From c4ac05c82c49e678ca78147b2716e0a26b103f8d Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Tue, 26 Mar 2019 14:37:43 -0400
Subject: [PATCH 01/14] Implement Const Buffer Accessor

---
 src/video_core/CMakeLists.txt                 |  2 ++
 src/video_core/const_buffer_accessor.cpp      | 35 +++++++++++++++++++
 src/video_core/const_buffer_accessor.h        | 21 +++++++++++
 .../renderer_opengl/gl_rasterizer.cpp         |  6 ++--
 .../renderer_opengl/gl_rasterizer.h           |  3 ++
 5 files changed, 65 insertions(+), 2 deletions(-)
 create mode 100644 src/video_core/const_buffer_accessor.cpp
 create mode 100644 src/video_core/const_buffer_accessor.h

diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 242a0d1cd..804395d38 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -1,4 +1,6 @@
 add_library(video_core STATIC
+    const_buffer_accessor.cpp
+    const_buffer_accessor.h
     dma_pusher.cpp
     dma_pusher.h
     debug_utils/debug_utils.cpp
diff --git a/src/video_core/const_buffer_accessor.cpp b/src/video_core/const_buffer_accessor.cpp
new file mode 100644
index 000000000..c89ab91c7
--- /dev/null
+++ b/src/video_core/const_buffer_accessor.cpp
@@ -0,0 +1,35 @@
+#pragma once
+
+#include <cstring>
+
+#include "common/common_types.h"
+#include "core/core.h"
+#include "core/memory.h"
+#include "video_core/const_buffer_accessor.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
+
+namespace Tegra {
+
+u32 ConstBufferAccessor::access32(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
+                                  u64 const_buffer, u64 offset) {
+    auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
+    const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<std::size_t>(stage)];
+    const auto& buffer = shader_stage.const_buffers[const_buffer];
+    u32 result;
+    std::memcpy(&result, memory_manager.GetPointer(buffer.address + offset * 4), sizeof(u32));
+    return result;
+}
+
+u64 ConstBufferAccessor::access64(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
+                                  u64 const_buffer, u64 offset) {
+    auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
+    const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<std::size_t>(stage)];
+    const auto& buffer = shader_stage.const_buffers[const_buffer];
+    u64 result;
+    std::memcpy(&result, memory_manager.GetPointer(buffer.address + offset * 4), sizeof(u64));
+    return result;
+}
+
+} // namespace Tegra
diff --git a/src/video_core/const_buffer_accessor.h b/src/video_core/const_buffer_accessor.h
new file mode 100644
index 000000000..2410f1483
--- /dev/null
+++ b/src/video_core/const_buffer_accessor.h
@@ -0,0 +1,21 @@
+#pragma once
+
+#include "common/common_types.h"
+#include "video_core/engines/maxwell_3d.h"
+
+namespace Tegra {
+
+class ConstBufferAccessor {
+public:
+    ConstBufferAccessor(Tegra::Engines::Maxwell3D& maxwell3d) : maxwell3d(maxwell3d) {}
+    ~ConstBufferAccessor() = default;
+
+    u32 access32(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, u64 const_buffer, u64 offset);
+
+    u64 access64(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, u64 const_buffer, u64 offset);
+
+private:
+    Tegra::Engines::Maxwell3D& maxwell3d;
+};
+
+} // namespace Tegra
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 7ff1e6737..f2ac5382d 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -98,9 +98,11 @@ struct FramebufferCacheKey {
     }
 };
 
-RasterizerOpenGL::RasterizerOpenGL(Core::System& system, ScreenInfo& info)
+RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, Core::System& system,
+                                   ScreenInfo& info)
     : res_cache{*this}, shader_cache{*this, system}, global_cache{*this}, system{system},
-      screen_info{info}, buffer_cache(*this, STREAM_BUFFER_SIZE) {
+      screen_info{info}, buffer_cache(*this, STREAM_BUFFER_SIZE),
+      const_buffer_accessor(system.GPU().Maxwell3D()) {
     // Create sampler objects
     for (std::size_t i = 0; i < texture_samplers.size(); ++i) {
         texture_samplers[i].Create();
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index 54fbf48aa..886e9c2b5 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -17,6 +17,7 @@
 #include <glad/glad.h>
 
 #include "common/common_types.h"
+#include "video_core/const_buffer_accessor.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/rasterizer_cache.h"
 #include "video_core/rasterizer_interface.h"
@@ -229,6 +230,8 @@ private:
     PrimitiveAssembler primitive_assembler{buffer_cache};
     GLint uniform_buffer_alignment;
 
+    Tegra::ConstBufferAccessor const_buffer_accessor;
+
     std::size_t CalculateVertexArraysSize() const;
 
     std::size_t CalculateIndexBufferSize() const;

From e28fd3d0a533695242d17350dd929ad3bb56c429 Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Tue, 26 Mar 2019 17:05:23 -0400
Subject: [PATCH 02/14] Implement Bindless Samplers and TEX_B in the IR.

---
 src/video_core/engines/shader_bytecode.h      |  2 +
 .../renderer_opengl/gl_shader_disk_cache.cpp  |  2 +-
 src/video_core/shader/decode/texture.cpp      | 58 +++++++++++++++++--
 src/video_core/shader/shader_ir.h             | 31 +++++++---
 4 files changed, 77 insertions(+), 16 deletions(-)

diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 7f613370b..2edd3245e 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -1309,6 +1309,7 @@ public:
         LDG, // Load from global memory
         STG, // Store in global memory
         TEX,
+        TEX_B,  // Texture Load Bindless
         TXQ,    // Texture Query
         TEXS,   // Texture Fetch with scalar/non-vec4 source/destinations
         TLDS,   // Texture Load with scalar/non-vec4 source/destinations
@@ -1577,6 +1578,7 @@ private:
             INST("1110111011010---", Id::LDG, Type::Memory, "LDG"),
             INST("1110111011011---", Id::STG, Type::Memory, "STG"),
             INST("110000----111---", Id::TEX, Type::Texture, "TEX"),
+            INST("1101111010111---", Id::TEX_B, Type::Texture, "TEX_B"),
             INST("1101111101001---", Id::TXQ, Type::Texture, "TXQ"),
             INST("1101-00---------", Id::TEXS, Type::Texture, "TEXS"),
             INST("1101101---------", Id::TLDS, Type::Texture, "TLDS"),
diff --git a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
index 8a43eb157..6a95af6f6 100644
--- a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
@@ -328,7 +328,7 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
         }
         entry.entries.samplers.emplace_back(
             static_cast<std::size_t>(offset), static_cast<std::size_t>(index),
-            static_cast<Tegra::Shader::TextureType>(type), is_array != 0, is_shadow != 0);
+            static_cast<Tegra::Shader::TextureType>(type), is_array != 0, is_shadow != 0, false);
     }
 
     u32 global_memory_count{};
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index a775b402b..23f2ad999 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -57,6 +57,23 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
             GetTexCode(instr, texture_type, process_mode, depth_compare, is_array, is_aoffi));
         break;
     }
+    case OpCode::Id::TEX_B: {
+        UNIMPLEMENTED_IF_MSG(instr.tex.UsesMiscMode(TextureMiscMode::AOFFI),
+                             "AOFFI is not implemented");
+
+        if (instr.tex.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TEX.NODEP implementation is incomplete");
+        }
+
+        const TextureType texture_type{instr.tex.texture_type};
+        const bool is_array = instr.tex.array != 0;
+        const bool depth_compare = instr.tex.UsesMiscMode(TextureMiscMode::DC);
+        const auto process_mode = instr.tex.GetTextureProcessMode();
+        WriteTexInstructionFloat(bb, instr,
+                                 GetTexCode(instr, texture_type, process_mode, depth_compare,
+                                            is_array, true, instr.gpr20));
+        break;
+    }
     case OpCode::Id::TEXS: {
         const TextureType texture_type{instr.texs.GetTextureType()};
         const bool is_array{instr.texs.IsArrayTexture()};
@@ -250,10 +267,36 @@ const Sampler& ShaderIR::GetSampler(const Tegra::Shader::Sampler& sampler, Textu
 
     // Otherwise create a new mapping for this sampler
     const std::size_t next_index = used_samplers.size();
-    const Sampler entry{offset, next_index, type, is_array, is_shadow};
+    const Sampler entry{offset, next_index, type, is_array, is_shadow, false};
     return *used_samplers.emplace(entry).first;
 }
 
