1 files changed, 198 insertions, 166 deletions

diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index f6b2c5a86..360fb0cd5 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -7,6 +7,7 @@
 
 #include "common/alignment.h"
 #include "common/assert.h"
+#include "common/logging/log.h"
 #include "common/microprofile.h"
 #include "common/scope_exit.h"
 #include "core/core.h"
@@ -51,10 +52,12 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     params.type = GetFormatType(params.pixel_format);
     params.width = Common::AlignUp(config.tic.Width(), GetCompressionFactor(params.pixel_format));
     params.height = Common::AlignUp(config.tic.Height(), GetCompressionFactor(params.pixel_format));
+    params.depth = config.tic.Depth();
     params.unaligned_height = config.tic.Height();
     params.size_in_bytes = params.SizeInBytes();
-    params.cache_width = Common::AlignUp(params.width, 16);
-    params.cache_height = Common::AlignUp(params.height, 16);
+    params.cache_width = Common::AlignUp(params.width, 8);
+    params.cache_height = Common::AlignUp(params.height, 8);
+    params.target = SurfaceTargetFromTextureType(config.tic.texture_type);
     return params;
 }
 
@@ -69,10 +72,12 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     params.type = GetFormatType(params.pixel_format);
     params.width = config.width;
     params.height = config.height;
+    params.depth = 1;
     params.unaligned_height = config.height;
     params.size_in_bytes = params.SizeInBytes();
-    params.cache_width = Common::AlignUp(params.width, 16);
-    params.cache_height = Common::AlignUp(params.height, 16);
+    params.cache_width = Common::AlignUp(params.width, 8);
+    params.cache_height = Common::AlignUp(params.height, 8);
+    params.target = SurfaceTarget::Texture2D;
     return params;
 }
 
@@ -86,13 +91,14 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
     params.pixel_format = PixelFormatFromDepthFormat(format);
     params.component_type = ComponentTypeFromDepthFormat(format);
     params.type = GetFormatType(params.pixel_format);
-    params.size_in_bytes = params.SizeInBytes();
     params.width = zeta_width;
     params.height = zeta_height;
+    params.depth = 1;
     params.unaligned_height = zeta_height;
     params.size_in_bytes = params.SizeInBytes();
-    params.cache_width = Common::AlignUp(params.width, 16);
-    params.cache_height = Common::AlignUp(params.height, 16);
+    params.cache_width = Common::AlignUp(params.width, 8);
+    params.cache_height = Common::AlignUp(params.height, 8);
+    params.target = SurfaceTarget::Texture2D;
     return params;
 }
 
@@ -166,6 +172,26 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form
      ComponentType::Float, false}, // Z32FS8
 }};
 
+static GLenum SurfaceTargetToGL(SurfaceParams::SurfaceTarget target) {
+    switch (target) {
+    case SurfaceParams::SurfaceTarget::Texture1D:
+        return GL_TEXTURE_1D;
+    case SurfaceParams::SurfaceTarget::Texture2D:
+        return GL_TEXTURE_2D;
+    case SurfaceParams::SurfaceTarget::Texture3D:
+        return GL_TEXTURE_3D;
+    case SurfaceParams::SurfaceTarget::Texture1DArray:
+        return GL_TEXTURE_1D_ARRAY;
+    case SurfaceParams::SurfaceTarget::Texture2DArray:
+        return GL_TEXTURE_2D_ARRAY;
+    case SurfaceParams::SurfaceTarget::TextureCubemap:
+        return GL_TEXTURE_CUBE_MAP;
+    }
+    LOG_CRITICAL(Render_OpenGL, "Unimplemented texture target={}", static_cast<u32>(target));
+    UNREACHABLE();
+    return {};
+}
+
 static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType component_type) {
     ASSERT(static_cast<size_t>(pixel_format) < tex_format_tuples.size());
     auto& format = tex_format_tuples[static_cast<unsigned int>(pixel_format)];
@@ -220,7 +246,8 @@ static bool IsFormatBCn(PixelFormat format) {
 }
 
 template <bool morton_to_gl, PixelFormat format>
-void MortonCopy(u32 stride, u32 block_height, u32 height, std::vector<u8>& gl_buffer, VAddr addr) {
+void MortonCopy(u32 stride, u32 block_height, u32 height, u8* gl_buffer, size_t gl_buffer_size,
+                VAddr addr) {
     constexpr u32 bytes_per_pixel = SurfaceParams::GetFormatBpp(format) / CHAR_BIT;
     constexpr u32 gl_bytes_per_pixel = CachedSurface::GetGLBytesPerPixel(format);
 
