7 files changed, 388 insertions, 141 deletions

diff --git a/externals/dynarmic b/externals/dynarmic
-Subproject 171d11659d760a4d4674d3a90698fe31ea407e2
+Subproject 4e6848d1c9e8dadc70595c15b5589f8b14aad47
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 1fcd13f04..587d9dffb 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -738,7 +738,7 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader,
         }
 
         texture_samplers[current_bindpoint].SyncWithConfig(texture.tsc);
-        Surface surface = res_cache.GetTextureSurface(texture);
+        Surface surface = res_cache.GetTextureSurface(texture, entry);
         if (surface != nullptr) {
             state.texture_units[current_bindpoint].texture = surface->Texture().handle;
             state.texture_units[current_bindpoint].target = surface->Target();
@@ -909,7 +909,10 @@ void RasterizerOpenGL::SyncTransformFeedback() {
 void RasterizerOpenGL::SyncPointState() {
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
 
-    state.point.size = regs.point_size;
+    // TODO(Rodrigo): Most games do not set a point size. I think this is a case of a
+    // register carrying a default value. For now, if the point size is zero, assume it's
+    // OpenGL's default (1).
+    state.point.size = regs.point_size == 0 ? 1 : regs.point_size;
 }
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 24a540258..ce967c4d6 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -41,7 +41,7 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
 }
 
 /*static*/ SurfaceParams SurfaceParams::CreateForTexture(
-    const Tegra::Texture::FullTextureInfo& config) {
+    const Tegra::Texture::FullTextureInfo& config, const GLShader::SamplerEntry& entry) {
     SurfaceParams params{};
     params.addr = TryGetCpuAddr(config.tic.Address());
     params.is_tiled = config.tic.IsTiled();
@@ -60,9 +60,23 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     case SurfaceTarget::Texture2D:
         params.depth = 1;
         break;
+    case SurfaceTarget::TextureCubemap:
+        params.depth = config.tic.Depth() * 6;
+        break;
     case SurfaceTarget::Texture3D:
+        params.depth = config.tic.Depth();
+        break;
     case SurfaceTarget::Texture2DArray:
         params.depth = config.tic.Depth();
+        if (!entry.IsArray()) {
+            // TODO(bunnei): We have seen games re-use a Texture2D as Texture2DArray with depth of
+            // one, but sample the texture in the shader as if it were not an array texture. This
+            // probably is valid on hardware, but we still need to write a test to confirm this. In
+            // emulation, the workaround here is to continue to treat this as a Texture2D. An
+            // example game that does this is Super Mario Odyssey (in Cloud Kingdom).
+            ASSERT(params.depth == 1);
+            params.target = SurfaceTarget::Texture2D;
+        }
         break;
     default:
         LOG_CRITICAL(HW_GPU, "Unknown depth for target={}", static_cast<u32>(params.target));
@@ -71,7 +85,11 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
         break;
     }
 
-    params.size_in_bytes = params.SizeInBytes();
+    params.size_in_bytes_total = params.SizeInBytesTotal();
+    params.size_in_bytes_2d = params.SizeInBytes2D();
+    params.max_mip_level = config.tic.max_mip_level + 1;
+    params.rt = {};
+
     return params;
 }
 
@@ -89,7 +107,16 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     params.unaligned_height = config.height;
     params.target = SurfaceTarget::Texture2D;
     params.depth = 1;
-    params.size_in_bytes = params.SizeInBytes();
+    params.size_in_bytes_total = params.SizeInBytesTotal();
+    params.size_in_bytes_2d = params.SizeInBytes2D();
+    params.max_mip_level = 0;
+
+    // Render target specific parameters, not used for caching
+    params.rt.index = static_cast<u32>(index);
+    params.rt.array_mode = config.array_mode;
+    params.rt.layer_stride = config.layer_stride;
+    params.rt.base_layer = config.base_layer;
+
     return params;
 }
 
@@ -108,7 +135,11 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     params.unaligned_height = zeta_height;
     params.target = SurfaceTarget::Texture2D;
     params.depth = 1;
-    params.size_in_bytes = params.SizeInBytes();
+    params.size_in_bytes_total = params.SizeInBytesTotal();
+    params.size_in_bytes_2d = params.SizeInBytes2D();
+    params.max_mip_level = 0;
+    params.rt = {};
+
     return params;
 }
 
@@ -400,9 +431,13 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, std::size_t, VAddr),
         // clang-format on
 };
 
