Fu Haohuan is the deputy director of the Wuxi National Supercomputing Center and the principal professor of the Department of Geosciences of Tsinghua University. "Post-80s" he has a "supercalculative youth" label on his body. This label is not simple: the Chinese supercomputing application team has won the "Golden Bell Award" known as the "Nobel Prize in the field of supercomputing applications" for a total of 4 times (including 1 international cooperation) so far, and Fu Haohuan has been deeply involved in it 3 times.
As the representative of the winning team, Fu Haohuan was a guest of Science Network on January 20, and in an exclusive interview with China Science News, he shared the anecdotes behind the team's challenge to this application.
Fu Haohuan himself provided the picture
"To confront each other on a battlefield demarcated by others"
China Science Daily: Your team won the 2021 Gordon Bell Award for supercomputing applications, which sounds very shocking: breaking google's claimed "quantum hegemony" that year. Why did the team do this supercomputing simulation of benchmarking quantum computing in the first place?
Fu Haohuan: First of all, "quantum hegemony" is now increasingly translated as "quantum superiority.". From the perspective of computing, quantum computers can reflect the huge advantages of supercomputers in some specific scenarios, and random quantum circuit sampling is one of them.
When Google's quantum computer "Planewood" came out, it claimed that the sampling speed of its random quantum circuit was 1 billion times faster than that of the fastest supercomputer at that time (the American "vertex" supercomputing), and the time it took to achieve sampling on the same scale was 200 seconds versus "ten thousand years". This number is so frightening that when Google first proposed "quantum supremacy", many people came to me and asked me, say supercomputer or not supercomputer? Is it super slow?
China Science Daily: So it is necessary for the research team to benchmark it.
Fu Haohuan: Why did we choose this topic to do it? First of all, we feel that quantum computers are a very potential computing paradigm, but they also need the simulation support of traditional computers. For example, the sampling of Google's "plane tree", its quantum scale manipulation after multiple qubits superimposed, has a large error, fidelity is only about 0.2%. Simulating a sampling result using supercomputing can provide a comparison standard for the results of a quantum computer. In fact, it does need such a comparison.
In fact, choosing such a question is also very challenging - after all, it is necessary to "fight in the battlefield delineated by others", but we feel that if we make it, it will also promote the algorithm and application of supercomputing. This is exactly the case, and the tensor network shrinkage technique we used in the title can be used in many other fields. In addition, we also proved in passing that supercomputing is still very fast.
Eight Easy paper title, the author's English name is "harmonic stem"
China Science Daily: Challenging quantum computers with supercomputing. Are you also feeling stressed?
Fu Haohuan: In fact, for whether our application can be realized, from the beginning of the heart is bottomless. As you may not imagine, the paper we submitted at that time was revised eight times in terms of title alone.
The final paper title was "Closing the 'Quantum Supermacy' Gap" (translated as" bridging the 'quantum hegemony' divide). But at first we weren't on the topic, starting with "Measuring the 'Quantum Supermacy' Gap", then "What is exactly...", and "Minimizing the 'Quantum Supermacy' Gap". Later, we did it step by step to determine the final topic.
Now that I think about it, this process is still quite interesting.
China Science Daily: In this process, what other interesting things will make you unforgettable?
Fu Haohuan: There are many. For example, when we published the paper, we specially retained the English names of the first three authors, namely Alexander, Lucy, and Nancy, who were our chief software architect Dr. Liu Yong, team leader Dr. Liu Xin, and Dr. Li Fang, a team member who was also from Wuxi Supercomputer. The English names of these three are translated as Chinese the consonants are "Alexander" (consonant "pressure mountain"), "Lucy" (harmonic "Lu Xi", taken from "Lu Long And Its Xiu Yuan Xi"), "Nancy" (harmonic "difficult to xi"? ”）。
2021 Golden Bell Prize author bysign
Retaining these 3 English names with "harmonic terriers" feels that this is a vivid reflection of the team's continuous exploration in the process. In fact, our cross-cutting team has gradually formed and cooperated in this exploration.
When I first joined the team, I went to the library of Tsinghua University to borrow several textbooks on quantum computing, found some tutorials on the Internet, and learned a lot of new things. Our final design also draws on the quantum computing team of the Post-Ali Dharma Academy, the Zhang Pan team of the Institute of Theoretical Physics of the Chinese Academy of Sciences, and many other work in the same field.
Quantum computing does not impact supercomputing, and the two have the potential to cooperate
China Science Daily: Will the development of quantum computers impact the development of supercomputers?
