From
[email protected] Wed Aug 14 02:55:35 1991
From:
[email protected] (Rick Schlichting)
Newsgroups: comp.research.japan,soc.culture.china
Subject: Kahaner Report: Computing and related S&T activities in China (PRC)
Date: 12 Aug 91 17:29:26 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 [
[email protected]]
Re: Computing and related S&T activities in China (PRC)
12 August 1991
This file is named "computing.ch"
ABSTRACT. An overview of computing and related activities in China (PRC).
INTRODUCTION AND SUMMARY.
This report is a complement to earlier reports on Chinese computing,
(xian.rpt, 14 Aug 1991; tiawan, 17 Jan 1991; hongkong.91, 11 July 1991).
It is based on one visit to PRC during June 1991 and reading of some
English literature and English translations.
In many ways mainland China (People's Republic of China-- PRC) is a
modern industrial nation, and in others a poor third world country.
High tech industries in China are struggling because of the usual
problems associated with partially developed countries. There are
additional problems related to restrictions on trade from the US and EC
because of political differences. Lack of good communication facilities
such as computer networks and fax machines means that information flow
is erratic; several of the scientists we spoke to were unaware of
activities in neighboring Chinese institutions, while others appeared to
be exceptionally well connected. Electronic mail is just beginning to
appear. Access to computers is generally erratic. There is a heavy
emphasis on simulation and theory when experimentation might be better.
Basic theoretical research (pencil and paper) is excellent, especially
in applied mathematical analysis. We saw no parallel computing activity
although reading available reference materials indicates that some work
is in progress. We were not able to assess research in computer
science either, except it was clear that computer engineering is very
active. Fuzzy logic, for example, has a tremendous following, and there
are a great many applications being developed for industrial processes,
perhaps more than in the US. (See the anecdotal list of applications
given below, as well as papers on the topic listed in the forthcoming
report (xian.rpt, 14 Aug 1991).
Access to Western research information also appears spotty. Some
lectures at a conference we attended in Xi'an suggested a very out of
date view of Western literature. On the other hand senior scientists
travel to and study in the West and read and write English fairly well.
But many Chinese scientists have not travelled outside of their country,
and have not had much opportunity to use spoken English. There also
appears to be substantial library facilities at least at some
institutions. Communicating in English seems to be more of a problem
than in Japan, but this is also very variable.
There is no doubt that China is trying very hard to push science and
technology as a means to move itself forward. Numerous studies have
emphasized that the country that invented paper, movable type, gunpowder,
and the compass has had no similar scientific achievements in recent
times. Four volumes of "S&T white papers" have been published since 1986
describing the Chinese government's significant policy and strategy
decisions, related materials, and regulations for scientists and others.
It is hoped, for example, that by the year 2000 high-tech industries
will become pillars of China's national economy, traditional industries
will improve their productivity, the internal consumer market will fuel
other improvements, and even military research institutes will shift
strategically toward civilian products, as is now occurring in the Soviet
Union.
Thus there is plenty of interest, ample reports, and relatively large
quantities of money to be spent on R&D, although it must be spread over
a huge infrastructure. For example, the State Science and Technology
Commission will funnel over $2B US into high-tech development during the
'90s. For better or worse, China has over 7500 scientific research
organizations at the county level or above, a high-tech work force of
more than 3 million including about 400,000 scientific research
personnel, almost as many as in the US or Japan, and more than in
England. Many of the policies for promoting S&T are common to other
industrializing countries and will not be detailed here. A typical
quote, "we should strengthen reverse engineering and digest, absorb,
improve, and develop imported technology, take advantage of our vast
population and raise design starting points. An example is importing and
digesting French engine technologies as a basis for reverse design and
absorbing new engine technologies from foreign countries for development
and innovation to design and develop our own new type of engine.."
