diff options
Diffstat (limited to 'src/video_core/renderer_opengl')
18 files changed, 558 insertions, 147 deletions
diff --git a/src/video_core/renderer_opengl/blit_image.cpp b/src/video_core/renderer_opengl/blit_image.cpp new file mode 100644 index 000000000..9a560a73b --- /dev/null +++ b/src/video_core/renderer_opengl/blit_image.cpp @@ -0,0 +1,59 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#include <algorithm> + +#include "video_core/host_shaders/blit_color_float_frag.h" +#include "video_core/host_shaders/full_screen_triangle_vert.h" +#include "video_core/renderer_opengl/blit_image.h" +#include "video_core/renderer_opengl/gl_shader_manager.h" +#include "video_core/renderer_opengl/gl_shader_util.h" + +namespace OpenGL { + +BlitImageHelper::BlitImageHelper(ProgramManager& program_manager_) + : program_manager(program_manager_), + full_screen_vert(CreateProgram(HostShaders::FULL_SCREEN_TRIANGLE_VERT, GL_VERTEX_SHADER)), + blit_color_to_color_frag( + CreateProgram(HostShaders::BLIT_COLOR_FLOAT_FRAG, GL_FRAGMENT_SHADER)) {} + +BlitImageHelper::~BlitImageHelper() = default; + +void BlitImageHelper::BlitColor(GLuint dst_framebuffer, GLuint src_image_view, GLuint src_sampler, + const Region2D& dst_region, const Region2D& src_region, + const Extent3D& src_size) { + 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); + + program_manager.BindPresentPrograms(full_screen_vert.handle, blit_color_to_color_frag.handle); + glProgramUniform2f(full_screen_vert.handle, 0, + static_cast<float>(src_region.end.x - src_region.start.x) / + static_cast<float>(src_size.width), + static_cast<float>(src_region.end.y - src_region.start.y) / + static_cast<float>(src_size.height)); + glProgramUniform2f(full_screen_vert.handle, 1, + static_cast<float>(src_region.start.x) / static_cast<float>(src_size.width), + static_cast<float>(src_region.start.y) / + static_cast<float>(src_size.height)); + glViewport(std::min(dst_region.start.x, dst_region.end.x), + std::min(dst_region.start.y, dst_region.end.y), + std::abs(dst_region.end.x - dst_region.start.x), + std::abs(dst_region.end.y - dst_region.start.y)); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, dst_framebuffer); + glBindSampler(0, src_sampler); + glBindTextureUnit(0, src_image_view); + glClear(GL_COLOR_BUFFER_BIT); + glDrawArrays(GL_TRIANGLES, 0, 3); +} +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/blit_image.h b/src/video_core/renderer_opengl/blit_image.h new file mode 100644 index 000000000..5a2b12d16 --- /dev/null +++ b/src/video_core/renderer_opengl/blit_image.h @@ -0,0 +1,38 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include <glad/glad.h> + +#include "video_core/engines/fermi_2d.h" +#include "video_core/renderer_opengl/gl_resource_manager.h" +#include "video_core/texture_cache/types.h" + +namespace OpenGL { + +using VideoCommon::Extent3D; +using VideoCommon::Offset2D; +using VideoCommon::Region2D; + +class ProgramManager; +class Framebuffer; +class ImageView; + +class BlitImageHelper { +public: + explicit BlitImageHelper(ProgramManager& program_manager); + ~BlitImageHelper(); + + void BlitColor(GLuint dst_framebuffer, GLuint src_image_view, GLuint src_sampler, + const Region2D& dst_region, const Region2D& src_region, + const Extent3D& src_size); + +private: + ProgramManager& program_manager; + + OGLProgram full_screen_vert; + OGLProgram blit_color_to_color_frag; +}; + +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_compute_pipeline.cpp b/src/video_core/renderer_opengl/gl_compute_pipeline.cpp index 26d066004..1a0cea9b7 100644 --- a/src/video_core/renderer_opengl/gl_compute_pipeline.cpp +++ b/src/video_core/renderer_opengl/gl_compute_pipeline.cpp @@ -30,7 +30,7 @@ bool ComputePipelineKey::operator==(const ComputePipelineKey& rhs) const noexcep ComputePipeline::ComputePipeline(const Device& device, TextureCache& texture_cache_, BufferCache& buffer_cache_, ProgramManager& program_manager_, const