From: rick@cs.arizona.edu (Rick Schlichting) Newsgroups: comp.research.japan,soc.culture.taiwan Subject: Kahaner Report: Computing in Taiwan Date: 19 Jan 91 03:20:09 GMT Followup-To: comp.research.japan [Dr. David Kahaner is a numerical analyst visiting Japan for two-years under the auspices of the Office of Naval Research-Asia (ONR/Asia). The following is the professional opinion of David Kahaner and in no way has the blessing of the US Government or any agency of it. All information is dated and of limited life time. This disclaimer should be noted on ANY attribution.] [Copies of previous reports written by Kahaner can be obtained from host cs.arizona.edu using anonymous FTP.] To: Distribution From: David K. Kahaner ONR Asia [kahaner@xroads.cc.u-tokyo.ac.jp] Re: Int Comp Symp & 1st WS on Parallel Proc (Nat Tsing Hua Univ, Taiwan), Institute of Information Industry (Taipei), 17-21 Dec 1990. 17 Jan 1991 ABSTRACT. Two meetings at National Tsing Hua University (Taiwan), International Computer Symposium, and the First Workshop on Parallel Processing, as well as a visit to the Institute for Information Industry, (Taipei) are described. 17-21 Dec 1990. SIGNIFICANT ITEMS IN THIS REPORT.---------------------------------------- Most interesting papers were from the West, computer related research in Taiwan seems much more theoretical than practical. Best paper, about optimal mapping of mesh onto hypercube (from U. Hong Kong), had some ONR support. Numerical computation papers almost nonexistent. NEC's VP Watanabe describes his vision of Science, Technology, Industry as inseparable and claims that supercomputer technology will be used in NEC's mainframe machines. A National Supercomputer Center has just been established. Taiwanese industry is moving away from PC's. Government has various plans for computer growth and research, including Hsinchu Science Park, Institute for Information Industry, and special manufacturing focuses. OVERVIEW.---------------------------------------------------------------- The International Computer Symposium, essentially Taiwan's Information Processing Society meeting, was held for three days 17-19 Dec 1990, at the National Tsing Hua University, in Hsinchu, about 70km south of Taipei. All presentations and papers were given in English. There were approximately 400 participants; about 160 papers were presented out of almost 400 submitted. This was immediately followed by a two day Workshop on Parallel Processing, at which all the presentations were in Chinese. The Workshop had over 200 participants. Both meetings have printed Proceedings in English. A title and author list is attached to this report. I have one copy of the Proceedings and will send copies of papers to requestors if I have the time. Additional copies can be obtained by writing Department of Computer Science National Tsing Hua University Hsinchu, Taiwan 30043, Republic of China. After the meetings I spent one day in Taipei, at the Institute for Information Industry. There were only two or three Western scientists in attendance at the International Symposium and none at the Parallel Processing Workshop. Most of the participants from outside Taiwan were Taiwanese from the US, or Japanese, or other Chinese speaking scientists from Hong Kong or Singapore. I saw no evidence of attendees from mainland China, Korea, or other Pacific Rim countries. Generally the best papers were from outside Taiwan, although most of these were by Taiwan-born scientists now living in the West. I did not see much evidence that these scientists are moving back to Taiwan into senior positions. My general impression, confirmed by most of the Chinese speaking attendees, and by all the non-Taiwanese, is that Taiwanese computer research activities are heavily theoretical. This reflects the fact that presently, there are few practical parallel processing projects in the country. (I was told that there are a few computer systems oriented, experimental research projects in Taiwan, but that they are behind in all practical areas, not just in practical parallel processing. I did not get to see any of these.) In that sense Taiwanese research is behind Korea, where a reasonable amount of catch-up research is in progress. However, a few theoretical papers were very good, although the best was by two Hong Kong computer scientists on mapping of a two dimensional mesh efficiently onto a hypercube. A handful of numerical computation papers were presented, but this seems a particularly weak area. TAIWAN-GENERAL BACKGROUND.----------------------------------------------- About two hours flying time south west of Tokyo, Taiwan straddles the Tropic of Cancer (as does Hawaii). It has about 2.5% of the land mass of mainland China, and a roughly proportionate population (20million). The country is densely populated and rugged, with only one third arable land and more than 60 over-3000-meter peaks . Taiwan is the original name of the island; Formosa was a Portuguese word meaning beautiful island, used until recent years. The present government considers Taiwan part of China (hence ROC=Republic of China) although many of the scientists I spoke to did not share that view. Restrictions on Taiwanese trips to the mainland have been reduced in the past few years and several hundred thousand have visited. The present Taiwanese government still has many tough sounding laws developed in an era when relations between them and the mainland were very tense, although I was told these are being relaxed. Nevertheless there is plenty of controversy; on TV I was astonished to see a free-for-all fist fight in the "parliament". Political issues (of which I have no knowledge) aside, if Taiwan and mainland China can resolve their differences, China will be a huge natural market for Taiwanese products and expertise. Taipei, the capital, has about 2.5 million people. It is choked with traffic and heavily polluted. There is no metro although one is being built. This is surprising, as Taiwan's per capita income is over $8K, about 50% more than Korea's which has a substantial metro system in Seoul. (Viewing Seoul and Taipei, I was not able to distinguish differences in the income levels of these countries. My Korean friends tell me that the differences are noticable in the countryside.) Nevertheless it is easy to find one's way, as most streets are on a rectangular grid. Hsinchu, the site of both meetings is also the location of Taiwan's largest science based industrial park, established in 1988. There are already over 100 manufacturers, close to 20,000 people, and more than 100 PhD's working in the Park. Unfortunately, as most of the computer-types were at these meetings there was no opportunity to visit any of their labs. This would be an important stop on any subsequent trip. Hsinchu is also the location of two major universities, National Tsing Hua (with about 4500 students and about 450 faculty), and National Chiao-Tung Univ. The model for growth here seems to be Tsukuba science city in Japan. COMPUTING IN TAIWAN.