-static bool BlitTextures(GLuint src_tex, const MathUtil::Rectangle<u32>& src_rect, GLuint dst_tex,
-                         const MathUtil::Rectangle<u32>& dst_rect, SurfaceType type,
-                         GLuint read_fb_handle, GLuint draw_fb_handle) {
+static bool BlitSurface(const Surface& src_surface, const Surface& dst_surface,
+                        GLuint read_fb_handle, GLuint draw_fb_handle, GLenum src_attachment = 0,
+                        GLenum dst_attachment = 0, std::size_t cubemap_face = 0) {
+
+    const auto& src_params{src_surface->GetSurfaceParams()};
+    const auto& dst_params{dst_surface->GetSurfaceParams()};
+
     OpenGLState prev_state{OpenGLState::GetCurState()};
     SCOPE_EXIT({ prev_state.Apply(); });
 
@@ -413,47 +448,203 @@ static bool BlitTextures(GLuint src_tex, const MathUtil::Rectangle<u32>& src_rec
 
     u32 buffers{};
 
-    if (type == SurfaceType::ColorTexture) {
-        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, src_tex,
-                               0);
-        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
-                               0);
+    if (src_params.type == SurfaceType::ColorTexture) {
+        switch (src_params.target) {
+        case SurfaceParams::SurfaceTarget::Texture2D:
+            glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                                   GL_TEXTURE_2D, src_surface->Texture().handle, 0);
+            glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                                   0, 0);
+            break;
+        case SurfaceParams::SurfaceTarget::TextureCubemap:
+            glFramebufferTexture2D(
+                GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cubemap_face),
+                src_surface->Texture().handle, 0);
+            glFramebufferTexture2D(
+                GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
+                static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cubemap_face), 0, 0);
+            break;
+        case SurfaceParams::SurfaceTarget::Texture2DArray:
+            glFramebufferTextureLayer(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                                      src_surface->Texture().handle, 0, 0);
+            glFramebufferTextureLayer(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, 0, 0, 0);
+            break;
+        case SurfaceParams::SurfaceTarget::Texture3D:
+            glFramebufferTexture3D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                                   SurfaceTargetToGL(src_params.target),
+                                   src_surface->Texture().handle, 0, 0);
+            glFramebufferTexture3D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
+                                   SurfaceTargetToGL(src_params.target), 0, 0, 0);
+            break;
+        default:
+            glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                                   GL_TEXTURE_2D, src_surface->Texture().handle, 0);
+            glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                                   0, 0);
+            break;
+        }
+
+        switch (dst_params.target) {
+        case SurfaceParams::SurfaceTarget::Texture2D:
+            glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                                   GL_TEXTURE_2D, dst_surface->Texture().handle, 0);
+            glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                                   0, 0);
+            break;
+        case SurfaceParams::SurfaceTarget::TextureCubemap:
+            glFramebufferTexture2D(
+                GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cubemap_face),
+                dst_surface->Texture().handle, 0);
+            glFramebufferTexture2D(
+                GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
+                static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cubemap_face), 0, 0);
+            break;
+        case SurfaceParams::SurfaceTarget::Texture2DArray:
+            glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                                      dst_surface->Texture().handle, 0, 0);
+            glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, 0, 0, 0);
+            break;
 
-        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex,
-                               0);
-        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
-                               0);
+        case SurfaceParams::SurfaceTarget::Texture3D:
+            glFramebufferTexture3D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                                   SurfaceTargetToGL(dst_params.target),
+                                   dst_surface->Texture().handle, 0, 0);
+            glFramebufferTexture3D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
+                                   SurfaceTargetToGL(dst_params.target), 0, 0, 0);
+            break;
+        default:
+            glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                                   GL_TEXTURE_2D, dst_surface->Texture().handle, 0);
+            glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                                   0, 0);
+            break;
+        }
 
         buffers = GL_COLOR_BUFFER_BIT;
-    } else if (type == SurfaceType::Depth) {
-        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
-        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, src_tex, 0);
+    } else if (src_params.type == SurfaceType::Depth) {
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                               GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D,
+                               src_surface->Texture().handle, 0);
         glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
 
-        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
-        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, dst_tex, 0);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                               GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D,
+                               dst_surface->Texture().handle, 0);
         glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
 
         buffers = GL_DEPTH_BUFFER_BIT;
-    } else if (type == SurfaceType::DepthStencil) {
-        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
+    } else if (src_params.type == SurfaceType::DepthStencil) {
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                               GL_TEXTURE_2D, 0, 0);
         glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
-                               src_tex, 0);
+                               src_surface->Texture().handle, 0);
 
