« iepriekšējāTurpināt »
I was told that Dr. Miyoshi's paper the first three declined to participate, WAVELETS SEMINARS proposed a system of more than 100 although the others have had some IN TOKYO processors, connected via a full cross- collaborations and a manuscript sumbar network (so that each processor marizing their findings might be avail- Five seminars on wavelets and their can communicate quickly with any able in May 1992. Presumably their applications were sponsored by SIAM other), allowing 400-MB/s communi- research is continuing--as we were Japan on 14 November 1991. Although cation between each pair, with a total leaving, we met Dr. Tadashi Watanabe, most of the talks did not contain any communication capacity of 80 GB/s. NEC's Assistant General Manager for new research announcements, two of Each processor should be faster than a EDP Product Planning, and a key fig- the presentations are worth mentioning. Fujitsu VP-400. This would be a dis- ure in the development of the SX-3. Hideki Kawahara, from the NTT tributed machine, with each processor Miyoshi explained that a system Basic Research Laboratories, discussed having 256 MB mail memory, 128-KB similar to the one I mentioned above techniques for illustrating sound using vector registers, and 64-KB cache. I might be possible with BiCMOS, as graphics. He covered everything from was also told that Fujitsu's researchers 1,000 pin-outs may be realizable, but standard black-and-white, twogave a few research talks about such a that problems with energy consump- dimensional (2D) graphs to threesystem and suggested that it might be tion and maintenance were uppermost dimensional (3D) color surfaces of able to be built within 3 years. The in his mind. Miyoshi, as the director of wavelets. Kawahara commented that symposium was held in Japanese and a a computing facility that must serve a he does not use the sophisticated 3D Proceedings has not yet been published. large number of scientists, seemed more rendering techniques used at IBM
. , I went to NAL to visit Dr. Miyoshi interested in capabilities for the many Research. He proceeded to explain the to hear more of the details. Miyoshi than for the few. However, in summary, biological aspects of voice generation speaks very little English and asked Dr. Miyoshi wanted to emphasize that and reception. In summary, Kawahara that I bring a translator so that there the parallel machine I asked him about presented physiological evidence on how would be no misunderstandings. He was simply a proposal, not a plan. wavelets are better suited for acoustiwas quite concerned about recent U.S./ NAL is roughly similar in mission cal signal processing than currently Japan frictions in the area of super- to the National Aeronautics and Space employed techniques. Sound decomcomputing. I was accompanied by Mr. Administration in the United States, position using wavelets was described, Frank Nagashima from my office. With but much smaller. Founded in 1955, its and both female and male voices were Miyoshi was Dr. Toshiyuki Iwamiya, current mission is to establish the used in a demonstration that showed senior researcher in Miyoshi's Compu- necessary technologies for future devel- how the amount of information needed tational Sciences Division. Iwamiya can opment of aircraft in Japan and to per- to convey human speech could be sigalso be reached by E-mail at form research in those areas necessary nificantly reduced but not detected by [firstname.lastname@example.org). for future Japanese space development. the human ear if wavelets are used
Miyoshi told us about the history of NAL is operated as one of the bodies rather than conventional techniques. computing at NAL, which is built around under the Science and Technology Researchers interested in this area Fujitsu equipment. NAL has been using Agency. NAL's main headquarters are might contact Kawahara at NTT. He is Fujitsu hardware since the late 1970s. in Chofu, about 30 minutes outside conducting interesting and original work (See a brief description of NAL below.) Tokyo. There is also a smaller research on sound processing with emphasis on He explained that his paper (which is center in Kakuda, near Sendai. NAL's voice (probably for phone transmisnot yet available) was simply a state- budget has been about $70M for a sion since he's from NTT). ment of some of his ideas and does not number of years; staffing has been slowly represent any concrete plan. He com- decreasing from nearly 500 in the late Hideki Kawahara mented that about 2 years ago NAL 1960s to 439 in 1991. Miyoshi's Com- Senior Research Scientist Supervisor approached a number of computer putational Sciences Division performs NTT Basic Research Laboratories vendors, including Cray Japan, IBM research in nonlinear phenomena, 9-11 Midori-cho 3-chome Japan, Unisys, NEC, Fujitsu, and other numerical simulation, artificial intelli- Musashino, Tokyo 180, Japan Japanese companies, for assistance in gence, and image understanding.--David Tel: +81-422-59-2276 developing, at least at the specification K Kahaner, ONRASIA
E-mail: email@example.com level, a machine that would be appropriate for numerical wind tunnel simu
Prof. Yamada, from the Department lation. He explained that at the time
of Civil Engineering, Kyoto University, discussed singularity analysis and
detection. The example used was the results from a study comparing Bi-CG, a Stanford University conference. (The standard 1/3, 1/3, 1/3 Cantor set from CGS, and Bi-CGSTAB methods for release date was purposely designed undergraduate mathematics. The talk solving the two-dimensional Laplacian
solving the two-dimensional Laplacian to coincide with the first distribution was based on research work by with an inhomogeneous sinusoidal term
with an inhomogeneous sinusoidal term day of Knuth's classic, TeX Book.) Vergassola using the Mexican Hat and on the unit square with general bound- Enhancement continued from versions French Hat wavelets. For singularities ary conditions.
