SPEAKERS

  • Zhiyuan Li, Purdue

  • Professor Zhiyuan Li received his Bachelor of Science degree from Department of Mathematics, Xiamen University, Fujian, in 1982. He received his MS and PhD degrees in 1985 and 1989 respectively from Department of Computer Science, University of Illinois, Urbana-Champaign.

    Dr. Li joined the Department of Computer Science, Purdue University at West Lafayette, Indiana in 1997 as an Associate Professor. He has been a Professor in the same department since August of 2006. Prior to joining Purdue, he taught in York University, Canada and University of Minnesota, Minneapolis-St. Paul.

    Dr. Li was a Visiting Faculty at Microprocessor Technology Lab, Intel Corporation at Santa Clara, California from May to June in 2004 and a Visiting Faculty at Software & Solutions Group, Intel Corporation at Santa Clara, California from May to June in 2006.

    Dr. Li’s technical contributions date back to as early as 1989 with research on data dependence analysis framework and methods. Over the last two decades, he has continued to push the envelope in compiler research for parallel and high performance computers. His two seminal papers in the inaugural issues of IEEE Transactions on Parallel and Distributed Computing (TPDS) in 1990 have influenced the design of dependence testing algorithms in both research and production compilers to this date. His result on data dependence analysis has been used by researchers in many institutions.

    Dr. Li’s 1992 publication on array privatization for parallel execution of loops gave the first array dataflow analysis framework for removing artificial dependences due to array rewriting which severely restricts parallelism in practical programs. He is invited to write a retrospective article on this work for the 25 year Anniversary of ACM International Conference on Supercomputing. With parallel processors (or multicores) becoming main-stream, interprocedural array dataflow analysis for parallelization, which Dr. Li pioneered with two other groups (Rice University and Ecole des Mines de Paris), have been adopted in various forms in production compilers for multicores. Dr. Li has extended his array dataflow analysis for enhancing data locality and memory access speed. His 1999 ACM TOPLAS paper on tiling methods to improve cache temporal locality continues to inspire today’s researchers in designing domain-specific languages and computation schemes for stencil computation.

    In recent years, Dr. Li has expanded his compiler research to address issues emerging in distributed embedded systems such as mobile devices and wireless sensor networks. His early work on computation offloading on server-supported mobile devices predated currently popular cloud-based mobile computing by nearly a decade. He has focused on his current effort on dynamic analysis techniques such as data dependence profiling and on techniques for locating the root causes of program errors.

    Dr. Li has a long record of service for the computing community by serving on the technical program committees for leading conferences sponsored by IEEE Computer Society and ACM, including ACM PPoPP, PLDI, ICS, LCTES, IEEE ISPASS, IPDPS, IPPS, International Conference on High Performance Computing, as well as the Programming Language Processing Workshop affiliated with IPDPS. He served for several years as the Steering Committee Chair for ACM/SIGPLAN International Symposium on Languages, Compilers, and Tools for Embedded Systems (LCTES).

  • 廖世偉 副教授, 台灣大學

  • Shih-wei Liao's lifelong goal is to make impacts through innovation and education, open-source and product-quality systems, so he worked at Google and Intel headquarters in Silicon Valley for more than a dozen years. He received his PhD and MS from Stanford University and his bachelor degree (台大資訊工程系76級, class of 1987) from National Taiwan University.