-        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                               GL_TEXTURE_2D, 0, 0);
         glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
-                               dst_tex, 0);
+                               dst_surface->Texture().handle, 0);
 
         buffers = GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT;
     }
 
-    glBlitFramebuffer(src_rect.left, src_rect.bottom, src_rect.right, src_rect.top, dst_rect.left,
-                      dst_rect.bottom, dst_rect.right, dst_rect.top, buffers,
+    const auto& rect{src_params.GetRect()};
+    glBlitFramebuffer(rect.left, rect.bottom, rect.right, rect.top, rect.left, rect.bottom,
+                      rect.right, rect.top, buffers,
                       buffers == GL_COLOR_BUFFER_BIT ? GL_LINEAR : GL_NEAREST);
 
     return true;
 }
 
+static void CopySurface(const Surface& src_surface, const Surface& dst_surface,
+                        GLuint copy_pbo_handle, GLenum src_attachment = 0,
+                        GLenum dst_attachment = 0, std::size_t cubemap_face = 0) {
+    ASSERT_MSG(dst_attachment == 0, "Unimplemented");
+
+    const auto& src_params{src_surface->GetSurfaceParams()};
+    const auto& dst_params{dst_surface->GetSurfaceParams()};
+
+    auto source_format = GetFormatTuple(src_params.pixel_format, src_params.component_type);
+    auto dest_format = GetFormatTuple(dst_params.pixel_format, dst_params.component_type);
+
+    std::size_t buffer_size =
+        std::max(src_params.size_in_bytes_total, dst_params.size_in_bytes_total);
+
+    glBindBuffer(GL_PIXEL_PACK_BUFFER, copy_pbo_handle);
+    glBufferData(GL_PIXEL_PACK_BUFFER, buffer_size, nullptr, GL_STREAM_DRAW_ARB);
+    if (source_format.compressed) {
+        glGetCompressedTextureImage(src_surface->Texture().handle, src_attachment,
+                                    static_cast<GLsizei>(src_params.size_in_bytes_total), nullptr);
+    } else {
+        glGetTextureImage(src_surface->Texture().handle, src_attachment, source_format.format,
+                          source_format.type, static_cast<GLsizei>(src_params.size_in_bytes_total),
+                          nullptr);
+    }
+    // If the new texture is bigger than the previous one, we need to fill in the rest with data
+    // from the CPU.
+    if (src_params.size_in_bytes_total < dst_params.size_in_bytes_total) {
+        // Upload the rest of the memory.
+        if (dst_params.is_tiled) {
+            // TODO(Subv): We might have to de-tile the subtexture and re-tile it with the rest
+            // of the data in this case. Games like Super Mario Odyssey seem to hit this case
+            // when drawing, it re-uses the memory of a previous texture as a bigger framebuffer
+            // but it doesn't clear it beforehand, the texture is already full of zeros.
+            LOG_DEBUG(HW_GPU, "Trying to upload extra texture data from the CPU during "
+                              "reinterpretation but the texture is tiled.");
+        }
+        std::size_t remaining_size =
+            dst_params.size_in_bytes_total - src_params.size_in_bytes_total;
+        std::vector<u8> data(remaining_size);
+        Memory::ReadBlock(dst_params.addr + src_params.size_in_bytes_total, data.data(),
+                          data.size());
+        glBufferSubData(GL_PIXEL_PACK_BUFFER, src_params.size_in_bytes_total, remaining_size,
+                        data.data());
+    }
+
+    glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
+
+    const GLsizei width{static_cast<GLsizei>(
+        std::min(src_params.GetRect().GetWidth(), dst_params.GetRect().GetWidth()))};
+    const GLsizei height{static_cast<GLsizei>(
+        std::min(src_params.GetRect().GetHeight(), dst_params.GetRect().GetHeight()))};
+
+    glBindBuffer(GL_PIXEL_UNPACK_BUFFER, copy_pbo_handle);
+    if (dest_format.compressed) {
+        LOG_CRITICAL(HW_GPU, "Compressed copy is unimplemented!");
+        UNREACHABLE();
+    } else {
+        switch (dst_params.target) {
+        case SurfaceParams::SurfaceTarget::Texture1D:
+            glTextureSubImage1D(dst_surface->Texture().handle, 0, 0, width, dest_format.format,
+                                dest_format.type, nullptr);
+            break;
+        case SurfaceParams::SurfaceTarget::Texture2D:
+            glTextureSubImage2D(dst_surface->Texture().handle, 0, 0, 0, width, height,
+                                dest_format.format, dest_format.type, nullptr);
+            break;
+        case SurfaceParams::SurfaceTarget::Texture3D:
+        case SurfaceParams::SurfaceTarget::Texture2DArray:
+            glTextureSubImage3D(dst_surface->Texture().handle, 0, 0, 0, 0, width, height,
+                                static_cast<GLsizei>(dst_params.depth), dest_format.format,
+                                dest_format.type, nullptr);
+            break;
+        case SurfaceParams::SurfaceTarget::TextureCubemap:
+            glTextureSubImage3D(dst_surface->Texture().handle, 0, 0, 0,
+                                static_cast<GLint>(cubemap_face), width, height, 1,
+                                dest_format.format, dest_format.type, nullptr);
+            break;
+        default:
+            LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
+                         static_cast<u32>(dst_params.target));
+            UNREACHABLE();
+        }
+        glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
+    }
+}
+
 CachedSurface::CachedSurface(const SurfaceParams& params)
     : params(params), gl_target(SurfaceTargetToGL(params.target)) {
     texture.Create();
@@ -481,6 +672,7 @@ CachedSurface::CachedSurface(const SurfaceParams& params)
                            rect.GetWidth());
             break;
         case SurfaceParams::SurfaceTarget::Texture2D:
+        case SurfaceParams::SurfaceTarget::TextureCubemap:
             glTexStorage2D(SurfaceTargetToGL(params.target), 1, format_tuple.internal_format,
                            rect.GetWidth(), rect.GetHeight());
             break;
@@ -585,29 +777,39 @@ void CachedSurface::LoadGLBuffer() {
 
