> y;
+                };
+                union {
+                    BitField< 0, 16, u32> z;
+                };
+
+                INSERT_PADDING_WORDS(0x3);
+
+                union {
+                    BitField<0, 1, u32> directional;
+                    BitField<1, 1, u32> two_sided_diffuse; // When disabled, clamp dot-product to 0
+                };
+            };
+
+            BitField<0, 20, u32> dist_atten_bias;
+            BitField<0, 20, u32> dist_atten_scale;
+
+            INSERT_PADDING_WORDS(0x4);
+        };
+        static_assert(sizeof(LightSrc) == 0x10 * sizeof(u32), "LightSrc structure must be 0x10 words");
+
+        LightSrc light[8];
+        LightColor global_ambient; // Emission + (material.ambient * lighting.ambient)
+        INSERT_PADDING_WORDS(0x1);
+        BitField<0, 3, u32> num_lights; // Number of enabled lights - 1
+
+        union {
+            BitField< 2, 2, LightingFresnelSelector> fresnel_selector;
+            BitField< 4, 4, LightingConfig> config;
+            BitField<22, 2, u32> bump_selector; // 0: Texture 0, 1: Texture 1, 2: Texture 2
+            BitField<27, 1, u32> clamp_highlights;
+            BitField<28, 2, LightingBumpMode> bump_mode;
+            BitField<30, 1, u32> disable_bump_renorm;
+        };
+
+        union {
+            BitField<16, 1, u32> disable_lut_d0;
+            BitField<17, 1, u32> disable_lut_d1;
+            BitField<19, 1, u32> disable_lut_fr;
+            BitField<20, 1, u32> disable_lut_rr;
+            BitField<21, 1, u32> disable_lut_rg;
+            BitField<22, 1, u32> disable_lut_rb;
+
+            // Each bit specifies whether distance attenuation should be applied for the
+            // corresponding light
+
+            BitField<24, 1, u32> disable_dist_atten_light_0;
+            BitField<25, 1, u32> disable_dist_atten_light_1;
+            BitField<26, 1, u32> disable_dist_atten_light_2;
+            BitField<27, 1, u32> disable_dist_atten_light_3;
+            BitField<28, 1, u32> disable_dist_atten_light_4;
+            BitField<29, 1, u32> disable_dist_atten_light_5;
+            BitField<30, 1, u32> disable_dist_atten_light_6;
+            BitField<31, 1, u32> disable_dist_atten_light_7;
+        };
+
+        bool IsDistAttenDisabled(unsigned index) const {
+            const unsigned disable[] = { disable_dist_atten_light_0, disable_dist_atten_light_1,
+                                         disable_dist_atten_light_2, disable_dist_atten_light_3,
+                                         disable_dist_atten_light_4, disable_dist_atten_light_5,
+                                         disable_dist_atten_light_6, disable_dist_atten_light_7 };
+            return disable[index] != 0;
+        }
+
+        union {
+            BitField<0, 8, u32> index;      ///< Index at which to set data in the LUT
+            BitField<8, 5, u32> type;       ///< Type of LUT for which to set data
+        } lut_config;
+
+        BitField<0, 1, u32> disable;
+        INSERT_PADDING_WORDS(0x1);
+
+        // When data is written to any of these registers, it gets written to the lookup table of
+        // the selected type at the selected index, specified above in the `lut_config` register.
+        // With each write, `lut_config.index` is incremented. It does not matter which of these
+        // registers is written to, the behavior will be the same.
+        u32 lut_data[8];
+
+        // These are used to specify if absolute (abs) value should be used for each LUT index. When
+        // abs mode is disabled, LUT indexes are in the range of (-1.0, 1.0). Otherwise, they are in
+        // the range of (0.0, 1.0).
+        union {
+            BitField< 1, 1, u32> disable_d0;
+            BitField< 5, 1, u32> disable_d1;
+            BitField< 9, 1, u32> disable_sp;
+            BitField<13, 1, u32> disable_fr;
+            BitField<17, 1, u32> disable_rb;
+            BitField<21, 1, u32> disable_rg;
+            BitField<25, 1, u32> disable_rr;
+        } abs_lut_input;
+
+        union {
+            BitField< 0, 3, LightingLutInput> d0;
+            BitField< 4, 3, LightingLutInput> d1;
+            BitField< 8, 3, LightingLutInput> sp;
+            BitField<12, 3, LightingLutInput> fr;
+            BitField<16, 3, LightingLutInput> rb;
+            BitField<20, 3, LightingLutInput> rg;
+            BitField<24, 3, LightingLutInput> rr;
+        } lut_input;
+
+        union {
+            BitField< 0, 3, LightingScale> d0;
+            BitField< 4, 3, LightingScale> d1;
+            BitField< 8, 3, LightingScale> sp;
+            BitField<12, 3, LightingScale> fr;
+            BitField<16, 3, LightingScale> rb;
+            BitField<20, 3, LightingScale> rg;
+            BitField<24, 3, LightingScale> rr;
+
+            static float GetScale(LightingScale scale) {
+                switch (scale) {
+                case LightingScale::Scale1:
+                    return 1.0f;
+                case LightingScale::Scale2:
+                    return 2.0f;
+                case LightingScale::Scale4:
+                    return 4.0f;
+                case LightingScale::Scale8:
+                    return 8.0f;
+                case LightingScale::Scale1_4:
+                    return 0.25f;
+                case LightingScale::Scale1_2:
+                    return 0.5f;
+                }
+                return 0.0f;
+            }
+        } lut_scale;
+
+        INSERT_PADDING_WORDS(0x6);
+
+        union {
+            // There are 8 light enable "slots", corresponding to the total number of lights
+            // supported by Pica. For N enabled lights (specified by register 0x1c2, or 'src_num'
+            // above), the first N slots below will be set to integers within the range of 0-7,
+            // corresponding to the actual light that is enabled for each slot.
+
+            BitField< 0, 3, u32> slot_0;
+            BitField< 4, 3, u32> slot_1;
+            BitField< 8, 3, u32> slot_2;
+            BitField<12, 3, u32> slot_3;
+            BitField<16, 3, u32> slot_4;
+            BitField<20, 3, u32> slot_5;
+            BitField<24, 3, u32> slot_6;
+            BitField<28, 3, u32> slot_7;
+
+            unsigned GetNum(unsigned index) const {
+                const unsigned enable_slots[] = { slot_0, slot_1, slot_2, slot_3, slot_4, slot_5, slot_6, slot_7 };
+                return enable_slots[index];
+            }
+        } light_enable;
+    } lighting;
+
+    INSERT_PADDING_WORDS(0x26);
 
     enum class VertexAttributeFormat : u64 {
         BYTE = 0,
@@ -990,6 +1254,7 @@ ASSERT_REG_POSITION(viewport_corner, 0x68);
 ASSERT_REG_POSITION(texture0_enable, 0x80);
 ASSERT_REG_POSITION(texture0, 0x81);
 ASSERT_REG_POSITION(texture0_format, 0x8e);
+ASSERT_REG_POSITION(fragment_lighting_enable, 0x8f);
 ASSERT_REG_POSITION(texture1, 0x91);
 ASSERT_REG_POSITION(texture1_format, 0x96);
 ASSERT_REG_POSITION(texture2, 0x99);
@@ -1004,6 +1269,7 @@ ASSERT_REG_POSITION(tev_stage5, 0xf8);
 ASSERT_REG_POSITION(tev_combiner_buffer_color, 0xfd);
 ASSERT_REG_POSITION(output_merger, 0x100);
 ASSERT_REG_POSITION(framebuffer, 0x110);
+ASSERT_REG_POSITION(lighting, 0x140);
 ASSERT_REG_POSITION(vertex_attributes, 0x200);
 ASSERT_REG_POSITION(index_array, 0x227);
 ASSERT_REG_POSITION(num_vertices, 0x228);
@@ -1026,118 +1292,6 @@ static_assert(sizeof(Regs::ShaderConfig) == 0x30 * sizeof(u32), "ShaderConfig st
 static_assert(sizeof(Regs) <= 0x300 * sizeof(u32), "Register set structure larger than it should be");
 static_assert(sizeof(Regs) >= 0x300 * sizeof(u32), "Register set structure smaller than it should be");
 
