A Python interface to the MAGMA libraries. This package is developed to provide an python interface to GPU accelerated matrix and vector operations - especially the sparse-iter functions from MAGMA. It provides easy access to iterative sparse solvers on the GPU. Therefore, two classes magma_matrix and magma_vector are added. Objects of these classes can be created with a scipy sparse matrix or numpy array, respectively and the class can be used to automatically transfer the magma matrix to GPU memory space and internally call functions corresponding to the present data type (z, c, d or s). The package currently supports MAGMA-2.0.2. It has not been tested for other versions of the MAGMA libraries. Parts of the code are taken from scikit-cuda written by Lev Givon.
The package is written and maintained by Carlo Holly for my research at RWTH Aachen University at the Chair for Laser Technology.
The MAGMA-2.0.2 libraries (libmagma.so and libmagma_sparse.so on a Linux system) have to be installed on your system. To download and install the libraries visit this site.
To install pymagma use...
This project is maintained by Carlo Holly. Please contact me via carlo.holly@rwth-aachen.de.
If you make use of pymagma in your research, please cite it. The BibTeX reference is
- @article{Holly_pymagma,
- author = {Carlo Holly and Lev E. Givon}, title = {pymagma 0.0.1 - a {Python} interface to the {MAGMA} libraries}, month = August, year = 2016, doi = "", url = "",
}
This software is licensed under the BSD License. See the included LICENSE file for more information.