Add option to choose Tracy instead of Optick for CPU profiling

CMakeLists.txt

@@ -9,7 +9,9 @@ option(NC_BUILD_TESTS "Enable building tests." OFF)
 option(NC_BUILD_INTEGRATION_TESTS "Enable building integration tests." OFF)
 option(NC_BUILD_NCCONVERT "Enable building nc-convert asset converter." ON)
 option(NC_RUNTIME_SHADER_COMPILATION "Enable compiling shaders at runtime" ON)
-option(NC_PROFILING_ENABLED "Enable profiling with Optick" OFF)
+option(NC_PROFILING_ENABLED "Enable profiling" OFF)
+set(NC_PROFILER "Tracy" CACHE STRING "Profiler to use (Tracy or Optick)")
+set_property(CACHE NC_PROFILER PROPERTY STRINGS "Tracy" "Optick")
 option(NC_PROD_BUILD "Only build NcEngine production binaries." OFF)
 
 # Variables
@@ -59,6 +61,11 @@ endif()
 
 if(NC_PROFILING_ENABLED)
     list(APPEND NC_COMPILE_DEFINITIONS NC_PROFILING_ENABLED)
+    if(NC_PROFILER STREQUAL "Tracy")
+        list(APPEND NC_COMPILE_DEFINITIONS NC_USE_TRACY)
+    elseif(NC_PROFILER STREQUAL "Optick")
+        list(APPEND NC_COMPILE_DEFINITIONS NC_USE_OPTICK)
+    endif()
 endif()
 
 # Get dependencies

cmake/FetchDependencies.cmake

@@ -72,7 +72,7 @@ FetchContent_Declare(taskflow
 
 # Optick
 set(OPTICK_INSTALL_TARGETS OFF CACHE BOOL "" FORCE)
-if(${NC_PROFILING_ENABLED})
+if(NC_PROFILING_ENABLED AND NC_PROFILER STREQUAL "Optick")
     set(OPTICK_ENABLED ON CACHE BOOL "" FORCE)
 else()
     set(OPTICK_ENABLED OFF CACHE BOOL "" FORCE)
@@ -84,6 +84,20 @@ FetchContent_Declare(optick
                      GIT_SHALLOW    TRUE
 )
 
+# Tracy
+if(NC_PROFILING_ENABLED AND NC_PROFILER STREQUAL "Tracy")
+    set(TRACY_ENABLE ON CACHE BOOL "" FORCE)
+else()
+    set(TRACY_ENABLE OFF CACHE BOOL "" FORCE)
+endif()
+set(TRACY_ON_DEMAND ON CACHE BOOL "" FORCE)
+
+FetchContent_Declare(tracy
+                     GIT_REPOSITORY https://github.com/wolfpld/tracy.git
+                     GIT_TAG        v0.13.1
+                     GIT_SHALLOW    TRUE
+)
+
 # Jolt
 set(CPP_EXCEPTIONS_ENABLED ON CACHE BOOL "" FORCE)
 set(CPP_RTTI_ENABLED ON CACHE BOOL "" FORCE)
@@ -156,7 +170,7 @@ FetchContent_Declare(fmt
 )
 
 # Fetch all required sources
-FetchContent_MakeAvailable(taskflow optick JoltPhysics DirectXMath fmt DiligentCore DiligentTools)
+FetchContent_MakeAvailable(taskflow optick tracy JoltPhysics DirectXMath fmt DiligentCore DiligentTools)
 
 # Silence warnings
 disable_warnings_for_headers(Taskflow)
@@ -172,10 +186,15 @@ disable_warnings_for_target(Diligent-Imgui)
 if(NC_PROFILING_ENABLED AND NOT CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
     target_compile_definitions(Jolt
         PUBLIC
-            -DJPH_EXTERNAL_PROFILE
+            JPH_EXTERNAL_PROFILE
     )
 endif()
 
