Planning

Planning

This section is about planning, decisions and trade-offs.

TOC

Graphite Rendering Strategy

Overview

Efficient rendering for math visualizations with large datasets and hot-reload support.

Core Concepts

1. SSBO (Shader Storage Buffers)

What: GPU-side storage for point data
Why: Large buffers, read/write from shaders, better performance than VBO
Use: Store all math function points on GPU

2. LOD (Level of Detail)

What: Adaptive point density based on zoom
Why: Render fewer points when zoomed out, more when zoomed in
Levels:

• Far view: Every 10th point
• Medium: Every 2nd point
• Close: All points

3. Chunking System

What: Split dataset into regions (~1000 points each)
Why: Update only changed chunks, not entire dataset
Benefit: Fast updates during hot-reload

4. Partial Updates

What: Only re-upload modified chunks to GPU
Why: Minimal CPU→GPU transfer
How: Track dirty chunks, use glBufferSubData()

Pipeline

CPU: Generate points → Organize into chunks → Mark dirty chunks GPU: Read from SSBO → Apply LOD → Render visible points Update: Recompute dirty chunks → Upload to GPU (partial)

Trade-offs

SSBO vs VBO:

• SSBO: Larger, flexible, compute-ready (requires OpenGL 4.3+)
• VBO: Simpler, widely supported

Decision: SSBO for scalability

Chunk Size:

• Small: Fine updates, more overhead
• Large: Less overhead, coarse updates

Decision: ~1000 points per chunk

Implementation Phases

1. Fix current bug (only last point renders)
2. Basic SSBO rendering all points
3. Add chunking + partial updates
4. Implement LOD system
5. Test & optimize

Performance Target

• 1M+ points at 60 FPS
• Smooth hot-reload updates
• Interactive zoom/pan

Status

Current: Refactoring (bug fix needed)
Next: SSBO implementation
Priority: Resume after rest