@@ -230,18 +257,18 @@ void MortonCopy(u32 stride, u32 block_height, u32 height, std::vector<u8>& gl_bu
         const u32 tile_size{IsFormatBCn(format) ? 4U : 1U};
         const std::vector<u8> data = Tegra::Texture::UnswizzleTexture(
             addr, tile_size, bytes_per_pixel, stride, height, block_height);
-        const size_t size_to_copy{std::min(gl_buffer.size(), data.size())};
-        gl_buffer.assign(data.begin(), data.begin() + size_to_copy);
+        const size_t size_to_copy{std::min(gl_buffer_size, data.size())};
+        memcpy(gl_buffer, data.data(), size_to_copy);
     } else {
         // TODO(bunnei): Assumes the default rendering GOB size of 16 (128 lines). We should
         // check the configuration for this and perform more generic un/swizzle
         LOG_WARNING(Render_OpenGL, "need to use correct swizzle/GOB parameters!");
         VideoCore::MortonCopyPixels128(stride, height, bytes_per_pixel, gl_bytes_per_pixel,
-                                       Memory::GetPointer(addr), gl_buffer.data(), morton_to_gl);
+                                       Memory::GetPointer(addr), gl_buffer, morton_to_gl);
     }
 }
 
-static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, VAddr),
+static constexpr std::array<void (*)(u32, u32, u32, u8*, size_t, VAddr),
                             SurfaceParams::MaxPixelFormat>
     morton_to_gl_fns = {
         // clang-format off
@@ -298,7 +325,7 @@ static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, VAddr),
         // clang-format on
 };
 
-static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, VAddr),
+static constexpr std::array<void (*)(u32, u32, u32, u8*, size_t, VAddr),
                             SurfaceParams::MaxPixelFormat>
     gl_to_morton_fns = {
         // clang-format off
@@ -357,33 +384,6 @@ static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, VAddr),
         // clang-format on
 };
 
-// Allocate an uninitialized texture of appropriate size and format for the surface
-static void AllocateSurfaceTexture(GLuint texture, const FormatTuple& format_tuple, u32 width,
-                                   u32 height) {
-    OpenGLState cur_state = OpenGLState::GetCurState();
-
-    // Keep track of previous texture bindings
-    GLuint old_tex = cur_state.texture_units[0].texture_2d;
-    cur_state.texture_units[0].texture_2d = texture;
-    cur_state.Apply();
-    glActiveTexture(GL_TEXTURE0);
-
-    if (!format_tuple.compressed) {
-        // Only pre-create the texture for non-compressed textures.
-        glTexImage2D(GL_TEXTURE_2D, 0, format_tuple.internal_format, width, height, 0,
-                     format_tuple.format, format_tuple.type, nullptr);
-    }
-
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 0);
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
-
-    // Restore previous texture bindings
-    cur_state.texture_units[0].texture_2d = old_tex;
-    cur_state.Apply();
-}
-
 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) {
@@ -438,12 +438,56 @@ static bool BlitTextures(GLuint src_tex, const MathUtil::Rectangle<u32>& src_rec
     return true;
 }
 
