Reproduce “GSPIM: A Temporal-Locality-Aware 4D Gaussian Splatting Accelerator with Polymorphic PIM and Multi-Frame Rendering Dataflow” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zirui Ma, Zhihua Fan, Haibin Wu, Dongrui Fan, Xiaochun Ye, Wenming Li
GSPIM reduces wasted data transfers in edge rendering of time-varying 3D scenes by selecting active Gaussian records near memory before their full attributes reach the GPU, then processing frame windows to reuse attributes and sorting work. In a GPU–PIM model calibrated to Orin across three methods and six scenes, it achieves 4.15× geometric-mean speedup and 4.99× energy-efficiency improvement, with small image-quality differences.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.