About DML-CZ | FAQ | Conditions of Use | Math Archives | Contact Us
  Previous |  Up |  Next

Article

Břichňáč, Vít ; Šístek, Jakub
Performance of parallel QR factorization methods on the NVIDIA Grace CPU Superchip. (English). In: Chleboun, J., Papež, J., Segeth, K., Šístek, J. and Vejchodský, T. (eds.): Programs and Algorithms of Numerical Mathematics. Proceedings of Seminar. Hejnice, June 23-28, 2024. , Prague, 2025. pp. 29-40
MSC: 65F05 | DOI: 10.21136/panm.2024.03
Keywords:
QR factorization; task-based programming; NVIDIA Grace CPU
Summary:
This article studies several algorithms for QR factorization based on hierarchical Householder reflectors organized into elimination trees, which are particularly suited for tall-and-skinny matrices and allow parallelization. We examine the effect of various parameters on the performance of the tree-based algorithms. The work is accompanied with a custom implementation that utilizes a task-based runtime system (OpenMP or StarPU). The same algorithm is implemented in the PLASMA library. The performance evaluation is done on the recent NVIDIA Grace CPU Superchip.
Partner of
EuDML logo