-struct float24 {
-    static float24 FromFloat32(float val) {
-        float24 ret;
-        ret.value = val;
-        return ret;
-    }
-
-    // 16 bit mantissa, 7 bit exponent, 1 bit sign
-    // TODO: No idea if this works as intended
-    static float24 FromRawFloat24(u32 hex) {
-        float24 ret;
-        if ((hex & 0xFFFFFF) == 0) {
-            ret.value = 0;
-        } else {
-            u32 mantissa = hex & 0xFFFF;
-            u32 exponent = (hex >> 16) & 0x7F;
-            u32 sign = hex >> 23;
-            ret.value = std::pow(2.0f, (float)exponent-63.0f) * (1.0f + mantissa * std::pow(2.0f, -16.f));
-            if (sign)
-                ret.value = -ret.value;
-        }
-        return ret;
-    }
-
-    static float24 Zero() {
-        return FromFloat32(0.f);
-    }
-
-    // Not recommended for anything but logging
-    float ToFloat32() const {
-        return value;
-    }
-
-    float24 operator * (const float24& flt) const {
-        if ((this->value == 0.f && !std::isnan(flt.value)) ||
-            (flt.value == 0.f && !std::isnan(this->value)))
-            // PICA gives 0 instead of NaN when multiplying by inf
-            return Zero();
-        return float24::FromFloat32(ToFloat32() * flt.ToFloat32());
-    }
-
-    float24 operator / (const float24& flt) const {
-        return float24::FromFloat32(ToFloat32() / flt.ToFloat32());
-    }
-
-    float24 operator + (const float24& flt) const {
-        return float24::FromFloat32(ToFloat32() + flt.ToFloat32());
-    }
-
-    float24 operator - (const float24& flt) const {
-        return float24::FromFloat32(ToFloat32() - flt.ToFloat32());
-    }
-
-    float24& operator *= (const float24& flt) {
-        if ((this->value == 0.f && !std::isnan(flt.value)) ||
-            (flt.value == 0.f && !std::isnan(this->value)))
-            // PICA gives 0 instead of NaN when multiplying by inf
-            *this = Zero();
-        else value *= flt.ToFloat32();
-        return *this;
-    }
-
-    float24& operator /= (const float24& flt) {
-        value /= flt.ToFloat32();
-        return *this;
-    }
-
-    float24& operator += (const float24& flt) {
-        value += flt.ToFloat32();
-        return *this;
-    }
-
-    float24& operator -= (const float24& flt) {
-        value -= flt.ToFloat32();
-        return *this;
-    }
-
-    float24 operator - () const {
-        return float24::FromFloat32(-ToFloat32());
-    }
-
-    bool operator < (const float24& flt) const {
-        return ToFloat32() < flt.ToFloat32();
-    }
-
-    bool operator > (const float24& flt) const {
-        return ToFloat32() > flt.ToFloat32();
-    }
-
-    bool operator >= (const float24& flt) const {
-        return ToFloat32() >= flt.ToFloat32();
-    }
-
-    bool operator <= (const float24& flt) const {
-        return ToFloat32() <= flt.ToFloat32();
-    }
-
-    bool operator == (const float24& flt) const {
-        return ToFloat32() == flt.ToFloat32();
-    }
-
-    bool operator != (const float24& flt) const {
-        return ToFloat32() != flt.ToFloat32();
-    }
-
-private:
-    // Stored as a regular float, merely for convenience
-    // TODO: Perform proper arithmetic on this!
-    float value;
-};
-static_assert(sizeof(float24) == sizeof(float), "Shader JIT assumes float24 is implemented as a 32-bit float");
-
 /// Struct used to describe current Pica state
 struct State {
     /// Pica registers
@@ -1163,6 +1317,25 @@ struct State {
     ShaderSetup vs;
     ShaderSetup gs;
 
+    struct {
+        union LutEntry {
+            // Used for raw access
+            u32 raw;
+
+            // LUT value, encoded as 12-bit fixed point, with 12 fraction bits
+            BitField< 0, 12, u32> value;
+
+            // Used by HW for efficient interpolation, Citra does not use these
+            BitField<12, 12, u32> difference;
+
+            float ToFloat() {
+                return static_cast<float>(value) / 4095.f;
+            }
+        };
+
+        std::array<std::array<LutEntry, 256>, 24> luts;
+    } lighting;
+
     /// Current Pica command list
     struct {
         const u32* head_ptr;

src/video_core/pica_types.h

@@ -0,0 +1,146 @@
+// Copyright 2015 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <cstring>
+
+#include "common/common_types.h"
+
+namespace Pica {
+
+/**
+ * Template class for converting arbitrary Pica float types to IEEE 754 32-bit single-precision
+ * floating point.
+ *
+ * When decoding, format is as follows:
+ *  - The first `M` bits are the mantissa
+ *  - The next `E` bits are the exponent
+ *  - The last bit is the sign bit
+ *
+ * @todo Verify on HW if this conversion is sufficiently accurate.
+ */
+template<unsigned M, unsigned E>
+struct Float {
+public:
+    static Float<M, E> FromFloat32(float val) {
+        Float<M, E> ret;
+        ret.value = val;
+        return ret;
+    }
+
+    static Float<M, E> FromRaw(u32 hex) {
+        Float<M, E> res;
+
+        const int width = M + E + 1;
+        const int bias = 128 - (1 << (E - 1));
+        const int exponent = (hex >> M) & ((1 << E) - 1);
+        const unsigned mantissa = hex & ((1 << M) - 1);
+
+        if (hex & ((1 << (width - 1)) - 1))
+            hex = ((hex >> (E + M)) << 31) | (mantissa << (23 - M)) | ((exponent + bias) << 23);
+        else
+            hex = ((hex >> (E + M)) << 31);
+
+        std::memcpy(&res.value, &hex, sizeof(float));
+
+        return res;
+    }
+
+    static Float<M, E> Zero() {
+        return FromFloat32(0.f);
+    }
+
+    // Not recommended for anything but logging
+    float ToFloat32() const {
+        return value;
+    }
+
+    Float<M, E> operator * (const Float<M, E>& flt) const {
+        if ((this->value == 0.f && !std::isnan(flt.value)) ||
+            (flt.value == 0.f && !std::isnan(this->value)))
+            // PICA gives 0 instead of NaN when multiplying by inf
+            return Zero();
+        return Float<M, E>::FromFloat32(ToFloat32() * flt.ToFloat32());
+    }
+
+    Float<M, E> operator / (const Float<M, E>& flt) const {
+        return Float<M, E>::FromFloat32(ToFloat32() / flt.ToFloat32());
+    }
+
+    Float<M, E> operator + (const Float<M, E>& flt) const {
+        return Float<M, E>::FromFloat32(ToFloat32() + flt.ToFloat32());
+    }
+
+    Float<M, E> operator - (const Float<M, E>& flt) const {
+        return Float<M, E>::FromFloat32(ToFloat32() - flt.ToFloat32());
+    }
+
+    Float<M, E>& operator *= (const Float<M, E>& flt) {
+        if ((this->value == 0.f && !std::isnan(flt.value)) ||
+            (flt.value == 0.f && !std::isnan(this->value)))
+            // PICA gives 0 instead of NaN when multiplying by inf
+            *this = Zero();
+        else value *= flt.ToFloat32();
+        return *this;
+    }
+
+    Float<M, E>& operator /= (const Float<M, E>& flt) {
+        value /= flt.ToFloat32();
+        return *this;
+    }
+
+    Float<M, E>& operator += (const Float<M, E>& flt) {
+        value += flt.ToFloat32();
+        return *this;
+    }
+
+    Float<M, E>& operator -= (const Float<M, E>& flt) {
+        value -= flt.ToFloat32();
+        return *this;
+    }
+
+    Float<M, E> operator - () const {
+        return Float<M, E>::FromFloat32(-ToFloat32());
+    }
+
+    bool operator < (const Float<M, E>& flt) const {
+        return ToFloat32() < flt.ToFloat32();
+    }
+
+    bool operator > (const Float<M, E>& flt) const {
+        return ToFloat32() > flt.ToFloat32();
+    }
+
+    bool operator >= (const Float<M, E>& flt) const {
+        return ToFloat32() >= flt.ToFloat32();
+    }
+
+    bool operator <= (const Float<M, E>& flt) const {
+        return ToFloat32() <= flt.ToFloat32();
+    }
+
+    bool operator == (const Float<M, E>& flt) const {
+        return ToFloat32() == flt.ToFloat32();
+    }
+
+    bool operator != (const Float<M, E>& flt) const {
+        return ToFloat32() != flt.ToFloat32();
+    }
+
+private:
+    static const unsigned MASK = (1 << (M + E + 1)) - 1;
+    static const unsigned MANTISSA_MASK = (1 << M) - 1;
+    static const unsigned EXPONENT_MASK = (1 << E) - 1;
+
+    // Stored as a regular float, merely for convenience
+    // TODO: Perform proper arithmetic on this!
+    float value;
+};
+
+using float24 = Float<16, 7>;
+using float20 = Float<12, 7>;
+using float16 = Float<10, 5>;
+
+} // namespace Pica

