10 files changed, 82 insertions, 43 deletions

diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 10114909b..9431d64ac 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -1006,7 +1006,8 @@ TextureBufferUsage RasterizerOpenGL::SetupComputeTextures(const Shader& kernel)
             }
             const auto cbuf = entry.GetBindlessCBuf();
             Tegra::Texture::TextureHandle tex_handle;
-            tex_handle.raw = compute.AccessConstBuffer32(Tegra::Engines::ShaderType::Compute, cbuf.first, cbuf.second);
+            tex_handle.raw = compute.AccessConstBuffer32(Tegra::Engines::ShaderType::Compute,
+                                                         cbuf.first, cbuf.second);
             return compute.GetTextureInfo(tex_handle, entry.GetOffset());
         }();
 
@@ -1051,7 +1052,8 @@ void RasterizerOpenGL::SetupComputeImages(const Shader& shader) {
             }
             const auto cbuf = entry.GetBindlessCBuf();
             Tegra::Texture::TextureHandle tex_handle;
-            tex_handle.raw = compute.AccessConstBuffer32(Tegra::Engines::ShaderType::Compute, cbuf.first, cbuf.second);
+            tex_handle.raw = compute.AccessConstBuffer32(Tegra::Engines::ShaderType::Compute,
+                                                         cbuf.first, cbuf.second);
             return compute.GetTextureInfo(tex_handle, entry.GetOffset()).tic;
         }();
         SetupImage(bindpoint, tic, entry);
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 42ca3b1bd..9e2799876 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -10,6 +10,7 @@
 #include "common/scope_exit.h"
 #include "core/core.h"
 #include "core/frontend/emu_window.h"
+#include "video_core/engines/kepler_compute.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/memory_manager.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
@@ -173,8 +174,9 @@ u64 GetUniqueIdentifier(ProgramType program_type, const ProgramCode& code,
 }
 
 /// Creates an unspecialized program from code streams
-GLShader::ProgramResult CreateProgram(const Device& device, ProgramType program_type,
-                                      ProgramCode program_code, ProgramCode program_code_b) {
+GLShader::ProgramResult CreateProgram(Core::System& system, const Device& device,
+                                      ProgramType program_type, ProgramCode program_code,
+                                      ProgramCode program_code_b) {
     GLShader::ShaderSetup setup(program_code);
     setup.program.size_a = CalculateProgramSize(program_code);
     setup.program.size_b = 0;
@@ -190,14 +192,25 @@ GLShader::ProgramResult CreateProgram(const Device& device, ProgramType program_
 
     switch (program_type) {
     case ProgramType::VertexA:
-    case ProgramType::VertexB:
-        return GLShader::GenerateVertexShader(device, setup);
-    case ProgramType::Geometry:
-        return GLShader::GenerateGeometryShader(device, setup);
-    case ProgramType::Fragment:
-        return GLShader::GenerateFragmentShader(device, setup);
-    case ProgramType::Compute:
-        return GLShader::GenerateComputeShader(device, setup);
+    case ProgramType::VertexB: {
+        VideoCommon::Shader::ConstBufferLocker locker{Tegra::Engines::ShaderType::Vertex,
+                                                      &(system.GPU().Maxwell3D())};
+        return GLShader::GenerateVertexShader(locker, device, setup);
+    }
+    case ProgramType::Geometry: {
+        VideoCommon::Shader::ConstBufferLocker locker{Tegra::Engines::ShaderType::Geometry,
+                                                      &(system.GPU().Maxwell3D())};
+        return GLShader::GenerateGeometryShader(locker, device, setup);
+    }
+    case ProgramType::Fragment: {
+        VideoCommon::Shader::ConstBufferLocker locker{Tegra::Engines::ShaderType::Fragment,
+                                                      &(system.GPU().Maxwell3D())};
+        return GLShader::GenerateFragmentShader(locker, device, setup);
+    }
+    case ProgramType::Compute: {
+        VideoCommon::Shader::ConstBufferLocker locker{Tegra::Engines::ShaderType::Compute, &(system.GPU().KeplerCompute())};
+        return GLShader::GenerateComputeShader(locker, device, setup);
+    }
     default:
         UNIMPLEMENTED_MSG("Unimplemented program_type={}", static_cast<u32>(program_type));
         return {};
@@ -307,8 +320,8 @@ Shader CachedShader::CreateStageFromMemory(const ShaderParameters& params,
                                            ProgramCode&& program_code_b) {
     const auto code_size{CalculateProgramSize(program_code)};
     const auto code_size_b{CalculateProgramSize(program_code_b)};
-    auto result{
-        CreateProgram(params.device, GetProgramType(program_type), program_code, program_code_b)};
+    auto result{CreateProgram(params.system, params.device, GetProgramType(program_type),
+                              program_code, program_code_b)};
     if (result.first.empty()) {
         // TODO(Rodrigo): Unimplemented shader stages hit here, avoid using these for now
         return {};
@@ -331,7 +344,7 @@ Shader CachedShader::CreateStageFromCache(const ShaderParameters& params,
 }
 
