13 files changed, 236 insertions, 166 deletions

diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.cpp b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
index ad0577a4f..e461e4c70 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
@@ -22,21 +22,53 @@ using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
 MICROPROFILE_DEFINE(OpenGL_Buffer_Download, "OpenGL", "Buffer Download", MP_RGB(192, 192, 128));
 
-Buffer::Buffer(VAddr cpu_addr, const std::size_t size) : VideoCommon::BufferBlock{cpu_addr, size} {
+Buffer::Buffer(const Device& device, VAddr cpu_addr, std::size_t size)
+    : VideoCommon::BufferBlock{cpu_addr, size} {
     gl_buffer.Create();
     glNamedBufferData(gl_buffer.handle, static_cast<GLsizeiptr>(size), nullptr, GL_DYNAMIC_DRAW);
+    if (device.HasVertexBufferUnifiedMemory()) {
+        glMakeNamedBufferResidentNV(gl_buffer.handle, GL_READ_WRITE);
+        glGetNamedBufferParameterui64vNV(gl_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, &gpu_address);
+    }
 }
 
 Buffer::~Buffer() = default;
 
+void Buffer::Upload(std::size_t offset, std::size_t size, const u8* data) {
+    glNamedBufferSubData(Handle(), static_cast<GLintptr>(offset), static_cast<GLsizeiptr>(size),
+                         data);
+}
+
+void Buffer::Download(std::size_t offset, std::size_t size, u8* data) {
+    MICROPROFILE_SCOPE(OpenGL_Buffer_Download);
+    const GLsizeiptr gl_size = static_cast<GLsizeiptr>(size);
+    const GLintptr gl_offset = static_cast<GLintptr>(offset);
+    if (read_buffer.handle == 0) {
+        read_buffer.Create();
+        glNamedBufferData(read_buffer.handle, static_cast<GLsizeiptr>(Size()), nullptr,
+                          GL_STREAM_READ);
+    }
+    glMemoryBarrier(GL_BUFFER_UPDATE_BARRIER_BIT);
+    glCopyNamedBufferSubData(gl_buffer.handle, read_buffer.handle, gl_offset, gl_offset, gl_size);
+    glGetNamedBufferSubData(read_buffer.handle, gl_offset, gl_size, data);
+}
+
+void Buffer::CopyFrom(const Buffer& src, std::size_t src_offset, std::size_t dst_offset,
+                      std::size_t size) {
+    glCopyNamedBufferSubData(src.Handle(), Handle(), static_cast<GLintptr>(src_offset),
+                             static_cast<GLintptr>(dst_offset), static_cast<GLsizeiptr>(size));
+}
+
 OGLBufferCache::OGLBufferCache(RasterizerOpenGL& rasterizer, Core::System& system,
-                               const Device& device, std::size_t stream_size)
-    : GenericBufferCache{rasterizer, system, std::make_unique<OGLStreamBuffer>(stream_size, true)} {
+                               const Device& device_, std::size_t stream_size)
+    : GenericBufferCache{rasterizer, system,
+                         std::make_unique<OGLStreamBuffer>(device_, stream_size, true)},
+      device{device_} {
     if (!device.HasFastBufferSubData()) {
         return;
     }
 
-    static constexpr auto size = static_cast<GLsizeiptr>(Maxwell::MaxConstBufferSize);
+    static constexpr GLsizeiptr size = static_cast<GLsizeiptr>(Maxwell::MaxConstBufferSize);
     glCreateBuffers(static_cast<GLsizei>(std::size(cbufs)), std::data(cbufs));
     for (const GLuint cbuf : cbufs) {
         glNamedBufferData(cbuf, size, nullptr, GL_STREAM_DRAW);
@@ -48,39 +80,20 @@ OGLBufferCache::~OGLBufferCache() {
 }
 
 std::shared_ptr<Buffer> OGLBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) {
-    return std::make_shared<Buffer>(cpu_addr, size);
+    return std::make_shared<Buffer>(device, cpu_addr, size);
 }
 
-GLuint OGLBufferCache::GetEmptyBuffer(std::size_t) {
-    return 0;
-}
-
-void OGLBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
-                                     const u8* data) {
-    glNamedBufferSubData(buffer.Handle(), static_cast<GLintptr>(offset),
-                         static_cast<GLsizeiptr>(size), data);
-}
-
-void OGLBufferCache::DownloadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
-                                       u8* data) {
-    MICROPROFILE_SCOPE(OpenGL_Buffer_Download);
-    glMemoryBarrier(GL_BUFFER_UPDATE_BARRIER_BIT);
-    glGetNamedBufferSubData(buffer.Handle(), static_cast<GLintptr>(offset),
-                            static_cast<GLsizeiptr>(size), data);
-}
-
-void OGLBufferCache::CopyBlock(const Buffer& src, const Buffer& dst, std::size_t src_offset,
-                               std::size_t dst_offset, std::size_t size) {
-    glCopyNamedBufferSubData(src.Handle(), dst.Handle(), static_cast<GLintptr>(src_offset),
-                             static_cast<GLintptr>(dst_offset), static_cast<GLsizeiptr>(size));
+OGLBufferCache::BufferInfo OGLBufferCache::GetEmptyBuffer(std::size_t) {
+    return {0, 0, 0};
 }
 
