From 656758fd81b679196e734224f07b133a95d51f0b Mon Sep 17 00:00:00 2001
From: Subv <subv2112@gmail.com>
Date: Sun, 19 Aug 2018 12:45:01 -0500
Subject: [PATCH 1/4] Shaders: Added decodings for TLD4 and TLD4S

---
 src/video_core/engines/shader_bytecode.h | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 8ae0e6df2c..7a21fba256 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -524,9 +524,11 @@ public:
         LDG, // Load from global memory
         STG, // Store in global memory
         TEX,
-        TEXQ, // Texture Query
-        TEXS, // Texture Fetch with scalar/non-vec4 source/destinations
-        TLDS, // Texture Load with scalar/non-vec4 source/destinations
+        TEXQ,  // Texture Query
+        TEXS,  // Texture Fetch with scalar/non-vec4 source/destinations
+        TLDS,  // Texture Load with scalar/non-vec4 source/destinations
+        TLD4,  // Texture Load 4
+        TLD4S, // Texture Load 4 with scalar / non - vec4 source / destinations
         EXIT,
         IPA,
         FFMA_IMM, // Fused Multiply and Add
@@ -740,6 +742,8 @@ private:
             INST("1101111101001---", Id::TEXQ, Type::Memory, "TEXQ"),
             INST("1101100---------", Id::TEXS, Type::Memory, "TEXS"),
             INST("1101101---------", Id::TLDS, Type::Memory, "TLDS"),
+            INST("110010----111---", Id::TLD4, Type::Memory, "TLD4"),
+            INST("1101111100------", Id::TLD4S, Type::Memory, "TLD4S"),
             INST("111000110000----", Id::EXIT, Type::Trivial, "EXIT"),
             INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),
             INST("0011001-1-------", Id::FFMA_IMM, Type::Ffma, "FFMA_IMM"),

From 73b937b190d2c9747af2bafd65b0e47b78dbc2e4 Mon Sep 17 00:00:00 2001
From: Subv <subv2112@gmail.com>
Date: Sun, 19 Aug 2018 12:46:35 -0500
Subject: [PATCH 2/4] Shader: Added bitfields for the texture type of the
 various sampling instructions.

---
 src/video_core/engines/shader_bytecode.h | 66 +++++++++++++++++++++++-
 1 file changed, 65 insertions(+), 1 deletion(-)

diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 7a21fba256..569225ad72 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -222,6 +222,13 @@ enum class PredicateResultMode : u64 {
     NotZero = 0x3,
 };
 
+enum class TextureType : u64 {
+    Texture1D = 0,
+    Texture2D = 1,
+    Texture3D = 2,
+    TextureCube = 3,
+};
+
 union Instruction {
     Instruction& operator=(const Instruction& instr) {
         value = instr.value;
@@ -429,6 +436,8 @@ union Instruction {
     } conversion;
 
     union {
+        BitField<28, 1, u64> array;
+        BitField<29, 2, TextureType> texture_type;
         BitField<31, 4, u64> component_mask;
 
         bool IsComponentEnabled(size_t component) const {
@@ -437,9 +446,39 @@ union Instruction {
     } tex;
 
     union {
-        BitField<50, 3, u64> component_mask_selector;
+        BitField<28, 1, u64> array;
+        BitField<29, 2, TextureType> texture_type;
+        BitField<56, 2, u64> component;
+    } tld4;
+
+    union {
+        BitField<52, 2, u64> component;
+    } tld4s;
+
+    union {
         BitField<0, 8, Register> gpr0;
         BitField<28, 8, Register> gpr28;
+        BitField<50, 3, u64> component_mask_selector;
+        BitField<53, 4, u64> texture_info;
+
+        TextureType GetTextureType() const {
+            // The TEXS instruction has a weird encoding for the texture type.
+            if (texture_info == 0)
+                return TextureType::Texture1D;
+            if (texture_info >= 1 && texture_info <= 9)
+                return TextureType::Texture2D;
+            if (texture_info >= 10 && texture_info <= 11)
+                return TextureType::Texture3D;
+            if (texture_info >= 12 && texture_info <= 13)
+                return TextureType::TextureCube;
+
+            UNIMPLEMENTED();
+        }
+
+        bool IsArrayTexture() const {
+            // TEXS only supports Texture2D arrays.
+            return texture_info >= 7 && texture_info <= 9;
+        }
 
