Panel Battle of the Accelerator Stars Panel Moderator

Panel: Battle of the Accelerator Stars Panel Moderator: Yong Chen

Battle of the Accelerator Stars � Computing accelerators have gained intensive attentions in recent years � Generally believed promising as a low power consumption and highly parallel computing solution � E. g. Stampede at the Texas Advanced Computing Center will produce 10 petaflop speeds, with 8 Peta. FLOPS from MIC accelerator � Witnessed proliferations of computing accelerators � NVIDIA TESLA, Quadro, Ge. Force, AMD Radeon, Fusion APU, Intel Larrabee, MIC (Knights corner), STI Cell, FPGA, etc. 2 � CUDA, Open. CL, Open. MP, Intel Cilk Plus, Open ACC, etc. P 2 S 2 -2012 10/2/2020

Battle of the Accelerator Stars (cont. ) � Invited experts in this field to discuss the future of computing accelerators � With a theme on hardware/software divergence in accelerator computing � Share with you their visions and experience � Panelists will address based on five questions asked and will interact with the audience � We have an Accelerated Computing session following too 3 P 2 S 2 -2012 10/2/2020

Questions We Asked Panelists � Q 1. How do you envision the role of accelerator in parallel computing and high performance computing in the next decade including the role in the exascale systems? � Q 2. How do you view the hardware/software divergence in accelerator Computing? � Q 3. Which accelerator (hardware) do you think will have advantages in the next 10 years and most likely win the battle in the next decade and why? 4 P 2 S 2 -2012 10/2/2020

Questions We Asked Panelists (cont. ) � Q 4. What programming model/library of accelerated computing do you think will have advantages in the next 10 years and most likely win the battle in the next decade and why? � Q 5. What research challenges do you envision will be most critical and should be addressed in the coming years for the success of accelerator computing? " 5 P 2 S 2 -2012 10/2/2020

Panelist – Pavan Balaji � Computer Scientist at the Argonne National Laboratory, and as a research fellow of the Computation Institute at the University of Chicago � Focuses on parallel programming models and runtime systems for communication and I/O, modern system architecture, cloud computing systems, and job scheduling and resource management 6 P 2 S 2 -2012 10/2/2020

Panelist - Taisuke Boku � Professor in the Computer Science Department of the University of Tsukuba � Focuses on large-scale supercomputing and high performance computing with GPGPUs 7 P 2 S 2 -2012 10/2/2020

Panelist - Volodymyr Kindratenko � Senior Research Scientist at the National Center for Supercomputing Applications (NCSA); Lecturer in the department of Electrical and Computer Engineering at the University of Illinois at Urbana. Champaign � Focuses on high-performance computing with accelerators and special-purpose computing architectures 8 P 2 S 2 -2012 10/2/2020

Panelist: Xipeng Shen � Adina Allen Term Distinguished Associate Professor in the Computer Science Department of The College of William and Mary. Currently on sabbatical leave, working as a visiting scientist at MIT Computer Science and Artificial Intelligent Lab � Focuses on compiler technology and programming systems to achieve high performance and programming productivity on parallel computing platforms 9 P 2 S 2 -2012 10/2/2020

Panelist: Kyle Spafford � Research and development staff member in the Future Technologies group of the Oak Ridge National Laboratory � Focuses on emerging architectures for extreme scale computing. A major contributor to the well-known SHOC benchmark suite and a member of the NSF Keeneland project 10 P 2 S 2 -2012 10/2/2020

Panel starting � Panelists start with position statements � Pavan Balaji � Taisuke Boku � Volodymyr Kindratenko � Xipeng Shen � Kyle Spafford � Please hold questions and panelists with answer any questions afterwards � Thank 11 You! P 2 S 2 -2012 10/2/2020