Numdiff

Installation and setup

To set up the project create a virtual environment, activate the environment, and install the dependencies.

python3 --version  # Ensure you have at least python 3.7 (May work with earlier versions, but not tested)
python3 -m venv venv
source venv/bin/activate  # Or the equivalent for your shell
pip3 install -r requirements.txt

For how to activate the environment on different shells see the venv docs.

You may need to install swig to be able to build scikit-umfpack. Alternatively you may remove it from requirements.txt, as it isn't used explicitly, but may be used by the solvers in scipy.sparse.linalg if present. Do note that during testing I had some issues with memory usage using the default solvers from scipy, while the umfpack solvers didn't seem to have that issue.

Configuring the project

numdiff/settings.py contains some variables that are read by the rest of the project that may be changed to alter the behaviour of the code. These are documented in the file itself. The most important of these may be FINE_PARAMETERS which toggles whether the tasks will use parameters that are easy to run (< 1 min), or that give the most detailed plots (>> 5 min).

Running the code

Running python main.py while in the virtual environment prompts you to enter the tasks you want to run. To run several tasks at once you may type a common prefix of the tasks you want (e.g. 1 to run all tasks related to task 1), and you may specify several prefixes space separated. To run the code non-interactively you may specify these parameters as commandline arguments instead of at the prompt.

Part 2 - task 3: The advection diffusion equation

Note: Part 2 - task 3 is in this project renamed to task 6 to fit the scheme.

To run the code comparison task (part 2 - task 3c) run task 6c_runtime e.g. python main.py 6c. This will generate a plot in the images folder with two figures: l2 error vs degrees of freedom and runtime vs degrees of freedom.

Editing the tasks

The tasks are loaded into main.py from the tasks directory. Each major task (1, 2, ...) has its own file in the tasks directory, while the functions for the different subtasks lie inside those files. The taskname used in main.py should, for the most part, coincide with the names of the functions.

To edit the behaviour or the parameters of any task, simply edit the corresponding function.

Amund211 / numdiff