Bringing Distributed Ordering to Heterogeneous Memory Consistency
IEEE Micro · 2026-09 · 46 (5) · 62-69
DOI: 10.1109/mm.2026.3701401 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Micro, 2026-09. Verfasst von Yanpeng Yu und 2 weiteren Autor:innen. Die Arbeit nennt 12 Literaturangaben. Gefördert von: National Science Foundation.
Abstract (Originalsprache)
Consistency ORdered at Directory (cord) is a novel cache-coherence protocol that applies distributed-systems techniques to tightly coupled heterogeneous processors, unlocking efficient, programmer-friendly shared memory for emerging heterogeneous workloads such as artificial intelligence/machine language. Cache-coherence protocols have historically enforced consistency at the source processor, inflating application latency and interconnect traffic. Central to cord’s efficiency is its reframing of heterogeneous coherence as a distributed-systems problem: Drawing on logical-timestamp algorithms for event ordering, cord assigns consistency-tailored timestamps to write-through stores and orders them at the cache directory—near each store’s completion point at the last-level cache. In our evaluation, cord approaches message-passing efficiency while incurring a fraction of the programming complexity, with negligible storage, area, and power overheads. The result is a concrete demonstration that long-standing architectural challenges in shared-memory designs can be tackled through the lens of distributed systems.
Förderung
- National Science Foundation
Zitieren
APA
Yu, Y., Oswald, N., & Khandelwal, A. (2026). Bringing Distributed Ordering to Heterogeneous Memory Consistency. IEEE Micro, 46(5), 62-69. https://doi.org/10.1109/mm.2026.3701401
Vancouver
Yu Y, Oswald N, Khandelwal A. Bringing Distributed Ordering to Heterogeneous Memory Consistency. IEEE Micro. 2026;46(5):62-69. doi:10.1109/mm.2026.3701401
BibTeX
@article{Yu2026,
author = {Yu, Yanpeng and Oswald, Nicolai and Khandelwal, Anurag},
title = {Bringing Distributed Ordering to Heterogeneous Memory Consistency},
journal = {IEEE Micro},
year = {2026},
volume = {46},
number = {5},
pages = {62--69},
doi = {10.1109/mm.2026.3701401}
}
EUVIMED