Merge pull request #2822 from wwylele/sw_lighting-2

Implement fragment lighting in the sw renderer (take 2)

8 files changed, 315 insertions, 9 deletions

diff --git a/src/common/quaternion.h b/src/common/quaternion.h
index 84ac82ed3..77f626bcb 100644
--- a/src/common/quaternion.h
+++ b/src/common/quaternion.h
@@ -30,6 +30,11 @@ public:
         return {xyz * other.w + other.xyz * w + Cross(xyz, other.xyz),
                 w * other.w - Dot(xyz, other.xyz)};
     }
+
+    Quaternion<T> Normalized() const {
+        T length = std::sqrt(xyz.Length2() + w * w);
+        return {xyz / length, w / length};
+    }
 };
 
 template <typename T>
diff --git a/src/common/vector_math.h b/src/common/vector_math.h
index c7a461a1e..6e2a5ad60 100644
--- a/src/common/vector_math.h
+++ b/src/common/vector_math.h
@@ -31,7 +31,6 @@
 #pragma once
 
 #include <cmath>
-#include <type_traits>
 
 namespace Math {
 
@@ -90,7 +89,7 @@ public:
         x -= other.x;
         y -= other.y;
     }
-    template <typename Q = T, class = typename std::enable_if<std::is_signed<Q>::value>::type>
+
     Vec2<decltype(-T{})> operator-() const {
         return MakeVec(-x, -y);
     }
@@ -247,7 +246,7 @@ public:
         y -= other.y;
         z -= other.z;
     }
-    template <typename Q = T, class = typename std::enable_if<std::is_signed<Q>::value>::type>
+
     Vec3<decltype(-T{})> operator-() const {
         return MakeVec(-x, -y, -z);
     }
@@ -462,7 +461,7 @@ public:
         z -= other.z;
         w -= other.w;
     }
-    template <typename Q = T, class = typename std::enable_if<std::is_signed<Q>::value>::type>
+
     Vec4<decltype(-T{})> operator-() const {
         return MakeVec(-x, -y, -z, -w);
     }
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 0961a3251..cffa4c952 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -15,6 +15,7 @@ set(SRCS
             shader/shader_interpreter.cpp
             swrasterizer/clipper.cpp
             swrasterizer/framebuffer.cpp
+            swrasterizer/lighting.cpp
             swrasterizer/proctex.cpp
             swrasterizer/rasterizer.cpp
             swrasterizer/swrasterizer.cpp
@@ -55,6 +56,7 @@ set(HEADERS
             shader/shader_interpreter.h
             swrasterizer/clipper.h
             swrasterizer/framebuffer.h
+            swrasterizer/lighting.h
             swrasterizer/proctex.h
             swrasterizer/rasterizer.h
             swrasterizer/swrasterizer.h
diff --git a/src/video_core/pica_state.h b/src/video_core/pica_state.h
index 2d23d34e6..864a2c9e6 100644
--- a/src/video_core/pica_state.h
+++ b/src/video_core/pica_state.h
@@ -79,7 +79,7 @@ struct State {
         std::array<ColorDifferenceEntry, 256> color_diff_table;
     } proctex;
 
-    struct {
+    struct Lighting {
         union LutEntry {
             // Used for raw access
             u32 raw;
diff --git a/src/video_core/swrasterizer/clipper.cpp b/src/video_core/swrasterizer/clipper.cpp
index 6fb923756..7537689b7 100644
--- a/src/video_core/swrasterizer/clipper.cpp
+++ b/src/video_core/swrasterizer/clipper.cpp
@@ -95,6 +95,17 @@ void ProcessTriangle(const OutputVertex& v0, const OutputVertex& v1, const Outpu
     static const size_t MAX_VERTICES = 9;
     static_vector<Vertex, MAX_VERTICES> buffer_a = {v0, v1, v2};
     static_vector<Vertex, MAX_VERTICES> buffer_b;
+
+    auto FlipQuaternionIfOpposite = [](auto& a, const auto& b) {
+        if (Math::Dot(a, b) < float24::Zero())
+            a = -a;
+    };
+
+    // Flip the quaternions if they are opposite to prevent interpolating them over the wrong
+    // direction.
+    FlipQuaternionIfOpposite(buffer_a[1].quat, buffer_a[0].quat);
+    FlipQuaternionIfOpposite(buffer_a[2].quat, buffer_a[0].quat);
+
     auto* output_list = &buffer_a;
     auto* input_list = &buffer_b;
 