At the same time that the Chinese government is trying to open itself to
the international scientific community there is also a substantial
amount of of old-line inertia and flowery governmental prose in
published documents. This is compounded by official concerns about
giving away secrets to foreign intelligence agents. However, all the
experiences during our recent visit suggest that individual Chinese
scientists are just as eager as those anywhere to engage in cooperation
and free scientific information exchange. Let us do everything possible
to encourage and support this attitude.
CHINESE KEY NATIONAL LABORATORIES.
A plan to build national key labs in universities and research
institutes began in 1984. The focus has been on building and operating
open labs, setting up a new research system, promoting collaboration and
cooperation, personnel buildup, establishing a good academic atmosphere,
and managing the labs. More than 60 national labs were built since 1984.
In addition, the Chinese Academy of Science (CAS) also has invested in
setting up a similar number of academic labs and institutes which are
university-like (including some students) but in a different
organizational chain. Further, more labs have been proposed, so that
there could be on the order of 150 by 1995.
The Chinese government is now focusing on the fact that building a high
class laboratory involves not only a large one-time construction cost,
but a high on-going operational cost. There is now a trend to
consolidate some labs to make their operations more cost effective, make
research funds less diffuse and make sure that research directions are
not divided too finely. There are still many problems related to "turf",
and cooperation as well as communication between labs needs improving.
There are also a number of issues related to personnel, such as getting
lab leaders of international stature (the target here is the Cavendish
Laboratory with Maxwell and Rayleigh as early directors), getting (and
keeping) enough young researchers and support technicians who will make
most of the actual breakthroughs. Of course the perennial problem of
logistic support occurs too. To operate the labs it is recommended that
"operation is more important than construction", "academic atmosphere is
more important that current academic standard", and "management is more
important than current experimental condition".
863 PROGRAM.
This is the major plan for S&T development, proposed in March 1986. It is
centered around five areas, with its main focus to catch up with advanced
technology. Five areas are of special interest.
* Biotechnology
* Information technology (see below)
* Automation technology (especially a demo production line at Qinghua
Univ, comprehensive automated manufacturing systems, high
performance sensor technology, multiphase system control technology,
and robots for precision work, work beyond 300 meters depth, and
work in difficult environments)
* New materials (especially for high temperature and shock resistance,
high malleability, for power equipment, aerospace materials,
micromaterials, thermoplastic resins, ceramic-base, semiconductor
photoelectric, optical memory, artificial crystals, membrane
materials, non-crystalline, superconducting materials, etc., to the
late 1980's levels.)
* Energy resource technology The Chinese government has also initiated a
"Torch Program" to ensure that any advances of 386 will be put into
commercial/industrial use.
INFORMATION TECHNOLOGY (IT).
A large number of articles in the Chinese official, popular, and
technical literature have emphasized the importance of IT as part of
China's High Technology Development Plan, also called 863 Plan, see
above. In some articles IT is described as the most important of all the
386 projects. The IT aspects are divided into three general areas.
* Information acquisition
The plan is to develop a broad variety of information acquisition
and processing technologies for industrial and agricultural needs
especially in the areas of infrared detection, adaptive optical
telescopes, imaging radar technology (including satellite-carried
synthetic aperture radar and infrared focal plane technology), high
speed real-time signal processing and graphics.
* Optoelectronics
The plan is to develop all kinds of novel optoelectronic devices
and associated system integration technology for sensing, computing
and communications, and study new ways to fabricate VLSI in order
to lay the technical and material groundwork for new information
acquisition systems, computers, and communication equipment. It is
generally accepted in China that if the 20th century is the era of
electronic information, then the 21st will be the era of
optoelectronic information.
* Intelligent computing
The plan is to put the best people and resources in advanced
computer technology and AI together to track the latest
developments in the world, investigate the theoretical basis and
key technology of intelligent computers, and promote the widespread
use of AI to push for an intelligent computer industry in China.
Processing Chinese characters is an important part of many
projects associated with intelligent computing.
All together there are about 300 projects, 2,500 scientists and
technicians, and about 80 organizations. The Chinese are definitely
attempting to pool resources into national research centers. For
example, the intelligent computing work has been centered in the CAS
Institute of Computing Technology. About one third of the projects have
been completed, and the Chinese estimate that about one quarter of these
are competitive with international scientific standards.
