Milan Klöwer

PhD student in Climate Computing
Atmospheric, Oceanic and Planetary Physics
University of Oxford
twitter @milankloewer
github @milankl

The speaker's profile picture


3.6x speedup on A64FX by squeezing ShallowWaters.jl into Float16
Milan Klöwer

ShallowWaters.jl, a fluid circulation model that was written with a focus on 16-bit arithmetics, runs on A64FX 3.6x faster in Float16 compared to Float64 without a significant model degradation. Calculations were systematically rescaled to fit into the very limited range of Float16 guided by Sherlogs.jl. ShallowWaters.jl shows that 16-bit calculations on A64FX are indeed a competitive way to accelerate Earth-system simulations on available hardware.