All seminars are free and open to the public.
Wednesday, March 21,
2007, 2:10 PM, 210 Neville Hall
CMT: THE NEW GENERATION MICROPROCESSOR ARCHITECTURE
Dr. John Heath
Consultant
ABSTRACT:
Traditionally, microprocessor designs have focused on improving single thread
performance, and dramatic performance gains have been achieved using multiple
parallel instruction units, out-of-order scheduling, deep pipelines, and
aggressive branch prediction. This approach, however, seems to have
reached the point of diminishing returns. Recent attempts to leverage
these techniques to achieve demonstrable gains on today's commercial workloads
have not been as successful. The cause is the relatively slower improvements in
memory latency; while microprocessor performance has doubled every two years,
memory speeds have doubled only every six years. As a consequence, the
performance bottleneck has shifted from the microprocessor to memory
latency. The result is that pipeline utilizations of today's
microprocessors are typically as low as 20%.
A new microprocessor design has been developed to overcome the memory latency
bottleneck and provide a continuation of the exponential performance gains seen
over the last three decades. This new microprocessor design, called Chip
Multi-threading (CMT), combines chip multiprocessing and thread-level
parallelism. For many applications CMT processors deliver several times
the throughput of comparable single-threaded processors, and require only half
the power of comparable systems. This new design has been exploited in a
Sun Microsystems' new 32-way chip-multithreaded microprocessor, called Niagara. This talk gives an overview of CMT design and the Niagara
architecture. We also discuss applications that have achieved very high
performance with CMT. Among these are genomic sequence analysis, TCP/IP
processing, and web serving.
BIOGRAPHY
Dr. Heath holds a BA and MA in Mathematics from the University of Maine, and an M.Sc. and Ph.D. in Computer and Information Science from the University of Minnesota. He taught in the Computer Science Department of USM from 1976 until
2000, at which point he went to SUN Microsystems where he stayed until 2006.
From 2004 until 2006 he was president of the International Communication
Sciences and Technology Association which organizes 25 international
conferences each year in the field of communication sciences and technology.
Wednesday, March 28,
2007, 2:10 PM, 210 Neville Hall
QUICK OR DEAD: ORGANIZATIONAL VELOCITY FOR AN IMPATIENT AGE
Tom DeMarco
Principal, Atlantic Systems Guild
ABSTRACT:
Tom DeMarco offers a new and chilling look at the way slow organizations work
and often fail to work, as well as a practical prescription to turn them around
and make them hum. He confronts major complaints of this impatient age: late
projects, work forever in catch-up mode, the paradox that in an era of
"Hurry Up," companies are slowing down, sometimes almost grinding to
a stop. Most of all, he develops a unique approach to achieving true
organizational velocity.
BIOGRAPHY
Tom DeMarco is a Principal of the Atlantic Systems Guild, and a Fellow of the
Cutter Consortium. He was the winner of the 1986 Warnier Prize for
"lifetime contribution to the field of computing," and the 1999
Stevens Award for "contribution to the methods of software
development."
His new book (with co-author Tim Lister), "Waltzing with Bears:
Managing Risk on Software Projects", was published by Dorset House in
March, 2003. This is the ultimate how-to book about risk management. In its
five parts it guides you in building a case for risk management, protecting
yourself against its possibly dangerous impact in a politically unready
environment, making the mechanics work, and making sure you derive all the
benefits, and testing your organization for risk readiness.
His previous business book is called "Slack, Getting Past Burnout,
Busywork, and the Myth of Total Efficiency." It was brought out by
Random House in 2001 and in a paperback edition in 2002. Slack answers the key
questions, “Why are all so damn busy?” and “Is it good for us and for
the companies we work for?”
Prior works include “The Deadline: A Novel About Project Management”,
published in the summer of 1997 by Dorset House. It is the story of a veteran
software manager who bets his life on a delivery date. How does he manage the
project with the stakes so high?
The classic, “Peopleware: Productive Projects and Teams” (with co-author
Tim Lister) is now out in its second edition from Dorset House, Tom's 1995 book
of essays was entitled “Why Does Software Cost So Much?, And Other
Puzzles of the Information Age”, also from Dorset.
Early works include, “Structured Analysis and System Specification”
[Prentice-Hall, 1979], “Controlling Software Projects: Management,
Measurement and Estimation,” [Prentice Hall, 1982], and more than one
hundred articles and papers about management and the system development
process. In 1990, he served with Tim Lister as editor of “Software: State of
the Art” [Dorset House, 1990].
