/brian2-benchmark

brian2 package benchmarking

Primary LanguagePythonMIT LicenseMIT

Benchmarking with brian2

# Build
docker build -t brian2-benchmark .

# Run
docker run -t --rm brian2-benchmark

# To debug
docker run -it --rm brian2-benchmark /bin/bash

# Get results on first machine
docker run -t --rm brian2-benchmark > results/cpu1.txt
# Get results on second machine
docker run -t --rm brian2-benchmark > results/cpu2.txt
# ...
# Format results in Markdown
ruby format_results.rb results/*

Brian-2 benchmarks are taken from the ModelDB record #222725 and associated with the paper Software for Brain Network Simulations: A Comparative Study

Hardware platforms

  1. MacBook Pro 14" M1 Max 10CPU 32GPU (2021, 64GB RAM)
  • M1 Max
    • 10 cores (8 performance, 2 efficient)
    • CPU TDP ~30W
  • VM, Docker for Mac 4.12.0, configured to 10 Cores, 16GB RAM, virtualization.framework
  • macOS Monterey 12.5.1
  1. MacBook Air M2 8CPU 10GPU (2022, 24GB RAM)
  • M2
    • 8 cores (4 performance, 4 efficient)
    • CPU TDP ~15W, passive cooling
  • VM, Docker for Mac 4.12.0, configured to 8 Cores, 8GB RAM, virtualization.framework
  • macOS Monterey 12.5.1
  1. FrameWork i7-1165G7 (2021, 64GB RAM)
  • 11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz
    • 4 cores / 8 threads
    • CPU TDP Turbo 60W
  • Docker version 20.10.14, build a224086
  • Pop!_OS 22.04 LTS

Tests

  1. 5000 Leaky Integrate-and-File neurons in a sparsely connected network.
  2. 500 Hodgkin-Huxly neurons with double-exponential synapses and random connections

Results

Graphs with results