ASC 学校介绍

1 University Supercomputing Overview

1.1 Software and Hardware Platforms for Supercomputing

Southwest Petroleum University is an engineering-focused teaching and research university with a strong multidisciplinary approach, excelling in the petroleum and natural gas sectors. The university is committed to advancing high-level scientific research and technological innovation, continuously producing skilled professionals and innovative results, while fostering the deep integration of industry, academia, and research to establish a thriving innovation ecosystem.

As the central support platform for the university’s research system, The Internet Information Center serves as the computational powerhouse driving both disciplinary development and technological innovation, primarily supporting engineering disciplines and cutting-edge interdisciplinary research. The center leverages advanced infrastructure to build a CPU+GPU heterogeneous architecture and ultra-high-speed data exchange network. In its trial operation in April 2023, the system demonstrated outstanding performance, achieving a theoretical peak computing capacity of 626.8 TFLOPS. The CPU cluster uses Inspur NF5688M6 nodes with Intel® Xeon® Gold 6348 Processors, providing a single-node peak performance of 638.4 GFLOPS. The GPU cluster consists of 4 nodes, equipped with 32 NVIDIA A100 80G GPUs, each offering a double-precision floating-point performance of 9.7 TFLOPS, with a total memory capacity exceeding 2.56 TB. The storage system provides 987 TB of capacity for high-concurrency data access, integrated with an InfiniBand/HDR 200G ultra-high-speed network architecture that ensures data transmission latency below 3 microseconds and peak bandwidth of 200 Gbps. These features fully meet the massive data processing needs of fields like oil and gas exploration, computational chemistry, and more.

In terms of platform construction, the college has built a multi-level research and practice platform: it has established an Image Processing and Parallel Computing Laboratory, equipped with advanced hardware such as Inspur AI server clusters; in-depth collaboration with the National Supercomputing Center in Chengdu to jointly establish an Intelligent Oil and Gas Laboratory and practice base, providing computational support for specialized research such as seismic exploration multiple-wave suppression technology development and oil and gas production capacity prediction model construction. The university-industry collaborative innovation platform supports more than 20 research projects annually, effectively bridging the gap between theoretical research and industrial application.

1.2 Courses, Training Programs, and Interest Groups for Supercomputing

The curriculum system highlights the characteristics of "cutting-edge interdisciplinary integration and industry-education collaboration." Since being approved as the first NVIDIA CUDA Education Center in Southwest China in 2013, the university has continuously deepened the teaching reform of GPU parallel programming technology, offering courses such as "Fundamentals of Parallel Programming (Bilingual)" and "Parallel Processing and Distributed Systems," which integrate undergraduate and master's degree programs. In 2018, the university was the first to introduce the NVIDIA DLI certification program, deeply integrating deep learning and parallel computing technologies. To date, it has trained over a thousand internationally certified supercomputing talents.

The practical teaching adopts a "research-driven teaching" model, creatively transforming the ASC Supercomputing Competition problems into six major practical modules. The focus is on developing skills in algorithm optimization and distributed system development. Leveraging university-industry joint laboratories, interdisciplinary engineering practices such as intelligent oil and gas prediction and seismic wavefield simulation are carried out, forming an advanced training pathway of "classroom learning - competition sharpening - project practice."

In July 2019, the university established the Supercomputing and Parallel Computing Team, which is part of the School of Computer and Software Engineering. The team primarily relies on the Image Processing and Parallel Computing Laboratory to carry out research activities. The team focuses on research and applications in the field of high-performance computing, with research directions including performance optimization of domestic supercomputing platforms, parallel algorithm development, and artificial intelligence inference acceleration. Under the guidance of teachers such as Bo Peng, Quan Zhang, and Yan Li, the team is dedicated to cultivating supercomputing professionals with an international perspective.

1.3 Research and Applications in Supercomputing

The team uses international high-level competitions as a practical platform, focusing on the ASC World Undergraduate Supercomputing Competition, as well as the ISC in Germany and the SC in the United States. We also actively participate in specialized competitions such as the China Parallel Application Challenge On Domestic CPU and the Priority Research Application Challenge. In recent years, the team has won more than 15 national-level competition awards, published 3 academic papers in the field of HPC, and obtained 2 authorized invention patents.

In terms of technological development, the team has carried out deep optimization of the domestic "Sunway" supercomputing platform, with a notable case where algorithm parallelism improvements boosted code execution efficiency from 95 seconds to 0.59 seconds. The team has also innovatively applied Winograd algorithms and block optimization techniques within parallel programming frameworks like OpenMP and MPI, significantly improving the computational efficiency of deep learning operators. Additionally, the team actively explores the intersection of high-performance computing and the petroleum industry, providing algorithm optimization support for the oil and gas exploration and development field.

1.4 Major Achievements in Supercomputing

The team has gained remarkable achievements in the field of supercomputing, including the development of parallel algorithms for complex simulations, the publication of multiple research papers in national and international journals and conferences, and collaboration with industry partners to solve real-world challenges through supercomputing solutions. Below are two of the key achievements.

[1] Bo Peng, Shasha Luo, Zhengqiu Xu, Jingfeng Jiang, "Accelerating 3-D GPU-based Motion Tracking for Ultrasound Strain Elastography Using Sum-Tables: Analysis and Initial Results" Appl. Sci. 2019, 9, 1991.

[2] Pengcheng Chen, Bo Peng, Anxin Zou, Luwen Xu, "Performance comparison of GPU-accelerated fast motion estimation method," 2019 IEEE International Symposium on Parallel and Distributed Processing with Applications (ISPA), Xiamen, 2019, (Accepted). (CCF C)