12-05-2017 | Jacob Alldredge: High Performance Computing Architectures Use at APL for Modeling

Title: HPC Incubator Speaker Series: High Performance Computing Architectures Use at APL for Modeling

Speaker: Jacob Alldredge, Senior Scientist, Johns Hopkins Applied Physics Laboratory

Date: Tuesday, December 5, 2017

Location: NASA/LaRC, Bldg. 1205, Room 222

Time: 10:00am-11:00am

Sponsor: Cara Campbell Leckey, NASA/LaRC

Abstract: There has been a rapid evolution in computing architectures in recent years. This has mainly been a response to the leveling off of single threaded performance due to fundamental technology limitations. In response to this there are now a wide range of architectures and programing tools from GPUs, to Xeon Phis, to FPGAs. Due to the diversity modeling tool companies have been slow to embrace many of these architectures, while they wait for the market to stabilize. As such, there has been a need for individual researchers to develop their own tools to take advantage of specific architectures that are suited to the problem at hand.

At the Johns Hopkins Applied Physics Laboratory we have explored several of these architectures to tackle a variety of problems that are of interest to national security. We have looked into different architectures and algorithms for fluid dynamics, heat transfer and finite element modeling. These algorithms and work on their applicability to different parallel architectures, have led us to develop specialized models for a variety of applications. First and foremost is our work on in situ modeling and monitoring of additive manufacturing process for non-destructive evaluation. Here I will talk about our various applications, lessons learned and our general philosophy that we try to follow when tackling projects.