Guide: "Shenwei · Taihu Light" for the first time has passed the 1 billion calculation of the integer level. E-level computing is still a peak to look at for the current computing technology capabilities of human beings. Conquering it is challenging. .

OFweek Electronic Engineering Network News "Shenwei Taihu Light" refreshed the world's fastest computer and continued computing speed shortly. On July 4, Zhongke Shuguang announced the launch of an E-class high-performance computer (referred to as "E-class supercomputing" for short). 10 billion calculations) prototype system development project. Because this "next crown" recognized by the supercomputing community can well support major applications in big data, deep learning, cloud computing and other fields, people's eager expectations for its potential role even exceed the desire for computing speed.

"E-level pre-research" three heads together

The reason why "supercomputing" is widely concerned is that it reflects the comprehensive national strength and scientific and technological innovation capabilities of various countries. When powerful countries around the world began to compete for E-level supercomputing, China also wrote 10 billion billion supercomputers and related technologies into the national "Thirteenth Five-Year Plan", and hopes to complete it around 2020. It is expected that Four years ahead of the United States.

The E-class supercomputing research and development plans in Europe and Japan will also set the completion time in 2020. Since "Tianhe II" topped the supercomputing TOP500 (the world's fastest 500 machines), the number one supercomputing powerhouse in the United States Has missed this ranking for 3 consecutive years. Is the United States behind? No.

The United States Department of Energy (DOE) is pursuing the development of ten billion billion computing systems through the National Strategic Computing Plan (NSCI) and cooperation with industry and academia. It is only scheduled to be after 2023.

In China, the “three heads together” approach is adopted, and a prototype system development project of E-level supercomputing is led by China Shuguang, National University of Defense Technology (National University of Defense Technology) and Jiangnan Institute of Computing Technology (Jiangnan Institute).

The technical strength of these three units in the field of supercomputing is needless to say. All the supercomputing equipment appearing in the top ten of the TOP500 list of global supercomputing are from these three companies. Their technology, products and market performance are unique and indispensable.

The most prominent feature of the Jiangnan Institute, which has "Shenwei · Taihu Light", is the use of national-produced CPUs, which are also strong in other independent technologies. The National University of Defense Technology has very good network technology and is unique in autonomous high-performance Internet networks. Adopt a general-purpose CPU, and its representative works are "Tianhe No. 1" and "Tianhe No. 2"; Shuguang is the only one of the three institutions that fully follows the commercialized market operation, and its overall plan is not limited to the realization of E-level supercomputing. It focuses on the applications and services of future E-class supercomputing.

In response to the inevitable competition in the development of prototype systems, Zhang Yunquan, a research fellow at the Chinese Academy of Sciences and a secretary general of the High Performance Computing Committee of the Chinese Computer Society, said that China has adopted a competitive system in the development of E-level supercomputing. After evaluation, the unit adopted the best solution to integrate into the national E-level supercomputing research and development system. By then, "the three units will likely have intersection or cooperation."

It is worth noting that these three units are currently receiving "E-level supercomputing prototype system development project", that is, pre-research work before the E-level supercomputing project officially enters the agenda. In this regard, Zhang Yunquan said that similar pre-research projects are "very necessary". The development of prototype systems can verify some key technical assumptions, test and improve some key technical difficulties, and clear obstacles for the final construction of all systems to avoid emergence. Big technical errors and problems.

Li Guojie, an academician of the Chinese Academy of Engineering, said: "Whether it is the" Tianhe "or the" Light of Taihu Lake ", it has reached the world's number one, indicating that China's high-performance computer design capabilities are already the world's leading." Not only that, this time another TOP500 announced another Important data shows that 167 of the TOP500 are made in China and 165 in the United States, indicating that the hardware level of China's high-performance computers is also among the highest in the world. These capabilities have laid the foundation for China to move towards E-level supercomputing.

"E-level challenge" break the wall first

For the first time, "Shenwei · Light of Taihu Lake" has passed the integer level of one billion calculations. The E-level calculation is still a peak to look at for the current computing technology capabilities of human beings. Conquering it is a challenge.

Li Jun, president of Zhongke Shuguang, said in an interview with reporters: "On the surface, the goal of 10 billion supercomputers is to increase speed, but to build it, the world is facing programming walls, power walls, and reliability walls. , Storage walls and several large walls of innovative E-level applications. "

Shao Zongyou, vice president of Zhongke Shuguang, said: "The Americans said after the top of the" Shenwei · Taihu Light "top 500, what can the stuff do? This actually gives us a proposition, which is an E-level super The application of computing is as important as the realization. "

Zhang Yunquan also said on many occasions that although the application and optimization of E-class computing are still rarely mentioned, this is a problem that the Chinese supercomputing industry must face and solve. It is urgent to find out the future application of E-class supercomputing. , As important as the construction of E-class supercomputing.

Some people have done calculations. The computing power achieved by "Shenwei Taihu Light" in one minute is equivalent to the continuous calculation of 32 years of calculation by a calculator of 7.2 billion people worldwide, and 10 times the computing power of "Shenwei Taihu Light" E-level computing can achieve tens of billions of calculations per second.

"When the computer performance reaches this level, it is difficult to find related issues in the traditional scientific computing field." Said Shao Zongyou, vice president of Zhongke Shuguang, "from the perspective of sales, one-third of high-performance computers have been available in five years. It is used for data processing. This is because massive data needs to be stored and deeply processed to realize its value. The rise of the era of big data has opened up a clear road for high-performance computers. The first half of E-class computing is very narrow and must be used. Finding application problems outside the field of scientific computing, and extending traditional supercomputer applications to cloud services and data processing is the most important application link to solve future E-class computing. "

Today, the reason why people spare no effort to challenge E-class supercomputing is that these future computers will not only be faster, but also provide new solutions to the increasingly severe problems of climate, energy, health, and imbalanced development facing humankind. Solutions.