Understanding flow-induced processes at a molecular level is a weighty challenge that could not be tackled without the use of supercomputing.
But this materials science research will impact many industrial applications in lubrication of gears, engine bearings and magnetic hard drives.
It is just one of the projects supported by the PaCER program, which helps researchers optimise their algorithms, codes and workflows for next-generation supercomputers such as Setonix.
PaCER – Pawsey Centre for Extreme Scale Readiness – is a collaboration between Pawsey and research groups across Australia. It allows researchers early access to Pawsey’s supercomputing tools and infrastructure, training and exclusive hackathons focused on HPC performance at scale.
The pending arrival of Setonix has elevated Pawsey to the front line of computational science and created a unique opportunity for the Australian research community.
Setonix is built on the same architecture used in world-leading exascale supercomputer projects in the US and Europe, and PaCER is working to ensure researchers workflows are exascale-ready for future requirements.
Ten research projects were successful in the first iteration, covering fields from radio astronomy to computational fluid dynamics, and the leap in compute power creates new opportunities to look at problems in more detail or process data at previously unavailable scale.
Dr Maciej Cytowski, Pawsey’s Head of Scientific Services and co-Chair of the PaCER Committee, said PaCER develops codes and techniques that will not only solve current defined challenges, but have significant impact on entire research domains.
“We are talking about extremely large-scale challenges that can be solved only through supercomputing, on the front-line of computational science,” he said.
Dr Cytowski said the program focuses on upskilling by building new expertise and high-performance computing software development capabilities between PaCER research groups.
“We have worked closely with all projects to identify their needs, in terms of programming tools and techniques as well as to understand details of their project plans,” he said.
“Based on this we have reached out to the international community through Pawsey’s network to facilitate training and presentations for PaCER researchers.
“The first sessions were in November, with AMD providing an introduction to the AMD CPU and GPU ecosystem, while HPE Cray provided us with a first look at the programming environment for Setonix.”
Between December 6 to 10, PaCER has organised an offering that includes lectures to share experiences on porting and scaling applications on AMD GPUs and panel discussion looking at:
- Programming models for Exascale
- ADIOS I/O library
- In situ visualisation
- Profiling and optimisation on Setonix
- OpenCL programming language
- Appentra software optimisation tools
“Many of the products created within PaCER will be made available to anyone in the HPC community,” Dr Cytowski said.
“The program is designed to upskill researchers to ensure that they are ready for Setonix and the next generation of supercomputers.”