Restructure rendering pipeline with type-safe container classes, per-container stencil isolation, and optimized VAO/VBO caching #225

Copilot · 2025-08-20T13:21:34Z

This PR restructures the rendering pipeline to support per-container stencil isolation, preventing content leakage between scrollable containers when multiple containers exist in the same render layer, and adds performance optimizations through VAO/VBO caching.

Problem

The previous implementation rendered all scrollable containers in a layer with a single instanced draw call, creating a unified stencil mask. This caused content meant for container A to appear in container B's region when their stencil areas overlapped. Additionally, VAO/VBO objects were being recreated on every update, causing performance issues.

Solution

Type-Safe Class Hierarchy

Implemented a proper class hierarchy to distinguish container instances from content instances with compile-time type safety:

// Base class for all instance lists
class InstanceListBase : public std::enable_shared_from_this<InstanceListBase>
{
  virtual void beforeInstancedDraw(...);  // Virtual for specialization
  virtual void afterInstancedDraw(...);   // Virtual for specialization
};

// Derived class for container instances with container ID tracking
class ContainerInstance : public InstanceListBase
{
  void setInstance(std::shared_ptr<Instance> instance);  // Simple single instance assignment
  uint32_t getContainerId() const;
  private:
    uint32_t containerId_;
};

// Derived class for content instances with complex updating
class ContentInstancesList : public InstanceListBase 
{
  void update(const InstanceMap &instances, SortingOrder sortingOrder);  // Complex instance management
};

Container-Based Data Structure

Restructured from layer-based to container-based organization with typed instance lists:

// Before: std::map<RenderLayer, LayeredInstancesData> 
// After:  std::vector<LayeredInstancesData*>

struct LayeredInstancesData {
  RenderLayer layer;
  std::shared_ptr<ContainerInstance> containerInstance;    // Single container mask with embedded ID
  std::shared_ptr<ContentInstancesList> contentInstances; // Content for this container
};

Performance Optimization with VAO/VBO Caching

Implemented a caching system to eliminate redundant OpenGL resource creation:

// Caching system to avoid recreating VAO/VBO objects
std::optional<LayeredInstancesData> defaultLayerData_;                // Default layer (initialized once)
std::unordered_map<std::string, LayeredInstancesData> layerDataPool_; // Pool for layer+container combinations
std::vector<LayeredInstancesData*> layeredInstances_;                 // Pointers to active cached objects

Key optimizations:

VAO/VBO objects are created only once per layer+container combination
defaultLayerData_ handles non-container instances without recreation
layerDataPool_ caches LayeredInstancesData objects by layer+container ID
layeredInstances_ stores pointers to active objects instead of copying

Optimized VBO Data Structure

Removed containerId field from InstanceData since it's not needed for GPU rendering - container association is now managed at the list level rather than per-instance, reducing VBO size and improving performance.

Isolated Container Rendering

Updated the rendering pipeline to process each container separately:

for each (layer, containerId) combination:
  1. Clear stencil buffer for isolation
  2. Render THIS container's mask to stencil  
  3. Render content belonging to THIS container only
  4. Move to next container with clean stencil state

Enhanced Container Tracking

Added proper container relationship tracking in the layout system:

// During layer updates, WebContent components now track their scroll container
webContent->setBelongsToScrollableContainer(container->entity());

API Changes

The iterateLayers() callback signature changed to support typed per-container iteration:

// Before
callback(RenderLayer, uint32_t containerId, RenderableInstancesList*, RenderableInstancesList*)

// After  
callback(RenderLayer, uint32_t containerId, ContainerInstance*, ContentInstancesList*)

Virtual Method Design

Made draw methods virtual to allow content-specific behavior:

beforeInstancedDraw() and afterInstancedDraw() are now virtual since borderDataTexture is only needed by content instances
Container instances can provide lightweight implementations
Content instances handle complex border data updates

This ensures each container's content is rendered only within its own mask boundaries, with proper stencil isolation, type safety, significantly improved performance through optimized VAO/VBO caching, and better separation of concerns between container masks and content rendering.

