diff options
Diffstat (limited to 'src')
40 files changed, 365 insertions, 238 deletions
diff --git a/src/common/color.h b/src/common/color.h index 381d6332e..bbcac858e 100644 --- a/src/common/color.h +++ b/src/common/color.h @@ -10,7 +10,7 @@ #include "common/swap.h" #include "common/vector_math.h" -namespace Color { +namespace Common::Color { /// Convert a 1-bit color component to 8 bit [[nodiscard]] constexpr u8 Convert1To8(u8 value) { @@ -268,4 +268,4 @@ inline void EncodeX24S8(u8 stencil, u8* bytes) { bytes[3] = stencil; } -} // namespace Color +} // namespace Common::Color diff --git a/src/core/cpu_manager.cpp b/src/core/cpu_manager.cpp index 358943429..ef0bae556 100644 --- a/src/core/cpu_manager.cpp +++ b/src/core/cpu_manager.cpp @@ -41,9 +41,9 @@ void CpuManager::Shutdown() { running_mode = false; Pause(false); if (is_multicore) { - for (std::size_t core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) { - core_data[core].host_thread->join(); - core_data[core].host_thread.reset(); + for (auto& data : core_data) { + data.host_thread->join(); + data.host_thread.reset(); } } else { core_data[0].host_thread->join(); @@ -166,25 +166,23 @@ void CpuManager::MultiCorePause(bool paused) { bool all_not_barrier = false; while (!all_not_barrier) { all_not_barrier = true; - for (std::size_t core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) { - all_not_barrier &= - !core_data[core].is_running.load() && core_data[core].initialized.load(); + for (const auto& data : core_data) { + all_not_barrier &= !data.is_running.load() && data.initialized.load(); } } - for (std::size_t core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) { - core_data[core].enter_barrier->Set(); + for (auto& data : core_data) { + data.enter_barrier->Set(); } if (paused_state.load()) { bool all_barrier = false; while (!all_barrier) { all_barrier = true; - for (std::size_t core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) { - all_barrier &= - core_data[core].is_paused.load() && core_data[core].initialized.load(); + for (const auto& data : core_data) { + all_barrier &= data.is_paused.load() && data.initialized.load(); } } - for (std::size_t core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) { - core_data[core].exit_barrier->Set(); + for (auto& data : core_data) { + data.exit_barrier->Set(); } } } else { @@ -192,9 +190,8 @@ void CpuManager::MultiCorePause(bool paused) { bool all_barrier = false; while (!all_barrier) { all_barrier = true; - for (std::size_t core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) { - all_barrier &= - core_data[core].is_paused.load() && core_data[core].initialized.load(); + for (const auto& data : core_data) { + all_barrier &= data.is_paused.load() && data.initialized.load(); } } /// Don't release the barrier diff --git a/src/core/file_sys/registered_cache.cpp b/src/core/file_sys/registered_cache.cpp index e42b677f7..da01002d5 100644 --- a/src/core/file_sys/registered_cache.cpp +++ b/src/core/file_sys/registered_cache.cpp @@ -257,8 +257,7 @@ std::vector<NcaID> PlaceholderCache::List() const { for (const auto& sdir : dir->GetSubdirectories()) { for (const auto& file : sdir->GetFiles()) { const auto name = file->GetName(); - if (name.length() == 36 && name[32] == '.' && name[33] == 'n' && name[34] == 'c' && - name[35] == 'a') { + if (name.length() == 36 && name.ends_with(".nca")) { out.push_back(Common::HexStringToArray<0x10>(name.substr(0, 32))); } } @@ -621,25 +620,25 @@ InstallResult RegisteredCache::InstallEntry(const NSP& nsp, bool overwrite_if_ex InstallResult RegisteredCache::InstallEntry(const NCA& nca, TitleType type, bool overwrite_if_exists, const VfsCopyFunction& copy) { - CNMTHeader header{ - nca.GetTitleId(), // Title ID - 0, // Ignore/Default title version - type, // Type - {}, // Padding - 0x10, // Default table offset - 1, // 1 Content Entry - 0, // No Meta Entries - {}, // Padding - {}, // Reserved 1 - 0, // Is committed - 0, // Required download system version - {}, // Reserved 2 + const CNMTHeader header{ + .title_id = nca.GetTitleId(), + .title_version = 0, + .type = type, + .reserved = {}, + .table_offset = 0x10, + .number_content_entries = 1, + .number_meta_entries = 0, + .attributes = 0, + .reserved2 = {}, + .is_committed = 0, + .required_download_system_version = 0, + .reserved3 = {}, }; - OptionalHeader opt_header{0, 0}; + const OptionalHeader opt_header{0, 0}; ContentRecord c_rec{{}, {}, {}, GetCRTypeFromNCAType(nca.GetType()), {}}; const auto& data = nca.GetBaseFile()->ReadBytes(0x100000); mbedtls_sha256_ret(data.data(), data.size(), c_rec.hash.data(), 0); - memcpy(&c_rec.nca_id, &c_rec.hash, 16); + std::memcpy(&c_rec.nca_id, &c_rec.hash, 16); const CNMT new_cnmt(header, opt_header, {c_rec}, {}); if (!RawInstallYuzuMeta(new_cnmt)) { return InstallResult::ErrorMetaFailed; diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index 3cd896a0f..d85f1e9d1 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -1,3 +1,5 @@ +add_subdirectory(host_shaders) + add_library(video_core STATIC buffer_cache/buffer_block.h buffer_cache/buffer_cache.h @@ -244,6 +246,9 @@ create_target_directory_groups(video_core) target_link_libraries(video_core PUBLIC common core) target_link_libraries(video_core PRIVATE glad xbyak) +add_dependencies(video_core host_shaders) +target_include_directories(video_core PRIVATE ${HOST_SHADERS_INCLUDE}) + if (ENABLE_VULKAN) target_include_directories(video_core PRIVATE sirit ../../externals/Vulkan-Headers/include) target_compile_definitions(video_core PRIVATE HAS_VULKAN) diff --git a/src/video_core/engines/fermi_2d.cpp b/src/video_core/engines/fermi_2d.cpp index ff10ff40d..6e50661a3 100644 --- a/src/video_core/engines/fermi_2d.cpp +++ b/src/video_core/engines/fermi_2d.cpp @@ -10,7 +10,13 @@ namespace Tegra::Engines { -Fermi2D::Fermi2D(VideoCore::RasterizerInterface& rasterizer) : rasterizer{rasterizer} {} +Fermi2D::Fermi2D() = default; + +Fermi2D::~Fermi2D() = default; + +void Fermi2D::BindRasterizer(VideoCore::RasterizerInterface& rasterizer_) { + rasterizer = &rasterizer_; +} void Fermi2D::CallMethod(u32 method, u32 method_argument, bool is_last_call) { ASSERT_MSG(method < Regs::NUM_REGS, @@ -87,7 +93,7 @@ void Fermi2D::HandleSurfaceCopy() { copy_config.src_rect = src_rect; copy_config.dst_rect = dst_rect; - if (!rasterizer.AccelerateSurfaceCopy(regs.src, regs.dst, copy_config)) { + if (!rasterizer->AccelerateSurfaceCopy(regs.src, regs.dst, copy_config)) { UNIMPLEMENTED(); } } diff --git a/src/video_core/engines/fermi_2d.h b/src/video_core/engines/fermi_2d.h index 8f37d053f..213abfaae 100644 --- a/src/video_core/engines/fermi_2d.h +++ b/src/video_core/engines/fermi_2d.h @@ -34,8 +34,11 @@ namespace Tegra::Engines { class Fermi2D final : public EngineInterface { public: - explicit Fermi2D(VideoCore::RasterizerInterface& rasterizer); - ~Fermi2D() = default; + explicit Fermi2D(); + ~Fermi2D(); + + /// Binds a rasterizer to this engine. + void BindRasterizer(VideoCore::RasterizerInterface& rasterizer); /// Write the value to the register identified by method. void CallMethod(u32 method, u32 method_argument, bool is_last_call) override; @@ -149,7 +152,7 @@ public: }; private: - VideoCore::RasterizerInterface& rasterizer; + VideoCore::RasterizerInterface* rasterizer; /// Performs the copy from the source surface to the destination surface as configured in the /// registers. diff --git a/src/video_core/engines/kepler_compute.cpp b/src/video_core/engines/kepler_compute.cpp index a82b06a38..898370739 100644 --- a/src/video_core/engines/kepler_compute.cpp +++ b/src/video_core/engines/kepler_compute.cpp @@ -16,14 +16,15 @@ namespace Tegra::Engines { -KeplerCompute::KeplerCompute(Core::System& system, VideoCore::RasterizerInterface& rasterizer, - MemoryManager& memory_manager) - : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager}, upload_state{ - memory_manager, - regs.upload} {} +KeplerCompute::KeplerCompute(Core::System& system_, MemoryManager& memory_manager_) + : system{system_}, memory_manager{memory_manager_}, upload_state{memory_manager, regs.upload} {} KeplerCompute::~KeplerCompute() = default; +void KeplerCompute::BindRasterizer(VideoCore::RasterizerInterface& rasterizer_) { + rasterizer = &rasterizer_; +} + void KeplerCompute::CallMethod(u32 method, u32 method_argument, bool is_last_call) { ASSERT_MSG(method < Regs::NUM_REGS, "Invalid KeplerCompute register, increase the size of the Regs structure"); @@ -104,11 +105,11 @@ SamplerDescriptor KeplerCompute::AccessSampler(u32 handle) const { } VideoCore::GuestDriverProfile& KeplerCompute::AccessGuestDriverProfile() { - return rasterizer.AccessGuestDriverProfile(); + return rasterizer->AccessGuestDriverProfile(); } const VideoCore::GuestDriverProfile& KeplerCompute::AccessGuestDriverProfile() const { - return rasterizer.AccessGuestDriverProfile(); + return rasterizer->AccessGuestDriverProfile(); } void KeplerCompute::ProcessLaunch() { @@ -119,7 +120,7 @@ void KeplerCompute::ProcessLaunch() { const GPUVAddr code_addr = regs.code_loc.Address() + launch_description.program_start; LOG_TRACE(HW_GPU, "Compute invocation launched at address 0x{:016x}", code_addr); - rasterizer.DispatchCompute(code_addr); + rasterizer->DispatchCompute(code_addr); } Texture::TICEntry KeplerCompute::GetTICEntry(u32 tic_index) const { diff --git a/src/video_core/engines/kepler_compute.h b/src/video_core/engines/kepler_compute.h index b7f668d88..7f2500aab 100644 --- a/src/video_core/engines/kepler_compute.h +++ b/src/video_core/engines/kepler_compute.h @@ -42,10 +42,12 @@ namespace Tegra::Engines { class KeplerCompute final : public ConstBufferEngineInterface, public EngineInterface { public: - explicit KeplerCompute(Core::System& system, VideoCore::RasterizerInterface& rasterizer, - MemoryManager& memory_manager); + explicit KeplerCompute(Core::System& system, MemoryManager& memory_manager); ~KeplerCompute(); + /// Binds a rasterizer to this engine. + void BindRasterizer(VideoCore::RasterizerInterface& rasterizer); + static constexpr std::size_t NumConstBuffers = 8; struct Regs { @@ -230,11 +232,6 @@ public: const VideoCore::GuestDriverProfile& AccessGuestDriverProfile() const override; private: - Core::System& system; - VideoCore::RasterizerInterface& rasterizer; - MemoryManager& memory_manager; - Upload::State upload_state; - void ProcessLaunch(); /// Retrieves information about a specific TIC entry from the TIC buffer. @@ -242,6 +239,11 @@ private: /// Retrieves information about a specific TSC entry from the TSC buffer. Texture::TSCEntry GetTSCEntry(u32 tsc_index) const; + + Core::System& system; + MemoryManager& memory_manager; + VideoCore::RasterizerInterface* rasterizer = nullptr; + Upload::State upload_state; }; #define ASSERT_REG_POSITION(field_name, position) \ diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index c01436295..33854445f 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -22,14 +22,19 @@ using VideoCore::QueryType; /// First register id that is actually a Macro call. constexpr u32 MacroRegistersStart = 0xE00; -Maxwell3D::Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& rasterizer, - MemoryManager& memory_manager) - : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager}, - macro_engine{GetMacroEngine(*this)}, upload_state{memory_manager, regs.upload} { +Maxwell3D::Maxwell3D(Core::System& system_, MemoryManager& memory_manager_) + : system{system_}, memory_manager{memory_manager_}, macro_engine{GetMacroEngine(*this)}, + upload_state{memory_manager, regs.upload} { dirty.flags.flip(); InitializeRegisterDefaults(); } +Maxwell3D::~Maxwell3D() = default; + +void Maxwell3D::BindRasterizer(VideoCore::RasterizerInterface& rasterizer_) { + rasterizer = &rasterizer_; +} + void Maxwell3D::InitializeRegisterDefaults() { // Initializes registers to their default values - what games expect them to be at boot. This is // for certain registers that may not be explicitly set by games. @@ -192,7 +197,7 @@ void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) { switch (method) { case MAXWELL3D_REG_INDEX(wait_for_idle): { - rasterizer.WaitForIdle(); + rasterizer->WaitForIdle(); break; } case MAXWELL3D_REG_INDEX(shadow_ram_control): { @@ -402,7 +407,7 @@ void Maxwell3D::FlushMMEInlineDraw() { const bool is_indexed = mme_draw.current_mode == MMEDrawMode::Indexed; if (ShouldExecute()) { - rasterizer.Draw(is_indexed, true); + rasterizer->Draw(is_indexed, true); } // TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if @@ -465,7 +470,7 @@ void Maxwell3D::ProcessQueryGet() { switch (regs.query.query_get.operation) { case Regs::QueryOperation::Release: if (regs.query.query_get.fence == 1) { - rasterizer.SignalSemaphore(regs.query.QueryAddress(), regs.query.query_sequence); + rasterizer->SignalSemaphore(regs.query.QueryAddress(), regs.query.query_sequence); } else { StampQueryResult(regs.query.query_sequence, regs.query.query_get.short_query == 0); } @@ -533,7 +538,7 @@ void Maxwell3D::ProcessQueryCondition() { void Maxwell3D::ProcessCounterReset() { switch (regs.counter_reset) { case Regs::CounterReset::SampleCnt: - rasterizer.ResetCounter(QueryType::SamplesPassed); + rasterizer->ResetCounter(QueryType::SamplesPassed); break; default: LOG_DEBUG(Render_OpenGL, "Unimplemented counter reset={}", @@ -547,7 +552,7 @@ void Maxwell3D::ProcessSyncPoint() { const u32 increment = regs.sync_info.increment.Value(); [[maybe_unused]] const u32 cache_flush = regs.sync_info.unknown.Value(); if (increment) { - rasterizer.SignalSyncPoint(sync_point); + rasterizer->SignalSyncPoint(sync_point); } } @@ -570,7 +575,7 @@ void Maxwell3D::DrawArrays() { const bool is_indexed{regs.index_array.count && !regs.vertex_buffer.count}; if (ShouldExecute()) { - rasterizer.Draw(is_indexed, false); + rasterizer->Draw(is_indexed, false); } // TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if @@ -590,8 +595,8 @@ std::optional<u64> Maxwell3D::GetQueryResult() { return 0; case Regs::QuerySelect::SamplesPassed: // Deferred. - rasterizer.Query(regs.query.QueryAddress(), VideoCore::QueryType::SamplesPassed, - system.GPU().GetTicks()); + rasterizer->Query(regs.query.QueryAddress(), VideoCore::QueryType::SamplesPassed, + system.GPU().GetTicks()); return {}; default: LOG_DEBUG(HW_GPU, "Unimplemented query select type {}", @@ -718,7 +723,7 @@ void Maxwell3D::ProcessClearBuffers() { regs.clear_buffers.R == regs.clear_buffers.B && regs.clear_buffers.R == regs.clear_buffers.A); - rasterizer.Clear(); + rasterizer->Clear(); } u32 Maxwell3D::AccessConstBuffer32(ShaderType stage, u64 const_buffer, u64 offset) const { @@ -752,11 +757,11 @@ SamplerDescriptor Maxwell3D::AccessSampler(u32 handle) const { } VideoCore::GuestDriverProfile& Maxwell3D::AccessGuestDriverProfile() { - return rasterizer.AccessGuestDriverProfile(); + return rasterizer->AccessGuestDriverProfile(); } const VideoCore::GuestDriverProfile& Maxwell3D::AccessGuestDriverProfile() const { - return rasterizer.AccessGuestDriverProfile(); + return rasterizer->AccessGuestDriverProfile(); } } // namespace Tegra::Engines diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h index c97eeb792..bc289c55d 100644 --- a/src/video_core/engines/maxwell_3d.h +++ b/src/video_core/engines/maxwell_3d.h @@ -51,9 +51,11 @@ namespace Tegra::Engines { class Maxwell3D final : public ConstBufferEngineInterface, public EngineInterface { public: - explicit Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& rasterizer, - MemoryManager& memory_manager); - ~Maxwell3D() = default; + explicit Maxwell3D(Core::System& system, MemoryManager& memory_manager); + ~Maxwell3D(); + + /// Binds a rasterizer to this engine. + void BindRasterizer(VideoCore::RasterizerInterface& rasterizer); /// Register structure of the Maxwell3D engine. /// TODO(Subv): This structure will need to be made bigger as more registers are discovered. @@ -1418,12 +1420,12 @@ public: return execute_on; } - VideoCore::RasterizerInterface& GetRasterizer() { - return rasterizer; + VideoCore::RasterizerInterface& Rasterizer() { + return *rasterizer; } - const VideoCore::RasterizerInterface& GetRasterizer() const { - return rasterizer; + const VideoCore::RasterizerInterface& Rasterizer() const { + return *rasterizer; } /// Notify a memory write has happened. @@ -1460,11 +1462,10 @@ private: void InitializeRegisterDefaults(); Core::System& system; - - VideoCore::RasterizerInterface& rasterizer; - MemoryManager& memory_manager; + VideoCore::RasterizerInterface* rasterizer = nullptr; + /// Start offsets of each macro in macro_memory std::array<u32, 0x80> macro_positions = {}; diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index 512578c8b..acb6e6d46 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -27,21 +27,28 @@ namespace Tegra { MICROPROFILE_DEFINE(GPU_wait, "GPU", "Wait for the GPU", MP_RGB(128, 128, 192)); -GPU::GPU(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer_, bool is_async) - : system{system}, renderer{std::move(renderer_)}, is_async{is_async} { - auto& rasterizer{renderer->Rasterizer()}; - memory_manager = std::make_unique<Tegra::MemoryManager>(system, rasterizer); - dma_pusher = std::make_unique<Tegra::DmaPusher>(system, *this); - maxwell_3d = std::make_unique<Engines::Maxwell3D>(system, rasterizer, *memory_manager); - fermi_2d = std::make_unique<Engines::Fermi2D>(rasterizer); - kepler_compute = std::make_unique<Engines::KeplerCompute>(system, rasterizer, *memory_manager); - maxwell_dma = std::make_unique<Engines::MaxwellDMA>(system, *memory_manager); - kepler_memory = std::make_unique<Engines::KeplerMemory>(system, *memory_manager); - shader_notify = std::make_unique<VideoCore::ShaderNotify>(); -} +GPU::GPU(Core::System& system_, bool is_async_) + : system{system_}, dma_pusher{std::make_unique<Tegra::DmaPusher>(system, *this)}, + memory_manager{std::make_unique<Tegra::MemoryManager>(system)}, + maxwell_3d{std::make_unique<Engines::Maxwell3D>(system, *memory_manager)}, + fermi_2d{std::make_unique<Engines::Fermi2D>()}, + kepler_compute{std::make_unique<Engines::KeplerCompute>(system, *memory_manager)}, + maxwell_dma{std::make_unique<Engines::MaxwellDMA>(system, *memory_manager)}, + kepler_memory{std::make_unique<Engines::KeplerMemory>(system, *memory_manager)}, + shader_notify{std::make_unique<VideoCore::ShaderNotify>()}, is_async{is_async_} {} GPU::~GPU() = default; +void GPU::BindRenderer(std::unique_ptr<VideoCore::RendererBase> renderer_) { + renderer = std::move(renderer_); + + VideoCore::RasterizerInterface& rasterizer = renderer->Rasterizer(); + memory_manager->BindRasterizer(rasterizer); + maxwell_3d->BindRasterizer(rasterizer); + fermi_2d->BindRasterizer(rasterizer); + kepler_compute->BindRasterizer(rasterizer); +} + Engines::Maxwell3D& GPU::Maxwell3D() { return *maxwell_3d; } diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h index ebfc7b0c7..c7d11deb2 100644 --- a/src/video_core/gpu.h +++ b/src/video_core/gpu.h @@ -142,11 +142,6 @@ class MemoryManager; class GPU { public: - explicit GPU(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer, - bool is_async); - - virtual ~GPU(); - struct MethodCall { u32 method{}; u32 argument{}; @@ -162,6 +157,12 @@ public: method_count(method_count) {} }; + explicit GPU(Core::System& system, bool is_async); + virtual ~GPU(); + + /// Binds a renderer to the GPU. + void BindRenderer(std::unique_ptr<VideoCore::RendererBase> renderer); + /// Calls a GPU method. void CallMethod(const MethodCall& method_call); @@ -345,8 +346,8 @@ private: bool ExecuteMethodOnEngine(u32 method); protected: - std::unique_ptr<Tegra::DmaPusher> dma_pusher; Core::System& system; + std::unique_ptr<Tegra::DmaPusher> dma_pusher; std::unique_ptr<VideoCore::RendererBase> renderer; private: diff --git a/src/video_core/gpu_asynch.cpp b/src/video_core/gpu_asynch.cpp index 7b855f63e..70a3d5738 100644 --- a/src/video_core/gpu_asynch.cpp +++ b/src/video_core/gpu_asynch.cpp @@ -10,16 +10,14 @@ namespace VideoCommon { -GPUAsynch::GPUAsynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer_, - std::unique_ptr<Core::Frontend::GraphicsContext>&& context) - : GPU(system, std::move(renderer_), true), gpu_thread{system}, - cpu_context(renderer->GetRenderWindow().CreateSharedContext()), - gpu_context(std::move(context)) {} +GPUAsynch::GPUAsynch(Core::System& system) : GPU{system, true}, gpu_thread{system} {} GPUAsynch::~GPUAsynch() = default; void GPUAsynch::Start() { - gpu_thread.StartThread(*renderer, *gpu_context, *dma_pusher); + gpu_thread.StartThread(*renderer, renderer->Context(), *dma_pusher); + cpu_context = renderer->GetRenderWindow().CreateSharedContext(); + cpu_context->MakeCurrent(); } void GPUAsynch::ObtainContext() { diff --git a/src/video_core/gpu_asynch.h b/src/video_core/gpu_asynch.h index 15e9f1d38..f89c855a5 100644 --- a/src/video_core/gpu_asynch.h +++ b/src/video_core/gpu_asynch.h @@ -20,8 +20,7 @@ namespace VideoCommon { /// Implementation of GPU interface that runs the GPU asynchronously class GPUAsynch final : public Tegra::GPU { public: - explicit GPUAsynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer, - std::unique_ptr<Core::Frontend::GraphicsContext>&& context); + explicit GPUAsynch(Core::System& system); ~GPUAsynch() override; void Start() override; @@ -42,7 +41,6 @@ protected: private: GPUThread::ThreadManager gpu_thread; std::unique_ptr<Core::Frontend::GraphicsContext> cpu_context; - std::unique_ptr<Core::Frontend::GraphicsContext> gpu_context; }; } // namespace VideoCommon diff --git a/src/video_core/gpu_synch.cpp b/src/video_core/gpu_synch.cpp index aaeb9811d..1ca47ddef 100644 --- a/src/video_core/gpu_synch.cpp +++ b/src/video_core/gpu_synch.cpp @@ -7,20 +7,18 @@ namespace VideoCommon { -GPUSynch::GPUSynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer, - std::unique_ptr<Core::Frontend::GraphicsContext>&& context) - : GPU(system, std::move(renderer), false), context{std::move(context)} {} +GPUSynch::GPUSynch(Core::System& system) : GPU{system, false} {} GPUSynch::~GPUSynch() = default; void GPUSynch::Start() {} void GPUSynch::ObtainContext() { - context->MakeCurrent(); + renderer->Context().MakeCurrent(); } void GPUSynch::ReleaseContext() { - context->DoneCurrent(); + renderer->Context().DoneCurrent(); } void GPUSynch::PushGPUEntries(Tegra::CommandList&& entries) { diff --git a/src/video_core/gpu_synch.h b/src/video_core/gpu_synch.h index 762c20aa5..297258cb1 100644 --- a/src/video_core/gpu_synch.h +++ b/src/video_core/gpu_synch.h @@ -19,8 +19,7 @@ namespace VideoCommon { /// Implementation of GPU interface that runs the GPU synchronously class GPUSynch final : public Tegra::GPU { public: - explicit GPUSynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase>&& renderer, - std::unique_ptr<Core::Frontend::GraphicsContext>&& context); + explicit GPUSynch(Core::System& system); ~GPUSynch() override; void Start() override; @@ -36,9 +35,6 @@ public: protected: void TriggerCpuInterrupt([[maybe_unused]] u32 syncpoint_id, [[maybe_unused]] u32 value) const override {} - -private: - std::unique_ptr<Core::Frontend::GraphicsContext> context; }; } // namespace VideoCommon diff --git a/src/video_core/host_shaders/CMakeLists.txt b/src/video_core/host_shaders/CMakeLists.txt new file mode 100644 index 000000000..aa62363a7 --- /dev/null +++ b/src/video_core/host_shaders/CMakeLists.txt @@ -0,0 +1,43 @@ +set(SHADER_FILES + opengl_present.frag + opengl_present.vert +) + +set(SHADER_INCLUDE ${CMAKE_CURRENT_BINARY_DIR}/include) +set(HOST_SHADERS_INCLUDE ${SHADER_INCLUDE} PARENT_SCOPE) + +set(SHADER_DIR ${SHADER_INCLUDE}/video_core/host_shaders) +add_custom_command( + OUTPUT + ${SHADER_DIR} + COMMAND + ${CMAKE_COMMAND} -E make_directory ${SHADER_DIR} +) + +set(INPUT_FILE ${CMAKE_CURRENT_SOURCE_DIR}/source_shader.h.in) +set(HEADER_GENERATOR ${CMAKE_CURRENT_SOURCE_DIR}/StringShaderHeader.cmake) + +foreach(FILENAME IN ITEMS ${SHADER_FILES}) + string(REPLACE "." "_" SHADER_NAME ${FILENAME}) + set(SOURCE_FILE ${CMAKE_CURRENT_SOURCE_DIR}/${FILENAME}) + set(HEADER_FILE ${SHADER_DIR}/${SHADER_NAME}.h) + add_custom_command( + OUTPUT + ${HEADER_FILE} + COMMAND + ${CMAKE_COMMAND} -P ${HEADER_GENERATOR} ${SOURCE_FILE} ${HEADER_FILE} ${INPUT_FILE} + MAIN_DEPENDENCY + ${SOURCE_FILE} + DEPENDS + ${HEADER_GENERATOR} + ${INPUT_FILE} + ) + set(SHADER_HEADERS ${SHADER_HEADERS} ${HEADER_FILE}) +endforeach() + +add_custom_target(host_shaders + DEPENDS + ${SHADER_HEADERS} + SOURCES + ${SHADER_FILES} +) diff --git a/src/video_core/host_shaders/StringShaderHeader.cmake b/src/video_core/host_shaders/StringShaderHeader.cmake new file mode 100644 index 000000000..368bce0ed --- /dev/null +++ b/src/video_core/host_shaders/StringShaderHeader.cmake @@ -0,0 +1,11 @@ +set(SOURCE_FILE ${CMAKE_ARGV3}) +set(HEADER_FILE ${CMAKE_ARGV4}) +set(INPUT_FILE ${CMAKE_ARGV5}) + +get_filename_component(CONTENTS_NAME ${SOURCE_FILE} NAME) +string(REPLACE "." "_" CONTENTS_NAME ${CONTENTS_NAME}) +string(TOUPPER ${CONTENTS_NAME} CONTENTS_NAME) + +file(READ ${SOURCE_FILE} CONTENTS) + +configure_file(${INPUT_FILE} ${HEADER_FILE} @ONLY) diff --git a/src/video_core/host_shaders/opengl_present.frag b/src/video_core/host_shaders/opengl_present.frag new file mode 100644 index 000000000..8a4cb024b --- /dev/null +++ b/src/video_core/host_shaders/opengl_present.frag @@ -0,0 +1,10 @@ +#version 430 core + +layout (location = 0) in vec2 frag_tex_coord; +layout (location = 0) out vec4 color; + +layout (binding = 0) uniform sampler2D color_texture; + +void main() { + color = vec4(texture(color_texture, frag_tex_coord).rgb, 1.0f); +} diff --git a/src/video_core/host_shaders/opengl_present.vert b/src/video_core/host_shaders/opengl_present.vert new file mode 100644 index 000000000..2235d31a4 --- /dev/null +++ b/src/video_core/host_shaders/opengl_present.vert @@ -0,0 +1,24 @@ +#version 430 core + +out gl_PerVertex { + vec4 gl_Position; +}; + +layout (location = 0) in vec2 vert_position; +layout (location = 1) in vec2 vert_tex_coord; +layout (location = 0) out vec2 frag_tex_coord; + +// This is a truncated 3x3 matrix for 2D transformations: +// The upper-left 2x2 submatrix performs scaling/rotation/mirroring. +// The third column performs translation. +// The third row could be used for projection, which we don't need in 2D. It hence is assumed to +// implicitly be [0, 0, 1] +layout (location = 0) uniform mat3x2 modelview_matrix; + +void main() { + // Multiply input position by the rotscale part of the matrix and then manually translate by + // the last column. This is equivalent to using a full 3x3 matrix and expanding the vector + // to `vec3(vert_position.xy, 1.0)` + gl_Position = vec4(mat2(modelview_matrix) * vert_position + modelview_matrix[2], 0.0, 1.0); + frag_tex_coord = vert_tex_coord; +} diff --git a/src/video_core/host_shaders/source_shader.h.in