HPC China | High-Throughput Computing Center Leads a New Era of Intelligent Nuclear
In October 19-21, 2017, 2017 National High Performance Academic Annual Meeting (HPC China 2017) was successfully concluded in Hefei, Anhui Province. The conference will be the theme of "application-driven and ecological co-construction". The conference will include the special report of the conference, thematic forums, thematic forums, application and optimization competition, exhibition and other aspects, top experts from home and abroad, national well-known institutions and scientific research Academics, business people and more than 1,500 participants gathered at the scene. High-throughput computer research center of Chinese Academy of Sciences Institute in this meeting was responsible for organizing a number of links, to exchange and discuss on some leading-edge multi-core technology, high-throughput computing research, developing trends and major applications.
On the 18th afternoon, as a special guest, center director Dongrui Fan, participated in the Panel Symposium on "High Performance Computing Middleware Seminar (HPCMid 2017)", which was jointly organized by the Institute of High Performance Numerical Simulation Software Center and the Beijing Institute of Applied Physics and Computational Mathematics.
In the seminar, Dongrui Fan analyzed and proposed new practical and guided viewpoints on the application status and existing problems of the current high-performance data center middleware, for the domain programming framework, program language, performance optimization and evaluation tools and other research directions to the corresponding solutions for scientific computing and the future development of large data processing system in the middle.
On the afternoon of October 20, at the HPC Plus forum, high-throughput center ungraduated doctor Tan Xu shared the paper "An Efficient Network-on-Chip Router for Dataflow Architecture" published at JCST.
Dr. Tan first introduced the advantages of the data flow architecture, then he introduced the application of the data flow execution mode on the SPU. Finally, the characteristics of SPU on-chip network, and the router optimization method based on this. Dr. Tan shared the data stream on the network method to adapt to the data stream execution mode, can effectively improve the data stream on the network transmission efficiency and effectively。
In his presentation on "Data Streams and Big Data - From Academic Research to Industrial Practice Forums" on the afternoon of 20, Dongrui Fan delivered a lecture entitled "SmarCo: Design High-End Processors with Dataflow Execution Model", which shared high-throughput Center of the use of data flow architecture for innovation, research and development progress and experience.
Mr. Dongrui Fan first explained the application of data flow in China's first nuclear processor Godson-T, Godson-T data transmission design using data flow ideas to effectively improve the data stream synchronization efficiency; then Dongrui Fan director introduced the high-pass SmarCo, SmarCo has the characteristics of high energy efficiency, good disposability and strong expansibility. Director Dongrui Fan said that as a highly efficient, vigorous structure, the data flow architecture will be in the future high-performance computing and artificial intelligence era play a vital role。
On the evening of 20th, the center celebrated the reception by Professor Guangrong Gao, who won the 2017 annual Ramakrishna Rau Award and became the seventh outstanding recipient of the world.
The wine party is hosted by the director of high-throughput computer research center of the Chinese Academy of Sciences Institute, Chairman of SmartCore(Beijing)co.,Ltd. Fan Dongrui researcher.
Professor Guoliang Chen, the director of the Chinese Academy of Sciences Ninghui Sun, Professor Jin Hai of Huazhong University of Science and Technology, Prof. Yin Haifei from Shanghai Jiaotong University, Prof. Yuyafei of Peking University, Professor Qian Depei of Sun Yat-sen University and other experts and scholars at home or abroad attended the reception, congratulat on the achievements made by the professor Gao!
In order to express the high-throughput center for the self-directed and impartial guidance of Professor Guangrong Gao, Dongrui Fan on behalf of the team, the professor Guangrong Gao issued a "high-throughput center for life-long guidance experts."
Professor Guangrong Gao personally cut the celebration of the cake and recalled the sharing of a lot of interesting things on the road of scientific research. The reception was completed in a relaxed atmosphere.
On the afternoon of the 21st, high-throughput computer research center is responsible for the "processor evaluation and optimization technology" thematic forum, director Fan Dongrui as the forum host, and Ye Xiaoshan, deputy researcher, etc. completed the organization of this forum。
The forum invited the researcher Weijun Zhong, the researcher of the Institute of Computing Technology of Wuxi Jiangnan Institute of Computing Technology, Professor Liang Yun of Peking University, Dr. Tang Shanjiang of Tianjin University, and Dr. Li Wenming, Institute of Computing Technology, Chinese Academy of Sciences, on the current processor evaluation and optimization criteria carried out in-depth analysis and discussion, at the same time on the future evaluation of the processor and optimize the direction of the development of practical insights, help the long-term healthy development of domestic processors。
In the "Processor Evaluation and Optimization Technology" thematic forum, Dr. Li Wenming gave a lecture entitled "Evaluation and Optimization Technology for High-Throughput Processor Architecture", which summarizes the high-throughput computer research center in the new high Flux processor research direction of the latest research results, mainly for the new high-throughput processor evaluation and optimization work to explain.
He first analyzed and explained the definition of high-throughput processors and the difference with traditional high-performance processors, as well as high-throughput applications for processor design to solve the main problem. In view of the design of future high-throughput processors, the report puts forward the three-dimensional evaluation standard centered on data "manufacturing", "transmission" and "consumption", which provides theoretical research direction for future high-throughput processor design. Based on this evaluation criteria, the report introduces the SimICT, a simulation tool developed by the High Throughput Computer Research Center, and the 256-core, large-scale verification platform based on FPGA, which can effectively support the evaluation and evaluation of the processor design process. Iterative optimization. At the same time, the report also proposed a set of high-throughput processor microarchitecture benchmarking program that can effectively evaluate the advantages and disadvantages of high-throughput processor design, providing a basis for further optimization. This report provides a set of evaluation, evaluating and optimizing hardware and software tools for the evaluation and optimization of high-throughput processor design, which can effectively drive the future design and optimization of high-throughput application processors.
At the PAC Awards 2010, Professor Guangrong Gao, a Distinguished Expert at the High Tester Center, attended the party and presented the award.
HPC China 2017 concluded successfully. The event highlights industry applications, focuses on the technological innovation, for the domestic high-performance computing academic development, inspire ideas, exchange of experience to provide an opportunity. As a leading global high-volume research team, the high-throughput center will further strengthen the cooperation and exchange of the industry through the participation of the conference, expand the influence of the center. The center will forge ahead in the future, and promote the in-depth study of data flow and the nuclear, regard high-throughput computing as the core, provide more solutions for the era of big data and artificial intelligence.