IGAnets

This repository provides Docker build scripts for creating customizable IGANets images. It supports both CPU-only and GPU-enabled configurations, with two build variants:

• Production: Provides a clean installation of IGANets and the configured modules.
• Development: Stops at the end of the build stage, preserving the complete source and build directories for further development and debugging.

Prebuilt Docker images

Pre-built Docker images are available through GitHub Container Registry (GHCR)

Image name	Description
ghcr.io/iganets/iganet:cpu	CPU-only build
ghcr.io/iganets/iganet:cpu-dev	CPU-only developer build
ghcr.io/iganets/iganet:cuda	CUDA-enabled build
ghcr.io/iganets/iganet:cuda-dev	CUDA-enabled developer build

Building the Docker image

Build the CPU-only Docker image from the repository root:

docker build \
  --file Dockerfile.cpu \
  --tag iganet:cpu \
  .

On Linux distributions that use Podman instead of Docker, run the equivalent command:

podman build \
  --file Dockerfile.cpu \
  --tag iganet:cpu \
  .

Here, --file selects the Dockerfile, --tag assigns a name and tag to the resulting image, and . sets the current directory as the build context.

The build process can be customized by passing one or more of the build arguments listed below using the --build-arg option:

Argument	Description	Default
CMAKE_VERSION	CMake version	4.4.0
LIBTORCH_VERSION	LibTorch version	2.13.0
NJOBS	Number of parallel build jobs	1
IGANET_BUILD_CPUONLY	Build CPU-only binaries, even when a GPU driver is detected	OFF
IGANET_BUILD_DOCS	Build IGANet documentation	OFF
IGANET_BUILD_PCH	Use precompiled headers	ON
IGANET_BUILD_TYPE	CMake build type	Release
IGANET_OPTIONAL	Optional IGANet modules to build	None
IGANET_ROOT	Installation prefix	/opt/iganet
IGANET_WITH_GISMO	Enable G+Smo support	OFF
IGANET_WITH_MATPLOT	Enable Matplotlib support	OFF
IGANET_WITH_MPI	Enable MPI support	OFF
IGANET_WITH_OPENMP	Enable OpenMP support	ON

Examples

To build IGANet with the optional examples module enabled, run:

docker build \
  --file Dockerfile.cpu \
  --tag iganet:cpu \
  --build-arg IGANET_OPTIONAL=examples \
  .

To build the optional examples module from a specific branch, run:

docker build \
  --file Dockerfile.cpu \
  --tag iganet:cpu \
  --build-arg IGANET_OPTIONAL=examples[branch] \
  .

To build multiple optional modules, provide their names as a semicolon-separated list:

docker build \
  --file Dockerfile.cpu \
  --tag iganet:cpu \
  --build-arg IGANET_OPTIONAL=module1;module2 \
  .

Optional modules

• Examples examples[main]
• Unit tests unittests[main]
• Performance tests perftests[main]
• Python bindings python[main]
• MATLAB bindings matlab[main]

Development builds

By default, the Docker images contain a clean installation of IGANets and the configured modules. If you want to have a development build, run:

docker build \
  --file Dockerfile.cpu \
  --tag iganet:cpu-dev \
  --target dev \
  .

Running the Docker image

Once the build process completed you can run the image:

docker run -ti --rm --name iganet iganet:cpu

This also works with podman under Redhat Linux:

podman run -ti --rm --name iganet iganet:cpu

You can also run executables directly provided that you enabled the optional module examples:

docker run --rm --name iganet iganet:cpu iganet_poisson

or

podman run --rm --name iganet iganet:cpu iganet_poisson