Triumvirate is a Python/C++ software package for measuring three-point (and
two-point) clustering statistics in large-scale structure (LSS) cosmological
analyses.
Comprehensive documentation including the scientific background, installation instructions, tutorials and API reference can be found at triumvirate.readthedocs.io.
Triumvirate as a Python package is distributed through PyPI and Conda. Instructions for installation
can be found on the Installation
page in the documentation.
Tip
CUDA variants of the Python package are/will be made available as
Triumvirate-CUDA on PyPI
and triumvirate-cuda through
Conda.
Triumvirate as either a static library or a binary executable can be
built using make. Instructions for compilation can be found on the
Installation
page in the documentation.
Both the Python package and the C++ library/program can be set up in
development mode with make, provided that dependency requirements are
satisfied (GSL and FFTW3 libraries are mandatory while an OpenMP library
is optional).
First git clone the desired branch/release from the GitHub repository and
change into the repository directory path:
git clone git@github.com:MikeSWang/Triumvirate.git --branch <branch-or-release>
cd TriumvirateThen, execute in shell:
make clean
make ([py|cpp]install)|(cpp[libinstall|appbuild]) [useomp=(true|1)] [usecuda=(true|1)]where cpplibinstall or cppappbuild respectively builds the C++
static library or binary executable only, cppinstall builds both,
pyinstall builds the Python package only, and install builds
all of the above. To enable OpenMP parallelisation, append useomp=true
or useomp=1 to the end of the second line as shown above. To enable
CUDA support, append usecuda=true or usecuda=1 to the end of the
second line as shown above.
Note
The latest release is on the main branch. The default
Makefile (located at the repository directory root) should
work in most build environments, but may need to be modified as appropriate.
Note
See the Installation page in the documentation for more details about dependency requirements.
Important
If enabling OpenMP, ensure the C++ compiler used supports it and is
configured accordingly. The default Makefile (located at the
repository directory root) assumes the GCC compiler and OpenMP library. See
the Installation
page in the documentation for more details.
Important
If enabling CUDA capability, ensure there is a CUDA-capable GPU with the
appropriate driver installed. For atypical CUDA Toolkit paths, you may
need to append the header and library paths to DEP_INCLUDES and
DEP_LDFLAGS in the default Makefile (located at the
repository directory root). See the Installation
page in the documentation for more details.
Tip
Pass option -j[N] -O to make to run multiple concurrent jobs
for parallel building (optional parameter N is the number of
parallel jobs; see GNU Make Manual).
To acknowledge the use of Triumvirate in your published research, please
cite the publications linked above; for convenience, you can refer to the
files CITATION.cff and CITATION.md
for the relevant information in different formats.
This project has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (Grant agreement ID: 853291).
Key underlying numerical algorithms were originally developed by
Naonori S Sugiyama, and are available in the GitHub repository hitomi.
We thank the JOSS reviewers, William Coulton (@wcoulton) and Alfonso Veropalumbo (@alfonso-veropalumbo), for their valuable feedback and suggestions (openjournals/joss-reviews#5571), which have improved the functionality and documentation of the code.
User feedback and contributions are very welcome. Please refer to the contribution guidelines.
A community forum for users and developers exists, where you can receive announcements, post questions, share ideas and get updates.
A wiki site collects wisdoms for specific use cases and user environments.
Release notes are included in the change log.
Triumvirate is made freely available under the GPLv3+ licence.
Please see LICENCE (located at the repository directory root)
for full terms and conditions.
© 2023 Mike S Wang & Naonori S Sugiyama