     const u32 bytes_per_pixel = GetGLBytesPerPixel(params.pixel_format);
     const u32 copy_size = params.width * params.height * bytes_per_pixel;
+    const std::size_t total_size = copy_size * params.depth;
 
     MICROPROFILE_SCOPE(OpenGL_SurfaceLoad);
 
     if (params.is_tiled) {
+        gl_buffer.resize(total_size);
+
         // TODO(bunnei): This only unswizzles and copies a 2D texture - we do not yet know how to do
         // this for 3D textures, etc.
         switch (params.target) {
         case SurfaceParams::SurfaceTarget::Texture2D:
             // Pass impl. to the fallback code below
             break;
+        case SurfaceParams::SurfaceTarget::Texture2DArray:
+        case SurfaceParams::SurfaceTarget::TextureCubemap:
+            for (std::size_t index = 0; index < params.depth; ++index) {
+                const std::size_t offset{index * copy_size};
+                morton_to_gl_fns[static_cast<std::size_t>(params.pixel_format)](
+                    params.width, params.block_height, params.height, gl_buffer.data() + offset,
+                    copy_size, params.addr + offset);
+            }
+            break;
         default:
             LOG_CRITICAL(HW_GPU, "Unimplemented tiled load for target={}",
                          static_cast<u32>(params.target));
             UNREACHABLE();
         }
 
-        gl_buffer.resize(static_cast<std::size_t>(params.depth) * copy_size);
         morton_to_gl_fns[static_cast<std::size_t>(params.pixel_format)](
             params.width, params.block_height, params.height, gl_buffer.data(), copy_size,
             params.addr);
     } else {
-        const u8* const texture_src_data_end{texture_src_data +
-                                             (static_cast<std::size_t>(params.depth) * copy_size)};
+        const u8* const texture_src_data_end{texture_src_data + total_size};
         gl_buffer.assign(texture_src_data, texture_src_data_end);
     }
 
@@ -634,7 +836,7 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
     // Load data from memory to the surface
     const GLint x0 = static_cast<GLint>(rect.left);
     const GLint y0 = static_cast<GLint>(rect.bottom);
-    const std::size_t buffer_offset =
+    std::size_t buffer_offset =
         static_cast<std::size_t>(static_cast<std::size_t>(y0) * params.width +
                                  static_cast<std::size_t>(x0)) *
         GetGLBytesPerPixel(params.pixel_format);
@@ -663,15 +865,25 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
             glCompressedTexImage2D(
                 SurfaceTargetToGL(params.target), 0, tuple.internal_format,
                 static_cast<GLsizei>(params.width), static_cast<GLsizei>(params.height), 0,
-                static_cast<GLsizei>(params.size_in_bytes), &gl_buffer[buffer_offset]);
+                static_cast<GLsizei>(params.size_in_bytes_2d), &gl_buffer[buffer_offset]);
             break;
         case SurfaceParams::SurfaceTarget::Texture3D:
         case SurfaceParams::SurfaceTarget::Texture2DArray:
             glCompressedTexImage3D(
                 SurfaceTargetToGL(params.target), 0, tuple.internal_format,
                 static_cast<GLsizei>(params.width), static_cast<GLsizei>(params.height),
-                static_cast<GLsizei>(params.depth), 0, static_cast<GLsizei>(params.size_in_bytes),
-                &gl_buffer[buffer_offset]);
+                static_cast<GLsizei>(params.depth), 0,
+                static_cast<GLsizei>(params.size_in_bytes_total), &gl_buffer[buffer_offset]);
+            break;
+        case SurfaceParams::SurfaceTarget::TextureCubemap:
+            for (std::size_t face = 0; face < params.depth; ++face) {
+                glCompressedTexImage2D(static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face),
+                                       0, tuple.internal_format, static_cast<GLsizei>(params.width),
+                                       static_cast<GLsizei>(params.height), 0,
+                                       static_cast<GLsizei>(params.size_in_bytes_2d),
+                                       &gl_buffer[buffer_offset]);
+                buffer_offset += params.size_in_bytes_2d;
+            }
             break;
         default:
             LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
@@ -679,8 +891,8 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
             UNREACHABLE();
             glCompressedTexImage2D(
                 GL_TEXTURE_2D, 0, tuple.internal_format, static_cast<GLsizei>(params.width),
-                static_cast<GLsizei>(params.height), 0, static_cast<GLsizei>(params.size_in_bytes),
-                &gl_buffer[buffer_offset]);
+                static_cast<GLsizei>(params.height), 0,
+                static_cast<GLsizei>(params.size_in_bytes_2d), &gl_buffer[buffer_offset]);
         }
     } else {
 
