Title：Irregular Computations on Modern Parallel Processors
Irregular computations exist in a number of computational problems in real-world applications. Researchers have been always looking for faster parallel algorithms for irregular problems in the last decades. Recently, the emergence of many-core processors has introduced more conflicts between algorithm efficiency and scalability. On one hand, many-core processors require a large amount of fine-grained tasks to saturate their resources, on the other, the irregular data structure brings difficulties to the task partitioning. This talk will use sparse matrix algorithms as a representative of irregular problems, and discuss the scalability of existing work and our newly designed data structures, such as the CSR5 format (ICS ’15), and algorithms, such as sparse matrix transposition (ICS '16), sparse matrix-vector multiplication (ICS ’15 and Parco), sparse triangular solve (Euro-Par ‘16) and sparse matrix-matrix multiplication (IPDPS ’14 and JPDC).
Furthermore, the newly introduced high-bandwidth on-package memory (OPM) is another interesting and helpful feature of modern parallel processors, because it innovates the conventional memory hierarchy by augmenting a new on-package layer between classic on-chip cache and off-chip DRAM. This talk will demonstrate our experimental results of a comprehensive evaluation for a wide spectrum of computational kernels, in particular irregular ones, with a large amount of representative inputs on two Intel OPMs: eDRAM on multicore Broadwell and MCDRAM on manycore Knights Landing. Guided by our general optimization models, we demonstrate OPM’s effectiveness for easing programmers’ tuning efforts to reach ideal throughput for both compute-bound and memory-bound applications (SC ’17).
Weifeng Liu is currently a Research Scientist at the Department of Computer Science of the Norwegian University of Science and Technology in Trondheim, Norway. He is also a Marie Curie Fellow of the European Union. He received his Ph.D. in 2016 from University of Copenhagen, Denmark. Before he moved to Copenhagen, he has been working as a Senior Researcher in high performance computing technology at SINOPEC Exploration & Production Research Institute for about six years (2006-2012). He also has been shortly working as a Research Associate at STFC Rutherford Appleton Laboratory, UK, in 2016. He received his B.E. degree and M.E. degree in computer science, both from China University of Petroleum, Beijing, in 2002 and 2006, respectively. His research interests include numerical linear algebra and parallel computing, particularly in designing parallel and scalable algorithms and data structures for sparse matrix computations on throughput-oriented architectures.