diff options
Diffstat (limited to 'src')
| -rw-r--r-- | src/video_core/gpu.cpp | 7 | ||||
| -rw-r--r-- | src/video_core/gpu.h | 4 | ||||
| -rw-r--r-- | src/video_core/gpu_thread.cpp | 30 | ||||
| -rw-r--r-- | src/video_core/gpu_thread.h | 5 |
4 files changed, 34 insertions, 12 deletions
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index f99a8a0de..6ab06775f 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp | |||
| @@ -38,7 +38,7 @@ GPU::GPU(Core::System& system_, bool is_async_, bool use_nvdec_) | |||
| 38 | maxwell_dma{std::make_unique<Engines::MaxwellDMA>(system, *memory_manager)}, | 38 | maxwell_dma{std::make_unique<Engines::MaxwellDMA>(system, *memory_manager)}, |
| 39 | kepler_memory{std::make_unique<Engines::KeplerMemory>(system, *memory_manager)}, | 39 | kepler_memory{std::make_unique<Engines::KeplerMemory>(system, *memory_manager)}, |
| 40 | shader_notify{std::make_unique<VideoCore::ShaderNotify>()}, is_async{is_async_}, | 40 | shader_notify{std::make_unique<VideoCore::ShaderNotify>()}, is_async{is_async_}, |
| 41 | gpu_thread{system_} {} | 41 | gpu_thread{system_, is_async_} {} |
| 42 | 42 | ||
| 43 | GPU::~GPU() = default; | 43 | GPU::~GPU() = default; |
| 44 | 44 | ||
| @@ -524,7 +524,10 @@ void GPU::WaitIdle() const { | |||
| 524 | } | 524 | } |
| 525 | 525 | ||
| 526 | void GPU::OnCommandListEnd() { | 526 | void GPU::OnCommandListEnd() { |
| 527 | gpu_thread.OnCommandListEnd(); | 527 | if (is_async) { |
| 528 | // This command only applies to asynchronous GPU mode | ||
| 529 | gpu_thread.OnCommandListEnd(); | ||
| 530 | } | ||
| 528 | } | 531 | } |
| 529 | 532 | ||
| 530 | } // namespace Tegra | 533 | } // namespace Tegra |
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h index a2bb4d82d..d81e38680 100644 --- a/src/video_core/gpu.h +++ b/src/video_core/gpu.h | |||
| @@ -406,10 +406,10 @@ private: | |||
| 406 | u64 last_flush_fence{}; | 406 | u64 last_flush_fence{}; |
| 407 | std::mutex flush_request_mutex; | 407 | std::mutex flush_request_mutex; |
| 408 | 408 | ||
| 409 | const bool is_async; | ||
| 410 | |||
| 409 | VideoCommon::GPUThread::ThreadManager gpu_thread; | 411 | VideoCommon::GPUThread::ThreadManager gpu_thread; |
| 410 | std::unique_ptr<Core::Frontend::GraphicsContext> cpu_context; | 412 | std::unique_ptr<Core::Frontend::GraphicsContext> cpu_context; |
| 411 | |||
| 412 | const bool is_async; | ||
| 413 | }; | 413 | }; |
| 414 | 414 | ||
| 415 | #define ASSERT_REG_POSITION(field_name, position) \ | 415 | #define ASSERT_REG_POSITION(field_name, position) \ |
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp index e27218b96..56b9621b1 100644 --- a/src/video_core/gpu_thread.cpp +++ b/src/video_core/gpu_thread.cpp | |||
| @@ -65,7 +65,8 @@ static void RunThread(Core::System& system, VideoCore::RendererBase& renderer, | |||
| 65 | } | 65 | } |
| 66 | } | 66 | } |
| 67 | 67 | ||
| 68 | ThreadManager::ThreadManager(Core::System& system_) : system{system_} {} | 68 | ThreadManager::ThreadManager(Core::System& system_, bool is_async_) |
| 69 | : system{system_}, is_async{is_async_} {} | ||
| 69 | 70 | ||
| 70 | ThreadManager::~ThreadManager() { | 71 | ThreadManager::~ThreadManager() { |
| 71 | if (!thread.joinable()) { | 72 | if (!thread.joinable()) { |
| @@ -97,19 +98,30 @@ void ThreadManager::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { | |||
| 97 | } | 98 | } |
| 98 | 99 | ||
| 99 | void ThreadManager::FlushRegion(VAddr addr, u64 size) { | 100 | void ThreadManager::FlushRegion(VAddr addr, u64 size) { |
| 100 | if (!Settings::IsGPULevelHigh()) { | 101 | if (!is_async) { |
| 102 | // Always flush with synchronous GPU mode | ||