-CachedSurface::CachedSurface(const SurfaceParams& params) : params(params) {
+CachedSurface::CachedSurface(const SurfaceParams& params)
+    : params(params), gl_target(SurfaceTargetToGL(params.target)) {
     texture.Create();
     const auto& rect{params.GetRect()};
-    AllocateSurfaceTexture(texture.handle,
-                           GetFormatTuple(params.pixel_format, params.component_type),
-                           rect.GetWidth(), rect.GetHeight());
+
+    // Keep track of previous texture bindings
+    OpenGLState cur_state = OpenGLState::GetCurState();
+    const auto& old_tex = cur_state.texture_units[0];
+    SCOPE_EXIT({
+        cur_state.texture_units[0] = old_tex;
+        cur_state.Apply();
+    });
+
+    cur_state.texture_units[0].texture = texture.handle;
+    cur_state.texture_units[0].target = SurfaceTargetToGL(params.target);
+    cur_state.Apply();
+    glActiveTexture(GL_TEXTURE0);
+
+    const auto& format_tuple = GetFormatTuple(params.pixel_format, params.component_type);
+    if (!format_tuple.compressed) {
+        // Only pre-create the texture for non-compressed textures.
+        switch (params.target) {
+        case SurfaceParams::SurfaceTarget::Texture1D:
+            glTexImage1D(SurfaceTargetToGL(params.target), 0, format_tuple.internal_format,
+                         rect.GetWidth(), 0, format_tuple.format, format_tuple.type, nullptr);
+            break;
+        case SurfaceParams::SurfaceTarget::Texture2D:
+            glTexImage2D(SurfaceTargetToGL(params.target), 0, format_tuple.internal_format,
+                         rect.GetWidth(), rect.GetHeight(), 0, format_tuple.format,
+                         format_tuple.type, nullptr);
+            break;
+        case SurfaceParams::SurfaceTarget::Texture3D:
+        case SurfaceParams::SurfaceTarget::Texture2DArray:
+            glTexImage3D(SurfaceTargetToGL(params.target), 0, format_tuple.internal_format,
+                         rect.GetWidth(), rect.GetHeight(), params.depth, 0, format_tuple.format,
+                         format_tuple.type, nullptr);
+            break;
+        default:
+            LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
+                         static_cast<u32>(params.target));
+            UNREACHABLE();
+            glTexImage2D(GL_TEXTURE_2D, 0, format_tuple.internal_format, rect.GetWidth(),
+                         rect.GetHeight(), 0, format_tuple.format, format_tuple.type, nullptr);
+        }
+    }
+
+    glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_MAX_LEVEL, 0);
+    glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_MIN_FILTER, GL_LINEAR);
+    glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
+    glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
 }
 
 static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height) {
@@ -514,23 +558,6 @@ static void ConvertFormatAsNeeded_LoadGLBuffer(std::vector<u8>& data, PixelForma
     }
 }
 
-/**
- * Helper function to perform software conversion (as needed) when flushing a buffer to Switch
- * memory. This is for Maxwell pixel formats that cannot be represented as-is in OpenGL or with
- * typical desktop GPUs.
- */
-static void ConvertFormatAsNeeded_FlushGLBuffer(std::vector<u8>& /*data*/, PixelFormat pixel_format,
-                                                u32 /*width*/, u32 /*height*/) {
-    switch (pixel_format) {
-    case PixelFormat::ASTC_2D_4X4:
-    case PixelFormat::S8Z24:
-        LOG_CRITICAL(Render_OpenGL, "Unimplemented pixel_format={}",
-                     static_cast<u32>(pixel_format));
-        UNREACHABLE();
-        break;
-    }
-}
-
 MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 64, 192));
 void CachedSurface::LoadGLBuffer() {
     ASSERT(params.type != SurfaceType::Fill);
@@ -545,13 +572,24 @@ void CachedSurface::LoadGLBuffer() {
     MICROPROFILE_SCOPE(OpenGL_SurfaceLoad);
 
     if (params.is_tiled) {
-        gl_buffer.resize(copy_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;
+        default:
+            LOG_CRITICAL(HW_GPU, "Unimplemented tiled load for target={}",
+                         static_cast<u32>(params.target));
+            UNREACHABLE();
+        }
 
+        gl_buffer.resize(params.depth * copy_size);
         morton_to_gl_fns[static_cast<size_t>(params.pixel_format)](
-            params.width, params.block_height, params.height, gl_buffer, params.addr);
+            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 + copy_size;
-
+        const u8* const texture_src_data_end{texture_src_data + (params.depth * copy_size)};
         gl_buffer.assign(texture_src_data, texture_src_data_end);
     }
 
@@ -560,23 +598,7 @@ void CachedSurface::LoadGLBuffer() {
 
 MICROPROFILE_DEFINE(OpenGL_SurfaceFlush, "OpenGL", "Surface Flush", MP_RGB(128, 192, 64));
 void CachedSurface::FlushGLBuffer() {
-    u8* const dst_buffer = Memory::GetPointer(params.addr);
-
-    ASSERT(dst_buffer);
-    ASSERT(gl_buffer.size() ==
-           params.width * params.height * GetGLBytesPerPixel(params.pixel_format));
-
-    MICROPROFILE_SCOPE(OpenGL_SurfaceFlush);
-
-    ConvertFormatAsNeeded_FlushGLBuffer(gl_buffer, params.pixel_format, params.width,
-                                        params.height);
-
-    if (!params.is_tiled) {
-        std::memcpy(dst_buffer, gl_buffer.data(), params.size_in_bytes);
-    } else {
-        gl_to_morton_fns[static_cast<size_t>(params.pixel_format)](
-            params.width, params.block_height, params.height, gl_buffer, params.addr);
-    }
+    ASSERT_MSG(false, "Unimplemented");
 }
 
