【字色： 红 蓝 褐 绿 黑 紫 粉红 深蓝】
【字体:8 7 6 5 4 3 2 1】
题目：Compiler and Runtime Optimizations for GPU Memory Systems
报告人简介：Bo Wu is an Assistant Professor in the EECS department at Colorado School of Mines. He received his Ph.D. degree in computer science from the College of William and Mary for his dissertation on "Matching Non-Uniformity for Program Optimizations on Heterogeneous Manycore Systems”. He received his M.S. degree in computer science and B.S. degree in mathematics both from Central South University in China. His research interests lie in the broad field of compilers and programming systems, with an emphasis on program optimizations for heterogeneous computing and emerging architectures. Most of his research activities have centered around data locality enhancement for heterogeneous computing systems. He is a recipient of the IBM Center of Advance Study fellowship, the NSF CRII Award, and the best paper award in Supercomputing’15.
内容提要：In recent year, we have seen a rapid rise of massive parallelism in modern processors, exemplified by GPUs, APUs and MIC processors. The emergence of massive parallelism in a single node brings growing demands for efficient data accesses. Optimizing memory throughput is hence critical for tapping into the full potential of the many-core systems. In the first half of the talk, I will present some theoretical and empirical results on optimizing memory coalescing for GPUs. In the second half of the talk, I will present our framework to automatically place data on GPUs for performance and portability.