diff --git a/src/video_core/swrasterizer/lighting.cpp b/src/video_core/swrasterizer/lighting.cpp
new file mode 100644
index 000000000..63088eee8
--- /dev/null
+++ b/src/video_core/swrasterizer/lighting.cpp
@@ -0,0 +1,250 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/math_util.h"
+#include "video_core/swrasterizer/lighting.h"
+
+namespace Pica {
+
+static float LookupLightingLut(const Pica::State::Lighting& lighting, size_t lut_index, u8 index,
+                               float delta) {
+    ASSERT_MSG(lut_index < lighting.luts.size(), "Out of range lut");
+    ASSERT_MSG(index < lighting.luts[lut_index].size(), "Out of range index");
+
+    const auto& lut = lighting.luts[lut_index][index];
+
+    float lut_value = lut.ToFloat();
+    float lut_diff = lut.DiffToFloat();
+
+    return lut_value + lut_diff * delta;
+}
+
+std::tuple<Math::Vec4<u8>, Math::Vec4<u8>> ComputeFragmentsColors(
+    const Pica::LightingRegs& lighting, const Pica::State::Lighting& lighting_state,
+    const Math::Quaternion<float>& normquat, const Math::Vec3<float>& view) {
+
+    // TODO(Subv): Bump mapping
+    Math::Vec3<float> surface_normal = {0.0f, 0.0f, 1.0f};
+
+    if (lighting.config0.bump_mode != LightingRegs::LightingBumpMode::None) {
+        LOG_CRITICAL(HW_GPU, "unimplemented bump mapping");
+        UNIMPLEMENTED();
+    }
+
+    // Use the normalized the quaternion when performing the rotation
+    auto normal = Math::QuaternionRotate(normquat, surface_normal);
+
+    Math::Vec4<float> diffuse_sum = {0.0f, 0.0f, 0.0f, 1.0f};
+    Math::Vec4<float> specular_sum = {0.0f, 0.0f, 0.0f, 1.0f};
+
+    for (unsigned light_index = 0; light_index <= lighting.max_light_index; ++light_index) {
+        unsigned num = lighting.light_enable.GetNum(light_index);
+        const auto& light_config = lighting.light[num];
+
+        Math::Vec3<float> refl_value = {};
+        Math::Vec3<float> position = {float16::FromRaw(light_config.x).ToFloat32(),
+                                      float16::FromRaw(light_config.y).ToFloat32(),
+                                      float16::FromRaw(light_config.z).ToFloat32()};
+        Math::Vec3<float> light_vector;
+
+        if (light_config.config.directional)
+            light_vector = position;
+        else
+            light_vector = position + view;
+
+        light_vector.Normalize();
+
+        float dist_atten = 1.0f;
+        if (!lighting.IsDistAttenDisabled(num)) {
+            auto distance = (-view - position).Length();
+            float scale = Pica::float20::FromRaw(light_config.dist_atten_scale).ToFloat32();
+            float bias = Pica::float20::FromRaw(light_config.dist_atten_bias).ToFloat32();
+            size_t lut =
+                static_cast<size_t>(LightingRegs::LightingSampler::DistanceAttenuation) + num;
+
+            float sample_loc = MathUtil::Clamp(scale * distance + bias, 0.0f, 1.0f);
+
+            u8 lutindex =
+                static_cast<u8>(MathUtil::Clamp(std::floor(sample_loc * 256.0f), 0.0f, 255.0f));
+            float delta = sample_loc * 256 - lutindex;
+            dist_atten = LookupLightingLut(lighting_state, lut, lutindex, delta);
+        }
+
+        auto GetLutValue = [&](LightingRegs::LightingLutInput input, bool abs,
+                               LightingRegs::LightingScale scale_enum,
+                               LightingRegs::LightingSampler sampler) {
+            Math::Vec3<float> norm_view = view.Normalized();
+            Math::Vec3<float> half_angle = (norm_view + light_vector).Normalized();
+            float result = 0.0f;
+
+            switch (input) {
+            case LightingRegs::LightingLutInput::NH:
+                result = Math::Dot(normal, half_angle);
+                break;
+
+            case LightingRegs::LightingLutInput::VH:
+                result = Math::Dot(norm_view, half_angle);
+                break;
+
+            case LightingRegs::LightingLutInput::NV:
+                result = Math::Dot(normal, norm_view);
+                break;
+
+            case LightingRegs::LightingLutInput::LN:
+                result = Math::Dot(light_vector, normal);
+                break;
+
+            default:
+                LOG_CRITICAL(HW_GPU, "Unknown lighting LUT input %u\n", static_cast<u32>(input));
+                UNIMPLEMENTED();
+                result = 0.0f;