----------------------------------------------------- More than 5,000 companies (about 90,000 people) are engaged in either manufacturing computer related hardware, or exporting it out of Taiwan Taiwan makes about 60% of the world's PC mother boards. However, until very recently they did not make complete systems. This has now changed. I tried several perfectly reasonable all-Taiwanese PCs, including several 386s. Some of these are beginning to be exported. I was told that Taiwan products now satisfy almost all internal PC demand. However, many of the best use Japanese monitors. I was also told that in the near future there will be a Taiwanese Unix-based workstation, for internal use. Almost all systems larger than PCs had been imported until recently, and the government has identified that as a major area for growth. There are also some strange properties of the Taiwan market. For example, until very recently there were more Wang systems installed than IBM's. An important part of computer growth in Taiwan is associated with the ability to input Chinese characters. This is a significant research activity both here and on the mainland. (Japanese is spoken by many older Taiwanese, as a result of the Japanese occupation of Taiwan during the first third of this century.) There is also a good deal of manufacturing for Japanese products; my Sony Walkman has cables clearly marked "made in Taiwan". During this trip I only was able to talk to academic computer scientists, but was repeatedly told that the best Taiwanese technology is associated with manufacturing. Apparently this is true, industrial growth has averaged over 13% per year for more than 30 years, more than 50% of the country's GNP is now generated through exports, and Taiwan is the second largest holder of US dollars (after Japan). Computer industry growth has also been very good, the only items now growing are printers and terminals. By 1995 research expenditures are projected to be about 2% of GNP, and 0.2% of the population designated as "researcher". For research in the information and electronics industries, 3% of researchers have PhD, 48% have MS, and 40% have BS. There are many government sponsored programs; discussing these in detail would require a separate report. The National Supercomputer Center (see below), Sub-Micron Parts Laboratory, and the Hsinchu Science Park are among the most recent and interesting. The ROC government has various plans for making computers even more important in Taiwan's economy. Some are general, and other are quite specific. For example, they have identified 14 specific manufactured items that they want to promote. These are - Cathode ray tube (CRT) monitors - Low-speed nonimpact printers - Low-speed impact printers - Medium speed printers - Linear scanners - Fixed (hard) disk drives - Optical readers - Chinese-English word processors - Micro and minicomputer systems with Chinese operating systems - Software for medium and small computers - Modems - Chinese language electrical typewriters - Floppy disk drives - Computer magnetic tapes An interesting area worth following is the Taiwanese development of systems larger than PCs. Already, with licenses from SUN, a workstation consortium has been developing a SPARC-compatible system. About two years ago an experimental 32-bit RISC multiprocessor machine (called MR-10), running UNIX, X-Windows, and suitable as a network file server, was introduced. This was done by the Industrial Technology Research Institute (ITRI) under the sponsorship of Ministry of Economic Affairs. I was told that the spinoffs from this project will be several special purpose machines, such as a database server, and more interestingly a vector processor, but I had no opportunity to learn about this in more detail. It seems that with possibly the exception of the SPARC-compatible workstations, all these projects are for acquiring basic computer technologies rather than for immediate commercial use. PAPERS OF INTEREST.------------------------------------------------------ The International Computer Symposium was organized into one day of tutorials followed by two days of four or five parallel sessions, each containing an invited presentation, and three or four submitted papers. With one exception the tutorials were all by Chinese now working in the US, as were many of the invited papers. The Parallel Processing Workshop consisted of only one track, but as mentioned above, all the talks were given in Chinese. Some of these were by speakers from the Symposium, presenting some new results, but also several overview papers, such as the one by Fang (Convex) on automatic parallelization techniques, or by C. Hsiung (Cray) on how supercomputers are going to fit into future computing systems. I concentrated on papers from non-Western scientists on the theory that others' work is already fairly well known. Ultimately only three that were most interesting to me are reported on here. (At any meeting with parallel sessions it is easy to overlook outstanding papers.) PAPERS (OPTIMAL MESH TO HYPERCUBE MAPPING).------------------------------ The best technical paper was by Dr. M.Y.Chan James Capel (Far East) Ltd Hong Kong and Prof. Francis Yuk-Lun Chin Department of Computer Science University of Hong Kong Pokfulam Road, Hong Kong Tel: +852 859-2178, Fax +852 559-8447. Email: CHIN@CSD.HKU.HK, or CHIN%CSD@HKUJNT.BITNET The problem the authors consider is extremely important. We are given a two dimensional mesh, and a hypercube computer. The mesh might come from a finite element calculation. A hypercube is a collection of individual computing processors (elements) that are connected in a very special way, so that the individual elements are thought of as at the vertices of a multi dimensional cube. A 3-hypercube contains 2**3 elements, a 5- hypercube contains 32 elements at the vertices of a 5-cube, etc. Hypercubes were the first commercially practical parallel computer; many systems are in use, and there is a tremendous amount of research on effective algorithms to take advantage of their structure. For a two dimensional finite element mesh it is natural to associate each mesh point with a processing element of the hypercube. But how, specifically, should this association be made? For example how should we associate 9 x 3 grid with 27 elements in a 5-hypercube (32 elements). Of course, any association will work, but an efficient one will recognize that neighbors on the grid should correspond to neighboring processors in the hypercube. The reason is that the only way two hypercube processors can communicate is along the edges of the cube. For a 3-cube this can mean that a message might need to pass through two intermediate processors to get between sender and receiver (along three edges); a 5-hypercube may require passage along five edges. This problem had already been studied by the present authors, as well as others, and Chan had provided an algorithm for doing this in an optimal way, in the process proving that it was always possible to arrange the association so that adjacent neighbors on the grid are separated by at most one intermediate processor. But that algorithm for deciding how to do this is slow and suffers from some other difficulties. The current paper describes a parallel algorithm associating any two dimensional grid to its optimal hypercube. The size of the grid is broadcast to each node of the hypercube which can then determine in constant time which grid node it will associate with. Moreover each node can also determine the communication paths it will take to reach the nodes associated with grid neighbors. Further, each node will act as an intermediate node for at most two pairs of grid-neighbors, reducing potential congestion at intermediate processors. This is a very excellent theoretical result that has substantial practical implications. The authors claim that they can extend the method to three dimensional meshes. Part of this work was sponsored by an ONR grant while Chin was at U Texas, Dallas, making it even more satisfying for me to report about. PAPERS (NEC'S VP KEYNOTE SPEECH ON SCIENCE, TECHNOLOGY, AND INDUSTRY).