 OGLBufferCache::BufferInfo OGLBufferCache::ConstBufferUpload(const void* raw_pointer,
                                                              std::size_t size) {
     DEBUG_ASSERT(cbuf_cursor < std::size(cbufs));
     const GLuint cbuf = cbufs[cbuf_cursor++];
+
     glNamedBufferSubData(cbuf, 0, static_cast<GLsizeiptr>(size), raw_pointer);
-    return {cbuf, 0};
+    return {cbuf, 0, 0};
 }
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.h b/src/video_core/renderer_opengl/gl_buffer_cache.h
index a49aaf9c4..88fdc0536 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.h
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.h
@@ -25,15 +25,28 @@ class RasterizerOpenGL;
 
 class Buffer : public VideoCommon::BufferBlock {
 public:
-    explicit Buffer(VAddr cpu_addr, const std::size_t size);
+    explicit Buffer(const Device& device, VAddr cpu_addr, std::size_t size);
     ~Buffer();
 
-    GLuint Handle() const {
+    void Upload(std::size_t offset, std::size_t size, const u8* data);
+
+    void Download(std::size_t offset, std::size_t size, u8* data);
+
+    void CopyFrom(const Buffer& src, std::size_t src_offset, std::size_t dst_offset,
+                  std::size_t size);
+
+    GLuint Handle() const noexcept {
         return gl_buffer.handle;
     }
 
+    u64 Address() const noexcept {
+        return gpu_address;
+    }
+
 private:
     OGLBuffer gl_buffer;
+    OGLBuffer read_buffer;
+    u64 gpu_address = 0;
 };
 
 using GenericBufferCache = VideoCommon::BufferCache<Buffer, GLuint, OGLStreamBuffer>;
@@ -43,7 +56,7 @@ public:
                             const Device& device, std::size_t stream_size);
     ~OGLBufferCache();
 
-    GLuint GetEmptyBuffer(std::size_t) override;
+    BufferInfo GetEmptyBuffer(std::size_t) override;
 
     void Acquire() noexcept {
         cbuf_cursor = 0;
@@ -52,22 +65,16 @@ public:
 protected:
     std::shared_ptr<Buffer> CreateBlock(VAddr cpu_addr, std::size_t size) override;
 
-    void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
-                         const u8* data) override;
-
-    void DownloadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
-                           u8* data) override;
-
-    void CopyBlock(const Buffer& src, const Buffer& dst, std::size_t src_offset,
-                   std::size_t dst_offset, std::size_t size) override;
-
     BufferInfo ConstBufferUpload(const void* raw_pointer, std::size_t size) override;
 
 private:
+    static constexpr std::size_t NUM_CBUFS = Tegra::Engines::Maxwell3D::Regs::MaxConstBuffers *
+                                             Tegra::Engines::Maxwell3D::Regs::MaxShaderProgram;
+
+    const Device& device;
+
     std::size_t cbuf_cursor = 0;
-    std::array<GLuint, Tegra::Engines::Maxwell3D::Regs::MaxConstBuffers *
-                           Tegra::Engines::Maxwell3D::Regs::MaxShaderProgram>
-        cbufs;
+    std::array<GLuint, NUM_CBUFS> cbufs{};
 };
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp
index b31d604e4..208fc6167 100644
--- a/src/video_core/renderer_opengl/gl_device.cpp
+++ b/src/video_core/renderer_opengl/gl_device.cpp
@@ -178,7 +178,7 @@ bool IsASTCSupported() {
         for (const GLenum format : formats) {
             for (const GLenum support : required_support) {
                 GLint value;
-                glGetInternalformativ(GL_TEXTURE_2D, format, support, 1, &value);
+                glGetInternalformativ(target, format, support, 1, &value);
                 if (value != GL_FULL_SUPPORT) {
                     return false;
                 }
@@ -193,6 +193,7 @@ bool IsASTCSupported() {
 Device::Device()
     : max_uniform_buffers{BuildMaxUniformBuffers()}, base_bindings{BuildBaseBindings()} {
     const std::string_view vendor = reinterpret_cast<const char*>(glGetString(GL_VENDOR));
+    const std::string_view renderer = reinterpret_cast<const char*>(glGetString(GL_RENDERER));
     const std::string_view version = reinterpret_cast<const char*>(glGetString(GL_VERSION));
     const std::vector extensions = GetExtensions();
 