 Shader CachedShader::CreateKernelFromMemory(const ShaderParameters& params, ProgramCode&& code) {
-    auto result{CreateProgram(params.device, ProgramType::Compute, code, {})};
+    auto result{CreateProgram(params.system, params.device, ProgramType::Compute, code, {})};
 
     const auto code_size{CalculateProgramSize(code)};
     params.disk_cache.SaveRaw(ShaderDiskCacheRaw(params.unique_identifier, ProgramType::Compute,
@@ -566,7 +579,7 @@ std::unordered_map<u64, UnspecializedShader> ShaderCacheOpenGL::GenerateUnspecia
             result = {stored_decompiled.code, stored_decompiled.entries};
         } else {
             // Otherwise decompile the shader at boot and save the result to the decompiled file
-            result = CreateProgram(device, raw.GetProgramType(), raw.GetProgramCode(),
+            result = CreateProgram(system, device, raw.GetProgramType(), raw.GetProgramCode(),
                                    raw.GetProgramCodeB());
             disk_cache.SaveDecompiled(unique_identifier, result.first, result.second);
         }
@@ -612,7 +625,7 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
     const auto unique_identifier =
         GetUniqueIdentifier(GetProgramType(program), program_code, program_code_b);
     const auto cpu_addr{*memory_manager.GpuToCpuAddress(program_addr)};
-    const ShaderParameters params{disk_cache, precompiled_programs, device, cpu_addr,
+    const ShaderParameters params{disk_cache, precompiled_programs, system, device, cpu_addr,
                                   host_ptr,   unique_identifier};
 
     const auto found = precompiled_shaders.find(unique_identifier);
@@ -639,7 +652,7 @@ Shader ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
     auto code{GetShaderCode(memory_manager, code_addr, host_ptr)};
     const auto unique_identifier{GetUniqueIdentifier(ProgramType::Compute, code, {})};
     const auto cpu_addr{*memory_manager.GpuToCpuAddress(code_addr)};
-    const ShaderParameters params{disk_cache, precompiled_programs, device, cpu_addr,
+    const ShaderParameters params{disk_cache, precompiled_programs, system, device, cpu_addr,
                                   host_ptr,   unique_identifier};
 
     const auto found = precompiled_shaders.find(unique_identifier);
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index de195cc5d..6ff78f005 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -45,6 +45,7 @@ using PrecompiledShaders = std::unordered_map<u64, GLShader::ProgramResult>;
 struct ShaderParameters {
     ShaderDiskCacheOpenGL& disk_cache;
     const PrecompiledPrograms& precompiled_programs;
+    Core::System& system;
     const Device& device;
     VAddr cpu_addr;
     u8* host_ptr;
diff --git a/src/video_core/renderer_opengl/gl_shader_gen.cpp b/src/video_core/renderer_opengl/gl_shader_gen.cpp
index b5a43e79e..817c6e12c 100644
--- a/src/video_core/renderer_opengl/gl_shader_gen.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_gen.cpp
@@ -21,7 +21,8 @@ static constexpr u32 COMPUTE_OFFSET = 0;
 
 static constexpr CompilerSettings settings{CompileDepth::NoFlowStack, true};
 
