StellarCreateOptimization

Stellar Create Optimization 1.0.5

Added

• Added comprehensive Mechanical Arm rendering optimizations on Forge 1.20.1 and NeoForge 1.21.1, including conservative motion-envelope culling, stable pose-matrix caching, bulk extended-vertex writes, and persistent GPU rendering for compatible arm parts and cogwheels.
• Added complete spatial caches for Transport Belts. Belt direction, slope, length, controller identity, path endpoints, transported-item bounds, and long-belt visibility groups are now reused until their actual dependencies change.
• Added geometry-state reuse and layered visibility culling for Chain Conveyors, reducing repeated connection reconstruction, distance calculations, LOD decisions, lighting queries, and vertex submissions.
• Added platform-native bulk vertex writers for belt bodies, pulleys, and funnel flaps. Forge uses Embeddium/Oculus layouts, while NeoForge uses Sodium/Iris layouts.
• Added persistent VBO, instanced rendering, and precise affine tongue rendering for Frogports on Forge 1.20.1.
• Added Forge per-instance color support for persistent kinetic VBOs. Overstress and recovery color transitions can reuse static geometry without creating a new VBO for every animation step.

Changed

• Belt render bounds now use topology and resource generations for precise invalidation instead of repeatedly rebuilding stable spatial data.
• Frogport shadow rendering on Forge can use FULL, FAR_SIMPLIFIED, or DISABLED; FULL remains the default.
• Unsupported shaders, vertex formats, dynamic models, third-party subclasses, virtual scenes, and incompatible rendering states continue using Create’s original rendering paths.

Fixed

• Fixed belt visibility groups being incorrectly culled during shader shadow passes by preventing reuse of the main-camera frustum in the shadow stage.
• Fixed stale belt render bounds after resource reloads, topology changes, controller changes, or same-length direction changes.
• Fixed the NeoForge render-bound cache invalidation access path.
• Fixed Forge far-belt and chute fallback texture lookups to use the correct block atlas.
• Fixed possible cache desynchronization between model rebakes and belt render bounds.
• Fixed game crashes caused by the coexistence of lithium mod

Performance

• Package Frogports: 25 FPS after optimization, compared with 0 FPS before optimization — at least approximately 2,400% faster, conservatively calculated using 1 FPS for the stalled baseline.
• Mechanical Arms: 102 FPS after optimization, compared with 6 FPS before optimization — approximately 1,600% faster.
• Mechanical Belts: 33 FPS after this optimization, compared with 21 FPS after the previous optimization — approximately 57% faster.
• Mechanical Pumps: 31 FPS after optimization, compared with 1 FPS before optimization — approximately 3,000% faster.

Stellar Create Optimization 1.0.5

新增

• 为 Forge 1.20.1 和 NeoForge 1.21.1 新增完整的机械动力臂渲染优化，包括保守运动包络剔除、稳定姿态矩阵缓存、扩展顶点批量写入，以及兼容部件和齿轮的持久 GPU 渲染。
• 新增完整的传送带空间缓存。方向、坡度、长度、控制器身份、路径端点、运输物品范围和长传送带可见性分组会持续复用，直到实际依赖发生变化。
• 新增锁链传动轮几何状态复用和分层可见性剔除，减少重复连接重建、距离计算、LOD 判断、光照查询和顶点提交。
• 新增传送带本体、滚轮和漏斗挡片的原生批量顶点写入。Forge 使用 Embeddium/Oculus 布局，NeoForge 使用 Sodium/Iris 布局。
• Forge 1.20.1 新增机械动力蛙港持久 VBO、实例化绘制及精确仿射舌头渲染。
• Forge 新增动力机械持久 VBO 的每实例颜色支持。进入过载和恢复运行时的颜色动画可以继续复用静态几何，不再为每个动画阶段建立新 VBO。

调整

• 传送带渲染范围使用拓扑代次和资源代次精确失效，不再重复重建稳定空间数据。
• Forge 蛙港阴影支持 FULL、FAR_SIMPLIFIED 和 DISABLED；默认仍为完整阴影 FULL。
• 不支持的光影、顶点格式、动态模型、第三方子类、虚拟场景和不兼容渲染状态继续使用 Create 原始渲染路径。

修复

• 修复传送带可见性分组在光影阴影阶段错误复用主相机视锥而被误剔除的问题。
• 修复资源重载、拓扑变化、控制器变化或相同长度下方向变化后，传送带渲染范围没有及时更新的问题。
• 修复 NeoForge 渲染范围缓存失效接口的访问问题。
• 修复 Forge 远距离传送带物品和溜槽回退渲染使用错误纹理图集的问题。
• 修复模型重新烘焙后传送带范围缓存可能与当前资源不同步的问题。
• 修复与锂lithium模组同时存在时导致的游戏崩溃

性能表现

• 机械动力蛙港（Package Frogport）：优化后 25 FPS，优化前显示为 0 FPS；按 1 FPS 保守折算，至少提升约 2,400%。
• 机械动力臂（Mechanical Arm）：优化后 102 FPS，优化前 6 FPS，提升约 1,600%。
• 械传送带（Mechanical Belt）：本次优化后 33 FPS，上一次优化后 21 FPS，提升约 57%。
• 机械动力泵（Mechanical Pump）：优化后 31 FPS，优化前 1 FPS，提升约 3,000%。