- Kyungjoo Kim
- Mauro Perego, Sandia National Labratories
- Nathan Roberts
Finite elements are widely used to solve continuum mechanics problems across a range of discretizations, from nodal continuous elements for elasticity to more complex mixed discretizations in fluid dynamics and elasticity applications with incompressible materials. Recent advances in hardware have led to opportunities for substantial acceleration of finite element computations; however, developing software and algorithms that take full advantage of the hardware, especially in a performance-portable fashion, remains a challenge. In this minisymposium we focus on efficient implementation of such finite element discretizations on different architectures (e.g. CPUs/GPUs).
Relevant topics include flexible and performant workflows for matrix assembly and matrix-free assembly and performance optimizations for special classes of elements/meshes (e.g. tensor product elements, hybrid elements, affine elements, and high-order/hierarchical bases).