China Supercomputer runs fast and performs well, but there is still a gap between software and ecology.
"Sunway TaihuLight" supercomputer. Visual China Information
He Liang
Recently, China's self-developed supercomputer, "Sunway TaihuLight," clinched the World Supercomputer Championship at the 2017 International High Performance Computing Conference, marking the third consecutive championship win for a domestically produced Chinese supercomputer system in the world supercomputing rankings. Coupled with the six-time victories of its predecessor, "Tianhe-2," China's supercomputers have shone brightly on the global stage.
Despite its impressive performance and speed, the spotlighted supercomputer has some shortcomings when it comes to its practical applications.
"Sailing out to sea" isn't always a guaranteed success.
Before the advent of "Sunway TaihuLight," China's supercomputer, "Tianhe-2," had once been the "dominant force" in high-performance computing worldwide. It won the top spot in the global Top 500 list six times and found extensive use in fields such as meteorology, climate studies, biomedicine, and new materials. However, it also faced criticism. The Tianhe series processors relied heavily on foreign components—Tianhe-1 used Intel CPUs paired with Nvidia GPUs, while Tianhe-2 utilized Intel GPUs combined with Intel many-core processors.
"The reliance on mature yet foreign chip technology and the rapid advancement of foreign components have been seen as a development strategy, but they prevented the realization of true autonomy and control," Dr. Gan Lin from Tsinghua University's Department of Computer Science told a reporter from Science and Technology Daily on July 28th. A series of measures announced by the U.S. regarding the embargo on China's core processors also directly obstructed the progress of the Tianhe system.
"Through 'borrowing from the sea,' Tianhe-2 surpassed U.S. systems like Titan, but in critical sectors such as national defense security, the country has mandated self-reliance," Dr. Gan Lin explained. While these systems demonstrated "long-term skills," their applications were significantly constrained at a crucial level.
China's Core Enables a Jedi "Counterattack"
Why did China's supercomputers celebrate so fervently following the emergence of "Sunway TaihuLight"? The reason lies in the fact that in response to unilateral sanctions imposed by the U.S., China's supercomputers have fully embraced domestically developed chips—the "Sunway 26010" heterogeneous multicore processor—achieving a Jedi-style "counterattack." This breakthrough is not just about speed; it represents a complete leap forward in terms of autonomous control, peak performance, continuous performance, energy efficiency, and applications.
The National Institute of Computational Fluid Dynamics (CCM) successfully completed large-scale parallel computations of the reentry flight status of the Tiangong-1 spacecraft using the "Unified Algorithm Numerical Simulation" software developed for "Sunway TaihuLight." By utilizing 16,384 processors (one-eighth of the entire TaihuLight system), they accomplished a task that would have taken 12 months in just 20 days, with the results aligning well with wind tunnel experiments. Furthermore, materials science, closely linked to the aerospace sector, requires simulations to determine the optimal elements for specific effects.
Now, Sunway TaihuLight boasts a peak computing performance of 12.54 petaflops per second. This means that in just one minute, Sunway TaihuLight performs calculations equivalent to 7.2 billion people using calculators nonstop for 32 years. Clearly, China's supercomputers are no slouch in terms of hardware capability. But what about their potential?
"Intermediate Interruptions" Limit Application Expansion
In 2016, an application project running on Sunway TaihuLight won the "Gordon Bell" prize, marking a 29-year milestone for high-performance applications in China.
The project titled "Ten Million-Core Scalable Global Atmospheric Dynamics Implicit Simulation" utilized over 10 million cores to perform a numerical simulation of the climate. Its standout feature was the algorithm—a novel "implicit" calculation method. This algorithm improves computational efficiency by nearly an order of magnitude compared to the current international benchmark. Yang Chao, leader of the award-winning team and a researcher at the Institute of Software of the Chinese Academy of Sciences, stated that the algorithm and performance were the primary reasons for their victory.