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -75,6 +75,12 @@ void RasterizerOpenGL::InitObjects() {
     glEnableVertexAttribArray(GLShader::ATTRIBUTE_TEXCOORD1);
     glEnableVertexAttribArray(GLShader::ATTRIBUTE_TEXCOORD2);
 
+    glVertexAttribPointer(GLShader::ATTRIBUTE_NORMQUAT, 4, GL_FLOAT, GL_FALSE, sizeof(HardwareVertex), (GLvoid*)offsetof(HardwareVertex, normquat));
+    glEnableVertexAttribArray(GLShader::ATTRIBUTE_NORMQUAT);
+
+    glVertexAttribPointer(GLShader::ATTRIBUTE_VIEW, 3, GL_FLOAT, GL_FALSE, sizeof(HardwareVertex), (GLvoid*)offsetof(HardwareVertex, view));
+    glEnableVertexAttribArray(GLShader::ATTRIBUTE_VIEW);
+
     SetShader();
 
     // Create textures for OGL framebuffer that will be rendered to, initially 1x1 to succeed in framebuffer creation
@@ -120,6 +126,19 @@ void RasterizerOpenGL::InitObjects() {
     glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, fb_color_texture.texture.handle, 0);
     glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, fb_depth_texture.texture.handle, 0);
 
+    for (size_t i = 0; i < lighting_lut.size(); ++i) {
+        lighting_lut[i].Create();
+        state.lighting_lut[i].texture_1d = lighting_lut[i].handle;
+
+        glActiveTexture(GL_TEXTURE3 + i);
+        glBindTexture(GL_TEXTURE_1D, state.lighting_lut[i].texture_1d);
+
+        glTexImage1D(GL_TEXTURE_1D, 0, GL_RGBA32F, 256, 0, GL_RGBA, GL_FLOAT, nullptr);
+        glTexParameteri(GL_TEXTURE_1D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
+        glTexParameteri(GL_TEXTURE_1D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
+    }
+    state.Apply();
+
     ASSERT_MSG(glCheckFramebufferStatus(GL_FRAMEBUFFER) == GL_FRAMEBUFFER_COMPLETE,
                "OpenGL rasterizer framebuffer setup failed, status %X", glCheckFramebufferStatus(GL_FRAMEBUFFER));
 }
@@ -139,12 +158,34 @@ void RasterizerOpenGL::Reset() {
     res_cache.InvalidateAll();
 }
 
+/**
+ * This is a helper function to resolve an issue with opposite quaternions being interpolated by
+ * OpenGL. See below for a detailed description of this issue (yuriks):
+ *
+ * For any rotation, there are two quaternions Q, and -Q, that represent the same rotation. If you
+ * interpolate two quaternions that are opposite, instead of going from one rotation to another
+ * using the shortest path, you'll go around the longest path. You can test if two quaternions are
+ * opposite by checking if Dot(Q1, W2) < 0. In that case, you can flip either of them, therefore
+ * making Dot(-Q1, W2) positive.
+ *
+ * NOTE: This solution corrects this issue per-vertex before passing the quaternions to OpenGL. This
+ * should be correct for nearly all cases, however a more correct implementation (but less trivial
+ * and perhaps unnecessary) would be to handle this per-fragment, by interpolating the quaternions
+ * manually using two Lerps, and doing this correction before each Lerp.
+ */
+static bool AreQuaternionsOpposite(Math::Vec4<Pica::float24> qa, Math::Vec4<Pica::float24> qb) {
+    Math::Vec4f a{ qa.x.ToFloat32(), qa.y.ToFloat32(), qa.z.ToFloat32(), qa.w.ToFloat32() };
+    Math::Vec4f b{ qb.x.ToFloat32(), qb.y.ToFloat32(), qb.z.ToFloat32(), qb.w.ToFloat32() };
+
+    return (Math::Dot(a, b) < 0.f);
+}
+
 void RasterizerOpenGL::AddTriangle(const Pica::Shader::OutputVertex& v0,
                                    const Pica::Shader::OutputVertex& v1,
                                    const Pica::Shader::OutputVertex& v2) {
-    vertex_batch.emplace_back(v0);
+    vertex_batch.emplace_back(v0, false);
-    vertex_batch.emplace_back(v1);
+    vertex_batch.emplace_back(v1, AreQuaternionsOpposite(v0.quat, v1.quat));
-    vertex_batch.emplace_back(v2);
+    vertex_batch.emplace_back(v2, AreQuaternionsOpposite(v0.quat, v2.quat));
 }
 
 void RasterizerOpenGL::DrawTriangles() {
@@ -156,6 +197,13 @@ void RasterizerOpenGL::DrawTriangles() {
         state.draw.shader_dirty = false;
     }
 
+    for (unsigned index = 0; index < lighting_lut.size(); index++) {
+        if (uniform_block_data.lut_dirty[index]) {
+            SyncLightingLUT(index);
+            uniform_block_data.lut_dirty[index] = false;
+        }
+    }
+
     if (uniform_block_data.dirty) {
         glBufferData(GL_UNIFORM_BUFFER, sizeof(UniformData), &uniform_block_data.data, GL_STATIC_DRAW);
         uniform_block_data.dirty = false;
@@ -283,6 +331,165 @@ void RasterizerOpenGL::NotifyPicaRegisterChanged(u32 id) {
     case PICA_REG_INDEX(tev_combiner_buffer_color):
         SyncCombinerColor();
         break;
+
+    // Fragment lighting specular 0 color
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[0].specular_0, 0x140 + 0 * 0x10):
+        SyncLightSpecular0(0);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[1].specular_0, 0x140 + 1 * 0x10):
+        SyncLightSpecular0(1);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[2].specular_0, 0x140 + 2 * 0x10):
+        SyncLightSpecular0(2);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[3].specular_0, 0x140 + 3 * 0x10):
+        SyncLightSpecular0(3);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[4].specular_0, 0x140 + 4 * 0x10):
+        SyncLightSpecular0(4);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[5].specular_0, 0x140 + 5 * 0x10):
+        SyncLightSpecular0(5);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[6].specular_0, 0x140 + 6 * 0x10):
+        SyncLightSpecular0(6);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[7].specular_0, 0x140 + 7 * 0x10):
+        SyncLightSpecular0(7);
+        break;
+
+    // Fragment lighting specular 1 color
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[0].specular_1, 0x141 + 0 * 0x10):
+        SyncLightSpecular1(0);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[1].specular_1, 0x141 + 1 * 0x10):
+        SyncLightSpecular1(1);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[2].specular_1, 0x141 + 2 * 0x10):
+        SyncLightSpecular1(2);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[3].specular_1, 0x141 + 3 * 0x10):
+        SyncLightSpecular1(3);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[4].specular_1, 0x141 + 4 * 0x10):
+        SyncLightSpecular1(4);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[5].specular_1, 0x141 + 5 * 0x10):
+        SyncLightSpecular1(5);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[6].specular_1, 0x141 + 6 * 0x10):
+        SyncLightSpecular1(6);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[7].specular_1, 0x141 + 7 * 0x10):
+        SyncLightSpecular1(7);
+        break;
+
+    // Fragment lighting diffuse color
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[0].diffuse, 0x142 + 0 * 0x10):
+        SyncLightDiffuse(0);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[1].diffuse, 0x142 + 1 * 0x10):
+        SyncLightDiffuse(1);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[2].diffuse, 0x142 + 2 * 0x10):
+        SyncLightDiffuse(2);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[3].diffuse, 0x142 + 3 * 0x10):
+        SyncLightDiffuse(3);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[4].diffuse, 0x142 + 4 * 0x10):
+        SyncLightDiffuse(4);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[5].diffuse, 0x142 + 5 * 0x10):
+        SyncLightDiffuse(5);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[6].diffuse, 0x142 + 6 * 0x10):
+        SyncLightDiffuse(6);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[7].diffuse, 0x142 + 7 * 0x10):
+        SyncLightDiffuse(7);
+        break;
+
+    // Fragment lighting ambient color
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[0].ambient, 0x143 + 0 * 0x10):
+        SyncLightAmbient(0);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[1].ambient, 0x143 + 1 * 0x10):
+        SyncLightAmbient(1);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[2].ambient, 0x143 + 2 * 0x10):
+        SyncLightAmbient(2);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[3].ambient, 0x143 + 3 * 0x10):
+        SyncLightAmbient(3);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[4].ambient, 0x143 + 4 * 0x10):
+        SyncLightAmbient(4);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[5].ambient, 0x143 + 5 * 0x10):
+        SyncLightAmbient(5);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[6].ambient, 0x143 + 6 * 0x10):
+        SyncLightAmbient(6);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[7].ambient, 0x143 + 7 * 0x10):
+        SyncLightAmbient(7);
+        break;
+
+     // Fragment lighting position
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[0].x, 0x144 + 0 * 0x10):
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[0].z, 0x145 + 0 * 0x10):
+        SyncLightPosition(0);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[1].x, 0x144 + 1 * 0x10):
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[1].z, 0x145 + 1 * 0x10):
+        SyncLightPosition(1);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[2].x, 0x144 + 2 * 0x10):
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[2].z, 0x145 + 2 * 0x10):
+        SyncLightPosition(2);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[3].x, 0x144 + 3 * 0x10):
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[3].z, 0x145 + 3 * 0x10):
+        SyncLightPosition(3);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[4].x, 0x144 + 4 * 0x10):
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[4].z, 0x145 + 4 * 0x10):
+        SyncLightPosition(4);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[5].x, 0x144 + 5 * 0x10):
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[5].z, 0x145 + 5 * 0x10):
+        SyncLightPosition(5);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[6].x, 0x144 + 6 * 0x10):
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[6].z, 0x145 + 6 * 0x10):
+        SyncLightPosition(6);
+        break;
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[7].x, 0x144 + 7 * 0x10):
+    case PICA_REG_INDEX_WORKAROUND(lighting.light[7].z, 0x145 + 7 * 0x10):
+        SyncLightPosition(7);
+        break;
+
+    // Fragment lighting global ambient color (emission + ambient * ambient)
+    case PICA_REG_INDEX_WORKAROUND(lighting.global_ambient, 0x1c0):
+        SyncGlobalAmbient();
+        break;
+
+    // Fragment lighting lookup tables
+    case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[0], 0x1c8):
+    case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[1], 0x1c9):
+    case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[2], 0x1ca):
+    case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[3], 0x1cb):
+    case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[4], 0x1cc):
+    case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[5], 0x1cd):
+    case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[6], 0x1ce):
+    case PICA_REG_INDEX_WORKAROUND(lighting.lut_data[7], 0x1cf):
+    {
+        auto& lut_config = regs.lighting.lut_config;
+        uniform_block_data.lut_dirty[lut_config.type / 4] = true;
+        break;
+    }
+
     }
 }
 
