refactor(渲染): 优化渲染排序和变换更新机制

- 将渲染命令排序键改为基于图层和深度，替代原有的材质纹理组合
- 引入变换版本号机制，避免冗余的世界变换计算
- 移除命令队列批次执行中的冗余全局UBO绑定
- 跳过仅含变换组件的节点渲染，提升渲染效率

include/renderer/command_queue.h

@@ -411,9 +411,8 @@ private:
    * @brief 执行单个批次
    * @param batchIndex 批次索引
    * @param batch 命令批次
-   * @param frameIndex 当前帧索引（用于双缓冲）
    */
-  void executeBatch(uint32_t batchIndex, const CommandBatch &batch, uint32_t frameIndex);
+  void executeBatch(uint32_t batchIndex, const CommandBatch &batch);
 
   /**
    * @brief 分配材质 UBO 空间（使用 CPU 缓冲区）

include/renderer/render_types.h

@@ -102,6 +102,28 @@ struct RenderCommand {
   Color getColor() const { return drawMesh.color; }
 };
 
+constexpr uint32_t SORT_KEY_LAYER_SHIFT = 24;
+constexpr uint32_t SORT_KEY_DEPTH_SHIFT = 8;
+constexpr uint32_t SORT_KEY_LAYER_MASK = 0xFF;
+constexpr uint32_t SORT_KEY_DEPTH_MASK = 0xFFFF;
+constexpr uint32_t SORT_KEY_FLAGS_MASK = 0xFF;
+
+inline uint32_t makeSortKey(uint8_t layer, uint16_t depth, uint8_t flags = 0) {
+  return (static_cast<uint32_t>(layer) << SORT_KEY_LAYER_SHIFT) |
+         (static_cast<uint32_t>(depth) << SORT_KEY_DEPTH_SHIFT) |
+         static_cast<uint32_t>(flags);
+}
+
+inline uint8_t extractSortLayer(uint32_t sortKey) {
+  return static_cast<uint8_t>((sortKey >> SORT_KEY_LAYER_SHIFT) &
+                              SORT_KEY_LAYER_MASK);
+}
+
+inline uint16_t extractSortDepth(uint32_t sortKey) {
+  return static_cast<uint16_t>((sortKey >> SORT_KEY_DEPTH_SHIFT) &
+                               SORT_KEY_DEPTH_MASK);
+}
+
 // 清除标志
 constexpr uint32_t CLEAR_COLOR_FLAG = 1 << 0;
 constexpr uint32_t CLEAR_DEPTH_FLAG = 1 << 1;

include/scene/components/transform_component.h

@@ -1,6 +1,7 @@
 #pragma once
 
 #include <scene/component.h>
+#include <cstdint>
 #include <types/math/transform.h>
 #include <types/math/vec2.h>
 #include <types/math/mat4.h>
@@ -150,6 +151,7 @@ public:
      * @return 是否脏
      */
     bool isDirty() const { return dirty_; }
+    uint64_t getWorldVersion() const;
 
 private:
     /**
@@ -160,6 +162,10 @@ private:
     Transform local_;
     mutable Transform world_;
     mutable bool dirty_ = true;
+    uint64_t localVersion_ = 1;
+    mutable uint64_t cachedLocalVersion_ = 0;
+    mutable uint64_t cachedParentVersion_ = 0;
+    mutable uint64_t worldVersion_ = 0;
     Vec2 anchor_ = Vec2(0.5f, 0.5f);
     Vec2 size_ = Vec2(100.0f, 100.0f);
 };

src/renderer/command_queue.cpp

@@ -366,8 +366,8 @@ void CommandQueue::submitDraw(Ptr<Material> material, Ptr<Mesh> mesh,
 
   // 构建排序键
   uint32_t materialId = getMaterialId(material.get());
-  uint16_t depth = static_cast<uint16_t>((sortKey >> 8) & 0xFFFF);
-  uint8_t layer = static_cast<uint8_t>(sortKey & 0xFF);
+  uint16_t depth = extractSortDepth(sortKey);
+  uint8_t layer = extractSortLayer(sortKey);
   cmd.key = DrawKey::make(materialId, depth, layer);
 
   // 设置管线
@@ -500,8 +500,16 @@ void CommandQueue::execute(uint32_t frameIndex) {
 
   // 执行批次
   const auto executeStart = Clock::now();
+  if (uboManager_) {
+    UniformBuffer *globalUBO = uboManager_->getGlobalUBO(frameIndex);
+    if (globalUBO) {
+      commandList_->setUniformBuffer(0, globalUBO->getRHIBuffer());
+      stats_.bufferBinds++;
+    }
+  }
+
   for (uint32_t i = 0; i < batcher_.getBatchCount(); ++i) {
-    executeBatch(i, batcher_.getBatch(i), frameIndex);
+    executeBatch(i, batcher_.getBatch(i));
   }
 
   // 提交命令到 GPU
@@ -521,8 +529,7 @@ void CommandQueue::execute(uint32_t frameIndex) {
       std::chrono::duration<float, std::milli>(executeEnd - totalStart).count();
 }
 