@@ -703,6 +915,15 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
                             static_cast<GLsizei>(rect.GetHeight()), params.depth, tuple.format,
                             tuple.type, &gl_buffer[buffer_offset]);
             break;
+        case SurfaceParams::SurfaceTarget::TextureCubemap:
+            for (std::size_t face = 0; face < params.depth; ++face) {
+                glTexSubImage2D(static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face), 0, x0,
+                                y0, static_cast<GLsizei>(rect.GetWidth()),
+                                static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
+                                &gl_buffer[buffer_offset]);
+                buffer_offset += params.size_in_bytes_2d;
+            }
+            break;
         default:
             LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
                          static_cast<u32>(params.target));
@@ -722,8 +943,9 @@ RasterizerCacheOpenGL::RasterizerCacheOpenGL() {
     copy_pbo.Create();
 }
 
-Surface RasterizerCacheOpenGL::GetTextureSurface(const Tegra::Texture::FullTextureInfo& config) {
-    return GetSurface(SurfaceParams::CreateForTexture(config));
+Surface RasterizerCacheOpenGL::GetTextureSurface(const Tegra::Texture::FullTextureInfo& config,
+                                                 const GLShader::SamplerEntry& entry) {
+    return GetSurface(SurfaceParams::CreateForTexture(config, entry));
 }
 
 Surface RasterizerCacheOpenGL::GetDepthBufferSurface(bool preserve_contents) {
@@ -811,98 +1033,69 @@ Surface RasterizerCacheOpenGL::GetUncachedSurface(const SurfaceParams& params) {
     return surface;
 }
 
-Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& surface,
+Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface,
                                                const SurfaceParams& new_params) {
     // Verify surface is compatible for blitting
-    const auto& params{surface->GetSurfaceParams()};
+    auto old_params{old_surface->GetSurfaceParams()};
 
     // Get a new surface with the new parameters, and blit the previous surface to it
     Surface new_surface{GetUncachedSurface(new_params)};
 
-    if (params.pixel_format == new_params.pixel_format ||
-        !Settings::values.use_accurate_framebuffers) {
-        // If the format is the same, just do a framebuffer blit. This is significantly faster than
-        // using PBOs. The is also likely less accurate, as textures will be converted rather than
-        // reinterpreted.
-
-        BlitTextures(surface->Texture().handle, params.GetRect(), new_surface->Texture().handle,
-                     params.GetRect(), params.type, read_framebuffer.handle,
-                     draw_framebuffer.handle);
-    } else {
-        // When use_accurate_framebuffers setting is enabled, perform a more accurate surface copy,
-        // where pixels are reinterpreted as a new format (without conversion). This code path uses
-        // OpenGL PBOs and is quite slow.
-
-        auto source_format = GetFormatTuple(params.pixel_format, params.component_type);
-        auto dest_format = GetFormatTuple(new_params.pixel_format, new_params.component_type);
-
-        std::size_t buffer_size = std::max(params.SizeInBytes(), new_params.SizeInBytes());
+    // If the format is the same, just do a framebuffer blit. This is significantly faster than
+    // using PBOs. The is also likely less accurate, as textures will be converted rather than
+    // reinterpreted. When use_accurate_framebuffers setting is enabled, perform a more accurate
+    // surface copy, where pixels are reinterpreted as a new format (without conversion). This
+    // code path uses OpenGL PBOs and is quite slow.
+    const bool is_blit{old_params.pixel_format == new_params.pixel_format ||
+                       !Settings::values.use_accurate_framebuffers};
 
