ANU, Fujitsu partner on supercomputing software

Fujitsu and the Australian National University (ANU) have partnered to develop software for modelling tsunamis and fusion reactors on petascale machines.

The three-year project aimed to produce mathematical techniques for parallel processing on next-generation high performance computers (HPCs).

According to Fujitsu team leader Dr Ross Nobes, ANU tsunami modelling and plasma physics applications would be used as “test-beds” for new technology.

Findings would be used to optimise Fujitsu’s latest HPCs with multi-core nodes and complex communication networks, addressing challenges of scalability, robustness, and fault tolerance.

“Current algorithms do not have the required scalability nor robustness required by the next generation of supercomputers,” ANU Professor Markus Hegland explained.

“New algorithms are required to deal with the challenges posed of the large numbers [up to millions] of parallel computational threads, their synchronisation and communication between the threads.”

The project was titled ‘Robust numerical solution of partial differential equations on petascale computer systems with applications to tsunami modelling and plasma physics’.

It was funded by a $446,000 Australian Research Council linkage grant, matched by $150,000 and in-kind contributions from Fujitsu, including access to prototype hardware and staff.

Researchers planned to spend the funds on employing a postdoctoral research associate and two PhD students. The funds will also support staff exchanges, internships and targeted collaboration with Fujitsu.

The project was part of Fujitsu’s Open Petascale Libraries (OPL) project, which aimed to develop general-purpose and application-specific numerical libraries for petascale HPCs.