b/src/video_core/host_shaders/source_shader.h.in new file mode 100644 index 000000000..ccdb0d2a9 --- /dev/null +++ b/src/video_core/host_shaders/source_shader.h.in @@ -0,0 +1,9 @@ +#pragma once + +#include <string_view> + +namespace HostShaders { + +constexpr std::string_view @CONTENTS_NAME@ = R"(@CONTENTS@)"; + +} // namespace HostShaders diff --git a/src/video_core/macro/macro_hle.cpp b/src/video_core/macro/macro_hle.cpp index 0c9ff59a4..df00b57df 100644 --- a/src/video_core/macro/macro_hle.cpp +++ b/src/video_core/macro/macro_hle.cpp @@ -24,7 +24,7 @@ void HLE_771BB18C62444DA0(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& maxwell3d.regs.index_array.first = parameters[4]; if (maxwell3d.ShouldExecute()) { - maxwell3d.GetRasterizer().Draw(true, true); + maxwell3d.Rasterizer().Draw(true, true); } maxwell3d.regs.index_array.count = 0; maxwell3d.mme_draw.instance_count = 0; @@ -42,7 +42,7 @@ void HLE_0D61FC9FAAC9FCAD(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& maxwell3d.mme_draw.instance_count = count; if (maxwell3d.ShouldExecute()) { - maxwell3d.GetRasterizer().Draw(false, true); + maxwell3d.Rasterizer().Draw(false, true); } maxwell3d.regs.vertex_buffer.count = 0; maxwell3d.mme_draw.instance_count = 0; @@ -65,7 +65,7 @@ void HLE_0217920100488FF7(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& maxwell3d.regs.draw.topology.Assign( static_cast<Tegra::Engines::Maxwell3D::Regs::PrimitiveTopology>(parameters[0])); if (maxwell3d.ShouldExecute()) { - maxwell3d.GetRasterizer().Draw(true, true); + maxwell3d.Rasterizer().Draw(true, true); } maxwell3d.regs.reg_array[0x446] = 0x0; // vertex id base? maxwell3d.regs.index_array.count = 0; diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp index 844164645..16b2aaa27 100644 --- a/src/video_core/memory_manager.cpp +++ b/src/video_core/memory_manager.cpp @@ -14,11 +14,15 @@ namespace Tegra { -MemoryManager::MemoryManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer) - : system{system}, rasterizer{rasterizer}, page_table(page_table_size) {} +MemoryManager::MemoryManager(Core::System& system_) + : system{system_}, page_table(page_table_size) {} MemoryManager::~MemoryManager() = default; +void MemoryManager::BindRasterizer(VideoCore::RasterizerInterface& rasterizer_) { + rasterizer = &rasterizer_; +} + GPUVAddr MemoryManager::UpdateRange(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size) { u64 remaining_size{size}; for (u64 offset{}; offset < size; offset += page_size) { @@ -217,7 +221,7 @@ void MemoryManager::ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, std::siz // Flush must happen on the rasterizer interface, such that memory is always synchronous // when it is read (even when in asynchronous GPU mode). Fixes Dead Cells title menu. - rasterizer.FlushRegion(src_addr, copy_amount); + rasterizer->FlushRegion(src_addr, copy_amount); system.Memory().ReadBlockUnsafe(src_addr, dest_buffer, copy_amount); } @@ -266,7 +270,7 @@ void MemoryManager::WriteBlock(GPUVAddr gpu_dest_addr, const void* src_buffer, s // Invalidate must happen on the rasterizer interface, such that memory is always // synchronous when it is written (even when in asynchronous GPU mode). - rasterizer.InvalidateRegion(dest_addr, copy_amount); + rasterizer->InvalidateRegion(dest_addr, copy_amount); system.Memory().WriteBlockUnsafe(dest_addr, src_buffer, copy_amount); } @@ -312,10 +316,10 @@ void MemoryManager::CopyBlockUnsafe(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_add WriteBlockUnsafe(gpu_dest_addr, tmp_buffer.data(), size); } -bool MemoryManager::IsGranularRange(GPUVAddr gpu_addr, std::size_t size) { +bool MemoryManager::IsGranularRange(GPUVAddr gpu_addr, std::size_t size) const { const auto cpu_addr{GpuToCpuAddress(gpu_addr)}; if (!cpu_addr) { - return {}; + return false; } const std::size_t page{(*cpu_addr & Core::Memory::PAGE_MASK) + size}; return page <= Core::Memory::PAGE_SIZE; diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h index 681bd9588..53c8d122a 100644 --- a/src/video_core/memory_manager.h +++ b/src/video_core/memory_manager.h @@ -31,19 +31,19 @@ public: constexpr PageEntry(State state) : state{state} {} constexpr PageEntry(VAddr addr) : state{static_cast<State>(addr >> ShiftBits)} {} - constexpr bool IsUnmapped() const { + [[nodiscard]] constexpr bool IsUnmapped() const { return state == State::Unmapped; } - constexpr bool IsAllocated() const { + [[nodiscard]] constexpr bool IsAllocated() const { return state == State::Allocated; } - constexpr bool IsValid() const { + [[nodiscard]] constexpr bool IsValid() const { return !IsUnmapped() && !IsAllocated(); } - constexpr VAddr ToAddress() const { + [[nodiscard]] constexpr VAddr ToAddress() const { if (!IsValid()) { return {}; } @@ -51,7 +51,7 @@ public: return static_cast<VAddr>(state) << ShiftBits; } - constexpr PageEntry operator+(u64 offset) { + [[nodiscard]] constexpr PageEntry operator+(u64 offset) const { // If this is a reserved value, offsets do not apply if (!IsValid()) { return *this; @@ -68,19 +68,22 @@ static_assert(sizeof(PageEntry) == 4, "PageEntry is too large"); class MemoryManager final { public: - explicit MemoryManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer); + explicit MemoryManager(Core::System& system); ~MemoryManager(); - std::optional<VAddr> GpuToCpuAddress(GPUVAddr addr) const; + /// Binds a renderer to the memory manager. + void BindRasterizer(VideoCore::RasterizerInterface& rasterizer); + + [[nodiscard]] std::optional<VAddr> GpuToCpuAddress(GPUVAddr addr) const; template <typename T> - T Read(GPUVAddr addr) const; + [[nodiscard]] T Read(GPUVAddr addr) const; template <typename T> void Write(GPUVAddr addr, T data); - u8* GetPointer(GPUVAddr addr); - const u8* GetPointer(GPUVAddr addr) const; + [[nodiscard]] u8* GetPointer(GPUVAddr addr); + [[nodiscard]] const u8* GetPointer(GPUVAddr addr) const; /** * ReadBlock and WriteBlock are full read and write operations over virtual @@ -109,24 +112,24 @@ public: /** * IsGranularRange checks if a gpu region can be simply read with a pointer. */ - bool IsGranularRange(GPUVAddr gpu_addr, std::size_t size); + [[nodiscard]] bool IsGranularRange(GPUVAddr gpu_addr, std::size_t size) const; - GPUVAddr Map(VAddr cpu_addr, GPUVAddr gpu_addr, std::size_t size); - GPUVAddr MapAllocate(VAddr cpu_addr, std::size_t size, std::size_t align); - std::optional<GPUVAddr> AllocateFixed(GPUVAddr gpu_addr, std::size_t size); - GPUVAddr Allocate(std::size_t size, std::size_t align); + [[nodiscard]] GPUVAddr Map(VAddr cpu_addr, GPUVAddr gpu_addr, std::size_t size); + [[nodiscard]] GPUVAddr MapAllocate(VAddr cpu_addr, std::size_t size, std::size_t