During this time, let Ouyang Zhen find someone to build the server group of Huayu Technology, and Qin Yu was preparing for this hole card. Pen? Interesting? Pavilion wWw. ｂiquge。 ｉｎfo

"The first generation of the brain!"

This computer drawing that Qin Yu had taken out a long time ago, when it was about to be forgotten, Qin Yu finally planned to play its important role.

For supercomputers, each country has its own international understanding and view!

Not long ago, "Sunway Taihu Light", which is manufactured using Huaxia's own chips, replaced "Tianhe-2" to the top of the global supercomputer 500 list. In fact, before this, another world record has also been "low-key" generated, relying on the "Sunway Taihu Light", using the self-developed parallel computing software SCETD-PF, the Computer Network Information Center of the Chinese Academy of Sciences (hereinafter referred to as the Network Center of the Chinese Academy of Sciences) has achieved the parallel computing scale of 9 million cores using the whole machine system, and the measured performance has reached 40% of the peak performance, which is currently the largest parallel computing in the world, far ahead of the international similar computer simulation level.

Perhaps most non-industry insiders do not fully understand the significance of this achievement, but on the way of China's supercomputer constantly hitting the world's first crown, in addition to praise and applause, doubts and reflections have not stopped, such as high energy consumption costs, low actual utilization rate, long application software development cycle, insufficient talent echelon reserves, etc. Some people even pointedly pointed out that possession and computing speed are certainly a "sounding" indicator, but if they can't be used, no matter how high the computing power is, it is just a fascinating number.

"Without software development and utilization, supercomputing is a pile of broken copper and iron. In the same way, if there is only a high-performance machine and no talent to do a high level of service, the machine will be obsolete in 5 years. In the view of Chi Xuebin, deputy director of the Network Center of the Chinese Academy of Sciences and director of the Center for Computing Science Application Research, in addition to doing a good job in basic research and cutting-edge research, it is also necessary to make breakthroughs in the software versatility, industry specificity and software engineering of supercomputing applications.

Parallel computing: 20 years of grinding a sword

Taking the United States as an example, the purpose of research and development of supercomputers is for their own use, and generally there is a need for computing first, and the system is designed according to the required computing amount, and the architecture of the supercomputer is designed according to the requirements. Huaxia, on the other hand, started construction and tried to improve performance as much as possible to meet higher computing needs.

"In fact, whether you use it before you build it, or build it before you use it, the computer that can meet the application requirements is the best computer. Especially with the advent of the petascale computing era, it is no longer possible to simply use more processing cores to execute the original parallel application, which requires deep parallel optimization of the application and the design of parallel algorithms with better performance. In Chi Xuebin's view, the advanced development of Huaxia supercomputing is an inevitable trend of technological development, but only by grasping the "bull's nose" of parallel computing can we improve computing efficiency and application fields.

Since 1997, the Network Center of the Chinese Academy of Sciences has been improving its basic algorithm library step by step and innovating in mathematical methods in the fields of atmosphere, physics, chemistry, ocean, and metal materials.

"Taking this on-site simulation as an example, in order to closely integrate with the 'Shenwei' system, it took nearly three years from program implementation, software development to the implementation and verification of large-scale parallel computing, and more than 10,000 lines of algorithms were written." Zhang Jian, a researcher at the Computing Science Application Research Center of the Chinese Academy of Sciences, told reporters.

"If a supercomputing center can solve a problem within its life cycle, it is possible to save a time span of more than ten years or decades for a certain field and a certain industry, and this value cannot be calculated with money." Chi Xuebin believes that the development of supercomputing applications requires long-term continuous investment and accumulation in order to see obvious results at a certain time in the future.

As of December 2015, the supercomputing environment of the Chinese Academy of Sciences has provided users with 87 parallel applications in multiple fields such as computational physics, computational chemistry, materials science, fluid mechanics, etc., with a total of 498 external accounts, more than 550,000 valid grid jobs submitted, and more than 120 million CPU hours used.

Targeting application: Seize E-class high ground

Opening the deployment diagram of the supercomputing network environment of the Chinese Academy of Sciences, a three-tier supercomputing environment has taken shape, with more than 100 billions of computing power aggregated by the "main center-sub-center-institute-level center", and each layer of the structure has established a system operation and maintenance and application support team for related fields.

"With the continuous maturity of the Shenwei 26010 software and hardware ecosystem, the demand for application platforms is more diversified, and the diversified development of the talent team is very important to meet the needs of different types of users. In the supercomputing center, each researcher can choose their favorite field, and we have no task indicators, no assessment, and no requirement for everyone to publish articles. Chi Xuebin told reporters.

In Zhang Jian's view, this combination is like a group of troops ready to fight at any time, although everyone focuses on different fields, there is one string that is always tightly pulled, that is, to carry out services and research around parallel computing, once there is a task, it can be topped.

Not seeking to be known and not publicizing is a true portrayal of this team, but it is precisely with this kind of support that many scientific research work inside and outside the Chinese Academy of Sciences has been fruitful.

According to the data, during the "Twelfth Five-Year Plan" period, the number of papers published by users of the supercomputing environment of the Chinese Academy of Sciences continued to rise, with a total of 2,724 papers of various types, most of which were SCI papers. What's even more gratifying is that the "Phase Field Simulation of Titanium Alloy Microstructure Evolution" based on "Sunway Taihu Light" of the Network Center of the Chinese Academy of Sciences has been successfully nominated for the Gordon Bell Prize, the highest award in the field of international high-performance computing applications.

"This is not only the first time that the Huaxia team has been shortlisted, but also breaks the monopoly of Western developed countries, which shows that our country can not only build good computers, but also use good computers." Although Chi Xuebin's mouth is only a subtle difference between "making" and "using", this word condenses the determination and persistence of the people of the Network Center of the Chinese Academy of Sciences not to forget their original intentions.

Talking about the future research focus, Chi Xuebin said, "E-level supercomputing is recognized as the 'next crown of the supercomputer industry' all over the world, and we must seize this strategic commanding heights to provide support services for it, and at the same time, we must also develop products for major application needs of the industry and major goals of enterprises, provide application computing services, develop parallel software, etc." ”

Of course, the above is just a supercomputer made by the Chinese country, and this first generation of intelligent brain is an intelligent computer made by Qin Yu based on the current process based on the memory of the planet Navisda.

When it is completely free from the limitations of code, it is time for a computer to have its own unique operating system, that is, when the computer opens a new door. (To be continued.) )