 MICROPROFILE_DEFINE(OpenGL_TextureUL, "OpenGL", "Texture Upload", MP_RGB(128, 64, 192));
@@ -587,7 +609,7 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
     MICROPROFILE_SCOPE(OpenGL_TextureUL);
 
     ASSERT(gl_buffer.size() ==
-           params.width * params.height * GetGLBytesPerPixel(params.pixel_format));
+           params.width * params.height * GetGLBytesPerPixel(params.pixel_format) * params.depth);
 
     const auto& rect{params.GetRect()};
 
@@ -600,8 +622,13 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
     GLuint target_tex = texture.handle;
     OpenGLState cur_state = OpenGLState::GetCurState();
 
-    GLuint old_tex = cur_state.texture_units[0].texture_2d;
-    cur_state.texture_units[0].texture_2d = target_tex;
+    const auto& old_tex = cur_state.texture_units[0];
+    SCOPE_EXIT({
+        cur_state.texture_units[0] = old_tex;
+        cur_state.Apply();
+    });
+    cur_state.texture_units[0].texture = target_tex;
+    cur_state.texture_units[0].target = SurfaceTargetToGL(params.target);
     cur_state.Apply();
 
     // Ensure no bad interactions with GL_UNPACK_ALIGNMENT
@@ -610,74 +637,68 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
 
     glActiveTexture(GL_TEXTURE0);
     if (tuple.compressed) {
-        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]);
+        switch (params.target) {
+        case SurfaceParams::SurfaceTarget::Texture2D:
+            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]);
+            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]);
+            break;
+        default:
+            LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
+                         static_cast<u32>(params.target));
+            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]);
+        }
     } else {
-        glTexSubImage2D(GL_TEXTURE_2D, 0, x0, y0, static_cast<GLsizei>(rect.GetWidth()),
-                        static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
-                        &gl_buffer[buffer_offset]);
-    }
-
-    glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
-
-    cur_state.texture_units[0].texture_2d = old_tex;
-    cur_state.Apply();
-}
-
-MICROPROFILE_DEFINE(OpenGL_TextureDL, "OpenGL", "Texture Download", MP_RGB(128, 192, 64));
-void CachedSurface::DownloadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle) {
-    if (params.type == SurfaceType::Fill)
-        return;
-
-    MICROPROFILE_SCOPE(OpenGL_TextureDL);
-
-    gl_buffer.resize(params.width * params.height * GetGLBytesPerPixel(params.pixel_format));
-
-    OpenGLState state = OpenGLState::GetCurState();
-    OpenGLState prev_state = state;
-    SCOPE_EXIT({ prev_state.Apply(); });
 
-    const FormatTuple& tuple = GetFormatTuple(params.pixel_format, params.component_type);
-
-    // Ensure no bad interactions with GL_PACK_ALIGNMENT
-    ASSERT(params.width * GetGLBytesPerPixel(params.pixel_format) % 4 == 0);
-    glPixelStorei(GL_PACK_ROW_LENGTH, static_cast<GLint>(params.width));
-
-    const auto& rect{params.GetRect()};
-    size_t buffer_offset =
-        (rect.bottom * params.width + rect.left) * GetGLBytesPerPixel(params.pixel_format);
-
-    state.UnbindTexture(texture.handle);
-    state.draw.read_framebuffer = read_fb_handle;
-    state.Apply();
-
-    if (params.type == SurfaceType::ColorTexture) {
-        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D,
-                               texture.handle, 0);
-        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
-                               0);
-    } else if (params.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,
-                               texture.handle, 0);
-        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
-    } else {
-        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
-        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
-                               texture.handle, 0);
+        switch (params.target) {
+        case SurfaceParams::SurfaceTarget::Texture1D:
+            glTexSubImage1D(SurfaceTargetToGL(params.target), 0, x0,
+                            static_cast<GLsizei>(rect.GetWidth()), tuple.format, tuple.type,
+                            &gl_buffer[buffer_offset]);
+            break;
+        case SurfaceParams::SurfaceTarget::Texture2D:
+            glTexSubImage2D(SurfaceTargetToGL(params.target), 0, x0, y0,
+                            static_cast<GLsizei>(rect.GetWidth()),
+                            static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
+                            &gl_buffer[buffer_offset]);
+            break;
+        case SurfaceParams::SurfaceTarget::Texture3D:
+        case SurfaceParams::SurfaceTarget::Texture2DArray:
+            glTexSubImage3D(SurfaceTargetToGL(params.target), 0, x0, y0, 0,
+                            static_cast<GLsizei>(rect.GetWidth()),
+                            static_cast<GLsizei>(rect.GetHeight()), params.depth, tuple.format,
+                            tuple.type, &gl_buffer[buffer_offset]);
+            break;
+        default:
+            LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
+                         static_cast<u32>(params.target));
+            UNREACHABLE();
+            glTexSubImage2D(GL_TEXTURE_2D, 0, x0, y0, static_cast<GLsizei>(rect.GetWidth()),
+                            static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
+                            &gl_buffer[buffer_offset]);
+        }
     }
-    glReadPixels(static_cast<GLint>(rect.left), static_cast<GLint>(rect.bottom),
-                 static_cast<GLsizei>(rect.GetWidth()), static_cast<GLsizei>(rect.GetHeight()),
-                 tuple.format, tuple.type, &gl_buffer[buffer_offset]);
 