1.0, 1.1, 1.2, 2.0, 2.99, to 3.0 in 1990, of a known order, wavelets can be used The second speaker, S. Fujino, also then 3.14 and upwards. (TeX is expected to verify the order. The remainder of from iCFD, summarized both his and to converge to pi upon Knuth's death the talk focused on the open question Zhan's talks in commenting that according to his will.) Note: the casual of whether wavelets can be used to Bi-CGSTAB is superior in that it does user will not be able to distinguish the determine the order of a singularity. not oscillate widely as it converges difference in versions 3.0 and above. In Vergassola's work shows how to “clean” towards the desired solution, whereas the same vein, Meta Font is being data to estimate the order but could the other two methods do and may upgraded from 2.0 to 2.7 and upwards not concretely answer the question.-- mislead a naive scientist. During the and is expected to converge to e upon Mei Kobayashi, IBM Japan
question-and-answer (Q&A) session Knuth's death. Nodera spent the latter he commented that Van der Vorst half discussing how the multi-and twoannounced an improved version, column styles differ and how to play
Bi-CGSTAB II, 2 weeks ago in Zurich; tricks to format a paper with an RESEARCH INSTITUTE OF however, implementation of this abstract to conform to the Information MATHEMATICAL SCIENCES upgraded version has not yet begun at Processing Society of Japan (IPSJ) doc(RIMS) WORKSHOP ON the Institute for Computational Fluid ument style. “Why not simply cut and NUMERICAL ALGORITHMS Dynamics (CFD).
paste?" asked an irritated listener. This
Mori's (University of Tokyo) talk was the beginning of a discussion on This workshop is one of a series of was on a three-dimensional (3D) exten- the quotation of the day, “We are in the sixsponsored annually by the Research sion of a two-dimensional (2D) approx- business of doing mathematics, not Institute of Mathematical Sciences imation of a blast furnace problem (2D publishing." (RIMS) of Kyoto University during the reference: M. Mori, M. Natori, and
, The next talk on Springer-Verlag's fall-winter period. The organizer of Zhang Guo-Feng, Intl. J. Num. Meth. TeX Macropackage was equally, if not the workshop was Fluids, 569-82 (1989)].
more, controversial. The speaker noted
The author's talk began with illus- the trend towards electronic submisProf. Taketomo Mitsui
trations of simulations using the latest sions to prestigious journals and book Nagoya University
IBM graphics technologies. Modelling publishers. Details on accessing and Dept of Information Engineering and rendering techniques were also using Springer's new package were Furo-cho, Chikusa-ku
illustrated through an award-winning outlined. The Q&A session became Nagoya 464-01, Japan
video, “Edo,” by Miyata, also of IBM. heated once again. To summarize, a Tel: +81-52-781-5111,5808 or 5810 The latter half of the talk was on the significant portion of the audience Fax: +81-52-782-9143
topic of processing of the graphics and seemed concerned about whether the E-mail: firstname.lastname@example.org.
image data using the Haar wavelet basis. use of certain formatting styles and nagoya-u.ac.jp
Although only two-dimensional results electronic submissions to journals would
were shown, the author explained that become mandatory. The announcement The talks on the first afternoon level sets of 10 2D layers have been that Springer asks for DVI rather than were rather unusual in that they empha- generated and will be released when 32 the original TeX file brought forth very sized computations and visualization or more 2D layers are used.
strong negative reactions from the of the results. Three color videos (by Two talks on TeX and LaTeX stirred audience. Concerns were also voiced S.-R. Zhan, M. Mori, and the author) up heated discussion (an unusual event regarding trouble by some Japanese showed how new technologies could be for a Japanese mathematical meeting). universities in recruiting young used to enhance understanding of math- The first TeX talk by T. Nodera (Keio researchers because of comparatively ematical and numerical work. Zhan University) began with the birth of TeX
University) began with the birth of TeX poor E-mail or networking services.-[Institute for Computational Fluid as a word processor for “A Christmas Mei Kobayashi, IBM Japan Dynamics (iCFD)], a recent graduate Carol” in 1978. The history of TeX was of Tsukuba University and a former outlined from version 0.5 to 0.99 until student of Mori, began the session with the release of TeX 1.0 in July of 1982 at
(2) Development of microprocessors and a TX2 and TX1ASSP announced.
