GauXC Copyright (c) 2020-2023, The Regents of the University of California, through Lawrence Berkeley National Laboratory (subject to receipt of any required approvals from the U.S. Dept. of Energy). All rights reserved.
If you have questions about your rights to use or distribute this software, please contact Berkeley Lab's Intellectual Property Office at IPO@lbl.gov.
NOTICE. This Software was developed under funding from the U.S. Department of Energy and the U.S. Government consequently retains certain rights. As such, the U.S. Government has been granted for itself and others acting on its behalf a paid-up, nonexclusive, irrevocable, worldwide license in the Software to reproduce, distribute copies to the public, prepare derivative works, and perform publicly and display publicly, and to permit others to do so.
GauXC is a modern, modular C++ library for the evaluation of quantities related to the exchange-correlation (XC) and exact-exchange (K) energy (e.g. potential, etc) in the Gaussian basis set discretization of Kohn-Sham density function theory (KS-DFT). GauXC provides efficient, scalable distributed memory XC and K integrators for both CPU and accelerator-based (GPU) architectures. Currently, GPU support is provided through the CUDA and HIP frameworks to target NVIDIA and AMD GPUs, respectively. Evaluation of the XC functional CPU/accelerator architectures is provided by the ExchCXX library. Quadratures are generated by the IntegratorXX.
GauXC is a work in progress. Its development has been funded by the U.S. Department of Energy Exascale Computing Project (NWChemEx).
- Provide a stable, portable and high-performance implementation of numerical integrators optimized for the evaluation of XC and K related quantities in Gaussian basis set KS-DFT on CPU and accelerator based architectures.
- Develop a modern, modular, extensible C++ software infrastructure which allows for flexible and agile development in the field of KS-DFT.
- CMake (3.20+)
- BLAS (for CPU integrators)
- ExchCXX
- IntegratorXX
- Gau2Grid (pregenerated source packaged with GauXC)
- MPI (Optional)
- OpenMP (CPU parallelism, Optional)
- Cereal (Optional)
- HDF5 (Optional)
- Eigen3 (Testing Only)
- CUDA/cuBLAS (Required only if CUDA enabled)
- HIP/ROCm (Required only if HIP enabled)
- MAGMA (Optional if CUDA/HIP enabled)
Primary Developer and Maintainer: David Williams-Young - LBNL (dbwy at lbl dot gov)
GauXC has received major contributions from the following developers (in no particular order):
- Thom Popovici (LBNL) - Optimized sn-K kernels for CPU and GPU architectures
- Teri Lambros (UW) - Unrestricted DFT
We have also receieved significant support from industry collaborators:
- David Clark (NVIDIA) - Optimization of critical kernels for NVIDIA architectures
- Damon McDougall (AMD) - Optimization of critical kernels for AMDGPU architectures
Please cite the following publications if GauXC was used in your publication:
% Distributed Memory Seminumerical Exact Exchange implementation
@article{williams2023distributed,
title = {Distributed memory, GPU accelerated Fock construction for hybrid, Gaussian basis density functional theory},
author = {Williams-Young, David B. and Asadchev, Andrey and Popovici, Doru Thom and Clark, David and Waldrop, Jonathan and
Windus, Theresa L. and Valeev, Edward F. and de Jong, Wibe A.},
journal = {The Journal of Chemical Physics},
volume = {158},
number = {23},
pages = {234104},
year = {2023},
doi = {10.1063/5.0151070},
url = {https://doi.org/10.1063/5.0151070}
}
% Performance Portability (HIP/SYCL implementations)
@article{williams2021achieving,
