Renderer: Don't sort z=0 Commands

If Command has z==0, then those elements won't be sorted.
Only Z !=0 will be sorted, and it will use `sort` instead of `stable_sort`
for z!=0, since it is faster

CHANGELOG

@@ -17,10 +17,11 @@ cocos2d-x-3.0final ?.? ?
      [FIX]          ControlSlider doesn't support to set selected thumb sprite.
      [FIX]          ControlButton doesn't support to set scale ratio of touchdown state.
      [FIX]          Particles: Crash was triggered if there is not `textureFileName`section in particle plist file.
-     [FIX]          Renderer: QuadCommand::init() does not copy the Quads, it only store a reference making the code faster
-     [FIX]          Renderer: Performance improved in Sprite and SpriteBatchNode (and subclasses) sprites in about 20%
-     [FIX]          Renderer: When note using VAO, call glBufferData() instead of glBufferSubData().
      [FIX]          Renderer: Uses a float as key with only the depth. Viewport, opaque are not needed now
+     [FIX]          Renderer Performance Fix: QuadCommand::init() does not copy the Quads, it only store a reference making the code faster
+     [FIX]          Renderer Performance Fix: Sprite and SpriteBatchNode (and subclasses) has much better performance
+     [FIX]          Renderer Performance Fix: When note using VAO, call glBufferData() instead of glBufferSubData().
+     [FIX]          Renderer Performance Fix: Doesn't sort z=0 elements. It also uses sort() instead of stable_sort() for z!=0.
      [FIX]          Sprite: removed _hasChildren optimization. It uses !_children.empty() now which is super fast as well
      [FIX]          Tests: TestCpp works with CMake on Windows.
      [FIX]          Tests: Sprites Performance Test has 4 new tests

cocos/2d/renderer/CCBatchCommand.cpp

@@ -54,7 +54,7 @@ BatchCommand::~BatchCommand()
 {
 }
 
-void BatchCommand::execute()
+void BatchCommand::execute() const
 {
     // Set material
     _shader->use();

cocos/2d/renderer/CCBatchCommand.h

@@ -45,7 +45,7 @@ public:
 
     void init(float depth, GLuint texutreID, GLProgram* shader, BlendFunc blendType, TextureAtlas *textureAtlas, const kmMat4& modelViewTransform);
 
-    void execute();
+    void execute() const;
 
 protected:
     int32_t _materialID;

cocos/2d/renderer/CCCustomCommand.cpp

@@ -42,7 +42,7 @@ CustomCommand::~CustomCommand()
 
 }
 
-void CustomCommand::execute()
+void CustomCommand::execute() const
 {
     if(func)
     {

cocos/2d/renderer/CCCustomCommand.h

@@ -41,7 +41,7 @@ public:
 
     void init(float depth);
 
-    void execute();
+    void execute() const;
 
     inline bool isTranslucent() { return true; }
     std::function<void()> func;

cocos/2d/renderer/CCGroupCommand.h

@@ -56,12 +56,9 @@ public:
     GroupCommand();
     ~GroupCommand();
     
-public:
-
     void init(float depth);
 
-    inline bool isTranslucent() {return true;}
-    inline int getRenderQueueID() {return _renderQueueID;}
+    inline int getRenderQueueID() const {return _renderQueueID;}
     
 protected:
     int _renderQueueID;

cocos/2d/renderer/CCQuadCommand.cpp

@@ -101,7 +101,7 @@ void QuadCommand::generateMaterialID()
             | (uint32_t)_textureID << 0;
 }
 
-void QuadCommand::useMaterial()
+void QuadCommand::useMaterial() const
 {
     _shader->use();
 

cocos/2d/renderer/CCQuadCommand.h

@@ -44,7 +44,7 @@ public:
     void init(float depth, GLuint texutreID, GLProgram* shader, BlendFunc blendType, V3F_C4B_T2F_Quad* quads, ssize_t quadCount,
               const kmMat4& mv);
 
-    void useMaterial();
+    void useMaterial() const;
 
     //TODO use material to decide if it is translucent
     inline bool isTranslucent() const { return true; }

cocos/2d/renderer/CCRenderCommand.h

@@ -50,10 +50,10 @@ public:
     };
 
     /** Get Render Command Id */
-    inline float getDepth() { return _depth; }
+    inline float getDepth() const { return _depth; }
 