--- The keynote speech of the Symposium was by Hitoshi Watanabe Vice President NEC Corporation. Watanabe explained that he had invented NEC's first computer. This had 48 bit floating point arithmetic and some parallel architecture, although his original design called for 64 bit floating point. He showed us a photo of a much younger man standing in front of a very large console which is now in NEC's museum; the console was the only part that would fit. The essence of his talk was his view that science (theory), technology (design), and industry must be an indivisible activity, what he called STI activity. He called STI a chain from fundamental research, applied research, development, production, to sales and marketing that must be balanced at each link, and that the mission of engineers is to consider all these parts at the same time. He went on to illustrate this by various examples. For instance in the 1910's theory consisted of Boolean algebra, Kirchoff's topological rules, etc., design was typified by research in loaded lines, and industry in electric power systems and telephony. He went on to illustrate the point during the 1920-1950's, and then 1960 to the present. He also showed "maps" of the information industry at various periods. Frankly I didn't get much out of these slides except that integration is here to stay. The slides were reproduced as part of the Symposium materials, but his lecture itself was not available in printed form, and many interesting details were given, rapid fire. Unfortunately Watanabe did not stay around long enough to allow me to corroborate my notes. The most sensible of these are given below. Watanabe described the workstation of the near future (2000) as having a 5.25" read/write video disk, a flat panel color display with 10 million pixel resolution, connected by optical fiber (by 2010 all fiber optic) to a broadband intelligent network. Communication will be by either ATM or STM. He said that executives don't like to use keyboards, so the workstation will a touch screen, mouse, or more sophisticated input. He showed an artist's sketch of a Japanese office with these workstations, screens neatly integrated into office furniture and a small server tucked away elsewhere. He emphasized flat panel technology repeatedly to display color and high definition television displays mounted on the wall, with scenes changed under computer control. (In the 1960's Richard Hamming made similar predictions but the technology had not been invented at the time.) Integration and globalization were key themes. One interesting comment was that NEC was definitely using its supercomputing technology downstream. Watanabe explained that the "same" high performance LSI technology used in NEC's SX/3 was also being used in its water cooled 3800/20 and 3800/60 mainframes. While this is no surprise it reinforces the economic justification for expensive R&D used in small unit sales products. Of course, until this year NEC has not sold many of its SX computers. He ended his talk with a very general description of several important future research directions. Future Industry : Integrated Information Networks Basic Theory : Unified modeling theory for information networks Practical Method: Efficient method of information network integration Final Target : Useful network compiler for large scale network (several hundred thousand sub-parts) Future Industry : Global Systems Basic Theory : Elegant modeling theory for social systems Practical Method: High performance simulation Final Target : Powerful simulator for complex global systems. This was a perfect presentation to inspire the scientists and students at this meeting, but there was no opportunity to dig out some of the most important details. After Watanabe's lecture, Prof Radu Popescu-Zeletin FOCUS Directorate Research Center for Open Communication Systems GMD Hardenbergplatz 2 D-1000 Berlin 12 Germany Tel: 49-30-25499-206, Fax: 49-30-25499-202 Email: ZELETIN@FOKUS.GMD.DBP.DE who was another of the invited speakers, commented that if we really wanted to see integrated computing and communications we should visit his laboratory in Berlin. PAPERS (TESTING VECTORIZED FFTS).---------------------------------------- One of the most heavily used algorithms is fast Fourier transform (FFT). In fact it has been called the most important computer algorithm ever devised, and plays a key role in a vast number of applications, from solving partial differential equations, to image processing. Although the first significant paper was in 1965, research continues even today. Every computer vendor that markets to scientific users provides a highly optimized collection of FFT routines. Prof Chung-Ping Chung Dept of Computer Science and Infor Engineering National Chiao-Tung Univ Hsinchu, Taiwan 30050 ROC Tel: (035) 712121, ext 3728, Fax: (035) 724176 reported on his studies of several variations of the FFT, with special emphasis on their implementation for large general purpose supercomputers. One of the least well known, Ruritanian's method, performed best when Chung timed in on a Convex C-1. Unfortunately Chung used his own implementations of these algorithms, rather than tuned implementations provided by vendors, or even carefully coded ones in the public domain such as by Schwarztrauber, etc. With the dearth of work in Taiwan in numerical computation, it is a pity that when someone wants to do something he is hampered by lack of access. In this case Chung would benefit very greatly by having access either to source programs of good FFT implementations, or at least an opportunity to run his programs against some of them. Western researchers who are interested in this topic should write to Chung directly. POST MEETING DISCUSSIONS (R.C.T. LEE, NATIONAL TSING HUA UNIVERSITY).---- I had a fascinating morning with Tsing Hua's Dean and University Provost, Prof. R.C.T. Lee Dean of Academic Affairs National Tsing Hua Univ Hsinchu, Taiwan 30043 ROC Tel: 886-35-719134 (O) 886-35-711610 (H, Fax: 886-35-722713 Email: RTCLEE@TWNMOE10.BITNET Lee was also General Chairman of the International Symposium. Lee is a computer scientist who is still active. He gave me a list of publications numbering about 50, including several books. (One is to be published early next year.) He still supervises several PhD students, and his office is filled with photos of former students in wedding poses, and others with newborn children. Hsinchu as a city is clearly moving toward science, and this emphasis is reflected in the university. Nevertheless, Lee is trying to expand literature and other liberal arts appreciation with special programs, exhibits, etc. We talked for a while about problems in Taiwan with industry, especially relative to Korea. As mentioned earlier, Korea has much less per capita income than Taiwan, but Korea also has several very large companies, such as Daiwa, Pohang, etc. On the other hand, Taiwan has mostly small to medium sized