+const Sampler& ShaderIR::GetBindlessSampler(const Tegra::Shader::Register& reg,
+                                            TextureType type, bool is_array, bool is_shadow) {
+
+    const Node sampler_register = GetRegister(reg);
+    const Node base_sampler = TrackCbuf(sampler_register, global_code, static_cast<s64>(global_code.size()));
+    const auto cbuf = std::get_if<CbufNode>(base_sampler);
+    const auto cbuf_offset_imm = std::get_if<ImmediateNode>(cbuf->GetOffset());
+    ASSERT(cbuf_offset_imm != nullptr);
+    const auto cbuf_offset = cbuf_offset_imm->GetValue();
+    const auto cbuf_index = cbuf->GetIndex();
+    const std::pair<u32, u32> cbuf_pair = {cbuf_index, cbuf_offset};
+
+    // If this sampler has already been used, return the existing mapping.
+    if (used_bindless_samplers.count(cbuf_pair) > 0) {
+        const auto& sampler = used_bindless_samplers[cbuf_pair];
+        ASSERT(sampler.GetType() == type && sampler.IsArray() == is_array &&
+               sampler.IsShadow() == is_shadow);
+        return sampler;
+    }
+
+    // Otherwise create a new mapping for this sampler
+    const std::size_t next_index = used_bindless_samplers.size();
+    const Sampler entry{0, next_index, type, is_array, is_shadow, true};
+    return (*used_bindless_samplers.emplace(std::make_pair(cbuf_pair, entry)).first).second;
+}
+
 void ShaderIR::WriteTexInstructionFloat(NodeBlock& bb, Instruction instr, const Node4& components) {
     u32 dest_elem = 0;
     for (u32 elem = 0; elem < 4; ++elem) {
@@ -325,8 +368,8 @@ void ShaderIR::WriteTexsInstructionHalfFloat(NodeBlock& bb, Instruction instr,
 
 Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
                                TextureProcessMode process_mode, std::vector<Node> coords,
-                               Node array, Node depth_compare, u32 bias_offset,
-                               std::vector<Node> aoffi) {
+                               Node array, Node depth_compare, u32 bias_offset, std::vector<Node> aoffi, bool is_bindless,
+                               Register bindless_reg) {
     const bool is_array = array;
     const bool is_shadow = depth_compare;
 
@@ -334,7 +377,9 @@ Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
                              (texture_type == TextureType::TextureCube && is_array && is_shadow),
                          "This method is not supported.");
 
-    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, is_shadow);
+    const auto& sampler = !is_bindless
+                              ? GetSampler(instr.sampler, texture_type, is_array, is_shadow)
+                              : GetBindlessSampler(bindless_reg, texture_type, is_array, is_shadow);
 
     const bool lod_needed = process_mode == TextureProcessMode::LZ ||
                             process_mode == TextureProcessMode::LL ||
@@ -384,7 +429,7 @@ Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
 
 Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
                            TextureProcessMode process_mode, bool depth_compare, bool is_array,
-                           bool is_aoffi) {
+                           bool is_aoffi, bool is_bindless, Register bindless_reg) {
     const bool lod_bias_enabled{
         (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ)};
 
@@ -423,7 +468,8 @@ Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
         dc = GetRegister(parameter_register++);
     }
 
-    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0, aoffi);
+    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0, aoffi, is_bindless,
+                          bindless_reg);
 }
 
 Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index 4888998d3..712dc3ddb 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -196,9 +196,12 @@ enum class ExitMethod {
 
 class Sampler {
 public:
+    Sampler() = default;
     explicit Sampler(std::size_t offset, std::size_t index, Tegra::Shader::TextureType type,
-                     bool is_array, bool is_shadow)
-        : offset{offset}, index{index}, type{type}, is_array{is_array}, is_shadow{is_shadow} {}
+                     bool is_array, bool is_shadow, bool is_bindless)
+        : offset{offset}, index{index}, type{type}, is_array{is_array}, is_shadow{is_shadow}, is_bindless{is_bindless} {}
+
+    ~Sampler() = default;
 
     std::size_t GetOffset() const {
         return offset;
@@ -233,6 +236,7 @@ private:
     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.
     bool is_shadow{}; ///< Whether the texture is being sampled as a depth texture or not.
+    bool is_bindless{}; ///< Whether this sampler belongs to a bindless texture or not.
 };
 
 class ConstBuffer {
@@ -730,6 +734,10 @@ private:
     const Sampler& GetSampler(const Tegra::Shader::Sampler& sampler,
                               Tegra::Shader::TextureType type, bool is_array, bool is_shadow);
 
+    // Accesses a texture sampler for a bindless texture.
+    const Sampler& GetBindlessSampler(const Tegra::Shader::Register& reg, Tegra::Shader::TextureType type,
+                                      bool is_array, bool is_shadow);
+
     /// Extracts a sequence of bits from a node
     Node BitfieldExtract(Node value, u32 offset, u32 bits);
 
@@ -741,9 +749,11 @@ private:
     void WriteTexsInstructionHalfFloat(NodeBlock& bb, Tegra::Shader::Instruction instr,
                                        const Node4& components);
 
-    Node4 GetTexCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
-                     Tegra::Shader::TextureProcessMode process_mode, bool depth_compare,
-                     bool is_array, bool is_aoffi);
+    Node4 GetTexCode(
+        Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
+        Tegra::Shader::TextureProcessMode process_mode, bool depth_compare, bool is_array,
+        bool is_aoffi, bool is_bindless = false,
+        Tegra::Shader::Register bindless_reg = static_cast<Tegra::Shader::Register>(0));
 
     Node4 GetTexsCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
                       Tegra::Shader::TextureProcessMode process_mode, bool depth_compare,
@@ -760,10 +770,12 @@ private:
         bool lod_bias_enabled, std::size_t max_coords, std::size_t max_inputs);
 
     std::vector<Node> GetAoffiCoordinates(Node aoffi_reg, std::size_t coord_count, bool is_tld4);
-
-    Node4 GetTextureCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
-                         Tegra::Shader::TextureProcessMode process_mode, std::vector<Node> coords,
-                         Node array, Node depth_compare, u32 bias_offset, std::vector<Node> aoffi);
+    
+    Node4 GetTextureCode(
+        Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
+        Tegra::Shader::TextureProcessMode process_mode, std::vector<Node> coords, Node array,
+        Node depth_compare, u32 bias_offset, std::vector<Node> aoffi, bool is_bindless = false,
+        Tegra::Shader::Register bindless_reg = static_cast<Tegra::Shader::Register>(0));
 
     Node GetVideoOperand(Node op, bool is_chunk, bool is_signed, Tegra::Shader::VideoType type,
                          u64 byte_height);
@@ -833,6 +845,7 @@ private:
     std::set<Tegra::Shader::Attribute::Index> used_output_attributes;
     std::map<u32, ConstBuffer> used_cbufs;
     std::set<Sampler> used_samplers;
+    std::map<std::pair<u32, u32>, Sampler> used_bindless_samplers;
     std::array<bool, Tegra::Engines::Maxwell3D::Regs::NumClipDistances> used_clip_distances{};
     std::set<GlobalMemoryBase> used_global_memory_bases;
 

From fe392fff2425c10c9683a4058c779d352b9855ec Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Tue, 26 Mar 2019 17:56:16 -0400
Subject: [PATCH 03/14] Unify both sampler types.

---
 .../renderer_opengl/gl_shader_decompiler.h    |  3 +-
 .../renderer_opengl/gl_shader_disk_cache.cpp  |  9 +++--
 src/video_core/shader/decode/texture.cpp      | 22 ++++++------
 src/video_core/shader/shader_ir.h             | 36 ++++++++++++++-----
 4 files changed, 48 insertions(+), 22 deletions(-)

diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.h b/src/video_core/renderer_opengl/gl_shader_decompiler.h
index 4e04ab2f8..9f7b7272e 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.h
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.h
@@ -58,6 +58,7 @@ private:
 struct ShaderEntries {
     std::vector<ConstBufferEntry> const_buffers;
     std::vector<SamplerEntry> samplers;
+    std::vector<SamplerEntry> bindless_samplers;
     std::vector<GlobalMemoryEntry> global_memory_entries;
     std::array<bool, Maxwell::NumClipDistances> clip_distances{};
     std::size_t shader_length{};
@@ -68,4 +69,4 @@ std::string GetCommonDeclarations();
 ProgramResult Decompile(const VideoCommon::Shader::ShaderIR& ir, Maxwell::ShaderStage stage,
                         const std::string& suffix);
 
