[Nrg-l] Special guest talk - November 26 @ 3pm

Jorge Londoño jmlon at cs.bu.edu
Tue Nov 20 10:47:14 EST 2007

Locality and Parallelism: Old Problems in a New Era

Chen Ding
University of Rochester and MIT


Increasingly in the past the most time and energy of computing systems
are consumed by storing and transferring data.  Today, memory
performance is a fundamental constraint limiting the scalability of
multi-core processors.  As a software solution, traditional program
analysis alone is increasingly inadequate as the compound memory usage
of computer programs is not amenable to modular analysis.  In this talk
I will present program behavior analysis, which uses trace- based or
on-line models to analyze composite and emergent patterns across tens of
billions of program actions.  The basic metrics of locality are the
frequency and distance of data reuse.  In recent years various
distance-based models have emerged as a unified basis for understanding
and controlling large-scale program behavior at program, system, and
hardware levels.  I will describe the five dimensions of locality and
their use in program transformation and memory management.  Recently in
collaboration with Professor Xipeng Shen at William and Mary we have
been developing a behavior-based interface for dynamic parallelization.
In the remaining part of the talk I will discuss our experience in
parallelizing large, existing applications on a multi-core,
multi-processor PC.


Chen Ding is an Associate Professor in the Computer Science Department
at the University of Rochester and presently an Visiting Associate
Professor in the EECS Department at MIT.  He is an receipient of the
Early Career Principal Investigator award from DoE, the CAREER award
from NSF, several CAS Faculty Fellowships from IBM, and a best-paper
award from the IEEE International Parallel and Distributed Processing
Symposium.  He co-founded the ACM SIGPLAN Workshop on Memory System
Performance and Correctness (MSPC) in 2002 and organized the Workshop on
High-Level Parallel Programming Models and Supportive Environments in
2007.  Between February and August 2007, he was a visiting researcher in
the Run-time Analysis and Design group at Microsoft Research.  More
information about his work can be found at

Have a nice thanks-giving!


More information about the Nrg-l mailing list