Repo for P-OPT: Practical Optimal Cache Replacement for Graph Analytics (HPCA 2021)
This repo contains cache simulators for the following cache replacement policies:
simulators
: Pin-based cache simulators for the different replacement policiesapplications
: Annotated graph applications designed to work with the simulatorsscripts
: Helper scripts to launch experiments (and plot results)The cache simulators are pintools designed to work with Pin-2.14 (Download script provided in scripts/download_pin.py
)
The pintools require an older version of g++ (v4.9)
All the applications and scripts were tested on Debian Stretch with g++-6.3.0 and python (v3.5.3)
The following scripts should be run from the scripts
directory:
download_pin.py
(to install Pin) followed by download_and_build_graphs.py
(to build all inputs) run_cache_sims.py
to start the cache simulationsplot_llcmiss_red.py
to plot LLC miss reduction from different policies simulators/README.md
contains information on defining new LLC replacement policies and the applications/README.md
directory provides information for creating new applications to evaluate the policies on.
If this work was useful to you, please consider citing our work:
@inproceedings{popt-hpca21,
title={P-OPT: Practical Optimal Cache Replacement for Graph Analytics},
author={Balaji, Vignesh and Crago, Neal and Jaleel, Aamer and Lucia, Brandon},
booktitle={2021 IEEE International Symposium on High-Performance Computer Architecture (HPCA)},
pages={668--681},
year={2021},
organization={IEEE}
}
Please create an issue for reporting bugs or providing feedback.