@@ -216,12 +217,18 @@ Device::Device()
     has_shader_ballot = GLAD_GL_ARB_shader_ballot;
     has_vertex_viewport_layer = GLAD_GL_ARB_shader_viewport_layer_array;
     has_image_load_formatted = HasExtension(extensions, "GL_EXT_shader_image_load_formatted");
+    has_texture_shadow_lod = HasExtension(extensions, "GL_EXT_texture_shadow_lod");
     has_astc = IsASTCSupported();
     has_variable_aoffi = TestVariableAoffi();
     has_component_indexing_bug = is_amd;
     has_precise_bug = TestPreciseBug();
-    has_fast_buffer_sub_data = is_nvidia && !disable_fast_buffer_sub_data;
     has_nv_viewport_array2 = GLAD_GL_NV_viewport_array2;
+    has_vertex_buffer_unified_memory = GLAD_GL_NV_vertex_buffer_unified_memory;
+
+    // At the moment of writing this, only Nvidia's driver optimizes BufferSubData on exclusive
+    // uniform buffers as "push constants"
+    has_fast_buffer_sub_data = is_nvidia && !disable_fast_buffer_sub_data;
+
     use_assembly_shaders = Settings::values.use_assembly_shaders && GLAD_GL_NV_gpu_program5 &&
                            GLAD_GL_NV_compute_program5 && GLAD_GL_NV_transform_feedback &&
                            GLAD_GL_NV_transform_feedback2;
@@ -245,6 +252,7 @@ Device::Device(std::nullptr_t) {
     has_shader_ballot = true;
     has_vertex_viewport_layer = true;
     has_image_load_formatted = true;
+    has_texture_shadow_lod = true;
     has_variable_aoffi = true;
 }
 
diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h
index 145347943..e1d811966 100644
--- a/src/video_core/renderer_opengl/gl_device.h
+++ b/src/video_core/renderer_opengl/gl_device.h
@@ -68,6 +68,14 @@ public:
         return has_image_load_formatted;
     }
 
+    bool HasTextureShadowLod() const {
+        return has_texture_shadow_lod;
+    }
+
+    bool HasVertexBufferUnifiedMemory() const {
+        return has_vertex_buffer_unified_memory;
+    }
+
     bool HasASTC() const {
         return has_astc;
     }
@@ -110,6 +118,8 @@ private:
     bool has_shader_ballot{};
     bool has_vertex_viewport_layer{};
     bool has_image_load_formatted{};
+    bool has_texture_shadow_lod{};
+    bool has_vertex_buffer_unified_memory{};
     bool has_astc{};
     bool has_variable_aoffi{};
     bool has_component_indexing_bug{};
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 2d6c11320..e960a0ef1 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -61,7 +61,8 @@ constexpr std::size_t NUM_CONST_BUFFERS_BYTES_PER_STAGE =
 constexpr std::size_t TOTAL_CONST_BUFFER_BYTES =
     NUM_CONST_BUFFERS_BYTES_PER_STAGE * Maxwell::MaxShaderStage;
 
-constexpr std::size_t NumSupportedVertexAttributes = 16;
+constexpr std::size_t NUM_SUPPORTED_VERTEX_ATTRIBUTES = 16;
+constexpr std::size_t NUM_SUPPORTED_VERTEX_BINDINGS = 16;
 
 template <typename Engine, typename Entry>
 Tegra::Texture::FullTextureInfo GetTextureInfo(const Engine& engine, const Entry& entry,
@@ -193,7 +194,7 @@ void RasterizerOpenGL::SetupVertexFormat() {
     // avoid OpenGL errors.
     // TODO(Subv): Analyze the shader to identify which attributes are actually used and don't
     // assume every shader uses them all.
-    for (std::size_t index = 0; index < NumSupportedVertexAttributes; ++index) {
+    for (std::size_t index = 0; index < NUM_SUPPORTED_VERTEX_ATTRIBUTES; ++index) {
         if (!flags[Dirty::VertexFormat0 + index]) {
             continue;
         }
@@ -212,9 +213,10 @@ void RasterizerOpenGL::SetupVertexFormat() {
         if (attrib.type == Maxwell::VertexAttribute::Type::SignedInt ||
             attrib.type == Maxwell::VertexAttribute::Type::UnsignedInt) {
             glVertexAttribIFormat(gl_index, attrib.ComponentCount(),
-                                  MaxwellToGL::VertexType(attrib), attrib.offset);
+                                  MaxwellToGL::VertexFormat(attrib), attrib.offset);
         } else {
-            glVertexAttribFormat(gl_index, attrib.ComponentCount(), MaxwellToGL::VertexType(attrib),
+            glVertexAttribFormat(gl_index, attrib.ComponentCount(),
+                                 MaxwellToGL::VertexFormat(attrib),
                                  attrib.IsNormalized() ? GL_TRUE : GL_FALSE, attrib.offset);
         }
         glVertexAttribBinding(gl_index, attrib.buffer);
@@ -231,9 +233,11 @@ void RasterizerOpenGL::SetupVertexBuffer() {
 
     MICROPROFILE_SCOPE(OpenGL_VB);
 