+            }
+
+            u8 index;
+            float delta;
+
+            if (abs) {
+                if (light_config.config.two_sided_diffuse)
+                    result = std::abs(result);
+                else
+                    result = std::max(result, 0.0f);
+
+                float flr = std::floor(result * 256.0f);
+                index = static_cast<u8>(MathUtil::Clamp(flr, 0.0f, 255.0f));
+                delta = result * 256 - index;
+            } else {
+                float flr = std::floor(result * 128.0f);
+                s8 signed_index = static_cast<s8>(MathUtil::Clamp(flr, -128.0f, 127.0f));
+                delta = result * 128.0f - signed_index;
+                index = static_cast<u8>(signed_index);
+            }
+
+            float scale = lighting.lut_scale.GetScale(scale_enum);
+            return scale *
+                   LookupLightingLut(lighting_state, static_cast<size_t>(sampler), index, delta);
+        };
+
+        // Specular 0 component
+        float d0_lut_value = 1.0f;
+        if (lighting.config1.disable_lut_d0 == 0 &&
+            LightingRegs::IsLightingSamplerSupported(
+                lighting.config0.config, LightingRegs::LightingSampler::Distribution0)) {
+            d0_lut_value =
+                GetLutValue(lighting.lut_input.d0, lighting.abs_lut_input.disable_d0 == 0,
+                            lighting.lut_scale.d0, LightingRegs::LightingSampler::Distribution0);
+        }
+
+        Math::Vec3<float> specular_0 = d0_lut_value * light_config.specular_0.ToVec3f();
+
+        // If enabled, lookup ReflectRed value, otherwise, 1.0 is used
+        if (lighting.config1.disable_lut_rr == 0 &&
+            LightingRegs::IsLightingSamplerSupported(lighting.config0.config,
+                                                     LightingRegs::LightingSampler::ReflectRed)) {
+            refl_value.x =
+                GetLutValue(lighting.lut_input.rr, lighting.abs_lut_input.disable_rr == 0,
+                            lighting.lut_scale.rr, LightingRegs::LightingSampler::ReflectRed);
+        } else {
+            refl_value.x = 1.0f;
+        }
+
+        // If enabled, lookup ReflectGreen value, otherwise, ReflectRed value is used
+        if (lighting.config1.disable_lut_rg == 0 &&
+            LightingRegs::IsLightingSamplerSupported(lighting.config0.config,
+                                                     LightingRegs::LightingSampler::ReflectGreen)) {
+            refl_value.y =
+                GetLutValue(lighting.lut_input.rg, lighting.abs_lut_input.disable_rg == 0,
+                            lighting.lut_scale.rg, LightingRegs::LightingSampler::ReflectGreen);
+        } else {
+            refl_value.y = refl_value.x;
+        }
+
+        // If enabled, lookup ReflectBlue value, otherwise, ReflectRed value is used
+        if (lighting.config1.disable_lut_rb == 0 &&
+            LightingRegs::IsLightingSamplerSupported(lighting.config0.config,
+                                                     LightingRegs::LightingSampler::ReflectBlue)) {
+            refl_value.z =
+                GetLutValue(lighting.lut_input.rb, lighting.abs_lut_input.disable_rb == 0,
+                            lighting.lut_scale.rb, LightingRegs::LightingSampler::ReflectBlue);
+        } else {
+            refl_value.z = refl_value.x;
+        }
+
+        // Specular 1 component
+        float d1_lut_value = 1.0f;
+        if (lighting.config1.disable_lut_d1 == 0 &&
+            LightingRegs::IsLightingSamplerSupported(
+                lighting.config0.config, LightingRegs::LightingSampler::Distribution1)) {
+            d1_lut_value =
+                GetLutValue(lighting.lut_input.d1, lighting.abs_lut_input.disable_d1 == 0,
+                            lighting.lut_scale.d1, LightingRegs::LightingSampler::Distribution1);
+        }
+
+        Math::Vec3<float> specular_1 =
+            d1_lut_value * refl_value * light_config.specular_1.ToVec3f();
+
+        // Fresnel
+        if (lighting.config1.disable_lut_fr == 0 &&
+            LightingRegs::IsLightingSamplerSupported(lighting.config0.config,
+                                                     LightingRegs::LightingSampler::Fresnel)) {
+
+            float lut_value =
+                GetLutValue(lighting.lut_input.fr, lighting.abs_lut_input.disable_fr == 0,