companies. Lee remarked that there were advantages and disadvantages with either. Of course, large companies can have the resources needed for big time research, but are often inflexible and slow to respond to changes. A Taiwanese scientist told a story about a friend who owned a company that manufactured electronic arcade games, but when the market for these slowed recently the company was able to change almost overnight, to producing PCs. Lee felt that the ROC government needs to be more proactive in helping industry. He told me about a new government program that he had a hand in promoting that tries to do this. In it, scientists (from universities and research laboratories) come together to define some products that need to be developed, along with fairly detailed specifications. Subsequently, private industry makes proposals for the development work (which involves some technology transfer from the universities). Half of the development funds come from the government and half directly from the winning company. Lee also mentioned another program which allows engineers and technicians from industry to spend a few weeks to a few months at a University being trained in some new technology; ICs were specifically mentioned in this context. Workstations were another area where Taiwanese technicians needed retraining. Until recently, there was a great deal of PC expertise, but not much Unix development experience. When this was recognized substantial numbers of Taiwanese were sent out of the country to learn the necessary skills. Generally though, Lee acknowledges that it will be impossible to compete directly with Japan at the cutting edge of high technology, and that Taiwan should focus on those areas for which it is best suited. In his opinion, small company manufacturing is done very well in Taiwan, and definitely surpasses Taiwanese skills in computer systems. There are some anomalies though. Taiwan has a strong trade surplus (these figures change yearly) but in the area of electrical components, including connectors, etc., the net flow is from Japan to Taiwan. Lee urged me to return and visit the industrial manufacturing facilities that he could arrange and also to meet the Minister of Economic Affairs, who he characterized as a scholarly and scientific person. Unfortunately, this trip was primarily to attend these conferences and for get acquainted purposes. Finally, I remarked to Professor Lee that I was disappointed by the lack of papers about numerical computation at either of the meetings. He agreed that this was a serious problem; researchers here are not much interested in computer modeling. Of course an important reason is the lack of general purpose high performance computers. [There was one session at the Symposium specifically designated for numerical computation. It promised to have one excellent invited talk on sparse linear systems by Prof. D.J. Evans (UK) who unfortunately was unable to attend at the last moment. Of the three remaining papers, one was by a physicist who had reinvented the method of lines, and the other two by students. Attendance at this session was very low, reflecting the general interest in these topics among the attendees.] POST MEETING DISCUSSIONS (K. WU, NATIONAL SUPERCOMPUTER CENTER TAIWAN).-- The government is hoping to increase research in numerical modeling by its development of a National Supercomputing Center. Dr. Kuo Wu, Visiting Specialist National Supercomputer Center National Taiwan University Taipei, Taiwan Tel: 363-0231 3221-221 is on leave from Cray Research to help establish the Center. Wu told me that they had just received $85 million US to be spent over five years. Half will be for equipment, and half for a staff of about 100, including half researchers. A building is now being constructed in the Hsinchu Science Park. The Center expects a machine to be delivered in 1992, essentially to provide supercomputer cycles to the academic community, a la the US NSF Centers. No decision has been made on specific hardware although their budget allows for a high-end system, such as a Cray Y/MP. Wu claims that they are looking to the future and so might be interested in an SSI. My own feeling is that prestige aside, they will ultimately go for a mature system, and let larger installations be test sites for new machines. At this stage of their activity Wu and his staff are just beginning to search out potential users to learn more about their needs. POST MEETING DISCUSSIONS (K. KIN, CONVEX COMPUTER, ASIA).---------------- I also had the opportunity to talk with Dr. Kenneth L. Kin Managing Director, Asia/Pacific Operations Convex Computer Pte Ltd 1 Scotts Road #25-06 Shaw Center Singapore 0922 Tel: (65) 7334355, Fax: (65) 7334354 about supercomputer and minisuper applications. I began by asking why his office was in Singapore instead of Japan, where there are already more than 50 Convex systems installed. He explained that Convex does have an office there but the company considers Japan a separate market from the remainder of the Pacific region. He felt that too many organizations centralize themselves in Japan, and then because of all the activity there, have less incentive to consider other countries where individual potential may not be as large. For Convex, business has been brisk this year in Taiwan, Malaysia, India, etc, while many other American firms remain fixated on Japan--a mistake he felt. Of course, the Japanese are well aware of this and are actively going after this market too. Kin claimed that he also had a substantial number of potential orders from (mainland) China, but until recently had been prevented from accepting them because of export restrictions. Having such a widespread customer base causes problems too. A complete set of spare parts is stocked centrally, in Convex's case in Taiwan, and this needs to be restocked from the factory in the US. Convex emphasizes the reliability of their equipment, nevertheless there are a few engineers in the Singapore office who respond to problems. For some installations, such as the one in Sumatra (Indonesia) the contract called for an engineer on site. This has turned out to be a real hardship assignment, as the engineer has repeatedly been ill, including a bout with malaria. Kin himself is on the road most of the time. In fact his family has remained in the DC area but he gets to see them about once a month. Several of the scientists at the meeting agreed that a medium sized Convex was a good bet for Departmental computing in many mid-sized companies or universities that could not justify a larger Cray-class machine. On the other hand, having a "real" Cray has advantages too, not the least of which is its name recognition, especially with politically savvy government ministers or corporate board members. Sometimes this matters quite a lot. I asked Kin if he could generalize about his user community. He explained that oil related applications have been important recently because of concern about oil supplies from the Gulf. He also confirmed my own observations that outside of Universities and some government institutes, almost all users are running packaged applications. Only within the University environment is any substantial programming being done. [Later, I was able to verify that in a recent survey of computer languages being used, within Taiwanese government