Page 51 First New Information Processing Technology Workshop '91 The purpose of this workshop was to listen to proposals from scientists interested Computer Security in Japan Jonathan D. Moffett The state of progress in computer security activities in Japan is assessed. Japanese Advances in Fuzzy Systems and Case-Based Reasoning This article presents a survey and assessment of fuzzy systems and case-based Mathematical Theory of Networks and Systems '91 (MTNS '91) The conference and site visits in Japan, Hong Kong, Macao, Singapore, and India Manufacturing Science State of the Art in Japanese Computer-Aided Design Methodologies for Japanese use of computers in design of mechanical products is surveyed, the state Materials Science Page Intelligent Materials Systems and Materials Science Research in Australia ........... 113 Iqbal Ahmad The concept of "smart/intelligent" materials systems is receiving increasing High Performance/High Temperature Materials in Japan Current research in Japan on intermetallic compounds, functionally gradient 119 Ocean Science United States-Japan Cooperative Program in Natural Resources (UJNR) A new subdivision of the UJNR program, Deep Marine Technology, is discussed. South Pacific Environmental Program (SPREP) Biodiversity Workshop and Marine Mammal Conservation Plan ....... Gregory Stone, Michael Donoghue, and Stephen Leatherwood The group identified environmental problems specific to island countries and .... 131 137 Cover: The humpback whale. One of the first priorities for the South Pacific Environmental Program's (SPREP) proposed Marine Mammal Action Plan is to access humpback whale breeding populations in the region. Historically, the SPREP region contained many breeding sites for both the northern and southern hemisphere stocks. Contemporary breeding sites are poorly understood. Research programs would include surveys to identify the sites and the use of natural markings for individual identification. Photo courtesy of Gregory Stone (see his article on SPREP on page 137). SUPERCOMPUTING BENCHMARK PROPOSAL and Recently, four Japanese scientists, Wong Weng Fai [wong@rkna50.riken. go.jp], Eiichi Goto, Yoshio Oyanagi [oyanagi@is.s.u-tokyo.ac.jp], Nobuaki Yoshida [nyoshida@rkna50. riken.go.jp], have proposed two sets of benchmark problems for supercomputers. Goto was the director of the University of Tokyo Computer Center until he retired in March 1991. He is now at Kanagawa University. Oyanagi is a central researcher in the QCDPAX parallel processing project at Tsukuba University, although he is now at the University of Tokyo. Their proposal allows for maximum freedom for programmers within the constraints of the given problem to select the language, programming style, algorithm, and optimization necessary to get best performance for solving the given problem on a given machine. Performance is measured in terms of number of outputs per unit time or time taken to produce a specified unit of output. Six basic problems are • Random number generation Elementary function generation • Fast Fourier transform ⚫ Dense matrix computation • Rule-specified sparse matrix computation • List-vector-specified sparse matrix computation that are meant to highlight subtle hardware characteristics not exposed by the basic problem set. The motivation for these proposed benchmarks is to reflect fundamental and frequently used capabilities of supercomputers; to reflect chaining of arithmetic logical units, which is a key feature of today's supercomputers; and to reflect the memory addressing capabilities of these machines. Benchmarking is an important subject both in the United States and European Community. Major projects are in progress at Los Alamos, the National Aeronautics and Space Administration, Illinois, and many other places. This summer there was a very small workshop on benchmarking between the United States (six persons) and Japan (four persons) on Kauai, Hawaii. Oyanagi told me that he felt the underlying trend was that the U.S. participants were more interested in benchmarking applications, while Japanese (including Oyanagi) stressed the importance of architecture. He felt that one reason might be that the U.S. participants were mostly related to big supercomputer centers and had different responsibilities than the Japanese. There are two upcoming supercomputing meetings, one in Fukuoka, Japan, from 7-9 November 1991 and another in Albuquerque, NM, from 18-22 November 1991. I think it would be valuable to have discussions on this proposal during these meetings. I also hope that some of the summer workshop participants will provide a summary. Oyanagi is hoping to participate in a panel discussion at the Fukuoka meeting and may present a paper as well, but he will not be in Albuquerque.