Exploiting Many-Core Architectures for Dimensionally Adaptive Sparse Grids

Beschreibung

Sparse grids are the only general-purpose grid-based discretization technique that is also suitable for settings in more than three dimensions. However, their hierarchical nature poses a challenge to modern multi- and many-core architectures. In this thesis, an efficient, contemporary, and hardware-aware set of co-designed data structures and algorithms is presented for dimensionally adaptive sparse grids. The solutions described are shown to exploit the performance-critical resources of modern parallel computing platforms and thus overcome performance problems observed in other sparse grid implementations. The applicability and competitiveness of the presented solutions is also demonstrated in relevant applications.

Autor*in

Gerrit Buse

Themen in »Exploiting Many-Core Architectures for Dimensionally Adaptive Sparse Grids«

Dimensionally Adaptive Sparse Grids Many-Core Architectures performance-critical resources

Stimmen zu »Exploiting Many-Core Architectures for Dimensionally Adaptive Sparse Grids«

Details

Link teilen