-        glBindBuffer(GL_PIXEL_PACK_BUFFER, copy_pbo.handle);
-        glBufferData(GL_PIXEL_PACK_BUFFER, buffer_size, nullptr, GL_STREAM_DRAW_ARB);
-        if (source_format.compressed) {
-            glGetCompressedTextureImage(surface->Texture().handle, 0,
-                                        static_cast<GLsizei>(params.SizeInBytes()), nullptr);
+    switch (new_params.target) {
+    case SurfaceParams::SurfaceTarget::Texture2D:
+        if (is_blit) {
+            BlitSurface(old_surface, new_surface, read_framebuffer.handle, draw_framebuffer.handle);
         } else {
-            glGetTextureImage(surface->Texture().handle, 0, source_format.format,
-                              source_format.type, static_cast<GLsizei>(params.SizeInBytes()),
-                              nullptr);
+            CopySurface(old_surface, new_surface, copy_pbo.handle);
         }
-        // If the new texture is bigger than the previous one, we need to fill in the rest with data
-        // from the CPU.
-        if (params.SizeInBytes() < new_params.SizeInBytes()) {
-            // Upload the rest of the memory.
-            if (new_params.is_tiled) {
-                // TODO(Subv): We might have to de-tile the subtexture and re-tile it with the rest
-                // of the data in this case. Games like Super Mario Odyssey seem to hit this case
-                // when drawing, it re-uses the memory of a previous texture as a bigger framebuffer
-                // but it doesn't clear it beforehand, the texture is already full of zeros.
-                LOG_DEBUG(HW_GPU, "Trying to upload extra texture data from the CPU during "
-                                  "reinterpretation but the texture is tiled.");
-            }
-            std::size_t remaining_size = new_params.SizeInBytes() - params.SizeInBytes();
-            std::vector<u8> data(remaining_size);
-            Memory::ReadBlock(new_params.addr + params.SizeInBytes(), data.data(), data.size());
-            glBufferSubData(GL_PIXEL_PACK_BUFFER, params.SizeInBytes(), remaining_size,
-                            data.data());
+        break;
+    case SurfaceParams::SurfaceTarget::TextureCubemap: {
+        if (old_params.rt.array_mode != 1) {
+            // TODO(bunnei): This is used by Breath of the Wild, I'm not sure how to implement this
+            // yet (array rendering used as a cubemap texture).
+            LOG_CRITICAL(HW_GPU, "Unhandled rendertarget array_mode {}", old_params.rt.array_mode);
+            UNREACHABLE();
+            return new_surface;
         }
 
-        glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
-
-        const auto& dest_rect{new_params.GetRect()};
-
-        glBindBuffer(GL_PIXEL_UNPACK_BUFFER, copy_pbo.handle);
-        if (dest_format.compressed) {
-            LOG_CRITICAL(HW_GPU, "Compressed copy is unimplemented!");
-            UNREACHABLE();
-        } else {
-            switch (new_params.target) {
-            case SurfaceParams::SurfaceTarget::Texture1D:
-                glTextureSubImage1D(new_surface->Texture().handle, 0, 0,
-                                    static_cast<GLsizei>(dest_rect.GetWidth()), dest_format.format,
-                                    dest_format.type, nullptr);
-                break;
-            case SurfaceParams::SurfaceTarget::Texture2D:
-                glTextureSubImage2D(new_surface->Texture().handle, 0, 0, 0,
-                                    static_cast<GLsizei>(dest_rect.GetWidth()),
-                                    static_cast<GLsizei>(dest_rect.GetHeight()), dest_format.format,
-                                    dest_format.type, nullptr);
-                break;
-            case SurfaceParams::SurfaceTarget::Texture3D:
-            case SurfaceParams::SurfaceTarget::Texture2DArray:
-                glTextureSubImage3D(new_surface->Texture().handle, 0, 0, 0, 0,
-                                    static_cast<GLsizei>(dest_rect.GetWidth()),
-                                    static_cast<GLsizei>(dest_rect.GetHeight()),
-                                    static_cast<GLsizei>(new_params.depth), dest_format.format,
-                                    dest_format.type, nullptr);
-                break;
-            default:
-                LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
-                             static_cast<u32>(params.target));
-                UNREACHABLE();
+        // This seems to be used for render-to-cubemap texture
+        ASSERT_MSG(old_params.target == SurfaceParams::SurfaceTarget::Texture2D, "Unexpected");
+        ASSERT_MSG(old_params.pixel_format == new_params.pixel_format, "Unexpected");
+        ASSERT_MSG(old_params.rt.base_layer == 0, "Unimplemented");
+
+        // TODO(bunnei): Verify the below - this stride seems to be in 32-bit words, not pixels.
+        // Tested with Splatoon 2, Super Mario Odyssey, and Breath of the Wild.
+        const std::size_t byte_stride{old_params.rt.layer_stride * sizeof(u32)};
+
+        for (std::size_t index = 0; index < new_params.depth; ++index) {
+            Surface face_surface{TryGetReservedSurface(old_params)};
+            ASSERT_MSG(face_surface, "Unexpected");
+
+            if (is_blit) {
+                BlitSurface(face_surface, new_surface, read_framebuffer.handle,
+                            draw_framebuffer.handle, face_surface->GetSurfaceParams().rt.index,
+                            new_params.rt.index, index);
+            } else {
+                CopySurface(face_surface, new_surface, copy_pbo.handle,
+                            face_surface->GetSurfaceParams().rt.index, new_params.rt.index, index);
             }
+
+            old_params.addr += byte_stride;
         }
-        glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
+        break;
+    }
+    default:
+        LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
+                     static_cast<u32>(new_params.target));
+        UNREACHABLE();
     }
 
     return new_surface;
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
index 80c5f324b..49025a3fe 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
@@ -9,12 +9,14 @@
 #include <memory>
 #include <vector>
 
