Lapas attēli
PDF
ePub

I was told that Dr. Miyoshi's paper proposed a system of more than 100 processors, connected via a full crossbar network (so that each processor can communicate quickly with any other), allowing 400-MB/s communication between each pair, with a total communication capacity of 80 GB/s. Each processor should be faster than a Fujitsu VP-400. This would be a distributed machine, with each processor having 256 MB mail memory, 128-KB vector registers, and 64-KB cache. I was also told that Fujitsu's researchers gave a few research talks about such a system and suggested that it might be able to be built within 3 years. The symposium was held in Japanese and a Proceedings has not yet been published.

I went to NAL to visit Dr. Miyoshi to hear more of the details. Miyoshi speaks very little English and asked that I bring a translator so that there would be no misunderstandings. He was quite concerned about recent U.S./ Japan frictions in the area of supercomputing. I was accompanied by Mr. Frank Nagashima from my office. With Miyoshi was Dr. Toshiyuki Iwamiya, senior researcher in Miyoshi's Computational Sciences Division. Iwamiya can also be reached by E-mail at [iwamiya@aerospace-lab.go.jp].

Miyoshi told us about the history of computing at NAL, which is built around Fujitsu equipment. NAL has been using Fujitsu hardware since the late 1970s. (See a brief description of NAL below.) He explained that his paper (which is not yet available) was simply a statement of some of his ideas and does not represent any concrete plan. He commented that about 2 years ago NAL approached a number of computer vendors, including Cray Japan, IBM Japan, Unisys, NEC, Fujitsu, and other Japanese companies, for assistance in developing, at least at the specification level, a machine that would be appropriate for numerical wind tunnel simulation. He explained that at the time

the first three declined to participate, WAVELETS SEMINARS although the others have had some IN TOKYO collaborations and a manuscript summarizing their findings might be available in May 1992. Presumably their research is continuing--as we were leaving, we met Dr. Tadashi Watanabe, NEC's Assistant General Manager for EDP Product Planning, and a key figure in the development of the SX-3.

Miyoshi explained that a system similar to the one I mentioned above might be possible with BiCMOS, as 1,000 pin-outs may be realizable, but that problems with energy consumption and maintenance were uppermost in his mind. Miyoshi, as the director of a computing facility that must serve a large number of scientists, seemed more interested in capabilities for the many than for the few. However, in summary, Dr. Miyoshi wanted to emphasize that the parallel machine I asked him about was simply a proposal, not a plan.

NAL is roughly similar in mission to the National Aeronautics and Space Administration in the United States, but much smaller. Founded in 1955, its current mission is to establish the necessary technologies for future development of aircraft in Japan and to perform research in those areas necessary for future Japanese space development. NAL is operated as one of the bodies under the Science and Technology Agency. NAL's main headquarters are in Chofu, about 30 minutes outside Tokyo. There is also a smaller research center in Kakuda, near Sendai. NAL's budget has been about $70M for a number of years; staffing has been slowly decreasing from nearly 500 in the late 1960s to 439 in 1991. Miyoshi's Computational Sciences Division performs research in nonlinear phenomena, numerical simulation, artificial intelligence, and image understanding.--David K. Kahaner, ONRASIA

Five seminars on wavelets and their applications were sponsored by SIAM Japan on 14 November 1991. Although most of the talks did not contain any new research announcements, two of the presentations are worth mentioning.

Hideki Kawahara, from the NTT Basic Research Laboratories, discussed techniques for illustrating sound using graphics. He covered everything from standard black-and-white, twodimensional (2D) graphs to threedimensional (3D) color surfaces of wavelets. Kawahara commented that he does not use the sophisticated 3D rendering techniques used at IBM Research. He proceeded to explain the biological aspects of voice generation and reception. In summary, Kawahara presented physiological evidence on how wavelets are better suited for acoustical signal processing than currently employed techniques. Sound decomposition using wavelets was described, and both female and male voices were used in a demonstration that showed how the amount of information needed to convey human speech could be significantly reduced but not detected by the human ear if wavelets are used rather than conventional techniques.

Researchers interested in this area might contact Kawahara at NTT. He is conducting interesting and original work on sound processing with emphasis on voice (probably for phone transmission since he's from NTT).