-ProgramResult GenerateVertexShader(const Device& device, const ShaderSetup& setup) {
+ProgramResult GenerateVertexShader(ConstBufferLocker& locker, const Device& device,
+                                   const ShaderSetup& setup) {
     const std::string id = fmt::format("{:016x}", setup.program.unique_identifier);
 
     std::string out = "// Shader Unique Id: VS" + id + "\n\n";
@@ -35,14 +36,15 @@ layout (std140, binding = EMULATION_UBO_BINDING) uniform vs_config {
 
 )";
 
-    const ShaderIR program_ir(setup.program.code, PROGRAM_OFFSET, setup.program.size_a, settings);
+    const ShaderIR program_ir(setup.program.code, PROGRAM_OFFSET, setup.program.size_a, settings,
+                              locker);
     const auto stage = setup.IsDualProgram() ? ProgramType::VertexA : ProgramType::VertexB;
     ProgramResult program = Decompile(device, program_ir, stage, "vertex");
     out += program.first;
 
     if (setup.IsDualProgram()) {
         const ShaderIR program_ir_b(setup.program.code_b, PROGRAM_OFFSET, setup.program.size_b,
-                                    settings);
+                                    settings, locker);
         ProgramResult program_b = Decompile(device, program_ir_b, ProgramType::VertexB, "vertex_b");
         out += program_b.first;
     }
@@ -71,7 +73,8 @@ void main() {
     return {std::move(out), std::move(program.second)};
 }
 
-ProgramResult GenerateGeometryShader(const Device& device, const ShaderSetup& setup) {
+ProgramResult GenerateGeometryShader(ConstBufferLocker& locker, const Device& device,
+                                     const ShaderSetup& setup) {
     const std::string id = fmt::format("{:016x}", setup.program.unique_identifier);
 
     std::string out = "// Shader Unique Id: GS" + id + "\n\n";
@@ -85,7 +88,8 @@ layout (std140, binding = EMULATION_UBO_BINDING) uniform gs_config {
 
 )";
 
-    const ShaderIR program_ir(setup.program.code, PROGRAM_OFFSET, setup.program.size_a, settings);
+    const ShaderIR program_ir(setup.program.code, PROGRAM_OFFSET, setup.program.size_a, settings,
+                              locker);
     ProgramResult program = Decompile(device, program_ir, ProgramType::Geometry, "geometry");
     out += program.first;
 
@@ -97,7 +101,8 @@ void main() {
     return {std::move(out), std::move(program.second)};
 }
 
-ProgramResult GenerateFragmentShader(const Device& device, const ShaderSetup& setup) {
+ProgramResult GenerateFragmentShader(ConstBufferLocker& locker, const Device& device,
+                                     const ShaderSetup& setup) {
     const std::string id = fmt::format("{:016x}", setup.program.unique_identifier);
 
     std::string out = "// Shader Unique Id: FS" + id + "\n\n";
@@ -120,7 +125,8 @@ layout (std140, binding = EMULATION_UBO_BINDING) uniform fs_config {
 
 )";
 
-    const ShaderIR program_ir(setup.program.code, PROGRAM_OFFSET, setup.program.size_a, settings);
+    const ShaderIR program_ir(setup.program.code, PROGRAM_OFFSET, setup.program.size_a, settings,
+                              locker);
     ProgramResult program = Decompile(device, program_ir, ProgramType::Fragment, "fragment");
     out += program.first;
 
@@ -133,13 +139,15 @@ void main() {
     return {std::move(out), std::move(program.second)};
 }
 
-ProgramResult GenerateComputeShader(const Device& device, const ShaderSetup& setup) {
+ProgramResult GenerateComputeShader(ConstBufferLocker& locker, const Device& device,
+                                    const ShaderSetup& setup) {
     const std::string id = fmt::format("{:016x}", setup.program.unique_identifier);
 
     std::string out = "// Shader Unique Id: CS" + id + "\n\n";
     out += GetCommonDeclarations();
 