@@ -491,18 +698,39 @@ void RasterizerOpenGL::SetShader() {
         uniform_tex = glGetUniformLocation(shader->shader.handle, "tex[2]");
         if (uniform_tex != -1) { glUniform1i(uniform_tex, 2); }
 
+        // Set the texture samplers to correspond to different lookup table texture units
+        GLuint uniform_lut = glGetUniformLocation(shader->shader.handle, "lut[0]");
+        if (uniform_lut != -1) { glUniform1i(uniform_lut, 3); }
+        uniform_lut = glGetUniformLocation(shader->shader.handle, "lut[1]");
+        if (uniform_lut != -1) { glUniform1i(uniform_lut, 4); }
+        uniform_lut = glGetUniformLocation(shader->shader.handle, "lut[2]");
+        if (uniform_lut != -1) { glUniform1i(uniform_lut, 5); }
+        uniform_lut = glGetUniformLocation(shader->shader.handle, "lut[3]");
+        if (uniform_lut != -1) { glUniform1i(uniform_lut, 6); }
+        uniform_lut = glGetUniformLocation(shader->shader.handle, "lut[4]");
+        if (uniform_lut != -1) { glUniform1i(uniform_lut, 7); }
+        uniform_lut = glGetUniformLocation(shader->shader.handle, "lut[5]");
+        if (uniform_lut != -1) { glUniform1i(uniform_lut, 8); }
+
         current_shader = shader_cache.emplace(config, std::move(shader)).first->second.get();
 
         unsigned int block_index = glGetUniformBlockIndex(current_shader->shader.handle, "shader_data");
         glUniformBlockBinding(current_shader->shader.handle, block_index, 0);
-    }
 
-    // Update uniforms
+        // Update uniforms
-    SyncAlphaTest();
+        SyncAlphaTest();
-    SyncCombinerColor();
+        SyncCombinerColor();
-    auto& tev_stages = Pica::g_state.regs.GetTevStages();
+        auto& tev_stages = Pica::g_state.regs.GetTevStages();
-    for (int index = 0; index < tev_stages.size(); ++index)
+        for (int index = 0; index < tev_stages.size(); ++index)
-        SyncTevConstColor(index, tev_stages[index]);
+            SyncTevConstColor(index, tev_stages[index]);
+
+        SyncGlobalAmbient();
+        for (int light_index = 0; light_index < 8; light_index++) {
+            SyncLightDiffuse(light_index);
+            SyncLightAmbient(light_index);
+            SyncLightPosition(light_index);
+        }
+    }
 }
 
 void RasterizerOpenGL::SyncFramebuffer() {
@@ -604,8 +832,8 @@ void RasterizerOpenGL::SyncCullMode() {
 }
 
 void RasterizerOpenGL::SyncDepthModifiers() {
-    float depth_scale = -Pica::float24::FromRawFloat24(Pica::g_state.regs.viewport_depth_range).ToFloat32();
+    float depth_scale = -Pica::float24::FromRaw(Pica::g_state.regs.viewport_depth_range).ToFloat32();
-    float depth_offset = Pica::float24::FromRawFloat24(Pica::g_state.regs.viewport_depth_far_plane).ToFloat32() / 2.0f;
+    float depth_offset = Pica::float24::FromRaw(Pica::g_state.regs.viewport_depth_far_plane).ToFloat32() / 2.0f;
 
     // TODO: Implement scale modifier
     uniform_block_data.data.depth_offset = depth_offset;
@@ -683,12 +911,81 @@ void RasterizerOpenGL::SyncTevConstColor(int stage_index, const Pica::Regs::TevS
     }
 }
 
+void RasterizerOpenGL::SyncGlobalAmbient() {
+    auto color = PicaToGL::LightColor(Pica::g_state.regs.lighting.global_ambient);
+    if (color != uniform_block_data.data.lighting_global_ambient) {
+        uniform_block_data.data.lighting_global_ambient = color;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerOpenGL::SyncLightingLUT(unsigned lut_index) {
+    std::array<GLvec4, 256> new_data;
+
+    for (unsigned offset = 0; offset < new_data.size(); ++offset) {
+        new_data[offset][0] = Pica::g_state.lighting.luts[(lut_index * 4) + 0][offset].ToFloat();
+        new_data[offset][1] = Pica::g_state.lighting.luts[(lut_index * 4) + 1][offset].ToFloat();
+        new_data[offset][2] = Pica::g_state.lighting.luts[(lut_index * 4) + 2][offset].ToFloat();
+        new_data[offset][3] = Pica::g_state.lighting.luts[(lut_index * 4) + 3][offset].ToFloat();
+    }
+
+    if (new_data != lighting_lut_data[lut_index]) {
+        lighting_lut_data[lut_index] = new_data;
+        glActiveTexture(GL_TEXTURE3 + lut_index);
+        glTexSubImage1D(GL_TEXTURE_1D, 0, 0, 256, GL_RGBA, GL_FLOAT, lighting_lut_data[lut_index].data());
+    }
+}
+
+void RasterizerOpenGL::SyncLightSpecular0(int light_index) {
+    auto color = PicaToGL::LightColor(Pica::g_state.regs.lighting.light[light_index].specular_0);
+    if (color != uniform_block_data.data.light_src[light_index].specular_0) {
+        uniform_block_data.data.light_src[light_index].specular_0 = color;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerOpenGL::SyncLightSpecular1(int light_index) {
+    auto color = PicaToGL::LightColor(Pica::g_state.regs.lighting.light[light_index].specular_1);
+    if (color != uniform_block_data.data.light_src[light_index].specular_1) {
+        uniform_block_data.data.light_src[light_index].specular_1 = color;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerOpenGL::SyncLightDiffuse(int light_index) {
+    auto color = PicaToGL::LightColor(Pica::g_state.regs.lighting.light[light_index].diffuse);
+    if (color != uniform_block_data.data.light_src[light_index].diffuse) {
+        uniform_block_data.data.light_src[light_index].diffuse = color;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerOpenGL::SyncLightAmbient(int light_index) {
+    auto color = PicaToGL::LightColor(Pica::g_state.regs.lighting.light[light_index].ambient);
+    if (color != uniform_block_data.data.light_src[light_index].ambient) {
+        uniform_block_data.data.light_src[light_index].ambient = color;
+        uniform_block_data.dirty = true;
+    }
+}
+
+void RasterizerOpenGL::SyncLightPosition(int light_index) {
+    GLvec3 position = {
+        Pica::float16::FromRaw(Pica::g_state.regs.lighting.light[light_index].x).ToFloat32(),
+        Pica::float16::FromRaw(Pica::g_state.regs.lighting.light[light_index].y).ToFloat32(),
+        Pica::float16::FromRaw(Pica::g_state.regs.lighting.light[light_index].z).ToFloat32() };
+
+    if (position != uniform_block_data.data.light_src[light_index].position) {
+        uniform_block_data.data.light_src[light_index].position = position;
+        uniform_block_data.dirty = true;
+    }
+}
+
 void RasterizerOpenGL::SyncDrawState() {
     const auto& regs = Pica::g_state.regs;
 