Hideki Kawahara

Senior Research Scientist Supervisor
NTT Basic Research Laboratories
9-11 Midori-cho 3-chome
Musashino, Tokyo 180, Japan
Tel: +81-422-59-2276
E-mail: kawahara@siva.ntt.jp

Prof. Yamada, from the Department of Civil Engineering, Kyoto University, discussed singularity analysis and

detection. The example used was the standard 1/3, 1/3, 1/3 Cantor set from undergraduate mathematics. The talk was based on research work by Vergassola using the Mexican Hat and French Hat wavelets. For singularities of a known order, wavelets can be used to verify the order. The remainder of the talk focused on the open question of whether wavelets can be used to determine the order of a singularity. Vergassola's work shows how to "clean" data to estimate the order but could not concretely answer the question.-Mei Kobayashi, IBM Japan

RESEARCH INSTITUTE OF MATHEMATICAL SCIENCES (RIMS) WORKSHOP ON NUMERICAL ALGORITHMS

This workshop is one of a series of six sponsored annually by the Research Institute of Mathematical Sciences (RIMS) of Kyoto University during the fall-winter period. The organizer of the workshop was

Prof. Taketomo Mitsui
Nagoya University

Dept of Information Engineering
Furo-cho, Chikusa-ku

Nagoya 464-01, Japan

results from a study comparing Bi-CG, CGS, and Bi-CGSTAB methods for solving the two-dimensional Laplacian solving the two-dimensional Laplacian with an inhomogeneous sinusoidal term on the unit square with general boundary conditions.

The second speaker, S. Fujino, also from iCFD, summarized both his and Zhan's talks in commenting that Bi-CGSTAB is superior in that it does not oscillate widely as it converges towards the desired solution, whereas the other two methods do and may mislead a naive scientist. During the question-and-answer (Q&A) session he commented that Van der Vorst announced an improved version, Bi-CGSTAB II, 2 weeks ago in Zurich; however, implementation of this upgraded version has not yet begun at the Institute for Computational Fluid Dynamics (iCFD).

Mori's (University of Tokyo) talk was on a three-dimensional (3D) extension of a two-dimensional (2D) approximation of a blast furnace problem [2D reference: M. Mori, M. Natori, and Zhang Guo-Feng, Intl. J. Num. Meth. Fluids, 569-82 (1989)].

The author's talk began with illustrations of simulations using the latest IBM graphics technologies. Modelling and rendering techniques were also illustrated through an award-winning video, "Edo," by Miyata, also of IBM.

Tel: +81-52-781-5111, x5808 or 5810 The latter half of the talk was on the
Fax: +81-52-782-9143
E-mail: a41794a@nucc.cc.

nagoya-u.ac.jp

The talks on the first afternoon were rather unusual in that they emphasized computations and visualization of the results. Three color videos (by S.-R. Zhan, M. Mori, and the author) showed how new technologies could be used to enhance understanding of mathematical and numerical work. Zhan [Institute for Computational Fluid Dynamics (iCFD)], a recent graduate of Tsukuba University and a former student of Mori, began the session with

topic of processing of the graphics and image data using the Haar wavelet basis. Although only two-dimensional results were shown, the author explained that level sets of 10 2D layers have been generated and will be released when 32 or more 2D layers are used.

Two talks on TeX and LaTeX stirred up heated discussion (an unusual event for a Japanese mathematical meeting). The first Tex talk by T. Nodera (Keio University) began with the birth of TeX as a word processor for "A Christmas Carol" in 1978. The history of TeX was outlined from version 0.5 to 0.99 until the release of TeX 1.0 in July of 1982 at

a Stanford University conference. (The release date was purposely designed to coincide with the first distribution day of Knuth's classic, TeX Book.) Enhancement continued from versions 1.0, 1.1, 1.2, 2.0, 2.99, to 3.0 in 1990, then 3.14 and upwards. (TeX is expected to converge to pi upon Knuth's death according to his will.) Note: the casual user will not be able to distinguish the difference in versions 3.0 and above. In the same vein, Meta Font is being upgraded from 2.0 to 2.7 and upwards and is expected to converge to e upon Knuth's death. Nodera spent the latter half discussing how the multi-and twocolumn styles differ and how to play tricks to format a paper with an abstract to conform to the Information Processing Society of Japan (IPSJ) document style. "Why not simply cut and paste?" asked an irritated listener. This was the beginning of a discussion on the quotation of the day, "We are in the business of doing mathematics, not publishing."

The next talk on Springer-Verlag's TeX Macropackage was equally, if not more, controversial. The speaker noted the trend towards electronic submissions to prestigious journals and book publishers. Details on accessing and using Springer's new package were outlined. The Q&A session became heated once again. To summarize, a significant portion of the audience seemed concerned about whether the use of certain formatting styles and electronic submissions to journals would become mandatory. The announcement that Springer asks for DVI rather than the original TeX file brought forth very strong negative reactions from the audience. Concerns were also voiced regarding trouble by some Japanese universities in recruiting young researchers because of comparatively poor E-mail or networking services.-Mei Kobayashi, IBM Japan

TRONSHOW '91

Introduction

(2) Development of microprocessors
for TRON OS (not compatible with
Intel or Motorola processors).

