From 80de01a5b4a7f57ec7850079fbd38fac76b9d08f Mon Sep 17 00:00:00 2001 From: Liam Date: Wed, 3 Jan 2024 22:46:59 -0500 Subject: video_core: simplify accelerated surface fetch and crop handling between APIs --- src/video_core/renderer_opengl/gl_fsr.cpp | 21 +- src/video_core/renderer_opengl/gl_fsr.h | 2 +- src/video_core/renderer_opengl/gl_rasterizer.cpp | 28 ++- src/video_core/renderer_opengl/gl_rasterizer.h | 13 +- src/video_core/renderer_opengl/renderer_opengl.cpp | 235 +++++++++------------ src/video_core/renderer_opengl/renderer_opengl.h | 32 +-- 6 files changed, 142 insertions(+), 189 deletions(-) (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/renderer_opengl/gl_fsr.cpp b/src/video_core/renderer_opengl/gl_fsr.cpp index 77262dcf1..429dcdc6c 100644 --- a/src/video_core/renderer_opengl/gl_fsr.cpp +++ b/src/video_core/renderer_opengl/gl_fsr.cpp @@ -25,7 +25,7 @@ FSR::~FSR() = default; void FSR::Draw(ProgramManager& program_manager, const Common::Rectangle& screen, u32 input_image_width, u32 input_image_height, - const Common::Rectangle& crop_rect) { + const Common::Rectangle& crop_rect) { const auto output_image_width = screen.GetWidth(); const auto output_image_height = screen.GetHeight(); @@ -57,14 +57,19 @@ void FSR::Draw(ProgramManager& program_manager, const Common::Rectangle& sc glViewportIndexedf(0, 0.0f, 0.0f, static_cast(output_image_width), static_cast(output_image_height)); - FsrConstants constants; - FsrEasuConOffset( - constants.data() + 0, constants.data() + 4, constants.data() + 8, constants.data() + 12, + const f32 input_width = static_cast(input_image_width); + const f32 input_height = static_cast(input_image_height); + const f32 output_width = static_cast(screen.GetWidth()); + const f32 output_height = static_cast(screen.GetHeight()); + const f32 viewport_width = (crop_rect.right - crop_rect.left) * input_width; + const f32 viewport_x = crop_rect.left * input_width; + const f32 viewport_height = (crop_rect.bottom - crop_rect.top) * input_height; + const f32 viewport_y = crop_rect.top * input_height; - static_cast(crop_rect.GetWidth()), static_cast(crop_rect.GetHeight()), - static_cast(input_image_width), static_cast(input_image_height), - static_cast(output_image_width), static_cast(output_image_height), - static_cast(crop_rect.left), static_cast(crop_rect.top)); + FsrConstants constants; + FsrEasuConOffset(constants.data() + 0, constants.data() + 4, constants.data() + 8, + constants.data() + 12, viewport_width, viewport_height, input_width, + input_height, output_width, output_height, viewport_x, viewport_y); glProgramUniform4uiv(fsr_easu_frag.handle, 0, sizeof(constants), std::data(constants)); diff --git a/src/video_core/renderer_opengl/gl_fsr.h b/src/video_core/renderer_opengl/gl_fsr.h index 1f6ae3115..a5092e396 100644 --- a/src/video_core/renderer_opengl/gl_fsr.h +++ b/src/video_core/renderer_opengl/gl_fsr.h @@ -22,7 +22,7 @@ public: void Draw(ProgramManager& program_manager, const Common::Rectangle& screen, u32 input_image_width, u32 input_image_height, - const Common::Rectangle& crop_rect); + const Common::Rectangle& crop_rect); void InitBuffers(); diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index d5354ef2d..050a74cca 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -71,10 +71,10 @@ std::optional MaxwellToVideoCoreQuery(VideoCommon::QueryTy RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_, Tegra::MaxwellDeviceMemoryManager& device_memory_, - const Device& device_, ScreenInfo& screen_info_, - ProgramManager& program_manager_, StateTracker& state_tracker_) - : gpu(gpu_), device_memory(device_memory_), device(device_), screen_info(screen_info_), - program_manager(program_manager_), state_tracker(state_tracker_), + const Device& device_, ProgramManager& program_manager_, + StateTracker& state_tracker_) + : gpu(gpu_), device_memory(device_memory_), device(device_), program_manager(program_manager_), + state_tracker(state_tracker_), texture_cache_runtime(device, program_manager, state_tracker, staging_buffer_pool), texture_cache(texture_cache_runtime, device_memory_), buffer_cache_runtime(device, staging_buffer_pool), @@ -739,10 +739,10 @@ void RasterizerOpenGL::AccelerateInlineToMemory(GPUVAddr address, size_t copy_si query_cache.InvalidateRegion(*cpu_addr, copy_size); } -bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config, - DAddr framebuffer_addr, u32 pixel_stride) { +std::optional RasterizerOpenGL::AccelerateDisplay( + const Tegra::FramebufferConfig& config, DAddr framebuffer_addr, u32 pixel_stride) { if (framebuffer_addr == 0) { - return false; + return {}; } MICROPROFILE_SCOPE(OpenGL_CacheManagement); @@ -750,16 +750,14 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config, ImageView* const image_view{ texture_cache.TryFindFramebufferImageView(config, framebuffer_addr)}; if (!image_view) { - return false; + return {}; } - // Verify that the cached surface is the same size and format as the requested framebuffer - // ASSERT_MSG(image_view->size.width == config.width, "Framebuffer width is different"); - // ASSERT_MSG(image_view->size.height == config.height, "Framebuffer height is different"); - screen_info.texture.width = image_view->size.width; - screen_info.texture.height = image_view->size.height; - screen_info.display_texture = image_view->Handle(Shader::TextureType::Color2D); - return true; + FramebufferTextureInfo info{}; + info.display_texture = image_view->Handle(Shader::TextureType::Color2D); + info.width = image_view->size.width; + info.height = image_view->size.height; + return info; } void RasterizerOpenGL::SyncState() { diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index 34aa73526..ee82d9f3a 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -37,7 +37,7 @@ class MemoryManager; namespace OpenGL { -struct ScreenInfo; +struct FramebufferTextureInfo; struct ShaderEntries; struct BindlessSSBO { @@ -76,8 +76,8 @@ class RasterizerOpenGL : public VideoCore::RasterizerInterface, public: explicit RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_, Tegra::MaxwellDeviceMemoryManager& device_memory_, - const Device& device_, ScreenInfo& screen_info_, - ProgramManager& program_manager_, StateTracker& state_tracker_); + const Device& device_, ProgramManager& program_manager_, + StateTracker& state_tracker_); ~RasterizerOpenGL() override; void Draw(bool is_indexed, u32 instance_count) override; @@ -122,8 +122,6 @@ public: Tegra::Engines::AccelerateDMAInterface& AccessAccelerateDMA() override; void AccelerateInlineToMemory(GPUVAddr address, size_t copy_size, std::span memory) override; - bool AccelerateDisplay(const Tegra::FramebufferConfig& config, DAddr framebuffer_addr, - u32 pixel_stride) override; void LoadDiskResources(u64 title_id, std::stop_token stop_loading, const VideoCore::DiskResourceLoadCallback& callback) override; @@ -144,6 +142,10 @@ public: return true; } + std::optional AccelerateDisplay(const Tegra::FramebufferConfig& config, + VAddr framebuffer_addr, + u32 pixel_stride); + private: static constexpr size_t MAX_TEXTURES = 192; static constexpr size_t MAX_IMAGES = 48; @@ -237,7 +239,6 @@ private: Tegra::MaxwellDeviceMemoryManager& device_memory; const Device& device; - ScreenInfo& screen_info; ProgramManager& program_manager; StateTracker& state_tracker; diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index b75376fdb..ea5ed3e2f 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -148,8 +148,7 @@ RendererOpenGL::RendererOpenGL(Core::TelemetrySession& telemetry_session_, : RendererBase{emu_window_, std::move(context_)}, telemetry_session{telemetry_session_}, emu_window{emu_window_}, device_memory{device_memory_}, gpu{gpu_}, device{emu_window_}, state_tracker{}, program_manager{device}, - rasterizer(emu_window, gpu, device_memory, device, screen_info, program_manager, - state_tracker) { + rasterizer(emu_window, gpu, device_memory, device, program_manager, state_tracker) { if (Settings::values.renderer_debug && GLAD_GL_KHR_debug) { glEnable(GL_DEBUG_OUTPUT); glEnable(GL_DEBUG_OUTPUT_SYNCHRONOUS); @@ -184,11 +183,11 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { if (!framebuffer) { return; } - PrepareRendertarget(framebuffer); - RenderScreenshot(); + + RenderScreenshot(*framebuffer); state_tracker.BindFramebuffer(0); - DrawScreen(emu_window.GetFramebufferLayout()); + DrawScreen(*framebuffer, emu_window.GetFramebufferLayout()); ++m_current_frame; @@ -199,41 +198,37 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { render_window.OnFrameDisplayed(); } -void RendererOpenGL::PrepareRendertarget(const Tegra::FramebufferConfig* framebuffer) { - if (!framebuffer) { - return; - } +FramebufferTextureInfo RendererOpenGL::PrepareRenderTarget( + const Tegra::FramebufferConfig& framebuffer) { // If framebuffer is provided, reload it from memory to a texture - if (screen_info.texture.width != static_cast(framebuffer->width) || - screen_info.texture.height != static_cast(framebuffer->height) || - screen_info.texture.pixel_format != framebuffer->pixel_format || + if (framebuffer_texture.width != static_cast(framebuffer.width) || + framebuffer_texture.height != static_cast(framebuffer.height) || + framebuffer_texture.pixel_format != framebuffer.pixel_format || gl_framebuffer_data.empty()) { // Reallocate texture if the framebuffer size has changed. // This is expected to not happen very often and hence should not be a // performance problem. - ConfigureFramebufferTexture(screen_info.texture, *framebuffer); + ConfigureFramebufferTexture(framebuffer); } - // Load the framebuffer from memory, draw it to the screen, and swap buffers - LoadFBToScreenInfo(*framebuffer); + // Load the framebuffer from memory if needed + return LoadFBToScreenInfo(framebuffer); } -void RendererOpenGL::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer) { - // Framebuffer orientation handling - framebuffer_transform_flags = framebuffer.transform_flags; - framebuffer_crop_rect = framebuffer.crop_rect; - framebuffer_width = framebuffer.width; - framebuffer_height = framebuffer.height; - +FramebufferTextureInfo RendererOpenGL::LoadFBToScreenInfo( + const Tegra::FramebufferConfig& framebuffer) { const VAddr framebuffer_addr{framebuffer.address + framebuffer.offset}; - screen_info.was_accelerated = + const auto accelerated_info = rasterizer.AccelerateDisplay(framebuffer, framebuffer_addr, framebuffer.stride); - if (screen_info.was_accelerated) { - return; + if (accelerated_info) { + return *accelerated_info; } // Reset the screen info's display texture to its own permanent texture - screen_info.display_texture = screen_info.texture.resource.handle; + FramebufferTextureInfo info{}; + info.display_texture = framebuffer_texture.resource.handle; + info.width = framebuffer.width; + info.height = framebuffer.height; // TODO(Rodrigo): Read this from HLE constexpr u32 block_height_log2 = 4; @@ -256,17 +251,13 @@ void RendererOpenGL::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuf // they differ from the LCD resolution. // TODO: Applications could theoretically crash yuzu here by specifying too large // framebuffer sizes. We should make sure that this cannot happen. - glTextureSubImage2D(screen_info.texture.resource.handle, 0, 0, 0, framebuffer.width, - framebuffer.height, screen_info.texture.gl_format, - screen_info.texture.gl_type, gl_framebuffer_data.data()); + glTextureSubImage2D(framebuffer_texture.resource.handle, 0, 0, 0, framebuffer.width, + framebuffer.height, framebuffer_texture.gl_format, + framebuffer_texture.gl_type, gl_framebuffer_data.data()); glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); -} -void RendererOpenGL::LoadColorToActiveGLTexture(u8 color_r, u8 color_g, u8 color_b, u8 color_a, - const TextureInfo& texture) { - const u8 framebuffer_data[4] = {color_a, color_b, color_g, color_r}; - glClearTexImage(texture.resource.handle, 0, GL_RGBA, GL_UNSIGNED_BYTE, framebuffer_data); + return info; } void RendererOpenGL::InitOpenGLObjects() { @@ -343,15 +334,15 @@ void RendererOpenGL::InitOpenGLObjects() { glNamedBufferData(vertex_buffer.handle, sizeof(ScreenRectVertex) * 4, nullptr, GL_STREAM_DRAW); // Allocate textures for the screen - screen_info.texture.resource.Create(GL_TEXTURE_2D); + framebuffer_texture.resource.Create(GL_TEXTURE_2D); - const GLuint texture = screen_info.texture.resource.handle; + const GLuint texture = framebuffer_texture.resource.handle; glTextureStorage2D(texture, 1, GL_RGBA8, 1, 1); - screen_info.display_texture = screen_info.texture.resource.handle; - // Clear screen to black - LoadColorToActiveGLTexture(0, 0, 0, 0, screen_info.texture); + const u8 framebuffer_data[4] = {0, 0, 0, 0}; + glClearTexImage(framebuffer_texture.resource.handle, 0, GL_RGBA, GL_UNSIGNED_BYTE, + framebuffer_data); aa_framebuffer.Create(); @@ -380,60 +371,65 @@ void RendererOpenGL::AddTelemetryFields() { telemetry_session.AddField(user_system, "GPU_OpenGL_Version", std::string(gl_version)); } -void RendererOpenGL::ConfigureFramebufferTexture(TextureInfo& texture, - const Tegra::FramebufferConfig& framebuffer) { - texture.width = framebuffer.width; - texture.height = framebuffer.height; - texture.pixel_format = framebuffer.pixel_format; +void RendererOpenGL::ConfigureFramebufferTexture(const Tegra::FramebufferConfig& framebuffer) { + framebuffer_texture.width = framebuffer.width; + framebuffer_texture.height = framebuffer.height; + framebuffer_texture.pixel_format = framebuffer.pixel_format; const auto pixel_format{ VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format)}; const u32 bytes_per_pixel{VideoCore::Surface::BytesPerBlock(pixel_format)}; - gl_framebuffer_data.resize(texture.width * texture.height * bytes_per_pixel); + gl_framebuffer_data.resize(framebuffer_texture.width * framebuffer_texture.height * + bytes_per_pixel); GLint internal_format; switch (framebuffer.pixel_format) { case Service::android::PixelFormat::Rgba8888: internal_format = GL_RGBA8; - texture.gl_format = GL_RGBA; - texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; + framebuffer_texture.gl_format = GL_RGBA; + framebuffer_texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; break; case Service::android::PixelFormat::Rgb565: internal_format = GL_RGB565; - texture.gl_format = GL_RGB; - texture.gl_type = GL_UNSIGNED_SHORT_5_6_5; + framebuffer_texture.gl_format = GL_RGB; + framebuffer_texture.gl_type = GL_UNSIGNED_SHORT_5_6_5; break; default: internal_format = GL_RGBA8; - texture.gl_format = GL_RGBA; - texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; + framebuffer_texture.gl_format = GL_RGBA; + framebuffer_texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; // UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}", // static_cast(framebuffer.pixel_format)); break; } - texture.resource.Release(); - texture.resource.Create(GL_TEXTURE_2D); - glTextureStorage2D(texture.resource.handle, 1, internal_format, texture.width, texture.height); + framebuffer_texture.resource.Release(); + framebuffer_texture.resource.Create(GL_TEXTURE_2D); + glTextureStorage2D(framebuffer_texture.resource.handle, 1, internal_format, + framebuffer_texture.width, framebuffer_texture.height); aa_texture.Release(); aa_texture.Create(GL_TEXTURE_2D); glTextureStorage2D(aa_texture.handle, 1, GL_RGBA16F, - Settings::values.resolution_info.ScaleUp(screen_info.texture.width), - Settings::values.resolution_info.ScaleUp(screen_info.texture.height)); + Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), + Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, aa_texture.handle, 0); smaa_edges_tex.Release(); smaa_edges_tex.Create(GL_TEXTURE_2D); glTextureStorage2D(smaa_edges_tex.handle, 1, GL_RG16F, - Settings::values.resolution_info.ScaleUp(screen_info.texture.width), - Settings::values.resolution_info.ScaleUp(screen_info.texture.height)); + Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), + Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); smaa_blend_tex.Release(); smaa_blend_tex.Create(GL_TEXTURE_2D); glTextureStorage2D(smaa_blend_tex.handle, 1, GL_RGBA16F, - Settings::values.resolution_info.ScaleUp(screen_info.texture.width), - Settings::values.resolution_info.ScaleUp(screen_info.texture.height)); + Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), + Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); } -void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { +void RendererOpenGL::DrawScreen(const Tegra::FramebufferConfig& framebuffer, + const Layout::FramebufferLayout& layout) { + FramebufferTextureInfo info = PrepareRenderTarget(framebuffer); + const auto crop = Tegra::NormalizeCrop(framebuffer, info.width, info.height); + // TODO: Signal state tracker about these changes state_tracker.NotifyScreenDrawVertexArray(); state_tracker.NotifyPolygonModes(); @@ -469,7 +465,7 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { glColorMaski(0, GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); glDepthRangeIndexed(0, 0.0, 0.0); - glBindTextureUnit(0, screen_info.display_texture); + glBindTextureUnit(0, info.display_texture); auto anti_aliasing = Settings::values.anti_aliasing.GetValue(); if (anti_aliasing >= Settings::AntiAliasing::MaxEnum) { @@ -480,22 +476,22 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { if (anti_aliasing != Settings::AntiAliasing::None) { glEnablei(GL_SCISSOR_TEST, 0); - auto viewport_width = screen_info.texture.width; - auto scissor_width = framebuffer_crop_rect.GetWidth(); + auto viewport_width = info.width; + auto scissor_width = static_cast(crop.GetWidth()); if (scissor_width <= 0) { scissor_width = viewport_width; } - auto viewport_height = screen_info.texture.height; - auto scissor_height = framebuffer_crop_rect.GetHeight(); + auto viewport_height = info.height; + auto scissor_height = static_cast(crop.GetHeight()); if (scissor_height <= 0) { scissor_height = viewport_height; } - if (screen_info.was_accelerated) { - viewport_width = Settings::values.resolution_info.ScaleUp(viewport_width); - scissor_width = Settings::values.resolution_info.ScaleUp(scissor_width); - viewport_height = Settings::values.resolution_info.ScaleUp(viewport_height); - scissor_height = Settings::values.resolution_info.ScaleUp(scissor_height); - } + + viewport_width = Settings::values.resolution_info.ScaleUp(viewport_width); + scissor_width = Settings::values.resolution_info.ScaleUp(scissor_width); + viewport_height = Settings::values.resolution_info.ScaleUp(viewport_height); + scissor_height = Settings::values.resolution_info.ScaleUp(scissor_height); + glScissorIndexed(0, 0, 0, scissor_width, scissor_height); glViewportIndexedf(0, 0.0f, 0.0f, static_cast(viewport_width), static_cast(viewport_height)); @@ -536,7 +532,7 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { smaa_blending_weight_calculation_frag.handle); glDrawArrays(GL_TRIANGLES, 0, 3); - glBindTextureUnit(0, screen_info.display_texture); + glBindTextureUnit(0, info.display_texture); glBindTextureUnit(1, smaa_blend_tex.handle); glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, aa_texture.handle, 0); @@ -561,18 +557,10 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { fsr->InitBuffers(); } - auto crop_rect = framebuffer_crop_rect; - if (crop_rect.GetWidth() == 0) { - crop_rect.right = framebuffer_width; - } - if (crop_rect.GetHeight() == 0) { - crop_rect.bottom = framebuffer_height; - } - crop_rect = crop_rect.Scale(Settings::values.resolution_info.up_factor); - const auto fsr_input_width = Settings::values.resolution_info.ScaleUp(framebuffer_width); - const auto fsr_input_height = Settings::values.resolution_info.ScaleUp(framebuffer_height); + const auto fsr_input_width = Settings::values.resolution_info.ScaleUp(info.width); + const auto fsr_input_height = Settings::values.resolution_info.ScaleUp(info.height); glBindSampler(0, present_sampler.handle); - fsr->Draw(program_manager, layout.screen, fsr_input_width, fsr_input_height, crop_rect); + fsr->Draw(program_manager, layout.screen, fsr_input_width, fsr_input_height, crop); } else { if (fsr->AreBuffersInitialized()) { fsr->ReleaseBuffers(); @@ -603,61 +591,34 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { glProgramUniformMatrix3x2fv(present_vertex.handle, ModelViewMatrixLocation, 1, GL_FALSE, ortho_matrix.data()); - const auto& texcoords = screen_info.display_texcoords; - auto left = texcoords.left; - auto right = texcoords.right; - if (framebuffer_transform_flags != Service::android::BufferTransformFlags::Unset) { - if (framebuffer_transform_flags == Service::android::BufferTransformFlags::FlipV) { - // Flip the framebuffer vertically - left = texcoords.right; - right = texcoords.left; - } else { - // Other transformations are unsupported - LOG_CRITICAL(Render_OpenGL, "Unsupported framebuffer_transform_flags={}", - framebuffer_transform_flags); - UNIMPLEMENTED(); - } - } - - ASSERT_MSG(framebuffer_crop_rect.left == 0, "Unimplemented"); - - f32 left_start{}; - if (framebuffer_crop_rect.Top() > 0) { - left_start = static_cast(framebuffer_crop_rect.Top()) / - static_cast(framebuffer_crop_rect.Bottom()); - } - f32 scale_u = static_cast(framebuffer_width) / static_cast(screen_info.texture.width); - f32 scale_v = - static_cast(framebuffer_height) / static_cast(screen_info.texture.height); - - if (Settings::values.scaling_filter.GetValue() != Settings::ScalingFilter::Fsr) { - // Scale the output by the crop width/height. This is commonly used with 1280x720 rendering - // (e.g. handheld mode) on a 1920x1080 framebuffer. - if (framebuffer_crop_rect.GetWidth() > 0) { - scale_u = static_cast(framebuffer_crop_rect.GetWidth()) / - static_cast(screen_info.texture.width); - } - if (framebuffer_crop_rect.GetHeight() > 0) { - scale_v = static_cast(framebuffer_crop_rect.GetHeight()) / - static_cast(screen_info.texture.height); - } - } - if (Settings::values.anti_aliasing.GetValue() == Settings::AntiAliasing::Fxaa && - !screen_info.was_accelerated) { - scale_u /= Settings::values.resolution_info.up_factor; - scale_v /= Settings::values.resolution_info.up_factor; + f32 left, top, right, bottom; + if (Settings::values.scaling_filter.GetValue() == Settings::ScalingFilter::Fsr) { + // FSR has already applied the crop, so we just want to render the image + // it has produced. + left = 0; + top = 0; + right = 1; + bottom = 1; + } else { + // Apply the precomputed crop. + left = crop.left; + top = crop.top; + right = crop.right; + bottom = crop.bottom; } + // Map the coordinates to the screen. const auto& screen = layout.screen; + const auto x = screen.left; + const auto y = screen.top; + const auto w = screen.GetWidth(); + const auto h = screen.GetHeight(); + const std::array vertices = { - ScreenRectVertex(screen.left, screen.top, texcoords.top * scale_u, - left_start + left * scale_v), - ScreenRectVertex(screen.right, screen.top, texcoords.bottom * scale_u, - left_start + left * scale_v), - ScreenRectVertex(screen.left, screen.bottom, texcoords.top * scale_u, - left_start + right * scale_v), - ScreenRectVertex(screen.right, screen.bottom, texcoords.bottom * scale_u, - left_start + right * scale_v), + ScreenRectVertex(x, y, left, top), + ScreenRectVertex(x + w, y, right, top), + ScreenRectVertex(x, y + h, left, bottom), + ScreenRectVertex(x + w, y + h, right, bottom), }; glNamedBufferSubData(vertex_buffer.handle, 0, sizeof(vertices), std::data(vertices)); @@ -701,7 +662,7 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { // program_manager.RestoreGuestPipeline(); } -void RendererOpenGL::RenderScreenshot() { +void RendererOpenGL::RenderScreenshot(const Tegra::FramebufferConfig& framebuffer) { if (!renderer_settings.screenshot_requested) { return; } @@ -723,7 +684,7 @@ void RendererOpenGL::RenderScreenshot() { glRenderbufferStorage(GL_RENDERBUFFER, GL_SRGB8, layout.width, layout.height); glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, renderbuffer); - DrawScreen(layout); + DrawScreen(framebuffer, layout); glBindBuffer(GL_PIXEL_PACK_BUFFER, 0); glPixelStorei(GL_PACK_ROW_LENGTH, 0); diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h index 18699610a..cde8c5702 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.h +++ b/src/video_core/renderer_opengl/renderer_opengl.h @@ -50,11 +50,10 @@ struct TextureInfo { }; /// Structure used for storing information about the display target for the Switch screen -struct ScreenInfo { +struct FramebufferTextureInfo { GLuint display_texture{}; - bool was_accelerated = false; - const Common::Rectangle display_texcoords{0.0f, 0.0f, 1.0f, 1.0f}; - TextureInfo texture; + u32 width; + u32 height; }; class RendererOpenGL final : public VideoCore::RendererBase { @@ -81,23 +80,18 @@ private: void AddTelemetryFields(); - void ConfigureFramebufferTexture(TextureInfo& texture, - const Tegra::FramebufferConfig& framebuffer); + void ConfigureFramebufferTexture(const Tegra::FramebufferConfig& framebuffer); /// Draws the emulated screens to the emulator window. - void DrawScreen(const Layout::FramebufferLayout& layout); + void DrawScreen(const Tegra::FramebufferConfig& framebuffer, + const Layout::FramebufferLayout& layout); - void RenderScreenshot(); + void RenderScreenshot(const Tegra::FramebufferConfig& framebuffer); /// Loads framebuffer from emulated memory into the active OpenGL texture. - void LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer); + FramebufferTextureInfo LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer); - /// Fills active OpenGL texture with the given RGB color.Since the color is solid, the texture - /// can be 1x1 but will stretch across whatever it's rendered on. - void LoadColorToActiveGLTexture(u8 color_r, u8 color_g, u8 color_b, u8 color_a, - const TextureInfo& texture); - - void PrepareRendertarget(const Tegra::FramebufferConfig* framebuffer); + FramebufferTextureInfo PrepareRenderTarget(const Tegra::FramebufferConfig& framebuffer); Core::TelemetrySession& telemetry_session; Core::Frontend::EmuWindow& emu_window; @@ -126,7 +120,7 @@ private: GLuint64EXT vertex_buffer_address = 0; /// Display information for Switch screen - ScreenInfo screen_info; + TextureInfo framebuffer_texture; OGLTexture aa_texture; OGLFramebuffer aa_framebuffer; @@ -145,12 +139,6 @@ private: /// OpenGL framebuffer data std::vector gl_framebuffer_data; - - /// Used for transforming the framebuffer orientation - Service::android::BufferTransformFlags framebuffer_transform_flags{}; - Common::Rectangle framebuffer_crop_rect; - u32 framebuffer_width; - u32 framebuffer_height; }; } // namespace OpenGL -- cgit v1.2.3 From 453091f61100effba637950dc840da41d95be477 Mon Sep 17 00:00:00 2001 From: Liam Date: Fri, 26 Jan 2024 22:27:34 -0500 Subject: video_core: consistently account for resolution scaling when rendering --- src/video_core/renderer_opengl/gl_rasterizer.cpp | 8 +++++-- .../renderer_opengl/gl_texture_cache.cpp | 4 ++++ src/video_core/renderer_opengl/gl_texture_cache.h | 2 ++ src/video_core/renderer_opengl/renderer_opengl.cpp | 28 +++++++--------------- src/video_core/renderer_opengl/renderer_opengl.h | 2 ++ src/video_core/renderer_vulkan/vk_blit_screen.cpp | 6 +++-- src/video_core/renderer_vulkan/vk_blit_screen.h | 2 ++ src/video_core/renderer_vulkan/vk_rasterizer.cpp | 6 ++++- src/video_core/texture_cache/texture_cache.h | 9 +++---- src/video_core/texture_cache/texture_cache_base.h | 4 ++-- 10 files changed, 40 insertions(+), 31 deletions(-) (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 050a74cca..b42fb110c 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -747,16 +747,20 @@ std::optional RasterizerOpenGL::AccelerateDisplay( MICROPROFILE_SCOPE(OpenGL_CacheManagement); std::scoped_lock lock{texture_cache.mutex}; - ImageView* const image_view{ - texture_cache.TryFindFramebufferImageView(config, framebuffer_addr)}; + const auto [image_view, scaled] = + texture_cache.TryFindFramebufferImageView(config, framebuffer_addr); if (!image_view) { return {}; } + const auto& resolution = Settings::values.resolution_info; + FramebufferTextureInfo info{}; info.display_texture = image_view->Handle(Shader::TextureType::Color2D); info.width = image_view->size.width; info.height = image_view->size.height; + info.scaled_width = scaled ? resolution.ScaleUp(info.width) : info.width; + info.scaled_height = scaled ? resolution.ScaleUp(info.height) : info.height; return info; } diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp index 66a5ca03e..be14494ca 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.cpp +++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp @@ -1051,6 +1051,10 @@ void Image::Scale(bool up_scale) { state_tracker.NotifyScissor0(); } +bool Image::IsRescaled() const { + return True(flags & ImageFlagBits::Rescaled); +} + bool Image::ScaleUp(bool ignore) { const auto& resolution = runtime->resolution; if (!resolution.active) { diff --git a/src/video_core/renderer_opengl/gl_texture_cache.h b/src/video_core/renderer_opengl/gl_texture_cache.h index 34870c81f..3e54edcc2 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.h +++ b/src/video_core/renderer_opengl/gl_texture_cache.h @@ -217,6 +217,8 @@ public: return gl_type; } + bool IsRescaled() const; + bool ScaleUp(bool ignore = false); bool ScaleDown(bool ignore = false); diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index ea5ed3e2f..2b9ebff92 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -229,6 +229,8 @@ FramebufferTextureInfo RendererOpenGL::LoadFBToScreenInfo( info.display_texture = framebuffer_texture.resource.handle; info.width = framebuffer.width; info.height = framebuffer.height; + info.scaled_width = framebuffer.width; + info.scaled_height = framebuffer.height; // TODO(Rodrigo): Read this from HLE constexpr u32 block_height_log2 = 4; @@ -476,25 +478,13 @@ void RendererOpenGL::DrawScreen(const Tegra::FramebufferConfig& framebuffer, if (anti_aliasing != Settings::AntiAliasing::None) { glEnablei(GL_SCISSOR_TEST, 0); - auto viewport_width = info.width; - auto scissor_width = static_cast(crop.GetWidth()); - if (scissor_width <= 0) { - scissor_width = viewport_width; - } - auto viewport_height = info.height; - auto scissor_height = static_cast(crop.GetHeight()); - if (scissor_height <= 0) { - scissor_height = viewport_height; - } - - viewport_width = Settings::values.resolution_info.ScaleUp(viewport_width); - scissor_width = Settings::values.resolution_info.ScaleUp(scissor_width); - viewport_height = Settings::values.resolution_info.ScaleUp(viewport_height); - scissor_height = Settings::values.resolution_info.ScaleUp(scissor_height); + auto scissor_width = Settings::values.resolution_info.ScaleUp(framebuffer_texture.width); + auto viewport_width = static_cast(scissor_width); + auto scissor_height = Settings::values.resolution_info.ScaleUp(framebuffer_texture.height); + auto viewport_height = static_cast(scissor_height); glScissorIndexed(0, 0, 0, scissor_width, scissor_height); - glViewportIndexedf(0, 0.0f, 0.0f, static_cast(viewport_width), - static_cast(viewport_height)); + glViewportIndexedf(0, 0.0f, 0.0f, viewport_width, viewport_height); glBindSampler(0, present_sampler.handle); GLint old_read_fb; @@ -557,10 +547,8 @@ void RendererOpenGL::DrawScreen(const Tegra::FramebufferConfig& framebuffer, fsr->InitBuffers(); } - const auto fsr_input_width = Settings::values.resolution_info.ScaleUp(info.width); - const auto fsr_input_height = Settings::values.resolution_info.ScaleUp(info.height); glBindSampler(0, present_sampler.handle); - fsr->Draw(program_manager, layout.screen, fsr_input_width, fsr_input_height, crop); + fsr->Draw(program_manager, layout.screen, info.scaled_width, info.scaled_height, crop); } else { if (fsr->AreBuffersInitialized()) { fsr->ReleaseBuffers(); diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h index cde8c5702..3a83a9b78 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.h +++ b/src/video_core/renderer_opengl/renderer_opengl.h @@ -54,6 +54,8 @@ struct FramebufferTextureInfo { GLuint display_texture{}; u32 width; u32 height; + u32 scaled_width; + u32 scaled_height; }; class RendererOpenGL final : public VideoCore::RendererBase { diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.cpp b/src/video_core/renderer_vulkan/vk_blit_screen.cpp index c21a9c8fe..24781860b 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.cpp +++ b/src/video_core/renderer_vulkan/vk_blit_screen.cpp @@ -152,6 +152,8 @@ void BlitScreen::Draw(RasterizerVulkan& rasterizer, const Tegra::FramebufferConf framebuffer, framebuffer.address + framebuffer.offset, framebuffer.stride); const u32 texture_width = texture_info ? texture_info->width : framebuffer.width; const u32 texture_height = texture_info ? texture_info->height : framebuffer.height; + const u32 scaled_width = texture_info ? texture_info->scaled_width : texture_width; + const u32 scaled_height = texture_info ? texture_info->scaled_height : texture_height; const bool use_accelerated = texture_info.has_value(); RefreshResources(framebuffer); @@ -363,8 +365,8 @@ void BlitScreen::Draw(RasterizerVulkan& rasterizer, const Tegra::FramebufferConf if (fsr) { const auto crop_rect = Tegra::NormalizeCrop(framebuffer, texture_width, texture_height); const VkExtent2D fsr_input_size{ - .width = Settings::values.resolution_info.ScaleUp(texture_width), - .height = Settings::values.resolution_info.ScaleUp(texture_height), + .width = scaled_width, + .height = scaled_height, }; VkImageView fsr_image_view = fsr->Draw(scheduler, image_index, source_image_view, fsr_input_size, crop_rect); diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.h b/src/video_core/renderer_vulkan/vk_blit_screen.h index 40338886a..56ac47f08 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.h +++ b/src/video_core/renderer_vulkan/vk_blit_screen.h @@ -47,6 +47,8 @@ struct FramebufferTextureInfo { VkImageView image_view{}; u32 width{}; u32 height{}; + u32 scaled_width{}; + u32 scaled_height{}; }; class BlitScreen { diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index e593d7225..aa0a027bb 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -788,18 +788,22 @@ std::optional RasterizerVulkan::AccelerateDisplay( return {}; } std::scoped_lock lock{texture_cache.mutex}; - ImageView* const image_view = + const auto [image_view, scaled] = texture_cache.TryFindFramebufferImageView(config, framebuffer_addr); if (!image_view) { return {}; } query_cache.NotifySegment(false); + const auto& resolution = Settings::values.resolution_info; + FramebufferTextureInfo info{}; info.image = image_view->ImageHandle(); info.image_view = image_view->Handle(Shader::TextureType::Color2D); info.width = image_view->size.width; info.height = image_view->size.height; + info.scaled_width = scaled ? resolution.ScaleUp(info.width) : info.width; + info.scaled_height = scaled ? resolution.ScaleUp(info.height) : info.height; return info; } diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h index a7400adfa..a20c956ff 100644 --- a/src/video_core/texture_cache/texture_cache.h +++ b/src/video_core/texture_cache/texture_cache.h @@ -713,12 +713,12 @@ bool TextureCache