| 101 | PushCommand(FlushRegionCommand(addr, size)); | 103 | PushCommand(FlushRegionCommand(addr, size)); |
| 102 | return; | 104 | return; |
| 103 | } | 105 | } |
| 104 | if (!Settings::IsGPULevelExtreme()) { | 106 | |
| 105 | return; | 107 | // Asynchronous GPU mode |
| 106 | } | 108 | switch (Settings::values.gpu_accuracy.GetValue()) { |
| 107 | if (system.Renderer().Rasterizer().MustFlushRegion(addr, size)) { | 109 | case Settings::GPUAccuracy::Normal: |
| 110 | PushCommand(FlushRegionCommand(addr, size)); | ||
| 111 | break; | ||
| 112 | case Settings::GPUAccuracy::High: | ||
| 113 | // TODO(bunnei): Is this right? Preserving existing behavior for now | ||
| 114 | break; | ||
| 115 | case Settings::GPUAccuracy::Extreme: { | ||
| 108 | auto& gpu = system.GPU(); | 116 | auto& gpu = system.GPU(); |
| 109 | u64 fence = gpu.RequestFlush(addr, size); | 117 | u64 fence = gpu.RequestFlush(addr, size); |
| 110 | PushCommand(GPUTickCommand()); | 118 | PushCommand(GPUTickCommand()); |
| 111 | while (fence > gpu.CurrentFlushRequestFence()) { | 119 | while (fence > gpu.CurrentFlushRequestFence()) { |
| 112 | } | 120 | } |
| 121 | break; | ||
| 122 | } | ||
| 123 | default: | ||
| 124 | UNIMPLEMENTED_MSG("Unsupported gpu_accuracy {}", Settings::values.gpu_accuracy.GetValue()); | ||
| 113 | } | 125 | } |
| 114 | } | 126 | } |
| 115 | 127 | ||
| @@ -134,6 +146,12 @@ void ThreadManager::OnCommandListEnd() { | |||
| 134 | u64 ThreadManager::PushCommand(CommandData&& command_data) { | 146 | u64 ThreadManager::PushCommand(CommandData&& command_data) { |
| 135 | const u64 fence{++state.last_fence}; | 147 | const u64 fence{++state.last_fence}; |
| 136 | state.queue.Push(CommandDataContainer(std::move(command_data), fence)); | 148 | state.queue.Push(CommandDataContainer(std::move(command_data), fence)); |
| 149 | |||
| 150 | if (!is_async) { | ||
| 151 | // In synchronous GPU mode, block the caller until the command has executed | ||
| 152 | WaitIdle(); | ||
| 153 | } | ||
| 154 | |||
| 137 | return fence; | 155 | return fence; |
| 138 | } | 156 | } |
| 139 | 157 | ||
diff --git a/src/video_core/gpu_thread.h b/src/video_core/gpu_thread.h index 0071195d6..2775629e7 100644 --- a/src/video_core/gpu_thread.h +++ b/src/video_core/gpu_thread.h | |||
| @@ -27,7 +27,7 @@ class System; | |||
| 27 | } // namespace Core | 27 | } // namespace Core |
| 28 | 28 | ||
| 29 | namespace VideoCore { | 29 | namespace VideoCore { |
| 30 | class RendererBase; | 30 | class RendererBase; |
| 31 | } // namespace VideoCore | 31 | } // namespace VideoCore |
| 32 | 32 | ||
| 33 | namespace VideoCommon::GPUThread { | 33 | namespace VideoCommon::GPUThread { |
| @@ -117,7 +117,7 @@ struct SynchState final { | |||
| 117 | /// Class used to manage the GPU thread | 117 | /// Class used to manage the GPU thread |
| 118 | class ThreadManager final { | 118 | class ThreadManager final { |
| 119 | public: | 119 | public: |
| 120 | explicit ThreadManager(Core::System& system_); | 120 | explicit ThreadManager(Core::System& system_, bool is_async_); |
| 121 | ~ThreadManager(); | 121 | ~ThreadManager(); |
| 122 | 122 | ||
| 123 | /// Creates and starts the GPU thread. | 123 | /// Creates and starts the GPU thread. |
| @@ -155,6 +155,7 @@ private: | |||
| 155 | Core::System& system; | 155 | Core::System& system; |
| 156 | std::thread thread; | 156 | std::thread thread; |
| 157 | std::thread::id thread_id; | 157 | std::thread::id thread_id; |
| 158 | const bool is_async; | ||
| 158 | }; | 159 | }; |
| 159 | 160 | ||
| 160 | } // namespace VideoCommon::GPUThread | 161 | } // namespace VideoCommon::GPUThread |