The Chinese have stated that "we should encourage international joint R&D
activities and strengthen cooperation among open laboratories, research
centers, research institutes and higher-learning institutions in China. A
unified, orderly, competitive and mutually complementary network of
cooperation should be established." There is also great interest in
having visiting researchers as well as sending Chinese students to other
countries. For this, as well as other international activities, however,
some non technical issues are also relevant. For example, "we must
intensify their ideological and political education to firm up their
patriotism, national pride and confidence, and faith in socialism," etc.
As of 9/90 the Chinese government has broken out its information
technology project accomplishments as follows.
Topic World Domestic | Intermed Design Disseminate Total
Level Leader | Results Final Applications
-----
Intelligent
Computing 12 35 | 21 21 5 47
Optoelectronics 7 31 | 33 3 2 38
Information
Acquisition
& Processing 3 10 | 8 0 5 13
-- -- -- -- -- --
Total 22 76 | 62 24 12 98
The descriptions below attempt to give a very brief picture of some of
these projects. The remarks are fragmented and have all been gathered
from available documentation. I have not had personal contact with these
projects, so the descriptions may not be accurate. I should note that the
Chinese scientists I spoke with were frank and realistic about the status
of their work. But in print, projects are often described in
exceptionally positive terms. This is especially true for summary
reports. If readers are interested I will attempt to obtain further
information about selected projects.
COMPUTING, GENERAL.
It is estimated that there are about half million PC type computers in
China, with about two-thirds of Chinese manufacture, compared to about
25% in 1981. In 1985 about $1.2B US was spent on computer imports; in
1989 this had been reduced to $389M US. It is now hoped that Chinese
systems can be exported, perhaps up to $1B US by 1995. Production
includes complete systems, peripherals, monitors, printers, magnetic
recording equipment, as well as system and application software. A major
thrust of work has been input of Chinese characters. Computers
manufactured in China are marketed under the names of Great Wall, Taiji,
Zijin, and Legend. For example, Legend (Beijing Legend Computer Group,
part of the Chinese Academy of Science CAS) markets 286, 386, 386SX, and
486 systems. A 386 system (33Mhz) has also been developed by CAS, with
66Mbit/sec transfer rate between memory and cpu, and a RISC based
floating point unit. This system can be purchased with an independently
designed 1280x1024 color card. In 1990 about 100K 286 mother boards were
shipped to the US and Europe. China's first electron tube calculators
were also made at the Chinese Academy of Science (CAS) in 1958--a copy of
a Soviet model; current production uses LSI technology. Changzou
Electronic Computer Plant also markets a laptop based on a 4.77MHz 80C88
and supports both Chinese and standard Ascii characters. Another model
uses memory card technology based on an independently designed and
developed card.
To the best of my knowledge China currently has no Western
supercomputers. However, I was told last fall by representatives of
Convex Computer that there were several sales close to finalization, and
in March 1991 it was announced that 4 C120 Convex minisupers were
approved for sale to arrive in late spring. Applications for these
computers are seismological, petroleum prospecting, simulation of oil
reserves, and weather. I do not know exactly where these systems are
going.
NEURAL NETS/EXPERT SYSTEMS.
China's first neural net conference was held Dec 1990 in Beijing with
representation from 100 Chinese institutions resulting in over 350
technical papers in the proceedings.
Chinese Academy of Science (Institute of Automation) reports that they
have applied a neural net model to diagnosing reactor faults, and have
actually used this in practice.
Naval Academy of Engineering professors have been promoting the use of
neural nets for inference engines, claiming that they can be more
flexible, fault-tolerant, and are more natural for learning. They have
also published very basic theoretical results on new learning algorithms
for multilayered nets which can find global minimums. Other work has
shown by simulation (on digitized data of the silhouette of a US Naval
and a Soviet Naval ship) that a simple neural net can be used for ship
silhouette recognition independent of translation, scale, rotation, or
aspect changes.
