About: Study dynamics of inertial active matter in potential flow (irrotational, incompressible).
This repository contains the molecular dynamics like engine written in C++ for our Journal of Fluid Mechanics paper.
Note that the current release is v1.0.3 and version 2 development is in progress for finitely-many bodies in potential flow.
Authors: Alec Glisman
GitHub actions:
Third-party services:
Code is documented using Doxygen and can be found in the docs directory.
Open the documentation locally via opening the main index page.
The project can be built and run inside a Docker container.
The docker-compose.yml file will specify all build options and Dockerfile builds the image using a Ubuntu 20.04 base image.
Data I/O is handled automatically with an attached volume.
Commands:
docker-compose --env-file [docker_environment_file] up --detach # build
docker-compose down # tear down
.vscode: Files relevant for developing the project in VSCode.
include: External dependencies required for the project.
Further information can be found at the end of this readme.
input: Data files that are used in Perl scripts to modify parameters of interest across a range of simulations.
profile: Scripts to profile code on various platforms and find performance improvement areas.
python: Python scripts to generate GSD files to input to simulation as well as analyze GSD files output from simulation.
requirements: Files and scripts relevant for loading C++, Perl, and Python dependencies.
scripts: Perl scripts to run many simulations simultaneously and analyze the results.
src: C++ code for the simulation.
test: C++ code containing unit test for simulation.
All data is input and output from simulation using the HOOMD GSD format. The schema is well-documented.
I will be making a few modifications and make extensive use of the log section of the schema to save additional data.
Many variables are stored as floats, but I output a number of variables as doubles in the logs section for accuracy in further computations.
Frame 0 is created using a Python script and passed into the C++ simulation. The simulation then updates parameters (such as kinematics), and outputs the "true" initial frame as Frame 1. For this reason, there could be issues when loading data from frame 0 and any data that is not an input parameter should not be used for further work.
The simulation system can be readily adapted for other configurations and constraints.
The code that must be changed inside the C++ framework is tagged with comments of the form @review_swimmer and are found in the Doxygen documentation page named "Altering Swimmer Design List".
The relevant classes to modify are the SystemData and GSDUtil.
Separate Python scripts for GSD initialization and numerical analysis must also be generated.
These can be found in python/initial_configurations and python/analysis directories, respectively.
Further information found in requirements directory readme.
- ZSH v5.8
- Intel oneAPI v2021.1.1
- CUDA v11.3
- CMake v3.16.3
- GCC v11.1.0
- Python (Miniconda3) v3.9.5
- Perl v5.30.0
Manifest is found at base of repository and command vcpkg/./vcpkg install will manage installation.
Note that the vcpkg Intel MKL will only verify MKL is installed and not install it on its own.
Therefore, oneAPI must still be downloaded and installed manually.
- Intel MKL (oneAPI v2021.1.1): Optimized mathematical instructions
- Intel TBB (oneAPI v2021.1.1): Parallel computing on multi-core processors
- OpenMP (4.5): Multi-platform shared-memory multiprocessing programming