Shulou(Shulou.com)11/24 Report--

How do Chinese supercomputers grow from scratch, from mainframes to supercomputers?

In March 1950, the Central people's Radio broadcast a letter saying, "although Liangyuan is good, he has not lived in his hometown for a long time."

The author of the letter is Hua Luogeng, a mathematician who has just arrived in Beijing via Hong Kong from the United States. this is written by him on his way back to China, with the aim of calling on overseas intellectuals to return to China to participate in socialist construction.

In the letter, Hua Luogeng shouted that enlightening sentence: "Science has no national boundaries, scientists have their own motherland."

"Liangyuan is good" comes from a feeling when Sima Xiangru bid farewell to King Liang Xiaowang in the Han Dynasty, which shows Sima Xiangru's lofty ambition in all directions.

At that time, many Chinese scientists went to Europe and the United States to avoid the flames of war, devoted themselves to scientific research, and received preferential treatment.

Hua Luogeng has been employed as a professor at the University of Illinois in the United States.

Under the call of the people's Republic of China, many scientists, including Hua Luogeng, resolutely chose to return to the motherland from overseas.

After returning, Hua Luogeng returned to Tsinghua Garden as head of the Mathematics Department of Tsinghua University.

In the summer of 1952, feeling that the United States was carrying out research on tube computers, Hua Luogeng advocated and took the lead in setting up China's first electronic computer research group in the Institute of Mathematics of the Chinese Academy of Sciences with Min Naida, Xia Peisu, Wang Chuanying and others.

At that time, China's computer field was deserted, and forerunners were holding torches to explore the unknown.

Min Naida wrote China's first "tentative ideas and plans for computer research", which opened the prelude to the development of electronic computers in China.

In March 1956, in the outline of the long-term Plan for the Development of Science and Technology 1956-1967 formulated by New China, the computer became one of the "four emergency measures" for the development of science and technology.

At the same time, Hua Luogeng was appointed director of the Preparatory Committee of the Institute of Computing Technology of the Chinese Academy of Sciences, developed a general-purpose digital computer, and sent people to learn from the "Big Brother" Soviet Union.

In 1958, under the principle of "based on domestic imitation and then self-design", China successfully imitated the 103 computers of the Soviet Union's Mmur3 mainframe computer, with a speed of 1500 operations per second, realizing a leap from zero to one.

Although China's computer research and development is only 12 years away from the invention of the world's first computing ENIAC by the United States, there is already a considerable distance between them in terms of talents and equipment.

The United States has carried out computer research since World War II, and there are hundreds of thousands of experienced researchers from the military to the civilian community, and they are well funded.

At this time in China, not to mention computer research, many people do not even know what a computer is, let alone see it with their own eyes, and their funds are even more stretched.

At that time, with the help of Bell Laboratories, the American IBM Company successfully developed the RCA501 type of computer constructed by transistors.

Due to the use of transistor logic elements and fast magnetic core memory, the computing speed of the RCA501 computer has increased from 5000 times per second to hundreds of thousands of times per second, and the storage capacity of the main memory has also increased from several thousand KB to more than 100, 000 KB. At the same time, there is also a simple operating system using high-level language and its compiler.

Compared with the traditional tube computer, the transistor can not only realize the function of the tube, but also has the advantages of small size, light weight, high efficiency and low power consumption.

In 1959, China developed 104 imitations of the Soviet Union's CM- Ⅱ computer, barely 10,000 times, but with 4200 tubes and 4000 diodes assembled, it covers an area of more than 400 square meters.

After the delivery of the 104 aircraft to the aerospace and military industrial departments, it completed a number of scientific research operations, including the first atomic bomb.

Although there is still a big gap between 104 and the American RCA501 computer, it is almost the same as the computer level of Britain and Japan which started in the same period.

2 in the face of the shortage of talents, the Preparatory Committee of the China Institute of Computing, in cooperation with many famous colleges and universities in China, held four computer training courses in succession, and successively trained more than 700 computer researchers. It has laid the talent foundation for China's computer industry.

