Challenges before E-level calculations

"Protein folding" is an important subject in the field of biological science research. The greatest challenge facing the health system in the world today is Alzheimer's disease caused by the misfolding of human protein molecules. For this unobserved “microscopic world”, the current way scientists are trying to understand the process of protein folding is to simulate the process of protein folding on a computer and compare it with the results of drug experimentation in order to find ways to cure the disease and simulate protein folding. The process must rely on the support of the HPC (High Performance Computing) system.

Chen Yifeng, associate dean of the Computer Department of Peking University’s School of Information and his colleagues, is one of the HPC technical teams that supports such research. Chen Yifeng told reporters that the supercomputers such as "Tianhe-2" and "Tianhe-A" have assisted domestic fields in biology, chemistry, physics, astronomy, engineering, engineering, etc. The research institutes achieved a number of major breakthroughs. From the point of view of the field of biological science research, "Sky Creek" and other supercomputers are more significant in sprints to E-grade calculations. Research like "protein folding" is very promising in the era of E-scale supercomputing. Breakthrough breakthrough to solve the problem of human genetic diseases. However, prior to this, there are still some classes in the HPC application field that must be filled in in order to allow HPC to play its due value in the research field.

Although many studies have a strong demand for HPC, there are not many that can be used now and make HPC use its due value. Chen Yifeng emphasized that the disciplines involved in the application of HPC are complex, the technical threshold and professional threshold are high, and the requirements for talents are higher. Judging from the current situation, there is still a certain gap between China and foreign countries in terms of personnel training. There are very few talents who understand both mathematics and computer science. The entire industry often "can't recruit people" and there are problems in the world.

In addition, the speed of hardware development far exceeds software. From the perspective of the evolution of CPU technology, the performance of HPC system hardware has changed dramatically in a year or two from single-core to dual-core to quad-core to multi-core, and now many-core, far surpassing biological, chemical, and physical properties. The speed of development of these disciplines. Although people have seen breakthrough opportunities that may arise from the increase in computing power and reduced computing costs, the development of appropriate software and good tools requires constant learning in order to serve application research. For an HPC center like Peking University, it is necessary to study a number of professional disciplines such as applied mathematics in order to dig out suitable methods to help the application. It is a long process from the emergence of HPC related new technologies to the real research areas.

"Laboratory mechanisms in some foreign countries are very worthy of our reference." Chen Yifeng told reporters that the United States and other countries have set up large-scale supercomputer national laboratories. They generally adopt multi-disciplinary scientists to jointly exchange and jointly solve HPC problems. Physicists, chemists, and computer experts will work together to solve the application problems of using HPC technology. Multidisciplinary experts will cross-study the use of HPC technology and make it easier to produce valuable results. This is the mechanism that is currently lacking in the domestic HPC application field.

In the list of global supercomputer TOP500 announced this year, "Tianhe 2" topped the list, and China Supercomputer has won for three consecutive years. In the HPC application hardware conditions, China has even surpassed many developed countries. However, how can such a hardware foundation be transformed into one breakthrough in the fields of biology, chemistry, and engineering? It is indeed a question that the HPC ecosystem should consider.

As the lead of the HPC hardware platform, Intel is also making every effort to compensate for the software lag caused by the rapid development of the hardware. In terms of personnel development, Intel has launched a multi-nuclear university cooperation program, organized professional training for university teachers with the Ministry of Education, and expanded the training scope of HPC talents by editing textbooks, setting up related courses in partner universities, and holding competitions. For software development difficulties, Intel has introduced a series of tools to help developers reduce development costs and improve development efficiency. Such as VTune, Thread Profiler and other multi-threaded tools, cluster tools, etc., but also on the MIC and tools on the CPU are also consistent, developers do not have to relearn. Through technology integration, Intel is gradually achieving full compatibility of the computing platform instruction set. In addition, Intel is also eliminating the issue of different instruction sets through the compiler to enable the future to target different hardware platforms, and applications can automatically adapt to changes.

With all kinds of efforts, HPC may soon be able to get rid of the image of “Spring Snow” and move from “small circles” to “large circles” to create miracles in more areas, even for next-generation applications such as big data and 3D printing. service.