Oak Ridge National Laboratory (ORNL) is working with Allinea Software to develop petascale debugging software, capable of handling 225,000 simultaneous processes and above, for next generation supercomputers.
The collaboration will begin by scaling up the capabilities of Allinea DDT (the Distributed Debugging Tool) on the Cray XT5 supercomputer, nicknamed Jaguar, which was installed in 2008 at the ORNL facility in Tennessee. Scientists at ORNL have already used the Cray system to set a new world record for a sustained performance of more than one petaflop (a quadrillion mathematical calculations per second) on two scientific applications.
The decision to select Allinea for the petascale debugging project was the result of an extended evaluation process. 'Our users liked the look and feel of Allinea's DDT debugging tool,' said Richard L Graham, Applications Performance Tools Group Leader, ORNL. 'We also knew that DDT was being used at several HPC computing centres, such as Lawrence Livermore Labs, NERSC, TACC, and several European supercomputing centres.'