CAS scientists have developed a Chinese speech recognition system using
artificial neural net concepts, and including a voice operated text input
system capable of recognizing over 20K dictionary entries. It is claimed
that tone-recognition is over 99% and word-recognition is over 90% even
with compound words.
CAS researchers have developed a general purpose expert-system language,
TUILI which can be referenced from either C or Prolog, and is claimed to
be superior to Prolog.
Qinghua Univ has Chinese character recognition system (THOCR-90) that can
handle various fonts, alphabets, numbers, and other symbols, and employs
neural-net techniques for pattern matching. Recognition speed is about 30
chars/sec on a 386/33. Another (somewhat faster) system has been
developed at Nankai Univ.
FUZZY.
Work in this area has been active since 1979. There have been numerous
practical applications and research results reported, perhaps more than
in any other country outside of Japan. In 1987 Chinese scientists
presented almost one quarter of the papers at an international fuzzy
society conference in Japan, equal to those presented by Japanese
attendees. Further, I was told that the (US) Journal of Fuzzy Sets and
Systems is being flooded with papers from Chinese scientists. In China
some specific applications include the following (these are all
anecdotal).
* CAS (Institute of Semiconductors) has built almost two dozen multi-
logic circuits (two-value, multi-value, continuous-value, e.g.
fuzzy). Chinese researchers are very much aware of work in Japan by
Yamakawa who has built fuzzy-logic circuits using CMOS technology and
in fact two researchers from Tsinghua Univ (Beijing) report that they
have also built (using 5 micron) CMOS, various basic circuits that
they claim are simpler and more reliable than Yamakawa's.
* Electronics Industry Computing Center has developed a fuzzy-
information processing command decision-support system designed
specifically for use in communications networks used for military
command decision support.
* Math Dept of Beijing Teachers' Univ (BTU) developed a fuzzy inference
engine capable of 15M inferences/second. BTU has also built hardware
claimed to be able handle 30K (sic) basic rules with 2000
input/output variables for fuzzy based inference, and that products
are being marketed this year.
* Beijing Normal Univ has proposed a new method for fuzzy inference,
called Truth-valued Flow Inference, which they claim can represent
knowledge more effectively than traditional fuzzy inference.
* North China Industrial Univ (NCIU) did a 6 year study on microcomputer
fuzzy control theory and applications.
* Various papers on fuzzy control simulation and adaptiveness of fuzzy
control.
* Staff & Workers College of Shanghai Instrument and Meter Industrial Co
have developed proportionality-factor-type fuzzy controller and
three-loop fuzzy controller.
* Hunan Univ is using fuzzy composite control in gas smelter control
systems.
* Wuha Fist Sci Research Inst of Light Industry has a fuzzy control
system for glass kilns.
* Handran Resin Plant has a fuzzy control system for PVC resin
polymerization process.
* Univ of Petroleum used fuzzy control in large hysteresis systems for
chemical engineering processes.
* Chemical Fertilizer Industrial Inst, Jilin Industrial Univ, and Metal
Products Inst have used fuzzy control for compressor regulation,
power factor compensation, DC reversible-speed regulation, and a
digital dual-frequency-channel amplitude frequency instrument.
* Air Force Inst of Engineering developed a fuzzy quantitative
evaluation expert system for aircraft maintenance.
* Nanjing Univ use fuzzy theory to analyze evaluation of sound quality
in high-fi sound systems.
* Armed Forces Eco Inst used natural fuzzy language in expert systems.
* Zhenjiang Shipping Inst used dBASE III to develop a fuzzy database
inquiry system.
* Central China Univ of S&T has used fuzzy quantization in knowledge
engineering.
* Univ of S&T for National Defense developed a PCB logic
diagram/manuscript pattern-recognition system in C, using fuzzy
relations among text areas. On a PC AT it is claimed to have a speed
of 912 characters/minute and an accuracy of 98.9%. Another
institute has also used fuzzy techniques and has complete system
including software and scanner for installation on a variety of
micros.