+                            lighting.lut_scale.fr, LightingRegs::LightingSampler::Fresnel);
+
+            // Enabled for diffuse lighting alpha component
+            if (lighting.config0.fresnel_selector ==
+                    LightingRegs::LightingFresnelSelector::PrimaryAlpha ||
+                lighting.config0.fresnel_selector == LightingRegs::LightingFresnelSelector::Both) {
+                diffuse_sum.a() *= lut_value;
+            }
+
+            // Enabled for the specular lighting alpha component
+            if (lighting.config0.fresnel_selector ==
+                    LightingRegs::LightingFresnelSelector::SecondaryAlpha ||
+                lighting.config0.fresnel_selector == LightingRegs::LightingFresnelSelector::Both) {
+                specular_sum.a() *= lut_value;
+            }
+        }
+
+        auto dot_product = Math::Dot(light_vector, normal);
+
+        // Calculate clamp highlights before applying the two-sided diffuse configuration to the dot
+        // product.
+        float clamp_highlights = 1.0f;
+        if (lighting.config0.clamp_highlights) {
+            if (dot_product <= 0.0f)
+                clamp_highlights = 0.0f;
+            else
+                clamp_highlights = 1.0f;
+        }
+
+        if (light_config.config.two_sided_diffuse)
+            dot_product = std::abs(dot_product);
+        else
+            dot_product = std::max(dot_product, 0.0f);
+
+        auto diffuse =
+            light_config.diffuse.ToVec3f() * dot_product + light_config.ambient.ToVec3f();
+        diffuse_sum += Math::MakeVec(diffuse * dist_atten, 0.0f);
+
+        specular_sum +=
+            Math::MakeVec((specular_0 + specular_1) * clamp_highlights * dist_atten, 0.0f);
+    }
+
+    diffuse_sum += Math::MakeVec(lighting.global_ambient.ToVec3f(), 0.0f);
+
+    auto diffuse = Math::MakeVec<float>(MathUtil::Clamp(diffuse_sum.x, 0.0f, 1.0f) * 255,
+                                        MathUtil::Clamp(diffuse_sum.y, 0.0f, 1.0f) * 255,
+                                        MathUtil::Clamp(diffuse_sum.z, 0.0f, 1.0f) * 255,
+                                        MathUtil::Clamp(diffuse_sum.w, 0.0f, 1.0f) * 255)
+                       .Cast<u8>();
+    auto specular = Math::MakeVec<float>(MathUtil::Clamp(specular_sum.x, 0.0f, 1.0f) * 255,
+                                         MathUtil::Clamp(specular_sum.y, 0.0f, 1.0f) * 255,
+                                         MathUtil::Clamp(specular_sum.z, 0.0f, 1.0f) * 255,
+                                         MathUtil::Clamp(specular_sum.w, 0.0f, 1.0f) * 255)
+                        .Cast<u8>();
+    return {diffuse, specular};
+}
+
+} // namespace Pica
diff --git a/src/video_core/swrasterizer/lighting.h b/src/video_core/swrasterizer/lighting.h
new file mode 100644
index 000000000..438dca926
--- /dev/null
+++ b/src/video_core/swrasterizer/lighting.h
@@ -0,0 +1,18 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <tuple>
+#include "common/quaternion.h"
+#include "common/vector_math.h"
+#include "video_core/pica_state.h"
+
+namespace Pica {
+
+std::tuple<Math::Vec4<u8>, Math::Vec4<u8>> ComputeFragmentsColors(
+    const Pica::LightingRegs& lighting, const Pica::State::Lighting& lighting_state,
+    const Math::Quaternion<float>& normquat, const Math::Vec3<float>& view);
+
+} // namespace Pica
diff --git a/src/video_core/swrasterizer/rasterizer.cpp b/src/video_core/swrasterizer/rasterizer.cpp
index 512e81c08..fdc1df199 100644
--- a/src/video_core/swrasterizer/rasterizer.cpp
+++ b/src/video_core/swrasterizer/rasterizer.cpp
@@ -13,6 +13,7 @@
 #include "common/logging/log.h"
 #include "common/math_util.h"
 #include "common/microprofile.h"
+#include "common/quaternion.h"
 #include "common/vector_math.h"
 #include "core/hw/gpu.h"
 #include "core/memory.h"
@@ -24,6 +25,7 @@
 #include "video_core/regs_texturing.h"
 #include "video_core/shader/shader.h"
 #include "video_core/swrasterizer/framebuffer.h"
+#include "video_core/swrasterizer/lighting.h"
 #include "video_core/swrasterizer/proctex.h"
 #include "video_core/swrasterizer/rasterizer.h"
 #include "video_core/swrasterizer/texturing.h"
@@ -419,6 +421,26 @@ static void ProcessTriangleInternal(const Vertex& v0, const Vertex& v1, const Ve
                 regs.texturing.tev_combiner_buffer_color.a,
             };
 
