2026/07/11 by Zhenxiang Ma, Zeyu He, Yuanzhen Zhou +6
Computer Science · Engineering · #3D Shape Modeling and Analysis #3D rendering #Advanced Vision and Imaging #Artificial neural network #Computer Graphics and Visualization Techniques #Deep neural networks #Metadata #Real-time rendering #Rendering (computer graphics) #Scheduling (production processes) #Speedup
paper · pdf · open access · doi:10.1145/3806645.3820072
openalex publication_date 2026/07/11 · openalex created_date 2026/07/12 · openalex updated_date 2026/08/01
We present Odin, a training system for distributed point-based neural rendering (PBNR) that accelerates training by removing unnecessary global barriers. Odin observes that two training steps need synchronization only when they update overlapping visible parameters. It uses lightweight visibility metadata and a two-phase design with ahead-of-time scheduling plus runtime refinement to handle irregular and dynamic behavior. Odin hides 82% of exposed communication and achieves up to 1.89 × throughput speedup on 64 GPUs while preserving baseline accuracy.