Citing Ginkgo#

If you use Ginkgo in academic work or in any publication, please cite it. Citations are the most direct way to support continued development — funding agencies look at them, and they help us justify the time we spend making the library faster and more capable.

Primary citation#

The canonical Ginkgo reference is the 2022 ACM TOMS paper:

@article{ginkgo-toms-2022,
  title   = {{Ginkgo: A Modern Linear Operator Algebra Framework for High Performance Computing}},
  author  = {Anzt, Hartwig and Cojean, Terry and Flegar, Goran and G\"{o}bel, Fritz and
             Gr\"{u}tzmacher, Thomas and Nayak, Pratik and Ribizel, Tobias and
             Tsai, Yuhsiang Mike and Quintana-Ort{\'\i}, Enrique S.},
  journal = {ACM Transactions on Mathematical Software},
  volume  = {48},
  number  = {1},
  year    = {2022},
  pages   = {1--33},
  doi     = {10.1145/3480935}
}

If you can only fit one citation in your paper, use that one — it covers the library’s overall design and serves as the default attribution.

Software citation (JOSS)#

A Journal of Open Source Software paper accompanies the software and is useful when you want to cite Ginkgo as software rather than the underlying methodology:

@article{ginkgo-joss-2020,
  title   = {{Ginkgo: A High Performance Numerical Linear Algebra Library}},
  author  = {Anzt, Hartwig and Cojean, Terry and Yen-Chen, Chen and Flegar, Goran and
             G\"{o}bel, Fritz and Gr\"{u}tzmacher, Thomas and Nayak, Pratik and
             Ribizel, Tobias and Tsai, Yuhsiang Mike},
  journal = {Journal of Open Source Software},
  volume  = {5},
  number  = {52},
  pages   = {2260},
  year    = {2020},
  doi     = {10.21105/joss.02260}
}

Citing a specific feature#

Several Ginkgo subsystems have their own peer-reviewed publications. If your work relies heavily on a particular feature, citing both the flagship TOMS paper and the subsystem paper is the most accurate attribution.

The relevant publications are listed at the bottom of the corresponding concept page in the user guide:

Distributed solvers — Distributed computing → Publications (also covers Schwarz, distributed Pgm).
Batched solvers — Batched functionality → Publications.
Mixed-precision techniques — Mixed precision → Publications.
Preconditioners (block Jacobi, ParILU, ParILUT, AMG) — Preconditioners → Publications.
Hardware-portable kernels and the executor model — the TOMS paper above covers this comprehensively.

Each page links the BibTeX entries directly, so you can copy them from there into your bibliography.

Citing a specific release#

Tagged releases on GitHub carry the version number and a date. For reproducibility-critical work, cite the exact version you used in addition to the paper:

Ginkgo X.Y.Z, https://github.com/ginkgo-project/ginkgo, tag vX.Y.Z, accessed YYYY-MM-DD.

If you need a DOI for a specific release for archival purposes, maintainers can mint one on Zenodo on request — ask in Discussions or by email.

Crediting individual authors#

The full list of authors who have contributed to Ginkgo is in the contributors.txt file. If you are writing a software-focused paper (rather than a method paper) and want to credit specific contributors beyond the citation list, the maintainer team is happy to advise — open a discussion or send an email.

Telling us about your work#

If you publish a paper or thesis that uses Ginkgo, we’d love to hear about it — it helps us understand which features are landing in practice, and it directly supports funding renewals. Drop a line in Discussions or email ginkgo.library@gmail.com with a link to the paper.