CRISTIANA BARBOSA BENTES
Cristiana Bentes received the B.S. degree in Mathematics from the Rio de Janeiro State University (UERJ), Brazil, in 1988, with specialization in Computer Science. She received the M.Sc. (1991) and Ph.D. (1998) degrees in Systems Engineering and Computer Science from the Federal University of Rio de Janeiro, Brazil. She is currently Full Professor at the Systems Engineering and Computer Science Department of the State University of Rio de Janeiro, Rio de Janeiro, Brazil. She worked on a number of research projects, collaborating with research groups in Brazil and France. Her research interests are High-performance Computing, Parallel Programming, Novel Parallel Architectures and Cloud Computing. Much of her recent work has been on exploiting the power of GPUs to accelerate a number of applications. She has published numerous journals and conference papers on these subjects, and she is an active member of her research community, co-organizing different HPC conferences.
Topic: Exploiting Concurrency in GPUs
Abstract: Graphics Processing Units (GPUs) have proven to be a powerful and efficient platform to accelerate a substantial class of compute-intensive applications. For this reason, many large scale data centers are based on heterogeneous architecture comprising multicore CPUs and GPUs to meet the requirements of high performance and data throughput. GPUs are also being used in computational clouds. Exploring GPUs in clouds, through GPU virtualization, allows physical devices to be logically decoupled from a computational node and shared by any application, resulting in monetary costs reduction, energy savings and more flexibility. In this scenario, efficiently sharing a GPU across different applications is an indispensable feature. In this talk, we will discuss how concurrent kernel execution can improve hardware utilization and system throughput. We will present the current challenges in exploiting this feature in GPUs and discuss our work and contributions in this area.