Supercomputing Rapidly Advancing Aerospace Engineering Written 29 June 2015

Panelists: Moderator Doug Cline, Lockheed Martin Corporation; William Gropp, University of Illinois at Urbana-Champaign; Brian Mitchell,senior principl engineer, GE Global Research; and Mark Seager, chief technology officer, High Performance Computing Ecosystem

by Lawrence Garrett, AIAA web editor

Panelists in the discussion, "Supercomputing: Roadmap and its Future Role in Aerospace Engineering," on Thursday, 25 June, at the 2015 AIAA AVIATION Forum.

Supercomputing has made significant contributions in aerospace engineering in recent decades, including advances in computational fluid dynamics that has fundamentally altered the way aircraft are designed.

And the relentless growth in high-performance computing power holds promise of huge leaps in engine performance and other aerospace technology.

That was a key message from a group of computer technology experts last Thursday at the AIAA AVIATION 2015 forum.

The panelists on stage, noted moderator Doug Cline of Lockheed Martin, represented “maybe 120-man years of aero-computing experience.” 

Mark Seager, chief technology officer of high performance computing ecosystem at Intel, noted that the Department of Energy’s (DOE) Argonne National Laboratory recently picked Intel to build the world’s biggest supercomputer. That machine is expected to have processing speed of 180 petaflops — capable of handling thousand trillion operations a second. 

“This is a system that’s designed to go after a broad swath of science and engineering challenge problems,” Seager said. “And the science that they’re actually interested in aren’t just the traditional DOE sciences; but also include biological sciences and renewable energy.”

Thanks largely to the “High Performance Computing Act” signed by President George H.W. Bush in 1991, Cline said, the high-performance computing system in the Unites States has “seen basically seven orders of magnitude performance increase” between 1992 and 2012.

“There’s hardly any other technology out there that has actualized this kind of price for performance improvement,” Cline said.

Federal funding for the program has increased every year. The Obama administration requested a little over $4 billion in its 2016 budget, a “large chunk” of which Cline said would go to high-end computing systems.

Brian Mitchell, senior principal engineer at GE Global Research called supercomputing technology a “key enabler as we work to improve our products to get radical improvements in performance.”

Mitchell said GE has been doing supercomputing for some 40 years, and that “the roots of computational fluid dynamics go back at least that far in the company.” Yet even after several decades, he said, the field remains very dynamic.

Mitchell said that spurred by competitive pressures, GE has seen a “wholesale change in the engine portfolios” that they’ve been fielding and that almost all of the engines that they manufacture are “in the process of being changed out by new radically improved technology.” 

William Gropp, director of the Parallel Computing Institute at the University of Illinois at Urbana-Champaign, said supercomputing hold tremendous opportunities as well as great challenges.

Referring to a “CFD Vision 2030 Study: A Path to Revolutionary Computational Aero Sciences,” Gropp said that systemcomplexity will increase and that obtaining ever-increasing performance improvements is “becoming more and more complicated.”

Gropp cautioned that a quest for a single universal software solution — “perhaps the single biggest reason that we have a software crisis’ in high-performance computing — is focused on the wrong problem, and it has cost us a tremendous amount of time and progress.”

Gropp cited three key challenges ahead forsupercomputing. 

First, he said, it must "match and work with features of the architecture.” 

In addition, algorithm changes are needed, with “higher-order approximations in numerical codes; more compact representations in general; and chained vector operations.” Finally, he said, “We need better software for realizing these algorithms,” adding that “we need to be able to give the experts the tools that they need to build fast codes,” while noting that “those [required] people are scarce.” 

Video

 

All 2015 AIAA AVIATION Forum Vidoes