+    const bool use_unified_memory = device.HasVertexBufferUnifiedMemory();
+
     // Upload all guest vertex arrays sequentially to our buffer
     const auto& regs = gpu.regs;
-    for (std::size_t index = 0; index < Maxwell::NumVertexArrays; ++index) {
+    for (std::size_t index = 0; index < NUM_SUPPORTED_VERTEX_BINDINGS; ++index) {
         if (!flags[Dirty::VertexBuffer0 + index]) {
             continue;
         }
@@ -246,16 +250,25 @@ void RasterizerOpenGL::SetupVertexBuffer() {
 
         const GPUVAddr start = vertex_array.StartAddress();
         const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
-
         ASSERT(end >= start);
+
+        const GLuint gl_index = static_cast<GLuint>(index);
         const u64 size = end - start;
         if (size == 0) {
-            glBindVertexBuffer(static_cast<GLuint>(index), 0, 0, vertex_array.stride);
+            glBindVertexBuffer(gl_index, 0, 0, vertex_array.stride);
+            if (use_unified_memory) {
+                glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, gl_index, 0, 0);
+            }
             continue;
         }
-        const auto [vertex_buffer, vertex_buffer_offset] = buffer_cache.UploadMemory(start, size);
-        glBindVertexBuffer(static_cast<GLuint>(index), vertex_buffer, vertex_buffer_offset,
-                           vertex_array.stride);
+        const auto info = buffer_cache.UploadMemory(start, size);
+        if (use_unified_memory) {
+            glBindVertexBuffer(gl_index, 0, 0, vertex_array.stride);
+            glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, gl_index,
+                                   info.address + info.offset, size);
+        } else {
+            glBindVertexBuffer(gl_index, info.handle, info.offset, vertex_array.stride);
+        }
     }
 }
 
@@ -268,7 +281,7 @@ void RasterizerOpenGL::SetupVertexInstances() {
     flags[Dirty::VertexInstances] = false;
 
     const auto& regs = gpu.regs;
-    for (std::size_t index = 0; index < NumSupportedVertexAttributes; ++index) {
+    for (std::size_t index = 0; index < NUM_SUPPORTED_VERTEX_ATTRIBUTES; ++index) {
         if (!flags[Dirty::VertexInstance0 + index]) {
             continue;
         }
@@ -285,9 +298,9 @@ GLintptr RasterizerOpenGL::SetupIndexBuffer() {
     MICROPROFILE_SCOPE(OpenGL_Index);
     const auto& regs = system.GPU().Maxwell3D().regs;
     const std::size_t size = CalculateIndexBufferSize();
-    const auto [buffer, offset] = buffer_cache.UploadMemory(regs.index_array.IndexStart(), size);
-    glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, buffer);
-    return offset;
+    const auto info = buffer_cache.UploadMemory(regs.index_array.IndexStart(), size);
+    glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, info.handle);
+    return info.offset;
 }
 
 void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
@@ -643,9 +656,9 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     if (!device.UseAssemblyShaders()) {
         MaxwellUniformData ubo;
         ubo.SetFromRegs(gpu);
-        const auto [buffer, offset] =
+        const auto info =
             buffer_cache.UploadHostMemory(&ubo, sizeof(ubo), device.GetUniformBufferAlignment());
-        glBindBufferRange(GL_UNIFORM_BUFFER, EmulationUniformBlockBinding, buffer, offset,
+        glBindBufferRange(GL_UNIFORM_BUFFER, EmulationUniformBlockBinding, info.handle, info.offset,
                           static_cast<GLsizeiptr>(sizeof(ubo)));
     }
 
@@ -956,8 +969,7 @@ void RasterizerOpenGL::SetupConstBuffer(GLenum stage, u32 binding,
         if (device.UseAssemblyShaders()) {
             glBindBufferRangeNV(stage, entry.GetIndex(), 0, 0, 0);
         } else {
-            glBindBufferRange(GL_UNIFORM_BUFFER, binding,
-                              buffer_cache.GetEmptyBuffer(sizeof(float)), 0, sizeof(float));
+            glBindBufferRange(GL_UNIFORM_BUFFER, binding, 0, 0, sizeof(float));
         }
         return;
     }
@@ -970,24 +982,25 @@ void RasterizerOpenGL::SetupConstBuffer(GLenum stage, u32 binding,
 
     const std::size_t alignment = use_unified ? 4 : device.GetUniformBufferAlignment();
     const GPUVAddr gpu_addr = buffer.address;
-    auto [cbuf, offset] = buffer_cache.UploadMemory(gpu_addr, size, alignment, false, fast_upload);
+    auto info = buffer_cache.UploadMemory(gpu_addr, size, alignment, false, fast_upload);
 
     if (device.UseAssemblyShaders()) {
         UNIMPLEMENTED_IF(use_unified);
-        if (offset != 0) {
+        if (info.offset != 0) {
             const GLuint staging_cbuf = staging_cbufs[current_cbuf++];
-            glCopyNamedBufferSubData(cbuf, staging_cbuf, offset, 0, size);
-            cbuf = staging_cbuf;
-            offset = 0;
+            glCopyNamedBufferSubData(info.handle, staging_cbuf, info.offset, 0, size);
+            info.handle = staging_cbuf;
+            info.offset = 0;
         }
-        glBindBufferRangeNV(stage, binding, cbuf, offset, size);
+        glBindBufferRangeNV(stage, binding, info.handle, info.offset, size);
         return;
     }
 
     if (use_unified) {
-        glCopyNamedBufferSubData(cbuf, unified_uniform_buffer.handle, offset, unified_offset, size);
+        glCopyNamedBufferSubData(info.handle, unified_uniform_buffer.handle, info.offset,
+                                 unified_offset, size);
     } else {
-        glBindBufferRange(GL_UNIFORM_BUFFER, binding, cbuf, offset, size);
+        glBindBufferRange(GL_UNIFORM_BUFFER, binding, info.handle, info.offset, size);
     }
 }
 