agencies Fortran is almost as popular as Cobol, twice as heavily used as Assembly, and three times as heavily used as C.] Also, like Cray, Convex is concerned about distributed memory parallel processing. Their conventional wisdom is that a reasonable model for many new installations will be to provide a parallel processor as a node on a heterogeneous multicomputer network. Finally, I mentioned to Kin my conversations with NEC's SX/3 designer Watanabe, who claimed NEC was not interested (at this time) in marketing a mini super machine. Needless to say, he was very happy to hear this. As far as Convex's current software activities are concerned, an excellent survey of automatic parallelization techniques was given by Mr. Jesse (Zhixi) Fang Software Development Program Manager Convex Computer Corporation 3000 Waterview Parkway P.O.Box 833851 Richardson, TX 75083-3851 Tel: (214) 497-4275, Fax: (214) 497-4500 Email: JFANG@CONVEX.COM Fang thought that Convex had every reason to hope that its Fortran compiler would shortly be better than Cray's at automatically finding and taking advantage of parallelizing opportunities, primarily because Convex is developing their compiler from scratch rather than having to make it compatible with older products. Personally, I am skeptical. INSTITUTE OF INFORMATION INDUSTRY (III), TAIPEI.------------------------- Institute for Information Industry 10th Floor, 116 Nanking E. Road Sec 2, Taipei, Taiwan, ROC Tel: 886-2-542-2540, Fax: 886-2-531-0760. This is a government financed organization focused on promoting information technology in Taiwan and planning its growth. It was founded in 1979, and currently has about 650 employees in half dozen locations within Taipei. It also has two Exhibition Centers in the US, one in Mountain View CA, and a second in Gaithersburg MD (about five minutes from my US home). I spent one day at III main office, but unfortunately did not get to meet any of the higher level or research staff. It would be useful to return in order to get a better picture. In addition to promotional and planning, III has several products or services that their staff has developed. Many of these are very closely related to corresponding items in the West, but have been fully absorbed by III's staff to the point where new design and implementation can be done locally. The most important of these is the Software Engineering Environment Development (SEED) project. Begun in 1989 it is hoped that this will create a good distributed software environment via integrated networks, all the related information, methods, standards, and tools needed for software development. The idea is to have as many of Taiwan's software development projects working under SEED, in order to improve productivity and reliability by using a standard environment. More specifically, SEED will build workstations, and provide communication, an operating system (Unix kernel, CHINEX), graphics window interface, database system, and mail system. The mail system, workstation tools, Form Master, Imagemaker are clearly derived from Western products, but have been modified to deal with Chinese characters. There is also a Chinese Input System built around phonetic input, much like Japanese "kana" input systems. The SEED software/hardware is developed by III staff and also with support from several Taiwanese computer companies. III is also developing a SEED Service Center where various services, such as network management, databank, consulting, training, etc., are located. SEED has also published its second version of a handbook, "Software Development Guidelines", to explain international software standards. Other countries have also established programs with some of the same goals as SEED. In Asia these include SIGMA (Japan) and SUPER (Korea). III has been involved in expert system research. This would not be considered at a terribly exciting level by most Western standards. Nevertheless they have build rule-shell, frame-shell, and Chinese-shell expert systems, and have been applying prototypes to taxation auditing, and auto malfunction reduction. All concerned now believe that more professional training needs to be done in order to get expert systems more firmly rooted in the country. III also publishes several excellent survey reports with detailed statistics of computing technology in Taiwan. I can provide copies of these to a limited number of requestors. FINAL NOTE.-------------------------------------------------------------- Taipei has the National Palace Museum, an astonishing collection of the best Chinese art and culture. English is used extensively in tours, exhibit labels, and lectures. Not to be missed! TITLES AND AUTHORS.------------------------------------------------------ INTERNATIONAL COMPUTER SYMPOSIUM Algorithms 1. Some NP-complete combinatorial problems in computational geometry (Invited) M. H. Alsuwaiyel and D. T. Lee, Northwestern Univ., U.S.A. 2. Flexible functional cell design for floor planning Y. C. Chang, T. Y. Ho and L. H. Hsu, Chiao Tung Univ., R.O.C. 3. Parallel algorithms for optimal routing between ring of terminals S. C. Chang, IBM T. J. Watson Research Center, U.S.A. 4. An average case analysis of a resolution principle algorithm in mechanical theorem proving H. Hu, C. Y. Tang and R. C. T. Lee, Tsing Hua Univ., R.O.C. Database 1. Algorithms for concurrency control on R-trees J. K. Chen and Y. F. Huang, Tatung Inst. of Tech., R.O.C. 2. Feasibility of index-scan join method on cube-connected multicomputers J. Y. Tien and W. P. Yang, Chiao Tung Univ., R.O.C. 3. A query evaluation scheme on general purpose MIMD parallel machines W. S. Lee, Purdue Univ., U.S.A. P. C-Y. Sheu, Rutgers Univ., U.S.A. 4. Designing a highly parallel database server using off-the shelf components K. A. Hua, IBM Mid-Hudson Lab., U.S.A. H. C. Young, IBM Almaden Research Center, U.S.A. Computer Communications 1. From broadband ISDN to multimedia computer networks (Invited) R. Popescu-Zeletin, Technical Univ. Berlin, Fed. Rep. of Germany 2. A solution for packet switching of broadband ISDN C. M. Weng and J. J. Li, Tsing Hua Univ., R.O.C. 3. On the statistical analysis of some continuous ARQ schemes with repeated transmissions J. K. Hu and T. H. Lee, Chiao Tung Univ., R.O.C. 4. A throughput analysis of Go-Back-N ARQ protocol with copy transmission under priority Y. Fukagawa, M. Ooki, Nishinippon Inst. of Tech., Japan S. Murakami and S. Yoshida, Kyushu Inst. of Tech., Japan Distributed Systems 1. Protocol conversion (Invited) M. T. Liu, Ohio State Univ., U.S.A. 2. A deadlock-free O( N) mutual exclusion algorithm for distributed systems Y. I. Chang, M. Singhal, and M. T. Liu, Ohio State Univ., U.S.A. 3. A distributed probabilistic scheme with unequal weights S. Y. Sung, Univ. of Singapore, Singapore 4. Remote tasking facilities in DMINIX system S. R. Tsai and W. J. Ueng, Cheng Kung Univ., R.O.C. Image Processing 1. A new access method based on signature file in text retrieval H. Fu and S. Y. Lee, Chiao Tung Univ., R.O.C. 2. Recognition of handwritten Chinese characters via short line segments B. Chen and H. J. Lee, Chiao Tung Univ., R.O.C. 3. Multi-sensor planning for target recognition with constant time complexity Y. L. Liu and L. C. Fu, Taiwan Univ., R.O.C. 4. Applications of ICM segmentation algorithm C. C. Chen, Tsing Hua Univ., R.O.C. 5. 