Scientific computing requires handling large composed or structured matrices. Fastmat is a framework for handling large composed or structured matrices. It allows expressing and using them in a mathematically intuitive way while storing and handling them internally in an efficient way. This approach allows huge savings in computational time and memory requirements compared to using dense matrix representations.
Binary wheels are provided for Python >=3.5 for linux, windows and mac, as well as for x86 and ARM architectures.
For all systems, for which no wheels are provided, you may still install fastmat from the soruce distribution.
If you use fastmat, or parts of it, for commercial purposes you are required to acknowledge the use of fastmat visibly to all users of your work and put a reference to the project and the EMS Group at TU Ilmenau.
If you use fastmat for your scientific work you are required to mention the EMS Group at TU Ilmenau and cite the following publication affiliated with the project:
Christoph W. Wagner and Sebastian Semper and Jan Kirchhof, fastmat: Efficient linear transforms in Python, SoftwareX, 2022, https://doi.org/10.1016/j.softx.2022.101013
@article{Wagner_2022,
doi = {10.1016/j.softx.2022.101013},
url = {https://doi.org/10.1016%2Fj.softx.2022.101013},
year = {2022},
month = {jun},
publisher = {Elsevier {BV}},
volume = {18},
pages = {101013},
author = {Christoph W. Wagner and Sebastian Semper and Jan Kirchhof},
title = {fastmat: Efficient linear transforms in Python},
journal = {{SoftwareX}}
}
fastmat currently supports Linux, Windows and Mac OS. Lately it also has been seen on ARM cores coming in a Xilinx ZYNQ FPGA SoC shell. We encourage you to go ahead trying other platforms as the aforementioned as well and are very happy if you share your experience with us, allowing us to keep the list updated.
fastmat is included in the Python Package Index (PyPI) and can be installed
from the commandline by running one easy and straightforward command:
pip install fastmat
When installing with pip all dependencies of the package will be installed along. With release 0.1.1 python wheels will be offered for many versions greatly improving installation time and effort.
Building binaries has been developed and tested for the use
pip install .
to install fastmat on your computermake install
to install fastmat on your computermake install MODE=--user
insteadmake
targets that directly or indirectly invoke Python:
make install PYTHON=python2
make compile PYTHON=python3
make compile
make install
is possible, provided the installation log file setup.files
has been preserved
Invoking make uninstall
without a local setup.files
causes another installation for generating the setup file log prior to uninstallingFeel free to have a look at the demos in the demo/
directory of the source
distribution. Please make sure to have fastmat already installed when running
these.
Please note that the edgeDetect demo requires the Python Imaging Library (PIL) installed and the SAFT demos do compile a cython-core of a user defined matrix class beforehand thus having a delaying the first time they're executed.
Please have a look at the documentation, which is included in the source
distribution at github or may be built locally on your machine by running
make doc
If you experience any trouble please do not hesitate to contact us or to open an issue on our github projectpage: https://github.com/EMS-TU-Ilmenau/fastmat
Please check out our project documentation at readthedocs.
Often, this is caused by missing header files. Unfortunately windows ships without a c-compiler and the header files necessary to compile native binary code. If you use the Intel Distribution for Python this can be resolved by installing the Visual Studio Build tools with the version as recommended by the version of the Intel Distribution for Python that you are using.
Please contact us or leave your bug report in the issue section. Thank You!