Guided research presentation. Matthias is advised by Yu Wang and Prof. Dr. Christian B. Mendl.
Previous talks at the SCCS Colloquium
Matthias Reumann: Enhanced Krylov Methods for Molecular Hamiltonians via Tensor Hypercontraction: Parallel Reduction Algorithm For Coulomb Term
SCCS Colloquium |
Recently Wang et al. proposed the THC-MPO algorithm for the more efficient computation of Krylov subspace methods in terms of memory and runtime complexity. This work studies the algorithm from the perspective of parallel computing. Preliminary benchmarks show that the speed-up of the parallel algorithm is nearly perfect for up to 64 cores.