         bool HasTwoDestinations() const {
             return gpr28.Value() != Register::ZeroIndex;
@@ -459,6 +498,31 @@ union Instruction {
         }
     } texs;
 
+    union {
+        BitField<53, 4, u64> texture_info;
+
+        TextureType GetTextureType() const {
+            // The TLDS instruction has a weird encoding for the texture type.
+            if (texture_info >= 0 && texture_info <= 1) {
+                return TextureType::Texture1D;
+            }
+            if (texture_info == 2 || texture_info == 8 || texture_info == 12 ||
+                texture_info >= 4 && texture_info <= 6) {
+                return TextureType::Texture2D;
+            }
+            if (texture_info == 7) {
+                return TextureType::Texture3D;
+            }
+
+            UNIMPLEMENTED();
+        }
+
+        bool IsArrayTexture() const {
+            // TEXS only supports Texture2D arrays.
+            return texture_info == 8;
+        }
+    } tlds;
+
     union {
         BitField<20, 24, u64> target;
         BitField<5, 1, u64> constant_buffer;

From 3ef4b3d4b445960576f10d1ba6521580d03e3da8 Mon Sep 17 00:00:00 2001
From: Subv <subv2112@gmail.com>
Date: Sun, 19 Aug 2018 12:54:19 -0500
Subject: [PATCH 3/4] Shader: Use the right sampler type in the TEX, TEXS and
 TLDS instructions.

Different sampler types have their parameters in different registers.
---
 .../renderer_opengl/gl_shader_decompiler.cpp  | 106 ++++++++++++++----
 .../renderer_opengl/gl_shader_gen.h           |  50 ++++++++-
 2 files changed, 127 insertions(+), 29 deletions(-)

diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 4007ecc028..6562be102e 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -439,13 +439,12 @@ public:
         }
         declarations.AddNewLine();
 
-        // Append the sampler2D array for the used textures.
-        size_t num_samplers = GetSamplers().size();
-        if (num_samplers > 0) {
-            declarations.AddLine("uniform sampler2D " + SamplerEntry::GetArrayName(stage) + '[' +
-                                 std::to_string(num_samplers) + "];");
-            declarations.AddNewLine();
+        const auto& samplers = GetSamplers();
+        for (const auto& sampler : samplers) {
+            declarations.AddLine("uniform " + sampler.GetTypeString() + ' ' + sampler.GetName() +
+                                 ';');
         }
+        declarations.AddNewLine();
     }
 
     /// Returns a list of constant buffer declarations
@@ -457,13 +456,14 @@ public:
     }
 
     /// Returns a list of samplers used in the shader
-    std::vector<SamplerEntry> GetSamplers() const {
+    const std::vector<SamplerEntry>& GetSamplers() const {
         return used_samplers;
     }
 
     /// Returns the GLSL sampler used for the input shader sampler, and creates a new one if
     /// necessary.
-    std::string AccessSampler(const Sampler& sampler) {
+    std::string AccessSampler(const Sampler& sampler, Tegra::Shader::TextureType type,
+                              bool is_array) {
         size_t offset = static_cast<size_t>(sampler.index.Value());
 
         // If this sampler has already been used, return the existing mapping.
@@ -472,12 +472,13 @@ public:
                          [&](const SamplerEntry& entry) { return entry.GetOffset() == offset; });
 
         if (itr != used_samplers.end()) {
+            ASSERT(itr->GetType() == type && itr->IsArray() == is_array);
             return itr->GetName();
         }
 
