Allinea MAPs Out New Performance Tool, Crowd-Sources Design

Nov. 12, 2012 - PRLog -- SALT LAKE CITY, Utah, 2012-11-12 - Allinea Software shifted the performance landscape at SC12 today with the announcement of a new performance analysis tool, Allinea MAP (http://www.allinea.com/products/map/).

Better known for their high-performance, scalable parallel debugger, Allinea DDT (http://www.allinea.com/products/ddt/ddt-features/), the company has taken an unusual tack in developing the new product.

"We worked with a range of leading HPC labs right from the word go - we sat down together with a rough whiteboard sketch and tried to figure out what we could do to get more people profiling their codes", recalls Mark O'Connor, Product Manager at Allinea Software.

This collaborative process resulted in some rather unique features - Allinea MAP eschews the classic instrumentation-based MPI timeline in favor of a dynamic sampling engine that claims to scale tens of thousands of processes whilst adding just 5% to the total runtime.

"We were able to cheat a bit by building on top of Allinea DDT's infrastructure," admits David Lecomber, CTO at Allinea Software, "we already had an infrastructure to launch and merge data proven at 275,000 processes. This gave us a common visual interface over the entire Allinea environment as well as saving development time, which let us focus much more on the user experience."

That user experience has been turning some influential heads. "I really liked seeing the performance information directly in the source code," said Helen He of the National Energy Research Supercomputer Center (NERSC) (http://www.nersc.gov/) at Lawrence Berkeley National Laboratory, Department of Energy (http://www.lbl.gov/). "It's very cool," agrees Richard Gerber, Deputy Group Lead for the NERSC User Services Group, "we need a tool that helps users to help themselves."

The rallying-cry around a more usable, reliable profiling tool was repeated across the Atlantic. "When we profile code, the simplest metrics are sometimes the trickiest to collect. A lightweight tool like Allinea MAP lets us get this key performance data faster," said Steven Jarvis (http://www2.warwick.ac.uk/fac/sci/dcs/people/stephen_jarvis/), head of the High Performance Systems Group at Warwick University, England, whose group reviewed and advised on several iterations of Allinea MAP.

Allinea's community-led experiment hasn't stopped with the release. Customers who sign up during the rest of SC12 (http://sc12.supercomputing.org/) are receiving exclusive invitations into an extended development phase to fit the product to their users' needs. Indeed, with Allinea DDT announcing support for Intel Xeon Phi (http://www.intel.com/content/www/us/en/high-performance-computing/high-performance-xeon-phi-coprocessor-brief.html), several key sites are expecting Allinea MAP support to follow in short order.

"Developing a product in parallel with key labs like this has been an experiment for us," remarks Mark O'Connor, "but parallelizing a process to get better results faster? That's what we do in HPC all day long."