align); + [[nodiscard]] std::optional<GPUVAddr> AllocateFixed(GPUVAddr gpu_addr, std::size_t size); + [[nodiscard]] GPUVAddr Allocate(std::size_t size, std::size_t align); void Unmap(GPUVAddr gpu_addr, std::size_t size); private: - PageEntry GetPageEntry(GPUVAddr gpu_addr) const; + [[nodiscard]] PageEntry GetPageEntry(GPUVAddr gpu_addr) const; void SetPageEntry(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size = page_size); GPUVAddr UpdateRange(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size); - std::optional<GPUVAddr> FindFreeRange(std::size_t size, std::size_t align) const; + [[nodiscard]] std::optional<GPUVAddr> FindFreeRange(std::size_t size, std::size_t align) const; void TryLockPage(PageEntry page_entry, std::size_t size); void TryUnlockPage(PageEntry page_entry, std::size_t size); - static constexpr std::size_t PageEntryIndex(GPUVAddr gpu_addr) { + [[nodiscard]] static constexpr std::size_t PageEntryIndex(GPUVAddr gpu_addr) { return (gpu_addr >> page_bits) & page_table_mask; } @@ -141,7 +144,7 @@ private: Core::System& system; - VideoCore::RasterizerInterface& rasterizer; + VideoCore::RasterizerInterface* rasterizer = nullptr; std::vector<PageEntry> page_table; }; diff --git a/src/video_core/renderer_base.cpp b/src/video_core/renderer_base.cpp index dfb06e87e..a93a1732c 100644 --- a/src/video_core/renderer_base.cpp +++ b/src/video_core/renderer_base.cpp @@ -9,7 +9,9 @@ namespace VideoCore { -RendererBase::RendererBase(Core::Frontend::EmuWindow& window) : render_window{window} { +RendererBase::RendererBase(Core::Frontend::EmuWindow& window_, + std::unique_ptr<Core::Frontend::GraphicsContext> context_) + : render_window{window_}, context{std::move(context_)} { RefreshBaseSettings(); } diff --git a/src/video_core/renderer_base.h b/src/video_core/renderer_base.h index 1d85219b6..649074acd 100644 --- a/src/video_core/renderer_base.h +++ b/src/video_core/renderer_base.h @@ -15,7 +15,8 @@ namespace Core::Frontend { class EmuWindow; -} +class GraphicsContext; +} // namespace Core::Frontend namespace VideoCore { @@ -25,14 +26,15 @@ struct RendererSettings { // Screenshot std::atomic<bool> screenshot_requested{false}; - void* screenshot_bits; + void* screenshot_bits{}; std::function<void()> screenshot_complete_callback; Layout::FramebufferLayout screenshot_framebuffer_layout; }; class RendererBase : NonCopyable { public: - explicit RendererBase(Core::Frontend::EmuWindow& window); + explicit RendererBase(Core::Frontend::EmuWindow& window, + std::unique_ptr<Core::Frontend::GraphicsContext> context); virtual ~RendererBase(); /// Initialize the renderer @@ -68,6 +70,14 @@ public: return *rasterizer; } + Core::Frontend::GraphicsContext& Context() { + return *context; + } + + const Core::Frontend::GraphicsContext& Context() const { + return *context; + } + Core::Frontend::EmuWindow& GetRenderWindow() { return render_window; } @@ -94,6 +104,7 @@ public: protected: Core::Frontend::EmuWindow& render_window; ///< Reference to the render window handle. std::unique_ptr<RasterizerInterface> rasterizer; + std::unique_ptr<Core::Frontend::GraphicsContext> context; f32 m_current_fps = 0.0f; ///< Current framerate, should be set by the renderer int m_current_frame = 0; ///< Current frame, should be set by the renderer diff --git a/src/video_core/renderer_opengl/gl_resource_manager.cpp b/src/video_core/renderer_opengl/gl_resource_manager.cpp index a787e27d2..0ebcec427 100644 --- a/src/video_core/renderer_opengl/gl_resource_manager.cpp +++ b/src/video_core/renderer_opengl/gl_resource_manager.cpp @@ -2,6 +2,7 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include <string_view> #include <utility> #include <glad/glad.h> #include "common/common_types.h" @@ -82,11 +83,13 @@ void OGLSampler::Release() { handle = 0; } -void OGLShader::Create(const char* source, GLenum type) { - if (handle != 0) +void OGLShader::Create(std::string_view source, GLenum type) { + if (handle != 0) { return; - if (source == nullptr) + } + if (source.empty()) { return; + } MICROPROFILE_SCOPE(OpenGL_ResourceCreation); handle = GLShader::LoadShader(source, type); diff --git a/src/video_core/renderer_opengl/gl_resource_manager.h b/src/video_core/renderer_opengl/gl_resource_manager.h index b05cb641c..f48398669 100644 --- a/src/video_core/renderer_opengl/gl_resource_manager.h +++ b/src/video_core/renderer_opengl/gl_resource_manager.h @@ -4,6 +4,7 @@ #pragma once +#include <string_view> #include <utility> #include <glad/glad.h> #include "common/common_types.h" @@ -127,7 +128,7 @@ public: return *this; } - void Create(const char* source, GLenum type); + void Create(std::string_view source, GLenum type); void Release(); diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp index eb49a36bf..a07d56ef0 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.cpp +++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp @@ -22,6 +22,7 @@ #include "video_core/memory_manager.h" #include "video_core/renderer_opengl/gl_arb_decompiler.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_cache.h" #include "video_core/renderer_opengl/gl_shader_decompiler.h" #include "video_core/renderer_opengl/gl_shader_disk_cache.h" diff --git a/src/video_core/renderer_opengl/gl_shader_util.cpp b/src/video_core/renderer_opengl/gl_shader_util.cpp index 9e74eda0d..4bf0d6090 100644 --- a/src/video_core/renderer_opengl/gl_shader_util.cpp +++ b/src/video_core/renderer_opengl/gl_shader_util.cpp @@ -2,6 +2,7 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include <string_view> #include <vector> #include <glad/glad.h> #include "common/assert.h" @@ -11,7 +12,8 @@ namespace OpenGL::GLShader { namespace { -const char* GetStageDebugName(GLenum type) { + +std::string_view StageDebugName(GLenum type) { switch (type) { case GL_VERTEX_SHADER: return "vertex"; @@ -25,12 +27,17 @@ const char* GetStageDebugName(GLenum type) { UNIMPLEMENTED(); return "unknown"; } + } // Anonymous namespace -GLuint LoadShader(const char* source, GLenum type) { - const char* debug_type = GetStageDebugName(type); +GLuint LoadShader(std::string_view source, GLenum type) { + const std::string_view debug_type = StageDebugName(type); const GLuint shader_id = glCreateShader(type); - glShaderSource(shader_id, 1, &source, nullptr); + + const GLchar* source_string = source.data(); + const GLint source_length = static_cast<GLint>(source.size()); + + glShaderSource(shader_id, 1, &source_string, &source_length); LOG_DEBUG(Render_OpenGL, "Compiling {} shader...", debug_type); glCompileShader(shader_id); diff --git a/src/video_core/renderer_opengl/gl_shader_util.h b/src/video_core/renderer_opengl/gl_shader_util.h index 03b7548c2..1b770532e 100644 --- a/src/video_core/renderer_opengl/gl_shader_util.h +++ b/src/video_core/renderer_opengl/gl_shader_util.h @@ -38,7 +38,7 @@ void LogShaderSource(T... shaders) { * @param source String of the GLSL shader program * @param type Type of the shader (GL_VERTEX_SHADER, GL_GEOMETRY_SHADER or GL_FRAGMENT_SHADER) */ -GLuint LoadShader(const char* source, GLenum type); +GLuint LoadShader(std::string_view source, GLenum type); /** * Utility function to create and compile an OpenGL GLSL shader program (vertex + fragment shader) diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index 14bbc3a1c..b759c2dba 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -21,6 +21,8 @@ #include "core/perf_stats.h" #include "core/settings.h" #include "core/telemetry_session.h" +#include "video_core/host_shaders/opengl_present_frag.h" +#include "video_core/host_shaders/opengl_present_vert.h" #include "video_core/morton.h" #include "video_core/renderer_opengl/gl_rasterizer.h" #include "video_core/renderer_opengl/gl_shader_manager.h" @@ -44,46 +46,6 @@ struct Frame { bool is_srgb{}; /// Framebuffer is sRGB or RGB }; -constexpr char VERTEX_SHADER[] = R"( -#version 430 core - -out gl_PerVertex { - vec4 gl_Position; -}; - -layout (location = 0) in vec2 vert_position; -layout (location = 1) in vec2 vert_tex_coord; -layout (location = 0) out vec2 frag_tex_coord; - -// This is a truncated 3x3 matrix for 2D transformations: -// The upper-left 2x2 submatrix performs scaling/rotation/mirroring. -// The third column performs translation. -// The third row could be used for projection, which we don't need in 2D. It hence is assumed to -// implicitly be [0, 0, 1] -layout (location = 0) uniform mat3x2 modelview_matrix; - -void main() { - // Multiply input position by the rotscale part of the matrix and then manually translate by - // the last column. This is equivalent to using a full 3x3 matrix and expanding the vector - // to `vec3(vert_position.xy, 1.0)` - gl_Position = vec4(mat2(modelview_matrix) * vert_position + modelview_matrix[2], 0.0, 1.0); - frag_tex_coord = vert_tex_coord; -} -)"; - -constexpr char FRAGMENT_SHADER[] = R"( -#version 430 core - -layout (location = 0) in vec2 frag_tex_coord; -layout (location = 0) out vec4 color; - -layout (binding = 0) uniform sampler2D color_texture; - -void main() { - color = vec4(texture(color_texture, frag_tex_coord).rgb, 1.0f); -} -)"; - constexpr GLint PositionLocation = 0; constexpr GLint TexCoordLocation = 1; constexpr GLint ModelViewMatrixLocation = 0; @@ -313,10 +275,11 @@ public: } }; -RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system, - Core::Frontend::GraphicsContext& context) - : RendererBase{emu_window}, emu_window{emu_window}, system{system}, context{context}, - program_manager{device}, has_debug_tool{HasDebugTool()} {} +RendererOpenGL::RendererOpenGL(Core::System& system_, Core::Frontend::EmuWindow& emu_window_, + Tegra::GPU& gpu_, + std::unique_ptr<Core::Frontend::GraphicsContext> context_) + : RendererBase{emu_window_, std::move(context_)}, system{system_}, + emu_window{emu_window_}, gpu{gpu_}, program_manager{device}, has_debug_tool{HasDebugTool()} {} RendererOpenGL::~RendererOpenGL() = default; @@ -384,7 +347,7 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { if (has_debug_tool) { glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0); Present(0); - context.SwapBuffers(); + context->SwapBuffers(); } } @@ -460,10 +423,10 @@ void RendererOpenGL::InitOpenGLObjects() { // Create shader programs OGLShader vertex_shader; - vertex_shader.Create(VERTEX_SHADER, GL_VERTEX_SHADER); + vertex_shader.Create(HostShaders::OPENGL_PRESENT_VERT, GL_VERTEX_SHADER); OGLShader fragment_shader; - fragment_shader.Create(FRAGMENT_SHADER, GL_FRAGMENT_SHADER); + fragment_shader.Create(HostShaders::OPENGL_PRESENT_FRAG, GL_FRAGMENT_SHADER); vertex_program.Create(true, false, vertex_shader.handle); fragment_program.Create(true, false, fragment_shader.handle); diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h index 8b18d32e6..52ea76b7d 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.h +++ b/src/video_core/renderer_opengl/renderer_opengl.h @@ -56,8 +56,9 @@ class FrameMailbox; class RendererOpenGL final : public VideoCore::RendererBase { public: - explicit RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system, - Core::Frontend::GraphicsContext& context); + explicit RendererOpenGL(Core::System& system, Core::Frontend::EmuWindow& emu_window, + Tegra::GPU& gpu, + std::unique_ptr<Core::Frontend::GraphicsContext> context); ~RendererOpenGL() override; bool Init() override; @@ -93,9 +94,9 @@ private: bool Present(int timeout_ms); - Core::Frontend::EmuWindow& emu_window; Core::System& system; - Core::Frontend::GraphicsContext& context; + Core::Frontend::EmuWindow& emu_window; + Tegra::GPU& gpu; const Device device; StateTracker state_tracker{system}; @@ -120,7 +121,7 @@ private: std::vector<u8> gl_framebuffer_data; /// Used for transforming the framebuffer orientation - Tegra::FramebufferConfig::TransformFlags framebuffer_transform_flags; + Tegra::FramebufferConfig::TransformFlags framebuffer_transform_flags{}; Common::Rectangle<int> framebuffer_crop_rect; /// Frame presentation mailbox diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp index 6e49699d0..ae46e0444 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp +++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp @@ -237,8 +237,10 @@ std::string BuildCommaSeparatedExtensions(std::vector<std::string> available_ext } // Anonymous namespace -RendererVulkan::RendererVulkan(Core::Frontend::EmuWindow& window, Core::System& system) - : RendererBase(window), system{system} {} +RendererVulkan::RendererVulkan(Core::System& system_, Core::Frontend::EmuWindow& emu_window, + Tegra::GPU& gpu_, + std::unique_ptr<Core::Frontend::GraphicsContext> context) + : RendererBase{emu_window, std::move(context)}, system{system_}, gpu{gpu_} {} RendererVulkan::~RendererVulkan() { ShutDown(); diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.h b/src/video_core/renderer_vulkan/renderer_vulkan.h index 522b5bff8..13debbbc0 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.h +++ b/src/video_core/renderer_vulkan/renderer_vulkan.h @@ -38,7 +38,9 @@ struct VKScreenInfo { class RendererVulkan final : public VideoCore::RendererBase { public: - explicit RendererVulkan(Core::Frontend::EmuWindow& window, Core::System& system); + explicit RendererVulkan(Core::System& system, Core::Frontend::EmuWindow& emu_window, + Tegra::GPU& gpu, + std::unique_ptr<Core::Frontend::GraphicsContext> context); ~RendererVulkan() override; bool Init() override; @@ -58,6 +60,7 @@ private: void Report() const; Core::System& system; + Tegra::GPU& gpu; Common::DynamicLibrary