2-D EAG Model for Describing and Identifying hand drawing objects M. Zhao, Asian Inst. of Tech., Thailand Algorithms 1. Heap-ordered trees, 2-partitions, and continued fractions W. C. Chen and W. C. Ni, Taiwan Univ., R.O.C. 2. Unordered tree contraction L. H. Hsu and J. J. J. Wang, Chiao Tung Univ., R.O.C. 3. An O(N*M*T) spatial match retrieval algorithm D. J. Buehrer and C. C. Chang, Chung Cheng Univ., R.O.C. 4. On a maximum stable matching problem J. J. M. Tan, Chiao Tung Univ., R.O.C. 5. Performance of heuristics for the longest common subsequences problem F. Chin and C. K. Poon, Univ. of Hong Kong, Hong Kong Database 1. An overview of the SunSet database management system F. Golshani and S. W. Dietrich, Arizona State Univ., U.S.A. 2. Inserting tuples into coexistent views C. H. Cho and Y. C. Lee, Univ. of Michigan, U.S.A. 3. Solving the view update problem in nested relations H. H. Yao, Yuan-Ze Inst. of Tech., R.O.C. 4. Using partial pictures to view agreement in database relations J. H. Leuchner, L. Miller and G. Slutzki, Iowa State Univ., U.S.A. Computer Communications 1. Performance analysis of packet switching networks with window control J. S. Wu and S. F. Yang, Central Univ., R.O.C. 2. Tree algorithm with nonuniform traffic pattern J. H. Huang and Y. C. Lai, Taiwan Univ., R.O.C. 3. Polling system with K=1 limited service discipline and noncyclic service order C. J. Chang, I. H. Chen, and L. C. Hwang, Chiao Tung Univ., R.O.C. 4. On the complexity of generating the test sequence without synchronization problem L. T. Liu, Telecommunication Lab., R.O.C. G. H. Chen, Taiwan Univ., R.O.C. C. S. Lu, Telecommunication Lab., R.O.C. 5. An executable rule-based specification and its incremental implementation for communication protocols C. M. Huang, Y. I. Chang, and M. T. Liu, Ohio State Univ., U.S.A. Distributed Systems 1. Resource allocation for fault tolerant systems using external backups Y. Huang, AT&T Bell Labs., Murray Hill, U.S.A. S. K. Tripathi, Univ. of Maryland, U.S.A. 2. A distributed self-stabilizing algorithm for finding bridges M. S. Tsai, C. C. Tsao, and S. T. Huang, Tsing Hua Univ., R.O.C. 3. A robust and efficient algorithm of replicated ID table management in GALAXY distributed operating system X. Jia, K. Shimizu, H. Nakano, and M. Maekawa, Univ. of Tokyo, Japan 4. An optimal diagnosis structure for distributed systems F. W. Shih, IBM Watson Research Center, U.S.A. K. Nakajima, Univ. of Maryland, U.S.A. 5. Localizing the effects of faults in distributed object based systems L. Lin, Unisys Corporation, San Jose, U.S.A. M. Ahamad, Georgia Inst. of Tech., U.S.A. Image Processing 1. Earley and MPM methods for attributed grammar parsing K. Y. Huang and C. H. Chen, Chiao Tung Univ., R.O.C. D. R. Leu, Univ. of Houston, U.S.A. 2. A flexible parallel architecture for relaxation labeling algorithms S. Y. Lin and Z. Chen, Chiao Tung Univ., R.O.C. 3. Array architecture design for video communication applications L. G. Chen, Y. S. Jehng, and T. D. Chiueh, Taiwan Univ., R.O.C. 4. The development of HIGIPS the High-speed General-purpose Image Processing System F. H. Yao, A. Tamaki, K. Kato, Kyushu Inst. of Tech., Japan 5. Prototyping a hypermedia document model with knowledge representation K. Utsumiya, K. Korida, K. Yoshida, and S. Futamura, Oita Univ., Japan Algorithms 1. The Most vital edges with respect to the number of spanning trees in two-terminal series-parallel graphs L. H. Hsu, R. H. Jan and Y. Y. Lee, Chiao Tung Univ., R.O.C. 2. Guarding problems in the gallery H. Y. Fang and C. Y. Tang, Tsing Hua Univ., R.O.C. 3. On the relocation problems with some constraints M. T. Lin and S. S. Tseng, Chiao Tung Univ., R.O.C. 4. An efficient annealing schedule for multiconstraint zero-one knapsack problem F. T. Lin, C. C. Hsu, and C. Y. Kao, Taiwan Univ., R.O.C 5. The complexity of periodic maintenance S. Baruah, L. Rosier, I. Tulchinsky and Univ. of Texas, Austin, U.S.A. Database 1. A binary implementation of single-key-lock scheme for dynamic access control C. C. Chang, Chung Cheng Univ., R.O.C. T. C. Wu, Chiao Tung Univ., R.O.C. 2. A record-oriented cryptosystem for database sharing C. H. Lin, Tsing Hua Univ., R.O.C. C. C. Chang, Chung Cheng Univ., R.O.C. R. C. T. Lee, Tsing Hua Univ., R.O.C. 3. Time-space optimal numerical database for large-scale scientific applications S. C. Park and J. P. Draayer, Louisiana State Univ., U.S.A. S. Q. Zheng, Louisiana State Univ., U.S.A. 4. Optimal MMI file systems for orthogonal range retrieval C. Y. Chen, Tsing Hua Univ., R.O.C. C. C. Chang, Chung Cheng Univ., R.O.C. R. C. T. Lee, Tsing Hua Univ., R.O.C. 5. A comprehensive model for fuzzy information management A. Ramer, Univ. of Oklahoma, U.S.A. Artificial Intelligence 1. Solve assumption-based truth maintenance problems by constructing deduction graphs H. L. Li, Chiao Tung Univ., R.O.C. 2. Paradox-free logics as the fundamental theory of knowledge Engineering J. Cheng, Kyushu Univ., Japan 3. General solutions of boolean equations C. S. Wu, Industrial Tech. Research Inst., R.O.C. 4. The modal-functional unification method for automating reasoning in a logic of branching time M. C. Chan, Univ. of New South Wales, Australia T. J. Richards, La Trobe Univ., Australia 5. Preliminary evaluation of LogDf: A data-driven parallel architecture for logic programs C. C. Tseng, Chiao Tung Univ., R.O.C. P. Biswas, Texas Instrument Corp., U.S.A. Distributed Systems 1. The optimal checkpoint interval in time Warp parallel simulation S. Y. Hwang, Chiao Tung Univ., R.O.C. E. D. Lazowska and Y. B. Lin, Univ. of Washington, U.S.A. 2. Algorithms for the assignment of pipelined and acyclic tasks in ring-structured distributed systems J. P. Sheu, Central Univ., R.O.C. C. Y. Chen, Tatung Inst. of Tech., R.O.C. 3. A transformational approach to process partitioning using timed petri nets G. M. Yee and C. M. Woodside, Carleton Univ., Canada 4. A teamware workbench for multimedia information management N. Matsuura, S. Ichimura, S. I. Hiraiwa, K. I. Okada, and Y. Matsushita, Keio Univ., Japan 5. A virtual machine based execution environment for stand-alone application systems T. H. Chang, C. S. Hwang, and S. R. Tsai, Cheng Kung Univ., R.O.C. Image Processing 1. Line segment matching for 3-D computer vision using a new iteration scheme S. L. Chou and W. H. Tsai, Chiao Tung Univ., R.O.C. 2. Line segment correspondence finding using path and shape coherences C. H. Wu and C. L. Huang, Tsing Hua Univ., R.O.C. 3. Moment preserving edge detection and its application to image data compression H. S. Hsu and W. H. Tsai, Chiao Tung Univ., R.O.C. 4. Image coding using eigendecomposition and its comparison with DCT P. W. Chin, J. F. Yang, and C. H. Chen, Cheng Kung Univ., R.O.C. 5. A new adaptive coder for runlength coding J. F. Wang, L. W. Lee, and M. J. Yang, Cheng Kung Univ., R.O.C. Algorithms 1. Design of a parallel theorem prover for first order logic W. T. Chen, T. R. Chou, H. J. Liu and K. R. Hsieh, Tsing Hua Univ., R.O.C. 2. A linear recurrence solver F. C. Lin and K. L. Chung, Taiwan Univ., R.O.C. Y. N. Yeh, Academia Sinica, R.O.C. 3. On the design of parallel algorithms for non-linear space-time representations H. A. Fencl and C. H. Huang, Ohio State Univ., U.S.A. 4. An O(1) time algorithm for generating computation tree forms B. F. Wang and G. H. Chen, Taiwan Univ., R.O.C. 5. Conflict-free broadcasting algorithms for graph traversal and its