Shader::Info& info_, std::string code, - std::vector<u32> code_v) + std::vector<u32> code_v, bool force_context_flush) : texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, program_manager{program_manager_}, info{info_} { switch (device.GetShaderBackend()) { @@ -63,6 +63,15 @@ ComputePipeline::ComputePipeline(const Device& device, TextureCache& texture_cac writes_global_memory = !use_storage_buffers && std::ranges::any_of(info.storage_buffers_descriptors, [](const auto& desc) { return desc.is_written; }); + if (force_context_flush) { + std::scoped_lock lock{built_mutex}; + built_fence.Create(); + // Flush this context to ensure compilation commands and fence are in the GPU pipe. + glFlush(); + built_condvar.notify_one(); + } else { + is_built = true; + } } void ComputePipeline::Configure() { @@ -142,6 +151,9 @@ void ComputePipeline::Configure() { } texture_cache.FillComputeImageViews(std::span(views.data(), views.size())); + if (!is_built) { + WaitForBuild(); + } if (assembly_program.handle != 0) { program_manager.BindComputeAssemblyProgram(assembly_program.handle); } else { @@ -223,4 +235,13 @@ void ComputePipeline::Configure() { } } +void ComputePipeline::WaitForBuild() { + if (built_fence.handle == 0) { + std::unique_lock lock{built_mutex}; + built_condvar.wait(lock, [this] { return built_fence.handle != 0; }); + } + ASSERT(glClientWaitSync(built_fence.handle, 0, GL_TIMEOUT_IGNORED) != GL_WAIT_FAILED); + is_built = true; +} + } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_compute_pipeline.h b/src/video_core/renderer_opengl/gl_compute_pipeline.h index 6534dec32..9bcc72b59 100644 --- a/src/video_core/renderer_opengl/gl_compute_pipeline.h +++ b/src/video_core/renderer_opengl/gl_compute_pipeline.h @@ -50,7 +50,8 @@ class ComputePipeline { public: explicit ComputePipeline(const Device& device, TextureCache& texture_cache_, BufferCache& buffer_cache_, ProgramManager& program_manager_, - const Shader::Info& info_, std::string code, std::vector<u32> code_v); + const Shader::Info& info_, std::string code, std::vector<u32> code_v, + bool force_context_flush = false); void Configure(); @@ -65,6 +66,8 @@ public: } private: + void WaitForBuild(); + TextureCache& texture_cache; BufferCache& buffer_cache; Tegra::MemoryManager* gpu_memory; @@ -81,6 +84,11 @@ private: bool use_storage_buffers{}; bool writes_global_memory{}; + + std::mutex built_mutex; + std::condition_variable built_condvar; + OGLSync built_fence{}; + bool is_built{false}; }; } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp index cee5c3247..22ed16ebf 100644 --- a/src/video_core/renderer_opengl/gl_device.cpp +++ b/src/video_core/renderer_opengl/gl_device.cpp @@ -166,6 +166,7 @@ Device::Device(Core::Frontend::EmuWindow& emu_window) { has_shader_int64 = HasExtension(extensions, "GL_ARB_gpu_shader_int64"); has_amd_shader_half_float = GLAD_GL_AMD_gpu_shader_half_float; has_sparse_texture_2 = GLAD_GL_ARB_sparse_texture2; + has_draw_texture = GLAD_GL_NV_draw_texture; warp_size_potentially_larger_than_guest = !is_nvidia && !is_intel; need_fastmath_off = is_nvidia; can_report_memory = GLAD_GL_NVX_gpu_memory_info; diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h index 2a72d84be..3ff8cad83 100644 --- a/src/video_core/renderer_opengl/gl_device.h +++ b/src/video_core/renderer_opengl/gl_device.h @@ -4,6 +4,8 @@ #pragma once #include <cstddef> +#include <string> + #include "common/common_types.h" #include "core/frontend/emu_window.h" #include "shader_recompiler/stage.h" @@ -146,6 +148,10 @@ public: return has_sparse_texture_2; } + bool HasDrawTexture() const { + return has_draw_texture; + } + bool IsWarpSizePotentiallyLargerThanGuest() const { return warp_size_potentially_larger_than_guest; } @@ -216,6 +222,7 @@ private: bool has_shader_int64{}; bool has_amd_shader_half_float{}; bool has_sparse_texture_2{}; + bool has_draw_texture{}; bool warp_size_potentially_larger_than_guest{}; bool