AbiDocker

Goals

This project is to build Docker images for ABINIT.

• build-environment/: Dockerfiles for the software environment in which ABINIT can be built and run. Does NOT include ABINIT source code or build process.
• with-abinit/: Dockerfiles with ABINIT installed (built from specific git versions/branches).
• abibuildbotworker/: Dockerfiles for making AbiBuildbot workers for the AbiBuildbot CI system.

Large language models (LLMs) are used to generate Dockerfiles for the above three goals. The agent skills are designed to generate Dockerfiles, where template Dockerfiles are used as example inputs, and instructions are given to generate Dockerfiles for each goal.

Naming Convention

Docker images follow the pattern:

• Build environment: abidocker/abienv:<os>-<compiler>-<mpi>-<mathlib>
• ABINIT installed: abidocker/abinit:<os>-<compiler>-<mpi>-<mathlib>-<version>
• AbiBuildbot worker: abidocker/abibuildbotworker:<os>-<compiler>-<mpi>-<mathlib>

Directory Structure

dockerfiles/
├── build-environment/                    # Software environments for building/running ABINIT
│   ├── ubuntu22.04-gcc-netlib/         # Serial build (no MPI)
│   │   ├── Dockerfile
│   │   ├── install-deps.sh             # Ubuntu 22.04 dependencies
│   │   └── README.md
│   ├── ubuntu22.04-gcc-openmpi-openblas/  # MPI build with OpenBLAS
│   │   ├── Dockerfile
│   │   ├── install-deps.sh             # Ubuntu 22.04 dependencies
│   │   └── README.md
│   ├── ubuntu26.04-gcc-openmpi-openblas/  # MPI build with OpenBLAS
│   │   ├── Dockerfile
│   │   ├── install-deps.sh             # Ubuntu 26.04 dependencies
│   │   └── README.md
│   └── ubuntu26.04-intel-oneapi-mkl/   # MPI build with Intel oneAPI and MKL
│       ├── Dockerfile
│       ├── install-deps.sh             # Ubuntu 26.04 Intel oneAPI dependencies
│       └── README.md
├── with-abinit/                        # ABINIT pre-installed images
│   └── README.md                       # Template and documentation
└── abibuildbotworker/                  # AbiBuildbot CI worker images
    ├── README.md                       # Template and documentation
    ├── ubuntu22.04-gcc-openmpi-openblas/  # MPI worker with OpenBLAS
    ├── ubuntu26.04-gcc-openmpi-openblas/  # MPI worker with OpenBLAS
    └── ubuntu26.04-intel-oneapi-mkl/   # MPI worker with Intel oneAPI and MKL

The List of Dockerfiles

Build Environments

Software environments for building and running ABINIT (without ABINIT pre-installed).

OS	Compiler	MPI	Math Lib	FFTW	LibXC	HDF5	NetCDF	Directory	Description	Status
Ubuntu 22.04	GCC	None	Netlib BLAS/LAPACK	FFTW3	Yes	Yes	Yes (Fortran)	build-environment/ubuntu22.04-gcc-netlib/	Minimal serial build environment with basic libraries	Available
Ubuntu 22.04	GCC	OpenMPI 4.x	OpenBLAS + ScaLAPACK	FFTW3	Yes	Yes	Yes (Fortran)	build-environment/ubuntu22.04-gcc-openmpi-openblas/	MPI parallel build with optimized linear algebra	Available
Ubuntu 26.04	GCC	OpenMPI 5.x	OpenBLAS + ScaLAPACK	FFTW3	Yes	Yes	Yes (Fortran)	build-environment/ubuntu26.04-gcc-openmpi-openblas/	MPI parallel build with current Ubuntu GCC/OpenMPI packages	Available
Ubuntu 26.04	Intel oneAPI	Intel MPI	Intel oneMKL	DFTI (MKL)	Yes	Yes	Yes (Fortran)	build-environment/ubuntu26.04-intel-oneapi-mkl/	MPI parallel build with Intel ifx, Intel MPI, and MKL	Available

Build Environment Details

ubuntu22.04-gcc-netlib

• Image tag: abidocker/abienv:ubuntu22.04-gcc-netlib
• Parallelization: OpenMP only (no MPI)
• Build time: ~20-30 minutes
• Image size: ~500 MB - 1 GB (runtime)

ubuntu22.04-gcc-openmpi-openblas

• Image tag: abidocker/abienv:ubuntu22.04-gcc-openmpi-openblas
• Parallelization: MPI + OpenMP hybrid
• Build time: ~25-40 minutes
• Image size: ~1.5 - 2 GB (runtime)

ubuntu26.04-gcc-openmpi-openblas

• Image tag: abidocker/abienv:ubuntu26.04-gcc-openmpi-openblas
• Parallelization: MPI + OpenMP hybrid
• Build time: ~25-40 minutes
• Image size: ~1.5 - 2 GB (runtime)

ubuntu26.04-intel-oneapi-mkl

• Image tag: abidocker/abienv:ubuntu26.04-intel-oneapi-mkl
• Parallelization: MPI + OpenMP hybrid
• Build time: ~40-60 minutes
• Image size: ~3 - 4 GB (runtime)

With-ABINIT

Docker images with ABINIT pre-installed. Each configuration specifies the build environment used and the ABINIT version.