for TRON OS (not compatible with In addition, there is a 16-bit control Introduction
Intel or Motorola processors). processor TLCS900 with a real-time
OS based on myITRON, called TR900. K Sakamura (University of Tokyo), This is not an exclusive relationship: Matsushita showed its MN10400 (32-bit) the originator of the TRON concept, TRON MPUs are also used for Unix MPU with 20 MIPS (maximum) for again had an impressive list of support implementations and TRON OSs are fast execution of TRON commands. ing and sponsoring organizations for successfully implemented on, e.g., Intel An OS ITRON is available. Also, a this second show: the Ministry of Inter- MPUS.
16-bit MPU MN10300 using myITRON national Trade and Industry (MITI),
was developed. the Ministry of Posts (MOP), and the Software Science and Technology Agency (STA);
Some Systems computer manufacturers; and construc- Fujitsu is marketing a real-time OS, tion companies. (See also the report by based on ITRON specifications, called The electric machinery maker D.K Kahaner, “TRON (The Real Time REALOS/Gmicro. Oki introduced its Meiden presented a workstation for Operating System Nucleus,” Scientific real-time OS RG68KS, based on factory automation. The machine is Information Bulletin 16(3), 11-19 (1991).] CTRON specifications. BTRON with based on NEC's 32-bit MPU V80, which The newly formed TRON Multimedia its object-oriented features and its runs NEC's RX-UX832 real-time Unix, Broadcasting Company (TMBC) was graphical user interface (GUI) is mak- which is composed of Unix V and taking on-the-spot shots to include them ing some progress in the direction as ITRONas real-time kernel. It provides in its multimedia show. All together OS for multimedia. The Multimedia peripheral boards for multichannel about 20 companies displayed prod. Working Group has proposed an exper- communication and Meiden real-time ucts and applications in booths. In imental system, connecting BTRON Basic (MRTB) for parallel execution. addition, a central theater was used to workstations via ISDN to form a con- A simpler system for factory automaexplain ideas and products through video ference system. On the other hand, Wind tion (FA), called TB-100, is produced demonstrations.
River Systems (U.S.) brought their real- by Nihon Minicomputer Systems. It Among the exhibitors, all Japanese time OS VXWORKS on the Gmicro uses the Gmicro/100 processor and computer manufacturers were present processors.
ITRON as OS. As an embedded syswith the notable exception of NEC.
tem, Mitsubishi Electric showed a test NEC has been reluctant to join the Microprocessors
version of a fax machine using myITRON TRON project because it is success
to handle all functions of the machine fully marketing MS-DOS/Windows/ There were no new developments and a color copier with an outline font Unix machines and because it has devel- except customizable versions of 32-bit
except customizable versions of 32-bit driver using Gmicro processors. Another oped its own series of Intel-compatible processors and downscaling to 16-bit application was a controller for a small microprocessors, the V-series. The processors. Within the “Gmicro” micro- robot. Originally a publishing house, multiprocessing units (MPUs) at the processor line (Fujitsu and Hitachi, Personal Media Co. presented a notelower end are already replacing the Oki), Fujitsu presented the Gmicro book computer (hardware (Intel 386) Intel processors due to their superior G32 series, 32-bit MPUs with the /300- by Matsushita) with BTRON as OS performance. On the software side, NEC version at the upper end, achieving with a Windows-like user interface. As is, however, using ITRON specifica- 24 MIPS at 33 MHz. There have been applications a simple wordprocessor, tions as real-time kernel in its real- plans for high performance MPUS/400 graphic editor, etc. are available. In time Unix operating system (OS), called and /500 for some time. A broad range addition, the company has developed a RX-UX832.
of peripheral chips by Fujitsu is avail- workstation using TRON microproThe activities of TRON comprise able. In addition, a multimedia board cessors and an OS "B2" based on two lines:
using a F32/300 (32 MIPS) processor BTRON specifications. Japan Airlines
developed by Genesys, using the TRON is using terminals with BTRON OS to (1) Development and implementation application databus (TAD), is able to access its passenger reservation system
of multitask, real-time Oss handle all kinds of media. Hitachi pre- AXESS, and Matsushita introduced its (CTRON, ITRON + subset sented the H32-series with a program educational computer based on myITRON, BTRON).
