EigenG-Batched
Overview
EigenG-Batched is a spinoff Batched version of EigenG, focusing on portability with EigenG and other Batched eigensolver such as cusolver 'cusolverDnSsyevjBatched' function. Along with the portability. it outperforms cusolver when the matrix size is middle (32x32) or larger cases.
Downloads
- EigenG-Batched version 1.2 (tar.gz, 45KB) (October 25, 2022)
Publications
- Hasegawa, Imamura. Ina, Onodera, Asahi, Idomura: GPU Optimization of Lattice Boltzmann Method with Local Ensemble Transform Kalman Filter, ScalAH2022 conj with SC22 at Dallas (accepted)