It is not possible to know just how successful these projects have been
as their descriptions are often brief, but the point is that Chinese
scientists are exploring applications aggressively. At the same time
there does not appear to be much emphasis on basic research for its own
sake in this area.
PARALLEL PROCESSING.
My visit did not lead to learning about any parallel computing except for
old ELXSI systems that appeared to be not seriously used, but the
literature does describe several projects of potential interest.
Wuhan Digital Engin Res Inst developed the 980 STAR systolic array
computer. This is composed of a host (Intel 310/286), image subsystem,
interface processor, and systolic computing array. The latter consists of
a 4x4 matrix of cells. Pipeline beat is 200ns, and each cell is capable
of 10mips. Maximum throughput is 80Mbytes/sec. Various system software
has also been developed. This is considered mid 1980's technology. A new
model is currently being developed, and is claimed to have 100 times
greater performance.
Programming on China's YH-1 (Galaxy-1) supercomputer has been via
vectorized Fortran. (I am trying to learn more about this.) Current work
is also in progress to develop a parallel programming capability, centered
on C. There is also research on analyzing blocks of vector Fortran for
vector block dependencies and a Pascal scheduler has been developed. It
is claimed that some Cray 1 calls such as VF.MUL (vector float mult) can
be reduced from 644 to 381 beats. The Galaxy is a 100mips machine. One
specific application has been has been for processing seismological data
from sections taken from petroleum prospecting areas.
East China Inst of Computing Tech has a microcomputer parallel processing
system based on 12 Inmos T800 transputers grouped into three clusters
(boards), and connected to a PC-AT. One processor on each cluster is the
main control processor. Peak performance is 80mips or 10MFlops.
Jiangnam Institute has developed and marketed a PC board insert with an
Inmos Transputer and 8Mbyte RAM, and software for development.
Qinghua Univ has developed boards with 1, 2, or 4 transputers (TTH-1A, -
2A, and -4A) that will plug into PC-ATs, and also developed a transputer
development system. These products are claimed to be fully compatible
with European Transputers at the late 1980's level.
Shantou & Nanjing Univ researchers have built an eight processor parallel
computer (Transcube) built around 8 Inmos T414 transputers and an
additional 1Mbyte of memory on each processor. The 8 PEs are the bottom
layer of the 3-layer system, in which the top layer is the PC AT host,
and the middle layer is an interface buffer controller. The researchers
claim that it is easy to expand the system beyond 8 PEs.
Graduate students at a research institute of Ministry of Machine-Building
& Electronics Industry have used two TI DSP chips to design a real-time
parallel signal processing circuit for a pulse doppler ground
surveillance (moving target) radar. They claim to have solved two key
problems which had caused bottlenecks related to data input and
synchronous coordination of both chips, and that the system performs at
20M multiplications/additions per second and consumes 5 watts.
Fudan Univ researchers have developed a sea-wave processing system based
on diffraction analysis of sea-waves. The system (which runs on an XT)
can estimate the spectrum of a 2D 512x512 image array with 16 azimuthal
inputs in 9 minutes with wave direction accuracy of 6 degrees, and 18
minutes for 32 inputs and accuracy of 3 degrees.
SOFTWARE.
RT/VMS is a Chinese developed military real-time operating system. The
Chinese military feels that a major threat of future wars will be from
computer viruses, rather than traditional weapons, and there has been a
substantial effort to study and eliminate computer viruses within China.
(Frankly I had not thought about this before, but it is an intriguing
view.) RT/VMS is claimed to have innovations in channel interfacing, job
scheduling, low system overhead, and hardware debugging technology.
The Commission of Science, Tech and Industry for National Defense has
developed an Ada compiler (in C), which is claimed to be the first Ada
compiler to simulate a multiprocessor environment on a single processor
Unix systems. More than 4000 modules in the compiler have passed ACVC1.10
verification as specified by DOD in 1989.
