A library for fast Prototyping of Numerical methods.
This is a small library that allows quick and dirty implementations of numerical methods for PDEs.
It started as a throw-away code to implement the cutHHO method (see this paper), however it grew more than expected.
At the moment it implements
- the 2D version of cutHHO (Fictitious Domain problem)
- the 2D version of an HHO method for obstacle problems
The library is only 2D and allows only quadrilateral meshes (it had to be quick and dirty, remember?).
The code is tested on Mac OS X and on Linux (sorry, I don't have the time to deal with the non-compliance to standards of Windows). To compile you need, apart of a decent C++14 (I'll move to c++17 soon) compiler that should be already available in your system, the following software:
- CMake
brew install cmakeon OS Xapt-get install cmakeon Linux
- SILO
brew install siloon OS Xapt-get install libsilo-devon Linux
- Lua
brew install luaon OS Xapt-get install liblua5.3-devon Linux
- Eigen3
brew install eigenon OS Xapt-get install libeigen3-devon Linux
If you want to just run the code, on OS X, you can tap my homebrew repo with
brew tap datafl4sh/code
and install with just
brew install cuthho
TO BE UPDATED
You will end up with two executables, cuthho_square and convergence_test.
The first is the cutHHO driver, the second does a convergence test of the standard cutHHO method.
The executable cuthho_square accepts the following parameters:
-M: the number of cells in x direction-N: the number of cells in y direction-k: the degree of the method-m: disable Point Displacement
Be sure to run cuthho_square in a specific directory, because it produces a bunch of files.