-} // namespace OpenGL::GLShader
\ No newline at end of file
+} // namespace OpenGL::GLShader
diff --git a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
index 6a95af6f6..e27740383 100644
--- a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
@@ -319,16 +319,18 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
         u32 type{};
         u8 is_array{};
         u8 is_shadow{};
+        u8 is_bindless{};
         if (file.ReadBytes(&offset, sizeof(u64)) != sizeof(u64) ||
             file.ReadBytes(&index, sizeof(u64)) != sizeof(u64) ||
             file.ReadBytes(&type, sizeof(u32)) != sizeof(u32) ||
             file.ReadBytes(&is_array, sizeof(u8)) != sizeof(u8) ||
-            file.ReadBytes(&is_shadow, sizeof(u8)) != sizeof(u8)) {
+            file.ReadBytes(&is_shadow, sizeof(u8)) != sizeof(u8) ||
+            file.ReadBytes(&is_bindless, sizeof(u8)) != sizeof(u8)) {
             return {};
         }
         entry.entries.samplers.emplace_back(
             static_cast<std::size_t>(offset), static_cast<std::size_t>(index),
-            static_cast<Tegra::Shader::TextureType>(type), is_array != 0, is_shadow != 0, false);
+            static_cast<Tegra::Shader::TextureType>(type), is_array != 0, is_shadow != 0, is_bindless != 0);
     }
 
     u32 global_memory_count{};
@@ -388,7 +390,8 @@ bool ShaderDiskCacheOpenGL::SaveDecompiledFile(FileUtil::IOFile& file, u64 uniqu
             file.WriteObject(static_cast<u64>(sampler.GetIndex())) != 1 ||
             file.WriteObject(static_cast<u32>(sampler.GetType())) != 1 ||
             file.WriteObject(static_cast<u8>(sampler.IsArray() ? 1 : 0)) != 1 ||
-            file.WriteObject(static_cast<u8>(sampler.IsShadow() ? 1 : 0)) != 1) {
+            file.WriteObject(static_cast<u8>(sampler.IsShadow() ? 1 : 0)) != 1 ||
+            file.WriteObject(static_cast<u8>(sampler.IsBindless() ? 1 : 0)) != 1) {
             return false;
         }
     }
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index 23f2ad999..3ac04f6b7 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -267,7 +267,7 @@ const Sampler& ShaderIR::GetSampler(const Tegra::Shader::Sampler& sampler, Textu
 
     // Otherwise create a new mapping for this sampler
     const std::size_t next_index = used_samplers.size();
-    const Sampler entry{offset, next_index, type, is_array, is_shadow, false};
+    const Sampler entry{offset, next_index, type, is_array, is_shadow};
     return *used_samplers.emplace(entry).first;
 }
 
@@ -281,20 +281,22 @@ const Sampler& ShaderIR::GetBindlessSampler(const Tegra::Shader::Register& reg,
     ASSERT(cbuf_offset_imm != nullptr);
     const auto cbuf_offset = cbuf_offset_imm->GetValue();
     const auto cbuf_index = cbuf->GetIndex();
-    const std::pair<u32, u32> cbuf_pair = {cbuf_index, cbuf_offset};
+    const u64 cbuf_key = (cbuf_index << 32) | cbuf_offset;
 
     // If this sampler has already been used, return the existing mapping.
-    if (used_bindless_samplers.count(cbuf_pair) > 0) {
-        const auto& sampler = used_bindless_samplers[cbuf_pair];
-        ASSERT(sampler.GetType() == type && sampler.IsArray() == is_array &&
-               sampler.IsShadow() == is_shadow);
-        return sampler;
+    const auto itr =
+        std::find_if(used_samplers.begin(), used_samplers.end(),
+                     [&](const Sampler& entry) { return entry.GetOffset() == cbuf_key; });
+    if (itr != used_samplers.end()) {
+        ASSERT(itr->GetType() == type && itr->IsArray() == is_array &&
+               itr->IsShadow() == is_shadow);
+        return *itr;
     }
 
     // Otherwise create a new mapping for this sampler
-    const std::size_t next_index = used_bindless_samplers.size();
-    const Sampler entry{0, next_index, type, is_array, is_shadow, true};
-    return (*used_bindless_samplers.emplace(std::make_pair(cbuf_pair, entry)).first).second;
+    const std::size_t next_index = used_samplers.size();
+    const Sampler entry{cbuf_index, cbuf_offset, next_index, type, is_array, is_shadow};
+    return *used_samplers.emplace(entry).first;
 }
 
 void ShaderIR::WriteTexInstructionFloat(NodeBlock& bb, Instruction instr, const Node4& components) {
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index 712dc3ddb..773c71fa5 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -196,12 +196,24 @@ enum class ExitMethod {
 
 class Sampler {
 public:
-    Sampler() = default;
+    // Use this constructor for binded Samplers
+    explicit Sampler(std::size_t offset, std::size_t index, Tegra::Shader::TextureType type,
+                     bool is_array, bool is_shadow)
+        : offset{offset}, index{index}, type{type}, is_array{is_array}, is_shadow{is_shadow},
+          is_bindless{false} {}
+
+    // Use this constructor for bindless Samplers
+    explicit Sampler(u32 cbuf_index, u32 cbuf_offset, std::size_t index,
+                     Tegra::Shader::TextureType type, bool is_array, bool is_shadow)
+        : offset{(static_cast<u64>(cbuf_index) << 32) | cbuf_offset}, index{index}, type{type}, is_array{is_array},
+          is_shadow{is_shadow}, is_bindless{true} {}
+
+    // Use this only for serialization/deserialization
     explicit Sampler(std::size_t offset, std::size_t index, Tegra::Shader::TextureType type,
                      bool is_array, bool is_shadow, bool is_bindless)
-        : offset{offset}, index{index}, type{type}, is_array{is_array}, is_shadow{is_shadow}, is_bindless{is_bindless} {}
+        : offset{offset}, index{index}, type{type}, is_array{is_array}, is_shadow{is_shadow},
+          is_bindless{is_bindless} {}
 
-    ~Sampler() = default;
 
     std::size_t GetOffset() const {
         return offset;
@@ -223,6 +235,14 @@ public:
         return is_shadow;
     }
 
+    bool IsBindless() const {
+        return is_bindless;
+    }
+
+    std::pair<u32, u32> GetBindlessCBuf() {
+        return {offset >> 32, offset & 0x00000000FFFFFFFFULL};
+    }
+
     bool operator<(const Sampler& rhs) const {
         return std::tie(offset, index, type, is_array, is_shadow) <
                std::tie(rhs.offset, rhs.index, rhs.type, rhs.is_array, rhs.is_shadow);
@@ -234,8 +254,8 @@ private:
     std::size_t offset{};
     std::size_t 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.
-    bool is_shadow{}; ///< Whether the texture is being sampled as a depth texture or not.
+    bool is_array{};    ///< Whether the texture is being sampled as an array texture or not.
+    bool is_shadow{};   ///< Whether the texture is being sampled as a depth texture or not.
     bool is_bindless{}; ///< Whether this sampler belongs to a bindless texture or not.
 };
 
@@ -735,8 +755,9 @@ private:
                               Tegra::Shader::TextureType type, bool is_array, bool is_shadow);
 
     // Accesses a texture sampler for a bindless texture.
-    const Sampler& GetBindlessSampler(const Tegra::Shader::Register& reg, Tegra::Shader::TextureType type,
-                                      bool is_array, bool is_shadow);
+    const Sampler& GetBindlessSampler(const Tegra::Shader::Register& reg,
+                                      Tegra::Shader::TextureType type, bool is_array,
+                                      bool is_shadow);
 
     /// Extracts a sequence of bits from a node
     Node BitfieldExtract(Node value, u32 offset, u32 bits);
@@ -845,7 +866,6 @@ private:
     std::set<Tegra::Shader::Attribute::Index> used_output_attributes;
     std::map<u32, ConstBuffer> used_cbufs;
     std::set<Sampler> used_samplers;
-    std::map<std::pair<u32, u32>, Sampler> used_bindless_samplers;
     std::array<bool, Tegra::Engines::Maxwell3D::Regs::NumClipDistances> used_clip_distances{};
     std::set<GlobalMemoryBase> used_global_memory_bases;
 

From 7af82ca022fd6f02583e5686d5c69baf0b6a3611 Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Tue, 26 Mar 2019 18:18:54 -0400
Subject: [PATCH 04/14] Implement Bindless Handling on SetupTexture

---
 src/video_core/engines/maxwell_3d.cpp         | 31 +++++++++++--------
 src/video_core/engines/maxwell_3d.h           |  4 +++
 .../renderer_opengl/gl_rasterizer.cpp         | 10 +++++-
 src/video_core/shader/shader_ir.h             |  7 ++---
 4 files changed, 34 insertions(+), 18 deletions(-)

diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp
index 74403eed4..079132135 100644
--- a/src/video_core/engines/maxwell_3d.cpp
+++ b/src/video_core/engines/maxwell_3d.cpp
@@ -482,19 +482,8 @@ std::vector<Texture::FullTextureInfo> Maxwell3D::GetStageTextures(Regs::ShaderSt
     return textures;
 }
 
-Texture::FullTextureInfo Maxwell3D::GetStageTexture(Regs::ShaderStage stage,
-                                                    std::size_t offset) const {
-    auto& shader = state.shader_stages[static_cast<std::size_t>(stage)];
-    auto& tex_info_buffer = shader.const_buffers[regs.tex_cb_index];
-    ASSERT(tex_info_buffer.enabled && tex_info_buffer.address != 0);
-
-    const GPUVAddr tex_info_address =
-        tex_info_buffer.address + offset * sizeof(Texture::TextureHandle);
-
-    ASSERT(tex_info_address < tex_info_buffer.address + tex_info_buffer.size);
-
-    const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(tex_info_address)};
-
+Texture::FullTextureInfo Maxwell3D::GetTextureInfo(const Texture::TextureHandle tex_handle,
+                                                   std::size_t offset) const {
     Texture::FullTextureInfo tex_info{};
     tex_info.index = static_cast<u32>(offset);
 