library; vk::InstanceDispatch dld; diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index 936f76195..ff1b52eab 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -1443,10 +1443,10 @@ void RasterizerVulkan::UpdateFrontFace(Tegra::Engines::Maxwell3D::Regs& regs) { } void RasterizerVulkan::UpdatePrimitiveTopology(Tegra::Engines::Maxwell3D::Regs& regs) { - if (!state_tracker.TouchPrimitiveTopology()) { + const Maxwell::PrimitiveTopology primitive_topology = regs.draw.topology.Value(); + if (!state_tracker.ChangePrimitiveTopology(primitive_topology)) { return; } - const Maxwell::PrimitiveTopology primitive_topology = regs.draw.topology.Value(); scheduler.Record([this, primitive_topology](vk::CommandBuffer cmdbuf) { cmdbuf.SetPrimitiveTopologyEXT(MaxwellToVK::PrimitiveTopology(device, primitive_topology)); }); diff --git a/src/video_core/renderer_vulkan/vk_state_tracker.cpp b/src/video_core/renderer_vulkan/vk_state_tracker.cpp index 9151d9fb1..4bd1009f9 100644 --- a/src/video_core/renderer_vulkan/vk_state_tracker.cpp +++ b/src/video_core/renderer_vulkan/vk_state_tracker.cpp @@ -42,7 +42,6 @@ Flags MakeInvalidationFlags() { flags[DepthWriteEnable] = true; flags[DepthCompareOp] = true; flags[FrontFace] = true; - flags[PrimitiveTopology] = true; flags[StencilOp] = true; flags[StencilTestEnable] = true; return flags; @@ -112,10 +111,6 @@ void SetupDirtyFrontFace(Tables& tables) { table[OFF(screen_y_control)] = FrontFace; } -void SetupDirtyPrimitiveTopology(Tables& tables) { - tables[0][OFF(draw.topology)] = PrimitiveTopology; -} - void SetupDirtyStencilOp(Tables& tables) { auto& table = tables[0]; table[OFF(stencil_front_op_fail)] = StencilOp; @@ -156,13 +151,13 @@ void StateTracker::Initialize() { SetupDirtyDepthWriteEnable(tables); SetupDirtyDepthCompareOp(tables); SetupDirtyFrontFace(tables); - SetupDirtyPrimitiveTopology(tables); SetupDirtyStencilOp(tables); SetupDirtyStencilTestEnable(tables); } void StateTracker::InvalidateCommandBufferState() { system.GPU().Maxwell3D().dirty.flags |= invalidation_flags; + current_topology = INVALID_TOPOLOGY; } } // namespace Vulkan diff --git a/src/video_core/renderer_vulkan/vk_state_tracker.h b/src/video_core/renderer_vulkan/vk_state_tracker.h index 54ca0d6c6..13a6ce786 100644 --- a/src/video_core/renderer_vulkan/vk_state_tracker.h +++ b/src/video_core/renderer_vulkan/vk_state_tracker.h @@ -32,7 +32,6 @@ enum : u8 { DepthWriteEnable, DepthCompareOp, FrontFace, - PrimitiveTopology, StencilOp, StencilTestEnable, @@ -43,6 +42,8 @@ static_assert(Last <= std::numeric_limits<u8>::max()); } // namespace Dirty class StateTracker { + using Maxwell = Tegra::Engines::Maxwell3D::Regs; + public: explicit StateTracker(Core::System& system); @@ -102,10 +103,6 @@ public: return Exchange(Dirty::FrontFace, false); } - bool TouchPrimitiveTopology() { - return Exchange(Dirty::PrimitiveTopology, false); - } - bool TouchStencilOp() { return Exchange(Dirty::StencilOp, false); } @@ -114,7 +111,15 @@ public: return Exchange(Dirty::StencilTestEnable, false); } + bool ChangePrimitiveTopology(Maxwell::PrimitiveTopology new_topology) { + const bool has_changed = current_topology != new_topology; + current_topology = new_topology; + return has_changed; + } + private: + static constexpr auto INVALID_TOPOLOGY = static_cast<Maxwell::PrimitiveTopology>(~0u); + bool Exchange(std::size_t id, bool new_value) const noexcept { auto& flags = system.GPU().Maxwell3D().dirty.flags; const bool is_dirty = flags[id]; @@ -124,6 +129,7 @@ private: Core::System& system; Tegra::Engines::Maxwell3D::DirtyState::Flags invalidation_flags; + Maxwell::PrimitiveTopology current_topology = INVALID_TOPOLOGY; }; } // namespace Vulkan diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp index e4739394d..e2bba88dd 100644 --- a/src/video_core/shader/decode/memory.cpp +++ b/src/video_core/shader/decode/memory.cpp @@ -386,7 +386,8 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { break; } case OpCode::Id::RED: { - UNIMPLEMENTED_IF_MSG(instr.red.type != GlobalAtomicType::U32); + UNIMPLEMENTED_IF_MSG(instr.red.type != GlobalAtomicType::U32, "type={}", + static_cast<int>(instr.red.type.Value())); const auto [real_address, base_address, descriptor] = TrackGlobalMemory(bb, instr, true, true); if (!real_address || !base_address) { diff --git a/src/video_core/video_core.cpp b/src/video_core/video_core.cpp index 45f360bdd..4e3a092c7 100644 --- a/src/video_core/video_core.cpp +++ b/src/video_core/video_core.cpp @@ -3,6 +3,7 @@ // Refer to the license.txt file included. #include <memory> + #include "common/logging/log.h" #include "core/core.h" #include "core/settings.h" @@ -16,37 +17,46 @@ #include "video_core/video_core.h" namespace { -std::unique_ptr<VideoCore::RendererBase> CreateRenderer(Core::Frontend::EmuWindow& emu_window, - Core::System& system, - Core::Frontend::GraphicsContext& context) { + +std::unique_ptr<VideoCore::RendererBase> CreateRenderer( + Core::System& system, Core::Frontend::EmuWindow& emu_window, Tegra::GPU& gpu, + std::unique_ptr<Core::Frontend::GraphicsContext> context) { switch (Settings::values.renderer_backend.GetValue()) { case Settings::RendererBackend::OpenGL: - return std::make_unique<OpenGL::RendererOpenGL>(emu_window, system, context); + return std::make_unique<OpenGL::RendererOpenGL>(system, emu_window, gpu, + std::move(context)); #ifdef HAS_VULKAN case Settings::RendererBackend::Vulkan: - return std::make_unique<Vulkan::RendererVulkan>(emu_window, system); + return std::make_unique<Vulkan::RendererVulkan>(system, emu_window, gpu, + std::move(context)); #endif default: return nullptr; } } + } // Anonymous namespace namespace VideoCore { std::unique_ptr<Tegra::GPU> CreateGPU(Core::Frontend::EmuWindow& emu_window, Core::System& system) { + std::unique_ptr<Tegra::GPU> gpu; + if (Settings::values.use_asynchronous_gpu_emulation.GetValue()) { + gpu = std::make_unique<VideoCommon::GPUAsynch>(system); + } else { + gpu = std::make_unique<VideoCommon::GPUSynch>(system); + } + auto context = emu_window.CreateSharedContext(); const auto scope = context->Acquire(); - auto renderer = CreateRenderer(emu_window, system, *context); + + auto renderer = CreateRenderer(system, emu_window, *gpu, std::move(context)); if (!renderer->Init()) { return nullptr; } - if (Settings::values.use_asynchronous_gpu_emulation.GetValue()) { - return std::make_unique<VideoCommon::GPUAsynch>(system, std::move(renderer), - std::move(context)); - } - return std::make_unique<VideoCommon::GPUSynch>(system, std::move(renderer), std::move(context)); + gpu->BindRenderer(std::move(renderer)); + return gpu; } u16 GetResolutionScaleFactor(const RendererBase& renderer) { |