Since 1987 - Covering the Fastest Computers in the World and the People Who Run Them

January 14, 2014

Crash Testing at Scale

Tiffany Trader

Over on Cray’s blog site, the company’s manufacturing segment manager, Greg Clifford, writes about the performance of explicit structural applications at scale as a follow-up post to a blog on extreme scaling in CAE applications,” which shows such applications successfully scaling to over 10,000 cores.

The new entry is concerned with the scaling performance of explicit structural analysis applications – aka “crash simulation” applications. Clifford notes that there has been increased emphasis on this application set which has led to significant improvements.

Some of the more popular explicit structural analysis codes include Abaqus/Explicit, LS-DYNA, PAM-CRASH, and RADIOSS. The codes are used across a wide range of industries, including automotive, aerospace, consumer products, materials industry and for many defense purposes. Because of this wide user base, Clifford contents that these explicit codes take up the largest number of high performance computing (HPC) cycles within the computer aided engineering (CAE) field.

The explicit codes became popular in the mid-1980s when they were mainly run on Cray vector computers to simulate automotive crashes. The subsequent growth and reliance on crash and safety simulation cemented the use of HPC in the automotive sector. Clifford maintains that while industry originally saw HPC as a way to reduce the number of physical crash tests, saving time and money in the process, the motivation has shifted to become increasingly safety-focused.

“The main benefit,” observes the Cray rep, “is that crash and safety simulation has enabled auto companies to meet increasingly demanding safety requirements and produce much safer vehicles and save thousands of lives every year.”

But these safety requirements are not static. There are new and increasingly complex crash test mandates as well as pressure to reduce vehicle weight. Addressing these realities in a cost-competitive way set the stage for HPC to become a core technology, necessary for running larger workloads and supporting more computationally complex simulations. Rising to this challenge requires improved parallel scaling.

“The initial MPI implementation of the crash codes gave a significant improvement in parallel performance,” writes Clifford. “The SMP version would typically scale well to four- to eight-way parallel, but the MPI version would go to dozens of cores, providing a large boost in performance over the SMP option. To scale to hundreds of cores, Amdahl’s Law requires that well over 99 percent of the computations scale efficiently. Explicit applications have grown into very large, general purpose programs with thousands of simulation options. Thus, it has been an ongoing effort over the years to improve the scaling throughout the application. This has been especially challenging in the “contact” area, but the complexity of the codes in general makes scaling to thousands of cores a project that requires a team effort.”

Clifford goes on to detail the work that Cray has undertaken in partnership with Livermore Software Technology Corporation (LSTC) to increase the scaling performance of LS-DYNA on thousands of cores. He says the goal is to get real production models running 10 to 100 times faster, a feat that requires running large complex simulations and profiling the performance on thousands of cores. The experiments can pinpoint the parallel bottlenecks and reveal what parts of the application need attention. There have already been “impressive results,” including a crash simulation model that scales well to over 4,000 cores, while most crash simulations use fewer than 100 cores. In another example, a large aerospace model is scaling to over 11,000 cores.

LS-DYNA-Performance 800x

“The goal,” says Clifford, “is to make scaling to thousands of cores the norm for large production impact simulations.” Cray plans to continue working with automotive and aerospace communities to obtain real models for testing at large core counts. “This cooperation between the application developers, key users and Cray is a proven method for enhancing performance and driving simulation to the next level,” according to the Cray rep.

Tags: , , ,
Share This