::BlitImage(const Tegra::Engines::Fermi2D::Surface& dst, } template -typename P::ImageView* TextureCache

::TryFindFramebufferImageView( +std::pair TextureCache

::TryFindFramebufferImageView( const Tegra::FramebufferConfig& config, DAddr cpu_addr) { // TODO: Properly implement this const auto it = page_table.find(cpu_addr >> YUZU_PAGEBITS); if (it == page_table.end()) { - return nullptr; + return {}; } const auto& image_map_ids = it->second; boost::container::small_vector valid_image_ids; @@ -747,7 +747,8 @@ typename P::ImageView* TextureCache

::TryFindFramebufferImageView( const auto GetImageViewForFramebuffer = [&](ImageId image_id) { const ImageViewInfo info{ImageViewType::e2D, view_format}; - return &slot_image_views[FindOrEmplaceImageView(image_id, info)]; + return std::make_pair(&slot_image_views[FindOrEmplaceImageView(image_id, info)], + slot_images[image_id].IsRescaled()); }; if (valid_image_ids.size() == 1) [[likely]] { @@ -761,7 +762,7 @@ typename P::ImageView* TextureCache

::TryFindFramebufferImageView( return GetImageViewForFramebuffer(*most_recent); } - return nullptr; + return {}; } template diff --git a/src/video_core/texture_cache/texture_cache_base.h b/src/video_core/texture_cache/texture_cache_base.h index f9aebb293..e7b910121 100644 --- a/src/video_core/texture_cache/texture_cache_base.h +++ b/src/video_core/texture_cache/texture_cache_base.h @@ -212,8 +212,8 @@ public: const Tegra::Engines::Fermi2D::Config& copy); /// Try to find a cached image view in the given CPU address - [[nodiscard]] ImageView* TryFindFramebufferImageView(const Tegra::FramebufferConfig& config, - DAddr cpu_addr); + [[nodiscard]] std::pair TryFindFramebufferImageView( + const Tegra::FramebufferConfig& config, DAddr cpu_addr); /// Return true when there are uncommitted images to be downloaded [[nodiscard]] bool HasUncommittedFlushes() const noexcept; -- cgit v1.2.3 From 2b1dd3bef511806aa479ec93e3d9b414db80d4a9 Mon Sep 17 00:00:00 2001 From: Liam Date: Fri, 12 Jan 2024 00:46:17 -0500 Subject: renderer_opengl: isolate core presentation code --- src/video_core/CMakeLists.txt | 2 + src/video_core/renderer_opengl/gl_blit_screen.cpp | 519 +++++++++++++++++++++ src/video_core/renderer_opengl/gl_blit_screen.h | 110 +++++ src/video_core/renderer_opengl/gl_rasterizer.h | 1 + src/video_core/renderer_opengl/renderer_opengl.cpp | 506 +------------------- src/video_core/renderer_opengl/renderer_opengl.h | 78 +--- 6 files changed, 639 insertions(+), 577 deletions(-) create mode 100644 src/video_core/renderer_opengl/gl_blit_screen.cpp create mode 100644 src/video_core/renderer_opengl/gl_blit_screen.h (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index 36aa7bb66..c158970f2 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -118,6 +118,8 @@ add_library(video_core STATIC renderer_null/renderer_null.h renderer_opengl/blit_image.cpp renderer_opengl/blit_image.h + renderer_opengl/gl_blit_screen.cpp + renderer_opengl/gl_blit_screen.h renderer_opengl/gl_buffer_cache_base.cpp renderer_opengl/gl_buffer_cache.cpp renderer_opengl/gl_buffer_cache.h diff --git a/src/video_core/renderer_opengl/gl_blit_screen.cpp b/src/video_core/renderer_opengl/gl_blit_screen.cpp new file mode 100644 index 000000000..88757ba38 --- /dev/null +++ b/src/video_core/renderer_opengl/gl_blit_screen.cpp @@ -0,0 +1,519 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#include "video_core/framebuffer_config.h" +#include "video_core/host_shaders/ffx_a_h.h" +#include "video_core/host_shaders/ffx_fsr1_h.h" +#include "video_core/host_shaders/full_screen_triangle_vert.h" +#include "video_core/host_shaders/fxaa_frag.h" +#include "video_core/host_shaders/fxaa_vert.h" +#include "video_core/host_shaders/opengl_fidelityfx_fsr_easu_frag.h" +#include "video_core/host_shaders/opengl_fidelityfx_fsr_frag.h" +#include "video_core/host_shaders/opengl_fidelityfx_fsr_rcas_frag.h" +#include "video_core/host_shaders/opengl_present_frag.h" +#include "video_core/host_shaders/opengl_present_scaleforce_frag.h" +#include "video_core/host_shaders/opengl_present_vert.h" +#include "video_core/host_shaders/opengl_smaa_glsl.h" +#include "video_core/host_shaders/present_bicubic_frag.h" +#include "video_core/host_shaders/present_gaussian_frag.h" +#include "video_core/host_shaders/smaa_blending_weight_calculation_frag.h" +#include "video_core/host_shaders/smaa_blending_weight_calculation_vert.h" +#include "video_core/host_shaders/smaa_edge_detection_frag.h" +#include "video_core/host_shaders/smaa_edge_detection_vert.h" +#include "video_core/host_shaders/smaa_neighborhood_blending_frag.h" +#include "video_core/host_shaders/smaa_neighborhood_blending_vert.h" +#include "video_core/renderer_opengl/gl_blit_screen.h" +#include "video_core/renderer_opengl/gl_rasterizer.h" +#include "video_core/renderer_opengl/gl_shader_manager.h" +#include "video_core/renderer_opengl/gl_shader_util.h" +#include "video_core/renderer_opengl/gl_state_tracker.h" +#include "video_core/smaa_area_tex.h" +#include "video_core/smaa_search_tex.h" +#include "video_core/textures/decoders.h" + +namespace OpenGL { + +namespace { +constexpr GLint PositionLocation = 0; +constexpr GLint TexCoordLocation = 1; +constexpr GLint ModelViewMatrixLocation = 0; + +struct ScreenRectVertex { + constexpr ScreenRectVertex(u32 x, u32 y, GLfloat u, GLfloat v) + : position{{static_cast(x), static_cast(y)}}, tex_coord{{u, v}} {} + + std::array position; + std::array tex_coord; +}; + +/** + * Defines a 1:1 pixel ortographic projection matrix with (0,0) on the top-left + * corner and (width, height) on the lower-bottom. + * + * The projection part of the matrix is trivial, hence these operations are represented + * by a 3x2 matrix. + */ +std::array MakeOrthographicMatrix(float width, float height) { + std::array matrix; // Laid out in column-major order + + // clang-format off + matrix[0] = 2.f / width; matrix[2] = 0.f; matrix[4] = -1.f; + matrix[1] = 0.f; matrix[3] = -2.f / height; matrix[5] = 1.f; + // Last matrix row is implicitly assumed to be [0, 0, 1]. + // clang-format on + + return matrix; +} +} // namespace + +BlitScreen::BlitScreen(RasterizerOpenGL& rasterizer_, + Tegra::MaxwellDeviceMemoryManager& device_memory_, + StateTracker& state_tracker_, ProgramManager& program_manager_, + Device& device_) + : rasterizer(rasterizer_), device_memory(device_memory_), state_tracker(state_tracker_), + program_manager(program_manager_), device(device_) { + // Create shader programs + fxaa_vertex = CreateProgram(HostShaders::FXAA_VERT, GL_VERTEX_SHADER); + fxaa_fragment = CreateProgram(HostShaders::FXAA_FRAG, GL_FRAGMENT_SHADER); + + const auto replace_include = [](std::string& shader_source, std::string_view include_name, + std::string_view include_content) { + const std::string include_string = fmt::format("#include \"{}\"", include_name); + const std::size_t pos = shader_source.find(include_string); + ASSERT(pos != std::string::npos); + shader_source.replace(pos, include_string.size(), include_content); + }; + + const auto SmaaShader = [&](std::string_view specialized_source, GLenum stage) { + std::string shader_source{specialized_source}; + replace_include(shader_source, "opengl_smaa.glsl", HostShaders::OPENGL_SMAA_GLSL); + return CreateProgram(shader_source, stage); + }; + + smaa_edge_detection_vert = SmaaShader(HostShaders::SMAA_EDGE_DETECTION_VERT, GL_VERTEX_SHADER); + smaa_edge_detection_frag = + SmaaShader(HostShaders::SMAA_EDGE_DETECTION_FRAG, GL_FRAGMENT_SHADER); + smaa_blending_weight_calculation_vert = + SmaaShader(HostShaders::SMAA_BLENDING_WEIGHT_CALCULATION_VERT, GL_VERTEX_SHADER); + smaa_blending_weight_calculation_frag = + SmaaShader(HostShaders::SMAA_BLENDING_WEIGHT_CALCULATION_FRAG, GL_FRAGMENT_SHADER); + smaa_neighborhood_blending_vert = + SmaaShader(HostShaders::SMAA_NEIGHBORHOOD_BLENDING_VERT, GL_VERTEX_SHADER); + smaa_neighborhood_blending_frag = + SmaaShader(HostShaders::SMAA_NEIGHBORHOOD_BLENDING_FRAG, GL_FRAGMENT_SHADER); + + present_vertex = CreateProgram(HostShaders::OPENGL_PRESENT_VERT, GL_VERTEX_SHADER); + present_bilinear_fragment = CreateProgram(HostShaders::OPENGL_PRESENT_FRAG, GL_FRAGMENT_SHADER); + present_bicubic_fragment = CreateProgram(HostShaders::PRESENT_BICUBIC_FRAG, GL_FRAGMENT_SHADER); + present_gaussian_fragment = + CreateProgram(HostShaders::PRESENT_GAUSSIAN_FRAG, GL_FRAGMENT_SHADER); + present_scaleforce_fragment = + CreateProgram(fmt::format("#version 460\n{}", HostShaders::OPENGL_PRESENT_SCALEFORCE_FRAG), + GL_FRAGMENT_SHADER); + + std::string fsr_source{HostShaders::OPENGL_FIDELITYFX_FSR_FRAG}; + replace_include(fsr_source, "ffx_a.h", HostShaders::FFX_A_H); + replace_include(fsr_source, "ffx_fsr1.h", HostShaders::FFX_FSR1_H); + + std::string fsr_easu_frag_source{HostShaders::OPENGL_FIDELITYFX_FSR_EASU_FRAG}; + std::string fsr_rcas_frag_source{HostShaders::OPENGL_FIDELITYFX_FSR_RCAS_FRAG}; + replace_include(fsr_easu_frag_source, "opengl_fidelityfx_fsr.frag", fsr_source); + replace_include(fsr_rcas_frag_source, "opengl_fidelityfx_fsr.frag", fsr_source); + + fsr = std::make_unique(HostShaders::FULL_SCREEN_TRIANGLE_VERT, fsr_easu_frag_source, + fsr_rcas_frag_source); + + // Generate presentation sampler + present_sampler.Create(); + glSamplerParameteri(present_sampler.handle, GL_TEXTURE_MIN_FILTER, GL_LINEAR); + glSamplerParameteri(present_sampler.handle, GL_TEXTURE_MAG_FILTER, GL_LINEAR); + glSamplerParameteri(present_sampler.handle, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glSamplerParameteri(present_sampler.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + glSamplerParameteri(present_sampler.handle, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE); + + present_sampler_nn.Create(); + glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_MIN_FILTER, GL_NEAREST); + glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_MAG_FILTER, GL_NEAREST); + glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE); + + // Generate VBO handle for drawing + vertex_buffer.Create(); + + // Attach vertex data to VAO + glNamedBufferData(vertex_buffer.handle, sizeof(ScreenRectVertex) * 4, nullptr, GL_STREAM_DRAW); + + // Allocate textures for the screen + framebuffer_texture.resource.Create(GL_TEXTURE_2D); + + const GLuint texture = framebuffer_texture.resource.handle; + glTextureStorage2D(texture, 1, GL_RGBA8, 1, 1); + + // Clear screen to black + const u8 framebuffer_data[4] = {0, 0, 0, 0}; + glClearTexImage(framebuffer_texture.resource.handle, 0, GL_RGBA, GL_UNSIGNED_BYTE, + framebuffer_data); + + aa_framebuffer.Create(); + + smaa_area_tex.Create(GL_TEXTURE_2D); + glTextureStorage2D(smaa_area_tex.handle, 1, GL_RG8, AREATEX_WIDTH, AREATEX_HEIGHT); + glTextureSubImage2D(smaa_area_tex.handle, 0, 0, 0, AREATEX_WIDTH, AREATEX_HEIGHT, GL_RG, + GL_UNSIGNED_BYTE, areaTexBytes); + smaa_search_tex.Create(GL_TEXTURE_2D); + glTextureStorage2D(smaa_search_tex.handle, 1, GL_R8, SEARCHTEX_WIDTH, SEARCHTEX_HEIGHT); + glTextureSubImage2D(smaa_search_tex.handle, 0, 0, 0, SEARCHTEX_WIDTH, SEARCHTEX_HEIGHT, GL_RED, + GL_UNSIGNED_BYTE, searchTexBytes); + + // Enable unified vertex attributes and query vertex buffer address when the driver supports it + if (device.HasVertexBufferUnifiedMemory()) { + glEnableClientState(GL_VERTEX_ATTRIB_ARRAY_UNIFIED_NV); + glEnableClientState(GL_ELEMENT_ARRAY_UNIFIED_NV); + glMakeNamedBufferResidentNV(vertex_buffer.handle, GL_READ_ONLY); + glGetNamedBufferParameterui64vNV(vertex_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, + &vertex_buffer_address); + } +} + +FramebufferTextureInfo BlitScreen::PrepareRenderTarget( + const Tegra::FramebufferConfig& framebuffer) { + // If framebuffer is provided, reload it from memory to a texture + if (framebuffer_texture.width != static_cast(framebuffer.width) || + framebuffer_texture.height != static_cast(framebuffer.height) || + framebuffer_texture.pixel_format != framebuffer.pixel_format || + gl_framebuffer_data.empty()) { + // Reallocate texture if the framebuffer size has changed. + // This is expected to not happen very often and hence should not be a + // performance problem. + ConfigureFramebufferTexture(framebuffer); + } + + // Load the framebuffer from memory if needed + return LoadFBToScreenInfo(framebuffer); +} + +FramebufferTextureInfo BlitScreen::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer) { + const DAddr framebuffer_addr{framebuffer.address + framebuffer.offset}; + const auto accelerated_info = + rasterizer.AccelerateDisplay(framebuffer, framebuffer_addr, framebuffer.stride); + if (accelerated_info) { + return *accelerated_info; + } + + // Reset the screen info's display texture to its own permanent texture + FramebufferTextureInfo info{}; + info.display_texture = framebuffer_texture.resource.handle; + info.width = framebuffer.width; + info.height = framebuffer.height; + info.scaled_width = framebuffer.width; + info.scaled_height = framebuffer.height; + + // TODO(Rodrigo): Read this from HLE + constexpr u32 block_height_log2 = 4; + const auto pixel_format{ + VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format)}; + const u32 bytes_per_pixel{VideoCore::Surface::BytesPerBlock(pixel_format)}; + const u64 size_in_bytes{Tegra::Texture::CalculateSize( + true, bytes_per_pixel, framebuffer.stride, framebuffer.height, 1, block_height_log2, 0)}; + const u8* const host_ptr{device_memory.GetPointer(framebuffer_addr)}; + const std::span input_data(host_ptr, size_in_bytes); + Tegra::Texture::UnswizzleTexture(gl_framebuffer_data, input_data, bytes_per_pixel, + framebuffer.width, framebuffer.height, 1, block_height_log2, + 0); + + glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0); + glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast(framebuffer.stride)); + + // Update existing texture + // TODO: Test what happens on hardware when you change the framebuffer dimensions so that + // they differ from the LCD resolution. + // TODO: Applications could theoretically crash yuzu here by specifying too large + // framebuffer sizes. We should make sure that this cannot happen. + glTextureSubImage2D(framebuffer_texture.resource.handle, 0, 0, 0, framebuffer.width, + framebuffer.height, framebuffer_texture.gl_format, + framebuffer_texture.gl_type, gl_framebuffer_data.data()); + + glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); + + return info; +} + +void BlitScreen::ConfigureFramebufferTexture(const Tegra::FramebufferConfig& framebuffer) { + framebuffer_texture.width = framebuffer.width; + framebuffer_texture.height = framebuffer.height; + framebuffer_texture.pixel_format = framebuffer.pixel_format; + + const auto pixel_format{ + VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format)}; + const u32 bytes_per_pixel{VideoCore::Surface::BytesPerBlock(pixel_format)}; + gl_framebuffer_data.resize(framebuffer_texture.width * framebuffer_texture.height * + bytes_per_pixel); + + GLint internal_format; + switch (framebuffer.pixel_format) { + case Service::android::PixelFormat::Rgba8888: + internal_format = GL_RGBA8; + framebuffer_texture.gl_format = GL_RGBA; + framebuffer_texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; + break; + case Service::android::PixelFormat::Rgb565: + internal_format = GL_RGB565; + framebuffer_texture.gl_format = GL_RGB; + framebuffer_texture.gl_type = GL_UNSIGNED_SHORT_5_6_5; + break; + default: + internal_format = GL_RGBA8; + framebuffer_texture.gl_format = GL_RGBA; + framebuffer_texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; + // UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}", + // static_cast(framebuffer.pixel_format)); + break; + } + + framebuffer_texture.resource.Release(); + framebuffer_texture.resource.Create(GL_TEXTURE_2D); + glTextureStorage2D(framebuffer_texture.resource.handle, 1, internal_format, + framebuffer_texture.width, framebuffer_texture.height); + aa_texture.Release(); + aa_texture.Create(GL_TEXTURE_2D); + glTextureStorage2D(aa_texture.handle, 1, GL_RGBA16F, + Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), + Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); + glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, aa_texture.handle, 0); + smaa_edges_tex.Release(); + smaa_edges_tex.Create(GL_TEXTURE_2D); + glTextureStorage2D(smaa_edges_tex.handle, 1, GL_RG16F, + Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), + Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); + smaa_blend_tex.Release(); + smaa_blend_tex.Create(GL_TEXTURE_2D); + glTextureStorage2D(smaa_blend_tex.handle, 1, GL_RGBA16F, + Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), + Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); +} + +void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, + const Layout::FramebufferLayout& layout) { + FramebufferTextureInfo info = PrepareRenderTarget(framebuffer); + const auto crop = Tegra::NormalizeCrop(framebuffer, info.width, info.height); + + // TODO: Signal state tracker about these changes + state_tracker.NotifyScreenDrawVertexArray(); + state_tracker.NotifyPolygonModes(); + state_tracker.NotifyViewport0(); + state_tracker.NotifyScissor0(); + state_tracker.NotifyColorMask(0); + state_tracker.NotifyBlend0(); + state_tracker.NotifyFramebuffer(); + state_tracker.NotifyFrontFace(); + state_tracker.NotifyCullTest(); + state_tracker.NotifyDepthTest(); + state_tracker.NotifyStencilTest(); + state_tracker.NotifyPolygonOffset(); + state_tracker.NotifyRasterizeEnable(); + state_tracker.NotifyFramebufferSRGB(); + state_tracker.NotifyLogicOp(); + state_tracker.NotifyClipControl(); + state_tracker.NotifyAlphaTest(); + + state_tracker.ClipControl(GL_LOWER_LEFT, GL_ZERO_TO_ONE); + + glEnable(GL_CULL_FACE); + glDisable(GL_COLOR_LOGIC_OP); + glDisable(GL_DEPTH_TEST); + glDisable(GL_STENCIL_TEST); + glDisable(GL_POLYGON_OFFSET_FILL); + glDisable(GL_RASTERIZER_DISCARD); + glDisable(GL_ALPHA_TEST); + glDisablei(GL_BLEND, 0); + glPolygonMode(GL_FRONT_AND_BACK, GL_FILL); + glCullFace(GL_BACK); + glFrontFace(GL_CW); + glColorMaski(0, GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); + glDepthRangeIndexed(0, 0.0, 0.0); + + glBindTextureUnit(0, info.display_texture); + + auto anti_aliasing = Settings::values.anti_aliasing.GetValue(); + if (anti_aliasing >= Settings::AntiAliasing::MaxEnum) { + LOG_ERROR(Render_OpenGL, "Invalid antialiasing option selected {}", anti_aliasing); + anti_aliasing = Settings::AntiAliasing::None; + Settings::values.anti_aliasing.SetValue(anti_aliasing); + } + + if (anti_aliasing != Settings::AntiAliasing::None) { + glEnablei(GL_SCISSOR_TEST, 0); + auto scissor_width = Settings::values.resolution_info.ScaleUp(framebuffer_texture.width); + auto viewport_width = static_cast(scissor_width); + auto scissor_height = Settings::values.resolution_info.ScaleUp(framebuffer_texture.height); + auto viewport_height = static_cast(scissor_height); + + glScissorIndexed(0, 0, 0, scissor_width, scissor_height); + glViewportIndexedf(0, 0.0f, 0.0f, viewport_width, viewport_height); + + glBindSampler(0, present_sampler.handle); + GLint old_read_fb; + GLint old_draw_fb; + glGetIntegerv(GL_READ_FRAMEBUFFER_BINDING, &old_read_fb); + glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb); + + switch (anti_aliasing) { + case Settings::AntiAliasing::Fxaa: { + program_manager.BindPresentPrograms(fxaa_vertex.handle, fxaa_fragment.handle); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, aa_framebuffer.handle); + glDrawArrays(GL_TRIANGLE_STRIP, 0, 4); + } break; + case Settings::AntiAliasing::Smaa: { + glClearColor(0, 0, 0, 0); + glFrontFace(GL_CCW); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, aa_framebuffer.handle); + glBindSampler(1, present_sampler.handle); + glBindSampler(2, present_sampler.handle); + + glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, + smaa_edges_tex.handle, 0); + glClear(GL_COLOR_BUFFER_BIT); + program_manager.BindPresentPrograms(smaa_edge_detection_vert.handle, + smaa_edge_detection_frag.handle); + glDrawArrays(GL_TRIANGLES, 0, 3); + + glBindTextureUnit(0, smaa_edges_tex.handle); + glBindTextureUnit(1, smaa_area_tex.handle); + glBindTextureUnit(2, smaa_search_tex.handle); + glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, + smaa_blend_tex.handle, 0); + glClear(GL_COLOR_BUFFER_BIT); + program_manager.BindPresentPrograms(smaa_blending_weight_calculation_vert.handle, + smaa_blending_weight_calculation_frag.handle); + glDrawArrays(GL_TRIANGLES, 0, 3); + + glBindTextureUnit(0, info.display_texture); + glBindTextureUnit(1, smaa_blend_tex.handle); + glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, + aa_texture.handle, 0); + program_manager.BindPresentPrograms(smaa_neighborhood_blending_vert.handle, + smaa_neighborhood_blending_frag.handle); + glDrawArrays(GL_TRIANGLES, 0, 3); + glFrontFace(GL_CW); + } break; + default: + UNREACHABLE(); + } + + glBindFramebuffer(GL_READ_FRAMEBUFFER, old_read_fb); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); + + glBindTextureUnit(0, aa_texture.handle); + } + glDisablei(GL_SCISSOR_TEST, 0); + + if (Settings::values.scaling_filter.GetValue() == Settings::ScalingFilter::Fsr) { + if (!fsr->AreBuffersInitialized()) { + fsr->InitBuffers(); + } + + glBindSampler(0, present_sampler.handle); + fsr->Draw(program_manager, layout.screen, info.scaled_width, info.scaled_height, crop); + } else { + if (fsr->AreBuffersInitialized()) { + fsr->ReleaseBuffers(); + } + } + + const std::array ortho_matrix = + MakeOrthographicMatrix(static_cast(layout.width), static_cast(layout.height)); + + const auto fragment_handle = [this]() { + switch (Settings::values.scaling_filter.GetValue()) { + case Settings::ScalingFilter::NearestNeighbor: + case Settings::ScalingFilter::Bilinear: + return present_bilinear_fragment.handle; + case Settings::ScalingFilter::Bicubic: + return present_bicubic_fragment.handle; + case Settings::ScalingFilter::Gaussian: + return present_gaussian_fragment.handle; + case Settings::ScalingFilter::ScaleForce: + return present_scaleforce_fragment.handle; + case Settings::ScalingFilter::Fsr: + return fsr->GetPresentFragmentProgram().handle; + default: + return present_bilinear_fragment.handle; + } + }(); + program_manager.BindPresentPrograms(present_vertex.handle, fragment_handle); + glProgramUniformMatrix3x2fv(present_vertex.handle, ModelViewMatrixLocation, 1, GL_FALSE, + ortho_matrix.data()); + + f32 left, top, right, bottom; + if (Settings::values.scaling_filter.GetValue() == Settings::ScalingFilter::Fsr) { + // FSR has already applied the crop, so we just want to render the image + // it has produced. + left = 0; + top = 0; + right = 1; + bottom = 1; + } else { + // Apply the precomputed crop. + left = crop.left; + top = crop.top; + right = crop.right; + bottom = crop.bottom; + } + + // Map the coordinates to the screen. + const auto& screen = layout.screen; + const auto x = screen.left; + const auto y = screen.top; + const auto w = screen.GetWidth(); + const auto h = screen.GetHeight(); + + const std::array vertices = { + ScreenRectVertex(x, y, left, top), + ScreenRectVertex(x + w, y, right, top), + ScreenRectVertex(x, y + h, left, bottom), + ScreenRectVertex(x + w, y + h, right, bottom), + }; + glNamedBufferSubData(vertex_buffer.handle, 0, sizeof(vertices), std::data(vertices)); + + glDisable(GL_FRAMEBUFFER_SRGB); + glViewportIndexedf(0, 0.0f, 0.0f, static_cast(layout.width), + static_cast(layout.height)); + + glEnableVertexAttribArray(PositionLocation); + glEnableVertexAttribArray(TexCoordLocation); + glVertexAttribDivisor(PositionLocation, 0); + glVertexAttribDivisor(TexCoordLocation, 0); + glVertexAttribFormat(PositionLocation, 2, GL_FLOAT, GL_FALSE, + offsetof(ScreenRectVertex, position)); + glVertexAttribFormat(TexCoordLocation, 2, GL_FLOAT, GL_FALSE, + offsetof(ScreenRectVertex, tex_coord)); + glVertexAttribBinding(PositionLocation, 0); + glVertexAttribBinding(TexCoordLocation, 0); + 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)); + } + + if (Settings::values.scaling_filter.GetValue() != Settings::ScalingFilter::NearestNeighbor) { + glBindSampler(0, present_sampler.handle); + } else { + glBindSampler(0, present_sampler_nn.handle); + } + + // Update background color before drawing + glClearColor(Settings::values.bg_red.GetValue() / 255.0f, + Settings::values.bg_green.GetValue() / 255.0f, + Settings::values.bg_blue.GetValue() / 255.0f, 1.0f); + + glClear(GL_COLOR_BUFFER_BIT); + glDrawArrays(GL_TRIANGLE_STRIP, 0, 4); + + // TODO + // program_manager.RestoreGuestPipeline(); +} + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_blit_screen.h b/src/video_core/renderer_opengl/gl_blit_screen.h new file mode 100644 index 000000000..13d769958 --- /dev/null +++ b/src/video_core/renderer_opengl/gl_blit_screen.h @@ -0,0 +1,110 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include +#include + +#include "core/hle/service/nvnflinger/pixel_format.h" +#include "video_core/host1x/gpu_device_memory_manager.h" +#include "video_core/renderer_opengl/gl_fsr.h" +#include "video_core/renderer_opengl/gl_resource_manager.h" + +namespace Layout { +struct FramebufferLayout; +} + +namespace Tegra { +struct FramebufferConfig; +} + +namespace OpenGL { + +class Device; +class RasterizerOpenGL; +class StateTracker; + +/// Structure used for storing information about the textures for the Switch screen +struct TextureInfo { + OGLTexture resource; + GLsizei width; + GLsizei height; + GLenum gl_format; + GLenum gl_type; + Service::android::PixelFormat pixel_format; +}; + +/// Structure used for storing information about the display target for the Switch screen +struct FramebufferTextureInfo { + GLuint display_texture{}; + u32 width; + u32 height; + u32 scaled_width; + u32 scaled_height; +}; + +class BlitScreen { +public: + explicit BlitScreen(RasterizerOpenGL& rasterizer, + Tegra::MaxwellDeviceMemoryManager& device_memory, + StateTracker& state_tracker, ProgramManager& program_manager, + Device& device); + + void ConfigureFramebufferTexture(const Tegra::FramebufferConfig& framebuffer); + + /// Draws the emulated screens to the emulator window. + void DrawScreen(const Tegra::FramebufferConfig& framebuffer, + const Layout::FramebufferLayout& layout); + + void RenderScreenshot(const Tegra::FramebufferConfig& framebuffer); + + /// Loads framebuffer from emulated memory into the active OpenGL texture. + FramebufferTextureInfo LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer); + + FramebufferTextureInfo PrepareRenderTarget(const Tegra::FramebufferConfig& framebuffer); + +private: + RasterizerOpenGL& rasterizer; + Tegra::MaxwellDeviceMemoryManager& device_memory; + StateTracker& state_tracker; + ProgramManager& program_manager; + Device& device; + + OGLSampler present_sampler; + OGLSampler present_sampler_nn; + OGLBuffer vertex_buffer; + OGLProgram fxaa_vertex; + OGLProgram fxaa_fragment; + OGLProgram present_vertex; + OGLProgram present_bilinear_fragment; + OGLProgram present_bicubic_fragment; + OGLProgram present_gaussian_fragment; + OGLProgram present_scaleforce_fragment; + + /// Display information for Switch screen + TextureInfo framebuffer_texture; + OGLTexture aa_texture; + OGLFramebuffer aa_framebuffer; + + OGLProgram smaa_edge_detection_vert; + OGLProgram smaa_blending_weight_calculation_vert; + OGLProgram smaa_neighborhood_blending_vert; + OGLProgram smaa_edge_detection_frag; + OGLProgram smaa_blending_weight_calculation_frag; + OGLProgram smaa_neighborhood_blending_frag; + OGLTexture smaa_area_tex; + OGLTexture smaa_search_tex; + OGLTexture smaa_edges_tex; + OGLTexture smaa_blend_tex; + + std::unique_ptr fsr; + + /// OpenGL framebuffer data + std::vector gl_framebuffer_data; + + // GPU address of the vertex buffer + GLuint64EXT vertex_buffer_address = 0; +}; + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index ee82d9f3a..6eae51ff7 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -16,6 +16,7 @@ #include "video_core/engines/maxwell_dma.h" #include "video_core/rasterizer_interface.h" #include "video_core/renderer_opengl/blit_image.h" +#include "video_core/renderer_opengl/gl_blit_screen.h" #include "video_core/renderer_opengl/gl_buffer_cache.h" #include "video_core/renderer_opengl/gl_device.h" #include "video_core/renderer_opengl/gl_fence_manager.h" diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index 2b9ebff92..38b0aacf4 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -16,68 +16,16 @@ #include "core/core_timing.h" #include "core/frontend/emu_window.h" #include "core/telemetry_session.h" -#include "video_core/host_shaders/ffx_a_h.h" -#include "video_core/host_shaders/ffx_fsr1_h.h" -#include "video_core/host_shaders/full_screen_triangle_vert.h" -#include "video_core/host_shaders/fxaa_frag.h" -#include "video_core/host_shaders/fxaa_vert.h" -#include "video_core/host_shaders/opengl_fidelityfx_fsr_easu_frag.h" -#include "video_core/host_shaders/opengl_fidelityfx_fsr_frag.h" -#include "video_core/host_shaders/opengl_fidelityfx_fsr_rcas_frag.h" -#include "video_core/host_shaders/opengl_present_frag.h" -#include "video_core/host_shaders/opengl_present_scaleforce_frag.h" -#include "video_core/host_shaders/opengl_present_vert.h" -#include "video_core/host_shaders/opengl_smaa_glsl.h" -#include "video_core/host_shaders/present_bicubic_frag.h" -#include "video_core/host_shaders/present_gaussian_frag.h" -#include "video_core/host_shaders/smaa_blending_weight_calculation_frag.h" -#include "video_core/host_shaders/smaa_blending_weight_calculation_vert.h" -#include "video_core/host_shaders/smaa_edge_detection_frag.h" -#include "video_core/host_shaders/smaa_edge_detection_vert.h" -#include "video_core/host_shaders/smaa_neighborhood_blending_frag.h" -#include "video_core/host_shaders/smaa_neighborhood_blending_vert.h" +#include "video_core/renderer_opengl/gl_blit_screen.h" #include "video_core/renderer_opengl/gl_fsr.h" #include "video_core/renderer_opengl/gl_rasterizer.h" #include "video_core/renderer_opengl/gl_shader_manager.h" #include "video_core/renderer_opengl/gl_shader_util.h" #include "video_core/renderer_opengl/renderer_opengl.h" -#include "video_core/smaa_area_tex.h" -#include "video_core/smaa_search_tex.h" #include "video_core/textures/decoders.h" namespace OpenGL { namespace { -constexpr GLint PositionLocation = 0; -constexpr GLint TexCoordLocation = 1; -constexpr GLint ModelViewMatrixLocation = 0; - -struct ScreenRectVertex { - constexpr ScreenRectVertex(u32 x, u32 y, GLfloat u, GLfloat v) - : position{{static_cast(x), static_cast(y)}}, tex_coord{{u, v}} {} - - std::array position; - std::array tex_coord; -}; - -/** - * Defines a 1:1 pixel ortographic projection matrix with (0,0) on the top-left - * corner and (width, height) on the lower-bottom. - * - * The projection part of the matrix is trivial, hence these operations are represented - * by a 3x2 matrix. - */ -std::array MakeOrthographicMatrix(float width, float height) { - std::array matrix; // Laid out in column-major order - - // clang-format off - matrix[0] = 2.f / width; matrix[2] = 0.f; matrix[4] = -1.f; - matrix[1] = 0.f; matrix[3] = -2.f / height; matrix[5] = 1.f; - // Last matrix row is implicitly assumed to be [0, 0, 1]. - // clang-format on - - return matrix; -} - const char* GetSource(GLenum source) { switch (source) { case GL_DEBUG_SOURCE_API: @@ -155,7 +103,6 @@ RendererOpenGL::RendererOpenGL(Core::TelemetrySession& telemetry_session_, glDebugMessageCallback(DebugHandler, nullptr); } AddTelemetryFields(); - InitOpenGLObjects(); // Initialize default attributes to match hardware's disabled attributes GLint max_attribs{}; @@ -167,14 +114,8 @@ RendererOpenGL::RendererOpenGL(Core::TelemetrySession& telemetry_session_, if (!GLAD_GL_ARB_seamless_cubemap_per_texture && !GLAD_GL_AMD_seamless_cubemap_per_texture) { glEnable(GL_TEXTURE_CUBE_MAP_SEAMLESS); } - // Enable unified vertex attributes and query vertex buffer address when the driver supports it - if (device.HasVertexBufferUnifiedMemory()) { - glEnableClientState(GL_VERTEX_ATTRIB_ARRAY_UNIFIED_NV); - glEnableClientState(GL_ELEMENT_ARRAY_UNIFIED_NV); - glMakeNamedBufferResidentNV(vertex_buffer.handle, GL_READ_ONLY); - glGetNamedBufferParameterui64vNV(vertex_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, - &vertex_buffer_address); - } + blit_screen = std::make_unique(rasterizer, device_memory, state_tracker, + program_manager, device); } RendererOpenGL::~RendererOpenGL() = default; @@ -187,7 +128,7 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { RenderScreenshot(*framebuffer); state_tracker.BindFramebuffer(0); - DrawScreen(*framebuffer, emu_window.GetFramebufferLayout()); + blit_screen->DrawScreen(*framebuffer, emu_window.GetFramebufferLayout()); ++m_current_frame; @@ -198,166 +139,6 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { render_window.OnFrameDisplayed(); } -FramebufferTextureInfo RendererOpenGL::PrepareRenderTarget( - const Tegra::FramebufferConfig& framebuffer) { - // If framebuffer is provided, reload it from memory to a texture - if (framebuffer_texture.width != static_cast(framebuffer.width) || - framebuffer_texture.height != static_cast(framebuffer.height) || - framebuffer_texture.pixel_format != framebuffer.pixel_format || - gl_framebuffer_data.empty()) { - // Reallocate texture if the framebuffer size has changed. - // This is expected to not happen very often and hence should not be a - // performance problem. - ConfigureFramebufferTexture(framebuffer); - } - - // Load the framebuffer from memory if needed - return LoadFBToScreenInfo(framebuffer); -} - -FramebufferTextureInfo RendererOpenGL::LoadFBToScreenInfo( - const Tegra::FramebufferConfig& framebuffer) { - const VAddr framebuffer_addr{framebuffer.address + framebuffer.offset}; - const auto accelerated_info = - rasterizer.AccelerateDisplay(framebuffer, framebuffer_addr, framebuffer.stride); - if (accelerated_info) { - return *accelerated_info; - } - - // Reset the screen info's display texture to its own permanent texture - FramebufferTextureInfo info{}; - info.display_texture = framebuffer_texture.resource.handle; - info.width = framebuffer.width; - info.height = framebuffer.height; - info.scaled_width = framebuffer.width; - info.scaled_height = framebuffer.height; - - // TODO(Rodrigo): Read this from HLE - constexpr u32 block_height_log2 = 4; - const auto pixel_format{ - VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format)}; - const u32 bytes_per_pixel{VideoCore::Surface::BytesPerBlock(pixel_format)}; - const u64 size_in_bytes{Tegra::Texture::CalculateSize( - true, bytes_per_pixel, framebuffer.stride, framebuffer.height, 1, block_height_log2, 0)}; - const u8* const host_ptr{device_memory.GetPointer(framebuffer_addr)}; - const std::span input_data(host_ptr, size_in_bytes); - Tegra::Texture::UnswizzleTexture(gl_framebuffer_data, input_data, bytes_per_pixel, - framebuffer.width, framebuffer.height, 1, block_height_log2, - 0); - - glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0); - glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast(framebuffer.stride)); - - // Update existing texture - // TODO: Test what happens on hardware when you change the framebuffer dimensions so that - // they differ from the LCD resolution. - // TODO: Applications could theoretically crash yuzu here by specifying too large - // framebuffer sizes. We should make sure that this cannot happen. - glTextureSubImage2D(framebuffer_texture.resource.handle, 0, 0, 0, framebuffer.width, - framebuffer.height, framebuffer_texture.gl_format, - framebuffer_texture.gl_type, gl_framebuffer_data.data()); - - glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); - - return info; -} - -void RendererOpenGL::InitOpenGLObjects() { - // Create shader programs - fxaa_vertex = CreateProgram(HostShaders::FXAA_VERT, GL_VERTEX_SHADER); - fxaa_fragment = CreateProgram(HostShaders::FXAA_FRAG, GL_FRAGMENT_SHADER); - - const auto replace_include = [](std::string& shader_source, std::string_view include_name, - std::string_view include_content) { - const std::string include_string = fmt::format("#include \"{}\"", include_name); - const std::size_t pos = shader_source.find(include_string); - ASSERT(pos != std::string::npos); - shader_source.replace(pos, include_string.size(), include_content); - }; - - const auto SmaaShader = [&](std::string_view specialized_source, GLenum stage) { - std::string shader_source{specialized_source}; - replace_include(shader_source, "opengl_smaa.glsl", HostShaders::OPENGL_SMAA_GLSL); - return CreateProgram(shader_source, stage); - }; - - smaa_edge_detection_vert = SmaaShader(HostShaders::SMAA_EDGE_DETECTION_VERT, GL_VERTEX_SHADER); - smaa_edge_detection_frag = - SmaaShader(HostShaders::SMAA_EDGE_DETECTION_FRAG, GL_FRAGMENT_SHADER); - smaa_blending_weight_calculation_vert = - SmaaShader(HostShaders::SMAA_BLENDING_WEIGHT_CALCULATION_VERT, GL_VERTEX_SHADER); - smaa_blending_weight_calculation_frag = - SmaaShader(HostShaders::SMAA_BLENDING_WEIGHT_CALCULATION_FRAG, GL_FRAGMENT_SHADER); - smaa_neighborhood_blending_vert = - SmaaShader(HostShaders::SMAA_NEIGHBORHOOD_BLENDING_VERT, GL_VERTEX_SHADER); - smaa_neighborhood_blending_frag = - SmaaShader(HostShaders::SMAA_NEIGHBORHOOD_BLENDING_FRAG, GL_FRAGMENT_SHADER); - - present_vertex = CreateProgram(HostShaders::OPENGL_PRESENT_VERT, GL_VERTEX_SHADER); - present_bilinear_fragment = CreateProgram(HostShaders::OPENGL_PRESENT_FRAG, GL_FRAGMENT_SHADER); - present_bicubic_fragment = CreateProgram(HostShaders::PRESENT_BICUBIC_FRAG, GL_FRAGMENT_SHADER); - present_gaussian_fragment = - CreateProgram(HostShaders::PRESENT_GAUSSIAN_FRAG, GL_FRAGMENT_SHADER); - present_scaleforce_fragment = - CreateProgram(fmt::format("#version 460\n{}", HostShaders::OPENGL_PRESENT_SCALEFORCE_FRAG), - GL_FRAGMENT_SHADER); - - std::string fsr_source{HostShaders::OPENGL_FIDELITYFX_FSR_FRAG}; - replace_include(fsr_source, "ffx_a.h", HostShaders::FFX_A_H); - replace_include(fsr_source, "ffx_fsr1.h", HostShaders::FFX_FSR1_H); - - std::string fsr_easu_frag_source{HostShaders::OPENGL_FIDELITYFX_FSR_EASU_FRAG}; - std::string fsr_rcas_frag_source{HostShaders::OPENGL_FIDELITYFX_FSR_RCAS_FRAG}; - replace_include(fsr_easu_frag_source, "opengl_fidelityfx_fsr.frag", fsr_source); - replace_include(fsr_rcas_frag_source, "opengl_fidelityfx_fsr.frag", fsr_source); - - fsr = std::make_unique(HostShaders::FULL_SCREEN_TRIANGLE_VERT, fsr_easu_frag_source, - fsr_rcas_frag_source); - - // Generate presentation sampler - present_sampler.Create(); - glSamplerParameteri(present_sampler.handle, GL_TEXTURE_MIN_FILTER, GL_LINEAR); - glSamplerParameteri(present_sampler.handle, GL_TEXTURE_MAG_FILTER, GL_LINEAR); - glSamplerParameteri(present_sampler.handle, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); - glSamplerParameteri(present_sampler.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); - glSamplerParameteri(present_sampler.handle, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE); - - present_sampler_nn.Create(); - glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_MIN_FILTER, GL_NEAREST); - glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_MAG_FILTER, GL_NEAREST); - glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); - glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); - glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE); - - // Generate VBO handle for drawing - vertex_buffer.Create(); - - // Attach vertex data to VAO - glNamedBufferData(vertex_buffer.handle, sizeof(ScreenRectVertex) * 4, nullptr, GL_STREAM_DRAW); - - // Allocate textures for the screen - framebuffer_texture.resource.Create(GL_TEXTURE_2D); - - const GLuint texture = framebuffer_texture.resource.handle; - glTextureStorage2D(texture, 1, GL_RGBA8, 1, 1); - - // Clear screen to black - const u8 framebuffer_data[4] = {0, 0, 0, 0}; - glClearTexImage(framebuffer_texture.resource.handle, 0, GL_RGBA, GL_UNSIGNED_BYTE, - framebuffer_data); - - aa_framebuffer.Create(); - - smaa_area_tex.Create(GL_TEXTURE_2D); - glTextureStorage2D(smaa_area_tex.handle, 1, GL_RG8, AREATEX_WIDTH, AREATEX_HEIGHT); - glTextureSubImage2D(smaa_area_tex.handle, 0, 0, 0, AREATEX_WIDTH, AREATEX_HEIGHT, GL_RG, - GL_UNSIGNED_BYTE, areaTexBytes); - smaa_search_tex.Create(GL_TEXTURE_2D); - glTextureStorage2D(smaa_search_tex.handle, 1, GL_R8, SEARCHTEX_WIDTH, SEARCHTEX_HEIGHT); - glTextureSubImage2D(smaa_search_tex.handle, 0, 0, 0, SEARCHTEX_WIDTH, SEARCHTEX_HEIGHT, GL_RED, - GL_UNSIGNED_BYTE, searchTexBytes); -} - void RendererOpenGL::AddTelemetryFields() { const char* const gl_version{reinterpret_cast(glGetString(GL_VERSION))}; const char* const gpu_vendor{reinterpret_cast(glGetString(GL_VENDOR))}; @@ -373,283 +154,6 @@ void RendererOpenGL::AddTelemetryFields() { telemetry_session.AddField(user_system, "GPU_OpenGL_Version", std::string(gl_version)); } -void RendererOpenGL::ConfigureFramebufferTexture(const Tegra::FramebufferConfig& framebuffer) { - framebuffer_texture.width = framebuffer.width; - framebuffer_texture.height = framebuffer.height; - framebuffer_texture.pixel_format = framebuffer.pixel_format; - - const auto pixel_format{ - VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format)}; - const u32 bytes_per_pixel{VideoCore::Surface::BytesPerBlock(pixel_format)}; - gl_framebuffer_data.resize(framebuffer_texture.width * framebuffer_texture.height * - bytes_per_pixel); - - GLint internal_format; - switch (framebuffer.pixel_format) { - case Service::android::PixelFormat::Rgba8888: - internal_format = GL_RGBA8; - framebuffer_texture.gl_format = GL_RGBA; - framebuffer_texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; - break; - case Service::android::PixelFormat::Rgb565: - internal_format = GL_RGB565; - framebuffer_texture.gl_format = GL_RGB; - framebuffer_texture.gl_type = GL_UNSIGNED_SHORT_5_6_5; - break; - default: - internal_format = GL_RGBA8; - framebuffer_texture.gl_format = GL_RGBA; - framebuffer_texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; - // UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}", - // static_cast(framebuffer.pixel_format)); - break; - } - - framebuffer_texture.resource.Release(); - framebuffer_texture.resource.Create(GL_TEXTURE_2D); - glTextureStorage2D(framebuffer_texture.resource.handle, 1, internal_format, - framebuffer_texture.width, framebuffer_texture.height); - aa_texture.Release(); - aa_texture.Create(GL_TEXTURE_2D); - glTextureStorage2D(aa_texture.handle, 1, GL_RGBA16F, - Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), - Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); - glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, aa_texture.handle, 0); - smaa_edges_tex.Release(); - smaa_edges_tex.Create(GL_TEXTURE_2D); - glTextureStorage2D(smaa_edges_tex.handle, 1, GL_RG16F, - Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), - Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); - smaa_blend_tex.Release(); - smaa_blend_tex.Create(GL_TEXTURE_2D); - glTextureStorage2D(smaa_blend_tex.handle, 1, GL_RGBA16F, - Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), - Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); -} - -void RendererOpenGL::DrawScreen(const Tegra::FramebufferConfig& framebuffer, - const Layout::FramebufferLayout& layout) { - FramebufferTextureInfo info = PrepareRenderTarget(framebuffer); - const auto crop = Tegra::NormalizeCrop(framebuffer, info.width, info.height); - - // TODO: Signal state tracker about these changes - state_tracker.NotifyScreenDrawVertexArray(); - state_tracker.NotifyPolygonModes(); - state_tracker.NotifyViewport0(); - state_tracker.NotifyScissor0(); - state_tracker.NotifyColorMask(0); - state_tracker.NotifyBlend0(); - state_tracker.NotifyFramebuffer(); - state_tracker.NotifyFrontFace(); - state_tracker.NotifyCullTest(); - state_tracker.NotifyDepthTest(); - state_tracker.NotifyStencilTest(); - state_tracker.NotifyPolygonOffset(); - state_tracker.NotifyRasterizeEnable(); - state_tracker.NotifyFramebufferSRGB(); - state_tracker.NotifyLogicOp(); - state_tracker.NotifyClipControl(); - state_tracker.NotifyAlphaTest(); - - state_tracker.ClipControl(GL_LOWER_LEFT, GL_ZERO_TO_ONE); - - glEnable(GL_CULL_FACE); - glDisable(GL_COLOR_LOGIC_OP); - glDisable(GL_DEPTH_TEST); - glDisable(GL_STENCIL_TEST); - glDisable(GL_POLYGON_OFFSET_FILL); - glDisable(GL_RASTERIZER_DISCARD); - glDisable(GL_ALPHA_TEST); - glDisablei(GL_BLEND, 0); - glPolygonMode(GL_FRONT_AND_BACK, GL_FILL); - glCullFace(GL_BACK); - glFrontFace(GL_CW); - glColorMaski(0, GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); - glDepthRangeIndexed(0, 0.0, 0.0); - - glBindTextureUnit(0, info.display_texture); - - auto anti_aliasing = Settings::values.anti_aliasing.GetValue(); - if (anti_aliasing >= Settings::AntiAliasing::MaxEnum) { - LOG_ERROR(Render_OpenGL, "Invalid antialiasing option selected {}", anti_aliasing); - anti_aliasing = Settings::AntiAliasing::None; - Settings::values.anti_aliasing.SetValue(anti_aliasing); - } - - if (anti_aliasing != Settings::AntiAliasing::None) { - glEnablei(GL_SCISSOR_TEST, 0); - auto scissor_width = Settings::values.resolution_info.ScaleUp(framebuffer_texture.width); - auto viewport_width = static_cast(scissor_width); - auto scissor_height = Settings::values.resolution_info.ScaleUp(framebuffer_texture.height); - auto viewport_height = static_cast(scissor_height); - - glScissorIndexed(0, 0, 0, scissor_width, scissor_height); - glViewportIndexedf(0, 0.0f, 0.0f, viewport_width, viewport_height); - - glBindSampler(0, present_sampler.handle); - GLint old_read_fb; - GLint old_draw_fb; - glGetIntegerv(GL_READ_FRAMEBUFFER_BINDING, &old_read_fb); - glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb); - - switch (anti_aliasing) { - case Settings::AntiAliasing::Fxaa: { - program_manager.BindPresentPrograms(fxaa_vertex.handle, fxaa_fragment.handle); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, aa_framebuffer.handle); - glDrawArrays(GL_TRIANGLE_STRIP, 0, 4); - } break; - case Settings::AntiAliasing::Smaa: { - glClearColor(0, 0, 0, 0); - glFrontFace(GL_CCW); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, aa_framebuffer.handle); - glBindSampler(1, present_sampler.handle); - glBindSampler(2, present_sampler.handle); - - glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, - smaa_edges_tex.handle, 0); - glClear(GL_COLOR_BUFFER_BIT); - program_manager.BindPresentPrograms(smaa_edge_detection_vert.handle, - smaa_edge_detection_frag.handle); - glDrawArrays(GL_TRIANGLES, 0, 3); - - glBindTextureUnit(0, smaa_edges_tex.handle); - glBindTextureUnit(1, smaa_area_tex.handle); - glBindTextureUnit(2, smaa_search_tex.handle); - glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, - smaa_blend_tex.handle, 0); - glClear(GL_COLOR_BUFFER_BIT); - program_manager.BindPresentPrograms(smaa_blending_weight_calculation_vert.handle, - smaa_blending_weight_calculation_frag.handle); - glDrawArrays(GL_TRIANGLES, 0, 3); - - glBindTextureUnit(0, info.display_texture); - glBindTextureUnit(1, smaa_blend_tex.handle); - glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, - aa_texture.handle, 0); - program_manager.BindPresentPrograms(smaa_neighborhood_blending_vert.handle, - smaa_neighborhood_blending_frag.handle); - glDrawArrays(GL_TRIANGLES, 0, 3); - glFrontFace(GL_CW); - } break; - default: - UNREACHABLE(); - } - - glBindFramebuffer(GL_READ_FRAMEBUFFER, old_read_fb); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); - - glBindTextureUnit(0, aa_texture.handle); - } - glDisablei(GL_SCISSOR_TEST, 0); - - if (Settings::values.scaling_filter.GetValue() == Settings::ScalingFilter::Fsr) { - if (!fsr->AreBuffersInitialized()) { - fsr->InitBuffers(); - } - - glBindSampler(0, present_sampler.handle); - fsr->Draw(program_manager, layout.screen, info.scaled_width, info.scaled_height, crop); - } else { - if (fsr->AreBuffersInitialized()) { - fsr->ReleaseBuffers(); - } - } - - const std::array ortho_matrix = - MakeOrthographicMatrix(static_cast(layout.width), static_cast(layout.height)); - - const auto fragment_handle = [this]() { - switch (Settings::values.scaling_filter.GetValue()) { - case Settings::ScalingFilter::NearestNeighbor: - case Settings::ScalingFilter::Bilinear: - return present_bilinear_fragment.handle; - case Settings::ScalingFilter::Bicubic: - return present_bicubic_fragment.handle; - case Settings::ScalingFilter::Gaussian: - return present_gaussian_fragment.handle; - case Settings::ScalingFilter::ScaleForce: - return present_scaleforce_fragment.handle; - case Settings::ScalingFilter::Fsr: - return fsr->GetPresentFragmentProgram().handle; - default: - return present_bilinear_fragment.handle; - } - }(); - program_manager.BindPresentPrograms(present_vertex.handle, fragment_handle); - glProgramUniformMatrix3x2fv(present_vertex.handle, ModelViewMatrixLocation, 1, GL_FALSE, - ortho_matrix.data()); - - f32 left, top, right, bottom; - if (Settings::values.scaling_filter.GetValue() == Settings::ScalingFilter::Fsr) { - // FSR has already applied the crop, so we just want to render the image - // it has produced. - left = 0; - top = 0; - right = 1; - bottom = 1; - } else { - // Apply the precomputed crop. - left = crop.left; - top = crop.top; - right = crop.right; - bottom = crop.bottom; - } - - // Map the coordinates to the screen. - const auto& screen = layout.screen; - const auto x = screen.left; - const auto y = screen.top; - const auto w = screen.GetWidth(); - const auto h = screen.GetHeight(); - - const std::array vertices = { - ScreenRectVertex(x, y, left, top), - ScreenRectVertex(x + w, y, right, top), - ScreenRectVertex(x, y + h, left, bottom), - ScreenRectVertex(x + w, y + h, right, bottom), - }; - glNamedBufferSubData(vertex_buffer.handle, 0, sizeof(vertices), std::data(vertices)); - - glDisable(GL_FRAMEBUFFER_SRGB); - glViewportIndexedf(0, 0.0f, 0.0f, static_cast(layout.width), - static_cast(layout.height)); - - glEnableVertexAttribArray(PositionLocation); - glEnableVertexAttribArray(TexCoordLocation); - glVertexAttribDivisor(PositionLocation, 0); - glVertexAttribDivisor(TexCoordLocation, 0); - glVertexAttribFormat(PositionLocation, 2, GL_FLOAT, GL_FALSE, - offsetof(ScreenRectVertex, position)); - glVertexAttribFormat(TexCoordLocation, 2, GL_FLOAT, GL_FALSE, - offsetof(ScreenRectVertex, tex_coord)); - glVertexAttribBinding(PositionLocation, 0); - glVertexAttribBinding(TexCoordLocation, 0); - 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)); - } - - if (Settings::values.scaling_filter.GetValue() != Settings::ScalingFilter::NearestNeighbor) { - glBindSampler(0, present_sampler.handle); - } else { - glBindSampler(0, present_sampler_nn.handle); - } - - // Update background color before drawing - glClearColor(Settings::values.bg_red.GetValue() / 255.0f, - Settings::values.bg_green.GetValue() / 255.0f, - Settings::values.bg_blue.GetValue() / 255.0f, 1.0f); - - glClear(GL_COLOR_BUFFER_BIT); - glDrawArrays(GL_TRIANGLE_STRIP, 0, 4); - - // TODO - // program_manager.RestoreGuestPipeline(); -} - void RendererOpenGL::RenderScreenshot(const Tegra::FramebufferConfig& framebuffer) { if (!renderer_settings.screenshot_requested) { return; @@ -672,7 +176,7 @@ void RendererOpenGL::RenderScreenshot(const Tegra::FramebufferConfig& framebuffe glRenderbufferStorage(GL_RENDERBUFFER, GL_SRGB8, layout.width, layout.height); glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, renderbuffer); - DrawScreen(framebuffer, layout); + blit_screen->DrawScreen(framebuffer, layout); glBindBuffer(GL_PIXEL_PACK_BUFFER, 0); glPixelStorei(GL_PACK_ROW_LENGTH, 0); diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h index 3a83a9b78..23aff055a 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.h +++ b/src/video_core/renderer_opengl/renderer_opengl.h @@ -25,38 +25,13 @@ namespace Core::Frontend { class EmuWindow; } -namespace Core::Memory { -class Memory; -} - -namespace Layout { -struct FramebufferLayout; -} - namespace Tegra { class GPU; } namespace OpenGL { -/// Structure used for storing information about the textures for the Switch screen -struct TextureInfo { - OGLTexture resource; - GLsizei width; - GLsizei height; - GLenum gl_format; - GLenum gl_type; - Service::android::PixelFormat pixel_format; -}; - -/// Structure used for storing information about the display target for the Switch screen -struct FramebufferTextureInfo { - GLuint display_texture{}; - u32 width; - u32 height; - u32 scaled_width; - u32 scaled_height; -}; +class BlitScreen; class RendererOpenGL final : public VideoCore::RendererBase { public: @@ -77,24 +52,9 @@ public: } private: - /// Initializes the OpenGL state and creates persistent objects. - void InitOpenGLObjects(); - void AddTelemetryFields(); - - void ConfigureFramebufferTexture(const Tegra::FramebufferConfig& framebuffer); - - /// Draws the emulated screens to the emulator window. - void DrawScreen(const Tegra::FramebufferConfig& framebuffer, - const Layout::FramebufferLayout& layout); - void RenderScreenshot(const Tegra::FramebufferConfig& framebuffer); - /// Loads framebuffer from emulated memory into the active OpenGL texture. - FramebufferTextureInfo LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer); - - FramebufferTextureInfo PrepareRenderTarget(const Tegra::FramebufferConfig& framebuffer); - Core::TelemetrySession& telemetry_session; Core::Frontend::EmuWindow& emu_window; Tegra::MaxwellDeviceMemoryManager& device_memory; @@ -104,43 +64,9 @@ private: StateTracker state_tracker; ProgramManager program_manager; RasterizerOpenGL rasterizer; - - // OpenGL object IDs - OGLSampler present_sampler; - OGLSampler present_sampler_nn; - OGLBuffer vertex_buffer; - OGLProgram fxaa_vertex; - OGLProgram fxaa_fragment; - OGLProgram present_vertex; - OGLProgram present_bilinear_fragment; - OGLProgram present_bicubic_fragment; - OGLProgram present_gaussian_fragment; - OGLProgram present_scaleforce_fragment; OGLFramebuffer screenshot_framebuffer; - // GPU address of the vertex buffer - GLuint64EXT vertex_buffer_address = 0; - - /// Display information for Switch screen - TextureInfo framebuffer_texture; - OGLTexture aa_texture; - OGLFramebuffer aa_framebuffer; - - OGLProgram smaa_edge_detection_vert; - OGLProgram smaa_blending_weight_calculation_vert; - OGLProgram smaa_neighborhood_blending_vert; - OGLProgram smaa_edge_detection_frag; - OGLProgram smaa_blending_weight_calculation_frag; - OGLProgram smaa_neighborhood_blending_frag; - OGLTexture smaa_area_tex; - OGLTexture smaa_search_tex; - OGLTexture smaa_edges_tex; - OGLTexture smaa_blend_tex; - - std::unique_ptr fsr; - - /// OpenGL framebuffer data - std::vector gl_framebuffer_data; + std::unique_ptr blit_screen; }; } // namespace OpenGL -- cgit v1.2.3 From b90eff4bc666548a77eb58ac152408c80ff952b3 Mon Sep 17 00:00:00 2001 From: Liam Date: Sun, 14 Jan 2024 21:11:28 -0500 Subject: renderer_opengl: split out SMAA --- src/video_core/CMakeLists.txt | 7 +- src/video_core/renderer_opengl/gl_blit_screen.cpp | 90 +++-------------- src/video_core/renderer_opengl/gl_blit_screen.h | 17 +--- src/video_core/renderer_opengl/gl_fsr.cpp | 106 -------------------- src/video_core/renderer_opengl/gl_fsr.h | 43 -------- src/video_core/renderer_opengl/present/fsr.cpp | 106 ++++++++++++++++++++ src/video_core/renderer_opengl/present/fsr.h | 43 ++++++++ src/video_core/renderer_opengl/present/smaa.cpp | 108 +++++++++++++++++++++ src/video_core/renderer_opengl/present/smaa.h | 35 +++++++ src/video_core/renderer_opengl/present/util.h | 32 ++++++ src/video_core/renderer_opengl/renderer_opengl.cpp | 1 - src/video_core/renderer_opengl/renderer_opengl.h | 1 - 12 files changed, 345 insertions(+), 244 deletions(-) delete mode 100644 src/video_core/renderer_opengl/gl_fsr.cpp delete mode 100644 src/video_core/renderer_opengl/gl_fsr.h create mode 100644 src/video_core/renderer_opengl/present/fsr.cpp create mode 100644 src/video_core/renderer_opengl/present/fsr.h create mode 100644 src/video_core/renderer_opengl/present/smaa.cpp create mode 100644 src/video_core/renderer_opengl/present/smaa.h create mode 100644 src/video_core/renderer_opengl/present/util.h (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index 825815ebd..524e2cae8 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -116,6 +116,11 @@ add_library(video_core STATIC renderer_null/null_rasterizer.h renderer_null/renderer_null.cpp renderer_null/renderer_null.h + renderer_opengl/present/fsr.cpp + renderer_opengl/present/fsr.h + renderer_opengl/present/smaa.cpp + renderer_opengl/present/smaa.h + renderer_opengl/present/util.h renderer_opengl/blit_image.cpp renderer_opengl/blit_image.h renderer_opengl/gl_blit_screen.cpp @@ -129,8 +134,6 @@ add_library(video_core STATIC renderer_opengl/gl_device.h renderer_opengl/gl_fence_manager.cpp renderer_opengl/gl_fence_manager.h - renderer_opengl/gl_fsr.cpp - renderer_opengl/gl_fsr.h renderer_opengl/gl_graphics_pipeline.cpp renderer_opengl/gl_graphics_pipeline.h renderer_opengl/gl_rasterizer.cpp diff --git a/src/video_core/renderer_opengl/gl_blit_screen.cpp b/src/video_core/renderer_opengl/gl_blit_screen.cpp index 88757ba38..cc343f171 100644 --- a/src/video_core/renderer_opengl/gl_blit_screen.cpp +++ b/src/video_core/renderer_opengl/gl_blit_screen.cpp @@ -13,22 +13,16 @@ #include "video_core/host_shaders/opengl_present_frag.h" #include "video_core/host_shaders/opengl_present_scaleforce_frag.h" #include "video_core/host_shaders/opengl_present_vert.h" -#include "video_core/host_shaders/opengl_smaa_glsl.h" #include "video_core/host_shaders/present_bicubic_frag.h" #include "video_core/host_shaders/present_gaussian_frag.h" -#include "video_core/host_shaders/smaa_blending_weight_calculation_frag.h" -#include "video_core/host_shaders/smaa_blending_weight_calculation_vert.h" -#include "video_core/host_shaders/smaa_edge_detection_frag.h" -#include "video_core/host_shaders/smaa_edge_detection_vert.h" -#include "video_core/host_shaders/smaa_neighborhood_blending_frag.h" -#include "video_core/host_shaders/smaa_neighborhood_blending_vert.h" + #include "video_core/renderer_opengl/gl_blit_screen.h" #include "video_core/renderer_opengl/gl_rasterizer.h" #include "video_core/renderer_opengl/gl_shader_manager.h" #include "video_core/renderer_opengl/gl_shader_util.h" #include "video_core/renderer_opengl/gl_state_tracker.h" -#include "video_core/smaa_area_tex.h" -#include "video_core/smaa_search_tex.h" +#include "video_core/renderer_opengl/present/fsr.h" +#include "video_core/renderer_opengl/present/smaa.h" #include "video_core/textures/decoders.h" namespace OpenGL { @@ -84,24 +78,6 @@ BlitScreen::BlitScreen(RasterizerOpenGL& rasterizer_, shader_source.replace(pos, include_string.size(), include_content); }; - const auto SmaaShader = [&](std::string_view specialized_source, GLenum stage) { - std::string shader_source{specialized_source}; - replace_include(shader_source, "opengl_smaa.glsl", HostShaders::OPENGL_SMAA_GLSL); - return CreateProgram(shader_source, stage); - }; - - smaa_edge_detection_vert = SmaaShader(HostShaders::SMAA_EDGE_DETECTION_VERT, GL_VERTEX_SHADER); - smaa_edge_detection_frag = - SmaaShader(HostShaders::SMAA_EDGE_DETECTION_FRAG, GL_FRAGMENT_SHADER); - smaa_blending_weight_calculation_vert = - SmaaShader(HostShaders::SMAA_BLENDING_WEIGHT_CALCULATION_VERT, GL_VERTEX_SHADER); - smaa_blending_weight_calculation_frag = - SmaaShader(HostShaders::SMAA_BLENDING_WEIGHT_CALCULATION_FRAG, GL_FRAGMENT_SHADER); - smaa_neighborhood_blending_vert = - SmaaShader(HostShaders::SMAA_NEIGHBORHOOD_BLENDING_VERT, GL_VERTEX_SHADER); - smaa_neighborhood_blending_frag = - SmaaShader(HostShaders::SMAA_NEIGHBORHOOD_BLENDING_FRAG, GL_FRAGMENT_SHADER); - present_vertex = CreateProgram(HostShaders::OPENGL_PRESENT_VERT, GL_VERTEX_SHADER); present_bilinear_fragment = CreateProgram(HostShaders::OPENGL_PRESENT_FRAG, GL_FRAGMENT_SHADER); present_bicubic_fragment = CreateProgram(HostShaders::PRESENT_BICUBIC_FRAG, GL_FRAGMENT_SHADER); @@ -157,15 +133,6 @@ BlitScreen::BlitScreen(RasterizerOpenGL& rasterizer_, aa_framebuffer.Create(); - smaa_area_tex.Create(GL_TEXTURE_2D); - glTextureStorage2D(smaa_area_tex.handle, 1, GL_RG8, AREATEX_WIDTH, AREATEX_HEIGHT); - glTextureSubImage2D(smaa_area_tex.handle, 0, 0, 0, AREATEX_WIDTH, AREATEX_HEIGHT, GL_RG, - GL_UNSIGNED_BYTE, areaTexBytes); - smaa_search_tex.Create(GL_TEXTURE_2D); - glTextureStorage2D(smaa_search_tex.handle, 1, GL_R8, SEARCHTEX_WIDTH, SEARCHTEX_HEIGHT); - glTextureSubImage2D(smaa_search_tex.handle, 