Metraipolis › archive › dc_83be0f1d41b7

Hardware Acceleration of Block-Diffusion LLM for Edge Devices — arXiv:2609.01084

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.01084
License (as archived)
CC0-1.0
Archived
2026-10-17T06:42:09.311Z
SHA-256 of the archived text
92a8a4a751a10883fdc17828c2573bab226123f3ece984dbe014f422ac22204e
Stable URL of this page
https://metraipolis.com/documento/dc_83be0f1d41b7

How to cite

Hardware Acceleration of Block-Diffusion LLM for Edge Devices — arXiv:2609.01084 — arXiv (en), https://arxiv.org/abs/2609.01084 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_83be0f1d41b7 (Metraipolis archive dc_83be0f1d41b7, sha256 92a8a4a751a10883fdc17828c2573bab226123f3ece984dbe014f422ac22204e)
BibTeX
@misc{dc83be0f1d41b7,
  title = {Hardware Acceleration of Block-Diffusion LLM for Edge Devices — arXiv:2609.01084},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.01084},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_83be0f1d41b7; sha256 92a8a4a751a10883fdc17828c2573bab226123f3ece984dbe014f422ac22204e},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.01084. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Single-stream (batch-one) edge inference cannot amortize weight traffic across requests. Full-attention diffusion LLMs recompute the entire sequence at every step; native block diffusion makes completed blocks immutable and exactly cacheable, yet refinement still streams prefix KV and FFN weights. We co-design WIFiV-LPDDR, a wide-I/O LPDDR system for precision-tagged reads, BRQ-KV for a canonical low-rank-plus-INT8-residual prefix with query-dependent per-entry precision, and DAT-FFN for drift-mapped canonical replacement, adjacent-stage-corrected low-bit delta, or cached-state carry while keeping live activations unquantized. Both map to an input-stationary mixed-precision systolic array. For the evaluated 1.5B/7B models on modeled Jetson-class platforms, the full stack provides arithmetic-mean energy-reduction factors of 3.79x/3.96x and arithmetic-mean latency speedups of 2.88x/4.44x at the reported DAT-FFN settings; every corresponding compressed model-benchmark score drops by less than one absolute percentage point from its baseline.

Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-09-02 in cs.AR).
Categorie arXiv: Hardware Architecture (cs.AR).
Pagina del lavoro: https://arxiv.org/abs/2609.01084 — 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 2026-10-17T06:34:40.987Z dalla pagina https://arxiv.org/catchup/cs.AR/2026-09-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).