A supercomputing framework for solving PDEs by hybrid parallelism.
SOLVCON: a multi-physics, supercomputing software framework for high-fidelity solutions of partial differential equations (PDEs) by hybrid parallelism.
SOLVCON uses the space-time Conservation Element and Solution Element (CESE) method to solve generic conservation laws. SOLVCON focuses on rapid development of high-performance computing (HPC) code for large-scale simulations. SOLVCON is developed by using Python for the main structure, to incorporate C, CUDA, or other programming languages for HPC.
- Pluggable multi-physics
- Unstructured meshes for modeling complex geometry
- Hybrid parallel computing
- Ready-to-use I/O formats
- Parallel I/O and in situ visualization
- Automated work flow
Installing SOLVCON requires building shared libraries by using SCons. SOLVCON uses C to implement HPC code. The C code in SOLVCON is designed to be standard shared libraries rather than Python extension modules. SOLVCON uses ctypes to access the libraries. As such, the binary code can be optimized more flexibly. It is recommended to run SOLVCON on 64-bit Linux.
SOLVCON depends on the following packages: (i) Python 2.6 or 2.7 (preferred), (ii) SCons, (iii) gcc (version 4.3 or higher) or icc, (iv) Numpy (version 1.6 or higher), (v) LAPACK, (vi) NetCDF (version 4 or higher), and (vii) METIS (version 4.0.3; SOLVCON will download it for you on building). Optional dependencies include: (i) SCOTCH (version 5.1 or higher) as an alternative of METIS, (ii) Nose for running unit tests, (iii) Epydoc for generating API documentation, (iv) VTK for in situ visualization, and (v) docutils and pygraphviz for Epydoc formatting. Debian or Ubuntu users can use the following command to install the dependencies:
$ sudo apt-get install scons build-essential gcc liblapack-pic libnetcdf-dev libnetcdf6 netcdf-bin libscotch-5.1 python2.7 python2.7-dev python-numpy python-vtk python-nose python-epydoc python-docutils python-pygraphviz
Another optional dependency is CUDA, which needs to be separately installed and configured. For using meshes with more then 35 million cells, SCOTCH-5.1 is recommended. METIS-4 has issues on memory allocation for large graphs.
There are three steps to install SOLVCON:
Obtain the latest release from https://bitbucket.org/yungyuc/solvcon/downloads . Unpack the source tarball.
Get into the source tree and run SCons to build the binary codes:
$ cd $SCSRC $ scons --download --extract
where $SCSRC indicates the root directory of unpacked source tree.
$ python setup.py install
The option --download used above lets the building script download necessary external source packages, e.g., METIS, from Internet. Option --extract extracts the downloaded packages.
Although not recommended, you can optionally install SOLVCON to your $HOME/.local directory. It is one of the workarounds when you don’t have the root permission on the system. To do this, add the --user when invoking the setup.py script:
$ python setup.py install --user
SOLVCON is designed to work without explicit installation. You can simply set the $PYTHONPATH environment variable to point to the unpacked source distribution ($SCSRC). Compilation of binary code by using SCons is still required.
To use the latest development version, you need to use Mercurial to access the source repository. Clone the repository:
$ sudo apt-get install mercurial $ hg clone https://bitbucket.org/yungyuc/solvcon
and follow steps 2 and 3 in Install.
If you want to rebuild and reinstall, you can run:
$ cd $SCSRC $ scons $ python setup.py install
without using the options --download and --extract. If you want a clean rebuild, run scons -c before scons.
If you have Nose installed, you can run:
inside the source tree for unit tests. To test the installed package, use the following command instead:
$ python -c 'import solvcon; solvcon.test()'
When testing the installed package, make sure your current directory does not have a sub-directory named as solvcon.
Because SOLVCON uses ssh as its default approach for remote procedure call (RPC), you need to set up the public key authentication for ssh, or some of the unit tests for RPC could fail. Some tests using optional libraries could be skipped (indicated by S), if you do not have the libraries installed. Everything else should pass.
Build and Install Dependencies (Optional)
SOLVCON depends on a number of external software packages. Although these dependencies should be taken care by OSes, it takes time to get the support personnels to install missing packages on a cluster/supercomputer. As such, SOLVCON provides a simple building system to facilitate the installation into a customizable location.
The $SCSRC/ground directory contains scripts to build most of the packages that SOLVCON depends on. The $SCSRC/ground/get script downloads the source packages to be built. The $SCSRC/ground/Makefile file has three default targets: binary, python, and vtk. The built files will be automatically installed into the path specified by the $SCROOT environment variable, which is set to $HOME/opt/scruntime by default. The $SCROOT/bin/scvars.sh script will be created to export necessary environment variables for the installed software, and the $SCROOT environment variable itself.
The $SCSRC/soil directory contains scripts to build gcc. The $SCROOT/bin/scgccvars.sh script will be created to export necessary environment variables for the self-compiled gcc. The enabled languages include only C, C++, and Fortran. The default value of $SCROOT remains to be $HOME/opt/scruntime, while the software will be installed into $SCROOT/soil. Note: (i) Do not use different $SCROOT when building $SCSRC/soil and $SCSRC/ground. (ii) On hyper-threading CPUs the NP environment variable should be set to the actual number of cores, or compilation of gcc could exhaust system memory.
$SCROOT/bin/scvars.sh and $SCROOT/bin/scgccvars.sh can be separately sourced. The two sets of packages reside in different directories and do not mix with each other nor system software. Users can disable these environments by not sourcing the two scripts.
$ sudo apt-get install build-essential gcc gfortran gcc-multilib m4 libreadline6 libreadline6-dev libncursesw5 libncurses5-dev libbz2-1.0 libbz2-dev libdb4.8 libdb-dev libgdbm3 libgdbm-dev libsqlite3-0 libsqlite3-dev libcurl4-gnutls-dev libhdf5-serial-dev libgl1-mesa-dev libxt-dev
These building scripts have only been tested with 64-bit Linux.