@@ -1023,9 +1036,8 @@ void RasterizerOpenGL::SetupComputeGlobalMemory(Shader* kernel) {
 void RasterizerOpenGL::SetupGlobalMemory(u32 binding, const GlobalMemoryEntry& entry,
                                          GPUVAddr gpu_addr, std::size_t size) {
     const auto alignment{device.GetShaderStorageBufferAlignment()};
-    const auto [ssbo, buffer_offset] =
-        buffer_cache.UploadMemory(gpu_addr, size, alignment, entry.is_written);
-    glBindBufferRange(GL_SHADER_STORAGE_BUFFER, binding, ssbo, buffer_offset,
+    const auto info = buffer_cache.UploadMemory(gpu_addr, size, alignment, entry.is_written);
+    glBindBufferRange(GL_SHADER_STORAGE_BUFFER, binding, info.handle, info.offset,
                       static_cast<GLsizeiptr>(size));
 }
 
@@ -1712,8 +1724,9 @@ void RasterizerOpenGL::EndTransformFeedback() {
         const GLuint handle = transform_feedback_buffers[index].handle;
         const GPUVAddr gpu_addr = binding.Address();
         const std::size_t size = binding.buffer_size;
-        const auto [dest_buffer, offset] = buffer_cache.UploadMemory(gpu_addr, size, 4, true);
-        glCopyNamedBufferSubData(handle, dest_buffer, 0, offset, static_cast<GLsizeiptr>(size));
+        const auto info = buffer_cache.UploadMemory(gpu_addr, size, 4, true);
+        glCopyNamedBufferSubData(handle, info.handle, 0, info.offset,
+                                 static_cast<GLsizeiptr>(size));
     }
 }
 
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 46e780a06..c6a3bf3a1 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -460,8 +460,9 @@ Shader* ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
         const u8* host_ptr_b = memory_manager.GetPointer(address_b);
         code_b = GetShaderCode(memory_manager, address_b, host_ptr_b, false);
     }
+    const std::size_t code_size = code.size() * sizeof(u64);
 
-    const auto unique_identifier = GetUniqueIdentifier(
+    const u64 unique_identifier = GetUniqueIdentifier(
         GetShaderType(program), program == Maxwell::ShaderProgram::VertexA, code, code_b);
 
     const ShaderParameters params{system,    disk_cache, device,
@@ -477,7 +478,7 @@ Shader* ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
 
     Shader* const result = shader.get();
     if (cpu_addr) {
-        Register(std::move(shader), *cpu_addr, code.size() * sizeof(u64));
+        Register(std::move(shader), *cpu_addr, code_size);
     } else {
         null_shader = std::move(shader);
     }
@@ -495,8 +496,9 @@ Shader* ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
 
     const auto host_ptr{memory_manager.GetPointer(code_addr)};
     // No kernel found, create a new one
-    auto code{GetShaderCode(memory_manager, code_addr, host_ptr, true)};
-    const auto unique_identifier{GetUniqueIdentifier(ShaderType::Compute, false, code)};
+    ProgramCode code{GetShaderCode(memory_manager, code_addr, host_ptr, true)};
+    const std::size_t code_size{code.size() * sizeof(u64)};
+    const u64 unique_identifier{GetUniqueIdentifier(ShaderType::Compute, false, code)};
 
     const ShaderParameters params{system,    disk_cache, device,
                                   *cpu_addr, host_ptr,   unique_identifier};
@@ -511,7 +513,7 @@ Shader* ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
 
     Shader* const result = kernel.get();
     if (cpu_addr) {
-        Register(std::move(kernel), *cpu_addr, code.size() * sizeof(u64));
+        Register(std::move(kernel), *cpu_addr, code_size);
     } else {
         null_kernel = std::move(kernel);
     }
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index 6848f1388..994aaeaf2 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -37,7 +37,6 @@ namespace OpenGL {
 
 class Device;
 class RasterizerOpenGL;
-struct UnspecializedShader;
 
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index d6e30b321..2c49aeaac 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -37,6 +37,7 @@ using Tegra::Shader::IpaMode;
 using Tegra::Shader::IpaSampleMode;
 using Tegra::Shader::PixelImap;
 using Tegra::Shader::Register;
+using Tegra::Shader::TextureType;
 using VideoCommon::Shader::BuildTransformFeedback;
 using VideoCommon::Shader::Registry;
 