     // Sync the viewport
-    GLsizei viewport_width = (GLsizei)Pica::float24::FromRawFloat24(regs.viewport_size_x).ToFloat32() * 2;
+    GLsizei viewport_width = (GLsizei)Pica::float24::FromRaw(regs.viewport_size_x).ToFloat32() * 2;
-    GLsizei viewport_height = (GLsizei)Pica::float24::FromRawFloat24(regs.viewport_size_y).ToFloat32() * 2;
+    GLsizei viewport_height = (GLsizei)Pica::float24::FromRaw(regs.viewport_size_y).ToFloat32() * 2;
 
     // OpenGL uses different y coordinates, so negate corner offset and flip origin
     // TODO: Ensure viewport_corner.x should not be negated or origin flipped

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -17,6 +17,7 @@
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_rasterizer_cache.h"
 #include "video_core/renderer_opengl/gl_state.h"
+#include "video_core/renderer_opengl/pica_to_gl.h"
 #include "video_core/shader/shader_interpreter.h"
 
 /**
@@ -71,6 +72,59 @@ struct PicaShaderConfig {
             regs.tev_combiner_buffer_input.update_mask_rgb.Value() |
             regs.tev_combiner_buffer_input.update_mask_a.Value() << 4;
 
+        // Fragment lighting
+
+        res.lighting.enable = !regs.lighting.disable;
+        res.lighting.src_num = regs.lighting.num_lights + 1;
+
+        for (unsigned light_index = 0; light_index < res.lighting.src_num; ++light_index) {
+            unsigned num = regs.lighting.light_enable.GetNum(light_index);
+            const auto& light = regs.lighting.light[num];
+            res.lighting.light[light_index].num = num;
+            res.lighting.light[light_index].directional = light.directional != 0;
+            res.lighting.light[light_index].two_sided_diffuse = light.two_sided_diffuse != 0;
+            res.lighting.light[light_index].dist_atten_enable = !regs.lighting.IsDistAttenDisabled(num);
+            res.lighting.light[light_index].dist_atten_bias = Pica::float20::FromRaw(light.dist_atten_bias).ToFloat32();
+            res.lighting.light[light_index].dist_atten_scale = Pica::float20::FromRaw(light.dist_atten_scale).ToFloat32();
+        }
+
+        res.lighting.lut_d0.enable = regs.lighting.disable_lut_d0 == 0;
+        res.lighting.lut_d0.abs_input = regs.lighting.abs_lut_input.disable_d0 == 0;
+        res.lighting.lut_d0.type = regs.lighting.lut_input.d0.Value();
+        res.lighting.lut_d0.scale = regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.d0);
+
+        res.lighting.lut_d1.enable = regs.lighting.disable_lut_d1 == 0;
+        res.lighting.lut_d1.abs_input = regs.lighting.abs_lut_input.disable_d1 == 0;
+        res.lighting.lut_d1.type = regs.lighting.lut_input.d1.Value();
+        res.lighting.lut_d1.scale = regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.d1);
+
+        res.lighting.lut_fr.enable = regs.lighting.disable_lut_fr == 0;
+        res.lighting.lut_fr.abs_input = regs.lighting.abs_lut_input.disable_fr == 0;
+        res.lighting.lut_fr.type = regs.lighting.lut_input.fr.Value();
+        res.lighting.lut_fr.scale = regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.fr);
+
+        res.lighting.lut_rr.enable = regs.lighting.disable_lut_rr == 0;
+        res.lighting.lut_rr.abs_input = regs.lighting.abs_lut_input.disable_rr == 0;
+        res.lighting.lut_rr.type = regs.lighting.lut_input.rr.Value();
+        res.lighting.lut_rr.scale = regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.rr);
+
+        res.lighting.lut_rg.enable = regs.lighting.disable_lut_rg == 0;
+        res.lighting.lut_rg.abs_input = regs.lighting.abs_lut_input.disable_rg == 0;
+        res.lighting.lut_rg.type = regs.lighting.lut_input.rg.Value();
+        res.lighting.lut_rg.scale = regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.rg);
+
+        res.lighting.lut_rb.enable = regs.lighting.disable_lut_rb == 0;
+        res.lighting.lut_rb.abs_input = regs.lighting.abs_lut_input.disable_rb == 0;
+        res.lighting.lut_rb.type = regs.lighting.lut_input.rb.Value();
+        res.lighting.lut_rb.scale = regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.rb);
+
+        res.lighting.config = regs.lighting.config;
+        res.lighting.fresnel_selector = regs.lighting.fresnel_selector;
+        res.lighting.bump_mode = regs.lighting.bump_mode;
+        res.lighting.bump_selector = regs.lighting.bump_selector;
+        res.lighting.bump_renorm = regs.lighting.disable_bump_renorm == 0;
+        res.lighting.clamp_highlights = regs.lighting.clamp_highlights != 0;
+
         return res;
     }
 
@@ -86,9 +140,37 @@ struct PicaShaderConfig {
         return std::memcmp(this, &o, sizeof(PicaShaderConfig)) == 0;
     };
 
-    Pica::Regs::CompareFunc alpha_test_func;
+    Pica::Regs::CompareFunc alpha_test_func = Pica::Regs::CompareFunc::Never;
     std::array<Pica::Regs::TevStageConfig, 6> tev_stages = {};
-    u8 combiner_buffer_input;
+    u8 combiner_buffer_input = 0;
+
+    struct {
+        struct {
+            unsigned num = 0;
+            bool directional = false;
+            bool two_sided_diffuse = false;
+            bool dist_atten_enable = false;
+            GLfloat dist_atten_scale = 0.0f;
+            GLfloat dist_atten_bias = 0.0f;
+        } light[8];
+
+        bool enable = false;
+        unsigned src_num = 0;
+        Pica::Regs::LightingBumpMode bump_mode = Pica::Regs::LightingBumpMode::None;
+        unsigned bump_selector = 0;
+        bool bump_renorm = false;
+        bool clamp_highlights = false;
+
+        Pica::Regs::LightingConfig config = Pica::Regs::LightingConfig::Config0;
+        Pica::Regs::LightingFresnelSelector fresnel_selector = Pica::Regs::LightingFresnelSelector::None;
+
+        struct {
+            bool enable = false;
+            bool abs_input = false;
+            Pica::Regs::LightingLutInput type = Pica::Regs::LightingLutInput::NH;
+            float scale = 1.0f;
+        } lut_d0, lut_d1, lut_fr, lut_rr, lut_rg, lut_rb;
+    } lighting;
 };
 
 namespace std {
@@ -167,7 +249,7 @@ private:
 
     /// Structure that the hardware rendered vertices are composed of
     struct HardwareVertex {
-        HardwareVertex(const Pica::Shader::OutputVertex& v) {
+        HardwareVertex(const Pica::Shader::OutputVertex& v, bool flip_quaternion) {
             position[0] = v.pos.x.ToFloat32();
             position[1] = v.pos.y.ToFloat32();
             position[2] = v.pos.z.ToFloat32();
@@ -182,6 +264,19 @@ private:
             tex_coord1[1] = v.tc1.y.ToFloat32();
             tex_coord2[0] = v.tc2.x.ToFloat32();
             tex_coord2[1] = v.tc2.y.ToFloat32();
+            normquat[0] = v.quat.x.ToFloat32();
+            normquat[1] = v.quat.y.ToFloat32();
+            normquat[2] = v.quat.z.ToFloat32();
+            normquat[3] = v.quat.w.ToFloat32();
+            view[0] = v.view.x.ToFloat32();
+            view[1] = v.view.y.ToFloat32();
+            view[2] = v.view.z.ToFloat32();
+
+            if (flip_quaternion) {
+                for (float& x : normquat) {
+                    x = -x;
+                }
+            }
         }
 