In the wild era, this small team, like the faint fluorescence in the long night, lit up China's computer industry.

After the end of the "honeymoon" between China and the Soviet Union, the Soviet Union withdrew all its aid to China, including computer experts and equipment.

China's computer researchers have not stopped, relying on collective wisdom and dedication to overcome difficulties in times of turmoil.

In 1960, Xia Peisu led a team to successfully develop China's first self-designed small general-purpose electronic digital computer 107, which was delivered to the University of Science and Technology of China, which was still in Beijing at that time, which was also the first scientific research computer in a university in China.

In 1964, China's first self-designed large-scale general-purpose digital tube computer 119 came out; in the same year, Fudan developed a 602 electronic digital computer programmed in machine language.

In June 1965, the Institute of Computing of the Chinese Academy of Sciences successfully developed the first transistor computer 109 B with a speed of 90,000 fixed-point operations per second and 60,000 floating-point operations per second.

The subsequent 109C machine played an important role in the trial production of "two bombs" and was known as the "meritorious machine".

Throughout the 1960s, because the domestic computer research and development was carried out around major national defense projects, it only pursued to continuously improve the computing speed, and did not consider much about the overall performance and popularity of the computer, and not only spent a lot of money, but also ignored the needs of social production and construction, not to mention the concept of mass production.

In the 1970s, the United States, the West and other developed countries have developed a miniaturized integrated circuit general-purpose digital electronic computer, and widely used in the private sector.

Therefore, in January 1973, the fourth Ministry of Machinery Industry held the "first Professional Conference on Electronic computers (7301 Conference)" in Beijing, which made it clear that in the future, "we must abandon the technical policy of simply pursuing higher computing speed and determine the policy of developing series machines."

After this meeting, China began the embryonic form of the development of the computer industry in the following decade.

Amid red flags and the noisy sound of gongs and drums, Chinese computer researchers are stepping up research and development while outlining the computer models needed for economic construction.

In August 1973, China's first million integrated circuit mainframe computer 150 was born, with a main memory of 130K. Equipped with multiple programs and operating systems, with a speed of up to 1 million operations per second.

Since then, China's mainframe computers have gradually shifted to the level of economic construction, shouldering a new mission in China's petroleum exploration, weather forecasting, scientific computing and other fields.

It is worth mentioning that the DJS-131 computer, which was produced in 1975, is a desktop computer with a memory of 4cm 32K (which can be expanded to 64K) and a computing speed of 500000 times per second.

DJS-131 minicomputer produced a total of 334 units, but it was the most widely used and most stable domestic digital computer at that time, and it was simultaneously used in 23 provinces and cities in the fields of post and telecommunications, electric power, railway, communications, medical treatment, earthquake, scientific research, transportation, industry and national defense construction.

At the same time, the North China Computing Institute organized 57 units across the country to jointly carry out the development and design of DJS-200 series computers and DJS-180 super minicomputers, which opened the prelude to the localization of microcomputers.

The subsequent emergence of NCI-2780 super minicomputer, TJ-2000 series computer and AP array processor and other products gradually make China's computer industry embark on the road of serialized mass production.

Always adhere to the "walking on two legs", China's computer industry continues to maintain gratifying achievements on the road of miniaturization and large-scale.

In November 1976, the Institute of Computing of the Chinese Academy of Sciences successfully developed a large-scale general-purpose integrated circuit general-purpose digital electronic computer with a speed of 10 million operations per second, which was then widely used in aerospace, national defense and other fields.

Those who participated in the research and development of these projects were all the students who came out of the training course of the China Institute of Computing at that time.

The once-swaying spark eventually became a scientific pedigree and spread out in various fields of computer application in China to open the inheritance.

(3) through independent research and development, China's computer industry has completed the transition from the second generation to the third generation, especially through application to promote computer research and development, the combination of production and use, and promote the localization of microcomputers.

Although China has been unremittingly carrying out research and development in the field of large computers, there is still a large technological gap between China and the United States.

