Predictable Shared Last-level Caches

Multicore with a shared LLC, from the ZeroCost-LLC paper (RTAS 2023).

A shared last-level cache (LLC) gives multicores extra caching capacity, but cores contend for it, and in safety-critical systems every memory access needs a worst-case latency (WCL) bound. The usual fix, giving each core its own partition, uses the cache poorly, and the problem gets worse as core counts grow.

This project designs shared LLCs that keep the WCL low while still delivering the performance benefits of sharing.

Sharing partitions. An approach to share LLC partitions among multiple cores while still providing low WCL bounds.
ZeroCost-LLC. A shared inclusive LLC that adds no cost to the WCL compared with having no LLC at all. It eliminates cache-line invalidations across the hierarchy and places no constraints on how cores use the LLC. The journal version adds ZCLLC-OPT, with optimized bus arbitration and overlapping memory requests, lowering the WCL further and improving average performance.
Exclusive hierarchies. PECC is a predictable exclusive coherence mechanism. An exclusive hierarchy eliminates back-invalidations, one of the largest contributors to the WCL of inclusive hierarchies, so PECC lowers the bound and improves average performance.
Isolation with shared data. ParRP partitions the cache’s replacement policy rather than its entries, which isolates each core’s cache space even when cores share data.

See also Predictable Cache Coherence.

Publications

RTAS

ParRP: Enabling Space Isolation in Caches with Shared Data

, Rodolfo Pellizzoni and Hiren Patel

In proceedings of IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS), pp. 1–11, May, 2025

DOI

BibTeX

@inproceedings{wang_25_parp,
  author = {Xinzhe Wang and Rodolfo Pellizzoni and Hiren Patel},
  booktitle = {{proceedings of IEEE Real-Time and Embedded Technology and
               Applications Symposium (RTAS)}},
  pages = {1–-11},
  numpages = {11},
  publisher = {IEEE},
  month = {may},
  doi = {10.1109/RTAS65571.2025.00020},
  title = {{ParRP: Enabling Space Isolation in Caches with Shared Data}},
  year = {2025}
}

TECS

High Performance and Predictable Shared Last-level Cache for Safety-Critical Systems

, Anirudh Mohan Kaushik and Hiren Patel

ACM Transactions on Embedded Computing Systems, pp. 1–30, September, 2024

DOI

BibTeX

@article{wu_24_tecs,
  author = {Zhuanhao Wu and Anirudh Mohan Kaushik and Hiren Patel},
  journal = {ACM Transactions on Embedded Computing Systems},
  volume = {23},
  number = {6},
  pages = {1--30},
  publisher = {ACM},
  month = {sep},
  doi = {10.1145/3687308},
  title = {{High Performance and Predictable Shared Last-level Cache for Safety-Critical Systems}},
  year = {2024}
}

RTAS

Exclusive Hierarchies for Predictable Sharing in Last-level Cache

, , Rodolfo Pellizzoni and Hiren Patel

In proceedings of IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS), pp. 1–11, May, 2024

PDF DOI

BibTeX

@inproceedings{wang_24_excl,
  author = {Xinzhe Wang and Zhuanhao Wu and Rodolfo Pellizzoni and Hiren Patel},
  booktitle = {{proceedings of IEEE Real-Time and Embedded Technology and
               Applications Symposium (RTAS)}},
  pages = {1–-11},
  numpages = {11},
  publisher = {IEEE},
  month = {may},
  doi = {10.1109/RTAS61025.2024.00023},
  title = {{Exclusive Hierarchies for Predictable Sharing in Last-level Cache}},
  year = {2024}
}

RTAS

ZeroCost-LLC: Shared LLCs at No Cost to WCL

, and Hiren Patel

In proceedings of IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS), pp. 249-261, May, 2023

PDF DOI

BibTeX

@inproceedings{wu_23_zcllc_rtas,
  author = {Zhuanhao Wu and Anirudh M. Kaushik and Hiren Patel},
  booktitle = {{proceedings of IEEE Real-Time and Embedded Technology and
               Applications Symposium (RTAS)}},
  pages = {249-261},
  numpages = {11},
  publisher = {IEEE},
  month = {may},
  doi = {10.1109/RTAS58335.2023.00027},
  title = {{ZeroCost-LLC: Shared LLCs at No Cost to WCL}},
  year = {2023}
}

DAC

Predictable Sharing of Last-level Cache Partitions for Multi-core Safety-critical Systems

and Hiren Patel

In proceedings of IEEE Design Automation Conference (DAC), pp. 1273–1278, 2022

ACM

BibTeX

@inproceedings{wu_22_llc_dac,
  author = {Zhuanhao Wu and Hiren Patel},
  booktitle = {{proceedings of IEEE Design Automation Conference (DAC)}},
  pages = {1273–1278},
  numpages = {6},
  publisher = {IEEE},
  title = {{Predictable Sharing of Last-level Cache Partitions for Multi-core
           Safety-critical Systems}},
  year = {2022}
}