freeaml

Description

FreeAML is an easy-to-use, parallel applied mathematics library written in C++. It boasts native integration with OpenMP and can therefore automatically use multiple cores to accelerate its implemented algorithms (e.g. matrix multiplication, vectorial dot product etc.) on shared-memory architectures.

FreeAML offers basic data structures tailored for mathematical applications (vectors and matrices) as well as:

• linear system solvers
• linear least-squares solvers
• matrix factorization algorithms.

Integrating freeAML into your project can be done in minutes since it is a header-only library. All you need to do is download the sources and include the desired header files into your project.

License

All code from this project is licensed under the FreeBSD license. See the LICENSE file for more information.

Available tools

The best way to learn how to use freeAML is by looking at its unit tests. Each unit test is written in a very simple manner and is a self-contained example of how a certain tool (e.g. a linear system solver) can be used. The following list shows the available data structures and algorithms in the library, with links pointing to their associated unit tests:

Data structures

• Matrix
• Sparse matrix
• Vector

Linear system solvers

• Biconjugate gradient stabilized (BiCGSTAB)
• Conjugate gradient (CG)
• Conjugate gradient with incomplete-Cholesky preconditioning
• Gaussian elimination
• Gauss-Seidel
• Generalized minimum residual (GMRES)
• Jacobi
• Minimum residual (MINRES)
• Steepest descent
• Successive over-relaxation (SOR)
• Weighted Jacobi

Linear least-squares solvers

• Normal equations
• QR solver

Matrix factorization algorithms

• Bidiagonal factorization
• Cholesky factorization
• Hessenberg factorization
• PLU factorization
• QR factorization

Installation and usage instructions

FreeAML can be downloaded with the following command:

git clone https://github.com/dassencio/freeaml.git

No installation is required: users can simply include the desired header files from freeAML in their program sources and start focusing on what matters to them.

If you wish to benefit from the native parallelization with OpenMP, make sure that it is installed on your system (if you are using a popular Linux distribution such as Ubuntu, it should already be there). Also, don’t forget to tell your compiler to link your application to the OpenMP library (for instance, with gcc, you must add the -fopenmp option).

To compile and run the unit tests, go to the directory where you cloned the freeAML sources, create a directory called build there, enter it and then run the following commands in sequence:

cmake -DUSE_OPENMP=ON ..
make
./run-tests

If you do not wish to use OpenMP, simply replace ON with OFF on the command above.

Doxygen documentation

FreeAML is thoroughly documented using the Doxygen format. To generate the documentation as an HTML document, go to the directory where you cloned the freeAML sources, create a directory called build there, enter it and then run the following commands in sequence:

cmake ..
make doc

This will create a subdirectory called html containing an index.html file. Open it on a browser to navigate through the documentation of the library classes and functions.

Contributors & contact information

Diego Assencio / diego@assencio.com

This site is open source. Improve this page.