Since the early 1960s, the United States has developed the third generation computer with integrated circuit (IC) as the main body, and began the process of commercialization.

Among them, IBM released IBM 360 system mainframe computer, which can run many different programs at the same time, and extend to the field of word processing and graphics processing, and the mainframe computer is many times ahead of China.

Take China's relatively advanced 104s at that time as an example. This large computer, which later participated in a number of large-scale national defense research projects, could perform 10,000 floating-point operations per second, while the 7030 computers designed by IBM for the US military in the same period had a floating-point performance of 600,000 per second.

The performance of floating-point computing determines the efficiency and accuracy of scientific research. China has tried to buy it for more than $7 million, but failed to do so under special historical circumstances.

Even in the field of minicomputers, the United States is a big step ahead.

In the 1970s, the United States, through DEC's minicomputer PDP-11, began to try to use Unix and C language for programming to meet the needs of various scenarios and really make computers enter the stage of "serving the people."

At that time, the performance and output of China's mainframe computers were still difficult to meet the various construction needs, and it was still a difficult journey.

In 1976, Clay Company of the United States launched the world's first supercomputer with a computing speed of 250 million times per second.

"mainframe computers" and "supercomputers" seem similar, but there is a big difference between the two.

To put it simply, large computers use special instruction systems and operating systems, and because they can only perform non-numerical calculations (data processing), their applications are generally limited in the commercial field; while supercomputers use general-purpose processors and UNIX or UNIX-like operating systems, are good at numerical computing (scientific computing), and can be widely used in more sophisticated and complex scenarios such as national defense, aerospace, meteorology, industry, and so on.

Supercomputer is the real "national weight", which brings together tens of thousands of processors and can carry out massive data high-speed operations. it is not only an important symbol to measure a country's comprehensive national strength, but also the fundamental guarantee of national information construction.

In the face of the realistic gap, the Chinese government has decided to learn and master the policy of cutting-edge supercomputer technology by purchasing equipment.

After China and the United States began to resume contacts, the United States exported two supercomputers to China as a gesture of friendship, but let Chinese computer researchers experience a history of humiliating "glass houses."

At the end of 1976, outgoing US President Ford signed an agreement on the export of two Cyber172 supercomputers to China.

However, the United States has tampered with this computer, and its computing performance is far from the actual level, and the United States also requires that the computer can only be used in geological exploration, not in other ways.

What makes Chinese scientists and technicians most angry and helpless is that the computer is set up in a special glass computer room, but the key is managed by the United States, and it must be approved by the Americans before each use.

After the completion of the operation, the US side will immediately block the glass house, and the operation log will be submitted to the US government for review on a regular basis.

After the introduction of Cyber172 from the United States, China bought three Hitachi M-series supercomputers from Japan for weather forecasting.

But the US side played a trick, not only asking the Japanese side to significantly reduce the performance of the export version to China, but also asking for the establishment of an audit system for personnel on duty and monitoring logs, similar to the practice of Cyber172.

This situation continued until the early 1980s, when China purchased various types of large-scale and supercomputer equipment from the United States, and the key technologies were closely blocked by each other.

For Chinese computer researchers, this is a terrible thing to look back on.

Yang Xuejun, academician of the Chinese Academy of Sciences and former president of the University of National Defense Science and Technology, said of this history: "this is an everlasting pain in the hearts of Chinese researchers, just as farmers have no food and mothers have no milk to feed their children."... "

Under the stimulation of "choking" again and again, Chinese computer researchers realize that it is an unrealistic fantasy to try to achieve self-improvement by buying advanced foreign equipment, and that self-reliance is the only way out.

4 March 1978: Deng Xiaoping listened to the computer development report and made it clear that the system of the Commission of Science, Technology and Industry for National Defense was responsible for the development of the first 100 million supercomputers, saying that "if China wants to carry out the four modernizations, it cannot do without supercomputers (supercomputers)."

In May of the same year, the project was named "785 supercomputer" at the meeting of China's supercomputer program, and General Zhang Aiping, then director of the Science and Technology Commission for National Defense, named it "Yinhe".