@@ -511,6 +500,22 @@ Texture::FullTextureInfo Maxwell3D::GetStageTexture(Regs::ShaderStage stage,
     return tex_info;
 }
 
+Texture::FullTextureInfo Maxwell3D::GetStageTexture(Regs::ShaderStage stage,
+                                                    std::size_t offset) const {
+    auto& shader = state.shader_stages[static_cast<std::size_t>(stage)];
+    auto& tex_info_buffer = shader.const_buffers[regs.tex_cb_index];
+    ASSERT(tex_info_buffer.enabled && tex_info_buffer.address != 0);
+
+    const GPUVAddr tex_info_address =
+        tex_info_buffer.address + offset * sizeof(Texture::TextureHandle);
+
+    ASSERT(tex_info_address < tex_info_buffer.address + tex_info_buffer.size);
+
+    const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(tex_info_address)};
+
+    return GetTextureInfo(tex_handle, offset);
+}
+
 u32 Maxwell3D::GetRegisterValue(u32 method) const {
     ASSERT_MSG(method < Regs::NUM_REGS, "Invalid Maxwell3D register");
     return regs.reg_array[method];
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index 321af3297..fd2c35a01 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -1131,6 +1131,10 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    /// Given a Texture Handle, returns the TSC and TIC entries.
+    Texture::FullTextureInfo GetTextureInfo(const Texture::TextureHandle tex_handle,
+                                            std::size_t offset) const;
+
     /// Returns a list of enabled textures for the specified shader stage.
     std::vector<Texture::FullTextureInfo> GetStageTextures(Regs::ShaderStage stage) const;
 
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index f2ac5382d..a20acfe8e 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -985,7 +985,15 @@ void RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, const Shader& s
 
     for (u32 bindpoint = 0; bindpoint < entries.size(); ++bindpoint) {
         const auto& entry = entries[bindpoint];
-        const auto texture = maxwell3d.GetStageTexture(stage, entry.GetOffset());
+        Tegra::Texture::FullTextureInfo texture;
+        if (!entry.IsBindless()) {
+            texture = maxwell3d.GetStageTexture(stage, entry.GetOffset());
+        } else {
+            const auto cbuf = entry.GetBindlessCBuf();
+            Tegra::Texture::TextureHandle tex_handle;
+            tex_handle.raw = const_buffer_accessor.access32(stage, cbuf.first, cbuf.second);
+            texture = maxwell3d.GetTextureInfo(tex_handle, entry.GetOffset());
+        }
         const u32 current_bindpoint = base_bindings.sampler + bindpoint;
 
         texture_samplers[current_bindpoint].SyncWithConfig(texture.tsc);
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index 773c71fa5..ed321cfe5 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -205,8 +205,8 @@ public:
     // Use this constructor for bindless Samplers
     explicit Sampler(u32 cbuf_index, u32 cbuf_offset, std::size_t index,
                      Tegra::Shader::TextureType type, bool is_array, bool is_shadow)
-        : offset{(static_cast<u64>(cbuf_index) << 32) | cbuf_offset}, index{index}, type{type}, is_array{is_array},
-          is_shadow{is_shadow}, is_bindless{true} {}
+        : offset{(static_cast<u64>(cbuf_index) << 32) | cbuf_offset}, index{index}, type{type},
+          is_array{is_array}, is_shadow{is_shadow}, is_bindless{true} {}
 
     // Use this only for serialization/deserialization
     explicit Sampler(std::size_t offset, std::size_t index, Tegra::Shader::TextureType type,
@@ -214,7 +214,6 @@ public:
         : offset{offset}, index{index}, type{type}, is_array{is_array}, is_shadow{is_shadow},
           is_bindless{is_bindless} {}
 
-
     std::size_t GetOffset() const {
         return offset;
     }
@@ -239,7 +238,7 @@ public:
         return is_bindless;
     }
 
-    std::pair<u32, u32> GetBindlessCBuf() {
+    std::pair<u32, u32> GetBindlessCBuf() const {
         return {offset >> 32, offset & 0x00000000FFFFFFFFULL};
     }
 

From 90d06acfedb4c58b7b62153059c97b05035fc979 Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Tue, 26 Mar 2019 18:46:46 -0400
Subject: [PATCH 05/14] Fixes to Const Buffer Accessor and Formatting

---
 src/video_core/const_buffer_accessor.cpp         | 12 ++++++++----
 src/video_core/const_buffer_accessor.h           |  5 +----
 src/video_core/renderer_opengl/gl_rasterizer.cpp |  3 +--
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/src/video_core/const_buffer_accessor.cpp b/src/video_core/const_buffer_accessor.cpp
index c89ab91c7..0e613e1f9 100644
--- a/src/video_core/const_buffer_accessor.cpp
+++ b/src/video_core/const_buffer_accessor.cpp
@@ -14,21 +14,25 @@ namespace Tegra {
 
 u32 ConstBufferAccessor::access32(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
                                   u64 const_buffer, u64 offset) {
-    auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
+    auto& gpu = Core::System::GetInstance().GPU();
+    auto& memory_manager = gpu.MemoryManager();
+    auto& maxwell3d = gpu.Maxwell3D();
     const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<std::size_t>(stage)];
     const auto& buffer = shader_stage.const_buffers[const_buffer];
     u32 result;
-    std::memcpy(&result, memory_manager.GetPointer(buffer.address + offset * 4), sizeof(u32));
+    std::memcpy(&result, memory_manager.GetPointer(buffer.address + offset), sizeof(u32));
     return result;
 }
 
 u64 ConstBufferAccessor::access64(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
                                   u64 const_buffer, u64 offset) {
-    auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
+    auto& gpu = Core::System::GetInstance().GPU();
+    auto& memory_manager = gpu.MemoryManager();
+    auto& maxwell3d = gpu.Maxwell3D();
     const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<std::size_t>(stage)];
     const auto& buffer = shader_stage.const_buffers[const_buffer];
     u64 result;
-    std::memcpy(&result, memory_manager.GetPointer(buffer.address + offset * 4), sizeof(u64));
+    std::memcpy(&result, memory_manager.GetPointer(buffer.address + offset), sizeof(u64));
     return result;
 }
 
diff --git a/src/video_core/const_buffer_accessor.h b/src/video_core/const_buffer_accessor.h
index 2410f1483..37d1ca767 100644
--- a/src/video_core/const_buffer_accessor.h
+++ b/src/video_core/const_buffer_accessor.h
@@ -7,15 +7,12 @@ namespace Tegra {
 
 class ConstBufferAccessor {
 public:
-    ConstBufferAccessor(Tegra::Engines::Maxwell3D& maxwell3d) : maxwell3d(maxwell3d) {}
+    ConstBufferAccessor() {}
     ~ConstBufferAccessor() = default;
 
     u32 access32(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, u64 const_buffer, u64 offset);
 
     u64 access64(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, u64 const_buffer, u64 offset);
-
-private:
-    Tegra::Engines::Maxwell3D& maxwell3d;
 };
 
 } // namespace Tegra
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index a20acfe8e..e847b75fb 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -101,8 +101,7 @@ struct FramebufferCacheKey {
 RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, Core::System& system,
                                    ScreenInfo& info)
     : res_cache{*this}, shader_cache{*this, system}, global_cache{*this}, system{system},
-      screen_info{info}, buffer_cache(*this, STREAM_BUFFER_SIZE),
-      const_buffer_accessor(system.GPU().Maxwell3D()) {
+      screen_info{info}, buffer_cache(*this, STREAM_BUFFER_SIZE), const_buffer_accessor() {
     // Create sampler objects
     for (std::size_t i = 0; i < texture_samplers.size(); ++i) {
         texture_samplers[i].Create();

From ac3ba9a33e0d1e14061fc0f341b69cb85ea2e6a6 Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Tue, 26 Mar 2019 19:02:24 -0400
Subject: [PATCH 06/14] Corrections to TEX_B

---
 src/video_core/engines/shader_bytecode.h | 32 ++++++++++++++++++++++++
 src/video_core/shader/decode/texture.cpp |  9 ++++---
 2 files changed, 37 insertions(+), 4 deletions(-)

diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 2edd3245e..71c22aff0 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -966,6 +966,38 @@ union Instruction {
         }
     } tex;
 
