PyPDEVS-and-JADEVS-DEVStone-benchmarks

devstone-comparative

Here you can find several DEVStone implementations for comparing the main DEVS engines.

Adapted to benchmark Dynamic Structure DEVS

Running the pypdevs benchmarks

First select a simulator:

• classic
• parallel
• dynamic
  • classic
  • number of models to generate per cycle
• minimal

Shared parameters:

• path: where the results will be stored
• filename:
• append: will append the data to the output file instead of overwriting it
• max width/depth: max model width/depth
• width/depth step size:
  • determines how fast the width/depth increases to the max specified value
  • the simulation will be run for every width and depth combination
  • the step size must be a divider of the max width/depth
  • if not provided the model will run only the max width/depth
• repetitions: number of times the exact same simulation will be repeated
• output
• memory: Measures the memory usage of a simulation.

• interval: The interval at which measurements are taken.

• model structure: each simulator has its own set of models

  • LI, HI, HO, HOmod (classic, parallel, dynamic, minimal)
  • HI2LI, LI2HI, dLI, dHI (dynamic)

For each simulation the parameters and the creation, setup and simulation time are saved to a csv file.

plotting

Measuring Memory

Run the following command, the width and depth step size parameters should be left out, to measure the memory usage of a single simulation, for the same reason the repetitions should be set to one.

uv run testPyPDEVS.py -md 100 -mw 100 -r 1 -O dynamic -ms dLI & PID=$!; python -m procpath record -i 0.1 -f stat -d output.sqlite --stop-without-result "$..children[?(@.stat.pid == $PID)]"

procpath plot -d output.sqlite -f out.svg -q rss