         // Otherwise create a new mapping for this sampler
         size_t next_index = used_samplers.size();
-        SamplerEntry entry{stage, offset, next_index};
+        SamplerEntry entry{stage, offset, next_index, type, is_array};
         used_samplers.emplace_back(entry);
         return entry.GetName();
     }
@@ -638,8 +639,8 @@ private:
     }
 
     /// Generates code representing a texture sampler.
-    std::string GetSampler(const Sampler& sampler) {
-        return regs.AccessSampler(sampler);
+    std::string GetSampler(const Sampler& sampler, Tegra::Shader::TextureType type, bool is_array) {
+        return regs.AccessSampler(sampler, type, is_array);
     }
 
     /**
@@ -1464,10 +1465,29 @@ private:
                 break;
             }
             case OpCode::Id::TEX: {
-                const std::string op_a = regs.GetRegisterAsFloat(instr.gpr8);
-                const std::string op_b = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
-                const std::string sampler = GetSampler(instr.sampler);
-                const std::string coord = "vec2 coords = vec2(" + op_a + ", " + op_b + ");";
+                ASSERT_MSG(instr.tex.array == 0, "TEX arrays unimplemented");
+                std::string coord{};
+
+                switch (instr.tex.texture_type) {
+                case Tegra::Shader::TextureType::Texture2D: {
+                    std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+                    std::string y = regs.GetRegisterAsFloat(instr.gpr20);
+                    coord = "vec2 coords = vec2(" + x + ", " + y + ");";
+                    break;
+                }
+                case Tegra::Shader::TextureType::Texture3D: {
+                    std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+                    std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+                    std::string z = regs.GetRegisterAsFloat(instr.gpr20);
+                    coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
+                    break;
+                }
+                default:
+                    UNIMPLEMENTED();
+                }
+
+                const std::string sampler =
+                    GetSampler(instr.sampler, instr.tex.texture_type, instr.tex.array);
                 // Add an extra scope and declare the texture coords inside to prevent
                 // overwriting them in case they are used as outputs of the texs instruction.
                 shader.AddLine("{");
@@ -1489,20 +1509,60 @@ private:
                 break;
             }
             case OpCode::Id::TEXS: {
-                const std::string op_a = regs.GetRegisterAsFloat(instr.gpr8);
-                const std::string op_b = regs.GetRegisterAsFloat(instr.gpr20);
-                const std::string sampler = GetSampler(instr.sampler);
-                const std::string coord = "vec2 coords = vec2(" + op_a + ", " + op_b + ");";
+                std::string coord{};
+
+                switch (instr.texs.GetTextureType()) {
+                case Tegra::Shader::TextureType::Texture2D: {
+                    if (instr.texs.IsArrayTexture()) {
+                        std::string index = regs.GetRegisterAsInteger(instr.gpr8);
+                        std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+                        std::string y = regs.GetRegisterAsFloat(instr.gpr20);
+                        coord = "vec3 coords = vec3(" + x + ", " + y + ", " + index + ");";
+                    } else {
+                        std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+                        std::string y = regs.GetRegisterAsFloat(instr.gpr20);
+                        coord = "vec2 coords = vec2(" + x + ", " + y + ");";
+                    }
+                    break;
+                }
+                case Tegra::Shader::TextureType::TextureCube: {
+                    std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+                    std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+                    std::string z = regs.GetRegisterAsFloat(instr.gpr20);
+                    coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
+                    break;
+                }
+                default:
+                    UNIMPLEMENTED();
+                }
+                const std::string sampler = GetSampler(instr.sampler, instr.texs.GetTextureType(),
+                                                       instr.texs.IsArrayTexture());
 
