ABSTRACT:
The
enormous and growing complexity of today's high-end systems have
increased the already significant challenges of obtaining high
performance on today's equally complex scientific applications.
In this talk, we discuss the role of compiler technology in supporting
application developers in a systematic approach to performance tuning
of key application computations. Based on scenarios taken from
development of scientific codes, we describe how compiler support can
enable the programmer to achieve the same or better performance result
in a much more productive way. We also examine the very structure
of today's compilers, which are currently monolithic and difficult to
retarget to new architectures, and propose a systematic and principled
strategy to compiler design and optimization.
BIO:
Dr. Mary Hall is currently a Project Leader at Information Sciences
Institute and jointly a Research Associate Professor in Computer
Science at University of Southern California, where she has been since
1996. Her research focuses on compiler support for high
performance, targeting a variety of architectures from
Field-Programmable Gate Arrays (FPGAs) and multimedia extension
architectures, to shared-memory multiprocessors and memory hierarchies
of high-end systems. Dr. Hall has published over 70 articles and
served on over 35 program committees in compilers and their interaction
with architecture, parallel computing, and embedded and reconfigurable
computing, including the 2005 program chair of the ACM SIGPLAN PLDI
conference, and the poster chair of SC'05.
# # #