need_fastmath_off{}; bool has_cbuf_ftou_bug{}; diff --git a/src/video_core/renderer_opengl/gl_fsr.cpp b/src/video_core/renderer_opengl/gl_fsr.cpp new file mode 100644 index 000000000..77262dcf1 --- /dev/null +++ b/src/video_core/renderer_opengl/gl_fsr.cpp @@ -0,0 +1,101 @@ +// 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<u32, 4 * 4>; + +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<u32>& screen, + u32 input_image_width, u32 input_image_height, + const Common::Rectangle<int>& 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<u32>(fsr_tex_width) != output_image_width || + static_cast<u32>(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<GLfloat>(output_image_width), + static_cast<GLfloat>(output_image_height)); + + FsrConstants constants; + FsrEasuConOffset( + constants.data() + 0, constants.data() + 4, constants.data() + 8, constants.data() + 12, + + static_cast<f32>(crop_rect.GetWidth()), static_cast<f32>(crop_rect.GetHeight()), + static_cast<f32>(input_image_width), static_cast<f32>(input_image_height), + static_cast<f32>(output_image_width), static_cast<f32>(output_image_height), + static_cast<f32>(crop_rect.left), static_cast<f32>(crop_rect.top)); + + 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<float>(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 new file mode 100644 index 000000000..1f6ae3115 --- /dev/null +++ b/src/video_core/renderer_opengl/gl_fsr.h @@ -0,0 +1,43 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include <string_view> + +#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<u32>& screen, + u32 input_image_width, u32 input_image_height, + const Common::Rectangle<int>& 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/gl_graphics_pipeline.cpp b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp index c115dabe1..29491e762 100644 --- a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp +++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp @@ -176,7 +176,7 @@ GraphicsPipeline::GraphicsPipeline(const Device& device, TextureCache& texture_c std::array<std::string, 5> sources, std::array<std::vector<u32>, 5> sources_spirv, const std::array<const Shader::Info*, 5>& infos, - const GraphicsPipelineKey& key_) + const GraphicsPipelineKey& key_, bool force_context_flush) : texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, program_manager{program_manager_}, state_tracker{state_tracker_}, key{key_} { if (shader_notify) { @@ -231,7 +231,8 @@ GraphicsPipeline::GraphicsPipeline(const Device& device, TextureCache& texture_c const bool in_parallel = thread_worker != nullptr; const auto backend = device.GetShaderBackend(); auto func{[this, sources = std::move(sources), sources_spirv = std::move(sources_spirv), - shader_notify, backend, in_parallel](ShaderContext::Context*) mutable { + shader_notify, backend, in_parallel, + force_context_flush](ShaderContext::Context*) mutable { for (size_t stage = 0; stage < 5; ++stage) { switch (backend) { case Settings::ShaderBackend::GLSL: @@ -251,7 +252,7 @@ GraphicsPipeline::GraphicsPipeline(const Device& device, TextureCache& texture_c break; } } - if (in_parallel) { + if (force_context_flush || in_parallel) { std::scoped_lock lock{built_mutex}; built_fence.Create(); // Flush this context to ensure compilation commands and fence are in the GPU pipe. diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.h b/src/video_core/renderer_opengl/gl_graphics_pipeline.h index 1c06b3655..7bab3be0a 100644 --- a/src/video_core/renderer_opengl/gl_graphics_pipeline.h +++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.h @@ -78,7 +78,7 @@ public: std::array<std::string, 5> sources, std::array<std::vector<u32>, 5> sources_spirv, const std::array<const Shader::Info*, 5>& infos, - const GraphicsPipelineKey& key_); + const GraphicsPipelineKey& key_, bool force_context_flush = false); void Configure(bool