News on FGCS Technology and related activities by Research Institute for Advanced Information Technology (AITEC), the successor of ICOT
cybernetics: comparative study of the automatic control system formed by the nervous system and brain and by mechanical-electrical communication systems (Hypertext Webster Interface)
This first KLIC Programming Contest was held for the purpose of promoting
the KL1 parallel logic programming language and its associated programming
environment, KLIC. The members of the contest committee included the
developers of KL1 and KLIC at ICOT, as well as users, and others asked to
help with the contest.
As related in previous issues of the AITEC NEWS, we received a total of 77
programs as follows:
Category 1 (Sequential Environment): 54 programs
Category 2 (Parallel Environment): 15 programs
Category 3 (Free Subject): 8 programs
We had a total of 57 applicants (excluding ones who applied to more than
one category and including group applicants).
Committee members as well as contest staff, who evaluated applications
programs in each category, attempted to do so with both accuracy and
fairness. They measured elapsed times for a given problem, compared the
`intelligence' of the various programs, and evaluated the documentation.
Special measures were required to be undertaken for programs entered in the
parallel environment category; a long time was required for program
analysis, and for acquisition of accurate data for evaluation by the
addition of test programs.
In this issue, we'll announce the contest results as judged by the
committee. Please visit our web page and click on the "cyber" award
ceremony for details. The URL for our web page is: http://www.icot.or.jp
(Article No. 7-1)
2. Activities of Working Groups in the Department of Future Technology
Studies (FTS)
=======================================================
We will report on the activities of two working groups from the Department
of Future Technology Studies (FTS). This department is engaged in investigation
of leading-edge information technology and the publication of research
reports. These reports assist government with the drafting of new
strategies and policies, as well as helping business leaders with their R&D
planning.
In carrying out these objectives, the Future Technology Studies Department
gathers and analyzes data on new Information Technologies (IT) being
developed around the world. Particular comparisons are made between
Japanese projects and research projects of foreign countries.
At the moment, the department has organized two working groups: the
"Petaflops Machine and super computing Technology Research Working Group
(PFM-WG)", and the "Network and AI-related New Technologies Research
Working Group (NAI-WG). Research on new information technologies conducted
by these working groups will increase in importance in future when
parallel processing, distributed computing, and network computing become
more prevalent than they are at present.
(Article No.7-2)
3. Report on IEEE Symposium "Frontiers '96" at Annapolis
=========================================================
As we reported in the last AITEC NEWS headlines, two researchers from the
FTS Department participated in the "The Sixth Symposium on Massively
Parallel Computation" subtitled "Frontiers '96" from October 27 to 31 at
Annapolis, Maryland in the U.S.A. Mr. Takahari, one of two conference
participants will detail his business trip to the U.S.A, reporting on the
conference themes, the general atmosphere as well as his impressions
"Frontiers '96."
(Article No.7-3)
This first KLIC Programming Contest was held for the purpose of promoting the KL1 parallel logic programming language and its programming environment, KLIC.
The contest consisted of three categories:
Category 1: Sequential Environment
Category 2: Parallel Environment
Category 3: Free Subject
We had 77 programs from a total of 57 applicants by the application
deadline, November 17.
Ex-ICOT members, who had been involved in the development of KL1 and KLIC,
as well as users were urged to join in support of the contest committee, to
help administer the contest.
Not only program elapsed time, but also quality of documentation (including
manuals) were judged by committee members and other technical staff.
Originally, judging was to have been finished by December 11. However, for
Category 2, Parallel programming, the committee was unable to determine the
winners by this date, due to the length of time required to obtain
satisfactory data from the programs. Accordingly, the committee judged the
winners on December 16. We confirmed the results with all the committee
members, and accordingly announced all the winners.