                 const std::string texture = "texture(" + sampler + ", coords)";
                 WriteTexsInstruction(instr, coord, texture);
                 break;
             }
             case OpCode::Id::TLDS: {
-                const std::string op_a = regs.GetRegisterAsInteger(instr.gpr8);
-                const std::string op_b = regs.GetRegisterAsInteger(instr.gpr20);
-                const std::string sampler = GetSampler(instr.sampler);
-                const std::string coord = "ivec2 coords = ivec2(" + op_a + ", " + op_b + ");";
+                ASSERT(instr.tlds.GetTextureType() == Tegra::Shader::TextureType::Texture2D);
+                ASSERT(instr.tlds.IsArrayTexture() == false);
+                std::string coord{};
+
+                switch (instr.tlds.GetTextureType()) {
+                case Tegra::Shader::TextureType::Texture2D: {
+                    if (instr.tlds.IsArrayTexture()) {
+                        UNIMPLEMENTED();
+                    } else {
+                        std::string x = regs.GetRegisterAsInteger(instr.gpr8);
+                        std::string y = regs.GetRegisterAsInteger(instr.gpr20);
+                        coord = "ivec2 coords = ivec2(" + x + ", " + y + ");";
+                    }
+                    break;
+                }
+                default:
+                    UNIMPLEMENTED();
+                }
+                const std::string sampler = GetSampler(instr.sampler, instr.tlds.GetTextureType(),
+                                                       instr.tlds.IsArrayTexture());
                 const std::string texture = "texelFetch(" + sampler + ", coords, 0)";
                 WriteTexsInstruction(instr, coord, texture);
                 break;
diff --git a/src/video_core/renderer_opengl/gl_shader_gen.h b/src/video_core/renderer_opengl/gl_shader_gen.h
index 4729ce0fc8..db48da645e 100644
--- a/src/video_core/renderer_opengl/gl_shader_gen.h
+++ b/src/video_core/renderer_opengl/gl_shader_gen.h
@@ -11,6 +11,7 @@
 #include <vector>
 #include "common/common_types.h"
 #include "common/hash.h"
+#include "video_core/engines/shader_bytecode.h"
 
 namespace GLShader {
 
@@ -72,8 +73,9 @@ class SamplerEntry {
     using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
 public:
-    SamplerEntry(Maxwell::ShaderStage stage, size_t offset, size_t index)
-        : offset(offset), stage(stage), sampler_index(index) {}
+    SamplerEntry(Maxwell::ShaderStage stage, size_t offset, size_t index,
+                 Tegra::Shader::TextureType type, bool is_array)
+        : offset(offset), stage(stage), sampler_index(index), type(type), is_array(is_array) {}
 
     size_t GetOffset() const {
         return offset;
@@ -88,8 +90,41 @@ public:
     }
 
     std::string GetName() const {
-        return std::string(TextureSamplerNames[static_cast<size_t>(stage)]) + '[' +
-               std::to_string(sampler_index) + ']';
+        return std::string(TextureSamplerNames[static_cast<size_t>(stage)]) + '_' +
+               std::to_string(sampler_index);
+    }
+
+    std::string GetTypeString() const {
+        using Tegra::Shader::TextureType;
+        std::string glsl_type;
+
+        switch (type) {
+        case TextureType::Texture1D:
+            glsl_type = "sampler1D";
+            break;
+        case TextureType::Texture2D:
+            glsl_type = "sampler2D";
+            break;
+        case TextureType::Texture3D:
+            glsl_type = "sampler3D";
+            break;
+        case TextureType::TextureCube:
+            glsl_type = "samplerCube";
+            break;
+        default:
+            UNIMPLEMENTED();
+        }
+        if (is_array)
+            glsl_type += "Array";
+        return glsl_type;
+    }
+
+    Tegra::Shader::TextureType GetType() const {
+        return type;
+    }
+
+    bool IsArray() const {
+        return is_array;
     }
 