+# Silence Tracy warnings
+if(NC_PROFILING_ENABLED AND NC_PROFILER STREQUAL "Tracy")
+    disable_warnings_for_headers(TracyClient)
+endif()
+
 # Optional Dependencies
 
 if(NC_BUILD_NCCONVERT)

include/ncengine/debug/Profile.h

@@ -6,11 +6,45 @@
 
 #ifdef NC_PROFILING_ENABLED
 
+#if defined(NC_USE_TRACY)
+
+#include "tracy/Tracy.hpp"
+
+namespace nc
+{
+/** @cond internal */
+namespace detail
+{
+struct TaskMeasurement
+{
+    explicit TaskMeasurement()
+    {
+        tracy::SetThreadName("NcEngine Worker");
+    }
+
+    ~TaskMeasurement() noexcept = default;
+};
+} // namespace detail
+} // namespace nc
+/** @endcond */
+
+/** @brief Mark frame boundary. */
+#define NC_PROFILE_FRAME(name) FrameMark
+
+/** @brief Profile a function or inner scope. */
+#define NC_PROFILE_SCOPE(name, category) ZoneScopedN(name)
+
+/** @brief Profile a task. Use this at the top level of a task instead of NC_PROFILE_SCOPE for proper thread tracking. */
+#define NC_PROFILE_TASK(name, category)                                        \
+    const auto _ncTaskMeasurement ## __LINE__ = nc::detail::TaskMeasurement{}; \
+    NC_PROFILE_SCOPE(name, category)
+
+#elif defined(NC_USE_OPTICK)
+
 #include "optick.h"
 
 namespace nc
 {
-/** @brief Color and filter for a profile event. */
 using ProfileCategory = Optick::Category;
 
 /** @cond internal */
@@ -32,15 +66,23 @@ struct TaskMeasurement
 } // namespace nc
 /** @endcond */
 
+/** @brief Mark frame boundary. */
+#define NC_PROFILE_FRAME(name) OPTICK_FRAME(name)
+
 /** @brief Profile a function or inner scope. */
-#define NC_PROFILE_SCOPE(name, category) OPTICK_CATEGORY(name, category);
+#define NC_PROFILE_SCOPE(name, category) OPTICK_CATEGORY(name, category)
 
 /** @brief Profile a task. Use this at the top level of a task instead of NC_PROFILE_SCOPE for proper thread tracking. */
 #define NC_PROFILE_TASK(name, category)                                        \
     const auto _ncTaskMeasurement ## __LINE__ = nc::detail::TaskMeasurement{}; \
-    NC_PROFILE_SCOPE(name, category);
+    NC_PROFILE_SCOPE(name, category)
 
-#else
+#endif // NC_USE_TRACY / NC_USE_OPTICK
+
+#else // NC_PROFILING_ENABLED not defined
+
+#define NC_PROFILE_FRAME(name)
 #define NC_PROFILE_SCOPE(name, category)
 #define NC_PROFILE_TASK(name, category)
-#endif
+
+#endif // NC_PROFILING_ENABLED

source/ncengine/CMakeLists.txt

@@ -30,6 +30,13 @@ set_target_properties(${NC_ENGINE_LIB}
     PROPERTIES CMAKE_DEBUG_POSTFIX d
 )
 
+# Profiler library selection
+if(NC_PROFILING_ENABLED AND NC_PROFILER STREQUAL "Tracy")
+    set(NC_PROFILER_LIB TracyClient)
+else()
+    set(NC_PROFILER_LIB OptickCore)
+endif()
+
 # Link libraries
 target_link_libraries(${NC_ENGINE_LIB}
     PUBLIC
@@ -40,7 +47,7 @@ target_link_libraries(${NC_ENGINE_LIB}
         glfw
         imgui
         Taskflow
-        OptickCore
+        ${NC_PROFILER_LIB}
         RtAudio
         ${DILIGENT_LIBRARIES}
 )

source/ncengine/engine/NcEngineImpl.cpp

@@ -9,7 +9,7 @@
 #include "utility/Log.h"
 #include "window/Window.h"
 