The Galaxy computer has a planned performance of 100 million floating-point operations per second, which is 100 times faster than the most advanced computer ever developed in China.

The director of the computer Research Institute of Changsha Institute of Technology (National University of Defense Science and Technology), Tzu Yun-Kui, who is in his twenties, issued a military writ: "even if we risk my life, we must get our own supercomputers out!"

Ci Yungui is a great master in Chinese computer industry. From tube computers to transistor computers to integrated circuit computers, he is a major participant in every upgrade of Chinese computers.

The 785 supercomputer project team also declared its position: 100 million times per second, not a lot at once! Six years without procrastination! The budget does not exceed a penny!

Grievances can be confided, difficulties must be overcome, and the rise of the scientific and technological strength of a great power will eventually depend on the unremitting efforts of a generation.

The research team first referred to the overall architecture of the most advanced supercomputer Cray-1 in the United States, and then began the division of labor and cooperation. If the domestic supporting equipment is directly used and cannot achieve independent research and development, it will be directly imported from countries with loose conditions for the export of spare parts.

But the team has a basic principle: the equipment can be used from other countries, and the technology must be your own.

In order to achieve this goal, the researchers checked one by one in the test tape piled into hills; in order to ensure the stability and reliability of the machine, 25000 winding wires, 120000 winding joints and more than 2 million solder joints were carefully examined.

Finally, Ciyungui team creatively put forward the structure of "two-way quantity array", which greatly improved the operation speed of the machine and realized the planning and development task one year ahead of schedule.

On December 26, 1983, China's first 100 million supercomputer, Yinhe-1, passed the national technical appraisal and was born.

Yinhe-1 is an important milestone in the development of high-speed computers in China, marking that China has become the third country after the United States and Japan to design and manufacture supercomputers independently.

After the celebration, each backbone of R & D staff gave a bonus of 500 yuan, but they did not ask for it, and all of it was donated to the computer project they were working on.

It is a pity that during the nearly 20 years of research and development of the "Galaxy" series, many people in the team died of illness due to high work intensity and lack of nutrition and medical conditions.

These researchers know that their research funds are limited and their own salaries are meagre. In order not to cause trouble to R & D, they have chosen to hide their illness and procrastinate again and again.

They are Professor Zhong Shixi, 49, Associate Professor Xu Xianfu, 43, Associate researcher Wang Yumin, 41, Lecturer Zhang Shusheng, 40, and Assistant researcher Yu Wulong, 35. They deserve to be remembered forever!

Among the relics of these researchers is the third Wave, a book by American writer Alvin Toffler, in which many people draw a line with red and blue pencils. "like the ancestors of the revolution, our mission is to create the future."

Time bows and salutes here, the world is heavy, the national scholar is unparalleled.

In the same year that "Galaxy No. 1" appeared, the "five-character font" of the 26-key scheme was born, which was the first version of the five-character input method that was later well-known at home and abroad, and the inventor Wang Yongmin was called the modern "Bi Sheng".

"five Stroke input" with domestic Dmur2000 Chinese character intelligent terminal and ZD-1110 character display terminal, ancient characters and computer systems representing modern science and technology meet. Those bouncing characters collide into sentences and are linked together to interpret the cultural world that belongs to the Chinese people on the screen.

This year, China's first large-scale vector computer system, 757, was also successfully developed, with 10 million vector operations per second and 2.8 million scalar operations per second.

Chinese scientific research institutions have also used the Great Wall 100 (DJS-0520 microcomputer), a microcomputer developed by six institutes of the Ministry of Electronics, which has initially acquired the main features of personal computers.

Outside the window is the world of information, the "third wave" is sweeping in, and human society has opened a new chapter.

Under the background that IBM's PC compatible computer has become the mainstream personal computer in the world, China's computer industry proposes to "follow the PC of IBM".

It's just that this kind of "follow-up" actually has no physical objects and no drawings, and it is entirely up to the computer manufacturers to explore on their own. In the end, most of the "touch" computers are irrelevant, and neither the performance nor the operating system can achieve mass production.

