Epic: Phase 6 — Mesh Optimization Pipeline

[fernandotonon]

Overview

Complete the game-dev export pipeline with interactive mesh optimization, texture management, and performance analysis tools. Essential for shipping assets to game engines.

Prerequisites

• Phase 1 (mesh buffer manipulation patterns from sub-mesh editing)
• Phase 4 (topology tools provide the mesh simplification foundation)

Scope

1. Interactive Mesh Decimation

• Slider to reduce polygon count with real-time viewport preview
• Target: triangle count, percentage, or vertex budget
• Algorithm: Quadric Error Metrics (QEM) — preserves visual quality
• Preserve UV seams and material boundaries during simplification
• Lock vertices: exclude selected vertices/edges from decimation
• Compare mode: split viewport showing original vs decimated side-by-side
• Export decimated mesh directly

2. Texture Atlas Generation

• Select multiple materials/textures → pack into a single atlas
• Configurable atlas resolution (512, 1024, 2048, 4096)
• UV remapping: automatically update mesh UVs to reference the atlas
• Channel-aware packing (don't mix normal maps with diffuse)
• Preview atlas layout before applying
• Support: diffuse, normal, specular, emissive atlases

3. Draco Mesh Compression

• Export to Draco-compressed glTF (.glb)
• Configurable compression level (1-10) and quantization bits
• Size comparison display (original vs compressed)
• Decompression preview to verify quality

4. Draw Call Analysis

• Overlay showing draw call count, material count, batch count
• Color-code objects by material (same color = same draw call potential)
• Suggestions: "Merge these 5 objects sharing MaterialX to reduce draw calls"
• One-click merge for objects sharing the same material

5. Vertex Cache Optimization

• Reorder index buffer for GPU vertex cache efficiency
• Before/after ACMR (Average Cache Miss Ratio) display
• Automatic optimization on export (optional toggle)
• Algorithm: Tom Forsyth's linear-speed vertex cache optimization

6. Memory & VRAM Usage Display

• Panel showing per-mesh: vertex count, index count, estimated GPU memory
• Per-texture: resolution, format, estimated VRAM
• Total scene budget with configurable target (e.g., "mobile: 50MB VRAM")
• Warning indicators when assets exceed budgets
• Export size estimation for each format

7. Batch Optimization Pipeline (CLI)

• qtmesh optimize model.fbx -o optimized.fbx — apply all optimizations
• qtmesh optimize model.fbx --target-tris 5000 — decimate to target
• qtmesh atlas scene.glb -o atlased.glb --resolution 2048
• qtmesh optimize model.fbx --draco -o model.glb
• JSON output for CI integration

Technical Notes

• QEM decimation: Ogre's MeshLodGenerator uses a variant; extend or replace with a full QEM implementation that preserves UV/bone weights
• Atlas packing: Use a simple shelf/skyline bin-packing algorithm; stb_rect_pack is a lightweight option
• Draco: Link against Google's Draco library (CMake FetchContent or bundled)
• Vertex cache opt: Implement Forsyth's algorithm (~200 lines) or use meshoptimizer library
• Memory estimation: vertex_count * stride + index_count * 2 (16-bit) or 4 (32-bit) for mesh; width * height * bpp for textures

Acceptance Criteria

• Polygon reduction slider with real-time preview preserving UV/material boundaries
• Texture atlas generation packs multiple textures and remaps UVs
• Draco-compressed glTF export with configurable quality
• Draw call analysis overlay with merge suggestions
• Vertex cache optimization with before/after ACMR display
• Memory usage panel shows per-asset and total scene budgets
• CLI subcommands: optimize, atlas work in headless mode
• All operations integrate with undo system where applicable