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 data movement in 4D Gaussian Splatting by selecting active records near memory before the GPU fetches full attributes, then rendering adjacent frames together to reuse shared data. It combines programmable processing-in-memory with cache-sized batching and pipeline overlap. Across three methods and six scenes, it achieves a 4.15× geometric-mean speedup, based on calibrated simulation and circuit implementation rather than a complete device.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.