CAS has moved all the Unix SVR3.1 source code from an ATT 3B2 to a Vax
11/750 by cross compilation and reverse engineering. They have also moved
the top layers of this system to a 386/Xenix. CAS claims that this is the
first domestic (Chinese) movement of a full Unix system.
ROBOTS.
* Chinese Academy of Science reports that in the early 1980s they proposed
establishment of a robot demonstration project which began operation in
1989. The cost was 58M Yuan plus $5.9M (US) foreign exchange. This is
now called CAS Shenyang Robotics Engineering R&D Center and occupies
34,000m^2 of building space. The plan is to use this a base to produce
robots and develop and manufacture more products. Many related projects
spring from this one.
* A medium-size underwater robot (Recon-IV-300-SIA-02) is being sold to
the US. This is claimed to be an improvement on a prototype brought from
the US. 90% of this robot's parts are of Chinese manufacture. Robot has
audio, video, sonar, two hands and can perform five or six functions.
Closed-loop control of depth and direction of navigation is included,
payload is 160kg, and maximum speed is 3knots.
* A light-weight underwater robot (Gold Fish) also made in China is
used in oil drilling, sea rescue, dam inspection and repair.
* Nankai and Tianjin University have jointly developed a robot with
vision, hearing and touch capability (NKRC-03), which can automatically
control the amount of force when it grabs an object. This is claimed to
be at the 1980s level world wide and state of the art in China. This
robot uses the NKV robot language to allow high level language
programming; a standard second generation robot.
FACTORY AUTOMATION.
China Huajiing Electronics group claims that after a two year effort by
more than 100 scientists they have developed China's first independently
designed integrated-circuit CAM system, consisting of 10 subsystems
(planning, production, management, process technique, quality control,
economics, operations, materials, plate making, statistics, system
management).
Beijing Inst of Machine Tools has built a physical simulation system for
studying the various software parts of a flexible manufacturing system
based on linear rail guided carts. The system can simulate transport,
loading/unloading, control, and specific aspects of a machining
operation.
The above suggests that there is are a great many projects in the area of
manufacturing. Most of these are related to automation and associated
machine tool and numerical control equipment. For example, last year the
Second Chinese Machine Tool and Instruments Expo was held in Beijing, and
various newly developed machine tools, numerical control systems,
sensors, CAD software systems, etc., were on display (226 metal cutting
machines, 67 lathes, 37 grinders, turning centers, etc). There are, in
fact over 2000 machine tool products manufactured in China, mostly for
domestic use. However the Chinese machine-too building industry is also
trying to design products that will be marketable internationally (some
have been already) and have identified several products that they claim
are of world class, for example a six-axis, four-link hobbing machine
computer numerical control system which can be used for machining
ellipsoidal gears. Another example is a unit for machining small diameter
holes which is claimed to be twice as fast and equal in other
specifications to a Raycon (US) product. There are also cooperative
agreements with major international companies such as Fanuc-Besc, GE,
Werner-Kolb, and Schiess (German). It is not necessary to be an expert in
this area to appreciate the main thrust, that the Chinese want not only
to become self sufficient but also to become net exporters of this
technology.
OPTOELECTRONICS.
* 140Mbit/s hybrid optical terminal, and 622 Mbit/s hybrid integrated-
component module and distributed feedback laser diode.
* 1.5micron single-mode narrow-line-width tunable semiconductor laser,
semiconductor travelling-wave optical amp, Er-doped glass fiber amp,
lithium-niobate optical-waveguide high speed modulator, 1.1-1.6micron
avalanche photodiode detector, and duplex frequency-division-
multiplexing coherent optical communications devices.
* Various quantum-well optoelectronic devices.
* Optoelectronic IC made domestically.
* Superlattice growth of II-VI semiconductors and fabrication of
optically bistable devices, bismuth-silicate spatial light modulator,
molecular beam epitaxy growth of GaAs, and GaAs/Si LEDs and MESFETs
have been fabricated.
-----------------------END OF REPORT----------------------------------
