Metraipolis › archive › dc_081db7edba35

Annotation-Driven Migration of CUDA Programs to Tenstorrent Blackhole — arXiv:2610.02658

Source
arXiv
Original
https://arxiv.org/abs/2610.02658
License (as archived)
CC0-1.0
Archived
2026-10-05T22:52:55.220Z
SHA-256 of the archived text
ebf6197392612a8b11ce37edbdc92f944dd3ab805936982e4d3ecf450006b284
Stable URL of this page
https://metraipolis.com/documento/dc_081db7edba35

How to cite

Annotation-Driven Migration of CUDA Programs to Tenstorrent Blackhole — arXiv:2610.02658 — arXiv, https://arxiv.org/abs/2610.02658 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_081db7edba35 (Metraipolis archive dc_081db7edba35, sha256 ebf6197392612a8b11ce37edbdc92f944dd3ab805936982e4d3ecf450006b284)
BibTeX
@misc{dc081db7edba35,
  title = {Annotation-Driven Migration of CUDA Programs to Tenstorrent Blackhole — arXiv:2610.02658},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.02658},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_081db7edba35; sha256 ebf6197392612a8b11ce37edbdc92f944dd3ab805936982e4d3ecf450006b284},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.02658. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Tenstorrent Blackhole combines distributed local memories, explicit inter-core communication, and Tensix cores decoupling data movement from computation. CUDA offers a substantial HPC software base but leaves physical data placement and scheduling largely implicit. Migrating CUDA HPC kernels to Blackhole requires spatial mapping across cores and per-core coordination of compute and data-movement kernels. We present an MLIR-based compiler deriving data placement, inter-core communication, and tile computation from a statically shaped affine CUDA subset. Declarative annotations express choices not fixed by the source, including compute/DM operation placement and streaming granularity. We evaluate it on Gaussian elimination, five-point stencil, and symmetric rank-k update. Relative to our compiler's default realizations, the best measured configurations achieve a 4.2x speedup for BF16 Gaussian and 2.1-2.2x for FP32 Gaussian and Jacobi. A choice's performance impact can reverse with surrounding policies, precision, and loop schedule, motivating comparison of alternative per-core realizations rather than independent policy selection.

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/2610.02658 — 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 2026-10-05T22:46:25.525Z dalla pagina https://arxiv.org/list/cs.DC/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).