On February 16, 1984, Deng Xiaoping touched the head of a young computer player at the National 10-year Science and Technology Achievement Exhibition held in Shanghai and said firmly, "computer popularization should start with dolls."

The old man's advice made computer courses like a blowout, and many schools offered computer courses one after another, from simple typing exercises to advanced Basic programming, which kindled Chinese people's enthusiasm for learning computers.

People are flying like fingers on the keyboard, as if they were pouring out the urgent voice of an era of progress.

In 1985, the Chinese realized the need for complete Chinese information processing on computers by loading the Chinese-made microcomputer Great Wall 0520CH with a Chinese operating system.

In the same year, the computer company that sold Lenovo cards changed its name to Lenovo.

Driven by brands such as the Great Wall and Lenovo, a large number of computer manufacturing enterprises have emerged in China, such as Sitong, founder, Tongchuang, Shida and so on, which have become the leaders in driving the development of China's computer industry. Let China's computer industry from the third generation to the fourth generation.

This is a rich and colorful history, but it is followed by a regrettable period of stagnation in China's computer industry.

With the gradual popularity of Sino-US relations, many advanced technologies can be introduced smoothly. Therefore, the development mode of China's computer and semiconductor electronics industry has changed from "innovation first, introduction as a supplement" and paying attention to basic research to simple introduction, giving up the catch-up plan for the research and development of semiconductor general circuits.

The integrated circuit and semiconductor industry, which was not far behind the world, was suspended due to lack of funds, a large number of scientific research teams were disbanded, and some researchers were even arranged to become computer room administrators, at a loss as to what to do.

In fact, at that time, China could already copy CPU chips such as 6800 and 8080, which was not inferior to many developed countries in technology.

However, due to the shrinking project funding and the mistake of R & D direction, the CPU research has not been carried out independently, thus laying a hidden danger for the computer chip industry in the future.

Fortunately, some experts in the Chinese scientific community have realized the gap between themselves and the international cutting-edge science and technology and kept calling for it, which enabled China's supercomputer projects to catch up with each other at a faster speed.

In March 1986, four scientists, Wang Daheng, Wang Yanchang, Yang Jiayun, and Chen Fangyun, based on the strong scientific and technological strength behind the Star Wars program put forward by the United States, put forward a proposal to catch up with the world's high and new technology.

Deng Xiaoping expressed support for the proposal, and soon after, the "High-tech Research and Development Plan" was released. Because the four scientists submitted the letter and Comrade Deng Xiaoping approved it in March 1986, the plan is also called the "863 Plan."

A year later, at the computer Application Technology Research Institute in Beijing, the Chinese sent their first email: "Across theGreat Wall we can reach every corner in the world."

From that moment on, China and the world have been closely linked.

In the 863 plan, "intelligent computer" is listed as a separate theme project (that is, 306 theme), which is undertaken by the Institute of Computing of the Chinese Academy of Sciences.

The Computing Institute of the Chinese Academy of Sciences is the cradle of computer development in China, where the first general-purpose digital computer and the first general-purpose CPU chip were born, which was an important behind-the-scenes contribution to the success of "two bombs and one satellite."

At that time, the world had just set off an upsurge of artificial intelligence, and Japan's "five-generation machine" was developing rapidly in this field. Therefore, the "306 theme" followed the strategy of Japan's "five-generation machine" and began the research of intelligent computers.

In the process of research and development, experts found that the main goal of Japan's "fifth generation machine" is to achieve natural language translation, but the application market is narrow and the maintenance cost is high. At the same time, the US R & D targets are personal computers, supercomputers and the Internet, and the prospect seems to be even broader.

There is no time to wait, and the expert group has resolutely adjusted its research direction from "intelligent computer" to "high-performance parallel computer system", which has spawned the accelerated development of China's supercomputer series.

Professor Li Guojie, who is in charge of the intelligent computer project, has just returned from studying in the United States. he realized that at that time, almost all domestic computer research and development was done behind closed doors, and the research and development time was extremely long, and often the machine development was already lagging behind. always one step behind the market.