     /** Returns the Command type */
-    inline Type getType() { return _type; }
+    inline Type getType() const { return _type; }
 
 protected:
     RenderCommand();

cocos/2d/renderer/CCRenderer.cpp

@@ -37,9 +37,65 @@
 #include <algorithm>    // for std::stable_sort
 
 NS_CC_BEGIN
-using namespace std;
+
+bool compareRenderCommand(RenderCommand* a, RenderCommand* b)
+{
+    return a->getDepth() < b->getDepth();
+}
+
+void RenderQueue::push_back(RenderCommand* command)
+{
+    float z = command->getDepth();
+    if(z < 0)
+        _queueNegZ.push_back(command);
+    if(z > 0)
+        _queuePosZ.push_back(command);
+    else
+        _queue0.push_back(command);
+}
+
+ssize_t RenderQueue::size() const
+{
+    return _queueNegZ.size() + _queue0.size() + _queuePosZ.size();
+}
+
+void RenderQueue::sort()
+{
+    // Don't sort _queue0, it already comes sorted
+    std::sort(std::begin(_queueNegZ), std::end(_queueNegZ), compareRenderCommand);
+    std::sort(std::begin(_queuePosZ), std::end(_queuePosZ), compareRenderCommand);
+}
+
+const RenderCommand* RenderQueue::operator[](ssize_t index) const
+{
+    if(index < _queueNegZ.size())
+        return _queueNegZ[index];
+
+    index -= _queueNegZ.size();
+
+    if(index < _queue0.size())
+        return _queue0[index];
+
+    index -= _queue0.size();
+
+    if(index < _queuePosZ.size())
+        return _queuePosZ[index];
+
+    CCASSERT(false, "invalid index");
+    return nullptr;
+}
+
+void RenderQueue::clear()
+{
+    _queueNegZ.clear();
+    _queue0.clear();
+    _queuePosZ.clear();
+}
 
 
+//
+//
+//
 #define DEFAULT_RENDER_QUEUE 0
 
 Renderer::Renderer()
@@ -205,11 +261,6 @@ int Renderer::createRenderQueue()
     return (int)_renderGroups.size() - 1;
 }
 
-bool compareRenderCommand(RenderCommand* a, RenderCommand* b)
-{
-    return a->getDepth() < b->getDepth();
-}
-
 void Renderer::render()
 {
     //Uncomment this once everything is rendered by new renderer
@@ -221,9 +272,9 @@ void Renderer::render()
     {
         //Process render commands
         //1. Sort render commands based on ID
-        for (auto it = _renderGroups.begin(); it != _renderGroups.end(); ++it)
+        for (auto &renderqueue : _renderGroups)
         {
-            std::stable_sort((*it).begin(), (*it).end(), compareRenderCommand);
+            renderqueue.sort();
         }
         
         while(!_renderStack.empty())
@@ -244,7 +295,7 @@ void Renderer::render()
                 
                 if(commandType == RenderCommand::Type::QUAD_COMMAND)
                 {
-                    QuadCommand* cmd = static_cast<QuadCommand*>(command);
+                    auto cmd = static_cast<const QuadCommand*>(command);
                     ssize_t cmdQuadCount = cmd->getQuadCount();
                     
                     //Batch quads
@@ -266,19 +317,19 @@ void Renderer::render()
                 else if(commandType == RenderCommand::Type::CUSTOM_COMMAND)
                 {
                     flush();
-                    CustomCommand* cmd = static_cast<CustomCommand*>(command);
+                    auto cmd = static_cast<const CustomCommand*>(command);
                     cmd->execute();
                 }
                 else if(commandType == RenderCommand::Type::BATCH_COMMAND)
                 {
                     flush();
-                    BatchCommand* cmd = static_cast<BatchCommand*>(command);
+                    auto cmd = static_cast<const BatchCommand*>(command);
                     cmd->execute();
                 }
                 else if(commandType == RenderCommand::Type::GROUP_COMMAND)
                 {
                     flush();
-                    GroupCommand* cmd = static_cast<GroupCommand*>(command);
+                    auto cmd = static_cast<const GroupCommand*>(command);
                     
                     _renderStack.top().currentIndex = i + 1;
                     
@@ -413,10 +464,10 @@ void Renderer::drawBatchedQuads()
     //Start drawing verties in batch
     for(ssize_t i = _firstCommand; i <= _lastCommand; i++)
     {
-        RenderCommand* command = _renderGroups[_renderStack.top().renderQueueID][i];
+        auto command = _renderGroups[_renderStack.top().renderQueueID][i];
         if (command->getType() == RenderCommand::Type::QUAD_COMMAND)
         {
-            QuadCommand* cmd = static_cast<QuadCommand*>(command);
+            auto cmd = static_cast<const QuadCommand*>(command);
             if(_lastMaterialID != cmd->getMaterialID())
             {
                 //Draw quads

cocos/2d/renderer/CCRenderer.h

@@ -37,7 +37,26 @@ NS_CC_BEGIN
 
 class EventListenerCustom;
 
-typedef std::vector<RenderCommand*> RenderQueue;
+/** Class that knows how to sort the Commands.
+ Since the commands that have z==0 are "pushed back" in
+ the correct order, the only Commands that need to be sorted, 
+ are the ones that have z <0 and z >0.
+ And that is what this class does.
+*/
+class RenderQueue {
+
+public:
+    void push_back(RenderCommand* command);
+    ssize_t size() const;
+    void sort();
+    const RenderCommand* operator[](ssize_t index) const;
+    void clear();
+
+protected:
+    std::vector<RenderCommand*> _queueNegZ;
+    std::vector<RenderCommand*> _queue0;
+    std::vector<RenderCommand*> _queuePosZ;
+};
 
 struct RenderStackElement
 {