In a 40960x40960 generalized eigenproblem on four A100 GPUs, ELPA 2024.05 reduces the matrix-matrix multiplication time from 294 to 18.8 seconds, producing a 3 to 4x overall speedup.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cond-mat.mtrl-sci 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Solvers for Large-Scale Electronic Structure Theory: ELPA and ELSI
In a 40960x40960 generalized eigenproblem on four A100 GPUs, ELPA 2024.05 reduces the matrix-matrix multiplication time from 294 to 18.8 seconds, producing a 3 to 4x overall speedup.