Cache Simulators for P-OPT

Repo for P-OPT: Practical Optimal Cache Replacement for Graph Analytics (HPCA 2021)

This repo contains cache simulators for the following cache replacement policies:

• LRU
• DRRIP
• P-OPT
• Transpose-driven Belady's OPT replacement policy (T-OPT)

Repo Organization

• simulators : Pin-based cache simulators for the different replacement policies
• applications : Annotated graph applications designed to work with the simulators
• scripts : Helper scripts to launch experiments (and plot results)

Requirements

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)

Usage Instructions

The following scripts should be run from the scripts directory:

• Step 1: Run download_pin.py (to install Pin) followed by download_and_build_graphs.py (to build all inputs)
• Step 2: Run run_cache_sims.py to start the cache simulations
• Step 3: Run plot_llcmiss_red.py to plot LLC miss reduction from different policies

Adding More Policies/Applications

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.

Citation

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}
}

Feedback

Please create an issue for reporting bugs or providing feedback.