is_indexed) { configure_func(this, is_indexed); diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 181857d9c..7bced675c 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -64,7 +64,8 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra shader_cache(*this, emu_window_, device, texture_cache, buffer_cache, program_manager, state_tracker, gpu.ShaderNotify()), query_cache(*this), accelerate_dma(buffer_cache), - fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache) {} + fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache), + blit_image(program_manager_) {} RasterizerOpenGL::~RasterizerOpenGL() = default; @@ -320,6 +321,47 @@ void RasterizerOpenGL::DrawIndirect() { buffer_cache.SetDrawIndirect(nullptr); } +void RasterizerOpenGL::DrawTexture() { + MICROPROFILE_SCOPE(OpenGL_Drawing); + + SCOPE_EXIT({ gpu.TickWork(); }); + query_cache.UpdateCounters(); + + texture_cache.SynchronizeGraphicsDescriptors(); + texture_cache.UpdateRenderTargets(false); + + SyncState(); + + const auto& draw_texture_state = maxwell3d->draw_manager->GetDrawTextureState(); + const auto& sampler = texture_cache.GetGraphicsSampler(draw_texture_state.src_sampler); + const auto& texture = texture_cache.GetImageView(draw_texture_state.src_texture); + + if (device.HasDrawTexture()) { + state_tracker.BindFramebuffer(texture_cache.GetFramebuffer()->Handle()); + + glDrawTextureNV(texture.DefaultHandle(), sampler->Handle(), draw_texture_state.dst_x0, + draw_texture_state.dst_y0, draw_texture_state.dst_x1, + draw_texture_state.dst_y1, 0, + draw_texture_state.src_x0 / static_cast<float>(texture.size.width), + draw_texture_state.src_y0 / static_cast<float>(texture.size.height), + draw_texture_state.src_x1 / static_cast<float>(texture.size.width), + draw_texture_state.src_y1 / static_cast<float>(texture.size.height)); + } else { + Region2D dst_region = {Offset2D{.x = static_cast<s32>(draw_texture_state.dst_x0), + .y = static_cast<s32>(draw_texture_state.dst_y0)}, + Offset2D{.x = static_cast<s32>(draw_texture_state.dst_x1), + .y = static_cast<s32>(draw_texture_state.dst_y1)}}; + Region2D src_region = {Offset2D{.x = static_cast<s32>(draw_texture_state.src_x0), + .y = static_cast<s32>(draw_texture_state.src_y0)}, + Offset2D{.x = static_cast<s32>(draw_texture_state.src_x1), + .y = static_cast<s32>(draw_texture_state.src_y1)}}; + blit_image.BlitColor(texture_cache.GetFramebuffer()->Handle(), texture.DefaultHandle(), + sampler->Handle(), dst_region, src_region, texture.size); + } + + ++num_queued_commands; +} + void RasterizerOpenGL::DispatchCompute() { gpu_memory->FlushCaching(); ComputePipeline* const pipeline{shader_cache.CurrentComputePipeline()}; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index be4f76c18..0c45832ae 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_accelerated.h" #include "video_core/rasterizer_interface.h" +#include "video_core/renderer_opengl/blit_image.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" @@ -70,6 +71,7 @@ public: void Draw(bool is_indexed, u32 instance_count) override; void DrawIndirect() override; + void DrawTexture() override; void Clear(u32 layer_count) override; void DispatchCompute() override; void ResetCounter(VideoCore::QueryType type) override; @@ -224,6 +226,8 @@ private: AccelerateDMA accelerate_dma; FenceManagerOpenGL fence_manager; + BlitImageHelper blit_image; + boost::container::static_vector<u32, MAX_IMAGE_VIEWS> image_view_indices; std::array<ImageViewId, MAX_IMAGE_VIEWS> image_view_ids; boost::container::static_vector<GLuint, MAX_TEXTURES> sampler_handles; diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp index 7dd854e0f..626ea7dcb 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.cpp +++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp @@ -286,7 +286,7 @@ void ShaderCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading, file.read(reinterpret_cast<char*>(&key), sizeof(key)); queue_work([this, key, env = std::move(env), &state, &callback](Context* ctx) mutable { ctx->pools.ReleaseContents(); - auto pipeline{CreateComputePipeline(ctx->pools, key, env)}; + auto pipeline{CreateComputePipeline(ctx->pools, key, env, true)}; std::scoped_lock lock{state.mutex}; if (pipeline) { compute_cache.emplace(key, std::move(pipeline)); @@ -307,7 +307,7 @@ void ShaderCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading, env_ptrs.push_back(&env); } ctx->pools.ReleaseContents(); - auto pipeline{CreateGraphicsPipeline(ctx->pools, key, MakeSpan(env_ptrs), false)}; + auto pipeline{CreateGraphicsPipeline(ctx->pools, key, MakeSpan(env_ptrs), false, true)}; std::scoped_lock lock{state.mutex}; if (pipeline) { graphics_cache.emplace(key, std::move(pipeline)); @@ -439,7 +439,8 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline() { std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline( ShaderContext::ShaderPools& pools, const GraphicsPipelineKey& key, - std::span<Shader::Environment* const> envs, bool build_in_parallel) try { + std::span<Shader::Environment* const> envs, bool use_shader_workers, + bool force_context_flush) try { LOG_INFO(Render_OpenGL, "0x{:016x}", key.Hash()); size_t env_index{}; u32 total_storage_buffers{}; @@ -531,10 +532,10 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline( } previous_program = &program; } - auto* const thread_worker{build_in_parallel ? workers.get() : nullptr}; + auto* const thread_worker{use_shader_workers ? workers.get() : nullptr}; return std::make_unique<GraphicsPipeline>(device, texture_cache, buffer_cache, program_manager, state_tracker, thread_worker, &shader_notify, sources, - sources_spirv, infos, key); + sources_spirv, infos, key, force_context_flush); } catch (Shader::Exception& exception) { LOG_ERROR(Render_OpenGL, "{}", exception.what()); @@ -559,8 +560,8 @@ std::unique_ptr<ComputePipeline> ShaderCache::CreateComputePipeline( } std::unique_ptr<ComputePipeline> ShaderCache::CreateComputePipeline( - ShaderContext::ShaderPools& pools, const ComputePipelineKey& key, - Shader::Environment& env) try { + ShaderContext::ShaderPools& pools, const ComputePipelineKey& key, Shader::Environment& env, + bool force_context_flush) try { LOG_INFO(Render_OpenGL, "0x{:016x}", key.Hash()); Shader::Maxwell::Flow::CFG cfg{env, pools.flow_block, env.StartAddress()}; @@ -589,7 +590,7 @@ std::unique_ptr<ComputePipeline> ShaderCache::CreateComputePipeline( } return std::make_unique<ComputePipeline>(device, texture_cache, buffer_cache, program_manager, - program.info, code, code_spirv); + program.info, code, code_spirv, force_context_flush); } catch (Shader::Exception& exception) { LOG_ERROR(Render_OpenGL, "{}", exception.what()); return nullptr; diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h index f82420592..6b9732fca 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.h +++ b/src/video_core/renderer_opengl/gl_shader_cache.h @@ -50,14 +50,16 @@ private: std::unique_ptr<GraphicsPipeline> CreateGraphicsPipeline( ShaderContext::ShaderPools& pools, const GraphicsPipelineKey& key, - std::span<Shader::Environment* const> envs, bool build_in_parallel); + std::span<Shader::Environment* const> envs, bool use_shader_workers, + bool force_context_flush = false); std::unique_ptr<ComputePipeline> CreateComputePipeline(const ComputePipelineKey& key, const VideoCommon::ShaderInfo* shader); std::unique_ptr<ComputePipeline> CreateComputePipeline(ShaderContext::ShaderPools& pools, const ComputePipelineKey& key, - Shader::Environment& env); + Shader::Environment& env, + bool force_context_flush = false); std::unique_ptr<ShaderWorker> CreateWorkers() const; diff --git a/src/video_core/renderer_opengl/gl_shader_manager.cpp b/src/video_core/renderer_opengl/gl_shader_manager.cpp index d9c29d8b7..98841ae65 100644 --- a/src/video_core/renderer_opengl/gl_shader_manager.cpp +++ b/src/video_core/renderer_opengl/gl_shader_manager.cpp @@ -1,2 +1,123 @@ // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project // SPDX-License-Identifier: GPL-2.0-or-later + +#include <glad/glad.h> + +#include "video_core/renderer_opengl/gl_shader_manager.h" + +namespace OpenGL { + +static constexpr std::array ASSEMBLY_PROGRAM_ENUMS{ + GL_VERTEX_PROGRAM_NV, GL_TESS_CONTROL_PROGRAM_NV, GL_TESS_EVALUATION_PROGRAM_NV, + GL_GEOMETRY_PROGRAM_NV, GL_FRAGMENT_PROGRAM_NV, +}; + +ProgramManager::ProgramManager(const Device& device) { + glCreateProgramPipelines(1, &pipeline.handle); + if (device.UseAssemblyShaders()) { + glEnable(GL_COMPUTE_PROGRAM_NV); + } +} + +void ProgramManager::BindComputeProgram(GLuint program) { + glUseProgram(program); + is_compute_bound = true; +} + +void ProgramManager::BindComputeAssemblyProgram(GLuint program) { + if (current_assembly_compute_program != program) { + current_assembly_compute_program = program; + glBindProgramARB(GL_COMPUTE_PROGRAM_NV, program); + } + UnbindPipeline(); +} + +void ProgramManager::BindSourcePrograms(std::span<const OGLProgram, NUM_STAGES> programs) { + static constexpr std::array<GLenum, 5> stage_enums{ + GL_VERTEX_SHADER_BIT, GL_TESS_CONTROL_SHADER_BIT, GL_TESS_EVALUATION_SHADER_BIT, + GL_GEOMETRY_SHADER_BIT, GL_FRAGMENT_SHADER_BIT, + }; + for (size_t stage = 0; stage < NUM_STAGES; ++stage) { + if (current_programs[stage] != programs[stage].handle) { + current_programs[stage] = programs[stage].handle; + glUseProgramStages(pipeline.handle, stage_enums[stage], programs[stage].handle); + } + } + BindPipeline(); +} + +void ProgramManager::BindPresentPrograms(GLuint vertex, GLuint fragment) { + if (current_programs[0] != vertex) { + current_programs[0] = vertex; + glUseProgramStages(pipeline.handle, GL_VERTEX_SHADER_BIT, vertex); + } + if (current_programs[4] != fragment) { + current_programs[4] = fragment; + glUseProgramStages(pipeline.handle, GL_FRAGMENT_SHADER_BIT, fragment); + } + glUseProgramStages( + pipeline.handle, + GL_TESS_CONTROL_SHADER_BIT | GL_TESS_EVALUATION_SHADER_BIT | GL_GEOMETRY_SHADER_BIT, 0); + current_programs[1] = 0; + current_programs[2] = 0; + current_programs[3] = 0; + + if (current_stage_mask != 0) { + current_stage_mask = 0; + for (const GLenum program_type : ASSEMBLY_PROGRAM_ENUMS) { + glDisable(program_type); + } + } + BindPipeline(); +} + +void ProgramManager::BindAssemblyPrograms(std::span<const OGLAssemblyProgram, NUM_STAGES> programs, + u32 stage_mask) { + const u32 changed_mask = current_stage_mask ^ stage_mask; + current_stage_mask = stage_mask; + + if (changed_mask != 0) { + for (size_t stage = 0; stage < NUM_STAGES; ++stage) { + if (((changed_mask >> stage) & 1) != 0) { + if (((stage_mask >> stage) & 1) != 0) { + glEnable(ASSEMBLY_PROGRAM_ENUMS[stage]); + } else { + glDisable(ASSEMBLY_PROGRAM_ENUMS[stage]); + } + } + } + } + for (size_t stage = 0; stage < NUM_STAGES; ++stage) { + if (current_programs[stage] != programs[stage].handle) { + current_programs[stage] = programs[stage].handle; + glBindProgramARB(ASSEMBLY_PROGRAM_ENUMS[stage], programs[stage].handle); + } + } + UnbindPipeline(); +} + +void ProgramManager::RestoreGuestCompute() {} + +void ProgramManager::BindPipeline() { + if (!is_pipeline_bound) { + is_pipeline_bound = true; + glBindProgramPipeline(pipeline.handle); + } + UnbindCompute(); +} + +void ProgramManager::UnbindPipeline() { + if (is_pipeline_bound) { + is_pipeline_bound = false; + glBindProgramPipeline(0); + } + UnbindCompute(); +} + +void ProgramManager::UnbindCompute() { + if (is_compute_bound) { + is_compute_bound = false; + glUseProgram(0); + } +} +} // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_shader_manager.h b/src/video_core/renderer_opengl/gl_shader_manager.h index a84f5aeb3..07ffab77f 100644 --- a/src/video_core/renderer_opengl/gl_shader_manager.h +++ b/src/video_core/renderer_opengl/gl_shader_manager.h @@ -6,8 +6,6 @@ #include <array> #include <span> -#include <glad/glad.h> - #include "video_core/renderer_opengl/gl_device.h" #include "video_core/renderer_opengl/gl_resource_manager.h" @@ -16,121 +14,28 @@ namespace OpenGL { class ProgramManager { static constexpr size_t NUM_STAGES = 5; - static constexpr std::array ASSEMBLY_PROGRAM_ENUMS{ - GL_VERTEX_PROGRAM_NV, GL_TESS_CONTROL_PROGRAM_NV, GL_TESS_EVALUATION_PROGRAM_NV, - GL_GEOMETRY_PROGRAM_NV, GL_FRAGMENT_PROGRAM_NV, - }; - public: - explicit ProgramManager(const Device& device) { - glCreateProgramPipelines(1, &pipeline.handle); - if (device.UseAssemblyShaders()) { - glEnable(GL_COMPUTE_PROGRAM_NV); - } - } - - void BindComputeProgram(GLuint program) { - glUseProgram(program); - is_compute_bound = true; - } - - void BindComputeAssemblyProgram(GLuint program) { - if (current_assembly_compute_program != program) { - current_assembly_compute_program = program; - glBindProgramARB(GL_COMPUTE_PROGRAM_NV, program); - } - UnbindPipeline(); - } - - void BindSourcePrograms(std::span<const OGLProgram, NUM_STAGES> programs) { - static constexpr std::array<GLenum, 5> stage_enums{ - GL_VERTEX_SHADER_BIT, GL_TESS_CONTROL_SHADER_BIT, GL_TESS_EVALUATION_SHADER_BIT, - GL_GEOMETRY_SHADER_BIT, GL_FRAGMENT_SHADER_BIT, - }; - for (size_t stage = 0; stage < NUM_STAGES; ++stage) { - if (current_programs[stage] != programs[stage].handle) { - current_programs[stage] = programs[stage].handle; - glUseProgramStages(pipeline.handle, stage_enums[stage], programs[stage].handle); - } - } - BindPipeline(); - } - - void BindPresentPrograms(GLuint vertex, GLuint fragment) { - if (current_programs[0] != vertex) { - current_programs[0] = vertex; - glUseProgramStages(pipeline.handle, GL_VERTEX_SHADER_BIT, vertex); - } - if (current_programs[4] != fragment) { - current_programs[4] = fragment; - glUseProgramStages(pipeline.handle, GL_FRAGMENT_SHADER_BIT, fragment); - } - glUseProgramStages( - pipeline.handle, - GL_TESS_CONTROL_SHADER_BIT | GL_TESS_EVALUATION_SHADER_BIT | GL_GEOMETRY_SHADER_BIT, 0); - current_programs[1] = 0; - current_programs[2] = 0; - current_programs[3] = 0; - - if (current_stage_mask != 0) { - current_stage_mask = 0; - for (const GLenum program_type : ASSEMBLY_PROGRAM_ENUMS) { - glDisable(program_type); - } - } - BindPipeline(); - } + explicit ProgramManager(const Device& device); + + void BindComputeProgram(GLuint program); + + void BindComputeAssemblyProgram(GLuint program); + + void BindSourcePrograms(std::span<const OGLProgram, NUM_STAGES> programs); + + void BindPresentPrograms(GLuint vertex, GLuint fragment); void BindAssemblyPrograms(std::span<const OGLAssemblyProgram, NUM_STAGES> programs, - u32 stage_mask) { - const u32 changed_mask = current_stage_mask ^ stage_mask; - current_stage_mask = stage_mask; - - if (changed_mask != 0) { - for (size_t stage = 0; stage < NUM_STAGES; ++stage) { - if (((changed_mask >> stage) & 1) != 0) { - if (((stage_mask >> stage) & 1) != 0) { - glEnable(ASSEMBLY_PROGRAM_ENUMS[stage]); - } else { - glDisable(ASSEMBLY_PROGRAM_ENUMS[stage]); - } - } - } - } - for (size_t stage = 0; stage < NUM_STAGES; ++stage) { - if (current_programs[stage] != programs[stage].handle) { - current_programs[stage] = programs[stage].handle; - glBindProgramARB(ASSEMBLY_PROGRAM_ENUMS[stage], programs[stage].handle); - } - } - UnbindPipeline(); - } - - void RestoreGuestCompute() {} + u32 stage_mask); + + void RestoreGuestCompute(); private: - void BindPipeline() { - if (!is_pipeline_bound) { - is_pipeline_bound = true; - glBindProgramPipeline(pipeline.handle); - } - UnbindCompute(); - } - - void UnbindPipeline() { - if (is_pipeline_bound) { - is_pipeline_bound = false; - glBindProgramPipeline(0); - } - UnbindCompute(); - } - - void UnbindCompute() { - if (is_compute_bound) { - is_compute_bound = false; - glUseProgram(0); - } - } + void BindPipeline(); + + void UnbindPipeline(); + + void UnbindCompute(); OGLPipeline pipeline; bool is_pipeline_bound{}; diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index de95f2634..2a74c1d05 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -17,8 +17,14 @@ #include "core/frontend/emu_window.h" #include "core/memory.