-    const ShaderIR program_ir(setup.program.code, COMPUTE_OFFSET, setup.program.size_a, settings);
+    const ShaderIR program_ir(setup.program.code, COMPUTE_OFFSET, setup.program.size_a, settings,
+                              locker);
     ProgramResult program = Decompile(device, program_ir, ProgramType::Compute, "compute");
     out += program.first;
 
diff --git a/src/video_core/renderer_opengl/gl_shader_gen.h b/src/video_core/renderer_opengl/gl_shader_gen.h
index 3833e88ab..05f157298 100644
--- a/src/video_core/renderer_opengl/gl_shader_gen.h
+++ b/src/video_core/renderer_opengl/gl_shader_gen.h
@@ -16,6 +16,7 @@ class Device;
 
 namespace OpenGL::GLShader {
 
+using VideoCommon::Shader::ConstBufferLocker;
 using VideoCommon::Shader::ProgramCode;
 
 struct ShaderSetup {
@@ -46,15 +47,19 @@ private:
 };
 
 /// Generates the GLSL vertex shader program source code for the given VS program
-ProgramResult GenerateVertexShader(const Device& device, const ShaderSetup& setup);
+ProgramResult GenerateVertexShader(ConstBufferLocker& locker, const Device& device,
+                                   const ShaderSetup& setup);
 
 /// Generates the GLSL geometry shader program source code for the given GS program
-ProgramResult GenerateGeometryShader(const Device& device, const ShaderSetup& setup);
+ProgramResult GenerateGeometryShader(ConstBufferLocker& locker, const Device& device,
+                                     const ShaderSetup& setup);
 
 /// Generates the GLSL fragment shader program source code for the given FS program
-ProgramResult GenerateFragmentShader(const Device& device, const ShaderSetup& setup);
+ProgramResult GenerateFragmentShader(ConstBufferLocker& locker, const Device& device,
+                                     const ShaderSetup& setup);
 
 /// Generates the GLSL compute shader program source code for the given CS program
-ProgramResult GenerateComputeShader(const Device& device, const ShaderSetup& setup);
+ProgramResult GenerateComputeShader(ConstBufferLocker& locker, const Device& device,
+                                    const ShaderSetup& setup);
 
 } // namespace OpenGL::GLShader
diff --git a/src/video_core/shader/control_flow.cpp b/src/video_core/shader/control_flow.cpp
index 70f758642..dac2e4272 100644
--- a/src/video_core/shader/control_flow.cpp
+++ b/src/video_core/shader/control_flow.cpp
@@ -57,8 +57,8 @@ struct BlockInfo {
 
 struct CFGRebuildState {
     explicit CFGRebuildState(const ProgramCode& program_code, const std::size_t program_size,
-                             const u32 start)
-        : start{start}, program_code{program_code}, program_size{program_size} {}
+                             const u32 start, ConstBufferLocker& locker)
+        : start{start}, program_code{program_code}, program_size{program_size}, locker{locker} {}
 
     u32 start{};
     std::vector<BlockInfo> block_info{};
@@ -72,6 +72,7 @@ struct CFGRebuildState {
     const ProgramCode& program_code;
     const std::size_t program_size;
     ASTManager* manager;
+    ConstBufferLocker& locker;
 };
 
 enum class BlockCollision : u32 { None, Found, Inside };
@@ -214,7 +215,7 @@ std::optional<BranchIndirectInfo> TrackBranchIndirectInfo(const CFGRebuildState&
         if (opcode->get().GetId() == OpCode::Id::IMNMX_IMM) {
             if (instr.gpr0.Value() == track_register) {
                 track_register = instr.gpr8.Value();
-                result.entries = instr.alu.GetSignedImm20_20();
+                result.entries = instr.alu.GetSignedImm20_20() + 1;
                 pos--;
                 found_track = true;
                 break;
@@ -406,8 +407,14 @@ std::pair<ParseResult, ParseInfo> ParseCode(CFGRebuildState& state, u32 address)
             auto tmp = TrackBranchIndirectInfo(state, address, offset);
             if (tmp) {
                 auto result = *tmp;
-                LOG_CRITICAL(HW_GPU, "Track Successful, BRX: buffer:{}, offset:{}, entries:{}",
-                             result.buffer, result.offset, result.entries);
+                std::string entries{};
+                for (u32 i = 0; i < result.entries; i++) {
+                    auto k = locker.ObtainKey(result.buffer, result.offset + i * 4);
+                    entries = entries + std::to_string(*k) + '\n';
+                }
+                LOG_CRITICAL(HW_GPU,
+                             "Track Successful, BRX: buffer:{}, offset:{}, entries:{}, inner:\n{}",
+                             result.buffer, result.offset, result.entries, entries);
             } else {
                 LOG_CRITICAL(HW_GPU, "Track Unsuccesful");
             }
@@ -588,14 +595,15 @@ void DecompileShader(CFGRebuildState& state) {
 