-#include "optick.h"
+#include "ncengine/debug/Profile.h"
 
 #include <sstream>
 #include <thread>
@@ -171,7 +171,7 @@ void NcEngineImpl::Run()
     auto* ncScene = m_modules->Get<NcScene>();
     auto update = [this, ncWindow = m_modules->Get<window::NcWindow>()](float dt)
     {
-        OPTICK_FRAME("Main Thread");
+        NC_PROFILE_FRAME("Main Thread");
         time::SetDeltaTime(dt);
         input::Flush();
         ncWindow->ProcessSystemMessages();

source/ncengine/graphics2/NcGraphicsImpl2.cpp

@@ -532,7 +532,7 @@ void NcGraphicsImpl2::OnBuildTaskGraph(task::UpdateTasks& update, task::RenderTa
 
 void NcGraphicsImpl2::Run()
 {
-    NC_PROFILE_TASK("Render", Optick::Category::Rendering);
+    NC_PROFILE_TASK("Render", ProfileCategory::Rendering);
 
     if (m_resizeNeeded)
     {

source/ncengine/graphics2/diligent/UIBackend.cpp

@@ -1,5 +1,6 @@
 #include "UIBackend.h"
 
+#include "ncengine/debug/Profile.h"
 #include "imgui.h"
 #include "ImGuizmo.h"
 
@@ -16,13 +17,15 @@ UIBackend::UIBackend(Diligent::IRenderDevice& device,
 
 void UIBackend::FrameBegin(Diligent::ISwapChain& swapChain)
 {
+    NC_PROFILE_SCOPE("UIBackend::FrameBegin", ProfileCategory::Rendering);
     const auto& scDesc = swapChain.GetDesc();
     m_imguiBackend.NewFrame(scDesc.Width, scDesc.Height, scDesc.PreTransform);
     ImGuizmo::BeginFrame();
 }
 
 void UIBackend::Render(Diligent::IDeviceContext& context)
 {
+    NC_PROFILE_SCOPE("UIBackend::Render", ProfileCategory::Rendering);
     m_imguiBackend.Render(&context);
 }
 

source/ncengine/graphics2/diligent/pass/PassBackend.cpp

@@ -212,6 +212,7 @@ void PassBackend::RenderShadowPass(IDeviceContext& context,
                                    const std::vector<Batch>& skinnedBatches,
                                    const std::span<const LightData>& lights)
 {
+    NC_PROFILE_SCOPE("PassBackend::RenderShadowPass", ProfileCategory::Rendering);
     auto& sinkIndexBuffer = perPassResourceSignature.GetSinkIndexBufferResource();
     auto& uniShadowMapsBuffer = perPassResourceSignature.GetUniShadowMapSinksResource();
     auto& pointShadowMapsBuffer = perPassResourceSignature.GetPointShadowMapSinksResource();
@@ -418,8 +419,8 @@ void PassBackend::RenderParticle(IDeviceContext& context,
                                  PerPassResourceSignature& perPassResourceSignature,
                                  const ParticleRenderState& state,
                                  const Viewport& viewport)
-
 {
+    NC_PROFILE_SCOPE("PassBackend::RenderParticle", ProfileCategory::Rendering);
     if (state.particleData.instances.empty() || !m_particlePass)
     {
         return;

source/ncengine/graphics2/diligent/resource/CubeMapBufferResource.cpp

@@ -1,4 +1,5 @@
 #include "CubeMapBufferResource.h"
+#include "ncengine/debug/Profile.h"
 #include "ResourceTypes.h"
 