As a result, Li Guojie sent a team to Silicon Valley to "jump the queue" to absorb foreign advanced technology and update it into the project at the same time. Moreover, he also developed and designed the domestic operating system SNIX on the basis of the original UNIX source code.

In October 1993, the peak computing speed reached 640 million times per second, and the "Dawn-1" supercomputer with fully symmetrical shared memory multi-processing structure was born. As a whole, it is similar to the architecture and technology of Intel Company in 1990.

Subsequently, the dawning 1000 supercomputer also began its market-oriented operation and entered the ranks of enterprise services.

In 1997, with the dawning Tianchao 1000A settled in Liaohe Oilfield, Chinese supercomputers completely broke the market monopoly of imported products and washed away the humiliation of the former "glass house" in the field of supercomputing.

During this period, with the continuous emergence of domestic computer brands, the price of personal computers that used to cost tens of thousands of yuan each has fallen to less than 10,000 yuan, popularized to the homes of ordinary people, and the door of the information age is officially open to the Chinese people.

In the two decades after crossing the new millennium, with the recovery of national strength and the layout of scientific research, Chinese supercomputer research institutions are catching up with each other and developing like a raging fire.

In 2008, the dawning 5000 supercomputer of the Chinese Academy of Sciences was successfully developed with a computing speed of more than 100 trillion times.

Only a year later, the "Tianhe-1" supercomputer of the University of National Defense Science and Technology appeared, and China became the second country in the world to successfully develop trillions of supercomputers after the United States.

In June 2010, the "Nebula" trillions of computers of the Chinese Academy of Sciences won the second place in the 35th TOP500 ranking of supercomputers, successfully crowding into the top three supercomputers in the world.

Six months later, the "Tianhe-1A" supercomputer of the University of National Defense Science and Technology directly won the first place on the list.

At the same time, the trillions supercomputer "Shenwei Blu-ray" of Jiangnan computer Research Institute took the lead in completing the localization of CPU.

Dawning, Tianhe and Shenwei have become the dazzling "three musketeers" of high-performance computing projects on the track of supercomputing in China.

With the deepening of supercomputing research, China has successively built seven institutions, including the National supercomputing Jinan Center, Shenzhen Center, Wuxi Center, Zhengzhou Center and so on.

During this period, cloud computing, big data, artificial intelligence and other innovations based on supercomputing architecture are constantly staging a new chapter.

In 2018, Dawn, Tianhe and Shenwei have entered the supercomputing research and development of class E (10 billion calculations per second) in the field of supercomputer competition, and gradually realize the national production of CPU and accelerators.

In 2021, in the 58th global supercomputer TOP500 ranking, 173Chinese supercomputers entered the list, accounting for 34.6%. The second place in the United States was 149, accounting for 29.8%.

However, although the number of supercomputers in China exceeds that of the United States, it still lags behind the United States and other countries in terms of comprehensive computing power.

China's Shenwei Taihu Lake Light ranked sixth in the global supercomputing TOP500 list in the first half of 2022, making it the best among the "three Musketeers".

But there is still a 10-fold gap in computing power before the world's first E-class supercomputer, the Border supercomputer (Frontier) of Oak Ridge National Laboratory in the United States.

It can be said that in the face of China's supercomputing system, the situation has never stopped.

In the 1960s, Gordon Moore, one of Intel's founders, predicted that the number of transistors on integrated circuits would double steadily every 18 months and maintain that momentum for decades to come.

Since then, Moore's Law has become the law of the times and an evolutionary metaphor for the unknown world.

From desolation to fertile fields, China's supercomputers have been in the middle of ancient chess games, becoming increasingly clear and gathering majestic forces in the changing circumstances of the times.

Shakespeare said in the Tempest: all the past is a preface.

In front of China's supercomputer is a rapidly changing wave of digitization, but there are no regrets and the outcome is still uncertain.

This article comes from the official account of Wechat: Guishi Technology (ID:lishikeji2016), author: Wang Jian

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.