Items to be filled in on the computing form:
Name of project:
Astrophysical Turbulence and Dynamo Action
Project numbers (see https://supr.snic.se/):
-
naiss2023-3-23 (2023) + naiss2023-4-20 (disk)
snic2022-3-23 (2023) + snic2022-4-20 (disk)
snic2021-3-29 (2022) + snic2020-4-12 (disk)
SNIC2020-3-30 (2021) + snic2020-4-12 (disk)
SNIC2019-2-22 (2020)
SNIC2018-2-41 (2019)
SNIC2017-12-57 (2018)
SNIC2016-34-38 (2017)
SNIC2015-16-51 (2016)
SNIC2015-1-189 (2015b)
SNIC2014-11-34 (2015)
SNIC2013-26-32 (2014)
SNIC025-12-36 (2013)
SNIC020-11-29 (2012)
SNIC014-10-2 (2011)
→ nscquota
Abstract:
We use the Pencil Code (http://pencil-code.googlecode.com),
which is a sixth order finite difference code with a third order time
stepping scheme.
The code uses MPI and is running on a range of different platforms
around the world and is designed to work on large Linux clusters.
Our production runs tend to have 512^3 meshpoints and
require at least 64 but preferentially more processors.
A typical run requires at least 250,000 time steps.
In previous studies we used several runs.
With 0.075 microsec per meshpoint and per timestep, this means 30 days of
wall clock time (using multiple restarts) for just one run.
This corresponds to a total of about 50,000 CPU hours.
In order to address properly the critical question of the dependence on the
magnetic Reynolds number we now have to move to runs with 1024^3 and
2048^3 meshpoints, which would require 400,000 CPU and 1,600,000 CPU hours,
respectively.
In order to satisfy these demands we estimate a use of about 80,000 CPU hours
per month.
Estimated maximum disk usage during the project time
100 GB for permanent storage (currently 20 Gb), and
1000 GB for temporary storage (currently 1000 Gb).
Estimated job memory usage per core (in Mbytes)
Estimated number of cores per job
http://nhpc.hi.is/content/nhpc-hardware
http://www.snic.vr.se/apply-for-resources/available-resources/c3se/?searchterm=glenn
Akka 100 x 1000 core hours/month
Abisko 200 x 1000 core hours/month
Lindgren 50 x 1000 core hours/month
Nordic HPC 50 x 1000 core hours/month (Iceland, managed by Uppsala)
Glenn 50 x 1000 core hours/month (Goethenburg)
Jukka Komminaho
NHPC, called Gardar (Iceland machine)
manage SNIC projects
Norweb
$Date: 2024/01/03 09:49:34 $, $Author: brandenb $, $Revision: 1.26 $