0, 0, 0, SEARCHTEX_WIDTH, SEARCHTEX_HEIGHT, GL_RED, - GL_UNSIGNED_BYTE, searchTexBytes); - // Enable unified vertex attributes and query vertex buffer address when the driver supports it if (device.HasVertexBufferUnifiedMemory()) { glEnableClientState(GL_VERTEX_ATTRIB_ARRAY_UNIFIED_NV); @@ -176,6 +143,8 @@ BlitScreen::BlitScreen(RasterizerOpenGL& rasterizer_, } } +BlitScreen::~BlitScreen() = default; + FramebufferTextureInfo BlitScreen::PrepareRenderTarget( const Tegra::FramebufferConfig& framebuffer) { // If framebuffer is provided, reload it from memory to a texture @@ -281,16 +250,10 @@ void BlitScreen::ConfigureFramebufferTexture(const Tegra::FramebufferConfig& fra Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, aa_texture.handle, 0); - smaa_edges_tex.Release(); - smaa_edges_tex.Create(GL_TEXTURE_2D); - glTextureStorage2D(smaa_edges_tex.handle, 1, GL_RG16F, - Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), - Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); - smaa_blend_tex.Release(); - smaa_blend_tex.Create(GL_TEXTURE_2D); - glTextureStorage2D(smaa_blend_tex.handle, 1, GL_RGBA16F, - Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), - Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); + + smaa = std::make_unique( + Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), + Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); } void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, @@ -363,39 +326,10 @@ void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, program_manager.BindPresentPrograms(fxaa_vertex.handle, fxaa_fragment.handle); glBindFramebuffer(GL_DRAW_FRAMEBUFFER, aa_framebuffer.handle); glDrawArrays(GL_TRIANGLE_STRIP, 0, 4); + glBindTextureUnit(0, aa_texture.handle); } break; case Settings::AntiAliasing::Smaa: { - glClearColor(0, 0, 0, 0); - glFrontFace(GL_CCW); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, aa_framebuffer.handle); - glBindSampler(1, present_sampler.handle); - glBindSampler(2, present_sampler.handle); - - glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, - smaa_edges_tex.handle, 0); - glClear(GL_COLOR_BUFFER_BIT); - program_manager.BindPresentPrograms(smaa_edge_detection_vert.handle, - smaa_edge_detection_frag.handle); - glDrawArrays(GL_TRIANGLES, 0, 3); - - glBindTextureUnit(0, smaa_edges_tex.handle); - glBindTextureUnit(1, smaa_area_tex.handle); - glBindTextureUnit(2, smaa_search_tex.handle); - glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, - smaa_blend_tex.handle, 0); - glClear(GL_COLOR_BUFFER_BIT); - program_manager.BindPresentPrograms(smaa_blending_weight_calculation_vert.handle, - smaa_blending_weight_calculation_frag.handle); - glDrawArrays(GL_TRIANGLES, 0, 3); - - glBindTextureUnit(0, info.display_texture); - glBindTextureUnit(1, smaa_blend_tex.handle); - glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, - aa_texture.handle, 0); - program_manager.BindPresentPrograms(smaa_neighborhood_blending_vert.handle, - smaa_neighborhood_blending_frag.handle); - glDrawArrays(GL_TRIANGLES, 0, 3); - glFrontFace(GL_CW); + glBindTextureUnit(0, smaa->Draw(program_manager, info.display_texture)); } break; default: UNREACHABLE(); @@ -403,8 +337,6 @@ void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, glBindFramebuffer(GL_READ_FRAMEBUFFER, old_read_fb); glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); - - glBindTextureUnit(0, aa_texture.handle); } glDisablei(GL_SCISSOR_TEST, 0); diff --git a/src/video_core/renderer_opengl/gl_blit_screen.h b/src/video_core/renderer_opengl/gl_blit_screen.h index 13d769958..945c7226a 100644 --- a/src/video_core/renderer_opengl/gl_blit_screen.h +++ b/src/video_core/renderer_opengl/gl_blit_screen.h @@ -8,7 +8,6 @@ #include "core/hle/service/nvnflinger/pixel_format.h" #include "video_core/host1x/gpu_device_memory_manager.h" -#include "video_core/renderer_opengl/gl_fsr.h" #include "video_core/renderer_opengl/gl_resource_manager.h" namespace Layout { @@ -22,7 +21,10 @@ struct FramebufferConfig; namespace OpenGL { class Device; +class FSR; +class ProgramManager; class RasterizerOpenGL; +class SMAA; class StateTracker; /// Structure used for storing information about the textures for the Switch screen @@ -50,6 +52,7 @@ public: Tegra::MaxwellDeviceMemoryManager& device_memory, StateTracker& state_tracker, ProgramManager& program_manager, Device& device); + ~BlitScreen(); void ConfigureFramebufferTexture(const Tegra::FramebufferConfig& framebuffer); @@ -87,18 +90,8 @@ private: OGLTexture aa_texture; OGLFramebuffer aa_framebuffer; - OGLProgram smaa_edge_detection_vert; - OGLProgram smaa_blending_weight_calculation_vert; - OGLProgram smaa_neighborhood_blending_vert; - OGLProgram smaa_edge_detection_frag; - OGLProgram smaa_blending_weight_calculation_frag; - OGLProgram smaa_neighborhood_blending_frag; - OGLTexture smaa_area_tex; - OGLTexture smaa_search_tex; - OGLTexture smaa_edges_tex; - OGLTexture smaa_blend_tex; - std::unique_ptr fsr; + std::unique_ptr smaa; /// OpenGL framebuffer data std::vector gl_framebuffer_data; diff --git a/src/video_core/renderer_opengl/gl_fsr.cpp b/src/video_core/renderer_opengl/gl_fsr.cpp deleted file mode 100644 index 429dcdc6c..000000000 --- a/src/video_core/renderer_opengl/gl_fsr.cpp +++ /dev/null @@ -1,106 +0,0 @@ -// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project -// SPDX-License-Identifier: GPL-2.0-or-later - -#include "common/settings.h" -#include "video_core/fsr.h" -#include "video_core/renderer_opengl/gl_fsr.h" -#include "video_core/renderer_opengl/gl_shader_manager.h" -#include "video_core/renderer_opengl/gl_shader_util.h" - -namespace OpenGL { -using namespace FSR; - -using FsrConstants = std::array; - -FSR::FSR(std::string_view fsr_vertex_source, std::string_view fsr_easu_source, - std::string_view fsr_rcas_source) - : fsr_vertex{CreateProgram(fsr_vertex_source, GL_VERTEX_SHADER)}, - fsr_easu_frag{CreateProgram(fsr_easu_source, GL_FRAGMENT_SHADER)}, - fsr_rcas_frag{CreateProgram(fsr_rcas_source, GL_FRAGMENT_SHADER)} { - glProgramUniform2f(fsr_vertex.handle, 0, 1.0f, 1.0f); - glProgramUniform2f(fsr_vertex.handle, 1, 0.0f, 0.0f); -} - -FSR::~FSR() = default; - -void FSR::Draw(ProgramManager& program_manager, const Common::Rectangle& screen, - u32 input_image_width, u32 input_image_height, - const Common::Rectangle& crop_rect) { - - const auto output_image_width = screen.GetWidth(); - const auto output_image_height = screen.GetHeight(); - - if (fsr_intermediate_tex.handle) { - GLint fsr_tex_width, fsr_tex_height; - glGetTextureLevelParameteriv(fsr_intermediate_tex.handle, 0, GL_TEXTURE_WIDTH, - &fsr_tex_width); - glGetTextureLevelParameteriv(fsr_intermediate_tex.handle, 0, GL_TEXTURE_HEIGHT, - &fsr_tex_height); - if (static_cast(fsr_tex_width) != output_image_width || - static_cast(fsr_tex_height) != output_image_height) { - fsr_intermediate_tex.Release(); - } - } - if (!fsr_intermediate_tex.handle) { - fsr_intermediate_tex.Create(GL_TEXTURE_2D); - glTextureStorage2D(fsr_intermediate_tex.handle, 1, GL_RGB16F, output_image_width, - output_image_height); - glNamedFramebufferTexture(fsr_framebuffer.handle, GL_COLOR_ATTACHMENT0, - fsr_intermediate_tex.handle, 0); - } - - GLint old_draw_fb; - glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb); - - glFrontFace(GL_CW); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, fsr_framebuffer.handle); - glViewportIndexedf(0, 0.0f, 0.0f, static_cast(output_image_width), - static_cast(output_image_height)); - - const f32 input_width = static_cast(input_image_width); - const f32 input_height = static_cast(input_image_height); - const f32 output_width = static_cast(screen.GetWidth()); - const f32 output_height = static_cast(screen.GetHeight()); - const f32 viewport_width = (crop_rect.right - crop_rect.left) * input_width; - const f32 viewport_x = crop_rect.left * input_width; - const f32 viewport_height = (crop_rect.bottom - crop_rect.top) * input_height; - const f32 viewport_y = crop_rect.top * input_height; - - FsrConstants constants; - FsrEasuConOffset(constants.data() + 0, constants.data() + 4, constants.data() + 8, - constants.data() + 12, viewport_width, viewport_height, input_width, - input_height, output_width, output_height, viewport_x, viewport_y); - - glProgramUniform4uiv(fsr_easu_frag.handle, 0, sizeof(constants), std::data(constants)); - - program_manager.BindPresentPrograms(fsr_vertex.handle, fsr_easu_frag.handle); - glDrawArrays(GL_TRIANGLES, 0, 3); - - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); - glBindTextureUnit(0, fsr_intermediate_tex.handle); - - const float sharpening = - static_cast(Settings::values.fsr_sharpening_slider.GetValue()) / 100.0f; - - FsrRcasCon(constants.data(), sharpening); - glProgramUniform4uiv(fsr_rcas_frag.handle, 0, sizeof(constants), std::data(constants)); -} - -void FSR::InitBuffers() { - fsr_framebuffer.Create(); -} - -void FSR::ReleaseBuffers() { - fsr_framebuffer.Release(); - fsr_intermediate_tex.Release(); -} - -const OGLProgram& FSR::GetPresentFragmentProgram() const noexcept { - return fsr_rcas_frag; -} - -bool FSR::AreBuffersInitialized() const noexcept { - return fsr_framebuffer.handle; -} - -} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_fsr.h b/src/video_core/renderer_opengl/gl_fsr.h deleted file mode 100644 index a5092e396..000000000 --- a/src/video_core/renderer_opengl/gl_fsr.h +++ /dev/null @@ -1,43 +0,0 @@ -// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project -// SPDX-License-Identifier: GPL-2.0-or-later - -#pragma once - -#include - -#include "common/common_types.h" -#include "common/math_util.h" -#include "video_core/fsr.h" -#include "video_core/renderer_opengl/gl_resource_manager.h" - -namespace OpenGL { - -class ProgramManager; - -class FSR { -public: - explicit FSR(std::string_view fsr_vertex_source, std::string_view fsr_easu_source, - std::string_view fsr_rcas_source); - ~FSR(); - - void Draw(ProgramManager& program_manager, const Common::Rectangle& screen, - u32 input_image_width, u32 input_image_height, - const Common::Rectangle& crop_rect); - - void InitBuffers(); - - void ReleaseBuffers(); - - [[nodiscard]] const OGLProgram& GetPresentFragmentProgram() const noexcept; - - [[nodiscard]] bool AreBuffersInitialized() const noexcept; - -private: - OGLFramebuffer fsr_framebuffer; - OGLProgram fsr_vertex; - OGLProgram fsr_easu_frag; - OGLProgram fsr_rcas_frag; - OGLTexture fsr_intermediate_tex; -}; - -} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/fsr.cpp b/src/video_core/renderer_opengl/present/fsr.cpp new file mode 100644 index 000000000..e5945b80b --- /dev/null +++ b/src/video_core/renderer_opengl/present/fsr.cpp @@ -0,0 +1,106 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#include "common/settings.h" +#include "video_core/fsr.h" +#include "video_core/renderer_opengl/gl_shader_manager.h" +#include "video_core/renderer_opengl/gl_shader_util.h" +#include "video_core/renderer_opengl/present/fsr.h" + +namespace OpenGL { +using namespace FSR; + +using FsrConstants = std::array; + +FSR::FSR(std::string_view fsr_vertex_source, std::string_view fsr_easu_source, + std::string_view fsr_rcas_source) + : fsr_vertex{CreateProgram(fsr_vertex_source, GL_VERTEX_SHADER)}, + fsr_easu_frag{CreateProgram(fsr_easu_source, GL_FRAGMENT_SHADER)}, + fsr_rcas_frag{CreateProgram(fsr_rcas_source, GL_FRAGMENT_SHADER)} { + glProgramUniform2f(fsr_vertex.handle, 0, 1.0f, 1.0f); + glProgramUniform2f(fsr_vertex.handle, 1, 0.0f, 0.0f); +} + +FSR::~FSR() = default; + +void FSR::Draw(ProgramManager& program_manager, const Common::Rectangle& screen, + u32 input_image_width, u32 input_image_height, + const Common::Rectangle& crop_rect) { + + const auto output_image_width = screen.GetWidth(); + const auto output_image_height = screen.GetHeight(); + + if (fsr_intermediate_tex.handle) { + GLint fsr_tex_width, fsr_tex_height; + glGetTextureLevelParameteriv(fsr_intermediate_tex.handle, 0, GL_TEXTURE_WIDTH, + &fsr_tex_width); + glGetTextureLevelParameteriv(fsr_intermediate_tex.handle, 0, GL_TEXTURE_HEIGHT, + &fsr_tex_height); + if (static_cast(fsr_tex_width) != output_image_width || + static_cast(fsr_tex_height) != output_image_height) { + fsr_intermediate_tex.Release(); + } + } + if (!fsr_intermediate_tex.handle) { + fsr_intermediate_tex.Create(GL_TEXTURE_2D); + glTextureStorage2D(fsr_intermediate_tex.handle, 1, GL_RGB16F, output_image_width, + output_image_height); + glNamedFramebufferTexture(fsr_framebuffer.handle, GL_COLOR_ATTACHMENT0, + fsr_intermediate_tex.handle, 0); + } + + GLint old_draw_fb; + glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb); + + glFrontFace(GL_CW); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, fsr_framebuffer.handle); + glViewportIndexedf(0, 0.0f, 0.0f, static_cast(output_image_width), + static_cast(output_image_height)); + + const f32 input_width = static_cast(input_image_width); + const f32 input_height = static_cast(input_image_height); + const f32 output_width = static_cast(screen.GetWidth()); + const f32 output_height = static_cast(screen.GetHeight()); + const f32 viewport_width = (crop_rect.right - crop_rect.left) * input_width; + const f32 viewport_x = crop_rect.left * input_width; + const f32 viewport_height = (crop_rect.bottom - crop_rect.top) * input_height; + const f32 viewport_y = crop_rect.top * input_height; + + FsrConstants constants; + FsrEasuConOffset(constants.data() + 0, constants.data() + 4, constants.data() + 8, + constants.data() + 12, viewport_width, viewport_height, input_width, + input_height, output_width, output_height, viewport_x, viewport_y); + + glProgramUniform4uiv(fsr_easu_frag.handle, 0, sizeof(constants), std::data(constants)); + + program_manager.BindPresentPrograms(fsr_vertex.handle, fsr_easu_frag.handle); + glDrawArrays(GL_TRIANGLES, 0, 3); + + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); + glBindTextureUnit(0, fsr_intermediate_tex.handle); + + const float sharpening = + static_cast(Settings::values.fsr_sharpening_slider.GetValue()) / 100.0f; + + FsrRcasCon(constants.data(), sharpening); + glProgramUniform4uiv(fsr_rcas_frag.handle, 0, sizeof(constants), std::data(constants)); +} + +void FSR::InitBuffers() { + fsr_framebuffer.Create(); +} + +void FSR::ReleaseBuffers() { + fsr_framebuffer.Release(); + fsr_intermediate_tex.Release(); +} + +const OGLProgram& FSR::GetPresentFragmentProgram() const noexcept { + return fsr_rcas_frag; +} + +bool FSR::AreBuffersInitialized() const noexcept { + return fsr_framebuffer.handle; +} + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/fsr.h b/src/video_core/renderer_opengl/present/fsr.h new file mode 100644 index 000000000..a5092e396 --- /dev/null +++ b/src/video_core/renderer_opengl/present/fsr.h @@ -0,0 +1,43 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include + +#include "common/common_types.h" +#include "common/math_util.h" +#include "video_core/fsr.h" +#include "video_core/renderer_opengl/gl_resource_manager.h" + +namespace OpenGL { + +class ProgramManager; + +class FSR { +public: + explicit FSR(std::string_view fsr_vertex_source, std::string_view fsr_easu_source, + std::string_view fsr_rcas_source); + ~FSR(); + + void Draw(ProgramManager& program_manager, const Common::Rectangle& screen, + u32 input_image_width, u32 input_image_height, + const Common::Rectangle& crop_rect); + + void InitBuffers(); + + void ReleaseBuffers(); + + [[nodiscard]] const OGLProgram& GetPresentFragmentProgram() const noexcept; + + [[nodiscard]] bool AreBuffersInitialized() const noexcept; + +private: + OGLFramebuffer fsr_framebuffer; + OGLProgram fsr_vertex; + OGLProgram fsr_easu_frag; + OGLProgram fsr_rcas_frag; + OGLTexture fsr_intermediate_tex; +}; + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/smaa.cpp b/src/video_core/renderer_opengl/present/smaa.cpp new file mode 100644 index 000000000..a9a0eb6c6 --- /dev/null +++ b/src/video_core/renderer_opengl/present/smaa.cpp @@ -0,0 +1,108 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#include "video_core/host_shaders/opengl_smaa_glsl.h" +#include "video_core/host_shaders/smaa_blending_weight_calculation_frag.h" +#include "video_core/host_shaders/smaa_blending_weight_calculation_vert.h" +#include "video_core/host_shaders/smaa_edge_detection_frag.h" +#include "video_core/host_shaders/smaa_edge_detection_vert.h" +#include "video_core/host_shaders/smaa_neighborhood_blending_frag.h" +#include "video_core/host_shaders/smaa_neighborhood_blending_vert.h" +#include "video_core/renderer_opengl/gl_shader_manager.h" +#include "video_core/renderer_opengl/gl_shader_util.h" +#include "video_core/renderer_opengl/present/smaa.h" +#include "video_core/renderer_opengl/present/util.h" +#include "video_core/smaa_area_tex.h" +#include "video_core/smaa_search_tex.h" + +namespace OpenGL { + +SMAA::SMAA(u32 width, u32 height) { + const auto SmaaShader = [&](std::string_view specialized_source, GLenum stage) { + std::string shader_source{specialized_source}; + ReplaceInclude(shader_source, "opengl_smaa.glsl", HostShaders::OPENGL_SMAA_GLSL); + return CreateProgram(shader_source, stage); + }; + + edge_detection_vert = SmaaShader(HostShaders::SMAA_EDGE_DETECTION_VERT, GL_VERTEX_SHADER); + edge_detection_frag = SmaaShader(HostShaders::SMAA_EDGE_DETECTION_FRAG, GL_FRAGMENT_SHADER); + blending_weight_calculation_vert = + SmaaShader(HostShaders::SMAA_BLENDING_WEIGHT_CALCULATION_VERT, GL_VERTEX_SHADER); + blending_weight_calculation_frag = + SmaaShader(HostShaders::SMAA_BLENDING_WEIGHT_CALCULATION_FRAG, GL_FRAGMENT_SHADER); + neighborhood_blending_vert = + SmaaShader(HostShaders::SMAA_NEIGHBORHOOD_BLENDING_VERT, GL_VERTEX_SHADER); + neighborhood_blending_frag = + SmaaShader(HostShaders::SMAA_NEIGHBORHOOD_BLENDING_FRAG, GL_FRAGMENT_SHADER); + + glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0); + glPixelStorei(GL_UNPACK_SWAP_BYTES, GL_FALSE); + glPixelStorei(GL_UNPACK_LSB_FIRST, GL_FALSE); + glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); + glPixelStorei(GL_UNPACK_IMAGE_HEIGHT, 0); + glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0); + glPixelStorei(GL_UNPACK_SKIP_ROWS, 0); + glPixelStorei(GL_UNPACK_ALIGNMENT, 4); + + area_tex.Create(GL_TEXTURE_2D); + glTextureStorage2D(area_tex.handle, 1, GL_RG8, AREATEX_WIDTH, AREATEX_HEIGHT); + glTextureSubImage2D(area_tex.handle, 0, 0, 0, AREATEX_WIDTH, AREATEX_HEIGHT, GL_RG, + GL_UNSIGNED_BYTE, areaTexBytes); + search_tex.Create(GL_TEXTURE_2D); + glTextureStorage2D(search_tex.handle, 1, GL_R8, SEARCHTEX_WIDTH, SEARCHTEX_HEIGHT); + glTextureSubImage2D(search_tex.handle, 0, 0, 0, SEARCHTEX_WIDTH, SEARCHTEX_HEIGHT, GL_RED, + GL_UNSIGNED_BYTE, searchTexBytes); + + edges_tex.Create(GL_TEXTURE_2D); + glTextureStorage2D(edges_tex.handle, 1, GL_RG16F, width, height); + + blend_tex.Create(GL_TEXTURE_2D); + glTextureStorage2D(blend_tex.handle, 1, GL_RGBA16F, width, height); + + sampler = CreateBilinearSampler(); + + framebuffer.Create(); + + texture.Create(GL_TEXTURE_2D); + glTextureStorage2D(texture.handle, 1, GL_RGBA16F, width, height); + glNamedFramebufferTexture(framebuffer.handle, GL_COLOR_ATTACHMENT0, texture.handle, 0); +} + +SMAA::~SMAA() = default; + +GLuint SMAA::Draw(ProgramManager& program_manager, GLuint input_texture) { + glClearColor(0, 0, 0, 0); + glFrontFace(GL_CCW); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, framebuffer.handle); + glBindSampler(0, sampler.handle); + glBindSampler(1, sampler.handle); + glBindSampler(2, sampler.handle); + + glBindTextureUnit(0, input_texture); + glNamedFramebufferTexture(framebuffer.handle, GL_COLOR_ATTACHMENT0, edges_tex.handle, 0); + glClear(GL_COLOR_BUFFER_BIT); + program_manager.BindPresentPrograms(edge_detection_vert.handle, edge_detection_frag.handle); + glDrawArrays(GL_TRIANGLES, 0, 3); + + glBindTextureUnit(0, edges_tex.handle); + glBindTextureUnit(1, area_tex.handle); + glBindTextureUnit(2, search_tex.handle); + glNamedFramebufferTexture(framebuffer.handle, GL_COLOR_ATTACHMENT0, blend_tex.handle, 0); + glClear(GL_COLOR_BUFFER_BIT); + program_manager.BindPresentPrograms(blending_weight_calculation_vert.handle, + blending_weight_calculation_frag.handle); + glDrawArrays(GL_TRIANGLES, 0, 3); + + glBindTextureUnit(0, input_texture); + glBindTextureUnit(1, blend_tex.handle); + glNamedFramebufferTexture(framebuffer.handle, GL_COLOR_ATTACHMENT0, texture.handle, 0); + program_manager.BindPresentPrograms(neighborhood_blending_vert.handle, + neighborhood_blending_frag.handle); + glClear(GL_COLOR_BUFFER_BIT); + glDrawArrays(GL_TRIANGLES, 0, 3); + glFrontFace(GL_CW); + + return texture.handle; +} + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/smaa.h b/src/video_core/renderer_opengl/present/smaa.h new file mode 100644 index 000000000..a48cb4fa9 --- /dev/null +++ b/src/video_core/renderer_opengl/present/smaa.h @@ -0,0 +1,35 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include "video_core/renderer_opengl/gl_resource_manager.h" + +namespace OpenGL { + +class ProgramManager; + +class SMAA { +public: + explicit SMAA(u32 width, u32 height); + ~SMAA(); + + GLuint Draw(ProgramManager& program_manager, GLuint input_texture); + +private: + OGLProgram edge_detection_vert; + OGLProgram blending_weight_calculation_vert; + OGLProgram neighborhood_blending_vert; + OGLProgram edge_detection_frag; + OGLProgram blending_weight_calculation_frag; + OGLProgram neighborhood_blending_frag; + OGLTexture area_tex; + OGLTexture search_tex; + OGLTexture edges_tex; + OGLTexture blend_tex; + OGLSampler sampler; + OGLFramebuffer framebuffer; + OGLTexture texture; +}; + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/util.h b/src/video_core/renderer_opengl/present/util.h new file mode 100644 index 000000000..0aa8b110c --- /dev/null +++ b/src/video_core/renderer_opengl/present/util.h @@ -0,0 +1,32 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include + +#include "common/assert.h" +#include "video_core/renderer_opengl/gl_resource_manager.h" + +namespace OpenGL { + +static inline void ReplaceInclude(std::string& shader_source, std::string_view include_name, + std::string_view include_content) { + const std::string include_string = fmt::format("#include \"{}\"", include_name); + const std::size_t pos = shader_source.find(include_string); + ASSERT(pos != std::string::npos); + shader_source.replace(pos, include_string.size(), include_content); +}; + +static inline OGLSampler CreateBilinearSampler() { + OGLSampler sampler; + sampler.Create(); + glSamplerParameteri(sampler.handle, GL_TEXTURE_MIN_FILTER, GL_LINEAR); + glSamplerParameteri(sampler.handle, GL_TEXTURE_MAG_FILTER, GL_LINEAR); + glSamplerParameteri(sampler.handle, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glSamplerParameteri(sampler.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + glSamplerParameteri(sampler.handle, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE); + return sampler; +} + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index 38b0aacf4..3d75fd17a 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -17,7 +17,6 @@ #include "core/frontend/emu_window.h" #include "core/telemetry_session.h" #include "video_core/renderer_opengl/gl_blit_screen.h" -#include "video_core/renderer_opengl/gl_fsr.h" #include "video_core/renderer_opengl/gl_rasterizer.h" #include "video_core/renderer_opengl/gl_shader_manager.h" #include "video_core/renderer_opengl/gl_shader_util.h" diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h index 23aff055a..7ab163372 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.h +++ b/src/video_core/renderer_opengl/renderer_opengl.h @@ -10,7 +10,6 @@ #include "video_core/renderer_base.h" #include "video_core/renderer_opengl/gl_device.h" -#include "video_core/renderer_opengl/gl_fsr.h" #include "video_core/renderer_opengl/gl_rasterizer.h" #include "video_core/renderer_opengl/gl_resource_manager.h" #include "video_core/renderer_opengl/gl_shader_manager.h" -- cgit v1.2.3 From 60ee29aac386c31b9a9add6a12f051273fa45dae Mon Sep 17 00:00:00 2001 From: Liam Date: Mon, 15 Jan 2024 00:09:34 -0500 Subject: renderer_opengl: split out FXAA --- src/video_core/CMakeLists.txt | 2 ++ src/video_core/host_shaders/fxaa.vert | 4 +-- src/video_core/renderer_opengl/gl_blit_screen.cpp | 22 +++---------- src/video_core/renderer_opengl/gl_blit_screen.h | 6 ++-- src/video_core/renderer_opengl/present/fxaa.cpp | 40 +++++++++++++++++++++++ src/video_core/renderer_opengl/present/fxaa.h | 27 +++++++++++++++ src/video_core/renderer_vulkan/present/fxaa.cpp | 2 +- 7 files changed, 79 insertions(+), 24 deletions(-) create mode 100644 src/video_core/renderer_opengl/present/fxaa.cpp create mode 100644 src/video_core/renderer_opengl/present/fxaa.h (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index 524e2cae8..9879c3ad7 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -118,6 +118,8 @@ add_library(video_core STATIC renderer_null/renderer_null.h renderer_opengl/present/fsr.cpp renderer_opengl/present/fsr.h + renderer_opengl/present/fxaa.cpp + renderer_opengl/present/fxaa.h renderer_opengl/present/smaa.cpp renderer_opengl/present/smaa.h renderer_opengl/present/util.h diff --git a/src/video_core/host_shaders/fxaa.vert b/src/video_core/host_shaders/fxaa.vert index c2717d90d..223ab785e 100644 --- a/src/video_core/host_shaders/fxaa.vert +++ b/src/video_core/host_shaders/fxaa.vert @@ -7,8 +7,8 @@ out gl_PerVertex { vec4 gl_Position; }; -const vec2 vertices[4] = - vec2[4](vec2(-1.0, 1.0), vec2(1.0, 1.0), vec2(-1.0, -1.0), vec2(1.0, -1.0)); +const vec2 vertices[3] = + vec2[3](vec2(-1,-1), vec2(3,-1), vec2(-1, 3)); layout (location = 0) out vec4 posPos; diff --git a/src/video_core/renderer_opengl/gl_blit_screen.cpp b/src/video_core/renderer_opengl/gl_blit_screen.cpp index cc343f171..44f6a0922 100644 --- a/src/video_core/renderer_opengl/gl_blit_screen.cpp +++ b/src/video_core/renderer_opengl/gl_blit_screen.cpp @@ -5,8 +5,6 @@ #include "video_core/host_shaders/ffx_a_h.h" #include "video_core/host_shaders/ffx_fsr1_h.h" #include "video_core/host_shaders/full_screen_triangle_vert.h" -#include "video_core/host_shaders/fxaa_frag.h" -#include "video_core/host_shaders/fxaa_vert.h" #include "video_core/host_shaders/opengl_fidelityfx_fsr_easu_frag.h" #include "video_core/host_shaders/opengl_fidelityfx_fsr_frag.h" #include "video_core/host_shaders/opengl_fidelityfx_fsr_rcas_frag.h" @@ -22,6 +20,7 @@ #include "video_core/renderer_opengl/gl_shader_util.h" #include "video_core/renderer_opengl/gl_state_tracker.h" #include "video_core/renderer_opengl/present/fsr.h" +#include "video_core/renderer_opengl/present/fxaa.h" #include "video_core/renderer_opengl/present/smaa.h" #include "video_core/textures/decoders.h" @@ -67,9 +66,6 @@ BlitScreen::BlitScreen(RasterizerOpenGL& rasterizer_, : rasterizer(rasterizer_), device_memory(device_memory_), state_tracker(state_tracker_), program_manager(program_manager_), device(device_) { // Create shader programs - fxaa_vertex = CreateProgram(HostShaders::FXAA_VERT, GL_VERTEX_SHADER); - fxaa_fragment = CreateProgram(HostShaders::FXAA_FRAG, GL_FRAGMENT_SHADER); - const auto replace_include = [](std::string& shader_source, std::string_view include_name, std::string_view include_content) { const std::string include_string = fmt::format("#include \"{}\"", include_name); @@ -131,8 +127,6 @@ BlitScreen::BlitScreen(RasterizerOpenGL& rasterizer_, glClearTexImage(framebuffer_texture.resource.handle, 0, GL_RGBA, GL_UNSIGNED_BYTE, framebuffer_data); - aa_framebuffer.Create(); - // Enable unified vertex attributes and query vertex buffer address when the driver supports it if (device.HasVertexBufferUnifiedMemory()) { glEnableClientState(GL_VERTEX_ATTRIB_ARRAY_UNIFIED_NV); @@ -244,13 +238,10 @@ void BlitScreen::ConfigureFramebufferTexture(const Tegra::FramebufferConfig& fra framebuffer_texture.resource.Create(GL_TEXTURE_2D); glTextureStorage2D(framebuffer_texture.resource.handle, 1, internal_format, framebuffer_texture.width, framebuffer_texture.height); - aa_texture.Release(); - aa_texture.Create(GL_TEXTURE_2D); - glTextureStorage2D(aa_texture.handle, 1, GL_RGBA16F, - Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), - Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); - glNamedFramebufferTexture(aa_framebuffer.handle, GL_COLOR_ATTACHMENT0, aa_texture.handle, 0); + fxaa = std::make_unique( + Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), + Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); smaa = std::make_unique( Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); @@ -323,10 +314,7 @@ void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, switch (anti_aliasing) { case Settings::AntiAliasing::Fxaa: { - program_manager.BindPresentPrograms(fxaa_vertex.handle, fxaa_fragment.handle); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, aa_framebuffer.handle); - glDrawArrays(GL_TRIANGLE_STRIP, 0, 4); - glBindTextureUnit(0, aa_texture.handle); + glBindTextureUnit(0, fxaa->Draw(program_manager, info.display_texture)); } break; case Settings::AntiAliasing::Smaa: { glBindTextureUnit(0, smaa->Draw(program_manager, info.display_texture)); diff --git a/src/video_core/renderer_opengl/gl_blit_screen.h b/src/video_core/renderer_opengl/gl_blit_screen.h index 945c7226a..2cb9a5015 100644 --- a/src/video_core/renderer_opengl/gl_blit_screen.h +++ b/src/video_core/renderer_opengl/gl_blit_screen.h @@ -22,6 +22,7 @@ namespace OpenGL { class Device; class FSR; +class FXAA; class ProgramManager; class RasterizerOpenGL; class SMAA; @@ -77,8 +78,6 @@ private: OGLSampler present_sampler; OGLSampler present_sampler_nn; OGLBuffer vertex_buffer; - OGLProgram fxaa_vertex; - OGLProgram fxaa_fragment; OGLProgram present_vertex; OGLProgram present_bilinear_fragment; OGLProgram present_bicubic_fragment; @@ -87,10 +86,9 @@ private: /// Display information for Switch screen TextureInfo framebuffer_texture; - OGLTexture aa_texture; - OGLFramebuffer aa_framebuffer; std::unique_ptr fsr; + std::unique_ptr fxaa; std::unique_ptr smaa; /// OpenGL framebuffer data diff --git a/src/video_core/renderer_opengl/present/fxaa.cpp b/src/video_core/renderer_opengl/present/fxaa.cpp new file mode 100644 index 000000000..9425c42fa --- /dev/null +++ b/src/video_core/renderer_opengl/present/fxaa.cpp @@ -0,0 +1,40 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#include "video_core/host_shaders/fxaa_frag.h" +#include "video_core/host_shaders/fxaa_vert.h" +#include "video_core/renderer_opengl/gl_shader_manager.h" +#include "video_core/renderer_opengl/gl_shader_util.h" +#include "video_core/renderer_opengl/present/fxaa.h" +#include "video_core/renderer_opengl/present/util.h" + +namespace OpenGL { + +FXAA::FXAA(u32 width, u32 height) { + vert_shader = CreateProgram(HostShaders::FXAA_VERT, GL_VERTEX_SHADER); + frag_shader = CreateProgram(HostShaders::FXAA_FRAG, GL_FRAGMENT_SHADER); + + sampler = CreateBilinearSampler(); + + framebuffer.Create(); + + texture.Create(GL_TEXTURE_2D); + glTextureStorage2D(texture.handle, 1, GL_RGBA16F, width, height); + glNamedFramebufferTexture(framebuffer.handle, GL_COLOR_ATTACHMENT0, texture.handle, 0); +} + +FXAA::~FXAA() = default; + +GLuint FXAA::Draw(ProgramManager& program_manager, GLuint input_texture) { + glFrontFace(GL_CCW); + + program_manager.BindPresentPrograms(vert_shader.handle, frag_shader.handle); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, framebuffer.handle); + glBindTextureUnit(0, input_texture); + glDrawArrays(GL_TRIANGLES, 0, 3); + glFrontFace(GL_CW); + + return texture.handle; +} + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/fxaa.h b/src/video_core/renderer_opengl/present/fxaa.h new file mode 100644 index 000000000..b898198f1 --- /dev/null +++ b/src/video_core/renderer_opengl/present/fxaa.h @@ -0,0 +1,27 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include "video_core/renderer_opengl/gl_resource_manager.h" + +namespace OpenGL { + +class ProgramManager; + +class FXAA { +public: + explicit FXAA(u32 width, u32 height); + ~FXAA(); + + GLuint Draw(ProgramManager& program_manager, GLuint input_texture); + +private: + OGLProgram vert_shader; + OGLProgram frag_shader; + OGLSampler sampler; + OGLFramebuffer framebuffer; + OGLTexture texture; +}; + +} // namespace OpenGL diff --git a/src/video_core/renderer_vulkan/present/fxaa.cpp b/src/video_core/renderer_vulkan/present/fxaa.cpp index 6f87ddebb..6c772ada3 100644 --- a/src/video_core/renderer_vulkan/present/fxaa.cpp +++ b/src/video_core/renderer_vulkan/present/fxaa.cpp @@ -133,7 +133,7 @@ VkImageView FXAA::Draw(Scheduler& scheduler, size_t image_index, VkImage source_ BeginRenderPass(cmdbuf, renderpass, framebuffer, extent); cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline); cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set, {}); - cmdbuf.Draw(4, 1, 0, 0); + cmdbuf.Draw(3, 1, 0, 0); cmdbuf.EndRenderPass(); TransitionImageLayout(cmdbuf, output_image, VK_IMAGE_LAYOUT_GENERAL); }); -- cgit v1.2.3 From b78900e9567a4c98c44ef06b5088ccb507464d66 Mon Sep 17 00:00:00 2001 From: Liam Date: Mon, 15 Jan 2024 00:36:54 -0500 Subject: renderer_opengl: move out FSR shader source construction --- src/video_core/renderer_opengl/gl_blit_screen.cpp | 20 +---------------- src/video_core/renderer_opengl/present/fsr.cpp | 26 ++++++++++++++++++----- src/video_core/renderer_opengl/present/fsr.h | 3 +-- 3 files changed, 23 insertions(+), 26 deletions(-) (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/renderer_opengl/gl_blit_screen.cpp b/src/video_core/renderer_opengl/gl_blit_screen.cpp index 44f6a0922..4e9d80d10 100644 --- a/src/video_core/renderer_opengl/gl_blit_screen.cpp +++ b/src/video_core/renderer_opengl/gl_blit_screen.cpp @@ -66,14 +66,6 @@ BlitScreen::BlitScreen(RasterizerOpenGL& rasterizer_, : rasterizer(rasterizer_), device_memory(device_memory_), state_tracker(state_tracker_), program_manager(program_manager_), device(device_) { // Create shader programs - const auto replace_include = [](std::string& shader_source, std::string_view include_name, - std::string_view include_content) { - const std::string include_string = fmt::format("#include \"{}\"", include_name); - const std::size_t pos = shader_source.find(include_string); - ASSERT(pos != std::string::npos); - shader_source.replace(pos, include_string.size(), include_content); - }; - present_vertex = CreateProgram(HostShaders::OPENGL_PRESENT_VERT, GL_VERTEX_SHADER); present_bilinear_fragment = CreateProgram(HostShaders::OPENGL_PRESENT_FRAG, GL_FRAGMENT_SHADER); present_bicubic_fragment = CreateProgram(HostShaders::PRESENT_BICUBIC_FRAG, GL_FRAGMENT_SHADER); @@ -83,17 +75,7 @@ BlitScreen::BlitScreen(RasterizerOpenGL& rasterizer_, CreateProgram(fmt::format("#version 460\n{}", HostShaders::OPENGL_PRESENT_SCALEFORCE_FRAG), GL_FRAGMENT_SHADER); - std::string fsr_source{HostShaders::OPENGL_FIDELITYFX_FSR_FRAG}; - replace_include(fsr_source, "ffx_a.h", HostShaders::FFX_A_H); - replace_include(fsr_source, "ffx_fsr1.h", HostShaders::FFX_FSR1_H); - - std::string fsr_easu_frag_source{HostShaders::OPENGL_FIDELITYFX_FSR_EASU_FRAG}; - std::string fsr_rcas_frag_source{HostShaders::OPENGL_FIDELITYFX_FSR_RCAS_FRAG}; - replace_include(fsr_easu_frag_source, "opengl_fidelityfx_fsr.frag", fsr_source); - replace_include(fsr_rcas_frag_source, "opengl_fidelityfx_fsr.frag", fsr_source); - - fsr = std::make_unique(HostShaders::FULL_SCREEN_TRIANGLE_VERT, fsr_easu_frag_source, - fsr_rcas_frag_source); + fsr = std::make_unique(); // Generate presentation sampler present_sampler.Create(); diff --git a/src/video_core/renderer_opengl/present/fsr.cpp b/src/video_core/renderer_opengl/present/fsr.cpp index e5945b80b..a5540bb0c 100644 --- a/src/video_core/renderer_opengl/present/fsr.cpp +++ b/src/video_core/renderer_opengl/present/fsr.cpp @@ -3,20 +3,36 @@ #include "common/settings.h" #include "video_core/fsr.h" +#include "video_core/host_shaders/ffx_a_h.h" +#include "video_core/host_shaders/ffx_fsr1_h.h" +#include "video_core/host_shaders/full_screen_triangle_vert.h" +#include "video_core/host_shaders/opengl_fidelityfx_fsr_easu_frag.h" +#include "video_core/host_shaders/opengl_fidelityfx_fsr_frag.h" +#include "video_core/host_shaders/opengl_fidelityfx_fsr_rcas_frag.h" #include "video_core/renderer_opengl/gl_shader_manager.h" #include "video_core/renderer_opengl/gl_shader_util.h" #include "video_core/renderer_opengl/present/fsr.h" +#include "video_core/renderer_opengl/present/util.h" namespace OpenGL { using namespace FSR; using FsrConstants = std::array; -FSR::FSR(std::string_view fsr_vertex_source, std::string_view fsr_easu_source, - std::string_view fsr_rcas_source) - : fsr_vertex{CreateProgram(fsr_vertex_source, GL_VERTEX_SHADER)}, - fsr_easu_frag{CreateProgram(fsr_easu_source, GL_FRAGMENT_SHADER)}, - fsr_rcas_frag{CreateProgram(fsr_rcas_source, GL_FRAGMENT_SHADER)} { +FSR::FSR() { + std::string fsr_source{HostShaders::OPENGL_FIDELITYFX_FSR_FRAG}; + ReplaceInclude(fsr_source, "ffx_a.h", HostShaders::FFX_A_H); + ReplaceInclude(fsr_source, "ffx_fsr1.h", HostShaders::FFX_FSR1_H); + + std::string fsr_easu_source{HostShaders::OPENGL_FIDELITYFX_FSR_EASU_FRAG}; + std::string fsr_rcas_source{HostShaders::OPENGL_FIDELITYFX_FSR_RCAS_FRAG}; + ReplaceInclude(fsr_easu_source, "opengl_fidelityfx_fsr.frag", fsr_source); + ReplaceInclude(fsr_rcas_source, "opengl_fidelityfx_fsr.frag", fsr_source); + + fsr_vertex = CreateProgram(HostShaders::FULL_SCREEN_TRIANGLE_VERT, GL_VERTEX_SHADER); + fsr_easu_frag = CreateProgram(fsr_easu_source, GL_FRAGMENT_SHADER); + fsr_rcas_frag = CreateProgram(fsr_rcas_source, GL_FRAGMENT_SHADER); + glProgramUniform2f(fsr_vertex.handle, 0, 1.0f, 1.0f); glProgramUniform2f(fsr_vertex.handle, 1, 0.0f, 0.0f); } diff --git a/src/video_core/renderer_opengl/present/fsr.h b/src/video_core/renderer_opengl/present/fsr.h index a5092e396..fa57c6f00 100644 --- a/src/video_core/renderer_opengl/present/fsr.h +++ b/src/video_core/renderer_opengl/present/fsr.h @@ -16,8 +16,7 @@ class ProgramManager; class FSR { public: - explicit FSR(std::string_view fsr_vertex_source, std::string_view fsr_easu_source, - std::string_view fsr_rcas_source); + explicit FSR(); ~FSR(); void Draw(ProgramManager& program_manager, const Common::Rectangle& screen, -- cgit v1.2.3 From dd2918efd83b586861ebc463dfee20c35e9d3bb3 Mon Sep 17 00:00:00 2001 From: Liam Date: Mon, 15 Jan 2024 14:28:03 -0500 Subject: renderer_opengl: move out ownership of FSR resources --- src/video_core/renderer_opengl/gl_blit_screen.cpp | 35 +++---- src/video_core/renderer_opengl/present/fsr.cpp | 110 +++++++++------------- src/video_core/renderer_opengl/present/fsr.h | 29 +++--- 3 files changed, 74 insertions(+), 100 deletions(-) (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/renderer_opengl/gl_blit_screen.cpp b/src/video_core/renderer_opengl/gl_blit_screen.cpp index 4e9d80d10..5f6221b9b 100644 --- a/src/video_core/renderer_opengl/gl_blit_screen.cpp +++ b/src/video_core/renderer_opengl/gl_blit_screen.cpp @@ -75,8 +75,6 @@ BlitScreen::BlitScreen(RasterizerOpenGL& rasterizer_, CreateProgram(fmt::format("#version 460\n{}", HostShaders::OPENGL_PRESENT_SCALEFORCE_FRAG), GL_FRAGMENT_SHADER); - fsr = std::make_unique(); - // Generate presentation sampler present_sampler.Create(); glSamplerParameteri(present_sampler.handle, GL_TEXTURE_MIN_FILTER, GL_LINEAR); @@ -269,7 +267,7 @@ void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, glColorMaski(0, GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); glDepthRangeIndexed(0, 0.0, 0.0); - glBindTextureUnit(0, info.display_texture); + GLuint texture = info.display_texture; auto anti_aliasing = Settings::values.anti_aliasing.GetValue(); if (anti_aliasing >= Settings::AntiAliasing::MaxEnum) { @@ -296,10 +294,10 @@ void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, switch (anti_aliasing) { case Settings::AntiAliasing::Fxaa: { - glBindTextureUnit(0, fxaa->Draw(program_manager, info.display_texture)); + texture = fxaa->Draw(program_manager, info.display_texture); } break; case Settings::AntiAliasing::Smaa: { - glBindTextureUnit(0, smaa->Draw(program_manager, info.display_texture)); + texture = smaa->Draw(program_manager, info.display_texture); } break; default: UNREACHABLE(); @@ -311,34 +309,37 @@ void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, glDisablei(GL_SCISSOR_TEST, 0); if (Settings::values.scaling_filter.GetValue() == Settings::ScalingFilter::Fsr) { - if (!fsr->AreBuffersInitialized()) { - fsr->InitBuffers(); - } + GLint old_read_fb; + GLint old_draw_fb; + glGetIntegerv(GL_READ_FRAMEBUFFER_BINDING, &old_read_fb); + glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb); - glBindSampler(0, present_sampler.handle); - fsr->Draw(program_manager, layout.screen, info.scaled_width, info.scaled_height, crop); - } else { - if (fsr->AreBuffersInitialized()) { - fsr->ReleaseBuffers(); + if (!fsr || fsr->NeedsRecreation(layout.screen)) { + fsr = std::make_unique(layout.screen.GetWidth(), layout.screen.GetHeight()); } + + texture = fsr->Draw(program_manager, texture, info.scaled_width, info.scaled_height, crop); + + glBindFramebuffer(GL_READ_FRAMEBUFFER, old_read_fb); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); } + glBindTextureUnit(0, texture); + const std::array ortho_matrix = MakeOrthographicMatrix(static_cast(layout.width), static_cast(layout.height)); const auto fragment_handle = [this]() { switch (Settings::values.scaling_filter.GetValue()) { - case Settings::ScalingFilter::NearestNeighbor: - case Settings::ScalingFilter::Bilinear: - return present_bilinear_fragment.handle; case Settings::ScalingFilter::Bicubic: return present_bicubic_fragment.handle; case Settings::ScalingFilter::Gaussian: return present_gaussian_fragment.handle; case Settings::ScalingFilter::ScaleForce: return present_scaleforce_fragment.handle; + case Settings::ScalingFilter::NearestNeighbor: + case Settings::ScalingFilter::Bilinear: case Settings::ScalingFilter::Fsr: - return fsr->GetPresentFragmentProgram().handle; default: return present_bilinear_fragment.handle; } diff --git a/src/video_core/renderer_opengl/present/fsr.cpp b/src/video_core/renderer_opengl/present/fsr.cpp index a5540bb0c..b764aadae 100644 --- a/src/video_core/renderer_opengl/present/fsr.cpp +++ b/src/video_core/renderer_opengl/present/fsr.cpp @@ -19,7 +19,7 @@ using namespace FSR; using FsrConstants = std::array; -FSR::FSR() { +FSR::FSR(u32 output_width_, u32 output_height_) : width(output_width_), height(output_height_) { std::string fsr_source{HostShaders::OPENGL_FIDELITYFX_FSR_FRAG}; ReplaceInclude(fsr_source, "ffx_a.h", HostShaders::FFX_A_H); ReplaceInclude(fsr_source, "ffx_fsr1.h", HostShaders::FFX_FSR1_H); @@ -29,94 +29,70 @@ FSR::FSR() { ReplaceInclude(fsr_easu_source, "opengl_fidelityfx_fsr.frag", fsr_source); ReplaceInclude(fsr_rcas_source, "opengl_fidelityfx_fsr.frag", fsr_source); - fsr_vertex = CreateProgram(HostShaders::FULL_SCREEN_TRIANGLE_VERT, GL_VERTEX_SHADER); - fsr_easu_frag = CreateProgram(fsr_easu_source, GL_FRAGMENT_SHADER); - fsr_rcas_frag = CreateProgram(fsr_rcas_source, GL_FRAGMENT_SHADER); + vert = CreateProgram(HostShaders::FULL_SCREEN_TRIANGLE_VERT, GL_VERTEX_SHADER); + easu_frag = CreateProgram(fsr_easu_source, GL_FRAGMENT_SHADER); + rcas_frag = CreateProgram(fsr_rcas_source, GL_FRAGMENT_SHADER); - glProgramUniform2f(fsr_vertex.handle, 0, 1.0f, 1.0f); - glProgramUniform2f(fsr_vertex.handle, 1, 0.0f, 0.0f); -} + glProgramUniform2f(vert.handle, 0, 1.0f, -1.0f); + glProgramUniform2f(vert.handle, 1, 0.0f, 1.0f); -FSR::~FSR() = default; + sampler = CreateBilinearSampler(); + framebuffer.Create(); -void FSR::Draw(ProgramManager& program_manager, const Common::Rectangle& screen, - u32 input_image_width, u32 input_image_height, - const Common::Rectangle& crop_rect) { - - const auto output_image_width = screen.GetWidth(); - const auto output_image_height = screen.GetHeight(); - - if (fsr_intermediate_tex.handle) { - GLint fsr_tex_width, fsr_tex_height; - glGetTextureLevelParameteriv(fsr_intermediate_tex.handle, 0, GL_TEXTURE_WIDTH, - &fsr_tex_width); - glGetTextureLevelParameteriv(fsr_intermediate_tex.handle, 0, GL_TEXTURE_HEIGHT, - &fsr_tex_height); - if (static_cast(fsr_tex_width) != output_image_width || - static_cast(fsr_tex_height) != output_image_height) { - fsr_intermediate_tex.Release(); - } - } - if (!fsr_intermediate_tex.handle) { - fsr_intermediate_tex.Create(GL_TEXTURE_2D); - glTextureStorage2D(fsr_intermediate_tex.handle, 1, GL_RGB16F, output_image_width, - output_image_height); - glNamedFramebufferTexture(fsr_framebuffer.handle, GL_COLOR_ATTACHMENT0, - fsr_intermediate_tex.handle, 0); - } - - GLint old_draw_fb; - glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb); + easu_tex.Create(GL_TEXTURE_2D); + glTextureStorage2D(easu_tex.handle, 1, GL_RGBA16F, width, height); - glFrontFace(GL_CW); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, fsr_framebuffer.handle); - glViewportIndexedf(0, 0.0f, 0.0f, static_cast(output_image_width), - static_cast(output_image_height)); + rcas_tex.Create(GL_TEXTURE_2D); + glTextureStorage2D(rcas_tex.handle, 1, GL_RGBA16F, width, height); +} +FSR::~FSR() = default; + +GLuint FSR::Draw(ProgramManager& program_manager, GLuint texture, u32 input_image_width, + u32 input_image_height, const Common::Rectangle& crop_rect) { const f32 input_width = static_cast(input_image_width); const f32 input_height = static_cast(input_image_height); - const f32 output_width = static_cast(screen.GetWidth()); - const f32 output_height = static_cast(screen.GetHeight()); + const f32 output_width = static_cast(width); + const f32 output_height = static_cast(height); const f32 viewport_width = (crop_rect.right - crop_rect.left) * input_width; const f32 viewport_x = crop_rect.left * input_width; const f32 viewport_height = (crop_rect.bottom - crop_rect.top) * input_height; const f32 viewport_y = crop_rect.top * input_height; - FsrConstants constants; - FsrEasuConOffset(constants.data() + 0, constants.data() + 4, constants.data() + 8, - constants.data() + 12, viewport_width, viewport_height, input_width, - input_height, output_width, output_height, viewport_x, viewport_y); - - glProgramUniform4uiv(fsr_easu_frag.handle, 0, sizeof(constants), std::data(constants)); - - program_manager.BindPresentPrograms(fsr_vertex.handle, fsr_easu_frag.handle); - glDrawArrays(GL_TRIANGLES, 0, 3); + FsrConstants easu_con{}; + FsrConstants rcas_con{}; - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); - glBindTextureUnit(0, fsr_intermediate_tex.handle); + FsrEasuConOffset(easu_con.data() + 0, easu_con.data() + 4, easu_con.data() + 8, + easu_con.data() + 12, viewport_width, viewport_height, input_width, + input_height, output_width, output_height, viewport_x, viewport_y); const float sharpening = static_cast(Settings::values.fsr_sharpening_slider.GetValue()) / 100.0f; - FsrRcasCon(constants.data(), sharpening); - glProgramUniform4uiv(fsr_rcas_frag.handle, 0, sizeof(constants), std::data(constants)); -} + FsrRcasCon(rcas_con.data(), sharpening); -void FSR::InitBuffers() { - fsr_framebuffer.Create(); -} + glProgramUniform4uiv(easu_frag.handle, 0, sizeof(easu_con), easu_con.data()); + glProgramUniform4uiv(rcas_frag.handle, 0, sizeof(rcas_con), rcas_con.data()); -void FSR::ReleaseBuffers() { - fsr_framebuffer.Release(); - fsr_intermediate_tex.Release(); -} + glFrontFace(GL_CW); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, framebuffer.handle); + glNamedFramebufferTexture(framebuffer.handle, GL_COLOR_ATTACHMENT0, easu_tex.handle, 0); + glViewportIndexedf(0, 0.0f, 0.0f, output_width, output_height); + program_manager.BindPresentPrograms(vert.handle, easu_frag.handle); + glBindTextureUnit(0, texture); + glBindSampler(0, sampler.handle); + glDrawArrays(GL_TRIANGLES, 0, 3); + + glNamedFramebufferTexture(framebuffer.handle, GL_COLOR_ATTACHMENT0, rcas_tex.handle, 0); + program_manager.BindPresentPrograms(vert.handle, rcas_frag.handle); + glBindTextureUnit(0, easu_tex.handle); + glDrawArrays(GL_TRIANGLES, 0, 3); -const OGLProgram& FSR::GetPresentFragmentProgram() const noexcept { - return fsr_rcas_frag; + return rcas_tex.handle; } -bool FSR::AreBuffersInitialized() const noexcept { - return fsr_framebuffer.handle; +bool FSR::NeedsRecreation(const Common::Rectangle& screen) { + return screen.GetWidth() != width || screen.GetHeight() != height; } } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/fsr.h b/src/video_core/renderer_opengl/present/fsr.h index fa57c6f00..606935a01 100644 --- a/src/video_core/renderer_opengl/present/fsr.h +++ b/src/video_core/renderer_opengl/present/fsr.h @@ -16,27 +16,24 @@ class ProgramManager; class FSR { public: - explicit FSR(); + explicit FSR(u32 output_width, u32 output_height); ~FSR(); - void Draw(ProgramManager& program_manager, const Common::Rectangle& screen, - u32 input_image_width, u32 input_image_height, - const Common::Rectangle& crop_rect); + GLuint Draw(ProgramManager& program_manager, GLuint texture, u32 input_image_width, + u32 input_image_height, const Common::Rectangle& crop_rect); - void InitBuffers(); - - void ReleaseBuffers(); - - [[nodiscard]] const OGLProgram& GetPresentFragmentProgram() const noexcept; - - [[nodiscard]] bool AreBuffersInitialized() const noexcept; + bool NeedsRecreation(const Common::Rectangle& screen); private: - OGLFramebuffer fsr_framebuffer; - OGLProgram fsr_vertex; - OGLProgram fsr_easu_frag; - OGLProgram fsr_rcas_frag; - OGLTexture fsr_intermediate_tex; + const u32 width; + const u32 height; + OGLFramebuffer framebuffer; + OGLSampler sampler; + OGLProgram vert; + OGLProgram easu_frag; + OGLProgram rcas_frag; + OGLTexture easu_tex; + OGLTexture rcas_tex; }; } // namespace OpenGL -- cgit v1.2.3 From d4de04584f14f3ea8fde4cd79102b887c084fbc2 Mon Sep 17 00:00:00 2001 From: Liam Date: Mon, 15 Jan 2024 15:08:21 -0500 Subject: renderer_opengl: split up blit screen resources into antialias and window adapt passes --- src/video_core/CMakeLists.txt | 4 + src/video_core/renderer_opengl/gl_blit_screen.cpp | 272 +++++---------------- src/video_core/renderer_opengl/gl_blit_screen.h | 26 +- src/video_core/renderer_opengl/present/filters.cpp | 39 +++ src/video_core/renderer_opengl/present/filters.h | 17 ++ src/video_core/renderer_opengl/present/fxaa.cpp | 1 + src/video_core/renderer_opengl/present/smaa.cpp | 6 - src/video_core/renderer_opengl/present/util.h | 11 + .../renderer_opengl/present/window_adapt_pass.cpp | 128 ++++++++++ .../renderer_opengl/present/window_adapt_pass.h | 39 +++ src/video_core/renderer_opengl/renderer_opengl.cpp | 6 + .../renderer_vulkan/present/window_adapt_pass.cpp | 6 +- src/video_core/renderer_vulkan/renderer_vulkan.cpp | 3 + 13 files changed, 329 insertions(+), 229 deletions(-) create mode 100644 src/video_core/renderer_opengl/present/filters.cpp create mode 100644 src/video_core/renderer_opengl/present/filters.h create mode 100644 src/video_core/renderer_opengl/present/window_adapt_pass.cpp create mode 100644 src/video_core/renderer_opengl/present/window_adapt_pass.h (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index 9879c3ad7..c6b0d628d 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -116,6 +116,8 @@ add_library(video_core STATIC renderer_null/null_rasterizer.h renderer_null/renderer_null.cpp renderer_null/renderer_null.h + renderer_opengl/present/filters.cpp + renderer_opengl/present/filters.h renderer_opengl/present/fsr.cpp renderer_opengl/present/fsr.h renderer_opengl/present/fxaa.cpp @@ -123,6 +125,8 @@ add_library(video_core STATIC renderer_opengl/present/smaa.cpp renderer_opengl/present/smaa.h renderer_opengl/present/util.h + renderer_opengl/present/window_adapt_pass.cpp + renderer_opengl/present/window_adapt_pass.h renderer_opengl/blit_image.cpp renderer_opengl/blit_image.h renderer_opengl/gl_blit_screen.cpp diff --git a/src/video_core/renderer_opengl/gl_blit_screen.cpp b/src/video_core/renderer_opengl/gl_blit_screen.cpp index 5f6221b9b..f9dbef0fc 100644 --- a/src/video_core/renderer_opengl/gl_blit_screen.cpp +++ b/src/video_core/renderer_opengl/gl_blit_screen.cpp @@ -2,100 +2,26 @@ // SPDX-License-Identifier: GPL-2.0-or-later #include "video_core/framebuffer_config.h" -#include "video_core/host_shaders/ffx_a_h.h" -#include "video_core/host_shaders/ffx_fsr1_h.h" -#include "video_core/host_shaders/full_screen_triangle_vert.h" -#include "video_core/host_shaders/opengl_fidelityfx_fsr_easu_frag.h" -#include "video_core/host_shaders/opengl_fidelityfx_fsr_frag.h" -#include "video_core/host_shaders/opengl_fidelityfx_fsr_rcas_frag.h" -#include "video_core/host_shaders/opengl_present_frag.h" -#include "video_core/host_shaders/opengl_present_scaleforce_frag.h" -#include "video_core/host_shaders/opengl_present_vert.h" -#include "video_core/host_shaders/present_bicubic_frag.h" -#include "video_core/host_shaders/present_gaussian_frag.h" - #include "video_core/renderer_opengl/gl_blit_screen.h" #include "video_core/renderer_opengl/gl_rasterizer.h" #include "video_core/renderer_opengl/gl_shader_manager.h" #include "video_core/renderer_opengl/gl_shader_util.h" #include "video_core/renderer_opengl/gl_state_tracker.h" +#include "video_core/renderer_opengl/present/filters.h" #include "video_core/renderer_opengl/present/fsr.h" #include "video_core/renderer_opengl/present/fxaa.h" #include "video_core/renderer_opengl/present/smaa.h" +#include "video_core/renderer_opengl/present/window_adapt_pass.h" #include "video_core/textures/decoders.h" namespace OpenGL { -namespace { -constexpr GLint PositionLocation = 0; -constexpr GLint TexCoordLocation = 1; -constexpr GLint ModelViewMatrixLocation = 0; - -struct ScreenRectVertex { - constexpr ScreenRectVertex(u32 x, u32 y, GLfloat u, GLfloat v) - : position{{static_cast(x), static_cast(y)}}, tex_coord{{u, v}} {} - - std::array position; - std::array tex_coord; -}; - -/** - * Defines a 1:1 pixel ortographic projection matrix with (0,0) on the top-left - * corner and (width, height) on the lower-bottom. - * - * The projection part of the matrix is trivial, hence these operations are represented - * by a 3x2 matrix. - */ -std::array MakeOrthographicMatrix(float width, float height) { - std::array matrix; // Laid out in column-major order - - // clang-format off - matrix[0] = 2.f / width; matrix[2] = 0.f; matrix[4] = -1.f; - matrix[1] = 0.f; matrix[3] = -2.f / height; matrix[5] = 1.f; - // Last matrix row is implicitly assumed to be [0, 0, 1]. - // clang-format on - - return matrix; -} -} // namespace - BlitScreen::BlitScreen(RasterizerOpenGL& rasterizer_, Tegra::MaxwellDeviceMemoryManager& device_memory_, StateTracker& state_tracker_, ProgramManager& program_manager_, Device& device_) : rasterizer(rasterizer_), device_memory(device_memory_), state_tracker(state_tracker_), program_manager(program_manager_), device(device_) { - // Create shader programs - present_vertex = CreateProgram(HostShaders::OPENGL_PRESENT_VERT, GL_VERTEX_SHADER); - present_bilinear_fragment = CreateProgram(HostShaders::OPENGL_PRESENT_FRAG, GL_FRAGMENT_SHADER); - present_bicubic_fragment = CreateProgram(HostShaders::PRESENT_BICUBIC_FRAG, GL_FRAGMENT_SHADER); - present_gaussian_fragment = - CreateProgram(HostShaders::PRESENT_GAUSSIAN_FRAG, GL_FRAGMENT_SHADER); - present_scaleforce_fragment = - CreateProgram(fmt::format("#version 460\n{}", HostShaders::OPENGL_PRESENT_SCALEFORCE_FRAG), - GL_FRAGMENT_SHADER); - - // Generate presentation sampler - present_sampler.Create(); - glSamplerParameteri(present_sampler.handle, GL_TEXTURE_MIN_FILTER, GL_LINEAR); - glSamplerParameteri(present_sampler.handle, GL_TEXTURE_MAG_FILTER, GL_LINEAR); - glSamplerParameteri(present_sampler.handle, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); - glSamplerParameteri(present_sampler.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); - glSamplerParameteri(present_sampler.handle, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE); - - present_sampler_nn.Create(); - glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_MIN_FILTER, GL_NEAREST); - glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_MAG_FILTER, GL_NEAREST); - glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); - glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); - glSamplerParameteri(present_sampler_nn.handle, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE); - - // Generate VBO handle for drawing - vertex_buffer.Create(); - - // Attach vertex data to VAO - glNamedBufferData(vertex_buffer.handle, sizeof(ScreenRectVertex) * 4, nullptr, GL_STREAM_DRAW); - // Allocate textures for the screen framebuffer_texture.resource.Create(GL_TEXTURE_2D); @@ -106,15 +32,6 @@ BlitScreen::BlitScreen(RasterizerOpenGL& rasterizer_, const u8 framebuffer_data[4] = {0, 0, 0, 0}; glClearTexImage(framebuffer_texture.resource.handle, 0, GL_RGBA, GL_UNSIGNED_BYTE, framebuffer_data); - - // Enable unified vertex attributes and query vertex buffer address when the driver supports it - if (device.HasVertexBufferUnifiedMemory()) { - glEnableClientState(GL_VERTEX_ATTRIB_ARRAY_UNIFIED_NV); - glEnableClientState(GL_ELEMENT_ARRAY_UNIFIED_NV); - glMakeNamedBufferResidentNV(vertex_buffer.handle, GL_READ_ONLY); - glGetNamedBufferParameterui64vNV(vertex_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, - &vertex_buffer_address); - } } BlitScreen::~BlitScreen() = default; @@ -219,18 +136,14 @@ void BlitScreen::ConfigureFramebufferTexture(const Tegra::FramebufferConfig& fra glTextureStorage2D(framebuffer_texture.resource.handle, 1, internal_format, framebuffer_texture.width, framebuffer_texture.height); - fxaa = std::make_unique( - Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), - Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); - smaa = std::make_unique( - Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), - Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); + fxaa.reset(); + smaa.reset(); } void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, const Layout::FramebufferLayout& layout) { FramebufferTextureInfo info = PrepareRenderTarget(framebuffer); - const auto crop = Tegra::NormalizeCrop(framebuffer, info.width, info.height); + auto crop = Tegra::NormalizeCrop(framebuffer, info.width, info.height); // TODO: Signal state tracker about these changes state_tracker.NotifyScreenDrawVertexArray(); @@ -267,15 +180,14 @@ void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, glColorMaski(0, GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); glDepthRangeIndexed(0, 0.0, 0.0); + GLint old_read_fb; + GLint old_draw_fb; + glGetIntegerv(GL_READ_FRAMEBUFFER_BINDING, &old_read_fb); + glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb); + GLuint texture = info.display_texture; auto anti_aliasing = Settings::values.anti_aliasing.GetValue(); - if (anti_aliasing >= Settings::AntiAliasing::MaxEnum) { - LOG_ERROR(Render_OpenGL, "Invalid antialiasing option selected {}", anti_aliasing); - anti_aliasing = Settings::AntiAliasing::None; - Settings::values.anti_aliasing.SetValue(anti_aliasing); - } - if (anti_aliasing != Settings::AntiAliasing::None) { glEnablei(GL_SCISSOR_TEST, 0); auto scissor_width = Settings::values.resolution_info.ScaleUp(framebuffer_texture.width); @@ -286,137 +198,83 @@ void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, glScissorIndexed(0, 0, 0, scissor_width, scissor_height); glViewportIndexedf(0, 0.0f, 0.0f, viewport_width, viewport_height); - glBindSampler(0, present_sampler.handle); - GLint old_read_fb; - GLint old_draw_fb; - glGetIntegerv(GL_READ_FRAMEBUFFER_BINDING, &old_read_fb); - glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb); - switch (anti_aliasing) { - case Settings::AntiAliasing::Fxaa: { + case Settings::AntiAliasing::Fxaa: + CreateFXAA(); texture = fxaa->Draw(program_manager, info.display_texture); - } break; - case Settings::AntiAliasing::Smaa: { - texture = smaa->Draw(program_manager, info.display_texture); - } break; + break; + case Settings::AntiAliasing::Smaa: default: - UNREACHABLE(); + CreateSMAA(); + texture = smaa->Draw(program_manager, info.display_texture); + break; } - - glBindFramebuffer(GL_READ_FRAMEBUFFER, old_read_fb); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); } + glDisablei(GL_SCISSOR_TEST, 0); if (Settings::values.scaling_filter.GetValue() == Settings::ScalingFilter::Fsr) { - GLint old_read_fb; - GLint old_draw_fb; - glGetIntegerv(GL_READ_FRAMEBUFFER_BINDING, &old_read_fb); - glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb); - if (!fsr || fsr->NeedsRecreation(layout.screen)) { fsr = std::make_unique(layout.screen.GetWidth(), layout.screen.GetHeight()); } texture = fsr->Draw(program_manager, texture, info.scaled_width, info.scaled_height, crop); - - glBindFramebuffer(GL_READ_FRAMEBUFFER, old_read_fb); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); + crop = {0, 0, 1, 1}; } - glBindTextureUnit(0, texture); - - const std::array ortho_matrix = - MakeOrthographicMatrix(static_cast(layout.width), static_cast(layout.height)); - - const auto fragment_handle = [this]() { - switch (Settings::values.scaling_filter.GetValue()) { - case Settings::ScalingFilter::Bicubic: - return present_bicubic_fragment.handle; - case Settings::ScalingFilter::Gaussian: - return present_gaussian_fragment.handle; - case Settings::ScalingFilter::ScaleForce: - return present_scaleforce_fragment.handle; - case Settings::ScalingFilter::NearestNeighbor: - case Settings::ScalingFilter::Bilinear: - case Settings::ScalingFilter::Fsr: - default: - return present_bilinear_fragment.handle; - } - }(); - program_manager.BindPresentPrograms(present_vertex.handle, fragment_handle); - glProgramUniformMatrix3x2fv(present_vertex.handle, ModelViewMatrixLocation, 1, GL_FALSE, - ortho_matrix.data()); + glBindFramebuffer(GL_READ_FRAMEBUFFER, old_read_fb); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); - f32 left, top, right, bottom; - if (Settings::values.scaling_filter.GetValue() == Settings::ScalingFilter::Fsr) { - // FSR has already applied the crop, so we just want to render the image - // it has produced. - left = 0; - top = 0; - right = 1; - bottom = 1; - } else { - // Apply the precomputed crop. - left = crop.left; - top = crop.top; - right = crop.right; - bottom = crop.bottom; - } + CreateWindowAdapt(); + window_adapt->DrawToFramebuffer(program_manager, texture, layout, crop); - // Map the coordinates to the screen. - const auto& screen = layout.screen; - const auto x = screen.left; - const auto y = screen.top; - const auto w = screen.GetWidth(); - const auto h = screen.GetHeight(); - - const std::array vertices = { - ScreenRectVertex(x, y, left, top), - ScreenRectVertex(x + w, y, right, top), - ScreenRectVertex(x, y + h, left, bottom), - ScreenRectVertex(x + w, y + h, right, bottom), - }; - glNamedBufferSubData(vertex_buffer.handle, 0, sizeof(vertices), std::data(vertices)); - - glDisable(GL_FRAMEBUFFER_SRGB); - glViewportIndexedf(0, 0.0f, 0.0f, static_cast(layout.width), - static_cast(layout.height)); - - glEnableVertexAttribArray(PositionLocation); - glEnableVertexAttribArray(TexCoordLocation); - glVertexAttribDivisor(PositionLocation, 0); - glVertexAttribDivisor(TexCoordLocation, 0); - glVertexAttribFormat(PositionLocation, 2, GL_FLOAT, GL_FALSE, - offsetof(ScreenRectVertex, position)); - glVertexAttribFormat(TexCoordLocation, 2, GL_FLOAT, GL_FALSE, - offsetof(ScreenRectVertex, tex_coord)); - glVertexAttribBinding(PositionLocation, 0); - glVertexAttribBinding(TexCoordLocation, 0); - 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)); - } + // TODO + // program_manager.RestoreGuestPipeline(); +} - if (Settings::values.scaling_filter.GetValue() != Settings::ScalingFilter::NearestNeighbor) { - glBindSampler(0, present_sampler.handle); - } else { - glBindSampler(0, present_sampler_nn.handle); +void BlitScreen::CreateFXAA() { + smaa.reset(); + if (!fxaa) { + fxaa = std::make_unique( + Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), + Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); } +} - // Update background color before drawing - glClearColor(Settings::values.bg_red.GetValue() / 255.0f, - Settings::values.bg_green.GetValue() / 255.0f, - Settings::values.bg_blue.GetValue() / 255.0f, 1.0f); +void BlitScreen::CreateSMAA() { + fxaa.reset(); + if (!smaa) { + smaa = std::make_unique( + Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), + Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); + } +} - glClear(GL_COLOR_BUFFER_BIT); - glDrawArrays(GL_TRIANGLE_STRIP, 0, 4); +void BlitScreen::CreateWindowAdapt() { + if (window_adapt && Settings::values.scaling_filter.GetValue() == current_window_adapt) { + return; + } - // TODO - // program_manager.RestoreGuestPipeline(); + current_window_adapt = Settings::values.scaling_filter.GetValue(); + switch (current_window_adapt) { + case Settings::ScalingFilter::NearestNeighbor: + window_adapt = MakeNearestNeighbor(device); + break; + case Settings::ScalingFilter::Bicubic: + window_adapt = MakeBicubic(device); + break; + case Settings::ScalingFilter::Gaussian: + window_adapt = MakeGaussian(device); + break; + case Settings::ScalingFilter::ScaleForce: + window_adapt = MakeScaleForce(device); + break; + case Settings::ScalingFilter::Fsr: + case Settings::ScalingFilter::Bilinear: + default: + window_adapt = MakeBilinear(device); + break; + } } } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_blit_screen.h b/src/video_core/renderer_opengl/gl_blit_screen.h index 2cb9a5015..f42f89dee 100644 --- a/src/video_core/renderer_opengl/gl_blit_screen.h +++ b/src/video_core/renderer_opengl/gl_blit_screen.h @@ -18,6 +18,10 @@ namespace Tegra { struct FramebufferConfig; } +namespace Settings { +enum class ScalingFilter : u32; +} + namespace OpenGL { class Device; @@ -27,6 +31,7 @@ class ProgramManager; class RasterizerOpenGL; class SMAA; class StateTracker; +class WindowAdaptPass; /// Structure used for storing information about the textures for the Switch screen struct TextureInfo { @@ -61,29 +66,22 @@ public: void DrawScreen(const Tegra::FramebufferConfig& framebuffer, const Layout::FramebufferLayout& layout); - void RenderScreenshot(const Tegra::FramebufferConfig& framebuffer); - /// Loads framebuffer from emulated memory into the active OpenGL texture. FramebufferTextureInfo LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer); FramebufferTextureInfo PrepareRenderTarget(const Tegra::FramebufferConfig& framebuffer); private: + void CreateFXAA(); + void CreateSMAA(); + void CreateWindowAdapt(); + RasterizerOpenGL& rasterizer; Tegra::MaxwellDeviceMemoryManager& device_memory; StateTracker& state_tracker; ProgramManager& program_manager; Device& device; - OGLSampler present_sampler; - OGLSampler present_sampler_nn; - OGLBuffer vertex_buffer; - OGLProgram present_vertex; - OGLProgram present_bilinear_fragment; - OGLProgram present_bicubic_fragment; - OGLProgram present_gaussian_fragment; - OGLProgram present_scaleforce_fragment; - /// Display information for Switch screen TextureInfo framebuffer_texture; @@ -91,11 +89,11 @@ private: std::unique_ptr fxaa; std::unique_ptr smaa; + Settings::ScalingFilter current_window_adapt{}; + std::unique_ptr window_adapt; + /// OpenGL framebuffer data std::vector gl_framebuffer_data; - - // GPU address of the vertex buffer - GLuint64EXT vertex_buffer_address = 0; }; } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/filters.cpp b/src/video_core/renderer_opengl/present/filters.cpp new file mode 100644 index 000000000..819e5d77f --- /dev/null +++ b/src/video_core/renderer_opengl/present/filters.cpp @@ -0,0 +1,39 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#include "video_core/host_shaders/opengl_present_frag.h" +#include "video_core/host_shaders/opengl_present_scaleforce_frag.h" +#include "video_core/host_shaders/present_bicubic_frag.h" +#include "video_core/host_shaders/present_gaussian_frag.h" +#include "video_core/renderer_opengl/present/filters.h" +#include "video_core/renderer_opengl/present/util.h" + +namespace OpenGL { + +std::unique_ptr MakeNearestNeighbor(const Device& device) { + return std::make_unique(device, CreateNearestNeighborSampler(), + HostShaders::OPENGL_PRESENT_FRAG); +} + +std::unique_ptr MakeBilinear(const Device& device) { + return std::make_unique(device, CreateBilinearSampler(), + HostShaders::OPENGL_PRESENT_FRAG); +} + +std::unique_ptr MakeBicubic(const Device& device) { + return std::make_unique(device, CreateBilinearSampler(), + HostShaders::PRESENT_BICUBIC_FRAG); +} + +std::unique_ptr MakeGaussian(const Device& device) { + return std::make_unique(device, CreateBilinearSampler(), + HostShaders::PRESENT_GAUSSIAN_FRAG); +} + +std::unique_ptr MakeScaleForce(const Device& device) { + return std::make_unique( + device, CreateBilinearSampler(), + fmt::format("#version 460\n{}", HostShaders::OPENGL_PRESENT_SCALEFORCE_FRAG)); +} + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/filters.h b/src/video_core/renderer_opengl/present/filters.h new file mode 100644 index 000000000..122ab7436 --- /dev/null +++ b/src/video_core/renderer_opengl/present/filters.h @@ -0,0 +1,17 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include +#include "video_core/renderer_opengl/present/window_adapt_pass.h" + +namespace OpenGL { + +std::unique_ptr MakeNearestNeighbor(const Device& device); +std::unique_ptr MakeBilinear(const Device& device); +std::unique_ptr MakeBicubic(const Device& device); +std::unique_ptr MakeGaussian(const Device& device); +std::unique_ptr MakeScaleForce(const Device& device); + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/fxaa.cpp b/src/video_core/renderer_opengl/present/fxaa.cpp index 9425c42fa..d9b58512d 100644 --- a/src/video_core/renderer_opengl/present/fxaa.cpp +++ b/src/video_core/renderer_opengl/present/fxaa.cpp @@ -31,6 +31,7 @@ GLuint FXAA::Draw(ProgramManager& program_manager, GLuint input_texture) { program_manager.BindPresentPrograms(vert_shader.handle, frag_shader.handle); glBindFramebuffer(GL_DRAW_FRAMEBUFFER, framebuffer.handle); glBindTextureUnit(0, input_texture); + glBindSampler(0, sampler.handle); glDrawArrays(GL_TRIANGLES, 0, 3); glFrontFace(GL_CW); diff --git a/src/video_core/renderer_opengl/present/smaa.cpp b/src/video_core/renderer_opengl/present/smaa.cpp index a9a0eb6c6..de7f6e502 100644 --- a/src/video_core/renderer_opengl/present/smaa.cpp +++ b/src/video_core/renderer_opengl/present/smaa.cpp @@ -36,13 +36,7 @@ SMAA::SMAA(u32 width, u32 height) { SmaaShader(HostShaders::SMAA_NEIGHBORHOOD_BLENDING_FRAG, GL_FRAGMENT_SHADER); glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0); - glPixelStorei(GL_UNPACK_SWAP_BYTES, GL_FALSE); - glPixelStorei(GL_UNPACK_LSB_FIRST, GL_FALSE); glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); - glPixelStorei(GL_UNPACK_IMAGE_HEIGHT, 0); - glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0); - glPixelStorei(GL_UNPACK_SKIP_ROWS, 0); - glPixelStorei(GL_UNPACK_ALIGNMENT, 4); area_tex.Create(GL_TEXTURE_2D); glTextureStorage2D(area_tex.handle, 1, GL_RG8, AREATEX_WIDTH, AREATEX_HEIGHT); diff --git a/src/video_core/renderer_opengl/present/util.h b/src/video_core/renderer_opengl/present/util.h index 0aa8b110c..67f03aa27 100644 --- a/src/video_core/renderer_opengl/present/util.h +++ b/src/video_core/renderer_opengl/present/util.h @@ -29,4 +29,15 @@ static inline OGLSampler CreateBilinearSampler() { return sampler; } +static inline OGLSampler CreateNearestNeighborSampler() { + OGLSampler sampler; + sampler.Create(); + glSamplerParameteri(sampler.handle, GL_TEXTURE_MIN_FILTER, GL_NEAREST); + glSamplerParameteri(sampler.handle, GL_TEXTURE_MAG_FILTER, GL_NEAREST); + glSamplerParameteri(sampler.handle, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glSamplerParameteri(sampler.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + glSamplerParameteri(sampler.handle, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE); + return sampler; +} + } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/window_adapt_pass.cpp b/src/video_core/renderer_opengl/present/window_adapt_pass.cpp new file mode 100644 index 000000000..168fa1aea --- /dev/null +++ b/src/video_core/renderer_opengl/present/window_adapt_pass.cpp @@ -0,0 +1,128 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#include "common/settings.h" +#include "video_core/host_shaders/opengl_present_vert.h" +#include "video_core/renderer_opengl/gl_device.h" +#include "video_core/renderer_opengl/gl_shader_manager.h" +#include "video_core/renderer_opengl/gl_shader_util.h" +#include "video_core/renderer_opengl/present/window_adapt_pass.h" + +namespace OpenGL { + +namespace { +constexpr GLint PositionLocation = 0; +constexpr GLint TexCoordLocation = 1; +constexpr GLint ModelViewMatrixLocation = 0; + +struct ScreenRectVertex { + constexpr ScreenRectVertex(u32 x, u32 y, GLfloat u, GLfloat v) + : position{{static_cast(x), static_cast(y)}}, tex_coord{{u, v}} {} + + std::array position; + std::array tex_coord; +}; + +/** + * Defines a 1:1 pixel orthographic projection matrix with (0,0) on the top-left + * corner and (width, height) on the lower-bottom. + * + * The projection part of the matrix is trivial, hence these operations are represented + * by a 3x2 matrix. + */ +std::array MakeOrthographicMatrix(float width, float height) { + std::array matrix; // Laid out in column-major order + + // clang-format off + matrix[0] = 2.f / width; matrix[2] = 0.f; matrix[4] = -1.f; + matrix[1] = 0.f; matrix[3] = -2.f / height; matrix[5] = 1.f; + // Last matrix row is implicitly assumed to be [0, 0, 1]. + // clang-format on + + return matrix; +} +} // namespace + +WindowAdaptPass::WindowAdaptPass(const Device& device_, OGLSampler&& sampler_, + std::string_view frag_source) + : device(device_), sampler(std::move(sampler_)) { + vert = CreateProgram(HostShaders::OPENGL_PRESENT_VERT, GL_VERTEX_SHADER); + frag = CreateProgram(frag_source, GL_FRAGMENT_SHADER); + + // Generate VBO handle for drawing + vertex_buffer.Create(); + + // Attach vertex data to VAO + glNamedBufferData(vertex_buffer.handle, sizeof(ScreenRectVertex) * 4, nullptr, GL_STREAM_DRAW); + + // Query vertex buffer address when the driver supports unified vertex attributes + if (device.HasVertexBufferUnifiedMemory()) { + glMakeNamedBufferResidentNV(vertex_buffer.handle, GL_READ_ONLY); + glGetNamedBufferParameterui64vNV(vertex_buffer.handle, GL_BUFFER_GPU_ADDRESS_NV, + &vertex_buffer_address); + } +} + +WindowAdaptPass::~WindowAdaptPass() = default; + +void WindowAdaptPass::DrawToFramebuffer(ProgramManager& program_manager, GLuint texture, + const Layout::FramebufferLayout& layout, + const Common::Rectangle& crop) { + glBindTextureUnit(0, texture); + + const std::array ortho_matrix = + MakeOrthographicMatrix(static_cast(layout.width), static_cast(layout.height)); + + program_manager.BindPresentPrograms(vert.handle, frag.handle); + glProgramUniformMatrix3x2fv(vert.handle, ModelViewMatrixLocation, 1, GL_FALSE, + ortho_matrix.data()); + + // Map the coordinates to the screen. + const auto& screen = layout.screen; + const auto x = screen.left; + const auto y = screen.top; + const auto w = screen.GetWidth(); + const auto h = screen.GetHeight(); + + const std::array vertices = { + ScreenRectVertex(x, y, crop.left, crop.top), + ScreenRectVertex(x + w, y, crop.right, crop.top), + ScreenRectVertex(x, y + h, crop.left, crop.bottom), + ScreenRectVertex(x + w, y + h, crop.right, crop.bottom), + }; + glNamedBufferSubData(vertex_buffer.handle, 0, sizeof(vertices), std::data(vertices)); + + glDisable(GL_FRAMEBUFFER_SRGB); + glViewportIndexedf(0, 0.0f, 0.0f, static_cast(layout.width), + static_cast(layout.height)); + + glEnableVertexAttribArray(PositionLocation); + glEnableVertexAttribArray(TexCoordLocation); + glVertexAttribDivisor(PositionLocation, 0); + glVertexAttribDivisor(TexCoordLocation, 0); + glVertexAttribFormat(PositionLocation, 2, GL_FLOAT, GL_FALSE, + offsetof(ScreenRectVertex, position)); + glVertexAttribFormat(TexCoordLocation, 2, GL_FLOAT, GL_FALSE, + offsetof(ScreenRectVertex, tex_coord)); + glVertexAttribBinding(PositionLocation, 0); + glVertexAttribBinding(TexCoordLocation, 0); + 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)); + } + + glBindSampler(0, sampler.handle); + + // Update background color before drawing + glClearColor(Settings::values.bg_red.GetValue() / 255.0f, + Settings::values.bg_green.GetValue() / 255.0f, + Settings::values.bg_blue.GetValue() / 255.0f, 1.0f); + + glClear(GL_COLOR_BUFFER_BIT); + glDrawArrays(GL_TRIANGLE_STRIP, 0, 4); +} + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/window_adapt_pass.h b/src/video_core/renderer_opengl/present/window_adapt_pass.h new file mode 100644 index 000000000..65dcd09ff --- /dev/null +++ b/src/video_core/renderer_opengl/present/window_adapt_pass.h @@ -0,0 +1,39 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include "common/math_util.h" +#include "video_core/renderer_opengl/gl_resource_manager.h" + +namespace Layout { +struct FramebufferLayout; +} + +namespace OpenGL { + +class Device; +class ProgramManager; + +class WindowAdaptPass final { +public: + explicit WindowAdaptPass(const Device& device, OGLSampler&& sampler, + std::string_view frag_source); + ~WindowAdaptPass(); + + void DrawToFramebuffer(ProgramManager& program_manager, GLuint texture, + const Layout::FramebufferLayout& layout, + const Common::Rectangle& crop); + +private: + const Device& device; + OGLSampler sampler; + OGLProgram vert; + OGLProgram frag; + OGLBuffer vertex_buffer; + + // GPU address of the vertex buffer + GLuint64EXT vertex_buffer_address = 0; +}; + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index 3d75fd17a..0d138c189 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -113,6 +113,12 @@ RendererOpenGL::RendererOpenGL(Core::TelemetrySession& telemetry_session_, if (!GLAD_GL_ARB_seamless_cubemap_per_texture && !GLAD_GL_AMD_seamless_cubemap_per_texture) { glEnable(GL_TEXTURE_CUBE_MAP_SEAMLESS); } + + // Enable unified vertex attributes when the driver supports it + if (device.HasVertexBufferUnifiedMemory()) { + glEnableClientState(GL_VERTEX_ATTRIB_ARRAY_UNIFIED_NV); + glEnableClientState(GL_ELEMENT_ARRAY_UNIFIED_NV); + } blit_screen = std::make_unique(rasterizer, device_memory, state_tracker, program_manager, device); } diff --git a/src/video_core/renderer_vulkan/present/window_adapt_pass.cpp b/src/video_core/renderer_vulkan/present/window_adapt_pass.cpp index 7fd9ecd22..1d1828a4c 100644 --- a/src/video_core/renderer_vulkan/present/window_adapt_pass.cpp +++ b/src/video_core/renderer_vulkan/present/window_adapt_pass.cpp @@ -92,7 +92,9 @@ void WindowAdaptPass::Draw(Scheduler& scheduler, size_t image_index, VkImageView const VkFramebuffer host_framebuffer{*dst->framebuffer}; const VkRenderPass renderpass{*render_pass}; const VkPipeline graphics_pipeline{*pipeline}; + const VkPipelineLayout graphics_pipeline_layout{*pipeline_layout}; const VkDescriptorSet descriptor_set{descriptor_sets[image_index]}; + const VkBuffer vertex_buffer{*buffer}; const VkExtent2D render_area{ .width = dst->width, .height = dst->height, @@ -134,8 +136,8 @@ void WindowAdaptPass::Draw(Scheduler& scheduler, size_t image_index, VkImageView cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, graphics_pipeline); cmdbuf.SetViewport(0, viewport); cmdbuf.SetScissor(0, scissor); - cmdbuf.BindVertexBuffer(0, *buffer, offsetof(BufferData, vertices)); - cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, *pipeline_layout, 0, + cmdbuf.BindVertexBuffer(0, vertex_buffer, offsetof(BufferData, vertices)); + cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, graphics_pipeline_layout, 0, descriptor_set, {}); cmdbuf.Draw(4, 1, 0, 0); cmdbuf.EndRenderPass(); diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp index 2912aaff6..a99ef08a5 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp +++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp @@ -222,6 +222,9 @@ void Vulkan::RendererVulkan::RenderScreenshot(const Tegra::FramebufferConfig& fr .image = std::move(staging_image), .image_view = std::move(dst_view), .framebuffer = std::move(screenshot_fb), + .cmdbuf{}, + .render_ready{}, + .present_done{}, }; }(); -- cgit v1.2.3 From 10cf0585180bcf2eab38ebf65dc593fecc4ddf92 Mon Sep 17 00:00:00 2001 From: Liam Date: Thu, 18 Jan 2024 20:47:50 -0500 Subject: renderer_opengl: implement layer stack composition --- src/video_core/CMakeLists.txt | 3 + src/video_core/renderer_opengl/gl_blit_screen.cpp | 198 +------------------ src/video_core/renderer_opengl/gl_blit_screen.h | 38 +--- src/video_core/renderer_opengl/present/layer.cpp | 215 +++++++++++++++++++++ src/video_core/renderer_opengl/present/layer.h | 80 ++++++++ .../renderer_opengl/present/present_uniforms.h | 43 +++++ .../renderer_opengl/present/window_adapt_pass.cpp | 91 ++++----- .../renderer_opengl/present/window_adapt_pass.h | 14 +- src/video_core/renderer_opengl/renderer_opengl.cpp | 8 +- src/video_core/renderer_opengl/renderer_opengl.h | 2 +- 10 files changed, 402 insertions(+), 290 deletions(-) create mode 100644 src/video_core/renderer_opengl/present/layer.cpp create mode 100644 src/video_core/renderer_opengl/present/layer.h create mode 100644 src/video_core/renderer_opengl/present/present_uniforms.h (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index 7526de699..16c905db9 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -122,6 +122,9 @@ add_library(video_core STATIC renderer_opengl/present/fsr.h renderer_opengl/present/fxaa.cpp renderer_opengl/present/fxaa.h + renderer_opengl/present/layer.cpp + renderer_opengl/present/layer.h + renderer_opengl/present/present_uniforms.h renderer_opengl/present/smaa.cpp renderer_opengl/present/smaa.h renderer_opengl/present/util.h diff --git a/src/video_core/renderer_opengl/gl_blit_screen.cpp b/src/video_core/renderer_opengl/gl_blit_screen.cpp index f9dbef0fc..6ba8b214b 100644 --- a/src/video_core/renderer_opengl/gl_blit_screen.cpp +++ b/src/video_core/renderer_opengl/gl_blit_screen.cpp @@ -1,18 +1,12 @@ // SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project // SPDX-License-Identifier: GPL-2.0-or-later -#include "video_core/framebuffer_config.h" +#include "common/settings.h" #include "video_core/renderer_opengl/gl_blit_screen.h" -#include "video_core/renderer_opengl/gl_rasterizer.h" -#include "video_core/renderer_opengl/gl_shader_manager.h" -#include "video_core/renderer_opengl/gl_shader_util.h" #include "video_core/renderer_opengl/gl_state_tracker.h" #include "video_core/renderer_opengl/present/filters.h" -#include "video_core/renderer_opengl/present/fsr.h" -#include "video_core/renderer_opengl/present/fxaa.h" -#include "video_core/renderer_opengl/present/smaa.h" +#include "video_core/renderer_opengl/present/layer.h" #include "video_core/renderer_opengl/present/window_adapt_pass.h" -#include "video_core/textures/decoders.h" namespace OpenGL { @@ -21,130 +15,12 @@ BlitScreen::BlitScreen(RasterizerOpenGL& rasterizer_, StateTracker& state_tracker_, ProgramManager& program_manager_, Device& device_) : rasterizer(rasterizer_), device_memory(device_memory_), state_tracker(state_tracker_), - program_manager(program_manager_), device(device_) { - // Allocate textures for the screen - framebuffer_texture.resource.Create(GL_TEXTURE_2D); - - const GLuint texture = framebuffer_texture.resource.handle; - glTextureStorage2D(texture, 1, GL_RGBA8, 1, 1); - - // Clear screen to black - const u8 framebuffer_data[4] = {0, 0, 0, 0}; - glClearTexImage(framebuffer_texture.resource.handle, 0, GL_RGBA, GL_UNSIGNED_BYTE, - framebuffer_data); -} + program_manager(program_manager_), device(device_) {} BlitScreen::~BlitScreen() = default; -FramebufferTextureInfo BlitScreen::PrepareRenderTarget( - const Tegra::FramebufferConfig& framebuffer) { - // If framebuffer is provided, reload it from memory to a texture - if (framebuffer_texture.width != static_cast(framebuffer.width) || - framebuffer_texture.height != static_cast(framebuffer.height) || - framebuffer_texture.pixel_format != framebuffer.pixel_format || - gl_framebuffer_data.empty()) { - // Reallocate texture if the framebuffer size has changed. - // This is expected to not happen very often and hence should not be a - // performance problem. - ConfigureFramebufferTexture(framebuffer); - } - - // Load the framebuffer from memory if needed - return LoadFBToScreenInfo(framebuffer); -} - -FramebufferTextureInfo BlitScreen::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer) { - const DAddr framebuffer_addr{framebuffer.address + framebuffer.offset}; - const auto accelerated_info = - rasterizer.AccelerateDisplay(framebuffer, framebuffer_addr, framebuffer.stride); - if (accelerated_info) { - return *accelerated_info; - } - - // Reset the screen info's display texture to its own permanent texture - FramebufferTextureInfo info{}; - info.display_texture = framebuffer_texture.resource.handle; - info.width = framebuffer.width; - info.height = framebuffer.height; - info.scaled_width = framebuffer.width; - info.scaled_height = framebuffer.height; - - // TODO(Rodrigo): Read this from HLE - constexpr u32 block_height_log2 = 4; - const auto pixel_format{ - VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format)}; - const u32 bytes_per_pixel{VideoCore::Surface::BytesPerBlock(pixel_format)}; - const u64 size_in_bytes{Tegra::Texture::CalculateSize( - true, bytes_per_pixel, framebuffer.stride, framebuffer.height, 1, block_height_log2, 0)}; - const u8* const host_ptr{device_memory.GetPointer(framebuffer_addr)}; - const std::span input_data(host_ptr, size_in_bytes); - Tegra::Texture::UnswizzleTexture(gl_framebuffer_data, input_data, bytes_per_pixel, - framebuffer.width, framebuffer.height, 1, block_height_log2, - 0); - - glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0); - glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast(framebuffer.stride)); - - // Update existing texture - // TODO: Test what happens on hardware when you change the framebuffer dimensions so that - // they differ from the LCD resolution. - // TODO: Applications could theoretically crash yuzu here by specifying too large - // framebuffer sizes. We should make sure that this cannot happen. - glTextureSubImage2D(framebuffer_texture.resource.handle, 0, 0, 0, framebuffer.width, - framebuffer.height, framebuffer_texture.gl_format, - framebuffer_texture.gl_type, gl_framebuffer_data.data()); - - glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); - - return info; -} - -void BlitScreen::ConfigureFramebufferTexture(const Tegra::FramebufferConfig& framebuffer) { - framebuffer_texture.width = framebuffer.width; - framebuffer_texture.height = framebuffer.height; - framebuffer_texture.pixel_format = framebuffer.pixel_format; - - const auto pixel_format{ - VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format)}; - const u32 bytes_per_pixel{VideoCore::Surface::BytesPerBlock(pixel_format)}; - gl_framebuffer_data.resize(framebuffer_texture.width * framebuffer_texture.height * - bytes_per_pixel); - - GLint internal_format; - switch (framebuffer.pixel_format) { - case Service::android::PixelFormat::Rgba8888: - internal_format = GL_RGBA8; - framebuffer_texture.gl_format = GL_RGBA; - framebuffer_texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; - break; - case Service::android::PixelFormat::Rgb565: - internal_format = GL_RGB565; - framebuffer_texture.gl_format = GL_RGB; - framebuffer_texture.gl_type = GL_UNSIGNED_SHORT_5_6_5; - break; - default: - internal_format = GL_RGBA8; - framebuffer_texture.gl_format = GL_RGBA; - framebuffer_texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; - // UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}", - // static_cast(framebuffer.pixel_format)); - break; - } - - framebuffer_texture.resource.Release(); - framebuffer_texture.resource.Create(GL_TEXTURE_2D); - glTextureStorage2D(framebuffer_texture.resource.handle, 1, internal_format, - framebuffer_texture.width, framebuffer_texture.height); - - fxaa.reset(); - smaa.reset(); -} - -void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, +void BlitScreen::DrawScreen(std::span framebuffers, const Layout::FramebufferLayout& layout) { - FramebufferTextureInfo info = PrepareRenderTarget(framebuffer); - auto crop = Tegra::NormalizeCrop(framebuffer, info.width, info.height); - // TODO: Signal state tracker about these changes state_tracker.NotifyScreenDrawVertexArray(); state_tracker.NotifyPolygonModes(); @@ -163,7 +39,6 @@ void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, state_tracker.NotifyLogicOp(); state_tracker.NotifyClipControl(); state_tracker.NotifyAlphaTest(); - state_tracker.ClipControl(GL_LOWER_LEFT, GL_ZERO_TO_ONE); glEnable(GL_CULL_FACE); @@ -180,76 +55,17 @@ void BlitScreen::DrawScreen(const Tegra::FramebufferConfig& framebuffer, glColorMaski(0, GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); glDepthRangeIndexed(0, 0.0, 0.0); - GLint old_read_fb; - GLint old_draw_fb; - glGetIntegerv(GL_READ_FRAMEBUFFER_BINDING, &old_read_fb); - glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb); - - GLuint texture = info.display_texture; - - auto anti_aliasing = Settings::values.anti_aliasing.GetValue(); - if (anti_aliasing != Settings::AntiAliasing::None) { - glEnablei(GL_SCISSOR_TEST, 0); - auto scissor_width = Settings::values.resolution_info.ScaleUp(framebuffer_texture.width); - auto viewport_width = static_cast(scissor_width); - auto scissor_height = Settings::values.resolution_info.ScaleUp(framebuffer_texture.height); - auto viewport_height = static_cast(scissor_height); - - glScissorIndexed(0, 0, 0, scissor_width, scissor_height); - glViewportIndexedf(0, 0.0f, 0.0f, viewport_width, viewport_height); - - switch (anti_aliasing) { - case Settings::AntiAliasing::Fxaa: - CreateFXAA(); - texture = fxaa->Draw(program_manager, info.display_texture); - break; - case Settings::AntiAliasing::Smaa: - default: - CreateSMAA(); - texture = smaa->Draw(program_manager, info.display_texture); - break; - } + while (layers.size() < framebuffers.size()) { + layers.emplace_back(rasterizer, device_memory); } - glDisablei(GL_SCISSOR_TEST, 0); - - if (Settings::values.scaling_filter.GetValue() == Settings::ScalingFilter::Fsr) { - if (!fsr || fsr->NeedsRecreation(layout.screen)) { - fsr = std::make_unique(layout.screen.GetWidth(), layout.screen.GetHeight()); - } - - texture = fsr->Draw(program_manager, texture, info.scaled_width, info.scaled_height, crop); - crop = {0, 0, 1, 1}; - } - - glBindFramebuffer(GL_READ_FRAMEBUFFER, old_read_fb); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); - CreateWindowAdapt(); - window_adapt->DrawToFramebuffer(program_manager, texture, layout, crop); + window_adapt->DrawToFramebuffer(program_manager, layers, framebuffers, layout); // TODO // program_manager.RestoreGuestPipeline(); } -void BlitScreen::CreateFXAA() { - smaa.reset(); - if (!fxaa) { - fxaa = std::make_unique( - Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), - Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); - } -} - -void BlitScreen::CreateSMAA() { - fxaa.reset(); - if (!smaa) { - smaa = std::make_unique( - Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), - Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); - } -} - void BlitScreen::CreateWindowAdapt() { if (window_adapt && Settings::values.scaling_filter.GetValue() == current_window_adapt) { return; diff --git a/src/video_core/renderer_opengl/gl_blit_screen.h b/src/video_core/renderer_opengl/gl_blit_screen.h index f42f89dee..0c3d838f1 100644 --- a/src/video_core/renderer_opengl/gl_blit_screen.h +++ b/src/video_core/renderer_opengl/gl_blit_screen.h @@ -3,8 +3,9 @@ #pragma once +#include #include -#include +#include #include "core/hle/service/nvnflinger/pixel_format.h" #include "video_core/host1x/gpu_device_memory_manager.h" @@ -25,24 +26,12 @@ enum class ScalingFilter : u32; namespace OpenGL { class Device; -class FSR; -class FXAA; +class Layer; class ProgramManager; class RasterizerOpenGL; -class SMAA; class StateTracker; class WindowAdaptPass; -/// Structure used for storing information about the textures for the Switch screen -struct TextureInfo { - OGLTexture resource; - GLsizei width; - GLsizei height; - GLenum gl_format; - GLenum gl_type; - Service::android::PixelFormat pixel_format; -}; - /// Structure used for storing information about the display target for the Switch screen struct FramebufferTextureInfo { GLuint display_texture{}; @@ -60,20 +49,11 @@ public: Device& device); ~BlitScreen(); - void ConfigureFramebufferTexture(const Tegra::FramebufferConfig& framebuffer); - /// Draws the emulated screens to the emulator window. - void DrawScreen(const Tegra::FramebufferConfig& framebuffer, + void DrawScreen(std::span framebuffers, const Layout::FramebufferLayout& layout); - /// Loads framebuffer from emulated memory into the active OpenGL texture. - FramebufferTextureInfo LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer); - - FramebufferTextureInfo PrepareRenderTarget(const Tegra::FramebufferConfig& framebuffer); - private: - void CreateFXAA(); - void CreateSMAA(); void CreateWindowAdapt(); RasterizerOpenGL& rasterizer; @@ -82,18 +62,10 @@ private: ProgramManager& program_manager; Device& device; - /// Display information for Switch screen - TextureInfo framebuffer_texture; - - std::unique_ptr fsr; - std::unique_ptr fxaa; - std::unique_ptr smaa; - Settings::ScalingFilter current_window_adapt{}; std::unique_ptr window_adapt; - /// OpenGL framebuffer data - std::vector gl_framebuffer_data; + std::list layers; }; } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/layer.cpp b/src/video_core/renderer_opengl/present/layer.cpp new file mode 100644 index 000000000..8643e07c6 --- /dev/null +++ b/src/video_core/renderer_opengl/present/layer.cpp @@ -0,0 +1,215 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#include "video_core/framebuffer_config.h" +#include "video_core/renderer_opengl/gl_blit_screen.h" +#include "video_core/renderer_opengl/gl_rasterizer.h" +#include "video_core/renderer_opengl/present/fsr.h" +#include "video_core/renderer_opengl/present/fxaa.h" +#include "video_core/renderer_opengl/present/layer.h" +#include "video_core/renderer_opengl/present/present_uniforms.h" +#include "video_core/renderer_opengl/present/smaa.h" +#include "video_core/surface.h" +#include "video_core/textures/decoders.h" + +namespace OpenGL { + +Layer::Layer(RasterizerOpenGL& rasterizer_, Tegra::MaxwellDeviceMemoryManager& device_memory_) + : rasterizer(rasterizer_), device_memory(device_memory_) { + // Allocate textures for the screen + framebuffer_texture.resource.Create(GL_TEXTURE_2D); + + const GLuint texture = framebuffer_texture.resource.handle; + glTextureStorage2D(texture, 1, GL_RGBA8, 1, 1); + + // Clear screen to black + const u8 framebuffer_data[4] = {0, 0, 0, 0}; + glClearTexImage(framebuffer_texture.resource.handle, 0, GL_RGBA, GL_UNSIGNED_BYTE, + framebuffer_data); +} + +Layer::~Layer() = default; + +GLuint Layer::ConfigureDraw(std::array& out_matrix, + std::array& out_vertices, + ProgramManager& program_manager, + const Tegra::FramebufferConfig& framebuffer, + const Layout::FramebufferLayout& layout) { + FramebufferTextureInfo info = PrepareRenderTarget(framebuffer); + auto crop = Tegra::NormalizeCrop(framebuffer, info.width, info.height); + GLuint texture = info.display_texture; + + auto anti_aliasing = Settings::values.anti_aliasing.GetValue(); + if (anti_aliasing != Settings::AntiAliasing::None) { + glEnablei(GL_SCISSOR_TEST, 0); + auto