[ModelLoader] Hybrid Parallel Model Loader

[juanchuletas]

Description:
Currently, FunGT loads models sequentially via Model::load(). Loading multiple large models (like Pixar assets) can block the main thread and result in noticeable delays. To improve performance and scalability, we should implement a hybrid parallel model loader similar to Blender and modern game engines.

Proposed Solution:

1. Internal parallelism inside Model::load()

   • Use Intel TBB (tbb::parallel_for) to process meshes, materials, and textures within a single model.
   • This speeds up loading large individual models.

2. External model loading manager

   • Create a ModelLoader class or task queue to handle multiple models concurrently.
   • Allows users to enqueue multiple models without manually spawning threads.
   • Provides methods like waitForAll() or progress tracking.

3. GPU synchronization

   • Keep GPU buffer creation on the main thread to ensure thread safety with OpenGL.
   • CPU parsing and data preparation remain parallel.

Benefits:

• Fast loading of large individual models.
• Allows multiple models to load concurrently without user pain.
• Mirrors Blender and game engine best practices.

Optional Enhancements:

• Lazy loading / streaming for very large assets.
• Limit concurrency via TBB task arena to prevent oversubscription.

References / Inspiration:

• Blender: Uses TBB for dependency graph, physics, modifiers, and asset import.
• Game engines (Unreal, Unity, Godot): Separate CPU parse and GPU upload; job/task systems for async model loading.

Suggested Labels:

• enhancement
• performance
• tbb
• model-loading

Suggested Milestone:

• v1.2 (or next performance-focused release)