         GLfloat position[4];
@@ -189,20 +284,31 @@ private:
         GLfloat tex_coord0[2];
         GLfloat tex_coord1[2];
         GLfloat tex_coord2[2];
+        GLfloat normquat[4];
+        GLfloat view[3];
+    };
+
+    struct LightSrc {
+        alignas(16) GLvec3 specular_0;
+        alignas(16) GLvec3 specular_1;
+        alignas(16) GLvec3 diffuse;
+        alignas(16) GLvec3 ambient;
+        alignas(16) GLvec3 position;
     };
 
     /// Uniform structure for the Uniform Buffer Object, all members must be 16-byte aligned
     struct UniformData {
         // A vec4 color for each of the six tev stages
-        std::array<GLfloat, 4> const_color[6];
+        GLvec4 const_color[6];
-        std::array<GLfloat, 4> tev_combiner_buffer_color;
+        GLvec4 tev_combiner_buffer_color;
         GLint alphatest_ref;
         GLfloat depth_offset;
-        INSERT_PADDING_BYTES(8);
+        alignas(16) GLvec3 lighting_global_ambient;
+        LightSrc light_src[8];
     };
 
-    static_assert(sizeof(UniformData) == 0x80, "The size of the UniformData structure has changed, update the structure in the shader");
+    static_assert(sizeof(UniformData) == 0x310, "The size of the UniformData structure has changed, update the structure in the shader");
-    static_assert(sizeof(UniformData) < 16000, "UniformData structure must be less than 16kb as per the OpenGL spec");
+    static_assert(sizeof(UniformData) < 16384, "UniformData structure must be less than 16kb as per the OpenGL spec");
 
     /// Reconfigure the OpenGL color texture to use the given format and dimensions
     void ReconfigureColorTexture(TextureInfo& texture, Pica::Regs::ColorFormat format, u32 width, u32 height);
@@ -249,6 +355,27 @@ private:
     /// Syncs the TEV combiner color buffer to match the PICA register
     void SyncCombinerColor();
 
+    /// Syncs the lighting global ambient color to match the PICA register
+    void SyncGlobalAmbient();
+
+    /// Syncs the lighting lookup tables
+    void SyncLightingLUT(unsigned index);
+
+    /// Syncs the specified light's diffuse color to match the PICA register
+    void SyncLightDiffuse(int light_index);
+
+    /// Syncs the specified light's ambient color to match the PICA register
+    void SyncLightAmbient(int light_index);
+
+    /// Syncs the specified light's position to match the PICA register
+    void SyncLightPosition(int light_index);
+
+    /// Syncs the specified light's specular 0 color to match the PICA register
+    void SyncLightSpecular0(int light_index);
+
+    /// Syncs the specified light's specular 1 color to match the PICA register
+    void SyncLightSpecular1(int light_index);
+
     /// Syncs the remaining OpenGL drawing state to match the current PICA state
     void SyncDrawState();
 
@@ -291,6 +418,7 @@ private:
 
     struct {
         UniformData data;
+        bool lut_dirty[6];
         bool dirty;
     } uniform_block_data;
 
@@ -298,4 +426,7 @@ private:
     OGLBuffer vertex_buffer;
     OGLBuffer uniform_buffer;
     OGLFramebuffer framebuffer;
+
+    std::array<OGLTexture, 6> lighting_lut;
+    std::array<std::array<GLvec4, 256>, 6> lighting_lut_data;
 };

src/video_core/renderer_opengl/gl_shader_gen.cpp

@@ -32,12 +32,10 @@ static void AppendSource(std::string& out, TevStageConfig::Source source,
         out += "primary_color";
         break;
     case Source::PrimaryFragmentColor:
-        // HACK: Until we implement fragment lighting, use primary_color
+        out += "primary_fragment_color";
-        out += "primary_color";
         break;
     case Source::SecondaryFragmentColor:
-        // HACK: Until we implement fragment lighting, use zero
+        out += "secondary_fragment_color";
-        out += "vec4(0.0)";
         break;
     case Source::Texture0:
         out += "texture(tex[0], texcoord[0])";
@@ -320,26 +318,229 @@ static void WriteTevStage(std::string& out, const PicaShaderConfig& config, unsi
         out += "next_combiner_buffer.a = last_tex_env_out.a;\n";
 }
 