--David K. Kahaner, ONRASIA APPLIED MATHEMATICS WORKSHOP The primary goal of the workshop, which was held from 11-12 August 1991 at Tokushima University, was to facilitate the exchange of ideas on recent developments in applied mathematics by bringing together active researchers from universities and industrial laboratories. In Japan, interaction between these two societies has been rather limited. One of the principal organizers, Prof. Kametaka of Tokushima University, expressed hope that opportunities for meeting, establishing friendships, and promoting understanding between different communities would increase, through small, informal workshops as this one at Tokushima. The workshop consisted of 14 50-minute talks on ordinary and partial differential equations (ODES and PDEs), numerical simulation, and modelling. The presentations ranged from very mathematically abstract material to practical problem solving through numerical simulations. Computationally oriented talks were given on the first day and more theoretical discussions on the second. Approximately 50 scientists attended. The three talks given by Japanese industrial researchers were on control theory and fuzzy logic (Fuji Electric), scientific visualization systems (IBM Tokyo Research Laboratory), and gassolid flow simulation of discharge of powder coal from a bin (Kobe Steel). The first talk consisted of some basics of control theory and feedback systems and mathematical modelling of systems and processes. Unfortunately, time limitations only allowed the speaker to mention fuzzy logic during the last few minutes, and specific applications of the theory to systems at Fuji Electric were not presented. During the questionand-answer session the speaker did, however, touch upon the difficulties of constructing more intelligent camera systems. distributed. Requests for copies should be addressed to H. Nakanishi and S. Toyoshima at the Mechanical Research Laboratory of Kobe Steel, Kobe, Japan. Following a Japanese tradition or initiation rite, three students (Kimura of Kyoto RIMS, Sasamoto of Chiba University, and Saw Win Maung of Chiba University) presented their recent findings in numerical analysis. Saw Win Maung is one of a growing number of foreign students studying in Japanese universities. His presentation was, quite impressively, given in Japanese. The remainder of the talks were from young academicians who often attend the Mathematical Society of Japan and Kyoto RIMS meetings. Their work is already widely known in the general mathematical community. Proceedings from the workshop will be compiled by Prof. Kametaka of Tokushima University and should be available by mid-October 1991. Inquiries should be addressed to Prof. Yoshinori Kametaka Tokushima University The second talk began with an overview of graphics tools developed at IBM. The capabilities of the tools were illustrated through specific examples, i.e., visualization of air and dust flow pattern simulation inside cleanrooms and memory disks, three-dimensional translucent imaging of brain tumors, and modelling and rendering of varieties of --Mei Kobayashi, IBM Japan textures. The overview concluded with a 3-minute video of Edo Castle that was presented at SIGGRAPH. The remainder of the talk consisted of a JAPANESE NATIONAL AEROSPACE LABORATORY'S (NAL) COMPUTING DIRECTIONS A major application of supercomputing is to the design of airplanes and related vehicles. In fact, the goal of aircraft designers is to be able to simulate with fine spatial resolution the air flow over a complete plane. Simulations can be done now, but computer time and memory limit the practical resolution that can be achieved. This is a worldwide research activity. Theoretical and algorithmic developments have been significant since the 1930s, earlier for some situations. A very excellent statement of at least one Japanese scientist's opinion of this field was recently given by Prof. Kunio Kuwahara in an interview with Jaap Hollenberg in Supercomputer, July 1991 (ISSN 0168-7875): "We need more power and less modeling." Kuwahara runs the Institute for Space and Astronomical Sciences (ISAS) in Tokyo. His view is a very conservative one: I am only a computer user not a computer architect. I am only interested in solving partial differential equations.... We just solve the Navier-Stokes equation, that is all; very easy. ... I do not like to model. I like a direct approach to that phenomenon [turbulence], and take a very fine grid, then it is not difficult anymore, the difficulty is only the computer power. When asked about special-purpose computers, Kuwahara remarked that these "will always become obsolete very soon, because you cannot change it. My vector machines are rather flexible." Recently, a colleague mentioned to me that at the Ninth Plane Computational Aerodynamics Symposium, 12-14 June 1991, held at the Japanese National Aerospace Laboratory, another view was presented in a paper titled "Specifications for a Parallel Computer for CFD," by Mr. Hajime Miyoshi National Aerospace Laboratory 7-44-1 Jindaiji-Higashimachi Chofu, Tokyo 182 Tel: +81-422-47-5911 Fax: +81-422-8569 |