 std::unique_ptr<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code,
                                                 std::size_t program_size, u32 start_address,
-                                                const CompilerSettings& settings) {
+                                                const CompilerSettings& settings,
+                                                ConstBufferLocker& locker) {
     auto result_out = std::make_unique<ShaderCharacteristics>();
     if (settings.depth == CompileDepth::BruteForce) {
         result_out->settings.depth = CompileDepth::BruteForce;
         return result_out;
     }
 
-    CFGRebuildState state{program_code, program_size, start_address};
+    CFGRebuildState state{program_code, program_size, start_address, locker};
     // Inspect Code and generate blocks
     state.labels.clear();
     state.labels.emplace(start_address);
diff --git a/src/video_core/shader/control_flow.h b/src/video_core/shader/control_flow.h
index 37e987d62..6d0e50d7c 100644
--- a/src/video_core/shader/control_flow.h
+++ b/src/video_core/shader/control_flow.h
@@ -78,6 +78,7 @@ struct ShaderCharacteristics {
 
 std::unique_ptr<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code,
                                                 std::size_t program_size, u32 start_address,
-                                                const CompilerSettings& settings);
+                                                const CompilerSettings& settings,
+                                                ConstBufferLocker& locker);
 
 } // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/decode.cpp b/src/video_core/shader/decode.cpp
index 2626b1616..3f87b87ca 100644
--- a/src/video_core/shader/decode.cpp
+++ b/src/video_core/shader/decode.cpp
@@ -102,7 +102,7 @@ void ShaderIR::Decode() {
     std::memcpy(&header, program_code.data(), sizeof(Tegra::Shader::Header));
 
     decompiled = false;
-    auto info = ScanFlow(program_code, program_size, main_offset, settings);
+    auto info = ScanFlow(program_code, program_size, main_offset, settings, locker);
     auto& shader_info = *info;
     coverage_begin = shader_info.start;
     coverage_end = shader_info.end;
diff --git a/src/video_core/shader/shader_ir.cpp b/src/video_core/shader/shader_ir.cpp
index c1f2b88c8..6430575ec 100644
--- a/src/video_core/shader/shader_ir.cpp
+++ b/src/video_core/shader/shader_ir.cpp
@@ -23,9 +23,9 @@ using Tegra::Shader::PredOperation;
 using Tegra::Shader::Register;
 
 ShaderIR::ShaderIR(const ProgramCode& program_code, u32 main_offset, const std::size_t size,
-                   CompilerSettings settings)
+                   CompilerSettings settings, ConstBufferLocker& locker)
     : program_code{program_code}, main_offset{main_offset}, program_size{size}, basic_blocks{},
-      program_manager{true, true}, settings{settings} {
+      program_manager{true, true}, settings{settings}, locker{locker} {
     Decode();
 }
 
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index 68818643c..e3b568d3e 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -68,7 +68,7 @@ struct GlobalMemoryUsage {
 class ShaderIR final {
 public:
     explicit ShaderIR(const ProgramCode& program_code, u32 main_offset, std::size_t size,
-                      CompilerSettings settings);
+                      CompilerSettings settings, ConstBufferLocker& locker);
     ~ShaderIR();
 
     const std::map<u32, NodeBlock>& GetBasicBlocks() const {
@@ -389,6 +389,7 @@ private:
     NodeBlock global_code;
     ASTManager program_manager;
     CompilerSettings settings{};
+    ConstBufferLocker& locker;
 
     std::set<u32> used_registers;
     std::set<Tegra::Shader::Pred> used_predicates;