+/// Writes the code to emulate fragment lighting
+static void WriteLighting(std::string& out, const PicaShaderConfig& config) {
+    // Define lighting globals
+    out += "vec4 diffuse_sum = vec4(0.0, 0.0, 0.0, 1.0);\n"
+           "vec4 specular_sum = vec4(0.0, 0.0, 0.0, 1.0);\n"
+           "vec3 light_vector = vec3(0.0);\n"
+           "vec3 refl_value = vec3(0.0);\n";
+
+    // Compute fragment normals
+    if (config.lighting.bump_mode == Pica::Regs::LightingBumpMode::NormalMap) {
+        // Bump mapping is enabled using a normal map, read perturbation vector from the selected texture
+        std::string bump_selector = std::to_string(config.lighting.bump_selector);
+        out += "vec3 surface_normal = 2.0 * texture(tex[" + bump_selector + "], texcoord[" + bump_selector + "]).rgb - 1.0;\n";
+
+        // Recompute Z-component of perturbation if 'renorm' is enabled, this provides a higher precision result
+        if (config.lighting.bump_renorm) {
+            std::string val = "(1.0 - (surface_normal.x*surface_normal.x + surface_normal.y*surface_normal.y))";
+            out += "surface_normal.z = sqrt(max(" + val + ", 0.0));\n";
+        }
+    } else if (config.lighting.bump_mode == Pica::Regs::LightingBumpMode::TangentMap) {
+        // Bump mapping is enabled using a tangent map
+        LOG_CRITICAL(HW_GPU, "unimplemented bump mapping mode (tangent mapping)");
+        UNIMPLEMENTED();
+    } else {
+        // No bump mapping - surface local normal is just a unit normal
+        out += "vec3 surface_normal = vec3(0.0, 0.0, 1.0);\n";
+    }
+
+    // Rotate the surface-local normal by the interpolated normal quaternion to convert it to eyespace
+    out += "vec3 normal = normalize(quaternion_rotate(normquat, surface_normal));\n";
+
+    // Gets the index into the specified lookup table for specular lighting
+    auto GetLutIndex = [config](unsigned light_num, Regs::LightingLutInput input, bool abs) {
+        const std::string half_angle = "normalize(normalize(view) + light_vector)";
+        std::string index;
+        switch (input) {
+        case Regs::LightingLutInput::NH:
+            index = "dot(normal, " + half_angle + ")";
+            break;
+
+        case Regs::LightingLutInput::VH:
+            index = std::string("dot(normalize(view), " + half_angle + ")");
+            break;
+
+        case Regs::LightingLutInput::NV:
+            index = std::string("dot(normal, normalize(view))");
+            break;
+
+        case Regs::LightingLutInput::LN:
+            index = std::string("dot(light_vector, normal)");
+            break;
+
+        default:
+            LOG_CRITICAL(HW_GPU, "Unknown lighting LUT input %d\n", (int)input);
+            UNIMPLEMENTED();
+            break;
+        }
+
+        if (abs) {
+            // LUT index is in the range of (0.0, 1.0)
+            index = config.lighting.light[light_num].two_sided_diffuse ? "abs(" + index + ")" : "max(" + index + ", 0.f)";
+            return "(FLOAT_255 * clamp(" + index + ", 0.0, 1.0))";
+        } else {
+            // LUT index is in the range of (-1.0, 1.0)
+            index = "clamp(" + index + ", -1.0, 1.0)";
+            return "(FLOAT_255 * ((" + index + " < 0) ? " + index + " + 2.0 : " + index + ") / 2.0)";
+        }
+
+        return std::string();
+    };
+
+    // Gets the lighting lookup table value given the specified sampler and index
+    auto GetLutValue = [](Regs::LightingSampler sampler, std::string lut_index) {
+        return std::string("texture(lut[" + std::to_string((unsigned)sampler / 4) + "], " +
+                           lut_index + ")[" + std::to_string((unsigned)sampler & 3) + "]");
+    };
+
+    // Write the code to emulate each enabled light
+    for (unsigned light_index = 0; light_index < config.lighting.src_num; ++light_index) {
+        const auto& light_config = config.lighting.light[light_index];
+        std::string light_src = "light_src[" + std::to_string(light_config.num) + "]";
+
+        // Compute light vector (directional or positional)
+        if (light_config.directional)
+            out += "light_vector = normalize(" + light_src + ".position);\n";
+        else
+            out += "light_vector = normalize(" + light_src + ".position + view);\n";
+
+        // Compute dot product of light_vector and normal, adjust if lighting is one-sided or two-sided
+        std::string dot_product = light_config.two_sided_diffuse ? "abs(dot(light_vector, normal))" : "max(dot(light_vector, normal), 0.0)";
+
+        // If enabled, compute distance attenuation value
+        std::string dist_atten = "1.0";
+        if (light_config.dist_atten_enable) {
+            std::string scale = std::to_string(light_config.dist_atten_scale);
+            std::string bias = std::to_string(light_config.dist_atten_bias);
+            std::string index = "(" + scale + " * length(-view - " + light_src + ".position) + " + bias + ")";
+            index = "((clamp(" + index + ", 0.0, FLOAT_255)))";
+            const unsigned lut_num = ((unsigned)Regs::LightingSampler::DistanceAttenuation + light_config.num);
+            dist_atten = GetLutValue((Regs::LightingSampler)lut_num, index);
+        }
+
+        // If enabled, clamp specular component if lighting result is negative
+        std::string clamp_highlights = config.lighting.clamp_highlights ? "(dot(light_vector, normal) <= 0.0 ? 0.0 : 1.0)" : "1.0";
+
+        // Specular 0 component
+        std::string d0_lut_value = "1.0";
+        if (config.lighting.lut_d0.enable && Pica::Regs::IsLightingSamplerSupported(config.lighting.config, Pica::Regs::LightingSampler::Distribution0)) {
+            // Lookup specular "distribution 0" LUT value
+            std::string index = GetLutIndex(light_config.num, config.lighting.lut_d0.type, config.lighting.lut_d0.abs_input);
+            d0_lut_value = "(" + std::to_string(config.lighting.lut_d0.scale) + " * " + GetLutValue(Regs::LightingSampler::Distribution0, index) + ")";
+        }
+        std::string specular_0 = "(" + d0_lut_value + " * " + light_src + ".specular_0)";
+
+        // If enabled, lookup ReflectRed value, otherwise, 1.0 is used
+        if (config.lighting.lut_rr.enable && Pica::Regs::IsLightingSamplerSupported(config.lighting.config, Pica::Regs::LightingSampler::ReflectRed)) {
+            std::string index = GetLutIndex(light_config.num, config.lighting.lut_rr.type, config.lighting.lut_rr.abs_input);
+            std::string value = "(" + std::to_string(config.lighting.lut_rr.scale) + " * " + GetLutValue(Regs::LightingSampler::ReflectRed, index) + ")";
+            out += "refl_value.r = " + value + ";\n";
+        } else {
+            out += "refl_value.r = 1.0;\n";
+        }
+
+        // If enabled, lookup ReflectGreen value, otherwise, ReflectRed value is used
+        if (config.lighting.lut_rg.enable && Pica::Regs::IsLightingSamplerSupported(config.lighting.config, Pica::Regs::LightingSampler::ReflectGreen)) {
+            std::string index = GetLutIndex(light_config.num, config.lighting.lut_rg.type, config.lighting.lut_rg.abs_input);
+            std::string value = "(" + std::to_string(config.lighting.lut_rg.scale) + " * " + GetLutValue(Regs::LightingSampler::ReflectGreen, index) + ")";
+            out += "refl_value.g = " + value + ";\n";
+        } else {
+            out += "refl_value.g = refl_value.r;\n";
+        }
+
+        // If enabled, lookup ReflectBlue value, otherwise, ReflectRed value is used
+        if (config.lighting.lut_rb.enable && Pica::Regs::IsLightingSamplerSupported(config.lighting.config, Pica::Regs::LightingSampler::ReflectBlue)) {
+            std::string index = GetLutIndex(light_config.num, config.lighting.lut_rb.type, config.lighting.lut_rb.abs_input);
+            std::string value = "(" + std::to_string(config.lighting.lut_rb.scale) + " * " + GetLutValue(Regs::LightingSampler::ReflectBlue, index) + ")";
+            out += "refl_value.b = " + value + ";\n";
+        } else {
+            out += "refl_value.b = refl_value.r;\n";
+        }
+
+        // Specular 1 component
+        std::string d1_lut_value = "1.0";
+        if (config.lighting.lut_d1.enable && Pica::Regs::IsLightingSamplerSupported(config.lighting.config, Pica::Regs::LightingSampler::Distribution1)) {
+            // Lookup specular "distribution 1" LUT value
+            std::string index = GetLutIndex(light_config.num, config.lighting.lut_d1.type, config.lighting.lut_d1.abs_input);
+            d1_lut_value = "(" + std::to_string(config.lighting.lut_d1.scale) + " * " + GetLutValue(Regs::LightingSampler::Distribution1, index) + ")";
+        }
+        std::string specular_1 = "(" + d1_lut_value + " * refl_value * " + light_src + ".specular_1)";
+
+        // Fresnel
+        if (config.lighting.lut_fr.enable && Pica::Regs::IsLightingSamplerSupported(config.lighting.config, Pica::Regs::LightingSampler::Fresnel)) {
+            // Lookup fresnel LUT value
+            std::string index = GetLutIndex(light_config.num, config.lighting.lut_fr.type, config.lighting.lut_fr.abs_input);
+            std::string value = "(" + std::to_string(config.lighting.lut_fr.scale) + " * " + GetLutValue(Regs::LightingSampler::Fresnel, index) + ")";
+
+            // Enabled for difffuse lighting alpha component
+            if (config.lighting.fresnel_selector == Pica::Regs::LightingFresnelSelector::PrimaryAlpha ||
+                config.lighting.fresnel_selector == Pica::Regs::LightingFresnelSelector::Both)
+                out += "diffuse_sum.a  *= " + value + ";\n";
+
+            // Enabled for the specular lighting alpha component
+            if (config.lighting.fresnel_selector == Pica::Regs::LightingFresnelSelector::SecondaryAlpha ||
+                config.lighting.fresnel_selector == Pica::Regs::LightingFresnelSelector::Both)
+                out += "specular_sum.a *= " + value + ";\n";
+        }
+
+        // Compute primary fragment color (diffuse lighting) function
+        out += "diffuse_sum.rgb += ((" + light_src + ".diffuse * " + dot_product + ") + " + light_src + ".ambient) * " + dist_atten + ";\n";
+
+        // Compute secondary fragment color (specular lighting) function
+        out += "specular_sum.rgb += (" + specular_0 + " + " + specular_1 + ") * " + clamp_highlights + " * " + dist_atten + ";\n";
+    }
+
+    // Sum final lighting result
+    out += "diffuse_sum.rgb += lighting_global_ambient;\n";
+    out += "primary_fragment_color = clamp(diffuse_sum, vec4(0.0), vec4(1.0));\n";
+    out += "secondary_fragment_color = clamp(specular_sum, vec4(0.0), vec4(1.0));\n";
+}
+
 std::string GenerateFragmentShader(const PicaShaderConfig& config) {
     std::string out = R"(
 #version 330 core
 #define NUM_TEV_STAGES 6
+#define NUM_LIGHTS 8
+#define LIGHTING_LUT_SIZE 256
+#define FLOAT_255 (255.0 / 256.0)
 
 in vec4 primary_color;
 in vec2 texcoord[3];
+in vec4 normquat;
+in vec3 view;
 
 out vec4 color;
 
+struct LightSrc {
+    vec3 specular_0;
+    vec3 specular_1;
+    vec3 diffuse;
+    vec3 ambient;
+    vec3 position;
+};
+
 layout (std140) uniform shader_data {
     vec4 const_color[NUM_TEV_STAGES];
     vec4 tev_combiner_buffer_color;
     int alphatest_ref;
     float depth_offset;
+    vec3 lighting_global_ambient;
+    LightSrc light_src[NUM_LIGHTS];
 };
 
 uniform sampler2D tex[3];
+uniform sampler1D lut[6];
+
+// Rotate the vector v by the quaternion q
+vec3 quaternion_rotate(vec4 q, vec3 v) {
+    return v + 2.0 * cross(q.xyz, cross(q.xyz, v) + q.w * v);
+}
 
 void main() {
+vec4 primary_fragment_color = vec4(0.0);
+vec4 secondary_fragment_color = vec4(0.0);
 )";
 
