RCOMPSs: A Task-based Parallel Programming Model in R
R is widely used for scientific and statistical computing, and its package ecosystem supports a broad range of reproducible analyses. This paper introduces RCOMPSs, an R binding to the COMPSs runtime for task-based parallel execution. Users write standard R code and mark selected functions as tasks. The runtime then tracks dependencies, schedules tasks, moves data, and executes them across available resources. RCOMPSs is intended to complement existing R parallel frameworks by supporting workflows that benefit from explicit task dependencies. We evaluate it with three representative algorithms, k-nearest neighbors classification, k-means clustering, and linear regression with predictions, on KAUST’s Shaheen III and BSC’s MareNostrum 5. The experiments examine execution-time scaling and workflow trade-offs in single-node comparisons and distributed-memory runs. Overall, RCOMPSs adds COMPSs task-based execution to the R parallel ecosystem for scalable statistical workflows.
Authors
- Rosa M. Badía (ORCID: https://orcid.org/0000-0003-2941-5499)
- David Elliot Keyes (ORCID: https://orcid.org/0000-0002-4052-7224)
- Marc G. Genton (ORCID: https://orcid.org/0000-0001-6467-2998)
- Javier Conejero (ORCID: https://orcid.org/0000-0001-6401-6229)
- Jorge Ejarque (ORCID: https://orcid.org/0000-0003-4725-5097)
- Sameh Abdulah (ORCID: https://orcid.org/0000-0002-8850-5753)
- Ying Sun (ORCID: https://orcid.org/0000-0001-6703-4270)
- Xiran Zhang
Institutions
- Barcelona Supercomputing Center (ES)
- King Abdullah University of Science and Technology (SA)
Publication Details
- Journal
- Journal of Computational and Graphical Statistics
- Published
- 2026-11-19
- DOI
- https://doi.org/10.1080/10618600.2026.2762063
- Citations
- 1
- Primary Topic
- Parallel Computing and Optimization Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00