This is not an exclusive relationship:
TRON MPUs are also used for Unix
TRON MPUs are also used for Unix
implementations and TRON OSS are
successfully implemented on, e.g., Intel
MPUs.

K. Sakamura (University of Tokyo), the originator of the TRON concept, again had an impressive list of supporting and sponsoring organizations for this second show: the Ministry of International Trade and Industry (MITI), the Ministry of Posts (MOP), and the Software Science and Technology Agency (STA); computer manufacturers; and construction companies. [See also the report by D.K. Kahaner, "TRON (The Real Time Operating System Nucleus," Scientific Information Bulletin 16(3), 11-19 (1991).] The newly formed TRON Multimedia Broadcasting Company (TMBC) was taking on-the-spot shots to include them in its multimedia show. All together about 20 companies displayed products and applications in booths. In addition, a central theater was used to explain ideas and products through video demonstrations.

Fujitsu is marketing a real-time OS, based on ITRON specifications, called REALOS/Gmicro. Oki introduced its real-time OS RG68KS, based on CTRON specifications. BTRON with its object-oriented features and its graphical user interface (GUI) is making some progress in the direction as OS for multimedia. The Multimedia Working Group has proposed an experimental system, connecting BTRON workstations via ISDN to form a conference system. On the other hand, Wind River Systems (U.S.) brought their realtime OS VxWORKS on the Gmicro processors.

Among the exhibitors, all Japanese computer manufacturers were present with the notable exception of NEC. NEC has been reluctant to join the Microprocessors TRON project because it is successfully marketing MS-DOS/Windows/ Unix machines and because it has developed its own series of Intel-compatible microprocessors, the V-series. The multiprocessing units (MPUs) at the lower end are already replacing the Intel processors due to their superior performance. On the software side, NEC is, however, using ITRON specifications as real-time kernel in its realtime Unix operating system (OS), called RX-UX832.

There were no new developments except customizable versions of 32-bit processors and downscaling to 16-bit processors. Within the “Gmicro" microprocessor line (Fujitsu and Hitachi, Oki), Fujitsu presented the Gmicro G32 series, 32-bit MPUs with the /300version at the upper end, achieving 24 MIPS at 33 MHz. There have been plans for high performance MPUs/400 and /500 for some time. A broad range of peripheral chips by Fujitsu is avail

The activities of TRON comprise able. In addition, a multimedia board two lines:

(1) Development and implementation

of multitask, real-time OSS (CTRON, ITRON + subset myITRON, BTRON).

using a F32/300 (32 MIPS) processor
developed by Genesys, using the TRON
application databus (TAD), is able to
handle all kinds of media. Hitachi pre-
sented the H32-series with a program
similar to that of Fujitsu. A separate
line is under development by Toshiba
with a 32-bit TX1 processor available

and a TX2 and TX1ASSP announced. In addition, there is a 16-bit control processor TLCS900 with a real-time OS based on myITRON, called TR900. Matsushita showed its MN10400 (32-bit) MPU with 20 MIPS (maximum) for fast execution of TRON commands. An OS ITRON is available. Also, a 16-bit MPU MN10300 using myITRON was developed.

Some Systems

The electric machinery maker Meiden presented a workstation for factory automation. The machine is based on NEC's 32-bit MPU V80, which runs NEC's RX-UX832 real-time Unix, which is composed of Unix V and ITRON as real-time kernel. It provides peripheral boards for multichannel communication and Meiden real-time Basic (MRTB) for parallel execution. A simpler system for factory automation (FA), called TB-100, is produced by Nihon Minicomputer Systems. It uses the Gmicro/100 processor and ITRON as OS. As an embedded system, Mitsubishi Electric showed a test version of a fax machine using myITRON to handle all functions of the machine and a color copier with an outline font driver using Gmicro processors. Another application was a controller for a small robot. Originally a publishing house, Personal Media Co. presented a notebook computer [hardware (Intel 386) by Matsushita] with BTRON as OS with a Windows-like user interface. As applications a simple wordprocessor, graphic editor, etc. are available. In addition, the company has developed a workstation using TRON microprocessors and an OS "B2" based on BTRON specifications. Japan Airlines is using terminals with BTRON OS to access its passenger reservation system AXESS, and Matsushita introduced its educational computer based on BTRON1 specification, called PanaCAL ET.