+    union {
+        BitField<28, 1, u64> array;
+        BitField<29, 2, TextureType> texture_type;
+        BitField<31, 4, u64> component_mask;
+        BitField<49, 1, u64> nodep_flag;
+        BitField<50, 1, u64> dc_flag;
+        BitField<36, 1, u64> aoffi_flag;
+        BitField<37, 3, TextureProcessMode> process_mode;
+
+        bool IsComponentEnabled(std::size_t component) const {
+            return ((1ull << component) & component_mask) != 0;
+        }
+
+        TextureProcessMode GetTextureProcessMode() const {
+            return process_mode;
+        }
+
+        bool UsesMiscMode(TextureMiscMode mode) const {
+            switch (mode) {
+            case TextureMiscMode::DC:
+                return dc_flag != 0;
+            case TextureMiscMode::NODEP:
+                return nodep_flag != 0;
+            case TextureMiscMode::AOFFI:
+                return aoffi_flag != 0;
+            default:
+                break;
+            }
+            return false;
+        }
+    } tex_b;
+
     union {
         BitField<22, 6, TextureQueryType> query_type;
         BitField<31, 4, u64> component_mask;
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index 3ac04f6b7..300f1abad 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -65,10 +65,10 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
             LOG_WARNING(HW_GPU, "TEX.NODEP implementation is incomplete");
         }
 
-        const TextureType texture_type{instr.tex.texture_type};
-        const bool is_array = instr.tex.array != 0;
-        const bool depth_compare = instr.tex.UsesMiscMode(TextureMiscMode::DC);
-        const auto process_mode = instr.tex.GetTextureProcessMode();
+        const TextureType texture_type{instr.tex_b.texture_type};
+        const bool is_array = instr.tex_b.array != 0;
+        const bool depth_compare = instr.tex_b.UsesMiscMode(TextureMiscMode::DC);
+        const auto process_mode = instr.tex_b.GetTextureProcessMode();
         WriteTexInstructionFloat(bb, instr,
                                  GetTexCode(instr, texture_type, process_mode, depth_compare,
                                             is_array, true, instr.gpr20));
@@ -462,6 +462,7 @@ Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
     if (is_aoffi) {
         aoffi = GetAoffiCoordinates(GetRegister(parameter_register++), coord_count, false);
     }
+    const u32 bindless_offset = (is_bindless ? 1 : 0);
 
     Node dc{};
     if (depth_compare) {

From 189bd1980cc405d512eb8b19307df76976344e6e Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Tue, 26 Mar 2019 19:13:53 -0400
Subject: [PATCH 07/14] Implement TMML_B

---
 src/video_core/shader/decode/texture.cpp | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index 300f1abad..ddb7755b8 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -40,7 +40,7 @@ static std::size_t GetCoordCount(TextureType texture_type) {
 u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
-
+    bool is_bindless = false;
     switch (opcode->get().GetId()) {
     case OpCode::Id::TEX: {
         if (instr.tex.UsesMiscMode(TextureMiscMode::NODEP)) {
@@ -185,6 +185,8 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
         }
         break;
     }
+    case OpCode::Id::TMML_B:
+        is_bindless = true;
     case OpCode::Id::TMML: {
         UNIMPLEMENTED_IF_MSG(instr.tmml.UsesMiscMode(Tegra::Shader::TextureMiscMode::NDV),
                              "NDV is not implemented");
@@ -195,7 +197,9 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
 
         auto texture_type = instr.tmml.texture_type.Value();
         const bool is_array = instr.tmml.array != 0;
-        const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
+        const auto& sampler = !is_bindless
+                                  ? GetSampler(instr.sampler, texture_type, is_array, false)
+                                  : GetBindlessSampler(instr.gpr20, texture_type, is_array, false);
 
         std::vector<Node> coords;
 
@@ -271,11 +275,12 @@ const Sampler& ShaderIR::GetSampler(const Tegra::Shader::Sampler& sampler, Textu
     return *used_samplers.emplace(entry).first;
 }
 
-const Sampler& ShaderIR::GetBindlessSampler(const Tegra::Shader::Register& reg,
-                                            TextureType type, bool is_array, bool is_shadow) {
+const Sampler& ShaderIR::GetBindlessSampler(const Tegra::Shader::Register& reg, TextureType type,
+                                            bool is_array, bool is_shadow) {
 
     const Node sampler_register = GetRegister(reg);
-    const Node base_sampler = TrackCbuf(sampler_register, global_code, static_cast<s64>(global_code.size()));
+    const Node base_sampler =
+        TrackCbuf(sampler_register, global_code, static_cast<s64>(global_code.size()));
     const auto cbuf = std::get_if<CbufNode>(base_sampler);
     const auto cbuf_offset_imm = std::get_if<ImmediateNode>(cbuf->GetOffset());
     ASSERT(cbuf_offset_imm != nullptr);

From 4841440382c72047d68bb2c0ce7a7defadab7d3d Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Tue, 26 Mar 2019 19:46:11 -0400
Subject: [PATCH 08/14] Implement TXQ_B

---
 src/video_core/engines/shader_bytecode.h |  2 ++
 src/video_core/shader/decode/texture.cpp | 10 ++++++++--
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 71c22aff0..f7ef9a32a 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -1343,6 +1343,7 @@ public:
         TEX,
         TEX_B,  // Texture Load Bindless
         TXQ,    // Texture Query
+        TXQ_B,  // Texture Query Bindless
         TEXS,   // Texture Fetch with scalar/non-vec4 source/destinations
         TLDS,   // Texture Load with scalar/non-vec4 source/destinations
         TLD4,   // Texture Load 4
@@ -1612,6 +1613,7 @@ private:
             INST("110000----111---", Id::TEX, Type::Texture, "TEX"),
             INST("1101111010111---", Id::TEX_B, Type::Texture, "TEX_B"),
             INST("1101111101001---", Id::TXQ, Type::Texture, "TXQ"),
+            INST("1101111101010---", Id::TXQ_B, Type::Texture, "TXQ_B"),
             INST("1101-00---------", Id::TEXS, Type::Texture, "TEXS"),
             INST("1101101---------", Id::TLDS, Type::Texture, "TLDS"),
             INST("110010----111---", Id::TLD4, Type::Texture, "TLD4"),
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index ddb7755b8..3eac75bef 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -151,6 +151,8 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
         WriteTexsInstructionFloat(bb, instr, values);
         break;
     }
+    case OpCode::Id::TXQ_B:
+        is_bindless = true;
     case OpCode::Id::TXQ: {
         if (instr.txq.UsesMiscMode(TextureMiscMode::NODEP)) {
             LOG_WARNING(HW_GPU, "TXQ.NODEP implementation is incomplete");
@@ -160,7 +162,10 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
         // Sadly, not all texture instructions specify the type of texture their sampler
         // uses. This must be fixed at a later instance.
         const auto& sampler =
-            GetSampler(instr.sampler, Tegra::Shader::TextureType::Texture2D, false, false);
+            !is_bindless
+                ? GetSampler(instr.sampler, Tegra::Shader::TextureType::Texture2D, false, false)
+                : GetBindlessSampler(instr.gpr8, Tegra::Shader::TextureType::Texture2D, false,
+                                     false);
 
         u32 indexer = 0;
         switch (instr.txq.query_type) {
@@ -171,7 +176,8 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
                 }
                 MetaTexture meta{sampler, {}, {}, {}, {}, {}, {}, element};
                 const Node value =
-                    Operation(OperationCode::TextureQueryDimensions, meta, GetRegister(instr.gpr8));
+                    Operation(OperationCode::TextureQueryDimensions, meta,
+                              GetRegister(instr.gpr8.Value() + (is_bindless ? 1 : 0)));
                 SetTemporal(bb, indexer++, value);
             }
             for (u32 i = 0; i < indexer; ++i) {

From fd4e994de3196dfdd2a3f2caf4ca8934e719c296 Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Wed, 27 Mar 2019 07:11:50 -0400
Subject: [PATCH 09/14] Refactor GetTextureCode and GetTexCode to use an
 optional instead of optional parameters

---
 src/video_core/shader/decode/texture.cpp | 47 ++++++++++++------------
 src/video_core/shader/shader_ir.h        | 20 +++++-----
 2 files changed, 33 insertions(+), 34 deletions(-)

diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index 3eac75bef..5d670b24e 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -54,7 +54,7 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
         const auto process_mode = instr.tex.GetTextureProcessMode();
         WriteTexInstructionFloat(
             bb, instr,
-            GetTexCode(instr, texture_type, process_mode, depth_compare, is_array, is_aoffi));
+            GetTexCode(instr, texture_type, process_mode, depth_compare, is_array, is_aoffi, {}));
         break;
     }
     case OpCode::Id::TEX_B: {
@@ -69,9 +69,9 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
         const bool is_array = instr.tex_b.array != 0;
         const bool depth_compare = instr.tex_b.UsesMiscMode(TextureMiscMode::DC);
         const auto process_mode = instr.tex_b.GetTextureProcessMode();
-        WriteTexInstructionFloat(bb, instr,
-                                 GetTexCode(instr, texture_type, process_mode, depth_compare,
-                                            is_array, true, instr.gpr20));
+        WriteTexInstructionFloat(
+            bb, instr,
+            GetTexCode(instr, texture_type, process_mode, depth_compare, is_array, {instr.gpr20}));
         break;
     }
     case OpCode::Id::TEXS: {
@@ -162,10 +162,10 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
         // Sadly, not all texture instructions specify the type of texture their sampler
         // uses. This must be fixed at a later instance.
         const auto& sampler =
-            !is_bindless
-                ? GetSampler(instr.sampler, Tegra::Shader::TextureType::Texture2D, false, false)
-                : GetBindlessSampler(instr.gpr8, Tegra::Shader::TextureType::Texture2D, false,
-                                     false);
+            is_bindless
+                ? GetBindlessSampler(instr.gpr8, Tegra::Shader::TextureType::Texture2D, false,
+                                     false)
+                : GetSampler(instr.sampler, Tegra::Shader::TextureType::Texture2D, false, false);
 
         u32 indexer = 0;
         switch (instr.txq.query_type) {
@@ -203,9 +203,9 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
 
         auto texture_type = instr.tmml.texture_type.Value();
         const bool is_array = instr.tmml.array != 0;
-        const auto& sampler = !is_bindless
-                                  ? GetSampler(instr.sampler, texture_type, is_array, false)
-                                  : GetBindlessSampler(instr.gpr20, texture_type, is_array, false);
+        const auto& sampler = is_bindless
+                                  ? GetBindlessSampler(instr.gpr20, texture_type, is_array, false)
+                                  : GetSampler(instr.sampler, texture_type, is_array, false);
 
         std::vector<Node> coords;
 
