Skip to content

Reproducibility.md

Latest commit

 

History

History
85 lines (54 loc) · 2.55 KB

Reproducibility.md

File metadata and controls

85 lines (54 loc) · 2.55 KB

Reproducibility PADS 2022

This documents describes how to reproduce the results discussed in the paper:

"Spatial/Temporal Locality-based Load-sharing in Speculative Discrete Event Simulation on Multi-core Machine"

To be published in ACM SIGSIM PADS 2022

Download

Authors & Contacts

Federica Montesano federica.montesano@alumni.uniroma2.eu Romolo Marotta romolo.marotta@gmail.com Francesco Quaglia francesco.quaglia@uniroma2.it

Requirements for running the tests

  • x86_64 hardware
  • Unix system with gcc toolchain
    • bash, gcc toolchain, make, libnuma-dev or numactl-devel

Requirements for generating figures

  • Python3, matplotlib and seaborn
  • bash

Configuration

USE does not need any manual configuration. Type the following

  • make config

and everything should go smoothly. Such a phase is required to generate necessary information about cache layout and clocks. In particular, the folder include-gen should contain two headers:

  • clock_constant.h
  • hpipe.h

Reproducing the results of the paper

The paper has 4 Figures that can be reproduced:

  • Figure 3;
  • Figure 4;
  • Figure 5;
  • Figure 6 (this figure has two subplots, one for each of the hardware platforms used for the experimental evaluation).

To generate all the figures, type the following:

  • ./run_and_gen_all.sh
  • ./process_data.sh

The Figures will be available in the figures folder.

There is one script for running tests and generating plots: run_and_gen_all.sh. However, run all the tests should require about 8 hours. If you prefer to split the running time, each Figure has associated two scripts for executing the simulations and generating data, figure*-run.sh and figure*-plot.sh respectively.

Notes

  • We uploaded the thread counts for targeting platform A (48 cores) discussed within the paper. If you want to define thread counts for your machine, please modify the thread.conf file as you prefer. In particular:

    • set MAX_THREADS to the number of cores (e.g., MAX_THREADS="<num_of_cores>")
    • set THREAD_list to the list of thread counts you want to evaluate. (e.g., on a 40 CPU-core machine set THREAD_list="1 10 20 30 40")

  • Figure 3 shows one subset of a more extensive exploration of the W parameter. The script generates 6 variants for different values of W. The one corresponding to Figure 3 of the paper is: figure3-lp_4096-ta_0.24-w_1.6.pdf.