__ __ _ _
| \ \ ___ | | ___ _| | _ _
| |/ ._>| |/ . \/ . || | |
|_|_|_|\___.|_|\___/\___|`_. |
<___' --- CXL performance analysis suite
- Time-series analysis code for SPA
- The full set scripts for all workloads in SPA
- Upload workload performance slowdown data
Melody is a framework for systematic characterization and analysis of CXL memory performance. Melody builds on an extensive evaluation spanning hundered of workloads on different CPU platforms with 4 CXL devices. It revealed several interesting insights: workload sensitivity to sub-microsecond CXL latencies (140–410 ns), the first disclosure of CXL tail latencies, CPU tolerance to CXL-induced delays, a novel approach SPA for pinpointing CXL bottlenecks, and CPU prefetcher inefficiencies caused by CXL's high latency.
It contains:
For more detail, please read the paper: Systematic CXL Memory Characterization and Performance Analysis at Scale [ASPLOS'25]
Platforms tested in the paper:
- SKX A: Two Intel Xeon Silver 4114 10-core CPUs at 2.20 GHz, 192GB DDR4 Memory
- SKX B: Eight Intel Xeon Platinum 8180 28-core CPUs at 2.50GHz, 376GB DDR4 Memory
- SPR: Two Intel Xeon Gold 6430 32-core CPUs at 2.10 GHz, 256GB DDR5 Memory
- EMR A: Two Intel Xeon Gold 6530 32-core CPUs at 2.10 GHz, 256GB DDR5 Memory
- EMR B: Two Intel Xeon Platinum 8573C 52-core CPUs at at 2.30 GHz, 3TB DDR5 Memory
@inproceedings{10.1145/3676641.3715987,
author = {Liu, Jinshu and Hadian, Hamid and Wang, Yuyue and Berger, Daniel S. and Nguyen, Marie and Jian, Xun and Noh, Sam H. and Li, Huaicheng},
title = {Systematic CXL Memory Characterization and Performance Analysis at Scale},
year = {2025},
isbn = {9798400710797},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3676641.3715987},
doi = {10.1145/3676641.3715987},
booktitle = {Proceedings of the 30th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2}
}
Maintainer: Jinshu Liu, [email protected]