viewport_width = Settings::values.resolution_info.ScaleUp(framebuffer_texture.width); + auto viewport_height = Settings::values.resolution_info.ScaleUp(framebuffer_texture.height); + + glScissorIndexed(0, 0, 0, viewport_width, viewport_height); + glViewportIndexedf(0, 0.0f, 0.0f, static_cast(viewport_width), + static_cast(viewport_height)); + + switch (anti_aliasing) { + case Settings::AntiAliasing::Fxaa: + CreateFXAA(); + texture = fxaa->Draw(program_manager, info.display_texture); + break; + case Settings::AntiAliasing::Smaa: + default: + CreateSMAA(); + texture = smaa->Draw(program_manager, info.display_texture); + break; + } + } + + glDisablei(GL_SCISSOR_TEST, 0); + + if (Settings::values.scaling_filter.GetValue() == Settings::ScalingFilter::Fsr) { + if (!fsr || fsr->NeedsRecreation(layout.screen)) { + fsr = std::make_unique(layout.screen.GetWidth(), layout.screen.GetHeight()); + } + + texture = fsr->Draw(program_manager, texture, info.scaled_width, info.scaled_height, crop); + crop = {0, 0, 1, 1}; + } + + out_matrix = + MakeOrthographicMatrix(static_cast(layout.width), static_cast(layout.height)); + + // Map the coordinates to the screen. + const auto& screen = layout.screen; + const auto x = screen.left; + const auto y = screen.top; + const auto w = screen.GetWidth(); + const auto h = screen.GetHeight(); + + out_vertices[0] = ScreenRectVertex(x, y, crop.left, crop.top); + out_vertices[1] = ScreenRectVertex(x + w, y, crop.right, crop.top); + out_vertices[2] = ScreenRectVertex(x, y + h, crop.left, crop.bottom); + out_vertices[3] = ScreenRectVertex(x + w, y + h, crop.right, crop.bottom); + + return texture; +} + +FramebufferTextureInfo Layer::PrepareRenderTarget(const Tegra::FramebufferConfig& framebuffer) { + // If framebuffer is provided, reload it from memory to a texture + if (framebuffer_texture.width != static_cast(framebuffer.width) || + framebuffer_texture.height != static_cast(framebuffer.height) || + framebuffer_texture.pixel_format != framebuffer.pixel_format || + gl_framebuffer_data.empty()) { + // Reallocate texture if the framebuffer size has changed. + // This is expected to not happen very often and hence should not be a + // performance problem. + ConfigureFramebufferTexture(framebuffer); + } + + // Load the framebuffer from memory if needed + return LoadFBToScreenInfo(framebuffer); +} + +FramebufferTextureInfo Layer::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer) { + const VAddr framebuffer_addr{framebuffer.address + framebuffer.offset}; + const auto accelerated_info = + rasterizer.AccelerateDisplay(framebuffer, framebuffer_addr, framebuffer.stride); + if (accelerated_info) { + return *accelerated_info; + } + + // Reset the screen info's display texture to its own permanent texture + FramebufferTextureInfo info{}; + info.display_texture = framebuffer_texture.resource.handle; + info.width = framebuffer.width; + info.height = framebuffer.height; + info.scaled_width = framebuffer.width; + info.scaled_height = framebuffer.height; + + // TODO(Rodrigo): Read this from HLE + constexpr u32 block_height_log2 = 4; + const auto pixel_format{ + VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format)}; + const u32 bytes_per_pixel{VideoCore::Surface::BytesPerBlock(pixel_format)}; + const u64 size_in_bytes{Tegra::Texture::CalculateSize( + true, bytes_per_pixel, framebuffer.stride, framebuffer.height, 1, block_height_log2, 0)}; + const u8* const host_ptr{device_memory.GetPointer(framebuffer_addr)}; + const std::span input_data(host_ptr, size_in_bytes); + Tegra::Texture::UnswizzleTexture(gl_framebuffer_data, input_data, bytes_per_pixel, + framebuffer.width, framebuffer.height, 1, block_height_log2, + 0); + + glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0); + glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast(framebuffer.stride)); + + // Update existing texture + // TODO: Test what happens on hardware when you change the framebuffer dimensions so that + // they differ from the LCD resolution. + // TODO: Applications could theoretically crash yuzu here by specifying too large + // framebuffer sizes. We should make sure that this cannot happen. + glTextureSubImage2D(framebuffer_texture.resource.handle, 0, 0, 0, framebuffer.width, + framebuffer.height, framebuffer_texture.gl_format, + framebuffer_texture.gl_type, gl_framebuffer_data.data()); + + glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); + + return info; +} + +void Layer::ConfigureFramebufferTexture(const Tegra::FramebufferConfig& framebuffer) { + framebuffer_texture.width = framebuffer.width; + framebuffer_texture.height = framebuffer.height; + framebuffer_texture.pixel_format = framebuffer.pixel_format; + + const auto pixel_format{ + VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format)}; + const u32 bytes_per_pixel{VideoCore::Surface::BytesPerBlock(pixel_format)}; + gl_framebuffer_data.resize(framebuffer_texture.width * framebuffer_texture.height * + bytes_per_pixel); + + GLint internal_format; + switch (framebuffer.pixel_format) { + case Service::android::PixelFormat::Rgba8888: + internal_format = GL_RGBA8; + framebuffer_texture.gl_format = GL_RGBA; + framebuffer_texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; + break; + case Service::android::PixelFormat::Rgb565: + internal_format = GL_RGB565; + framebuffer_texture.gl_format = GL_RGB; + framebuffer_texture.gl_type = GL_UNSIGNED_SHORT_5_6_5; + break; + default: + internal_format = GL_RGBA8; + framebuffer_texture.gl_format = GL_RGBA; + framebuffer_texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; + // UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}", + // static_cast(framebuffer.pixel_format)); + break; + } + + framebuffer_texture.resource.Release(); + framebuffer_texture.resource.Create(GL_TEXTURE_2D); + glTextureStorage2D(framebuffer_texture.resource.handle, 1, internal_format, + framebuffer_texture.width, framebuffer_texture.height); + + fxaa.reset(); + smaa.reset(); +} + +void Layer::CreateFXAA() { + smaa.reset(); + if (!fxaa) { + fxaa = std::make_unique( + Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), + Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); + } +} + +void Layer::CreateSMAA() { + fxaa.reset(); + if (!smaa) { + smaa = std::make_unique( + Settings::values.resolution_info.ScaleUp(framebuffer_texture.width), + Settings::values.resolution_info.ScaleUp(framebuffer_texture.height)); + } +} + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/layer.h b/src/video_core/renderer_opengl/present/layer.h new file mode 100644 index 000000000..ef1055abf --- /dev/null +++ b/src/video_core/renderer_opengl/present/layer.h @@ -0,0 +1,80 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include +#include + +#include "video_core/host1x/gpu_device_memory_manager.h" +#include "video_core/renderer_opengl/gl_resource_manager.h" + +namespace Layout { +struct FramebufferLayout; +} + +namespace Service::android { +enum class PixelFormat : u32; +}; + +namespace Tegra { +struct FramebufferConfig; +} + +namespace OpenGL { + +struct FramebufferTextureInfo; +class FSR; +class FXAA; +class ProgramManager; +class RasterizerOpenGL; +class SMAA; + +/// Structure used for storing information about the textures for the Switch screen +struct TextureInfo { + OGLTexture resource; + GLsizei width; + GLsizei height; + GLenum gl_format; + GLenum gl_type; + Service::android::PixelFormat pixel_format; +}; + +struct ScreenRectVertex; + +class Layer { +public: + explicit Layer(RasterizerOpenGL& rasterizer, Tegra::MaxwellDeviceMemoryManager& device_memory); + ~Layer(); + + GLuint ConfigureDraw(std::array& out_matrix, + std::array& out_vertices, + ProgramManager& program_manager, + const Tegra::FramebufferConfig& framebuffer, + const Layout::FramebufferLayout& layout); + +private: + /// Loads framebuffer from emulated memory into the active OpenGL texture. + FramebufferTextureInfo LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer); + FramebufferTextureInfo PrepareRenderTarget(const Tegra::FramebufferConfig& framebuffer); + void ConfigureFramebufferTexture(const Tegra::FramebufferConfig& framebuffer); + + void CreateFXAA(); + void CreateSMAA(); + +private: + RasterizerOpenGL& rasterizer; + Tegra::MaxwellDeviceMemoryManager& device_memory; + + /// OpenGL framebuffer data + std::vector gl_framebuffer_data; + + /// Display information for Switch screen + TextureInfo framebuffer_texture; + + std::unique_ptr fsr; + std::unique_ptr fxaa; + std::unique_ptr smaa; +}; + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/present_uniforms.h b/src/video_core/renderer_opengl/present/present_uniforms.h new file mode 100644 index 000000000..3a19f05c7 --- /dev/null +++ b/src/video_core/renderer_opengl/present/present_uniforms.h @@ -0,0 +1,43 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include "video_core/renderer_opengl/gl_resource_manager.h" + +namespace OpenGL { + +constexpr GLint PositionLocation = 0; +constexpr GLint TexCoordLocation = 1; +constexpr GLint ModelViewMatrixLocation = 0; + +struct ScreenRectVertex { + constexpr ScreenRectVertex() = default; + + constexpr ScreenRectVertex(u32 x, u32 y, GLfloat u, GLfloat v) + : position{{static_cast(x), static_cast(y)}}, tex_coord{{u, v}} {} + + std::array position{}; + std::array tex_coord{}; +}; + +/** + * Defines a 1:1 pixel orthographic projection matrix with (0,0) on the top-left + * corner and (width, height) on the lower-bottom. + * + * The projection part of the matrix is trivial, hence these operations are represented + * by a 3x2 matrix. + */ +static inline std::array MakeOrthographicMatrix(float width, float height) { + std::array matrix; // Laid out in column-major order + + // clang-format off + matrix[0] = 2.f / width; matrix[2] = 0.f; matrix[4] = -1.f; + matrix[1] = 0.f; matrix[3] = -2.f / height; matrix[5] = 1.f; + // Last matrix row is implicitly assumed to be [0, 0, 1]. + // clang-format on + + return matrix; +} + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/window_adapt_pass.cpp b/src/video_core/renderer_opengl/present/window_adapt_pass.cpp index 168fa1aea..4d681606b 100644 --- a/src/video_core/renderer_opengl/present/window_adapt_pass.cpp +++ b/src/video_core/renderer_opengl/present/window_adapt_pass.cpp @@ -2,47 +2,17 @@ // SPDX-License-Identifier: GPL-2.0-or-later #include "common/settings.h" +#include "video_core/framebuffer_config.h" #include "video_core/host_shaders/opengl_present_vert.h" #include "video_core/renderer_opengl/gl_device.h" #include "video_core/renderer_opengl/gl_shader_manager.h" #include "video_core/renderer_opengl/gl_shader_util.h" +#include "video_core/renderer_opengl/present/layer.h" +#include "video_core/renderer_opengl/present/present_uniforms.h" #include "video_core/renderer_opengl/present/window_adapt_pass.h" namespace OpenGL { -namespace { -constexpr GLint PositionLocation = 0; -constexpr GLint TexCoordLocation = 1; -constexpr GLint ModelViewMatrixLocation = 0; - -struct ScreenRectVertex { - constexpr ScreenRectVertex(u32 x, u32 y, GLfloat u, GLfloat v) - : position{{static_cast(x), static_cast(y)}}, tex_coord{{u, v}} {} - - std::array position; - std::array tex_coord; -}; - -/** - * Defines a 1:1 pixel orthographic projection matrix with (0,0) on the top-left - * corner and (width, height) on the lower-bottom. - * - * The projection part of the matrix is trivial, hence these operations are represented - * by a 3x2 matrix. - */ -std::array MakeOrthographicMatrix(float width, float height) { - std::array matrix; // Laid out in column-major order - - // clang-format off - matrix[0] = 2.f / width; matrix[2] = 0.f; matrix[4] = -1.f; - matrix[1] = 0.f; matrix[3] = -2.f / height; matrix[5] = 1.f; - // Last matrix row is implicitly assumed to be [0, 0, 1]. - // clang-format on - - return matrix; -} -} // namespace - WindowAdaptPass::WindowAdaptPass(const Device& device_, OGLSampler&& sampler_, std::string_view frag_source) : device(device_), sampler(std::move(sampler_)) { @@ -65,32 +35,30 @@ WindowAdaptPass::WindowAdaptPass(const Device& device_, OGLSampler&& sampler_, WindowAdaptPass::~WindowAdaptPass() = default; -void WindowAdaptPass::DrawToFramebuffer(ProgramManager& program_manager, GLuint texture, - const Layout::FramebufferLayout& layout, - const Common::Rectangle& crop) { - glBindTextureUnit(0, texture); +void WindowAdaptPass::DrawToFramebuffer(ProgramManager& program_manager, std::list& layers, + std::span framebuffers, + const Layout::FramebufferLayout& layout) { + GLint old_read_fb; + GLint old_draw_fb; + glGetIntegerv(GL_READ_FRAMEBUFFER_BINDING, &old_read_fb); + glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb); + + const size_t layer_count = framebuffers.size(); + std::vector textures(layer_count); + std::vector> matrices(layer_count); + std::vector> vertices(layer_count); + + auto layer_it = layers.begin(); + for (size_t i = 0; i < layer_count; i++) { + textures[i] = layer_it->ConfigureDraw(matrices[i], vertices[i], program_manager, + framebuffers[i], layout); + layer_it++; + } - const std::array ortho_matrix = - MakeOrthographicMatrix(static_cast(layout.width), static_cast(layout.height)); + glBindFramebuffer(GL_READ_FRAMEBUFFER, old_read_fb); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb); program_manager.BindPresentPrograms(vert.handle, frag.handle); - glProgramUniformMatrix3x2fv(vert.handle, ModelViewMatrixLocation, 1, GL_FALSE, - ortho_matrix.data()); - - // Map the coordinates to the screen. - const auto& screen = layout.screen; - const auto x = screen.left; - const auto y = screen.top; - const auto w = screen.GetWidth(); - const auto h = screen.GetHeight(); - - const std::array vertices = { - ScreenRectVertex(x, y, crop.left, crop.top), - ScreenRectVertex(x + w, y, crop.right, crop.top), - ScreenRectVertex(x, y + h, crop.left, crop.bottom), - ScreenRectVertex(x + w, y + h, crop.right, crop.bottom), - }; - glNamedBufferSubData(vertex_buffer.handle, 0, sizeof(vertices), std::data(vertices)); glDisable(GL_FRAMEBUFFER_SRGB); glViewportIndexedf(0, 0.0f, 0.0f, static_cast(layout.width), @@ -109,7 +77,7 @@ void WindowAdaptPass::DrawToFramebuffer(ProgramManager& program_manager, GLuint if (device.HasVertexBufferUnifiedMemory()) { glBindVertexBuffer(0, 0, 0, sizeof(ScreenRectVertex)); glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, 0, vertex_buffer_address, - sizeof(vertices)); + sizeof(decltype(vertices)::value_type)); } else { glBindVertexBuffer(0, vertex_buffer.handle, 0, sizeof(ScreenRectVertex)); } @@ -122,7 +90,14 @@ void WindowAdaptPass::DrawToFramebuffer(ProgramManager& program_manager, GLuint Settings::values.bg_blue.GetValue() / 255.0f, 1.0f); glClear(GL_COLOR_BUFFER_BIT); - glDrawArrays(GL_TRIANGLE_STRIP, 0, 4); + + for (size_t i = 0; i < layer_count; i++) { + glBindTextureUnit(0, textures[i]); + glProgramUniformMatrix3x2fv(vert.handle, ModelViewMatrixLocation, 1, GL_FALSE, + matrices[i].data()); + glNamedBufferSubData(vertex_buffer.handle, 0, sizeof(vertices[i]), std::data(vertices[i])); + glDrawArrays(GL_TRIANGLE_STRIP, 0, 4); + } } } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/present/window_adapt_pass.h b/src/video_core/renderer_opengl/present/window_adapt_pass.h index 65dcd09ff..00975a9c6 100644 --- a/src/video_core/renderer_opengl/present/window_adapt_pass.h +++ b/src/video_core/renderer_opengl/present/window_adapt_pass.h @@ -3,6 +3,9 @@ #pragma once +#include +#include + #include "common/math_util.h" #include "video_core/renderer_opengl/gl_resource_manager.h" @@ -10,9 +13,14 @@ namespace Layout { struct FramebufferLayout; } +namespace Tegra { +struct FramebufferConfig; +} + namespace OpenGL { class Device; +class Layer; class ProgramManager; class WindowAdaptPass final { @@ -21,9 +29,9 @@ public: std::string_view frag_source); ~WindowAdaptPass(); - void DrawToFramebuffer(ProgramManager& program_manager, GLuint texture, - const Layout::FramebufferLayout& layout, - const Common::Rectangle& crop); + void DrawToFramebuffer(ProgramManager& program_manager, std::list& layers, + std::span framebuffers, + const Layout::FramebufferLayout& layout); private: const Device& device; diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index 0d138c189..10a9f973c 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -130,10 +130,10 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { return; } - RenderScreenshot(*framebuffer); + RenderScreenshot(framebuffer); state_tracker.BindFramebuffer(0); - blit_screen->DrawScreen(*framebuffer, emu_window.GetFramebufferLayout()); + blit_screen->DrawScreen(std::span(framebuffer, 1), emu_window.GetFramebufferLayout()); ++m_current_frame; @@ -159,7 +159,7 @@ void RendererOpenGL::AddTelemetryFields() { telemetry_session.AddField(user_system, "GPU_OpenGL_Version", std::string(gl_version)); } -void RendererOpenGL::RenderScreenshot(const Tegra::FramebufferConfig& framebuffer) { +void RendererOpenGL::RenderScreenshot(const Tegra::FramebufferConfig* framebuffer) { if (!renderer_settings.screenshot_requested) { return; } @@ -181,7 +181,7 @@ void RendererOpenGL::RenderScreenshot(const Tegra::FramebufferConfig& framebuffe glRenderbufferStorage(GL_RENDERBUFFER, GL_SRGB8, layout.width, layout.height); glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, renderbuffer); - blit_screen->DrawScreen(framebuffer, layout); + blit_screen->DrawScreen(std::span(framebuffer, 1), layout); glBindBuffer(GL_PIXEL_PACK_BUFFER, 0); glPixelStorei(GL_PACK_ROW_LENGTH, 0); diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h index 7ab163372..df76d3d05 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.h +++ b/src/video_core/renderer_opengl/renderer_opengl.h @@ -52,7 +52,7 @@ public: private: void AddTelemetryFields(); - void RenderScreenshot(const Tegra::FramebufferConfig& framebuffer); + void RenderScreenshot(const Tegra::FramebufferConfig* framebuffer); Core::TelemetrySession& telemetry_session; Core::Frontend::EmuWindow& emu_window; -- cgit v1.2.3 From a595e9e8a7a6a742481b1cd05455d3c639095413 Mon Sep 17 00:00:00 2001 From: Liam Date: Mon, 22 Jan 2024 12:40:50 -0500 Subject: nvnflinger/gpu: implement layer stack composition --- src/core/CMakeLists.txt | 3 + .../hle/service/nvdrv/devices/nvdisp_disp0.cpp | 35 ++-- src/core/hle/service/nvdrv/devices/nvdisp_disp0.h | 10 +- src/core/hle/service/nvnflinger/buffer_item.h | 2 +- .../hle/service/nvnflinger/hardware_composer.cpp | 190 +++++++++++++++++++++ .../hle/service/nvnflinger/hardware_composer.h | 59 +++++++ src/core/hle/service/nvnflinger/hwc_layer.h | 27 +++ src/core/hle/service/nvnflinger/nvnflinger.cpp | 32 +--- src/core/hle/service/nvnflinger/nvnflinger.h | 1 + src/core/hle/service/vi/display/vi_display.cpp | 4 +- src/core/hle/service/vi/display/vi_display.h | 15 +- src/core/hle/service/vi/vi.cpp | 5 +- src/video_core/framebuffer_config.h | 3 +- src/video_core/gpu.cpp | 29 ++-- src/video_core/gpu.h | 4 +- src/video_core/gpu_thread.cpp | 6 - src/video_core/gpu_thread.h | 15 +- src/video_core/renderer_base.h | 2 +- src/video_core/renderer_null/renderer_null.cpp | 4 +- src/video_core/renderer_null/renderer_null.h | 2 +- src/video_core/renderer_opengl/renderer_opengl.cpp | 12 +- src/video_core/renderer_opengl/renderer_opengl.h | 4 +- src/video_core/renderer_vulkan/present/util.cpp | 78 +++++++++ src/video_core/renderer_vulkan/present/util.h | 4 + src/video_core/renderer_vulkan/renderer_vulkan.cpp | 167 +++--------------- src/video_core/renderer_vulkan/renderer_vulkan.h | 4 +- src/video_core/renderer_vulkan/vk_blit_screen.cpp | 2 +- src/video_core/renderer_vulkan/vk_blit_screen.h | 2 +- 28 files changed, 469 insertions(+), 252 deletions(-) create mode 100644 src/core/hle/service/nvnflinger/hardware_composer.cpp create mode 100644 src/core/hle/service/nvnflinger/hardware_composer.h create mode 100644 src/core/hle/service/nvnflinger/hwc_layer.h (limited to 'src/video_core/renderer_opengl') diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt index 570acb193..eb8f643a2 100644 --- a/src/core/CMakeLists.txt +++ b/src/core/CMakeLists.txt @@ -775,6 +775,9 @@ add_library(core STATIC hle/service/nvnflinger/graphic_buffer_producer.h hle/service/nvnflinger/hos_binder_driver_server.cpp hle/service/nvnflinger/hos_binder_driver_server.h + hle/service/nvnflinger/hardware_composer.cpp + hle/service/nvnflinger/hardware_composer.h + hle/service/nvnflinger/hwc_layer.h hle/service/nvnflinger/nvnflinger.cpp hle/service/nvnflinger/nvnflinger.h hle/service/nvnflinger/parcel.h diff --git a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp index c1ebbd62d..abe95303e 100644 --- a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp +++ b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp @@ -1,6 +1,8 @@ // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project // SPDX-License-Identifier: GPL-2.0-or-later +#include + #include "common/assert.h" #include "common/logging/log.h" #include "core/core.h" @@ -38,19 +40,30 @@ NvResult nvdisp_disp0::Ioctl3(DeviceFD fd, Ioctl command, std::span in void nvdisp_disp0::OnOpen(NvCore::SessionId session_id, DeviceFD fd) {} void nvdisp_disp0::OnClose(DeviceFD fd) {} -void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, android::PixelFormat format, u32 width, - u32 height, u32 stride, android::BufferTransformFlags transform, - const Common::Rectangle& crop_rect, - std::array& fences, u32 num_fences) { - const DAddr addr = nvmap.GetHandleAddress(buffer_handle); - LOG_TRACE(Service, - "Drawing from address {:X} offset {:08X} Width {} Height {} Stride {} Format {}", - addr, offset, width, height, stride, format); +void nvdisp_disp0::Composite(std::span sorted_layers) { + std::vector output_layers; + std::vector output_fences; + output_layers.reserve(sorted_layers.size()); + output_fences.reserve(sorted_layers.size()); + + for (auto& layer : sorted_layers) { + output_layers.emplace_back(Tegra::FramebufferConfig{ + .address = nvmap.GetHandleAddress(layer.buffer_handle), + .offset = layer.offset, + .width = layer.width, + .height = layer.height, + .stride = layer.stride, + .pixel_format = layer.format, + .transform_flags = layer.transform, + .crop_rect = layer.crop_rect, + }); - const Tegra::FramebufferConfig framebuffer{addr, offset, width, height, - stride, format, transform, crop_rect}; + for (size_t i = 0; i < layer.acquire_fence.num_fences; i++) { + output_fences.push_back(layer.acquire_fence.fences[i]); + } + } - system.GPU().RequestSwapBuffers(&framebuffer, fences, num_fences); + system.GPU().RequestComposite(std::move(output_layers), std::move(output_fences)); system.SpeedLimiter().DoSpeedLimiting(system.CoreTiming().GetGlobalTimeUs()); system.GetPerfStats().EndSystemFrame(); system.GetPerfStats().BeginSystemFrame(); diff --git a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h index 5f13a50a2..1082b85c2 100644 --- a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h +++ b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h @@ -8,8 +8,7 @@ #include "common/common_types.h" #include "common/math_util.h" #include "core/hle/service/nvdrv/devices/nvdevice.h" -#include "core/hle/service/nvnflinger/buffer_transform_flags.h" -#include "core/hle/service/nvnflinger/pixel_format.h" +#include "core/hle/service/nvnflinger/hwc_layer.h" namespace Service::Nvidia::NvCore { class Container; @@ -35,11 +34,8 @@ public: void OnOpen(NvCore::SessionId session_id, DeviceFD fd) override; void OnClose(DeviceFD fd) override; - /// Performs a screen flip, drawing the buffer pointed to by the handle. - void flip(u32 buffer_handle, u32 offset, android::PixelFormat format, u32 width, u32 height, - u32 stride, android::BufferTransformFlags transform, - const Common::Rectangle& crop_rect, - std::array& fences, u32 num_fences); + /// Performs a screen flip, compositing each buffer. + void Composite(std::span sorted_layers); Kernel::KEvent* QueryEvent(u32 event_id) override; diff --git a/src/core/hle/service/nvnflinger/buffer_item.h b/src/core/hle/service/nvnflinger/buffer_item.h index 7fd808f54..f9f262628 100644 --- a/src/core/hle/service/nvnflinger/buffer_item.h +++ b/src/core/hle/service/nvnflinger/buffer_item.h @@ -40,7 +40,7 @@ public: bool is_droppable{}; bool acquire_called{}; bool transform_to_display_inverse{}; - s32 swap_interval{}; + u32 swap_interval{}; }; } // namespace Service::android diff --git a/src/core/hle/service/nvnflinger/hardware_composer.cpp b/src/core/hle/service/nvnflinger/hardware_composer.cpp new file mode 100644 index 000000000..54889bb4f --- /dev/null +++ b/src/core/hle/service/nvnflinger/hardware_composer.cpp @@ -0,0 +1,190 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + +#include + +#include "common/microprofile.h" +#include "core/hle/service/nvdrv/devices/nvdisp_disp0.h" +#include "core/hle/service/nvnflinger/buffer_item.h" +#include "core/hle/service/nvnflinger/buffer_item_consumer.h" +#include "core/hle/service/nvnflinger/buffer_queue_producer.h" +#include "core/hle/service/nvnflinger/hardware_composer.h" +#include "core/hle/service/nvnflinger/hwc_layer.h" +#include "core/hle/service/nvnflinger/ui/graphic_buffer.h" +#include "core/hle/service/vi/display/vi_display.h" +#include "core/hle/service/vi/layer/vi_layer.h" + +namespace Service::Nvnflinger { + +HardwareComposer::HardwareComposer() = default; +HardwareComposer::~HardwareComposer() = default; + +u32 HardwareComposer::ComposeLocked(VI::Display& display, Nvidia::Devices::nvdisp_disp0& nvdisp, + u32 frame_advance) { + boost::container::small_vector composition_stack; + + m_frame_number += frame_advance; + + // Release any necessary framebuffers. + for (auto& [layer_id, framebuffer] : m_framebuffers) { + if (framebuffer.release_frame_number > m_frame_number) { + // Not yet ready to release this framebuffer. + continue; + } + + if (!framebuffer.is_acquired) { + // Already released. + continue; + } + + if (auto* layer = display.FindLayer(layer_id); layer != nullptr) { + // TODO: support release fence + // This is needed to prevent screen tearing + layer->GetConsumer().ReleaseBuffer(framebuffer.item, android::Fence::NoFence()); + framebuffer.is_acquired = false; + } + } + + // Determine the number of vsync periods to wait before composing again. + std::optional swap_interval{}; + bool has_acquired_buffer{}; + + // Acquire all necessary framebuffers. + for (size_t i = 0; i < display.GetNumLayers(); i++) { + auto& layer = display.GetLayer(i); + auto layer_id = layer.GetLayerId(); + + // Try to fetch the framebuffer (either new or stale). + const auto result = this->CacheFramebufferLocked(layer, layer_id); + + // If we failed, skip this layer. + if (result == CacheStatus::NoBufferAvailable) { + continue; + } + + // If we acquired a new buffer, we need to present. + if (result == CacheStatus::BufferAcquired) { + has_acquired_buffer = true; + } + + const auto& buffer = m_framebuffers[layer_id]; + const auto& item = buffer.item; + const auto& igbp_buffer = *item.graphic_buffer; + + // TODO: get proper Z-index from layer + composition_stack.emplace_back(HwcLayer{ + .buffer_handle = igbp_buffer.BufferId(), + .offset = igbp_buffer.Offset(), + .format = igbp_buffer.ExternalFormat(), + .width = igbp_buffer.Width(), + .height = igbp_buffer.Height(), + .stride = igbp_buffer.Stride(), + .z_index = 0, + .transform = static_cast(item.transform), + .crop_rect = item.crop, + .acquire_fence = item.fence, + }); + + // We need to compose again either before this frame is supposed to + // be released, or exactly on the vsync period it should be released. + // + // TODO: handle cases where swap intervals are relatively prime. So far, + // only swap intervals of 0, 1 and 2 have been observed, but if 3 were + // to be introduced, this would cause an issue. + if (swap_interval) { + swap_interval = std::min(*swap_interval, item.swap_interval); + } else { + swap_interval = item.swap_interval; + } + } + + // If any new buffers were acquired, we can present. + if (has_acquired_buffer) { + // Sort by Z-index. + std::stable_sort(composition_stack.begin(), composition_stack.end(), + [&](auto& l, auto& r) { return l.z_index < r.z_index; }); + + // Composite. + nvdisp.Composite(composition_stack); + } + + // Render MicroProfile. + MicroProfileFlip(); + + // If we advanced, then advance by at least 1 frame. + if (swap_interval) { + return std::max(*swap_interval, 1U); + } + + // Otherwise, advance by exactly one frame. + return 1U; +} + +void HardwareComposer::RemoveLayerLocked(VI::Display& display, LayerId layer_id) { + // Check if we are tracking a slot with this layer_id. + const auto it = m_framebuffers.find(layer_id); + if (it == m_framebuffers.end()) { + return; + } + + // Try to release the buffer item. + auto* const layer = display.FindLayer(layer_id); + if (layer && it->second.is_acquired) { + layer->GetConsumer().ReleaseBuffer(it->second.item, android::Fence::NoFence()); + } + + // Erase the slot. + m_framebuffers.erase(it); +} + +bool HardwareComposer::TryAcquireFramebufferLocked(VI::Layer& layer, Framebuffer& framebuffer) { + // Attempt the update. + const auto status = layer.GetConsumer().AcquireBuffer(&framebuffer.item, {}, false); + if (status != android::Status::NoError) { + return false; + } + + // We succeeded, so set the new release frame info. + framebuffer.release_frame_number = + m_frame_number + std::max(1U, framebuffer.item.swap_interval); + framebuffer.is_acquired = true; + + return true; +} + +HardwareComposer::CacheStatus HardwareComposer::CacheFramebufferLocked(VI::Layer& layer, + LayerId layer_id) { + // Check if this framebuffer is already present. + const auto it = m_framebuffers.find(layer_id); + if (it != m_framebuffers.end()) { + // If it's currently still acquired, we are done. + if (it->second.is_acquired) { + return CacheStatus::CachedBufferReused; + } + + // Try to acquire a new item. + if (this->TryAcquireFramebufferLocked(layer, it->second)) { + // We got a new item. + return CacheStatus::BufferAcquired; + } else { + // We didn't acquire a new item, but we can reuse the slot. + return CacheStatus::CachedBufferReused; + } + } + + // Framebuffer is not present, so try to create it. + Framebuffer framebuffer{}; + + if (this->TryAcquireFramebufferLocked(layer, framebuffer)) { + // Move the buffer item into a new slot. + m_framebuffers.emplace(layer_id, std::move(framebuffer)); + + // We succeeded. + return CacheStatus::BufferAcquired; + } + + // We couldn't acquire the buffer item, so don't create a slot. + return CacheStatus::NoBufferAvailable; +} + +} // namespace Service::Nvnflinger diff --git a/src/core/hle/service/nvnflinger/hardware_composer.h b/src/core/hle/service/nvnflinger/hardware_composer.h new file mode 100644 index 000000000..611afc169 --- /dev/null +++ b/src/core/hle/service/nvnflinger/hardware_composer.h @@ -0,0 +1,59 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + +#pragma once + +#include +#include + +#include "core/hle/service/nvnflinger/buffer_item.h" + +namespace Service::Nvidia::Devices { +class nvdisp_disp0; +} + +namespace Service::VI { +class Display; +class Layer; +} // namespace Service::VI + +namespace Service::Nvnflinger { + +using LayerId = u64; + +class HardwareComposer { +public: + explicit HardwareComposer(); + ~HardwareComposer(); + + u32 ComposeLocked(VI::Display& display, Nvidia::Devices::nvdisp_disp0& nvdisp, + u32 frame_advance); + void RemoveLayerLocked(VI::Display& display, LayerId layer_id); + +private: + // TODO: do we want to track frame number in vi instead? + u64 m_frame_number{0}; + +private: + using ReleaseFrameNumber = u64; + + struct Framebuffer { + android::BufferItem item{}; + ReleaseFrameNumber release_frame_number{}; + bool is_acquired{false}; + }; + + enum class CacheStatus : u32 { + NoBufferAvailable, + BufferAcquired, + CachedBufferReused, + }; + + boost::container::flat_map m_framebuffers{}; + +private: + bool TryAcquireFramebufferLocked(VI::Layer& layer, Framebuffer& framebuffer); + CacheStatus CacheFramebufferLocked(VI::Layer& layer, LayerId layer_id); +}; + +} // namespace Service::Nvnflinger diff --git a/src/core/hle/service/nvnflinger/hwc_layer.h b/src/core/hle/service/nvnflinger/hwc_layer.h new file mode 100644 index 000000000..3af668a25 --- /dev/null +++ b/src/core/hle/service/nvnflinger/hwc_layer.h @@ -0,0 +1,27 @@ +// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + +#pragma once + +#include "common/math_util.h" +#include "core/hle/service/nvdrv/nvdata.h" +#include "core/hle/service/nvnflinger/buffer_transform_flags.h" +#include "core/hle/service/nvnflinger/pixel_format.h" +#include "core/hle/service/nvnflinger/ui/fence.h" + +namespace Service::Nvnflinger { + +struct HwcLayer { + u32 buffer_handle; + u32 offset; + android::PixelFormat format; + u32 width; + u32 height; + u32 stride; + s32 z_index; + android::BufferTransformFlags transform; + Common::Rectangle crop_rect; + android::Fence acquire_fence; +}; + +} // namespace Service::Nvnflinger diff --git a/src/core/hle/service/nvnflinger/nvnflinger.cpp b/src/core/hle/service/nvnflinger/nvnflinger.cpp index 51133853c..e775a2ca8 100644 --- a/src/core/hle/service/nvnflinger/nvnflinger.cpp +++ b/src/core/hle/service/nvnflinger/nvnflinger.cpp @@ -18,6 +18,7 @@ #include "core/hle/service/nvnflinger/buffer_item_consumer.h" #include "core/hle/service/nvnflinger/buffer_queue_core.h" #include "core/hle/service/nvnflinger/fb_share_buffer_manager.h" +#include "core/hle/service/nvnflinger/hardware_composer.h" #include "core/hle/service/nvnflinger/hos_binder_driver_server.h" #include "core/hle/service/nvnflinger/nvnflinger.h" #include "core/hle/service/nvnflinger/ui/graphic_buffer.h" @@ -279,45 +280,18 @@ void Nvnflinger::Compose() { SCOPE_EXIT({ display.SignalVSyncEvent(); }); // Don't do anything for displays without layers. - if (!display.HasLayers()) - continue; - - // TODO(Subv): Support more than 1 layer. - VI::Layer& layer = display.GetLayer(0); - - android::BufferItem buffer{}; - const auto status = layer.GetConsumer().AcquireBuffer(&buffer, {}, false); - - if (status != android::Status::NoError) { + if (!display.HasLayers()) { continue; } - const auto& igbp_buffer = *buffer.graphic_buffer; - if (!system.IsPoweredOn()) { return; // We are likely shutting down } - // Now send the buffer to the GPU for drawing. - // TODO(Subv): Support more than just disp0. The display device selection is probably based - // on which display we're drawing (Default, Internal, External, etc) auto nvdisp = nvdrv->GetDevice(disp_fd); ASSERT(nvdisp); - Common::Rectangle crop_rect{ - static_cast(buffer.crop.Left()), static_cast(buffer.crop.Top()), - static_cast(buffer.crop.Right()), static_cast(buffer.crop.Bottom())}; - - nvdisp->flip(igbp_buffer.BufferId(), igbp_buffer.Offset(), igbp_buffer.ExternalFormat(), - igbp_buffer.Width(), igbp_buffer.Height(), igbp_buffer.Stride(), - static_cast(buffer.transform), crop_rect, - buffer.fence.fences, buffer.fence.num_fences); - - MicroProfileFlip(); - - swap_interval = buffer.swap_interval; - - layer.GetConsumer().ReleaseBuffer(buffer, android::Fence::NoFence()); + swap_interval = display.GetComposer().ComposeLocked(display, *nvdisp, swap_interval); } } diff --git a/src/core/hle/service/nvnflinger/nvnflinger.h b/src/core/hle/service/nvnflinger/nvnflinger.h index 369439142..73ff36620 100644 --- a/src/core/hle/service/nvnflinger/nvnflinger.h +++ b/src/core/hle/service/nvnflinger/nvnflinger.h @@ -46,6 +46,7 @@ class BufferQueueProducer; namespace Service::Nvnflinger { class FbShareBufferManager; +class HardwareComposer; class HosBinderDriverServer; class Nvnflinger final { diff --git a/src/core/hle/service/vi/display/vi_display.cpp b/src/core/hle/service/vi/display/vi_display.cpp index dab1905cc..7f2af9acc 100644 --- a/src/core/hle/service/vi/display/vi_display.cpp +++ b/src/core/hle/service/vi/display/vi_display.cpp @@ -16,6 +16,7 @@ #include "core/hle/service/nvnflinger/buffer_queue_consumer.h" #include "core/hle/service/nvnflinger/buffer_queue_core.h" #include "core/hle/service/nvnflinger/buffer_queue_producer.h" +#include "core/hle/service/nvnflinger/hardware_composer.h" #include "core/hle/service/nvnflinger/hos_binder_driver_server.h" #include "core/hle/service/vi/display/vi_display.h" #include "core/hle/service/vi/layer/vi_layer.h" @@ -43,6 +44,7 @@ Display::Display(u64 id, std::string name_, KernelHelpers::ServiceContext& service_context_, Core::System& system_) : display_id{id}, name{std::move(name_)}, hos_binder_driver_server{hos_binder_driver_server_}, service_context{service_context_} { + hardware_composer = std::make_unique(); vsync_event = service_context.CreateEvent(fmt::format("Display VSync Event {}", id)); } @@ -81,8 +83,6 @@ void Display::SignalVSyncEvent() { void Display::CreateLayer(u64 layer_id, u32 binder_id, Service::Nvidia::NvCore::Container& nv_core) { - ASSERT_MSG(layers.empty(), "Only one layer is supported per display at the moment"); - auto [core, producer, consumer] = CreateBufferQueue(service_context, nv_core.GetNvMapFile()); auto buffer_item_consumer = std::make_shared(std::move(consumer)); diff --git a/src/core/hle/service/vi/display/vi_display.h b/src/core/hle/service/vi/display/vi_display.h index 8eb8a5155..220292cff 100644 --- a/src/core/hle/service/vi/display/vi_display.h +++ b/src/core/hle/service/vi/display/vi_display.h @@ -11,9 +11,14 @@ #include "common/common_types.h" #include "core/hle/result.h" +namespace Core { +class System; +} + namespace Kernel { class KEvent; -} +class KReadableEvent; +} // namespace Kernel namespace Service::android { class BufferQueueProducer; @@ -24,8 +29,9 @@ class ServiceContext; } namespace Service::Nvnflinger { +class HardwareComposer; class HosBinderDriverServer; -} +} // namespace Service::Nvnflinger namespace Service::Nvidia::NvCore { class Container; @@ -118,6 +124,10 @@ public: /// const Layer* FindLayer(u64 layer_id) const; + Nvnflinger::HardwareComposer& GetComposer() const { + return *hardware_composer; + } + private: u64 display_id; std::string name; @@ -125,6 +135,7 @@ private: KernelHelpers::ServiceContext& service_context; std::vector> layers; + std::unique_ptr hardware_composer; Kernel::KEvent* vsync_event{}; bool is_abandoned{}; }; diff --git a/src/core/hle/service/vi/vi.cpp b/src/core/hle/service/vi/vi.cpp index 73058db9a..d508ed28c 100644 --- a/src/core/hle/service/vi/vi.cpp +++ b/src/core/hle/service/vi/vi.cpp @@ -195,8 +195,9 @@ private: void GetSharedBufferMemoryHandleId(HLERequestContext& ctx) { IPC::RequestParser rp{ctx}; const u64 buffer_id = rp.PopRaw(); + const u64 aruid = ctx.GetPID(); - LOG_INFO(Service_VI, "called. buffer_id={:#x}", buffer_id); + LOG_INFO(Service_VI, "called. buffer_id={:#x}, aruid={:#x}", buffer_id, aruid); struct OutputParameters { s32 nvmap_handle; @@ -206,7 +207,7 @@ private: OutputParameters out{}; Nvnflinger::SharedMemoryPoolLayout layout{}; const auto result = nvnflinger.GetSystemBufferManager().GetSharedBufferMemoryHandleId( - &out.size, &out.nvmap_handle, &layout, buffer_id, 0); + &out.size, &out.nvmap_handle, &layout, buffer_id, aruid); ctx.WriteBuffer(&layout, sizeof(layout)); diff --git a/src/video_core/framebuffer_config.h b/src/video_core/framebuffer_config.h index 10ddc75a7..6a18b76fb 100644 --- a/src/video_core/framebuffer_config.h +++ b/src/video_core/framebuffer_config.h @@ -7,6 +7,7 @@ #include "common/math_util.h" #include "core/hle/service/nvnflinger/buffer_transform_flags.h" #include "core/hle/service/nvnflinger/pixel_format.h" +#include "core/hle/service/nvnflinger/ui/fence.h" namespace Tegra { @@ -21,7 +22,7 @@ struct FramebufferConfig { u32 stride{}; Service::android::PixelFormat pixel_format{}; Service::android::BufferTransformFlags transform_flags{}; - Common::Rectangle crop_rect; + Common::Rectangle crop_rect{}; }; Common::Rectangle NormalizeCrop(const FramebufferConfig& framebuffer, u32 texture_width, diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index 609704b33..f4a5d831c 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -274,11 +274,6 @@ struct GPU::Impl { } } - /// Swap buffers (render frame) - void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { - gpu_thread.SwapBuffers(framebuffer); - } - /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory void FlushRegion(DAddr addr, u64 size) { gpu_thread.FlushRegion(addr, size); @@ -313,8 +308,9 @@ struct GPU::Impl { gpu_thread.FlushAndInvalidateRegion(addr, size); } - void RequestSwapBuffers(const Tegra::FramebufferConfig* framebuffer, - std::array& fences, size_t num_fences) { + void RequestComposite(std::vector&& layers, + std::vector&& fences) { + size_t num_fences{fences.size()}; size_t current_request_counter{}; { std::unique_lock lk(request_swap_mutex); @@ -328,13 +324,12 @@ struct GPU::Impl { } } const auto wait_fence = - RequestSyncOperation([this, current_request_counter, framebuffer, fences, num_fences] { + RequestSyncOperation([this, current_request_counter, &layers, &fences, num_fences] { auto& syncpoint_manager = host1x.GetSyncpointManager(); if (num_fences == 0) { - renderer->SwapBuffers(framebuffer); + renderer->Composite(layers); } - const auto executer = [this, current_request_counter, - framebuffer_copy = *framebuffer]() { + const auto executer = [this, current_request_counter, layers_copy = layers]() { { std::unique_lock lk(request_swap_mutex); if (--request_swap_counters[current_request_counter] != 0) { @@ -342,7 +337,7 @@ struct GPU::Impl { } free_swap_counters.push_back(current_request_counter); } - renderer->SwapBuffers(&framebuffer_copy); + renderer->Composite(layers_copy); }; for (size_t i = 0; i < num_fences; i++) { syncpoint_manager.RegisterGuestAction(fences[i].id, fences[i].value, executer); @@ -505,9 +500,9 @@ const VideoCore::ShaderNotify& GPU::ShaderNotify() const { return impl->ShaderNotify(); } -void GPU::RequestSwapBuffers(const Tegra::FramebufferConfig* framebuffer, - std::array& fences, size_t num_fences) { - impl->RequestSwapBuffers(framebuffer, fences, num_fences); +void GPU::RequestComposite(std::vector&& layers, + std::vector&& fences) { + impl->RequestComposite(std::move(layers), std::move(fences)); } u64 GPU::GetTicks() const { @@ -554,10 +549,6 @@ void GPU::ClearCdmaInstance(u32 id) { impl->ClearCdmaInstance(id); } -void GPU::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { - impl->SwapBuffers(framebuffer); -} - VideoCore::RasterizerDownloadArea GPU::OnCPURead(PAddr addr, u64 size) { return impl->OnCPURead(addr, size); } diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h index b3c1d15bd..c4602ca37 100644 --- a/src/video_core/gpu.h +++ b/src/video_core/gpu.h @@ -212,8 +212,8 @@ public: void RendererFrameEndNotify(); - void RequestSwapBuffers(const Tegra::FramebufferConfig* framebuffer, - std::array& fences, size_t num_fences); + void RequestComposite(std::vector&& layers, + std::vector&& fences); /// Performs any additional setup necessary in order to begin GPU emulation. /// This can be used to launch any necessary threads and register any necessary diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp index 788d4f61e..58d8110b8 100644 --- a/src/video_core/gpu_thread.cpp +++ b/src/video_core/gpu_thread.cpp @@ -40,8 +40,6 @@ static void RunThread(std::stop_token stop_token, Core::System& system, } if (auto* submit_list = std::get_if(&next.data)) { scheduler.Push(submit_list->channel, std::move(submit_list->entries)); - } else if (const auto* data = std::get_if(&next.data)) { - renderer.SwapBuffers(data->framebuffer ? &*data->framebuffer : nullptr); } else if (std::holds_alternative(next.data)) { system.GPU().TickWork(); } else if (const auto* flush = std::get_if(&next.data)) { @@ -78,10 +76,6 @@ void ThreadManager::SubmitList(s32 channel, Tegra::CommandList&& entries) { PushCommand(SubmitListCommand(channel, std::move(entries))); } -void ThreadManager::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { - PushCommand(SwapBuffersCommand(framebuffer ? std::make_optional(*framebuffer) : std::nullopt)); -} - void ThreadManager::FlushRegion(DAddr addr, u64 size) { if (!is_async) { // Always flush with synchronous GPU mode diff --git a/src/video_core/gpu_thread.h b/src/video_core/gpu_thread.h index 2de25e9ef..dc0fce9f8 100644 --- a/src/video_core/gpu_thread.h +++ b/src/video_core/gpu_thread.h @@ -44,14 +44,6 @@ struct SubmitListCommand final { Tegra::CommandList entries; }; -/// Command to signal to the GPU thread that a swap buffers is pending -struct SwapBuffersCommand final { - explicit SwapBuffersCommand(std::optional framebuffer_) - : framebuffer{std::move(framebuffer_)} {} - - std::optional framebuffer; -}; - /// Command to signal to the GPU thread to flush a region struct FlushRegionCommand final { explicit constexpr FlushRegionCommand(DAddr addr_, u64 size_) : addr{addr_}, size{size_} {} @@ -81,8 +73,8 @@ struct FlushAndInvalidateRegionCommand final { struct GPUTickCommand final {}; using CommandData = - std::variant; + std::variant; struct CommandDataContainer { CommandDataContainer() = default; @@ -118,9 +110,6 @@ public: /// Push GPU command entries to be processed void SubmitList(s32 channel, Tegra::CommandList&& entries); - /// Swap buffers (render frame) - void SwapBuffers(const Tegra::FramebufferConfig* framebuffer); - /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory void FlushRegion(DAddr addr, u64 size); diff --git a/src/video_core/renderer_base.h b/src/video_core/renderer_base.h index 78ea5208b..3ad180f67 100644 --- a/src/video_core/renderer_base.h +++ b/src/video_core/renderer_base.h @@ -38,7 +38,7 @@ public: virtual ~RendererBase(); /// Finalize rendering the guest frame and draw into the presentation texture - virtual void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) = 0; + virtual void Composite(std::span layers) = 0; [[nodiscard]] virtual RasterizerInterface* ReadRasterizer() = 0; diff --git a/src/video_core/renderer_null/renderer_null.cpp b/src/video_core/renderer_null/renderer_null.cpp index 078feb925..c89daff53 100644 --- a/src/video_core/renderer_null/renderer_null.cpp +++ b/src/video_core/renderer_null/renderer_null.cpp @@ -13,8 +13,8 @@ RendererNull::RendererNull(Core::Frontend::EmuWindow& emu_window, Tegra::GPU& gp RendererNull::~RendererNull() = default; -void RendererNull::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { - if (!framebuffer) { +void RendererNull::Composite(std::span framebuffers) { + if (framebuffers.empty()) { return; } diff --git a/src/video_core/renderer_null/renderer_null.h b/src/video_core/renderer_null/renderer_null.h index 9531b43f6..063b476bb 100644 --- a/src/video_core/renderer_null/renderer_null.h +++ b/src/video_core/renderer_null/renderer_null.h @@ -17,7 +17,7 @@ public: std::unique_ptr context); ~RendererNull() override; - void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override; + void Composite(std::span framebuffer) override; VideoCore::RasterizerInterface* ReadRasterizer() override { return &m_rasterizer; diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index 10a9f973c..e33a32592 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -125,15 +125,15 @@ RendererOpenGL::RendererOpenGL(Core::TelemetrySession& telemetry_session_, RendererOpenGL::~RendererOpenGL() = default; -void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { - if (!framebuffer) { +void RendererOpenGL::Composite(std::span framebuffers) { + if (framebuffers.empty()) { return; } - RenderScreenshot(framebuffer); + RenderScreenshot(framebuffers); state_tracker.BindFramebuffer(0); - blit_screen->DrawScreen(std::span(framebuffer, 1), emu_window.GetFramebufferLayout()); + blit_screen->DrawScreen(framebuffers, emu_window.GetFramebufferLayout()); ++m_current_frame; @@ -159,7 +159,7 @@ void RendererOpenGL::AddTelemetryFields() { telemetry_session.AddField(user_system, "GPU_OpenGL_Version", std::string(gl_version)); } -void RendererOpenGL::RenderScreenshot(const Tegra::FramebufferConfig* framebuffer) { +void RendererOpenGL::RenderScreenshot(std::span framebuffers) { if (!renderer_settings.screenshot_requested) { return; } @@ -181,7 +181,7 @@ void RendererOpenGL::RenderScreenshot(const Tegra::FramebufferConfig* framebuffe glRenderbufferStorage(GL_RENDERBUFFER, GL_SRGB8, layout.width, layout.height); glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, renderbuffer); - blit_screen->DrawScreen(std::span(framebuffer, 1), layout); + blit_screen->DrawScreen(framebuffers, layout); glBindBuffer(GL_PIXEL_PACK_BUFFER, 0); glPixelStorei(GL_PACK_ROW_LENGTH, 0); diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h index df76d3d05..c4625c96e 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.h +++ b/src/video_core/renderer_opengl/renderer_opengl.h @@ -40,7 +40,7 @@ public: std::unique_ptr context_); ~RendererOpenGL() override; - void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override; + void Composite(std::span framebuffers) override; VideoCore::RasterizerInterface* ReadRasterizer() override { return &rasterizer; @@ -52,7 +52,7 @@ public: private: void AddTelemetryFields(); - void RenderScreenshot(const Tegra::FramebufferConfig* framebuffer); + void RenderScreenshot(std::span framebuffers); Core::TelemetrySession& telemetry_session; Core::Frontend::EmuWindow& emu_window; diff --git a/src/video_core/renderer_vulkan/present/util.cpp b/src/video_core/renderer_vulkan/present/util.cpp index 7bff1c436..6ee16595d 100644 --- a/src/video_core/renderer_vulkan/present/util.cpp +++ b/src/video_core/renderer_vulkan/present/util.cpp @@ -7,6 +7,20 @@ namespace Vulkan { +vk::Buffer CreateWrappedBuffer(MemoryAllocator& allocator, VkDeviceSize size, MemoryUsage usage) { + const VkBufferCreateInfo dst_buffer_info{ + .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO, + .pNext = nullptr, + .flags = 0, + .size = size, + .usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT, + .sharingMode = VK_SHARING_MODE_EXCLUSIVE, + .queueFamilyIndexCount = 0, + .pQueueFamilyIndices = nullptr, + }; + return allocator.CreateBuffer(dst_buffer_info, usage); +} + vk::Image CreateWrappedImage(MemoryAllocator& allocator, VkExtent2D dimensions, VkFormat format) { const VkImageCreateInfo image_ci{ .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO, @@ -96,6 +110,70 @@ void UploadImage(const Device& device, MemoryAllocator& allocator, Scheduler& sc scheduler.Finish(); } +void DownloadColorImage(vk::CommandBuffer& cmdbuf, VkImage image, VkBuffer buffer, + VkExtent3D extent) { + const VkImageMemoryBarrier read_barrier{ + .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, + .pNext = nullptr, + .srcAccessMask = VK_ACCESS_MEMORY_WRITE_BIT, + .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT, + .oldLayout = VK_IMAGE_LAYOUT_GENERAL, + .newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, + .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, + .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, + .image = image, + .subresourceRange{ + .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, + .baseMipLevel = 0, + .levelCount = VK_REMAINING_MIP_LEVELS, + .baseArrayLayer = 0, + .layerCount = VK_REMAINING_ARRAY_LAYERS, + }, + }; + const VkImageMemoryBarrier image_write_barrier{ + .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, + .pNext = nullptr, + .srcAccessMask = 0, + .dstAccessMask = VK_ACCESS_MEMORY_WRITE_BIT, + .oldLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, + .newLayout = VK_IMAGE_LAYOUT_GENERAL, + .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, + .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, + .image = image, + .subresourceRange{ + .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, + .baseMipLevel = 0, + .levelCount = VK_REMAINING_MIP_LEVELS, + .baseArrayLayer = 0, + .layerCount = VK_REMAINING_ARRAY_LAYERS, + }, + }; + static constexpr VkMemoryBarrier memory_write_barrier{ + .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER, + .pNext = nullptr, + .srcAccessMask = VK_ACCESS_MEMORY_WRITE_BIT, + .dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT, + }; + const VkBufferImageCopy copy{ + .bufferOffset = 0, + .bufferRowLength = 0, + .bufferImageHeight = 0, + .imageSubresource{ + .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, + .mipLevel = 0, + .baseArrayLayer = 0, + .layerCount = 1, + }, + .imageOffset{.x = 0, .y = 0, .z = 0}, + .imageExtent{extent}, + }; + cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, + read_barrier); + cmdbuf.CopyImageToBuffer(image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, buffer, copy); + cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, + memory_write_barrier, nullptr, image_write_barrier); +} + vk::ImageView CreateWrappedImageView(const Device& device, vk::Image& image, VkFormat format) { return device.GetLogical().CreateImageView(VkImageViewCreateInfo{ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO, diff --git a/src/video_core/renderer_vulkan/present/util.h b/src/video_core/renderer_vulkan/present/util.h index fb4e4a8e4..1104aaa15 100644 --- a/src/video_core/renderer_vulkan/present/util.h +++ b/src/video_core/renderer_vulkan/present/util.h @@ -11,12 +11,16 @@ namespace Vulkan { #define ARRAY_TO_SPAN(a) std::span(a, (sizeof(a) / sizeof(a[0]))) +vk::Buffer CreateWrappedBuffer(MemoryAllocator& allocator, VkDeviceSize size, MemoryUsage usage); + vk::Image CreateWrappedImage(MemoryAllocator& allocator, VkExtent2D dimensions, VkFormat format); void TransitionImageLayout(vk::CommandBuffer& cmdbuf, VkImage image, VkImageLayout target_layout, VkImageLayout source_layout = VK_IMAGE_LAYOUT_GENERAL); void UploadImage(const Device& device, MemoryAllocator& allocator, Scheduler& scheduler, vk::Image& image, VkExtent2D dimensions, VkFormat format, std::span initial_contents = {}); +void DownloadColorImage(vk::CommandBuffer& cmdbuf, VkImage image, VkBuffer buffer, + VkExtent3D extent); void ClearColorImage(vk::CommandBuffer& cmdbuf, VkImage image); vk::ImageView CreateWrappedImageView(const Device& device, vk::Image& image, VkFormat format); diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp index 77837adde..48a105327 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp +++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp @@ -20,12 +20,14 @@ #include "core/frontend/graphics_context.h" #include "core/telemetry_session.h" #include "video_core/gpu.h" +#include "video_core/renderer_vulkan/present/util.h" #include "video_core/renderer_vulkan/renderer_vulkan.h" #include "video_core/renderer_vulkan/vk_blit_screen.h" #include "video_core/renderer_vulkan/vk_rasterizer.h" #include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_state_tracker.h" #include "video_core/renderer_vulkan/vk_swapchain.h" +#include "video_core/textures/decoders.h" #include "video_core/vulkan_common/vulkan_debug_callback.h" #include "video_core/vulkan_common/vulkan_device.h" #include "video_core/vulkan_common/vulkan_instance.h" @@ -116,18 +118,20 @@ RendererVulkan::~RendererVulkan() { void(device.GetLogical().WaitIdle()); } -void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { - if (!framebuffer) { +void RendererVulkan::Composite(std::span framebuffers) { + if (framebuffers.empty()) { return; } + SCOPE_EXIT({ render_window.OnFrameDisplayed(); }); + if (!render_window.IsShown()) { return; } - RenderScreenshot(framebuffer); + RenderScreenshot(framebuffers); Frame* frame = present_manager.GetRenderFrame(); - blit_swapchain.DrawToFrame(rasterizer, frame, std::span(framebuffer, 1), + blit_swapchain.DrawToFrame(rasterizer, frame, framebuffers, render_window.GetFramebufferLayout(), swapchain.GetImageCount(), swapchain.GetImageViewFormat()); scheduler.Flush(*frame->render_ready); @@ -163,156 +167,37 @@ void RendererVulkan::Report() const { telemetry_session.AddField(field, "GPU_Vulkan_Extensions", extensions); } -void Vulkan::RendererVulkan::RenderScreenshot(const Tegra::FramebufferConfig* framebuffer) { +void Vulkan::RendererVulkan::RenderScreenshot( + std::span framebuffers) { if (!renderer_settings.screenshot_requested) { return; } + + constexpr VkFormat ScreenshotFormat{VK_FORMAT_B8G8R8A8_UNORM}; const Layout::FramebufferLayout layout{renderer_settings.screenshot_framebuffer_layout}; - auto frame = [&]() { - vk::Image staging_image = memory_allocator.CreateImage(VkImageCreateInfo{ - .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO, - .pNext = nullptr, - .flags = VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT, - .imageType = VK_IMAGE_TYPE_2D, - .format = VK_FORMAT_B8G8R8A8_UNORM, - .extent = - { - .width = layout.width, - .height = layout.height, - .depth = 1, - }, - .mipLevels = 1, - .arrayLayers = 1, - .samples = VK_SAMPLE_COUNT_1_BIT, - .tiling = VK_IMAGE_TILING_OPTIMAL, - .usage = VK_IMAGE_USAGE_TRANSFER_SRC_BIT | VK_IMAGE_USAGE_TRANSFER_DST_BIT | - VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT, - .sharingMode = VK_SHARING_MODE_EXCLUSIVE, - .queueFamilyIndexCount = 0, - .pQueueFamilyIndices = nullptr, - .initialLayout = VK_IMAGE_LAYOUT_UNDEFINED, - }); - vk::ImageView dst_view = device.GetLogical().CreateImageView(VkImageViewCreateInfo{ - .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO, - .pNext = nullptr, - .flags = 0, - .image = *staging_image, - .viewType = VK_IMAGE_VIEW_TYPE_2D, - .format = VK_FORMAT_B8G8R8A8_UNORM, - .components{ - .r = VK_COMPONENT_SWIZZLE_IDENTITY, - .g = VK_COMPONENT_SWIZZLE_IDENTITY, - .b = VK_COMPONENT_SWIZZLE_IDENTITY, - .a = VK_COMPONENT_SWIZZLE_IDENTITY, - }, - .subresourceRange{ - .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, - .baseMipLevel = 0, - .levelCount = 1, - .baseArrayLayer = 0, - .layerCount = VK_REMAINING_ARRAY_LAYERS, - }, - }); - vk::Framebuffer screenshot_fb = - blit_screenshot.CreateFramebuffer(layout, *dst_view, VK_FORMAT_B8G8R8A8_UNORM); - return Frame{ - .width = layout.width, - .height = layout.height, - .image = std::move(staging_image), - .image_view = std::move(dst_view), - .framebuffer = std::move(screenshot_fb), - .cmdbuf{}, - .render_ready{}, - .present_done{}, - }; + auto frame = [&]() { + Frame f{}; + f.image = CreateWrappedImage(memory_allocator, VkExtent2D{layout.width, layout.height}, + ScreenshotFormat); + f.image_view = CreateWrappedImageView(device, f.image, ScreenshotFormat); + f.framebuffer = blit_screenshot.CreateFramebuffer(layout, *f.image_view, ScreenshotFormat); + return f; }(); - blit_screenshot.DrawToFrame(rasterizer, &frame, std::span(framebuffer, 1), layout, 1, + blit_screenshot.DrawToFrame(rasterizer, &frame, framebuffers, layout, 1, VK_FORMAT_B8G8R8A8_UNORM); - const auto buffer_size = static_cast(layout.width * layout.height * 4); - const VkBufferCreateInfo dst_buffer_info{ - .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO, - .pNext = nullptr, - .flags = 0, - .size = buffer_size, - .usage = VK_BUFFER_USAGE_TRANSFER_DST_BIT, - .sharingMode = VK_SHARING_MODE_EXCLUSIVE, - .queueFamilyIndexCount = 0, - .pQueueFamilyIndices = nullptr, - }; - const vk::Buffer dst_buffer = - memory_allocator.CreateBuffer(dst_buffer_info, MemoryUsage::Download); + const auto dst_buffer = CreateWrappedBuffer( + memory_allocator, static_cast(layout.width * layout.height * 4), + MemoryUsage::Download); scheduler.RequestOutsideRenderPassOperationContext(); scheduler.Record([&](vk::CommandBuffer cmdbuf) { - const VkImageMemoryBarrier read_barrier{ - .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, - .pNext = nullptr, - .srcAccessMask = VK_ACCESS_MEMORY_WRITE_BIT, - .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT, - .oldLayout = VK_IMAGE_LAYOUT_GENERAL, - .newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, - .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, - .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, - .image = *frame.image, - .subresourceRange{ - .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, - .baseMipLevel = 0, - .levelCount = VK_REMAINING_MIP_LEVELS, - .baseArrayLayer = 0, - .layerCount = VK_REMAINING_ARRAY_LAYERS, - }, - }; - const VkImageMemoryBarrier image_write_barrier{ - .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER, - .pNext = nullptr, - .srcAccessMask = 0, - .dstAccessMask = VK_ACCESS_MEMORY_WRITE_BIT, - .oldLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, - .newLayout = VK_IMAGE_LAYOUT_GENERAL, - .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, - .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED, - .image = *frame.image, - .subresourceRange{ - .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, - .baseMipLevel = 0, - .levelCount = VK_REMAINING_MIP_LEVELS, - .baseArrayLayer = 0, - .layerCount = VK_REMAINING_ARRAY_LAYERS, - }, - }; - static constexpr VkMemoryBarrier memory_write_barrier{ - .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER, - .pNext = nullptr, - .srcAccessMask = VK_ACCESS_MEMORY_WRITE_BIT, - .dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT, - }; - const VkBufferImageCopy copy{ - .bufferOffset = 0, - .bufferRowLength = 0, - .bufferImageHeight = 0, - .imageSubresource{ - .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT, - .mipLevel = 0, - .baseArrayLayer = 0, - .layerCount = 1, - }, - .imageOffset{.x = 0, .y = 0, .z = 0}, - .imageExtent{ - .width = layout.width, - .height = layout.height, - .depth = 1, - }, - }; - cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, - 0, read_barrier); - cmdbuf.CopyImageToBuffer(*frame.image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, *dst_buffer, - copy); - cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, - 0, memory_write_barrier, nullptr, image_write_barrier); + DownloadColorImage(cmdbuf, *frame.image, *dst_buffer, + VkExtent3D{layout.width, layout.height, 1}); }); + // Ensure the copy is fully completed before saving the screenshot scheduler.Finish(); diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.h b/src/video_core/renderer_vulkan/renderer_vulkan.h index bdeb43a54..c6d8a0f21 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.h +++ b/src/video_core/renderer_vulkan/renderer_vulkan.h @@ -46,7 +46,7 @@ public: std::unique_ptr context_); ~RendererVulkan() override; - void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override; + void Composite(std::span framebuffers) override; VideoCore::RasterizerInterface* ReadRasterizer() override { return &rasterizer; @@ -59,7 +59,7 @@ public: private: void Report() const; - void RenderScreenshot(const Tegra::FramebufferConfig* framebuffer); + void RenderScreenshot(std::span framebuffers); Core::TelemetrySession& telemetry_session; Tegra::MaxwellDeviceMemoryManager& device_memory; diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.cpp b/src/video_core/renderer_vulkan/vk_blit_screen.cpp index b2dcbf80b..2275fcc46 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.cpp +++ b/src/video_core/renderer_vulkan/vk_blit_screen.cpp @@ -115,7 +115,7 @@ void BlitScreen::DrawToFrame(RasterizerVulkan& rasterizer, Frame* frame, } vk::Framebuffer BlitScreen::CreateFramebuffer(const Layout::FramebufferLayout& layout, - const VkImageView& image_view, + VkImageView image_view, VkFormat current_view_format) { const bool format_updated = std::exchange(swapchain_view_format, current_view_format) != current_view_format; diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.h b/src/video_core/renderer_vulkan/vk_blit_screen.h index 9a3476c77..cbdf2d5d0 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.h +++ b/src/video_core/renderer_vulkan/vk_blit_screen.h @@ -56,7 +56,7 @@ public: VkFormat current_swapchain_view_format); [[nodiscard]] vk::Framebuffer CreateFramebuffer(const Layout::FramebufferLayout& layout, - const VkImageView& image_view, + VkImageView image_view, VkFormat current_view_format); private: -- cgit v1.2.3