The award winners in each category are as follows:
Category 1: | Sequential Environment | |
First prize: | Yasuyuki Mimatsu | |
(Japan Advanced Institute of Science and Technology) | ||
Second prize: | Haruhiko Usa (The University of Tokyo) | |
Honorable mention: | Katsutoshi Teranishi (Tohoku University) | |
Category 2: | Parallel Environment | |
First prize: | No Winner in this Category | |
Second prize: | Haruhiko Usa (The University of Tokyo) | |
Honorable mentions: | Kazuhiko Ono (Kyoto University), | |
Naomaru Kawasaki (Tokyo Institute of Technology) | ||
Category 3: | Free Subject | |
First prize: | Kenta Cho (Waseda University) | |
Second prize: | Takuya Araki (The University of Tokyo) | |
Honorable mention: | Tomoaki Tsumura (Kyoto University) | |
Special prize: | Shin Natori (Tokyo Institute of Technology) |
At the contest committee meeting, we decided to set out three categories,
namely:
Category 1 (Sequential Environment)
Category 2 (Parallel Environment)
Category 3 (Free Subject)
We also established a contest timetable and took care of other
administrative matters. Contest announcements were sent out through e-mail
and via the Internet, and also published in some magazines. We received 77
programs in total (Category 1: 54 programs, Category 2: 15 programs,
Category 3: 8 programs). As there were multiple applicants, the number of
applicants in total (including those who applied in groups) was 57. I think
that the number of applicants was quite satisfactory considering the
relative lack of parallel processing machines available in Japan at the
moment.
Members of the contest committee included those who were involved in the
development of KL1 and KLIC; KL1 users and many other technical staff
helped to measure the elapsed times and evaluate the documentation.
Now, let's briefly review the results of the judging in each category.
For Category 1, the programs were initially judged based on their elapsed
time with several different test data. Next, the structure of the programs,
and quality of documents (including manuals) were judged. Because every
programmer had done debugging and fine-tuning in the KLIC sequential
environment, this category was not too difficult to judge.
For Category 2, however, it proved unexpectedly difficult to judge the
winners. If adequate access to parallel machines and KLIC systems had been
available to applicants in Category 2, it would have proven easier for the
applicants to debug and fine-tune their programs by themselves, and
therefore a less difficult task for the examiners to judge them.
Unfortunately, however, most of the programs proved not to be fine-tuned
when confronted with genuine parallel environments.
Programs in Category 2 were first examined with simple test data in a
sequential processing environment followed by testing with more complicated
data to measure the elapsed time in a parallel processing environment. In
cases where program execution was not stable, the program contents were
carefully examined to determine why the program would not execute
satisfactorily. As well as examination of the programs themselves, the
quality of documentation (including manuals) was also evaluated. After all
was said and done, we decided to award no first prize, one second prize and
two honorable mentions.
Consequently, from this experience we learned that we need to provide
prospective applicants with real parallel machines for debugging and
fine-tuning purposes for the next contest. However, despite this, we
nevertheless very much appreciate the efforts of the applicants who tackled
Category 2 in spite of their difficult circumstances, without access to
parallel machines.
For Category 3, we received a number of interesting programs. Most of the
programs were excellent, so we could hardly decide which one was best. We
ended up choosing the winners by majority vote from among the twenty
committee members responsible for choosing the award winners.
All in all, I was gratified, as chairperson, to be in a position to judge
all these excellent programs, which were thoughtfully designed so as to
demonstrate the unique characteristics of KL1. Again, I would like to thank
all the contest applicants for their efforts.
As the judging process wore on, we decided to give every applicant a copy
of "Bit" magazine from July, 1993, a special issue on the topic of "Fifth
Generation Parallel Computing" as a token of our appreciation.
In preparation for the next contest, I would encourage you all to install
KLIC, if you have access to a parallel processing machine, and to try your
hand at a parallel environment.
To conclude my remarks, I would like again to thank all the applicants and
staff for their great efforts in making the contest a success. I look
forward to your entry for the next contest in 1997.
-----------------------------------------------------
Thanks to all the applicants, the Chairperson of the Contest Committee, the
Chairperson of Contest Administration and all the other staff members, we
were able to make the first KLIC Programming Contest a success. We would
like to make the second contest even more interesting. We would appreciate
it if you would send us your comments about the first contest to:
kaneda@icot.or.jp (Kiyoshi Kaneda)
2. Reports of the 2 Working Groups at Dept. of Future Technology Studies
====================================================
2.1) Petaflops Machines Technology Research Working Group (PFM-WG)
-----------------------------------------------
This is the first report on the PFM-WG at AITEC. We have previously
introduced both the nature and activities of this working group on the
AITEC homepage.