     static std::string GetArrayName(Maxwell::ShaderStage stage) {
@@ -100,11 +135,14 @@ private:
     static constexpr std::array<const char*, Maxwell::MaxShaderStage> TextureSamplerNames = {
         "tex_vs", "tex_tessc", "tex_tesse", "tex_gs", "tex_fs",
     };
+
     /// Offset in TSC memory from which to read the sampler object, as specified by the sampling
     /// instruction.
     size_t offset;
-    Maxwell::ShaderStage stage; ///< Shader stage where this sampler was used.
-    size_t sampler_index;       ///< Value used to index into the generated GLSL sampler array.
+    Maxwell::ShaderStage stage;      ///< Shader stage where this sampler was used.
+    size_t sampler_index;            ///< Value used to index into the generated GLSL sampler array.
+    Tegra::Shader::TextureType type; ///< The type used to sample this texture (Texture2D, etc)
+    bool is_array; ///< Whether the texture is being sampled as an array texture or not.
 };
 
 struct ShaderEntries {

From 7fb406c3fca2f7c8266a5cc1988aa8d741921310 Mon Sep 17 00:00:00 2001
From: Subv <subv2112@gmail.com>
Date: Sun, 19 Aug 2018 12:55:50 -0500
Subject: [PATCH 4/4] Shader: Implemented the TLD4 and TLD4S opcodes using
 GLSL's textureGather.

It is unknown how TLD4S determines the sampler type, more research is needed.
---
 .../renderer_opengl/gl_shader_decompiler.cpp  | 51 +++++++++++++++++++
 1 file changed, 51 insertions(+)

diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 6562be102e..f466b94271 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -1567,6 +1567,57 @@ private:
                 WriteTexsInstruction(instr, coord, texture);
                 break;
             }
+            case OpCode::Id::TLD4: {
+                ASSERT(instr.tld4.texture_type == Tegra::Shader::TextureType::Texture2D);
+                ASSERT(instr.tld4.array == 0);
+                std::string coord{};
+
+                switch (instr.tld4.texture_type) {
+                case Tegra::Shader::TextureType::Texture2D: {
+                    std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+                    std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+                    coord = "vec2 coords = vec2(" + x + ", " + y + ");";
+                    break;
+                }
+                default:
+                    UNIMPLEMENTED();
+                }
+
+                const std::string sampler =
+                    GetSampler(instr.sampler, instr.tld4.texture_type, instr.tld4.array);
+                // Add an extra scope and declare the texture coords inside to prevent
+                // overwriting them in case they are used as outputs of the texs instruction.
+                shader.AddLine("{");
+                ++shader.scope;
+                shader.AddLine(coord);
+                const std::string texture = "textureGather(" + sampler + ", coords, " +
+                                            std::to_string(instr.tld4.component) + ')';
+
+                size_t dest_elem{};
+                for (size_t elem = 0; elem < 4; ++elem) {
+                    if (!instr.tex.IsComponentEnabled(elem)) {
+                        // Skip disabled components
+                        continue;
+                    }
+                    regs.SetRegisterToFloat(instr.gpr0, elem, texture, 1, 4, false, dest_elem);
+                    ++dest_elem;
+                }
+                --shader.scope;
+                shader.AddLine("}");
+                break;
+            }
+            case OpCode::Id::TLD4S: {
+                const std::string op_a = regs.GetRegisterAsFloat(instr.gpr8);
+                const std::string op_b = regs.GetRegisterAsFloat(instr.gpr20);
+                // TODO(Subv): Figure out how the sampler type is encoded in the TLD4S instruction.
+                const std::string sampler =
+                    GetSampler(instr.sampler, Tegra::Shader::TextureType::Texture2D, false);
+                const std::string coord = "vec2 coords = vec2(" + op_a + ", " + op_b + ");";
+                const std::string texture = "textureGather(" + sampler + ", coords, " +
+                                            std::to_string(instr.tld4s.component) + ')';
+                WriteTexsInstruction(instr, coord, texture);
+                break;
+            }
             default: {
                 LOG_CRITICAL(HW_GPU, "Unhandled memory instruction: {}", opcode->GetName());
                 UNREACHABLE();