+            Math::Vec4<u8> primary_fragment_color = {0, 0, 0, 0};
+            Math::Vec4<u8> secondary_fragment_color = {0, 0, 0, 0};
+
+            if (!g_state.regs.lighting.disable) {
+                Math::Quaternion<float> normquat = Math::Quaternion<float>{
+                    {GetInterpolatedAttribute(v0.quat.x, v1.quat.x, v2.quat.x).ToFloat32(),
+                     GetInterpolatedAttribute(v0.quat.y, v1.quat.y, v2.quat.y).ToFloat32(),
+                     GetInterpolatedAttribute(v0.quat.z, v1.quat.z, v2.quat.z).ToFloat32()},
+                    GetInterpolatedAttribute(v0.quat.w, v1.quat.w, v2.quat.w).ToFloat32(),
+                }.Normalized();
+
+                Math::Vec3<float> view{
+                    GetInterpolatedAttribute(v0.view.x, v1.view.x, v2.view.x).ToFloat32(),
+                    GetInterpolatedAttribute(v0.view.y, v1.view.y, v2.view.y).ToFloat32(),
+                    GetInterpolatedAttribute(v0.view.z, v1.view.z, v2.view.z).ToFloat32(),
+                };
+                std::tie(primary_fragment_color, secondary_fragment_color) =
+                    ComputeFragmentsColors(g_state.regs.lighting, g_state.lighting, normquat, view);
+            }
+
             for (unsigned tev_stage_index = 0; tev_stage_index < tev_stages.size();
                  ++tev_stage_index) {
                 const auto& tev_stage = tev_stages[tev_stage_index];
@@ -427,14 +449,13 @@ static void ProcessTriangleInternal(const Vertex& v0, const Vertex& v1, const Ve
                 auto GetSource = [&](Source source) -> Math::Vec4<u8> {
                     switch (source) {
                     case Source::PrimaryColor:
+                        return primary_color;
 
-                    // HACK: Until we implement fragment lighting, use primary_color
                     case Source::PrimaryFragmentColor:
-                        return primary_color;
+                        return primary_fragment_color;
 
-                    // HACK: Until we implement fragment lighting, use zero
                     case Source::SecondaryFragmentColor:
-                        return {0, 0, 0, 0};
+                        return secondary_fragment_color;
 
                     case Source::Texture0:
                         return texture_color[0];