Reproduce “DISTWAR: Fast Differentiable Rendering on Raster-based Rendering Pipelines” 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
Sankeerth Durvasula, Adrian Zhao, Fan Chen, Ruofan Liang, Pawan Kumar Sanjaya, Nandita Vijaykumar
DISTWAR reduces atomic updates during gradient computation in raster-based differentiable rendering. It combines contributions from nearby GPU threads when their number meets a tuned threshold, while smaller groups use existing L2 atomic units. Across evaluated 3DGS, NvDiffRec, and Pulsar workloads on RTX 4090, it achieves 2.44× average gradient-computation speedup, but full-pipeline speedup averages 1.41×.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.