energy-toolkit

Energy Toolkit

energy-toolkit is a Python module designed to facilitate the execution of programs while measuring their energy consumption and execution time. This toolkit provides users with valuable insights into the performance of their code, allowing for better resource management and optimization.

Features

• Execute Python functions or scripts
• Measure energy usage during execution
• Measure execution time
• Return a comprehensive set of metrics upon completion

Requirements

• Python 3.6 or a newer version
• Linux OS
• Intel/AMD CPU offering RAPL registers
• Google Chrome (For Ubuntu see this guide)

Installation

You can install the energy-toolkit module via pip:

pip install energy-toolkit

Command Line Usage

The energy-toolkit also provides a powerful command-line interface (CLI) for measuring energy consumption and validating program configurations. This interface is built with Click, offering a clean and intuitive user experience.

Overview

To use the CLI, simply run:

energy-toolkit [COMMAND] [OPTIONS]

Run the following command to view all available options:

energy-toolkit --help

1. Measure Command

The measure command is used to measure the energy consumption and execution performance of target executables defined in a program configuration file (programs.yaml).

energy-toolkit measure PROGRAMS [OPTIONS]

Arguments

• PROGRAMS Path to a YAML configuration file describing the programs to be measured. The file must include details such as executable paths, arguments, and input data.

Options

Option	Short	Type	Default	Description
--core	-c	Integer	0	CPU core on which the measurement should be performed.
--repetitions	-r	Integer	100	Number of repetitions to average each measurement.
--datapoints	-d	Integer	100	Number of measurement datapoints to collect.
--output	-o	Path	./results	Directory where results will be stored.
--verbose	-v	Flag	-	Enables debug logging and detailed output.
--stats	-s	Flag	-	Prints statistics after measurement completion.

Usage Example

sudo energy-toolkit measure ./programs.yaml -c 2 -r 50 -d 200 -o ./output --stats --verbose

Notes

• This command must be run with elevated privileges (e.g., using sudo) to access energy measurement interfaces like RAPL.
• Results and statistics are saved automatically in the specified output directory.
• Running with --verbose prints detailed runtime logs with timestamps.

Example Output

[12:32:05] Validating programs configuration
[12:32:05] Configuration valid.
[12:32:05] Running analysis on core 2.
[12:32:05] Measurement will record 200 datapoints.
[12:32:05] Each datapoint will be averaged over 50 repetitions.
[12:32:05] Resulting files will be saved at /home/user/output
[12:32:06] Starting measurements! Grab a coffee...
[12:36:44] Measurements finished.
[12:36:44] Saving results!

2. Validate Command

The validate command checks whether a given program configuration file (programs.yaml) is properly formatted and contains all required fields.

energy-toolkit validate PROGRAMS [OPTIONS]

Arguments

• PROGRAMS Path to the YAML file describing the programs to measure.

Options

Option	Short	Type	Default	Description
--verbose	-v	Flag	-	Prints debug messages during validation.

Usage Example

energy-toolkit validate ./programs.yaml --verbose

Example Output

[09:45:12] Validating programs file ./programs.yaml
[09:45:12] Configuration valid.

3. Plot Command

The plot command is used to visualize the results generated by energy-toolkit. It reads the measurement data from a specified results folder and generates plots in either bar or line style.

energy-toolkit plot RESULTS [OPTIONS]

Arguments

• RESULTS Path to the folder containing results.csv files generated by previous measurements. Expects a layout of the following format:

  Example:

  ├── 0
  │   ├── results.csv
  │   └── statistics.csv
  └── 1
      ├── results.csv
      └── statistics.csv
  

Options

Option	Short	Type	Default	Description
--mode	-	Choice	bar	Choose the chart style: bar or line.
--headless	-h	Flag	-	If set, saves the plot as a PDF without displaying it. If not set, it generates an interactive .html file.

Usage Examples

# Show a bar plot interactively
energy-toolkit plot ./results --mode bar

# Save a line plot as PDF without displaying it
energy-toolkit plot ./results --mode line --headless

Example Output

• Interactive plot: creates a .html file in the results directory that alows interaction with the generated data. Open it in your favourite browser.
• Headless mode: creates a PDF file in the current directory with the plot.

4. Example programs.yaml File

Below is a minimal example of a configuration file for defining the executables to be measured:

programs:
  - executeable: ./my_program
    args: ["--input", "data.txt", "--mode", "fast"]
    input: null

  - executeable: /usr/bin/python3
    args: ["script.py", "--option", "value"]
    input: input_data.txt

Each entry defines:

• executeable: Path to the program or script.
• args: Optional list of command-line arguments.
• input: Optional input file or data stream.

Summary

Command	Purpose
measure	Runs the configured programs and records energy consumption data.
validate	Validates program configuration files before measurement.

The CLI provides a flexible and scriptable way to benchmark energy efficiency, making it ideal for automated testing or performance evaluation workflows.

Usage as Library

In addition to the command-line interface, energy-toolkit can also be used directly as a Python library to measure the energy consumption and execution time of programs within your own code or tests.

Example

The following example demonstrates how to use the Energy_Toolkit class to:

• Create an energy measurement toolkit.
• Define two programs to be measured.
• Run the measurements.
• Write results and statistics to disk.

from energy_toolkit.energy_toolkit import Energy_Toolkit, Program

# Create an Energy_Toolkit instance
# Parameters: datapoints=3, repetitions=2
toolkit = Energy_Toolkit(datapoints=3, repetitions=2)

# Define the programs to be measured
program1 = Program("./program_a", ["--mode", "fast"], "")
program2 = Program("./program_b", ["--input", "data.txt"], "")

# Add programs to the toolkit
toolkit.add_program(program1)
toolkit.add_program(program2)

# Run the measurement process
toolkit.measure()

# Write results and statistics to files
toolkit.write_results()
toolkit.write_statistics()

# Optionally print statistics to the console
toolkit.print_statistics()

Explanation

• Energy_Toolkit(datapoints, repetitions) Initializes the toolkit with the number of datapoints to collect and repetitions to average per datapoint.

• Program(executable, args, input) Defines an executable program, its command-line arguments, and optional input.

• add_program() Adds the program to the measurement queue.

• measure() Executes all added programs and records energy and timing data.

• write_results() / write_statistics() Save detailed measurement results and computed statistics to the output directory.

• print_statistics() Prints a summary of the measurement results to the console.

Metrics Returned

After executing a measurement - whether through the command-line interface or the Python library - the energy-toolkit automatically generates a structured set of result files in the specified output directory.

Output Structure

Each measured program is assigned a unique program ID (pid), numbered in ascending order based on how the programs were added to the toolkit or defined in the programs.yaml configuration file.

For each program, a dedicated subdirectory is created:

results/<pid>/

Within each <pid> directory, the following files are generated:

File	Description
results.csv	Contains raw measurement data, including the recorded energy consumption and execution time for each datapoint.
statistics.csv	Contains aggregated metrics derived from the raw data, such as mean, variance, and standard deviation for both energy and time measurements.

Example Directory Layout

results/
├── 0/
│   ├── results.csv
│   └── statistics.csv
└── 1/
    ├── results.csv
    └── statistics.csv

This structure makes it easy to analyze individual program runs or combine results for larger performance and energy efficiency studies.

Contributing

Contributions to the energy-toolkit are welcome! If you encounter any bugs, unexpected behavior, or have feature suggestions, please open an issue on the project’s GitHub repository.

When reporting a bug, include as much detail as possible — such as the command used, environment information, and any error messages — to help us reproduce and resolve the issue quickly.