We have a growing group of modellers using the Ice-sheet and Sea-level System Model (ISSM) in Australia. It’d be great to have a system-wide build of ISSM - accessible to any gadi users - that is maintained and optimised. It’d also be great to get some support from ACCESS-NRI with benchmarking of ISSM for parallel performance.
We’ve had some issues with recent versions of ISSM that don’t seem to be behaving well on gadi with the system petsc modules. I’d be happy to share the latest install script that we’ve used for installing ISSM on gadi, and hopefully go from there.
ISSM is just one of a few ice sheet models used by the Australian community, so I recognise that this is not an issue unique to our group, and that there might be other ice sheet modellers out there with a similar need.
Cheers,
Felicity
2 Likes
Aidan
(Aidan Heerdegen, ACCESS-NRI Release Team Lead)
2
That sounds like a good idea and something ACCESS-NRI would like to be involved with, subject to resource constraints of course.
First step is to create an issue on our spack_packages repository outlining what you need, including links to the source code repository and your build script if that is available.
This repository is where we publish our spack package definitions which we used to build software on gadi.
There is already a PETSc package included in spack. This means it can be built with spack, but I was trying that a while back and ran into some issues, and PETSc involves quite a bundle of software, so it may not be straightforward.
The reason I was trying to build petsc was to collaborate with the Underworld folks who build petsc and wrap it all in a container to deploy on gadi. So we might have even more reason to collaborate with them if we can help each other, and they have a great deal of experience with petsc scaling.
Hi @Aidan, thanks for the input on the github page. I’m wondering about how best to proceed with this installation of ISSM - are you able to provide guidance about install location for ISSM and the dependencies? Thanks heaps
Aidan
(Aidan Heerdegen, ACCESS-NRI Release Team Lead)
6
We’re in a difficult position at the moment as far as this is concerned, as we are still in a position of working through the best way to do some of these ourselves.
I would be worried that suggestions might not be well tested, or not optimal, and end up wasting more time than it saves.
Do you have a specific timeframe you need for this work to be complete?
Thanks for getting back to me @Aidan. We have no specific timeframe - the sooner the better would be great, but we definitely don’t want to waste time or resources so happy to wait for your advice. Thanks very much!