A petaflops machine is, as the name implies, thousands of times more
powerful than the world's largest massively parallel teraflops computer; it
is forecast that its total computational capacity will exceed all currently
existing computers in the U.S.A.
It will be employed for a wide variety of end-uses such as: military,
economic predictions, nuclear physics, astronomy, biology (DNA sequence
analysis and protein synthesis), pharmacology, mechanical engineering,
artificial-intelligence (AI) and database/knowledge base processing. It
will also be used to study the production of application software for all
the above end-uses.
From the study of such leading-edge technologies, coming down the pipe,
many useful new spin-offs can be derived. Especially with respect to
software, new useful parallel-processing algorithms are expected to be
developed.
At the department of FTS, we have set up a working group to study and analyze
ongoing petaflops research in the U.S.A., which is the leader in the study of
high-end computing. The chairperson, Dr. Yoshinori Yamaguchi of the
Electrotechnical Laboratory, and 11 other members meet once a month at
AITEC to hold discussions on recent research.
The PFM-WG plans to meet a total of 5 times; it is also expected to produce
a report on the importance of PFM to Japanese computer research as well as
to stress its importance to the future of the Japanese computer industry.
The first WG meeting dealt with a variety of topics with respect to
architectures, software and applications, as well as discussing several
areas such as "reliability of parallel machines," "difficulty in
development of general parallel machines," and "initiatives in software and
hardware."
Their second meeting was held on November 14, 1996. Two AITEC staff members
presented reports on their business trip wherein they participated in the
petaflops machine conference, "Frontiers '96". Furthermore, Dr. Amano, a
member of the WG, presented an explanation and his comments on network
connecting processors, and several problems in relationship to these.
Scalable general-purpose machines can be realized by making use of CC-NUMA
with DSM, after which it is required for the network to be able to manage
the DSMs. Though petaflops-class special purpose machines will undoubtedly
be realized, most WG members concluded that the construction of
petaflops-class general purpose machines would be extremely difficult with
present technologies. However, in the next 15 years our dreams may come
true--especially when we think about how technology has improved over the
last 15 years.
After the development of teraflops-class machines, we find that we are often
impressed with U.S. capabilities when they challenge a new objective.
The WG plans further discussion and investigations with respect to the
feasibility of petaflops machines, paying particular attention to
applications and systems, software, architectures, semiconductor devices
and so on. (Kazumi Kasai)
2.2) Network and AI-related New Technologies Research Working Group (NAI-WG)
--------------------------------------------------
The Internet is currently a hot topic. Many books about the Internet are to
be found in bookstores, as well as other topics such as multi-modal user
interfaces, virtual reality, and mobile computing, even including books
describing new computing paradigms such as agent-oriented computing. The
areas of networks and new computing types are quite lively.
In the last several years, use of both the Internet and computers have
mushroomed, and a variety of services using these are a frequent topic of
discussion. One can only describe the situation as chaotic, which is
ironic, considering that the 21st century is virtually upon us.
Are new computers, the Internet and other new technology as convenient,
easy to use, and indeed, useful as they are advertised now?
I would have to say "No." I say this because in order to make use of any
of this new technology, we have to read manuals with lots of unfamiliar
terms--the technology is anything but "user friendly." To render devices
more useful and easier-to-use, we require more advanced technology and a
better understanding of the ideal man-machine interface.
Keeping the above in mind, the Network and AI-related New Technologies
Research Working Group (NAI-WG) began discussions of basic technologies and
research subjects, out of which will be conceived the foundation of the
future information industry.
For instance, we will look at new technologies related to networks such as
the Internet, (and various tools and services on it), as well as AI
technology related to man-machine interfaces. We also plan to discuss
provisions for infrastructure to promote efficient research, as well as
the impact that such investments in research and development will likely
have on society in future. The chairperson is Dr. Hiroshi Okuno of NTT
Basic Research Laboratories; in addition, there are 10 members whose areas
of study are as follows:
a) New computer technologies and applications such as mobile computing
b) New application types such as Computer Supported Cooperative Tasks
c) Database technology including network processing
d) Multimodal interface technology and related AI technology
e) Social services and required information technology
In their first meeting, Dr. Uchida, the Director of AITEC, greeted the
attendees, after which each member introduced himself. Presentations and
demonstrations of multimodal interactive systems were given by a working
group member. It was interesting to see human-like agents created by
computer graphics techniques making legal arguments with a human in the
demonstration. After this, there were lively discussions on research
evaluation.