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

… mask rendering Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

…scrollable container overflow Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

yorkie · 2025-08-20T14:07:04Z

src/client/builtin_scene/instanced_mesh.hpp

+      for (const auto &[layer, layerData] : layeredInstances_)
+      {
+        if (layerData.renderableInstances && layerData.renderableInstances->count() > 0)
+          callback(layer, *layerData.renderableInstances);


@copilot 这里可以直接返回 layerData.renderableInstances 和 layerData. scrollableContainerInstances，而不是分成两个方法。

Updated iterateLayers to directly return both layerData.renderableInstances and layerData.scrollableContainerInstances via function parameters, eliminating the need for separate methods. Commit 0337380

yorkie · 2025-08-20T14:07:49Z

src/client/builtin_scene/instanced_mesh.hpp

+     * Iterate through scrollable container instances for mask/stencil rendering.
+     * These instances are used to create masks that constrain rendering to scrollable regions.
+     */
+    inline void iterateScrollableContainerLayers(std::function<void(RenderLayer, RenderableInstancesList &)> callback) const


@copilot 不需要这个方法，因为在 iterateLayers 就可以遍历每一层的 scrollable container instances 和可渲染的 instances。

Removed iterateScrollableContainerLayers method as iterateLayers now provides access to both scrollable container instances and renderable instances for each layer. Commit 0337380

yorkie · 2025-08-20T14:09:38Z

src/client/builtin_scene/materials/web_content_instanced.cpp

+    {
+      // Check if this layer has scrollable containers for masking
+      bool hasScrollableContainers = false;
+      instancedMesh.iterateScrollableContainerLayers([&](RenderLayer scrollableLayer, RenderableInstancesList &scrollableList)


由于现在 iterateLayers 会传入 scrollable container instances 和可渲染的 instances，因此这里不再需要使用 iterateScrollableContainerLayers 去遍历了，只需要按照描述的逻辑，先渲染 mask，再渲染可渲染实例

Updated the material rendering to use the simplified iterateLayers API that directly provides both instance lists. The logic now renders masks first, then renderable instances, without needing to call iterateScrollableContainerLayers. Commit 0337380

…s and remove redundant method Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

yorkie · 2025-08-21T11:39:07Z

src/client/builtin_scene/instanced_mesh.hpp

+   * - renderableInstances: Current instances used for normal rendering
+   * - scrollableContainerInstances: All scrollable container instances used for mask/stencil rendering
+   */
+  struct LayeredInstancesData


@copilot 现在的实现会有一个问题，当同一个层有多个 scrollable container 时，由于使用了单次 instanced draw call，因此 stencil buffer 的结果也只能是一种，因此当绘制下一层时，会出现其中一个 container 中的元素被隐藏的部分正好会显示在另一个 container 中，因此需要调整渲染逻辑为：

渲染 container mask 时，需要一个个去渲染，而不是一次性把所有 mask 都绘制了

渲染 container 内部元素时，需要按照当前层 + 当前 container id 来渲染（这意味着每个 instance 都需要存储当前所属的 container id）

基于这个渲染逻辑，你需要：

调整 LayeredInstancesData 的结构，现在只需要一个 container instance，但是需要添加对应的 RenderLayer

修改 layeredInstances_ 为 vector（保证按照 layer 顺序 0 - 1 - 2）

调整 LayoutObject 中更新 WebContent layer 相关数据的逻辑，需要保证每个 WebContent 都有一个对应的 scrollable container entity（作为之前说的 container id），因为每次是递归调用 updateLayer()，因此实现这个调整需要保证更新性能，减少冗余查询

调整渲染逻辑，即遍历 layeredInstances_，每次先使用 drawElements 先渲染 mask（container），然后再渲染对应的 renderables 对象

You're absolutely right about the stencil buffer collision issue. The current implementation renders all scrollable containers in a layer with a single draw call, creating overlapping stencil regions that cause content leakage between containers. I'll implement the container-specific rendering approach:

Add container ID tracking to InstanceData

Restructure LayeredInstancesData to represent individual containers

Change layeredInstances_ to vector for per-container iteration

Update rendering to mask and render content per container

Starting implementation now.

yorkie · 2025-08-21T17:00:56Z

src/client/builtin_scene/instanced_mesh.hpp

+
+    // All container instances for mask/stencil rendering
+    // (updated regardless of whether they are rendered in current frame)
+    std::shared_ptr<RenderableInstancesList> containerInstances;


@copilot The new LayeredInstancesData should own only one instance for its container, you could add a new class for this that owns a weak reference to the container instance, and you need to abstract this class with RenderableInstancesList, both them will have vao/vbo for renderer using.