     // Do not do any sort of processing if it's obvious we're not going to pass the alpha test
@@ -348,6 +549,9 @@ void main() {
         return out;
     }
 
+    if (config.lighting.enable)
+        WriteLighting(out, config);
+
     out += "vec4 combiner_buffer = vec4(0.0);\n";
     out += "vec4 next_combiner_buffer = tev_combiner_buffer_color;\n";
     out += "vec4 last_tex_env_out = vec4(0.0);\n";
@@ -369,21 +573,28 @@ void main() {
 
 std::string GenerateVertexShader() {
     std::string out = "#version 330 core\n";
+
     out += "layout(location = " + std::to_string((int)ATTRIBUTE_POSITION)  + ") in vec4 vert_position;\n";
     out += "layout(location = " + std::to_string((int)ATTRIBUTE_COLOR)     + ") in vec4 vert_color;\n";
     out += "layout(location = " + std::to_string((int)ATTRIBUTE_TEXCOORD0) + ") in vec2 vert_texcoord0;\n";
     out += "layout(location = " + std::to_string((int)ATTRIBUTE_TEXCOORD1) + ") in vec2 vert_texcoord1;\n";
     out += "layout(location = " + std::to_string((int)ATTRIBUTE_TEXCOORD2) + ") in vec2 vert_texcoord2;\n";
+    out += "layout(location = " + std::to_string((int)ATTRIBUTE_NORMQUAT)  + ") in vec4 vert_normquat;\n";
+    out += "layout(location = " + std::to_string((int)ATTRIBUTE_VIEW)      + ") in vec3 vert_view;\n";
 
     out += R"(
 out vec4 primary_color;
 out vec2 texcoord[3];
+out vec4 normquat;
+out vec3 view;
 
 void main() {
     primary_color = vert_color;
     texcoord[0] = vert_texcoord0;
     texcoord[1] = vert_texcoord1;
     texcoord[2] = vert_texcoord2;
+    normquat = vert_normquat;
+    view = vert_view;
     gl_Position = vec4(vert_position.x, vert_position.y, -vert_position.z, vert_position.w);
 }
 )";

src/video_core/renderer_opengl/gl_shader_util.h

@@ -14,6 +14,8 @@ enum Attributes {
     ATTRIBUTE_TEXCOORD0,
     ATTRIBUTE_TEXCOORD1,
     ATTRIBUTE_TEXCOORD2,
+    ATTRIBUTE_NORMQUAT,
+    ATTRIBUTE_VIEW,
 };
 
 /**

src/video_core/renderer_opengl/gl_state.cpp

@@ -170,6 +170,14 @@ void OpenGLState::Apply() {
         }
     }
 
+    // Lighting LUTs
+    for (unsigned i = 0; i < ARRAY_SIZE(lighting_lut); ++i) {
+        if (lighting_lut[i].texture_1d != cur_state.lighting_lut[i].texture_1d) {
+            glActiveTexture(GL_TEXTURE3 + i);
+            glBindTexture(GL_TEXTURE_1D, lighting_lut[i].texture_1d);
+        }
+    }
+
     // Framebuffer
     if (draw.framebuffer != cur_state.draw.framebuffer) {
         glBindFramebuffer(GL_FRAMEBUFFER, draw.framebuffer);

src/video_core/renderer_opengl/gl_state.h

@@ -61,6 +61,10 @@ public:
         GLuint sampler; // GL_SAMPLER_BINDING
     } texture_units[3];
 
+    struct {
+        GLuint texture_1d; // GL_TEXTURE_BINDING_1D
+    } lighting_lut[6];
+
     struct {
         GLuint framebuffer; // GL_DRAW_FRAMEBUFFER_BINDING
         GLuint vertex_array; // GL_VERTEX_ARRAY_BINDING

src/video_core/renderer_opengl/pica_to_gl.h

@@ -10,6 +10,9 @@
 
 #include "video_core/pica.h"
 
+using GLvec3 = std::array<GLfloat, 3>;
+using GLvec4 = std::array<GLfloat, 4>;
+
 namespace PicaToGL {
 
 inline GLenum TextureFilterMode(Pica::Regs::TextureConfig::TextureFilter mode) {
@@ -175,7 +178,7 @@ inline GLenum StencilOp(Pica::Regs::StencilAction action) {
     return stencil_op_table[(unsigned)action];
 }
 
-inline std::array<GLfloat, 4> ColorRGBA8(const u32 color) {
+inline GLvec4 ColorRGBA8(const u32 color) {
     return { { (color >>  0 & 0xFF) / 255.0f,
                (color >>  8 & 0xFF) / 255.0f,
                (color >> 16 & 0xFF) / 255.0f,
@@ -183,4 +186,11 @@ inline std::array<GLfloat, 4> ColorRGBA8(const u32 color) {
            } };
 }
 
+inline std::array<GLfloat, 3> LightColor(const Pica::Regs::LightColor& color) {
+    return { { color.r / 255.0f,
+               color.g / 255.0f,
+               color.b / 255.0f
+           } };
+}
+
 } // namespace

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -81,8 +81,8 @@ struct ScreenRectVertex {
  * The projection part of the matrix is trivial, hence these operations are represented
  * by a 3x2 matrix.
  */
-static std::array<GLfloat, 3*2> MakeOrthographicMatrix(const float width, const float height) {
+static std::array<GLfloat, 3 * 2> MakeOrthographicMatrix(const float width, const float height) {
-    std::array<GLfloat, 3*2> matrix;
+    std::array<GLfloat, 3 * 2> matrix;
 
     matrix[0] = 2.f / width; matrix[2] = 0.f;           matrix[4] = -1.f;
     matrix[1] = 0.f;         matrix[3] = -2.f / height; matrix[5] = 1.f;

src/video_core/shader/shader.cpp

@@ -134,11 +134,13 @@ OutputVertex Run(UnitState<false>& state, const InputVertex& input, int num_attr
             std::fmin(std::fabs(ret.color[i].ToFloat32()), 1.0f));
     }
 
-    LOG_TRACE(Render_Software, "Output vertex: pos (%.2f, %.2f, %.2f, %.2f), quat (%.2f, %.2f, %.2f, %.2f), col(%.2f, %.2f, %.2f, %.2f), tc0(%.2f, %.2f)",
+    LOG_TRACE(Render_Software, "Output vertex: pos(%.2f, %.2f, %.2f, %.2f), quat(%.2f, %.2f, %.2f, %.2f), "
+        "col(%.2f, %.2f, %.2f, %.2f), tc0(%.2f, %.2f), view(%.2f, %.2f, %.2f)",
         ret.pos.x.ToFloat32(), ret.pos.y.ToFloat32(), ret.pos.z.ToFloat32(), ret.pos.w.ToFloat32(),
         ret.quat.x.ToFloat32(), ret.quat.y.ToFloat32(), ret.quat.z.ToFloat32(), ret.quat.w.ToFloat32(),
         ret.color.x.ToFloat32(), ret.color.y.ToFloat32(), ret.color.z.ToFloat32(), ret.color.w.ToFloat32(),
-        ret.tc0.u().ToFloat32(), ret.tc0.v().ToFloat32());
+        ret.tc0.u().ToFloat32(), ret.tc0.v().ToFloat32(),
+        ret.view.x.ToFloat32(), ret.view.y.ToFloat32(), ret.view.z.ToFloat32());
 
     return ret;
 }

src/video_core/shader/shader.h

@@ -37,17 +37,19 @@ struct OutputVertex {
     Math::Vec4<float24> color;
     Math::Vec2<float24> tc0;
     Math::Vec2<float24> tc1;
-    float24 pad[6];
+    INSERT_PADDING_WORDS(2);
+    Math::Vec3<float24> view;
+    INSERT_PADDING_WORDS(1);
     Math::Vec2<float24> tc2;
 
     // Padding for optimal alignment
-    float24 pad2[4];
+    INSERT_PADDING_WORDS(4);
 
     // Attributes used to store intermediate results
 
     // position after perspective divide
     Math::Vec3<float24> screenpos;
-    float24 pad3;
+    INSERT_PADDING_WORDS(1);
 
     // Linear interpolation
     // factor: 0=this, 1=vtx