Researchers at University College London (UCL) modeling intracranial blood flow have been able to reach a major milestone by using development tools from Allinea Software. According to a Allinea news release, a massive simulation using 50,000 processor cores on the UK’s largest supercomputer, ARCHER, became possible after solving an application crash that only occurred at this extreme scale.
UCL’s HemeLB software applies computational fluid dynamics to model blood flow around cerebral vessels and simulate pressure at points of weakness such as aneurysms. Using data from MRI scans of a patient’s blood vessels, the group anticipates that simulations will one day help to decide the best clinical option for individual patients. The complexity of the simulations requires the UCL team to use ARCHER, the UK’s flagship Cray XC30 system, which is managed by Edinburgh Parallel Computing Centre (EPCC) on behalf of EPSRC and other UK research councils.
The UCL team used Allinea Software’s performance profiling tool, Allinea MAP, to increase the performance of HemeLB on the system. Having already improved performance on some test cases by over 25% they wanted to run a larger simulation.
We’d never been able to look at this many cores – and get a clear view of how the time was being used – we were keen to see it in Allinea MAP,” UCL Post-doctoral Researcher Derek Groen said. However, the application crashed when using 50,000 processor cores, which stopped them in their tracks.
The crash was totally unexpected. I didn’t know how I would diagnose or fix it at that scale – it was beyond anything I had tried to do before.” says Groen. “Allinea Software helped us straight away – they knew that if we could run the simulation with their debugger, we would find the problem.”
The UCL team and Allinea Software were brought together by the EU CRESTA project – which is preparing applications for future extreme-scale computing. Using Allinea DDT, they resolved the problem quickly by debugging all 50,000 application processes simultaneously.