Summary

There is a full line of 32-bit microprocessors available that are free of foreign property rights, and in the next stage 64-bit processors are to be expected. On the other side, 16-bit processors have been introduced for cost efficient process control. There seems to be a rising interest in using ITRON and myITRON in multifunctional fax, video recorders, video cameras, printers, etc., using both MPUs with TRON specifications and nonTRON processors. Business applications, where BTRON is competing with existing Unix software, are still in an experimental stage. Multimedia applications, however, should help to promote BTRON with its real-time processing potential. In addition, Sakamura's ideas concerning TRON houses, TRON buildings, and the TRON Computer City in Chiba Prefecture will further promote the TRON standards with their coherent design for all kinds of applications ranging from embedded microcomputers to large-scale distributed computer systems.--Ulrich Wattenberg, GMD

FRIEND21

The 1991 International Symposium on Next Generation Human Interface was held from 25-27 November 1991 at Keidanren Kaikan, Tokyo. FRIEND21 is a Ministry of International Trade and Industry (MITI) project related to human-computer interface technology that started in 1988. After the domestic symposium in June, this meeting was turned international by the invited speakers. [See also the report by D.K. Kahaner, "Virtual Reality," Scientific Information Bulletin 16(4), 43-45 (1991).] The audience of 600

was Japanese with a very few nonJapanese participants. From abroad the presenters were: William Buxton (University of Toronto, Canada); Philip Cohen (SRI, U.S.); Gloranna Davenport (Massachusetts Institute of Technology, U.S.); Jonathan Grudin (University of California at Irvine, U.S.); D. Austin Henderson (Xerox, U.S.); Marilyn Mantei (University of Toronto, Canada); Gary Olson (University of Michigan, U.S.); William Verplank (IDEO Product Dev., U.S.); Eric Werner (PLATO Co., Germany); and Richard Young (Medical Research Council, U.K.). In addition, 13 Japanese speakers gave papers. It was the second international meeting after the first one at the start of the FRIEND21 project in September 1988. Papers were presented in the following sections: Status of the FRIEND21 Project, Communication and Human Interface, Cognitive Models, Interface Interface, Cognitive Models, Interface Design, Communication and Cooperation, Computer-Supported Cooperative Work, and Multimedia.

FRIEND21, as a MITI project, is pursued within a central institution, the Institution for Personalized Environment (PIE), which is in contact with the 14 sponsoring companies (computer manufacturers, home electronic corporations, and publishing or printing companies). At the institute, which is staffed with a dozen researchers, a prototype human interface (HI) is under development, targeting the untrained casual user rather than the professional. The key developers at PIE, Hirotada Ueda (PIE/Hitachi) and Hajime Nonogaki (until recently at PIE and now back at Fujitsu), presented their model: agents (including the human user) for different tasks, which exchange messages through a "studio." For presentation of and access to information, three metaphors, "newspaper," "TV/ video tape recorder,” and “database,” are explored.

Broader aspects of HI were taken up in the other papers presented at the symposium. Hiroshi Ishii (NTT Human Interface Laboratories) presented his concept of an open shared workspace, intending to overcome acceptance problems by not forcing users in a completely new environment. To make best use of the limited screen space, translucent overlay of individual workspace images was introduced. With this overlay technology, the actual faces of the participants could be shown. Prof. Yuzuru Tanaka (Hokkaido University) explained ideas, partly based on Tsichritzis' object-oriented proposals, for open systems. He showed an impressive video of his “IntelligentPad” system, which relies on a generic toolkit, synthetic programming, open platform, and integrated management. His objects could be easily combined, cut into pieces, rearranged, etc. His examples included also an application to computer-aided instruction (CAI) in classical mechanics. Two companies presented results in the field of accessing multimedia. Miyatake (PIE/Hitachi) showed a well-advanced digitized video tape editor that includes automatic shot separation, iconization, and editing tools. Watanabe (PIE/Sony) explained automatic shot separation and investigations of TV quiz programs for development of scenario-based interfaces.

The final panel discussion on "Human Interface in the Future" was led by Prof. Mario Tokoro (Keio University, Sony). In his introduction he gave the picture of a “sea of computers" where everybody can move freely, contacting everybody else with a pocketsize computer. Other panelists gave less futuristic examples and actually the discussion went back to the question, What position should computers have within the society? Two approaches were apparent: (1) making computers useful for everybody, and (2) looking

for needs to support or replace human activities by computers. Both approaches are leading at the end to marketing strategies, problems which are best solved within or together with companies. Philosopher Shun Tsuchiya (Chiba University) came to the conclusion that, rather than worrying about interfaces, school education should provide children at an early stage with knowledge about their own responsibility when using computers. Computers may be faulty or may invite infringements on personal data, copyright material etc.

This symposium was well organized by the people behind FRIEND21 working at PIE, who brought together researchers from various fields and countries and did not avoid critical contributions. PIE is a good platform for discussion and coordination of research on HI related issues. Its own research must be seen as homework to be done within quite a strict budget and time restrictions.--Ulrich Wattenberg, GMD

« iepriekšējāTurpināt »