@@ -526,6 +527,9 @@ private:
         if (device.HasImageLoadFormatted()) {
             code.AddLine("#extension GL_EXT_shader_image_load_formatted : require");
         }
+        if (device.HasTextureShadowLod()) {
+            code.AddLine("#extension GL_EXT_texture_shadow_lod : require");
+        }
         if (device.HasWarpIntrinsics()) {
             code.AddLine("#extension GL_NV_gpu_shader5 : require");
             code.AddLine("#extension GL_NV_shader_thread_group : require");
@@ -909,13 +913,13 @@ private:
                     return "samplerBuffer";
                 }
                 switch (sampler.type) {
-                case Tegra::Shader::TextureType::Texture1D:
+                case TextureType::Texture1D:
                     return "sampler1D";
-                case Tegra::Shader::TextureType::Texture2D:
+                case TextureType::Texture2D:
                     return "sampler2D";
-                case Tegra::Shader::TextureType::Texture3D:
+                case TextureType::Texture3D:
                     return "sampler3D";
-                case Tegra::Shader::TextureType::TextureCube:
+                case TextureType::TextureCube:
                     return "samplerCube";
                 default:
                     UNREACHABLE();
@@ -1380,8 +1384,19 @@ private:
         const std::size_t count = operation.GetOperandsCount();
         const bool has_array = meta->sampler.is_array;
         const bool has_shadow = meta->sampler.is_shadow;
+        const bool workaround_lod_array_shadow_as_grad =
+            !device.HasTextureShadowLod() && function_suffix == "Lod" && meta->sampler.is_shadow &&
+            ((meta->sampler.type == TextureType::Texture2D && meta->sampler.is_array) ||
+             meta->sampler.type == TextureType::TextureCube);
+
+        std::string expr = "texture";
+
+        if (workaround_lod_array_shadow_as_grad) {
+            expr += "Grad";
+        } else {
+            expr += function_suffix;
+        }
 
-        std::string expr = "texture" + function_suffix;
         if (!meta->aoffi.empty()) {
             expr += "Offset";
         } else if (!meta->ptp.empty()) {
@@ -1415,6 +1430,16 @@ private:
             expr += ')';
         }
 
+        if (workaround_lod_array_shadow_as_grad) {
+            switch (meta->sampler.type) {
+            case TextureType::Texture2D:
+                return expr + ", vec2(0.0), vec2(0.0))";
+            case TextureType::TextureCube:
+                return expr + ", vec3(0.0), vec3(0.0))";
+            }
+            UNREACHABLE();
+        }
+
         for (const auto& variant : extras) {
             if (const auto argument = std::get_if<TextureArgument>(&variant)) {
                 expr += GenerateTextureArgument(*argument);
@@ -2041,8 +2066,19 @@ private:
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
-        std::string expr = GenerateTexture(
-            operation, "Lod", {TextureArgument{Type::Float, meta->lod}, TextureOffset{}});
+        std::string expr{};
+
+        if (!device.HasTextureShadowLod() && meta->sampler.is_shadow &&
+            ((meta->sampler.type == TextureType::Texture2D && meta->sampler.is_array) ||
+             meta->sampler.type == TextureType::TextureCube)) {
+            LOG_ERROR(Render_OpenGL,
+                      "Device lacks GL_EXT_texture_shadow_lod, using textureGrad as a workaround");
+            expr = GenerateTexture(operation, "Lod", {});
+        } else {
+            expr = GenerateTexture(operation, "Lod",
+                                   {TextureArgument{Type::Float, meta->lod}, TextureOffset{}});
+        }
+
         if (meta->sampler.is_shadow) {
             expr = "vec4(" + expr + ')';
         }
diff --git a/src/video_core/renderer_opengl/gl_stream_buffer.cpp b/src/video_core/renderer_opengl/gl_stream_buffer.cpp
index 932a2f69e..3655ff629 100644
--- a/src/video_core/renderer_opengl/gl_stream_buffer.cpp
+++ b/src/video_core/renderer_opengl/gl_stream_buffer.cpp
@@ -2,11 +2,13 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <deque>
+#include <tuple>
 #include <vector>
+
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/microprofile.h"
+#include "video_core/renderer_opengl/gl_device.h"
 #include "video_core/renderer_opengl/gl_stream_buffer.h"
 
 MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",
@@ -14,8 +16,7 @@ MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",
 
 namespace OpenGL {
 
-OGLStreamBuffer::OGLStreamBuffer(GLsizeiptr size, bool vertex_data_usage, bool prefer_coherent,
-                                 bool use_persistent)
+OGLStreamBuffer::OGLStreamBuffer(const Device& device, GLsizeiptr size, bool vertex_data_usage)
     : buffer_size(size) {
     gl_buffer.Create();
 
@@ -29,23 +30,19 @@ OGLStreamBuffer::OGLStreamBuffer(GLsizeiptr size, bool vertex_data_usage, bool p
         allocate_size *= 2;
     }
 
-    if (use_persistent) {
-        persistent = true;
-        coherent = prefer_coherent;
-        const GLbitfield flags =
-            GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT | (coherent ? GL_MAP_COHERENT_BIT : 0);
-        glNamedBufferStorage(gl_buffer.handle, allocate_size, nullptr, flags);
-        mapped_ptr = static_cast<u8*>(glMapNamedBufferRange(
-            gl_buffer.handle, 0, buffer_size, flags | (coherent ? 0 : GL_MAP_FLUSH_EXPLICIT_BIT)));
-    } else {
-        glNamedBufferData(gl_buffer.handle, allocate_size, nullptr, GL_STREAM_DRAW);
+    static constexpr GLbitfield flags = GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT;
+    glNamedBufferStorage(gl_buffer.handle, allocate_size, nullptr, flags);
+    mapped_ptr = static_cast<u8*>(
+        glMapNamedBufferRange(gl_buffer.handle, 0, buffer_size, flags | GL_MAP_FLUSH_EXPLICIT_BIT));
+
+    if (device.HasVertexBufferUnifiedMemory()) {
+        glMakeNamedBufferResidentNV(gl_buffer.handle, GL_READ_ONLY);
+        glGetNamedBufferParameterui64vNV(gl_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, &gpu_address);
     }
 }
 
