Metraipolis › archive › dc_d7150a2c591e

PolyCIM: Improving Data Reuse in Digital CIM Accelerators with Polyhedral-Based Compilation — arXiv:2609.34351

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.34351
License (as archived)
CC0-1.0
Archived
2026-10-07T22:13:00.351Z
SHA-256 of the archived text
47c79dc485b851b5859c13f7746721df8a35f702ad4e0b691c7f344693b20c36
Stable URL of this page
https://metraipolis.com/documento/dc_d7150a2c591e

How to cite

PolyCIM: Improving Data Reuse in Digital CIM Accelerators with Polyhedral-Based Compilation — arXiv:2609.34351 — arXiv (en), https://arxiv.org/abs/2609.34351 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d7150a2c591e (Metraipolis archive dc_d7150a2c591e, sha256 47c79dc485b851b5859c13f7746721df8a35f702ad4e0b691c7f344693b20c36)
BibTeX
@misc{dcd7150a2c591e,
  title = {PolyCIM: Improving Data Reuse in Digital CIM Accelerators with Polyhedral-Based Compilation — arXiv:2609.34351},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.34351},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d7150a2c591e; sha256 47c79dc485b851b5859c13f7746721df8a35f702ad4e0b691c7f344693b20c36},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34351. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Digital Compute-in-Memory (CIM) presents a promising solution for accelerating deep neural networks (DNNs) through the integration of computational logic directly within memory arrays. However, mapping modern DNN operators to CIM accelerators often results in severe array underutilization, due to the strict data reuse constraints imposed by the rigid CIM array structure. We observe that data reuse in modern DNNs forms hyperplane structures often oriented along non-axial directions, rendering them invisible to conventional mapping methods that only exploit axis-aligned reuse. In this work, we propose PolyCIM, a polyhedral-based compilation framework for CIM architectures that systematically exposes and realigns these hyperplanes through affine transformations. PolyCIM provides a unified abstraction capable of efficiently representing both diverse DNN workloads and digital CIM architectures. Through data reuse exposure, computation mapping, and data movement optimization, PolyCIM generates mappings for CIM architectures that achieve superior array utilization and performance. Experimental results show that PolyCIM delivers up to $4\times$ improvement in macro utilization and $3.2\times$ speedup, effectively bridging the gap between modern DNN operators and CIM architectures.

Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AR).
Categorie arXiv: Hardware Architecture (cs.AR).
Pagina del lavoro: https://arxiv.org/abs/2609.34351 — 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 2026-10-07T21:54:31.000Z dalla pagina https://arxiv.org/catchup/cs.AR/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).