Breaking Google's quantum hegemony, China supercomputing team won the Gordon Bell Award in 2021

Early this morning, Beijing time, at the SC international supercomputing conference in 2021, the Chinese supercomputing team composed of 14 people from Tsinghua University, Zhejiang University and Wuxi Supercomputing Center won the ACM (International Computer Society) Gordon Bell Award for its application "super large scale quantum random circuit real-time simulation" (swqsim) based on China's new generation Shenwei supercomputer.

Gordon Bell award is the highest academic award in the field of high-performance computing applications in the world, known as the "Nobel Prize in supercomputing".

The title of the award-winning paper is: bridging the gap of "quantum hegemony": closing the "quantum supreme" gap: achieving real-time simulation of a random quantum circuit using a new Sunway supercomputer. In the article "three studies in China show that classical computers will completely disintegrate Google's quantum hegemony" on November 5, the photon box reported this research.

The 14 authors are as follows:
Zhijiang laboratory and Wuxi National Supercomputing Center: Liu Yong, Liu Xin, Li Fang,Yang Yuling, song Jiawei, Zhao pengpeng, Wang Zhen, Peng Dajia, Chen Huarong

Tsinghua University and Wuxi National Supercomputing Center: Fu haohuan, Chen Dexun
Wuxi National Supercomputing Center: Wu Wenzhao

Shanghai Quantum Science Research Center: Huang Heliang, Guo Chu

Corresponding authors: Liu Xin, Fu haohuan, Guo Chu, Chen Dexun

In the work of the team that won the Gordon Bell Award, the researchers introduced a system design process, covering the algorithm, parallelization and architecture required for simulation. Using a new generation of Shenwei supercomputer, the researchers effectively simulated a 10 × 10 qubits and 1 + 40 + 1 layer depth random quantum circuit. In the simulation, the research team used 41.9 million processors to achieve the single precision of 1.2 eflops (10 billion floating-point operations per second) or mixed precision of 4.4 eflops.

For the 53 qubits of Google's "sycamore" and the random circuit of 20 layer cycle, Liu Xin of Zhijiang laboratory and the National Supercomputing Wuxi center, the corresponding author of the paper, said that compared with the "sycamore" completing the million 0.2% fidelity sampling task in 200 seconds, swqsim can obtain millions of higher fidelity correlation samples within 304 seconds, Get the same number of unrelated samples within a week, breaking its claimed "quantum hegemony". In addition, swqsim can complete more than 1000 times more complex quantum circuit simulation than "sycamore" in 60 hours, and realize single amplitude and multi amplitude simulation of 100-400 bit quantum circuit algorithm, which provides a solid simulation support for the development of quantum computing in the future.

Liu Xin said: "we can definitely answer that the so-called 'quantum hegemony' of the random quantum circuit sampling task proposed by Google is not tenable. Our work shows this: classical supercomputers can also do it." [1]

Provided by researchers, the main summary of classical random quantum circuit simulation. In the figure, the x-axis represents the number of qubits and the y-axis represents the corresponding storage space required. The sizes of circles and rectangles represent the complexity / depth of the circuit.

In the article "original technology" of the official account, Bao Yungang, a researcher, doctoral supervisor, assistant director of the Chinese Academy of Sciences, and professor of China University of science and technology, commented: congratulations on China's team winning the Gordon Bell prize for the third time!

This award has three meanings:
The first is scientific significance. The award-winning application is "real-time simulation of super large-scale quantum random circuits". This application can complete the single amplitude simulation of 100 qubits and 40 layers of quantum random circuits within 60 hours and complete the simulation of "quantum hegemony" of Google "sycamore" in 304 seconds (Google claimed that it would take 10000 years to get similar results using summit, the fastest supercomputer at that time). This achievement effectively counterattacks the recent myth that quantum computers have obtained quantum hegemony over traditional supercomputing on specific issues.

The second significance is to show the strength of Chinese supercomputers. At present, China, the United States and Japan are attacking E-class supercomputers (10 billion operations per second). Overseas reports have reported that China has taken the lead in developing E-class supercomputers, and there are two [4] However, for some reasons, China has not included the latest supercomputer in the top 500 list, so we still see that Fuyue of Japan ranks first and Shenwei Taihu light of China ranks fourth. This Gordon Bell award shows that this quantum random circuit real-time simulation work is based on "Shenwei new generation supercomputer" It is a kind of Versailles not participating in the top 500 ranking, but winning the annual Gordon Bell award with the application of E-class super computing.

The third significance is that China's previous situation of "making the strong and using the weak" in the supercomputing field has begun to turn around. This year, three supercomputing applications in China have been shortlisted for the Gordon Bell Award (6 shortlisted in the world). In addition to the applications of the winning teams, the other two applications are "ten million core scalable first principle bit Mann spectral simulation" and "Multi architecture large-scale parallel symplectic preserving structure electromagnetic full dynamic plasma simulation".

Link:
[1] https://mp.weixin.qq.com/s/814oGFphuuCaMDemTYW8Ew
[2] https://www.hpcwire.com/2021/11/18/2021-gordon-bell-prize-goes-to-exascale-powered-quantum-supremacy-challenge/
[3] https://www.nextplatform.com/2021/11/18/chinas-exascale-quantum-simulation-not-all-it-appears/
[4] https://www.nextplatform.com/2021/10/26/china-has-already-reached-exascale-on-two-separate-systems/