Adaptive Supercomputing

Luiz DeRose

Cray, Inc.

April 19th, 12-1pm
FAB 310-18

ABSTRACT

Scientific applications have become much more complex, increasing the demand for better performance, which can no longer be achieved through conventional processor improvements and a one-size-fits-all mentality. The challenges in today's high performance computing market are driving changes in the way supercomputers should be designed. Application developers need systems that can adapt to the application needs, rather than having to constantly redesign their applications to adapt to the systems. In this talk, I will discuss Cray's adaptive supercomputing strategy, which integrates scalar processing, vector processing, multithreading and hardware accelerators in a single high-performance computing platform. These "adaptive supercomputing" systems will be supported by intelligent compilers, performance tools, and runtime systems, that can examine an application, determine which processing technique will work best with it and then handle the application accordingly; thus providing a single computing resource that can solve multidisciplinary and multiscale problems, while also achieving high levels of performance and scalability.

BIOGRAPHY

Dr. Luiz DeRose has a Ph.D. in Computer Science from the University of Illinois at Urbana-Champaign. He is a Senior Principal Engineer and the Programming Environments Director at Cray, Inc., where he is responsible for the software development environment strategy for all Cray systems. Before joining Cray in 2004, Dr. DeRose was a research staff member at IBM Research where he focused his efforts on designing and developing performance tools for high performance computing (HPC) systems. In particular, Dr. DeRose was responsible for developing the Hardware Performance Monitoring (HPM) Toolkit, which is used worldwide on IBM HPC systems. Dr. DeRose has more than 20 years of experience in high-performance computing, working in the area of software support for high-performance computation. He participated in the design and development of performance tools, such as FALCON, SvPablo, SIGMA, the HPM Toolkit, the CrayPat Performance Collector and the Cray Apprentice2 Performance Analyzer. With a deep knowledge of the HPC programming environment, Dr. DeRose has published more than 40 peer- review articles in scientific publications, primarily on the topic of compilers and performance tools over the span of his career.

HOST
Dr. Karen Karavanic