applications G. H. Chen, W. W. Liang, and L. T. Liu, Taiwan Univ., R.O.C. Z. C. Yeh, Tatung Inst. of Tech., R.O.C. Computer Vision 1. Corner Detection and Segmented Regression J. S. Huang, Academia Sinica, R.O.C. 2. Automatic thresholding for edge detection based on human visual perception J. S. Chen, W. N. Lie, and Y. C. Chen, Tsing Hua Univ., R.O.C. 3. Collision avoidance for monitoring multiple operating robot manipulators by computer vision using 2-D camera views W. T. Chao and W. H. Tsai, Chiao Tung Univ., R.O.C. 4. A new Model-based approach for industrial visual inspection Y. N. Sun and C. T. Tsai, Cheng Kung Univ., R.O.C. 5. Fast object detection for a mobile robot using structured light in an indoor environment X. W. Tu and B. Dubuisson, Compiegne Univ., France Artificial Intelligence 1. TCGD A Time-Constrained approximate guided Depth-first search algorithm (Invited) B. W. Wah and L. C. Chu, Univ. of Illinois, U.S.A. 2. A generalized learning problem T. P. Hong and S. S. Tseng, Chiao Tung Univ., R.O.C. 3. Using both feature-based and Constraint-based Knowledge in the evidential relaxation labeling scheme C. C. Hsu and V. W. Soo, Tsing Hua Univ., R.O.C. 4. A direction-selective chart and a head-driven chart parser for Chinese sentence analysis L. F. Chien and L. S. Lee, Taiwan Univ., R.O.C. K. J. Chen, Academia Sinica, R.O.C. Parallel Processing 1. Parallelized simulation of grids on hypercubes (Invited) M. Y. Chan and Y. L. Chin, Univ. of Hong Kong, Hong Kong 2. 3-D optical multistage interconnection networks: topology and routing K. S. Huang and J. K. Peir, IBM T. J. Watson Research Center, U.S.A. 3. The generalized cube connected cycles network C. S. Yang and W. Y. Chen, Sun Yat-Sen Univ., R.O.C. 4. Optimal broadcasting on SIMD hypercubes C. T. Ho, IBM Almaden Research Center, U.S.A. VLSI/CAD 1. On the design of concurrent error detectable multiply and divide arrays C. L. Wey, Michigan State Univ., U.S.A. T. Y. Chang, Tsing Hua Univ., R.O.C. 2. On easily testable array multipliers K. J. Lin and C. W. Wu, Tsing Hua Univ., R.O.C. 3. Test pattern generation for sequential circuits by dirceted-search method J. F. Wang, P. C. Chen, C. G. Wang, B. D. Liu and T. Y. Kuo, Cheng Kung Univ., R.O.C. 4. A new hybrid sense algorithm for three-layer via minimization with practical constraints S. C. Fang, Taiwan Univ., R.O.C. K. E. Chang, Taiwan Normal Univ., R.O.C. W. S. Feng, Taiwan Univ., R.O.C. 5. Worst case analysis of greedy heuristic to find a minimum total-jogs layout for river routing K. H. Teo and T. C. Tuan, Univ. of Oklahoma, U.S.A. 6. A new net distribution algorithm for cell layout P. W. Huang, H. R. Lin, and Y. C. Hsu, Tsing Hua Univ., R.O.C. Algorithms 1. Voronoi diagrams for moving objects (Invited) H. Imai and K. Imai, Univ. of Tokyo, Japan 2. Constant time algorithms for sorting and computing convex hulls B. F. Wang and G. H. Chen, Taiwan Univ., R.O.C. 3. Computing minimum spanning trees and relative neighborhood graphs in the polygonal Minkowski metric S. N. Yang and C. H. Cheng, Tsing Hua Univ., R.O.C. 4. The NP-hardness of two connected cell layout problems on cellular radio networks N. F. Huang and R. S. Lin, Tsing Hua Univ., R.O.C. Computer Vision 1. Thinning and stroke segmentation for document symbols R. H. Cheng and Z. Chen, Chiao Tung Univ., R.O.C. 2. A mathematical morphology approach to euclidean distance transformation F. Y. Shih, New Jersey Inst. of Tech., U.S.A. O. R. Mitchell, Univ. of Texas, U.S.A. 3. Skeletonization using the newtonian potential J. H. Chuang, Univ. of Illinois, U.S.A. Artificial Intelligence 1. Design and implementation of a generalized blackboard structuring subsystem C. S. Ho, R. C. Chen, and J. M. Lin, Taiwan Inst. of Tech., R.O.C. 2. An adaptive expert system for control decision in battle ship simulation W. W. Loa and C. L. Wang, Chung San Inst. of Science and Technology, R.O.C. 3. Z-III: a PC-based fuzzy expert system shell K. S. Leung, Y. T. So, A. Leung, and W. S. F. Wong, Chinese Univ. of Hong Kong, Hong Kong 4. Hardware encoding of concept hierarchy H. H. Chung, M. T. Hsieh, and S. Y. Hwang, Chiao Tung Univ., R.O.C. Parallel Processing 1. An interactive technique to analyze the performance of finite-buffered multistage interconnection networks H. S. Jang, M. Singhal, K. Y. Lee and M. T. Liu, Ohio State Univ., U.S.A. 2. Refining the diffusion approximation for modelling multiprocessor systems J. S. Wu and J. Y. Wang, Central Univ., R.O.C. 3. A fault-tolerant VLSI interconnection network using cellular array H. Mei and T. C. Lin, Univ. of Texas, U.S.A. 4. GALAXY A parallel computer for large data/knowledge base systems K. A. Hua, Univ. of Central Florida, U.S.A. C. Lee, IBM Mid-Hudson Laboratories, U.S.A. 5. A decomposition strategy for extending double precision and reducing ROM cost in conversion between binary and binary logarithm J. F. Hsueh and H. Y. Lo, Feng Chia Univ., R.O.C. VLSI/CAD 1. A new approach to CMOS operational amplifier synthesis C. Y. Kuo, L. G. Chen and T. M. Parng, Taiwan Univ. R.O.C. 2. Automated synthesis of asynchronous digital circuits from graphic timing specification W. K. Chia and Y. H. Kuo, Cheng Kung Univ., R.O.C 3. A new synthesizer of microprogrammed controllers for multicycling operations J. F. Wang, M. H. Sheu, Y. L. Jeang, W. Y. Lee and J. Y. Lee, Cheng Kung Univ., R.O.C. 4. FIPOC: A finite state machine synthesizer J. P. Wang and C. S. Lin, Taiwan Univ., R.O.C. 5. A technology-driven simulation system based on code-generation design methodology A. T. Yang and J. T. Yao, Univ. of Washington, U.S.A. Algorithms 1. Parallel algorithms for sparse linear systems (Invited) D. J. Evans, Loughborough Univ. of Tech., U.K. 2. Solving diffusion equations using vector/parallel algorithm L. C. Chien, Academia Sinica, R.O.C. H. C. Hsu, Taichung Inst. of Commercial, R.O.C. 3. Solving large sparse linear system by iterative method on multiprocessor N. S. Chang, J. S. Wang and C. Y. Tang, Tsing Hua Univ., R.O.C. 4. An efficient parallel algorithm for Jacobian computation J. Z. C. Lai and J. S. Kuo, Feng Chia Univ., R.O.C. Neural Networks 1. Fuzzy combinatorial optimization by neural networks M. Sakawa, Hiroshima Univ., Japan 2. A neural network for solving the satisfiability problems K. T. Sun and H. C. Fu, Chiao Tung Univ., R.O.C. 3. Solving the travelling salesman problem using hopfield nets with time-varying energy function 763 S. D. Wang and C. M. Tsai, Taiwan Univ., R.O.C. 4. A neural network approach to constrained task allocation problems Y. S. Tseng, J. L. Wu and J. H. Huang, Taiwan Univ., R.O.C. Software Engineering 1. Software engineering in the year 2000 (Invited) C. V. Ramamoorthy, Univ. of California, Berkeley, U.S.A. 2. A requirement organization approach for object-based construction of software systems D. J. Chen, S. K. Huang and J. Y. Chen, Chiao Tung Univ., R.O.C. 3. An approach to module invocation based performance dependency analysis for Ada programs P. S. E. Chang and C. Y. Hsieh, Univ. of Oklahoma., U.S.A. 4. Rule-based software performance evaluation C. C. Lien, C. C. Yang and S. J. Hwang, Taiwan Inst. of Tech., R.O.C. Parallel Processing 1. Heterogeneous Multicomputers (Invited) H. T. Kung, Carnegie Mellon Univ., U.S.A. 2. VLSI architecture of the MPA-I superscalar microprocessor B.T. Chang, Y. J. Oyang and F. C. Lin, Taiwan Univ., R.O.C. 3. Analytical and simulated performance