Thesis scripts

This repository includes a set of scripts for automating my thesis experimental tasks, namely:

• running a modified full-system gem5 version;
• parsing configs, traces and run metrics;
• performing a power/energy impact analysis using McPAT;
• analysing the results and plotting graphs (jupyter notebooks).

Contents

• rungem5-fs.py: a wrapper script for running full-system gem5 simulations with different configurations;
• rungem5-se.py: a similar script for system call emulation mode;
• env/: sample environment path files, with all the paths needed to run gem5 simulations (used by rungem5-fs.py and rungem5-se.py);
• sim-scripts/: collection of gem5 full-system run scripts for a variety of public and custom benchmarks;
• o3-configs/: set of O3CPU ARMv8 configurations used for my thesis experimental work;
• mcpat-templates/: corresponding McPAT templates for these OoO ARMv8 CPU configurations;
• notebooks/: jupyter notebooks for analysing the results and generating plots;
• util/:
  • util/gen_data.py: a script for generating an integer dataset following a random probability distribution;
  • util/analysis_width.py: a script for analysing the bit-width distribution of an integer dataset.
  • util/disk/: utility scripts for (un)mounting and "chrooting" full-system disk images;
  • util/analysis/: collection of scripts for parsing gem5 output metrics into csv files;
  • util/mcpat/: scripts for using McPAT for power estimation
    • util/mcpat/GEM5toMcPAT.py: versatile parser from gem5 stats to McPAT xml input. Originally developed by Daya Khudia (GEM5toMCPAT) and improved by Andreas Brokalakis (cMcPAT)
    • util/mcpat/print_energy.py: script to calculate energy dissipation. By Andreas Brokalakis (cMcPAT).
• experimental_pipeline.svg: overview of how these scripts connect in the experimental pipeline;

Experimental Pipeline

Benchmark Scripts

• Most Splash-2 benchmarks;
• Several PARSEC-3.0 benchmarks;
• Two SPEC2006 benchmarks (libquantum and hmmer);
• Custom set of thesis kernels and mini-apps.

TODO

• Refactor get-stats script to a data pipeline/build system;
• Handle dependencies better: requirements.txt, env...;
• Update python2 scripts to python3.
• Add more thesis kernels sizes (normal, large)