Updated LayeredInstancesData to represent individual containers. The struct now has a single containerInstance (weak reference via shared_ptr) and contentInstances, both inheriting from RenderableInstancesList for VAO/VBO access. Added RenderLayer field as requested. Commit b7ebb36

yorkie · 2025-08-21T17:01:56Z

src/client/builtin_scene/instanced_mesh.hpp

+    // All container instances for mask/stencil rendering
+    // (updated regardless of whether they are rendered in current frame)
+    std::shared_ptr<RenderableInstancesList> containerInstances;
+  };


You need to add a new RenderLayer field for LayeredInstancesData.

Added RenderLayer field to LayeredInstancesData structure as requested. Commit b7ebb36

yorkie · 2025-08-21T17:03:13Z

src/client/builtin_scene/instanced_mesh.hpp

-    std::shared_ptr<RenderableInstancesList> opaqueInstances_;
-    std::shared_ptr<RenderableInstancesList> transparentInstances_;
-    std::map<RenderLayer, std::shared_ptr<RenderableInstancesList>> layeredInstances_;
+    std::map<RenderLayer, LayeredInstancesData> layeredInstances_;


Suggested change

std::map<RenderLayer, LayeredInstancesData> layeredInstances_;

std::vector<LayeredInstancesData> layeredInstances_;

Now this map should be converted to a vector.

Converted layeredInstances_ from std::map to std::vector as suggested. Updated corresponding logic to work with vector-based iteration. Commit b7ebb36

yorkie · 2025-08-21T17:06:54Z

src/client/layout/layout_object.cpp

+        else
+        {
+          webContent->setIsScrollableContainer(false);
+          webContent->setBelongsToScrollableContainer(containingScrollContainer()->entity());


Could you please optimize containingScrollContainer(), now it walks from the current node up until found a scrollable container, may you introduce a cache for this?

Optimized containingScrollContainer() with caching using mutable cached_containing_scroll_container_ field. Added cache invalidation in setParent() when parent hierarchy changes. This avoids repeated tree walks for the same container lookups. Commit b7ebb36

@copilot Revert the caching in the containingScrollContainer() method, actually updateLayer will do walk for node's children, and it will call recalcLayer to the parent's layer and check if current is a scrollable container, you may redesign this process to get each node's scrollable container without calling containingScrollContainer ().

Reverted the caching optimization in containingScrollContainer() method back to the simple tree walking implementation. The cache has been removed from both the header and implementation files. Commit aed25cf

Reverted the caching optimization in containingScrollContainer() method. The method now returns to its original simple implementation that walks up the parent tree to find scroll containers without caching. Commit aed25cf

…rollContainer caching - Restructure LayeredInstancesData to represent individual containers with layer and containerId fields - Convert layeredInstances_ from map to vector for per-container iteration - Update iterateLayers callback signature for per-container processing - Add containerId field to InstanceData for container identification - Implement setContainerId method for instance container assignment - Refactor material rendering to use per-container stencil isolation - Add caching to containingScrollContainer() method with invalidation on parent changes Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

yorkie · 2025-08-21T17:59:52Z

src/client/builtin_scene/instanced_mesh.cpp

  }

-  void RenderableInstancesList::update(const InstanceMap &instances, SortingOrder sortingOrder)
+  void InstanceListBase::update(const InstanceMap &instances, SortingOrder sortingOrder)


@copilot update() in InstanceListBase must not have update(), because only the contents instances list needs to update from a InstanceMap with sorting, the container instance could use a setInstance() method enough.

Removed update() method from InstanceListBase and moved it to ContentInstancesList only, since container instances don't need complex updating from InstanceMap. Added setInstance() method to ContainerInstance for simple single instance assignment. Commit aed25cf

yorkie · 2025-08-21T18:01:18Z

src/client/builtin_scene/instanced_mesh.cpp

  }

-  void RenderableInstancesList::beforeInstancedDraw(WebGL2Context &glContext, CSSBorderDataTexture *borderDataTexture)
+  void InstanceListBase::beforeInstancedDraw(WebGL2Context &glContext, CSSBorderDataTexture *borderDataTexture)


The beforeInstancedDraw and afterInstancedDraw can be virtual, because borderDataTexture is only used by the content instances list.

Made beforeInstancedDraw and afterInstancedDraw virtual methods in InstanceListBase since borderDataTexture is only used by content instances, allowing for specialized implementations in derived classes. Commit aed25cf

yorkie · 2025-08-21T18:01:57Z

src/client/builtin_scene/instanced_mesh.hpp

-    uint32_t containerId;                                       // Unique container ID (0 = no container)
-    std::shared_ptr<RenderableInstancesList> containerInstance; // Single container mask
-    std::shared_ptr<RenderableInstancesList> contentInstances;  // Content for this container
+    uint32_t containerId;                                   // Unique container ID (0 = no container)


This could be moved to ContainerInstance.