of the USC orthogonal multiprocessor C. M. Cheng and K. Hwang, Univ. of Southern California, U.S.A. 4. Correlation between cache size and coherence protocol overhead J. C. Wang and M. Dubois, Univ. of Southern California, U.S.A. Signal Processing 1. Computing pseudo wigner distribution by the fast Hartley transform S. C. Pei and I. I. Yang, Taiwan Univ., R.O.C. 2. The unconstraint minimization approach of the pisarenko spectral estimation S. J. Chern and T. M. Lan, Sen Yat-sen Univ., R.O.C. 3. An independence technique and its application to transform coding C. J. Kuo, Chung Cheng Univ., R.O.C. 4. Signal reconstruction from Hartley transform magnitudes G. S. Chen, J. L. Wu and L. S. Lee, Taiwan Univ., R.O.C. 5. A new implementation of generalized order statistic filter by threshold decomposition L. W. Chang and S. S. Yu, Tsing Hua Univ., R.O.C. Computer Graphics 1. Approximation to planar curves by the minimum number of piecewise parabolas H. M. Sun, Academia Sinica, R.O.C. M. C. K. Yang, Academia Sinica, R.O.C. and Univ. of Florida, U.S.A. 2. A developmental model of herbaceous plants with flowers--the growth of orchid Y. J. Shih and Z. C. Shih, Chiao Tung Univ., R.O.C. 3. Polar form for surface modelings T. V. To and H. N. Phien, Asian Inst. of Tech., Thailand 4. A CG technique to generate pictures of bird's-eye view of terrain models for constructing holographic stereogram Y. Aoki, T. Naruse and Y. Sakai, Hokkaido Univ., Japan 5. Modeling deformable objects using a linkage model P. Y. Ku, Inst. of Information Industry, R.O.C. Y. T. Ching, Academia Sinica and Chiao Tung Univ., R.O.C. Neural Networks 1. Decision support using fuzzy neural logic network F. H. Nah and S. C. Chan, Univ. of Singapore, Singapore 2. A smart spell checking system - a neural net approach P. Y. Chen and J. S. Wang, Tsing Hua Univ., R.O.C. 3. A numerical stable pipeline net VLSI architecture for the isomorphic Hopfield model P. R. Chang and K. S. Hwang, Industrial Tech. Research Inst., R.O.C. 4. An improved pocket learning of neural network Y. P. Chu, S. T. Wang and C. M. Hsieh, Chung Hsing Univ., R.O.C. Software Engineering 1. Automatic transformation of high-level logic specifications into efficient imperative programs A. Pan and B. R. Bryant, Univ. of Alabama, Birmingham, U.S.A. 2. Design and implementation of the EC++ language H. J. Chen and B. J. Liu, Central Univ., R.O.C. 3. Software reliability measurement during operation phase based on a test-effort dependent reliability growth model J. Hishitani, S. Yamada, M. Kimura and S. Osaki, Hiroshima Univ., Japan 4. An efficient lookahead algorithm for disk cache T. M. Chang, Dongguk Univ., Korea C. S. Jhon, Seoul Univ., Korea 5. Trio: an object-oriented operating system for education P. J. Lee and F. J. Wang, Chiao Tung Univ., R.O.C. Parallel Processing 1. A dependence-based loop partitioning method for multitasking in a vetor computer I. K. Chou, C. P. Chung and C. Chen, Chiao Tung Univ., R.O.C. 2. Trapezoid self-scheduling: a practical scheduling scheme for parallel compilers T. H. Tzen and L. M. Ni, Michigan state Univ., U.S.A. 3. STRAD: A tool for automatic synthesis of regular array algorithms Y. C. Lin, Taiwan Inst. of Tech., R.O.C. 4. Modeling the performance of grouping on multicomputers C. T. King, W. D. Ju and I. R. Kau, New Jersey Institute of Technology, U.S.A. 5. Study of the vectorization of FFT C. P. Chung and C. H. Tsai, Chiao Tung Univ., R.O.C. Speech Recognition 1. Recognition of phonetic labels of the TIMIT speech corpus by means of an artificial neural network J. X. Wu and C. Chan, Univ. of Hong Kong, Hong Kong 2. On the use of matrix quantization to text-constraint speaker identification M. S. Chen, P. H. Lin and H. C. Wang, Tsing Hua Univ., R.O.C. 3. A study on two-dimensional cepstrum approach for the speech recognition H. F. Pai and H. C. Wang, Tsing Hua Univ., R.O.C. 4. An initial study on large vocabulary continuous mandarin speech recognition C. H. Tseng and L. S. Lee, Taiwan Univ., R.O.C. ============================================================================== FIRST WORKSHOP ON PARALLEL PROCESSING National Tsing Hua University Dec 20-21 1990 Hsinchu, Taiwan ROC PROGRAM Keynote Speech: How to Move Parallel Processing into the Mainstream, H.T. Kung (Carnegie Mellon Univ USA) Whatever Happened to the Promised MFLOPs? C.C. Hsiuing (Cray Research, USA) Orthogonal Multiprocessor for Integrated Image Understqnding K. Hwang & D.K. Panda (Univ Southern Cal USA) On Multistage Interconnection Networks T. Feng (Penn State Univ, USA) The Role of Protocol Engineering in Parallel Processing M.T. Liu (Ohio State U, USA) Communication Issues in Multicomputers L.M. Ni (Michigan State Univ, USA) Design and Experimentation of Routing Schemes for 2D Torus Message Passing Network T.W. Hou, L.M. Tseng, W.G. Lin, S.R. Tsai, C.K. Shie (National Cheng Kung Univ) SEE: A System Evaluation Environment for Parallel Systems P.C. Yew & J.D. Bruner (U. Illinois at Urbana-Champaign USA) Parallel Algorithms for Computing the Minimum Visible Vertex Distance Between Two Polygons F.R. Hsu, R.C.T. Lee, R.C. Chang (National Chiao Tung Univ) Configurational Computation on the Processor Array with a Reconfigurable Bus System B.F. Wang, G.H. Chen, H. Li (National Taiwan Univ) An Optimal Parallel Algorithm for Generating Permutations on a Linear Array B.Y. Wu & C.Y. Tang (National Tsing Hua Univ) Cache Organizations in Shared Memory Parallel Systems K. So (IBM T.J. Watson Research Center USA) Parasight: An Integrated Programming Environment for Sequential and Parallel Programs P.Y. Chen, Z. Aral, I. Gertner, G. Schaffer, J. Grier (Encore Co USA) Research and Education of Software Engineering for Concurrent Systems K.C. Tai (US National Science Foundation, & North Carolina State Univ, USA) A 3n-state Self Stabilizing Protocol for Uniform Rings S.T. Huang (National Tsing Hua Univ) Parallel Computing and Parallel Compilers J. Fang (Convex Computer Corp) Extract Parallelism within a Nested Loop by Cycle Breaking and Loop Reordering C.M. Wang, S.D. Wang (National Taiwan Univ) Synthesizing Nested Loop Algorithms Using Nonlinear Transformation Method J.P. Sheu, C.Y. Chang (National Central Univ) Computer Scheduling Algorithms: Past, Present, and Future K.M. Baumgartner, B.W. Wah (Univ of Illinois at Urbana-Champaign) Scheduling Support for Parallel Programs on Multiprocessors H. Chang (IBM T.J. Watson Research Center) An Optimal Partitioning in the Star Network with Limited Processors M.T. Chen, S.S. Tseng (National Chiao Tung Univ) Exploitation of Multiple-Instruction-Stream Superscalar Microprocessor Architecture C Wu, R. G. Prasadh (Univ Texas at Austin) Major Architectural Features of the Spectra-I Superscale Microprocessor Y. Oyang, C. Wen, Y. Chen (National Taiwan Univ) Parallel Inference Model Based on Multi RISC-style Processing System C. Chen, C. Chung, C. Chiang, H. Fu, T. Chang, R. Ou, S. Wang (National Chiao Tung Univ) The Design of an Integrated Parallel Processing Unit with a Reconfigurable Bus System W. Chen, C. Liu, M. Fang (National Tsing Hua Univ) Two-Level Fault-Tolerant Boolean Hypercube Architecture C. Yang, S. Wu (National Sun Yat-Sen Univ) The VP/30 Compute Server: The First Mini-supercomputer in Taiwan H. Yeh, J. Tsau, P. Shiue, B. Lin, A. Chu, J. Chang (CCL/ITRI) ------------------------END OF REPORT-----------------------------------