@@ -381,25 +381,26 @@ void ShaderIR::WriteTexsInstructionHalfFloat(NodeBlock& bb, Instruction instr,
 
 Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
                                TextureProcessMode process_mode, std::vector<Node> coords,
-                               Node array, Node depth_compare, u32 bias_offset, std::vector<Node> aoffi, bool is_bindless,
-                               Register bindless_reg) {
+                               Node array, Node depth_compare, u32 bias_offset, std::vector<Node> aoffi, std::optional<Tegra::Shader::Register> bindless_reg) {
     const bool is_array = array;
     const bool is_shadow = depth_compare;
+    const bool is_bindless = bindless_reg.has_value();
 
     UNIMPLEMENTED_IF_MSG((texture_type == TextureType::Texture3D && (is_array || is_shadow)) ||
                              (texture_type == TextureType::TextureCube && is_array && is_shadow),
                          "This method is not supported.");
 
-    const auto& sampler = !is_bindless
-                              ? GetSampler(instr.sampler, texture_type, is_array, is_shadow)
-                              : GetBindlessSampler(bindless_reg, texture_type, is_array, is_shadow);
+    const auto& sampler = is_bindless
+                              ? GetBindlessSampler(*bindless_reg, texture_type, is_array, is_shadow)
+                              : GetSampler(instr.sampler, texture_type, is_array, is_shadow);
 
     const bool lod_needed = process_mode == TextureProcessMode::LZ ||
                             process_mode == TextureProcessMode::LL ||
                             process_mode == TextureProcessMode::LLA;
 
-    // LOD selection (either via bias or explicit textureLod) not supported in GL for
-    // sampler2DArrayShadow and samplerCubeArrayShadow.
+    // LOD selection (either via bias or explicit textureLod) not
+    // supported in GL for sampler2DArrayShadow and
+    // samplerCubeArrayShadow.
     const bool gl_lod_supported =
         !((texture_type == Tegra::Shader::TextureType::Texture2D && is_array && is_shadow) ||
           (texture_type == Tegra::Shader::TextureType::TextureCube && is_array && is_shadow));
@@ -417,8 +418,9 @@ Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
             lod = Immediate(0.0f);
             break;
         case TextureProcessMode::LB:
-            // If present, lod or bias are always stored in the register indexed by the gpr20
-            // field with an offset depending on the usage of the other registers
+            // If present, lod or bias are always stored in the register
+            // indexed by the gpr20 field with an offset depending on the
+            // usage of the other registers
             bias = GetRegister(instr.gpr20.Value() + bias_offset);
             break;
         case TextureProcessMode::LL:
@@ -442,7 +444,7 @@ Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
 
 Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
                            TextureProcessMode process_mode, bool depth_compare, bool is_array,
-                           bool is_aoffi, bool is_bindless, Register bindless_reg) {
+                           bool is_aoffi, std::optional<Tegra::Shader::Register> bindless_reg) {
     const bool lod_bias_enabled{
         (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ)};
 
@@ -482,8 +484,7 @@ Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
         dc = GetRegister(parameter_register++);
     }
 
-    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0, aoffi, is_bindless,
-                          bindless_reg);
+    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0, aoffi, bindless_reg);
 }
 
 Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index ed321cfe5..11495799f 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -769,11 +769,10 @@ private:
     void WriteTexsInstructionHalfFloat(NodeBlock& bb, Tegra::Shader::Instruction instr,
                                        const Node4& components);
 
-    Node4 GetTexCode(
-        Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
-        Tegra::Shader::TextureProcessMode process_mode, bool depth_compare, bool is_array,
-        bool is_aoffi, bool is_bindless = false,
-        Tegra::Shader::Register bindless_reg = static_cast<Tegra::Shader::Register>(0));
+    Node4 GetTexCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
+                     Tegra::Shader::TextureProcessMode process_mode, bool depth_compare,
+                     bool is_array, bool is_aoffi,
+                     std::optional<Tegra::Shader::Register> bindless_reg);
 
     Node4 GetTexsCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
                       Tegra::Shader::TextureProcessMode process_mode, bool depth_compare,
@@ -790,12 +789,11 @@ private:
         bool lod_bias_enabled, std::size_t max_coords, std::size_t max_inputs);
 
     std::vector<Node> GetAoffiCoordinates(Node aoffi_reg, std::size_t coord_count, bool is_tld4);
-    
-    Node4 GetTextureCode(
-        Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
-        Tegra::Shader::TextureProcessMode process_mode, std::vector<Node> coords, Node array,
-        Node depth_compare, u32 bias_offset, std::vector<Node> aoffi, bool is_bindless = false,
-        Tegra::Shader::Register bindless_reg = static_cast<Tegra::Shader::Register>(0));
+
+    Node4 GetTextureCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
+                         Tegra::Shader::TextureProcessMode process_mode, std::vector<Node> coords,
+                         Node array, Node depth_compare, u32 bias_offset, std::vector<Node> aoffi,
+                         std::optional<Tegra::Shader::Register> bindless_reg);
 
     Node GetVideoOperand(Node op, bool is_chunk, bool is_signed, Tegra::Shader::VideoType type,
                          u64 byte_height);

From a77e9a27b0e82192cd17eea255e6b04893ccafa7 Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Thu, 28 Mar 2019 14:54:52 -0400
Subject: [PATCH 10/14] Simplify ConstBufferAccessor

---
 src/video_core/CMakeLists.txt                 |  1 -
 src/video_core/const_buffer_accessor.cpp      | 39 -------------------
 src/video_core/const_buffer_accessor.h        | 26 +++++++++----
 .../renderer_opengl/gl_rasterizer.cpp         |  6 ++-
 .../renderer_opengl/gl_rasterizer.h           |  3 --
 5 files changed, 22 insertions(+), 53 deletions(-)
 delete mode 100644 src/video_core/const_buffer_accessor.cpp

diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 804395d38..c58f51f18 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -1,5 +1,4 @@
 add_library(video_core STATIC
-    const_buffer_accessor.cpp
     const_buffer_accessor.h
     dma_pusher.cpp
     dma_pusher.h
diff --git a/src/video_core/const_buffer_accessor.cpp b/src/video_core/const_buffer_accessor.cpp
deleted file mode 100644
index 0e613e1f9..000000000
--- a/src/video_core/const_buffer_accessor.cpp
+++ /dev/null
@@ -1,39 +0,0 @@
-#pragma once
-
-#include <cstring>
-
-#include "common/common_types.h"
-#include "core/core.h"
-#include "core/memory.h"
-#include "video_core/const_buffer_accessor.h"
-#include "video_core/engines/maxwell_3d.h"
-#include "video_core/gpu.h"
-#include "video_core/memory_manager.h"
-
-namespace Tegra {
-
-u32 ConstBufferAccessor::access32(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
-                                  u64 const_buffer, u64 offset) {
-    auto& gpu = Core::System::GetInstance().GPU();
-    auto& memory_manager = gpu.MemoryManager();
-    auto& maxwell3d = gpu.Maxwell3D();
-    const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<std::size_t>(stage)];
-    const auto& buffer = shader_stage.const_buffers[const_buffer];
-    u32 result;
-    std::memcpy(&result, memory_manager.GetPointer(buffer.address + offset), sizeof(u32));
-    return result;
-}
-
-u64 ConstBufferAccessor::access64(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
-                                  u64 const_buffer, u64 offset) {
-    auto& gpu = Core::System::GetInstance().GPU();
-    auto& memory_manager = gpu.MemoryManager();
-    auto& maxwell3d = gpu.Maxwell3D();
-    const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<std::size_t>(stage)];
-    const auto& buffer = shader_stage.const_buffers[const_buffer];
-    u64 result;
-    std::memcpy(&result, memory_manager.GetPointer(buffer.address + offset), sizeof(u64));
-    return result;
-}
-
-} // namespace Tegra
diff --git a/src/video_core/const_buffer_accessor.h b/src/video_core/const_buffer_accessor.h
index 37d1ca767..01524673b 100644
--- a/src/video_core/const_buffer_accessor.h
+++ b/src/video_core/const_buffer_accessor.h
@@ -1,18 +1,28 @@
 #pragma once
 
