Merge pull request #5278 from samcake/yellow

Adding more NSight instrumentation

interface/src/Application.cpp

@@ -1183,7 +1183,6 @@ bool Application::event(QEvent* event) {
 }
 
 bool Application::eventFilter(QObject* object, QEvent* event) {
-
     if (event->type() == QEvent::ShortcutOverride) {
         if (DependencyManager::get<OffscreenUi>()->shouldSwallowShortcut(event)) {
             event->accept();
@@ -1788,6 +1787,7 @@ void Application::checkFPS() {
 }
 
 void Application::idle() {
+    PROFILE_RANGE(__FUNCTION__);
     static SimpleAverage<float> interIdleDurations;
     static uint64_t lastIdleEnd{ 0 };
 

interface/src/GLCanvas.cpp

@@ -58,6 +58,7 @@ void GLCanvas::initializeGL() {
 }
 
 void GLCanvas::paintGL() {
+    PROFILE_RANGE(__FUNCTION__);
     if (!_throttleRendering && !Application::getInstance()->getWindow()->isMinimized()) {
         Application::getInstance()->paintGL();
     }

libraries/entities-renderer/src/RenderableModelEntityItem.cpp

@@ -161,6 +161,7 @@ namespace render {
     template <> void payloadRender(const RenderableModelEntityItemMeta::Pointer& payload, RenderArgs* args) {
         if (args) {
             if (payload && payload->entity) {
+                PROFILE_RANGE("MetaModelRender");
                 payload->entity->render(args);
             }
         }

libraries/gpu/src/gpu/Batch.cpp

@@ -12,6 +12,17 @@
 
 #include <QDebug>
 
+#if defined(NSIGHT_FOUND)
+#include "nvToolsExt.h"
+
+ProfileRange::ProfileRange(const char *name) {
+    nvtxRangePush(name);
+}
+ProfileRange::~ProfileRange() {
+    nvtxRangePop();
+}
+#endif
+
 #define ADD_COMMAND(call) _commands.push_back(COMMAND_##call); _commandOffsets.push_back(_params.size());
 
 using namespace gpu;

libraries/gpu/src/gpu/Batch.h

@@ -26,17 +26,11 @@
 #include "Framebuffer.h"
 
 #if defined(NSIGHT_FOUND)
-    #include "nvToolsExt.h"
     class ProfileRange {
     public:
-        ProfileRange(const char *name) {
-            nvtxRangePush(name);
-        }
-        ~ProfileRange() {
-            nvtxRangePop();
-        }
+        ProfileRange(const char *name);
+        ~ProfileRange();
     };
-
     #define PROFILE_RANGE(name) ProfileRange profileRangeThis(name);
 #else
 #define PROFILE_RANGE(name)

libraries/gpu/src/gpu/Context.cpp

@@ -33,9 +33,11 @@ bool Context::makeProgram(Shader& shader, const Shader::BindingSet& bindings) {
 }
 
 void Context::render(Batch& batch) {
+    PROFILE_RANGE(__FUNCTION__);
     _backend->render(batch);
 }
 
 void Context::syncCache() {
+    PROFILE_RANGE(__FUNCTION__);
     _backend->syncCache();
 }

libraries/render-utils/src/Model.cpp

@@ -417,6 +417,7 @@ void Model::reset() {
 }
 
 bool Model::updateGeometry() {
+    PROFILE_RANGE(__FUNCTION__);
     bool needFullUpdate = false;
 
     bool needToRebuild = false;
@@ -690,6 +691,7 @@ void Model::recalculateMeshPartOffsets() {
 // entity-scripts to call.  I think it would be best to do the picking once-per-frame (in cpu, or gpu if possible)
 // and then the calls use the most recent such result. 
 void Model::recalculateMeshBoxes(bool pickAgainstTriangles) {
+    PROFILE_RANGE(__FUNCTION__);
     bool calculatedMeshTrianglesNeeded = pickAgainstTriangles && !_calculatedMeshTrianglesValid;
 
     if (!_calculatedMeshBoxesValid || calculatedMeshTrianglesNeeded || (!_calculatedMeshPartBoxesValid && pickAgainstTriangles) ) {
@@ -1281,6 +1283,7 @@ Blender::Blender(Model* model, int blendNumber, const QWeakPointer<NetworkGeomet
 }
 
 void Blender::run() {
+    PROFILE_RANGE(__FUNCTION__);
     QVector<glm::vec3> vertices, normals;
     if (!_model.isNull()) {
         int offset = 0;
@@ -1392,6 +1395,7 @@ void Model::snapToRegistrationPoint() {
 }
 
 void Model::simulate(float deltaTime, bool fullUpdate) {
+    PROFILE_RANGE(__FUNCTION__);
     fullUpdate = updateGeometry() || fullUpdate || (_scaleToFit && !_scaledToFit)
                     || (_snapModelToRegistrationPoint && !_snappedToRegistrationPoint);
                     