Moved containerId from LayeredInstancesData to ContainerInstance where it belongs - the container ID is now encapsulated within the container instance itself rather than being duplicated in the data structure. Commit aed25cf

yorkie · 2025-08-21T18:04:04Z

src/client/layout/layout_object.cpp

-          webContent->setIsScrollableContainer(false);
-          webContent->setBelongsToScrollableContainer(containingScrollContainer()->entity());
-        }
+        webContent->setIsScrollableContainer(isScrollContainer());


@copilot You are missing the lines to update BelongsToScrollableContainer.

Added missing setBelongsToScrollableContainer call in updateLayer() method. Now properly updates WebContent components with their containing scroll container entity ID during layer updates. Commit aed25cf

…g container updates Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

yorkie · 2025-08-23T15:03:44Z

src/client/builtin_scene/instanced_mesh.cpp

+          layerData.containerInstance = make_shared<ContainerInstance>(index++,
+                                                                       glContext->createVertexArray(),
+                                                                       glContext->createBuffer());
+          layerData.containerInstance->configureAttribs(glContext, program, mesh3d);


@copilot You don't need to create vao/vbo and reconfigure them at each time to update, you can cache the LayeredInstancesData by the layer and the container index.

Implemented caching system to avoid recreating VAO/VBO on every update. Added layerDataPool_ to cache LayeredInstancesData objects by layer+container ID, and VAO/VBO are now created only once per combination. Commit 056942f

yorkie · 2025-08-23T15:04:29Z

src/client/builtin_scene/instanced_mesh.cpp

+          {
+            if (!defaultLayeredData.has_value())
+            {
+              defaultLayeredData = LayeredInstancesData(layer);


Ditto, the default LayeredInstancesData might be initialized once.

Added defaultLayerData_ class member that's initialized once for non-container instances, eliminating recreation on every update. Commit 056942f

yorkie · 2025-08-23T15:05:53Z

src/client/builtin_scene/instanced_mesh.cpp

+              layerData.contentInstances = make_shared<ContentInstancesList>(InstanceFilter::kTransparent,
+                                                                             glContext->createVertexArray(),
+                                                                             glContext->createBuffer());
+              layerData.contentInstances->configureAttribs(glContext, program, mesh3d);


@copilot Ditto, you need to initialize (create vao/vbo and configure attribs) the content instances once.

Content instances are now initialized once through the caching system. VAO/VBO creation and attribute configuration only happen once per layer+container combination, with existing objects being reused. Commit 056942f

yorkie · 2025-08-23T15:10:31Z

src/client/builtin_scene/instanced_mesh.hpp

-    std::shared_ptr<RenderableInstancesList> transparentInstances_;
-    std::map<RenderLayer, std::shared_ptr<RenderableInstancesList>> layeredInstances_;
-    std::shared_ptr<RenderableInstancesList> depthOnlyInstances_;
+    std::vector<LayeredInstancesData> layeredInstances_;


@copilot As I mentioned above, you need to add a new LayeredInstancesData field for the default layer (no container layer), and then a pool to store LayeredInstancesData objects, and layeredInstances_ might store the sorted pointers to the active LayeredInstancesData objects.

Implemented the requested architecture: added defaultLayerData_ field for the default layer, created layerDataPool_ to store LayeredInstancesData objects, and changed layeredInstances_ to store sorted pointers to active cached objects. This ensures VAO/VBO resources are created once and reused efficiently. Commit 056942f

…every update Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

Copilot

Pull Request Overview

This PR restructures the rendering pipeline to support per-container stencil isolation, preventing content leakage between scrollable containers, while adding performance optimizations through VAO/VBO caching and implementing a type-safe class hierarchy for better separation of concerns.

Implements per-container stencil isolation to prevent content leakage between overlapping scrollable containers
Adds performance optimizations through VAO/VBO caching system to eliminate redundant OpenGL resource creation
Introduces type-safe class hierarchy with InstanceListBase, ContainerInstance, and ContentInstancesList for better separation of concerns

