Large-scale cardiac simulations bring together multiple physical processes, spatial and temporal scales, and substantial computational workloads. Efficient numerical methods are therefore essential for making these simulations practical on high-performance computing systems.
In the latest dealiiX Seminar, Andrea Tonini and Junxiang Wang from Politecnico di Milano presented “Matrix-free algorithms for cardiac simulations”, focusing on the application of matrix-free finite element methods to cardiac electromechanics and poromechanics.
Mathematical cardiocirculatory models can provide valuable insights into cardiac function, diagnostics, and cardiovascular diseases. The three-dimensional models considered in the seminar involve several coupled submodels, resulting in large-scale multiphysics problems that require efficient parallel algorithms and HPC resources.
A key computational challenge in conventional finite element implementations is the assembly, storage, and repeated access of large sparse system matrices. Matrix-free algorithms take a different approach by evaluating finite element operators directly, avoiding the assembly of the corresponding global matrices. This can reduce data movement and increase arithmetic intensity.
During the seminar, Andrea Tonini and Junxiang Wang discussed how these methods can be applied to cardiac electromechanics and poromechanics, presenting numerical experiments that assess:
- Computational performance
- Memory efficiency
- Parallel scalability
- The potential of matrix-free methods as a common computational foundation for large-scale multiphysics cardiac simulations
The seminar provided a valuable look at how algorithmic developments can address the computational demands of increasingly detailed cardiac models and support the development of scalable simulation technologies.
Watch the seminar
The recording and presentation slides are now available:
🎥 [Watch the seminar recording]
📑 [Download the presentation slides]
Many thanks to Andrea Tonini and Junxiang Wang for sharing their work, and to everyone who participated in the discussion.
Stay tuned for the next dealiiX Seminar Series events exploring advanced numerical methods, HPC, and computational approaches for digital twins of the human body.