The topics of the WG will, in the main, be determined by the members and
will be discussed in the meetings.
(Kouichi Takeda)
3. The Report of "Frontiers '96" (Business Trip to the USA)
--------------------------------------
As part of my petaflops machine research at AITEC, I attended the Sixth
Symposium on the Frontiers of Massively Parallel Computation by IEEE
(Frontiers '96) held at Annapolis, Maryland from the 27th to the 31st of
October, 1996.
The symposium was held at the Loews Annapolis Hotel, which is an elegant
hotel located in the historic quarter.
About 130 people attended the symposium, seven of whom were from Japan
(including myself.)
While the attendees were of various nationalities, many of them were
resident in the U.S. according to the symposium's registration list.
The main presentations at Frontiers '96 concerned projects funded by the
National Science Foundation.
The title of the session was "Petaflops Computing/Point Design Studies".
There were eight research presentations in this session. Most of them were
the reports concerning the realization of 100 teraflops machines by the
year 2007.
I cannot predict whether 100 teraflops machines will actually be achieved
or not. However, I suppose that U.S. researchers may be able to carry it off.
If, in fact, they are successful, they may succeed in constructing
petaflops-class machines by the year 2014 as far as hardware is concerned.
There were four panel sessions. At each session, applications program
researchers appeared cheerful, while on the other hand, researchers into
basic system software appeared cheerless. The reason, I suppose, was
because the latter researchers have not as yet achieved any innovative
breakthroughs.
On the other hand, the presentations on superconducting device technologies
attracted the attention of the audience, far beyond what I would have
expected. I believe it is only natural that researchers are pinning their
hopes on superconducting devices, because more than one million CMOS
processing units are necessary to construct a petaflops machine. The concept
of integrating one million CMOS processors seems almost impossible now.
However, it may be the case that superconducting technology will achieve
this goal in place of CMOS devices. While only some products currently make
use of super-conductive devices, for example digital-analogue converters,
both technologies will undoubtedly evolve further by the year 2014. The
presentations made me consider that superconducting devices may provably have
some potential in making petaflops machines a reality.
My overall impression of Frontiers '96 was that nothing definite had been
decided so far, but petaflops machines will be an both a large and
interesting challenge for the 21st century. Furthermore, many spin-offs
will result from this research.
One day, at lunch time, there was a ceremony transferring an older-type MPP
(massively parallel processor) from NASA's Goddard Space Flight Center to
the Smithsonian Institution. At the ceremony, a speaker talked about the
history of parallel computers. He stated that Japanese fifth generation
computers demonstrated the effectiveness of parallel machines. I found
myself wishing that Japan could have a project to develop a petaflops machine
as well as the U.S. does.
At the end of the presentation, someone asked the speaker a question about
collaborative efforts with Japanese researchers. I expected that his answer
would be positive. Unfortunately, however, his answer was non-committal. I
believe that since research areas (such as petaflops machines) are
scientific and non-competitive in nature, that this would be an appropriate
arena for collaborative efforts.
It appeared to me that supercomputer hardware made in Japan was as good as
that made in the U.S. at present. However, in future, I think the U.S. will
gain the upper hand in this regard, because more supercomputers will be
provided as research tools and also because software researchers can get
more funds from the U.S. government than their counterparts can from the
Japanese government.
(Kunio Takahari)
********************************************************************** * * * A I T E C N E W S Issue #7 * * AITEC NEWS Editorial Team: * * Makiko Sato, Chie Takahashi, Akira Aiba * * Hiroshi Sato, Shunichi Uchida * * AITEC NEWS English Version Team: * * Masayo Fukushima, Shunichi Uchida * * Issued on: December 20, 1996(Japanese Version) * * February 25, 1997(English Version) * * By: Research Institute for Advanced Information * * Technology (AITEC), a subcenter of * * Japan Information Processing Development * * Center (JIPDEC) * * 2-3-3, Minato-ku, Shiba, Tokyo 105, Japan * * Tel: 03-3456-3191 FAX: 03-3455-4877 * * E-mail: aitec-news@icot.or.jp * * * **********************************************************************