Implementing high-performance computing to achieve practical applications isn't solely about the supercomputer itself. It's akin to a chain, where software bridges the gap between upstream application scenarios and downstream hardware. Professor Yang Guangwen, director of the National Supercomputing Wuxi Center and a professor at Tsinghua University, noted in an interview with Science and Technology Daily that researchers in the computer field cannot design superior algorithms and application software without understanding the applications. Conversely, users of application software often lack access to appropriate computers due to unfamiliarity with the technology. The ability to produce high-quality software... Even if the technology is advanced and demand is strong, a lack of software will hinder the application of supercomputing power.
Moreover, China faces a greater challenge: not only is software lagging behind, but there is also a gap in the software ecosystem compared to foreign counterparts. For instance, Gan Lin pointed out: The sophisticated processor of Sunway TaihuLight is akin to coordinating and pulling heavy objects with 10 million ants simultaneously. If one ant "falls," the tasks assigned to that ant must be redistributed to other software. Ensuring smooth operations requires robust software mechanisms for fault tolerance and error correction. Although China has such mechanisms, they are not as mature as those abroad. Most programs running on Sunway TaihuLight require code modifications to optimize the hardware's unique features. The tuning process depends heavily on software tools.
"Even though we've won the 'Gordon Bell' award, there's still a long way to go before we fully enhance our nation's supercomputing application level," Professor Yang Guangwen remarked candidly. This is akin to taking something out of the lab—its efficacy is predictable, but implementation remains untested. Professor Yang believes that researchers must adopt an "artisan spirit" to excel in supercomputing.
From hardware to software, China has been closing the gap and working tirelessly to catch up. At the national level, several projects under major national science and technology special projects are receiving funding support in the field of supercomputing, with investments continuing to rise. Professor Yang Guangwen stated that systemic policies, projects, and talent support will drive the development of ultra-numerical R&D and nurturing industries, promoting national industrial transformation and upgrading.
Quantum Computing and Hypercomputing Are Not Competitors
At the "2017 International Big Data and E-Class Computing Symposium," Professor Yang Guangwen disclosed that the development of the next-generation exascale supercomputers based on "Sunway TaihuLight" has been included in the national "13th Five-Year Plan" and is currently progressing rapidly. The focus is on developing three prototypes: "Sunway," "Tianhe," and "Dawning." It is anticipated that the first domestically produced exascale supercomputer will be launched around 2020.
Scaling from billions to quintillions presents immense challenges. Professor Yang told Science and Technology Daily that increasing computing power leads to rising power consumption. Current efforts are focused on designing low-power chips, adopting efficient cooling technologies, and optimizing system-wide management.
Compared to emerging quantum computers, Ms. Gan Lin believes that the next generation of supercomputers will not compete with them for decades. Traditional computers represent each bit as either 0 or 1, whereas quantum computers leverage quantum uncertainty to enable multistate representations. For example, a qubit can exist in multiple states, potentially making quantum computers far more powerful than traditional ones at the same scale. However, the uncertainty in molding these possibilities introduces additional variables.
"For quantum computers to become mature, they require a robust algorithm capable of simulating applications and matching quantum uncertainty to yield desired results," said Gan Lin. This limits quantum computers to specific application domains.
"In the next phase of supercomputing development and competition, China must not only win the race of speed but more importantly win the race of applications!" Professor Yang Guangwen emphasized. The lag in domestic application software and the shortage of talent have severely impeded supercomputing progress. With the ultimate goal of "applications," the "Sunway TaihuLight" team plans to organize both domestic and international strengths to build a domestic chip ecosystem and enhance application software development, particularly in fields such as weather and climate services, advanced manufacturing, and big data analysis. By addressing application bottlenecks and filling gaps, China aims to elevate the capabilities of its supercomputers. "
HQD Vape pen, Wholesale HQD Vape, HQD Manufacture
Shenzhen Xcool Vapor Technology Co.,Ltd , http://www.xcoolvapor.com