-    glPixelStorei(GL_PACK_ROW_LENGTH, 0);
+    glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
 }
 
 RasterizerCacheOpenGL::RasterizerCacheOpenGL() {
     read_framebuffer.Create();
     draw_framebuffer.Create();
+    copy_pbo.Create();
 }
 
 Surface RasterizerCacheOpenGL::GetTextureSurface(const Tegra::Texture::FullTextureInfo& config) {
@@ -748,7 +769,6 @@ void RasterizerCacheOpenGL::LoadSurface(const Surface& surface) {
 }
 
 void RasterizerCacheOpenGL::FlushSurface(const Surface& surface) {
-    surface->DownloadGLTexture(read_framebuffer.handle, draw_framebuffer.handle);
     surface->FlushGLBuffer();
 }
 
@@ -809,8 +829,8 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& surface,
     // If format is unchanged, we can do a faster blit without reinterpreting pixel data
     if (params.pixel_format == new_params.pixel_format) {
         BlitTextures(surface->Texture().handle, params.GetRect(), new_surface->Texture().handle,
-                     new_surface->GetSurfaceParams().GetRect(), params.type,
-                     read_framebuffer.handle, draw_framebuffer.handle);
+                     params.GetRect(), params.type, read_framebuffer.handle,
+                     draw_framebuffer.handle);
         return new_surface;
     }
 
@@ -821,12 +841,7 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& surface,
 
         size_t buffer_size = std::max(params.SizeInBytes(), new_params.SizeInBytes());
 
-        // Use a Pixel Buffer Object to download the previous texture and then upload it to the new
-        // one using the new format.
-        OGLBuffer pbo;
-        pbo.Create();
-
-        glBindBuffer(GL_PIXEL_PACK_BUFFER, pbo.handle);
+        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,
@@ -845,8 +860,8 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& surface,
                 // 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_CRITICAL(HW_GPU, "Trying to upload extra texture data from the CPU during "
-                                     "reinterpretation but the texture is tiled.");
+                LOG_DEBUG(HW_GPU, "Trying to upload extra texture data from the CPU during "
+                                  "reinterpretation but the texture is tiled.");
             }
             size_t remaining_size = new_params.SizeInBytes() - params.SizeInBytes();
             std::vector<u8> data(remaining_size);
@@ -859,21 +874,38 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& surface,
 
         const auto& dest_rect{new_params.GetRect()};
 
-        glBindBuffer(GL_PIXEL_UNPACK_BUFFER, pbo.handle);
+        glBindBuffer(GL_PIXEL_UNPACK_BUFFER, copy_pbo.handle);
         if (dest_format.compressed) {
-            glCompressedTexSubImage2D(
-                GL_TEXTURE_2D, 0, 0, 0, static_cast<GLsizei>(dest_rect.GetWidth()),
-                static_cast<GLsizei>(dest_rect.GetHeight()), dest_format.format,
-                static_cast<GLsizei>(new_params.SizeInBytes()), nullptr);
+            LOG_CRITICAL(HW_GPU, "Compressed copy is unimplemented!");
+            UNREACHABLE();
         } else {
-            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);
+            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();
+            }
         }
         glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
-
-        pbo.Release();
     }
 
     return new_surface;