SPORCO-CUDA provides GPU-accelerated versions of selected convolutional sparse coding algorithms in the SPORCO package. It is a component of SPORCO, and is subject to the same license, but is made available as an optional extension to avoid complicating the prerequisites and build/install procedure for the main part of SPORCO. If you use this software for published work, please cite it.
Documentation is available online at Read the Docs, or can be built from the root directory of the source distribution by the command
python setup.py build_sphinx
in which case the HTML documentation can be found in the build/sphinx/html directory (the top-level document is index.html).
Scripts illustrating usage of the package can be found in the examples directory of the source distribution. These examples can be run from the root directory of the package by, for example
python examples/cmp_cbpdn.py
To run these scripts prior to installing the package, it is necessary to build it in place, which involves the following steps:
Install the requirements described below
If
nvccis not already in the executable search path, add it; e.gexport PATH=$PATH:/usr/local/cuda-10.1/bin
where
/usr/local/cuda-10.1/binis the path for CUDA compilernvcc.Build the
sporco-cudapackage in place:python setup.py build_ext --inplace
Set the
PYTHONPATHenvironment variable to include the root directory of the package. For example, in abashshellexport PYTHONPATH=$PYTHONPATH:`pwd`
from the root directory of the package.
If the
sporcopackage is not installed, create a symlink from the SPORCO-CUDA package root directory to thesporcodirectory in the SPORCO package.
The primary requirements are Python, sporco and its requirements, Cython, and the CUDA Toolkit.
See the installation instructions in the online documentation.
SPORCO-CUDA is part of the SPORCO package and is distributed with the same 3-Clause BSD license; see the LICENSE file for details.