fdaPDE Docker Environment

This docker image provides an environment for compiling and running the C++ library fdaPDE-cpp or its core.

docker pull aldoclemente/fdapde-docker:latest

Getting started

Cloning the repository

Assume you have cloned fdaPDE-cpp repository into a directory <dir> on your host machine.

Mounting the directory

To test or develop within the Docker container, you need to mount <dir>. Run the following command from the terminal, ensuring your current working directory is <dir>:

 docker run --rm -v $(pwd)/../<dir>:/root/<dir> -ti aldoclemente/fdapde-docker /bin/bash

Notes

• The -v parameter flags the absolute path to <dir> on your host machine to root/<dir> in the container.

• You can modify files in <dir> using your favorite text editor on the host machine while the container is running.

• The container is automatically removed once the session ends (i.e., after you run exit in the terminal).

Running Tests

Once inside the docker container, you can run the library's predefined tests as follows:

cd 
cd <dir>/test/
./run_test.sh

Developing and compiling your own scripts

To compile a custom C++ script inside the container, use the following command:

g++ -o script script.cpp -I/path/to/<dir> -I/path/to/<dir>/fdaPDE/core/ -I/usr/include/eigen3 -O2 -std=c++20 -g -march=native -DFDAPDE_NO_DEBUG

Ensure that every custom script contains #include<Eigen/Dense> along with other necessary headers from the fdaPDE library.

LBFGS++

The Docker image includes LBFGS++, a header-only C++ library implementing both LBFGS and LBFGSB (i.e. the version with box constraints). The library depends only on the Eigen library. To use LBFGS++, compile your C++ source with the appropriate include paths:

g++ -o script script.cpp -I/usr/include/eigen3 -I/usr/include/LBFGSpp/include -O2 -std=c++20 

MPI support

The Docker image also provides MPI support for parallel computing with MPI. To compile a script with MPI support, use the following command:

mpicxx -o output script.cpp -I/path/to/<dir>/ -I/path/to/<dir>/fdaPDE/core/ -I/usr/include/eigen3 -O2 -std=c++20 -g -march=native -DFDAPDE_NO_DEBUG

To execute the compiled script using MPI, run:

mpiexec --allow-run-as-root -np <processes> ./script

Notes

• Replace <processes> with the desired number of processes.

• --allow-run-as-root flag is required to run MPI as root inside the Docker container.

heaptrack support

The Docker image includes the memory profiler heaptrack, which allows you to analyze the memory usage of your executables.

To use heaptrack inside the container while enabling graphical output, run:

xhost +local:docker
docker run --rm -v $(pwd)/../<dir>:/root/<dir> \ 
	-v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=$DISPLAY \ 
	-ti aldoclemente/fdapde-docker /bin/bash

Notes

• The -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=$DISPLAY options allow graphical output.

• Users on Wayland may need alternative setup (e.g., XDG_RUNTIME_DIR and WAYLAND_DISPLAY environment variables).

Once inside the container, compile your C++ script as explained in the previous section. Then, run heaptrack as follows:

heaptrack ./<exec>

This command generates a compressed .zst file containing the profiling data. The GUI will also attempt to launch automatically. If the GUI does not launch immediately or if you prefer to analyze the data later, you can manually open the report using:

heaptrack_gui <exec>.zst

After closing the container, restore X server permissions by running:

xhost -local:docker