Beginning his career at Bell Telephone Laboratories, he was part of the cutover
team of the now-legendary ESS-1 project. In later years, he managed real-time
projects for La CEGOS Informatique in France, and was responsible for
distributed on-line banking systems installed in Sweden, Holland, France and Finland. He has lectured and consulted throughout the Americas, Europe, Africa, Australia and the Far East. His particular areas of interest these days are project management,
change facilitation, and litigation of software-intensive contracts.
Tom DeMarco has a BSEE degree from Cornell University, an M.S. from Columbia University and a diplome from the University of Paris at the Sorbonne. He is a member
of the ACM and a Fellow of the IEEE. He makes his home in Camden, Maine.
For more information, visit http://systemsguild.com/GuildSite/TDM/TDMBio.html.
Thursday, April 26, 2007, 2:10 PM, 227 Neville Hall
UNCOMPUTABLE
NUMBERS: TURING'S 1936 PAPER REVISITED
Greg Chaitin
IBM Research
It is usually forgotten that the key point of Turing's famous 1936 paper "On computable numbers with an application to the Entscheidungsproblem" was to exhibit an uncomputable real number. In fact, Borel in 1927 had already done this. I'll discuss the relationship between the ideas of Turing and those of Borel and also my own uncomputable real number, the halting probability Omega.
Reference:
http://cs.umaine.edu/~chaitin/mjm.pdf
BIOGRAPHY
Gregory Chaitin is at the IBM Watson Research Center in New York, and is
the discoverer of the remarkable Omega number. He is the author of ten
books on mathematics, including a book for the general public, "Meta
Math!," currently available in hardcover and softcover. For more
information, see
http://www.umcs.maine.edu/~chaitin/bio.html
Tuesday, May 1, 2007,
2:10 PM, 123 Barrows Hall
SECURING THE HOMELAND FROM ITSELF: THE PRECARIOUS FATE OF SOUTHERN LOUISIANA
Joshua Anchors
FEMA Consultant
ABSTRACT:
Southern Louisiana is disappearing. Every thirty-eight minutes, one football
field worth of coastal wetlands washes away into the Gulf of Mexico.
Twenty-four square miles of land are lost every year. Two hundred and seventeen
square miles lost during Hurricane Katrina alone. By 2050, scientists predict
that 500 additional square miles will be lost. New Orleans will be extremely
vulnerable to hurricanes. Millions of Louisianans will be displaced. One of our
nation's most diverse and delicate bioregions will be irrevocably lost.
Southern Louisiana’s coastal land loss is one of the most significant
environmental crises that the U.S. faces today. Using New Orleans as a case
study, this presentation will review the history of coastal land loss in Louisiana and will address the recent developments to restore and protect Louisiana's
coast.
BIOGRAPHY
Joshua Anchors has worked with the Federal Emergency Management Agency on
coastal restoration and protection issues in Louisiana, Mississippi, and Alabama for the last eighteen months. He is the author of “Regarding Hwange and Other
Matters of Perception” and the co-editor of the forthcoming anthology “Watershed
Literature: Writings from the Gulf of Maine”. Mr. Anchors received his BA
and MA in English from the University of Maine.
Thursday, May 3, 2007, 2:10 PM, 227 Neville Hall
ENGINEERING THE PRACTICAL USE OF AUGMENTED REALITY: AR MACRO AND MICRO NAVIGATION TECHNOLOGIES, HUMAN FACTORS AND AR COMPUTING SYSTEMS
Vince Quintana
Engineer, General Dynamics
ABSTRACT:
Augmented Reality Systems will be the final step at which the full promise of contextually relevant, real-time computer data is provided to a user ubiquitously and on demand. For many emergency responders, that information must be situationally accurate, correctly displayed and constantly updated to ensure user confidence especially in adverse environments. The development of computer systems, navigational aids, feducial systems and peripherals that lay outside the Personal Area Network (PAN) will be the backbone of this effort. This lecture will concentrate on methods and systems developed to date, and the developments that must be achieved to take this technology to its true realization, and the benefits it will provide.
BIOGRAPHY
Vince Quintana is currently the principle test engineer for Human
Systems Integration on the US Navy's DDG 1000 Project. He is retired
from the US Navy, a graduate of the University of Southern Illinois,
and a post graduate student at UMaine. He holds several patents for
the development of Human Borne Computing systems and specializes in
the development of those systems for extreme terrestrial
environments. He is supporting several Maine companies in the
development of Augmented Reality systems and applications for First
Responder and military personnel.