+#include "common/alignment.h"
 #include "common/common_types.h"
 #include "common/hash.h"
 #include "common/math_util.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/rasterizer_cache.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
+#include "video_core/renderer_opengl/gl_shader_gen.h"
 #include "video_core/textures/texture.h"
 
 namespace OpenGL {
@@ -126,6 +128,8 @@ struct SurfaceParams {
         case Tegra::Texture::TextureType::Texture2D:
         case Tegra::Texture::TextureType::Texture2DNoMipmap:
             return SurfaceTarget::Texture2D;
+        case Tegra::Texture::TextureType::TextureCubemap:
+            return SurfaceTarget::TextureCubemap;
         case Tegra::Texture::TextureType::Texture1DArray:
             return SurfaceTarget::Texture1DArray;
         case Tegra::Texture::TextureType::Texture2DArray:
@@ -689,17 +693,23 @@ struct SurfaceParams {
     /// Returns the rectangle corresponding to this surface
     MathUtil::Rectangle<u32> GetRect() const;
 
-    /// Returns the size of this surface in bytes, adjusted for compression
-    std::size_t SizeInBytes() const {
+    /// Returns the size of this surface as a 2D texture in bytes, adjusted for compression
+    std::size_t SizeInBytes2D() const {
         const u32 compression_factor{GetCompressionFactor(pixel_format)};
         ASSERT(width % compression_factor == 0);
         ASSERT(height % compression_factor == 0);
         return (width / compression_factor) * (height / compression_factor) *
-               GetFormatBpp(pixel_format) * depth / CHAR_BIT;
+               GetFormatBpp(pixel_format) / CHAR_BIT;
+    }
+
+    /// Returns the total size of this surface in bytes, adjusted for compression
+    std::size_t SizeInBytesTotal() const {
+        return SizeInBytes2D() * depth;
     }
 
     /// Creates SurfaceParams from a texture configuration
-    static SurfaceParams CreateForTexture(const Tegra::Texture::FullTextureInfo& config);
+    static SurfaceParams CreateForTexture(const Tegra::Texture::FullTextureInfo& config,
+                                          const GLShader::SamplerEntry& entry);
 
     /// Creates SurfaceParams from a framebuffer configuration
     static SurfaceParams CreateForFramebuffer(std::size_t index);
@@ -711,8 +721,9 @@ struct SurfaceParams {
 
     /// Checks if surfaces are compatible for caching
     bool IsCompatibleSurface(const SurfaceParams& other) const {
-        return std::tie(pixel_format, type, width, height) ==
-               std::tie(other.pixel_format, other.type, other.width, other.height);
+        return std::tie(pixel_format, type, width, height, target, depth) ==
+               std::tie(other.pixel_format, other.type, other.width, other.height, other.target,
+                        other.depth);
     }
 
     VAddr addr;
@@ -725,8 +736,18 @@ struct SurfaceParams {
     u32 height;
     u32 depth;
     u32 unaligned_height;
-    std::size_t size_in_bytes;
+    std::size_t size_in_bytes_total;
+    std::size_t size_in_bytes_2d;
     SurfaceTarget target;
+    u32 max_mip_level;
+
+    // Render target specific parameters, not used in caching
+    struct {
+        u32 index;
+        u32 array_mode;
+        u32 layer_stride;
+        u32 base_layer;
+    } rt;
 };
 
 }; // namespace OpenGL
@@ -736,6 +757,7 @@ struct SurfaceReserveKey : Common::HashableStruct<OpenGL::SurfaceParams> {
     static SurfaceReserveKey Create(const OpenGL::SurfaceParams& params) {
         SurfaceReserveKey res;
         res.state = params;
+        res.state.rt = {}; // Ignore rt config in caching
         return res;
     }
 };
@@ -759,7 +781,7 @@ public:
     }
 
     std::size_t GetSizeInBytes() const {
-        return params.size_in_bytes;
+        return params.size_in_bytes_total;
     }
 
     const OGLTexture& Texture() const {
@@ -800,7 +822,8 @@ public:
     RasterizerCacheOpenGL();
 
     /// Get a surface based on the texture configuration
-    Surface GetTextureSurface(const Tegra::Texture::FullTextureInfo& config);
+    Surface GetTextureSurface(const Tegra::Texture::FullTextureInfo& config,
+                              const GLShader::SamplerEntry& entry);
 
     /// Get the depth surface based on the framebuffer configuration
     Surface GetDepthBufferSurface(bool preserve_contents);
@@ -822,7 +845,7 @@ private:
     Surface GetUncachedSurface(const SurfaceParams& params);
 
     /// Recreates a surface with new parameters
-    Surface RecreateSurface(const Surface& surface, const SurfaceParams& new_params);
+    Surface RecreateSurface(const Surface& old_surface, const SurfaceParams& new_params);
 
