The Parallel Sparse Computation Toolkit (PSCToolkit): bringing linear algebra to the extreme scale

• Date: 23. 4. 2026
• Time: 14:00 - 15:30
• Address:
  Sokolovská 83, Praha
• Room: K1
• Speaker: Fabio Durastante

The increasing demand for high-fidelity simulations in science and engineering is driving computational methods toward extreme-scale architectures, where efficiency, scalability, and robustness of linear algebra operations become critical. This talk presents the Parallel Sparse Computation Toolkit (PSCToolkit), a suite of libraries designed to address these challenges by enabling efficient sparse linear algebra computations on modern high-performance computing systems. We will first outline the core design and implementation strategies behind PSCToolkit, including its data structures, parallelization model, and communication patterns tailored for distributed-memory environments. Emphasis will be placed on how the toolkit achieves scalability while maintaining flexibility and ease of use. The talk will then showcase representative use cases, demonstrating how PSCToolkit can be employed to solve large-scale problems arising in scientific computing. These examples highlight both performance characteristics and practical integration into existing workflows. Finally, we will discuss the integration of PSCToolkit within the deal.II finite element library, illustrating how advanced sparse linear algebra capabilities can be seamlessly incorporated into established numerical frameworks. This integration enables users to leverage extreme-scale performance in real-world finite element applications. Overall, the presentation aims to provide insight into the role of modern sparse computation tools in pushing the boundaries of large-scale numerical simulation. The material presented in this talk is the result of collaborative work with Pasqua D'Ambra (IAC-CNR), Salvatore Filippone, and Marco Feder (University of Rome "Tor Vergata").