+#include <cstring>
+
 #include "common/common_types.h"
+#include "core/core.h"
 #include "video_core/engines/maxwell_3d.h"
+#include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
 
 namespace Tegra {
 
-class ConstBufferAccessor {
-public:
-    ConstBufferAccessor() {}
-    ~ConstBufferAccessor() = default;
+namespace ConstBufferAccessor {
 
-    u32 access32(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, u64 const_buffer, u64 offset);
-
-    u64 access64(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, u64 const_buffer, u64 offset);
-};
+template <typename T>
+T access(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, u64 const_buffer, u64 offset) {
+    auto& gpu = Core::System::GetInstance().GPU();
+    auto& memory_manager = gpu.MemoryManager();
+    auto& maxwell3d = gpu.Maxwell3D();
+    const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<std::size_t>(stage)];
+    const auto& buffer = shader_stage.const_buffers[const_buffer];
+    T result;
+    std::memcpy(&result, memory_manager.GetPointer(buffer.address + offset), sizeof(T));
+    return result;
+}
 
+} // namespace ConstBufferAccessor
 } // namespace Tegra
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index e847b75fb..30cad484a 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -19,6 +19,7 @@
 #include "core/core.h"
 #include "core/hle/kernel/process.h"
 #include "core/settings.h"
+#include "video_core/const_buffer_accessor.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_shader_cache.h"
@@ -101,7 +102,7 @@ struct FramebufferCacheKey {
 RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, Core::System& system,
                                    ScreenInfo& info)
     : res_cache{*this}, shader_cache{*this, system}, global_cache{*this}, system{system},
-      screen_info{info}, buffer_cache(*this, STREAM_BUFFER_SIZE), const_buffer_accessor() {
+      screen_info{info}, buffer_cache(*this, STREAM_BUFFER_SIZE) {
     // Create sampler objects
     for (std::size_t i = 0; i < texture_samplers.size(); ++i) {
         texture_samplers[i].Create();
@@ -990,7 +991,8 @@ void RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, const Shader& s
         } else {
             const auto cbuf = entry.GetBindlessCBuf();
             Tegra::Texture::TextureHandle tex_handle;
-            tex_handle.raw = const_buffer_accessor.access32(stage, cbuf.first, cbuf.second);
+            tex_handle.raw =
+                Tegra::ConstBufferAccessor::access<u32>(stage, cbuf.first, cbuf.second);
             texture = maxwell3d.GetTextureInfo(tex_handle, entry.GetOffset());
         }
         const u32 current_bindpoint = base_bindings.sampler + bindpoint;
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index 886e9c2b5..54fbf48aa 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -17,7 +17,6 @@
 #include <glad/glad.h>
 
 #include "common/common_types.h"
-#include "video_core/const_buffer_accessor.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/rasterizer_cache.h"
 #include "video_core/rasterizer_interface.h"
@@ -230,8 +229,6 @@ private:
     PrimitiveAssembler primitive_assembler{buffer_cache};
     GLint uniform_buffer_alignment;
 
-    Tegra::ConstBufferAccessor const_buffer_accessor;
-
     std::size_t CalculateVertexArraysSize() const;
 
     std::size_t CalculateIndexBufferSize() const;

From c60b0b8432953fbbd7434578fd0858073b908392 Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Thu, 28 Mar 2019 19:45:19 -0400
Subject: [PATCH 11/14] Fix TMML

---
 src/video_core/shader/decode/texture.cpp | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index 5d670b24e..99385c46e 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -226,17 +226,19 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
             coords.push_back(GetRegister(instr.gpr8.Value() + 1));
             texture_type = TextureType::Texture2D;
         }
-
+        u32 indexer = 0;
         for (u32 element = 0; element < 2; ++element) {
+            if (!instr.tmml.IsComponentEnabled(element)) {
+                continue;
+            }
             auto params = coords;
             MetaTexture meta{sampler, {}, {}, {}, {}, {}, {}, element};
             const Node value = Operation(OperationCode::TextureQueryLod, meta, std::move(params));
-            SetTemporal(bb, element, value);
+            SetTemporal(bb, indexer++, value);
         }
-        for (u32 element = 0; element < 2; ++element) {
-            SetRegister(bb, instr.gpr0.Value() + element, GetTemporal(element));
+        for (u32 i = 0; i < indexer; ++i) {
+            SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i));
         }
-
         break;
     }
     case OpCode::Id::TLDS: {

From 797e351bf816f86f689ee2704eb95c63411d4002 Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Fri, 5 Apr 2019 19:48:16 -0400
Subject: [PATCH 12/14] Fix bad rebase

---
 src/video_core/renderer_opengl/gl_rasterizer.cpp | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 30cad484a..ed1e97a73 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -99,8 +99,7 @@ struct FramebufferCacheKey {
     }
 };
 
-RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, Core::System& system,
-                                   ScreenInfo& info)
+RasterizerOpenGL::RasterizerOpenGL(Core::System& system, ScreenInfo& info)
     : res_cache{*this}, shader_cache{*this, system}, global_cache{*this}, system{system},
       screen_info{info}, buffer_cache(*this, STREAM_BUFFER_SIZE) {
     // Create sampler objects

From 492040bd9ce40f86f9845699d68104d31d272155 Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Sun, 7 Apr 2019 08:30:26 -0400
Subject: [PATCH 13/14] Move ConstBufferAccessor to Maxwell3d, correct mistakes
 and clang format.

---
 src/video_core/CMakeLists.txt                 |  1 -
 src/video_core/const_buffer_accessor.h        | 28 -------------------
 src/video_core/engines/maxwell_3d.cpp         | 12 ++++++--
 src/video_core/engines/maxwell_3d.h           |  2 ++
 src/video_core/engines/shader_bytecode.h      |  2 +-
 .../renderer_opengl/gl_rasterizer.cpp         | 10 +++----
 .../renderer_opengl/gl_shader_disk_cache.cpp  |  7 +++--
 src/video_core/shader/decode/texture.cpp      |  3 +-
 src/video_core/shader/shader_ir.h             |  4 +--
 9 files changed, 25 insertions(+), 44 deletions(-)
 delete mode 100644 src/video_core/const_buffer_accessor.h

diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index c58f51f18..242a0d1cd 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -1,5 +1,4 @@
 add_library(video_core STATIC
-    const_buffer_accessor.h
     dma_pusher.cpp
     dma_pusher.h
     debug_utils/debug_utils.cpp
diff --git a/src/video_core/const_buffer_accessor.h b/src/video_core/const_buffer_accessor.h
deleted file mode 100644
index 01524673b..000000000
--- a/src/video_core/const_buffer_accessor.h
+++ /dev/null
@@ -1,28 +0,0 @@
-#pragma once
-
-#include <cstring>
-
-#include "common/common_types.h"
-#include "core/core.h"
-#include "video_core/engines/maxwell_3d.h"
-#include "video_core/gpu.h"
-#include "video_core/memory_manager.h"
-
-namespace Tegra {
-
-namespace ConstBufferAccessor {
-
-template <typename T>
-T access(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, u64 const_buffer, u64 offset) {
-    auto& gpu = Core::System::GetInstance().GPU();
-    auto& memory_manager = gpu.MemoryManager();
-    auto& maxwell3d = gpu.Maxwell3D();
-    const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<std::size_t>(stage)];
-    const auto& buffer = shader_stage.const_buffers[const_buffer];
-    T result;
-    std::memcpy(&result, memory_manager.GetPointer(buffer.address + offset), sizeof(T));
-    return result;
-}
-
-} // namespace ConstBufferAccessor
-} // namespace Tegra
diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp
index 079132135..b198793bc 100644
--- a/src/video_core/engines/maxwell_3d.cpp
+++ b/src/video_core/engines/maxwell_3d.cpp
@@ -502,8 +502,8 @@ Texture::FullTextureInfo Maxwell3D::GetTextureInfo(const Texture::TextureHandle
 
 Texture::FullTextureInfo Maxwell3D::GetStageTexture(Regs::ShaderStage stage,
                                                     std::size_t offset) const {
-    auto& shader = state.shader_stages[static_cast<std::size_t>(stage)];
-    auto& tex_info_buffer = shader.const_buffers[regs.tex_cb_index];
+    const auto& shader = state.shader_stages[static_cast<std::size_t>(stage)];
+    const auto& tex_info_buffer = shader.const_buffers[regs.tex_cb_index];
     ASSERT(tex_info_buffer.enabled && tex_info_buffer.address != 0);
 
     const GPUVAddr tex_info_address =
@@ -529,4 +529,12 @@ void Maxwell3D::ProcessClearBuffers() {
     rasterizer.Clear();
 }
 
