This repository provides two Docker images for running your LCM and eCAL Python-based benchmarking scripts:

1. debian-arm64

   • Base: Debian (arm64)
   • Builds eCAL from source
   • Python + LCM installed via pip

2. ubuntu-x86_64

   • Base: Ubuntu (x86_64)
   • Installs eCAL from PPA
   • Python + LCM installed via pip

Each image bundles your benchmarking scripts and allows you to run:

• run-publisher-benchmark.bash
• run-subscriber-benchmark.bash

• Docker (>= 20.10)
• Git (to clone this repo)

Make the generator script executable and run it from the project root:

chmod +x generate-docker-image.bash
./generate-docker-image.bash

This will detect your host architecture and build both images:

• bench-debian-arm64
• bench-ubuntu-x86_64

You can also force a specific architecture:

ARCH=arm64 ./generate-docker-image.bash   # only build Debian-arm64
ARCH=amd64 ./generate-docker-image.bash   # only build Ubuntu-x86_64

Mount your workspace into the container and execute the scripts:

docker run --rm -it --privileged --network=host --ipc=host \
  -v "$(pwd)":/workspace \
  bench-ubuntu-x86_64 \
  bash -lc "cd /workspace && ./run-publisher-benchmark.bash"

In a separate terminal, start the subscriber:

docker run --rm -it --privileged --network=host --ipc=host \
  -v "$(pwd)":/workspace \
  bench-ubuntu-x86_64 \
  bash -lc "cd /workspace && ./run-subscriber-benchmark.bash"

Replace bench-debian-arm64 with bench-ubuntu-x86_64 if desired.

Both scripts will write their YAML reports into the results/ directory on your host.

• Dockerfile.debian_arm64 – builds eCAL from source on Debian/arm64
• Dockerfile.ubuntu_x86_64 – installs eCAL from PPA on Ubuntu/amd64
• generate-docker-image.bash – builds images based on host arch
• run-publisher-benchmark.bash – executes publisher script across message sizes
• run-subscriber-benchmark.bash – executes subscriber script
• benchmark_publisher.py – Python publisher benchmark
• benchmark_subscriber.py – Python subscriber benchmark
• benchmark.py – common utilities (serialization, stats)
• bench_pb2.py – Protobuf definitions
• lcmtypes/bench_t.lcm – LCM type definitions

• The first message in each run is ignored in saved reports to avoid skew from startup overhead.
• Scripts use KiB (1024 bytes) units for message sizes.
• Ensure clocks are synchronized (NTP/PTP) if comparing one-way latencies across machines.