similar to that of Fujitsu. A separate BTRON1 specification, called
was Japanese with a very few non- Broader aspects of HI were taken
Japanese participants. From abroad the up in the other papers presented at the There is a full line of 32-bit micro- presenters were: William Buxton symposium. Hiroshi Ishii (NTT Human processors available that are free of (University of Toronto, Canada); Philip Interface Laboratories) presented his foreign property rights, and in the next Cohen (SRI, U.S.); Gloranna Davenport concept of an open shared workspace, stage 64-bit processors are to be (Massachusetts Institute of Technology, intending to overcome acceptance expected. On the other side, 16-bit U.S.); Jonathan Grudin (University of problems by not forcing users in a processors have been introduced for
California at Irvine, U.S.); D. Austin completely new environment. To make cost efficient process control. There Henderson (Xerox, U.S.); Marilyn best use of the limited screen space, seems to be a rising interest in using Mantei (University of Toronto, Canada); translucent overlay of individual workITRON and myITRON in multifunc- Gary Olson (University of Michigan, space images was introduced. With tional fax, video recorders, video U.S.); William Verplank (IDEO Product this overlay technology, the actual faces cameras, printers, etc, using both MPUS Dev., U.S.); Eric Werner (PLATO CO., of the participants could be shown. with TRON specifications and non- Germany); and Richard Young (Medical Prof. Yuzuru Tanaka (Hokkaido UniTRON processors. Business applica- Research Council, U.K.). In addition, versity) explained ideas, partly based tions, where BTRON is competing with 13 Japanese speakers gave papers. It on Tsichritzis'object-oriented proposexisting Unix software, are still in an was the second international meeting als, for open systems. He showed an experimental stage. Multimedia appli- after the first one at the start of the impressive video of his “IntelligentPad” cations, however, should help to pro- FRIEND21 project in September 1988. system, which relies on a generic toolmote BTRON with its real-time pro- Papers were presented in the following kit, synthetic programming, open platcessing potential. In addition, sections: Status of the FRIEND21 form, and integrated management. His Sakamura's ideas concerning TRON Project, Communication and Human objects could be easily combined, cut houses, TRON buildings, and the TRON Interface, Cognitive Models, Interface into pieces, rearranged, etc. His examComputer City in Chiba Prefecture will Design, Communication and Coopera- ples included also an application to further promote the TRON standards tion, Computer-Supported Cooperative computer-aided instruction (CAI) in with their coherent design for all kinds Work, and Multimedia.
classical mechanics. Two companies of applications ranging from embedded FRIEND21, as a MITI project, is presented results in the field of accessmicrocomputers to large-scale dis- pursued within a central institution, ing multimedia. Miyatake (PIE/Hitachi) tributed computer systems.--Ulrich the Institution for Personalized Envi- showed a well-advanced digitized video Wattenberg, GMD
ronment (PIE), which is in contact with tape editor that includes automatic shot the 14 sponsoring companies (com- separation, iconization, and editing puter manufacturers, home electronic tools. Watanabe (PIE/Sony) explained
corporations, and publishing or print- automatic shot separation and investiFRIEND21
ing companies). At the institute, which gations of TV quiz programs for devel
is staffed with a dozen researchers, a opment of scenario-based interfaces. The 1991 International Symposium prototype human interface (HI) is under The final panel discussion on on Next Generation Human Interface development, targeting the untrained "Human Interface in the Future” was was held from 25-27 November 1991 at casual user rather than the professional. led by Prof. Mario Tokoro (Keio UniKeidanren Kaikan, Tokyo. FRIEND21 The key developers at PIE, Hirotada versity, Sony). In his introduction he is a Ministry of International Trade Ueda (PIE/Hitachi) and Hajime gave the picture of a “sea of computand Industry (MITI) project related to Nonogaki (until recently at PIE and ers" where everybody can move freely, human-computer interface technology now back at Fujitsu), presented their contacting everybody else with a pocketthat started in 1988. After the domestic model: agents (including the human size computer. Other panelists gave symposium in June, this meeting was user) for different tasks, which exchange less futuristic examples and actually turned international by the invited messages through a “studio.” For pre- the discussion went back to the quesspeakers. (See also the report by sentation of and access to information, tion, What position should computers D.K. Kahaner, "Virtual Reality,"
Reality,” three metaphors, “newspaper,” “TVI have within the society? Two approaches Scientific Information Bulletin 16(4), video tape recorder,” and “database,” were apparent: (1) making computers 43-45 (1991).] The audience of 600 are explored.
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