     /// Reserves a unique surface that can be reused later
     void ReserveSurface(const Surface& surface);
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index b3e95187e..579a78702 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -2000,6 +2000,14 @@ private:
                     }
                     break;
                 }
+                case Tegra::Shader::TextureType::TextureCube: {
+                    ASSERT_MSG(!is_array, "Unimplemented");
+                    std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+                    std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+                    std::string z = regs.GetRegisterAsFloat(instr.gpr20);
+                    coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
+                    break;
+                }
                 default:
                     LOG_CRITICAL(HW_GPU, "Unhandled texture type {}",
                                  static_cast<u32>(texture_type));
@@ -2018,9 +2026,12 @@ private:
                 break;
             }
             case OpCode::Id::TLDS: {
-                ASSERT(instr.tlds.GetTextureType() == Tegra::Shader::TextureType::Texture2D);
-                ASSERT(instr.tlds.IsArrayTexture() == false);
                 std::string coord;
+                const Tegra::Shader::TextureType texture_type{instr.tlds.GetTextureType()};
+                const bool is_array{instr.tlds.IsArrayTexture()};
+
+                ASSERT(texture_type == Tegra::Shader::TextureType::Texture2D);
+                ASSERT(is_array == false);
 
                 ASSERT_MSG(!instr.tlds.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
                            "NODEP is not implemented");
@@ -2029,9 +2040,14 @@ private:
                 ASSERT_MSG(!instr.tlds.UsesMiscMode(Tegra::Shader::TextureMiscMode::MZ),
                            "MZ is not implemented");
 
-                switch (instr.tlds.GetTextureType()) {
+                switch (texture_type) {
+                case Tegra::Shader::TextureType::Texture1D: {
+                    const std::string x = regs.GetRegisterAsInteger(instr.gpr8);
+                    coord = "int coords = " + x + ';';
+                    break;
+                }
                 case Tegra::Shader::TextureType::Texture2D: {
-                    if (instr.tlds.IsArrayTexture()) {
+                    if (is_array) {
                         LOG_CRITICAL(HW_GPU, "Unhandled 2d array texture");
                         UNREACHABLE();
                     } else {
@@ -2043,11 +2059,11 @@ private:
                 }
                 default:
                     LOG_CRITICAL(HW_GPU, "Unhandled texture type {}",
-                                 static_cast<u32>(instr.tlds.GetTextureType()));
+                                 static_cast<u32>(texture_type));
                     UNREACHABLE();
                 }
-                const std::string sampler = GetSampler(instr.sampler, instr.tlds.GetTextureType(),
-                                                       instr.tlds.IsArrayTexture());
+
+                const std::string sampler = GetSampler(instr.sampler, texture_type, is_array);
                 const std::string texture = "texelFetch(" + sampler + ", coords, 0)";
                 WriteTexsInstruction(instr, coord, texture);
                 break;
diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h
index c2fb824b2..14aea4838 100644
--- a/src/video_core/textures/texture.h
+++ b/src/video_core/textures/texture.h
@@ -165,6 +165,8 @@ struct TICEntry {
 
         // High 16 bits of the pitch value
         BitField<0, 16, u32> pitch_high;
+
+        BitField<28, 4, u32> max_mip_level;
     };
     union {
         BitField<0, 16, u32> width_minus_1;
diff --git a/src/yuzu/main.cpp b/src/yuzu/main.cpp
index 681758ad2..27015d02c 100644
--- a/src/yuzu/main.cpp
+++ b/src/yuzu/main.cpp
@@ -1055,11 +1055,21 @@ void GMainWindow::OnMenuInstallToNAND() {
             return;
         }
 
-        if (index >= 5)
-            index += 0x7B;
+        // If index is equal to or past Game, add the jump in TitleType.
+        if (index >= 5) {
+            index += static_cast<size_t>(FileSys::TitleType::Application) -
+                     static_cast<size_t>(FileSys::TitleType::FirmwarePackageB);
+        }
+
+        FileSys::InstallResult res;
+        if (index >= static_cast<size_t>(FileSys::TitleType::Application)) {
+            res = Service::FileSystem::GetUserNANDContents()->InstallEntry(
+                nca, static_cast<FileSys::TitleType>(index), false, qt_raw_copy);
+        } else {
+            res = Service::FileSystem::GetSystemNANDContents()->InstallEntry(
+                nca, static_cast<FileSys::TitleType>(index), false, qt_raw_copy);
+        }
 
-        const auto res = Service::FileSystem::GetUserNANDContents()->InstallEntry(
-            nca, static_cast<FileSys::TitleType>(index), false, qt_raw_copy);
         if (res == FileSys::InstallResult::Success) {
             success();
         } else if (res == FileSys::InstallResult::ErrorAlreadyExists) {