 #include "TextureLoader.h"
@@ -36,6 +37,7 @@ void CubeMapBufferResource::Load(std::span<const asset::CubeMapWithId> cubeMaps,
                               IDeviceContext& context,
                               IRenderDevice& device)
 {
+    NC_PROFILE_SCOPE("CubeMapBufferResource::Load", ProfileCategory::Rendering);
     const auto cubeMapCount = cubeMaps.size();
     if (cubeMapCount == 0)
     {

source/ncengine/graphics2/diligent/resource/EnvironmentBufferResource.cpp

@@ -1,5 +1,6 @@
 #include "EnvironmentBufferResource.h"
 #include "graphics2/frontend/subsystem/CameraRenderState.h"
+#include "ncengine/debug/Profile.h"
 #include "graphics2/frontend/subsystem/EnvironmentRenderState.h"
 #include "graphics2/frontend/subsystem/LightRenderState.h"
 #include "ncengine/graphics/Environment.h"
@@ -25,6 +26,7 @@ void EnvironmentBufferResource::Update(Diligent::IDeviceContext& context,
                                        const LightRenderState& lightRenderState,
                                        const EnvironmentRenderState& environmentRenderState)
 {
+    NC_PROFILE_SCOPE("EnvironmentBufferResource::Update", ProfileCategory::Rendering);
     const auto data = GlobalEnvironmentData{
         .cameraViewProjection = cameraState.viewProjection,
         .cameraInvProjection = cameraState.invProjection,

source/ncengine/graphics2/diligent/resource/SinkBufferResource.cpp

@@ -1,5 +1,6 @@
 #include "SinkBufferResource.h"
 #include "graphics2/diligent/pass/PassTypes.h"
+#include "ncengine/debug/Profile.h"
 #include "ResourceTypes.h"
 
 #include "ncutility/NcError.h"
@@ -210,6 +211,7 @@ void SinkBufferResource::Clear()
 
 void SinkBufferResource::Update()
 {
+    NC_PROFILE_SCOPE("SinkBufferResource::Update", ProfileCategory::Rendering);
     SetArrayRegion(m_variable, std::span<Diligent::IDeviceObject*>(m_shaderResourceViews), 0u, m_shaderResourceViews.size());
 }
 } // namespace nc::graphics

source/ncengine/graphics2/diligent/resource/TextureBufferResource.cpp

@@ -1,4 +1,5 @@
 #include "TextureBufferResource.h"
+#include "ncengine/debug/Profile.h"
 #include "ResourceTypes.h"
 
 #include "TextureLoader.h"
@@ -31,6 +32,7 @@ void TextureBufferResource::Load(std::span<const asset::TextureWithId> textures,
                                  Diligent::IDeviceContext& context,
                                  Diligent::IRenderDevice& device)
 {
+    NC_PROFILE_SCOPE("TextureBufferResource::Load", ProfileCategory::Rendering);
     if (!m_initialLoadComplete)
     {
         InitializeArray(context, device, m_variable, m_maxTextures);

source/ncengine/graphics2/diligent/resource/WireframeBufferResource.cpp

@@ -1,4 +1,5 @@
 #include "WireframeBufferResource.h"
+#include "ncengine/debug/Profile.h"
 
 namespace nc::graphics
 {
@@ -19,6 +20,7 @@ WireframeBufferResource::WireframeBufferResource(Diligent::IDeviceContext& conte
 void WireframeBufferResource::Update(Diligent::IDeviceContext& context,
                                      const WireframeData& data)
 {
+    NC_PROFILE_SCOPE("WireframeBufferResource::Update", ProfileCategory::Rendering);
     m_buffer.Write(context, data);
 }
 } // namespace nc::graphics

source/ncengine/graphics2/frontend/GraphicsFrontend.cpp

@@ -1,11 +1,13 @@
 #include "GraphicsFrontend.h"
 #include "FrontendRenderState.h"
+#include "ncengine/debug/Profile.h"
 #include "ncengine/ecs/Ecs.h"
 
 namespace nc::graphics
 {
 auto GraphicsFrontend::BuildRenderState(ecs::Ecs world) -> FrontendRenderState
 {
+    NC_PROFILE_SCOPE("GraphicsFrontend::BuildRenderState", ProfileCategory::Rendering);
     return FrontendRenderState{
         .cameraState = m_cameraSystem.BuildState(world),
         .environmentRenderState = m_environmentSystem.BuildState(),

source/ncengine/graphics2/frontend/subsystem/CameraSubsystem.cpp

@@ -1,6 +1,7 @@
 #include "CameraSubsystem.h"
 #include "CameraRenderState.h"
 
