Forked from https://github.com/paranumal/libparanumal
git clone https://github.com/E-Stroem/libparanumal
cd libparanumal/
git clone https://github.com/noelchalmers/occa
cd occa
export OCCA_DIR=`pwd`
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$OCCA_DIR/lib
make -j
cd ../
OCCA must be build on a system with GPU access for GPU support.
See RUN.bsub for CPU. (Build occa on CPU) See RUNGPU.bsub for GPU. (Build Occa on GPU)
Anders Melander - [email protected]
Emil Strøm - [email protected]
An experimental set of finite element flow solvers for heterogeneous (GPU/CPU) systems. The initial development of libParanumal was performed by the Parallel Numerical Algorithms Group at Virginia Tech.
libParanumal is funded in part by the US Department of Energy as part of the activities of the Center for Efficient Exscale Discretizations.
Why is it called libParanumal ?: the high-order finite-element implementations in libParanumal are spectrally accurate and rely heavily on ghost elements for MPI communications.
If you use libParanumal as part of a research project see Section 8 below for papers to reference.
Brief summary of major features:
A. Supported elements:
- Triangles, quadrilaterals, tetrahedra, hexahedra.
- Lagrange basis functions up to degree 15.
- Partial support for Bezier-Bernstein basis functions.
B. Mesh wrangling:
- Gmsh format file loaders.
- Load balanced geometric partitioning using space filling curves (Hilbert or Morton ordering).
- Clustered partitioning for multirate time stepping.
C. Elliptic solver:
- Linear Poisson and screened Poisson potential solvers.
- GPU optimized matrix-vector products.
- Hybrid p-type multigrid and algebraic multigrid preconditioned conjugate gradient solver.
- Sparse matrix or nearly matrix-free algebraic multigrid for coarse levels of multigrid hierarchy.
D. Heterogeneous accelerated flow solvers:
- Linearized Euler equations.
- Isothermal compressible Navier-Stokes solver with:
- Upwind discontinuous Galerkin discretization in space.
- Dormand-Prince adaptive Runge-Kutta integration in time.
- Isothermal Galerkin-Boltzmann gas dynamics solver with:
- Penalty flux DG discretization in space.
- Adaptive semi-analytic (pointwise exponential) integration in time.
- Multiaxial quasi-perfectly matched absorbing layer far field boundary condition.
- Incompressible Navier-Stokes solver with:
- Choice of continuous FEM or interior penalty DG in space.
- Extrapolation-BDF integration in time.
- Sub-cycling (Operator Integration Factor Splitting) for advection.
E. Dependencies:
- Message Passing Interface (MPI).
- The libParanumal makefiles assume that mpic++ are installed and visible in your path.
- Open Concurrent Compute Abstraction (OCCA)
- OCCA must be installed.
- OCCA will try to detect if any of these execution models are installed OpenMP, CUDA, OpenCL, HIP.
- If OCCA does not detect any of these it will default to Serial execution.
- You will need to adjust the libParnumal setup input files to choose the execution model and compute device appropriate for your system.
- The OCCA github repo is here
- The OCCA webpage is here
git clone https://github.com/paranumal/libparanumal
git clone https://github.com/noelchalmers/occa
cd occa
export OCCA_DIR=`pwd`
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$OCCA_DIR/lib
make -j
cd ../
The elliptic solver and flow solvers reside in sub-directories of the solver directory. Each sub-directory includes makefile, src directory, data directory (including header files for defining boundary conditions), okl kernel directory, and setups directory. The setups directory includes a number of example input files that specify input parameters for the solver.
cd libparanumal/solvers/elliptic
make -j
./ellipticMain setups/setupQuad2D.rc
mpiexec -n 2 ./ellipticMain setups/setupQuad2D.rc
The MIT License (MIT)
Copyright (c) 2017 Tim Warburton, Noel Chalmers, Jesse Chan, Ali Karakus
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Discontinuous Galerkin Boltzmann (bns) solver arXiv version: Karakus, A., Chalmers, N., Hesthaven, J.S. and Warburton, T., 2018. Discontinuous Galerkin Discretizations of the Boltzmann Equations in 2D: semi-analytic time stepping and absorbing boundary layers. arXiv preprint arXiv:1805.02082.
Incompressible Navier-Stokes (discontinuous) Galerkin (ins) solver arXiv version: Karakus, A., Chalmers, N., Swirydowicz, K. and Warburton, T., 2017. GPU Acceleration of a High-Order Discontinuous Galerkin Incompressible Flow Solver. arXiv preprint arXiv:1801.00246.
Optimization of elliptic mat-vec operations for (elliptic) solver on hexes arXiv version: Świrydowicz, K., Chalmers, N., Karakus, A. and Warburton, T., 2017. Acceleration of tensor-product operations for high-order finite element methods. arXiv preprint arXiv:1711.00903.