@@ -1829,6 +1833,7 @@ void Model::setupBatchTransform(gpu::Batch& batch, RenderArgs* args) {
 }
 
 AABox Model::getPartBounds(int meshIndex, int partIndex) {
+
     if (meshIndex < _meshStates.size()) {
         const MeshState& state = _meshStates.at(meshIndex);
         bool isSkinned = state.clusterMatrices.size() > 1;
@@ -1859,6 +1864,7 @@ AABox Model::getPartBounds(int meshIndex, int partIndex) {
 }
 
 void Model::renderPart(RenderArgs* args, int meshIndex, int partIndex, bool translucent) {
+ //   PROFILE_RANGE(__FUNCTION__);
     PerformanceTimer perfTimer("Model::renderPart");
     if (!_readyWhenAdded) {
         return; // bail asap
@@ -1933,7 +1939,9 @@ void Model::renderPart(RenderArgs* args, int meshIndex, int partIndex, bool tran
     pickPrograms(batch, mode, translucentMesh, alphaThreshold, hasLightmap, hasTangents, hasSpecular, isSkinned, wireframe,
                  args, locations);
 
-    updateVisibleJointStates();
+    {
+        updateVisibleJointStates();
+    }
 
     // if our index is ever out of range for either meshes or networkMeshes, then skip it, and set our _meshGroupsKnown
     // to false to rebuild out mesh groups.
@@ -2076,9 +2084,13 @@ void Model::renderPart(RenderArgs* args, int meshIndex, int partIndex, bool tran
         }
     }
     
-    _mutex.lock();
-    qint64 offset = _calculatedMeshPartOffset[QPair<int,int>(meshIndex, partIndex)];
-    _mutex.unlock();
+    qint64 offset;
+    {
+        // FIXME_STUTTER: We should n't have any lock here
+        _mutex.lock();
+        offset = _calculatedMeshPartOffset[QPair<int,int>(meshIndex, partIndex)];
+        _mutex.unlock();
+    }
 
     if (part.quadIndices.size() > 0) {
         batch.drawIndexed(gpu::QUADS, part.quadIndices.size(), offset);

libraries/render/CMakeLists.txt

@@ -9,4 +9,16 @@ add_dependency_external_projects(glm)
 find_package(GLM REQUIRED)
 target_include_directories(${TARGET_NAME} PUBLIC ${GLM_INCLUDE_DIRS})
 
-link_hifi_libraries(shared gpu model)
+link_hifi_libraries(shared gpu model)
+
+if (WIN32)
+  if (USE_NSIGHT)
+    # try to find the Nsight package and add it to the build if we find it
+    find_package(NSIGHT)
+    if (NSIGHT_FOUND)
+      include_directories(${NSIGHT_INCLUDE_DIRS})
+      add_definitions(-DNSIGHT_FOUND)
+      target_link_libraries(${TARGET_NAME} "${NSIGHT_LIBRARIES}")
+    endif ()
+  endif()
+endif (WIN32)

libraries/render/src/render/Scene.cpp

@@ -11,6 +11,7 @@
 #include "Scene.h"
 
 #include <numeric>
+#include "gpu/Batch.h"
 
 using namespace render;
 
@@ -167,6 +168,7 @@ void consolidateChangeQueue(PendingChangesQueue& queue, PendingChanges& singleBa
 }
  
 void Scene::processPendingChangesQueue() {
+    PROFILE_RANGE(__FUNCTION__);
     _changeQueueMutex.lock();
     PendingChanges consolidatedPendingChanges;
     consolidateChangeQueue(_changeQueue, consolidatedPendingChanges);

libraries/render/src/render/drawItemStatus.slv

@@ -22,10 +22,10 @@ uniform vec3 inBoundPos;
 uniform vec3 inBoundDim;
 uniform ivec4 inStatus;
 
-vec3 paintRainbow(float nv) {
-    float v = nv * 5.f;
+vec3 paintRainbow(float normalizedHue) {
+    float v = normalizedHue * 6.f;
     if (v < 0.f) {
-        return vec3(0.f, 0.f, 0.f);
+        return vec3(1.f, 0.f, 0.f);
     } else if (v < 1.f) {
         return vec3(1.f, v, 0.f);
     } else if (v < 2.f) {
@@ -36,8 +36,10 @@ vec3 paintRainbow(float nv) {
         return vec3(0.f, 1.f - (v-3.f), 1.f );
     } else if (v < 5.f) {
         return vec3((v-4.f), 0.f, 1.f );
+    } else if (v < 6.f) {
+        return vec3(1.f, 0.f, 1.f - (v-5.f));
     } else {
-        return vec3(1.f, 1.f, 1.f);
+        return vec3(1.f, 0.f, 0.f);
     }
 }