+u32 Maxwell3D::AccessConstBuffer32(Regs::ShaderStage stage, u64 const_buffer, u64 offset) const {
+    const auto& shader_stage = state.shader_stages[static_cast<std::size_t>(stage)];
+    const auto& buffer = shader_stage.const_buffers[const_buffer];
+    u32 result;
+    std::memcpy(&result, memory_manager.GetPointer(buffer.address + offset), sizeof(u32));
+    return result;
+}
+
 } // namespace Tegra::Engines
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index fd2c35a01..cc2424d38 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -1141,6 +1141,8 @@ public:
     /// Returns the texture information for a specific texture in a specific shader stage.
     Texture::FullTextureInfo GetStageTexture(Regs::ShaderStage stage, std::size_t offset) const;
 
+    u32 AccessConstBuffer32(Regs::ShaderStage stage, u64 const_buffer, u64 offset) const;
+
     /// Memory for macro code - it's undetermined how big this is, however 1MB is much larger than
     /// we've seen used.
     using MacroMemory = std::array<u32, 0x40000>;
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index f7ef9a32a..a7ef5da9a 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -976,7 +976,7 @@ union Instruction {
         BitField<37, 3, TextureProcessMode> process_mode;
 
         bool IsComponentEnabled(std::size_t component) const {
-            return ((1ull << component) & component_mask) != 0;
+            return ((1ULL << component) & component_mask) != 0;
         }
 
         TextureProcessMode GetTextureProcessMode() const {
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index ed1e97a73..6f3bcccec 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -19,7 +19,6 @@
 #include "core/core.h"
 #include "core/hle/kernel/process.h"
 #include "core/settings.h"
-#include "video_core/const_buffer_accessor.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_shader_cache.h"
@@ -985,14 +984,13 @@ void RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, const Shader& s
     for (u32 bindpoint = 0; bindpoint < entries.size(); ++bindpoint) {
         const auto& entry = entries[bindpoint];
         Tegra::Texture::FullTextureInfo texture;
-        if (!entry.IsBindless()) {
-            texture = maxwell3d.GetStageTexture(stage, entry.GetOffset());
-        } else {
+        if (entry.IsBindless()) {
             const auto cbuf = entry.GetBindlessCBuf();
             Tegra::Texture::TextureHandle tex_handle;
-            tex_handle.raw =
-                Tegra::ConstBufferAccessor::access<u32>(stage, cbuf.first, cbuf.second);
+            tex_handle.raw = maxwell3d.AccessConstBuffer32(stage, cbuf.first, cbuf.second);
             texture = maxwell3d.GetTextureInfo(tex_handle, entry.GetOffset());
+        } else {
+            texture = maxwell3d.GetStageTexture(stage, entry.GetOffset());
         }
         const u32 current_bindpoint = base_bindings.sampler + bindpoint;
 
diff --git a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
index e27740383..08603b7a5 100644
--- a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
@@ -328,9 +328,10 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
             file.ReadBytes(&is_bindless, sizeof(u8)) != sizeof(u8)) {
             return {};
         }
-        entry.entries.samplers.emplace_back(
-            static_cast<std::size_t>(offset), static_cast<std::size_t>(index),
-            static_cast<Tegra::Shader::TextureType>(type), is_array != 0, is_shadow != 0, is_bindless != 0);
+        entry.entries.samplers.emplace_back(static_cast<std::size_t>(offset),
+                                            static_cast<std::size_t>(index),
+                                            static_cast<Tegra::Shader::TextureType>(type),
+                                            is_array != 0, is_shadow != 0, is_bindless != 0);
     }
 
     u32 global_memory_count{};
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index 99385c46e..dd5310c36 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -153,6 +153,7 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
     }
     case OpCode::Id::TXQ_B:
         is_bindless = true;
+        [[fallthrough]];
     case OpCode::Id::TXQ: {
         if (instr.txq.UsesMiscMode(TextureMiscMode::NODEP)) {
             LOG_WARNING(HW_GPU, "TXQ.NODEP implementation is incomplete");
@@ -193,6 +194,7 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
     }
     case OpCode::Id::TMML_B:
         is_bindless = true;
+        [[fallthrough]];
     case OpCode::Id::TMML: {
         UNIMPLEMENTED_IF_MSG(instr.tmml.UsesMiscMode(Tegra::Shader::TextureMiscMode::NDV),
                              "NDV is not implemented");
@@ -285,7 +287,6 @@ const Sampler& ShaderIR::GetSampler(const Tegra::Shader::Sampler& sampler, Textu
 
 const Sampler& ShaderIR::GetBindlessSampler(const Tegra::Shader::Register& reg, TextureType type,
                                             bool is_array, bool is_shadow) {
-
     const Node sampler_register = GetRegister(reg);
     const Node base_sampler =
         TrackCbuf(sampler_register, global_code, static_cast<s64>(global_code.size()));
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index 11495799f..249024167 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -196,7 +196,7 @@ enum class ExitMethod {
 
 class Sampler {
 public:
-    // Use this constructor for binded Samplers
+    // Use this constructor for bounded Samplers
     explicit Sampler(std::size_t offset, std::size_t index, Tegra::Shader::TextureType type,
                      bool is_array, bool is_shadow)
         : offset{offset}, index{index}, type{type}, is_array{is_array}, is_shadow{is_shadow},
@@ -239,7 +239,7 @@ public:
     }
 
     std::pair<u32, u32> GetBindlessCBuf() const {
-        return {offset >> 32, offset & 0x00000000FFFFFFFFULL};
+        return {static_cast<u32>(offset >> 32), static_cast<u32>(offset)};
     }
 
     bool operator<(const Sampler& rhs) const {

From ef8be408d321f4f15f0731c46118834bb757be1a Mon Sep 17 00:00:00 2001
From: Fernando Sahmkow <fsahmkow27@gmail.com>
Date: Mon, 8 Apr 2019 12:07:56 -0400
Subject: [PATCH 14/14] Adapt Bindless to work with AOFFI

---
 src/video_core/shader/decode/texture.cpp | 25 +++++++++++++++++-------
 1 file changed, 18 insertions(+), 7 deletions(-)

diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index dd5310c36..fa65ac9a9 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -67,11 +67,12 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) {
 
         const TextureType texture_type{instr.tex_b.texture_type};
         const bool is_array = instr.tex_b.array != 0;
+        const bool is_aoffi = instr.tex.UsesMiscMode(TextureMiscMode::AOFFI);
         const bool depth_compare = instr.tex_b.UsesMiscMode(TextureMiscMode::DC);
         const auto process_mode = instr.tex_b.GetTextureProcessMode();
-        WriteTexInstructionFloat(
-            bb, instr,
-            GetTexCode(instr, texture_type, process_mode, depth_compare, is_array, {instr.gpr20}));
+        WriteTexInstructionFloat(bb, instr,
+                                 GetTexCode(instr, texture_type, process_mode, depth_compare,
+                                            is_array, is_aoffi, {instr.gpr20}));
         break;
     }
     case OpCode::Id::TEXS: {
@@ -384,7 +385,9 @@ void ShaderIR::WriteTexsInstructionHalfFloat(NodeBlock& bb, Instruction instr,
 
 Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
                                TextureProcessMode process_mode, std::vector<Node> coords,
-                               Node array, Node depth_compare, u32 bias_offset, std::vector<Node> aoffi, std::optional<Tegra::Shader::Register> bindless_reg) {
+                               Node array, Node depth_compare, u32 bias_offset,
+                               std::vector<Node> aoffi,
+                               std::optional<Tegra::Shader::Register> bindless_reg) {
     const bool is_array = array;
     const bool is_shadow = depth_compare;
     const bool is_bindless = bindless_reg.has_value();
@@ -451,7 +454,14 @@ Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
     const bool lod_bias_enabled{
         (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ)};
 
+    const bool is_bindless = bindless_reg.has_value();
+
     u64 parameter_register = instr.gpr20.Value();
+    if (is_bindless) {
+        ++parameter_register;
+    }
+
+    const u32 bias_lod_offset = (is_bindless ? 1 : 0);
     if (lod_bias_enabled) {
         ++parameter_register;
     }
@@ -478,7 +488,6 @@ Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
     if (is_aoffi) {
         aoffi = GetAoffiCoordinates(GetRegister(parameter_register++), coord_count, false);
     }
-    const u32 bindless_offset = (is_bindless ? 1 : 0);
 
     Node dc{};
     if (depth_compare) {
@@ -487,7 +496,8 @@ Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
         dc = GetRegister(parameter_register++);
     }
 
-    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0, aoffi, bindless_reg);
+    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, bias_lod_offset,
+                          aoffi, bindless_reg);
 }
 
 Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
@@ -523,7 +533,8 @@ Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
         dc = GetRegister(depth_register);
     }
 
-    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, bias_offset, {});
+    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, bias_offset, {},
+                          {});
 }
 
 Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool depth_compare,