 OGLStreamBuffer::~OGLStreamBuffer() {
-    if (persistent) {
-        glUnmapNamedBuffer(gl_buffer.handle);
-    }
+    glUnmapNamedBuffer(gl_buffer.handle);
     gl_buffer.Release();
 }
 
@@ -60,36 +57,21 @@ std::tuple<u8*, GLintptr, bool> OGLStreamBuffer::Map(GLsizeiptr size, GLintptr a
 
     bool invalidate = false;
     if (buffer_pos + size > buffer_size) {
+        MICROPROFILE_SCOPE(OpenGL_StreamBuffer);
+        glInvalidateBufferData(gl_buffer.handle);
+
         buffer_pos = 0;
         invalidate = true;
-
-        if (persistent) {
-            glUnmapNamedBuffer(gl_buffer.handle);
-        }
     }
 
-    if (invalidate || !persistent) {
-        MICROPROFILE_SCOPE(OpenGL_StreamBuffer);
-        GLbitfield flags = GL_MAP_WRITE_BIT | (persistent ? GL_MAP_PERSISTENT_BIT : 0) |
-                           (coherent ? GL_MAP_COHERENT_BIT : GL_MAP_FLUSH_EXPLICIT_BIT) |
-                           (invalidate ? GL_MAP_INVALIDATE_BUFFER_BIT : GL_MAP_UNSYNCHRONIZED_BIT);
-        mapped_ptr = static_cast<u8*>(
-            glMapNamedBufferRange(gl_buffer.handle, buffer_pos, buffer_size - buffer_pos, flags));
-        mapped_offset = buffer_pos;
-    }
-
-    return std::make_tuple(mapped_ptr + buffer_pos - mapped_offset, buffer_pos, invalidate);
+    return std::make_tuple(mapped_ptr + buffer_pos, buffer_pos, invalidate);
 }
 
 void OGLStreamBuffer::Unmap(GLsizeiptr size) {
     ASSERT(size <= mapped_size);
 
-    if (!coherent && size > 0) {
-        glFlushMappedNamedBufferRange(gl_buffer.handle, buffer_pos - mapped_offset, size);
-    }
-
-    if (!persistent) {
-        glUnmapNamedBuffer(gl_buffer.handle);
+    if (size > 0) {
+        glFlushMappedNamedBufferRange(gl_buffer.handle, buffer_pos, size);
     }
 
     buffer_pos += size;
diff --git a/src/video_core/renderer_opengl/gl_stream_buffer.h b/src/video_core/renderer_opengl/gl_stream_buffer.h
index 866da3594..307a67113 100644
--- a/src/video_core/renderer_opengl/gl_stream_buffer.h
+++ b/src/video_core/renderer_opengl/gl_stream_buffer.h
@@ -11,10 +11,11 @@
 
 namespace OpenGL {
 
+class Device;
+
 class OGLStreamBuffer : private NonCopyable {
 public:
-    explicit OGLStreamBuffer(GLsizeiptr size, bool vertex_data_usage, bool prefer_coherent = false,
-                             bool use_persistent = true);
+    explicit OGLStreamBuffer(const Device& device, GLsizeiptr size, bool vertex_data_usage);
     ~OGLStreamBuffer();
 
     /*
@@ -33,19 +34,20 @@ public:
         return gl_buffer.handle;
     }
 
-    GLsizeiptr Size() const {
+    u64 Address() const {
+        return gpu_address;
+    }
+
+    GLsizeiptr Size() const noexcept {
         return buffer_size;
     }
 
 private:
     OGLBuffer gl_buffer;
 
-    bool coherent = false;
-    bool persistent = false;
-
+    GLuint64EXT gpu_address = 0;
     GLintptr buffer_pos = 0;
     GLsizeiptr buffer_size = 0;
-    GLintptr mapped_offset = 0;
     GLsizeiptr mapped_size = 0;
     u8* mapped_ptr = nullptr;
 };
diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h
index 35e329240..fe9bd4b5a 100644
--- a/src/video_core/renderer_opengl/maxwell_to_gl.h
+++ b/src/video_core/renderer_opengl/maxwell_to_gl.h
@@ -24,10 +24,11 @@ namespace MaxwellToGL {
 