Configuration Parameters

Parameter	Description	Default
ABINIT_REPO	Git repository URL	https://github.com/abinit/abinit.git
ABINIT_BRANCH	Git branch or tag	master
RUN_TESTS	Run test suite during build	yes
TEST_SUITE	Test suite name (fast, v1, paral, etc.)	Config-dependent
MPI_IMPL	MPI implementation (openmpi, mpich)	Config-dependent

Available Configurations

OS	Compiler	MPI	Math Lib	ABINIT Version	Directory	Description	Status
-	-	-	-	-	with-abinit/	(TODO: Add configurations)	-

AbiBuildbot Workers

Docker images for running AbiBuildbot CI workers. Includes build environment + worker tools (Python, Git, Docker CLI, Buildbot worker runtime).

What's Included

Component	Description
Build Environment	Based on build-environment configurations
Python	Python 3.x for worker operation
Git	Git client for repository operations
Docker	Docker CLI for host Docker socket access when mounted
Worker protocol	Direct Buildbot worker protocol runtime
AbiBuildbot	Provided at runtime through mounted checkouts or configuration

Available Configurations

OS	Compiler	MPI	Math Lib	Directory	Description	Status
Ubuntu 22.04	GCC	OpenMPI 4.x	OpenBLAS + ScaLAPACK	abibuildbotworker/ubuntu22.04-gcc-openmpi-openblas/	MPI worker with Buildbot runtime tooling	Available
Ubuntu 26.04	GCC	OpenMPI 5.x	OpenBLAS + ScaLAPACK	abibuildbotworker/ubuntu26.04-gcc-openmpi-openblas/	MPI worker with current Ubuntu GCC/OpenMPI packages	Available
Ubuntu 26.04	Intel oneAPI	Intel MPI	Intel oneMKL	abibuildbotworker/ubuntu26.04-intel-oneapi-mkl/	MPI worker with Intel ifx, Intel MPI, and MKL	Available

SKILLS

The skills for generating Dockerfiles include instructions, knowledge, and templates for generating Dockerfiles for the software environment.

During development of the Dockerfiles, the generated files may not be perfect and may need improvement. This can be done by:

1. Running the Dockerfiles
2. Checking for errors during build or run
3. Using error messages as feedback to improve the Dockerfiles
4. Iterating with the agent to improve the files

Improved Dockerfiles can then be used as templates for the next generation.

Documentation

• Build Environment Overview
• With-ABINIT Overview
• AbiBuildbot Worker Overview

Quick Start

Build a Build Environment

# Serial build environment
docker build \
  -f dockerfiles/build-environment/ubuntu22.04-gcc-netlib/Dockerfile \
  -t abidocker/abienv:ubuntu22.04-gcc-netlib \
  dockerfiles/build-environment/ubuntu22.04-gcc-netlib

# MPI build environment
docker build \
  -f dockerfiles/build-environment/ubuntu22.04-gcc-openmpi-openblas/Dockerfile \
  -t abidocker/abienv:ubuntu22.04-gcc-openmpi-openblas \
  dockerfiles/build-environment/ubuntu22.04-gcc-openmpi-openblas

Build ABINIT with a Build Environment

# Using the MPI build environment
docker build \
  -f dockerfiles/build-environment/ubuntu22.04-gcc-openmpi-openblas/Dockerfile \
  --build-arg ABINIT_BRANCH=9.10.3 \
  -t abidocker/abinit:ubuntu22.04-gcc-openmpi-openblas-9.10.3 \
  dockerfiles/build-environment/ubuntu22.04-gcc-openmpi-openblas

Run ABINIT

# Serial run
docker run --rm -v $(pwd):/workspace abidocker/abienv:ubuntu22.04-gcc-netlib abinit input.in

# Parallel run
docker run --rm -v $(pwd):/workspace abidocker/abienv:ubuntu22.04-gcc-openmpi-openblas \
  mpirun -np 4 abinit input.in