Metraipolis › archive › dc_fa2a62330942
Characterizing CUDA Locality Domains on Blackwell GPUs — arXiv:2611.07535
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.07535
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-06T14:43:27.632Z
- SHA-256 of the archived text
edf84cdcfeebebfea8c5c004793bd85de3da336b80c80c0cab3e6dfa9e632f59- Stable URL of this page
https://metraipolis.com/documento/dc_fa2a62330942
How to cite
Characterizing CUDA Locality Domains on Blackwell GPUs — arXiv:2611.07535 — arXiv (en), https://arxiv.org/abs/2611.07535 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fa2a62330942 (Metraipolis archive dc_fa2a62330942, sha256 edf84cdcfeebebfea8c5c004793bd85de3da336b80c80c0cab3e6dfa9e632f59)
BibTeX
@misc{dcfa2a62330942,
title = {Characterizing CUDA Locality Domains on Blackwell GPUs — arXiv:2611.07535},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.07535},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fa2a62330942; sha256 edf84cdcfeebebfea8c5c004793bd85de3da336b80c80c0cab3e6dfa9e632f59},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.07535. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Low latency is critical to user experience. As NVIDIA Blackwell and Rubin GPUs advance multi-die scaling, data placement and cross-die communication become increasingly important. However, the latency implications and underlying costs of cross-die communication and die-aware execution remain largely unexplored. In this report, we present a comprehensive evaluation of the Locality Domain (LD) feature introduced in CUDA 13.4. Conducted on a single node equipped with eight NVIDIA B300 GPUs, our study examines three aspects: memory management and access, computation, and communication, to identify the sources of latency and, more importantly, explore how LD can mitigate them. Our measurements show that the benefits of LD depend on data access patterns. Through die-aware adaptations and integration with widely used libraries and tools, we achieve speedups of up to 43.5% for NCCL 2.32.3 (AllGather), 111.3% for nvbandwidth (DtoD), and 37% for DeepGEMM (EP). We also identify non-negligible LD-related GPU memory allocation overhead when allocation sizes exceed 32 MB due to barrier and TLB invalidation, we propose a fix to the GPU driver kernel module (submitted to NVIDIA) and show that overhead could be reduced 7x ~ 10x without requiring any user-code changes. Tema degli esploratori: calcolo e infrastrutture (elenco arXiv cs.DC, nuovi lavori del giorno). Categorie arXiv: Distributed, Parallel, and Cluster Computing (cs.DC). Pagina del lavoro: https://arxiv.org/abs/2611.07535 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-11-06T14:19:12.618Z dalla pagina https://arxiv.org/list/cs.DC/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).