GPU-Driven Pipeline & Automatic Instancing
How Helio's GPU-driven rendering pipeline works end to end — from CPU object submission to automatic draw call batching, frustum culling, and multi_draw_indexed_indirect — with zero per-object CPU iteration
Updated March 23, 2026•30 min read•
gpu-driveninstancingindirect-draw
Related Documentation
Indirect Dispatch Pass
GPU frustum culling and DrawIndexedIndirect command generation — the compute pass that builds the complete indirect draw buffer each frame with zero CPU iteration
Virtual Geometry Pass
GPU-driven per-meshlet visibility culling and G-buffer draw — Helio's scalable geometry pipeline for scenes with tens of millions of triangles
Flush & Dirty Tracking
How Helio's Scene::flush() orchestrates GPU uploads — dirty flags, rebuild_instance_buffers, O(1) steady-state guarantees, and the full per-frame data flow