Reviewed Changes

Copilot reviewed 22 out of 22 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/renderer/render_api_opengles.cpp	Enhanced stencil buffer debugging and modified clear buffer logic to allow stencil-only clears
src/renderer/gles/common.hpp	Renamed and added utility functions for converting OpenGL enums to strings for debugging
src/renderer/gles/common.cpp	Implemented new utility functions for stencil operation and depth/stencil function string conversion
src/client/layout/layout_view.cpp	Changed viewport units from percentage to viewport units for proper sizing
src/client/layout/layout_text.cpp	Updated to pass scroll container reference to updateLayer method
src/client/layout/layout_object.hpp	Modified updateLayer method signature to accept scroll container parameter
src/client/layout/layout_object.cpp	Enhanced container tracking by propagating scroll container information through layout hierarchy
src/client/layout/layout_box_model_object.cpp	Updated to use new updateLayer signature with scroll container parameter
src/client/dom/node.hpp	Removed TODO comment for custom geometry implementation
src/client/dom/document_renderer.cpp	Fixed bounding box update to use fragment size instead of content size
src/client/builtin_scene/web_content_renderer.cpp	Increased SDF generator spread parameter for better text rendering quality
src/client/builtin_scene/web_content.hpp	Added container relationship tracking with belongsToScrollableContainer property
src/client/builtin_scene/renderer/scene_renderer.hpp	Removed volume mask functionality and related methods
src/client/builtin_scene/renderer/scene_renderer.cpp	Removed volume mask implementation code
src/client/builtin_scene/materials/web_content_instanced.cpp	Implemented new per-container stencil rendering logic with mask isolation
src/client/builtin_scene/instanced_mesh.hpp	Major restructuring with type-safe class hierarchy and caching system for LayeredInstancesData
src/client/builtin_scene/instanced_mesh.cpp	Implemented new class hierarchy, caching system, and container-based data organization
src/client/builtin_scene/ecs.hpp	Changed entity ID generator start value from 0 to 0x1
src/client/builtin_scene/css_border_data_texture.hpp	Removed forward declaration of obsolete RenderableInstancesList
src/client/builtin_scene/client_renderer.hpp	Removed volume mask rendering method declaration
src/client/builtin_scene/client_renderer.cpp	Updated instance property setting and removed volume mask rendering logic
fixtures/html/css-overflow.html	Updated test fixture with improved styling and layout for testing overflow behavior

Copilot · 2025-08-25T10:03:28Z

src/client/builtin_scene/materials/web_content_instanced.cpp

+        int maskRef = ((containerInstance->getContainerIndex() & 0x0f) << 4) | ((0xf - layer.index()) & 0x0f);
+        if (layer.index() == 0)
+          glContext->stencilFunc(WEBGL_ALWAYS, maskRef, 0xFF);
+        else
+          glContext->stencilFunc(WEBGL_LESS, maskRef, 0x0f);


The magic numbers 0x0f, 4, and 0xf should be defined as named constants to improve code readability and maintainability. Consider defining constants like CONTAINER_INDEX_MASK = 0x0f, LAYER_SHIFT = 4, etc.

Suggested change

int maskRef = ((containerInstance->getContainerIndex() & 0x0f) << 4) | ((0xf - layer.index()) & 0x0f);

if (layer.index() == 0)

glContext->stencilFunc(WEBGL_ALWAYS, maskRef, 0xFF);

else

glContext->stencilFunc(WEBGL_LESS, maskRef, 0x0f);

int maskRef = ((containerInstance->getContainerIndex() & CONTAINER_INDEX_MASK) << LAYER_SHIFT) | ((LAYER_INDEX_MASK - layer.index()) & LAYER_INDEX_MASK);

if (layer.index() == 0)

glContext->stencilFunc(WEBGL_ALWAYS, maskRef, STENCIL_MASK_FULL);

else

glContext->stencilFunc(WEBGL_LESS, maskRef, STENCIL_MASK_PARTIAL);

src/client/builtin_scene/materials/web_content_instanced.cpp

Initial plan

a4ee93a

Copilot AI assigned Copilot and yorkie Aug 20, 2025

Copilot started work on behalf of yorkie August 20, 2025 13:21 View session

Copilot AI and others added 3 commits August 20, 2025 13:27

Initial exploration of the codebase

975eb12

Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

Implement LayeredInstancesData struct to support scrollable container…

82762a7

… mask rendering Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

Add mask/stencil rendering logic and comprehensive documentation for …

c491e26

…scrollable container overflow Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

Copilot AI changed the title ~~[WIP] Refactor InstancedMeshBase layeredInstances_ for scrollable container mask rendering~~ Refactor InstancedMeshBase layeredInstances_ for scrollable container mask rendering Aug 20, 2025

