The easy-to-use debugging tool is installed on Canada's fastest supercomputer, hosted at Toronto's SciNet, the country's largest supercomputer centre. The facility has a multi-platform system that combines IBM Power, x86, and Blue Gene/Q architectures plus GPUs.
Dr. Daniel Gruner, CTO-Software, SciNet, commented: "Allinea DDT is widely used here. It was the only tool that fit the bill with our multiple platforms and GPUs and we've had great success with it. We have been able to quickly find bugs in user codes, and even in system libraries."
SciNet has also purchased an Allinea MAP license for its x86 systems. Allinea MAP uses the same interface as Allinea DDT so scientists can analyze code performance without having to learn a new tool.
With the addition of the new Blue Gene/Q machine, SciNet now has almost one petaflop of computing power. The new machine has 32,768 cores and sits 77th out of the world's top 500 fastest supercomputers, as of June 2013.
This new supercomputer will give Canadian researchers enough compute power to tackle previously unattainable Big Data challenges.
"Canada needs more knowledge-based industries to diversify our national economic portfolio beyond the current over-weighting of commodities and natural resources, and to help eliminate our identified innovation gap", stated Dr. David Naylor, president of the University of Toronto.
When bugs and performance issues arise, SciNet staff and Canada's top researchers can be confident they are working with world-class tools - Allinea DDT and Allinea MAP.
"SciNet have a highly respected software team which is playing a major role in developing the skills and software needed to exploit HPC in the national interest", explained David Lecomber, COO of Allinea Software. "Our software tools, through the unified interface for debugging and profiling with Allinea DDT and Allinea MAP, are able to play a significant part in enabling both the skills to be acquired quickly and the supercomputers to deliver results."
You can read a case study on the work with SciNet.