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" @@ -31,6 +37,7 @@ #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_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" @@ -268,12 +275,17 @@ void RendererOpenGL::InitOpenGLObjects() { fxaa_vertex = CreateProgram(HostShaders::FXAA_VERT, GL_VERTEX_SHADER); fxaa_fragment = CreateProgram(HostShaders::FXAA_FRAG, GL_FRAGMENT_SHADER); - const auto SmaaShader = [](std::string_view specialized_source, GLenum stage) { - std::string shader_source{specialized_source}; - constexpr std::string_view include_string = "#include \"opengl_smaa.glsl\""; + 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(), HostShaders::OPENGL_SMAA_GLSL); + 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); }; @@ -298,14 +310,32 @@ void RendererOpenGL::InitOpenGLObjects() { 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<FSR>(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(); @@ -525,6 +555,31 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { glBindTextureUnit(0, aa_texture.handle); } + glDisablei(GL_SCISSOR_TEST, 0); + + if (Settings::values.scaling_filter.GetValue() == Settings::ScalingFilter::Fsr) { + if (!fsr->AreBuffersInitialized()) { + 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); + glBindSampler(0, present_sampler.handle); + fsr->Draw(program_manager, layout.screen, fsr_input_width, fsr_input_height, crop_rect); + } else { + if (fsr->AreBuffersInitialized()) { + fsr->ReleaseBuffers(); + } + } + const std::array ortho_matrix = MakeOrthographicMatrix(static_cast<float>(layout.width), static_cast<float>(layout.height)); @@ -540,10 +595,7 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { case Settings::ScalingFilter::ScaleForce: return present_scaleforce_fragment.handle; case Settings::ScalingFilter::Fsr: - LOG_WARNING( - Render_OpenGL, - "FidelityFX Super Resolution is not supported in OpenGL, changing to ScaleForce"); - return present_scaleforce_fragment.handle; + return fsr->GetPresentFragmentProgram().handle; default: return present_bilinear_fragment.handle; } @@ -578,15 +630,18 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { f32 scale_u = static_cast<f32>(framebuffer_width) / static_cast<f32>(screen_info.texture.width); f32 scale_v = static_cast<f32>(framebuffer_height) / static_cast<f32>(screen_info.texture.height); - // 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<f32>(framebuffer_crop_rect.GetWidth()) / - static_cast<f32>(screen_info.texture.width); - } - if (framebuffer_crop_rect.GetHeight() > 0) { - scale_v = static_cast<f32>(framebuffer_crop_rect.GetHeight()) / - static_cast<f32>(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<f32>(framebuffer_crop_rect.GetWidth()) / + static_cast<f32>(screen_info.texture.width); + } + if (framebuffer_crop_rect.GetHeight() > 0) { + scale_v = static_cast<f32>(framebuffer_crop_rect.GetHeight()) / + static_cast<f32>(screen_info.texture.height); + } } if (Settings::values.anti_aliasing.GetValue() == Settings::AntiAliasing::Fxaa && !screen_info.was_accelerated) { @@ -612,7 +667,6 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) { } else { glDisable(GL_FRAMEBUFFER_SRGB); } - glDisablei(GL_SCISSOR_TEST, 0); glViewportIndexedf(0, 0.0f, 0.0f, static_cast<GLfloat>(layout.width), static_cast<GLfloat>(layout.height)); diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h index cc97d7b26..f1d5fd954 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.h +++ b/src/video_core/renderer_opengl/renderer_opengl.h @@ -10,6 +10,7 @@ #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" @@ -141,6 +142,8 @@ private: OGLTexture smaa_edges_tex; OGLTexture smaa_blend_tex; + std::unique_ptr<FSR> fsr; + /// OpenGL framebuffer data std::vector<u8> gl_framebuffer_data; |