Unique code-generating software makes weather and climate forecasting easier
Researchers at the UK’s Science and Technology Facilities Council (STFC) are demonstrating a software framework that can automatically generate computer code which enables scientific applications to run on supercomputers.
Dr Mike Ashworth from STFC’s Scientific Computing Department will showcase PSyclone, a software framework that automatically generates the parts of the code necessary to run on supercomputers. He will present the work of the STFC team that developed it - Rupert Ford, Andrew Porter, and Karthee Sivalingam - at the 2016 MultiCore 6 workshop on 13-14 September in Boulder, Colorado.
PSyclone was developed for the UK Met Office and is now a part of the build system for Dynamo, the dynamical core currently in development for the Met Office’s ‘next generation’ weather and climate model software.
By generating the code needed to make use of thousands of processors, PSyclone leaves the Met Office scientists free to concentrate on improving the accuracy and precision of the model and not how to run it efficiently at scale. This means that they will not have to change their code from something that works on a single processing unit (or core) to something that runs on many thousands of cores.
‘This is quite a radical approach to the challenge of running models on different kinds of computer architectures,’ said Dr Ashworth. ‘It significantly reduces the potential for errors and will make weather forecasting and climate modelling simpler and more efficient on the high-performance computers of the future.’
It will be of particular interest to those developers who focus on trying to adapt such models for novel and emerging highly parallel computer architectures’ concluded Ashworth.
PSyclone can users to parallelise and optimise the code for a particular architecture. Its ability to generate both MPI- and OpenMP-parallel code has enabled the UK Met Office’s prototype dynamical core to go from serial execution to running on more than 50,000 cores with no change to the scientific code at the base of the application.
Dr Ashworth commented: ‘We are now exploring opportunities to make the PSyclone software framework available for other areas of science and technology, and would welcome discussions with anyone interested in using it.’
Hosted by the US National Center for Atmospheric Research (NCAR), the 2016 MultiCore 6 workshop provides a forum for open discussion to better understand the application of new high-performance computing technologies for the next generation of weather, climate, and earth-system models. The new generation of high-performance computer architectures present significant challenges to the communities working on these models.
PSyclone’s development was funded by the STFC Hartree Centre as part of the GungHo project, part of a collaboration between STFC, the UK Met Office and the Natural Environment Research Council (NERC).