perf: optimize instanced buffer batching

[llsansun]

Re: #

Changelog

• Optimized instanced buffer batching by caching compatible instance groups to avoid repeatedly scanning all existing instance groups.
• Reduced instance buffer upload size by uploading only the valid instance data range instead of the whole allocated buffer.
• Refactored instance creation and append logic to make the instanced buffer update path clearer and easier to maintain.

Performance

This change optimizes the existing instancing path without changing when instancing is enabled. Materials still need to enable USE_INSTANCING explicitly.

Before this change, each incoming instanced model searched through all existing instance groups linearly to find a compatible group. As the number of instance groups increased, the CPU cost of InstancedBuffer.merge() grew accordingly.

After this change, compatible instance groups are cached by a key generated from the input assembler, lighting map, reflection probe textures, reflection probe type, and instance attribute stride. This allows the merge path to quickly find candidate groups and only scan groups that share the same compatibility key.

In addition, buffer uploads now use the valid instance data size, instance.count * instance.stride, instead of uploading the whole allocated backing buffer. This avoids uploading unused capacity when the instance buffer has grown but is not fully occupied.

Before profiling result:

After profiling result:

---

Continuous Integration

This pull request:

• needs automatic test cases check.

  Manual trigger with @cocos-robot run test cases afterward.

• does not change any runtime related code or build configuration

  If any reviewer thinks the CI checks are needed, please uncheck this option, then close and reopen the issue.

---

Compatibility Check

This pull request:

• changes public API, and have ensured backward compatibility with deprecated features.
• affects platform compatibility, e.g. system version, browser version, platform sdk version, platform toolchain, language version, hardware compatibility etc.
• affects file structure of the build package or build configuration which requires user project upgrade.
• introduces breaking changes, please list all changes, affected features and the scope of violation.

Test Plan

• Verified a scene with many mesh renderers using the same mesh and material with USE_INSTANCING enabled.
• Verified that all instances render correctly before and after this change.
• Verified that non-instancing materials are not affected.
• Verified that instances with different IA, lighting map, reflection probe textures, reflection probe type, or instance attribute stride are not merged together.
• Verified that exceeding one instanced buffer capacity creates another compatible instance group correctly.
• Verified that partial buffer upload works correctly after the buffer grows and the valid instance count becomes smaller in later frames.
• Compared Chrome DevTools CPU profiles before and after this change and confirmed that InstancedBuffer.merge() cost is reduced.

[github-actions]

Code Size Check Report

Wechat (WASM)	Before	After	Diff
2D Empty (legacy pipeline)	1015126 bytes	1015084 bytes	👍 -42 bytes
2D All (legacy pipeline)	2682844 bytes	2682802 bytes	👍 -42 bytes
2D All (new pipeline)	2774681 bytes	2774639 bytes	👍 -42 bytes
(2D + 3D) All	10032007 bytes	10031965 bytes	👍 -42 bytes

Web (WASM + ASMJS)	Before	After	Diff
(2D + 3D) All	16868425 bytes	16868383 bytes	👍 -42 bytes

Interface Check Report

This pull request does not change any public interfaces !

[llsansun]

@cocos-robot run test cases

[github-actions]

@llsansun, Please check the result of run test cases:

• Test Platform: PR-Test
• Editor Version: v4.0.0
• Task Result: NA
• Task URL: https://cctest.cocos.org/#/reportview/PR-TEST/5625a9f5-2dc6-413a-b7e4-7f3059f506a6/-1
• GitHub Action: https://github.com/cocos/cocos4/actions/runs/31770310537

Task Details

[github-actions]

@llsansun, Please check the result of run test cases:

• Test Platform: PR-Test
• Editor Version: v4.0.0
• Task Result: PASS
• Task URL: https://cctest.cocos.org/#/reportview/PR-TEST/75900c11-1b60-4976-800f-1a6b4b6328b4/-1
• GitHub Action: https://github.com/cocos/cocos4/actions/runs/31770310537

Task Details

[llsansun]

@cocos-robot run test cases

[github-actions]

@llsansun, Please check the result of run test cases:

• Test Platform: PR-Test
• Editor Version: v4.0.0
• Task Result: NA
• Task URL: https://cctest.cocos.org/#/reportview/PR-TEST/55796739-f279-44d8-a89a-1128abc745e5/-1
• GitHub Action: https://github.com/cocos/cocos4/actions/runs/31778976224

Task Details

[github-actions]

@llsansun, Please check the result of run test cases:

• Test Platform: PR-Test
• Editor Version: v4.0.0
• Task Result: PASS
• Task URL: https://cctest.cocos.org/#/reportview/PR-TEST/06f89bce-9c8a-4370-ad6d-82665cb82546/-1
• GitHub Action: https://github.com/cocos/cocos4/actions/runs/31778976224

Task Details