1st IAU-G5 seminar: The revolution of task-based computing, applied to stellar and planetary atmospheres

Authors
Prof.
ºAke Nordlund
Date and time
18 May 2021 - 17:00 Europe/London
Address

Online

Talk language
English
Slides language
English
Serie number
1
Description

(This seminar is organized by the IAU G5 commission on stellar and planetary atmospheres) 

Task-based computing is a method where computational problems are split
   into a large number of semi-independent tasks (cf.
   2018MNRAS.477..624N). The method is a general one, with application not
   limited to traditional grid-based simulations; it can be applied with
   advantages also to particle-based and hybrid simulations, which involve
   both particles and fields. The main advantages emerge when doing
   simulations of very complex and / or multi-scale systems, where the
   cost of updating is very unevenly distributed in space, with perhaps
   large volumes with very low update cost and small but important regions
   with large update costs.

   Possible applications in the context of stellar atmospheres include
   modelling that covers large scales, such as whole active regions on the
   Sun or even the entire Sun, while at the same time allows resolving
   small-scale details in the photosphere, chromosphere, and corona. In
   the context of planetary atmospheres, models of pebble-accreting hot
   primordial atmospheres that cover all scales, from the surfaces of
   Mars- and Earth-size embryos to the scale heights of the surrounding
   protoplanetary disks, have already been computed (2018MNRAS.479.5136P,
   2019MNRAS.482L.107P), and one can envision a number of applications
   where the task-based computing advantage is leveraged, for example to
   selectively do the detailed chemistry necessary to treat atmospheres
   saturated with evaporated solids, or to do complex cloud chemistry
   combined with 3-D radiative transfer.

   In the talk I will give a quick overview of the principles behind
   task-based computing, and then use both already published and still
   on-going work to illustrate how this may be used in practice. I will
   finish by discussing how these methods could be applied with great
   advantage to problems such as non-equilibrium ionization, non-LTE
   radiative transfer, and partial redistribution diagnostics of spectral
   lines.