Fu Haohuan: The development of quantum computers cannot talk about and will not impact the development of supercomputing, and the development of the two is not contradictory.
As mentioned above, our supercomputing application proves that it is indeed capable of simulating the problems that quantum computing is good at, and it also shows that the algorithm simulation of supercomputing can provide certain assistance and promotion for the development of quantum computing. This is exactly what we did in the first place, otherwise the experts in quantum computing (including the Shanghai Quantum Science Research Center in the participating team) would not have participated in our project.
China Science News: In the application of the "Gordon Bell Award" in 2021, does it reveal that quantum computing and traditional supercomputing have a mutual promotion trend?
Fu Haohuan: That's right. Compared with the so-called "shock", we are actually more concerned about the visible future, it is possible to achieve the connection or interaction between supercomputing systems and quantum computing.
Judging from the results of our supercomputing applications, the speed difference between the two is not as large as in the "legend", which also gives us confidence that supercomputing and quantum computing have the potential to cooperate. What we know now is that quantum computing has advantages in some specific scenarios, and supercomputing also has areas of expertise, and if the two complement each other, there will be more and more interesting things coming out.
There are many "unsung heroes" behind the awards
"China Science News": In recent years, China's supercomputing applications have shined on the "stage" of the "Golden bell award", and you have also participated in several awards. What kind of award is this from a professional perspective?
Fu Haohuan: The Gordon Bell Award was originally established by Gordon Bell in 1987 to encourage the use of parallel computing applications, and later in 2006 in cooperation with the International Computer Association (ACM), which was responsible for the management and operation of the award. The Gordon Bell Prize emphasizes how top-notch supercomputers can be used to solve an application challenge in science, engineering, or large-scale data analysis. This reflects the typical characteristics of supercomputing and has gradually become a benchmark in the field of high-performance computing applications.
There are two important aspects to the Gordon Bell Prize: whether it mobilizes the power of the top supercomputer system-wide, and whether it solves a difficult problem in applications such as science or engineering and innovates. It can be said that the development of continental supercomputing in recent years has provided us with the necessary conditions for impacting this award.
"China Science News": In the 2016, 2017, 2021 and other years of the Gordon Bell Award, you have your presence, you can be said to be a "regular visitor". Can you summarize the tips for winning the award?
Fu Haohuan: First of all, I think I am very lucky to have met these excellent collaborators. In the case of fiscal 2021, we are a team of 14 people with very cross-cutting fields. Many supercomputing applications are actually very complex problems, and to make breakthroughs requires multidisciplinary interdisciplinary thinking and the joint participation of experts in different fields. In fact, behind the application of our annual impact on the Gordon Bell Award, there are many units cooperating.
On the basis of the cross-edge problem challenge, we must also realize the efficient operation of the application model in supercomputing, which is the "intersection" of the upper and lower layers of supercomputing. For example, "Shenwei Taihu Light" is the world's first supercomputer with 1 billion floating-point operations per second, which has tens of millions of cores. Coordinating tens of millions of nuclei to do one thing requires highly systematic design and implementation.
Of course, the biggest foundation for us to impact the Gordon Bell Prize is that the continental supercomputing system has made great progress, achieving many scientific and engineering breakthroughs at the hardware level. It can be said that the supercomputing hardware developers, system maintenance parties, etc. have played a very large role and made their own contributions behind the scenes. Therefore, the Gordon Bell Award belongs to everyone and belongs to the unsung heroes behind China's supercomputing.
Quick Q&A:
Q1, you have done a lot of work in supercomputing to simulate earthquakes, can supercomputing predict earthquakes?
A: No, but simulating earthquakes is very helpful for us to understand the geological structure and activity.
Q2, Tonga undersea volcanic eruption attracted the world's attention, can supercomputing simulate its impact?
A: There are now corresponding estimates from peers in the climate field, and it seems that the impact on the climate is not too large at present.
Q3. Did you win the "Gordon Bell Award" unexpectedly? Who did you share the award with first time after receiving the award?
A: Every time it's exciting, after all, every finalist is possible. The first time, this message was sent to the team's project cooperation group in the form of photos.
Q4. The results were selected as "2021 Top Ten Domestic Science and Technology Progress News" by academicians of the two academies, what do you want to express most?
A: Thanks to scientists in different fields for recognizing this work.
Q5、 Two keywords for the future development of China's supercomputing?
A: Independent and controllable, cross-talent training.