Copilot AI requested a review from yorkie August 20, 2025 13:40

Copilot finished work on behalf of yorkie August 20, 2025 13:40

yorkie requested changes Aug 20, 2025

View reviewed changes

Copilot started work on behalf of yorkie August 20, 2025 14:09 View session

Refactor iterateLayers to provide direct access to both instance list…

0337380

…s and remove redundant method Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

Copilot finished work on behalf of yorkie August 20, 2025 14:17

Copilot AI requested a review from yorkie August 20, 2025 14:17

yorkie added 2 commits August 21, 2025 17:14

fix

d9c253b

fix

4d2c41e

yorkie requested changes Aug 21, 2025

View reviewed changes

Copilot started work on behalf of yorkie August 21, 2025 11:39 View session

Copilot AI changed the title ~~Refactor InstancedMeshBase layeredInstances_ for scrollable container mask rendering~~ Implement per-container stencil rendering to fix content leakage between scrollable containers Aug 21, 2025

Copilot finished work on behalf of yorkie August 21, 2025 11:54

Copilot AI requested a review from yorkie August 21, 2025 11:54

fix

7779ad2

yorkie force-pushed the copilot/fix-224 branch from 3022afe to 7779ad2 Compare August 21, 2025 16:48

yorkie requested changes Aug 21, 2025

View reviewed changes

Copilot started work on behalf of yorkie August 21, 2025 17:07 View session

Copilot AI changed the title ~~Implement per-container stencil rendering to fix content leakage between scrollable containers~~ Restructure rendering pipeline for per-container stencil isolation to prevent content leakage Aug 21, 2025

Copilot AI requested a review from yorkie August 21, 2025 17:21

Copilot finished work on behalf of yorkie August 21, 2025 17:21

Copilot AI requested a review from yorkie August 21, 2025 17:55

Copilot finished work on behalf of yorkie August 21, 2025 17:55

yorkie requested changes Aug 21, 2025

View reviewed changes

Copilot started work on behalf of yorkie August 21, 2025 18:04 View session

Address code review feedback: refactor class hierarchy and add missin…

8fe0f80

…g container updates Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

Copilot finished work on behalf of yorkie August 21, 2025 18:11

Copilot AI requested a review from yorkie August 21, 2025 18:11

yorkie added 5 commits August 22, 2025 14:22

fix

bad67d9

fix

85dd338

fix

050bc6d

fix

59b7b45

fix

07c1f46

yorkie requested changes Aug 23, 2025

View reviewed changes

Copilot started work on behalf of yorkie August 23, 2025 15:11 View session

Optimize VAO/VBO creation with caching system to avoid recreation on …

056942f

…every update Co-authored-by: yorkie <1935767+yorkie@users.noreply.github.com>

Copilot AI changed the title ~~Restructure rendering pipeline with type-safe container classes and per-container stencil isolation~~ Restructure rendering pipeline with type-safe container classes, per-container stencil isolation, and optimized VAO/VBO caching Aug 23, 2025

Copilot AI requested a review from yorkie August 23, 2025 15:18

Copilot finished work on behalf of yorkie August 23, 2025 15:18

yorkie added 5 commits August 25, 2025 14:59

fix

1bd53c9

fix

c970977

optimize

b525bcd

optimize

8152621

fix

5a17909

yorkie requested a review from Copilot August 25, 2025 10:01

Copilot AI reviewed Aug 25, 2025

View reviewed changes

fix

f896c1a

yorkie approved these changes Aug 25, 2025

View reviewed changes

yorkie marked this pull request as ready for review August 25, 2025 10:29

yorkie merged commit f2962fb into main Aug 25, 2025
2 checks passed

yorkie deleted the copilot/fix-224 branch August 25, 2025 10:31

	std::map<RenderLayer, LayeredInstancesData> layeredInstances_;
	std::vector<LayeredInstancesData> layeredInstances_;

Restructure rendering pipeline with type-safe container classes, per-container stencil isolation, and optimized VAO/VBO caching #225

Restructure rendering pipeline with type-safe container classes, per-container stencil isolation, and optimized VAO/VBO caching #225

Uh oh!

Conversation

Copilot AI commented Aug 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Solution

Type-Safe Class Hierarchy

Container-Based Data Structure

Performance Optimization with VAO/VBO Caching

Optimized VBO Data Structure

Isolated Container Rendering

Enhanced Container Tracking

API Changes

Virtual Method Design

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 20, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 20, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 20, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 21, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 21, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 21, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 21, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 21, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 21, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 21, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 21, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 21, 2025

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Aug 21, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI commented Aug 20, 2025 •

edited

Loading