title={Achieving performance portability in Gaussian basis set density functional
theory on accelerator based architectures in NWChemEx},
author={Williams-Young, David B and Bagusetty, Abhishek and de Jong, Wibe A and
Doerfler, Douglas and van Dam, Hubertus JJ and V{\'a}zquez-Mayagoitia, {\'A}lvaro and
Windus, Theresa L and Yang, Chao},
journal={Parallel Computing},
volume={108},
pages={102829},
year={2021},
doi={10.1016/j.parco.2021.102829},
url={https://www.sciencedirect.com/science/article/pii/S0167819121000776?via%3Dihub}
}
% CUDA and distributed memory implementation
@article{williams20on,
author={David B. Williams--Young and Wibe A. de Jong and Hubertus J.J. van Dam and
Chao Yang},
title={On the Efficient Evaluation of the Exchange Correlation Potential on
Graphics Processing Unit Clusters},
journal={Frontiers in Chemistry},
volume={8},
pages={581058},
year={2020},
doi={10.3389/fchem.2020.581058},
url={https://www.frontiersin.org/articles/10.3389/fchem.2020.581058/abstract},
preprint={https://arxiv.org/abs/2007.03143}
}
% Algorithm for XC potential assembly and shared-memory CPU implementation
@article{petrone18an,
author={Alessio Petrone and David B. Williams--Young and Shichao Sun and
Torin F. Stetina and Xiaosong Li},
title={An Efficient Implementation of Two-Component Relativistic Density
Functional Theory with Torque-Free Auxiliary Variables},
journal={The European Physical Journal B},
volume={91},
number={169},
pages={},
year={2018},
doi={10.1140/epjb/e2018-90170-1},
url={https://link.springer.com/article/10.1140/epjb/e2018-90170-1}
}
GauXC provides a CMake build system with automatic dependency management (through FetchContent). As such, a simple CMake invocation will often suffice for most purposes
cmake -S /path/to/gauxc -B /path/to/build [GauXC configure options]
cmake --build /path/to/build
GauXC is linkable both as an installed library as well as a CMake subproject via FetchContent
# GauXC Discovery
find_package( gauxc REQUIRED )
target_link_libraries( my_target PUBLIC gauxc::gauxc )
# GauXC as CMake Subproject
include(FetchContent)
# Set GauXC CMake options (see below)
# Pull master branch of GauXC
FetchContent_Declare( gauxc
GIT_REPOSITORY https://github/com/wavefunction91/GauXC.git
GIT_TAG master
)
FetchContent_MakeAvailable( gauxc )
# Link to target
target_link_libraries( my_target PUBLIC gauxc::gauxc )
Variable Name | Description | Default |
---|---|---|
GAUXC_ENABLE_TESTS |
Enable Testing Framework (Catch2) | ON |
GAUXC_ENABLE_HOST |
Enable HOST integrators | ON |
GAUXC_ENABLE_CUDA |
Enable CUDA integrators | OFF |
GAUXC_ENABLE_HIP |
Enable HIP integrators | OFF |
GAUXC_ENABLE_MAGMA |
Enable MAGMA for batched BLAS (No effect if no GPU) | ON |
GAUXC_ENABLE_CUTLASS |
Enable CUTLASS for batched BLAS (No effect if no CUDA) | OFF |
GAUXC_ENABLE_NCCL |
Enable NCCL bindings for topology aware GPU reductions | OFF |
GAUXC_ENABLE_MPI |
Enable MPI Bindings | ON |
GAUXC_ENABLE_OPENMP |
Enable OpenMP Bindings | ON |
CMAKE_CUDA_ARCHITECTURES |
CUDA architechtures (e.g. 70 for Volta, 80 for Ampere) | -- |
BLAS_LIBRARIES |
Full BLAS linker. | -- |
MAGMA_ROOT_DIR |
Install prefix for MAGMA. | -- |
Coming Soon.... See test/standalone_driver.cxx
for an example end-to-end invocation of GauXC for various integrands.
GauXC is made freely available under the terms of a modified 3-Clause BSD license. See LICENSE.txt for details.
The development of GauXC is supported by the Exascale Computing Project (17-SC-20-SC), a collaborative effort of the U.S. Department of Energy Office of Science and the National Nuclear Security Administration.