+#include "ncengine/debug/Profile.h"
 #include "ncengine/ecs/Ecs.h"
 #include "ncengine/graphics/Camera.h"
 #include "ncengine/window/Window.h"
@@ -20,6 +21,7 @@ namespace nc::graphics
 {
 auto CameraSubsystem::BuildState(ecs::ExplicitEcs<Transform> ecs) -> CameraRenderState
 {
+    NC_PROFILE_SCOPE("CameraSubsystem::BuildState", ProfileCategory::Rendering);
     if (m_mainCamera)
     {
         const auto& transform = ecs.Get<Transform>(m_mainCamera->ParentEntity());

source/ncengine/graphics2/frontend/subsystem/InstanceCache.inl

@@ -133,6 +133,7 @@ void InstanceCache<T>::CommitPendingChanges()
 template<class T>
 auto InstanceCache<T>::BuildState() -> BufferUpdateInfo<T>
 {
+    NC_PROFILE_SCOPE("InstanceCache::BuildState", ProfileCategory::Rendering);
     // For simplicity, we're only tracking one dirty range from the first modified index to the buffer
     // end. This can result in more copies than necessary for non-shifting insertions and removals, but the
     // data is pretty cheap to copy.
@@ -181,6 +182,7 @@ auto InstanceCache<T>::BuildBatches(std::span<const MaterialPassFlag::type> pass
 template<class T>
 void InstanceCache<T>::Purge() noexcept
 {
+    NC_PROFILE_SCOPE("InstanceCache::Purge", ProfileCategory::Rendering);
     CommitPendingChanges();
     m_dirtyBegin = 0u;
     auto bufferEnd = 0u;

source/ncengine/graphics2/frontend/subsystem/LightSubsystem.cpp

@@ -1,5 +1,6 @@
 #include "LightSubsystem.h"
 
+#include "ncengine/debug/Profile.h"
 #include "ncengine/ecs/Ecs.h"
 #include "ncengine/ecs/Transform.h"
 #include "ncengine/graphics/Light.h"
@@ -37,6 +38,7 @@ namespace nc::graphics
 {
 auto LightSubsystem::BuildState(ecs::ExplicitEcs<DirectionalLight, PointLight, SpotLight, Transform> ecs) -> LightRenderState
 {
+    NC_PROFILE_SCOPE("LightSubsystem::BuildState", ProfileCategory::Rendering);
     m_lightData.clear();
     m_lightMatrixData.clear();
     auto lightMatrixIndex = 0u;

source/ncengine/graphics2/frontend/subsystem/PostProcessSubsystem.cpp

@@ -1,4 +1,5 @@
 #include "PostProcessSubsystem.h"
+#include "ncengine/debug/Profile.h"
 #include "ncengine/graphics/GraphicsUtility.h"
 
 #include "ncutility/NcError.h"
@@ -121,6 +122,7 @@ void PostProcessSubsystem::SetProperties(PostProcessEffectId effectId,
 
 auto PostProcessSubsystem::BuildState() -> PostProcessState
 {
+    NC_PROFILE_SCOPE("PostProcessSubsystem::BuildState", ProfileCategory::Rendering);
     return PostProcessState{
         .toggledEffects = std::move(m_toggledEffects),
         .modifiedProperties = std::move(m_modifiedProperties)

source/ncengine/graphics2/frontend/subsystem/TransformCache.cpp

@@ -33,6 +33,7 @@ void TransformCache::RemoveInstance(TransformDataHandle instance)
 
 void TransformCache::CommitPendingChanges()
 {
+    NC_PROFILE_SCOPE("TransformCache::CommitPendingChanges", ProfileCategory::Rendering);
     m_buffer.CommitPendingChanges();
 }