EvoStencils is an open source tool for the grammar-based design of geometric multigrid methods for solving partial differential equations (PDEs) with stencil-based discretizations. With the use of grammar-guided genetic programming (GGGP) it can alter a multigrid cycle's components on each discretization level individually and is, therefore, capable of evolving algorithms whose construction has been previously impossible.
EvoStencils is purely written in Python and utilizes the DEAP framework as genetic programming backend. To automatically generate efficient implementations of a multigrid solver, EvoStencils builds upon the ExaStencils framework, which is capable of generating highly optimized and scalable solver code that utilizes the hardware resources of recent supercomputers.
EvoStencils has been developed at the Chair for System Simulation of Friedrich-Alexander University Erlangen-Nürnberg (FAU). In case you run into any troubles using EvoStencils, please contact [email protected].
- Clone and build the ExaStencils framework
git clone https://i10git.cs.fau.de/exastencils/release exastencils
cd exastencils
sbt compile
sbt assembly
- Install required Python packages
pip install deap sympy mpi4py
- Install LFA Lab (optional) Follow these instructions: https://hrittich.github.io/lfa-lab/install.html
- Clone EvoStencils
git clone https://github.com/jonas-schmitt/evostencils
- Set up the environment
cd evostencils
source ./setup.sh
An example script for running an optimization can be found in the scripts folder.
python scripts/optimize.py
To refer to EvoStencils, please cite the following publications.
@inproceedings{evostencils2022,
author = {Schmitt, Jonas and K\"{o}stler, Harald},
title = {Evolving Generalizable Multigrid-Based Helmholtz Preconditioners with Grammar-Guided Genetic Programming},
year = {2022},
isbn = {9781450392372},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3512290.3528688},
doi = {10.1145/3512290.3528688},
booktitle = {Proceedings of the Genetic and Evolutionary Computation Conference},
pages = {1009–1018},
numpages = {10},
location = {Boston, Massachusetts},
series = {GECCO '22}
}
@article{evostencils2021,
author={Schmitt, Jonas and Kuckuk, Sebastian and K{\"o}stler, Harald},
title={EvoStencils: a grammar-based genetic programming approach for constructing efficient geometric multigrid methods},
journal={Genetic Programming and Evolvable Machines},
year={2021},
month={Sep},
day={03},
issn={1573-7632},
doi={10.1007/s10710-021-09412-w},
url={https://doi.org/10.1007/s10710-021-09412-w}
}
@inproceedings{evostencils2020,
author = {Schmitt, Jonas and Kuckuk, Sebastian and K\"{o}stler, Harald},
title = {Constructing Efficient Multigrid Solvers with Genetic Programming},
year = {2020},
isbn = {9781450371285},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3377930.3389811},
doi = {10.1145/3377930.3389811},
booktitle = {Proceedings of the 2020 Genetic and Evolutionary Computation Conference},
pages = {1012–1020},
numpages = {9},
location = {Canc\'{u}n, Mexico},
series = {GECCO '20}
}
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent