Reproduce “Uncovering the Memory Hierarchy of Modern NVIDIA GPUs for Accurate Simulation” 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
Ziwen Wang, Jiapeng Zhou, Changxi Liu, Dongsheng Wang, Yun Chen
GemAtlas probes GPU cache behavior with controlled SASS instructions and incorporates the inferred memory hierarchy into AccelSim. Its model captures cache-set routing, replacement, and coherence mechanisms, including a distributed L2 model. Across H20, H100, A800, and RTX 4090, MAPE falls from 31.69% to 22.24%, indicating more accurate cycle predictions, not faster kernels or accuracy for every workload.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.