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
-inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
+inline GLenum VertexFormat(Maxwell::VertexAttribute attrib) {
     switch (attrib.type) {
-    case Maxwell::VertexAttribute::Type::UnsignedInt:
     case Maxwell::VertexAttribute::Type::UnsignedNorm:
+    case Maxwell::VertexAttribute::Type::UnsignedScaled:
+    case Maxwell::VertexAttribute::Type::UnsignedInt:
         switch (attrib.size) {
         case Maxwell::VertexAttribute::Size::Size_8:
         case Maxwell::VertexAttribute::Size::Size_8_8:
@@ -48,8 +49,9 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
             return GL_UNSIGNED_INT_2_10_10_10_REV;
         }
         break;
-    case Maxwell::VertexAttribute::Type::SignedInt:
     case Maxwell::VertexAttribute::Type::SignedNorm:
+    case Maxwell::VertexAttribute::Type::SignedScaled:
+    case Maxwell::VertexAttribute::Type::SignedInt:
         switch (attrib.size) {
         case Maxwell::VertexAttribute::Size::Size_8:
         case Maxwell::VertexAttribute::Size::Size_8_8:
@@ -84,36 +86,8 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
             return GL_FLOAT;
         }
         break;
-    case Maxwell::VertexAttribute::Type::UnsignedScaled:
-        switch (attrib.size) {
-        case Maxwell::VertexAttribute::Size::Size_8:
-        case Maxwell::VertexAttribute::Size::Size_8_8:
-        case Maxwell::VertexAttribute::Size::Size_8_8_8:
-        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
-            return GL_UNSIGNED_BYTE;
-        case Maxwell::VertexAttribute::Size::Size_16:
-        case Maxwell::VertexAttribute::Size::Size_16_16:
-        case Maxwell::VertexAttribute::Size::Size_16_16_16:
-        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
-            return GL_UNSIGNED_SHORT;
-        }
-        break;
-    case Maxwell::VertexAttribute::Type::SignedScaled:
-        switch (attrib.size) {
-        case Maxwell::VertexAttribute::Size::Size_8:
-        case Maxwell::VertexAttribute::Size::Size_8_8:
-        case Maxwell::VertexAttribute::Size::Size_8_8_8:
-        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
-            return GL_BYTE;
-        case Maxwell::VertexAttribute::Size::Size_16:
-        case Maxwell::VertexAttribute::Size::Size_16_16:
-        case Maxwell::VertexAttribute::Size::Size_16_16_16:
-        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
-            return GL_SHORT;
-        }
-        break;
     }
-    UNIMPLEMENTED_MSG("Unimplemented vertex type={} and size={}", attrib.TypeString(),
+    UNIMPLEMENTED_MSG("Unimplemented vertex format of type={} and size={}", attrib.TypeString(),
                       attrib.SizeString());
     return {};
 }
@@ -217,6 +191,12 @@ inline GLenum WrapMode(Tegra::Texture::WrapMode wrap_mode) {
         } else {
             return GL_MIRROR_CLAMP_TO_EDGE;
         }
+    case Tegra::Texture::WrapMode::MirrorOnceClampOGL:
+        if (GL_EXT_texture_mirror_clamp) {
+            return GL_MIRROR_CLAMP_EXT;
+        } else {
+            return GL_MIRROR_CLAMP_TO_EDGE;
+        }
     }
     UNIMPLEMENTED_MSG("Unimplemented texture wrap mode={}", static_cast<u32>(wrap_mode));
     return GL_REPEAT;
diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp
index 6214fcbc3..c40adb6e7 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.cpp
+++ b/src/video_core/renderer_opengl/renderer_opengl.cpp
@@ -488,6 +488,15 @@ void RendererOpenGL::InitOpenGLObjects() {
 
     // Clear screen to black
     LoadColorToActiveGLTexture(0, 0, 0, 0, screen_info.texture);
+
+    // Enable unified vertex attributes and query vertex buffer address when the driver supports it
+    if (device.HasVertexBufferUnifiedMemory()) {
+        glEnableClientState(GL_VERTEX_ATTRIB_ARRAY_UNIFIED_NV);
+
+        glMakeNamedBufferResidentNV(vertex_buffer.handle, GL_READ_ONLY);
+        glGetNamedBufferParameterui64vNV(vertex_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV,
+                                         &vertex_buffer_address);
+    }
 }
 
 void RendererOpenGL::AddTelemetryFields() {
@@ -656,7 +665,13 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
                          offsetof(ScreenRectVertex, tex_coord));
     glVertexAttribBinding(PositionLocation, 0);
     glVertexAttribBinding(TexCoordLocation, 0);
-    glBindVertexBuffer(0, vertex_buffer.handle, 0, sizeof(ScreenRectVertex));
+    if (device.HasVertexBufferUnifiedMemory()) {
+        glBindVertexBuffer(0, 0, 0, sizeof(ScreenRectVertex));
+        glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, 0, vertex_buffer_address,
+                               sizeof(vertices));
+    } else {
+        glBindVertexBuffer(0, vertex_buffer.handle, 0, sizeof(ScreenRectVertex));
+    }
 
     glBindTextureUnit(0, screen_info.display_texture);
     glBindSampler(0, 0);
diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h
index 61bf507f4..8b18d32e6 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.h
+++ b/src/video_core/renderer_opengl/renderer_opengl.h
@@ -107,6 +107,9 @@ private:
     OGLPipeline pipeline;
     OGLFramebuffer screenshot_framebuffer;
 
+    // GPU address of the vertex buffer
+    GLuint64EXT vertex_buffer_address = 0;
+
     /// Display information for Switch screen
     ScreenInfo screen_info;