-void CommandQueue::executeBatch(uint32_t batchIndex, const CommandBatch &batch,
-                                uint32_t frameIndex) {
+void CommandQueue::executeBatch(uint32_t batchIndex, const CommandBatch &batch) {
   if (!commandList_) {
     return;
   }
@@ -535,15 +542,6 @@ void CommandQueue::executeBatch(uint32_t batchIndex, const CommandBatch &batch,
     E2D_WARN("批次没有有效的管线！");
   }
 
-  // 绑定全局 UBO (binding = 0) - 使用双缓冲
-  if (uboManager_) {
-    UniformBuffer *globalUBO = uboManager_->getGlobalUBO(frameIndex);
-    if (globalUBO) {
-      commandList_->setUniformBuffer(0, globalUBO->getRHIBuffer());
-      stats_.bufferBinds++;
-    }
-  }
-
   // 绑定纹理
   if (batch.textureCount > 0) {
     for (uint32_t i = 0; i < batch.textureCount; ++i) {

src/scene/components/sprite_renderer.cpp

@@ -29,9 +29,9 @@ void SpriteRenderer::render() {
   RenderCommand cmd;
   cmd.type = RenderCommandType::DrawMesh;
 
-  uint32_t materialId = material_.index();
-  uint32_t textureId = texture_.index();
-  cmd.sortKey = (materialId << 16) | (textureId & 0xFFFF);
+  int32 layerTag = owner_->getTag();
+  uint8_t layer = layerTag < 0 ? 0 : static_cast<uint8_t>(layerTag & 0xFF);
+  cmd.sortKey = makeSortKey(layer, 0);
 
   // 如果没有指定网格，使用默认的四边形网格
   cmd.drawMesh.mesh = Handle<Mesh>::invalid();  // RendererModule 会使用默认网格

src/scene/components/transform_component.cpp

@@ -70,14 +70,7 @@ Vec2 TransformComponent::getAnchorOffset() const {
 
 void TransformComponent::setDirty() {
     dirty_ = true;
-    // 递归标记所有子节点为脏
-    if (owner_) {
-        for (auto& child : owner_->getChildren()) {
-            if (auto* transform = child->getTransform()) {
-                transform->setDirty();
-            }
-        }
-    }
+    ++localVersion_;
 }
 
 /**
@@ -87,7 +80,18 @@ void TransformComponent::setDirty() {
  * 默认1x1的四边形配合64x64的尺寸会正确显示为64x64像素大小。
  */
 void TransformComponent::updateWorldTransform() const {
-    if (!dirty_) return;
+    uint64_t parentVersion = 0;
+    Transform parentWorld;
+    if (owner_ && owner_->getParent()) {
+        if (auto* parentTransform = owner_->getParent()->getTransform()) {
+            parentWorld = parentTransform->getWorldTransform();
+            parentVersion = parentTransform->getWorldVersion();
+        }
+    }
+    if (!dirty_ && cachedLocalVersion_ == localVersion_ &&
+        cachedParentVersion_ == parentVersion) {
+        return;
+    }
 
     // 计算本地旋转（度数转弧度）
     float rotRad = local_.rot * glm::pi<float>() / 180.0f;
@@ -100,8 +104,6 @@ void TransformComponent::updateWorldTransform() const {
     // 计算世界变换
     if (owner_ && owner_->getParent()) {
         // 有父节点，需要组合变换
-        Transform parentWorld = owner_->getParent()->getWorldTransform();
-
         // 本地旋转后的锚点偏移
         float rotatedOffsetX = anchorOffset.x * c - anchorOffset.y * s;
         float rotatedOffsetY = anchorOffset.x * s + anchorOffset.y * c;
@@ -135,6 +137,9 @@ void TransformComponent::updateWorldTransform() const {
     }
 
     dirty_ = false;
+    cachedLocalVersion_ = localVersion_;
+    cachedParentVersion_ = parentVersion;
+    ++worldVersion_;
 }
 
 Transform TransformComponent::getWorldTransform() const {
@@ -142,6 +147,11 @@ Transform TransformComponent::getWorldTransform() const {
     return world_;
 }
 
+uint64_t TransformComponent::getWorldVersion() const {
+    updateWorldTransform();
+    return worldVersion_;
+}
+
 Mat4 TransformComponent::getWorldMatrix() const {
     updateWorldTransform();
 

src/scene/node.cpp

@@ -175,14 +175,12 @@ void Node::onExit() {
 }
 
 void Node::update(float dt) {
-    // 更新所有组件
     for (auto& comp : components_) {
-        if (comp->isEnabled()) {
+        if (comp.get() != transform_ && comp->isEnabled()) {
             comp->update(dt);
         }
     }
 
-    // 递归更新子节点
     for (auto& child : children_) {
         child->update(dt);
     }
@@ -191,14 +189,16 @@ void Node::update(float dt) {
 void Node::render() {
     if (!visible_) return;
 
-    // 渲染所有组件
+    if (children_.empty() && components_.size() == 1) {
+        return;
+    }
+
     for (auto& comp : components_) {
-        if (comp->isEnabled()) {
+        if (comp.get() != transform_ && comp->isEnabled()) {
             comp->render();
         }
     }
 
-    // 递归渲染子节点
     for (auto& child : children_) {
         child->render();
     }