The group focus on research crossing multiple software layers, from OS, runtimes and low-level APIs to programming models, tools, and applications.
The System Tools and Advanced Runtimes (STAR) group of the Barcelona Supercomputing Center-Centro Nacional de Supercomputación (BSC-CNS) has announced the release of the STAR software stack (version 2023.05), featuring several updates and enhancements. This release includes the OmpSs-2 programming model, Task-Aware libraries (TA-X), the ovni instrumentation and the nOS-V tasking library.
Here is a summary of the changes:
- First release of the new low-level nOS-V threading and tasking library, enabling co-execution of applications [5].
- The new NODES runtime, built on top of nOS-V, which implements the taskiter construct [1] and leverages directed cyclic task graphs (DCTG) to optimize the execution of iterative applications. Users now have the option to select between Nanos6 and NODES runtimes with a new compiler option.
- We are also excited to unveil for the first time the new TACUDA, TAHIP and TASYCL task-aware libraries (TA-X). They seamlessly integrate the CUDA, HIP and SYCL APIs with the OmpSs-2 and OpenMP tasking models.
- Performance and usability improvements in the Nanos6 runtime which also added instrumentation support for the idle CPUs.
- Several bug fixes and performance improvements in the TAMPI and TAGASPI communication libraries. Added support for the ovni instrumentation in TAMPI.
- New graph-based design in the ovni instrumentation to support complex models like the new breakdown timeline:
For more information and access to the repositories, please visit: https://github.com/bsc-pm/ompss-2-releases
Recent publications:
- D. Álvarez and V. Beltran, “Optimizing Iterative Data-flow Scientific Applications using Directed Cyclic Graphs,” in IEEE Access (2023), doi: 10.1109/ACCESS.2023.3269902.”
- M. Maroñas, A. Navarro, E. Ayguadé and V. Beltran, “Mitigating the NUMA effect on task-based runtime systems,” in the Journal of Supercomputing (2023), doi: 10.1007/s11227-023-05164-9
- PJ. Martinez-Ferrer, T. Arslan and V. Beltran, “Improving the performance of classical linear algebra iterative methods via hybrid parallelism,” in the Journal of Parallel and Distributed Computing (2023), doi: 10.1016/j.jpdc.20204.012
- S. Macià, PJ. Martinez-Ferrer, E. Ayguadé and V. Beltran, “Assessing Saiph, a Task-Based DSL for High-Performance Computational Fluid Dynamics,” in Future Generation Computer Systems (2023), doi: 10.1016/j.future.2023.0035
- D. Álvarez, K. Sala and V. Beltran, “nOS-V: Co-Executing HPC Applications Using System-Wide Task Scheduling, ” in arXiv, doi: 10.48550/arXiv.2204.10768
Funding and Acknowledges:
The STAR research group has been funded by Huawei SoW1 and the European Union’s Horizon 2020/EuroHPC research and innovation program under grants agreement Nº 955606 (DEEP-SEA) and Nº 956416 (ExaFOAM). Projects PCI2021-121958 and PCI2021-121961 have received funding from MCIN/AEI /10.13039/501100011033 and European Union NextGenerationEU/PRTR. This work has also received financial support from the Ramón y Cajal program [grant number RYC2019-027592-I] and the Severo Ochoa Centre of Excellence accreditation [grant number CEX2021-001148-S], both funded by MCIN/AEI /10.13039/501100011033. The Departament de Recerca i Universitats de la Generalitat de Catalunya has also provided funding for the Programming Models research group at BSC- UPC [grant number 2021 SGR01007].
