Two papers from Prof. Yun (Eric) Liang’s Group were accepted by the top conferences in computer architecture


Recently, Prof. Yun (Eric) Liang’s works: "Poly: efficient heterogeneous system and application management for interactive applications" and "GPU-based partitioning and batching" are accepted by the 25th Annual IEEE International Symposium on High-Performance Computer Architecture (HPCA) and the 24th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming (PPoPP), respectively.

In the previous paper, Yun (Eric) Liang and his Ph.D. candidate Shuo Wang proposed an efficient runtime management framework Poly for heterogeneous computing systems integrating multiple GPUs and FPGAs; the framework requires high real-time interaction. The application can improve the overall energy efficiency ratio of complex systems under the premise of ensuring real-time performance, thus providing an energy-efficient computing resource management and capacity expansion solution for cloud computing, data center, and other application scenarios.

In the latter paper, Yun (Eric) Liang and his Ph.D. student Xiuhong Li and the research and development staff of Beijing Shangtang Technology Development Co., Ltd. designed a new matrix block decision algorithm and batch execution decision algorithm to significantly improve the small matrix algorithm. The computational efficiency overcomes the bottleneck that many small matrix algorithms in real-world applications cannot fully exploit GPU computing power.

HPCA and PPoPP are the top conferences in the field of computer architecture. They are listed as the Class A conferences in the field of computer systems and high-performance computing by the China Computer Society Recommended International Academic Conference and Journal Catalogue. They have high academic influence at home and abroad. The paper is strict,and the admission rate is low (the former is about 45 papers/year, the latter is about 30 papers/year, and the average admission rate is 15%-20%).